Re: EPT: Misconfiguration

[Avi Kivity <avi@redhat.com>]

On 01/25/2011 08:29 PM, Ruben Kerkhof wrote:
> >  When you say "suddenly", this was with no changes to software and hardware?
>
> The host software and hardware hasn't changed in the two months since
> the machine has been running. 2.6.34.7 kernel and qemu-kvm 0.13.
>
> We host customer vms on it though, so virtual machines come and go.
> Various operating systems, a mixture of Linux, FreeBSD and Windows
> 2008 R2. We have other machines with the same config without these
> problems though.

Are those other machines running a similar workload?

The traces look awfully like bad hardware, though that can also be 
explained by random memory corruption due to a bug.

> This time I have a few different messages though:
>
> 2011-01-25T11:58:50.001208+01:00 phy005 kernel: general protection fault: 0000 [#1] SMP
>
> RSI: 0000000000000000 RDI: 1603a07305001568
>
> 2011-01-25T11:58:50.001486+01:00 phy005 kernel: Code: ff ff 41 8b 46
> 08 41 29 06 4c 89 e7 57 9d 0f 1f 44 00 00 48 83 c4 18 5b 41 5c 41 5d
> 41 5e 41 5f c9 c3 55 48 89 e5 0f 1f 44 00 00<f0>  ff 4f 08 0f 94 c0 84
> c0 74 10 85 f6 75 07 e8 63 fe ff ff eb

lock decl 0x8(%rdi)

%rdi is completely crap, looks like corruption again.  Strangely, it is 
similar to the bad spte from the previous trace: 0x1603a0730500d277.  
The upper 48 bits are identical, the lower 16 bits are different.:
> 2011-01-25T12:06:32.673937+01:00 phy005 kernel: qemu-kvm: Corrupted
> page table at address 7f37b37ff000
> 2011-01-25T12:06:32.673959+01:00 phy005 kernel: PGD c201d1067 PUD
> 94e538067 PMD 61e5bf067 PTE 1603a0730500e067

Here are those magic 48 bits again, in the PTE entry.
> 2011-01-25T12:38:49.416943+01:00 phy005 kernel: EPT: Misconfiguration.
> 2011-01-25T12:38:49.417518+01:00 phy005 kernel: EPT: GPA: 0x2abff038
> 2011-01-25T12:38:49.417526+01:00 phy005 kernel:
> ept_misconfig_inspect_spte: spte 0x5f49e9007 level 4
> 2011-01-25T12:38:49.417532+01:00 phy005 kernel:
> ept_misconfig_inspect_spte: spte 0x5db595007 level 3
> 2011-01-25T12:38:49.417553+01:00 phy005 kernel:
> ept_misconfig_inspect_spte: spte 0x5d5da7007 level 2
> 2011-01-25T12:38:49.417558+01:00 phy005 kernel:
> ept_misconfig_inspect_spte: spte 0x1603a07305006277 level 1

Again.

> 2011-01-25T13:16:58.192440+01:00 phy005 kernel: BUG: Bad page map in
> process qemu-kvm  pte:1603a0730500d067 pmd:61059f067

Again.

However, these all came from a single boot, yes?  If so they can be the 
same corruption.  Please collect more traces, with reboots in between.

-- 
error compiling committee.c: too many arguments to function