[Qemu-devel] VM stops responding, 100% CPU, switching to monitor fails

[Qemu-devel] VM stops responding, 100% CPU, switching to monitor fails

[Steve Fosdick]

I am currently experiencing an issue where the guest VM stops
responding, attempting to switch to the monitor does not do anything
either and the host CPU usage shows as 100% system.

So far the only solution I have found is to kill the qemu process.  A
normal SIGTERM signal seems to be sufficient.

This is with kernel 2.6.28.2, kqmeu 1.3.0~pre11 and qemu 0.9.1.

Do you have any tips on what I can do to get to the bottom of this?

I am thinking as the CPU usage is all system time that there maybe an
infinite loop running within kqemu though I would not know if the loop
was in code that is part of kqemu itself or the guest OS.  If the guest
had got stuck in an infinite loop though I would have hoped the monitor
would still work.

Regards,
Steve.

Re: [Qemu-devel] VM stops responding, 100% CPU, switching to monitor fails

[Nathan]

Hi Steve,

This may be the issue I just posted about.  To test, in the host, run
strace on the qemu process.

If the output shows many SIGALRM signals between fewer calls to
"clone" which return ERESTARTNOINTR, then it's the same issue and I
can advise you.

Otherwise, I'm not much help (I'm new to qemu).

regards,
Nathan Wilcox


On Tue, Feb 3, 2009 at 5:10 AM, Steve Fosdick
<lists@pelvoux.nildram.co.uk> wrote:
> I am currently experiencing an issue where the guest VM stops
> responding, attempting to switch to the monitor does not do anything
> either and the host CPU usage shows as 100% system.
>
> So far the only solution I have found is to kill the qemu process.  A
> normal SIGTERM signal seems to be sufficient.
>
> This is with kernel 2.6.28.2, kqmeu 1.3.0~pre11 and qemu 0.9.1.
>
> Do you have any tips on what I can do to get to the bottom of this?
>
> I am thinking as the CPU usage is all system time that there maybe an
> infinite loop running within kqemu though I would not know if the loop
> was in code that is part of kqemu itself or the guest OS.  If the guest
> had got stuck in an infinite loop though I would have hoped the monitor
> would still work.
>
> Regards,
> Steve.
>
>
>

Re: [Qemu-devel] VM stops responding, 100% CPU, switching to monitor fails

[Steve Fosdick]

On Tue, 2009-02-03 at 09:11 -0800, Nathan wrote:
> Hi Steve,
> 
> This may be the issue I just posted about.  To test, in the host, run
> strace on the qemu process.
> 
> If the output shows many SIGALRM signals between fewer calls to
> "clone" which return ERESTARTNOINTR, then it's the same issue and I
> can advise you.
> 
> Otherwise, I'm not much help (I'm new to qemu).
> 
> regards,
> Nathan Wilcox

Interestingly if I attach to the qemu process with strace the guest VM
starts working again and continues to work after I have stopped the
strace command with control-C.

Here is a sample of what strace was giving:

--- SIGALRM (Alarm clock) @ 0 (0) ---
rt_sigreturn(0x63d24020)                = 0
select(16, [11 14 15], [], [], {0, 0})  = 0 (Timeout)
gettimeofday({1233684242, 162854}, NULL) = 0
clock_gettime(CLOCK_MONOTONIC, {87714, 363355675}) = 0
clock_gettime(CLOCK_MONOTONIC, {87714, 363373833}) = 0
clock_gettime(CLOCK_MONOTONIC, {87714, 363390595}) = 0
clock_gettime(CLOCK_MONOTONIC, {87714, 363406519}) = 0
clock_gettime(CLOCK_MONOTONIC, {87714, 363421884}) = 0
timer_gettime(0, {it_interval={0, 0}, it_value={0, 0}}) = 0
timer_settime(0, 0, {it_interval={0, 0}, it_value={0, 952000}}, NULL) =
0
ioctl(10, SPIOCSTYPE or VTXIOCGETINFO, 0x7fffba4f51d0) = 0
ioctl(10, SPIOCSTYPE or VTXIOCGETINFO, 0x7fffba4f51d0) = 0
ioctl(10, SPIOCSTYPE or VTXIOCGETINFO, 0x7fffba4f51d0) = 0
ioctl(10, SPIOCSTYPE or VTXIOCGETINFO, 0x7fffba4f51d0) = 0
ioctl(10, SPIOCSTYPE or VTXIOCGETINFO, 0x7fffba4f51d0) = 0

Regards,
Steve.