Re: NOHZ: local_softirq_pending 08

[Thomas Gleixner <tglx@linutronix.de>]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 1059 bytes --]

On Fri, 8 Jun 2012, Dave Jones wrote:

> On Fri, Jun 08, 2012 at 09:57:03AM +0200, Thomas Gleixner wrote:
>  > On Thu, 7 Jun 2012, Dave Jones wrote:
>  > > 
>  > > The log is still huge though. (68M uncompressed) at
>  > > http://www.codemonkey.org.uk/junk/r8169-trace.txt.xz
>  > 
>  > Following that URL gives me a nice picture.
>  > 
>  > This is somewhat embarrassing, isn’t it? :)
> 
> indeed ;-) Should be fixed.

I can't find the point where the warning is issued, but I think I
found the cause of the problem.

static void rtl_slow_event_work(struct rtl8169_private *tp)
{
	.....
	napi_schedule(&tp->napi);
	  --> __napi_schedule();
	      -->	list_add_tail(&napi->poll_list, &sd->poll_list);
	              __raise_softirq_irqoff(NET_RX_SOFTIRQ);

		      This merily sets the softirq bit.


So this code is really wrong. It's called from full preemptible
context of the workqueue. And if the next thing is a context switch to
idle then the pending softirq check will trigger.

I let the network folks sort out the proper solution.

Thanks,

	tglx