ucc_geth: Fix empty TX queue processing
authorAnton Vorontsov <avorontsov@ru.mvista.com>
Thu, 24 Dec 2009 05:31:03 +0000 (05:31 +0000)
committerDavid S. Miller <davem@davemloft.net>
Sun, 27 Dec 2009 04:24:42 +0000 (20:24 -0800)
Following oops was seen with the ucc_geth driver:

 Unable to handle kernel paging request for data at address 0x00000058
 Faulting instruction address: 0xc024f2fc
 Oops: Kernel access of bad area, sig: 11 [#1]
 [...]
 NIP [c024f2fc] skb_recycle_check+0x14/0x100
 LR [e30aa0a4] ucc_geth_poll+0xd8/0x4e0 [ucc_geth_driver]
 Call Trace:
 [df857d50] [c000b03c] __ipipe_grab_irq+0x3c/0xa4 (unreliable)
 [df857d60] [e30aa0a4] ucc_geth_poll+0xd8/0x4e0 [ucc_geth_driver]
 [df857dd0] [c0258cf8] net_rx_action+0xf8/0x1b8
 [df857e10] [c0032a38] __do_softirq+0xb8/0x13c
 [df857e60] [c00065cc] do_softirq+0xa0/0xac
 [...]

This is because ucc_geth_tx() tries to process an empty queue when
queues are logically stopped. Stopping the queues doesn't disable
polling, and since nowadays ucc_geth_tx() is actually called from
the polling routine, the oops above might pop up.

Fix this by removing 'netif_queue_stopped() == 0' check.

Reported-by: Lennart Sorensen <lsorense@csclub.uwaterloo.ca>
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Tested-by: Lennart Sorensen <lsorense@csclub.uwaterloo.ca>
Cc: Stable <stable@vger.kernel.org> [2.6.32]
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ucc_geth.c

index afaf088b72eaab23d1a6847f0e91abcae49ee2d6..0f8c99e845dd3cc98cf9005a59d5cfd4ac6ab729 100644 (file)
@@ -3273,7 +3273,7 @@ static int ucc_geth_tx(struct net_device *dev, u8 txQ)
                /* Handle the transmitted buffer and release */
                /* the BD to be used with the current frame  */
 
-               if ((bd == ugeth->txBd[txQ]) && (netif_queue_stopped(dev) == 0))
+               if (bd == ugeth->txBd[txQ]) /* queue empty? */
                        break;
 
                dev->stats.tx_packets++;