H.J. Lu [Fri, 18 May 2018 21:23:41 +0000 (14:23 -0700)]
x86: Don't set eh->local_ref to 1 for versioned symbol
bfd_hide_sym_by_version can't be used to check if a versioned symbol is
hidden. It has to be synced with _bfd_elf_link_assign_sym_version to
get the correct answer.
bfd/
PR ld/23194
* elfxx-x86.c (_bfd_x86_elf_link_symbol_references_local): Don't
set eh->local_ref to 1 if a symbol is versioned and there is a
version script.
H.J. Lu [Fri, 18 May 2018 13:43:19 +0000 (06:43 -0700)]
x86: Don't set eh->local_ref to 1 for linker defined symbols
Since symbols created by HIDDEN and PROVIDE_HIDDEN assignments in
linker script may be marked as defined, but not hidden, we can't
set eh->local_ref to 1 in _bfd_x86_elf_link_symbol_references_local.
Also R_386_GOT32X should be handled as just like R_386_GOT32 when
relocating a section. The input R_386_GOT32X relocations, which
can be relaxed, should have been converted to R_386_PC32, R_386_32
or R_386_GOTOFF.
bfd/
PR ld/23189
* elf32-i386.c (elf_i386_relocate_section): Handle R_386_GOT32X
like R_386_GOT32.
* elfxx-x86.c (_bfd_x86_elf_link_symbol_references_local): Don't
set eh->local_ref to 1 for linker defined symbols.
Alan Modra [Fri, 18 May 2018 07:39:45 +0000 (17:09 +0930)]
PR23199, Invalid SHT_GROUP entry leads to group confusion
This patch prevents elf_next_in_group list pointer confusion when
SHT_GROUP sections specify other SHT_GROUP sections in their list of
group sections.
PR 23199
* elf.c (setup_group): Formatting. Check that SHT_GROUP entries
don't point at other SHT_GROUP sections. Set shdr corresponding
to invalid entry, to NULL rather than section 0. Identify
SHT_GROUP section index when reporting an error. Cope with NULL
shdr pointer.
gdb/doc/gdb.texinfo:2498: warning: `.' or `,' must follow @xref, not `@'.
gdb/doc/gdb.texinfo:2517: warning: `.' or `,' must follow @xref, not `@'.
gdb/doc/gdb.texinfo:43443: Node `gdb-add-index man' requires a sectioning command (e.g., @unnumberedsubsec).
gdb/doc/gdb.texinfo:43443: `gdb-add-index man' has no Up field (perhaps incorrect sectioning?).
gdb/doc/gdb.texinfo:43350: Next field of node `gdbinit man' not pointed to (perhaps incorrect sectioning?).
gdb/doc/gdb.texinfo:43443: This node (gdb-add-index man) has the bad Prev.
Alan Modra [Fri, 27 Apr 2018 06:16:40 +0000 (15:46 +0930)]
PR23123, PowerPC32 ifunc regression
Two of the gcc ifunc tests fail for ppc32, due to my pr22374 fix being
a little too enthusiastic in trimming PLT entries. ppc64 doesn't have
the same failures because ppc64_elf_check_relocs happens to set
needs_plt for any ifunc reloc.
PR 23123
PR 22374
* elf32-ppc.c (ppc_elf_adjust_dynamic_symbol): Don't drop plt
relocs for ifuncs.
* elf64-ppc.c (ppc64_elf_adjust_dynamic_symbol): Comment fixes.
H.J. Lu [Thu, 8 Feb 2018 21:52:22 +0000 (13:52 -0800)]
x86: Keep the unused _GLOBAL_OFFSET_TABLE_ for Solaris
Solaris requires to keep _GLOBAL_OFFSET_TABLE_ even if it isn't used.
This patch detects Solaris target and keeps _GLOBAL_OFFSET_TABLE_ for
Solaris.
* elf32-i386.c (elf32_i386_copy_solaris_special_section_fields):
New prototype.
(elf_i386_link_setup_gnu_properties): Set need_global_offset_table
for Solaris.
* elf64-x86-64.c (elf64_x86_64_copy_solaris_special_section_fields):
New prototype.
(elf_x86_64_link_setup_gnu_properties): Set
need_global_offset_table for Solaris.
* elfxx-x86.c (_bfd_x86_elf_size_dynamic_sections): Keep the
unused _GLOBAL_OFFSET_TABLE_ for Solaris.
(_bfd_x86_elf_link_setup_gnu_properties): Copy
need_global_offset_table.
* elfxx-x86.h (elf_x86_link_hash_table): Add
need_global_offset_table.
(elf_x86_init_table): Likewise.
H.J. Lu [Mon, 5 Feb 2018 16:38:16 +0000 (08:38 -0800)]
x86: Remove the unused _GLOBAL_OFFSET_TABLE_
Since _GLOBAL_OFFSET_TABLE_ may be referenced implicitly on x86,
checking ref_regular_nonweak leaves the unused _GLOBAL_OFFSET_TABLE_
in output. This patch checks explicit GOT references instead.
ld-i386/discarded1.s and ld-x86-64/discarded1.s are updated to avoid
linker optimization which removes GOT references.
bfd/
PR ld/22782
* elf32-i386.c (elf_i386_check_relocs): Set got_referenced if
_GLOBAL_OFFSET_TABLE_ is referenced or GOT is needed to resolve
undefined weak symbol to 0.
* elf64-x86-64.c (elf_x86_64_check_relocs): Set got_referenced
if _GLOBAL_OFFSET_TABLE_ is referenced.
* elfxx-x86.c (_bfd_x86_elf_size_dynamic_sections): Check
got_referenced instead of ref_regular_nonweak. Remove the
unused _GLOBAL_OFFSET_TABLE_ from symbol table.
* elfxx-x86.h (elf_x86_link_hash_table): Add got_referenced.
ld/
PR ld/22782
* testsuite/ld-i386/discarded1.s: Replace mov with div.
* testsuite/ld-x86-64/discarded1.s: Likewise.
* testsuite/ld-i386/i386.exp: Run pr22782.
* testsuite/ld-i386/load1-nacl.d: Updated for removing
_GLOBAL_OFFSET_TABLE_ from output.
* testsuite/ld-i386/load1.d: Likewise.
* testsuite/ld-x86-64/load1a-nacl.d: Likewise.
* testsuite/ld-x86-64/load1a.d: Likewise.
* testsuite/ld-x86-64/load1b-nacl.d: Likewise.
* testsuite/ld-x86-64/load1b.d: Likewise.
* testsuite/ld-i386/pr22782.d: New file.
* testsuite/ld-i386/pr22782.s: Likewise.
* testsuite/ld-x86-64/pr22782.s: Likewise.
* testsuite/ld-x86-64/pr22782a.d: Likewise.
* testsuite/ld-x86-64/pr22782b.d: Likewise.
* testsuite/ld-x86-64/x86-64.exp: Run pr22782a and pr22782b.
Nick Clifton [Fri, 6 Apr 2018 14:04:50 +0000 (15:04 +0100)]
Import patch from mainline sources to stop the linker from removing IA64 unwind sections when garbage collecting.
PR 23030
* emulparams/elf64_ia64.sh (OTHER_READONLY_SECTIONS): Make sure
that the .IA_64.unwind_info and .IA_64.unwind sections are not
subject to garbage collection.
H.J. Lu [Wed, 4 Apr 2018 11:36:44 +0000 (04:36 -0700)]
i386: Clear vex instead of vex.evex
"vex" has many fields to control how to decode an instruction. Clear
all fields in "vex" before decoding an instruction to avoid using values
left from the previous instruction.
gas/
PR binutils/23025
* testsuite/gas/i386/prefix.s: Add tests for vcvtpd2dq with
VEX and EVEX prefixes.
* testsuite/gas/i386/prefix.d: Updated.
opcodes/
PR binutils/23025
* i386-dis.c (get_valid_dis386): Don't set vex.prefix nor vex.w
to 0.
(print_insn): Clear vex instead of vex.evex.
Nick Clifton [Thu, 29 Mar 2018 11:31:04 +0000 (12:31 +0100)]
Import patch from the mainline that fixes the ARM assembler's parsing of an illegal ORR instruction.
PR 22773
* config/tc-arm.c (md_apply_fix): Test Rn field of Thumb ORR
instruction before assuming that it is a MOV instruction.
* testsuite/gas/arm/pr22773.s: New test.
* testsuite/gas/arm/pr22773.d: New test driver.
* testsuite/gas/arm/pr22773.l: New expected output.
Cary Coutant [Fri, 23 Mar 2018 16:03:34 +0000 (09:03 -0700)]
Fix case where IR file provides symbol visibility but replacement file does not.
In PR 22868, two IR files provide conflicting visibility for a symbol.
When a def with PROTECTED visibility is seen after a def with DEFAULT
visibility, gold does not override the visibility. Later, if the
replacement object define the symbol with DEFAULT visibility, the symbol
remains DEFAULT. This was caused by a recent change to allow multiply-defined
absolute symbols, combined with the fact that the plugin framework was using
SHN_ABS as the section index for placeholder symbols. The solution is to
use a real (but arbitrary) section index.
gold/
PR gold/22868
* plugin.cc (Sized_pluginobj::do_add_symbols): Use a real section
index instead of SHN_ABS for defined symbols.
* testsuite/Makefile.am (plugin_pr22868): New test case.
* testsuite/Makefile.in: Regenerate
* testsuite/plugin_pr22868.sh: New test script.
* testsuite/plugin_pr22868_a.c: New source file.
* testsuite/plugin_pr22868_b.c: New source file.
Eric Botcazou [Wed, 28 Mar 2018 10:17:15 +0000 (12:17 +0200)]
PR ld/22972 on SPARC.
This is a regression for the corner case of a hidden symbol in a PIC/PIE
binary which is subject to both a new-style GOTDATA relocation and an
old-style GOT relocation. In this case, depending on the link order,
the R_SPARC_RELATIVE dynamic relocation for the GOT slot needed because
of the old-style relocation can be replaced with R_SPARC_NONE coming
from the GOTDATA relocation.
The fix simply records whether an old-style GOT relocation is seen for a
symbol and prevents the R_SPARC_NONE from being generated in this case.
bfd/
* elfxx-sparc.c (struct _bfd_sparc_elf_link_hash_entry): Add new flag
has_old_style_got_reloc.
(_bfd_sparc_elf_check_relocs) <GOT relocations>: Set it for old-style
relocations. Fix a couple of long lines.
(_bfd_sparc_elf_relocate_section) <R_SPARC_GOTDATA_OP>: Do not generate
a R_SPARC_NONE for the GOT slot if the symbol is also subject to
old-style GOT relocations.
ld/
* testsuite/ld-sparc/sparc.exp: Add test for mixed GOTDATA/GOT relocs.
* testsuite/ld-sparc/gotop-hidden.c: New file.
* testsuite/ld-sparc/got-hidden32.s: Likewise.
* testsuite/ld-sparc/got-hidden64.s: Likewise.
* testsuite/ld-sparc/pass.out: Likewise.
H.J. Lu [Mon, 26 Mar 2018 10:57:01 +0000 (03:57 -0700)]
x86-64: Add ENDBR64 to the TLSDESC PLT entry
The TLSDESC entry in a lazy procedure linkage table is called indirectly
with "callq *(%rax)". This patch adds an ENDBR64 to support indirect
branch tracking in Intel CET. The TLSDESC PLT entry now looks like:
to adjust from the original out-of-line save/restore function address
in sfpr to a copy at the end of stub_sec goes wrong when stub_sec is
padded, because the copy is no longer at the end of stub_sec. The
solution is to pad before copying sfpr, so the copy is always at the
end of stub_sec.
* elf64-ppc.c (sfpr_define): Adjust for stub_sec size having
sfpr size added before defining alias symbols.
(ppc64_elf_build_stubs): Add stub section padding before
copying sfpr contents and defining save/restore alias symbols.
Alan Modra [Wed, 7 Mar 2018 23:47:41 +0000 (10:17 +1030)]
Remove unnecessary power9 group terminating nop
Power9 doesn't have a group terminating nop, so we may as well emit a
normal nop for power9. Not that it matters a great deal, I believe
ori 2,2,0 will be treated exactly as ori 0,0,0 by the hardware.
* config/tc-ppc.c (ppc_handle_align): Don't emit a group
terminating nop for power9.
Alan Modra [Tue, 13 Feb 2018 03:39:48 +0000 (14:09 +1030)]
PR22836, "-r -s" doesn't work with -g3 using GCC 7
This fixes the case where all of a group is removed with ld -r, the
situation in the PR, and failures where part of a group is removed
that contain relocs.
bfd/
PR 22836
* elf.c (_bfd_elf_fixup_group_sections): Account for removed
relocation sections. If size reduces to just the flag word,
remove that too and mark with SEC_EXCLUDE.
* elflink.c (bfd_elf_final_link): Strip empty group sections.
binutils/
* testsuite/binutils-all/group-7.s,
* testsuite/binutils-all/group-7a.d,
* testsuite/binutils-all/group-7b.d,
* testsuite/binutils-all/group-7c.d: New tests.
* testsuite/binutils-all/objcopy.exp: Run them.
ld/
* testsuite/ld-elf/pr22836-2.d,
* testsuite/ld-elf/pr22836-2.s: New test.