locking/atomic/x86: Introduce arch_try_cmpxchg64() for !CONFIG_X86_CMPXCHG64
authorUros Bizjak <ubizjak@gmail.com>
Mon, 8 Apr 2024 09:13:58 +0000 (11:13 +0200)
committerIngo Molnar <mingo@kernel.org>
Tue, 9 Apr 2024 07:51:03 +0000 (09:51 +0200)
commitaef95dac9ce4f271cc43195ffc175114ed934cbe
treef94e5309223aa88569dc21294d4542240a4d9eb5
parent7016cc5def44b9dcb28089efae4412fa0d6c78c2
locking/atomic/x86: Introduce arch_try_cmpxchg64() for !CONFIG_X86_CMPXCHG64

Commit:

  6d12c8d308e68 ("percpu: Wire up cmpxchg128")

improved emulated cmpxchg8b_emu() library function to return
success/failure in a ZF flag.

Define arch_try_cmpxchg64() for !CONFIG_X86_CMPXCHG64 targets
to override the generic archy_try_cmpxchg() with an optimized
target specific implementation that handles ZF flag.

The assembly code at the call sites improves from:

   bf56d: e8 fc ff ff ff        call   cmpxchg8b_emu
   bf572: 8b 74 24 28           mov    0x28(%esp),%esi
   bf576: 89 c3                 mov    %eax,%ebx
   bf578: 89 d1                 mov    %edx,%ecx
   bf57a: 8b 7c 24 2c           mov    0x2c(%esp),%edi
   bf57e: 89 f0                 mov    %esi,%eax
   bf580: 89 fa                 mov    %edi,%edx
   bf582: 31 d8                 xor    %ebx,%eax
   bf584: 31 ca                 xor    %ecx,%edx
   bf586: 09 d0                 or     %edx,%eax
   bf588: 0f 84 e3 01 00 00     je     bf771 <...>

to:

   bf572: e8 fc ff ff ff        call   cmpxchg8b_emu
   bf577: 0f 84 b6 01 00 00     je     bf733 <...>

Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Link: https://lore.kernel.org/r/20240408091547.90111-4-ubizjak@gmail.com
arch/x86/include/asm/cmpxchg_32.h