[PATCH] pnfs: layout_driver MUST set .free_deviceid_node if using dev-cache

From: Boaz Harrosh <bharrosh@panasas.com>
To: Benny Halevy <bhalevy@panasas.com>,
	Trond Myklebust <Trond.Myklebust@netapp.com>,
	NFS list <linux-nfs@vger.kernel.org>
Subject: [PATCH] pnfs: layout_driver MUST set .free_deviceid_node if using dev-cache
Date: Tue, 24 May 2011 13:54:32 +0300	[thread overview]
Message-ID: <4DDB8E68.2030504@panasas.com> (raw)
In-Reply-To: <4DDB8DD1.20101@panasas.com>


A device cache is not a matter of memory store. It is a matter
of mounting/login and unmounting/logout. So it is not logical
to not set free_deviceid_node. Who will do the unmount?

It is better to crash the developer then let him leak mounts.

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
 fs/nfs/pnfs_dev.c |   16 ++++------------
 1 files changed, 4 insertions(+), 12 deletions(-)

diff --git a/fs/nfs/pnfs_dev.c b/fs/nfs/pnfs_dev.c
index 37ca215..1b592d9 100644
--- a/fs/nfs/pnfs_dev.c
+++ b/fs/nfs/pnfs_dev.c
@@ -163,10 +163,7 @@ nfs4_delete_deviceid(const struct pnfs_layoutdriver_type *ld,
 	d = nfs4_unhash_put_deviceid(ld, clp, id);
 	if (!d)
 		return;
-	if (d->ld->free_deviceid_node)
-		d->ld->free_deviceid_node(d);
-	else
-		kfree(d);
+	d->ld->free_deviceid_node(d);
 }
 EXPORT_SYMBOL_GPL(nfs4_delete_deviceid);
 
@@ -232,8 +229,7 @@ nfs4_put_deviceid_node(struct nfs4_deviceid_node *d)
 	hlist_del_init_rcu(&d->node);
 	spin_unlock(&nfs4_deviceid_lock);
 	synchronize_rcu();
-	if (d->ld->free_deviceid_node)
-		d->ld->free_deviceid_node(d);
+	d->ld->free_deviceid_node(d);
 	return true;
 }
 EXPORT_SYMBOL_GPL(nfs4_put_deviceid_node);
@@ -258,12 +254,8 @@ _deviceid_purge_client(const struct nfs_client *clp, long hash)
 
 	synchronize_rcu();
 	hlist_for_each_entry_safe(d, n, next, &tmp, node)
-		if (atomic_dec_and_test(&d->ref)) {
-			if (d->ld->free_deviceid_node)
-				d->ld->free_deviceid_node(d);
-			else
-				kfree(d);
-		}
+		if (atomic_dec_and_test(&d->ref))
+			d->ld->free_deviceid_node(d);
 }
 
 void
-- 
1.7.2.3


