[NET]: Prepare for zerocopy NFS and IPSEC.

- Import va10-hwchecksum-2.5.36.patch
- Import va11-udpsendfile-2.5.36.patch
- Implement new encapsulation friendly ipv4 output path.

include/linux/ip.h

@@ -137,8 +137,24 @@ struct inet_opt {
 	int			mc_index;	/* Multicast device index */
 	__u32			mc_addr;
 	struct ip_mc_socklist	*mc_list;	/* Group array */
+	struct page		*sndmsg_page;	/* Cached page for sendmsg */
+	u32			sndmsg_off;	/* Cached offset for sendmsg */
+	/*
+	 * Following members are used to retain the infomation to build
+	 * an ip header on each ip fragmentation while the socket is corked.
+	 */
+	struct {
+		unsigned int		flags;
+		unsigned int		fragsize;
+		struct ip_options	*opt;
+		struct rtable		*rt;
+		int			length; /* Total length of all frames */
+		u32			addr;
+	} cork;
 };
 
+#define IPCORK_OPT	1	/* ip-options has been held in ipcork.opt */
+
 struct ipv6_pinfo;
 
 /* WARNING: don't change the layout of the members in inet_sock! */

include/linux/skbuff.h

@@ -765,6 +765,15 @@ static inline int skb_headlen(const struct sk_buff *skb)
 	return skb->len - skb->data_len;
 }
 
+static inline int skb_pagelen(const struct sk_buff *skb)
+{
+	int i, len = 0;
+
+	for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
+		len += skb_shinfo(skb)->frags[i].size;
+	return len + skb_headlen(skb);
+}
+
 #define SKB_PAGE_ASSERT(skb) do { if (skb_shinfo(skb)->nr_frags) \
 					BUG(); } while (0)
 #define SKB_FRAG_ASSERT(skb) do { if (skb_shinfo(skb)->frag_list) \

include/linux/tcp.h

@@ -285,8 +285,6 @@ struct tcp_opt {
 
 	struct tcp_func		*af_specific;	/* Operations which are AF_INET{4,6} specific	*/
 	struct sk_buff		*send_head;	/* Front of stuff to transmit			*/
-	struct page		*sndmsg_page;	/* Cached page for sendmsg			*/
-	u32			sndmsg_off;	/* Cached offset for sendmsg			*/
 
  	__u32	rcv_wnd;	/* Current receiver window		*/
 	__u32	rcv_wup;	/* rcv_nxt on last window update sent	*/

include/linux/udp.h

@@ -17,6 +17,9 @@
 #ifndef _LINUX_UDP_H
 #define _LINUX_UDP_H
 
+#include <asm/byteorder.h>
+#include <net/sock.h>
+#include <linux/ip.h>
 
 struct udphdr {
 	__u16	source;
@@ -25,5 +28,33 @@ struct udphdr {
 	__u16	check;
 };
 
+/* UDP socket options */
+#define UDP_CORK	1	/* Never send partially complete segments */
+
+struct udp_opt {
+	int		pending;	/* Any pending frames ? */
+	unsigned int	corkflag;	/* Cork is required */
+	/*
+	 * Following members retains the infomation to create a UDP header
+	 * when the socket is uncorked.
+	 */
+	u32		saddr;		/* source address */
+	u32		daddr;		/* destination address */
+	__u16		sport;		/* source port */
+	__u16		dport;		/* destination port */
+	__u16		len;		/* total length of pending frames */
+};
+
+/* WARNING: don't change the layout of the members in udp_sock! */
+struct udp_sock {
+	struct sock	  sk;
+#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
+	struct ipv6_pinfo *pinet6;
+#endif
+	struct inet_opt	  inet;
+	struct udp_opt	  udp;
+};
+
+#define udp_sk(__sk) (&((struct udp_sock *)__sk)->udp)
 
 #endif	/* _LINUX_UDP_H */

include/net/dst.h

@@ -29,6 +29,7 @@ struct dst_entry
 	struct dst_entry        *next;
 	atomic_t		__refcnt;	/* client references	*/
 	int			__use;
+	struct dst_entry	*child;
 	struct net_device       *dev;
 	int			obsolete;
 	int			flags;
@@ -36,6 +37,8 @@ struct dst_entry
 	unsigned long		lastuse;
 	unsigned long		expires;
 
+	unsigned		header_len;	/* more space at head required */
+
 	unsigned		mxlock;
 	unsigned		pmtu;
 	unsigned		window;
@@ -108,17 +111,29 @@ void dst_release(struct dst_entry * dst)
 		atomic_dec(&dst->__refcnt);
 }
 
+/* Children define the path of the packet through the
+ * Linux networking.  Thus, destinations are stackable.
+ */
+
+static inline struct dst_entry *dst_pop(struct dst_entry *dst)
+{
+	struct dst_entry *child = dst_clone(dst->child);
+
+	dst_release(dst);
+	return child;
+}
+
 extern void * dst_alloc(struct dst_ops * ops);
 extern void __dst_free(struct dst_entry * dst);
-extern void dst_destroy(struct dst_entry * dst);
+extern struct dst_entry *dst_destroy(struct dst_entry * dst);
 
-static inline
-void dst_free(struct dst_entry * dst)
+static inline void dst_free(struct dst_entry * dst)
 {
 	if (dst->obsolete > 1)
 		return;
 	if (!atomic_read(&dst->__refcnt)) {
-		dst_destroy(dst);
+		dst = dst_destroy(dst);
+		if (!dst)
 			return;
 	}
 	__dst_free(dst);
@@ -155,6 +170,37 @@ static inline void dst_set_expires(struct dst_entry *dst, int timeout)
 		dst->expires = expires;
 }
 
+/* Output packet to network from transport.  */
+static inline int dst_output(struct sk_buff *skb)
+{
+	int err;
+
+	for (;;) {
+		err = skb->dst->output(skb);
+
+		if (likely(err == 0))
+			return err;
+		if (unlikely(err != NET_XMIT_BYPASS))
+			return err;
+	}
+}
+
+/* Input packet from network to transport.  */
+static inline int dst_input(struct sk_buff *skb)
+{
+	int err;
+
+	for (;;) {
+		err = skb->dst->input(skb);
+
+		if (likely(err == 0))
+			return err;
+		/* Oh, Jamal... Seems, I will not forgive you this mess. :-) */
+		if (unlikely(err != NET_XMIT_BYPASS))
+			return err;
+	}
+}
+
 extern void		dst_init(void);
 
 #endif

include/net/ip.h

@@ -102,12 +102,26 @@ extern int		ip_build_xmit(struct sock *sk,
 				      int getfrag (const void *,
 						   char *,
 						   unsigned int,
-						   unsigned int),
+						   unsigned int,
+						   struct sk_buff *),
 				      const void *frag,
 				      unsigned length,
 				      struct ipcm_cookie *ipc,
 				      struct rtable *rt,
 				      int flags);
+extern int		ip_append_data(struct sock *sk,
+				       int getfrag(void *from, char *to, int offset, int len,
+						   int odd, struct sk_buff *skb),
+				void *from, int len, int protolen,
+				struct ipcm_cookie *ipc,
+				struct rtable *rt,
+				unsigned int flags);
+extern int		generic_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb);
+extern ssize_t		ip_append_page(struct sock *sk, struct page *page,
+				int offset, size_t size, int flags);
+extern int		ip_push_pending_frames(struct sock *sk);
+extern void		ip_flush_pending_frames(struct sock *sk);
+
 
 /*
  *	Map a multicast IP onto multicast MAC for type Token Ring.

include/net/sock.h

@@ -249,6 +249,8 @@ struct proto {
 					   struct msghdr *msg,
 					int len, int noblock, int flags, 
 					int *addr_len);
+	int			(*sendpage)(struct sock *sk, struct page *page,
+					int offset, size_t size, int flags);
 	int			(*bind)(struct sock *sk, 
 					struct sockaddr *uaddr, int addr_len);
 

include/net/tcp.h

@@ -1851,7 +1851,7 @@ static inline void tcp_v4_setup_caps(struct sock *sk, struct dst_entry *dst)
 {
 	sk->route_caps = dst->dev->features;
 	if (sk->route_caps & NETIF_F_TSO) {
-		if (sk->no_largesend)
+		if (sk->no_largesend || dst->header_len)
 			sk->route_caps &= ~NETIF_F_TSO;
 	}
 }

include/net/udp.h

@@ -76,6 +76,4 @@ extern struct udp_mib udp_statistics[NR_CPUS*2];
 #define UDP_INC_STATS_BH(field)		SNMP_INC_STATS_BH(udp_statistics, field)
 #define UDP_INC_STATS_USER(field) 	SNMP_INC_STATS_USER(udp_statistics, field)
 
-#define udp_sock inet_sock
-
 #endif	/* _UDP_H */

net/core/dst.c

@@ -40,7 +40,6 @@ static void dst_run_gc(unsigned long);
 static struct timer_list dst_gc_timer =
 	{ data: DST_GC_MIN, function: dst_run_gc };
 
-
 static void dst_run_gc(unsigned long dummy)
 {
 	int    delayed = 0;
@@ -60,6 +59,10 @@ static void dst_run_gc(unsigned long dummy)
 			delayed++;
 			continue;
 		}
+		if (dst->child) {
+			dst->child->next = dst->next;
+			*dstp = dst->child;
+		} else
 			*dstp = dst->next;
 		dst_destroy(dst);
 	}
@@ -141,10 +144,16 @@ void __dst_free(struct dst_entry * dst)
 	spin_unlock_bh(&dst_lock);
 }
 
-void dst_destroy(struct dst_entry * dst)
+struct dst_entry *dst_destroy(struct dst_entry * dst)
 {
-	struct neighbour *neigh = dst->neighbour;
-	struct hh_cache *hh = dst->hh;
+	struct dst_entry *child;
+	struct neighbour *neigh;
+	struct hh_cache *hh;
+
+again:
+	neigh = dst->neighbour;
+	hh = dst->hh;
+	child = dst->child;
 
 	dst->hh = NULL;
 	if (hh && atomic_dec_and_test(&hh->hh_refcnt))
@@ -165,6 +174,12 @@ void dst_destroy(struct dst_entry * dst)
 	atomic_dec(&dst_total);
 #endif
 	kmem_cache_free(dst->ops->kmem_cachep, dst);
+
+	dst = child;
+	if (dst && !atomic_read(&dst->__refcnt))
+		goto again;
+
+	return dst;
 }
 
 static int dst_dev_event(struct notifier_block *this, unsigned long event, void *ptr)

net/ipv4/af_inet.c

@@ -774,6 +774,21 @@ int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
 	return sk->prot->sendmsg(iocb, sk, msg, size);
 }
 
+
+ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
+{
+	struct sock *sk = sock->sk;
+
+	/* We may need to bind the socket. */
+	if (!inet_sk(sk)->num && inet_autobind(sk))
+		return -EAGAIN;
+
+	if (sk->prot->sendpage)
+		return sk->prot->sendpage(sk, page, offset, size, flags);
+	return sock_no_sendpage(sock, page, offset, size, flags);
+}
+
+
 int inet_shutdown(struct socket *sock, int how)
 {
 	struct sock *sk = sock->sk;
@@ -977,7 +992,7 @@ struct proto_ops inet_dgram_ops = {
 	.sendmsg =	inet_sendmsg,
 	.recvmsg =	inet_recvmsg,
 	.mmap =		sock_no_mmap,
-	.sendpage =	sock_no_sendpage,
+	.sendpage =	inet_sendpage,
 };
 
 struct net_proto_family inet_family_ops = {

net/ipv4/icmp.c

@@ -357,12 +357,14 @@ static void icmp_out_count(int type)
  *	checksum.
  */
 static int icmp_glue_bits(const void *p, char *to, unsigned int offset,
-			  unsigned int fraglen)
+			  unsigned int fraglen, struct sk_buff *skb)
 {
 	struct icmp_bxm *icmp_param = (struct icmp_bxm *)p;
 	struct icmphdr *icmph;
 	unsigned int csum;
 
+	skb->ip_summed = CHECKSUM_NONE;
+
 	if (offset) {
 		icmp_param->csum =
 			skb_copy_and_csum_bits(icmp_param->skb,

net/ipv4/ip_sockglue.c

@@ -437,8 +437,10 @@ int ip_setsockopt(struct sock *sk, int level, int optname, char *optval, int opt
 				    (!((1<<sk->state)&(TCPF_LISTEN|TCPF_CLOSE))
 				     && inet->daddr != LOOPBACK4_IPV6)) {
 #endif
+					if (inet->opt)
+						tp->ext_header_len -= inet->opt->optlen;
 					if (opt)
-						tp->ext_header_len = opt->optlen;
+						tp->ext_header_len += opt->optlen;
 					tcp_sync_mss(sk, tp->pmtu_cookie);
 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 				}

net/ipv4/raw.c

@@ -259,9 +259,10 @@ struct rawfakehdr
  */
   
 static int raw_getfrag(const void *p, char *to, unsigned int offset,
-			unsigned int fraglen)
+			unsigned int fraglen, struct sk_buff *skb)
 {
 	struct rawfakehdr *rfh = (struct rawfakehdr *) p;
+	skb->ip_summed = CHECKSUM_NONE; /* Is there any good place to set it? */
 	return memcpy_fromiovecend(to, rfh->iov, offset, fraglen);
 }
 
@@ -270,10 +271,12 @@ static int raw_getfrag(const void *p, char *to, unsigned int offset,
  */
  
 static int raw_getrawfrag(const void *p, char *to, unsigned int offset,
-				unsigned int fraglen)
+				unsigned int fraglen, struct sk_buff *skb)
 {
 	struct rawfakehdr *rfh = (struct rawfakehdr *) p;
 
+	skb->ip_summed = CHECKSUM_NONE; /* Is there any good place to set it? */
+
 	if (memcpy_fromiovecend(to, rfh->iov, offset, fraglen))
 		return -EFAULT;
 

net/ipv4/tcp.c

@@ -204,6 +204,8 @@
  *		Andi Kleen 	:	Make poll agree with SIGIO
  *	Salvatore Sanfilippo	:	Support SO_LINGER with linger == 1 and
  *					lingertime == 0 (RFC 793 ABORT Call)
+ *	Hirokazu Takahashi	:	Use copy_from_user() instead of
+ *					csum_and_copy_from_user() if possible.
  *
  *		This program is free software; you can redistribute it and/or
  *		modify it under the terms of the GNU General Public License
@@ -958,8 +960,8 @@ ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset,
 	return res;
 }
 
-#define TCP_PAGE(sk)	(tcp_sk(sk)->sndmsg_page)
-#define TCP_OFF(sk)	(tcp_sk(sk)->sndmsg_off)
+#define TCP_PAGE(sk)	(inet_sk(sk)->sndmsg_page)
+#define TCP_OFF(sk)	(inet_sk(sk)->sndmsg_off)
 
 static inline int tcp_copy_to_page(struct sock *sk, char *from,
 				   struct sk_buff *skb, struct page *page,
@@ -968,18 +970,22 @@ static inline int tcp_copy_to_page(struct sock *sk, char *from,
 	int err = 0;
 	unsigned int csum;
 
+	if (skb->ip_summed == CHECKSUM_NONE) {
 		csum = csum_and_copy_from_user(from, page_address(page) + off,
 				       copy, 0, &err);
-	if (!err) {
-		if (skb->ip_summed == CHECKSUM_NONE)
+		if (err) return err;
 		skb->csum = csum_block_add(skb->csum, csum, skb->len);
+	} else {
+		if (copy_from_user(page_address(page) + off, from, copy))
+			return -EFAULT;
+	}
+
 	skb->len += copy;
 	skb->data_len += copy;
 	skb->truesize += copy;
 	sk->wmem_queued += copy;
 	sk->forward_alloc -= copy;
-	}
-	return err;
+	return 0;
 }
 
 static inline int skb_add_data(struct sk_buff *skb, char *from, int copy)
@@ -988,12 +994,17 @@ static inline int skb_add_data(struct sk_buff *skb, char *from, int copy)
 	unsigned int csum;
 	int off = skb->len;
 
+	if (skb->ip_summed == CHECKSUM_NONE) {
 		csum = csum_and_copy_from_user(from, skb_put(skb, copy),
 				       copy, 0, &err);
 		if (!err) {
 			skb->csum = csum_block_add(skb->csum, csum, off);
 			return 0;
 		}
+	} else {
+		if (!copy_from_user(skb_put(skb, copy), from, copy))
+			return 0;
+	}
 
 	__skb_trim(skb, off);
 	return -EFAULT;
@@ -1075,6 +1086,12 @@ int tcp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 				if (!skb)
 					goto wait_for_memory;
 
+				/*
+				 * Check whether we can use HW checksum.
+				 */
+				if (sk->route_caps & (NETIF_F_IP_CSUM|NETIF_F_NO_CSUM|NETIF_F_HW_CSUM))
+					skb->ip_summed = CHECKSUM_HW;
+
 				skb_entail(sk, tp, skb);
 				copy = mss_now;
 			}

net/ipv4/tcp_ipv4.c

@@ -781,6 +781,7 @@ int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 
 	__sk_dst_set(sk, &rt->u.dst);
 	tcp_v4_setup_caps(sk, &rt->u.dst);
+	tp->ext_header_len += rt->u.dst.header_len;
 
 	if (!inet->opt || !inet->opt->srr)
 		daddr = rt->rt_dst;
@@ -1577,6 +1578,7 @@ struct sock *tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
 	newtp->ext_header_len = 0;
 	if (newinet->opt)
 		newtp->ext_header_len = newinet->opt->optlen;
+	newtp->ext_header_len += dst->header_len;
 	newinet->id = newtp->write_seq ^ jiffies;
 
 	tcp_sync_mss(newsk, dst->pmtu);
@@ -2087,8 +2089,8 @@ static int tcp_v4_destroy_sock(struct sock *sk)
 		tcp_put_port(sk);
 
 	/* If sendmsg cached page exists, toss it. */
-	if (tp->sndmsg_page)
-		__free_page(tp->sndmsg_page);
+	if (inet_sk(sk)->sndmsg_page)
+		__free_page(inet_sk(sk)->sndmsg_page);
 
 	atomic_dec(&tcp_sockets_allocated);
 

net/ipv6/tcp_ipv6.c

@@ -1876,6 +1876,7 @@ static int tcp_v6_init_sock(struct sock *sk)
 static int tcp_v6_destroy_sock(struct sock *sk)
 {
 	struct tcp_opt *tp = tcp_sk(sk);
+	struct inet_opt *inet = inet_sk(sk);
 
 	tcp_clear_xmit_timers(sk);
 
@@ -1893,8 +1894,8 @@ static int tcp_v6_destroy_sock(struct sock *sk)
 		tcp_put_port(sk);
 
 	/* If sendmsg cached page exists, toss it. */
-	if (tp->sndmsg_page != NULL)
-		__free_page(tp->sndmsg_page);
+	if (inet->sndmsg_page != NULL)
+		__free_page(inet->sndmsg_page);
 
 	atomic_dec(&tcp_sockets_allocated);