[NET] gso: Fix up GSO packets with broken checksums

Certain subsystems in the stack (e.g., netfilter) can break the partial
checksum on GSO packets.  Until they're fixed, this patch allows this to
work by recomputing the partial checksums through the GSO mechanism.

Once they've all been converted to update the partial checksum instead of
clearing it, this workaround can be removed.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

5 files changed, 166 insertions, 32 deletions

diff --git a/net/core/dev.c b/net/core/dev.c
index 0096349ee38..4d2b5167d7f 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -1162,9 +1162,17 @@ int skb_checksum_help(struct sk_buff *skb, int inward)
 	unsigned int csum;
 	int ret = 0, offset = skb->h.raw - skb->data;
 
-	if (inward) {
-		skb->ip_summed = CHECKSUM_NONE;
-		goto out;
+	if (inward)
+		goto out_set_summed;
+
+	if (unlikely(skb_shinfo(skb)->gso_size)) {
+		static int warned;
+
+		WARN_ON(!warned);
+		warned = 1;
+
+		/* Let GSO fix up the checksum. */
+		goto out_set_summed;
 	}
 
 	if (skb_cloned(skb)) {
@@ -1181,6 +1189,8 @@ int skb_checksum_help(struct sk_buff *skb, int inward)
 	BUG_ON(skb->csum + 2 > offset);
 
 	*(u16*)(skb->h.raw + skb->csum) = csum_fold(csum);
+
+out_set_summed:
 	skb->ip_summed = CHECKSUM_NONE;
 out:	
 	return ret;
@@ -1201,17 +1211,35 @@ struct sk_buff *skb_gso_segment(struct sk_buff *skb, int features)
 	struct sk_buff *segs = ERR_PTR(-EPROTONOSUPPORT);
 	struct packet_type *ptype;
 	int type = skb->protocol;
+	int err;
 
 	BUG_ON(skb_shinfo(skb)->frag_list);
-	BUG_ON(skb->ip_summed != CHECKSUM_HW);
 
 	skb->mac.raw = skb->data;
 	skb->mac_len = skb->nh.raw - skb->data;
 	__skb_pull(skb, skb->mac_len);
 
+	if (unlikely(skb->ip_summed != CHECKSUM_HW)) {
+		static int warned;
+
+		WARN_ON(!warned);
+		warned = 1;
+
+		if (skb_header_cloned(skb) &&
+		    (err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
+			return ERR_PTR(err);
+	}
+
 	rcu_read_lock();
 	list_for_each_entry_rcu(ptype, &ptype_base[ntohs(type) & 15], list) {
 		if (ptype->type == type && !ptype->dev && ptype->gso_segment) {
+			if (unlikely(skb->ip_summed != CHECKSUM_HW)) {
+				err = ptype->gso_send_check(skb);
+				segs = ERR_PTR(err);
+				if (err || skb_gso_ok(skb, features))
+					break;
+				__skb_push(skb, skb->data - skb->nh.raw);
+			}
 			segs = ptype->gso_segment(skb, features);
 			break;
 		}
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 318d4674faa..c84a32070f8 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -1097,6 +1097,40 @@ int inet_sk_rebuild_header(struct sock *sk)
 
 EXPORT_SYMBOL(inet_sk_rebuild_header);
 
+static int inet_gso_send_check(struct sk_buff *skb)
+{
+	struct iphdr *iph;
+	struct net_protocol *ops;
+	int proto;
+	int ihl;
+	int err = -EINVAL;
+
+	if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
+		goto out;
+
+	iph = skb->nh.iph;
+	ihl = iph->ihl * 4;
+	if (ihl < sizeof(*iph))
+		goto out;
+
+	if (unlikely(!pskb_may_pull(skb, ihl)))
+		goto out;
+
+	skb->h.raw = __skb_pull(skb, ihl);
+	iph = skb->nh.iph;
+	proto = iph->protocol & (MAX_INET_PROTOS - 1);
+	err = -EPROTONOSUPPORT;
+
+	rcu_read_lock();
+	ops = rcu_dereference(inet_protos[proto]);
+	if (likely(ops && ops->gso_send_check))
+		err = ops->gso_send_check(skb);
+	rcu_read_unlock();
+
+out:
+	return err;
+}
+
 static struct sk_buff *inet_gso_segment(struct sk_buff *skb, int features)
 {
 	struct sk_buff *segs = ERR_PTR(-EINVAL);
@@ -1162,6 +1196,7 @@ static struct net_protocol igmp_protocol = {
 static struct net_protocol tcp_protocol = {
 	.handler =	tcp_v4_rcv,
 	.err_handler =	tcp_v4_err,
+	.gso_send_check = tcp_v4_gso_send_check,
 	.gso_segment =	tcp_tso_segment,
 	.no_policy =	1,
 };
@@ -1208,6 +1243,7 @@ static int ipv4_proc_init(void);
 static struct packet_type ip_packet_type = {
 	.type = __constant_htons(ETH_P_IP),
 	.func = ip_rcv,
+	.gso_send_check = inet_gso_send_check,
 	.gso_segment = inet_gso_segment,
 };
 
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 5a886e6efbb..a891133f00e 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -496,6 +496,24 @@ void tcp_v4_send_check(struct sock *sk, int len, struct sk_buff *skb)
 	}
 }
 
+int tcp_v4_gso_send_check(struct sk_buff *skb)
+{
+	struct iphdr *iph;
+	struct tcphdr *th;
+
+	if (!pskb_may_pull(skb, sizeof(*th)))
+		return -EINVAL;
+
+	iph = skb->nh.iph;
+	th = skb->h.th;
+
+	th->check = 0;
+	th->check = ~tcp_v4_check(th, skb->len, iph->saddr, iph->daddr, 0);
+	skb->csum = offsetof(struct tcphdr, check);
+	skb->ip_summed = CHECKSUM_HW;
+	return 0;
+}
+
 /*
  *	This routine will send an RST to the other tcp.
  *
diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c
index 0c17dec11c8..43327264e69 100644
--- a/net/ipv6/ipv6_sockglue.c
+++ b/net/ipv6/ipv6_sockglue.c
@@ -57,29 +57,11 @@
 
 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
 
-static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
+static struct inet6_protocol *ipv6_gso_pull_exthdrs(struct sk_buff *skb,
+						    int proto)
 {
-	struct sk_buff *segs = ERR_PTR(-EINVAL);
-	struct ipv6hdr *ipv6h;
-	struct inet6_protocol *ops;
-	int proto;
+	struct inet6_protocol *ops = NULL;
 
-	if (unlikely(skb_shinfo(skb)->gso_type &
-		     ~(SKB_GSO_UDP |
-		       SKB_GSO_DODGY |
-		       SKB_GSO_TCP_ECN |
-		       SKB_GSO_TCPV6 |
-		       0)))
-		goto out;
-
-	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
-		goto out;
-
-	ipv6h = skb->nh.ipv6h;
-	proto = ipv6h->nexthdr;
-	__skb_pull(skb, sizeof(*ipv6h));
-
-	rcu_read_lock();
 	for (;;) {
 		struct ipv6_opt_hdr *opth;
 		int len;
@@ -88,30 +70,80 @@ static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
 			ops = rcu_dereference(inet6_protos[proto]);
 
 			if (unlikely(!ops))
-				goto unlock;
+				break;
 
 			if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
 				break;
 		}
 
 		if (unlikely(!pskb_may_pull(skb, 8)))
-			goto unlock;
+			break;
 
 		opth = (void *)skb->data;
 		len = opth->hdrlen * 8 + 8;
 
 		if (unlikely(!pskb_may_pull(skb, len)))
-			goto unlock;
+			break;
 
 		proto = opth->nexthdr;
 		__skb_pull(skb, len);
 	}
 
-	skb->h.raw = skb->data;
-	if (likely(ops->gso_segment))
-		segs = ops->gso_segment(skb, features);
+	return ops;
+}
+
+static int ipv6_gso_send_check(struct sk_buff *skb)
+{
+	struct ipv6hdr *ipv6h;
+	struct inet6_protocol *ops;
+	int err = -EINVAL;
+
+	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
+		goto out;
 
-unlock:
+	ipv6h = skb->nh.ipv6h;
+	__skb_pull(skb, sizeof(*ipv6h));
+	err = -EPROTONOSUPPORT;
+
+	rcu_read_lock();
+	ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
+	if (likely(ops && ops->gso_send_check)) {
+		skb->h.raw = skb->data;
+		err = ops->gso_send_check(skb);
+	}
+	rcu_read_unlock();
+
+out:
+	return err;
+}
+
+static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
+{
+	struct sk_buff *segs = ERR_PTR(-EINVAL);
+	struct ipv6hdr *ipv6h;
+	struct inet6_protocol *ops;
+
+	if (unlikely(skb_shinfo(skb)->gso_type &
+		     ~(SKB_GSO_UDP |
+		       SKB_GSO_DODGY |
+		       SKB_GSO_TCP_ECN |
+		       SKB_GSO_TCPV6 |
+		       0)))
+		goto out;
+
+	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
+		goto out;
+
+	ipv6h = skb->nh.ipv6h;
+	__skb_pull(skb, sizeof(*ipv6h));
+	segs = ERR_PTR(-EPROTONOSUPPORT);
+
+	rcu_read_lock();
+	ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
+	if (likely(ops && ops->gso_segment)) {
+		skb->h.raw = skb->data;
+		segs = ops->gso_segment(skb, features);
+	}
 	rcu_read_unlock();
 
 	if (unlikely(IS_ERR(segs)))
@@ -130,6 +162,7 @@ out:
 static struct packet_type ipv6_packet_type = {
 	.type = __constant_htons(ETH_P_IPV6), 
 	.func = ipv6_rcv,
+	.gso_send_check = ipv6_gso_send_check,
 	.gso_segment = ipv6_gso_segment,
 };
 
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 5bdcb9002cf..923989d0520 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -552,6 +552,24 @@ static void tcp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb)
 	}
 }
 
+static int tcp_v6_gso_send_check(struct sk_buff *skb)
+{
+	struct ipv6hdr *ipv6h;
+	struct tcphdr *th;
+
+	if (!pskb_may_pull(skb, sizeof(*th)))
+		return -EINVAL;
+
+	ipv6h = skb->nh.ipv6h;
+	th = skb->h.th;
+
+	th->check = 0;
+	th->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len,
+				     IPPROTO_TCP, 0);
+	skb->csum = offsetof(struct tcphdr, check);
+	skb->ip_summed = CHECKSUM_HW;
+	return 0;
+}
 
 static void tcp_v6_send_reset(struct sk_buff *skb)
 {
@@ -1603,6 +1621,7 @@ struct proto tcpv6_prot = {
 static struct inet6_protocol tcpv6_protocol = {
 	.handler	=	tcp_v6_rcv,
 	.err_handler	=	tcp_v6_err,
+	.gso_send_check	=	tcp_v6_gso_send_check,
 	.gso_segment	=	tcp_tso_segment,
 	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
 };