[PATCH] [2.6.26] [CAN] Fix can_send() handling on dev_queue_xmit() failures

[Oliver Hartkopp <oliver@hartkopp.net>]

[-- Attachment #1: Type: text/plain, Size: 641 bytes --]

from Oliver Hartkopp <oliver@hartkopp.net>

The tx packet counting and the local loopback of CAN frames should
only happen in the case that the CAN frame has been enqueued to the
netdevice tx queue successfully.

Thanks to Andre Naujoks <nautsch@gmail.com> for reporting this issue.

Signed-off-by: Oliver Hartkopp <oliver@hartkopp.net>
Signed-off-by: Urs Thuermann <urs@isnogud.escape.de>

---

Hello Dave,

this patch should go into 2.6.26 and also applies fine on 2.6.25.1.

IMO this issue is not that 'critical' that it *requires* to become
part of 2.6.25.2. Please forward the patch to Greg at your own opinion.

Many Thanks!
Oliver




[-- Attachment #2: can_send.diff --]
[-- Type: text/x-diff, Size: 1312 bytes --]

diff --git a/net/can/af_can.c b/net/can/af_can.c
index 2759b76..7e8ca28 100644
--- a/net/can/af_can.c
+++ b/net/can/af_can.c
@@ -208,6 +208,7 @@ static int can_create(struct net *net, struct socket *sock, int protocol)
  */
 int can_send(struct sk_buff *skb, int loop)
 {
+	struct sk_buff *newskb = NULL;
 	int err;
 
 	if (skb->dev->type != ARPHRD_CAN) {
@@ -244,8 +245,7 @@ int can_send(struct sk_buff *skb, int loop)
 			 * If the interface is not capable to do loopback
 			 * itself, we do it here.
 			 */
-			struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
-
+			newskb = skb_clone(skb, GFP_ATOMIC);
 			if (!newskb) {
 				kfree_skb(skb);
 				return -ENOMEM;
@@ -254,7 +254,6 @@ int can_send(struct sk_buff *skb, int loop)
 			newskb->sk = skb->sk;
 			newskb->ip_summed = CHECKSUM_UNNECESSARY;
 			newskb->pkt_type = PACKET_BROADCAST;
-			netif_rx(newskb);
 		}
 	} else {
 		/* indication for the CAN driver: no loopback required */
@@ -266,11 +265,20 @@ int can_send(struct sk_buff *skb, int loop)
 	if (err > 0)
 		err = net_xmit_errno(err);
 
+	if (err) {
+		if (newskb)
+			kfree_skb(newskb);
+		return err;
+	}
+
+	if (newskb)
+		netif_rx(newskb);
+
 	/* update statistics */
 	can_stats.tx_frames++;
 	can_stats.tx_frames_delta++;
 
-	return err;
+	return 0;
 }
 EXPORT_SYMBOL(can_send);