All of lore.kernel.org
 help / color / mirror / Atom feed
From: Alexandros Batsakis <batsakis@netapp.com>
To: linux-nfs@vger.kernel.org
Cc: trond@netapp.com, root <vadmin.root@vadmin (none) >,
	Alexandros Batsakis <batsakis@netapp.com>
Subject: [PATCH 2/5] nfs: prevent backlogging of renewd requests
Date: Fri,  5 Feb 2010 03:32:31 -0800	[thread overview]
Message-ID: <1265369554-2105-3-git-send-email-batsakis@netapp.com> (raw)
In-Reply-To: <1265369554-2105-2-git-send-email-batsakis@netapp.com>

From: root <root@vadmin.(none)>

If the renewd send queue gets backlogged (e.g., if the server goes down),
we will keep filling the queue with periodic RENEW/SEQUENCE requests.

This patch schedules a new renewd request if and only if the previous one
returns (either success or failure)

Signed-off-by: Alexandros Batsakis <batsakis@netapp.com>
---
 fs/nfs/nfs4proc.c   |    3 +++
 fs/nfs/nfs4renewd.c |   24 +++++++-----------------
 2 files changed, 10 insertions(+), 17 deletions(-)

diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 84b53d3..ce44b5a 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -3162,6 +3162,7 @@ static void nfs4_renew_done(struct rpc_task *task, void *data)
 	if (time_before(clp->cl_last_renewal,timestamp))
 		clp->cl_last_renewal = timestamp;
 	spin_unlock(&clp->cl_lock);
+	nfs4_schedule_state_renewal(clp);
 }
 
 static const struct rpc_call_ops nfs4_renew_ops = {
@@ -5040,6 +5041,8 @@ void nfs41_sequence_call_done(struct rpc_task *task, void *data)
 	}
 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
 
+	nfs4_schedule_state_renewal(clp);
+
 	kfree(task->tk_msg.rpc_argp);
 	kfree(task->tk_msg.rpc_resp);
 
diff --git a/fs/nfs/nfs4renewd.c b/fs/nfs/nfs4renewd.c
index 0156c01..d87f103 100644
--- a/fs/nfs/nfs4renewd.c
+++ b/fs/nfs/nfs4renewd.c
@@ -36,11 +36,6 @@
  * as an rpc_task, not a real kernel thread, so it always runs in rpciod's
  * context.  There is one renewd per nfs_server.
  *
- * TODO: If the send queue gets backlogged (e.g., if the server goes down),
- * we will keep filling the queue with periodic RENEW requests.  We need a
- * mechanism for ensuring that if renewd successfully sends off a request,
- * then it only wakes up when the request is finished.  Maybe use the
- * child task framework of the RPC layer?
  */
 
 #include <linux/mm.h>
@@ -63,7 +58,7 @@ nfs4_renew_state(struct work_struct *work)
 	struct nfs_client *clp =
 		container_of(work, struct nfs_client, cl_renewd.work);
 	struct rpc_cred *cred;
-	long lease, timeout;
+	long lease;
 	unsigned long last, now;
 
 	ops = nfs4_state_renewal_ops[clp->cl_minorversion];
@@ -75,7 +70,6 @@ nfs4_renew_state(struct work_struct *work)
 	lease = clp->cl_lease_time;
 	last = clp->cl_last_renewal;
 	now = jiffies;
-	timeout = (2 * lease) / 3 + (long)last - (long)now;
 	/* Are we close to a lease timeout? */
 	if (time_after(now, last + lease/3)) {
 		cred = ops->get_state_renewal_cred_locked(clp);
@@ -90,19 +84,15 @@ nfs4_renew_state(struct work_struct *work)
 			/* Queue an asynchronous RENEW. */
 			ops->sched_state_renewal(clp, cred);
 			put_rpccred(cred);
+			goto out_exp;
 		}
-		timeout = (2 * lease) / 3;
-		spin_lock(&clp->cl_lock);
-	} else
+	} else {
 		dprintk("%s: failed to call renewd. Reason: lease not expired \n",
 				__func__);
-	if (timeout < 5 * HZ)    /* safeguard */
-		timeout = 5 * HZ;
-	dprintk("%s: requeueing work. Lease period = %ld\n",
-			__func__, (timeout + HZ - 1) / HZ);
-	cancel_delayed_work(&clp->cl_renewd);
-	schedule_delayed_work(&clp->cl_renewd, timeout);
-	spin_unlock(&clp->cl_lock);
+		spin_unlock(&clp->cl_lock);
+	}
+	nfs4_schedule_state_renewal(clp);
+out_exp:
 	nfs_expire_unreferenced_delegations(clp);
 out:
 	dprintk("%s: done\n", __func__);
-- 
1.6.2.5


  reply	other threads:[~2010-02-11 20:41 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-02-05 11:32 [PATCH 0/5] nfs: fix umount - renewd race Alexandros Batsakis
2010-02-05 11:32 ` [PATCH 1/5] nfs: kill renewd before clearing client minor version Alexandros Batsakis
2010-02-05 11:32   ` Alexandros Batsakis [this message]
2010-02-05 11:32     ` [PATCH 3/5] nfs41: renewd sequence operations should take/put client reference Alexandros Batsakis
2010-02-05 11:32       ` [PATCH 4/5] nfs4: renewd renew operations should take/put a " Alexandros Batsakis
2010-02-05 11:32         ` [PATCH 5/5] nfs4: adjust rpc timeout for nfs_client->rpcclient based on the lease_time Alexandros Batsakis

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1265369554-2105-3-git-send-email-batsakis@netapp.com \
    --to=batsakis@netapp.com \
    --cc=linux-nfs@vger.kernel.org \
    --cc=trond@netapp.com \
    --cc=vadmin.root@vadmin \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.