[Linux-cluster] no version for "gfs2_unmount_lockproto"

Steven Whitehouse swhiteho at redhat.com
Tue Feb 12 19:52:05 UTC 2008


Hi,

On Tue, 2008-02-12 at 13:06 -0600, Bob Peterson wrote:
> On Tue, 2008-02-12 at 12:46 -0600, Brian Kroth wrote:
> > I'm curious.  Are there plans to include this patch in the mainline?  Any
> > reasons why or why not?
> > 
> > Thanks,
> > Brian
> 
> Hi Brian,
> 
> That's a gfs2 patch, and gfs2 "mainline" is the upstream git tree from
> kernel.org.  We've never tried to add that patch to the upstream kernel
> because we figured it would be rejected by the upstream community.
> Those people likely wouldn't accept the suggestion that it should be
> added "just in case someone wants to run out-of-tree gfs on top of
> gfs2."  After all, the patch is really only useful if gfs (vers. 1) is
> required, but gfs is not part of that upstream kernel.
> 
> >From what I've heard, the Fedora community doesn't really accept
> anything along those lines anymore either.  So far they've let us get
> away with it, but I've heard that it's unacceptable for us going
> forward (Fedora 9 or 10?  I'm not sure where the cutoff is). So lately
> we've been discussing whether/how to separate out the gfs1 and gfs2
> infrastructure so that there isn't that crossover.  That issue is still
> work-in-progress.
> 
> Regards,
> 
> Bob Peterson
> Red Hat GFS
> 
Just to confirm some of the details, F-8 is the last one which will have
our exports in it. The reason that we haven't pushed the export patch
upstream is that in the future for GFS2 we intend to get rid of that
particular interface, so we didn't want to export it in case someone
else started using it and thus prevented or slowed that process. Also
exports for use by out-of-tree code tend to be frowned upon.

The plan is that the interface between the filesystem and the lock
manager will be the DLM's interface, so we will still allow pluggable
lock managers, but just at a slightly different level in the code.

So at some stage we'll need to duplicate the lock modules for GFS
anyway, and about now seems as good a time as any to start that process,

Steve.






More information about the Linux-cluster mailing list