vmwrite error after upgrade to pve-kernel-2.6.32-22-pve

[hugo]

Hi all,

I've just updated my proxmox servers with the latest updates. This included the update of the kernel from pve-kernel-2.6.32-20-pve to pve-kernel-2.6.32-22-pve.
At 2 of the 4 nodes this kernel message is logged when starting an vm on the node (or migrating to).

Aug 5 09:39:44 lbk-gc-pve02 kernel: vmwrite error: reg 401e value a06f4bc0 (err 12)
Aug 5 09:39:44 lbk-gc-pve02 kernel: Pid: 185672, comm: kvm veid: 0 Not tainted 2.6.32-22-pve #1
Aug 5 09:39:44 lbk-gc-pve02 kernel: Call Trace:
Aug 5 09:39:44 lbk-gc-pve02 kernel: [<ffffffffa06f306b>] ? vmwrite_error+0x2c/0x2e [kvm_intel]
Aug 5 09:39:44 lbk-gc-pve02 kernel: [<ffffffffa06ec4d0>] ? vmcs_writel+0x20/0x30 [kvm_intel]
Aug 5 09:39:44 lbk-gc-pve02 kernel: [<ffffffffa06ecdc3>] ? vmx_cpuid_update+0x53/0x100 [kvm_intel]
Aug 5 09:39:44 lbk-gc-pve02 kernel: [<ffffffffa0697ea7>] ? kvm_arch_vcpu_ioctl+0x9f7/0x12a0 [kvm]
Aug 5 09:39:44 lbk-gc-pve02 kernel: [<ffffffff8115fc69>] ? handle_mm_fault+0x239/0x310
Aug 5 09:39:44 lbk-gc-pve02 kernel: [<ffffffffa068468c>] ? kvm_vcpu_ioctl+0x8c/0x580 [kvm]
Aug 5 09:39:44 lbk-gc-pve02 kernel: [<ffffffffa0688850>] ? kvm_dev_ioctl+0x50/0x4c0 [kvm]
Aug 5 09:39:44 lbk-gc-pve02 kernel: [<ffffffff811b352a>] ? vfs_ioctl+0x2a/0xa0
Aug 5 09:39:44 lbk-gc-pve02 kernel: [<ffffffff811b3b5e>] ? do_vfs_ioctl+0x7e/0x570
Aug 5 09:39:44 lbk-gc-pve02 kernel: [<ffffffff811b409f>] ? sys_ioctl+0x4f/0x80
Aug 5 09:39:44 lbk-gc-pve02 kernel: [<ffffffff8100b182>] ? system_call_fastpath+0x16/0x1b

When booting with the previous kernel, pve-kernel-2.6.32-20-pve, no errors occur.
Although, even the error occurs, the vm's run fine.

The 4 nodes are not identical hardware.
The nodes with the errors have (older) Xeon 5130 cpu's. The nodes without has (less older) Xeon E5420 cpu's.

I found this: http://git.kernel.org/cgit/linux/ke.../?id=29282fde80d44e587f8c152b10049a56e61659f0
Which seems to be the exact issue. Will this patch be included in future pve kernel versions?

Regards,
Hugo

 

[tom]

I also see this on one of our old servers, with Intel Xeon 3050, but also all VMs are running without issues on this box.

 

[Oer2001]

I also see this on one of our old servers, with Intel Xeon 3050, but also all VMs are running without issues on this box.

you are right. But online migration to this host does not work for me.
Also kernel 2.6.32-23-pve is affected.

Regards,
Oer

 

[dduh]

Hi!
Similar. 2 X Intel Xeon 5130. Kernel 2.6.32-23-pve

kernel: vmwrite error: reg 401e value a05f6ca0 (err 12)
kernel: Pid: 18154, comm: kvm veid: 0 Not tainted 2.6.32-23-pve #1
kernel: Call Trace:
kernel: [<ffffffffa05f510b>] ? vmwrite_error+0x2c/0x2e [kvm_intel]
kernel: [<ffffffffa05ee4d0>] ? vmcs_writel+0x20/0x30 [kvm_intel]
kernel: [<ffffffffa05eed03>] ? vmx_cpuid_update+0x53/0x100 [kvm_intel]
kernel: [<ffffffffa0599f17>] ? kvm_arch_vcpu_ioctl+0x9f7/0x12a0 [kvm]
kernel: [<ffffffff8116fd8c>] ? __vunmap.isra.16+0x9c/0x120
kernel: [<ffffffffa058668c>] ? kvm_vcpu_ioctl+0x8c/0x580 [kvm]
kernel: [<ffffffffa058a7e0>] ? kvm_dev_ioctl+0x50/0x4c0 [kvm]
kernel: [<ffffffff811b379a>] ? vfs_ioctl+0x2a/0xa0
kernel: [<ffffffff811b3dce>] ? do_vfs_ioctl+0x7e/0x570
kernel: [<ffffffff811b430f>] ? sys_ioctl+0x4f/0x80
kernel: [<ffffffff8100b182>] ? system_call_fastpath+0x16/0x1b

Will not problems?

Maybe at older hardware, go back to version 2.3?

 

[hugo]

I did not encouter any issues and therefore ignoring these messages. Also live migration works without issues.
The patch mentioned on: http://git.kernel.org/cgit/linux/ke.../?id=29282fde80d44e587f8c152b10049a56e61659f0 seems to fix the issue (not tested myself). However, this patch does not seemed to be applied in the proxmox kernel.
Is it possible to include this patch in the pve kernels?

 

[dietmar]

Is it possible to include this patch in the pve kernels?

Will include that in the next upload.

 

[tom]

its now uploaded to pve-no-subscription repo.