This patch adds the following relocation support into binutils bfd linker.
BFD_RELOC_AARCH64_TLSLE_LDST16_TPREL_LO12,
BFD_RELOC_AARCH64_TLSLE_LDST16_TPREL_LO12_NC,
BFD_RELOC_AARCH64_TLSLE_LDST32_TPREL_LO12,
BFD_RELOC_AARCH64_TLSLE_LDST32_TPREL_LO12_NC,
BFD_RELOC_AARCH64_TLSLE_LDST64_TPREL_LO12,
BFD_RELOC_AARCH64_TLSLE_LDST64_TPREL_LO12_NC,
BFD_RELOC_AARCH64_TLSLE_LDST8_TPREL_LO12,
BFD_RELOC_AARCH64_TLSLE_LDST8_TPREL_LO12_NC.
Those relocations includes both ip64 and ilp32 variant.
For absolute symbols which are forced local or not dynamic, the ABS relocation
should be resolved at static linking time.
Originally, an RELATIVE/ABS relocation will be generated even for absolution
symbols for the dynamic linker to resolve.
bfd/
2018-03-07 Renlin Li <renlin.li@arm.com>
PR ld/20402
* elfnn-aarch64.c (elfNN_aarch64_final_link_relocate): Check absolute symbol,
and don't emit relocation in specific case.
ld/
2018-03-07 Renlin Li <renlin.li@arm.com>
PR ld/20402
* testsuite/ld-aarch64/aarch64-elf.exp: Run new test.
* testsuite/ld-aarch64/pr20402.s: New.
* testsuite/ld-aarch64/pr20402.d: New.
The assumption that R_AARCH64_ABS16 and R_AARCH64_ABS32 relocation in LP64 abi
will be used to generate an address does not hold for absolute symbol.
In this case, it is a value fixed at static linking time.
The condition to check the relocations is relax to allow absolute symbol and
undefined symbol case.
bfd/
2018-02-05 Renlin Li <renlin.li@arm.com>
PR ld/22764
* elfnn-aarch64.c (elfNN_aarch64_check_relocs): Relax the
R_AARCH64_ABS32 and R_AARCH64_ABS16 for absolute symbol. Apply the
check for writeable section as well.
ld/
2018-02-05 Renlin Li <renlin.li@arm.com>
PR ld/22764
* testsuite/ld-aarch64/emit-relocs-258.s: Define symbol as an address.
* testsuite/ld-aarch64/emit-relocs-259.s: Likewise.
* testsuite/ld-aarch64/aarch64-elf.exp: Run new test.
* testsuite/ld-aarch64/pr22764.s: New.
* testsuite/ld-aarch64/pr22764.d: New.
R_AARCH64_ABS64, R_AARCH64_ABS32 and R_AARCH64_ABS16 are data relocations
supported in AArch64 elf ABI.
R_AARCH64_ABS64 under LP64 is allowed in shared object and a dynamic relocation entry
will be generated. This allows the dynamic linker to do further symbol resolution.
R_AARCH64_ABS32 likewise is allowed in shared object, however under ILP32 abi.
The original behavior for R_AARCH64_ABS32 under LP64 is that, it's allowed
in shared object and silently resolved at static linking time.
No dynamic relocation entry is generate for it.
R_AARCH64_ABS16 is allowed in shared object under both L64 and ILP32.
It's resolved at static linking time as well.
Under LP64, the address should be 64-bit. R_AARCH64_ABS32 relocation indicates
an address that is only sized 32 bits which is meaningless in LP64 shared object.
It's useful to error out.
I have checked glibc dynamic linker code, R_AARCH64_ABS16 is not supported at all. So
R_AARCH64_ABS16 should be reject in shared object completely.
In this patch, R_AARCH64_ABS32 is rejected under LP64 in constant section of shared object.
R_AARCH64_ABS16 is rejected in constant section of shared object in both ABI.
This will sometimes provide useful information for buggy code.
For some pc-relative relocations we want to allow them under PIC mode while
a normal global symbol defined in the same dynamic object can still bind
externally through copy relocation. So, we should not allow pc-relative
relocation against such symbol.
SYMBOL_REFERENCES_LOCAL should be used and is more accurate than the original
individual checks.
bfd/
* elfnn-aarch64.c (elfNN_aarch64_final_link_relocate): Use
SYMBOL_REFERENCES_LOCAL.
ld/
* testsuite/ld-aarch64/aarch64-elf.exp: Update test name
* testsuite/ld-aarch64/pcrel.s: Add new testcases.
* testsuite/ld-aarch64/pcrel_pic_undefined.d: Update the expected
warnings.
* testsuite/ld-aarch64/pcrel_pic_defined_local.d: Rename ...
* testsuite/ld-aarch64/pcrel_pic_defined.d: ... to this.
Update expected warnings.
As discussed at the PR, this patch tries to avoid COPY relocation generation
and propagate the original relocation into runtime if it was relocating on
writable section. The ELIMINATE_COPY_RELOCS has been set to true and it's
underlying infrastructure has been improved so that the COPY reloc elimination
at least working on absoluate relocations (ABS64) on AArch64.
BFD linker copy relocation elimination framwork requires the backend to always
allocate dynrelocs for all those relocation types that are possible to introduce
copy relocations. This is for adjust_dynamic_symbol hook to be able to get all
symbol reference information. Should one symbol is referenced by more than one
relocations, if there is any of them needs copy relocation then linker should
generate it.
bfd/
PR ld/21532
* elfnn-aarch64.c (ELIMINATE_COPY_RELOCS): Set to 1.
(elfNN_aarch64_final_link_relocate): Also propagate relocations to
runtime for if there needs copy relocation elimination.
(need_copy_relocation_p): New function. Return true for symbol with
pc-relative references and if it's against read-only sections.
(elfNN_aarch64_adjust_dynamic_symbol): Use need_copy_relocation_p.
(elfNN_aarch64_check_relocs): Allocate dynrelocs for relocation types
that are related with accessing external objects.
(elfNN_aarch64_gc_sweep_hook): Sync the relocation types with the change
in elfNN_aarch64_check_relocs.
ld/
* testsuite/ld-aarch64/copy-reloc-exe-2.s: New test source file.
* testsuite/ld-aarch64/copy-reloc-2.d: New test.
* testsuite/ld-aarch64/copy-reloc-exe-eliminate.s: New test source file.
* testsuite/ld-aarch64/copy-reloc-eliminate.d: New test.
* testsuite/ld-aarch64/copy-reloc-so.s: Define new global objects.
* testsuite/ld-aarch64/aarch64-elf.exp: Run new tests.
This reverts commit bc327528fd.
This patch can only be committed after PC-relative relocation types
support on copy relocation elimination is also completed.
As discussed at the PR, this patch tries to avoid COPY relocation generation
and propagate the original relocation into runtime if it was relocating on
writable section. The ELIMINATE_COPY_RELOCS has been set to true and it's
underlying infrastructure has been improved so that the COPY reloc elimination
at least working on absoluate relocations (ABS64) after this patch.
bfd/
PR ld/21532
* elfnn-aarch64.c (ELIMINATE_COPY_RELOCS): Set to 1.
(elfNN_aarch64_final_link_relocate): Also propagate relocations to
runtime for copy relocation elimination cases.
(alias_readonly_dynrelocs): New function.
(elfNN_aarch64_adjust_dynamic_symbol): Keep the dynamic relocs instead
of generating copy relocation if it is not against read-only sections.
(elfNN_aarch64_check_relocs): Likewise.
ld/
* testsuite/ld-aarch64/copy-reloc-eliminate.d: New test.
* testsuite/ld-aarch64/copy-reloc-exe-eliminate.s: New test source file.
* testsuite/ld-aarch64/aarch64-elf.exp: Run new testcase.
bfd * elfnn-aarch64.c: Fix relaxations for ILP32 mode.
ld * testsuite/ld-aarch64/aarch64-elf.exp: Run new tests.
* testsuite/ld-aarch64/tls-desc-ie-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-all-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-gd-le-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-gdesc-le-2-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-gdesc-le-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-ie-le-2-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-ie-le-3-ilp32.d: New test.
* testsuite/ld-aarch64/tls-relax-ie-le-ilp32.d: New test.
* testsuite/ld-aarch64/tls-tiny-desc-ie-ilp32.d: New test.
* testsuite/ld-aarch64/tls-tiny-desc-le-ilp32.d: New test.
* testsuite/ld-aarch64/tls-tiny-gd-ie-ilp32.d: New test.
* testsuite/ld-aarch64/tls-tiny-gd-le-ilp32.d: New test.
bfd/
* elfnn-aarch64.c (elf_aarch64_hash_symbol): New function.
(elf_backend_hash_symbol): Define.
ld/
* testsuite/ld-aarch64/aarch64-elf.exp (aarch64elflinktests): New tests.
* testsuite/ld-aarch64/func-in-so.s: New test source file.
* testsuite/ld-aarch64/func-sym-hash-opt.s: Likewise.
* testsuite/ld-aarch64/func-sym-hash-opt.d: New expected test result.
TCB_SIZE is 2*sizeof(void *), which is 0x10 for lp64, and 0x8 for
ilp32. During relaxation, ld goes to do a replace:
bl __tls_get_addr => add R0, R0, TCB_SIZE
But actual implementation is:
bfd_putl32 (0x91004000, contents + rel->r_offset + 4);
Which is equivalent of add x0, x0, 0x10. This is wrong for ilp32.
The possible fix for it is:
bfd_putl32 (0x91000000 | (TCB_SIZE<<10), contents + rel->r_offset + 4);
But ilp32 also needs w-registers, so it's simpler to put proper
instruction in #if/#else condition.
There are 2 such relaxations in elfNN_aarch64_tls_relax(), and so 2 new
tests added for ilp32 mode to test it.
Yury
* bfd/elfnn-aarch64.c: fix TLS relaxations for ilp32 where
TCB_SIZE is used.
* ld/testsuite/ld-aarch64/aarch64-elf.exp: Add tests for the case.
* ld/testsuite/ld-aarch64/tls-relax-ld-le-small-ilp32.d: New file.
* ld/testsuite/ld-aarch64/tls-relax-ld-le-tiny-ilp32.d: New file.
Signed-off-by: Yury Norov <ynorov@caviumnetworks.com>
TCB_SIZE is 2*sizeof(void *), which is 0x10 for lp64, and 0x8 for
ilp32. During relaxation, ld goes to do a replace:
bl __tls_get_addr => add R0, R0, TCB_SIZE
But actual implementation is:
bfd_putl32 (0x91004000, contents + rel->r_offset + 4);
Which is equivalent of add x0, x0, 0x10. This is wrong for ilp32.
The possible fix for it is:
bfd_putl32 (0x91000000 | (TCB_SIZE<<10), contents + rel->r_offset + 4);
But ilp32 also needs w-registers, so it's simpler to put proper
instruction in #if/#else condition.
THere are 2 such relaxations in elfNN_aarch64_tls_relax(), and so 2 new
tests added for ilp32 mode to test it.
Yury
PR ld/20868
bfd * elfnn-aarch64.c (elfNN_aarch64_tls_relax): Use 32-bit accesses
to the GOT when operating in 32-bit mode.
ld * testsuite/ld-aarch64/tls-relax-gd-ie-ilp32.d: New test.
* testsuite/ld-aarch64/relocs-ilp32.ld: Linker script for the new
test.
* testsuite/ld-aarch64/aarch64-elf.exp: Run the new test.
bfd/
PR target/20737
* elfnn-aarch64.c (elfNN_aarch64_final_link_relocate): Bind defined
symbol locally in PIE.
ld/
* testsuite/ld-aarch64/pie-bind-locally-a.s: New test source.
* testsuite/ld-aarch64/pie-bind-locally-b.s: Likewise.
* testsuite/ld-aarch64/pie-bind-locally.d: New testcase.
* testsuite/ld-aarch64/aarch64-elf.exp: Run new testcase.
When handling absolute relocations for global symbols bind within the
shared object, AArch64 will generate one dynamic RELATIVE relocation,
but won't apply the value for this absolution relocations at static
linking stage. This is different from AArch64 gold linker and x86-64.
This is not a bug as AArch64 is RELA, there is only guarantee that
relocation addend is placed in the relocation entry. But some
system softwares originally writen for x86-64 might assume AArch64
bfd linker gets the same behavior as x86-64, then they could take
advantage of this buy skipping those RELATIVE dynamic relocations
if the load address is the same as the static linking address.
This patch makes AArch64 BFD linker applies absolution relocations at
static linking stage for scenario described above. Meanwhile old AArch64
android loader has a bug (PR19163) which relies on current linker behavior
as a workaround, so the same option --no-apply-dynamic-relocs added.
2016-03-07 Jiong Wang <jiong.wang@arm.com>
bfd/
* elfnn-aarch64.c (elfNN_aarch64_check_relocs): Always create .got section
if the symbol "_GLOBAL_OFFSET_TABLE_" referenced.
ld/
* testsuite/ld-aarch64/implicit_got_section_1.s: New test source file.
* testsuite/ld-aarch64/implicit_got_section_1.d: New test expected result.
* testsuite/ld-aarch64/aarch64-elf.exp: Run new test.
For these three relocations, 17 bit signed value should be used, instead of
16 bit. The bitsize field is changed from 16 to 17, this field in aarch64
backend is used for overflow check only.
bfd/
2016-02-26 Renlin Li <renlin.li@arm.com>
* elfnn-aarch64.c (elfNN_aarch64_howto_table): Fix signed overflow
check for MOVW_SABS_G0, MOVW_SABS_G1, MOVW_SABS_G2.
ld/
2016-02-26 Renlin Li <renlin.li@arm.com>
* testsuite/ld-aarch64/aarch64-elf.exp: Run new testcases.
* testsuite/ld-aarch64/emit-relocs-270.d: Update to use new boundary.
* testsuite/ld-aarch64/emit-relocs-271.d: Likewise.
* testsuite/ld-aarch64/emit-relocs-272.d: Likewise.
* testsuite/ld-aarch64/emit-relocs-270-overflow.d: New.
* testsuite/ld-aarch64/emit-relocs-270-overflow.s: New.
* testsuite/ld-aarch64/emit-relocs-271-overflow.d: New.
* testsuite/ld-aarch64/emit-relocs-271-overflow.s: New.
* testsuite/ld-aarch64/emit-relocs-272-overflow.d: New.
* testsuite/ld-aarch64/emit-relocs-272-overflow.s: New.
bfd * elfnn-aarch64.c (elfNN_aarch64_relocate_section): Add a more
helpful warning message to explain why certain AArch64 relocs
might overflow.
ld * testsuite/ld-aarch64/reloc-overflow-bad.d: New test.
* testsuite/ld-aarch64/reloc-overflow-1.s: New source file.
* testsuite/ld-aarch64/reloc-overflow-2.s: New source file.
* testsuite/ld-aarch64/aarch64-elf.exp: Run the new test.
bfd * elfnn-aarch64.c (elfNN_aarch64_relocate_section): Add a more
helpful warning message to explain why certain AArch64 relocs
might overflow.
ld * testsuite/ld-aarch64/reloc-overflow-bad.d: New test.
* testsuite/ld-aarch64/reloc-overflow-1.s: New source file.
* testsuite/ld-aarch64/reloc-overflow-2.s: New source file.
* testsuite/ld-aarch64/aarch64-elf.exp: Run the new test.
As defined at AArch64 ELF Specification (4.6.7 Call and Jump
relocations), symbol with type of non STT_FUNC but in different input
section with relocation place should insert long branch veneer also.
Meanwhile the current long branch veneer infrastructure havn't considered
the situation where the branch destination is "sym_value + rela->addend".
This was OK because we only insert veneer for long call destination is
STT_FUNC symbol for which the addend is always zero. But as we relax the
support to other situations by this patch, we need to handle addend be
non-zero value. For example, for static function, relocation against
"local symbol" are turned into relocation against "section symbol + offset"
where there is a valid addend.
bfd/
* elfnn-aarch64.c (aarch64_type_of_stub): Allow insert long branch
veneer for sym_sec != input_sec.
(elfNN_aarch64_size_stub): Support STT_SECTION symbol.
(elfNN_aarch64_final_link_relocate): Take rela addend into account when
calculation destination.
ld/
* testsuite/ld-aarch64/farcall-section.d: Delete.
* testsuite/ld-aarch64/farcall-section.s: Delete.
* testsuite/ld-aarch64/farcall-b-section.d: New expectation file.
* testsuite/ld-aarch64/farcall-bl-section.d: Likewise.
* testsuite/ld-aarch64/farcall-b-section.s: New testcase.
* testsuite/ld-aarch64/farcall-bl-section.s: Likewise.
* testsuite/ld-aarch64/aarch64-elf.exp: Likewise.
bfd/
2015-10-02 Renlin Li <renlin.li@arm.com>
* elfnn-aarch64.c (IS_AARCH64_TLS_RELAX_RELOC): Add
TLSIE_MOVW_GOTTPREL_G1.
(aarch64_tls_transition_without_check): Add
TLSIE_MOVW_GOTTPREL_G1 to TLSLE_MOVW_TPREL_G2
transition for local symbol.
(elfNN_aarch64_tls_relax): Add a argument to pass tp offset.
Add TLSIE_MOVW_GOTTPREL_G1 relaxation.
(elfNN_aarch64_relocate_section): Call elfNN_aarch64_tls_relax
with new argument.
ld/testsuite/
2015-10-02 Renlin Li <renlin.li@arm.com>
* ld-aarch64/aarch64-elf.exp (tls-relax-large-le-ie): Run new test.
* ld-aarch64/tls-relax-large-ie-le.d: New.
* ld-aarch64/tls-relax-large-ie-le.s: New.
bfd/
2015-10-02 Renlin Li <renlin.li@arm.com>
* elfnn-aarch64.c(IS_AARCH64_TLS_RELAX_RELOC):
Add relaxation support for TLSGD_MOVW_G0_NC and TLSGD_MOVW_G1.
(aarch64_tls_transition_without_check): Likewise
(elfNN_aarch64_tls_relax): Likwise.
ld/testsuite/
2015-10-02 Renlin Li <renlin.li@arm.com>
* ld-aarch64/aarch64-elf.exp: run new test
* ld-aarch64/tls-relax-large-gd-ie.d: New.
* ld-aarch64/tls-relax-large-gd-ie.s: New.
* ld-aarch64/tls-relax-large-gd-le.d: New.
* ld-aarch64/tls-relax-large-gd-le.s: New.
This patch fixed those failures on elf configuration by:
* Improve the ILP32 target selector "aarch64_choose_ilp32_emul",
makes it more robust. Target triples copied from configure.tgt
* Updated emit-relocs-86/-overflow.d to use aarch64_choose_ilp32_emul
which is following what have done with emit-relocs-28.
* Those instruction encoding mismatch is because those encoding
contains pc-relative address. As for elf, we may have different
start address. relaxed encodind check, especially for
aarch64-farcall-b/bl-plt, as the main purpose of those check are
ELF text/data layout, we just want to make sure veneer to plt stub
is generated.
2015-08-12 Jiong Wang <jiong.wang@arm.com>
ld/testsuite/
* ld-aarch64/aarch64-elf.exp (aarch64_choose_ilp32_emul): Support all
four triple shapes: aarch64-*-linux*, aarch64-*-elf,
aarch64_be-*-linux*, aarch64_be-*-elf.
* ld-aarch64/emit-relocs-86.d: Use aarch64_choose_ilp32_emul.
* ld-aarch64/emit-relocs-86-overflow.d: Likewise.
* ld-aarch64/ld-aarch64/farcall-b-plt.d: Relax instrucion encoding
check when they reflect address.
* ld-aarch64/ld-aarch64/farcall-bl-plt.d: Likewise.
2015-08-11 Jiong Wang <jiong.wang@arm.com>
bfd/
* bfd/elfnn-aarch64.c (aarch64_type_of_stub): New parameter "sym_sec".
Loose the check for symbol from ABS section.
(elfNN_aarch64_size_stubs): Pass sym_sec.
ld/testsuite/
* ld-aarch64/farcall-b-defsym.s: New test.
* ld-aarch64/farcall-bl-defsym.s: Likewise.
* ld-aarch64/farcall-b-defsym.d: New expectation.
* ld-aarch64/farcall-bl-defsym.d: Likewise.
