This patch is part of a series of patches to add support for Armv8.1-M Mainline
instructions to binutils.
This adds infrastructure for the Branch Future instructions (BF, BFX, BFL, BFLX,
BFCSEL). These are the first instructions in ARM that have more than one
relocations in them.
This is the first infrastructure patch that adds a new bfd_reloc_code_real enum
for the fallback branch offset.
This is common for all such instructions and needs to be resolvable by the
assembler.
ChangeLog entries are as follows :
*** bfd/ChangeLog ***
2019-04-15 Sudakshina Das <sudi.das@arm.com>
* reloc.c (BFD_RELOC_THUMB_PCREL_BRANCH5): New enum.
* bfd-in2.h: Regenerate.
* libbfd.h: Regenerate.
*** gas/ChangeLog ***
2019-04-15 Sudakshina Das <sudi.das@arm.com>
* config/tc-arm.c (md_pcrel_from_section): New switch case
for BFD_RELOC_THUMB_PCREL_BRANCH5.
(v8_1_branch_value_check): New function to check branch
offsets.
(md_appdy_fix): New switch case for
BFD_RELOC_THUMB_PCREL_BRANCH5.
(tc_gen_reloc): Likewise.
*** opcodes/ChangeLog ***
2019-04-15 Sudakshina Das <sudi.das@arm.com>
* arm-dis.c (print_insn_thumb32): Updated to accept new %G pattern.
This patch implements the dsp, fp and fp.dp extensions for Armv8.1-M Mainline.
This patch also removes the fp-armv8 check from the half-precision move
instructions 'do_neon_movhf', as checking that the FP16 instructions extension
feature bit is enabled 'ARM_EXT2_FP16_INST' is enough.
gas/ChangeLog:
2019-04-15 Andre Vieira <andre.simoesdiasvieira@arm.com>
* config/tc-arm.c (do_neon_movhf): Remove fp-armv8 check.
(armv8_1m_main_ext_table): New extension table.
(arm_archs): Use the new extension table.
* doc/c-arm.texi: Add missing arch and document new extensions.
* testsuite/gas/arm/armv8.1-m.main-fp.d: New.
* testsuite/gas/arm/armv8.1-m.main-fp-dp.d: New.
* testsuite/gas/arm/armv8.1-m.main-hp.d: New.
The patch is straightforward, it does the following:
- support the new Tag_CPU_arch build attribute value, ie.:
+ declare the new value
+ update all the asserts forcing logic to be reviewed for new
architectures
+ create a corresponding bfd_mach_arm_8_1M_MAIN enumerator in bfd and
add mapping from Tag_CPU_arch to it
+ teach readelf about new Tag_CPU_arch value
- declare armv8.1-m.main as a supported architecture value
- define Armv8.1-M Mainline in terms of feature bits available
- tell objdump mapping from bfd_mach_arm_8_1M_MAIN enumerator to feature
bits available
- update architecture-specific logic in gas and bfd guarded by the
asserts mentioned above.
- tests for all the above
ChangeLog entries are as follows:
*** bfd/ChangeLog ***
2019-04-15 Thomas Preud'homme <thomas.preudhomme@arm.com>
* archures.c (bfd_mach_arm_8_1M_MAIN): Define.
* bfd-in2.h: Regenerate.
* cpu-arm.c (arch_info_struct): Add entry for Armv8.1-M Mainline.
* elf32-arm.c (using_thumb_only): Return true for Armv8.1-M Mainline
and update assert.
(using_thumb2): Likewise.
(using_thumb2_bl): Update assert.
(arch_has_arm_nop): Likewise.
(bfd_arm_get_mach_from_attributes): Add case for Armv8.1-M Mainline.
(tag_cpu_arch_combine): Add logic for Armv8.1-M Mainline merging.
*** binutils/ChangeLog ***
2019-04-15 Thomas Preud'homme <thomas.preudhomme@arm.com>
* readelf.c (arm_attr_tag_CPU_arch): Add entry for Armv8.1-M Mainline.
*** gas/ChangeLog ***
2019-04-15 Thomas Preud'homme <thomas.preudhomme@arm.com>
* config/tc-arm.c (cpu_arch_ver): Add entry for Armv8.1-M Mainline
Tag_CPU_arch build attribute value. Reindent.
(get_aeabi_cpu_arch_from_fset): Update assert.
(aeabi_set_public_attributes): Update assert for Tag_DIV_use logic.
* testsuite/gas/arm/attr-march-armv8_1-m.main.d: New test.
*** include/ChangeLog ***
2019-04-15 Thomas Preud'homme <thomas.preudhomme@arm.com>
* elf/arm.h (TAG_CPU_ARCH_V8_1M_MAIN): new macro.
(MAX_TAG_CPU_ARCH): Set value to above macro.
* opcode/arm.h (ARM_EXT2_V8_1M_MAIN): New macro.
(ARM_AEXT_V8_1M_MAIN): Likewise.
(ARM_AEXT2_V8_1M_MAIN): Likewise.
(ARM_ARCH_V8_1M_MAIN): Likewise.
*** ld/ChangeLog ***
2019-04-15 Thomas Preud'homme <thomas.preudhomme@arm.com>
* testsuite/ld-arm/attr-merge-13.attr: New test.
* testsuite/ld-arm/attr-merge-13a.s: New test.
* testsuite/ld-arm/attr-merge-13b.s: New test.
*** opcodes/ChangeLog ***
2019-04-15 Thomas Preud'homme <thomas.preudhomme@arm.com>
* arm-dis.c (select_arm_features): Add logic for Armv8.1-M Mainline.
When using writes to memory through a struct to merge and extract
multi-word value, it is the endianness of the host, not the target
that affects which order the component words need to be written into
the structure.
Of the 5 functions adjusted here 4 of them are unused. The 5th,
JOINSIDF will soon be used by the or1k target.
For or1k, simulated on x86-64, this change fixes this function so that
the correct bytes are now returned.
sim/common/ChangeLog:
* cgen-ops.h (SUBWORDXFSI): Compare HOST_BYTE_ORDER not
CURRENT_TARGET_BYTE_ORDER.
(SUBWORDTFSI): Likewise.
(JOINSIDF): Likewise.
(JOINSIXF): Likewise.
(JOINSITF): Likewise.
gas/
* config/tc-mips.c (mips_cpu_info_table): Add i6500. Update
default ASEs for i6400.
* doc/c-mips.texi (-march): Document i6500.
* testsuite/gas/mips/elf_mach_i6400.d: New test.
* testsuite/gas/mips/elf_mach_i6500.d: New test.
* testsuite/gas/mips/mips.exp: Run the new tests.
GAS does not enable implicit ASEs for most MIPS processors.
The rework of option handling done as part of .module implementation
left the implicit ASE logic broken and default enabled ASEs for
most processors did not get applied. This patch ensures the ASE
information is carried forward to the point where it is required.
gas/
* config/tc-mips.c (mips_set_options) <init_ase>: New field.
(file_mips_opts, mips_opts) <init_ase>: Initialize new field.
(file_mips_check_options): Propagate initial ASE settings.
(mips_after_parse_args, parse_code_option): Track the initial
ASE settings for a CPU.
(s_mipsset): Restore the initial ASE settings when reverting
to the default arch.
* testsuite/gas/mips/elf_mach_p6600.d: New test.
* testsuite/gas/mips/mips.exp: Run the new test.
It is no longer true that autoconf/automake cannot substitute vars
with embedded new-lines.
* configure.ac (TDIRS): Build up tdirs in this variable and
AC_SUBST, also using AM_SUBST_NOTMAKE.
* configure: Regenerate.
* Makefile.am (DISTCLEANFILES): Remove tdirs.
* Makefile.in: Regenerate.
The x86-32 ABI specifies 96-bit long double, this was causing a
failure on the test gdb.base/complex-parts.exp.
The problem is that GDB tries to find a builtin floating point type of
the correct size in order to reuse the name of that type as the name
for the components of the complex type being built.
Previously GDB was only aware of floating point types sized 32, 64, or
128 bits. This patch teaches GDB how to handle 96 bit floating point
type.
gdb/ChangeLog:
* dwarf2read.c (dwarf2_init_complex_target_type): Handle complex
target types of size 96-bits, add some additional comments, and
check that the builtin type we found was the correct size.
gdb/ChangeLog:
2019-04-12 Eli Zaretskii <eliz@gnu.org>
* utils.c (prompt_for_continue): Don't restore the styling at the
end, as applied_style has the wrong value. This fixes styling in
long lists of file names that are interrupted by the "Continue?"
prompt.
The local board file ensures that the sysroot is always set to load
files from the local filesystem.
Add a gdbserver test to explicitly test the sysroot set to both the
remote target and the local filesystem.
gdb/testsuite/ChangeLog:
* gdb.server/sysroot.c: New test.
* gdb.server/sysroot.exp: New file.
* lib/gdbserver-support.exp (gdb_target_cmd): Add additional text
matching param.
The language_defn structure has an la_magic field, this used to be
used as a basic check that the language_defn structure had the
expected layout - at least the end of the structure was where we
expected it to be.
This feature only really makes sense if we imagine GDB dynamically
loading language support from dynamic libraries, where a version
mismatch might cause problems.
However, in current GDB language support is statically built into GDB,
and since this commit:
commit 47e77640be
Date: Thu Jul 20 18:28:01 2017 +0100
Make language_def O(1)
the existing (if pointless) check of the la_magic field was removed.
There now appears to be no use of the la_magic field, and I propose
that we delete it.
There should be no user visible changes after this commit.
gdb/ChangeLog:
* ada-lang.c (ada_language_defn): Remove use of LANG_MAGIC.
* c-lang.c (c_language_defn): Likewise.
(cplus_language_defn): Likewise.
(asm_language_defn): Likewise.
(minimal_language_defn): Likewise.
* d-lang.c (d_language_defn): Likewise.
* f-lang.c (f_language_defn): Likewise.
* go-lang.c (go_language_defn): Likewise.
* language.c (unknown_language_defn): Likewise.
(auto_language_defn): Likewise.
* language.h (struct language_defn): Remove la_magic field.
(LANG_MAGIC): Delete.
* m2-lang.c (m2_language_defn): Remove use of LANG_MAGIC.
* objc-lang.c (objc_language_defn): Likewise.
* opencl-lang.c (opencl_language_defn): Likewise.
* p-lang.c (pascal_language_defn): Likewise.
* rust-lang.c (rust_language_defn): Likewise.
Make use of the type_align function and remove riscv_type_alignment as
it is no longer needed. I tested this against a number of RV32 and
RV64 targets, and I also ran the tests with an assertion in place
checking that the old riscv_type_alignment function gives the same
answer as the common type_align function - it does, and all the tests
still pass.
gdb/ChangeLog:
* riscv-tdep.c (riscv_type_align): New function.
(riscv_type_alignment): Delete.
(riscv_arg_location): Use 'type_align'.
(riscv_gdbarch_init): Register riscv_type_align gdbarch function.
The current code in gdbtypes.c:type_align incorrectly returns 0 as the
alignment for a structure containing only static fields. After this
patch the correct value of 1 is returned. The gdb.base/align.exp test
is extended to cover this case.
gdb/ChangeLog:
* gdbtypes.c (type_align): A struct with no non-static fields also
has alignment of 1.
gdb/testsuite/ChangeLog:
* gdb.base/align.exp: Extend test to cover structures containing
only static fields.
This commit resolves a large number of failures in the test script
gdb.base/infcall-nested-structs.exp which were caused by GDB (for
RISC-V) incorrectly handling empty C++ structures when preparing
arguments for a dummy call, or collecting a return value.
The issue is further complicated in that there was a bug in GCC, such
that in some cases GCC would generate incorrect code when passing a
small structure that contained empty sub-structures. This was fixed
in GCC trunk on 5-March-2019, so in order to see the best results with
this patch you'll need a recent version of GCC.
Anything that used to work should continue to work after this patch,
regardless of GCC version being used.
The fix in this commit is that GDB now pays more attention to the
offset of fields within a structure when preparing arguments as in C++
an empty structure has a non-zero size, this is an example:
struct s1 { struct s2 { } empty; int f; };
We previously assumed that 'f' was at offset 0 inside type 's1',
however this is not the case in C++ as 's2' has size 1, and with
alignment 'f' is likely at some even bigger offset inside 's1'.
gdb/ChangeLog:
* riscv-tdep.c (riscv_call_arg_complex_float): Fix offset of first
component to 0.
(riscv_struct_info::riscv_struct_info): Initialise m_offsets
member.
(riscv_struct_info::analyse): New implementation using new
analyse_inner member function.
(riscv_struct_info::field_offset): New member function.
(riscv_struct_info::m_offsets): New member variable.
(riscv_struct_info::analyse_inner): New private member function,
takes the old implementation of riscv_struct_info::analyse but
extended to track field offsets.
(riscv_call_arg_struct): Update the struct folding special cases
to handle cases where empty C++ structs, which are non-zero
length, are found.
(riscv_arg_location): Initialise the length of each location, a
non-zero length now indicates the location is in use.
(riscv_push_dummy_call): Allow for the first location having a
non-zero offset when setting up arguments.
(riscv_return_value): Likewise, but for return values.
I noticed that the "msg" variable in internal_vproblem could be
"const". This seems like an improvement because it can wind up in
rodata.
Tested by rebuilding.
gdb/ChangeLog
2019-04-11 Tom Tromey <tromey@adacore.com>
* utils.c (internal_vproblem): Make "msg" const.
gas/
2019-04-11 Max Filippov <jcmvbkbc@gmail.com>
* testsuite/gas/xtensa/loop-relax-2.d: New test definition.
* testsuite/gas/xtensa/loop-relax.d: New test definition.
* testsuite/gas/xtensa/loop-relax.s: New test source.
* testsuite/gas/xtensa/text-section-literals-1a.d: New test
definition.
* testsuite/gas/xtensa/text-section-literals-2.d: New test
definition.
* testsuite/gas/xtensa/text-section-literals-2.s: New test
source.
* testsuite/gas/xtensa/text-section-literals-2a.d: New test
definition.
* testsuite/gas/xtensa/text-section-literals-3.d: New test
definition.
* testsuite/gas/xtensa/text-section-literals-3.s: New test
source.
* testsuite/gas/xtensa/text-section-literals-4.d: New test
definition.
* testsuite/gas/xtensa/text-section-literals-4.s: New test
source.
* testsuite/gas/xtensa/text-section-literals-4a.d: New test
definition.
gas/
2019-04-11 Max Filippov <jcmvbkbc@gmail.com>
* testsuite/gas/xtensa/all.exp: Remove all expect-based
tests and all explicit run_dump_test / run_list_test
invocations. Add run_dump_tests for all .d files in the
test subdirectory.
* testsuite/gas/xtensa/entry_align.d: New test definition.
* testsuite/gas/xtensa/entry_align.l: New test output.
* testsuite/gas/xtensa/entry_misalign.d: New test definition.
* testsuite/gas/xtensa/entry_misalign2.d: New test definition.
* testsuite/gas/xtensa/j_too_far.d: New test definition.
* testsuite/gas/xtensa/j_too_far.l: New test output.
* testsuite/gas/xtensa/loop_align.d: New test definition.
* testsuite/gas/xtensa/loop_misalign.d: New test definition.
* testsuite/gas/xtensa/trampoline-2.d: New test definition.
* testsuite/gas/xtensa/trampoline-2.l: Remove empty output.
* testsuite/gas/xtensa/xtensa-err.exp: Use positive logic.
gas/
2019-04-11 Max Filippov <jcmvbkbc@gmail.com>
* config/tc-xtensa.c (xtensa_literal_pseudo): Drop code that has
no effect.
(get_literal_pool_location): Only search for the literal pool
when auto litpools is used, otherwise take one recorded in the
tc_segment_info_data.
(xtensa_assign_litpool_addresses): New function.
(xtensa_move_literals): Don't duplicate 'literal pool location
required...' error message. Call xtensa_assign_litpool_addresses.
Provide literal position at the beginning of each section for literal
space reserved by relaxations when text-section-literals or
auto-litpools options are used. Remove code that adds fill frag to the
literal section for every .literal_position directive to avoid creation
of empty literal sections.
Fix auto-litpools tests that got literal pool address changes.
gas/
2019-04-11 Max Filippov <jcmvbkbc@gmail.com>
* config/tc-xtensa.c (xtensa_is_init_fini): Add declaration.
(xtensa_mark_literal_pool_location): Don't add fill frag to literal
section that records literal pool location.
(md_begin): Call xtensa_mark_literal_pool_location when text
section literals or auto litpools are used.
(xtensa_elf_section_change_hook): Call
xtensa_mark_literal_pool_location when text section literals or
auto litpools are used, there's no literal pool location defined
for the current section and it's not .init or .fini.
* testsuite/gas/xtensa/auto-litpools-first1.d: Fix up addresses.
* testsuite/gas/xtensa/auto-litpools-first2.d: Likewise.
* testsuite/gas/xtensa/auto-litpools.d: Likewise.
Add a new option to disable the listing of discarded sections
in map file output. The use case stems from a large application
built with -ffunction-sections --gc-sections where the list of
discarded sections blows up the map file output. The default
behaviour remains to print discarded sections, but the new option
allows us to disable it.
ld/
* NEWS: Mention new option --no-print-map-discarded.
* ld.h (ld_config_type) <print_map_discarded>: New field.
* ldlang.c (lang_map): Conditionally output discarded sections
in map files based on configuration option.
* ldlex.h (option_values) <OPTION_PRINT_MAP_DISCARDED,
OPTION_NO_PRINT_MAP_DISCARDED>: New.
* ldmain.c (main): Enabled print_map_discarded by default.
* lexsup.c (ld_options): Add new command-line options.
(parse_args) <OPTION_NO_PRINT_MAP_DISCARDED,
OPTION_PRINT_MAP_DISCARDED>: New cases.
* ld.texi: Document new options.
* testsuite/ld-gc/gc.exp: Add new test.
* testsuite/ld-gc/skip-map-discarded.s: New file.
* testsuite/ld-gc/skip-map-discarded.d: New file.
* testsuite/ld-gc/skip-map-discarded.map: New file.
We can use CC_WITH_TWEAKS_FLAGS when cd-ing into the gdb build subdir and
invoking make check:
...
$ cd $objdir/gdb
$ make check \
RUNTESTFLAGS='--target_board=cc-with-tweaks' \
CC_WITH_TWEAKS_FLAGS='-z'
...
But when cd-ing into the top-level build dir and invoking make check-gdb
instead:
...
$ cd $objdir
$ make check-gdb \
RUNTESTFLAGS='--target_board=cc-with-tweaks' \
CC_WITH_TWEAKS_FLAGS='-z'
...
using CC_WITH_TWEAKS_FLAGS has no effect, because CC_WITH_TWEAKS_FLAGS is not
passed down from the top level Makefile.
Add cc-with-dwz.exp and cc-with-dwz-m.exp, that don't require
CC_WITH_TWEAKS_FLAGS to be set in the make invocation, allowing us to run these
test configurations from the toplevel build dir:
...
$ cd $objdir
$ make check-gdb \
RUNTESTFLAGS='--target_board=cc-with-dwz'
...
Tested on x86_64-linux.
gdb/testsuite/ChangeLog:
2019-04-11 Tom de Vries <tdevries@suse.de>
* boards/cc-with-dwz-m.exp: New file.
* boards/cc-with-dwz.exp: New file.
* boards/cc-with-tweaks.exp: Note that check-gdb doesn't work.
Add -static-pie tests for DT_INIT_ARRAY, DT_FINI_ARRAY, DT_PREINIT_ARRAY
and IFUNC.
* testsuite/config/default.exp (STATIC_PIE_LDFLAGS): New. Set
to "-static-pie" if target compiler supports it.
* testsuite/ld-elf/elf.exp: Run -static-pie tests if
$STATIC_PIE_LDFLAGS isn't empty.
* testsuite/ld-ifunc/ifunc.exp: Likewise.
When using DF_BIND_NOW on AArch64 we don't reserve the GOT slot for a TLSDESC,
but we still emitted DT_TLSDESC_GOT and DT_TLSDESC_PLT. This caused random
memory corruption as the "special" value of (bfd_vma)-1 would be set for
dt_tlsdesc_got.
Since we don't have a value of dt_tlsdesc_got I also don't emit DT_TLSDESC_PLT
now becuase it would point to an incomplete PLT. To be able to write the PLT
entry DT_TLSDESC_GOT is needed and since we don't have one we can't write the
PLT entry either.
It is my understanding that GLIBC doesn't need these two entries when not lazy
loading. Conversely AArch32 does not reserve neither the GOT not the PLT slot
when doing DF_BIND_NOW.
AArch32 does not need these checks because these values are initialized to 0
and so the if (...) checks don't pass, but on AArch64 these are initialized
to (bfd_vma)-1 and thus we need some extra checks.
bfd/ChangeLog:
PR ld/24302
* elfnn-aarch64.c (elfNN_aarch64_size_dynamic_sections): Don't emit
DT_TLSDESC_GOT and DT_TLSDESC_PLT when DF_BIND_NOW.
(elfNN_aarch64_finish_dynamic_sections): Don't write PLT if DF_BIND_NOW.
ld/ChangeLog:
PR ld/24302
* testsuite/ld-aarch64/aarch64-elf.exp: Add new test.
* testsuite/ld-aarch64/tls-relax-gdesc-le-now.d: New test.
This patch updates the Store allocation tags instructions in
Armv8.5-A Memory Tagging Extension. This is part of the changes
that have been introduced recently in the 00bet10 release
All of these instructions have an updated register operand (Xt -> <Xt|SP>)
- STG <Xt|SP>, [<Xn|SP>, #<simm>]
- STG <Xt|SP>, [<Xn|SP>, #<simm>]!
- STG <Xt|SP>, [<Xn|SP>], #<simm>
- STZG <Xt|SP>, [<Xn|SP>, #<simm>]
- STZG <Xt|SP>, [<Xn|SP>, #<simm>]!
- STZG <Xt|SP>, [<Xn|SP>], #<simm>
- ST2G <Xt|SP>, [<Xn|SP>, #<simm>]
- ST2G <Xt|SP>, [<Xn|SP>, #<simm>]!
- ST2G <Xt|SP>, [<Xn|SP>], #<simm>
- STZ2G <Xt|SP>, [<Xn|SP>, #<simm>]
- STZ2G <Xt|SP>, [<Xn|SP>, #<simm>]!
- STZ2G <Xt|SP>, [<Xn|SP>], #<simm>
In order to accept <Rt|SP> a new operand type Rt_SP is introduced which has
the same field as FLD_Rt but follows other semantics of Rn_SP.
*** gas/ChangeLog ***
2019-04-11 Sudakshina Das <sudi.das@arm.com>
* config/tc-aarch64.c (process_omitted_operand): Add case for
AARCH64_OPND_Rt_SP.
(parse_operands): Likewise.
* testsuite/gas/aarch64/armv8_5-a-memtag.d: Update tests.
* testsuite/gas/aarch64/armv8_5-a-memtag.s: Likewise.
* testsuite/gas/aarch64/illegal-memtag.l: Likewise.
* testsuite/gas/aarch64/illegal-memtag.s: Likewise.
*** include/ChangeLog ***
2019-04-11 Sudakshina Das <sudi.das@arm.com>
* opcode/aarch64.h (enum aarch64_opnd): Add AARCH64_OPND_Rt_SP.
*** opcodes/ChangeLog ***
2019-04-11 Sudakshina Das <sudi.das@arm.com>
* aarch64-opc.c (aarch64_print_operand): Add case for
AARCH64_OPND_Rt_SP.
(verify_constraints): Likewise.
* aarch64-tbl.h (QL_LDST_AT): Update to add SP qualifier.
(struct aarch64_opcode): Update stg, stzg, st2g, stz2g instructions
to accept Rt|SP as first operand.
(AARCH64_OPERANDS): Add new Rt_SP.
* aarch64-asm-2.c: Regenerated.
* aarch64-dis-2.c: Regenerated.
* aarch64-opc-2.c: Regenerated.
This patch adds the new LDGM/STGM instructions of the
Armv8.5-A Memory Tagging Extension. This is part of the changes
that have been introduced recently in the 00bet10 release
The instructions are as follows:
LDGM Xt, [<Xn|SP>]
STGM Xt, [<Xn|SP>]
*** gas/ChangeLog ***
2019-04-11 Sudakshina Das <sudi.das@arm.com>
* testsuite/gas/aarch64/armv8_5-a-memtag.d: New tests for ldgm and stgm.
* testsuite/gas/aarch64/armv8_5-a-memtag.s: Likewise.
* testsuite/gas/aarch64/illegal-memtag.l: Likewise.
* testsuite/gas/aarch64/illegal-memtag.s: Likewise.
*** opcodes/ChangeLog ***
2019-04-11 Sudakshina Das <sudi.das@arm.com>
* aarch64-asm-2.c: Regenerated.
* aarch64-dis-2.c: Likewise.
* aarch64-opc-2.c: Likewise.
* aarch64-tbl.h (aarch64_opcode): Add new ldgm and stgm.
A recent change made the AArch64 self tests resuse the saved regs
cache, rather than creating a new one. Ensure it is reset to default
values between tests.
Do this by splitting the reset functionality from trad_frame_alloc_saved_regs
into a new function.
Fixes selftest on AArch64.
gdb/ChangeLog:
* aarch64-tdep.c (aarch64_analyze_prologue_test): Reset saved regs.
* trad-frame.c (trad_frame_reset_saved_regs): New function.
(trad_frame_alloc_saved_regs): Call trad_frame_reset_saved_regs.
* trad-frame.h (trad_frame_reset_saved_regs): New declaration.
__rel[a]_iplt_start and __rel[a]_iplt_end are defined to handle IFUNC in
static executables. For PIE, since IFUNC is handled as the part of the
normal dynamic relocation processing, there is no need to define these
symbols in PIE.
* scripttempl/elf.sc (CREATE_PIC): New. Set for CREATE_SHLIB or
CREATE_PIE.
(__rel_iplt_start): Don't define for CREATE_PIC.
(__rel_iplt_end): Likewise.
(__rela_iplt_start): Likewise.
(__rela_iplt_end): Likewise.
This commit fixes some failures in gdb.base/interrupt.exp
when debugging a 32-bit i386 linux inferior from an amd64 host.
When running the following test...
make check RUNTESTFLAGS="--target_board unix/-m32 interrupt.exp"
... without this commit, I see the following output:
FAIL: gdb.base/interrupt.exp: continue (the program exited)
FAIL: gdb.base/interrupt.exp: echo data
FAIL: gdb.base/interrupt.exp: Send Control-C, second time
FAIL: gdb.base/interrupt.exp: signal SIGINT (the program is no longer running)
ERROR: Undefined command "".
ERROR: GDB process no longer exists
=== gdb Summary ===
When the test is run with this commit in place, we see 12 passes
instead. This is the desired behavior.
Analysis:
On Linux, when a syscall is interrupted by a signal, the syscall
may return -ERESTARTSYS when a signal occurs. Doing so indicates that
the syscall is restartable. Then, depending on settings associated
with the signal handler, and after the signal handler is called, the
kernel can then either return -EINTR or can cause the syscall to be
restarted. In this discussion, we are concerned with the latter
case.
On i386, the kernel returns this status via the EAX register.
When debugging a 32-bit (i386) process from a 64-bit (amd64)
GDB, the debugger fetches 64-bit registers even though the
process being debugged is 32-bit. Since we're debugging a 32-bit
target, only 32 bits are being saved in the register cache.
Now, ideally, GDB would save all 64-bits in the regcache and
then would be able to restore those same values when it comes
time to continue the target. I've looked into doing this, but
it's not easy and I don't see many benefits to doing so. One
benefit, however, would be that EAX would appear as a negative
value for doing syscall restarts.
At the moment, GDB is setting the high 32 bits of RAX (and other
registers too) to 0. So, when GDB restores EAX just prior to
a syscall restart, the high 32 bits of RAX are zeroed, thus making
it look like a positive value. For this particular purpose, we
need to sign extend EAX so that RAX will appear as a negative
value when EAX is set to -ERESTARTSYS. This in turn will cause
the signal handling code in the kernel to recognize -ERESTARTSYS
which will in turn cause the syscall to be restarted.
This commit is based on work by Jan Kratochvil from 2009:
https://sourceware.org/ml/gdb-patches/2009-11/msg00592.html
Jan's patch had the sign extension code in amd64-nat.c. Several
other native targets make use of this code, so it seemed better
to move the sign extension code to a linux specific file. I
also added similar code to gdbserver.
Another approach is to fix the problem in the kernel. Hui Zhu
tried to get a fix into the kernel back in 2014, but it was not
accepted. Discussion regarding this approach may be found here:
https://lore.kernel.org/patchwork/patch/457841/
Even if a fix were to be put into the kernel, we'd still need
some kind of fix in GDB in order to support older kernels.
Finally, I'll note that Fedora has been carrying a similar patch for
at least nine years. Other distributions, including RHEL and CentOS
have picked up this change and have been using it too.
gdb/ChangeLog:
* amd64-linux-nat.c (amd64_linux_collect_native_gregset): New
function.
(fill_gregset): Call amd64_linux_collect_native_gregset instead
of amd64_collect_native_gregset.
(amd64_linux_nat_target::store_registers): Likewise.
gdb/gdbserver/ChangeLog:
* linux-x86-low.c (x86_fill_gregset): Sign extend EAX value
when using a 64-bit gdbserver.
PR 89394
* cp-demangle.c (cplus_demangle_fill_name): Reject negative
lengths.
(d_count_templates_scopes): Replace num_templates and num_scopes
parameters with a struct d_print_info pointer parameter. Adjust
body of the function accordingly. Add recursion counter and check
that the recursion limit is not reached.
(d_print_init): Pass dpi parameter to d_count_templates_scopes.
Reset recursion counter afterwards, unless the recursion limit was
reached.
This introduces a new iterator and range adapter for iteration over
the separate debug files of a given objfile. As in the current
approach, the requested objfile is returned first, followed by the
separate debug objfiles.
gdb/ChangeLog
2019-04-10 Tom Tromey <tom@tromey.com>
* symtab.c (lookup_global_symbol_from_objfile)
(lookup_symbol_in_objfile_from_linkage_name): Use the iterator.
* objfiles.h (class separate_debug_iterator): New.
(class separate_debug_range): New.
(struct objfile) <separate_debug_objfiles>: New method.
(objfile_separate_debug_iterate): Don't declare.
* objfiles.c (separate_debug_iterator::operator++): Rename from
objfile_separate_debug_iterate.
(objfile_relocate, objfile_rebase, objfile_has_symbols): Use the
iterator.
* minsyms.c (lookup_minimal_symbol_by_pc_section): Use the
iterator.
While working on objfiles I noticed a typo in one comment, and another
comment that, as far as I can tell, has been obsolete for a very long
time.
gdb/ChangeLog
2019-04-10 Tom Tromey <tom@tromey.com>
* symfile.c (reread_symbols): Remove old comment.
* objfiles.c (free_all_objfiles): Fix a typo.
The "object_files" macro is sometimes used when iterating over
objfiles. This patch removes a few such uses in favor of the new
range adapter.
gdb/ChangeLog
2019-04-10 Tom Tromey <tom@tromey.com>
* ia64-tdep.c (ia64_get_dyn_info_list): Use foreach.
* minsyms.c (lookup_minimal_symbol): Use foreach.
(lookup_minimal_symbol_text, lookup_minimal_symbol_by_pc_name)
(lookup_minimal_symbol_solib_trampoline): Likewise.
* symfile.c (reread_symbols): Use foreach.
The fix H.J. implemented for PR gas/22791 in the thread starting at
[PATCH] x86-64: Treat PC32 relocation with branch as PLT32
https://sourceware.org/ml/binutils/2018-02/msg00065.html
is causing problems on Solaris/x86. The native linker is strongly
preferred there, and there's no intention of implementing the linker
optimization he plans there. Besides, the kernel runtime linker,
otherwise has no need to deal with that reloc at all, and instead of
adding (possibly even more) workarounds with no benefit, it seems
appropriate to disable the R_X86_64_PLT32 generation as branch marker on
Solaris/x86 in the first place.
The patch itself is trivial, the only complication is adapting the
testsuite. Since I've found no way to have conditional sections in the
.d files, I've instead used the solution already found elsewhere of
having separate .d files for the affected tests in an i386/solaris
subdirectory and skipping the original ones.
Tested on amd64-pc-solaris2.11 and x86_64-pc-linux-gnu without
regressions.
* config/tc-i386.c (need_plt32_p) [TE_SOLARIS]: Return FALSE.
* testsuite/gas/i386/solaris/solaris.exp: New driver.
* testsuite/gas/i386/solaris/reloc64.d,
testsuite/gas/i386/solaris/x86-64-jump.d,
testsuite/gas/i386/solaris/x86-64-mpx-branch-1.d,
testsuite/gas/i386/solaris/x86-64-mpx-branch-2.d,
testsuite/gas/i386/solaris/x86-64-nop-3.d,
testsuite/gas/i386/solaris/x86-64-nop-4.d,
testsuite/gas/i386/solaris/x86-64-nop-5.d,
testsuite/gas/i386/solaris/x86-64-relax-2.d,
testsuite/gas/i386/solaris/x86-64-relax-3.d: New tests.
* testsuite/gas/i386/reloc64.d,
testsuite/gas/i386/x86-64-jump.d,
testsuite/gas/i386/x86-64-mpx-branch-1.d,
testsuite/gas/i386/x86-64-mpx-branch-2.d,
testsuite/gas/i386/x86-64-nop-3.d,
testsuite/gas/i386/x86-64-nop-4.d,
testsuite/gas/i386/x86-64-nop-5.d,
testsuite/gas/i386/x86-64-relax-2.d,
testsuite/gas/i386/x86-64-relax-3.d: Skip on *-*-solaris*.
This fixes the csky-elf ld-elf/pr21884 and ld-unique/pr21529 failures,
by disabling branch stubs for binary (and other non-csky) output.
The csky-linux target gets branch stubs off by default because
presumably there are reasons why branch stubs were disabled, but
rather than killing the support completely it can be enabled by
--branch-stub.
* emultempl/cskyelf.em (csk_elf_before_parse): New function,
setting use_branch_stub false for linux.
(csky_elf_create_output_section_statements): Do emit this
function and all others in the file for linux, plus the branch
option control. Disable branch stubs when non-ELF.
