Re: [PATCH] perf, x86: Add hw_watchdog_set_attr() in a sake of nmi-watchdog on P4

[Cyrill Gorcunov <gorcunov@gmail.com>]

On Tue, Jun 28, 2011 at 05:53:31PM +0200, Stephane Eranian wrote:
> Cyrill,
> 
> #ifdef CONFIG_HARDLOCKUP_DETECTOR
> +void __weak hw_nmi_watchdog_set_attr(struct perf_event_attr *wd_attr) { }
> 
> The weak function has to remain inside the #ifdef. Remember it's just an
> override in case you cannot using the generic cycle event. It is only needed
> when you have the HARDLOCK detector, i.e., are using the PMU to detect
> deadlocks.
> 
> I suspect P4 may be the only one so far which exhibited a problem there.
> 

heh ;) Guys, maybe we're talking about different things? Don, when you said
"Though I do wonder about this call in the watchdog" you meant the 

 	wd_attr->sample_period = hw_nmi_get_sample_period(watchdog_thresh);
-->	hw_nmi_watchdog_set_attr(wd_attr);

itself? Ie you suspect some different point where to call it?

When I said not a "best place" I meant about __weak function bare implementation
placed that near to call (which is looked somehow suspicious for me from overall
code structure), but I didn't mean the call sequence itself ;)

	Cyrill