Ubuntu 2.6.24-29.94-server
We've encountered what appears to be the same issue as #415632, but on Hardy. The patch attached to #415632 will not apply on Hardy. In specific circumstances, attempting to call mkdir() from within an app with an active AppArmor profile results in failure of the mkdir() and the following log message:
Oct 12 13:33:06 node02 kernel: [ 1094.498142] audit(1318444386.721:15): type=1503 operation="inode_mkdir" info="Failed name resolution - object not a valid entry" error=-2 requested_mask="w::" denied_mask="w::" pid=7283 profile="/usr/sbin/apache2//www.example.com" namespace="default"
Repro steps:
1) Create an apparmor hat that allows rw on an NFS-mounted directory and its subdirectories.
2) Execute code that attempts to create new subdirectories under this directory.
3) Observe that, unreliably, some combinations of mkdir()/rmdir() will result in failure of the mkdir() call and the above AppArmor log entry.
The order of events necessary to trigger the bug seems to vary depending on how mkdir() is called (via the languages built-in mkdir interface or via a system() call). Multiple hosts mounting the same NFS filesystem seem to affect each other. Calling mkdir() or rmdir() on one host affects whether or not it will subsequently succeed on another host.
Unfortunately I can't be more specific as there appears to be a random element that causes carefully elicited patterns to suddenly and spontaneously change, causing the sysadmins to scream in agony and tear at their hair.
This bug is missing log files that will aid in diagnosing the problem. From a terminal window please run:
apport-collect 874544
and then change the status of the bug to 'Confirmed'.
If, due to the nature of the issue you have encountered, you are unable to run this command, please add a comment stating that fact and change the bug status to 'Confirmed'.
This change has been made by an automated script, maintained by the Ubuntu Kernel Team.