[Patch v2] Handle .TOC. sections during gdb-compile for rs6000 target.

[Patch v2] Handle .TOC. sections during gdb-compile for rs6000 target.

[will schmidt]

[gdb] Handle .TOC. sections during gdb-compile for rs6000 target.
    
Hi,

[V2]  Multiple changes and cleanups per feedback.  Added logic
to iterate through the sections to find one with SEC_ALLOC flags
set to serve as a substitite for an absent .toc section.
This was/is exercised with the test suite running in a power9 LE
environment.   Per that testing, I don't believe I've got a case
where we had to resort as far as using the bfd_abs_section_ptr
as a fallback, so additional scrutiny may be required for that
portion of this patch.

[V1, updated]

  When we encounter a .TOC. symbol in the object we are loading,
we need to associate this with the .toc section in order to
properly resolve other symbols in the object.  IF a .toc section
is not found, iterate the sections until we find one with the
SEC_ALLOC flag.   If that also fails, fall back to using
bfd_abs_section_ptr.

This is part of a short series of patches that allows gdb-compile
support to function for the ppc64le target.

OK for trunk?

Thanks
-Will
    
    
    gdb/Changelog:
    yyyy-mm-dd  Will Schmidt  <will_schmidt@vnet.ibm.com>
    
            * gdb/compile/compile-object-load.c (compile_object_load):
            Add handling for the .TOC. symbol.

diff --git a/gdb/compile/compile-object-load.c b/gdb/compile/compile-object-load.c
index 1c512801bcd..70742bf9b8e 100644
--- a/gdb/compile/compile-object-load.c
+++ b/gdb/compile/compile-object-load.c
@@ -720,10 +720,52 @@ compile_object_load (const compile_file_names &file_names,
 	     need for _GLOBAL_OFFSET_TABLE_.  Together with -fPIE the data
 	     remain PC-relative even with _GLOBAL_OFFSET_TABLE_ as zero.  */
 	  sym->value = 0;
 	  continue;
 	}
+      if (strcmp (sym->name, ".TOC.") == 0)
+	{
+	  /* Handle the .TOC. symbol as the linker would do.  Set the .TOC.
+	     sections value and the section to point to the ".toc" section,
+	     and pass the toc->vma value into bfd_set_gp_value().
+	     If the .toc section is not found, use the first section
+	     with the SEC_ALLOC flag set.   Im the unlikely case that
+	     we still do not have a section identified, fall back to using
+	     bfd_abs_section_ptr.  */
+	  asection *toc_fallback = bfd_get_section_by_name(abfd.get(), ".toc");
+	  if (toc_fallback == NULL) {
+	    for (asection *tsect = abfd->sections; tsect != nullptr;
+	         tsect = tsect->next) {
+	      if (bfd_section_flags (tsect) & SEC_ALLOC) {
+		if (compile_debug)
+		  fprintf_unfiltered (gdb_stdlog,
+				      "Using section [%s] as .toc fallback.\n",
+				      tsect->name);
+		toc_fallback = tsect;
+		break;
+	      }
+	    }
+	  }
+	  if (toc_fallback == NULL) {
+            /*  If we've gotten here, we have not found a section usable
+		as a backup for the .toc section.  Use bfd_abs_section_ptr. */
+            if (compile_debug)
+	      fprintf_unfiltered (gdb_stdlog,
+				  "Using bfd_abs_section_ptr as .toc fallback.\n");
+            toc_fallback = bfd_abs_section_ptr;
+          }
+          sym->section = toc_fallback;
+          sym->value = 0x8000;
+          bfd_set_gp_value(abfd.get(), toc_fallback->vma);
+          if (compile_debug)
+            fprintf_unfiltered (gdb_stdlog,
+                                "Connectiong ELF symbol \"%s\" to the .toc section (%s)\n",
+                                sym->name,
+                                paddress (target_gdbarch (), sym->value));
+          continue;
+        }
+
       bmsym = lookup_minimal_symbol (sym->name, NULL, NULL);
       switch (bmsym.minsym == NULL
 	      ? mst_unknown : MSYMBOL_TYPE (bmsym.minsym))
 	{
 	case mst_text:

Re: [Patch v2] Handle .TOC. sections during gdb-compile for rs6000 target.

[Ulrich Weigand]

"will schmidt" <will_schmidt@vnet.ibm.com> wrote on 28.07.2021 21:08:21:

> +          sym->section = toc_fallback;
> +          sym->value = 0x8000;
> +          bfd_set_gp_value(abfd.get(), toc_fallback->vma);

For compatibility with BFD, shouldn't this be rather:

        bfd_set_gp_value(abfd.get(), toc_fallback->vma + sym->value);

Alan, can you confirm?


Bye,
Ulrich

Re: [Patch v2] Handle .TOC. sections during gdb-compile for rs6000 target.

[Alan Modra]

On Thu, Jul 29, 2021 at 09:18:18AM +0200, Ulrich Weigand wrote:
> 
> 
> "will schmidt" <will_schmidt@vnet.ibm.com> wrote on 28.07.2021 21:08:21:
> 
> > +          sym->section = toc_fallback;
> > +          sym->value = 0x8000;
> > +          bfd_set_gp_value(abfd.get(), toc_fallback->vma);
> 
> For compatibility with BFD, shouldn't this be rather:
> 
>         bfd_set_gp_value(abfd.get(), toc_fallback->vma + sym->value);
> 
> Alan, can you confirm?

No, see bfd/elf64-ppc.c:ppc64_elf_sec_toc.  A long time ago I chose
the value of elf_gp used in elf64-ppc.c to be the start of the toc
section rather than the value of ".TOC.".  The value is somewhat
arbitrary.  Yes, I could change it to be the value of .TOC., but for
now the value above is correct.

-- 
Alan Modra
Australia Development Lab, IBM

Re: [Patch v2] Handle .TOC. sections during gdb-compile for rs6000 target.

[Ulrich Weigand]

"Alan Modra" <amodra@gmail.com> wrote on 29.07.2021 10:26:11:
> On Thu, Jul 29, 2021 at 09:18:18AM +0200, Ulrich Weigand wrote:
> >
> >
> > "will schmidt" <will_schmidt@vnet.ibm.com> wrote on 28.07.2021
21:08:21:
> >
> > > +          sym->section = toc_fallback;
> > > +          sym->value = 0x8000;
> > > +          bfd_set_gp_value(abfd.get(), toc_fallback->vma);
> >
> > For compatibility with BFD, shouldn't this be rather:
> >
> >         bfd_set_gp_value(abfd.get(), toc_fallback->vma + sym->value);
> >
> > Alan, can you confirm?
>
> No, see bfd/elf64-ppc.c:ppc64_elf_sec_toc.  A long time ago I chose
> the value of elf_gp used in elf64-ppc.c to be the start of the toc
> section rather than the value of ".TOC.".  The value is somewhat
> arbitrary.  Yes, I could change it to be the value of .TOC., but for
> now the value above is correct.

I see, thanks for the explanation.

Will, the patch is OK for trunk then.

Thanks,
Ulrich

Re: [Patch v2] Handle .TOC. sections during gdb-compile for rs6000 target.

[will schmidt]

On Thu, 2021-07-29 at 13:49 +0200, Ulrich Weigand wrote:
> "Alan Modra" <amodra@gmail.com> wrote on 29.07.2021 10:26:11:
> > On Thu, Jul 29, 2021 at 09:18:18AM +0200, Ulrich Weigand wrote:
> > > 
> > > 
> > > "will schmidt" <will_schmidt@vnet.ibm.com> wrote on 28.07.2021
> 21:08:21:
> > > 
> > > > +          sym->section = toc_fallback;
> > > > +          sym->value = 0x8000;
> > > > +          bfd_set_gp_value(abfd.get(), toc_fallback->vma);
> > > 
> > > For compatibility with BFD, shouldn't this be rather:
> > > 
> > >         bfd_set_gp_value(abfd.get(), toc_fallback->vma + sym-
> >value);
> > > 
> > > Alan, can you confirm?
> > 
> > No, see bfd/elf64-ppc.c:ppc64_elf_sec_toc.  A long time ago I chose
> > the value of elf_gp used in elf64-ppc.c to be the start of the toc
> > section rather than the value of ".TOC.".  The value is somewhat
> > arbitrary.  Yes, I could change it to be the value of .TOC., but
> for
> > now the value above is correct.
> 
> I see, thanks for the explanation.
> 
> Will, the patch is OK for trunk then.

Committed with some additional indentation and cosmetic tweaks,
Thanks
-WIll


> 
> Thanks,
> Ulrich
> 
>