Update from further testing in QEMU guests. Details and the proposed patch are in https://github.com/openzfs/zfs/issues/19659
- 26.10 is affected too: zfs 2.4.4-1ubuntu4 on kernel 7.3.0-9-generic hits the same oops in vdev_reopen through ZED (Comm: agents). 24.04 with zfs 2.2.2-0ubuntu9.5 is not affected and returns "operation not supported". - On 26.04, a pool imported from /dev/disk/by-uuid with autoexpand=on can crash the module with no resize while the VM is up and no zpool command involved. A VM whose disk had been grown before it booted hit this by itself 603 s after boot. With zfs-zed masked it didn't happen. The exact trigger is unknown. - Workaround: zpool set autoexpand=off on such a pool, and don't pass the numeric name from zpool status to zpool online. A pool re-imported from /dev/disk/by-id or /dev/disk/by-path no longer has a numeric name, and can be expanded. - Recovery after the crash: systemctl mask zfs-zed (this works while the pool is hung, but sync hangs), hard reset, zpool set autoexpand=off, unmask, and start zfs-zed. ** Bug watch added: github.com/openzfs/zfs/issues #19659 https://github.com/openzfs/zfs/issues/19659 -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2172684 Title: zpool online with the pool GUID oopses in vdev_reopen and hangs ZFS until reboot To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2172684/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
