On Tue, Mar 12, 2019 at 16:13:18 +0100, Michal Privoznik wrote:
https://bugzilla.redhat.com/show_bug.cgi?id=1623389
Steps to reproduce:
1) cat shmem.xml <shmem name='my_shmem0'> <model type='ivshmem-plain'/> <size unit='M'>4</size> <alias name='ua-123'/> </shmem>
2) virsh attach-device vm1 shmem.xml
3) virsh detach-device-alias vm1 ua-123; virsh detach-device vm1 shmem.xml
4) observe that the device is still in the domain: virsh dumpxml vm1 <shmem name='my_shmem0'> <model type='ivshmem-plain'/> <size unit='M'>4</size> <alias name='ua-123'/> <address type='pci' domain='0x0000' bus='0x00' slot='0x0a' function='0x0'/> </shmem>
5) But qemu has the device no more: virsh detach-device-alias vm1 ua-123 error: Failed to detach device with alias ua-123 error: internal error: unable to execute QEMU command 'device_del': Device 'ua-123' not found
This reproducer is to make sure that DELETE_DEVICE event arrives while monitor is unlocked. It is very hard to time qemu and libvirt so that the event comes exactly at the time when detach-device from 3) is doing the monitor call. Simulate this by unlocking monitor, waiting a few seconds and locking the monitor again.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
I strongly suggest to not add sign-off to patches which are not meant to be pushed. This will actually make the otherwise useless sign-off somewhat usefull as the push-hook will reject pushing such a series.