Fix PR ld/33199 SEGV in _bfd_x86_elf_create_sframe_plt
Currently, the selection for sframe_plt was not being done (and simply
set to NULL) for the case when !normal_target, causing SEGV on Solaris.
Initialize sframe_plt to init_table->sframe_lazy_plt when lazy_plt is
true, and NULL otherwise. This is in line with htab->non_lazy_plt being
set to NULL for !normal_target.
bfd/
PR ld/33199
* elfxx-x86.c (_bfd_x86_elf_link_setup_gnu_properties):
Setup sframe_plt for !normal_target.
In a later commit I'd like to add support to GDB for including the
NT_386_TLS note in the core files that GDB creates (using 'gcore'
command).
To achieve this we need some standard boilerplate code added to bfd.
The only part of this patch which I think needs consideration is the
name I selected for the pseudo section to hold the note contents when
a core file is loaded. I chose '.reg-i386-tls'. The '.reg' prefix is
the standard used by most other pseudo sections, and the '-i386-tls'
suffix seemed to match the note name, though I added the 'i' to
'i386', instead of just using '.reg-386-tls'. I thought 'i386' seemed
clearer.
There's no test included here, but when I merge the NT_386_TLS
creation to GDB it will depend on this and act as a test. I plan to
post that work to the GDB list once this patch is merged.
Recognise ECOFF archives and reject them so that the ecoff archive
support has a chance to handle the archive. Also use memcmp rather
than startswith (strncmp) as we know the string length.
* archive.c (bfd_slurp_armap): Recognize ECOFF armap. Use memcmp
to match names, and tidy buffer sizes.
This tidies objcopy/strip handling of IR objects, in the process of
removing the unnecessary is_strip_input flag.
The first thing I noticed when looking at is_strip_input code was that
the abfd->my_archive test in bfd_check_format_matches meant that
plugins were disabled when reading archive elements. We can instead
disable plugins by setting bfd_no_plugin, so there doesn't seem to be
a need for is_strip_input in objcopy.c:copy_archive. This isn't
exactly the same, because bfd_no_plugin prevents the plugin target
recognising archive elements in the bfd_check_format_matches loop over
all targets as well as just the first !target_defaulted test. But
that turns out to be fine. IR code is handled in copy_archive as for
other unknown format files. In fact, the only need for the plugin
target when copying archives is when reading symbols for the archive
map. I've made that plain by moving the plugin target override and
commenting on why it is really needed.
So on to plain object files. Here, IR code is also copied unchanged,
so there doesn't seem a need for the plugin target there either. It
isn't quite so simple though, because the objcopy/strip code handling
object files wants to verify the format of the object file. Allowing
objcopy/strip to copy unknown format files would be a change in
behaviour (and results in mmix testsuite fails, ld-mmix/b-badfil1 and
others). However, always excluding the plugin target results in a
fail of tests added in commit c2729c37f1. So I've enabled a plugin
format check only for files that are otherwise unrecognised, and
commented why this is done. I question the need to objcopy LLVM
bytecode files.
bfd/
* bfd.c (struct bfd<is_strip_input>): Delete.
* format.c (bfd_check_format_matches): Delete is_strip_input
special case code.
* bfd-in2.h: Regenerate.
binutils/
* objcopy.c (copy_archive): Don't set is_strip_input. Always
set bfd_plugin_no when reading elements. Enable plugins when
opening copied elements.
(check_format_object): Delete.
(copy_file): Don't enable plugin target here. Don't set
is_strip_input. Set bfd_plugin_no. Move bfd_core handling
code earlier to remove goto. Enable plugin for llvm bytecode.
Copy slim IR files as unknown objects.
If all external symbol accesses are indirect, we can treat protected
symbols as local since there will be no copy relocation for data and
external function pointer access will go through GOT, instead of PLT.
No PLT slot should be used for external function pointer in executable.
bfd/
PR ld/33260
* elfxx-x86.h (COPY_INPUT_RELOC_P): Treat protected symbols with
indirect external access as local.
ld/
PR ld/33260
* testsuite/ld-i386/i386.exp: Run PR ld/33260 test.
* testsuite/ld-x86-64/x86-64.exp: Likewise.
* testsuite/ld-i386/pr33260.d: New file.
* testsuite/ld-i386/pr33260.s: Likewise.
* testsuite/ld-x86-64/pr33260-x32.d: Likewise.
* testsuite/ld-x86-64/pr33260.d: Likewise.
* testsuite/ld-x86-64/pr33260.s: Likewise.
Signed-off-by: H.J. Lu <hjl.tools@gmail.com>
There's no reason to have the compiler materialize objects onto the
stack. And there's also no reason to allow comb[] and name_table[] to be
modifiable.
There's no reason for riscv_all_supported_ext[] to appear in libbfd.so's
dynamic symbol table. There's also no reason for various pieces of data
to live in .data, when .data.rel.ro or even .rodata can do.
... and const. There's no reason to have the compiler copy anonymous
objects onto the stack. And there's also no reason to allow the arrays
to be modifiable.
It isn't needed anywhere except plugin.c. The typedef can disappear.
Also make a forward declaraion for ld_plugin_input_file in plugin.h
so that this header can be used without first including plugin-api.h.
bfd/
* plugin.h (struct ld_plugin_input_file): Forward declare.
(struct plugin_data_struct): Move to..
* plugin.c: ..here.
(add_symbols): Size plugin_data without using type.
* archive.c: Don't include plugin-api.h.
* elflink.c: Likewise.
* format.c: Likewise.
binutils/
* ar.c: Don't include plugin-api.h or ansidecl.h. Only
include plugin.h when BFD_SUPPORTS_PLUGINS.
* nm.c: Don't include plugin-api.h. Only include plugin.h
when BFD_SUPPORTS_PLUGINS.
* objcopy.c: Likewise.
ld/
* ldfile.c: Don't include plugin-api.h.
* ldmain.c: Likewise.
There's a logic error in loongarch_relax_perform_deletes: when there's
not any delete operation of which the start address is strictly smaller
than the symbol address, splay_tree_predecessor() will return nullptr
and the symbol size will be unchanged even if some bytes of it are
removed.
Make the logic more complete to fix this issue. Also factor out the
symbol size adjustment logic into a function to avoid code bloating.
Tested-by: WANG Xuerui <git@xen0n.name>
Signed-off-by: Xi Ruoyao <xry111@xry111.site>
Fat IR objects contains both regular sections and IR sections. After
commit 717a38e9a0
Author: H.J. Lu <hjl.tools@gmail.com>
Date: Sun May 4 05:12:46 2025 +0800
strip: Add GCC LTO IR support
"strip --strip-debug" no longer strips debug sections in fat IR objects
since fat IR objects are recognized as plugin object and copied as unknown
objects. Add a is_strip_input field to bfd to indicate called from strip.
Update bfd_check_format_matches not to treat archive member nor standalone
fat IR object as IR object so that strip can remove debug and IR sections
in fat IR object. For archive member, it is copied as an unknown object
if the plugin target is in use or it is a slim IR object. For standalone
fat IR object, it is copied as non-IR object.
bfd/
PR binutils/33246
* archive.c: Include "plugin-api.h" and "plugin.h" if plugin is
enabled.
(_bfd_compute_and_write_armap): Don't complain plugin is needed
when the plugin target is in use.
* bfd-in2.h: Regenerated.
* bfd.c (bfd): Add is_strip_input.
* format.c (bfd_set_lto_type): If there is .llvm.lto section,
set LTO type to lto_fat_ir_object.
(bfd_check_format_matches): Don't set LTO type when setting
format. When called from strip, don't treat archive member nor
standalone fat IR object as an IR object.
* plugin.c (bfd_plugin_get_symbols_in_object_only): Copy LTO
type derived from input sections.
nm/
PR binutils/33246
* nm.c (filter_symbols): Don't complain plugin is needed when
the plugin target is in use.
(display_rel_file): Likewise.
* objcopy.c (copy_archive): Set the BFD is_strip_input field of
archive member to 1 to indicate called from strip. Also copy
slim IR archive member as unknown object.
(copy_file): Set the BFD is_strip_input field of input bfd to
1 to indicate called from strip.
(strip_main): Keep .gnu.debuglto_* sections unless all GCC LTO
sections will be removed.
ld/
PR binutils/33246
* testsuite/ld-plugin/lto-binutils.exp (run_pr33246_test): New.
Run binutils/33246 tests with GCC and Clang.
* testsuite/ld-plugin/pr33246.c: New file.
Signed-off-by: H.J. Lu <hjl.tools@gmail.com>
Tidy changes to bfd_check_format_matches made by commit 9b854f169d
which added a bfd_plugin_specified_p test and commit f752be8f91
which added an lto_sections_removed arg. Both of these changes are
unnecessary if plugin_format is set to bfd_plugin_no before calling
bfd_check_format. bfd_plugin_no will prevent the plugin object_p
function from returning a match (and in the first case from a segfault
when loading plugins while a plugin is running). The plugin object_p
function already protected itself from recursive calls by setting
bfd_plugin_no before loading a plugin, but commit 9b854f169d opened
new bfds so they were unprotected.
It isn't strictly necessary to test for bfd_plugin_no in
bfd_check_format_matches but I kept the check to minimise functional
changes. Close inspection of the patch will notice I've added an
is_linker_input test too. That also isn't strictly necessary, I
think, but the match_count test was for the linker. See commit
999d6dff80.
PR 12291
PR 12430
PR 13298
PR 33198
bfd/
* format.c (bfd_check_format_lto): Revert to bfd_check_format.
(bfd_check_format_matches_lto): Revert to bfd_check_format_matches.
Correct comments. Manage both the lto_sections_removed and
bfd_plugin_specified_p cases by testing for bfd_plugin_no.
* plugin.c (bfd_plugin_get_symbols_in_object_only): Set
plugin_format to bfd_plugin_no before checking new bfds.
(try_load_plugin): Comment setting bfd_plugin_no.
(bfd_plugin_specified_p): Delete.
* plugin.h (bfd_plugin_specified_p): Delete.
* bfd-in2.h: Regenerate.
binutils/
* objcopy.c (copy_archive): Replace bfd_check_format_lto calls
with bfd_check_format using plugin_format set to bfd_plugin_no.
(check_format_object): New function.
(copy_file): Use it.
ppc32 isn't susceptible to the PR33223 segfault, but could hit a
_bfd_clear_contents segfault with a carefully crafted invalid object.
* elf32-ppc.c (ARRAY_SIZE): Define.
(ppc_elf_howto_init): Use ARRAY_SIZE.
(ppc_elf_reloc_name_lookup): Likewise.
(ppc_elf_info_to_howto): Likewise, and consolidate error
handling.
(ppc_elf_check_relocs): Guard against segfaults caused by a NULL
howto passed to _bfd_clear_contents. Use ARRAY_SIZE.
Bounds check accesses to ppc64_elf_howto_table and don't dereference a
NULL howto. I think this catches all cases where that might happen.
PR 33223
bfd/
* elf64-ppc.c (ppc64_elf_info_to_howto): Consolidate error handling.
(ppc64_elf_check_relocs): Tidy error messages.
(ppc64_elf_relocate_section): Don't segfault when attempting to
report an unsupported relocation. Don't pass a NULL howto to
_bfd_clear_contents.
ld/
* testsuite/ld-powerpc/elfv2-2so.d: Adjust to suit error message
change.
commit 717a38e9a0
Author: H.J. Lu <hjl.tools@gmail.com>
Date: Sun May 4 05:12:46 2025 +0800
strip: Add GCC LTO IR support
added "-R .gnu.lto_.*" to strip to remove all GCC LTO sections. When
"-R .gnu.lto_.*" is used, the plugin target is ignored so that all LTO
sections are stripped as the regular sections. It works for the slim
GCC LTO IR since the GCC LTO IR is stored in the regular sections. When
the plugin target is ignored, the GCC LTO IR can be recognized as the
normal object files. But it doesn't work for the slim LLVM IR which
is stored in a standalone file.
1. Add bfd_check_format_matches_lto and bfd_check_format_lto to take an
argument, lto_sections_removed, to indicate if all LTO sections should
be removed.
2. Update strip to always enable the plugin target so that the plugin
target is enabled when checking for bfd_archive.
3. Update strip to ignore the plugin target for bfd_object when all LTO
sections should be removed. If the object is unknown, copy it as an
unknown file without any messages.
4. Treat the "-R .llvm.lto" strip option as removing all LTO sections.
bfd/
PR binutils/33198
* format.c (bfd_check_format_lto): New function.
(bfd_check_format): Call bfd_check_format_matches_lto.
(bfd_check_format_matches): Renamed to ...
(bfd_check_format_matches_lto): This. Add an argument,
lto_sections_removed, to indicate if all LTO sections should be
removed and don't match the plugin target if lto_sections_removed
is true.
(bfd_check_format_matches): Call bfd_check_format_matches_lto.
* bfd-in2.h: Regenerated.
binutils/
PR binutils/33198
* objcopy.c (copy_archive): Call bfd_check_format_lto, instead
of bfd_check_format, and pass lto_sections_removed. Remove the
non-fatal message on unknown element since it will be copied as
an unknown file.
(copy_file): Don't check lto_sections_removed when enabling LTO
plugin in strip.
(copy_file): Ignore the plugin target first if all LTO sections
should be removed. Try with the plugin target next if ignoring
the plugin target failed to match the format.
(strip_main): Also set lto_sections_removed for -R .llvm.lto.
* testsuite/binutils-all/x86-64/pr33198.c: New file.
* testsuite/binutils-all/x86-64/x86-64.exp (run_pr33198_test):
New.
Run binutils/33198 tests.
* testsuite/lib/binutils-common.exp (llvm_plug_opt): New.
(CLANG_FOR_TARGET): New. Set to "clang" for native build if
"clang -v" reports "clang version".
Signed-off-by: H.J. Lu <hjl.tools@gmail.com>
Considering the following case,
% cat tmp.s
.option pic
.text
.global _start
_start:
nop
.section .discard.s, "ax"
la x1, _start
% cat tmp.ld
OUTPUT_ARCH(riscv)
ENTRY(_start)
SECTIONS
{
/DISCARD/ : { *(.discard.*) }
. = 0x10000;
.text : { *(.text) }
. = 0x20000;
.got : { *(.got) *(.got.plt)}
. = 0x30000;
.data : { *(.data) *(.data.*) }
}
% riscv64-unknown-linux-gnu-as tmp.s -o tmp.o
% riscv64-unknown-linux-gnu-ld -pie -Ttmp.ld tmp.o
riscv64-unknown-linux-gnu-ld: BFD (GNU Binutils) 2.44.50.20250624 assertion fail binutils-gdb/bfd/elfnn-riscv.c:3638
This happens when pie and the input sections, which refers to the global
symbol by got, are all discarded. Since referenced sections are all discarded,
we won't go into relocate_section for those sections, the got entry also won't
be initialized. Therefore, we will get assert fail when adding the RELATIVE
reloc in the finish_dynamic_symbol.
After seeing other target codes, there are two root causes as follows,
1. risc-v may call bfd_elf_link_record_dynamic_symbol in the allocate_dynrelocs
for not only undefweak symbols.
2. risc-v is missing the code to add RELATIVE to R_RISCV_GOT entries in the
relocate_section if a symbol is not dynamic and is not undefined weak under
pic and pie.
If we call bfd_elf_link_record_dynamic_symbol, then the global symbol will be
forced to dynamic, so the h->dynindx will forced to be a number rather than -1,
even it should be -1. Once h->dynindx != -1 and pic/pie, it will go into
finish_dynamic_symbol and insert RELATIVE/64 relocs for the got entry; For the
above case there are two issues,
1. The global symbol _start is forced to be dynamic in the allocate_dynrelocs.
when pie and all the referenced section are discarded, it won't go into
relocate_section to initialize the got entry, so it will cause assert fail
when adding RELATIVE reloc in the finish_dynamic_symbol. The assert fail
represents another problem - if we don't initialize the got entry in the
relocate_section under pie, which means we don't need to go into the
finish_dynamic_symbol and don't need a RELATIVE reloc for the got entry,
it should be NONE reloc.
2. Without linking any discarded section, it originally forces every RELATIVE
relocs added for every got by the finish_dynamic_symbol. Even The final
result looks correct under pie (genearte a RELATIVE reloc for got entry),
not sure if it may cause other problems for some special cases, excpet the
above one.
Therefore, this patch try to fix the above assert fail, and also clarify the
behavior of the allocate_dynrelocs which should only call bfd_elf_link_record_dynamic_symbol
for undefweak symbols, and add the missing code to generate RELATIVE reloc to
R_RISCV_GOT entries in the relocate_section if a symbol is not dynamic and is
not undefined weak under pic and pie.
Passed the gcc/binutils regressions of riscv-gnu-toolchain at least.
