[Bug 769064] [NEW] [Maverick] autofs5 leaves too many orphaned mount points in /proc/mounts and reboots takes forever

[Leonardo Borda]

Public bug reported:

Binary package hint: autofs5

Overview:

In an environment with over 11K mount points in the table (/proc/mounts) and using autofs5 with autofs5-ldap support, rebooting the machine takes forever since autofs leaves many orphaned entries in /proc/mounts.
Running automount -dvf, then hitting ctrl-c does the right thing and unmounts all mounts fine. 

Reproducible: 100% 
$ sudo stop autofs
Password:
autofs stop/waiting
$ cat /proc/mounts | wc -l
11269

Workaround:
Use Lucid. Behaviour does not happen in Lucid however it does happen in Maverick and in the new Natty release.

Useful information:
strace output:
[pid 1063] ioctl(3, AUTOFS_IOC_ASKUMOUNT, 0x7f8fca2b0988) = 0
[pid 1063] ioctl(3, AUTOFS_IOC_CATATONIC, 0) = 0
[pid 1063] close(3) = 0
[pid 1063] sched_yield() = 0
[pid 1063] umount("/path/changed", 0 <unfinished ...>
[pid 1036] <... ???? resumed> ) = ? <unavailable>
[pid 1036] +++ killed by SIGKILL +++
PANIC: handle_group_exit: 1036 leader 1034
[pid 1035] +++ killed by SIGKILL +++
PANIC: handle_group_exit: 1035 leader 1034
[pid 1063] +++ killed by SIGKILL +++
PANIC: handle_group_exit: 1063 leader 1034
+++ killed by SIGKILL +++

** Affects: autofs5 (Ubuntu)
     Importance: Undecided
         Status: New

-- 
You received this bug notification because you are a member of Ubuntu
Server Team, which is subscribed to autofs5 in Ubuntu.
https://bugs.launchpad.net/bugs/769064

Title:
  [Maverick] autofs5 leaves too many orphaned mount points in
  /proc/mounts and reboots takes forever