Re: [PATCH v1] security_manager: Ensure top lock is acquired before nested locks

Yes, we have recently encountered this issue, and it is exactly the same as the scenario described in the bug report. We investigated the code and found this potential problem. The issue occurs when collecting information through the ‘virsh capabilities' while concurrently creating vm. The higher the frequency and concurrency, the more likely the issue will reproduce. We believe this is a very rare bug; our environment has been running for three years, and we've only encountered this issue once. To successfully reproduce the problem, we constructed a test case by adding a sleep between the lock nested and unlock. But we still need to try to fix this issue, even though the probability is very low.