Patchwork r8169: fix problem with TSO (TX_BUFFS_AVAIL negative value)

login
register
mail settings
Submitter françois romieu
Date May 7, 2012, 11:42 p.m.
Message ID <20120507234205.GA2230@electric-eye.fr.zoreil.com>
Download mbox | patch
Permalink /patch/157490/
State RFC
Delegated to: David Miller
Headers show

Comments

françois romieu - May 7, 2012, 11:42 p.m.
Julien Ducourthial <jducourt@free.fr> :
> The r8169 may get stuck or show bad behaviour after activating TSO :
> the net_device is not stopped when it has no more TX descriptors.
> This problem comes from TX_BUFS_AVAIL which may reach -1 when all
> transmit descriptors are in use. The patch simply tries to keep positive
> values.

It seems more than good.

Alex, Thomas, can you check if Julien's patch below fixes your broken
kernels as well ? 

    The RTL chips use a 64 element hash table based on the Ethernet CRC.
*/
@@ -5115,7 +5119,7 @@ static netdev_tx_t rtl8169_start_xmit(struct
sk_buff *skb,
 	u32 opts[2];
 	int frags;
 
-	if (unlikely(TX_BUFFS_AVAIL(tp) < skb_shinfo(skb)->nr_frags)) {
+	if (unlikely(!TX_FRAGS_READY_FOR(tp, skb_shinfo(skb)->nr_frags))) {
 		netif_err(tp, drv, dev, "BUG! Tx Ring full when queue awake!\n");
 		goto err_stop_0;
 	}
@@ -5169,7 +5173,7 @@ static netdev_tx_t rtl8169_start_xmit(struct
sk_buff *skb,
 
 	mmiowb();
 
-	if (TX_BUFFS_AVAIL(tp) < MAX_SKB_FRAGS) {
+	if (!TX_FRAGS_READY_FOR(tp, MAX_SKB_FRAGS)) {
 		/* Avoid wrongly optimistic queue wake-up: rtl_tx thread must
 		 * not miss a ring update when it notices a stopped queue.
 		 */
@@ -5183,7 +5187,7 @@ static netdev_tx_t rtl8169_start_xmit(struct
sk_buff *skb,
 		 * can't.
 		 */
 		smp_mb();
-		if (TX_BUFFS_AVAIL(tp) >= MAX_SKB_FRAGS)
+		if (TX_FRAGS_READY_FOR(tp, MAX_SKB_FRAGS))
 			netif_wake_queue(dev);
 	}
 
@@ -5306,7 +5310,7 @@ static void rtl_tx(struct net_device *dev, struct
rtl8169_private *tp)
 		 */
 		smp_mb();
 		if (netif_queue_stopped(dev) &&
-		    (TX_BUFFS_AVAIL(tp) >= MAX_SKB_FRAGS)) {
+		    TX_FRAGS_READY_FOR(tp, MAX_SKB_FRAGS)) {
 			netif_wake_queue(dev);
 		}
 		/*
Alex Villacís Lasso - May 9, 2012, 4:12 a.m.
El 07/05/12 18:42, Francois Romieu escribió:
> Julien Ducourthial<jducourt@free.fr>  :
>> The r8169 may get stuck or show bad behaviour after activating TSO :
>> the net_device is not stopped when it has no more TX descriptors.
>> This problem comes from TX_BUFS_AVAIL which may reach -1 when all
>> transmit descriptors are in use. The patch simply tries to keep positive
>> values.
> It seems more than good.
>
> Alex, Thomas, can you check if Julien's patch below fixes your broken
> kernels as well ?
>
No luck. The backtrace still appears after using the patched driver.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/drivers/net/ethernet/realtek/r8169.c
b/drivers/net/ethernet/realtek/r8169.c
index f545093..d1e3c51 100644
--- a/drivers/net/ethernet/realtek/r8169.c
+++ b/drivers/net/ethernet/realtek/r8169.c
@@ -61,8 +61,12 @@ 
 #define R8169_MSG_DEFAULT \
 	(NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN)
 
-#define TX_BUFFS_AVAIL(tp) \
-	(tp->dirty_tx + NUM_TX_DESC - tp->cur_tx - 1)
+#define TX_SLOTS_AVAIL(tp) \
+	(tp->dirty_tx + NUM_TX_DESC - tp->cur_tx)
+
+/* A skbuff with nr_frags needs nr_frags+1 entries in the tx queue */
+#define TX_FRAGS_READY_FOR(tp,nr_frags) \
+	(TX_SLOTS_AVAIL(tp) >= (nr_frags + 1))
 
 /* Maximum number of multicast addresses to filter (vs.
Rx-all-multicast).