On 20.01.2009, at 21:07, Avi Kivity <avi@xxxxxxxxxx> wrote:
Alexander Graf wrote:
Hi list,
recently I've been hitting some KVM bugs others seem to have
reported as
well, including
- CIFS timeouts
- Stuck ?? errors
- Random segmentation faults in the guest
so I figured, I'll put together a stress test that can be used to
reproduce these issues. This is done by using a CIFS mount on the
host
and unpacking data from that mount to the mount. I have been able to
bring kvm down to its knees a lot just by doing this.
Simply run the test in an endless-loop. FWIW enabling NPT helps
triggering the issue.
Are the problems specific to AMD?
I don't know, as all machines I tried it on were AMD so far. But
judging from user reports on the ml, it happens on Intel too.
What does "helps triggering" mean - does it happen with NPT disabled?
It seems like the chances for breakage are higher with NPT enabled. I
do see them without as well though.
The guest kernels included here are openSUSE 11.0 (2.6.25) and 11.1
(2.6.27) kernels.
Find the tests here: http://alex.csgraf.de/kvm-tests.tar.bz2
And some logs here (NPT enabled): http://alex.csgraf.de/kvm-logs.tar.bz2
I'm somewhat lost on the reason for these failures, so if you do have
some time on your hands, please give me a hand debugging this! If I'd
had to guess, I'd say it's either an APIC issue and/or guest memory
corruption.
I'd guess memory corruption.
Does running a uniprocessor guest help? What about a uniprocessor
guest pinned to one host core?
I'll try to start tests tomorrow.
Alex
--
Do not meddle in the internals of kernels, for they are subtle and
quick to panic.
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
