This was broken when I changed how we compute the value for the gp register.
It used to be computed inside the sdata section. Now it is computed at the
end which makes it an abs section symbol. There is code that tries to use
the alignment of the section that the gp value is in, but this does not work
if it is in the abs section, as the abs section has alignment of 1 byte.
There are people using alternative linker scripts that still define it in the
sdata section, so the code is still useful. Thus adding a check to disable
this when gp is in the abs section.
bfd/
* elfnn-riscv.c (_bfd_riscv_relax_lui): Add check to exclude abs
section when setting max_alignment. Update comment.
(_bfd_riscv_relax_pc): Likewise.
BFD was leaking memory in bfd_check_format_matches. As part of
deciding the proper format of an archive, BFD looks at the format of
the first file stored. That file's bfd was left open for reasons
given in a comment removed in git commit 0e71e4955c that said:
/* We ought to close `first' here, but we can't, because
we have no way to remove it from the archive cache.
It's close to impossible to figure out when we can
release bfd_ardata. FIXME. */
Well, things have changed since that comment was true and we now can
remove files from the archive cache. Closing the first file is good
and cures some of the leaks. Other leaks are caused by
bfd_check_format_matches throwing away bfd tdata before trying a new
match. That lost the element cache set up when format checking the
first element in the archive. The easiest and cleanest fix is to
simply disable the caching when checking the first element.
PR 24891
* bfd.c (struct bfd): Add no_element_cache.
* archive.c (_bfd_get_elt_at_filepos): Don't add element to
archive cache when no_element_cache.
(bfd_generic_archive_p): Set no_element_cache when opening first
element to check format. Close first element too.
(do_slurp_bsd_armap): Don't zero ardata->cache here.
* bfd-in2.h: Regenerate.
Found on a GOT reference to __ehdr_start, which is tweaked to be
undefined weak at some stages of linking. SYMBOL_REFERENCES_LOCAL
isn't a sufficient test.
* elf64-ppc.c (ppc64_elf_edit_toc): Exclude undefined weak
symbols from GOT optimisation.
The PLT GOT entry should point to the first PLT entry which contains the
runtime linker function. It was pointing back to the symbol PLT entry
causing an infinite loop.
I found this when testing the OpenRISC glibc port which uses the runtime
dynamic linker. It seems other libc's we use so far have not been
making use of the initial PLT GOT entries.
bfd/ChangeLog:
* elf32-or1k.c (or1k_elf_finish_dynamic_symbol): Use correct value for
PLT GOT entries.
PR 24456
* elf.c (bfd_section_from_shdr): Issue an informative warning
message and continue processing other sections after encountering
a reloc section for a section which already has other relocs
associated with it.
This patch adds support for following CPUs:
Cortex-M35P, Cortex-A77, Cortex-A76AE.
Related specifications can be found at https://developer.arm.com/ip-products/processors.
gas/ChangeLog:
* config/tc-arm.c: New entries for Cortex-M35P, Cortex-A77,
and Cortex-A76AE.
* doc/c-arm.texi: Document new processors.
* testsuite/gas/arm/cpu-cortex-a76ae.d: New test.
* testsuite/gas/arm/cpu-cortex-a77.d: New test.
* testsuite/gas/arm/cpu-cortex-m35p.d: New test.
bfd/ChangeLog:
* cpu-arm.c: New entries for Cortex-M35P, Cortex-A77, Cortex-A76AE.
This patch fixes a few linker crashes due to TLS code reaching an assert when it
shouldn't.
The first scenario is with weak TLS symbols that remain weak during linking. In
this case the mid-end would not have seen a TLS symbol and so wouldn't have
allocated the TLS section. We currently assert here and the linker crashes with
a not very useful message.
This patch changes this to return the value 0 for the TLS symbol in question
emulating what lld and gold and other BFD targets do. However because weak TLS
is implementation defined and we don't define any behavior for it I also emit a
warning to the user to inform them of such.
Secondly when a strong TLS reference is undefined. The linker crashes even after
it correctly reported that there is an undefined reference. This changes it so
that it gracefully exits and reports a useful error.
bfd/ChangeLog:
PR ld/24601
* elfnn-aarch64.c (aarch64_relocate): Handle weak TLS and undefined TLS.
Also Pass input_bfd to _bfd_aarch64_elf_resolve_relocation.
* elfxx-aarch64.c (_bfd_aarch64_elf_resolve_relocation): Use it.
* elfxx-aarch64.h (_bfd_aarch64_elf_resolve_relocation): Emit warning
for weak TLS.
ld/ChangeLog:
PR ld/24601
* testsuite/ld-aarch64/aarch64-elf.exp (undef-tls, weak-tls): New.
* testsuite/ld-aarch64/undef-tls.d: New test.
* testsuite/ld-aarch64/undef-tls.s: New test.
* testsuite/ld-aarch64/weak-tls.d: New test.
* testsuite/ld-aarch64/weak-tls.s: New test.
This patch removes use of st_target_internal to cache the result of
comparing symbol names against CMSE_PREFIX. The problem with setting
a bit in st_target_internal in swap_symbol_in is that calling
bfd_elf_sym_name from swap_symbol_in requires symtab_hdr, and you
don't know for sure whether swap_symbol_in is operating on dynsyms
(and thus elf_tdata (abfd)->dynsymtab_hdr should be used) or on the
normal symtab (thus elf_tdata (abfd)->symtab_hdr). You can make an
educated guess based on abfd->flags & DYNAMIC but that relies on
knowing a lot about calls to bfd_elf_get_elf_syms, and is fragile in
the face of possible future changes.
include/
* elf/arm.h (ARM_GET_SYM_CMSE_SPCL, ARM_SET_SYM_CMSE_SPCL): Delete.
bfd/
* elf32-arm.c (cmse_scan): Don't use ARM_GET_SYM_CMSE_SPCL,
instead recognize CMSE_PREFIX in symbol name.
(elf32_arm_gc_mark_extra_sections): Likewise.
(elf32_arm_filter_cmse_symbols): Don't test ARM_GET_SYM_CMSE_SPCL.
(elf32_arm_swap_symbol_in): Don't invoke ARM_SET_SYM_CMSE_SPCL.
Related specifications can be found at
https://developer.arm.com/ip-products/processors.
gas * NEWS: Mention the Arm and AArch64 new processors.
* config/tc-aarch64.c: New entries for Cortex-A34, Cortex-A65,
Cortex-A77, cortex-A65AE, and Cortex-A76AE.
* doc/c-aarch64.texi: Document new CPUs.
* testsuite/gas/aarch64/cpu-cortex-a34.d: New test.
* testsuite/gas/aarch64/cpu-cortex-a65.d: New test.
* testsuite/gas/aarch64/cpu-cortex-a65ae.d: New test.
* testsuite/gas/aarch64/cpu-cortex-a76ae.d: New test.
* testsuite/gas/aarch64/cpu-cortex-a77.d: New test.
* testsuite/gas/aarch64/nop-asm.s: New test.
bfd * cpu-aarch64.c: New entries for Cortex-A34, Cortex-A65,
Cortex-A77, cortex-A65AE, and Cortex-A76AE.
We currently use a padding NOP after a Thumb to Arm interworking veneer (BX pc).
The NOP is never executed but may result in a performance penalty on some cores.
For this reason this patch changes the NOPs after Thumb to Arm veneers into B .-2
and adds a note to this in the source code for future reference.
bfd/ChangeLog:
* elf32-arm.c (elf32_thumb2_plt_entry, elf32_arm_plt_thumb_stub,
elf32_arm_stub_long_branch_v4t_thumb_thumb,
elf32_arm_stub_long_branch_v4t_thumb_arm,
elf32_arm_stub_short_branch_v4t_thumb_arm,
elf32_arm_stub_long_branch_v4t_thumb_arm_pic,
elf32_arm_stub_long_branch_v4t_thumb_thumb_pic,
elf32_arm_stub_long_branch_v4t_thumb_tls_pic): Change nop to branch to
previous instruction.
ld/ChangeLog:
* testsuite/ld-arm/cortex-a8-fix-b-plt.d: Update Testcase.
* testsuite/ld-arm/cortex-a8-fix-b-rel-arm.d: Likewise.
* testsuite/ld-arm/cortex-a8-fix-bcc-plt.d: Likewise.
* testsuite/ld-arm/farcall-cond-thumb-arm.d: Likewise.
* testsuite/ld-arm/farcall-mixed-app.d: Likewise.
* testsuite/ld-arm/farcall-mixed-app2.d: Likewise.
* testsuite/ld-arm/farcall-mixed-lib-v4t.d: Likewise.
* testsuite/ld-arm/farcall-thumb-arm-pic-veneer.d: Likewise.
* testsuite/ld-arm/farcall-thumb-arm-short.d: Likewise.
* testsuite/ld-arm/farcall-thumb-arm.d: Likewise.
* testsuite/ld-arm/farcall-thumb-thumb-pic-veneer.d: Likewise.
* testsuite/ld-arm/farcall-thumb-thumb.d: Likewise.
* testsuite/ld-arm/fix-arm1176-on.d: Likewise.
* testsuite/ld-arm/ifunc-10.dd: Likewise.
* testsuite/ld-arm/ifunc-2.dd: Likewise.
* testsuite/ld-arm/ifunc-4.dd: Likewise.
* testsuite/ld-arm/ifunc-6.dd: Likewise.
* testsuite/ld-arm/ifunc-8.dd: Likewise.
* testsuite/ld-arm/jump-reloc-veneers-long.d: Likewise.
* testsuite/ld-arm/mixed-app.d: Likewise.
* testsuite/ld-arm/thumb2-b-interwork.d: Likewise.
* testsuite/ld-arm/tls-longplt.d: Likewise.
* testsuite/ld-arm/tls-thumb1.d: Likewise.
A customer reported a case where addr2line was very slow. We tracked
this down to some N^2 behavior in _bfd_dwarf2_find_symbol_bias in the
unusual case where no function can be found.
This patch fixes the bug, and reduces the runtime for a particular
request from 127 seconds to 1 second.
bfd/ChangeLog
2019-08-19 Tom Tromey <tromey@adacore.com>
* dwarf2.c (_bfd_dwarf2_find_symbol_bias): Create hash table
holding symbols.
These are done in ppc64_elf_edit_toc, which now also garbage collects
unused GOT entries. The checks for legitimate instructions weren't
being done for the GOT relocs, unless the file also happened to have a
toc section.
* elf64-ppc.c (struct ppc64_elf_obj_tdata): Rename has_gotrel
to has_optrel.
(struct _ppc64_elf_section_data): Likewise.
(ppc64_elf_check_relocs): Set has_optrel for more relocs.
(ppc64_elf_edit_toc): Do ha/lo insn checks in GOT loop rather
than TOC loop. Check PLT16 insns too.
commit 83924b3846
Author: H.J. Lu <hjl.tools@gmail.com>
Date: Tue Feb 5 18:45:23 2019 -0800
x86-64: Restore PIC check for PCREL reloc against protected symbol
moved PIC check for PC-relative relocations to elf_x86_64_check_relocs.
Since linker defined symbols may not be processed at the time, we need
to move the check back to elf_x86_64_relocate_section.
bfd/
PR ld/24905
* elf64-x86-64.c (elf_x86_64_check_relocs): Move PIC check for
PC-relative relocations back to ...
(elf_x86_64_relocate_section): Here.
ld/
PR ld/24905
* testsuite/ld-x86-64/pr24905-x32.d: New file.
* testsuite/ld-x86-64/pr24905.d: Likewise.
* testsuite/ld-x86-64/pr24905.s: Likewise.
* testsuite/ld-x86-64/pr24905.t: Likewise.
* testsuite/ld-x86-64/x86-64.exp: Run pr24905 and pr24905-x32.
2019-08-16 Martin Liska <mliska@suse.cz>
PR ld/24912
* elflink.c: Report error only for not relocatable.
* linker.c (_bfd_generic_link_add_one_symbol): Do not handle
here lto_slim_object as it's handled in caller.
2019-08-16 Martin Liska <mliska@suse.cz>
PR ld/24912
* object.cc (big_endian>::do_layout): Do not report error,
but only set a flag.
(big_endian>::do_add_symbols): Report error only for when
relocatable.
Given R_PPC_ADDR32 or R_PPC_UADDR32 relocs, this patch generates
R_PPC_ADDR32 or R_PPC_UADDR32 dynamic relocs from either type
depending on whether r_offset is 4-byte aligned, and similarly for
R_PPC_ADDR16/R_PPC_UADDR16.
* elf32-ppc.c (ppc_elf_relocate_section): Optimize unaligned relocs.
This fixes a problem originally reported at
https://github.com/riscv/riscv-binutils-gdb/issues/173
If you have code linked at address zero, you can have a lui instruction
loading a value 0x800 which gets relaxed to a c.lui which is valid (c.lui 0x1
followed by addi -0x800). Relaxation can reduce the value below 0x800 at which
point the c.lui 0x0 is no longer valid. We can fix this by converting the
c.lui to a c.li which can load 0.
bfd/
* elfnn-riscv.c (perform_relocation) <R_RISCV_RVC_LUI>: If
RISCV_CONST_HIGH_PART (value) is zero, then convert c.lui instruction
to c.li instruction, and use ENCODE_RVC_IMM to set value.
ld/
* testsuite/ld-riscv-elf/c-lui-2.d: New.
* testsuite/ld-riscv-elf/c-lui-2.ld: New.
* testsuite/ld-riscv-elf/c-lui-2.s: New.
* testsuite/ld-riscv-elf/ld-riscv-elf.exp: Run the c-lui-2 test.
A customer program had a DWARF CU that consisted of just a CU DIE,
without any children. In this situation, scan_unit_for_symbols will
try to read past the end of the current CU, and will take use the
first bytes of the next CU as an abbrev, printing an error message.
This patch fixes the bug by changing scan_unit_for_symbols to stop at
the end of the CU rather than the end of the .debug_info section.
bfd/ChangeLog
2019-08-15 Tom Tromey <tromey@adacore.com>
* dwarf2.c (scan_unit_for_symbols): Check for end of CU, not end
of section.
PR 24623
* dwarf2.c (stash_comp_unit): New function, extracted from..
(_bfd_dwarf2_find_nearest_line): ..here.
(find_abstract_instance): Parse comp units and decode line info
as needed.
This parameter might appear to be used to set up offset_size, but
since git commit 024b2372f5 offset_size is either set from the
debug_info data or is set to 4.
* dwarf2.c (_bfd_dwarf2_find_nearest_line): Remove addr_size parameter.
* libbfd-in.h (_bfd_dwarf2_find_nearest_line): Update prototype.
* coffgen.c (coff_find_nearest_line_with_names): Adjust
_bfd_dwarf2_find_nearest_line calls.
* elf.c (_bfd_elf_find_nearest_line, _bfd_elf_find_line): Likewise.
* elf32-arm.c (elf32_arm_find_nearest_line): Likewise.
* elf64-alpha.c (elf64_alpha_find_nearest_line): Likewise.
* elfnn-aarch64.c (elfNN_aarch64_find_nearest_line): Likewise.
* elfxx-mips.c (_bfd_mips_elf_find_nearest_line): Likewise.
* mach-o.c (bfd_mach_o_find_nearest_line): Likewise.
* libbfd.h: Regenerate.
This patch is a reimplementation of [1] which was submitted in 2015 by
Neil Schellenberger. Copyright issue was sorted out [2] last year.
It proposed a new section (.gnu.xhash) and related dynamic tag
(DT_GNU_XHASH). The new section would be virtually identical to the
existing .gnu.hash except for the translation table (xlat) which would
contain correct MIPS .dynsym indexes corresponding to the hashvals in
chains. This is because MIPS ABI imposes a different ordering on the
dynsyms than the one expected by the .gnu.hash section. Another addition
would be a leading word (ngnusyms) which would contain the number of
entries in the translation table.
In this patch, the new section name and dynamic tag are changed to
reflect the fact that the section should be treated as MIPS-specific
(.MIPS.xhash and DT_MIPS_XHASH).
This patch addresses the alignment issue as reported in [3], which is
caused by the leading word added to the .MIPS.xhash section. Leading word
is removed in this patch, and the number of entries in the translation
table is now calculated using DT_MIPS_SYMTABNO dynamic tag (this is
addressed by the corresponding glibc patch).
Suggestions on coding style in [4] were taken into account. Existing
GNU hash testcase was covered, and another one was added in the MIPS
part of the testsuite.
The other major change is reserving MIPS ABI version 5 for .MIPS.xhash,
marking the need of support for .MIPS.xhash in the dynamic linker (again,
addressed in the corresponding glibc patch). This is something which I
am not sure of, especially after reading [5]. I am confused on whether
this ABI version is reserved for IFUNC, or it can be used for this
purpose.
Already mentioned glibc patch is submitted at:
https://sourceware.org/ml/libc-alpha/2019-06/msg00456.html
[1] https://sourceware.org/ml/binutils/2015-10/msg00057.html
[2] https://sourceware.org/ml/binutils/2018-03/msg00025.html
[3] https://sourceware.org/ml/binutils/2016-01/msg00006.html
[4] https://sourceware.org/ml/binutils/2016-02/msg00097.html
[5] https://sourceware.org/ml/libc-alpha/2016-12/msg00853.html
ld * emulparams/elf32bmip.sh: Add .MIPS.xhash section.
* emulparams/elf32bmipn32-defs.sh: Add .MIPS.xhash section.
* emulparams/elf64bmip-defs.sh: Add .MIPS.xhash section.
* emultempl/mipself.em: Remove mips_after_parse function.
* testsuite/ld-elf/hash.d: Update comment.
* testsuite/ld-mips-elf/hash1.d: New test.
* testsuite/ld-mips-elf/hash1.s: Ditto.
* testsuite/ld-mips-elf/hash1a.d: Remove.
* testsuite/ld-mips-elf/hash1b.d: Ditto.
* testsuite/ld-mips-elf/hash1c.d: Ditto
* testsuite/ld-mips-elf/hash2.d: New test.
* testsuite/ld-mips-elf/mips-elf.exp: New tests.
* testsuite/ld-mips-elf/start.s: New test.
bfd * elf-bfd.h (struct elf_backend_data): New members.
* elflink.c (_bfd_elf_link_create_dynamic_sections): Create
.gnu.hash section if necessary.
(struct collect_gnu_hash_codes): New member.
(elf_gnu_hash_process_symidx): New function name.
(elf_renumber_gnu_hash_syms): Ignore local and undefined
symbols. Record xlat location for every symbol which should have
a .MIPS.xhash entry.
(bfd_elf_size_dynamic_sections): Add DT_GNU_HASH dynamic tag to
dynamic section if necessary.
(GNU_HASH_SECTION_NAME): New define.
(bfd_elf_size_dynsym_hash_dynstr): Get .MIPS.xhash section.
Update the section size info.
* elfxx-mips.c (struct mips_elf_hash_sort_data): New members.
(struct mips_elf_link_hash_entry): New member.
(mips_elf_link_hash_newfunc): Initialize .MIPS.xhash translation
table location.
(mips_elf_sort_hash_table): Initialize the pointer to the
.MIPS.xhash section.
(mips_elf_sort_hash_table_f): Populate the .MIPS.xhash
translation table entry with the symbol dynindx.
(_bfd_mips_elf_section_from_shdr): Add SHT_MIPS_XHASH.
(_bfd_mips_elf_fake_sections): Initialize .MIPS.xhash section
info.
(_bfd_mips_elf_create_dynamic_sections): Create .MIPS.xhash
section.
(_bfd_mips_elf_size_dynamic_sections): Add DT_MIPS_XHASH tag to
dynamic section.
(_bfd_mips_elf_finish_synamic_sections): Add DT_MIPS_XHASH.
(_bfd_mips_elf_final_write_processing): Set .MIPS.xhash section
sh_link info.
(_bfd_mips_elf_get_target_dtag): Get DT_MIPS_XHASH tag.
(MIPS_LIBC_ABI_XHASH): New ABI version enum value.
(_bfd_mips_post_process_headers): Mark the ABI version as
MIPS_LIBC_ABI_XHASH if there exists a .MIPS.xhash section,
but not a .hash section.
(_bfd_mips_elf_record_xhash_symbol): New function. Record a
position in the translation table, associated with the hash
entry.
* elfxx-mips.h (literal_reloc_p): Define
elf_backend_record_xhash_symbol backend hook.
* elfxx-target.h: Initialize elf_backend_record_xhash_symbol
backend hook.
include * elf/mips.h (SHT_GNU_XHASH): New define.
(DT_GNU_XHASH): New define.
binutils * readelf.c (get_mips_dynamic_type): Return MIPS_XHASH dynamic type.
(get_mips_section_type_name): Return MI{S_XHASH name string.
(dynamic_section_mips_val): Initialize the .MIPS.xhash dynamic
info.
(process_symbol_table): Initialize the .MIPS.xhash section
pointer. Adjust the readelf output to support the new section.
(process_object): Set the .MIPS.xhash dynamic info to zero.
This patch changes the eBPF linker to provide a relocate_section
function instead of relying on using special functions in relocation
howtos.
Tested in x86_64 host.
No regressions.
bfd/ChangeLog:
2019-08-07 Jose E. Marchesi <jose.marchesi@oracle.com>
* elf64-bpf.c (bpf_elf_relocate_section): New function.
(bpf_elf_insn_disp_reloc): Delete function.
(elf_backend_relocate_section): Define.
bfd/ChangeLog
Ilia Diachkov <ilia.diachkov@optimitech.com>
* elfnn-riscv.c (_bfd_riscv_relax_lui): Set lui relax safety area to
two pages in relro presence.
check_loop_aligned is used during link time relaxation to only allow
transformations that don't violate loop body alignment requirements.
Assembler can relax loops that have too long body by adding instructions
between the loop instruction and the loop body. check_loop_aligned must
check alignment of the first instruction of the actual loop body.
Detect loop / rsr.lend / wsr.lbeg sequence used in assembly time
relaxation and adjust alignment check when it's detected.
bfd/
2019-08-01 Max Filippov <jcmvbkbc@gmail.com>
* elf32-xtensa.c (insn_num_slots, get_rsr_lend_opcode)
(get_wsr_lbeg_opcode): New functions.
(check_loop_aligned): Detect relaxed loops and adjust loop_len
and insn_len for the first actual instruction of the loop.
This fixes two cases where elf_link_add_object_symbols returns an
error, setting the catch-all bfd_error_bad_value without explaining
the error. The second one is an internal error that can only be
caused by a target elf_add_symbol_hook, so make that one abort. The
first one is my PR24339 fix. PR24339 is another of those fuzzing bugs
and the fix I made catches the problem when loading symbols, rather
than when symbols are used in relocs. While ld is correct to reject
the object file as not complying with the ELF standard, let's be a
little more forgiving for dynamic objects.
PR 24857
PR 24339
* elflink.c (elf_link_add_object_symbols): Report an informative
error on finding local symbols with index equal or greater than
symbol table sh_info. Correct comment. Allow such symbols in
dynamic objects. Abort on NULL section for symbol.
This field effectively became usused a long time ago, perhaps as early
as 1994.
* elf-bfd.h (struct output_elf_obj_tdata): Delete "linker" field.
(elf_linker): Don't define.
* elflink.c (bfd_elf_final_link): Don't set elf_linker.
Add linker relaxation. The first relaxation added is converting
GOTPC32 to PCREL relocations. This relaxation doesn't change the size of
the binary.
bfd/
xxxx-xx-xx Claudiu Zissulescu <claziss@synopsys.com>
* elf32-arc.c (bfd_get_32_me): New function.
(bfd_put_32_me): Likewise.
(arc_elf_relax_section): Likewise.
(bfd_elf32_bfd_relax_section): Define.
ld/testsuite/
xxxx-xx-xx Claudiu Zissulescu <claziss@synopsys.com>
* ld-arc/relax-local-pic.d: New test.
* ld-arc/relax-local-pic.s: New file.
I missed some early exits from final_write_processing that mean
_bfd_elf_final_write_processing could be missed.
* elf-vxworks.c (elf_vxworks_final_write_processing): Don't return
early.
* elf32-arc.c (arc_elf_final_write_processing): Likewise.
* elf32-xtensa.c (elf_xtensa_final_write_processing): Likewise.
When SHF_GNU_MBIND was added in the SHF_LOOS to SHF_HIOS range, it
should have required ELFOSABI_GNU since these flags are already in use
by other OSes. HPUX SHF_HP_TLS in fact has the same value. That
means no place in binutils should test SHF_GNU_MBIND without first
checking OSABI, and SHF_GNU_MBIND should not be set without also
setting OSABI. At least, that's the ideal, but the patch accepts
SHF_GNU_MBIND on ELFOSABI_NONE object files since gas didn't always
set OSABI. However, to reinforce the fact that SHF_GNU_MBIND isn't
proper without a non-zero OSABI, readelf will display the flag as
LOOS+0 if OSABI isn't set.
The clash with SHF_HP_TLS means that hppa64-linux either has that flag
on .tbss sections or supports GNU_MBIND, not both. (hppa64-linux
users, if there are any, may have noticed that GNU ld since 2017
mysteriously aligned their .tbss sections to a 4k boundary. That was
one consequence of SHF_HP_TLS being blindly interpreted as
SHF_GNU_MBIND.) Since it seems that binutils, gdb, gcc, glibc, and
the linux kernel don't care about SHF_HP_TLS I took that flag out of
.tbss for hppa64-linux.
bfd/
* elf-bfd.h (enum elf_gnu_osabi): Add elf_gnu_osabi_mbind.
* elf.c (_bfd_elf_make_section_from_shdr): Set elf_gnu_osabi_mbind.
(get_program_header_size): Formatting. Only test SH_GNU_MBIND
when elf_gnu_osabi_mbind is set.
(_bfd_elf_map_sections_to_segments): Likewise.
(_bfd_elf_init_private_section_data): Likewise.
(_bfd_elf_final_write_processing): Update comment.
* elf64-hppa.c (elf64_hppa_special_sections): Move .tbss entry.
(elf_backend_special_sections): Define without .tbss for linux.
binutils/
* readelf.c (get_parisc_segment_type): Split off hpux entries..
(get_ia64_segment_type): ..and these..
(get_hpux_segment_type): ..to here.
(get_segment_type): Condition GNU_MBIND on osabi. Use
get_hpux_segment_type.
(get_symbol_binding): Do not print UNIQUE for ELFOSABI_NONE.
(get_symbol_type): Do not print IFUNC for ELFOSABI_NONE.
gas/
* config/obj-elf.c (obj_elf_change_section): Don't emit a fatal
error for non-SHF_ALLOC SHF_GNU_MBIND here.
(obj_elf_parse_section_letters): Return SHF_GNU_MBIND in new
gnu_attr param.
(obj_elf_section): Adjust obj_elf_parse_section_letters call.
Formatting. Set SHF_GNU_MBIND and elf_osabi from gnu_attr.
Emit normal error for non-SHF_ALLOC SHF_GNU_MBIND and wrong osabi.
(obj_elf_type): Set elf_osabi for ifunc.
* testsuite/gas/elf/section12a.d: xfail msp430 and hpux.
* testsuite/gas/elf/section12b.d: Likewise.
* testsuite/gas/elf/section13.d: Likewise.
* testsuite/gas/elf/section13.l: Adjust expected error.
ld/
* emultempl/elf32.em (gld${EMULATION_NAME}_place_orphan): Condition
SHF_GNU_MBIND on osabi. Set output elf_gnu_osabi_mbind.
This patch supports using pcrel instructions in TLS code sequences. A
number of new relocations are needed, gas operand modifiers to
generate those relocations, and new TLS optimisation. For
optimisation it turns out that the new pcrel GD and LD sequences can
be distinguished from the non-pcrel GD and LD sequences by there being
different relocations on the new sequence. The final "add ra,rb,13"
on IE sequences similarly needs a new relocation, or as I chose, a
modification of R_PPC64_TLS. On pcrel IE code, the R_PPC64_TLS points
one byte into the "add" instruction rather than being on the
instruction boundary.
GD:
pla 3,z@got@tlsgd@pcrel # R_PPC64_GOT_TLSGD34
bl __tls_get_addr@notoc(z@tlsgd) # R_PPC64_TLSGD and R_PPC64_REL24_NOTOC
edited to IE
pld 3,z@got@tprel@pcrel
add 3,3,13
edited to LE
paddi 3,13,z@tprel
nop
LD:
pla 3,z@got@tlsld@pcrel # R_PPC64_GOT_TLSLD34
bl __tls_get_addr@notoc(z@tlsld) # R_PPC64_TLSLD and R_PPC64_REL24_NOTOC
..
paddi 9,3,z2@dtprel
pld 10,z3@got@dtprel@pcrel
add 10,10,3
edited to LE
paddi 3,13,0x1000
nop
IE:
pld 9,z@got@tprel@pcrel # R_PPC64_GOT_TPREL34
add 3,9,z@tls@pcrel # R_PPC64_TLS at insn+1
ldx 4,9,z@tls@pcrel
lwax 5,9,z@tls@pcrel
stdx 5,9,z@tls@pcrel
edited to LE
paddi 9,13,z@tprel
nop
ld 4,0(9)
lwa 5,0(9)
std 5,0(9)
LE:
paddi 10,13,z@tprel
include/
* elf/ppc64.h (R_PPC64_TPREL34, R_PPC64_DTPREL34),
(R_PPC64_GOT_TLSGD34, R_PPC64_GOT_TLSLD34),
(R_PPC64_GOT_TPREL34, R_PPC64_GOT_DTPREL34): Define.
(IS_PPC64_TLS_RELOC): Include new tls relocs.
bfd/
* reloc.c (BFD_RELOC_PPC64_TPREL34, BFD_RELOC_PPC64_DTPREL34),
(BFD_RELOC_PPC64_GOT_TLSGD34, BFD_RELOC_PPC64_GOT_TLSLD34),
(BFD_RELOC_PPC64_GOT_TPREL34, BFD_RELOC_PPC64_GOT_DTPREL34),
(BFD_RELOC_PPC64_TLS_PCREL): New pcrel tls relocs.
* elf64-ppc.c (ppc64_elf_howto_raw): Add howtos for pcrel tls relocs.
(ppc64_elf_reloc_type_lookup): Translate pcrel tls relocs.
(must_be_dyn_reloc, dec_dynrel_count): Add R_PPC64_TPREL64.
(ppc64_elf_check_relocs): Support pcrel tls relocs.
(ppc64_elf_tls_optimize, ppc64_elf_relocate_section): Likewise.
* bfd-in2.h: Regenerate.
* libbfd.h: Regenerate.
gas/
* config/tc-ppc.c (ppc_elf_suffix): Map "tls@pcrel", "got@tlsgd@pcrel",
"got@tlsld@pcrel", "got@tprel@pcrel", and "got@dtprel@pcrel".
(fixup_size, md_assemble): Handle pcrel tls relocs.
(ppc_force_relocation, ppc_fix_adjustable): Likewise.
(md_apply_fix, tc_gen_reloc): Likewise.
ld/
* testsuite/ld-powerpc/tlsgd.d,
* testsuite/ld-powerpc/tlsgd.s,
* testsuite/ld-powerpc/tlsie.d,
* testsuite/ld-powerpc/tlsie.s,
* testsuite/ld-powerpc/tlsld.d,
* testsuite/ld-powerpc/tlsld.s: New tests.
* testsuite/ld-powerpc/powerpc.exp: Run them.
It's not correct to use non-STT_TLS symbols with TLS relocation, not
that it matters much when editing relocs, but this edited reloc can be
output by --emit-relocs. So don't use a symbol on the reloc.
* elf64-ppc.c (ppc64_elf_relocate_section): Don't bother selecting
a TLS section symbol for edited relocs. Tighten TLS symbol/reloc
match test.
This saves a bit in tls_mask, and fixes a bug that could be triggered
in the unlikely case that both @got (usual ELF style) and @toc
(PowerOpen style) code was used to set up args for __tls_get_addr.
* elf64-ppc.c (TLS_EXPLICIT): Define as 256.
(ppc64_elf_check_relocs): Don't store TLS_EXPLICIT even if char
is more than 8 bits.
(ppc64_elf_tls_optimize): Likewise. Make tls_set, tls_clear, and
tls_type vars unsigned int.
(ppc64_elf_relocate_section): Use r_type rather than TLS_EXPLICIT
to select r_type edit.
Choose a better name, that reflects why the flag is set (GD to IE
optimisation) rather than what the flag produces (TPREL64 reloc on
a single GOT entry replacing a tls_index pair).
* elf32-ppc.c (TLS_GDIE): Rename from TLS_TPRELGD throughout file.
Correct comment.
* elf64-ppc.c (TLS_GDIE): Likewise.
I don't expect anyone will have hit this bug. You'd need a TLS
segment of 2G before you'd notice.
* elf64-ppc.c (ppc64_elf_tls_optimize): Correct test for allowed
range of tp-relative offsets.
> Building LLVM 6.0 on FreeBSD/powerpc (devel/llvm60 port) the assertion
> in the subject trips (displays twice) when linking libLTO.so.1. The
> issue has been filed in FreeBSD's bugzilla, at
> https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=237068 . It appears
> the 'llvm::hashing::detail::get_execution_seed()::seed@@JL_LLVM_6.0'
> symbol is being weakly aliased to an indirect symbol
> __bss_start@@JL_LLVM_6.0. Since __bss_start@@JL_LLVM_6.0 is an
> indirect symbol, it fails the assertion.
I haven't looked under a debugger at your testcase but I think I know
what is going on here. You have a shared library with a weakly
defined llvm::hashing::detail::get_execution_seed()::seed which
happens to be at the same location as __bss_start in that library. At
the time the linker loads symbols for that library, it sees they are
both versioned and thus introduces non-versioned indirect symbols for
them. The linker considers the symbols as possibly being aliases,
setting up h->u.alias and h->is_weakalias such that
__bss_start@@JL_LLVM_6.0 is the definition. No real problem so far,
the definition is bfd_link_hash_defined, except that the zero size, no
type __bss_start symbol possibly should not be considered an alias in
the first place.
Later, __bss_start as defined by the linker script is entered into the
linker symbol table. This is similar to __bss_start being defined by
a regular object file in that ELF symbol resolution rules say that the
value of __bss_start in the library is overridden by __bss_start in
the executable/library being produced. So to accomplish the override,
ld flips __bss_start from being an indirect symbol pointing at
__bss_start@@JL_LLVM_6.0 to __bss_start@@JL_LLVM_6.0 being an indirect
symbol pointing at __bss_start. That's how we get an unexpected
indirect symbol and hit the assert.
What should happen I think, is for the def->def_regular code above the
assert to run in this case. The symbols are no longer aliases.
* elflink.c (_bfd_elf_fix_symbol_flags): If the def for an
alias is no longer bfd_link_hash_defined, clear the alias.
It is possible to create shared libraries on PowerPC using
-ftls-model=inital-exec or -ftls-model=local-exec. The first is half
reasonable, getting you a shared library that can't be dlopen'd but
otherwise is reasonable. The second is quite bad. Not only do you
lose being able to dlopen, the library also has dynamic text
relocations. Worse, the TPREL16_LO, TPREL16_HA and other TPREL16
dynamic relocs emitted were wrong, resulting in wrong values being
applied by ld.so.
Using the first TLS section symbol in dynamic relocations for local
TLS symbols doesn't work. It's wrong because TLS symbols used by TLS
relocs have values relative to the TLS segment, whereas the TLS
section symbols are addresses. This patch instead uses a symbol index
of zero which is used elsewhere by PowerPC on dynamic TLS relocs.
It's not strictly ABI compliant to use a non-TLS symbol with TLS
relocs but symbol index zero can be interpreted as "no symbol". Not
using the first TLS section symbol means it doesn't need to be dynamic.
The patch also fixes a further problem with PowerPC32 dynamic TPREL16*
relocs, which shouldn't have the symbol value in the addend as we do
for non-TLS symbols.
bfd/
* elflink.c (_bfd_elf_omit_section_dynsym_default): Don't keep
tls_sec.
(_bfd_elf_init_1_index_section): Prefer not using TLS sections.
(_bfd_elf_init_2_index_sections): Likewise.
* elf64-ppc.c (ppc64_elf_relocate_section): When emitting dynamic
relocations for local TLS symbols, use STN_UNDEF as the relocation
symbol.
* elf32-ppc.c (ppc_elf_relocate_section): Likewise, and don't
leave TLS symbol value in the addend.
ld/
* testsuite/ld-powerpc/tlsso.r: Update.
* testsuite/ld-powerpc/tlsso32.g: Update.
* testsuite/ld-powerpc/tlsso32.r: Update.
* testsuite/ld-powerpc/tlstocso.r: Update.
* testsuite/ld-cris/tls-dso-dtpoffd2.d: Update.
* testsuite/ld-cris/tls-dso-dtpoffd4.d: Update.
* testsuite/ld-cris/tls-dso-tpoffgotcomm1.d: Update.
* testsuite/ld-cris/tls-gd-1.d: Update.
* testsuite/ld-cris/tls-gd-1h.d: Update.
* testsuite/ld-cris/tls-gd-2.d: Update.
* testsuite/ld-cris/tls-gd-2h.d: Update.
* testsuite/ld-cris/tls-ie-10.d: Update.
* testsuite/ld-cris/tls-ie-11.d: Update.
* testsuite/ld-cris/tls-ie-8.d: Update.
* testsuite/ld-cris/tls-ie-9.d: Update.
* testsuite/ld-cris/tls-js1.d: Update.
* testsuite/ld-cris/tls-ld-4.d: Update.
* testsuite/ld-cris/tls-ld-5.d: Update.
* testsuite/ld-cris/tls-ld-6.d: Update.
* testsuite/ld-cris/tls-ld-7.d: Update.
* testsuite/ld-cris/tls-ldgd-14.d: Update.
* testsuite/ld-cris/tls-ldgd-15.d: Update.
* testsuite/ld-cris/tls-ldgdx-14.d: Update.
* testsuite/ld-cris/tls-ldgdx-15.d: Update.
* testsuite/ld-cris/tls-local-54.d: Update.
* testsuite/ld-cris/tls-local-60.d: Update.
* testsuite/ld-cris/tls-local-61.d: Update.
* testsuite/ld-cris/tls-local-63.d: Update.
* testsuite/ld-cris/tls-local-64.d: Update.
* testsuite/ld-cris/tls-ok-30.d: Update.
* testsuite/ld-cris/tls-ok-32.d: Update.
* testsuite/ld-cris/tls-ok-34.d: Update.
* testsuite/ld-mips-elf/tls-multi-got-1.got: Update.
* testsuite/ld-mips-elf/tls-multi-got-1.r: Update.
* testsuite/ld-mips-elf/tlsdyn-pie-o32.d: Update.
* testsuite/ld-mips-elf/tlsdyn-pie-o32.got: Update.
* testsuite/ld-mips-elf/tlslib-o32-hidden.got: Update.
* testsuite/ld-mips-elf/tlslib-o32-ver.got: Update.
* testsuite/ld-mips-elf/tlslib-o32.got: Update.
* testsuite/ld-s390/tlspic.rd: Update.
* testsuite/ld-s390/tlspic_64.rd: Update.
* testsuite/ld-sparc/tlssunnopic32.rd: Update.
* testsuite/ld-sparc/tlssunnopic64.rd: Update.
* testsuite/ld-sparc/tlssunpic32.rd: Update.
* testsuite/ld-sparc/tlssunpic64.rd: Update.
