We don't need a PLT entry when function pointer initialization in a
read/write section is the only reference to a given function symbol.
This patch prevents the unnecessary PLT entry, and ensures no dynamic
relocs are emitted when UNDEFWEAK_NO_DYNAMIC_RELOC says so.
bfd/
PR 22374
* elf32-ppc.c (ppc_elf_adjust_dynamic_symbol): Don't create a plt
entry when just a dynamic reloc can serve. Ensure no dynamic
relocations when UNDEFWEAK_NO_DYNAMIC_RELOC by setting non_got_ref.
Expand and move the non_got_ref comment.
* elf64-ppc.c (ppc64_elf_adjust_dynamic_symbol): Likewise.
ld/
* testsuite/ld-powerpc/ambiguousv2.d: Remove FIXME.
Function pointer references in .data ought to use a dynamic reloc.
There shouldn't be any need for a PLT entry and definitely no copy
reloc.
This test fails on quite a few targets, but isn't something that
anyone should worry about too much. It's really just a missed
optimization.
PR 22374
* testsuite/ld-elf/pr22374a.s,
* testsuite/ld-elf/pr22374b.s,
* testsuite/ld-elf/pr22374-1.r,
* testsuite/ld-elf/pr22374-2.r: New test.
* testsuite/ld-elf/elf.exp: Run it.
* testsuite/ld-elf/elf.exp: Merge some conditionals, a better
name than "shared library for next test", and use xfail arg
of run_ld_link_tests in a few places.
xfail tests for certain targets.
Check shared library support for shared test.
Relax pr21703-r.sd and pr21703-shared.sd with additional "#..." pattern lines.
ld/
* testsuite/ld-elf/elf.exp: xfail pr21703 tests on specific targets.
Only run shared lib test for targets which support it.
* testsuite/ld-elf/pr21703-r.sd: Adjust the expected output.
* testsuite/ld-elf/pr21703-shared.sd: Likewise.
Checked with an arm-none-eabi tester and an arm-none-linux-gnueabi tester with no issues.
* testsuite/ld-arm/cortex-a8-far.d: Update expected disassembly.
* testsuite/ld-arm/farcall-group-size2: Likewise.
* testsuite/ld-arm/farcall-group.d: Likewise.
Some targets prefix global symbols with "_".
bfd/
* archive.c (_bfd_compute_and_write_armap): Match "__gnu_lto_slim"
optionally prefixed with "_".
* linker.c (_bfd_generic_link_add_one_symbol): Likewise.
binutils/
* nm.c (filter_symbols): Match "__gnu_lto_slim" optionally prefixed
with "_".
gold/
* symtab.cc (Symbol_table::add_from_relobj): Match "__gnu_lto_slim"
optionally prefixed with "_".
ld/
* testsuite/ld-plugin/lto-3r.d: Match "__gnu_lto_v" optionally
prefixed with "_".
* testsuite/ld-plugin/lto-5r.d: Likewise.
This matches the ISA specification. This also adds two tests: one to
make sure the assembler rejects invalid 'c.lui's, and one to make sure
we only relax valid 'c.lui's.
bfd/ChangeLog
2017-10-24 Andrew Waterman <andrew@sifive.com>
* elfnn-riscv.c (_bfd_riscv_relax_lui): Don't relax to c.lui
when rd is x0.
include/ChangeLog
2017-10-24 Andrew Waterman <andrew@sifive.com>
* opcode/riscv.h (VALID_RVC_LUI_IMM): c.lui can't load the
immediate 0.
gas/ChangeLog
2017-10-24 Andrew Waterman <andrew@sifive.com>
* testsuite/gas/riscv/c-lui-fail.d: New testcase.
gas/testsuite/gas/riscv/c-lui-fail.l: Likewise.
gas/testsuite/gas/riscv/c-lui-fail.s: Likewise.
gas/testsuite/gas/riscv/riscv.exp: Likewise.
ld/ChangeLog
2017-10-24 Andrew Waterman <andrew@sifive.com>
* ld/testsuite/ld-riscv-elf/c-lui.d: New testcase.
ld/testsuite/ld-riscv-elf/c-lui.s: Likewise.
ld/testsuite/ld-riscv-elf/ld-riscv-elf.exp: New test suite.
The behavior of _bfd_elf_merge_symbol and _bfd_generic_link_add_one_symbol is
inconsistent.
In multiple definition case, _bfd_elf_merge_symbol decided to override the old
symbol definition with the new defintion, (size, type, target data)
In _bfd_generic_link_add_one_symbol, it simply return without doing anything
because of allow-multiple-definition is provided.
This leaves the symbol in a wrong state.
Here, following the documentation, I made this patch to force the old definition
override the new definition if the old symbol is not dynamic or weak.
Because, in those two cases, it's expected to do some merge. I have checked
that, those two cases are properly handled.
bfd/
PR ld/21703
* elflink.c (_bfd_elf_merge_symbol): Handle multiple definition case.
ld/
PR ld/21703
* testsuite/ld-elf/elf.exp: Run new tests.
* testsuite/ld-elf/pr21703-1.s: New.
* testsuite/ld-elf/pr21703-2.s: New.
* testsuite/ld-elf/pr21703-3.s: New.
* testsuite/ld-elf/pr21703-4.s: New.
* testsuite/ld-elf/pr21703-r.sd: New.
* testsuite/ld-elf/pr21703-shared.sd: New.
* testsuite/ld-elf/pr21703.sd: New.
* testsuite/ld-elf/pr21703.ver: New.
For 32-bit BFD, since elf64-x86-64.o isn't compiled in, "-m elf32_x86_64"
never worked. Don't add elf32_x86_64 to supported emulations with 32-bit
BFD.
* configure.tgt (i[3-7]86-*-linux-*): Move elf32_x86_64 from
targ_extra_libpath to targ64_extra_libpath.
PR 22319
bfd * elflink.c (elf_link_output_extsym): Keep global undefined
symbols if they have been marked as needed.
ld * testsuite/ld-elf/pr22310.s: New test source file.
* testsuite/ld-elf/pr22310.d: New test driver.
* testsuite/ld-mmix/undef-3.d: Update expected output from readelf.
* testsuite/ld-elf/shared.exp: Remove kfails.
The test-cases started passing with 5c3261b0e8,
"ELF: Call check_relocs after opening all inputs".
The lists could now be re-concatenated (see other run_ld_link_tests
calls in shared.exp), but are for now left separate to simplify future
kfail/xfailing.
This compile-time test requires a target C compiler to run. It fails
on many targets where ELF backend linkers fail to check undefined weak
symbol in static PIE via UNDEFWEAK_NO_DYNAMIC_RELOC.
PR ld/22269
* testsuite/ld-elf/pr22269-1.rd: New file.
* testsuite/ld-elf/pr22269-1.c: Likewise.
* testsuite/ld-elf/shared.exp: Run pr22269-1.
We currently delete bytes by shifting an entire BFD backwards to
overwrite the bytes we no longer need. The result is that relaxing a
BFD is quadratic time.
This patch adds an additional relocation that specifies a byte range
that will be deleted from the final object file, and adds a relaxation
pass (between the existing passes that delete bytes and the alignment
pass) that actually deletes the bytes. Note that deletion is still
quadratic time, and nothing uses R_RISCV_DELETE yet.
I've been meaning to go convert all the other relaxations to use
R_RISCV_DELETE and then make it faster, but this patch has been sitting
around for months so it looks like that won't happen for a bit. The
PCREL->GPREL relaxation that comes next uses this, and since we've been
using these two patches out of tree since I wrote them months ago I
figure it's better to just get them in now. I (or someone else :)) can
convert all the relocations later...
R_RISCV_DELETE will never be emitted into ELF objects, so therefor isn't
exposed to the rest of binutils. As such, we're not considering this as
part of the ABI.
bfd/ChangeLog
2017-10-19 Palmer Dabbelt <palmer@dabbelt.com>
* elfnn-riscv (R_RISCV_DELETE): New define.
(_bfd_riscv_relax_delete): New function.
(perform_relocation): Handle R_RISCV_DELETE.
(_bfd_riscv_relax_section): Likewise.
ld/ChangeLog
2017-10-19 Palmer Dabbelt <palmer@dabbelt.com>
* emultempl/riscvelf.em (riscv_elf_before_allocation): Add a
third relaxation pass.
-z text applies to all dynamic binaries, not just shared libraries.
A lot of the other options needed attention too.
* ld.texinfo (-z): Combine negative options with corresponding
positive option. Sort the table. Expand and correct "combreloc",
"common", "common-page-size", "interpose", "loadfltr",
"max-page-size", "muldefs", "nodefaultlib", "nodelete", "nodlopen",
"nodump", "noextern-protected-data", "now", "origin", and "text".
Don't generate dynamic relocation against weak undefined symbol if it
is resolved to zero. FIXME: UNDEFWEAK_NO_DYNAMIC_RELOC may need to be
checked in more places.
bfd/
PR ld/22269
* elf32-cris.c (cris_elf_relocate_section): Don't generate
dynamic relocation if UNDEFWEAK_NO_DYNAMIC_RELOC is true.
(cris_elf_check_relocs): Don't allocate dynamic relocation
if UNDEFWEAK_NO_DYNAMIC_RELOC is true.
ld/
PR ld/22269
* testsuite/ld-cris/weakhiddso.d: Update and remove R_CRIS_NONE.
A number of targets need dynamic relocs in PIEs for reasons other than
relocating thread variables. The PR is about text relocations, and,
reading between the lines, unnecessary dynamic tprel relocations.
Change the test to check for those conditions rather than no dynamic
relocations
PR ld/22263
* testsuite/ld-elf/tls.exp: Link with -z text.
* testsuite/ld-elf/pr22263-1.rd: Test for tprel relocs.
This compile-time test requires a target C compiler to run. It fails
on many targets where ELF backend linkers incorrectly check bfd_link_pic
for TLS relocations, which should check bfd_link_executable instead.
PR ld/22263
* testsuite/ld-elf/pr22263-1.rd: New file.
* testsuite/ld-elf/pr22263-1a.c: Likewise.
* testsuite/ld-elf/pr22263-1b.c: Likewise.
* testsuite/ld-elf/tls.exp: Likewise.
* reloc.c (enum bfd_reloc_status): Start values at 2.
* bfd-in2.h: Regenerate.
* elfnn-aarch64.c (aarch64_relocate): Invert sense of function, so
that a TRUE return indicates success. Compare the result of
calling _bfd_aarch64_elf_put_addend against bfd_reloc_ok.
(build_one_stub): Change sense of tests against aarch64_relocate
return value.
(elfNN_aarch64_tls_relax): Return bfd_reloc_notsupported, rather
than FALSE, when an error is detected.
(elfNN_aarch64_final_link_relocate): Likewise.
* testsuite/ld-aarch64/pcrel_pic_defined.d: Expect errors not
warnings. Expect errors about unsupported relocations.
* testsuite/ld-aarch64/pcrel_pic_undefined.d: Likewise.
I believe we should be warning if ld is given both --no-dynamic-linker
and -z dynamic-undefined-weak. The two options are contradictory, the
first says an executable has no dynamic interpreter to resolve dynamic
symbols, while the second is asking for dynamic symbols to be emitted.
(And even if a static PIE's relocation code, which is needed to
process R_*_RELATIVE relocs, could process symbols, there are no
DT_NEEDED dynamic objects to define such symbols.)
I also think that dynamic_undefined_weak is the right flag to control
whether undefined weaks are made dynamic, whether in static PIEs or
anywhere else. So force it to 0 for static PIEs, fixing PR 22269 for
powerpc and any other target where the backend usually defaults to
undefined weaks being made dynamic.
This patch introduces regressions. I'd normally not do that, but
these are all in very recently added test cases, or expose bugs in the
x86 backend. The test cases were added after I'd made it known that
this patch or one like it was imminent.
PR 22269
* emultempl/elf32.em (after_parse): Warn on --no-dynamic-linker
-z dynamic-undefined-weak combination. Set dynamic_undefined_weak
to zero when nointerp.
This function shouldn't be called directly, except from backend code.
bfd/
* elflink.c (_bfd_elf_adjust_dynamic_symbol): Call
elf_backend_hide_symbol, not _bfd_elf_link_hash_hide_symbol.
(bfd_elf_define_start_stop): Likewise.
ld/
* emultempl/elf32.em (before_allocation): Call
elf_backend_hide_symbol, not _bfd_elf_link_hash_hide_symbol.
Formatting.
Since undefined weak symbols in static PIE are always resolved to 0 at
run-time, linker should resolve them as 0 at link-time, regardless of
whether "-z dynamic-undefined-weak" is used. "-z dynamic-undefined-weak"
only makes undefined weak symbols dynamic, but doesn't change undefined
weak symbol resolution in static PIE at link-time. These tests currently
pass on x86, but fails on many other targets.
The framework to resolve weak symbols in static PE at link-time is
posted at
https://sourceware.org/ml/binutils/2017-10/msg00087.html
which requires users/hjl/check_relocs branch to call check_relocs after
opening all inputs. I will submit backend patches for failling targets
after merging users/hjl/check_relocs branch next.
* PR ld/22269
* testsuite/ld-elf/pr22269.s: New file.
* testsuite/ld-elf/pr22269a.d: Likewise.
* testsuite/ld-elf/pr22269b.d: Likewise.
Similar as aarch64 backend, arm backend only overrides the decision on undefined
weak symbols. arm backend part already emits necessary relative relocation for
this case.
bfd/
PR ld/21402
* elf32-arm.c (allocate_dynrelocs_for_symbol): Only make undefined weak
symbols into dynamic.
ld/
PR ld/21402
* testsuite/ld-arm/tls-app.d: Update address.
* testsuite/ld-arm/tls-app.r: Remove relocations.
* testsuite/ld-arm/unresolved-1-dyn.d: Update.
There were two copies of earmelf_fuchsia.c dependencies.
* Makefile.am (earmelfb_fuchsia.c): Rename rule from earmelf_fuchsia.c.
* Makefile.in: Regenerate.
Don't allow '~' as the first character in symbol name in linker script.
PR ld/22267
* ldlex.l (SYMBOLNAMECHAR1) New.
(DEFSYMEXP): Replace FILENAMECHAR1 with SYMBOLNAMECHAR1.
(EXPRESSION): Likewise.
* testsuite/ld-scripts/expr.exp: Run pr22267.
* testsuite/ld-scripts/pr22267.d: New file.
* testsuite/ld-scripts/pr22267.s: Likewise.
* testsuite/ld-scripts/pr22267.t: Likewise.
This fixes a wart I've known about for years, but haven't done
anything about because BFD treats relocation sections as an adjunct to
the section they relocate. SHF_GROUP on the section thus implicitly
applies to its relocation section(s), but it is an error that the
reloc sections aren't part of the group.
Like many patches to gas, this wasn't as straightforward as it could
be due to a number of backends, i386, cr16 and others, removing relocs
in tc_get_reloc rather than marking them as "done" earlier in
md_apply_reloc. So it isn't possible for the group support to
reliably detect the presence of relocs by looking at fixups earlier
than write_relocs. However the group support needs to create
signature symbols, and that must be done before the symbol table is
frozen, before write_relocs. So split off the group sizing from
elf_adjust_symtab and put it in elf_frob_file_after_relocs.
bfd/
PR 21167
* elf.c (_bfd_elf_setup_sections): Don't trim reloc sections from
groups.
(_bfd_elf_init_reloc_shdr): Pass sec_hdr, use it to copy SHF_GROUP
flag from section.
(elf_fake_sections): Adjust calls. Exit immediately on failure.
(bfd_elf_set_group_contents): Add associated reloc section indices
to group contents
gas/
PR 21167
* config/obj-elf.c (struct group_list): Delete elt_count.
(groups): New static.
(build_group_lists): Don't count elements.
(elf_adjust_symtab): Use groups rather than auto list. Set up
pointer from group member to SHT_GROUP section. Don't size
SHT_GROUP section or clean up here..
(elf_frob_file_after_relocs): ..do so here instead.
* testsuite/gas/arc/jli-1.d,
* testsuite/gas/elf/groupautob.d,
* testsuite/gas/mips/compact-eh-eb-2.d,
* testsuite/gas/mips/compact-eh-eb-5.d,
* testsuite/gas/mips/compact-eh-el-2.d,
* testsuite/gas/mips/compact-eh-el-5.d: Adjust.
ld/
PR 21167
* testsuite/ld-elf/group9b.d: Adjust for relocs included in group.
* ld.texinfo (-z relro): Expand description.
(DATA_SEGMENT_ALIGN): Note that -z relro is not effective when
running with system page size larger than commonpagesize.
(DATA_SEGMENT_RELRO_END): Be explicit about the alignment.
I am applying a patch that has been lying around in the Fedora
binutils sources for a while. It skips the PR14918 linker test for
ARM based targets. This test checks that libgcc is not included in a
link of an empty executable. This works for most targets, but on the
ARM the crt1.o startup code calls __libc_csu_init which is in
/usr/lib/libc_nonshared.a(elf-init.oS). This in turn needs
__aeabi_unwind_cpp_pr0@@GCC_3.5 which is provided by libgcc_s.so.1,
and so the test fails.
non_ir_ref_dynamic wasn't being set in the case where we have a
versioned dynamic symbol definition with a non-versioned matching IR
symbol.
bfd/
PR 22220
* elflink.c (_bfd_elf_merge_symbol): Set non_ir_ref_dynamic in
a case where plugin_notice isn't called.
ld/
* testsuite/ld-plugin/pr22220.h,
* testsuite/ld-plugin/pr22220lib.cc,
* testsuite/ld-plugin/pr22220lib.ver,
* testsuite/ld-plugin/pr22220main.cc: New test.
* testsuite/ld-plugin/lto.exp: Run it.
Fix commit 647d4de92e ("Test undefined symbols in shared libraries")
and remove "undefined symbols in shared lib (dyn reloc)" test failures
for MIPS and S+core targets.
This test assumes 3 dynamic relocation entries, however the respective
psABIs define an extra single dummy *_NONE relocation at the beginning
of the dynamic relocation table, which results in test output like the
following for MIPS:
Relocation section '.rel.dyn' at offset 0x33c contains 4 entries:
Offset Info Type Sym.Value Sym. Name
00000000 00000000 R_MIPS_NONE
00010364 00000903 R_MIPS_REL32 00000000 undef_pfun
00010360 00000a03 R_MIPS_REL32 00000000 undef_data
00010368 00000b03 R_MIPS_REL32 00000000 undef_notype
and S+core:
Relocation section '.rel.dyn' at offset 0x274 contains 4 entries:
Offset Info Type Sym.Value Sym. Name
00000000 00000000 R_SCORE_NONE
5ffe0464 00000e12 R_SCORE_REL32 00000000 undef_notype
5ffe045c 00000f12 R_SCORE_REL32 00000000 undef_data
5ffe0460 00001012 R_SCORE_REL32 00000000 undef_pfun
targets respectively, and also n64 MIPS targets pad relocation triplets
they use with extra *_NONE relocations, resulting in output like this:
Relocation section '.rel.dyn' at offset 0x4c0 contains 4 entries:
Offset Info Type Sym. Value Sym. Name
000000000000 000000000000 R_MIPS_NONE
Type2: R_MIPS_NONE
Type3: R_MIPS_NONE
000000010508 000900001203 R_MIPS_REL32 0000000000000000 undef_pfun
Type2: R_MIPS_64
Type3: R_MIPS_NONE
000000010500 000a00001203 R_MIPS_REL32 0000000000000000 undef_data
Type2: R_MIPS_64
Type3: R_MIPS_NONE
000000010510 000b00001203 R_MIPS_REL32 0000000000000000 undef_notype
Type2: R_MIPS_64
Type3: R_MIPS_NONE
Expect the right number of *_NONE relocations then, and adjust the total
number of relocations expected accordingly.
ld/
* testsuite/ld-undefined/undefined.exp: Correct the dyn reloc
test for MIPS and S+core targets.
PowerPC64 .cfi directives use DW_EH_PE_sdata4 encoding for .eh_frame,
so there is no real reason why .eh_frame should be 8 byte aligned.
gas/
* config/tc-ppc.h (EH_FRAME_ALIGNMENT): Define.
ld/
* testsuite/ld-powerpc/tlsopt5.wf: Update for reduced alignment.
Complement commit 26eebcf553 ("Update OpenBSD/mips64 support"),
<https://sourceware.org/ml/binutils/2005-04/msg00382.html>, which added
OpenBSD/mips64 support to GAS, and also add it to LD, avoiding a build
failure at the configuration stage, like:
*** ld does not support target mips64-unknown-openbsd
*** see ld/configure.tgt for supported targets
make[1]: *** [configure-ld] Error 1
As per OS support only include n64 MIPS emulations, and use the
traditional ones, matching the choice already made with the addition of
GAS support.
ld/
* configure.tgt <mips64el-*-openbsd*, mips64-*-openbsd*>: New
targets.
With -PIE on x86-64, we get
0x000000006ffffffb (FLAGS_1) Flags: GLOBAUDIT PIE
We should allow additional bits in DT_FLAGS_1.
PR ld/22139
* testsuite/ld-elf/globalaudit.rd: Allow additional bits in
DT_FLAGS_1.
bfd/
PR ld/22135
* elf32-i386.c (elf_i386_convert_load_reloc): Add an argument
to indicate if conversion is performed.
(elf_i386_check_relocs): Cache section contents and relocations
if conversion is performed.
* elf64-x86-64.c (elf_x86_64_check_relocs): Cache section
contents and relocations if conversion is performed.
ld/
PR ld/22135
* testsuite/ld-i386/i386.exp: Run pr22135.
* testsuite/ld-x86-64/x86-64.exp: Likewise.
* testsuite/ld-i386/pr22135.d: New file.
* testsuite/ld-i386/pr22135.s: Likewise.
* testsuite/ld-x86-64/pr22135.d: Likewise.
* testsuite/ld-x86-64/pr22135.s: Likewise.
