Red Hat Bugzilla – Bug 656984
TSC offset of virtual machines is not initialized correctly by 'kvm_amd' kernel module.
Last modified: 2013-01-10 22:36:16 EST
This bug has been copied from bug #642659 and has been proposed
to be backported to 5.5 z-stream (EUS).
Using the bz642659 comment26 test on kvm-83-164.el5_5.29,failed.
---snip of output---
crash> px ((struct vcpu_svm *)0xffff810133e841c0)->vcpu.arch.host_tsc
$9 = 0x1322fe6d16801
crash> px ((struct vcpu_svm *)0xffff810133e841c0)->vmcb.control.tsc_offset
$8 = 0xfffd9ba044aaf031
crash> px 0-0xfffd9ba044aaf031
$10 = 0x2645fbb550fcf
crash> pd 0x1322fe6d16801-0x2645fbb550fcf
$11 = -336655986960334
-----The difference between host TSC and guest TSC offset is
Set this issue status as assigned.
Fixes have not been ported to 5.5.z yet, so failure is expected.
(In reply to comment #8)
> Fixes have not been ported to 5.5.z yet, so failure is expected.
Which fixes Zach? comment #4 states that the fixes are there. Are there more?
(In reply to comment #9)
> (In reply to comment #8)
> > Fixes have not been ported to 5.5.z yet, so failure is expected.
> Which fixes Zach? comment #4 states that the fixes are there. Are there more?
I checked the 5.5.z git tree and no changes had been ported. I haven't worked with z-stream before so I'm not 100% familiar with the process.
Now I can't access git at all.
[zamsden@mysore rhel5.5.z]$ git pull
fatal: The remote end hung up unexpectedly
On second thought, perhaps I'm a bit confused. The changes would be to the KVM module, which isn't part of the 5.5.z git tree. In fact, I don't know where the KVM tree corresponding to 5.5.z lives. Is there a kvm-rhel5.5.z.git tree to which these should be ported?
Also, there were two patches, the first required a correction. Perhaps one of them is missing.
I'll check as soon as git access is restored.
Okay, it looks as though the kvm-rhel5.5.z kmod is built from the branch rhel55/next in the kvm-rhel5 git tree, and that Eduardo just pulled the second fix into that branch. All should be well now.
Verified on kvm-83-164.el5_5.30,passed.
---snip of output---
crash> px ((struct vcpu_svm *)0xffff81011e19c040)->vcpu.arch.host_tsc
$6 = 0x1ef3ea5660f29
crash> px ((struct vcpu_svm *)0xffff81011e19c040)->vmcb.control.tsc_offset
$7 = 0xfffe10c16374676f
crash> px 0-0xfffe10c16374676f
$8 = 0x1ef3e9c8b9891
crash> pd 0x1ef3ea5660f29-0x1ef3e9c8b9891
$9 = 148534936
The difference between host TSC and guest TSC offset
is only 148534936 cycles.
An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on therefore solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.