Hello,
IIUC the migration thread should always see valid migration object, as it
takes one refcount at the entrance of migration_thread():
object_ref(OBJECT(s));
Could the migration have failed before ? in migrate_fd_connect()
So it's not yet clear to me on why the mutex was destroyed, if the main
object should still be there; logically it was only destroyed in the
finalize phase (migration_instance_finalize), but that should be after
migration thread quits and releases the refcount..
Avihai, would you have a reproducer ? or did you keep some logs ?
a backtrace maybe ?
Thanks,
C.