Re: [lock validator] inet6_destroy_sock(): soft-safe -> soft-unsafe lock dependency

[Ingo Molnar <mingo@elte.hu>]

* Ingo Molnar <mingo@elte.hu> wrote:

> hm, i got a new one:
> 
> ============================================
> [ BUG: circular locking deadlock detected! ]
> --------------------------------------------
> sshd/28997 is trying to acquire lock:
>  (&sk->sk_lock.slock){-+}, at: [<c0c6be28>] packet_rcv+0xbf/0x34b
> 
> but task is already holding lock:
>  (&dev->xmit_lock){-+}, at: [<c0bb04ec>] qdisc_restart+0x46/0x207
> 
> which lock already depends on the new lock,
> which could lead to circular deadlocks!
> 
> the dependency chain (in reverse order) is:
> -> #2 (&dev->xmit_lock){-+}: [<c0bb04ec>] qdisc_restart+0x46/0x207
> -> #1 (&dev->queue_lock){-+}: [<c0b98137>] dev_queue_xmit+0xc3/0x21e
> -> #0 (&sk->sk_lock.slock){-+}: [<c0c6be28>] packet_rcv+0xbf/0x34b

i think this might be a false positive, caused by my (ill-advised) 
change below? [i did the change to clean up an unlock ordering 
assymetry, but apparently i thus also introduced the xmit_lock -> 
queue_lock dependency.]

	Ingo

Index: linux/net/sched/sch_generic.c
===================================================================
--- linux.orig/net/sched/sch_generic.c
+++ linux/net/sched/sch_generic.c
@@ -98,6 +98,8 @@ int qdisc_restart(struct net_device *dev
 	/* Dequeue packet */
 	if ((skb = q->dequeue(q)) != NULL) {
 		unsigned nolock = (dev->features & NETIF_F_LLTX);
+		/* release queue */
+		spin_unlock(&dev->queue_lock);
 		/*
 		 * When the driver has LLTX set it does its own locking
 		 * in start_xmit. No need to add additional overhead by
@@ -121,9 +123,11 @@ int qdisc_restart(struct net_device *dev
 					kfree_skb(skb);
 					if (net_ratelimit())
 						printk(KERN_DEBUG "Dead loop on netdevice %s, fix it urgently!\n", dev->name);
+					spin_lock(&dev->queue_lock);
 					return -1;
 				}
 				__get_cpu_var(netdev_rx_stat).cpu_collision++;
+				spin_lock(&dev->queue_lock);
 				goto requeue;
 			}
 			/* Remember that the driver is grabbed by us. */
@@ -131,8 +135,6 @@ int qdisc_restart(struct net_device *dev
 		}
 		
 		{
-			/* And release queue */
-			spin_unlock(&dev->queue_lock);
 
 			if (!netif_queue_stopped(dev)) {
 				int ret;
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/