2.6.10, lockup on lvrename of a snapshot volume
device-mapper snapshots lock up the kernel if you lvrename the snapshot volume. The original volume has to be mounted for this to happen.
pvcreate /dev/hdb1
vgcreate testvg /dev/hdb1
lvcreate -L10G -n test testvg
mke2fs -j /dev/mapper/testvg-test
mount /dev/mapper/testvg-test /mnt
lvcreate -s -l 128 -n test_backup testvg/test
lvrename testvg test_backup test_backup.tmp
<lockup>
SysRq Trace of lvrename:dm_unplug_all [dm_mod]io_shedulesync_buffersync_buffer__wait_on_bitout_of_line_wait_on_bitsync_bufferwake_bit_functionwake_bit_functionsync_dirty_bufferext3_unlockfs [ext3]thaw_bdev__unlock_fs [dm_mod]dm_resume [dm_mod]do_resume [dm_mod]dev_suspend [dm_mod]ctl_ioctl [dm_mod]sys_ioctlsyscall_call
LVM2 version is 2.00.24 (Debian Sarge userland), kernel 2.6.10 from kernel.org. It has been reported that this affects 2.6.9 also, but I have not verified this myself.
Quoted 2 lines · Andre Tomt
device-mapper snapshots lock up the kernel if you lvrename the snapshot volume. The original volume has to be mounted for this to happen.
[...]
Quoted 3 lines
LVM2 version is 2.00.24 (Debian Sarge userland), kernel 2.6.10 from kernel.org. It has been reported that this affects 2.6.9 also, but I have not verified this myself.
I experienced the same with 2.6.9-rc2-mm4 (also running Sarge). The system runs fine, but the lvrename process gets stuck and the lvm subsystem isn't working properly after, a reboot is required.
LLAP, Martin