I saw your boot option, in KVMGT you should add ‘ignore_loglevel’ in grub if you want to get all the log info. ‘loglvl=all’ is for XenGT.
In this log, I saw there is one restart of VM2, it’s auto restart? Or your test case trigger auto restart. Could you help to share the event log and dump file in VM2 when you got windows guest trigger reset?
In your last email you mentioned:
After few days test, I found that the E3V5 host crashed after 60~70 hours, which crashing is "freeze".
Then I checked htop and dmesg watch monitor and found two things:
1. There is no error log when crash, only kernel trace about i915 driver when starting VM.
2. When testing, the memory usage is increasing.
I am not sure ‘freeze’ mean, you can still use htop and dmesg after you got ‘freeze’? so host still can response?
Could you help to share the memory utilization when you got ‘freeze’? if you still can run command in this scenario, could you also help to dump the call trace in all CPU cores?