From: Greg Banks <gnb@melbourne.sgi.com>
Subject: [PATCH 005 of 11] knfsd: test and set SK_BUSY atomically
Date: Tue, 25 Jul 2006 15:09:53 +1000
Message-ID: <1153804193.21040.11.camel@hole.melbourne.sgi.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Cc: Linux NFS Mailing List <nfs@lists.sourceforge.net>
To: Neil Brown <neilb@suse.de>
Sender: nfs-bounces@lists.sourceforge.net
Errors-To: nfs-bounces@lists.sourceforge.net

knfsd: The SK_BUSY bit in svc_sock->sk_flags ensures that we do not
attempt to enqueue a socket twice.  Currently, setting and clearing
the bit is protected by svc_serv->sv_lock.  As I intend to reduce the
data that the lock protects so it's not held when svc_sock_enqueue()
tests and sets SK_BUSY, that test and set needs to be atomic.

Signed-off-by: Greg Banks <gnb@melbourne.sgi.com>
---

 net/sunrpc/svcsock.c |   18 +++++++++---------
 1 files changed, 9 insertions(+), 9 deletions(-)

Index: linus-git/net/sunrpc/svcsock.c
===================================================================
--- linus-git.orig/net/sunrpc/svcsock.c	2006-07-24 17:16:25.698821516 +1000
+++ linus-git/net/sunrpc/svcsock.c	2006-07-24 17:20:26.031867832 +1000
@@ -47,12 +47,11 @@
  *
  * 	svc_serv->sv_lock protects most stuff for that service.
  *	svc_sock->sk_defer_lock protects the svc_sock->sk_deferred list
+ *      svc_sock->sk_flags.SK_BUSY prevents a svc_sock being enqueued multiply.
  *
  *	Some flags can be set to certain values at any time
  *	providing that certain rules are followed:
  *
- *	SK_BUSY  can be set to 0 at any time.  
- *		svc_sock_enqueue must be called afterwards
  *	SK_CONN, SK_DATA, can be set or cleared at any time.
  *		after a set, svc_sock_enqueue must be called.	
  *		after a clear, the socket must be read/accepted
@@ -169,8 +168,13 @@ svc_sock_enqueue(struct svc_sock *svsk)
 		goto out_unlock;
 	}
 
-	if (test_bit(SK_BUSY, &svsk->sk_flags)) {
-		/* Don't enqueue socket while daemon is receiving */
+	/* Mark socket as busy. It will remain in this state until the
+	 * server has processed all pending data and put the socket back
+	 * on the idle list.  We update SK_BUSY atomically because
+	 * it also guards against trying to enqueue the svc_sock twice.
+	 */
+	if (test_and_set_bit(SK_BUSY, &svsk->sk_flags)) {
+		/* Don't enqueue socket while already enqueued */
 		dprintk("svc: socket %p busy, not enqueued\n", svsk->sk_sk);
 		goto out_unlock;
 	}
@@ -184,15 +188,11 @@ svc_sock_enqueue(struct svc_sock *svsk)
 		dprintk("svc: socket %p  no space, %d*2 > %ld, not enqueued\n",
 			svsk->sk_sk, atomic_read(&svsk->sk_reserved)+serv->sv_bufsz,
 			svc_sock_wspace(svsk));
+		clear_bit(SK_BUSY, &svsk->sk_flags);
 		goto out_unlock;
 	}
 	clear_bit(SOCK_NOSPACE, &svsk->sk_sock->flags);
 
-	/* Mark socket as busy. It will remain in this state until the
-	 * server has processed all pending data and put the socket back
-	 * on the idle list.
-	 */
-	set_bit(SK_BUSY, &svsk->sk_flags);
 
 	if (!list_empty(&serv->sv_threads)) {
 		rqstp = list_entry(serv->sv_threads.next,

-- 
Greg Banks, R&D Software Engineer, SGI Australian Software Group.
I don't speak for SGI.


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs