I've had this wierdity off and on where cpu bound tasks won't respond to ^C
or a kill running another higher priority CPU bound task will finally make it
see the signal - this seems to have gone away with 2.2.0 but I still do have
some strange something happening on one of my over night jobs it does the
same sort of thing - it takes over the cpu - as soon as I touch a key in the
morning everything comes right and all I see to show for it is a silly load
average.
So I went and looked at the signal delivery code and I realised that
if you try and send a signal to a process that's running on another CPU
it doesn't see the signal untill the cpu next gets rescheduled - whether this
has anything to do with the problems I've been seeing I don't know - (I'd expect
that signals would eventually be delivered at the next timer interrupt).
Anyway I've produced the following patch to force a reschedule to a CPU running
a process that has a signal delivered to it - to my purely subjective eye X
seems to be running faster but I suspect that that's all in my head :-)
While I have a lot of Unix kernel experience this is my first time hacking on
Linux so please show me any stupid mistakes in the following:
thanks
Paul Campbell
[EMAIL PROTECTED]
--- linux-2.2.0-final/kernel/signal.c Fri Dec 18 09:37:22 1998
+++ linux/kernel/signal.c Mon Feb 1 22:17:49 1999
@@ -363,8 +363,31 @@
}
sigaddset(&t->signal, sig);
- if (!sigismember(&t->blocked, sig))
+ if (!sigismember(&t->blocked, sig)) {
t->sigpending = 1;
+#ifdef __SMP__
+ //
+ // if the task is running on a different CPU
+ // force a reschedule on the other CPU - note that
+ // the code below is a tad loose and might occasionally
+ // wake the wrong CPU if we catch the process in the
+ // process of changing - but no harm is done by that
+ // other than forcing a rechedule of the wrong process
+ //
+ {
+ unsigned long flags; //pwc
+ extern spinlock_t runqueue_lock; //pwc
+ //pwc
+ spin_lock_irqsave(&runqueue_lock, flags); //pwc
+ if (t->has_cpu && t->processor != smp_processor_id()) { //pwc
+ t->need_resched = 1; //pwc
+ smp_send_reschedule(t->processor); //pwc
+ } //pwc
+ spin_unlock_irqrestore(&runqueue_lock, flags); //pwc
+ } //pwc
+
+#endif /* __SMP__ */
+ }
out:
spin_unlock_irqrestore(&t->sigmask_lock, flags);
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
