Red Hat Bugzilla – Bug 124116
mount --bind works, then later the directory contents diverge
Last modified: 2007-04-18 13:07:57 EDT
Description of problem:
Loopback mount of a directory tree works as expected for a long time
(over seventy days of uptime in this case), then for some unknown
reason the directory contents diverge: the original location displays
the (properly) updated files, yet the loopback-mount location
('path_to_B' in the case of 'mount --bind /path_to_A /path_to_B') does
NOT contain the updated files; rather, it contains a slightly (half a
day perhaps?) stale listing of the files, and subsequent updates on
files in /path_to_A does not affect files in /path_to_B. The 'mount'
command reports that the directory is still properly loopback-mounted,
and all files except for the recently updated ones appear in both
/path_to_A and /path_to_B as expected. If I manually unmount the
loopback-mount, /path_to_B displays no files at all (as expected), so
the mount was definitely still active.
Unfortunately this problem occurred on a production webserver, so I
could not spend much time investigating the problem at the time it
happened. I ran 'umount /path_to_B ; mount --bind /path_to_A
/path_to_B' and the behavior returned to normal; the updated copies of
files appeared in both /path_to_A and /path_to_B, and
creating/modifying/removing subsequent files continue to function as
I am attaching a (slightly sanitized for security reasons) log of the
event; it should show that 1) a number of files on both mounts indeed
differed; 2) unmounting then mounting the filesystem fixes the problem
(at least for now); and 3) the directory /path_to_B contained no files
when no loopback-mount was active.
Machine is a generic motherboard / Celeron 700 / 128MB RAM / 1.0GB swap.
Possible factors: Until recently, httpd was not restarted often.
Now, crond reloads httpd every five minutes ('service httpd reload').
Apache uses the /var/www/htdocs/ directory. (I need to use 'mount
--bind' because Apache only enables SuExec for files inside the
Version-Release number of selected component (if applicable):
kernel-2.4.20-20.9 (uniprocessor version)
Haven't been able to yet. Will try on non-production systems.
Created attachment 100485 [details]
sanitized transcript of shell output for initial discovery of the problem
Ack, this machine is running Redhat 9.0 rather than FC1. Kernel is
Thanks for the bug report. However, Red Hat no longer maintains this version of
the product. Please upgrade to the latest version and open a new bug if the problem
The Fedora Legacy project (http://fedoralegacy.org/) maintains some older releases,
and if you believe this bug is interesting to them, please report the problem in
the bug tracker at: http://bugzilla.fedora.us/