ip_output.c 37.6 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 52
 *		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 <asm/system.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
53
#include <linux/highmem.h>
54
#include <linux/slab.h>
L
Linus Torvalds 已提交
55 56 57 58 59 60 61 62 63 64 65 66 67 68 69

#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>
70
#include <net/xfrm.h>
L
Linus Torvalds 已提交
71 72 73 74 75 76 77 78 79 80 81
#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>
82
#include <linux/tcp.h>
L
Linus Torvalds 已提交
83

84
int sysctl_ip_default_ttl __read_mostly = IPDEFTTL;
85
EXPORT_SYMBOL(sysctl_ip_default_ttl);
L
Linus Torvalds 已提交
86 87 88 89 90 91 92

/* Generate a checksum for an outgoing IP datagram. */
__inline__ void ip_send_check(struct iphdr *iph)
{
	iph->check = 0;
	iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
}
E
Eric Dumazet 已提交
93
EXPORT_SYMBOL(ip_send_check);
L
Linus Torvalds 已提交
94

H
Herbert Xu 已提交
95 96 97 98 99 100
int __ip_local_out(struct sk_buff *skb)
{
	struct iphdr *iph = ip_hdr(skb);

	iph->tot_len = htons(skb->len);
	ip_send_check(iph);
101 102
	return nf_hook(NFPROTO_IPV4, NF_INET_LOCAL_OUT, skb, NULL,
		       skb_dst(skb)->dev, dst_output);
H
Herbert Xu 已提交
103 104 105 106 107 108 109 110 111 112 113 114 115 116
}

int ip_local_out(struct sk_buff *skb)
{
	int err;

	err = __ip_local_out(skb);
	if (likely(err == 1))
		err = dst_output(skb);

	return err;
}
EXPORT_SYMBOL_GPL(ip_local_out);

L
Linus Torvalds 已提交
117 118 119
/* dev_loopback_xmit for use with netfilter. */
static int ip_dev_loopback_xmit(struct sk_buff *newskb)
{
120
	skb_reset_mac_header(newskb);
121
	__skb_pull(newskb, skb_network_offset(newskb));
L
Linus Torvalds 已提交
122 123
	newskb->pkt_type = PACKET_LOOPBACK;
	newskb->ip_summed = CHECKSUM_UNNECESSARY;
E
Eric Dumazet 已提交
124
	WARN_ON(!skb_dst(newskb));
125
	skb_dst_force(newskb);
E
Eric Dumazet 已提交
126
	netif_rx_ni(newskb);
L
Linus Torvalds 已提交
127 128 129 130 131 132 133 134
	return 0;
}

static inline int ip_select_ttl(struct inet_sock *inet, struct dst_entry *dst)
{
	int ttl = inet->uc_ttl;

	if (ttl < 0)
135
		ttl = ip4_dst_hoplimit(dst);
L
Linus Torvalds 已提交
136 137 138
	return ttl;
}

139
/*
L
Linus Torvalds 已提交
140 141 142 143
 *		Add an ip header to a skbuff and send it out.
 *
 */
int ip_build_and_send_pkt(struct sk_buff *skb, struct sock *sk,
144
			  __be32 saddr, __be32 daddr, struct ip_options_rcu *opt)
L
Linus Torvalds 已提交
145 146
{
	struct inet_sock *inet = inet_sk(sk);
E
Eric Dumazet 已提交
147
	struct rtable *rt = skb_rtable(skb);
L
Linus Torvalds 已提交
148 149 150
	struct iphdr *iph;

	/* Build the IP header. */
151
	skb_push(skb, sizeof(struct iphdr) + (opt ? opt->opt.optlen : 0));
152
	skb_reset_network_header(skb);
153
	iph = ip_hdr(skb);
L
Linus Torvalds 已提交
154 155 156
	iph->version  = 4;
	iph->ihl      = 5;
	iph->tos      = inet->tos;
157
	if (ip_dont_fragment(sk, &rt->dst))
L
Linus Torvalds 已提交
158 159 160
		iph->frag_off = htons(IP_DF);
	else
		iph->frag_off = 0;
161
	iph->ttl      = ip_select_ttl(inet, &rt->dst);
162 163
	iph->daddr    = (opt && opt->opt.srr ? opt->opt.faddr : daddr);
	iph->saddr    = saddr;
L
Linus Torvalds 已提交
164
	iph->protocol = sk->sk_protocol;
165
	ip_select_ident(iph, &rt->dst, sk);
L
Linus Torvalds 已提交
166

167 168 169
	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
170 171 172
	}

	skb->priority = sk->sk_priority;
173
	skb->mark = sk->sk_mark;
L
Linus Torvalds 已提交
174 175

	/* Send it out. */
H
Herbert Xu 已提交
176
	return ip_local_out(skb);
L
Linus Torvalds 已提交
177
}
178 179
EXPORT_SYMBOL_GPL(ip_build_and_send_pkt);

L
Linus Torvalds 已提交
180 181
static inline int ip_finish_output2(struct sk_buff *skb)
{
E
Eric Dumazet 已提交
182
	struct dst_entry *dst = skb_dst(skb);
183
	struct rtable *rt = (struct rtable *)dst;
L
Linus Torvalds 已提交
184
	struct net_device *dev = dst->dev;
185
	unsigned int hh_len = LL_RESERVED_SPACE(dev);
186
	struct neighbour *neigh;
L
Linus Torvalds 已提交
187

188 189 190 191
	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);
192

L
Linus Torvalds 已提交
193
	/* Be paranoid, rather than too clever. */
194
	if (unlikely(skb_headroom(skb) < hh_len && dev->header_ops)) {
L
Linus Torvalds 已提交
195 196 197 198 199 200 201 202 203 204 205 206 207
		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);
		kfree_skb(skb);
		skb = skb2;
	}

208
	rcu_read_lock();
209
	neigh = dst_get_neighbour(dst);
210 211 212 213 214 215 216
	if (neigh) {
		int res = neigh_output(neigh, skb);

		rcu_read_unlock();
		return res;
	}
	rcu_read_unlock();
217

L
Linus Torvalds 已提交
218 219 220 221 222 223
	if (net_ratelimit())
		printk(KERN_DEBUG "ip_finish_output2: No header cache and no neighbour!\n");
	kfree_skb(skb);
	return -EINVAL;
}

J
John Heffner 已提交
224 225 226 227 228
static inline int ip_skb_dst_mtu(struct sk_buff *skb)
{
	struct inet_sock *inet = skb->sk ? inet_sk(skb->sk) : NULL;

	return (inet && inet->pmtudisc == IP_PMTUDISC_PROBE) ?
E
Eric Dumazet 已提交
229
	       skb_dst(skb)->dev->mtu : dst_mtu(skb_dst(skb));
J
John Heffner 已提交
230 231
}

P
Patrick McHardy 已提交
232
static int ip_finish_output(struct sk_buff *skb)
L
Linus Torvalds 已提交
233
{
234 235
#if defined(CONFIG_NETFILTER) && defined(CONFIG_XFRM)
	/* Policy lookup after SNAT yielded a new policy */
E
Eric Dumazet 已提交
236
	if (skb_dst(skb)->xfrm != NULL) {
237 238 239
		IPCB(skb)->flags |= IPSKB_REROUTED;
		return dst_output(skb);
	}
240
#endif
J
John Heffner 已提交
241
	if (skb->len > ip_skb_dst_mtu(skb) && !skb_is_gso(skb))
242 243 244
		return ip_fragment(skb, ip_finish_output2);
	else
		return ip_finish_output2(skb);
L
Linus Torvalds 已提交
245 246 247 248 249
}

int ip_mc_output(struct sk_buff *skb)
{
	struct sock *sk = skb->sk;
E
Eric Dumazet 已提交
250
	struct rtable *rt = skb_rtable(skb);
251
	struct net_device *dev = rt->dst.dev;
L
Linus Torvalds 已提交
252 253 254 255

	/*
	 *	If the indicated interface is up and running, send the packet.
	 */
256
	IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len);
L
Linus Torvalds 已提交
257 258 259 260 261 262 263 264 265

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

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

	if (rt->rt_flags&RTCF_MULTICAST) {
266
		if (sk_mc_loop(sk)
L
Linus Torvalds 已提交
267 268 269 270 271 272 273 274 275
#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.
		 */
276 277 278
		    &&
		    ((rt->rt_flags & RTCF_LOCAL) ||
		     !(IPCB(skb)->flags & IPSKB_FORWARDED))
L
Linus Torvalds 已提交
279
#endif
280
		   ) {
L
Linus Torvalds 已提交
281 282
			struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
			if (newskb)
283 284
				NF_HOOK(NFPROTO_IPV4, NF_INET_POST_ROUTING,
					newskb, NULL, newskb->dev,
L
Linus Torvalds 已提交
285 286 287 288 289
					ip_dev_loopback_xmit);
		}

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

290
		if (ip_hdr(skb)->ttl == 0) {
L
Linus Torvalds 已提交
291 292 293 294 295 296 297 298
			kfree_skb(skb);
			return 0;
		}
	}

	if (rt->rt_flags&RTCF_BROADCAST) {
		struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
		if (newskb)
299 300
			NF_HOOK(NFPROTO_IPV4, NF_INET_POST_ROUTING, newskb,
				NULL, newskb->dev, ip_dev_loopback_xmit);
L
Linus Torvalds 已提交
301 302
	}

303 304
	return NF_HOOK_COND(NFPROTO_IPV4, NF_INET_POST_ROUTING, skb, NULL,
			    skb->dev, ip_finish_output,
305
			    !(IPCB(skb)->flags & IPSKB_REROUTED));
L
Linus Torvalds 已提交
306 307 308 309
}

int ip_output(struct sk_buff *skb)
{
E
Eric Dumazet 已提交
310
	struct net_device *dev = skb_dst(skb)->dev;
311

312
	IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len);
L
Linus Torvalds 已提交
313

314 315 316
	skb->dev = dev;
	skb->protocol = htons(ETH_P_IP);

317
	return NF_HOOK_COND(NFPROTO_IPV4, NF_INET_POST_ROUTING, skb, NULL, dev,
318
			    ip_finish_output,
319
			    !(IPCB(skb)->flags & IPSKB_REROUTED));
L
Linus Torvalds 已提交
320 321
}

322
int ip_queue_xmit(struct sk_buff *skb, struct flowi *fl)
L
Linus Torvalds 已提交
323
{
324
	struct sock *sk = skb->sk;
L
Linus Torvalds 已提交
325
	struct inet_sock *inet = inet_sk(sk);
326
	struct ip_options_rcu *inet_opt;
327
	struct flowi4 *fl4;
L
Linus Torvalds 已提交
328 329
	struct rtable *rt;
	struct iphdr *iph;
330
	int res;
L
Linus Torvalds 已提交
331 332 333 334

	/* Skip all of this if the packet is already routed,
	 * f.e. by something like SCTP.
	 */
335
	rcu_read_lock();
336
	inet_opt = rcu_dereference(inet->inet_opt);
337
	fl4 = &fl->u.ip4;
E
Eric Dumazet 已提交
338
	rt = skb_rtable(skb);
L
Linus Torvalds 已提交
339 340 341 342 343 344
	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
345
		__be32 daddr;
L
Linus Torvalds 已提交
346 347

		/* Use correct destination address if we have options. */
E
Eric Dumazet 已提交
348
		daddr = inet->inet_daddr;
349 350
		if (inet_opt && inet_opt->opt.srr)
			daddr = inet_opt->opt.faddr;
L
Linus Torvalds 已提交
351

352 353 354 355
		/* If this fails, retransmit mechanism of transport layer will
		 * keep trying until route appears or the connection times
		 * itself out.
		 */
356
		rt = ip_route_output_ports(sock_net(sk), fl4, sk,
357 358 359 360 361 362 363 364
					   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;
365
		sk_setup_caps(sk, &rt->dst);
L
Linus Torvalds 已提交
366
	}
367
	skb_dst_set_noref(skb, &rt->dst);
L
Linus Torvalds 已提交
368 369

packet_routed:
370
	if (inet_opt && inet_opt->opt.is_strictroute && fl4->daddr != rt->rt_gateway)
L
Linus Torvalds 已提交
371 372 373
		goto no_route;

	/* OK, we know where to send it, allocate and build IP header. */
374
	skb_push(skb, sizeof(struct iphdr) + (inet_opt ? inet_opt->opt.optlen : 0));
375
	skb_reset_network_header(skb);
376
	iph = ip_hdr(skb);
377
	*((__be16 *)iph) = htons((4 << 12) | (5 << 8) | (inet->tos & 0xff));
378
	if (ip_dont_fragment(sk, &rt->dst) && !skb->local_df)
L
Linus Torvalds 已提交
379 380 381
		iph->frag_off = htons(IP_DF);
	else
		iph->frag_off = 0;
382
	iph->ttl      = ip_select_ttl(inet, &rt->dst);
L
Linus Torvalds 已提交
383
	iph->protocol = sk->sk_protocol;
384 385
	iph->saddr    = fl4->saddr;
	iph->daddr    = fl4->daddr;
L
Linus Torvalds 已提交
386 387
	/* Transport layer set skb->h.foo itself. */

388 389 390
	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
391 392
	}

393
	ip_select_ident_more(iph, &rt->dst, sk,
394
			     (skb_shinfo(skb)->gso_segs ?: 1) - 1);
L
Linus Torvalds 已提交
395 396

	skb->priority = sk->sk_priority;
397
	skb->mark = sk->sk_mark;
L
Linus Torvalds 已提交
398

399 400 401
	res = ip_local_out(skb);
	rcu_read_unlock();
	return res;
L
Linus Torvalds 已提交
402 403

no_route:
404
	rcu_read_unlock();
P
Pavel Emelyanov 已提交
405
	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
L
Linus Torvalds 已提交
406 407 408
	kfree_skb(skb);
	return -EHOSTUNREACH;
}
E
Eric Dumazet 已提交
409
EXPORT_SYMBOL(ip_queue_xmit);
L
Linus Torvalds 已提交
410 411 412 413 414 415 416


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
417
	skb_dst_drop(to);
418
	skb_dst_copy(to, from);
L
Linus Torvalds 已提交
419
	to->dev = from->dev;
T
Thomas Graf 已提交
420
	to->mark = from->mark;
L
Linus Torvalds 已提交
421 422 423 424 425 426 427

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

#ifdef CONFIG_NET_SCHED
	to->tc_index = from->tc_index;
#endif
428
	nf_copy(to, from);
429 430 431 432
#if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
    defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
	to->nf_trace = from->nf_trace;
#endif
433 434
#if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
	to->ipvs_property = from->ipvs_property;
L
Linus Torvalds 已提交
435
#endif
436
	skb_copy_secmark(to, from);
L
Linus Torvalds 已提交
437 438 439 440 441 442 443 444 445
}

/*
 *	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.
 */

446
int ip_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
L
Linus Torvalds 已提交
447 448 449 450 451
{
	struct iphdr *iph;
	int ptr;
	struct net_device *dev;
	struct sk_buff *skb2;
452
	unsigned int mtu, hlen, left, len, ll_rs;
L
Linus Torvalds 已提交
453
	int offset;
454
	__be16 not_last_frag;
E
Eric Dumazet 已提交
455
	struct rtable *rt = skb_rtable(skb);
L
Linus Torvalds 已提交
456 457
	int err = 0;

458
	dev = rt->dst.dev;
L
Linus Torvalds 已提交
459 460 461 462 463

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

464
	iph = ip_hdr(skb);
L
Linus Torvalds 已提交
465 466

	if (unlikely((iph->frag_off & htons(IP_DF)) && !skb->local_df)) {
P
Pavel Emelyanov 已提交
467
		IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
L
Linus Torvalds 已提交
468
		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
J
John Heffner 已提交
469
			  htonl(ip_skb_dst_mtu(skb)));
L
Linus Torvalds 已提交
470 471 472 473 474 475 476 477 478
		kfree_skb(skb);
		return -EMSGSIZE;
	}

	/*
	 *	Setup starting values.
	 */

	hlen = iph->ihl * 4;
479
	mtu = dst_mtu(&rt->dst) - hlen;	/* Size of data space */
480 481 482 483
#ifdef CONFIG_BRIDGE_NETFILTER
	if (skb->nf_bridge)
		mtu -= nf_bridge_mtu_reduction(skb);
#endif
H
Herbert Xu 已提交
484
	IPCB(skb)->flags |= IPSKB_FRAG_COMPLETE;
L
Linus Torvalds 已提交
485 486 487 488 489 490 491 492

	/* 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.
	 */
493
	if (skb_has_frag_list(skb)) {
494
		struct sk_buff *frag, *frag2;
L
Linus Torvalds 已提交
495 496 497 498
		int first_len = skb_pagelen(skb);

		if (first_len - hlen > mtu ||
		    ((first_len - hlen) & 7) ||
499
		    ip_is_fragment(iph) ||
L
Linus Torvalds 已提交
500 501 502
		    skb_cloned(skb))
			goto slow_path;

503
		skb_walk_frags(skb, frag) {
L
Linus Torvalds 已提交
504 505 506 507
			/* Correct geometry. */
			if (frag->len > mtu ||
			    ((frag->len & 7) && frag->next) ||
			    skb_headroom(frag) < hlen)
508
				goto slow_path_clean;
L
Linus Torvalds 已提交
509 510 511

			/* Partially cloned skb? */
			if (skb_shared(frag))
512
				goto slow_path_clean;
513 514 515 516 517 518

			BUG_ON(frag->sk);
			if (skb->sk) {
				frag->sk = skb->sk;
				frag->destructor = sock_wfree;
			}
519
			skb->truesize -= frag->truesize;
L
Linus Torvalds 已提交
520 521 522 523 524 525 526
		}

		/* Everything is OK. Generate! */

		err = 0;
		offset = 0;
		frag = skb_shinfo(skb)->frag_list;
527
		skb_frag_list_init(skb);
L
Linus Torvalds 已提交
528 529 530 531 532 533 534 535 536 537 538
		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;
539
				skb_reset_transport_header(frag);
540 541
				__skb_push(frag, hlen);
				skb_reset_network_header(frag);
542
				memcpy(skb_network_header(frag), iph, hlen);
543
				iph = ip_hdr(frag);
L
Linus Torvalds 已提交
544 545 546 547 548 549 550 551 552 553 554 555 556 557
				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);

558
			if (!err)
P
Pavel Emelyanov 已提交
559
				IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGCREATES);
L
Linus Torvalds 已提交
560 561 562 563 564 565 566 567 568
			if (err || !frag)
				break;

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

		if (err == 0) {
P
Pavel Emelyanov 已提交
569
			IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGOKS);
L
Linus Torvalds 已提交
570 571 572 573 574 575 576 577
			return 0;
		}

		while (frag) {
			skb = frag->next;
			kfree_skb(frag);
			frag = skb;
		}
P
Pavel Emelyanov 已提交
578
		IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
L
Linus Torvalds 已提交
579
		return err;
580 581 582 583 584 585 586 587 588

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
589 590 591 592
	}

slow_path:
	left = skb->len - hlen;		/* Space per frame */
593
	ptr = hlen;		/* Where to start from */
L
Linus Torvalds 已提交
594 595

	/* for bridged IP traffic encapsulated inside f.e. a vlan header,
596 597
	 * we need to make room for the encapsulating header
	 */
598
	ll_rs = LL_RESERVED_SPACE_EXTRA(rt->dst.dev, nf_bridge_pad(skb));
599

L
Linus Torvalds 已提交
600 601 602 603 604 605 606 607 608 609 610
	/*
	 *	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
611
	while (left > 0) {
L
Linus Torvalds 已提交
612 613 614 615
		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
616
		/* IF: we are not sending up to and including the packet end
L
Linus Torvalds 已提交
617 618 619 620 621 622 623 624 625
		   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) {
626
			NETDEBUG(KERN_INFO "IP: frag: no memory for new fragment!\n");
L
Linus Torvalds 已提交
627 628 629 630 631 632 633 634 635 636 637
			err = -ENOMEM;
			goto fail;
		}

		/*
		 *	Set up data on packet
		 */

		ip_copy_metadata(skb2, skb);
		skb_reserve(skb2, ll_rs);
		skb_put(skb2, len + hlen);
638
		skb_reset_network_header(skb2);
639
		skb2->transport_header = skb2->network_header + hlen;
L
Linus Torvalds 已提交
640 641 642 643 644 645 646 647 648 649 650 651 652

		/*
		 *	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.
		 */

653
		skb_copy_from_linear_data(skb, skb_network_header(skb2), hlen);
L
Linus Torvalds 已提交
654 655 656 657

		/*
		 *	Copy a block of the IP datagram.
		 */
658
		if (skb_copy_bits(skb, ptr, skb_transport_header(skb2), len))
L
Linus Torvalds 已提交
659 660 661 662 663 664
			BUG();
		left -= len;

		/*
		 *	Fill in the new header fields.
		 */
665
		iph = ip_hdr(skb2);
L
Linus Torvalds 已提交
666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695
		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;
696

P
Pavel Emelyanov 已提交
697
		IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGCREATES);
L
Linus Torvalds 已提交
698 699
	}
	kfree_skb(skb);
P
Pavel Emelyanov 已提交
700
	IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGOKS);
L
Linus Torvalds 已提交
701 702 703
	return err;

fail:
704
	kfree_skb(skb);
P
Pavel Emelyanov 已提交
705
	IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
L
Linus Torvalds 已提交
706 707
	return err;
}
708 709
EXPORT_SYMBOL(ip_fragment);

L
Linus Torvalds 已提交
710 711 712 713 714
int
ip_generic_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
{
	struct iovec *iov = from;

715
	if (skb->ip_summed == CHECKSUM_PARTIAL) {
L
Linus Torvalds 已提交
716 717 718
		if (memcpy_fromiovecend(to, iov, offset, len) < 0)
			return -EFAULT;
	} else {
719
		__wsum csum = 0;
L
Linus Torvalds 已提交
720 721 722 723 724 725
		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
726
EXPORT_SYMBOL(ip_generic_getfrag);
L
Linus Torvalds 已提交
727

728
static inline __wsum
L
Linus Torvalds 已提交
729 730 731
csum_page(struct page *page, int offset, int copy)
{
	char *kaddr;
732
	__wsum csum;
L
Linus Torvalds 已提交
733 734 735 736 737 738
	kaddr = kmap(page);
	csum = csum_partial(kaddr + offset, copy, 0);
	kunmap(page);
	return csum;
}

A
Adrian Bunk 已提交
739
static inline int ip_ufo_append_data(struct sock *sk,
740
			struct sk_buff_head *queue,
741 742 743
			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,
744
			int transhdrlen, int maxfraglen, unsigned int flags)
745 746 747 748 749 750 751 752
{
	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
	 */
753
	if ((skb = skb_peek_tail(queue)) == NULL) {
754 755 756 757 758 759 760 761 762 763 764
		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 */
765
		skb_put(skb, fragheaderlen + transhdrlen);
766 767

		/* initialize network header pointer */
768
		skb_reset_network_header(skb);
769 770

		/* initialize protocol header pointer */
771
		skb->transport_header = skb->network_header + fragheaderlen;
772

773
		skb->ip_summed = CHECKSUM_PARTIAL;
774 775
		skb->csum = 0;

776
		/* specify the length of each IP datagram fragment */
777
		skb_shinfo(skb)->gso_size = maxfraglen - fragheaderlen;
H
Herbert Xu 已提交
778
		skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
779
		__skb_queue_tail(queue, skb);
780
	}
781 782 783

	return skb_append_datato_frags(sk, skb, getfrag, from,
				       (length - transhdrlen));
784 785
}

786 787 788
static int __ip_append_data(struct sock *sk,
			    struct flowi4 *fl4,
			    struct sk_buff_head *queue,
789 790 791 792 793
			    struct inet_cork *cork,
			    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
794 795 796 797
{
	struct inet_sock *inet = inet_sk(sk);
	struct sk_buff *skb;

798
	struct ip_options *opt = cork->opt;
L
Linus Torvalds 已提交
799 800 801 802 803 804 805 806
	int hh_len;
	int exthdrlen;
	int mtu;
	int copy;
	int err;
	int offset = 0;
	unsigned int maxfraglen, fragheaderlen;
	int csummode = CHECKSUM_NONE;
807
	struct rtable *rt = (struct rtable *)cork->dst;
L
Linus Torvalds 已提交
808

809 810 811
	skb = skb_peek_tail(queue);

	exthdrlen = !skb ? rt->dst.header_len : 0;
812
	mtu = cork->fragsize;
L
Linus Torvalds 已提交
813

814
	hh_len = LL_RESERVED_SPACE(rt->dst.dev);
L
Linus Torvalds 已提交
815 816 817 818

	fragheaderlen = sizeof(struct iphdr) + (opt ? opt->optlen : 0);
	maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen;

819
	if (cork->length + length > 0xFFFF - fragheaderlen) {
820
		ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
E
Eric Dumazet 已提交
821
			       mtu-exthdrlen);
L
Linus Torvalds 已提交
822 823 824 825 826 827 828 829 830
		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 &&
831
	    rt->dst.dev->features & NETIF_F_V4_CSUM &&
L
Linus Torvalds 已提交
832
	    !exthdrlen)
833
		csummode = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
834

835
	cork->length += length;
836
	if (((length > mtu) || (skb && skb_is_gso(skb))) &&
837
	    (sk->sk_protocol == IPPROTO_UDP) &&
838
	    (rt->dst.dev->features & NETIF_F_UFO) && !rt->dst.header_len) {
839 840
		err = ip_ufo_append_data(sk, queue, getfrag, from, length,
					 hh_len, fragheaderlen, transhdrlen,
841
					 maxfraglen, flags);
842
		if (err)
843 844 845
			goto error;
		return 0;
	}
L
Linus Torvalds 已提交
846 847 848 849 850 851 852 853

	/* 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.
	 */

854
	if (!skb)
L
Linus Torvalds 已提交
855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884
		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;

885
			if ((flags & MSG_MORE) &&
886
			    !(rt->dst.dev->features&NETIF_F_SG))
L
Linus Torvalds 已提交
887 888
				alloclen = mtu;
			else
889
				alloclen = fraglen;
L
Linus Torvalds 已提交
890

891 892
			alloclen += exthdrlen;

L
Linus Torvalds 已提交
893 894 895 896 897
			/* 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.
			 */
898
			if (datalen == length + fraggap)
899
				alloclen += rt->dst.trailer_len;
900

L
Linus Torvalds 已提交
901
			if (transhdrlen) {
902
				skb = sock_alloc_send_skb(sk,
L
Linus Torvalds 已提交
903 904 905 906 907 908
						alloclen + hh_len + 15,
						(flags & MSG_DONTWAIT), &err);
			} else {
				skb = NULL;
				if (atomic_read(&sk->sk_wmem_alloc) <=
				    2 * sk->sk_sndbuf)
909
					skb = sock_wmalloc(sk,
L
Linus Torvalds 已提交
910 911 912 913
							   alloclen + hh_len + 15, 1,
							   sk->sk_allocation);
				if (unlikely(skb == NULL))
					err = -ENOBUFS;
914 915 916
				else
					/* only the initial fragment is
					   time stamped */
917
					cork->tx_flags = 0;
L
Linus Torvalds 已提交
918 919 920 921 922 923 924 925 926 927
			}
			if (skb == NULL)
				goto error;

			/*
			 *	Fill in the control structures
			 */
			skb->ip_summed = csummode;
			skb->csum = 0;
			skb_reserve(skb, hh_len);
928
			skb_shinfo(skb)->tx_flags = cork->tx_flags;
L
Linus Torvalds 已提交
929 930 931 932

			/*
			 *	Find where to start putting bytes.
			 */
933
			data = skb_put(skb, fraglen + exthdrlen);
934
			skb_set_network_header(skb, exthdrlen);
935 936
			skb->transport_header = (skb->network_header +
						 fragheaderlen);
937
			data += fragheaderlen + exthdrlen;
L
Linus Torvalds 已提交
938 939 940 941 942 943 944 945

			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;
946
				pskb_trim_unique(skb_prev, maxfraglen);
L
Linus Torvalds 已提交
947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964
			}

			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.
			 */
965
			__skb_queue_tail(queue, skb);
L
Linus Torvalds 已提交
966 967 968 969 970 971
			continue;
		}

		if (copy > length)
			copy = length;

972
		if (!(rt->dst.dev->features&NETIF_F_SG)) {
L
Linus Torvalds 已提交
973 974 975
			unsigned int off;

			off = skb->len;
976
			if (getfrag(from, skb_put(skb, copy),
L
Linus Torvalds 已提交
977 978 979 980 981 982 983 984
					offset, copy, off, skb) < 0) {
				__skb_trim(skb, off);
				err = -EFAULT;
				goto error;
			}
		} else {
			int i = skb_shinfo(skb)->nr_frags;
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i-1];
985 986
			struct page *page = cork->page;
			int off = cork->off;
L
Linus Torvalds 已提交
987 988 989 990 991
			unsigned int left;

			if (page && (left = PAGE_SIZE - off) > 0) {
				if (copy >= left)
					copy = left;
992
				if (page != skb_frag_page(frag)) {
L
Linus Torvalds 已提交
993 994 995 996
					if (i == MAX_SKB_FRAGS) {
						err = -EMSGSIZE;
						goto error;
					}
997
					skb_fill_page_desc(skb, i, page, off, 0);
998
					skb_frag_ref(skb, i);
L
Linus Torvalds 已提交
999 1000 1001 1002 1003 1004 1005 1006 1007 1008
					frag = &skb_shinfo(skb)->frags[i];
				}
			} else if (i < MAX_SKB_FRAGS) {
				if (copy > PAGE_SIZE)
					copy = PAGE_SIZE;
				page = alloc_pages(sk->sk_allocation, 0);
				if (page == NULL)  {
					err = -ENOMEM;
					goto error;
				}
1009 1010
				cork->page = page;
				cork->off = 0;
L
Linus Torvalds 已提交
1011 1012 1013 1014 1015 1016 1017

				skb_fill_page_desc(skb, i, page, 0, 0);
				frag = &skb_shinfo(skb)->frags[i];
			} else {
				err = -EMSGSIZE;
				goto error;
			}
E
Eric Dumazet 已提交
1018
			if (getfrag(from, skb_frag_address(frag)+skb_frag_size(frag),
1019
				    offset, copy, skb->len, skb) < 0) {
L
Linus Torvalds 已提交
1020 1021 1022
				err = -EFAULT;
				goto error;
			}
1023
			cork->off += copy;
E
Eric Dumazet 已提交
1024
			skb_frag_size_add(frag, copy);
L
Linus Torvalds 已提交
1025 1026
			skb->len += copy;
			skb->data_len += copy;
1027 1028
			skb->truesize += copy;
			atomic_add(copy, &sk->sk_wmem_alloc);
L
Linus Torvalds 已提交
1029 1030 1031 1032 1033 1034 1035 1036
		}
		offset += copy;
		length -= copy;
	}

	return 0;

error:
1037
	cork->length -= length;
P
Pavel Emelyanov 已提交
1038
	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTDISCARDS);
1039
	return err;
L
Linus Torvalds 已提交
1040 1041
}

1042 1043 1044 1045
static int ip_setup_cork(struct sock *sk, struct inet_cork *cork,
			 struct ipcm_cookie *ipc, struct rtable **rtp)
{
	struct inet_sock *inet = inet_sk(sk);
1046
	struct ip_options_rcu *opt;
1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
	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;
		}
1060
		memcpy(cork->opt, &opt->opt, sizeof(struct ip_options) + opt->opt.optlen);
1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071
		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;
	cork->fragsize = inet->pmtudisc == IP_PMTUDISC_PROBE ?
1072
			 rt->dst.dev->mtu : dst_mtu(&rt->dst);
1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092
	cork->dst = &rt->dst;
	cork->length = 0;
	cork->tx_flags = ipc->tx_flags;
	cork->page = NULL;
	cork->off = 0;

	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.
 */
1093
int ip_append_data(struct sock *sk, struct flowi4 *fl4,
1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106
		   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)) {
1107
		err = ip_setup_cork(sk, &inet->cork.base, ipc, rtp);
1108 1109 1110 1111 1112 1113
		if (err)
			return err;
	} else {
		transhdrlen = 0;
	}

1114
	return __ip_append_data(sk, fl4, &sk->sk_write_queue, &inet->cork.base, getfrag,
1115 1116 1117
				from, length, transhdrlen, flags);
}

1118
ssize_t	ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
L
Linus Torvalds 已提交
1119 1120 1121 1122 1123 1124
		       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;
1125
	struct inet_cork *cork;
L
Linus Torvalds 已提交
1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140
	int hh_len;
	int mtu;
	int len;
	int err;
	unsigned int maxfraglen, fragheaderlen, fraggap;

	if (inet->hdrincl)
		return -EPERM;

	if (flags&MSG_PROBE)
		return 0;

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

1141 1142 1143 1144
	cork = &inet->cork.base;
	rt = (struct rtable *)cork->dst;
	if (cork->flags & IPCORK_OPT)
		opt = cork->opt;
L
Linus Torvalds 已提交
1145

1146
	if (!(rt->dst.dev->features&NETIF_F_SG))
L
Linus Torvalds 已提交
1147 1148
		return -EOPNOTSUPP;

1149
	hh_len = LL_RESERVED_SPACE(rt->dst.dev);
1150
	mtu = cork->fragsize;
L
Linus Torvalds 已提交
1151 1152 1153 1154

	fragheaderlen = sizeof(struct iphdr) + (opt ? opt->optlen : 0);
	maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen;

1155
	if (cork->length + size > 0xFFFF - fragheaderlen) {
1156
		ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport, mtu);
L
Linus Torvalds 已提交
1157 1158 1159 1160 1161 1162
		return -EMSGSIZE;
	}

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

1163
	cork->length += size;
1164 1165
	if ((size + skb->len > mtu) &&
	    (sk->sk_protocol == IPPROTO_UDP) &&
1166
	    (rt->dst.dev->features & NETIF_F_UFO)) {
1167
		skb_shinfo(skb)->gso_size = mtu - fragheaderlen;
H
Herbert Xu 已提交
1168
		skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
1169
	}
1170

L
Linus Torvalds 已提交
1171 1172 1173 1174

	while (size > 0) {
		int i;

H
Herbert Xu 已提交
1175
		if (skb_is_gso(skb))
1176 1177 1178 1179 1180 1181 1182 1183
			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
1184 1185 1186 1187 1188
		if (len <= 0) {
			struct sk_buff *skb_prev;
			int alloclen;

			skb_prev = skb;
1189
			fraggap = skb_prev->len - maxfraglen;
L
Linus Torvalds 已提交
1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207

			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.
			 */
1208
			skb_put(skb, fragheaderlen + fraggap);
1209
			skb_reset_network_header(skb);
1210 1211
			skb->transport_header = (skb->network_header +
						 fragheaderlen);
L
Linus Torvalds 已提交
1212
			if (fraggap) {
1213 1214
				skb->csum = skb_copy_and_csum_bits(skb_prev,
								   maxfraglen,
1215
						    skb_transport_header(skb),
1216
								   fraggap, 0);
L
Linus Torvalds 已提交
1217 1218
				skb_prev->csum = csum_sub(skb_prev->csum,
							  skb->csum);
1219
				pskb_trim_unique(skb_prev, maxfraglen);
L
Linus Torvalds 已提交
1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232
			}

			/*
			 * 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
1233
			skb_frag_size_add(&skb_shinfo(skb)->frags[i-1], len);
L
Linus Torvalds 已提交
1234 1235 1236 1237 1238 1239 1240 1241 1242
		} 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) {
1243
			__wsum csum;
L
Linus Torvalds 已提交
1244 1245 1246 1247 1248 1249
			csum = csum_page(page, offset, len);
			skb->csum = csum_block_add(skb->csum, csum, skb->len);
		}

		skb->len += len;
		skb->data_len += len;
1250 1251
		skb->truesize += len;
		atomic_add(len, &sk->sk_wmem_alloc);
L
Linus Torvalds 已提交
1252 1253 1254 1255 1256 1257
		offset += len;
		size -= len;
	}
	return 0;

error:
1258
	cork->length -= size;
P
Pavel Emelyanov 已提交
1259
	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTDISCARDS);
L
Linus Torvalds 已提交
1260 1261 1262
	return err;
}

1263
static void ip_cork_release(struct inet_cork *cork)
1264
{
1265 1266 1267 1268 1269
	cork->flags &= ~IPCORK_OPT;
	kfree(cork->opt);
	cork->opt = NULL;
	dst_release(cork->dst);
	cork->dst = NULL;
1270 1271
}

L
Linus Torvalds 已提交
1272 1273 1274 1275
/*
 *	Combined all pending IP fragments on the socket as one IP datagram
 *	and push them out.
 */
1276
struct sk_buff *__ip_make_skb(struct sock *sk,
1277
			      struct flowi4 *fl4,
1278 1279
			      struct sk_buff_head *queue,
			      struct inet_cork *cork)
L
Linus Torvalds 已提交
1280 1281 1282 1283
{
	struct sk_buff *skb, *tmp_skb;
	struct sk_buff **tail_skb;
	struct inet_sock *inet = inet_sk(sk);
1284
	struct net *net = sock_net(sk);
L
Linus Torvalds 已提交
1285
	struct ip_options *opt = NULL;
1286
	struct rtable *rt = (struct rtable *)cork->dst;
L
Linus Torvalds 已提交
1287
	struct iphdr *iph;
1288
	__be16 df = 0;
L
Linus Torvalds 已提交
1289 1290
	__u8 ttl;

1291
	if ((skb = __skb_dequeue(queue)) == NULL)
L
Linus Torvalds 已提交
1292 1293 1294 1295
		goto out;
	tail_skb = &(skb_shinfo(skb)->frag_list);

	/* move skb->data to ip header from ext header */
1296
	if (skb->data < skb_network_header(skb))
1297
		__skb_pull(skb, skb_network_offset(skb));
1298
	while ((tmp_skb = __skb_dequeue(queue)) != NULL) {
1299
		__skb_pull(tmp_skb, skb_network_header_len(skb));
L
Linus Torvalds 已提交
1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312
		*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.
	 */
J
John Heffner 已提交
1313
	if (inet->pmtudisc < IP_PMTUDISC_DO)
L
Linus Torvalds 已提交
1314 1315 1316 1317 1318
		skb->local_df = 1;

	/* DF bit is set when we want to see DF on outgoing frames.
	 * If local_df is set too, we still allow to fragment this frame
	 * locally. */
J
John Heffner 已提交
1319
	if (inet->pmtudisc >= IP_PMTUDISC_DO ||
1320 1321
	    (skb->len <= dst_mtu(&rt->dst) &&
	     ip_dont_fragment(sk, &rt->dst)))
L
Linus Torvalds 已提交
1322 1323
		df = htons(IP_DF);

1324 1325
	if (cork->flags & IPCORK_OPT)
		opt = cork->opt;
L
Linus Torvalds 已提交
1326 1327 1328 1329

	if (rt->rt_type == RTN_MULTICAST)
		ttl = inet->mc_ttl;
	else
1330
		ttl = ip_select_ttl(inet, &rt->dst);
L
Linus Torvalds 已提交
1331 1332 1333 1334 1335 1336

	iph = (struct iphdr *)skb->data;
	iph->version = 4;
	iph->ihl = 5;
	iph->tos = inet->tos;
	iph->frag_off = df;
1337
	ip_select_ident(iph, &rt->dst, sk);
L
Linus Torvalds 已提交
1338 1339
	iph->ttl = ttl;
	iph->protocol = sk->sk_protocol;
1340 1341
	iph->saddr = fl4->saddr;
	iph->daddr = fl4->daddr;
L
Linus Torvalds 已提交
1342

1343 1344 1345 1346 1347
	if (opt) {
		iph->ihl += opt->optlen>>2;
		ip_options_build(skb, opt, cork->addr, rt, 0);
	}

L
Linus Torvalds 已提交
1348
	skb->priority = sk->sk_priority;
1349
	skb->mark = sk->sk_mark;
1350 1351 1352 1353
	/*
	 * Steal rt from cork.dst to avoid a pair of atomic_inc/atomic_dec
	 * on dst refcount
	 */
1354
	cork->dst = NULL;
1355
	skb_dst_set(skb, &rt->dst);
L
Linus Torvalds 已提交
1356

1357
	if (iph->protocol == IPPROTO_ICMP)
1358
		icmp_out_count(net, ((struct icmphdr *)
1359 1360
			skb_transport_header(skb))->type);

1361 1362 1363 1364 1365 1366 1367 1368 1369 1370
	ip_cork_release(cork);
out:
	return skb;
}

int ip_send_skb(struct sk_buff *skb)
{
	struct net *net = sock_net(skb->sk);
	int err;

H
Herbert Xu 已提交
1371
	err = ip_local_out(skb);
L
Linus Torvalds 已提交
1372 1373
	if (err) {
		if (err > 0)
E
Eric Dumazet 已提交
1374
			err = net_xmit_errno(err);
L
Linus Torvalds 已提交
1375
		if (err)
1376
			IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
L
Linus Torvalds 已提交
1377 1378 1379 1380 1381
	}

	return err;
}

1382
int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4)
1383
{
1384 1385
	struct sk_buff *skb;

1386
	skb = ip_finish_skb(sk, fl4);
1387 1388 1389 1390 1391
	if (!skb)
		return 0;

	/* Netfilter gets whole the not fragmented skb. */
	return ip_send_skb(skb);
1392 1393
}

L
Linus Torvalds 已提交
1394 1395 1396
/*
 *	Throw away all pending data on the socket.
 */
1397 1398 1399
static void __ip_flush_pending_frames(struct sock *sk,
				      struct sk_buff_head *queue,
				      struct inet_cork *cork)
L
Linus Torvalds 已提交
1400 1401 1402
{
	struct sk_buff *skb;

1403
	while ((skb = __skb_dequeue_tail(queue)) != NULL)
L
Linus Torvalds 已提交
1404 1405
		kfree_skb(skb);

1406 1407 1408 1409 1410
	ip_cork_release(cork);
}

void ip_flush_pending_frames(struct sock *sk)
{
1411
	__ip_flush_pending_frames(sk, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
L
Linus Torvalds 已提交
1412 1413
}

1414
struct sk_buff *ip_make_skb(struct sock *sk,
1415
			    struct flowi4 *fl4,
1416 1417 1418 1419 1420 1421
			    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)
{
1422
	struct inet_cork cork;
1423 1424 1425 1426 1427 1428 1429 1430
	struct sk_buff_head queue;
	int err;

	if (flags & MSG_PROBE)
		return NULL;

	__skb_queue_head_init(&queue);

1431 1432
	cork.flags = 0;
	cork.addr = 0;
1433
	cork.opt = NULL;
1434 1435 1436 1437
	err = ip_setup_cork(sk, &cork, ipc, rtp);
	if (err)
		return ERR_PTR(err);

1438
	err = __ip_append_data(sk, fl4, &queue, &cork, getfrag,
1439 1440 1441 1442 1443 1444
			       from, length, transhdrlen, flags);
	if (err) {
		__ip_flush_pending_frames(sk, &queue, &cork);
		return ERR_PTR(err);
	}

1445
	return __ip_make_skb(sk, fl4, &queue, &cork);
1446
}
L
Linus Torvalds 已提交
1447 1448 1449 1450

/*
 *	Fetch data from kernel space and fill in checksum if needed.
 */
1451
static int ip_reply_glue_bits(void *dptr, char *to, int offset,
L
Linus Torvalds 已提交
1452 1453
			      int len, int odd, struct sk_buff *skb)
{
1454
	__wsum csum;
L
Linus Torvalds 已提交
1455 1456 1457

	csum = csum_partial_copy_nocheck(dptr+offset, to, len, 0);
	skb->csum = csum_block_add(skb->csum, csum, odd);
1458
	return 0;
L
Linus Torvalds 已提交
1459 1460
}

1461
/*
L
Linus Torvalds 已提交
1462 1463 1464
 *	Generic function to send a packet as reply to another packet.
 *	Used to send TCP resets so far. ICMP should use this function too.
 *
1465
 *	Should run single threaded per socket because it uses the sock
L
Linus Torvalds 已提交
1466 1467
 *     	structure to pass arguments.
 */
1468 1469
void ip_send_reply(struct sock *sk, struct sk_buff *skb, __be32 daddr,
		   struct ip_reply_arg *arg, unsigned int len)
L
Linus Torvalds 已提交
1470 1471
{
	struct inet_sock *inet = inet_sk(sk);
1472
	struct ip_options_data replyopts;
L
Linus Torvalds 已提交
1473
	struct ipcm_cookie ipc;
1474
	struct flowi4 fl4;
E
Eric Dumazet 已提交
1475
	struct rtable *rt = skb_rtable(skb);
L
Linus Torvalds 已提交
1476

1477
	if (ip_options_echo(&replyopts.opt.opt, skb))
L
Linus Torvalds 已提交
1478 1479
		return;

1480
	ipc.addr = daddr;
L
Linus Torvalds 已提交
1481
	ipc.opt = NULL;
1482
	ipc.tx_flags = 0;
L
Linus Torvalds 已提交
1483

1484
	if (replyopts.opt.opt.optlen) {
L
Linus Torvalds 已提交
1485 1486
		ipc.opt = &replyopts.opt;

1487 1488
		if (replyopts.opt.opt.srr)
			daddr = replyopts.opt.opt.faddr;
L
Linus Torvalds 已提交
1489 1490
	}

1491 1492 1493 1494 1495 1496 1497 1498 1499 1500
	flowi4_init_output(&fl4, arg->bound_dev_if, 0,
			   RT_TOS(ip_hdr(skb)->tos),
			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
			   ip_reply_arg_flowi_flags(arg),
			   daddr, rt->rt_spec_dst,
			   tcp_hdr(skb)->source, tcp_hdr(skb)->dest);
	security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
	rt = ip_route_output_key(sock_net(sk), &fl4);
	if (IS_ERR(rt))
		return;
L
Linus Torvalds 已提交
1501 1502 1503 1504 1505 1506 1507 1508

	/* And let IP do all the hard work.

	   This chunk is not reenterable, hence spinlock.
	   Note that it uses the fact, that this function is called
	   with locally disabled BH and that sk cannot be already spinlocked.
	 */
	bh_lock_sock(sk);
1509
	inet->tos = ip_hdr(skb)->tos;
L
Linus Torvalds 已提交
1510
	sk->sk_priority = skb->priority;
1511
	sk->sk_protocol = ip_hdr(skb)->protocol;
1512
	sk->sk_bound_dev_if = arg->bound_dev_if;
1513
	ip_append_data(sk, &fl4, ip_reply_glue_bits, arg->iov->iov_base, len, 0,
1514
		       &ipc, &rt, MSG_DONTWAIT);
L
Linus Torvalds 已提交
1515 1516
	if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) {
		if (arg->csumoffset >= 0)
1517 1518 1519
			*((__sum16 *)skb_transport_header(skb) +
			  arg->csumoffset) = csum_fold(csum_add(skb->csum,
								arg->csum));
L
Linus Torvalds 已提交
1520
		skb->ip_summed = CHECKSUM_NONE;
1521
		ip_push_pending_frames(sk, &fl4);
L
Linus Torvalds 已提交
1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537
	}

	bh_unlock_sock(sk);

	ip_rt_put(rt);
}

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

#if defined(CONFIG_IP_MULTICAST) && defined(CONFIG_PROC_FS)
	igmp_mc_proc_init();
#endif
}