We noticed that there is a Handle leak issue observed in RHEL cifs client and there are two scenarios:
1) When file metadata is cached on the client but the file was renamed into another existing file on the server by another client. The client accessing the file gets FileID from the server which doesn't match the cached one. Then the client fails open system call with an handle being opened on the cloud side. Other rare cases (e.g. low memory) might cause a similar behavior.
2) When a process waiting for a SMB2_CREATE response inside the kernel was interrupted by the signal (ex: The application was killed). In that case the client doesn’t recognize a response when it finally came thus leaking an open handle on the cloud side.
- Red Hat Enterprise Linux 7
- Seen on 3.10.0-1062.4.1.el7 and 3.10.0-1062.1.2.el7
- cifs mount with "-o nofail,vers=3.0,sec=ntlmssp"
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase of over 48,000 articles and solutions.