On Thu, Dec 27, 2018 at 12:20:42PM +0800, Su Hua wrote: > Hi Brian： > > We use kernel version: 4.4.163 > > Our test department found that this version of the kernel still has the > problem that the umount operation will enter the D state after the disk is > unplugged. The information is as follows: > > [Mon Dec 17 13:41:17 2018] INFO: task umount:19009 blocked for more than > 120 seconds. > [Mon Dec 17 13:41:17 2018] Tainted: G W O 4.4.163 #4 > [Mon Dec 17 13:41:17 2018] "echo 0 > > /proc/sys/kernel/hung_task_timeout_secs" disables this message. > [Mon Dec 17 13:41:17 2018] umount D ffffffc000084974 0 19009 > 8925 0x00000008 > [Mon Dec 17 13:41:17 2018] Call trace: > [Mon Dec 17 13:41:17 2018] [<ffffffc000084974>] __switch_to+0x60/0x6c > [Mon Dec 17 13:41:17 2018] [<ffffffc0007ed9d8>] __schedule+0x1a4/0x564 > [Mon Dec 17 13:41:17 2018] [<ffffffc0007eddd0>] schedule+0x38/0x90 > [Mon Dec 17 13:41:17 2018] [<ffffffc00035a10c>] > xfs_ail_push_all_sync+0xac/0xf8 > [Mon Dec 17 13:41:17 2018] [<ffffffc000343bd4>] xfs_unmountfs+0x48/0x110 > [Mon Dec 17 13:41:17 2018] [<ffffffc000346db4>] xfs_fs_put_super+0x30/0x84 > [Mon Dec 17 13:41:17 2018] [<ffffffc0001896e0>] > generic_shutdown_super+0x68/0xdc > [Mon Dec 17 13:41:17 2018] [<ffffffc000189a34>] kill_block_super+0x1c/0x5c > [Mon Dec 17 13:41:17 2018] [<ffffffc000189bb0>] > deactivate_locked_super+0x5c/0x88 > [Mon Dec 17 13:41:17 2018] [<ffffffc00018a948>] deactivate_super+0xa8/0xb4 > [Mon Dec 17 13:41:17 2018] [<ffffffc0001a4614>] cleanup_mnt+0x3c/0x74 > [Mon Dec 17 13:41:17 2018] [<ffffffc0001a46b0>] __cleanup_mnt+0x10/0x18 > [Mon Dec 17 13:41:17 2018] [<ffffffc0000b3f90>] task_work_run+0xa8/0xc4 > [Mon Dec 17 13:41:17 2018] [<ffffffc000086458>] do_notify_resume+0x6c/0x74 > [Mon Dec 17 13:41:17 2018] [<ffffffc0000839e0>] work_pending+0x1c/0x20 > > And the trace info like this: > xfs_umount_D.txt > <https://github.com/zlsh007/Ohters/blob/master/Bug/xfs_umount_D.txt> > > Can you give me some advice, how can I solve this problem? > I'd probably start with commit e6b3bb78962e6 ("xfs: add "fail at unmount" error handling configuration") which went into v4.7. That may or may not depend on earlier error configuration work as well. Brian > Best regards
