Re: [PATCH] nmi_watchdog=2 and P4-HT

[Andrew Morton]

Philippe Elie <phil.el@xxxxxxxxxx> wrote:
>
> Hi,
> 
> Actually with nmi_watchdog=2 and a P4 ht box the nmi is reflected
> only on logical processor 0, it's better to get it on both.

What do you mean by "reflected"?  That the NMi is delivered to both
siblings but only appears in /proc/interrupts as being delivered to the
zeroeth?

> Note, if you test this patch, than on all x86 SMP and nmi_watchdog=2
> nmi occurs at 1000 hz (if the cpu is loaded) not at the intended 1 hz
> rate but that's a distinct problem.

nmi_watchdog=2 is local apic, and nmi_watchdog=1 is I/O apic, is that
correct?

I am showing the current behaviour:


nmi_watchdog=1: 1000 NMI/second, accounted to both siblings
nmi_watchdog=2: one NMI/second, accounted to sibling 0 only.

with your patch:

nmi_watchdog=1: 1000 NMI/second, accounted to both siblings
nmi_watchdog=2: 1000 NMI/second, accounted to both siblings

All of these are wrong, aren't they?  We'd like to see one NMI per second,
on all siblings.  I gues that's not possible for the IO APIC?

>From the above it appears that you have a solution planned for the local
APIC at least, yes?
-
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/