Description of problem:
Due to some foolishness on my part, I was attempting to bind mount a file onto /proc/self/cgroup, instead of /proc/THE-ACTUAL-PID/cgroup. I noticed that when I do this, I get a stale, non-removeable entry in the kernel mount table.

# cp /proc/self/cgroup  /root/cgroup

# mount --bind /root/cgroup /proc/self/cgroup 

# grep proc /proc/mounts 
proc /proc proc rw,relatime 0 0
systemd-1 /proc/sys/fs/binfmt_misc autofs rw,relatime,fd=33,pgrp=1,timeout=300,minproto=5,maxproto=5,direct 0 0
binfmt_misc /proc/sys/fs/binfmt_misc binfmt_misc rw,relatime 0 0
sunrpc /proc/fs/nfsd nfsd rw,relatime 0 0
/dev/mapper/vg_t500wlan-lv_root /proc/26770/cgroup ext4 rw,seclabel,relatime,data=ordered 0 0

# umount -f /proc/26770/cgroup
umount: /proc/26770/cgroup: not found

In fact if I use /proc/1234/cgroup, i'll hit the same problem when process 1234 exits.

So the kernel doesn't appear to cleanup bind mounts under any /proc/NNNNN/ location when the process NNNN exits.

Version-Release number of selected component (if applicable):
kernel-3.9.9-302.fc19.x86_64

How reproducible:
Always

Steps to Reproduce:
1. Bind mount a file onto /proc/NNNN/YYYY, where NNN is any process ID, or 'self'
2. Let the process NNN exit
3. Look for the mount bind

Actual results:
The bind mount still exists & cannot be removed, since /proc/NNN no longer exists

Expected results:
Any mounts under /proc/NNNN are purged when process NNN exits.

Additional info: