[Bug ld/32219] ELF orphan placement doesn't work well without .interp
https://sourceware.org/bugzilla/show_bug.cgi?id=32219 Nick Clifton changed: What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED --- Comment #6 from Nick Clifton --- patch applied -- You are receiving this mail because: You are on the CC list for the bug.
[Bug ld/32219] ELF orphan placement doesn't work well without .interp
https://sourceware.org/bugzilla/show_bug.cgi?id=32219 --- Comment #5 from Sourceware Commits --- The master branch has been updated by Nick Clifton : https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=7ecf0250f7f0868f3b828ffadd47d0e5f2dbd875 commit 7ecf0250f7f0868f3b828ffadd47d0e5f2dbd875 Author: Nick Clifton Date: Mon Sep 30 14:41:11 2024 +0100 Improve the placement of orphan note sections. PR 32219 -- You are receiving this mail because: You are on the CC list for the bug.
[Bug ld/32219] ELF orphan placement doesn't work well without .interp
https://sourceware.org/bugzilla/show_bug.cgi?id=32219
--- Comment #4 from H.J. Lu ---
(In reply to Nick Clifton from comment #3)
> Created attachment 15722 [details]
> Proposed patch
>
> Hi H.J.
>
> How about this proposed patch instead ?
>
> It modifies the orphan placement algorithm to test to see if the anchor
> section for orphaned NOTES exists before using it:
>
> +{
> + /* PR 32219: Check that the .interp section
> + exists before attaching orphans to it. */
> + if (lang_output_section_find (hold[orphan_interp].name))
> + place = &hold[orphan_interp];
> + /* Next best place: after .rodata. */
> + else if (lang_output_section_find (hold[orphan_rodata].name))
> + place = &hold[orphan_rodata];
> + /* Last attempt: the .text section. */
> + else
> + place = &hold[orphan_text];
> +}
>
> It also tweaks various linker script files for targets that might benefit
> and then adds XFAILS to the pr23658-1e test for the remaining unhandled
> targets.
>
> What do you think ?
>
It looks good to me. Thanks.
--
You are receiving this mail because:
You are on the CC list for the bug.
[Bug ld/32219] ELF orphan placement doesn't work well without .interp
https://sourceware.org/bugzilla/show_bug.cgi?id=32219 Sam James changed: What|Removed |Added CC||sam at gentoo dot org -- You are receiving this mail because: You are on the CC list for the bug.
[Bug ld/32219] ELF orphan placement doesn't work well without .interp
https://sourceware.org/bugzilla/show_bug.cgi?id=32219
--- Comment #3 from Nick Clifton ---
Created attachment 15722
--> https://sourceware.org/bugzilla/attachment.cgi?id=15722&action=edit
Proposed patch
Hi H.J.
How about this proposed patch instead ?
It modifies the orphan placement algorithm to test to see if the anchor
section for orphaned NOTES exists before using it:
+{
+ /* PR 32219: Check that the .interp section
+exists before attaching orphans to it. */
+ if (lang_output_section_find (hold[orphan_interp].name))
+ place = &hold[orphan_interp];
+ /* Next best place: after .rodata. */
+ else if (lang_output_section_find (hold[orphan_rodata].name))
+ place = &hold[orphan_rodata];
+ /* Last attempt: the .text section. */
+ else
+ place = &hold[orphan_text];
+}
It also tweaks various linker script files for targets that might benefit and
then adds XFAILS to the pr23658-1e test for the remaining unhandled targets.
What do you think ?
Cheers
Nick
--
You are receiving this mail because:
You are on the CC list for the bug.
[Bug ld/32219] ELF orphan placement doesn't work well without .interp
https://sourceware.org/bugzilla/show_bug.cgi?id=32219
--- Comment #2 from H.J. Lu ---
This patch removes .interp and uses .rodata for note sections instead:
diff --git a/ld/ldelf.c b/ld/ldelf.c
index 09691609fc0..b09ba054ed5 100644
--- a/ld/ldelf.c
+++ b/ld/ldelf.c
@@ -2121,9 +2121,6 @@ ldelf_place_orphan (asection *s, const char *secname, int
constraint)
{ 0,
SEC_HAS_CONTENTS | SEC_ALLOC | SEC_LOAD | SEC_READONLY | SEC_DATA,
0, 0, 0, 0 },
- { ".interp",
- SEC_HAS_CONTENTS | SEC_ALLOC | SEC_LOAD | SEC_READONLY | SEC_DATA,
- 0, 0, 0, 0 },
{ ".sdata",
SEC_HAS_CONTENTS | SEC_ALLOC | SEC_LOAD | SEC_DATA | SEC_SMALL_DATA,
0, 0, 0, 0 },
@@ -2139,7 +2136,6 @@ ldelf_place_orphan (asection *s, const char *secname, int
constraint)
orphan_data,
orphan_bss,
orphan_rel,
- orphan_interp,
orphan_sdata,
orphan_nonalloc
};
@@ -2328,7 +2324,7 @@ ldelf_place_orphan (asection *s, const char *secname, int
constraint)
&& (elfinput
? sh_type == SHT_NOTE
: startswith (secname, ".note")))
-place = &hold[orphan_interp];
+place = &hold[orphan_rodata];
else if ((flags & (SEC_LOAD | SEC_HAS_CONTENTS | SEC_THREAD_LOCAL)) == 0)
place = &hold[orphan_bss];
else if ((flags & SEC_SMALL_DATA) != 0)
But -z separate-code still doesn't work for many targets.
--
You are receiving this mail because:
You are on the CC list for the bug.
[Bug ld/32219] ELF orphan placement doesn't work well without .interp
https://sourceware.org/bugzilla/show_bug.cgi?id=32219 Nick Clifton changed: What|Removed |Added CC||nickc at redhat dot com --- Comment #1 from Nick Clifton --- The problem also affects: arc-linux-uclibc avr-elf d10v-elf h8300-linux-gnu ip2k-elf m32r-elf mep-elf microblaze-elf moxie-elf spu-elf v850-elf xtensa-elf -- You are receiving this mail because: You are on the CC list for the bug.
