Merge [email protected]:/pub/scm/linux/kernel/git/davem/net-2.6

Documentation/networking/ip-sysctl.txt

@@ -294,15 +294,15 @@ tcp_rmem - vector of 3 INTEGERs: min, default, max
 	Default: 87380*2 bytes.
 
 tcp_mem - vector of 3 INTEGERs: min, pressure, max
-	low: below this number of pages TCP is not bothered about its
+	min: below this number of pages TCP is not bothered about its
 	memory appetite.
 
 	pressure: when amount of memory allocated by TCP exceeds this number
 	of pages, TCP moderates its memory consumption and enters memory
 	pressure mode, which is exited when memory consumption falls
-	under "low".
+	under "min".
 
-	high: number of pages allowed for queueing by all TCP sockets.
+	max: number of pages allowed for queueing by all TCP sockets.
 
 	Defaults are calculated at boot time from amount of available
 	memory.

drivers/net/bnx2.c

@@ -56,8 +56,8 @@
 
 #define DRV_MODULE_NAME		"bnx2"
 #define PFX DRV_MODULE_NAME	": "
-#define DRV_MODULE_VERSION	"1.4.43"
-#define DRV_MODULE_RELDATE	"June 28, 2006"
+#define DRV_MODULE_VERSION	"1.4.44"
+#define DRV_MODULE_RELDATE	"August 10, 2006"
 
 #define RUN_AT(x) (jiffies + (x))
 
@@ -209,8 +209,10 @@ MODULE_DEVICE_TABLE(pci, bnx2_pci_tbl);
 
 static inline u32 bnx2_tx_avail(struct bnx2 *bp)
 {
-	u32 diff = TX_RING_IDX(bp->tx_prod) - TX_RING_IDX(bp->tx_cons);
+	u32 diff;
 
+	smp_mb();
+	diff = TX_RING_IDX(bp->tx_prod) - TX_RING_IDX(bp->tx_cons);
 	if (diff > MAX_TX_DESC_CNT)
 		diff = (diff & MAX_TX_DESC_CNT) - 1;
 	return (bp->tx_ring_size - diff);
@@ -1569,7 +1571,7 @@ bnx2_alloc_rx_skb(struct bnx2 *bp, u16 index)
 	struct rx_bd *rxbd = &bp->rx_desc_ring[RX_RING(index)][RX_IDX(index)];
 	unsigned long align;
 
-	skb = dev_alloc_skb(bp->rx_buf_size);
+	skb = netdev_alloc_skb(bp->dev, bp->rx_buf_size);
 	if (skb == NULL) {
 		return -ENOMEM;
 	}
@@ -1578,7 +1580,6 @@ bnx2_alloc_rx_skb(struct bnx2 *bp, u16 index)
 		skb_reserve(skb, 8 - align);
 	}
 
-	skb->dev = bp->dev;
 	mapping = pci_map_single(bp->pdev, skb->data, bp->rx_buf_use_size,
 		PCI_DMA_FROMDEVICE);
 
@@ -1686,15 +1687,20 @@ bnx2_tx_int(struct bnx2 *bp)
 	}
 
 	bp->tx_cons = sw_cons;
+	/* Need to make the tx_cons update visible to bnx2_start_xmit()
+	 * before checking for netif_queue_stopped().  Without the
+	 * memory barrier, there is a small possibility that bnx2_start_xmit()
+	 * will miss it and cause the queue to be stopped forever.
+	 */
+	smp_mb();
 
-	if (unlikely(netif_queue_stopped(bp->dev))) {
-		spin_lock(&bp->tx_lock);
+	if (unlikely(netif_queue_stopped(bp->dev)) &&
+		     (bnx2_tx_avail(bp) > bp->tx_wake_thresh)) {
+		netif_tx_lock(bp->dev);
 		if ((netif_queue_stopped(bp->dev)) &&
-		    (bnx2_tx_avail(bp) > MAX_SKB_FRAGS)) {
-
+		    (bnx2_tx_avail(bp) > bp->tx_wake_thresh))
 			netif_wake_queue(bp->dev);
-		}
-		spin_unlock(&bp->tx_lock);
+		netif_tx_unlock(bp->dev);
 	}
 }
 
@@ -1786,7 +1792,7 @@ bnx2_rx_int(struct bnx2 *bp, int budget)
 		if ((bp->dev->mtu > 1500) && (len <= RX_COPY_THRESH)) {
 			struct sk_buff *new_skb;
 
-			new_skb = dev_alloc_skb(len + 2);
+			new_skb = netdev_alloc_skb(bp->dev, len + 2);
 			if (new_skb == NULL)
 				goto reuse_rx;
 
@@ -1797,7 +1803,6 @@ bnx2_rx_int(struct bnx2 *bp, int budget)
 
 			skb_reserve(new_skb, 2);
 			skb_put(new_skb, len);
-			new_skb->dev = bp->dev;
 
 			bnx2_reuse_rx_skb(bp, skb,
 				sw_ring_cons, sw_ring_prod);
@@ -3503,6 +3508,8 @@ bnx2_init_tx_ring(struct bnx2 *bp)
 	struct tx_bd *txbd;
 	u32 val;
 
+	bp->tx_wake_thresh = bp->tx_ring_size / 2;
+
 	txbd = &bp->tx_desc_ring[MAX_TX_DESC_CNT];
 
 	txbd->tx_bd_haddr_hi = (u64) bp->tx_desc_mapping >> 32;
@@ -3952,7 +3959,7 @@ bnx2_run_loopback(struct bnx2 *bp, int loopback_mode)
 		return -EINVAL;
 
 	pkt_size = 1514;
-	skb = dev_alloc_skb(pkt_size);
+	skb = netdev_alloc_skb(bp->dev, pkt_size);
 	if (!skb)
 		return -ENOMEM;
 	packet = skb_put(skb, pkt_size);
@@ -4390,10 +4397,8 @@ bnx2_vlan_rx_kill_vid(struct net_device *dev, uint16_t vid)
 #endif
 
 /* Called with netif_tx_lock.
- * hard_start_xmit is pseudo-lockless - a lock is only required when
- * the tx queue is full. This way, we get the benefit of lockless
- * operations most of the time without the complexities to handle
- * netif_stop_queue/wake_queue race conditions.
+ * bnx2_tx_int() runs without netif_tx_lock unless it needs to call
+ * netif_wake_queue().
  */
 static int
 bnx2_start_xmit(struct sk_buff *skb, struct net_device *dev)
@@ -4512,12 +4517,9 @@ bnx2_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	dev->trans_start = jiffies;
 
 	if (unlikely(bnx2_tx_avail(bp) <= MAX_SKB_FRAGS)) {
-		spin_lock(&bp->tx_lock);
 		netif_stop_queue(dev);
-
-		if (bnx2_tx_avail(bp) > MAX_SKB_FRAGS)
+		if (bnx2_tx_avail(bp) > bp->tx_wake_thresh)
 			netif_wake_queue(dev);
-		spin_unlock(&bp->tx_lock);
 	}
 
 	return NETDEV_TX_OK;
@@ -5628,7 +5630,6 @@ bnx2_init_board(struct pci_dev *pdev, struct net_device *dev)
 	bp->pdev = pdev;
 
 	spin_lock_init(&bp->phy_lock);
-	spin_lock_init(&bp->tx_lock);
 	INIT_WORK(&bp->reset_task, bnx2_reset_task, bp);
 
 	dev->base_addr = dev->mem_start = pci_resource_start(pdev, 0);
@@ -5751,7 +5752,7 @@ bnx2_init_board(struct pci_dev *pdev, struct net_device *dev)
 	bp->mac_addr[5] = (u8) reg;
 
 	bp->tx_ring_size = MAX_TX_DESC_CNT;
-	bnx2_set_rx_ring_size(bp, 100);
+	bnx2_set_rx_ring_size(bp, 255);
 
 	bp->rx_csum = 1;
 

drivers/net/bnx2.h

@@ -3890,10 +3890,6 @@ struct bnx2 {
 	u32		tx_prod_bseq __attribute__((aligned(L1_CACHE_BYTES)));
 	u16		tx_prod;
 
-	struct tx_bd	*tx_desc_ring;
-	struct sw_bd	*tx_buf_ring;
-	int		tx_ring_size;
-
 	u16		tx_cons __attribute__((aligned(L1_CACHE_BYTES)));
 	u16		hw_tx_cons;
 
@@ -3916,9 +3912,11 @@ struct bnx2 {
 	struct sw_bd		*rx_buf_ring;
 	struct rx_bd		*rx_desc_ring[MAX_RX_RINGS];
 
-	/* Only used to synchronize netif_stop_queue/wake_queue when tx */
-	/* ring is full */
-	spinlock_t		tx_lock;
+	/* TX constants */
+	struct tx_bd	*tx_desc_ring;
+	struct sw_bd	*tx_buf_ring;
+	int		tx_ring_size;
+	u32		tx_wake_thresh;
 
 	/* End of fields used in the performance code paths. */
 

drivers/net/ppp_generic.c

@@ -192,7 +192,7 @@ struct cardmap {
 	void *ptr[CARDMAP_WIDTH];
 };
 static void *cardmap_get(struct cardmap *map, unsigned int nr);
-static void cardmap_set(struct cardmap **map, unsigned int nr, void *ptr);
+static int cardmap_set(struct cardmap **map, unsigned int nr, void *ptr);
 static unsigned int cardmap_find_first_free(struct cardmap *map);
 static void cardmap_destroy(struct cardmap **map);
 
@@ -1995,10 +1995,9 @@ ppp_register_channel(struct ppp_channel *chan)
 {
 	struct channel *pch;
 
-	pch = kmalloc(sizeof(struct channel), GFP_KERNEL);
+	pch = kzalloc(sizeof(struct channel), GFP_KERNEL);
 	if (pch == 0)
 		return -ENOMEM;
-	memset(pch, 0, sizeof(struct channel));
 	pch->ppp = NULL;
 	pch->chan = chan;
 	chan->ppp = pch;
@@ -2408,13 +2407,12 @@ ppp_create_interface(int unit, int *retp)
 	int ret = -ENOMEM;
 	int i;
 
-	ppp = kmalloc(sizeof(struct ppp), GFP_KERNEL);
+	ppp = kzalloc(sizeof(struct ppp), GFP_KERNEL);
 	if (!ppp)
 		goto out;
 	dev = alloc_netdev(0, "", ppp_setup);
 	if (!dev)
 		goto out1;
-	memset(ppp, 0, sizeof(struct ppp));
 
 	ppp->mru = PPP_MRU;
 	init_ppp_file(&ppp->file, INTERFACE);
@@ -2454,11 +2452,16 @@ ppp_create_interface(int unit, int *retp)
 	}
 
 	atomic_inc(&ppp_unit_count);
-	cardmap_set(&all_ppp_units, unit, ppp);
+	ret = cardmap_set(&all_ppp_units, unit, ppp);
+	if (ret != 0)
+		goto out3;
+
 	mutex_unlock(&all_ppp_mutex);
 	*retp = 0;
 	return ppp;
 
+out3:
+	atomic_dec(&ppp_unit_count);
 out2:
 	mutex_unlock(&all_ppp_mutex);
 	free_netdev(dev);
@@ -2695,7 +2698,7 @@ static void *cardmap_get(struct cardmap *map, unsigned int nr)
 	return NULL;
 }
 
-static void cardmap_set(struct cardmap **pmap, unsigned int nr, void *ptr)
+static int cardmap_set(struct cardmap **pmap, unsigned int nr, void *ptr)
 {
 	struct cardmap *p;
 	int i;
@@ -2704,8 +2707,9 @@ static void cardmap_set(struct cardmap **pmap, unsigned int nr, void *ptr)
 	if (p == NULL || (nr >> p->shift) >= CARDMAP_WIDTH) {
 		do {
 			/* need a new top level */
-			struct cardmap *np = kmalloc(sizeof(*np), GFP_KERNEL);
-			memset(np, 0, sizeof(*np));
+			struct cardmap *np = kzalloc(sizeof(*np), GFP_KERNEL);
+			if (!np)
+				goto enomem;
 			np->ptr[0] = p;
 			if (p != NULL) {
 				np->shift = p->shift + CARDMAP_ORDER;
@@ -2719,8 +2723,9 @@ static void cardmap_set(struct cardmap **pmap, unsigned int nr, void *ptr)
 	while (p->shift > 0) {
 		i = (nr >> p->shift) & CARDMAP_MASK;
 		if (p->ptr[i] == NULL) {
-			struct cardmap *np = kmalloc(sizeof(*np), GFP_KERNEL);
-			memset(np, 0, sizeof(*np));
+			struct cardmap *np = kzalloc(sizeof(*np), GFP_KERNEL);
+			if (!np)
+				goto enomem;
 			np->shift = p->shift - CARDMAP_ORDER;
 			np->parent = p;
 			p->ptr[i] = np;
@@ -2735,6 +2740,9 @@ static void cardmap_set(struct cardmap **pmap, unsigned int nr, void *ptr)
 		set_bit(i, &p->inuse);
 	else
 		clear_bit(i, &p->inuse);
+	return 0;
+ enomem:
+	return -ENOMEM;
 }
 
 static unsigned int cardmap_find_first_free(struct cardmap *map)

include/linux/if_vlan.h

@@ -155,6 +155,11 @@ static inline int __vlan_hwaccel_rx(struct sk_buff *skb,
 {
 	struct net_device_stats *stats;
 
+	if (skb_bond_should_drop(skb)) {
+		dev_kfree_skb_any(skb);
+		return NET_RX_DROP;
+	}
+
 	skb->dev = grp->vlan_devices[vlan_tag & VLAN_VID_MASK];
 	if (skb->dev == NULL) {
 		dev_kfree_skb_any(skb);

include/linux/netdevice.h

@@ -320,6 +320,9 @@ struct net_device
 #define NETIF_F_TSO_ECN		(SKB_GSO_TCP_ECN << NETIF_F_GSO_SHIFT)
 #define NETIF_F_TSO6		(SKB_GSO_TCPV6 << NETIF_F_GSO_SHIFT)
 
+	/* List of features with software fallbacks. */
+#define NETIF_F_GSO_SOFTWARE	(NETIF_F_TSO | NETIF_F_TSO_ECN | NETIF_F_TSO6)
+
 #define NETIF_F_GEN_CSUM	(NETIF_F_NO_CSUM | NETIF_F_HW_CSUM)
 #define NETIF_F_ALL_CSUM	(NETIF_F_IP_CSUM | NETIF_F_GEN_CSUM)
 
@@ -1012,6 +1015,30 @@ static inline int netif_needs_gso(struct net_device *dev, struct sk_buff *skb)
 		unlikely(skb->ip_summed != CHECKSUM_HW));
 }
 
+/* On bonding slaves other than the currently active slave, suppress
+ * duplicates except for 802.3ad ETH_P_SLOW and alb non-mcast/bcast.
+ */
+static inline int skb_bond_should_drop(struct sk_buff *skb)
+{
+	struct net_device *dev = skb->dev;
+	struct net_device *master = dev->master;
+
+	if (master &&
+	    (dev->priv_flags & IFF_SLAVE_INACTIVE)) {
+		if (master->priv_flags & IFF_MASTER_ALB) {
+			if (skb->pkt_type != PACKET_BROADCAST &&
+			    skb->pkt_type != PACKET_MULTICAST)
+				return 0;
+		}
+		if (master->priv_flags & IFF_MASTER_8023AD &&
+		    skb->protocol == __constant_htons(ETH_P_SLOW))
+			return 0;
+
+		return 1;
+	}
+	return 0;
+}
+
 #endif /* __KERNEL__ */
 
 #endif	/* _LINUX_DEV_H */

net/atm/proc.c

@@ -507,7 +507,7 @@ int __init atm_proc_init(void)
 	goto out;
 }
 
-void __exit atm_proc_exit(void)
+void atm_proc_exit(void)
 {
 	atm_proc_dirs_remove();
 }

net/bridge/br_if.c

@@ -386,12 +386,17 @@ void br_features_recompute(struct net_bridge *br)
 			checksum = 0;
 
 		if (feature & NETIF_F_GSO)
-			feature |= NETIF_F_TSO;
+			feature |= NETIF_F_GSO_SOFTWARE;
 		feature |= NETIF_F_GSO;
 
 		features &= feature;
 	}
 
+	if (!(checksum & NETIF_F_ALL_CSUM))
+		features &= ~NETIF_F_SG;
+	if (!(features & NETIF_F_SG))
+		features &= ~NETIF_F_GSO_MASK;
+
 	br->dev->features = features | checksum | NETIF_F_LLTX |
 			    NETIF_F_GSO_ROBUST;
 }