tags 604604 + moreinfo
thanks

23.11.2010 01:20, Jakub Wilk wrote:
> Package: qemu-kvm
> Version: 0.13.0+dfsg-2
> Severity: grave
> Justification: renders package unusable
> 
> After 0.12.5+dfsg-5 -> 0.13.0+dfsg-2 upgrade kvm doesn't start anymore.
> I get an error immediately:
> 
> $ kvm
> kvm: vm entry failed with error 0xffffffff
> 
> kvm_run returned -22

> model name    : AMD Athlon(tm) 64 X2 Dual Core Processor 4400+
> Architecture: i386 (i686)
> Kernel: Linux 2.6.32-5-686 (SMP w/2 CPU cores)
> linux-image-2.6.32-5-686 [lin 2.6.32-27  Linux 2.6.32 for modern PCs

This looks like a kernel problem which is fixed by
a recent -stable upstream kernel (2.6.32.26):

>From [email protected]  Fri Oct 29 15:18:43 2010
From: Joerg Roedel <[email protected]>
Date: Thu, 28 Oct 2010 16:48:09 -0200
Subject: KVM: SVM: Fix wrong intercept masks on 32 bit
To: [email protected]
Cc: [email protected], [email protected], [email protected], Jan Kiszka 
<[email protected]>, Gleb Natapov <[email protected]>, Joerg Roedel 
<[email protected]>
Message-ID: <[email protected]>
Content-Disposition: inline; 
filename=0001-KVM-SVM-Fix-wrong-intercept-masks-on-32-bit.patch

From: Joerg Roedel <[email protected]>

commit 061e2fd16863009c8005b4b5fdfb75c7215c0b99 upstream.

This patch makes KVM on 32 bit SVM working again by
correcting the masks used for iret interception. With the
wrong masks the upper 32 bits of the intercepts are masked
out which leaves vmrun unintercepted. This is not legal on
svm and the vmrun fails.
Bug was introduced by commits 95ba827313 and 3cfc3092.

Cc: Jan Kiszka <[email protected]>
Cc: Gleb Natapov <[email protected]>
Signed-off-by: Joerg Roedel <[email protected]>
Signed-off-by: Avi Kivity <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

---
 arch/x86/kvm/svm.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)


The problem case happens on 32bit kernel only and does
not happen on 64bit.  Please try with amd64 kernel
(linux-image-2.6.32-5-amd64) - it should just work.

Note that almost no one tests kvm on 32bits nowadays,
consider running 64bit kernel with either 32 or 64
bit userspace.  I run 64bit kernel and 32bit userspace
for a long time and now finally I can say most probs
in this configuration has been fixed - for more than
a year I haven't encountered one :)

And please also consider switching to AthlonII - your
processor lacks npt (nested page tables) support and
TSC syncronisation, both affects performance and stability
of kvm guest in significant way.  Not related to this
bugreport, just a general suggestion.

Tagging as 'moreinfo' for now, if the problem turns
out to be what I think it is, I'll reassign it to
kernel-image and downgrade to normal.

Thanks!

/mjt



-- 
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]

Reply via email to