Re: [PATCH -tip] kgdb, x86: Pull up NMI notifier handler priority

[Cyrill Gorcunov]

On 03/24/2011 12:16 AM, Don Zickus wrote:
> On Wed, Mar 23, 2011 at 11:32:33PM +0300, Cyrill Gorcunov wrote:
>> kgdb needs IPI to be sent and handled before perf
>> or anything else NMI, otherwise kgdb hangs with bootup
>> self-tests (found on P4 HT SMP machine). Raise its priority
>> so that we're called first in a notifier chain.
> 
> This is only because P4 perf swallows all the nmis.  If that is the case
> you are arguing to make the perf nmi at the bottom of the priority list,
> which is probably not where it should be due to its volume.

The problem is that there IPI wait cycle inside kgdb and we are to be sure
to handle it early. And perf eventually can consume kgdb NMI which would
lead to infinite wait loop so I don't see any easier way to deal with it.

> 
> I am stuck debugging P4 problems again for RHEL-6 and I noticed a small
> change that is needed (didn't help my problem though) but it looked like
> an oversight that might help your case.
> 
> Cheers,
> Don
> 
> 
> diff --git a/arch/x86/kernel/cpu/perf_event_p4.c b/arch/x86/kernel/cpu/perf_event_p4.c
> index 3769ac8..d945314 100644
> --- a/arch/x86/kernel/cpu/perf_event_p4.c
> +++ b/arch/x86/kernel/cpu/perf_event_p4.c
> @@ -777,6 +787,7 @@ static inline int p4_pmu_clear_cccr_ovf(struct hw_perf_event *hwc)
>  	 * the counter has reached zero value and continued counting before
>  	 * real NMI signal was received:
>  	 */
> +	rdmsrl(hwc->event_base, v);
>  	if (!(v & ARCH_P4_UNFLAGGED_BIT))
>  		return 1;
>  

Good catch! Ack! It seems to be escaped in first place (I fear I forgot to
refresh patch before send it). Mind to send the full patch to Ingo?

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