On Tue, Mar 23, 2010 at 05:33:38PM -0400, Don Zickus wrote:
quoted text
> +/* Callback function for perf event subsystem */
> +void watchdog_overflow_callback(struct perf_event *event, int nmi,
> +		 struct perf_sample_data *data,
> +		 struct pt_regs *regs)
> +{
> +	int this_cpu = smp_processor_id();
> +	unsigned long touch_ts = per_cpu(watchdog_touch_ts, this_cpu);
> +	int duration;
> +
> +	if (touch_ts == 0) {
> +		__touch_watchdog();
> +		return;
> +	}
> +
> +	/* check for a hardlockup
> +	 * This is done by making sure our timer interrupt
> +	 * is incrementing.  The timer interrupt should have
> +	 * fired multiple times before we overflow'd.  If it hasn't
> +	 * then this is a good indication the cpu is stuck
> +	 */
> +	if (is_hardlockup(this_cpu)) {
> +		if (hardlockup_panic)
> +			panic("Watchdog detected hard LOCKUP on cpu %d", this_cpu);
> +		else
> +			WARN(1, "Watchdog detected hard LOCKUP on cpu %d", this_cpu);



panic is going to endless loop so this is fine.
But if you only warn, the path continues, and if you have a
hardlockup then you also have a softlockup, which will probably
warn in turn, making the hardlockup report to vanish. But if
any hardlockup, its report is much more important as it is the real point,
a softlockup that follows is only a consequence of the hardlockup.

Btw, you don't have any cpumask that keeps track of the cpus
that have warned already?


quoted text
> +static int watchdog_enable(int cpu)
> +{
> +	struct perf_event_attr *wd_attr;
> +	struct perf_event *event = per_cpu(watchdog_ev, cpu);
> +	struct task_struct *p = per_cpu(softlockup_watchdog, cpu);
> +
> +	/* is it already setup and enabled? */
> +	if (event && event->state > PERF_EVENT_STATE_OFF)
> +		goto out;
> +
> +	/* it is setup but not enabled */
> +	if (event != NULL)
> +		goto out_enable;
> +
> +	/* Try to register using hardware perf events first */
> +	wd_attr = &wd_hw_attr;
> +	wd_attr->sample_period = hw_nmi_get_sample_period();
> +	event = perf_event_create_kernel_counter(wd_attr, cpu, -1, watchdog_overflow_callback);
> +	if (!IS_ERR(event)) {
> +		printk(KERN_INFO "NMI watchdog enabled, takes one hw-pmu counter.\n");
> +		goto out_save;
> +	}
> +
> +	/* hardware doesn't exist or not supported, fallback to software events */
> +	printk(KERN_INFO "NMI watchdog: hardware not available, trying software events\n");
> +	wd_attr = &wd_sw_attr;
> +	wd_attr->sample_period = softlockup_thresh * NSEC_PER_SEC;
> +	event = perf_event_create_kernel_counter(wd_attr, cpu, -1, watchdog_overflow_callback);



I fear the cpu clock is not going to help you detecting any hard lockups.
If you're stuck in an interrupt or an irq disabled loop, your cpu clock is
not going to fire.

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