blob: ec562713db9b16e3ee6a3b03ca17a13af2024e77 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * IPv6 output functions
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09003 * Linux INET6 implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 *
5 * Authors:
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09006 * Pedro Roque <roque@di.fc.ul.pt>
Linus Torvalds1da177e2005-04-16 15:20:36 -07007 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * Based on linux/net/ipv4/ip_output.c
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
14 *
15 * Changes:
16 * A.N.Kuznetsov : airthmetics in fragmentation.
17 * extension headers are implemented.
18 * route changes now work.
19 * ip6_forward does not confuse sniffers.
20 * etc.
21 *
22 * H. von Brand : Added missing #include <linux/string.h>
23 * Imran Patel : frag id should be in NBO
24 * Kazunori MIYAZAWA @USAGI
25 * : add ip6_append_data and related functions
26 * for datagram xmit
27 */
28
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/errno.h>
Herbert Xuef76bc22008-01-11 19:15:08 -080030#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <linux/string.h>
32#include <linux/socket.h>
33#include <linux/net.h>
34#include <linux/netdevice.h>
35#include <linux/if_arp.h>
36#include <linux/in6.h>
37#include <linux/tcp.h>
38#include <linux/route.h>
Herbert Xub59f45d2006-05-27 23:05:54 -070039#include <linux/module.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090040#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
42#include <linux/netfilter.h>
43#include <linux/netfilter_ipv6.h>
44
45#include <net/sock.h>
46#include <net/snmp.h>
47
48#include <net/ipv6.h>
49#include <net/ndisc.h>
50#include <net/protocol.h>
51#include <net/ip6_route.h>
52#include <net/addrconf.h>
53#include <net/rawv6.h>
54#include <net/icmp.h>
55#include <net/xfrm.h>
56#include <net/checksum.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090057#include <linux/mroute6.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070058
David Stevensad0081e2010-12-17 11:42:42 +000059int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *));
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
Herbert Xuef76bc22008-01-11 19:15:08 -080061int __ip6_local_out(struct sk_buff *skb)
62{
63 int len;
64
65 len = skb->len - sizeof(struct ipv6hdr);
66 if (len > IPV6_MAXPLEN)
67 len = 0;
68 ipv6_hdr(skb)->payload_len = htons(len);
69
Jan Engelhardtb2e0b382010-03-23 04:09:07 +010070 return nf_hook(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
71 skb_dst(skb)->dev, dst_output);
Herbert Xuef76bc22008-01-11 19:15:08 -080072}
73
74int ip6_local_out(struct sk_buff *skb)
75{
76 int err;
77
78 err = __ip6_local_out(skb);
79 if (likely(err == 1))
80 err = dst_output(skb);
81
82 return err;
83}
84EXPORT_SYMBOL_GPL(ip6_local_out);
85
Linus Torvalds1da177e2005-04-16 15:20:36 -070086/* dev_loopback_xmit for use with netfilter. */
87static int ip6_dev_loopback_xmit(struct sk_buff *newskb)
88{
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -070089 skb_reset_mac_header(newskb);
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -030090 __skb_pull(newskb, skb_network_offset(newskb));
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 newskb->pkt_type = PACKET_LOOPBACK;
92 newskb->ip_summed = CHECKSUM_UNNECESSARY;
Eric Dumazetadf30902009-06-02 05:19:30 +000093 WARN_ON(!skb_dst(newskb));
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
Eric Dumazete30b38c2010-04-15 09:13:03 +000095 netif_rx_ni(newskb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070096 return 0;
97}
98
Jan Engelhardt9e508492010-04-13 15:28:11 +020099static int ip6_finish_output2(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100{
Eric Dumazetadf30902009-06-02 05:19:30 +0000101 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102 struct net_device *dev = dst->dev;
David S. Millerf6b72b62011-07-14 07:53:20 -0700103 struct neighbour *neigh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104
105 skb->protocol = htons(ETH_P_IPV6);
106 skb->dev = dev;
107
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700108 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000109 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110
Octavian Purdila7ad68482010-01-06 20:37:01 -0800111 if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(skb->sk) &&
Patrick McHardyd1db2752010-05-11 14:40:55 +0200112 ((mroute6_socket(dev_net(dev), skb) &&
Benjamin Therybd91b8b2008-12-10 16:07:08 -0800113 !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) ||
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900114 ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
115 &ipv6_hdr(skb)->saddr))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
117
118 /* Do not check for IFF_ALLMULTI; multicast routing
119 is not supported in any case.
120 */
121 if (newskb)
Jan Engelhardtb2e0b382010-03-23 04:09:07 +0100122 NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING,
123 newskb, NULL, newskb->dev,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124 ip6_dev_loopback_xmit);
125
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700126 if (ipv6_hdr(skb)->hop_limit == 0) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700127 IP6_INC_STATS(dev_net(dev), idev,
128 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 kfree_skb(skb);
130 return 0;
131 }
132 }
133
Neil Hormanedf391f2009-04-27 02:45:02 -0700134 IP6_UPD_PO_STATS(dev_net(dev), idev, IPSTATS_MIB_OUTMCAST,
135 skb->len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 }
137
Eric Dumazetf2c31e32011-07-29 19:00:53 +0000138 rcu_read_lock();
David S. Miller69cce1d2011-07-17 23:09:49 -0700139 neigh = dst_get_neighbour(dst);
Eric Dumazetf2c31e32011-07-29 19:00:53 +0000140 if (neigh) {
141 int res = neigh_output(neigh, skb);
David S. Miller05e3aa02011-07-16 17:26:00 -0700142
Eric Dumazetf2c31e32011-07-29 19:00:53 +0000143 rcu_read_unlock();
144 return res;
145 }
146 rcu_read_unlock();
Jan Engelhardt9e508492010-04-13 15:28:11 +0200147 IP6_INC_STATS_BH(dev_net(dst->dev),
148 ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
149 kfree_skb(skb);
150 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151}
152
Jan Engelhardt9e508492010-04-13 15:28:11 +0200153static int ip6_finish_output(struct sk_buff *skb)
154{
155 if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
156 dst_allfrag(skb_dst(skb)))
157 return ip6_fragment(skb, ip6_finish_output2);
158 else
159 return ip6_finish_output2(skb);
160}
161
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162int ip6_output(struct sk_buff *skb)
163{
Jan Engelhardt9e508492010-04-13 15:28:11 +0200164 struct net_device *dev = skb_dst(skb)->dev;
Eric Dumazetadf30902009-06-02 05:19:30 +0000165 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900166 if (unlikely(idev->cnf.disable_ipv6)) {
Jan Engelhardt9e508492010-04-13 15:28:11 +0200167 IP6_INC_STATS(dev_net(dev), idev,
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700168 IPSTATS_MIB_OUTDISCARDS);
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900169 kfree_skb(skb);
170 return 0;
171 }
172
Jan Engelhardt9c6eb282010-04-13 15:32:16 +0200173 return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING, skb, NULL, dev,
174 ip6_finish_output,
175 !(IP6CB(skb)->flags & IP6SKB_REROUTED));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176}
177
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178/*
Shan Weib5d43992010-04-15 16:48:48 +0000179 * xmit an sk_buff (used by TCP, SCTP and DCCP)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180 */
181
David S. Miller4c9483b2011-03-12 16:22:43 -0500182int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
Eric Dumazetb903d322011-10-27 00:44:35 -0400183 struct ipv6_txoptions *opt, int tclass)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184{
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700185 struct net *net = sock_net(sk);
Patrick McHardyb30bd282006-03-23 01:17:25 -0800186 struct ipv6_pinfo *np = inet6_sk(sk);
David S. Miller4c9483b2011-03-12 16:22:43 -0500187 struct in6_addr *first_hop = &fl6->daddr;
Eric Dumazetadf30902009-06-02 05:19:30 +0000188 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 struct ipv6hdr *hdr;
David S. Miller4c9483b2011-03-12 16:22:43 -0500190 u8 proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 int seg_len = skb->len;
Gerrit Renkere651f032009-08-09 08:12:48 +0000192 int hlimit = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 u32 mtu;
194
195 if (opt) {
Chuck Leverc2636b42007-10-23 21:07:32 -0700196 unsigned int head_room;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197
198 /* First: exthdrs may take lots of space (~8K for now)
199 MAX_HEADER is not enough.
200 */
201 head_room = opt->opt_nflen + opt->opt_flen;
202 seg_len += head_room;
203 head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
204
205 if (skb_headroom(skb) < head_room) {
206 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900207 if (skb2 == NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000208 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900209 IPSTATS_MIB_OUTDISCARDS);
210 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 return -ENOBUFS;
212 }
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900213 kfree_skb(skb);
214 skb = skb2;
Dan Carpenter83d7eb22010-04-30 16:42:08 -0700215 skb_set_owner_w(skb, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 }
217 if (opt->opt_flen)
218 ipv6_push_frag_opts(skb, opt, &proto);
219 if (opt->opt_nflen)
220 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
221 }
222
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700223 skb_push(skb, sizeof(struct ipv6hdr));
224 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700225 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226
227 /*
228 * Fill in the IPv6 header
229 */
Eric Dumazetb903d322011-10-27 00:44:35 -0400230 if (np)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 hlimit = np->hop_limit;
232 if (hlimit < 0)
YOSHIFUJI Hideaki6b75d092008-03-10 06:00:30 -0400233 hlimit = ip6_dst_hoplimit(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234
David S. Miller4c9483b2011-03-12 16:22:43 -0500235 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | fl6->flowlabel;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900236
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 hdr->payload_len = htons(seg_len);
238 hdr->nexthdr = proto;
239 hdr->hop_limit = hlimit;
240
David S. Miller4c9483b2011-03-12 16:22:43 -0500241 ipv6_addr_copy(&hdr->saddr, &fl6->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 ipv6_addr_copy(&hdr->daddr, first_hop);
243
Patrick McHardya2c20642006-01-08 22:37:26 -0800244 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800245 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -0800246
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 mtu = dst_mtu(dst);
Wei Yongjun283d07ac2008-08-03 21:15:59 -0700248 if ((skb->len <= mtu) || skb->local_df || skb_is_gso(skb)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000249 IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)),
Neil Hormanedf391f2009-04-27 02:45:02 -0700250 IPSTATS_MIB_OUT, skb->len);
Jan Engelhardtb2e0b382010-03-23 04:09:07 +0100251 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
252 dst->dev, dst_output);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253 }
254
255 if (net_ratelimit())
256 printk(KERN_DEBUG "IPv6: sending pkt_too_big to self\n");
257 skb->dev = dst->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000258 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazetadf30902009-06-02 05:19:30 +0000259 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 kfree_skb(skb);
261 return -EMSGSIZE;
262}
263
YOSHIFUJI Hideaki71590392007-02-22 22:05:40 +0900264EXPORT_SYMBOL(ip6_xmit);
265
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266/*
267 * To avoid extra problems ND packets are send through this
268 * routine. It's code duplication but I really want to avoid
269 * extra checks since ipv6_build_header is used by TCP (which
270 * is for us performance critical)
271 */
272
273int ip6_nd_hdr(struct sock *sk, struct sk_buff *skb, struct net_device *dev,
YOSHIFUJI Hideaki9acd9f32008-04-10 15:42:10 +0900274 const struct in6_addr *saddr, const struct in6_addr *daddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 int proto, int len)
276{
277 struct ipv6_pinfo *np = inet6_sk(sk);
278 struct ipv6hdr *hdr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279
280 skb->protocol = htons(ETH_P_IPV6);
281 skb->dev = dev;
282
Arnaldo Carvalho de Melo55f79cc2007-03-14 21:05:03 -0300283 skb_reset_network_header(skb);
284 skb_put(skb, sizeof(struct ipv6hdr));
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700285 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286
Al Viroae08e1f2006-11-08 00:27:11 -0800287 *(__be32*)hdr = htonl(0x60000000);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288
289 hdr->payload_len = htons(len);
290 hdr->nexthdr = proto;
291 hdr->hop_limit = np->hop_limit;
292
293 ipv6_addr_copy(&hdr->saddr, saddr);
294 ipv6_addr_copy(&hdr->daddr, daddr);
295
296 return 0;
297}
298
299static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
300{
301 struct ip6_ra_chain *ra;
302 struct sock *last = NULL;
303
304 read_lock(&ip6_ra_lock);
305 for (ra = ip6_ra_chain; ra; ra = ra->next) {
306 struct sock *sk = ra->sk;
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700307 if (sk && ra->sel == sel &&
308 (!sk->sk_bound_dev_if ||
309 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 if (last) {
311 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
312 if (skb2)
313 rawv6_rcv(last, skb2);
314 }
315 last = sk;
316 }
317 }
318
319 if (last) {
320 rawv6_rcv(last, skb);
321 read_unlock(&ip6_ra_lock);
322 return 1;
323 }
324 read_unlock(&ip6_ra_lock);
325 return 0;
326}
327
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700328static int ip6_forward_proxy_check(struct sk_buff *skb)
329{
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700330 struct ipv6hdr *hdr = ipv6_hdr(skb);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700331 u8 nexthdr = hdr->nexthdr;
332 int offset;
333
334 if (ipv6_ext_hdr(nexthdr)) {
335 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr);
336 if (offset < 0)
337 return 0;
338 } else
339 offset = sizeof(struct ipv6hdr);
340
341 if (nexthdr == IPPROTO_ICMPV6) {
342 struct icmp6hdr *icmp6;
343
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700344 if (!pskb_may_pull(skb, (skb_network_header(skb) +
345 offset + 1 - skb->data)))
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700346 return 0;
347
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700348 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700349
350 switch (icmp6->icmp6_type) {
351 case NDISC_ROUTER_SOLICITATION:
352 case NDISC_ROUTER_ADVERTISEMENT:
353 case NDISC_NEIGHBOUR_SOLICITATION:
354 case NDISC_NEIGHBOUR_ADVERTISEMENT:
355 case NDISC_REDIRECT:
356 /* For reaction involving unicast neighbor discovery
357 * message destined to the proxied address, pass it to
358 * input function.
359 */
360 return 1;
361 default:
362 break;
363 }
364 }
365
Ville Nuorvala74553b02006-09-22 14:42:18 -0700366 /*
367 * The proxying router can't forward traffic sent to a link-local
368 * address, so signal the sender and discard the packet. This
369 * behavior is clarified by the MIPv6 specification.
370 */
371 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
372 dst_link_failure(skb);
373 return -1;
374 }
375
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700376 return 0;
377}
378
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379static inline int ip6_forward_finish(struct sk_buff *skb)
380{
381 return dst_output(skb);
382}
383
384int ip6_forward(struct sk_buff *skb)
385{
Eric Dumazetadf30902009-06-02 05:19:30 +0000386 struct dst_entry *dst = skb_dst(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700387 struct ipv6hdr *hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 struct inet6_skb_parm *opt = IP6CB(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900389 struct net *net = dev_net(dst->dev);
David S. Miller69cce1d2011-07-17 23:09:49 -0700390 struct neighbour *n;
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800391 u32 mtu;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900392
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700393 if (net->ipv6.devconf_all->forwarding == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 goto error;
395
Ben Hutchings4497b072008-06-19 16:22:28 -0700396 if (skb_warn_if_lro(skb))
397 goto drop;
398
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700400 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 goto drop;
402 }
403
Alexey Kuznetsov72b43d02011-01-12 08:34:08 +0000404 if (skb->pkt_type != PACKET_HOST)
405 goto drop;
406
Herbert Xu35fc92a2007-03-26 23:22:20 -0700407 skb_forward_csum(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408
409 /*
410 * We DO NOT make any processing on
411 * RA packets, pushing them to user level AS IS
412 * without ane WARRANTY that application will be able
413 * to interpret them. The reason is that we
414 * cannot make anything clever here.
415 *
416 * We are not end-node, so that if packet contains
417 * AH/ESP, we cannot make anything.
418 * Defragmentation also would be mistake, RA packets
419 * cannot be fragmented, because there is no warranty
420 * that different fragments will go along one path. --ANK
421 */
422 if (opt->ra) {
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700423 u8 *ptr = skb_network_header(skb) + opt->ra;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 if (ip6_call_ra_chain(skb, (ptr[2]<<8) + ptr[3]))
425 return 0;
426 }
427
428 /*
429 * check and decrement ttl
430 */
431 if (hdr->hop_limit <= 1) {
432 /* Force OUTPUT device used as source address */
433 skb->dev = dst->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000434 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0);
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700435 IP6_INC_STATS_BH(net,
436 ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
438 kfree_skb(skb);
439 return -ETIMEDOUT;
440 }
441
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -0700442 /* XXX: idev->cnf.proxy_ndp? */
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700443 if (net->ipv6.devconf_all->proxy_ndp &&
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800444 pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
Ville Nuorvala74553b02006-09-22 14:42:18 -0700445 int proxied = ip6_forward_proxy_check(skb);
446 if (proxied > 0)
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700447 return ip6_input(skb);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700448 else if (proxied < 0) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700449 IP6_INC_STATS(net, ip6_dst_idev(dst),
450 IPSTATS_MIB_INDISCARDS);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700451 goto drop;
452 }
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700453 }
454
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 if (!xfrm6_route_forward(skb)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700456 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 goto drop;
458 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000459 dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460
461 /* IPv6 specs say nothing about it, but it is clear that we cannot
462 send redirects to source routed frames.
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900463 We don't send redirects to frames decapsulated from IPsec.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 */
David S. Miller69cce1d2011-07-17 23:09:49 -0700465 n = dst_get_neighbour(dst);
466 if (skb->dev == dst->dev && n && opt->srcrt == 0 && !skb_sec_path(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 struct in6_addr *target = NULL;
468 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
470 /*
471 * incoming and outgoing devices are the same
472 * send a redirect.
473 */
474
475 rt = (struct rt6_info *) dst;
476 if ((rt->rt6i_flags & RTF_GATEWAY))
477 target = (struct in6_addr*)&n->primary_key;
478 else
479 target = &hdr->daddr;
480
David S. Miller92d86822011-02-04 15:55:25 -0800481 if (!rt->rt6i_peer)
482 rt6_bind_peer(rt, 1);
483
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 /* Limit redirects both by destination (here)
485 and by source (inside ndisc_send_redirect)
486 */
David S. Miller92d86822011-02-04 15:55:25 -0800487 if (inet_peer_xrlim_allow(rt->rt6i_peer, 1*HZ))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 ndisc_send_redirect(skb, n, target);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700489 } else {
490 int addrtype = ipv6_addr_type(&hdr->saddr);
491
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 /* This check is security critical. */
YOSHIFUJI Hideakif81b2e72008-06-25 16:55:26 +0900493 if (addrtype == IPV6_ADDR_ANY ||
494 addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK))
David L Stevens5bb1ab02007-05-09 13:53:44 -0700495 goto error;
496 if (addrtype & IPV6_ADDR_LINKLOCAL) {
497 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000498 ICMPV6_NOT_NEIGHBOUR, 0);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700499 goto error;
500 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 }
502
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800503 mtu = dst_mtu(dst);
504 if (mtu < IPV6_MIN_MTU)
505 mtu = IPV6_MIN_MTU;
506
Herbert Xu0aa68272010-05-27 16:14:30 -0700507 if (skb->len > mtu && !skb_is_gso(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 /* Again, force OUTPUT device used as source address */
509 skb->dev = dst->dev;
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800510 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700511 IP6_INC_STATS_BH(net,
512 ip6_dst_idev(dst), IPSTATS_MIB_INTOOBIGERRORS);
513 IP6_INC_STATS_BH(net,
514 ip6_dst_idev(dst), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 kfree_skb(skb);
516 return -EMSGSIZE;
517 }
518
519 if (skb_cow(skb, dst->dev->hard_header_len)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700520 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 goto drop;
522 }
523
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700524 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525
526 /* Mangling hops number delayed to point after skb COW */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900527
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 hdr->hop_limit--;
529
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700530 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
Jan Engelhardtb2e0b382010-03-23 04:09:07 +0100531 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dst->dev,
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800532 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
534error:
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700535 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536drop:
537 kfree_skb(skb);
538 return -EINVAL;
539}
540
541static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
542{
543 to->pkt_type = from->pkt_type;
544 to->priority = from->priority;
545 to->protocol = from->protocol;
Eric Dumazetadf30902009-06-02 05:19:30 +0000546 skb_dst_drop(to);
547 skb_dst_set(to, dst_clone(skb_dst(from)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 to->dev = from->dev;
Thomas Graf82e91ff2006-11-09 15:19:14 -0800549 to->mark = from->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550
551#ifdef CONFIG_NET_SCHED
552 to->tc_index = from->tc_index;
553#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -0700554 nf_copy(to, from);
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700555#if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
556 defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
557 to->nf_trace = from->nf_trace;
558#endif
James Morris984bc162006-06-09 00:29:17 -0700559 skb_copy_secmark(to, from);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560}
561
562int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr)
563{
564 u16 offset = sizeof(struct ipv6hdr);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700565 struct ipv6_opt_hdr *exthdr =
566 (struct ipv6_opt_hdr *)(ipv6_hdr(skb) + 1);
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700567 unsigned int packet_len = skb->tail - skb->network_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 int found_rhdr = 0;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700569 *nexthdr = &ipv6_hdr(skb)->nexthdr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570
571 while (offset + 1 <= packet_len) {
572
573 switch (**nexthdr) {
574
575 case NEXTHDR_HOP:
Masahide NAKAMURA27637df2006-08-23 19:29:47 -0700576 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 case NEXTHDR_ROUTING:
Masahide NAKAMURA27637df2006-08-23 19:29:47 -0700578 found_rhdr = 1;
579 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 case NEXTHDR_DEST:
Masahide NAKAMURA59fbb3a62007-06-26 23:56:32 -0700581#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
Masahide NAKAMURA27637df2006-08-23 19:29:47 -0700582 if (ipv6_find_tlv(skb, offset, IPV6_TLV_HAO) >= 0)
583 break;
584#endif
585 if (found_rhdr)
586 return offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 break;
588 default :
589 return offset;
590 }
Masahide NAKAMURA27637df2006-08-23 19:29:47 -0700591
592 offset += ipv6_optlen(exthdr);
593 *nexthdr = &exthdr->nexthdr;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700594 exthdr = (struct ipv6_opt_hdr *)(skb_network_header(skb) +
595 offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 }
597
598 return offset;
599}
600
Eric Dumazet87c48fa2011-07-21 21:25:58 -0700601void ipv6_select_ident(struct frag_hdr *fhdr, struct rt6_info *rt)
602{
603 static atomic_t ipv6_fragmentation_id;
604 int old, new;
605
Eric Dumazete688a602011-12-22 04:15:53 +0000606 if (rt && !(rt->dst.flags & DST_NOPEER)) {
Eric Dumazet87c48fa2011-07-21 21:25:58 -0700607 struct inet_peer *peer;
608
609 if (!rt->rt6i_peer)
610 rt6_bind_peer(rt, 1);
611 peer = rt->rt6i_peer;
612 if (peer) {
613 fhdr->identification = htonl(inet_getid(peer, 0));
614 return;
615 }
616 }
617 do {
618 old = atomic_read(&ipv6_fragmentation_id);
619 new = old + 1;
620 if (!new)
621 new = 1;
622 } while (atomic_cmpxchg(&ipv6_fragmentation_id, old, new) != old);
623 fhdr->identification = htonl(new);
624}
625
David Stevensad0081e2010-12-17 11:42:42 +0000626int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 struct sk_buff *frag;
Eric Dumazetadf30902009-06-02 05:19:30 +0000629 struct rt6_info *rt = (struct rt6_info*)skb_dst(skb);
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800630 struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 struct ipv6hdr *tmp_hdr;
632 struct frag_hdr *fh;
633 unsigned int mtu, hlen, left, len;
Al Viroae08e1f2006-11-08 00:27:11 -0800634 __be32 frag_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 int ptr, offset = 0, err=0;
636 u8 *prevhdr, nexthdr = 0;
Eric Dumazetadf30902009-06-02 05:19:30 +0000637 struct net *net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639 hlen = ip6_find_1stfragopt(skb, &prevhdr);
640 nexthdr = *prevhdr;
641
John Heffner628a5c52007-04-20 15:53:27 -0700642 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700643
644 /* We must not fragment if the socket is set to force MTU discovery
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800645 * or if the skb it not generated by a local socket.
John Heffnerb881ef72007-04-20 15:52:39 -0700646 */
Shan Weif2228f72010-04-18 16:58:22 +0000647 if (!skb->local_df && skb->len > mtu) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000648 skb->dev = skb_dst(skb)->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000649 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazetadf30902009-06-02 05:19:30 +0000650 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700651 IPSTATS_MIB_FRAGFAILS);
John Heffnerb881ef72007-04-20 15:52:39 -0700652 kfree_skb(skb);
653 return -EMSGSIZE;
654 }
655
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800656 if (np && np->frag_size < mtu) {
657 if (np->frag_size)
658 mtu = np->frag_size;
659 }
660 mtu -= hlen + sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661
David S. Miller21dc3302010-08-23 00:13:46 -0700662 if (skb_has_frag_list(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 int first_len = skb_pagelen(skb);
Eric Dumazet3d130082010-09-21 08:47:45 +0000664 struct sk_buff *frag2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665
666 if (first_len - hlen > mtu ||
667 ((first_len - hlen) & 7) ||
668 skb_cloned(skb))
669 goto slow_path;
670
David S. Miller4d9092b2009-06-09 00:20:05 -0700671 skb_walk_frags(skb, frag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 /* Correct geometry. */
673 if (frag->len > mtu ||
674 ((frag->len & 7) && frag->next) ||
675 skb_headroom(frag) < hlen)
Eric Dumazet3d130082010-09-21 08:47:45 +0000676 goto slow_path_clean;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 /* Partially cloned skb? */
679 if (skb_shared(frag))
Eric Dumazet3d130082010-09-21 08:47:45 +0000680 goto slow_path_clean;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700681
682 BUG_ON(frag->sk);
683 if (skb->sk) {
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700684 frag->sk = skb->sk;
685 frag->destructor = sock_wfree;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700686 }
Eric Dumazet3d130082010-09-21 08:47:45 +0000687 skb->truesize -= frag->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 }
689
690 err = 0;
691 offset = 0;
692 frag = skb_shinfo(skb)->frag_list;
David S. Miller4d9092b2009-06-09 00:20:05 -0700693 skb_frag_list_init(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 /* BUILD HEADER */
695
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800696 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700697 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 if (!tmp_hdr) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000699 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700700 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 return -ENOMEM;
702 }
703
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 __skb_pull(skb, hlen);
705 fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700706 __skb_push(skb, hlen);
707 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700708 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709
Eric Dumazet87c48fa2011-07-21 21:25:58 -0700710 ipv6_select_ident(fh, rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 fh->nexthdr = nexthdr;
712 fh->reserved = 0;
713 fh->frag_off = htons(IP6_MF);
714 frag_id = fh->identification;
715
716 first_len = skb_pagelen(skb);
717 skb->data_len = first_len - skb_headlen(skb);
718 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700719 ipv6_hdr(skb)->payload_len = htons(first_len -
720 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900721
Changli Gaod8d1f302010-06-10 23:31:35 -0700722 dst_hold(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723
724 for (;;) {
725 /* Prepare header of the next frame,
726 * before previous one went down. */
727 if (frag) {
728 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300729 skb_reset_transport_header(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700731 __skb_push(frag, hlen);
732 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700733 memcpy(skb_network_header(frag), tmp_hdr,
734 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 offset += skb->len - hlen - sizeof(struct frag_hdr);
736 fh->nexthdr = nexthdr;
737 fh->reserved = 0;
738 fh->frag_off = htons(offset);
739 if (frag->next != NULL)
740 fh->frag_off |= htons(IP6_MF);
741 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700742 ipv6_hdr(frag)->payload_len =
743 htons(frag->len -
744 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 ip6_copy_metadata(frag, skb);
746 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900747
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 err = output(skb);
Wei Dongdafee492006-08-02 13:41:21 -0700749 if(!err)
Changli Gaod8d1f302010-06-10 23:31:35 -0700750 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700751 IPSTATS_MIB_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700752
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 if (err || !frag)
754 break;
755
756 skb = frag;
757 frag = skb->next;
758 skb->next = NULL;
759 }
760
Jesper Juhla51482b2005-11-08 09:41:34 -0800761 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762
763 if (err == 0) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700764 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700765 IPSTATS_MIB_FRAGOKS);
Changli Gaod8d1f302010-06-10 23:31:35 -0700766 dst_release(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 return 0;
768 }
769
770 while (frag) {
771 skb = frag->next;
772 kfree_skb(frag);
773 frag = skb;
774 }
775
Changli Gaod8d1f302010-06-10 23:31:35 -0700776 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700777 IPSTATS_MIB_FRAGFAILS);
Changli Gaod8d1f302010-06-10 23:31:35 -0700778 dst_release(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 return err;
Eric Dumazet3d130082010-09-21 08:47:45 +0000780
781slow_path_clean:
782 skb_walk_frags(skb, frag2) {
783 if (frag2 == frag)
784 break;
785 frag2->sk = NULL;
786 frag2->destructor = NULL;
787 skb->truesize += frag2->truesize;
788 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 }
790
791slow_path:
792 left = skb->len - hlen; /* Space per frame */
793 ptr = hlen; /* Where to start from */
794
795 /*
796 * Fragment the datagram.
797 */
798
799 *prevhdr = NEXTHDR_FRAGMENT;
800
801 /*
802 * Keep copying data until we run out.
803 */
804 while(left > 0) {
805 len = left;
806 /* IF: it doesn't fit, use 'mtu' - the data space left */
807 if (len > mtu)
808 len = mtu;
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300809 /* IF: we are not sending up to and including the packet end
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 then align the next start on an eight byte boundary */
811 if (len < left) {
812 len &= ~7;
813 }
814 /*
815 * Allocate buffer.
816 */
817
Changli Gaod8d1f302010-06-10 23:31:35 -0700818 if ((frag = alloc_skb(len+hlen+sizeof(struct frag_hdr)+LL_ALLOCATED_SPACE(rt->dst.dev), GFP_ATOMIC)) == NULL) {
Patrick McHardy64ce2072005-08-09 20:50:53 -0700819 NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n");
Eric Dumazetadf30902009-06-02 05:19:30 +0000820 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900821 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 err = -ENOMEM;
823 goto fail;
824 }
825
826 /*
827 * Set up data on packet
828 */
829
830 ip6_copy_metadata(frag, skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700831 skb_reserve(frag, LL_RESERVED_SPACE(rt->dst.dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700833 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300834 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700835 frag->transport_header = (frag->network_header + hlen +
836 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837
838 /*
839 * Charge the memory for the fragment to any owner
840 * it might possess
841 */
842 if (skb->sk)
843 skb_set_owner_w(frag, skb->sk);
844
845 /*
846 * Copy the packet header into the new buffer.
847 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300848 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
850 /*
851 * Build fragment header.
852 */
853 fh->nexthdr = nexthdr;
854 fh->reserved = 0;
Yan Zhengf36d6ab2005-10-03 14:19:15 -0700855 if (!frag_id) {
Eric Dumazet87c48fa2011-07-21 21:25:58 -0700856 ipv6_select_ident(fh, rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 frag_id = fh->identification;
858 } else
859 fh->identification = frag_id;
860
861 /*
862 * Copy a block of the IP datagram.
863 */
Wei Yongjun8984e412007-08-21 20:59:08 -0700864 if (skb_copy_bits(skb, ptr, skb_transport_header(frag), len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 BUG();
866 left -= len;
867
868 fh->frag_off = htons(offset);
869 if (left > 0)
870 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700871 ipv6_hdr(frag)->payload_len = htons(frag->len -
872 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873
874 ptr += len;
875 offset += len;
876
877 /*
878 * Put this fragment into the sending queue.
879 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 err = output(frag);
881 if (err)
882 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700883
Eric Dumazetadf30902009-06-02 05:19:30 +0000884 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700885 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000887 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900888 IPSTATS_MIB_FRAGOKS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 return err;
891
892fail:
Eric Dumazetadf30902009-06-02 05:19:30 +0000893 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900894 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900895 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 return err;
897}
898
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000899static inline int ip6_rt_check(const struct rt6key *rt_key,
900 const struct in6_addr *fl_addr,
901 const struct in6_addr *addr_cache)
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700902{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000903 return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
904 (addr_cache == NULL || !ipv6_addr_equal(fl_addr, addr_cache));
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700905}
906
Herbert Xu497c6152006-07-30 20:19:33 -0700907static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
908 struct dst_entry *dst,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000909 const struct flowi6 *fl6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910{
Herbert Xu497c6152006-07-30 20:19:33 -0700911 struct ipv6_pinfo *np = inet6_sk(sk);
912 struct rt6_info *rt = (struct rt6_info *)dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913
Herbert Xu497c6152006-07-30 20:19:33 -0700914 if (!dst)
915 goto out;
916
917 /* Yes, checking route validity in not connected
918 * case is not very simple. Take into account,
919 * that we do not support routing by source, TOS,
920 * and MSG_DONTROUTE --ANK (980726)
921 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700922 * 1. ip6_rt_check(): If route was host route,
923 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700924 * If it is network route, we still may
925 * check its validity using saved pointer
926 * to the last used address: daddr_cache.
927 * We do not want to save whole address now,
928 * (because main consumer of this service
929 * is tcp, which has not this problem),
930 * so that the last trick works only on connected
931 * sockets.
932 * 2. oif also should be the same.
933 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500934 if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700935#ifdef CONFIG_IPV6_SUBTREES
David S. Miller4c9483b2011-03-12 16:22:43 -0500936 ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700937#endif
David S. Miller4c9483b2011-03-12 16:22:43 -0500938 (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex)) {
Herbert Xu497c6152006-07-30 20:19:33 -0700939 dst_release(dst);
940 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 }
942
Herbert Xu497c6152006-07-30 20:19:33 -0700943out:
944 return dst;
945}
946
947static int ip6_dst_lookup_tail(struct sock *sk,
David S. Miller4c9483b2011-03-12 16:22:43 -0500948 struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700949{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900950 struct net *net = sock_net(sk);
David S. Miller69cce1d2011-07-17 23:09:49 -0700951#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
952 struct neighbour *n;
953#endif
954 int err;
Herbert Xu497c6152006-07-30 20:19:33 -0700955
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 if (*dst == NULL)
David S. Miller4c9483b2011-03-12 16:22:43 -0500957 *dst = ip6_route_output(net, sk, fl6);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958
959 if ((err = (*dst)->error))
960 goto out_err_release;
961
David S. Miller4c9483b2011-03-12 16:22:43 -0500962 if (ipv6_addr_any(&fl6->saddr)) {
Daniel Walterc3968a82011-04-13 21:10:57 +0000963 struct rt6_info *rt = (struct rt6_info *) *dst;
964 err = ip6_route_get_saddr(net, rt, &fl6->daddr,
965 sk ? inet6_sk(sk)->srcprefs : 0,
966 &fl6->saddr);
Olaf Hering44456d32005-07-27 11:45:17 -0700967 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 goto out_err_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 }
970
Neil Horman95c385b2007-04-25 17:08:10 -0700971#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700972 /*
973 * Here if the dst entry we've looked up
974 * has a neighbour entry that is in the INCOMPLETE
975 * state and the src address from the flow is
976 * marked as OPTIMISTIC, we release the found
977 * dst entry and replace it instead with the
978 * dst entry of the nexthop router
979 */
Eric Dumazetf2c31e32011-07-29 19:00:53 +0000980 rcu_read_lock();
David S. Miller69cce1d2011-07-17 23:09:49 -0700981 n = dst_get_neighbour(*dst);
982 if (n && !(n->nud_state & NUD_VALID)) {
Neil Hormane550dfb2008-09-09 13:51:35 -0700983 struct inet6_ifaddr *ifp;
David S. Miller4c9483b2011-03-12 16:22:43 -0500984 struct flowi6 fl_gw6;
Neil Hormane550dfb2008-09-09 13:51:35 -0700985 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -0700986
Eric Dumazetf2c31e32011-07-29 19:00:53 +0000987 rcu_read_unlock();
David S. Miller4c9483b2011-03-12 16:22:43 -0500988 ifp = ipv6_get_ifaddr(net, &fl6->saddr,
Neil Hormane550dfb2008-09-09 13:51:35 -0700989 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -0700990
Neil Hormane550dfb2008-09-09 13:51:35 -0700991 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
992 if (ifp)
993 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -0700994
Neil Hormane550dfb2008-09-09 13:51:35 -0700995 if (redirect) {
996 /*
997 * We need to get the dst entry for the
998 * default router instead
999 */
1000 dst_release(*dst);
David S. Miller4c9483b2011-03-12 16:22:43 -05001001 memcpy(&fl_gw6, fl6, sizeof(struct flowi6));
1002 memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr));
1003 *dst = ip6_route_output(net, sk, &fl_gw6);
Neil Hormane550dfb2008-09-09 13:51:35 -07001004 if ((err = (*dst)->error))
1005 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -07001006 }
Eric Dumazetf2c31e32011-07-29 19:00:53 +00001007 } else {
1008 rcu_read_unlock();
Neil Hormane550dfb2008-09-09 13:51:35 -07001009 }
Neil Horman95c385b2007-04-25 17:08:10 -07001010#endif
1011
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 return 0;
1013
1014out_err_release:
Mitsuru Chinenca46f9c2007-12-05 22:31:47 -08001015 if (err == -ENETUNREACH)
Denis V. Lunev483a47d2008-10-08 11:09:27 -07001016 IP6_INC_STATS_BH(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 dst_release(*dst);
1018 *dst = NULL;
1019 return err;
1020}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001021
Herbert Xu497c6152006-07-30 20:19:33 -07001022/**
1023 * ip6_dst_lookup - perform route lookup on flow
1024 * @sk: socket which provides route info
1025 * @dst: pointer to dst_entry * for result
David S. Miller4c9483b2011-03-12 16:22:43 -05001026 * @fl6: flow to lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001027 *
1028 * This function performs a route lookup on the given flow.
1029 *
1030 * It returns zero on success, or a standard errno code on error.
1031 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001032int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -07001033{
1034 *dst = NULL;
David S. Miller4c9483b2011-03-12 16:22:43 -05001035 return ip6_dst_lookup_tail(sk, dst, fl6);
Herbert Xu497c6152006-07-30 20:19:33 -07001036}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -08001037EXPORT_SYMBOL_GPL(ip6_dst_lookup);
1038
Herbert Xu497c6152006-07-30 20:19:33 -07001039/**
David S. Miller68d0c6d2011-03-01 13:19:07 -08001040 * ip6_dst_lookup_flow - perform route lookup on flow with ipsec
1041 * @sk: socket which provides route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001042 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001043 * @final_dst: final destination address for ipsec lookup
David S. Millera1414712011-03-01 14:32:04 -08001044 * @can_sleep: we are in a sleepable context
David S. Miller68d0c6d2011-03-01 13:19:07 -08001045 *
1046 * This function performs a route lookup on the given flow.
1047 *
1048 * It returns a valid dst pointer on success, or a pointer encoded
1049 * error code.
1050 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001051struct dst_entry *ip6_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
David S. Miller68d0c6d2011-03-01 13:19:07 -08001052 const struct in6_addr *final_dst,
David S. Millera1414712011-03-01 14:32:04 -08001053 bool can_sleep)
David S. Miller68d0c6d2011-03-01 13:19:07 -08001054{
1055 struct dst_entry *dst = NULL;
1056 int err;
1057
David S. Miller4c9483b2011-03-12 16:22:43 -05001058 err = ip6_dst_lookup_tail(sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001059 if (err)
1060 return ERR_PTR(err);
1061 if (final_dst)
David S. Miller4c9483b2011-03-12 16:22:43 -05001062 ipv6_addr_copy(&fl6->daddr, final_dst);
David S. Miller2774c132011-03-01 14:59:04 -08001063 if (can_sleep)
David S. Miller4c9483b2011-03-12 16:22:43 -05001064 fl6->flowi6_flags |= FLOWI_FLAG_CAN_SLEEP;
David S. Miller2774c132011-03-01 14:59:04 -08001065
David S. Miller4c9483b2011-03-12 16:22:43 -05001066 return xfrm_lookup(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001067}
1068EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow);
1069
1070/**
1071 * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow
1072 * @sk: socket which provides the dst cache and route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001073 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001074 * @final_dst: final destination address for ipsec lookup
David S. Millera1414712011-03-01 14:32:04 -08001075 * @can_sleep: we are in a sleepable context
Herbert Xu497c6152006-07-30 20:19:33 -07001076 *
1077 * This function performs a route lookup on the given flow with the
1078 * possibility of using the cached route in the socket if it is valid.
1079 * It will take the socket dst lock when operating on the dst cache.
1080 * As a result, this function can only be used in process context.
1081 *
David S. Miller68d0c6d2011-03-01 13:19:07 -08001082 * It returns a valid dst pointer on success, or a pointer encoded
1083 * error code.
Herbert Xu497c6152006-07-30 20:19:33 -07001084 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001085struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
David S. Miller68d0c6d2011-03-01 13:19:07 -08001086 const struct in6_addr *final_dst,
David S. Millera1414712011-03-01 14:32:04 -08001087 bool can_sleep)
Herbert Xu497c6152006-07-30 20:19:33 -07001088{
David S. Miller68d0c6d2011-03-01 13:19:07 -08001089 struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
1090 int err;
Herbert Xu497c6152006-07-30 20:19:33 -07001091
David S. Miller4c9483b2011-03-12 16:22:43 -05001092 dst = ip6_sk_dst_check(sk, dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001093
David S. Miller4c9483b2011-03-12 16:22:43 -05001094 err = ip6_dst_lookup_tail(sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001095 if (err)
1096 return ERR_PTR(err);
1097 if (final_dst)
David S. Miller4c9483b2011-03-12 16:22:43 -05001098 ipv6_addr_copy(&fl6->daddr, final_dst);
David S. Miller2774c132011-03-01 14:59:04 -08001099 if (can_sleep)
David S. Miller4c9483b2011-03-12 16:22:43 -05001100 fl6->flowi6_flags |= FLOWI_FLAG_CAN_SLEEP;
David S. Miller2774c132011-03-01 14:59:04 -08001101
David S. Miller4c9483b2011-03-12 16:22:43 -05001102 return xfrm_lookup(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
Herbert Xu497c6152006-07-30 20:19:33 -07001103}
David S. Miller68d0c6d2011-03-01 13:19:07 -08001104EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
Herbert Xu497c6152006-07-30 20:19:33 -07001105
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001106static inline int ip6_ufo_append_data(struct sock *sk,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001107 int getfrag(void *from, char *to, int offset, int len,
1108 int odd, struct sk_buff *skb),
1109 void *from, int length, int hh_len, int fragheaderlen,
Eric Dumazet87c48fa2011-07-21 21:25:58 -07001110 int transhdrlen, int mtu,unsigned int flags,
1111 struct rt6_info *rt)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001112
1113{
1114 struct sk_buff *skb;
1115 int err;
1116
1117 /* There is support for UDP large send offload by network
1118 * device, so create one single skb packet containing complete
1119 * udp datagram
1120 */
1121 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) {
1122 skb = sock_alloc_send_skb(sk,
1123 hh_len + fragheaderlen + transhdrlen + 20,
1124 (flags & MSG_DONTWAIT), &err);
1125 if (skb == NULL)
Zheng Yan504744e2011-10-28 00:26:00 -04001126 return err;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001127
1128 /* reserve space for Hardware header */
1129 skb_reserve(skb, hh_len);
1130
1131 /* create space for UDP/IP header */
1132 skb_put(skb,fragheaderlen + transhdrlen);
1133
1134 /* initialize network header pointer */
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001135 skb_reset_network_header(skb);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001136
1137 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001138 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001139
Patrick McHardy84fa7932006-08-29 16:44:56 -07001140 skb->ip_summed = CHECKSUM_PARTIAL;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001141 skb->csum = 0;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001142 }
1143
1144 err = skb_append_datato_frags(sk,skb, getfrag, from,
1145 (length - transhdrlen));
1146 if (!err) {
1147 struct frag_hdr fhdr;
1148
Sridhar Samudralac31d5322009-07-09 08:09:58 +00001149 /* Specify the length of each IPv6 datagram fragment.
1150 * It has to be a multiple of 8.
1151 */
1152 skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
1153 sizeof(struct frag_hdr)) & ~7;
Herbert Xuf83ef8c2006-06-30 13:37:03 -07001154 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Eric Dumazet87c48fa2011-07-21 21:25:58 -07001155 ipv6_select_ident(&fhdr, rt);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001156 skb_shinfo(skb)->ip6_frag_id = fhdr.identification;
1157 __skb_queue_tail(&sk->sk_write_queue, skb);
1158
1159 return 0;
1160 }
1161 /* There is not enough support do UPD LSO,
1162 * so follow normal path
1163 */
1164 kfree_skb(skb);
1165
1166 return err;
1167}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168
Herbert Xu0178b692009-02-05 15:15:50 -08001169static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
1170 gfp_t gfp)
1171{
1172 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1173}
1174
1175static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
1176 gfp_t gfp)
1177{
1178 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1179}
1180
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001181int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
1182 int offset, int len, int odd, struct sk_buff *skb),
1183 void *from, int length, int transhdrlen,
David S. Miller4c9483b2011-03-12 16:22:43 -05001184 int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi6 *fl6,
Brian Haley13b52cd2010-04-23 11:26:08 +00001185 struct rt6_info *rt, unsigned int flags, int dontfrag)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186{
1187 struct inet_sock *inet = inet_sk(sk);
1188 struct ipv6_pinfo *np = inet6_sk(sk);
David S. Millerbdc712b2011-05-06 15:02:07 -07001189 struct inet_cork *cork;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190 struct sk_buff *skb;
1191 unsigned int maxfraglen, fragheaderlen;
1192 int exthdrlen;
Steffen Klassert299b0762011-10-11 01:43:33 +00001193 int dst_exthdrlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 int hh_len;
1195 int mtu;
1196 int copy;
1197 int err;
1198 int offset = 0;
1199 int csummode = CHECKSUM_NONE;
Anders Berggrena693e692011-02-28 12:32:11 -08001200 __u8 tx_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201
1202 if (flags&MSG_PROBE)
1203 return 0;
David S. Millerbdc712b2011-05-06 15:02:07 -07001204 cork = &inet->cork.base;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 if (skb_queue_empty(&sk->sk_write_queue)) {
1206 /*
1207 * setup for corking
1208 */
1209 if (opt) {
Herbert Xu0178b692009-02-05 15:15:50 -08001210 if (WARN_ON(np->cork.opt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 return -EINVAL;
Herbert Xu0178b692009-02-05 15:15:50 -08001212
1213 np->cork.opt = kmalloc(opt->tot_len, sk->sk_allocation);
1214 if (unlikely(np->cork.opt == NULL))
1215 return -ENOBUFS;
1216
1217 np->cork.opt->tot_len = opt->tot_len;
1218 np->cork.opt->opt_flen = opt->opt_flen;
1219 np->cork.opt->opt_nflen = opt->opt_nflen;
1220
1221 np->cork.opt->dst0opt = ip6_opt_dup(opt->dst0opt,
1222 sk->sk_allocation);
1223 if (opt->dst0opt && !np->cork.opt->dst0opt)
1224 return -ENOBUFS;
1225
1226 np->cork.opt->dst1opt = ip6_opt_dup(opt->dst1opt,
1227 sk->sk_allocation);
1228 if (opt->dst1opt && !np->cork.opt->dst1opt)
1229 return -ENOBUFS;
1230
1231 np->cork.opt->hopopt = ip6_opt_dup(opt->hopopt,
1232 sk->sk_allocation);
1233 if (opt->hopopt && !np->cork.opt->hopopt)
1234 return -ENOBUFS;
1235
1236 np->cork.opt->srcrt = ip6_rthdr_dup(opt->srcrt,
1237 sk->sk_allocation);
1238 if (opt->srcrt && !np->cork.opt->srcrt)
1239 return -ENOBUFS;
1240
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 /* need source address above miyazawa*/
1242 }
Changli Gaod8d1f302010-06-10 23:31:35 -07001243 dst_hold(&rt->dst);
David S. Millerbdc712b2011-05-06 15:02:07 -07001244 cork->dst = &rt->dst;
David S. Miller4c9483b2011-03-12 16:22:43 -05001245 inet->cork.fl.u.ip6 = *fl6;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 np->cork.hop_limit = hlimit;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001247 np->cork.tclass = tclass;
John Heffner628a5c52007-04-20 15:53:27 -07001248 mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ?
Steffen Klassert299b0762011-10-11 01:43:33 +00001249 rt->dst.dev->mtu : dst_mtu(&rt->dst);
Dave Jonesc7503602006-03-20 22:44:52 -08001250 if (np->frag_size < mtu) {
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -08001251 if (np->frag_size)
1252 mtu = np->frag_size;
1253 }
David S. Millerbdc712b2011-05-06 15:02:07 -07001254 cork->fragsize = mtu;
Changli Gaod8d1f302010-06-10 23:31:35 -07001255 if (dst_allfrag(rt->dst.path))
David S. Millerbdc712b2011-05-06 15:02:07 -07001256 cork->flags |= IPCORK_ALLFRAG;
1257 cork->length = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 sk->sk_sndmsg_page = NULL;
1259 sk->sk_sndmsg_off = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001260 exthdrlen = (opt ? opt->opt_flen : 0) - rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 length += exthdrlen;
1262 transhdrlen += exthdrlen;
Steffen Klassert299b0762011-10-11 01:43:33 +00001263 dst_exthdrlen = rt->dst.header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264 } else {
David S. Millerbdc712b2011-05-06 15:02:07 -07001265 rt = (struct rt6_info *)cork->dst;
David S. Miller4c9483b2011-03-12 16:22:43 -05001266 fl6 = &inet->cork.fl.u.ip6;
Herbert Xu0178b692009-02-05 15:15:50 -08001267 opt = np->cork.opt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 transhdrlen = 0;
1269 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001270 dst_exthdrlen = 0;
David S. Millerbdc712b2011-05-06 15:02:07 -07001271 mtu = cork->fragsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272 }
1273
Changli Gaod8d1f302010-06-10 23:31:35 -07001274 hh_len = LL_RESERVED_SPACE(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001276 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001277 (opt ? opt->opt_nflen : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr);
1279
1280 if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
David S. Millerbdc712b2011-05-06 15:02:07 -07001281 if (cork->length + length > sizeof(struct ipv6hdr) + IPV6_MAXPLEN - fragheaderlen) {
David S. Miller4c9483b2011-03-12 16:22:43 -05001282 ipv6_local_error(sk, EMSGSIZE, fl6, mtu-exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 return -EMSGSIZE;
1284 }
1285 }
1286
Anders Berggrena693e692011-02-28 12:32:11 -08001287 /* For UDP, check if TX timestamp is enabled */
1288 if (sk->sk_type == SOCK_DGRAM) {
1289 err = sock_tx_timestamp(sk, &tx_flags);
1290 if (err)
1291 goto error;
1292 }
1293
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294 /*
1295 * Let's try using as much space as possible.
1296 * Use MTU if total length of the message fits into the MTU.
1297 * Otherwise, we need to reserve fragment header and
1298 * fragment alignment (= 8-15 octects, in total).
1299 *
1300 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001301 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302 * the message.
1303 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001304 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 * at once if non-fragmentable extension headers
1306 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001307 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308 */
1309
David S. Millerbdc712b2011-05-06 15:02:07 -07001310 cork->length += length;
Brian Haley4b340ae2010-04-23 11:26:09 +00001311 if (length > mtu) {
1312 int proto = sk->sk_protocol;
1313 if (dontfrag && (proto == IPPROTO_UDP || proto == IPPROTO_RAW)){
David S. Miller4c9483b2011-03-12 16:22:43 -05001314 ipv6_local_rxpmtu(sk, fl6, mtu-exthdrlen);
Brian Haley4b340ae2010-04-23 11:26:09 +00001315 return -EMSGSIZE;
1316 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001317
Brian Haley4b340ae2010-04-23 11:26:09 +00001318 if (proto == IPPROTO_UDP &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001319 (rt->dst.dev->features & NETIF_F_UFO)) {
Brian Haley4b340ae2010-04-23 11:26:09 +00001320
1321 err = ip6_ufo_append_data(sk, getfrag, from, length,
1322 hh_len, fragheaderlen,
Eric Dumazet87c48fa2011-07-21 21:25:58 -07001323 transhdrlen, mtu, flags, rt);
Brian Haley4b340ae2010-04-23 11:26:09 +00001324 if (err)
1325 goto error;
1326 return 0;
1327 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001328 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329
1330 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL)
1331 goto alloc_new_skb;
1332
1333 while (length > 0) {
1334 /* Check if the remaining data fits into current packet. */
David S. Millerbdc712b2011-05-06 15:02:07 -07001335 copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 if (copy < length)
1337 copy = maxfraglen - skb->len;
1338
1339 if (copy <= 0) {
1340 char *data;
1341 unsigned int datalen;
1342 unsigned int fraglen;
1343 unsigned int fraggap;
1344 unsigned int alloclen;
1345 struct sk_buff *skb_prev;
1346alloc_new_skb:
1347 skb_prev = skb;
1348
1349 /* There's no room in the current skb */
1350 if (skb_prev)
1351 fraggap = skb_prev->len - maxfraglen;
1352 else
1353 fraggap = 0;
1354
1355 /*
1356 * If remaining data exceeds the mtu,
1357 * we know we need more fragment(s).
1358 */
1359 datalen = length + fraggap;
David S. Millerbdc712b2011-05-06 15:02:07 -07001360 if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 datalen = maxfraglen - fragheaderlen;
1362
1363 fraglen = datalen + fragheaderlen;
1364 if ((flags & MSG_MORE) &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001365 !(rt->dst.dev->features&NETIF_F_SG))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366 alloclen = mtu;
1367 else
1368 alloclen = datalen + fragheaderlen;
1369
Steffen Klassert299b0762011-10-11 01:43:33 +00001370 alloclen += dst_exthdrlen;
1371
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 /*
1373 * The last fragment gets additional space at tail.
1374 * Note: we overallocate on fragments with MSG_MODE
1375 * because we have no idea if we're the last one.
1376 */
1377 if (datalen == length + fraggap)
Changli Gaod8d1f302010-06-10 23:31:35 -07001378 alloclen += rt->dst.trailer_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379
1380 /*
1381 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001382 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 * (without MSG_MORE) fits into the MTU.
1384 */
1385 alloclen += sizeof(struct frag_hdr);
1386
1387 if (transhdrlen) {
1388 skb = sock_alloc_send_skb(sk,
1389 alloclen + hh_len,
1390 (flags & MSG_DONTWAIT), &err);
1391 } else {
1392 skb = NULL;
1393 if (atomic_read(&sk->sk_wmem_alloc) <=
1394 2 * sk->sk_sndbuf)
1395 skb = sock_wmalloc(sk,
1396 alloclen + hh_len, 1,
1397 sk->sk_allocation);
1398 if (unlikely(skb == NULL))
1399 err = -ENOBUFS;
Anders Berggrena693e692011-02-28 12:32:11 -08001400 else {
1401 /* Only the initial fragment
1402 * is time stamped.
1403 */
1404 tx_flags = 0;
1405 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 }
1407 if (skb == NULL)
1408 goto error;
1409 /*
1410 * Fill in the control structures
1411 */
1412 skb->ip_summed = csummode;
1413 skb->csum = 0;
1414 /* reserve for fragmentation */
1415 skb_reserve(skb, hh_len+sizeof(struct frag_hdr));
1416
Anders Berggrena693e692011-02-28 12:32:11 -08001417 if (sk->sk_type == SOCK_DGRAM)
1418 skb_shinfo(skb)->tx_flags = tx_flags;
1419
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 /*
1421 * Find where to start putting bytes
1422 */
Steffen Klassert299b0762011-10-11 01:43:33 +00001423 data = skb_put(skb, fraglen + dst_exthdrlen);
1424 skb_set_network_header(skb, exthdrlen + dst_exthdrlen);
1425 data += fragheaderlen + dst_exthdrlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001426 skb->transport_header = (skb->network_header +
1427 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 if (fraggap) {
1429 skb->csum = skb_copy_and_csum_bits(
1430 skb_prev, maxfraglen,
1431 data + transhdrlen, fraggap, 0);
1432 skb_prev->csum = csum_sub(skb_prev->csum,
1433 skb->csum);
1434 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001435 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436 }
1437 copy = datalen - transhdrlen - fraggap;
Steffen Klassert299b0762011-10-11 01:43:33 +00001438
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 if (copy < 0) {
1440 err = -EINVAL;
1441 kfree_skb(skb);
1442 goto error;
1443 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1444 err = -EFAULT;
1445 kfree_skb(skb);
1446 goto error;
1447 }
1448
1449 offset += copy;
1450 length -= datalen - fraggap;
1451 transhdrlen = 0;
1452 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001453 dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 csummode = CHECKSUM_NONE;
1455
1456 /*
1457 * Put the packet on the pending queue
1458 */
1459 __skb_queue_tail(&sk->sk_write_queue, skb);
1460 continue;
1461 }
1462
1463 if (copy > length)
1464 copy = length;
1465
Changli Gaod8d1f302010-06-10 23:31:35 -07001466 if (!(rt->dst.dev->features&NETIF_F_SG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 unsigned int off;
1468
1469 off = skb->len;
1470 if (getfrag(from, skb_put(skb, copy),
1471 offset, copy, off, skb) < 0) {
1472 __skb_trim(skb, off);
1473 err = -EFAULT;
1474 goto error;
1475 }
1476 } else {
1477 int i = skb_shinfo(skb)->nr_frags;
1478 skb_frag_t *frag = &skb_shinfo(skb)->frags[i-1];
1479 struct page *page = sk->sk_sndmsg_page;
1480 int off = sk->sk_sndmsg_off;
1481 unsigned int left;
1482
1483 if (page && (left = PAGE_SIZE - off) > 0) {
1484 if (copy >= left)
1485 copy = left;
Ian Campbell408dadf2011-08-22 23:45:00 +00001486 if (page != skb_frag_page(frag)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 if (i == MAX_SKB_FRAGS) {
1488 err = -EMSGSIZE;
1489 goto error;
1490 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 skb_fill_page_desc(skb, i, page, sk->sk_sndmsg_off, 0);
Ian Campbell408dadf2011-08-22 23:45:00 +00001492 skb_frag_ref(skb, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 frag = &skb_shinfo(skb)->frags[i];
1494 }
1495 } else if(i < MAX_SKB_FRAGS) {
1496 if (copy > PAGE_SIZE)
1497 copy = PAGE_SIZE;
1498 page = alloc_pages(sk->sk_allocation, 0);
1499 if (page == NULL) {
1500 err = -ENOMEM;
1501 goto error;
1502 }
1503 sk->sk_sndmsg_page = page;
1504 sk->sk_sndmsg_off = 0;
1505
1506 skb_fill_page_desc(skb, i, page, 0, 0);
1507 frag = &skb_shinfo(skb)->frags[i];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508 } else {
1509 err = -EMSGSIZE;
1510 goto error;
1511 }
Eric Dumazet9e903e02011-10-18 21:00:24 +00001512 if (getfrag(from,
1513 skb_frag_address(frag) + skb_frag_size(frag),
Ian Campbell408dadf2011-08-22 23:45:00 +00001514 offset, copy, skb->len, skb) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515 err = -EFAULT;
1516 goto error;
1517 }
1518 sk->sk_sndmsg_off += copy;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001519 skb_frag_size_add(frag, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 skb->len += copy;
1521 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001522 skb->truesize += copy;
1523 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 }
1525 offset += copy;
1526 length -= copy;
1527 }
1528 return 0;
1529error:
David S. Millerbdc712b2011-05-06 15:02:07 -07001530 cork->length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001531 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 return err;
1533}
1534
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001535static void ip6_cork_release(struct inet_sock *inet, struct ipv6_pinfo *np)
1536{
Herbert Xu0178b692009-02-05 15:15:50 -08001537 if (np->cork.opt) {
1538 kfree(np->cork.opt->dst0opt);
1539 kfree(np->cork.opt->dst1opt);
1540 kfree(np->cork.opt->hopopt);
1541 kfree(np->cork.opt->srcrt);
1542 kfree(np->cork.opt);
1543 np->cork.opt = NULL;
1544 }
1545
David S. Millerbdc712b2011-05-06 15:02:07 -07001546 if (inet->cork.base.dst) {
1547 dst_release(inet->cork.base.dst);
1548 inet->cork.base.dst = NULL;
1549 inet->cork.base.flags &= ~IPCORK_ALLFRAG;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001550 }
1551 memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
1552}
1553
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554int ip6_push_pending_frames(struct sock *sk)
1555{
1556 struct sk_buff *skb, *tmp_skb;
1557 struct sk_buff **tail_skb;
1558 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
1559 struct inet_sock *inet = inet_sk(sk);
1560 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001561 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562 struct ipv6hdr *hdr;
1563 struct ipv6_txoptions *opt = np->cork.opt;
David S. Millerbdc712b2011-05-06 15:02:07 -07001564 struct rt6_info *rt = (struct rt6_info *)inet->cork.base.dst;
David S. Miller4c9483b2011-03-12 16:22:43 -05001565 struct flowi6 *fl6 = &inet->cork.fl.u.ip6;
1566 unsigned char proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 int err = 0;
1568
1569 if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL)
1570 goto out;
1571 tail_skb = &(skb_shinfo(skb)->frag_list);
1572
1573 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001574 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001575 __skb_pull(skb, skb_network_offset(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001577 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578 *tail_skb = tmp_skb;
1579 tail_skb = &(tmp_skb->next);
1580 skb->len += tmp_skb->len;
1581 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 skb->truesize += tmp_skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 tmp_skb->destructor = NULL;
1584 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585 }
1586
Herbert Xu28a894532008-02-12 18:07:27 -08001587 /* Allow local fragmentation. */
Herbert Xub5c15fc2008-02-14 23:49:37 -08001588 if (np->pmtudisc < IPV6_PMTUDISC_DO)
Herbert Xu28a894532008-02-12 18:07:27 -08001589 skb->local_df = 1;
1590
David S. Miller4c9483b2011-03-12 16:22:43 -05001591 ipv6_addr_copy(final_dst, &fl6->daddr);
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001592 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 if (opt && opt->opt_flen)
1594 ipv6_push_frag_opts(skb, opt, &proto);
1595 if (opt && opt->opt_nflen)
1596 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1597
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001598 skb_push(skb, sizeof(struct ipv6hdr));
1599 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001600 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001601
David S. Miller4c9483b2011-03-12 16:22:43 -05001602 *(__be32*)hdr = fl6->flowlabel |
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001603 htonl(0x60000000 | ((int)np->cork.tclass << 20));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 hdr->hop_limit = np->cork.hop_limit;
1606 hdr->nexthdr = proto;
David S. Miller4c9483b2011-03-12 16:22:43 -05001607 ipv6_addr_copy(&hdr->saddr, &fl6->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 ipv6_addr_copy(&hdr->daddr, final_dst);
1609
Patrick McHardya2c20642006-01-08 22:37:26 -08001610 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001611 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001612
Changli Gaod8d1f302010-06-10 23:31:35 -07001613 skb_dst_set(skb, dst_clone(&rt->dst));
Neil Hormanedf391f2009-04-27 02:45:02 -07001614 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
David L Stevens14878f72007-09-16 16:52:35 -07001615 if (proto == IPPROTO_ICMPV6) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001616 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
David L Stevens14878f72007-09-16 16:52:35 -07001617
Denis V. Lunev5a57d4c2008-10-08 10:34:14 -07001618 ICMP6MSGOUT_INC_STATS_BH(net, idev, icmp6_hdr(skb)->icmp6_type);
Denis V. Luneve41b5362008-10-08 10:33:26 -07001619 ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001620 }
1621
Herbert Xuef76bc22008-01-11 19:15:08 -08001622 err = ip6_local_out(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 if (err) {
1624 if (err > 0)
Eric Dumazet6ce9e7b2009-09-02 18:05:33 -07001625 err = net_xmit_errno(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626 if (err)
1627 goto error;
1628 }
1629
1630out:
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001631 ip6_cork_release(inet, np);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632 return err;
1633error:
Eric Dumazet06254912009-09-01 18:37:16 -07001634 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 goto out;
1636}
1637
1638void ip6_flush_pending_frames(struct sock *sk)
1639{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 struct sk_buff *skb;
1641
1642 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001643 if (skb_dst(skb))
1644 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001645 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 kfree_skb(skb);
1647 }
1648
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001649 ip6_cork_release(inet_sk(sk), inet6_sk(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650}