btrfs qemu-ga - multiple mounts block fsfreeze

[Dominique Ramaekers]

[Impact]

 * backport upstream fix to avoid double unmounts (e.g. bind mounts)
   hanging the freeze action.

[Test Case]

 * Prepare a guest
   $ uvt-simplestreams-libvirt --verbose sync --source http://cloud-images.ubuntu.com/daily arch=amd64 release=xenial label=daily
uvt-kvm create --password ubuntu x-freeze arch=amd64 release=xenial label=daily

 * Shut it down and use virsh edit to add a channel for the agent. Add this somewhere under <devices>:
<channel type='unix'>
   <source mode='bind'/>
   <target type='virtio' name='org.qemu.guest_agent.0'/>
</channel>

  * Start the guest again, in it install qemu-guest-agent

  *

  * On the Host then freeze and thaw the guest
    $ virsh domfsfreeze x-freeze
    $ virsh domfsthaw x-freeze

  # The above works, then trigger the issue, in the guest do
    $ sudo mkdir /mnt/test
    $ sudo mount -o bind / /mnt/test/

  * Now the freeze/thaw will fail unless the fix is applied to the guest.

[Regression Potential]

 * From looking at the code alone I could think of a different cause that
   makes it return EBUSY.
   The biggest risk I could see due to that is the agent reporting a
   successful freeze due to accepting EBUSY which is not actually true.
   But we have that in Ubuntu since Artful and
   upstream since January 2017 and there are no such reports.

[Other Info]

 * n/a

---

Having two mounts of the same device makes fsfreeze through qemu-ga impossible.
root@CmsrvMTA:/# mount -l | grep /dev/vda2
/dev/vda2 on / type btrfs (rw,relatime,space_cache,subvolid=257,subvol=/@)
/dev/vda2 on /home type btrfs (rw,relatime,space_cache,subvolid=258,subvol=/@home)

Having two mounts is rather common with btrfs, so the feature fsfreeze is unusable on these systems.

Below more information about how we encountered this issue...

Message send to <email address hidden>:

Message 1:
----------
I use external snapshots to backup my guests. I use the 'quiesce' option to flush and frees the guest file system with the qemu guest agent.

With the exeption of one guest, this procedure works fine. On the 'unwilling' guest, I get this error message:
"ERROR 2016-05-25 00:51:19 | T25-bakVMSCmsrvVH2 | fout: internal error: unable to execute QEMU agent command 'guest-fsfreeze-freeze': failed to freeze /: Device or resource busy"

I don't think this is not some sort of time-out error, because activation of the fsfreeze and the error message happen immediately after each other:

$ grep qemu-ga syslog.1
May 25 00:51:19 CmsrvMTA qemu-ga: info: guest-fsfreeze called

This is the only entry of the qemu guest agent in syslog.

$ sudo virsh version
Compiled against library: libvirt 1.3.1
Using library: libvirt 1.3.1
Gebruikte API: QEMU 1.3.1
Draaiende hypervisor: QEMU 2.5.0

$ virsh qemu-agent-command CmsrvMTA '{"execute": "guest-info"}'
{"return":{"version":"2.5.0", ... ,{"enabled":true,"name":"guest-fstrim","success-response":true},{"enabled":true,"name":"guest-fsfreeze-thaw","success-response":true},{"enabled":true,"name":"guest-fsfreeze-status","success-response":true},{"enabled":true,"name":"guest-fsfreeze-freeze-list","success-response":true},{"enabled":true,"name":"guest-fsfreeze-freeze","success-response":true}, ... }

For making an external snapshot, I use this command:
$ virsh snapshot-create-as --domain CmsrvMTA sn1 --disk-only --atomic --quiesce --no-metadata --diskspec vda,file=/srv/poolVMS/CmsrvMTA.sn1

Piece of reply 1:
-----------------
I have encountered this before. Some operating systems
 may have bind-mounts that let a device appear multiple times in the mount list. Unfortunately the guest agent is not smart enough to consider a device that has been frozen as succesfull and keep going. This causes this specific error.

Piece of reply 2:
-----------------
Ok, that seems to be it.

I’ve got the ‘/’ and ‘/home’ on the same device formatted as btrfs on two separate sub volumes.

[Christian Theune (ctheune)]

I'm the responder from the Qemu list. I reviewed that C code a while ago when I stumbled over it. If someone helps me through the Qemu patch acceptance process then I'm willing to provide a patch.

[Tim Rose (iamtimde)]

I have a related Problem with a libvirt Guest running some Docker-Containers.

root@docker1:~# df -h
Filesystem Size Used Avail Use% Mounted on
/dev/sda1 7.9G 2.9G 4.6G 39% /
udev 10M 0 10M 0% /dev
tmpfs 794M 8.6M 785M 2% /run
tmpfs 2.0G 672K 2.0G 1% /dev/shm
tmpfs 5.0M 0 5.0M 0% /run/lock
tmpfs 2.0G 0 2.0G 0% /sys/fs/cgroup
none 7.9G 2.9G 4.6G 39% /var/lib/docker/aufs/mnt/8b510e9eb71523cf9afc723e11831a169c822b7251049f95ebed5e978b22184b
shm 64M 0 64M 0% /var/lib/docker/containers/3d5542423739bc1a4467ce5e15a3b82224d2d1b315af75d7d91a0cc04e1e7b07/shm
none 7.9G 2.9G 4.6G 39% /var/lib/docker/aufs/mnt/5d186a68a1fca2b58a440e6dbdf293853011a9526b819ac1dbcba7f91272535f
shm 64M 0 64M 0% /var/lib/docker/containers/a4355f5e2a66b25a564c5027f818b4385cd417f85d8aaf79700ca4086fb86e63/shm
none 7.9G 2.9G 4.6G 39% /var/lib/docker/aufs/mnt/e972052c347006cd8c79dc5bf9aaba07fdba2edeadc01e18e903d04ca4b05a86
none 7.9G 2.9G 4.6G 39% /var/lib/docker/aufs/mnt/e22995e962919ea50fcf7b4a5e52916e6839f76708d22e57e81de687af2f0dd0
shm 64M 0 64M 0% /var/lib/docker/containers/00e3490e37ded90e7421eb19e650cc684dec2b65a848a203c21212a549bf9e0e/shm
shm 64M 0 64M 0% /var/lib/docker/containers/631b37644a0ac106b7d7403c2fcc3760d4647cb96141915b0aa7ae78f98e4e05/shm
root@docker1:~# mount -l | grep /dev/vda2
root@docker1:~# mount -l | grep /dev/sda
/dev/sda1 on / type ext4 (rw,relatime,discard,errors=remount-ro,data=ordered)
/dev/sda1 on /var/lib/docker/aufs type ext4 (rw,relatime,discard,errors=remount-ro,data=ordered)

I am not able to take snapshots:

error: internal error: unable to execute QEMU agent command 'guest-fsfreeze-freeze': failed to freeze /: Device or resource busy

Thank you
Tim

[Alexandre (totalworlddomination)]

Strangely I've started having the same problem as Tim this week, beside regular upgrade I did reboot since last backup... (Ubuntu LTS 16.04)

if I stop docker, the backup will proceed, otherwise i get the same failed to freeze / error message, even with no container running, just the daemon. That's both true with 1.11.2 and 1.12.1 docker dameon version.

[Alexandre (totalworlddomination)]

I also have a path mounted twice, but this time it's devicemapper and not AUFS like tim, so maybe not fs type related.

virsh version:
  Compiled against library: libvirt 1.3.1
  Using library: libvirt 1.3.1
  Using API: QEMU 1.3.1
  Running hypervisor: QEMU 2.5.0

KVM/QEMU Host command:
    virsh snapshot-create-as $VM_NAME \
            ${VM_NAME}_snapshot \
            "${VM_NAME} snapshot $(date +%Y-%m-%d-%H%M%S)" \
        --diskspec vda,file=${VM_QCOW_FILE}-snap.qcow2 \
        --disk-only \
        --atomic \
        --no-metadata \
        --quiesce

VM similar mount source:
  /dev/mapper/delve--vg-root on / type ext4 (rw,relatime,discard,errors=remount-ro,data=ordered)
  /dev/mapper/delve--vg-root on /var/lib/docker/devicemapper type ext4 (rw,relatime,discard,errors=remount-ro,data=ordered)

In the VM logs:
  Sep 8 12:48:45 xxxxxxxx qemu-ga: info: guest-fsfreeze called

[Christian Theune (ctheune)]

This has been merged in upstream Qemu a while ago.

[Thomas Huth (th-huth)]

Was that this commit here:
http://git.qemu.org/?p=qemu.git;a=commitdiff;h=ce2eb6c4a044d809c
?
That was included in QEMU v2.9.0, so I think we can set the status to "Fix released" now?

[Alexandre (totalworlddomination)]

Any chance this will ever get backported to 2.5 on Xenial?
cheers!

[Thomas Huth (th-huth)]

You've got to open this (or a new bug) against Ubuntu's qemu package to get it included there. So far, this bug here is only opened against the upstream QEMU project.

[Janåke Rönnblom (jan-ake)]

This affects Ubuntu 16.04 as in #4

[Christian Ehrhardt  (paelzer)]

Fix is known and seems backportable, marking as server-next and subscribing for somebody to take a look.
Also as mentioned it is in 2.9, so newer releases are already fixed.

[Christian Ehrhardt  (paelzer)]

Steps to reproduce

uvt-simplestreams-libvirt --verbose sync --source http://cloud-images.ubuntu.com/daily arch=amd64 release=xenial label=daily
uvt-kvm create --password ubuntu x-freeze arch=amd64 release=xenial label=daily

Add this to the guest definition and restart it
<channel type='unix'>
   <source mode='bind'/>
   <target type='virtio' name='org.qemu.guest_agent.0'/>
</channel>

From the host freeze and thaw it:
$ virsh domfsfreeze x-freeze
$ virsh domfsthaw x-freeze

# The above works

In the Guest then add a bind mount to trigger the issue
$ sudo mkdir /mnt/test
$ sudo mount -o bind / /mnt/test/

Then again try to freeze/unfreeze
$ virsh domfsfreeze x-freeze
error: Unable to freeze filesystems
error: internal error: unable to execute QEMU agent command 'guest-fsfreeze-freeze': failed to freeze /: Device or resource busy

I installed a fix from my PPA [1] and with that it works just fine now.
Proposing an MP for review by the team, to afterwards move on into SRU processing.

[1]: https://launchpad.net/~ci-train-ppa-service/+archive/ubuntu/3310

[Christian Ehrhardt  (paelzer)]

Tests are all ok, MP review was acked and case confirmed.
No Security Update since then in between, so sponsoring into SRU queue now ...

[Robie Basak (racb)]

Hello Dominique, or anyone else affected,

Accepted qemu into xenial-proposed. The package will build now and be available at https://launchpad.net/ubuntu/+source/qemu/1:2.5+dfsg-5ubuntu10.31 in a few hours, and then in the -proposed repository.

Please help us by testing this new package. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation on how to enable and use -proposed.Your feedback will aid us getting this update out to other Ubuntu users.

If this package fixes the bug for you, please add a comment to this bug, mentioning the version of the package you tested and change the tag from verification-needed-xenial to verification-done-xenial. If it does not fix the bug for you, please add a comment stating that, and change the tag to verification-failed-xenial. In either case, details of your testing will help us make a better decision.

Further information regarding the verification process can be found at https://wiki.ubuntu.com/QATeam/PerformingSRUVerification . Thank you in advance!

[Christian Ehrhardt  (paelzer)]

Prior to the update:

$ virsh domfsfreeze x-freeze; virsh domfsthaw x-freeze
error: Unable to freeze filesystems
error: internal error: unable to execute QEMU agent command 'guest-fsfreeze-freeze': failed to freeze /: Device or resource busy

Thawed 0 filesystem(s)

$ sudo apt install qemu-guest-agent
Reading package lists... Done
Building dependency tree
Reading state information... Done
The following packages will be upgraded:
  qemu-guest-agent
1 upgraded, 0 newly installed, 0 to remove and 27 not upgraded.
Need to get 132 kB of archives.
After this operation, 0 B of additional disk space will be used.
Get:1 http://archive.ubuntu.com/ubuntu xenial-proposed/universe amd64 qemu-guest-agent amd64 1:2.5+dfsg-5ubuntu10.31 [132 kB]
Fetched 132 kB in 0s (1.161 kB/s)
(Reading database ... 54125 files and directories currently installed.)
Preparing to unpack .../qemu-guest-agent_1%3a2.5+dfsg-5ubuntu10.31_amd64.deb ...
Unpacking qemu-guest-agent (1:2.5+dfsg-5ubuntu10.31) over (1:2.5+dfsg-5ubuntu10.30) ...
Processing triggers for man-db (2.7.5-1) ...
Processing triggers for systemd (229-4ubuntu21.2) ...
Processing triggers for ureadahead (0.100.0-19) ...
Setting up qemu-guest-agent (1:2.5+dfsg-5ubuntu10.31) ...

Retesting the above:
$ virsh domfsfreeze x-freeze; virsh domfsthaw x-freeze
Froze 1 filesystem(s)

Thawed 1 filesystem(s)

Found no other hickups due to the update, setting verified.

[Łukasz Zemczak (sil2100)]

The verification of the Stable Release Update for qemu has completed successfully and the package has now been released to -updates. Subsequently, the Ubuntu Stable Release Updates Team is being unsubscribed and will not receive messages about this bug report. In the event that you encounter a regression using the package from -updates please report a new bug using ubuntu-bug and tag the bug report regression-update so we can easily find any regressions.

[Launchpad Janitor (janitor)]

This bug was fixed in the package qemu - 1:2.5+dfsg-5ubuntu10.31

---------------
qemu (1:2.5+dfsg-5ubuntu10.31) xenial; urgency=medium

  * d/p/ubuntu/lp-1587065-qga-ignore-EBUSY-when-freezing-a-filesystem.patch:
    Fix qemu-guest-agent failing to freeze filesystems that were mounted
    multiple times like bind mounts. (LP: #1587065).

 -- Christian Ehrhardt <email address hidden> Thu, 28 Jun 2018 11:35:05 +0200