Re: ath5k: kernel timing screwed - due to unserialised register access?

[Thomas Gleixner <tglx@linutronix.de>]

On Tue, 7 Oct 2008, Thomas Gleixner wrote:
> On Tue, 7 Oct 2008, Elias Oltmanns wrote:
> > Thomas Gleixner <tglx@linutronix.de> wrote:
> > > On Mon, 6 Oct 2008, Elias Oltmanns wrote:
> > >> Make sure that event1 is the right device. chktimer usually reports
> > >> several premature timer expiries in less than a minute.
> > [...]
> > > Your measuring method is wrong. You really want to measure the delta
> > > of the timer events in the kernel via ktime_get(), not the delta of
> > > something else in userspace.
> > 
> > Alright, here is a stripped down version of the test case. This time,
> > you only need to load the timer-test module and start up the ath5k
> > interface. The glitch is triggered slightly less reliably, but I can
> 
> -ENOATH5KHARDWARE
> 
> > still easily verify that the problem is present when running 2.6.27-rc9
> > on my system.
> 
> Hmm. Can you please add some real info to the printk :)
> 
> > +	if (timespec_to_ns(&diff) < TSTM_THRESH)
> > +		printk(KERN_INFO "Timer expired prematurely.\n");
> 
> e.g.:
> 
>   		printk(KERN_INFO "b: %6ld.%09ld n: %6ld.09ld e: %lu j: %lu\n",
> 			before.tv_sec, before.tv_nsec, now.tv_sec, now.tv_nsec,
> 			tstm_timer.expires, jiffies);
> 
> Does this happen w/o the ath5k driver as well ?

Some more questions:

Does this happen with any on the combinations of highres/nohz
enabled/disabled ?

Can you please add:

    sysrq_timer_list_show();

    after the printk. You have to export it in kernel/time/timer_list.c
    to use it with a module

Please do cat /proc/timer_list right before you start the module as well.

Thanks,

	tglx