Red Hat Bugzilla – Bug 135109
Nfs mount problem whith complex /etc/exports
Last modified: 2008-05-09 13:34:24 EDT
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.2)
Description of problem:
I was unable to mount an exported NFS directory.
On the Server (hostname = raid) I used the following /etc/exports:
So the /ma directory is exported to the whole class A subnet
192.168.1.0 and in addition to the 2 distinct clients 192.168.6.126
and 192.168.6.162 which will subsequently be called "exceptions".
/stud is exported to the whole class B net 192.168.0.0 .
Now an attempt to mount /stud from e.g. client 192.168.6.126 fails.
After some time the client's /var/log/messages shows a "NFS server
raid not responding, still trying" line. In the server's
/var/log/messages I can see :
raid rpc.mountd: authenticated mount request from
r133-001.pool.mintern.mi.hdm-stuttgart.de:898 for /stud (/stud)
If I change the client's IP adress to any value in the same subnet but
distinct from the two "exception" adresses in the /stud rule of
/etc/exports I have no problem at all. And finally, if I remove the
two exception rules in /etc/exports, i.e.:
In this case I have no mounting problem at all.
I tried older and newer Versions of nfs-utils (FC1, FC3-test2,
FC-development) which all show the same behaviour.
BTW, non-technical issue: I recently read that it is not a good idea
to call a directory /stud due to its meaning as a "four letter word".
At our site this is just an abbreviation for the german word /student,
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1.Create a corresponding /etc/exports
2.Try to mount from a client with "exception rule IP"
Sorry for a small error:
The line "distinct from the two "exception" adresses in the /stud rule
of" should be replaced by "distinct from the two "exception" adresses
in the /ma rule of"
What happens if you add "no_subtree_check" option into all rules in
/etc/exports? I may have a similar problem. It seems that 2.6 kernel
NFS sometimes does not work correctly when you export two subdirs of
the same filesystem (are your /stud and /ma on the same FS?). The
no_subtree_check goes around this problem, but then you export
something more than you have originally intended to.
Does the attached patch fix this to you? For description see my post
to LKML and firstname.lastname@example.org from today (or to the patch itself).
Created attachment 105509 [details]
Patch to linux/net/sunrpc/svcauth_unix.c
This is similar to bug 127521
Refering to comment no. 3 the patch did it's job i.e. I am able to
mount the share in question. Will this patch find its way into the
next updated FC2 Kernel?
*** This bug has been marked as a duplicate of 127521 ***
Changed to 'CLOSED' state since 'RESOLVED' has been deprecated.