Re: [patch 2/4] s390: fix preempt_count of idle thread with cpu hotplug

[Heiko Carstens]

> >Set preempt_count of idle_thread to zero before switching off cpu.
> >Otherwise the preempt_count will be wrong if the cpu is switched on again
> >since the thread will be reused.
> 
> I had a similar discussion back in November, that one about
> /proc/interrupts stats.  Rather than do that all over again below is a cut
> and paste of my reply to that discussion.  The executive summary is I 
> rather like the current behavior as is.
> 
> -------------------------------------------------------------
> 
> > When CPU2 is off-lined, the statistics for CPU2 do not appear
> >(expected). However when you look at the before picture (all CPUs
> >present) and after picture (all cpus present after CPU2 re-added), you
> >see that the original data was returned and has incremented:

Think we're talking about different things. preempt_count is not statistical
stuff. Actually if you have preempt_count != 0 preemption is disabled.
At least on s390 we didn't set this counter to 0 in case of cpu hotplug thus
ending with a preempt_count > 0 if a cpu gets reenabled. Which gives us a lot
of warnings ala "scheduling while atomic: swapper/0x00000001/0".
So this is not a restriction that should be documented somewhere but a bug.

No idea how powerpc deals with this and if there is the same issue. Just
tried to give a hint since cpu hotplug code for powerpc and s390 looks quite
the same.

Heiko
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/