Bug 207217
| Summary: | Kernel panic when using NFS with fscache and doing any read() operation | ||
|---|---|---|---|
| Product: | File System | Reporter: | David Flemström (david.flemstrom) |
| Component: | NFS | Assignee: | David Howells (dhowells) |
| Status: | ASSIGNED --- | ||
| Severity: | blocking | CC: | jelle, maxim.doucet, trondmy |
| Priority: | P1 | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Kernel Version: | 5.6.3 | Subsystem: | |
| Regression: | No | Bisected commit-id: | |
| Attachments: | dmesg output | ||
Maybe not too surprisingly, removing the fsc mount option causes the panic to no longer happen, but of course with degraded performance. I misread the stack trace, apparently the crash happens with a write() and then doing a read() shortly after This seems to be a duplicate of bug 208883 where a fix was proposed |
Created attachment 288395 [details] dmesg output Mounting a NFS volume and then performing a write() operation on any file causes this panic to occur. The write() of course never completes, and the calling process hangs in an unkillable state. The mount options are: rw,vers=4.1,fsc,async,hard,rsize=1048576,wsize=1048576,timeo=900,retrans=5,_netdev,auto,x-systemd.automount,x-systemd.device-timeout=10,timeo=14,x-systemd.requires=network-online.target,nodev,nosuid,noatime,nodiratime See dmesg output attached