Add generic and target specific support for local{,64}_try_cmpxchg
and wire up support for all targets that use local_t infrastructure.
The patch enables x86 targets to emit special instruction for
local_try_cmpxchg and also local64_try_cmpxchg for x86_64.
The last patch changes __perf_output_begin in events/ring_buffer
to use new locking primitive and improves code from
4b3: 48 8b 82 e8 00 00 00 mov 0xe8(%rdx),%rax
4ba: 48 8b b8 08 04 00 00 mov 0x408(%rax),%rdi
4c1: 8b 42 1c mov 0x1c(%rdx),%eax
4c4: 48 8b 4a 28 mov 0x28(%rdx),%rcx
4c8: 85 c0 test %eax,%eax
...
4ef: 48 89 c8 mov %rcx,%rax
4f2: 48 0f b1 7a 28 cmpxchg %rdi,0x28(%rdx)
4f7: 48 39 c1 cmp %rax,%rcx
4fa: 75 b7 jne 4b3 <...>
to
4b2: 48 8b 4a 28 mov 0x28(%rdx),%rcx
4b6: 48 8b 82 e8 00 00 00 mov 0xe8(%rdx),%rax
4bd: 48 8b b0 08 04 00 00 mov 0x408(%rax),%rsi
4c4: 8b 42 1c mov 0x1c(%rdx),%eax
4c7: 85 c0 test %eax,%eax
...
4d4: 48 89 c8 mov %rcx,%rax
4d7: 48 0f b1 72 28 cmpxchg %rsi,0x28(%rdx)
4dc: 0f 85 d0 00 00 00 jne 5b2 <...>
...
5b2: 48 89 c1 mov %rax,%rcx
5b5: e9 fc fe ff ff jmp 4b6 <...>
Please note that in addition to removed compare, the load from
0x28(%rdx) gets moved out of the loop and the code is rearranged
according to likely/unlikely tags in the source.
Cc: Richard Henderson <richard.henderson@xxxxxxxxxx>
Cc: Ivan Kokshaysky <ink@xxxxxxxxxxxxxxxxxxxx>
Cc: Matt Turner <mattst88@xxxxxxxxx>
Cc: Huacai Chen <chenhuacai@xxxxxxxxxx>
Cc: WANG Xuerui <kernel@xxxxxxxxxx>
Cc: Thomas Bogendoerfer <tsbogend@xxxxxxxxxxxxxxxx>
Cc: Michael Ellerman <mpe@xxxxxxxxxxxxxx>
Cc: Nicholas Piggin <npiggin@xxxxxxxxx>
Cc: Christophe Leroy <christophe.leroy@xxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Ingo Molnar <mingo@xxxxxxxxxx>
Cc: Borislav Petkov <bp@xxxxxxxxx>
Cc: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>
Cc: x86@xxxxxxxxxx
Cc: "H. Peter Anvin" <hpa@xxxxxxxxx>
Cc: Arnd Bergmann <arnd@xxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
Cc: Mark Rutland <mark.rutland@xxxxxxx>
Cc: Alexander Shishkin <alexander.shishkin@xxxxxxxxxxxxxxx>
Cc: Jiri Olsa <jolsa@xxxxxxxxxx>
Cc: Namhyung Kim <namhyung@xxxxxxxxxx>
Cc: Ian Rogers <irogers@xxxxxxxxxx>
Cc: Will Deacon <will@xxxxxxxxxx>
Cc: Boqun Feng <boqun.feng@xxxxxxxxx>
Cc: Jiaxun Yang <jiaxun.yang@xxxxxxxxxxx>
Cc: Jun Yi <yijun@xxxxxxxxxxx>
Uros Bizjak (10):
locking/atomic: Add missing cast to try_cmpxchg() fallbacks
locking/atomic: Add generic try_cmpxchg{,64}_local support
locking/alpha: Wire up local_try_cmpxchg
locking/loongarch: Wire up local_try_cmpxchg
locking/mips: Wire up local_try_cmpxchg
locking/powerpc: Wire up local_try_cmpxchg
locking/x86: Wire up local_try_cmpxchg
locking/generic: Wire up local{,64}_try_cmpxchg
locking/x86: Enable local{,64}_try_cmpxchg
perf/ring_buffer: use local_try_cmpxchg in __perf_output_begin
arch/alpha/include/asm/local.h | 2 ++
arch/loongarch/include/asm/local.h | 2 ++
arch/mips/include/asm/local.h | 2 ++
arch/powerpc/include/asm/local.h | 11 ++++++
arch/x86/include/asm/cmpxchg.h | 6 ++++
arch/x86/include/asm/local.h | 2 ++
include/asm-generic/local.h | 1 +
include/asm-generic/local64.h | 2 ++
include/linux/atomic/atomic-arch-fallback.h | 40 ++++++++++++++++-----
include/linux/atomic/atomic-instrumented.h | 20 ++++++++++-
kernel/events/ring_buffer.c | 5 +--
scripts/atomic/gen-atomic-fallback.sh | 6 +++-
scripts/atomic/gen-atomic-instrumented.sh | 2 +-
13 files changed, 87 insertions(+), 14 deletions(-)
--
2.39.2
