If you do
/etc/rc.d/init.d/nfs stop
/etc/rc.d/init.d/nfs restart
and then some computer tries to mount a volume of the server, it locks.
Tracking down the problem, we could see that it happens if all the following
conditions are true:
- NFS server daemons [nfsd] are down;
- rpc.mountd is up;
- you run exportfs -a or exportfs -r; and of course
- if you try to mount some volume.
When a client tries to mount a NFS volume, it provokes the following function
calls in knfsd:
exp_rootfh -> fh_put -> add_to_fhcache -> find_fhe
find_fhe must return a file handle that already lies in the cache, OR an empty
space so add_to_fhcache can put the file handle in the cache. But, if NFS
server is down, the cache size becomes zero, and find_fhe does not return any
file handle nor an empty cache entry.
So add_to_fhcache tries to invalidade some old file handle (it will fail
because cache size is zero anyway) and will call find_fhe again, that will fail,
so it will try to invalidate something again and... we have an infinite loop
that locks the server.
The following patch makes add_to_fhcache test if nfsd_nservers is zero; if it
is, it will not loop because it knows the allocation will be unsucessfull
anyway. (It applies for kernel 2.2.14)
--- nfsfh.c.orig Tue Jan 4 16:12:23 2000
+++ nfsfh.c Fri Mar 10 13:23:04 2000
@@ -676,6 +676,15 @@
return 1;
}
+ /* if nfsd_server is zero, NFSD_MAXFH will be zero too, so
+ * find_fhe() will NEVER find the file handle NOR an empty space,
+ * and expire_slot will not be able to expire any file handle,
+ * because NFSD_MAXFH is zero ... */
+
+ if (nfsd_nservers <= 0) {
+ return 0;
+ }
+
expire_slot(cache);
goto repeat;
}
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Mar 15 2000 - 21:00:26 EST