[PATCH] : ir256_sock_connect_cli.diff

From: Jean Tourrilhes <jt@bougret.hpl.hp.com>
To: Linus Torvalds <torvalds@transmeta.com>,
	irda-users@lists.sourceforge.net,
	Linux kernel mailing list <linux-kernel@vger.kernel.org>
Subject: [PATCH] : ir256_sock_connect_cli.diff
Date: Tue, 5 Mar 2002 14:34:31 -0800	[thread overview]
Message-ID: <20020305143431.C1254@bougret.hpl.hp.com> (raw)

ir256_sock_connect_cli.diff :
---------------------------
	o [CRITICA] Fix socket connect to remove dangerous cli()
	<Tested on SMP>

--- linux/net/irda/af_irda.d4.c	Mon Mar  4 14:54:38 2002
+++ linux/net/irda/af_irda.c	Mon Mar  4 17:07:45 2002
@@ -1022,28 +1022,26 @@ static int irda_connect(struct socket *s
 	/* Now the loop */
 	if (sk->state != TCP_ESTABLISHED && (flags & O_NONBLOCK))
 		return -EINPROGRESS;
-		
-	cli();	/* To avoid races on the sleep */
-	
-	/* A Connect Ack with Choke or timeout or failed routing will go to
-	 * closed.  */
+
+	/* Here, there is a race condition : the state may change between
+	 * our test and the sleep, via irda_connect_confirm().
+	 * The way to workaround that is to sleep with a timeout, so that
+	 * we don't sleep forever and check the state when waking up.
+	 * 50ms is plenty good enough, because the LAP is already connected.
+	 * Jean II */
 	while (sk->state == TCP_SYN_SENT) {
-		interruptible_sleep_on(sk->sleep);
+		interruptible_sleep_on_timeout(sk->sleep, HZ/20);
 		if (signal_pending(current)) {
-			sti();
 			return -ERESTARTSYS;
 		}
 	}
 	
 	if (sk->state != TCP_ESTABLISHED) {
-		sti();
 		sock->state = SS_UNCONNECTED;
 		return sock_error(sk);	/* Always set at this point */
 	}
 	
 	sock->state = SS_CONNECTED;
-	
-	sti();
 	
 	/* At this point, IrLMP has assigned our source address */
 	self->saddr = irttp_get_saddr(self->tsap);
