On Tue, Jun 13, 2017 at 10:19:47AM +0200, Alexander Stein wrote:
> On Tuesday 13 June 2017 15:58:45, Dong Aisheng wrote:
> > diff --git a/drivers/clocksource/timer-imx-tpm.c
> > b/drivers/clocksource/timer-imx-tpm.c new file mode 100644
> > index 0000000..940a4f75
> > --- /dev/null
> > +++ b/drivers/clocksource/timer-imx-tpm.c
> > @@ -0,0 +1,227 @@
> > [...]
> > +static int tpm_set_next_event(unsigned long delta,
> > + struct clock_event_device *evt)
> > +{
> > + unsigned long next, now;
> > +
> > + next = readl(timer_base + TPM_CNT) + delta;
> > + writel(next, timer_base + TPM_C0V);
> > + now = readl(timer_base + TPM_CNT);
>
> What about:
> > now = readl(timer_base + TPM_CNT);
> > next = now + delta;
> > writel(next, timer_base + TPM_C0V);
> > return 0;
>
> > + return (int)((next - now) <= 0) ? -ETIME : 0;
>
> Can this error actually happen, even with your implementation?
Yes, i did observe some -ETIME when testing with nanosleep or
hrtimer during programing min_delta event, especially when system
is high loading. e.g. run GPU stress test.
Please also refer to another mail for details i just replied to Daniel.
Regards
Dong Aisheng
