在 2010-01-21四的 11:57 +0800，Li Yi写道： > On Wed, 2010-01-20 at 05:56 -0500, [email protected] wrote: > > > > --- trunk/arch/blackfin/kernel/time-ts.c 2010-01-20 07:25:31 UTC (rev > > 8203) > > +++ trunk/arch/blackfin/kernel/time-ts.c 2010-01-20 10:56:24 UTC (rev > > 8204) > > @@ -21,6 +21,7 @@ > > #include <asm/blackfin.h> > > #include <asm/time.h> > > #include <asm/gptimers.h> > > +#include <asm/nmi.h> > > > > /* Accelerators for sched_clock() > > * convert from cycles(64bits) => nanoseconds (64bits) > > @@ -313,6 +314,9 @@ > > > > smp_mb(); > > evt->event_handler(evt); > > + > > + check_nmi_watchdog(cpu); > > + > > return IRQ_HANDLED; > > } > > > Is timer interrupt handler a good place to touch watchdog? In High > resolution mode, timer irq may be triggered frequently (e.g, 0.1 ms), > and check_nmi_watchdog() would be a heavy overhead. I agree.
> > It seems the periodic tick handler (i.e, jiffies) is a better place, but > I am not sure how to add it without changing common code. Maybe need more investigation. > > And, include/linux/nmi.h defines an API touch_nmi_watchdog(), it can be > used by common code. I think check_nmi_watchdog() should be renamed to > touch_nmi_watchdog()? There is some difference in implementing nmi watchdog on bf561 compare to x86. On x86, the touch_nmi_watchdog() only set a per-cpu variable named touched. The nmi handler, do_nmi(), will periodical(5 sec) check this variable and reset the watchdog. On bf561, all above work is done by check_nmi_watchdog(). The do_nmi() will never be called until there is a lockup. Currently, the nmi handler need to save/restore all core registers, turn off/on D-cache, (I-cache seems not necessary), so I think its a heavy overhead. If every one think its acceptable to do_nmi() every 5 seconds, I will change this behavior. -Graff > > -Yi > > > > _______________________________________________ Linux-kernel-commits mailing list [email protected] https://blackfin.uclinux.org/mailman/listinfo/linux-kernel-commits
