[pnfs] nfs_callback_down in nfs4_fill_super

[William A. (Andy) Adamson]

On 5/23/07, William A. (Andy) Adamson <andros at citi.umich.edu> wrote:
>
>
>
> On 5/23/07, Iyer, Rahul <Rahul.Iyer at netapp.com> wrote:
> >
> > Well, actually, it is. This is what deallocates the mempool that was
> > created in the xprt_setup_backchannel that was called in
> > nfs_callback_up(). If it works for now, remove the call and I'll see why
> > it's bombing.
>
>
> i think this is what is happening.
> xprt_setup_backchannel does not mempool_alloc an rpc_rqst buffer for
> xprt->bc_mempool.
> but nfs_callback_down calls xprot_destroy_backchannel which tries to
> mempool_destroy the bc_mempool and thus the BUG_ON in mempool_destroy.
>


hmm what i said above must be wrong{ it would seem likely that a
mempool_create (called in xprt_setup_backchannel) followed by a
mempool_destroy (called in nfs_callback_down)  would work. does
nfs41_callback_up call mempool_alloc?

-->Andy

-->Andy
>
> Regards
> > Rahul
> >
> >
> > > -----Original Message-----
> > > From: William A. (Andy) Adamson [mailto:andros at citi.umich.edu]
> > > Sent: Wednesday, May 23, 2007 12:49 PM
> > > To: Iyer, Rahul; pnfs at linux-nfs.org
> > > Subject: nfs_callback_down in nfs4_fill_super
> > >
> > > Hi Rahul
> > >
> > > I'm in the sessions branch, where nothing is working (!). The
> > > client is in nfs4_fill_super() calling
> > > rpc_ops->setup_session, which fails on exchange_id.
> > >
> > >               if (server->rpc_ops->setup_session) {
> > >                         lock_kernel();
> > >                         down_write(&server->nfs4_state->cl_sem);
> > >                         err =
> > > server->rpc_ops->setup_session(server->nfs4_state);
> > >                         up_write(&server->nfs4_state->cl_sem);
> > >                         unlock_kernel();
> > >
> > >                         if (err) {
> > >                                 printk(KERN_EMERG "Couldn't
> > > mount using minorversion %d\n", i);
> > >                                 /*
> > >                                  * Shut down the callback service
> > >                                  */
> > >                                 nfs_callback_down(i,
> > > server->client->cl_xprt);
> > >
> > >                                 rpc_shutdown_client(server->client);
> > >                         }
> > >
> > > The client hits a BUG in mm/mempool due to the call to
> > > nfs_callback_down. I don't think the callback channel has not
> > > been created, so this call is not needed. Correct?
> > >
> > > -->Andy
> > >
> > >
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://linux-nfs.org/pipermail/pnfs/attachments/20070523/78668824/attachment.htm