* Mips target in gold - revision 3 - part 2
@ 2014-04-15 13:53 Sasa Stankovic
2014-04-15 20:46 ` Cary Coutant
0 siblings, 1 reply; 9+ messages in thread
From: Sasa Stankovic @ 2014-04-15 13:53 UTC (permalink / raw)
To: Cary Coutant [ccoutant@google.com]
Cc: Rich Fuhler, binutils, iant, Petar Jovanovic
[-- Attachment #1: Type: text/plain, Size: 7058 bytes --]
Hi Cary,
Attached is the second part of the patch that implements Mips target in gold. This part contains second part of the mips.cc file.
> > // In some cases GCC dead code elimination removes the LO16 but
> > // keeps the corresponding HI16. This is strictly speaking a
> > // violation of the ABI but not immediately harmful.
>
> If I understand the code correctly, in this case the HI16 reloc
> will remain on the got16_addends_ list forever. Is there a point
> where you know it's safe to remove it? Can you clear the list
> at the end of each section?
If LO16 is removed by GCC, than HI16 reloc is paired with the next found LO16
reloc (this means that two or more HI16 reloc will be paired with a single LO16 reloc).
This is what GNU ld does - it always tries to find LO16 reloc for HI16 reloc, and if it
can't find it, error is reported.
When trying to find LO16 reloc, GNU ld examines next relocations in the section until
it finds it, or reports an error. I implemented it differently in Gold - when HI16 or GOT16
reloc is found, it is recorded in the got16_addends_ list, and the normal scaning of
relocations continues. Then when the LO16 reloc is found, the got16_addends_ list is
examined and all pending matching HI16 and GOT16 relocs are removed. The error when
HI16 or GOT16 reloc doesn't have the LO16 part is detected by checking whether the
section of the pending HI16 or GOT16 relocation is different from the section of the current
LO16 relocation. The check whether the final section that is scanned has HI16 or GOT16
without the LO16 part is in the do_finalize_sections method - I just check whether the
got16_addends_ list is empty - if it isn't, error is reported.
> > //gold_assert(static_cast<section_size_type>(pov - oview) == oview_size);
>
> Was this commented out while debugging? If you can't leave it
> enabled, please add a TODO noting what the problem is. At the
> least, you should make sure you didn't overrun the oview, and if
> you underran it, you should fill with zeroes.
I enabled it again.
> > typename std::list<got16_addend<size, big_endian> > got16_addends_;
>
> Do you realy want to use a linked list here? Wouldn't a
> vector be better?
I used std::list because erase is called on it.
> > layout->add_output_section_data(".got", elfcpp::SHT_PROGBITS,
> > (elfcpp::SHF_ALLOC | elfcpp::SHF_WRITE |
> > elfcpp::SHF_MIPS_GPREL),
> > this->got_, ORDER_DATA, false);
> > ...
> > // If there is no .got section, gp should be based on .sdata.
> > // TODO(sasa): If there are both .got and .sdata sections, they must be
> > // together, with .got comming first.
>
> If you want .got and .sdata together, shouldn't you define .got
> with ORDER_SMALL_DATA? (To guarantee that it precedes .sdata, you
> may need to add ORDER_SMALL_DATA_FIRST to layout.h.)
I removed this TODO and added a new TODO to check whether .sdata is accessed
using gp-relative addressing. This was the case for older MIPS architectures
which accessed both .got and .sdata section using gp-relative addressing (so that
items from the .sdata were read/written using just one instruction instead of
two). But I think that modern Mips Linux ELF architectures don't access .sdata
using gp-relative addressing. So the code that initializes gp to point on
.sdata if there is no .got will probably be removed in the future.
> > // big-endian file, the result is the same; in a little-endian
> > // file, the two 16-bit halves of the 32 bit value are swapped.>
> Do you mean that the bytes within each are swapped, but the
> 16-bit words are stored in order, or that the 16-bit words are
> swapped (i.e., the second written before the first)? I'm guessing
> that you mean that the byte-swapping takes place within each
> 16-bit word, and the order of the two words is unaffected.
Yes. On MIPS16 and microMIPS (whose instruction sets have both 16-bit and 32-bit
instructions) [31:16] part of the instruction is always stored first, and [15:0] is stored after;
bytes within each 16-bit word are swapped based on endianess.
> > // To put it in MIPS ABI terms, the relocation field is T-targ26-16,
> > // defined as
> > //
> > // big-endian:
> > // +-------- +----------------------+
> > // | | |
> > // | | targ26-16 |
> > // |31 26|25 0|
> > // +-------- +----------------------+
> > //
> > // little-endian:
> > // +----------+------+-------------+
> > // | | | |
> > // | sub1 | | sub2 |
> > // |0 9|10 15|16 31|
> > // +----------+--------------------+
> > // where targ26-16 is sub1 followed by sub2 (i.e., the addend field A is
> > // ((sub1 << 16) | sub2)).
>
> I can't figure out what this diagram is saying. In particular,
> why are the bits numbered right-to-left for big endian and
> left-to-right for little endian? Most processors I'm familiar
> with use the same bit numbering convention for both big- and
> little-endian, but I certainly wouldn't expect to see this.
That whole comment (and many others) is copied from the GNU ld (file bfd/elfxx-mips.c).
32-bit JAL and JALX instructions on MIPS16 have the format
+--------------+--------------------------------+-----------
| JALX | X| Imm 20:16 | Imm 25:21 |
+--------------+--------------------------------+-----------
| Immediate 15:0 |
+-----------------------------------------------+------------
First row represents the bits [31:16], second row represents the bits [15:0]. Note that two
5-bit immediates in the first word are swapped.
When producing a relocatable object file (either by gas or by giving -r option to linker),
JAL(X) MIPS16 instruction is stored like its MIPS32 equivalent:
+--------------+----------------------------------------+
| JAL(X) | Immediate 25:0 |
+--------------+----------------------------------------+
So basically, 3 immediates are first reordered into full 26-bit target of the JAL instruction,
and then the instruction is stored as two 16-bit parts (like all the other 32-bit instructions on
MIPS16 and microMIPS). This means that in the little-endian picture from your question,
sub1 represents Imm[25:16] and sub2 represents Imm[15:0] (I agree that bit numbering on
the little-endian picture is confusing).
Note that this only applies for relocatable files. For final link, the JAL(X) instruction is stored
based on the instruction format (with bits [31:16] always stored first). MIPS16 JAL(X) instruction
is the only instruction whose layout differs in relocatable and executable files.
Regards,
Sasa
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: mips3.patch --]
[-- Type: text/x-patch; name="mips3.patch", Size: 196095 bytes --]
diff --git a/gold/mips.cc b/gold/mips.cc
index 1a8e0db..c344096 100644
--- a/gold/mips.cc
+++ b/gold/mips.cc
@@ -5227,3 +5227,5231 @@ Mips_got_info<size, big_endian>::count_got_symbols(Symbol_table* symtab)
}
}
}
+
+// Return the offset of GOT page entry for VALUE. Initialize the entry with
+// VALUE if it is not initialized.
+
+template<int size, bool big_endian>
+unsigned int
+Mips_got_info<size, big_endian>::get_got_page_offset(Mips_address value,
+ Mips_output_data_got<size, big_endian>* got)
+{
+ typename Got_page_offsets::iterator it = this->got_page_offsets_.find(value);
+ if (it != this->got_page_offsets_.end())
+ return it->second;
+
+ gold_assert(this->got_page_offset_next_ < this->got_page_offset_start_
+ + (size/8) * this->page_gotno_);
+
+ unsigned int got_offset = this->got_page_offset_next_;
+ this->got_page_offsets_[value] = got_offset;
+ this->got_page_offset_next_ += size/8;
+ got->update_got_entry(got_offset, value);
+ return got_offset;
+}
+
+// Remove lazy-binding stubs for global symbols in this GOT.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::remove_lazy_stubs(
+ Target_mips<size, big_endian>* target)
+{
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (entry->is_for_global_symbol())
+ target->remove_lazy_stub_entry(entry->sym());
+ }
+}
+
+// Count the number of GOT entries required.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::count_got_entries()
+{
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ this->count_got_entry(*p);
+ }
+}
+
+// Count the number of GOT entries required by ENTRY. Accumulate the result.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::count_got_entry(
+ Mips_got_entry<size, big_endian>* entry)
+{
+ if (entry->is_tls_entry())
+ this->tls_gotno_ += mips_tls_got_entries(entry->tls_type());
+ else if (entry->is_for_local_symbol()
+ || entry->sym()->global_got_area() == GGA_NONE)
+ ++this->local_gotno_;
+ else
+ ++this->global_gotno_;
+}
+
+// Add FROM's GOT entries.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_got_entries(
+ Mips_got_info<size, big_endian>* from)
+{
+ for (typename Got_entry_set::iterator
+ p = from->got_entries_.begin();
+ p != from->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (this->got_entries_.find(entry) == this->got_entries_.end())
+ {
+ Mips_got_entry<size, big_endian>* entry2 =
+ new Mips_got_entry<size, big_endian>(*entry);
+ this->got_entries_.insert(entry2);
+ this->count_got_entry(entry);
+ }
+ }
+}
+
+// Add FROM's GOT page entries.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_got_page_entries(
+ Mips_got_info<size, big_endian>* from)
+{
+ for (typename Got_page_entry_set::iterator
+ p = from->got_page_entries_.begin();
+ p != from->got_page_entries_.end();
+ ++p)
+ {
+ Got_page_entry* entry = *p;
+ if (this->got_page_entries_.find(entry) == this->got_page_entries_.end())
+ {
+ Got_page_entry* entry2 = new Got_page_entry(*entry);
+ this->got_page_entries_.insert(entry2);
+ this->page_gotno_ += entry->num_pages;
+ }
+ }
+}
+
+// Mips_output_data_got methods.
+
+// Lay out the GOT. Add local, global and TLS entries. If GOT is
+// larger than 64K, create multi-GOT.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_got<size, big_endian>::lay_out_got(Layout* layout,
+ Symbol_table* symtab, const Input_objects* input_objects)
+{
+ // Decide which symbols need to go in the global part of the GOT and
+ // count the number of reloc-only GOT symbols.
+ this->master_got_info_->count_got_symbols(symtab);
+
+ // Count the number of GOT entries.
+ this->master_got_info_->count_got_entries();
+
+ unsigned int got_size = this->master_got_info_->got_size();
+ if (got_size > Target_mips<size, big_endian>::MIPS_GOT_MAX_SIZE)
+ this->lay_out_multi_got(layout, input_objects);
+ else
+ {
+ // Record that all objects use single GOT.
+ for (Input_objects::Relobj_iterator p = input_objects->relobj_begin();
+ p != input_objects->relobj_end();
+ ++p)
+ {
+ Mips_relobj<size, big_endian>* object =
+ Mips_relobj<size, big_endian>::as_mips_relobj(*p);
+ if (object->get_got_info() != NULL)
+ object->set_got_info(this->master_got_info_);
+ }
+
+ this->master_got_info_->add_local_entries(this->target_, layout);
+ this->master_got_info_->add_global_entries(this->target_, layout,
+ /*not used*/-1U);
+ this->master_got_info_->add_tls_entries(this->target_, layout);
+ }
+}
+
+// Create multi-GOT. For every GOT, add local, global and TLS entries.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_got<size, big_endian>::lay_out_multi_got(Layout* layout,
+ const Input_objects* input_objects)
+{
+ // Try to merge the GOTs of input objects together, as long as they
+ // don't seem to exceed the maximum GOT size, choosing one of them
+ // to be the primary GOT.
+ this->merge_gots(input_objects);
+
+ // Every symbol that is referenced in a dynamic relocation must be
+ // present in the primary GOT.
+ this->primary_got_->set_global_gotno(this->master_got_info_->global_gotno());
+
+ // Add GOT entries.
+ unsigned int i = 0;
+ unsigned int offset = 0;
+ Mips_got_info<size, big_endian>* g = this->primary_got_;
+ do
+ {
+ g->set_index(i);
+ g->set_offset(offset);
+
+ g->add_local_entries(this->target_, layout);
+ if (i == 0)
+ g->add_global_entries(this->target_, layout,
+ (this->master_got_info_->global_gotno()
+ - this->master_got_info_->reloc_only_gotno()));
+ else
+ g->add_global_entries(this->target_, layout, /*not used*/-1U);
+ g->add_tls_entries(this->target_, layout);
+
+ // Forbid global symbols in every non-primary GOT from having
+ // lazy-binding stubs.
+ if (i > 0)
+ g->remove_lazy_stubs(this->target_);
+
+ ++i;
+ offset += g->got_size();
+ g = g->next();
+ }
+ while (g);
+}
+
+// Attempt to merge GOTs of different input objects. Try to use as much as
+// possible of the primary GOT, since it doesn't require explicit dynamic
+// relocations, but don't use objects that would reference global symbols
+// out of the addressable range. Failing the primary GOT, attempt to merge
+// with the current GOT, or finish the current GOT and then make make the new
+// GOT current.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_got<size, big_endian>::merge_gots(
+ const Input_objects* input_objects)
+{
+ gold_assert(this->primary_got_ == NULL);
+ Mips_got_info<size, big_endian>* current = NULL;
+
+ for (Input_objects::Relobj_iterator p = input_objects->relobj_begin();
+ p != input_objects->relobj_end();
+ ++p)
+ {
+ Mips_relobj<size, big_endian>* object =
+ Mips_relobj<size, big_endian>::as_mips_relobj(*p);
+
+ Mips_got_info<size, big_endian>* g = object->get_got_info();
+ if (g == NULL)
+ continue;
+
+ g->count_got_entries();
+
+ // Work out the number of page, local and TLS entries.
+ unsigned int estimate = this->master_got_info_->page_gotno();
+ if (estimate > g->page_gotno())
+ estimate = g->page_gotno();
+ estimate += g->local_gotno() + g->tls_gotno();
+
+ // We place TLS GOT entries after both locals and globals. The globals
+ // for the primary GOT may overflow the normal GOT size limit, so be
+ // sure not to merge a GOT which requires TLS with the primary GOT in that
+ // case. This doesn't affect non-primary GOTs.
+ estimate += (g->tls_gotno() > 0 ? this->master_got_info_->global_gotno()
+ : g->global_gotno());
+
+ unsigned int max_count =
+ Target_mips<size, big_endian>::MIPS_GOT_MAX_SIZE / (size/8) - 2;
+ if (estimate <= max_count)
+ {
+ // If we don't have a primary GOT, use it as
+ // a starting point for the primary GOT.
+ if (!this->primary_got_)
+ {
+ this->primary_got_ = g;
+ continue;
+ }
+
+ // Try merging with the primary GOT.
+ if (this->merge_got_with(g, object, this->primary_got_))
+ continue;
+ }
+
+ // If we can merge with the last-created GOT, do it.
+ if (current && this->merge_got_with(g, object, current))
+ continue;
+
+ // Well, we couldn't merge, so create a new GOT. Don't check if it
+ // fits; if it turns out that it doesn't, we'll get relocation
+ // overflows anyway.
+ g->set_next(current);
+ current = g;
+ }
+
+ // If we do not find any suitable primary GOT, create an empty one.
+ if (this->primary_got_ == NULL)
+ this->primary_got_ = new Mips_got_info<size, big_endian>();
+
+ // Link primary GOT with secondary GOTs.
+ this->primary_got_->set_next(current);
+}
+
+// Consider merging FROM, which is OBJECT's GOT, into TO. Return false if
+// this would lead to overflow, true if they were merged successfully.
+
+template<int size, bool big_endian>
+bool
+Mips_output_data_got<size, big_endian>::merge_got_with(
+ Mips_got_info<size, big_endian>* from,
+ Mips_relobj<size, big_endian>* object,
+ Mips_got_info<size, big_endian>* to)
+{
+ // Work out how many page entries we would need for the combined GOT.
+ unsigned int estimate = this->master_got_info_->page_gotno();
+ if (estimate >= from->page_gotno() + to->page_gotno())
+ estimate = from->page_gotno() + to->page_gotno();
+
+ // Conservatively estimate how many local and TLS entries would be needed.
+ estimate += from->local_gotno() + to->local_gotno();
+ estimate += from->tls_gotno() + to->tls_gotno();
+
+ // If we're merging with the primary got, any TLS relocations will
+ // come after the full set of global entries. Otherwise estimate those
+ // conservatively as well.
+ if (to == this->primary_got_ && (from->tls_gotno() + to->tls_gotno()) > 0)
+ estimate += this->master_got_info_->global_gotno();
+ else
+ estimate += from->global_gotno() + to->global_gotno();
+
+ // Bail out if the combined GOT might be too big.
+ unsigned int max_count =
+ Target_mips<size, big_endian>::MIPS_GOT_MAX_SIZE / (size/8) - 2;
+ if (estimate > max_count)
+ return false;
+
+ // Transfer the object's GOT information from FROM to TO.
+ to->add_got_entries(from);
+ to->add_got_page_entries(from);
+
+ // Record that OBJECT should use output GOT TO.
+ object->set_got_info(to);
+
+ return true;
+}
+
+// Write out the GOT.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_got<size, big_endian>::do_write(Output_file* of)
+{
+ // Call parent to write out GOT.
+ Output_data_got<size, big_endian>::do_write(of);
+
+ const off_t offset = this->offset();
+ const section_size_type oview_size =
+ convert_to_section_size_type(this->data_size());
+ unsigned char* const oview = of->get_output_view(offset, oview_size);
+
+ // Needed for fixing values of .got section.
+ this->got_view_ = oview;
+
+ // Write lazy stub addresses.
+ for (typename Unordered_set<Mips_symbol<size>*>::iterator
+ p = this->master_got_info_->global_got_symbols().begin();
+ p != this->master_got_info_->global_got_symbols().end();
+ ++p)
+ {
+ Mips_symbol<size>* mips_sym = *p;
+ if (mips_sym->has_lazy_stub())
+ {
+ Valtype* wv = reinterpret_cast<Valtype*>(
+ oview + this->get_primary_got_offset(mips_sym));
+ Valtype value =
+ this->target_->mips_stubs_section()->stub_address(mips_sym);
+ elfcpp::Swap<size, big_endian>::writeval(wv, value);
+ }
+ }
+
+ // Add +1 to GGA_NONE nonzero MIPS16 and microMIPS entries.
+ for (typename Unordered_set<Mips_symbol<size>*>::iterator
+ p = this->master_got_info_->global_got_symbols().begin();
+ p != this->master_got_info_->global_got_symbols().end();
+ ++p)
+ {
+ Mips_symbol<size>* mips_sym = *p;
+ if (!this->multi_got()
+ && (mips_sym->is_mips16() || mips_sym->is_micromips())
+ && mips_sym->global_got_area() == GGA_NONE
+ && mips_sym->has_got_offset(GOT_TYPE_STANDARD))
+ {
+ Valtype* wv = reinterpret_cast<Valtype*>(
+ oview + mips_sym->got_offset(GOT_TYPE_STANDARD));
+ Valtype value = elfcpp::Swap<size, big_endian>::readval(wv);
+ if (value != 0)
+ {
+ value |= 1;
+ elfcpp::Swap<size, big_endian>::writeval(wv, value);
+ }
+ }
+ }
+
+ if (!this->secondary_got_relocs_.empty())
+ {
+ // Fixup for the secondary GOT R_MIPS_REL32 relocs. For global
+ // secondary GOT entries with non-zero initial value copy the value
+ // to the corresponding primary GOT entry, and set the secondary GOT
+ // entry to zero.
+ // TODO(sasa): This is workaround. It needs to be investigated further.
+
+ for (size_t i = 0; i < this->secondary_got_relocs_.size(); ++i)
+ {
+ Static_reloc& reloc(this->secondary_got_relocs_[i]);
+ if (reloc.symbol_is_global())
+ {
+ Mips_symbol<size>* gsym = reloc.symbol();
+ gold_assert(gsym != NULL);
+
+ unsigned got_offset = reloc.got_offset();
+ gold_assert(got_offset < oview_size);
+
+ // Find primary GOT entry.
+ Valtype* wv_prim = reinterpret_cast<Valtype*>(
+ oview + this->get_primary_got_offset(gsym));
+
+ // Find secondary GOT entry.
+ Valtype* wv_sec = reinterpret_cast<Valtype*>(oview + got_offset);
+
+ Valtype value = elfcpp::Swap<size, big_endian>::readval(wv_sec);
+ if (value != 0)
+ {
+ elfcpp::Swap<size, big_endian>::writeval(wv_prim, value);
+ elfcpp::Swap<size, big_endian>::writeval(wv_sec, 0);
+ gsym->set_applied_secondary_got_fixup();
+ }
+ }
+ }
+
+ of->write_output_view(offset, oview_size, oview);
+ }
+
+ // We are done if there is no fix up.
+ if (this->static_relocs_.empty())
+ return;
+
+ Output_segment* tls_segment = this->layout_->tls_segment();
+ gold_assert(tls_segment != NULL);
+
+ for (size_t i = 0; i < this->static_relocs_.size(); ++i)
+ {
+ Static_reloc& reloc(this->static_relocs_[i]);
+
+ Mips_address value;
+ if (!reloc.symbol_is_global())
+ {
+ Sized_relobj_file<size, big_endian>* object = reloc.relobj();
+ const Symbol_value<size>* psymval =
+ object->local_symbol(reloc.index());
+
+ // We are doing static linking. Issue an error and skip this
+ // relocation if the symbol is undefined or in a discarded_section.
+ bool is_ordinary;
+ unsigned int shndx = psymval->input_shndx(&is_ordinary);
+ if ((shndx == elfcpp::SHN_UNDEF)
+ || (is_ordinary
+ && shndx != elfcpp::SHN_UNDEF
+ && !object->is_section_included(shndx)
+ && !this->symbol_table_->is_section_folded(object, shndx)))
+ {
+ gold_error(_("undefined or discarded local symbol %u from "
+ " object %s in GOT"),
+ reloc.index(), reloc.relobj()->name().c_str());
+ continue;
+ }
+
+ value = psymval->value(object, 0);
+ }
+ else
+ {
+ const Mips_symbol<size>* gsym = reloc.symbol();
+ gold_assert(gsym != NULL);
+
+ // We are doing static linking. Issue an error and skip this
+ // relocation if the symbol is undefined or in a discarded_section
+ // unless it is a weakly_undefined symbol.
+ if ((gsym->is_defined_in_discarded_section() || gsym->is_undefined())
+ && !gsym->is_weak_undefined())
+ {
+ gold_error(_("undefined or discarded symbol %s in GOT"),
+ gsym->name());
+ continue;
+ }
+
+ if (!gsym->is_weak_undefined())
+ value = gsym->value();
+ else
+ value = 0;
+ }
+
+ unsigned got_offset = reloc.got_offset();
+ gold_assert(got_offset < oview_size);
+
+ Valtype* wv = reinterpret_cast<Valtype*>(oview + got_offset);
+ Valtype x;
+
+ switch (reloc.r_type())
+ {
+ case elfcpp::R_MIPS_TLS_DTPMOD32:
+ case elfcpp::R_MIPS_TLS_DTPMOD64:
+ x = value;
+ break;
+ case elfcpp::R_MIPS_TLS_DTPREL32:
+ case elfcpp::R_MIPS_TLS_DTPREL64:
+ x = value - elfcpp::DTP_OFFSET;
+ break;
+ case elfcpp::R_MIPS_TLS_TPREL32:
+ case elfcpp::R_MIPS_TLS_TPREL64:
+ x = value - elfcpp::TP_OFFSET;
+ break;
+ default:
+ gold_unreachable();
+ break;
+ }
+
+ elfcpp::Swap<size, big_endian>::writeval(wv, x);
+ }
+
+ of->write_output_view(offset, oview_size, oview);
+}
+
+// Mips_relobj methods.
+
+// Count the local symbols. The Mips backend needs to know if a symbol
+// is a MIPS16 or microMIPS function or not. For global symbols, it is easy
+// because the Symbol object keeps the ELF symbol type and st_other field.
+// For local symbol it is harder because we cannot access this information.
+// So we override the do_count_local_symbol in parent and scan local symbols to
+// mark MIPS16 and microMIPS functions. This is not the most efficient way but
+// I do not want to slow down other ports by calling a per symbol target hook
+// inside Sized_relobj_file<size, big_endian>::do_count_local_symbols.
+
+template<int size, bool big_endian>
+void
+Mips_relobj<size, big_endian>::do_count_local_symbols(
+ Stringpool_template<char>* pool,
+ Stringpool_template<char>* dynpool)
+{
+ // Ask parent to count the local symbols.
+ Sized_relobj_file<size, big_endian>::do_count_local_symbols(pool, dynpool);
+ const unsigned int loccount = this->local_symbol_count();
+ if (loccount == 0)
+ return;
+
+ // Initialize the mips16 and micromips function bit-vector.
+ this->local_symbol_is_mips16_.resize(loccount, false);
+ this->local_symbol_is_micromips_.resize(loccount, false);
+
+ // Read the symbol table section header.
+ const unsigned int symtab_shndx = this->symtab_shndx();
+ elfcpp::Shdr<size, big_endian>
+ symtabshdr(this, this->elf_file()->section_header(symtab_shndx));
+ gold_assert(symtabshdr.get_sh_type() == elfcpp::SHT_SYMTAB);
+
+ // Read the local symbols.
+ const int sym_size = elfcpp::Elf_sizes<size>::sym_size;
+ gold_assert(loccount == symtabshdr.get_sh_info());
+ off_t locsize = loccount * sym_size;
+ const unsigned char* psyms = this->get_view(symtabshdr.get_sh_offset(),
+ locsize, true, true);
+
+ // Loop over the local symbols and mark any MIPS16 or microMIPS local symbols.
+
+ // Skip the first dummy symbol.
+ psyms += sym_size;
+ for (unsigned int i = 1; i < loccount; ++i, psyms += sym_size)
+ {
+ elfcpp::Sym<size, big_endian> sym(psyms);
+ unsigned char st_other = sym.get_st_other();
+ this->local_symbol_is_mips16_[i] = elfcpp::elf_st_is_mips16(st_other);
+ this->local_symbol_is_micromips_[i] =
+ elfcpp::elf_st_is_micromips(st_other);
+ }
+}
+
+// Read the symbol information.
+
+template<int size, bool big_endian>
+void
+Mips_relobj<size, big_endian>::do_read_symbols(Read_symbols_data* sd)
+{
+ // Call parent class to read symbol information.
+ Sized_relobj_file<size, big_endian>::do_read_symbols(sd);
+
+ // Read processor-specific flags in ELF file header.
+ const unsigned char* pehdr = this->get_view(elfcpp::file_header_offset,
+ elfcpp::Elf_sizes<size>::ehdr_size,
+ true, false);
+ elfcpp::Ehdr<size, big_endian> ehdr(pehdr);
+ this->processor_specific_flags_ = ehdr.get_e_flags();
+
+ // Get the section names.
+ const unsigned char* pnamesu = sd->section_names->data();
+ const char* pnames = reinterpret_cast<const char*>(pnamesu);
+
+ // Initialize the mips16 stub section bit-vectors.
+ this->section_is_mips16_fn_stub_.resize(this->shnum(), false);
+ this->section_is_mips16_call_stub_.resize(this->shnum(), false);
+ this->section_is_mips16_call_fp_stub_.resize(this->shnum(), false);
+
+ const size_t shdr_size = elfcpp::Elf_sizes<size>::shdr_size;
+ const unsigned char* pshdrs = sd->section_headers->data();
+ const unsigned char* ps = pshdrs + shdr_size;
+ for (unsigned int i = 1; i < this->shnum(); ++i, ps += shdr_size)
+ {
+ elfcpp::Shdr<size, big_endian> shdr(ps);
+
+ if (shdr.get_sh_type() == elfcpp::SHT_MIPS_REGINFO)
+ {
+ // Read the gp value that was used to create this object. We need the
+ // gp value while processing relocs. The .reginfo section is not used
+ // in the 64-bit MIPS ELF ABI.
+ section_offset_type section_offset = shdr.get_sh_offset();
+ section_size_type section_size =
+ convert_to_section_size_type(shdr.get_sh_size());
+ const unsigned char* view =
+ this->get_view(section_offset, section_size, true, false);
+
+ this->gp_ = elfcpp::Swap<size, big_endian>::readval(view + 20);
+ }
+
+ const char* name = pnames + shdr.get_sh_name();
+ this->section_is_mips16_fn_stub_[i] = is_prefix_of(".mips16.fn", name);
+ this->section_is_mips16_call_stub_[i] =
+ is_prefix_of(".mips16.call.", name);
+ this->section_is_mips16_call_fp_stub_[i] =
+ is_prefix_of(".mips16.call.fp.", name);
+
+ if (strcmp(name, ".pdr") == 0)
+ {
+ gold_assert(this->pdr_shndx_ == -1U);
+ this->pdr_shndx_ = i;
+ }
+ }
+}
+
+// Discard MIPS16 stub secions that are not needed.
+
+template<int size, bool big_endian>
+void
+Mips_relobj<size, big_endian>::discard_mips16_stub_sections(Symbol_table* symtab)
+{
+ for (typename Mips16_stubs_int_map::const_iterator
+ it = this->mips16_stub_sections_.begin();
+ it != this->mips16_stub_sections_.end(); ++it)
+ {
+ Mips16_stub_section<size, big_endian>* stub_section = it->second;
+ if (!stub_section->is_target_found())
+ {
+ gold_error(_("no relocation found in mips16 stub section '%s'"),
+ stub_section->object()
+ ->section_name(stub_section->shndx()).c_str());
+ }
+
+ bool discard = false;
+ if (stub_section->is_for_local_function())
+ {
+ if (stub_section->is_fn_stub())
+ {
+ // This stub is for a local symbol. This stub will only
+ // be needed if there is some relocation in this object,
+ // other than a 16 bit function call, which refers to this
+ // symbol.
+ if (!this->has_local_non_16bit_call_relocs(stub_section->r_sym()))
+ discard = true;
+ else
+ this->add_local_mips16_fn_stub(stub_section);
+ }
+ else
+ {
+ // This stub is for a local symbol. This stub will only
+ // be needed if there is some relocation (R_MIPS16_26) in
+ // this object that refers to this symbol.
+ gold_assert(stub_section->is_call_stub()
+ || stub_section->is_call_fp_stub());
+ if (!this->has_local_16bit_call_relocs(stub_section->r_sym()))
+ discard = true;
+ else
+ this->add_local_mips16_call_stub(stub_section);
+ }
+ }
+ else
+ {
+ Mips_symbol<size>* gsym = stub_section->gsym();
+ if (stub_section->is_fn_stub())
+ {
+ if (gsym->has_mips16_fn_stub())
+ // We already have a stub for this function.
+ discard = true;
+ else
+ {
+ gsym->set_mips16_fn_stub(stub_section);
+ if (gsym->should_add_dynsym_entry(symtab))
+ {
+ // If we have a MIPS16 function with a stub, the
+ // dynamic symbol must refer to the stub, since only
+ // the stub uses the standard calling conventions.
+ gsym->set_need_fn_stub();
+ if (gsym->is_from_dynobj())
+ gsym->set_needs_dynsym_value();
+ }
+ }
+ if (!gsym->need_fn_stub())
+ discard = true;
+ }
+ else if (stub_section->is_call_stub())
+ {
+ if (gsym->is_mips16())
+ // We don't need the call_stub; this is a 16 bit
+ // function, so calls from other 16 bit functions are
+ // OK.
+ discard = true;
+ else if (gsym->has_mips16_call_stub())
+ // We already have a stub for this function.
+ discard = true;
+ else
+ gsym->set_mips16_call_stub(stub_section);
+ }
+ else
+ {
+ gold_assert(stub_section->is_call_fp_stub());
+ if (gsym->is_mips16())
+ // We don't need the call_stub; this is a 16 bit
+ // function, so calls from other 16 bit functions are
+ // OK.
+ discard = true;
+ else if (gsym->has_mips16_call_fp_stub())
+ // We already have a stub for this function.
+ discard = true;
+ else
+ gsym->set_mips16_call_fp_stub(stub_section);
+ }
+ }
+ if (discard)
+ this->set_output_section(stub_section->shndx(), NULL);
+ }
+}
+
+// Mips_output_data_la25_stub methods.
+
+// Template for standard LA25 stub.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_la25_stub<size, big_endian>::la25_stub_entry[] =
+{
+ 0x3c190000, // lui $25,%hi(func)
+ 0x08000000, // j func
+ 0x27390000, // add $25,$25,%lo(func)
+ 0x00000000 // nop
+};
+
+// Template for microMIPS LA25 stub.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_la25_stub<size, big_endian>::la25_stub_micromips_entry[] =
+{
+ 0x41b9, 0x0000, // lui t9,%hi(func)
+ 0xd400, 0x0000, // j func
+ 0x3339, 0x0000, // addiu t9,t9,%lo(func)
+ 0x0000, 0x0000 // nop
+};
+
+// Create la25 stub for a symbol.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_la25_stub<size, big_endian>::create_la25_stub(
+ Symbol_table* symtab, Target_mips<size, big_endian>* target,
+ Mips_symbol<size>* gsym)
+{
+ if (!gsym->has_la25_stub())
+ {
+ gsym->set_la25_stub_offset(this->symbols_.size() * 16);
+ this->symbols_.insert(gsym);
+ this->create_stub_symbol(gsym, symtab, target, 16);
+ }
+}
+
+// Create a symbol for SYM stub's value and size, to help make the disassembly
+// easier to read.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_la25_stub<size, big_endian>::create_stub_symbol(
+ Mips_symbol<size>* sym, Symbol_table* symtab,
+ Target_mips<size, big_endian>* target, uint64_t symsize)
+{
+ std::string name(".pic.");
+ name += sym->name();
+
+ unsigned int offset = sym->la25_stub_offset();
+ if (sym->is_micromips())
+ offset |= 1;
+
+ // Make it a local function.
+ Symbol* new_sym = symtab->define_in_output_data(name.c_str(), NULL,
+ Symbol_table::PREDEFINED,
+ target->la25_stub_section(),
+ offset, symsize, elfcpp::STT_FUNC,
+ elfcpp::STB_LOCAL,
+ elfcpp::STV_DEFAULT, 0,
+ false, false);
+ new_sym->set_is_forced_local();
+}
+
+// Write out la25 stubs. This uses the hand-coded instructions above,
+// and adjusts them as needed.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_la25_stub<size, big_endian>::do_write(Output_file* of)
+{
+ const off_t offset = this->offset();
+ const section_size_type oview_size =
+ convert_to_section_size_type(this->data_size());
+ unsigned char* const oview = of->get_output_view(offset, oview_size);
+
+ for (typename Unordered_set<Mips_symbol<size>*>::iterator
+ p = this->symbols_.begin();
+ p != this->symbols_.end();
+ ++p)
+ {
+ Mips_symbol<size>* sym = *p;
+ unsigned char* pov = oview + sym->la25_stub_offset();
+
+ Mips_address target = sym->value();
+ if (!sym->is_micromips())
+ {
+ elfcpp::Swap<32, big_endian>::writeval(pov,
+ la25_stub_entry[0] | (((target + 0x8000) >> 16) & 0xffff));
+ elfcpp::Swap<32, big_endian>::writeval(pov + 4,
+ la25_stub_entry[1] | ((target >> 2) & 0x3ffffff));
+ elfcpp::Swap<32, big_endian>::writeval(pov + 8,
+ la25_stub_entry[2] | (target & 0xffff));
+ elfcpp::Swap<32, big_endian>::writeval(pov + 12, la25_stub_entry[3]);
+ }
+ else
+ {
+ target |= 1;
+ // First stub instruction. Paste high 16-bits of the target.
+ elfcpp::Swap<16, big_endian>::writeval(pov,
+ la25_stub_micromips_entry[0]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2,
+ ((target + 0x8000) >> 16) & 0xffff);
+ // Second stub instruction. Paste low 26-bits of the target, shifted
+ // right by 1.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 4,
+ la25_stub_micromips_entry[2] | ((target >> 17) & 0x3ff));
+ elfcpp::Swap<16, big_endian>::writeval(pov + 6,
+ la25_stub_micromips_entry[3] | ((target >> 1) & 0xffff));
+ // Third stub instruction. Paste low 16-bits of the target.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 8,
+ la25_stub_micromips_entry[4]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 10, target & 0xffff);
+ // Fourth stub instruction.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 12,
+ la25_stub_micromips_entry[6]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 14,
+ la25_stub_micromips_entry[7]);
+ }
+ }
+
+ of->write_output_view(offset, oview_size, oview);
+}
+
+// Mips_output_data_plt methods.
+
+// The format of the first PLT entry in an O32 executable.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::plt0_entry_o32[] =
+{
+ 0x3c1c0000, // lui $28, %hi(&GOTPLT[0])
+ 0x8f990000, // lw $25, %lo(&GOTPLT[0])($28)
+ 0x279c0000, // addiu $28, $28, %lo(&GOTPLT[0])
+ 0x031cc023, // subu $24, $24, $28
+ 0x03e07821, // move $15, $31 # 32-bit move (addu)
+ 0x0018c082, // srl $24, $24, 2
+ 0x0320f809, // jalr $25
+ 0x2718fffe // subu $24, $24, 2
+};
+
+// The format of the first PLT entry in an N32 executable. Different
+// because gp ($28) is not available; we use t2 ($14) instead.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::plt0_entry_n32[] =
+{
+ 0x3c0e0000, // lui $14, %hi(&GOTPLT[0])
+ 0x8dd90000, // lw $25, %lo(&GOTPLT[0])($14)
+ 0x25ce0000, // addiu $14, $14, %lo(&GOTPLT[0])
+ 0x030ec023, // subu $24, $24, $14
+ 0x03e07821, // move $15, $31 # 32-bit move (addu)
+ 0x0018c082, // srl $24, $24, 2
+ 0x0320f809, // jalr $25
+ 0x2718fffe // subu $24, $24, 2
+};
+
+// The format of the first PLT entry in an N64 executable. Different
+// from N32 because of the increased size of GOT entries.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::plt0_entry_n64[] =
+{
+ 0x3c0e0000, // lui $14, %hi(&GOTPLT[0])
+ 0xddd90000, // ld $25, %lo(&GOTPLT[0])($14)
+ 0x25ce0000, // addiu $14, $14, %lo(&GOTPLT[0])
+ 0x030ec023, // subu $24, $24, $14
+ 0x03e07821, // move $15, $31 # 64-bit move (daddu)
+ 0x0018c0c2, // srl $24, $24, 3
+ 0x0320f809, // jalr $25
+ 0x2718fffe // subu $24, $24, 2
+};
+
+// The format of the microMIPS first PLT entry in an O32 executable.
+// We rely on v0 ($2) rather than t8 ($24) to contain the address
+// of the GOTPLT entry handled, so this stub may only be used when
+// all the subsequent PLT entries are microMIPS code too.
+//
+// The trailing NOP is for alignment and correct disassembly only.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::
+plt0_entry_micromips_o32[] =
+{
+ 0x7980, 0x0000, // addiupc $3, (&GOTPLT[0]) - .
+ 0xff23, 0x0000, // lw $25, 0($3)
+ 0x0535, // subu $2, $2, $3
+ 0x2525, // srl $2, $2, 2
+ 0x3302, 0xfffe, // subu $24, $2, 2
+ 0x0dff, // move $15, $31
+ 0x45f9, // jalrs $25
+ 0x0f83, // move $28, $3
+ 0x0c00 // nop
+};
+
+// The format of the microMIPS first PLT entry in an O32 executable
+// in the insn32 mode.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::
+plt0_entry_micromips32_o32[] =
+{
+ 0x41bc, 0x0000, // lui $28, %hi(&GOTPLT[0])
+ 0xff3c, 0x0000, // lw $25, %lo(&GOTPLT[0])($28)
+ 0x339c, 0x0000, // addiu $28, $28, %lo(&GOTPLT[0])
+ 0x0398, 0xc1d0, // subu $24, $24, $28
+ 0x001f, 0x7950, // move $15, $31
+ 0x0318, 0x1040, // srl $24, $24, 2
+ 0x03f9, 0x0f3c, // jalr $25
+ 0x3318, 0xfffe // subu $24, $24, 2
+};
+
+// The format of subsequent standard entries in the PLT.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::plt_entry[] =
+{
+ 0x3c0f0000, // lui $15, %hi(.got.plt entry)
+ 0x8df90000, // l[wd] $25, %lo(.got.plt entry)($15)
+ 0x03200008, // jr $25
+ 0x25f80000 // addiu $24, $15, %lo(.got.plt entry)
+};
+
+// The format of subsequent MIPS16 o32 PLT entries. We use v1 ($3) as a
+// temporary because t8 ($24) and t9 ($25) are not directly addressable.
+// Note that this differs from the GNU ld which uses both v0 ($2) and v1 ($3).
+// We cannot use v0 because MIPS16 call stubs from the CS toolchain expect
+// target function address in register v0.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::plt_entry_mips16_o32[] =
+{
+ 0xb303, // lw $3, 12($pc)
+ 0x651b, // move $24, $3
+ 0x9b60, // lw $3, 0($3)
+ 0xeb00, // jr $3
+ 0x653b, // move $25, $3
+ 0x6500, // nop
+ 0x0000, 0x0000 // .word (.got.plt entry)
+};
+
+// The format of subsequent microMIPS o32 PLT entries. We use v0 ($2)
+// as a temporary because t8 ($24) is not addressable with ADDIUPC.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::
+plt_entry_micromips_o32[] =
+{
+ 0x7900, 0x0000, // addiupc $2, (.got.plt entry) - .
+ 0xff22, 0x0000, // lw $25, 0($2)
+ 0x4599, // jr $25
+ 0x0f02 // move $24, $2
+};
+
+// The format of subsequent microMIPS o32 PLT entries in the insn32 mode.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::
+plt_entry_micromips32_o32[] =
+{
+ 0x41af, 0x0000, // lui $15, %hi(.got.plt entry)
+ 0xff2f, 0x0000, // lw $25, %lo(.got.plt entry)($15)
+ 0x0019, 0x0f3c, // jr $25
+ 0x330f, 0x0000 // addiu $24, $15, %lo(.got.plt entry)
+};
+
+// Add an entry to the PLT for a symbol referenced by r_type relocation.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_plt<size, big_endian>::add_entry(Mips_symbol<size>* gsym,
+ unsigned int r_type)
+{
+ gold_assert(!gsym->has_plt_offset());
+
+ // Final PLT offset for a symbol will be set in method set_plt_offsets().
+ gsym->set_plt_offset(this->entry_count() * sizeof(plt_entry)
+ + sizeof(plt0_entry_o32));
+ this->symbols_.push_back(gsym);
+
+ // Record whether the relocation requires a standard MIPS
+ // or a compressed code entry.
+ if (jal_reloc(r_type))
+ {
+ if (r_type == elfcpp::R_MIPS_26)
+ gsym->set_needs_mips_plt(true);
+ else
+ gsym->set_needs_comp_plt(true);
+ }
+
+ section_offset_type got_offset = this->got_plt_->current_data_size();
+
+ // Every PLT entry needs a GOT entry which points back to the PLT
+ // entry (this will be changed by the dynamic linker, normally
+ // lazily when the function is called).
+ this->got_plt_->set_current_data_size(got_offset + size/8);
+
+ gsym->set_needs_dynsym_entry();
+ this->rel_->add_global(gsym, elfcpp::R_MIPS_JUMP_SLOT, this->got_plt_,
+ got_offset);
+}
+
+// Set final PLT offsets. For each symbol, determine whether standard or
+// compressed (MIPS16 or microMIPS) PLT entry is used.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_plt<size, big_endian>::set_plt_offsets()
+{
+ // The sizes of individual PLT entries.
+ unsigned int plt_mips_entry_size = this->standard_plt_entry_size();
+ unsigned int plt_comp_entry_size = (!this->target_->is_output_newabi()
+ ? this->compressed_plt_entry_size() : 0);
+
+ for (typename std::vector<Mips_symbol<size>*>::const_iterator
+ p = this->symbols_.begin(); p != this->symbols_.end(); ++p)
+ {
+ Mips_symbol<size>* mips_sym = *p;
+
+ // There are no defined MIPS16 or microMIPS PLT entries for n32 or n64,
+ // so always use a standard entry there.
+ //
+ // If the symbol has a MIPS16 call stub and gets a PLT entry, then
+ // all MIPS16 calls will go via that stub, and there is no benefit
+ // to having a MIPS16 entry. And in the case of call_stub a
+ // standard entry actually has to be used as the stub ends with a J
+ // instruction.
+ if (this->target_->is_output_newabi()
+ || mips_sym->has_mips16_call_stub()
+ || mips_sym->has_mips16_call_fp_stub())
+ {
+ mips_sym->set_needs_mips_plt(true);
+ mips_sym->set_needs_comp_plt(false);
+ }
+
+ // Otherwise, if there are no direct calls to the function, we
+ // have a free choice of whether to use standard or compressed
+ // entries. Prefer microMIPS entries if the object is known to
+ // contain microMIPS code, so that it becomes possible to create
+ // pure microMIPS binaries. Prefer standard entries otherwise,
+ // because MIPS16 ones are no smaller and are usually slower.
+ if (!mips_sym->needs_mips_plt() && !mips_sym->needs_comp_plt())
+ {
+ if (this->target_->is_output_micromips())
+ mips_sym->set_needs_comp_plt(true);
+ else
+ mips_sym->set_needs_mips_plt(true);
+ }
+
+ if (mips_sym->needs_mips_plt())
+ {
+ mips_sym->set_mips_plt_offset(this->plt_mips_offset_);
+ this->plt_mips_offset_ += plt_mips_entry_size;
+ }
+ if (mips_sym->needs_comp_plt())
+ {
+ mips_sym->set_comp_plt_offset(this->plt_comp_offset_);
+ this->plt_comp_offset_ += plt_comp_entry_size;
+ }
+ }
+
+ // Figure out the size of the PLT header if we know that we are using it.
+ if (this->plt_mips_offset_ + this->plt_comp_offset_ != 0)
+ this->plt_header_size_ = this->get_plt_header_size();
+}
+
+// Write out the PLT. This uses the hand-coded instructions above,
+// and adjusts them as needed.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_plt<size, big_endian>::do_write(Output_file* of)
+{
+ const off_t offset = this->offset();
+ const section_size_type oview_size =
+ convert_to_section_size_type(this->data_size());
+ unsigned char* const oview = of->get_output_view(offset, oview_size);
+
+ const off_t gotplt_file_offset = this->got_plt_->offset();
+ const section_size_type gotplt_size =
+ convert_to_section_size_type(this->got_plt_->data_size());
+ unsigned char* const gotplt_view = of->get_output_view(gotplt_file_offset,
+ gotplt_size);
+ unsigned char* pov = oview;
+
+ Mips_address plt_address = this->address();
+
+ // Calculate the address of .got.plt.
+ Mips_address gotplt_addr = this->got_plt_->address();
+ Mips_address gotplt_addr_high = ((gotplt_addr + 0x8000) >> 16) & 0xffff;
+ Mips_address gotplt_addr_low = gotplt_addr & 0xffff;
+
+ // The PLT sequence is not safe for N64 if .got.plt's address can
+ // not be loaded in two instructions.
+ gold_assert((gotplt_addr & ~(Mips_address) 0x7fffffff) == 0
+ || ~(gotplt_addr | 0x7fffffff) == 0);
+
+ // Write the PLT header.
+ const uint32_t* plt0_entry = this->get_plt_header_entry();
+ if (plt0_entry == plt0_entry_micromips_o32)
+ {
+ // Write microMIPS PLT header.
+ gold_assert(gotplt_addr % 4 == 0);
+
+ Mips_address gotpc_offset = gotplt_addr - ((plt_address | 3) ^ 3);
+
+ // ADDIUPC has a span of +/-16MB, check we're in range.
+ if (gotpc_offset + 0x1000000 >= 0x2000000)
+ {
+ gold_error(_(".got.plt offset of %ld from .plt beyond the range of "
+ "ADDIUPC"), (long)gotpc_offset);
+ return;
+ }
+
+ elfcpp::Swap<16, big_endian>::writeval(pov,
+ plt0_entry[0] | ((gotpc_offset >> 18) & 0x7f));
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2,
+ (gotpc_offset >> 2) & 0xffff);
+ pov += 4;
+ for (unsigned int i = 2;
+ i < (sizeof(plt0_entry_micromips_o32)
+ / sizeof(plt0_entry_micromips_o32[0]));
+ i++)
+ {
+ elfcpp::Swap<16, big_endian>::writeval(pov, plt0_entry[i]);
+ pov += 2;
+ }
+ }
+ else if (plt0_entry == plt0_entry_micromips32_o32)
+ {
+ // Write microMIPS PLT header in insn32 mode.
+ elfcpp::Swap<16, big_endian>::writeval(pov, plt0_entry[0]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2, gotplt_addr_high);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 4, plt0_entry[2]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 6, gotplt_addr_low);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 8, plt0_entry[4]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 10, gotplt_addr_low);
+ pov += 12;
+ for (unsigned int i = 6;
+ i < (sizeof(plt0_entry_micromips32_o32)
+ / sizeof(plt0_entry_micromips32_o32[0]));
+ i++)
+ {
+ elfcpp::Swap<16, big_endian>::writeval(pov, plt0_entry[i]);
+ pov += 2;
+ }
+ }
+ else
+ {
+ // Write standard PLT header.
+ elfcpp::Swap<32, big_endian>::writeval(pov,
+ plt0_entry[0] | gotplt_addr_high);
+ elfcpp::Swap<32, big_endian>::writeval(pov + 4,
+ plt0_entry[1] | gotplt_addr_low);
+ elfcpp::Swap<32, big_endian>::writeval(pov + 8,
+ plt0_entry[2] | gotplt_addr_low);
+ pov += 12;
+ for (int i = 3; i < 8; i++)
+ {
+ elfcpp::Swap<32, big_endian>::writeval(pov, plt0_entry[i]);
+ pov += 4;
+ }
+ }
+
+
+ unsigned char* gotplt_pov = gotplt_view;
+ unsigned int got_entry_size = size/8; // TODO(sasa): MIPS_ELF_GOT_SIZE
+
+ // The first two entries in .got.plt are reserved.
+ elfcpp::Swap<size, big_endian>::writeval(gotplt_pov, 0);
+ elfcpp::Swap<size, big_endian>::writeval(gotplt_pov + got_entry_size, 0);
+
+ unsigned int gotplt_offset = 2 * got_entry_size;
+ gotplt_pov += 2 * got_entry_size;
+
+ // Calculate the address of the PLT header.
+ Mips_address header_address = (plt_address
+ + (this->is_plt_header_compressed() ? 1 : 0));
+
+ // Initialize compressed PLT area view.
+ unsigned char* pov2 = pov + this->plt_mips_offset_;
+
+ // Write the PLT entries.
+ for (typename std::vector<Mips_symbol<size>*>::const_iterator
+ p = this->symbols_.begin();
+ p != this->symbols_.end();
+ ++p, gotplt_pov += got_entry_size, gotplt_offset += got_entry_size)
+ {
+ Mips_symbol<size>* mips_sym = *p;
+
+ // Calculate the address of the .got.plt entry.
+ uint32_t gotplt_entry_addr = (gotplt_addr + gotplt_offset);
+ uint32_t gotplt_entry_addr_hi = (((gotplt_entry_addr + 0x8000) >> 16)
+ & 0xffff);
+ uint32_t gotplt_entry_addr_lo = gotplt_entry_addr & 0xffff;
+
+ // Initially point the .got.plt entry at the PLT header.
+ if (this->target_->is_output_n64())
+ elfcpp::Swap<64, big_endian>::writeval(gotplt_pov, header_address);
+ else
+ elfcpp::Swap<32, big_endian>::writeval(gotplt_pov, header_address);
+
+ // Now handle the PLT itself. First the standard entry.
+ if (mips_sym->has_mips_plt_offset())
+ {
+ // Pick the load opcode (LW or LD).
+ uint64_t load = this->target_->is_output_n64() ? 0xdc000000
+ : 0x8c000000;
+
+ // Fill in the PLT entry itself.
+ elfcpp::Swap<32, big_endian>::writeval(pov,
+ plt_entry[0] | gotplt_entry_addr_hi);
+ elfcpp::Swap<32, big_endian>::writeval(pov + 4,
+ plt_entry[1] | gotplt_entry_addr_lo | load);
+ elfcpp::Swap<32, big_endian>::writeval(pov + 8, plt_entry[2]);
+ elfcpp::Swap<32, big_endian>::writeval(pov + 12,
+ plt_entry[3] | gotplt_entry_addr_lo);
+ pov += 16;
+ }
+
+ // Now the compressed entry. They come after any standard ones.
+ if (mips_sym->has_comp_plt_offset())
+ {
+ if (!this->target_->is_output_micromips())
+ {
+ // Write MIPS16 PLT entry.
+ const uint32_t* plt_entry = plt_entry_mips16_o32;
+
+ elfcpp::Swap<16, big_endian>::writeval(pov2, plt_entry[0]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 2, plt_entry[1]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 4, plt_entry[2]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 6, plt_entry[3]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 8, plt_entry[4]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 10, plt_entry[5]);
+ elfcpp::Swap<32, big_endian>::writeval(pov2 + 12,
+ gotplt_entry_addr);
+ pov2 += 16;
+ }
+ else if (this->target_->use_32bit_micromips_instructions())
+ {
+ // Write microMIPS PLT entry in insn32 mode.
+ const uint32_t* plt_entry = plt_entry_micromips32_o32;
+
+ elfcpp::Swap<16, big_endian>::writeval(pov2, plt_entry[0]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 2,
+ gotplt_entry_addr_hi);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 4, plt_entry[2]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 6,
+ gotplt_entry_addr_lo);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 8, plt_entry[4]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 10, plt_entry[5]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 12, plt_entry[6]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 14,
+ gotplt_entry_addr_lo);
+ pov2 += 16;
+ }
+ else
+ {
+ // Write microMIPS PLT entry.
+ const uint32_t* plt_entry = plt_entry_micromips_o32;
+
+ gold_assert(gotplt_entry_addr % 4 == 0);
+
+ Mips_address loc_address = plt_address + pov2 - oview;
+ int gotpc_offset = gotplt_entry_addr - ((loc_address | 3) ^ 3);
+
+ // ADDIUPC has a span of +/-16MB, check we're in range.
+ if (gotpc_offset + 0x1000000 >= 0x2000000)
+ {
+ gold_error(_(".got.plt offset of %ld from .plt beyond the "
+ "range of ADDIUPC"), (long)gotpc_offset);
+ return;
+ }
+
+ elfcpp::Swap<16, big_endian>::writeval(pov2,
+ plt_entry[0] | ((gotpc_offset >> 18) & 0x7f));
+ elfcpp::Swap<16, big_endian>::writeval(
+ pov2 + 2, (gotpc_offset >> 2) & 0xffff);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 4, plt_entry[2]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 6, plt_entry[3]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 8, plt_entry[4]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 10, plt_entry[5]);
+ pov2 += 12;
+ }
+ }
+ }
+
+ // Check the number of bytes written for standard entries.
+ gold_assert(static_cast<section_size_type>(
+ pov - oview - this->plt_header_size_) == this->plt_mips_offset_);
+ // Check the number of bytes written for compressed entries.
+ gold_assert((static_cast<section_size_type>(pov2 - pov)
+ == this->plt_comp_offset_));
+ // Check the total number of bytes written.
+ gold_assert(static_cast<section_size_type>(pov2 - oview) == oview_size);
+
+ gold_assert(static_cast<section_size_type>(gotplt_pov - gotplt_view)
+ == gotplt_size);
+
+ of->write_output_view(offset, oview_size, oview);
+ of->write_output_view(gotplt_file_offset, gotplt_size, gotplt_view);
+}
+
+// Mips_output_data_mips_stubs methods.
+
+// The format of the lazy binding stub when dynamic symbol count is less than
+// 64K, dynamic symbol index is less than 32K, and ABI is not N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_normal_1[4] =
+{
+ 0x8f998010, // lw t9,0x8010(gp)
+ 0x03e07821, // addu t7,ra,zero
+ 0x0320f809, // jalr t9,ra
+ 0x24180000 // addiu t8,zero,DYN_INDEX sign extended
+};
+
+// The format of the lazy binding stub when dynamic symbol count is less than
+// 64K, dynamic symbol index is less than 32K, and ABI is N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_normal_1_n64[4] =
+{
+ 0xdf998010, // ld t9,0x8010(gp)
+ 0x03e0782d, // daddu t7,ra,zero
+ 0x0320f809, // jalr t9,ra
+ 0x64180000 // daddiu t8,zero,DYN_INDEX sign extended
+};
+
+// The format of the lazy binding stub when dynamic symbol count is less than
+// 64K, dynamic symbol index is between 32K and 64K, and ABI is not N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_normal_2[4] =
+{
+ 0x8f998010, // lw t9,0x8010(gp)
+ 0x03e07821, // addu t7,ra,zero
+ 0x0320f809, // jalr t9,ra
+ 0x34180000 // ori t8,zero,DYN_INDEX unsigned
+};
+
+// The format of the lazy binding stub when dynamic symbol count is less than
+// 64K, dynamic symbol index is between 32K and 64K, and ABI is N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_normal_2_n64[4] =
+{
+ 0xdf998010, // ld t9,0x8010(gp)
+ 0x03e0782d, // daddu t7,ra,zero
+ 0x0320f809, // jalr t9,ra
+ 0x34180000 // ori t8,zero,DYN_INDEX unsigned
+};
+
+// The format of the lazy binding stub when dynamic symbol count is greater than
+// 64K, and ABI is not N64.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_big[5] =
+{
+ 0x8f998010, // lw t9,0x8010(gp)
+ 0x03e07821, // addu t7,ra,zero
+ 0x3c180000, // lui t8,DYN_INDEX
+ 0x0320f809, // jalr t9,ra
+ 0x37180000 // ori t8,t8,DYN_INDEX
+};
+
+// The format of the lazy binding stub when dynamic symbol count is greater than
+// 64K, and ABI is N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_big_n64[5] =
+{
+ 0xdf998010, // ld t9,0x8010(gp)
+ 0x03e0782d, // daddu t7,ra,zero
+ 0x3c180000, // lui t8,DYN_INDEX
+ 0x0320f809, // jalr t9,ra
+ 0x37180000 // ori t8,t8,DYN_INDEX
+};
+
+// microMIPS stubs.
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// less than 64K, dynamic symbol index is less than 32K, and ABI is not N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_micromips_normal_1[] =
+{
+ 0xff3c, 0x8010, // lw t9,0x8010(gp)
+ 0x0dff, // move t7,ra
+ 0x45d9, // jalr t9
+ 0x3300, 0x0000 // addiu t8,zero,DYN_INDEX sign extended
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// less than 64K, dynamic symbol index is less than 32K, and ABI is N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::
+lazy_stub_micromips_normal_1_n64[] =
+{
+ 0xdf3c, 0x8010, // ld t9,0x8010(gp)
+ 0x0dff, // move t7,ra
+ 0x45d9, // jalr t9
+ 0x5f00, 0x0000 // daddiu t8,zero,DYN_INDEX sign extended
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol
+// count is less than 64K, dynamic symbol index is between 32K and 64K,
+// and ABI is not N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_micromips_normal_2[] =
+{
+ 0xff3c, 0x8010, // lw t9,0x8010(gp)
+ 0x0dff, // move t7,ra
+ 0x45d9, // jalr t9
+ 0x5300, 0x0000 // ori t8,zero,DYN_INDEX unsigned
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol
+// count is less than 64K, dynamic symbol index is between 32K and 64K,
+// and ABI is N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::
+lazy_stub_micromips_normal_2_n64[] =
+{
+ 0xdf3c, 0x8010, // ld t9,0x8010(gp)
+ 0x0dff, // move t7,ra
+ 0x45d9, // jalr t9
+ 0x5300, 0x0000 // ori t8,zero,DYN_INDEX unsigned
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// greater than 64K, and ABI is not N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_micromips_big[] =
+{
+ 0xff3c, 0x8010, // lw t9,0x8010(gp)
+ 0x0dff, // move t7,ra
+ 0x41b8, 0x0000, // lui t8,DYN_INDEX
+ 0x45d9, // jalr t9
+ 0x5318, 0x0000 // ori t8,t8,DYN_INDEX
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// greater than 64K, and ABI is N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_micromips_big_n64[] =
+{
+ 0xdf3c, 0x8010, // ld t9,0x8010(gp)
+ 0x0dff, // move t7,ra
+ 0x41b8, 0x0000, // lui t8,DYN_INDEX
+ 0x45d9, // jalr t9
+ 0x5318, 0x0000 // ori t8,t8,DYN_INDEX
+};
+
+// 32-bit microMIPS stubs.
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// less than 64K, dynamic symbol index is less than 32K, ABI is not N64, and we
+// can use only 32-bit instructions.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::
+lazy_stub_micromips32_normal_1[] =
+{
+ 0xff3c, 0x8010, // lw t9,0x8010(gp)
+ 0x001f, 0x7950, // addu t7,ra,zero
+ 0x03f9, 0x0f3c, // jalr ra,t9
+ 0x3300, 0x0000 // addiu t8,zero,DYN_INDEX sign extended
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// less than 64K, dynamic symbol index is less than 32K, ABI is N64, and we can
+// use only 32-bit instructions.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::
+lazy_stub_micromips32_normal_1_n64[] =
+{
+ 0xdf3c, 0x8010, // ld t9,0x8010(gp)
+ 0x581f, 0x7950, // daddu t7,ra,zero
+ 0x03f9, 0x0f3c, // jalr ra,t9
+ 0x5f00, 0x0000 // daddiu t8,zero,DYN_INDEX sign extended
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol
+// count is less than 64K, dynamic symbol index is between 32K and 64K,
+// ABI is not N64, and we can use only 32-bit instructions.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::
+lazy_stub_micromips32_normal_2[] =
+{
+ 0xff3c, 0x8010, // lw t9,0x8010(gp)
+ 0x001f, 0x7950, // addu t7,ra,zero
+ 0x03f9, 0x0f3c, // jalr ra,t9
+ 0x5300, 0x0000 // ori t8,zero,DYN_INDEX unsigned
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol
+// count is less than 64K, dynamic symbol index is between 32K and 64K,
+// ABI is N64, and we can use only 32-bit instructions.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::
+lazy_stub_micromips32_normal_2_n64[] =
+{
+ 0xdf3c, 0x8010, // ld t9,0x8010(gp)
+ 0x581f, 0x7950, // daddu t7,ra,zero
+ 0x03f9, 0x0f3c, // jalr ra,t9
+ 0x5300, 0x0000 // ori t8,zero,DYN_INDEX unsigned
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// greater than 64K, ABI is not N64, and we can use only 32-bit instructions.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_micromips32_big[] =
+{
+ 0xff3c, 0x8010, // lw t9,0x8010(gp)
+ 0x001f, 0x7950, // addu t7,ra,zero
+ 0x41b8, 0x0000, // lui t8,DYN_INDEX
+ 0x03f9, 0x0f3c, // jalr ra,t9
+ 0x5318, 0x0000 // ori t8,t8,DYN_INDEX
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// greater than 64K, ABI is N64, and we can use only 32-bit instructions.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_micromips32_big_n64[] =
+{
+ 0xdf3c, 0x8010, // ld t9,0x8010(gp)
+ 0x581f, 0x7950, // daddu t7,ra,zero
+ 0x41b8, 0x0000, // lui t8,DYN_INDEX
+ 0x03f9, 0x0f3c, // jalr ra,t9
+ 0x5318, 0x0000 // ori t8,t8,DYN_INDEX
+};
+
+// Create entry for a symbol.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_mips_stubs<size, big_endian>::make_entry(
+ Mips_symbol<size>* gsym)
+{
+ if (!gsym->has_lazy_stub() && !gsym->has_plt_offset())
+ {
+ this->symbols_.insert(gsym);
+ gsym->set_has_lazy_stub(true);
+ }
+}
+
+// Remove entry for a symbol.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_mips_stubs<size, big_endian>::remove_entry(
+ Mips_symbol<size>* gsym)
+{
+ if (gsym->has_lazy_stub())
+ {
+ this->symbols_.erase(gsym);
+ gsym->set_has_lazy_stub(false);
+ }
+}
+
+// Set stub offsets for symbols. This method expects that the number of
+// entries in dynamic symbol table is set.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_mips_stubs<size, big_endian>::set_lazy_stub_offsets()
+{
+ gold_assert(this->dynsym_count_ != -1U);
+
+ if (this->stub_offsets_are_set_)
+ return;
+
+ unsigned int stub_size = this->stub_size();
+ unsigned int offset = 0;
+ for (typename Unordered_set<Mips_symbol<size>*>::const_iterator
+ p = this->symbols_.begin();
+ p != this->symbols_.end();
+ ++p, offset += stub_size)
+ {
+ Mips_symbol<size>* mips_sym = *p;
+ mips_sym->set_lazy_stub_offset(offset);
+ }
+ this->stub_offsets_are_set_ = true;
+}
+
+template<int size, bool big_endian>
+void
+Mips_output_data_mips_stubs<size, big_endian>::set_needs_dynsym_value()
+{
+ for (typename Unordered_set<Mips_symbol<size>*>::const_iterator
+ p = this->symbols_.begin(); p != this->symbols_.end(); ++p)
+ {
+ Mips_symbol<size>* sym = *p;
+ if (sym->is_from_dynobj())
+ sym->set_needs_dynsym_value();
+ }
+}
+
+// Write out the .MIPS.stubs. This uses the hand-coded instructions and
+// adjusts them as needed.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_mips_stubs<size, big_endian>::do_write(Output_file* of)
+{
+ const off_t offset = this->offset();
+ const section_size_type oview_size =
+ convert_to_section_size_type(this->data_size());
+ unsigned char* const oview = of->get_output_view(offset, oview_size);
+
+ bool big_stub = this->dynsym_count_ > 0x10000;
+
+ unsigned char* pov = oview;
+ for (typename Unordered_set<Mips_symbol<size>*>::const_iterator
+ p = this->symbols_.begin(); p != this->symbols_.end(); ++p)
+ {
+ Mips_symbol<size>* sym = *p;
+ const uint32_t* lazy_stub;
+ bool n64 = this->target_->is_output_n64();
+
+ if (!this->target_->is_output_micromips())
+ {
+ // Write standard (non-microMIPS) stub.
+ if (!big_stub)
+ {
+ if (sym->dynsym_index() & ~0x7fff)
+ // Dynsym index is between 32K and 64K.
+ lazy_stub = n64 ? lazy_stub_normal_2_n64 : lazy_stub_normal_2;
+ else
+ // Dynsym index is less than 32K.
+ lazy_stub = n64 ? lazy_stub_normal_1_n64 : lazy_stub_normal_1;
+ }
+ else
+ lazy_stub = n64 ? lazy_stub_big_n64 : lazy_stub_big;
+
+ unsigned int i = 0;
+ elfcpp::Swap<32, big_endian>::writeval(pov, lazy_stub[i]);
+ elfcpp::Swap<32, big_endian>::writeval(pov + 4, lazy_stub[i + 1]);
+ pov += 8;
+
+ i += 2;
+ if (big_stub)
+ {
+ // LUI instruction of the big stub. Paste high 16 bits of the
+ // dynsym index.
+ elfcpp::Swap<32, big_endian>::writeval(pov,
+ lazy_stub[i] | ((sym->dynsym_index() >> 16) & 0x7fff));
+ pov += 4;
+ i += 1;
+ }
+ elfcpp::Swap<32, big_endian>::writeval(pov, lazy_stub[i]);
+ // Last stub instruction. Paste low 16 bits of the dynsym index.
+ elfcpp::Swap<32, big_endian>::writeval(pov + 4,
+ lazy_stub[i + 1] | (sym->dynsym_index() & 0xffff));
+ pov += 8;
+ }
+ else if (this->target_->use_32bit_micromips_instructions())
+ {
+ // Write microMIPS stub in insn32 mode.
+ if (!big_stub)
+ {
+ if (sym->dynsym_index() & ~0x7fff)
+ // Dynsym index is between 32K and 64K.
+ lazy_stub = n64 ? lazy_stub_micromips32_normal_2_n64
+ : lazy_stub_micromips32_normal_2;
+ else
+ // Dynsym index is less than 32K.
+ lazy_stub = n64 ? lazy_stub_micromips32_normal_1_n64
+ : lazy_stub_micromips32_normal_1;
+ }
+ else
+ lazy_stub = n64 ? lazy_stub_micromips32_big_n64
+ : lazy_stub_micromips32_big;
+
+ unsigned int i = 0;
+ // First stub instruction. We emit 32-bit microMIPS instructions by
+ // emitting two 16-bit parts because on microMIPS the 16-bit part of
+ // the instruction where the opcode is must always come first, for
+ // both little and big endian.
+ elfcpp::Swap<16, big_endian>::writeval(pov, lazy_stub[i]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2, lazy_stub[i + 1]);
+ // Second stub instruction.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 4, lazy_stub[i + 2]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 6, lazy_stub[i + 3]);
+ pov += 8;
+ i += 4;
+ if (big_stub)
+ {
+ // LUI instruction of the big stub. Paste high 16 bits of the
+ // dynsym index.
+ elfcpp::Swap<16, big_endian>::writeval(pov, lazy_stub[i]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2,
+ (sym->dynsym_index() >> 16) & 0x7fff);
+ pov += 4;
+ i += 2;
+ }
+ elfcpp::Swap<16, big_endian>::writeval(pov, lazy_stub[i]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2, lazy_stub[i + 1]);
+ // Last stub instruction. Paste low 16 bits of the dynsym index.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 4, lazy_stub[i + 2]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 6,
+ sym->dynsym_index() & 0xffff);
+ pov += 8;
+ }
+ else
+ {
+ // Write microMIPS stub.
+ if (!big_stub)
+ {
+ if (sym->dynsym_index() & ~0x7fff)
+ // Dynsym index is between 32K and 64K.
+ lazy_stub = n64 ? lazy_stub_micromips_normal_2_n64
+ : lazy_stub_micromips_normal_2;
+ else
+ // Dynsym index is less than 32K.
+ lazy_stub = n64 ? lazy_stub_micromips_normal_1_n64
+ : lazy_stub_micromips_normal_1;
+ }
+ else
+ lazy_stub = n64 ? lazy_stub_micromips_big_n64
+ : lazy_stub_micromips_big;
+
+ unsigned int i = 0;
+ // First stub instruction. We emit 32-bit microMIPS instructions by
+ // emitting two 16-bit parts because on microMIPS the 16-bit part of
+ // the instruction where the opcode is must always come first, for
+ // both little and big endian.
+ elfcpp::Swap<16, big_endian>::writeval(pov, lazy_stub[i]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2, lazy_stub[i + 1]);
+ // Second stub instruction.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 4, lazy_stub[i + 2]);
+ pov += 6;
+ i += 3;
+ if (big_stub)
+ {
+ // LUI instruction of the big stub. Paste high 16 bits of the
+ // dynsym index.
+ elfcpp::Swap<16, big_endian>::writeval(pov, lazy_stub[i]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2,
+ (sym->dynsym_index() >> 16) & 0x7fff);
+ pov += 4;
+ i += 2;
+ }
+ elfcpp::Swap<16, big_endian>::writeval(pov, lazy_stub[i]);
+ // Last stub instruction. Paste low 16 bits of the dynsym index.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2, lazy_stub[i + 1]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 4,
+ sym->dynsym_index() & 0xffff);
+ pov += 6;
+ }
+ }
+
+ // We always allocate 20 bytes for every stub, because final dynsym count is
+ // not known in method do_finalize_sections. There are 4 unused bytes per
+ // stub if final dynsym count is less than 0x10000.
+ unsigned int used = pov - oview;
+ unsigned int unused = big_stub ? 0 : this->symbols_.size() * 4;
+ gold_assert(static_cast<section_size_type>(used + unused) == oview_size);
+
+ // Fill the unused space with zeroes.
+ // TODO(sasa): Can we strip unused bytes during the relaxation?
+ unsigned char* end = oview + oview_size;
+ while (pov < end)
+ {
+ elfcpp::Swap<32, big_endian>::writeval(pov, 0);
+ pov += 4;
+ }
+
+ of->write_output_view(offset, oview_size, oview);
+}
+
+// Mips_output_section_reginfo methods.
+
+template<int size, bool big_endian>
+void
+Mips_output_section_reginfo<size, big_endian>::do_write(Output_file* of)
+{
+ Valtype gprmask = 0;
+ Valtype cprmask1 = 0;
+ Valtype cprmask2 = 0;
+ Valtype cprmask3 = 0;
+ Valtype cprmask4 = 0;
+
+ for (Input_section_list::const_iterator p = this->input_sections().begin();
+ p != this->input_sections().end();
+ ++p)
+ {
+ Relobj* relobj = p->relobj();
+ unsigned int shndx = p->shndx();
+
+ section_size_type section_size;
+ const unsigned char* section_contents =
+ relobj->section_contents(shndx, §ion_size, false);
+
+ gprmask |= elfcpp::Swap<size, big_endian>::readval(section_contents);
+ cprmask1 |= elfcpp::Swap<size, big_endian>::readval(section_contents + 4);
+ cprmask2 |= elfcpp::Swap<size, big_endian>::readval(section_contents + 8);
+ cprmask3 |=
+ elfcpp::Swap<size, big_endian>::readval(section_contents + 12);
+ cprmask4 |=
+ elfcpp::Swap<size, big_endian>::readval(section_contents + 16);
+ }
+
+ off_t offset = this->offset();
+ off_t data_size = this->data_size();
+
+ unsigned char* view = of->get_output_view(offset, data_size);
+ elfcpp::Swap<size, big_endian>::writeval(view, gprmask);
+ elfcpp::Swap<size, big_endian>::writeval(view + 4, cprmask1);
+ elfcpp::Swap<size, big_endian>::writeval(view + 8, cprmask2);
+ elfcpp::Swap<size, big_endian>::writeval(view + 12, cprmask3);
+ elfcpp::Swap<size, big_endian>::writeval(view + 16, cprmask4);
+ // Write the gp value.
+ elfcpp::Swap<size, big_endian>::writeval(view + 20,
+ this->target_->gp_value());
+
+ of->write_output_view(offset, data_size, view);
+}
+
+// Mips_copy_relocs methods.
+
+// Emit any saved relocs.
+
+template<int sh_type, int size, bool big_endian>
+void
+Mips_copy_relocs<sh_type, size, big_endian>::emit_mips(
+ Output_data_reloc<sh_type, true, size, big_endian>* reloc_section,
+ Symbol_table* symtab, Layout* layout, Target_mips<size, big_endian>* target)
+{
+ for (typename Copy_relocs<sh_type, size, big_endian>::
+ Copy_reloc_entries::iterator p = this->entries_.begin();
+ p != this->entries_.end();
+ ++p)
+ emit_entry(*p, reloc_section, symtab, layout, target);
+
+ // We no longer need the saved information.
+ this->entries_.clear();
+}
+
+// Emit the reloc if appropriate.
+
+template<int sh_type, int size, bool big_endian>
+void
+Mips_copy_relocs<sh_type, size, big_endian>::emit_entry(
+ Copy_reloc_entry& entry,
+ Output_data_reloc<sh_type, true, size, big_endian>* reloc_section,
+ Symbol_table* symtab, Layout* layout, Target_mips<size, big_endian>* target)
+{
+ // If the symbol is no longer defined in a dynamic object, then we
+ // emitted a COPY relocation, and we do not want to emit this
+ // dynamic relocation.
+ if (!entry.sym_->is_from_dynobj())
+ return;
+
+ bool can_make_dynamic = (entry.reloc_type_ == elfcpp::R_MIPS_32
+ || entry.reloc_type_ == elfcpp::R_MIPS_REL32
+ || entry.reloc_type_ == elfcpp::R_MIPS_64);
+
+ Mips_symbol<size>* sym = Mips_symbol<size>::as_mips_sym(entry.sym_);
+ if (can_make_dynamic && !sym->has_static_relocs())
+ {
+ Mips_relobj<size, big_endian>* object =
+ Mips_relobj<size, big_endian>::as_mips_relobj(entry.relobj_);
+ target->got_section(symtab, layout)->record_global_got_symbol(
+ sym, object, entry.reloc_type_, true, false);
+ if (!symbol_references_local(sym, sym->should_add_dynsym_entry(symtab)))
+ target->rel_dyn_section(layout)->add_global(sym, elfcpp::R_MIPS_REL32,
+ entry.output_section_, entry.relobj_, entry.shndx_, entry.address_);
+ else
+ target->rel_dyn_section(layout)->add_symbolless_global_addend(
+ sym, elfcpp::R_MIPS_REL32, entry.output_section_, entry.relobj_,
+ entry.shndx_, entry.address_);
+ }
+ else
+ this->make_copy_reloc(symtab, layout,
+ static_cast<Sized_symbol<size>*>(entry.sym_),
+ reloc_section);
+}
+
+// Target_mips methods.
+
+// Return the value to use for a dynamic symbol which requires special
+// treatment. This is how we support equality comparisons of function
+// pointers across shared library boundaries, as described in the
+// processor specific ABI supplement.
+
+template<int size, bool big_endian>
+uint64_t
+Target_mips<size, big_endian>::do_dynsym_value(const Symbol* gsym) const
+{
+ uint64_t value = 0;
+ const Mips_symbol<size>* mips_sym = Mips_symbol<size>::as_mips_sym(gsym);
+
+ if (!mips_sym->has_lazy_stub())
+ {
+ if (mips_sym->has_plt_offset())
+ {
+ // We distinguish between PLT entries and lazy-binding stubs by
+ // giving the former an st_other value of STO_MIPS_PLT. Set the
+ // value to the stub address if there are any relocations in the
+ // binary where pointer equality matters.
+ if (mips_sym->pointer_equality_needed())
+ {
+ // Prefer a standard MIPS PLT entry.
+ if (mips_sym->has_mips_plt_offset())
+ value = this->plt_section()->mips_entry_address(mips_sym);
+ else
+ value = this->plt_section()->comp_entry_address(mips_sym) + 1;
+ }
+ else
+ value = 0;
+ }
+ }
+ else
+ {
+ // First, set stub offsets for symbols. This method expects that the
+ // number of entries in dynamic symbol table is set.
+ this->mips_stubs_section()->set_lazy_stub_offsets();
+
+ // The run-time linker uses the st_value field of the symbol
+ // to reset the global offset table entry for this external
+ // to its stub address when unlinking a shared object.
+ value = this->mips_stubs_section()->stub_address(mips_sym);
+ }
+
+ if (mips_sym->has_mips16_fn_stub())
+ {
+ // If we have a MIPS16 function with a stub, the dynamic symbol must
+ // refer to the stub, since only the stub uses the standard calling
+ // conventions.
+ value = mips_sym->template
+ get_mips16_fn_stub<big_endian>()->output_address();
+ }
+
+ return value;
+}
+
+// Get the dynamic reloc section, creating it if necessary. It's always
+// .rel.dyn, even for MIPS64.
+
+template<int size, bool big_endian>
+typename Target_mips<size, big_endian>::Reloc_section*
+Target_mips<size, big_endian>::rel_dyn_section(Layout* layout)
+{
+ if (this->rel_dyn_ == NULL)
+ {
+ gold_assert(layout != NULL);
+ this->rel_dyn_ = new Reloc_section(parameters->options().combreloc());
+ layout->add_output_section_data(".rel.dyn", elfcpp::SHT_REL,
+ elfcpp::SHF_ALLOC, this->rel_dyn_,
+ ORDER_DYNAMIC_RELOCS, false);
+
+ // First entry in .rel.dyn has to be null.
+ // This is hack - we define dummy output data and set its address to 0,
+ // and define absolute R_MIPS_NONE relocation with offset 0 against it.
+ // This ensures that the entry is null.
+ Output_data* od = new Output_data_zero_fill(0, 0);
+ od->set_address(0);
+ this->rel_dyn_->add_absolute(elfcpp::R_MIPS_NONE, od, 0);
+ }
+ return this->rel_dyn_;
+}
+
+// Get the GOT section, creating it if necessary.
+
+template<int size, bool big_endian>
+Mips_output_data_got<size, big_endian>*
+Target_mips<size, big_endian>::got_section(Symbol_table* symtab,
+ Layout* layout)
+{
+ if (this->got_ == NULL)
+ {
+ gold_assert(symtab != NULL && layout != NULL);
+
+ this->got_ = new Mips_output_data_got<size, big_endian>(this, symtab,
+ layout);
+ layout->add_output_section_data(".got", elfcpp::SHT_PROGBITS,
+ (elfcpp::SHF_ALLOC | elfcpp::SHF_WRITE |
+ elfcpp::SHF_MIPS_GPREL),
+ this->got_, ORDER_DATA, false);
+
+ // Define _GLOBAL_OFFSET_TABLE_ at the start of the .got section.
+ symtab->define_in_output_data("_GLOBAL_OFFSET_TABLE_", NULL,
+ Symbol_table::PREDEFINED,
+ this->got_,
+ 0, 0, elfcpp::STT_OBJECT,
+ elfcpp::STB_GLOBAL,
+ elfcpp::STV_DEFAULT, 0,
+ false, false);
+ }
+
+ return this->got_;
+}
+
+// Calculate value of _gp symbol.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::set_gp(Layout* layout, Symbol_table* symtab)
+{
+ if (this->gp_ != NULL)
+ return;
+
+ Output_data* section = layout->find_output_section(".got");
+ if (section == NULL)
+ {
+ // If there is no .got section, gp should be based on .sdata.
+ // TODO(sasa): This is probably not needed. This was needed for older
+ // MIPS architectures which accessed both GOT and .sdata section using
+ // gp-relative addressing. Modern Mips Linux ELF architectures don't
+ // access .sdata using gp-relative addressing.
+ for (Layout::Section_list::const_iterator
+ p = layout->section_list().begin();
+ p != layout->section_list().end();
+ ++p)
+ {
+ if (strcmp((*p)->name(), ".sdata") == 0)
+ {
+ section = *p;
+ break;
+ }
+ }
+ }
+
+ Sized_symbol<size>* gp =
+ static_cast<Sized_symbol<size>*>(symtab->lookup("_gp"));
+ if (gp != NULL)
+ {
+ if (gp->source() != Symbol::IS_CONSTANT && section != NULL)
+ gp->init_output_data(gp->name(), NULL, section, MIPS_GP_OFFSET, 0,
+ elfcpp::STT_OBJECT,
+ elfcpp::STB_GLOBAL,
+ elfcpp::STV_DEFAULT, 0,
+ false, false);
+ this->gp_ = gp;
+ }
+ else if (section != NULL)
+ {
+ gp = static_cast<Sized_symbol<size>*>(symtab->define_in_output_data(
+ "_gp", NULL, Symbol_table::PREDEFINED,
+ section, MIPS_GP_OFFSET, 0,
+ elfcpp::STT_OBJECT,
+ elfcpp::STB_GLOBAL,
+ elfcpp::STV_DEFAULT,
+ 0, false, false));
+ this->gp_ = gp;
+ }
+}
+
+// Set the dynamic symbol indexes. INDEX is the index of the first
+// global dynamic symbol. Pointers to the symbols are stored into the
+// vector SYMS. The names are added to DYNPOOL. This returns an
+// updated dynamic symbol index.
+
+template<int size, bool big_endian>
+unsigned int
+Target_mips<size, big_endian>::do_set_dynsym_indexes(
+ std::vector<Symbol*>* dyn_symbols, unsigned int index,
+ std::vector<Symbol*>* syms, Stringpool* dynpool,
+ Versions* versions, Symbol_table* symtab) const
+{
+ std::vector<Symbol*> non_got_symbols;
+ std::vector<Symbol*> got_symbols;
+
+ reorder_dyn_symbols<size, big_endian>(dyn_symbols, &non_got_symbols,
+ &got_symbols);
+
+ for (std::vector<Symbol*>::iterator p = non_got_symbols.begin();
+ p != non_got_symbols.end();
+ ++p)
+ {
+ Symbol* sym = *p;
+
+ // Note that SYM may already have a dynamic symbol index, since
+ // some symbols appear more than once in the symbol table, with
+ // and without a version.
+
+ if (!sym->has_dynsym_index())
+ {
+ sym->set_dynsym_index(index);
+ ++index;
+ syms->push_back(sym);
+ dynpool->add(sym->name(), false, NULL);
+
+ // Record any version information.
+ if (sym->version() != NULL)
+ versions->record_version(symtab, dynpool, sym);
+
+ // If the symbol is defined in a dynamic object and is
+ // referenced in a regular object, then mark the dynamic
+ // object as needed. This is used to implement --as-needed.
+ if (sym->is_from_dynobj() && sym->in_reg())
+ sym->object()->set_is_needed();
+ }
+ }
+
+ for (std::vector<Symbol*>::iterator p = got_symbols.begin();
+ p != got_symbols.end();
+ ++p)
+ {
+ Symbol* sym = *p;
+ if (!sym->has_dynsym_index())
+ {
+ // Record any version information.
+ if (sym->version() != NULL)
+ versions->record_version(symtab, dynpool, sym);
+ }
+ }
+
+ index = versions->finalize(symtab, index, syms);
+
+ int got_sym_count = 0;
+ for (std::vector<Symbol*>::iterator p = got_symbols.begin();
+ p != got_symbols.end();
+ ++p)
+ {
+ Symbol* sym = *p;
+
+ if (!sym->has_dynsym_index())
+ {
+ ++got_sym_count;
+ sym->set_dynsym_index(index);
+ ++index;
+ syms->push_back(sym);
+ dynpool->add(sym->name(), false, NULL);
+
+ // If the symbol is defined in a dynamic object and is
+ // referenced in a regular object, then mark the dynamic
+ // object as needed. This is used to implement --as-needed.
+ if (sym->is_from_dynobj() && sym->in_reg())
+ sym->object()->set_is_needed();
+ }
+ }
+
+ // Set index of the first symbol that has .got entry.
+ this->got_->set_first_global_got_dynsym_index(
+ got_sym_count > 0 ? index - got_sym_count : -1U);
+
+ if (this->mips_stubs_ != NULL)
+ this->mips_stubs_->set_dynsym_count(index);
+
+ return index;
+}
+
+// Create a PLT entry for a global symbol referenced by r_type relocation.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::make_plt_entry(Symbol_table* symtab,
+ Layout* layout,
+ Mips_symbol<size>* gsym,
+ unsigned int r_type)
+{
+ if (gsym->has_lazy_stub() || gsym->has_plt_offset())
+ return;
+
+ if (this->plt_ == NULL)
+ {
+ // Create the GOT section first.
+ this->got_section(symtab, layout);
+
+ this->got_plt_ = new Output_data_space(4, "** GOT PLT");
+ layout->add_output_section_data(".got.plt", elfcpp::SHT_PROGBITS,
+ (elfcpp::SHF_ALLOC | elfcpp::SHF_WRITE),
+ this->got_plt_, ORDER_DATA, false);
+
+ // The first two entries are reserved.
+ this->got_plt_->set_current_data_size(2 * size/8);
+
+ this->plt_ = new Mips_output_data_plt<size, big_endian>(layout,
+ this->got_plt_,
+ this);
+ layout->add_output_section_data(".plt", elfcpp::SHT_PROGBITS,
+ (elfcpp::SHF_ALLOC
+ | elfcpp::SHF_EXECINSTR),
+ this->plt_, ORDER_PLT, false);
+ }
+
+ this->plt_->add_entry(gsym, r_type);
+}
+
+
+// Get the .MIPS.stubs section, creating it if necessary.
+
+template<int size, bool big_endian>
+Mips_output_data_mips_stubs<size, big_endian>*
+Target_mips<size, big_endian>::mips_stubs_section(Layout* layout)
+{
+ if (this->mips_stubs_ == NULL)
+ {
+ this->mips_stubs_ =
+ new Mips_output_data_mips_stubs<size, big_endian>(this);
+ layout->add_output_section_data(".MIPS.stubs", elfcpp::SHT_PROGBITS,
+ (elfcpp::SHF_ALLOC
+ | elfcpp::SHF_EXECINSTR),
+ this->mips_stubs_, ORDER_PLT, false);
+ }
+ return this->mips_stubs_;
+}
+
+// Get the LA25 stub section, creating it if necessary.
+
+template<int size, bool big_endian>
+Mips_output_data_la25_stub<size, big_endian>*
+Target_mips<size, big_endian>::la25_stub_section(Layout* layout)
+{
+ if (this->la25_stub_ == NULL)
+ {
+ this->la25_stub_ = new Mips_output_data_la25_stub<size, big_endian>();
+ layout->add_output_section_data(".text", elfcpp::SHT_PROGBITS,
+ (elfcpp::SHF_ALLOC
+ | elfcpp::SHF_EXECINSTR),
+ this->la25_stub_, ORDER_TEXT, false);
+ }
+ return this->la25_stub_;
+}
+
+// Process the relocations to determine unreferenced sections for
+// garbage collection.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::gc_process_relocs(
+ Symbol_table* symtab,
+ Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ unsigned int,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ size_t local_symbol_count,
+ const unsigned char* plocal_symbols)
+{
+ typedef Target_mips<size, big_endian> Mips;
+ typedef typename Target_mips<size, big_endian>::Scan Scan;
+
+ gold::gc_process_relocs<size, big_endian, Mips, elfcpp::SHT_REL, Scan,
+ typename Target_mips::Relocatable_size_for_reloc>(
+ symtab,
+ layout,
+ this,
+ object,
+ data_shndx,
+ prelocs,
+ reloc_count,
+ output_section,
+ needs_special_offset_handling,
+ local_symbol_count,
+ plocal_symbols);
+}
+
+// Scan relocations for a section.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::scan_relocs(
+ Symbol_table* symtab,
+ Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ size_t local_symbol_count,
+ const unsigned char* plocal_symbols)
+{
+ typedef Target_mips<size, big_endian> Mips;
+ typedef typename Target_mips<size, big_endian>::Scan Scan;
+
+ if (sh_type == elfcpp::SHT_REL)
+ gold::scan_relocs<size, big_endian, Mips, elfcpp::SHT_REL, Scan>(
+ symtab,
+ layout,
+ this,
+ object,
+ data_shndx,
+ prelocs,
+ reloc_count,
+ output_section,
+ needs_special_offset_handling,
+ local_symbol_count,
+ plocal_symbols);
+ else if (sh_type == elfcpp::SHT_RELA)
+ gold::scan_relocs<size, big_endian, Mips, elfcpp::SHT_RELA, Scan>(
+ symtab,
+ layout,
+ this,
+ object,
+ data_shndx,
+ prelocs,
+ reloc_count,
+ output_section,
+ needs_special_offset_handling,
+ local_symbol_count,
+ plocal_symbols);
+}
+
+template<int size, bool big_endian>
+bool
+Target_mips<size, big_endian>::mips_32bit_flags(elfcpp::Elf_Word flags)
+{
+ return ((flags & elfcpp::EF_MIPS_32BITMODE) != 0
+ || (flags & elfcpp::EF_MIPS_ABI) == elfcpp::E_MIPS_ABI_O32
+ || (flags & elfcpp::EF_MIPS_ABI) == elfcpp::E_MIPS_ABI_EABI32
+ || (flags & elfcpp::EF_MIPS_ARCH) == elfcpp::E_MIPS_ARCH_1
+ || (flags & elfcpp::EF_MIPS_ARCH) == elfcpp::E_MIPS_ARCH_2
+ || (flags & elfcpp::EF_MIPS_ARCH) == elfcpp::E_MIPS_ARCH_32
+ || (flags & elfcpp::EF_MIPS_ARCH) == elfcpp::E_MIPS_ARCH_32R2);
+}
+
+// Return the MACH for a MIPS e_flags value.
+template<int size, bool big_endian>
+unsigned int
+Target_mips<size, big_endian>::elf_mips_mach(elfcpp::Elf_Word flags)
+{
+ switch (flags & elfcpp::EF_MIPS_MACH)
+ {
+ case elfcpp::E_MIPS_MACH_3900:
+ return mach_mips3900;
+
+ case elfcpp::E_MIPS_MACH_4010:
+ return mach_mips4010;
+
+ case elfcpp::E_MIPS_MACH_4100:
+ return mach_mips4100;
+
+ case elfcpp::E_MIPS_MACH_4111:
+ return mach_mips4111;
+
+ case elfcpp::E_MIPS_MACH_4120:
+ return mach_mips4120;
+
+ case elfcpp::E_MIPS_MACH_4650:
+ return mach_mips4650;
+
+ case elfcpp::E_MIPS_MACH_5400:
+ return mach_mips5400;
+
+ case elfcpp::E_MIPS_MACH_5500:
+ return mach_mips5500;
+
+ case elfcpp::E_MIPS_MACH_9000:
+ return mach_mips9000;
+
+ case elfcpp::E_MIPS_MACH_SB1:
+ return mach_mips_sb1;
+
+ case elfcpp::E_MIPS_MACH_LS2E:
+ return mach_mips_loongson_2e;
+
+ case elfcpp::E_MIPS_MACH_LS2F:
+ return mach_mips_loongson_2f;
+
+ case elfcpp::E_MIPS_MACH_LS3A:
+ return mach_mips_loongson_3a;
+
+ case elfcpp::E_MIPS_MACH_OCTEON2:
+ return mach_mips_octeon2;
+
+ case elfcpp::E_MIPS_MACH_OCTEON:
+ return mach_mips_octeon;
+
+ case elfcpp::E_MIPS_MACH_XLR:
+ return mach_mips_xlr;
+
+ default:
+ switch (flags & elfcpp::EF_MIPS_ARCH)
+ {
+ default:
+ case elfcpp::E_MIPS_ARCH_1:
+ return mach_mips3000;
+
+ case elfcpp::E_MIPS_ARCH_2:
+ return mach_mips6000;
+
+ case elfcpp::E_MIPS_ARCH_3:
+ return mach_mips4000;
+
+ case elfcpp::E_MIPS_ARCH_4:
+ return mach_mips8000;
+
+ case elfcpp::E_MIPS_ARCH_5:
+ return mach_mips5;
+
+ case elfcpp::E_MIPS_ARCH_32:
+ return mach_mipsisa32;
+
+ case elfcpp::E_MIPS_ARCH_64:
+ return mach_mipsisa64;
+
+ case elfcpp::E_MIPS_ARCH_32R2:
+ return mach_mipsisa32r2;
+
+ case elfcpp::E_MIPS_ARCH_64R2:
+ return mach_mipsisa64r2;
+ }
+ }
+
+ return 0;
+}
+
+// Check whether machine EXTENSION is an extension of machine BASE.
+template<int size, bool big_endian>
+bool
+Target_mips<size, big_endian>::mips_mach_extends(unsigned int base,
+ unsigned int extension)
+{
+ if (extension == base)
+ return true;
+
+ if ((base == mach_mipsisa32)
+ && this->mips_mach_extends(mach_mipsisa64, extension))
+ return true;
+
+ if ((base == mach_mipsisa32r2)
+ && this->mips_mach_extends(mach_mipsisa64r2, extension))
+ return true;
+
+ for (unsigned int i = 0; i < this->mips_mach_extensions_.size(); ++i)
+ if (extension == this->mips_mach_extensions_[i].first)
+ {
+ extension = this->mips_mach_extensions_[i].second;
+ if (extension == base)
+ return true;
+ }
+
+ return false;
+}
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::merge_processor_specific_flags(
+ const std::string& name, elfcpp::Elf_Word in_flags,
+ unsigned char in_ei_class, bool dyn_obj)
+{
+ // If flags are not set yet, just copy them.
+ if (!this->are_processor_specific_flags_set())
+ {
+ this->set_processor_specific_flags(in_flags);
+ this->ei_class_ = in_ei_class;
+ this->mach_ = this->elf_mips_mach(in_flags);
+ return;
+ }
+
+ elfcpp::Elf_Word new_flags = in_flags;
+ elfcpp::Elf_Word old_flags = this->processor_specific_flags();
+ elfcpp::Elf_Word merged_flags = this->processor_specific_flags();
+ merged_flags |= new_flags & elfcpp::EF_MIPS_NOREORDER;
+
+ // Check flag compatibility.
+ new_flags &= ~elfcpp::EF_MIPS_NOREORDER;
+ old_flags &= ~elfcpp::EF_MIPS_NOREORDER;
+
+ // Some IRIX 6 BSD-compatibility objects have this bit set. It
+ // doesn't seem to matter.
+ new_flags &= ~elfcpp::EF_MIPS_XGOT;
+ old_flags &= ~elfcpp::EF_MIPS_XGOT;
+
+ // MIPSpro generates ucode info in n64 objects. Again, we should
+ // just be able to ignore this.
+ new_flags &= ~elfcpp::EF_MIPS_UCODE;
+ old_flags &= ~elfcpp::EF_MIPS_UCODE;
+
+ // DSOs should only be linked with CPIC code.
+ if (dyn_obj)
+ new_flags |= elfcpp::EF_MIPS_PIC | elfcpp::EF_MIPS_CPIC;
+
+ if (new_flags == old_flags)
+ {
+ this->set_processor_specific_flags(merged_flags);
+ return;
+ }
+
+ if (((new_flags & (elfcpp::EF_MIPS_PIC | elfcpp::EF_MIPS_CPIC)) != 0)
+ != ((old_flags & (elfcpp::EF_MIPS_PIC | elfcpp::EF_MIPS_CPIC)) != 0))
+ gold_warning(_("%s: linking abicalls files with non-abicalls files"),
+ name.c_str());
+
+ if (new_flags & (elfcpp::EF_MIPS_PIC | elfcpp::EF_MIPS_CPIC))
+ merged_flags |= elfcpp::EF_MIPS_CPIC;
+ if (!(new_flags & elfcpp::EF_MIPS_PIC))
+ merged_flags &= ~elfcpp::EF_MIPS_PIC;
+
+ new_flags &= ~(elfcpp::EF_MIPS_PIC | elfcpp::EF_MIPS_CPIC);
+ old_flags &= ~(elfcpp::EF_MIPS_PIC | elfcpp::EF_MIPS_CPIC);
+
+ // Compare the ISAs.
+ if (mips_32bit_flags(old_flags) != mips_32bit_flags(new_flags))
+ gold_error(_("%s: linking 32-bit code with 64-bit code"), name.c_str());
+ else if (!this->mips_mach_extends(this->elf_mips_mach(in_flags), this->mach_))
+ {
+ // Output ISA isn't the same as, or an extension of, input ISA.
+ if (this->mips_mach_extends(this->mach_, this->elf_mips_mach(in_flags)))
+ {
+ // Copy the architecture info from input object to output. Also copy
+ // the 32-bit flag (if set) so that we continue to recognise
+ // output as a 32-bit binary.
+ this->mach_ = this->elf_mips_mach(in_flags);
+ merged_flags &= ~(elfcpp::EF_MIPS_ARCH | elfcpp::EF_MIPS_MACH);
+ merged_flags |= (new_flags & (elfcpp::EF_MIPS_ARCH
+ | elfcpp::EF_MIPS_MACH | elfcpp::EF_MIPS_32BITMODE));
+
+ // Copy across the ABI flags if output doesn't use them
+ // and if that was what caused us to treat input object as 32-bit.
+ if ((old_flags & elfcpp::EF_MIPS_ABI) == 0
+ && this->mips_32bit_flags(new_flags)
+ && !this->mips_32bit_flags(new_flags & ~elfcpp::EF_MIPS_ABI))
+ merged_flags |= new_flags & elfcpp::EF_MIPS_ABI;
+ }
+ else
+ // The ISAs aren't compatible.
+ gold_error(_("%s: linking %s module with previous %s modules"),
+ name.c_str(), this->elf_mips_mach_name(in_flags),
+ this->elf_mips_mach_name(merged_flags));
+ }
+
+ new_flags &= (~(elfcpp::EF_MIPS_ARCH | elfcpp::EF_MIPS_MACH
+ | elfcpp::EF_MIPS_32BITMODE));
+ old_flags &= (~(elfcpp::EF_MIPS_ARCH | elfcpp::EF_MIPS_MACH
+ | elfcpp::EF_MIPS_32BITMODE));
+
+ // Compare ABIs. The 64-bit ABI does not use EF_MIPS_ABI. But, it does set
+ // EI_CLASS differently from any 32-bit ABI.
+ if ((new_flags & elfcpp::EF_MIPS_ABI) != (old_flags & elfcpp::EF_MIPS_ABI)
+ || (in_ei_class != this->ei_class_))
+ {
+ // Only error if both are set (to different values).
+ if (((new_flags & elfcpp::EF_MIPS_ABI)
+ && (old_flags & elfcpp::EF_MIPS_ABI))
+ || (in_ei_class != this->ei_class_))
+ gold_error(_("%s: ABI mismatch: linking %s module with "
+ "previous %s modules"), name.c_str(),
+ this->elf_mips_abi_name(in_flags, in_ei_class),
+ this->elf_mips_abi_name(merged_flags, this->ei_class_));
+
+ new_flags &= ~elfcpp::EF_MIPS_ABI;
+ old_flags &= ~elfcpp::EF_MIPS_ABI;
+ }
+
+ // Compare ASEs. Forbid linking MIPS16 and microMIPS ASE modules together
+ // and allow arbitrary mixing of the remaining ASEs (retain the union).
+ if ((new_flags & elfcpp::EF_MIPS_ARCH_ASE)
+ != (old_flags & elfcpp::EF_MIPS_ARCH_ASE))
+ {
+ int old_micro = old_flags & elfcpp::EF_MIPS_ARCH_ASE_MICROMIPS;
+ int new_micro = new_flags & elfcpp::EF_MIPS_ARCH_ASE_MICROMIPS;
+ int old_m16 = old_flags & elfcpp::EF_MIPS_ARCH_ASE_M16;
+ int new_m16 = new_flags & elfcpp::EF_MIPS_ARCH_ASE_M16;
+ int micro_mis = old_m16 && new_micro;
+ int m16_mis = old_micro && new_m16;
+
+ if (m16_mis || micro_mis)
+ gold_error(_("%s: ASE mismatch: linking %s module with "
+ "previous %s modules"), name.c_str(),
+ m16_mis ? "MIPS16" : "microMIPS",
+ m16_mis ? "microMIPS" : "MIPS16");
+
+ merged_flags |= new_flags & elfcpp::EF_MIPS_ARCH_ASE;
+
+ new_flags &= ~ elfcpp::EF_MIPS_ARCH_ASE;
+ old_flags &= ~ elfcpp::EF_MIPS_ARCH_ASE;
+ }
+
+ // Warn about any other mismatches.
+ if (new_flags != old_flags)
+ gold_error(_("%s: uses different e_flags (0x%x) fields than previous "
+ "modules (0x%x)"), name.c_str(), new_flags, old_flags);
+
+ this->set_processor_specific_flags(merged_flags);
+}
+
+// Adjust ELF file header.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::do_adjust_elf_header(
+ unsigned char* view,
+ int len)
+{
+ gold_assert(len == elfcpp::Elf_sizes<size>::ehdr_size);
+
+ elfcpp::Ehdr<size, big_endian> ehdr(view);
+ unsigned char e_ident[elfcpp::EI_NIDENT];
+ memcpy(e_ident, ehdr.get_e_ident(), elfcpp::EI_NIDENT);
+
+ e_ident[elfcpp::EI_CLASS] = this->ei_class_;
+
+ elfcpp::Ehdr_write<size, big_endian> oehdr(view);
+ oehdr.put_e_ident(e_ident);
+ if (this->entry_symbol_is_compressed_)
+ oehdr.put_e_entry(ehdr.get_e_entry() + 1);
+}
+
+// do_make_elf_object to override the same function in the base class.
+// We need to use a target-specific sub-class of
+// Sized_relobj_file<size, big_endian> to store Mips specific information.
+// Hence we need to have our own ELF object creation.
+
+template<int size, bool big_endian>
+Object*
+Target_mips<size, big_endian>::do_make_elf_object(
+ const std::string& name,
+ Input_file* input_file,
+ off_t offset, const elfcpp::Ehdr<size, big_endian>& ehdr)
+{
+ int et = ehdr.get_e_type();
+ // ET_EXEC files are valid input for --just-symbols/-R,
+ // and we treat them as relocatable objects.
+ if (et == elfcpp::ET_REL
+ || (et == elfcpp::ET_EXEC && input_file->just_symbols()))
+ {
+ Mips_relobj<size, big_endian>* obj =
+ new Mips_relobj<size, big_endian>(name, input_file, offset, ehdr);
+ obj->setup();
+ return obj;
+ }
+ else if (et == elfcpp::ET_DYN)
+ {
+ // TODO(sasa): Should we create Mips_dynobj?
+ return Target::do_make_elf_object(name, input_file, offset, ehdr);
+ }
+ else
+ {
+ gold_error(_("%s: unsupported ELF file type %d"),
+ name.c_str(), et);
+ return NULL;
+ }
+}
+
+// Finalize the sections.
+
+template <int size, bool big_endian>
+void
+Target_mips<size, big_endian>::do_finalize_sections(Layout* layout,
+ const Input_objects* input_objects,
+ Symbol_table* symtab)
+{
+ // Add +1 to MIPS16 and microMIPS init_ and _fini symbols so that DT_INIT and
+ // DT_FINI have correct values.
+ Mips_symbol<size>* init = static_cast<Mips_symbol<size>*>(
+ symtab->lookup(parameters->options().init()));
+ if (init != NULL && (init->is_mips16() || init->is_micromips()))
+ init->set_value(init->value() | 1);
+ Mips_symbol<size>* fini = static_cast<Mips_symbol<size>*>(
+ symtab->lookup(parameters->options().fini()));
+ if (fini != NULL && (fini->is_mips16() || fini->is_micromips()))
+ fini->set_value(fini->value() | 1);
+
+ // Check whether the entry symbol is mips16 or micromips. This is needed to
+ // adjust entry address in ELF header.
+ Mips_symbol<size>* entry =
+ static_cast<Mips_symbol<size>*>(symtab->lookup(this->entry_symbol_name()));
+ this->entry_symbol_is_compressed_ = (entry != NULL && (entry->is_mips16()
+ || entry->is_micromips()));
+
+ if (!parameters->doing_static_link()
+ && (strcmp(parameters->options().hash_style(), "gnu") == 0
+ || strcmp(parameters->options().hash_style(), "both") == 0))
+ {
+ // .gnu.hash and the MIPS ABI require .dynsym to be sorted in different
+ // ways. .gnu.hash needs symbols to be grouped by hash code whereas the
+ // MIPS ABI requires a mapping between the GOT and the symbol table.
+ gold_error(".gnu.hash is incompatible with the MIPS ABI");
+ }
+
+ // Check whether the final section that was scaned has HI16 or GOT16
+ // relocations without the corresponding LO16 part.
+ if (this->got16_addends_.size() > 0)
+ gold_error("Can't find matching LO16 reloc");
+
+ // Set _gp value.
+ this->set_gp(layout, symtab);
+
+ // Check for any mips16 stub sections that we can discard.
+ if (!parameters->options().relocatable())
+ {
+ for (Input_objects::Relobj_iterator p = input_objects->relobj_begin();
+ p != input_objects->relobj_end();
+ ++p)
+ {
+ Mips_relobj<size, big_endian>* object =
+ Mips_relobj<size, big_endian>::as_mips_relobj(*p);
+ object->discard_mips16_stub_sections(symtab);
+ }
+ }
+
+ // Merge processor-specific flags.
+ for (Input_objects::Relobj_iterator p = input_objects->relobj_begin();
+ p != input_objects->relobj_end();
+ ++p)
+ {
+ Mips_relobj<size, big_endian>* relobj =
+ Mips_relobj<size, big_endian>::as_mips_relobj(*p);
+
+ Input_file::Format format = relobj->input_file()->format();
+ if (format == Input_file::FORMAT_ELF)
+ {
+ // Read processor-specific flags in ELF file header.
+ const unsigned char* pehdr = relobj->get_view(
+ elfcpp::file_header_offset,
+ elfcpp::Elf_sizes<size>::ehdr_size,
+ true, false);
+
+ elfcpp::Ehdr<size, big_endian> ehdr(pehdr);
+ elfcpp::Elf_Word in_flags = ehdr.get_e_flags();
+ unsigned char ei_class = ehdr.get_e_ident()[elfcpp::EI_CLASS];
+
+ this->merge_processor_specific_flags(relobj->name(), in_flags,
+ ei_class, false);
+ }
+ }
+
+ for (Input_objects::Dynobj_iterator p = input_objects->dynobj_begin();
+ p != input_objects->dynobj_end();
+ ++p)
+ {
+ Sized_dynobj<size, big_endian>* dynobj =
+ static_cast<Sized_dynobj<size, big_endian>*>(*p);
+
+ // Read processor-specific flags.
+ const unsigned char* pehdr = dynobj->get_view(elfcpp::file_header_offset,
+ elfcpp::Elf_sizes<size>::ehdr_size,
+ true, false);
+
+ elfcpp::Ehdr<size, big_endian> ehdr(pehdr);
+ elfcpp::Elf_Word in_flags = ehdr.get_e_flags();
+ unsigned char ei_class = ehdr.get_e_ident()[elfcpp::EI_CLASS];
+
+ this->merge_processor_specific_flags(dynobj->name(), in_flags, ei_class,
+ true);
+ }
+
+ if (this->plt_ != NULL)
+ {
+ // Set final PLT offsets for symbols.
+ this->plt_section()->set_plt_offsets();
+
+ // Define _PROCEDURE_LINKAGE_TABLE_ at the start of the .plt section.
+ // Set STO_MICROMIPS flag if the output has microMIPS code, but only if
+ // there are no standard PLT entries present.
+ unsigned char nonvis = 0;
+ if (this->is_output_micromips()
+ && !this->plt_section()->has_standard_entries())
+ nonvis = elfcpp::STO_MICROMIPS >> 2;
+ symtab->define_in_output_data("_PROCEDURE_LINKAGE_TABLE_", NULL,
+ Symbol_table::PREDEFINED,
+ this->plt_,
+ 0, 0, elfcpp::STT_FUNC,
+ elfcpp::STB_LOCAL,
+ elfcpp::STV_DEFAULT, nonvis,
+ false, false);
+ }
+
+ if (this->mips_stubs_ != NULL)
+ {
+ // Define _MIPS_STUBS_ at the start of the .MIPS.stubs section.
+ unsigned char nonvis = 0;
+ if (this->is_output_micromips())
+ nonvis = elfcpp::STO_MICROMIPS >> 2;
+ symtab->define_in_output_data("_MIPS_STUBS_", NULL,
+ Symbol_table::PREDEFINED,
+ this->mips_stubs_,
+ 0, 0, elfcpp::STT_FUNC,
+ elfcpp::STB_LOCAL,
+ elfcpp::STV_DEFAULT, nonvis,
+ false, false);
+ }
+
+ if (!parameters->options().relocatable() && !parameters->doing_static_link())
+ // In case there is no .got section, create one.
+ this->got_section(symtab, layout);
+
+ // Emit any relocs we saved in an attempt to avoid generating COPY
+ // relocs.
+ if (this->copy_relocs_.any_saved_relocs())
+ this->copy_relocs_.emit_mips(this->rel_dyn_section(layout), symtab, layout,
+ this);
+
+ // Emit dynamic relocs.
+ for (typename std::vector<Dyn_reloc>::iterator p = this->dyn_relocs_.begin();
+ p != this->dyn_relocs_.end();
+ ++p)
+ p->emit(this->rel_dyn_section(layout), this->got_section(), symtab);
+
+ if (this->has_got_section())
+ this->got_section()->lay_out_got(layout, symtab, input_objects);
+
+ if (this->mips_stubs_ != NULL)
+ this->mips_stubs_->set_needs_dynsym_value();
+
+ // Check for functions that might need $25 to be valid on entry.
+ // TODO(sasa): Can we do this without iterating over all symbols?
+ typedef Symbol_visitor_check_symbols<size, big_endian> Symbol_visitor;
+ symtab->for_all_symbols<size, Symbol_visitor>(Symbol_visitor(this, layout,
+ symtab));
+
+ // Add NULL segment.
+ if (!parameters->options().relocatable())
+ layout->make_output_segment(elfcpp::PT_NULL, 0);
+
+ for (Layout::Section_list::const_iterator p = layout->section_list().begin();
+ p != layout->section_list().end();
+ ++p)
+ {
+ if ((*p)->type() == elfcpp::SHT_MIPS_REGINFO)
+ {
+ Mips_output_section_reginfo<size, big_endian>* reginfo =
+ Mips_output_section_reginfo<size, big_endian>::
+ as_mips_output_section_reginfo(*p);
+
+ if (!parameters->options().relocatable())
+ {
+ Output_segment* reginfo_segment =
+ layout->make_output_segment(elfcpp::PT_MIPS_REGINFO,
+ elfcpp::PF_R);
+ reginfo_segment->add_output_section_to_nonload(reginfo,
+ elfcpp::PF_R);
+ }
+ }
+ }
+
+ // Fill in some more dynamic tags.
+ // TODO(sasa): Add more dynamic tags.
+ const Reloc_section* rel_plt = (this->plt_ == NULL
+ ? NULL : this->plt_->rel_plt());
+ layout->add_target_dynamic_tags(true, this->got_, rel_plt,
+ this->rel_dyn_, true, false);
+
+ Output_data_dynamic* const odyn = layout->dynamic_data();
+ if (odyn != NULL
+ && !parameters->options().relocatable()
+ && !parameters->doing_static_link())
+ {
+ unsigned int d_val;
+ // This element holds a 32-bit version id for the Runtime
+ // Linker Interface. This will start at integer value 1.
+ d_val = 0x01;
+ odyn->add_constant(elfcpp::DT_MIPS_RLD_VERSION, d_val);
+
+ // Dynamic flags
+ d_val = elfcpp::RHF_NOTPOT;
+ odyn->add_constant(elfcpp::DT_MIPS_FLAGS, d_val);
+
+ // Save layout for using when emiting custom dynamic tags.
+ this->layout_ = layout;
+
+ // This member holds the base address of the segment.
+ odyn->add_custom(elfcpp::DT_MIPS_BASE_ADDRESS);
+
+ // This member holds the number of entries in the .dynsym section.
+ odyn->add_custom(elfcpp::DT_MIPS_SYMTABNO);
+
+ // This member holds the index of the first dynamic symbol
+ // table entry that corresponds to an entry in the global offset table.
+ odyn->add_custom(elfcpp::DT_MIPS_GOTSYM);
+
+ // This member holds the number of local GOT entries.
+ odyn->add_constant(elfcpp::DT_MIPS_LOCAL_GOTNO,
+ this->got_->get_local_gotno());
+
+ if (this->plt_ != NULL)
+ // DT_MIPS_PLTGOT dynamic tag
+ odyn->add_section_address(elfcpp::DT_MIPS_PLTGOT, this->got_plt_);
+ }
+ }
+
+// Get the custom dynamic tag value.
+template<int size, bool big_endian>
+unsigned int
+Target_mips<size, big_endian>::do_dynamic_tag_custom_value(elfcpp::DT tag) const
+{
+ switch (tag)
+ {
+ case elfcpp::DT_MIPS_BASE_ADDRESS:
+ {
+ // The base address of the segment.
+ // At this point, the segment list has been sorted into final order,
+ // so just return vaddr of the first readable PT_LOAD segment.
+ Output_segment* seg =
+ this->layout_->find_output_segment(elfcpp::PT_LOAD, elfcpp::PF_R, 0);
+ gold_assert(seg != NULL);
+ return seg->vaddr();
+ }
+
+ case elfcpp::DT_MIPS_SYMTABNO:
+ // The number of entries in the .dynsym section.
+ return this->get_dt_mips_symtabno();
+
+ case elfcpp::DT_MIPS_GOTSYM:
+ {
+ // The index of the first dynamic symbol table entry that corresponds
+ // to an entry in the GOT.
+ if (this->got_->first_global_got_dynsym_index() != -1U)
+ return this->got_->first_global_got_dynsym_index();
+ else
+ // In case if we don't have global GOT symbols we default to setting
+ // DT_MIPS_GOTSYM to the same value as DT_MIPS_SYMTABNO.
+ return this->get_dt_mips_symtabno();
+ }
+
+ default:
+ gold_error(_("Unknown dynamic tag 0x%x"), (unsigned int)tag);
+ }
+
+ return (unsigned int)-1;
+}
+
+// Relocate section data.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::relocate_section(
+ const Relocate_info<size, big_endian>* relinfo,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ unsigned char* view,
+ Mips_address address,
+ section_size_type view_size,
+ const Reloc_symbol_changes* reloc_symbol_changes)
+{
+ typedef Target_mips<size, big_endian> Mips;
+ typedef typename Target_mips<size, big_endian>::Relocate Mips_relocate;
+
+ if (sh_type == elfcpp::SHT_REL)
+ gold::relocate_section<size, big_endian, Mips, elfcpp::SHT_REL,
+ Mips_relocate, gold::Default_comdat_behavior>(
+ relinfo,
+ this,
+ prelocs,
+ reloc_count,
+ output_section,
+ needs_special_offset_handling,
+ view,
+ address,
+ view_size,
+ reloc_symbol_changes);
+ else if (sh_type == elfcpp::SHT_RELA)
+ gold::relocate_section<size, big_endian, Mips, elfcpp::SHT_RELA,
+ Mips_relocate, gold::Default_comdat_behavior>(
+ relinfo,
+ this,
+ prelocs,
+ reloc_count,
+ output_section,
+ needs_special_offset_handling,
+ view,
+ address,
+ view_size,
+ reloc_symbol_changes);
+}
+
+// Return the size of a relocation while scanning during a relocatable
+// link.
+
+template<int size, bool big_endian>
+unsigned int
+Target_mips<size, big_endian>::Relocatable_size_for_reloc::get_size_for_reloc(
+ unsigned int r_type,
+ Relobj* object)
+{
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_NONE:
+ case elfcpp::R_MIPS_TLS_DTPMOD64:
+ case elfcpp::R_MIPS_TLS_DTPREL64:
+ case elfcpp::R_MIPS_TLS_TPREL64:
+ return 0;
+
+ case elfcpp::R_MIPS_32:
+ case elfcpp::R_MIPS_TLS_DTPMOD32:
+ case elfcpp::R_MIPS_TLS_DTPREL32:
+ case elfcpp::R_MIPS_TLS_TPREL32:
+ case elfcpp::R_MIPS_REL32:
+ case elfcpp::R_MIPS_PC32:
+ case elfcpp::R_MIPS_GPREL32:
+ case elfcpp::R_MIPS_JALR:
+ return 4;
+
+ case elfcpp::R_MIPS_16:
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MIPS_LO16:
+ case elfcpp::R_MIPS_GPREL16:
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MIPS16_LO16:
+ case elfcpp::R_MIPS_PC16:
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MIPS_TLS_DTPREL_HI16:
+ case elfcpp::R_MIPS_TLS_DTPREL_LO16:
+ case elfcpp::R_MIPS_TLS_TPREL_HI16:
+ case elfcpp::R_MIPS_TLS_TPREL_LO16:
+ case elfcpp::R_MIPS16_GPREL:
+ case elfcpp::R_MIPS_GOT_DISP:
+ case elfcpp::R_MIPS_LITERAL:
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MIPS_GOT_OFST:
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ return 2;
+
+ // These relocations are not byte sized
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MIPS16_26:
+ return 4;
+
+ case elfcpp::R_MIPS_COPY:
+ case elfcpp::R_MIPS_JUMP_SLOT:
+ object->error(_("unexpected reloc %u in object file"), r_type);
+ return 0;
+
+ default:
+ object->error(_("unsupported reloc %u in object file"), r_type);
+ return 0;
+ }
+}
+
+// Scan the relocs during a relocatable link.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::scan_relocatable_relocs(
+ Symbol_table* symtab,
+ Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ size_t local_symbol_count,
+ const unsigned char* plocal_symbols,
+ Relocatable_relocs* rr)
+{
+ gold_assert(sh_type == elfcpp::SHT_REL);
+
+ typedef Mips_scan_relocatable_relocs<big_endian, elfcpp::SHT_REL,
+ Relocatable_size_for_reloc> Scan_relocatable_relocs;
+
+ gold::scan_relocatable_relocs<size, big_endian, elfcpp::SHT_REL,
+ Scan_relocatable_relocs>(
+ symtab,
+ layout,
+ object,
+ data_shndx,
+ prelocs,
+ reloc_count,
+ output_section,
+ needs_special_offset_handling,
+ local_symbol_count,
+ plocal_symbols,
+ rr);
+}
+
+// Emit relocations for a section.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::relocate_relocs(
+ const Relocate_info<size, big_endian>* relinfo,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ typename elfcpp::Elf_types<size>::Elf_Off
+ offset_in_output_section,
+ const Relocatable_relocs* rr,
+ unsigned char* view,
+ Mips_address view_address,
+ section_size_type view_size,
+ unsigned char* reloc_view,
+ section_size_type reloc_view_size)
+{
+ gold_assert(sh_type == elfcpp::SHT_REL);
+
+ gold::relocate_relocs<size, big_endian, elfcpp::SHT_REL>(
+ relinfo,
+ prelocs,
+ reloc_count,
+ output_section,
+ offset_in_output_section,
+ rr,
+ view,
+ view_address,
+ view_size,
+ reloc_view,
+ reloc_view_size);
+}
+
+// Perform target-specific processing in a relocatable link. This is
+// only used if we use the relocation strategy RELOC_SPECIAL.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::relocate_special_relocatable(
+ const Relocate_info<size, big_endian>* relinfo,
+ unsigned int sh_type,
+ const unsigned char* preloc_in,
+ size_t relnum,
+ Output_section* output_section,
+ typename elfcpp::Elf_types<size>::Elf_Off offset_in_output_section,
+ unsigned char* view,
+ Mips_address view_address,
+ section_size_type,
+ unsigned char* preloc_out)
+{
+ // We can only handle REL type relocation sections.
+ gold_assert(sh_type == elfcpp::SHT_REL);
+
+ typedef typename Reloc_types<elfcpp::SHT_REL, size, big_endian>::Reloc
+ Reltype;
+ typedef typename Reloc_types<elfcpp::SHT_REL, size, big_endian>::Reloc_write
+ Reltype_write;
+
+ typedef Mips_relocate_functions<size, big_endian> Reloc_funcs;
+
+ const Mips_address invalid_address = static_cast<Mips_address>(0) - 1;
+
+ Mips_relobj<size, big_endian>* object =
+ Mips_relobj<size, big_endian>::as_mips_relobj(relinfo->object);
+ const unsigned int local_count = object->local_symbol_count();
+
+ Reltype reloc(preloc_in);
+ Reltype_write reloc_write(preloc_out);
+
+ elfcpp::Elf_types<32>::Elf_WXword r_info = reloc.get_r_info();
+ const unsigned int r_sym = elfcpp::elf_r_sym<size>(r_info);
+ const unsigned int r_type = elfcpp::elf_r_type<size>(r_info);
+
+ // Get the new symbol index.
+ // We only use RELOC_SPECIAL strategy in local relocations.
+ gold_assert(r_sym < local_count);
+
+ // We are adjusting a section symbol. We need to find
+ // the symbol table index of the section symbol for
+ // the output section corresponding to input section
+ // in which this symbol is defined.
+ bool is_ordinary;
+ unsigned int shndx = object->local_symbol_input_shndx(r_sym, &is_ordinary);
+ gold_assert(is_ordinary);
+ Output_section* os = object->output_section(shndx);
+ gold_assert(os != NULL);
+ gold_assert(os->needs_symtab_index());
+ unsigned int new_symndx = os->symtab_index();
+
+ // Get the new offset--the location in the output section where
+ // this relocation should be applied.
+
+ Mips_address offset = reloc.get_r_offset();
+ Mips_address new_offset;
+ if (offset_in_output_section != invalid_address)
+ new_offset = offset + offset_in_output_section;
+ else
+ {
+ section_offset_type sot_offset =
+ convert_types<section_offset_type, Mips_address>(offset);
+ section_offset_type new_sot_offset =
+ output_section->output_offset(object, relinfo->data_shndx,
+ sot_offset);
+ gold_assert(new_sot_offset != -1);
+ new_offset = new_sot_offset;
+ }
+
+ // In an object file, r_offset is an offset within the section.
+ // In an executable or dynamic object, generated by
+ // --emit-relocs, r_offset is an absolute address.
+ if (!parameters->options().relocatable())
+ {
+ new_offset += view_address;
+ if (offset_in_output_section != invalid_address)
+ new_offset -= offset_in_output_section;
+ }
+
+ reloc_write.put_r_offset(new_offset);
+ reloc_write.put_r_info(elfcpp::elf_r_info<32>(new_symndx, r_type));
+
+ // Handle the reloc addend.
+ // The relocation uses a section symbol in the input file.
+ // We are adjusting it to use a section symbol in the output
+ // file. The input section symbol refers to some address in
+ // the input section. We need the relocation in the output
+ // file to refer to that same address. This adjustment to
+ // the addend is the same calculation we use for a simple
+ // absolute relocation for the input section symbol.
+
+ const Symbol_value<size>* psymval = object->local_symbol(r_sym);
+
+ unsigned char* paddend = view + offset;
+ typename Reloc_funcs::Status reloc_status = Reloc_funcs::STATUS_OKAY;
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_26:
+ reloc_status = Reloc_funcs::rel26(paddend, object, psymval,
+ offset_in_output_section, true, 0, sh_type == elfcpp::SHT_REL, NULL,
+ false /*TODO(sasa): cross mode jump*/, r_type, this->jal_to_bal());
+ break;
+
+ default:
+ gold_unreachable();
+ }
+
+ // Report any errors.
+ switch (reloc_status)
+ {
+ case Reloc_funcs::STATUS_OKAY:
+ break;
+ case Reloc_funcs::STATUS_OVERFLOW:
+ gold_error_at_location(relinfo, relnum, reloc.get_r_offset(),
+ _("relocation overflow"));
+ break;
+ case Reloc_funcs::STATUS_BAD_RELOC:
+ gold_error_at_location(relinfo, relnum, reloc.get_r_offset(),
+ _("unexpected opcode while processing relocation"));
+ break;
+ default:
+ gold_unreachable();
+ }
+}
+
+// Optimize the TLS relocation type based on what we know about the
+// symbol. IS_FINAL is true if the final address of this symbol is
+// known at link time.
+
+template<int size, bool big_endian>
+tls::Tls_optimization
+Target_mips<size, big_endian>::optimize_tls_reloc(bool, int)
+{
+ // FIXME: Currently we do not do any TLS optimization.
+ return tls::TLSOPT_NONE;
+}
+
+// Scan a relocation for a local symbol.
+
+template<int size, bool big_endian>
+inline void
+Target_mips<size, big_endian>::Scan::local(
+ Symbol_table* symtab,
+ Layout* layout,
+ Target_mips<size, big_endian>* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>* rela,
+ const elfcpp::Rel<size, big_endian>* rel,
+ unsigned int rel_type,
+ unsigned int r_type,
+ const elfcpp::Sym<size, big_endian>& lsym,
+ bool is_discarded)
+{
+ if (is_discarded)
+ return;
+
+ Mips_address r_offset;
+ typename elfcpp::Elf_types<size>::Elf_WXword r_info;
+ typename elfcpp::Elf_types<size>::Elf_Swxword r_addend;
+
+ if (rel_type == elfcpp::SHT_RELA)
+ {
+ r_offset = rela->get_r_offset();
+ r_info = rela->get_r_info();
+ r_addend = rela->get_r_addend();
+ }
+ else
+ {
+ r_offset = rel->get_r_offset();
+ r_info = rel->get_r_info();
+ r_addend = 0;
+ }
+
+ unsigned int r_sym = elfcpp::elf_r_sym<size>(r_info);
+ Mips_relobj<size, big_endian>* mips_obj =
+ Mips_relobj<size, big_endian>::as_mips_relobj(object);
+
+ if (mips_obj->is_mips16_stub_section(data_shndx))
+ {
+ mips_obj->get_mips16_stub_section(data_shndx)
+ ->new_local_reloc_found(r_type, r_sym);
+ }
+
+ if (r_type == elfcpp::R_MIPS_NONE)
+ // R_MIPS_NONE is used in mips16 stub sections, to define the target of the
+ // mips16 stub.
+ return;
+
+ if (!mips16_call_reloc(r_type)
+ && !mips_obj->section_allows_mips16_refs(data_shndx))
+ // This reloc would need to refer to a MIPS16 hard-float stub, if
+ // there is one. We ignore MIPS16 stub sections and .pdr section when
+ // looking for relocs that would need to refer to MIPS16 stubs.
+ mips_obj->add_local_non_16bit_call(r_sym);
+
+ if (r_type == elfcpp::R_MIPS16_26
+ && !mips_obj->section_allows_mips16_refs(data_shndx))
+ mips_obj->add_local_16bit_call(r_sym);
+
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MIPS_GOT_OFST:
+ case elfcpp::R_MIPS_GOT_DISP:
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_GOT16:
+ case elfcpp::R_MICROMIPS_CALL16:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_GOT_PAGE:
+ case elfcpp::R_MICROMIPS_GOT_OFST:
+ case elfcpp::R_MICROMIPS_GOT_DISP:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ // We need a GOT section.
+ target->got_section(symtab, layout);
+ break;
+
+ default:
+ break;
+ }
+
+ if (call_lo16_reloc(r_type)
+ || got_lo16_reloc(r_type)
+ || got_disp_reloc(r_type))
+ {
+ // We may need a local GOT entry for this relocation. We
+ // don't count R_MIPS_GOT_PAGE because we can estimate the
+ // maximum number of pages needed by looking at the size of
+ // the segment. Similar comments apply to R_MIPS*_GOT16 and
+ // R_MIPS*_CALL16. We don't count R_MIPS_GOT_HI16, or
+ // R_MIPS_CALL_HI16 because these are always followed by an
+ // R_MIPS_GOT_LO16 or R_MIPS_CALL_LO16.
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ unsigned int r_sym = elfcpp::elf_r_sym<size>(r_info);
+ got->record_local_got_symbol(mips_obj, r_sym, r_addend, r_type, -1U);
+ }
+
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MICROMIPS_CALL16:
+ gold_error(_("CALL16 reloc at 0x%lx not against global symbol "),
+ (unsigned long)r_offset);
+ return;
+
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MICROMIPS_GOT_PAGE:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_GOT16:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ {
+ // This relocation needs a page entry in the GOT.
+ // Get the section contents.
+ section_size_type view_size = 0;
+ const unsigned char* view = object->section_contents(data_shndx,
+ &view_size, false);
+ view += r_offset;
+
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(view);
+ Valtype32 addend = (rel_type == elfcpp::SHT_REL ? val & 0xffff
+ : r_addend);
+
+ if (rel_type == elfcpp::SHT_REL && got16_reloc(r_type))
+ target->got16_addends_.push_back(got16_addend<size, big_endian>(
+ object, data_shndx, r_type, r_sym, addend));
+ else
+ target->got_section()->record_got_page_entry(mips_obj, r_sym, addend);
+ break;
+ }
+
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MICROMIPS_HI16:
+ // Record the reloc so that we can check whether the corresponding LO16
+ // part exists.
+ if (rel_type == elfcpp::SHT_REL)
+ target->got16_addends_.push_back(got16_addend<size, big_endian>(
+ object, data_shndx, r_type, r_sym, 0));
+ break;
+
+ case elfcpp::R_MIPS_LO16:
+ case elfcpp::R_MIPS16_LO16:
+ case elfcpp::R_MICROMIPS_LO16:
+ {
+ if (rel_type != elfcpp::SHT_REL)
+ break;
+
+ // Find corresponding GOT16/HI16 relocation.
+
+ // According to the MIPS ELF ABI, the R_MIPS_LO16 relocation must
+ // be immediately following. However, for the IRIX6 ABI, the next
+ // relocation may be a composed relocation consisting of several
+ // relocations for the same address. In that case, the R_MIPS_LO16
+ // relocation may occur as one of these. We permit a similar
+ // extension in general, as that is useful for GCC.
+
+ // In some cases GCC dead code elimination removes the LO16 but
+ // keeps the corresponding HI16. This is strictly speaking a
+ // violation of the ABI but not immediately harmful.
+
+ typename std::list<got16_addend<size, big_endian> >::iterator it =
+ target->got16_addends_.begin();
+ while (it != target->got16_addends_.end())
+ {
+ got16_addend<size, big_endian> _got16_addend = *it;
+
+ // TODO(sasa): Split got16_addends_ list into two lists - one for
+ // GOT16 relocs and the other for HI16 relocs.
+
+ // Report an error if we find HI16 or GOT16 reloc from the
+ // previous section without the matching LO16 part.
+ if (_got16_addend.object != object
+ || _got16_addend.shndx != data_shndx)
+ {
+ gold_error("Can't find matching LO16 reloc");
+ break;
+ }
+
+ if (_got16_addend.r_sym != r_sym
+ || !is_matching_lo16_reloc(_got16_addend.r_type, r_type))
+ {
+ ++it;
+ continue;
+ }
+
+ // We found a matching HI16 or GOT16 reloc for this LO16 reloc.
+ // For GOT16, we need to calculate combined addend and record GOT page
+ // entry.
+ if (got16_reloc(_got16_addend.r_type))
+ {
+
+ section_size_type view_size = 0;
+ const unsigned char* view = object->section_contents(data_shndx,
+ &view_size,
+ false);
+ view += r_offset;
+
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(view);
+ int32_t addend = Bits<16>::sign_extend32(val & 0xffff);
+
+ addend = (_got16_addend.addend << 16) + addend;
+ target->got_section()->record_got_page_entry(mips_obj, r_sym,
+ addend);
+ }
+
+ it = target->got16_addends_.erase(it);
+ }
+ break;
+ }
+ }
+
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_32:
+ case elfcpp::R_MIPS_REL32:
+ case elfcpp::R_MIPS_64:
+ {
+ if (parameters->options().output_is_position_independent())
+ {
+ // If building a shared library (or a position-independent
+ // executable), we need to create a dynamic relocation for
+ // this location.
+ Reloc_section* rel_dyn = target->rel_dyn_section(layout);
+ unsigned int r_sym = elfcpp::elf_r_sym<32>(r_info);
+ rel_dyn->add_symbolless_local_addend(object, r_sym,
+ elfcpp::R_MIPS_REL32,
+ output_section, data_shndx,
+ r_offset);
+ }
+ break;
+ }
+
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ {
+ unsigned int r_sym = elfcpp::elf_r_sym<size>(r_info);
+ bool output_is_shared = parameters->options().shared();
+ const tls::Tls_optimization optimized_type
+ = Target_mips<size, big_endian>::optimize_tls_reloc(
+ !output_is_shared, r_type);
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ if (optimized_type == tls::TLSOPT_NONE)
+ {
+ // Create a pair of GOT entries for the module index and
+ // dtv-relative offset.
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ unsigned int shndx = lsym.get_st_shndx();
+ bool is_ordinary;
+ shndx = object->adjust_sym_shndx(r_sym, shndx, &is_ordinary);
+ if (!is_ordinary)
+ {
+ object->error(_("local symbol %u has bad shndx %u"),
+ r_sym, shndx);
+ break;
+ }
+ got->record_local_got_symbol(mips_obj, r_sym, r_addend, r_type,
+ shndx);
+ }
+ else
+ {
+ // FIXME: TLS optimization not supported yet.
+ gold_unreachable();
+ }
+ break;
+
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ if (optimized_type == tls::TLSOPT_NONE)
+ {
+ // We always record LDM symbols as local with index 0.
+ target->got_section()->record_local_got_symbol(mips_obj, 0,
+ r_addend, r_type,
+ -1U);
+ }
+ else
+ {
+ // FIXME: TLS optimization not supported yet.
+ gold_unreachable();
+ }
+ break;
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ layout->set_has_static_tls();
+ if (optimized_type == tls::TLSOPT_NONE)
+ {
+ // Create a GOT entry for the tp-relative offset.
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ got->record_local_got_symbol(mips_obj, r_sym, r_addend, r_type,
+ -1U);
+ }
+ else
+ {
+ // FIXME: TLS optimization not supported yet.
+ gold_unreachable();
+ }
+ break;
+
+ default:
+ gold_unreachable();
+ }
+ }
+ break;
+
+ default:
+ break;
+ }
+
+ // Refuse some position-dependent relocations when creating a
+ // shared library. Do not refuse R_MIPS_32 / R_MIPS_64; they're
+ // not PIC, but we can create dynamic relocations and the result
+ // will be fine. Also do not refuse R_MIPS_LO16, which can be
+ // combined with R_MIPS_GOT16.
+ if (parameters->options().shared())
+ {
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MICROMIPS_HI16:
+ // Don't refuse a high part relocation if it's against
+ // no symbol (e.g. part of a compound relocation).
+ if (r_sym == 0)
+ break;
+
+ // FALLTHROUGH
+
+ case elfcpp::R_MIPS16_26:
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MICROMIPS_26_S1:
+ gold_error(_("%s: relocation %u against `%s' can not be used when "
+ "making a shared object; recompile with -fPIC"),
+ object->name().c_str(), r_type, "a local symbol");
+ default:
+ break;
+ }
+ }
+}
+
+template<int size, bool big_endian>
+inline void
+Target_mips<size, big_endian>::Scan::local(
+ Symbol_table* symtab,
+ Layout* layout,
+ Target_mips<size, big_endian>* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rel<size, big_endian>& reloc,
+ unsigned int r_type,
+ const elfcpp::Sym<size, big_endian>& lsym,
+ bool is_discarded)
+{
+ if (is_discarded)
+ return;
+
+ local(
+ symtab,
+ layout,
+ target,
+ object,
+ data_shndx,
+ output_section,
+ (const elfcpp::Rela<size, big_endian>*) NULL,
+ &reloc,
+ elfcpp::SHT_REL,
+ r_type,
+ lsym, is_discarded);
+}
+
+
+template<int size, bool big_endian>
+inline void
+Target_mips<size, big_endian>::Scan::local(
+ Symbol_table* symtab,
+ Layout* layout,
+ Target_mips<size, big_endian>* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>& reloc,
+ unsigned int r_type,
+ const elfcpp::Sym<size, big_endian>& lsym,
+ bool is_discarded)
+{
+ if (is_discarded)
+ return;
+
+ local(
+ symtab,
+ layout,
+ target,
+ object,
+ data_shndx,
+ output_section,
+ &reloc,
+ (const elfcpp::Rel<size, big_endian>*) NULL,
+ elfcpp::SHT_RELA,
+ r_type,
+ lsym, is_discarded);
+}
+
+// Scan a relocation for a global symbol.
+
+template<int size, bool big_endian>
+inline void
+Target_mips<size, big_endian>::Scan::global(
+ Symbol_table* symtab,
+ Layout* layout,
+ Target_mips<size, big_endian>* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>* rela,
+ const elfcpp::Rel<size, big_endian>* rel,
+ unsigned int rel_type,
+ unsigned int r_type,
+ Symbol* gsym)
+{
+ Mips_address r_offset;
+ typename elfcpp::Elf_types<size>::Elf_WXword r_info;
+ typename elfcpp::Elf_types<size>::Elf_Swxword r_addend;
+
+ if (rel_type == elfcpp::SHT_RELA)
+ {
+ r_offset = rela->get_r_offset();
+ r_info = rela->get_r_info();
+ r_addend = rela->get_r_addend();
+ }
+ else
+ {
+ r_offset = rel->get_r_offset();
+ r_info = rel->get_r_info();
+ r_addend = 0;
+ }
+
+ unsigned int r_sym = elfcpp::elf_r_sym<size>(r_info);
+ Mips_relobj<size, big_endian>* mips_obj =
+ Mips_relobj<size, big_endian>::as_mips_relobj(object);
+ Mips_symbol<size>* mips_sym = Mips_symbol<size>::as_mips_sym(gsym);
+
+ if (mips_obj->is_mips16_stub_section(data_shndx))
+ {
+ mips_obj->get_mips16_stub_section(data_shndx)
+ ->new_global_reloc_found(r_type, mips_sym);
+ }
+
+ if (r_type == elfcpp::R_MIPS_NONE)
+ // R_MIPS_NONE is used in mips16 stub sections, to define the target of the
+ // mips16 stub.
+ return;
+
+ if (!mips16_call_reloc(r_type)
+ && !mips_obj->section_allows_mips16_refs(data_shndx))
+ // This reloc would need to refer to a MIPS16 hard-float stub, if
+ // there is one. We ignore MIPS16 stub sections and .pdr section when
+ // looking for relocs that would need to refer to MIPS16 stubs.
+ mips_sym->set_need_fn_stub();
+
+ // A reference to _GLOBAL_OFFSET_TABLE_ implies that we need a got
+ // section. We check here to avoid creating a dynamic reloc against
+ // _GLOBAL_OFFSET_TABLE_.
+ if (!target->has_got_section()
+ && strcmp(gsym->name(), "_GLOBAL_OFFSET_TABLE_") == 0)
+ target->got_section(symtab, layout);
+
+ // We need PLT entries if there are static-only relocations against
+ // an externally-defined function. This can technically occur for
+ // shared libraries if there are branches to the symbol, although it
+ // is unlikely that this will be used in practice due to the short
+ // ranges involved. It can occur for any relative or absolute relocation
+ // in executables; in that case, the PLT entry becomes the function's
+ // canonical address.
+ bool static_reloc = false;
+
+ // Set CAN_MAKE_DYNAMIC to true if we can convert this
+ // relocation into a dynamic one.
+ bool can_make_dynamic = false;
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MIPS_GOT_OFST:
+ case elfcpp::R_MIPS_GOT_DISP:
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_GOT16:
+ case elfcpp::R_MICROMIPS_CALL16:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_GOT_PAGE:
+ case elfcpp::R_MICROMIPS_GOT_OFST:
+ case elfcpp::R_MICROMIPS_GOT_DISP:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ // We need a GOT section.
+ target->got_section(symtab, layout);
+ break;
+
+ // This is just a hint; it can safely be ignored. Don't set
+ // has_static_relocs for the corresponding symbol.
+ case elfcpp::R_MIPS_JALR:
+ case elfcpp::R_MICROMIPS_JALR:
+ break;
+
+ case elfcpp::R_MIPS_GPREL16:
+ case elfcpp::R_MIPS_GPREL32:
+ case elfcpp::R_MIPS16_GPREL:
+ case elfcpp::R_MICROMIPS_GPREL16:
+ // TODO(sasa)
+ // GP-relative relocations always resolve to a definition in a
+ // regular input file, ignoring the one-definition rule. This is
+ // important for the GP setup sequence in NewABI code, which
+ // always resolves to a local function even if other relocations
+ // against the symbol wouldn't.
+ //constrain_symbol_p = FALSE;
+ break;
+
+ case elfcpp::R_MIPS_32:
+ case elfcpp::R_MIPS_REL32:
+ case elfcpp::R_MIPS_64:
+ if (parameters->options().shared()
+ || strcmp(gsym->name(), "__gnu_local_gp") != 0)
+ {
+ if (r_type != elfcpp::R_MIPS_REL32)
+ {
+ static_reloc = true;
+ mips_sym->set_pointer_equality_needed();
+ }
+ can_make_dynamic = true;
+ break;
+ }
+ // Fall through.
+
+ default:
+ // Most static relocations require pointer equality, except
+ // for branches.
+ mips_sym->set_pointer_equality_needed();
+
+ // Fall through.
+
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MIPS_PC16:
+ case elfcpp::R_MIPS16_26:
+ case elfcpp::R_MICROMIPS_26_S1:
+ case elfcpp::R_MICROMIPS_PC7_S1:
+ case elfcpp::R_MICROMIPS_PC10_S1:
+ case elfcpp::R_MICROMIPS_PC16_S1:
+ case elfcpp::R_MICROMIPS_PC23_S2:
+ static_reloc = true;
+ mips_sym->set_has_static_relocs();
+ break;
+ }
+
+ // If there are call relocations against an externally-defined symbol,
+ // see whether we can create a MIPS lazy-binding stub for it. We can
+ // only do this if all references to the function are through call
+ // relocations, and in that case, the traditional lazy-binding stubs
+ // are much more efficient than PLT entries.
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MIPS_JALR:
+ case elfcpp::R_MICROMIPS_CALL16:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ case elfcpp::R_MICROMIPS_JALR:
+ if (!mips_sym->no_lazy_stub())
+ {
+ if ((mips_sym->needs_plt_entry() && mips_sym->is_from_dynobj())
+ // Calls from shared objects to undefined symbols of type
+ // STT_NOTYPE need lazy-binding stub.
+ || (mips_sym->is_undefined() && parameters->options().shared()))
+ target->mips_stubs_section(layout)->make_entry(mips_sym);
+ }
+ break;
+ default:
+ {
+ // We must not create a stub for a symbol that has relocations
+ // related to taking the function's address.
+ mips_sym->set_no_lazy_stub();
+ target->remove_lazy_stub_entry(mips_sym);
+ break;
+ }
+ }
+
+ if (relocation_needs_la25_stub<size, big_endian>(mips_obj, r_type,
+ mips_sym->is_mips16()))
+ mips_sym->set_has_nonpic_branches();
+
+ // R_MIPS_HI16 against _gp_disp is used for $gp setup,
+ // and has a special meaning.
+ bool gp_disp_against_hi16 = (!mips_obj->is_newabi()
+ && strcmp(gsym->name(), "_gp_disp") == 0
+ && (hi16_reloc(r_type) || lo16_reloc(r_type)));
+ if (static_reloc && gsym->needs_plt_entry())
+ {
+ target->make_plt_entry(symtab, layout, mips_sym, r_type);
+
+ // Since this is not a PC-relative relocation, we may be
+ // taking the address of a function. In that case we need to
+ // set the entry in the dynamic symbol table to the address of
+ // the PLT entry.
+ if (gsym->is_from_dynobj() && !parameters->options().shared())
+ {
+ gsym->set_needs_dynsym_value();
+ // We distinguish between PLT entries and lazy-binding stubs by
+ // giving the former an st_other value of STO_MIPS_PLT. Set the
+ // flag if there are any relocations in the binary where pointer
+ // equality matters.
+ if (mips_sym->pointer_equality_needed())
+ mips_sym->set_mips_plt();
+ }
+ }
+ if ((static_reloc || can_make_dynamic) && !gp_disp_against_hi16)
+ {
+ // Absolute addressing relocations.
+ // Make a dynamic relocation if necessary.
+ if (gsym->needs_dynamic_reloc(Scan::get_reference_flags(r_type)))
+ {
+ if (gsym->may_need_copy_reloc())
+ {
+ target->copy_reloc(symtab, layout, object,
+ data_shndx, output_section, gsym, *rel);
+ }
+ else if (can_make_dynamic)
+ {
+ // Create .rel.dyn section.
+ target->rel_dyn_section(layout);
+ target->dynamic_reloc(mips_sym, elfcpp::R_MIPS_REL32, mips_obj,
+ data_shndx, output_section, r_offset);
+ }
+ else
+ gold_error(_("non-dynamic relocations refer to dynamic symbol %s"),
+ gsym->name());
+ }
+ }
+
+ bool for_call = false;
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MICROMIPS_CALL16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ for_call = true;
+ // Fall through.
+
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_GOT16:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ case elfcpp::R_MIPS_GOT_DISP:
+ case elfcpp::R_MICROMIPS_GOT_DISP:
+ {
+ // The symbol requires a GOT entry.
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ got->record_global_got_symbol(mips_sym, mips_obj, r_type, false,
+ for_call);
+ mips_sym->set_global_got_area(GGA_NORMAL);
+ }
+ break;
+
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MICROMIPS_GOT_PAGE:
+ {
+ // This relocation needs a page entry in the GOT.
+ // Get the section contents.
+ section_size_type view_size = 0;
+ const unsigned char* view =
+ object->section_contents(data_shndx, &view_size, false);
+ view += r_offset;
+
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(view);
+ Valtype32 addend = (rel_type == elfcpp::SHT_REL ? val & 0xffff
+ : r_addend);
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ got->record_got_page_entry(mips_obj, r_sym, addend);
+
+ // If this is a global, overridable symbol, GOT_PAGE will
+ // decay to GOT_DISP, so we'll need a GOT entry for it.
+ bool def_regular = (mips_sym->source() == Symbol::FROM_OBJECT
+ && !mips_sym->object()->is_dynamic()
+ && !mips_sym->is_undefined());
+ if (!def_regular
+ || (parameters->options().output_is_position_independent()
+ && !parameters->options().Bsymbolic()
+ && !mips_sym->is_forced_local()))
+ {
+ got->record_global_got_symbol(mips_sym, mips_obj, r_type, false,
+ for_call);
+ mips_sym->set_global_got_area(GGA_NORMAL);
+ }
+ }
+ break;
+
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ {
+ const bool is_final = gsym->final_value_is_known();
+ const tls::Tls_optimization optimized_type =
+ Target_mips<size, big_endian>::optimize_tls_reloc(is_final, r_type);
+
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ if (optimized_type == tls::TLSOPT_NONE)
+ {
+ // Create a pair of GOT entries for the module index and
+ // dtv-relative offset.
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ got->record_global_got_symbol(mips_sym, mips_obj, r_type, false,
+ false);
+ }
+ else
+ {
+ // FIXME: TLS optimization not supported yet.
+ gold_unreachable();
+ }
+ break;
+
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ if (optimized_type == tls::TLSOPT_NONE)
+ {
+ // We always record LDM symbols as local with index 0.
+ target->got_section()->record_local_got_symbol(mips_obj, 0,
+ r_addend, r_type,
+ -1U);
+ }
+ else
+ {
+ // FIXME: TLS optimization not supported yet.
+ gold_unreachable();
+ }
+ break;
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ layout->set_has_static_tls();
+ if (optimized_type == tls::TLSOPT_NONE)
+ {
+ // Create a GOT entry for the tp-relative offset.
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ got->record_global_got_symbol(mips_sym, mips_obj, r_type, false,
+ false);
+ }
+ else
+ {
+ // FIXME: TLS optimization not supported yet.
+ gold_unreachable();
+ }
+ break;
+
+ default:
+ gold_unreachable();
+ }
+ }
+ break;
+ case elfcpp::R_MIPS_COPY:
+ case elfcpp::R_MIPS_JUMP_SLOT:
+ // These are relocations which should only be seen by the
+ // dynamic linker, and should never be seen here.
+ gold_error(_("%s: unexpected reloc %u in object file"),
+ object->name().c_str(), r_type);
+ break;
+
+ default:
+ break;
+ }
+
+ // Refuse some position-dependent relocations when creating a
+ // shared library. Do not refuse R_MIPS_32 / R_MIPS_64; they're
+ // not PIC, but we can create dynamic relocations and the result
+ // will be fine. Also do not refuse R_MIPS_LO16, which can be
+ // combined with R_MIPS_GOT16.
+ if (parameters->options().shared())
+ {
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MICROMIPS_HI16:
+ // Don't refuse a high part relocation if it's against
+ // no symbol (e.g. part of a compound relocation).
+ if (r_sym == 0)
+ break;
+
+ // R_MIPS_HI16 against _gp_disp is used for $gp setup,
+ // and has a special meaning.
+ if (!mips_obj->is_newabi() && strcmp(gsym->name(), "_gp_disp") == 0)
+ break;
+
+ // FALLTHROUGH
+
+ case elfcpp::R_MIPS16_26:
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MICROMIPS_26_S1:
+ gold_error(_("%s: relocation %u against `%s' can not be used when "
+ "making a shared object; recompile with -fPIC"),
+ object->name().c_str(), r_type, gsym->name());
+ default:
+ break;
+ }
+ }
+}
+
+template<int size, bool big_endian>
+inline void
+Target_mips<size, big_endian>::Scan::global(
+ Symbol_table* symtab,
+ Layout* layout,
+ Target_mips<size, big_endian>* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>& reloc,
+ unsigned int r_type,
+ Symbol* gsym)
+{
+ global(
+ symtab,
+ layout,
+ target,
+ object,
+ data_shndx,
+ output_section,
+ &reloc,
+ (const elfcpp::Rel<size, big_endian>*) NULL,
+ elfcpp::SHT_RELA,
+ r_type,
+ gsym);
+}
+
+template<int size, bool big_endian>
+inline void
+Target_mips<size, big_endian>::Scan::global(
+ Symbol_table* symtab,
+ Layout* layout,
+ Target_mips<size, big_endian>* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rel<size, big_endian>& reloc,
+ unsigned int r_type,
+ Symbol* gsym)
+{
+ global(
+ symtab,
+ layout,
+ target,
+ object,
+ data_shndx,
+ output_section,
+ (const elfcpp::Rela<size, big_endian>*) NULL,
+ &reloc,
+ elfcpp::SHT_REL,
+ r_type,
+ gsym);
+}
+
+// Return whether a R_MIPS_32 relocation needs to be applied.
+
+template<int size, bool big_endian>
+inline bool
+Target_mips<size, big_endian>::Relocate::should_apply_r_mips_32_reloc(
+ const Mips_symbol<size>* gsym,
+ unsigned int r_type,
+ Output_section* output_section,
+ Target_mips* target)
+{
+ // If the output section is not allocated, then we didn't call
+ // scan_relocs, we didn't create a dynamic reloc, and we must apply
+ // the reloc here.
+ if ((output_section->flags() & elfcpp::SHF_ALLOC) == 0)
+ return true;
+
+ if (gsym == NULL)
+ return true;
+ else
+ {
+ // For global symbols, we use the same helper routines used in the
+ // scan pass.
+ if (gsym->needs_dynamic_reloc(Scan::get_reference_flags(r_type))
+ && !gsym->may_need_copy_reloc())
+ {
+ // We have generated dynamic reloc (R_MIPS_REL32).
+
+ bool multi_got = false;
+ if (target->has_got_section())
+ multi_got = target->got_section()->multi_got();
+ bool has_got_offset;
+ if (!multi_got)
+ has_got_offset = gsym->has_got_offset(GOT_TYPE_STANDARD);
+ else
+ has_got_offset = gsym->global_gotoffset() != -1U;
+ if (!has_got_offset)
+ return true;
+ else
+ // Apply the relocation only if the symbol is in the local got.
+ // Do not apply the relocation if the symbol is in the global
+ // got.
+ return symbol_references_local(gsym, gsym->has_dynsym_index());
+ }
+ else
+ // We have not generated dynamic reloc.
+ return true;
+ }
+}
+
+// Perform a relocation.
+
+template<int size, bool big_endian>
+inline bool
+Target_mips<size, big_endian>::Relocate::relocate(
+ const Relocate_info<size, big_endian>* relinfo,
+ Target_mips* target,
+ Output_section* output_section,
+ size_t relnum,
+ const elfcpp::Rela<size, big_endian>* rela,
+ const elfcpp::Rel<size, big_endian>* rel,
+ unsigned int rel_type,
+ unsigned int r_type,
+ const Sized_symbol<size>* gsym,
+ const Symbol_value<size>* psymval,
+ unsigned char* view,
+ Mips_address address,
+ section_size_type)
+{
+ Mips_address r_offset;
+ typename elfcpp::Elf_types<size>::Elf_WXword r_info;
+ typename elfcpp::Elf_types<size>::Elf_Swxword r_addend;
+
+ if (rel_type == elfcpp::SHT_RELA)
+ {
+ r_offset = rela->get_r_offset();
+ r_info = rela->get_r_info();
+ r_addend = rela->get_r_addend();
+ }
+ else
+ {
+ r_offset = rel->get_r_offset();
+ r_info = rel->get_r_info();
+ r_addend = 0;
+ }
+
+ typedef Mips_relocate_functions<size, big_endian> Reloc_funcs;
+ typename Reloc_funcs::Status reloc_status = Reloc_funcs::STATUS_OKAY;
+
+ Mips_relobj<size, big_endian>* object =
+ Mips_relobj<size, big_endian>::as_mips_relobj(relinfo->object);
+
+ unsigned int r_sym = elfcpp::elf_r_sym<size>(r_info);
+ bool target_is_16_bit_code = false;
+ bool target_is_micromips_code = false;
+ bool cross_mode_jump;
+
+ Symbol_value<size> symval;
+
+ const Mips_symbol<size>* mips_sym = Mips_symbol<size>::as_mips_sym(gsym);
+
+ bool changed_symbol_value = false;
+ if (gsym == NULL)
+ {
+ target_is_16_bit_code = object->local_symbol_is_mips16(r_sym);
+ target_is_micromips_code = object->local_symbol_is_micromips(r_sym);
+ if (target_is_16_bit_code || target_is_micromips_code)
+ {
+ // MIPS16/microMIPS text labels should be treated as odd.
+ symval.set_output_value(psymval->value(object, 1));
+ psymval = &symval;
+ changed_symbol_value = true;
+ }
+ }
+ else
+ {
+ target_is_16_bit_code = mips_sym->is_mips16();
+ target_is_micromips_code = mips_sym->is_micromips();
+
+ // If this is a mips16/microMIPS text symbol, add 1 to the value to make
+ // it odd. This will cause something like .word SYM to come up with
+ // the right value when it is loaded into the PC.
+
+ if ((mips_sym->is_mips16() || mips_sym->is_micromips())
+ && psymval->value(object, 0) != 0)
+ {
+ symval.set_output_value(psymval->value(object, 0) | 1);
+ psymval = &symval;
+ changed_symbol_value = true;
+ }
+
+ // Pick the value to use for symbols defined in shared objects.
+ if (mips_sym->use_plt_offset(Scan::get_reference_flags(r_type))
+ || mips_sym->has_lazy_stub())
+ {
+ Mips_address value;
+ if (!mips_sym->has_lazy_stub())
+ {
+ // Prefer a standard MIPS PLT entry.
+ if (mips_sym->has_mips_plt_offset())
+ {
+ value = target->plt_section()->mips_entry_address(mips_sym);
+ target_is_micromips_code = false;
+ target_is_16_bit_code = false;
+ }
+ else
+ {
+ value = (target->plt_section()->comp_entry_address(mips_sym)
+ + 1);
+ if (target->is_output_micromips())
+ target_is_micromips_code = true;
+ else
+ target_is_16_bit_code = true;
+ }
+ }
+ else
+ value = target->mips_stubs_section()->stub_address(mips_sym);
+
+ symval.set_output_value(value);
+ psymval = &symval;
+ }
+ }
+
+ // TRUE if the symbol referred to by this relocation is "_gp_disp".
+ // Note that such a symbol must always be a global symbol.
+ bool gp_disp = (gsym != NULL && (strcmp(gsym->name(), "_gp_disp") == 0)
+ && !object->is_newabi());
+
+ // TRUE if the symbol referred to by this relocation is "__gnu_local_gp".
+ // Note that such a symbol must always be a global symbol.
+ bool gnu_local_gp = gsym && (strcmp(gsym->name(), "__gnu_local_gp") == 0);
+
+
+ if (gp_disp)
+ {
+ if (!hi16_reloc(r_type) && !lo16_reloc(r_type))
+ gold_error_at_location(relinfo, relnum, r_offset,
+ _("relocations against _gp_disp are permitted only"
+ " with R_MIPS_HI16 and R_MIPS_LO16 relocations."));
+ }
+ else if (gnu_local_gp)
+ {
+ // __gnu_local_gp is _gp symbol.
+ symval.set_output_value(target->adjusted_gp_value(object));
+ psymval = &symval;
+ }
+
+ // If this is a reference to a 16-bit function with a stub, we need
+ // to redirect the relocation to the stub unless:
+ //
+ // (a) the relocation is for a MIPS16 JAL;
+ //
+ // (b) the relocation is for a MIPS16 PIC call, and there are no
+ // non-MIPS16 uses of the GOT slot; or
+ //
+ // (c) the section allows direct references to MIPS16 functions.
+ if (r_type != elfcpp::R_MIPS16_26
+ && !parameters->options().relocatable()
+ && ((mips_sym != NULL
+ && mips_sym->has_mips16_fn_stub()
+ && (r_type != elfcpp::R_MIPS16_CALL16 || mips_sym->need_fn_stub()))
+ || (mips_sym == NULL
+ && object->get_local_mips16_fn_stub(r_sym) != NULL))
+ && !object->section_allows_mips16_refs(relinfo->data_shndx))
+ {
+ // This is a 32- or 64-bit call to a 16-bit function. We should
+ // have already noticed that we were going to need the
+ // stub.
+ Mips_address value;
+ if (mips_sym == NULL)
+ value = object->get_local_mips16_fn_stub(r_sym)->output_address();
+ else
+ {
+ gold_assert(mips_sym->need_fn_stub());
+ if (mips_sym->has_la25_stub())
+ value = target->la25_stub_section()->stub_address(mips_sym);
+ else
+ {
+ value = mips_sym->template
+ get_mips16_fn_stub<big_endian>()->output_address();
+ }
+ }
+ symval.set_output_value(value);
+ psymval = &symval;
+ changed_symbol_value = true;
+
+ // The target is 16-bit, but the stub isn't.
+ target_is_16_bit_code = false;
+ }
+ // If this is a MIPS16 call with a stub, that is made through the PLT or
+ // to a standard MIPS function, we need to redirect the call to the stub.
+ // Note that we specifically exclude R_MIPS16_CALL16 from this behavior;
+ // indirect calls should use an indirect stub instead.
+ else if (r_type == elfcpp::R_MIPS16_26 && !parameters->options().relocatable()
+ && ((mips_sym != NULL
+ && (mips_sym->has_mips16_call_stub()
+ || mips_sym->has_mips16_call_fp_stub()))
+ || (mips_sym == NULL
+ && object->get_local_mips16_call_stub(r_sym) != NULL))
+ && ((mips_sym != NULL && mips_sym->has_plt_offset())
+ || !target_is_16_bit_code))
+ {
+ Mips16_stub_section<size, big_endian>* call_stub;
+ if (mips_sym == NULL)
+ call_stub = object->get_local_mips16_call_stub(r_sym);
+ else
+ {
+ // If both call_stub and call_fp_stub are defined, we can figure
+ // out which one to use by checking which one appears in the input
+ // file.
+ if (mips_sym->has_mips16_call_stub()
+ && mips_sym->has_mips16_call_fp_stub())
+ {
+ call_stub = NULL;
+ for (unsigned int i = 1; i < object->shnum(); ++i)
+ {
+ if (object->is_mips16_call_fp_stub_section(i))
+ {
+ call_stub = mips_sym->template
+ get_mips16_call_fp_stub<big_endian>();
+ break;
+ }
+
+ }
+ if (call_stub == NULL)
+ call_stub =
+ mips_sym->template get_mips16_call_stub<big_endian>();
+ }
+ else if (mips_sym->has_mips16_call_stub())
+ call_stub = mips_sym->template get_mips16_call_stub<big_endian>();
+ else
+ call_stub = mips_sym->template get_mips16_call_fp_stub<big_endian>();
+ }
+
+ symval.set_output_value(call_stub->output_address());
+ psymval = &symval;
+ changed_symbol_value = true;
+ }
+ // If this is a direct call to a PIC function, redirect to the
+ // non-PIC stub.
+ else if (mips_sym != NULL
+ && mips_sym->has_la25_stub()
+ && relocation_needs_la25_stub<size, big_endian>(
+ object, r_type, target_is_16_bit_code))
+ {
+ Mips_address value = target->la25_stub_section()->stub_address(mips_sym);
+ if (mips_sym->is_micromips())
+ value += 1;
+ symval.set_output_value(value);
+ psymval = &symval;
+ }
+ // For direct MIPS16 and microMIPS calls make sure the compressed PLT
+ // entry is used if a standard PLT entry has also been made.
+ else if ((r_type == elfcpp::R_MIPS16_26
+ || r_type == elfcpp::R_MICROMIPS_26_S1)
+ && !parameters->options().relocatable()
+ && mips_sym != NULL
+ && mips_sym->has_plt_offset()
+ && mips_sym->has_comp_plt_offset()
+ && mips_sym->has_mips_plt_offset())
+ {
+ Mips_address value = (target->plt_section()->comp_entry_address(mips_sym)
+ + 1);
+ symval.set_output_value(value);
+ psymval = &symval;
+
+ target_is_16_bit_code = !target->is_output_micromips();
+ target_is_micromips_code = target->is_output_micromips();
+ }
+
+ // Make sure MIPS16 and microMIPS are not used together.
+ if ((r_type == elfcpp::R_MIPS16_26 && target_is_micromips_code)
+ || (micromips_branch_reloc(r_type) && target_is_16_bit_code))
+ {
+ gold_error(_("MIPS16 and microMIPS functions cannot call each other"));
+ }
+
+ // Calls from 16-bit code to 32-bit code and vice versa require the
+ // mode change. However, we can ignore calls to undefined weak symbols,
+ // which should never be executed at runtime. This exception is important
+ // because the assembly writer may have "known" that any definition of the
+ // symbol would be 16-bit code, and that direct jumps were therefore
+ // acceptable.
+ cross_mode_jump =
+ (!parameters->options().relocatable()
+ && !(gsym != NULL && gsym->is_weak_undefined())
+ && ((r_type == elfcpp::R_MIPS16_26 && !target_is_16_bit_code)
+ || (r_type == elfcpp::R_MICROMIPS_26_S1 && !target_is_micromips_code)
+ || ((r_type == elfcpp::R_MIPS_26 || r_type == elfcpp::R_MIPS_JALR)
+ && (target_is_16_bit_code || target_is_micromips_code))));
+
+ bool local = (mips_sym == NULL
+ || (mips_sym->got_only_for_calls()
+ ? symbol_calls_local(mips_sym, mips_sym->has_dynsym_index())
+ : symbol_references_local(mips_sym,
+ mips_sym->has_dynsym_index())));
+
+ // Global R_MIPS_GOT_PAGE/R_MICROMIPS_GOT_PAGE relocations are equivalent
+ // to R_MIPS_GOT_DISP/R_MICROMIPS_GOT_DISP. The addend is applied by the
+ // corresponding R_MIPS_GOT_OFST/R_MICROMIPS_GOT_OFST.
+ if (got_page_reloc(r_type) && !local)
+ r_type = (micromips_reloc(r_type) ? elfcpp::R_MICROMIPS_GOT_DISP
+ : elfcpp::R_MIPS_GOT_DISP);
+
+ unsigned int got_offset = 0;
+ int gp_offset = 0;
+
+ bool update_got_entry = false;
+ bool extract_addend = rel_type == elfcpp::SHT_REL;
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_NONE:
+ break;
+ case elfcpp::R_MIPS_16:
+ reloc_status = Reloc_funcs::rel16(view, object, psymval, r_addend,
+ extract_addend, r_type);
+ break;
+
+ case elfcpp::R_MIPS_32:
+ if (should_apply_r_mips_32_reloc(mips_sym, r_type, output_section,
+ target))
+ reloc_status = Reloc_funcs::rel32(view, object, psymval, r_addend,
+ extract_addend, r_type);
+ if (mips_sym != NULL
+ && (mips_sym->is_mips16() || mips_sym->is_micromips())
+ && mips_sym->global_got_area() == GGA_RELOC_ONLY)
+ {
+ // If mips_sym->has_mips16_fn_stub() is false, symbol value is
+ // already updated by adding +1.
+ if (mips_sym->has_mips16_fn_stub())
+ {
+ gold_assert(mips_sym->need_fn_stub());
+ Mips16_stub_section<size, big_endian>* fn_stub =
+ mips_sym->template get_mips16_fn_stub<big_endian>();
+
+ symval.set_output_value(fn_stub->output_address());
+ psymval = &symval;
+ }
+ got_offset = mips_sym->global_gotoffset();
+ update_got_entry = true;
+ }
+ break;
+
+ case elfcpp::R_MIPS_REL32:
+ gold_unreachable();
+
+ case elfcpp::R_MIPS_PC32:
+ reloc_status = Reloc_funcs::relpc32(view, object, psymval, address,
+ r_addend, extract_addend, r_type);
+ break;
+
+ case elfcpp::R_MIPS16_26:
+ // The calculation for R_MIPS16_26 is just the same as for an
+ // R_MIPS_26. It's only the storage of the relocated field into
+ // the output file that's different. So, we just fall through to the
+ // R_MIPS_26 case here.
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MICROMIPS_26_S1:
+ reloc_status = Reloc_funcs::rel26(view, object, psymval, address,
+ gsym == NULL, r_addend, extract_addend, gsym, cross_mode_jump, r_type,
+ target->jal_to_bal());
+ break;
+
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MICROMIPS_HI16:
+ reloc_status = Reloc_funcs::relhi16(view, object, psymval, r_addend,
+ address, gp_disp, r_type,
+ extract_addend);
+ break;
+
+ case elfcpp::R_MIPS_LO16:
+ case elfcpp::R_MIPS16_LO16:
+ case elfcpp::R_MICROMIPS_LO16:
+ case elfcpp::R_MICROMIPS_HI0_LO16:
+ reloc_status = Reloc_funcs::rello16(target, view, object, psymval,
+ r_addend, extract_addend, address,
+ gp_disp, r_type);
+ break;
+
+ case elfcpp::R_MIPS_LITERAL:
+ case elfcpp::R_MICROMIPS_LITERAL:
+ // Because we don't merge literal sections, we can handle this
+ // just like R_MIPS_GPREL16. In the long run, we should merge
+ // shared literals, and then we will need to additional work
+ // here.
+
+ // Fall through.
+
+ case elfcpp::R_MIPS_GPREL16:
+ case elfcpp::R_MIPS16_GPREL:
+ case elfcpp::R_MICROMIPS_GPREL7_S2:
+ case elfcpp::R_MICROMIPS_GPREL16:
+ reloc_status = Reloc_funcs::relgprel(view, object, psymval,
+ target->adjusted_gp_value(object),
+ r_addend, extract_addend,
+ gsym == NULL, r_type);
+ break;
+
+ case elfcpp::R_MIPS_PC16:
+ reloc_status = Reloc_funcs::relpc16(view, object, psymval, address,
+ r_addend, extract_addend, r_type);
+ break;
+ case elfcpp::R_MICROMIPS_PC7_S1:
+ reloc_status = Reloc_funcs::relmicromips_pc7_s1(view, object, psymval,
+ address, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MICROMIPS_PC10_S1:
+ reloc_status = Reloc_funcs::relmicromips_pc10_s1(view, object, psymval,
+ address, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MICROMIPS_PC16_S1:
+ reloc_status = Reloc_funcs::relmicromips_pc16_s1(view, object, psymval,
+ address, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_GPREL32:
+ reloc_status = Reloc_funcs::relgprel32(view, object, psymval,
+ target->adjusted_gp_value(object),
+ r_addend, extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ if (gsym != NULL)
+ got_offset = target->got_section()->got_offset(gsym, GOT_TYPE_STANDARD,
+ object);
+ else
+ got_offset = target->got_section()->got_offset(r_sym, GOT_TYPE_STANDARD,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot_hi16(view, gp_offset, r_type);
+ update_got_entry = changed_symbol_value;
+ break;
+
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ if (gsym != NULL)
+ got_offset = target->got_section()->got_offset(gsym, GOT_TYPE_STANDARD,
+ object);
+ else
+ got_offset = target->got_section()->got_offset(r_sym, GOT_TYPE_STANDARD,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot_lo16(view, gp_offset, r_type);
+ update_got_entry = changed_symbol_value;
+ break;
+
+ case elfcpp::R_MIPS_GOT_DISP:
+ case elfcpp::R_MICROMIPS_GOT_DISP:
+ if (gsym != NULL)
+ got_offset = target->got_section()->got_offset(gsym, GOT_TYPE_STANDARD,
+ object);
+ else
+ got_offset = target->got_section()->got_offset(r_sym, GOT_TYPE_STANDARD,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot(view, gp_offset, r_type);
+ break;
+
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MICROMIPS_CALL16:
+ gold_assert(gsym != NULL);
+ got_offset = target->got_section()->got_offset(gsym, GOT_TYPE_STANDARD,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot(view, gp_offset, r_type);
+ // TODO(sasa): We should also initialize update_got_entry in other places
+ // where relgot is called.
+ update_got_entry = changed_symbol_value;
+ break;
+
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MICROMIPS_GOT16:
+ if (gsym != NULL)
+ {
+ got_offset = target->got_section()->got_offset(gsym,
+ GOT_TYPE_STANDARD,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot(view, gp_offset, r_type);
+ }
+ else
+ reloc_status = Reloc_funcs::relgot16_local(view, object, psymval,
+ r_addend, extract_addend,
+ r_type);
+ update_got_entry = changed_symbol_value;
+ break;
+
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ if (gsym != NULL)
+ got_offset = target->got_section()->got_offset(gsym, GOT_TYPE_TLS_PAIR,
+ object);
+ else
+ got_offset = target->got_section()->got_offset(r_sym, GOT_TYPE_TLS_PAIR,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot(view, gp_offset, r_type);
+ break;
+
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ if (gsym != NULL)
+ got_offset = target->got_section()->got_offset(gsym,
+ GOT_TYPE_TLS_OFFSET,
+ object);
+ else
+ got_offset = target->got_section()->got_offset(r_sym,
+ GOT_TYPE_TLS_OFFSET,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot(view, gp_offset, r_type);
+ break;
+
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ // Relocate the field with the offset of the GOT entry for
+ // the module index.
+ got_offset = target->got_section()->tls_ldm_offset(object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot(view, gp_offset, r_type);
+ break;
+
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MICROMIPS_GOT_PAGE:
+ reloc_status = Reloc_funcs::relgotpage(target, view, object, psymval,
+ r_addend, extract_addend, r_type);
+ break;
+
+ case elfcpp::R_MIPS_GOT_OFST:
+ case elfcpp::R_MICROMIPS_GOT_OFST:
+ reloc_status = Reloc_funcs::relgotofst(target, view, object, psymval,
+ r_addend, extract_addend, local,
+ r_type);
+ break;
+
+ case elfcpp::R_MIPS_JALR:
+ case elfcpp::R_MICROMIPS_JALR:
+ // This relocation is only a hint. In some cases, we optimize
+ // it into a bal instruction. But we don't try to optimize
+ // when the symbol does not resolve locally.
+ if (gsym == NULL || symbol_calls_local(gsym, gsym->has_dynsym_index()))
+ reloc_status = Reloc_funcs::reljalr(view, object, psymval, address,
+ r_addend, extract_addend,
+ cross_mode_jump, r_type,
+ target->jalr_to_bal(),
+ target->jr_to_b());
+ break;
+
+ case elfcpp::R_MIPS_TLS_DTPREL_HI16:
+ case elfcpp::R_MIPS16_TLS_DTPREL_HI16:
+ case elfcpp::R_MICROMIPS_TLS_DTPREL_HI16:
+ reloc_status = Reloc_funcs::tlsrelhi16(view, object, psymval,
+ elfcpp::DTP_OFFSET, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_TLS_DTPREL_LO16:
+ case elfcpp::R_MIPS16_TLS_DTPREL_LO16:
+ case elfcpp::R_MICROMIPS_TLS_DTPREL_LO16:
+ reloc_status = Reloc_funcs::tlsrello16(view, object, psymval,
+ elfcpp::DTP_OFFSET, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_TLS_DTPREL32:
+ case elfcpp::R_MIPS_TLS_DTPREL64:
+ reloc_status = Reloc_funcs::tlsrel32(view, object, psymval,
+ elfcpp::DTP_OFFSET, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_TLS_TPREL_HI16:
+ case elfcpp::R_MIPS16_TLS_TPREL_HI16:
+ case elfcpp::R_MICROMIPS_TLS_TPREL_HI16:
+ reloc_status = Reloc_funcs::tlsrelhi16(view, object, psymval,
+ elfcpp::TP_OFFSET, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_TLS_TPREL_LO16:
+ case elfcpp::R_MIPS16_TLS_TPREL_LO16:
+ case elfcpp::R_MICROMIPS_TLS_TPREL_LO16:
+ reloc_status = Reloc_funcs::tlsrello16(view, object, psymval,
+ elfcpp::TP_OFFSET, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_TLS_TPREL32:
+ case elfcpp::R_MIPS_TLS_TPREL64:
+ reloc_status = Reloc_funcs::tlsrel32(view, object, psymval,
+ elfcpp::TP_OFFSET, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_SUB:
+ case elfcpp::R_MICROMIPS_SUB:
+ reloc_status = Reloc_funcs::relsub(view, object, psymval, r_addend,
+ extract_addend, r_type);
+ break;
+ default:
+ gold_error_at_location(relinfo, relnum, r_offset,
+ _("unsupported reloc %u"), r_type);
+ break;
+ }
+
+ if (update_got_entry)
+ {
+ Mips_output_data_got<size, big_endian>* got = target->got_section();
+ if (mips_sym != NULL && mips_sym->get_applied_secondary_got_fixup())
+ got->update_got_entry(got->get_primary_got_offset(mips_sym),
+ psymval->value(object, 0));
+ else
+ got->update_got_entry(got_offset, psymval->value(object, 0));
+ }
+
+ // Report any errors.
+ switch (reloc_status)
+ {
+ case Reloc_funcs::STATUS_OKAY:
+ break;
+ case Reloc_funcs::STATUS_OVERFLOW:
+ gold_error_at_location(relinfo, relnum, r_offset,
+ _("relocation overflow"));
+ break;
+ case Reloc_funcs::STATUS_BAD_RELOC:
+ gold_error_at_location(relinfo, relnum, r_offset,
+ _("unexpected opcode while processing relocation"));
+ break;
+ default:
+ gold_unreachable();
+ }
+
+ return true;
+}
+
+template<int size, bool big_endian>
+inline bool
+Target_mips<size, big_endian>::Relocate::relocate(
+ const Relocate_info<size, big_endian>* relinfo,
+ Target_mips* target,
+ Output_section* output_section,
+ size_t relnum,
+ const elfcpp::Rela<size, big_endian>& reloc,
+ unsigned int r_type,
+ const Sized_symbol<size>* gsym,
+ const Symbol_value<size>* psymval,
+ unsigned char* view,
+ Mips_address address,
+ section_size_type view_size)
+{
+ return relocate(
+ relinfo,
+ target,
+ output_section,
+ relnum,
+ &reloc,
+ (const elfcpp::Rel<size, big_endian>*) NULL,
+ elfcpp::SHT_RELA,
+ r_type,
+ gsym,
+ psymval,
+ view,
+ address,
+ view_size);
+}
+
+template<int size, bool big_endian>
+inline bool
+Target_mips<size, big_endian>::Relocate::relocate(
+ const Relocate_info<size, big_endian>* relinfo,
+ Target_mips* target,
+ Output_section* output_section,
+ size_t relnum,
+ const elfcpp::Rel<size, big_endian>& reloc,
+ unsigned int r_type,
+ const Sized_symbol<size>* gsym,
+ const Symbol_value<size>* psymval,
+ unsigned char* view,
+ Mips_address address,
+ section_size_type view_size)
+{
+ return relocate(
+ relinfo,
+ target,
+ output_section,
+ relnum,
+ (const elfcpp::Rela<size, big_endian>*) NULL,
+ &reloc,
+ elfcpp::SHT_REL,
+ r_type,
+ gsym,
+ psymval,
+ view,
+ address,
+ view_size);
+}
+
+// Get the Reference_flags for a particular relocation.
+
+template<int size, bool big_endian>
+int
+Target_mips<size, big_endian>::Scan::get_reference_flags(
+ unsigned int r_type)
+{
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_NONE:
+ // No symbol reference.
+ return 0;
+
+ case elfcpp::R_MIPS_16:
+ case elfcpp::R_MIPS_32:
+ case elfcpp::R_MIPS_64:
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MIPS_LO16:
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MIPS16_LO16:
+ case elfcpp::R_MICROMIPS_HI16:
+ case elfcpp::R_MICROMIPS_LO16:
+ return Symbol::ABSOLUTE_REF;
+
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MIPS16_26:
+ case elfcpp::R_MICROMIPS_26_S1:
+ return Symbol::FUNCTION_CALL | Symbol::ABSOLUTE_REF;
+
+ case elfcpp::R_MIPS_GPREL32:
+ case elfcpp::R_MIPS_GPREL16:
+ case elfcpp::R_MIPS_REL32:
+ case elfcpp::R_MIPS16_GPREL:
+ return Symbol::RELATIVE_REF;
+
+ case elfcpp::R_MIPS_PC16:
+ case elfcpp::R_MIPS_PC32:
+ case elfcpp::R_MIPS_JALR:
+ case elfcpp::R_MICROMIPS_JALR:
+ return Symbol::FUNCTION_CALL | Symbol::RELATIVE_REF;
+
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS_GOT_DISP:
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MIPS_LITERAL:
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MIPS_GOT_OFST:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MICROMIPS_GOT16:
+ case elfcpp::R_MICROMIPS_CALL16:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ // Absolute in GOT.
+ return Symbol::RELATIVE_REF;
+
+ case elfcpp::R_MIPS_TLS_DTPMOD32:
+ case elfcpp::R_MIPS_TLS_DTPREL32:
+ case elfcpp::R_MIPS_TLS_DTPMOD64:
+ case elfcpp::R_MIPS_TLS_DTPREL64:
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS_TLS_DTPREL_HI16:
+ case elfcpp::R_MIPS_TLS_DTPREL_LO16:
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS_TLS_TPREL32:
+ case elfcpp::R_MIPS_TLS_TPREL64:
+ case elfcpp::R_MIPS_TLS_TPREL_HI16:
+ case elfcpp::R_MIPS_TLS_TPREL_LO16:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_TPREL_HI16:
+ case elfcpp::R_MICROMIPS_TLS_TPREL_LO16:
+ return Symbol::TLS_REF;
+
+ case elfcpp::R_MIPS_COPY:
+ case elfcpp::R_MIPS_JUMP_SLOT:
+ default:
+ gold_unreachable();
+ // Not expected. We will give an error later.
+ return 0;
+ }
+}
+
+// Report an unsupported relocation against a local symbol.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::Scan::unsupported_reloc_local(
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int r_type)
+{
+ gold_error(_("%s: unsupported reloc %u against local symbol"),
+ object->name().c_str(), r_type);
+}
+
+// Report an unsupported relocation against a global symbol.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::Scan::unsupported_reloc_global(
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int r_type,
+ Symbol* gsym)
+{
+ gold_error(_("%s: unsupported reloc %u against global symbol %s"),
+ object->name().c_str(), r_type, gsym->demangled_name().c_str());
+}
+
+// Return printable name for ABI.
+template<int size, bool big_endian>
+const char*
+Target_mips<size, big_endian>::elf_mips_abi_name(elfcpp::Elf_Word e_flags,
+ unsigned char ei_class)
+{
+ switch (e_flags & elfcpp::EF_MIPS_ABI)
+ {
+ case 0:
+ if ((e_flags & elfcpp::EF_MIPS_ABI2) != 0)
+ return "N32";
+ else if (elfcpp::abi_64(ei_class))
+ return "64";
+ else
+ return "none";
+ case elfcpp::E_MIPS_ABI_O32:
+ return "O32";
+ case elfcpp::E_MIPS_ABI_O64:
+ return "O64";
+ case elfcpp::E_MIPS_ABI_EABI32:
+ return "EABI32";
+ case elfcpp::E_MIPS_ABI_EABI64:
+ return "EABI64";
+ default:
+ return "unknown abi";
+ }
+}
+
+template<int size, bool big_endian>
+const char*
+Target_mips<size, big_endian>::elf_mips_mach_name(elfcpp::Elf_Word e_flags)
+{
+ switch (e_flags & elfcpp::EF_MIPS_MACH)
+ {
+ case elfcpp::E_MIPS_MACH_3900:
+ return "mips:3900";
+ case elfcpp::E_MIPS_MACH_4010:
+ return "mips:4010";
+ case elfcpp::E_MIPS_MACH_4100:
+ return "mips:4100";
+ case elfcpp::E_MIPS_MACH_4111:
+ return "mips:4111";
+ case elfcpp::E_MIPS_MACH_4120:
+ return "mips:4120";
+ case elfcpp::E_MIPS_MACH_4650:
+ return "mips:4650";
+ case elfcpp::E_MIPS_MACH_5400:
+ return "mips:5400";
+ case elfcpp::E_MIPS_MACH_5500:
+ return "mips:5500";
+ case elfcpp::E_MIPS_MACH_SB1:
+ return "mips:sb1";
+ case elfcpp::E_MIPS_MACH_9000:
+ return "mips:9000";
+ case elfcpp::E_MIPS_MACH_LS2E:
+ return "mips:loongson-2e";
+ case elfcpp::E_MIPS_MACH_LS2F:
+ return "mips:loongson-2f";
+ case elfcpp::E_MIPS_MACH_LS3A:
+ return "mips:loongson-3a";
+ case elfcpp::E_MIPS_MACH_OCTEON:
+ return "mips:octeon";
+ case elfcpp::E_MIPS_MACH_OCTEON2:
+ return "mips:octeon2";
+ case elfcpp::E_MIPS_MACH_XLR:
+ return "mips:xlr";
+ default:
+ switch (e_flags & elfcpp::EF_MIPS_ARCH)
+ {
+ default:
+ case elfcpp::E_MIPS_ARCH_1:
+ return "mips:3000";
+
+ case elfcpp::E_MIPS_ARCH_2:
+ return "mips:6000";
+
+ case elfcpp::E_MIPS_ARCH_3:
+ return "mips:4000";
+
+ case elfcpp::E_MIPS_ARCH_4:
+ return "mips:8000";
+
+ case elfcpp::E_MIPS_ARCH_5:
+ return "mips:mips5";
+
+ case elfcpp::E_MIPS_ARCH_32:
+ return "mips:isa32";
+
+ case elfcpp::E_MIPS_ARCH_64:
+ return "mips:isa64";
+
+ case elfcpp::E_MIPS_ARCH_32R2:
+ return "mips:isa32r2";
+
+ case elfcpp::E_MIPS_ARCH_64R2:
+ return "mips:isa64r2";
+ }
+ }
+ return "unknown CPU";
+}
+
+template<int size, bool big_endian>
+Target::Target_info Target_mips<size, big_endian>::mips_info =
+{
+ size, // size
+ big_endian, // is_big_endian
+ elfcpp::EM_MIPS, // machine_code
+ true, // has_make_symbol
+ false, // has_resolve
+ false, // has_code_fill
+ true, // is_default_stack_executable
+ false, // can_icf_inline_merge_sections
+ '\0', // wrap_char
+ "/lib/ld.so.1", // dynamic_linker
+ 0x400000, // default_text_segment_address
+ 64 * 1024, // abi_pagesize (overridable by -z max-page-size)
+ 4 * 1024, // common_pagesize (overridable by -z common-page-size)
+ false, // isolate_execinstr
+ 0, // rosegment_gap
+ elfcpp::SHN_UNDEF, // small_common_shndx
+ elfcpp::SHN_UNDEF, // large_common_shndx
+ 0, // small_common_section_flags
+ 0, // large_common_section_flags
+ NULL, // attributes_section
+ NULL, // attributes_vendor
+ "__start" // entry_symbol_name
+};
+
+// The selector for mips object files.
+
+template<int size, bool big_endian>
+class Target_selector_mips : public Target_selector
+{
+public:
+ Target_selector_mips()
+ : Target_selector(elfcpp::EM_MIPS, size, big_endian,
+ (size == 64 ?
+ (big_endian ? "elf64-tradbigmips" : "elf64-tradlittlemips") :
+ (big_endian ? "elf32-tradbigmips" : "elf32-tradlittlemips")),
+ (size == 64 ?
+ (big_endian ? "elf64-tradbigmips" : "elf64-tradlittlemips") :
+ (big_endian ? "elf32-tradbigmips" : "elf32-tradlittlemips")))
+ { }
+
+ Target* do_instantiate_target()
+ { return new Target_mips<size, big_endian>(); }
+};
+
+Target_selector_mips<32, true> target_selector_mips32be;
+Target_selector_mips<32, false> target_selector_mips32;
+Target_selector_mips<64, true> target_selector_mips64be;
+Target_selector_mips<64, false> target_selector_mips64;
+
+
+} // End anonymous namespace.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Mips target in gold - revision 3 - part 2
2014-04-15 13:53 Mips target in gold - revision 3 - part 2 Sasa Stankovic
@ 2014-04-15 20:46 ` Cary Coutant
2014-04-15 21:08 ` Cary Coutant
` (2 more replies)
0 siblings, 3 replies; 9+ messages in thread
From: Cary Coutant @ 2014-04-15 20:46 UTC (permalink / raw)
To: Sasa Stankovic; +Cc: Rich Fuhler, binutils, iant, Petar Jovanovic
>> If I understand the code correctly, in this case the HI16 reloc
>> will remain on the got16_addends_ list forever. Is there a point
>> where you know it's safe to remove it? Can you clear the list
>> at the end of each section?
>
> If LO16 is removed by GCC, than HI16 reloc is paired with the next found LO16
> reloc (this means that two or more HI16 reloc will be paired with a single LO16 reloc).
> This is what GNU ld does - it always tries to find LO16 reloc for HI16 reloc, and if it
> can't find it, error is reported.
>
> When trying to find LO16 reloc, GNU ld examines next relocations in the section until
> it finds it, or reports an error. I implemented it differently in Gold - when HI16 or GOT16
> reloc is found, it is recorded in the got16_addends_ list, and the normal scaning of
> relocations continues. Then when the LO16 reloc is found, the got16_addends_ list is
> examined and all pending matching HI16 and GOT16 relocs are removed. The error when
> HI16 or GOT16 reloc doesn't have the LO16 part is detected by checking whether the
> section of the pending HI16 or GOT16 relocation is different from the section of the current
> LO16 relocation. The check whether the final section that is scanned has HI16 or GOT16
> without the LO16 part is in the do_finalize_sections method - I just check whether the
> got16_addends_ list is empty - if it isn't, error is reported.
Ah, I see.
>> > typename std::list<got16_addend<size, big_endian> > got16_addends_;
>>
>> Do you realy want to use a linked list here? Wouldn't a
>> vector be better?
>
> I used std::list because erase is called on it.
Sorry, I missed that. And now I understand that you don't expect the
list to ever grow very long.
> // Check whether the final section that was scaned has HI16 or GOT16
Typo: "scanned".
> // Fill the unused space with zeroes.
> // TODO(sasa): Can we strip unused bytes during the relaxation?
> unsigned char* end = oview + oview_size;
> while (pov < end)
> {
> elfcpp::Swap<32, big_endian>::writeval(pov, 0);
> pov += 4;
> }
You can just use memset here:
if (unused > 0)
memset(pov, 0, unused);
If that's OK with you, I'll commit it with that change.
Thanks for working on this, and for putting up with the long review process!
-cary
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Mips target in gold - revision 3 - part 2
2014-04-15 20:46 ` Cary Coutant
@ 2014-04-15 21:08 ` Cary Coutant
2014-06-04 21:01 ` Sasa Stankovic
2014-06-04 21:03 ` Sasa Stankovic
2 siblings, 0 replies; 9+ messages in thread
From: Cary Coutant @ 2014-04-15 21:08 UTC (permalink / raw)
To: Sasa Stankovic; +Cc: Rich Fuhler, binutils, iant, Petar Jovanovic
One more issue I just noticed...
In Mips_output_section_reginfo::do_write:
> const unsigned char* section_contents =
> relobj->section_contents(shndx, §ion_size, false);
>
> gprmask |= elfcpp::Swap<size, big_endian>::readval(section_contents);
> cprmask1 |= elfcpp::Swap<size, big_endian>::readval(section_contents + 4);
> cprmask2 |= elfcpp::Swap<size, big_endian>::readval(section_contents + 8);
> cprmask3 |=
> elfcpp::Swap<size, big_endian>::readval(section_contents + 12);
> cprmask4 |=
> elfcpp::Swap<size, big_endian>::readval(section_contents + 16);
Going back to the input sections at this point could be a performance
issue, especially for very large links. This is essentially another
pass over the input files, and if we've run out of file descriptors,
we'll have to close and reopen all the input files one more time each.
It would be better to read these values and accumulate the five masks
in do_read_symbols(), and you won't have to call
set_always_keeps_input_sections() for Mips_output_section_reginfo.
-cary
^ permalink raw reply [flat|nested] 9+ messages in thread
* RE: Mips target in gold - revision 3 - part 2
2014-04-15 20:46 ` Cary Coutant
2014-04-15 21:08 ` Cary Coutant
@ 2014-06-04 21:01 ` Sasa Stankovic
2014-06-04 21:32 ` Sasa Stankovic
2014-06-04 21:03 ` Sasa Stankovic
2 siblings, 1 reply; 9+ messages in thread
From: Sasa Stankovic @ 2014-06-04 21:01 UTC (permalink / raw)
To: Cary Coutant; +Cc: Rich Fuhler, binutils, Petar Jovanovic
[-- Attachment #1: Type: text/plain, Size: 4389 bytes --]
Hi Cary,
Attached is the first part of the updated patch. Sorry for the delay; I didn't respond immediately after your last message because I thought there will be more reviews.
> > // Fill the unused space with zeroes.
> > // TODO(sasa): Can we strip unused bytes during the relaxation?
> > unsigned char* end = oview + oview_size;
> > while (pov < end)
> > {
> > elfcpp::Swap<32, big_endian>::writeval(pov, 0);
> > pov += 4;
> > }
>
> You can just use memset here:
>
> if (unused > 0)
> memset(pov, 0, unused);
Done.
> One more issue I just noticed...
>
> In Mips_output_section_reginfo::do_write:
>
> > const unsigned char* section_contents =
> > relobj->section_contents(shndx, §ion_size, false);
> >
> > gprmask |= elfcpp::Swap<size, big_endian>::readval(section_contents);
> > cprmask1 |= elfcpp::Swap<size, big_endian>::readval(section_contents + 4);
> > cprmask2 |= elfcpp::Swap<size, big_endian>::readval(section_contents + 8);
> > cprmask3 |=
> > elfcpp::Swap<size, big_endian>::readval(section_contents + 12);
> > cprmask4 |=
> > elfcpp::Swap<size, big_endian>::readval(section_contents + 16);
>
> Going back to the input sections at this point could be a performance
> issue, especially for very large links. This is essentially another
> pass over the input files, and if we've run out of file descriptors,
> we'll have to close and reopen all the input files one more time each.
> It would be better to read these values and accumulate the five masks
> in do_read_symbols(), and you won't have to call
> set_always_keeps_input_sections() for Mips_output_section_reginfo.
Done.
Regards,
Sasa
________________________________________
From: Cary Coutant [ccoutant@google.com]
Sent: Tuesday, April 15, 2014 10:46 PM
To: Sasa Stankovic
Cc: Rich Fuhler; binutils@sourceware.org; iant@google.com; Petar Jovanovic
Subject: Re: Mips target in gold - revision 3 - part 2
>> If I understand the code correctly, in this case the HI16 reloc
>> will remain on the got16_addends_ list forever. Is there a point
>> where you know it's safe to remove it? Can you clear the list
>> at the end of each section?
>
> If LO16 is removed by GCC, than HI16 reloc is paired with the next found LO16
> reloc (this means that two or more HI16 reloc will be paired with a single LO16 reloc).
> This is what GNU ld does - it always tries to find LO16 reloc for HI16 reloc, and if it
> can't find it, error is reported.
>
> When trying to find LO16 reloc, GNU ld examines next relocations in the section until
> it finds it, or reports an error. I implemented it differently in Gold - when HI16 or GOT16
> reloc is found, it is recorded in the got16_addends_ list, and the normal scaning of
> relocations continues. Then when the LO16 reloc is found, the got16_addends_ list is
> examined and all pending matching HI16 and GOT16 relocs are removed. The error when
> HI16 or GOT16 reloc doesn't have the LO16 part is detected by checking whether the
> section of the pending HI16 or GOT16 relocation is different from the section of the current
> LO16 relocation. The check whether the final section that is scanned has HI16 or GOT16
> without the LO16 part is in the do_finalize_sections method - I just check whether the
> got16_addends_ list is empty - if it isn't, error is reported.
Ah, I see.
>> > typename std::list<got16_addend<size, big_endian> > got16_addends_;
>>
>> Do you realy want to use a linked list here? Wouldn't a
>> vector be better?
>
> I used std::list because erase is called on it.
Sorry, I missed that. And now I understand that you don't expect the
list to ever grow very long.
> // Check whether the final section that was scaned has HI16 or GOT16
Typo: "scanned".
> // Fill the unused space with zeroes.
> // TODO(sasa): Can we strip unused bytes during the relaxation?
> unsigned char* end = oview + oview_size;
> while (pov < end)
> {
> elfcpp::Swap<32, big_endian>::writeval(pov, 0);
> pov += 4;
> }
You can just use memset here:
if (unused > 0)
memset(pov, 0, unused);
If that's OK with you, I'll commit it with that change.
Thanks for working on this, and for putting up with the long review process!
-cary
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: config.patch --]
[-- Type: text/x-patch; name="config.patch", Size: 2761 bytes --]
diff --git a/elfcpp/mips.h b/elfcpp/mips.h
index 91f7270..ccb017f 100644
--- a/elfcpp/mips.h
+++ b/elfcpp/mips.h
@@ -1,7 +1,7 @@
// mips.h -- ELF definitions specific to EM_MIPS -*- C++ -*-
// Copyright (C) 2012-2014 Free Software Foundation, Inc.
-// Written by Sasa Stankovic <sasa.stankovic@rt-rk.com>
+// Written by Sasa Stankovic <sasa.stankovic@imgtec.com>
// and Aleksandar Simeonov <aleksandar.simeonov@rt-rk.com>.
// This file is part of elfcpp.
diff --git a/gold/ChangeLog b/gold/ChangeLog
index 5e15f5e..a7b8bd4 100644
--- a/gold/ChangeLog
+++ b/gold/ChangeLog
@@ -1,3 +1,13 @@
+2014-04-09 Sasa Stankovic <Sasa.Stankovic@imgtec.com>
+
+ * mips.cc: New file.
+ * Makefile.am (TARGETSOURCES): Add mips.cc
+ (ALL_TARGETOBJS): Add mips.$(OBJEXT)
+ * configure.tgt: Add entries for mips*.
+ * configure.ac: Likewise.
+ * Makefile.in: Regenerate.
+ * configure: Likewise.
+
2014-06-03 Alan Modra <amodra@gmail.com>
* powerpc.cc (addis_12_2): Define.
diff --git a/gold/Makefile.am b/gold/Makefile.am
index 78f0271..c2cfe3f 100644
--- a/gold/Makefile.am
+++ b/gold/Makefile.am
@@ -163,11 +163,12 @@ DEFFILES = arm-reloc.def
EXTRA_DIST = yyscript.c yyscript.h
TARGETSOURCES = \
- i386.cc x86_64.cc sparc.cc powerpc.cc arm.cc arm-reloc-property.cc tilegx.cc
+ i386.cc x86_64.cc sparc.cc powerpc.cc arm.cc arm-reloc-property.cc tilegx.cc \
+ mips.cc
ALL_TARGETOBJS = \
i386.$(OBJEXT) x86_64.$(OBJEXT) sparc.$(OBJEXT) powerpc.$(OBJEXT) \
- arm.$(OBJEXT) arm-reloc-property.$(OBJEXT) tilegx.$(OBJEXT)
+ arm.$(OBJEXT) arm-reloc-property.$(OBJEXT) tilegx.$(OBJEXT) mips.$(OBJEXT)
libgold_a_SOURCES = $(CCFILES) $(HFILES) $(YFILES) $(DEFFILES)
libgold_a_LIBADD = $(LIBOBJS)
diff --git a/gold/configure.ac b/gold/configure.ac
index 7ad7302..14ecd72 100644
--- a/gold/configure.ac
+++ b/gold/configure.ac
@@ -205,6 +205,7 @@ for targ in $target $canon_targets; do
AM_CONDITIONAL(DEFAULT_TARGET_SPARC, test "$targ_obj" = "sparc")
AM_CONDITIONAL(DEFAULT_TARGET_X86_64, test "$targ_obj" = "x86_64")
AM_CONDITIONAL(DEFAULT_TARGET_TILEGX, test "$targ_obj" = "tilegx")
+ AM_CONDITIONAL(DEFAULT_TARGET_MIPS, test "$targ_obj" = "mips")
DEFAULT_TARGET=${targ_obj}
AC_SUBST(DEFAULT_TARGET)
fi
diff --git a/gold/configure.tgt b/gold/configure.tgt
index aad4f54..7add4c4 100644
--- a/gold/configure.tgt
+++ b/gold/configure.tgt
@@ -144,6 +144,20 @@ arm*-*-*)
targ_big_endian=false
targ_extra_big_endian=true
;;
+mips*-*-*)
+ targ_obj=mips
+ targ_machine=EM_MIPS
+ targ_size=32
+ targ_big_endian=true
+ targ_extra_big_endian=false
+ ;;
+mips*el*-*-*|mips*le*-*-*)
+ targ_obj=mips
+ targ_machine=EM_MIPS_RS3_LE
+ targ_size=32
+ targ_big_endian=false
+ targ_extra_big_endian=true
+ ;;
*)
targ_obj=UNKNOWN
;;
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #3: mips1.patch --]
[-- Type: text/x-patch; name="mips1.patch", Size: 187154 bytes --]
diff --git a/gold/mips.cc b/gold/mips.cc
new file mode 100644
index 0000000..b0fcc93
--- /dev/null
+++ b/gold/mips.cc
@@ -0,0 +1,5254 @@
+// mips.cc -- mips target support for gold.
+
+// Copyright (C) 2011-2014 Free Software Foundation, Inc.
+// Written by Sasa Stankovic <sasa.stankovic@imgtec.com>
+// and Aleksandar Simeonov <aleksandar.simeonov@rt-rk.com>.
+// This file contains borrowed and adapted code from bfd/elfxx-mips.c.
+
+// This file is part of gold.
+
+// This program is free software; you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation; either version 3 of the License, or
+// (at your option) any later version.
+
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU General Public License for more details.
+
+// You should have received a copy of the GNU General Public License
+// along with this program; if not, write to the Free Software
+// Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
+// MA 02110-1301, USA.
+
+#include "gold.h"
+
+#include <algorithm>
+#include <set>
+#include <sstream>
+#include "demangle.h"
+
+#include "elfcpp.h"
+#include "parameters.h"
+#include "reloc.h"
+#include "mips.h"
+#include "object.h"
+#include "symtab.h"
+#include "layout.h"
+#include "output.h"
+#include "copy-relocs.h"
+#include "target.h"
+#include "target-reloc.h"
+#include "target-select.h"
+#include "tls.h"
+#include "errors.h"
+#include "gc.h"
+
+namespace
+{
+using namespace gold;
+
+template<int size, bool big_endian>
+class Mips_output_data_plt;
+
+template<int size, bool big_endian>
+class Mips_output_data_got;
+
+template<int size, bool big_endian>
+class Target_mips;
+
+template<int size, bool big_endian>
+class Mips_output_section_reginfo;
+
+template<int size, bool big_endian>
+class Mips_output_data_la25_stub;
+
+template<int size, bool big_endian>
+class Mips_output_data_mips_stubs;
+
+template<int size>
+class Mips_symbol;
+
+template<int size, bool big_endian>
+class Mips_got_info;
+
+template<int size, bool big_endian>
+class Mips_relobj;
+
+class Mips16_stub_section_base;
+
+template<int size, bool big_endian>
+class Mips16_stub_section;
+
+// The ABI says that every symbol used by dynamic relocations must have
+// a global GOT entry. Among other things, this provides the dynamic
+// linker with a free, directly-indexed cache. The GOT can therefore
+// contain symbols that are not referenced by GOT relocations themselves
+// (in other words, it may have symbols that are not referenced by things
+// like R_MIPS_GOT16 and R_MIPS_GOT_PAGE).
+
+// GOT relocations are less likely to overflow if we put the associated
+// GOT entries towards the beginning. We therefore divide the global
+// GOT entries into two areas: "normal" and "reloc-only". Entries in
+// the first area can be used for both dynamic relocations and GP-relative
+// accesses, while those in the "reloc-only" area are for dynamic
+// relocations only.
+
+// These GGA_* ("Global GOT Area") values are organised so that lower
+// values are more general than higher values. Also, non-GGA_NONE
+// values are ordered by the position of the area in the GOT.
+
+enum Global_got_area
+{
+ GGA_NORMAL = 0,
+ GGA_RELOC_ONLY = 1,
+ GGA_NONE = 2
+};
+
+// The types of GOT entries needed for this platform.
+// These values are exposed to the ABI in an incremental link.
+// Do not renumber existing values without changing the version
+// number of the .gnu_incremental_inputs section.
+enum Got_type
+{
+ GOT_TYPE_STANDARD = 0, // GOT entry for a regular symbol
+ GOT_TYPE_TLS_OFFSET = 1, // GOT entry for TLS offset
+ GOT_TYPE_TLS_PAIR = 2, // GOT entry for TLS module/offset pair
+
+ // GOT entries for multi-GOT. We support up to 1024 GOTs in multi-GOT links.
+ GOT_TYPE_STANDARD_MULTIGOT = 3,
+ GOT_TYPE_TLS_OFFSET_MULTIGOT = GOT_TYPE_STANDARD_MULTIGOT + 1024,
+ GOT_TYPE_TLS_PAIR_MULTIGOT = GOT_TYPE_TLS_OFFSET_MULTIGOT + 1024
+};
+
+// TLS type of GOT entry.
+enum Got_tls_type
+{
+ GOT_TLS_NONE = 0,
+ GOT_TLS_GD = 1,
+ GOT_TLS_LDM = 2,
+ GOT_TLS_IE = 4
+};
+
+// Return TRUE if a relocation of type R_TYPE from OBJECT might
+// require an la25 stub. See also local_pic_function, which determines
+// whether the destination function ever requires a stub.
+template<int size, bool big_endian>
+static inline bool
+relocation_needs_la25_stub(Mips_relobj<size, big_endian>* object,
+ unsigned int r_type, bool target_is_16_bit_code)
+{
+ // We specifically ignore branches and jumps from EF_PIC objects,
+ // where the onus is on the compiler or programmer to perform any
+ // necessary initialization of $25. Sometimes such initialization
+ // is unnecessary; for example, -mno-shared functions do not use
+ // the incoming value of $25, and may therefore be called directly.
+ if (object->is_pic())
+ return false;
+
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MIPS_PC16:
+ case elfcpp::R_MICROMIPS_26_S1:
+ case elfcpp::R_MICROMIPS_PC7_S1:
+ case elfcpp::R_MICROMIPS_PC10_S1:
+ case elfcpp::R_MICROMIPS_PC16_S1:
+ case elfcpp::R_MICROMIPS_PC23_S2:
+ return true;
+
+ case elfcpp::R_MIPS16_26:
+ return !target_is_16_bit_code;
+
+ default:
+ return false;
+ }
+}
+
+// Return true if SYM is a locally-defined PIC function, in the sense
+// that it or its fn_stub might need $25 to be valid on entry.
+// Note that MIPS16 functions set up $gp using PC-relative instructions,
+// so they themselves never need $25 to be valid. Only non-MIPS16
+// entry points are of interest here.
+template<int size, bool big_endian>
+static inline bool
+local_pic_function(Mips_symbol<size>* sym)
+{
+ bool def_regular = (sym->source() == Symbol::FROM_OBJECT
+ && !sym->object()->is_dynamic()
+ && !sym->is_undefined());
+
+ if (sym->is_defined() && def_regular)
+ {
+ Mips_relobj<size, big_endian>* object =
+ static_cast<Mips_relobj<size, big_endian>*>(sym->object());
+
+ if ((object->is_pic() || sym->is_pic())
+ && (!sym->is_mips16()
+ || (sym->has_mips16_fn_stub() && sym->need_fn_stub())))
+ return true;
+ }
+ return false;
+}
+
+static inline bool
+hi16_reloc(int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_HI16
+ || r_type == elfcpp::R_MIPS16_HI16
+ || r_type == elfcpp::R_MICROMIPS_HI16);
+}
+
+static inline bool
+lo16_reloc(int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_LO16
+ || r_type == elfcpp::R_MIPS16_LO16
+ || r_type == elfcpp::R_MICROMIPS_LO16);
+}
+
+static inline bool
+got16_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_GOT16
+ || r_type == elfcpp::R_MIPS16_GOT16
+ || r_type == elfcpp::R_MICROMIPS_GOT16);
+}
+
+static inline bool
+call_lo16_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_CALL_LO16
+ || r_type == elfcpp::R_MICROMIPS_CALL_LO16);
+}
+
+static inline bool
+got_lo16_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_GOT_LO16
+ || r_type == elfcpp::R_MICROMIPS_GOT_LO16);
+}
+
+static inline bool
+got_disp_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_GOT_DISP
+ || r_type == elfcpp::R_MICROMIPS_GOT_DISP);
+}
+
+static inline bool
+got_page_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_GOT_PAGE
+ || r_type == elfcpp::R_MICROMIPS_GOT_PAGE);
+}
+
+static inline bool
+tls_gd_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_TLS_GD
+ || r_type == elfcpp::R_MIPS16_TLS_GD
+ || r_type == elfcpp::R_MICROMIPS_TLS_GD);
+}
+
+static inline bool
+tls_gottprel_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_TLS_GOTTPREL
+ || r_type == elfcpp::R_MIPS16_TLS_GOTTPREL
+ || r_type == elfcpp::R_MICROMIPS_TLS_GOTTPREL);
+}
+
+static inline bool
+tls_ldm_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_TLS_LDM
+ || r_type == elfcpp::R_MIPS16_TLS_LDM
+ || r_type == elfcpp::R_MICROMIPS_TLS_LDM);
+}
+
+static inline bool
+mips16_call_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS16_26
+ || r_type == elfcpp::R_MIPS16_CALL16);
+}
+
+static inline bool
+jal_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_26
+ || r_type == elfcpp::R_MIPS16_26
+ || r_type == elfcpp::R_MICROMIPS_26_S1);
+}
+
+static inline bool
+micromips_branch_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MICROMIPS_26_S1
+ || r_type == elfcpp::R_MICROMIPS_PC16_S1
+ || r_type == elfcpp::R_MICROMIPS_PC10_S1
+ || r_type == elfcpp::R_MICROMIPS_PC7_S1);
+}
+
+// Check if R_TYPE is a MIPS16 reloc.
+static inline bool
+mips16_reloc(unsigned int r_type)
+{
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS16_26:
+ case elfcpp::R_MIPS16_GPREL:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MIPS16_LO16:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MIPS16_TLS_DTPREL_HI16:
+ case elfcpp::R_MIPS16_TLS_DTPREL_LO16:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_TPREL_HI16:
+ case elfcpp::R_MIPS16_TLS_TPREL_LO16:
+ return true;
+
+ default:
+ return false;
+ }
+}
+
+// Check if R_TYPE is a microMIPS reloc.
+static inline bool
+micromips_reloc(unsigned int r_type)
+{
+ switch (r_type)
+ {
+ case elfcpp::R_MICROMIPS_26_S1:
+ case elfcpp::R_MICROMIPS_HI16:
+ case elfcpp::R_MICROMIPS_LO16:
+ case elfcpp::R_MICROMIPS_GPREL16:
+ case elfcpp::R_MICROMIPS_LITERAL:
+ case elfcpp::R_MICROMIPS_GOT16:
+ case elfcpp::R_MICROMIPS_PC7_S1:
+ case elfcpp::R_MICROMIPS_PC10_S1:
+ case elfcpp::R_MICROMIPS_PC16_S1:
+ case elfcpp::R_MICROMIPS_CALL16:
+ case elfcpp::R_MICROMIPS_GOT_DISP:
+ case elfcpp::R_MICROMIPS_GOT_PAGE:
+ case elfcpp::R_MICROMIPS_GOT_OFST:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_SUB:
+ case elfcpp::R_MICROMIPS_HIGHER:
+ case elfcpp::R_MICROMIPS_HIGHEST:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ case elfcpp::R_MICROMIPS_SCN_DISP:
+ case elfcpp::R_MICROMIPS_JALR:
+ case elfcpp::R_MICROMIPS_HI0_LO16:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ case elfcpp::R_MICROMIPS_TLS_DTPREL_HI16:
+ case elfcpp::R_MICROMIPS_TLS_DTPREL_LO16:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_TPREL_HI16:
+ case elfcpp::R_MICROMIPS_TLS_TPREL_LO16:
+ case elfcpp::R_MICROMIPS_GPREL7_S2:
+ case elfcpp::R_MICROMIPS_PC23_S2:
+ return true;
+
+ default:
+ return false;
+ }
+}
+
+static inline bool
+is_matching_lo16_reloc(unsigned int high_reloc, unsigned int lo16_reloc)
+{
+ switch (high_reloc)
+ {
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MIPS_GOT16:
+ return lo16_reloc == elfcpp::R_MIPS_LO16;
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MIPS16_GOT16:
+ return lo16_reloc == elfcpp::R_MIPS16_LO16;
+ case elfcpp::R_MICROMIPS_HI16:
+ case elfcpp::R_MICROMIPS_GOT16:
+ return lo16_reloc == elfcpp::R_MICROMIPS_LO16;
+ default:
+ return false;
+ }
+}
+
+// This class is used to hold information about one GOT entry.
+// There are three types of entry:
+//
+// (1) a SYMBOL + OFFSET address, where SYMBOL is local to an input object
+// (object != NULL, symndx >= 0, tls_type != GOT_TLS_LDM)
+// (2) a SYMBOL address, where SYMBOL is not local to an input object
+// (object != NULL, symndx == -1)
+// (3) a TLS LDM slot
+// (object != NULL, symndx == 0, tls_type == GOT_TLS_LDM)
+
+template<int size, bool big_endian>
+class Mips_got_entry
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+
+ public:
+ Mips_got_entry(Mips_relobj<size, big_endian>* object, unsigned int symndx,
+ Mips_address addend, unsigned char tls_type,
+ unsigned int shndx)
+ : object_(object), symndx_(symndx), tls_type_(tls_type), shndx_(shndx)
+ { this->d.addend = addend; }
+
+ Mips_got_entry(Mips_relobj<size, big_endian>* object, Mips_symbol<size>* sym,
+ unsigned char tls_type)
+ : object_(object), symndx_(-1U), tls_type_(tls_type), shndx_(-1U)
+ { this->d.sym = sym; }
+
+ // Return whether this entry is for a local symbol.
+ bool
+ is_for_local_symbol() const
+ { return this->symndx_ != -1U; }
+
+ // Return whether this entry is for a global symbol.
+ bool
+ is_for_global_symbol() const
+ { return this->symndx_ == -1U; }
+
+ // Return the hash of this entry.
+ size_t
+ hash() const
+ {
+ if (this->tls_type_ == GOT_TLS_LDM)
+ return this->symndx_ + (1 << 18);
+ if (this->symndx_ != -1U)
+ {
+ uintptr_t object_id = reinterpret_cast<uintptr_t>(this->object());
+ return this->symndx_ + object_id + this->d.addend;
+ }
+ else
+ {
+ uintptr_t sym_id = reinterpret_cast<uintptr_t>(this->d.sym);
+ return this->symndx_ + sym_id;
+ }
+ }
+
+ // Return whether this entry is equal to OTHER.
+ bool
+ equals(Mips_got_entry<size, big_endian>* other) const
+ {
+ if (this->symndx_ != other->symndx_
+ || this->tls_type_ != other->tls_type_)
+ return false;
+ if (this->tls_type_ == GOT_TLS_LDM)
+ return true;
+ if (this->symndx_ != -1U)
+ return (this->object() == other->object()
+ && this->d.addend == other->d.addend);
+ else
+ return this->d.sym == other->d.sym;
+ }
+
+ // Return input object that needs this GOT entry.
+ Mips_relobj<size, big_endian>*
+ object() const
+ {
+ gold_assert(this->object_ != NULL);
+ return this->object_;
+ }
+
+ // Return local symbol index for local GOT entries.
+ unsigned int
+ symndx() const
+ {
+ gold_assert(this->symndx_ != -1U);
+ return this->symndx_;
+ }
+
+ // Return the relocation addend for local GOT entries.
+ Mips_address
+ addend() const
+ {
+ gold_assert(this->symndx_ != -1U);
+ return this->d.addend;
+ }
+
+ // Return global symbol for global GOT entries.
+ Mips_symbol<size>*
+ sym() const
+ {
+ gold_assert(this->symndx_ == -1U);
+ return this->d.sym;
+ }
+
+ // Return whether this is a TLS GOT entry.
+ bool
+ is_tls_entry() const
+ { return this->tls_type_ != GOT_TLS_NONE; }
+
+ // Return TLS type of this GOT entry.
+ unsigned char
+ tls_type() const
+ { return this->tls_type_; }
+
+ // Return section index of the local symbol for local GOT entries.
+ unsigned int
+ shndx() const
+ { return this->shndx_; }
+
+ private:
+ // The input object that needs the GOT entry.
+ Mips_relobj<size, big_endian>* object_;
+ // The index of the symbol if we have a local symbol; -1 otherwise.
+ unsigned int symndx_;
+
+ union
+ {
+ // If symndx != -1, the addend of the relocation that should be added to the
+ // symbol value.
+ Mips_address addend;
+ // If symndx == -1, the global symbol corresponding to this GOT entry. The
+ // symbol's entry is in the local area if mips_sym->global_got_area is
+ // GGA_NONE, otherwise it is in the global area.
+ Mips_symbol<size>* sym;
+ } d;
+
+ // The TLS type of this GOT entry. An LDM GOT entry will be a local
+ // symbol entry with r_symndx == 0.
+ unsigned char tls_type_;
+
+ // For local GOT entries, section index of the local symbol.
+ unsigned int shndx_;
+};
+
+// Hash for Mips_got_entry.
+
+template<int size, bool big_endian>
+class Mips_got_entry_hash
+{
+ public:
+ size_t
+ operator()(Mips_got_entry<size, big_endian>* entry) const
+ { return entry->hash(); }
+};
+
+// Equality for Mips_got_entry.
+
+template<int size, bool big_endian>
+class Mips_got_entry_eq
+{
+ public:
+ bool
+ operator()(Mips_got_entry<size, big_endian>* e1,
+ Mips_got_entry<size, big_endian>* e2) const
+ { return e1->equals(e2); }
+};
+
+// Got_page_range. This class describes a range of addends: [MIN_ADDEND,
+// MAX_ADDEND]. The instances form a non-overlapping list that is sorted by
+// increasing MIN_ADDEND.
+
+struct Got_page_range
+{
+ Got_page_range()
+ : next(NULL), min_addend(0), max_addend(0)
+ { }
+
+ Got_page_range* next;
+ int min_addend;
+ int max_addend;
+
+ // Return the maximum number of GOT page entries required.
+ int
+ get_max_pages()
+ { return (this->max_addend - this->min_addend + 0x1ffff) >> 16; }
+};
+
+// Got_page_entry. This class describes the range of addends that are applied
+// to page relocations against a given symbol.
+
+struct Got_page_entry
+{
+ Got_page_entry()
+ : object(NULL), symndx(-1U), ranges(NULL), num_pages(0)
+ { }
+
+ Got_page_entry(Object* object_, unsigned int symndx_)
+ : object(object_), symndx(symndx_), ranges(NULL), num_pages(0)
+ { }
+
+ // The input object that needs the GOT page entry.
+ Object* object;
+ // The index of the symbol, as stored in the relocation r_info.
+ unsigned int symndx;
+ // The ranges for this page entry.
+ Got_page_range* ranges;
+ // The maximum number of page entries needed for RANGES.
+ unsigned int num_pages;
+};
+
+// Hash for Got_page_entry.
+
+struct Got_page_entry_hash
+{
+ size_t
+ operator()(Got_page_entry* entry) const
+ { return reinterpret_cast<uintptr_t>(entry->object) + entry->symndx; }
+};
+
+// Equality for Got_page_entry.
+
+struct Got_page_entry_eq
+{
+ bool
+ operator()(Got_page_entry* entry1, Got_page_entry* entry2) const
+ {
+ return entry1->object == entry2->object && entry1->symndx == entry2->symndx;
+ }
+};
+
+// This class is used to hold .got information when linking.
+
+template<int size, bool big_endian>
+class Mips_got_info
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+ typedef Output_data_reloc<elfcpp::SHT_REL, true, size, big_endian>
+ Reloc_section;
+ typedef Unordered_map<unsigned int, unsigned int> Got_page_offsets;
+
+ // Unordered set of GOT entries.
+ typedef Unordered_set<Mips_got_entry<size, big_endian>*,
+ Mips_got_entry_hash<size, big_endian>,
+ Mips_got_entry_eq<size, big_endian> > Got_entry_set;
+
+ // Unordered set of GOT page entries.
+ typedef Unordered_set<Got_page_entry*,
+ Got_page_entry_hash, Got_page_entry_eq> Got_page_entry_set;
+
+ public:
+ Mips_got_info()
+ : local_gotno_(0), page_gotno_(0), global_gotno_(0), reloc_only_gotno_(0),
+ tls_gotno_(0), tls_ldm_offset_(-1U), global_got_symbols_(),
+ got_entries_(), got_page_entries_(), got_page_offset_start_(0),
+ got_page_offset_next_(0), got_page_offsets_(), next_(NULL), index_(-1U),
+ offset_(0)
+ { }
+
+ // Reserve GOT entry for a GOT relocation of type R_TYPE against symbol
+ // SYMNDX + ADDEND, where SYMNDX is a local symbol in section SHNDX in OBJECT.
+ void
+ record_local_got_symbol(Mips_relobj<size, big_endian>* object,
+ unsigned int symndx, Mips_address addend,
+ unsigned int r_type, unsigned int shndx);
+
+ // Reserve GOT entry for a GOT relocation of type R_TYPE against MIPS_SYM,
+ // in OBJECT. FOR_CALL is true if the caller is only interested in
+ // using the GOT entry for calls. DYN_RELOC is true if R_TYPE is a dynamic
+ // relocation.
+ void
+ record_global_got_symbol(Mips_symbol<size>* mips_sym,
+ Mips_relobj<size, big_endian>* object,
+ unsigned int r_type, bool dyn_reloc, bool for_call);
+
+ // Add ENTRY to master GOT and to OBJECT's GOT.
+ void
+ record_got_entry(Mips_got_entry<size, big_endian>* entry,
+ Mips_relobj<size, big_endian>* object);
+
+ // Record that OBJECT has a page relocation against symbol SYMNDX and
+ // that ADDEND is the addend for that relocation.
+ void
+ record_got_page_entry(Mips_relobj<size, big_endian>* object,
+ unsigned int symndx, int addend);
+
+ // Create all entries that should be in the local part of the GOT.
+ void
+ add_local_entries(Target_mips<size, big_endian>* target, Layout* layout);
+
+ // Create GOT page entries.
+ void
+ add_page_entries(Target_mips<size, big_endian>* target, Layout* layout);
+
+ // Create global GOT entries, both GGA_NORMAL and GGA_RELOC_ONLY.
+ void
+ add_global_entries(Target_mips<size, big_endian>* target, Layout* layout,
+ unsigned int non_reloc_only_global_gotno);
+
+ // Create global GOT entries that should be in the GGA_RELOC_ONLY area.
+ void
+ add_reloc_only_entries(Mips_output_data_got<size, big_endian>* got);
+
+ // Create TLS GOT entries.
+ void
+ add_tls_entries(Target_mips<size, big_endian>* target, Layout* layout);
+
+ // Decide whether the symbol needs an entry in the global part of the primary
+ // GOT, setting global_got_area accordingly. Count the number of global
+ // symbols that are in the primary GOT only because they have dynamic
+ // relocations R_MIPS_REL32 against them (reloc_only_gotno).
+ void
+ count_got_symbols(Symbol_table* symtab);
+
+ // Return the offset of GOT page entry for VALUE.
+ unsigned int
+ get_got_page_offset(Mips_address value,
+ Mips_output_data_got<size, big_endian>* got);
+
+ // Count the number of GOT entries required.
+ void
+ count_got_entries();
+
+ // Count the number of GOT entries required by ENTRY. Accumulate the result.
+ void
+ count_got_entry(Mips_got_entry<size, big_endian>* entry);
+
+ // Add FROM's GOT entries.
+ void
+ add_got_entries(Mips_got_info<size, big_endian>* from);
+
+ // Add FROM's GOT page entries.
+ void
+ add_got_page_entries(Mips_got_info<size, big_endian>* from);
+
+ // Return GOT size.
+ unsigned int
+ got_size() const
+ { return ((2 + this->local_gotno_ + this->page_gotno_ + this->global_gotno_
+ + this->tls_gotno_) * size/8);
+ }
+
+ // Return the number of local GOT entries.
+ unsigned int
+ local_gotno() const
+ { return this->local_gotno_; }
+
+ // Return the maximum number of page GOT entries needed.
+ unsigned int
+ page_gotno() const
+ { return this->page_gotno_; }
+
+ // Return the number of global GOT entries.
+ unsigned int
+ global_gotno() const
+ { return this->global_gotno_; }
+
+ // Set the number of global GOT entries.
+ void
+ set_global_gotno(unsigned int global_gotno)
+ { this->global_gotno_ = global_gotno; }
+
+ // Return the number of GGA_RELOC_ONLY global GOT entries.
+ unsigned int
+ reloc_only_gotno() const
+ { return this->reloc_only_gotno_; }
+
+ // Return the number of TLS GOT entries.
+ unsigned int
+ tls_gotno() const
+ { return this->tls_gotno_; }
+
+ // Return the GOT type for this GOT. Used for multi-GOT links only.
+ unsigned int
+ multigot_got_type(unsigned int got_type) const
+ {
+ switch (got_type)
+ {
+ case GOT_TYPE_STANDARD:
+ return GOT_TYPE_STANDARD_MULTIGOT + this->index_;
+ case GOT_TYPE_TLS_OFFSET:
+ return GOT_TYPE_TLS_OFFSET_MULTIGOT + this->index_;
+ case GOT_TYPE_TLS_PAIR:
+ return GOT_TYPE_TLS_PAIR_MULTIGOT + this->index_;
+ default:
+ gold_unreachable();
+ }
+ }
+
+ // Remove lazy-binding stubs for global symbols in this GOT.
+ void
+ remove_lazy_stubs(Target_mips<size, big_endian>* target);
+
+ // Return offset of this GOT from the start of .got section.
+ unsigned int
+ offset() const
+ { return this->offset_; }
+
+ // Set offset of this GOT from the start of .got section.
+ void
+ set_offset(unsigned int offset)
+ { this->offset_ = offset; }
+
+ // Set index of this GOT in multi-GOT links.
+ void
+ set_index(unsigned int index)
+ { this->index_ = index; }
+
+ // Return next GOT in multi-GOT links.
+ Mips_got_info<size, big_endian>*
+ next() const
+ { return this->next_; }
+
+ // Set next GOT in multi-GOT links.
+ void
+ set_next(Mips_got_info<size, big_endian>* next)
+ { this->next_ = next; }
+
+ // Return the offset of TLS LDM entry for this GOT.
+ unsigned int
+ tls_ldm_offset() const
+ { return this->tls_ldm_offset_; }
+
+ // Set the offset of TLS LDM entry for this GOT.
+ void
+ set_tls_ldm_offset(unsigned int tls_ldm_offset)
+ { this->tls_ldm_offset_ = tls_ldm_offset; }
+
+ Unordered_set<Mips_symbol<size>*>&
+ global_got_symbols()
+ { return this->global_got_symbols_; }
+
+ // Return the GOT_TLS_* type required by relocation type R_TYPE.
+ static int
+ mips_elf_reloc_tls_type(unsigned int r_type)
+ {
+ if (tls_gd_reloc(r_type))
+ return GOT_TLS_GD;
+
+ if (tls_ldm_reloc(r_type))
+ return GOT_TLS_LDM;
+
+ if (tls_gottprel_reloc(r_type))
+ return GOT_TLS_IE;
+
+ return GOT_TLS_NONE;
+ }
+
+ // Return the number of GOT slots needed for GOT TLS type TYPE.
+ static int
+ mips_tls_got_entries(unsigned int type)
+ {
+ switch (type)
+ {
+ case GOT_TLS_GD:
+ case GOT_TLS_LDM:
+ return 2;
+
+ case GOT_TLS_IE:
+ return 1;
+
+ case GOT_TLS_NONE:
+ return 0;
+
+ default:
+ gold_unreachable();
+ }
+ }
+
+ private:
+ // The number of local GOT entries.
+ unsigned int local_gotno_;
+ // The maximum number of page GOT entries needed.
+ unsigned int page_gotno_;
+ // The number of global GOT entries.
+ unsigned int global_gotno_;
+ // The number of global GOT entries that are in the GGA_RELOC_ONLY area.
+ unsigned int reloc_only_gotno_;
+ // The number of TLS GOT entries.
+ unsigned int tls_gotno_;
+ // The offset of TLS LDM entry for this GOT.
+ unsigned int tls_ldm_offset_;
+ // All symbols that have global GOT entry.
+ Unordered_set<Mips_symbol<size>*> global_got_symbols_;
+ // A hash table holding GOT entries.
+ Got_entry_set got_entries_;
+ // A hash table of GOT page entries.
+ Got_page_entry_set got_page_entries_;
+ // The offset of first GOT page entry for this GOT.
+ unsigned int got_page_offset_start_;
+ // The offset of next available GOT page entry for this GOT.
+ unsigned int got_page_offset_next_;
+ // A hash table that maps GOT page entry value to the GOT offset where
+ // the entry is located.
+ Got_page_offsets got_page_offsets_;
+ // In multi-GOT links, a pointer to the next GOT.
+ Mips_got_info<size, big_endian>* next_;
+ // Index of this GOT in multi-GOT links.
+ unsigned int index_;
+ // The offset of this GOT in multi-GOT links.
+ unsigned int offset_;
+};
+
+// This is a helper class used during relocation scan. It records GOT16 addend.
+
+template<int size, bool big_endian>
+struct got16_addend
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+
+ got16_addend(const Sized_relobj_file<size, big_endian>* _object,
+ unsigned int _shndx, unsigned int _r_type, unsigned int _r_sym,
+ Mips_address _addend)
+ : object(_object), shndx(_shndx), r_type(_r_type), r_sym(_r_sym),
+ addend(_addend)
+ { }
+
+ const Sized_relobj_file<size, big_endian>* object;
+ unsigned int shndx;
+ unsigned int r_type;
+ unsigned int r_sym;
+ Mips_address addend;
+};
+
+// Mips_symbol class. Holds additional symbol information needed for Mips.
+
+template<int size>
+class Mips_symbol : public Sized_symbol<size>
+{
+ public:
+ Mips_symbol()
+ : need_fn_stub_(false), has_nonpic_branches_(false), la25_stub_offset_(-1U),
+ has_static_relocs_(false), no_lazy_stub_(false), lazy_stub_offset_(0),
+ pointer_equality_needed_(false), global_got_area_(GGA_NONE),
+ global_gotoffset_(-1U), got_only_for_calls_(true), has_lazy_stub_(false),
+ needs_mips_plt_(false), needs_comp_plt_(false), mips_plt_offset_(-1U),
+ comp_plt_offset_(-1U), mips16_fn_stub_(NULL), mips16_call_stub_(NULL),
+ mips16_call_fp_stub_(NULL), applied_secondary_got_fixup_(false)
+ { }
+
+ // Return whether this is a MIPS16 symbol.
+ bool
+ is_mips16() const
+ {
+ // (st_other & STO_MIPS16) == STO_MIPS16
+ return ((this->nonvis() & (elfcpp::STO_MIPS16 >> 2))
+ == elfcpp::STO_MIPS16 >> 2);
+ }
+
+ // Return whether this is a microMIPS symbol.
+ bool
+ is_micromips() const
+ {
+ // (st_other & STO_MIPS_ISA) == STO_MICROMIPS
+ return ((this->nonvis() & (elfcpp::STO_MIPS_ISA >> 2))
+ == elfcpp::STO_MICROMIPS >> 2);
+ }
+
+ // Return whether the symbol needs MIPS16 fn_stub.
+ bool
+ need_fn_stub() const
+ { return this->need_fn_stub_; }
+
+ // Set that the symbol needs MIPS16 fn_stub.
+ void
+ set_need_fn_stub()
+ { this->need_fn_stub_ = true; }
+
+ // Return whether this symbol is referenced by branch relocations from
+ // any non-PIC input file.
+ bool
+ has_nonpic_branches() const
+ { return this->has_nonpic_branches_; }
+
+ // Set that this symbol is referenced by branch relocations from
+ // any non-PIC input file.
+ void
+ set_has_nonpic_branches()
+ { this->has_nonpic_branches_ = true; }
+
+ // Return the offset of the la25 stub for this symbol from the start of the
+ // la25 stub section.
+ unsigned int
+ la25_stub_offset() const
+ { return this->la25_stub_offset_; }
+
+ // Set the offset of the la25 stub for this symbol from the start of the
+ // la25 stub section.
+ void
+ set_la25_stub_offset(unsigned int offset)
+ { this->la25_stub_offset_ = offset; }
+
+ // Return whether the symbol has la25 stub. This is true if this symbol is
+ // for a PIC function, and there are non-PIC branches and jumps to it.
+ bool
+ has_la25_stub() const
+ { return this->la25_stub_offset_ != -1U; }
+
+ // Return whether there is a relocation against this symbol that must be
+ // resolved by the static linker (that is, the relocation cannot possibly
+ // be made dynamic).
+ bool
+ has_static_relocs() const
+ { return this->has_static_relocs_; }
+
+ // Set that there is a relocation against this symbol that must be resolved
+ // by the static linker (that is, the relocation cannot possibly be made
+ // dynamic).
+ void
+ set_has_static_relocs()
+ { this->has_static_relocs_ = true; }
+
+ // Return whether we must not create a lazy-binding stub for this symbol.
+ bool
+ no_lazy_stub() const
+ { return this->no_lazy_stub_; }
+
+ // Set that we must not create a lazy-binding stub for this symbol.
+ void
+ set_no_lazy_stub()
+ { this->no_lazy_stub_ = true; }
+
+ // Return the offset of the lazy-binding stub for this symbol from the start
+ // of .MIPS.stubs section.
+ unsigned int
+ lazy_stub_offset() const
+ { return this->lazy_stub_offset_; }
+
+ // Set the offset of the lazy-binding stub for this symbol from the start
+ // of .MIPS.stubs section.
+ void
+ set_lazy_stub_offset(unsigned int offset)
+ { this->lazy_stub_offset_ = offset; }
+
+ // Return whether there are any relocations for this symbol where
+ // pointer equality matters.
+ bool
+ pointer_equality_needed() const
+ { return this->pointer_equality_needed_; }
+
+ // Set that there are relocations for this symbol where pointer equality
+ // matters.
+ void
+ set_pointer_equality_needed()
+ { this->pointer_equality_needed_ = true; }
+
+ // Return global GOT area where this symbol in located.
+ Global_got_area
+ global_got_area() const
+ { return this->global_got_area_; }
+
+ // Set global GOT area where this symbol in located.
+ void
+ set_global_got_area(Global_got_area global_got_area)
+ { this->global_got_area_ = global_got_area; }
+
+ // Return the global GOT offset for this symbol. For multi-GOT links, this
+ // returns the offset from the start of .got section to the first GOT entry
+ // for the symbol. Note that in multi-GOT links the symbol can have entry
+ // in more than one GOT.
+ unsigned int
+ global_gotoffset() const
+ { return this->global_gotoffset_; }
+
+ // Set the global GOT offset for this symbol. Note that in multi-GOT links
+ // the symbol can have entry in more than one GOT. This method will set
+ // the offset only if it is less than current offset.
+ void
+ set_global_gotoffset(unsigned int offset)
+ {
+ if (this->global_gotoffset_ == -1U || offset < this->global_gotoffset_)
+ this->global_gotoffset_ = offset;
+ }
+
+ // Return whether all GOT relocations for this symbol are for calls.
+ bool
+ got_only_for_calls() const
+ { return this->got_only_for_calls_; }
+
+ // Set that there is a GOT relocation for this symbol that is not for call.
+ void
+ set_got_not_only_for_calls()
+ { this->got_only_for_calls_ = false; }
+
+ // Return whether this is a PIC symbol.
+ bool
+ is_pic() const
+ {
+ // (st_other & STO_MIPS_FLAGS) == STO_MIPS_PIC
+ return ((this->nonvis() & (elfcpp::STO_MIPS_FLAGS >> 2))
+ == (elfcpp::STO_MIPS_PIC >> 2));
+ }
+
+ // Set the flag in st_other field that marks this symbol as PIC.
+ void
+ set_pic()
+ {
+ if (this->is_mips16())
+ // (st_other & ~(STO_MIPS16 | STO_MIPS_FLAGS)) | STO_MIPS_PIC
+ this->set_nonvis((this->nonvis()
+ & ~((elfcpp::STO_MIPS16 >> 2)
+ | (elfcpp::STO_MIPS_FLAGS >> 2)))
+ | (elfcpp::STO_MIPS_PIC >> 2));
+ else
+ // (other & ~STO_MIPS_FLAGS) | STO_MIPS_PIC
+ this->set_nonvis((this->nonvis() & ~(elfcpp::STO_MIPS_FLAGS >> 2))
+ | (elfcpp::STO_MIPS_PIC >> 2));
+ }
+
+ // Set the flag in st_other field that marks this symbol as PLT.
+ void
+ set_mips_plt()
+ {
+ if (this->is_mips16())
+ // (st_other & (STO_MIPS16 | ~STO_MIPS_FLAGS)) | STO_MIPS_PLT
+ this->set_nonvis((this->nonvis()
+ & ((elfcpp::STO_MIPS16 >> 2)
+ | ~(elfcpp::STO_MIPS_FLAGS >> 2)))
+ | (elfcpp::STO_MIPS_PLT >> 2));
+
+ else
+ // (st_other & ~STO_MIPS_FLAGS) | STO_MIPS_PLT
+ this->set_nonvis((this->nonvis() & ~(elfcpp::STO_MIPS_FLAGS >> 2))
+ | (elfcpp::STO_MIPS_PLT >> 2));
+ }
+
+ // Downcast a base pointer to a Mips_symbol pointer.
+ static Mips_symbol<size>*
+ as_mips_sym(Symbol* sym)
+ { return static_cast<Mips_symbol<size>*>(sym); }
+
+ // Downcast a base pointer to a Mips_symbol pointer.
+ static const Mips_symbol<size>*
+ as_mips_sym(const Symbol* sym)
+ { return static_cast<const Mips_symbol<size>*>(sym); }
+
+ // Return whether the symbol has lazy-binding stub.
+ bool
+ has_lazy_stub() const
+ { return this->has_lazy_stub_; }
+
+ // Set whether the symbol has lazy-binding stub.
+ void
+ set_has_lazy_stub(bool has_lazy_stub)
+ { this->has_lazy_stub_ = has_lazy_stub; }
+
+ // Return whether the symbol needs a standard PLT entry.
+ bool
+ needs_mips_plt() const
+ { return this->needs_mips_plt_; }
+
+ // Set whether the symbol needs a standard PLT entry.
+ void
+ set_needs_mips_plt(bool needs_mips_plt)
+ { this->needs_mips_plt_ = needs_mips_plt; }
+
+ // Return whether the symbol needs a compressed (MIPS16 or microMIPS) PLT
+ // entry.
+ bool
+ needs_comp_plt() const
+ { return this->needs_comp_plt_; }
+
+ // Set whether the symbol needs a compressed (MIPS16 or microMIPS) PLT entry.
+ void
+ set_needs_comp_plt(bool needs_comp_plt)
+ { this->needs_comp_plt_ = needs_comp_plt; }
+
+ // Return standard PLT entry offset, or -1 if none.
+ unsigned int
+ mips_plt_offset() const
+ { return this->mips_plt_offset_; }
+
+ // Set standard PLT entry offset.
+ void
+ set_mips_plt_offset(unsigned int mips_plt_offset)
+ { this->mips_plt_offset_ = mips_plt_offset; }
+
+ // Return whether the symbol has standard PLT entry.
+ bool
+ has_mips_plt_offset() const
+ { return this->mips_plt_offset_ != -1U; }
+
+ // Return compressed (MIPS16 or microMIPS) PLT entry offset, or -1 if none.
+ unsigned int
+ comp_plt_offset() const
+ { return this->comp_plt_offset_; }
+
+ // Set compressed (MIPS16 or microMIPS) PLT entry offset.
+ void
+ set_comp_plt_offset(unsigned int comp_plt_offset)
+ { this->comp_plt_offset_ = comp_plt_offset; }
+
+ // Return whether the symbol has compressed (MIPS16 or microMIPS) PLT entry.
+ bool
+ has_comp_plt_offset() const
+ { return this->comp_plt_offset_ != -1U; }
+
+ // Return MIPS16 fn stub for a symbol.
+ template<bool big_endian>
+ Mips16_stub_section<size, big_endian>*
+ get_mips16_fn_stub() const
+ {
+ return static_cast<Mips16_stub_section<size, big_endian>*>(mips16_fn_stub_);
+ }
+
+ // Set MIPS16 fn stub for a symbol.
+ void
+ set_mips16_fn_stub(Mips16_stub_section_base* stub)
+ { this->mips16_fn_stub_ = stub; }
+
+ // Return whether symbol has MIPS16 fn stub.
+ bool
+ has_mips16_fn_stub() const
+ { return this->mips16_fn_stub_ != NULL; }
+
+ // Return MIPS16 call stub for a symbol.
+ template<bool big_endian>
+ Mips16_stub_section<size, big_endian>*
+ get_mips16_call_stub() const
+ {
+ return static_cast<Mips16_stub_section<size, big_endian>*>(
+ mips16_call_stub_);
+ }
+
+ // Set MIPS16 call stub for a symbol.
+ void
+ set_mips16_call_stub(Mips16_stub_section_base* stub)
+ { this->mips16_call_stub_ = stub; }
+
+ // Return whether symbol has MIPS16 call stub.
+ bool
+ has_mips16_call_stub() const
+ { return this->mips16_call_stub_ != NULL; }
+
+ // Return MIPS16 call_fp stub for a symbol.
+ template<bool big_endian>
+ Mips16_stub_section<size, big_endian>*
+ get_mips16_call_fp_stub() const
+ {
+ return static_cast<Mips16_stub_section<size, big_endian>*>(
+ mips16_call_fp_stub_);
+ }
+
+ // Set MIPS16 call_fp stub for a symbol.
+ void
+ set_mips16_call_fp_stub(Mips16_stub_section_base* stub)
+ { this->mips16_call_fp_stub_ = stub; }
+
+ // Return whether symbol has MIPS16 call_fp stub.
+ bool
+ has_mips16_call_fp_stub() const
+ { return this->mips16_call_fp_stub_ != NULL; }
+
+ bool
+ get_applied_secondary_got_fixup() const
+ { return applied_secondary_got_fixup_; }
+
+ void
+ set_applied_secondary_got_fixup()
+ { this->applied_secondary_got_fixup_ = true; }
+
+ private:
+ // Whether the symbol needs MIPS16 fn_stub. This is true if this symbol
+ // appears in any relocs other than a 16 bit call.
+ bool need_fn_stub_;
+
+ // True if this symbol is referenced by branch relocations from
+ // any non-PIC input file. This is used to determine whether an
+ // la25 stub is required.
+ bool has_nonpic_branches_;
+
+ // The offset of the la25 stub for this symbol from the start of the
+ // la25 stub section.
+ unsigned int la25_stub_offset_;
+
+ // True if there is a relocation against this symbol that must be
+ // resolved by the static linker (that is, the relocation cannot
+ // possibly be made dynamic).
+ bool has_static_relocs_;
+
+ // Whether we must not create a lazy-binding stub for this symbol.
+ // This is true if the symbol has relocations related to taking the
+ // function's address.
+ bool no_lazy_stub_;
+
+ // The offset of the lazy-binding stub for this symbol from the start of
+ // .MIPS.stubs section.
+ unsigned int lazy_stub_offset_;
+
+ // True if there are any relocations for this symbol where pointer equality
+ // matters.
+ bool pointer_equality_needed_;
+
+ // Global GOT area where this symbol in located, or GGA_NONE if symbol is not
+ // in the global part of the GOT.
+ Global_got_area global_got_area_;
+
+ // The global GOT offset for this symbol. For multi-GOT links, this is offset
+ // from the start of .got section to the first GOT entry for the symbol.
+ // Note that in multi-GOT links the symbol can have entry in more than one GOT.
+ unsigned int global_gotoffset_;
+
+ // Whether all GOT relocations for this symbol are for calls.
+ bool got_only_for_calls_;
+ // Whether the symbol has lazy-binding stub.
+ bool has_lazy_stub_;
+ // Whether the symbol needs a standard PLT entry.
+ bool needs_mips_plt_;
+ // Whether the symbol needs a compressed (MIPS16 or microMIPS) PLT entry.
+ bool needs_comp_plt_;
+ // Standard PLT entry offset, or -1 if none.
+ unsigned int mips_plt_offset_;
+ // Compressed (MIPS16 or microMIPS) PLT entry offset, or -1 if none.
+ unsigned int comp_plt_offset_;
+ // MIPS16 fn stub for a symbol.
+ Mips16_stub_section_base* mips16_fn_stub_;
+ // MIPS16 call stub for a symbol.
+ Mips16_stub_section_base* mips16_call_stub_;
+ // MIPS16 call_fp stub for a symbol.
+ Mips16_stub_section_base* mips16_call_fp_stub_;
+
+ bool applied_secondary_got_fixup_;
+};
+
+// Mips16_stub_section class.
+
+// The mips16 compiler uses a couple of special sections to handle
+// floating point arguments.
+
+// Section names that look like .mips16.fn.FNNAME contain stubs that
+// copy floating point arguments from the fp regs to the gp regs and
+// then jump to FNNAME. If any 32 bit function calls FNNAME, the
+// call should be redirected to the stub instead. If no 32 bit
+// function calls FNNAME, the stub should be discarded. We need to
+// consider any reference to the function, not just a call, because
+// if the address of the function is taken we will need the stub,
+// since the address might be passed to a 32 bit function.
+
+// Section names that look like .mips16.call.FNNAME contain stubs
+// that copy floating point arguments from the gp regs to the fp
+// regs and then jump to FNNAME. If FNNAME is a 32 bit function,
+// then any 16 bit function that calls FNNAME should be redirected
+// to the stub instead. If FNNAME is not a 32 bit function, the
+// stub should be discarded.
+
+// .mips16.call.fp.FNNAME sections are similar, but contain stubs
+// which call FNNAME and then copy the return value from the fp regs
+// to the gp regs. These stubs store the return address in $18 while
+// calling FNNAME; any function which might call one of these stubs
+// must arrange to save $18 around the call. (This case is not
+// needed for 32 bit functions that call 16 bit functions, because
+// 16 bit functions always return floating point values in both
+// $f0/$f1 and $2/$3.)
+
+// Note that in all cases FNNAME might be defined statically.
+// Therefore, FNNAME is not used literally. Instead, the relocation
+// information will indicate which symbol the section is for.
+
+// We record any stubs that we find in the symbol table.
+
+// TODO(sasa): All mips16 stub sections should be emitted in the .text section.
+
+class Mips16_stub_section_base { };
+
+template<int size, bool big_endian>
+class Mips16_stub_section : public Mips16_stub_section_base
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+
+ public:
+ Mips16_stub_section(Mips_relobj<size, big_endian>* object, unsigned int shndx)
+ : object_(object), shndx_(shndx), r_sym_(0), gsym_(NULL),
+ found_r_mips_none_(false)
+ {
+ gold_assert(object->is_mips16_fn_stub_section(shndx)
+ || object->is_mips16_call_stub_section(shndx)
+ || object->is_mips16_call_fp_stub_section(shndx));
+ }
+
+ // Return the object of this stub section.
+ Mips_relobj<size, big_endian>*
+ object() const
+ { return this->object_; }
+
+ // Return the size of a section.
+ uint64_t
+ section_size() const
+ { return this->object_->section_size(this->shndx_); }
+
+ // Return section index of this stub section.
+ unsigned int
+ shndx() const
+ { return this->shndx_; }
+
+ // Return symbol index, if stub is for a local function.
+ unsigned int
+ r_sym() const
+ { return this->r_sym_; }
+
+ // Return symbol, if stub is for a global function.
+ Mips_symbol<size>*
+ gsym() const
+ { return this->gsym_; }
+
+ // Return whether stub is for a local function.
+ bool
+ is_for_local_function() const
+ { return this->gsym_ == NULL; }
+
+ // This method is called when a new relocation R_TYPE for local symbol R_SYM
+ // is found in the stub section. Try to find stub target.
+ void
+ new_local_reloc_found(unsigned int r_type, unsigned int r_sym)
+ {
+ // To find target symbol for this stub, trust the first R_MIPS_NONE
+ // relocation, if any. Otherwise trust the first relocation, whatever
+ // its kind.
+ if (this->found_r_mips_none_)
+ return;
+ if (r_type == elfcpp::R_MIPS_NONE)
+ {
+ this->r_sym_ = r_sym;
+ this->gsym_ = NULL;
+ this->found_r_mips_none_ = true;
+ }
+ else if (!is_target_found())
+ this->r_sym_ = r_sym;
+ }
+
+ // This method is called when a new relocation R_TYPE for global symbol GSYM
+ // is found in the stub section. Try to find stub target.
+ void
+ new_global_reloc_found(unsigned int r_type, Mips_symbol<size>* gsym)
+ {
+ // To find target symbol for this stub, trust the first R_MIPS_NONE
+ // relocation, if any. Otherwise trust the first relocation, whatever
+ // its kind.
+ if (this->found_r_mips_none_)
+ return;
+ if (r_type == elfcpp::R_MIPS_NONE)
+ {
+ this->gsym_ = gsym;
+ this->r_sym_ = 0;
+ this->found_r_mips_none_ = true;
+ }
+ else if (!is_target_found())
+ this->gsym_ = gsym;
+ }
+
+ // Return whether we found the stub target.
+ bool
+ is_target_found() const
+ { return this->r_sym_ != 0 || this->gsym_ != NULL; }
+
+ // Return whether this is a fn stub.
+ bool
+ is_fn_stub() const
+ { return this->object_->is_mips16_fn_stub_section(this->shndx_); }
+
+ // Return whether this is a call stub.
+ bool
+ is_call_stub() const
+ { return this->object_->is_mips16_call_stub_section(this->shndx_); }
+
+ // Return whether this is a call_fp stub.
+ bool
+ is_call_fp_stub() const
+ { return this->object_->is_mips16_call_fp_stub_section(this->shndx_); }
+
+ // Return the output address.
+ Mips_address
+ output_address() const
+ {
+ return (this->object_->output_section(this->shndx_)->address()
+ + this->object_->output_section_offset(this->shndx_));
+ }
+
+ private:
+ // The object of this stub section.
+ Mips_relobj<size, big_endian>* object_;
+ // The section index of this stub section.
+ unsigned int shndx_;
+ // The symbol index, if stub is for a local function.
+ unsigned int r_sym_;
+ // The symbol, if stub is for a global function.
+ Mips_symbol<size>* gsym_;
+ // True if we found R_MIPS_NONE relocation in this stub.
+ bool found_r_mips_none_;
+};
+
+// Mips_relobj class.
+
+template<int size, bool big_endian>
+class Mips_relobj : public Sized_relobj_file<size, big_endian>
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+ typedef std::map<unsigned int, Mips16_stub_section<size, big_endian>*>
+ Mips16_stubs_int_map;
+ typedef typename elfcpp::Swap<size, big_endian>::Valtype Valtype;
+
+ public:
+ Mips_relobj(const std::string& name, Input_file* input_file, off_t offset,
+ const typename elfcpp::Ehdr<size, big_endian>& ehdr)
+ : Sized_relobj_file<size, big_endian>(name, input_file, offset, ehdr),
+ processor_specific_flags_(0), local_symbol_is_mips16_(),
+ local_symbol_is_micromips_(), mips16_stub_sections_(),
+ local_non_16bit_calls_(), local_16bit_calls_(), local_mips16_fn_stubs_(),
+ local_mips16_call_stubs_(), gp_(0), got_info_(NULL),
+ section_is_mips16_fn_stub_(), section_is_mips16_call_stub_(),
+ section_is_mips16_call_fp_stub_(), pdr_shndx_(-1U), gprmask_(0),
+ cprmask1_(0), cprmask2_(0), cprmask3_(0), cprmask4_(0)
+ {
+ this->is_pic_ = (ehdr.get_e_flags() & elfcpp::EF_MIPS_PIC) != 0;
+ this->is_n32_ = elfcpp::abi_n32(ehdr.get_e_flags());
+ this->is_n64_ = elfcpp::abi_64(ehdr.get_e_ident()[elfcpp::EI_CLASS]);
+ }
+
+ ~Mips_relobj()
+ { }
+
+ // Downcast a base pointer to a Mips_relobj pointer. This is
+ // not type-safe but we only use Mips_relobj not the base class.
+ static Mips_relobj<size, big_endian>*
+ as_mips_relobj(Relobj* relobj)
+ { return static_cast<Mips_relobj<size, big_endian>*>(relobj); }
+
+ // Downcast a base pointer to a Mips_relobj pointer. This is
+ // not type-safe but we only use Mips_relobj not the base class.
+ static const Mips_relobj<size, big_endian>*
+ as_mips_relobj(const Relobj* relobj)
+ { return static_cast<const Mips_relobj<size, big_endian>*>(relobj); }
+
+ // Processor-specific flags in ELF file header. This is valid only after
+ // reading symbols.
+ elfcpp::Elf_Word
+ processor_specific_flags() const
+ { return this->processor_specific_flags_; }
+
+ // Whether a local symbol is MIPS16 symbol. R_SYM is the symbol table
+ // index. This is only valid after do_count_local_symbol is called.
+ bool
+ local_symbol_is_mips16(unsigned int r_sym) const
+ {
+ gold_assert(r_sym < this->local_symbol_is_mips16_.size());
+ return this->local_symbol_is_mips16_[r_sym];
+ }
+
+ // Whether a local symbol is microMIPS symbol. R_SYM is the symbol table
+ // index. This is only valid after do_count_local_symbol is called.
+ bool
+ local_symbol_is_micromips(unsigned int r_sym) const
+ {
+ gold_assert(r_sym < this->local_symbol_is_micromips_.size());
+ return this->local_symbol_is_micromips_[r_sym];
+ }
+
+ // Get or create MIPS16 stub section.
+ Mips16_stub_section<size, big_endian>*
+ get_mips16_stub_section(unsigned int shndx)
+ {
+ typename Mips16_stubs_int_map::const_iterator it =
+ this->mips16_stub_sections_.find(shndx);
+ if (it != this->mips16_stub_sections_.end())
+ return (*it).second;
+
+ Mips16_stub_section<size, big_endian>* stub_section =
+ new Mips16_stub_section<size, big_endian>(this, shndx);
+ this->mips16_stub_sections_.insert(
+ std::pair<unsigned int, Mips16_stub_section<size, big_endian>*>(
+ stub_section->shndx(), stub_section));
+ return stub_section;
+ }
+
+ // Return MIPS16 fn stub section for local symbol R_SYM, or NULL if this
+ // object doesn't have fn stub for R_SYM.
+ Mips16_stub_section<size, big_endian>*
+ get_local_mips16_fn_stub(unsigned int r_sym) const
+ {
+ typename Mips16_stubs_int_map::const_iterator it =
+ this->local_mips16_fn_stubs_.find(r_sym);
+ if (it != this->local_mips16_fn_stubs_.end())
+ return (*it).second;
+ return NULL;
+ }
+
+ // Record that this object has MIPS16 fn stub for local symbol. This method
+ // is only called if we decided not to discard the stub.
+ void
+ add_local_mips16_fn_stub(Mips16_stub_section<size, big_endian>* stub)
+ {
+ gold_assert(stub->is_for_local_function());
+ unsigned int r_sym = stub->r_sym();
+ this->local_mips16_fn_stubs_.insert(
+ std::pair<unsigned int, Mips16_stub_section<size, big_endian>*>(
+ r_sym, stub));
+ }
+
+ // Return MIPS16 call stub section for local symbol R_SYM, or NULL if this
+ // object doesn't have call stub for R_SYM.
+ Mips16_stub_section<size, big_endian>*
+ get_local_mips16_call_stub(unsigned int r_sym) const
+ {
+ typename Mips16_stubs_int_map::const_iterator it =
+ this->local_mips16_call_stubs_.find(r_sym);
+ if (it != this->local_mips16_call_stubs_.end())
+ return (*it).second;
+ return NULL;
+ }
+
+ // Record that this object has MIPS16 call stub for local symbol. This method
+ // is only called if we decided not to discard the stub.
+ void
+ add_local_mips16_call_stub(Mips16_stub_section<size, big_endian>* stub)
+ {
+ gold_assert(stub->is_for_local_function());
+ unsigned int r_sym = stub->r_sym();
+ this->local_mips16_call_stubs_.insert(
+ std::pair<unsigned int, Mips16_stub_section<size, big_endian>*>(
+ r_sym, stub));
+ }
+
+ // Record that we found "non 16-bit" call relocation against local symbol
+ // SYMNDX. This reloc would need to refer to a MIPS16 fn stub, if there
+ // is one.
+ void
+ add_local_non_16bit_call(unsigned int symndx)
+ { this->local_non_16bit_calls_.insert(symndx); }
+
+ // Return true if there is any "non 16-bit" call relocation against local
+ // symbol SYMNDX in this object.
+ bool
+ has_local_non_16bit_call_relocs(unsigned int symndx)
+ {
+ return (this->local_non_16bit_calls_.find(symndx)
+ != this->local_non_16bit_calls_.end());
+ }
+
+ // Record that we found 16-bit call relocation R_MIPS16_26 against local
+ // symbol SYMNDX. Local MIPS16 call or call_fp stubs will only be needed
+ // if there is some R_MIPS16_26 relocation that refers to the stub symbol.
+ void
+ add_local_16bit_call(unsigned int symndx)
+ { this->local_16bit_calls_.insert(symndx); }
+
+ // Return true if there is any 16-bit call relocation R_MIPS16_26 against local
+ // symbol SYMNDX in this object.
+ bool
+ has_local_16bit_call_relocs(unsigned int symndx)
+ {
+ return (this->local_16bit_calls_.find(symndx)
+ != this->local_16bit_calls_.end());
+ }
+
+ // Get gp value that was used to create this object.
+ Mips_address
+ gp_value() const
+ { return this->gp_; }
+
+ // Return whether the object is a PIC object.
+ bool
+ is_pic() const
+ { return this->is_pic_; }
+
+ // Return whether the object uses N32 ABI.
+ bool
+ is_n32() const
+ { return this->is_n32_; }
+
+ // Return whether the object uses N64 ABI.
+ bool
+ is_n64() const
+ { return this->is_n64_; }
+
+ // Return whether the object uses NewABI conventions.
+ bool
+ is_newabi() const
+ { return this->is_n32_ || this->is_n64_; }
+
+ // Return Mips_got_info for this object.
+ Mips_got_info<size, big_endian>*
+ get_got_info() const
+ { return this->got_info_; }
+
+ // Return Mips_got_info for this object. Create new info if it doesn't exist.
+ Mips_got_info<size, big_endian>*
+ get_or_create_got_info()
+ {
+ if (!this->got_info_)
+ this->got_info_ = new Mips_got_info<size, big_endian>();
+ return this->got_info_;
+ }
+
+ // Set Mips_got_info for this object.
+ void
+ set_got_info(Mips_got_info<size, big_endian>* got_info)
+ { this->got_info_ = got_info; }
+
+ // Whether a section SHDNX is a MIPS16 stub section. This is only valid
+ // after do_read_symbols is called.
+ bool
+ is_mips16_stub_section(unsigned int shndx)
+ {
+ return (is_mips16_fn_stub_section(shndx)
+ || is_mips16_call_stub_section(shndx)
+ || is_mips16_call_fp_stub_section(shndx));
+ }
+
+ // Return TRUE if relocations in section SHNDX can refer directly to a
+ // MIPS16 function rather than to a hard-float stub. This is only valid
+ // after do_read_symbols is called.
+ bool
+ section_allows_mips16_refs(unsigned int shndx)
+ {
+ return (this->is_mips16_stub_section(shndx) || shndx == this->pdr_shndx_);
+ }
+
+ // Whether a section SHDNX is a MIPS16 fn stub section. This is only valid
+ // after do_read_symbols is called.
+ bool
+ is_mips16_fn_stub_section(unsigned int shndx)
+ {
+ gold_assert(shndx < this->section_is_mips16_fn_stub_.size());
+ return this->section_is_mips16_fn_stub_[shndx];
+ }
+
+ // Whether a section SHDNX is a MIPS16 call stub section. This is only valid
+ // after do_read_symbols is called.
+ bool
+ is_mips16_call_stub_section(unsigned int shndx)
+ {
+ gold_assert(shndx < this->section_is_mips16_call_stub_.size());
+ return this->section_is_mips16_call_stub_[shndx];
+ }
+
+ // Whether a section SHDNX is a MIPS16 call_fp stub section. This is only
+ // valid after do_read_symbols is called.
+ bool
+ is_mips16_call_fp_stub_section(unsigned int shndx)
+ {
+ gold_assert(shndx < this->section_is_mips16_call_fp_stub_.size());
+ return this->section_is_mips16_call_fp_stub_[shndx];
+ }
+
+ // Discard MIPS16 stub secions that are not needed.
+ void
+ discard_mips16_stub_sections(Symbol_table* symtab);
+
+ // Return gprmask from the .reginfo section of this object.
+ Valtype
+ gprmask() const
+ { return this->gprmask_; }
+
+ // Return cprmask1 from the .reginfo section of this object.
+ Valtype
+ cprmask1() const
+ { return this->cprmask1_; }
+
+ // Return cprmask2 from the .reginfo section of this object.
+ Valtype
+ cprmask2() const
+ { return this->cprmask2_; }
+
+ // Return cprmask3 from the .reginfo section of this object.
+ Valtype
+ cprmask3() const
+ { return this->cprmask3_; }
+
+ // Return cprmask4 from the .reginfo section of this object.
+ Valtype
+ cprmask4() const
+ { return this->cprmask4_; }
+
+ protected:
+ // Count the local symbols.
+ void
+ do_count_local_symbols(Stringpool_template<char>*,
+ Stringpool_template<char>*);
+
+ // Read the symbol information.
+ void
+ do_read_symbols(Read_symbols_data* sd);
+
+ private:
+ // processor-specific flags in ELF file header.
+ elfcpp::Elf_Word processor_specific_flags_;
+
+ // Bit vector to tell if a local symbol is a MIPS16 symbol or not.
+ // This is only valid after do_count_local_symbol is called.
+ std::vector<bool> local_symbol_is_mips16_;
+
+ // Bit vector to tell if a local symbol is a microMIPS symbol or not.
+ // This is only valid after do_count_local_symbol is called.
+ std::vector<bool> local_symbol_is_micromips_;
+
+ // Map from section index to the MIPS16 stub for that section. This contains
+ // all stubs found in this object.
+ Mips16_stubs_int_map mips16_stub_sections_;
+
+ // Local symbols that have "non 16-bit" call relocation. This relocation
+ // would need to refer to a MIPS16 fn stub, if there is one.
+ std::set<unsigned int> local_non_16bit_calls_;
+
+ // Local symbols that have 16-bit call relocation R_MIPS16_26. Local MIPS16
+ // call or call_fp stubs will only be needed if there is some R_MIPS16_26
+ // relocation that refers to the stub symbol.
+ std::set<unsigned int> local_16bit_calls_;
+
+ // Map from local symbol index to the MIPS16 fn stub for that symbol.
+ // This contains only the stubs that we decided not to discard.
+ Mips16_stubs_int_map local_mips16_fn_stubs_;
+
+ // Map from local symbol index to the MIPS16 call stub for that symbol.
+ // This contains only the stubs that we decided not to discard.
+ Mips16_stubs_int_map local_mips16_call_stubs_;
+
+ // gp value that was used to create this object.
+ Mips_address gp_;
+ // Whether the object is a PIC object.
+ bool is_pic_ : 1;
+ // Whether the object uses N32 ABI.
+ bool is_n32_ : 1;
+ // Whether the object uses N64 ABI.
+ bool is_n64_ : 1;
+ // The Mips_got_info for this object.
+ Mips_got_info<size, big_endian>* got_info_;
+
+ // Bit vector to tell if a section is a MIPS16 fn stub section or not.
+ // This is only valid after do_read_symbols is called.
+ std::vector<bool> section_is_mips16_fn_stub_;
+
+ // Bit vector to tell if a section is a MIPS16 call stub section or not.
+ // This is only valid after do_read_symbols is called.
+ std::vector<bool> section_is_mips16_call_stub_;
+
+ // Bit vector to tell if a section is a MIPS16 call_fp stub section or not.
+ // This is only valid after do_read_symbols is called.
+ std::vector<bool> section_is_mips16_call_fp_stub_;
+
+ // .pdr section index.
+ unsigned int pdr_shndx_;
+
+ // gprmask from the .reginfo section of this object.
+ Valtype gprmask_;
+ // cprmask1 from the .reginfo section of this object.
+ Valtype cprmask1_;
+ // cprmask2 from the .reginfo section of this object.
+ Valtype cprmask2_;
+ // cprmask3 from the .reginfo section of this object.
+ Valtype cprmask3_;
+ // cprmask4 from the .reginfo section of this object.
+ Valtype cprmask4_;
+};
+
+// Mips_output_data_got class.
+
+template<int size, bool big_endian>
+class Mips_output_data_got : public Output_data_got<size, big_endian>
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+ typedef Output_data_reloc<elfcpp::SHT_REL, true, size, big_endian>
+ Reloc_section;
+ typedef typename elfcpp::Swap<size, big_endian>::Valtype Valtype;
+
+ public:
+ Mips_output_data_got(Target_mips<size, big_endian>* target,
+ Symbol_table* symtab, Layout* layout)
+ : Output_data_got<size, big_endian>(), target_(target),
+ symbol_table_(symtab), layout_(layout), static_relocs_(), got_view_(NULL),
+ first_global_got_dynsym_index_(-1U), primary_got_(NULL),
+ secondary_got_relocs_()
+ {
+ this->master_got_info_ = new Mips_got_info<size, big_endian>();
+ this->set_addralign(16);
+ }
+
+ // Reserve GOT entry for a GOT relocation of type R_TYPE against symbol
+ // SYMNDX + ADDEND, where SYMNDX is a local symbol in section SHNDX in OBJECT.
+ void
+ record_local_got_symbol(Mips_relobj<size, big_endian>* object,
+ unsigned int symndx, Mips_address addend,
+ unsigned int r_type, unsigned int shndx)
+ {
+ this->master_got_info_->record_local_got_symbol(object, symndx, addend,
+ r_type, shndx);
+ }
+
+ // Reserve GOT entry for a GOT relocation of type R_TYPE against MIPS_SYM,
+ // in OBJECT. FOR_CALL is true if the caller is only interested in
+ // using the GOT entry for calls. DYN_RELOC is true if R_TYPE is a dynamic
+ // relocation.
+ void
+ record_global_got_symbol(Mips_symbol<size>* mips_sym,
+ Mips_relobj<size, big_endian>* object,
+ unsigned int r_type, bool dyn_reloc, bool for_call)
+ {
+ this->master_got_info_->record_global_got_symbol(mips_sym, object, r_type,
+ dyn_reloc, for_call);
+ }
+
+ // Record that OBJECT has a page relocation against symbol SYMNDX and
+ // that ADDEND is the addend for that relocation.
+ void
+ record_got_page_entry(Mips_relobj<size, big_endian>* object,
+ unsigned int symndx, int addend)
+ { this->master_got_info_->record_got_page_entry(object, symndx, addend); }
+
+ // Add a static entry for the GOT entry at OFFSET. GSYM is a global
+ // symbol and R_TYPE is the code of a dynamic relocation that needs to be
+ // applied in a static link.
+ void
+ add_static_reloc(unsigned int got_offset, unsigned int r_type,
+ Mips_symbol<size>* gsym)
+ { this->static_relocs_.push_back(Static_reloc(got_offset, r_type, gsym)); }
+
+ // Add a static reloc for the GOT entry at OFFSET. RELOBJ is an object
+ // defining a local symbol with INDEX. R_TYPE is the code of a dynamic
+ // relocation that needs to be applied in a static link.
+ void
+ add_static_reloc(unsigned int got_offset, unsigned int r_type,
+ Sized_relobj_file<size, big_endian>* relobj,
+ unsigned int index)
+ {
+ this->static_relocs_.push_back(Static_reloc(got_offset, r_type, relobj,
+ index));
+ }
+
+ // Record that global symbol GSYM has R_TYPE dynamic relocation in the
+ // secondary GOT at OFFSET.
+ void
+ add_secondary_got_reloc(unsigned int got_offset, unsigned int r_type,
+ Mips_symbol<size>* gsym)
+ {
+ this->secondary_got_relocs_.push_back(Static_reloc(got_offset,
+ r_type, gsym));
+ }
+
+ // Update GOT entry at OFFSET with VALUE.
+ void
+ update_got_entry(unsigned int offset, Mips_address value)
+ {
+ elfcpp::Swap<size, big_endian>::writeval(this->got_view_ + offset, value);
+ }
+
+ // Return the number of entries in local part of the GOT. This includes
+ // local entries, page entries and 2 reserved entries.
+ unsigned int
+ get_local_gotno() const
+ {
+ if (!this->multi_got())
+ {
+ return (2 + this->master_got_info_->local_gotno()
+ + this->master_got_info_->page_gotno());
+ }
+ else
+ return 2 + this->primary_got_->local_gotno() + this->primary_got_->page_gotno();
+ }
+
+ // Return dynamic symbol table index of the first symbol with global GOT
+ // entry.
+ unsigned int
+ first_global_got_dynsym_index() const
+ { return this->first_global_got_dynsym_index_; }
+
+ // Set dynamic symbol table index of the first symbol with global GOT entry.
+ void
+ set_first_global_got_dynsym_index(unsigned int index)
+ { this->first_global_got_dynsym_index_ = index; }
+
+ // Lay out the GOT. Add local, global and TLS entries. If GOT is
+ // larger than 64K, create multi-GOT.
+ void
+ lay_out_got(Layout* layout, Symbol_table* symtab,
+ const Input_objects* input_objects);
+
+ // Create multi-GOT. For every GOT, add local, global and TLS entries.
+ void
+ lay_out_multi_got(Layout* layout, const Input_objects* input_objects);
+
+ // Attempt to merge GOTs of different input objects.
+ void
+ merge_gots(const Input_objects* input_objects);
+
+ // Consider merging FROM, which is OBJECT's GOT, into TO. Return false if
+ // this would lead to overflow, true if they were merged successfully.
+ bool
+ merge_got_with(Mips_got_info<size, big_endian>* from,
+ Mips_relobj<size, big_endian>* object,
+ Mips_got_info<size, big_endian>* to);
+
+ // Return the offset of GOT page entry for VALUE. For multi-GOT links,
+ // use OBJECT's GOT.
+ unsigned int
+ get_got_page_offset(Mips_address value,
+ const Mips_relobj<size, big_endian>* object)
+ {
+ Mips_got_info<size, big_endian>* g = (!this->multi_got()
+ ? this->master_got_info_
+ : object->get_got_info());
+ gold_assert(g != NULL);
+ return g->get_got_page_offset(value, this);
+ }
+
+ // Return the GOT offset of type GOT_TYPE of the global symbol
+ // GSYM. For multi-GOT links, use OBJECT's GOT.
+ unsigned int got_offset(const Symbol* gsym, unsigned int got_type,
+ Mips_relobj<size, big_endian>* object) const
+ {
+ if (!this->multi_got())
+ return gsym->got_offset(got_type);
+ else
+ {
+ Mips_got_info<size, big_endian>* g = object->get_got_info();
+ gold_assert(g != NULL);
+ return gsym->got_offset(g->multigot_got_type(got_type));
+ }
+ }
+
+ // Return the GOT offset of type GOT_TYPE of the local symbol
+ // SYMNDX.
+ unsigned int
+ got_offset(unsigned int symndx, unsigned int got_type,
+ Sized_relobj_file<size, big_endian>* object) const
+ { return object->local_got_offset(symndx, got_type); }
+
+ // Return the offset of TLS LDM entry. For multi-GOT links, use OBJECT's GOT.
+ unsigned int
+ tls_ldm_offset(Mips_relobj<size, big_endian>* object) const
+ {
+ Mips_got_info<size, big_endian>* g = (!this->multi_got()
+ ? this->master_got_info_
+ : object->get_got_info());
+ gold_assert(g != NULL);
+ return g->tls_ldm_offset();
+ }
+
+ // Set the offset of TLS LDM entry. For multi-GOT links, use OBJECT's GOT.
+ void
+ set_tls_ldm_offset(unsigned int tls_ldm_offset,
+ Mips_relobj<size, big_endian>* object)
+ {
+ Mips_got_info<size, big_endian>* g = (!this->multi_got()
+ ? this->master_got_info_
+ : object->get_got_info());
+ gold_assert(g != NULL);
+ g->set_tls_ldm_offset(tls_ldm_offset);
+ }
+
+ // Return true for multi-GOT links.
+ bool
+ multi_got() const
+ { return this->primary_got_ != NULL; }
+
+ // Return the offset of OBJECT's GOT from the start of .got section.
+ unsigned int
+ get_got_offset(const Mips_relobj<size, big_endian>* object)
+ {
+ if (!this->multi_got())
+ return 0;
+ else
+ {
+ Mips_got_info<size, big_endian>* g = object->get_got_info();
+ return g != NULL ? g->offset() : 0;
+ }
+ }
+
+ // Create global GOT entries that should be in the GGA_RELOC_ONLY area.
+ void
+ add_reloc_only_entries()
+ { this->master_got_info_->add_reloc_only_entries(this); }
+
+ // Return offset of the primary GOT's entry for global symbol.
+ unsigned int
+ get_primary_got_offset(const Mips_symbol<size>* sym) const
+ {
+ gold_assert(sym->global_got_area() != GGA_NONE);
+ return (this->get_local_gotno() + sym->dynsym_index()
+ - this->first_global_got_dynsym_index()) * size/8;
+ }
+
+ // For the entry at offset GOT_OFFSET, return its offset from the gp.
+ // Input argument GOT_OFFSET is always global offset from the start of
+ // .got section, for both single and multi-GOT links.
+ // For single GOT links, this returns GOT_OFFSET - 0x7FF0. For multi-GOT
+ // links, the return value is object_got_offset - 0x7FF0, where
+ // object_got_offset is offset in the OBJECT's GOT.
+ int
+ gp_offset(unsigned int got_offset,
+ const Mips_relobj<size, big_endian>* object) const
+ {
+ return (this->address() + got_offset
+ - this->target_->adjusted_gp_value(object));
+ }
+
+ protected:
+ // Write out the GOT table.
+ void
+ do_write(Output_file*);
+
+ private:
+
+ // This class represent dynamic relocations that need to be applied by
+ // gold because we are using TLS relocations in a static link.
+ class Static_reloc
+ {
+ public:
+ Static_reloc(unsigned int got_offset, unsigned int r_type,
+ Mips_symbol<size>* gsym)
+ : got_offset_(got_offset), r_type_(r_type), symbol_is_global_(true)
+ { this->u_.global.symbol = gsym; }
+
+ Static_reloc(unsigned int got_offset, unsigned int r_type,
+ Sized_relobj_file<size, big_endian>* relobj, unsigned int index)
+ : got_offset_(got_offset), r_type_(r_type), symbol_is_global_(false)
+ {
+ this->u_.local.relobj = relobj;
+ this->u_.local.index = index;
+ }
+
+ // Return the GOT offset.
+ unsigned int
+ got_offset() const
+ { return this->got_offset_; }
+
+ // Relocation type.
+ unsigned int
+ r_type() const
+ { return this->r_type_; }
+
+ // Whether the symbol is global or not.
+ bool
+ symbol_is_global() const
+ { return this->symbol_is_global_; }
+
+ // For a relocation against a global symbol, the global symbol.
+ Mips_symbol<size>*
+ symbol() const
+ {
+ gold_assert(this->symbol_is_global_);
+ return this->u_.global.symbol;
+ }
+
+ // For a relocation against a local symbol, the defining object.
+ Sized_relobj_file<size, big_endian>*
+ relobj() const
+ {
+ gold_assert(!this->symbol_is_global_);
+ return this->u_.local.relobj;
+ }
+
+ // For a relocation against a local symbol, the local symbol index.
+ unsigned int
+ index() const
+ {
+ gold_assert(!this->symbol_is_global_);
+ return this->u_.local.index;
+ }
+
+ private:
+ // GOT offset of the entry to which this relocation is applied.
+ unsigned int got_offset_;
+ // Type of relocation.
+ unsigned int r_type_;
+ // Whether this relocation is against a global symbol.
+ bool symbol_is_global_;
+ // A global or local symbol.
+ union
+ {
+ struct
+ {
+ // For a global symbol, the symbol itself.
+ Mips_symbol<size>* symbol;
+ } global;
+ struct
+ {
+ // For a local symbol, the object defining object.
+ Sized_relobj_file<size, big_endian>* relobj;
+ // For a local symbol, the symbol index.
+ unsigned int index;
+ } local;
+ } u_;
+ };
+
+ // The target.
+ Target_mips<size, big_endian>* target_;
+ // The symbol table.
+ Symbol_table* symbol_table_;
+ // The layout.
+ Layout* layout_;
+ // Static relocs to be applied to the GOT.
+ std::vector<Static_reloc> static_relocs_;
+ // .got section view.
+ unsigned char* got_view_;
+ // The dynamic symbol table index of the first symbol with global GOT entry.
+ unsigned int first_global_got_dynsym_index_;
+ // The master GOT information.
+ Mips_got_info<size, big_endian>* master_got_info_;
+ // The primary GOT information.
+ Mips_got_info<size, big_endian>* primary_got_;
+ // Secondary GOT fixups.
+ std::vector<Static_reloc> secondary_got_relocs_;
+};
+
+// A class to handle LA25 stubs - non-PIC interface to a PIC function. There are
+// two ways of creating these interfaces. The first is to add:
+//
+// lui $25,%hi(func)
+// j func
+// addiu $25,$25,%lo(func)
+//
+// to a separate trampoline section. The second is to add:
+//
+// lui $25,%hi(func)
+// addiu $25,$25,%lo(func)
+//
+// immediately before a PIC function "func", but only if a function is at the
+// beginning of the section, and the section is not too heavily aligned (i.e we
+// would need to add no more than 2 nops before the stub.)
+//
+// We only create stubs of the first type.
+
+template<int size, bool big_endian>
+class Mips_output_data_la25_stub : public Output_section_data
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+
+ public:
+ Mips_output_data_la25_stub()
+ : Output_section_data(size == 32 ? 4 : 8), symbols_()
+ { }
+
+ // Create LA25 stub for a symbol.
+ void
+ create_la25_stub(Symbol_table* symtab, Target_mips<size, big_endian>* target,
+ Mips_symbol<size>* gsym);
+
+ // Return output address of a stub.
+ Mips_address
+ stub_address(const Mips_symbol<size>* sym) const
+ {
+ gold_assert(sym->has_la25_stub());
+ return this->address() + sym->la25_stub_offset();
+ }
+
+ protected:
+ void
+ do_adjust_output_section(Output_section* os)
+ { os->set_entsize(0); }
+
+ private:
+ // Template for standard LA25 stub.
+ static const uint32_t la25_stub_entry[];
+ // Template for microMIPS LA25 stub.
+ static const uint32_t la25_stub_micromips_entry[];
+
+ // Set the final size.
+ void
+ set_final_data_size()
+ { this->set_data_size(this->symbols_.size() * 16); }
+
+ // Create a symbol for SYM stub's value and size, to help make the
+ // disassembly easier to read.
+ void
+ create_stub_symbol(Mips_symbol<size>* sym, Symbol_table* symtab,
+ Target_mips<size, big_endian>* target, uint64_t symsize);
+
+ // Write out the LA25 stub section.
+ void
+ do_write(Output_file*);
+
+ // Symbols that have LA25 stubs.
+ Unordered_set<Mips_symbol<size>*> symbols_;
+};
+
+// A class to handle the PLT data.
+
+template<int size, bool big_endian>
+class Mips_output_data_plt : public Output_section_data
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+ typedef Output_data_reloc<elfcpp::SHT_REL, true,
+ size, big_endian> Reloc_section;
+
+ public:
+ // Create the PLT section. The ordinary .got section is an argument,
+ // since we need to refer to the start.
+ Mips_output_data_plt(Layout* layout, Output_data_space* got_plt,
+ Target_mips<size, big_endian>* target)
+ : Output_section_data(size == 32 ? 4 : 8), got_plt_(got_plt), symbols_(),
+ plt_mips_offset_(0), plt_comp_offset_(0), plt_header_size_(0),
+ target_(target)
+ {
+ this->rel_ = new Reloc_section(false);
+ layout->add_output_section_data(".rel.plt", elfcpp::SHT_REL,
+ elfcpp::SHF_ALLOC, this->rel_,
+ ORDER_DYNAMIC_PLT_RELOCS, false);
+ }
+
+ // Add an entry to the PLT for a symbol referenced by r_type relocation.
+ void
+ add_entry(Mips_symbol<size>* gsym, unsigned int r_type);
+
+ // Return the .rel.plt section data.
+ const Reloc_section*
+ rel_plt() const
+ { return this->rel_; }
+
+ // Return the number of PLT entries.
+ unsigned int
+ entry_count() const
+ { return this->symbols_.size(); }
+
+ // Return the offset of the first non-reserved PLT entry.
+ unsigned int
+ first_plt_entry_offset() const
+ { return sizeof(plt0_entry_o32); }
+
+ // Return the size of a PLT entry.
+ unsigned int
+ plt_entry_size() const
+ { return sizeof(plt_entry); }
+
+ // Set final PLT offsets. For each symbol, determine whether standard or
+ // compressed (MIPS16 or microMIPS) PLT entry is used.
+ void
+ set_plt_offsets();
+
+ // Return the offset of the first standard PLT entry.
+ unsigned int
+ first_mips_plt_offset() const
+ { return this->plt_header_size_; }
+
+ // Return the offset of the first compressed PLT entry.
+ unsigned int
+ first_comp_plt_offset() const
+ { return this->plt_header_size_ + this->plt_mips_offset_; }
+
+ // Return whether there are any standard PLT entries.
+ bool
+ has_standard_entries() const
+ { return this->plt_mips_offset_ > 0; }
+
+ // Return the output address of standard PLT entry.
+ Mips_address
+ mips_entry_address(const Mips_symbol<size>* sym) const
+ {
+ gold_assert (sym->has_mips_plt_offset());
+ return (this->address() + this->first_mips_plt_offset()
+ + sym->mips_plt_offset());
+ }
+
+ // Return the output address of compressed (MIPS16 or microMIPS) PLT entry.
+ Mips_address
+ comp_entry_address(const Mips_symbol<size>* sym) const
+ {
+ gold_assert (sym->has_comp_plt_offset());
+ return (this->address() + this->first_comp_plt_offset()
+ + sym->comp_plt_offset());
+ }
+
+ protected:
+ void
+ do_adjust_output_section(Output_section* os)
+ { os->set_entsize(0); }
+
+ // Write to a map file.
+ void
+ do_print_to_mapfile(Mapfile* mapfile) const
+ { mapfile->print_output_data(this, _(".plt")); }
+
+ private:
+ // Template for the first PLT entry.
+ static const uint32_t plt0_entry_o32[];
+ static const uint32_t plt0_entry_n32[];
+ static const uint32_t plt0_entry_n64[];
+ static const uint32_t plt0_entry_micromips_o32[];
+ static const uint32_t plt0_entry_micromips32_o32[];
+
+ // Template for subsequent PLT entries.
+ static const uint32_t plt_entry[];
+ static const uint32_t plt_entry_mips16_o32[];
+ static const uint32_t plt_entry_micromips_o32[];
+ static const uint32_t plt_entry_micromips32_o32[];
+
+ // Set the final size.
+ void
+ set_final_data_size()
+ {
+ this->set_data_size(this->plt_header_size_ + this->plt_mips_offset_
+ + this->plt_comp_offset_);
+ }
+
+ // Write out the PLT data.
+ void
+ do_write(Output_file*);
+
+ // Return whether the plt header contains microMIPS code. For the sake of
+ // cache alignment always use a standard header whenever any standard entries
+ // are present even if microMIPS entries are present as well. This also lets
+ // the microMIPS header rely on the value of $v0 only set by microMIPS
+ // entries, for a small size reduction.
+ bool
+ is_plt_header_compressed() const
+ {
+ gold_assert(this->plt_mips_offset_ + this->plt_comp_offset_ != 0);
+ return this->target_->is_output_micromips() && this->plt_mips_offset_ == 0;
+ }
+
+ // Return the size of the PLT header.
+ unsigned int
+ get_plt_header_size() const
+ {
+ if (this->target_->is_output_n64())
+ return 4 * sizeof(plt0_entry_n64) / sizeof(plt0_entry_n64[0]);
+ else if (this->target_->is_output_n32())
+ return 4 * sizeof(plt0_entry_n32) / sizeof(plt0_entry_n32[0]);
+ else if (!this->is_plt_header_compressed())
+ return 4 * sizeof(plt0_entry_o32) / sizeof(plt0_entry_o32[0]);
+ else if (this->target_->use_32bit_micromips_instructions())
+ return (2 * sizeof(plt0_entry_micromips32_o32)
+ / sizeof(plt0_entry_micromips32_o32[0]));
+ else
+ return (2 * sizeof(plt0_entry_micromips_o32)
+ / sizeof(plt0_entry_micromips_o32[0]));
+ }
+
+ // Return the PLT header entry.
+ const uint32_t*
+ get_plt_header_entry() const
+ {
+ if (this->target_->is_output_n64())
+ return plt0_entry_n64;
+ else if (this->target_->is_output_n32())
+ return plt0_entry_n32;
+ else if (!this->is_plt_header_compressed())
+ return plt0_entry_o32;
+ else if (this->target_->use_32bit_micromips_instructions())
+ return plt0_entry_micromips32_o32;
+ else
+ return plt0_entry_micromips_o32;
+ }
+
+ // Return the size of the standard PLT entry.
+ unsigned int
+ standard_plt_entry_size() const
+ { return 4 * sizeof(plt_entry) / sizeof(plt_entry[0]); }
+
+ // Return the size of the compressed PLT entry.
+ unsigned int
+ compressed_plt_entry_size() const
+ {
+ gold_assert(!this->target_->is_output_newabi());
+
+ if (!this->target_->is_output_micromips())
+ return (2 * sizeof(plt_entry_mips16_o32)
+ / sizeof(plt_entry_mips16_o32[0]));
+ else if (this->target_->use_32bit_micromips_instructions())
+ return (2 * sizeof(plt_entry_micromips32_o32)
+ / sizeof(plt_entry_micromips32_o32[0]));
+ else
+ return (2 * sizeof(plt_entry_micromips_o32)
+ / sizeof(plt_entry_micromips_o32[0]));
+ }
+
+ // The reloc section.
+ Reloc_section* rel_;
+ // The .got.plt section.
+ Output_data_space* got_plt_;
+ // Symbols that have PLT entry.
+ std::vector<Mips_symbol<size>*> symbols_;
+ // The offset of the next standard PLT entry to create.
+ unsigned int plt_mips_offset_;
+ // The offset of the next compressed PLT entry to create.
+ unsigned int plt_comp_offset_;
+ // The size of the PLT header in bytes.
+ unsigned int plt_header_size_;
+ // The target.
+ Target_mips<size, big_endian>* target_;
+};
+
+// A class to handle the .MIPS.stubs data.
+
+template<int size, bool big_endian>
+class Mips_output_data_mips_stubs : public Output_section_data
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+
+ public:
+ Mips_output_data_mips_stubs(Target_mips<size, big_endian>* target)
+ : Output_section_data(size == 32 ? 4 : 8), symbols_(), dynsym_count_(-1U),
+ stub_offsets_are_set_(false), target_(target)
+ { }
+
+ // Create entry for a symbol.
+ void
+ make_entry(Mips_symbol<size>*);
+
+ // Remove entry for a symbol.
+ void
+ remove_entry(Mips_symbol<size>* gsym);
+
+ // Set stub offsets for symbols. This method expects that the number of
+ // entries in dynamic symbol table is set.
+ void
+ set_lazy_stub_offsets();
+
+ void
+ set_needs_dynsym_value();
+
+ // Set the number of entries in dynamic symbol table.
+ void
+ set_dynsym_count(unsigned int dynsym_count)
+ { this->dynsym_count_ = dynsym_count; }
+
+ // Return maximum size of the stub, ie. the stub size if the dynamic symbol
+ // count is greater than 0x10000. If the dynamic symbol count is less than
+ // 0x10000, the stub will be 4 bytes smaller.
+ // There's no disadvantage from using microMIPS code here, so for the sake of
+ // pure-microMIPS binaries we prefer it whenever there's any microMIPS code in
+ // output produced at all. This has a benefit of stubs being shorter by
+ // 4 bytes each too, unless in the insn32 mode.
+ unsigned int
+ stub_max_size() const
+ {
+ if (!this->target_->is_output_micromips()
+ || this->target_->use_32bit_micromips_instructions())
+ return 20;
+ else
+ return 16;
+ }
+
+ // Return the size of the stub. This method expects that the final dynsym
+ // count is set.
+ unsigned int
+ stub_size() const
+ {
+ gold_assert(this->dynsym_count_ != -1U);
+ if (this->dynsym_count_ > 0x10000)
+ return this->stub_max_size();
+ else
+ return this->stub_max_size() - 4;
+ }
+
+ // Return output address of a stub.
+ Mips_address
+ stub_address(const Mips_symbol<size>* sym) const
+ {
+ gold_assert(sym->has_lazy_stub());
+ return this->address() + sym->lazy_stub_offset();
+ }
+
+ protected:
+ void
+ do_adjust_output_section(Output_section* os)
+ { os->set_entsize(0); }
+
+ // Write to a map file.
+ void
+ do_print_to_mapfile(Mapfile* mapfile) const
+ { mapfile->print_output_data(this, _(".MIPS.stubs")); }
+
+ private:
+ static const uint32_t lazy_stub_normal_1[];
+ static const uint32_t lazy_stub_normal_1_n64[];
+ static const uint32_t lazy_stub_normal_2[];
+ static const uint32_t lazy_stub_normal_2_n64[];
+ static const uint32_t lazy_stub_big[];
+ static const uint32_t lazy_stub_big_n64[];
+
+ static const uint32_t lazy_stub_micromips_normal_1[];
+ static const uint32_t lazy_stub_micromips_normal_1_n64[];
+ static const uint32_t lazy_stub_micromips_normal_2[];
+ static const uint32_t lazy_stub_micromips_normal_2_n64[];
+ static const uint32_t lazy_stub_micromips_big[];
+ static const uint32_t lazy_stub_micromips_big_n64[];
+
+ static const uint32_t lazy_stub_micromips32_normal_1[];
+ static const uint32_t lazy_stub_micromips32_normal_1_n64[];
+ static const uint32_t lazy_stub_micromips32_normal_2[];
+ static const uint32_t lazy_stub_micromips32_normal_2_n64[];
+ static const uint32_t lazy_stub_micromips32_big[];
+ static const uint32_t lazy_stub_micromips32_big_n64[];
+
+ // Set the final size.
+ void
+ set_final_data_size()
+ { this->set_data_size(this->symbols_.size() * this->stub_max_size()); }
+
+ // Write out the .MIPS.stubs data.
+ void
+ do_write(Output_file*);
+
+ // .MIPS.stubs symbols
+ Unordered_set<Mips_symbol<size>*> symbols_;
+ // Number of entries in dynamic symbol table.
+ unsigned int dynsym_count_;
+ // Whether the stub offsets are set.
+ bool stub_offsets_are_set_;
+ // The target.
+ Target_mips<size, big_endian>* target_;
+};
+
+// This class handles Mips .reginfo output section.
+
+template<int size, bool big_endian>
+class Mips_output_section_reginfo : public Output_section
+{
+ typedef typename elfcpp::Swap<size, big_endian>::Valtype Valtype;
+
+ public:
+ Mips_output_section_reginfo(const char* name, elfcpp::Elf_Word type,
+ elfcpp::Elf_Xword flags,
+ Target_mips<size, big_endian>* target)
+ : Output_section(name, type, flags), target_(target), gprmask_(0),
+ cprmask1_(0), cprmask2_(0), cprmask3_(0), cprmask4_(0)
+ {
+ this->set_always_keeps_input_sections();
+ }
+
+ // Downcast a base pointer to a Mips_output_section_reginfo pointer.
+ static Mips_output_section_reginfo<size, big_endian>*
+ as_mips_output_section_reginfo(Output_section* os)
+ { return static_cast<Mips_output_section_reginfo<size, big_endian>*>(os); }
+
+ // Set masks of the output .reginfo section.
+ void
+ set_masks(Valtype gprmask, Valtype cprmask1, Valtype cprmask2,
+ Valtype cprmask3, Valtype cprmask4)
+ {
+ this->gprmask_ = gprmask;
+ this->cprmask1_ = cprmask1;
+ this->cprmask2_ = cprmask2;
+ this->cprmask3_ = cprmask3;
+ this->cprmask4_ = cprmask4;
+ }
+
+ protected:
+ // Set the final data size.
+ void
+ set_final_data_size()
+ { this->set_data_size(24); }
+
+ // Write out reginfo section.
+ void
+ do_write(Output_file* of);
+
+ private:
+ Target_mips<size, big_endian>* target_;
+
+ // gprmask of the output .reginfo section.
+ Valtype gprmask_;
+ // cprmask1 of the output .reginfo section.
+ Valtype cprmask1_;
+ // cprmask2 of the output .reginfo section.
+ Valtype cprmask2_;
+ // cprmask3 of the output .reginfo section.
+ Valtype cprmask3_;
+ // cprmask4 of the output .reginfo section.
+ Valtype cprmask4_;
+};
+
+// The MIPS target has relocation types which default handling of relocatable
+// relocation cannot process. So we have to extend the default code.
+
+template<bool big_endian, int sh_type, typename Classify_reloc>
+class Mips_scan_relocatable_relocs :
+ public Default_scan_relocatable_relocs<sh_type, Classify_reloc>
+{
+ public:
+ // Return the strategy to use for a local symbol which is a section
+ // symbol, given the relocation type.
+ inline Relocatable_relocs::Reloc_strategy
+ local_section_strategy(unsigned int r_type, Relobj* object)
+ {
+ if (sh_type == elfcpp::SHT_RELA)
+ return Relocatable_relocs::RELOC_ADJUST_FOR_SECTION_RELA;
+ else
+ {
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_26:
+ return Relocatable_relocs::RELOC_SPECIAL;
+
+ default:
+ return Default_scan_relocatable_relocs<sh_type, Classify_reloc>::
+ local_section_strategy(r_type, object);
+ }
+ }
+ }
+};
+
+// Mips_copy_relocs class. The only difference from the base class is the
+// method emit_mips, which should be called instead of Copy_reloc_entry::emit.
+// Mips cannot convert all relocation types to dynamic relocs. If a reloc
+// cannot be made dynamic, a COPY reloc is emitted.
+
+template<int sh_type, int size, bool big_endian>
+class Mips_copy_relocs : public Copy_relocs<sh_type, size, big_endian>
+{
+ public:
+ Mips_copy_relocs()
+ : Copy_relocs<sh_type, size, big_endian>(elfcpp::R_MIPS_COPY)
+ { }
+
+ // Emit any saved relocations which turn out to be needed. This is
+ // called after all the relocs have been scanned.
+ void
+ emit_mips(Output_data_reloc<sh_type, true, size, big_endian>*,
+ Symbol_table*, Layout*, Target_mips<size, big_endian>*);
+
+ private:
+ typedef typename Copy_relocs<sh_type, size, big_endian>::Copy_reloc_entry
+ Copy_reloc_entry;
+
+ // Emit this reloc if appropriate. This is called after we have
+ // scanned all the relocations, so we know whether we emitted a
+ // COPY relocation for SYM_.
+ void
+ emit_entry(Copy_reloc_entry& entry,
+ Output_data_reloc<sh_type, true, size, big_endian>* reloc_section,
+ Symbol_table* symtab, Layout* layout,
+ Target_mips<size, big_endian>* target);
+};
+
+
+// Return true if the symbol SYM should be considered to resolve local
+// to the current module, and false otherwise. The logic is taken from
+// GNU ld's method _bfd_elf_symbol_refs_local_p.
+static bool
+symbol_refs_local(const Symbol* sym, bool has_dynsym_entry,
+ bool local_protected)
+{
+ // If it's a local sym, of course we resolve locally.
+ if (sym == NULL)
+ return true;
+
+ // STV_HIDDEN or STV_INTERNAL ones must be local.
+ if (sym->visibility() == elfcpp::STV_HIDDEN
+ || sym->visibility() == elfcpp::STV_INTERNAL)
+ return true;
+
+ // If we don't have a definition in a regular file, then we can't
+ // resolve locally. The sym is either undefined or dynamic.
+ if (sym->source() != Symbol::FROM_OBJECT || sym->object()->is_dynamic()
+ || sym->is_undefined())
+ return false;
+
+ // Forced local symbols resolve locally.
+ if (sym->is_forced_local())
+ return true;
+
+ // As do non-dynamic symbols.
+ if (!has_dynsym_entry)
+ return true;
+
+ // At this point, we know the symbol is defined and dynamic. In an
+ // executable it must resolve locally, likewise when building symbolic
+ // shared libraries.
+ if (parameters->options().output_is_executable()
+ || parameters->options().Bsymbolic())
+ return true;
+
+ // Now deal with defined dynamic symbols in shared libraries. Ones
+ // with default visibility might not resolve locally.
+ if (sym->visibility() == elfcpp::STV_DEFAULT)
+ return false;
+
+ // STV_PROTECTED non-function symbols are local.
+ if (sym->type() != elfcpp::STT_FUNC)
+ return true;
+
+ // Function pointer equality tests may require that STV_PROTECTED
+ // symbols be treated as dynamic symbols. If the address of a
+ // function not defined in an executable is set to that function's
+ // plt entry in the executable, then the address of the function in
+ // a shared library must also be the plt entry in the executable.
+ return local_protected;
+}
+
+// Return TRUE if references to this symbol always reference the symbol in this
+// object.
+static bool
+symbol_references_local(const Symbol* sym, bool has_dynsym_entry)
+{
+ return symbol_refs_local(sym, has_dynsym_entry, false);
+}
+
+// Return TRUE if calls to this symbol always call the version in this object.
+static bool
+symbol_calls_local(const Symbol* sym, bool has_dynsym_entry)
+{
+ return symbol_refs_local(sym, has_dynsym_entry, true);
+}
+
+// Compare GOT offsets of two symbols.
+
+template<int size, bool big_endian>
+static bool
+got_offset_compare(Symbol* sym1, Symbol* sym2)
+{
+ Mips_symbol<size>* mips_sym1 = Mips_symbol<size>::as_mips_sym(sym1);
+ Mips_symbol<size>* mips_sym2 = Mips_symbol<size>::as_mips_sym(sym2);
+ unsigned int area1 = mips_sym1->global_got_area();
+ unsigned int area2 = mips_sym2->global_got_area();
+ gold_assert(area1 != GGA_NONE && area1 != GGA_NONE);
+
+ // GGA_NORMAL entries always come before GGA_RELOC_ONLY.
+ if (area1 != area2)
+ return area1 < area2;
+
+ return mips_sym1->global_gotoffset() < mips_sym2->global_gotoffset();
+}
+
+// This method divides dynamic symbols into symbols that have GOT entry, and
+// symbols that don't have GOT entry. It also sorts symbols with the GOT entry.
+// Mips ABI requires that symbols with the GOT entry must be at the end of
+// dynamic symbol table, and the order in dynamic symbol table must match the
+// order in GOT.
+
+template<int size, bool big_endian>
+static void
+reorder_dyn_symbols(std::vector<Symbol*>* dyn_symbols,
+ std::vector<Symbol*>* non_got_symbols,
+ std::vector<Symbol*>* got_symbols)
+{
+ for (std::vector<Symbol*>::iterator p = dyn_symbols->begin();
+ p != dyn_symbols->end();
+ ++p)
+ {
+ Mips_symbol<size>* mips_sym = Mips_symbol<size>::as_mips_sym(*p);
+ if (mips_sym->global_got_area() == GGA_NORMAL
+ || mips_sym->global_got_area() == GGA_RELOC_ONLY)
+ got_symbols->push_back(mips_sym);
+ else
+ non_got_symbols->push_back(mips_sym);
+ }
+
+ std::sort(got_symbols->begin(), got_symbols->end(),
+ got_offset_compare<size, big_endian>);
+}
+
+// Functor class for processing the global symbol table.
+
+template<int size, bool big_endian>
+class Symbol_visitor_check_symbols
+{
+ public:
+ Symbol_visitor_check_symbols(Target_mips<size, big_endian>* target,
+ Layout* layout, Symbol_table* symtab)
+ : target_(target), layout_(layout), symtab_(symtab)
+ { }
+
+ void
+ operator()(Sized_symbol<size>* sym)
+ {
+ Mips_symbol<size>* mips_sym = Mips_symbol<size>::as_mips_sym(sym);
+ if (local_pic_function<size, big_endian>(mips_sym))
+ {
+ // SYM is a function that might need $25 to be valid on entry.
+ // If we're creating a non-PIC relocatable object, mark SYM as
+ // being PIC. If we're creating a non-relocatable object with
+ // non-PIC branches and jumps to SYM, make sure that SYM has an la25
+ // stub.
+ if (parameters->options().relocatable())
+ {
+ if (!parameters->options().output_is_position_independent())
+ mips_sym->set_pic();
+ }
+ else if (mips_sym->has_nonpic_branches())
+ {
+ this->target_->la25_stub_section(layout_)
+ ->create_la25_stub(this->symtab_, this->target_, mips_sym);
+ }
+ }
+ }
+
+ private:
+ Target_mips<size, big_endian>* target_;
+ Layout* layout_;
+ Symbol_table* symtab_;
+};
+
+template<int size, bool big_endian>
+class Target_mips : public Sized_target<size, big_endian>
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+ typedef Output_data_reloc<elfcpp::SHT_REL, true, size, big_endian>
+ Reloc_section;
+ typedef Output_data_reloc<elfcpp::SHT_RELA, true, size, big_endian>
+ Reloca_section;
+ typedef typename elfcpp::Swap<32, big_endian>::Valtype Valtype32;
+ typedef typename elfcpp::Swap<size, big_endian>::Valtype Valtype;
+
+ public:
+ Target_mips(const Target::Target_info* info = &mips_info)
+ : Sized_target<size, big_endian>(info), got_(NULL), gp_(NULL), plt_(NULL),
+ got_plt_(NULL), rel_dyn_(NULL), copy_relocs_(),
+ dyn_relocs_(), la25_stub_(NULL), mips_mach_extensions_(),
+ mips_stubs_(NULL), ei_class_(0), mach_(0), layout_(NULL),
+ got16_addends_(), entry_symbol_is_compressed_(false), insn32_(false)
+ {
+ this->add_machine_extensions();
+ }
+
+ // The offset of $gp from the beginning of the .got section.
+ static const unsigned int MIPS_GP_OFFSET = 0x7ff0;
+
+ // The maximum size of the GOT for it to be addressable using 16-bit
+ // offsets from $gp.
+ static const unsigned int MIPS_GOT_MAX_SIZE = MIPS_GP_OFFSET + 0x7fff;
+
+ // Make a new symbol table entry for the Mips target.
+ Sized_symbol<size>*
+ make_symbol() const
+ { return new Mips_symbol<size>(); }
+
+ // Process the relocations to determine unreferenced sections for
+ // garbage collection.
+ void
+ gc_process_relocs(Symbol_table* symtab,
+ Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ size_t local_symbol_count,
+ const unsigned char* plocal_symbols);
+
+ // Scan the relocations to look for symbol adjustments.
+ void
+ scan_relocs(Symbol_table* symtab,
+ Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ size_t local_symbol_count,
+ const unsigned char* plocal_symbols);
+
+ // Finalize the sections.
+ void
+ do_finalize_sections(Layout*, const Input_objects*, Symbol_table*);
+
+ // Relocate a section.
+ void
+ relocate_section(const Relocate_info<size, big_endian>*,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ unsigned char* view,
+ Mips_address view_address,
+ section_size_type view_size,
+ const Reloc_symbol_changes*);
+
+ // Scan the relocs during a relocatable link.
+ void
+ scan_relocatable_relocs(Symbol_table* symtab,
+ Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ size_t local_symbol_count,
+ const unsigned char* plocal_symbols,
+ Relocatable_relocs*);
+
+ // Emit relocations for a section.
+ void
+ relocate_relocs(const Relocate_info<size, big_endian>*,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ typename elfcpp::Elf_types<size>::Elf_Off
+ offset_in_output_section,
+ const Relocatable_relocs*,
+ unsigned char* view,
+ Mips_address view_address,
+ section_size_type view_size,
+ unsigned char* reloc_view,
+ section_size_type reloc_view_size);
+
+ // Perform target-specific processing in a relocatable link. This is
+ // only used if we use the relocation strategy RELOC_SPECIAL.
+ void
+ relocate_special_relocatable(const Relocate_info<size, big_endian>* relinfo,
+ unsigned int sh_type,
+ const unsigned char* preloc_in,
+ size_t relnum,
+ Output_section* output_section,
+ typename elfcpp::Elf_types<size>::Elf_Off
+ offset_in_output_section,
+ unsigned char* view,
+ Mips_address view_address,
+ section_size_type view_size,
+ unsigned char* preloc_out);
+
+ // Return whether SYM is defined by the ABI.
+ bool
+ do_is_defined_by_abi(const Symbol* sym) const
+ {
+ return ((strcmp(sym->name(), "__gnu_local_gp") == 0)
+ || (strcmp(sym->name(), "_gp_disp") == 0)
+ || (strcmp(sym->name(), "___tls_get_addr") == 0));
+ }
+
+ // Return the number of entries in the GOT.
+ unsigned int
+ got_entry_count() const
+ {
+ if (!this->has_got_section())
+ return 0;
+ return this->got_size() / (size/8);
+ }
+
+ // Return the number of entries in the PLT.
+ unsigned int
+ plt_entry_count() const
+ {
+ if (this->plt_ == NULL)
+ return 0;
+ return this->plt_->entry_count();
+ }
+
+ // Return the offset of the first non-reserved PLT entry.
+ unsigned int
+ first_plt_entry_offset() const
+ { return this->plt_->first_plt_entry_offset(); }
+
+ // Return the size of each PLT entry.
+ unsigned int
+ plt_entry_size() const
+ { return this->plt_->plt_entry_size(); }
+
+ // Get the GOT section, creating it if necessary.
+ Mips_output_data_got<size, big_endian>*
+ got_section(Symbol_table*, Layout*);
+
+ // Get the GOT section.
+ Mips_output_data_got<size, big_endian>*
+ got_section() const
+ {
+ gold_assert(this->got_ != NULL);
+ return this->got_;
+ }
+
+ // Get the .MIPS.stubs section, creating it if necessary.
+ Mips_output_data_mips_stubs<size, big_endian>*
+ mips_stubs_section(Layout* layout);
+
+ // Get the .MIPS.stubs section.
+ Mips_output_data_mips_stubs<size, big_endian>*
+ mips_stubs_section() const
+ {
+ gold_assert(this->mips_stubs_ != NULL);
+ return this->mips_stubs_;
+ }
+
+ // Get the LA25 stub section, creating it if necessary.
+ Mips_output_data_la25_stub<size, big_endian>*
+ la25_stub_section(Layout*);
+
+ // Get the LA25 stub section.
+ Mips_output_data_la25_stub<size, big_endian>*
+ la25_stub_section()
+ {
+ gold_assert(this->la25_stub_ != NULL);
+ return this->la25_stub_;
+ }
+
+ // Get gp value. It has the value of .got + 0x7FF0.
+ Mips_address
+ gp_value() const
+ {
+ if (this->gp_ != NULL)
+ return this->gp_->value();
+ return 0;
+ }
+
+ // Get gp value. It has the value of .got + 0x7FF0. Adjust it for
+ // multi-GOT links so that OBJECT's GOT + 0x7FF0 is returned.
+ Mips_address
+ adjusted_gp_value(const Mips_relobj<size, big_endian>* object)
+ {
+ if (this->gp_ == NULL)
+ return 0;
+
+ bool multi_got = false;
+ if (this->has_got_section())
+ multi_got = this->got_section()->multi_got();
+ if (!multi_got)
+ return this->gp_->value();
+ else
+ return this->gp_->value() + this->got_section()->get_got_offset(object);
+ }
+
+ // Get the dynamic reloc section, creating it if necessary.
+ Reloc_section*
+ rel_dyn_section(Layout*);
+
+ bool
+ do_has_custom_set_dynsym_indexes() const
+ { return true; }
+
+ // Don't emit input .reginfo sections to output .reginfo.
+ bool
+ do_should_include_section(elfcpp::Elf_Word sh_type) const
+ { return sh_type != elfcpp::SHT_MIPS_REGINFO; }
+
+ // Set the dynamic symbol indexes. INDEX is the index of the first
+ // global dynamic symbol. Pointers to the symbols are stored into the
+ // vector SYMS. The names are added to DYNPOOL. This returns an
+ // updated dynamic symbol index.
+ unsigned int
+ do_set_dynsym_indexes(std::vector<Symbol*>* dyn_symbols, unsigned int index,
+ std::vector<Symbol*>* syms, Stringpool* dynpool,
+ Versions* versions, Symbol_table* symtab) const;
+
+ // Remove .MIPS.stubs entry for a symbol.
+ void
+ remove_lazy_stub_entry(Mips_symbol<size>* sym)
+ {
+ if (this->mips_stubs_ != NULL)
+ this->mips_stubs_->remove_entry(sym);
+ }
+
+ // The value to write into got[1] for SVR4 targets, to identify it is
+ // a GNU object. The dynamic linker can then use got[1] to store the
+ // module pointer.
+ uint64_t
+ mips_elf_gnu_got1_mask()
+ {
+ if (this->is_output_n64())
+ return (uint64_t)1 << 63;
+ else
+ return 1 << 31;
+ }
+
+ // Whether the output has microMIPS code. This is valid only after
+ // merge_processor_specific_flags() is called.
+ bool
+ is_output_micromips() const
+ {
+ gold_assert(this->are_processor_specific_flags_set());
+ return elfcpp::is_micromips(this->processor_specific_flags());
+ }
+
+ // Whether the output uses N32 ABI. This is valid only after
+ // merge_processor_specific_flags() is called.
+ bool
+ is_output_n32() const
+ {
+ gold_assert(this->are_processor_specific_flags_set());
+ return elfcpp::abi_n32(this->processor_specific_flags());
+ }
+
+ // Whether the output uses N64 ABI. This is valid only after
+ // merge_processor_specific_flags() is called.
+ bool
+ is_output_n64() const
+ {
+ gold_assert(this->are_processor_specific_flags_set());
+ return elfcpp::abi_64(this->ei_class_);
+ }
+
+ // Whether the output uses NEWABI. This is valid only after
+ // merge_processor_specific_flags() is called.
+ bool
+ is_output_newabi() const
+ { return this->is_output_n32() || this->is_output_n64(); }
+
+ // Whether we can only use 32-bit microMIPS instructions.
+ bool
+ use_32bit_micromips_instructions() const
+ { return this->insn32_; }
+
+ protected:
+ // Return the value to use for a dynamic symbol which requires special
+ // treatment. This is how we support equality comparisons of function
+ // pointers across shared library boundaries, as described in the
+ // processor specific ABI supplement.
+ uint64_t
+ do_dynsym_value(const Symbol* gsym) const;
+
+ // Make an ELF object.
+ Object*
+ do_make_elf_object(const std::string&, Input_file*, off_t,
+ const elfcpp::Ehdr<size, big_endian>& ehdr);
+
+ Object*
+ do_make_elf_object(const std::string&, Input_file*, off_t,
+ const elfcpp::Ehdr<size, !big_endian>&)
+ { gold_unreachable(); }
+
+ // Make an output section.
+ Output_section*
+ do_make_output_section(const char* name, elfcpp::Elf_Word type,
+ elfcpp::Elf_Xword flags)
+ {
+ if (type == elfcpp::SHT_MIPS_REGINFO)
+ return new Mips_output_section_reginfo<size, big_endian>(name, type,
+ flags, this);
+ else
+ return new Output_section(name, type, flags);
+ }
+
+ // Adjust ELF file header.
+ void
+ do_adjust_elf_header(unsigned char* view, int len);
+
+ // Get the custom dynamic tag value.
+ unsigned int
+ do_dynamic_tag_custom_value(elfcpp::DT) const;
+
+ // Adjust the value written to the dynamic symbol table.
+ virtual void
+ do_adjust_dyn_symbol(const Symbol* sym, unsigned char* view) const
+ {
+ elfcpp::Sym<size, big_endian> isym(view);
+ elfcpp::Sym_write<size, big_endian> osym(view);
+ const Mips_symbol<size>* mips_sym = Mips_symbol<size>::as_mips_sym(sym);
+
+ // Keep dynamic compressed symbols odd. This allows the dynamic linker
+ // to treat compressed symbols like any other.
+ Mips_address value = isym.get_st_value();
+ if (mips_sym->is_mips16() && value != 0)
+ {
+ if (!mips_sym->has_mips16_fn_stub())
+ value |= 1;
+ else
+ {
+ // If we have a MIPS16 function with a stub, the dynamic symbol
+ // must refer to the stub, since only the stub uses the standard
+ // calling conventions. Stub contains MIPS32 code, so don't add +1
+ // in this case.
+
+ // There is a code which does this in the method
+ // Target_mips::do_dynsym_value, but that code will only be
+ // executed if the symbol is from dynobj.
+ // TODO(sasa): GNU ld also changes the value in non-dynamic symbol
+ // table.
+
+ Mips16_stub_section<size, big_endian>* fn_stub =
+ mips_sym->template get_mips16_fn_stub<big_endian>();
+ value = fn_stub->output_address();
+ osym.put_st_size(fn_stub->section_size());
+ }
+
+ osym.put_st_value(value);
+ osym.put_st_other(elfcpp::elf_st_other(sym->visibility(),
+ mips_sym->nonvis() - (elfcpp::STO_MIPS16 >> 2)));
+ }
+ else if ((mips_sym->is_micromips()
+ // Stubs are always microMIPS if there is any microMIPS code in
+ // the output.
+ || (this->is_output_micromips() && mips_sym->has_lazy_stub()))
+ && value != 0)
+ {
+ osym.put_st_value(value | 1);
+ osym.put_st_other(elfcpp::elf_st_other(sym->visibility(),
+ mips_sym->nonvis() - (elfcpp::STO_MICROMIPS >> 2)));
+ }
+ }
+
+ private:
+ // The class which scans relocations.
+ class Scan
+ {
+ public:
+ Scan()
+ { }
+
+ static inline int
+ get_reference_flags(unsigned int r_type);
+
+ inline void
+ local(Symbol_table* symtab, Layout* layout, Target_mips* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rel<size, big_endian>& reloc, unsigned int r_type,
+ const elfcpp::Sym<size, big_endian>& lsym,
+ bool is_discarded);
+
+ inline void
+ local(Symbol_table* symtab, Layout* layout, Target_mips* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>& reloc, unsigned int r_type,
+ const elfcpp::Sym<size, big_endian>& lsym,
+ bool is_discarded);
+
+ inline void
+ local(Symbol_table* symtab, Layout* layout, Target_mips* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>* rela,
+ const elfcpp::Rel<size, big_endian>* rel,
+ unsigned int rel_type,
+ unsigned int r_type,
+ const elfcpp::Sym<size, big_endian>& lsym,
+ bool is_discarded);
+
+ inline void
+ global(Symbol_table* symtab, Layout* layout, Target_mips* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rel<size, big_endian>& reloc, unsigned int r_type,
+ Symbol* gsym);
+
+ inline void
+ global(Symbol_table* symtab, Layout* layout, Target_mips* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>& reloc, unsigned int r_type,
+ Symbol* gsym);
+
+ inline void
+ global(Symbol_table* symtab, Layout* layout, Target_mips* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>* rela,
+ const elfcpp::Rel<size, big_endian>* rel,
+ unsigned int rel_type,
+ unsigned int r_type,
+ Symbol* gsym);
+
+ inline bool
+ local_reloc_may_be_function_pointer(Symbol_table* , Layout*,
+ Target_mips*,
+ Sized_relobj_file<size, big_endian>*,
+ unsigned int,
+ Output_section*,
+ const elfcpp::Rel<size, big_endian>&,
+ unsigned int,
+ const elfcpp::Sym<size, big_endian>&)
+ { return false; }
+
+ inline bool
+ global_reloc_may_be_function_pointer(Symbol_table*, Layout*,
+ Target_mips*,
+ Sized_relobj_file<size, big_endian>*,
+ unsigned int,
+ Output_section*,
+ const elfcpp::Rel<size, big_endian>&,
+ unsigned int, Symbol*)
+ { return false; }
+
+ inline bool
+ local_reloc_may_be_function_pointer(Symbol_table*, Layout*,
+ Target_mips*,
+ Sized_relobj_file<size, big_endian>*,
+ unsigned int,
+ Output_section*,
+ const elfcpp::Rela<size, big_endian>&,
+ unsigned int,
+ const elfcpp::Sym<size, big_endian>&)
+ { return false; }
+
+ inline bool
+ global_reloc_may_be_function_pointer(Symbol_table*, Layout*,
+ Target_mips*,
+ Sized_relobj_file<size, big_endian>*,
+ unsigned int,
+ Output_section*,
+ const elfcpp::Rela<size, big_endian>&,
+ unsigned int, Symbol*)
+ { return false; }
+ private:
+ static void
+ unsupported_reloc_local(Sized_relobj_file<size, big_endian>*,
+ unsigned int r_type);
+
+ static void
+ unsupported_reloc_global(Sized_relobj_file<size, big_endian>*,
+ unsigned int r_type, Symbol*);
+ };
+
+ // The class which implements relocation.
+ class Relocate
+ {
+ public:
+ Relocate()
+ { }
+
+ ~Relocate()
+ { }
+
+ // Return whether the R_MIPS_32 relocation needs to be applied.
+ inline bool
+ should_apply_r_mips_32_reloc(const Mips_symbol<size>* gsym,
+ unsigned int r_type,
+ Output_section* output_section,
+ Target_mips* target);
+
+ // Do a relocation. Return false if the caller should not issue
+ // any warnings about this relocation.
+ inline bool
+ relocate(const Relocate_info<size, big_endian>*, Target_mips*,
+ Output_section*, size_t relnum,
+ const elfcpp::Rela<size, big_endian>*,
+ const elfcpp::Rel<size, big_endian>*,
+ unsigned int,
+ unsigned int, const Sized_symbol<size>*,
+ const Symbol_value<size>*,
+ unsigned char*,
+ Mips_address,
+ section_size_type);
+
+ inline bool
+ relocate(const Relocate_info<size, big_endian>*, Target_mips*,
+ Output_section*, size_t relnum,
+ const elfcpp::Rel<size, big_endian>&,
+ unsigned int, const Sized_symbol<size>*,
+ const Symbol_value<size>*,
+ unsigned char*,
+ Mips_address,
+ section_size_type);
+
+ inline bool
+ relocate(const Relocate_info<size, big_endian>*, Target_mips*,
+ Output_section*, size_t relnum,
+ const elfcpp::Rela<size, big_endian>&,
+ unsigned int, const Sized_symbol<size>*,
+ const Symbol_value<size>*,
+ unsigned char*,
+ Mips_address,
+ section_size_type);
+ };
+
+ // A class which returns the size required for a relocation type,
+ // used while scanning relocs during a relocatable link.
+ class Relocatable_size_for_reloc
+ {
+ public:
+ unsigned int
+ get_size_for_reloc(unsigned int, Relobj*);
+ };
+
+ // This POD class holds the dynamic relocations that should be emitted instead
+ // of R_MIPS_32, R_MIPS_REL32 and R_MIPS_64 relocations. We will emit these
+ // relocations if it turns out that the symbol does not have static
+ // relocations.
+ class Dyn_reloc
+ {
+ public:
+ Dyn_reloc(Mips_symbol<size>* sym, unsigned int r_type,
+ Mips_relobj<size, big_endian>* relobj, unsigned int shndx,
+ Output_section* output_section, Mips_address r_offset)
+ : sym_(sym), r_type_(r_type), relobj_(relobj),
+ shndx_(shndx), output_section_(output_section),
+ r_offset_(r_offset)
+ { }
+
+ // Emit this reloc if appropriate. This is called after we have
+ // scanned all the relocations, so we know whether the symbol has
+ // static relocations.
+ void
+ emit(Reloc_section* rel_dyn, Mips_output_data_got<size, big_endian>* got,
+ Symbol_table* symtab)
+ {
+ if (!this->sym_->has_static_relocs())
+ {
+ got->record_global_got_symbol(this->sym_, this->relobj_,
+ this->r_type_, true, false);
+ if (!symbol_references_local(this->sym_,
+ this->sym_->should_add_dynsym_entry(symtab)))
+ rel_dyn->add_global(this->sym_, this->r_type_,
+ this->output_section_, this->relobj_,
+ this->shndx_, this->r_offset_);
+ else
+ rel_dyn->add_symbolless_global_addend(this->sym_, this->r_type_,
+ this->output_section_, this->relobj_,
+ this->shndx_, this->r_offset_);
+ }
+ }
+
+ private:
+ Mips_symbol<size>* sym_;
+ unsigned int r_type_;
+ Mips_relobj<size, big_endian>* relobj_;
+ unsigned int shndx_;
+ Output_section* output_section_;
+ Mips_address r_offset_;
+ };
+
+ // Adjust TLS relocation type based on the options and whether this
+ // is a local symbol.
+ static tls::Tls_optimization
+ optimize_tls_reloc(bool is_final, int r_type);
+
+ // Return whether there is a GOT section.
+ bool
+ has_got_section() const
+ { return this->got_ != NULL; }
+
+ // Check whether the given ELF header flags describe a 32-bit binary.
+ bool
+ mips_32bit_flags(elfcpp::Elf_Word);
+
+ enum Mips_mach {
+ mach_mips3000 = 3000,
+ mach_mips3900 = 3900,
+ mach_mips4000 = 4000,
+ mach_mips4010 = 4010,
+ mach_mips4100 = 4100,
+ mach_mips4111 = 4111,
+ mach_mips4120 = 4120,
+ mach_mips4300 = 4300,
+ mach_mips4400 = 4400,
+ mach_mips4600 = 4600,
+ mach_mips4650 = 4650,
+ mach_mips5000 = 5000,
+ mach_mips5400 = 5400,
+ mach_mips5500 = 5500,
+ mach_mips6000 = 6000,
+ mach_mips7000 = 7000,
+ mach_mips8000 = 8000,
+ mach_mips9000 = 9000,
+ mach_mips10000 = 10000,
+ mach_mips12000 = 12000,
+ mach_mips14000 = 14000,
+ mach_mips16000 = 16000,
+ mach_mips16 = 16,
+ mach_mips5 = 5,
+ mach_mips_loongson_2e = 3001,
+ mach_mips_loongson_2f = 3002,
+ mach_mips_loongson_3a = 3003,
+ mach_mips_sb1 = 12310201, // octal 'SB', 01
+ mach_mips_octeon = 6501,
+ mach_mips_octeonp = 6601,
+ mach_mips_octeon2 = 6502,
+ mach_mips_xlr = 887682, // decimal 'XLR'
+ mach_mipsisa32 = 32,
+ mach_mipsisa32r2 = 33,
+ mach_mipsisa64 = 64,
+ mach_mipsisa64r2 = 65,
+ mach_mips_micromips = 96
+ };
+
+ // Return the MACH for a MIPS e_flags value.
+ unsigned int
+ elf_mips_mach(elfcpp::Elf_Word);
+
+ // Check whether machine EXTENSION is an extension of machine BASE.
+ bool
+ mips_mach_extends(unsigned int, unsigned int);
+
+ // Merge processor specific flags.
+ void
+ merge_processor_specific_flags(const std::string&, elfcpp::Elf_Word,
+ unsigned char, bool);
+
+ // True if we are linking for CPUs that are faster if JAL is converted to BAL.
+ static inline bool
+ jal_to_bal()
+ { return false; }
+
+ // True if we are linking for CPUs that are faster if JALR is converted to
+ // BAL. This should be safe for all architectures. We enable this predicate
+ // for all CPUs.
+ static inline bool
+ jalr_to_bal()
+ { return true; }
+
+ // True if we are linking for CPUs that are faster if JR is converted to B.
+ // This should be safe for all architectures. We enable this predicate for
+ // all CPUs.
+ static inline bool
+ jr_to_b()
+ { return true; }
+
+ // Return the size of the GOT section.
+ section_size_type
+ got_size() const
+ {
+ gold_assert(this->got_ != NULL);
+ return this->got_->data_size();
+ }
+
+ // Create a PLT entry for a global symbol referenced by r_type relocation.
+ void
+ make_plt_entry(Symbol_table*, Layout*, Mips_symbol<size>*,
+ unsigned int r_type);
+
+ // Get the PLT section.
+ Mips_output_data_plt<size, big_endian>*
+ plt_section() const
+ {
+ gold_assert(this->plt_ != NULL);
+ return this->plt_;
+ }
+
+ // Get the GOT PLT section.
+ const Mips_output_data_plt<size, big_endian>*
+ got_plt_section() const
+ {
+ gold_assert(this->got_plt_ != NULL);
+ return this->got_plt_;
+ }
+
+ // Copy a relocation against a global symbol.
+ void
+ copy_reloc(Symbol_table* symtab, Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int shndx, Output_section* output_section,
+ Symbol* sym, const elfcpp::Rel<size, big_endian>& reloc)
+ {
+ this->copy_relocs_.copy_reloc(symtab, layout,
+ symtab->get_sized_symbol<size>(sym),
+ object, shndx, output_section,
+ reloc, this->rel_dyn_section(layout));
+ }
+
+ void
+ dynamic_reloc(Mips_symbol<size>* sym, unsigned int r_type,
+ Mips_relobj<size, big_endian>* relobj,
+ unsigned int shndx, Output_section* output_section,
+ Mips_address r_offset)
+ {
+ this->dyn_relocs_.push_back(Dyn_reloc(sym, r_type, relobj, shndx,
+ output_section, r_offset));
+ }
+
+ // Calculate value of _gp symbol.
+ void
+ set_gp(Layout*, Symbol_table*);
+
+ const char*
+ elf_mips_abi_name(elfcpp::Elf_Word e_flags, unsigned char ei_class);
+ const char*
+ elf_mips_mach_name(elfcpp::Elf_Word e_flags);
+
+ // Adds entries that describe how machines relate to one another. The entries
+ // are ordered topologically with MIPS I extensions listed last. First
+ // element is extension, second element is base.
+ void
+ add_machine_extensions()
+ {
+ // MIPS64r2 extensions.
+ this->add_extension(mach_mips_octeon2, mach_mips_octeonp);
+ this->add_extension(mach_mips_octeonp, mach_mips_octeon);
+ this->add_extension(mach_mips_octeon, mach_mipsisa64r2);
+
+ // MIPS64 extensions.
+ this->add_extension(mach_mipsisa64r2, mach_mipsisa64);
+ this->add_extension(mach_mips_sb1, mach_mipsisa64);
+ this->add_extension(mach_mips_xlr, mach_mipsisa64);
+ this->add_extension(mach_mips_loongson_3a, mach_mipsisa64);
+
+ // MIPS V extensions.
+ this->add_extension(mach_mipsisa64, mach_mips5);
+
+ // R10000 extensions.
+ this->add_extension(mach_mips12000, mach_mips10000);
+ this->add_extension(mach_mips14000, mach_mips10000);
+ this->add_extension(mach_mips16000, mach_mips10000);
+
+ // R5000 extensions. Note: the vr5500 ISA is an extension of the core
+ // vr5400 ISA, but doesn't include the multimedia stuff. It seems
+ // better to allow vr5400 and vr5500 code to be merged anyway, since
+ // many libraries will just use the core ISA. Perhaps we could add
+ // some sort of ASE flag if this ever proves a problem.
+ this->add_extension(mach_mips5500, mach_mips5400);
+ this->add_extension(mach_mips5400, mach_mips5000);
+
+ // MIPS IV extensions.
+ this->add_extension(mach_mips5, mach_mips8000);
+ this->add_extension(mach_mips10000, mach_mips8000);
+ this->add_extension(mach_mips5000, mach_mips8000);
+ this->add_extension(mach_mips7000, mach_mips8000);
+ this->add_extension(mach_mips9000, mach_mips8000);
+
+ // VR4100 extensions.
+ this->add_extension(mach_mips4120, mach_mips4100);
+ this->add_extension(mach_mips4111, mach_mips4100);
+
+ // MIPS III extensions.
+ this->add_extension(mach_mips_loongson_2e, mach_mips4000);
+ this->add_extension(mach_mips_loongson_2f, mach_mips4000);
+ this->add_extension(mach_mips8000, mach_mips4000);
+ this->add_extension(mach_mips4650, mach_mips4000);
+ this->add_extension(mach_mips4600, mach_mips4000);
+ this->add_extension(mach_mips4400, mach_mips4000);
+ this->add_extension(mach_mips4300, mach_mips4000);
+ this->add_extension(mach_mips4100, mach_mips4000);
+ this->add_extension(mach_mips4010, mach_mips4000);
+
+ // MIPS32 extensions.
+ this->add_extension(mach_mipsisa32r2, mach_mipsisa32);
+
+ // MIPS II extensions.
+ this->add_extension(mach_mips4000, mach_mips6000);
+ this->add_extension(mach_mipsisa32, mach_mips6000);
+
+ // MIPS I extensions.
+ this->add_extension(mach_mips6000, mach_mips3000);
+ this->add_extension(mach_mips3900, mach_mips3000);
+ }
+
+ // Add value to MIPS extenstions.
+ void
+ add_extension(unsigned int base, unsigned int extension)
+ {
+ std::pair<unsigned int, unsigned int> ext(base, extension);
+ this->mips_mach_extensions_.push_back(ext);
+ }
+
+ // Return the number of entries in the .dynsym section.
+ unsigned int get_dt_mips_symtabno() const
+ {
+ return ((unsigned int)(this->layout_->dynsym_section()->data_size()
+ / elfcpp::Elf_sizes<size>::sym_size));
+ // TODO(sasa): Entry size is MIPS_ELF_SYM_SIZE.
+ }
+
+ // Information about this specific target which we pass to the
+ // general Target structure.
+ static Target::Target_info mips_info;
+ // The GOT section.
+ Mips_output_data_got<size, big_endian>* got_;
+ // gp symbol. It has the value of .got + 0x7FF0.
+ Sized_symbol<size>* gp_;
+ // The PLT section.
+ Mips_output_data_plt<size, big_endian>* plt_;
+ // The GOT PLT section.
+ Output_data_space* got_plt_;
+ // The dynamic reloc section.
+ Reloc_section* rel_dyn_;
+ // Relocs saved to avoid a COPY reloc.
+ Mips_copy_relocs<elfcpp::SHT_REL, size, big_endian> copy_relocs_;
+
+ // A list of dyn relocs to be saved.
+ std::vector<Dyn_reloc> dyn_relocs_;
+
+ // The LA25 stub section.
+ Mips_output_data_la25_stub<size, big_endian>* la25_stub_;
+ // Architecture extensions.
+ std::vector<std::pair<unsigned int, unsigned int> > mips_mach_extensions_;
+ // .MIPS.stubs
+ Mips_output_data_mips_stubs<size, big_endian>* mips_stubs_;
+
+ unsigned char ei_class_;
+ unsigned int mach_;
+ Layout* layout_;
+
+ typename std::list<got16_addend<size, big_endian> > got16_addends_;
+
+ // Whether the entry symbol is mips16 or micromips.
+ bool entry_symbol_is_compressed_;
+
+ // Whether we can use only 32-bit microMIPS instructions.
+ // TODO(sasa): This should be a linker option.
+ bool insn32_;
+};
+
+
+// Helper structure for R_MIPS*_HI16/LO16 and R_MIPS*_GOT16/LO16 relocations.
+// It records high part of the relocation pair.
+
+template<int size, bool big_endian>
+struct reloc_high
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+
+ reloc_high(unsigned char* _view, const Mips_relobj<size, big_endian>* _object,
+ const Symbol_value<size>* _psymval, Mips_address _addend,
+ unsigned int _r_type, bool _extract_addend,
+ Mips_address _address = 0, bool _gp_disp = false)
+ : view(_view), object(_object), psymval(_psymval), addend(_addend),
+ r_type(_r_type), extract_addend(_extract_addend), address(_address),
+ gp_disp(_gp_disp)
+ { }
+
+ unsigned char* view;
+ const Mips_relobj<size, big_endian>* object;
+ const Symbol_value<size>* psymval;
+ Mips_address addend;
+ unsigned int r_type;
+ bool extract_addend;
+ Mips_address address;
+ bool gp_disp;
+};
+
+template<int size, bool big_endian>
+class Mips_relocate_functions : public Relocate_functions<size, big_endian>
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+ typedef typename elfcpp::Swap<16, big_endian>::Valtype Valtype16;
+ typedef typename elfcpp::Swap<32, big_endian>::Valtype Valtype32;
+
+ public:
+ typedef enum
+ {
+ STATUS_OKAY, // No error during relocation.
+ STATUS_OVERFLOW, // Relocation overflow.
+ STATUS_BAD_RELOC // Relocation cannot be applied.
+ } Status;
+
+ private:
+ typedef Relocate_functions<size, big_endian> Base;
+ typedef Mips_relocate_functions<size, big_endian> This;
+
+ static typename std::list<reloc_high<size, big_endian> > hi16_relocs;
+ static typename std::list<reloc_high<size, big_endian> > got16_relocs;
+
+ // R_MIPS16_26 is used for the mips16 jal and jalx instructions.
+ // Most mips16 instructions are 16 bits, but these instructions
+ // are 32 bits.
+ //
+ // The format of these instructions is:
+ //
+ // +--------------+--------------------------------+
+ // | JALX | X| Imm 20:16 | Imm 25:21 |
+ // +--------------+--------------------------------+
+ // | Immediate 15:0 |
+ // +-----------------------------------------------+
+ //
+ // JALX is the 5-bit value 00011. X is 0 for jal, 1 for jalx.
+ // Note that the immediate value in the first word is swapped.
+ //
+ // When producing a relocatable object file, R_MIPS16_26 is
+ // handled mostly like R_MIPS_26. In particular, the addend is
+ // stored as a straight 26-bit value in a 32-bit instruction.
+ // (gas makes life simpler for itself by never adjusting a
+ // R_MIPS16_26 reloc to be against a section, so the addend is
+ // always zero). However, the 32 bit instruction is stored as 2
+ // 16-bit values, rather than a single 32-bit value. In a
+ // big-endian file, the result is the same; in a little-endian
+ // file, the two 16-bit halves of the 32 bit value are swapped.
+ // This is so that a disassembler can recognize the jal
+ // instruction.
+ //
+ // When doing a final link, R_MIPS16_26 is treated as a 32 bit
+ // instruction stored as two 16-bit values. The addend A is the
+ // contents of the targ26 field. The calculation is the same as
+ // R_MIPS_26. When storing the calculated value, reorder the
+ // immediate value as shown above, and don't forget to store the
+ // value as two 16-bit values.
+ //
+ // To put it in MIPS ABI terms, the relocation field is T-targ26-16,
+ // defined as
+ //
+ // big-endian:
+ // +--------+----------------------+
+ // | | |
+ // | | targ26-16 |
+ // |31 26|25 0|
+ // +--------+----------------------+
+ //
+ // little-endian:
+ // +----------+------+-------------+
+ // | | | |
+ // | sub1 | | sub2 |
+ // |0 9|10 15|16 31|
+ // +----------+--------------------+
+ // where targ26-16 is sub1 followed by sub2 (i.e., the addend field A is
+ // ((sub1 << 16) | sub2)).
+ //
+ // When producing a relocatable object file, the calculation is
+ // (((A < 2) | ((P + 4) & 0xf0000000) + S) >> 2)
+ // When producing a fully linked file, the calculation is
+ // let R = (((A < 2) | ((P + 4) & 0xf0000000) + S) >> 2)
+ // ((R & 0x1f0000) << 5) | ((R & 0x3e00000) >> 5) | (R & 0xffff)
+ //
+ // The table below lists the other MIPS16 instruction relocations.
+ // Each one is calculated in the same way as the non-MIPS16 relocation
+ // given on the right, but using the extended MIPS16 layout of 16-bit
+ // immediate fields:
+ //
+ // R_MIPS16_GPREL R_MIPS_GPREL16
+ // R_MIPS16_GOT16 R_MIPS_GOT16
+ // R_MIPS16_CALL16 R_MIPS_CALL16
+ // R_MIPS16_HI16 R_MIPS_HI16
+ // R_MIPS16_LO16 R_MIPS_LO16
+ //
+ // A typical instruction will have a format like this:
+ //
+ // +--------------+--------------------------------+
+ // | EXTEND | Imm 10:5 | Imm 15:11 |
+ // +--------------+--------------------------------+
+ // | Major | rx | ry | Imm 4:0 |
+ // +--------------+--------------------------------+
+ //
+ // EXTEND is the five bit value 11110. Major is the instruction
+ // opcode.
+ //
+ // All we need to do here is shuffle the bits appropriately.
+ // As above, the two 16-bit halves must be swapped on a
+ // little-endian system.
+
+ // Similar to MIPS16, the two 16-bit halves in microMIPS must be swapped
+ // on a little-endian system. This does not apply to R_MICROMIPS_PC7_S1
+ // and R_MICROMIPS_PC10_S1 relocs that apply to 16-bit instructions.
+
+ static inline bool
+ should_shuffle_micromips_reloc(unsigned int r_type)
+ {
+ return (micromips_reloc(r_type)
+ && r_type != elfcpp::R_MICROMIPS_PC7_S1
+ && r_type != elfcpp::R_MICROMIPS_PC10_S1);
+ }
+
+ static void
+ mips_reloc_unshuffle(unsigned char* view, unsigned int r_type,
+ bool jal_shuffle)
+ {
+ if (!mips16_reloc(r_type)
+ && !should_shuffle_micromips_reloc(r_type))
+ return;
+
+ // Pick up the first and second halfwords of the instruction.
+ Valtype16 first = elfcpp::Swap<16, big_endian>::readval(view);
+ Valtype16 second = elfcpp::Swap<16, big_endian>::readval(view + 2);
+ Valtype32 val;
+
+ if (micromips_reloc(r_type)
+ || (r_type == elfcpp::R_MIPS16_26 && !jal_shuffle))
+ val = first << 16 | second;
+ else if (r_type != elfcpp::R_MIPS16_26)
+ val = (((first & 0xf800) << 16) | ((second & 0xffe0) << 11)
+ | ((first & 0x1f) << 11) | (first & 0x7e0) | (second & 0x1f));
+ else
+ val = (((first & 0xfc00) << 16) | ((first & 0x3e0) << 11)
+ | ((first & 0x1f) << 21) | second);
+
+ elfcpp::Swap<32, big_endian>::writeval(view, val);
+ }
+
+ static void
+ mips_reloc_shuffle(unsigned char* view, unsigned int r_type, bool jal_shuffle)
+ {
+ if (!mips16_reloc(r_type)
+ && !should_shuffle_micromips_reloc(r_type))
+ return;
+
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(view);
+ Valtype16 first, second;
+
+ if (micromips_reloc(r_type)
+ || (r_type == elfcpp::R_MIPS16_26 && !jal_shuffle))
+ {
+ second = val & 0xffff;
+ first = val >> 16;
+ }
+ else if (r_type != elfcpp::R_MIPS16_26)
+ {
+ second = ((val >> 11) & 0xffe0) | (val & 0x1f);
+ first = ((val >> 16) & 0xf800) | ((val >> 11) & 0x1f) | (val & 0x7e0);
+ }
+ else
+ {
+ second = val & 0xffff;
+ first = ((val >> 16) & 0xfc00) | ((val >> 11) & 0x3e0)
+ | ((val >> 21) & 0x1f);
+ }
+
+ elfcpp::Swap<16, big_endian>::writeval(view + 2, second);
+ elfcpp::Swap<16, big_endian>::writeval(view, first);
+ }
+
+ public:
+ // R_MIPS_16: S + sign-extend(A)
+ static inline typename This::Status
+ rel16(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype16* wv = reinterpret_cast<Valtype16*>(view);
+ Valtype16 val = elfcpp::Swap<16, big_endian>::readval(wv);
+
+ Valtype32 addend = (extract_addend ? Bits<16>::sign_extend32(val)
+ : Bits<16>::sign_extend32(addend_a));
+
+ Valtype32 x = psymval->value(object, addend);
+ val = Bits<16>::bit_select32(val, x, 0xffffU);
+ elfcpp::Swap<16, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<16>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_32: S + A
+ static inline typename This::Status
+ rel32(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 addend = (extract_addend
+ ? elfcpp::Swap<32, big_endian>::readval(wv)
+ : Bits<32>::sign_extend32(addend_a));
+ Valtype32 x = psymval->value(object, addend);
+ elfcpp::Swap<32, big_endian>::writeval(wv, x);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_JALR, R_MICROMIPS_JALR
+ static inline typename This::Status
+ reljalr(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ Mips_address addend_a, bool extract_addend, bool cross_mode_jump,
+ unsigned int r_type, bool jalr_to_bal, bool jr_to_b)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 addend = extract_addend ? 0 : addend_a;
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ // Try converting J(AL)R to B(AL), if the target is in range.
+ if (!parameters->options().relocatable()
+ && r_type == elfcpp::R_MIPS_JALR
+ && !cross_mode_jump
+ && ((jalr_to_bal && val == 0x0320f809) // jalr t9
+ || (jr_to_b && val == 0x03200008))) // jr t9
+ {
+ int offset = psymval->value(object, addend) - (address + 4);
+ if (!Bits<18>::has_overflow32(offset))
+ {
+ if (val == 0x03200008) // jr t9
+ val = 0x10000000 | (((Valtype32)offset >> 2) & 0xffff); // b addr
+ else
+ val = 0x04110000 | (((Valtype32)offset >> 2) & 0xffff); //bal addr
+ }
+ }
+
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_PC32: S + A - P
+ static inline typename This::Status
+ relpc32(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ Mips_address addend_a, bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 addend = (extract_addend
+ ? elfcpp::Swap<32, big_endian>::readval(wv)
+ : Bits<32>::sign_extend32(addend_a));
+ Valtype32 x = psymval->value(object, addend) - address;
+ elfcpp::Swap<32, big_endian>::writeval(wv, x);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_26, R_MIPS16_26, R_MICROMIPS_26_S1
+ static inline typename This::Status
+ rel26(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ bool local, Mips_address addend_a, bool extract_addend,
+ const Symbol* gsym, bool cross_mode_jump, unsigned int r_type,
+ bool jal_to_bal)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend;
+ if (extract_addend)
+ {
+ if (r_type == elfcpp::R_MICROMIPS_26_S1)
+ addend = (val & 0x03ffffff) << 1;
+ else
+ addend = (val & 0x03ffffff) << 2;
+ }
+ else
+ addend = addend_a;
+
+ // Make sure the target of JALX is word-aligned. Bit 0 must be
+ // the correct ISA mode selector and bit 1 must be 0.
+ if (cross_mode_jump
+ && (psymval->value(object, 0) & 3) != (r_type == elfcpp::R_MIPS_26))
+ {
+ gold_warning(_("JALX to a non-word-aligned address"));
+ mips_reloc_shuffle(view, r_type, !parameters->options().relocatable());
+ return This::STATUS_BAD_RELOC;
+ }
+
+ // Shift is 2, unusually, for microMIPS JALX.
+ unsigned int shift =
+ (!cross_mode_jump && r_type == elfcpp::R_MICROMIPS_26_S1) ? 1 : 2;
+
+ Valtype32 x;
+ if (local)
+ x = addend | ((address + 4) & (0xfc000000 << shift));
+ else
+ {
+ if (shift == 1)
+ x = Bits<27>::sign_extend32(addend);
+ else
+ x = Bits<28>::sign_extend32(addend);
+ }
+ x = psymval->value(object, x) >> shift;
+
+ if (!local && !gsym->is_weak_undefined())
+ {
+ if ((x >> 26) != ((address + 4) >> (26 + shift)))
+ {
+ gold_error(_("relocation truncated to fit: %u against '%s'"),
+ r_type, gsym->name());
+ return This::STATUS_OVERFLOW;
+ }
+ }
+
+ val = Bits<32>::bit_select32(val, x, 0x03ffffff);
+
+ // If required, turn JAL into JALX.
+ if (cross_mode_jump)
+ {
+ bool ok;
+ Valtype32 opcode = val >> 26;
+ Valtype32 jalx_opcode;
+
+ // Check to see if the opcode is already JAL or JALX.
+ if (r_type == elfcpp::R_MIPS16_26)
+ {
+ ok = (opcode == 0x6) || (opcode == 0x7);
+ jalx_opcode = 0x7;
+ }
+ else if (r_type == elfcpp::R_MICROMIPS_26_S1)
+ {
+ ok = (opcode == 0x3d) || (opcode == 0x3c);
+ jalx_opcode = 0x3c;
+ }
+ else
+ {
+ ok = (opcode == 0x3) || (opcode == 0x1d);
+ jalx_opcode = 0x1d;
+ }
+
+ // If the opcode is not JAL or JALX, there's a problem. We cannot
+ // convert J or JALS to JALX.
+ if (!ok)
+ {
+ gold_error(_("Unsupported jump between ISA modes; consider "
+ "recompiling with interlinking enabled."));
+ return This::STATUS_BAD_RELOC;
+ }
+
+ // Make this the JALX opcode.
+ val = (val & ~(0x3f << 26)) | (jalx_opcode << 26);
+ }
+
+ // Try converting JAL to BAL, if the target is in range.
+ if (!parameters->options().relocatable()
+ && !cross_mode_jump
+ && ((jal_to_bal
+ && r_type == elfcpp::R_MIPS_26
+ && (val >> 26) == 0x3))) // jal addr
+ {
+ Valtype32 dest = (x << 2) | (((address + 4) >> 28) << 28);
+ int offset = dest - (address + 4);
+ if (!Bits<18>::has_overflow32(offset))
+ {
+ if (val == 0x03200008) // jr t9
+ val = 0x10000000 | (((Valtype32)offset >> 2) & 0xffff); // b addr
+ else
+ val = 0x04110000 | (((Valtype32)offset >> 2) & 0xffff); //bal addr
+ }
+ }
+
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, !parameters->options().relocatable());
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_PC16
+ static inline typename This::Status
+ relpc16(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ Mips_address addend_a, bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = extract_addend ? (val & 0xffff) << 2 : addend_a;
+ addend = Bits<18>::sign_extend32(addend);
+
+ Valtype32 x = psymval->value(object, addend) - address;
+ val = Bits<16>::bit_select32(val, x >> 2, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<18>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MICROMIPS_PC7_S1
+ static inline typename This::Status
+ relmicromips_pc7_s1(unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ Mips_address addend_a, bool extract_addend,
+ unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = extract_addend ? (val & 0x7f) << 1 : addend_a;
+ addend = Bits<8>::sign_extend32(addend);
+
+ Valtype32 x = psymval->value(object, addend) - address;
+ val = Bits<16>::bit_select32(val, x >> 1, 0x7f);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<8>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MICROMIPS_PC10_S1
+ static inline typename This::Status
+ relmicromips_pc10_s1(unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ Mips_address addend_a, bool extract_addend,
+ unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = extract_addend ? (val & 0x3ff) << 1 : addend_a;
+ addend = Bits<11>::sign_extend32(addend);
+
+ Valtype32 x = psymval->value(object, addend) - address;
+ val = Bits<16>::bit_select32(val, x >> 1, 0x3ff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<11>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MICROMIPS_PC16_S1
+ static inline typename This::Status
+ relmicromips_pc16_s1(unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ Mips_address addend_a, bool extract_addend,
+ unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = extract_addend ? (val & 0xffff) << 1 : addend_a;
+ addend = Bits<17>::sign_extend32(addend);
+
+ Valtype32 x = psymval->value(object, addend) - address;
+ val = Bits<16>::bit_select32(val, x >> 1, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<17>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_HI16, R_MIPS16_HI16, R_MICROMIPS_HI16,
+ static inline typename This::Status
+ relhi16(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend,
+ Mips_address address, bool gp_disp, unsigned int r_type,
+ bool extract_addend)
+ {
+ // Record the relocation. It will be resolved when we find lo16 part.
+ hi16_relocs.push_back(reloc_high<size, big_endian>(view, object, psymval,
+ addend, r_type, extract_addend, address, gp_disp));
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_HI16, R_MIPS16_HI16, R_MICROMIPS_HI16,
+ static inline typename This::Status
+ do_relhi16(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_hi,
+ Mips_address address, bool is_gp_disp, unsigned int r_type,
+ bool extract_addend, Valtype32 addend_lo,
+ Target_mips<size, big_endian>* target)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = (extract_addend ? ((val & 0xffff) << 16) + addend_lo
+ : addend_hi);
+
+ Valtype32 value;
+ if (!is_gp_disp)
+ value = psymval->value(object, addend);
+ else
+ {
+ // For MIPS16 ABI code we generate this sequence
+ // 0: li $v0,%hi(_gp_disp)
+ // 4: addiupc $v1,%lo(_gp_disp)
+ // 8: sll $v0,16
+ // 12: addu $v0,$v1
+ // 14: move $gp,$v0
+ // So the offsets of hi and lo relocs are the same, but the
+ // base $pc is that used by the ADDIUPC instruction at $t9 + 4.
+ // ADDIUPC clears the low two bits of the instruction address,
+ // so the base is ($t9 + 4) & ~3.
+ Valtype32 gp_disp;
+ if (r_type == elfcpp::R_MIPS16_HI16)
+ gp_disp = (target->adjusted_gp_value(object)
+ - ((address + 4) & ~0x3));
+ // The microMIPS .cpload sequence uses the same assembly
+ // instructions as the traditional psABI version, but the
+ // incoming $t9 has the low bit set.
+ else if (r_type == elfcpp::R_MICROMIPS_HI16)
+ gp_disp = target->adjusted_gp_value(object) - address - 1;
+ else
+ gp_disp = target->adjusted_gp_value(object) - address;
+ value = gp_disp + addend;
+ }
+ Valtype32 x = ((value + 0x8000) >> 16) & 0xffff;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (is_gp_disp && Bits<16>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_GOT16, R_MIPS16_GOT16, R_MICROMIPS_GOT16
+ static inline typename This::Status
+ relgot16_local(unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, unsigned int r_type)
+ {
+ // Record the relocation. It will be resolved when we find lo16 part.
+ got16_relocs.push_back(reloc_high<size, big_endian>(view, object, psymval,
+ addend_a, r_type, extract_addend));
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_GOT16, R_MIPS16_GOT16, R_MICROMIPS_GOT16
+ static inline typename This::Status
+ do_relgot16_local(unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_hi,
+ unsigned int r_type, bool extract_addend,
+ Valtype32 addend_lo, Target_mips<size, big_endian>* target)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = (extract_addend ? ((val & 0xffff) << 16) + addend_lo
+ : addend_hi);
+
+ // Find GOT page entry.
+ Mips_address value = ((psymval->value(object, addend) + 0x8000) >> 16)
+ & 0xffff;
+ value <<= 16;
+ unsigned int got_offset =
+ target->got_section()->get_got_page_offset(value, object);
+
+ // Resolve the relocation.
+ Valtype32 x = target->got_section()->gp_offset(got_offset, object);
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<16>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_LO16, R_MIPS16_LO16, R_MICROMIPS_LO16, R_MICROMIPS_HI0_LO16
+ static inline typename This::Status
+ rello16(Target_mips<size, big_endian>* target, unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, Mips_address address, bool is_gp_disp,
+ unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = (extract_addend ? Bits<16>::sign_extend32(val & 0xffff)
+ : addend_a);
+
+ // Resolve pending R_MIPS_HI16 relocations.
+ typename std::list<reloc_high<size, big_endian> >::iterator it =
+ hi16_relocs.begin();
+ while (it != hi16_relocs.end())
+ {
+ reloc_high<size, big_endian> hi16 = *it;
+ if (hi16.psymval->value(hi16.object, 0) == psymval->value(object, 0))
+ {
+ if (do_relhi16(hi16.view, hi16.object, hi16.psymval, hi16.addend,
+ hi16.address, hi16.gp_disp, hi16.r_type,
+ hi16.extract_addend, addend, target)
+ == This::STATUS_OVERFLOW)
+ return This::STATUS_OVERFLOW;
+ it = hi16_relocs.erase(it);
+ }
+ else
+ ++it;
+ }
+
+ // Resolve pending local R_MIPS_GOT16 relocations.
+ typename std::list<reloc_high<size, big_endian> >::iterator it2 =
+ got16_relocs.begin();
+ while (it2 != got16_relocs.end())
+ {
+ reloc_high<size, big_endian> got16 = *it2;
+ if (got16.psymval->value(got16.object, 0) == psymval->value(object, 0))
+ {
+ if (do_relgot16_local(got16.view, got16.object, got16.psymval,
+ got16.addend, got16.r_type,
+ got16.extract_addend, addend,
+ target) == This::STATUS_OVERFLOW)
+ return This::STATUS_OVERFLOW;
+ it2 = got16_relocs.erase(it2);
+ }
+ else
+ ++it2;
+ }
+
+ // Resolve R_MIPS_LO16 relocation.
+ Valtype32 x;
+ if (!is_gp_disp)
+ x = psymval->value(object, addend);
+ else
+ {
+ // See the comment for R_MIPS16_HI16 above for the reason
+ // for this conditional.
+ Valtype32 gp_disp;
+ if (r_type == elfcpp::R_MIPS16_LO16)
+ gp_disp = target->adjusted_gp_value(object) - (address & ~0x3);
+ else if (r_type == elfcpp::R_MICROMIPS_LO16
+ || r_type == elfcpp::R_MICROMIPS_HI0_LO16)
+ gp_disp = target->adjusted_gp_value(object) - address + 3;
+ else
+ gp_disp = target->adjusted_gp_value(object) - address + 4;
+ // The MIPS ABI requires checking the R_MIPS_LO16 relocation
+ // for overflow. Relocations against _gp_disp are normally
+ // generated from the .cpload pseudo-op. It generates code
+ // that normally looks like this:
+
+ // lui $gp,%hi(_gp_disp)
+ // addiu $gp,$gp,%lo(_gp_disp)
+ // addu $gp,$gp,$t9
+
+ // Here $t9 holds the address of the function being called,
+ // as required by the MIPS ELF ABI. The R_MIPS_LO16
+ // relocation can easily overflow in this situation, but the
+ // R_MIPS_HI16 relocation will handle the overflow.
+ // Therefore, we consider this a bug in the MIPS ABI, and do
+ // not check for overflow here.
+ x = gp_disp + addend;
+ }
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_CALL16, R_MIPS16_CALL16, R_MICROMIPS_CALL16
+ // R_MIPS_GOT16, R_MIPS16_GOT16, R_MICROMIPS_GOT16
+ // R_MIPS_TLS_GD, R_MIPS16_TLS_GD, R_MICROMIPS_TLS_GD
+ // R_MIPS_TLS_GOTTPREL, R_MIPS16_TLS_GOTTPREL, R_MICROMIPS_TLS_GOTTPREL
+ // R_MIPS_TLS_LDM, R_MIPS16_TLS_LDM, R_MICROMIPS_TLS_LDM
+ // R_MIPS_GOT_DISP, R_MICROMIPS_GOT_DISP
+ static inline typename This::Status
+ relgot(unsigned char* view, int gp_offset, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 x = gp_offset;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<16>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_GOT_PAGE, R_MICROMIPS_GOT_PAGE
+ static inline typename This::Status
+ relgotpage(Target_mips<size, big_endian>* target, unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(view);
+ Valtype32 addend = extract_addend ? val & 0xffff : addend_a;
+
+ // Find a GOT page entry that points to within 32KB of symbol + addend.
+ Mips_address value = (psymval->value(object, addend) + 0x8000) & ~0xffff;
+ unsigned int got_offset =
+ target->got_section()->get_got_page_offset(value, object);
+
+ Valtype32 x = target->got_section()->gp_offset(got_offset, object);
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<16>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_GOT_OFST, R_MICROMIPS_GOT_OFST
+ static inline typename This::Status
+ relgotofst(Target_mips<size, big_endian>* target, unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, bool local, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(view);
+ Valtype32 addend = extract_addend ? val & 0xffff : addend_a;
+
+ // For a local symbol, find a GOT page entry that points to within 32KB of
+ // symbol + addend. Relocation value is the offset of the GOT page entry's
+ // value from symbol + addend.
+ // For a global symbol, relocation value is addend.
+ Valtype32 x;
+ if (local)
+ {
+ // Find GOT page entry.
+ Mips_address value = ((psymval->value(object, addend) + 0x8000)
+ & ~0xffff);
+ target->got_section()->get_got_page_offset(value, object);
+
+ x = psymval->value(object, addend) - value;
+ }
+ else
+ x = addend;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<16>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_GOT_HI16, R_MIPS_CALL_HI16,
+ // R_MICROMIPS_GOT_HI16, R_MICROMIPS_CALL_HI16
+ static inline typename This::Status
+ relgot_hi16(unsigned char* view, int gp_offset, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 x = gp_offset;
+ x = ((x + 0x8000) >> 16) & 0xffff;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_GOT_LO16, R_MIPS_CALL_LO16,
+ // R_MICROMIPS_GOT_LO16, R_MICROMIPS_CALL_LO16
+ static inline typename This::Status
+ relgot_lo16(unsigned char* view, int gp_offset, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 x = gp_offset;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_GPREL16, R_MIPS16_GPREL, R_MIPS_LITERAL, R_MICROMIPS_LITERAL
+ // R_MICROMIPS_GPREL7_S2, R_MICROMIPS_GPREL16
+ static inline typename This::Status
+ relgprel(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address gp,
+ Mips_address addend_a, bool extract_addend, bool local,
+ unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend;
+ if (extract_addend)
+ {
+ if (r_type == elfcpp::R_MICROMIPS_GPREL7_S2)
+ addend = (val & 0x7f) << 2;
+ else
+ addend = val & 0xffff;
+ // Only sign-extend the addend if it was extracted from the
+ // instruction. If the addend was separate, leave it alone,
+ // otherwise we may lose significant bits.
+ addend = Bits<16>::sign_extend32(addend);
+ }
+ else
+ addend = addend_a;
+
+ Valtype32 x = psymval->value(object, addend) - gp;
+
+ // If the symbol was local, any earlier relocatable links will
+ // have adjusted its addend with the gp offset, so compensate
+ // for that now. Don't do it for symbols forced local in this
+ // link, though, since they won't have had the gp offset applied
+ // to them before.
+ if (local)
+ x += object->gp_value();
+
+ if (r_type == elfcpp::R_MICROMIPS_GPREL7_S2)
+ val = Bits<32>::bit_select32(val, x, 0x7f);
+ else
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ if (Bits<16>::has_overflow32(x))
+ {
+ gold_error(_("small-data section exceeds 64KB; lower small-data size "
+ "limit (see option -G)"));
+ return This::STATUS_OVERFLOW;
+ }
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_GPREL32
+ static inline typename This::Status
+ relgprel32(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address gp,
+ Mips_address addend_a, bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 addend = extract_addend ? val : addend_a;
+
+ // R_MIPS_GPREL32 relocations are defined for local symbols only.
+ Valtype32 x = psymval->value(object, addend) + object->gp_value() - gp;
+ elfcpp::Swap<32, big_endian>::writeval(wv, x);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_TLS_TPREL_HI16, R_MIPS16_TLS_TPREL_HI16, R_MICROMIPS_TLS_TPREL_HI16
+ // R_MIPS_TLS_DTPREL_HI16, R_MIPS16_TLS_DTPREL_HI16,
+ // R_MICROMIPS_TLS_DTPREL_HI16
+ static inline typename This::Status
+ tlsrelhi16(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Valtype32 tp_offset,
+ Mips_address addend_a, bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 addend = extract_addend ? val & 0xffff : addend_a;
+
+ // tls symbol values are relative to tls_segment()->vaddr()
+ Valtype32 x = ((psymval->value(object, addend) - tp_offset) + 0x8000) >> 16;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_TLS_TPREL_LO16, R_MIPS16_TLS_TPREL_LO16, R_MICROMIPS_TLS_TPREL_LO16,
+ // R_MIPS_TLS_DTPREL_LO16, R_MIPS16_TLS_DTPREL_LO16,
+ // R_MICROMIPS_TLS_DTPREL_LO16,
+ static inline typename This::Status
+ tlsrello16(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Valtype32 tp_offset,
+ Mips_address addend_a, bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 addend = extract_addend ? val & 0xffff : addend_a;
+
+ // tls symbol values are relative to tls_segment()->vaddr()
+ Valtype32 x = psymval->value(object, addend) - tp_offset;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_TLS_TPREL32, R_MIPS_TLS_TPREL64,
+ // R_MIPS_TLS_DTPREL32, R_MIPS_TLS_DTPREL64
+ static inline typename This::Status
+ tlsrel32(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Valtype32 tp_offset,
+ Mips_address addend_a, bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 addend = extract_addend ? val : addend_a;
+
+ // tls symbol values are relative to tls_segment()->vaddr()
+ Valtype32 x = psymval->value(object, addend) - tp_offset;
+ elfcpp::Swap<32, big_endian>::writeval(wv, x);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_SUB, R_MICROMIPS_SUB
+ static inline typename This::Status
+ relsub(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 addend = extract_addend ? val : addend_a;
+
+ Valtype32 x = psymval->value(object, -addend);
+ elfcpp::Swap<32, big_endian>::writeval(wv, x);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+};
+
+template<int size, bool big_endian>
+typename std::list<reloc_high<size, big_endian> >
+ Mips_relocate_functions<size, big_endian>::hi16_relocs;
+
+template<int size, bool big_endian>
+typename std::list<reloc_high<size, big_endian> >
+ Mips_relocate_functions<size, big_endian>::got16_relocs;
+
+// Mips_got_info methods.
+
+// Reserve GOT entry for a GOT relocation of type R_TYPE against symbol
+// SYMNDX + ADDEND, where SYMNDX is a local symbol in section SHNDX in OBJECT.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::record_local_got_symbol(
+ Mips_relobj<size, big_endian>* object, unsigned int symndx,
+ Mips_address addend, unsigned int r_type, unsigned int shndx)
+{
+ Mips_got_entry<size, big_endian>* entry =
+ new Mips_got_entry<size, big_endian>(object, symndx, addend,
+ mips_elf_reloc_tls_type(r_type),
+ shndx);
+ this->record_got_entry(entry, object);
+}
+
+// Reserve GOT entry for a GOT relocation of type R_TYPE against MIPS_SYM,
+// in OBJECT. FOR_CALL is true if the caller is only interested in
+// using the GOT entry for calls. DYN_RELOC is true if R_TYPE is a dynamic
+// relocation.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::record_global_got_symbol(
+ Mips_symbol<size>* mips_sym, Mips_relobj<size, big_endian>* object,
+ unsigned int r_type, bool dyn_reloc, bool for_call)
+{
+ if (!for_call)
+ mips_sym->set_got_not_only_for_calls();
+
+ // A global symbol in the GOT must also be in the dynamic symbol table.
+ if (!mips_sym->needs_dynsym_entry())
+ {
+ switch (mips_sym->visibility())
+ {
+ case elfcpp::STV_INTERNAL:
+ case elfcpp::STV_HIDDEN:
+ mips_sym->set_is_forced_local();
+ break;
+ default:
+ mips_sym->set_needs_dynsym_entry();
+ break;
+ }
+ }
+
+ unsigned char tls_type = mips_elf_reloc_tls_type(r_type);
+ if (tls_type == GOT_TLS_NONE)
+ this->global_got_symbols_.insert(mips_sym);
+
+ if (dyn_reloc)
+ {
+ if (mips_sym->global_got_area() == GGA_NONE)
+ mips_sym->set_global_got_area(GGA_RELOC_ONLY);
+ return;
+ }
+
+ Mips_got_entry<size, big_endian>* entry =
+ new Mips_got_entry<size, big_endian>(object, mips_sym, tls_type);
+
+ this->record_got_entry(entry, object);
+}
+
+// Add ENTRY to master GOT and to OBJECT's GOT.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::record_got_entry(
+ Mips_got_entry<size, big_endian>* entry,
+ Mips_relobj<size, big_endian>* object)
+{
+ if (this->got_entries_.find(entry) == this->got_entries_.end())
+ this->got_entries_.insert(entry);
+
+ // Create the GOT entry for the OBJECT's GOT.
+ Mips_got_info<size, big_endian>* g = object->get_or_create_got_info();
+ Mips_got_entry<size, big_endian>* entry2 =
+ new Mips_got_entry<size, big_endian>(*entry);
+
+ if (g->got_entries_.find(entry2) == g->got_entries_.end())
+ g->got_entries_.insert(entry2);
+}
+
+// Record that OBJECT has a page relocation against symbol SYMNDX and
+// that ADDEND is the addend for that relocation.
+// This function creates an upper bound on the number of GOT slots
+// required; no attempt is made to combine references to non-overridable
+// global symbols across multiple input files.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::record_got_page_entry(
+ Mips_relobj<size, big_endian>* object, unsigned int symndx, int addend)
+{
+ struct Got_page_range **range_ptr, *range;
+ int old_pages, new_pages;
+
+ // Find the Got_page_entry for this symbol.
+ Got_page_entry* entry = new Got_page_entry(object, symndx);
+ typename Got_page_entry_set::iterator it =
+ this->got_page_entries_.find(entry);
+ if (it != this->got_page_entries_.end())
+ entry = *it;
+ else
+ this->got_page_entries_.insert(entry);
+
+ // Add the same entry to the OBJECT's GOT.
+ Got_page_entry* entry2 = NULL;
+ Mips_got_info<size, big_endian>* g2 = object->get_or_create_got_info();
+ if (g2->got_page_entries_.find(entry) == g2->got_page_entries_.end())
+ {
+ entry2 = new Got_page_entry(*entry);
+ g2->got_page_entries_.insert(entry2);
+ }
+
+ // Skip over ranges whose maximum extent cannot share a page entry
+ // with ADDEND.
+ range_ptr = &entry->ranges;
+ while (*range_ptr && addend > (*range_ptr)->max_addend + 0xffff)
+ range_ptr = &(*range_ptr)->next;
+
+ // If we scanned to the end of the list, or found a range whose
+ // minimum extent cannot share a page entry with ADDEND, create
+ // a new singleton range.
+ range = *range_ptr;
+ if (!range || addend < range->min_addend - 0xffff)
+ {
+ range = new Got_page_range();
+ range->next = *range_ptr;
+ range->min_addend = addend;
+ range->max_addend = addend;
+
+ *range_ptr = range;
+ ++entry->num_pages;
+ if (entry2 != NULL)
+ ++entry2->num_pages;
+ ++this->page_gotno_;
+ ++g2->page_gotno_;
+ return;
+ }
+
+ // Remember how many pages the old range contributed.
+ old_pages = range->get_max_pages();
+
+ // Update the ranges.
+ if (addend < range->min_addend)
+ range->min_addend = addend;
+ else if (addend > range->max_addend)
+ {
+ if (range->next && addend >= range->next->min_addend - 0xffff)
+ {
+ old_pages += range->next->get_max_pages();
+ range->max_addend = range->next->max_addend;
+ range->next = range->next->next;
+ }
+ else
+ range->max_addend = addend;
+ }
+
+ // Record any change in the total estimate.
+ new_pages = range->get_max_pages();
+ if (old_pages != new_pages)
+ {
+ entry->num_pages += new_pages - old_pages;
+ if (entry2 != NULL)
+ entry2->num_pages += new_pages - old_pages;
+ this->page_gotno_ += new_pages - old_pages;
+ g2->page_gotno_ += new_pages - old_pages;
+ }
+}
+
+// Create all entries that should be in the local part of the GOT.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_local_entries(
+ Target_mips<size, big_endian>* target, Layout* layout)
+{
+ Mips_output_data_got<size, big_endian>* got = target->got_section();
+ // First two GOT entries are reserved. The first entry will be filled at
+ // runtime. The second entry will be used by some runtime loaders.
+ got->add_constant(0);
+ got->add_constant(target->mips_elf_gnu_got1_mask());
+
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (entry->is_for_local_symbol() && !entry->is_tls_entry())
+ {
+ got->add_local(entry->object(), entry->symndx(),
+ GOT_TYPE_STANDARD);
+ unsigned int got_offset = entry->object()->local_got_offset(
+ entry->symndx(), GOT_TYPE_STANDARD);
+ if (got->multi_got() && this->index_ > 0
+ && parameters->options().output_is_position_independent())
+ target->rel_dyn_section(layout)->add_local(entry->object(),
+ entry->symndx(), elfcpp::R_MIPS_REL32, got, got_offset);
+ }
+ }
+
+ this->add_page_entries(target, layout);
+
+ // Add global entries that should be in the local area.
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (!entry->is_for_global_symbol())
+ continue;
+
+ Mips_symbol<size>* mips_sym = entry->sym();
+ if (mips_sym->global_got_area() == GGA_NONE && !entry->is_tls_entry())
+ {
+ unsigned int got_type;
+ if (!got->multi_got())
+ got_type = GOT_TYPE_STANDARD;
+ else
+ got_type = GOT_TYPE_STANDARD_MULTIGOT + this->index_;
+ if (got->add_global(mips_sym, got_type))
+ {
+ mips_sym->set_global_gotoffset(mips_sym->got_offset(got_type));
+ if (got->multi_got() && this->index_ > 0
+ && parameters->options().output_is_position_independent())
+ target->rel_dyn_section(layout)->add_symbolless_global_addend(
+ mips_sym, elfcpp::R_MIPS_REL32, got,
+ mips_sym->got_offset(got_type));
+ }
+ }
+ }
+}
+
+// Create GOT page entries.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_page_entries(
+ Target_mips<size, big_endian>* target, Layout* layout)
+{
+ if (this->page_gotno_ == 0)
+ return;
+
+ Mips_output_data_got<size, big_endian>* got = target->got_section();
+ this->got_page_offset_start_ = got->add_constant(0);
+ if (got->multi_got() && this->index_ > 0
+ && parameters->options().output_is_position_independent())
+ target->rel_dyn_section(layout)->add_absolute(elfcpp::R_MIPS_REL32, got,
+ this->got_page_offset_start_);
+ int num_entries = this->page_gotno_;
+ unsigned int prev_offset = this->got_page_offset_start_;
+ while (--num_entries > 0)
+ {
+ unsigned int next_offset = got->add_constant(0);
+ if (got->multi_got() && this->index_ > 0
+ && parameters->options().output_is_position_independent())
+ target->rel_dyn_section(layout)->add_absolute(elfcpp::R_MIPS_REL32, got,
+ next_offset);
+ gold_assert(next_offset == prev_offset + size/8);
+ prev_offset = next_offset;
+ }
+ this->got_page_offset_next_ = this->got_page_offset_start_;
+}
+
+// Create global GOT entries, both GGA_NORMAL and GGA_RELOC_ONLY.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_global_entries(
+ Target_mips<size, big_endian>* target, Layout* layout,
+ unsigned int non_reloc_only_global_gotno)
+{
+ Mips_output_data_got<size, big_endian>* got = target->got_section();
+ // Add GGA_NORMAL entries.
+ unsigned int count = 0;
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (!entry->is_for_global_symbol())
+ continue;
+
+ Mips_symbol<size>* mips_sym = entry->sym();
+ if (mips_sym->global_got_area() != GGA_NORMAL)
+ continue;
+
+ unsigned int got_type;
+ if (!got->multi_got())
+ got_type = GOT_TYPE_STANDARD;
+ else
+ // In multi-GOT links, global symbol can be in both primary and
+ // secondary GOT(s). By creating custom GOT type
+ // (GOT_TYPE_STANDARD_MULTIGOT + got_index) we ensure that symbol
+ // is added to secondary GOT(s).
+ got_type = GOT_TYPE_STANDARD_MULTIGOT + this->index_;
+ if (!got->add_global(mips_sym, got_type))
+ continue;
+
+ mips_sym->set_global_gotoffset(mips_sym->got_offset(got_type));
+ if (got->multi_got() && this->index_ == 0)
+ count++;
+ if (got->multi_got() && this->index_ > 0)
+ {
+ if (parameters->options().output_is_position_independent()
+ || (!parameters->doing_static_link()
+ && mips_sym->is_from_dynobj() && !mips_sym->is_undefined()))
+ {
+ target->rel_dyn_section(layout)->add_global(
+ mips_sym, elfcpp::R_MIPS_REL32, got,
+ mips_sym->got_offset(got_type));
+ got->add_secondary_got_reloc(mips_sym->got_offset(got_type),
+ elfcpp::R_MIPS_REL32, mips_sym);
+ }
+ }
+ }
+
+ if (!got->multi_got() || this->index_ == 0)
+ {
+ if (got->multi_got())
+ {
+ // We need to allocate space in the primary GOT for GGA_NORMAL entries
+ // of secondary GOTs, to ensure that GOT offsets of GGA_RELOC_ONLY
+ // entries correspond to dynamic symbol indexes.
+ while (count < non_reloc_only_global_gotno)
+ {
+ got->add_constant(0);
+ ++count;
+ }
+ }
+
+ // Add GGA_RELOC_ONLY entries.
+ got->add_reloc_only_entries();
+ }
+}
+
+// Create global GOT entries that should be in the GGA_RELOC_ONLY area.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_reloc_only_entries(
+ Mips_output_data_got<size, big_endian>* got)
+{
+ for (typename Unordered_set<Mips_symbol<size>*>::iterator
+ p = this->global_got_symbols_.begin();
+ p != this->global_got_symbols_.end();
+ ++p)
+ {
+ Mips_symbol<size>* mips_sym = *p;
+ if (mips_sym->global_got_area() == GGA_RELOC_ONLY)
+ {
+ unsigned int got_type;
+ if (!got->multi_got())
+ got_type = GOT_TYPE_STANDARD;
+ else
+ got_type = GOT_TYPE_STANDARD_MULTIGOT;
+ if (got->add_global(mips_sym, got_type))
+ mips_sym->set_global_gotoffset(mips_sym->got_offset(got_type));
+ }
+ }
+}
+
+// Create TLS GOT entries.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_tls_entries(
+ Target_mips<size, big_endian>* target, Layout* layout)
+{
+ Mips_output_data_got<size, big_endian>* got = target->got_section();
+ // Add local tls entries.
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (!entry->is_tls_entry() || !entry->is_for_local_symbol())
+ continue;
+
+ if (entry->tls_type() == GOT_TLS_GD)
+ {
+ unsigned int got_type = GOT_TYPE_TLS_PAIR;
+ unsigned int r_type1 = (size == 32 ? elfcpp::R_MIPS_TLS_DTPMOD32
+ : elfcpp::R_MIPS_TLS_DTPMOD64);
+ unsigned int r_type2 = (size == 32 ? elfcpp::R_MIPS_TLS_DTPREL32
+ : elfcpp::R_MIPS_TLS_DTPREL64);
+
+ if (!parameters->doing_static_link())
+ {
+ got->add_local_pair_with_rel(entry->object(), entry->symndx(),
+ entry->shndx(), got_type,
+ target->rel_dyn_section(layout),
+ r_type1);
+ unsigned int got_offset =
+ entry->object()->local_got_offset(entry->symndx(), got_type);
+ got->add_static_reloc(got_offset + size/8, r_type2,
+ entry->object(), entry->symndx());
+ }
+ else
+ {
+ // We are doing a static link. Mark it as belong to module 1,
+ // the executable.
+ unsigned int got_offset = got->add_constant(1);
+ entry->object()->set_local_got_offset(entry->symndx(), got_type,
+ got_offset);
+ got->add_constant(0);
+ got->add_static_reloc(got_offset + size/8, r_type2,
+ entry->object(), entry->symndx());
+ }
+ }
+ else if (entry->tls_type() == GOT_TLS_IE)
+ {
+ unsigned int got_type = GOT_TYPE_TLS_OFFSET;
+ unsigned int r_type = (size == 32 ? elfcpp::R_MIPS_TLS_TPREL32
+ : elfcpp::R_MIPS_TLS_TPREL64);
+ if (!parameters->doing_static_link())
+ got->add_local_with_rel(entry->object(), entry->symndx(), got_type,
+ target->rel_dyn_section(layout), r_type);
+ else
+ {
+ got->add_local(entry->object(), entry->symndx(), got_type);
+ unsigned int got_offset =
+ entry->object()->local_got_offset(entry->symndx(), got_type);
+ got->add_static_reloc(got_offset, r_type, entry->object(),
+ entry->symndx());
+ }
+ }
+ else if (entry->tls_type() == GOT_TLS_LDM)
+ {
+ unsigned int r_type = (size == 32 ? elfcpp::R_MIPS_TLS_DTPMOD32
+ : elfcpp::R_MIPS_TLS_DTPMOD64);
+ unsigned int got_offset;
+ if (!parameters->doing_static_link())
+ {
+ got_offset = got->add_constant(0);
+ target->rel_dyn_section(layout)->add_local(
+ entry->object(), 0, r_type, got, got_offset);
+ }
+ else
+ // We are doing a static link. Just mark it as belong to module 1,
+ // the executable.
+ got_offset = got->add_constant(1);
+
+ got->add_constant(0);
+ got->set_tls_ldm_offset(got_offset, entry->object());
+ }
+ else
+ gold_unreachable();
+ }
+
+ // Add global tls entries.
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (!entry->is_tls_entry() || !entry->is_for_global_symbol())
+ continue;
+
+ Mips_symbol<size>* mips_sym = entry->sym();
+ if (entry->tls_type() == GOT_TLS_GD)
+ {
+ unsigned int got_type;
+ if (!got->multi_got())
+ got_type = GOT_TYPE_TLS_PAIR;
+ else
+ got_type = GOT_TYPE_TLS_PAIR_MULTIGOT + this->index_;
+ unsigned int r_type1 = (size == 32 ? elfcpp::R_MIPS_TLS_DTPMOD32
+ : elfcpp::R_MIPS_TLS_DTPMOD64);
+ unsigned int r_type2 = (size == 32 ? elfcpp::R_MIPS_TLS_DTPREL32
+ : elfcpp::R_MIPS_TLS_DTPREL64);
+ if (!parameters->doing_static_link())
+ got->add_global_pair_with_rel(mips_sym, got_type,
+ target->rel_dyn_section(layout), r_type1, r_type2);
+ else
+ {
+ // Add a GOT pair for for R_MIPS_TLS_GD. The creates a pair of
+ // GOT entries. The first one is initialized to be 1, which is the
+ // module index for the main executable and the second one 0. A
+ // reloc of the type R_MIPS_TLS_DTPREL32/64 will be created for
+ // the second GOT entry and will be applied by gold.
+ unsigned int got_offset = got->add_constant(1);
+ mips_sym->set_got_offset(got_type, got_offset);
+ got->add_constant(0);
+ got->add_static_reloc(got_offset + size/8, r_type2, mips_sym);
+ }
+ }
+ else if (entry->tls_type() == GOT_TLS_IE)
+ {
+ unsigned int got_type;
+ if (!got->multi_got())
+ got_type = GOT_TYPE_TLS_OFFSET;
+ else
+ got_type = GOT_TYPE_TLS_OFFSET_MULTIGOT + this->index_;
+ unsigned int r_type = (size == 32 ? elfcpp::R_MIPS_TLS_TPREL32
+ : elfcpp::R_MIPS_TLS_TPREL64);
+ if (!parameters->doing_static_link())
+ got->add_global_with_rel(mips_sym, got_type,
+ target->rel_dyn_section(layout), r_type);
+ else
+ {
+ got->add_global(mips_sym, got_type);
+ unsigned int got_offset = mips_sym->got_offset(got_type);
+ got->add_static_reloc(got_offset, r_type, mips_sym);
+ }
+ }
+ else
+ gold_unreachable();
+ }
+}
^ permalink raw reply [flat|nested] 9+ messages in thread
* RE: Mips target in gold - revision 3 - part 2
2014-04-15 20:46 ` Cary Coutant
2014-04-15 21:08 ` Cary Coutant
2014-06-04 21:01 ` Sasa Stankovic
@ 2014-06-04 21:03 ` Sasa Stankovic
2 siblings, 0 replies; 9+ messages in thread
From: Sasa Stankovic @ 2014-06-04 21:03 UTC (permalink / raw)
To: Cary Coutant; +Cc: Rich Fuhler, binutils, Petar Jovanovic
[-- Attachment #1: Type: text/plain, Size: 2704 bytes --]
Attached is the second part of the updated patch.
Regards,
Sasa
________________________________________
From: Cary Coutant [ccoutant@google.com]
Sent: Tuesday, April 15, 2014 10:46 PM
To: Sasa Stankovic
Cc: Rich Fuhler; binutils@sourceware.org; iant@google.com; Petar Jovanovic
Subject: Re: Mips target in gold - revision 3 - part 2
>> If I understand the code correctly, in this case the HI16 reloc
>> will remain on the got16_addends_ list forever. Is there a point
>> where you know it's safe to remove it? Can you clear the list
>> at the end of each section?
>
> If LO16 is removed by GCC, than HI16 reloc is paired with the next found LO16
> reloc (this means that two or more HI16 reloc will be paired with a single LO16 reloc).
> This is what GNU ld does - it always tries to find LO16 reloc for HI16 reloc, and if it
> can't find it, error is reported.
>
> When trying to find LO16 reloc, GNU ld examines next relocations in the section until
> it finds it, or reports an error. I implemented it differently in Gold - when HI16 or GOT16
> reloc is found, it is recorded in the got16_addends_ list, and the normal scaning of
> relocations continues. Then when the LO16 reloc is found, the got16_addends_ list is
> examined and all pending matching HI16 and GOT16 relocs are removed. The error when
> HI16 or GOT16 reloc doesn't have the LO16 part is detected by checking whether the
> section of the pending HI16 or GOT16 relocation is different from the section of the current
> LO16 relocation. The check whether the final section that is scanned has HI16 or GOT16
> without the LO16 part is in the do_finalize_sections method - I just check whether the
> got16_addends_ list is empty - if it isn't, error is reported.
Ah, I see.
>> > typename std::list<got16_addend<size, big_endian> > got16_addends_;
>>
>> Do you realy want to use a linked list here? Wouldn't a
>> vector be better?
>
> I used std::list because erase is called on it.
Sorry, I missed that. And now I understand that you don't expect the
list to ever grow very long.
> // Check whether the final section that was scaned has HI16 or GOT16
Typo: "scanned".
> // Fill the unused space with zeroes.
> // TODO(sasa): Can we strip unused bytes during the relaxation?
> unsigned char* end = oview + oview_size;
> while (pov < end)
> {
> elfcpp::Swap<32, big_endian>::writeval(pov, 0);
> pov += 4;
> }
You can just use memset here:
if (unused > 0)
memset(pov, 0, unused);
If that's OK with you, I'll commit it with that change.
Thanks for working on this, and for putting up with the long review process!
-cary
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: mips2.patch --]
[-- Type: text/x-patch; name="mips2.patch", Size: 197885 bytes --]
diff --git a/gold/mips.cc b/gold/mips.cc
index b0fcc93..bb7a796 100644
--- a/gold/mips.cc
+++ b/gold/mips.cc
@@ -5252,3 +5252,5268 @@ Mips_got_info<size, big_endian>::add_tls_entries(
gold_unreachable();
}
}
+
+// Decide whether the symbol needs an entry in the global part of the primary
+// GOT, setting global_got_area accordingly. Count the number of global
+// symbols that are in the primary GOT only because they have dynamic
+// relocations R_MIPS_REL32 against them (reloc_only_gotno).
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::count_got_symbols(Symbol_table* symtab)
+{
+ for (typename Unordered_set<Mips_symbol<size>*>::iterator
+ p = this->global_got_symbols_.begin();
+ p != this->global_got_symbols_.end();
+ ++p)
+ {
+ Mips_symbol<size>* sym = *p;
+ // Make a final decision about whether the symbol belongs in the
+ // local or global GOT. Symbols that bind locally can (and in the
+ // case of forced-local symbols, must) live in the local GOT.
+ // Those that are aren't in the dynamic symbol table must also
+ // live in the local GOT.
+
+ if (!sym->should_add_dynsym_entry(symtab)
+ || (sym->got_only_for_calls()
+ ? symbol_calls_local(sym, sym->should_add_dynsym_entry(symtab))
+ : symbol_references_local(sym,
+ sym->should_add_dynsym_entry(symtab))))
+ // The symbol belongs in the local GOT. We no longer need this
+ // entry if it was only used for relocations; those relocations
+ // will be against the null or section symbol instead.
+ sym->set_global_got_area(GGA_NONE);
+ else if (sym->global_got_area() == GGA_RELOC_ONLY)
+ {
+ ++this->reloc_only_gotno_;
+ ++this->global_gotno_ ;
+ }
+ }
+}
+
+// Return the offset of GOT page entry for VALUE. Initialize the entry with
+// VALUE if it is not initialized.
+
+template<int size, bool big_endian>
+unsigned int
+Mips_got_info<size, big_endian>::get_got_page_offset(Mips_address value,
+ Mips_output_data_got<size, big_endian>* got)
+{
+ typename Got_page_offsets::iterator it = this->got_page_offsets_.find(value);
+ if (it != this->got_page_offsets_.end())
+ return it->second;
+
+ gold_assert(this->got_page_offset_next_ < this->got_page_offset_start_
+ + (size/8) * this->page_gotno_);
+
+ unsigned int got_offset = this->got_page_offset_next_;
+ this->got_page_offsets_[value] = got_offset;
+ this->got_page_offset_next_ += size/8;
+ got->update_got_entry(got_offset, value);
+ return got_offset;
+}
+
+// Remove lazy-binding stubs for global symbols in this GOT.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::remove_lazy_stubs(
+ Target_mips<size, big_endian>* target)
+{
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (entry->is_for_global_symbol())
+ target->remove_lazy_stub_entry(entry->sym());
+ }
+}
+
+// Count the number of GOT entries required.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::count_got_entries()
+{
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ this->count_got_entry(*p);
+ }
+}
+
+// Count the number of GOT entries required by ENTRY. Accumulate the result.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::count_got_entry(
+ Mips_got_entry<size, big_endian>* entry)
+{
+ if (entry->is_tls_entry())
+ this->tls_gotno_ += mips_tls_got_entries(entry->tls_type());
+ else if (entry->is_for_local_symbol()
+ || entry->sym()->global_got_area() == GGA_NONE)
+ ++this->local_gotno_;
+ else
+ ++this->global_gotno_;
+}
+
+// Add FROM's GOT entries.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_got_entries(
+ Mips_got_info<size, big_endian>* from)
+{
+ for (typename Got_entry_set::iterator
+ p = from->got_entries_.begin();
+ p != from->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (this->got_entries_.find(entry) == this->got_entries_.end())
+ {
+ Mips_got_entry<size, big_endian>* entry2 =
+ new Mips_got_entry<size, big_endian>(*entry);
+ this->got_entries_.insert(entry2);
+ this->count_got_entry(entry);
+ }
+ }
+}
+
+// Add FROM's GOT page entries.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_got_page_entries(
+ Mips_got_info<size, big_endian>* from)
+{
+ for (typename Got_page_entry_set::iterator
+ p = from->got_page_entries_.begin();
+ p != from->got_page_entries_.end();
+ ++p)
+ {
+ Got_page_entry* entry = *p;
+ if (this->got_page_entries_.find(entry) == this->got_page_entries_.end())
+ {
+ Got_page_entry* entry2 = new Got_page_entry(*entry);
+ this->got_page_entries_.insert(entry2);
+ this->page_gotno_ += entry->num_pages;
+ }
+ }
+}
+
+// Mips_output_data_got methods.
+
+// Lay out the GOT. Add local, global and TLS entries. If GOT is
+// larger than 64K, create multi-GOT.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_got<size, big_endian>::lay_out_got(Layout* layout,
+ Symbol_table* symtab, const Input_objects* input_objects)
+{
+ // Decide which symbols need to go in the global part of the GOT and
+ // count the number of reloc-only GOT symbols.
+ this->master_got_info_->count_got_symbols(symtab);
+
+ // Count the number of GOT entries.
+ this->master_got_info_->count_got_entries();
+
+ unsigned int got_size = this->master_got_info_->got_size();
+ if (got_size > Target_mips<size, big_endian>::MIPS_GOT_MAX_SIZE)
+ this->lay_out_multi_got(layout, input_objects);
+ else
+ {
+ // Record that all objects use single GOT.
+ for (Input_objects::Relobj_iterator p = input_objects->relobj_begin();
+ p != input_objects->relobj_end();
+ ++p)
+ {
+ Mips_relobj<size, big_endian>* object =
+ Mips_relobj<size, big_endian>::as_mips_relobj(*p);
+ if (object->get_got_info() != NULL)
+ object->set_got_info(this->master_got_info_);
+ }
+
+ this->master_got_info_->add_local_entries(this->target_, layout);
+ this->master_got_info_->add_global_entries(this->target_, layout,
+ /*not used*/-1U);
+ this->master_got_info_->add_tls_entries(this->target_, layout);
+ }
+}
+
+// Create multi-GOT. For every GOT, add local, global and TLS entries.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_got<size, big_endian>::lay_out_multi_got(Layout* layout,
+ const Input_objects* input_objects)
+{
+ // Try to merge the GOTs of input objects together, as long as they
+ // don't seem to exceed the maximum GOT size, choosing one of them
+ // to be the primary GOT.
+ this->merge_gots(input_objects);
+
+ // Every symbol that is referenced in a dynamic relocation must be
+ // present in the primary GOT.
+ this->primary_got_->set_global_gotno(this->master_got_info_->global_gotno());
+
+ // Add GOT entries.
+ unsigned int i = 0;
+ unsigned int offset = 0;
+ Mips_got_info<size, big_endian>* g = this->primary_got_;
+ do
+ {
+ g->set_index(i);
+ g->set_offset(offset);
+
+ g->add_local_entries(this->target_, layout);
+ if (i == 0)
+ g->add_global_entries(this->target_, layout,
+ (this->master_got_info_->global_gotno()
+ - this->master_got_info_->reloc_only_gotno()));
+ else
+ g->add_global_entries(this->target_, layout, /*not used*/-1U);
+ g->add_tls_entries(this->target_, layout);
+
+ // Forbid global symbols in every non-primary GOT from having
+ // lazy-binding stubs.
+ if (i > 0)
+ g->remove_lazy_stubs(this->target_);
+
+ ++i;
+ offset += g->got_size();
+ g = g->next();
+ }
+ while (g);
+}
+
+// Attempt to merge GOTs of different input objects. Try to use as much as
+// possible of the primary GOT, since it doesn't require explicit dynamic
+// relocations, but don't use objects that would reference global symbols
+// out of the addressable range. Failing the primary GOT, attempt to merge
+// with the current GOT, or finish the current GOT and then make make the new
+// GOT current.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_got<size, big_endian>::merge_gots(
+ const Input_objects* input_objects)
+{
+ gold_assert(this->primary_got_ == NULL);
+ Mips_got_info<size, big_endian>* current = NULL;
+
+ for (Input_objects::Relobj_iterator p = input_objects->relobj_begin();
+ p != input_objects->relobj_end();
+ ++p)
+ {
+ Mips_relobj<size, big_endian>* object =
+ Mips_relobj<size, big_endian>::as_mips_relobj(*p);
+
+ Mips_got_info<size, big_endian>* g = object->get_got_info();
+ if (g == NULL)
+ continue;
+
+ g->count_got_entries();
+
+ // Work out the number of page, local and TLS entries.
+ unsigned int estimate = this->master_got_info_->page_gotno();
+ if (estimate > g->page_gotno())
+ estimate = g->page_gotno();
+ estimate += g->local_gotno() + g->tls_gotno();
+
+ // We place TLS GOT entries after both locals and globals. The globals
+ // for the primary GOT may overflow the normal GOT size limit, so be
+ // sure not to merge a GOT which requires TLS with the primary GOT in that
+ // case. This doesn't affect non-primary GOTs.
+ estimate += (g->tls_gotno() > 0 ? this->master_got_info_->global_gotno()
+ : g->global_gotno());
+
+ unsigned int max_count =
+ Target_mips<size, big_endian>::MIPS_GOT_MAX_SIZE / (size/8) - 2;
+ if (estimate <= max_count)
+ {
+ // If we don't have a primary GOT, use it as
+ // a starting point for the primary GOT.
+ if (!this->primary_got_)
+ {
+ this->primary_got_ = g;
+ continue;
+ }
+
+ // Try merging with the primary GOT.
+ if (this->merge_got_with(g, object, this->primary_got_))
+ continue;
+ }
+
+ // If we can merge with the last-created GOT, do it.
+ if (current && this->merge_got_with(g, object, current))
+ continue;
+
+ // Well, we couldn't merge, so create a new GOT. Don't check if it
+ // fits; if it turns out that it doesn't, we'll get relocation
+ // overflows anyway.
+ g->set_next(current);
+ current = g;
+ }
+
+ // If we do not find any suitable primary GOT, create an empty one.
+ if (this->primary_got_ == NULL)
+ this->primary_got_ = new Mips_got_info<size, big_endian>();
+
+ // Link primary GOT with secondary GOTs.
+ this->primary_got_->set_next(current);
+}
+
+// Consider merging FROM, which is OBJECT's GOT, into TO. Return false if
+// this would lead to overflow, true if they were merged successfully.
+
+template<int size, bool big_endian>
+bool
+Mips_output_data_got<size, big_endian>::merge_got_with(
+ Mips_got_info<size, big_endian>* from,
+ Mips_relobj<size, big_endian>* object,
+ Mips_got_info<size, big_endian>* to)
+{
+ // Work out how many page entries we would need for the combined GOT.
+ unsigned int estimate = this->master_got_info_->page_gotno();
+ if (estimate >= from->page_gotno() + to->page_gotno())
+ estimate = from->page_gotno() + to->page_gotno();
+
+ // Conservatively estimate how many local and TLS entries would be needed.
+ estimate += from->local_gotno() + to->local_gotno();
+ estimate += from->tls_gotno() + to->tls_gotno();
+
+ // If we're merging with the primary got, any TLS relocations will
+ // come after the full set of global entries. Otherwise estimate those
+ // conservatively as well.
+ if (to == this->primary_got_ && (from->tls_gotno() + to->tls_gotno()) > 0)
+ estimate += this->master_got_info_->global_gotno();
+ else
+ estimate += from->global_gotno() + to->global_gotno();
+
+ // Bail out if the combined GOT might be too big.
+ unsigned int max_count =
+ Target_mips<size, big_endian>::MIPS_GOT_MAX_SIZE / (size/8) - 2;
+ if (estimate > max_count)
+ return false;
+
+ // Transfer the object's GOT information from FROM to TO.
+ to->add_got_entries(from);
+ to->add_got_page_entries(from);
+
+ // Record that OBJECT should use output GOT TO.
+ object->set_got_info(to);
+
+ return true;
+}
+
+// Write out the GOT.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_got<size, big_endian>::do_write(Output_file* of)
+{
+ // Call parent to write out GOT.
+ Output_data_got<size, big_endian>::do_write(of);
+
+ const off_t offset = this->offset();
+ const section_size_type oview_size =
+ convert_to_section_size_type(this->data_size());
+ unsigned char* const oview = of->get_output_view(offset, oview_size);
+
+ // Needed for fixing values of .got section.
+ this->got_view_ = oview;
+
+ // Write lazy stub addresses.
+ for (typename Unordered_set<Mips_symbol<size>*>::iterator
+ p = this->master_got_info_->global_got_symbols().begin();
+ p != this->master_got_info_->global_got_symbols().end();
+ ++p)
+ {
+ Mips_symbol<size>* mips_sym = *p;
+ if (mips_sym->has_lazy_stub())
+ {
+ Valtype* wv = reinterpret_cast<Valtype*>(
+ oview + this->get_primary_got_offset(mips_sym));
+ Valtype value =
+ this->target_->mips_stubs_section()->stub_address(mips_sym);
+ elfcpp::Swap<size, big_endian>::writeval(wv, value);
+ }
+ }
+
+ // Add +1 to GGA_NONE nonzero MIPS16 and microMIPS entries.
+ for (typename Unordered_set<Mips_symbol<size>*>::iterator
+ p = this->master_got_info_->global_got_symbols().begin();
+ p != this->master_got_info_->global_got_symbols().end();
+ ++p)
+ {
+ Mips_symbol<size>* mips_sym = *p;
+ if (!this->multi_got()
+ && (mips_sym->is_mips16() || mips_sym->is_micromips())
+ && mips_sym->global_got_area() == GGA_NONE
+ && mips_sym->has_got_offset(GOT_TYPE_STANDARD))
+ {
+ Valtype* wv = reinterpret_cast<Valtype*>(
+ oview + mips_sym->got_offset(GOT_TYPE_STANDARD));
+ Valtype value = elfcpp::Swap<size, big_endian>::readval(wv);
+ if (value != 0)
+ {
+ value |= 1;
+ elfcpp::Swap<size, big_endian>::writeval(wv, value);
+ }
+ }
+ }
+
+ if (!this->secondary_got_relocs_.empty())
+ {
+ // Fixup for the secondary GOT R_MIPS_REL32 relocs. For global
+ // secondary GOT entries with non-zero initial value copy the value
+ // to the corresponding primary GOT entry, and set the secondary GOT
+ // entry to zero.
+ // TODO(sasa): This is workaround. It needs to be investigated further.
+
+ for (size_t i = 0; i < this->secondary_got_relocs_.size(); ++i)
+ {
+ Static_reloc& reloc(this->secondary_got_relocs_[i]);
+ if (reloc.symbol_is_global())
+ {
+ Mips_symbol<size>* gsym = reloc.symbol();
+ gold_assert(gsym != NULL);
+
+ unsigned got_offset = reloc.got_offset();
+ gold_assert(got_offset < oview_size);
+
+ // Find primary GOT entry.
+ Valtype* wv_prim = reinterpret_cast<Valtype*>(
+ oview + this->get_primary_got_offset(gsym));
+
+ // Find secondary GOT entry.
+ Valtype* wv_sec = reinterpret_cast<Valtype*>(oview + got_offset);
+
+ Valtype value = elfcpp::Swap<size, big_endian>::readval(wv_sec);
+ if (value != 0)
+ {
+ elfcpp::Swap<size, big_endian>::writeval(wv_prim, value);
+ elfcpp::Swap<size, big_endian>::writeval(wv_sec, 0);
+ gsym->set_applied_secondary_got_fixup();
+ }
+ }
+ }
+
+ of->write_output_view(offset, oview_size, oview);
+ }
+
+ // We are done if there is no fix up.
+ if (this->static_relocs_.empty())
+ return;
+
+ Output_segment* tls_segment = this->layout_->tls_segment();
+ gold_assert(tls_segment != NULL);
+
+ for (size_t i = 0; i < this->static_relocs_.size(); ++i)
+ {
+ Static_reloc& reloc(this->static_relocs_[i]);
+
+ Mips_address value;
+ if (!reloc.symbol_is_global())
+ {
+ Sized_relobj_file<size, big_endian>* object = reloc.relobj();
+ const Symbol_value<size>* psymval =
+ object->local_symbol(reloc.index());
+
+ // We are doing static linking. Issue an error and skip this
+ // relocation if the symbol is undefined or in a discarded_section.
+ bool is_ordinary;
+ unsigned int shndx = psymval->input_shndx(&is_ordinary);
+ if ((shndx == elfcpp::SHN_UNDEF)
+ || (is_ordinary
+ && shndx != elfcpp::SHN_UNDEF
+ && !object->is_section_included(shndx)
+ && !this->symbol_table_->is_section_folded(object, shndx)))
+ {
+ gold_error(_("undefined or discarded local symbol %u from "
+ " object %s in GOT"),
+ reloc.index(), reloc.relobj()->name().c_str());
+ continue;
+ }
+
+ value = psymval->value(object, 0);
+ }
+ else
+ {
+ const Mips_symbol<size>* gsym = reloc.symbol();
+ gold_assert(gsym != NULL);
+
+ // We are doing static linking. Issue an error and skip this
+ // relocation if the symbol is undefined or in a discarded_section
+ // unless it is a weakly_undefined symbol.
+ if ((gsym->is_defined_in_discarded_section() || gsym->is_undefined())
+ && !gsym->is_weak_undefined())
+ {
+ gold_error(_("undefined or discarded symbol %s in GOT"),
+ gsym->name());
+ continue;
+ }
+
+ if (!gsym->is_weak_undefined())
+ value = gsym->value();
+ else
+ value = 0;
+ }
+
+ unsigned got_offset = reloc.got_offset();
+ gold_assert(got_offset < oview_size);
+
+ Valtype* wv = reinterpret_cast<Valtype*>(oview + got_offset);
+ Valtype x;
+
+ switch (reloc.r_type())
+ {
+ case elfcpp::R_MIPS_TLS_DTPMOD32:
+ case elfcpp::R_MIPS_TLS_DTPMOD64:
+ x = value;
+ break;
+ case elfcpp::R_MIPS_TLS_DTPREL32:
+ case elfcpp::R_MIPS_TLS_DTPREL64:
+ x = value - elfcpp::DTP_OFFSET;
+ break;
+ case elfcpp::R_MIPS_TLS_TPREL32:
+ case elfcpp::R_MIPS_TLS_TPREL64:
+ x = value - elfcpp::TP_OFFSET;
+ break;
+ default:
+ gold_unreachable();
+ break;
+ }
+
+ elfcpp::Swap<size, big_endian>::writeval(wv, x);
+ }
+
+ of->write_output_view(offset, oview_size, oview);
+}
+
+// Mips_relobj methods.
+
+// Count the local symbols. The Mips backend needs to know if a symbol
+// is a MIPS16 or microMIPS function or not. For global symbols, it is easy
+// because the Symbol object keeps the ELF symbol type and st_other field.
+// For local symbol it is harder because we cannot access this information.
+// So we override the do_count_local_symbol in parent and scan local symbols to
+// mark MIPS16 and microMIPS functions. This is not the most efficient way but
+// I do not want to slow down other ports by calling a per symbol target hook
+// inside Sized_relobj_file<size, big_endian>::do_count_local_symbols.
+
+template<int size, bool big_endian>
+void
+Mips_relobj<size, big_endian>::do_count_local_symbols(
+ Stringpool_template<char>* pool,
+ Stringpool_template<char>* dynpool)
+{
+ // Ask parent to count the local symbols.
+ Sized_relobj_file<size, big_endian>::do_count_local_symbols(pool, dynpool);
+ const unsigned int loccount = this->local_symbol_count();
+ if (loccount == 0)
+ return;
+
+ // Initialize the mips16 and micromips function bit-vector.
+ this->local_symbol_is_mips16_.resize(loccount, false);
+ this->local_symbol_is_micromips_.resize(loccount, false);
+
+ // Read the symbol table section header.
+ const unsigned int symtab_shndx = this->symtab_shndx();
+ elfcpp::Shdr<size, big_endian>
+ symtabshdr(this, this->elf_file()->section_header(symtab_shndx));
+ gold_assert(symtabshdr.get_sh_type() == elfcpp::SHT_SYMTAB);
+
+ // Read the local symbols.
+ const int sym_size = elfcpp::Elf_sizes<size>::sym_size;
+ gold_assert(loccount == symtabshdr.get_sh_info());
+ off_t locsize = loccount * sym_size;
+ const unsigned char* psyms = this->get_view(symtabshdr.get_sh_offset(),
+ locsize, true, true);
+
+ // Loop over the local symbols and mark any MIPS16 or microMIPS local symbols.
+
+ // Skip the first dummy symbol.
+ psyms += sym_size;
+ for (unsigned int i = 1; i < loccount; ++i, psyms += sym_size)
+ {
+ elfcpp::Sym<size, big_endian> sym(psyms);
+ unsigned char st_other = sym.get_st_other();
+ this->local_symbol_is_mips16_[i] = elfcpp::elf_st_is_mips16(st_other);
+ this->local_symbol_is_micromips_[i] =
+ elfcpp::elf_st_is_micromips(st_other);
+ }
+}
+
+// Read the symbol information.
+
+template<int size, bool big_endian>
+void
+Mips_relobj<size, big_endian>::do_read_symbols(Read_symbols_data* sd)
+{
+ // Call parent class to read symbol information.
+ Sized_relobj_file<size, big_endian>::do_read_symbols(sd);
+
+ // Read processor-specific flags in ELF file header.
+ const unsigned char* pehdr = this->get_view(elfcpp::file_header_offset,
+ elfcpp::Elf_sizes<size>::ehdr_size,
+ true, false);
+ elfcpp::Ehdr<size, big_endian> ehdr(pehdr);
+ this->processor_specific_flags_ = ehdr.get_e_flags();
+
+ // Get the section names.
+ const unsigned char* pnamesu = sd->section_names->data();
+ const char* pnames = reinterpret_cast<const char*>(pnamesu);
+
+ // Initialize the mips16 stub section bit-vectors.
+ this->section_is_mips16_fn_stub_.resize(this->shnum(), false);
+ this->section_is_mips16_call_stub_.resize(this->shnum(), false);
+ this->section_is_mips16_call_fp_stub_.resize(this->shnum(), false);
+
+ const size_t shdr_size = elfcpp::Elf_sizes<size>::shdr_size;
+ const unsigned char* pshdrs = sd->section_headers->data();
+ const unsigned char* ps = pshdrs + shdr_size;
+ for (unsigned int i = 1; i < this->shnum(); ++i, ps += shdr_size)
+ {
+ elfcpp::Shdr<size, big_endian> shdr(ps);
+
+ if (shdr.get_sh_type() == elfcpp::SHT_MIPS_REGINFO)
+ {
+ // Read the gp value that was used to create this object. We need the
+ // gp value while processing relocs. The .reginfo section is not used
+ // in the 64-bit MIPS ELF ABI.
+ section_offset_type section_offset = shdr.get_sh_offset();
+ section_size_type section_size =
+ convert_to_section_size_type(shdr.get_sh_size());
+ const unsigned char* view =
+ this->get_view(section_offset, section_size, true, false);
+
+ this->gp_ = elfcpp::Swap<size, big_endian>::readval(view + 20);
+
+ // Read the rest of .reginfo.
+ this->gprmask_ = elfcpp::Swap<size, big_endian>::readval(view);
+ this->cprmask1_ = elfcpp::Swap<size, big_endian>::readval(view + 4);
+ this->cprmask2_ = elfcpp::Swap<size, big_endian>::readval(view + 8);
+ this->cprmask3_ = elfcpp::Swap<size, big_endian>::readval(view + 12);
+ this->cprmask4_ = elfcpp::Swap<size, big_endian>::readval(view + 16);
+ }
+
+ const char* name = pnames + shdr.get_sh_name();
+ this->section_is_mips16_fn_stub_[i] = is_prefix_of(".mips16.fn", name);
+ this->section_is_mips16_call_stub_[i] =
+ is_prefix_of(".mips16.call.", name);
+ this->section_is_mips16_call_fp_stub_[i] =
+ is_prefix_of(".mips16.call.fp.", name);
+
+ if (strcmp(name, ".pdr") == 0)
+ {
+ gold_assert(this->pdr_shndx_ == -1U);
+ this->pdr_shndx_ = i;
+ }
+ }
+}
+
+// Discard MIPS16 stub secions that are not needed.
+
+template<int size, bool big_endian>
+void
+Mips_relobj<size, big_endian>::discard_mips16_stub_sections(Symbol_table* symtab)
+{
+ for (typename Mips16_stubs_int_map::const_iterator
+ it = this->mips16_stub_sections_.begin();
+ it != this->mips16_stub_sections_.end(); ++it)
+ {
+ Mips16_stub_section<size, big_endian>* stub_section = it->second;
+ if (!stub_section->is_target_found())
+ {
+ gold_error(_("no relocation found in mips16 stub section '%s'"),
+ stub_section->object()
+ ->section_name(stub_section->shndx()).c_str());
+ }
+
+ bool discard = false;
+ if (stub_section->is_for_local_function())
+ {
+ if (stub_section->is_fn_stub())
+ {
+ // This stub is for a local symbol. This stub will only
+ // be needed if there is some relocation in this object,
+ // other than a 16 bit function call, which refers to this
+ // symbol.
+ if (!this->has_local_non_16bit_call_relocs(stub_section->r_sym()))
+ discard = true;
+ else
+ this->add_local_mips16_fn_stub(stub_section);
+ }
+ else
+ {
+ // This stub is for a local symbol. This stub will only
+ // be needed if there is some relocation (R_MIPS16_26) in
+ // this object that refers to this symbol.
+ gold_assert(stub_section->is_call_stub()
+ || stub_section->is_call_fp_stub());
+ if (!this->has_local_16bit_call_relocs(stub_section->r_sym()))
+ discard = true;
+ else
+ this->add_local_mips16_call_stub(stub_section);
+ }
+ }
+ else
+ {
+ Mips_symbol<size>* gsym = stub_section->gsym();
+ if (stub_section->is_fn_stub())
+ {
+ if (gsym->has_mips16_fn_stub())
+ // We already have a stub for this function.
+ discard = true;
+ else
+ {
+ gsym->set_mips16_fn_stub(stub_section);
+ if (gsym->should_add_dynsym_entry(symtab))
+ {
+ // If we have a MIPS16 function with a stub, the
+ // dynamic symbol must refer to the stub, since only
+ // the stub uses the standard calling conventions.
+ gsym->set_need_fn_stub();
+ if (gsym->is_from_dynobj())
+ gsym->set_needs_dynsym_value();
+ }
+ }
+ if (!gsym->need_fn_stub())
+ discard = true;
+ }
+ else if (stub_section->is_call_stub())
+ {
+ if (gsym->is_mips16())
+ // We don't need the call_stub; this is a 16 bit
+ // function, so calls from other 16 bit functions are
+ // OK.
+ discard = true;
+ else if (gsym->has_mips16_call_stub())
+ // We already have a stub for this function.
+ discard = true;
+ else
+ gsym->set_mips16_call_stub(stub_section);
+ }
+ else
+ {
+ gold_assert(stub_section->is_call_fp_stub());
+ if (gsym->is_mips16())
+ // We don't need the call_stub; this is a 16 bit
+ // function, so calls from other 16 bit functions are
+ // OK.
+ discard = true;
+ else if (gsym->has_mips16_call_fp_stub())
+ // We already have a stub for this function.
+ discard = true;
+ else
+ gsym->set_mips16_call_fp_stub(stub_section);
+ }
+ }
+ if (discard)
+ this->set_output_section(stub_section->shndx(), NULL);
+ }
+}
+
+// Mips_output_data_la25_stub methods.
+
+// Template for standard LA25 stub.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_la25_stub<size, big_endian>::la25_stub_entry[] =
+{
+ 0x3c190000, // lui $25,%hi(func)
+ 0x08000000, // j func
+ 0x27390000, // add $25,$25,%lo(func)
+ 0x00000000 // nop
+};
+
+// Template for microMIPS LA25 stub.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_la25_stub<size, big_endian>::la25_stub_micromips_entry[] =
+{
+ 0x41b9, 0x0000, // lui t9,%hi(func)
+ 0xd400, 0x0000, // j func
+ 0x3339, 0x0000, // addiu t9,t9,%lo(func)
+ 0x0000, 0x0000 // nop
+};
+
+// Create la25 stub for a symbol.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_la25_stub<size, big_endian>::create_la25_stub(
+ Symbol_table* symtab, Target_mips<size, big_endian>* target,
+ Mips_symbol<size>* gsym)
+{
+ if (!gsym->has_la25_stub())
+ {
+ gsym->set_la25_stub_offset(this->symbols_.size() * 16);
+ this->symbols_.insert(gsym);
+ this->create_stub_symbol(gsym, symtab, target, 16);
+ }
+}
+
+// Create a symbol for SYM stub's value and size, to help make the disassembly
+// easier to read.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_la25_stub<size, big_endian>::create_stub_symbol(
+ Mips_symbol<size>* sym, Symbol_table* symtab,
+ Target_mips<size, big_endian>* target, uint64_t symsize)
+{
+ std::string name(".pic.");
+ name += sym->name();
+
+ unsigned int offset = sym->la25_stub_offset();
+ if (sym->is_micromips())
+ offset |= 1;
+
+ // Make it a local function.
+ Symbol* new_sym = symtab->define_in_output_data(name.c_str(), NULL,
+ Symbol_table::PREDEFINED,
+ target->la25_stub_section(),
+ offset, symsize, elfcpp::STT_FUNC,
+ elfcpp::STB_LOCAL,
+ elfcpp::STV_DEFAULT, 0,
+ false, false);
+ new_sym->set_is_forced_local();
+}
+
+// Write out la25 stubs. This uses the hand-coded instructions above,
+// and adjusts them as needed.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_la25_stub<size, big_endian>::do_write(Output_file* of)
+{
+ const off_t offset = this->offset();
+ const section_size_type oview_size =
+ convert_to_section_size_type(this->data_size());
+ unsigned char* const oview = of->get_output_view(offset, oview_size);
+
+ for (typename Unordered_set<Mips_symbol<size>*>::iterator
+ p = this->symbols_.begin();
+ p != this->symbols_.end();
+ ++p)
+ {
+ Mips_symbol<size>* sym = *p;
+ unsigned char* pov = oview + sym->la25_stub_offset();
+
+ Mips_address target = sym->value();
+ if (!sym->is_micromips())
+ {
+ elfcpp::Swap<32, big_endian>::writeval(pov,
+ la25_stub_entry[0] | (((target + 0x8000) >> 16) & 0xffff));
+ elfcpp::Swap<32, big_endian>::writeval(pov + 4,
+ la25_stub_entry[1] | ((target >> 2) & 0x3ffffff));
+ elfcpp::Swap<32, big_endian>::writeval(pov + 8,
+ la25_stub_entry[2] | (target & 0xffff));
+ elfcpp::Swap<32, big_endian>::writeval(pov + 12, la25_stub_entry[3]);
+ }
+ else
+ {
+ target |= 1;
+ // First stub instruction. Paste high 16-bits of the target.
+ elfcpp::Swap<16, big_endian>::writeval(pov,
+ la25_stub_micromips_entry[0]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2,
+ ((target + 0x8000) >> 16) & 0xffff);
+ // Second stub instruction. Paste low 26-bits of the target, shifted
+ // right by 1.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 4,
+ la25_stub_micromips_entry[2] | ((target >> 17) & 0x3ff));
+ elfcpp::Swap<16, big_endian>::writeval(pov + 6,
+ la25_stub_micromips_entry[3] | ((target >> 1) & 0xffff));
+ // Third stub instruction. Paste low 16-bits of the target.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 8,
+ la25_stub_micromips_entry[4]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 10, target & 0xffff);
+ // Fourth stub instruction.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 12,
+ la25_stub_micromips_entry[6]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 14,
+ la25_stub_micromips_entry[7]);
+ }
+ }
+
+ of->write_output_view(offset, oview_size, oview);
+}
+
+// Mips_output_data_plt methods.
+
+// The format of the first PLT entry in an O32 executable.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::plt0_entry_o32[] =
+{
+ 0x3c1c0000, // lui $28, %hi(&GOTPLT[0])
+ 0x8f990000, // lw $25, %lo(&GOTPLT[0])($28)
+ 0x279c0000, // addiu $28, $28, %lo(&GOTPLT[0])
+ 0x031cc023, // subu $24, $24, $28
+ 0x03e07821, // move $15, $31 # 32-bit move (addu)
+ 0x0018c082, // srl $24, $24, 2
+ 0x0320f809, // jalr $25
+ 0x2718fffe // subu $24, $24, 2
+};
+
+// The format of the first PLT entry in an N32 executable. Different
+// because gp ($28) is not available; we use t2 ($14) instead.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::plt0_entry_n32[] =
+{
+ 0x3c0e0000, // lui $14, %hi(&GOTPLT[0])
+ 0x8dd90000, // lw $25, %lo(&GOTPLT[0])($14)
+ 0x25ce0000, // addiu $14, $14, %lo(&GOTPLT[0])
+ 0x030ec023, // subu $24, $24, $14
+ 0x03e07821, // move $15, $31 # 32-bit move (addu)
+ 0x0018c082, // srl $24, $24, 2
+ 0x0320f809, // jalr $25
+ 0x2718fffe // subu $24, $24, 2
+};
+
+// The format of the first PLT entry in an N64 executable. Different
+// from N32 because of the increased size of GOT entries.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::plt0_entry_n64[] =
+{
+ 0x3c0e0000, // lui $14, %hi(&GOTPLT[0])
+ 0xddd90000, // ld $25, %lo(&GOTPLT[0])($14)
+ 0x25ce0000, // addiu $14, $14, %lo(&GOTPLT[0])
+ 0x030ec023, // subu $24, $24, $14
+ 0x03e07821, // move $15, $31 # 64-bit move (daddu)
+ 0x0018c0c2, // srl $24, $24, 3
+ 0x0320f809, // jalr $25
+ 0x2718fffe // subu $24, $24, 2
+};
+
+// The format of the microMIPS first PLT entry in an O32 executable.
+// We rely on v0 ($2) rather than t8 ($24) to contain the address
+// of the GOTPLT entry handled, so this stub may only be used when
+// all the subsequent PLT entries are microMIPS code too.
+//
+// The trailing NOP is for alignment and correct disassembly only.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::
+plt0_entry_micromips_o32[] =
+{
+ 0x7980, 0x0000, // addiupc $3, (&GOTPLT[0]) - .
+ 0xff23, 0x0000, // lw $25, 0($3)
+ 0x0535, // subu $2, $2, $3
+ 0x2525, // srl $2, $2, 2
+ 0x3302, 0xfffe, // subu $24, $2, 2
+ 0x0dff, // move $15, $31
+ 0x45f9, // jalrs $25
+ 0x0f83, // move $28, $3
+ 0x0c00 // nop
+};
+
+// The format of the microMIPS first PLT entry in an O32 executable
+// in the insn32 mode.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::
+plt0_entry_micromips32_o32[] =
+{
+ 0x41bc, 0x0000, // lui $28, %hi(&GOTPLT[0])
+ 0xff3c, 0x0000, // lw $25, %lo(&GOTPLT[0])($28)
+ 0x339c, 0x0000, // addiu $28, $28, %lo(&GOTPLT[0])
+ 0x0398, 0xc1d0, // subu $24, $24, $28
+ 0x001f, 0x7950, // move $15, $31
+ 0x0318, 0x1040, // srl $24, $24, 2
+ 0x03f9, 0x0f3c, // jalr $25
+ 0x3318, 0xfffe // subu $24, $24, 2
+};
+
+// The format of subsequent standard entries in the PLT.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::plt_entry[] =
+{
+ 0x3c0f0000, // lui $15, %hi(.got.plt entry)
+ 0x8df90000, // l[wd] $25, %lo(.got.plt entry)($15)
+ 0x03200008, // jr $25
+ 0x25f80000 // addiu $24, $15, %lo(.got.plt entry)
+};
+
+// The format of subsequent MIPS16 o32 PLT entries. We use v1 ($3) as a
+// temporary because t8 ($24) and t9 ($25) are not directly addressable.
+// Note that this differs from the GNU ld which uses both v0 ($2) and v1 ($3).
+// We cannot use v0 because MIPS16 call stubs from the CS toolchain expect
+// target function address in register v0.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::plt_entry_mips16_o32[] =
+{
+ 0xb303, // lw $3, 12($pc)
+ 0x651b, // move $24, $3
+ 0x9b60, // lw $3, 0($3)
+ 0xeb00, // jr $3
+ 0x653b, // move $25, $3
+ 0x6500, // nop
+ 0x0000, 0x0000 // .word (.got.plt entry)
+};
+
+// The format of subsequent microMIPS o32 PLT entries. We use v0 ($2)
+// as a temporary because t8 ($24) is not addressable with ADDIUPC.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::
+plt_entry_micromips_o32[] =
+{
+ 0x7900, 0x0000, // addiupc $2, (.got.plt entry) - .
+ 0xff22, 0x0000, // lw $25, 0($2)
+ 0x4599, // jr $25
+ 0x0f02 // move $24, $2
+};
+
+// The format of subsequent microMIPS o32 PLT entries in the insn32 mode.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_plt<size, big_endian>::
+plt_entry_micromips32_o32[] =
+{
+ 0x41af, 0x0000, // lui $15, %hi(.got.plt entry)
+ 0xff2f, 0x0000, // lw $25, %lo(.got.plt entry)($15)
+ 0x0019, 0x0f3c, // jr $25
+ 0x330f, 0x0000 // addiu $24, $15, %lo(.got.plt entry)
+};
+
+// Add an entry to the PLT for a symbol referenced by r_type relocation.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_plt<size, big_endian>::add_entry(Mips_symbol<size>* gsym,
+ unsigned int r_type)
+{
+ gold_assert(!gsym->has_plt_offset());
+
+ // Final PLT offset for a symbol will be set in method set_plt_offsets().
+ gsym->set_plt_offset(this->entry_count() * sizeof(plt_entry)
+ + sizeof(plt0_entry_o32));
+ this->symbols_.push_back(gsym);
+
+ // Record whether the relocation requires a standard MIPS
+ // or a compressed code entry.
+ if (jal_reloc(r_type))
+ {
+ if (r_type == elfcpp::R_MIPS_26)
+ gsym->set_needs_mips_plt(true);
+ else
+ gsym->set_needs_comp_plt(true);
+ }
+
+ section_offset_type got_offset = this->got_plt_->current_data_size();
+
+ // Every PLT entry needs a GOT entry which points back to the PLT
+ // entry (this will be changed by the dynamic linker, normally
+ // lazily when the function is called).
+ this->got_plt_->set_current_data_size(got_offset + size/8);
+
+ gsym->set_needs_dynsym_entry();
+ this->rel_->add_global(gsym, elfcpp::R_MIPS_JUMP_SLOT, this->got_plt_,
+ got_offset);
+}
+
+// Set final PLT offsets. For each symbol, determine whether standard or
+// compressed (MIPS16 or microMIPS) PLT entry is used.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_plt<size, big_endian>::set_plt_offsets()
+{
+ // The sizes of individual PLT entries.
+ unsigned int plt_mips_entry_size = this->standard_plt_entry_size();
+ unsigned int plt_comp_entry_size = (!this->target_->is_output_newabi()
+ ? this->compressed_plt_entry_size() : 0);
+
+ for (typename std::vector<Mips_symbol<size>*>::const_iterator
+ p = this->symbols_.begin(); p != this->symbols_.end(); ++p)
+ {
+ Mips_symbol<size>* mips_sym = *p;
+
+ // There are no defined MIPS16 or microMIPS PLT entries for n32 or n64,
+ // so always use a standard entry there.
+ //
+ // If the symbol has a MIPS16 call stub and gets a PLT entry, then
+ // all MIPS16 calls will go via that stub, and there is no benefit
+ // to having a MIPS16 entry. And in the case of call_stub a
+ // standard entry actually has to be used as the stub ends with a J
+ // instruction.
+ if (this->target_->is_output_newabi()
+ || mips_sym->has_mips16_call_stub()
+ || mips_sym->has_mips16_call_fp_stub())
+ {
+ mips_sym->set_needs_mips_plt(true);
+ mips_sym->set_needs_comp_plt(false);
+ }
+
+ // Otherwise, if there are no direct calls to the function, we
+ // have a free choice of whether to use standard or compressed
+ // entries. Prefer microMIPS entries if the object is known to
+ // contain microMIPS code, so that it becomes possible to create
+ // pure microMIPS binaries. Prefer standard entries otherwise,
+ // because MIPS16 ones are no smaller and are usually slower.
+ if (!mips_sym->needs_mips_plt() && !mips_sym->needs_comp_plt())
+ {
+ if (this->target_->is_output_micromips())
+ mips_sym->set_needs_comp_plt(true);
+ else
+ mips_sym->set_needs_mips_plt(true);
+ }
+
+ if (mips_sym->needs_mips_plt())
+ {
+ mips_sym->set_mips_plt_offset(this->plt_mips_offset_);
+ this->plt_mips_offset_ += plt_mips_entry_size;
+ }
+ if (mips_sym->needs_comp_plt())
+ {
+ mips_sym->set_comp_plt_offset(this->plt_comp_offset_);
+ this->plt_comp_offset_ += plt_comp_entry_size;
+ }
+ }
+
+ // Figure out the size of the PLT header if we know that we are using it.
+ if (this->plt_mips_offset_ + this->plt_comp_offset_ != 0)
+ this->plt_header_size_ = this->get_plt_header_size();
+}
+
+// Write out the PLT. This uses the hand-coded instructions above,
+// and adjusts them as needed.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_plt<size, big_endian>::do_write(Output_file* of)
+{
+ const off_t offset = this->offset();
+ const section_size_type oview_size =
+ convert_to_section_size_type(this->data_size());
+ unsigned char* const oview = of->get_output_view(offset, oview_size);
+
+ const off_t gotplt_file_offset = this->got_plt_->offset();
+ const section_size_type gotplt_size =
+ convert_to_section_size_type(this->got_plt_->data_size());
+ unsigned char* const gotplt_view = of->get_output_view(gotplt_file_offset,
+ gotplt_size);
+ unsigned char* pov = oview;
+
+ Mips_address plt_address = this->address();
+
+ // Calculate the address of .got.plt.
+ Mips_address gotplt_addr = this->got_plt_->address();
+ Mips_address gotplt_addr_high = ((gotplt_addr + 0x8000) >> 16) & 0xffff;
+ Mips_address gotplt_addr_low = gotplt_addr & 0xffff;
+
+ // The PLT sequence is not safe for N64 if .got.plt's address can
+ // not be loaded in two instructions.
+ gold_assert((gotplt_addr & ~(Mips_address) 0x7fffffff) == 0
+ || ~(gotplt_addr | 0x7fffffff) == 0);
+
+ // Write the PLT header.
+ const uint32_t* plt0_entry = this->get_plt_header_entry();
+ if (plt0_entry == plt0_entry_micromips_o32)
+ {
+ // Write microMIPS PLT header.
+ gold_assert(gotplt_addr % 4 == 0);
+
+ Mips_address gotpc_offset = gotplt_addr - ((plt_address | 3) ^ 3);
+
+ // ADDIUPC has a span of +/-16MB, check we're in range.
+ if (gotpc_offset + 0x1000000 >= 0x2000000)
+ {
+ gold_error(_(".got.plt offset of %ld from .plt beyond the range of "
+ "ADDIUPC"), (long)gotpc_offset);
+ return;
+ }
+
+ elfcpp::Swap<16, big_endian>::writeval(pov,
+ plt0_entry[0] | ((gotpc_offset >> 18) & 0x7f));
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2,
+ (gotpc_offset >> 2) & 0xffff);
+ pov += 4;
+ for (unsigned int i = 2;
+ i < (sizeof(plt0_entry_micromips_o32)
+ / sizeof(plt0_entry_micromips_o32[0]));
+ i++)
+ {
+ elfcpp::Swap<16, big_endian>::writeval(pov, plt0_entry[i]);
+ pov += 2;
+ }
+ }
+ else if (plt0_entry == plt0_entry_micromips32_o32)
+ {
+ // Write microMIPS PLT header in insn32 mode.
+ elfcpp::Swap<16, big_endian>::writeval(pov, plt0_entry[0]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2, gotplt_addr_high);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 4, plt0_entry[2]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 6, gotplt_addr_low);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 8, plt0_entry[4]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 10, gotplt_addr_low);
+ pov += 12;
+ for (unsigned int i = 6;
+ i < (sizeof(plt0_entry_micromips32_o32)
+ / sizeof(plt0_entry_micromips32_o32[0]));
+ i++)
+ {
+ elfcpp::Swap<16, big_endian>::writeval(pov, plt0_entry[i]);
+ pov += 2;
+ }
+ }
+ else
+ {
+ // Write standard PLT header.
+ elfcpp::Swap<32, big_endian>::writeval(pov,
+ plt0_entry[0] | gotplt_addr_high);
+ elfcpp::Swap<32, big_endian>::writeval(pov + 4,
+ plt0_entry[1] | gotplt_addr_low);
+ elfcpp::Swap<32, big_endian>::writeval(pov + 8,
+ plt0_entry[2] | gotplt_addr_low);
+ pov += 12;
+ for (int i = 3; i < 8; i++)
+ {
+ elfcpp::Swap<32, big_endian>::writeval(pov, plt0_entry[i]);
+ pov += 4;
+ }
+ }
+
+
+ unsigned char* gotplt_pov = gotplt_view;
+ unsigned int got_entry_size = size/8; // TODO(sasa): MIPS_ELF_GOT_SIZE
+
+ // The first two entries in .got.plt are reserved.
+ elfcpp::Swap<size, big_endian>::writeval(gotplt_pov, 0);
+ elfcpp::Swap<size, big_endian>::writeval(gotplt_pov + got_entry_size, 0);
+
+ unsigned int gotplt_offset = 2 * got_entry_size;
+ gotplt_pov += 2 * got_entry_size;
+
+ // Calculate the address of the PLT header.
+ Mips_address header_address = (plt_address
+ + (this->is_plt_header_compressed() ? 1 : 0));
+
+ // Initialize compressed PLT area view.
+ unsigned char* pov2 = pov + this->plt_mips_offset_;
+
+ // Write the PLT entries.
+ for (typename std::vector<Mips_symbol<size>*>::const_iterator
+ p = this->symbols_.begin();
+ p != this->symbols_.end();
+ ++p, gotplt_pov += got_entry_size, gotplt_offset += got_entry_size)
+ {
+ Mips_symbol<size>* mips_sym = *p;
+
+ // Calculate the address of the .got.plt entry.
+ uint32_t gotplt_entry_addr = (gotplt_addr + gotplt_offset);
+ uint32_t gotplt_entry_addr_hi = (((gotplt_entry_addr + 0x8000) >> 16)
+ & 0xffff);
+ uint32_t gotplt_entry_addr_lo = gotplt_entry_addr & 0xffff;
+
+ // Initially point the .got.plt entry at the PLT header.
+ if (this->target_->is_output_n64())
+ elfcpp::Swap<64, big_endian>::writeval(gotplt_pov, header_address);
+ else
+ elfcpp::Swap<32, big_endian>::writeval(gotplt_pov, header_address);
+
+ // Now handle the PLT itself. First the standard entry.
+ if (mips_sym->has_mips_plt_offset())
+ {
+ // Pick the load opcode (LW or LD).
+ uint64_t load = this->target_->is_output_n64() ? 0xdc000000
+ : 0x8c000000;
+
+ // Fill in the PLT entry itself.
+ elfcpp::Swap<32, big_endian>::writeval(pov,
+ plt_entry[0] | gotplt_entry_addr_hi);
+ elfcpp::Swap<32, big_endian>::writeval(pov + 4,
+ plt_entry[1] | gotplt_entry_addr_lo | load);
+ elfcpp::Swap<32, big_endian>::writeval(pov + 8, plt_entry[2]);
+ elfcpp::Swap<32, big_endian>::writeval(pov + 12,
+ plt_entry[3] | gotplt_entry_addr_lo);
+ pov += 16;
+ }
+
+ // Now the compressed entry. They come after any standard ones.
+ if (mips_sym->has_comp_plt_offset())
+ {
+ if (!this->target_->is_output_micromips())
+ {
+ // Write MIPS16 PLT entry.
+ const uint32_t* plt_entry = plt_entry_mips16_o32;
+
+ elfcpp::Swap<16, big_endian>::writeval(pov2, plt_entry[0]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 2, plt_entry[1]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 4, plt_entry[2]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 6, plt_entry[3]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 8, plt_entry[4]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 10, plt_entry[5]);
+ elfcpp::Swap<32, big_endian>::writeval(pov2 + 12,
+ gotplt_entry_addr);
+ pov2 += 16;
+ }
+ else if (this->target_->use_32bit_micromips_instructions())
+ {
+ // Write microMIPS PLT entry in insn32 mode.
+ const uint32_t* plt_entry = plt_entry_micromips32_o32;
+
+ elfcpp::Swap<16, big_endian>::writeval(pov2, plt_entry[0]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 2,
+ gotplt_entry_addr_hi);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 4, plt_entry[2]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 6,
+ gotplt_entry_addr_lo);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 8, plt_entry[4]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 10, plt_entry[5]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 12, plt_entry[6]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 14,
+ gotplt_entry_addr_lo);
+ pov2 += 16;
+ }
+ else
+ {
+ // Write microMIPS PLT entry.
+ const uint32_t* plt_entry = plt_entry_micromips_o32;
+
+ gold_assert(gotplt_entry_addr % 4 == 0);
+
+ Mips_address loc_address = plt_address + pov2 - oview;
+ int gotpc_offset = gotplt_entry_addr - ((loc_address | 3) ^ 3);
+
+ // ADDIUPC has a span of +/-16MB, check we're in range.
+ if (gotpc_offset + 0x1000000 >= 0x2000000)
+ {
+ gold_error(_(".got.plt offset of %ld from .plt beyond the "
+ "range of ADDIUPC"), (long)gotpc_offset);
+ return;
+ }
+
+ elfcpp::Swap<16, big_endian>::writeval(pov2,
+ plt_entry[0] | ((gotpc_offset >> 18) & 0x7f));
+ elfcpp::Swap<16, big_endian>::writeval(
+ pov2 + 2, (gotpc_offset >> 2) & 0xffff);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 4, plt_entry[2]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 6, plt_entry[3]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 8, plt_entry[4]);
+ elfcpp::Swap<16, big_endian>::writeval(pov2 + 10, plt_entry[5]);
+ pov2 += 12;
+ }
+ }
+ }
+
+ // Check the number of bytes written for standard entries.
+ gold_assert(static_cast<section_size_type>(
+ pov - oview - this->plt_header_size_) == this->plt_mips_offset_);
+ // Check the number of bytes written for compressed entries.
+ gold_assert((static_cast<section_size_type>(pov2 - pov)
+ == this->plt_comp_offset_));
+ // Check the total number of bytes written.
+ gold_assert(static_cast<section_size_type>(pov2 - oview) == oview_size);
+
+ gold_assert(static_cast<section_size_type>(gotplt_pov - gotplt_view)
+ == gotplt_size);
+
+ of->write_output_view(offset, oview_size, oview);
+ of->write_output_view(gotplt_file_offset, gotplt_size, gotplt_view);
+}
+
+// Mips_output_data_mips_stubs methods.
+
+// The format of the lazy binding stub when dynamic symbol count is less than
+// 64K, dynamic symbol index is less than 32K, and ABI is not N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_normal_1[4] =
+{
+ 0x8f998010, // lw t9,0x8010(gp)
+ 0x03e07821, // addu t7,ra,zero
+ 0x0320f809, // jalr t9,ra
+ 0x24180000 // addiu t8,zero,DYN_INDEX sign extended
+};
+
+// The format of the lazy binding stub when dynamic symbol count is less than
+// 64K, dynamic symbol index is less than 32K, and ABI is N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_normal_1_n64[4] =
+{
+ 0xdf998010, // ld t9,0x8010(gp)
+ 0x03e0782d, // daddu t7,ra,zero
+ 0x0320f809, // jalr t9,ra
+ 0x64180000 // daddiu t8,zero,DYN_INDEX sign extended
+};
+
+// The format of the lazy binding stub when dynamic symbol count is less than
+// 64K, dynamic symbol index is between 32K and 64K, and ABI is not N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_normal_2[4] =
+{
+ 0x8f998010, // lw t9,0x8010(gp)
+ 0x03e07821, // addu t7,ra,zero
+ 0x0320f809, // jalr t9,ra
+ 0x34180000 // ori t8,zero,DYN_INDEX unsigned
+};
+
+// The format of the lazy binding stub when dynamic symbol count is less than
+// 64K, dynamic symbol index is between 32K and 64K, and ABI is N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_normal_2_n64[4] =
+{
+ 0xdf998010, // ld t9,0x8010(gp)
+ 0x03e0782d, // daddu t7,ra,zero
+ 0x0320f809, // jalr t9,ra
+ 0x34180000 // ori t8,zero,DYN_INDEX unsigned
+};
+
+// The format of the lazy binding stub when dynamic symbol count is greater than
+// 64K, and ABI is not N64.
+template<int size, bool big_endian>
+const uint32_t Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_big[5] =
+{
+ 0x8f998010, // lw t9,0x8010(gp)
+ 0x03e07821, // addu t7,ra,zero
+ 0x3c180000, // lui t8,DYN_INDEX
+ 0x0320f809, // jalr t9,ra
+ 0x37180000 // ori t8,t8,DYN_INDEX
+};
+
+// The format of the lazy binding stub when dynamic symbol count is greater than
+// 64K, and ABI is N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_big_n64[5] =
+{
+ 0xdf998010, // ld t9,0x8010(gp)
+ 0x03e0782d, // daddu t7,ra,zero
+ 0x3c180000, // lui t8,DYN_INDEX
+ 0x0320f809, // jalr t9,ra
+ 0x37180000 // ori t8,t8,DYN_INDEX
+};
+
+// microMIPS stubs.
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// less than 64K, dynamic symbol index is less than 32K, and ABI is not N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_micromips_normal_1[] =
+{
+ 0xff3c, 0x8010, // lw t9,0x8010(gp)
+ 0x0dff, // move t7,ra
+ 0x45d9, // jalr t9
+ 0x3300, 0x0000 // addiu t8,zero,DYN_INDEX sign extended
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// less than 64K, dynamic symbol index is less than 32K, and ABI is N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::
+lazy_stub_micromips_normal_1_n64[] =
+{
+ 0xdf3c, 0x8010, // ld t9,0x8010(gp)
+ 0x0dff, // move t7,ra
+ 0x45d9, // jalr t9
+ 0x5f00, 0x0000 // daddiu t8,zero,DYN_INDEX sign extended
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol
+// count is less than 64K, dynamic symbol index is between 32K and 64K,
+// and ABI is not N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_micromips_normal_2[] =
+{
+ 0xff3c, 0x8010, // lw t9,0x8010(gp)
+ 0x0dff, // move t7,ra
+ 0x45d9, // jalr t9
+ 0x5300, 0x0000 // ori t8,zero,DYN_INDEX unsigned
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol
+// count is less than 64K, dynamic symbol index is between 32K and 64K,
+// and ABI is N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::
+lazy_stub_micromips_normal_2_n64[] =
+{
+ 0xdf3c, 0x8010, // ld t9,0x8010(gp)
+ 0x0dff, // move t7,ra
+ 0x45d9, // jalr t9
+ 0x5300, 0x0000 // ori t8,zero,DYN_INDEX unsigned
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// greater than 64K, and ABI is not N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_micromips_big[] =
+{
+ 0xff3c, 0x8010, // lw t9,0x8010(gp)
+ 0x0dff, // move t7,ra
+ 0x41b8, 0x0000, // lui t8,DYN_INDEX
+ 0x45d9, // jalr t9
+ 0x5318, 0x0000 // ori t8,t8,DYN_INDEX
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// greater than 64K, and ABI is N64.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_micromips_big_n64[] =
+{
+ 0xdf3c, 0x8010, // ld t9,0x8010(gp)
+ 0x0dff, // move t7,ra
+ 0x41b8, 0x0000, // lui t8,DYN_INDEX
+ 0x45d9, // jalr t9
+ 0x5318, 0x0000 // ori t8,t8,DYN_INDEX
+};
+
+// 32-bit microMIPS stubs.
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// less than 64K, dynamic symbol index is less than 32K, ABI is not N64, and we
+// can use only 32-bit instructions.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::
+lazy_stub_micromips32_normal_1[] =
+{
+ 0xff3c, 0x8010, // lw t9,0x8010(gp)
+ 0x001f, 0x7950, // addu t7,ra,zero
+ 0x03f9, 0x0f3c, // jalr ra,t9
+ 0x3300, 0x0000 // addiu t8,zero,DYN_INDEX sign extended
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// less than 64K, dynamic symbol index is less than 32K, ABI is N64, and we can
+// use only 32-bit instructions.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::
+lazy_stub_micromips32_normal_1_n64[] =
+{
+ 0xdf3c, 0x8010, // ld t9,0x8010(gp)
+ 0x581f, 0x7950, // daddu t7,ra,zero
+ 0x03f9, 0x0f3c, // jalr ra,t9
+ 0x5f00, 0x0000 // daddiu t8,zero,DYN_INDEX sign extended
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol
+// count is less than 64K, dynamic symbol index is between 32K and 64K,
+// ABI is not N64, and we can use only 32-bit instructions.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::
+lazy_stub_micromips32_normal_2[] =
+{
+ 0xff3c, 0x8010, // lw t9,0x8010(gp)
+ 0x001f, 0x7950, // addu t7,ra,zero
+ 0x03f9, 0x0f3c, // jalr ra,t9
+ 0x5300, 0x0000 // ori t8,zero,DYN_INDEX unsigned
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol
+// count is less than 64K, dynamic symbol index is between 32K and 64K,
+// ABI is N64, and we can use only 32-bit instructions.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::
+lazy_stub_micromips32_normal_2_n64[] =
+{
+ 0xdf3c, 0x8010, // ld t9,0x8010(gp)
+ 0x581f, 0x7950, // daddu t7,ra,zero
+ 0x03f9, 0x0f3c, // jalr ra,t9
+ 0x5300, 0x0000 // ori t8,zero,DYN_INDEX unsigned
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// greater than 64K, ABI is not N64, and we can use only 32-bit instructions.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_micromips32_big[] =
+{
+ 0xff3c, 0x8010, // lw t9,0x8010(gp)
+ 0x001f, 0x7950, // addu t7,ra,zero
+ 0x41b8, 0x0000, // lui t8,DYN_INDEX
+ 0x03f9, 0x0f3c, // jalr ra,t9
+ 0x5318, 0x0000 // ori t8,t8,DYN_INDEX
+};
+
+// The format of the microMIPS lazy binding stub when dynamic symbol count is
+// greater than 64K, ABI is N64, and we can use only 32-bit instructions.
+template<int size, bool big_endian>
+const uint32_t
+Mips_output_data_mips_stubs<size, big_endian>::lazy_stub_micromips32_big_n64[] =
+{
+ 0xdf3c, 0x8010, // ld t9,0x8010(gp)
+ 0x581f, 0x7950, // daddu t7,ra,zero
+ 0x41b8, 0x0000, // lui t8,DYN_INDEX
+ 0x03f9, 0x0f3c, // jalr ra,t9
+ 0x5318, 0x0000 // ori t8,t8,DYN_INDEX
+};
+
+// Create entry for a symbol.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_mips_stubs<size, big_endian>::make_entry(
+ Mips_symbol<size>* gsym)
+{
+ if (!gsym->has_lazy_stub() && !gsym->has_plt_offset())
+ {
+ this->symbols_.insert(gsym);
+ gsym->set_has_lazy_stub(true);
+ }
+}
+
+// Remove entry for a symbol.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_mips_stubs<size, big_endian>::remove_entry(
+ Mips_symbol<size>* gsym)
+{
+ if (gsym->has_lazy_stub())
+ {
+ this->symbols_.erase(gsym);
+ gsym->set_has_lazy_stub(false);
+ }
+}
+
+// Set stub offsets for symbols. This method expects that the number of
+// entries in dynamic symbol table is set.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_mips_stubs<size, big_endian>::set_lazy_stub_offsets()
+{
+ gold_assert(this->dynsym_count_ != -1U);
+
+ if (this->stub_offsets_are_set_)
+ return;
+
+ unsigned int stub_size = this->stub_size();
+ unsigned int offset = 0;
+ for (typename Unordered_set<Mips_symbol<size>*>::const_iterator
+ p = this->symbols_.begin();
+ p != this->symbols_.end();
+ ++p, offset += stub_size)
+ {
+ Mips_symbol<size>* mips_sym = *p;
+ mips_sym->set_lazy_stub_offset(offset);
+ }
+ this->stub_offsets_are_set_ = true;
+}
+
+template<int size, bool big_endian>
+void
+Mips_output_data_mips_stubs<size, big_endian>::set_needs_dynsym_value()
+{
+ for (typename Unordered_set<Mips_symbol<size>*>::const_iterator
+ p = this->symbols_.begin(); p != this->symbols_.end(); ++p)
+ {
+ Mips_symbol<size>* sym = *p;
+ if (sym->is_from_dynobj())
+ sym->set_needs_dynsym_value();
+ }
+}
+
+// Write out the .MIPS.stubs. This uses the hand-coded instructions and
+// adjusts them as needed.
+
+template<int size, bool big_endian>
+void
+Mips_output_data_mips_stubs<size, big_endian>::do_write(Output_file* of)
+{
+ const off_t offset = this->offset();
+ const section_size_type oview_size =
+ convert_to_section_size_type(this->data_size());
+ unsigned char* const oview = of->get_output_view(offset, oview_size);
+
+ bool big_stub = this->dynsym_count_ > 0x10000;
+
+ unsigned char* pov = oview;
+ for (typename Unordered_set<Mips_symbol<size>*>::const_iterator
+ p = this->symbols_.begin(); p != this->symbols_.end(); ++p)
+ {
+ Mips_symbol<size>* sym = *p;
+ const uint32_t* lazy_stub;
+ bool n64 = this->target_->is_output_n64();
+
+ if (!this->target_->is_output_micromips())
+ {
+ // Write standard (non-microMIPS) stub.
+ if (!big_stub)
+ {
+ if (sym->dynsym_index() & ~0x7fff)
+ // Dynsym index is between 32K and 64K.
+ lazy_stub = n64 ? lazy_stub_normal_2_n64 : lazy_stub_normal_2;
+ else
+ // Dynsym index is less than 32K.
+ lazy_stub = n64 ? lazy_stub_normal_1_n64 : lazy_stub_normal_1;
+ }
+ else
+ lazy_stub = n64 ? lazy_stub_big_n64 : lazy_stub_big;
+
+ unsigned int i = 0;
+ elfcpp::Swap<32, big_endian>::writeval(pov, lazy_stub[i]);
+ elfcpp::Swap<32, big_endian>::writeval(pov + 4, lazy_stub[i + 1]);
+ pov += 8;
+
+ i += 2;
+ if (big_stub)
+ {
+ // LUI instruction of the big stub. Paste high 16 bits of the
+ // dynsym index.
+ elfcpp::Swap<32, big_endian>::writeval(pov,
+ lazy_stub[i] | ((sym->dynsym_index() >> 16) & 0x7fff));
+ pov += 4;
+ i += 1;
+ }
+ elfcpp::Swap<32, big_endian>::writeval(pov, lazy_stub[i]);
+ // Last stub instruction. Paste low 16 bits of the dynsym index.
+ elfcpp::Swap<32, big_endian>::writeval(pov + 4,
+ lazy_stub[i + 1] | (sym->dynsym_index() & 0xffff));
+ pov += 8;
+ }
+ else if (this->target_->use_32bit_micromips_instructions())
+ {
+ // Write microMIPS stub in insn32 mode.
+ if (!big_stub)
+ {
+ if (sym->dynsym_index() & ~0x7fff)
+ // Dynsym index is between 32K and 64K.
+ lazy_stub = n64 ? lazy_stub_micromips32_normal_2_n64
+ : lazy_stub_micromips32_normal_2;
+ else
+ // Dynsym index is less than 32K.
+ lazy_stub = n64 ? lazy_stub_micromips32_normal_1_n64
+ : lazy_stub_micromips32_normal_1;
+ }
+ else
+ lazy_stub = n64 ? lazy_stub_micromips32_big_n64
+ : lazy_stub_micromips32_big;
+
+ unsigned int i = 0;
+ // First stub instruction. We emit 32-bit microMIPS instructions by
+ // emitting two 16-bit parts because on microMIPS the 16-bit part of
+ // the instruction where the opcode is must always come first, for
+ // both little and big endian.
+ elfcpp::Swap<16, big_endian>::writeval(pov, lazy_stub[i]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2, lazy_stub[i + 1]);
+ // Second stub instruction.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 4, lazy_stub[i + 2]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 6, lazy_stub[i + 3]);
+ pov += 8;
+ i += 4;
+ if (big_stub)
+ {
+ // LUI instruction of the big stub. Paste high 16 bits of the
+ // dynsym index.
+ elfcpp::Swap<16, big_endian>::writeval(pov, lazy_stub[i]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2,
+ (sym->dynsym_index() >> 16) & 0x7fff);
+ pov += 4;
+ i += 2;
+ }
+ elfcpp::Swap<16, big_endian>::writeval(pov, lazy_stub[i]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2, lazy_stub[i + 1]);
+ // Last stub instruction. Paste low 16 bits of the dynsym index.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 4, lazy_stub[i + 2]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 6,
+ sym->dynsym_index() & 0xffff);
+ pov += 8;
+ }
+ else
+ {
+ // Write microMIPS stub.
+ if (!big_stub)
+ {
+ if (sym->dynsym_index() & ~0x7fff)
+ // Dynsym index is between 32K and 64K.
+ lazy_stub = n64 ? lazy_stub_micromips_normal_2_n64
+ : lazy_stub_micromips_normal_2;
+ else
+ // Dynsym index is less than 32K.
+ lazy_stub = n64 ? lazy_stub_micromips_normal_1_n64
+ : lazy_stub_micromips_normal_1;
+ }
+ else
+ lazy_stub = n64 ? lazy_stub_micromips_big_n64
+ : lazy_stub_micromips_big;
+
+ unsigned int i = 0;
+ // First stub instruction. We emit 32-bit microMIPS instructions by
+ // emitting two 16-bit parts because on microMIPS the 16-bit part of
+ // the instruction where the opcode is must always come first, for
+ // both little and big endian.
+ elfcpp::Swap<16, big_endian>::writeval(pov, lazy_stub[i]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2, lazy_stub[i + 1]);
+ // Second stub instruction.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 4, lazy_stub[i + 2]);
+ pov += 6;
+ i += 3;
+ if (big_stub)
+ {
+ // LUI instruction of the big stub. Paste high 16 bits of the
+ // dynsym index.
+ elfcpp::Swap<16, big_endian>::writeval(pov, lazy_stub[i]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2,
+ (sym->dynsym_index() >> 16) & 0x7fff);
+ pov += 4;
+ i += 2;
+ }
+ elfcpp::Swap<16, big_endian>::writeval(pov, lazy_stub[i]);
+ // Last stub instruction. Paste low 16 bits of the dynsym index.
+ elfcpp::Swap<16, big_endian>::writeval(pov + 2, lazy_stub[i + 1]);
+ elfcpp::Swap<16, big_endian>::writeval(pov + 4,
+ sym->dynsym_index() & 0xffff);
+ pov += 6;
+ }
+ }
+
+ // We always allocate 20 bytes for every stub, because final dynsym count is
+ // not known in method do_finalize_sections. There are 4 unused bytes per
+ // stub if final dynsym count is less than 0x10000.
+ unsigned int used = pov - oview;
+ unsigned int unused = big_stub ? 0 : this->symbols_.size() * 4;
+ gold_assert(static_cast<section_size_type>(used + unused) == oview_size);
+
+ // Fill the unused space with zeroes.
+ // TODO(sasa): Can we strip unused bytes during the relaxation?
+ if (unused > 0)
+ memset(pov, 0, unused);
+
+ of->write_output_view(offset, oview_size, oview);
+}
+
+// Mips_output_section_reginfo methods.
+
+template<int size, bool big_endian>
+void
+Mips_output_section_reginfo<size, big_endian>::do_write(Output_file* of)
+{
+ off_t offset = this->offset();
+ off_t data_size = this->data_size();
+
+ unsigned char* view = of->get_output_view(offset, data_size);
+ elfcpp::Swap<size, big_endian>::writeval(view, this->gprmask_);
+ elfcpp::Swap<size, big_endian>::writeval(view + 4, this->cprmask1_);
+ elfcpp::Swap<size, big_endian>::writeval(view + 8, this->cprmask2_);
+ elfcpp::Swap<size, big_endian>::writeval(view + 12, this->cprmask3_);
+ elfcpp::Swap<size, big_endian>::writeval(view + 16, this->cprmask4_);
+ // Write the gp value.
+ elfcpp::Swap<size, big_endian>::writeval(view + 20,
+ this->target_->gp_value());
+
+ of->write_output_view(offset, data_size, view);
+}
+
+// Mips_copy_relocs methods.
+
+// Emit any saved relocs.
+
+template<int sh_type, int size, bool big_endian>
+void
+Mips_copy_relocs<sh_type, size, big_endian>::emit_mips(
+ Output_data_reloc<sh_type, true, size, big_endian>* reloc_section,
+ Symbol_table* symtab, Layout* layout, Target_mips<size, big_endian>* target)
+{
+ for (typename Copy_relocs<sh_type, size, big_endian>::
+ Copy_reloc_entries::iterator p = this->entries_.begin();
+ p != this->entries_.end();
+ ++p)
+ emit_entry(*p, reloc_section, symtab, layout, target);
+
+ // We no longer need the saved information.
+ this->entries_.clear();
+}
+
+// Emit the reloc if appropriate.
+
+template<int sh_type, int size, bool big_endian>
+void
+Mips_copy_relocs<sh_type, size, big_endian>::emit_entry(
+ Copy_reloc_entry& entry,
+ Output_data_reloc<sh_type, true, size, big_endian>* reloc_section,
+ Symbol_table* symtab, Layout* layout, Target_mips<size, big_endian>* target)
+{
+ // If the symbol is no longer defined in a dynamic object, then we
+ // emitted a COPY relocation, and we do not want to emit this
+ // dynamic relocation.
+ if (!entry.sym_->is_from_dynobj())
+ return;
+
+ bool can_make_dynamic = (entry.reloc_type_ == elfcpp::R_MIPS_32
+ || entry.reloc_type_ == elfcpp::R_MIPS_REL32
+ || entry.reloc_type_ == elfcpp::R_MIPS_64);
+
+ Mips_symbol<size>* sym = Mips_symbol<size>::as_mips_sym(entry.sym_);
+ if (can_make_dynamic && !sym->has_static_relocs())
+ {
+ Mips_relobj<size, big_endian>* object =
+ Mips_relobj<size, big_endian>::as_mips_relobj(entry.relobj_);
+ target->got_section(symtab, layout)->record_global_got_symbol(
+ sym, object, entry.reloc_type_, true, false);
+ if (!symbol_references_local(sym, sym->should_add_dynsym_entry(symtab)))
+ target->rel_dyn_section(layout)->add_global(sym, elfcpp::R_MIPS_REL32,
+ entry.output_section_, entry.relobj_, entry.shndx_, entry.address_);
+ else
+ target->rel_dyn_section(layout)->add_symbolless_global_addend(
+ sym, elfcpp::R_MIPS_REL32, entry.output_section_, entry.relobj_,
+ entry.shndx_, entry.address_);
+ }
+ else
+ this->make_copy_reloc(symtab, layout,
+ static_cast<Sized_symbol<size>*>(entry.sym_),
+ reloc_section);
+}
+
+// Target_mips methods.
+
+// Return the value to use for a dynamic symbol which requires special
+// treatment. This is how we support equality comparisons of function
+// pointers across shared library boundaries, as described in the
+// processor specific ABI supplement.
+
+template<int size, bool big_endian>
+uint64_t
+Target_mips<size, big_endian>::do_dynsym_value(const Symbol* gsym) const
+{
+ uint64_t value = 0;
+ const Mips_symbol<size>* mips_sym = Mips_symbol<size>::as_mips_sym(gsym);
+
+ if (!mips_sym->has_lazy_stub())
+ {
+ if (mips_sym->has_plt_offset())
+ {
+ // We distinguish between PLT entries and lazy-binding stubs by
+ // giving the former an st_other value of STO_MIPS_PLT. Set the
+ // value to the stub address if there are any relocations in the
+ // binary where pointer equality matters.
+ if (mips_sym->pointer_equality_needed())
+ {
+ // Prefer a standard MIPS PLT entry.
+ if (mips_sym->has_mips_plt_offset())
+ value = this->plt_section()->mips_entry_address(mips_sym);
+ else
+ value = this->plt_section()->comp_entry_address(mips_sym) + 1;
+ }
+ else
+ value = 0;
+ }
+ }
+ else
+ {
+ // First, set stub offsets for symbols. This method expects that the
+ // number of entries in dynamic symbol table is set.
+ this->mips_stubs_section()->set_lazy_stub_offsets();
+
+ // The run-time linker uses the st_value field of the symbol
+ // to reset the global offset table entry for this external
+ // to its stub address when unlinking a shared object.
+ value = this->mips_stubs_section()->stub_address(mips_sym);
+ }
+
+ if (mips_sym->has_mips16_fn_stub())
+ {
+ // If we have a MIPS16 function with a stub, the dynamic symbol must
+ // refer to the stub, since only the stub uses the standard calling
+ // conventions.
+ value = mips_sym->template
+ get_mips16_fn_stub<big_endian>()->output_address();
+ }
+
+ return value;
+}
+
+// Get the dynamic reloc section, creating it if necessary. It's always
+// .rel.dyn, even for MIPS64.
+
+template<int size, bool big_endian>
+typename Target_mips<size, big_endian>::Reloc_section*
+Target_mips<size, big_endian>::rel_dyn_section(Layout* layout)
+{
+ if (this->rel_dyn_ == NULL)
+ {
+ gold_assert(layout != NULL);
+ this->rel_dyn_ = new Reloc_section(parameters->options().combreloc());
+ layout->add_output_section_data(".rel.dyn", elfcpp::SHT_REL,
+ elfcpp::SHF_ALLOC, this->rel_dyn_,
+ ORDER_DYNAMIC_RELOCS, false);
+
+ // First entry in .rel.dyn has to be null.
+ // This is hack - we define dummy output data and set its address to 0,
+ // and define absolute R_MIPS_NONE relocation with offset 0 against it.
+ // This ensures that the entry is null.
+ Output_data* od = new Output_data_zero_fill(0, 0);
+ od->set_address(0);
+ this->rel_dyn_->add_absolute(elfcpp::R_MIPS_NONE, od, 0);
+ }
+ return this->rel_dyn_;
+}
+
+// Get the GOT section, creating it if necessary.
+
+template<int size, bool big_endian>
+Mips_output_data_got<size, big_endian>*
+Target_mips<size, big_endian>::got_section(Symbol_table* symtab,
+ Layout* layout)
+{
+ if (this->got_ == NULL)
+ {
+ gold_assert(symtab != NULL && layout != NULL);
+
+ this->got_ = new Mips_output_data_got<size, big_endian>(this, symtab,
+ layout);
+ layout->add_output_section_data(".got", elfcpp::SHT_PROGBITS,
+ (elfcpp::SHF_ALLOC | elfcpp::SHF_WRITE |
+ elfcpp::SHF_MIPS_GPREL),
+ this->got_, ORDER_DATA, false);
+
+ // Define _GLOBAL_OFFSET_TABLE_ at the start of the .got section.
+ symtab->define_in_output_data("_GLOBAL_OFFSET_TABLE_", NULL,
+ Symbol_table::PREDEFINED,
+ this->got_,
+ 0, 0, elfcpp::STT_OBJECT,
+ elfcpp::STB_GLOBAL,
+ elfcpp::STV_DEFAULT, 0,
+ false, false);
+ }
+
+ return this->got_;
+}
+
+// Calculate value of _gp symbol.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::set_gp(Layout* layout, Symbol_table* symtab)
+{
+ if (this->gp_ != NULL)
+ return;
+
+ Output_data* section = layout->find_output_section(".got");
+ if (section == NULL)
+ {
+ // If there is no .got section, gp should be based on .sdata.
+ // TODO(sasa): This is probably not needed. This was needed for older
+ // MIPS architectures which accessed both GOT and .sdata section using
+ // gp-relative addressing. Modern Mips Linux ELF architectures don't
+ // access .sdata using gp-relative addressing.
+ for (Layout::Section_list::const_iterator
+ p = layout->section_list().begin();
+ p != layout->section_list().end();
+ ++p)
+ {
+ if (strcmp((*p)->name(), ".sdata") == 0)
+ {
+ section = *p;
+ break;
+ }
+ }
+ }
+
+ Sized_symbol<size>* gp =
+ static_cast<Sized_symbol<size>*>(symtab->lookup("_gp"));
+ if (gp != NULL)
+ {
+ if (gp->source() != Symbol::IS_CONSTANT && section != NULL)
+ gp->init_output_data(gp->name(), NULL, section, MIPS_GP_OFFSET, 0,
+ elfcpp::STT_OBJECT,
+ elfcpp::STB_GLOBAL,
+ elfcpp::STV_DEFAULT, 0,
+ false, false);
+ this->gp_ = gp;
+ }
+ else if (section != NULL)
+ {
+ gp = static_cast<Sized_symbol<size>*>(symtab->define_in_output_data(
+ "_gp", NULL, Symbol_table::PREDEFINED,
+ section, MIPS_GP_OFFSET, 0,
+ elfcpp::STT_OBJECT,
+ elfcpp::STB_GLOBAL,
+ elfcpp::STV_DEFAULT,
+ 0, false, false));
+ this->gp_ = gp;
+ }
+}
+
+// Set the dynamic symbol indexes. INDEX is the index of the first
+// global dynamic symbol. Pointers to the symbols are stored into the
+// vector SYMS. The names are added to DYNPOOL. This returns an
+// updated dynamic symbol index.
+
+template<int size, bool big_endian>
+unsigned int
+Target_mips<size, big_endian>::do_set_dynsym_indexes(
+ std::vector<Symbol*>* dyn_symbols, unsigned int index,
+ std::vector<Symbol*>* syms, Stringpool* dynpool,
+ Versions* versions, Symbol_table* symtab) const
+{
+ std::vector<Symbol*> non_got_symbols;
+ std::vector<Symbol*> got_symbols;
+
+ reorder_dyn_symbols<size, big_endian>(dyn_symbols, &non_got_symbols,
+ &got_symbols);
+
+ for (std::vector<Symbol*>::iterator p = non_got_symbols.begin();
+ p != non_got_symbols.end();
+ ++p)
+ {
+ Symbol* sym = *p;
+
+ // Note that SYM may already have a dynamic symbol index, since
+ // some symbols appear more than once in the symbol table, with
+ // and without a version.
+
+ if (!sym->has_dynsym_index())
+ {
+ sym->set_dynsym_index(index);
+ ++index;
+ syms->push_back(sym);
+ dynpool->add(sym->name(), false, NULL);
+
+ // Record any version information.
+ if (sym->version() != NULL)
+ versions->record_version(symtab, dynpool, sym);
+
+ // If the symbol is defined in a dynamic object and is
+ // referenced in a regular object, then mark the dynamic
+ // object as needed. This is used to implement --as-needed.
+ if (sym->is_from_dynobj() && sym->in_reg())
+ sym->object()->set_is_needed();
+ }
+ }
+
+ for (std::vector<Symbol*>::iterator p = got_symbols.begin();
+ p != got_symbols.end();
+ ++p)
+ {
+ Symbol* sym = *p;
+ if (!sym->has_dynsym_index())
+ {
+ // Record any version information.
+ if (sym->version() != NULL)
+ versions->record_version(symtab, dynpool, sym);
+ }
+ }
+
+ index = versions->finalize(symtab, index, syms);
+
+ int got_sym_count = 0;
+ for (std::vector<Symbol*>::iterator p = got_symbols.begin();
+ p != got_symbols.end();
+ ++p)
+ {
+ Symbol* sym = *p;
+
+ if (!sym->has_dynsym_index())
+ {
+ ++got_sym_count;
+ sym->set_dynsym_index(index);
+ ++index;
+ syms->push_back(sym);
+ dynpool->add(sym->name(), false, NULL);
+
+ // If the symbol is defined in a dynamic object and is
+ // referenced in a regular object, then mark the dynamic
+ // object as needed. This is used to implement --as-needed.
+ if (sym->is_from_dynobj() && sym->in_reg())
+ sym->object()->set_is_needed();
+ }
+ }
+
+ // Set index of the first symbol that has .got entry.
+ this->got_->set_first_global_got_dynsym_index(
+ got_sym_count > 0 ? index - got_sym_count : -1U);
+
+ if (this->mips_stubs_ != NULL)
+ this->mips_stubs_->set_dynsym_count(index);
+
+ return index;
+}
+
+// Create a PLT entry for a global symbol referenced by r_type relocation.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::make_plt_entry(Symbol_table* symtab,
+ Layout* layout,
+ Mips_symbol<size>* gsym,
+ unsigned int r_type)
+{
+ if (gsym->has_lazy_stub() || gsym->has_plt_offset())
+ return;
+
+ if (this->plt_ == NULL)
+ {
+ // Create the GOT section first.
+ this->got_section(symtab, layout);
+
+ this->got_plt_ = new Output_data_space(4, "** GOT PLT");
+ layout->add_output_section_data(".got.plt", elfcpp::SHT_PROGBITS,
+ (elfcpp::SHF_ALLOC | elfcpp::SHF_WRITE),
+ this->got_plt_, ORDER_DATA, false);
+
+ // The first two entries are reserved.
+ this->got_plt_->set_current_data_size(2 * size/8);
+
+ this->plt_ = new Mips_output_data_plt<size, big_endian>(layout,
+ this->got_plt_,
+ this);
+ layout->add_output_section_data(".plt", elfcpp::SHT_PROGBITS,
+ (elfcpp::SHF_ALLOC
+ | elfcpp::SHF_EXECINSTR),
+ this->plt_, ORDER_PLT, false);
+ }
+
+ this->plt_->add_entry(gsym, r_type);
+}
+
+
+// Get the .MIPS.stubs section, creating it if necessary.
+
+template<int size, bool big_endian>
+Mips_output_data_mips_stubs<size, big_endian>*
+Target_mips<size, big_endian>::mips_stubs_section(Layout* layout)
+{
+ if (this->mips_stubs_ == NULL)
+ {
+ this->mips_stubs_ =
+ new Mips_output_data_mips_stubs<size, big_endian>(this);
+ layout->add_output_section_data(".MIPS.stubs", elfcpp::SHT_PROGBITS,
+ (elfcpp::SHF_ALLOC
+ | elfcpp::SHF_EXECINSTR),
+ this->mips_stubs_, ORDER_PLT, false);
+ }
+ return this->mips_stubs_;
+}
+
+// Get the LA25 stub section, creating it if necessary.
+
+template<int size, bool big_endian>
+Mips_output_data_la25_stub<size, big_endian>*
+Target_mips<size, big_endian>::la25_stub_section(Layout* layout)
+{
+ if (this->la25_stub_ == NULL)
+ {
+ this->la25_stub_ = new Mips_output_data_la25_stub<size, big_endian>();
+ layout->add_output_section_data(".text", elfcpp::SHT_PROGBITS,
+ (elfcpp::SHF_ALLOC
+ | elfcpp::SHF_EXECINSTR),
+ this->la25_stub_, ORDER_TEXT, false);
+ }
+ return this->la25_stub_;
+}
+
+// Process the relocations to determine unreferenced sections for
+// garbage collection.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::gc_process_relocs(
+ Symbol_table* symtab,
+ Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ unsigned int,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ size_t local_symbol_count,
+ const unsigned char* plocal_symbols)
+{
+ typedef Target_mips<size, big_endian> Mips;
+ typedef typename Target_mips<size, big_endian>::Scan Scan;
+
+ gold::gc_process_relocs<size, big_endian, Mips, elfcpp::SHT_REL, Scan,
+ typename Target_mips::Relocatable_size_for_reloc>(
+ symtab,
+ layout,
+ this,
+ object,
+ data_shndx,
+ prelocs,
+ reloc_count,
+ output_section,
+ needs_special_offset_handling,
+ local_symbol_count,
+ plocal_symbols);
+}
+
+// Scan relocations for a section.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::scan_relocs(
+ Symbol_table* symtab,
+ Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ size_t local_symbol_count,
+ const unsigned char* plocal_symbols)
+{
+ typedef Target_mips<size, big_endian> Mips;
+ typedef typename Target_mips<size, big_endian>::Scan Scan;
+
+ if (sh_type == elfcpp::SHT_REL)
+ gold::scan_relocs<size, big_endian, Mips, elfcpp::SHT_REL, Scan>(
+ symtab,
+ layout,
+ this,
+ object,
+ data_shndx,
+ prelocs,
+ reloc_count,
+ output_section,
+ needs_special_offset_handling,
+ local_symbol_count,
+ plocal_symbols);
+ else if (sh_type == elfcpp::SHT_RELA)
+ gold::scan_relocs<size, big_endian, Mips, elfcpp::SHT_RELA, Scan>(
+ symtab,
+ layout,
+ this,
+ object,
+ data_shndx,
+ prelocs,
+ reloc_count,
+ output_section,
+ needs_special_offset_handling,
+ local_symbol_count,
+ plocal_symbols);
+}
+
+template<int size, bool big_endian>
+bool
+Target_mips<size, big_endian>::mips_32bit_flags(elfcpp::Elf_Word flags)
+{
+ return ((flags & elfcpp::EF_MIPS_32BITMODE) != 0
+ || (flags & elfcpp::EF_MIPS_ABI) == elfcpp::E_MIPS_ABI_O32
+ || (flags & elfcpp::EF_MIPS_ABI) == elfcpp::E_MIPS_ABI_EABI32
+ || (flags & elfcpp::EF_MIPS_ARCH) == elfcpp::E_MIPS_ARCH_1
+ || (flags & elfcpp::EF_MIPS_ARCH) == elfcpp::E_MIPS_ARCH_2
+ || (flags & elfcpp::EF_MIPS_ARCH) == elfcpp::E_MIPS_ARCH_32
+ || (flags & elfcpp::EF_MIPS_ARCH) == elfcpp::E_MIPS_ARCH_32R2);
+}
+
+// Return the MACH for a MIPS e_flags value.
+template<int size, bool big_endian>
+unsigned int
+Target_mips<size, big_endian>::elf_mips_mach(elfcpp::Elf_Word flags)
+{
+ switch (flags & elfcpp::EF_MIPS_MACH)
+ {
+ case elfcpp::E_MIPS_MACH_3900:
+ return mach_mips3900;
+
+ case elfcpp::E_MIPS_MACH_4010:
+ return mach_mips4010;
+
+ case elfcpp::E_MIPS_MACH_4100:
+ return mach_mips4100;
+
+ case elfcpp::E_MIPS_MACH_4111:
+ return mach_mips4111;
+
+ case elfcpp::E_MIPS_MACH_4120:
+ return mach_mips4120;
+
+ case elfcpp::E_MIPS_MACH_4650:
+ return mach_mips4650;
+
+ case elfcpp::E_MIPS_MACH_5400:
+ return mach_mips5400;
+
+ case elfcpp::E_MIPS_MACH_5500:
+ return mach_mips5500;
+
+ case elfcpp::E_MIPS_MACH_9000:
+ return mach_mips9000;
+
+ case elfcpp::E_MIPS_MACH_SB1:
+ return mach_mips_sb1;
+
+ case elfcpp::E_MIPS_MACH_LS2E:
+ return mach_mips_loongson_2e;
+
+ case elfcpp::E_MIPS_MACH_LS2F:
+ return mach_mips_loongson_2f;
+
+ case elfcpp::E_MIPS_MACH_LS3A:
+ return mach_mips_loongson_3a;
+
+ case elfcpp::E_MIPS_MACH_OCTEON2:
+ return mach_mips_octeon2;
+
+ case elfcpp::E_MIPS_MACH_OCTEON:
+ return mach_mips_octeon;
+
+ case elfcpp::E_MIPS_MACH_XLR:
+ return mach_mips_xlr;
+
+ default:
+ switch (flags & elfcpp::EF_MIPS_ARCH)
+ {
+ default:
+ case elfcpp::E_MIPS_ARCH_1:
+ return mach_mips3000;
+
+ case elfcpp::E_MIPS_ARCH_2:
+ return mach_mips6000;
+
+ case elfcpp::E_MIPS_ARCH_3:
+ return mach_mips4000;
+
+ case elfcpp::E_MIPS_ARCH_4:
+ return mach_mips8000;
+
+ case elfcpp::E_MIPS_ARCH_5:
+ return mach_mips5;
+
+ case elfcpp::E_MIPS_ARCH_32:
+ return mach_mipsisa32;
+
+ case elfcpp::E_MIPS_ARCH_64:
+ return mach_mipsisa64;
+
+ case elfcpp::E_MIPS_ARCH_32R2:
+ return mach_mipsisa32r2;
+
+ case elfcpp::E_MIPS_ARCH_64R2:
+ return mach_mipsisa64r2;
+ }
+ }
+
+ return 0;
+}
+
+// Check whether machine EXTENSION is an extension of machine BASE.
+template<int size, bool big_endian>
+bool
+Target_mips<size, big_endian>::mips_mach_extends(unsigned int base,
+ unsigned int extension)
+{
+ if (extension == base)
+ return true;
+
+ if ((base == mach_mipsisa32)
+ && this->mips_mach_extends(mach_mipsisa64, extension))
+ return true;
+
+ if ((base == mach_mipsisa32r2)
+ && this->mips_mach_extends(mach_mipsisa64r2, extension))
+ return true;
+
+ for (unsigned int i = 0; i < this->mips_mach_extensions_.size(); ++i)
+ if (extension == this->mips_mach_extensions_[i].first)
+ {
+ extension = this->mips_mach_extensions_[i].second;
+ if (extension == base)
+ return true;
+ }
+
+ return false;
+}
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::merge_processor_specific_flags(
+ const std::string& name, elfcpp::Elf_Word in_flags,
+ unsigned char in_ei_class, bool dyn_obj)
+{
+ // If flags are not set yet, just copy them.
+ if (!this->are_processor_specific_flags_set())
+ {
+ this->set_processor_specific_flags(in_flags);
+ this->ei_class_ = in_ei_class;
+ this->mach_ = this->elf_mips_mach(in_flags);
+ return;
+ }
+
+ elfcpp::Elf_Word new_flags = in_flags;
+ elfcpp::Elf_Word old_flags = this->processor_specific_flags();
+ elfcpp::Elf_Word merged_flags = this->processor_specific_flags();
+ merged_flags |= new_flags & elfcpp::EF_MIPS_NOREORDER;
+
+ // Check flag compatibility.
+ new_flags &= ~elfcpp::EF_MIPS_NOREORDER;
+ old_flags &= ~elfcpp::EF_MIPS_NOREORDER;
+
+ // Some IRIX 6 BSD-compatibility objects have this bit set. It
+ // doesn't seem to matter.
+ new_flags &= ~elfcpp::EF_MIPS_XGOT;
+ old_flags &= ~elfcpp::EF_MIPS_XGOT;
+
+ // MIPSpro generates ucode info in n64 objects. Again, we should
+ // just be able to ignore this.
+ new_flags &= ~elfcpp::EF_MIPS_UCODE;
+ old_flags &= ~elfcpp::EF_MIPS_UCODE;
+
+ // DSOs should only be linked with CPIC code.
+ if (dyn_obj)
+ new_flags |= elfcpp::EF_MIPS_PIC | elfcpp::EF_MIPS_CPIC;
+
+ if (new_flags == old_flags)
+ {
+ this->set_processor_specific_flags(merged_flags);
+ return;
+ }
+
+ if (((new_flags & (elfcpp::EF_MIPS_PIC | elfcpp::EF_MIPS_CPIC)) != 0)
+ != ((old_flags & (elfcpp::EF_MIPS_PIC | elfcpp::EF_MIPS_CPIC)) != 0))
+ gold_warning(_("%s: linking abicalls files with non-abicalls files"),
+ name.c_str());
+
+ if (new_flags & (elfcpp::EF_MIPS_PIC | elfcpp::EF_MIPS_CPIC))
+ merged_flags |= elfcpp::EF_MIPS_CPIC;
+ if (!(new_flags & elfcpp::EF_MIPS_PIC))
+ merged_flags &= ~elfcpp::EF_MIPS_PIC;
+
+ new_flags &= ~(elfcpp::EF_MIPS_PIC | elfcpp::EF_MIPS_CPIC);
+ old_flags &= ~(elfcpp::EF_MIPS_PIC | elfcpp::EF_MIPS_CPIC);
+
+ // Compare the ISAs.
+ if (mips_32bit_flags(old_flags) != mips_32bit_flags(new_flags))
+ gold_error(_("%s: linking 32-bit code with 64-bit code"), name.c_str());
+ else if (!this->mips_mach_extends(this->elf_mips_mach(in_flags), this->mach_))
+ {
+ // Output ISA isn't the same as, or an extension of, input ISA.
+ if (this->mips_mach_extends(this->mach_, this->elf_mips_mach(in_flags)))
+ {
+ // Copy the architecture info from input object to output. Also copy
+ // the 32-bit flag (if set) so that we continue to recognise
+ // output as a 32-bit binary.
+ this->mach_ = this->elf_mips_mach(in_flags);
+ merged_flags &= ~(elfcpp::EF_MIPS_ARCH | elfcpp::EF_MIPS_MACH);
+ merged_flags |= (new_flags & (elfcpp::EF_MIPS_ARCH
+ | elfcpp::EF_MIPS_MACH | elfcpp::EF_MIPS_32BITMODE));
+
+ // Copy across the ABI flags if output doesn't use them
+ // and if that was what caused us to treat input object as 32-bit.
+ if ((old_flags & elfcpp::EF_MIPS_ABI) == 0
+ && this->mips_32bit_flags(new_flags)
+ && !this->mips_32bit_flags(new_flags & ~elfcpp::EF_MIPS_ABI))
+ merged_flags |= new_flags & elfcpp::EF_MIPS_ABI;
+ }
+ else
+ // The ISAs aren't compatible.
+ gold_error(_("%s: linking %s module with previous %s modules"),
+ name.c_str(), this->elf_mips_mach_name(in_flags),
+ this->elf_mips_mach_name(merged_flags));
+ }
+
+ new_flags &= (~(elfcpp::EF_MIPS_ARCH | elfcpp::EF_MIPS_MACH
+ | elfcpp::EF_MIPS_32BITMODE));
+ old_flags &= (~(elfcpp::EF_MIPS_ARCH | elfcpp::EF_MIPS_MACH
+ | elfcpp::EF_MIPS_32BITMODE));
+
+ // Compare ABIs. The 64-bit ABI does not use EF_MIPS_ABI. But, it does set
+ // EI_CLASS differently from any 32-bit ABI.
+ if ((new_flags & elfcpp::EF_MIPS_ABI) != (old_flags & elfcpp::EF_MIPS_ABI)
+ || (in_ei_class != this->ei_class_))
+ {
+ // Only error if both are set (to different values).
+ if (((new_flags & elfcpp::EF_MIPS_ABI)
+ && (old_flags & elfcpp::EF_MIPS_ABI))
+ || (in_ei_class != this->ei_class_))
+ gold_error(_("%s: ABI mismatch: linking %s module with "
+ "previous %s modules"), name.c_str(),
+ this->elf_mips_abi_name(in_flags, in_ei_class),
+ this->elf_mips_abi_name(merged_flags, this->ei_class_));
+
+ new_flags &= ~elfcpp::EF_MIPS_ABI;
+ old_flags &= ~elfcpp::EF_MIPS_ABI;
+ }
+
+ // Compare ASEs. Forbid linking MIPS16 and microMIPS ASE modules together
+ // and allow arbitrary mixing of the remaining ASEs (retain the union).
+ if ((new_flags & elfcpp::EF_MIPS_ARCH_ASE)
+ != (old_flags & elfcpp::EF_MIPS_ARCH_ASE))
+ {
+ int old_micro = old_flags & elfcpp::EF_MIPS_ARCH_ASE_MICROMIPS;
+ int new_micro = new_flags & elfcpp::EF_MIPS_ARCH_ASE_MICROMIPS;
+ int old_m16 = old_flags & elfcpp::EF_MIPS_ARCH_ASE_M16;
+ int new_m16 = new_flags & elfcpp::EF_MIPS_ARCH_ASE_M16;
+ int micro_mis = old_m16 && new_micro;
+ int m16_mis = old_micro && new_m16;
+
+ if (m16_mis || micro_mis)
+ gold_error(_("%s: ASE mismatch: linking %s module with "
+ "previous %s modules"), name.c_str(),
+ m16_mis ? "MIPS16" : "microMIPS",
+ m16_mis ? "microMIPS" : "MIPS16");
+
+ merged_flags |= new_flags & elfcpp::EF_MIPS_ARCH_ASE;
+
+ new_flags &= ~ elfcpp::EF_MIPS_ARCH_ASE;
+ old_flags &= ~ elfcpp::EF_MIPS_ARCH_ASE;
+ }
+
+ // Warn about any other mismatches.
+ if (new_flags != old_flags)
+ gold_error(_("%s: uses different e_flags (0x%x) fields than previous "
+ "modules (0x%x)"), name.c_str(), new_flags, old_flags);
+
+ this->set_processor_specific_flags(merged_flags);
+}
+
+// Adjust ELF file header.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::do_adjust_elf_header(
+ unsigned char* view,
+ int len)
+{
+ gold_assert(len == elfcpp::Elf_sizes<size>::ehdr_size);
+
+ elfcpp::Ehdr<size, big_endian> ehdr(view);
+ unsigned char e_ident[elfcpp::EI_NIDENT];
+ memcpy(e_ident, ehdr.get_e_ident(), elfcpp::EI_NIDENT);
+
+ e_ident[elfcpp::EI_CLASS] = this->ei_class_;
+
+ elfcpp::Ehdr_write<size, big_endian> oehdr(view);
+ oehdr.put_e_ident(e_ident);
+ if (this->entry_symbol_is_compressed_)
+ oehdr.put_e_entry(ehdr.get_e_entry() + 1);
+}
+
+// do_make_elf_object to override the same function in the base class.
+// We need to use a target-specific sub-class of
+// Sized_relobj_file<size, big_endian> to store Mips specific information.
+// Hence we need to have our own ELF object creation.
+
+template<int size, bool big_endian>
+Object*
+Target_mips<size, big_endian>::do_make_elf_object(
+ const std::string& name,
+ Input_file* input_file,
+ off_t offset, const elfcpp::Ehdr<size, big_endian>& ehdr)
+{
+ int et = ehdr.get_e_type();
+ // ET_EXEC files are valid input for --just-symbols/-R,
+ // and we treat them as relocatable objects.
+ if (et == elfcpp::ET_REL
+ || (et == elfcpp::ET_EXEC && input_file->just_symbols()))
+ {
+ Mips_relobj<size, big_endian>* obj =
+ new Mips_relobj<size, big_endian>(name, input_file, offset, ehdr);
+ obj->setup();
+ return obj;
+ }
+ else if (et == elfcpp::ET_DYN)
+ {
+ // TODO(sasa): Should we create Mips_dynobj?
+ return Target::do_make_elf_object(name, input_file, offset, ehdr);
+ }
+ else
+ {
+ gold_error(_("%s: unsupported ELF file type %d"),
+ name.c_str(), et);
+ return NULL;
+ }
+}
+
+// Finalize the sections.
+
+template <int size, bool big_endian>
+void
+Target_mips<size, big_endian>::do_finalize_sections(Layout* layout,
+ const Input_objects* input_objects,
+ Symbol_table* symtab)
+{
+ // Add +1 to MIPS16 and microMIPS init_ and _fini symbols so that DT_INIT and
+ // DT_FINI have correct values.
+ Mips_symbol<size>* init = static_cast<Mips_symbol<size>*>(
+ symtab->lookup(parameters->options().init()));
+ if (init != NULL && (init->is_mips16() || init->is_micromips()))
+ init->set_value(init->value() | 1);
+ Mips_symbol<size>* fini = static_cast<Mips_symbol<size>*>(
+ symtab->lookup(parameters->options().fini()));
+ if (fini != NULL && (fini->is_mips16() || fini->is_micromips()))
+ fini->set_value(fini->value() | 1);
+
+ // Check whether the entry symbol is mips16 or micromips. This is needed to
+ // adjust entry address in ELF header.
+ Mips_symbol<size>* entry =
+ static_cast<Mips_symbol<size>*>(symtab->lookup(this->entry_symbol_name()));
+ this->entry_symbol_is_compressed_ = (entry != NULL && (entry->is_mips16()
+ || entry->is_micromips()));
+
+ if (!parameters->doing_static_link()
+ && (strcmp(parameters->options().hash_style(), "gnu") == 0
+ || strcmp(parameters->options().hash_style(), "both") == 0))
+ {
+ // .gnu.hash and the MIPS ABI require .dynsym to be sorted in different
+ // ways. .gnu.hash needs symbols to be grouped by hash code whereas the
+ // MIPS ABI requires a mapping between the GOT and the symbol table.
+ gold_error(".gnu.hash is incompatible with the MIPS ABI");
+ }
+
+ // Check whether the final section that was scanned has HI16 or GOT16
+ // relocations without the corresponding LO16 part.
+ if (this->got16_addends_.size() > 0)
+ gold_error("Can't find matching LO16 reloc");
+
+ // Set _gp value.
+ this->set_gp(layout, symtab);
+
+ // Check for any mips16 stub sections that we can discard.
+ if (!parameters->options().relocatable())
+ {
+ for (Input_objects::Relobj_iterator p = input_objects->relobj_begin();
+ p != input_objects->relobj_end();
+ ++p)
+ {
+ Mips_relobj<size, big_endian>* object =
+ Mips_relobj<size, big_endian>::as_mips_relobj(*p);
+ object->discard_mips16_stub_sections(symtab);
+ }
+ }
+
+ // Merge processor-specific flags.
+ for (Input_objects::Relobj_iterator p = input_objects->relobj_begin();
+ p != input_objects->relobj_end();
+ ++p)
+ {
+ Mips_relobj<size, big_endian>* relobj =
+ Mips_relobj<size, big_endian>::as_mips_relobj(*p);
+
+ Input_file::Format format = relobj->input_file()->format();
+ if (format == Input_file::FORMAT_ELF)
+ {
+ // Read processor-specific flags in ELF file header.
+ const unsigned char* pehdr = relobj->get_view(
+ elfcpp::file_header_offset,
+ elfcpp::Elf_sizes<size>::ehdr_size,
+ true, false);
+
+ elfcpp::Ehdr<size, big_endian> ehdr(pehdr);
+ elfcpp::Elf_Word in_flags = ehdr.get_e_flags();
+ unsigned char ei_class = ehdr.get_e_ident()[elfcpp::EI_CLASS];
+
+ this->merge_processor_specific_flags(relobj->name(), in_flags,
+ ei_class, false);
+ }
+ }
+
+ for (Input_objects::Dynobj_iterator p = input_objects->dynobj_begin();
+ p != input_objects->dynobj_end();
+ ++p)
+ {
+ Sized_dynobj<size, big_endian>* dynobj =
+ static_cast<Sized_dynobj<size, big_endian>*>(*p);
+
+ // Read processor-specific flags.
+ const unsigned char* pehdr = dynobj->get_view(elfcpp::file_header_offset,
+ elfcpp::Elf_sizes<size>::ehdr_size,
+ true, false);
+
+ elfcpp::Ehdr<size, big_endian> ehdr(pehdr);
+ elfcpp::Elf_Word in_flags = ehdr.get_e_flags();
+ unsigned char ei_class = ehdr.get_e_ident()[elfcpp::EI_CLASS];
+
+ this->merge_processor_specific_flags(dynobj->name(), in_flags, ei_class,
+ true);
+ }
+
+ // Merge .reginfo contents of input objects.
+ Valtype gprmask = 0;
+ Valtype cprmask1 = 0;
+ Valtype cprmask2 = 0;
+ Valtype cprmask3 = 0;
+ Valtype cprmask4 = 0;
+ for (Input_objects::Relobj_iterator p = input_objects->relobj_begin();
+ p != input_objects->relobj_end();
+ ++p)
+ {
+ Mips_relobj<size, big_endian>* relobj =
+ Mips_relobj<size, big_endian>::as_mips_relobj(*p);
+
+ gprmask |= relobj->gprmask();
+ cprmask1 |= relobj->cprmask1();
+ cprmask2 |= relobj->cprmask2();
+ cprmask3 |= relobj->cprmask3();
+ cprmask4 |= relobj->cprmask4();
+ }
+
+ if (this->plt_ != NULL)
+ {
+ // Set final PLT offsets for symbols.
+ this->plt_section()->set_plt_offsets();
+
+ // Define _PROCEDURE_LINKAGE_TABLE_ at the start of the .plt section.
+ // Set STO_MICROMIPS flag if the output has microMIPS code, but only if
+ // there are no standard PLT entries present.
+ unsigned char nonvis = 0;
+ if (this->is_output_micromips()
+ && !this->plt_section()->has_standard_entries())
+ nonvis = elfcpp::STO_MICROMIPS >> 2;
+ symtab->define_in_output_data("_PROCEDURE_LINKAGE_TABLE_", NULL,
+ Symbol_table::PREDEFINED,
+ this->plt_,
+ 0, 0, elfcpp::STT_FUNC,
+ elfcpp::STB_LOCAL,
+ elfcpp::STV_DEFAULT, nonvis,
+ false, false);
+ }
+
+ if (this->mips_stubs_ != NULL)
+ {
+ // Define _MIPS_STUBS_ at the start of the .MIPS.stubs section.
+ unsigned char nonvis = 0;
+ if (this->is_output_micromips())
+ nonvis = elfcpp::STO_MICROMIPS >> 2;
+ symtab->define_in_output_data("_MIPS_STUBS_", NULL,
+ Symbol_table::PREDEFINED,
+ this->mips_stubs_,
+ 0, 0, elfcpp::STT_FUNC,
+ elfcpp::STB_LOCAL,
+ elfcpp::STV_DEFAULT, nonvis,
+ false, false);
+ }
+
+ if (!parameters->options().relocatable() && !parameters->doing_static_link())
+ // In case there is no .got section, create one.
+ this->got_section(symtab, layout);
+
+ // Emit any relocs we saved in an attempt to avoid generating COPY
+ // relocs.
+ if (this->copy_relocs_.any_saved_relocs())
+ this->copy_relocs_.emit_mips(this->rel_dyn_section(layout), symtab, layout,
+ this);
+
+ // Emit dynamic relocs.
+ for (typename std::vector<Dyn_reloc>::iterator p = this->dyn_relocs_.begin();
+ p != this->dyn_relocs_.end();
+ ++p)
+ p->emit(this->rel_dyn_section(layout), this->got_section(), symtab);
+
+ if (this->has_got_section())
+ this->got_section()->lay_out_got(layout, symtab, input_objects);
+
+ if (this->mips_stubs_ != NULL)
+ this->mips_stubs_->set_needs_dynsym_value();
+
+ // Check for functions that might need $25 to be valid on entry.
+ // TODO(sasa): Can we do this without iterating over all symbols?
+ typedef Symbol_visitor_check_symbols<size, big_endian> Symbol_visitor;
+ symtab->for_all_symbols<size, Symbol_visitor>(Symbol_visitor(this, layout,
+ symtab));
+
+ // Add NULL segment.
+ if (!parameters->options().relocatable())
+ layout->make_output_segment(elfcpp::PT_NULL, 0);
+
+ for (Layout::Section_list::const_iterator p = layout->section_list().begin();
+ p != layout->section_list().end();
+ ++p)
+ {
+ if ((*p)->type() == elfcpp::SHT_MIPS_REGINFO)
+ {
+ Mips_output_section_reginfo<size, big_endian>* reginfo =
+ Mips_output_section_reginfo<size, big_endian>::
+ as_mips_output_section_reginfo(*p);
+
+ reginfo->set_masks(gprmask, cprmask1, cprmask2, cprmask3, cprmask4);
+
+ if (!parameters->options().relocatable())
+ {
+ Output_segment* reginfo_segment =
+ layout->make_output_segment(elfcpp::PT_MIPS_REGINFO,
+ elfcpp::PF_R);
+ reginfo_segment->add_output_section_to_nonload(reginfo,
+ elfcpp::PF_R);
+ }
+ }
+ }
+
+ // Fill in some more dynamic tags.
+ // TODO(sasa): Add more dynamic tags.
+ const Reloc_section* rel_plt = (this->plt_ == NULL
+ ? NULL : this->plt_->rel_plt());
+ layout->add_target_dynamic_tags(true, this->got_, rel_plt,
+ this->rel_dyn_, true, false);
+
+ Output_data_dynamic* const odyn = layout->dynamic_data();
+ if (odyn != NULL
+ && !parameters->options().relocatable()
+ && !parameters->doing_static_link())
+ {
+ unsigned int d_val;
+ // This element holds a 32-bit version id for the Runtime
+ // Linker Interface. This will start at integer value 1.
+ d_val = 0x01;
+ odyn->add_constant(elfcpp::DT_MIPS_RLD_VERSION, d_val);
+
+ // Dynamic flags
+ d_val = elfcpp::RHF_NOTPOT;
+ odyn->add_constant(elfcpp::DT_MIPS_FLAGS, d_val);
+
+ // Save layout for using when emiting custom dynamic tags.
+ this->layout_ = layout;
+
+ // This member holds the base address of the segment.
+ odyn->add_custom(elfcpp::DT_MIPS_BASE_ADDRESS);
+
+ // This member holds the number of entries in the .dynsym section.
+ odyn->add_custom(elfcpp::DT_MIPS_SYMTABNO);
+
+ // This member holds the index of the first dynamic symbol
+ // table entry that corresponds to an entry in the global offset table.
+ odyn->add_custom(elfcpp::DT_MIPS_GOTSYM);
+
+ // This member holds the number of local GOT entries.
+ odyn->add_constant(elfcpp::DT_MIPS_LOCAL_GOTNO,
+ this->got_->get_local_gotno());
+
+ if (this->plt_ != NULL)
+ // DT_MIPS_PLTGOT dynamic tag
+ odyn->add_section_address(elfcpp::DT_MIPS_PLTGOT, this->got_plt_);
+ }
+ }
+
+// Get the custom dynamic tag value.
+template<int size, bool big_endian>
+unsigned int
+Target_mips<size, big_endian>::do_dynamic_tag_custom_value(elfcpp::DT tag) const
+{
+ switch (tag)
+ {
+ case elfcpp::DT_MIPS_BASE_ADDRESS:
+ {
+ // The base address of the segment.
+ // At this point, the segment list has been sorted into final order,
+ // so just return vaddr of the first readable PT_LOAD segment.
+ Output_segment* seg =
+ this->layout_->find_output_segment(elfcpp::PT_LOAD, elfcpp::PF_R, 0);
+ gold_assert(seg != NULL);
+ return seg->vaddr();
+ }
+
+ case elfcpp::DT_MIPS_SYMTABNO:
+ // The number of entries in the .dynsym section.
+ return this->get_dt_mips_symtabno();
+
+ case elfcpp::DT_MIPS_GOTSYM:
+ {
+ // The index of the first dynamic symbol table entry that corresponds
+ // to an entry in the GOT.
+ if (this->got_->first_global_got_dynsym_index() != -1U)
+ return this->got_->first_global_got_dynsym_index();
+ else
+ // In case if we don't have global GOT symbols we default to setting
+ // DT_MIPS_GOTSYM to the same value as DT_MIPS_SYMTABNO.
+ return this->get_dt_mips_symtabno();
+ }
+
+ default:
+ gold_error(_("Unknown dynamic tag 0x%x"), (unsigned int)tag);
+ }
+
+ return (unsigned int)-1;
+}
+
+// Relocate section data.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::relocate_section(
+ const Relocate_info<size, big_endian>* relinfo,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ unsigned char* view,
+ Mips_address address,
+ section_size_type view_size,
+ const Reloc_symbol_changes* reloc_symbol_changes)
+{
+ typedef Target_mips<size, big_endian> Mips;
+ typedef typename Target_mips<size, big_endian>::Relocate Mips_relocate;
+
+ if (sh_type == elfcpp::SHT_REL)
+ gold::relocate_section<size, big_endian, Mips, elfcpp::SHT_REL,
+ Mips_relocate, gold::Default_comdat_behavior>(
+ relinfo,
+ this,
+ prelocs,
+ reloc_count,
+ output_section,
+ needs_special_offset_handling,
+ view,
+ address,
+ view_size,
+ reloc_symbol_changes);
+ else if (sh_type == elfcpp::SHT_RELA)
+ gold::relocate_section<size, big_endian, Mips, elfcpp::SHT_RELA,
+ Mips_relocate, gold::Default_comdat_behavior>(
+ relinfo,
+ this,
+ prelocs,
+ reloc_count,
+ output_section,
+ needs_special_offset_handling,
+ view,
+ address,
+ view_size,
+ reloc_symbol_changes);
+}
+
+// Return the size of a relocation while scanning during a relocatable
+// link.
+
+template<int size, bool big_endian>
+unsigned int
+Target_mips<size, big_endian>::Relocatable_size_for_reloc::get_size_for_reloc(
+ unsigned int r_type,
+ Relobj* object)
+{
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_NONE:
+ case elfcpp::R_MIPS_TLS_DTPMOD64:
+ case elfcpp::R_MIPS_TLS_DTPREL64:
+ case elfcpp::R_MIPS_TLS_TPREL64:
+ return 0;
+
+ case elfcpp::R_MIPS_32:
+ case elfcpp::R_MIPS_TLS_DTPMOD32:
+ case elfcpp::R_MIPS_TLS_DTPREL32:
+ case elfcpp::R_MIPS_TLS_TPREL32:
+ case elfcpp::R_MIPS_REL32:
+ case elfcpp::R_MIPS_PC32:
+ case elfcpp::R_MIPS_GPREL32:
+ case elfcpp::R_MIPS_JALR:
+ return 4;
+
+ case elfcpp::R_MIPS_16:
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MIPS_LO16:
+ case elfcpp::R_MIPS_GPREL16:
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MIPS16_LO16:
+ case elfcpp::R_MIPS_PC16:
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MIPS_TLS_DTPREL_HI16:
+ case elfcpp::R_MIPS_TLS_DTPREL_LO16:
+ case elfcpp::R_MIPS_TLS_TPREL_HI16:
+ case elfcpp::R_MIPS_TLS_TPREL_LO16:
+ case elfcpp::R_MIPS16_GPREL:
+ case elfcpp::R_MIPS_GOT_DISP:
+ case elfcpp::R_MIPS_LITERAL:
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MIPS_GOT_OFST:
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ return 2;
+
+ // These relocations are not byte sized
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MIPS16_26:
+ return 4;
+
+ case elfcpp::R_MIPS_COPY:
+ case elfcpp::R_MIPS_JUMP_SLOT:
+ object->error(_("unexpected reloc %u in object file"), r_type);
+ return 0;
+
+ default:
+ object->error(_("unsupported reloc %u in object file"), r_type);
+ return 0;
+ }
+}
+
+// Scan the relocs during a relocatable link.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::scan_relocatable_relocs(
+ Symbol_table* symtab,
+ Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ size_t local_symbol_count,
+ const unsigned char* plocal_symbols,
+ Relocatable_relocs* rr)
+{
+ gold_assert(sh_type == elfcpp::SHT_REL);
+
+ typedef Mips_scan_relocatable_relocs<big_endian, elfcpp::SHT_REL,
+ Relocatable_size_for_reloc> Scan_relocatable_relocs;
+
+ gold::scan_relocatable_relocs<size, big_endian, elfcpp::SHT_REL,
+ Scan_relocatable_relocs>(
+ symtab,
+ layout,
+ object,
+ data_shndx,
+ prelocs,
+ reloc_count,
+ output_section,
+ needs_special_offset_handling,
+ local_symbol_count,
+ plocal_symbols,
+ rr);
+}
+
+// Emit relocations for a section.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::relocate_relocs(
+ const Relocate_info<size, big_endian>* relinfo,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ typename elfcpp::Elf_types<size>::Elf_Off
+ offset_in_output_section,
+ const Relocatable_relocs* rr,
+ unsigned char* view,
+ Mips_address view_address,
+ section_size_type view_size,
+ unsigned char* reloc_view,
+ section_size_type reloc_view_size)
+{
+ gold_assert(sh_type == elfcpp::SHT_REL);
+
+ gold::relocate_relocs<size, big_endian, elfcpp::SHT_REL>(
+ relinfo,
+ prelocs,
+ reloc_count,
+ output_section,
+ offset_in_output_section,
+ rr,
+ view,
+ view_address,
+ view_size,
+ reloc_view,
+ reloc_view_size);
+}
+
+// Perform target-specific processing in a relocatable link. This is
+// only used if we use the relocation strategy RELOC_SPECIAL.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::relocate_special_relocatable(
+ const Relocate_info<size, big_endian>* relinfo,
+ unsigned int sh_type,
+ const unsigned char* preloc_in,
+ size_t relnum,
+ Output_section* output_section,
+ typename elfcpp::Elf_types<size>::Elf_Off offset_in_output_section,
+ unsigned char* view,
+ Mips_address view_address,
+ section_size_type,
+ unsigned char* preloc_out)
+{
+ // We can only handle REL type relocation sections.
+ gold_assert(sh_type == elfcpp::SHT_REL);
+
+ typedef typename Reloc_types<elfcpp::SHT_REL, size, big_endian>::Reloc
+ Reltype;
+ typedef typename Reloc_types<elfcpp::SHT_REL, size, big_endian>::Reloc_write
+ Reltype_write;
+
+ typedef Mips_relocate_functions<size, big_endian> Reloc_funcs;
+
+ const Mips_address invalid_address = static_cast<Mips_address>(0) - 1;
+
+ Mips_relobj<size, big_endian>* object =
+ Mips_relobj<size, big_endian>::as_mips_relobj(relinfo->object);
+ const unsigned int local_count = object->local_symbol_count();
+
+ Reltype reloc(preloc_in);
+ Reltype_write reloc_write(preloc_out);
+
+ elfcpp::Elf_types<32>::Elf_WXword r_info = reloc.get_r_info();
+ const unsigned int r_sym = elfcpp::elf_r_sym<size>(r_info);
+ const unsigned int r_type = elfcpp::elf_r_type<size>(r_info);
+
+ // Get the new symbol index.
+ // We only use RELOC_SPECIAL strategy in local relocations.
+ gold_assert(r_sym < local_count);
+
+ // We are adjusting a section symbol. We need to find
+ // the symbol table index of the section symbol for
+ // the output section corresponding to input section
+ // in which this symbol is defined.
+ bool is_ordinary;
+ unsigned int shndx = object->local_symbol_input_shndx(r_sym, &is_ordinary);
+ gold_assert(is_ordinary);
+ Output_section* os = object->output_section(shndx);
+ gold_assert(os != NULL);
+ gold_assert(os->needs_symtab_index());
+ unsigned int new_symndx = os->symtab_index();
+
+ // Get the new offset--the location in the output section where
+ // this relocation should be applied.
+
+ Mips_address offset = reloc.get_r_offset();
+ Mips_address new_offset;
+ if (offset_in_output_section != invalid_address)
+ new_offset = offset + offset_in_output_section;
+ else
+ {
+ section_offset_type sot_offset =
+ convert_types<section_offset_type, Mips_address>(offset);
+ section_offset_type new_sot_offset =
+ output_section->output_offset(object, relinfo->data_shndx,
+ sot_offset);
+ gold_assert(new_sot_offset != -1);
+ new_offset = new_sot_offset;
+ }
+
+ // In an object file, r_offset is an offset within the section.
+ // In an executable or dynamic object, generated by
+ // --emit-relocs, r_offset is an absolute address.
+ if (!parameters->options().relocatable())
+ {
+ new_offset += view_address;
+ if (offset_in_output_section != invalid_address)
+ new_offset -= offset_in_output_section;
+ }
+
+ reloc_write.put_r_offset(new_offset);
+ reloc_write.put_r_info(elfcpp::elf_r_info<32>(new_symndx, r_type));
+
+ // Handle the reloc addend.
+ // The relocation uses a section symbol in the input file.
+ // We are adjusting it to use a section symbol in the output
+ // file. The input section symbol refers to some address in
+ // the input section. We need the relocation in the output
+ // file to refer to that same address. This adjustment to
+ // the addend is the same calculation we use for a simple
+ // absolute relocation for the input section symbol.
+
+ const Symbol_value<size>* psymval = object->local_symbol(r_sym);
+
+ unsigned char* paddend = view + offset;
+ typename Reloc_funcs::Status reloc_status = Reloc_funcs::STATUS_OKAY;
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_26:
+ reloc_status = Reloc_funcs::rel26(paddend, object, psymval,
+ offset_in_output_section, true, 0, sh_type == elfcpp::SHT_REL, NULL,
+ false /*TODO(sasa): cross mode jump*/, r_type, this->jal_to_bal());
+ break;
+
+ default:
+ gold_unreachable();
+ }
+
+ // Report any errors.
+ switch (reloc_status)
+ {
+ case Reloc_funcs::STATUS_OKAY:
+ break;
+ case Reloc_funcs::STATUS_OVERFLOW:
+ gold_error_at_location(relinfo, relnum, reloc.get_r_offset(),
+ _("relocation overflow"));
+ break;
+ case Reloc_funcs::STATUS_BAD_RELOC:
+ gold_error_at_location(relinfo, relnum, reloc.get_r_offset(),
+ _("unexpected opcode while processing relocation"));
+ break;
+ default:
+ gold_unreachable();
+ }
+}
+
+// Optimize the TLS relocation type based on what we know about the
+// symbol. IS_FINAL is true if the final address of this symbol is
+// known at link time.
+
+template<int size, bool big_endian>
+tls::Tls_optimization
+Target_mips<size, big_endian>::optimize_tls_reloc(bool, int)
+{
+ // FIXME: Currently we do not do any TLS optimization.
+ return tls::TLSOPT_NONE;
+}
+
+// Scan a relocation for a local symbol.
+
+template<int size, bool big_endian>
+inline void
+Target_mips<size, big_endian>::Scan::local(
+ Symbol_table* symtab,
+ Layout* layout,
+ Target_mips<size, big_endian>* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>* rela,
+ const elfcpp::Rel<size, big_endian>* rel,
+ unsigned int rel_type,
+ unsigned int r_type,
+ const elfcpp::Sym<size, big_endian>& lsym,
+ bool is_discarded)
+{
+ if (is_discarded)
+ return;
+
+ Mips_address r_offset;
+ typename elfcpp::Elf_types<size>::Elf_WXword r_info;
+ typename elfcpp::Elf_types<size>::Elf_Swxword r_addend;
+
+ if (rel_type == elfcpp::SHT_RELA)
+ {
+ r_offset = rela->get_r_offset();
+ r_info = rela->get_r_info();
+ r_addend = rela->get_r_addend();
+ }
+ else
+ {
+ r_offset = rel->get_r_offset();
+ r_info = rel->get_r_info();
+ r_addend = 0;
+ }
+
+ unsigned int r_sym = elfcpp::elf_r_sym<size>(r_info);
+ Mips_relobj<size, big_endian>* mips_obj =
+ Mips_relobj<size, big_endian>::as_mips_relobj(object);
+
+ if (mips_obj->is_mips16_stub_section(data_shndx))
+ {
+ mips_obj->get_mips16_stub_section(data_shndx)
+ ->new_local_reloc_found(r_type, r_sym);
+ }
+
+ if (r_type == elfcpp::R_MIPS_NONE)
+ // R_MIPS_NONE is used in mips16 stub sections, to define the target of the
+ // mips16 stub.
+ return;
+
+ if (!mips16_call_reloc(r_type)
+ && !mips_obj->section_allows_mips16_refs(data_shndx))
+ // This reloc would need to refer to a MIPS16 hard-float stub, if
+ // there is one. We ignore MIPS16 stub sections and .pdr section when
+ // looking for relocs that would need to refer to MIPS16 stubs.
+ mips_obj->add_local_non_16bit_call(r_sym);
+
+ if (r_type == elfcpp::R_MIPS16_26
+ && !mips_obj->section_allows_mips16_refs(data_shndx))
+ mips_obj->add_local_16bit_call(r_sym);
+
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MIPS_GOT_OFST:
+ case elfcpp::R_MIPS_GOT_DISP:
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_GOT16:
+ case elfcpp::R_MICROMIPS_CALL16:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_GOT_PAGE:
+ case elfcpp::R_MICROMIPS_GOT_OFST:
+ case elfcpp::R_MICROMIPS_GOT_DISP:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ // We need a GOT section.
+ target->got_section(symtab, layout);
+ break;
+
+ default:
+ break;
+ }
+
+ if (call_lo16_reloc(r_type)
+ || got_lo16_reloc(r_type)
+ || got_disp_reloc(r_type))
+ {
+ // We may need a local GOT entry for this relocation. We
+ // don't count R_MIPS_GOT_PAGE because we can estimate the
+ // maximum number of pages needed by looking at the size of
+ // the segment. Similar comments apply to R_MIPS*_GOT16 and
+ // R_MIPS*_CALL16. We don't count R_MIPS_GOT_HI16, or
+ // R_MIPS_CALL_HI16 because these are always followed by an
+ // R_MIPS_GOT_LO16 or R_MIPS_CALL_LO16.
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ unsigned int r_sym = elfcpp::elf_r_sym<size>(r_info);
+ got->record_local_got_symbol(mips_obj, r_sym, r_addend, r_type, -1U);
+ }
+
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MICROMIPS_CALL16:
+ gold_error(_("CALL16 reloc at 0x%lx not against global symbol "),
+ (unsigned long)r_offset);
+ return;
+
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MICROMIPS_GOT_PAGE:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_GOT16:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ {
+ // This relocation needs a page entry in the GOT.
+ // Get the section contents.
+ section_size_type view_size = 0;
+ const unsigned char* view = object->section_contents(data_shndx,
+ &view_size, false);
+ view += r_offset;
+
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(view);
+ Valtype32 addend = (rel_type == elfcpp::SHT_REL ? val & 0xffff
+ : r_addend);
+
+ if (rel_type == elfcpp::SHT_REL && got16_reloc(r_type))
+ target->got16_addends_.push_back(got16_addend<size, big_endian>(
+ object, data_shndx, r_type, r_sym, addend));
+ else
+ target->got_section()->record_got_page_entry(mips_obj, r_sym, addend);
+ break;
+ }
+
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MICROMIPS_HI16:
+ // Record the reloc so that we can check whether the corresponding LO16
+ // part exists.
+ if (rel_type == elfcpp::SHT_REL)
+ target->got16_addends_.push_back(got16_addend<size, big_endian>(
+ object, data_shndx, r_type, r_sym, 0));
+ break;
+
+ case elfcpp::R_MIPS_LO16:
+ case elfcpp::R_MIPS16_LO16:
+ case elfcpp::R_MICROMIPS_LO16:
+ {
+ if (rel_type != elfcpp::SHT_REL)
+ break;
+
+ // Find corresponding GOT16/HI16 relocation.
+
+ // According to the MIPS ELF ABI, the R_MIPS_LO16 relocation must
+ // be immediately following. However, for the IRIX6 ABI, the next
+ // relocation may be a composed relocation consisting of several
+ // relocations for the same address. In that case, the R_MIPS_LO16
+ // relocation may occur as one of these. We permit a similar
+ // extension in general, as that is useful for GCC.
+
+ // In some cases GCC dead code elimination removes the LO16 but
+ // keeps the corresponding HI16. This is strictly speaking a
+ // violation of the ABI but not immediately harmful.
+
+ typename std::list<got16_addend<size, big_endian> >::iterator it =
+ target->got16_addends_.begin();
+ while (it != target->got16_addends_.end())
+ {
+ got16_addend<size, big_endian> _got16_addend = *it;
+
+ // TODO(sasa): Split got16_addends_ list into two lists - one for
+ // GOT16 relocs and the other for HI16 relocs.
+
+ // Report an error if we find HI16 or GOT16 reloc from the
+ // previous section without the matching LO16 part.
+ if (_got16_addend.object != object
+ || _got16_addend.shndx != data_shndx)
+ {
+ gold_error("Can't find matching LO16 reloc");
+ break;
+ }
+
+ if (_got16_addend.r_sym != r_sym
+ || !is_matching_lo16_reloc(_got16_addend.r_type, r_type))
+ {
+ ++it;
+ continue;
+ }
+
+ // We found a matching HI16 or GOT16 reloc for this LO16 reloc.
+ // For GOT16, we need to calculate combined addend and record GOT page
+ // entry.
+ if (got16_reloc(_got16_addend.r_type))
+ {
+
+ section_size_type view_size = 0;
+ const unsigned char* view = object->section_contents(data_shndx,
+ &view_size,
+ false);
+ view += r_offset;
+
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(view);
+ int32_t addend = Bits<16>::sign_extend32(val & 0xffff);
+
+ addend = (_got16_addend.addend << 16) + addend;
+ target->got_section()->record_got_page_entry(mips_obj, r_sym,
+ addend);
+ }
+
+ it = target->got16_addends_.erase(it);
+ }
+ break;
+ }
+ }
+
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_32:
+ case elfcpp::R_MIPS_REL32:
+ case elfcpp::R_MIPS_64:
+ {
+ if (parameters->options().output_is_position_independent())
+ {
+ // If building a shared library (or a position-independent
+ // executable), we need to create a dynamic relocation for
+ // this location.
+ Reloc_section* rel_dyn = target->rel_dyn_section(layout);
+ unsigned int r_sym = elfcpp::elf_r_sym<32>(r_info);
+ rel_dyn->add_symbolless_local_addend(object, r_sym,
+ elfcpp::R_MIPS_REL32,
+ output_section, data_shndx,
+ r_offset);
+ }
+ break;
+ }
+
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ {
+ unsigned int r_sym = elfcpp::elf_r_sym<size>(r_info);
+ bool output_is_shared = parameters->options().shared();
+ const tls::Tls_optimization optimized_type
+ = Target_mips<size, big_endian>::optimize_tls_reloc(
+ !output_is_shared, r_type);
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ if (optimized_type == tls::TLSOPT_NONE)
+ {
+ // Create a pair of GOT entries for the module index and
+ // dtv-relative offset.
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ unsigned int shndx = lsym.get_st_shndx();
+ bool is_ordinary;
+ shndx = object->adjust_sym_shndx(r_sym, shndx, &is_ordinary);
+ if (!is_ordinary)
+ {
+ object->error(_("local symbol %u has bad shndx %u"),
+ r_sym, shndx);
+ break;
+ }
+ got->record_local_got_symbol(mips_obj, r_sym, r_addend, r_type,
+ shndx);
+ }
+ else
+ {
+ // FIXME: TLS optimization not supported yet.
+ gold_unreachable();
+ }
+ break;
+
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ if (optimized_type == tls::TLSOPT_NONE)
+ {
+ // We always record LDM symbols as local with index 0.
+ target->got_section()->record_local_got_symbol(mips_obj, 0,
+ r_addend, r_type,
+ -1U);
+ }
+ else
+ {
+ // FIXME: TLS optimization not supported yet.
+ gold_unreachable();
+ }
+ break;
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ layout->set_has_static_tls();
+ if (optimized_type == tls::TLSOPT_NONE)
+ {
+ // Create a GOT entry for the tp-relative offset.
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ got->record_local_got_symbol(mips_obj, r_sym, r_addend, r_type,
+ -1U);
+ }
+ else
+ {
+ // FIXME: TLS optimization not supported yet.
+ gold_unreachable();
+ }
+ break;
+
+ default:
+ gold_unreachable();
+ }
+ }
+ break;
+
+ default:
+ break;
+ }
+
+ // Refuse some position-dependent relocations when creating a
+ // shared library. Do not refuse R_MIPS_32 / R_MIPS_64; they're
+ // not PIC, but we can create dynamic relocations and the result
+ // will be fine. Also do not refuse R_MIPS_LO16, which can be
+ // combined with R_MIPS_GOT16.
+ if (parameters->options().shared())
+ {
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MICROMIPS_HI16:
+ // Don't refuse a high part relocation if it's against
+ // no symbol (e.g. part of a compound relocation).
+ if (r_sym == 0)
+ break;
+
+ // FALLTHROUGH
+
+ case elfcpp::R_MIPS16_26:
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MICROMIPS_26_S1:
+ gold_error(_("%s: relocation %u against `%s' can not be used when "
+ "making a shared object; recompile with -fPIC"),
+ object->name().c_str(), r_type, "a local symbol");
+ default:
+ break;
+ }
+ }
+}
+
+template<int size, bool big_endian>
+inline void
+Target_mips<size, big_endian>::Scan::local(
+ Symbol_table* symtab,
+ Layout* layout,
+ Target_mips<size, big_endian>* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rel<size, big_endian>& reloc,
+ unsigned int r_type,
+ const elfcpp::Sym<size, big_endian>& lsym,
+ bool is_discarded)
+{
+ if (is_discarded)
+ return;
+
+ local(
+ symtab,
+ layout,
+ target,
+ object,
+ data_shndx,
+ output_section,
+ (const elfcpp::Rela<size, big_endian>*) NULL,
+ &reloc,
+ elfcpp::SHT_REL,
+ r_type,
+ lsym, is_discarded);
+}
+
+
+template<int size, bool big_endian>
+inline void
+Target_mips<size, big_endian>::Scan::local(
+ Symbol_table* symtab,
+ Layout* layout,
+ Target_mips<size, big_endian>* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>& reloc,
+ unsigned int r_type,
+ const elfcpp::Sym<size, big_endian>& lsym,
+ bool is_discarded)
+{
+ if (is_discarded)
+ return;
+
+ local(
+ symtab,
+ layout,
+ target,
+ object,
+ data_shndx,
+ output_section,
+ &reloc,
+ (const elfcpp::Rel<size, big_endian>*) NULL,
+ elfcpp::SHT_RELA,
+ r_type,
+ lsym, is_discarded);
+}
+
+// Scan a relocation for a global symbol.
+
+template<int size, bool big_endian>
+inline void
+Target_mips<size, big_endian>::Scan::global(
+ Symbol_table* symtab,
+ Layout* layout,
+ Target_mips<size, big_endian>* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>* rela,
+ const elfcpp::Rel<size, big_endian>* rel,
+ unsigned int rel_type,
+ unsigned int r_type,
+ Symbol* gsym)
+{
+ Mips_address r_offset;
+ typename elfcpp::Elf_types<size>::Elf_WXword r_info;
+ typename elfcpp::Elf_types<size>::Elf_Swxword r_addend;
+
+ if (rel_type == elfcpp::SHT_RELA)
+ {
+ r_offset = rela->get_r_offset();
+ r_info = rela->get_r_info();
+ r_addend = rela->get_r_addend();
+ }
+ else
+ {
+ r_offset = rel->get_r_offset();
+ r_info = rel->get_r_info();
+ r_addend = 0;
+ }
+
+ unsigned int r_sym = elfcpp::elf_r_sym<size>(r_info);
+ Mips_relobj<size, big_endian>* mips_obj =
+ Mips_relobj<size, big_endian>::as_mips_relobj(object);
+ Mips_symbol<size>* mips_sym = Mips_symbol<size>::as_mips_sym(gsym);
+
+ if (mips_obj->is_mips16_stub_section(data_shndx))
+ {
+ mips_obj->get_mips16_stub_section(data_shndx)
+ ->new_global_reloc_found(r_type, mips_sym);
+ }
+
+ if (r_type == elfcpp::R_MIPS_NONE)
+ // R_MIPS_NONE is used in mips16 stub sections, to define the target of the
+ // mips16 stub.
+ return;
+
+ if (!mips16_call_reloc(r_type)
+ && !mips_obj->section_allows_mips16_refs(data_shndx))
+ // This reloc would need to refer to a MIPS16 hard-float stub, if
+ // there is one. We ignore MIPS16 stub sections and .pdr section when
+ // looking for relocs that would need to refer to MIPS16 stubs.
+ mips_sym->set_need_fn_stub();
+
+ // A reference to _GLOBAL_OFFSET_TABLE_ implies that we need a got
+ // section. We check here to avoid creating a dynamic reloc against
+ // _GLOBAL_OFFSET_TABLE_.
+ if (!target->has_got_section()
+ && strcmp(gsym->name(), "_GLOBAL_OFFSET_TABLE_") == 0)
+ target->got_section(symtab, layout);
+
+ // We need PLT entries if there are static-only relocations against
+ // an externally-defined function. This can technically occur for
+ // shared libraries if there are branches to the symbol, although it
+ // is unlikely that this will be used in practice due to the short
+ // ranges involved. It can occur for any relative or absolute relocation
+ // in executables; in that case, the PLT entry becomes the function's
+ // canonical address.
+ bool static_reloc = false;
+
+ // Set CAN_MAKE_DYNAMIC to true if we can convert this
+ // relocation into a dynamic one.
+ bool can_make_dynamic = false;
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MIPS_GOT_OFST:
+ case elfcpp::R_MIPS_GOT_DISP:
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_GOT16:
+ case elfcpp::R_MICROMIPS_CALL16:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_GOT_PAGE:
+ case elfcpp::R_MICROMIPS_GOT_OFST:
+ case elfcpp::R_MICROMIPS_GOT_DISP:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ // We need a GOT section.
+ target->got_section(symtab, layout);
+ break;
+
+ // This is just a hint; it can safely be ignored. Don't set
+ // has_static_relocs for the corresponding symbol.
+ case elfcpp::R_MIPS_JALR:
+ case elfcpp::R_MICROMIPS_JALR:
+ break;
+
+ case elfcpp::R_MIPS_GPREL16:
+ case elfcpp::R_MIPS_GPREL32:
+ case elfcpp::R_MIPS16_GPREL:
+ case elfcpp::R_MICROMIPS_GPREL16:
+ // TODO(sasa)
+ // GP-relative relocations always resolve to a definition in a
+ // regular input file, ignoring the one-definition rule. This is
+ // important for the GP setup sequence in NewABI code, which
+ // always resolves to a local function even if other relocations
+ // against the symbol wouldn't.
+ //constrain_symbol_p = FALSE;
+ break;
+
+ case elfcpp::R_MIPS_32:
+ case elfcpp::R_MIPS_REL32:
+ case elfcpp::R_MIPS_64:
+ if (parameters->options().shared()
+ || strcmp(gsym->name(), "__gnu_local_gp") != 0)
+ {
+ if (r_type != elfcpp::R_MIPS_REL32)
+ {
+ static_reloc = true;
+ mips_sym->set_pointer_equality_needed();
+ }
+ can_make_dynamic = true;
+ break;
+ }
+ // Fall through.
+
+ default:
+ // Most static relocations require pointer equality, except
+ // for branches.
+ mips_sym->set_pointer_equality_needed();
+
+ // Fall through.
+
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MIPS_PC16:
+ case elfcpp::R_MIPS16_26:
+ case elfcpp::R_MICROMIPS_26_S1:
+ case elfcpp::R_MICROMIPS_PC7_S1:
+ case elfcpp::R_MICROMIPS_PC10_S1:
+ case elfcpp::R_MICROMIPS_PC16_S1:
+ case elfcpp::R_MICROMIPS_PC23_S2:
+ static_reloc = true;
+ mips_sym->set_has_static_relocs();
+ break;
+ }
+
+ // If there are call relocations against an externally-defined symbol,
+ // see whether we can create a MIPS lazy-binding stub for it. We can
+ // only do this if all references to the function are through call
+ // relocations, and in that case, the traditional lazy-binding stubs
+ // are much more efficient than PLT entries.
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MIPS_JALR:
+ case elfcpp::R_MICROMIPS_CALL16:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ case elfcpp::R_MICROMIPS_JALR:
+ if (!mips_sym->no_lazy_stub())
+ {
+ if ((mips_sym->needs_plt_entry() && mips_sym->is_from_dynobj())
+ // Calls from shared objects to undefined symbols of type
+ // STT_NOTYPE need lazy-binding stub.
+ || (mips_sym->is_undefined() && parameters->options().shared()))
+ target->mips_stubs_section(layout)->make_entry(mips_sym);
+ }
+ break;
+ default:
+ {
+ // We must not create a stub for a symbol that has relocations
+ // related to taking the function's address.
+ mips_sym->set_no_lazy_stub();
+ target->remove_lazy_stub_entry(mips_sym);
+ break;
+ }
+ }
+
+ if (relocation_needs_la25_stub<size, big_endian>(mips_obj, r_type,
+ mips_sym->is_mips16()))
+ mips_sym->set_has_nonpic_branches();
+
+ // R_MIPS_HI16 against _gp_disp is used for $gp setup,
+ // and has a special meaning.
+ bool gp_disp_against_hi16 = (!mips_obj->is_newabi()
+ && strcmp(gsym->name(), "_gp_disp") == 0
+ && (hi16_reloc(r_type) || lo16_reloc(r_type)));
+ if (static_reloc && gsym->needs_plt_entry())
+ {
+ target->make_plt_entry(symtab, layout, mips_sym, r_type);
+
+ // Since this is not a PC-relative relocation, we may be
+ // taking the address of a function. In that case we need to
+ // set the entry in the dynamic symbol table to the address of
+ // the PLT entry.
+ if (gsym->is_from_dynobj() && !parameters->options().shared())
+ {
+ gsym->set_needs_dynsym_value();
+ // We distinguish between PLT entries and lazy-binding stubs by
+ // giving the former an st_other value of STO_MIPS_PLT. Set the
+ // flag if there are any relocations in the binary where pointer
+ // equality matters.
+ if (mips_sym->pointer_equality_needed())
+ mips_sym->set_mips_plt();
+ }
+ }
+ if ((static_reloc || can_make_dynamic) && !gp_disp_against_hi16)
+ {
+ // Absolute addressing relocations.
+ // Make a dynamic relocation if necessary.
+ if (gsym->needs_dynamic_reloc(Scan::get_reference_flags(r_type)))
+ {
+ if (gsym->may_need_copy_reloc())
+ {
+ target->copy_reloc(symtab, layout, object,
+ data_shndx, output_section, gsym, *rel);
+ }
+ else if (can_make_dynamic)
+ {
+ // Create .rel.dyn section.
+ target->rel_dyn_section(layout);
+ target->dynamic_reloc(mips_sym, elfcpp::R_MIPS_REL32, mips_obj,
+ data_shndx, output_section, r_offset);
+ }
+ else
+ gold_error(_("non-dynamic relocations refer to dynamic symbol %s"),
+ gsym->name());
+ }
+ }
+
+ bool for_call = false;
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MICROMIPS_CALL16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ for_call = true;
+ // Fall through.
+
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_GOT16:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ case elfcpp::R_MIPS_GOT_DISP:
+ case elfcpp::R_MICROMIPS_GOT_DISP:
+ {
+ // The symbol requires a GOT entry.
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ got->record_global_got_symbol(mips_sym, mips_obj, r_type, false,
+ for_call);
+ mips_sym->set_global_got_area(GGA_NORMAL);
+ }
+ break;
+
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MICROMIPS_GOT_PAGE:
+ {
+ // This relocation needs a page entry in the GOT.
+ // Get the section contents.
+ section_size_type view_size = 0;
+ const unsigned char* view =
+ object->section_contents(data_shndx, &view_size, false);
+ view += r_offset;
+
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(view);
+ Valtype32 addend = (rel_type == elfcpp::SHT_REL ? val & 0xffff
+ : r_addend);
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ got->record_got_page_entry(mips_obj, r_sym, addend);
+
+ // If this is a global, overridable symbol, GOT_PAGE will
+ // decay to GOT_DISP, so we'll need a GOT entry for it.
+ bool def_regular = (mips_sym->source() == Symbol::FROM_OBJECT
+ && !mips_sym->object()->is_dynamic()
+ && !mips_sym->is_undefined());
+ if (!def_regular
+ || (parameters->options().output_is_position_independent()
+ && !parameters->options().Bsymbolic()
+ && !mips_sym->is_forced_local()))
+ {
+ got->record_global_got_symbol(mips_sym, mips_obj, r_type, false,
+ for_call);
+ mips_sym->set_global_got_area(GGA_NORMAL);
+ }
+ }
+ break;
+
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ {
+ const bool is_final = gsym->final_value_is_known();
+ const tls::Tls_optimization optimized_type =
+ Target_mips<size, big_endian>::optimize_tls_reloc(is_final, r_type);
+
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ if (optimized_type == tls::TLSOPT_NONE)
+ {
+ // Create a pair of GOT entries for the module index and
+ // dtv-relative offset.
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ got->record_global_got_symbol(mips_sym, mips_obj, r_type, false,
+ false);
+ }
+ else
+ {
+ // FIXME: TLS optimization not supported yet.
+ gold_unreachable();
+ }
+ break;
+
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ if (optimized_type == tls::TLSOPT_NONE)
+ {
+ // We always record LDM symbols as local with index 0.
+ target->got_section()->record_local_got_symbol(mips_obj, 0,
+ r_addend, r_type,
+ -1U);
+ }
+ else
+ {
+ // FIXME: TLS optimization not supported yet.
+ gold_unreachable();
+ }
+ break;
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ layout->set_has_static_tls();
+ if (optimized_type == tls::TLSOPT_NONE)
+ {
+ // Create a GOT entry for the tp-relative offset.
+ Mips_output_data_got<size, big_endian>* got =
+ target->got_section(symtab, layout);
+ got->record_global_got_symbol(mips_sym, mips_obj, r_type, false,
+ false);
+ }
+ else
+ {
+ // FIXME: TLS optimization not supported yet.
+ gold_unreachable();
+ }
+ break;
+
+ default:
+ gold_unreachable();
+ }
+ }
+ break;
+ case elfcpp::R_MIPS_COPY:
+ case elfcpp::R_MIPS_JUMP_SLOT:
+ // These are relocations which should only be seen by the
+ // dynamic linker, and should never be seen here.
+ gold_error(_("%s: unexpected reloc %u in object file"),
+ object->name().c_str(), r_type);
+ break;
+
+ default:
+ break;
+ }
+
+ // Refuse some position-dependent relocations when creating a
+ // shared library. Do not refuse R_MIPS_32 / R_MIPS_64; they're
+ // not PIC, but we can create dynamic relocations and the result
+ // will be fine. Also do not refuse R_MIPS_LO16, which can be
+ // combined with R_MIPS_GOT16.
+ if (parameters->options().shared())
+ {
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MICROMIPS_HI16:
+ // Don't refuse a high part relocation if it's against
+ // no symbol (e.g. part of a compound relocation).
+ if (r_sym == 0)
+ break;
+
+ // R_MIPS_HI16 against _gp_disp is used for $gp setup,
+ // and has a special meaning.
+ if (!mips_obj->is_newabi() && strcmp(gsym->name(), "_gp_disp") == 0)
+ break;
+
+ // FALLTHROUGH
+
+ case elfcpp::R_MIPS16_26:
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MICROMIPS_26_S1:
+ gold_error(_("%s: relocation %u against `%s' can not be used when "
+ "making a shared object; recompile with -fPIC"),
+ object->name().c_str(), r_type, gsym->name());
+ default:
+ break;
+ }
+ }
+}
+
+template<int size, bool big_endian>
+inline void
+Target_mips<size, big_endian>::Scan::global(
+ Symbol_table* symtab,
+ Layout* layout,
+ Target_mips<size, big_endian>* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>& reloc,
+ unsigned int r_type,
+ Symbol* gsym)
+{
+ global(
+ symtab,
+ layout,
+ target,
+ object,
+ data_shndx,
+ output_section,
+ &reloc,
+ (const elfcpp::Rel<size, big_endian>*) NULL,
+ elfcpp::SHT_RELA,
+ r_type,
+ gsym);
+}
+
+template<int size, bool big_endian>
+inline void
+Target_mips<size, big_endian>::Scan::global(
+ Symbol_table* symtab,
+ Layout* layout,
+ Target_mips<size, big_endian>* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rel<size, big_endian>& reloc,
+ unsigned int r_type,
+ Symbol* gsym)
+{
+ global(
+ symtab,
+ layout,
+ target,
+ object,
+ data_shndx,
+ output_section,
+ (const elfcpp::Rela<size, big_endian>*) NULL,
+ &reloc,
+ elfcpp::SHT_REL,
+ r_type,
+ gsym);
+}
+
+// Return whether a R_MIPS_32 relocation needs to be applied.
+
+template<int size, bool big_endian>
+inline bool
+Target_mips<size, big_endian>::Relocate::should_apply_r_mips_32_reloc(
+ const Mips_symbol<size>* gsym,
+ unsigned int r_type,
+ Output_section* output_section,
+ Target_mips* target)
+{
+ // If the output section is not allocated, then we didn't call
+ // scan_relocs, we didn't create a dynamic reloc, and we must apply
+ // the reloc here.
+ if ((output_section->flags() & elfcpp::SHF_ALLOC) == 0)
+ return true;
+
+ if (gsym == NULL)
+ return true;
+ else
+ {
+ // For global symbols, we use the same helper routines used in the
+ // scan pass.
+ if (gsym->needs_dynamic_reloc(Scan::get_reference_flags(r_type))
+ && !gsym->may_need_copy_reloc())
+ {
+ // We have generated dynamic reloc (R_MIPS_REL32).
+
+ bool multi_got = false;
+ if (target->has_got_section())
+ multi_got = target->got_section()->multi_got();
+ bool has_got_offset;
+ if (!multi_got)
+ has_got_offset = gsym->has_got_offset(GOT_TYPE_STANDARD);
+ else
+ has_got_offset = gsym->global_gotoffset() != -1U;
+ if (!has_got_offset)
+ return true;
+ else
+ // Apply the relocation only if the symbol is in the local got.
+ // Do not apply the relocation if the symbol is in the global
+ // got.
+ return symbol_references_local(gsym, gsym->has_dynsym_index());
+ }
+ else
+ // We have not generated dynamic reloc.
+ return true;
+ }
+}
+
+// Perform a relocation.
+
+template<int size, bool big_endian>
+inline bool
+Target_mips<size, big_endian>::Relocate::relocate(
+ const Relocate_info<size, big_endian>* relinfo,
+ Target_mips* target,
+ Output_section* output_section,
+ size_t relnum,
+ const elfcpp::Rela<size, big_endian>* rela,
+ const elfcpp::Rel<size, big_endian>* rel,
+ unsigned int rel_type,
+ unsigned int r_type,
+ const Sized_symbol<size>* gsym,
+ const Symbol_value<size>* psymval,
+ unsigned char* view,
+ Mips_address address,
+ section_size_type)
+{
+ Mips_address r_offset;
+ typename elfcpp::Elf_types<size>::Elf_WXword r_info;
+ typename elfcpp::Elf_types<size>::Elf_Swxword r_addend;
+
+ if (rel_type == elfcpp::SHT_RELA)
+ {
+ r_offset = rela->get_r_offset();
+ r_info = rela->get_r_info();
+ r_addend = rela->get_r_addend();
+ }
+ else
+ {
+ r_offset = rel->get_r_offset();
+ r_info = rel->get_r_info();
+ r_addend = 0;
+ }
+
+ typedef Mips_relocate_functions<size, big_endian> Reloc_funcs;
+ typename Reloc_funcs::Status reloc_status = Reloc_funcs::STATUS_OKAY;
+
+ Mips_relobj<size, big_endian>* object =
+ Mips_relobj<size, big_endian>::as_mips_relobj(relinfo->object);
+
+ unsigned int r_sym = elfcpp::elf_r_sym<size>(r_info);
+ bool target_is_16_bit_code = false;
+ bool target_is_micromips_code = false;
+ bool cross_mode_jump;
+
+ Symbol_value<size> symval;
+
+ const Mips_symbol<size>* mips_sym = Mips_symbol<size>::as_mips_sym(gsym);
+
+ bool changed_symbol_value = false;
+ if (gsym == NULL)
+ {
+ target_is_16_bit_code = object->local_symbol_is_mips16(r_sym);
+ target_is_micromips_code = object->local_symbol_is_micromips(r_sym);
+ if (target_is_16_bit_code || target_is_micromips_code)
+ {
+ // MIPS16/microMIPS text labels should be treated as odd.
+ symval.set_output_value(psymval->value(object, 1));
+ psymval = &symval;
+ changed_symbol_value = true;
+ }
+ }
+ else
+ {
+ target_is_16_bit_code = mips_sym->is_mips16();
+ target_is_micromips_code = mips_sym->is_micromips();
+
+ // If this is a mips16/microMIPS text symbol, add 1 to the value to make
+ // it odd. This will cause something like .word SYM to come up with
+ // the right value when it is loaded into the PC.
+
+ if ((mips_sym->is_mips16() || mips_sym->is_micromips())
+ && psymval->value(object, 0) != 0)
+ {
+ symval.set_output_value(psymval->value(object, 0) | 1);
+ psymval = &symval;
+ changed_symbol_value = true;
+ }
+
+ // Pick the value to use for symbols defined in shared objects.
+ if (mips_sym->use_plt_offset(Scan::get_reference_flags(r_type))
+ || mips_sym->has_lazy_stub())
+ {
+ Mips_address value;
+ if (!mips_sym->has_lazy_stub())
+ {
+ // Prefer a standard MIPS PLT entry.
+ if (mips_sym->has_mips_plt_offset())
+ {
+ value = target->plt_section()->mips_entry_address(mips_sym);
+ target_is_micromips_code = false;
+ target_is_16_bit_code = false;
+ }
+ else
+ {
+ value = (target->plt_section()->comp_entry_address(mips_sym)
+ + 1);
+ if (target->is_output_micromips())
+ target_is_micromips_code = true;
+ else
+ target_is_16_bit_code = true;
+ }
+ }
+ else
+ value = target->mips_stubs_section()->stub_address(mips_sym);
+
+ symval.set_output_value(value);
+ psymval = &symval;
+ }
+ }
+
+ // TRUE if the symbol referred to by this relocation is "_gp_disp".
+ // Note that such a symbol must always be a global symbol.
+ bool gp_disp = (gsym != NULL && (strcmp(gsym->name(), "_gp_disp") == 0)
+ && !object->is_newabi());
+
+ // TRUE if the symbol referred to by this relocation is "__gnu_local_gp".
+ // Note that such a symbol must always be a global symbol.
+ bool gnu_local_gp = gsym && (strcmp(gsym->name(), "__gnu_local_gp") == 0);
+
+
+ if (gp_disp)
+ {
+ if (!hi16_reloc(r_type) && !lo16_reloc(r_type))
+ gold_error_at_location(relinfo, relnum, r_offset,
+ _("relocations against _gp_disp are permitted only"
+ " with R_MIPS_HI16 and R_MIPS_LO16 relocations."));
+ }
+ else if (gnu_local_gp)
+ {
+ // __gnu_local_gp is _gp symbol.
+ symval.set_output_value(target->adjusted_gp_value(object));
+ psymval = &symval;
+ }
+
+ // If this is a reference to a 16-bit function with a stub, we need
+ // to redirect the relocation to the stub unless:
+ //
+ // (a) the relocation is for a MIPS16 JAL;
+ //
+ // (b) the relocation is for a MIPS16 PIC call, and there are no
+ // non-MIPS16 uses of the GOT slot; or
+ //
+ // (c) the section allows direct references to MIPS16 functions.
+ if (r_type != elfcpp::R_MIPS16_26
+ && !parameters->options().relocatable()
+ && ((mips_sym != NULL
+ && mips_sym->has_mips16_fn_stub()
+ && (r_type != elfcpp::R_MIPS16_CALL16 || mips_sym->need_fn_stub()))
+ || (mips_sym == NULL
+ && object->get_local_mips16_fn_stub(r_sym) != NULL))
+ && !object->section_allows_mips16_refs(relinfo->data_shndx))
+ {
+ // This is a 32- or 64-bit call to a 16-bit function. We should
+ // have already noticed that we were going to need the
+ // stub.
+ Mips_address value;
+ if (mips_sym == NULL)
+ value = object->get_local_mips16_fn_stub(r_sym)->output_address();
+ else
+ {
+ gold_assert(mips_sym->need_fn_stub());
+ if (mips_sym->has_la25_stub())
+ value = target->la25_stub_section()->stub_address(mips_sym);
+ else
+ {
+ value = mips_sym->template
+ get_mips16_fn_stub<big_endian>()->output_address();
+ }
+ }
+ symval.set_output_value(value);
+ psymval = &symval;
+ changed_symbol_value = true;
+
+ // The target is 16-bit, but the stub isn't.
+ target_is_16_bit_code = false;
+ }
+ // If this is a MIPS16 call with a stub, that is made through the PLT or
+ // to a standard MIPS function, we need to redirect the call to the stub.
+ // Note that we specifically exclude R_MIPS16_CALL16 from this behavior;
+ // indirect calls should use an indirect stub instead.
+ else if (r_type == elfcpp::R_MIPS16_26 && !parameters->options().relocatable()
+ && ((mips_sym != NULL
+ && (mips_sym->has_mips16_call_stub()
+ || mips_sym->has_mips16_call_fp_stub()))
+ || (mips_sym == NULL
+ && object->get_local_mips16_call_stub(r_sym) != NULL))
+ && ((mips_sym != NULL && mips_sym->has_plt_offset())
+ || !target_is_16_bit_code))
+ {
+ Mips16_stub_section<size, big_endian>* call_stub;
+ if (mips_sym == NULL)
+ call_stub = object->get_local_mips16_call_stub(r_sym);
+ else
+ {
+ // If both call_stub and call_fp_stub are defined, we can figure
+ // out which one to use by checking which one appears in the input
+ // file.
+ if (mips_sym->has_mips16_call_stub()
+ && mips_sym->has_mips16_call_fp_stub())
+ {
+ call_stub = NULL;
+ for (unsigned int i = 1; i < object->shnum(); ++i)
+ {
+ if (object->is_mips16_call_fp_stub_section(i))
+ {
+ call_stub = mips_sym->template
+ get_mips16_call_fp_stub<big_endian>();
+ break;
+ }
+
+ }
+ if (call_stub == NULL)
+ call_stub =
+ mips_sym->template get_mips16_call_stub<big_endian>();
+ }
+ else if (mips_sym->has_mips16_call_stub())
+ call_stub = mips_sym->template get_mips16_call_stub<big_endian>();
+ else
+ call_stub = mips_sym->template get_mips16_call_fp_stub<big_endian>();
+ }
+
+ symval.set_output_value(call_stub->output_address());
+ psymval = &symval;
+ changed_symbol_value = true;
+ }
+ // If this is a direct call to a PIC function, redirect to the
+ // non-PIC stub.
+ else if (mips_sym != NULL
+ && mips_sym->has_la25_stub()
+ && relocation_needs_la25_stub<size, big_endian>(
+ object, r_type, target_is_16_bit_code))
+ {
+ Mips_address value = target->la25_stub_section()->stub_address(mips_sym);
+ if (mips_sym->is_micromips())
+ value += 1;
+ symval.set_output_value(value);
+ psymval = &symval;
+ }
+ // For direct MIPS16 and microMIPS calls make sure the compressed PLT
+ // entry is used if a standard PLT entry has also been made.
+ else if ((r_type == elfcpp::R_MIPS16_26
+ || r_type == elfcpp::R_MICROMIPS_26_S1)
+ && !parameters->options().relocatable()
+ && mips_sym != NULL
+ && mips_sym->has_plt_offset()
+ && mips_sym->has_comp_plt_offset()
+ && mips_sym->has_mips_plt_offset())
+ {
+ Mips_address value = (target->plt_section()->comp_entry_address(mips_sym)
+ + 1);
+ symval.set_output_value(value);
+ psymval = &symval;
+
+ target_is_16_bit_code = !target->is_output_micromips();
+ target_is_micromips_code = target->is_output_micromips();
+ }
+
+ // Make sure MIPS16 and microMIPS are not used together.
+ if ((r_type == elfcpp::R_MIPS16_26 && target_is_micromips_code)
+ || (micromips_branch_reloc(r_type) && target_is_16_bit_code))
+ {
+ gold_error(_("MIPS16 and microMIPS functions cannot call each other"));
+ }
+
+ // Calls from 16-bit code to 32-bit code and vice versa require the
+ // mode change. However, we can ignore calls to undefined weak symbols,
+ // which should never be executed at runtime. This exception is important
+ // because the assembly writer may have "known" that any definition of the
+ // symbol would be 16-bit code, and that direct jumps were therefore
+ // acceptable.
+ cross_mode_jump =
+ (!parameters->options().relocatable()
+ && !(gsym != NULL && gsym->is_weak_undefined())
+ && ((r_type == elfcpp::R_MIPS16_26 && !target_is_16_bit_code)
+ || (r_type == elfcpp::R_MICROMIPS_26_S1 && !target_is_micromips_code)
+ || ((r_type == elfcpp::R_MIPS_26 || r_type == elfcpp::R_MIPS_JALR)
+ && (target_is_16_bit_code || target_is_micromips_code))));
+
+ bool local = (mips_sym == NULL
+ || (mips_sym->got_only_for_calls()
+ ? symbol_calls_local(mips_sym, mips_sym->has_dynsym_index())
+ : symbol_references_local(mips_sym,
+ mips_sym->has_dynsym_index())));
+
+ // Global R_MIPS_GOT_PAGE/R_MICROMIPS_GOT_PAGE relocations are equivalent
+ // to R_MIPS_GOT_DISP/R_MICROMIPS_GOT_DISP. The addend is applied by the
+ // corresponding R_MIPS_GOT_OFST/R_MICROMIPS_GOT_OFST.
+ if (got_page_reloc(r_type) && !local)
+ r_type = (micromips_reloc(r_type) ? elfcpp::R_MICROMIPS_GOT_DISP
+ : elfcpp::R_MIPS_GOT_DISP);
+
+ unsigned int got_offset = 0;
+ int gp_offset = 0;
+
+ bool update_got_entry = false;
+ bool extract_addend = rel_type == elfcpp::SHT_REL;
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_NONE:
+ break;
+ case elfcpp::R_MIPS_16:
+ reloc_status = Reloc_funcs::rel16(view, object, psymval, r_addend,
+ extract_addend, r_type);
+ break;
+
+ case elfcpp::R_MIPS_32:
+ if (should_apply_r_mips_32_reloc(mips_sym, r_type, output_section,
+ target))
+ reloc_status = Reloc_funcs::rel32(view, object, psymval, r_addend,
+ extract_addend, r_type);
+ if (mips_sym != NULL
+ && (mips_sym->is_mips16() || mips_sym->is_micromips())
+ && mips_sym->global_got_area() == GGA_RELOC_ONLY)
+ {
+ // If mips_sym->has_mips16_fn_stub() is false, symbol value is
+ // already updated by adding +1.
+ if (mips_sym->has_mips16_fn_stub())
+ {
+ gold_assert(mips_sym->need_fn_stub());
+ Mips16_stub_section<size, big_endian>* fn_stub =
+ mips_sym->template get_mips16_fn_stub<big_endian>();
+
+ symval.set_output_value(fn_stub->output_address());
+ psymval = &symval;
+ }
+ got_offset = mips_sym->global_gotoffset();
+ update_got_entry = true;
+ }
+ break;
+
+ case elfcpp::R_MIPS_REL32:
+ gold_unreachable();
+
+ case elfcpp::R_MIPS_PC32:
+ reloc_status = Reloc_funcs::relpc32(view, object, psymval, address,
+ r_addend, extract_addend, r_type);
+ break;
+
+ case elfcpp::R_MIPS16_26:
+ // The calculation for R_MIPS16_26 is just the same as for an
+ // R_MIPS_26. It's only the storage of the relocated field into
+ // the output file that's different. So, we just fall through to the
+ // R_MIPS_26 case here.
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MICROMIPS_26_S1:
+ reloc_status = Reloc_funcs::rel26(view, object, psymval, address,
+ gsym == NULL, r_addend, extract_addend, gsym, cross_mode_jump, r_type,
+ target->jal_to_bal());
+ break;
+
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MICROMIPS_HI16:
+ reloc_status = Reloc_funcs::relhi16(view, object, psymval, r_addend,
+ address, gp_disp, r_type,
+ extract_addend);
+ break;
+
+ case elfcpp::R_MIPS_LO16:
+ case elfcpp::R_MIPS16_LO16:
+ case elfcpp::R_MICROMIPS_LO16:
+ case elfcpp::R_MICROMIPS_HI0_LO16:
+ reloc_status = Reloc_funcs::rello16(target, view, object, psymval,
+ r_addend, extract_addend, address,
+ gp_disp, r_type);
+ break;
+
+ case elfcpp::R_MIPS_LITERAL:
+ case elfcpp::R_MICROMIPS_LITERAL:
+ // Because we don't merge literal sections, we can handle this
+ // just like R_MIPS_GPREL16. In the long run, we should merge
+ // shared literals, and then we will need to additional work
+ // here.
+
+ // Fall through.
+
+ case elfcpp::R_MIPS_GPREL16:
+ case elfcpp::R_MIPS16_GPREL:
+ case elfcpp::R_MICROMIPS_GPREL7_S2:
+ case elfcpp::R_MICROMIPS_GPREL16:
+ reloc_status = Reloc_funcs::relgprel(view, object, psymval,
+ target->adjusted_gp_value(object),
+ r_addend, extract_addend,
+ gsym == NULL, r_type);
+ break;
+
+ case elfcpp::R_MIPS_PC16:
+ reloc_status = Reloc_funcs::relpc16(view, object, psymval, address,
+ r_addend, extract_addend, r_type);
+ break;
+ case elfcpp::R_MICROMIPS_PC7_S1:
+ reloc_status = Reloc_funcs::relmicromips_pc7_s1(view, object, psymval,
+ address, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MICROMIPS_PC10_S1:
+ reloc_status = Reloc_funcs::relmicromips_pc10_s1(view, object, psymval,
+ address, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MICROMIPS_PC16_S1:
+ reloc_status = Reloc_funcs::relmicromips_pc16_s1(view, object, psymval,
+ address, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_GPREL32:
+ reloc_status = Reloc_funcs::relgprel32(view, object, psymval,
+ target->adjusted_gp_value(object),
+ r_addend, extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ if (gsym != NULL)
+ got_offset = target->got_section()->got_offset(gsym, GOT_TYPE_STANDARD,
+ object);
+ else
+ got_offset = target->got_section()->got_offset(r_sym, GOT_TYPE_STANDARD,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot_hi16(view, gp_offset, r_type);
+ update_got_entry = changed_symbol_value;
+ break;
+
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ if (gsym != NULL)
+ got_offset = target->got_section()->got_offset(gsym, GOT_TYPE_STANDARD,
+ object);
+ else
+ got_offset = target->got_section()->got_offset(r_sym, GOT_TYPE_STANDARD,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot_lo16(view, gp_offset, r_type);
+ update_got_entry = changed_symbol_value;
+ break;
+
+ case elfcpp::R_MIPS_GOT_DISP:
+ case elfcpp::R_MICROMIPS_GOT_DISP:
+ if (gsym != NULL)
+ got_offset = target->got_section()->got_offset(gsym, GOT_TYPE_STANDARD,
+ object);
+ else
+ got_offset = target->got_section()->got_offset(r_sym, GOT_TYPE_STANDARD,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot(view, gp_offset, r_type);
+ break;
+
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MICROMIPS_CALL16:
+ gold_assert(gsym != NULL);
+ got_offset = target->got_section()->got_offset(gsym, GOT_TYPE_STANDARD,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot(view, gp_offset, r_type);
+ // TODO(sasa): We should also initialize update_got_entry in other places
+ // where relgot is called.
+ update_got_entry = changed_symbol_value;
+ break;
+
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MICROMIPS_GOT16:
+ if (gsym != NULL)
+ {
+ got_offset = target->got_section()->got_offset(gsym,
+ GOT_TYPE_STANDARD,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot(view, gp_offset, r_type);
+ }
+ else
+ reloc_status = Reloc_funcs::relgot16_local(view, object, psymval,
+ r_addend, extract_addend,
+ r_type);
+ update_got_entry = changed_symbol_value;
+ break;
+
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ if (gsym != NULL)
+ got_offset = target->got_section()->got_offset(gsym, GOT_TYPE_TLS_PAIR,
+ object);
+ else
+ got_offset = target->got_section()->got_offset(r_sym, GOT_TYPE_TLS_PAIR,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot(view, gp_offset, r_type);
+ break;
+
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ if (gsym != NULL)
+ got_offset = target->got_section()->got_offset(gsym,
+ GOT_TYPE_TLS_OFFSET,
+ object);
+ else
+ got_offset = target->got_section()->got_offset(r_sym,
+ GOT_TYPE_TLS_OFFSET,
+ object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot(view, gp_offset, r_type);
+ break;
+
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ // Relocate the field with the offset of the GOT entry for
+ // the module index.
+ got_offset = target->got_section()->tls_ldm_offset(object);
+ gp_offset = target->got_section()->gp_offset(got_offset, object);
+ reloc_status = Reloc_funcs::relgot(view, gp_offset, r_type);
+ break;
+
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MICROMIPS_GOT_PAGE:
+ reloc_status = Reloc_funcs::relgotpage(target, view, object, psymval,
+ r_addend, extract_addend, r_type);
+ break;
+
+ case elfcpp::R_MIPS_GOT_OFST:
+ case elfcpp::R_MICROMIPS_GOT_OFST:
+ reloc_status = Reloc_funcs::relgotofst(target, view, object, psymval,
+ r_addend, extract_addend, local,
+ r_type);
+ break;
+
+ case elfcpp::R_MIPS_JALR:
+ case elfcpp::R_MICROMIPS_JALR:
+ // This relocation is only a hint. In some cases, we optimize
+ // it into a bal instruction. But we don't try to optimize
+ // when the symbol does not resolve locally.
+ if (gsym == NULL || symbol_calls_local(gsym, gsym->has_dynsym_index()))
+ reloc_status = Reloc_funcs::reljalr(view, object, psymval, address,
+ r_addend, extract_addend,
+ cross_mode_jump, r_type,
+ target->jalr_to_bal(),
+ target->jr_to_b());
+ break;
+
+ case elfcpp::R_MIPS_TLS_DTPREL_HI16:
+ case elfcpp::R_MIPS16_TLS_DTPREL_HI16:
+ case elfcpp::R_MICROMIPS_TLS_DTPREL_HI16:
+ reloc_status = Reloc_funcs::tlsrelhi16(view, object, psymval,
+ elfcpp::DTP_OFFSET, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_TLS_DTPREL_LO16:
+ case elfcpp::R_MIPS16_TLS_DTPREL_LO16:
+ case elfcpp::R_MICROMIPS_TLS_DTPREL_LO16:
+ reloc_status = Reloc_funcs::tlsrello16(view, object, psymval,
+ elfcpp::DTP_OFFSET, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_TLS_DTPREL32:
+ case elfcpp::R_MIPS_TLS_DTPREL64:
+ reloc_status = Reloc_funcs::tlsrel32(view, object, psymval,
+ elfcpp::DTP_OFFSET, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_TLS_TPREL_HI16:
+ case elfcpp::R_MIPS16_TLS_TPREL_HI16:
+ case elfcpp::R_MICROMIPS_TLS_TPREL_HI16:
+ reloc_status = Reloc_funcs::tlsrelhi16(view, object, psymval,
+ elfcpp::TP_OFFSET, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_TLS_TPREL_LO16:
+ case elfcpp::R_MIPS16_TLS_TPREL_LO16:
+ case elfcpp::R_MICROMIPS_TLS_TPREL_LO16:
+ reloc_status = Reloc_funcs::tlsrello16(view, object, psymval,
+ elfcpp::TP_OFFSET, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_TLS_TPREL32:
+ case elfcpp::R_MIPS_TLS_TPREL64:
+ reloc_status = Reloc_funcs::tlsrel32(view, object, psymval,
+ elfcpp::TP_OFFSET, r_addend,
+ extract_addend, r_type);
+ break;
+ case elfcpp::R_MIPS_SUB:
+ case elfcpp::R_MICROMIPS_SUB:
+ reloc_status = Reloc_funcs::relsub(view, object, psymval, r_addend,
+ extract_addend, r_type);
+ break;
+ default:
+ gold_error_at_location(relinfo, relnum, r_offset,
+ _("unsupported reloc %u"), r_type);
+ break;
+ }
+
+ if (update_got_entry)
+ {
+ Mips_output_data_got<size, big_endian>* got = target->got_section();
+ if (mips_sym != NULL && mips_sym->get_applied_secondary_got_fixup())
+ got->update_got_entry(got->get_primary_got_offset(mips_sym),
+ psymval->value(object, 0));
+ else
+ got->update_got_entry(got_offset, psymval->value(object, 0));
+ }
+
+ // Report any errors.
+ switch (reloc_status)
+ {
+ case Reloc_funcs::STATUS_OKAY:
+ break;
+ case Reloc_funcs::STATUS_OVERFLOW:
+ gold_error_at_location(relinfo, relnum, r_offset,
+ _("relocation overflow"));
+ break;
+ case Reloc_funcs::STATUS_BAD_RELOC:
+ gold_error_at_location(relinfo, relnum, r_offset,
+ _("unexpected opcode while processing relocation"));
+ break;
+ default:
+ gold_unreachable();
+ }
+
+ return true;
+}
+
+template<int size, bool big_endian>
+inline bool
+Target_mips<size, big_endian>::Relocate::relocate(
+ const Relocate_info<size, big_endian>* relinfo,
+ Target_mips* target,
+ Output_section* output_section,
+ size_t relnum,
+ const elfcpp::Rela<size, big_endian>& reloc,
+ unsigned int r_type,
+ const Sized_symbol<size>* gsym,
+ const Symbol_value<size>* psymval,
+ unsigned char* view,
+ Mips_address address,
+ section_size_type view_size)
+{
+ return relocate(
+ relinfo,
+ target,
+ output_section,
+ relnum,
+ &reloc,
+ (const elfcpp::Rel<size, big_endian>*) NULL,
+ elfcpp::SHT_RELA,
+ r_type,
+ gsym,
+ psymval,
+ view,
+ address,
+ view_size);
+}
+
+template<int size, bool big_endian>
+inline bool
+Target_mips<size, big_endian>::Relocate::relocate(
+ const Relocate_info<size, big_endian>* relinfo,
+ Target_mips* target,
+ Output_section* output_section,
+ size_t relnum,
+ const elfcpp::Rel<size, big_endian>& reloc,
+ unsigned int r_type,
+ const Sized_symbol<size>* gsym,
+ const Symbol_value<size>* psymval,
+ unsigned char* view,
+ Mips_address address,
+ section_size_type view_size)
+{
+ return relocate(
+ relinfo,
+ target,
+ output_section,
+ relnum,
+ (const elfcpp::Rela<size, big_endian>*) NULL,
+ &reloc,
+ elfcpp::SHT_REL,
+ r_type,
+ gsym,
+ psymval,
+ view,
+ address,
+ view_size);
+}
+
+// Get the Reference_flags for a particular relocation.
+
+template<int size, bool big_endian>
+int
+Target_mips<size, big_endian>::Scan::get_reference_flags(
+ unsigned int r_type)
+{
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_NONE:
+ // No symbol reference.
+ return 0;
+
+ case elfcpp::R_MIPS_16:
+ case elfcpp::R_MIPS_32:
+ case elfcpp::R_MIPS_64:
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MIPS_LO16:
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MIPS16_LO16:
+ case elfcpp::R_MICROMIPS_HI16:
+ case elfcpp::R_MICROMIPS_LO16:
+ return Symbol::ABSOLUTE_REF;
+
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MIPS16_26:
+ case elfcpp::R_MICROMIPS_26_S1:
+ return Symbol::FUNCTION_CALL | Symbol::ABSOLUTE_REF;
+
+ case elfcpp::R_MIPS_GPREL32:
+ case elfcpp::R_MIPS_GPREL16:
+ case elfcpp::R_MIPS_REL32:
+ case elfcpp::R_MIPS16_GPREL:
+ return Symbol::RELATIVE_REF;
+
+ case elfcpp::R_MIPS_PC16:
+ case elfcpp::R_MIPS_PC32:
+ case elfcpp::R_MIPS_JALR:
+ case elfcpp::R_MICROMIPS_JALR:
+ return Symbol::FUNCTION_CALL | Symbol::RELATIVE_REF;
+
+ case elfcpp::R_MIPS_GOT16:
+ case elfcpp::R_MIPS_CALL16:
+ case elfcpp::R_MIPS_GOT_DISP:
+ case elfcpp::R_MIPS_GOT_HI16:
+ case elfcpp::R_MIPS_GOT_LO16:
+ case elfcpp::R_MIPS_CALL_HI16:
+ case elfcpp::R_MIPS_CALL_LO16:
+ case elfcpp::R_MIPS_LITERAL:
+ case elfcpp::R_MIPS_GOT_PAGE:
+ case elfcpp::R_MIPS_GOT_OFST:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MICROMIPS_GOT16:
+ case elfcpp::R_MICROMIPS_CALL16:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ // Absolute in GOT.
+ return Symbol::RELATIVE_REF;
+
+ case elfcpp::R_MIPS_TLS_DTPMOD32:
+ case elfcpp::R_MIPS_TLS_DTPREL32:
+ case elfcpp::R_MIPS_TLS_DTPMOD64:
+ case elfcpp::R_MIPS_TLS_DTPREL64:
+ case elfcpp::R_MIPS_TLS_GD:
+ case elfcpp::R_MIPS_TLS_LDM:
+ case elfcpp::R_MIPS_TLS_DTPREL_HI16:
+ case elfcpp::R_MIPS_TLS_DTPREL_LO16:
+ case elfcpp::R_MIPS_TLS_GOTTPREL:
+ case elfcpp::R_MIPS_TLS_TPREL32:
+ case elfcpp::R_MIPS_TLS_TPREL64:
+ case elfcpp::R_MIPS_TLS_TPREL_HI16:
+ case elfcpp::R_MIPS_TLS_TPREL_LO16:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_TPREL_HI16:
+ case elfcpp::R_MICROMIPS_TLS_TPREL_LO16:
+ return Symbol::TLS_REF;
+
+ case elfcpp::R_MIPS_COPY:
+ case elfcpp::R_MIPS_JUMP_SLOT:
+ default:
+ gold_unreachable();
+ // Not expected. We will give an error later.
+ return 0;
+ }
+}
+
+// Report an unsupported relocation against a local symbol.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::Scan::unsupported_reloc_local(
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int r_type)
+{
+ gold_error(_("%s: unsupported reloc %u against local symbol"),
+ object->name().c_str(), r_type);
+}
+
+// Report an unsupported relocation against a global symbol.
+
+template<int size, bool big_endian>
+void
+Target_mips<size, big_endian>::Scan::unsupported_reloc_global(
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int r_type,
+ Symbol* gsym)
+{
+ gold_error(_("%s: unsupported reloc %u against global symbol %s"),
+ object->name().c_str(), r_type, gsym->demangled_name().c_str());
+}
+
+// Return printable name for ABI.
+template<int size, bool big_endian>
+const char*
+Target_mips<size, big_endian>::elf_mips_abi_name(elfcpp::Elf_Word e_flags,
+ unsigned char ei_class)
+{
+ switch (e_flags & elfcpp::EF_MIPS_ABI)
+ {
+ case 0:
+ if ((e_flags & elfcpp::EF_MIPS_ABI2) != 0)
+ return "N32";
+ else if (elfcpp::abi_64(ei_class))
+ return "64";
+ else
+ return "none";
+ case elfcpp::E_MIPS_ABI_O32:
+ return "O32";
+ case elfcpp::E_MIPS_ABI_O64:
+ return "O64";
+ case elfcpp::E_MIPS_ABI_EABI32:
+ return "EABI32";
+ case elfcpp::E_MIPS_ABI_EABI64:
+ return "EABI64";
+ default:
+ return "unknown abi";
+ }
+}
+
+template<int size, bool big_endian>
+const char*
+Target_mips<size, big_endian>::elf_mips_mach_name(elfcpp::Elf_Word e_flags)
+{
+ switch (e_flags & elfcpp::EF_MIPS_MACH)
+ {
+ case elfcpp::E_MIPS_MACH_3900:
+ return "mips:3900";
+ case elfcpp::E_MIPS_MACH_4010:
+ return "mips:4010";
+ case elfcpp::E_MIPS_MACH_4100:
+ return "mips:4100";
+ case elfcpp::E_MIPS_MACH_4111:
+ return "mips:4111";
+ case elfcpp::E_MIPS_MACH_4120:
+ return "mips:4120";
+ case elfcpp::E_MIPS_MACH_4650:
+ return "mips:4650";
+ case elfcpp::E_MIPS_MACH_5400:
+ return "mips:5400";
+ case elfcpp::E_MIPS_MACH_5500:
+ return "mips:5500";
+ case elfcpp::E_MIPS_MACH_SB1:
+ return "mips:sb1";
+ case elfcpp::E_MIPS_MACH_9000:
+ return "mips:9000";
+ case elfcpp::E_MIPS_MACH_LS2E:
+ return "mips:loongson-2e";
+ case elfcpp::E_MIPS_MACH_LS2F:
+ return "mips:loongson-2f";
+ case elfcpp::E_MIPS_MACH_LS3A:
+ return "mips:loongson-3a";
+ case elfcpp::E_MIPS_MACH_OCTEON:
+ return "mips:octeon";
+ case elfcpp::E_MIPS_MACH_OCTEON2:
+ return "mips:octeon2";
+ case elfcpp::E_MIPS_MACH_XLR:
+ return "mips:xlr";
+ default:
+ switch (e_flags & elfcpp::EF_MIPS_ARCH)
+ {
+ default:
+ case elfcpp::E_MIPS_ARCH_1:
+ return "mips:3000";
+
+ case elfcpp::E_MIPS_ARCH_2:
+ return "mips:6000";
+
+ case elfcpp::E_MIPS_ARCH_3:
+ return "mips:4000";
+
+ case elfcpp::E_MIPS_ARCH_4:
+ return "mips:8000";
+
+ case elfcpp::E_MIPS_ARCH_5:
+ return "mips:mips5";
+
+ case elfcpp::E_MIPS_ARCH_32:
+ return "mips:isa32";
+
+ case elfcpp::E_MIPS_ARCH_64:
+ return "mips:isa64";
+
+ case elfcpp::E_MIPS_ARCH_32R2:
+ return "mips:isa32r2";
+
+ case elfcpp::E_MIPS_ARCH_64R2:
+ return "mips:isa64r2";
+ }
+ }
+ return "unknown CPU";
+}
+
+template<int size, bool big_endian>
+Target::Target_info Target_mips<size, big_endian>::mips_info =
+{
+ size, // size
+ big_endian, // is_big_endian
+ elfcpp::EM_MIPS, // machine_code
+ true, // has_make_symbol
+ false, // has_resolve
+ false, // has_code_fill
+ true, // is_default_stack_executable
+ false, // can_icf_inline_merge_sections
+ '\0', // wrap_char
+ "/lib/ld.so.1", // dynamic_linker
+ 0x400000, // default_text_segment_address
+ 64 * 1024, // abi_pagesize (overridable by -z max-page-size)
+ 4 * 1024, // common_pagesize (overridable by -z common-page-size)
+ false, // isolate_execinstr
+ 0, // rosegment_gap
+ elfcpp::SHN_UNDEF, // small_common_shndx
+ elfcpp::SHN_UNDEF, // large_common_shndx
+ 0, // small_common_section_flags
+ 0, // large_common_section_flags
+ NULL, // attributes_section
+ NULL, // attributes_vendor
+ "__start" // entry_symbol_name
+};
+
+// The selector for mips object files.
+
+template<int size, bool big_endian>
+class Target_selector_mips : public Target_selector
+{
+public:
+ Target_selector_mips()
+ : Target_selector(elfcpp::EM_MIPS, size, big_endian,
+ (size == 64 ?
+ (big_endian ? "elf64-tradbigmips" : "elf64-tradlittlemips") :
+ (big_endian ? "elf32-tradbigmips" : "elf32-tradlittlemips")),
+ (size == 64 ?
+ (big_endian ? "elf64-tradbigmips" : "elf64-tradlittlemips") :
+ (big_endian ? "elf32-tradbigmips" : "elf32-tradlittlemips")))
+ { }
+
+ Target* do_instantiate_target()
+ { return new Target_mips<size, big_endian>(); }
+};
+
+Target_selector_mips<32, true> target_selector_mips32be;
+Target_selector_mips<32, false> target_selector_mips32;
+Target_selector_mips<64, true> target_selector_mips64be;
+Target_selector_mips<64, false> target_selector_mips64;
+
+
+} // End anonymous namespace.
^ permalink raw reply [flat|nested] 9+ messages in thread
* RE: Mips target in gold - revision 3 - part 2
2014-06-04 21:01 ` Sasa Stankovic
@ 2014-06-04 21:32 ` Sasa Stankovic
2014-06-23 19:10 ` Cary Coutant
0 siblings, 1 reply; 9+ messages in thread
From: Sasa Stankovic @ 2014-06-04 21:32 UTC (permalink / raw)
To: Cary Coutant; +Cc: Rich Fuhler, binutils, Petar Jovanovic
[-- Attachment #1: Type: text/plain, Size: 4800 bytes --]
I forgot to remove the call to Mips_output_section_reginfo::set_always_keeps_input_sections(). Attached is the updated patch where the call is removed.
Regards,
Sasa
________________________________________
From: Sasa Stankovic
Sent: Wednesday, June 04, 2014 11:00 PM
To: Cary Coutant
Cc: Rich Fuhler; binutils@sourceware.org; Petar Jovanovic
Subject: RE: Mips target in gold - revision 3 - part 2
Hi Cary,
Attached is the first part of the updated patch. Sorry for the delay; I didn't respond immediately after your last message because I thought there will be more reviews.
> > // Fill the unused space with zeroes.
> > // TODO(sasa): Can we strip unused bytes during the relaxation?
> > unsigned char* end = oview + oview_size;
> > while (pov < end)
> > {
> > elfcpp::Swap<32, big_endian>::writeval(pov, 0);
> > pov += 4;
> > }
>
> You can just use memset here:
>
> if (unused > 0)
> memset(pov, 0, unused);
Done.
> One more issue I just noticed...
>
> In Mips_output_section_reginfo::do_write:
>
> > const unsigned char* section_contents =
> > relobj->section_contents(shndx, §ion_size, false);
> >
> > gprmask |= elfcpp::Swap<size, big_endian>::readval(section_contents);
> > cprmask1 |= elfcpp::Swap<size, big_endian>::readval(section_contents + 4);
> > cprmask2 |= elfcpp::Swap<size, big_endian>::readval(section_contents + 8);
> > cprmask3 |=
> > elfcpp::Swap<size, big_endian>::readval(section_contents + 12);
> > cprmask4 |=
> > elfcpp::Swap<size, big_endian>::readval(section_contents + 16);
>
> Going back to the input sections at this point could be a performance
> issue, especially for very large links. This is essentially another
> pass over the input files, and if we've run out of file descriptors,
> we'll have to close and reopen all the input files one more time each.
> It would be better to read these values and accumulate the five masks
> in do_read_symbols(), and you won't have to call
> set_always_keeps_input_sections() for Mips_output_section_reginfo.
Done.
Regards,
Sasa
________________________________________
From: Cary Coutant [ccoutant@google.com]
Sent: Tuesday, April 15, 2014 10:46 PM
To: Sasa Stankovic
Cc: Rich Fuhler; binutils@sourceware.org; iant@google.com; Petar Jovanovic
Subject: Re: Mips target in gold - revision 3 - part 2
>> If I understand the code correctly, in this case the HI16 reloc
>> will remain on the got16_addends_ list forever. Is there a point
>> where you know it's safe to remove it? Can you clear the list
>> at the end of each section?
>
> If LO16 is removed by GCC, than HI16 reloc is paired with the next found LO16
> reloc (this means that two or more HI16 reloc will be paired with a single LO16 reloc).
> This is what GNU ld does - it always tries to find LO16 reloc for HI16 reloc, and if it
> can't find it, error is reported.
>
> When trying to find LO16 reloc, GNU ld examines next relocations in the section until
> it finds it, or reports an error. I implemented it differently in Gold - when HI16 or GOT16
> reloc is found, it is recorded in the got16_addends_ list, and the normal scaning of
> relocations continues. Then when the LO16 reloc is found, the got16_addends_ list is
> examined and all pending matching HI16 and GOT16 relocs are removed. The error when
> HI16 or GOT16 reloc doesn't have the LO16 part is detected by checking whether the
> section of the pending HI16 or GOT16 relocation is different from the section of the current
> LO16 relocation. The check whether the final section that is scanned has HI16 or GOT16
> without the LO16 part is in the do_finalize_sections method - I just check whether the
> got16_addends_ list is empty - if it isn't, error is reported.
Ah, I see.
>> > typename std::list<got16_addend<size, big_endian> > got16_addends_;
>>
>> Do you realy want to use a linked list here? Wouldn't a
>> vector be better?
>
> I used std::list because erase is called on it.
Sorry, I missed that. And now I understand that you don't expect the
list to ever grow very long.
> // Check whether the final section that was scaned has HI16 or GOT16
Typo: "scanned".
> // Fill the unused space with zeroes.
> // TODO(sasa): Can we strip unused bytes during the relaxation?
> unsigned char* end = oview + oview_size;
> while (pov < end)
> {
> elfcpp::Swap<32, big_endian>::writeval(pov, 0);
> pov += 4;
> }
You can just use memset here:
if (unused > 0)
memset(pov, 0, unused);
If that's OK with you, I'll commit it with that change.
Thanks for working on this, and for putting up with the long review process!
-cary
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: mips1.patch --]
[-- Type: text/x-patch; name="mips1.patch", Size: 187105 bytes --]
diff --git a/gold/mips.cc b/gold/mips.cc
new file mode 100644
index 0000000..b0fcc93
--- /dev/null
+++ b/gold/mips.cc
@@ -0,0 +1,5252 @@
+// mips.cc -- mips target support for gold.
+
+// Copyright (C) 2011-2014 Free Software Foundation, Inc.
+// Written by Sasa Stankovic <sasa.stankovic@imgtec.com>
+// and Aleksandar Simeonov <aleksandar.simeonov@rt-rk.com>.
+// This file contains borrowed and adapted code from bfd/elfxx-mips.c.
+
+// This file is part of gold.
+
+// This program is free software; you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation; either version 3 of the License, or
+// (at your option) any later version.
+
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU General Public License for more details.
+
+// You should have received a copy of the GNU General Public License
+// along with this program; if not, write to the Free Software
+// Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
+// MA 02110-1301, USA.
+
+#include "gold.h"
+
+#include <algorithm>
+#include <set>
+#include <sstream>
+#include "demangle.h"
+
+#include "elfcpp.h"
+#include "parameters.h"
+#include "reloc.h"
+#include "mips.h"
+#include "object.h"
+#include "symtab.h"
+#include "layout.h"
+#include "output.h"
+#include "copy-relocs.h"
+#include "target.h"
+#include "target-reloc.h"
+#include "target-select.h"
+#include "tls.h"
+#include "errors.h"
+#include "gc.h"
+
+namespace
+{
+using namespace gold;
+
+template<int size, bool big_endian>
+class Mips_output_data_plt;
+
+template<int size, bool big_endian>
+class Mips_output_data_got;
+
+template<int size, bool big_endian>
+class Target_mips;
+
+template<int size, bool big_endian>
+class Mips_output_section_reginfo;
+
+template<int size, bool big_endian>
+class Mips_output_data_la25_stub;
+
+template<int size, bool big_endian>
+class Mips_output_data_mips_stubs;
+
+template<int size>
+class Mips_symbol;
+
+template<int size, bool big_endian>
+class Mips_got_info;
+
+template<int size, bool big_endian>
+class Mips_relobj;
+
+class Mips16_stub_section_base;
+
+template<int size, bool big_endian>
+class Mips16_stub_section;
+
+// The ABI says that every symbol used by dynamic relocations must have
+// a global GOT entry. Among other things, this provides the dynamic
+// linker with a free, directly-indexed cache. The GOT can therefore
+// contain symbols that are not referenced by GOT relocations themselves
+// (in other words, it may have symbols that are not referenced by things
+// like R_MIPS_GOT16 and R_MIPS_GOT_PAGE).
+
+// GOT relocations are less likely to overflow if we put the associated
+// GOT entries towards the beginning. We therefore divide the global
+// GOT entries into two areas: "normal" and "reloc-only". Entries in
+// the first area can be used for both dynamic relocations and GP-relative
+// accesses, while those in the "reloc-only" area are for dynamic
+// relocations only.
+
+// These GGA_* ("Global GOT Area") values are organised so that lower
+// values are more general than higher values. Also, non-GGA_NONE
+// values are ordered by the position of the area in the GOT.
+
+enum Global_got_area
+{
+ GGA_NORMAL = 0,
+ GGA_RELOC_ONLY = 1,
+ GGA_NONE = 2
+};
+
+// The types of GOT entries needed for this platform.
+// These values are exposed to the ABI in an incremental link.
+// Do not renumber existing values without changing the version
+// number of the .gnu_incremental_inputs section.
+enum Got_type
+{
+ GOT_TYPE_STANDARD = 0, // GOT entry for a regular symbol
+ GOT_TYPE_TLS_OFFSET = 1, // GOT entry for TLS offset
+ GOT_TYPE_TLS_PAIR = 2, // GOT entry for TLS module/offset pair
+
+ // GOT entries for multi-GOT. We support up to 1024 GOTs in multi-GOT links.
+ GOT_TYPE_STANDARD_MULTIGOT = 3,
+ GOT_TYPE_TLS_OFFSET_MULTIGOT = GOT_TYPE_STANDARD_MULTIGOT + 1024,
+ GOT_TYPE_TLS_PAIR_MULTIGOT = GOT_TYPE_TLS_OFFSET_MULTIGOT + 1024
+};
+
+// TLS type of GOT entry.
+enum Got_tls_type
+{
+ GOT_TLS_NONE = 0,
+ GOT_TLS_GD = 1,
+ GOT_TLS_LDM = 2,
+ GOT_TLS_IE = 4
+};
+
+// Return TRUE if a relocation of type R_TYPE from OBJECT might
+// require an la25 stub. See also local_pic_function, which determines
+// whether the destination function ever requires a stub.
+template<int size, bool big_endian>
+static inline bool
+relocation_needs_la25_stub(Mips_relobj<size, big_endian>* object,
+ unsigned int r_type, bool target_is_16_bit_code)
+{
+ // We specifically ignore branches and jumps from EF_PIC objects,
+ // where the onus is on the compiler or programmer to perform any
+ // necessary initialization of $25. Sometimes such initialization
+ // is unnecessary; for example, -mno-shared functions do not use
+ // the incoming value of $25, and may therefore be called directly.
+ if (object->is_pic())
+ return false;
+
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_26:
+ case elfcpp::R_MIPS_PC16:
+ case elfcpp::R_MICROMIPS_26_S1:
+ case elfcpp::R_MICROMIPS_PC7_S1:
+ case elfcpp::R_MICROMIPS_PC10_S1:
+ case elfcpp::R_MICROMIPS_PC16_S1:
+ case elfcpp::R_MICROMIPS_PC23_S2:
+ return true;
+
+ case elfcpp::R_MIPS16_26:
+ return !target_is_16_bit_code;
+
+ default:
+ return false;
+ }
+}
+
+// Return true if SYM is a locally-defined PIC function, in the sense
+// that it or its fn_stub might need $25 to be valid on entry.
+// Note that MIPS16 functions set up $gp using PC-relative instructions,
+// so they themselves never need $25 to be valid. Only non-MIPS16
+// entry points are of interest here.
+template<int size, bool big_endian>
+static inline bool
+local_pic_function(Mips_symbol<size>* sym)
+{
+ bool def_regular = (sym->source() == Symbol::FROM_OBJECT
+ && !sym->object()->is_dynamic()
+ && !sym->is_undefined());
+
+ if (sym->is_defined() && def_regular)
+ {
+ Mips_relobj<size, big_endian>* object =
+ static_cast<Mips_relobj<size, big_endian>*>(sym->object());
+
+ if ((object->is_pic() || sym->is_pic())
+ && (!sym->is_mips16()
+ || (sym->has_mips16_fn_stub() && sym->need_fn_stub())))
+ return true;
+ }
+ return false;
+}
+
+static inline bool
+hi16_reloc(int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_HI16
+ || r_type == elfcpp::R_MIPS16_HI16
+ || r_type == elfcpp::R_MICROMIPS_HI16);
+}
+
+static inline bool
+lo16_reloc(int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_LO16
+ || r_type == elfcpp::R_MIPS16_LO16
+ || r_type == elfcpp::R_MICROMIPS_LO16);
+}
+
+static inline bool
+got16_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_GOT16
+ || r_type == elfcpp::R_MIPS16_GOT16
+ || r_type == elfcpp::R_MICROMIPS_GOT16);
+}
+
+static inline bool
+call_lo16_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_CALL_LO16
+ || r_type == elfcpp::R_MICROMIPS_CALL_LO16);
+}
+
+static inline bool
+got_lo16_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_GOT_LO16
+ || r_type == elfcpp::R_MICROMIPS_GOT_LO16);
+}
+
+static inline bool
+got_disp_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_GOT_DISP
+ || r_type == elfcpp::R_MICROMIPS_GOT_DISP);
+}
+
+static inline bool
+got_page_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_GOT_PAGE
+ || r_type == elfcpp::R_MICROMIPS_GOT_PAGE);
+}
+
+static inline bool
+tls_gd_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_TLS_GD
+ || r_type == elfcpp::R_MIPS16_TLS_GD
+ || r_type == elfcpp::R_MICROMIPS_TLS_GD);
+}
+
+static inline bool
+tls_gottprel_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_TLS_GOTTPREL
+ || r_type == elfcpp::R_MIPS16_TLS_GOTTPREL
+ || r_type == elfcpp::R_MICROMIPS_TLS_GOTTPREL);
+}
+
+static inline bool
+tls_ldm_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_TLS_LDM
+ || r_type == elfcpp::R_MIPS16_TLS_LDM
+ || r_type == elfcpp::R_MICROMIPS_TLS_LDM);
+}
+
+static inline bool
+mips16_call_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS16_26
+ || r_type == elfcpp::R_MIPS16_CALL16);
+}
+
+static inline bool
+jal_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MIPS_26
+ || r_type == elfcpp::R_MIPS16_26
+ || r_type == elfcpp::R_MICROMIPS_26_S1);
+}
+
+static inline bool
+micromips_branch_reloc(unsigned int r_type)
+{
+ return (r_type == elfcpp::R_MICROMIPS_26_S1
+ || r_type == elfcpp::R_MICROMIPS_PC16_S1
+ || r_type == elfcpp::R_MICROMIPS_PC10_S1
+ || r_type == elfcpp::R_MICROMIPS_PC7_S1);
+}
+
+// Check if R_TYPE is a MIPS16 reloc.
+static inline bool
+mips16_reloc(unsigned int r_type)
+{
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS16_26:
+ case elfcpp::R_MIPS16_GPREL:
+ case elfcpp::R_MIPS16_GOT16:
+ case elfcpp::R_MIPS16_CALL16:
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MIPS16_LO16:
+ case elfcpp::R_MIPS16_TLS_GD:
+ case elfcpp::R_MIPS16_TLS_LDM:
+ case elfcpp::R_MIPS16_TLS_DTPREL_HI16:
+ case elfcpp::R_MIPS16_TLS_DTPREL_LO16:
+ case elfcpp::R_MIPS16_TLS_GOTTPREL:
+ case elfcpp::R_MIPS16_TLS_TPREL_HI16:
+ case elfcpp::R_MIPS16_TLS_TPREL_LO16:
+ return true;
+
+ default:
+ return false;
+ }
+}
+
+// Check if R_TYPE is a microMIPS reloc.
+static inline bool
+micromips_reloc(unsigned int r_type)
+{
+ switch (r_type)
+ {
+ case elfcpp::R_MICROMIPS_26_S1:
+ case elfcpp::R_MICROMIPS_HI16:
+ case elfcpp::R_MICROMIPS_LO16:
+ case elfcpp::R_MICROMIPS_GPREL16:
+ case elfcpp::R_MICROMIPS_LITERAL:
+ case elfcpp::R_MICROMIPS_GOT16:
+ case elfcpp::R_MICROMIPS_PC7_S1:
+ case elfcpp::R_MICROMIPS_PC10_S1:
+ case elfcpp::R_MICROMIPS_PC16_S1:
+ case elfcpp::R_MICROMIPS_CALL16:
+ case elfcpp::R_MICROMIPS_GOT_DISP:
+ case elfcpp::R_MICROMIPS_GOT_PAGE:
+ case elfcpp::R_MICROMIPS_GOT_OFST:
+ case elfcpp::R_MICROMIPS_GOT_HI16:
+ case elfcpp::R_MICROMIPS_GOT_LO16:
+ case elfcpp::R_MICROMIPS_SUB:
+ case elfcpp::R_MICROMIPS_HIGHER:
+ case elfcpp::R_MICROMIPS_HIGHEST:
+ case elfcpp::R_MICROMIPS_CALL_HI16:
+ case elfcpp::R_MICROMIPS_CALL_LO16:
+ case elfcpp::R_MICROMIPS_SCN_DISP:
+ case elfcpp::R_MICROMIPS_JALR:
+ case elfcpp::R_MICROMIPS_HI0_LO16:
+ case elfcpp::R_MICROMIPS_TLS_GD:
+ case elfcpp::R_MICROMIPS_TLS_LDM:
+ case elfcpp::R_MICROMIPS_TLS_DTPREL_HI16:
+ case elfcpp::R_MICROMIPS_TLS_DTPREL_LO16:
+ case elfcpp::R_MICROMIPS_TLS_GOTTPREL:
+ case elfcpp::R_MICROMIPS_TLS_TPREL_HI16:
+ case elfcpp::R_MICROMIPS_TLS_TPREL_LO16:
+ case elfcpp::R_MICROMIPS_GPREL7_S2:
+ case elfcpp::R_MICROMIPS_PC23_S2:
+ return true;
+
+ default:
+ return false;
+ }
+}
+
+static inline bool
+is_matching_lo16_reloc(unsigned int high_reloc, unsigned int lo16_reloc)
+{
+ switch (high_reloc)
+ {
+ case elfcpp::R_MIPS_HI16:
+ case elfcpp::R_MIPS_GOT16:
+ return lo16_reloc == elfcpp::R_MIPS_LO16;
+ case elfcpp::R_MIPS16_HI16:
+ case elfcpp::R_MIPS16_GOT16:
+ return lo16_reloc == elfcpp::R_MIPS16_LO16;
+ case elfcpp::R_MICROMIPS_HI16:
+ case elfcpp::R_MICROMIPS_GOT16:
+ return lo16_reloc == elfcpp::R_MICROMIPS_LO16;
+ default:
+ return false;
+ }
+}
+
+// This class is used to hold information about one GOT entry.
+// There are three types of entry:
+//
+// (1) a SYMBOL + OFFSET address, where SYMBOL is local to an input object
+// (object != NULL, symndx >= 0, tls_type != GOT_TLS_LDM)
+// (2) a SYMBOL address, where SYMBOL is not local to an input object
+// (object != NULL, symndx == -1)
+// (3) a TLS LDM slot
+// (object != NULL, symndx == 0, tls_type == GOT_TLS_LDM)
+
+template<int size, bool big_endian>
+class Mips_got_entry
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+
+ public:
+ Mips_got_entry(Mips_relobj<size, big_endian>* object, unsigned int symndx,
+ Mips_address addend, unsigned char tls_type,
+ unsigned int shndx)
+ : object_(object), symndx_(symndx), tls_type_(tls_type), shndx_(shndx)
+ { this->d.addend = addend; }
+
+ Mips_got_entry(Mips_relobj<size, big_endian>* object, Mips_symbol<size>* sym,
+ unsigned char tls_type)
+ : object_(object), symndx_(-1U), tls_type_(tls_type), shndx_(-1U)
+ { this->d.sym = sym; }
+
+ // Return whether this entry is for a local symbol.
+ bool
+ is_for_local_symbol() const
+ { return this->symndx_ != -1U; }
+
+ // Return whether this entry is for a global symbol.
+ bool
+ is_for_global_symbol() const
+ { return this->symndx_ == -1U; }
+
+ // Return the hash of this entry.
+ size_t
+ hash() const
+ {
+ if (this->tls_type_ == GOT_TLS_LDM)
+ return this->symndx_ + (1 << 18);
+ if (this->symndx_ != -1U)
+ {
+ uintptr_t object_id = reinterpret_cast<uintptr_t>(this->object());
+ return this->symndx_ + object_id + this->d.addend;
+ }
+ else
+ {
+ uintptr_t sym_id = reinterpret_cast<uintptr_t>(this->d.sym);
+ return this->symndx_ + sym_id;
+ }
+ }
+
+ // Return whether this entry is equal to OTHER.
+ bool
+ equals(Mips_got_entry<size, big_endian>* other) const
+ {
+ if (this->symndx_ != other->symndx_
+ || this->tls_type_ != other->tls_type_)
+ return false;
+ if (this->tls_type_ == GOT_TLS_LDM)
+ return true;
+ if (this->symndx_ != -1U)
+ return (this->object() == other->object()
+ && this->d.addend == other->d.addend);
+ else
+ return this->d.sym == other->d.sym;
+ }
+
+ // Return input object that needs this GOT entry.
+ Mips_relobj<size, big_endian>*
+ object() const
+ {
+ gold_assert(this->object_ != NULL);
+ return this->object_;
+ }
+
+ // Return local symbol index for local GOT entries.
+ unsigned int
+ symndx() const
+ {
+ gold_assert(this->symndx_ != -1U);
+ return this->symndx_;
+ }
+
+ // Return the relocation addend for local GOT entries.
+ Mips_address
+ addend() const
+ {
+ gold_assert(this->symndx_ != -1U);
+ return this->d.addend;
+ }
+
+ // Return global symbol for global GOT entries.
+ Mips_symbol<size>*
+ sym() const
+ {
+ gold_assert(this->symndx_ == -1U);
+ return this->d.sym;
+ }
+
+ // Return whether this is a TLS GOT entry.
+ bool
+ is_tls_entry() const
+ { return this->tls_type_ != GOT_TLS_NONE; }
+
+ // Return TLS type of this GOT entry.
+ unsigned char
+ tls_type() const
+ { return this->tls_type_; }
+
+ // Return section index of the local symbol for local GOT entries.
+ unsigned int
+ shndx() const
+ { return this->shndx_; }
+
+ private:
+ // The input object that needs the GOT entry.
+ Mips_relobj<size, big_endian>* object_;
+ // The index of the symbol if we have a local symbol; -1 otherwise.
+ unsigned int symndx_;
+
+ union
+ {
+ // If symndx != -1, the addend of the relocation that should be added to the
+ // symbol value.
+ Mips_address addend;
+ // If symndx == -1, the global symbol corresponding to this GOT entry. The
+ // symbol's entry is in the local area if mips_sym->global_got_area is
+ // GGA_NONE, otherwise it is in the global area.
+ Mips_symbol<size>* sym;
+ } d;
+
+ // The TLS type of this GOT entry. An LDM GOT entry will be a local
+ // symbol entry with r_symndx == 0.
+ unsigned char tls_type_;
+
+ // For local GOT entries, section index of the local symbol.
+ unsigned int shndx_;
+};
+
+// Hash for Mips_got_entry.
+
+template<int size, bool big_endian>
+class Mips_got_entry_hash
+{
+ public:
+ size_t
+ operator()(Mips_got_entry<size, big_endian>* entry) const
+ { return entry->hash(); }
+};
+
+// Equality for Mips_got_entry.
+
+template<int size, bool big_endian>
+class Mips_got_entry_eq
+{
+ public:
+ bool
+ operator()(Mips_got_entry<size, big_endian>* e1,
+ Mips_got_entry<size, big_endian>* e2) const
+ { return e1->equals(e2); }
+};
+
+// Got_page_range. This class describes a range of addends: [MIN_ADDEND,
+// MAX_ADDEND]. The instances form a non-overlapping list that is sorted by
+// increasing MIN_ADDEND.
+
+struct Got_page_range
+{
+ Got_page_range()
+ : next(NULL), min_addend(0), max_addend(0)
+ { }
+
+ Got_page_range* next;
+ int min_addend;
+ int max_addend;
+
+ // Return the maximum number of GOT page entries required.
+ int
+ get_max_pages()
+ { return (this->max_addend - this->min_addend + 0x1ffff) >> 16; }
+};
+
+// Got_page_entry. This class describes the range of addends that are applied
+// to page relocations against a given symbol.
+
+struct Got_page_entry
+{
+ Got_page_entry()
+ : object(NULL), symndx(-1U), ranges(NULL), num_pages(0)
+ { }
+
+ Got_page_entry(Object* object_, unsigned int symndx_)
+ : object(object_), symndx(symndx_), ranges(NULL), num_pages(0)
+ { }
+
+ // The input object that needs the GOT page entry.
+ Object* object;
+ // The index of the symbol, as stored in the relocation r_info.
+ unsigned int symndx;
+ // The ranges for this page entry.
+ Got_page_range* ranges;
+ // The maximum number of page entries needed for RANGES.
+ unsigned int num_pages;
+};
+
+// Hash for Got_page_entry.
+
+struct Got_page_entry_hash
+{
+ size_t
+ operator()(Got_page_entry* entry) const
+ { return reinterpret_cast<uintptr_t>(entry->object) + entry->symndx; }
+};
+
+// Equality for Got_page_entry.
+
+struct Got_page_entry_eq
+{
+ bool
+ operator()(Got_page_entry* entry1, Got_page_entry* entry2) const
+ {
+ return entry1->object == entry2->object && entry1->symndx == entry2->symndx;
+ }
+};
+
+// This class is used to hold .got information when linking.
+
+template<int size, bool big_endian>
+class Mips_got_info
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+ typedef Output_data_reloc<elfcpp::SHT_REL, true, size, big_endian>
+ Reloc_section;
+ typedef Unordered_map<unsigned int, unsigned int> Got_page_offsets;
+
+ // Unordered set of GOT entries.
+ typedef Unordered_set<Mips_got_entry<size, big_endian>*,
+ Mips_got_entry_hash<size, big_endian>,
+ Mips_got_entry_eq<size, big_endian> > Got_entry_set;
+
+ // Unordered set of GOT page entries.
+ typedef Unordered_set<Got_page_entry*,
+ Got_page_entry_hash, Got_page_entry_eq> Got_page_entry_set;
+
+ public:
+ Mips_got_info()
+ : local_gotno_(0), page_gotno_(0), global_gotno_(0), reloc_only_gotno_(0),
+ tls_gotno_(0), tls_ldm_offset_(-1U), global_got_symbols_(),
+ got_entries_(), got_page_entries_(), got_page_offset_start_(0),
+ got_page_offset_next_(0), got_page_offsets_(), next_(NULL), index_(-1U),
+ offset_(0)
+ { }
+
+ // Reserve GOT entry for a GOT relocation of type R_TYPE against symbol
+ // SYMNDX + ADDEND, where SYMNDX is a local symbol in section SHNDX in OBJECT.
+ void
+ record_local_got_symbol(Mips_relobj<size, big_endian>* object,
+ unsigned int symndx, Mips_address addend,
+ unsigned int r_type, unsigned int shndx);
+
+ // Reserve GOT entry for a GOT relocation of type R_TYPE against MIPS_SYM,
+ // in OBJECT. FOR_CALL is true if the caller is only interested in
+ // using the GOT entry for calls. DYN_RELOC is true if R_TYPE is a dynamic
+ // relocation.
+ void
+ record_global_got_symbol(Mips_symbol<size>* mips_sym,
+ Mips_relobj<size, big_endian>* object,
+ unsigned int r_type, bool dyn_reloc, bool for_call);
+
+ // Add ENTRY to master GOT and to OBJECT's GOT.
+ void
+ record_got_entry(Mips_got_entry<size, big_endian>* entry,
+ Mips_relobj<size, big_endian>* object);
+
+ // Record that OBJECT has a page relocation against symbol SYMNDX and
+ // that ADDEND is the addend for that relocation.
+ void
+ record_got_page_entry(Mips_relobj<size, big_endian>* object,
+ unsigned int symndx, int addend);
+
+ // Create all entries that should be in the local part of the GOT.
+ void
+ add_local_entries(Target_mips<size, big_endian>* target, Layout* layout);
+
+ // Create GOT page entries.
+ void
+ add_page_entries(Target_mips<size, big_endian>* target, Layout* layout);
+
+ // Create global GOT entries, both GGA_NORMAL and GGA_RELOC_ONLY.
+ void
+ add_global_entries(Target_mips<size, big_endian>* target, Layout* layout,
+ unsigned int non_reloc_only_global_gotno);
+
+ // Create global GOT entries that should be in the GGA_RELOC_ONLY area.
+ void
+ add_reloc_only_entries(Mips_output_data_got<size, big_endian>* got);
+
+ // Create TLS GOT entries.
+ void
+ add_tls_entries(Target_mips<size, big_endian>* target, Layout* layout);
+
+ // Decide whether the symbol needs an entry in the global part of the primary
+ // GOT, setting global_got_area accordingly. Count the number of global
+ // symbols that are in the primary GOT only because they have dynamic
+ // relocations R_MIPS_REL32 against them (reloc_only_gotno).
+ void
+ count_got_symbols(Symbol_table* symtab);
+
+ // Return the offset of GOT page entry for VALUE.
+ unsigned int
+ get_got_page_offset(Mips_address value,
+ Mips_output_data_got<size, big_endian>* got);
+
+ // Count the number of GOT entries required.
+ void
+ count_got_entries();
+
+ // Count the number of GOT entries required by ENTRY. Accumulate the result.
+ void
+ count_got_entry(Mips_got_entry<size, big_endian>* entry);
+
+ // Add FROM's GOT entries.
+ void
+ add_got_entries(Mips_got_info<size, big_endian>* from);
+
+ // Add FROM's GOT page entries.
+ void
+ add_got_page_entries(Mips_got_info<size, big_endian>* from);
+
+ // Return GOT size.
+ unsigned int
+ got_size() const
+ { return ((2 + this->local_gotno_ + this->page_gotno_ + this->global_gotno_
+ + this->tls_gotno_) * size/8);
+ }
+
+ // Return the number of local GOT entries.
+ unsigned int
+ local_gotno() const
+ { return this->local_gotno_; }
+
+ // Return the maximum number of page GOT entries needed.
+ unsigned int
+ page_gotno() const
+ { return this->page_gotno_; }
+
+ // Return the number of global GOT entries.
+ unsigned int
+ global_gotno() const
+ { return this->global_gotno_; }
+
+ // Set the number of global GOT entries.
+ void
+ set_global_gotno(unsigned int global_gotno)
+ { this->global_gotno_ = global_gotno; }
+
+ // Return the number of GGA_RELOC_ONLY global GOT entries.
+ unsigned int
+ reloc_only_gotno() const
+ { return this->reloc_only_gotno_; }
+
+ // Return the number of TLS GOT entries.
+ unsigned int
+ tls_gotno() const
+ { return this->tls_gotno_; }
+
+ // Return the GOT type for this GOT. Used for multi-GOT links only.
+ unsigned int
+ multigot_got_type(unsigned int got_type) const
+ {
+ switch (got_type)
+ {
+ case GOT_TYPE_STANDARD:
+ return GOT_TYPE_STANDARD_MULTIGOT + this->index_;
+ case GOT_TYPE_TLS_OFFSET:
+ return GOT_TYPE_TLS_OFFSET_MULTIGOT + this->index_;
+ case GOT_TYPE_TLS_PAIR:
+ return GOT_TYPE_TLS_PAIR_MULTIGOT + this->index_;
+ default:
+ gold_unreachable();
+ }
+ }
+
+ // Remove lazy-binding stubs for global symbols in this GOT.
+ void
+ remove_lazy_stubs(Target_mips<size, big_endian>* target);
+
+ // Return offset of this GOT from the start of .got section.
+ unsigned int
+ offset() const
+ { return this->offset_; }
+
+ // Set offset of this GOT from the start of .got section.
+ void
+ set_offset(unsigned int offset)
+ { this->offset_ = offset; }
+
+ // Set index of this GOT in multi-GOT links.
+ void
+ set_index(unsigned int index)
+ { this->index_ = index; }
+
+ // Return next GOT in multi-GOT links.
+ Mips_got_info<size, big_endian>*
+ next() const
+ { return this->next_; }
+
+ // Set next GOT in multi-GOT links.
+ void
+ set_next(Mips_got_info<size, big_endian>* next)
+ { this->next_ = next; }
+
+ // Return the offset of TLS LDM entry for this GOT.
+ unsigned int
+ tls_ldm_offset() const
+ { return this->tls_ldm_offset_; }
+
+ // Set the offset of TLS LDM entry for this GOT.
+ void
+ set_tls_ldm_offset(unsigned int tls_ldm_offset)
+ { this->tls_ldm_offset_ = tls_ldm_offset; }
+
+ Unordered_set<Mips_symbol<size>*>&
+ global_got_symbols()
+ { return this->global_got_symbols_; }
+
+ // Return the GOT_TLS_* type required by relocation type R_TYPE.
+ static int
+ mips_elf_reloc_tls_type(unsigned int r_type)
+ {
+ if (tls_gd_reloc(r_type))
+ return GOT_TLS_GD;
+
+ if (tls_ldm_reloc(r_type))
+ return GOT_TLS_LDM;
+
+ if (tls_gottprel_reloc(r_type))
+ return GOT_TLS_IE;
+
+ return GOT_TLS_NONE;
+ }
+
+ // Return the number of GOT slots needed for GOT TLS type TYPE.
+ static int
+ mips_tls_got_entries(unsigned int type)
+ {
+ switch (type)
+ {
+ case GOT_TLS_GD:
+ case GOT_TLS_LDM:
+ return 2;
+
+ case GOT_TLS_IE:
+ return 1;
+
+ case GOT_TLS_NONE:
+ return 0;
+
+ default:
+ gold_unreachable();
+ }
+ }
+
+ private:
+ // The number of local GOT entries.
+ unsigned int local_gotno_;
+ // The maximum number of page GOT entries needed.
+ unsigned int page_gotno_;
+ // The number of global GOT entries.
+ unsigned int global_gotno_;
+ // The number of global GOT entries that are in the GGA_RELOC_ONLY area.
+ unsigned int reloc_only_gotno_;
+ // The number of TLS GOT entries.
+ unsigned int tls_gotno_;
+ // The offset of TLS LDM entry for this GOT.
+ unsigned int tls_ldm_offset_;
+ // All symbols that have global GOT entry.
+ Unordered_set<Mips_symbol<size>*> global_got_symbols_;
+ // A hash table holding GOT entries.
+ Got_entry_set got_entries_;
+ // A hash table of GOT page entries.
+ Got_page_entry_set got_page_entries_;
+ // The offset of first GOT page entry for this GOT.
+ unsigned int got_page_offset_start_;
+ // The offset of next available GOT page entry for this GOT.
+ unsigned int got_page_offset_next_;
+ // A hash table that maps GOT page entry value to the GOT offset where
+ // the entry is located.
+ Got_page_offsets got_page_offsets_;
+ // In multi-GOT links, a pointer to the next GOT.
+ Mips_got_info<size, big_endian>* next_;
+ // Index of this GOT in multi-GOT links.
+ unsigned int index_;
+ // The offset of this GOT in multi-GOT links.
+ unsigned int offset_;
+};
+
+// This is a helper class used during relocation scan. It records GOT16 addend.
+
+template<int size, bool big_endian>
+struct got16_addend
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+
+ got16_addend(const Sized_relobj_file<size, big_endian>* _object,
+ unsigned int _shndx, unsigned int _r_type, unsigned int _r_sym,
+ Mips_address _addend)
+ : object(_object), shndx(_shndx), r_type(_r_type), r_sym(_r_sym),
+ addend(_addend)
+ { }
+
+ const Sized_relobj_file<size, big_endian>* object;
+ unsigned int shndx;
+ unsigned int r_type;
+ unsigned int r_sym;
+ Mips_address addend;
+};
+
+// Mips_symbol class. Holds additional symbol information needed for Mips.
+
+template<int size>
+class Mips_symbol : public Sized_symbol<size>
+{
+ public:
+ Mips_symbol()
+ : need_fn_stub_(false), has_nonpic_branches_(false), la25_stub_offset_(-1U),
+ has_static_relocs_(false), no_lazy_stub_(false), lazy_stub_offset_(0),
+ pointer_equality_needed_(false), global_got_area_(GGA_NONE),
+ global_gotoffset_(-1U), got_only_for_calls_(true), has_lazy_stub_(false),
+ needs_mips_plt_(false), needs_comp_plt_(false), mips_plt_offset_(-1U),
+ comp_plt_offset_(-1U), mips16_fn_stub_(NULL), mips16_call_stub_(NULL),
+ mips16_call_fp_stub_(NULL), applied_secondary_got_fixup_(false)
+ { }
+
+ // Return whether this is a MIPS16 symbol.
+ bool
+ is_mips16() const
+ {
+ // (st_other & STO_MIPS16) == STO_MIPS16
+ return ((this->nonvis() & (elfcpp::STO_MIPS16 >> 2))
+ == elfcpp::STO_MIPS16 >> 2);
+ }
+
+ // Return whether this is a microMIPS symbol.
+ bool
+ is_micromips() const
+ {
+ // (st_other & STO_MIPS_ISA) == STO_MICROMIPS
+ return ((this->nonvis() & (elfcpp::STO_MIPS_ISA >> 2))
+ == elfcpp::STO_MICROMIPS >> 2);
+ }
+
+ // Return whether the symbol needs MIPS16 fn_stub.
+ bool
+ need_fn_stub() const
+ { return this->need_fn_stub_; }
+
+ // Set that the symbol needs MIPS16 fn_stub.
+ void
+ set_need_fn_stub()
+ { this->need_fn_stub_ = true; }
+
+ // Return whether this symbol is referenced by branch relocations from
+ // any non-PIC input file.
+ bool
+ has_nonpic_branches() const
+ { return this->has_nonpic_branches_; }
+
+ // Set that this symbol is referenced by branch relocations from
+ // any non-PIC input file.
+ void
+ set_has_nonpic_branches()
+ { this->has_nonpic_branches_ = true; }
+
+ // Return the offset of the la25 stub for this symbol from the start of the
+ // la25 stub section.
+ unsigned int
+ la25_stub_offset() const
+ { return this->la25_stub_offset_; }
+
+ // Set the offset of the la25 stub for this symbol from the start of the
+ // la25 stub section.
+ void
+ set_la25_stub_offset(unsigned int offset)
+ { this->la25_stub_offset_ = offset; }
+
+ // Return whether the symbol has la25 stub. This is true if this symbol is
+ // for a PIC function, and there are non-PIC branches and jumps to it.
+ bool
+ has_la25_stub() const
+ { return this->la25_stub_offset_ != -1U; }
+
+ // Return whether there is a relocation against this symbol that must be
+ // resolved by the static linker (that is, the relocation cannot possibly
+ // be made dynamic).
+ bool
+ has_static_relocs() const
+ { return this->has_static_relocs_; }
+
+ // Set that there is a relocation against this symbol that must be resolved
+ // by the static linker (that is, the relocation cannot possibly be made
+ // dynamic).
+ void
+ set_has_static_relocs()
+ { this->has_static_relocs_ = true; }
+
+ // Return whether we must not create a lazy-binding stub for this symbol.
+ bool
+ no_lazy_stub() const
+ { return this->no_lazy_stub_; }
+
+ // Set that we must not create a lazy-binding stub for this symbol.
+ void
+ set_no_lazy_stub()
+ { this->no_lazy_stub_ = true; }
+
+ // Return the offset of the lazy-binding stub for this symbol from the start
+ // of .MIPS.stubs section.
+ unsigned int
+ lazy_stub_offset() const
+ { return this->lazy_stub_offset_; }
+
+ // Set the offset of the lazy-binding stub for this symbol from the start
+ // of .MIPS.stubs section.
+ void
+ set_lazy_stub_offset(unsigned int offset)
+ { this->lazy_stub_offset_ = offset; }
+
+ // Return whether there are any relocations for this symbol where
+ // pointer equality matters.
+ bool
+ pointer_equality_needed() const
+ { return this->pointer_equality_needed_; }
+
+ // Set that there are relocations for this symbol where pointer equality
+ // matters.
+ void
+ set_pointer_equality_needed()
+ { this->pointer_equality_needed_ = true; }
+
+ // Return global GOT area where this symbol in located.
+ Global_got_area
+ global_got_area() const
+ { return this->global_got_area_; }
+
+ // Set global GOT area where this symbol in located.
+ void
+ set_global_got_area(Global_got_area global_got_area)
+ { this->global_got_area_ = global_got_area; }
+
+ // Return the global GOT offset for this symbol. For multi-GOT links, this
+ // returns the offset from the start of .got section to the first GOT entry
+ // for the symbol. Note that in multi-GOT links the symbol can have entry
+ // in more than one GOT.
+ unsigned int
+ global_gotoffset() const
+ { return this->global_gotoffset_; }
+
+ // Set the global GOT offset for this symbol. Note that in multi-GOT links
+ // the symbol can have entry in more than one GOT. This method will set
+ // the offset only if it is less than current offset.
+ void
+ set_global_gotoffset(unsigned int offset)
+ {
+ if (this->global_gotoffset_ == -1U || offset < this->global_gotoffset_)
+ this->global_gotoffset_ = offset;
+ }
+
+ // Return whether all GOT relocations for this symbol are for calls.
+ bool
+ got_only_for_calls() const
+ { return this->got_only_for_calls_; }
+
+ // Set that there is a GOT relocation for this symbol that is not for call.
+ void
+ set_got_not_only_for_calls()
+ { this->got_only_for_calls_ = false; }
+
+ // Return whether this is a PIC symbol.
+ bool
+ is_pic() const
+ {
+ // (st_other & STO_MIPS_FLAGS) == STO_MIPS_PIC
+ return ((this->nonvis() & (elfcpp::STO_MIPS_FLAGS >> 2))
+ == (elfcpp::STO_MIPS_PIC >> 2));
+ }
+
+ // Set the flag in st_other field that marks this symbol as PIC.
+ void
+ set_pic()
+ {
+ if (this->is_mips16())
+ // (st_other & ~(STO_MIPS16 | STO_MIPS_FLAGS)) | STO_MIPS_PIC
+ this->set_nonvis((this->nonvis()
+ & ~((elfcpp::STO_MIPS16 >> 2)
+ | (elfcpp::STO_MIPS_FLAGS >> 2)))
+ | (elfcpp::STO_MIPS_PIC >> 2));
+ else
+ // (other & ~STO_MIPS_FLAGS) | STO_MIPS_PIC
+ this->set_nonvis((this->nonvis() & ~(elfcpp::STO_MIPS_FLAGS >> 2))
+ | (elfcpp::STO_MIPS_PIC >> 2));
+ }
+
+ // Set the flag in st_other field that marks this symbol as PLT.
+ void
+ set_mips_plt()
+ {
+ if (this->is_mips16())
+ // (st_other & (STO_MIPS16 | ~STO_MIPS_FLAGS)) | STO_MIPS_PLT
+ this->set_nonvis((this->nonvis()
+ & ((elfcpp::STO_MIPS16 >> 2)
+ | ~(elfcpp::STO_MIPS_FLAGS >> 2)))
+ | (elfcpp::STO_MIPS_PLT >> 2));
+
+ else
+ // (st_other & ~STO_MIPS_FLAGS) | STO_MIPS_PLT
+ this->set_nonvis((this->nonvis() & ~(elfcpp::STO_MIPS_FLAGS >> 2))
+ | (elfcpp::STO_MIPS_PLT >> 2));
+ }
+
+ // Downcast a base pointer to a Mips_symbol pointer.
+ static Mips_symbol<size>*
+ as_mips_sym(Symbol* sym)
+ { return static_cast<Mips_symbol<size>*>(sym); }
+
+ // Downcast a base pointer to a Mips_symbol pointer.
+ static const Mips_symbol<size>*
+ as_mips_sym(const Symbol* sym)
+ { return static_cast<const Mips_symbol<size>*>(sym); }
+
+ // Return whether the symbol has lazy-binding stub.
+ bool
+ has_lazy_stub() const
+ { return this->has_lazy_stub_; }
+
+ // Set whether the symbol has lazy-binding stub.
+ void
+ set_has_lazy_stub(bool has_lazy_stub)
+ { this->has_lazy_stub_ = has_lazy_stub; }
+
+ // Return whether the symbol needs a standard PLT entry.
+ bool
+ needs_mips_plt() const
+ { return this->needs_mips_plt_; }
+
+ // Set whether the symbol needs a standard PLT entry.
+ void
+ set_needs_mips_plt(bool needs_mips_plt)
+ { this->needs_mips_plt_ = needs_mips_plt; }
+
+ // Return whether the symbol needs a compressed (MIPS16 or microMIPS) PLT
+ // entry.
+ bool
+ needs_comp_plt() const
+ { return this->needs_comp_plt_; }
+
+ // Set whether the symbol needs a compressed (MIPS16 or microMIPS) PLT entry.
+ void
+ set_needs_comp_plt(bool needs_comp_plt)
+ { this->needs_comp_plt_ = needs_comp_plt; }
+
+ // Return standard PLT entry offset, or -1 if none.
+ unsigned int
+ mips_plt_offset() const
+ { return this->mips_plt_offset_; }
+
+ // Set standard PLT entry offset.
+ void
+ set_mips_plt_offset(unsigned int mips_plt_offset)
+ { this->mips_plt_offset_ = mips_plt_offset; }
+
+ // Return whether the symbol has standard PLT entry.
+ bool
+ has_mips_plt_offset() const
+ { return this->mips_plt_offset_ != -1U; }
+
+ // Return compressed (MIPS16 or microMIPS) PLT entry offset, or -1 if none.
+ unsigned int
+ comp_plt_offset() const
+ { return this->comp_plt_offset_; }
+
+ // Set compressed (MIPS16 or microMIPS) PLT entry offset.
+ void
+ set_comp_plt_offset(unsigned int comp_plt_offset)
+ { this->comp_plt_offset_ = comp_plt_offset; }
+
+ // Return whether the symbol has compressed (MIPS16 or microMIPS) PLT entry.
+ bool
+ has_comp_plt_offset() const
+ { return this->comp_plt_offset_ != -1U; }
+
+ // Return MIPS16 fn stub for a symbol.
+ template<bool big_endian>
+ Mips16_stub_section<size, big_endian>*
+ get_mips16_fn_stub() const
+ {
+ return static_cast<Mips16_stub_section<size, big_endian>*>(mips16_fn_stub_);
+ }
+
+ // Set MIPS16 fn stub for a symbol.
+ void
+ set_mips16_fn_stub(Mips16_stub_section_base* stub)
+ { this->mips16_fn_stub_ = stub; }
+
+ // Return whether symbol has MIPS16 fn stub.
+ bool
+ has_mips16_fn_stub() const
+ { return this->mips16_fn_stub_ != NULL; }
+
+ // Return MIPS16 call stub for a symbol.
+ template<bool big_endian>
+ Mips16_stub_section<size, big_endian>*
+ get_mips16_call_stub() const
+ {
+ return static_cast<Mips16_stub_section<size, big_endian>*>(
+ mips16_call_stub_);
+ }
+
+ // Set MIPS16 call stub for a symbol.
+ void
+ set_mips16_call_stub(Mips16_stub_section_base* stub)
+ { this->mips16_call_stub_ = stub; }
+
+ // Return whether symbol has MIPS16 call stub.
+ bool
+ has_mips16_call_stub() const
+ { return this->mips16_call_stub_ != NULL; }
+
+ // Return MIPS16 call_fp stub for a symbol.
+ template<bool big_endian>
+ Mips16_stub_section<size, big_endian>*
+ get_mips16_call_fp_stub() const
+ {
+ return static_cast<Mips16_stub_section<size, big_endian>*>(
+ mips16_call_fp_stub_);
+ }
+
+ // Set MIPS16 call_fp stub for a symbol.
+ void
+ set_mips16_call_fp_stub(Mips16_stub_section_base* stub)
+ { this->mips16_call_fp_stub_ = stub; }
+
+ // Return whether symbol has MIPS16 call_fp stub.
+ bool
+ has_mips16_call_fp_stub() const
+ { return this->mips16_call_fp_stub_ != NULL; }
+
+ bool
+ get_applied_secondary_got_fixup() const
+ { return applied_secondary_got_fixup_; }
+
+ void
+ set_applied_secondary_got_fixup()
+ { this->applied_secondary_got_fixup_ = true; }
+
+ private:
+ // Whether the symbol needs MIPS16 fn_stub. This is true if this symbol
+ // appears in any relocs other than a 16 bit call.
+ bool need_fn_stub_;
+
+ // True if this symbol is referenced by branch relocations from
+ // any non-PIC input file. This is used to determine whether an
+ // la25 stub is required.
+ bool has_nonpic_branches_;
+
+ // The offset of the la25 stub for this symbol from the start of the
+ // la25 stub section.
+ unsigned int la25_stub_offset_;
+
+ // True if there is a relocation against this symbol that must be
+ // resolved by the static linker (that is, the relocation cannot
+ // possibly be made dynamic).
+ bool has_static_relocs_;
+
+ // Whether we must not create a lazy-binding stub for this symbol.
+ // This is true if the symbol has relocations related to taking the
+ // function's address.
+ bool no_lazy_stub_;
+
+ // The offset of the lazy-binding stub for this symbol from the start of
+ // .MIPS.stubs section.
+ unsigned int lazy_stub_offset_;
+
+ // True if there are any relocations for this symbol where pointer equality
+ // matters.
+ bool pointer_equality_needed_;
+
+ // Global GOT area where this symbol in located, or GGA_NONE if symbol is not
+ // in the global part of the GOT.
+ Global_got_area global_got_area_;
+
+ // The global GOT offset for this symbol. For multi-GOT links, this is offset
+ // from the start of .got section to the first GOT entry for the symbol.
+ // Note that in multi-GOT links the symbol can have entry in more than one GOT.
+ unsigned int global_gotoffset_;
+
+ // Whether all GOT relocations for this symbol are for calls.
+ bool got_only_for_calls_;
+ // Whether the symbol has lazy-binding stub.
+ bool has_lazy_stub_;
+ // Whether the symbol needs a standard PLT entry.
+ bool needs_mips_plt_;
+ // Whether the symbol needs a compressed (MIPS16 or microMIPS) PLT entry.
+ bool needs_comp_plt_;
+ // Standard PLT entry offset, or -1 if none.
+ unsigned int mips_plt_offset_;
+ // Compressed (MIPS16 or microMIPS) PLT entry offset, or -1 if none.
+ unsigned int comp_plt_offset_;
+ // MIPS16 fn stub for a symbol.
+ Mips16_stub_section_base* mips16_fn_stub_;
+ // MIPS16 call stub for a symbol.
+ Mips16_stub_section_base* mips16_call_stub_;
+ // MIPS16 call_fp stub for a symbol.
+ Mips16_stub_section_base* mips16_call_fp_stub_;
+
+ bool applied_secondary_got_fixup_;
+};
+
+// Mips16_stub_section class.
+
+// The mips16 compiler uses a couple of special sections to handle
+// floating point arguments.
+
+// Section names that look like .mips16.fn.FNNAME contain stubs that
+// copy floating point arguments from the fp regs to the gp regs and
+// then jump to FNNAME. If any 32 bit function calls FNNAME, the
+// call should be redirected to the stub instead. If no 32 bit
+// function calls FNNAME, the stub should be discarded. We need to
+// consider any reference to the function, not just a call, because
+// if the address of the function is taken we will need the stub,
+// since the address might be passed to a 32 bit function.
+
+// Section names that look like .mips16.call.FNNAME contain stubs
+// that copy floating point arguments from the gp regs to the fp
+// regs and then jump to FNNAME. If FNNAME is a 32 bit function,
+// then any 16 bit function that calls FNNAME should be redirected
+// to the stub instead. If FNNAME is not a 32 bit function, the
+// stub should be discarded.
+
+// .mips16.call.fp.FNNAME sections are similar, but contain stubs
+// which call FNNAME and then copy the return value from the fp regs
+// to the gp regs. These stubs store the return address in $18 while
+// calling FNNAME; any function which might call one of these stubs
+// must arrange to save $18 around the call. (This case is not
+// needed for 32 bit functions that call 16 bit functions, because
+// 16 bit functions always return floating point values in both
+// $f0/$f1 and $2/$3.)
+
+// Note that in all cases FNNAME might be defined statically.
+// Therefore, FNNAME is not used literally. Instead, the relocation
+// information will indicate which symbol the section is for.
+
+// We record any stubs that we find in the symbol table.
+
+// TODO(sasa): All mips16 stub sections should be emitted in the .text section.
+
+class Mips16_stub_section_base { };
+
+template<int size, bool big_endian>
+class Mips16_stub_section : public Mips16_stub_section_base
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+
+ public:
+ Mips16_stub_section(Mips_relobj<size, big_endian>* object, unsigned int shndx)
+ : object_(object), shndx_(shndx), r_sym_(0), gsym_(NULL),
+ found_r_mips_none_(false)
+ {
+ gold_assert(object->is_mips16_fn_stub_section(shndx)
+ || object->is_mips16_call_stub_section(shndx)
+ || object->is_mips16_call_fp_stub_section(shndx));
+ }
+
+ // Return the object of this stub section.
+ Mips_relobj<size, big_endian>*
+ object() const
+ { return this->object_; }
+
+ // Return the size of a section.
+ uint64_t
+ section_size() const
+ { return this->object_->section_size(this->shndx_); }
+
+ // Return section index of this stub section.
+ unsigned int
+ shndx() const
+ { return this->shndx_; }
+
+ // Return symbol index, if stub is for a local function.
+ unsigned int
+ r_sym() const
+ { return this->r_sym_; }
+
+ // Return symbol, if stub is for a global function.
+ Mips_symbol<size>*
+ gsym() const
+ { return this->gsym_; }
+
+ // Return whether stub is for a local function.
+ bool
+ is_for_local_function() const
+ { return this->gsym_ == NULL; }
+
+ // This method is called when a new relocation R_TYPE for local symbol R_SYM
+ // is found in the stub section. Try to find stub target.
+ void
+ new_local_reloc_found(unsigned int r_type, unsigned int r_sym)
+ {
+ // To find target symbol for this stub, trust the first R_MIPS_NONE
+ // relocation, if any. Otherwise trust the first relocation, whatever
+ // its kind.
+ if (this->found_r_mips_none_)
+ return;
+ if (r_type == elfcpp::R_MIPS_NONE)
+ {
+ this->r_sym_ = r_sym;
+ this->gsym_ = NULL;
+ this->found_r_mips_none_ = true;
+ }
+ else if (!is_target_found())
+ this->r_sym_ = r_sym;
+ }
+
+ // This method is called when a new relocation R_TYPE for global symbol GSYM
+ // is found in the stub section. Try to find stub target.
+ void
+ new_global_reloc_found(unsigned int r_type, Mips_symbol<size>* gsym)
+ {
+ // To find target symbol for this stub, trust the first R_MIPS_NONE
+ // relocation, if any. Otherwise trust the first relocation, whatever
+ // its kind.
+ if (this->found_r_mips_none_)
+ return;
+ if (r_type == elfcpp::R_MIPS_NONE)
+ {
+ this->gsym_ = gsym;
+ this->r_sym_ = 0;
+ this->found_r_mips_none_ = true;
+ }
+ else if (!is_target_found())
+ this->gsym_ = gsym;
+ }
+
+ // Return whether we found the stub target.
+ bool
+ is_target_found() const
+ { return this->r_sym_ != 0 || this->gsym_ != NULL; }
+
+ // Return whether this is a fn stub.
+ bool
+ is_fn_stub() const
+ { return this->object_->is_mips16_fn_stub_section(this->shndx_); }
+
+ // Return whether this is a call stub.
+ bool
+ is_call_stub() const
+ { return this->object_->is_mips16_call_stub_section(this->shndx_); }
+
+ // Return whether this is a call_fp stub.
+ bool
+ is_call_fp_stub() const
+ { return this->object_->is_mips16_call_fp_stub_section(this->shndx_); }
+
+ // Return the output address.
+ Mips_address
+ output_address() const
+ {
+ return (this->object_->output_section(this->shndx_)->address()
+ + this->object_->output_section_offset(this->shndx_));
+ }
+
+ private:
+ // The object of this stub section.
+ Mips_relobj<size, big_endian>* object_;
+ // The section index of this stub section.
+ unsigned int shndx_;
+ // The symbol index, if stub is for a local function.
+ unsigned int r_sym_;
+ // The symbol, if stub is for a global function.
+ Mips_symbol<size>* gsym_;
+ // True if we found R_MIPS_NONE relocation in this stub.
+ bool found_r_mips_none_;
+};
+
+// Mips_relobj class.
+
+template<int size, bool big_endian>
+class Mips_relobj : public Sized_relobj_file<size, big_endian>
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+ typedef std::map<unsigned int, Mips16_stub_section<size, big_endian>*>
+ Mips16_stubs_int_map;
+ typedef typename elfcpp::Swap<size, big_endian>::Valtype Valtype;
+
+ public:
+ Mips_relobj(const std::string& name, Input_file* input_file, off_t offset,
+ const typename elfcpp::Ehdr<size, big_endian>& ehdr)
+ : Sized_relobj_file<size, big_endian>(name, input_file, offset, ehdr),
+ processor_specific_flags_(0), local_symbol_is_mips16_(),
+ local_symbol_is_micromips_(), mips16_stub_sections_(),
+ local_non_16bit_calls_(), local_16bit_calls_(), local_mips16_fn_stubs_(),
+ local_mips16_call_stubs_(), gp_(0), got_info_(NULL),
+ section_is_mips16_fn_stub_(), section_is_mips16_call_stub_(),
+ section_is_mips16_call_fp_stub_(), pdr_shndx_(-1U), gprmask_(0),
+ cprmask1_(0), cprmask2_(0), cprmask3_(0), cprmask4_(0)
+ {
+ this->is_pic_ = (ehdr.get_e_flags() & elfcpp::EF_MIPS_PIC) != 0;
+ this->is_n32_ = elfcpp::abi_n32(ehdr.get_e_flags());
+ this->is_n64_ = elfcpp::abi_64(ehdr.get_e_ident()[elfcpp::EI_CLASS]);
+ }
+
+ ~Mips_relobj()
+ { }
+
+ // Downcast a base pointer to a Mips_relobj pointer. This is
+ // not type-safe but we only use Mips_relobj not the base class.
+ static Mips_relobj<size, big_endian>*
+ as_mips_relobj(Relobj* relobj)
+ { return static_cast<Mips_relobj<size, big_endian>*>(relobj); }
+
+ // Downcast a base pointer to a Mips_relobj pointer. This is
+ // not type-safe but we only use Mips_relobj not the base class.
+ static const Mips_relobj<size, big_endian>*
+ as_mips_relobj(const Relobj* relobj)
+ { return static_cast<const Mips_relobj<size, big_endian>*>(relobj); }
+
+ // Processor-specific flags in ELF file header. This is valid only after
+ // reading symbols.
+ elfcpp::Elf_Word
+ processor_specific_flags() const
+ { return this->processor_specific_flags_; }
+
+ // Whether a local symbol is MIPS16 symbol. R_SYM is the symbol table
+ // index. This is only valid after do_count_local_symbol is called.
+ bool
+ local_symbol_is_mips16(unsigned int r_sym) const
+ {
+ gold_assert(r_sym < this->local_symbol_is_mips16_.size());
+ return this->local_symbol_is_mips16_[r_sym];
+ }
+
+ // Whether a local symbol is microMIPS symbol. R_SYM is the symbol table
+ // index. This is only valid after do_count_local_symbol is called.
+ bool
+ local_symbol_is_micromips(unsigned int r_sym) const
+ {
+ gold_assert(r_sym < this->local_symbol_is_micromips_.size());
+ return this->local_symbol_is_micromips_[r_sym];
+ }
+
+ // Get or create MIPS16 stub section.
+ Mips16_stub_section<size, big_endian>*
+ get_mips16_stub_section(unsigned int shndx)
+ {
+ typename Mips16_stubs_int_map::const_iterator it =
+ this->mips16_stub_sections_.find(shndx);
+ if (it != this->mips16_stub_sections_.end())
+ return (*it).second;
+
+ Mips16_stub_section<size, big_endian>* stub_section =
+ new Mips16_stub_section<size, big_endian>(this, shndx);
+ this->mips16_stub_sections_.insert(
+ std::pair<unsigned int, Mips16_stub_section<size, big_endian>*>(
+ stub_section->shndx(), stub_section));
+ return stub_section;
+ }
+
+ // Return MIPS16 fn stub section for local symbol R_SYM, or NULL if this
+ // object doesn't have fn stub for R_SYM.
+ Mips16_stub_section<size, big_endian>*
+ get_local_mips16_fn_stub(unsigned int r_sym) const
+ {
+ typename Mips16_stubs_int_map::const_iterator it =
+ this->local_mips16_fn_stubs_.find(r_sym);
+ if (it != this->local_mips16_fn_stubs_.end())
+ return (*it).second;
+ return NULL;
+ }
+
+ // Record that this object has MIPS16 fn stub for local symbol. This method
+ // is only called if we decided not to discard the stub.
+ void
+ add_local_mips16_fn_stub(Mips16_stub_section<size, big_endian>* stub)
+ {
+ gold_assert(stub->is_for_local_function());
+ unsigned int r_sym = stub->r_sym();
+ this->local_mips16_fn_stubs_.insert(
+ std::pair<unsigned int, Mips16_stub_section<size, big_endian>*>(
+ r_sym, stub));
+ }
+
+ // Return MIPS16 call stub section for local symbol R_SYM, or NULL if this
+ // object doesn't have call stub for R_SYM.
+ Mips16_stub_section<size, big_endian>*
+ get_local_mips16_call_stub(unsigned int r_sym) const
+ {
+ typename Mips16_stubs_int_map::const_iterator it =
+ this->local_mips16_call_stubs_.find(r_sym);
+ if (it != this->local_mips16_call_stubs_.end())
+ return (*it).second;
+ return NULL;
+ }
+
+ // Record that this object has MIPS16 call stub for local symbol. This method
+ // is only called if we decided not to discard the stub.
+ void
+ add_local_mips16_call_stub(Mips16_stub_section<size, big_endian>* stub)
+ {
+ gold_assert(stub->is_for_local_function());
+ unsigned int r_sym = stub->r_sym();
+ this->local_mips16_call_stubs_.insert(
+ std::pair<unsigned int, Mips16_stub_section<size, big_endian>*>(
+ r_sym, stub));
+ }
+
+ // Record that we found "non 16-bit" call relocation against local symbol
+ // SYMNDX. This reloc would need to refer to a MIPS16 fn stub, if there
+ // is one.
+ void
+ add_local_non_16bit_call(unsigned int symndx)
+ { this->local_non_16bit_calls_.insert(symndx); }
+
+ // Return true if there is any "non 16-bit" call relocation against local
+ // symbol SYMNDX in this object.
+ bool
+ has_local_non_16bit_call_relocs(unsigned int symndx)
+ {
+ return (this->local_non_16bit_calls_.find(symndx)
+ != this->local_non_16bit_calls_.end());
+ }
+
+ // Record that we found 16-bit call relocation R_MIPS16_26 against local
+ // symbol SYMNDX. Local MIPS16 call or call_fp stubs will only be needed
+ // if there is some R_MIPS16_26 relocation that refers to the stub symbol.
+ void
+ add_local_16bit_call(unsigned int symndx)
+ { this->local_16bit_calls_.insert(symndx); }
+
+ // Return true if there is any 16-bit call relocation R_MIPS16_26 against local
+ // symbol SYMNDX in this object.
+ bool
+ has_local_16bit_call_relocs(unsigned int symndx)
+ {
+ return (this->local_16bit_calls_.find(symndx)
+ != this->local_16bit_calls_.end());
+ }
+
+ // Get gp value that was used to create this object.
+ Mips_address
+ gp_value() const
+ { return this->gp_; }
+
+ // Return whether the object is a PIC object.
+ bool
+ is_pic() const
+ { return this->is_pic_; }
+
+ // Return whether the object uses N32 ABI.
+ bool
+ is_n32() const
+ { return this->is_n32_; }
+
+ // Return whether the object uses N64 ABI.
+ bool
+ is_n64() const
+ { return this->is_n64_; }
+
+ // Return whether the object uses NewABI conventions.
+ bool
+ is_newabi() const
+ { return this->is_n32_ || this->is_n64_; }
+
+ // Return Mips_got_info for this object.
+ Mips_got_info<size, big_endian>*
+ get_got_info() const
+ { return this->got_info_; }
+
+ // Return Mips_got_info for this object. Create new info if it doesn't exist.
+ Mips_got_info<size, big_endian>*
+ get_or_create_got_info()
+ {
+ if (!this->got_info_)
+ this->got_info_ = new Mips_got_info<size, big_endian>();
+ return this->got_info_;
+ }
+
+ // Set Mips_got_info for this object.
+ void
+ set_got_info(Mips_got_info<size, big_endian>* got_info)
+ { this->got_info_ = got_info; }
+
+ // Whether a section SHDNX is a MIPS16 stub section. This is only valid
+ // after do_read_symbols is called.
+ bool
+ is_mips16_stub_section(unsigned int shndx)
+ {
+ return (is_mips16_fn_stub_section(shndx)
+ || is_mips16_call_stub_section(shndx)
+ || is_mips16_call_fp_stub_section(shndx));
+ }
+
+ // Return TRUE if relocations in section SHNDX can refer directly to a
+ // MIPS16 function rather than to a hard-float stub. This is only valid
+ // after do_read_symbols is called.
+ bool
+ section_allows_mips16_refs(unsigned int shndx)
+ {
+ return (this->is_mips16_stub_section(shndx) || shndx == this->pdr_shndx_);
+ }
+
+ // Whether a section SHDNX is a MIPS16 fn stub section. This is only valid
+ // after do_read_symbols is called.
+ bool
+ is_mips16_fn_stub_section(unsigned int shndx)
+ {
+ gold_assert(shndx < this->section_is_mips16_fn_stub_.size());
+ return this->section_is_mips16_fn_stub_[shndx];
+ }
+
+ // Whether a section SHDNX is a MIPS16 call stub section. This is only valid
+ // after do_read_symbols is called.
+ bool
+ is_mips16_call_stub_section(unsigned int shndx)
+ {
+ gold_assert(shndx < this->section_is_mips16_call_stub_.size());
+ return this->section_is_mips16_call_stub_[shndx];
+ }
+
+ // Whether a section SHDNX is a MIPS16 call_fp stub section. This is only
+ // valid after do_read_symbols is called.
+ bool
+ is_mips16_call_fp_stub_section(unsigned int shndx)
+ {
+ gold_assert(shndx < this->section_is_mips16_call_fp_stub_.size());
+ return this->section_is_mips16_call_fp_stub_[shndx];
+ }
+
+ // Discard MIPS16 stub secions that are not needed.
+ void
+ discard_mips16_stub_sections(Symbol_table* symtab);
+
+ // Return gprmask from the .reginfo section of this object.
+ Valtype
+ gprmask() const
+ { return this->gprmask_; }
+
+ // Return cprmask1 from the .reginfo section of this object.
+ Valtype
+ cprmask1() const
+ { return this->cprmask1_; }
+
+ // Return cprmask2 from the .reginfo section of this object.
+ Valtype
+ cprmask2() const
+ { return this->cprmask2_; }
+
+ // Return cprmask3 from the .reginfo section of this object.
+ Valtype
+ cprmask3() const
+ { return this->cprmask3_; }
+
+ // Return cprmask4 from the .reginfo section of this object.
+ Valtype
+ cprmask4() const
+ { return this->cprmask4_; }
+
+ protected:
+ // Count the local symbols.
+ void
+ do_count_local_symbols(Stringpool_template<char>*,
+ Stringpool_template<char>*);
+
+ // Read the symbol information.
+ void
+ do_read_symbols(Read_symbols_data* sd);
+
+ private:
+ // processor-specific flags in ELF file header.
+ elfcpp::Elf_Word processor_specific_flags_;
+
+ // Bit vector to tell if a local symbol is a MIPS16 symbol or not.
+ // This is only valid after do_count_local_symbol is called.
+ std::vector<bool> local_symbol_is_mips16_;
+
+ // Bit vector to tell if a local symbol is a microMIPS symbol or not.
+ // This is only valid after do_count_local_symbol is called.
+ std::vector<bool> local_symbol_is_micromips_;
+
+ // Map from section index to the MIPS16 stub for that section. This contains
+ // all stubs found in this object.
+ Mips16_stubs_int_map mips16_stub_sections_;
+
+ // Local symbols that have "non 16-bit" call relocation. This relocation
+ // would need to refer to a MIPS16 fn stub, if there is one.
+ std::set<unsigned int> local_non_16bit_calls_;
+
+ // Local symbols that have 16-bit call relocation R_MIPS16_26. Local MIPS16
+ // call or call_fp stubs will only be needed if there is some R_MIPS16_26
+ // relocation that refers to the stub symbol.
+ std::set<unsigned int> local_16bit_calls_;
+
+ // Map from local symbol index to the MIPS16 fn stub for that symbol.
+ // This contains only the stubs that we decided not to discard.
+ Mips16_stubs_int_map local_mips16_fn_stubs_;
+
+ // Map from local symbol index to the MIPS16 call stub for that symbol.
+ // This contains only the stubs that we decided not to discard.
+ Mips16_stubs_int_map local_mips16_call_stubs_;
+
+ // gp value that was used to create this object.
+ Mips_address gp_;
+ // Whether the object is a PIC object.
+ bool is_pic_ : 1;
+ // Whether the object uses N32 ABI.
+ bool is_n32_ : 1;
+ // Whether the object uses N64 ABI.
+ bool is_n64_ : 1;
+ // The Mips_got_info for this object.
+ Mips_got_info<size, big_endian>* got_info_;
+
+ // Bit vector to tell if a section is a MIPS16 fn stub section or not.
+ // This is only valid after do_read_symbols is called.
+ std::vector<bool> section_is_mips16_fn_stub_;
+
+ // Bit vector to tell if a section is a MIPS16 call stub section or not.
+ // This is only valid after do_read_symbols is called.
+ std::vector<bool> section_is_mips16_call_stub_;
+
+ // Bit vector to tell if a section is a MIPS16 call_fp stub section or not.
+ // This is only valid after do_read_symbols is called.
+ std::vector<bool> section_is_mips16_call_fp_stub_;
+
+ // .pdr section index.
+ unsigned int pdr_shndx_;
+
+ // gprmask from the .reginfo section of this object.
+ Valtype gprmask_;
+ // cprmask1 from the .reginfo section of this object.
+ Valtype cprmask1_;
+ // cprmask2 from the .reginfo section of this object.
+ Valtype cprmask2_;
+ // cprmask3 from the .reginfo section of this object.
+ Valtype cprmask3_;
+ // cprmask4 from the .reginfo section of this object.
+ Valtype cprmask4_;
+};
+
+// Mips_output_data_got class.
+
+template<int size, bool big_endian>
+class Mips_output_data_got : public Output_data_got<size, big_endian>
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+ typedef Output_data_reloc<elfcpp::SHT_REL, true, size, big_endian>
+ Reloc_section;
+ typedef typename elfcpp::Swap<size, big_endian>::Valtype Valtype;
+
+ public:
+ Mips_output_data_got(Target_mips<size, big_endian>* target,
+ Symbol_table* symtab, Layout* layout)
+ : Output_data_got<size, big_endian>(), target_(target),
+ symbol_table_(symtab), layout_(layout), static_relocs_(), got_view_(NULL),
+ first_global_got_dynsym_index_(-1U), primary_got_(NULL),
+ secondary_got_relocs_()
+ {
+ this->master_got_info_ = new Mips_got_info<size, big_endian>();
+ this->set_addralign(16);
+ }
+
+ // Reserve GOT entry for a GOT relocation of type R_TYPE against symbol
+ // SYMNDX + ADDEND, where SYMNDX is a local symbol in section SHNDX in OBJECT.
+ void
+ record_local_got_symbol(Mips_relobj<size, big_endian>* object,
+ unsigned int symndx, Mips_address addend,
+ unsigned int r_type, unsigned int shndx)
+ {
+ this->master_got_info_->record_local_got_symbol(object, symndx, addend,
+ r_type, shndx);
+ }
+
+ // Reserve GOT entry for a GOT relocation of type R_TYPE against MIPS_SYM,
+ // in OBJECT. FOR_CALL is true if the caller is only interested in
+ // using the GOT entry for calls. DYN_RELOC is true if R_TYPE is a dynamic
+ // relocation.
+ void
+ record_global_got_symbol(Mips_symbol<size>* mips_sym,
+ Mips_relobj<size, big_endian>* object,
+ unsigned int r_type, bool dyn_reloc, bool for_call)
+ {
+ this->master_got_info_->record_global_got_symbol(mips_sym, object, r_type,
+ dyn_reloc, for_call);
+ }
+
+ // Record that OBJECT has a page relocation against symbol SYMNDX and
+ // that ADDEND is the addend for that relocation.
+ void
+ record_got_page_entry(Mips_relobj<size, big_endian>* object,
+ unsigned int symndx, int addend)
+ { this->master_got_info_->record_got_page_entry(object, symndx, addend); }
+
+ // Add a static entry for the GOT entry at OFFSET. GSYM is a global
+ // symbol and R_TYPE is the code of a dynamic relocation that needs to be
+ // applied in a static link.
+ void
+ add_static_reloc(unsigned int got_offset, unsigned int r_type,
+ Mips_symbol<size>* gsym)
+ { this->static_relocs_.push_back(Static_reloc(got_offset, r_type, gsym)); }
+
+ // Add a static reloc for the GOT entry at OFFSET. RELOBJ is an object
+ // defining a local symbol with INDEX. R_TYPE is the code of a dynamic
+ // relocation that needs to be applied in a static link.
+ void
+ add_static_reloc(unsigned int got_offset, unsigned int r_type,
+ Sized_relobj_file<size, big_endian>* relobj,
+ unsigned int index)
+ {
+ this->static_relocs_.push_back(Static_reloc(got_offset, r_type, relobj,
+ index));
+ }
+
+ // Record that global symbol GSYM has R_TYPE dynamic relocation in the
+ // secondary GOT at OFFSET.
+ void
+ add_secondary_got_reloc(unsigned int got_offset, unsigned int r_type,
+ Mips_symbol<size>* gsym)
+ {
+ this->secondary_got_relocs_.push_back(Static_reloc(got_offset,
+ r_type, gsym));
+ }
+
+ // Update GOT entry at OFFSET with VALUE.
+ void
+ update_got_entry(unsigned int offset, Mips_address value)
+ {
+ elfcpp::Swap<size, big_endian>::writeval(this->got_view_ + offset, value);
+ }
+
+ // Return the number of entries in local part of the GOT. This includes
+ // local entries, page entries and 2 reserved entries.
+ unsigned int
+ get_local_gotno() const
+ {
+ if (!this->multi_got())
+ {
+ return (2 + this->master_got_info_->local_gotno()
+ + this->master_got_info_->page_gotno());
+ }
+ else
+ return 2 + this->primary_got_->local_gotno() + this->primary_got_->page_gotno();
+ }
+
+ // Return dynamic symbol table index of the first symbol with global GOT
+ // entry.
+ unsigned int
+ first_global_got_dynsym_index() const
+ { return this->first_global_got_dynsym_index_; }
+
+ // Set dynamic symbol table index of the first symbol with global GOT entry.
+ void
+ set_first_global_got_dynsym_index(unsigned int index)
+ { this->first_global_got_dynsym_index_ = index; }
+
+ // Lay out the GOT. Add local, global and TLS entries. If GOT is
+ // larger than 64K, create multi-GOT.
+ void
+ lay_out_got(Layout* layout, Symbol_table* symtab,
+ const Input_objects* input_objects);
+
+ // Create multi-GOT. For every GOT, add local, global and TLS entries.
+ void
+ lay_out_multi_got(Layout* layout, const Input_objects* input_objects);
+
+ // Attempt to merge GOTs of different input objects.
+ void
+ merge_gots(const Input_objects* input_objects);
+
+ // Consider merging FROM, which is OBJECT's GOT, into TO. Return false if
+ // this would lead to overflow, true if they were merged successfully.
+ bool
+ merge_got_with(Mips_got_info<size, big_endian>* from,
+ Mips_relobj<size, big_endian>* object,
+ Mips_got_info<size, big_endian>* to);
+
+ // Return the offset of GOT page entry for VALUE. For multi-GOT links,
+ // use OBJECT's GOT.
+ unsigned int
+ get_got_page_offset(Mips_address value,
+ const Mips_relobj<size, big_endian>* object)
+ {
+ Mips_got_info<size, big_endian>* g = (!this->multi_got()
+ ? this->master_got_info_
+ : object->get_got_info());
+ gold_assert(g != NULL);
+ return g->get_got_page_offset(value, this);
+ }
+
+ // Return the GOT offset of type GOT_TYPE of the global symbol
+ // GSYM. For multi-GOT links, use OBJECT's GOT.
+ unsigned int got_offset(const Symbol* gsym, unsigned int got_type,
+ Mips_relobj<size, big_endian>* object) const
+ {
+ if (!this->multi_got())
+ return gsym->got_offset(got_type);
+ else
+ {
+ Mips_got_info<size, big_endian>* g = object->get_got_info();
+ gold_assert(g != NULL);
+ return gsym->got_offset(g->multigot_got_type(got_type));
+ }
+ }
+
+ // Return the GOT offset of type GOT_TYPE of the local symbol
+ // SYMNDX.
+ unsigned int
+ got_offset(unsigned int symndx, unsigned int got_type,
+ Sized_relobj_file<size, big_endian>* object) const
+ { return object->local_got_offset(symndx, got_type); }
+
+ // Return the offset of TLS LDM entry. For multi-GOT links, use OBJECT's GOT.
+ unsigned int
+ tls_ldm_offset(Mips_relobj<size, big_endian>* object) const
+ {
+ Mips_got_info<size, big_endian>* g = (!this->multi_got()
+ ? this->master_got_info_
+ : object->get_got_info());
+ gold_assert(g != NULL);
+ return g->tls_ldm_offset();
+ }
+
+ // Set the offset of TLS LDM entry. For multi-GOT links, use OBJECT's GOT.
+ void
+ set_tls_ldm_offset(unsigned int tls_ldm_offset,
+ Mips_relobj<size, big_endian>* object)
+ {
+ Mips_got_info<size, big_endian>* g = (!this->multi_got()
+ ? this->master_got_info_
+ : object->get_got_info());
+ gold_assert(g != NULL);
+ g->set_tls_ldm_offset(tls_ldm_offset);
+ }
+
+ // Return true for multi-GOT links.
+ bool
+ multi_got() const
+ { return this->primary_got_ != NULL; }
+
+ // Return the offset of OBJECT's GOT from the start of .got section.
+ unsigned int
+ get_got_offset(const Mips_relobj<size, big_endian>* object)
+ {
+ if (!this->multi_got())
+ return 0;
+ else
+ {
+ Mips_got_info<size, big_endian>* g = object->get_got_info();
+ return g != NULL ? g->offset() : 0;
+ }
+ }
+
+ // Create global GOT entries that should be in the GGA_RELOC_ONLY area.
+ void
+ add_reloc_only_entries()
+ { this->master_got_info_->add_reloc_only_entries(this); }
+
+ // Return offset of the primary GOT's entry for global symbol.
+ unsigned int
+ get_primary_got_offset(const Mips_symbol<size>* sym) const
+ {
+ gold_assert(sym->global_got_area() != GGA_NONE);
+ return (this->get_local_gotno() + sym->dynsym_index()
+ - this->first_global_got_dynsym_index()) * size/8;
+ }
+
+ // For the entry at offset GOT_OFFSET, return its offset from the gp.
+ // Input argument GOT_OFFSET is always global offset from the start of
+ // .got section, for both single and multi-GOT links.
+ // For single GOT links, this returns GOT_OFFSET - 0x7FF0. For multi-GOT
+ // links, the return value is object_got_offset - 0x7FF0, where
+ // object_got_offset is offset in the OBJECT's GOT.
+ int
+ gp_offset(unsigned int got_offset,
+ const Mips_relobj<size, big_endian>* object) const
+ {
+ return (this->address() + got_offset
+ - this->target_->adjusted_gp_value(object));
+ }
+
+ protected:
+ // Write out the GOT table.
+ void
+ do_write(Output_file*);
+
+ private:
+
+ // This class represent dynamic relocations that need to be applied by
+ // gold because we are using TLS relocations in a static link.
+ class Static_reloc
+ {
+ public:
+ Static_reloc(unsigned int got_offset, unsigned int r_type,
+ Mips_symbol<size>* gsym)
+ : got_offset_(got_offset), r_type_(r_type), symbol_is_global_(true)
+ { this->u_.global.symbol = gsym; }
+
+ Static_reloc(unsigned int got_offset, unsigned int r_type,
+ Sized_relobj_file<size, big_endian>* relobj, unsigned int index)
+ : got_offset_(got_offset), r_type_(r_type), symbol_is_global_(false)
+ {
+ this->u_.local.relobj = relobj;
+ this->u_.local.index = index;
+ }
+
+ // Return the GOT offset.
+ unsigned int
+ got_offset() const
+ { return this->got_offset_; }
+
+ // Relocation type.
+ unsigned int
+ r_type() const
+ { return this->r_type_; }
+
+ // Whether the symbol is global or not.
+ bool
+ symbol_is_global() const
+ { return this->symbol_is_global_; }
+
+ // For a relocation against a global symbol, the global symbol.
+ Mips_symbol<size>*
+ symbol() const
+ {
+ gold_assert(this->symbol_is_global_);
+ return this->u_.global.symbol;
+ }
+
+ // For a relocation against a local symbol, the defining object.
+ Sized_relobj_file<size, big_endian>*
+ relobj() const
+ {
+ gold_assert(!this->symbol_is_global_);
+ return this->u_.local.relobj;
+ }
+
+ // For a relocation against a local symbol, the local symbol index.
+ unsigned int
+ index() const
+ {
+ gold_assert(!this->symbol_is_global_);
+ return this->u_.local.index;
+ }
+
+ private:
+ // GOT offset of the entry to which this relocation is applied.
+ unsigned int got_offset_;
+ // Type of relocation.
+ unsigned int r_type_;
+ // Whether this relocation is against a global symbol.
+ bool symbol_is_global_;
+ // A global or local symbol.
+ union
+ {
+ struct
+ {
+ // For a global symbol, the symbol itself.
+ Mips_symbol<size>* symbol;
+ } global;
+ struct
+ {
+ // For a local symbol, the object defining object.
+ Sized_relobj_file<size, big_endian>* relobj;
+ // For a local symbol, the symbol index.
+ unsigned int index;
+ } local;
+ } u_;
+ };
+
+ // The target.
+ Target_mips<size, big_endian>* target_;
+ // The symbol table.
+ Symbol_table* symbol_table_;
+ // The layout.
+ Layout* layout_;
+ // Static relocs to be applied to the GOT.
+ std::vector<Static_reloc> static_relocs_;
+ // .got section view.
+ unsigned char* got_view_;
+ // The dynamic symbol table index of the first symbol with global GOT entry.
+ unsigned int first_global_got_dynsym_index_;
+ // The master GOT information.
+ Mips_got_info<size, big_endian>* master_got_info_;
+ // The primary GOT information.
+ Mips_got_info<size, big_endian>* primary_got_;
+ // Secondary GOT fixups.
+ std::vector<Static_reloc> secondary_got_relocs_;
+};
+
+// A class to handle LA25 stubs - non-PIC interface to a PIC function. There are
+// two ways of creating these interfaces. The first is to add:
+//
+// lui $25,%hi(func)
+// j func
+// addiu $25,$25,%lo(func)
+//
+// to a separate trampoline section. The second is to add:
+//
+// lui $25,%hi(func)
+// addiu $25,$25,%lo(func)
+//
+// immediately before a PIC function "func", but only if a function is at the
+// beginning of the section, and the section is not too heavily aligned (i.e we
+// would need to add no more than 2 nops before the stub.)
+//
+// We only create stubs of the first type.
+
+template<int size, bool big_endian>
+class Mips_output_data_la25_stub : public Output_section_data
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+
+ public:
+ Mips_output_data_la25_stub()
+ : Output_section_data(size == 32 ? 4 : 8), symbols_()
+ { }
+
+ // Create LA25 stub for a symbol.
+ void
+ create_la25_stub(Symbol_table* symtab, Target_mips<size, big_endian>* target,
+ Mips_symbol<size>* gsym);
+
+ // Return output address of a stub.
+ Mips_address
+ stub_address(const Mips_symbol<size>* sym) const
+ {
+ gold_assert(sym->has_la25_stub());
+ return this->address() + sym->la25_stub_offset();
+ }
+
+ protected:
+ void
+ do_adjust_output_section(Output_section* os)
+ { os->set_entsize(0); }
+
+ private:
+ // Template for standard LA25 stub.
+ static const uint32_t la25_stub_entry[];
+ // Template for microMIPS LA25 stub.
+ static const uint32_t la25_stub_micromips_entry[];
+
+ // Set the final size.
+ void
+ set_final_data_size()
+ { this->set_data_size(this->symbols_.size() * 16); }
+
+ // Create a symbol for SYM stub's value and size, to help make the
+ // disassembly easier to read.
+ void
+ create_stub_symbol(Mips_symbol<size>* sym, Symbol_table* symtab,
+ Target_mips<size, big_endian>* target, uint64_t symsize);
+
+ // Write out the LA25 stub section.
+ void
+ do_write(Output_file*);
+
+ // Symbols that have LA25 stubs.
+ Unordered_set<Mips_symbol<size>*> symbols_;
+};
+
+// A class to handle the PLT data.
+
+template<int size, bool big_endian>
+class Mips_output_data_plt : public Output_section_data
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+ typedef Output_data_reloc<elfcpp::SHT_REL, true,
+ size, big_endian> Reloc_section;
+
+ public:
+ // Create the PLT section. The ordinary .got section is an argument,
+ // since we need to refer to the start.
+ Mips_output_data_plt(Layout* layout, Output_data_space* got_plt,
+ Target_mips<size, big_endian>* target)
+ : Output_section_data(size == 32 ? 4 : 8), got_plt_(got_plt), symbols_(),
+ plt_mips_offset_(0), plt_comp_offset_(0), plt_header_size_(0),
+ target_(target)
+ {
+ this->rel_ = new Reloc_section(false);
+ layout->add_output_section_data(".rel.plt", elfcpp::SHT_REL,
+ elfcpp::SHF_ALLOC, this->rel_,
+ ORDER_DYNAMIC_PLT_RELOCS, false);
+ }
+
+ // Add an entry to the PLT for a symbol referenced by r_type relocation.
+ void
+ add_entry(Mips_symbol<size>* gsym, unsigned int r_type);
+
+ // Return the .rel.plt section data.
+ const Reloc_section*
+ rel_plt() const
+ { return this->rel_; }
+
+ // Return the number of PLT entries.
+ unsigned int
+ entry_count() const
+ { return this->symbols_.size(); }
+
+ // Return the offset of the first non-reserved PLT entry.
+ unsigned int
+ first_plt_entry_offset() const
+ { return sizeof(plt0_entry_o32); }
+
+ // Return the size of a PLT entry.
+ unsigned int
+ plt_entry_size() const
+ { return sizeof(plt_entry); }
+
+ // Set final PLT offsets. For each symbol, determine whether standard or
+ // compressed (MIPS16 or microMIPS) PLT entry is used.
+ void
+ set_plt_offsets();
+
+ // Return the offset of the first standard PLT entry.
+ unsigned int
+ first_mips_plt_offset() const
+ { return this->plt_header_size_; }
+
+ // Return the offset of the first compressed PLT entry.
+ unsigned int
+ first_comp_plt_offset() const
+ { return this->plt_header_size_ + this->plt_mips_offset_; }
+
+ // Return whether there are any standard PLT entries.
+ bool
+ has_standard_entries() const
+ { return this->plt_mips_offset_ > 0; }
+
+ // Return the output address of standard PLT entry.
+ Mips_address
+ mips_entry_address(const Mips_symbol<size>* sym) const
+ {
+ gold_assert (sym->has_mips_plt_offset());
+ return (this->address() + this->first_mips_plt_offset()
+ + sym->mips_plt_offset());
+ }
+
+ // Return the output address of compressed (MIPS16 or microMIPS) PLT entry.
+ Mips_address
+ comp_entry_address(const Mips_symbol<size>* sym) const
+ {
+ gold_assert (sym->has_comp_plt_offset());
+ return (this->address() + this->first_comp_plt_offset()
+ + sym->comp_plt_offset());
+ }
+
+ protected:
+ void
+ do_adjust_output_section(Output_section* os)
+ { os->set_entsize(0); }
+
+ // Write to a map file.
+ void
+ do_print_to_mapfile(Mapfile* mapfile) const
+ { mapfile->print_output_data(this, _(".plt")); }
+
+ private:
+ // Template for the first PLT entry.
+ static const uint32_t plt0_entry_o32[];
+ static const uint32_t plt0_entry_n32[];
+ static const uint32_t plt0_entry_n64[];
+ static const uint32_t plt0_entry_micromips_o32[];
+ static const uint32_t plt0_entry_micromips32_o32[];
+
+ // Template for subsequent PLT entries.
+ static const uint32_t plt_entry[];
+ static const uint32_t plt_entry_mips16_o32[];
+ static const uint32_t plt_entry_micromips_o32[];
+ static const uint32_t plt_entry_micromips32_o32[];
+
+ // Set the final size.
+ void
+ set_final_data_size()
+ {
+ this->set_data_size(this->plt_header_size_ + this->plt_mips_offset_
+ + this->plt_comp_offset_);
+ }
+
+ // Write out the PLT data.
+ void
+ do_write(Output_file*);
+
+ // Return whether the plt header contains microMIPS code. For the sake of
+ // cache alignment always use a standard header whenever any standard entries
+ // are present even if microMIPS entries are present as well. This also lets
+ // the microMIPS header rely on the value of $v0 only set by microMIPS
+ // entries, for a small size reduction.
+ bool
+ is_plt_header_compressed() const
+ {
+ gold_assert(this->plt_mips_offset_ + this->plt_comp_offset_ != 0);
+ return this->target_->is_output_micromips() && this->plt_mips_offset_ == 0;
+ }
+
+ // Return the size of the PLT header.
+ unsigned int
+ get_plt_header_size() const
+ {
+ if (this->target_->is_output_n64())
+ return 4 * sizeof(plt0_entry_n64) / sizeof(plt0_entry_n64[0]);
+ else if (this->target_->is_output_n32())
+ return 4 * sizeof(plt0_entry_n32) / sizeof(plt0_entry_n32[0]);
+ else if (!this->is_plt_header_compressed())
+ return 4 * sizeof(plt0_entry_o32) / sizeof(plt0_entry_o32[0]);
+ else if (this->target_->use_32bit_micromips_instructions())
+ return (2 * sizeof(plt0_entry_micromips32_o32)
+ / sizeof(plt0_entry_micromips32_o32[0]));
+ else
+ return (2 * sizeof(plt0_entry_micromips_o32)
+ / sizeof(plt0_entry_micromips_o32[0]));
+ }
+
+ // Return the PLT header entry.
+ const uint32_t*
+ get_plt_header_entry() const
+ {
+ if (this->target_->is_output_n64())
+ return plt0_entry_n64;
+ else if (this->target_->is_output_n32())
+ return plt0_entry_n32;
+ else if (!this->is_plt_header_compressed())
+ return plt0_entry_o32;
+ else if (this->target_->use_32bit_micromips_instructions())
+ return plt0_entry_micromips32_o32;
+ else
+ return plt0_entry_micromips_o32;
+ }
+
+ // Return the size of the standard PLT entry.
+ unsigned int
+ standard_plt_entry_size() const
+ { return 4 * sizeof(plt_entry) / sizeof(plt_entry[0]); }
+
+ // Return the size of the compressed PLT entry.
+ unsigned int
+ compressed_plt_entry_size() const
+ {
+ gold_assert(!this->target_->is_output_newabi());
+
+ if (!this->target_->is_output_micromips())
+ return (2 * sizeof(plt_entry_mips16_o32)
+ / sizeof(plt_entry_mips16_o32[0]));
+ else if (this->target_->use_32bit_micromips_instructions())
+ return (2 * sizeof(plt_entry_micromips32_o32)
+ / sizeof(plt_entry_micromips32_o32[0]));
+ else
+ return (2 * sizeof(plt_entry_micromips_o32)
+ / sizeof(plt_entry_micromips_o32[0]));
+ }
+
+ // The reloc section.
+ Reloc_section* rel_;
+ // The .got.plt section.
+ Output_data_space* got_plt_;
+ // Symbols that have PLT entry.
+ std::vector<Mips_symbol<size>*> symbols_;
+ // The offset of the next standard PLT entry to create.
+ unsigned int plt_mips_offset_;
+ // The offset of the next compressed PLT entry to create.
+ unsigned int plt_comp_offset_;
+ // The size of the PLT header in bytes.
+ unsigned int plt_header_size_;
+ // The target.
+ Target_mips<size, big_endian>* target_;
+};
+
+// A class to handle the .MIPS.stubs data.
+
+template<int size, bool big_endian>
+class Mips_output_data_mips_stubs : public Output_section_data
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+
+ public:
+ Mips_output_data_mips_stubs(Target_mips<size, big_endian>* target)
+ : Output_section_data(size == 32 ? 4 : 8), symbols_(), dynsym_count_(-1U),
+ stub_offsets_are_set_(false), target_(target)
+ { }
+
+ // Create entry for a symbol.
+ void
+ make_entry(Mips_symbol<size>*);
+
+ // Remove entry for a symbol.
+ void
+ remove_entry(Mips_symbol<size>* gsym);
+
+ // Set stub offsets for symbols. This method expects that the number of
+ // entries in dynamic symbol table is set.
+ void
+ set_lazy_stub_offsets();
+
+ void
+ set_needs_dynsym_value();
+
+ // Set the number of entries in dynamic symbol table.
+ void
+ set_dynsym_count(unsigned int dynsym_count)
+ { this->dynsym_count_ = dynsym_count; }
+
+ // Return maximum size of the stub, ie. the stub size if the dynamic symbol
+ // count is greater than 0x10000. If the dynamic symbol count is less than
+ // 0x10000, the stub will be 4 bytes smaller.
+ // There's no disadvantage from using microMIPS code here, so for the sake of
+ // pure-microMIPS binaries we prefer it whenever there's any microMIPS code in
+ // output produced at all. This has a benefit of stubs being shorter by
+ // 4 bytes each too, unless in the insn32 mode.
+ unsigned int
+ stub_max_size() const
+ {
+ if (!this->target_->is_output_micromips()
+ || this->target_->use_32bit_micromips_instructions())
+ return 20;
+ else
+ return 16;
+ }
+
+ // Return the size of the stub. This method expects that the final dynsym
+ // count is set.
+ unsigned int
+ stub_size() const
+ {
+ gold_assert(this->dynsym_count_ != -1U);
+ if (this->dynsym_count_ > 0x10000)
+ return this->stub_max_size();
+ else
+ return this->stub_max_size() - 4;
+ }
+
+ // Return output address of a stub.
+ Mips_address
+ stub_address(const Mips_symbol<size>* sym) const
+ {
+ gold_assert(sym->has_lazy_stub());
+ return this->address() + sym->lazy_stub_offset();
+ }
+
+ protected:
+ void
+ do_adjust_output_section(Output_section* os)
+ { os->set_entsize(0); }
+
+ // Write to a map file.
+ void
+ do_print_to_mapfile(Mapfile* mapfile) const
+ { mapfile->print_output_data(this, _(".MIPS.stubs")); }
+
+ private:
+ static const uint32_t lazy_stub_normal_1[];
+ static const uint32_t lazy_stub_normal_1_n64[];
+ static const uint32_t lazy_stub_normal_2[];
+ static const uint32_t lazy_stub_normal_2_n64[];
+ static const uint32_t lazy_stub_big[];
+ static const uint32_t lazy_stub_big_n64[];
+
+ static const uint32_t lazy_stub_micromips_normal_1[];
+ static const uint32_t lazy_stub_micromips_normal_1_n64[];
+ static const uint32_t lazy_stub_micromips_normal_2[];
+ static const uint32_t lazy_stub_micromips_normal_2_n64[];
+ static const uint32_t lazy_stub_micromips_big[];
+ static const uint32_t lazy_stub_micromips_big_n64[];
+
+ static const uint32_t lazy_stub_micromips32_normal_1[];
+ static const uint32_t lazy_stub_micromips32_normal_1_n64[];
+ static const uint32_t lazy_stub_micromips32_normal_2[];
+ static const uint32_t lazy_stub_micromips32_normal_2_n64[];
+ static const uint32_t lazy_stub_micromips32_big[];
+ static const uint32_t lazy_stub_micromips32_big_n64[];
+
+ // Set the final size.
+ void
+ set_final_data_size()
+ { this->set_data_size(this->symbols_.size() * this->stub_max_size()); }
+
+ // Write out the .MIPS.stubs data.
+ void
+ do_write(Output_file*);
+
+ // .MIPS.stubs symbols
+ Unordered_set<Mips_symbol<size>*> symbols_;
+ // Number of entries in dynamic symbol table.
+ unsigned int dynsym_count_;
+ // Whether the stub offsets are set.
+ bool stub_offsets_are_set_;
+ // The target.
+ Target_mips<size, big_endian>* target_;
+};
+
+// This class handles Mips .reginfo output section.
+
+template<int size, bool big_endian>
+class Mips_output_section_reginfo : public Output_section
+{
+ typedef typename elfcpp::Swap<size, big_endian>::Valtype Valtype;
+
+ public:
+ Mips_output_section_reginfo(const char* name, elfcpp::Elf_Word type,
+ elfcpp::Elf_Xword flags,
+ Target_mips<size, big_endian>* target)
+ : Output_section(name, type, flags), target_(target), gprmask_(0),
+ cprmask1_(0), cprmask2_(0), cprmask3_(0), cprmask4_(0)
+ { }
+
+ // Downcast a base pointer to a Mips_output_section_reginfo pointer.
+ static Mips_output_section_reginfo<size, big_endian>*
+ as_mips_output_section_reginfo(Output_section* os)
+ { return static_cast<Mips_output_section_reginfo<size, big_endian>*>(os); }
+
+ // Set masks of the output .reginfo section.
+ void
+ set_masks(Valtype gprmask, Valtype cprmask1, Valtype cprmask2,
+ Valtype cprmask3, Valtype cprmask4)
+ {
+ this->gprmask_ = gprmask;
+ this->cprmask1_ = cprmask1;
+ this->cprmask2_ = cprmask2;
+ this->cprmask3_ = cprmask3;
+ this->cprmask4_ = cprmask4;
+ }
+
+ protected:
+ // Set the final data size.
+ void
+ set_final_data_size()
+ { this->set_data_size(24); }
+
+ // Write out reginfo section.
+ void
+ do_write(Output_file* of);
+
+ private:
+ Target_mips<size, big_endian>* target_;
+
+ // gprmask of the output .reginfo section.
+ Valtype gprmask_;
+ // cprmask1 of the output .reginfo section.
+ Valtype cprmask1_;
+ // cprmask2 of the output .reginfo section.
+ Valtype cprmask2_;
+ // cprmask3 of the output .reginfo section.
+ Valtype cprmask3_;
+ // cprmask4 of the output .reginfo section.
+ Valtype cprmask4_;
+};
+
+// The MIPS target has relocation types which default handling of relocatable
+// relocation cannot process. So we have to extend the default code.
+
+template<bool big_endian, int sh_type, typename Classify_reloc>
+class Mips_scan_relocatable_relocs :
+ public Default_scan_relocatable_relocs<sh_type, Classify_reloc>
+{
+ public:
+ // Return the strategy to use for a local symbol which is a section
+ // symbol, given the relocation type.
+ inline Relocatable_relocs::Reloc_strategy
+ local_section_strategy(unsigned int r_type, Relobj* object)
+ {
+ if (sh_type == elfcpp::SHT_RELA)
+ return Relocatable_relocs::RELOC_ADJUST_FOR_SECTION_RELA;
+ else
+ {
+ switch (r_type)
+ {
+ case elfcpp::R_MIPS_26:
+ return Relocatable_relocs::RELOC_SPECIAL;
+
+ default:
+ return Default_scan_relocatable_relocs<sh_type, Classify_reloc>::
+ local_section_strategy(r_type, object);
+ }
+ }
+ }
+};
+
+// Mips_copy_relocs class. The only difference from the base class is the
+// method emit_mips, which should be called instead of Copy_reloc_entry::emit.
+// Mips cannot convert all relocation types to dynamic relocs. If a reloc
+// cannot be made dynamic, a COPY reloc is emitted.
+
+template<int sh_type, int size, bool big_endian>
+class Mips_copy_relocs : public Copy_relocs<sh_type, size, big_endian>
+{
+ public:
+ Mips_copy_relocs()
+ : Copy_relocs<sh_type, size, big_endian>(elfcpp::R_MIPS_COPY)
+ { }
+
+ // Emit any saved relocations which turn out to be needed. This is
+ // called after all the relocs have been scanned.
+ void
+ emit_mips(Output_data_reloc<sh_type, true, size, big_endian>*,
+ Symbol_table*, Layout*, Target_mips<size, big_endian>*);
+
+ private:
+ typedef typename Copy_relocs<sh_type, size, big_endian>::Copy_reloc_entry
+ Copy_reloc_entry;
+
+ // Emit this reloc if appropriate. This is called after we have
+ // scanned all the relocations, so we know whether we emitted a
+ // COPY relocation for SYM_.
+ void
+ emit_entry(Copy_reloc_entry& entry,
+ Output_data_reloc<sh_type, true, size, big_endian>* reloc_section,
+ Symbol_table* symtab, Layout* layout,
+ Target_mips<size, big_endian>* target);
+};
+
+
+// Return true if the symbol SYM should be considered to resolve local
+// to the current module, and false otherwise. The logic is taken from
+// GNU ld's method _bfd_elf_symbol_refs_local_p.
+static bool
+symbol_refs_local(const Symbol* sym, bool has_dynsym_entry,
+ bool local_protected)
+{
+ // If it's a local sym, of course we resolve locally.
+ if (sym == NULL)
+ return true;
+
+ // STV_HIDDEN or STV_INTERNAL ones must be local.
+ if (sym->visibility() == elfcpp::STV_HIDDEN
+ || sym->visibility() == elfcpp::STV_INTERNAL)
+ return true;
+
+ // If we don't have a definition in a regular file, then we can't
+ // resolve locally. The sym is either undefined or dynamic.
+ if (sym->source() != Symbol::FROM_OBJECT || sym->object()->is_dynamic()
+ || sym->is_undefined())
+ return false;
+
+ // Forced local symbols resolve locally.
+ if (sym->is_forced_local())
+ return true;
+
+ // As do non-dynamic symbols.
+ if (!has_dynsym_entry)
+ return true;
+
+ // At this point, we know the symbol is defined and dynamic. In an
+ // executable it must resolve locally, likewise when building symbolic
+ // shared libraries.
+ if (parameters->options().output_is_executable()
+ || parameters->options().Bsymbolic())
+ return true;
+
+ // Now deal with defined dynamic symbols in shared libraries. Ones
+ // with default visibility might not resolve locally.
+ if (sym->visibility() == elfcpp::STV_DEFAULT)
+ return false;
+
+ // STV_PROTECTED non-function symbols are local.
+ if (sym->type() != elfcpp::STT_FUNC)
+ return true;
+
+ // Function pointer equality tests may require that STV_PROTECTED
+ // symbols be treated as dynamic symbols. If the address of a
+ // function not defined in an executable is set to that function's
+ // plt entry in the executable, then the address of the function in
+ // a shared library must also be the plt entry in the executable.
+ return local_protected;
+}
+
+// Return TRUE if references to this symbol always reference the symbol in this
+// object.
+static bool
+symbol_references_local(const Symbol* sym, bool has_dynsym_entry)
+{
+ return symbol_refs_local(sym, has_dynsym_entry, false);
+}
+
+// Return TRUE if calls to this symbol always call the version in this object.
+static bool
+symbol_calls_local(const Symbol* sym, bool has_dynsym_entry)
+{
+ return symbol_refs_local(sym, has_dynsym_entry, true);
+}
+
+// Compare GOT offsets of two symbols.
+
+template<int size, bool big_endian>
+static bool
+got_offset_compare(Symbol* sym1, Symbol* sym2)
+{
+ Mips_symbol<size>* mips_sym1 = Mips_symbol<size>::as_mips_sym(sym1);
+ Mips_symbol<size>* mips_sym2 = Mips_symbol<size>::as_mips_sym(sym2);
+ unsigned int area1 = mips_sym1->global_got_area();
+ unsigned int area2 = mips_sym2->global_got_area();
+ gold_assert(area1 != GGA_NONE && area1 != GGA_NONE);
+
+ // GGA_NORMAL entries always come before GGA_RELOC_ONLY.
+ if (area1 != area2)
+ return area1 < area2;
+
+ return mips_sym1->global_gotoffset() < mips_sym2->global_gotoffset();
+}
+
+// This method divides dynamic symbols into symbols that have GOT entry, and
+// symbols that don't have GOT entry. It also sorts symbols with the GOT entry.
+// Mips ABI requires that symbols with the GOT entry must be at the end of
+// dynamic symbol table, and the order in dynamic symbol table must match the
+// order in GOT.
+
+template<int size, bool big_endian>
+static void
+reorder_dyn_symbols(std::vector<Symbol*>* dyn_symbols,
+ std::vector<Symbol*>* non_got_symbols,
+ std::vector<Symbol*>* got_symbols)
+{
+ for (std::vector<Symbol*>::iterator p = dyn_symbols->begin();
+ p != dyn_symbols->end();
+ ++p)
+ {
+ Mips_symbol<size>* mips_sym = Mips_symbol<size>::as_mips_sym(*p);
+ if (mips_sym->global_got_area() == GGA_NORMAL
+ || mips_sym->global_got_area() == GGA_RELOC_ONLY)
+ got_symbols->push_back(mips_sym);
+ else
+ non_got_symbols->push_back(mips_sym);
+ }
+
+ std::sort(got_symbols->begin(), got_symbols->end(),
+ got_offset_compare<size, big_endian>);
+}
+
+// Functor class for processing the global symbol table.
+
+template<int size, bool big_endian>
+class Symbol_visitor_check_symbols
+{
+ public:
+ Symbol_visitor_check_symbols(Target_mips<size, big_endian>* target,
+ Layout* layout, Symbol_table* symtab)
+ : target_(target), layout_(layout), symtab_(symtab)
+ { }
+
+ void
+ operator()(Sized_symbol<size>* sym)
+ {
+ Mips_symbol<size>* mips_sym = Mips_symbol<size>::as_mips_sym(sym);
+ if (local_pic_function<size, big_endian>(mips_sym))
+ {
+ // SYM is a function that might need $25 to be valid on entry.
+ // If we're creating a non-PIC relocatable object, mark SYM as
+ // being PIC. If we're creating a non-relocatable object with
+ // non-PIC branches and jumps to SYM, make sure that SYM has an la25
+ // stub.
+ if (parameters->options().relocatable())
+ {
+ if (!parameters->options().output_is_position_independent())
+ mips_sym->set_pic();
+ }
+ else if (mips_sym->has_nonpic_branches())
+ {
+ this->target_->la25_stub_section(layout_)
+ ->create_la25_stub(this->symtab_, this->target_, mips_sym);
+ }
+ }
+ }
+
+ private:
+ Target_mips<size, big_endian>* target_;
+ Layout* layout_;
+ Symbol_table* symtab_;
+};
+
+template<int size, bool big_endian>
+class Target_mips : public Sized_target<size, big_endian>
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+ typedef Output_data_reloc<elfcpp::SHT_REL, true, size, big_endian>
+ Reloc_section;
+ typedef Output_data_reloc<elfcpp::SHT_RELA, true, size, big_endian>
+ Reloca_section;
+ typedef typename elfcpp::Swap<32, big_endian>::Valtype Valtype32;
+ typedef typename elfcpp::Swap<size, big_endian>::Valtype Valtype;
+
+ public:
+ Target_mips(const Target::Target_info* info = &mips_info)
+ : Sized_target<size, big_endian>(info), got_(NULL), gp_(NULL), plt_(NULL),
+ got_plt_(NULL), rel_dyn_(NULL), copy_relocs_(),
+ dyn_relocs_(), la25_stub_(NULL), mips_mach_extensions_(),
+ mips_stubs_(NULL), ei_class_(0), mach_(0), layout_(NULL),
+ got16_addends_(), entry_symbol_is_compressed_(false), insn32_(false)
+ {
+ this->add_machine_extensions();
+ }
+
+ // The offset of $gp from the beginning of the .got section.
+ static const unsigned int MIPS_GP_OFFSET = 0x7ff0;
+
+ // The maximum size of the GOT for it to be addressable using 16-bit
+ // offsets from $gp.
+ static const unsigned int MIPS_GOT_MAX_SIZE = MIPS_GP_OFFSET + 0x7fff;
+
+ // Make a new symbol table entry for the Mips target.
+ Sized_symbol<size>*
+ make_symbol() const
+ { return new Mips_symbol<size>(); }
+
+ // Process the relocations to determine unreferenced sections for
+ // garbage collection.
+ void
+ gc_process_relocs(Symbol_table* symtab,
+ Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ size_t local_symbol_count,
+ const unsigned char* plocal_symbols);
+
+ // Scan the relocations to look for symbol adjustments.
+ void
+ scan_relocs(Symbol_table* symtab,
+ Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ size_t local_symbol_count,
+ const unsigned char* plocal_symbols);
+
+ // Finalize the sections.
+ void
+ do_finalize_sections(Layout*, const Input_objects*, Symbol_table*);
+
+ // Relocate a section.
+ void
+ relocate_section(const Relocate_info<size, big_endian>*,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ unsigned char* view,
+ Mips_address view_address,
+ section_size_type view_size,
+ const Reloc_symbol_changes*);
+
+ // Scan the relocs during a relocatable link.
+ void
+ scan_relocatable_relocs(Symbol_table* symtab,
+ Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ bool needs_special_offset_handling,
+ size_t local_symbol_count,
+ const unsigned char* plocal_symbols,
+ Relocatable_relocs*);
+
+ // Emit relocations for a section.
+ void
+ relocate_relocs(const Relocate_info<size, big_endian>*,
+ unsigned int sh_type,
+ const unsigned char* prelocs,
+ size_t reloc_count,
+ Output_section* output_section,
+ typename elfcpp::Elf_types<size>::Elf_Off
+ offset_in_output_section,
+ const Relocatable_relocs*,
+ unsigned char* view,
+ Mips_address view_address,
+ section_size_type view_size,
+ unsigned char* reloc_view,
+ section_size_type reloc_view_size);
+
+ // Perform target-specific processing in a relocatable link. This is
+ // only used if we use the relocation strategy RELOC_SPECIAL.
+ void
+ relocate_special_relocatable(const Relocate_info<size, big_endian>* relinfo,
+ unsigned int sh_type,
+ const unsigned char* preloc_in,
+ size_t relnum,
+ Output_section* output_section,
+ typename elfcpp::Elf_types<size>::Elf_Off
+ offset_in_output_section,
+ unsigned char* view,
+ Mips_address view_address,
+ section_size_type view_size,
+ unsigned char* preloc_out);
+
+ // Return whether SYM is defined by the ABI.
+ bool
+ do_is_defined_by_abi(const Symbol* sym) const
+ {
+ return ((strcmp(sym->name(), "__gnu_local_gp") == 0)
+ || (strcmp(sym->name(), "_gp_disp") == 0)
+ || (strcmp(sym->name(), "___tls_get_addr") == 0));
+ }
+
+ // Return the number of entries in the GOT.
+ unsigned int
+ got_entry_count() const
+ {
+ if (!this->has_got_section())
+ return 0;
+ return this->got_size() / (size/8);
+ }
+
+ // Return the number of entries in the PLT.
+ unsigned int
+ plt_entry_count() const
+ {
+ if (this->plt_ == NULL)
+ return 0;
+ return this->plt_->entry_count();
+ }
+
+ // Return the offset of the first non-reserved PLT entry.
+ unsigned int
+ first_plt_entry_offset() const
+ { return this->plt_->first_plt_entry_offset(); }
+
+ // Return the size of each PLT entry.
+ unsigned int
+ plt_entry_size() const
+ { return this->plt_->plt_entry_size(); }
+
+ // Get the GOT section, creating it if necessary.
+ Mips_output_data_got<size, big_endian>*
+ got_section(Symbol_table*, Layout*);
+
+ // Get the GOT section.
+ Mips_output_data_got<size, big_endian>*
+ got_section() const
+ {
+ gold_assert(this->got_ != NULL);
+ return this->got_;
+ }
+
+ // Get the .MIPS.stubs section, creating it if necessary.
+ Mips_output_data_mips_stubs<size, big_endian>*
+ mips_stubs_section(Layout* layout);
+
+ // Get the .MIPS.stubs section.
+ Mips_output_data_mips_stubs<size, big_endian>*
+ mips_stubs_section() const
+ {
+ gold_assert(this->mips_stubs_ != NULL);
+ return this->mips_stubs_;
+ }
+
+ // Get the LA25 stub section, creating it if necessary.
+ Mips_output_data_la25_stub<size, big_endian>*
+ la25_stub_section(Layout*);
+
+ // Get the LA25 stub section.
+ Mips_output_data_la25_stub<size, big_endian>*
+ la25_stub_section()
+ {
+ gold_assert(this->la25_stub_ != NULL);
+ return this->la25_stub_;
+ }
+
+ // Get gp value. It has the value of .got + 0x7FF0.
+ Mips_address
+ gp_value() const
+ {
+ if (this->gp_ != NULL)
+ return this->gp_->value();
+ return 0;
+ }
+
+ // Get gp value. It has the value of .got + 0x7FF0. Adjust it for
+ // multi-GOT links so that OBJECT's GOT + 0x7FF0 is returned.
+ Mips_address
+ adjusted_gp_value(const Mips_relobj<size, big_endian>* object)
+ {
+ if (this->gp_ == NULL)
+ return 0;
+
+ bool multi_got = false;
+ if (this->has_got_section())
+ multi_got = this->got_section()->multi_got();
+ if (!multi_got)
+ return this->gp_->value();
+ else
+ return this->gp_->value() + this->got_section()->get_got_offset(object);
+ }
+
+ // Get the dynamic reloc section, creating it if necessary.
+ Reloc_section*
+ rel_dyn_section(Layout*);
+
+ bool
+ do_has_custom_set_dynsym_indexes() const
+ { return true; }
+
+ // Don't emit input .reginfo sections to output .reginfo.
+ bool
+ do_should_include_section(elfcpp::Elf_Word sh_type) const
+ { return sh_type != elfcpp::SHT_MIPS_REGINFO; }
+
+ // Set the dynamic symbol indexes. INDEX is the index of the first
+ // global dynamic symbol. Pointers to the symbols are stored into the
+ // vector SYMS. The names are added to DYNPOOL. This returns an
+ // updated dynamic symbol index.
+ unsigned int
+ do_set_dynsym_indexes(std::vector<Symbol*>* dyn_symbols, unsigned int index,
+ std::vector<Symbol*>* syms, Stringpool* dynpool,
+ Versions* versions, Symbol_table* symtab) const;
+
+ // Remove .MIPS.stubs entry for a symbol.
+ void
+ remove_lazy_stub_entry(Mips_symbol<size>* sym)
+ {
+ if (this->mips_stubs_ != NULL)
+ this->mips_stubs_->remove_entry(sym);
+ }
+
+ // The value to write into got[1] for SVR4 targets, to identify it is
+ // a GNU object. The dynamic linker can then use got[1] to store the
+ // module pointer.
+ uint64_t
+ mips_elf_gnu_got1_mask()
+ {
+ if (this->is_output_n64())
+ return (uint64_t)1 << 63;
+ else
+ return 1 << 31;
+ }
+
+ // Whether the output has microMIPS code. This is valid only after
+ // merge_processor_specific_flags() is called.
+ bool
+ is_output_micromips() const
+ {
+ gold_assert(this->are_processor_specific_flags_set());
+ return elfcpp::is_micromips(this->processor_specific_flags());
+ }
+
+ // Whether the output uses N32 ABI. This is valid only after
+ // merge_processor_specific_flags() is called.
+ bool
+ is_output_n32() const
+ {
+ gold_assert(this->are_processor_specific_flags_set());
+ return elfcpp::abi_n32(this->processor_specific_flags());
+ }
+
+ // Whether the output uses N64 ABI. This is valid only after
+ // merge_processor_specific_flags() is called.
+ bool
+ is_output_n64() const
+ {
+ gold_assert(this->are_processor_specific_flags_set());
+ return elfcpp::abi_64(this->ei_class_);
+ }
+
+ // Whether the output uses NEWABI. This is valid only after
+ // merge_processor_specific_flags() is called.
+ bool
+ is_output_newabi() const
+ { return this->is_output_n32() || this->is_output_n64(); }
+
+ // Whether we can only use 32-bit microMIPS instructions.
+ bool
+ use_32bit_micromips_instructions() const
+ { return this->insn32_; }
+
+ protected:
+ // Return the value to use for a dynamic symbol which requires special
+ // treatment. This is how we support equality comparisons of function
+ // pointers across shared library boundaries, as described in the
+ // processor specific ABI supplement.
+ uint64_t
+ do_dynsym_value(const Symbol* gsym) const;
+
+ // Make an ELF object.
+ Object*
+ do_make_elf_object(const std::string&, Input_file*, off_t,
+ const elfcpp::Ehdr<size, big_endian>& ehdr);
+
+ Object*
+ do_make_elf_object(const std::string&, Input_file*, off_t,
+ const elfcpp::Ehdr<size, !big_endian>&)
+ { gold_unreachable(); }
+
+ // Make an output section.
+ Output_section*
+ do_make_output_section(const char* name, elfcpp::Elf_Word type,
+ elfcpp::Elf_Xword flags)
+ {
+ if (type == elfcpp::SHT_MIPS_REGINFO)
+ return new Mips_output_section_reginfo<size, big_endian>(name, type,
+ flags, this);
+ else
+ return new Output_section(name, type, flags);
+ }
+
+ // Adjust ELF file header.
+ void
+ do_adjust_elf_header(unsigned char* view, int len);
+
+ // Get the custom dynamic tag value.
+ unsigned int
+ do_dynamic_tag_custom_value(elfcpp::DT) const;
+
+ // Adjust the value written to the dynamic symbol table.
+ virtual void
+ do_adjust_dyn_symbol(const Symbol* sym, unsigned char* view) const
+ {
+ elfcpp::Sym<size, big_endian> isym(view);
+ elfcpp::Sym_write<size, big_endian> osym(view);
+ const Mips_symbol<size>* mips_sym = Mips_symbol<size>::as_mips_sym(sym);
+
+ // Keep dynamic compressed symbols odd. This allows the dynamic linker
+ // to treat compressed symbols like any other.
+ Mips_address value = isym.get_st_value();
+ if (mips_sym->is_mips16() && value != 0)
+ {
+ if (!mips_sym->has_mips16_fn_stub())
+ value |= 1;
+ else
+ {
+ // If we have a MIPS16 function with a stub, the dynamic symbol
+ // must refer to the stub, since only the stub uses the standard
+ // calling conventions. Stub contains MIPS32 code, so don't add +1
+ // in this case.
+
+ // There is a code which does this in the method
+ // Target_mips::do_dynsym_value, but that code will only be
+ // executed if the symbol is from dynobj.
+ // TODO(sasa): GNU ld also changes the value in non-dynamic symbol
+ // table.
+
+ Mips16_stub_section<size, big_endian>* fn_stub =
+ mips_sym->template get_mips16_fn_stub<big_endian>();
+ value = fn_stub->output_address();
+ osym.put_st_size(fn_stub->section_size());
+ }
+
+ osym.put_st_value(value);
+ osym.put_st_other(elfcpp::elf_st_other(sym->visibility(),
+ mips_sym->nonvis() - (elfcpp::STO_MIPS16 >> 2)));
+ }
+ else if ((mips_sym->is_micromips()
+ // Stubs are always microMIPS if there is any microMIPS code in
+ // the output.
+ || (this->is_output_micromips() && mips_sym->has_lazy_stub()))
+ && value != 0)
+ {
+ osym.put_st_value(value | 1);
+ osym.put_st_other(elfcpp::elf_st_other(sym->visibility(),
+ mips_sym->nonvis() - (elfcpp::STO_MICROMIPS >> 2)));
+ }
+ }
+
+ private:
+ // The class which scans relocations.
+ class Scan
+ {
+ public:
+ Scan()
+ { }
+
+ static inline int
+ get_reference_flags(unsigned int r_type);
+
+ inline void
+ local(Symbol_table* symtab, Layout* layout, Target_mips* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rel<size, big_endian>& reloc, unsigned int r_type,
+ const elfcpp::Sym<size, big_endian>& lsym,
+ bool is_discarded);
+
+ inline void
+ local(Symbol_table* symtab, Layout* layout, Target_mips* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>& reloc, unsigned int r_type,
+ const elfcpp::Sym<size, big_endian>& lsym,
+ bool is_discarded);
+
+ inline void
+ local(Symbol_table* symtab, Layout* layout, Target_mips* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>* rela,
+ const elfcpp::Rel<size, big_endian>* rel,
+ unsigned int rel_type,
+ unsigned int r_type,
+ const elfcpp::Sym<size, big_endian>& lsym,
+ bool is_discarded);
+
+ inline void
+ global(Symbol_table* symtab, Layout* layout, Target_mips* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rel<size, big_endian>& reloc, unsigned int r_type,
+ Symbol* gsym);
+
+ inline void
+ global(Symbol_table* symtab, Layout* layout, Target_mips* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>& reloc, unsigned int r_type,
+ Symbol* gsym);
+
+ inline void
+ global(Symbol_table* symtab, Layout* layout, Target_mips* target,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int data_shndx,
+ Output_section* output_section,
+ const elfcpp::Rela<size, big_endian>* rela,
+ const elfcpp::Rel<size, big_endian>* rel,
+ unsigned int rel_type,
+ unsigned int r_type,
+ Symbol* gsym);
+
+ inline bool
+ local_reloc_may_be_function_pointer(Symbol_table* , Layout*,
+ Target_mips*,
+ Sized_relobj_file<size, big_endian>*,
+ unsigned int,
+ Output_section*,
+ const elfcpp::Rel<size, big_endian>&,
+ unsigned int,
+ const elfcpp::Sym<size, big_endian>&)
+ { return false; }
+
+ inline bool
+ global_reloc_may_be_function_pointer(Symbol_table*, Layout*,
+ Target_mips*,
+ Sized_relobj_file<size, big_endian>*,
+ unsigned int,
+ Output_section*,
+ const elfcpp::Rel<size, big_endian>&,
+ unsigned int, Symbol*)
+ { return false; }
+
+ inline bool
+ local_reloc_may_be_function_pointer(Symbol_table*, Layout*,
+ Target_mips*,
+ Sized_relobj_file<size, big_endian>*,
+ unsigned int,
+ Output_section*,
+ const elfcpp::Rela<size, big_endian>&,
+ unsigned int,
+ const elfcpp::Sym<size, big_endian>&)
+ { return false; }
+
+ inline bool
+ global_reloc_may_be_function_pointer(Symbol_table*, Layout*,
+ Target_mips*,
+ Sized_relobj_file<size, big_endian>*,
+ unsigned int,
+ Output_section*,
+ const elfcpp::Rela<size, big_endian>&,
+ unsigned int, Symbol*)
+ { return false; }
+ private:
+ static void
+ unsupported_reloc_local(Sized_relobj_file<size, big_endian>*,
+ unsigned int r_type);
+
+ static void
+ unsupported_reloc_global(Sized_relobj_file<size, big_endian>*,
+ unsigned int r_type, Symbol*);
+ };
+
+ // The class which implements relocation.
+ class Relocate
+ {
+ public:
+ Relocate()
+ { }
+
+ ~Relocate()
+ { }
+
+ // Return whether the R_MIPS_32 relocation needs to be applied.
+ inline bool
+ should_apply_r_mips_32_reloc(const Mips_symbol<size>* gsym,
+ unsigned int r_type,
+ Output_section* output_section,
+ Target_mips* target);
+
+ // Do a relocation. Return false if the caller should not issue
+ // any warnings about this relocation.
+ inline bool
+ relocate(const Relocate_info<size, big_endian>*, Target_mips*,
+ Output_section*, size_t relnum,
+ const elfcpp::Rela<size, big_endian>*,
+ const elfcpp::Rel<size, big_endian>*,
+ unsigned int,
+ unsigned int, const Sized_symbol<size>*,
+ const Symbol_value<size>*,
+ unsigned char*,
+ Mips_address,
+ section_size_type);
+
+ inline bool
+ relocate(const Relocate_info<size, big_endian>*, Target_mips*,
+ Output_section*, size_t relnum,
+ const elfcpp::Rel<size, big_endian>&,
+ unsigned int, const Sized_symbol<size>*,
+ const Symbol_value<size>*,
+ unsigned char*,
+ Mips_address,
+ section_size_type);
+
+ inline bool
+ relocate(const Relocate_info<size, big_endian>*, Target_mips*,
+ Output_section*, size_t relnum,
+ const elfcpp::Rela<size, big_endian>&,
+ unsigned int, const Sized_symbol<size>*,
+ const Symbol_value<size>*,
+ unsigned char*,
+ Mips_address,
+ section_size_type);
+ };
+
+ // A class which returns the size required for a relocation type,
+ // used while scanning relocs during a relocatable link.
+ class Relocatable_size_for_reloc
+ {
+ public:
+ unsigned int
+ get_size_for_reloc(unsigned int, Relobj*);
+ };
+
+ // This POD class holds the dynamic relocations that should be emitted instead
+ // of R_MIPS_32, R_MIPS_REL32 and R_MIPS_64 relocations. We will emit these
+ // relocations if it turns out that the symbol does not have static
+ // relocations.
+ class Dyn_reloc
+ {
+ public:
+ Dyn_reloc(Mips_symbol<size>* sym, unsigned int r_type,
+ Mips_relobj<size, big_endian>* relobj, unsigned int shndx,
+ Output_section* output_section, Mips_address r_offset)
+ : sym_(sym), r_type_(r_type), relobj_(relobj),
+ shndx_(shndx), output_section_(output_section),
+ r_offset_(r_offset)
+ { }
+
+ // Emit this reloc if appropriate. This is called after we have
+ // scanned all the relocations, so we know whether the symbol has
+ // static relocations.
+ void
+ emit(Reloc_section* rel_dyn, Mips_output_data_got<size, big_endian>* got,
+ Symbol_table* symtab)
+ {
+ if (!this->sym_->has_static_relocs())
+ {
+ got->record_global_got_symbol(this->sym_, this->relobj_,
+ this->r_type_, true, false);
+ if (!symbol_references_local(this->sym_,
+ this->sym_->should_add_dynsym_entry(symtab)))
+ rel_dyn->add_global(this->sym_, this->r_type_,
+ this->output_section_, this->relobj_,
+ this->shndx_, this->r_offset_);
+ else
+ rel_dyn->add_symbolless_global_addend(this->sym_, this->r_type_,
+ this->output_section_, this->relobj_,
+ this->shndx_, this->r_offset_);
+ }
+ }
+
+ private:
+ Mips_symbol<size>* sym_;
+ unsigned int r_type_;
+ Mips_relobj<size, big_endian>* relobj_;
+ unsigned int shndx_;
+ Output_section* output_section_;
+ Mips_address r_offset_;
+ };
+
+ // Adjust TLS relocation type based on the options and whether this
+ // is a local symbol.
+ static tls::Tls_optimization
+ optimize_tls_reloc(bool is_final, int r_type);
+
+ // Return whether there is a GOT section.
+ bool
+ has_got_section() const
+ { return this->got_ != NULL; }
+
+ // Check whether the given ELF header flags describe a 32-bit binary.
+ bool
+ mips_32bit_flags(elfcpp::Elf_Word);
+
+ enum Mips_mach {
+ mach_mips3000 = 3000,
+ mach_mips3900 = 3900,
+ mach_mips4000 = 4000,
+ mach_mips4010 = 4010,
+ mach_mips4100 = 4100,
+ mach_mips4111 = 4111,
+ mach_mips4120 = 4120,
+ mach_mips4300 = 4300,
+ mach_mips4400 = 4400,
+ mach_mips4600 = 4600,
+ mach_mips4650 = 4650,
+ mach_mips5000 = 5000,
+ mach_mips5400 = 5400,
+ mach_mips5500 = 5500,
+ mach_mips6000 = 6000,
+ mach_mips7000 = 7000,
+ mach_mips8000 = 8000,
+ mach_mips9000 = 9000,
+ mach_mips10000 = 10000,
+ mach_mips12000 = 12000,
+ mach_mips14000 = 14000,
+ mach_mips16000 = 16000,
+ mach_mips16 = 16,
+ mach_mips5 = 5,
+ mach_mips_loongson_2e = 3001,
+ mach_mips_loongson_2f = 3002,
+ mach_mips_loongson_3a = 3003,
+ mach_mips_sb1 = 12310201, // octal 'SB', 01
+ mach_mips_octeon = 6501,
+ mach_mips_octeonp = 6601,
+ mach_mips_octeon2 = 6502,
+ mach_mips_xlr = 887682, // decimal 'XLR'
+ mach_mipsisa32 = 32,
+ mach_mipsisa32r2 = 33,
+ mach_mipsisa64 = 64,
+ mach_mipsisa64r2 = 65,
+ mach_mips_micromips = 96
+ };
+
+ // Return the MACH for a MIPS e_flags value.
+ unsigned int
+ elf_mips_mach(elfcpp::Elf_Word);
+
+ // Check whether machine EXTENSION is an extension of machine BASE.
+ bool
+ mips_mach_extends(unsigned int, unsigned int);
+
+ // Merge processor specific flags.
+ void
+ merge_processor_specific_flags(const std::string&, elfcpp::Elf_Word,
+ unsigned char, bool);
+
+ // True if we are linking for CPUs that are faster if JAL is converted to BAL.
+ static inline bool
+ jal_to_bal()
+ { return false; }
+
+ // True if we are linking for CPUs that are faster if JALR is converted to
+ // BAL. This should be safe for all architectures. We enable this predicate
+ // for all CPUs.
+ static inline bool
+ jalr_to_bal()
+ { return true; }
+
+ // True if we are linking for CPUs that are faster if JR is converted to B.
+ // This should be safe for all architectures. We enable this predicate for
+ // all CPUs.
+ static inline bool
+ jr_to_b()
+ { return true; }
+
+ // Return the size of the GOT section.
+ section_size_type
+ got_size() const
+ {
+ gold_assert(this->got_ != NULL);
+ return this->got_->data_size();
+ }
+
+ // Create a PLT entry for a global symbol referenced by r_type relocation.
+ void
+ make_plt_entry(Symbol_table*, Layout*, Mips_symbol<size>*,
+ unsigned int r_type);
+
+ // Get the PLT section.
+ Mips_output_data_plt<size, big_endian>*
+ plt_section() const
+ {
+ gold_assert(this->plt_ != NULL);
+ return this->plt_;
+ }
+
+ // Get the GOT PLT section.
+ const Mips_output_data_plt<size, big_endian>*
+ got_plt_section() const
+ {
+ gold_assert(this->got_plt_ != NULL);
+ return this->got_plt_;
+ }
+
+ // Copy a relocation against a global symbol.
+ void
+ copy_reloc(Symbol_table* symtab, Layout* layout,
+ Sized_relobj_file<size, big_endian>* object,
+ unsigned int shndx, Output_section* output_section,
+ Symbol* sym, const elfcpp::Rel<size, big_endian>& reloc)
+ {
+ this->copy_relocs_.copy_reloc(symtab, layout,
+ symtab->get_sized_symbol<size>(sym),
+ object, shndx, output_section,
+ reloc, this->rel_dyn_section(layout));
+ }
+
+ void
+ dynamic_reloc(Mips_symbol<size>* sym, unsigned int r_type,
+ Mips_relobj<size, big_endian>* relobj,
+ unsigned int shndx, Output_section* output_section,
+ Mips_address r_offset)
+ {
+ this->dyn_relocs_.push_back(Dyn_reloc(sym, r_type, relobj, shndx,
+ output_section, r_offset));
+ }
+
+ // Calculate value of _gp symbol.
+ void
+ set_gp(Layout*, Symbol_table*);
+
+ const char*
+ elf_mips_abi_name(elfcpp::Elf_Word e_flags, unsigned char ei_class);
+ const char*
+ elf_mips_mach_name(elfcpp::Elf_Word e_flags);
+
+ // Adds entries that describe how machines relate to one another. The entries
+ // are ordered topologically with MIPS I extensions listed last. First
+ // element is extension, second element is base.
+ void
+ add_machine_extensions()
+ {
+ // MIPS64r2 extensions.
+ this->add_extension(mach_mips_octeon2, mach_mips_octeonp);
+ this->add_extension(mach_mips_octeonp, mach_mips_octeon);
+ this->add_extension(mach_mips_octeon, mach_mipsisa64r2);
+
+ // MIPS64 extensions.
+ this->add_extension(mach_mipsisa64r2, mach_mipsisa64);
+ this->add_extension(mach_mips_sb1, mach_mipsisa64);
+ this->add_extension(mach_mips_xlr, mach_mipsisa64);
+ this->add_extension(mach_mips_loongson_3a, mach_mipsisa64);
+
+ // MIPS V extensions.
+ this->add_extension(mach_mipsisa64, mach_mips5);
+
+ // R10000 extensions.
+ this->add_extension(mach_mips12000, mach_mips10000);
+ this->add_extension(mach_mips14000, mach_mips10000);
+ this->add_extension(mach_mips16000, mach_mips10000);
+
+ // R5000 extensions. Note: the vr5500 ISA is an extension of the core
+ // vr5400 ISA, but doesn't include the multimedia stuff. It seems
+ // better to allow vr5400 and vr5500 code to be merged anyway, since
+ // many libraries will just use the core ISA. Perhaps we could add
+ // some sort of ASE flag if this ever proves a problem.
+ this->add_extension(mach_mips5500, mach_mips5400);
+ this->add_extension(mach_mips5400, mach_mips5000);
+
+ // MIPS IV extensions.
+ this->add_extension(mach_mips5, mach_mips8000);
+ this->add_extension(mach_mips10000, mach_mips8000);
+ this->add_extension(mach_mips5000, mach_mips8000);
+ this->add_extension(mach_mips7000, mach_mips8000);
+ this->add_extension(mach_mips9000, mach_mips8000);
+
+ // VR4100 extensions.
+ this->add_extension(mach_mips4120, mach_mips4100);
+ this->add_extension(mach_mips4111, mach_mips4100);
+
+ // MIPS III extensions.
+ this->add_extension(mach_mips_loongson_2e, mach_mips4000);
+ this->add_extension(mach_mips_loongson_2f, mach_mips4000);
+ this->add_extension(mach_mips8000, mach_mips4000);
+ this->add_extension(mach_mips4650, mach_mips4000);
+ this->add_extension(mach_mips4600, mach_mips4000);
+ this->add_extension(mach_mips4400, mach_mips4000);
+ this->add_extension(mach_mips4300, mach_mips4000);
+ this->add_extension(mach_mips4100, mach_mips4000);
+ this->add_extension(mach_mips4010, mach_mips4000);
+
+ // MIPS32 extensions.
+ this->add_extension(mach_mipsisa32r2, mach_mipsisa32);
+
+ // MIPS II extensions.
+ this->add_extension(mach_mips4000, mach_mips6000);
+ this->add_extension(mach_mipsisa32, mach_mips6000);
+
+ // MIPS I extensions.
+ this->add_extension(mach_mips6000, mach_mips3000);
+ this->add_extension(mach_mips3900, mach_mips3000);
+ }
+
+ // Add value to MIPS extenstions.
+ void
+ add_extension(unsigned int base, unsigned int extension)
+ {
+ std::pair<unsigned int, unsigned int> ext(base, extension);
+ this->mips_mach_extensions_.push_back(ext);
+ }
+
+ // Return the number of entries in the .dynsym section.
+ unsigned int get_dt_mips_symtabno() const
+ {
+ return ((unsigned int)(this->layout_->dynsym_section()->data_size()
+ / elfcpp::Elf_sizes<size>::sym_size));
+ // TODO(sasa): Entry size is MIPS_ELF_SYM_SIZE.
+ }
+
+ // Information about this specific target which we pass to the
+ // general Target structure.
+ static Target::Target_info mips_info;
+ // The GOT section.
+ Mips_output_data_got<size, big_endian>* got_;
+ // gp symbol. It has the value of .got + 0x7FF0.
+ Sized_symbol<size>* gp_;
+ // The PLT section.
+ Mips_output_data_plt<size, big_endian>* plt_;
+ // The GOT PLT section.
+ Output_data_space* got_plt_;
+ // The dynamic reloc section.
+ Reloc_section* rel_dyn_;
+ // Relocs saved to avoid a COPY reloc.
+ Mips_copy_relocs<elfcpp::SHT_REL, size, big_endian> copy_relocs_;
+
+ // A list of dyn relocs to be saved.
+ std::vector<Dyn_reloc> dyn_relocs_;
+
+ // The LA25 stub section.
+ Mips_output_data_la25_stub<size, big_endian>* la25_stub_;
+ // Architecture extensions.
+ std::vector<std::pair<unsigned int, unsigned int> > mips_mach_extensions_;
+ // .MIPS.stubs
+ Mips_output_data_mips_stubs<size, big_endian>* mips_stubs_;
+
+ unsigned char ei_class_;
+ unsigned int mach_;
+ Layout* layout_;
+
+ typename std::list<got16_addend<size, big_endian> > got16_addends_;
+
+ // Whether the entry symbol is mips16 or micromips.
+ bool entry_symbol_is_compressed_;
+
+ // Whether we can use only 32-bit microMIPS instructions.
+ // TODO(sasa): This should be a linker option.
+ bool insn32_;
+};
+
+
+// Helper structure for R_MIPS*_HI16/LO16 and R_MIPS*_GOT16/LO16 relocations.
+// It records high part of the relocation pair.
+
+template<int size, bool big_endian>
+struct reloc_high
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+
+ reloc_high(unsigned char* _view, const Mips_relobj<size, big_endian>* _object,
+ const Symbol_value<size>* _psymval, Mips_address _addend,
+ unsigned int _r_type, bool _extract_addend,
+ Mips_address _address = 0, bool _gp_disp = false)
+ : view(_view), object(_object), psymval(_psymval), addend(_addend),
+ r_type(_r_type), extract_addend(_extract_addend), address(_address),
+ gp_disp(_gp_disp)
+ { }
+
+ unsigned char* view;
+ const Mips_relobj<size, big_endian>* object;
+ const Symbol_value<size>* psymval;
+ Mips_address addend;
+ unsigned int r_type;
+ bool extract_addend;
+ Mips_address address;
+ bool gp_disp;
+};
+
+template<int size, bool big_endian>
+class Mips_relocate_functions : public Relocate_functions<size, big_endian>
+{
+ typedef typename elfcpp::Elf_types<size>::Elf_Addr Mips_address;
+ typedef typename elfcpp::Swap<16, big_endian>::Valtype Valtype16;
+ typedef typename elfcpp::Swap<32, big_endian>::Valtype Valtype32;
+
+ public:
+ typedef enum
+ {
+ STATUS_OKAY, // No error during relocation.
+ STATUS_OVERFLOW, // Relocation overflow.
+ STATUS_BAD_RELOC // Relocation cannot be applied.
+ } Status;
+
+ private:
+ typedef Relocate_functions<size, big_endian> Base;
+ typedef Mips_relocate_functions<size, big_endian> This;
+
+ static typename std::list<reloc_high<size, big_endian> > hi16_relocs;
+ static typename std::list<reloc_high<size, big_endian> > got16_relocs;
+
+ // R_MIPS16_26 is used for the mips16 jal and jalx instructions.
+ // Most mips16 instructions are 16 bits, but these instructions
+ // are 32 bits.
+ //
+ // The format of these instructions is:
+ //
+ // +--------------+--------------------------------+
+ // | JALX | X| Imm 20:16 | Imm 25:21 |
+ // +--------------+--------------------------------+
+ // | Immediate 15:0 |
+ // +-----------------------------------------------+
+ //
+ // JALX is the 5-bit value 00011. X is 0 for jal, 1 for jalx.
+ // Note that the immediate value in the first word is swapped.
+ //
+ // When producing a relocatable object file, R_MIPS16_26 is
+ // handled mostly like R_MIPS_26. In particular, the addend is
+ // stored as a straight 26-bit value in a 32-bit instruction.
+ // (gas makes life simpler for itself by never adjusting a
+ // R_MIPS16_26 reloc to be against a section, so the addend is
+ // always zero). However, the 32 bit instruction is stored as 2
+ // 16-bit values, rather than a single 32-bit value. In a
+ // big-endian file, the result is the same; in a little-endian
+ // file, the two 16-bit halves of the 32 bit value are swapped.
+ // This is so that a disassembler can recognize the jal
+ // instruction.
+ //
+ // When doing a final link, R_MIPS16_26 is treated as a 32 bit
+ // instruction stored as two 16-bit values. The addend A is the
+ // contents of the targ26 field. The calculation is the same as
+ // R_MIPS_26. When storing the calculated value, reorder the
+ // immediate value as shown above, and don't forget to store the
+ // value as two 16-bit values.
+ //
+ // To put it in MIPS ABI terms, the relocation field is T-targ26-16,
+ // defined as
+ //
+ // big-endian:
+ // +--------+----------------------+
+ // | | |
+ // | | targ26-16 |
+ // |31 26|25 0|
+ // +--------+----------------------+
+ //
+ // little-endian:
+ // +----------+------+-------------+
+ // | | | |
+ // | sub1 | | sub2 |
+ // |0 9|10 15|16 31|
+ // +----------+--------------------+
+ // where targ26-16 is sub1 followed by sub2 (i.e., the addend field A is
+ // ((sub1 << 16) | sub2)).
+ //
+ // When producing a relocatable object file, the calculation is
+ // (((A < 2) | ((P + 4) & 0xf0000000) + S) >> 2)
+ // When producing a fully linked file, the calculation is
+ // let R = (((A < 2) | ((P + 4) & 0xf0000000) + S) >> 2)
+ // ((R & 0x1f0000) << 5) | ((R & 0x3e00000) >> 5) | (R & 0xffff)
+ //
+ // The table below lists the other MIPS16 instruction relocations.
+ // Each one is calculated in the same way as the non-MIPS16 relocation
+ // given on the right, but using the extended MIPS16 layout of 16-bit
+ // immediate fields:
+ //
+ // R_MIPS16_GPREL R_MIPS_GPREL16
+ // R_MIPS16_GOT16 R_MIPS_GOT16
+ // R_MIPS16_CALL16 R_MIPS_CALL16
+ // R_MIPS16_HI16 R_MIPS_HI16
+ // R_MIPS16_LO16 R_MIPS_LO16
+ //
+ // A typical instruction will have a format like this:
+ //
+ // +--------------+--------------------------------+
+ // | EXTEND | Imm 10:5 | Imm 15:11 |
+ // +--------------+--------------------------------+
+ // | Major | rx | ry | Imm 4:0 |
+ // +--------------+--------------------------------+
+ //
+ // EXTEND is the five bit value 11110. Major is the instruction
+ // opcode.
+ //
+ // All we need to do here is shuffle the bits appropriately.
+ // As above, the two 16-bit halves must be swapped on a
+ // little-endian system.
+
+ // Similar to MIPS16, the two 16-bit halves in microMIPS must be swapped
+ // on a little-endian system. This does not apply to R_MICROMIPS_PC7_S1
+ // and R_MICROMIPS_PC10_S1 relocs that apply to 16-bit instructions.
+
+ static inline bool
+ should_shuffle_micromips_reloc(unsigned int r_type)
+ {
+ return (micromips_reloc(r_type)
+ && r_type != elfcpp::R_MICROMIPS_PC7_S1
+ && r_type != elfcpp::R_MICROMIPS_PC10_S1);
+ }
+
+ static void
+ mips_reloc_unshuffle(unsigned char* view, unsigned int r_type,
+ bool jal_shuffle)
+ {
+ if (!mips16_reloc(r_type)
+ && !should_shuffle_micromips_reloc(r_type))
+ return;
+
+ // Pick up the first and second halfwords of the instruction.
+ Valtype16 first = elfcpp::Swap<16, big_endian>::readval(view);
+ Valtype16 second = elfcpp::Swap<16, big_endian>::readval(view + 2);
+ Valtype32 val;
+
+ if (micromips_reloc(r_type)
+ || (r_type == elfcpp::R_MIPS16_26 && !jal_shuffle))
+ val = first << 16 | second;
+ else if (r_type != elfcpp::R_MIPS16_26)
+ val = (((first & 0xf800) << 16) | ((second & 0xffe0) << 11)
+ | ((first & 0x1f) << 11) | (first & 0x7e0) | (second & 0x1f));
+ else
+ val = (((first & 0xfc00) << 16) | ((first & 0x3e0) << 11)
+ | ((first & 0x1f) << 21) | second);
+
+ elfcpp::Swap<32, big_endian>::writeval(view, val);
+ }
+
+ static void
+ mips_reloc_shuffle(unsigned char* view, unsigned int r_type, bool jal_shuffle)
+ {
+ if (!mips16_reloc(r_type)
+ && !should_shuffle_micromips_reloc(r_type))
+ return;
+
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(view);
+ Valtype16 first, second;
+
+ if (micromips_reloc(r_type)
+ || (r_type == elfcpp::R_MIPS16_26 && !jal_shuffle))
+ {
+ second = val & 0xffff;
+ first = val >> 16;
+ }
+ else if (r_type != elfcpp::R_MIPS16_26)
+ {
+ second = ((val >> 11) & 0xffe0) | (val & 0x1f);
+ first = ((val >> 16) & 0xf800) | ((val >> 11) & 0x1f) | (val & 0x7e0);
+ }
+ else
+ {
+ second = val & 0xffff;
+ first = ((val >> 16) & 0xfc00) | ((val >> 11) & 0x3e0)
+ | ((val >> 21) & 0x1f);
+ }
+
+ elfcpp::Swap<16, big_endian>::writeval(view + 2, second);
+ elfcpp::Swap<16, big_endian>::writeval(view, first);
+ }
+
+ public:
+ // R_MIPS_16: S + sign-extend(A)
+ static inline typename This::Status
+ rel16(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype16* wv = reinterpret_cast<Valtype16*>(view);
+ Valtype16 val = elfcpp::Swap<16, big_endian>::readval(wv);
+
+ Valtype32 addend = (extract_addend ? Bits<16>::sign_extend32(val)
+ : Bits<16>::sign_extend32(addend_a));
+
+ Valtype32 x = psymval->value(object, addend);
+ val = Bits<16>::bit_select32(val, x, 0xffffU);
+ elfcpp::Swap<16, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<16>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_32: S + A
+ static inline typename This::Status
+ rel32(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 addend = (extract_addend
+ ? elfcpp::Swap<32, big_endian>::readval(wv)
+ : Bits<32>::sign_extend32(addend_a));
+ Valtype32 x = psymval->value(object, addend);
+ elfcpp::Swap<32, big_endian>::writeval(wv, x);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_JALR, R_MICROMIPS_JALR
+ static inline typename This::Status
+ reljalr(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ Mips_address addend_a, bool extract_addend, bool cross_mode_jump,
+ unsigned int r_type, bool jalr_to_bal, bool jr_to_b)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 addend = extract_addend ? 0 : addend_a;
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ // Try converting J(AL)R to B(AL), if the target is in range.
+ if (!parameters->options().relocatable()
+ && r_type == elfcpp::R_MIPS_JALR
+ && !cross_mode_jump
+ && ((jalr_to_bal && val == 0x0320f809) // jalr t9
+ || (jr_to_b && val == 0x03200008))) // jr t9
+ {
+ int offset = psymval->value(object, addend) - (address + 4);
+ if (!Bits<18>::has_overflow32(offset))
+ {
+ if (val == 0x03200008) // jr t9
+ val = 0x10000000 | (((Valtype32)offset >> 2) & 0xffff); // b addr
+ else
+ val = 0x04110000 | (((Valtype32)offset >> 2) & 0xffff); //bal addr
+ }
+ }
+
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_PC32: S + A - P
+ static inline typename This::Status
+ relpc32(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ Mips_address addend_a, bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 addend = (extract_addend
+ ? elfcpp::Swap<32, big_endian>::readval(wv)
+ : Bits<32>::sign_extend32(addend_a));
+ Valtype32 x = psymval->value(object, addend) - address;
+ elfcpp::Swap<32, big_endian>::writeval(wv, x);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_26, R_MIPS16_26, R_MICROMIPS_26_S1
+ static inline typename This::Status
+ rel26(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ bool local, Mips_address addend_a, bool extract_addend,
+ const Symbol* gsym, bool cross_mode_jump, unsigned int r_type,
+ bool jal_to_bal)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend;
+ if (extract_addend)
+ {
+ if (r_type == elfcpp::R_MICROMIPS_26_S1)
+ addend = (val & 0x03ffffff) << 1;
+ else
+ addend = (val & 0x03ffffff) << 2;
+ }
+ else
+ addend = addend_a;
+
+ // Make sure the target of JALX is word-aligned. Bit 0 must be
+ // the correct ISA mode selector and bit 1 must be 0.
+ if (cross_mode_jump
+ && (psymval->value(object, 0) & 3) != (r_type == elfcpp::R_MIPS_26))
+ {
+ gold_warning(_("JALX to a non-word-aligned address"));
+ mips_reloc_shuffle(view, r_type, !parameters->options().relocatable());
+ return This::STATUS_BAD_RELOC;
+ }
+
+ // Shift is 2, unusually, for microMIPS JALX.
+ unsigned int shift =
+ (!cross_mode_jump && r_type == elfcpp::R_MICROMIPS_26_S1) ? 1 : 2;
+
+ Valtype32 x;
+ if (local)
+ x = addend | ((address + 4) & (0xfc000000 << shift));
+ else
+ {
+ if (shift == 1)
+ x = Bits<27>::sign_extend32(addend);
+ else
+ x = Bits<28>::sign_extend32(addend);
+ }
+ x = psymval->value(object, x) >> shift;
+
+ if (!local && !gsym->is_weak_undefined())
+ {
+ if ((x >> 26) != ((address + 4) >> (26 + shift)))
+ {
+ gold_error(_("relocation truncated to fit: %u against '%s'"),
+ r_type, gsym->name());
+ return This::STATUS_OVERFLOW;
+ }
+ }
+
+ val = Bits<32>::bit_select32(val, x, 0x03ffffff);
+
+ // If required, turn JAL into JALX.
+ if (cross_mode_jump)
+ {
+ bool ok;
+ Valtype32 opcode = val >> 26;
+ Valtype32 jalx_opcode;
+
+ // Check to see if the opcode is already JAL or JALX.
+ if (r_type == elfcpp::R_MIPS16_26)
+ {
+ ok = (opcode == 0x6) || (opcode == 0x7);
+ jalx_opcode = 0x7;
+ }
+ else if (r_type == elfcpp::R_MICROMIPS_26_S1)
+ {
+ ok = (opcode == 0x3d) || (opcode == 0x3c);
+ jalx_opcode = 0x3c;
+ }
+ else
+ {
+ ok = (opcode == 0x3) || (opcode == 0x1d);
+ jalx_opcode = 0x1d;
+ }
+
+ // If the opcode is not JAL or JALX, there's a problem. We cannot
+ // convert J or JALS to JALX.
+ if (!ok)
+ {
+ gold_error(_("Unsupported jump between ISA modes; consider "
+ "recompiling with interlinking enabled."));
+ return This::STATUS_BAD_RELOC;
+ }
+
+ // Make this the JALX opcode.
+ val = (val & ~(0x3f << 26)) | (jalx_opcode << 26);
+ }
+
+ // Try converting JAL to BAL, if the target is in range.
+ if (!parameters->options().relocatable()
+ && !cross_mode_jump
+ && ((jal_to_bal
+ && r_type == elfcpp::R_MIPS_26
+ && (val >> 26) == 0x3))) // jal addr
+ {
+ Valtype32 dest = (x << 2) | (((address + 4) >> 28) << 28);
+ int offset = dest - (address + 4);
+ if (!Bits<18>::has_overflow32(offset))
+ {
+ if (val == 0x03200008) // jr t9
+ val = 0x10000000 | (((Valtype32)offset >> 2) & 0xffff); // b addr
+ else
+ val = 0x04110000 | (((Valtype32)offset >> 2) & 0xffff); //bal addr
+ }
+ }
+
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, !parameters->options().relocatable());
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_PC16
+ static inline typename This::Status
+ relpc16(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ Mips_address addend_a, bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = extract_addend ? (val & 0xffff) << 2 : addend_a;
+ addend = Bits<18>::sign_extend32(addend);
+
+ Valtype32 x = psymval->value(object, addend) - address;
+ val = Bits<16>::bit_select32(val, x >> 2, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<18>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MICROMIPS_PC7_S1
+ static inline typename This::Status
+ relmicromips_pc7_s1(unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ Mips_address addend_a, bool extract_addend,
+ unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = extract_addend ? (val & 0x7f) << 1 : addend_a;
+ addend = Bits<8>::sign_extend32(addend);
+
+ Valtype32 x = psymval->value(object, addend) - address;
+ val = Bits<16>::bit_select32(val, x >> 1, 0x7f);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<8>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MICROMIPS_PC10_S1
+ static inline typename This::Status
+ relmicromips_pc10_s1(unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ Mips_address addend_a, bool extract_addend,
+ unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = extract_addend ? (val & 0x3ff) << 1 : addend_a;
+ addend = Bits<11>::sign_extend32(addend);
+
+ Valtype32 x = psymval->value(object, addend) - address;
+ val = Bits<16>::bit_select32(val, x >> 1, 0x3ff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<11>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MICROMIPS_PC16_S1
+ static inline typename This::Status
+ relmicromips_pc16_s1(unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address address,
+ Mips_address addend_a, bool extract_addend,
+ unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = extract_addend ? (val & 0xffff) << 1 : addend_a;
+ addend = Bits<17>::sign_extend32(addend);
+
+ Valtype32 x = psymval->value(object, addend) - address;
+ val = Bits<16>::bit_select32(val, x >> 1, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<17>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_HI16, R_MIPS16_HI16, R_MICROMIPS_HI16,
+ static inline typename This::Status
+ relhi16(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend,
+ Mips_address address, bool gp_disp, unsigned int r_type,
+ bool extract_addend)
+ {
+ // Record the relocation. It will be resolved when we find lo16 part.
+ hi16_relocs.push_back(reloc_high<size, big_endian>(view, object, psymval,
+ addend, r_type, extract_addend, address, gp_disp));
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_HI16, R_MIPS16_HI16, R_MICROMIPS_HI16,
+ static inline typename This::Status
+ do_relhi16(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_hi,
+ Mips_address address, bool is_gp_disp, unsigned int r_type,
+ bool extract_addend, Valtype32 addend_lo,
+ Target_mips<size, big_endian>* target)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = (extract_addend ? ((val & 0xffff) << 16) + addend_lo
+ : addend_hi);
+
+ Valtype32 value;
+ if (!is_gp_disp)
+ value = psymval->value(object, addend);
+ else
+ {
+ // For MIPS16 ABI code we generate this sequence
+ // 0: li $v0,%hi(_gp_disp)
+ // 4: addiupc $v1,%lo(_gp_disp)
+ // 8: sll $v0,16
+ // 12: addu $v0,$v1
+ // 14: move $gp,$v0
+ // So the offsets of hi and lo relocs are the same, but the
+ // base $pc is that used by the ADDIUPC instruction at $t9 + 4.
+ // ADDIUPC clears the low two bits of the instruction address,
+ // so the base is ($t9 + 4) & ~3.
+ Valtype32 gp_disp;
+ if (r_type == elfcpp::R_MIPS16_HI16)
+ gp_disp = (target->adjusted_gp_value(object)
+ - ((address + 4) & ~0x3));
+ // The microMIPS .cpload sequence uses the same assembly
+ // instructions as the traditional psABI version, but the
+ // incoming $t9 has the low bit set.
+ else if (r_type == elfcpp::R_MICROMIPS_HI16)
+ gp_disp = target->adjusted_gp_value(object) - address - 1;
+ else
+ gp_disp = target->adjusted_gp_value(object) - address;
+ value = gp_disp + addend;
+ }
+ Valtype32 x = ((value + 0x8000) >> 16) & 0xffff;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (is_gp_disp && Bits<16>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_GOT16, R_MIPS16_GOT16, R_MICROMIPS_GOT16
+ static inline typename This::Status
+ relgot16_local(unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, unsigned int r_type)
+ {
+ // Record the relocation. It will be resolved when we find lo16 part.
+ got16_relocs.push_back(reloc_high<size, big_endian>(view, object, psymval,
+ addend_a, r_type, extract_addend));
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_GOT16, R_MIPS16_GOT16, R_MICROMIPS_GOT16
+ static inline typename This::Status
+ do_relgot16_local(unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_hi,
+ unsigned int r_type, bool extract_addend,
+ Valtype32 addend_lo, Target_mips<size, big_endian>* target)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = (extract_addend ? ((val & 0xffff) << 16) + addend_lo
+ : addend_hi);
+
+ // Find GOT page entry.
+ Mips_address value = ((psymval->value(object, addend) + 0x8000) >> 16)
+ & 0xffff;
+ value <<= 16;
+ unsigned int got_offset =
+ target->got_section()->get_got_page_offset(value, object);
+
+ // Resolve the relocation.
+ Valtype32 x = target->got_section()->gp_offset(got_offset, object);
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<16>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_LO16, R_MIPS16_LO16, R_MICROMIPS_LO16, R_MICROMIPS_HI0_LO16
+ static inline typename This::Status
+ rello16(Target_mips<size, big_endian>* target, unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, Mips_address address, bool is_gp_disp,
+ unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend = (extract_addend ? Bits<16>::sign_extend32(val & 0xffff)
+ : addend_a);
+
+ // Resolve pending R_MIPS_HI16 relocations.
+ typename std::list<reloc_high<size, big_endian> >::iterator it =
+ hi16_relocs.begin();
+ while (it != hi16_relocs.end())
+ {
+ reloc_high<size, big_endian> hi16 = *it;
+ if (hi16.psymval->value(hi16.object, 0) == psymval->value(object, 0))
+ {
+ if (do_relhi16(hi16.view, hi16.object, hi16.psymval, hi16.addend,
+ hi16.address, hi16.gp_disp, hi16.r_type,
+ hi16.extract_addend, addend, target)
+ == This::STATUS_OVERFLOW)
+ return This::STATUS_OVERFLOW;
+ it = hi16_relocs.erase(it);
+ }
+ else
+ ++it;
+ }
+
+ // Resolve pending local R_MIPS_GOT16 relocations.
+ typename std::list<reloc_high<size, big_endian> >::iterator it2 =
+ got16_relocs.begin();
+ while (it2 != got16_relocs.end())
+ {
+ reloc_high<size, big_endian> got16 = *it2;
+ if (got16.psymval->value(got16.object, 0) == psymval->value(object, 0))
+ {
+ if (do_relgot16_local(got16.view, got16.object, got16.psymval,
+ got16.addend, got16.r_type,
+ got16.extract_addend, addend,
+ target) == This::STATUS_OVERFLOW)
+ return This::STATUS_OVERFLOW;
+ it2 = got16_relocs.erase(it2);
+ }
+ else
+ ++it2;
+ }
+
+ // Resolve R_MIPS_LO16 relocation.
+ Valtype32 x;
+ if (!is_gp_disp)
+ x = psymval->value(object, addend);
+ else
+ {
+ // See the comment for R_MIPS16_HI16 above for the reason
+ // for this conditional.
+ Valtype32 gp_disp;
+ if (r_type == elfcpp::R_MIPS16_LO16)
+ gp_disp = target->adjusted_gp_value(object) - (address & ~0x3);
+ else if (r_type == elfcpp::R_MICROMIPS_LO16
+ || r_type == elfcpp::R_MICROMIPS_HI0_LO16)
+ gp_disp = target->adjusted_gp_value(object) - address + 3;
+ else
+ gp_disp = target->adjusted_gp_value(object) - address + 4;
+ // The MIPS ABI requires checking the R_MIPS_LO16 relocation
+ // for overflow. Relocations against _gp_disp are normally
+ // generated from the .cpload pseudo-op. It generates code
+ // that normally looks like this:
+
+ // lui $gp,%hi(_gp_disp)
+ // addiu $gp,$gp,%lo(_gp_disp)
+ // addu $gp,$gp,$t9
+
+ // Here $t9 holds the address of the function being called,
+ // as required by the MIPS ELF ABI. The R_MIPS_LO16
+ // relocation can easily overflow in this situation, but the
+ // R_MIPS_HI16 relocation will handle the overflow.
+ // Therefore, we consider this a bug in the MIPS ABI, and do
+ // not check for overflow here.
+ x = gp_disp + addend;
+ }
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_CALL16, R_MIPS16_CALL16, R_MICROMIPS_CALL16
+ // R_MIPS_GOT16, R_MIPS16_GOT16, R_MICROMIPS_GOT16
+ // R_MIPS_TLS_GD, R_MIPS16_TLS_GD, R_MICROMIPS_TLS_GD
+ // R_MIPS_TLS_GOTTPREL, R_MIPS16_TLS_GOTTPREL, R_MICROMIPS_TLS_GOTTPREL
+ // R_MIPS_TLS_LDM, R_MIPS16_TLS_LDM, R_MICROMIPS_TLS_LDM
+ // R_MIPS_GOT_DISP, R_MICROMIPS_GOT_DISP
+ static inline typename This::Status
+ relgot(unsigned char* view, int gp_offset, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 x = gp_offset;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<16>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_GOT_PAGE, R_MICROMIPS_GOT_PAGE
+ static inline typename This::Status
+ relgotpage(Target_mips<size, big_endian>* target, unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(view);
+ Valtype32 addend = extract_addend ? val & 0xffff : addend_a;
+
+ // Find a GOT page entry that points to within 32KB of symbol + addend.
+ Mips_address value = (psymval->value(object, addend) + 0x8000) & ~0xffff;
+ unsigned int got_offset =
+ target->got_section()->get_got_page_offset(value, object);
+
+ Valtype32 x = target->got_section()->gp_offset(got_offset, object);
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<16>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_GOT_OFST, R_MICROMIPS_GOT_OFST
+ static inline typename This::Status
+ relgotofst(Target_mips<size, big_endian>* target, unsigned char* view,
+ const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, bool local, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(view);
+ Valtype32 addend = extract_addend ? val & 0xffff : addend_a;
+
+ // For a local symbol, find a GOT page entry that points to within 32KB of
+ // symbol + addend. Relocation value is the offset of the GOT page entry's
+ // value from symbol + addend.
+ // For a global symbol, relocation value is addend.
+ Valtype32 x;
+ if (local)
+ {
+ // Find GOT page entry.
+ Mips_address value = ((psymval->value(object, addend) + 0x8000)
+ & ~0xffff);
+ target->got_section()->get_got_page_offset(value, object);
+
+ x = psymval->value(object, addend) - value;
+ }
+ else
+ x = addend;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return (Bits<16>::has_overflow32(x)
+ ? This::STATUS_OVERFLOW
+ : This::STATUS_OKAY);
+ }
+
+ // R_MIPS_GOT_HI16, R_MIPS_CALL_HI16,
+ // R_MICROMIPS_GOT_HI16, R_MICROMIPS_CALL_HI16
+ static inline typename This::Status
+ relgot_hi16(unsigned char* view, int gp_offset, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 x = gp_offset;
+ x = ((x + 0x8000) >> 16) & 0xffff;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_GOT_LO16, R_MIPS_CALL_LO16,
+ // R_MICROMIPS_GOT_LO16, R_MICROMIPS_CALL_LO16
+ static inline typename This::Status
+ relgot_lo16(unsigned char* view, int gp_offset, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 x = gp_offset;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_GPREL16, R_MIPS16_GPREL, R_MIPS_LITERAL, R_MICROMIPS_LITERAL
+ // R_MICROMIPS_GPREL7_S2, R_MICROMIPS_GPREL16
+ static inline typename This::Status
+ relgprel(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address gp,
+ Mips_address addend_a, bool extract_addend, bool local,
+ unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+
+ Valtype32 addend;
+ if (extract_addend)
+ {
+ if (r_type == elfcpp::R_MICROMIPS_GPREL7_S2)
+ addend = (val & 0x7f) << 2;
+ else
+ addend = val & 0xffff;
+ // Only sign-extend the addend if it was extracted from the
+ // instruction. If the addend was separate, leave it alone,
+ // otherwise we may lose significant bits.
+ addend = Bits<16>::sign_extend32(addend);
+ }
+ else
+ addend = addend_a;
+
+ Valtype32 x = psymval->value(object, addend) - gp;
+
+ // If the symbol was local, any earlier relocatable links will
+ // have adjusted its addend with the gp offset, so compensate
+ // for that now. Don't do it for symbols forced local in this
+ // link, though, since they won't have had the gp offset applied
+ // to them before.
+ if (local)
+ x += object->gp_value();
+
+ if (r_type == elfcpp::R_MICROMIPS_GPREL7_S2)
+ val = Bits<32>::bit_select32(val, x, 0x7f);
+ else
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ if (Bits<16>::has_overflow32(x))
+ {
+ gold_error(_("small-data section exceeds 64KB; lower small-data size "
+ "limit (see option -G)"));
+ return This::STATUS_OVERFLOW;
+ }
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_GPREL32
+ static inline typename This::Status
+ relgprel32(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address gp,
+ Mips_address addend_a, bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 addend = extract_addend ? val : addend_a;
+
+ // R_MIPS_GPREL32 relocations are defined for local symbols only.
+ Valtype32 x = psymval->value(object, addend) + object->gp_value() - gp;
+ elfcpp::Swap<32, big_endian>::writeval(wv, x);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_TLS_TPREL_HI16, R_MIPS16_TLS_TPREL_HI16, R_MICROMIPS_TLS_TPREL_HI16
+ // R_MIPS_TLS_DTPREL_HI16, R_MIPS16_TLS_DTPREL_HI16,
+ // R_MICROMIPS_TLS_DTPREL_HI16
+ static inline typename This::Status
+ tlsrelhi16(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Valtype32 tp_offset,
+ Mips_address addend_a, bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 addend = extract_addend ? val & 0xffff : addend_a;
+
+ // tls symbol values are relative to tls_segment()->vaddr()
+ Valtype32 x = ((psymval->value(object, addend) - tp_offset) + 0x8000) >> 16;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_TLS_TPREL_LO16, R_MIPS16_TLS_TPREL_LO16, R_MICROMIPS_TLS_TPREL_LO16,
+ // R_MIPS_TLS_DTPREL_LO16, R_MIPS16_TLS_DTPREL_LO16,
+ // R_MICROMIPS_TLS_DTPREL_LO16,
+ static inline typename This::Status
+ tlsrello16(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Valtype32 tp_offset,
+ Mips_address addend_a, bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 addend = extract_addend ? val & 0xffff : addend_a;
+
+ // tls symbol values are relative to tls_segment()->vaddr()
+ Valtype32 x = psymval->value(object, addend) - tp_offset;
+ val = Bits<32>::bit_select32(val, x, 0xffff);
+ elfcpp::Swap<32, big_endian>::writeval(wv, val);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_TLS_TPREL32, R_MIPS_TLS_TPREL64,
+ // R_MIPS_TLS_DTPREL32, R_MIPS_TLS_DTPREL64
+ static inline typename This::Status
+ tlsrel32(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Valtype32 tp_offset,
+ Mips_address addend_a, bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 addend = extract_addend ? val : addend_a;
+
+ // tls symbol values are relative to tls_segment()->vaddr()
+ Valtype32 x = psymval->value(object, addend) - tp_offset;
+ elfcpp::Swap<32, big_endian>::writeval(wv, x);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+
+ // R_MIPS_SUB, R_MICROMIPS_SUB
+ static inline typename This::Status
+ relsub(unsigned char* view, const Mips_relobj<size, big_endian>* object,
+ const Symbol_value<size>* psymval, Mips_address addend_a,
+ bool extract_addend, unsigned int r_type)
+ {
+ mips_reloc_unshuffle(view, r_type, false);
+ Valtype32* wv = reinterpret_cast<Valtype32*>(view);
+ Valtype32 val = elfcpp::Swap<32, big_endian>::readval(wv);
+ Valtype32 addend = extract_addend ? val : addend_a;
+
+ Valtype32 x = psymval->value(object, -addend);
+ elfcpp::Swap<32, big_endian>::writeval(wv, x);
+ mips_reloc_shuffle(view, r_type, false);
+ return This::STATUS_OKAY;
+ }
+};
+
+template<int size, bool big_endian>
+typename std::list<reloc_high<size, big_endian> >
+ Mips_relocate_functions<size, big_endian>::hi16_relocs;
+
+template<int size, bool big_endian>
+typename std::list<reloc_high<size, big_endian> >
+ Mips_relocate_functions<size, big_endian>::got16_relocs;
+
+// Mips_got_info methods.
+
+// Reserve GOT entry for a GOT relocation of type R_TYPE against symbol
+// SYMNDX + ADDEND, where SYMNDX is a local symbol in section SHNDX in OBJECT.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::record_local_got_symbol(
+ Mips_relobj<size, big_endian>* object, unsigned int symndx,
+ Mips_address addend, unsigned int r_type, unsigned int shndx)
+{
+ Mips_got_entry<size, big_endian>* entry =
+ new Mips_got_entry<size, big_endian>(object, symndx, addend,
+ mips_elf_reloc_tls_type(r_type),
+ shndx);
+ this->record_got_entry(entry, object);
+}
+
+// Reserve GOT entry for a GOT relocation of type R_TYPE against MIPS_SYM,
+// in OBJECT. FOR_CALL is true if the caller is only interested in
+// using the GOT entry for calls. DYN_RELOC is true if R_TYPE is a dynamic
+// relocation.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::record_global_got_symbol(
+ Mips_symbol<size>* mips_sym, Mips_relobj<size, big_endian>* object,
+ unsigned int r_type, bool dyn_reloc, bool for_call)
+{
+ if (!for_call)
+ mips_sym->set_got_not_only_for_calls();
+
+ // A global symbol in the GOT must also be in the dynamic symbol table.
+ if (!mips_sym->needs_dynsym_entry())
+ {
+ switch (mips_sym->visibility())
+ {
+ case elfcpp::STV_INTERNAL:
+ case elfcpp::STV_HIDDEN:
+ mips_sym->set_is_forced_local();
+ break;
+ default:
+ mips_sym->set_needs_dynsym_entry();
+ break;
+ }
+ }
+
+ unsigned char tls_type = mips_elf_reloc_tls_type(r_type);
+ if (tls_type == GOT_TLS_NONE)
+ this->global_got_symbols_.insert(mips_sym);
+
+ if (dyn_reloc)
+ {
+ if (mips_sym->global_got_area() == GGA_NONE)
+ mips_sym->set_global_got_area(GGA_RELOC_ONLY);
+ return;
+ }
+
+ Mips_got_entry<size, big_endian>* entry =
+ new Mips_got_entry<size, big_endian>(object, mips_sym, tls_type);
+
+ this->record_got_entry(entry, object);
+}
+
+// Add ENTRY to master GOT and to OBJECT's GOT.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::record_got_entry(
+ Mips_got_entry<size, big_endian>* entry,
+ Mips_relobj<size, big_endian>* object)
+{
+ if (this->got_entries_.find(entry) == this->got_entries_.end())
+ this->got_entries_.insert(entry);
+
+ // Create the GOT entry for the OBJECT's GOT.
+ Mips_got_info<size, big_endian>* g = object->get_or_create_got_info();
+ Mips_got_entry<size, big_endian>* entry2 =
+ new Mips_got_entry<size, big_endian>(*entry);
+
+ if (g->got_entries_.find(entry2) == g->got_entries_.end())
+ g->got_entries_.insert(entry2);
+}
+
+// Record that OBJECT has a page relocation against symbol SYMNDX and
+// that ADDEND is the addend for that relocation.
+// This function creates an upper bound on the number of GOT slots
+// required; no attempt is made to combine references to non-overridable
+// global symbols across multiple input files.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::record_got_page_entry(
+ Mips_relobj<size, big_endian>* object, unsigned int symndx, int addend)
+{
+ struct Got_page_range **range_ptr, *range;
+ int old_pages, new_pages;
+
+ // Find the Got_page_entry for this symbol.
+ Got_page_entry* entry = new Got_page_entry(object, symndx);
+ typename Got_page_entry_set::iterator it =
+ this->got_page_entries_.find(entry);
+ if (it != this->got_page_entries_.end())
+ entry = *it;
+ else
+ this->got_page_entries_.insert(entry);
+
+ // Add the same entry to the OBJECT's GOT.
+ Got_page_entry* entry2 = NULL;
+ Mips_got_info<size, big_endian>* g2 = object->get_or_create_got_info();
+ if (g2->got_page_entries_.find(entry) == g2->got_page_entries_.end())
+ {
+ entry2 = new Got_page_entry(*entry);
+ g2->got_page_entries_.insert(entry2);
+ }
+
+ // Skip over ranges whose maximum extent cannot share a page entry
+ // with ADDEND.
+ range_ptr = &entry->ranges;
+ while (*range_ptr && addend > (*range_ptr)->max_addend + 0xffff)
+ range_ptr = &(*range_ptr)->next;
+
+ // If we scanned to the end of the list, or found a range whose
+ // minimum extent cannot share a page entry with ADDEND, create
+ // a new singleton range.
+ range = *range_ptr;
+ if (!range || addend < range->min_addend - 0xffff)
+ {
+ range = new Got_page_range();
+ range->next = *range_ptr;
+ range->min_addend = addend;
+ range->max_addend = addend;
+
+ *range_ptr = range;
+ ++entry->num_pages;
+ if (entry2 != NULL)
+ ++entry2->num_pages;
+ ++this->page_gotno_;
+ ++g2->page_gotno_;
+ return;
+ }
+
+ // Remember how many pages the old range contributed.
+ old_pages = range->get_max_pages();
+
+ // Update the ranges.
+ if (addend < range->min_addend)
+ range->min_addend = addend;
+ else if (addend > range->max_addend)
+ {
+ if (range->next && addend >= range->next->min_addend - 0xffff)
+ {
+ old_pages += range->next->get_max_pages();
+ range->max_addend = range->next->max_addend;
+ range->next = range->next->next;
+ }
+ else
+ range->max_addend = addend;
+ }
+
+ // Record any change in the total estimate.
+ new_pages = range->get_max_pages();
+ if (old_pages != new_pages)
+ {
+ entry->num_pages += new_pages - old_pages;
+ if (entry2 != NULL)
+ entry2->num_pages += new_pages - old_pages;
+ this->page_gotno_ += new_pages - old_pages;
+ g2->page_gotno_ += new_pages - old_pages;
+ }
+}
+
+// Create all entries that should be in the local part of the GOT.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_local_entries(
+ Target_mips<size, big_endian>* target, Layout* layout)
+{
+ Mips_output_data_got<size, big_endian>* got = target->got_section();
+ // First two GOT entries are reserved. The first entry will be filled at
+ // runtime. The second entry will be used by some runtime loaders.
+ got->add_constant(0);
+ got->add_constant(target->mips_elf_gnu_got1_mask());
+
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (entry->is_for_local_symbol() && !entry->is_tls_entry())
+ {
+ got->add_local(entry->object(), entry->symndx(),
+ GOT_TYPE_STANDARD);
+ unsigned int got_offset = entry->object()->local_got_offset(
+ entry->symndx(), GOT_TYPE_STANDARD);
+ if (got->multi_got() && this->index_ > 0
+ && parameters->options().output_is_position_independent())
+ target->rel_dyn_section(layout)->add_local(entry->object(),
+ entry->symndx(), elfcpp::R_MIPS_REL32, got, got_offset);
+ }
+ }
+
+ this->add_page_entries(target, layout);
+
+ // Add global entries that should be in the local area.
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (!entry->is_for_global_symbol())
+ continue;
+
+ Mips_symbol<size>* mips_sym = entry->sym();
+ if (mips_sym->global_got_area() == GGA_NONE && !entry->is_tls_entry())
+ {
+ unsigned int got_type;
+ if (!got->multi_got())
+ got_type = GOT_TYPE_STANDARD;
+ else
+ got_type = GOT_TYPE_STANDARD_MULTIGOT + this->index_;
+ if (got->add_global(mips_sym, got_type))
+ {
+ mips_sym->set_global_gotoffset(mips_sym->got_offset(got_type));
+ if (got->multi_got() && this->index_ > 0
+ && parameters->options().output_is_position_independent())
+ target->rel_dyn_section(layout)->add_symbolless_global_addend(
+ mips_sym, elfcpp::R_MIPS_REL32, got,
+ mips_sym->got_offset(got_type));
+ }
+ }
+ }
+}
+
+// Create GOT page entries.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_page_entries(
+ Target_mips<size, big_endian>* target, Layout* layout)
+{
+ if (this->page_gotno_ == 0)
+ return;
+
+ Mips_output_data_got<size, big_endian>* got = target->got_section();
+ this->got_page_offset_start_ = got->add_constant(0);
+ if (got->multi_got() && this->index_ > 0
+ && parameters->options().output_is_position_independent())
+ target->rel_dyn_section(layout)->add_absolute(elfcpp::R_MIPS_REL32, got,
+ this->got_page_offset_start_);
+ int num_entries = this->page_gotno_;
+ unsigned int prev_offset = this->got_page_offset_start_;
+ while (--num_entries > 0)
+ {
+ unsigned int next_offset = got->add_constant(0);
+ if (got->multi_got() && this->index_ > 0
+ && parameters->options().output_is_position_independent())
+ target->rel_dyn_section(layout)->add_absolute(elfcpp::R_MIPS_REL32, got,
+ next_offset);
+ gold_assert(next_offset == prev_offset + size/8);
+ prev_offset = next_offset;
+ }
+ this->got_page_offset_next_ = this->got_page_offset_start_;
+}
+
+// Create global GOT entries, both GGA_NORMAL and GGA_RELOC_ONLY.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_global_entries(
+ Target_mips<size, big_endian>* target, Layout* layout,
+ unsigned int non_reloc_only_global_gotno)
+{
+ Mips_output_data_got<size, big_endian>* got = target->got_section();
+ // Add GGA_NORMAL entries.
+ unsigned int count = 0;
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (!entry->is_for_global_symbol())
+ continue;
+
+ Mips_symbol<size>* mips_sym = entry->sym();
+ if (mips_sym->global_got_area() != GGA_NORMAL)
+ continue;
+
+ unsigned int got_type;
+ if (!got->multi_got())
+ got_type = GOT_TYPE_STANDARD;
+ else
+ // In multi-GOT links, global symbol can be in both primary and
+ // secondary GOT(s). By creating custom GOT type
+ // (GOT_TYPE_STANDARD_MULTIGOT + got_index) we ensure that symbol
+ // is added to secondary GOT(s).
+ got_type = GOT_TYPE_STANDARD_MULTIGOT + this->index_;
+ if (!got->add_global(mips_sym, got_type))
+ continue;
+
+ mips_sym->set_global_gotoffset(mips_sym->got_offset(got_type));
+ if (got->multi_got() && this->index_ == 0)
+ count++;
+ if (got->multi_got() && this->index_ > 0)
+ {
+ if (parameters->options().output_is_position_independent()
+ || (!parameters->doing_static_link()
+ && mips_sym->is_from_dynobj() && !mips_sym->is_undefined()))
+ {
+ target->rel_dyn_section(layout)->add_global(
+ mips_sym, elfcpp::R_MIPS_REL32, got,
+ mips_sym->got_offset(got_type));
+ got->add_secondary_got_reloc(mips_sym->got_offset(got_type),
+ elfcpp::R_MIPS_REL32, mips_sym);
+ }
+ }
+ }
+
+ if (!got->multi_got() || this->index_ == 0)
+ {
+ if (got->multi_got())
+ {
+ // We need to allocate space in the primary GOT for GGA_NORMAL entries
+ // of secondary GOTs, to ensure that GOT offsets of GGA_RELOC_ONLY
+ // entries correspond to dynamic symbol indexes.
+ while (count < non_reloc_only_global_gotno)
+ {
+ got->add_constant(0);
+ ++count;
+ }
+ }
+
+ // Add GGA_RELOC_ONLY entries.
+ got->add_reloc_only_entries();
+ }
+}
+
+// Create global GOT entries that should be in the GGA_RELOC_ONLY area.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_reloc_only_entries(
+ Mips_output_data_got<size, big_endian>* got)
+{
+ for (typename Unordered_set<Mips_symbol<size>*>::iterator
+ p = this->global_got_symbols_.begin();
+ p != this->global_got_symbols_.end();
+ ++p)
+ {
+ Mips_symbol<size>* mips_sym = *p;
+ if (mips_sym->global_got_area() == GGA_RELOC_ONLY)
+ {
+ unsigned int got_type;
+ if (!got->multi_got())
+ got_type = GOT_TYPE_STANDARD;
+ else
+ got_type = GOT_TYPE_STANDARD_MULTIGOT;
+ if (got->add_global(mips_sym, got_type))
+ mips_sym->set_global_gotoffset(mips_sym->got_offset(got_type));
+ }
+ }
+}
+
+// Create TLS GOT entries.
+
+template<int size, bool big_endian>
+void
+Mips_got_info<size, big_endian>::add_tls_entries(
+ Target_mips<size, big_endian>* target, Layout* layout)
+{
+ Mips_output_data_got<size, big_endian>* got = target->got_section();
+ // Add local tls entries.
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (!entry->is_tls_entry() || !entry->is_for_local_symbol())
+ continue;
+
+ if (entry->tls_type() == GOT_TLS_GD)
+ {
+ unsigned int got_type = GOT_TYPE_TLS_PAIR;
+ unsigned int r_type1 = (size == 32 ? elfcpp::R_MIPS_TLS_DTPMOD32
+ : elfcpp::R_MIPS_TLS_DTPMOD64);
+ unsigned int r_type2 = (size == 32 ? elfcpp::R_MIPS_TLS_DTPREL32
+ : elfcpp::R_MIPS_TLS_DTPREL64);
+
+ if (!parameters->doing_static_link())
+ {
+ got->add_local_pair_with_rel(entry->object(), entry->symndx(),
+ entry->shndx(), got_type,
+ target->rel_dyn_section(layout),
+ r_type1);
+ unsigned int got_offset =
+ entry->object()->local_got_offset(entry->symndx(), got_type);
+ got->add_static_reloc(got_offset + size/8, r_type2,
+ entry->object(), entry->symndx());
+ }
+ else
+ {
+ // We are doing a static link. Mark it as belong to module 1,
+ // the executable.
+ unsigned int got_offset = got->add_constant(1);
+ entry->object()->set_local_got_offset(entry->symndx(), got_type,
+ got_offset);
+ got->add_constant(0);
+ got->add_static_reloc(got_offset + size/8, r_type2,
+ entry->object(), entry->symndx());
+ }
+ }
+ else if (entry->tls_type() == GOT_TLS_IE)
+ {
+ unsigned int got_type = GOT_TYPE_TLS_OFFSET;
+ unsigned int r_type = (size == 32 ? elfcpp::R_MIPS_TLS_TPREL32
+ : elfcpp::R_MIPS_TLS_TPREL64);
+ if (!parameters->doing_static_link())
+ got->add_local_with_rel(entry->object(), entry->symndx(), got_type,
+ target->rel_dyn_section(layout), r_type);
+ else
+ {
+ got->add_local(entry->object(), entry->symndx(), got_type);
+ unsigned int got_offset =
+ entry->object()->local_got_offset(entry->symndx(), got_type);
+ got->add_static_reloc(got_offset, r_type, entry->object(),
+ entry->symndx());
+ }
+ }
+ else if (entry->tls_type() == GOT_TLS_LDM)
+ {
+ unsigned int r_type = (size == 32 ? elfcpp::R_MIPS_TLS_DTPMOD32
+ : elfcpp::R_MIPS_TLS_DTPMOD64);
+ unsigned int got_offset;
+ if (!parameters->doing_static_link())
+ {
+ got_offset = got->add_constant(0);
+ target->rel_dyn_section(layout)->add_local(
+ entry->object(), 0, r_type, got, got_offset);
+ }
+ else
+ // We are doing a static link. Just mark it as belong to module 1,
+ // the executable.
+ got_offset = got->add_constant(1);
+
+ got->add_constant(0);
+ got->set_tls_ldm_offset(got_offset, entry->object());
+ }
+ else
+ gold_unreachable();
+ }
+
+ // Add global tls entries.
+ for (typename Got_entry_set::iterator
+ p = this->got_entries_.begin();
+ p != this->got_entries_.end();
+ ++p)
+ {
+ Mips_got_entry<size, big_endian>* entry = *p;
+ if (!entry->is_tls_entry() || !entry->is_for_global_symbol())
+ continue;
+
+ Mips_symbol<size>* mips_sym = entry->sym();
+ if (entry->tls_type() == GOT_TLS_GD)
+ {
+ unsigned int got_type;
+ if (!got->multi_got())
+ got_type = GOT_TYPE_TLS_PAIR;
+ else
+ got_type = GOT_TYPE_TLS_PAIR_MULTIGOT + this->index_;
+ unsigned int r_type1 = (size == 32 ? elfcpp::R_MIPS_TLS_DTPMOD32
+ : elfcpp::R_MIPS_TLS_DTPMOD64);
+ unsigned int r_type2 = (size == 32 ? elfcpp::R_MIPS_TLS_DTPREL32
+ : elfcpp::R_MIPS_TLS_DTPREL64);
+ if (!parameters->doing_static_link())
+ got->add_global_pair_with_rel(mips_sym, got_type,
+ target->rel_dyn_section(layout), r_type1, r_type2);
+ else
+ {
+ // Add a GOT pair for for R_MIPS_TLS_GD. The creates a pair of
+ // GOT entries. The first one is initialized to be 1, which is the
+ // module index for the main executable and the second one 0. A
+ // reloc of the type R_MIPS_TLS_DTPREL32/64 will be created for
+ // the second GOT entry and will be applied by gold.
+ unsigned int got_offset = got->add_constant(1);
+ mips_sym->set_got_offset(got_type, got_offset);
+ got->add_constant(0);
+ got->add_static_reloc(got_offset + size/8, r_type2, mips_sym);
+ }
+ }
+ else if (entry->tls_type() == GOT_TLS_IE)
+ {
+ unsigned int got_type;
+ if (!got->multi_got())
+ got_type = GOT_TYPE_TLS_OFFSET;
+ else
+ got_type = GOT_TYPE_TLS_OFFSET_MULTIGOT + this->index_;
+ unsigned int r_type = (size == 32 ? elfcpp::R_MIPS_TLS_TPREL32
+ : elfcpp::R_MIPS_TLS_TPREL64);
+ if (!parameters->doing_static_link())
+ got->add_global_with_rel(mips_sym, got_type,
+ target->rel_dyn_section(layout), r_type);
+ else
+ {
+ got->add_global(mips_sym, got_type);
+ unsigned int got_offset = mips_sym->got_offset(got_type);
+ got->add_static_reloc(got_offset, r_type, mips_sym);
+ }
+ }
+ else
+ gold_unreachable();
+ }
+}
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Mips target in gold - revision 3 - part 2
2014-06-04 21:32 ` Sasa Stankovic
@ 2014-06-23 19:10 ` Cary Coutant
2014-06-26 2:14 ` Alan Modra
0 siblings, 1 reply; 9+ messages in thread
From: Cary Coutant @ 2014-06-23 19:10 UTC (permalink / raw)
To: Sasa Stankovic; +Cc: Rich Fuhler, binutils, Petar Jovanovic, Heiher
I've committed these changes on your behalf.
I also included a small fix to configure.tgt (thanks to
admin@heiher.info), which moves the catch-all "mips*-*-*" case below
the more specific "mips*el*-*-*|mips*le*-*-*" case.
Gold now support MIPS! Thanks for all the work on this!
-cary
2014-06-23 Sasa Stankovic <Sasa.Stankovic@imgtec.com>
* mips.cc: New file.
* Makefile.am (TARGETSOURCES): Add mips.cc
(ALL_TARGETOBJS): Add mips.$(OBJEXT)
* configure.tgt: Add entries for mips*.
* configure.ac: Likewise.
* Makefile.in: Regenerate.
* configure: Likewise.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Mips target in gold - revision 3 - part 2
2014-06-23 19:10 ` Cary Coutant
@ 2014-06-26 2:14 ` Alan Modra
2014-06-26 16:23 ` Cary Coutant
0 siblings, 1 reply; 9+ messages in thread
From: Alan Modra @ 2014-06-26 2:14 UTC (permalink / raw)
To: Cary Coutant
Cc: Sasa Stankovic, Rich Fuhler, binutils, Petar Jovanovic, Heiher
On Mon, Jun 23, 2014 at 12:10:10PM -0700, Cary Coutant wrote:
> I've committed these changes on your behalf.
>
> I also included a small fix to configure.tgt (thanks to
> admin@heiher.info), which moves the catch-all "mips*-*-*" case below
> the more specific "mips*el*-*-*|mips*le*-*-*" case.
>
> Gold now support MIPS! Thanks for all the work on this!
mips.o: In function `emit_entry':
/src/binutils-gdb/gold/mips.cc:7162: undefined reference to `gold::Symbol::should_add_dynsym_entry(gold::Symbol_table*) const'
mips.o: In function `emit':
/src/binutils-gdb/gold/mips.cc:3461: undefined reference to `gold::Symbol::should_add_dynsym_entry(gold::Symbol_table*) const'
mips.o: In function `count_got_symbols':
/src/binutils-gdb/gold/mips.cc:5275: undefined reference to `gold::Symbol::should_add_dynsym_entry(gold::Symbol_table*) const'
/src/binutils-gdb/gold/mips.cc:5275: undefined reference to `gold::Symbol::should_add_dynsym_entry(gold::Symbol_table*) const'
/src/binutils-gdb/gold/mips.cc:5275: undefined reference to `gold::Symbol::should_add_dynsym_entry(gold::Symbol_table*) const'
mips.o:/src/binutils-gdb/gold/mips.cc:5977: more undefined references to `gold::Symbol::should_add_dynsym_entry(gold::Symbol_table*) const' follow
OK to apply?
* symtab.cc (Symbol::should_add_dynsym_entry): Don't make inline.
diff --git a/gold/symtab.cc b/gold/symtab.cc
index 4e8afb1..b329bb6 100644
--- a/gold/symtab.cc
+++ b/gold/symtab.cc
@@ -334,7 +334,7 @@ Sized_symbol<size>::allocate_common(Output_data* od, Value_type value)
// Return true if this symbol should be added to the dynamic symbol
// table.
-inline bool
+bool
Symbol::should_add_dynsym_entry(Symbol_table* symtab) const
{
// If the symbol is only present on plugin files, the plugin decided we
--
Alan Modra
Australia Development Lab, IBM
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Mips target in gold - revision 3 - part 2
2014-06-26 2:14 ` Alan Modra
@ 2014-06-26 16:23 ` Cary Coutant
0 siblings, 0 replies; 9+ messages in thread
From: Cary Coutant @ 2014-06-26 16:23 UTC (permalink / raw)
To: Cary Coutant, Sasa Stankovic, Rich Fuhler, binutils,
Petar Jovanovic, Heiher
> OK to apply?
>
> * symtab.cc (Symbol::should_add_dynsym_entry): Don't make inline.
Yes, thanks!
-cary
^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2014-06-26 16:23 UTC | newest]
Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-04-15 13:53 Mips target in gold - revision 3 - part 2 Sasa Stankovic
2014-04-15 20:46 ` Cary Coutant
2014-04-15 21:08 ` Cary Coutant
2014-06-04 21:01 ` Sasa Stankovic
2014-06-04 21:32 ` Sasa Stankovic
2014-06-23 19:10 ` Cary Coutant
2014-06-26 2:14 ` Alan Modra
2014-06-26 16:23 ` Cary Coutant
2014-06-04 21:03 ` Sasa Stankovic
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).