ifb: add multiqueue operation

Add multiqueue capabilities to ifb netdevice.

This removes last bottleneck for ingress when mq qdisc can be used
to shard load from multiple RX queues on physical device.

Tested:

# netem based setup, installed at receiver side
ETH=eth0
IFB=ifb10
EST="est 1sec 4sec" # Optional rate estimator
RTT_HALF=2ms
#REORDER=20us
#LOSS="loss 1"
TXQ=8

ip link add ifb10 numtxqueues $TXQ type ifb
ip link set dev $IFB up

tc qdisc add dev $ETH ingress 2>/dev/null

tc filter add dev $ETH parent ffff: \
   protocol ip u32 match u32 0 0 flowid 1:1 \
	action mirred egress redirect dev $IFB

tc qdisc del dev $IFB root 2>/dev/null

tc qdisc add dev $IFB root handle 1: mq
for i in `seq 1 $TXQ`
do
 slot=$( printf %x $(( i )) )
 tc qd add dev $IFB parent 1:$slot $EST netem \
	limit 100000 delay $RTT_HALF $REORDER $LOSS
done

lpaa24:~# tc -s -d qd sh dev ifb10
qdisc mq 1: root
 Sent 316544766 bytes 5265927 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 98880b 1648p requeues 0
qdisc netem 8002: parent 1:1 limit 100000 delay 2.0ms
 Sent 39601416 bytes 658721 pkt (dropped 0, overlimits 0 requeues 0)
 rate 38235Kbit 79657pps backlog 12240b 204p requeues 0
qdisc netem 8003: parent 1:2 limit 100000 delay 2.0ms
 Sent 39472866 bytes 657227 pkt (dropped 0, overlimits 0 requeues 0)
 rate 38234Kbit 79655pps backlog 10620b 176p requeues 0
qdisc netem 8004: parent 1:3 limit 100000 delay 2.0ms
 Sent 39703417 bytes 659699 pkt (dropped 0, overlimits 0 requeues 0)
 rate 38320Kbit 79831pps backlog 12780b 213p requeues 0
qdisc netem 8005: parent 1:4 limit 100000 delay 2.0ms
 Sent 39565149 bytes 658011 pkt (dropped 0, overlimits 0 requeues 0)
 rate 38174Kbit 79530pps backlog 11880b 198p requeues 0
qdisc netem 8006: parent 1:5 limit 100000 delay 2.0ms
 Sent 39506078 bytes 657354 pkt (dropped 0, overlimits 0 requeues 0)
 rate 38195Kbit 79571pps backlog 12480b 208p requeues 0
qdisc netem 8007: parent 1:6 limit 100000 delay 2.0ms
 Sent 39675994 bytes 658849 pkt (dropped 0, overlimits 0 requeues 0)
 rate 38323Kbit 79838pps backlog 12600b 210p requeues 0
qdisc netem 8008: parent 1:7 limit 100000 delay 2.0ms
 Sent 39532042 bytes 658367 pkt (dropped 0, overlimits 0 requeues 0)
 rate 38177Kbit 79536pps backlog 13140b 219p requeues 0
qdisc netem 8009: parent 1:8 limit 100000 delay 2.0ms
 Sent 39488164 bytes 657705 pkt (dropped 0, overlimits 0 requeues 0)
 rate 38192Kbit 79568pps backlog 13Kb 222p requeues 0

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Alexei Starovoitov <ast@plumgrid.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ifb.c

@@ -38,69 +38,68 @@
 #include <net/net_namespace.h>
 
 #define TX_Q_LIMIT    32
-struct ifb_private {
+struct ifb_q_private {
+	struct net_device	*dev;
 	struct tasklet_struct   ifb_tasklet;
-	int     tasklet_pending;
-
-	struct u64_stats_sync	rsync;
+	int			tasklet_pending;
+	int			txqnum;
 	struct sk_buff_head     rq;
-	u64 rx_packets;
-	u64 rx_bytes;
+	u64			rx_packets;
+	u64			rx_bytes;
+	struct u64_stats_sync	rsync;
 
 	struct u64_stats_sync	tsync;
+	u64			tx_packets;
+	u64			tx_bytes;
 	struct sk_buff_head     tq;
-	u64 tx_packets;
-	u64 tx_bytes;
-};
+} ____cacheline_aligned_in_smp;
 
-static int numifbs = 2;
+struct ifb_dev_private {
+	struct ifb_q_private *tx_private;
+};
 
-static void ri_tasklet(unsigned long dev);
 static netdev_tx_t ifb_xmit(struct sk_buff *skb, struct net_device *dev);
 static int ifb_open(struct net_device *dev);
 static int ifb_close(struct net_device *dev);
 
-static void ri_tasklet(unsigned long dev)
+static void ifb_ri_tasklet(unsigned long _txp)
 {
-	struct net_device *_dev = (struct net_device *)dev;
-	struct ifb_private *dp = netdev_priv(_dev);
+	struct ifb_q_private *txp = (struct ifb_q_private *)_txp;
 	struct netdev_queue *txq;
 	struct sk_buff *skb;
 
-	txq = netdev_get_tx_queue(_dev, 0);
-	if ((skb = skb_peek(&dp->tq)) == NULL) {
-		if (__netif_tx_trylock(txq)) {
-			skb_queue_splice_tail_init(&dp->rq, &dp->tq);
-			__netif_tx_unlock(txq);
-		} else {
-			/* reschedule */
+	txq = netdev_get_tx_queue(txp->dev, txp->txqnum);
+	skb = skb_peek(&txp->tq);
+	if (!skb) {
+		if (!__netif_tx_trylock(txq))
 			goto resched;
-		}
+		skb_queue_splice_tail_init(&txp->rq, &txp->tq);
+		__netif_tx_unlock(txq);
 	}
 
-	while ((skb = __skb_dequeue(&dp->tq)) != NULL) {
+	while ((skb = __skb_dequeue(&txp->tq)) != NULL) {
 		u32 from = G_TC_FROM(skb->tc_verd);
 
 		skb->tc_verd = 0;
 		skb->tc_verd = SET_TC_NCLS(skb->tc_verd);
 
-		u64_stats_update_begin(&dp->tsync);
-		dp->tx_packets++;
-		dp->tx_bytes += skb->len;
-		u64_stats_update_end(&dp->tsync);
+		u64_stats_update_begin(&txp->tsync);
+		txp->tx_packets++;
+		txp->tx_bytes += skb->len;
+		u64_stats_update_end(&txp->tsync);
 
 		rcu_read_lock();
-		skb->dev = dev_get_by_index_rcu(dev_net(_dev), skb->skb_iif);
+		skb->dev = dev_get_by_index_rcu(dev_net(txp->dev), skb->skb_iif);
 		if (!skb->dev) {
 			rcu_read_unlock();
 			dev_kfree_skb(skb);
-			_dev->stats.tx_dropped++;
-			if (skb_queue_len(&dp->tq) != 0)
+			txp->dev->stats.tx_dropped++;
+			if (skb_queue_len(&txp->tq) != 0)
 				goto resched;
 			break;
 		}
 		rcu_read_unlock();
-		skb->skb_iif = _dev->ifindex;
+		skb->skb_iif = txp->dev->ifindex;
 
 		if (from & AT_EGRESS) {
 			dev_queue_xmit(skb);
@@ -112,10 +111,11 @@ static void ri_tasklet(unsigned long dev)
 	}
 
 	if (__netif_tx_trylock(txq)) {
-		if ((skb = skb_peek(&dp->rq)) == NULL) {
-			dp->tasklet_pending = 0;
-			if (netif_queue_stopped(_dev))
-				netif_wake_queue(_dev);
+		skb = skb_peek(&txp->rq);
+		if (!skb) {
+			txp->tasklet_pending = 0;
+			if (netif_tx_queue_stopped(txq))
+				netif_tx_wake_queue(txq);
 		} else {
 			__netif_tx_unlock(txq);
 			goto resched;
@@ -123,8 +123,8 @@ static void ri_tasklet(unsigned long dev)
 		__netif_tx_unlock(txq);
 	} else {
 resched:
-		dp->tasklet_pending = 1;
-		tasklet_schedule(&dp->ifb_tasklet);
+		txp->tasklet_pending = 1;
+		tasklet_schedule(&txp->ifb_tasklet);
 	}
 
 }
@@ -132,29 +132,58 @@ resched:
 static struct rtnl_link_stats64 *ifb_stats64(struct net_device *dev,
 					     struct rtnl_link_stats64 *stats)
 {
-	struct ifb_private *dp = netdev_priv(dev);
+	struct ifb_dev_private *dp = netdev_priv(dev);
+	struct ifb_q_private *txp = dp->tx_private;
 	unsigned int start;
-
-	do {
-		start = u64_stats_fetch_begin_irq(&dp->rsync);
-		stats->rx_packets = dp->rx_packets;
-		stats->rx_bytes = dp->rx_bytes;
-	} while (u64_stats_fetch_retry_irq(&dp->rsync, start));
-
-	do {
-		start = u64_stats_fetch_begin_irq(&dp->tsync);
-
-		stats->tx_packets = dp->tx_packets;
-		stats->tx_bytes = dp->tx_bytes;
-
-	} while (u64_stats_fetch_retry_irq(&dp->tsync, start));
-
+	u64 packets, bytes;
+	int i;
+
+	for (i = 0; i < dev->num_tx_queues; i++,txp++) {
+		do {
+			start = u64_stats_fetch_begin_irq(&txp->rsync);
+			packets = txp->rx_packets;
+			bytes = txp->rx_bytes;
+		} while (u64_stats_fetch_retry_irq(&txp->rsync, start));
+		stats->rx_packets += packets;
+		stats->rx_bytes += bytes;
+
+		do {
+			start = u64_stats_fetch_begin_irq(&txp->tsync);
+			packets = txp->tx_packets;
+			bytes = txp->tx_bytes;
+		} while (u64_stats_fetch_retry_irq(&txp->tsync, start));
+		stats->tx_packets += packets;
+		stats->tx_bytes += bytes;
+	}
 	stats->rx_dropped = dev->stats.rx_dropped;
 	stats->tx_dropped = dev->stats.tx_dropped;
 
 	return stats;
 }
 
+static int ifb_dev_init(struct net_device *dev)
+{
+	struct ifb_dev_private *dp = netdev_priv(dev);
+	struct ifb_q_private *txp;
+	int i;
+
+	txp = kcalloc(dev->num_tx_queues, sizeof(*txp), GFP_KERNEL);
+	if (!txp)
+		return -ENOMEM;
+	dp->tx_private = txp;
+	for (i = 0; i < dev->num_tx_queues; i++,txp++) {
+		txp->txqnum = i;
+		txp->dev = dev;
+		__skb_queue_head_init(&txp->rq);
+		__skb_queue_head_init(&txp->tq);
+		u64_stats_init(&txp->rsync);
+		u64_stats_init(&txp->tsync);
+		tasklet_init(&txp->ifb_tasklet, ifb_ri_tasklet,
+			     (unsigned long)txp);
+		netif_tx_start_queue(netdev_get_tx_queue(dev, i));
+	}
+	return 0;
+}
 
 static const struct net_device_ops ifb_netdev_ops = {
 	.ndo_open	= ifb_open,
@@ -162,6 +191,7 @@ static const struct net_device_ops ifb_netdev_ops = {
 	.ndo_get_stats64 = ifb_stats64,
 	.ndo_start_xmit	= ifb_xmit,
 	.ndo_validate_addr = eth_validate_addr,
+	.ndo_init	= ifb_dev_init,
 };
 
 #define IFB_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG  | NETIF_F_FRAGLIST	| \
@@ -169,10 +199,24 @@ static const struct net_device_ops ifb_netdev_ops = {
 		      NETIF_F_HIGHDMA | NETIF_F_HW_VLAN_CTAG_TX		| \
 		      NETIF_F_HW_VLAN_STAG_TX)
 
+static void ifb_dev_free(struct net_device *dev)
+{
+	struct ifb_dev_private *dp = netdev_priv(dev);
+	struct ifb_q_private *txp = dp->tx_private;
+	int i;
+
+	for (i = 0; i < dev->num_tx_queues; i++,txp++) {
+		tasklet_kill(&txp->ifb_tasklet);
+		__skb_queue_purge(&txp->rq);
+		__skb_queue_purge(&txp->tq);
+	}
+	kfree(dp->tx_private);
+	free_netdev(dev);
+}
+
 static void ifb_setup(struct net_device *dev)
 {
 	/* Initialize the device structure. */
-	dev->destructor = free_netdev;
 	dev->netdev_ops = &ifb_netdev_ops;
 
 	/* Fill in device structure with ethernet-generic values. */
@@ -188,17 +232,19 @@ static void ifb_setup(struct net_device *dev)
 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
 	netif_keep_dst(dev);
 	eth_hw_addr_random(dev);
+	dev->destructor = ifb_dev_free;
 }
 
 static netdev_tx_t ifb_xmit(struct sk_buff *skb, struct net_device *dev)
 {
-	struct ifb_private *dp = netdev_priv(dev);
+	struct ifb_dev_private *dp = netdev_priv(dev);
 	u32 from = G_TC_FROM(skb->tc_verd);
+	struct ifb_q_private *txp = dp->tx_private + skb_get_queue_mapping(skb);
 
-	u64_stats_update_begin(&dp->rsync);
-	dp->rx_packets++;
-	dp->rx_bytes += skb->len;
-	u64_stats_update_end(&dp->rsync);
+	u64_stats_update_begin(&txp->rsync);
+	txp->rx_packets++;
+	txp->rx_bytes += skb->len;
+	u64_stats_update_end(&txp->rsync);
 
 	if (!(from & (AT_INGRESS|AT_EGRESS)) || !skb->skb_iif) {
 		dev_kfree_skb(skb);
@@ -206,14 +252,13 @@ static netdev_tx_t ifb_xmit(struct sk_buff *skb, struct net_device *dev)
 		return NETDEV_TX_OK;
 	}
 
-	if (skb_queue_len(&dp->rq) >= dev->tx_queue_len) {
-		netif_stop_queue(dev);
-	}
+	if (skb_queue_len(&txp->rq) >= dev->tx_queue_len)
+		netif_tx_stop_queue(netdev_get_tx_queue(dev, txp->txqnum));
 
-	__skb_queue_tail(&dp->rq, skb);
-	if (!dp->tasklet_pending) {
-		dp->tasklet_pending = 1;
-		tasklet_schedule(&dp->ifb_tasklet);
+	__skb_queue_tail(&txp->rq, skb);
+	if (!txp->tasklet_pending) {
+		txp->tasklet_pending = 1;
+		tasklet_schedule(&txp->ifb_tasklet);
 	}
 
 	return NETDEV_TX_OK;
@@ -221,24 +266,13 @@ static netdev_tx_t ifb_xmit(struct sk_buff *skb, struct net_device *dev)
 
 static int ifb_close(struct net_device *dev)
 {
-	struct ifb_private *dp = netdev_priv(dev);
-
-	tasklet_kill(&dp->ifb_tasklet);
-	netif_stop_queue(dev);
-	__skb_queue_purge(&dp->rq);
-	__skb_queue_purge(&dp->tq);
+	netif_tx_stop_all_queues(dev);
 	return 0;
 }
 
 static int ifb_open(struct net_device *dev)
 {
-	struct ifb_private *dp = netdev_priv(dev);
-
-	tasklet_init(&dp->ifb_tasklet, ri_tasklet, (unsigned long)dev);
-	__skb_queue_head_init(&dp->rq);
-	__skb_queue_head_init(&dp->tq);
-	netif_start_queue(dev);
-
+	netif_tx_start_all_queues(dev);
 	return 0;
 }
 
@@ -255,31 +289,30 @@ static int ifb_validate(struct nlattr *tb[], struct nlattr *data[])
 
 static struct rtnl_link_ops ifb_link_ops __read_mostly = {
 	.kind		= "ifb",
-	.priv_size	= sizeof(struct ifb_private),
+	.priv_size	= sizeof(struct ifb_dev_private),
 	.setup		= ifb_setup,
 	.validate	= ifb_validate,
 };
 
-/* Number of ifb devices to be set up by this module. */
+/* Number of ifb devices to be set up by this module.
+ * Note that these legacy devices have one queue.
+ * Prefer something like : ip link add ifb10 numtxqueues 8 type ifb
+ */
+static int numifbs = 2;
 module_param(numifbs, int, 0);
 MODULE_PARM_DESC(numifbs, "Number of ifb devices");
 
 static int __init ifb_init_one(int index)
 {
 	struct net_device *dev_ifb;
-	struct ifb_private *dp;
 	int err;
 
-	dev_ifb = alloc_netdev(sizeof(struct ifb_private), "ifb%d",
+	dev_ifb = alloc_netdev(sizeof(struct ifb_dev_private), "ifb%d",
 			       NET_NAME_UNKNOWN, ifb_setup);
 
 	if (!dev_ifb)
 		return -ENOMEM;
 
-	dp = netdev_priv(dev_ifb);
-	u64_stats_init(&dp->rsync);
-	u64_stats_init(&dp->tsync);
-
 	dev_ifb->rtnl_link_ops = &ifb_link_ops;
 	err = register_netdevice(dev_ifb);
 	if (err < 0)