blob: cb9df0eb40237065e696dd1013180b6c82a9c2dc [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
Eric Dumazetaad88722014-04-15 13:47:15 -0400135int ip6_output(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136{
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);
WANG Cong60ff7462014-05-04 16:39:18 -0700222 if ((skb->len <= mtu) || skb->ignore_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
Florian Westphalfe6cc552014-02-13 23:09:12 +0100345static bool ip6_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
346{
Florian Westphal418a3152014-05-05 00:03:34 +0200347 if (skb->len <= mtu)
Florian Westphalfe6cc552014-02-13 23:09:12 +0100348 return false;
349
WANG Cong60ff7462014-05-04 16:39:18 -0700350 /* ipv6 conntrack defrag sets max_frag_size + ignore_df */
Florian Westphalfe6cc552014-02-13 23:09:12 +0100351 if (IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu)
352 return true;
353
WANG Cong60ff7462014-05-04 16:39:18 -0700354 if (skb->ignore_df)
Florian Westphal418a3152014-05-05 00:03:34 +0200355 return false;
356
Florian Westphalfe6cc552014-02-13 23:09:12 +0100357 if (skb_is_gso(skb) && skb_gso_network_seglen(skb) <= mtu)
358 return false;
359
360 return true;
361}
362
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363int ip6_forward(struct sk_buff *skb)
364{
Eric Dumazetadf30902009-06-02 05:19:30 +0000365 struct dst_entry *dst = skb_dst(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700366 struct ipv6hdr *hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 struct inet6_skb_parm *opt = IP6CB(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900368 struct net *net = dev_net(dst->dev);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800369 u32 mtu;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900370
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700371 if (net->ipv6.devconf_all->forwarding == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 goto error;
373
Li RongQing090f1162014-03-11 10:40:08 +0800374 if (skb->pkt_type != PACKET_HOST)
375 goto drop;
376
Ben Hutchings4497b072008-06-19 16:22:28 -0700377 if (skb_warn_if_lro(skb))
378 goto drop;
379
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800381 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
382 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 goto drop;
384 }
385
Herbert Xu35fc92a2007-03-26 23:22:20 -0700386 skb_forward_csum(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387
388 /*
389 * We DO NOT make any processing on
390 * RA packets, pushing them to user level AS IS
391 * without ane WARRANTY that application will be able
392 * to interpret them. The reason is that we
393 * cannot make anything clever here.
394 *
395 * We are not end-node, so that if packet contains
396 * AH/ESP, we cannot make anything.
397 * Defragmentation also would be mistake, RA packets
398 * cannot be fragmented, because there is no warranty
399 * that different fragments will go along one path. --ANK
400 */
YOSHIFUJI Hideaki / 吉藤英明ab4eb352013-06-22 11:13:13 +0900401 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
402 if (ip6_call_ra_chain(skb, ntohs(opt->ra)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 return 0;
404 }
405
406 /*
407 * check and decrement ttl
408 */
409 if (hdr->hop_limit <= 1) {
410 /* Force OUTPUT device used as source address */
411 skb->dev = dst->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000412 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0);
Eric Dumazet15c77d82013-12-05 21:38:06 -0800413 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
414 IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415
416 kfree_skb(skb);
417 return -ETIMEDOUT;
418 }
419
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -0700420 /* XXX: idev->cnf.proxy_ndp? */
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700421 if (net->ipv6.devconf_all->proxy_ndp &&
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800422 pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
Ville Nuorvala74553b02006-09-22 14:42:18 -0700423 int proxied = ip6_forward_proxy_check(skb);
424 if (proxied > 0)
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700425 return ip6_input(skb);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700426 else if (proxied < 0) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800427 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
428 IPSTATS_MIB_INDISCARDS);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700429 goto drop;
430 }
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700431 }
432
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 if (!xfrm6_route_forward(skb)) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800434 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
435 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 goto drop;
437 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000438 dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439
440 /* IPv6 specs say nothing about it, but it is clear that we cannot
441 send redirects to source routed frames.
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900442 We don't send redirects to frames decapsulated from IPsec.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 */
David S. Millerc45a3df2012-01-27 15:32:19 -0800444 if (skb->dev == dst->dev && opt->srcrt == 0 && !skb_sec_path(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 struct in6_addr *target = NULL;
David S. Millerfbfe95a2012-06-08 23:24:18 -0700446 struct inet_peer *peer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448
449 /*
450 * incoming and outgoing devices are the same
451 * send a redirect.
452 */
453
454 rt = (struct rt6_info *) dst;
David S. Millerc45a3df2012-01-27 15:32:19 -0800455 if (rt->rt6i_flags & RTF_GATEWAY)
456 target = &rt->rt6i_gateway;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 else
458 target = &hdr->daddr;
459
David S. Miller1d861aa2012-07-10 03:58:16 -0700460 peer = inet_getpeer_v6(net->ipv6.peers, &rt->rt6i_dst.addr, 1);
David S. Miller92d86822011-02-04 15:55:25 -0800461
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 /* Limit redirects both by destination (here)
463 and by source (inside ndisc_send_redirect)
464 */
David S. Millerfbfe95a2012-06-08 23:24:18 -0700465 if (inet_peer_xrlim_allow(peer, 1*HZ))
David S. Miller49919692012-01-27 15:30:48 -0800466 ndisc_send_redirect(skb, target);
David S. Miller1d861aa2012-07-10 03:58:16 -0700467 if (peer)
468 inet_putpeer(peer);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700469 } else {
470 int addrtype = ipv6_addr_type(&hdr->saddr);
471
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 /* This check is security critical. */
YOSHIFUJI Hideakif81b2e72008-06-25 16:55:26 +0900473 if (addrtype == IPV6_ADDR_ANY ||
474 addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK))
David L Stevens5bb1ab02007-05-09 13:53:44 -0700475 goto error;
476 if (addrtype & IPV6_ADDR_LINKLOCAL) {
477 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000478 ICMPV6_NOT_NEIGHBOUR, 0);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700479 goto error;
480 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 }
482
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100483 mtu = ip6_dst_mtu_forward(dst);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800484 if (mtu < IPV6_MIN_MTU)
485 mtu = IPV6_MIN_MTU;
486
Florian Westphalfe6cc552014-02-13 23:09:12 +0100487 if (ip6_pkt_too_big(skb, mtu)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 /* Again, force OUTPUT device used as source address */
489 skb->dev = dst->dev;
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800490 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazet15c77d82013-12-05 21:38:06 -0800491 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
492 IPSTATS_MIB_INTOOBIGERRORS);
493 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
494 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 kfree_skb(skb);
496 return -EMSGSIZE;
497 }
498
499 if (skb_cow(skb, dst->dev->hard_header_len)) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800500 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
501 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 goto drop;
503 }
504
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700505 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506
507 /* Mangling hops number delayed to point after skb COW */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900508
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 hdr->hop_limit--;
510
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700511 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
Vincent Bernat2d8dbb02012-06-05 03:41:42 +0000512 IP6_ADD_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len);
Jan Engelhardtb2e0b382010-03-23 04:09:07 +0100513 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dst->dev,
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800514 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
516error:
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700517 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518drop:
519 kfree_skb(skb);
520 return -EINVAL;
521}
522
523static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
524{
525 to->pkt_type = from->pkt_type;
526 to->priority = from->priority;
527 to->protocol = from->protocol;
Eric Dumazetadf30902009-06-02 05:19:30 +0000528 skb_dst_drop(to);
529 skb_dst_set(to, dst_clone(skb_dst(from)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 to->dev = from->dev;
Thomas Graf82e91ff2006-11-09 15:19:14 -0800531 to->mark = from->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532
533#ifdef CONFIG_NET_SCHED
534 to->tc_index = from->tc_index;
535#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -0700536 nf_copy(to, from);
James Morris984bc162006-06-09 00:29:17 -0700537 skb_copy_secmark(to, from);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538}
539
Eric Dumazet73f156a2014-06-02 05:26:03 -0700540static void ipv6_select_ident(struct frag_hdr *fhdr, struct rt6_info *rt)
541{
542 static u32 ip6_idents_hashrnd __read_mostly;
543 u32 hash, id;
544
545 net_get_random_once(&ip6_idents_hashrnd, sizeof(ip6_idents_hashrnd));
546
547 hash = __ipv6_addr_jhash(&rt->rt6i_dst.addr, ip6_idents_hashrnd);
548 id = ip_idents_reserve(hash, 1);
549 fhdr->identification = htonl(id);
550}
551
David Stevensad0081e2010-12-17 11:42:42 +0000552int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 struct sk_buff *frag;
Eric Dumazetadf30902009-06-02 05:19:30 +0000555 struct rt6_info *rt = (struct rt6_info*)skb_dst(skb);
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800556 struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 struct ipv6hdr *tmp_hdr;
558 struct frag_hdr *fh;
559 unsigned int mtu, hlen, left, len;
Herbert Xua7ae1992011-11-18 02:20:04 +0000560 int hroom, troom;
Al Viroae08e1f2006-11-08 00:27:11 -0800561 __be32 frag_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 int ptr, offset = 0, err=0;
563 u8 *prevhdr, nexthdr = 0;
Eric Dumazetadf30902009-06-02 05:19:30 +0000564 struct net *net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 hlen = ip6_find_1stfragopt(skb, &prevhdr);
567 nexthdr = *prevhdr;
568
John Heffner628a5c52007-04-20 15:53:27 -0700569 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700570
571 /* We must not fragment if the socket is set to force MTU discovery
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800572 * or if the skb it not generated by a local socket.
John Heffnerb881ef72007-04-20 15:52:39 -0700573 */
WANG Cong60ff7462014-05-04 16:39:18 -0700574 if (unlikely(!skb->ignore_df && skb->len > mtu) ||
Patrick McHardy4cdd34082012-08-26 19:13:58 +0200575 (IP6CB(skb)->frag_max_size &&
576 IP6CB(skb)->frag_max_size > mtu)) {
Eric Dumazeta34a1012012-05-18 21:51:44 +0000577 if (skb->sk && dst_allfrag(skb_dst(skb)))
578 sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK);
579
Eric Dumazetadf30902009-06-02 05:19:30 +0000580 skb->dev = skb_dst(skb)->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000581 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazetadf30902009-06-02 05:19:30 +0000582 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700583 IPSTATS_MIB_FRAGFAILS);
John Heffnerb881ef72007-04-20 15:52:39 -0700584 kfree_skb(skb);
585 return -EMSGSIZE;
586 }
587
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800588 if (np && np->frag_size < mtu) {
589 if (np->frag_size)
590 mtu = np->frag_size;
591 }
592 mtu -= hlen + sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593
David S. Miller21dc3302010-08-23 00:13:46 -0700594 if (skb_has_frag_list(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 int first_len = skb_pagelen(skb);
Eric Dumazet3d130082010-09-21 08:47:45 +0000596 struct sk_buff *frag2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597
598 if (first_len - hlen > mtu ||
599 ((first_len - hlen) & 7) ||
600 skb_cloned(skb))
601 goto slow_path;
602
David S. Miller4d9092b2009-06-09 00:20:05 -0700603 skb_walk_frags(skb, frag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 /* Correct geometry. */
605 if (frag->len > mtu ||
606 ((frag->len & 7) && frag->next) ||
607 skb_headroom(frag) < hlen)
Eric Dumazet3d130082010-09-21 08:47:45 +0000608 goto slow_path_clean;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610 /* Partially cloned skb? */
611 if (skb_shared(frag))
Eric Dumazet3d130082010-09-21 08:47:45 +0000612 goto slow_path_clean;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700613
614 BUG_ON(frag->sk);
615 if (skb->sk) {
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700616 frag->sk = skb->sk;
617 frag->destructor = sock_wfree;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700618 }
Eric Dumazet3d130082010-09-21 08:47:45 +0000619 skb->truesize -= frag->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 }
621
622 err = 0;
623 offset = 0;
624 frag = skb_shinfo(skb)->frag_list;
David S. Miller4d9092b2009-06-09 00:20:05 -0700625 skb_frag_list_init(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 /* BUILD HEADER */
627
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800628 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700629 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 if (!tmp_hdr) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000631 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700632 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 return -ENOMEM;
634 }
635
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 __skb_pull(skb, hlen);
637 fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700638 __skb_push(skb, hlen);
639 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700640 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641
Eric Dumazet87c48fa2011-07-21 21:25:58 -0700642 ipv6_select_ident(fh, rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 fh->nexthdr = nexthdr;
644 fh->reserved = 0;
645 fh->frag_off = htons(IP6_MF);
646 frag_id = fh->identification;
647
648 first_len = skb_pagelen(skb);
649 skb->data_len = first_len - skb_headlen(skb);
650 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700651 ipv6_hdr(skb)->payload_len = htons(first_len -
652 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900653
Changli Gaod8d1f302010-06-10 23:31:35 -0700654 dst_hold(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
656 for (;;) {
657 /* Prepare header of the next frame,
658 * before previous one went down. */
659 if (frag) {
660 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300661 skb_reset_transport_header(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700663 __skb_push(frag, hlen);
664 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700665 memcpy(skb_network_header(frag), tmp_hdr,
666 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 offset += skb->len - hlen - sizeof(struct frag_hdr);
668 fh->nexthdr = nexthdr;
669 fh->reserved = 0;
670 fh->frag_off = htons(offset);
671 if (frag->next != NULL)
672 fh->frag_off |= htons(IP6_MF);
673 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700674 ipv6_hdr(frag)->payload_len =
675 htons(frag->len -
676 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 ip6_copy_metadata(frag, skb);
678 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900679
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 err = output(skb);
Wei Dongdafee492006-08-02 13:41:21 -0700681 if(!err)
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_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700684
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 if (err || !frag)
686 break;
687
688 skb = frag;
689 frag = skb->next;
690 skb->next = NULL;
691 }
692
Jesper Juhla51482b2005-11-08 09:41:34 -0800693 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694
695 if (err == 0) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700696 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700697 IPSTATS_MIB_FRAGOKS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000698 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 return 0;
700 }
701
702 while (frag) {
703 skb = frag->next;
704 kfree_skb(frag);
705 frag = skb;
706 }
707
Changli Gaod8d1f302010-06-10 23:31:35 -0700708 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700709 IPSTATS_MIB_FRAGFAILS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000710 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 return err;
Eric Dumazet3d130082010-09-21 08:47:45 +0000712
713slow_path_clean:
714 skb_walk_frags(skb, frag2) {
715 if (frag2 == frag)
716 break;
717 frag2->sk = NULL;
718 frag2->destructor = NULL;
719 skb->truesize += frag2->truesize;
720 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 }
722
723slow_path:
Eric Dumazet72e843b2012-05-18 11:02:15 +0000724 if ((skb->ip_summed == CHECKSUM_PARTIAL) &&
725 skb_checksum_help(skb))
726 goto fail;
727
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 left = skb->len - hlen; /* Space per frame */
729 ptr = hlen; /* Where to start from */
730
731 /*
732 * Fragment the datagram.
733 */
734
735 *prevhdr = NEXTHDR_FRAGMENT;
Herbert Xua7ae1992011-11-18 02:20:04 +0000736 hroom = LL_RESERVED_SPACE(rt->dst.dev);
737 troom = rt->dst.dev->needed_tailroom;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738
739 /*
740 * Keep copying data until we run out.
741 */
742 while(left > 0) {
743 len = left;
744 /* IF: it doesn't fit, use 'mtu' - the data space left */
745 if (len > mtu)
746 len = mtu;
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300747 /* IF: we are not sending up to and including the packet end
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 then align the next start on an eight byte boundary */
749 if (len < left) {
750 len &= ~7;
751 }
752 /*
753 * Allocate buffer.
754 */
755
Herbert Xua7ae1992011-11-18 02:20:04 +0000756 if ((frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) +
757 hroom + troom, GFP_ATOMIC)) == NULL) {
Patrick McHardy64ce2072005-08-09 20:50:53 -0700758 NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n");
Eric Dumazetadf30902009-06-02 05:19:30 +0000759 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900760 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 err = -ENOMEM;
762 goto fail;
763 }
764
765 /*
766 * Set up data on packet
767 */
768
769 ip6_copy_metadata(frag, skb);
Herbert Xua7ae1992011-11-18 02:20:04 +0000770 skb_reserve(frag, hroom);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700772 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300773 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700774 frag->transport_header = (frag->network_header + hlen +
775 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
777 /*
778 * Charge the memory for the fragment to any owner
779 * it might possess
780 */
781 if (skb->sk)
782 skb_set_owner_w(frag, skb->sk);
783
784 /*
785 * Copy the packet header into the new buffer.
786 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300787 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
789 /*
790 * Build fragment header.
791 */
792 fh->nexthdr = nexthdr;
793 fh->reserved = 0;
Yan Zhengf36d6ab2005-10-03 14:19:15 -0700794 if (!frag_id) {
Eric Dumazet87c48fa2011-07-21 21:25:58 -0700795 ipv6_select_ident(fh, rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 frag_id = fh->identification;
797 } else
798 fh->identification = frag_id;
799
800 /*
801 * Copy a block of the IP datagram.
802 */
Wei Yongjun8984e412007-08-21 20:59:08 -0700803 if (skb_copy_bits(skb, ptr, skb_transport_header(frag), len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 BUG();
805 left -= len;
806
807 fh->frag_off = htons(offset);
808 if (left > 0)
809 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700810 ipv6_hdr(frag)->payload_len = htons(frag->len -
811 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812
813 ptr += len;
814 offset += len;
815
816 /*
817 * Put this fragment into the sending queue.
818 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 err = output(frag);
820 if (err)
821 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700822
Eric Dumazetadf30902009-06-02 05:19:30 +0000823 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700824 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000826 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900827 IPSTATS_MIB_FRAGOKS);
Eric Dumazet808db802012-04-24 10:17:59 +0000828 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 return err;
830
831fail:
Eric Dumazetadf30902009-06-02 05:19:30 +0000832 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900833 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900834 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 return err;
836}
837
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000838static inline int ip6_rt_check(const struct rt6key *rt_key,
839 const struct in6_addr *fl_addr,
840 const struct in6_addr *addr_cache)
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700841{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000842 return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
843 (addr_cache == NULL || !ipv6_addr_equal(fl_addr, addr_cache));
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700844}
845
Herbert Xu497c6152006-07-30 20:19:33 -0700846static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
847 struct dst_entry *dst,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000848 const struct flowi6 *fl6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849{
Herbert Xu497c6152006-07-30 20:19:33 -0700850 struct ipv6_pinfo *np = inet6_sk(sk);
Eric Dumazeta963a372013-06-26 04:15:07 -0700851 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852
Herbert Xu497c6152006-07-30 20:19:33 -0700853 if (!dst)
854 goto out;
855
Eric Dumazeta963a372013-06-26 04:15:07 -0700856 if (dst->ops->family != AF_INET6) {
857 dst_release(dst);
858 return NULL;
859 }
860
861 rt = (struct rt6_info *)dst;
Herbert Xu497c6152006-07-30 20:19:33 -0700862 /* Yes, checking route validity in not connected
863 * case is not very simple. Take into account,
864 * that we do not support routing by source, TOS,
865 * and MSG_DONTROUTE --ANK (980726)
866 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700867 * 1. ip6_rt_check(): If route was host route,
868 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700869 * If it is network route, we still may
870 * check its validity using saved pointer
871 * to the last used address: daddr_cache.
872 * We do not want to save whole address now,
873 * (because main consumer of this service
874 * is tcp, which has not this problem),
875 * so that the last trick works only on connected
876 * sockets.
877 * 2. oif also should be the same.
878 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500879 if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700880#ifdef CONFIG_IPV6_SUBTREES
David S. Miller4c9483b2011-03-12 16:22:43 -0500881 ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700882#endif
David S. Miller4c9483b2011-03-12 16:22:43 -0500883 (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex)) {
Herbert Xu497c6152006-07-30 20:19:33 -0700884 dst_release(dst);
885 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 }
887
Herbert Xu497c6152006-07-30 20:19:33 -0700888out:
889 return dst;
890}
891
892static int ip6_dst_lookup_tail(struct sock *sk,
David S. Miller4c9483b2011-03-12 16:22:43 -0500893 struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700894{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900895 struct net *net = sock_net(sk);
David S. Miller69cce1d2011-07-17 23:09:49 -0700896#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
897 struct neighbour *n;
David S. Miller97cac082012-07-02 22:43:47 -0700898 struct rt6_info *rt;
David S. Miller69cce1d2011-07-17 23:09:49 -0700899#endif
900 int err;
Herbert Xu497c6152006-07-30 20:19:33 -0700901
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 if (*dst == NULL)
David S. Miller4c9483b2011-03-12 16:22:43 -0500903 *dst = ip6_route_output(net, sk, fl6);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904
905 if ((err = (*dst)->error))
906 goto out_err_release;
907
David S. Miller4c9483b2011-03-12 16:22:43 -0500908 if (ipv6_addr_any(&fl6->saddr)) {
Daniel Walterc3968a82011-04-13 21:10:57 +0000909 struct rt6_info *rt = (struct rt6_info *) *dst;
910 err = ip6_route_get_saddr(net, rt, &fl6->daddr,
911 sk ? inet6_sk(sk)->srcprefs : 0,
912 &fl6->saddr);
Olaf Hering44456d32005-07-27 11:45:17 -0700913 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 goto out_err_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 }
916
Neil Horman95c385b2007-04-25 17:08:10 -0700917#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700918 /*
919 * Here if the dst entry we've looked up
920 * has a neighbour entry that is in the INCOMPLETE
921 * state and the src address from the flow is
922 * marked as OPTIMISTIC, we release the found
923 * dst entry and replace it instead with the
924 * dst entry of the nexthop router
925 */
Eric Dumazetc56bf6f2012-07-06 09:19:05 +0200926 rt = (struct rt6_info *) *dst;
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000927 rcu_read_lock_bh();
Julian Anastasov550bab42013-10-20 15:43:04 +0300928 n = __ipv6_neigh_lookup_noref(rt->dst.dev, rt6_nexthop(rt));
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000929 err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0;
930 rcu_read_unlock_bh();
931
932 if (err) {
Neil Hormane550dfb2008-09-09 13:51:35 -0700933 struct inet6_ifaddr *ifp;
David S. Miller4c9483b2011-03-12 16:22:43 -0500934 struct flowi6 fl_gw6;
Neil Hormane550dfb2008-09-09 13:51:35 -0700935 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -0700936
David S. Miller4c9483b2011-03-12 16:22:43 -0500937 ifp = ipv6_get_ifaddr(net, &fl6->saddr,
Neil Hormane550dfb2008-09-09 13:51:35 -0700938 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -0700939
Neil Hormane550dfb2008-09-09 13:51:35 -0700940 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
941 if (ifp)
942 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -0700943
Neil Hormane550dfb2008-09-09 13:51:35 -0700944 if (redirect) {
945 /*
946 * We need to get the dst entry for the
947 * default router instead
948 */
949 dst_release(*dst);
David S. Miller4c9483b2011-03-12 16:22:43 -0500950 memcpy(&fl_gw6, fl6, sizeof(struct flowi6));
951 memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr));
952 *dst = ip6_route_output(net, sk, &fl_gw6);
Neil Hormane550dfb2008-09-09 13:51:35 -0700953 if ((err = (*dst)->error))
954 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -0700955 }
Neil Hormane550dfb2008-09-09 13:51:35 -0700956 }
Neil Horman95c385b2007-04-25 17:08:10 -0700957#endif
958
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 return 0;
960
961out_err_release:
Mitsuru Chinenca46f9c2007-12-05 22:31:47 -0800962 if (err == -ENETUNREACH)
John Stultz5ac68e72013-10-07 15:52:01 -0700963 IP6_INC_STATS(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 dst_release(*dst);
965 *dst = NULL;
966 return err;
967}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -0800968
Herbert Xu497c6152006-07-30 20:19:33 -0700969/**
970 * ip6_dst_lookup - perform route lookup on flow
971 * @sk: socket which provides route info
972 * @dst: pointer to dst_entry * for result
David S. Miller4c9483b2011-03-12 16:22:43 -0500973 * @fl6: flow to lookup
Herbert Xu497c6152006-07-30 20:19:33 -0700974 *
975 * This function performs a route lookup on the given flow.
976 *
977 * It returns zero on success, or a standard errno code on error.
978 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500979int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700980{
981 *dst = NULL;
David S. Miller4c9483b2011-03-12 16:22:43 -0500982 return ip6_dst_lookup_tail(sk, dst, fl6);
Herbert Xu497c6152006-07-30 20:19:33 -0700983}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -0800984EXPORT_SYMBOL_GPL(ip6_dst_lookup);
985
Herbert Xu497c6152006-07-30 20:19:33 -0700986/**
David S. Miller68d0c6d2011-03-01 13:19:07 -0800987 * ip6_dst_lookup_flow - perform route lookup on flow with ipsec
988 * @sk: socket which provides route info
David S. Miller4c9483b2011-03-12 16:22:43 -0500989 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -0800990 * @final_dst: final destination address for ipsec lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -0800991 *
992 * This function performs a route lookup on the given flow.
993 *
994 * It returns a valid dst pointer on success, or a pointer encoded
995 * error code.
996 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500997struct dst_entry *ip6_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +0200998 const struct in6_addr *final_dst)
David S. Miller68d0c6d2011-03-01 13:19:07 -0800999{
1000 struct dst_entry *dst = NULL;
1001 int err;
1002
David S. Miller4c9483b2011-03-12 16:22:43 -05001003 err = ip6_dst_lookup_tail(sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001004 if (err)
1005 return ERR_PTR(err);
1006 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001007 fl6->daddr = *final_dst;
David S. Miller2774c132011-03-01 14:59:04 -08001008
David S. Miller4c9483b2011-03-12 16:22:43 -05001009 return xfrm_lookup(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001010}
1011EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow);
1012
1013/**
1014 * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow
1015 * @sk: socket which provides the dst cache and route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001016 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001017 * @final_dst: final destination address for ipsec lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001018 *
1019 * This function performs a route lookup on the given flow with the
1020 * possibility of using the cached route in the socket if it is valid.
1021 * It will take the socket dst lock when operating on the dst cache.
1022 * As a result, this function can only be used in process context.
1023 *
David S. Miller68d0c6d2011-03-01 13:19:07 -08001024 * It returns a valid dst pointer on success, or a pointer encoded
1025 * error code.
Herbert Xu497c6152006-07-30 20:19:33 -07001026 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001027struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001028 const struct in6_addr *final_dst)
Herbert Xu497c6152006-07-30 20:19:33 -07001029{
David S. Miller68d0c6d2011-03-01 13:19:07 -08001030 struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
1031 int err;
Herbert Xu497c6152006-07-30 20:19:33 -07001032
David S. Miller4c9483b2011-03-12 16:22:43 -05001033 dst = ip6_sk_dst_check(sk, dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001034
David S. Miller4c9483b2011-03-12 16:22:43 -05001035 err = ip6_dst_lookup_tail(sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001036 if (err)
1037 return ERR_PTR(err);
1038 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001039 fl6->daddr = *final_dst;
David S. Miller2774c132011-03-01 14:59:04 -08001040
David S. Miller4c9483b2011-03-12 16:22:43 -05001041 return xfrm_lookup(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
Herbert Xu497c6152006-07-30 20:19:33 -07001042}
David S. Miller68d0c6d2011-03-01 13:19:07 -08001043EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
Herbert Xu497c6152006-07-30 20:19:33 -07001044
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001045static inline int ip6_ufo_append_data(struct sock *sk,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001046 int getfrag(void *from, char *to, int offset, int len,
1047 int odd, struct sk_buff *skb),
1048 void *from, int length, int hh_len, int fragheaderlen,
Eric Dumazet87c48fa2011-07-21 21:25:58 -07001049 int transhdrlen, int mtu,unsigned int flags,
1050 struct rt6_info *rt)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001051
1052{
1053 struct sk_buff *skb;
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001054 struct frag_hdr fhdr;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001055 int err;
1056
1057 /* There is support for UDP large send offload by network
1058 * device, so create one single skb packet containing complete
1059 * udp datagram
1060 */
1061 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) {
1062 skb = sock_alloc_send_skb(sk,
1063 hh_len + fragheaderlen + transhdrlen + 20,
1064 (flags & MSG_DONTWAIT), &err);
1065 if (skb == NULL)
Zheng Yan504744e2011-10-28 00:26:00 -04001066 return err;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001067
1068 /* reserve space for Hardware header */
1069 skb_reserve(skb, hh_len);
1070
1071 /* create space for UDP/IP header */
1072 skb_put(skb,fragheaderlen + transhdrlen);
1073
1074 /* initialize network header pointer */
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001075 skb_reset_network_header(skb);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001076
1077 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001078 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001079
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001080 skb->protocol = htons(ETH_P_IPV6);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001081 skb->csum = 0;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001082
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001083 __skb_queue_tail(&sk->sk_write_queue, skb);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001084 } else if (skb_is_gso(skb)) {
1085 goto append;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001086 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001087
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001088 skb->ip_summed = CHECKSUM_PARTIAL;
1089 /* Specify the length of each IPv6 datagram fragment.
1090 * It has to be a multiple of 8.
1091 */
1092 skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
1093 sizeof(struct frag_hdr)) & ~7;
1094 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
1095 ipv6_select_ident(&fhdr, rt);
1096 skb_shinfo(skb)->ip6_frag_id = fhdr.identification;
1097
1098append:
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001099 return skb_append_datato_frags(sk, skb, getfrag, from,
1100 (length - transhdrlen));
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001101}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102
Herbert Xu0178b692009-02-05 15:15:50 -08001103static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
1104 gfp_t gfp)
1105{
1106 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1107}
1108
1109static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
1110 gfp_t gfp)
1111{
1112 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1113}
1114
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001115static void ip6_append_data_mtu(unsigned int *mtu,
Gao feng0c183372012-05-26 01:30:53 +00001116 int *maxfraglen,
1117 unsigned int fragheaderlen,
1118 struct sk_buff *skb,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001119 struct rt6_info *rt,
luciene367c2d2014-03-17 12:51:01 +08001120 unsigned int orig_mtu)
Gao feng0c183372012-05-26 01:30:53 +00001121{
1122 if (!(rt->dst.flags & DST_XFRM_TUNNEL)) {
1123 if (skb == NULL) {
1124 /* first fragment, reserve header_len */
luciene367c2d2014-03-17 12:51:01 +08001125 *mtu = orig_mtu - rt->dst.header_len;
Gao feng0c183372012-05-26 01:30:53 +00001126
1127 } else {
1128 /*
1129 * this fragment is not first, the headers
1130 * space is regarded as data space.
1131 */
luciene367c2d2014-03-17 12:51:01 +08001132 *mtu = orig_mtu;
Gao feng0c183372012-05-26 01:30:53 +00001133 }
1134 *maxfraglen = ((*mtu - fragheaderlen) & ~7)
1135 + fragheaderlen - sizeof(struct frag_hdr);
1136 }
1137}
1138
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001139int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
1140 int offset, int len, int odd, struct sk_buff *skb),
1141 void *from, int length, int transhdrlen,
David S. Miller4c9483b2011-03-12 16:22:43 -05001142 int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi6 *fl6,
Brian Haley13b52cd2010-04-23 11:26:08 +00001143 struct rt6_info *rt, unsigned int flags, int dontfrag)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144{
1145 struct inet_sock *inet = inet_sk(sk);
1146 struct ipv6_pinfo *np = inet6_sk(sk);
David S. Millerbdc712b2011-05-06 15:02:07 -07001147 struct inet_cork *cork;
Gao feng0c183372012-05-26 01:30:53 +00001148 struct sk_buff *skb, *skb_prev = NULL;
luciene367c2d2014-03-17 12:51:01 +08001149 unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150 int exthdrlen;
Steffen Klassert299b0762011-10-11 01:43:33 +00001151 int dst_exthdrlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 int hh_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 int copy;
1154 int err;
1155 int offset = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001156 __u8 tx_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157
1158 if (flags&MSG_PROBE)
1159 return 0;
David S. Millerbdc712b2011-05-06 15:02:07 -07001160 cork = &inet->cork.base;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 if (skb_queue_empty(&sk->sk_write_queue)) {
1162 /*
1163 * setup for corking
1164 */
1165 if (opt) {
Herbert Xu0178b692009-02-05 15:15:50 -08001166 if (WARN_ON(np->cork.opt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167 return -EINVAL;
Herbert Xu0178b692009-02-05 15:15:50 -08001168
Eric Dumazet284041e2013-05-17 04:53:13 +00001169 np->cork.opt = kzalloc(opt->tot_len, sk->sk_allocation);
Herbert Xu0178b692009-02-05 15:15:50 -08001170 if (unlikely(np->cork.opt == NULL))
1171 return -ENOBUFS;
1172
1173 np->cork.opt->tot_len = opt->tot_len;
1174 np->cork.opt->opt_flen = opt->opt_flen;
1175 np->cork.opt->opt_nflen = opt->opt_nflen;
1176
1177 np->cork.opt->dst0opt = ip6_opt_dup(opt->dst0opt,
1178 sk->sk_allocation);
1179 if (opt->dst0opt && !np->cork.opt->dst0opt)
1180 return -ENOBUFS;
1181
1182 np->cork.opt->dst1opt = ip6_opt_dup(opt->dst1opt,
1183 sk->sk_allocation);
1184 if (opt->dst1opt && !np->cork.opt->dst1opt)
1185 return -ENOBUFS;
1186
1187 np->cork.opt->hopopt = ip6_opt_dup(opt->hopopt,
1188 sk->sk_allocation);
1189 if (opt->hopopt && !np->cork.opt->hopopt)
1190 return -ENOBUFS;
1191
1192 np->cork.opt->srcrt = ip6_rthdr_dup(opt->srcrt,
1193 sk->sk_allocation);
1194 if (opt->srcrt && !np->cork.opt->srcrt)
1195 return -ENOBUFS;
1196
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197 /* need source address above miyazawa*/
1198 }
Changli Gaod8d1f302010-06-10 23:31:35 -07001199 dst_hold(&rt->dst);
David S. Millerbdc712b2011-05-06 15:02:07 -07001200 cork->dst = &rt->dst;
David S. Miller4c9483b2011-03-12 16:22:43 -05001201 inet->cork.fl.u.ip6 = *fl6;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 np->cork.hop_limit = hlimit;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001203 np->cork.tclass = tclass;
Gao feng0c183372012-05-26 01:30:53 +00001204 if (rt->dst.flags & DST_XFRM_TUNNEL)
Hannes Frederic Sowa93b36cf2013-12-15 03:41:14 +01001205 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
Gao feng0c183372012-05-26 01:30:53 +00001206 rt->dst.dev->mtu : dst_mtu(&rt->dst);
1207 else
Hannes Frederic Sowa93b36cf2013-12-15 03:41:14 +01001208 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
Gao feng0c183372012-05-26 01:30:53 +00001209 rt->dst.dev->mtu : dst_mtu(rt->dst.path);
Dave Jonesc7503602006-03-20 22:44:52 -08001210 if (np->frag_size < mtu) {
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -08001211 if (np->frag_size)
1212 mtu = np->frag_size;
1213 }
David S. Millerbdc712b2011-05-06 15:02:07 -07001214 cork->fragsize = mtu;
Changli Gaod8d1f302010-06-10 23:31:35 -07001215 if (dst_allfrag(rt->dst.path))
David S. Millerbdc712b2011-05-06 15:02:07 -07001216 cork->flags |= IPCORK_ALLFRAG;
1217 cork->length = 0;
Romain KUNTZ7efdba52013-01-16 12:47:40 +00001218 exthdrlen = (opt ? opt->opt_flen : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 length += exthdrlen;
1220 transhdrlen += exthdrlen;
Romain KUNTZ7efdba52013-01-16 12:47:40 +00001221 dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222 } else {
David S. Millerbdc712b2011-05-06 15:02:07 -07001223 rt = (struct rt6_info *)cork->dst;
David S. Miller4c9483b2011-03-12 16:22:43 -05001224 fl6 = &inet->cork.fl.u.ip6;
Herbert Xu0178b692009-02-05 15:15:50 -08001225 opt = np->cork.opt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226 transhdrlen = 0;
1227 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001228 dst_exthdrlen = 0;
David S. Millerbdc712b2011-05-06 15:02:07 -07001229 mtu = cork->fragsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 }
luciene367c2d2014-03-17 12:51:01 +08001231 orig_mtu = mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232
Changli Gaod8d1f302010-06-10 23:31:35 -07001233 hh_len = LL_RESERVED_SPACE(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001235 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001236 (opt ? opt->opt_nflen : 0);
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001237 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen -
1238 sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239
1240 if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001241 unsigned int maxnonfragsize, headersize;
1242
1243 headersize = sizeof(struct ipv6hdr) +
Hannes Frederic Sowa3a1cebe2014-05-11 23:01:13 +02001244 (opt ? opt->opt_flen + opt->opt_nflen : 0) +
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001245 (dst_allfrag(&rt->dst) ?
1246 sizeof(struct frag_hdr) : 0) +
1247 rt->rt6i_nfheader_len;
1248
WANG Cong60ff7462014-05-04 16:39:18 -07001249 if (ip6_sk_ignore_df(sk))
Hannes Frederic Sowa0b952272014-02-26 01:20:43 +01001250 maxnonfragsize = sizeof(struct ipv6hdr) + IPV6_MAXPLEN;
1251 else
1252 maxnonfragsize = mtu;
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001253
1254 /* dontfrag active */
1255 if ((cork->length + length > mtu - headersize) && dontfrag &&
1256 (sk->sk_protocol == IPPROTO_UDP ||
1257 sk->sk_protocol == IPPROTO_RAW)) {
1258 ipv6_local_rxpmtu(sk, fl6, mtu - headersize +
1259 sizeof(struct ipv6hdr));
1260 goto emsgsize;
1261 }
1262
1263 if (cork->length + length > maxnonfragsize - headersize) {
1264emsgsize:
1265 ipv6_local_error(sk, EMSGSIZE, fl6,
1266 mtu - headersize +
1267 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 return -EMSGSIZE;
1269 }
1270 }
1271
Anders Berggrena693e692011-02-28 12:32:11 -08001272 /* For UDP, check if TX timestamp is enabled */
Daniel Borkmannbf84a012013-04-14 08:08:13 +00001273 if (sk->sk_type == SOCK_DGRAM)
1274 sock_tx_timestamp(sk, &tx_flags);
Anders Berggrena693e692011-02-28 12:32:11 -08001275
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 /*
1277 * Let's try using as much space as possible.
1278 * Use MTU if total length of the message fits into the MTU.
1279 * Otherwise, we need to reserve fragment header and
1280 * fragment alignment (= 8-15 octects, in total).
1281 *
1282 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001283 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 * the message.
1285 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001286 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 * at once if non-fragmentable extension headers
1288 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001289 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290 */
1291
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001292 skb = skb_peek_tail(&sk->sk_write_queue);
1293 cork->length += length;
1294 if (((length > mtu) ||
1295 (skb && skb_is_gso(skb))) &&
1296 (sk->sk_protocol == IPPROTO_UDP) &&
1297 (rt->dst.dev->features & NETIF_F_UFO)) {
1298 err = ip6_ufo_append_data(sk, getfrag, from, length,
1299 hh_len, fragheaderlen,
1300 transhdrlen, mtu, flags, rt);
1301 if (err)
1302 goto error;
1303 return 0;
1304 }
1305
1306 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 goto alloc_new_skb;
1308
1309 while (length > 0) {
1310 /* Check if the remaining data fits into current packet. */
David S. Millerbdc712b2011-05-06 15:02:07 -07001311 copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 if (copy < length)
1313 copy = maxfraglen - skb->len;
1314
1315 if (copy <= 0) {
1316 char *data;
1317 unsigned int datalen;
1318 unsigned int fraglen;
1319 unsigned int fraggap;
1320 unsigned int alloclen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321alloc_new_skb:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322 /* There's no room in the current skb */
Gao feng0c183372012-05-26 01:30:53 +00001323 if (skb)
1324 fraggap = skb->len - maxfraglen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 else
1326 fraggap = 0;
Gao feng0c183372012-05-26 01:30:53 +00001327 /* update mtu and maxfraglen if necessary */
1328 if (skb == NULL || skb_prev == NULL)
1329 ip6_append_data_mtu(&mtu, &maxfraglen,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001330 fragheaderlen, skb, rt,
luciene367c2d2014-03-17 12:51:01 +08001331 orig_mtu);
Gao feng0c183372012-05-26 01:30:53 +00001332
1333 skb_prev = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334
1335 /*
1336 * If remaining data exceeds the mtu,
1337 * we know we need more fragment(s).
1338 */
1339 datalen = length + fraggap;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340
Gao feng0c183372012-05-26 01:30:53 +00001341 if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1342 datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 if ((flags & MSG_MORE) &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001344 !(rt->dst.dev->features&NETIF_F_SG))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345 alloclen = mtu;
1346 else
1347 alloclen = datalen + fragheaderlen;
1348
Steffen Klassert299b0762011-10-11 01:43:33 +00001349 alloclen += dst_exthdrlen;
1350
Gao feng0c183372012-05-26 01:30:53 +00001351 if (datalen != length + fraggap) {
1352 /*
1353 * this is not the last fragment, the trailer
1354 * space is regarded as data space.
1355 */
1356 datalen += rt->dst.trailer_len;
1357 }
1358
1359 alloclen += rt->dst.trailer_len;
1360 fraglen = datalen + fragheaderlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361
1362 /*
1363 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001364 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 * (without MSG_MORE) fits into the MTU.
1366 */
1367 alloclen += sizeof(struct frag_hdr);
1368
1369 if (transhdrlen) {
1370 skb = sock_alloc_send_skb(sk,
1371 alloclen + hh_len,
1372 (flags & MSG_DONTWAIT), &err);
1373 } else {
1374 skb = NULL;
1375 if (atomic_read(&sk->sk_wmem_alloc) <=
1376 2 * sk->sk_sndbuf)
1377 skb = sock_wmalloc(sk,
1378 alloclen + hh_len, 1,
1379 sk->sk_allocation);
1380 if (unlikely(skb == NULL))
1381 err = -ENOBUFS;
Anders Berggrena693e692011-02-28 12:32:11 -08001382 else {
1383 /* Only the initial fragment
1384 * is time stamped.
1385 */
1386 tx_flags = 0;
1387 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 }
1389 if (skb == NULL)
1390 goto error;
1391 /*
1392 * Fill in the control structures
1393 */
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001394 skb->protocol = htons(ETH_P_IPV6);
Eric Dumazetd7f7c0a2012-05-18 05:37:56 +00001395 skb->ip_summed = CHECKSUM_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 skb->csum = 0;
Gao feng1f858512012-03-19 22:36:10 +00001397 /* reserve for fragmentation and ipsec header */
1398 skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
1399 dst_exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400
Anders Berggrena693e692011-02-28 12:32:11 -08001401 if (sk->sk_type == SOCK_DGRAM)
1402 skb_shinfo(skb)->tx_flags = tx_flags;
1403
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 /*
1405 * Find where to start putting bytes
1406 */
Gao feng1f858512012-03-19 22:36:10 +00001407 data = skb_put(skb, fraglen);
1408 skb_set_network_header(skb, exthdrlen);
1409 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001410 skb->transport_header = (skb->network_header +
1411 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 if (fraggap) {
1413 skb->csum = skb_copy_and_csum_bits(
1414 skb_prev, maxfraglen,
1415 data + transhdrlen, fraggap, 0);
1416 skb_prev->csum = csum_sub(skb_prev->csum,
1417 skb->csum);
1418 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001419 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 }
1421 copy = datalen - transhdrlen - fraggap;
Steffen Klassert299b0762011-10-11 01:43:33 +00001422
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423 if (copy < 0) {
1424 err = -EINVAL;
1425 kfree_skb(skb);
1426 goto error;
1427 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1428 err = -EFAULT;
1429 kfree_skb(skb);
1430 goto error;
1431 }
1432
1433 offset += copy;
1434 length -= datalen - fraggap;
1435 transhdrlen = 0;
1436 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001437 dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438
1439 /*
1440 * Put the packet on the pending queue
1441 */
1442 __skb_queue_tail(&sk->sk_write_queue, skb);
1443 continue;
1444 }
1445
1446 if (copy > length)
1447 copy = length;
1448
Changli Gaod8d1f302010-06-10 23:31:35 -07001449 if (!(rt->dst.dev->features&NETIF_F_SG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450 unsigned int off;
1451
1452 off = skb->len;
1453 if (getfrag(from, skb_put(skb, copy),
1454 offset, copy, off, skb) < 0) {
1455 __skb_trim(skb, off);
1456 err = -EFAULT;
1457 goto error;
1458 }
1459 } else {
1460 int i = skb_shinfo(skb)->nr_frags;
Eric Dumazet5640f762012-09-23 23:04:42 +00001461 struct page_frag *pfrag = sk_page_frag(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462
Eric Dumazet5640f762012-09-23 23:04:42 +00001463 err = -ENOMEM;
1464 if (!sk_page_frag_refill(sk, pfrag))
1465 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466
Eric Dumazet5640f762012-09-23 23:04:42 +00001467 if (!skb_can_coalesce(skb, i, pfrag->page,
1468 pfrag->offset)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 err = -EMSGSIZE;
Eric Dumazet5640f762012-09-23 23:04:42 +00001470 if (i == MAX_SKB_FRAGS)
1471 goto error;
1472
1473 __skb_fill_page_desc(skb, i, pfrag->page,
1474 pfrag->offset, 0);
1475 skb_shinfo(skb)->nr_frags = ++i;
1476 get_page(pfrag->page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001478 copy = min_t(int, copy, pfrag->size - pfrag->offset);
Eric Dumazet9e903e02011-10-18 21:00:24 +00001479 if (getfrag(from,
Eric Dumazet5640f762012-09-23 23:04:42 +00001480 page_address(pfrag->page) + pfrag->offset,
1481 offset, copy, skb->len, skb) < 0)
1482 goto error_efault;
1483
1484 pfrag->offset += copy;
1485 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 skb->len += copy;
1487 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001488 skb->truesize += copy;
1489 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 }
1491 offset += copy;
1492 length -= copy;
1493 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001494
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 return 0;
Eric Dumazet5640f762012-09-23 23:04:42 +00001496
1497error_efault:
1498 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499error:
David S. Millerbdc712b2011-05-06 15:02:07 -07001500 cork->length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001501 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 return err;
1503}
Chris Elstona495f832012-04-29 21:48:53 +00001504EXPORT_SYMBOL_GPL(ip6_append_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001506static void ip6_cork_release(struct inet_sock *inet, struct ipv6_pinfo *np)
1507{
Herbert Xu0178b692009-02-05 15:15:50 -08001508 if (np->cork.opt) {
1509 kfree(np->cork.opt->dst0opt);
1510 kfree(np->cork.opt->dst1opt);
1511 kfree(np->cork.opt->hopopt);
1512 kfree(np->cork.opt->srcrt);
1513 kfree(np->cork.opt);
1514 np->cork.opt = NULL;
1515 }
1516
David S. Millerbdc712b2011-05-06 15:02:07 -07001517 if (inet->cork.base.dst) {
1518 dst_release(inet->cork.base.dst);
1519 inet->cork.base.dst = NULL;
1520 inet->cork.base.flags &= ~IPCORK_ALLFRAG;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001521 }
1522 memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
1523}
1524
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525int ip6_push_pending_frames(struct sock *sk)
1526{
1527 struct sk_buff *skb, *tmp_skb;
1528 struct sk_buff **tail_skb;
1529 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
1530 struct inet_sock *inet = inet_sk(sk);
1531 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001532 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 struct ipv6hdr *hdr;
1534 struct ipv6_txoptions *opt = np->cork.opt;
David S. Millerbdc712b2011-05-06 15:02:07 -07001535 struct rt6_info *rt = (struct rt6_info *)inet->cork.base.dst;
David S. Miller4c9483b2011-03-12 16:22:43 -05001536 struct flowi6 *fl6 = &inet->cork.fl.u.ip6;
1537 unsigned char proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 int err = 0;
1539
1540 if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL)
1541 goto out;
1542 tail_skb = &(skb_shinfo(skb)->frag_list);
1543
1544 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001545 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001546 __skb_pull(skb, skb_network_offset(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547 while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001548 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 *tail_skb = tmp_skb;
1550 tail_skb = &(tmp_skb->next);
1551 skb->len += tmp_skb->len;
1552 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553 skb->truesize += tmp_skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554 tmp_skb->destructor = NULL;
1555 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556 }
1557
Herbert Xu28a894532008-02-12 18:07:27 -08001558 /* Allow local fragmentation. */
WANG Cong60ff7462014-05-04 16:39:18 -07001559 skb->ignore_df = ip6_sk_ignore_df(sk);
Herbert Xu28a894532008-02-12 18:07:27 -08001560
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001561 *final_dst = fl6->daddr;
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001562 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 if (opt && opt->opt_flen)
1564 ipv6_push_frag_opts(skb, opt, &proto);
1565 if (opt && opt->opt_nflen)
1566 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1567
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001568 skb_push(skb, sizeof(struct ipv6hdr));
1569 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001570 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001571
YOSHIFUJI Hideaki / 吉藤英明3e4e4c12013-01-13 05:01:39 +00001572 ip6_flow_hdr(hdr, np->cork.tclass, fl6->flowlabel);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 hdr->hop_limit = np->cork.hop_limit;
1574 hdr->nexthdr = proto;
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001575 hdr->saddr = fl6->saddr;
1576 hdr->daddr = *final_dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577
Patrick McHardya2c20642006-01-08 22:37:26 -08001578 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001579 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001580
Changli Gaod8d1f302010-06-10 23:31:35 -07001581 skb_dst_set(skb, dst_clone(&rt->dst));
Neil Hormanedf391f2009-04-27 02:45:02 -07001582 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
David L Stevens14878f72007-09-16 16:52:35 -07001583 if (proto == IPPROTO_ICMPV6) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001584 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
David L Stevens14878f72007-09-16 16:52:35 -07001585
Hannes Frederic Sowa43a43b62014-03-31 20:14:10 +02001586 ICMP6MSGOUT_INC_STATS(net, idev, icmp6_hdr(skb)->icmp6_type);
1587 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001588 }
1589
Herbert Xuef76bc22008-01-11 19:15:08 -08001590 err = ip6_local_out(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591 if (err) {
1592 if (err > 0)
Eric Dumazet6ce9e7b2009-09-02 18:05:33 -07001593 err = net_xmit_errno(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 if (err)
1595 goto error;
1596 }
1597
1598out:
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001599 ip6_cork_release(inet, np);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 return err;
1601error:
Eric Dumazet06254912009-09-01 18:37:16 -07001602 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 goto out;
1604}
Chris Elstona495f832012-04-29 21:48:53 +00001605EXPORT_SYMBOL_GPL(ip6_push_pending_frames);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606
1607void ip6_flush_pending_frames(struct sock *sk)
1608{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 struct sk_buff *skb;
1610
1611 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001612 if (skb_dst(skb))
1613 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001614 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615 kfree_skb(skb);
1616 }
1617
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001618 ip6_cork_release(inet_sk(sk), inet6_sk(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619}
Chris Elstona495f832012-04-29 21:48:53 +00001620EXPORT_SYMBOL_GPL(ip6_flush_pending_frames);