platform/upstream/binutils.git
15 months agobpf: gas: add field overflow checking to the BPF assembler
Jose E. Marchesi [Sun, 30 Jul 2023 19:01:03 +0000 (21:01 +0200)]
bpf: gas: add field overflow checking to the BPF assembler

This patch makes the BPF assembler to throughfully check for overflow
in immediates.  This includes relaxed instructions.

Tested in bpf-unknown-none.

gas/ChangeLog:

2023-07-30  Jose E. Marchesi  <jose.marchesi@oracle.com>

* config/tc-bpf.c (signed_overflow): Copy function from
tc-aarch64.c.
(encode_insn): Check for overflow in constant immediates.
(add_relaxed_insn): Pass relax argument to encode_insn.
(add_fixed_insn): Likewise.
* testsuite/gas/bpf/disp16-overflow.d: New file.
* testsuite/gas/bpf/disp16-overflow.s: Likewise.
* testsuite/gas/bpf/disp16-overflow.l: Likewise.
* testsuite/gas/bpf/disp32-overflow.d: Likewise.
* testsuite/gas/bpf/disp32-overflow.s: Likewise.
* testsuite/gas/bpf/disp32-overflow.l: Likewise.
* testsuite/gas/bpf/imm32-overflow.d: Likewise.
* testsuite/gas/bpf/imm32-overflow.s: Likewise.
* testsuite/gas/bpf/imm32-overflow.l: Likewise.
* testsuite/gas/bpf/offset16-overflow.d: Likewise.
* testsuite/gas/bpf/offset16-overflow.s: Likewise.
* testsuite/gas/bpf/offset16-overflow.l: Likewise.
* testsuite/gas/bpf/disp16-overflow-relax.d: Likewise.
* testsuite/gas/bpf/disp16-overflow-relax.l: Likewise.
* testsuite/gas/bpf/disp16-overflow-relax.s: Likewise.
* testsuite/gas/bpf/jump-relax-jump-be.d: New file.
* testsuite/gas/bpf/bpf.exp: Run new tests.

15 months agoUpdate how to make a release document after the 2.41 release
Nick Clifton [Sun, 30 Jul 2023 15:07:09 +0000 (16:07 +0100)]
Update how to make a release document after the 2.41 release

15 months agoAutomatic date update in version.in
GDB Administrator [Sun, 30 Jul 2023 00:00:25 +0000 (00:00 +0000)]
Automatic date update in version.in

15 months agoAutomatic date update in version.in
GDB Administrator [Sat, 29 Jul 2023 00:00:20 +0000 (00:00 +0000)]
Automatic date update in version.in

15 months agobpf: remove spurious comment from tc-bpf.c
Jose E. Marchesi [Fri, 28 Jul 2023 20:20:41 +0000 (22:20 +0200)]
bpf: remove spurious comment from tc-bpf.c

15 months agobpf: gas: support relaxation of V4 jump instructions
Jose E. Marchesi [Thu, 27 Jul 2023 16:17:35 +0000 (18:17 +0200)]
bpf: gas: support relaxation of V4 jump instructions

The BPF jump-always instruction (JA), like all other jump instructions
in the ISA, get a signed 16-bit displacement target argument denoted
in number of 64-bit words minus one.  This can sometimes be overflown.

The BPF V4 ISA thus introduced support for a jump-always
instruction (JAL) that gets a signed 32-bit displacement instead.

This patch makes the BPF assembler to perform the following
relaxations when the disp16 field gets overflown, unless the option
-mno-relax is specified:

  JA disp16  -> JAL disp32
  Jxx disp16 -> Jxx +1; JA +1; JAL disp32

Documentation and tests added.
Tested in bpf-unknown-none.

gas/ChangeLog:

2023-07-28  Jose E. Marchesi  <jose.marchesi@oracle.com>

PR gas/30690
* config/tc-bpf.c (struct bpf_insn): Add fields is_relaxable and
relaxed_exp.
(enum options): Add OPTION_NO_RELAX.
(md_longopts): Likewise for -mno-relax.
(do_relax): New global.
(md_parse_option): Handle OPTION_NO_RELAX.
(RELAX_BRANCH_ENCODE): Define.
(RELAX_BRANCH_P): Likewise.
(RELAX_BRANCH_LENGTH): Likewise.
(RELAX_BRANCH_CONST): Likewise.
(RELAX_BRANCH_UNCOND): Likewise.
(relaxed_branch_length): New function.
(md_estimate_size_before_relax): Likewise.
(read_insn_word): Likewise.
(encode_int16): Likewise.
(encode_int32): Likewise.
(write_insn_bytes): Likewise.
(md_convert_frag): Likewise.
(encode_insn): Likewise.
(install_insn_fixups): Likewise.
(add_fixed_insn): Likewise.
(add_relaxed_insn): Likewise.
(md_assemble): Move instruction encoding logic to the above
new functions.
* testsuite/gas/bpf/jump-relax-ja.d: New test.
* testsuite/gas/bpf/jump-relax-ja-be.d: Likewise.
* testsuite/gas/bpf/jump-relax-ja.s: And corresponding source.
* testsuite/gas/bpf/jump-relax-jump.d: New test.
* testsuite/gas/bpf/jump-relax-jump-be.d: Likewise.
* testsuite/gas/bpf/jump-relax-jump.s: And corresponding source.
* testsuite/gas/bpf/bpf.exp: Run new tests.
* doc/c-bpf.texi (BPF Options): Document -mno-relax.

15 months ago[gdb] Rename variable main_thread to main_thread_id
Tom de Vries [Fri, 28 Jul 2023 14:01:51 +0000 (16:01 +0200)]
[gdb] Rename variable main_thread to main_thread_id

I noticed that the variable main_thread:
...
/* The main thread.  */

static std::thread::id main_thread;
...
has a confusing name and corresponding comment, because it doesn't contain the
main thread, but rather the main thread's std::thread::id.

Fix this by renaming to main_thread_id.

Tested on x86_64-linux.

Approved-By: Tom Tromey <tom@tromey.com>
15 months agoRe-acquire GIL earlier in gdbpy_parse_and_eval
Tom Tromey [Thu, 27 Jul 2023 13:54:18 +0000 (07:54 -0600)]
Re-acquire GIL earlier in gdbpy_parse_and_eval

Tom de Vries filed a bug about an intermittent gdb DAP failure -- and
coincidentally, at the same time, Alexandra Hájková sent email about a
somewhat similar failure.

After looking into this for a while (with no results) using ASan and
valgrind, I found that setting PYTHONMALLOC=malloc_debug found the bug
instantly.

The problem is that gdbpy_parse_and_eval releases the GIL while
calling parse_and_eval, but fails to re-acquire it before calling
value_to_value_object.  This is easily fixed by introducing a new
scope.

I wonder whether the test suite should unconditionally set
PYTHONMALLOC=malloc_debug.

Tested-by: Tom de Vries <tdevries@suse.de>
Reviewed-By: Tom de Vries <tdevries@suse.de>
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30686

15 months agogas: amend X_unsigned uses
Jan Beulich [Fri, 28 Jul 2023 10:34:30 +0000 (12:34 +0200)]
gas: amend X_unsigned uses

PR gas/30688

X_unsigned being clear does not indicate a negative number; it merely
indicates a signed one (whose sign may still be clear). Amend two uses
by an actual value check.

15 months agoMIPS: Support `-gnuabi64' target triplet suffix for 64-bit Linux targets
YunQiang Su [Fri, 28 Jul 2023 05:00:37 +0000 (06:00 +0100)]
MIPS: Support `-gnuabi64' target triplet suffix for 64-bit Linux targets

Make the n64 ABI the default for 64-bit Linux targets specified with
`-gnuabi64' suffix included in the target triplet, for configurations
such as the Debian mips64el and mips64r6el ports.  Adjust testsuite
configuration accordingly.

There are the following regressions with the new target triplet:

mips64-linux-gnuabi64  +FAIL: readelf -S bintest
mips64-linux-gnuabi64  +FAIL: MIPS reloc estimation 1
mips64el-linux-gnuabi64  +FAIL: readelf -S bintest
mips64el-linux-gnuabi64  +FAIL: MIPS reloc estimation 1

The `readelf' issue comes from a difference in section headers produced
that the `binutils/testsuite/binutils-all/readelf.s-64' pattern template
does not match.  While there has been a precedent it does not appear to
me that there is a clear advantage from adding more and more variations
to the template rather than forking the existing template into multiple
ones for a more exact match.  So this is best deferred to a separate
discussion.

The MIPS reloc estimation issue is an actual bug in `objdump', which
discards a number of trailing entries from output here for n64 composed
relocations:

DYNAMIC RELOCATION RECORDS
OFFSET           TYPE              VALUE
0000000000000000 R_MIPS_NONE       *ABS*
0000000000000000 R_MIPS_NONE       *ABS*

and consequently `ld/testsuite/ld-mips-elf/reloc-estimate-1.d' does not
match even though ELF output produced is correct according to `readelf':

Relocation section '.rel.dyn' at offset 0x10400 contains 2 entries:
  Offset          Info           Type           Sym. Value    Sym. Name
000000000000  000000000000 R_MIPS_NONE
                    Type2: R_MIPS_NONE
                    Type3: R_MIPS_NONE
000000010000  000300001203 R_MIPS_REL32      0000000000010010 foo@@V2
                    Type2: R_MIPS_64
                    Type3: R_MIPS_NONE

As a genuine bug this has to be handled separately.

Co-Authored by: Maciej W. Rozycki <macro@orcam.me.uk>

bfd/
* config.bfd: Add `mips64*el-*-linux*-gnuabi64' and
`mips64*-*-linux*-gnuabi64' targets.

binutils/
* testsuite/binutils-all/mips/mips.exp: Handle `*-*-*-gnuabi64'
targets.
* testsuite/binutils-all/objcopy.exp: Handle
`mips64*-*-*-gnuabi64' targets.
* testsuite/binutils-all/remove-relocs-01.d: Likewise.
* testsuite/binutils-all/remove-relocs-04.d: Likewise.
* testsuite/binutils-all/remove-relocs-05.d: Likewise.
* testsuite/binutils-all/remove-relocs-06.d: Likewise.

gas/
* configure.ac: Handle `mips64*-linux-gnuabi64' targets.
* configure: Regenerate.
* testsuite/gas/mips/compact-eh-eb-7.d: Handle
`mips64*-*-*-gnuabi64' targets.
* testsuite/gas/mips/compact-eh-el-7.d: Likewise.

ld/
* configure.tgt: Add `mips64*el-*-linux-gnuabi64' and
`mips64*-*-linux-gnuabi64' targets.
* testsuite/ld-undefined/undefined.exp: Handle
`mips64*-*-*-gnuabi64' targets.
* testsuite/ld-mips-elf/attr-gnu-4-10.d: Likewise.
* testsuite/ld-mips-elf/compact-eh6.d: Likewise.
* testsuite/ld-mips-elf/mips-elf.exp: Handle `*-*-*-gnuabi64'
targets.

15 months agoMIPS/GAS/testsuite: Fix n64 compact EH failures
YunQiang Su [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
MIPS/GAS/testsuite: Fix n64 compact EH failures

Expect a `.MIPS.options' section alternatively to `.reginfo' and ignore
contents of either as irrelevant for all the affected compact EH tests,
removing these regressions:

mips64-openbsd  -FAIL: Compact EH EB #1 with personality ID and FDE data
mips64-openbsd  -FAIL: Compact EH EB #2 with personality routine and FDE data
mips64-openbsd  -FAIL: Compact EH EB #3 with personality id and large FDE data
mips64-openbsd  -FAIL: Compact EH EB #4 with personality id, FDE data and LSDA
mips64-openbsd  -FAIL: Compact EH EB #5 with personality routine, FDE data and LSDA
mips64-openbsd  -FAIL: Compact EH EB #6 with personality id, LSDA and large FDE data
mips64-openbsd  -FAIL: Compact EH EL #1 with personality ID and FDE data
mips64-openbsd  -FAIL: Compact EH EL #2 with personality routine and FDE data
mips64-openbsd  -FAIL: Compact EH EL #3 with personality id and large FDE data
mips64-openbsd  -FAIL: Compact EH EL #4 with personality id, FDE data and LSDA
mips64-openbsd  -FAIL: Compact EH EL #5 with personality routine, FDE data and LSDA
mips64-openbsd  -FAIL: Compact EH EL #6 with personality id, LSDA and large FDE data
mips64el-openbsd  -FAIL: Compact EH EB #1 with personality ID and FDE data
mips64el-openbsd  -FAIL: Compact EH EB #2 with personality routine and FDE data
mips64el-openbsd  -FAIL: Compact EH EB #3 with personality id and large FDE data
mips64el-openbsd  -FAIL: Compact EH EB #4 with personality id, FDE data and LSDA
mips64el-openbsd  -FAIL: Compact EH EB #5 with personality routine, FDE data and LSDA
mips64el-openbsd  -FAIL: Compact EH EB #6 with personality id, LSDA and large FDE data
mips64el-openbsd  -FAIL: Compact EH EL #1 with personality ID and FDE data
mips64el-openbsd  -FAIL: Compact EH EL #2 with personality routine and FDE data
mips64el-openbsd  -FAIL: Compact EH EL #3 with personality id and large FDE data
mips64el-openbsd  -FAIL: Compact EH EL #4 with personality id, FDE data and LSDA
mips64el-openbsd  -FAIL: Compact EH EL #5 with personality routine, FDE data and LSDA
mips64el-openbsd  -FAIL: Compact EH EL #6 with personality id, LSDA and large FDE data

Co-Authored-By: Maciej W. Rozycki <macro@orcam.me.uk>
gas/
* testsuite/gas/mips/compact-eh-eb-1.d: Accept `.MIPS.options'
section as an alternative to `.reginfo' and ignore contents of
either.
* testsuite/gas/mips/compact-eh-eb-2.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-3.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-4.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-5.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-6.d: Likewise.
* testsuite/gas/mips/compact-eh-el-1.d: Likewise.
* testsuite/gas/mips/compact-eh-el-2.d: Likewise.
* testsuite/gas/mips/compact-eh-el-3.d: Likewise.
* testsuite/gas/mips/compact-eh-el-4.d: Likewise.
* testsuite/gas/mips/compact-eh-el-5.d: Likewise.
* testsuite/gas/mips/compact-eh-el-6.d: Likewise.

15 months agotestsuite: Handle composed R_MIPS_NONE relocations
YunQiang Su [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
testsuite: Handle composed R_MIPS_NONE relocations

MIPS n64 ABI has a peculiarity where all relocations are composed of
three, with subsequent relocation types set to R_MIPS_NONE if further
calculation is not required.  Example output produced by `readelf' and
`objdump' for such relocations is:

  Offset          Info           Type           Sym. Value    Sym. Name + Addend
000000000000  000800000002 R_MIPS_32         0000000000000000 foo + 0
                    Type2: R_MIPS_NONE
                    Type3: R_MIPS_NONE

and:

OFFSET           TYPE              VALUE
0000000000000000 R_MIPS_32         foo
0000000000000000 R_MIPS_NONE       *ABS*
0000000000000000 R_MIPS_NONE       *ABS*

respectively.  The presence of these extra R_MIPS_NONE entries is not
relevant for generic or even some MIPS tests, so optionally match them
with the respective dump patterns, also discarding `xfail' annotation
for MIPS/OpenBSD targets from gas/elf/missing-build-notes.d, removing
these regressions:

mips64-openbsd  -FAIL: readelf -r bintest
mips64-openbsd  -FAIL: forward expression
mips64-openbsd  -FAIL: assignment tests
mips64-openbsd  -FAIL: gas/all/none
mips64-openbsd  -XFAIL: gas/elf/missing-build-notes
mips64-openbsd  -FAIL: macro test 2
mips64-openbsd  -FAIL: macro irp
mips64-openbsd  -FAIL: macro rept
mips64-openbsd  -FAIL: nested irp/irpc/rept
mips64-openbsd  -FAIL: macro vararg
mips64-openbsd  -FAIL: mips jalx
mips64-openbsd  -FAIL: ST Microelectronics Loongson-2F workarounds of Jump Instruction issue
mips64el-openbsd  -FAIL: readelf -r bintest
mips64el-openbsd  -FAIL: forward expression
mips64el-openbsd  -FAIL: assignment tests
mips64el-openbsd  -FAIL: gas/all/none
mips64el-openbsd  -XFAIL: gas/elf/missing-build-notes
mips64el-openbsd  -FAIL: macro test 2
mips64el-openbsd  -FAIL: macro irp
mips64el-openbsd  -FAIL: macro rept
mips64el-openbsd  -FAIL: nested irp/irpc/rept
mips64el-openbsd  -FAIL: macro vararg
mips64el-openbsd  -FAIL: mips jalx
mips64el-openbsd  -FAIL: ST Microelectronics Loongson-2F workarounds of Jump Instruction issue

Co-Authored-By: Maciej W. Rozycki <macro@orcam.me.uk>
binutils/
* testsuite/binutils-all/readelf.r-64: Optionally match extra
R_MIPS_NONE pairs.

gas/
* testsuite/gas/all/assign.d: Optionally match extra
R_MIPS_NONE pairs.
* testsuite/gas/all/fwdexp.d: Likewise.
* testsuite/gas/all/none.d: Likewise.
* testsuite/gas/macros/irp.d: Likewise.
* testsuite/gas/macros/repeat.d: Likewise.
* testsuite/gas/macros/rept.d: Likewise.
* testsuite/gas/macros/test2.d: Likewise.
* testsuite/gas/macros/vararg.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-1.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-2.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-3.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-4.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-5.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-6.d: Likewise.
* testsuite/gas/mips/compact-eh-el-1.d: Likewise.
* testsuite/gas/mips/compact-eh-el-2.d: Likewise.
* testsuite/gas/mips/compact-eh-el-3.d: Likewise.
* testsuite/gas/mips/compact-eh-el-4.d: Likewise.
* testsuite/gas/mips/compact-eh-el-5.d: Likewise.
* testsuite/gas/mips/compact-eh-el-6.d: Likewise.
* testsuite/gas/mips/loongson-2f-3.d: Likewise.
* testsuite/gas/mips/mips-jalx.d: Likewise.
* testsuite/gas/elf/missing-build-notes.d: Likewise.  Remove
the `xfail' tag.

ld/
* testsuite/ld-mips-elf/reloc-estimate-1.d: Optionally match
extra R_MIPS_NONE pairs.

15 months agoMIPS/testsuite: Handle 64-bit addresses
YunQiang Su [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
MIPS/testsuite: Handle 64-bit addresses

Several MIPS test cases are suitable for the n64 ABI if not for the
extra leading zeros or spaces in addresses not handled by dump patterns.
Match the characters then, removing these regressions:

mips64-openbsd  -FAIL: .set arch=FOO
mips64-openbsd  -FAIL: ST Microelectronics Loongson-2F workarounds of nop issue
mips64-openbsd  -FAIL: MIPS DSP ASE for MIPS64
mips64-openbsd  -FAIL: gas/mips/align2
mips64-openbsd  -FAIL: gas/mips/align2-el
mips64-openbsd  -FAIL: Locally-resolvable PC-relative code references
mips64-openbsd  -FAIL: MIPS jalx-1
mips64-openbsd  -FAIL: JAL overflow 2
mips64el-openbsd  -FAIL: .set arch=FOO
mips64el-openbsd  -FAIL: ST Microelectronics Loongson-2F workarounds of nop issue
mips64el-openbsd  -FAIL: MIPS DSP ASE for MIPS64
mips64el-openbsd  -FAIL: gas/mips/align2
mips64el-openbsd  -FAIL: gas/mips/align2-el
mips64el-openbsd  -FAIL: Locally-resolvable PC-relative code references
mips64el-openbsd  -FAIL: MIPS jalx-1
mips64el-openbsd  -FAIL: JAL overflow 2

Co-Authored-By: Maciej W. Rozycki <macro@orcam.me.uk>
gas/
* testsuite/gas/mips/align2-el.d: Match extra leading zeros
with addresses.
* testsuite/gas/mips/align2.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-1.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-2.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-3.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-4.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-5.d: Likewise.
* testsuite/gas/mips/compact-eh-eb-6.d: Likewise.
* testsuite/gas/mips/compact-eh-el-1.d: Likewise.
* testsuite/gas/mips/compact-eh-el-2.d: Likewise.
* testsuite/gas/mips/compact-eh-el-3.d: Likewise.
* testsuite/gas/mips/compact-eh-el-4.d: Likewise.
* testsuite/gas/mips/compact-eh-el-5.d: Likewise.
* testsuite/gas/mips/compact-eh-el-6.d: Likewise.
* testsuite/gas/mips/loongson-2f-2.d: Likewise.
* testsuite/gas/mips/loongson-2f-3.d: Likewise.
* testsuite/gas/mips/mips-jalx.d: Likewise.
* testsuite/gas/mips/mips64-dsp.d: Likewise.
* testsuite/gas/mips/pcrel-1.d: Likewise.
* testsuite/gas/mips/set-arch.d: Likewise.

ld/
* testsuite/ld-mips-elf/jaloverflow-2.d: Match extra leading
zeros and spaces with addresses as appropriate.
* testsuite/ld-mips-elf/jalx-1.d: Likewise.
* testsuite/ld-mips-elf/reloc-estimate-1.d: Likewise.

15 months agotestsuite: Also discard the `.MIPS.options' section
YunQiang Su [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
testsuite: Also discard the `.MIPS.options' section

Also discard the `.MIPS.options' section, used with n64 MIPS binaries,
along with similar other MIPS sections (`.reginfo', `.MIPS.abiflags')
not relevant for the test cases concerned, fixing these regressions:

mips64-openbsd  -FAIL: ld-elf/group3a
mips64-openbsd  -FAIL: ld-elf/group3b
mips64-openbsd  -FAIL: Place orphan sections (map file check)
mips64-openbsd  -FAIL: ld-elf/orphan-region
mips64-openbsd  -FAIL: ld-elf/orphan
mips64-openbsd  -FAIL: overlay size (map file check)
mips64-openbsd  -FAIL: overlay size
mips64el-openbsd  -FAIL: ld-elf/group3a
mips64el-openbsd  -FAIL: ld-elf/group3b
mips64el-openbsd  -FAIL: Place orphan sections (map file check)
mips64el-openbsd  -FAIL: ld-elf/orphan-region
mips64el-openbsd  -FAIL: ld-elf/orphan
mips64el-openbsd  -FAIL: overlay size (map file check)
mips64el-openbsd  -FAIL: overlay size

Co-Authored-By: Maciej W. Rozycki <macro@orcam.me.uk>
binutils/
* testsuite/binutils-all/strip-3.d: Add `-R .MIPS.options' to
the `strip' tag.

ld/
* testsuite/ld-elf/group.ld: Also discard `.MIPS.options'.
* testsuite/ld-elf/orphan-region.ld: Likewise.
* testsuite/ld-elf/orphan.ld: Likewise.
* testsuite/ld-mips-elf/got-page-1.ld: Likewise.
* testsuite/ld-scripts/overlay-size.t: Likewise.

15 months agoMIPS/LD/testsuite: Fix MIPS16 interlinking test IRIX 6 regressions
Maciej W. Rozycki [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
MIPS/LD/testsuite: Fix MIPS16 interlinking test IRIX 6 regressions

IRIX 6 does not have MIPS16 stub section support in its n32 linker
scripts, causing such input sections to be propagated to the respective
output sections rather than `.text', causing dump pattern mismatches.

Expect IRIX 6 to fail with n32 testing then, removing this regression:

mips-sgi-irix6  -FAIL: MIPS16 interlinking for local functions 1 (n32)

We may choose to update IRIX 6 n32 linker scripts sometime, as it seems
a harmless change.

ld/
* testsuite/ld-mips-elf/mips-elf.exp: Expect IRIX 6 to fail with
n32 `mips16-local-stubs-1' testing.

15 months agoMIPS/LD/testsuite: Fix MIPS16 interlinking test n64 regressions
YunQiang Su [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
MIPS/LD/testsuite: Fix MIPS16 interlinking test n64 regressions

The MIPS16 interlinking test for local functions expects to be assembled
with 32-bit addressing, otherwise causing assembly warnings:

.../ld/testsuite/ld-mips-elf/mips16-local-stubs-1.s: Assembler messages:
.../ld/testsuite/ld-mips-elf/mips16-local-stubs-1.s:16: Warning: la used to load 64-bit address; recommend using dla instead

Use the per-ABI framework then to run the test explicitly for o32 and
n32 ABIs only, replacing the `-mips4' option from the `as' tag with
`.module mips4' pseudo-op within the source itself so as to avoid
assembly errors:

Assembler messages:
Error: -mips4 conflicts with the other architecture options, which imply -mips3

with n32 testing for some targets, and ultimately removing these
regressions:

mips64-openbsd  -FAIL: MIPS16 interlinking for local functions 1
mips64el-openbsd  -FAIL: MIPS16 interlinking for local functions 1

Co-Authored-By: Maciej W. Rozycki <macro@orcam.me.uk>
ld/
* testsuite/ld-mips-elf/mips16-local-stubs-1.d: Remove `-mips4'
from the `as' tag.
* testsuite/ld-mips-elf/mips16-local-stubs-1.s: Add `.module
mips4'.
* testsuite/ld-mips-elf/mips-elf.exp: Run `mips16-local-stubs-1'
for o32 and n32 ABIs only.

15 months agoMIPS/GAS/testsuite: Force o32 for tests expecting 32-bit addressing
YunQiang Su [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
MIPS/GAS/testsuite: Force o32 for tests expecting 32-bit addressing

A few GAS tests expect to be assembled with 32-bit addressing, otherwise
causing an assembly warning:

.../gas/testsuite/gas/mips/fix-rm7000-2.s:11: Warning: la used to load 64-bit address; recommend using dla instead

or pattern dump mismatches against 32-bit address calculations, however
these tests do not enforce their expectation in any.  For none of them
the specific ABI used is of any relevance however, so select the o32 ABI
unconditionally, removing these failures with OpenBSD targets:

mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (micromips)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips3)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips4)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips5)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips64)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips64r2)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips64r3)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips64r5)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (octeon)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (octeon2)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (octeon3)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (octeonp)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (r4000)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (sb1)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (vr5400)
mips64-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (xlr)
mips64-openbsd  -FAIL: MIPS-OCTEON octeon_saa_saad (octeon2)
mips64-openbsd  -FAIL: MIPS-OCTEON octeon_saa_saad (octeon3)
mips64-openbsd  -FAIL: MIPS-OCTEON octeon_saa_saad (octeonp)
mips64-openbsd  -FAIL: Full MIPS R5900
mips64-openbsd  -FAIL: MIPS R5900 VU0
mips64-openbsd  -FAIL: Paired LL/SC for mips64r6 (mips64r6)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (micromips)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips3)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips4)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips5)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips64)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips64r2)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips64r3)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (mips64r5)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (octeon)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (octeon2)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (octeon3)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (octeonp)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (r4000)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (sb1)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (vr5400)
mips64el-openbsd  -FAIL: MIPS RM7000 workarounds test 2 (xlr)
mips64el-openbsd  -FAIL: MIPS-OCTEON octeon_saa_saad (octeon2)
mips64el-openbsd  -FAIL: MIPS-OCTEON octeon_saa_saad (octeon3)
mips64el-openbsd  -FAIL: MIPS-OCTEON octeon_saa_saad (octeonp)
mips64el-openbsd  -FAIL: Full MIPS R5900
mips64el-openbsd  -FAIL: MIPS R5900 VU0
mips64el-openbsd  -FAIL: Paired LL/SC for mips64r6 (mips64r6)

Co-Authored-By: Maciej W. Rozycki <macro@orcam.me.uk>
gas/
* testsuite/gas/mips/fix-rm7000-2.d: Add `-32' to the `as' tag.
* testsuite/gas/mips/micromips@fix-rm7000-2.d: Likewise.
* testsuite/gas/mips/r5900-full.d: Likewise.
* testsuite/gas/mips/r5900-vu0.d: Likewise.
* testsuite/gas/mips/llpscp-64.d: Add `as' tag with `-32'.
* testsuite/gas/mips/octeon-saa-saad.d: Likewise.

15 months agoMIPS/LD/testsuite: Run `got-dump-1' for o32/n32 ABIs
Maciej W. Rozycki [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
MIPS/LD/testsuite: Run `got-dump-1' for o32/n32 ABIs

The `got-dump-1' test case uses 32-bit addressing, so it makes no sense
to run it with the n64 ABI.  And there is a corresponding `got-dump-2'
test already for the n64 ABI.

Use the per-ABI framework then to run the `got-dump-1' test explicitly
for o32 and n32 ABIs only.

ld/
* testsuite/ld-mips-elf/mips-elf.exp: Run `got-dump-1' for o32
and n32 ABIs only.

15 months agoMIPS/LD/testsuite: Fix `attr-gnu-4-10' failures with OpenBSD targets
Maciej W. Rozycki [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
MIPS/LD/testsuite: Fix `attr-gnu-4-10' failures with OpenBSD targets

OpenBSD targets produce ELF64 files while the pattern dump expects ELF32
output and specific header sizes.  Disable it for `mips64*-*-openbsd*'
for these targets then, removing these failures:

mips64-openbsd  -FAIL: ld-mips-elf/attr-gnu-4-10
mips64el-openbsd  -FAIL: ld-mips-elf/attr-gnu-4-10

ld/
* testsuite/ld-mips-elf/attr-gnu-4-10.d: Add `notarget' tag with
`mips64*-*-openbsd*'.

15 months agoMIPS/LD/testsuite: Fix `attr-gnu-4-10' failures with IRIX targets
Maciej W. Rozycki [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
MIPS/LD/testsuite: Fix `attr-gnu-4-10' failures with IRIX targets

IRIX targets do not enable the production of a `.pdr' section in GAS by
default, which causes a failure with the `attr-gnu-4-10' test case due
to a difference resulting in the number and indices of sections produced
in linker output.

As the presence or absence of this section is not relevant to this test
case, just enable it unconditionally, fixing these regressions:

mips-sgi-irix5  -FAIL: ld-mips-elf/attr-gnu-4-10
mips-sgi-irix6  -FAIL: ld-mips-elf/attr-gnu-4-10

ld/
* testsuite/ld-mips-elf/attr-gnu-4-10.d: Add `as' tag with
`-mpdr'.

15 months agoMIPS/LD/testsuite: Fix JALR relaxation test failure with IRIX 6
Maciej W. Rozycki [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
MIPS/LD/testsuite: Fix JALR relaxation test failure with IRIX 6

The `mips-sgi-irix6' target only supports IRIX linker emulations, but
most JALR relaxation tests request the relevant traditional emulation
instead, causing a link failure:

./ld-new: unrecognised emulation mode: elf32btsmipn32
Supported emulations: elf32bmipn32 elf32bsmip elf64bmip

This is clearly an omission from the conversion to use the per-ABI
framework made with commit 78da84f99405 ("MIPS/LD/testsuite: Correct
mips-elf.exp test ABI/emul/endian arrangement").  These tests are also
endianness agnostic, which was missed in the conversion as well.

Remove the unnecessary explicit ABI and endianness options then and rely
on the per-ABI framework to get things right, removing this regression:

mips-sgi-irix6  -FAIL: MIPS relax-jalr-shared n32

ld/
* testsuite/ld-mips-elf/relax-jalr-n32-shared.d: Remove flags
related to ABI and endianness selection from the `as' and `ld'
tags.
* testsuite/ld-mips-elf/relax-jalr-n64.d: Likewise.
* testsuite/ld-mips-elf/relax-jalr-n64-shared.d: Likewise.
* testsuite/ld-mips-elf/mips-elf.exp: Remove `as' and `ld' tag
additions from the invocation of JALR relaxation tests.

15 months agoMIPS/LD/testsuite: Fix unaligned JALX failures with OpenBSD targets
Maciej W. Rozycki [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
MIPS/LD/testsuite: Fix unaligned JALX failures with OpenBSD targets

There are only n64 linker emulations included with `mips64*-*-openbsd*'
targets, however the unaligned JALX tests insist on running across all
targets and force the n32 ABI, causing link errors with the targets
concerned, e.g.:

./ld-new: tmpdir/unaligned-jalx-0.o: ABI is incompatible with that of the selected emulation
./ld-new: failed to merge target specific data of file tmpdir/unaligned-jalx-0.o
./ld-new: tmpdir/unaligned-insn.o: ABI is incompatible with that of the selected emulation
./ld-new: failed to merge target specific data of file tmpdir/unaligned-insn.o

Convert the tests then to use the per-ABI framework and run them for the
o32 and n32 ABIs, removing these regressions:

mips64-openbsd  -FAIL: MIPS JALX to unaligned symbol 0
mips64-openbsd  -FAIL: MIPS JALX to unaligned symbol 1
mips64-openbsd  -FAIL: MIPS JALX to unaligned symbol 2
mips64-openbsd  -FAIL: MIPS JALX to unaligned symbol 3
mips64-openbsd  -FAIL: MIPS16 JALX to unaligned symbol 0
mips64-openbsd  -FAIL: MIPS16 JALX to unaligned symbol 1
mips64-openbsd  -FAIL: microMIPS JALX to unaligned symbol 0
mips64-openbsd  -FAIL: microMIPS JALX to unaligned symbol 1
mips64el-openbsd  -FAIL: MIPS JALX to unaligned symbol 0
mips64el-openbsd  -FAIL: MIPS JALX to unaligned symbol 1
mips64el-openbsd  -FAIL: MIPS JALX to unaligned symbol 2
mips64el-openbsd  -FAIL: MIPS JALX to unaligned symbol 3
mips64el-openbsd  -FAIL: MIPS16 JALX to unaligned symbol 0
mips64el-openbsd  -FAIL: MIPS16 JALX to unaligned symbol 1
mips64el-openbsd  -FAIL: microMIPS JALX to unaligned symbol 0
mips64el-openbsd  -FAIL: microMIPS JALX to unaligned symbol 1

Similar tests for the n64 ABI can be added separately, using suitable
dump patterns.

ld/
* testsuite/ld-mips-elf/unaligned-jalx-0.d: Remove `-32' from
the `as' tag.
* testsuite/ld-mips-elf/unaligned-jalx-1.d: Likewise.
* testsuite/ld-mips-elf/unaligned-jalx-2.d: Likewise.
* testsuite/ld-mips-elf/unaligned-jalx-3.d: Likewise.
* testsuite/ld-mips-elf/unaligned-jalx-mips16-0.d: Likewise.
* testsuite/ld-mips-elf/unaligned-jalx-mips16-1.d: Likewise.
* testsuite/ld-mips-elf/unaligned-jalx-micromips-0.d: Likewise.
* testsuite/ld-mips-elf/unaligned-jalx-micromips-1.d: Likewise.
* testsuite/ld-mips-elf/mips-elf.exp: Run unaligned JALX tests
with `run_dump_test_o32' and `run_dump_test_n32' rather than
`run_dump_test'.

15 months agoMIPS/GAS/testsuite: Disable compact EH #7 tests with OpenBSD targets
Maciej W. Rozycki [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
MIPS/GAS/testsuite: Disable compact EH #7 tests with OpenBSD targets

Compact EH #7 tests use output templates that are not suitable for the
n64 ABI, which `mips64*-*-openbsd*' targets use by default, because the
contents of the sections examined are expected to be differnt.  Disable
the tests then, removing these regressions:

mips64-openbsd  -FAIL: Compact EH EB #7 with personality id and fallback FDE
mips64-openbsd  -FAIL: Compact EH EL #7 with personality id and fallback FDE
mips64el-openbsd  -FAIL: Compact EH EB #7 with personality id and fallback FDE
mips64el-openbsd  -FAIL: Compact EH EL #7 with personality id and fallback FDE

Suitable corresponding tests for the n64 ABI can be added separately.

gas/
* testsuite/gas/mips/compact-eh-eb-7.d: Exclude for
`mips64*-*-openbsd*'.
* testsuite/gas/mips/compact-eh-el-7.d: Likewise.

15 months agoMIPS/LD: Include n64 `.interp' with INITIAL_READONLY_SECTIONS
YunQiang Su [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
MIPS/LD: Include n64 `.interp' with INITIAL_READONLY_SECTIONS

In ld/emulparams/elf64bmip-defs.sh there is no explicit handling of the
`.interp' section, which causes it to be positioned in output at an odd
place.

Let's include it with INITIAL_READONLY_SECTIONS, just like o32/n32 do,
fixing a regression from commit 5a8e7be242f3 ("INITIAL_READONLY_SECTIONS
in elf.sc"), where the handling of n64 was missed due to an unfortunate
sequence of events where ld/emulparams/elf64bmip-defs.sh was only added
with commit 94bb04b3c611 ("Use .reginfo rather than .MIPS.options in n32
linker scripts") the day before.

Add test cases covering section ordering across the three ABIs.  This
change also fixes ld/pr23658-2:

FAIL: Build pr23658-2

Co-Authored-By: Maciej W. Rozycki <macro@orcam.me.uk>
ld/ChangeLog:
* emulparams/elf64bmip-defs.sh: Include `.interp' with
INITIAL_READONLY_SECTIONS.
* testsuite/ld-mips-elf/pie-n64.d: Adjust addresses.
* testsuite/ld-mips-elf/sections-1-o32.rd: New test.
* testsuite/ld-mips-elf/sections-1-o32t.rd: New test.
* testsuite/ld-mips-elf/sections-1-n32.rd: New test.
* testsuite/ld-mips-elf/sections-1-n32t.rd: New test.
* testsuite/ld-mips-elf/sections-1-n32p.rd: New test.
* testsuite/ld-mips-elf/sections-1-n64.rd: New test.
* testsuite/ld-mips-elf/sections-1-n64t.rd: New test.
* testsuite/ld-mips-elf/sections-2-o32.rd: New test.
* testsuite/ld-mips-elf/sections-2-o32t.rd: New test.
* testsuite/ld-mips-elf/sections-2-n32.rd: New test.
* testsuite/ld-mips-elf/sections-2-n32t.rd: New test.
* testsuite/ld-mips-elf/sections-2-n32p.rd: New test.
* testsuite/ld-mips-elf/sections-2-n64.rd: New test.
* testsuite/ld-mips-elf/sections-2-n64t.rd: New test.
* testsuite/ld-mips-elf/sections.s: New test source.
* testsuite/ld-mips-elf/mips-elf.exp: Run the new tests.

15 months agoRevert "MIPS: support mips*64 as CPU and gnuabi64 as ABI"
Maciej W. Rozycki [Fri, 28 Jul 2023 05:00:36 +0000 (06:00 +0100)]
Revert "MIPS: support mips*64 as CPU and gnuabi64 as ABI"

This reverts commit 32f1c80375ebe8ad25d9805ee5889f0006c51e59.  It had
two unrelated changes lumped together, one of which changed the meaning
of the `mipsisa64*-*-linux*' target triplets, which was not properly
evaluated.

15 months agoldscripts/empty-address vs. xcoff
Alan Modra [Fri, 28 Jul 2023 03:36:17 +0000 (13:06 +0930)]
ldscripts/empty-address vs. xcoff

The empty-address tests check that if a section is removed by ld due
to being empty then properties of that section don't affect following
addresses.  The xcoff backend doesn't remove the empty .data section
created by empty-address-2* and empty-address-3* for some reason, and
therefore fails the test.

* testsuite/ld-scripts/empty-address-1.d: Accept more symbols.
* testsuite/ld-scripts/empty-address-2a.d: xfail for xcoff.
* testsuite/ld-scripts/empty-address-2b.d: Likewise.
* testsuite/ld-scripts/empty-address-3a.d: Likewise.
* testsuite/ld-scripts/empty-address-3b.d: Likewise.

15 months agoFix recent x86 pe/coff testsuite regressions
Alan Modra [Fri, 28 Jul 2023 03:33:46 +0000 (13:03 +0930)]
Fix recent x86 pe/coff testsuite regressions

* testsuite/gas/i386/sha512-intel.d: Accept section nop padding.
* testsuite/gas/i386/sha512.d: Likewise.
* testsuite/gas/i386/sm3-intel.d: Likewise.
* testsuite/gas/i386/sm3.d: Likewise.
* testsuite/gas/i386/x86-64-pbndkb-intel.d: Likewise.
* testsuite/gas/i386/x86-64-pbndkb.d: Likewise.
* testsuite/gas/i386/x86-64-sha512-intel.d: Likewise.
* testsuite/gas/i386/x86-64-sha512.d: Likewise.
* testsuite/gas/i386/x86-64-sm3-intel.d: Likewise.
* testsuite/gas/i386/x86-64-sm3.d: Likewise.

15 months agocoff/pe/xcoff and --extract-symbols
Alan Modra [Fri, 28 Jul 2023 03:14:17 +0000 (12:44 +0930)]
coff/pe/xcoff and --extract-symbols

This fixes failure of the "extract symbols" test for rs6000, where
--extract-symbols generates a non-zero sized .text.  By the look of
coffcode.h the same problem might occur for coff/pe too, but doesn't
happen to trigger a test failure.

bfd/
* coffcode.h (coff_compute_section_file_positions): Don't
adjust size of !SEC_LOAD sections.
binutils/
* objcopy.c (setup_section): Clear SEC_LOAD for --extract-symbol.

15 months agoRISC-V: Add actual 'Zvkt' extension support
Tsukasa OI [Wed, 26 Jul 2023 01:39:44 +0000 (01:39 +0000)]
RISC-V: Add actual 'Zvkt' extension support

The 'Zvkt' extension is listed on the added extensions in the GNU Binutils
version 2.41 (see binutils/NEWS).  However, the support of this extension
was actually missing.

This commit adds actual support of this extension and adds implications
from 'Zvkn' and 'Zvks' superset extensions.

bfd/ChangeLog:

* elfxx-riscv.c (riscv_implicit_subsets) Add implications from
'Zvkn' and 'Zvks'.  (riscv_supported_std_z_ext): Add 'Zvkt' to
the supported extension list.

15 months agoFix typo in riscv-dis.c comment
Tsukasa OI [Fri, 28 Jul 2023 03:33:39 +0000 (03:33 +0000)]
Fix typo in riscv-dis.c comment

Don't go "past" the start of the section.

15 months agoAutomatic date update in version.in
GDB Administrator [Fri, 28 Jul 2023 00:00:19 +0000 (00:00 +0000)]
Automatic date update in version.in

15 months agogdb: remove trailing empty line in target-delegates.c
Simon Marchi [Thu, 27 Jul 2023 17:32:38 +0000 (13:32 -0400)]
gdb: remove trailing empty line in target-delegates.c

In a review [1], I pointed out that applying the patch, git would say:

    .git/rebase-apply/patch:147: new blank line at EOF.

However, since the empty line is in target-delegates.c (a generated
file), there's nothing the author can do about it.  To avoid this
comment coming up again in the future, change make-target-delegates.py
to avoid the trailing empty line.  Do this by making it output empty
lines before each entity, not after.

Since this needs removing a newline output in gdbcopyright, adjust
ada-unicode.py and gdbarch.py to avoid changes in the files they
generate.

[1] https://inbox.sourceware.org/gdb-patches/20230427210113.45380-1-jhb@FreeBSD.org/T/#m083598405bef19157f67c9d97846d3dd90dc7d1c

Change-Id: Ic4c648f06443b432168cb76603402c918aa6e5d2
Approved-By: Tom Tromey <tom@tromey.com>
15 months agoReport supportsBreakpointLocationsRequest
Tom Tromey [Thu, 27 Jul 2023 16:31:32 +0000 (10:31 -0600)]
Report supportsBreakpointLocationsRequest

While looking at the DAP spec, I noticed that the breakpointLocations
request is gated behind a capability.  This patch changes gdb to
report this capability.

I've also added a comment to explain the fact that arguments to
breakpointLocations are not optional, even though the spec says they
are.

15 months ago/DISCARD/ in ld testsuite
Alan Modra [Thu, 27 Jul 2023 08:33:49 +0000 (18:03 +0930)]
/DISCARD/ in ld testsuite

The canonical form to discard all sections not mentioned earlier in
the script is
  /DISCARD/ : { *(*) }
not
  /DISCARD/ : { *(.*) }
".*" happens to work with the usual section names starting with a dot,
but let's not promote something not quite right.

15 months agosh: uninitialised sh_operand_info.type in get_specific
Alan Modra [Wed, 26 Jul 2023 23:43:56 +0000 (09:13 +0930)]
sh: uninitialised sh_operand_info.type in get_specific

Seen when running gas/testsuite/gas/sh/err-at.s

* config/tc-sh.c (get_operands): Always init operand type.
* testsuite/gas/sh/err-at.s: Expect unnecessary extra errors.

15 months agoSupport Intel PBNDKB
Hu, Lin1 [Mon, 24 Jul 2023 03:10:03 +0000 (11:10 +0800)]
Support Intel PBNDKB

gas/ChangeLog:

* NEWS: Support Intel PBNDKB.
* config/tc-i386.c: Add pbndkb.
* doc/c-i386.texi: Document .pbndkb.
* testsuite/gas/i386/i386.exp: Add PBNDKB tests.
* testsuite/gas/i386/x86-64.exp: Ditto.
* testsuite/gas/i386/pbndkb-inval.l: New test.
* testsuite/gas/i386/pbndkb-inval.s: Ditto.
* testsuite/gas/i386/x86-64-pbndkb-intel.d: Ditto.
* testsuite/gas/i386/x86-64-pbndkb.d: Ditto.
* testsuite/gas/i386/x86-64-pbndkb.s: Ditto.

opcodes/ChangeLog:

* i386-dis.c (X86_64_0F01_REG_0_MOD_3_RM_7): New.
(X86_64_0F01_REG_0_MOD_3_RM_7_P_0): Ditto.
(prefix_table): Add PREFIX_0F01_REG_0_MOD_3_RM_7.
(x86_64_table): Add X86_64_0F01_REG_0_MOD_3_RM_7_P_0.
(rm_table): New entry for pbndkb.
* i386-gen.c (cpu_flag): Add PBNDKB.
* i386-init.h: Regenerated.
* i386-mnem.h: Ditto.
* i386-opc.h (CpuPBNDKB): New.
(i386_cpu_flags): Add cpupbndkb.
* i386-opc.tbl: Add PBNDKB instructions.
* i386-tbl.h: Regenerated.

15 months agoSupport Intel SM4
Haochen Jiang [Mon, 24 Jul 2023 03:09:57 +0000 (11:09 +0800)]
Support Intel SM4

gas/ChangeLog:

* NEWS: Support Intel SM4.
* config/tc-i386.c: Add sm4.
* doc/c-i386.texi: Document .sm4.
* testsuite/gas/i386/i386.exp: Run SM4 tests.
* testsuite/gas/i386/x86-64.exp: Ditto.
* testsuite/gas/i386/sm4-intel.d: Add SM4 tests.
* testsuite/gas/i386/sm4.d: Ditto.
* testsuite/gas/i386/sm4.s: Ditto.
* testsuite/gas/i386/x86-64-sm4-intel.d: Ditto.
* testsuite/gas/i386/x86-64-sm4.d: Ditto.
* testsuite/gas/i386/x86-64-sm4.s: Ditto.

opcodes/ChangeLog:

* i386-dis.c (prefix_table): Add SM4 instructions.
* i386-gen.c (isa_dependencies): Add SM4.
(cpu_flags): Ditto.
* i386-init.h: Regenerated.
* i386-mnem.h: Ditto.
* i386-opc.h (CpuSM4): New.
(i386_cpu_flags): Add cpusm4.
* i386-opc.tbl: Add SM4 instructions.
* i386-tbl.h: Regenerated.

15 months agoSupport Intel SM3
Haochen Jiang [Mon, 24 Jul 2023 03:09:49 +0000 (11:09 +0800)]
Support Intel SM3

gas/ChangeLog:

* NEWS: Support Intel SM3.
* config/tc-i386.c: Add sm3.
* doc/c-i386.texi: Document .sm3.
* testsuite/gas/i386/i386.exp: Run sm3 tests.
* testsuite/gas/i386/x86-64.exp: Ditto.
* testsuite/gas/i386/sm3-intel.d: New test.
* testsuite/gas/i386/sm3.d: Ditto.
* testsuite/gas/i386/sm3.s: Ditto.
* testsuite/gas/i386/x86-64-sm3-intel.d: Ditto.
* testsuite/gas/i386/x86-64-sm3.d: Ditto.
* testsuite/gas/i386/x86-64-sm3.s: Ditto.

opcodes/ChangeLog:

* i386-dis.c (PREFIX_VEX_0F38DA_W_0): New.
(VEX_LEN_0F38DA_W_0_P_0): Ditto.
(VEX_LEN_0F38DA_W_0_P_2): Ditto.
(VEX_LEN_0F3ADE_W_0): Ditto.
(VEX_W_0F38DA): Ditto.
(VEX_W_0F3ADE): Ditto.
(prefix_table): Add PREFIX_VEX_0F38DA_W_0.
(vex_len_table): Add VEX_LEN_0F38DA_W_0_P_0,
VEX_LEN_0F38DA_W_0_P_2, VEX_LEN_0F3ADE_W_0.
(vex_w_table): Add VEX_W_0F38DA, VEX_W_0F3ADE.
* i386-gen.c (isa_dependencies): Add SM3.
(cpu_flags): Ditto.
* i386-init.h: Regenerated.
* i386-mnem.h: Ditto.
* i386-opc.h (CpuSM3): New.
(i386_cpu_flags): Add cpusm3.
* i386-opc.tbl: Add SM3 instructions.
* i386-tbl.h: Regenerated.

15 months agoSupport Intel SHA512
Haochen Jiang [Mon, 24 Jul 2023 03:09:43 +0000 (11:09 +0800)]
Support Intel SHA512

gas/ChangeLog:

* NEWS: Support Intel SHA512.
* config/tc-i386.c: Add sha512.
* doc/c-i386.texi: Document .sha512.
* testsuite/gas/i386/disassem.d: Add SHA512 tests.
* testsuite/gas/i386/disassem.s: Ditto.
* testsuite/gas/i386/i386.exp: Run SHA512 tests.
* testsuite/gas/i386/x86-64.exp: Ditto.
* testsuite/gas/i386/sha512-intel.d: New test.
* testsuite/gas/i386/sha512-inval.l: Ditto.
* testsuite/gas/i386/sha512-inval.s: Ditto.
* testsuite/gas/i386/sha512.d: Ditto.
* testsuite/gas/i386/sha512.s: Ditto.
* testsuite/gas/i386/x86-64-sha512-intel.d: Ditto.
* testsuite/gas/i386/x86-64-sha512-inval.l: Ditto.
* testsuite/gas/i386/x86-64-sha512-inval.s: Ditto.
* testsuite/gas/i386/x86-64-sha512.d: Ditto.
* testsuite/gas/i386/x86-64-sha512.s: Ditto.

opcodes/ChangeLog:

* i386-dis.c (Rxmmq): New.
(Rymm): Ditto.
(PREFIX_VEX_0F38CB): Ditto.
(PREFIX_VEX_0F38CC): Ditto.
(PREFIX_VEX_0F38CD): Ditto.
(VEX_LEN_0F38CB_P_3_W_0): Ditto.
(VEX_LEN_0F38CC_P_3_W_0): Ditto.
(VEX_LEN_0F38CD_P_3_W_0): Ditto.
(VEX_W_0F38CB_P_3): Ditto.
(VEX_W_0F38CC_P_3): Ditto.
(VEX_W_0F38CD_P_3): Ditto.
(prefix_table): Add PREFIX_VEX_0F38CB, PREFIX_VEX_0F38CC,
PREFIX_VEX_0F38CD.
(vex_len_table): Add VEX_LEN_0F38CB_P_3_W_0,
VEX_LEN_0F38CC_P_3_W_0, VEX_LEN_0F38CD_P_3_W_0.
(vex_w_table): Add VEX_W_0F38CB_P_3, VEX_W_0F38CC_P_3, VEX_W_0F38CD_P_3.
* i386-gen.c (isa_dependencies): Add SHA512.
(cpu_flags): Ditto.
* i386-init.h: Regenerated.
* i386-mnem.h: Ditto.
* i386-opc.h (CpuSHA512): New.
(i386_cpu_flags): Add cpusha512.
* i386-opc.tbl: Add SHA512 instructions.
* i386-tbl.h: Regenerated.

15 months agoSupport Intel AVX-VNNI-INT16
konglin1 [Mon, 24 Jul 2023 03:09:35 +0000 (11:09 +0800)]
Support Intel AVX-VNNI-INT16

gas/ChangeLog:

* NEWS: Support Intel AVX-VNNI-INT16.
* config/tc-i386.c: Add avx_vnni_int16.
* doc/c-i386.texi: Document avx_vnni_int16.
* testsuite/gas/i386/i386.exp: Run AVX VNNI INT16 tests.
* testsuite/gas/i386/x86-64.exp: Ditto.
* testsuite/gas/i386/avx-vnni-int16-intel.d: New test.
* testsuite/gas/i386/avx-vnni-int16.d: New test.
* testsuite/gas/i386/avx-vnni-int16.s: New test.
* testsuite/gas/i386/x86-64-avx-vnni-int16-intel.d: New test.
* testsuite/gas/i386/x86-64-avx-vnni-int16.d: New test.
* testsuite/gas/i386/x86-64-avx-vnni-int16.s: New test.

opcodes/ChangeLog:

* i386-dis.c (PREFIX_VEX_0F38D2_W_0): New.
(PREFIX_VEX_0F38D3_W_0): Ditto.
(VEX_W_0F38D2_P_0): Ditto.
(VEX_W_0F38D2_P_1): Ditto.
(VEX_W_0F38D2_P_2): Ditto.
(VEX_W_0F38D3_P_0): Ditto.
(VEX_W_0F38D3_P_1): Ditto.
(VEX_W_0F38D3_P_2): Ditto.
(prefix_table): Add PREFIX_VEX_0F38D2_W_0 and
PREFIX_VEX_0F38D3_W_0.
(vex_table): Add VEX_W_0F38D2 and VEX_W_0F38D3.
(vex_w_table): Ditto.
* i386-gen.c (isa_dependencies): Add AVX_VNNI_INT16.
(cpu_flag): Ditto.
* i386-init.h: Regenerated.
* i386-mnem.h: Ditto.
* i386-opc.h: (CpuAVX_VNNI_INT16): New.
* i386-opc.tbl: Add Intel AVX_VNNI_INT16 instructions.
* i386-tbl.h: Regenerated.

15 months agoAutomatic date update in version.in
GDB Administrator [Thu, 27 Jul 2023 00:00:41 +0000 (00:00 +0000)]
Automatic date update in version.in

15 months ago[gdb/testsuite] Fix gdb.python/py-thread-exited.exp
Tom de Vries [Wed, 26 Jul 2023 19:40:01 +0000 (21:40 +0200)]
[gdb/testsuite] Fix gdb.python/py-thread-exited.exp

Two fixes in gdb.python/py-thread-exited.exp:
- fix the copyright notice validity range (PR testsuite/30687):
  2022-202 -> 2022-2023, and
- add missing "require allow_python_tests".

Tested on x86_64-linux.

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30687

15 months agobpf: accept # as an inline comment char
David Faust [Mon, 24 Jul 2023 21:57:35 +0000 (14:57 -0700)]
bpf: accept # as an inline comment char

This little patch makes the BPF assembler accept '#' as an inline
comment character, which clang -S seems to use.

gas/
* config/tc-bpf.c (comment_chars): Add '#'.
* doc/c-bpf.texi (BPF Special Characters): Add note that '#' may
be used for inline comments.

15 months ago[gdb/build] Fix Wstringop-truncation in coff_getfilename
Tom de Vries [Wed, 26 Jul 2023 15:06:23 +0000 (17:06 +0200)]
[gdb/build] Fix Wstringop-truncation in coff_getfilename

When building gdb with -O2 -fsanitize-threads, I ran into
a Werror=stringop-truncation.

The problem is here in coff_getfilename in coffread.c:
...
      strncpy (buffer, aux_entry->x_file.x_n.x_fname, FILNMLEN);
      buffer[FILNMLEN] = '\0';
...

The constant FILNMLEN is expected to designate the size of
aux_entry->x_file.x_n.x_fname, but that's no longer the case since commit
60ebc257517 ("Fixes a buffer overflow when compiling assembler for the MinGW
targets.").

Fix this by using "sizeof (aux_entry->x_file.x_n.x_fname)" instead.

Likewise in xcoffread.c.

Tested on x86_64-linux.

Approved-By: Tom Tromey <tom@tromey.com>
PR build/30669
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30669

15 months agobpf: gas: add negi and neg32i tests
Jose E. Marchesi [Wed, 26 Jul 2023 13:02:08 +0000 (15:02 +0200)]
bpf: gas: add negi and neg32i tests

gas/ChangeLog:

2023-07-26  Jose E. Marchesi  <jose.marchesi@oracle.com>

* testsuite/gas/bpf/alu.s: Add test for NEGI and NEG32I.
* testsuite/gas/bpf/alu32.s: Likewise.
* testsuite/gas/bpf/alu-pseudoc.s: Likewise.
* testsuite/gas/bpf/alu32-pseudoc.s: Likewise.
* testsuite/gas/bpf/alu.d: Add expected results.
* testsuite/gas/bpf/alu-be.d: Likewise.
* testsuite/gas/bpf/alu-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu-be-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu32.d: Likewise.
* testsuite/gas/bpf/alu32-be.d: Likewise.
* testsuite/gas/bpf/alu32-be-pseudoc.d: Likewise.

15 months ago[gdb/testsuite] Drop -nostdlib in gdb.dwarf2/typeddwarf.exp
Tom de Vries [Wed, 26 Jul 2023 11:42:32 +0000 (13:42 +0200)]
[gdb/testsuite] Drop -nostdlib in gdb.dwarf2/typeddwarf.exp

As reported in PR testsuite/30633, when running test-case
gdb.dwarf2/typeddwarf.exp with target board native-gdbserver on Ubuntu
22.04.2, we run into:
...
(gdb) continue^M
Continuing.^M
^M
Program received signal SIGSEGV, Segmentation fault.^M
0x0000000000000001 in ?? ()^M
(gdb) FAIL: gdb.dwarf2/typeddwarf.exp: runto: run to main
...

We run into the FAIL as follows:
- due to using gdbserver, we attach at the point of the first instruction, in
  _start
- we then set a breakpoint at main
- the test-case is a .s file, that has main renamed to _start in the assembly,
  but not in the debuginfo
- setting a breakpoint at main sets the breakpoint at the same instruction
  we're currently stopped at
- continue doesn't hit the breakpoint, and we return out of _start, which
  causes a sigsegv

Note that this is for the amd64 case (using gdb.dwarf2/typeddwarf-amd64.S).
For the i386 case (using gdb.dwarf2/typeddwarf.S), setting a breakpoint in
main sets it one insn after function entry, and consequently the problem does
not occur.

The FAIL is a regression since commit 90cce6c0551 ("[gdb/testsuite] Add nopie
in a few test-cases").

Without nopie the executable is PIE, with nopie it's static instead.

In the PIE case, we attach at the point of _start in the dynamic linker, and
consequently we do not skip the breakpoint in main, and also don't run into
the FAIL.

Fix this by:
- removing the -nostdlib setting, and
- renaming _start to main in both .S files.

The change to use -nostdlib and rename main to _start was originally added
in commit 6edba76fe8b (submitted here:
https://sourceware.org/pipermail/gdb-patches/2011-May/082657.html ) , I assume
to fix the problem now fixed by using nopie.

Tested on x86_64-linux.

Reported-By: Simon Marchi <simon.marchi@efficios.com>
Tested-By: Simon Marchi <simon.marchi@efficios.com>
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30633

15 months ago[gdb/tui] Fix secondary prompt
Tom de Vries [Wed, 26 Jul 2023 11:31:53 +0000 (13:31 +0200)]
[gdb/tui] Fix secondary prompt

With CLI, a session defining a command looks like:
...
(gdb) define foo
Type commands for definition of "foo".
End with a line saying just "end".
>bar
>end
(gdb)
...

With TUI however, we get the same secondary prompts, and type the same, but
are left with:
...
(gdb) define foo
Type commands for definition of "foo".
End with a line saying just "end".
(gdb)
...

Fix this by calling tui_inject_newline_into_command_window in
gdb_readline_wrapper_line, as is done in tui_command_line_handler.

Tested on x86_64-linux.

PR tui/30636
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30636

15 months ago[gdb/testsuite] Fix gdb.gdb/python-helper.exp with -O2 -flto=auto and gcc 7.5.0 some...
Tom de Vries [Wed, 26 Jul 2023 11:28:15 +0000 (13:28 +0200)]
[gdb/testsuite] Fix gdb.gdb/python-helper.exp with -O2 -flto=auto and gcc 7.5.0 some more

With a gdb build with -O2 -flto=auto and gcc 7.5.0 and test-case
gdb.gdb/python-helper.exp I run into:
...
(outer-gdb) continue^M
Continuing.^M
print 1^M
^M
Thread 1 "xgdb" hit Breakpoint 2, \
  _Z11value_printP5valueP7ui_filePK19value_print_options (val=0x22e2590, \
  stream=0x1f65480, options=0x7fffffffcdc0) at gdb/valprint.c:1193^M
1193    {^M
(outer-gdb) FAIL: gdb.gdb/python-helper.exp: hit breakpoint in outer gdb
...

This is the "value_print" variant of the problem with "c_print_type" I fixed
in commit 0d332f11122 ("[gdb/testsuite] Fix gdb.gdb/python-helper.exp with -O2
 -flto=auto and gcc 7.5.0").

Fix this likewise.

Tested on x86_64-linux.

15 months ago[gdb/tui] Fix assert in ~gdbpy_tui_window_maker
Tom de Vries [Wed, 26 Jul 2023 10:29:28 +0000 (12:29 +0200)]
[gdb/tui] Fix assert in ~gdbpy_tui_window_maker

In gdb/tui/tui-layout.c, we have:
...
static window_types_map known_window_types;
...
and in gdb/python/py-tui.c:
...
  /* A global list of all gdbpy_tui_window_maker objects.  */
  static intrusive_list<gdbpy_tui_window_maker> m_window_maker_list;
};

/* See comment in class declaration above.  */

intrusive_list<gdbpy_tui_window_maker>
  gdbpy_tui_window_maker::m_window_maker_list;
...

With a gdb build with -O0 or -O2, the static destructor calling order seems to be:
- first gdb/tui/tui-layout.c,
- then gdb/python/py-tui.c.

So when running test-case gdb.python/tui-window-factory.exp, we see the
following order of events:
- the destructor for known_window_types is called, which triggers calling the
  destructor for the only element E of m_window_maker_list.  The destructor
  destroys E, and also removes E from m_window_maker_list, leaving it empty.
- the destructor for m_window_maker_list is called.  It's empty, so it's a nop.

However, when building gdb with -O2 -flto=auto, the static destructor calling
order seems to be reversed.

Instead, we have these events:
- the destructor for m_window_maker_list is called.  This doesn't destroy it's
  only element E, but it does make m_window_maker_list empty.
- the destructor for known_window_types is called, which triggers calling the
  destructor for E.  An attempt is done to remove E from m_window_maker_list,
  but we run into an assertion failure, because the list is empty.

Fix this by checking is_linked () before attempting to remove from
m_window_maker_list, similar to how things were addressed in commit 995a34b1772
("Guard against frame.c destructors running before frame-info.c's").

Tested on x86_64-linux.

PR tui/30646
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30646

15 months ago[gdb/testsuite] Fix regexps in gdb.base/step-over-syscall.exp
Tom de Vries [Wed, 26 Jul 2023 09:53:31 +0000 (11:53 +0200)]
[gdb/testsuite] Fix regexps in gdb.base/step-over-syscall.exp

When running test-case gdb.base/step-over-syscall.exp without glibc debuginfo
installed, I get:
...
(gdb) continue^M
Continuing.^M
^M
Breakpoint 2, 0x00007ffff7d4405e in vfork () from /lib64/libc.so.6^M
(gdb) PASS: gdb.base/step-over-syscall.exp: vfork: displaced=off: \
  continue to vfork (1st time)
...
but with glibc debuginfo installed I get instead:
...
(gdb) continue^M
Continuing.^M
^M
Breakpoint 2, 0x00007ffff7d4405e in __libc_vfork () at \
  ../sysdeps/unix/sysv/linux/x86_64/vfork.S:44^M
44      ENTRY (__vfork)^M
(gdb) FAIL: gdb.base/step-over-syscall.exp: vfork: displaced=off: \
  continue to vfork (1st time)
...

The FAIL is due to a mismatch with regexp:
...
  "Breakpoint \[0-9\]+, (.* in |__libc_|)$syscall \\(\\).*"
...
because it cannot match both ".* in " and the __libc_ prefix.

Fix this by using instead the regexp:
...
  "Breakpoint \[0-9\]+, (.* in )?(__libc_)?$syscall \\(\\).*"
...

Tested on x86_64-linux.

15 months agobpf: fix neg and neg32 BPF instructions in simulator
Jose E. Marchesi [Wed, 26 Jul 2023 09:44:20 +0000 (11:44 +0200)]
bpf: fix neg and neg32 BPF instructions in simulator

This patch fixes the semantics of the neg and neg32 BPF instructions
in the simulator, and also updates the corresponding tests
accordingly.

Tested in target bpf-unknown-none.

15 months agobpf: fix register NEG[32] instructions
Jose E. Marchesi [Wed, 26 Jul 2023 09:38:04 +0000 (11:38 +0200)]
bpf: fix register NEG[32] instructions

This patch fixes the BPF_INSN_NEGR and BPF_INSN_NEG32R BPF
instructions to not use their source registers.

Tested in bpf-unknown-none.

opcodes/ChangeLog:

2023-07-26  Jose E. Marchesi  <jose.marchesi@oracle.com>

* bpf-opc.c (bpf_opcodes): Fix BPF_INSN_NEGR to not use a src
register.

gas/ChangeLog:

2023-07-26  Jose E. Marchesi  <jose.marchesi@oracle.com>

* testsuite/gas/bpf/alu.s: The register neg instruction gets only
one argument.
* testsuite/gas/bpf/alu32-be-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu32-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu32-pseudoc.s: Likewise.
* testsuite/gas/bpf/alu-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu-be-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu-pseudoc.s: Likewise.
* testsuite/gas/bpf/alu-be.d: Likewise.
* testsuite/gas/bpf/alu.d: Likewise.
* testsuite/gas/bpf/alu32-be.d: Likewise.
* testsuite/gas/bpf/alu32.d: Likewise.
* testsuite/gas/bpf/alu32.s: Likewise.
* doc/c-bpf.texi (BPF Instructions): Update accordingly.

15 months agoPR30657, gprof heap buffer overflow
Alan Modra [Wed, 26 Jul 2023 00:24:03 +0000 (09:54 +0930)]
PR30657, gprof heap buffer overflow

PR 30657
* cg_arcs.c (cg_assemble): Sanity check find_call addresses.
* i386.c (i386_find_call): Don't access past end of core_text_space.
* aarch64.c (aarch64_find_call): Round up lowpc, round down highpc.
* alpha.c (alpha_find_call): Likewise.
* mips.c (mips_find_call): Likewise.
* sparc.c (sparc_find_call): Likewise.
* vax.c (vax_find_call): Sanity check core_text_space accesses.

15 months ago[GOLD] reporting local symbol names
Alan Modra [Fri, 21 Jul 2023 07:27:18 +0000 (16:57 +0930)]
[GOLD] reporting local symbol names

get_symbol_name currently returns "" for the usual STT_SECTION symbols
generated by gas.  That's not very helpful, return the section name.
Demangle local symbols too, fixing an inconsistency in
issue_discarded_error where global symbols are demangled.

* object.cc (Sized_relobj_file::get_symbol_name): Return a
std::string.  Return section name for STT_SECTION symbols with
zero st_name.  Sanity check st_name, and don't run off the end
of an improperly terminated .strtab.  Demangle sym names.
* object.h (Sized_relobj_file::get_symbol_name): Update decl.
* target-reloc.h (issue_discarded_error): Adjust.
* powerpc.cc (Target_powerpc::Relocate::relocate): Report reloc
type and symbol for relocation overflows.

15 months agoDon't warn on .attach_to_group to same group
Alan Modra [Fri, 21 Jul 2023 22:44:38 +0000 (08:14 +0930)]
Don't warn on .attach_to_group to same group

* config/obj-elf.c (obj_elf_attach_to_group): Don't warn if
group name matches current group for section.

15 months agobpf: format not a string literal
Alan Modra [Tue, 25 Jul 2023 06:41:51 +0000 (16:11 +0930)]
bpf: format not a string literal

* config/tc-bpf.c (md_assemble): Correct as_bad call.

15 months agoRegen bpf opcodes POTFILE
Alan Modra [Fri, 21 Jul 2023 12:19:27 +0000 (21:49 +0930)]
Regen bpf opcodes POTFILE

15 months agoAutomatic date update in version.in
GDB Administrator [Wed, 26 Jul 2023 00:00:23 +0000 (00:00 +0000)]
Automatic date update in version.in

15 months agobpf: Add atomic compare-and-exchange instructions
David Faust [Mon, 24 Jul 2023 21:50:34 +0000 (14:50 -0700)]
bpf: Add atomic compare-and-exchange instructions

This patch adds the two remaining BPF v3 atomic instructions:
- BPF_INSN_ACMP{,32}: atomic compare-and-swap
- BPF_INSN_AXCHG{,32}: atomic (non-conditional) exchange

Tests and documentation are also updated.

gas/
* doc/c-bpf.texi (BPF Instructions): Document atomic exchange and
atomic compare-and-swap instructions.
* testsuite/gas/bpf/atomic.s: Test ACMP, ACMP32, AXCHG, AXCGH32
instructions.
* testsuite/gas/bpf/atomic.d: Likewise.
* testsuite/gas/bpf/atomic-be.d: Likewise.
* testsuite/gas/bpf/atomic-pseudoc.s: Likewise.
* testsuite/gas/bpf/atomic-pseudoc.d: Likewise.
* testsuite/gas/bpf/atomic-be-pseudoc.d: Likewise.

include/
* opcode/bpf.h (BPF_IMM32_ACMP): Fix typo.
(enum bpf_insn_id): New entries for BPF_INSN_ACMP{,32} and
BPF_INSN_AXCHG{,32}.

opcodes/
* bpf-opc.c (bpf_opcodes): Add entries for ACMP{,32} and
AXCHG{,32} instructions.

15 months agobpf: Update atomic instruction pseudo-C syntax
David Faust [Tue, 25 Jul 2023 19:51:47 +0000 (12:51 -0700)]
bpf: Update atomic instruction pseudo-C syntax

This patch updates the pseudo-C dialect templates for the BPF v3 atomic
instructions.  The templates match the strings emitted by clang -S for
these instructions.

The tests and documentation are updated accordingly.

gas/
* doc/c-bpf.texi (BPF Instructions): Update entries for atomic
and 32-bit atomic instructions.
* testsuite/gas/bpf/atomic.s: Test AAND, AAND32, AOR, AOR32,
AXOR, AXOR32, AFADD, AFADD32, AFAND, AFAND32, AFOR, AFOR32,
AFXOR and AFXOR32 instructions.
* testsuite/gas/bpf/atomic.d: Likewise.
* testsuite/gas/bpf/atomic-be.d: Likewise.
* testsuite/gas/bpf/atomic-pseudoc.s: Likewise.
* testsuite/gas/bpf/atomic-pseudoc.d: Likewise.
* testsuite/gas/bpf/atomic-be-pseudoc.d: Likewise.
* testsuite/gas/bpf/atomic-v1.s: New test.
* testsuite/gas/bpf/atomic-v1.d: Likewise.
* testuiste/gas/bpf/atomic-v1-be.d: Likewise.
* testuiste/gas/bpf/bpf.exp: Run new tests.

opcodes/
* bpf-opc.c (bpf_opcodes): Update pseudo-C dialect templates for:
BPF_INSN_AADD, BPF_INSN_AOR, BPF_INSN_AAND, BPF_INSN_AXOR,
BPF_INSN_AFADD, BPF_INSN_AFOR, BPF_INSN_AFAND, BPF_INSN_AFXOR,
BPF_INSN_AADD32, BPF_INSN_AOR32, BPF_INSN_AAND32,
BPF_INSN_AXOR32, BPF_INSN_AFADD32, BPF_INSN_AFOR32,
BPF_INSN_AFAND32, and BPF_INSN_AFXOR32 instructions.

15 months agoRISC-V: Enable RVC on ".option arch, +zca" etc.
Tsukasa OI [Tue, 25 Jul 2023 01:40:09 +0000 (01:40 +0000)]
RISC-V: Enable RVC on ".option arch, +zca" etc.

Since the 'Zca' extension is the new base of the compressed instructions,
this commit enables RVC *also* when the 'Zca' extension is enabled
via ".option arch" directive.

gas/ChangeLog:

* config/tc-riscv.c (s_riscv_option): Enable RVC also when the
'Zca' extension is enabled after an ".option arch" directive.

15 months agoAutomatic date update in version.in
GDB Administrator [Tue, 25 Jul 2023 00:00:25 +0000 (00:00 +0000)]
Automatic date update in version.in

15 months agoRISC-V: Implications from 'Zc[fd]' extensions
Tsukasa OI [Mon, 24 Jul 2023 05:09:39 +0000 (05:09 +0000)]
RISC-V: Implications from 'Zc[fd]' extensions

The version 1.0.4-1 of the code size reduction specification clarifies
that 'Zcf' implies 'F' and 'Zcd' implies 'D'.

cf:
<https://github.com/riscv/riscv-code-size-reduction/releases/tag/v1.0.4-1>

This commit adds those implications.

bfd/ChangeLog:

* elfxx-riscv.c (riscv_implicit_subsets): Add two implications,
'Zcf' -> 'F' and 'Zcd' -> 'D'.

gas/ChangeLog:

* testsuite/gas/riscv/march-imply-zcd.d: New test.
* testsuite/gas/riscv/march-imply-zcf.d: New test.

15 months agoRISC-V: Prohibit the 'Zcf' extension on RV64
Tsukasa OI [Mon, 24 Jul 2023 05:01:12 +0000 (05:01 +0000)]
RISC-V: Prohibit the 'Zcf' extension on RV64

As per:
<https://github.com/riscv/riscv-code-size-reduction/issues/221>,
the 'Zcf' extension does not exist on RV64.  This is reflected on the
version 1.0.4-1 of the code size reduction specification:
<https://github.com/riscv/riscv-code-size-reduction/releases/tag/v1.0.4-1>.

This commit prohibits the combination: RV64 (or any ISA with XLEN > 32)
and the 'Zcf' extension.

bfd/ChangeLog:

* elfxx-riscv.c (riscv_parse_check_conflicts): Prohibit
combination of RV64 and 'Zcf'.

gas/ChangeLog:

* testsuite/gas/riscv/march-fail-rv64i_zcf.d: New test.
* testsuite/gas/riscv/march-fail-rv64i_zcf.l: Likewise.

15 months agoobjcopy embeds the current time and ignores SOURCE_DATE_EPOCH making the output unrep...
Johannes Schauer Marin Rodrigues [Mon, 24 Jul 2023 15:59:19 +0000 (16:59 +0100)]
objcopy embeds the current time and ignores SOURCE_DATE_EPOCH making the output unreproducible.

bfd
  * peXXigen.c (_bfd_XXi_only_swap_filehdr_out): If inserting a timestamp, use the value held in the SOURCE_DATE_EPOCH environment variable, if it is defined.
binutils
  * doc/binutils.texi (objcopy): Document change in behaviour of objcopy's --preserve-dates command line option.
ld
  * pe-dll.c (fill_edata): If inserting a timestamp, use the value held in the SOURCE_DATE_EPOCH environment variable, if it is defined.
  * ld.texi (--insert-timestamp): Document change in behaviour.

15 months agoUpdated translations for bfd, gold and opcodes
Nick Clifton [Mon, 24 Jul 2023 08:13:29 +0000 (09:13 +0100)]
Updated translations for bfd, gold and opcodes

15 months agoLoongArch: ld: Simplify inserting IRELATIVE relocations to .rela.dyn
mengqinggang [Tue, 11 Jul 2023 03:21:18 +0000 (11:21 +0800)]
LoongArch: ld: Simplify inserting IRELATIVE relocations to .rela.dyn

In LoongArch, the R_LARCH_IRELATIVE relocations for local ifunc symbols are
in .rela.dyn. Before, this is done by loongarch_elf_finish_dynamic_sections.
But this function is called after elf_link_sort_relocs, it need to find a
null slot to insert IRELATIVE relocation.

Now, it is processed by elf_loongarch_output_arch_local_syms before
elf_link_sort_relocs, just need to call loongarch_elf_append_rela to
insert IRELATIVE relocation.

bfd/ChangeLog:

* elfnn-loongarch.c (elfNN_allocate_local_ifunc_dynrelocs): Return
type change to int.
(loongarch_elf_size_dynamic_sections): Delete (void *).
(loongarch_elf_finish_dynamic_symbol): Use loongarch_elf_append_rela
insert IRELATIVE relocation to .rela.dyn.
(elfNN_loongarch_finish_local_dynamic_symbol): Return type change to
int.
(loongarch_elf_finish_dynamic_sections): Delete process of local
ifunc symbols.
(elf_backend_output_arch_local_syms): New.

ld/ChangeLog:

* testsuite/ld-loongarch-elf/local-ifunc-reloc.d: Regenerated.

15 months agoLoongArch: Fix immediate overflow check bug
mengqinggang [Sat, 15 Jul 2023 09:56:07 +0000 (17:56 +0800)]
LoongArch: Fix immediate overflow check bug

For B16/B21/B26/PCREL20_S2 relocations, if immediate overflow check after
rightshift, and the mask need to include sign bit.

Now, the immediate overflow check before rightshift for easier understand.

bfd/ChangeLog:

* elfxx-loongarch.c (reloc_bits_pcrel20_s2): Delete.
(reloc_bits_b16): Delete.
(reloc_bits_b21): Delete.
(reloc_bits_b26): Delete.
(reloc_sign_bits): New.

15 months agoLoongArch: Fix instruction immediate bug caused by sign-extend
mengqinggang [Sat, 15 Jul 2023 09:17:12 +0000 (17:17 +0800)]
LoongArch: Fix instruction immediate bug caused by sign-extend

For extreme code mode, the instruction sequences is
    pcalau12i $t0, hi20
    addi.d $t1, $zero, lo12
    lu32i.d $t1, lo20
    lu52i.d $t1, hi12
    add.d $t1, $t0, $t1

If lo12 > 0x7ff, hi20 need to add 0x1, lo20 need to sub 0x1.
If hi20 > 0x7ffff, lo20 need to add 0x1.

bfd/ChangeLog:

* elfnn-loongarch.c (RELOCATE_CALC_PC32_HI20): Redefined.
(RELOCATE_CALC_PC64_HI32): Redefined.

15 months agobpf: gas,include,opcode: add suppor for instructions BSWAP{16,32,64}
Jose E. Marchesi [Mon, 24 Jul 2023 00:54:06 +0000 (02:54 +0200)]
bpf: gas,include,opcode: add suppor for instructions BSWAP{16,32,64}

This patch adds support for the BPF V4 ISA byte swap instructions to
opcodes, assembler and disassembler.

Tested in bpf-unknown-none.

include/ChangeLog:

2023-07-24  Jose E. Marchesi  <jose.marchesi@oracle.com>

* opcode/bpf.h (BPF_IMM32_BSWAP16): Define.
(BPF_IMM32_BSWAP32): Likewise.
(BPF_IMM32_BSWAP64): Likewise.
(enum bpf_insn_id): New entries BPF_INSN_BSWAP{16,32,64}.

opcodes/ChangeLog:

2023-07-24  Jose E. Marchesi  <jose.marchesi@oracle.com>

* bpf-opc.c (bpf_opcodes): Add entries for the BSWAP*
instructions.

gas/ChangeLog:

2023-07-24  Jose E. Marchesi  <jose.marchesi@oracle.com>

* doc/c-bpf.texi (BPF Instructions): Document BSWAP* instructions.
* testsuite/gas/bpf/alu.s: Test BSWAP{16,32,64} instructions.
* testsuite/gas/bpf/alu.d: Likewise.
* testsuite/gas/bpf/alu-be.d: Likewise.
* testsuite/gas/bpf/alu-pseudoc.s: Likewise.
* testsuite/gas/bpf/alu-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu-be-pseudoc.d: Likewise.

15 months agobpf: gas: fix in manual that MOVS* pseudoc syntax uses = instead of s=
Jose E. Marchesi [Mon, 24 Jul 2023 00:20:55 +0000 (02:20 +0200)]
bpf: gas: fix in manual that MOVS* pseudoc syntax uses = instead of s=

gas/ChangeLog:

2023-07-24  Jose E. Marchesi  <jose.marchesi@oracle.com>

* doc/c-bpf.texi (BPF Instructions): The pseudoc syntax for MOVS*
doesn't use `s=' but `='.

15 months agobpf: gas,opcodes: fix pseudoc syntax for MOVS* and LDXS* insns
Jose E. Marchesi [Mon, 24 Jul 2023 00:11:34 +0000 (02:11 +0200)]
bpf: gas,opcodes: fix pseudoc syntax for MOVS* and LDXS* insns

This patch fixes the pseudoc syntax of the V4 instructions MOVS* and
LDXS* in order to reflect https://reviews.llvm.org/D144829.

opcodes/ChangeLog:

2023-07-24  Jose E. Marchesi  <jose.marchesi@oracle.com>

* bpf-opc.c (bpf_opcodes): Fix pseudo-c syntax for MOVS* and LDXS*
instructions.

gas/ChangeLog:

2023-07-24  Jose E. Marchesi  <jose.marchesi@oracle.com>

* doc/c-bpf.texi (BPF Instructions): Fix pseudoc syntax for MOVS*
and LDXS* instructions.
* testsuite/gas/bpf/mem-pseudoc.d: Likewise.
* testsuite/gas/bpf/mem-be-pseudoc.d: Likewise.
* testsuite/gas/bpf/mem-pseudoc.s: Likewise.
* testsuite/gas/bpf/alu-pseudoc.s: Likewise.
* testsuite/gas/bpf/alu-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu-be-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu32-pseudoc.s: Likewise.
* testsuite/gas/bpf/alu32-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu32-be-pseudoc.d: Likewise.

15 months agoAutomatic date update in version.in
GDB Administrator [Mon, 24 Jul 2023 00:00:23 +0000 (00:00 +0000)]
Automatic date update in version.in

15 months agobpf: add support for jal/gotol jump instruction with 32-bit target
Jose E. Marchesi [Sun, 23 Jul 2023 23:15:08 +0000 (01:15 +0200)]
bpf: add support for jal/gotol jump instruction with 32-bit target

This patch adds support for the V4 BPF instruction jal/gotol, which is
like ja/goto but it supports a signed 32-bit PC-relative (in number of
64-bit words minus one) target operand instead of the 16-bit signed
operand of the other instruction.  This greatly increases the jump
range in BPF programs.

Tested in bpf-unkown-none.

bfd/ChangeLog:

2023-07-24  Jose E. Marchesi  <jose.marchesi@oracle.com>

* reloc.c: New reloc BFD_RELOC_BPF_DISPCALL32.
* elf64-bpf.c (bpf_reloc_type_lookup): Handle the new reloc.
* libbfd.h (bfd_reloc_code_real_names): Regenerate.

gas/ChangeLog:

2023-07-24  Jose E. Marchesi  <jose.marchesi@oracle.com>

* config/tc-bpf.c (struct bpf_insn): New field `id'.
(md_assemble): Save the ids of successfully parsed instructions
and use the new BFD_RELOC_BPF_DISPCALL32 whenever appropriate.
(md_apply_fix): Adapt to the new BFD reloc.
* testsuite/gas/bpf/jump.s: Test JAL.
* testsuite/gas/bpf/jump.d: Likewise.
* testsuite/gas/bpf/jump-pseudoc.d: Likewise.
* testsuite/gas/bpf/jump-be.d: Likewise.
* testsuite/gas/bpf/jump-be-pseudoc.d: Likewise.
* doc/c-bpf.texi (BPF Instructions): Document new instruction
jal/gotol.
Document new operand type disp32.

include/ChangeLog:

2023-07-24  Jose E. Marchesi  <jose.marchesi@oracle.com>

* opcode/bpf.h (enum bpf_insn_id): Add entry BPF_INSN_JAL.
(enum bpf_insn_id): Remove spurious entry BPF_INSN_CALLI.

opcodes/ChangeLog:

2023-07-23  Jose E. Marchesi  <jose.marchesi@oracle.com>

* bpf-opc.c (bpf_opcodes): Add entry for jal.

15 months agoUse 'name' in DAP start_thread function
Tom Tromey [Wed, 28 Jun 2023 21:23:54 +0000 (15:23 -0600)]
Use 'name' in DAP start_thread function

The DAP start_thread helper function has a 'name' parameter that is
unused.  Apparently I forgot to hook it up to the thread constructor.
This patch fixes the oversight.

15 months agoExport gdb.block_signals and create gdb.Thread
Tom Tromey [Tue, 4 Jul 2023 15:15:54 +0000 (09:15 -0600)]
Export gdb.block_signals and create gdb.Thread

While working on an experiment, I realized that I needed the DAP
block_signals function.  I figured other developers may need it as
well, so this patch moves it from DAP to the gdb module and exports
it.

I also added a new subclass of threading.Thread that ensures that
signals are blocked in the new thread.

Finally, this patch slightly rearranges the documentation so that
gdb-side threading issues and functions are all discussed in a single
node.

15 months agogdb: two changes to linux_nat_debug_printf calls in linux-nat.c
Andrew Burgess [Sat, 22 Jul 2023 14:28:28 +0000 (15:28 +0100)]
gdb: two changes to linux_nat_debug_printf calls in linux-nat.c

This commit adjusts some of the debug output in linux-nat.c, but makes
no other functional changes to GDB.

In resume_lwp I've added the word "sibling" to one of the debug
messages.  All the other debug messages in this function talk about
operating on the sibling thread, so I think it makes sense, for
consistency, if the message I've updated also talks about the sibling
thread.

In resume_stopped_resumed_lwps I've reordered the condition checks so
that the vfork-parent check now happens after the checks for whether
the thread is already resumed or not.  This makes no functional
difference to GDB, but does, I think, mean we see more helpful debug
messages first.

Consider the situation where a vfork-parent thread is already resumed,
and resume_stopped_resumed_lwps is called.  Previously the message
saying that the thread was not being resumed due to being a
vfork-parent, was printed.  This might give the impression that the
thread is left in a not resumed state, which is misleading.

After this change we now get a message saying that the thread is not
being resumed due to it not being stopped (i.e. is already resumed).
With this message the already resumed nature of the thread is much
clearer.

I found this change helpful when debugging some vfork related issues.

15 months agogdb/testsuite: replace $testfile with $binfile in one case
Andrew Burgess [Sat, 22 Jul 2023 14:33:23 +0000 (15:33 +0100)]
gdb/testsuite: replace $testfile with $binfile in one case

For *reasons* I was hacking on gdb.base/foll-vfork.exp and wanted to
change the name of the binary that was created.  Should be easy, I
adjusted the global $binfile variable .... but that didn't work.

In one place the script uses $testfile instead of $binfile.

Fixed this to use $binfile, now I can easily change the name of the
generated binary, and the test still works.

There's no change in what is tested after this commit.

15 months agoAutomatic date update in version.in
GDB Administrator [Sun, 23 Jul 2023 00:00:19 +0000 (00:00 +0000)]
Automatic date update in version.in

15 months ago[gdb/testsuite] Improve gdb.arch/arm-pthread_cond_timedwait-bt.exp
Tom de Vries [Sat, 22 Jul 2023 09:00:25 +0000 (11:00 +0200)]
[gdb/testsuite] Improve gdb.arch/arm-pthread_cond_timedwait-bt.exp

I noticed in test-case gdb.arch/arm-pthread_cond_timedwait-bt.exp that
prepare_for_testing is used, followed by a clean_restart.

This calls clean_restart twice in a row.

Fix this by using build_executable instead.

Also, I noticed that the test-case requires an SVC instruction, so add a
require to limit the test-case to supported architectures.

While we're at it, run M-x indent-region in emacs to fix indentation.

Tested on x86_64-linux.

15 months ago[gdb/testsuite] Use proc readnow in two test-cases
Tom de Vries [Sat, 22 Jul 2023 09:00:25 +0000 (11:00 +0200)]
[gdb/testsuite] Use proc readnow in two test-cases

Use "require !readnow" in two test-cases, instead of the written-out variant.

Tested on x86_64-linux, with target boards unix and readnow.

15 months agoAutomatic date update in version.in
GDB Administrator [Sat, 22 Jul 2023 00:00:19 +0000 (00:00 +0000)]
Automatic date update in version.in

15 months agoFix crash with DW_FORM_implicit_const
Tom Tromey [Thu, 20 Jul 2023 21:48:46 +0000 (15:48 -0600)]
Fix crash with DW_FORM_implicit_const

Jakub pointed out that using DW_FORM_implicit_const with
DW_AT_bit_size would cause gdb to crash.  This happened because
DW_FORM_implicit_const is not an "unsigned" form, causing as_unsigned
to assert.  This patch fixes the problem.

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30651
Approved-By: Andrew Burgess <aburgess@redhat.com>
15 months agobpf: disasemble offsets of value 0 as "+0"
David Faust [Fri, 21 Jul 2023 18:41:43 +0000 (11:41 -0700)]
bpf: disasemble offsets of value 0 as "+0"

This tiny patch makes the BPF disassembler to emit, e.g.

  ldxdw %r1, [%r0+0]

instead of

  ldxdw %r1, [%r00]

when the offset is 0, to avoid confusion.

opcodes/

* bpf-dis.c (print_insn_bpf): Print offsets with value 0 as "+0".

gas/

* testsuite/gas/bpf/mem.s: Add tests with offset 0.
* testsuite/gas/bpf/mem-pseudoc.s: Likewise.
* testsuite/gas/bpf/mem.d: Update accordingly.
* testsuite/gas/bpf/mem-be.d: Likewise.
* testsuite/gas/bpf/mem-pseudoc.d: Likewise.
* testsuite/gas/bpf/mem-be-pseudoc.d: Likewise.

15 months agoImplement DAP modules request
Tom Tromey [Mon, 22 May 2023 14:09:18 +0000 (08:09 -0600)]
Implement DAP modules request

This implements the DAP "modules" request, and also arranges to add
the module ID to stack frames.

15 months agoAdd Progspace.objfile_for_address
Tom Tromey [Mon, 22 May 2023 17:40:10 +0000 (11:40 -0600)]
Add Progspace.objfile_for_address

This adds a new objfile_for_address method to gdb.Progspace.  This
makes it easy to find the objfile for a given address.

There's a related PR; and while this change would have been sufficient
for my original need, it's not clear to me whether I should close the
bug.  Nevertheless I think it makes sense to at least mention it here.

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=19288
Reviewed-By: Eli Zaretskii <eliz@gnu.org>
15 months agoRemove unused imports
Tom Tromey [Mon, 22 May 2023 19:43:31 +0000 (13:43 -0600)]
Remove unused imports

I noticed an unused import in dap/evaluate.py; and also I found out
that my recent changes to use frame filters from DAP left some unused
imports in dap/bt.py.

15 months agoDocument array indexing for Python gdb.Value
Tom Tromey [Tue, 18 Jul 2023 14:38:56 +0000 (08:38 -0600)]
Document array indexing for Python gdb.Value

I noticed that the documentation for gdb.Value doesn't mention array
indexing.

Approved-By: Eli Zaretskii <eliz@gnu.org>
15 months agobpf: opcodes, gas: support for signed load V4 instructions
Jose E. Marchesi [Fri, 21 Jul 2023 17:47:49 +0000 (19:47 +0200)]
bpf: opcodes, gas: support for signed load V4 instructions

This commit adds the signed load to register (ldxs*) instructions
introduced in the BPF ISA version 4, including opcodes and assembler
tests.

Tested in bpf-unknown-none.

include/ChangeLog:

2023-07-21  Jose E. Marchesi  <jose.marchesi@oracle.com>

* opcode/bpf.h (enum bpf_insn_id): Add entries for signed load
instructions.
(BPF_MODE_SMEM): Define.

opcodes/ChangeLog:

2023-07-21  Jose E. Marchesi  <jose.marchesi@oracle.com>

* bpf-opc.c (bpf_opcodes): Add entries for LDXS{B,W,H,DW}
instructions.

gas/ChangeLog:

2023-07-21  Jose E. Marchesi  <jose.marchesi@oracle.com>

* testsuite/gas/bpf/mem.s: Add signed load instructions.
* testsuite/gas/bpf/mem-pseudoc.s: Likewise.
* testsuite/gas/bpf/mem.d: Likewise.
* testsuite/gas/bpf/mem-pseudoc.d: Likewise.
* testsuite/gas/bpf/mem-be.d: Likewise.
* doc/c-bpf.texi (BPF Instructions): Document the signed load
instructions.

15 months agobpf: opcodes, gas: support for signed register move V4 instructions
Jose E. Marchesi [Fri, 21 Jul 2023 15:22:58 +0000 (17:22 +0200)]
bpf: opcodes, gas: support for signed register move V4 instructions

This commit adds the signed register move (movs) instructions
introduced in the BPF ISA version 4, including opcodes and assembler
tests.

Tested in bpf-unknown-none.

include/ChangeLog:

2023-07-21  Jose E. Marchesi  <jose.marchesi@oracle.com>

* opcode/bpf.h (BPF_OFFSET16_MOVS8): Define.
(BPF_OFFSET16_MOVS16): Likewise.
(BPF_OFFSET16_MOVS32): Likewise.
(enum bpf_insn_id): Add entries for MOVS{8,16,32}R and
MOVS32{8,16,32}R.

opcodes/ChangeLog:

2023-07-21  Jose E. Marchesi  <jose.marchesi@oracle.com>

* bpf-opc.c (bpf_opcodes): Add entries for MOVS{8,16,32}R and
MOVS32{8,16,32}R instructions.  and MOVS32I instructions.

gas/ChangeLog:

2023-07-21  Jose E. Marchesi  <jose.marchesi@oracle.com>

* testsuite/gas/bpf/alu.s: Test movs instructions.
* testsuite/gas/bpf/alu-pseudoc.s: Likewise.
* testsuite/gas/bpf/alu32.s: Likewise for movs32 instruction.
* testsuite/gas/bpf/alu32-pseudoc.s: Likewise.
* testsuite/gas/bpf/alu.d: Add expected results.
* testsuite/gas/bpf/alu32.d: Likewise.
* testsuite/gas/bpf/alu-be.d: Likewise.
* testsuite/gas/bpf/alu32-be.d: Likewise.
* testsuite/gas/bpf/alu-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu32-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu-be-pseudoc.d: Likewise.
* testsuite/gas/bpf/alu32-be-pseudoc.d: Likewise.

15 months agoRemove redundant @findex from python.texi
Tom Tromey [Fri, 21 Jul 2023 16:02:13 +0000 (10:02 -0600)]
Remove redundant @findex from python.texi

In a review, Eli pointed out that @findex is redundant when used with
@defun.  This patch removes all such uses from python.texi, plus a
couple uses before @defvar that are also unnecessary.

Approved-By: Eli Zaretskii <eliz@gnu.org>
15 months agoFix typo in py-type.c docstring
Tom Tromey [Fri, 21 Jul 2023 16:33:07 +0000 (10:33 -0600)]
Fix typo in py-type.c docstring

I noticed that a doc string py-type.c says "an signed".
This patch corrects it to "a signed".

15 months agoImplement Ada target name symbol
Tom Tromey [Fri, 23 Jun 2023 12:38:55 +0000 (06:38 -0600)]
Implement Ada target name symbol

Ada 2022 adds the "target name symbol", which can be used on the right
hand side of an assignment to refer to the left hand side.  This
allows for convenient updates.  This patch implements this for gdb's
Ada expression parser.

Reviewed-By: Eli Zaretskii <eliz@gnu.org>
15 months agoAdd instruction bytes to DAP disassembly response
Tom Tromey [Wed, 28 Jun 2023 12:57:16 +0000 (06:57 -0600)]
Add instruction bytes to DAP disassembly response

The DAP disassemble command lets the client return the underlying
bytes of the instruction in an implementation-defined format.  This
patch updates gdb to return this, and simply uses a hex string of the
bytes as the format.

Reviewed-By: Eli Zaretskii <eliz@gnu.org>
15 months agoRemove ancient Ada workaround
Tom Tromey [Tue, 18 Jul 2023 16:45:14 +0000 (10:45 -0600)]
Remove ancient Ada workaround

I ran across this very old code in gdb's Ada support.  After a bit of
archaeology, we couldn't determine what bug this might have been
working around.  It is no longer needed, so this patch removes it.

As this is entirely Ada-specific and was reviewed and tested at
AdaCore, I'm checking it in.

15 months agobpf: add missing bpf-dis.c to opcodes/Makefile.am
Jose E. Marchesi [Fri, 21 Jul 2023 12:38:19 +0000 (14:38 +0200)]
bpf: add missing bpf-dis.c to opcodes/Makefile.am

This was breaking --enable-targets=all builds.

opcodes/ChangeLog:

2023-07-21  Jose E. Marchesi  <jose.marchesi@oracle.com>

* Makefile.am (TARGET64_LIBOPCODES_CFILES): Add missing bpf-dis.c
* Makefile.in: Regenerate.

15 months agosim/bpf: desCGENization of the BPF simulator
Jose E. Marchesi [Mon, 17 Jul 2023 16:35:22 +0000 (18:35 +0200)]
sim/bpf: desCGENization of the BPF simulator

The BPF port in binutils has been rewritten (commit
d218e7fedc74d67837d2134120917f4ac877454c) in order to not be longer
based on CGEN.  Please see that commit log for more information.

This patch updates the BPF simulator accordingly.  The new
implementation is much simpler and it is based on the new BPF opcodes.

Tested with target bpf-unknown-none with both 64-bit little-endian
host and 32-bit little-endian host.

Note that I have not tested in a big-endian host yet.  I will do so
once this lands upstream so I can use the GCC compiler farm.

15 months agoDesCGENization of the BPF binutils port
Jose E. Marchesi [Fri, 14 Jul 2023 22:50:14 +0000 (00:50 +0200)]
DesCGENization of the BPF binutils port

CGEN is cool, but the BPF architecture is simply too bizarre for it.

The weird way of BPF to handle endianness in instruction encoding, the
weird C-like alternative assembly syntax, the weird abuse of
multi-byte (or infra-byte) instruction fields as opcodes, the unusual
presence of opcodes beyond the first 32-bits of some instructions, are
all examples of what makes it a PITA to continue using CGEN for this
port.  The bpf.cpu file is becoming so complex and so nested with
p-macros that it is very difficult to read, and quite challenging to
update.  Also, every time we are forced to change something in CGEN to
accommodate BPF requirements (which is often) we have to do extensive
testing to make sure we do not break any other target using CGEN.

This is getting un-maintenable.

So I have decided to bite the bullet and revamp/rewrite the port so it
no longer uses CGEN.  Overall, this involved:

* To remove the cpu/bpf.{cpu,opc} descriptions.

* To remove the CGEN generated files.

* To replace the CGEN generated opcodes table with a new hand-written
  opcodes table for BPF.

* To replace the CGEN generated disassembler wih a new disassembler
  that uses the new opcodes.

* To replace the CGEN generated assembler with a new assembler that uses the
  new opcodes.

* To replace the CGEN generated simulator with a new simulator that uses the
  new opcodes. [This is pushed in GDB in another patch.]

* To adapt the build systems to the new situation.

Additionally, this patch introduces some extensions and improvements:

* A new BPF relocation BPF_RELOC_BPF_DISP16 plus corresponding ELF
  relocation R_BPF_GNU_64_16 are added to the BPF BFD port.  These
  relocations are used for section-relative 16-bit offsets used in
  load/store instructions.

* The disassembler now has support for the "pseudo-c" assembly syntax of
  BPF.  What dialect to use when disassembling is controlled by a command
  line option.

* The disassembler now has support for dumping instruction immediates in
  either octal, hexadecimal or decimal.  The used output base is controlled
  by a new command-line option.

* The GAS BPF test suite has been re-structured and expanded in order to
  test the disassembler pseudoc syntax support.  Minor bugs have been also
  fixed there.  The assembler generic tests that were disabled for bpf-*-*
  targets due to the previous implementation of pseudoc syntax are now
  re-enabled.  Additional tests have been added to test the new features of
  the assembler.  .dump files are no longer used.

* The linker BPF test suite has been adapted to the command line options
  used by the new disassembler.

The result is very satisfactory.  This patchs adds 3448 lines of code
and removes 10542 lines of code.

Tested in:

* Target bpf-unknown-none with 64-bit little-endian host and 32-bit
  little-endian host.

* Target x86-64-linux-gnu with --enable-targets=all

Note that I have not tested in a big-endian host yet.  I will do so
once this lands upstream so I can use the GCC compiler farm.

I have not included ChangeLog entries in this patch: these would be
massive and not very useful, considering this is pretty much a rewrite
of the port.  I beg the indulgence of the global maintainers.

15 months agogdb/solib-rocm: limit the number of opened file descriptors
Lancelot Six [Thu, 20 Jul 2023 10:15:50 +0000 (10:15 +0000)]
gdb/solib-rocm: limit the number of opened file descriptors

ROCm programs can load a high number of compute kernels on GPU devices,
especially if lazy code-object loading have been disabled.  Each code
object containing such program is loaded once for each device available,
and each instance is reported by GDB as an individual shared library.

We came across situations where the number of shared libraries opened by
GDB gets higher than the allowed number of opened files for the process.
Increasing the opened files limit works around the problem, but there is a
better way this patch proposes to follow.

Under the hood, the GPU code objects are embedded inside the host
application binary and shared library binaries.  GDB currently opens the
underlying file once for each shared library it sees.  That means that
the same file is re-opened every time a code object is loaded on a GPU.

This patch proposes to only open each underlying file once.  This is
done by implementing a reference counting mechanism so the underlying
file is opened when the underlying file first needs to be opened, and
closed when the last BFD using the underlying file is closed.

On a program where GDB used to open about 1500 files to load all shared
libraries, this patch makes it so only 54 opened file descriptors are
needed.

I have tested this patch on downstream ROCgdb's full testsuite and
upstream GDB testsuite with no regression.

Approved-By: Pedro Alves <pedro@palves.net>
15 months agox86: adjust disassembly of insns operating on selector values
Jan Beulich [Fri, 21 Jul 2023 06:57:24 +0000 (08:57 +0200)]
x86: adjust disassembly of insns operating on selector values

Bring disassembly back in line with what the assembler accepts, thus
also making it self-consistent (with, in particular selector load/store
insns). While there further add D to all affected insns except ARPL
(where S is used, matching LAR/LSL), to also behave correctly in suffix-
always mode.

While there also hook up the Intel variant of the LKGS test.