On 18.07.20 15:48, John David Anglin wrote: > On 2020-07-18 9:08 a.m., John David Anglin wrote: >>>> diff --git a/arch/parisc/kernel/syscall.S b/arch/parisc/kernel/syscall.S >>>> index f05c9d5b6b9e..4ba43a3c7c67 100644 >>>> --- a/arch/parisc/kernel/syscall.S >>>> +++ b/arch/parisc/kernel/syscall.S >>>> @@ -659,8 +678,15 @@ cas_action: >>>> /* Error occurred on load or store */ >>>> /* Free lock */ >>>> #ifdef CONFIG_SMP >>>> -98: LDCW 0(%sr2,%r20), %r1 /* Barrier */ >>>> +98: sync /* Barrier */ >>>> 99: ALTERNATIVE(98b, 99b, ALT_COND_NO_SMP, INSN_NOP) >>>> + nop >>>> + nop >>>> + nop >>>> + nop >>>> + nop >>>> + nop >>>> + nop >>> I think you should put those nop's before the 99:ALTERNATIVE() instruction above. >>> That way they all get replaced by one jump instead of multiple nops (when >>> there is only one CPU in the SMP system). >> Okay. I'll test. Thought the replacement was just nops. > Do we know that a jump is faster than 8 nops? I don't know if it's faster. > We need ALTERNATIVE conditions for the barrier defines. Yes, that's Ok. Helge
