* RFC: Add DWARF 2 support for section group
@ 2004-04-29 0:49 H. J. Lu
2004-04-29 2:33 ` Zack Weinberg
0 siblings, 1 reply; 4+ messages in thread
From: H. J. Lu @ 2004-04-29 0:49 UTC (permalink / raw)
To: binutils; +Cc: GDB
The current ELF linker doesn't support DWARF for section group. We
have
/* DWARF 1.1 and DWARF 2 */
.debug_aranges 0 : { *(.debug_aranges) }
.debug_pubnames 0 : { *(.debug_pubnames) }
/* DWARF 2 */
.debug_info 0 : { *(.debug_info .gnu.linkonce.wi.*) }
.debug_abbrev 0 : { *(.debug_abbrev) }
.debug_line 0 : { *(.debug_line) }
.debug_frame 0 : { *(.debug_frame) }
.debug_str 0 : { *(.debug_str) }
.debug_loc 0 : { *(.debug_loc) }
.debug_macinfo 0 : { *(.debug_macinfo) }
For each section, it can be in a section group. I am thinking
something like. Also we need to support exception for section group.
Do we need something like
--- scripttempl/elf.sc.debug 2004-04-22 08:22:22.000000000 -0700
+++ scripttempl/elf.sc 2004-04-28 17:30:38.000000000 -0700
@@ -283,7 +283,7 @@ cat <<EOF
${CREATE_SHLIB-${SDATA2}}
${CREATE_SHLIB-${SBSS2}}
${OTHER_READONLY_SECTIONS}
- .eh_frame_hdr : { *(.eh_frame_hdr) }
+ .eh_frame_hdr : { *(.eh_frame_hdr .gnu.linkonce.efhdr.*) }
/* Adjust the address for the data segment. We want to adjust up to
the same address within the page on the next page up. */
@@ -318,7 +318,7 @@ cat <<EOF
.data1 ${RELOCATING-0} : { *(.data1) }
.tdata ${RELOCATING-0} : { *(.tdata${RELOCATING+ .tdata.* .gnu.linkonce.td.*}) }
.tbss ${RELOCATING-0} : { *(.tbss${RELOCATING+ .tbss.* .gnu.linkonce.tb.*})${RELOCATING+ *(.tcommon)} }
- .eh_frame ${RELOCATING-0} : { KEEP (*(.eh_frame)) }
+ .eh_frame ${RELOCATING-0} : { KEEP (*(.eh_frame .gnu.linkonce.ef.*)) }
.gcc_except_table ${RELOCATING-0} : { KEEP (*(.gcc_except_table)) *(.gcc_except_table.*) }
${WRITABLE_RODATA+${RODATA}}
${OTHER_READWRITE_SECTIONS}
@@ -380,17 +380,17 @@ cat <<EOF
.debug_sfnames 0 : { *(.debug_sfnames) }
/* DWARF 1.1 and DWARF 2 */
- .debug_aranges 0 : { *(.debug_aranges) }
- .debug_pubnames 0 : { *(.debug_pubnames) }
+ .debug_aranges 0 : { *(.debug_aranges .gnu.linkonce.wa.*) }
+ .debug_pubnames 0 : { *(.debug_pubnames .gnu.linkonce.wp.*) }
/* DWARF 2 */
.debug_info 0 : { *(.debug_info .gnu.linkonce.wi.*) }
- .debug_abbrev 0 : { *(.debug_abbrev) }
- .debug_line 0 : { *(.debug_line) }
- .debug_frame 0 : { *(.debug_frame) }
- .debug_str 0 : { *(.debug_str) }
- .debug_loc 0 : { *(.debug_loc) }
- .debug_macinfo 0 : { *(.debug_macinfo) }
+ .debug_abbrev 0 : { *(.debug_abbrev .gnu.linkonce.wA.*) }
+ .debug_line 0 : { *(.debug_line .gnu.linkonce.wl.*) }
+ .debug_frame 0 : { *(.debug_frame .gnu.linkonce.wf.*) }
+ .debug_str 0 : { *(.debug_str .gnu.linkonce.ws.*) }
+ .debug_loc 0 : { *(.debug_loc .gnu.linkonce.wL.*) }
+ .debug_macinfo 0 : { *(.debug_macinfo .gnu.linkonce.wm.*) }
/* SGI/MIPS DWARF 2 extensions */
.debug_weaknames 0 : { *(.debug_weaknames) }
H.J.
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: RFC: Add DWARF 2 support for section group
2004-04-29 0:49 RFC: Add DWARF 2 support for section group H. J. Lu
@ 2004-04-29 2:33 ` Zack Weinberg
2004-04-29 15:57 ` PATCH: Support multiple sections with same name H. J. Lu
0 siblings, 1 reply; 4+ messages in thread
From: Zack Weinberg @ 2004-04-29 2:33 UTC (permalink / raw)
To: H. J. Lu; +Cc: binutils, GDB
"H. J. Lu" <hjl@lucon.org> writes:
> The current ELF linker doesn't support DWARF for section group.
[...]
Barf. Why are we continuing to extend the .gnu.linkonce hack, when we
should be using proper ELF comdat groups?
zw
^ permalink raw reply [flat|nested] 4+ messages in thread
* PATCH: Support multiple sections with same name
2004-04-29 2:33 ` Zack Weinberg
@ 2004-04-29 15:57 ` H. J. Lu
2004-04-30 14:17 ` Alan Modra
0 siblings, 1 reply; 4+ messages in thread
From: H. J. Lu @ 2004-04-29 15:57 UTC (permalink / raw)
To: Zack Weinberg; +Cc: binutils
[-- Attachment #1: Type: text/plain, Size: 410 bytes --]
On Wed, Apr 28, 2004 at 05:49:10PM -0700, Zack Weinberg wrote:
> "H. J. Lu" <hjl@lucon.org> writes:
>
> > The current ELF linker doesn't support DWARF for section group.
> [...]
>
> Barf. Why are we continuing to extend the .gnu.linkonce hack, when we
> should be using proper ELF comdat groups?
>
I was wrong to say that gas supported multiple sections with same name.
This patch seems to fix it.
H.J.
[-- Attachment #2: gas-elf-dup-1.patch --]
[-- Type: text/plain, Size: 6766 bytes --]
gas/
2004-04-28 H.J. Lu <hongjiu.lu@intel.com>
* config/obj-elf.c (get_section): New function.
(obj_elf_change_section): Support multiple sections with same
name.
gas/testsuite/
2004-04-28 H.J. Lu <hongjiu.lu@intel.com>
* gas/elf/elf.exp: Remove group1, add group1a and group1b for
section group.
* gas/elf/group1a.d: New file.
* gas/elf/group1b.d: Likewise.
* gas/elf/group1.e: Removed.
--- gas/config/obj-elf.c.same 2004-04-26 21:12:45.000000000 -0700
+++ gas/config/obj-elf.c 2004-04-28 23:17:01.000000000 -0700
@@ -469,6 +469,34 @@ struct section_stack
static struct section_stack *section_stack;
+struct section_group
+{
+ const char *name;
+ const char *group_name;
+ asection *section;
+};
+
+static void
+get_section (bfd *abfd ATTRIBUTE_UNUSED, asection *sec, void *inf)
+{
+ struct section_group *group = inf;
+ const char *group_name = elf_group_name (sec);
+
+ /* Check if we have found the section we are looking for. */
+ if (group->section)
+ return;
+
+ if ((sec->name == group->name
+ || (sec->name != NULL
+ && group->name != NULL
+ && strcmp (sec->name, group->name) == 0))
+ && (group_name == group->group_name
+ || (group_name != NULL
+ && group->group_name != NULL
+ && strcmp (group_name, group->group_name) == 0)))
+ group->section = sec;
+}
+
/* Handle the .section pseudo-op. This code supports two different
syntaxes.
@@ -496,10 +524,10 @@ obj_elf_change_section (const char *name
int linkonce,
int push)
{
- asection *old_sec;
segT sec;
flagword flags;
const struct bfd_elf_special_section *ssect;
+ struct section_group group;
#ifdef md_flush_pending_output
md_flush_pending_output ();
@@ -520,8 +548,19 @@ obj_elf_change_section (const char *name
previous_section = now_seg;
previous_subsection = now_subseg;
- old_sec = bfd_get_section_by_name (stdoutput, name);
- sec = subseg_new (name, 0);
+ group.name = name;
+ group.group_name = group_name;
+ group.section = NULL;
+ bfd_map_over_sections (stdoutput, get_section, &group);
+
+ if (group.section)
+ {
+ sec = group.section;
+ subseg_set (sec, 0);
+ }
+ else
+ sec = subseg_force_new (name, 0);
+
ssect = _bfd_elf_get_sec_type_attr (stdoutput, name);
if (ssect != NULL)
@@ -532,7 +571,7 @@ obj_elf_change_section (const char *name
type = ssect->type;
else if (type != ssect->type)
{
- if (old_sec == NULL
+ if (group.section == NULL
/* FIXME: gcc, as of 2002-10-22, will emit
.section .init_array,"aw",@progbits
@@ -556,7 +595,7 @@ obj_elf_change_section (const char *name
}
}
- if (old_sec == NULL && (attr & ~ssect->attr) != 0)
+ if (group.section == NULL && (attr & ~ssect->attr) != 0)
{
/* As a GNU extension, we permit a .note section to be
allocatable. If the linker sees an allocatable .note
@@ -582,12 +621,13 @@ obj_elf_change_section (const char *name
override = TRUE;
else
{
- as_warn (_("setting incorrect section attributes for %s"),
- name);
+ if (group_name == NULL)
+ as_warn (_("setting incorrect section attributes for %s"),
+ name);
override = TRUE;
}
}
- if (!override && old_sec == NULL)
+ if (!override && group.section == NULL)
attr |= ssect->attr;
}
@@ -609,7 +649,10 @@ obj_elf_change_section (const char *name
flags = md_elf_section_flags (flags, attr, type);
#endif
- if (old_sec == NULL)
+ if (linkonce)
+ flags |= SEC_LINK_ONCE | SEC_LINK_DUPLICATES_DISCARD;
+
+ if (group.section == NULL)
{
symbolS *secsym;
@@ -617,8 +660,6 @@ obj_elf_change_section (const char *name
if (type == SHT_NOBITS)
seg_info (sec)->bss = 1;
- if (linkonce)
- flags |= SEC_LINK_ONCE | SEC_LINK_DUPLICATES_DISCARD;
bfd_set_section_flags (stdoutput, sec, flags);
if (flags & SEC_MERGE)
sec->entsize = entsize;
@@ -636,18 +677,15 @@ obj_elf_change_section (const char *name
/* If section attributes are specified the second time we see a
particular section, then check that they are the same as we
saw the first time. */
- if (((old_sec->flags ^ flags)
+ if (((group.section->flags ^ flags)
& (SEC_ALLOC | SEC_LOAD | SEC_READONLY | SEC_CODE
| SEC_EXCLUDE | SEC_SORT_ENTRIES | SEC_MERGE | SEC_STRINGS
| SEC_LINK_ONCE | SEC_LINK_DUPLICATES_DISCARD
| SEC_THREAD_LOCAL)))
as_warn (_("ignoring changed section attributes for %s"), name);
- if ((flags & SEC_MERGE) && old_sec->entsize != (unsigned) entsize)
+ if ((flags & SEC_MERGE)
+ && group.section->entsize != (unsigned) entsize)
as_warn (_("ignoring changed section entity size for %s"), name);
- if ((attr & SHF_GROUP) != 0
- && (elf_group_name (old_sec) == NULL
- || strcmp (elf_group_name (old_sec), group_name) != 0))
- as_warn (_("ignoring new section group for %s"), name);
}
#ifdef md_elf_section_change_hook
--- gas/testsuite/gas/elf/elf.exp.same 2004-04-26 21:13:06.000000000 -0700
+++ gas/testsuite/gas/elf/elf.exp 2004-04-28 22:56:42.000000000 -0700
@@ -50,7 +50,8 @@ if { ([istarget "*-*-elf*"]
run_dump_test "ehopt0"
run_dump_test "group0a"
run_dump_test "group0b"
- run_list_test "group1" "" "" ""
+ run_dump_test "group1a"
+ run_dump_test "group1b"
run_dump_test "section0"
run_dump_test "section1"
run_list_test "section2" "$target_machine" "-al" ""
--- gas/testsuite/gas/elf/group1.e.same 2004-04-26 21:04:50.000000000 -0700
+++ gas/testsuite/gas/elf/group1.e 2004-04-28 23:18:19.000000000 -0700
@@ -1,7 +0,0 @@
-
-Symbol table '.symtab' contains 4 entries:
- Num: Value[ ]* Size Type Bind Vis Ndx Name
- 0: 0+0 0 NOTYPE LOCAL DEFAULT UND
- 1: 0+0 0 SECTION LOCAL DEFAULT 1
- 2: 0+0 0 SECTION LOCAL DEFAULT 2
- 3: 0+0 0 SECTION LOCAL DEFAULT 3
--- gas/testsuite/gas/elf/group1a.d.same 2004-04-28 22:56:15.000000000 -0700
+++ gas/testsuite/gas/elf/group1a.d 2004-04-28 23:01:39.000000000 -0700
@@ -0,0 +1,10 @@
+#readelf: -SW
+#name: group section with multiple sections of same name
+#source: group1.s
+
+#...
+[ ]*\[.*\][ ]+\.text[ ]+PROGBITS.*[ ]+AX[ ]+.*
+#...
+[ ]*\[.*\][ ]+\.text[ ]+PROGBITS.*[ ]+AXG[ ]+.*
+[ ]*\[.*\][ ]+\.foo_group[ ]+GROUP.*
+#pass
--- gas/testsuite/gas/elf/group1b.d.same 2004-04-28 22:56:18.000000000 -0700
+++ gas/testsuite/gas/elf/group1b.d 2004-04-28 23:02:08.000000000 -0700
@@ -0,0 +1,9 @@
+#readelf: -g
+#name: group section with multiple sections of same name
+#source: group1.s
+
+#...
+COMDAT group section `.foo_group' \[.foo_group\] contains 1 sections:
+[ ]+\[Index\][ ]+Name
+[ ]+\[.*\][ ]+.text
+#pass
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: PATCH: Support multiple sections with same name
2004-04-29 15:57 ` PATCH: Support multiple sections with same name H. J. Lu
@ 2004-04-30 14:17 ` Alan Modra
0 siblings, 0 replies; 4+ messages in thread
From: Alan Modra @ 2004-04-30 14:17 UTC (permalink / raw)
To: H. J. Lu; +Cc: Zack Weinberg, binutils
On Wed, Apr 28, 2004 at 11:25:07PM -0700, H. J. Lu wrote:
> gas/
>
> 2004-04-28 H.J. Lu <hongjiu.lu@intel.com>
>
> * config/obj-elf.c (get_section): New function.
> (obj_elf_change_section): Support multiple sections with same
> name.
>
> gas/testsuite/
>
> 2004-04-28 H.J. Lu <hongjiu.lu@intel.com>
>
> * gas/elf/elf.exp: Remove group1, add group1a and group1b for
> section group.
>
> * gas/elf/group1a.d: New file.
> * gas/elf/group1b.d: Likewise.
>
> * gas/elf/group1.e: Removed.
OK.
--
Alan Modra
IBM OzLabs - Linux Technology Centre
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2004-04-30 12:28 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2004-04-29 0:49 RFC: Add DWARF 2 support for section group H. J. Lu
2004-04-29 2:33 ` Zack Weinberg
2004-04-29 15:57 ` PATCH: Support multiple sections with same name H. J. Lu
2004-04-30 14:17 ` Alan Modra
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).