[PATCH] macb: Don't re-enable interrupts while in polling mode

[Joshua Hoke]

From: Joshua Hoke <joshua.hoke@xxxxxxxxxx>

[PATCH] macb: Don't re-enable interrupts while in polling mode

On a busy network, the macb driver could get stuck in the interrupt
handler, quickly triggering the watchdog, due to a confluence of
factors:

 1. macb_poll re-enables interrupts unconditionally, even when it will
    be called again because it exhausted its rx budget

 2. macb_interrupt only disables interrupts after scheduling
    macb_poll, but scheduling fails when macb_poll is already scheduled
    because it didn't call napi_complete

 3. macb_interrupt loops until the interrupt status register is clear,
    which will never happen in this case if the driver doesn't disable
    the RX interrupt

Since macb_interrupt runs in interrupt context, this effectively locks
up the machine, triggering the hardware watchdog.

This issue was readily reproducible on a flooded network with a
modified 2.6.27.48 kernel. The same problem appears to still be in the
2.6.36-rc8 driver code, so I am submitting this patch against that
version. I have not tested this version of the patch except to make
sure the kernel compiles.

Signed-off-by: Joshua Hoke <joshua.hoke@xxxxxxxxxx>

---

I'm submitting this at the request of Andrew Morton in this bug:

  https://bugzilla.kernel.org/show_bug.cgi?id=20732

This version of the patch applies to 2.6.36-rc8 but has not been
tested. In particular I am assuming that napi_schedule_prep() behaves
the same as netif_rx_schedule_prep() did by failing when the macb_poll
callback is already scheduled.

diff --git a/drivers/net/macb.c b/drivers/net/macb.c
index ff2f158..36cf594 100644
--- a/drivers/net/macb.c
+++ b/drivers/net/macb.c
@@ -515,14 +515,15 @@ static int macb_poll(struct napi_struct *napi, int
budget)
 		(unsigned long)status, budget);
 
 	work_done = macb_rx(bp, budget);
-	if (work_done < budget)
+	if (work_done < budget) {
 		napi_complete(napi);
 
-	/*
-	 * We've done what we can to clean the buffers. Make sure we
-	 * get notified when new packets arrive.
-	 */
-	macb_writel(bp, IER, MACB_RX_INT_FLAGS);
+		/*
+		 * We've done what we can to clean the buffers. Make
sure we
+		 * get notified when new packets arrive.
+		 */
+		macb_writel(bp, IER, MACB_RX_INT_FLAGS);
+	}
 
 	/* TODO: Handle errors */
 
@@ -550,12 +551,16 @@ static irqreturn_t macb_interrupt(int irq, void
*dev_id)
 		}
 
 		if (status & MACB_RX_INT_FLAGS) {
+			/*
+			 * There's no point taking any more interrupts
+			 * until we have processed the buffers. The
+			 * scheduling call may fail if the poll routine
+			 * is already scheduled, so disable interrupts
+			 * now.
+			 */
+			macb_writel(bp, IDR, MACB_RX_INT_FLAGS);
+
 			if (napi_schedule_prep(&bp->napi)) {
-				/*
-				 * There's no point taking any more
interrupts
-				 * until we have processed the buffers
-				 */
-				macb_writel(bp, IDR, MACB_RX_INT_FLAGS);
 				dev_dbg(&bp->pdev->dev,
 					"scheduling RX softirq\n");
 				__napi_schedule(&bp->napi);
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/