net: Add asynchronous callbacks for xfrm on layer 2.

This patch implements asynchronous crypto callbacks
and a backlog handler that can be used when IPsec
is done at layer 2 in the TX path. It also extends
the skb validate functions so that we can update
the driver transmit return codes based on async
crypto operation or to indicate that we queued the
packet in a backlog queue.

Joint work with: Aviv Heller <avivh@mellanox.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>

include/linux/netdevice.h

@@ -2793,7 +2793,9 @@ struct softnet_data {
 	struct Qdisc		*output_queue;
 	struct Qdisc		**output_queue_tailp;
 	struct sk_buff		*completion_queue;
-
+#ifdef CONFIG_XFRM_OFFLOAD
+	struct sk_buff_head	xfrm_backlog;
+#endif
 #ifdef CONFIG_RPS
 	/* input_queue_head should be written by cpu owning this struct,
 	 * and only read by other cpus. Worth using a cache line.
@@ -3325,7 +3327,7 @@ int dev_get_phys_port_id(struct net_device *dev,
 int dev_get_phys_port_name(struct net_device *dev,
 			   char *name, size_t len);
 int dev_change_proto_down(struct net_device *dev, bool proto_down);
-struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
+struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
 				    struct netdev_queue *txq, int *ret);
 

include/net/xfrm.h

@@ -1051,6 +1051,7 @@ struct xfrm_offload {
 #define	XFRM_GSO_SEGMENT	16
 #define	XFRM_GRO		32
 #define	XFRM_ESP_NO_TRAILER	64
+#define	XFRM_DEV_RESUME		128
 
 	__u32			status;
 #define CRYPTO_SUCCESS				1
@@ -1874,21 +1875,28 @@ static inline struct xfrm_state *xfrm_input_state(struct sk_buff *skb)
 {
 	return skb->sp->xvec[skb->sp->len - 1];
 }
+#endif
+
 static inline struct xfrm_offload *xfrm_offload(struct sk_buff *skb)
 {
+#ifdef CONFIG_XFRM
 	struct sec_path *sp = skb->sp;
 
 	if (!sp || !sp->olen || sp->len != sp->olen)
 		return NULL;
 
 	return &sp->ovec[sp->olen - 1];
-}
+#else
+	return NULL;
 #endif
+}
 
 void __net_init xfrm_dev_init(void);
 
 #ifdef CONFIG_XFRM_OFFLOAD
-struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features);
+void xfrm_dev_resume(struct sk_buff *skb);
+void xfrm_dev_backlog(struct softnet_data *sd);
+struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features, bool *again);
 int xfrm_dev_state_add(struct net *net, struct xfrm_state *x,
 		       struct xfrm_user_offload *xuo);
 bool xfrm_dev_offload_ok(struct sk_buff *skb, struct xfrm_state *x);
@@ -1929,7 +1937,15 @@ static inline void xfrm_dev_state_free(struct xfrm_state *x)
 	}
 }
 #else
-static inline struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features)
+static inline void xfrm_dev_resume(struct sk_buff *skb)
+{
+}
+
+static inline void xfrm_dev_backlog(struct softnet_data *sd)
+{
+}
+
+static inline struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features, bool *again)
 {
 	return skb;
 }

net/core/dev.c

@@ -3059,7 +3059,7 @@ int skb_csum_hwoffload_help(struct sk_buff *skb,
 }
 EXPORT_SYMBOL(skb_csum_hwoffload_help);
 
-static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device *dev)
+static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device *dev, bool *again)
 {
 	netdev_features_t features;
 
@@ -3099,7 +3099,7 @@ static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device
 		}
 	}
 
-	skb = validate_xmit_xfrm(skb, features);
+	skb = validate_xmit_xfrm(skb, features, again);
 
 	return skb;
 
@@ -3110,7 +3110,7 @@ static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device
 	return NULL;
 }
 
-struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev)
+struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again)
 {
 	struct sk_buff *next, *head = NULL, *tail;
 
@@ -3121,7 +3121,7 @@ struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *d
 		/* in case skb wont be segmented, point to itself */
 		skb->prev = skb;
 
-		skb = validate_xmit_skb(skb, dev);
+		skb = validate_xmit_skb(skb, dev, again);
 		if (!skb)
 			continue;
 
@@ -3448,6 +3448,7 @@ static int __dev_queue_xmit(struct sk_buff *skb, void *accel_priv)
 	struct netdev_queue *txq;
 	struct Qdisc *q;
 	int rc = -ENOMEM;
+	bool again = false;
 
 	skb_reset_mac_header(skb);
 
@@ -3509,7 +3510,7 @@ static int __dev_queue_xmit(struct sk_buff *skb, void *accel_priv)
 				     XMIT_RECURSION_LIMIT))
 				goto recursion_alert;
 
-			skb = validate_xmit_skb(skb, dev);
+			skb = validate_xmit_skb(skb, dev, &again);
 			if (!skb)
 				goto out;
 
@@ -4193,6 +4194,8 @@ static __latent_entropy void net_tx_action(struct softirq_action *h)
 				spin_unlock(root_lock);
 		}
 	}
+
+	xfrm_dev_backlog(sd);
 }
 
 #if IS_ENABLED(CONFIG_BRIDGE) && IS_ENABLED(CONFIG_ATM_LANE)
@@ -8874,6 +8877,9 @@ static int __init net_dev_init(void)
 
 		skb_queue_head_init(&sd->input_pkt_queue);
 		skb_queue_head_init(&sd->process_queue);
+#ifdef CONFIG_XFRM_OFFLOAD
+		skb_queue_head_init(&sd->xfrm_backlog);
+#endif
 		INIT_LIST_HEAD(&sd->poll_list);
 		sd->output_queue_tailp = &sd->output_queue;
 #ifdef CONFIG_RPS

net/ipv4/esp4.c

@@ -121,14 +121,32 @@ static void esp_ssg_unref(struct xfrm_state *x, void *tmp)
 static void esp_output_done(struct crypto_async_request *base, int err)
 {
 	struct sk_buff *skb = base->data;
+	struct xfrm_offload *xo = xfrm_offload(skb);
 	void *tmp;
-	struct dst_entry *dst = skb_dst(skb);
-	struct xfrm_state *x = dst->xfrm;
+	struct xfrm_state *x;
+
+	if (xo && (xo->flags & XFRM_DEV_RESUME))
+		x = skb->sp->xvec[skb->sp->len - 1];
+	else
+		x = skb_dst(skb)->xfrm;
 
 	tmp = ESP_SKB_CB(skb)->tmp;
 	esp_ssg_unref(x, tmp);
 	kfree(tmp);
-	xfrm_output_resume(skb, err);
+
+	if (xo && (xo->flags & XFRM_DEV_RESUME)) {
+		if (err) {
+			XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
+			kfree_skb(skb);
+			return;
+		}
+
+		skb_push(skb, skb->data - skb_mac_header(skb));
+		secpath_reset(skb);
+		xfrm_dev_resume(skb);
+	} else {
+		xfrm_output_resume(skb, err);
+	}
 }
 
 /* Move ESP header back into place. */

net/ipv6/esp6.c

@@ -141,14 +141,32 @@ static void esp_ssg_unref(struct xfrm_state *x, void *tmp)
 static void esp_output_done(struct crypto_async_request *base, int err)
 {
 	struct sk_buff *skb = base->data;
+	struct xfrm_offload *xo = xfrm_offload(skb);
 	void *tmp;
-	struct dst_entry *dst = skb_dst(skb);
-	struct xfrm_state *x = dst->xfrm;
+	struct xfrm_state *x;
+
+	if (xo && (xo->flags & XFRM_DEV_RESUME))
+		x = skb->sp->xvec[skb->sp->len - 1];
+	else
+		x = skb_dst(skb)->xfrm;
 
 	tmp = ESP_SKB_CB(skb)->tmp;
 	esp_ssg_unref(x, tmp);
 	kfree(tmp);
-	xfrm_output_resume(skb, err);
+
+	if (xo && (xo->flags & XFRM_DEV_RESUME)) {
+		if (err) {
+			XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
+			kfree_skb(skb);
+			return;
+		}
+
+		skb_push(skb, skb->data - skb_mac_header(skb));
+		secpath_reset(skb);
+		xfrm_dev_resume(skb);
+	} else {
+		xfrm_output_resume(skb, err);
+	}
 }
 
 /* Move ESP header back into place. */

net/packet/af_packet.c

@@ -247,12 +247,13 @@ static int packet_direct_xmit(struct sk_buff *skb)
 	struct sk_buff *orig_skb = skb;
 	struct netdev_queue *txq;
 	int ret = NETDEV_TX_BUSY;
+	bool again = false;
 
 	if (unlikely(!netif_running(dev) ||
 		     !netif_carrier_ok(dev)))
 		goto drop;
 
-	skb = validate_xmit_skb_list(skb, dev);
+	skb = validate_xmit_skb_list(skb, dev, &again);
 	if (skb != orig_skb)
 		goto drop;
 

net/sched/sch_generic.c

@@ -32,6 +32,7 @@
 #include <net/pkt_sched.h>
 #include <net/dst.h>
 #include <trace/events/qdisc.h>
+#include <net/xfrm.h>
 
 /* Qdisc to use by default */
 const struct Qdisc_ops *default_qdisc_ops = &pfifo_fast_ops;
@@ -230,6 +231,8 @@ static struct sk_buff *dequeue_skb(struct Qdisc *q, bool *validate,
 
 		/* skb in gso_skb were already validated */
 		*validate = false;
+		if (xfrm_offload(skb))
+			*validate = true;
 		/* check the reason of requeuing without tx lock first */
 		txq = skb_get_tx_queue(txq->dev, skb);
 		if (!netif_xmit_frozen_or_stopped(txq)) {
@@ -285,6 +288,7 @@ bool sch_direct_xmit(struct sk_buff *skb, struct Qdisc *q,
 		     spinlock_t *root_lock, bool validate)
 {
 	int ret = NETDEV_TX_BUSY;
+	bool again = false;
 
 	/* And release qdisc */
 	if (root_lock)
@@ -292,7 +296,17 @@ bool sch_direct_xmit(struct sk_buff *skb, struct Qdisc *q,
 
 	/* Note that we validate skb (GSO, checksum, ...) outside of locks */
 	if (validate)
-		skb = validate_xmit_skb_list(skb, dev);
+		skb = validate_xmit_skb_list(skb, dev, &again);
+
+#ifdef CONFIG_XFRM_OFFLOAD
+	if (unlikely(again)) {
+		if (root_lock)
+			spin_lock(root_lock);
+
+		dev_requeue_skb(skb, q);
+		return false;
+	}
+#endif
 
 	if (likely(skb)) {
 		HARD_TX_LOCK(dev, txq, smp_processor_id());

net/xfrm/xfrm_device.c

@@ -23,12 +23,13 @@
 #include <linux/notifier.h>
 
 #ifdef CONFIG_XFRM_OFFLOAD
-struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features)
+struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features, bool *again)
 {
 	int err;
-	__u32 seq;
+	unsigned long flags;
 	struct xfrm_state *x;
 	struct sk_buff *skb2;
+	struct softnet_data *sd;
 	netdev_features_t esp_features = features;
 	struct xfrm_offload *xo = xfrm_offload(skb);
 
@@ -42,6 +43,16 @@ struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t featur
 	if (xo->flags & XFRM_GRO || x->xso.flags & XFRM_OFFLOAD_INBOUND)
 		return skb;
 
+	local_irq_save(flags);
+	sd = this_cpu_ptr(&softnet_data);
+	err = !skb_queue_empty(&sd->xfrm_backlog);
+	local_irq_restore(flags);
+
+	if (err) {
+		*again = true;
+		return skb;
+	}
+
 	if (skb_is_gso(skb)) {
 		struct net_device *dev = skb->dev;
 
@@ -54,23 +65,26 @@ struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t featur
 
 			segs = skb_gso_segment(skb, esp_features);
 			if (IS_ERR(segs)) {
-				XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
 				kfree_skb(skb);
+				atomic_long_inc(&dev->tx_dropped);
 				return NULL;
 			} else {
 				consume_skb(skb);
 				skb = segs;
 			}
-		} else {
-			return skb;
 		}
 	}
 
 	if (!skb->next) {
 		x->outer_mode->xmit(x, skb);
 
+		xo->flags |= XFRM_DEV_RESUME;
+
 		err = x->type_offload->xmit(x, skb, esp_features);
 		if (err) {
+			if (err == -EINPROGRESS)
+				return NULL;
+
 			XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
 			kfree_skb(skb);
 			return NULL;
@@ -82,36 +96,37 @@ struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t featur
 	}
 
 	skb2 = skb;
-	seq = xo->seq.low;
 
 	do {
 		struct sk_buff *nskb = skb2->next;
+		skb2->next = NULL;
 
 		xo = xfrm_offload(skb2);
-		xo->flags |= XFRM_GSO_SEGMENT;
-		xo->seq.low = seq;
-		xo->seq.hi = xfrm_replay_seqhi(x, seq);
-
-		if(!(features & NETIF_F_HW_ESP))
-			xo->flags |= CRYPTO_FALLBACK;
+		xo->flags |= XFRM_DEV_RESUME;
 
 		x->outer_mode->xmit(x, skb2);
 
 		err = x->type_offload->xmit(x, skb2, esp_features);
-		if (err) {
+		if (!err) {
+			skb2->next = nskb;
+		} else if (err != -EINPROGRESS) {
 			XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
 			skb2->next = nskb;
 			kfree_skb_list(skb2);
 			return NULL;
-		}
+		} else {
+			if (skb == skb2)
+				skb = nskb;
+
+			if (!skb)
+				return NULL;
 
-		if (!skb_is_gso(skb2))
-			seq++;
-		else
-			seq += skb_shinfo(skb2)->gso_segs;
+			goto skip_push;
+		}
 
 		skb_push(skb2, skb2->data - skb_mac_header(skb2));
 
+skip_push:
 		skb2 = nskb;
 	} while (skb2);
 
@@ -207,6 +222,55 @@ bool xfrm_dev_offload_ok(struct sk_buff *skb, struct xfrm_state *x)
 	return true;
 }
 EXPORT_SYMBOL_GPL(xfrm_dev_offload_ok);
+
+void xfrm_dev_resume(struct sk_buff *skb)
+{
+	struct net_device *dev = skb->dev;
+	int ret = NETDEV_TX_BUSY;
+	struct netdev_queue *txq;
+	struct softnet_data *sd;
+	unsigned long flags;
+
+	rcu_read_lock();
+	txq = netdev_pick_tx(dev, skb, NULL);
+
+	HARD_TX_LOCK(dev, txq, smp_processor_id());
+	if (!netif_xmit_frozen_or_stopped(txq))
+		skb = dev_hard_start_xmit(skb, dev, txq, &ret);
+	HARD_TX_UNLOCK(dev, txq);
+
+	if (!dev_xmit_complete(ret)) {
+		local_irq_save(flags);
+		sd = this_cpu_ptr(&softnet_data);
+		skb_queue_tail(&sd->xfrm_backlog, skb);
+		raise_softirq_irqoff(NET_TX_SOFTIRQ);
+		local_irq_restore(flags);
+	}
+	rcu_read_unlock();
+}
+EXPORT_SYMBOL_GPL(xfrm_dev_resume);
+
+void xfrm_dev_backlog(struct softnet_data *sd)
+{
+	struct sk_buff_head *xfrm_backlog = &sd->xfrm_backlog;
+	struct sk_buff_head list;
+	struct sk_buff *skb;
+
+	if (skb_queue_empty(xfrm_backlog))
+		return;
+
+	__skb_queue_head_init(&list);
+
+	spin_lock(&xfrm_backlog->lock);
+	skb_queue_splice_init(xfrm_backlog, &list);
+	spin_unlock(&xfrm_backlog->lock);
+
+	while (!skb_queue_empty(&list)) {
+		skb = __skb_dequeue(&list);
+		xfrm_dev_resume(skb);
+	}
+
+}
 #endif
 
 static int xfrm_dev_register(struct net_device *dev)