[Top] [All Lists]

xfs umount hang in xfs_ail_push_all_sync on i/o error

To: xfs@xxxxxxxxxxx
Subject: xfs umount hang in xfs_ail_push_all_sync on i/o error
From: Bob Mastors <bob.mastors@xxxxxxxxxxxxx>
Date: Mon, 28 Apr 2014 16:29:02 -0600
Delivered-to: xfs@xxxxxxxxxxx

I have an xfs umount hang caused by forcing the block device to return
i/o errors while copying files to the filesystem.
Detailed steps to reproduce the problem on virtualbox are below.

The linux version is a recent pull and reports as 3.15.0-rc3.

[ 2040.248096] INFO: task umount:10303 blocked for more than 120 seconds.
[ 2040.323947]       Not tainted 3.15.0-rc3 #4
[ 2040.343423] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 2040.352665] umount          D ffffffff8180fe40     0 10303   8691 0x00000000
[ 2040.404918]  ffff88001e33dd58 0000000000000086 ffff88001e33dd48 ffffffff81080f82
[ 2040.489901]  ffff88001b311900 0000000000013180 ffff88001e33dfd8 0000000000013180
[ 2040.534772]  ffff88003daa3200 ffff88001b311900 ffff88002421aec0 ffff88002421ae80
[ 2040.587450] Call Trace:
[ 2040.592176]  [<ffffffff81080f82>] ? try_to_wake_up+0x232/0x2b0
[ 2040.620212]  [<ffffffff816c54d9>] schedule+0x29/0x70
[ 2040.627685]  [<ffffffffa04c4cd6>] xfs_ail_push_all_sync+0x96/0xd0 [xfs]
[ 2040.632236]  [<ffffffff81092230>] ? __wake_up_sync+0x20/0x20
[ 2040.659105]  [<ffffffffa04731a3>] xfs_unmountfs+0x63/0x160 [xfs]
[ 2040.691774]  [<ffffffffa0478f65>] ? kmem_free+0x35/0x40 [xfs]
[ 2040.698610]  [<ffffffffa0474cf5>] xfs_fs_put_super+0x25/0x60 [xfs]
[ 2040.706838]  [<ffffffff8119561e>] generic_shutdown_super+0x7e/0x100
[ 2040.723958]  [<ffffffff811956d0>] kill_block_super+0x30/0x80
[ 2040.734963]  [<ffffffff8119591d>] deactivate_locked_super+0x4d/0x80
[ 2040.745485]  [<ffffffff8119652e>] deactivate_super+0x4e/0x70
[ 2040.751274]  [<ffffffff811b1d42>] mntput_no_expire+0xd2/0x160
[ 2040.755894]  [<ffffffff811b2fff>] SyS_umount+0xaf/0x3b0
[ 2040.761032]  [<ffffffff816d1592>] system_call_fastpath+0x16/0x1b
[ .060058] XFS (sdb): xfs_log_force: error 5 returned.
[ 268059] XFS (sdb): xfs_log_force: error 5 returned.

I took a look at xfs_ail_push_all_sync and it is pretty easy to see
the hang. But it is not obvious to me how to fix it.
Any ideas would be appreciated.

I am available to run additional tests or capture more logging
or whatever if that would help.


virtual box setup
1 GB ram
8 GB disk
4 processors
Operating System: Linux
Version: Debian (64 bit)
Bridged networking
iso: debian-7.4.0-amd64-netinst.iso
Software selection
    select SSH server
    select Standard system utilities
    deselect everything else
dhcp ip address:

upgrade linux to 3.15.0-rc3
cd /mnt/ksrc/linux
make modules_install
make install

install tools
apt-get install xfsprogs
apt-get install open-iscsi
apt-get install --no-install-recommends targetcli python-urwid

reproduce bug
    cd backstores
    fileio/ create name=file_backend2 file_or_dev=my_disk2 size=1G
    /iscsi create
    portals/ create
    cd <
    luns/ create /backstores/fileio/file_backend2
    cd <
    set attribute authentication=0 demo_mode_write_protect=0 generate_node_acls=1 cache_dynamic_acls=1

iscsiadm -m discovery -t sendtargets -p
iscsiadm -m node --login --target iqn.from_previous_iscsiadm_cmd_output
fdisk -l
mkfs -t xfs -f /dev/sdXX
mkdir /mnt/vola

put following lines in a shell script "foo.sh", fix up target and block device
    set -x
    iscsiadm -m node --target iqn.2003-01.org.linux-iscsi.debian.x8664:sn.51ab79ac3ce2 --login
    sleep 1
    mount -t xfs /dev/sdb /mnt/vola
    sleep 10
    iscsiadm -m node -U all
    sleep 1
    umount /mnt/vola

run the shell script in a loop like:
    while true; do ./foo.sh; done

in another terminal run the following:
    while true; do cp -r /usr/lib /mnt/vola; done

The hang seems to take from two to ten minutes to occur.

<Prev in Thread] Current Thread [Next in Thread>