GAS always places section groups (SHT_GROUP) before the rest of the
sections in the output file. However, other assemblers may place
section groups after the group members.
This patch fixes handlign such situations, and removes some duplicated
logic.
bfd/ChangeLog:
2017-06-06 Jose E. Marchesi <jose.marchesi@oracle.com>
* elf.c (setup_group): Make sure BFD sections are created for all
group sections in the input file when processing SHF_GROUP
sections.
(bfd_section_from_shdr): Avoid duplicating logic already
implemented in `setup_group'.
This commit adds a new linker feature: the ability to resolve section
groups as part of a relocatable link.
Currently section groups are automatically resolved when performing a
final link, and are carried through when performing a relocatable link.
By carried through this means that one copy of each section group (from
all the copies that might be found in all the input files) is placed
into the output file. Sections that are part of a section group will
not match input section specifiers within a linker script and are
forcibly kept as separate sections.
There is a slight resemblance between section groups and common
section. Like section groups, common sections are carried through when
performing a relocatable link, and resolved (allocated actual space)
only at final link time.
However, with common sections there is an ability to force the linker to
allocate space for the common sections when performing a relocatable
link, there's currently no such ability for section groups.
This commit adds such a mechanism. This new facility can be accessed in
two ways, first there's a command line switch --force-group-allocation,
second, there's a new linker script command FORCE_GROUP_ALLOCATION. If
one of these is used when performing a relocatable link then the linker
will resolve the section groups as though it were performing a final
link, the section group will be deleted, and the members of the group
will be placed like normal input sections. If there are multiple copies
of the group (from multiple input files) then only one copy of the group
members will be placed, the duplicate copies will be discarded.
Unlike common sections that have the --no-define-common command line
flag, and INHIBIT_COMMON_ALLOCATION linker script command there is no
way to prevent group resolution during a final link, this is because the
ELF gABI specifically prohibits the presence of SHT_GROUP sections in a
fully linked executable. However, the code as written should make
adding such a feature trivial, setting the new resolve_section_groups
flag to false during a final link should work as you'd expect.
bfd/ChangeLog:
* elf.c (_bfd_elf_make_section_from_shdr): Don't initially mark
SEC_GROUP sections as SEC_EXCLUDE.
(bfd_elf_set_group_contents): Replace use of abort with an assert.
(assign_section_numbers): Use resolve_section_groups flag instead
of relocatable link type.
(_bfd_elf_init_private_section_data): Use resolve_section_groups
flag instead of checking the final_link flag for part of the
checks in here. Fix white space as a result.
* elflink.c (elf_link_input_bfd): Use resolve_section_groups flag
instead of relocatable link type.
(bfd_elf_final_link): Likewise.
include/ChangeLog:
* bfdlink.h (struct bfd_link_info): Add new resolve_section_groups
flag.
ld/ChangeLog:
* ld.h (struct args_type): Add force_group_allocation field.
* ldgram.y: Add support for FORCE_GROUP_ALLOCATION.
* ldlex.h: Likewise.
* ldlex.l: Likewise.
* lexsup.c: Likewise.
* ldlang.c (unique_section_p): Check resolve_section_groups flag
not the relaxable link flag.
(lang_add_section): Discard section groups when we're resolving
groups. Clear the SEC_LINK_ONCE flag if we're resolving section
groups.
* ldmain.c (main): Initialise resolve_section_groups flag in
link_info based on command line flags.
* testsuite/ld-elf/group11.d: New file.
* testsuite/ld-elf/group12.d: New file.
* testsuite/ld-elf/group12.ld: New file.
* NEWS: Mention new features.
* ld.texinfo (Options): Document --force-group-allocation.
(Miscellaneous Commands): Document FORCE_GROUP_ALLOCATION.
Correct a commit e5713223cb ("MIPS/BFD: For n64 hold the number of
internal relocs in `->reloc_count'") regression and change internal
relocation handling in the generic ELF BFD linker code such that, except
in the presence of R_SPARC_OLO10 relocations, a section's `reloc_count'
holds the number of internal rather than external relocations, making
the handling more consistent between GAS, which sets `->reloc_count'
with a call to `bfd_set_reloc', and LD, which sets `->reloc_count' as it
reads input sections.
The handling of dynamic relocations remains unchanged and they continue
holding the number of external relocations in `->reloc_count'; they are
also not converted to the internal form except in `elf_link_sort_relocs'
(which does not handle the general, i.e. non-n64-MIPS case of composed
relocations correctly as per the ELF gABI, though it does not seem to
matter for the targets we currently support).
The n64 MIPS backend is the only one with `int_rels_per_ext_rel' set to
non-one, and consequently the change is trivial for all the remaining
backends and targets.
bfd/
* elf-bfd.h (RELOC_AGAINST_DISCARDED_SECTION): Subtract `count'
from `reloc_count' rather than decrementing it.
* elf.c (bfd_section_from_shdr): Multiply the adjustment to
`reloc_count' by `int_rels_per_ext_rel'.
* elf32-score.c (score_elf_final_link_relocate): Do not multiply
`reloc_count' by `int_rels_per_ext_rel' for last relocation
entry determination.
(s3_bfd_score_elf_check_relocs): Likewise.
* elf32-score7.c (score_elf_final_link_relocate): Likewise.
(s7_bfd_score_elf_relocate_section): Likewise.
(s7_bfd_score_elf_check_relocs): Likewise.
* elf64-mips.c (mips_elf64_get_reloc_upper_bound): Remove
prototype and function.
(mips_elf64_slurp_one_reloc_table): Do not update `reloc_count'.
(mips_elf64_slurp_reloc_table): Assert that `reloc_count' is
triple rather than once the sum of REL and RELA relocation entry
counts.
(bfd_elf64_get_reloc_upper_bound): Remove macro.
* elflink.c (_bfd_elf_link_read_relocs): Do not multiply
`reloc_count' by `int_rels_per_ext_rel' for internal relocation
storage allocation size determination.
(elf_link_input_bfd): Multiply `.ctors' and `.dtors' section's
size by `int_rels_per_ext_rel'. Do not multiply `reloc_count'
by `int_rels_per_ext_rel' for last relocation entry
determination.
(bfd_elf_final_link): Do not multiply `reloc_count' by
`int_rels_per_ext_rel' for internal relocation storage
allocation size determination.
(init_reloc_cookie_rels): Do not multiply `reloc_count' by
`int_rels_per_ext_rel' for last relocation entry determination.
(elf_gc_smash_unused_vtentry_relocs): Likewise.
* elfxx-mips.c (_bfd_mips_elf_check_relocs): Likewise.
(_bfd_mips_elf_relocate_section): Likewise.
FreeBSD ELF cores contain data structures with that have two different
layouts: one for ILP32 platforms and a second for LP64 platforms.
Previously, the code used 'bits_per_word' from 'arch_info', but this
field is not a reliable indicator of the format for FreeBSD MIPS cores
in particular.
I had originally posted this patch back in November because process
cores for FreeBSD MIPS contained an e_flags value of 0 in the header
which resulted in a bfd_arch which always had 'bits_per_word' set to
32. This permitted reading o32 cores, but not n64 cores. The feedback
I received then was to try to change n64 cores to use a different
default bfd_arch that had a 64-bit 'bits_per_word' when e_flags was zero.
I submitted a patch to that effect but it was never approved. Instead,
I changed FreeBSD's kernel and gcore commands to preserve the e_flags
field from an executable when generating process cores. With a proper
e_flags field in process cores, n64 cores now use a 64-bit bfd_arch and
now work fine. However, the change to include e_flags in the process
cores had the unintended side effect of breaking handling of o32
process cores. Specifically, FreeBSD MIPS builds o32 with a default
MIPS architecture of 'mips3', thus FreeBSD process cores with a non-zero
e_flags match the 'mips3' bfd_arch which has 64 'bits_per_word'.
From this, it seems that 'bits_per_word' for FreeBSD MIPS is not likely
to ever be completely correct. However, FreeBSD core dumps do
reliably set the ELF class to ELFCLASS32 for cores using ILP32 and
ELFCLASS64 for cores using LP64. As such, I think my original patch of
using the ELF class instead of 'bits_per_word' is probably the simplest
and most reliable approach for detecting the note structure layout.
bfd/ChangeLog:
* elf.c (elfcore_grok_freebsd_psinfo): Use ELF header class to
determine structure sizes.
(elfcore_grok_freebsd_prstatus): Likewise.
ELFv2 functions with localentry:0 are those with a single entry point,
ie. global entry == local entry, and that have no requirement on r2 or
r12, and guarantee r2 is unchanged on return. Such an external
function can be called via the PLT without saving r2 or restoring it
on return, avoiding a common load-hit-store for small functions. The
optimization is attractive. The TOC pointer load-hit-store is a major
reason why calls to small functions that need no register saves, or
with shrink-wrap, no register saves on a fast path, are slow on
powerpc64le.
To be safe, this optimization needs ld.so support to check that the
run-time matches link-time function implementation. If a function
in a shared library with st_other localentry non-zero is called
without saving and restoring r2, r2 will be trashed on return, leading
to segfaults. For that reason the optimization does not happen for
weak functions since a weak definition is a fairly solid hint that the
function will likely be overridden. I'm also not enabling the
optimization by default unless glibc-2.26 is detected, which should
have the ld.so checks implemented.
bfd/
* elf64-ppc.c (struct ppc_link_hash_table): Add has_plt_localentry0.
(ppc64_elf_merge_symbol_attribute): Merge localentry bits from
dynamic objects.
(is_elfv2_localentry0): New function.
(ppc64_elf_tls_setup): Default params->plt_localentry0.
(plt_stub_size): Adjust size for tls_get_addr_opt stub.
(build_tls_get_addr_stub): Use a simpler stub when r2 is not saved.
(ppc64_elf_size_stubs): Leave stub_type as ppc_stub_plt_call for
optimized localentry:0 stubs.
(ppc64_elf_build_stubs): Save r2 in ELFv2 __glink_PLTresolve.
(ppc64_elf_relocate_section): Leave nop unchanged for optimized
localentry:0 stubs.
(ppc64_elf_finish_dynamic_sections): Set PPC64_OPT_LOCALENTRY in
DT_PPC64_OPT.
* elf64-ppc.h (struct ppc64_elf_params): Add plt_localentry0.
include/
* elf/ppc64.h (PPC64_OPT_LOCALENTRY): Define.
ld/
* emultempl/ppc64elf.em (params): Init plt_localentry0 field.
(enum ppc64_opt): New, replacing OPTION_* defines. Add
OPTION_PLT_LOCALENTRY, and OPTION_NO_PLT_LOCALENTRY.
(PARSE_AND_LIST_*): Support --plt-localentry and --no-plt-localentry.
* testsuite/ld-powerpc/elfv2so.d: Update.
* testsuite/ld-powerpc/powerpc.exp (TLS opt 5): Use --no-plt-localentry.
* testsuite/ld-powerpc/tlsopt5.d: Update.
The general rule for bfd_arch_info_type->compatible (A, B) is that if A and B
are compatible, then this function should return architecture that is more
"feature-rich", that is, can run both A and B. ARCv2, EM and HS all has same
mach number, so bfd_default_compatible assumes they are the same, and returns
an A. That causes issues with GDB, because GDB assumes that if machines are
compatible, then "compatible ()" always returns same machine regardless of
argument order. As a result GDB gets confused because, for example,
compatible(ARCv2, EM) returns ARCv2, but compatible(EM, ARCv2) returns EM,
hence GDB is not sure if they are compatible and prints a warning.
bfd/ChangeLog:
yyyy-mm-dd Anton Kolesov Anton.Kolesov@synopsys.com
cpu-arc.c (arc_compatible): New function.
ARC600 is already defined as the head of the bfd_arch_arc.
bfd/ChangeLog:
yyyy-mm-dd Anton Kolesov <anton.kolesov@synopsys.com>
* cpu-arc.c (arch_info_struct): Remove duplicate ARC600 entry.
We can't use stat() to get archive element size. Add bfd_get_file_size
to get size for both normal files and archive elements.
bfd/
PR binutils/21519
* bfdio.c (bfd_get_file_size): New function.
* bfd-in2.h: Regenerated.
binutils/
PR binutils/21519
* objdump.c (dump_relocs_in_section): Replace get_file_size
with bfd_get_file_size to get archive element size.
* testsuite/binutils-all/objdump.exp (test_objdump_f): New
proc.
(test_objdump_h): Likewise.
(test_objdump_t): Likewise.
(test_objdump_r): Likewise.
(test_objdump_s): Likewise.
Add objdump tests on archive.
Revert parts of commit fee24f1c5b ("objdump improvements for mips
elf64"), <https://sourceware.org/ml/binutils/2003-03/msg00108.html>, and
make the `->reloc_count' member of `struct bfd_section' hold the actual
number of internal relocations stored in its `->relocation' vector. To
do so adjust `mips_elf64_slurp_one_reloc_table' to set `->reloc_count'
to the actual number of internal relocations retrieved and discard
`mips_elf64_canonicalize_reloc', `mips_elf64_canonicalize_dynamic_reloc'
and their corresponding target macros. Contrary to the description of
`mips_elf64_slurp_one_reloc_table', adjusted appropriately, this makes
generic relocation processing code happy and satisfies the "merge notes
section" binutils test case.
Add extra binutils test cases to expand the coverage of the generic
"merge notes section" test case, now passing with the n64 ABI, across
the MIPS o32, n32 and n64 ABIs regardless of the default ABI selected in
target configuration, and also to verify correctness of the relocations
produced. Conversely, do not provide any additional test cases for the
original issue addressed with the commit referred:
- objdump would display only 1/3 of the total number of relocations,
because it used the external relocation count, but each external
relocation is brought in as 3 internal relocations.
as n64 ABI relocation processing with `objdump -r' and `objdump -R' is
already widely covered across the GAS and LD test suites.
bfd/
* elf64-mips.c (mips_elf64_canonicalize_reloc): Remove prototype
and function.
(mips_elf64_canonicalize_dynamic_reloc): Likewise.
(mips_elf64_slurp_one_reloc_table): Set `reloc_count' to the
actual number of internal relocations retrieved. Adjust
function description.
(bfd_elf64_canonicalize_reloc): Remove macro.
(bfd_elf64_canonicalize_dynamic_reloc): Likewise.
binutils/
* testsuite/binutils-all/mips/mips-note-2.d: New test.
* testsuite/binutils-all/mips/mips-note-2r.d: New test.
* testsuite/binutils-all/mips/mips-note-2-n32.d: New test.
* testsuite/binutils-all/mips/mips-note-2-n64.d: New test.
* testsuite/binutils-all/mips/mips-note-2r-n32.d: New test.
* testsuite/binutils-all/mips/mips-note-2r-n64.d: New test.
* testsuite/binutils-all/mips/mips.exp: Define `has_newabi'.
Run the new tests.
The pr20882 testcase fails on a number of targets that add attribute
or note sections to object files, and the default linker script says
those sections should be kept. This patch changes --gc-sections to
drop debug and special sections like .comment when no SEC_ALLOC
section in an object file is kept. The assumption is that debug
sections are describing code and data that will become part of the
final image in memory.
* elflink.c (_bfd_elf_gc_mark_extra_sections): Don't keep
debug and special sections when no non-note alloc sections in an
object are kept.
If a debug section is referenced by a kept debug section, it should
also be kept.
Some targets, like mips, keep input files when there are some special
sections, like .gnu.attributes, even if input file is unused otherwise.
In this case, all debug sections are kept. The new test will fail on
such targets. We can either fix those targets or XFAIL the test.
bfd/
PR ld/20882
* elflink.c (elf_gc_mark_debug_section): New function.
(_bfd_elf_gc_mark_extra_sections): Mark any debug sections
referenced by kept debug sections.
ld/
PR ld/20882
* testsuite/ld-gc/gc.exp: Run pr20882.
* testsuite/ld-gc/pr20882.d: New file.
* testsuite/ld-gc/pr20882a.s: Likewise.
* testsuite/ld-gc/pr20882b.s: Likewise.
* testsuite/ld-gc/pr20882c.s: Likewise.
dynamic_ref_after_ir_def is a little odd compared to other symbol
flags in that as the name suggests, it is set only for certain
references after a definition. It turns out that setting a flag for
any non-ir reference from a dynamic object can be used to solve the
problem for which this flag was invented, which I think is a cleaner.
This patch does that, and sets non_ir_ref only for regular object
references.
include/
* bfdlink.h (struct bfd_link_hash_entry): Update non_ir_ref
comment. Rename dynamic_ref_after_ir_def to non_ir_ref_dynamic.
ld/
* plugin.c (is_visible_from_outside): Use non_ir_ref_dynamic.
(plugin_notice): Set non_ir_ref for references from regular
objects, non_ir_ref_dynamic for references from dynamic objects.
bfd/
* elf64-ppc.c (add_symbol_adjust): Transfer non_ir_ref_dynamic.
* elflink.c (elf_link_add_object_symbols): Update to use
non_ir_ref_dynamic.
(elf_link_input_bfd): Test non_ir_ref_dynamic in addition to
non_ir_ref.
* linker.c (_bfd_generic_link_add_one_symbol): Likewise.
Add MIPS16e2 ASE support as per the architecture specification[1],
including in particular:
1. A new ELF ASE flag to mark MIPS16e2 binaries.
2. MIPS16e2 instruction assembly support, including a relaxation update
to use LUI rather than an LI/SLL instruction pair for loading the
high part of 32-bit addresses.
3. MIPS16e2 instruction disassembly support, including updated rules for
extended forms of instructions that are now subdecoded and therefore
do not alias to the original MIPS16 ISA revision instructions even
for encodings that are not valid in the MIPS16e2 instruction set.
Add `-mmips16e2' and `-mno-mips16e2' GAS command-line options and their
corresponding `mips16e2' and `no-mips16e2' settings for the `.set' and
`.module' pseudo-ops. Control the availability of the MT ASE subset of
the MIPS16e2 instruction set with a combination of these controls and
the preexisting MT ASE controls.
Parts of this change by Matthew Fortune and Andrew Bennett.
References:
[1] "MIPS32 Architecture for Programmers: MIPS16e2 Application-Specific
Extension Technical Reference Manual", Imagination Technologies
Ltd., Document Number: MD01172, Revision 01.00, April 26, 2016
include/
* elf/mips.h (AFL_ASE_MIPS16E2): New macro.
(AFL_ASE_MASK): Adjust accordingly.
* opcode/mips.h: Document new operand codes defined.
(mips_operand_type): Add OP_REG28 enum value.
(INSN2_SHORT_ONLY): Update description.
(ASE_MIPS16E2, ASE_MIPS16E2_MT): New macros.
bfd/
* elfxx-mips.c (print_mips_ases): Handle MIPS16e2 ASE.
opcodes/
* mips-dis.c (mips_arch_choices): Add ASE_MIPS16E2 and
ASE_MIPS16E2_MT flags to the unnamed MIPS16 entry.
(mips_convert_abiflags_ases): Handle the AFL_ASE_MIPS16E2 flag.
(print_insn_arg) <OP_REG28>: Add handler.
(validate_insn_args) <OP_REG28>: Handle.
(print_mips16_insn_arg): Handle MIPS16 instructions that require
32-bit encoding and 9-bit immediates.
(print_insn_mips16): Handle MIPS16 instructions that require
32-bit encoding and MFC0/MTC0 operand decoding.
* mips16-opc.c (decode_mips16_operand) <'>', '9', 'G', 'N', 'O'>
<'Q', 'T', 'b', 'c', 'd', 'r', 'u'>: Add handlers.
(RD_C0, WR_C0, E2, E2MT): New macros.
(mips16_opcodes): Add entries for MIPS16e2 instructions:
GP-relative "addiu" and its "addu" spelling, "andi", "cache",
"di", "ehb", "ei", "ext", "ins", GP-relative "lb", "lbu", "lh",
"lhu", and "lw" instructions, "ll", "lui", "lwl", "lwr", "mfc0",
"movn", "movtn", "movtz", "movz", "mtc0", "ori", "pause",
"pref", "rdhwr", "sc", GP-relative "sb", "sh" and "sw"
instructions, "swl", "swr", "sync" and its "sync_acquire",
"sync_mb", "sync_release", "sync_rmb" and "sync_wmb" aliases,
"xori", "dmt", "dvpe", "emt" and "evpe". Add split
regular/extended entries for original MIPS16 ISA revision
instructions whose extended forms are subdecoded in the MIPS16e2
ISA revision: "li", "sll" and "srl".
binutils/
* readelf.c (print_mips_ases): Handle MIPS16e2 ASE.
* NEWS: Mention MIPS16e2 ASE support.
gas/
* config/tc-mips.c (RELAX_MIPS16_ENCODE): Add `e2' flag.
(RELAX_MIPS16_E2): New macro.
(RELAX_MIPS16_PIC, RELAX_MIPS16_SYM32, RELAX_MIPS16_NOMACRO)
(RELAX_MIPS16_USER_SMALL, RELAX_MIPS16_USER_EXT)
(RELAX_MIPS16_DSLOT, RELAX_MIPS16_JAL_DSLOT)
(RELAX_MIPS16_EXTENDED, RELAX_MIPS16_MARK_EXTENDED)
(RELAX_MIPS16_CLEAR_EXTENDED, RELAX_MIPS16_ALWAYS_EXTENDED)
(RELAX_MIPS16_MARK_ALWAYS_EXTENDED)
(RELAX_MIPS16_CLEAR_ALWAYS_EXTENDED, RELAX_MIPS16_MACRO)
(RELAX_MIPS16_MARK_MACRO, RELAX_MIPS16_CLEAR_MACRO): Shift bits.
(mips16_immed_extend): New prototype.
(options): Add OPTION_MIPS16E2 and OPTION_NO_MIPS16E2 enum
values.
(md_longopts): Add "mmips16e2" and "mno-mips16e2" options.
(mips_ases): Add "mips16e2" entry.
(mips_set_ase): Handle MIPS16e2 ASE.
(insn_insert_operand): Explicitly handle immediates with MIPS16
instructions that require 32-bit encoding.
(is_opcode_valid_16): Pass enabled ASE bitmask on to
`opcode_is_member'.
(validate_mips_insn): Explicitly handle immediates with MIPS16
instructions that require 32-bit encoding.
(operand_reg_mask) <OP_REG28>: Add handler.
(match_reg28_operand): New function.
(match_operand) <OP_REG28>: Add handler.
(append_insn): Pass ASE_MIPS16E2 setting to RELAX_MIPS16_ENCODE.
(match_mips16_insn): Handle MIPS16 instructions that require
32-bit encoding and `V' and `u' operand codes.
(mips16_ip): Allow any characters except from `.' in opcodes.
(mips16_immed_extend): Handle 9-bit immediates. Do not shuffle
immediates whose width is not one of these listed.
(md_estimate_size_before_relax): Handle MIPS16e2 relaxation.
(mips_relax_frag): Likewise.
(md_convert_frag): Likewise.
(mips_convert_ase_flags): Handle MIPS16e2 ASE.
* doc/as.texinfo (Target MIPS options): Add `-mmips16e2' and
`-mno-mips16e2' options.
(-mmips16e2, -mno-mips16e2): New options.
* doc/c-mips.texi (MIPS Options): Add `-mmips16e2' and
`-mno-mips16e2' options.
(MIPS ASE Instruction Generation Overrides): Add `.set mips16e2'
and `.set nomips16e2'.