SMP syncronization on AMD processors (broken?)
In-Reply-To: <[email protected]>
Quoted 3 lines · Kirill Korotaev
The question is whether concurrent spin_lock()'s should acquire it in more or less "fair" fashinon or one of CPUs can starve any arbitrary time while others do reacquire it in a loop.
You neglected to say what CPU type you compiled the kernel for.
If it wasn't Pentium Pro maybe you could patch include/asm-i386/spinlock.h
line 82 (or the same place in x86-64) like this:
___
* (PPro errata 66, 92)
*/
-#if !defined(CONFIG_X86_OOSTORE) && !defined(CONFIG_X86_PPRO_FENCE) +#if 0
#define __raw_spin_unlock_string \
"movb $1,%0" \___
The data might not make it out of the CPU write buffer without a locking instruction doing the update.
__
Chuck