[PATCH 2/4] forcedeth: xmit lock fix

[PATCH 2/4] forcedeth: xmit lock fix

[Ayaz Abdulla]

[-- Attachment #1: Type: text/plain, Size: 386 bytes --]

This patch fixes a potential race condition between xmit thread and xmit 
completion thread. The calculation of empty tx descriptors is not 
performed under the lock. This could cause it to set the stop flag while 
the completion thread finishes all tx's. This will result in the tx 
queue in stopped state and no one to wake it up.

Signed-off-by: Ayaz Abdulla <aabdulla@nvidia.com>



[-- Attachment #2: patch-forcedeth-xmitlockfix --]
[-- Type: text/plain, Size: 1123 bytes --]

--- old/drivers/net/forcedeth.c	2009-01-09 14:51:19.000000000 -0800
+++ new/drivers/net/forcedeth.c	2009-01-09 15:10:15.000000000 -0800
@@ -2096,14 +2096,15 @@
 			   ((skb_shinfo(skb)->frags[i].size & (NV_TX2_TSO_MAX_SIZE-1)) ? 1 : 0);
 	}
 
+	spin_lock_irqsave(&np->lock, flags);
 	empty_slots = nv_get_empty_tx_slots(np);
 	if (unlikely(empty_slots <= entries)) {
-		spin_lock_irqsave(&np->lock, flags);
 		netif_stop_queue(dev);
 		np->tx_stop = 1;
 		spin_unlock_irqrestore(&np->lock, flags);
 		return NETDEV_TX_BUSY;
 	}
+	spin_unlock_irqrestore(&np->lock, flags);
 
 	start_tx = put_tx = np->put_tx.orig;
 
@@ -2214,14 +2215,15 @@
 			   ((skb_shinfo(skb)->frags[i].size & (NV_TX2_TSO_MAX_SIZE-1)) ? 1 : 0);
 	}
 
+	spin_lock_irqsave(&np->lock, flags);
 	empty_slots = nv_get_empty_tx_slots(np);
 	if (unlikely(empty_slots <= entries)) {
-		spin_lock_irqsave(&np->lock, flags);
 		netif_stop_queue(dev);
 		np->tx_stop = 1;
 		spin_unlock_irqrestore(&np->lock, flags);
 		return NETDEV_TX_BUSY;
 	}
+	spin_unlock_irqrestore(&np->lock, flags);
 
 	start_tx = put_tx = np->put_tx.ex;
 	start_tx_ctx = np->put_tx_ctx;

Re: [PATCH 2/4] forcedeth: xmit lock fix

[David Miller]

From: Ayaz Abdulla <aabdulla@nvidia.com>
Date: Fri, 09 Jan 2009 16:03:44 -0500

> This patch fixes a potential race condition between xmit thread and xmit completion thread. The calculation of empty tx descriptors is not performed under the lock. This could cause it to set the stop flag while the completion thread finishes all tx's. This will result in the tx queue in stopped state and no one to wake it up.
> 
> Signed-off-by: Ayaz Abdulla <aabdulla@nvidia.com>

Applied.