Re: QEMU 7.2 non-deterministic migration-test CI failure

["Dr. David Alan Gilbert" <dgilbert@redhat.com>]

* Stefan Hajnoczi (stefanha@redhat.com) wrote:
> Hi,
> The following cross-i386-tci migration-test failure was triggered by CI:
> 
> >>> MALLOC_PERTURB_=133 G_TEST_DBUS_DAEMON=/builds/qemu-project/qemu/tests/dbus-vmstate-daemon.sh QTEST_QEMU_BINARY=./qemu-system-i386 /builds/qemu-project/qemu/build/tests/qtest/migration-test --tap -k
> ――――――――――――――――――――――――――――――――――――― ✀  ―――――――――――――――――――――――――――――――――――――
> stderr:
> Could not access KVM kernel module: No such file or directory
> Could not access KVM kernel module: No such file or directory
> Could not access KVM kernel module: No such file or directory
> Could not access KVM kernel module: No such file or directory
> qemu-system-i386: Unknown savevm section type 126
> ../tests/qtest/libqtest.c:188: kill_qemu() detected QEMU death from signal 11 (Segmentation fault) (core dumped)
> TAP parsing error: Too few tests run (expected 14, got 2)
> (test program exited with status code -6)
> ――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――
> 
> https://gitlab.com/qemu-project/qemu/-/jobs/3397205431
> 
> When I retried the test succeeded. I haven't managed to reproduce it
> locally yet either.
> 
> Any ideas?

Hmm, that's not good; I've not seen that error before, but:

QEMU_VM_SECTION_FOOTER= 0x7e

It probably means the destination read the wrong amount on a previous
section; probably something conditional on a state which is why
it only triggers sometimes.

Although,  that would normally trigger a
'Missing section footer for ...'

rather than end up hitting that error.

I can try letting it soak on one of my test machines.

Dave


> Stefan


-- 
Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK