Bug 838609

Summary:	guestmount + fusermount allows a race condition when unmounting and immediately using the disk image
Product:	Red Hat Enterprise Linux 6	Reporter:	Richard W.M. Jones <rjones>
Component:	libguestfs	Assignee:	Richard W.M. Jones <rjones>
Status:	CLOSED ERRATA	QA Contact:	Virtualization Bugs <virt-bugs>
Severity:	high	Docs Contact:
Priority:	unspecified
Version:	6.3	CC:	dnaori, dyasny, leiwang, mbooth, moli, pbrady, qguan, rjones
Target Milestone:	rc
Target Release:	---
Hardware:	x86_64
OS:	Linux
Whiteboard:
Fixed In Version:	libguestfs-1.16.30-2.el6	Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:	838592	Environment:
Last Closed:	2013-02-21 08:38:33 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:	838592
Bug Blocks:	835466

Description Richard W.M. Jones 2012-07-09 15:17:54 UTC

+++ This bug was initially created as a clone of Bug #838592 +++

The current design of guestmount allows a race condition when
unmounting a disk image:

  guestmount -a disk.img -i /mnt
  # copy things into /mnt
  fusermount -u /mnt
  # immediately try to use 'disk.img' <-- UNSAFE

The problem is that guestmount cleanup happens asynchronously
after fusermount has returned.  It is not safe to use 'disk.img'
until guestmount has exited, but there is no simple way to tell
when it has exited.

Suggest adding 'guestmount --pid-file=...' option so that a
PID file can be written, and then programs can wait for the PID
to exit after the fusermount.

Also we should document this.

--- Additional comment from pbrady on 2012-07-09 10:44:37 EDT ---

The same issue is referenced here:
http://code.google.com/p/s3ql/issues/detail?id=159
which in turn references this thread on the subject from 2006:
http://thread.gmane.org/gmane.comp.file-systems.fuse.devel/3903
And another case from 2008:
http://permalink.gmane.org/gmane.comp.file-systems.fuse.devel/6502

So it's been on the TODO list for 6 years.

--- Additional comment from rjones on 2012-07-09 11:14:46 EDT ---

Patches posted:

https://www.redhat.com/archives/libguestfs/2012-July/msg00027.html

Only patches 1, 3 and 4 are really relevant to this bug.
Patch 2 is a doc fix in a related part of the man page.

Comment 2 Mohua Li 2012-11-08 08:50:47 UTC

the time might quite short, after fusermount return, and before guestmount finished cleanup, so not quite easy to reproduce, any suggestion for this, Rich?

Comment 3 Richard W.M. Jones 2012-11-08 08:58:52 UTC

I think you just need to check that RHEL 6.4 guestmount
has the --pid-file option.

The race condition has *not* been fixed (and cannot be fixed).
The fix is for callers -- OpenStack in this case -- to use
the guestmount --pid-file option.

$ guestmount --version
guestmount 1.16.32rhel=6,release=3.el6
$ guestmount --pid-file
guestmount: option '--pid-file' requires an argument
Try `guestmount --help' for more information.

Comment 4 Richard W.M. Jones 2012-11-08 09:01:25 UTC

.. and check that --pid-file actually writes a PID file!

Comment 5 Mohua Li 2012-11-08 11:48:56 UTC

[root@station2 tmp]# guestmount --version
guestmount 1.16.32rhel=6,release=3.el6


the solution is "Suggest adding 'guestmount --pid-file=...' option so that a
PID file can be written, and then programs can wait for the PID
to exit after the fusermount."

[root@station2 tmp]# guestmount --pid-file pid  -a RHEL-Server-5.9-32-pv.raw -i test 
[root@station2 tmp]# cat pid
6043

so other program could get the pid from file, and wait for the $pid process to exit, to avoid the race condition,

Comment 7 errata-xmlrpc 2013-02-21 08:38:33 UTC

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

http://rhn.redhat.com/errata/RHBA-2013-0324.html