set_bit() called by set_cpu_cap() is a locked bit set instruction for atomic operation.
Since the c->x86_capability is not aligned to cache line depending on compiler, the locked bit set instruction falls into split lock situation which causes #AC exception when #AC exception is enabled by split locked accesses. But set_cpu_cap() is only called in early init phase on a CPU and therefore there is no contention for set_cpu_cap(). So it's unnecessary to be atomic operation. Using __set_bit(), which is not a locked instruction, can fix the kernel split lock issue and is faster than locked instruction. Signed-off-by: Fenghua Yu <[email protected]> Acked-by: Dave Hansen <[email protected]> --- arch/x86/include/asm/cpufeature.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/arch/x86/include/asm/cpufeature.h b/arch/x86/include/asm/cpufeature.h index b27da9602a6d..604f4c514efa 100644 --- a/arch/x86/include/asm/cpufeature.h +++ b/arch/x86/include/asm/cpufeature.h @@ -128,7 +128,8 @@ extern const char * const x86_bug_flags[NBUGINTS*32]; #define boot_cpu_has(bit) cpu_has(&boot_cpu_data, bit) -#define set_cpu_cap(c, bit) set_bit(bit, (unsigned long *)((c)->x86_capability)) +#define set_cpu_cap(c, bit) \ + __set_bit(bit, (unsigned long *)((c)->x86_capability)) extern void setup_clear_cpu_cap(unsigned int bit); extern void clear_cpu_cap(struct cpuinfo_x86 *c, unsigned int bit); -- 2.5.0
