ip_output.c 39.0 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7
/*
 * INET		An implementation of the TCP/IP protocol suite for the LINUX
 *		operating system.  INET is implemented using the  BSD Socket
 *		interface as the means of communication with the user level.
 *
 *		The Internet Protocol (IP) output module.
 *
8
 * Authors:	Ross Biro
L
Linus Torvalds 已提交
9 10 11 12 13 14 15 16 17 18 19 20 21 22
 *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 *		Donald Becker, <becker@super.org>
 *		Alan Cox, <Alan.Cox@linux.org>
 *		Richard Underwood
 *		Stefan Becker, <stefanb@yello.ping.de>
 *		Jorge Cwik, <jorge@laser.satlink.net>
 *		Arnt Gulbrandsen, <agulbra@nvg.unit.no>
 *		Hirokazu Takahashi, <taka@valinux.co.jp>
 *
 *	See ip_input.c for original log
 *
 *	Fixes:
 *		Alan Cox	:	Missing nonblock feature in ip_build_xmit.
 *		Mike Kilburn	:	htons() missing in ip_build_xmit.
23
 *		Bradford Johnson:	Fix faulty handling of some frames when
L
Linus Torvalds 已提交
24 25 26 27 28 29 30 31 32 33
 *					no route is found.
 *		Alexander Demenshin:	Missing sk/skb free in ip_queue_xmit
 *					(in case if packet not accepted by
 *					output firewall rules)
 *		Mike McLagan	:	Routing by source
 *		Alexey Kuznetsov:	use new route cache
 *		Andi Kleen:		Fix broken PMTU recovery and remove
 *					some redundant tests.
 *	Vitaly E. Lavrov	:	Transparent proxy revived after year coma.
 *		Andi Kleen	: 	Replace ip_reply with ip_send_reply.
34 35 36
 *		Andi Kleen	:	Split fast and slow ip_build_xmit path
 *					for decreased register pressure on x86
 *					and more readibility.
L
Linus Torvalds 已提交
37 38 39 40 41 42 43 44 45 46 47 48 49 50 51
 *		Marc Boucher	:	When call_out_firewall returns FW_QUEUE,
 *					silently drop skb instead of failing with -EPERM.
 *		Detlev Wengorz	:	Copy protocol for fragments.
 *		Hirokazu Takahashi:	HW checksumming for outgoing UDP
 *					datagrams.
 *		Hirokazu Takahashi:	sendfile() on UDP works now.
 */

#include <asm/uaccess.h>
#include <linux/module.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/string.h>
#include <linux/errno.h>
A
Al Viro 已提交
52
#include <linux/highmem.h>
53
#include <linux/slab.h>
L
Linus Torvalds 已提交
54 55 56 57 58 59 60 61 62 63 64 65 66 67 68

#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/in.h>
#include <linux/inet.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/proc_fs.h>
#include <linux/stat.h>
#include <linux/init.h>

#include <net/snmp.h>
#include <net/ip.h>
#include <net/protocol.h>
#include <net/route.h>
69
#include <net/xfrm.h>
L
Linus Torvalds 已提交
70 71 72 73 74 75 76 77 78 79 80
#include <linux/skbuff.h>
#include <net/sock.h>
#include <net/arp.h>
#include <net/icmp.h>
#include <net/checksum.h>
#include <net/inetpeer.h>
#include <linux/igmp.h>
#include <linux/netfilter_ipv4.h>
#include <linux/netfilter_bridge.h>
#include <linux/mroute.h>
#include <linux/netlink.h>
81
#include <linux/tcp.h>
L
Linus Torvalds 已提交
82

83
int sysctl_ip_default_ttl __read_mostly = IPDEFTTL;
84
EXPORT_SYMBOL(sysctl_ip_default_ttl);
L
Linus Torvalds 已提交
85 86

/* Generate a checksum for an outgoing IP datagram. */
87
void ip_send_check(struct iphdr *iph)
L
Linus Torvalds 已提交
88 89 90 91
{
	iph->check = 0;
	iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
}
E
Eric Dumazet 已提交
92
EXPORT_SYMBOL(ip_send_check);
L
Linus Torvalds 已提交
93

H
Herbert Xu 已提交
94 95 96 97 98 99
int __ip_local_out(struct sk_buff *skb)
{
	struct iphdr *iph = ip_hdr(skb);

	iph->tot_len = htons(skb->len);
	ip_send_check(iph);
100 101
	return nf_hook(NFPROTO_IPV4, NF_INET_LOCAL_OUT, skb, NULL,
		       skb_dst(skb)->dev, dst_output);
H
Herbert Xu 已提交
102 103
}

104
int ip_local_out_sk(struct sock *sk, struct sk_buff *skb)
H
Herbert Xu 已提交
105 106 107 108 109
{
	int err;

	err = __ip_local_out(skb);
	if (likely(err == 1))
110
		err = dst_output_sk(sk, skb);
H
Herbert Xu 已提交
111 112 113

	return err;
}
114
EXPORT_SYMBOL_GPL(ip_local_out_sk);
H
Herbert Xu 已提交
115

L
Linus Torvalds 已提交
116 117 118 119 120
static inline int ip_select_ttl(struct inet_sock *inet, struct dst_entry *dst)
{
	int ttl = inet->uc_ttl;

	if (ttl < 0)
121
		ttl = ip4_dst_hoplimit(dst);
L
Linus Torvalds 已提交
122 123 124
	return ttl;
}

125
/*
L
Linus Torvalds 已提交
126 127 128 129
 *		Add an ip header to a skbuff and send it out.
 *
 */
int ip_build_and_send_pkt(struct sk_buff *skb, struct sock *sk,
130
			  __be32 saddr, __be32 daddr, struct ip_options_rcu *opt)
L
Linus Torvalds 已提交
131 132
{
	struct inet_sock *inet = inet_sk(sk);
E
Eric Dumazet 已提交
133
	struct rtable *rt = skb_rtable(skb);
L
Linus Torvalds 已提交
134 135 136
	struct iphdr *iph;

	/* Build the IP header. */
137
	skb_push(skb, sizeof(struct iphdr) + (opt ? opt->opt.optlen : 0));
138
	skb_reset_network_header(skb);
139
	iph = ip_hdr(skb);
L
Linus Torvalds 已提交
140 141 142
	iph->version  = 4;
	iph->ihl      = 5;
	iph->tos      = inet->tos;
143
	if (ip_dont_fragment(sk, &rt->dst))
L
Linus Torvalds 已提交
144 145 146
		iph->frag_off = htons(IP_DF);
	else
		iph->frag_off = 0;
147
	iph->ttl      = ip_select_ttl(inet, &rt->dst);
148 149
	iph->daddr    = (opt && opt->opt.srr ? opt->opt.faddr : daddr);
	iph->saddr    = saddr;
L
Linus Torvalds 已提交
150
	iph->protocol = sk->sk_protocol;
E
Eric Dumazet 已提交
151
	ip_select_ident(skb, sk);
L
Linus Torvalds 已提交
152

153 154 155
	if (opt && opt->opt.optlen) {
		iph->ihl += opt->opt.optlen>>2;
		ip_options_build(skb, &opt->opt, daddr, rt, 0);
L
Linus Torvalds 已提交
156 157 158
	}

	skb->priority = sk->sk_priority;
159
	skb->mark = sk->sk_mark;
L
Linus Torvalds 已提交
160 161

	/* Send it out. */
H
Herbert Xu 已提交
162
	return ip_local_out(skb);
L
Linus Torvalds 已提交
163
}
164 165
EXPORT_SYMBOL_GPL(ip_build_and_send_pkt);

L
Linus Torvalds 已提交
166 167
static inline int ip_finish_output2(struct sk_buff *skb)
{
E
Eric Dumazet 已提交
168
	struct dst_entry *dst = skb_dst(skb);
169
	struct rtable *rt = (struct rtable *)dst;
L
Linus Torvalds 已提交
170
	struct net_device *dev = dst->dev;
171
	unsigned int hh_len = LL_RESERVED_SPACE(dev);
172
	struct neighbour *neigh;
173
	u32 nexthop;
L
Linus Torvalds 已提交
174

175 176 177 178
	if (rt->rt_type == RTN_MULTICAST) {
		IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUTMCAST, skb->len);
	} else if (rt->rt_type == RTN_BROADCAST)
		IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUTBCAST, skb->len);
179

L
Linus Torvalds 已提交
180
	/* Be paranoid, rather than too clever. */
181
	if (unlikely(skb_headroom(skb) < hh_len && dev->header_ops)) {
L
Linus Torvalds 已提交
182 183 184 185 186 187 188 189 190
		struct sk_buff *skb2;

		skb2 = skb_realloc_headroom(skb, LL_RESERVED_SPACE(dev));
		if (skb2 == NULL) {
			kfree_skb(skb);
			return -ENOMEM;
		}
		if (skb->sk)
			skb_set_owner_w(skb2, skb->sk);
191
		consume_skb(skb);
L
Linus Torvalds 已提交
192 193 194
		skb = skb2;
	}

195
	rcu_read_lock_bh();
J
Julian Anastasov 已提交
196
	nexthop = (__force u32) rt_nexthop(rt, ip_hdr(skb)->daddr);
197 198 199
	neigh = __ipv4_neigh_lookup_noref(dev, nexthop);
	if (unlikely(!neigh))
		neigh = __neigh_create(&arp_tbl, &nexthop, dev, false);
200
	if (!IS_ERR(neigh)) {
201
		int res = dst_neigh_output(dst, neigh, skb);
202

203
		rcu_read_unlock_bh();
204 205
		return res;
	}
206
	rcu_read_unlock_bh();
207

208 209
	net_dbg_ratelimited("%s: No header cache and no neighbour!\n",
			    __func__);
L
Linus Torvalds 已提交
210 211 212 213
	kfree_skb(skb);
	return -EINVAL;
}

214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255
static int ip_finish_output_gso(struct sk_buff *skb)
{
	netdev_features_t features;
	struct sk_buff *segs;
	int ret = 0;

	/* common case: locally created skb or seglen is <= mtu */
	if (((IPCB(skb)->flags & IPSKB_FORWARDED) == 0) ||
	      skb_gso_network_seglen(skb) <= ip_skb_dst_mtu(skb))
		return ip_finish_output2(skb);

	/* Slowpath -  GSO segment length is exceeding the dst MTU.
	 *
	 * This can happen in two cases:
	 * 1) TCP GRO packet, DF bit not set
	 * 2) skb arrived via virtio-net, we thus get TSO/GSO skbs directly
	 * from host network stack.
	 */
	features = netif_skb_features(skb);
	segs = skb_gso_segment(skb, features & ~NETIF_F_GSO_MASK);
	if (IS_ERR(segs)) {
		kfree_skb(skb);
		return -ENOMEM;
	}

	consume_skb(skb);

	do {
		struct sk_buff *nskb = segs->next;
		int err;

		segs->next = NULL;
		err = ip_fragment(segs, ip_finish_output2);

		if (err && ret == 0)
			ret = err;
		segs = nskb;
	} while (segs);

	return ret;
}

P
Patrick McHardy 已提交
256
static int ip_finish_output(struct sk_buff *skb)
L
Linus Torvalds 已提交
257
{
258 259
#if defined(CONFIG_NETFILTER) && defined(CONFIG_XFRM)
	/* Policy lookup after SNAT yielded a new policy */
E
Eric Dumazet 已提交
260
	if (skb_dst(skb)->xfrm != NULL) {
261 262 263
		IPCB(skb)->flags |= IPSKB_REROUTED;
		return dst_output(skb);
	}
264
#endif
265 266 267 268
	if (skb_is_gso(skb))
		return ip_finish_output_gso(skb);

	if (skb->len > ip_skb_dst_mtu(skb))
269
		return ip_fragment(skb, ip_finish_output2);
270 271

	return ip_finish_output2(skb);
L
Linus Torvalds 已提交
272 273
}

274
int ip_mc_output(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
275
{
E
Eric Dumazet 已提交
276
	struct rtable *rt = skb_rtable(skb);
277
	struct net_device *dev = rt->dst.dev;
L
Linus Torvalds 已提交
278 279 280 281

	/*
	 *	If the indicated interface is up and running, send the packet.
	 */
282
	IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len);
L
Linus Torvalds 已提交
283 284 285 286 287 288 289 290 291

	skb->dev = dev;
	skb->protocol = htons(ETH_P_IP);

	/*
	 *	Multicasts are looped back for other local users
	 */

	if (rt->rt_flags&RTCF_MULTICAST) {
292
		if (sk_mc_loop(sk)
L
Linus Torvalds 已提交
293 294 295 296 297 298 299 300 301
#ifdef CONFIG_IP_MROUTE
		/* Small optimization: do not loopback not local frames,
		   which returned after forwarding; they will be  dropped
		   by ip_mr_input in any case.
		   Note, that local frames are looped back to be delivered
		   to local recipients.

		   This check is duplicated in ip_mr_input at the moment.
		 */
302 303 304
		    &&
		    ((rt->rt_flags & RTCF_LOCAL) ||
		     !(IPCB(skb)->flags & IPSKB_FORWARDED))
L
Linus Torvalds 已提交
305
#endif
306
		   ) {
L
Linus Torvalds 已提交
307 308
			struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
			if (newskb)
309 310
				NF_HOOK(NFPROTO_IPV4, NF_INET_POST_ROUTING,
					newskb, NULL, newskb->dev,
311
					dev_loopback_xmit);
L
Linus Torvalds 已提交
312 313 314 315
		}

		/* Multicasts with ttl 0 must not go beyond the host */

316
		if (ip_hdr(skb)->ttl == 0) {
L
Linus Torvalds 已提交
317 318 319 320 321 322 323 324
			kfree_skb(skb);
			return 0;
		}
	}

	if (rt->rt_flags&RTCF_BROADCAST) {
		struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
		if (newskb)
325
			NF_HOOK(NFPROTO_IPV4, NF_INET_POST_ROUTING, newskb,
326
				NULL, newskb->dev, dev_loopback_xmit);
L
Linus Torvalds 已提交
327 328
	}

329 330
	return NF_HOOK_COND(NFPROTO_IPV4, NF_INET_POST_ROUTING, skb, NULL,
			    skb->dev, ip_finish_output,
331
			    !(IPCB(skb)->flags & IPSKB_REROUTED));
L
Linus Torvalds 已提交
332 333
}

334
int ip_output(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
335
{
E
Eric Dumazet 已提交
336
	struct net_device *dev = skb_dst(skb)->dev;
337

338
	IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len);
L
Linus Torvalds 已提交
339

340 341 342
	skb->dev = dev;
	skb->protocol = htons(ETH_P_IP);

343
	return NF_HOOK_COND(NFPROTO_IPV4, NF_INET_POST_ROUTING, skb, NULL, dev,
344
			    ip_finish_output,
345
			    !(IPCB(skb)->flags & IPSKB_REROUTED));
L
Linus Torvalds 已提交
346 347
}

348 349 350 351 352 353 354 355 356 357 358 359 360 361
/*
 * copy saddr and daddr, possibly using 64bit load/stores
 * Equivalent to :
 *   iph->saddr = fl4->saddr;
 *   iph->daddr = fl4->daddr;
 */
static void ip_copy_addrs(struct iphdr *iph, const struct flowi4 *fl4)
{
	BUILD_BUG_ON(offsetof(typeof(*fl4), daddr) !=
		     offsetof(typeof(*fl4), saddr) + sizeof(fl4->saddr));
	memcpy(&iph->saddr, &fl4->saddr,
	       sizeof(fl4->saddr) + sizeof(fl4->daddr));
}

362 363
/* Note: skb->sk can be different from sk, in case of tunnels */
int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl)
L
Linus Torvalds 已提交
364 365
{
	struct inet_sock *inet = inet_sk(sk);
366
	struct ip_options_rcu *inet_opt;
367
	struct flowi4 *fl4;
L
Linus Torvalds 已提交
368 369
	struct rtable *rt;
	struct iphdr *iph;
370
	int res;
L
Linus Torvalds 已提交
371 372 373 374

	/* Skip all of this if the packet is already routed,
	 * f.e. by something like SCTP.
	 */
375
	rcu_read_lock();
376
	inet_opt = rcu_dereference(inet->inet_opt);
377
	fl4 = &fl->u.ip4;
E
Eric Dumazet 已提交
378
	rt = skb_rtable(skb);
L
Linus Torvalds 已提交
379 380 381 382 383 384
	if (rt != NULL)
		goto packet_routed;

	/* Make sure we can route this packet. */
	rt = (struct rtable *)__sk_dst_check(sk, 0);
	if (rt == NULL) {
A
Al Viro 已提交
385
		__be32 daddr;
L
Linus Torvalds 已提交
386 387

		/* Use correct destination address if we have options. */
E
Eric Dumazet 已提交
388
		daddr = inet->inet_daddr;
389 390
		if (inet_opt && inet_opt->opt.srr)
			daddr = inet_opt->opt.faddr;
L
Linus Torvalds 已提交
391

392 393 394 395
		/* If this fails, retransmit mechanism of transport layer will
		 * keep trying until route appears or the connection times
		 * itself out.
		 */
396
		rt = ip_route_output_ports(sock_net(sk), fl4, sk,
397 398 399 400 401 402 403 404
					   daddr, inet->inet_saddr,
					   inet->inet_dport,
					   inet->inet_sport,
					   sk->sk_protocol,
					   RT_CONN_FLAGS(sk),
					   sk->sk_bound_dev_if);
		if (IS_ERR(rt))
			goto no_route;
405
		sk_setup_caps(sk, &rt->dst);
L
Linus Torvalds 已提交
406
	}
407
	skb_dst_set_noref(skb, &rt->dst);
L
Linus Torvalds 已提交
408 409

packet_routed:
J
Julian Anastasov 已提交
410
	if (inet_opt && inet_opt->opt.is_strictroute && rt->rt_uses_gateway)
L
Linus Torvalds 已提交
411 412 413
		goto no_route;

	/* OK, we know where to send it, allocate and build IP header. */
414
	skb_push(skb, sizeof(struct iphdr) + (inet_opt ? inet_opt->opt.optlen : 0));
415
	skb_reset_network_header(skb);
416
	iph = ip_hdr(skb);
417
	*((__be16 *)iph) = htons((4 << 12) | (5 << 8) | (inet->tos & 0xff));
W
WANG Cong 已提交
418
	if (ip_dont_fragment(sk, &rt->dst) && !skb->ignore_df)
L
Linus Torvalds 已提交
419 420 421
		iph->frag_off = htons(IP_DF);
	else
		iph->frag_off = 0;
422
	iph->ttl      = ip_select_ttl(inet, &rt->dst);
L
Linus Torvalds 已提交
423
	iph->protocol = sk->sk_protocol;
424 425
	ip_copy_addrs(iph, fl4);

L
Linus Torvalds 已提交
426 427
	/* Transport layer set skb->h.foo itself. */

428 429 430
	if (inet_opt && inet_opt->opt.optlen) {
		iph->ihl += inet_opt->opt.optlen >> 2;
		ip_options_build(skb, &inet_opt->opt, inet->inet_daddr, rt, 0);
L
Linus Torvalds 已提交
431 432
	}

E
Eric Dumazet 已提交
433
	ip_select_ident_segs(skb, sk, skb_shinfo(skb)->gso_segs ?: 1);
L
Linus Torvalds 已提交
434

435
	/* TODO : should we use skb->sk here instead of sk ? */
L
Linus Torvalds 已提交
436
	skb->priority = sk->sk_priority;
437
	skb->mark = sk->sk_mark;
L
Linus Torvalds 已提交
438

439 440 441
	res = ip_local_out(skb);
	rcu_read_unlock();
	return res;
L
Linus Torvalds 已提交
442 443

no_route:
444
	rcu_read_unlock();
P
Pavel Emelyanov 已提交
445
	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
L
Linus Torvalds 已提交
446 447 448
	kfree_skb(skb);
	return -EHOSTUNREACH;
}
E
Eric Dumazet 已提交
449
EXPORT_SYMBOL(ip_queue_xmit);
L
Linus Torvalds 已提交
450 451 452 453 454 455 456


static void ip_copy_metadata(struct sk_buff *to, struct sk_buff *from)
{
	to->pkt_type = from->pkt_type;
	to->priority = from->priority;
	to->protocol = from->protocol;
E
Eric Dumazet 已提交
457
	skb_dst_drop(to);
458
	skb_dst_copy(to, from);
L
Linus Torvalds 已提交
459
	to->dev = from->dev;
T
Thomas Graf 已提交
460
	to->mark = from->mark;
L
Linus Torvalds 已提交
461 462 463 464 465 466 467

	/* Copy the flags to each fragment. */
	IPCB(to)->flags = IPCB(from)->flags;

#ifdef CONFIG_NET_SCHED
	to->tc_index = from->tc_index;
#endif
468
	nf_copy(to, from);
469 470
#if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
	to->ipvs_property = from->ipvs_property;
L
Linus Torvalds 已提交
471
#endif
472
	skb_copy_secmark(to, from);
L
Linus Torvalds 已提交
473 474 475 476 477 478 479 480 481
}

/*
 *	This IP datagram is too large to be sent in one piece.  Break it up into
 *	smaller pieces (each of size equal to IP header plus
 *	a block of the data of the original IP data part) that will yet fit in a
 *	single device frame, and queue such a frame for sending.
 */

482
int ip_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
L
Linus Torvalds 已提交
483 484 485 486 487
{
	struct iphdr *iph;
	int ptr;
	struct net_device *dev;
	struct sk_buff *skb2;
488
	unsigned int mtu, hlen, left, len, ll_rs;
L
Linus Torvalds 已提交
489
	int offset;
490
	__be16 not_last_frag;
E
Eric Dumazet 已提交
491
	struct rtable *rt = skb_rtable(skb);
L
Linus Torvalds 已提交
492 493
	int err = 0;

494
	dev = rt->dst.dev;
L
Linus Torvalds 已提交
495 496 497 498 499

	/*
	 *	Point into the IP datagram header.
	 */

500
	iph = ip_hdr(skb);
L
Linus Torvalds 已提交
501

502
	mtu = ip_skb_dst_mtu(skb);
W
WANG Cong 已提交
503
	if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
504
		     (IPCB(skb)->frag_max_size &&
505
		      IPCB(skb)->frag_max_size > mtu))) {
P
Pavel Emelyanov 已提交
506
		IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
L
Linus Torvalds 已提交
507
		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
508
			  htonl(mtu));
L
Linus Torvalds 已提交
509 510 511 512 513 514 515 516 517
		kfree_skb(skb);
		return -EMSGSIZE;
	}

	/*
	 *	Setup starting values.
	 */

	hlen = iph->ihl * 4;
518
	mtu = mtu - hlen;	/* Size of data space */
519 520 521 522
#ifdef CONFIG_BRIDGE_NETFILTER
	if (skb->nf_bridge)
		mtu -= nf_bridge_mtu_reduction(skb);
#endif
H
Herbert Xu 已提交
523
	IPCB(skb)->flags |= IPSKB_FRAG_COMPLETE;
L
Linus Torvalds 已提交
524 525 526 527 528 529 530 531

	/* When frag_list is given, use it. First, check its validity:
	 * some transformers could create wrong frag_list or break existing
	 * one, it is not prohibited. In this case fall back to copying.
	 *
	 * LATER: this step can be merged to real generation of fragments,
	 * we can switch to copy when see the first bad fragment.
	 */
532
	if (skb_has_frag_list(skb)) {
533
		struct sk_buff *frag, *frag2;
L
Linus Torvalds 已提交
534 535 536 537
		int first_len = skb_pagelen(skb);

		if (first_len - hlen > mtu ||
		    ((first_len - hlen) & 7) ||
538
		    ip_is_fragment(iph) ||
L
Linus Torvalds 已提交
539 540 541
		    skb_cloned(skb))
			goto slow_path;

542
		skb_walk_frags(skb, frag) {
L
Linus Torvalds 已提交
543 544 545 546
			/* Correct geometry. */
			if (frag->len > mtu ||
			    ((frag->len & 7) && frag->next) ||
			    skb_headroom(frag) < hlen)
547
				goto slow_path_clean;
L
Linus Torvalds 已提交
548 549 550

			/* Partially cloned skb? */
			if (skb_shared(frag))
551
				goto slow_path_clean;
552 553 554 555 556 557

			BUG_ON(frag->sk);
			if (skb->sk) {
				frag->sk = skb->sk;
				frag->destructor = sock_wfree;
			}
558
			skb->truesize -= frag->truesize;
L
Linus Torvalds 已提交
559 560 561 562 563 564 565
		}

		/* Everything is OK. Generate! */

		err = 0;
		offset = 0;
		frag = skb_shinfo(skb)->frag_list;
566
		skb_frag_list_init(skb);
L
Linus Torvalds 已提交
567 568 569 570 571 572 573 574 575 576 577
		skb->data_len = first_len - skb_headlen(skb);
		skb->len = first_len;
		iph->tot_len = htons(first_len);
		iph->frag_off = htons(IP_MF);
		ip_send_check(iph);

		for (;;) {
			/* Prepare header of the next frame,
			 * before previous one went down. */
			if (frag) {
				frag->ip_summed = CHECKSUM_NONE;
578
				skb_reset_transport_header(frag);
579 580
				__skb_push(frag, hlen);
				skb_reset_network_header(frag);
581
				memcpy(skb_network_header(frag), iph, hlen);
582
				iph = ip_hdr(frag);
L
Linus Torvalds 已提交
583 584 585 586 587 588 589 590 591 592 593 594 595 596
				iph->tot_len = htons(frag->len);
				ip_copy_metadata(frag, skb);
				if (offset == 0)
					ip_options_fragment(frag);
				offset += skb->len - hlen;
				iph->frag_off = htons(offset>>3);
				if (frag->next != NULL)
					iph->frag_off |= htons(IP_MF);
				/* Ready, complete checksum */
				ip_send_check(iph);
			}

			err = output(skb);

597
			if (!err)
P
Pavel Emelyanov 已提交
598
				IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGCREATES);
L
Linus Torvalds 已提交
599 600 601 602 603 604 605 606 607
			if (err || !frag)
				break;

			skb = frag;
			frag = skb->next;
			skb->next = NULL;
		}

		if (err == 0) {
P
Pavel Emelyanov 已提交
608
			IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGOKS);
L
Linus Torvalds 已提交
609 610 611 612 613 614 615 616
			return 0;
		}

		while (frag) {
			skb = frag->next;
			kfree_skb(frag);
			frag = skb;
		}
P
Pavel Emelyanov 已提交
617
		IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
L
Linus Torvalds 已提交
618
		return err;
619 620 621 622 623 624 625 626 627

slow_path_clean:
		skb_walk_frags(skb, frag2) {
			if (frag2 == frag)
				break;
			frag2->sk = NULL;
			frag2->destructor = NULL;
			skb->truesize += frag2->truesize;
		}
L
Linus Torvalds 已提交
628 629 630
	}

slow_path:
631 632 633
	/* for offloaded checksums cleanup checksum before fragmentation */
	if ((skb->ip_summed == CHECKSUM_PARTIAL) && skb_checksum_help(skb))
		goto fail;
634
	iph = ip_hdr(skb);
635

L
Linus Torvalds 已提交
636
	left = skb->len - hlen;		/* Space per frame */
637
	ptr = hlen;		/* Where to start from */
L
Linus Torvalds 已提交
638 639

	/* for bridged IP traffic encapsulated inside f.e. a vlan header,
640 641
	 * we need to make room for the encapsulating header
	 */
642
	ll_rs = LL_RESERVED_SPACE_EXTRA(rt->dst.dev, nf_bridge_pad(skb));
643

L
Linus Torvalds 已提交
644 645 646 647 648 649 650 651 652 653 654
	/*
	 *	Fragment the datagram.
	 */

	offset = (ntohs(iph->frag_off) & IP_OFFSET) << 3;
	not_last_frag = iph->frag_off & htons(IP_MF);

	/*
	 *	Keep copying data until we run out.
	 */

S
Stephen Hemminger 已提交
655
	while (left > 0) {
L
Linus Torvalds 已提交
656 657 658 659
		len = left;
		/* IF: it doesn't fit, use 'mtu' - the data space left */
		if (len > mtu)
			len = mtu;
L
Lucas De Marchi 已提交
660
		/* IF: we are not sending up to and including the packet end
L
Linus Torvalds 已提交
661 662 663 664 665 666 667 668 669
		   then align the next start on an eight byte boundary */
		if (len < left)	{
			len &= ~7;
		}
		/*
		 *	Allocate buffer.
		 */

		if ((skb2 = alloc_skb(len+hlen+ll_rs, GFP_ATOMIC)) == NULL) {
670
			NETDEBUG(KERN_INFO "IP: frag: no memory for new fragment!\n");
L
Linus Torvalds 已提交
671 672 673 674 675 676 677 678 679 680 681
			err = -ENOMEM;
			goto fail;
		}

		/*
		 *	Set up data on packet
		 */

		ip_copy_metadata(skb2, skb);
		skb_reserve(skb2, ll_rs);
		skb_put(skb2, len + hlen);
682
		skb_reset_network_header(skb2);
683
		skb2->transport_header = skb2->network_header + hlen;
L
Linus Torvalds 已提交
684 685 686 687 688 689 690 691 692 693 694 695 696

		/*
		 *	Charge the memory for the fragment to any owner
		 *	it might possess
		 */

		if (skb->sk)
			skb_set_owner_w(skb2, skb->sk);

		/*
		 *	Copy the packet header into the new buffer.
		 */

697
		skb_copy_from_linear_data(skb, skb_network_header(skb2), hlen);
L
Linus Torvalds 已提交
698 699 700 701

		/*
		 *	Copy a block of the IP datagram.
		 */
702
		if (skb_copy_bits(skb, ptr, skb_transport_header(skb2), len))
L
Linus Torvalds 已提交
703 704 705 706 707 708
			BUG();
		left -= len;

		/*
		 *	Fill in the new header fields.
		 */
709
		iph = ip_hdr(skb2);
L
Linus Torvalds 已提交
710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739
		iph->frag_off = htons((offset >> 3));

		/* ANK: dirty, but effective trick. Upgrade options only if
		 * the segment to be fragmented was THE FIRST (otherwise,
		 * options are already fixed) and make it ONCE
		 * on the initial skb, so that all the following fragments
		 * will inherit fixed options.
		 */
		if (offset == 0)
			ip_options_fragment(skb);

		/*
		 *	Added AC : If we are fragmenting a fragment that's not the
		 *		   last fragment then keep MF on each bit
		 */
		if (left > 0 || not_last_frag)
			iph->frag_off |= htons(IP_MF);
		ptr += len;
		offset += len;

		/*
		 *	Put this fragment into the sending queue.
		 */
		iph->tot_len = htons(len + hlen);

		ip_send_check(iph);

		err = output(skb2);
		if (err)
			goto fail;
740

P
Pavel Emelyanov 已提交
741
		IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGCREATES);
L
Linus Torvalds 已提交
742
	}
743
	consume_skb(skb);
P
Pavel Emelyanov 已提交
744
	IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGOKS);
L
Linus Torvalds 已提交
745 746 747
	return err;

fail:
748
	kfree_skb(skb);
P
Pavel Emelyanov 已提交
749
	IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
L
Linus Torvalds 已提交
750 751
	return err;
}
752 753
EXPORT_SYMBOL(ip_fragment);

L
Linus Torvalds 已提交
754 755 756 757 758
int
ip_generic_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
{
	struct iovec *iov = from;

759
	if (skb->ip_summed == CHECKSUM_PARTIAL) {
L
Linus Torvalds 已提交
760 761 762
		if (memcpy_fromiovecend(to, iov, offset, len) < 0)
			return -EFAULT;
	} else {
763
		__wsum csum = 0;
L
Linus Torvalds 已提交
764 765 766 767 768 769
		if (csum_partial_copy_fromiovecend(to, iov, offset, len, &csum) < 0)
			return -EFAULT;
		skb->csum = csum_block_add(skb->csum, csum, odd);
	}
	return 0;
}
E
Eric Dumazet 已提交
770
EXPORT_SYMBOL(ip_generic_getfrag);
L
Linus Torvalds 已提交
771

772
static inline __wsum
L
Linus Torvalds 已提交
773 774 775
csum_page(struct page *page, int offset, int copy)
{
	char *kaddr;
776
	__wsum csum;
L
Linus Torvalds 已提交
777 778 779 780 781 782
	kaddr = kmap(page);
	csum = csum_partial(kaddr + offset, copy, 0);
	kunmap(page);
	return csum;
}

A
Adrian Bunk 已提交
783
static inline int ip_ufo_append_data(struct sock *sk,
784
			struct sk_buff_head *queue,
785 786 787
			int getfrag(void *from, char *to, int offset, int len,
			       int odd, struct sk_buff *skb),
			void *from, int length, int hh_len, int fragheaderlen,
788
			int transhdrlen, int maxfraglen, unsigned int flags)
789 790 791 792 793 794 795 796
{
	struct sk_buff *skb;
	int err;

	/* There is support for UDP fragmentation offload by network
	 * device, so create one single skb packet containing complete
	 * udp datagram
	 */
797
	if ((skb = skb_peek_tail(queue)) == NULL) {
798 799 800 801 802 803 804 805 806 807 808
		skb = sock_alloc_send_skb(sk,
			hh_len + fragheaderlen + transhdrlen + 20,
			(flags & MSG_DONTWAIT), &err);

		if (skb == NULL)
			return err;

		/* reserve space for Hardware header */
		skb_reserve(skb, hh_len);

		/* create space for UDP/IP header */
809
		skb_put(skb, fragheaderlen + transhdrlen);
810 811

		/* initialize network header pointer */
812
		skb_reset_network_header(skb);
813 814

		/* initialize protocol header pointer */
815
		skb->transport_header = skb->network_header + fragheaderlen;
816 817 818

		skb->csum = 0;

819

820
		__skb_queue_tail(queue, skb);
821 822
	} else if (skb_is_gso(skb)) {
		goto append;
823
	}
824

825 826 827 828 829 830
	skb->ip_summed = CHECKSUM_PARTIAL;
	/* specify the length of each IP datagram fragment */
	skb_shinfo(skb)->gso_size = maxfraglen - fragheaderlen;
	skb_shinfo(skb)->gso_type = SKB_GSO_UDP;

append:
831 832
	return skb_append_datato_frags(sk, skb, getfrag, from,
				       (length - transhdrlen));
833 834
}

835 836 837
static int __ip_append_data(struct sock *sk,
			    struct flowi4 *fl4,
			    struct sk_buff_head *queue,
838
			    struct inet_cork *cork,
839
			    struct page_frag *pfrag,
840 841 842 843
			    int getfrag(void *from, char *to, int offset,
					int len, int odd, struct sk_buff *skb),
			    void *from, int length, int transhdrlen,
			    unsigned int flags)
L
Linus Torvalds 已提交
844 845 846 847
{
	struct inet_sock *inet = inet_sk(sk);
	struct sk_buff *skb;

848
	struct ip_options *opt = cork->opt;
L
Linus Torvalds 已提交
849 850 851 852 853 854
	int hh_len;
	int exthdrlen;
	int mtu;
	int copy;
	int err;
	int offset = 0;
855
	unsigned int maxfraglen, fragheaderlen, maxnonfragsize;
L
Linus Torvalds 已提交
856
	int csummode = CHECKSUM_NONE;
857
	struct rtable *rt = (struct rtable *)cork->dst;
L
Linus Torvalds 已提交
858

859 860 861
	skb = skb_peek_tail(queue);

	exthdrlen = !skb ? rt->dst.header_len : 0;
862
	mtu = cork->fragsize;
L
Linus Torvalds 已提交
863

864
	hh_len = LL_RESERVED_SPACE(rt->dst.dev);
L
Linus Torvalds 已提交
865 866 867

	fragheaderlen = sizeof(struct iphdr) + (opt ? opt->optlen : 0);
	maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen;
W
WANG Cong 已提交
868
	maxnonfragsize = ip_sk_ignore_df(sk) ? 0xFFFF : mtu;
L
Linus Torvalds 已提交
869

870
	if (cork->length + length > maxnonfragsize - fragheaderlen) {
871
		ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
872
			       mtu - (opt ? opt->optlen : 0));
L
Linus Torvalds 已提交
873 874 875 876 877 878 879 880 881
		return -EMSGSIZE;
	}

	/*
	 * transhdrlen > 0 means that this is the first fragment and we wish
	 * it won't be fragmented in the future.
	 */
	if (transhdrlen &&
	    length + fragheaderlen <= mtu &&
882
	    rt->dst.dev->features & NETIF_F_V4_CSUM &&
L
Linus Torvalds 已提交
883
	    !exthdrlen)
884
		csummode = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
885

886
	cork->length += length;
887
	if (((length > mtu) || (skb && skb_is_gso(skb))) &&
888
	    (sk->sk_protocol == IPPROTO_UDP) &&
889
	    (rt->dst.dev->features & NETIF_F_UFO) && !rt->dst.header_len) {
890 891
		err = ip_ufo_append_data(sk, queue, getfrag, from, length,
					 hh_len, fragheaderlen, transhdrlen,
892
					 maxfraglen, flags);
893
		if (err)
894 895 896
			goto error;
		return 0;
	}
L
Linus Torvalds 已提交
897 898 899 900 901 902 903 904

	/* So, what's going on in the loop below?
	 *
	 * We use calculated fragment length to generate chained skb,
	 * each of segments is IP fragment ready for sending to network after
	 * adding appropriate IP header.
	 */

905
	if (!skb)
L
Linus Torvalds 已提交
906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935
		goto alloc_new_skb;

	while (length > 0) {
		/* Check if the remaining data fits into current packet. */
		copy = mtu - skb->len;
		if (copy < length)
			copy = maxfraglen - skb->len;
		if (copy <= 0) {
			char *data;
			unsigned int datalen;
			unsigned int fraglen;
			unsigned int fraggap;
			unsigned int alloclen;
			struct sk_buff *skb_prev;
alloc_new_skb:
			skb_prev = skb;
			if (skb_prev)
				fraggap = skb_prev->len - maxfraglen;
			else
				fraggap = 0;

			/*
			 * If remaining data exceeds the mtu,
			 * we know we need more fragment(s).
			 */
			datalen = length + fraggap;
			if (datalen > mtu - fragheaderlen)
				datalen = maxfraglen - fragheaderlen;
			fraglen = datalen + fragheaderlen;

936
			if ((flags & MSG_MORE) &&
937
			    !(rt->dst.dev->features&NETIF_F_SG))
L
Linus Torvalds 已提交
938 939
				alloclen = mtu;
			else
940
				alloclen = fraglen;
L
Linus Torvalds 已提交
941

942 943
			alloclen += exthdrlen;

L
Linus Torvalds 已提交
944 945 946 947 948
			/* The last fragment gets additional space at tail.
			 * Note, with MSG_MORE we overallocate on fragments,
			 * because we have no idea what fragment will be
			 * the last.
			 */
949
			if (datalen == length + fraggap)
950
				alloclen += rt->dst.trailer_len;
951

L
Linus Torvalds 已提交
952
			if (transhdrlen) {
953
				skb = sock_alloc_send_skb(sk,
L
Linus Torvalds 已提交
954 955 956 957 958 959
						alloclen + hh_len + 15,
						(flags & MSG_DONTWAIT), &err);
			} else {
				skb = NULL;
				if (atomic_read(&sk->sk_wmem_alloc) <=
				    2 * sk->sk_sndbuf)
960
					skb = sock_wmalloc(sk,
L
Linus Torvalds 已提交
961 962 963 964
							   alloclen + hh_len + 15, 1,
							   sk->sk_allocation);
				if (unlikely(skb == NULL))
					err = -ENOBUFS;
965 966 967
				else
					/* only the initial fragment is
					   time stamped */
968
					cork->tx_flags = 0;
L
Linus Torvalds 已提交
969 970 971 972 973 974 975 976 977 978
			}
			if (skb == NULL)
				goto error;

			/*
			 *	Fill in the control structures
			 */
			skb->ip_summed = csummode;
			skb->csum = 0;
			skb_reserve(skb, hh_len);
979
			skb_shinfo(skb)->tx_flags = cork->tx_flags;
L
Linus Torvalds 已提交
980 981 982 983

			/*
			 *	Find where to start putting bytes.
			 */
984
			data = skb_put(skb, fraglen + exthdrlen);
985
			skb_set_network_header(skb, exthdrlen);
986 987
			skb->transport_header = (skb->network_header +
						 fragheaderlen);
988
			data += fragheaderlen + exthdrlen;
L
Linus Torvalds 已提交
989 990 991 992 993 994 995 996

			if (fraggap) {
				skb->csum = skb_copy_and_csum_bits(
					skb_prev, maxfraglen,
					data + transhdrlen, fraggap, 0);
				skb_prev->csum = csum_sub(skb_prev->csum,
							  skb->csum);
				data += fraggap;
997
				pskb_trim_unique(skb_prev, maxfraglen);
L
Linus Torvalds 已提交
998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
			}

			copy = datalen - transhdrlen - fraggap;
			if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
				err = -EFAULT;
				kfree_skb(skb);
				goto error;
			}

			offset += copy;
			length -= datalen - fraggap;
			transhdrlen = 0;
			exthdrlen = 0;
			csummode = CHECKSUM_NONE;

			/*
			 * Put the packet on the pending queue.
			 */
1016
			__skb_queue_tail(queue, skb);
L
Linus Torvalds 已提交
1017 1018 1019 1020 1021 1022
			continue;
		}

		if (copy > length)
			copy = length;

1023
		if (!(rt->dst.dev->features&NETIF_F_SG)) {
L
Linus Torvalds 已提交
1024 1025 1026
			unsigned int off;

			off = skb->len;
1027
			if (getfrag(from, skb_put(skb, copy),
L
Linus Torvalds 已提交
1028 1029 1030 1031 1032 1033 1034 1035
					offset, copy, off, skb) < 0) {
				__skb_trim(skb, off);
				err = -EFAULT;
				goto error;
			}
		} else {
			int i = skb_shinfo(skb)->nr_frags;

1036 1037
			err = -ENOMEM;
			if (!sk_page_frag_refill(sk, pfrag))
L
Linus Torvalds 已提交
1038
				goto error;
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049

			if (!skb_can_coalesce(skb, i, pfrag->page,
					      pfrag->offset)) {
				err = -EMSGSIZE;
				if (i == MAX_SKB_FRAGS)
					goto error;

				__skb_fill_page_desc(skb, i, pfrag->page,
						     pfrag->offset, 0);
				skb_shinfo(skb)->nr_frags = ++i;
				get_page(pfrag->page);
L
Linus Torvalds 已提交
1050
			}
1051 1052 1053 1054 1055 1056 1057 1058
			copy = min_t(int, copy, pfrag->size - pfrag->offset);
			if (getfrag(from,
				    page_address(pfrag->page) + pfrag->offset,
				    offset, copy, skb->len, skb) < 0)
				goto error_efault;

			pfrag->offset += copy;
			skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
L
Linus Torvalds 已提交
1059 1060
			skb->len += copy;
			skb->data_len += copy;
1061 1062
			skb->truesize += copy;
			atomic_add(copy, &sk->sk_wmem_alloc);
L
Linus Torvalds 已提交
1063 1064 1065 1066 1067 1068 1069
		}
		offset += copy;
		length -= copy;
	}

	return 0;

1070 1071
error_efault:
	err = -EFAULT;
L
Linus Torvalds 已提交
1072
error:
1073
	cork->length -= length;
P
Pavel Emelyanov 已提交
1074
	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTDISCARDS);
1075
	return err;
L
Linus Torvalds 已提交
1076 1077
}

1078 1079 1080
static int ip_setup_cork(struct sock *sk, struct inet_cork *cork,
			 struct ipcm_cookie *ipc, struct rtable **rtp)
{
1081
	struct ip_options_rcu *opt;
1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
	struct rtable *rt;

	/*
	 * setup for corking.
	 */
	opt = ipc->opt;
	if (opt) {
		if (cork->opt == NULL) {
			cork->opt = kmalloc(sizeof(struct ip_options) + 40,
					    sk->sk_allocation);
			if (unlikely(cork->opt == NULL))
				return -ENOBUFS;
		}
1095
		memcpy(cork->opt, &opt->opt, sizeof(struct ip_options) + opt->opt.optlen);
1096 1097 1098 1099 1100 1101 1102 1103 1104 1105
		cork->flags |= IPCORK_OPT;
		cork->addr = ipc->addr;
	}
	rt = *rtp;
	if (unlikely(!rt))
		return -EFAULT;
	/*
	 * We steal reference to this route, caller should not release it
	 */
	*rtp = NULL;
1106 1107
	cork->fragsize = ip_sk_use_pmtu(sk) ?
			 dst_mtu(&rt->dst) : rt->dst.dev->mtu;
1108 1109
	cork->dst = &rt->dst;
	cork->length = 0;
1110 1111 1112
	cork->ttl = ipc->ttl;
	cork->tos = ipc->tos;
	cork->priority = ipc->priority;
1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128
	cork->tx_flags = ipc->tx_flags;

	return 0;
}

/*
 *	ip_append_data() and ip_append_page() can make one large IP datagram
 *	from many pieces of data. Each pieces will be holded on the socket
 *	until ip_push_pending_frames() is called. Each piece can be a page
 *	or non-page data.
 *
 *	Not only UDP, other transport protocols - e.g. raw sockets - can use
 *	this interface potentially.
 *
 *	LATER: length must be adjusted by pad at tail, when it is required.
 */
1129
int ip_append_data(struct sock *sk, struct flowi4 *fl4,
1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
		   int getfrag(void *from, char *to, int offset, int len,
			       int odd, struct sk_buff *skb),
		   void *from, int length, int transhdrlen,
		   struct ipcm_cookie *ipc, struct rtable **rtp,
		   unsigned int flags)
{
	struct inet_sock *inet = inet_sk(sk);
	int err;

	if (flags&MSG_PROBE)
		return 0;

	if (skb_queue_empty(&sk->sk_write_queue)) {
1143
		err = ip_setup_cork(sk, &inet->cork.base, ipc, rtp);
1144 1145 1146 1147 1148 1149
		if (err)
			return err;
	} else {
		transhdrlen = 0;
	}

1150 1151
	return __ip_append_data(sk, fl4, &sk->sk_write_queue, &inet->cork.base,
				sk_page_frag(sk), getfrag,
1152 1153 1154
				from, length, transhdrlen, flags);
}

1155
ssize_t	ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
L
Linus Torvalds 已提交
1156 1157 1158 1159 1160 1161
		       int offset, size_t size, int flags)
{
	struct inet_sock *inet = inet_sk(sk);
	struct sk_buff *skb;
	struct rtable *rt;
	struct ip_options *opt = NULL;
1162
	struct inet_cork *cork;
L
Linus Torvalds 已提交
1163 1164 1165 1166
	int hh_len;
	int mtu;
	int len;
	int err;
1167
	unsigned int maxfraglen, fragheaderlen, fraggap, maxnonfragsize;
L
Linus Torvalds 已提交
1168 1169 1170 1171 1172 1173 1174 1175 1176 1177

	if (inet->hdrincl)
		return -EPERM;

	if (flags&MSG_PROBE)
		return 0;

	if (skb_queue_empty(&sk->sk_write_queue))
		return -EINVAL;

1178 1179 1180 1181
	cork = &inet->cork.base;
	rt = (struct rtable *)cork->dst;
	if (cork->flags & IPCORK_OPT)
		opt = cork->opt;
L
Linus Torvalds 已提交
1182

1183
	if (!(rt->dst.dev->features&NETIF_F_SG))
L
Linus Torvalds 已提交
1184 1185
		return -EOPNOTSUPP;

1186
	hh_len = LL_RESERVED_SPACE(rt->dst.dev);
1187
	mtu = cork->fragsize;
L
Linus Torvalds 已提交
1188 1189 1190

	fragheaderlen = sizeof(struct iphdr) + (opt ? opt->optlen : 0);
	maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen;
W
WANG Cong 已提交
1191
	maxnonfragsize = ip_sk_ignore_df(sk) ? 0xFFFF : mtu;
L
Linus Torvalds 已提交
1192

1193
	if (cork->length + size > maxnonfragsize - fragheaderlen) {
1194 1195
		ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
			       mtu - (opt ? opt->optlen : 0));
L
Linus Torvalds 已提交
1196 1197 1198 1199 1200 1201
		return -EMSGSIZE;
	}

	if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL)
		return -EINVAL;

1202
	cork->length += size;
1203 1204
	if ((size + skb->len > mtu) &&
	    (sk->sk_protocol == IPPROTO_UDP) &&
1205
	    (rt->dst.dev->features & NETIF_F_UFO)) {
1206
		skb_shinfo(skb)->gso_size = mtu - fragheaderlen;
H
Herbert Xu 已提交
1207
		skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
1208
	}
1209

L
Linus Torvalds 已提交
1210 1211 1212 1213

	while (size > 0) {
		int i;

H
Herbert Xu 已提交
1214
		if (skb_is_gso(skb))
1215 1216 1217 1218 1219 1220 1221 1222
			len = size;
		else {

			/* Check if the remaining data fits into current packet. */
			len = mtu - skb->len;
			if (len < size)
				len = maxfraglen - skb->len;
		}
L
Linus Torvalds 已提交
1223 1224 1225 1226 1227
		if (len <= 0) {
			struct sk_buff *skb_prev;
			int alloclen;

			skb_prev = skb;
1228
			fraggap = skb_prev->len - maxfraglen;
L
Linus Torvalds 已提交
1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246

			alloclen = fragheaderlen + hh_len + fraggap + 15;
			skb = sock_wmalloc(sk, alloclen, 1, sk->sk_allocation);
			if (unlikely(!skb)) {
				err = -ENOBUFS;
				goto error;
			}

			/*
			 *	Fill in the control structures
			 */
			skb->ip_summed = CHECKSUM_NONE;
			skb->csum = 0;
			skb_reserve(skb, hh_len);

			/*
			 *	Find where to start putting bytes.
			 */
1247
			skb_put(skb, fragheaderlen + fraggap);
1248
			skb_reset_network_header(skb);
1249 1250
			skb->transport_header = (skb->network_header +
						 fragheaderlen);
L
Linus Torvalds 已提交
1251
			if (fraggap) {
1252 1253
				skb->csum = skb_copy_and_csum_bits(skb_prev,
								   maxfraglen,
1254
						    skb_transport_header(skb),
1255
								   fraggap, 0);
L
Linus Torvalds 已提交
1256 1257
				skb_prev->csum = csum_sub(skb_prev->csum,
							  skb->csum);
1258
				pskb_trim_unique(skb_prev, maxfraglen);
L
Linus Torvalds 已提交
1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271
			}

			/*
			 * Put the packet on the pending queue.
			 */
			__skb_queue_tail(&sk->sk_write_queue, skb);
			continue;
		}

		i = skb_shinfo(skb)->nr_frags;
		if (len > size)
			len = size;
		if (skb_can_coalesce(skb, i, page, offset)) {
E
Eric Dumazet 已提交
1272
			skb_frag_size_add(&skb_shinfo(skb)->frags[i-1], len);
L
Linus Torvalds 已提交
1273 1274 1275 1276 1277 1278 1279 1280 1281
		} else if (i < MAX_SKB_FRAGS) {
			get_page(page);
			skb_fill_page_desc(skb, i, page, offset, len);
		} else {
			err = -EMSGSIZE;
			goto error;
		}

		if (skb->ip_summed == CHECKSUM_NONE) {
1282
			__wsum csum;
L
Linus Torvalds 已提交
1283 1284 1285 1286 1287 1288
			csum = csum_page(page, offset, len);
			skb->csum = csum_block_add(skb->csum, csum, skb->len);
		}

		skb->len += len;
		skb->data_len += len;
1289 1290
		skb->truesize += len;
		atomic_add(len, &sk->sk_wmem_alloc);
L
Linus Torvalds 已提交
1291 1292 1293 1294 1295 1296
		offset += len;
		size -= len;
	}
	return 0;

error:
1297
	cork->length -= size;
P
Pavel Emelyanov 已提交
1298
	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTDISCARDS);
L
Linus Torvalds 已提交
1299 1300 1301
	return err;
}

1302
static void ip_cork_release(struct inet_cork *cork)
1303
{
1304 1305 1306 1307 1308
	cork->flags &= ~IPCORK_OPT;
	kfree(cork->opt);
	cork->opt = NULL;
	dst_release(cork->dst);
	cork->dst = NULL;
1309 1310
}

L
Linus Torvalds 已提交
1311 1312 1313 1314
/*
 *	Combined all pending IP fragments on the socket as one IP datagram
 *	and push them out.
 */
1315
struct sk_buff *__ip_make_skb(struct sock *sk,
1316
			      struct flowi4 *fl4,
1317 1318
			      struct sk_buff_head *queue,
			      struct inet_cork *cork)
L
Linus Torvalds 已提交
1319 1320 1321 1322
{
	struct sk_buff *skb, *tmp_skb;
	struct sk_buff **tail_skb;
	struct inet_sock *inet = inet_sk(sk);
1323
	struct net *net = sock_net(sk);
L
Linus Torvalds 已提交
1324
	struct ip_options *opt = NULL;
1325
	struct rtable *rt = (struct rtable *)cork->dst;
L
Linus Torvalds 已提交
1326
	struct iphdr *iph;
1327
	__be16 df = 0;
L
Linus Torvalds 已提交
1328 1329
	__u8 ttl;

1330
	if ((skb = __skb_dequeue(queue)) == NULL)
L
Linus Torvalds 已提交
1331 1332 1333 1334
		goto out;
	tail_skb = &(skb_shinfo(skb)->frag_list);

	/* move skb->data to ip header from ext header */
1335
	if (skb->data < skb_network_header(skb))
1336
		__skb_pull(skb, skb_network_offset(skb));
1337
	while ((tmp_skb = __skb_dequeue(queue)) != NULL) {
1338
		__skb_pull(tmp_skb, skb_network_header_len(skb));
L
Linus Torvalds 已提交
1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351
		*tail_skb = tmp_skb;
		tail_skb = &(tmp_skb->next);
		skb->len += tmp_skb->len;
		skb->data_len += tmp_skb->len;
		skb->truesize += tmp_skb->truesize;
		tmp_skb->destructor = NULL;
		tmp_skb->sk = NULL;
	}

	/* Unless user demanded real pmtu discovery (IP_PMTUDISC_DO), we allow
	 * to fragment the frame generated here. No matter, what transforms
	 * how transforms change size of the packet, it will come out.
	 */
W
WANG Cong 已提交
1352
	skb->ignore_df = ip_sk_ignore_df(sk);
L
Linus Torvalds 已提交
1353 1354

	/* DF bit is set when we want to see DF on outgoing frames.
W
WANG Cong 已提交
1355
	 * If ignore_df is set too, we still allow to fragment this frame
L
Linus Torvalds 已提交
1356
	 * locally. */
1357 1358
	if (inet->pmtudisc == IP_PMTUDISC_DO ||
	    inet->pmtudisc == IP_PMTUDISC_PROBE ||
1359 1360
	    (skb->len <= dst_mtu(&rt->dst) &&
	     ip_dont_fragment(sk, &rt->dst)))
L
Linus Torvalds 已提交
1361 1362
		df = htons(IP_DF);

1363 1364
	if (cork->flags & IPCORK_OPT)
		opt = cork->opt;
L
Linus Torvalds 已提交
1365

1366 1367 1368
	if (cork->ttl != 0)
		ttl = cork->ttl;
	else if (rt->rt_type == RTN_MULTICAST)
L
Linus Torvalds 已提交
1369 1370
		ttl = inet->mc_ttl;
	else
1371
		ttl = ip_select_ttl(inet, &rt->dst);
L
Linus Torvalds 已提交
1372

1373
	iph = ip_hdr(skb);
L
Linus Torvalds 已提交
1374 1375
	iph->version = 4;
	iph->ihl = 5;
1376
	iph->tos = (cork->tos != -1) ? cork->tos : inet->tos;
L
Linus Torvalds 已提交
1377 1378 1379
	iph->frag_off = df;
	iph->ttl = ttl;
	iph->protocol = sk->sk_protocol;
1380
	ip_copy_addrs(iph, fl4);
E
Eric Dumazet 已提交
1381
	ip_select_ident(skb, sk);
L
Linus Torvalds 已提交
1382

1383 1384 1385 1386 1387
	if (opt) {
		iph->ihl += opt->optlen>>2;
		ip_options_build(skb, opt, cork->addr, rt, 0);
	}

1388
	skb->priority = (cork->tos != -1) ? cork->priority: sk->sk_priority;
1389
	skb->mark = sk->sk_mark;
1390 1391 1392 1393
	/*
	 * Steal rt from cork.dst to avoid a pair of atomic_inc/atomic_dec
	 * on dst refcount
	 */
1394
	cork->dst = NULL;
1395
	skb_dst_set(skb, &rt->dst);
L
Linus Torvalds 已提交
1396

1397
	if (iph->protocol == IPPROTO_ICMP)
1398
		icmp_out_count(net, ((struct icmphdr *)
1399 1400
			skb_transport_header(skb))->type);

1401 1402 1403 1404 1405
	ip_cork_release(cork);
out:
	return skb;
}

E
Eric Dumazet 已提交
1406
int ip_send_skb(struct net *net, struct sk_buff *skb)
1407 1408 1409
{
	int err;

H
Herbert Xu 已提交
1410
	err = ip_local_out(skb);
L
Linus Torvalds 已提交
1411 1412
	if (err) {
		if (err > 0)
E
Eric Dumazet 已提交
1413
			err = net_xmit_errno(err);
L
Linus Torvalds 已提交
1414
		if (err)
1415
			IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
L
Linus Torvalds 已提交
1416 1417 1418 1419 1420
	}

	return err;
}

1421
int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4)
1422
{
1423 1424
	struct sk_buff *skb;

1425
	skb = ip_finish_skb(sk, fl4);
1426 1427 1428 1429
	if (!skb)
		return 0;

	/* Netfilter gets whole the not fragmented skb. */
E
Eric Dumazet 已提交
1430
	return ip_send_skb(sock_net(sk), skb);
1431 1432
}

L
Linus Torvalds 已提交
1433 1434 1435
/*
 *	Throw away all pending data on the socket.
 */
1436 1437 1438
static void __ip_flush_pending_frames(struct sock *sk,
				      struct sk_buff_head *queue,
				      struct inet_cork *cork)
L
Linus Torvalds 已提交
1439 1440 1441
{
	struct sk_buff *skb;

1442
	while ((skb = __skb_dequeue_tail(queue)) != NULL)
L
Linus Torvalds 已提交
1443 1444
		kfree_skb(skb);

1445 1446 1447 1448 1449
	ip_cork_release(cork);
}

void ip_flush_pending_frames(struct sock *sk)
{
1450
	__ip_flush_pending_frames(sk, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
L
Linus Torvalds 已提交
1451 1452
}

1453
struct sk_buff *ip_make_skb(struct sock *sk,
1454
			    struct flowi4 *fl4,
1455 1456 1457 1458 1459 1460
			    int getfrag(void *from, char *to, int offset,
					int len, int odd, struct sk_buff *skb),
			    void *from, int length, int transhdrlen,
			    struct ipcm_cookie *ipc, struct rtable **rtp,
			    unsigned int flags)
{
1461
	struct inet_cork cork;
1462 1463 1464 1465 1466 1467 1468 1469
	struct sk_buff_head queue;
	int err;

	if (flags & MSG_PROBE)
		return NULL;

	__skb_queue_head_init(&queue);

1470 1471
	cork.flags = 0;
	cork.addr = 0;
1472
	cork.opt = NULL;
1473 1474 1475 1476
	err = ip_setup_cork(sk, &cork, ipc, rtp);
	if (err)
		return ERR_PTR(err);

1477 1478
	err = __ip_append_data(sk, fl4, &queue, &cork,
			       &current->task_frag, getfrag,
1479 1480 1481 1482 1483 1484
			       from, length, transhdrlen, flags);
	if (err) {
		__ip_flush_pending_frames(sk, &queue, &cork);
		return ERR_PTR(err);
	}

1485
	return __ip_make_skb(sk, fl4, &queue, &cork);
1486
}
L
Linus Torvalds 已提交
1487 1488 1489 1490

/*
 *	Fetch data from kernel space and fill in checksum if needed.
 */
1491
static int ip_reply_glue_bits(void *dptr, char *to, int offset,
L
Linus Torvalds 已提交
1492 1493
			      int len, int odd, struct sk_buff *skb)
{
1494
	__wsum csum;
L
Linus Torvalds 已提交
1495 1496 1497

	csum = csum_partial_copy_nocheck(dptr+offset, to, len, 0);
	skb->csum = csum_block_add(skb->csum, csum, odd);
1498
	return 0;
L
Linus Torvalds 已提交
1499 1500
}

1501
/*
L
Linus Torvalds 已提交
1502
 *	Generic function to send a packet as reply to another packet.
1503
 *	Used to send some TCP resets/acks so far.
L
Linus Torvalds 已提交
1504
 *
1505
 *	Use a fake percpu inet socket to avoid false sharing and contention.
L
Linus Torvalds 已提交
1506
 */
1507 1508 1509 1510 1511 1512 1513 1514 1515
static DEFINE_PER_CPU(struct inet_sock, unicast_sock) = {
	.sk = {
		.__sk_common = {
			.skc_refcnt = ATOMIC_INIT(1),
		},
		.sk_wmem_alloc	= ATOMIC_INIT(1),
		.sk_allocation	= GFP_ATOMIC,
		.sk_flags	= (1UL << SOCK_USE_WRITE_QUEUE),
	},
1516 1517
	.pmtudisc	= IP_PMTUDISC_WANT,
	.uc_ttl		= -1,
1518 1519 1520
};

void ip_send_unicast_reply(struct net *net, struct sk_buff *skb, __be32 daddr,
1521 1522
			   __be32 saddr, const struct ip_reply_arg *arg,
			   unsigned int len)
L
Linus Torvalds 已提交
1523
{
1524
	struct ip_options_data replyopts;
L
Linus Torvalds 已提交
1525
	struct ipcm_cookie ipc;
1526
	struct flowi4 fl4;
E
Eric Dumazet 已提交
1527
	struct rtable *rt = skb_rtable(skb);
1528 1529 1530
	struct sk_buff *nskb;
	struct sock *sk;
	struct inet_sock *inet;
L
Linus Torvalds 已提交
1531

1532
	if (ip_options_echo(&replyopts.opt.opt, skb))
L
Linus Torvalds 已提交
1533 1534
		return;

1535
	ipc.addr = daddr;
L
Linus Torvalds 已提交
1536
	ipc.opt = NULL;
1537
	ipc.tx_flags = 0;
1538 1539
	ipc.ttl = 0;
	ipc.tos = -1;
L
Linus Torvalds 已提交
1540

1541
	if (replyopts.opt.opt.optlen) {
L
Linus Torvalds 已提交
1542 1543
		ipc.opt = &replyopts.opt;

1544 1545
		if (replyopts.opt.opt.srr)
			daddr = replyopts.opt.opt.faddr;
L
Linus Torvalds 已提交
1546 1547
	}

1548 1549
	flowi4_init_output(&fl4, arg->bound_dev_if,
			   IP4_REPLY_MARK(net, skb->mark),
1550
			   RT_TOS(arg->tos),
1551
			   RT_SCOPE_UNIVERSE, ip_hdr(skb)->protocol,
1552
			   ip_reply_arg_flowi_flags(arg),
1553
			   daddr, saddr,
1554 1555
			   tcp_hdr(skb)->source, tcp_hdr(skb)->dest);
	security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
1556
	rt = ip_route_output_key(net, &fl4);
1557 1558
	if (IS_ERR(rt))
		return;
L
Linus Torvalds 已提交
1559

1560
	inet = &get_cpu_var(unicast_sock);
L
Linus Torvalds 已提交
1561

1562
	inet->tos = arg->tos;
1563
	sk = &inet->sk;
L
Linus Torvalds 已提交
1564
	sk->sk_priority = skb->priority;
1565
	sk->sk_protocol = ip_hdr(skb)->protocol;
1566
	sk->sk_bound_dev_if = arg->bound_dev_if;
1567 1568 1569
	sock_net_set(sk, net);
	__skb_queue_head_init(&sk->sk_write_queue);
	sk->sk_sndbuf = sysctl_wmem_default;
1570
	ip_append_data(sk, &fl4, ip_reply_glue_bits, arg->iov->iov_base, len, 0,
1571
		       &ipc, &rt, MSG_DONTWAIT);
1572 1573
	nskb = skb_peek(&sk->sk_write_queue);
	if (nskb) {
L
Linus Torvalds 已提交
1574
		if (arg->csumoffset >= 0)
1575 1576
			*((__sum16 *)skb_transport_header(nskb) +
			  arg->csumoffset) = csum_fold(csum_add(nskb->csum,
1577
								arg->csum));
1578
		nskb->ip_summed = CHECKSUM_NONE;
1579
		skb_orphan(nskb);
1580
		skb_set_queue_mapping(nskb, skb_get_queue_mapping(skb));
1581
		ip_push_pending_frames(sk, &fl4);
L
Linus Torvalds 已提交
1582 1583
	}

1584
	put_cpu_var(unicast_sock);
L
Linus Torvalds 已提交
1585 1586 1587 1588 1589 1590 1591 1592 1593

	ip_rt_put(rt);
}

void __init ip_init(void)
{
	ip_rt_init();
	inet_initpeers();

1594 1595
#if defined(CONFIG_IP_MULTICAST)
	igmp_mc_init();
L
Linus Torvalds 已提交
1596 1597
#endif
}