Re: [PATCH 1/5] HWPOISON: define VM_FAULT_HWPOISON to 0 whenfeature is disabled

From: Alan Cox
Date: Fri Jun 12 2009 - 14:08:02 EST


> HWPOISON is a reliability enabling feature - if it enables prevalent
> of crappy hardwares, let's celebrate changing the world~~

HWPOISON is not in the most part a reliability enabling feature. Nothing
of the sort.

The existing behaviour is that your system goes kerblam on such a serious
error. The replacement behaviour is that bits of your machine go kerblam
in unpredictable ways.

In both cases you actually improve your reliability with clustering and
failover.

There are a few special cases its potentially useful - lots of VMs being
one where you have some chance of a controlled failure of a bounded
system. But even in that case I know if I was admin my scripts would read

if (hwpoison_error)
migrate_all_guests()
mail admin
schedule replacement of the machine

I'm not actually sure teaching hwpoison to handle anything but losing
entire guest OS systems is useful.
--
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/