Lots of targets already fixed the TEXTREL problem for TLS in PIE.
* For PR ld/25694,
In the check_reloc, refer to spare and loongarch, they don't need to reserve
any local dynamic reloc for TLS LE in pie/pde, and similar to other targets.
So it seems like riscv was too conservative to estimate the TLS LE before.
Just break and don't goto static_reloc for TLS LE in pie/pde can fix the
TEXTREL problem.
* For PR ld/22263,
The risc-v code for TLS GD/IE in the relocate_section seems same as MIPS port.
So similar to MIPS, pr22570, commits 9143e72c6d and 1cb83cac9a, it seems
also the right way to do the same thing for risc-v.
On risc-v, fixes
FAIL: Build pr22263-1
RISC-V haven't supported the TLS transitions, so will need the same fix (use
bfd_link_dll) in the future.
bfd/
PR ld/22263
PR ld/25694
* elfnn-riscv.c (riscv_elf_check_relocs): Replace bfd_link_pic with
bfd_link_dll for TLS IE. Don't need to reserve the local dynamic
relocation for TLS LE in pie/pde, and report error in pic just like
before.
(riscv_elf_relocate_section): For TLS GD/IE, use bfd_link_dll rather
than !bfd_link_pic in determining the dynamic symbol index. Avoid
the index of -1.
It's not worth avoiding an fseek to the same position, and can cause
problems if the linker's output file (which is opened "w+") is read,
because that can result in writing, reading, then writing again.
POSIX.1-2017 (IEEE Std 1003.1) says of fopen:
"When a file is opened with update mode ('+' as the second or third
character in the mode argument), both input and output may be
performed on the associated stream. However, the application shall
ensure that output is not directly followed by input without an
intervening call to fflush() or to a file positioning function
(fseek(), fsetpos(), or rewind()), and input is not directly followed
by output without an intervening call to a file positioning function,
unless the input operation encounters end-of-file."
* bfdio.c (bfd_seek): Always call iovec->bseek.
normally RELA relocs in BFD should not consider the contents of the
relocated place. The aarch64 psABI is even stricter, it specifies
(section 5.7.16) that all RELA relocs _must_ be idempotent.
Since the inception of the aarch64 BFD backend all the relocs have a
non-zero src_mask, and hence break this invariant. It's normally not
a very visible problem as one can see it only when the relocated place
already contains a non-zero value, which usually only happens sometimes
when using 'ld -r' (or as in the testcase when jumping through hoops to
generate the relocations). Or with alternative toolchains that do encode
stuff in the relocated places with the assumption that a relocation
to that place ignores whatever is there (as they can according to
the psABI).
Golang is such a toolchain and https://github.com/golang/go/issues/39927
is ultimately caused by this problem: the testcase testGCData failing
is caused by the garbage collection data-structure to describe a type
containing pointers to be wrong. It's wrong because a field that's
supposed to contain a file-relative offset (to some gcbits) has a
relocation applied and that relocation has an addend which also is
already part of the go-produced object file (so the addend is
implicitely applied twice).
bfd/
PR ld/30437
* elfnn-aarch64.c (elfNN_aarch64_howto_table): Clear src_mask
if all relocation descriptors.
ld/
* testsuite/ld-aarch64/rela-idempotent.s: New testcase.
* testsuite/ld-aarch64/rela-idempotent.d: New.
* testsuite/ld-aarch64/aarch64-elf.exp: Run it.
The fix for pr23686 had a hole in the reloc address sanity check,
the calculation could overflow. Note that stabsize is known to be a
non-zero multiple of 12 so stabsize - 4 can't underflow.
PR 23686
* syms.c (_bfd_stab_section_find_nearest_line): Correct
r->address sanity check.
I started down the path of attempting to fix
https://sourceware.org/pipermail/binutils/2023-April/127263.html but
decided after a while that I didn't want to mess with this code..
This patch is a just a few things that I thought worth doing, the main
one being reporting of errors up the call chain. The while loop to
for loop change is shamelessly stolen from Oleg.
* coffcode.h (handle_COMDAT): Return bool. Make sec_flags a
flagword*, and adjust to suit. Replace while loop with for
loop. Check isym.n_numaux before reading aux entries. Alloc
coff_comdat_info and name in one call to bfd_alloc. Remove
goto breakloop.
(styp_to_sec_flags): Adjust handle_COMDAT call.
The tic54x backend provides its own coff_set_arch_mach, but wants to
use the standard coff_set_section_contents. BFD_JUMP_TABLE_WRITE
defines both of these functions, so the code also provides a wrapper
for coff_set_section_contents. This is all quite OK, but I was on a
mission to remove unnecessary declarations in coffcode.h, and on
deleting the one for coff_set_arch_mach ran into a warning about the
function being unused. I could have kept that declaration with its
ATTRIBUTE_UNUSED or written "static bool ATTRIBUTE_UNUSED" on the
definition but the latter is not usual and looks odd to me. So I
had a closer look at tic54x_set_arch_mach and decided the function is
very likely wrong to allow bfd_arch_unknown. Thus the backend should
be using the standard coff_set_arch_mach.
* coff-tic54x.c: Use BFD_JUMP_TABLE_WRITE (coff) in target vecs.
(tic54x_coff_set_arch_mach): Delete.
(tic54x_set_section_contents): Delete.
* coffcode.h: Delete unnecessary forward declarations.
96d6e190e9
There are some known limitations for now,
* Do not shrink the length of the uleb128 value, even if the value is reduced
after relaxations. Also reports error if the length grows up.
* The R_RISCV_SET_ULEB128 needs to be paired with and be placed before the
R_RISCV_SUB_ULEB128.
bfd/
* bfd-in2.h: Regenerated.
* elfnn-riscv.c (perform_relocation): Perform R_RISCV_SUB_ULEB128 and
R_RISCV_SET_ULEB128 relocations. Do not shrink the length of the
uleb128 value, and report error if the length grows up. Called the
generic functions, _bfd_read_unsigned_leb128 and _bfd_write_unsigned_leb128,
to encode the uleb128 into the section contents.
(riscv_elf_relocate_section): Make sure that the R_RISCV_SET_ULEB128
must be paired with and be placed before the R_RISCV_SUB_ULEB128.
* elfxx-riscv.c (howto_table): Added R_RISCV_SUB_ULEB128 and
R_RISCV_SET_ULEB128.
(riscv_reloc_map): Likewise.
(riscv_elf_ignore_reloc): New function.
* libbfd.h: Regenerated.
* reloc.c (BFD_RELOC_RISCV_SET_ULEB128, BFD_RELOC_RISCV_SUB_ULEB128):
New relocations to support .uleb128 subtraction.
gas/
* config/tc-riscv.c (md_apply_fix): Added BFD_RELOC_RISCV_SET_ULEB128
and BFD_RELOC_RISCV_SUB_ULEB128.
(s_riscv_leb128): Updated to allow uleb128 subtraction.
(riscv_insert_uleb128_fixes): New function, scan uleb128 subtraction
expressions and insert fixups for them.
(riscv_md_finish): Called riscv_insert_uleb128_fixes for all sections.
include/
* elf/riscv.h ((R_RISCV_SET_ULEB128, (R_RISCV_SUB_ULEB128): Defined.
ld/
* testsuite/ld-riscv-elf/ld-riscv-elf.exp: Updated.
* testsuite/ld-riscv-elf/uleb128*: New testcase for uleb128 subtraction.
binutils/
* testsuite/binutils-all/nm.exp: Updated since RISCV supports .uleb128.
Another thing, section target_index is renumbered in
coff_compute_section_file_positions and _bfd_xcoff_bfd_final_link. I
don't know that there is currently any way that the output bfd
section_by_target_index could be populated before this point but
clear them out so no one need worry about it.
* coffcode.h (coff_compute_section_file_positions): Clear
section_by_target_index hash table when changing target_index.
(_bfd_xcoff_bfd_final_link): Likewise.
Commit 0e759f232b regressed these tests:
rs6000-aix7.2 +FAIL: Garbage collection test 1 (32-bit)
rs6000-aix7.2 +FAIL: Garbage collection test 1 (64-bit)
rs6000-aix7.2 +FAIL: Glink test 1 (32-bit)
rs6000-aix7.2 +FAIL: Glink test 1 (64-bit)
Investigation showed segfaults in coff_section_from_bfd_index called
by xcoff_write_global_symbol due to the hash table pointer being
NULL. Well, yes, the hash table isn't initialised for the output bfd.
mkobject_hook is the wrong place to do that.
* coffcode.h: Revert 0e759f232b changes.
* peicode.h: Likewise.
* coff-x86_64.c (htab_hash_section_index, htab_eq_section_index):
Moved here from coffcode.h.
(coff_amd64_rtype_to_howto): Create section_by_index htab.
* coffgen.c (htab_hash_section_target_index),
(htab_eq_section_target_index): Moved here from coffcode.h.
(coff_section_from_bfd_index): Create section_by_target_index
htab. Stash newly created sections in htab.
Trying to build binutils with an older gcc currently fails. Working
around these gcc bugs is not onerous so let's fix them.
bfd/
* elf32-csky.c (csky_elf_size_dynamic_sections): Don't type-pun
pointer.
* elf32-rl78.c (rl78_compute_complex_reloc): Rename "stat"
variable to "status".
gas/
* compress-debug.c (compress_finish): Supply all fields in
ZSTD_inBuffer initialisation.
include/
* xtensa-dynconfig.h (xtensa_isa_internal): Delete unnecessary
forward declaration.
opcodes/
* loongarch-opc.c: Supply all fields of zero struct initialisation
in various opcode tables.
* libcoff-in.h (struct coff_tdata): Add section_by_index and section_by_target_index hash tables.
* libcoff.h: Regenerate.
* coffcode.h (htab_hash_section_index): New function. (htab_eq_section_index): New function. (htab_hash_section_target_index): New function. (htab_eq_section_target_index): New function. (coff_mkobject_hool): Create the hash tables.
* peicode.h: Add the same new functions. (pe_mkobject_hook): Create the hash tables.
* coff-x86_64.c (coff_amd64_rtype_to_howto): Use the new tables to speed up lookups.
* coffgen.c (coff_section_from_bfd_index): Likewise. (_bfd_coff_close_and_cleanup): Delete the hash tables.
If a function symbol only get its address by la.global, without
directly called by bl instruction, the PLT entry is not required.
bfd/ChangeLog:
* elfnn-loongarch.c (loongarch_elf_adjust_dynamic_symbol): Fix PLT
entry generate bug.
ld/ChangeLog:
* testsuite/ld-elf/shared.exp: Clear xfail for LoongArch.
This is one part of the fix for GCC PR109128, along with a
corresponding GCC change. Without this patch, what happens in the
linker, when an unused object in a .a file has offload data, is that
elf_link_is_defined_archive_symbol calls bfd_link_plugin_object_p,
which ends up calling the plugin's claim_file_handler, which then
records the object as one with offload data. That is, the linker never
decides to use the object in the first place, but use of this _p
interface (called as part of trying to decide whether to use the
object) results in the plugin deciding to use its offload data (and a
consequent mismatch in the offload data present at runtime).
The new hook allows the linker plugin to distinguish calls to
claim_file_handler that know the object is being used by the linker
(from ldmain.c:add_archive_element), from calls that don't know it's
being used by the linker (from elf_link_is_defined_archive_symbol); in
the latter case, the plugin should avoid recording the object as one
with offload data.
bfd/
* plugin.c (struct plugin_list_entry): Add claim_file_v2.
(register_claim_file_v2): New.
(try_load_plugin): Use LDPT_REGISTER_CLAIM_FILE_HOOK_V2.
(ld_plugin_object_p): Take second argument.
(bfd_link_plugin_object_p): Update call to ld_plugin_object_p.
(register_ld_plugin_object_p): Update argument prototype.
(bfd_plugin_object_p): Update call to ld_plugin_object_p.
* plugin.h (register_ld_plugin_object_p): Update argument
prototype.
include/
* plugin.api.h (ld_plugin_claim_file_handler_v2)
(ld_plugin_register_claim_file_v2)
(LDPT_REGISTER_CLAIM_FILE_HOOK_V2): New.
(struct ld_plugin_tv): Add tv_register_claim_file_v2.
ld/
* plugin.c (struct plugin): Add claim_file_handler_v2.
(LDPT_REGISTER_CLAIM_FILE_HOOK_V2): New.
(plugin_object_p): Add second argument. Update call to
plugin_call_claim_file.
(register_claim_file_v2): New.
(set_tv_header): Handle LDPT_REGISTER_CLAIM_FILE_HOOK_V2.
(plugin_call_claim_file): Add argument known_used.
(plugin_maybe_claim): Update call to plugin_object_p.
* testplug.c, testplug2.c, testplug3.c, testplug4.c: Handle
LDPT_REGISTER_CLAIM_FILE_HOOK_V2.
* testsuite/ld-plugin/plugin-1.d, testsuite/ld-plugin/plugin-10.d,
testsuite/ld-plugin/plugin-11.d, testsuite/ld-plugin/plugin-13.d,
testsuite/ld-plugin/plugin-14.d, testsuite/ld-plugin/plugin-15.d,
testsuite/ld-plugin/plugin-16.d, testsuite/ld-plugin/plugin-17.d,
testsuite/ld-plugin/plugin-18.d, testsuite/ld-plugin/plugin-19.d,
testsuite/ld-plugin/plugin-2.d, testsuite/ld-plugin/plugin-26.d,
testsuite/ld-plugin/plugin-3.d, testsuite/ld-plugin/plugin-30.d,
testsuite/ld-plugin/plugin-4.d, testsuite/ld-plugin/plugin-5.d,
testsuite/ld-plugin/plugin-6.d, testsuite/ld-plugin/plugin-7.d,
testsuite/ld-plugin/plugin-8.d, testsuite/ld-plugin/plugin-9.d:
Update test expectations.
PR 30354
* elf32-arm.c (elf32_arm_gc_mark_extra_sections): If any debug sections are marked then rerun the extra marking in order to pick up any dependencies.
* coffcode.h (coff_read_word_from_buffer): New function.
* coffcode.h (COFF_CHECKSUM_BUFFER_SIZE): New constant.
* coffcode.h (coff_compute_checksum): Improve speed by reducing the number of seeks and reads used.
bfd/doc/chew.c extracts documentation from source code comments
annotated with keywords, and generates much of bfd.h and libbfd.h from
those same comments. The docs have suffered from people (me too)
adding things like CODE_FRAGMENT to the source to put code into bfd.h
without realising that CODE_FRAGMENT also puts @example around said
code into the docs. So we have random senseless things in the docs.
This patch fixes that problem (well, the senseless things from
CODE_FRAGMENT), moves most of the code out of bfd-in.h, and improves a
few chew.c features. libbfd.h now automatically gets ATTRIBUTE_HIDDEN
prototypes, and indentation in bfd.h and libbfd.h is better.
These functions don't belong in opncls.c.
* libbfd-in.h (bfd_release): Delete prototype.
* opncls.c (bfd_alloc, bfd_zalloc, bfd_release): Move to..
* libbfd.c: ..here. Include objalloc.c and provide bfd_release
with a FUNCTION comment.
* bfd-in2.h: Regenerate.
* libbfd.h: Regenerate.
