[IPv4]: More output path work.

- Fix a bug noted by Maxim Giryaev (gem@asplinux.ru)
when corking on a non-SG interface
- Convert ICMP and ip_send_reply to ip_append_foo.
- Rename generic_getfrag to more namespace friendly
ip_generic_getfrag.
- Kill n_iov member from ip_reply_arg.

include/net/ip.h

@@ -116,7 +116,7 @@ extern int		ip_append_data(struct sock *sk,
 				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 int		ip_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);
@@ -141,8 +141,7 @@ static inline void ip_tr_mc_map(u32 addr, char *buf)
 }
 
 struct ip_reply_arg {
-	struct iovec iov[2];   
-	int          n_iov;    /* redundant */
+	struct iovec iov[1];   
 	u32 	     csum; 
 	int	     csumoffset; /* u16 offset of csum in iov[0].iov_base */
 				 /* -1 if not needed */ 

net/ipv4/icmp.c

@@ -101,7 +101,6 @@ struct icmp_bxm {
 	int offset;
 	int data_len;
 
-	unsigned int csum;
 	struct {
 		struct icmphdr icmph;
 		__u32	       times[3];
@@ -356,41 +355,47 @@ static void icmp_out_count(int type)
  *	Checksum each fragment, and on the first include the headers and final
  *	checksum.
  */
-static int icmp_glue_bits(const void *p, char *to, unsigned int offset,
-			  unsigned int fraglen, struct sk_buff *skb)
+int
+icmp_glue_bits(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
 {
-	struct icmp_bxm *icmp_param = (struct icmp_bxm *)p;
-	struct icmphdr *icmph;
+	struct icmp_bxm *icmp_param = (struct icmp_bxm *)from;
 	unsigned int csum;
 
-	skb->ip_summed = CHECKSUM_NONE;
-
-	if (offset) {
-		icmp_param->csum =
-			skb_copy_and_csum_bits(icmp_param->skb,
-					       icmp_param->offset +
-					       (offset - icmp_param->head_len),
-						to, fraglen, icmp_param->csum);
-		goto out;
-	}
+	csum = skb_copy_and_csum_bits(icmp_param->skb,
+				      icmp_param->offset + offset,
+				      to, len, 0);
 
-	/*
-	 *	First fragment includes header. Note that we've done
-	 *	the other fragments first, so that we get the checksum
-	 *	for the whole packet here.
-	 */
-	csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
-					 to, icmp_param->head_len,
-					 icmp_param->csum);
-	csum = skb_copy_and_csum_bits(icmp_param->skb, icmp_param->offset,
-				      to + icmp_param->head_len,
-				      fraglen - icmp_param->head_len, csum);
-	icmph = (struct icmphdr *)to;
-	icmph->checksum = csum_fold(csum);
-out:
+	skb->csum = csum_block_add(skb->csum, csum, odd);
 	return 0;
 }
 
+static void
+icmp_push_reply(struct icmp_bxm *icmp_param, struct ipcm_cookie *ipc, struct rtable *rt)
+{
+	struct sk_buff *skb;
+
+	ip_append_data(icmp_socket->sk, icmp_glue_bits, icmp_param,
+		       icmp_param->data_len+icmp_param->head_len,
+		       icmp_param->head_len,
+		       ipc, rt, MSG_DONTWAIT);
+
+	if ((skb = skb_peek(&icmp_socket->sk->write_queue)) != NULL) {
+		struct icmphdr *icmph = skb->h.icmph;
+		unsigned int csum = 0;
+		struct sk_buff *skb1;
+
+		skb_queue_walk(&icmp_socket->sk->write_queue, skb1) {
+			csum = csum_add(csum, skb1->csum);
+		}
+		csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
+						 (char*)icmph, icmp_param->head_len,	
+						 csum);
+		icmph->checksum = csum_fold(csum);
+		skb->ip_summed = CHECKSUM_NONE;
+		ip_push_pending_frames(icmp_socket->sk);
+	}
+}
+
 /*
  *	Driving logic for building and sending ICMP messages.
  */
@@ -408,7 +413,6 @@ static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
 		goto out;
 
 	icmp_param->data.icmph.checksum = 0;
-	icmp_param->csum = 0;
 	icmp_out_count(icmp_param->data.icmph.type);
 
 	inet->tos = skb->nh.iph->tos;
@@ -429,11 +433,8 @@ static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
 			goto out_unlock;
 	}
 	if (icmpv4_xrlim_allow(rt, icmp_param->data.icmph.type,
-			       icmp_param->data.icmph.code)) {
-		ip_build_xmit(sk, icmp_glue_bits, icmp_param,
-			      icmp_param->data_len+icmp_param->head_len,
-			      &ipc, rt, MSG_DONTWAIT);
-	}
+			       icmp_param->data.icmph.code))
+		icmp_push_reply(icmp_param, &ipc, rt);
 	ip_rt_put(rt);
 out_unlock:
 	icmp_xmit_unlock_bh();
@@ -565,7 +566,6 @@ void icmp_send(struct sk_buff *skb_in, int type, int code, u32 info)
 	icmp_param.data.icmph.code	 = code;
 	icmp_param.data.icmph.un.gateway = info;
 	icmp_param.data.icmph.checksum	 = 0;
-	icmp_param.csum	  = 0;
 	icmp_param.skb	  = skb_in;
 	icmp_param.offset = skb_in->nh.raw - skb_in->data;
 	icmp_out_count(icmp_param.data.icmph.type);
@@ -599,9 +599,7 @@ void icmp_send(struct sk_buff *skb_in, int type, int code, u32 info)
 		icmp_param.data_len = room;
 	icmp_param.head_len = sizeof(struct icmphdr);
 
-	ip_build_xmit(icmp_socket->sk, icmp_glue_bits, &icmp_param,
-		      icmp_param.data_len + sizeof(struct icmphdr),
-		      &ipc, rt, MSG_DONTWAIT);
+	icmp_push_reply(&icmp_param, &ipc, rt);
 ende:
 	ip_rt_put(rt);
 out_unlock:

net/ipv4/ip_output.c

@@ -1003,7 +1003,7 @@ int ip_fragment(struct sk_buff *skb, int (*output)(struct sk_buff*))
 }
 
 int
-generic_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
+ip_generic_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
 {
 	struct iovec *iov = from;
 
@@ -1095,7 +1095,7 @@ int ip_append_data(struct sock *sk,
 		opt = ipc->opt;
 		if (opt) {
 			if (inet->cork.opt == NULL)
-				inet->cork.opt = kmalloc(sizeof(struct ip_options)+40, GFP_KERNEL);
+				inet->cork.opt = kmalloc(sizeof(struct ip_options)+40, sk->allocation);
 			memcpy(inet->cork.opt, opt, sizeof(struct ip_options)+opt->optlen);
 			inet->cork.flags |= IPCORK_OPT;
 			inet->cork.addr = ipc->addr;
@@ -1129,7 +1129,6 @@ int ip_append_data(struct sock *sk,
 		return -EMSGSIZE;
 	}
 
-#if 0 /* Not now */
 	/*
 	 * transhdrlen > 0 means that this is the first fragment and we wish
 	 * it won't be fragmented in the future.
@@ -1139,7 +1138,6 @@ int ip_append_data(struct sock *sk,
 	    rt->u.dst.dev->features&(NETIF_F_IP_CSUM|NETIF_F_NO_CSUM|NETIF_F_HW_CSUM) &&
 	    !exthdrlen)
 		csummode = CHECKSUM_HW;
-#endif
 
 	inet->cork.length += length;
 
@@ -1151,6 +1149,7 @@ int ip_append_data(struct sock *sk,
 			char *data;
 			unsigned int datalen;
 			unsigned int fraglen;
+			unsigned int alloclen;
 			BUG_TRAP(copy == 0);
 
 alloc_new_skb:
@@ -1159,12 +1158,19 @@ int ip_append_data(struct sock *sk,
 				datalen = length;
 
 			fraglen = datalen + fragheaderlen;
+			if ((flags & MSG_MORE) && 
+			    !(rt->u.dst.dev->features&NETIF_F_SG))
+				alloclen = maxfraglen;
+			else
+				alloclen = datalen + fragheaderlen;
 			if (!(flags & MSG_DONTWAIT) || transhdrlen) {
-				skb = sock_alloc_send_skb(sk, fraglen + hh_len + 15,
-							  (flags & MSG_DONTWAIT), &err);
+				skb = sock_alloc_send_skb(sk, 
+						alloclen + hh_len + 15,
+						(flags & MSG_DONTWAIT), &err);
 			} else {
-				skb = sock_wmalloc(sk, fraglen + hh_len + 15, 1,
-						   sk->allocation);
+				skb = sock_wmalloc(sk, 
+						alloclen + hh_len + 15, 1,
+						sk->allocation);
 				if (unlikely(skb == NULL))
 					err = -ENOBUFS;
 			}
@@ -1206,20 +1212,15 @@ int ip_append_data(struct sock *sk,
 			continue;
 		}
 
+		if (copy > length)
+			copy = length;
+
 		if (!(rt->u.dst.dev->features&NETIF_F_SG)) {
-			int off;
-			if (!((skb->len - fragheaderlen) & 7))
-				goto alloc_new_skb;
+			unsigned int off;
 
-			/* 
-			 * Align the start address of the next IP fragment
-			 * on 8 byte boundary.
-			 */
-			copy = 8 - ((skb->len - fragheaderlen) & 7);
 			off = skb->len;
-			if (copy > length)
-				copy = length;
-			if (getfrag(from, skb_put(skb, copy), offset, copy, off, skb) < 0) {
+			if (getfrag(from, skb_put(skb, copy), 
+					offset, copy, off, skb) < 0) {
 				__skb_trim(skb, off);
 				err = -EFAULT;
 				goto error;
@@ -1231,9 +1232,6 @@ int ip_append_data(struct sock *sk,
 			int off = inet->sndmsg_off;
 			unsigned int left;
 
-			if (copy > length)
-				copy = length;
-
 			if (page && (left = PAGE_SIZE - off) > 0) {
 				if (copy >= left)
 					copy = left;
@@ -1518,39 +1516,14 @@ void ip_flush_pending_frames(struct sock *sk)
 /*
  *	Fetch data from kernel space and fill in checksum if needed.
  */
-static int ip_reply_glue_bits(const void *dptr, char *to, unsigned int offset, 
-			      unsigned int fraglen, struct sk_buff *skb)
+static int ip_reply_glue_bits(void *dptr, char *to, int offset, 
+			      int len, int odd, struct sk_buff *skb)
 {
-        struct ip_reply_arg *dp = (struct ip_reply_arg*)dptr;
-	u16 *pktp = (u16 *)to;
-	struct iovec *iov; 
-	int len; 
-	int hdrflag = 1; 
-
-	iov = &dp->iov[0]; 
-	if (offset >= iov->iov_len) { 
-		offset -= iov->iov_len;
-		iov++; 
-		hdrflag = 0; 
-	}
-	len = iov->iov_len - offset;
-	if (fraglen > len) { /* overlapping. */ 
-		dp->csum = csum_partial_copy_nocheck(iov->iov_base+offset, to, len,
-					     dp->csum);
-		offset = 0;
-		fraglen -= len; 
-		to += len; 
-		iov++;
-	}
-
-	dp->csum = csum_partial_copy_nocheck(iov->iov_base+offset, to, fraglen, 
-					     dp->csum); 
-
-	if (hdrflag && dp->csumoffset)
-		*(pktp + dp->csumoffset) = csum_fold(dp->csum); /* fill in checksum */
-	skb->ip_summed = CHECKSUM_NONE;
+	unsigned int csum;
 
-	return 0;	       
+	csum = csum_partial_copy_nocheck(dptr+offset, to, len, 0);
+	skb->csum = csum_block_add(skb->csum, csum, odd);
+	return 0;  
 }
 
 /* 
@@ -1606,7 +1579,15 @@ void ip_send_reply(struct sock *sk, struct sk_buff *skb, struct ip_reply_arg *ar
 	inet->tos = skb->nh.iph->tos;
 	sk->priority = skb->priority;
 	sk->protocol = skb->nh.iph->protocol;
-	ip_build_xmit(sk, ip_reply_glue_bits, arg, len, &ipc, rt, MSG_DONTWAIT);
+	ip_append_data(sk, ip_reply_glue_bits, arg->iov->iov_base, len, 0,
+		       &ipc, rt, MSG_DONTWAIT);
+	if ((skb = skb_peek(&sk->write_queue)) != NULL) {
+		if (arg->csumoffset >= 0)
+			*((u16 *)skb->h.raw + arg->csumoffset) = csum_fold(csum_add(skb->csum, arg->csum));
+		skb->ip_summed = CHECKSUM_NONE;
+		ip_push_pending_frames(sk);
+	}
+
 	bh_unlock_sock(sk);
 
 	ip_rt_put(rt);

net/ipv4/tcp_ipv4.c

@@ -1188,7 +1188,6 @@ static void tcp_v4_send_reset(struct sk_buff *skb)
 	arg.csum = csum_tcpudp_nofold(skb->nh.iph->daddr,
 				      skb->nh.iph->saddr, /*XXX*/
 				      sizeof(struct tcphdr), IPPROTO_TCP, 0);
-	arg.n_iov = 1;
 	arg.csumoffset = offsetof(struct tcphdr, check) / 2;
 
 	inet_sk(tcp_socket->sk)->ttl = sysctl_ip_default_ttl;
@@ -1217,7 +1216,6 @@ static void tcp_v4_send_ack(struct sk_buff *skb, u32 seq, u32 ack,
 
 	arg.iov[0].iov_base = (unsigned char *)&rep;
 	arg.iov[0].iov_len  = sizeof(rep.th);
-	arg.n_iov = 1;
 	if (ts) {
 		rep.tsopt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
 				     (TCPOPT_TIMESTAMP << 8) |

net/ipv4/udp.c

@@ -484,7 +484,7 @@ static int udp_getfrag(const void *p, char * to, unsigned int offset, unsigned i
 		if (skb->ip_summed == CHECKSUM_HW) {
 			skb->csum = offsetof(struct udphdr, check);
 			ufh->uh.check = ~csum_tcpudp_magic(ufh->saddr, ufh->daddr, 
-					  ntohs(ufh->uh.len), IPPROTO_UDP, ufh->wcheck);
+					  ntohs(ufh->uh.len), IPPROTO_UDP, 0);
 			memcpy(to, ufh, sizeof(struct udphdr));
 			return memcpy_fromiovecend(to+sizeof(struct udphdr), ufh->iov, offset,
 					   fraglen-sizeof(struct udphdr));
@@ -730,7 +730,9 @@ int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 
 do_append_data:
 	up->len += ulen;
-	err = ip_append_data(sk, generic_getfrag, msg->msg_iov, ulen, sizeof(struct udphdr), &ipc, rt, msg->msg_flags);
+	err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, ulen, 
+			sizeof(struct udphdr), &ipc, rt, 
+			corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
 	if (err)
 		udp_flush_pending_frames(sk);
 	else if (!corkreq)