Sporadic deadlocks during rsync #4554
Comments
At the very least, the zfs_secpolicy_write_perms ioctl security policy callback, which calls dsl_dataset_hold(), can require freeing memory and, therefore, re-enter ZFS. This patch enables PF_FSTRANS for all of the security policy callbacks similarly to the manner in which it's enabled for the actual ioctl callback. May-fix: openzfs#4554
|
Thanks for the quick feedback. I'll give that a try and let you know how things turn out. I typically see a crash once or twice a week, right now, so shouldn't be long to know if it's doing better. |
|
Two and a half days of uptime so far, and no crashes. It's too soon to say for sure whether things are fixed, but your patch doesn't appear to have made things worse. I'll keep an eye on things and report back in a few days. Thanks again for the quick patch. |
|
@varju Thanks for the feedback. Your stack trace makes it pretty clear we need this so I'll post it as a PR today. |
|
@varju could you comment on if you've seen the issue after a week now? |
|
7 days of uptime, and no issues so far. |
At the very least, the zfs_secpolicy_write_perms ioctl security policy callback, which calls dsl_dataset_hold(), can require freeing memory and, therefore, re-enter ZFS. This patch enables PF_FSTRANS for all of the security policy callbacks similarly to the manner in which it's enabled for the actual ioctl callback. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #4554
|
Thank you both. My server is no longer crashing and I'm a happy camper. |
At the very least, the zfs_secpolicy_write_perms ioctl security policy callback, which calls dsl_dataset_hold(), can require freeing memory and, therefore, re-enter ZFS. This patch enables PF_FSTRANS for all of the security policy callbacks similarly to the manner in which it's enabled for the actual ioctl callback. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes openzfs#4554
At the very least, the zfs_secpolicy_write_perms ioctl security policy callback, which calls dsl_dataset_hold(), can require freeing memory and, therefore, re-enter ZFS. This patch enables PF_FSTRANS for all of the security policy callbacks similarly to the manner in which it's enabled for the actual ioctl callback. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes openzfs#4554
At the very least, the zfs_secpolicy_write_perms ioctl security policy callback, which calls dsl_dataset_hold(), can require freeing memory and, therefore, re-enter ZFS. This patch enables PF_FSTRANS for all of the security policy callbacks similarly to the manner in which it's enabled for the actual ioctl callback. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes openzfs#4554
For a number of months now I've been seeing deadlocks once or twice a week. Typically rsync will be running when I notice the problem, but this might not be the actual trigger.
Pool configuration:
Stack traces that include zfs:
The text was updated successfully, but these errors were encountered: