Re: [Xen-users] Lots of IOAPIC or GSI related errors

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Mon, 2013-01-07 at 15:13 +0000, Sven KÃhler wrote:
> Am 07.01.2013 12:23, schrieb Ian Campbell:
> > On Sun, 2013-01-06 at 07:34 +0000, Sven KÃhler wrote:
> >> Hi,
> >>
> >> I'm getting lots of IOAPIC and GSI related errors, both in xen's and the
> >> dom0 kernel's dmesg (see below).
> >>
> >> It might have to do with the fact, that I added "acpi=ht" to my xen
> >> command line. Previously to adding that, the errors didn't show up. But
> >> I added it, because it was said to fix my reboot problems. But along
> >> with adding acpi=ht, I also upgraded dom0 to kernel 3.7.1.
> > 
> > Any chance you could separate those two changes and isolatethe one which
> > causes the problem?
> 
> It's a remote server. I'm pretty sure I'm already annoying the employees
> of my hoster enough, as the machine enters some weired state during
> reboot, from which it does never recover - even pushing the reset button
> (which I can do remotely) doesn't work. Hence the hoster's employees
> always have to turn the machine off and on again.
> 
> > The docs for acpi=ht say:
> >         Using `acpi=ht` causes Xen to parse the ACPI tables enough to
> >         enumerate all CPUs, but will not use other ACPI features.  This is 
> > not
> >         common, and only has an effect if your system is blacklisted.
> > 
> > I doubt this is what you want for a reboot issue.
> > 
> > You should probably investigate the reboot= options instead.
> 
> Did that already. Didn't help.
> 
> The problem is actually known to the hoster. They recommended acpi=ht.
> The hoster blamed ACPI and said that disabling ACPI would fix the
> problem. However, since I added acpi=ht, I haven't rebootet the machine yet.
> 
> 
> Do you have any clue what these APIC/GSI error are all about? I mean,
> even if you can't explain why they happen, what do they actually mean?

GSIs are interrupts, the ACPI tables contain mappings from interrupts to
APICs, which are the interrupt controllers. The message is basically
saying it can't determine which APIC is responsible for the given
interrupt -- since you've explicitly asked Xen not to use ACPI for
anything other than hyperthreading that isn't unexpected.

I think there is an older pre-ACPI table which is supposed to also
contains a subset of this mapping, but they are often incomplete or
buggy on systems which supports ACPI (or more specifically, where
Windows would use ACPI). That's getting a bit outside the stuff I can
recall off the top of my head though I'm afraid.

Turning off ACPI on any half-way modern system is generally not going to
result in everything working properly.

Ian.



_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users