This patch fixes the poll routine for the natsemi driver so that if the
driver detects an RX state machine lockup then no interrupts will be
lost while the driver recovers from that.
Signed-Off-By: Mark Brown <broonie@sirena.org.uk>
Index: linux-2.6/drivers/net/natsemi.c
===================================================================
--- linux-2.6.orig/drivers/net/natsemi.c 2007-03-13 19:38:31.000000000 +0000
+++ linux-2.6/drivers/net/natsemi.c 2007-03-13 19:39:08.000000000 +0000
@@ -2169,6 +2169,14 @@
dev->name, np->intr_status,
readl(ioaddr + IntrMask));
+ /* netdev_rx() may read IntrStatus again if the RX state
+ * machine falls over so do it first. */
+ if (np->intr_status &
+ (IntrRxDone | IntrRxIntr | RxStatusFIFOOver |
+ IntrRxErr | IntrRxOverrun)) {
+ netdev_rx(dev, &work_done, work_to_do);
+ }
+
if (np->intr_status &
(IntrTxDone | IntrTxIntr | IntrTxIdle | IntrTxErr)) {
spin_lock(&np->lock);
@@ -2180,12 +2188,6 @@
if (np->intr_status & IntrAbnormalSummary)
netdev_error(dev, np->intr_status);
- if (np->intr_status &
- (IntrRxDone | IntrRxIntr | RxStatusFIFOOver |
- IntrRxErr | IntrRxOverrun)) {
- netdev_rx(dev, &work_done, work_to_do);
- }
-
*budget -= work_done;
dev->quota -= work_done;
--
"You grabbed my hand and we fell into it, like a daydream - or a fever."
-
