blob: ef02b26ccf812e57e794e9b151746600bcb9a8f0 [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
Jan Engelhardt9e508492010-04-13 15:28:11 +020059static int ip6_finish_output2(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -070060{
Eric Dumazetadf30902009-06-02 05:19:30 +000061 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070062 struct net_device *dev = dst->dev;
David S. Millerf6b72b62011-07-14 07:53:20 -070063 struct neighbour *neigh;
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +000064 struct in6_addr *nexthop;
65 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
67 skb->protocol = htons(ETH_P_IPV6);
68 skb->dev = dev;
69
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070070 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) {
Eric Dumazetadf30902009-06-02 05:19:30 +000071 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -070072
Octavian Purdila7ad68482010-01-06 20:37:01 -080073 if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(skb->sk) &&
Patrick McHardyd1db2752010-05-11 14:40:55 +020074 ((mroute6_socket(dev_net(dev), skb) &&
Benjamin Therybd91b8b2008-12-10 16:07:08 -080075 !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) ||
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090076 ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
77 &ipv6_hdr(skb)->saddr))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
79
80 /* Do not check for IFF_ALLMULTI; multicast routing
81 is not supported in any case.
82 */
83 if (newskb)
Jan Engelhardtb2e0b382010-03-23 04:09:07 +010084 NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING,
85 newskb, NULL, newskb->dev,
Michel Machado95603e22012-06-12 10:16:35 +000086 dev_loopback_xmit);
Linus Torvalds1da177e2005-04-16 15:20:36 -070087
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070088 if (ipv6_hdr(skb)->hop_limit == 0) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -070089 IP6_INC_STATS(dev_net(dev), idev,
90 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 kfree_skb(skb);
92 return 0;
93 }
94 }
95
Neil Hormanedf391f2009-04-27 02:45:02 -070096 IP6_UPD_PO_STATS(dev_net(dev), idev, IPSTATS_MIB_OUTMCAST,
97 skb->len);
Hannes Frederic Sowadd408512013-02-10 02:33:35 +000098
99 if (IPV6_ADDR_MC_SCOPE(&ipv6_hdr(skb)->daddr) <=
100 IPV6_ADDR_SCOPE_NODELOCAL &&
101 !(dev->flags & IFF_LOOPBACK)) {
102 kfree_skb(skb);
103 return 0;
104 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 }
106
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000107 rcu_read_lock_bh();
Julian Anastasov550bab42013-10-20 15:43:04 +0300108 nexthop = rt6_nexthop((struct rt6_info *)dst);
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000109 neigh = __ipv6_neigh_lookup_noref(dst->dev, nexthop);
110 if (unlikely(!neigh))
111 neigh = __neigh_create(&nd_tbl, nexthop, dst->dev, false);
112 if (!IS_ERR(neigh)) {
113 ret = dst_neigh_output(dst, neigh, skb);
114 rcu_read_unlock_bh();
115 return ret;
116 }
117 rcu_read_unlock_bh();
David S. Miller05e3aa02011-07-16 17:26:00 -0700118
Hannes Frederic Sowa7f88c6b2013-11-29 06:39:44 +0100119 IP6_INC_STATS(dev_net(dst->dev),
120 ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200121 kfree_skb(skb);
122 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123}
124
Jan Engelhardt9e508492010-04-13 15:28:11 +0200125static int ip6_finish_output(struct sk_buff *skb)
126{
127 if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
Jiri Pirko9037c352013-11-06 17:52:19 +0100128 dst_allfrag(skb_dst(skb)) ||
129 (IP6CB(skb)->frag_max_size && skb->len > IP6CB(skb)->frag_max_size))
Jan Engelhardt9e508492010-04-13 15:28:11 +0200130 return ip6_fragment(skb, ip6_finish_output2);
131 else
132 return ip6_finish_output2(skb);
133}
134
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135int ip6_output(struct sk_buff *skb)
136{
Jan Engelhardt9e508492010-04-13 15:28:11 +0200137 struct net_device *dev = skb_dst(skb)->dev;
Eric Dumazetadf30902009-06-02 05:19:30 +0000138 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900139 if (unlikely(idev->cnf.disable_ipv6)) {
Jan Engelhardt9e508492010-04-13 15:28:11 +0200140 IP6_INC_STATS(dev_net(dev), idev,
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700141 IPSTATS_MIB_OUTDISCARDS);
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900142 kfree_skb(skb);
143 return 0;
144 }
145
Jan Engelhardt9c6eb282010-04-13 15:32:16 +0200146 return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING, skb, NULL, dev,
147 ip6_finish_output,
148 !(IP6CB(skb)->flags & IP6SKB_REROUTED));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149}
150
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151/*
Shan Weib5d43992010-04-15 16:48:48 +0000152 * xmit an sk_buff (used by TCP, SCTP and DCCP)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 */
154
David S. Miller4c9483b2011-03-12 16:22:43 -0500155int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
Eric Dumazetb903d322011-10-27 00:44:35 -0400156 struct ipv6_txoptions *opt, int tclass)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157{
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700158 struct net *net = sock_net(sk);
Patrick McHardyb30bd282006-03-23 01:17:25 -0800159 struct ipv6_pinfo *np = inet6_sk(sk);
David S. Miller4c9483b2011-03-12 16:22:43 -0500160 struct in6_addr *first_hop = &fl6->daddr;
Eric Dumazetadf30902009-06-02 05:19:30 +0000161 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 struct ipv6hdr *hdr;
David S. Miller4c9483b2011-03-12 16:22:43 -0500163 u8 proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164 int seg_len = skb->len;
Gerrit Renkere651f032009-08-09 08:12:48 +0000165 int hlimit = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 u32 mtu;
167
168 if (opt) {
Chuck Leverc2636b42007-10-23 21:07:32 -0700169 unsigned int head_room;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170
171 /* First: exthdrs may take lots of space (~8K for now)
172 MAX_HEADER is not enough.
173 */
174 head_room = opt->opt_nflen + opt->opt_flen;
175 seg_len += head_room;
176 head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
177
178 if (skb_headroom(skb) < head_room) {
179 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900180 if (skb2 == NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000181 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900182 IPSTATS_MIB_OUTDISCARDS);
183 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 return -ENOBUFS;
185 }
Eric Dumazet808db802012-04-24 10:17:59 +0000186 consume_skb(skb);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900187 skb = skb2;
Dan Carpenter83d7eb22010-04-30 16:42:08 -0700188 skb_set_owner_w(skb, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 }
190 if (opt->opt_flen)
191 ipv6_push_frag_opts(skb, opt, &proto);
192 if (opt->opt_nflen)
193 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
194 }
195
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700196 skb_push(skb, sizeof(struct ipv6hdr));
197 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700198 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199
200 /*
201 * Fill in the IPv6 header
202 */
Eric Dumazetb903d322011-10-27 00:44:35 -0400203 if (np)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 hlimit = np->hop_limit;
205 if (hlimit < 0)
YOSHIFUJI Hideaki6b75d092008-03-10 06:00:30 -0400206 hlimit = ip6_dst_hoplimit(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207
YOSHIFUJI Hideaki / 吉藤英明3e4e4c12013-01-13 05:01:39 +0000208 ip6_flow_hdr(hdr, tclass, fl6->flowlabel);
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900209
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 hdr->payload_len = htons(seg_len);
211 hdr->nexthdr = proto;
212 hdr->hop_limit = hlimit;
213
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000214 hdr->saddr = fl6->saddr;
215 hdr->daddr = *first_hop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +0200217 skb->protocol = htons(ETH_P_IPV6);
Patrick McHardya2c20642006-01-08 22:37:26 -0800218 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800219 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -0800220
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 mtu = dst_mtu(dst);
Wei Yongjun283d07ac2008-08-03 21:15:59 -0700222 if ((skb->len <= mtu) || skb->local_df || skb_is_gso(skb)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000223 IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)),
Neil Hormanedf391f2009-04-27 02:45:02 -0700224 IPSTATS_MIB_OUT, skb->len);
Jan Engelhardtb2e0b382010-03-23 04:09:07 +0100225 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
226 dst->dev, dst_output);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 }
228
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 skb->dev = dst->dev;
Steffen Klassertf4e53e22013-02-05 01:48:05 +0000230 ipv6_local_error(sk, EMSGSIZE, fl6, mtu);
Eric Dumazetadf30902009-06-02 05:19:30 +0000231 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 kfree_skb(skb);
233 return -EMSGSIZE;
234}
235
YOSHIFUJI Hideaki71590392007-02-22 22:05:40 +0900236EXPORT_SYMBOL(ip6_xmit);
237
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
239{
240 struct ip6_ra_chain *ra;
241 struct sock *last = NULL;
242
243 read_lock(&ip6_ra_lock);
244 for (ra = ip6_ra_chain; ra; ra = ra->next) {
245 struct sock *sk = ra->sk;
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700246 if (sk && ra->sel == sel &&
247 (!sk->sk_bound_dev_if ||
248 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 if (last) {
250 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
251 if (skb2)
252 rawv6_rcv(last, skb2);
253 }
254 last = sk;
255 }
256 }
257
258 if (last) {
259 rawv6_rcv(last, skb);
260 read_unlock(&ip6_ra_lock);
261 return 1;
262 }
263 read_unlock(&ip6_ra_lock);
264 return 0;
265}
266
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700267static int ip6_forward_proxy_check(struct sk_buff *skb)
268{
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700269 struct ipv6hdr *hdr = ipv6_hdr(skb);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700270 u8 nexthdr = hdr->nexthdr;
Jesse Gross75f28112011-11-30 17:05:51 -0800271 __be16 frag_off;
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700272 int offset;
273
274 if (ipv6_ext_hdr(nexthdr)) {
Jesse Gross75f28112011-11-30 17:05:51 -0800275 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700276 if (offset < 0)
277 return 0;
278 } else
279 offset = sizeof(struct ipv6hdr);
280
281 if (nexthdr == IPPROTO_ICMPV6) {
282 struct icmp6hdr *icmp6;
283
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700284 if (!pskb_may_pull(skb, (skb_network_header(skb) +
285 offset + 1 - skb->data)))
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700286 return 0;
287
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700288 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700289
290 switch (icmp6->icmp6_type) {
291 case NDISC_ROUTER_SOLICITATION:
292 case NDISC_ROUTER_ADVERTISEMENT:
293 case NDISC_NEIGHBOUR_SOLICITATION:
294 case NDISC_NEIGHBOUR_ADVERTISEMENT:
295 case NDISC_REDIRECT:
296 /* For reaction involving unicast neighbor discovery
297 * message destined to the proxied address, pass it to
298 * input function.
299 */
300 return 1;
301 default:
302 break;
303 }
304 }
305
Ville Nuorvala74553b02006-09-22 14:42:18 -0700306 /*
307 * The proxying router can't forward traffic sent to a link-local
308 * address, so signal the sender and discard the packet. This
309 * behavior is clarified by the MIPv6 specification.
310 */
311 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
312 dst_link_failure(skb);
313 return -1;
314 }
315
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700316 return 0;
317}
318
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319static inline int ip6_forward_finish(struct sk_buff *skb)
320{
321 return dst_output(skb);
322}
323
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100324static unsigned int ip6_dst_mtu_forward(const struct dst_entry *dst)
325{
326 unsigned int mtu;
327 struct inet6_dev *idev;
328
329 if (dst_metric_locked(dst, RTAX_MTU)) {
330 mtu = dst_metric_raw(dst, RTAX_MTU);
331 if (mtu)
332 return mtu;
333 }
334
335 mtu = IPV6_MIN_MTU;
336 rcu_read_lock();
337 idev = __in6_dev_get(dst->dev);
338 if (idev)
339 mtu = idev->cnf.mtu6;
340 rcu_read_unlock();
341
342 return mtu;
343}
344
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345int ip6_forward(struct sk_buff *skb)
346{
Eric Dumazetadf30902009-06-02 05:19:30 +0000347 struct dst_entry *dst = skb_dst(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700348 struct ipv6hdr *hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 struct inet6_skb_parm *opt = IP6CB(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900350 struct net *net = dev_net(dst->dev);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800351 u32 mtu;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900352
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700353 if (net->ipv6.devconf_all->forwarding == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 goto error;
355
Ben Hutchings4497b072008-06-19 16:22:28 -0700356 if (skb_warn_if_lro(skb))
357 goto drop;
358
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800360 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
361 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 goto drop;
363 }
364
Alexey Kuznetsov72b43d02011-01-12 08:34:08 +0000365 if (skb->pkt_type != PACKET_HOST)
366 goto drop;
367
Herbert Xu35fc92a2007-03-26 23:22:20 -0700368 skb_forward_csum(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369
370 /*
371 * We DO NOT make any processing on
372 * RA packets, pushing them to user level AS IS
373 * without ane WARRANTY that application will be able
374 * to interpret them. The reason is that we
375 * cannot make anything clever here.
376 *
377 * We are not end-node, so that if packet contains
378 * AH/ESP, we cannot make anything.
379 * Defragmentation also would be mistake, RA packets
380 * cannot be fragmented, because there is no warranty
381 * that different fragments will go along one path. --ANK
382 */
YOSHIFUJI Hideaki / 吉藤英明ab4eb352013-06-22 11:13:13 +0900383 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
384 if (ip6_call_ra_chain(skb, ntohs(opt->ra)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 return 0;
386 }
387
388 /*
389 * check and decrement ttl
390 */
391 if (hdr->hop_limit <= 1) {
392 /* Force OUTPUT device used as source address */
393 skb->dev = dst->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000394 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0);
Eric Dumazet15c77d82013-12-05 21:38:06 -0800395 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
396 IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397
398 kfree_skb(skb);
399 return -ETIMEDOUT;
400 }
401
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -0700402 /* XXX: idev->cnf.proxy_ndp? */
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700403 if (net->ipv6.devconf_all->proxy_ndp &&
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800404 pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
Ville Nuorvala74553b02006-09-22 14:42:18 -0700405 int proxied = ip6_forward_proxy_check(skb);
406 if (proxied > 0)
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700407 return ip6_input(skb);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700408 else if (proxied < 0) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800409 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
410 IPSTATS_MIB_INDISCARDS);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700411 goto drop;
412 }
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700413 }
414
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 if (!xfrm6_route_forward(skb)) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800416 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
417 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
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100465 mtu = ip6_dst_mtu_forward(dst);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800466 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);
Eric Dumazet15c77d82013-12-05 21:38:06 -0800474 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
475 IPSTATS_MIB_INTOOBIGERRORS);
476 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
477 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)) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800483 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
484 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 goto drop;
486 }
487
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700488 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489
490 /* Mangling hops number delayed to point after skb COW */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900491
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 hdr->hop_limit--;
493
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700494 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
Vincent Bernat2d8dbb02012-06-05 03:41:42 +0000495 IP6_ADD_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len);
Jan Engelhardtb2e0b382010-03-23 04:09:07 +0100496 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dst->dev,
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800497 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498
499error:
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700500 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501drop:
502 kfree_skb(skb);
503 return -EINVAL;
504}
505
506static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
507{
508 to->pkt_type = from->pkt_type;
509 to->priority = from->priority;
510 to->protocol = from->protocol;
Eric Dumazetadf30902009-06-02 05:19:30 +0000511 skb_dst_drop(to);
512 skb_dst_set(to, dst_clone(skb_dst(from)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 to->dev = from->dev;
Thomas Graf82e91ff2006-11-09 15:19:14 -0800514 to->mark = from->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
516#ifdef CONFIG_NET_SCHED
517 to->tc_index = from->tc_index;
518#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -0700519 nf_copy(to, from);
Amerigo Wang07a93622012-10-29 16:23:10 +0000520#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700521 to->nf_trace = from->nf_trace;
522#endif
James Morris984bc162006-06-09 00:29:17 -0700523 skb_copy_secmark(to, from);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524}
525
David Stevensad0081e2010-12-17 11:42:42 +0000526int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 struct sk_buff *frag;
Eric Dumazetadf30902009-06-02 05:19:30 +0000529 struct rt6_info *rt = (struct rt6_info*)skb_dst(skb);
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800530 struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 struct ipv6hdr *tmp_hdr;
532 struct frag_hdr *fh;
533 unsigned int mtu, hlen, left, len;
Herbert Xua7ae1992011-11-18 02:20:04 +0000534 int hroom, troom;
Al Viroae08e1f2006-11-08 00:27:11 -0800535 __be32 frag_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 int ptr, offset = 0, err=0;
537 u8 *prevhdr, nexthdr = 0;
Eric Dumazetadf30902009-06-02 05:19:30 +0000538 struct net *net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 hlen = ip6_find_1stfragopt(skb, &prevhdr);
541 nexthdr = *prevhdr;
542
John Heffner628a5c52007-04-20 15:53:27 -0700543 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700544
545 /* We must not fragment if the socket is set to force MTU discovery
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800546 * or if the skb it not generated by a local socket.
John Heffnerb881ef72007-04-20 15:52:39 -0700547 */
Patrick McHardy4cdd34082012-08-26 19:13:58 +0200548 if (unlikely(!skb->local_df && skb->len > mtu) ||
549 (IP6CB(skb)->frag_max_size &&
550 IP6CB(skb)->frag_max_size > mtu)) {
Eric Dumazeta34a1012012-05-18 21:51:44 +0000551 if (skb->sk && dst_allfrag(skb_dst(skb)))
552 sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK);
553
Eric Dumazetadf30902009-06-02 05:19:30 +0000554 skb->dev = skb_dst(skb)->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000555 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazetadf30902009-06-02 05:19:30 +0000556 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700557 IPSTATS_MIB_FRAGFAILS);
John Heffnerb881ef72007-04-20 15:52:39 -0700558 kfree_skb(skb);
559 return -EMSGSIZE;
560 }
561
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800562 if (np && np->frag_size < mtu) {
563 if (np->frag_size)
564 mtu = np->frag_size;
565 }
566 mtu -= hlen + sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567
David S. Miller21dc3302010-08-23 00:13:46 -0700568 if (skb_has_frag_list(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 int first_len = skb_pagelen(skb);
Eric Dumazet3d130082010-09-21 08:47:45 +0000570 struct sk_buff *frag2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571
572 if (first_len - hlen > mtu ||
573 ((first_len - hlen) & 7) ||
574 skb_cloned(skb))
575 goto slow_path;
576
David S. Miller4d9092b2009-06-09 00:20:05 -0700577 skb_walk_frags(skb, frag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 /* Correct geometry. */
579 if (frag->len > mtu ||
580 ((frag->len & 7) && frag->next) ||
581 skb_headroom(frag) < hlen)
Eric Dumazet3d130082010-09-21 08:47:45 +0000582 goto slow_path_clean;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 /* Partially cloned skb? */
585 if (skb_shared(frag))
Eric Dumazet3d130082010-09-21 08:47:45 +0000586 goto slow_path_clean;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700587
588 BUG_ON(frag->sk);
589 if (skb->sk) {
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700590 frag->sk = skb->sk;
591 frag->destructor = sock_wfree;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700592 }
Eric Dumazet3d130082010-09-21 08:47:45 +0000593 skb->truesize -= frag->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 }
595
596 err = 0;
597 offset = 0;
598 frag = skb_shinfo(skb)->frag_list;
David S. Miller4d9092b2009-06-09 00:20:05 -0700599 skb_frag_list_init(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 /* BUILD HEADER */
601
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800602 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700603 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 if (!tmp_hdr) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000605 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700606 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 return -ENOMEM;
608 }
609
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610 __skb_pull(skb, hlen);
611 fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700612 __skb_push(skb, hlen);
613 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700614 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615
Eric Dumazet87c48fa2011-07-21 21:25:58 -0700616 ipv6_select_ident(fh, rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 fh->nexthdr = nexthdr;
618 fh->reserved = 0;
619 fh->frag_off = htons(IP6_MF);
620 frag_id = fh->identification;
621
622 first_len = skb_pagelen(skb);
623 skb->data_len = first_len - skb_headlen(skb);
624 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700625 ipv6_hdr(skb)->payload_len = htons(first_len -
626 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900627
Changli Gaod8d1f302010-06-10 23:31:35 -0700628 dst_hold(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629
630 for (;;) {
631 /* Prepare header of the next frame,
632 * before previous one went down. */
633 if (frag) {
634 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300635 skb_reset_transport_header(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700637 __skb_push(frag, hlen);
638 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700639 memcpy(skb_network_header(frag), tmp_hdr,
640 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 offset += skb->len - hlen - sizeof(struct frag_hdr);
642 fh->nexthdr = nexthdr;
643 fh->reserved = 0;
644 fh->frag_off = htons(offset);
645 if (frag->next != NULL)
646 fh->frag_off |= htons(IP6_MF);
647 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700648 ipv6_hdr(frag)->payload_len =
649 htons(frag->len -
650 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 ip6_copy_metadata(frag, skb);
652 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900653
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 err = output(skb);
Wei Dongdafee492006-08-02 13:41:21 -0700655 if(!err)
Changli Gaod8d1f302010-06-10 23:31:35 -0700656 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700657 IPSTATS_MIB_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700658
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 if (err || !frag)
660 break;
661
662 skb = frag;
663 frag = skb->next;
664 skb->next = NULL;
665 }
666
Jesper Juhla51482b2005-11-08 09:41:34 -0800667 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668
669 if (err == 0) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700670 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700671 IPSTATS_MIB_FRAGOKS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000672 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 return 0;
674 }
675
676 while (frag) {
677 skb = frag->next;
678 kfree_skb(frag);
679 frag = skb;
680 }
681
Changli Gaod8d1f302010-06-10 23:31:35 -0700682 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700683 IPSTATS_MIB_FRAGFAILS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000684 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 return err;
Eric Dumazet3d130082010-09-21 08:47:45 +0000686
687slow_path_clean:
688 skb_walk_frags(skb, frag2) {
689 if (frag2 == frag)
690 break;
691 frag2->sk = NULL;
692 frag2->destructor = NULL;
693 skb->truesize += frag2->truesize;
694 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 }
696
697slow_path:
Eric Dumazet72e843b2012-05-18 11:02:15 +0000698 if ((skb->ip_summed == CHECKSUM_PARTIAL) &&
699 skb_checksum_help(skb))
700 goto fail;
701
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 left = skb->len - hlen; /* Space per frame */
703 ptr = hlen; /* Where to start from */
704
705 /*
706 * Fragment the datagram.
707 */
708
709 *prevhdr = NEXTHDR_FRAGMENT;
Herbert Xua7ae1992011-11-18 02:20:04 +0000710 hroom = LL_RESERVED_SPACE(rt->dst.dev);
711 troom = rt->dst.dev->needed_tailroom;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712
713 /*
714 * Keep copying data until we run out.
715 */
716 while(left > 0) {
717 len = left;
718 /* IF: it doesn't fit, use 'mtu' - the data space left */
719 if (len > mtu)
720 len = mtu;
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300721 /* IF: we are not sending up to and including the packet end
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 then align the next start on an eight byte boundary */
723 if (len < left) {
724 len &= ~7;
725 }
726 /*
727 * Allocate buffer.
728 */
729
Herbert Xua7ae1992011-11-18 02:20:04 +0000730 if ((frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) +
731 hroom + troom, GFP_ATOMIC)) == NULL) {
Patrick McHardy64ce2072005-08-09 20:50:53 -0700732 NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n");
Eric Dumazetadf30902009-06-02 05:19:30 +0000733 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900734 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 err = -ENOMEM;
736 goto fail;
737 }
738
739 /*
740 * Set up data on packet
741 */
742
743 ip6_copy_metadata(frag, skb);
Herbert Xua7ae1992011-11-18 02:20:04 +0000744 skb_reserve(frag, hroom);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700746 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300747 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700748 frag->transport_header = (frag->network_header + hlen +
749 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
751 /*
752 * Charge the memory for the fragment to any owner
753 * it might possess
754 */
755 if (skb->sk)
756 skb_set_owner_w(frag, skb->sk);
757
758 /*
759 * Copy the packet header into the new buffer.
760 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300761 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762
763 /*
764 * Build fragment header.
765 */
766 fh->nexthdr = nexthdr;
767 fh->reserved = 0;
Yan Zhengf36d6ab2005-10-03 14:19:15 -0700768 if (!frag_id) {
Eric Dumazet87c48fa2011-07-21 21:25:58 -0700769 ipv6_select_ident(fh, rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 frag_id = fh->identification;
771 } else
772 fh->identification = frag_id;
773
774 /*
775 * Copy a block of the IP datagram.
776 */
Wei Yongjun8984e412007-08-21 20:59:08 -0700777 if (skb_copy_bits(skb, ptr, skb_transport_header(frag), len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 BUG();
779 left -= len;
780
781 fh->frag_off = htons(offset);
782 if (left > 0)
783 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700784 ipv6_hdr(frag)->payload_len = htons(frag->len -
785 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786
787 ptr += len;
788 offset += len;
789
790 /*
791 * Put this fragment into the sending queue.
792 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 err = output(frag);
794 if (err)
795 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700796
Eric Dumazetadf30902009-06-02 05:19:30 +0000797 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700798 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000800 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900801 IPSTATS_MIB_FRAGOKS);
Eric Dumazet808db802012-04-24 10:17:59 +0000802 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 return err;
804
805fail:
Eric Dumazetadf30902009-06-02 05:19:30 +0000806 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900807 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900808 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 return err;
810}
811
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000812static inline int ip6_rt_check(const struct rt6key *rt_key,
813 const struct in6_addr *fl_addr,
814 const struct in6_addr *addr_cache)
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700815{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000816 return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
817 (addr_cache == NULL || !ipv6_addr_equal(fl_addr, addr_cache));
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700818}
819
Herbert Xu497c6152006-07-30 20:19:33 -0700820static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
821 struct dst_entry *dst,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000822 const struct flowi6 *fl6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823{
Herbert Xu497c6152006-07-30 20:19:33 -0700824 struct ipv6_pinfo *np = inet6_sk(sk);
Eric Dumazeta963a372013-06-26 04:15:07 -0700825 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
Herbert Xu497c6152006-07-30 20:19:33 -0700827 if (!dst)
828 goto out;
829
Eric Dumazeta963a372013-06-26 04:15:07 -0700830 if (dst->ops->family != AF_INET6) {
831 dst_release(dst);
832 return NULL;
833 }
834
835 rt = (struct rt6_info *)dst;
Herbert Xu497c6152006-07-30 20:19:33 -0700836 /* Yes, checking route validity in not connected
837 * case is not very simple. Take into account,
838 * that we do not support routing by source, TOS,
839 * and MSG_DONTROUTE --ANK (980726)
840 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700841 * 1. ip6_rt_check(): If route was host route,
842 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700843 * If it is network route, we still may
844 * check its validity using saved pointer
845 * to the last used address: daddr_cache.
846 * We do not want to save whole address now,
847 * (because main consumer of this service
848 * is tcp, which has not this problem),
849 * so that the last trick works only on connected
850 * sockets.
851 * 2. oif also should be the same.
852 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500853 if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700854#ifdef CONFIG_IPV6_SUBTREES
David S. Miller4c9483b2011-03-12 16:22:43 -0500855 ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700856#endif
David S. Miller4c9483b2011-03-12 16:22:43 -0500857 (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex)) {
Herbert Xu497c6152006-07-30 20:19:33 -0700858 dst_release(dst);
859 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 }
861
Herbert Xu497c6152006-07-30 20:19:33 -0700862out:
863 return dst;
864}
865
866static int ip6_dst_lookup_tail(struct sock *sk,
David S. Miller4c9483b2011-03-12 16:22:43 -0500867 struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700868{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900869 struct net *net = sock_net(sk);
David S. Miller69cce1d2011-07-17 23:09:49 -0700870#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
871 struct neighbour *n;
David S. Miller97cac082012-07-02 22:43:47 -0700872 struct rt6_info *rt;
David S. Miller69cce1d2011-07-17 23:09:49 -0700873#endif
874 int err;
Herbert Xu497c6152006-07-30 20:19:33 -0700875
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 if (*dst == NULL)
David S. Miller4c9483b2011-03-12 16:22:43 -0500877 *dst = ip6_route_output(net, sk, fl6);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
879 if ((err = (*dst)->error))
880 goto out_err_release;
881
David S. Miller4c9483b2011-03-12 16:22:43 -0500882 if (ipv6_addr_any(&fl6->saddr)) {
Daniel Walterc3968a82011-04-13 21:10:57 +0000883 struct rt6_info *rt = (struct rt6_info *) *dst;
884 err = ip6_route_get_saddr(net, rt, &fl6->daddr,
885 sk ? inet6_sk(sk)->srcprefs : 0,
886 &fl6->saddr);
Olaf Hering44456d32005-07-27 11:45:17 -0700887 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 goto out_err_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 }
890
Neil Horman95c385b2007-04-25 17:08:10 -0700891#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700892 /*
893 * Here if the dst entry we've looked up
894 * has a neighbour entry that is in the INCOMPLETE
895 * state and the src address from the flow is
896 * marked as OPTIMISTIC, we release the found
897 * dst entry and replace it instead with the
898 * dst entry of the nexthop router
899 */
Eric Dumazetc56bf6f2012-07-06 09:19:05 +0200900 rt = (struct rt6_info *) *dst;
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000901 rcu_read_lock_bh();
Julian Anastasov550bab42013-10-20 15:43:04 +0300902 n = __ipv6_neigh_lookup_noref(rt->dst.dev, rt6_nexthop(rt));
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000903 err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0;
904 rcu_read_unlock_bh();
905
906 if (err) {
Neil Hormane550dfb2008-09-09 13:51:35 -0700907 struct inet6_ifaddr *ifp;
David S. Miller4c9483b2011-03-12 16:22:43 -0500908 struct flowi6 fl_gw6;
Neil Hormane550dfb2008-09-09 13:51:35 -0700909 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -0700910
David S. Miller4c9483b2011-03-12 16:22:43 -0500911 ifp = ipv6_get_ifaddr(net, &fl6->saddr,
Neil Hormane550dfb2008-09-09 13:51:35 -0700912 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -0700913
Neil Hormane550dfb2008-09-09 13:51:35 -0700914 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
915 if (ifp)
916 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -0700917
Neil Hormane550dfb2008-09-09 13:51:35 -0700918 if (redirect) {
919 /*
920 * We need to get the dst entry for the
921 * default router instead
922 */
923 dst_release(*dst);
David S. Miller4c9483b2011-03-12 16:22:43 -0500924 memcpy(&fl_gw6, fl6, sizeof(struct flowi6));
925 memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr));
926 *dst = ip6_route_output(net, sk, &fl_gw6);
Neil Hormane550dfb2008-09-09 13:51:35 -0700927 if ((err = (*dst)->error))
928 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -0700929 }
Neil Hormane550dfb2008-09-09 13:51:35 -0700930 }
Neil Horman95c385b2007-04-25 17:08:10 -0700931#endif
932
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933 return 0;
934
935out_err_release:
Mitsuru Chinenca46f9c2007-12-05 22:31:47 -0800936 if (err == -ENETUNREACH)
John Stultz5ac68e72013-10-07 15:52:01 -0700937 IP6_INC_STATS(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 dst_release(*dst);
939 *dst = NULL;
940 return err;
941}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -0800942
Herbert Xu497c6152006-07-30 20:19:33 -0700943/**
944 * ip6_dst_lookup - perform route lookup on flow
945 * @sk: socket which provides route info
946 * @dst: pointer to dst_entry * for result
David S. Miller4c9483b2011-03-12 16:22:43 -0500947 * @fl6: flow to lookup
Herbert Xu497c6152006-07-30 20:19:33 -0700948 *
949 * This function performs a route lookup on the given flow.
950 *
951 * It returns zero on success, or a standard errno code on error.
952 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500953int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700954{
955 *dst = NULL;
David S. Miller4c9483b2011-03-12 16:22:43 -0500956 return ip6_dst_lookup_tail(sk, dst, fl6);
Herbert Xu497c6152006-07-30 20:19:33 -0700957}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -0800958EXPORT_SYMBOL_GPL(ip6_dst_lookup);
959
Herbert Xu497c6152006-07-30 20:19:33 -0700960/**
David S. Miller68d0c6d2011-03-01 13:19:07 -0800961 * ip6_dst_lookup_flow - perform route lookup on flow with ipsec
962 * @sk: socket which provides route info
David S. Miller4c9483b2011-03-12 16:22:43 -0500963 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -0800964 * @final_dst: final destination address for ipsec lookup
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,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +0200972 const struct in6_addr *final_dst)
David S. Miller68d0c6d2011-03-01 13:19:07 -0800973{
974 struct dst_entry *dst = NULL;
975 int err;
976
David S. Miller4c9483b2011-03-12 16:22:43 -0500977 err = ip6_dst_lookup_tail(sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -0800978 if (err)
979 return ERR_PTR(err);
980 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000981 fl6->daddr = *final_dst;
David S. Miller2774c132011-03-01 14:59:04 -0800982
David S. Miller4c9483b2011-03-12 16:22:43 -0500983 return xfrm_lookup(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
David S. Miller68d0c6d2011-03-01 13:19:07 -0800984}
985EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow);
986
987/**
988 * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow
989 * @sk: socket which provides the dst cache and route info
David S. Miller4c9483b2011-03-12 16:22:43 -0500990 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -0800991 * @final_dst: final destination address for ipsec lookup
Herbert Xu497c6152006-07-30 20:19:33 -0700992 *
993 * This function performs a route lookup on the given flow with the
994 * possibility of using the cached route in the socket if it is valid.
995 * It will take the socket dst lock when operating on the dst cache.
996 * As a result, this function can only be used in process context.
997 *
David S. Miller68d0c6d2011-03-01 13:19:07 -0800998 * It returns a valid dst pointer on success, or a pointer encoded
999 * error code.
Herbert Xu497c6152006-07-30 20:19:33 -07001000 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001001struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001002 const struct in6_addr *final_dst)
Herbert Xu497c6152006-07-30 20:19:33 -07001003{
David S. Miller68d0c6d2011-03-01 13:19:07 -08001004 struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
1005 int err;
Herbert Xu497c6152006-07-30 20:19:33 -07001006
David S. Miller4c9483b2011-03-12 16:22:43 -05001007 dst = ip6_sk_dst_check(sk, dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001008
David S. Miller4c9483b2011-03-12 16:22:43 -05001009 err = ip6_dst_lookup_tail(sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001010 if (err)
1011 return ERR_PTR(err);
1012 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001013 fl6->daddr = *final_dst;
David S. Miller2774c132011-03-01 14:59:04 -08001014
David S. Miller4c9483b2011-03-12 16:22:43 -05001015 return xfrm_lookup(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
Herbert Xu497c6152006-07-30 20:19:33 -07001016}
David S. Miller68d0c6d2011-03-01 13:19:07 -08001017EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
Herbert Xu497c6152006-07-30 20:19:33 -07001018
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001019static inline int ip6_ufo_append_data(struct sock *sk,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001020 int getfrag(void *from, char *to, int offset, int len,
1021 int odd, struct sk_buff *skb),
1022 void *from, int length, int hh_len, int fragheaderlen,
Eric Dumazet87c48fa2011-07-21 21:25:58 -07001023 int transhdrlen, int mtu,unsigned int flags,
1024 struct rt6_info *rt)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001025
1026{
1027 struct sk_buff *skb;
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001028 struct frag_hdr fhdr;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001029 int err;
1030
1031 /* There is support for UDP large send offload by network
1032 * device, so create one single skb packet containing complete
1033 * udp datagram
1034 */
1035 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) {
1036 skb = sock_alloc_send_skb(sk,
1037 hh_len + fragheaderlen + transhdrlen + 20,
1038 (flags & MSG_DONTWAIT), &err);
1039 if (skb == NULL)
Zheng Yan504744e2011-10-28 00:26:00 -04001040 return err;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001041
1042 /* reserve space for Hardware header */
1043 skb_reserve(skb, hh_len);
1044
1045 /* create space for UDP/IP header */
1046 skb_put(skb,fragheaderlen + transhdrlen);
1047
1048 /* initialize network header pointer */
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001049 skb_reset_network_header(skb);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001050
1051 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001052 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001053
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001054 skb->protocol = htons(ETH_P_IPV6);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001055 skb->csum = 0;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001056
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001057 __skb_queue_tail(&sk->sk_write_queue, skb);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001058 } else if (skb_is_gso(skb)) {
1059 goto append;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001060 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001061
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001062 skb->ip_summed = CHECKSUM_PARTIAL;
1063 /* Specify the length of each IPv6 datagram fragment.
1064 * It has to be a multiple of 8.
1065 */
1066 skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
1067 sizeof(struct frag_hdr)) & ~7;
1068 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
1069 ipv6_select_ident(&fhdr, rt);
1070 skb_shinfo(skb)->ip6_frag_id = fhdr.identification;
1071
1072append:
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001073 return skb_append_datato_frags(sk, skb, getfrag, from,
1074 (length - transhdrlen));
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001075}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076
Herbert Xu0178b692009-02-05 15:15:50 -08001077static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
1078 gfp_t gfp)
1079{
1080 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1081}
1082
1083static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
1084 gfp_t gfp)
1085{
1086 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1087}
1088
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001089static void ip6_append_data_mtu(unsigned int *mtu,
Gao feng0c183372012-05-26 01:30:53 +00001090 int *maxfraglen,
1091 unsigned int fragheaderlen,
1092 struct sk_buff *skb,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001093 struct rt6_info *rt,
1094 bool pmtuprobe)
Gao feng0c183372012-05-26 01:30:53 +00001095{
1096 if (!(rt->dst.flags & DST_XFRM_TUNNEL)) {
1097 if (skb == NULL) {
1098 /* first fragment, reserve header_len */
1099 *mtu = *mtu - rt->dst.header_len;
1100
1101 } else {
1102 /*
1103 * this fragment is not first, the headers
1104 * space is regarded as data space.
1105 */
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001106 *mtu = min(*mtu, pmtuprobe ?
1107 rt->dst.dev->mtu :
1108 dst_mtu(rt->dst.path));
Gao feng0c183372012-05-26 01:30:53 +00001109 }
1110 *maxfraglen = ((*mtu - fragheaderlen) & ~7)
1111 + fragheaderlen - sizeof(struct frag_hdr);
1112 }
1113}
1114
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001115int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
1116 int offset, int len, int odd, struct sk_buff *skb),
1117 void *from, int length, int transhdrlen,
David S. Miller4c9483b2011-03-12 16:22:43 -05001118 int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi6 *fl6,
Brian Haley13b52cd2010-04-23 11:26:08 +00001119 struct rt6_info *rt, unsigned int flags, int dontfrag)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120{
1121 struct inet_sock *inet = inet_sk(sk);
1122 struct ipv6_pinfo *np = inet6_sk(sk);
David S. Millerbdc712b2011-05-06 15:02:07 -07001123 struct inet_cork *cork;
Gao feng0c183372012-05-26 01:30:53 +00001124 struct sk_buff *skb, *skb_prev = NULL;
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001125 unsigned int maxfraglen, fragheaderlen, mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 int exthdrlen;
Steffen Klassert299b0762011-10-11 01:43:33 +00001127 int dst_exthdrlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 int hh_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 int copy;
1130 int err;
1131 int offset = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001132 __u8 tx_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133
1134 if (flags&MSG_PROBE)
1135 return 0;
David S. Millerbdc712b2011-05-06 15:02:07 -07001136 cork = &inet->cork.base;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 if (skb_queue_empty(&sk->sk_write_queue)) {
1138 /*
1139 * setup for corking
1140 */
1141 if (opt) {
Herbert Xu0178b692009-02-05 15:15:50 -08001142 if (WARN_ON(np->cork.opt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 return -EINVAL;
Herbert Xu0178b692009-02-05 15:15:50 -08001144
Eric Dumazet284041e2013-05-17 04:53:13 +00001145 np->cork.opt = kzalloc(opt->tot_len, sk->sk_allocation);
Herbert Xu0178b692009-02-05 15:15:50 -08001146 if (unlikely(np->cork.opt == NULL))
1147 return -ENOBUFS;
1148
1149 np->cork.opt->tot_len = opt->tot_len;
1150 np->cork.opt->opt_flen = opt->opt_flen;
1151 np->cork.opt->opt_nflen = opt->opt_nflen;
1152
1153 np->cork.opt->dst0opt = ip6_opt_dup(opt->dst0opt,
1154 sk->sk_allocation);
1155 if (opt->dst0opt && !np->cork.opt->dst0opt)
1156 return -ENOBUFS;
1157
1158 np->cork.opt->dst1opt = ip6_opt_dup(opt->dst1opt,
1159 sk->sk_allocation);
1160 if (opt->dst1opt && !np->cork.opt->dst1opt)
1161 return -ENOBUFS;
1162
1163 np->cork.opt->hopopt = ip6_opt_dup(opt->hopopt,
1164 sk->sk_allocation);
1165 if (opt->hopopt && !np->cork.opt->hopopt)
1166 return -ENOBUFS;
1167
1168 np->cork.opt->srcrt = ip6_rthdr_dup(opt->srcrt,
1169 sk->sk_allocation);
1170 if (opt->srcrt && !np->cork.opt->srcrt)
1171 return -ENOBUFS;
1172
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 /* need source address above miyazawa*/
1174 }
Changli Gaod8d1f302010-06-10 23:31:35 -07001175 dst_hold(&rt->dst);
David S. Millerbdc712b2011-05-06 15:02:07 -07001176 cork->dst = &rt->dst;
David S. Miller4c9483b2011-03-12 16:22:43 -05001177 inet->cork.fl.u.ip6 = *fl6;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 np->cork.hop_limit = hlimit;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001179 np->cork.tclass = tclass;
Gao feng0c183372012-05-26 01:30:53 +00001180 if (rt->dst.flags & DST_XFRM_TUNNEL)
Hannes Frederic Sowa93b36cf2013-12-15 03:41:14 +01001181 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
Gao feng0c183372012-05-26 01:30:53 +00001182 rt->dst.dev->mtu : dst_mtu(&rt->dst);
1183 else
Hannes Frederic Sowa93b36cf2013-12-15 03:41:14 +01001184 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
Gao feng0c183372012-05-26 01:30:53 +00001185 rt->dst.dev->mtu : dst_mtu(rt->dst.path);
Dave Jonesc7503602006-03-20 22:44:52 -08001186 if (np->frag_size < mtu) {
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -08001187 if (np->frag_size)
1188 mtu = np->frag_size;
1189 }
David S. Millerbdc712b2011-05-06 15:02:07 -07001190 cork->fragsize = mtu;
Changli Gaod8d1f302010-06-10 23:31:35 -07001191 if (dst_allfrag(rt->dst.path))
David S. Millerbdc712b2011-05-06 15:02:07 -07001192 cork->flags |= IPCORK_ALLFRAG;
1193 cork->length = 0;
Romain KUNTZ7efdba52013-01-16 12:47:40 +00001194 exthdrlen = (opt ? opt->opt_flen : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195 length += exthdrlen;
1196 transhdrlen += exthdrlen;
Romain KUNTZ7efdba52013-01-16 12:47:40 +00001197 dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 } else {
David S. Millerbdc712b2011-05-06 15:02:07 -07001199 rt = (struct rt6_info *)cork->dst;
David S. Miller4c9483b2011-03-12 16:22:43 -05001200 fl6 = &inet->cork.fl.u.ip6;
Herbert Xu0178b692009-02-05 15:15:50 -08001201 opt = np->cork.opt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 transhdrlen = 0;
1203 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001204 dst_exthdrlen = 0;
David S. Millerbdc712b2011-05-06 15:02:07 -07001205 mtu = cork->fragsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 }
1207
Changli Gaod8d1f302010-06-10 23:31:35 -07001208 hh_len = LL_RESERVED_SPACE(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001210 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001211 (opt ? opt->opt_nflen : 0);
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001212 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen -
1213 sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214
1215 if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001216 unsigned int maxnonfragsize, headersize;
1217
1218 headersize = sizeof(struct ipv6hdr) +
1219 (opt ? opt->tot_len : 0) +
1220 (dst_allfrag(&rt->dst) ?
1221 sizeof(struct frag_hdr) : 0) +
1222 rt->rt6i_nfheader_len;
1223
1224 maxnonfragsize = (np->pmtudisc >= IPV6_PMTUDISC_DO) ?
1225 mtu : sizeof(struct ipv6hdr) + IPV6_MAXPLEN;
1226
1227 /* dontfrag active */
1228 if ((cork->length + length > mtu - headersize) && dontfrag &&
1229 (sk->sk_protocol == IPPROTO_UDP ||
1230 sk->sk_protocol == IPPROTO_RAW)) {
1231 ipv6_local_rxpmtu(sk, fl6, mtu - headersize +
1232 sizeof(struct ipv6hdr));
1233 goto emsgsize;
1234 }
1235
1236 if (cork->length + length > maxnonfragsize - headersize) {
1237emsgsize:
1238 ipv6_local_error(sk, EMSGSIZE, fl6,
1239 mtu - headersize +
1240 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 return -EMSGSIZE;
1242 }
1243 }
1244
Anders Berggrena693e692011-02-28 12:32:11 -08001245 /* For UDP, check if TX timestamp is enabled */
Daniel Borkmannbf84a012013-04-14 08:08:13 +00001246 if (sk->sk_type == SOCK_DGRAM)
1247 sock_tx_timestamp(sk, &tx_flags);
Anders Berggrena693e692011-02-28 12:32:11 -08001248
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 /*
1250 * Let's try using as much space as possible.
1251 * Use MTU if total length of the message fits into the MTU.
1252 * Otherwise, we need to reserve fragment header and
1253 * fragment alignment (= 8-15 octects, in total).
1254 *
1255 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001256 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257 * the message.
1258 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001259 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 * at once if non-fragmentable extension headers
1261 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001262 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 */
1264
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001265 skb = skb_peek_tail(&sk->sk_write_queue);
1266 cork->length += length;
1267 if (((length > mtu) ||
1268 (skb && skb_is_gso(skb))) &&
1269 (sk->sk_protocol == IPPROTO_UDP) &&
1270 (rt->dst.dev->features & NETIF_F_UFO)) {
1271 err = ip6_ufo_append_data(sk, getfrag, from, length,
1272 hh_len, fragheaderlen,
1273 transhdrlen, mtu, flags, rt);
1274 if (err)
1275 goto error;
1276 return 0;
1277 }
1278
1279 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280 goto alloc_new_skb;
1281
1282 while (length > 0) {
1283 /* Check if the remaining data fits into current packet. */
David S. Millerbdc712b2011-05-06 15:02:07 -07001284 copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 if (copy < length)
1286 copy = maxfraglen - skb->len;
1287
1288 if (copy <= 0) {
1289 char *data;
1290 unsigned int datalen;
1291 unsigned int fraglen;
1292 unsigned int fraggap;
1293 unsigned int alloclen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294alloc_new_skb:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 /* There's no room in the current skb */
Gao feng0c183372012-05-26 01:30:53 +00001296 if (skb)
1297 fraggap = skb->len - maxfraglen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 else
1299 fraggap = 0;
Gao feng0c183372012-05-26 01:30:53 +00001300 /* update mtu and maxfraglen if necessary */
1301 if (skb == NULL || skb_prev == NULL)
1302 ip6_append_data_mtu(&mtu, &maxfraglen,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001303 fragheaderlen, skb, rt,
Hannes Frederic Sowa93b36cf2013-12-15 03:41:14 +01001304 np->pmtudisc >=
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001305 IPV6_PMTUDISC_PROBE);
Gao feng0c183372012-05-26 01:30:53 +00001306
1307 skb_prev = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308
1309 /*
1310 * If remaining data exceeds the mtu,
1311 * we know we need more fragment(s).
1312 */
1313 datalen = length + fraggap;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314
Gao feng0c183372012-05-26 01:30:53 +00001315 if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1316 datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317 if ((flags & MSG_MORE) &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001318 !(rt->dst.dev->features&NETIF_F_SG))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 alloclen = mtu;
1320 else
1321 alloclen = datalen + fragheaderlen;
1322
Steffen Klassert299b0762011-10-11 01:43:33 +00001323 alloclen += dst_exthdrlen;
1324
Gao feng0c183372012-05-26 01:30:53 +00001325 if (datalen != length + fraggap) {
1326 /*
1327 * this is not the last fragment, the trailer
1328 * space is regarded as data space.
1329 */
1330 datalen += rt->dst.trailer_len;
1331 }
1332
1333 alloclen += rt->dst.trailer_len;
1334 fraglen = datalen + fragheaderlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335
1336 /*
1337 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001338 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 * (without MSG_MORE) fits into the MTU.
1340 */
1341 alloclen += sizeof(struct frag_hdr);
1342
1343 if (transhdrlen) {
1344 skb = sock_alloc_send_skb(sk,
1345 alloclen + hh_len,
1346 (flags & MSG_DONTWAIT), &err);
1347 } else {
1348 skb = NULL;
1349 if (atomic_read(&sk->sk_wmem_alloc) <=
1350 2 * sk->sk_sndbuf)
1351 skb = sock_wmalloc(sk,
1352 alloclen + hh_len, 1,
1353 sk->sk_allocation);
1354 if (unlikely(skb == NULL))
1355 err = -ENOBUFS;
Anders Berggrena693e692011-02-28 12:32:11 -08001356 else {
1357 /* Only the initial fragment
1358 * is time stamped.
1359 */
1360 tx_flags = 0;
1361 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 }
1363 if (skb == NULL)
1364 goto error;
1365 /*
1366 * Fill in the control structures
1367 */
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001368 skb->protocol = htons(ETH_P_IPV6);
Eric Dumazetd7f7c0a2012-05-18 05:37:56 +00001369 skb->ip_summed = CHECKSUM_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 skb->csum = 0;
Gao feng1f858512012-03-19 22:36:10 +00001371 /* reserve for fragmentation and ipsec header */
1372 skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
1373 dst_exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374
Anders Berggrena693e692011-02-28 12:32:11 -08001375 if (sk->sk_type == SOCK_DGRAM)
1376 skb_shinfo(skb)->tx_flags = tx_flags;
1377
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 /*
1379 * Find where to start putting bytes
1380 */
Gao feng1f858512012-03-19 22:36:10 +00001381 data = skb_put(skb, fraglen);
1382 skb_set_network_header(skb, exthdrlen);
1383 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001384 skb->transport_header = (skb->network_header +
1385 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 if (fraggap) {
1387 skb->csum = skb_copy_and_csum_bits(
1388 skb_prev, maxfraglen,
1389 data + transhdrlen, fraggap, 0);
1390 skb_prev->csum = csum_sub(skb_prev->csum,
1391 skb->csum);
1392 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001393 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 }
1395 copy = datalen - transhdrlen - fraggap;
Steffen Klassert299b0762011-10-11 01:43:33 +00001396
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397 if (copy < 0) {
1398 err = -EINVAL;
1399 kfree_skb(skb);
1400 goto error;
1401 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1402 err = -EFAULT;
1403 kfree_skb(skb);
1404 goto error;
1405 }
1406
1407 offset += copy;
1408 length -= datalen - fraggap;
1409 transhdrlen = 0;
1410 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001411 dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412
1413 /*
1414 * Put the packet on the pending queue
1415 */
1416 __skb_queue_tail(&sk->sk_write_queue, skb);
1417 continue;
1418 }
1419
1420 if (copy > length)
1421 copy = length;
1422
Changli Gaod8d1f302010-06-10 23:31:35 -07001423 if (!(rt->dst.dev->features&NETIF_F_SG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424 unsigned int off;
1425
1426 off = skb->len;
1427 if (getfrag(from, skb_put(skb, copy),
1428 offset, copy, off, skb) < 0) {
1429 __skb_trim(skb, off);
1430 err = -EFAULT;
1431 goto error;
1432 }
1433 } else {
1434 int i = skb_shinfo(skb)->nr_frags;
Eric Dumazet5640f762012-09-23 23:04:42 +00001435 struct page_frag *pfrag = sk_page_frag(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436
Eric Dumazet5640f762012-09-23 23:04:42 +00001437 err = -ENOMEM;
1438 if (!sk_page_frag_refill(sk, pfrag))
1439 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440
Eric Dumazet5640f762012-09-23 23:04:42 +00001441 if (!skb_can_coalesce(skb, i, pfrag->page,
1442 pfrag->offset)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 err = -EMSGSIZE;
Eric Dumazet5640f762012-09-23 23:04:42 +00001444 if (i == MAX_SKB_FRAGS)
1445 goto error;
1446
1447 __skb_fill_page_desc(skb, i, pfrag->page,
1448 pfrag->offset, 0);
1449 skb_shinfo(skb)->nr_frags = ++i;
1450 get_page(pfrag->page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001452 copy = min_t(int, copy, pfrag->size - pfrag->offset);
Eric Dumazet9e903e02011-10-18 21:00:24 +00001453 if (getfrag(from,
Eric Dumazet5640f762012-09-23 23:04:42 +00001454 page_address(pfrag->page) + pfrag->offset,
1455 offset, copy, skb->len, skb) < 0)
1456 goto error_efault;
1457
1458 pfrag->offset += copy;
1459 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460 skb->len += copy;
1461 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001462 skb->truesize += copy;
1463 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 }
1465 offset += copy;
1466 length -= copy;
1467 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001468
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 return 0;
Eric Dumazet5640f762012-09-23 23:04:42 +00001470
1471error_efault:
1472 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473error:
David S. Millerbdc712b2011-05-06 15:02:07 -07001474 cork->length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001475 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 return err;
1477}
Chris Elstona495f832012-04-29 21:48:53 +00001478EXPORT_SYMBOL_GPL(ip6_append_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001480static void ip6_cork_release(struct inet_sock *inet, struct ipv6_pinfo *np)
1481{
Herbert Xu0178b692009-02-05 15:15:50 -08001482 if (np->cork.opt) {
1483 kfree(np->cork.opt->dst0opt);
1484 kfree(np->cork.opt->dst1opt);
1485 kfree(np->cork.opt->hopopt);
1486 kfree(np->cork.opt->srcrt);
1487 kfree(np->cork.opt);
1488 np->cork.opt = NULL;
1489 }
1490
David S. Millerbdc712b2011-05-06 15:02:07 -07001491 if (inet->cork.base.dst) {
1492 dst_release(inet->cork.base.dst);
1493 inet->cork.base.dst = NULL;
1494 inet->cork.base.flags &= ~IPCORK_ALLFRAG;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001495 }
1496 memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
1497}
1498
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499int ip6_push_pending_frames(struct sock *sk)
1500{
1501 struct sk_buff *skb, *tmp_skb;
1502 struct sk_buff **tail_skb;
1503 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
1504 struct inet_sock *inet = inet_sk(sk);
1505 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001506 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 struct ipv6hdr *hdr;
1508 struct ipv6_txoptions *opt = np->cork.opt;
David S. Millerbdc712b2011-05-06 15:02:07 -07001509 struct rt6_info *rt = (struct rt6_info *)inet->cork.base.dst;
David S. Miller4c9483b2011-03-12 16:22:43 -05001510 struct flowi6 *fl6 = &inet->cork.fl.u.ip6;
1511 unsigned char proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 int err = 0;
1513
1514 if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL)
1515 goto out;
1516 tail_skb = &(skb_shinfo(skb)->frag_list);
1517
1518 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001519 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001520 __skb_pull(skb, skb_network_offset(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001522 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 *tail_skb = tmp_skb;
1524 tail_skb = &(tmp_skb->next);
1525 skb->len += tmp_skb->len;
1526 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 skb->truesize += tmp_skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 tmp_skb->destructor = NULL;
1529 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 }
1531
Herbert Xu28a894532008-02-12 18:07:27 -08001532 /* Allow local fragmentation. */
Herbert Xub5c15fc2008-02-14 23:49:37 -08001533 if (np->pmtudisc < IPV6_PMTUDISC_DO)
Herbert Xu28a894532008-02-12 18:07:27 -08001534 skb->local_df = 1;
1535
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001536 *final_dst = fl6->daddr;
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001537 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 if (opt && opt->opt_flen)
1539 ipv6_push_frag_opts(skb, opt, &proto);
1540 if (opt && opt->opt_nflen)
1541 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1542
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001543 skb_push(skb, sizeof(struct ipv6hdr));
1544 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001545 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001546
YOSHIFUJI Hideaki / 吉藤英明3e4e4c12013-01-13 05:01:39 +00001547 ip6_flow_hdr(hdr, np->cork.tclass, fl6->flowlabel);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 hdr->hop_limit = np->cork.hop_limit;
1549 hdr->nexthdr = proto;
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001550 hdr->saddr = fl6->saddr;
1551 hdr->daddr = *final_dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552
Patrick McHardya2c20642006-01-08 22:37:26 -08001553 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001554 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001555
Changli Gaod8d1f302010-06-10 23:31:35 -07001556 skb_dst_set(skb, dst_clone(&rt->dst));
Neil Hormanedf391f2009-04-27 02:45:02 -07001557 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
David L Stevens14878f72007-09-16 16:52:35 -07001558 if (proto == IPPROTO_ICMPV6) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001559 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
David L Stevens14878f72007-09-16 16:52:35 -07001560
Denis V. Lunev5a57d4c2008-10-08 10:34:14 -07001561 ICMP6MSGOUT_INC_STATS_BH(net, idev, icmp6_hdr(skb)->icmp6_type);
Denis V. Luneve41b5362008-10-08 10:33:26 -07001562 ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001563 }
1564
Herbert Xuef76bc22008-01-11 19:15:08 -08001565 err = ip6_local_out(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566 if (err) {
1567 if (err > 0)
Eric Dumazet6ce9e7b2009-09-02 18:05:33 -07001568 err = net_xmit_errno(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 if (err)
1570 goto error;
1571 }
1572
1573out:
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001574 ip6_cork_release(inet, np);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575 return err;
1576error:
Eric Dumazet06254912009-09-01 18:37:16 -07001577 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578 goto out;
1579}
Chris Elstona495f832012-04-29 21:48:53 +00001580EXPORT_SYMBOL_GPL(ip6_push_pending_frames);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581
1582void ip6_flush_pending_frames(struct sock *sk)
1583{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584 struct sk_buff *skb;
1585
1586 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001587 if (skb_dst(skb))
1588 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001589 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590 kfree_skb(skb);
1591 }
1592
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001593 ip6_cork_release(inet_sk(sk), inet6_sk(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594}
Chris Elstona495f832012-04-29 21:48:53 +00001595EXPORT_SYMBOL_GPL(ip6_flush_pending_frames);