blob: 9428345d3a078786f6662e9384e75c97b9c3e245 [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>
Ian Morris67ba4152014-08-24 21:53:10 +010023 * Imran Patel : frag id should be in NBO
Linus Torvalds1da177e2005-04-16 15:20:36 -070024 * 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>
David Ahernca254492015-10-12 11:47:10 -070058#include <net/l3mdev.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070059
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -050060static int ip6_finish_output2(struct net *net, struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -070061{
Eric Dumazetadf30902009-06-02 05:19:30 +000062 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 struct net_device *dev = dst->dev;
David S. Millerf6b72b62011-07-14 07:53:20 -070064 struct neighbour *neigh;
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +000065 struct in6_addr *nexthop;
66 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070067
68 skb->protocol = htons(ETH_P_IPV6);
69 skb->dev = dev;
70
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070071 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) {
Eric Dumazetadf30902009-06-02 05:19:30 +000072 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
David Miller7026b1d2015-04-05 22:19:04 -040074 if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(sk) &&
Eric W. Biederman78126c42015-09-15 20:04:09 -050075 ((mroute6_socket(net, skb) &&
Benjamin Therybd91b8b2008-12-10 16:07:08 -080076 !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) ||
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090077 ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
78 &ipv6_hdr(skb)->saddr))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
80
81 /* Do not check for IFF_ALLMULTI; multicast routing
82 is not supported in any case.
83 */
84 if (newskb)
Jan Engelhardtb2e0b382010-03-23 04:09:07 +010085 NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING,
Eric W. Biederman29a26a52015-09-15 20:04:16 -050086 net, sk, newskb, NULL, newskb->dev,
Michel Machado95603e22012-06-12 10:16:35 +000087 dev_loopback_xmit);
Linus Torvalds1da177e2005-04-16 15:20:36 -070088
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070089 if (ipv6_hdr(skb)->hop_limit == 0) {
Eric W. Biederman78126c42015-09-15 20:04:09 -050090 IP6_INC_STATS(net, idev,
Denis V. Lunev3bd653c2008-10-08 10:54:51 -070091 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 kfree_skb(skb);
93 return 0;
94 }
95 }
96
Eric W. Biederman78126c42015-09-15 20:04:09 -050097 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUTMCAST, 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();
Martin KaFai Lau2647a9b2015-05-22 20:55:58 -0700108 nexthop = rt6_nexthop((struct rt6_info *)dst, &ipv6_hdr(skb)->daddr);
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
Eric W. Biederman78126c42015-09-15 20:04:09 -0500119 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200120 kfree_skb(skb);
121 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122}
123
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -0500124static int ip6_finish_output(struct net *net, struct sock *sk, struct sk_buff *skb)
Jan Engelhardt9e508492010-04-13 15:28:11 +0200125{
126 if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
Jiri Pirko9037c352013-11-06 17:52:19 +0100127 dst_allfrag(skb_dst(skb)) ||
128 (IP6CB(skb)->frag_max_size && skb->len > IP6CB(skb)->frag_max_size))
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500129 return ip6_fragment(net, sk, skb, ip6_finish_output2);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200130 else
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500131 return ip6_finish_output2(net, sk, skb);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200132}
133
Eric W. Biedermanede20592015-10-07 16:48:47 -0500134int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135{
Jan Engelhardt9e508492010-04-13 15:28:11 +0200136 struct net_device *dev = skb_dst(skb)->dev;
Eric Dumazetadf30902009-06-02 05:19:30 +0000137 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Eric W. Biedermanbe10de02015-09-17 17:21:31 -0500138
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900139 if (unlikely(idev->cnf.disable_ipv6)) {
Eric W. Biederman19a06442015-09-15 20:04:10 -0500140 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900141 kfree_skb(skb);
142 return 0;
143 }
144
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500145 return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING,
146 net, sk, skb, NULL, dev,
Jan Engelhardt9c6eb282010-04-13 15:32:16 +0200147 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/*
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700152 * xmit an sk_buff (used by TCP, SCTP and DCCP)
153 * Note : socket lock is not held for SYNACK packets, but might be modified
154 * by calls to skb_set_owner_w() and ipv6_local_error(),
155 * which are using proper atomic operations or spinlocks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 */
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700157int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
Eric Dumazetb903d322011-10-27 00:44:35 -0400158 struct ipv6_txoptions *opt, int tclass)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159{
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700160 struct net *net = sock_net(sk);
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700161 const struct ipv6_pinfo *np = inet6_sk(sk);
David S. Miller4c9483b2011-03-12 16:22:43 -0500162 struct in6_addr *first_hop = &fl6->daddr;
Eric Dumazetadf30902009-06-02 05:19:30 +0000163 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164 struct ipv6hdr *hdr;
David S. Miller4c9483b2011-03-12 16:22:43 -0500165 u8 proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 int seg_len = skb->len;
Gerrit Renkere651f032009-08-09 08:12:48 +0000167 int hlimit = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 u32 mtu;
169
170 if (opt) {
Chuck Leverc2636b42007-10-23 21:07:32 -0700171 unsigned int head_room;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172
173 /* First: exthdrs may take lots of space (~8K for now)
174 MAX_HEADER is not enough.
175 */
176 head_room = opt->opt_nflen + opt->opt_flen;
177 seg_len += head_room;
178 head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
179
180 if (skb_headroom(skb) < head_room) {
181 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
Ian Morris63159f22015-03-29 14:00:04 +0100182 if (!skb2) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000183 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900184 IPSTATS_MIB_OUTDISCARDS);
185 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 return -ENOBUFS;
187 }
Eric Dumazet808db802012-04-24 10:17:59 +0000188 consume_skb(skb);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900189 skb = skb2;
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700190 /* skb_set_owner_w() changes sk->sk_wmem_alloc atomically,
191 * it is safe to call in our context (socket lock not held)
192 */
193 skb_set_owner_w(skb, (struct sock *)sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 }
195 if (opt->opt_flen)
196 ipv6_push_frag_opts(skb, opt, &proto);
197 if (opt->opt_nflen)
198 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
199 }
200
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700201 skb_push(skb, sizeof(struct ipv6hdr));
202 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700203 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204
205 /*
206 * Fill in the IPv6 header
207 */
Eric Dumazetb903d322011-10-27 00:44:35 -0400208 if (np)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 hlimit = np->hop_limit;
210 if (hlimit < 0)
YOSHIFUJI Hideaki6b75d092008-03-10 06:00:30 -0400211 hlimit = ip6_dst_hoplimit(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212
Tom Herbertcb1ce2e2014-07-01 21:33:10 -0700213 ip6_flow_hdr(hdr, tclass, ip6_make_flowlabel(net, skb, fl6->flowlabel,
Tom Herbert67800f92015-07-31 16:52:11 -0700214 np->autoflowlabel, fl6));
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900215
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 hdr->payload_len = htons(seg_len);
217 hdr->nexthdr = proto;
218 hdr->hop_limit = hlimit;
219
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000220 hdr->saddr = fl6->saddr;
221 hdr->daddr = *first_hop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +0200223 skb->protocol = htons(ETH_P_IPV6);
Patrick McHardya2c20642006-01-08 22:37:26 -0800224 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800225 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -0800226
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 mtu = dst_mtu(dst);
WANG Cong60ff7462014-05-04 16:39:18 -0700228 if ((skb->len <= mtu) || skb->ignore_df || skb_is_gso(skb)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000229 IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)),
Neil Hormanedf391f2009-04-27 02:45:02 -0700230 IPSTATS_MIB_OUT, skb->len);
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700231 /* hooks should never assume socket lock is held.
232 * we promote our socket to non const
233 */
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500234 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700235 net, (struct sock *)sk, skb, NULL, dst->dev,
Eric W. Biederman13206b62015-10-07 16:48:35 -0500236 dst_output);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 }
238
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 skb->dev = dst->dev;
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700240 /* ipv6_local_error() does not require socket lock,
241 * we promote our socket to non const
242 */
243 ipv6_local_error((struct sock *)sk, EMSGSIZE, fl6, mtu);
244
Eric Dumazetadf30902009-06-02 05:19:30 +0000245 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 kfree_skb(skb);
247 return -EMSGSIZE;
248}
YOSHIFUJI Hideaki71590392007-02-22 22:05:40 +0900249EXPORT_SYMBOL(ip6_xmit);
250
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
252{
253 struct ip6_ra_chain *ra;
254 struct sock *last = NULL;
255
256 read_lock(&ip6_ra_lock);
257 for (ra = ip6_ra_chain; ra; ra = ra->next) {
258 struct sock *sk = ra->sk;
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700259 if (sk && ra->sel == sel &&
260 (!sk->sk_bound_dev_if ||
261 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 if (last) {
263 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
264 if (skb2)
265 rawv6_rcv(last, skb2);
266 }
267 last = sk;
268 }
269 }
270
271 if (last) {
272 rawv6_rcv(last, skb);
273 read_unlock(&ip6_ra_lock);
274 return 1;
275 }
276 read_unlock(&ip6_ra_lock);
277 return 0;
278}
279
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700280static int ip6_forward_proxy_check(struct sk_buff *skb)
281{
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700282 struct ipv6hdr *hdr = ipv6_hdr(skb);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700283 u8 nexthdr = hdr->nexthdr;
Jesse Gross75f28112011-11-30 17:05:51 -0800284 __be16 frag_off;
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700285 int offset;
286
287 if (ipv6_ext_hdr(nexthdr)) {
Jesse Gross75f28112011-11-30 17:05:51 -0800288 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700289 if (offset < 0)
290 return 0;
291 } else
292 offset = sizeof(struct ipv6hdr);
293
294 if (nexthdr == IPPROTO_ICMPV6) {
295 struct icmp6hdr *icmp6;
296
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700297 if (!pskb_may_pull(skb, (skb_network_header(skb) +
298 offset + 1 - skb->data)))
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700299 return 0;
300
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700301 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700302
303 switch (icmp6->icmp6_type) {
304 case NDISC_ROUTER_SOLICITATION:
305 case NDISC_ROUTER_ADVERTISEMENT:
306 case NDISC_NEIGHBOUR_SOLICITATION:
307 case NDISC_NEIGHBOUR_ADVERTISEMENT:
308 case NDISC_REDIRECT:
309 /* For reaction involving unicast neighbor discovery
310 * message destined to the proxied address, pass it to
311 * input function.
312 */
313 return 1;
314 default:
315 break;
316 }
317 }
318
Ville Nuorvala74553b02006-09-22 14:42:18 -0700319 /*
320 * The proxying router can't forward traffic sent to a link-local
321 * address, so signal the sender and discard the packet. This
322 * behavior is clarified by the MIPv6 specification.
323 */
324 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
325 dst_link_failure(skb);
326 return -1;
327 }
328
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700329 return 0;
330}
331
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -0500332static inline int ip6_forward_finish(struct net *net, struct sock *sk,
333 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334{
Eric W. Biederman13206b62015-10-07 16:48:35 -0500335 return dst_output(net, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336}
337
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100338static unsigned int ip6_dst_mtu_forward(const struct dst_entry *dst)
339{
340 unsigned int mtu;
341 struct inet6_dev *idev;
342
343 if (dst_metric_locked(dst, RTAX_MTU)) {
344 mtu = dst_metric_raw(dst, RTAX_MTU);
345 if (mtu)
346 return mtu;
347 }
348
349 mtu = IPV6_MIN_MTU;
350 rcu_read_lock();
351 idev = __in6_dev_get(dst->dev);
352 if (idev)
353 mtu = idev->cnf.mtu6;
354 rcu_read_unlock();
355
356 return mtu;
357}
358
Florian Westphalfe6cc552014-02-13 23:09:12 +0100359static bool ip6_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
360{
Florian Westphal418a3152014-05-05 00:03:34 +0200361 if (skb->len <= mtu)
Florian Westphalfe6cc552014-02-13 23:09:12 +0100362 return false;
363
WANG Cong60ff7462014-05-04 16:39:18 -0700364 /* ipv6 conntrack defrag sets max_frag_size + ignore_df */
Florian Westphalfe6cc552014-02-13 23:09:12 +0100365 if (IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu)
366 return true;
367
WANG Cong60ff7462014-05-04 16:39:18 -0700368 if (skb->ignore_df)
Florian Westphal418a3152014-05-05 00:03:34 +0200369 return false;
370
Florian Westphalfe6cc552014-02-13 23:09:12 +0100371 if (skb_is_gso(skb) && skb_gso_network_seglen(skb) <= mtu)
372 return false;
373
374 return true;
375}
376
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377int ip6_forward(struct sk_buff *skb)
378{
Eric Dumazetadf30902009-06-02 05:19:30 +0000379 struct dst_entry *dst = skb_dst(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700380 struct ipv6hdr *hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 struct inet6_skb_parm *opt = IP6CB(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900382 struct net *net = dev_net(dst->dev);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800383 u32 mtu;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900384
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700385 if (net->ipv6.devconf_all->forwarding == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 goto error;
387
Li RongQing090f1162014-03-11 10:40:08 +0800388 if (skb->pkt_type != PACKET_HOST)
389 goto drop;
390
Hannes Frederic Sowa9ef2e962015-10-08 18:19:53 +0200391 if (unlikely(skb->sk))
392 goto drop;
393
Ben Hutchings4497b072008-06-19 16:22:28 -0700394 if (skb_warn_if_lro(skb))
395 goto drop;
396
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800398 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
399 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 goto drop;
401 }
402
Herbert Xu35fc92a2007-03-26 23:22:20 -0700403 skb_forward_csum(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404
405 /*
406 * We DO NOT make any processing on
407 * RA packets, pushing them to user level AS IS
408 * without ane WARRANTY that application will be able
409 * to interpret them. The reason is that we
410 * cannot make anything clever here.
411 *
412 * We are not end-node, so that if packet contains
413 * AH/ESP, we cannot make anything.
414 * Defragmentation also would be mistake, RA packets
415 * cannot be fragmented, because there is no warranty
416 * that different fragments will go along one path. --ANK
417 */
YOSHIFUJI Hideaki / 吉藤英明ab4eb352013-06-22 11:13:13 +0900418 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
419 if (ip6_call_ra_chain(skb, ntohs(opt->ra)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 return 0;
421 }
422
423 /*
424 * check and decrement ttl
425 */
426 if (hdr->hop_limit <= 1) {
427 /* Force OUTPUT device used as source address */
428 skb->dev = dst->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000429 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0);
Eric Dumazet15c77d82013-12-05 21:38:06 -0800430 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
431 IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432
433 kfree_skb(skb);
434 return -ETIMEDOUT;
435 }
436
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -0700437 /* XXX: idev->cnf.proxy_ndp? */
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700438 if (net->ipv6.devconf_all->proxy_ndp &&
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800439 pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
Ville Nuorvala74553b02006-09-22 14:42:18 -0700440 int proxied = ip6_forward_proxy_check(skb);
441 if (proxied > 0)
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700442 return ip6_input(skb);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700443 else if (proxied < 0) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800444 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
445 IPSTATS_MIB_INDISCARDS);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700446 goto drop;
447 }
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700448 }
449
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 if (!xfrm6_route_forward(skb)) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800451 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
452 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 goto drop;
454 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000455 dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456
457 /* IPv6 specs say nothing about it, but it is clear that we cannot
458 send redirects to source routed frames.
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900459 We don't send redirects to frames decapsulated from IPsec.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 */
David S. Millerc45a3df2012-01-27 15:32:19 -0800461 if (skb->dev == dst->dev && opt->srcrt == 0 && !skb_sec_path(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 struct in6_addr *target = NULL;
David S. Millerfbfe95a2012-06-08 23:24:18 -0700463 struct inet_peer *peer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
466 /*
467 * incoming and outgoing devices are the same
468 * send a redirect.
469 */
470
471 rt = (struct rt6_info *) dst;
David S. Millerc45a3df2012-01-27 15:32:19 -0800472 if (rt->rt6i_flags & RTF_GATEWAY)
473 target = &rt->rt6i_gateway;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 else
475 target = &hdr->daddr;
476
Martin KaFai Laufd0273d2015-05-22 20:55:57 -0700477 peer = inet_getpeer_v6(net->ipv6.peers, &hdr->daddr, 1);
David S. Miller92d86822011-02-04 15:55:25 -0800478
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 /* Limit redirects both by destination (here)
480 and by source (inside ndisc_send_redirect)
481 */
David S. Millerfbfe95a2012-06-08 23:24:18 -0700482 if (inet_peer_xrlim_allow(peer, 1*HZ))
David S. Miller49919692012-01-27 15:30:48 -0800483 ndisc_send_redirect(skb, target);
David S. Miller1d861aa2012-07-10 03:58:16 -0700484 if (peer)
485 inet_putpeer(peer);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700486 } else {
487 int addrtype = ipv6_addr_type(&hdr->saddr);
488
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 /* This check is security critical. */
YOSHIFUJI Hideakif81b2e72008-06-25 16:55:26 +0900490 if (addrtype == IPV6_ADDR_ANY ||
491 addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK))
David L Stevens5bb1ab02007-05-09 13:53:44 -0700492 goto error;
493 if (addrtype & IPV6_ADDR_LINKLOCAL) {
494 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000495 ICMPV6_NOT_NEIGHBOUR, 0);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700496 goto error;
497 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 }
499
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100500 mtu = ip6_dst_mtu_forward(dst);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800501 if (mtu < IPV6_MIN_MTU)
502 mtu = IPV6_MIN_MTU;
503
Florian Westphalfe6cc552014-02-13 23:09:12 +0100504 if (ip6_pkt_too_big(skb, mtu)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 /* Again, force OUTPUT device used as source address */
506 skb->dev = dst->dev;
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800507 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazet15c77d82013-12-05 21:38:06 -0800508 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
509 IPSTATS_MIB_INTOOBIGERRORS);
510 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
511 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 kfree_skb(skb);
513 return -EMSGSIZE;
514 }
515
516 if (skb_cow(skb, dst->dev->hard_header_len)) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800517 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
518 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 goto drop;
520 }
521
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700522 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523
524 /* Mangling hops number delayed to point after skb COW */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900525
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 hdr->hop_limit--;
527
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700528 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
Vincent Bernat2d8dbb02012-06-05 03:41:42 +0000529 IP6_ADD_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len);
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500530 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
531 net, NULL, skb, skb->dev, dst->dev,
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800532 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
534error:
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700535 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536drop:
537 kfree_skb(skb);
538 return -EINVAL;
539}
540
541static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
542{
543 to->pkt_type = from->pkt_type;
544 to->priority = from->priority;
545 to->protocol = from->protocol;
Eric Dumazetadf30902009-06-02 05:19:30 +0000546 skb_dst_drop(to);
547 skb_dst_set(to, dst_clone(skb_dst(from)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 to->dev = from->dev;
Thomas Graf82e91ff2006-11-09 15:19:14 -0800549 to->mark = from->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550
551#ifdef CONFIG_NET_SCHED
552 to->tc_index = from->tc_index;
553#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -0700554 nf_copy(to, from);
James Morris984bc162006-06-09 00:29:17 -0700555 skb_copy_secmark(to, from);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556}
557
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500558int ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
559 int (*output)(struct net *, struct sock *, struct sk_buff *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 struct sk_buff *frag;
Ian Morris67ba4152014-08-24 21:53:10 +0100562 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +0200563 struct ipv6_pinfo *np = skb->sk && !dev_recursion_level() ?
564 inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 struct ipv6hdr *tmp_hdr;
566 struct frag_hdr *fh;
567 unsigned int mtu, hlen, left, len;
Herbert Xua7ae1992011-11-18 02:20:04 +0000568 int hroom, troom;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700569 __be32 frag_id;
Ian Morris67ba4152014-08-24 21:53:10 +0100570 int ptr, offset = 0, err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 u8 *prevhdr, nexthdr = 0;
572
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 hlen = ip6_find_1stfragopt(skb, &prevhdr);
574 nexthdr = *prevhdr;
575
John Heffner628a5c52007-04-20 15:53:27 -0700576 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700577
578 /* We must not fragment if the socket is set to force MTU discovery
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800579 * or if the skb it not generated by a local socket.
John Heffnerb881ef72007-04-20 15:52:39 -0700580 */
Florian Westphal485fca62015-05-22 00:44:16 +0200581 if (unlikely(!skb->ignore_df && skb->len > mtu))
582 goto fail_toobig;
Eric Dumazeta34a1012012-05-18 21:51:44 +0000583
Florian Westphal485fca62015-05-22 00:44:16 +0200584 if (IP6CB(skb)->frag_max_size) {
585 if (IP6CB(skb)->frag_max_size > mtu)
586 goto fail_toobig;
587
588 /* don't send fragments larger than what we received */
589 mtu = IP6CB(skb)->frag_max_size;
590 if (mtu < IPV6_MIN_MTU)
591 mtu = IPV6_MIN_MTU;
John Heffnerb881ef72007-04-20 15:52:39 -0700592 }
593
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800594 if (np && np->frag_size < mtu) {
595 if (np->frag_size)
596 mtu = np->frag_size;
597 }
Hannes Frederic Sowa89bc7842015-10-28 13:21:04 +0100598 if (mtu < hlen + sizeof(struct frag_hdr) + 8)
Hannes Frederic Sowab72a2b02015-10-16 11:32:43 +0200599 goto fail_toobig;
Hannes Frederic Sowa1e0d69a2015-10-28 13:21:03 +0100600 mtu -= hlen + sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601
Martin KaFai Laufd0273d2015-05-22 20:55:57 -0700602 frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
603 &ipv6_hdr(skb)->saddr);
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700604
Hannes Frederic Sowa405c92f2015-10-27 22:40:42 +0100605 if (skb->ip_summed == CHECKSUM_PARTIAL &&
606 (err = skb_checksum_help(skb)))
607 goto fail;
608
Florian Westphal1d325d22015-09-16 17:26:14 +0200609 hroom = LL_RESERVED_SPACE(rt->dst.dev);
David S. Miller21dc3302010-08-23 00:13:46 -0700610 if (skb_has_frag_list(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 int first_len = skb_pagelen(skb);
Eric Dumazet3d130082010-09-21 08:47:45 +0000612 struct sk_buff *frag2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613
614 if (first_len - hlen > mtu ||
615 ((first_len - hlen) & 7) ||
Florian Westphal1d325d22015-09-16 17:26:14 +0200616 skb_cloned(skb) ||
617 skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 goto slow_path;
619
David S. Miller4d9092b2009-06-09 00:20:05 -0700620 skb_walk_frags(skb, frag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 /* Correct geometry. */
622 if (frag->len > mtu ||
623 ((frag->len & 7) && frag->next) ||
Florian Westphal1d325d22015-09-16 17:26:14 +0200624 skb_headroom(frag) < (hlen + hroom + sizeof(struct frag_hdr)))
Eric Dumazet3d130082010-09-21 08:47:45 +0000625 goto slow_path_clean;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 /* Partially cloned skb? */
628 if (skb_shared(frag))
Eric Dumazet3d130082010-09-21 08:47:45 +0000629 goto slow_path_clean;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700630
631 BUG_ON(frag->sk);
632 if (skb->sk) {
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700633 frag->sk = skb->sk;
634 frag->destructor = sock_wfree;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700635 }
Eric Dumazet3d130082010-09-21 08:47:45 +0000636 skb->truesize -= frag->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 }
638
639 err = 0;
640 offset = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 /* BUILD HEADER */
642
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800643 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700644 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645 if (!tmp_hdr) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000646 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700647 IPSTATS_MIB_FRAGFAILS);
Florian Westphal1d325d22015-09-16 17:26:14 +0200648 err = -ENOMEM;
649 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 }
Florian Westphal1d325d22015-09-16 17:26:14 +0200651 frag = skb_shinfo(skb)->frag_list;
652 skb_frag_list_init(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 __skb_pull(skb, hlen);
Ian Morris67ba4152014-08-24 21:53:10 +0100655 fh = (struct frag_hdr *)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700656 __skb_push(skb, hlen);
657 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700658 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 fh->nexthdr = nexthdr;
661 fh->reserved = 0;
662 fh->frag_off = htons(IP6_MF);
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700663 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664
665 first_len = skb_pagelen(skb);
666 skb->data_len = first_len - skb_headlen(skb);
667 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700668 ipv6_hdr(skb)->payload_len = htons(first_len -
669 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900670
Changli Gaod8d1f302010-06-10 23:31:35 -0700671 dst_hold(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672
673 for (;;) {
674 /* Prepare header of the next frame,
675 * before previous one went down. */
676 if (frag) {
677 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300678 skb_reset_transport_header(frag);
Ian Morris67ba4152014-08-24 21:53:10 +0100679 fh = (struct frag_hdr *)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700680 __skb_push(frag, hlen);
681 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700682 memcpy(skb_network_header(frag), tmp_hdr,
683 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 offset += skb->len - hlen - sizeof(struct frag_hdr);
685 fh->nexthdr = nexthdr;
686 fh->reserved = 0;
687 fh->frag_off = htons(offset);
Ian Morris53b24b82015-03-29 14:00:05 +0100688 if (frag->next)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 fh->frag_off |= htons(IP6_MF);
690 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700691 ipv6_hdr(frag)->payload_len =
692 htons(frag->len -
693 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 ip6_copy_metadata(frag, skb);
695 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900696
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500697 err = output(net, sk, skb);
Ian Morris67ba4152014-08-24 21:53:10 +0100698 if (!err)
Changli Gaod8d1f302010-06-10 23:31:35 -0700699 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700700 IPSTATS_MIB_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700701
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 if (err || !frag)
703 break;
704
705 skb = frag;
706 frag = skb->next;
707 skb->next = NULL;
708 }
709
Jesper Juhla51482b2005-11-08 09:41:34 -0800710 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711
712 if (err == 0) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700713 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700714 IPSTATS_MIB_FRAGOKS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000715 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 return 0;
717 }
718
Florian Westphal46cfd722014-09-10 01:08:46 +0200719 kfree_skb_list(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720
Changli Gaod8d1f302010-06-10 23:31:35 -0700721 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700722 IPSTATS_MIB_FRAGFAILS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000723 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 return err;
Eric Dumazet3d130082010-09-21 08:47:45 +0000725
726slow_path_clean:
727 skb_walk_frags(skb, frag2) {
728 if (frag2 == frag)
729 break;
730 frag2->sk = NULL;
731 frag2->destructor = NULL;
732 skb->truesize += frag2->truesize;
733 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 }
735
736slow_path:
737 left = skb->len - hlen; /* Space per frame */
738 ptr = hlen; /* Where to start from */
739
740 /*
741 * Fragment the datagram.
742 */
743
744 *prevhdr = NEXTHDR_FRAGMENT;
Herbert Xua7ae1992011-11-18 02:20:04 +0000745 troom = rt->dst.dev->needed_tailroom;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746
747 /*
748 * Keep copying data until we run out.
749 */
Ian Morris67ba4152014-08-24 21:53:10 +0100750 while (left > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 len = left;
752 /* IF: it doesn't fit, use 'mtu' - the data space left */
753 if (len > mtu)
754 len = mtu;
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300755 /* IF: we are not sending up to and including the packet end
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 then align the next start on an eight byte boundary */
757 if (len < left) {
758 len &= ~7;
759 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760
Joe Perchescbffccc2014-11-05 14:39:21 -0800761 /* Allocate buffer */
762 frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) +
763 hroom + troom, GFP_ATOMIC);
764 if (!frag) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000765 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900766 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 err = -ENOMEM;
768 goto fail;
769 }
770
771 /*
772 * Set up data on packet
773 */
774
775 ip6_copy_metadata(frag, skb);
Herbert Xua7ae1992011-11-18 02:20:04 +0000776 skb_reserve(frag, hroom);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700778 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300779 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700780 frag->transport_header = (frag->network_header + hlen +
781 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782
783 /*
784 * Charge the memory for the fragment to any owner
785 * it might possess
786 */
787 if (skb->sk)
788 skb_set_owner_w(frag, skb->sk);
789
790 /*
791 * Copy the packet header into the new buffer.
792 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300793 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
795 /*
796 * Build fragment header.
797 */
798 fh->nexthdr = nexthdr;
799 fh->reserved = 0;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700800 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801
802 /*
803 * Copy a block of the IP datagram.
804 */
Himangi Saraogie3f0b862014-07-12 01:57:17 +0530805 BUG_ON(skb_copy_bits(skb, ptr, skb_transport_header(frag),
806 len));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 left -= len;
808
809 fh->frag_off = htons(offset);
810 if (left > 0)
811 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700812 ipv6_hdr(frag)->payload_len = htons(frag->len -
813 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814
815 ptr += len;
816 offset += len;
817
818 /*
819 * Put this fragment into the sending queue.
820 */
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500821 err = output(net, sk, frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 if (err)
823 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700824
Eric Dumazetadf30902009-06-02 05:19:30 +0000825 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700826 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000828 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900829 IPSTATS_MIB_FRAGOKS);
Eric Dumazet808db802012-04-24 10:17:59 +0000830 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 return err;
832
Florian Westphal485fca62015-05-22 00:44:16 +0200833fail_toobig:
834 if (skb->sk && dst_allfrag(skb_dst(skb)))
835 sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK);
836
837 skb->dev = skb_dst(skb)->dev;
838 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
839 err = -EMSGSIZE;
840
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841fail:
Eric Dumazetadf30902009-06-02 05:19:30 +0000842 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900843 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900844 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 return err;
846}
847
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000848static inline int ip6_rt_check(const struct rt6key *rt_key,
849 const struct in6_addr *fl_addr,
850 const struct in6_addr *addr_cache)
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700851{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000852 return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
Ian Morris63159f22015-03-29 14:00:04 +0100853 (!addr_cache || !ipv6_addr_equal(fl_addr, addr_cache));
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700854}
855
Herbert Xu497c6152006-07-30 20:19:33 -0700856static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
857 struct dst_entry *dst,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000858 const struct flowi6 *fl6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859{
Herbert Xu497c6152006-07-30 20:19:33 -0700860 struct ipv6_pinfo *np = inet6_sk(sk);
Eric Dumazeta963a372013-06-26 04:15:07 -0700861 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862
Herbert Xu497c6152006-07-30 20:19:33 -0700863 if (!dst)
864 goto out;
865
Eric Dumazeta963a372013-06-26 04:15:07 -0700866 if (dst->ops->family != AF_INET6) {
867 dst_release(dst);
868 return NULL;
869 }
870
871 rt = (struct rt6_info *)dst;
Herbert Xu497c6152006-07-30 20:19:33 -0700872 /* Yes, checking route validity in not connected
873 * case is not very simple. Take into account,
874 * that we do not support routing by source, TOS,
Ian Morris67ba4152014-08-24 21:53:10 +0100875 * and MSG_DONTROUTE --ANK (980726)
Herbert Xu497c6152006-07-30 20:19:33 -0700876 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700877 * 1. ip6_rt_check(): If route was host route,
878 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700879 * If it is network route, we still may
880 * check its validity using saved pointer
881 * to the last used address: daddr_cache.
882 * We do not want to save whole address now,
883 * (because main consumer of this service
884 * is tcp, which has not this problem),
885 * so that the last trick works only on connected
886 * sockets.
887 * 2. oif also should be the same.
888 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500889 if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700890#ifdef CONFIG_IPV6_SUBTREES
David S. Miller4c9483b2011-03-12 16:22:43 -0500891 ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700892#endif
David Ahernca254492015-10-12 11:47:10 -0700893 (!(fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF) &&
894 (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex))) {
Herbert Xu497c6152006-07-30 20:19:33 -0700895 dst_release(dst);
896 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 }
898
Herbert Xu497c6152006-07-30 20:19:33 -0700899out:
900 return dst;
901}
902
Eric Dumazet3aef9342015-09-25 07:39:12 -0700903static int ip6_dst_lookup_tail(struct net *net, const struct sock *sk,
David S. Miller4c9483b2011-03-12 16:22:43 -0500904 struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700905{
David S. Miller69cce1d2011-07-17 23:09:49 -0700906#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
907 struct neighbour *n;
David S. Miller97cac082012-07-02 22:43:47 -0700908 struct rt6_info *rt;
David S. Miller69cce1d2011-07-17 23:09:49 -0700909#endif
910 int err;
Paolo Abeni6f21c962016-01-29 12:30:19 +0100911 int flags = 0;
Herbert Xu497c6152006-07-30 20:19:33 -0700912
Markus Stenberge16e8882015-05-05 13:36:59 +0300913 /* The correct way to handle this would be to do
914 * ip6_route_get_saddr, and then ip6_route_output; however,
915 * the route-specific preferred source forces the
916 * ip6_route_output call _before_ ip6_route_get_saddr.
917 *
918 * In source specific routing (no src=any default route),
919 * ip6_route_output will fail given src=any saddr, though, so
920 * that's why we try it again later.
921 */
922 if (ipv6_addr_any(&fl6->saddr) && (!*dst || !(*dst)->error)) {
923 struct rt6_info *rt;
924 bool had_dst = *dst != NULL;
925
926 if (!had_dst)
927 *dst = ip6_route_output(net, sk, fl6);
928 rt = (*dst)->error ? NULL : (struct rt6_info *)*dst;
929 err = ip6_route_get_saddr(net, rt, &fl6->daddr,
930 sk ? inet6_sk(sk)->srcprefs : 0,
931 &fl6->saddr);
932 if (err)
933 goto out_err_release;
934
935 /* If we had an erroneous initial result, pretend it
936 * never existed and let the SA-enabled version take
937 * over.
938 */
939 if (!had_dst && (*dst)->error) {
940 dst_release(*dst);
941 *dst = NULL;
942 }
Paolo Abeni6f21c962016-01-29 12:30:19 +0100943
944 if (fl6->flowi6_oif)
945 flags |= RT6_LOOKUP_F_IFACE;
Markus Stenberge16e8882015-05-05 13:36:59 +0300946 }
947
Ian Morris63159f22015-03-29 14:00:04 +0100948 if (!*dst)
Paolo Abeni6f21c962016-01-29 12:30:19 +0100949 *dst = ip6_route_output_flags(net, sk, fl6, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950
Ian Morrise5d08d72014-11-23 21:28:43 +0000951 err = (*dst)->error;
952 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953 goto out_err_release;
954
Neil Horman95c385b2007-04-25 17:08:10 -0700955#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700956 /*
957 * Here if the dst entry we've looked up
958 * has a neighbour entry that is in the INCOMPLETE
959 * state and the src address from the flow is
960 * marked as OPTIMISTIC, we release the found
961 * dst entry and replace it instead with the
962 * dst entry of the nexthop router
963 */
Eric Dumazetc56bf6f2012-07-06 09:19:05 +0200964 rt = (struct rt6_info *) *dst;
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000965 rcu_read_lock_bh();
Martin KaFai Lau2647a9b2015-05-22 20:55:58 -0700966 n = __ipv6_neigh_lookup_noref(rt->dst.dev,
967 rt6_nexthop(rt, &fl6->daddr));
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000968 err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0;
969 rcu_read_unlock_bh();
970
971 if (err) {
Neil Hormane550dfb2008-09-09 13:51:35 -0700972 struct inet6_ifaddr *ifp;
David S. Miller4c9483b2011-03-12 16:22:43 -0500973 struct flowi6 fl_gw6;
Neil Hormane550dfb2008-09-09 13:51:35 -0700974 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -0700975
David S. Miller4c9483b2011-03-12 16:22:43 -0500976 ifp = ipv6_get_ifaddr(net, &fl6->saddr,
Neil Hormane550dfb2008-09-09 13:51:35 -0700977 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -0700978
Neil Hormane550dfb2008-09-09 13:51:35 -0700979 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
980 if (ifp)
981 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -0700982
Neil Hormane550dfb2008-09-09 13:51:35 -0700983 if (redirect) {
984 /*
985 * We need to get the dst entry for the
986 * default router instead
987 */
988 dst_release(*dst);
David S. Miller4c9483b2011-03-12 16:22:43 -0500989 memcpy(&fl_gw6, fl6, sizeof(struct flowi6));
990 memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr));
991 *dst = ip6_route_output(net, sk, &fl_gw6);
Ian Morrise5d08d72014-11-23 21:28:43 +0000992 err = (*dst)->error;
993 if (err)
Neil Hormane550dfb2008-09-09 13:51:35 -0700994 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -0700995 }
Neil Hormane550dfb2008-09-09 13:51:35 -0700996 }
Neil Horman95c385b2007-04-25 17:08:10 -0700997#endif
998
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 return 0;
1000
1001out_err_release:
Mitsuru Chinenca46f9c2007-12-05 22:31:47 -08001002 if (err == -ENETUNREACH)
John Stultz5ac68e72013-10-07 15:52:01 -07001003 IP6_INC_STATS(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 dst_release(*dst);
1005 *dst = NULL;
1006 return err;
1007}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001008
Herbert Xu497c6152006-07-30 20:19:33 -07001009/**
1010 * ip6_dst_lookup - perform route lookup on flow
1011 * @sk: socket which provides route info
1012 * @dst: pointer to dst_entry * for result
David S. Miller4c9483b2011-03-12 16:22:43 -05001013 * @fl6: flow to lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001014 *
1015 * This function performs a route lookup on the given flow.
1016 *
1017 * It returns zero on success, or a standard errno code on error.
1018 */
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001019int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1020 struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -07001021{
1022 *dst = NULL;
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001023 return ip6_dst_lookup_tail(net, sk, dst, fl6);
Herbert Xu497c6152006-07-30 20:19:33 -07001024}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -08001025EXPORT_SYMBOL_GPL(ip6_dst_lookup);
1026
Herbert Xu497c6152006-07-30 20:19:33 -07001027/**
David S. Miller68d0c6d2011-03-01 13:19:07 -08001028 * ip6_dst_lookup_flow - perform route lookup on flow with ipsec
1029 * @sk: socket which provides route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001030 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001031 * @final_dst: final destination address for ipsec lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001032 *
1033 * This function performs a route lookup on the given flow.
1034 *
1035 * It returns a valid dst pointer on success, or a pointer encoded
1036 * error code.
1037 */
Eric Dumazet3aef9342015-09-25 07:39:12 -07001038struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001039 const struct in6_addr *final_dst)
David S. Miller68d0c6d2011-03-01 13:19:07 -08001040{
1041 struct dst_entry *dst = NULL;
1042 int err;
1043
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001044 err = ip6_dst_lookup_tail(sock_net(sk), sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001045 if (err)
1046 return ERR_PTR(err);
1047 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001048 fl6->daddr = *final_dst;
Phil Suttera0a9f332015-07-15 21:56:26 +02001049 if (!fl6->flowi6_oif)
David Ahernca254492015-10-12 11:47:10 -07001050 fl6->flowi6_oif = l3mdev_fib_oif(dst->dev);
David S. Miller2774c132011-03-01 14:59:04 -08001051
Steffen Klassertf92ee612014-09-16 10:08:40 +02001052 return xfrm_lookup_route(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001053}
1054EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow);
1055
1056/**
1057 * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow
1058 * @sk: socket which provides the dst cache and route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001059 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001060 * @final_dst: final destination address for ipsec lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001061 *
1062 * This function performs a route lookup on the given flow with the
1063 * possibility of using the cached route in the socket if it is valid.
1064 * It will take the socket dst lock when operating on the dst cache.
1065 * As a result, this function can only be used in process context.
1066 *
David S. Miller68d0c6d2011-03-01 13:19:07 -08001067 * It returns a valid dst pointer on success, or a pointer encoded
1068 * error code.
Herbert Xu497c6152006-07-30 20:19:33 -07001069 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001070struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001071 const struct in6_addr *final_dst)
Herbert Xu497c6152006-07-30 20:19:33 -07001072{
David S. Miller68d0c6d2011-03-01 13:19:07 -08001073 struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
1074 int err;
Herbert Xu497c6152006-07-30 20:19:33 -07001075
David S. Miller4c9483b2011-03-12 16:22:43 -05001076 dst = ip6_sk_dst_check(sk, dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001077
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001078 err = ip6_dst_lookup_tail(sock_net(sk), sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001079 if (err)
1080 return ERR_PTR(err);
1081 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001082 fl6->daddr = *final_dst;
David S. Miller2774c132011-03-01 14:59:04 -08001083
Steffen Klassertf92ee612014-09-16 10:08:40 +02001084 return xfrm_lookup_route(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
Herbert Xu497c6152006-07-30 20:19:33 -07001085}
David S. Miller68d0c6d2011-03-01 13:19:07 -08001086EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
Herbert Xu497c6152006-07-30 20:19:33 -07001087
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001088static inline int ip6_ufo_append_data(struct sock *sk,
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001089 struct sk_buff_head *queue,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001090 int getfrag(void *from, char *to, int offset, int len,
1091 int odd, struct sk_buff *skb),
1092 void *from, int length, int hh_len, int fragheaderlen,
Ian Morris67ba4152014-08-24 21:53:10 +01001093 int transhdrlen, int mtu, unsigned int flags,
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001094 const struct flowi6 *fl6)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001095
1096{
1097 struct sk_buff *skb;
1098 int err;
1099
1100 /* There is support for UDP large send offload by network
1101 * device, so create one single skb packet containing complete
1102 * udp datagram
1103 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001104 skb = skb_peek_tail(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001105 if (!skb) {
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001106 skb = sock_alloc_send_skb(sk,
1107 hh_len + fragheaderlen + transhdrlen + 20,
1108 (flags & MSG_DONTWAIT), &err);
Ian Morris63159f22015-03-29 14:00:04 +01001109 if (!skb)
Zheng Yan504744e2011-10-28 00:26:00 -04001110 return err;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001111
1112 /* reserve space for Hardware header */
1113 skb_reserve(skb, hh_len);
1114
1115 /* create space for UDP/IP header */
Ian Morris67ba4152014-08-24 21:53:10 +01001116 skb_put(skb, fragheaderlen + transhdrlen);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001117
1118 /* initialize network header pointer */
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001119 skb_reset_network_header(skb);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001120
1121 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001122 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001123
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001124 skb->protocol = htons(ETH_P_IPV6);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001125 skb->csum = 0;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001126
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001127 __skb_queue_tail(queue, skb);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001128 } else if (skb_is_gso(skb)) {
1129 goto append;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001130 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001131
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001132 skb->ip_summed = CHECKSUM_PARTIAL;
1133 /* Specify the length of each IPv6 datagram fragment.
1134 * It has to be a multiple of 8.
1135 */
1136 skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
1137 sizeof(struct frag_hdr)) & ~7;
1138 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001139 skb_shinfo(skb)->ip6_frag_id = ipv6_select_ident(sock_net(sk),
1140 &fl6->daddr,
1141 &fl6->saddr);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001142
1143append:
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001144 return skb_append_datato_frags(sk, skb, getfrag, from,
1145 (length - transhdrlen));
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001146}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147
Herbert Xu0178b692009-02-05 15:15:50 -08001148static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
1149 gfp_t gfp)
1150{
1151 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1152}
1153
1154static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
1155 gfp_t gfp)
1156{
1157 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1158}
1159
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001160static void ip6_append_data_mtu(unsigned int *mtu,
Gao feng0c183372012-05-26 01:30:53 +00001161 int *maxfraglen,
1162 unsigned int fragheaderlen,
1163 struct sk_buff *skb,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001164 struct rt6_info *rt,
luciene367c2d2014-03-17 12:51:01 +08001165 unsigned int orig_mtu)
Gao feng0c183372012-05-26 01:30:53 +00001166{
1167 if (!(rt->dst.flags & DST_XFRM_TUNNEL)) {
Ian Morris63159f22015-03-29 14:00:04 +01001168 if (!skb) {
Gao feng0c183372012-05-26 01:30:53 +00001169 /* first fragment, reserve header_len */
luciene367c2d2014-03-17 12:51:01 +08001170 *mtu = orig_mtu - rt->dst.header_len;
Gao feng0c183372012-05-26 01:30:53 +00001171
1172 } else {
1173 /*
1174 * this fragment is not first, the headers
1175 * space is regarded as data space.
1176 */
luciene367c2d2014-03-17 12:51:01 +08001177 *mtu = orig_mtu;
Gao feng0c183372012-05-26 01:30:53 +00001178 }
1179 *maxfraglen = ((*mtu - fragheaderlen) & ~7)
1180 + fragheaderlen - sizeof(struct frag_hdr);
1181 }
1182}
1183
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001184static int ip6_setup_cork(struct sock *sk, struct inet_cork_full *cork,
1185 struct inet6_cork *v6_cork,
1186 int hlimit, int tclass, struct ipv6_txoptions *opt,
1187 struct rt6_info *rt, struct flowi6 *fl6)
1188{
1189 struct ipv6_pinfo *np = inet6_sk(sk);
1190 unsigned int mtu;
1191
1192 /*
1193 * setup for corking
1194 */
1195 if (opt) {
1196 if (WARN_ON(v6_cork->opt))
1197 return -EINVAL;
1198
1199 v6_cork->opt = kzalloc(opt->tot_len, sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001200 if (unlikely(!v6_cork->opt))
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001201 return -ENOBUFS;
1202
1203 v6_cork->opt->tot_len = opt->tot_len;
1204 v6_cork->opt->opt_flen = opt->opt_flen;
1205 v6_cork->opt->opt_nflen = opt->opt_nflen;
1206
1207 v6_cork->opt->dst0opt = ip6_opt_dup(opt->dst0opt,
1208 sk->sk_allocation);
1209 if (opt->dst0opt && !v6_cork->opt->dst0opt)
1210 return -ENOBUFS;
1211
1212 v6_cork->opt->dst1opt = ip6_opt_dup(opt->dst1opt,
1213 sk->sk_allocation);
1214 if (opt->dst1opt && !v6_cork->opt->dst1opt)
1215 return -ENOBUFS;
1216
1217 v6_cork->opt->hopopt = ip6_opt_dup(opt->hopopt,
1218 sk->sk_allocation);
1219 if (opt->hopopt && !v6_cork->opt->hopopt)
1220 return -ENOBUFS;
1221
1222 v6_cork->opt->srcrt = ip6_rthdr_dup(opt->srcrt,
1223 sk->sk_allocation);
1224 if (opt->srcrt && !v6_cork->opt->srcrt)
1225 return -ENOBUFS;
1226
1227 /* need source address above miyazawa*/
1228 }
1229 dst_hold(&rt->dst);
1230 cork->base.dst = &rt->dst;
1231 cork->fl.u.ip6 = *fl6;
1232 v6_cork->hop_limit = hlimit;
1233 v6_cork->tclass = tclass;
1234 if (rt->dst.flags & DST_XFRM_TUNNEL)
1235 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
1236 rt->dst.dev->mtu : dst_mtu(&rt->dst);
1237 else
1238 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
1239 rt->dst.dev->mtu : dst_mtu(rt->dst.path);
1240 if (np->frag_size < mtu) {
1241 if (np->frag_size)
1242 mtu = np->frag_size;
1243 }
1244 cork->base.fragsize = mtu;
1245 if (dst_allfrag(rt->dst.path))
1246 cork->base.flags |= IPCORK_ALLFRAG;
1247 cork->base.length = 0;
1248
1249 return 0;
1250}
1251
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001252static int __ip6_append_data(struct sock *sk,
1253 struct flowi6 *fl6,
1254 struct sk_buff_head *queue,
1255 struct inet_cork *cork,
1256 struct inet6_cork *v6_cork,
1257 struct page_frag *pfrag,
1258 int getfrag(void *from, char *to, int offset,
1259 int len, int odd, struct sk_buff *skb),
1260 void *from, int length, int transhdrlen,
1261 unsigned int flags, int dontfrag)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262{
Gao feng0c183372012-05-26 01:30:53 +00001263 struct sk_buff *skb, *skb_prev = NULL;
luciene367c2d2014-03-17 12:51:01 +08001264 unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001265 int exthdrlen = 0;
1266 int dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 int hh_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 int copy;
1269 int err;
1270 int offset = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001271 __u8 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001272 u32 tskey = 0;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001273 struct rt6_info *rt = (struct rt6_info *)cork->dst;
1274 struct ipv6_txoptions *opt = v6_cork->opt;
Vlad Yasevich32dce962015-01-31 10:40:18 -05001275 int csummode = CHECKSUM_NONE;
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001276 unsigned int maxnonfragsize, headersize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001278 skb = skb_peek_tail(queue);
1279 if (!skb) {
1280 exthdrlen = opt ? opt->opt_flen : 0;
Romain KUNTZ7efdba52013-01-16 12:47:40 +00001281 dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 }
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001283
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001284 mtu = cork->fragsize;
luciene367c2d2014-03-17 12:51:01 +08001285 orig_mtu = mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286
Changli Gaod8d1f302010-06-10 23:31:35 -07001287 hh_len = LL_RESERVED_SPACE(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001289 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001290 (opt ? opt->opt_nflen : 0);
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001291 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen -
1292 sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001294 headersize = sizeof(struct ipv6hdr) +
1295 (opt ? opt->opt_flen + opt->opt_nflen : 0) +
1296 (dst_allfrag(&rt->dst) ?
1297 sizeof(struct frag_hdr) : 0) +
1298 rt->rt6i_nfheader_len;
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001299
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001300 if (cork->length + length > mtu - headersize && dontfrag &&
1301 (sk->sk_protocol == IPPROTO_UDP ||
1302 sk->sk_protocol == IPPROTO_RAW)) {
1303 ipv6_local_rxpmtu(sk, fl6, mtu - headersize +
1304 sizeof(struct ipv6hdr));
1305 goto emsgsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306 }
1307
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001308 if (ip6_sk_ignore_df(sk))
1309 maxnonfragsize = sizeof(struct ipv6hdr) + IPV6_MAXPLEN;
1310 else
1311 maxnonfragsize = mtu;
1312
1313 if (cork->length + length > maxnonfragsize - headersize) {
1314emsgsize:
1315 ipv6_local_error(sk, EMSGSIZE, fl6,
1316 mtu - headersize +
1317 sizeof(struct ipv6hdr));
1318 return -EMSGSIZE;
1319 }
1320
1321 /* CHECKSUM_PARTIAL only with no extension headers and when
1322 * we are not going to fragment
1323 */
1324 if (transhdrlen && sk->sk_protocol == IPPROTO_UDP &&
1325 headersize == sizeof(struct ipv6hdr) &&
1326 length < mtu - headersize &&
1327 !(flags & MSG_MORE) &&
Tom Herbertc8cd0982015-12-14 11:19:44 -08001328 rt->dst.dev->features & (NETIF_F_IPV6_CSUM | NETIF_F_HW_CSUM))
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001329 csummode = CHECKSUM_PARTIAL;
1330
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001331 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_RAW) {
Daniel Borkmannbf84a012013-04-14 08:08:13 +00001332 sock_tx_timestamp(sk, &tx_flags);
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001333 if (tx_flags & SKBTX_ANY_SW_TSTAMP &&
1334 sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)
1335 tskey = sk->sk_tskey++;
1336 }
Anders Berggrena693e692011-02-28 12:32:11 -08001337
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 /*
1339 * Let's try using as much space as possible.
1340 * Use MTU if total length of the message fits into the MTU.
1341 * Otherwise, we need to reserve fragment header and
1342 * fragment alignment (= 8-15 octects, in total).
1343 *
1344 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001345 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 * the message.
1347 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001348 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349 * at once if non-fragmentable extension headers
1350 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001351 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352 */
1353
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001354 cork->length += length;
1355 if (((length > mtu) ||
1356 (skb && skb_is_gso(skb))) &&
1357 (sk->sk_protocol == IPPROTO_UDP) &&
Michal Kubečekacf8dd02015-03-02 18:27:11 +01001358 (rt->dst.dev->features & NETIF_F_UFO) &&
Michal Kubeček40ba3302016-01-11 07:50:30 +01001359 (sk->sk_type == SOCK_DGRAM) && !udp_get_no_check6_tx(sk)) {
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001360 err = ip6_ufo_append_data(sk, queue, getfrag, from, length,
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001361 hh_len, fragheaderlen,
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001362 transhdrlen, mtu, flags, fl6);
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001363 if (err)
1364 goto error;
1365 return 0;
1366 }
1367
1368 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 goto alloc_new_skb;
1370
1371 while (length > 0) {
1372 /* Check if the remaining data fits into current packet. */
David S. Millerbdc712b2011-05-06 15:02:07 -07001373 copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 if (copy < length)
1375 copy = maxfraglen - skb->len;
1376
1377 if (copy <= 0) {
1378 char *data;
1379 unsigned int datalen;
1380 unsigned int fraglen;
1381 unsigned int fraggap;
1382 unsigned int alloclen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383alloc_new_skb:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 /* There's no room in the current skb */
Gao feng0c183372012-05-26 01:30:53 +00001385 if (skb)
1386 fraggap = skb->len - maxfraglen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 else
1388 fraggap = 0;
Gao feng0c183372012-05-26 01:30:53 +00001389 /* update mtu and maxfraglen if necessary */
Ian Morris63159f22015-03-29 14:00:04 +01001390 if (!skb || !skb_prev)
Gao feng0c183372012-05-26 01:30:53 +00001391 ip6_append_data_mtu(&mtu, &maxfraglen,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001392 fragheaderlen, skb, rt,
luciene367c2d2014-03-17 12:51:01 +08001393 orig_mtu);
Gao feng0c183372012-05-26 01:30:53 +00001394
1395 skb_prev = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396
1397 /*
1398 * If remaining data exceeds the mtu,
1399 * we know we need more fragment(s).
1400 */
1401 datalen = length + fraggap;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402
Gao feng0c183372012-05-26 01:30:53 +00001403 if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1404 datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 if ((flags & MSG_MORE) &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001406 !(rt->dst.dev->features&NETIF_F_SG))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 alloclen = mtu;
1408 else
1409 alloclen = datalen + fragheaderlen;
1410
Steffen Klassert299b0762011-10-11 01:43:33 +00001411 alloclen += dst_exthdrlen;
1412
Gao feng0c183372012-05-26 01:30:53 +00001413 if (datalen != length + fraggap) {
1414 /*
1415 * this is not the last fragment, the trailer
1416 * space is regarded as data space.
1417 */
1418 datalen += rt->dst.trailer_len;
1419 }
1420
1421 alloclen += rt->dst.trailer_len;
1422 fraglen = datalen + fragheaderlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423
1424 /*
1425 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001426 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427 * (without MSG_MORE) fits into the MTU.
1428 */
1429 alloclen += sizeof(struct frag_hdr);
1430
1431 if (transhdrlen) {
1432 skb = sock_alloc_send_skb(sk,
1433 alloclen + hh_len,
1434 (flags & MSG_DONTWAIT), &err);
1435 } else {
1436 skb = NULL;
1437 if (atomic_read(&sk->sk_wmem_alloc) <=
1438 2 * sk->sk_sndbuf)
1439 skb = sock_wmalloc(sk,
1440 alloclen + hh_len, 1,
1441 sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001442 if (unlikely(!skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 err = -ENOBUFS;
1444 }
Ian Morris63159f22015-03-29 14:00:04 +01001445 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446 goto error;
1447 /*
1448 * Fill in the control structures
1449 */
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001450 skb->protocol = htons(ETH_P_IPV6);
Vlad Yasevich32dce962015-01-31 10:40:18 -05001451 skb->ip_summed = csummode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452 skb->csum = 0;
Gao feng1f858512012-03-19 22:36:10 +00001453 /* reserve for fragmentation and ipsec header */
1454 skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
1455 dst_exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456
Willem de Bruijn11878b42014-07-14 17:55:06 -04001457 /* Only the initial fragment is time stamped */
1458 skb_shinfo(skb)->tx_flags = tx_flags;
1459 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001460 skb_shinfo(skb)->tskey = tskey;
1461 tskey = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001462
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 /*
1464 * Find where to start putting bytes
1465 */
Gao feng1f858512012-03-19 22:36:10 +00001466 data = skb_put(skb, fraglen);
1467 skb_set_network_header(skb, exthdrlen);
1468 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001469 skb->transport_header = (skb->network_header +
1470 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 if (fraggap) {
1472 skb->csum = skb_copy_and_csum_bits(
1473 skb_prev, maxfraglen,
1474 data + transhdrlen, fraggap, 0);
1475 skb_prev->csum = csum_sub(skb_prev->csum,
1476 skb->csum);
1477 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001478 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 }
1480 copy = datalen - transhdrlen - fraggap;
Steffen Klassert299b0762011-10-11 01:43:33 +00001481
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 if (copy < 0) {
1483 err = -EINVAL;
1484 kfree_skb(skb);
1485 goto error;
1486 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1487 err = -EFAULT;
1488 kfree_skb(skb);
1489 goto error;
1490 }
1491
1492 offset += copy;
1493 length -= datalen - fraggap;
1494 transhdrlen = 0;
1495 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001496 dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497
1498 /*
1499 * Put the packet on the pending queue
1500 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001501 __skb_queue_tail(queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 continue;
1503 }
1504
1505 if (copy > length)
1506 copy = length;
1507
Changli Gaod8d1f302010-06-10 23:31:35 -07001508 if (!(rt->dst.dev->features&NETIF_F_SG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 unsigned int off;
1510
1511 off = skb->len;
1512 if (getfrag(from, skb_put(skb, copy),
1513 offset, copy, off, skb) < 0) {
1514 __skb_trim(skb, off);
1515 err = -EFAULT;
1516 goto error;
1517 }
1518 } else {
1519 int i = skb_shinfo(skb)->nr_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520
Eric Dumazet5640f762012-09-23 23:04:42 +00001521 err = -ENOMEM;
1522 if (!sk_page_frag_refill(sk, pfrag))
1523 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524
Eric Dumazet5640f762012-09-23 23:04:42 +00001525 if (!skb_can_coalesce(skb, i, pfrag->page,
1526 pfrag->offset)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 err = -EMSGSIZE;
Eric Dumazet5640f762012-09-23 23:04:42 +00001528 if (i == MAX_SKB_FRAGS)
1529 goto error;
1530
1531 __skb_fill_page_desc(skb, i, pfrag->page,
1532 pfrag->offset, 0);
1533 skb_shinfo(skb)->nr_frags = ++i;
1534 get_page(pfrag->page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001536 copy = min_t(int, copy, pfrag->size - pfrag->offset);
Eric Dumazet9e903e02011-10-18 21:00:24 +00001537 if (getfrag(from,
Eric Dumazet5640f762012-09-23 23:04:42 +00001538 page_address(pfrag->page) + pfrag->offset,
1539 offset, copy, skb->len, skb) < 0)
1540 goto error_efault;
1541
1542 pfrag->offset += copy;
1543 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 skb->len += copy;
1545 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001546 skb->truesize += copy;
1547 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 }
1549 offset += copy;
1550 length -= copy;
1551 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001552
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553 return 0;
Eric Dumazet5640f762012-09-23 23:04:42 +00001554
1555error_efault:
1556 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557error:
David S. Millerbdc712b2011-05-06 15:02:07 -07001558 cork->length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001559 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 return err;
1561}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001562
1563int ip6_append_data(struct sock *sk,
1564 int getfrag(void *from, char *to, int offset, int len,
1565 int odd, struct sk_buff *skb),
1566 void *from, int length, int transhdrlen, int hlimit,
1567 int tclass, struct ipv6_txoptions *opt, struct flowi6 *fl6,
1568 struct rt6_info *rt, unsigned int flags, int dontfrag)
1569{
1570 struct inet_sock *inet = inet_sk(sk);
1571 struct ipv6_pinfo *np = inet6_sk(sk);
1572 int exthdrlen;
1573 int err;
1574
1575 if (flags&MSG_PROBE)
1576 return 0;
1577 if (skb_queue_empty(&sk->sk_write_queue)) {
1578 /*
1579 * setup for corking
1580 */
1581 err = ip6_setup_cork(sk, &inet->cork, &np->cork, hlimit,
1582 tclass, opt, rt, fl6);
1583 if (err)
1584 return err;
1585
1586 exthdrlen = (opt ? opt->opt_flen : 0);
1587 length += exthdrlen;
1588 transhdrlen += exthdrlen;
1589 } else {
1590 fl6 = &inet->cork.fl.u.ip6;
1591 transhdrlen = 0;
1592 }
1593
1594 return __ip6_append_data(sk, fl6, &sk->sk_write_queue, &inet->cork.base,
1595 &np->cork, sk_page_frag(sk), getfrag,
1596 from, length, transhdrlen, flags, dontfrag);
1597}
Chris Elstona495f832012-04-29 21:48:53 +00001598EXPORT_SYMBOL_GPL(ip6_append_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001600static void ip6_cork_release(struct inet_cork_full *cork,
1601 struct inet6_cork *v6_cork)
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001602{
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001603 if (v6_cork->opt) {
1604 kfree(v6_cork->opt->dst0opt);
1605 kfree(v6_cork->opt->dst1opt);
1606 kfree(v6_cork->opt->hopopt);
1607 kfree(v6_cork->opt->srcrt);
1608 kfree(v6_cork->opt);
1609 v6_cork->opt = NULL;
Herbert Xu0178b692009-02-05 15:15:50 -08001610 }
1611
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001612 if (cork->base.dst) {
1613 dst_release(cork->base.dst);
1614 cork->base.dst = NULL;
1615 cork->base.flags &= ~IPCORK_ALLFRAG;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001616 }
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001617 memset(&cork->fl, 0, sizeof(cork->fl));
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001618}
1619
Vlad Yasevich64223982015-01-31 10:40:15 -05001620struct sk_buff *__ip6_make_skb(struct sock *sk,
1621 struct sk_buff_head *queue,
1622 struct inet_cork_full *cork,
1623 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624{
1625 struct sk_buff *skb, *tmp_skb;
1626 struct sk_buff **tail_skb;
1627 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001629 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 struct ipv6hdr *hdr;
Vlad Yasevich64223982015-01-31 10:40:15 -05001631 struct ipv6_txoptions *opt = v6_cork->opt;
1632 struct rt6_info *rt = (struct rt6_info *)cork->base.dst;
1633 struct flowi6 *fl6 = &cork->fl.u.ip6;
David S. Miller4c9483b2011-03-12 16:22:43 -05001634 unsigned char proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635
Vlad Yasevich64223982015-01-31 10:40:15 -05001636 skb = __skb_dequeue(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001637 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 goto out;
1639 tail_skb = &(skb_shinfo(skb)->frag_list);
1640
1641 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001642 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001643 __skb_pull(skb, skb_network_offset(skb));
Vlad Yasevich64223982015-01-31 10:40:15 -05001644 while ((tmp_skb = __skb_dequeue(queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001645 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 *tail_skb = tmp_skb;
1647 tail_skb = &(tmp_skb->next);
1648 skb->len += tmp_skb->len;
1649 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 skb->truesize += tmp_skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 tmp_skb->destructor = NULL;
1652 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 }
1654
Herbert Xu28a894532008-02-12 18:07:27 -08001655 /* Allow local fragmentation. */
WANG Cong60ff7462014-05-04 16:39:18 -07001656 skb->ignore_df = ip6_sk_ignore_df(sk);
Herbert Xu28a894532008-02-12 18:07:27 -08001657
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001658 *final_dst = fl6->daddr;
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001659 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660 if (opt && opt->opt_flen)
1661 ipv6_push_frag_opts(skb, opt, &proto);
1662 if (opt && opt->opt_nflen)
1663 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1664
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001665 skb_push(skb, sizeof(struct ipv6hdr));
1666 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001667 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001668
Vlad Yasevich64223982015-01-31 10:40:15 -05001669 ip6_flow_hdr(hdr, v6_cork->tclass,
Tom Herbertcb1ce2e2014-07-01 21:33:10 -07001670 ip6_make_flowlabel(net, skb, fl6->flowlabel,
Tom Herbert67800f92015-07-31 16:52:11 -07001671 np->autoflowlabel, fl6));
Vlad Yasevich64223982015-01-31 10:40:15 -05001672 hdr->hop_limit = v6_cork->hop_limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 hdr->nexthdr = proto;
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001674 hdr->saddr = fl6->saddr;
1675 hdr->daddr = *final_dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676
Patrick McHardya2c20642006-01-08 22:37:26 -08001677 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001678 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001679
Changli Gaod8d1f302010-06-10 23:31:35 -07001680 skb_dst_set(skb, dst_clone(&rt->dst));
Neil Hormanedf391f2009-04-27 02:45:02 -07001681 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
David L Stevens14878f72007-09-16 16:52:35 -07001682 if (proto == IPPROTO_ICMPV6) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001683 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
David L Stevens14878f72007-09-16 16:52:35 -07001684
Hannes Frederic Sowa43a43b62014-03-31 20:14:10 +02001685 ICMP6MSGOUT_INC_STATS(net, idev, icmp6_hdr(skb)->icmp6_type);
1686 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001687 }
1688
Vlad Yasevich64223982015-01-31 10:40:15 -05001689 ip6_cork_release(cork, v6_cork);
1690out:
1691 return skb;
1692}
1693
1694int ip6_send_skb(struct sk_buff *skb)
1695{
1696 struct net *net = sock_net(skb->sk);
1697 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
1698 int err;
1699
Eric W. Biederman33224b12015-10-07 16:48:46 -05001700 err = ip6_local_out(net, skb->sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701 if (err) {
1702 if (err > 0)
Eric Dumazet6ce9e7b2009-09-02 18:05:33 -07001703 err = net_xmit_errno(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 if (err)
Vlad Yasevich64223982015-01-31 10:40:15 -05001705 IP6_INC_STATS(net, rt->rt6i_idev,
1706 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 }
1708
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709 return err;
Vlad Yasevich64223982015-01-31 10:40:15 -05001710}
1711
1712int ip6_push_pending_frames(struct sock *sk)
1713{
1714 struct sk_buff *skb;
1715
1716 skb = ip6_finish_skb(sk);
1717 if (!skb)
1718 return 0;
1719
1720 return ip6_send_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721}
Chris Elstona495f832012-04-29 21:48:53 +00001722EXPORT_SYMBOL_GPL(ip6_push_pending_frames);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001724static void __ip6_flush_pending_frames(struct sock *sk,
Vlad Yasevich64223982015-01-31 10:40:15 -05001725 struct sk_buff_head *queue,
1726 struct inet_cork_full *cork,
1727 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 struct sk_buff *skb;
1730
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001731 while ((skb = __skb_dequeue_tail(queue)) != NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001732 if (skb_dst(skb))
1733 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001734 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 kfree_skb(skb);
1736 }
1737
Vlad Yasevich64223982015-01-31 10:40:15 -05001738 ip6_cork_release(cork, v6_cork);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001740
1741void ip6_flush_pending_frames(struct sock *sk)
1742{
Vlad Yasevich64223982015-01-31 10:40:15 -05001743 __ip6_flush_pending_frames(sk, &sk->sk_write_queue,
1744 &inet_sk(sk)->cork, &inet6_sk(sk)->cork);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001745}
Chris Elstona495f832012-04-29 21:48:53 +00001746EXPORT_SYMBOL_GPL(ip6_flush_pending_frames);
Vlad Yasevich64223982015-01-31 10:40:15 -05001747
1748struct sk_buff *ip6_make_skb(struct sock *sk,
1749 int getfrag(void *from, char *to, int offset,
1750 int len, int odd, struct sk_buff *skb),
1751 void *from, int length, int transhdrlen,
1752 int hlimit, int tclass,
1753 struct ipv6_txoptions *opt, struct flowi6 *fl6,
1754 struct rt6_info *rt, unsigned int flags,
1755 int dontfrag)
1756{
1757 struct inet_cork_full cork;
1758 struct inet6_cork v6_cork;
1759 struct sk_buff_head queue;
1760 int exthdrlen = (opt ? opt->opt_flen : 0);
1761 int err;
1762
1763 if (flags & MSG_PROBE)
1764 return NULL;
1765
1766 __skb_queue_head_init(&queue);
1767
1768 cork.base.flags = 0;
1769 cork.base.addr = 0;
1770 cork.base.opt = NULL;
1771 v6_cork.opt = NULL;
1772 err = ip6_setup_cork(sk, &cork, &v6_cork, hlimit, tclass, opt, rt, fl6);
1773 if (err)
1774 return ERR_PTR(err);
1775
1776 if (dontfrag < 0)
1777 dontfrag = inet6_sk(sk)->dontfrag;
1778
1779 err = __ip6_append_data(sk, fl6, &queue, &cork.base, &v6_cork,
1780 &current->task_frag, getfrag, from,
1781 length + exthdrlen, transhdrlen + exthdrlen,
1782 flags, dontfrag);
1783 if (err) {
1784 __ip6_flush_pending_frames(sk, &queue, &cork, &v6_cork);
1785 return ERR_PTR(err);
1786 }
1787
1788 return __ip6_make_skb(sk, &queue, &cork, &v6_cork);
1789}