GFS inside Xen

Bojan Smojver bojan at
Fri Feb 25 03:29:37 UTC 2005

Quoting Bojan Smojver <bojan at>:

> This may sound like a very strange thing to do, but it may be useful in
> situations where you have only one physical machine and want to play with GFS
> based clusters. So the real question is, did anyone try this (with current
> development branch) and if yes, was there some kind of "magic" setup involved
> or did it just work?

OK, kernel 2.6.10-1.1149_FC4xenU, got GFS to compile with a minor patch:

diff -ruN gfs-kernel-2.6.9-5-vanilla/src/gfs/quota.c gfs-kernel-2.6.9-5/src/gfs/
--- gfs-kernel-2.6.9-5-vanilla/src/gfs/quota.c  2004-12-20 10:14:33.000000000 -0
+++ gfs-kernel-2.6.9-5/src/gfs/quota.c  2005-02-24 20:28:07.000000000 -0500
@@ -954,7 +954,7 @@
        if (current->signal) {
                tty = current->signal->tty;
                if (tty && tty->driver->write)
-                       tty->driver->write(tty, 0, line, len);
+                       tty->driver->write(tty, line, len);


GFS loads and I can make the file system, no worries. However, mounting still
doesn't work, unless lock_nolock is used (which is kinda pointless for what I'm
trying to do):

mount -v -t gfs /dev/sda2 /gfs
GFS: Trying to join cluster "lock_gulm", "cluster:gfs"
lock_gulm: Unknown symbol in6addr_loopback
lock_harness:  can't find protocol lock_gulm
GFS: can't mount proto = lock_gulm, table = cluster:gfs, hostdata =
mount: No such file or directory

I'm guessing recompile of the xenU kernel is in order to get the correct symbols
exported out of ipv6?


More information about the fedora-devel-list mailing list