Merge branch 'inet_csums'

Tom Herbert says:

====================
net: Checksum offload changes

I am working on overhauling RX checksum offload. Goals of this effort
are:

- Specify what exactly it means when driver returns CHECKSUM_UNNECESSARY
- Preserve CHECKSUM_COMPLETE through encapsulation layers
- Don't do skb_checksum more than once per packet
- Unify GRO and non-GRO csum verification as much as possible
- Unify the checksum functions (checksum_init)
- Simply code

What is in this first patch set:

- Create a common "checksum_init" function which is called from
  TCPv{4,6} and UDPv{4,6}
- Add some for RFC6936, UDP/IPv6 zero checksums
- Add architecture support for csum_add and provide implementations
  for x86_64 and Sparc 32 and 64 bit (please test the latter)

Please review carefully and test if possible, mucking with basic
checksum functions is always a little precarious :-)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

arch/sparc/include/asm/checksum_32.h

@@ -238,4 +238,16 @@ static inline __sum16 ip_compute_csum(const void *buff, int len)
 	return csum_fold(csum_partial(buff, len, 0));
 }
 
+#define HAVE_ARCH_CSUM_ADD
+static inline __wsum csum_add(__wsum csum, __wsum addend)
+{
+	__asm__ __volatile__(
+		"addcc   %0, %1, %0\n"
+		"addx    %0, %%g0, %0"
+		: "=r" (csum)
+		: "r" (addend), "0" (csum));
+
+	return csum;
+}
+
 #endif /* !(__SPARC_CHECKSUM_H) */

arch/sparc/include/asm/checksum_64.h

@@ -164,4 +164,16 @@ static inline __sum16 ip_compute_csum(const void *buff, int len)
 	return csum_fold(csum_partial(buff, len, 0));
 }
 
+#define HAVE_ARCH_CSUM_ADD
+static inline __wsum csum_add(__wsum csum, __wsum addend)
+{
+	__asm__ __volatile__(
+		"addcc   %0, %1, %0\n"
+		"addx    %0, %%g0, %0"
+		: "=r" (csum)
+		: "r" (addend), "0" (csum));
+
+	return csum;
+}
+
 #endif /* !(__SPARC64_CHECKSUM_H) */

arch/x86/include/asm/checksum_64.h

@@ -184,8 +184,15 @@ static inline unsigned add32_with_carry(unsigned a, unsigned b)
 	asm("addl %2,%0\n\t"
 	    "adcl $0,%0"
 	    : "=r" (a)
-	    : "0" (a), "r" (b));
+	    : "0" (a), "rm" (b));
 	return a;
 }
 
+#define HAVE_ARCH_CSUM_ADD
+static inline __wsum csum_add(__wsum csum, __wsum addend)
+{
+	return (__force __wsum)add32_with_carry((__force unsigned)csum,
+						(__force unsigned)addend);
+}
+
 #endif /* _ASM_X86_CHECKSUM_64_H */

include/linux/skbuff.h

@@ -2741,6 +2741,99 @@ static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
 	       0 : __skb_checksum_complete(skb);
 }
 
+/* Check if we need to perform checksum complete validation.
+ *
+ * Returns true if checksum complete is needed, false otherwise
+ * (either checksum is unnecessary or zero checksum is allowed).
+ */
+static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
+						  bool zero_okay,
+						  __sum16 check)
+{
+	if (skb_csum_unnecessary(skb)) {
+		return false;
+	} else if (zero_okay && !check) {
+		skb->ip_summed = CHECKSUM_UNNECESSARY;
+		return false;
+	}
+
+	return true;
+}
+
+/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
+ * in checksum_init.
+ */
+#define CHECKSUM_BREAK 76
+
+/* Validate (init) checksum based on checksum complete.
+ *
+ * Return values:
+ *   0: checksum is validated or try to in skb_checksum_complete. In the latter
+ *	case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
+ *	checksum is stored in skb->csum for use in __skb_checksum_complete
+ *   non-zero: value of invalid checksum
+ *
+ */
+static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
+						       bool complete,
+						       __wsum psum)
+{
+	if (skb->ip_summed == CHECKSUM_COMPLETE) {
+		if (!csum_fold(csum_add(psum, skb->csum))) {
+			skb->ip_summed = CHECKSUM_UNNECESSARY;
+			return 0;
+		}
+	}
+
+	skb->csum = psum;
+
+	if (complete || skb->len <= CHECKSUM_BREAK)
+		return __skb_checksum_complete(skb);
+
+	return 0;
+}
+
+static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
+{
+	return 0;
+}
+
+/* Perform checksum validate (init). Note that this is a macro since we only
+ * want to calculate the pseudo header which is an input function if necessary.
+ * First we try to validate without any computation (checksum unnecessary) and
+ * then calculate based on checksum complete calling the function to compute
+ * pseudo header.
+ *
+ * Return values:
+ *   0: checksum is validated or try to in skb_checksum_complete
+ *   non-zero: value of invalid checksum
+ */
+#define __skb_checksum_validate(skb, proto, complete,			\
+				zero_okay, check, compute_pseudo)	\
+({									\
+	__sum16 __ret = 0;						\
+	if (__skb_checksum_validate_needed(skb, zero_okay, check))	\
+		__ret = __skb_checksum_validate_complete(skb,		\
+				complete, compute_pseudo(skb, proto));	\
+	__ret;								\
+})
+
+#define skb_checksum_init(skb, proto, compute_pseudo)			\
+	__skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
+
+#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo)	\
+	__skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
+
+#define skb_checksum_validate(skb, proto, compute_pseudo)		\
+	__skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
+
+#define skb_checksum_validate_zero_check(skb, proto, check,		\
+					 compute_pseudo)		\
+	__skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
+
+#define skb_checksum_simple_validate(skb)				\
+	__skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
+
 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
 void nf_conntrack_destroy(struct nf_conntrack *nfct);
 static inline void nf_conntrack_put(struct nf_conntrack *nfct)

include/net/checksum.h

@@ -57,12 +57,14 @@ static __inline__ __wsum csum_and_copy_to_user
 }
 #endif
 
+#ifndef HAVE_ARCH_CSUM_ADD
 static inline __wsum csum_add(__wsum csum, __wsum addend)
 {
 	u32 res = (__force u32)csum;
 	res += (__force u32)addend;
 	return (__force __wsum)(res + (res < (__force u32)addend));
 }
+#endif
 
 static inline __wsum csum_sub(__wsum csum, __wsum addend)
 {

include/net/ip.h

@@ -342,6 +342,12 @@ static inline void ip_select_ident_more(struct sk_buff *skb, struct dst_entry *d
 		__ip_select_ident(iph, dst, more);
 }
 
+static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
+{
+	return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
+				  skb->len, proto, 0);
+}
+
 /*
  *	Map a multicast IP onto multicast MAC for type ethernet.
  */

include/net/ip6_checksum.h

@@ -41,6 +41,13 @@ __sum16 csum_ipv6_magic(const struct in6_addr *saddr,
 			__wsum csum);
 #endif
 
+static inline __wsum ip6_compute_pseudo(struct sk_buff *skb, int proto)
+{
+	return ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
+					    &ipv6_hdr(skb)->daddr,
+					    skb->len, proto, 0));
+}
+
 static __inline__ __sum16 tcp_v6_check(int len,
 				   const struct in6_addr *saddr,
 				   const struct in6_addr *daddr,

net/ipv4/tcp_ipv4.c

@@ -1744,28 +1744,6 @@ static struct sock *tcp_v4_hnd_req(struct sock *sk, struct sk_buff *skb)
 	return sk;
 }
 
-static __sum16 tcp_v4_checksum_init(struct sk_buff *skb)
-{
-	const struct iphdr *iph = ip_hdr(skb);
-
-	if (skb->ip_summed == CHECKSUM_COMPLETE) {
-		if (!tcp_v4_check(skb->len, iph->saddr,
-				  iph->daddr, skb->csum)) {
-			skb->ip_summed = CHECKSUM_UNNECESSARY;
-			return 0;
-		}
-	}
-
-	skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
-				       skb->len, IPPROTO_TCP, 0);
-
-	if (skb->len <= 76) {
-		return __skb_checksum_complete(skb);
-	}
-	return 0;
-}
-
-
 /* The socket must have it's spinlock held when we get
  * here.
  *
@@ -1960,7 +1938,8 @@ int tcp_v4_rcv(struct sk_buff *skb)
 	 * Packet length and doff are validated by header prediction,
 	 * provided case of th->doff==0 is eliminated.
 	 * So, we defer the checks. */
-	if (!skb_csum_unnecessary(skb) && tcp_v4_checksum_init(skb))
+
+	if (skb_checksum_init(skb, IPPROTO_TCP, inet_compute_pseudo))
 		goto csum_error;
 
 	th = tcp_hdr(skb);

net/ipv4/udp.c

@@ -1672,7 +1672,6 @@ static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
 static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
 				 int proto)
 {
-	const struct iphdr *iph;
 	int err;
 
 	UDP_SKB_CB(skb)->partial_cov = 0;
@@ -1684,22 +1683,8 @@ static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
 			return err;
 	}
 
-	iph = ip_hdr(skb);
-	if (uh->check == 0) {
-		skb->ip_summed = CHECKSUM_UNNECESSARY;
-	} else if (skb->ip_summed == CHECKSUM_COMPLETE) {
-		if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len,
-				      proto, skb->csum))
-			skb->ip_summed = CHECKSUM_UNNECESSARY;
-	}
-	if (!skb_csum_unnecessary(skb))
-		skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
-					       skb->len, proto, 0);
-	/* Probably, we should checksum udp header (it should be in cache
-	 * in any case) and data in tiny packets (< rx copybreak).
-	 */
-
-	return 0;
+	return skb_checksum_init_zero_check(skb, proto, uh->check,
+					    inet_compute_pseudo);
 }
 
 /*

net/ipv6/ip6_checksum.c

@@ -75,25 +75,12 @@ int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh, int proto)
 			return err;
 	}
 
-	if (uh->check == 0) {
-		/* RFC 2460 section 8.1 says that we SHOULD log
-		   this error. Well, it is reasonable.
-		 */
-		LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
-			       &ipv6_hdr(skb)->saddr, ntohs(uh->source),
-			       &ipv6_hdr(skb)->daddr, ntohs(uh->dest));
-		return 1;
-	}
-	if (skb->ip_summed == CHECKSUM_COMPLETE &&
-	    !csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
-			     skb->len, proto, skb->csum))
-		skb->ip_summed = CHECKSUM_UNNECESSARY;
-
-	if (!skb_csum_unnecessary(skb))
-		skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
-							 &ipv6_hdr(skb)->daddr,
-							 skb->len, proto, 0));
-
-	return 0;
+	/* To support RFC 6936 (allow zero checksum in UDP/IPV6 for tunnels)
+	 * we accept a checksum of zero here. When we find the socket
+	 * for the UDP packet we'll check if that socket allows zero checksum
+	 * for IPv6 (set by socket option).
+	 */
+	return skb_checksum_init_zero_check(skb, proto, uh->check,
+					   ip6_compute_pseudo);
 }
 EXPORT_SYMBOL(udp6_csum_init);

net/ipv6/tcp_ipv6.c

@@ -1294,25 +1294,6 @@ static struct sock *tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
 	return NULL;
 }
 
-static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
-{
-	if (skb->ip_summed == CHECKSUM_COMPLETE) {
-		if (!tcp_v6_check(skb->len, &ipv6_hdr(skb)->saddr,
-				  &ipv6_hdr(skb)->daddr, skb->csum)) {
-			skb->ip_summed = CHECKSUM_UNNECESSARY;
-			return 0;
-		}
-	}
-
-	skb->csum = ~csum_unfold(tcp_v6_check(skb->len,
-					      &ipv6_hdr(skb)->saddr,
-					      &ipv6_hdr(skb)->daddr, 0));
-
-	if (skb->len <= 76)
-		return __skb_checksum_complete(skb);
-	return 0;
-}
-
 /* The socket must have it's spinlock held when we get
  * here.
  *
@@ -1486,7 +1467,7 @@ static int tcp_v6_rcv(struct sk_buff *skb)
 	if (!pskb_may_pull(skb, th->doff*4))
 		goto discard_it;
 
-	if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb))
+	if (skb_checksum_init(skb, IPPROTO_TCP, ip6_compute_pseudo))
 		goto csum_error;
 
 	th = tcp_hdr(skb);