> I'm chasing down one SMP problem that apparently started triggering with
> the (fixed) 2.1.112. It might be the lock, for all I know, but it's
> interesting to note that the new lock code _should_ do the same thing as
> the old one with the exception that the new code won't ever work with
> interrupts trying to get the kernel lock (which is a no-no and has been
> for a long time, but maybe there was something there still doing it).
there is a difference, back half a year i had some very rare lockups
occuring when some (buggy) code did a lock_kernel() with IRQs turned off.
(this might block a pending APIC irq, typically it was the timer
interrupt, and if another CPU is waiting for jiffies to increase, we were
cooked.)
the fix for the old code was to add a 'sti' and 'cli' to the slow path of
lock_kernel(). This is not preserved in the 2.1.112 version. I suppose you
debugging version checked for IRQs turned-on, so this probably is a
nonproblem.
-- mingo
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html