blob: 6e3ddf806ec20abcfd4952dfe97d2b521138e055 [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
Herbert Xuef76bc22008-01-11 19:15:08 -080059int __ip6_local_out(struct sk_buff *skb)
60{
61 int len;
62
63 len = skb->len - sizeof(struct ipv6hdr);
64 if (len > IPV6_MAXPLEN)
65 len = 0;
66 ipv6_hdr(skb)->payload_len = htons(len);
67
Jan Engelhardtb2e0b382010-03-23 04:09:07 +010068 return nf_hook(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
69 skb_dst(skb)->dev, dst_output);
Herbert Xuef76bc22008-01-11 19:15:08 -080070}
71
72int ip6_local_out(struct sk_buff *skb)
73{
74 int err;
75
76 err = __ip6_local_out(skb);
77 if (likely(err == 1))
78 err = dst_output(skb);
79
80 return err;
81}
82EXPORT_SYMBOL_GPL(ip6_local_out);
83
Jan Engelhardt9e508492010-04-13 15:28:11 +020084static int ip6_finish_output2(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -070085{
Eric Dumazetadf30902009-06-02 05:19:30 +000086 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070087 struct net_device *dev = dst->dev;
David S. Millerf6b72b62011-07-14 07:53:20 -070088 struct neighbour *neigh;
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +000089 struct in6_addr *nexthop;
90 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
92 skb->protocol = htons(ETH_P_IPV6);
93 skb->dev = dev;
94
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070095 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) {
Eric Dumazetadf30902009-06-02 05:19:30 +000096 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -070097
Octavian Purdila7ad68482010-01-06 20:37:01 -080098 if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(skb->sk) &&
Patrick McHardyd1db2752010-05-11 14:40:55 +020099 ((mroute6_socket(dev_net(dev), skb) &&
Benjamin Therybd91b8b2008-12-10 16:07:08 -0800100 !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) ||
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900101 ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
102 &ipv6_hdr(skb)->saddr))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
104
105 /* Do not check for IFF_ALLMULTI; multicast routing
106 is not supported in any case.
107 */
108 if (newskb)
Jan Engelhardtb2e0b382010-03-23 04:09:07 +0100109 NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING,
110 newskb, NULL, newskb->dev,
Michel Machado95603e22012-06-12 10:16:35 +0000111 dev_loopback_xmit);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700113 if (ipv6_hdr(skb)->hop_limit == 0) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700114 IP6_INC_STATS(dev_net(dev), idev,
115 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116 kfree_skb(skb);
117 return 0;
118 }
119 }
120
Neil Hormanedf391f2009-04-27 02:45:02 -0700121 IP6_UPD_PO_STATS(dev_net(dev), idev, IPSTATS_MIB_OUTMCAST,
122 skb->len);
Hannes Frederic Sowadd408512013-02-10 02:33:35 +0000123
124 if (IPV6_ADDR_MC_SCOPE(&ipv6_hdr(skb)->daddr) <=
125 IPV6_ADDR_SCOPE_NODELOCAL &&
126 !(dev->flags & IFF_LOOPBACK)) {
127 kfree_skb(skb);
128 return 0;
129 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 }
131
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000132 rcu_read_lock_bh();
133 nexthop = rt6_nexthop((struct rt6_info *)dst, &ipv6_hdr(skb)->daddr);
134 neigh = __ipv6_neigh_lookup_noref(dst->dev, nexthop);
135 if (unlikely(!neigh))
136 neigh = __neigh_create(&nd_tbl, nexthop, dst->dev, false);
137 if (!IS_ERR(neigh)) {
138 ret = dst_neigh_output(dst, neigh, skb);
139 rcu_read_unlock_bh();
140 return ret;
141 }
142 rcu_read_unlock_bh();
David S. Miller05e3aa02011-07-16 17:26:00 -0700143
Jan Engelhardt9e508492010-04-13 15:28:11 +0200144 IP6_INC_STATS_BH(dev_net(dst->dev),
145 ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
146 kfree_skb(skb);
147 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148}
149
Jan Engelhardt9e508492010-04-13 15:28:11 +0200150static int ip6_finish_output(struct sk_buff *skb)
151{
152 if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
153 dst_allfrag(skb_dst(skb)))
154 return ip6_fragment(skb, ip6_finish_output2);
155 else
156 return ip6_finish_output2(skb);
157}
158
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159int ip6_output(struct sk_buff *skb)
160{
Jan Engelhardt9e508492010-04-13 15:28:11 +0200161 struct net_device *dev = skb_dst(skb)->dev;
Eric Dumazetadf30902009-06-02 05:19:30 +0000162 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900163 if (unlikely(idev->cnf.disable_ipv6)) {
Jan Engelhardt9e508492010-04-13 15:28:11 +0200164 IP6_INC_STATS(dev_net(dev), idev,
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700165 IPSTATS_MIB_OUTDISCARDS);
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900166 kfree_skb(skb);
167 return 0;
168 }
169
Jan Engelhardt9c6eb282010-04-13 15:32:16 +0200170 return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING, skb, NULL, dev,
171 ip6_finish_output,
172 !(IP6CB(skb)->flags & IP6SKB_REROUTED));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173}
174
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175/*
Shan Weib5d43992010-04-15 16:48:48 +0000176 * xmit an sk_buff (used by TCP, SCTP and DCCP)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 */
178
David S. Miller4c9483b2011-03-12 16:22:43 -0500179int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
Eric Dumazetb903d322011-10-27 00:44:35 -0400180 struct ipv6_txoptions *opt, int tclass)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181{
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700182 struct net *net = sock_net(sk);
Patrick McHardyb30bd282006-03-23 01:17:25 -0800183 struct ipv6_pinfo *np = inet6_sk(sk);
David S. Miller4c9483b2011-03-12 16:22:43 -0500184 struct in6_addr *first_hop = &fl6->daddr;
Eric Dumazetadf30902009-06-02 05:19:30 +0000185 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 struct ipv6hdr *hdr;
David S. Miller4c9483b2011-03-12 16:22:43 -0500187 u8 proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 int seg_len = skb->len;
Gerrit Renkere651f032009-08-09 08:12:48 +0000189 int hlimit = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 u32 mtu;
191
192 if (opt) {
Chuck Leverc2636b42007-10-23 21:07:32 -0700193 unsigned int head_room;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194
195 /* First: exthdrs may take lots of space (~8K for now)
196 MAX_HEADER is not enough.
197 */
198 head_room = opt->opt_nflen + opt->opt_flen;
199 seg_len += head_room;
200 head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
201
202 if (skb_headroom(skb) < head_room) {
203 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900204 if (skb2 == NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000205 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900206 IPSTATS_MIB_OUTDISCARDS);
207 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 return -ENOBUFS;
209 }
Eric Dumazet808db802012-04-24 10:17:59 +0000210 consume_skb(skb);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900211 skb = skb2;
Dan Carpenter83d7eb22010-04-30 16:42:08 -0700212 skb_set_owner_w(skb, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 }
214 if (opt->opt_flen)
215 ipv6_push_frag_opts(skb, opt, &proto);
216 if (opt->opt_nflen)
217 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
218 }
219
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700220 skb_push(skb, sizeof(struct ipv6hdr));
221 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700222 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223
224 /*
225 * Fill in the IPv6 header
226 */
Eric Dumazetb903d322011-10-27 00:44:35 -0400227 if (np)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 hlimit = np->hop_limit;
229 if (hlimit < 0)
YOSHIFUJI Hideaki6b75d092008-03-10 06:00:30 -0400230 hlimit = ip6_dst_hoplimit(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231
YOSHIFUJI Hideaki / 吉藤英明3e4e4c12013-01-13 05:01:39 +0000232 ip6_flow_hdr(hdr, tclass, fl6->flowlabel);
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900233
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 hdr->payload_len = htons(seg_len);
235 hdr->nexthdr = proto;
236 hdr->hop_limit = hlimit;
237
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000238 hdr->saddr = fl6->saddr;
239 hdr->daddr = *first_hop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240
Patrick McHardya2c20642006-01-08 22:37:26 -0800241 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800242 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -0800243
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 mtu = dst_mtu(dst);
Wei Yongjun283d07ac2008-08-03 21:15:59 -0700245 if ((skb->len <= mtu) || skb->local_df || skb_is_gso(skb)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000246 IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)),
Neil Hormanedf391f2009-04-27 02:45:02 -0700247 IPSTATS_MIB_OUT, skb->len);
Jan Engelhardtb2e0b382010-03-23 04:09:07 +0100248 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
249 dst->dev, dst_output);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 }
251
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 skb->dev = dst->dev;
Steffen Klassertf4e53e22013-02-05 01:48:05 +0000253 ipv6_local_error(sk, EMSGSIZE, fl6, mtu);
Eric Dumazetadf30902009-06-02 05:19:30 +0000254 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 kfree_skb(skb);
256 return -EMSGSIZE;
257}
258
YOSHIFUJI Hideaki71590392007-02-22 22:05:40 +0900259EXPORT_SYMBOL(ip6_xmit);
260
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
262{
263 struct ip6_ra_chain *ra;
264 struct sock *last = NULL;
265
266 read_lock(&ip6_ra_lock);
267 for (ra = ip6_ra_chain; ra; ra = ra->next) {
268 struct sock *sk = ra->sk;
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700269 if (sk && ra->sel == sel &&
270 (!sk->sk_bound_dev_if ||
271 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272 if (last) {
273 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
274 if (skb2)
275 rawv6_rcv(last, skb2);
276 }
277 last = sk;
278 }
279 }
280
281 if (last) {
282 rawv6_rcv(last, skb);
283 read_unlock(&ip6_ra_lock);
284 return 1;
285 }
286 read_unlock(&ip6_ra_lock);
287 return 0;
288}
289
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700290static int ip6_forward_proxy_check(struct sk_buff *skb)
291{
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700292 struct ipv6hdr *hdr = ipv6_hdr(skb);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700293 u8 nexthdr = hdr->nexthdr;
Jesse Gross75f28112011-11-30 17:05:51 -0800294 __be16 frag_off;
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700295 int offset;
296
297 if (ipv6_ext_hdr(nexthdr)) {
Jesse Gross75f28112011-11-30 17:05:51 -0800298 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700299 if (offset < 0)
300 return 0;
301 } else
302 offset = sizeof(struct ipv6hdr);
303
304 if (nexthdr == IPPROTO_ICMPV6) {
305 struct icmp6hdr *icmp6;
306
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700307 if (!pskb_may_pull(skb, (skb_network_header(skb) +
308 offset + 1 - skb->data)))
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700309 return 0;
310
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700311 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700312
313 switch (icmp6->icmp6_type) {
314 case NDISC_ROUTER_SOLICITATION:
315 case NDISC_ROUTER_ADVERTISEMENT:
316 case NDISC_NEIGHBOUR_SOLICITATION:
317 case NDISC_NEIGHBOUR_ADVERTISEMENT:
318 case NDISC_REDIRECT:
319 /* For reaction involving unicast neighbor discovery
320 * message destined to the proxied address, pass it to
321 * input function.
322 */
323 return 1;
324 default:
325 break;
326 }
327 }
328
Ville Nuorvala74553b02006-09-22 14:42:18 -0700329 /*
330 * The proxying router can't forward traffic sent to a link-local
331 * address, so signal the sender and discard the packet. This
332 * behavior is clarified by the MIPv6 specification.
333 */
334 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
335 dst_link_failure(skb);
336 return -1;
337 }
338
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700339 return 0;
340}
341
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342static inline int ip6_forward_finish(struct sk_buff *skb)
343{
344 return dst_output(skb);
345}
346
347int ip6_forward(struct sk_buff *skb)
348{
Eric Dumazetadf30902009-06-02 05:19:30 +0000349 struct dst_entry *dst = skb_dst(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700350 struct ipv6hdr *hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 struct inet6_skb_parm *opt = IP6CB(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900352 struct net *net = dev_net(dst->dev);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800353 u32 mtu;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900354
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700355 if (net->ipv6.devconf_all->forwarding == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 goto error;
357
Ben Hutchings4497b072008-06-19 16:22:28 -0700358 if (skb_warn_if_lro(skb))
359 goto drop;
360
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700362 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 goto drop;
364 }
365
Alexey Kuznetsov72b43d02011-01-12 08:34:08 +0000366 if (skb->pkt_type != PACKET_HOST)
367 goto drop;
368
Herbert Xu35fc92a2007-03-26 23:22:20 -0700369 skb_forward_csum(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370
371 /*
372 * We DO NOT make any processing on
373 * RA packets, pushing them to user level AS IS
374 * without ane WARRANTY that application will be able
375 * to interpret them. The reason is that we
376 * cannot make anything clever here.
377 *
378 * We are not end-node, so that if packet contains
379 * AH/ESP, we cannot make anything.
380 * Defragmentation also would be mistake, RA packets
381 * cannot be fragmented, because there is no warranty
382 * that different fragments will go along one path. --ANK
383 */
YOSHIFUJI Hideaki / 吉藤英明ab4eb352013-06-22 11:13:13 +0900384 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
385 if (ip6_call_ra_chain(skb, ntohs(opt->ra)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 return 0;
387 }
388
389 /*
390 * check and decrement ttl
391 */
392 if (hdr->hop_limit <= 1) {
393 /* Force OUTPUT device used as source address */
394 skb->dev = dst->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000395 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0);
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700396 IP6_INC_STATS_BH(net,
397 ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398
399 kfree_skb(skb);
400 return -ETIMEDOUT;
401 }
402
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -0700403 /* XXX: idev->cnf.proxy_ndp? */
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700404 if (net->ipv6.devconf_all->proxy_ndp &&
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800405 pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
Ville Nuorvala74553b02006-09-22 14:42:18 -0700406 int proxied = ip6_forward_proxy_check(skb);
407 if (proxied > 0)
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700408 return ip6_input(skb);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700409 else if (proxied < 0) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700410 IP6_INC_STATS(net, ip6_dst_idev(dst),
411 IPSTATS_MIB_INDISCARDS);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700412 goto drop;
413 }
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700414 }
415
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 if (!xfrm6_route_forward(skb)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700417 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 goto drop;
419 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000420 dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421
422 /* IPv6 specs say nothing about it, but it is clear that we cannot
423 send redirects to source routed frames.
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900424 We don't send redirects to frames decapsulated from IPsec.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 */
David S. Millerc45a3df2012-01-27 15:32:19 -0800426 if (skb->dev == dst->dev && opt->srcrt == 0 && !skb_sec_path(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 struct in6_addr *target = NULL;
David S. Millerfbfe95a2012-06-08 23:24:18 -0700428 struct inet_peer *peer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430
431 /*
432 * incoming and outgoing devices are the same
433 * send a redirect.
434 */
435
436 rt = (struct rt6_info *) dst;
David S. Millerc45a3df2012-01-27 15:32:19 -0800437 if (rt->rt6i_flags & RTF_GATEWAY)
438 target = &rt->rt6i_gateway;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 else
440 target = &hdr->daddr;
441
David S. Miller1d861aa2012-07-10 03:58:16 -0700442 peer = inet_getpeer_v6(net->ipv6.peers, &rt->rt6i_dst.addr, 1);
David S. Miller92d86822011-02-04 15:55:25 -0800443
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 /* Limit redirects both by destination (here)
445 and by source (inside ndisc_send_redirect)
446 */
David S. Millerfbfe95a2012-06-08 23:24:18 -0700447 if (inet_peer_xrlim_allow(peer, 1*HZ))
David S. Miller49919692012-01-27 15:30:48 -0800448 ndisc_send_redirect(skb, target);
David S. Miller1d861aa2012-07-10 03:58:16 -0700449 if (peer)
450 inet_putpeer(peer);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700451 } else {
452 int addrtype = ipv6_addr_type(&hdr->saddr);
453
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 /* This check is security critical. */
YOSHIFUJI Hideakif81b2e72008-06-25 16:55:26 +0900455 if (addrtype == IPV6_ADDR_ANY ||
456 addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK))
David L Stevens5bb1ab02007-05-09 13:53:44 -0700457 goto error;
458 if (addrtype & IPV6_ADDR_LINKLOCAL) {
459 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000460 ICMPV6_NOT_NEIGHBOUR, 0);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700461 goto error;
462 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 }
464
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800465 mtu = dst_mtu(dst);
466 if (mtu < IPV6_MIN_MTU)
467 mtu = IPV6_MIN_MTU;
468
Patrick McHardy4cdd34082012-08-26 19:13:58 +0200469 if ((!skb->local_df && skb->len > mtu && !skb_is_gso(skb)) ||
470 (IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 /* Again, force OUTPUT device used as source address */
472 skb->dev = dst->dev;
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800473 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700474 IP6_INC_STATS_BH(net,
475 ip6_dst_idev(dst), IPSTATS_MIB_INTOOBIGERRORS);
476 IP6_INC_STATS_BH(net,
477 ip6_dst_idev(dst), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 kfree_skb(skb);
479 return -EMSGSIZE;
480 }
481
482 if (skb_cow(skb, dst->dev->hard_header_len)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700483 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 goto drop;
485 }
486
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700487 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488
489 /* Mangling hops number delayed to point after skb COW */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900490
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 hdr->hop_limit--;
492
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700493 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
Vincent Bernat2d8dbb02012-06-05 03:41:42 +0000494 IP6_ADD_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len);
Jan Engelhardtb2e0b382010-03-23 04:09:07 +0100495 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dst->dev,
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800496 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497
498error:
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700499 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500drop:
501 kfree_skb(skb);
502 return -EINVAL;
503}
504
505static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
506{
507 to->pkt_type = from->pkt_type;
508 to->priority = from->priority;
509 to->protocol = from->protocol;
Eric Dumazetadf30902009-06-02 05:19:30 +0000510 skb_dst_drop(to);
511 skb_dst_set(to, dst_clone(skb_dst(from)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 to->dev = from->dev;
Thomas Graf82e91ff2006-11-09 15:19:14 -0800513 to->mark = from->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514
515#ifdef CONFIG_NET_SCHED
516 to->tc_index = from->tc_index;
517#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -0700518 nf_copy(to, from);
Amerigo Wang07a93622012-10-29 16:23:10 +0000519#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700520 to->nf_trace = from->nf_trace;
521#endif
James Morris984bc162006-06-09 00:29:17 -0700522 skb_copy_secmark(to, from);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523}
524
David Stevensad0081e2010-12-17 11:42:42 +0000525int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 struct sk_buff *frag;
Eric Dumazetadf30902009-06-02 05:19:30 +0000528 struct rt6_info *rt = (struct rt6_info*)skb_dst(skb);
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800529 struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 struct ipv6hdr *tmp_hdr;
531 struct frag_hdr *fh;
532 unsigned int mtu, hlen, left, len;
Herbert Xua7ae1992011-11-18 02:20:04 +0000533 int hroom, troom;
Al Viroae08e1f2006-11-08 00:27:11 -0800534 __be32 frag_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 int ptr, offset = 0, err=0;
536 u8 *prevhdr, nexthdr = 0;
Eric Dumazetadf30902009-06-02 05:19:30 +0000537 struct net *net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 hlen = ip6_find_1stfragopt(skb, &prevhdr);
540 nexthdr = *prevhdr;
541
John Heffner628a5c52007-04-20 15:53:27 -0700542 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700543
544 /* We must not fragment if the socket is set to force MTU discovery
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800545 * or if the skb it not generated by a local socket.
John Heffnerb881ef72007-04-20 15:52:39 -0700546 */
Patrick McHardy4cdd34082012-08-26 19:13:58 +0200547 if (unlikely(!skb->local_df && skb->len > mtu) ||
548 (IP6CB(skb)->frag_max_size &&
549 IP6CB(skb)->frag_max_size > mtu)) {
Eric Dumazeta34a1012012-05-18 21:51:44 +0000550 if (skb->sk && dst_allfrag(skb_dst(skb)))
551 sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK);
552
Eric Dumazetadf30902009-06-02 05:19:30 +0000553 skb->dev = skb_dst(skb)->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000554 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazetadf30902009-06-02 05:19:30 +0000555 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700556 IPSTATS_MIB_FRAGFAILS);
John Heffnerb881ef72007-04-20 15:52:39 -0700557 kfree_skb(skb);
558 return -EMSGSIZE;
559 }
560
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800561 if (np && np->frag_size < mtu) {
562 if (np->frag_size)
563 mtu = np->frag_size;
564 }
565 mtu -= hlen + sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566
David S. Miller21dc3302010-08-23 00:13:46 -0700567 if (skb_has_frag_list(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 int first_len = skb_pagelen(skb);
Eric Dumazet3d130082010-09-21 08:47:45 +0000569 struct sk_buff *frag2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570
571 if (first_len - hlen > mtu ||
572 ((first_len - hlen) & 7) ||
573 skb_cloned(skb))
574 goto slow_path;
575
David S. Miller4d9092b2009-06-09 00:20:05 -0700576 skb_walk_frags(skb, frag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 /* Correct geometry. */
578 if (frag->len > mtu ||
579 ((frag->len & 7) && frag->next) ||
580 skb_headroom(frag) < hlen)
Eric Dumazet3d130082010-09-21 08:47:45 +0000581 goto slow_path_clean;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 /* Partially cloned skb? */
584 if (skb_shared(frag))
Eric Dumazet3d130082010-09-21 08:47:45 +0000585 goto slow_path_clean;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700586
587 BUG_ON(frag->sk);
588 if (skb->sk) {
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700589 frag->sk = skb->sk;
590 frag->destructor = sock_wfree;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700591 }
Eric Dumazet3d130082010-09-21 08:47:45 +0000592 skb->truesize -= frag->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 }
594
595 err = 0;
596 offset = 0;
597 frag = skb_shinfo(skb)->frag_list;
David S. Miller4d9092b2009-06-09 00:20:05 -0700598 skb_frag_list_init(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 /* BUILD HEADER */
600
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800601 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700602 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 if (!tmp_hdr) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000604 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700605 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 return -ENOMEM;
607 }
608
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 __skb_pull(skb, hlen);
610 fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700611 __skb_push(skb, hlen);
612 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700613 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614
Eric Dumazet87c48fa2011-07-21 21:25:58 -0700615 ipv6_select_ident(fh, rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 fh->nexthdr = nexthdr;
617 fh->reserved = 0;
618 fh->frag_off = htons(IP6_MF);
619 frag_id = fh->identification;
620
621 first_len = skb_pagelen(skb);
622 skb->data_len = first_len - skb_headlen(skb);
623 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700624 ipv6_hdr(skb)->payload_len = htons(first_len -
625 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900626
Changli Gaod8d1f302010-06-10 23:31:35 -0700627 dst_hold(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628
629 for (;;) {
630 /* Prepare header of the next frame,
631 * before previous one went down. */
632 if (frag) {
633 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300634 skb_reset_transport_header(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700636 __skb_push(frag, hlen);
637 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700638 memcpy(skb_network_header(frag), tmp_hdr,
639 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 offset += skb->len - hlen - sizeof(struct frag_hdr);
641 fh->nexthdr = nexthdr;
642 fh->reserved = 0;
643 fh->frag_off = htons(offset);
644 if (frag->next != NULL)
645 fh->frag_off |= htons(IP6_MF);
646 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700647 ipv6_hdr(frag)->payload_len =
648 htons(frag->len -
649 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 ip6_copy_metadata(frag, skb);
651 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900652
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 err = output(skb);
Wei Dongdafee492006-08-02 13:41:21 -0700654 if(!err)
Changli Gaod8d1f302010-06-10 23:31:35 -0700655 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700656 IPSTATS_MIB_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700657
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 if (err || !frag)
659 break;
660
661 skb = frag;
662 frag = skb->next;
663 skb->next = NULL;
664 }
665
Jesper Juhla51482b2005-11-08 09:41:34 -0800666 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
668 if (err == 0) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700669 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700670 IPSTATS_MIB_FRAGOKS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000671 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 return 0;
673 }
674
675 while (frag) {
676 skb = frag->next;
677 kfree_skb(frag);
678 frag = skb;
679 }
680
Changli Gaod8d1f302010-06-10 23:31:35 -0700681 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700682 IPSTATS_MIB_FRAGFAILS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000683 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 return err;
Eric Dumazet3d130082010-09-21 08:47:45 +0000685
686slow_path_clean:
687 skb_walk_frags(skb, frag2) {
688 if (frag2 == frag)
689 break;
690 frag2->sk = NULL;
691 frag2->destructor = NULL;
692 skb->truesize += frag2->truesize;
693 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 }
695
696slow_path:
Eric Dumazet72e843b2012-05-18 11:02:15 +0000697 if ((skb->ip_summed == CHECKSUM_PARTIAL) &&
698 skb_checksum_help(skb))
699 goto fail;
700
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 left = skb->len - hlen; /* Space per frame */
702 ptr = hlen; /* Where to start from */
703
704 /*
705 * Fragment the datagram.
706 */
707
708 *prevhdr = NEXTHDR_FRAGMENT;
Herbert Xua7ae1992011-11-18 02:20:04 +0000709 hroom = LL_RESERVED_SPACE(rt->dst.dev);
710 troom = rt->dst.dev->needed_tailroom;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711
712 /*
713 * Keep copying data until we run out.
714 */
715 while(left > 0) {
716 len = left;
717 /* IF: it doesn't fit, use 'mtu' - the data space left */
718 if (len > mtu)
719 len = mtu;
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300720 /* IF: we are not sending up to and including the packet end
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 then align the next start on an eight byte boundary */
722 if (len < left) {
723 len &= ~7;
724 }
725 /*
726 * Allocate buffer.
727 */
728
Herbert Xua7ae1992011-11-18 02:20:04 +0000729 if ((frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) +
730 hroom + troom, GFP_ATOMIC)) == NULL) {
Patrick McHardy64ce2072005-08-09 20:50:53 -0700731 NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n");
Eric Dumazetadf30902009-06-02 05:19:30 +0000732 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900733 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 err = -ENOMEM;
735 goto fail;
736 }
737
738 /*
739 * Set up data on packet
740 */
741
742 ip6_copy_metadata(frag, skb);
Herbert Xua7ae1992011-11-18 02:20:04 +0000743 skb_reserve(frag, hroom);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700745 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300746 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700747 frag->transport_header = (frag->network_header + hlen +
748 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749
750 /*
751 * Charge the memory for the fragment to any owner
752 * it might possess
753 */
754 if (skb->sk)
755 skb_set_owner_w(frag, skb->sk);
756
757 /*
758 * Copy the packet header into the new buffer.
759 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300760 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
762 /*
763 * Build fragment header.
764 */
765 fh->nexthdr = nexthdr;
766 fh->reserved = 0;
Yan Zhengf36d6ab2005-10-03 14:19:15 -0700767 if (!frag_id) {
Eric Dumazet87c48fa2011-07-21 21:25:58 -0700768 ipv6_select_ident(fh, rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 frag_id = fh->identification;
770 } else
771 fh->identification = frag_id;
772
773 /*
774 * Copy a block of the IP datagram.
775 */
Wei Yongjun8984e412007-08-21 20:59:08 -0700776 if (skb_copy_bits(skb, ptr, skb_transport_header(frag), len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 BUG();
778 left -= len;
779
780 fh->frag_off = htons(offset);
781 if (left > 0)
782 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700783 ipv6_hdr(frag)->payload_len = htons(frag->len -
784 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785
786 ptr += len;
787 offset += len;
788
789 /*
790 * Put this fragment into the sending queue.
791 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 err = output(frag);
793 if (err)
794 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700795
Eric Dumazetadf30902009-06-02 05:19:30 +0000796 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700797 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000799 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900800 IPSTATS_MIB_FRAGOKS);
Eric Dumazet808db802012-04-24 10:17:59 +0000801 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 return err;
803
804fail:
Eric Dumazetadf30902009-06-02 05:19:30 +0000805 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900806 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900807 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 return err;
809}
810
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000811static inline int ip6_rt_check(const struct rt6key *rt_key,
812 const struct in6_addr *fl_addr,
813 const struct in6_addr *addr_cache)
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700814{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000815 return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
816 (addr_cache == NULL || !ipv6_addr_equal(fl_addr, addr_cache));
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700817}
818
Herbert Xu497c6152006-07-30 20:19:33 -0700819static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
820 struct dst_entry *dst,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000821 const struct flowi6 *fl6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822{
Herbert Xu497c6152006-07-30 20:19:33 -0700823 struct ipv6_pinfo *np = inet6_sk(sk);
Eric Dumazeta963a372013-06-26 04:15:07 -0700824 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825
Herbert Xu497c6152006-07-30 20:19:33 -0700826 if (!dst)
827 goto out;
828
Eric Dumazeta963a372013-06-26 04:15:07 -0700829 if (dst->ops->family != AF_INET6) {
830 dst_release(dst);
831 return NULL;
832 }
833
834 rt = (struct rt6_info *)dst;
Herbert Xu497c6152006-07-30 20:19:33 -0700835 /* Yes, checking route validity in not connected
836 * case is not very simple. Take into account,
837 * that we do not support routing by source, TOS,
838 * and MSG_DONTROUTE --ANK (980726)
839 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700840 * 1. ip6_rt_check(): If route was host route,
841 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700842 * If it is network route, we still may
843 * check its validity using saved pointer
844 * to the last used address: daddr_cache.
845 * We do not want to save whole address now,
846 * (because main consumer of this service
847 * is tcp, which has not this problem),
848 * so that the last trick works only on connected
849 * sockets.
850 * 2. oif also should be the same.
851 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500852 if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700853#ifdef CONFIG_IPV6_SUBTREES
David S. Miller4c9483b2011-03-12 16:22:43 -0500854 ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700855#endif
David S. Miller4c9483b2011-03-12 16:22:43 -0500856 (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex)) {
Herbert Xu497c6152006-07-30 20:19:33 -0700857 dst_release(dst);
858 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 }
860
Herbert Xu497c6152006-07-30 20:19:33 -0700861out:
862 return dst;
863}
864
865static int ip6_dst_lookup_tail(struct sock *sk,
David S. Miller4c9483b2011-03-12 16:22:43 -0500866 struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700867{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900868 struct net *net = sock_net(sk);
David S. Miller69cce1d2011-07-17 23:09:49 -0700869#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
870 struct neighbour *n;
David S. Miller97cac082012-07-02 22:43:47 -0700871 struct rt6_info *rt;
David S. Miller69cce1d2011-07-17 23:09:49 -0700872#endif
873 int err;
Herbert Xu497c6152006-07-30 20:19:33 -0700874
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 if (*dst == NULL)
David S. Miller4c9483b2011-03-12 16:22:43 -0500876 *dst = ip6_route_output(net, sk, fl6);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877
878 if ((err = (*dst)->error))
879 goto out_err_release;
880
David S. Miller4c9483b2011-03-12 16:22:43 -0500881 if (ipv6_addr_any(&fl6->saddr)) {
Daniel Walterc3968a82011-04-13 21:10:57 +0000882 struct rt6_info *rt = (struct rt6_info *) *dst;
883 err = ip6_route_get_saddr(net, rt, &fl6->daddr,
884 sk ? inet6_sk(sk)->srcprefs : 0,
885 &fl6->saddr);
Olaf Hering44456d32005-07-27 11:45:17 -0700886 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 goto out_err_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 }
889
Neil Horman95c385b2007-04-25 17:08:10 -0700890#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700891 /*
892 * Here if the dst entry we've looked up
893 * has a neighbour entry that is in the INCOMPLETE
894 * state and the src address from the flow is
895 * marked as OPTIMISTIC, we release the found
896 * dst entry and replace it instead with the
897 * dst entry of the nexthop router
898 */
Eric Dumazetc56bf6f2012-07-06 09:19:05 +0200899 rt = (struct rt6_info *) *dst;
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000900 rcu_read_lock_bh();
901 n = __ipv6_neigh_lookup_noref(rt->dst.dev, rt6_nexthop(rt, &fl6->daddr));
902 err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0;
903 rcu_read_unlock_bh();
904
905 if (err) {
Neil Hormane550dfb2008-09-09 13:51:35 -0700906 struct inet6_ifaddr *ifp;
David S. Miller4c9483b2011-03-12 16:22:43 -0500907 struct flowi6 fl_gw6;
Neil Hormane550dfb2008-09-09 13:51:35 -0700908 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -0700909
David S. Miller4c9483b2011-03-12 16:22:43 -0500910 ifp = ipv6_get_ifaddr(net, &fl6->saddr,
Neil Hormane550dfb2008-09-09 13:51:35 -0700911 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -0700912
Neil Hormane550dfb2008-09-09 13:51:35 -0700913 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
914 if (ifp)
915 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -0700916
Neil Hormane550dfb2008-09-09 13:51:35 -0700917 if (redirect) {
918 /*
919 * We need to get the dst entry for the
920 * default router instead
921 */
922 dst_release(*dst);
David S. Miller4c9483b2011-03-12 16:22:43 -0500923 memcpy(&fl_gw6, fl6, sizeof(struct flowi6));
924 memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr));
925 *dst = ip6_route_output(net, sk, &fl_gw6);
Neil Hormane550dfb2008-09-09 13:51:35 -0700926 if ((err = (*dst)->error))
927 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -0700928 }
Neil Hormane550dfb2008-09-09 13:51:35 -0700929 }
Neil Horman95c385b2007-04-25 17:08:10 -0700930#endif
931
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 return 0;
933
934out_err_release:
Mitsuru Chinenca46f9c2007-12-05 22:31:47 -0800935 if (err == -ENETUNREACH)
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700936 IP6_INC_STATS_BH(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 dst_release(*dst);
938 *dst = NULL;
939 return err;
940}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -0800941
Herbert Xu497c6152006-07-30 20:19:33 -0700942/**
943 * ip6_dst_lookup - perform route lookup on flow
944 * @sk: socket which provides route info
945 * @dst: pointer to dst_entry * for result
David S. Miller4c9483b2011-03-12 16:22:43 -0500946 * @fl6: flow to lookup
Herbert Xu497c6152006-07-30 20:19:33 -0700947 *
948 * This function performs a route lookup on the given flow.
949 *
950 * It returns zero on success, or a standard errno code on error.
951 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500952int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700953{
954 *dst = NULL;
David S. Miller4c9483b2011-03-12 16:22:43 -0500955 return ip6_dst_lookup_tail(sk, dst, fl6);
Herbert Xu497c6152006-07-30 20:19:33 -0700956}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -0800957EXPORT_SYMBOL_GPL(ip6_dst_lookup);
958
Herbert Xu497c6152006-07-30 20:19:33 -0700959/**
David S. Miller68d0c6d2011-03-01 13:19:07 -0800960 * ip6_dst_lookup_flow - perform route lookup on flow with ipsec
961 * @sk: socket which provides route info
David S. Miller4c9483b2011-03-12 16:22:43 -0500962 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -0800963 * @final_dst: final destination address for ipsec lookup
David S. Millera1414712011-03-01 14:32:04 -0800964 * @can_sleep: we are in a sleepable context
David S. Miller68d0c6d2011-03-01 13:19:07 -0800965 *
966 * This function performs a route lookup on the given flow.
967 *
968 * It returns a valid dst pointer on success, or a pointer encoded
969 * error code.
970 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500971struct dst_entry *ip6_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
David S. Miller68d0c6d2011-03-01 13:19:07 -0800972 const struct in6_addr *final_dst,
David S. Millera1414712011-03-01 14:32:04 -0800973 bool can_sleep)
David S. Miller68d0c6d2011-03-01 13:19:07 -0800974{
975 struct dst_entry *dst = NULL;
976 int err;
977
David S. Miller4c9483b2011-03-12 16:22:43 -0500978 err = ip6_dst_lookup_tail(sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -0800979 if (err)
980 return ERR_PTR(err);
981 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000982 fl6->daddr = *final_dst;
David S. Miller2774c132011-03-01 14:59:04 -0800983 if (can_sleep)
David S. Miller4c9483b2011-03-12 16:22:43 -0500984 fl6->flowi6_flags |= FLOWI_FLAG_CAN_SLEEP;
David S. Miller2774c132011-03-01 14:59:04 -0800985
David S. Miller4c9483b2011-03-12 16:22:43 -0500986 return xfrm_lookup(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
David S. Miller68d0c6d2011-03-01 13:19:07 -0800987}
988EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow);
989
990/**
991 * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow
992 * @sk: socket which provides the dst cache and route info
David S. Miller4c9483b2011-03-12 16:22:43 -0500993 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -0800994 * @final_dst: final destination address for ipsec lookup
David S. Millera1414712011-03-01 14:32:04 -0800995 * @can_sleep: we are in a sleepable context
Herbert Xu497c6152006-07-30 20:19:33 -0700996 *
997 * This function performs a route lookup on the given flow with the
998 * possibility of using the cached route in the socket if it is valid.
999 * It will take the socket dst lock when operating on the dst cache.
1000 * As a result, this function can only be used in process context.
1001 *
David S. Miller68d0c6d2011-03-01 13:19:07 -08001002 * It returns a valid dst pointer on success, or a pointer encoded
1003 * error code.
Herbert Xu497c6152006-07-30 20:19:33 -07001004 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001005struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
David S. Miller68d0c6d2011-03-01 13:19:07 -08001006 const struct in6_addr *final_dst,
David S. Millera1414712011-03-01 14:32:04 -08001007 bool can_sleep)
Herbert Xu497c6152006-07-30 20:19:33 -07001008{
David S. Miller68d0c6d2011-03-01 13:19:07 -08001009 struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
1010 int err;
Herbert Xu497c6152006-07-30 20:19:33 -07001011
David S. Miller4c9483b2011-03-12 16:22:43 -05001012 dst = ip6_sk_dst_check(sk, dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001013
David S. Miller4c9483b2011-03-12 16:22:43 -05001014 err = ip6_dst_lookup_tail(sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001015 if (err)
1016 return ERR_PTR(err);
1017 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001018 fl6->daddr = *final_dst;
David S. Miller2774c132011-03-01 14:59:04 -08001019 if (can_sleep)
David S. Miller4c9483b2011-03-12 16:22:43 -05001020 fl6->flowi6_flags |= FLOWI_FLAG_CAN_SLEEP;
David S. Miller2774c132011-03-01 14:59:04 -08001021
David S. Miller4c9483b2011-03-12 16:22:43 -05001022 return xfrm_lookup(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
Herbert Xu497c6152006-07-30 20:19:33 -07001023}
David S. Miller68d0c6d2011-03-01 13:19:07 -08001024EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
Herbert Xu497c6152006-07-30 20:19:33 -07001025
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001026static inline int ip6_ufo_append_data(struct sock *sk,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001027 int getfrag(void *from, char *to, int offset, int len,
1028 int odd, struct sk_buff *skb),
1029 void *from, int length, int hh_len, int fragheaderlen,
Eric Dumazet87c48fa2011-07-21 21:25:58 -07001030 int transhdrlen, int mtu,unsigned int flags,
1031 struct rt6_info *rt)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001032
1033{
1034 struct sk_buff *skb;
1035 int err;
1036
1037 /* There is support for UDP large send offload by network
1038 * device, so create one single skb packet containing complete
1039 * udp datagram
1040 */
1041 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) {
1042 skb = sock_alloc_send_skb(sk,
1043 hh_len + fragheaderlen + transhdrlen + 20,
1044 (flags & MSG_DONTWAIT), &err);
1045 if (skb == NULL)
Zheng Yan504744e2011-10-28 00:26:00 -04001046 return err;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001047
1048 /* reserve space for Hardware header */
1049 skb_reserve(skb, hh_len);
1050
1051 /* create space for UDP/IP header */
1052 skb_put(skb,fragheaderlen + transhdrlen);
1053
1054 /* initialize network header pointer */
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001055 skb_reset_network_header(skb);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001056
1057 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001058 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001059
Patrick McHardy84fa7932006-08-29 16:44:56 -07001060 skb->ip_summed = CHECKSUM_PARTIAL;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001061 skb->csum = 0;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001062 }
1063
1064 err = skb_append_datato_frags(sk,skb, getfrag, from,
1065 (length - transhdrlen));
1066 if (!err) {
1067 struct frag_hdr fhdr;
1068
Sridhar Samudralac31d5322009-07-09 08:09:58 +00001069 /* Specify the length of each IPv6 datagram fragment.
1070 * It has to be a multiple of 8.
1071 */
1072 skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
1073 sizeof(struct frag_hdr)) & ~7;
Herbert Xuf83ef8c2006-06-30 13:37:03 -07001074 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Eric Dumazet87c48fa2011-07-21 21:25:58 -07001075 ipv6_select_ident(&fhdr, rt);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001076 skb_shinfo(skb)->ip6_frag_id = fhdr.identification;
1077 __skb_queue_tail(&sk->sk_write_queue, skb);
1078
1079 return 0;
1080 }
1081 /* There is not enough support do UPD LSO,
1082 * so follow normal path
1083 */
1084 kfree_skb(skb);
1085
1086 return err;
1087}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088
Herbert Xu0178b692009-02-05 15:15:50 -08001089static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
1090 gfp_t gfp)
1091{
1092 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1093}
1094
1095static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
1096 gfp_t gfp)
1097{
1098 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1099}
1100
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001101static void ip6_append_data_mtu(unsigned int *mtu,
Gao feng0c183372012-05-26 01:30:53 +00001102 int *maxfraglen,
1103 unsigned int fragheaderlen,
1104 struct sk_buff *skb,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001105 struct rt6_info *rt,
1106 bool pmtuprobe)
Gao feng0c183372012-05-26 01:30:53 +00001107{
1108 if (!(rt->dst.flags & DST_XFRM_TUNNEL)) {
1109 if (skb == NULL) {
1110 /* first fragment, reserve header_len */
1111 *mtu = *mtu - rt->dst.header_len;
1112
1113 } else {
1114 /*
1115 * this fragment is not first, the headers
1116 * space is regarded as data space.
1117 */
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001118 *mtu = min(*mtu, pmtuprobe ?
1119 rt->dst.dev->mtu :
1120 dst_mtu(rt->dst.path));
Gao feng0c183372012-05-26 01:30:53 +00001121 }
1122 *maxfraglen = ((*mtu - fragheaderlen) & ~7)
1123 + fragheaderlen - sizeof(struct frag_hdr);
1124 }
1125}
1126
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001127int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
1128 int offset, int len, int odd, struct sk_buff *skb),
1129 void *from, int length, int transhdrlen,
David S. Miller4c9483b2011-03-12 16:22:43 -05001130 int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi6 *fl6,
Brian Haley13b52cd2010-04-23 11:26:08 +00001131 struct rt6_info *rt, unsigned int flags, int dontfrag)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132{
1133 struct inet_sock *inet = inet_sk(sk);
1134 struct ipv6_pinfo *np = inet6_sk(sk);
David S. Millerbdc712b2011-05-06 15:02:07 -07001135 struct inet_cork *cork;
Gao feng0c183372012-05-26 01:30:53 +00001136 struct sk_buff *skb, *skb_prev = NULL;
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001137 unsigned int maxfraglen, fragheaderlen, mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 int exthdrlen;
Steffen Klassert299b0762011-10-11 01:43:33 +00001139 int dst_exthdrlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 int hh_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 int copy;
1142 int err;
1143 int offset = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001144 __u8 tx_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145
1146 if (flags&MSG_PROBE)
1147 return 0;
David S. Millerbdc712b2011-05-06 15:02:07 -07001148 cork = &inet->cork.base;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 if (skb_queue_empty(&sk->sk_write_queue)) {
1150 /*
1151 * setup for corking
1152 */
1153 if (opt) {
Herbert Xu0178b692009-02-05 15:15:50 -08001154 if (WARN_ON(np->cork.opt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155 return -EINVAL;
Herbert Xu0178b692009-02-05 15:15:50 -08001156
Eric Dumazet284041e2013-05-17 04:53:13 +00001157 np->cork.opt = kzalloc(opt->tot_len, sk->sk_allocation);
Herbert Xu0178b692009-02-05 15:15:50 -08001158 if (unlikely(np->cork.opt == NULL))
1159 return -ENOBUFS;
1160
1161 np->cork.opt->tot_len = opt->tot_len;
1162 np->cork.opt->opt_flen = opt->opt_flen;
1163 np->cork.opt->opt_nflen = opt->opt_nflen;
1164
1165 np->cork.opt->dst0opt = ip6_opt_dup(opt->dst0opt,
1166 sk->sk_allocation);
1167 if (opt->dst0opt && !np->cork.opt->dst0opt)
1168 return -ENOBUFS;
1169
1170 np->cork.opt->dst1opt = ip6_opt_dup(opt->dst1opt,
1171 sk->sk_allocation);
1172 if (opt->dst1opt && !np->cork.opt->dst1opt)
1173 return -ENOBUFS;
1174
1175 np->cork.opt->hopopt = ip6_opt_dup(opt->hopopt,
1176 sk->sk_allocation);
1177 if (opt->hopopt && !np->cork.opt->hopopt)
1178 return -ENOBUFS;
1179
1180 np->cork.opt->srcrt = ip6_rthdr_dup(opt->srcrt,
1181 sk->sk_allocation);
1182 if (opt->srcrt && !np->cork.opt->srcrt)
1183 return -ENOBUFS;
1184
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 /* need source address above miyazawa*/
1186 }
Changli Gaod8d1f302010-06-10 23:31:35 -07001187 dst_hold(&rt->dst);
David S. Millerbdc712b2011-05-06 15:02:07 -07001188 cork->dst = &rt->dst;
David S. Miller4c9483b2011-03-12 16:22:43 -05001189 inet->cork.fl.u.ip6 = *fl6;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190 np->cork.hop_limit = hlimit;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001191 np->cork.tclass = tclass;
Gao feng0c183372012-05-26 01:30:53 +00001192 if (rt->dst.flags & DST_XFRM_TUNNEL)
1193 mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ?
1194 rt->dst.dev->mtu : dst_mtu(&rt->dst);
1195 else
1196 mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ?
1197 rt->dst.dev->mtu : dst_mtu(rt->dst.path);
Dave Jonesc7503602006-03-20 22:44:52 -08001198 if (np->frag_size < mtu) {
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -08001199 if (np->frag_size)
1200 mtu = np->frag_size;
1201 }
David S. Millerbdc712b2011-05-06 15:02:07 -07001202 cork->fragsize = mtu;
Changli Gaod8d1f302010-06-10 23:31:35 -07001203 if (dst_allfrag(rt->dst.path))
David S. Millerbdc712b2011-05-06 15:02:07 -07001204 cork->flags |= IPCORK_ALLFRAG;
1205 cork->length = 0;
Romain KUNTZ7efdba52013-01-16 12:47:40 +00001206 exthdrlen = (opt ? opt->opt_flen : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207 length += exthdrlen;
1208 transhdrlen += exthdrlen;
Romain KUNTZ7efdba52013-01-16 12:47:40 +00001209 dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 } else {
David S. Millerbdc712b2011-05-06 15:02:07 -07001211 rt = (struct rt6_info *)cork->dst;
David S. Miller4c9483b2011-03-12 16:22:43 -05001212 fl6 = &inet->cork.fl.u.ip6;
Herbert Xu0178b692009-02-05 15:15:50 -08001213 opt = np->cork.opt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 transhdrlen = 0;
1215 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001216 dst_exthdrlen = 0;
David S. Millerbdc712b2011-05-06 15:02:07 -07001217 mtu = cork->fragsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 }
1219
Changli Gaod8d1f302010-06-10 23:31:35 -07001220 hh_len = LL_RESERVED_SPACE(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001222 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001223 (opt ? opt->opt_nflen : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr);
1225
1226 if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
David S. Millerbdc712b2011-05-06 15:02:07 -07001227 if (cork->length + length > sizeof(struct ipv6hdr) + IPV6_MAXPLEN - fragheaderlen) {
David S. Miller4c9483b2011-03-12 16:22:43 -05001228 ipv6_local_error(sk, EMSGSIZE, fl6, mtu-exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 return -EMSGSIZE;
1230 }
1231 }
1232
Anders Berggrena693e692011-02-28 12:32:11 -08001233 /* For UDP, check if TX timestamp is enabled */
Daniel Borkmannbf84a012013-04-14 08:08:13 +00001234 if (sk->sk_type == SOCK_DGRAM)
1235 sock_tx_timestamp(sk, &tx_flags);
Anders Berggrena693e692011-02-28 12:32:11 -08001236
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 /*
1238 * Let's try using as much space as possible.
1239 * Use MTU if total length of the message fits into the MTU.
1240 * Otherwise, we need to reserve fragment header and
1241 * fragment alignment (= 8-15 octects, in total).
1242 *
1243 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001244 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 * the message.
1246 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001247 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 * at once if non-fragmentable extension headers
1249 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001250 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 */
1252
David S. Millerbdc712b2011-05-06 15:02:07 -07001253 cork->length += length;
Brian Haley4b340ae2010-04-23 11:26:09 +00001254 if (length > mtu) {
1255 int proto = sk->sk_protocol;
1256 if (dontfrag && (proto == IPPROTO_UDP || proto == IPPROTO_RAW)){
David S. Miller4c9483b2011-03-12 16:22:43 -05001257 ipv6_local_rxpmtu(sk, fl6, mtu-exthdrlen);
Brian Haley4b340ae2010-04-23 11:26:09 +00001258 return -EMSGSIZE;
1259 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001260
Brian Haley4b340ae2010-04-23 11:26:09 +00001261 if (proto == IPPROTO_UDP &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001262 (rt->dst.dev->features & NETIF_F_UFO)) {
Brian Haley4b340ae2010-04-23 11:26:09 +00001263
1264 err = ip6_ufo_append_data(sk, getfrag, from, length,
1265 hh_len, fragheaderlen,
Eric Dumazet87c48fa2011-07-21 21:25:58 -07001266 transhdrlen, mtu, flags, rt);
Brian Haley4b340ae2010-04-23 11:26:09 +00001267 if (err)
1268 goto error;
1269 return 0;
1270 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001271 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272
1273 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL)
1274 goto alloc_new_skb;
1275
1276 while (length > 0) {
1277 /* Check if the remaining data fits into current packet. */
David S. Millerbdc712b2011-05-06 15:02:07 -07001278 copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 if (copy < length)
1280 copy = maxfraglen - skb->len;
1281
1282 if (copy <= 0) {
1283 char *data;
1284 unsigned int datalen;
1285 unsigned int fraglen;
1286 unsigned int fraggap;
1287 unsigned int alloclen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288alloc_new_skb:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 /* There's no room in the current skb */
Gao feng0c183372012-05-26 01:30:53 +00001290 if (skb)
1291 fraggap = skb->len - maxfraglen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 else
1293 fraggap = 0;
Gao feng0c183372012-05-26 01:30:53 +00001294 /* update mtu and maxfraglen if necessary */
1295 if (skb == NULL || skb_prev == NULL)
1296 ip6_append_data_mtu(&mtu, &maxfraglen,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001297 fragheaderlen, skb, rt,
1298 np->pmtudisc ==
1299 IPV6_PMTUDISC_PROBE);
Gao feng0c183372012-05-26 01:30:53 +00001300
1301 skb_prev = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302
1303 /*
1304 * If remaining data exceeds the mtu,
1305 * we know we need more fragment(s).
1306 */
1307 datalen = length + fraggap;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308
Gao feng0c183372012-05-26 01:30:53 +00001309 if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1310 datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 if ((flags & MSG_MORE) &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001312 !(rt->dst.dev->features&NETIF_F_SG))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313 alloclen = mtu;
1314 else
1315 alloclen = datalen + fragheaderlen;
1316
Steffen Klassert299b0762011-10-11 01:43:33 +00001317 alloclen += dst_exthdrlen;
1318
Gao feng0c183372012-05-26 01:30:53 +00001319 if (datalen != length + fraggap) {
1320 /*
1321 * this is not the last fragment, the trailer
1322 * space is regarded as data space.
1323 */
1324 datalen += rt->dst.trailer_len;
1325 }
1326
1327 alloclen += rt->dst.trailer_len;
1328 fraglen = datalen + fragheaderlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329
1330 /*
1331 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001332 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333 * (without MSG_MORE) fits into the MTU.
1334 */
1335 alloclen += sizeof(struct frag_hdr);
1336
1337 if (transhdrlen) {
1338 skb = sock_alloc_send_skb(sk,
1339 alloclen + hh_len,
1340 (flags & MSG_DONTWAIT), &err);
1341 } else {
1342 skb = NULL;
1343 if (atomic_read(&sk->sk_wmem_alloc) <=
1344 2 * sk->sk_sndbuf)
1345 skb = sock_wmalloc(sk,
1346 alloclen + hh_len, 1,
1347 sk->sk_allocation);
1348 if (unlikely(skb == NULL))
1349 err = -ENOBUFS;
Anders Berggrena693e692011-02-28 12:32:11 -08001350 else {
1351 /* Only the initial fragment
1352 * is time stamped.
1353 */
1354 tx_flags = 0;
1355 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 }
1357 if (skb == NULL)
1358 goto error;
1359 /*
1360 * Fill in the control structures
1361 */
Eric Dumazetd7f7c0a2012-05-18 05:37:56 +00001362 skb->ip_summed = CHECKSUM_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 skb->csum = 0;
Gao feng1f858512012-03-19 22:36:10 +00001364 /* reserve for fragmentation and ipsec header */
1365 skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
1366 dst_exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367
Anders Berggrena693e692011-02-28 12:32:11 -08001368 if (sk->sk_type == SOCK_DGRAM)
1369 skb_shinfo(skb)->tx_flags = tx_flags;
1370
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 /*
1372 * Find where to start putting bytes
1373 */
Gao feng1f858512012-03-19 22:36:10 +00001374 data = skb_put(skb, fraglen);
1375 skb_set_network_header(skb, exthdrlen);
1376 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001377 skb->transport_header = (skb->network_header +
1378 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 if (fraggap) {
1380 skb->csum = skb_copy_and_csum_bits(
1381 skb_prev, maxfraglen,
1382 data + transhdrlen, fraggap, 0);
1383 skb_prev->csum = csum_sub(skb_prev->csum,
1384 skb->csum);
1385 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001386 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 }
1388 copy = datalen - transhdrlen - fraggap;
Steffen Klassert299b0762011-10-11 01:43:33 +00001389
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 if (copy < 0) {
1391 err = -EINVAL;
1392 kfree_skb(skb);
1393 goto error;
1394 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1395 err = -EFAULT;
1396 kfree_skb(skb);
1397 goto error;
1398 }
1399
1400 offset += copy;
1401 length -= datalen - fraggap;
1402 transhdrlen = 0;
1403 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001404 dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405
1406 /*
1407 * Put the packet on the pending queue
1408 */
1409 __skb_queue_tail(&sk->sk_write_queue, skb);
1410 continue;
1411 }
1412
1413 if (copy > length)
1414 copy = length;
1415
Changli Gaod8d1f302010-06-10 23:31:35 -07001416 if (!(rt->dst.dev->features&NETIF_F_SG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417 unsigned int off;
1418
1419 off = skb->len;
1420 if (getfrag(from, skb_put(skb, copy),
1421 offset, copy, off, skb) < 0) {
1422 __skb_trim(skb, off);
1423 err = -EFAULT;
1424 goto error;
1425 }
1426 } else {
1427 int i = skb_shinfo(skb)->nr_frags;
Eric Dumazet5640f762012-09-23 23:04:42 +00001428 struct page_frag *pfrag = sk_page_frag(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429
Eric Dumazet5640f762012-09-23 23:04:42 +00001430 err = -ENOMEM;
1431 if (!sk_page_frag_refill(sk, pfrag))
1432 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433
Eric Dumazet5640f762012-09-23 23:04:42 +00001434 if (!skb_can_coalesce(skb, i, pfrag->page,
1435 pfrag->offset)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436 err = -EMSGSIZE;
Eric Dumazet5640f762012-09-23 23:04:42 +00001437 if (i == MAX_SKB_FRAGS)
1438 goto error;
1439
1440 __skb_fill_page_desc(skb, i, pfrag->page,
1441 pfrag->offset, 0);
1442 skb_shinfo(skb)->nr_frags = ++i;
1443 get_page(pfrag->page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001445 copy = min_t(int, copy, pfrag->size - pfrag->offset);
Eric Dumazet9e903e02011-10-18 21:00:24 +00001446 if (getfrag(from,
Eric Dumazet5640f762012-09-23 23:04:42 +00001447 page_address(pfrag->page) + pfrag->offset,
1448 offset, copy, skb->len, skb) < 0)
1449 goto error_efault;
1450
1451 pfrag->offset += copy;
1452 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 skb->len += copy;
1454 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001455 skb->truesize += copy;
1456 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 }
1458 offset += copy;
1459 length -= copy;
1460 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001461
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 return 0;
Eric Dumazet5640f762012-09-23 23:04:42 +00001463
1464error_efault:
1465 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466error:
David S. Millerbdc712b2011-05-06 15:02:07 -07001467 cork->length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001468 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 return err;
1470}
Chris Elstona495f832012-04-29 21:48:53 +00001471EXPORT_SYMBOL_GPL(ip6_append_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001473static void ip6_cork_release(struct inet_sock *inet, struct ipv6_pinfo *np)
1474{
Herbert Xu0178b692009-02-05 15:15:50 -08001475 if (np->cork.opt) {
1476 kfree(np->cork.opt->dst0opt);
1477 kfree(np->cork.opt->dst1opt);
1478 kfree(np->cork.opt->hopopt);
1479 kfree(np->cork.opt->srcrt);
1480 kfree(np->cork.opt);
1481 np->cork.opt = NULL;
1482 }
1483
David S. Millerbdc712b2011-05-06 15:02:07 -07001484 if (inet->cork.base.dst) {
1485 dst_release(inet->cork.base.dst);
1486 inet->cork.base.dst = NULL;
1487 inet->cork.base.flags &= ~IPCORK_ALLFRAG;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001488 }
1489 memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
1490}
1491
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492int ip6_push_pending_frames(struct sock *sk)
1493{
1494 struct sk_buff *skb, *tmp_skb;
1495 struct sk_buff **tail_skb;
1496 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
1497 struct inet_sock *inet = inet_sk(sk);
1498 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001499 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 struct ipv6hdr *hdr;
1501 struct ipv6_txoptions *opt = np->cork.opt;
David S. Millerbdc712b2011-05-06 15:02:07 -07001502 struct rt6_info *rt = (struct rt6_info *)inet->cork.base.dst;
David S. Miller4c9483b2011-03-12 16:22:43 -05001503 struct flowi6 *fl6 = &inet->cork.fl.u.ip6;
1504 unsigned char proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 int err = 0;
1506
1507 if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL)
1508 goto out;
1509 tail_skb = &(skb_shinfo(skb)->frag_list);
1510
1511 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001512 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001513 __skb_pull(skb, skb_network_offset(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001515 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 *tail_skb = tmp_skb;
1517 tail_skb = &(tmp_skb->next);
1518 skb->len += tmp_skb->len;
1519 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 skb->truesize += tmp_skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 tmp_skb->destructor = NULL;
1522 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 }
1524
Herbert Xu28a894532008-02-12 18:07:27 -08001525 /* Allow local fragmentation. */
Herbert Xub5c15fc2008-02-14 23:49:37 -08001526 if (np->pmtudisc < IPV6_PMTUDISC_DO)
Herbert Xu28a894532008-02-12 18:07:27 -08001527 skb->local_df = 1;
1528
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001529 *final_dst = fl6->daddr;
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001530 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 if (opt && opt->opt_flen)
1532 ipv6_push_frag_opts(skb, opt, &proto);
1533 if (opt && opt->opt_nflen)
1534 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1535
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001536 skb_push(skb, sizeof(struct ipv6hdr));
1537 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001538 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001539
YOSHIFUJI Hideaki / 吉藤英明3e4e4c12013-01-13 05:01:39 +00001540 ip6_flow_hdr(hdr, np->cork.tclass, fl6->flowlabel);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 hdr->hop_limit = np->cork.hop_limit;
1542 hdr->nexthdr = proto;
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001543 hdr->saddr = fl6->saddr;
1544 hdr->daddr = *final_dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545
Patrick McHardya2c20642006-01-08 22:37:26 -08001546 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001547 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001548
Changli Gaod8d1f302010-06-10 23:31:35 -07001549 skb_dst_set(skb, dst_clone(&rt->dst));
Neil Hormanedf391f2009-04-27 02:45:02 -07001550 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
David L Stevens14878f72007-09-16 16:52:35 -07001551 if (proto == IPPROTO_ICMPV6) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001552 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
David L Stevens14878f72007-09-16 16:52:35 -07001553
Denis V. Lunev5a57d4c2008-10-08 10:34:14 -07001554 ICMP6MSGOUT_INC_STATS_BH(net, idev, icmp6_hdr(skb)->icmp6_type);
Denis V. Luneve41b5362008-10-08 10:33:26 -07001555 ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001556 }
1557
Herbert Xuef76bc22008-01-11 19:15:08 -08001558 err = ip6_local_out(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559 if (err) {
1560 if (err > 0)
Eric Dumazet6ce9e7b2009-09-02 18:05:33 -07001561 err = net_xmit_errno(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562 if (err)
1563 goto error;
1564 }
1565
1566out:
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001567 ip6_cork_release(inet, np);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568 return err;
1569error:
Eric Dumazet06254912009-09-01 18:37:16 -07001570 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 goto out;
1572}
Chris Elstona495f832012-04-29 21:48:53 +00001573EXPORT_SYMBOL_GPL(ip6_push_pending_frames);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574
1575void ip6_flush_pending_frames(struct sock *sk)
1576{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 struct sk_buff *skb;
1578
1579 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001580 if (skb_dst(skb))
1581 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001582 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 kfree_skb(skb);
1584 }
1585
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001586 ip6_cork_release(inet_sk(sk), inet6_sk(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587}
Chris Elstona495f832012-04-29 21:48:53 +00001588EXPORT_SYMBOL_GPL(ip6_flush_pending_frames);