With the changes of an added flag SFRAME_F_FDE_FUNC_START_PCREL, s390x
support and new section type SHT_GNU_SFRAME, indicate that this document
specifies the errata 1 of SFrame version 2. This will help distinguish
the document / specification better from previous releases.
libsframe/doc/
* sframe-spec.texi: Mention errata 1 of SFrame version 2.
Fix PR binutils/33186 - No SFrame dump if section name is not .sframe
When the section name is not ".sframe", ensure that readelf and objdump
are able to dump a section of type SHT_GNU_SFRAME and not fail if the
user specifies the new section name.
For objdump, in dump_dwarf_section (), use the match string of ".sframe"
to find the corresponding debug_displays[] item for SFrame section.
Doing this ensures that any call to dump_dwarf_section () with the
section pointing to the SFrame section (with name possibly different
from ".sframe") will successfully dump the SFrame section.
If the SFrame section is named anything but ".sframe", and user does not
specify the name of the SFrame section either, the documented behaviour
is that the default section name is assumed to be ".sframe". So the
following (albeit counter intuitive) is expected at this time:
$ readelf -S sort | grep sframe
[NN] .sframe2 GNU_SFRAME 0000000000NNNNNN 0000NNNN
(Note section name .sframe2).
$ objdump --sframe sort
sort: file format elf64-x86-64
No .sframe section present
(Similarly for readelf as well).
For objdump, set dump_sframe_section_name to ".sframe" if user specifies
no section name. In the error checking done in dump_sframe_section, add
the case when user specifies a valid section name but one that does not
contain SFrame section data. For sections generated with Binutils >=
2.45, this can be checked with section type of SHT_GNU_SFRAME.
Previously these sections were SHT_PROGBITS with name ".sframe".
Similar changes in readelf.
Add a test each for objdump and readelf to dump a renamed section. Use
gas_sframe_check to limit the execution of these tests only when a gas
supporting SFrame format is present.
binutils/
PR binutils/33186
* objdump.c (dump_dwarf_section): Set match to ".sframe" which
corresponds to the name in the debug_displays[] entry for
SFrame section.
(dump_sframe_section): Check if the user specified section name
contains SFrame data.
(main): Set default section name to ".sframe".
* readelf.c (display_debug_section): Adjust checks to find the
debug_diplays[] item for the input arg SFrame section.
Use id instead of i, as it is more readable.
binutils/testsuite/
PR binutils/33186
* binutils-all/x86-64/objdump-sframe-01.d: New test.
* binutils-all/x86-64/readelf-sframe-01.d: New test.
* binutils-all/x86-64/sframe-func.s: New test.
Fix PR gas/33175 sframe: --gsframe=no does not disable when .cfi_sections directive with .sframe
--gsframe=no should also disable generation of SFrame section when explicit CFI directive: .cfi_sections .sframe is specified in the input.
This means we need to track whether SFrame generation was explicitly disabled by the user.
Introduce a new enum to facilitate disambiguation between GEN_SFRAME_DEFAULT_NONE and GEN_SFRAME_DISABLED.
While fixing the bug by adding the enum, keep the upcoming requirement in mind: we will also need to disambiguate between --enable-default-sframe and user-specified --gsframe/--gsframe=yes.
The intent is to not display SFrame related warnings or errors like: as_bad (_(".sframe not supported for target")); for unsupported targets if --enable-default-sframe is in effect.
This implies we need to have a four state enum ( GEN_SFRAME_DEFAULT_NONE, GEN_SFRAME_CONFIG_ENABLED, GEN_SFRAME_DISABLED, GEN_SFRAME_ENABLED) gas
Add faminmax, move a couple of misplaced entries, and improve a few
other entries.
The documentation now lists every recognised extension name, with the
exception of a couple of aliases that are deliberately undocumented.
As described in PR ld/32580, when using SHELL=/bin/sh or /bin/ksh on
Solaris, the generated linker scripts get corrupted. So far, the only
workaround is to enforce /bin/bash instead.
This is a major nuisance for developers and users alike, so this patch
automates this by overriding SHELL in ld/configure.ac.
Tested on amd64-pc-solaris2.11 in three configurations:
* CONFIG_SHELL unset
* CONFIG_SHELL=/bin/ksh
* CONFIG_SHELL='/bin/bash --norc'
In the first two cases, SHELL was set to /bin/bash as desired, while in
the third it was left unchanged.
2025-07-24 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
ld:
PR ld/32580
* configure.ac <*-*-solaris2*>: Enforce SHELL=/bin/bash.
* configure: Regenerate.
(cherry picked from commit 96ad2fd3c0c0414110fe58ed4ee511f49768fa3d)
SFrame generation code assumes that since DW_CFA_restore means
restoration of the state of the register to the one at the beginning of
the function, there must be a state to restore to (hence the gas_assert
(cie_fre)).
This assumption needs adjustment. DW_CFA_restore may be present in the
very beginning of a (e.g., cold) function, with no initialized state for
SFrame functions to restore to.
gas/
PR gas/33170
* gas/gen-sframe.c (sframe_xlate_do_restore): Use current FRE if
CIE FRE is not yet setup.
gas/testsuite/
PR gas/33170
* gas/cfi-sframe/cfi-sframe.exp: New test.
* gas/cfi-sframe/cfi-sframe-x86_64-pr33170.d: New test.
* gas/cfi-sframe/cfi-sframe-x86_64-pr33170.s: New test.
(cherry picked from commit 83eeaf917873a550656faf9a38cd14e0f4c521b1)
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>
(cherry picked from commit f752be8f916efa70aea9c2e4f664c75690fd136c)
Rearrange scripttempl/avr.sc to avoid oddities of shells expanding
${RELOCATING+stuff} in here documents where "stuff" contains quoted
strings. Also I think it is better to avoid multi-line "stuff" as it
can be tricky to spot the ending brace.
(cherry picked from commit ae114fb523efe908f9e807359e2f494ee64d2801)
To rename a file on Windows, the target name cannot exist. Removing file
prior to renaming ensures this is handled.
To remove a file on Windows, the file cannot be open. Closing the bfd
handle ensures this is handled.
Moved call to free on isympp / osympp to after bfd is closed to align
with comment earlier in the cmdline_add_object_only_section function.
Signed-off-by: Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
(cherry picked from commit 233cd5946413108bf4902b22a9cb23ad0a468f5e)
PR gprofng/33151
gprofng ignores functions that are compiled as weak symbols. This
heavily affects C++ class methods that are always compiled by g++
and clang++ as weak symbols. In this case 'gprofng display text'
just displays <static>@ADDRESS(<FILENAME>) instead of proper method
name.
The bug has been introduced in the commit 470a0288a8.
Commit d7f343eaad3f testsuite change resulted in a regression for
s390x-linux. This extends the x86_64 fix to other targets.
PR ld/33156
* elf-bfd.h (RELOC_AGAINST_DISCARDED_SECTION): Remove .sframe
relocs too.
(cherry picked from commit fcf7470408aa0508fbe99abb99547757a348383d)
Fix PR ld/33131 Failed assertion when linking gccgo
Make amendments in both sframe_decoder_get_fre and
sframe_encoder_add_fre.
Since GNU as and the dw2gencfi code generally accepts such CFI, its best
to allow in SFrame FREs too.
libsframe/
PR ld/33131.
* sframe.c (sframe_decoder_get_fre): Relax the assertion a bit.
(sframe_encoder_add_fre): Likewise.
(cherry picked from commit 387efef5fef727cbe52099dcd5012905c4205be3)
I omitted documentation in 8aad677a12832885acd5be1de8f41e740b8e713d in
error. Rectify that with:
1) changing ---help to mention bare `--gsframe` too, as we're not
getting rid of that;
2) adding the new --gsframe=[no|yes] form to as.texi.
PR gas/33125
* gas/as.c (parse_args): Tweak --gsframe= help text.
* gas/doc/as.texi: Document --gsframe=[no|yes].
(cherry picked from commit 50c1c57426db6e1c7b44b4d05f0b07fcba91f890)
Being able to explicitly disable SFrames on the command line is useful,
especially when looking at a gas that enables SFrames by default. The
binutils testsuite will benefit from this as there's testcases that don't
expect their presence.
In summary:
* Nothing is passed => no SFrames (no change from before)
* --gsframe is passed => SFrames (no change from before)
* --gsframe=yes is passed => SFrames (previously rejected)
* --gsframe-no is passed => no SFrames (previously rejected)
PR gas/33125
* gas/as.c (parse_args): Accept --gsframe=no, --gsframe=yes.
(cherry picked from commit 8aad677a12832885acd5be1de8f41e740b8e713d)
Since unlike eh_frame editing code, sframe editing code keeps
R_X86_64_NONE reloc as is, its r_offset is wrong, we must not
generate R_X86_64_NONE reloc in sframe section against discarded
sections for "ld -r".
bfd/
PR ld/33156
* elf64-x86-64.c (elf_x86_64_relocate_section): Also remove
sframe relocations against discarded sections for "ld -r".
ld/
PR ld/33156
* testsuite/ld-elf/eh-group.exp (as_gsframe): New.
Assemble eh-group.o with $as_gsframe.
Signed-off-by: H.J. Lu <hjl.tools@gmail.com>
(cherry picked from commit d7f343eaad3f34c76657b9996e6253b4f9a218d5)
"ld -r" generates R_*_NONE relocations in sframe section if input
relocations in sframe section are against discarded section. Allow
input R_*_NONE relocations if there are more relocation entries than
SFrame entries, instead of assuming number of SFrame entries == number
of relocation entries.
bfd/
PR ld/33127
* elf-sframe.c (sframe_decoder_init_func_bfdinfo): Allow input
R_*_NONE relocations if there are more relocation entries than
SFrame entries.
ld/
PR ld/33127
* testsuite/ld-x86-64/sframe-reloc-2a.s: New file.
* testsuite/ld-x86-64/sframe-reloc-2b.s: Likewise.
* testsuite/ld-x86-64/x86-64.exp: Run PR ld/33127 tests.
Signed-off-by: H.J. Lu <hjl.tools@gmail.com>
(cherry picked from commit 5f9bf0cf711a153a0a20d6ff88181e9a6775bdba)
Clear map_head_is_link_order when generating .gnu_object_only section so
that lang_add_section can add new sections and .sframe sections will be
properly merged by _bfd_elf_merge_section_sframe.
PR ld/33146
* ldlang.c (cmdline_emit_object_only_section): Clear
map_head_is_link_order.
* testsuite/ld-plugin/lto.exp (as_gsframe): New.
(lto_link_tests): Add $as_gsframe to compile lto-4b.o and
lto-4c.o.
Signed-off-by: H.J. Lu <hjl.tools@gmail.com>
(cherry picked from commit 939eb467b21de5d18ee703755fb9704a525cfe21)
PR ld/33146
Correct TCL errors trying to access error output file in commit
ef7a634dc01d. In fact, get rid of the output file test entirely since
gas exit status is sufficient.
Also there is no need to firstly check for ELF support.
Set check_as_sframe_result, and remove ld-lib.exp check_as_sframe.
(cherry picked from commit a57a3a169ea9ec79977949c9c8dccd3a2a615fae)
Move gas_sframe_check to binutils-common.exp so that it can be used in
linker tests to check if a target assembler supports --gsframe.
binutils/
PR ld/33146
* testsuite/lib/binutils-common.exp (gas_sframe_check): Moved
from cfi-sframe.exp. Replace gas_host_run with remote_exec.
gas/
PR ld/33146
* testsuite/gas/cfi-sframe/cfi-sframe.exp (gas_sframe_check):
Moved to binutils-common.exp.
Signed-off-by: H.J. Lu <hjl.tools@gmail.com>
(cherry picked from commit ef7a634dc01df3d78f208c93316b52937d3fe8f4)
PR mi/32571 reports the following problem:
...
$ gdb -q -batch -ex "b bla.c:100"
<random output>
Make breakpoint pending on future shared library load? (y or [n]) \
[answered N; input not from terminal]
...
while this is expected:
...
$ gdb -q -batch -ex "b bla.c:100"
No symbol table is loaded. Use the "file" command.
Make breakpoint pending on future shared library load? (y or [n]) \
[answered N; input not from terminal]
...
A few factors in reproducing this are building gdb using gcc 14,
"-O2 -flto=auto" and --disable-nls. For more details, see the PR.
This turns out to be caused by a GCC PR [1], more specifically a problem in
ipa-modref.
Work around this by disabling ipa-modref for GCC versions 12-15 and 16.0,
assuming the GCC 16.1 release will contain a fix.
Tested on aarch64-linux and x86_64-linux.
Approved-By: Andrew Burgess <aburgess@redhat.com>
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=32571
[1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120987
Add new command line option -z memtag-stack for aarch64 elf. This
option instructs the linker to generate the necessary dynamic tag
DT_AARCH64_MEMTAG_STACK, which the dynamic loader can then use to
protect the stack memory with PROT_MTE. Linker issues an
'unrecognized option' error when -z memtag-stack is specified for
non-aarch64 based emulations.
readelf displays the dynamic tag when present:
$ readelf -d <exectutable>
Dynamic section at offset 0xfdd8 contains XX entries:
Tag Type Name/Value
0x0000000000000001 (NEEDED) Shared library: [libc.so.6]
0x000000000000000c (INIT) 0x400520
0x000000000000000d (FINI) 0x400b64
0x0000000000000019 (INIT_ARRAY) 0x41fdc8
... ... ...
0x000000007000000c (AARCH64_MEMTAG_STACK) 0x1
... ... ...
ChangeLog:
* bfd/elfnn-aarch64.c (elfNN_aarch64_late_size_sections): Emit
DT_AARCH64_MEMTAG_STACK dynamic tag.
* bfd/elfxx-aarch64.h (struct aarch64_memtag_opts): Add new
member for tracking whether stack access uses MTE insns.
* binutils/readelf.c (get_aarch64_dynamic_type): Handle
DT_AARCH64_MEMTAG_STACK.
* ld/emultempl/aarch64elf.em: Add new command line option.
* ld/ld.texi: Add documentation for -z memtag-stack.
* ld/testsuite/ld-aarch64/aarch64-elf.exp: Add new test.
* ld/testsuite/ld-aarch64/dt-memtag-stack.d: New test.
include/ChangeLog:
* elf/aarch64.h (DT_AARCH64_MEMTAG_STACK): New definition.
Add new command line option -z memtag-mode=<mode> to aarch64 elf,
where <mode> can be one of none, sync, or async. For mode of sync or
async, a DT_AARCH64_MEMTAG_MODE dynamic tag with a value of 0 or 1
respectively is emitted.
readelf displays the dynamic tag when present:
$ readelf -d <exectutable>
Dynamic section at offset 0xfdd8 contains XX entries:
Tag Type Name/Value
0x0000000000000001 (NEEDED) Shared library: [libc.so.6]
0x000000000000000c (INIT) 0x400520
0x000000000000000d (FINI) 0x400b64
0x0000000000000019 (INIT_ARRAY) 0x41fdc8
... ... ...
0x0000000070000009 (AARCH64_MEMTAG_MODE) 0x1
... ... ...
Note that this patch doesn't add support for the "asymm" MTE mode,
which is an Armv8.7 extension.
ChangeLog:
* bfd/elfnn-aarch64.c (struct elf_aarch64_link_hash_table): Add
new member for memtag properties.
(bfd_elfNN_aarch64_set_options): New argument to pass memtag
properties.
(elfNN_aarch64_late_size_sections): Emit DT_AARCH64_MEMTAG_MODE
dynamic tag.
* bfd/elfxx-aarch64.h: New definition for the various memtag
properties.
* binutils/readelf.c (get_aarch64_dynamic_type): Handle
DT_AARCH64_MEMTAG_MODE.
* ld/emultempl/aarch64elf.em: Likewise.
* ld/ld.texi: Add documentation for the new option
-z memtag-mode.
* ld/testsuite/ld-aarch64/aarch64-elf.exp: New test.
* ld/testsuite/ld-aarch64/dt-memtag.d: New test.
* ld/testsuite/ld-aarch64/dt-memtag-mode.s: New test.
include/ChangeLog:
* elf/aarch64.h (DT_AARCH64_MEMTAG_MODE): New definition.
As per the DWARF for the Arm 64-bit Architecture (AArch64)
specification, the augmentation char 'G' indicates that associated
frames may modify MTE tags on the stack space they use.
Add knowledge of the 'G' augmentation char to the EH Frame parsing
code.
ChangeLog:
* bfd/elf-eh-frame.c (_bfd_elf_parse_eh_frame): Accommodate
augmentation char 'G'.
* ld/testsuite/ld-aarch64/aarch64-elf.exp: New test.
* ld/testsuite/ld-aarch64/mte-tagged-frame-bar.s: New test.
* ld/testsuite/ld-aarch64/mte-tagged-frame-foo.s: New test.
* ld/testsuite/ld-aarch64/mte-tagged-frame.d: New test.
