From Bugzilla Helper:
User-Agent: Mozilla/5.0 (compatible; MSIE 6.0; Windows; U; AIIEEEE!; Win98; Windows 98; en-US; Gecko masquerading as IE; should it matter?; rv:1.8b) Gecko/20050217

Description of problem:
There is code in /etc/rc.d/init.d/named that attempts to bind-mount /proc if it isn't already mounted under chrooted (sic) hierarchy.
It does this by grep'ing /proc/mounts for "/proc ${ROOTDIR}/proc".

This is wrong because it presumes that the real path of ${ROOTDIR}/proc is the same as the ${ROOTDIR}/proc.


Version-Release number of selected component (if applicable):
bind-9.3.1-14_FC4

How reproducible:
Always

Steps to Reproduce:
1.mv /var/named/chroot /X
2.ln -s /X /var/named/chroot
3./etc/rc.d/init.d/named start
4./etc/rc.d/init.d/named restart
5./etc/rc.d/init.d/named restart
6.cat /proc/mounts
 

Actual Results:  ....
/proc /X/proc proc rw,nodiratime 0 0
/proc /X/proc proc rw,nodiratime 0 0
/proc /X/proc proc rw,nodiratime 0 0


Expected Results:  ....
/proc /X/proc proc rw,nodiratime 0 0 #just once


Additional info:

There is a lot of horrible grep'ing in RedHat/Fedora scripts.
This is just another example amongst many.

It would be better to a use more robust and reliable way to determine whether a node is mounted, especially where aliasing can occur because of symlinks.

Possibilities that come to mind:
1. df ${ROOTDIR}/proc >/dev/null; use $?
2. somehow invoke a realpath(3) translation of ${ROOTDIR}/proc, then grep it.
3. mount -o remount,rw ${ROOTDIR}/proc 2>/dev/null || mount /proc ${ROOTDIR}/proc

I think 3) is the one that I will be patching for.