xfrm: Separate ESP handling from segmentation for GRO packets.

We change the ESP GSO handlers to only segment the packets.
The ESP handling and encryption is defered to validate_xmit_xfrm()
where this is done for non GRO packets too. This makes the code
more robust and prepares for asynchronous crypto handling.
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>

include/net/xfrm.h

@@ -1888,7 +1888,7 @@ static inline struct xfrm_offload *xfrm_offload(struct sk_buff *skb)
 void __net_init xfrm_dev_init(void);
 
 #ifdef CONFIG_XFRM_OFFLOAD
-int 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);
 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);
@@ -1929,9 +1929,9 @@ static inline void xfrm_dev_state_free(struct xfrm_state *x)
 	}
 }
 #else
-static inline int validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features)
+static inline struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features)
 {
-	return 0;
+	return skb;
 }
 
 static inline int xfrm_dev_state_add(struct net *net, struct xfrm_state *x, struct xfrm_user_offload *xuo)

net/core/dev.c

@@ -3083,9 +3083,6 @@ static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device
 		    __skb_linearize(skb))
 			goto out_kfree_skb;
 
-		if (validate_xmit_xfrm(skb, features))
-			goto out_kfree_skb;
-
 		/* If packet is not checksummed and device does not
 		 * support checksumming for this protocol, complete
 		 * checksumming here.
@@ -3102,6 +3099,8 @@ static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device
 		}
 	}
 
+	skb = validate_xmit_xfrm(skb, features);
+
 	return skb;
 
 out_kfree_skb:

net/ipv4/esp4_offload.c

@@ -108,75 +108,36 @@ static void esp4_gso_encap(struct xfrm_state *x, struct sk_buff *skb)
 static struct sk_buff *esp4_gso_segment(struct sk_buff *skb,
 				        netdev_features_t features)
 {
-	__u32 seq;
-	int err = 0;
-	struct sk_buff *skb2;
 	struct xfrm_state *x;
 	struct ip_esp_hdr *esph;
 	struct crypto_aead *aead;
-	struct sk_buff *segs = ERR_PTR(-EINVAL);
 	netdev_features_t esp_features = features;
 	struct xfrm_offload *xo = xfrm_offload(skb);
 
 	if (!xo)
-		goto out;
-
-	seq = xo->seq.low;
+		return ERR_PTR(-EINVAL);
 
 	x = skb->sp->xvec[skb->sp->len - 1];
 	aead = x->data;
 	esph = ip_esp_hdr(skb);
 
 	if (esph->spi != x->id.spi)
-		goto out;
+		return ERR_PTR(-EINVAL);
 
 	if (!pskb_may_pull(skb, sizeof(*esph) + crypto_aead_ivsize(aead)))
-		goto out;
+		return ERR_PTR(-EINVAL);
 
 	__skb_pull(skb, sizeof(*esph) + crypto_aead_ivsize(aead));
 
 	skb->encap_hdr_csum = 1;
 
-	if (!(features & NETIF_F_HW_ESP))
+	if (!(features & NETIF_F_HW_ESP) || !x->xso.offload_handle ||
+	    (x->xso.dev != skb->dev))
 		esp_features = features & ~(NETIF_F_SG | NETIF_F_CSUM_MASK);
 
-	segs = x->outer_mode->gso_segment(x, skb, esp_features);
-	if (IS_ERR_OR_NULL(segs))
-		goto out;
-
-	__skb_pull(skb, skb->data - skb_mac_header(skb));
-
-	skb2 = segs;
-	do {
-		struct sk_buff *nskb = skb2->next;
-
-		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;
-
-		x->outer_mode->xmit(x, skb2);
-
-		err = x->type_offload->xmit(x, skb2, esp_features);
-		if (err) {
-			kfree_skb_list(segs);
-			return ERR_PTR(err);
-		}
-
-		if (!skb_is_gso(skb2))
-			seq++;
-		else
-			seq += skb_shinfo(skb2)->gso_segs;
-
-		skb_push(skb2, skb2->mac_len);
-		skb2 = nskb;
-	} while (skb2);
 
-out:
-	return segs;
+	return x->outer_mode->gso_segment(x, skb, esp_features);
 }
 
 static int esp_input_tail(struct xfrm_state *x, struct sk_buff *skb)
@@ -203,6 +164,7 @@ static int esp_xmit(struct xfrm_state *x, struct sk_buff *skb,  netdev_features_
 	struct crypto_aead *aead;
 	struct esp_info esp;
 	bool hw_offload = true;
+	__u32 seq;
 
 	esp.inplace = true;
 
@@ -241,23 +203,30 @@ static int esp_xmit(struct xfrm_state *x, struct sk_buff *skb,  netdev_features_
 			return esp.nfrags;
 	}
 
+	seq = xo->seq.low;
+
 	esph = esp.esph;
 	esph->spi = x->id.spi;
 
 	skb_push(skb, -skb_network_offset(skb));
 
 	if (xo->flags & XFRM_GSO_SEGMENT) {
-		esph->seq_no = htonl(xo->seq.low);
-	} else {
+		esph->seq_no = htonl(seq);
+
+		if (!skb_is_gso(skb))
+			xo->seq.low++;
+		else
+			xo->seq.low += skb_shinfo(skb)->gso_segs;
+	}
+
+	esp.seqno = cpu_to_be64(seq + ((u64)xo->seq.hi << 32));
+
 	ip_hdr(skb)->tot_len = htons(skb->len);
 	ip_send_check(ip_hdr(skb));
-	}
 
 	if (hw_offload)
 		return 0;
 
-	esp.seqno = cpu_to_be64(xo->seq.low + ((u64)xo->seq.hi << 32));
-
 	err = esp_output_tail(x, skb, &esp);
 	if (err)
 		return err;

net/ipv4/xfrm4_mode_tunnel.c

@@ -105,18 +105,15 @@ static struct sk_buff *xfrm4_mode_tunnel_gso_segment(struct xfrm_state *x,
 {
 	__skb_push(skb, skb->mac_len);
 	return skb_mac_gso_segment(skb, features);
-
 }
 
 static void xfrm4_mode_tunnel_xmit(struct xfrm_state *x, struct sk_buff *skb)
 {
 	struct xfrm_offload *xo = xfrm_offload(skb);
 
-	if (xo->flags & XFRM_GSO_SEGMENT) {
-		skb->network_header = skb->network_header - x->props.header_len;
+	if (xo->flags & XFRM_GSO_SEGMENT)
 		skb->transport_header = skb->network_header +
 					sizeof(struct iphdr);
-	}
 
 	skb_reset_mac_len(skb);
 	pskb_pull(skb, skb->mac_len + x->props.header_len);

net/ipv6/esp6_offload.c

@@ -135,75 +135,36 @@ static void esp6_gso_encap(struct xfrm_state *x, struct sk_buff *skb)
 static struct sk_buff *esp6_gso_segment(struct sk_buff *skb,
 				        netdev_features_t features)
 {
-	__u32 seq;
-	int err = 0;
-	struct sk_buff *skb2;
 	struct xfrm_state *x;
 	struct ip_esp_hdr *esph;
 	struct crypto_aead *aead;
-	struct sk_buff *segs = ERR_PTR(-EINVAL);
 	netdev_features_t esp_features = features;
 	struct xfrm_offload *xo = xfrm_offload(skb);
 
 	if (!xo)
-		goto out;
-
-	seq = xo->seq.low;
+		return ERR_PTR(-EINVAL);
 
 	x = skb->sp->xvec[skb->sp->len - 1];
 	aead = x->data;
 	esph = ip_esp_hdr(skb);
 
 	if (esph->spi != x->id.spi)
-		goto out;
+		return ERR_PTR(-EINVAL);
 
 	if (!pskb_may_pull(skb, sizeof(*esph) + crypto_aead_ivsize(aead)))
-		goto out;
+		return ERR_PTR(-EINVAL);
 
 	__skb_pull(skb, sizeof(*esph) + crypto_aead_ivsize(aead));
 
 	skb->encap_hdr_csum = 1;
 
-	if (!(features & NETIF_F_HW_ESP))
+	if (!(features & NETIF_F_HW_ESP) || !x->xso.offload_handle ||
+	    (x->xso.dev != skb->dev))
 		esp_features = features & ~(NETIF_F_SG | NETIF_F_CSUM_MASK);
 
-	segs = x->outer_mode->gso_segment(x, skb, esp_features);
-	if (IS_ERR_OR_NULL(segs))
-		goto out;
-
-	__skb_pull(skb, skb->data - skb_mac_header(skb));
-
-	skb2 = segs;
-	do {
-		struct sk_buff *nskb = skb2->next;
-
-		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;
 
-		x->outer_mode->xmit(x, skb2);
-
-		err = x->type_offload->xmit(x, skb2, esp_features);
-		if (err) {
-			kfree_skb_list(segs);
-			return ERR_PTR(err);
-		}
-
-		if (!skb_is_gso(skb2))
-			seq++;
-		else
-			seq += skb_shinfo(skb2)->gso_segs;
-
-		skb_push(skb2, skb2->mac_len);
-		skb2 = nskb;
-	} while (skb2);
-
-out:
-	return segs;
+	return x->outer_mode->gso_segment(x, skb, esp_features);
 }
 
 static int esp6_input_tail(struct xfrm_state *x, struct sk_buff *skb)
@@ -222,6 +183,7 @@ static int esp6_input_tail(struct xfrm_state *x, struct sk_buff *skb)
 
 static int esp6_xmit(struct xfrm_state *x, struct sk_buff *skb,  netdev_features_t features)
 {
+	int len;
 	int err;
 	int alen;
 	int blksize;
@@ -230,6 +192,7 @@ static int esp6_xmit(struct xfrm_state *x, struct sk_buff *skb,  netdev_features
 	struct crypto_aead *aead;
 	struct esp_info esp;
 	bool hw_offload = true;
+	__u32 seq;
 
 	esp.inplace = true;
 
@@ -265,28 +228,33 @@ static int esp6_xmit(struct xfrm_state *x, struct sk_buff *skb,  netdev_features
 			return esp.nfrags;
 	}
 
+	seq = xo->seq.low;
+
 	esph = ip_esp_hdr(skb);
 	esph->spi = x->id.spi;
 
 	skb_push(skb, -skb_network_offset(skb));
 
 	if (xo->flags & XFRM_GSO_SEGMENT) {
-		esph->seq_no = htonl(xo->seq.low);
-	} else {
-		int len;
+		esph->seq_no = htonl(seq);
+
+		if (!skb_is_gso(skb))
+			xo->seq.low++;
+		else
+			xo->seq.low += skb_shinfo(skb)->gso_segs;
+	}
+
+	esp.seqno = cpu_to_be64(xo->seq.low + ((u64)xo->seq.hi << 32));
 
 	len = skb->len - sizeof(struct ipv6hdr);
 	if (len > IPV6_MAXPLEN)
 		len = 0;
 
 	ipv6_hdr(skb)->payload_len = htons(len);
-	}
 
 	if (hw_offload)
 		return 0;
 
-	esp.seqno = cpu_to_be64(xo->seq.low + ((u64)xo->seq.hi << 32));
-
 	err = esp6_output_tail(x, skb, &esp);
 	if (err)
 		return err;

net/ipv6/xfrm6_mode_tunnel.c

@@ -105,17 +105,14 @@ static struct sk_buff *xfrm6_mode_tunnel_gso_segment(struct xfrm_state *x,
 {
 	__skb_push(skb, skb->mac_len);
 	return skb_mac_gso_segment(skb, features);
-
 }
 
 static void xfrm6_mode_tunnel_xmit(struct xfrm_state *x, struct sk_buff *skb)
 {
 	struct xfrm_offload *xo = xfrm_offload(skb);
 
-	if (xo->flags & XFRM_GSO_SEGMENT) {
-		skb->network_header = skb->network_header - x->props.header_len;
+	if (xo->flags & XFRM_GSO_SEGMENT)
 		skb->transport_header = skb->network_header + sizeof(struct ipv6hdr);
-	}
 
 	skb_reset_mac_len(skb);
 	pskb_pull(skb, skb->mac_len + x->props.header_len);

net/xfrm/xfrm_device.c

@@ -23,32 +23,99 @@
 #include <linux/notifier.h>
 
 #ifdef CONFIG_XFRM_OFFLOAD
-int 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)
 {
 	int err;
+	__u32 seq;
 	struct xfrm_state *x;
+	struct sk_buff *skb2;
+	netdev_features_t esp_features = features;
 	struct xfrm_offload *xo = xfrm_offload(skb);
 
-	if (skb_is_gso(skb))
-		return 0;
+	if (!xo)
+		return skb;
+
+	if (!(features & NETIF_F_HW_ESP))
+		esp_features = features & ~(NETIF_F_SG | NETIF_F_CSUM_MASK);
 
-	if (xo) {
 	x = skb->sp->xvec[skb->sp->len - 1];
 	if (xo->flags & XFRM_GRO || x->xso.flags & XFRM_OFFLOAD_INBOUND)
-			return 0;
+		return skb;
+
+	if (skb_is_gso(skb)) {
+		struct net_device *dev = skb->dev;
+
+		if (unlikely(!x->xso.offload_handle || (x->xso.dev != dev))) {
+			struct sk_buff *segs;
+
+			/* Packet got rerouted, fixup features and segment it. */
+			esp_features = esp_features & ~(NETIF_F_HW_ESP
+							| NETIF_F_GSO_ESP);
+
+			segs = skb_gso_segment(skb, esp_features);
+			if (IS_ERR(segs)) {
+				XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
+				kfree_skb(skb);
+				return NULL;
+			} else {
+				consume_skb(skb);
+				skb = segs;
+			}
+		} else {
+			return skb;
+		}
+	}
 
+	if (!skb->next) {
 		x->outer_mode->xmit(x, skb);
 
-		err = x->type_offload->xmit(x, skb, features);
+		err = x->type_offload->xmit(x, skb, esp_features);
 		if (err) {
 			XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
-			return err;
+			kfree_skb(skb);
+			return NULL;
 		}
 
 		skb_push(skb, skb->data - skb_mac_header(skb));
+
+		return skb;
 	}
 
-	return 0;
+	skb2 = skb;
+	seq = xo->seq.low;
+
+	do {
+		struct sk_buff *nskb = skb2->next;
+
+		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;
+
+		x->outer_mode->xmit(x, skb2);
+
+		err = x->type_offload->xmit(x, skb2, esp_features);
+		if (err) {
+			XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
+			skb2->next = nskb;
+			kfree_skb_list(skb2);
+			return NULL;
+		}
+
+		if (!skb_is_gso(skb2))
+			seq++;
+		else
+			seq += skb_shinfo(skb2)->gso_segs;
+
+		skb_push(skb2, skb2->data - skb_mac_header(skb2));
+
+		skb2 = nskb;
+	} while (skb2);
+
+	return skb;
 }
 EXPORT_SYMBOL_GPL(validate_xmit_xfrm);