[Linux-cluster] no version for "gfs2_unmount_lockproto"

[Ferenc Wagner]

Bob Peterson <rpeterso at redhat.com> writes:

> On Tue, 2008-02-12 at 19:13 +0100, Wagner Ferenc wrote:
> 
>> I've compiled cluster-2.01.00 against Linux 2.6.23.16.  On modprobe
>> gfs I got the following two kernel messages:
>> 
>> gfs: no version for "gfs2_unmount_lockproto" found: kernel tainted.
>> GFS 2.01.00 (built Feb 12 2008 14:42:50) installed
>
> The HEAD / RHEL5 / (similar) versions of GFS use part of gfs2's
> locking infrastructure.  For RHEL5, we did a patch to export
> those symbols from GFS2.  The patch looks like the one I have
> below. [...]
>
> --- a/fs/gfs2/locking.c	2008-02-11 11:10:57.000000000 -0600
> +++ b/fs/gfs2/locking.c	2008-02-08 14:10:36.000000000 -0600
> @@ -181,4 +181,6 @@ void gfs2_withdraw_lockproto(struct lm_l
>  
>  EXPORT_SYMBOL_GPL(gfs2_register_lockproto);
>  EXPORT_SYMBOL_GPL(gfs2_unregister_lockproto);
> -
> +EXPORT_SYMBOL_GPL(gfs2_withdraw_lockproto);
> +EXPORT_SYMBOL_GPL(gfs2_mount_lockproto);
> +EXPORT_SYMBOL_GPL(gfs2_unmount_lockproto);

Actually, I also patched my kernel tree like this.  In cases when I
forgot it, I wasn't even allowed to load the gfs module into the
kernel.  In this case the "tainted" warning was related to a slight
vermagic mismatch, and after recompiling everything properly, it went
away.

But the issue remained: the mount command just sits there, consuming
some CPU, and by now I've got the following console output (with my
notes in the brackets):

[modprobe gfs]
GFS 2.01.00 (built Feb 12 2008 22:07:48) installed
[starting the cluster infrastructure]
dlm: Using TCP for communications
dlm: connecting to 3
dlm: got connection from 3
[mount /mnt]
Trying to join cluster "lock_dlm", "pilot:test"
Joined cluster. Now mounting FS...
GFS: fsid=pilot:test.4294967295: can't mount journal #4294967295
GFS: fsid=pilot:test.4294967295: there are only 6 journals (0 - 5)
[a couple of minutes passed here]
GFS: fsid=pilot:test.4294967295: Unmount seems to be stalled. Dumping lock state...
Glock (2, 25)
  gl_flags = 
  gl_count = 2
  gl_state = 0
  req_gh = no
  req_bh = no
  lvb_count = 0
  object = yes
  new_le = no
  incore_le = no
  reclaim = no
  aspace = 0
  ail_bufs = no
  Inode:
    num = 25/25
    type = 1
    i_count = 1
    i_flags = 
    vnode = no
Glock (5, 25)
  gl_flags = 
  gl_count = 2
  gl_state = 3
  req_gh = no
  req_bh = no
  lvb_count = 0
  object = yes
  new_le = no
  incore_le = no
  reclaim = no
  aspace = no
  ail_bufs = no
  Holder
    owner = -1
    gh_state = 3
    gh_flags = 5 7 
    error = 0
    gh_iflags = 1 6 7 

Now, mount is still stalled, and still consumes 6% of CPU.
-- 
Regards,
Feri.