[pnfs] callback problem

[William A. (Andy) Adamson]

i built the 4.1-sessions branch on two laptops - i couldn't mount due to a
client hang which turns out to be the runtime lock correctness validator
(kernel/lockdep.c) printing in a loop - all i get on the console is 24 equal
signs over and over..

========================

this is triggered in fs/nfs/nfs4proc.c:nfs4_proc_setup_sessions() by the rpc
call to exchange_id.

here is the locking in nfs4_fill_super and sessions...

nfs4_fill_super
        lock_kernel()
        down_write(&server->nfs4_state->cl_sem);
        server->rpc_ops->setup_session()
                down(&clp->cl_session->session_sem);

/* if i bail here, no validator problem */
                exchange_id (rpc_call_sync)
/* if i try to bail here, client hangs with repeated validator printk's */

                create-session (rpc_call_sync)
                nfs4_proc_get_lease_time()
                        _nfs41_proc_setup_sequence
                        nfs4_find_slot
                                spin_lock(&slot->slot_lock);
                                spin_unlock(&slot->slot_lock
                                perhaps a wait_on_bit
                        rpc_call_sync
                        nfs41_proc_sequence_done
                                spin_lock(&clp->cl_lock);
                                set renewal
                                spin_unlock(&clp->cl_lock);
                                wake_up_bit
                spin_lock(&clp->cl_lock)
                set lease time
                spin_unlock(&clp->cl_lock);
                nfs4_schedule_state_renewal
                        spin_lock(&clp->cl_lock);
                        spin_unlock(&clp->cl_lock);
                up(&clp->cl_session->session_sem);
        up_write(&server->nfs4_state->cl_sem);
        unlock_kernel();


i don't know if the runtime validator has a bug, or if our code has a
bug....

i'm turning off the validator and will try to mount.

-->Andy



On 5/23/07, Marc Eshel <eshel at almaden.ibm.com> wrote:
>
> Trying to mount from a pNFS client I get the following trace on the server
> and the client hangs. It looks like the rpc_clnt was used before it was
> initialized as you can tell from the "nfs4_cb: server 0.0.0.0" msg. Was
> anyone able to mount with the latest code ?
> Marc.
>
> May 23 07:50:53 fin18 kernel: NFSD: move_to_confirm nfs4_client f6473400
> May 23 07:50:53 fin18 kernel: renewing client (clientid 465454c2/00000001)
> May 23 07:50:53 fin18 kernel: NFSD:     looking up UNIX cred
> May 23 07:50:53 fin18 kernel: hash_sessionid IDX: 412 csum 1587934240
> May 23 07:50:54 fin18 kernel: nfsd4_create_session returns 0 0
> May 23 07:50:58 fin18 kernel: nfs4_cb: server 0.0.0.0 not responding,
> timed out
> May 23 07:50:58 fin18 kernel: NFSD: nfs4_cb_null task->tk_status -5
> May 23 07:50:59 fin18 kernel: NFSD: callback establishment to client
> 9.1.72.239/9.1.72.227 AUTH_SYS 0 failed
>
> _______________________________________________
> pNFS mailing list
> pNFS at linux-nfs.org
> http://linux-nfs.org/cgi-bin/mailman/listinfo/pnfs
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://linux-nfs.org/pipermail/pnfs/attachments/20070523/02249ba6/attachment.htm