blob: 993fd9666f1b3178370ba1d6b7fa93e5792d9104 [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>
Roopa Prabhu14972cb2016-08-24 20:10:43 -070059#include <net/lwtunnel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -050061static int ip6_finish_output2(struct net *net, struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -070062{
Eric Dumazetadf30902009-06-02 05:19:30 +000063 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070064 struct net_device *dev = dst->dev;
David S. Millerf6b72b62011-07-14 07:53:20 -070065 struct neighbour *neigh;
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +000066 struct in6_addr *nexthop;
67 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070068
69 skb->protocol = htons(ETH_P_IPV6);
70 skb->dev = dev;
71
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070072 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) {
Eric Dumazetadf30902009-06-02 05:19:30 +000073 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
David Miller7026b1d2015-04-05 22:19:04 -040075 if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(sk) &&
Eric W. Biederman78126c42015-09-15 20:04:09 -050076 ((mroute6_socket(net, skb) &&
Benjamin Therybd91b8b2008-12-10 16:07:08 -080077 !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) ||
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090078 ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
79 &ipv6_hdr(skb)->saddr))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
81
82 /* Do not check for IFF_ALLMULTI; multicast routing
83 is not supported in any case.
84 */
85 if (newskb)
Jan Engelhardtb2e0b382010-03-23 04:09:07 +010086 NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING,
Eric W. Biederman29a26a52015-09-15 20:04:16 -050087 net, sk, newskb, NULL, newskb->dev,
Michel Machado95603e22012-06-12 10:16:35 +000088 dev_loopback_xmit);
Linus Torvalds1da177e2005-04-16 15:20:36 -070089
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070090 if (ipv6_hdr(skb)->hop_limit == 0) {
Eric W. Biederman78126c42015-09-15 20:04:09 -050091 IP6_INC_STATS(net, idev,
Denis V. Lunev3bd653c2008-10-08 10:54:51 -070092 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -070093 kfree_skb(skb);
94 return 0;
95 }
96 }
97
Eric W. Biederman78126c42015-09-15 20:04:09 -050098 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUTMCAST, skb->len);
Hannes Frederic Sowadd408512013-02-10 02:33:35 +000099
100 if (IPV6_ADDR_MC_SCOPE(&ipv6_hdr(skb)->daddr) <=
101 IPV6_ADDR_SCOPE_NODELOCAL &&
102 !(dev->flags & IFF_LOOPBACK)) {
103 kfree_skb(skb);
104 return 0;
105 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 }
107
Roopa Prabhu14972cb2016-08-24 20:10:43 -0700108 if (lwtunnel_xmit_redirect(dst->lwtstate)) {
109 int res = lwtunnel_xmit(skb);
110
111 if (res < 0 || res == LWTUNNEL_XMIT_DONE)
112 return res;
113 }
114
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000115 rcu_read_lock_bh();
Martin KaFai Lau2647a9b2015-05-22 20:55:58 -0700116 nexthop = rt6_nexthop((struct rt6_info *)dst, &ipv6_hdr(skb)->daddr);
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000117 neigh = __ipv6_neigh_lookup_noref(dst->dev, nexthop);
118 if (unlikely(!neigh))
119 neigh = __neigh_create(&nd_tbl, nexthop, dst->dev, false);
120 if (!IS_ERR(neigh)) {
121 ret = dst_neigh_output(dst, neigh, skb);
122 rcu_read_unlock_bh();
123 return ret;
124 }
125 rcu_read_unlock_bh();
David S. Miller05e3aa02011-07-16 17:26:00 -0700126
Eric W. Biederman78126c42015-09-15 20:04:09 -0500127 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200128 kfree_skb(skb);
129 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130}
131
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -0500132static int ip6_finish_output(struct net *net, struct sock *sk, struct sk_buff *skb)
Jan Engelhardt9e508492010-04-13 15:28:11 +0200133{
134 if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
Jiri Pirko9037c352013-11-06 17:52:19 +0100135 dst_allfrag(skb_dst(skb)) ||
136 (IP6CB(skb)->frag_max_size && skb->len > IP6CB(skb)->frag_max_size))
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500137 return ip6_fragment(net, sk, skb, ip6_finish_output2);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200138 else
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500139 return ip6_finish_output2(net, sk, skb);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200140}
141
Eric W. Biedermanede20592015-10-07 16:48:47 -0500142int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143{
Jan Engelhardt9e508492010-04-13 15:28:11 +0200144 struct net_device *dev = skb_dst(skb)->dev;
Eric Dumazetadf30902009-06-02 05:19:30 +0000145 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Eric W. Biedermanbe10de02015-09-17 17:21:31 -0500146
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900147 if (unlikely(idev->cnf.disable_ipv6)) {
Eric W. Biederman19a06442015-09-15 20:04:10 -0500148 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900149 kfree_skb(skb);
150 return 0;
151 }
152
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500153 return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING,
154 net, sk, skb, NULL, dev,
Jan Engelhardt9c6eb282010-04-13 15:32:16 +0200155 ip6_finish_output,
156 !(IP6CB(skb)->flags & IP6SKB_REROUTED));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157}
158
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159/*
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700160 * xmit an sk_buff (used by TCP, SCTP and DCCP)
161 * Note : socket lock is not held for SYNACK packets, but might be modified
162 * by calls to skb_set_owner_w() and ipv6_local_error(),
163 * which are using proper atomic operations or spinlocks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164 */
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700165int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
Eric Dumazetb903d322011-10-27 00:44:35 -0400166 struct ipv6_txoptions *opt, int tclass)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167{
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700168 struct net *net = sock_net(sk);
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700169 const struct ipv6_pinfo *np = inet6_sk(sk);
David S. Miller4c9483b2011-03-12 16:22:43 -0500170 struct in6_addr *first_hop = &fl6->daddr;
Eric Dumazetadf30902009-06-02 05:19:30 +0000171 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 struct ipv6hdr *hdr;
David S. Miller4c9483b2011-03-12 16:22:43 -0500173 u8 proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 int seg_len = skb->len;
Gerrit Renkere651f032009-08-09 08:12:48 +0000175 int hlimit = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 u32 mtu;
177
178 if (opt) {
Chuck Leverc2636b42007-10-23 21:07:32 -0700179 unsigned int head_room;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180
181 /* First: exthdrs may take lots of space (~8K for now)
182 MAX_HEADER is not enough.
183 */
184 head_room = opt->opt_nflen + opt->opt_flen;
185 seg_len += head_room;
186 head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
187
188 if (skb_headroom(skb) < head_room) {
189 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
Ian Morris63159f22015-03-29 14:00:04 +0100190 if (!skb2) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000191 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900192 IPSTATS_MIB_OUTDISCARDS);
193 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 return -ENOBUFS;
195 }
Eric Dumazet808db802012-04-24 10:17:59 +0000196 consume_skb(skb);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900197 skb = skb2;
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700198 /* skb_set_owner_w() changes sk->sk_wmem_alloc atomically,
199 * it is safe to call in our context (socket lock not held)
200 */
201 skb_set_owner_w(skb, (struct sock *)sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 }
203 if (opt->opt_flen)
204 ipv6_push_frag_opts(skb, opt, &proto);
205 if (opt->opt_nflen)
206 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
207 }
208
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700209 skb_push(skb, sizeof(struct ipv6hdr));
210 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700211 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212
213 /*
214 * Fill in the IPv6 header
215 */
Eric Dumazetb903d322011-10-27 00:44:35 -0400216 if (np)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 hlimit = np->hop_limit;
218 if (hlimit < 0)
YOSHIFUJI Hideaki6b75d092008-03-10 06:00:30 -0400219 hlimit = ip6_dst_hoplimit(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220
Tom Herbertcb1ce2e2014-07-01 21:33:10 -0700221 ip6_flow_hdr(hdr, tclass, ip6_make_flowlabel(net, skb, fl6->flowlabel,
Tom Herbert67800f92015-07-31 16:52:11 -0700222 np->autoflowlabel, fl6));
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 hdr->payload_len = htons(seg_len);
225 hdr->nexthdr = proto;
226 hdr->hop_limit = hlimit;
227
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000228 hdr->saddr = fl6->saddr;
229 hdr->daddr = *first_hop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +0200231 skb->protocol = htons(ETH_P_IPV6);
Patrick McHardya2c20642006-01-08 22:37:26 -0800232 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800233 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -0800234
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 mtu = dst_mtu(dst);
WANG Cong60ff7462014-05-04 16:39:18 -0700236 if ((skb->len <= mtu) || skb->ignore_df || skb_is_gso(skb)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000237 IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)),
Neil Hormanedf391f2009-04-27 02:45:02 -0700238 IPSTATS_MIB_OUT, skb->len);
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700239 /* hooks should never assume socket lock is held.
240 * we promote our socket to non const
241 */
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500242 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700243 net, (struct sock *)sk, skb, NULL, dst->dev,
Eric W. Biederman13206b62015-10-07 16:48:35 -0500244 dst_output);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 }
246
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 skb->dev = dst->dev;
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700248 /* ipv6_local_error() does not require socket lock,
249 * we promote our socket to non const
250 */
251 ipv6_local_error((struct sock *)sk, EMSGSIZE, fl6, mtu);
252
Eric Dumazetadf30902009-06-02 05:19:30 +0000253 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 kfree_skb(skb);
255 return -EMSGSIZE;
256}
YOSHIFUJI Hideaki71590392007-02-22 22:05:40 +0900257EXPORT_SYMBOL(ip6_xmit);
258
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
260{
261 struct ip6_ra_chain *ra;
262 struct sock *last = NULL;
263
264 read_lock(&ip6_ra_lock);
265 for (ra = ip6_ra_chain; ra; ra = ra->next) {
266 struct sock *sk = ra->sk;
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700267 if (sk && ra->sel == sel &&
268 (!sk->sk_bound_dev_if ||
269 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 if (last) {
271 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
272 if (skb2)
273 rawv6_rcv(last, skb2);
274 }
275 last = sk;
276 }
277 }
278
279 if (last) {
280 rawv6_rcv(last, skb);
281 read_unlock(&ip6_ra_lock);
282 return 1;
283 }
284 read_unlock(&ip6_ra_lock);
285 return 0;
286}
287
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700288static int ip6_forward_proxy_check(struct sk_buff *skb)
289{
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700290 struct ipv6hdr *hdr = ipv6_hdr(skb);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700291 u8 nexthdr = hdr->nexthdr;
Jesse Gross75f28112011-11-30 17:05:51 -0800292 __be16 frag_off;
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700293 int offset;
294
295 if (ipv6_ext_hdr(nexthdr)) {
Jesse Gross75f28112011-11-30 17:05:51 -0800296 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700297 if (offset < 0)
298 return 0;
299 } else
300 offset = sizeof(struct ipv6hdr);
301
302 if (nexthdr == IPPROTO_ICMPV6) {
303 struct icmp6hdr *icmp6;
304
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700305 if (!pskb_may_pull(skb, (skb_network_header(skb) +
306 offset + 1 - skb->data)))
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700307 return 0;
308
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700309 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700310
311 switch (icmp6->icmp6_type) {
312 case NDISC_ROUTER_SOLICITATION:
313 case NDISC_ROUTER_ADVERTISEMENT:
314 case NDISC_NEIGHBOUR_SOLICITATION:
315 case NDISC_NEIGHBOUR_ADVERTISEMENT:
316 case NDISC_REDIRECT:
317 /* For reaction involving unicast neighbor discovery
318 * message destined to the proxied address, pass it to
319 * input function.
320 */
321 return 1;
322 default:
323 break;
324 }
325 }
326
Ville Nuorvala74553b02006-09-22 14:42:18 -0700327 /*
328 * The proxying router can't forward traffic sent to a link-local
329 * address, so signal the sender and discard the packet. This
330 * behavior is clarified by the MIPv6 specification.
331 */
332 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
333 dst_link_failure(skb);
334 return -1;
335 }
336
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700337 return 0;
338}
339
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -0500340static inline int ip6_forward_finish(struct net *net, struct sock *sk,
341 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342{
Eric W. Biederman13206b62015-10-07 16:48:35 -0500343 return dst_output(net, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344}
345
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100346static unsigned int ip6_dst_mtu_forward(const struct dst_entry *dst)
347{
348 unsigned int mtu;
349 struct inet6_dev *idev;
350
351 if (dst_metric_locked(dst, RTAX_MTU)) {
352 mtu = dst_metric_raw(dst, RTAX_MTU);
353 if (mtu)
354 return mtu;
355 }
356
357 mtu = IPV6_MIN_MTU;
358 rcu_read_lock();
359 idev = __in6_dev_get(dst->dev);
360 if (idev)
361 mtu = idev->cnf.mtu6;
362 rcu_read_unlock();
363
364 return mtu;
365}
366
Florian Westphalfe6cc552014-02-13 23:09:12 +0100367static bool ip6_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
368{
Florian Westphal418a3152014-05-05 00:03:34 +0200369 if (skb->len <= mtu)
Florian Westphalfe6cc552014-02-13 23:09:12 +0100370 return false;
371
WANG Cong60ff7462014-05-04 16:39:18 -0700372 /* ipv6 conntrack defrag sets max_frag_size + ignore_df */
Florian Westphalfe6cc552014-02-13 23:09:12 +0100373 if (IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu)
374 return true;
375
WANG Cong60ff7462014-05-04 16:39:18 -0700376 if (skb->ignore_df)
Florian Westphal418a3152014-05-05 00:03:34 +0200377 return false;
378
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -0300379 if (skb_is_gso(skb) && skb_gso_validate_mtu(skb, mtu))
Florian Westphalfe6cc552014-02-13 23:09:12 +0100380 return false;
381
382 return true;
383}
384
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385int ip6_forward(struct sk_buff *skb)
386{
Eric Dumazetadf30902009-06-02 05:19:30 +0000387 struct dst_entry *dst = skb_dst(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700388 struct ipv6hdr *hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 struct inet6_skb_parm *opt = IP6CB(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900390 struct net *net = dev_net(dst->dev);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800391 u32 mtu;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900392
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700393 if (net->ipv6.devconf_all->forwarding == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 goto error;
395
Li RongQing090f1162014-03-11 10:40:08 +0800396 if (skb->pkt_type != PACKET_HOST)
397 goto drop;
398
Hannes Frederic Sowa9ef2e962015-10-08 18:19:53 +0200399 if (unlikely(skb->sk))
400 goto drop;
401
Ben Hutchings4497b072008-06-19 16:22:28 -0700402 if (skb_warn_if_lro(skb))
403 goto drop;
404
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700406 __IP6_INC_STATS(net, ip6_dst_idev(dst),
407 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 goto drop;
409 }
410
Herbert Xu35fc92a2007-03-26 23:22:20 -0700411 skb_forward_csum(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412
413 /*
414 * We DO NOT make any processing on
415 * RA packets, pushing them to user level AS IS
416 * without ane WARRANTY that application will be able
417 * to interpret them. The reason is that we
418 * cannot make anything clever here.
419 *
420 * We are not end-node, so that if packet contains
421 * AH/ESP, we cannot make anything.
422 * Defragmentation also would be mistake, RA packets
423 * cannot be fragmented, because there is no warranty
424 * that different fragments will go along one path. --ANK
425 */
YOSHIFUJI Hideaki / 吉藤英明ab4eb352013-06-22 11:13:13 +0900426 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
427 if (ip6_call_ra_chain(skb, ntohs(opt->ra)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 return 0;
429 }
430
431 /*
432 * check and decrement ttl
433 */
434 if (hdr->hop_limit <= 1) {
435 /* Force OUTPUT device used as source address */
436 skb->dev = dst->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000437 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0);
Eric Dumazet1d015502016-04-27 16:44:40 -0700438 __IP6_INC_STATS(net, ip6_dst_idev(dst),
439 IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440
441 kfree_skb(skb);
442 return -ETIMEDOUT;
443 }
444
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -0700445 /* XXX: idev->cnf.proxy_ndp? */
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700446 if (net->ipv6.devconf_all->proxy_ndp &&
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800447 pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
Ville Nuorvala74553b02006-09-22 14:42:18 -0700448 int proxied = ip6_forward_proxy_check(skb);
449 if (proxied > 0)
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700450 return ip6_input(skb);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700451 else if (proxied < 0) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700452 __IP6_INC_STATS(net, ip6_dst_idev(dst),
453 IPSTATS_MIB_INDISCARDS);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700454 goto drop;
455 }
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700456 }
457
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 if (!xfrm6_route_forward(skb)) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700459 __IP6_INC_STATS(net, ip6_dst_idev(dst),
460 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 goto drop;
462 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000463 dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464
465 /* IPv6 specs say nothing about it, but it is clear that we cannot
466 send redirects to source routed frames.
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900467 We don't send redirects to frames decapsulated from IPsec.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 */
David S. Millerc45a3df2012-01-27 15:32:19 -0800469 if (skb->dev == dst->dev && opt->srcrt == 0 && !skb_sec_path(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 struct in6_addr *target = NULL;
David S. Millerfbfe95a2012-06-08 23:24:18 -0700471 struct inet_peer *peer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473
474 /*
475 * incoming and outgoing devices are the same
476 * send a redirect.
477 */
478
479 rt = (struct rt6_info *) dst;
David S. Millerc45a3df2012-01-27 15:32:19 -0800480 if (rt->rt6i_flags & RTF_GATEWAY)
481 target = &rt->rt6i_gateway;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 else
483 target = &hdr->daddr;
484
Martin KaFai Laufd0273d2015-05-22 20:55:57 -0700485 peer = inet_getpeer_v6(net->ipv6.peers, &hdr->daddr, 1);
David S. Miller92d86822011-02-04 15:55:25 -0800486
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 /* Limit redirects both by destination (here)
488 and by source (inside ndisc_send_redirect)
489 */
David S. Millerfbfe95a2012-06-08 23:24:18 -0700490 if (inet_peer_xrlim_allow(peer, 1*HZ))
David S. Miller49919692012-01-27 15:30:48 -0800491 ndisc_send_redirect(skb, target);
David S. Miller1d861aa2012-07-10 03:58:16 -0700492 if (peer)
493 inet_putpeer(peer);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700494 } else {
495 int addrtype = ipv6_addr_type(&hdr->saddr);
496
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 /* This check is security critical. */
YOSHIFUJI Hideakif81b2e72008-06-25 16:55:26 +0900498 if (addrtype == IPV6_ADDR_ANY ||
499 addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK))
David L Stevens5bb1ab02007-05-09 13:53:44 -0700500 goto error;
501 if (addrtype & IPV6_ADDR_LINKLOCAL) {
502 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000503 ICMPV6_NOT_NEIGHBOUR, 0);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700504 goto error;
505 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 }
507
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100508 mtu = ip6_dst_mtu_forward(dst);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800509 if (mtu < IPV6_MIN_MTU)
510 mtu = IPV6_MIN_MTU;
511
Florian Westphalfe6cc552014-02-13 23:09:12 +0100512 if (ip6_pkt_too_big(skb, mtu)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 /* Again, force OUTPUT device used as source address */
514 skb->dev = dst->dev;
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800515 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazet1d015502016-04-27 16:44:40 -0700516 __IP6_INC_STATS(net, ip6_dst_idev(dst),
517 IPSTATS_MIB_INTOOBIGERRORS);
518 __IP6_INC_STATS(net, ip6_dst_idev(dst),
519 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 kfree_skb(skb);
521 return -EMSGSIZE;
522 }
523
524 if (skb_cow(skb, dst->dev->hard_header_len)) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700525 __IP6_INC_STATS(net, ip6_dst_idev(dst),
526 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 goto drop;
528 }
529
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700530 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531
532 /* Mangling hops number delayed to point after skb COW */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900533
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 hdr->hop_limit--;
535
Eric Dumazet1d015502016-04-27 16:44:40 -0700536 __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
537 __IP6_ADD_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len);
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500538 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
539 net, NULL, skb, skb->dev, dst->dev,
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800540 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541
542error:
Eric Dumazet1d015502016-04-27 16:44:40 -0700543 __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544drop:
545 kfree_skb(skb);
546 return -EINVAL;
547}
548
549static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
550{
551 to->pkt_type = from->pkt_type;
552 to->priority = from->priority;
553 to->protocol = from->protocol;
Eric Dumazetadf30902009-06-02 05:19:30 +0000554 skb_dst_drop(to);
555 skb_dst_set(to, dst_clone(skb_dst(from)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 to->dev = from->dev;
Thomas Graf82e91ff2006-11-09 15:19:14 -0800557 to->mark = from->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558
559#ifdef CONFIG_NET_SCHED
560 to->tc_index = from->tc_index;
561#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -0700562 nf_copy(to, from);
James Morris984bc162006-06-09 00:29:17 -0700563 skb_copy_secmark(to, from);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564}
565
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500566int ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
567 int (*output)(struct net *, struct sock *, struct sk_buff *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 struct sk_buff *frag;
Ian Morris67ba4152014-08-24 21:53:10 +0100570 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +0200571 struct ipv6_pinfo *np = skb->sk && !dev_recursion_level() ?
572 inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 struct ipv6hdr *tmp_hdr;
574 struct frag_hdr *fh;
575 unsigned int mtu, hlen, left, len;
Herbert Xua7ae1992011-11-18 02:20:04 +0000576 int hroom, troom;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700577 __be32 frag_id;
Ian Morris67ba4152014-08-24 21:53:10 +0100578 int ptr, offset = 0, err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 u8 *prevhdr, nexthdr = 0;
580
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 hlen = ip6_find_1stfragopt(skb, &prevhdr);
582 nexthdr = *prevhdr;
583
John Heffner628a5c52007-04-20 15:53:27 -0700584 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700585
586 /* We must not fragment if the socket is set to force MTU discovery
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800587 * or if the skb it not generated by a local socket.
John Heffnerb881ef72007-04-20 15:52:39 -0700588 */
Florian Westphal485fca62015-05-22 00:44:16 +0200589 if (unlikely(!skb->ignore_df && skb->len > mtu))
590 goto fail_toobig;
Eric Dumazeta34a1012012-05-18 21:51:44 +0000591
Florian Westphal485fca62015-05-22 00:44:16 +0200592 if (IP6CB(skb)->frag_max_size) {
593 if (IP6CB(skb)->frag_max_size > mtu)
594 goto fail_toobig;
595
596 /* don't send fragments larger than what we received */
597 mtu = IP6CB(skb)->frag_max_size;
598 if (mtu < IPV6_MIN_MTU)
599 mtu = IPV6_MIN_MTU;
John Heffnerb881ef72007-04-20 15:52:39 -0700600 }
601
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800602 if (np && np->frag_size < mtu) {
603 if (np->frag_size)
604 mtu = np->frag_size;
605 }
Hannes Frederic Sowa89bc7842015-10-28 13:21:04 +0100606 if (mtu < hlen + sizeof(struct frag_hdr) + 8)
Hannes Frederic Sowab72a2b02015-10-16 11:32:43 +0200607 goto fail_toobig;
Hannes Frederic Sowa1e0d69a2015-10-28 13:21:03 +0100608 mtu -= hlen + sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609
Martin KaFai Laufd0273d2015-05-22 20:55:57 -0700610 frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
611 &ipv6_hdr(skb)->saddr);
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700612
Hannes Frederic Sowa405c92f2015-10-27 22:40:42 +0100613 if (skb->ip_summed == CHECKSUM_PARTIAL &&
614 (err = skb_checksum_help(skb)))
615 goto fail;
616
Florian Westphal1d325d22015-09-16 17:26:14 +0200617 hroom = LL_RESERVED_SPACE(rt->dst.dev);
David S. Miller21dc3302010-08-23 00:13:46 -0700618 if (skb_has_frag_list(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 int first_len = skb_pagelen(skb);
Eric Dumazet3d130082010-09-21 08:47:45 +0000620 struct sk_buff *frag2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621
622 if (first_len - hlen > mtu ||
623 ((first_len - hlen) & 7) ||
Florian Westphal1d325d22015-09-16 17:26:14 +0200624 skb_cloned(skb) ||
625 skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 goto slow_path;
627
David S. Miller4d9092b2009-06-09 00:20:05 -0700628 skb_walk_frags(skb, frag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 /* Correct geometry. */
630 if (frag->len > mtu ||
631 ((frag->len & 7) && frag->next) ||
Florian Westphal1d325d22015-09-16 17:26:14 +0200632 skb_headroom(frag) < (hlen + hroom + sizeof(struct frag_hdr)))
Eric Dumazet3d130082010-09-21 08:47:45 +0000633 goto slow_path_clean;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 /* Partially cloned skb? */
636 if (skb_shared(frag))
Eric Dumazet3d130082010-09-21 08:47:45 +0000637 goto slow_path_clean;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700638
639 BUG_ON(frag->sk);
640 if (skb->sk) {
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700641 frag->sk = skb->sk;
642 frag->destructor = sock_wfree;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700643 }
Eric Dumazet3d130082010-09-21 08:47:45 +0000644 skb->truesize -= frag->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645 }
646
647 err = 0;
648 offset = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 /* BUILD HEADER */
650
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800651 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700652 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 if (!tmp_hdr) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000654 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700655 IPSTATS_MIB_FRAGFAILS);
Florian Westphal1d325d22015-09-16 17:26:14 +0200656 err = -ENOMEM;
657 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 }
Florian Westphal1d325d22015-09-16 17:26:14 +0200659 frag = skb_shinfo(skb)->frag_list;
660 skb_frag_list_init(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 __skb_pull(skb, hlen);
Ian Morris67ba4152014-08-24 21:53:10 +0100663 fh = (struct frag_hdr *)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700664 __skb_push(skb, hlen);
665 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700666 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 fh->nexthdr = nexthdr;
669 fh->reserved = 0;
670 fh->frag_off = htons(IP6_MF);
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700671 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672
673 first_len = skb_pagelen(skb);
674 skb->data_len = first_len - skb_headlen(skb);
675 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700676 ipv6_hdr(skb)->payload_len = htons(first_len -
677 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900678
Changli Gaod8d1f302010-06-10 23:31:35 -0700679 dst_hold(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680
681 for (;;) {
682 /* Prepare header of the next frame,
683 * before previous one went down. */
684 if (frag) {
685 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300686 skb_reset_transport_header(frag);
Ian Morris67ba4152014-08-24 21:53:10 +0100687 fh = (struct frag_hdr *)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700688 __skb_push(frag, hlen);
689 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700690 memcpy(skb_network_header(frag), tmp_hdr,
691 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 offset += skb->len - hlen - sizeof(struct frag_hdr);
693 fh->nexthdr = nexthdr;
694 fh->reserved = 0;
695 fh->frag_off = htons(offset);
Ian Morris53b24b82015-03-29 14:00:05 +0100696 if (frag->next)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 fh->frag_off |= htons(IP6_MF);
698 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700699 ipv6_hdr(frag)->payload_len =
700 htons(frag->len -
701 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 ip6_copy_metadata(frag, skb);
703 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900704
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500705 err = output(net, sk, skb);
Ian Morris67ba4152014-08-24 21:53:10 +0100706 if (!err)
Changli Gaod8d1f302010-06-10 23:31:35 -0700707 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700708 IPSTATS_MIB_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700709
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 if (err || !frag)
711 break;
712
713 skb = frag;
714 frag = skb->next;
715 skb->next = NULL;
716 }
717
Jesper Juhla51482b2005-11-08 09:41:34 -0800718 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719
720 if (err == 0) {
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_FRAGOKS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000723 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 return 0;
725 }
726
Florian Westphal46cfd722014-09-10 01:08:46 +0200727 kfree_skb_list(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728
Changli Gaod8d1f302010-06-10 23:31:35 -0700729 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700730 IPSTATS_MIB_FRAGFAILS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000731 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 return err;
Eric Dumazet3d130082010-09-21 08:47:45 +0000733
734slow_path_clean:
735 skb_walk_frags(skb, frag2) {
736 if (frag2 == frag)
737 break;
738 frag2->sk = NULL;
739 frag2->destructor = NULL;
740 skb->truesize += frag2->truesize;
741 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 }
743
744slow_path:
745 left = skb->len - hlen; /* Space per frame */
746 ptr = hlen; /* Where to start from */
747
748 /*
749 * Fragment the datagram.
750 */
751
752 *prevhdr = NEXTHDR_FRAGMENT;
Herbert Xua7ae1992011-11-18 02:20:04 +0000753 troom = rt->dst.dev->needed_tailroom;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754
755 /*
756 * Keep copying data until we run out.
757 */
Ian Morris67ba4152014-08-24 21:53:10 +0100758 while (left > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 len = left;
760 /* IF: it doesn't fit, use 'mtu' - the data space left */
761 if (len > mtu)
762 len = mtu;
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300763 /* IF: we are not sending up to and including the packet end
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 then align the next start on an eight byte boundary */
765 if (len < left) {
766 len &= ~7;
767 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768
Joe Perchescbffccc2014-11-05 14:39:21 -0800769 /* Allocate buffer */
770 frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) +
771 hroom + troom, GFP_ATOMIC);
772 if (!frag) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000773 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900774 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 err = -ENOMEM;
776 goto fail;
777 }
778
779 /*
780 * Set up data on packet
781 */
782
783 ip6_copy_metadata(frag, skb);
Herbert Xua7ae1992011-11-18 02:20:04 +0000784 skb_reserve(frag, hroom);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700786 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300787 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700788 frag->transport_header = (frag->network_header + hlen +
789 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790
791 /*
792 * Charge the memory for the fragment to any owner
793 * it might possess
794 */
795 if (skb->sk)
796 skb_set_owner_w(frag, skb->sk);
797
798 /*
799 * Copy the packet header into the new buffer.
800 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300801 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802
803 /*
804 * Build fragment header.
805 */
806 fh->nexthdr = nexthdr;
807 fh->reserved = 0;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700808 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809
810 /*
811 * Copy a block of the IP datagram.
812 */
Himangi Saraogie3f0b862014-07-12 01:57:17 +0530813 BUG_ON(skb_copy_bits(skb, ptr, skb_transport_header(frag),
814 len));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 left -= len;
816
817 fh->frag_off = htons(offset);
818 if (left > 0)
819 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700820 ipv6_hdr(frag)->payload_len = htons(frag->len -
821 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822
823 ptr += len;
824 offset += len;
825
826 /*
827 * Put this fragment into the sending queue.
828 */
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500829 err = output(net, sk, frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 if (err)
831 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700832
Eric Dumazetadf30902009-06-02 05:19:30 +0000833 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700834 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000836 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900837 IPSTATS_MIB_FRAGOKS);
Eric Dumazet808db802012-04-24 10:17:59 +0000838 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 return err;
840
Florian Westphal485fca62015-05-22 00:44:16 +0200841fail_toobig:
842 if (skb->sk && dst_allfrag(skb_dst(skb)))
843 sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK);
844
845 skb->dev = skb_dst(skb)->dev;
846 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
847 err = -EMSGSIZE;
848
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849fail:
Eric Dumazetadf30902009-06-02 05:19:30 +0000850 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900851 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900852 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 return err;
854}
855
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000856static inline int ip6_rt_check(const struct rt6key *rt_key,
857 const struct in6_addr *fl_addr,
858 const struct in6_addr *addr_cache)
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700859{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000860 return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
Ian Morris63159f22015-03-29 14:00:04 +0100861 (!addr_cache || !ipv6_addr_equal(fl_addr, addr_cache));
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700862}
863
Herbert Xu497c6152006-07-30 20:19:33 -0700864static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
865 struct dst_entry *dst,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000866 const struct flowi6 *fl6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867{
Herbert Xu497c6152006-07-30 20:19:33 -0700868 struct ipv6_pinfo *np = inet6_sk(sk);
Eric Dumazeta963a372013-06-26 04:15:07 -0700869 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870
Herbert Xu497c6152006-07-30 20:19:33 -0700871 if (!dst)
872 goto out;
873
Eric Dumazeta963a372013-06-26 04:15:07 -0700874 if (dst->ops->family != AF_INET6) {
875 dst_release(dst);
876 return NULL;
877 }
878
879 rt = (struct rt6_info *)dst;
Herbert Xu497c6152006-07-30 20:19:33 -0700880 /* Yes, checking route validity in not connected
881 * case is not very simple. Take into account,
882 * that we do not support routing by source, TOS,
Ian Morris67ba4152014-08-24 21:53:10 +0100883 * and MSG_DONTROUTE --ANK (980726)
Herbert Xu497c6152006-07-30 20:19:33 -0700884 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700885 * 1. ip6_rt_check(): If route was host route,
886 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700887 * If it is network route, we still may
888 * check its validity using saved pointer
889 * to the last used address: daddr_cache.
890 * We do not want to save whole address now,
891 * (because main consumer of this service
892 * is tcp, which has not this problem),
893 * so that the last trick works only on connected
894 * sockets.
895 * 2. oif also should be the same.
896 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500897 if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700898#ifdef CONFIG_IPV6_SUBTREES
David S. Miller4c9483b2011-03-12 16:22:43 -0500899 ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700900#endif
David Ahernca254492015-10-12 11:47:10 -0700901 (!(fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF) &&
902 (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex))) {
Herbert Xu497c6152006-07-30 20:19:33 -0700903 dst_release(dst);
904 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 }
906
Herbert Xu497c6152006-07-30 20:19:33 -0700907out:
908 return dst;
909}
910
Eric Dumazet3aef9342015-09-25 07:39:12 -0700911static int ip6_dst_lookup_tail(struct net *net, const struct sock *sk,
David S. Miller4c9483b2011-03-12 16:22:43 -0500912 struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700913{
David S. Miller69cce1d2011-07-17 23:09:49 -0700914#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
915 struct neighbour *n;
David S. Miller97cac082012-07-02 22:43:47 -0700916 struct rt6_info *rt;
David S. Miller69cce1d2011-07-17 23:09:49 -0700917#endif
918 int err;
Paolo Abeni6f21c962016-01-29 12:30:19 +0100919 int flags = 0;
Herbert Xu497c6152006-07-30 20:19:33 -0700920
David Ahern0d240e72016-06-16 16:24:25 -0700921 if (ipv6_addr_any(&fl6->saddr) && fl6->flowi6_oif &&
922 (!*dst || !(*dst)->error)) {
923 err = l3mdev_get_saddr6(net, sk, fl6);
924 if (err)
925 goto out_err;
926 }
927
Markus Stenberge16e8882015-05-05 13:36:59 +0300928 /* The correct way to handle this would be to do
929 * ip6_route_get_saddr, and then ip6_route_output; however,
930 * the route-specific preferred source forces the
931 * ip6_route_output call _before_ ip6_route_get_saddr.
932 *
933 * In source specific routing (no src=any default route),
934 * ip6_route_output will fail given src=any saddr, though, so
935 * that's why we try it again later.
936 */
937 if (ipv6_addr_any(&fl6->saddr) && (!*dst || !(*dst)->error)) {
938 struct rt6_info *rt;
939 bool had_dst = *dst != NULL;
940
941 if (!had_dst)
942 *dst = ip6_route_output(net, sk, fl6);
943 rt = (*dst)->error ? NULL : (struct rt6_info *)*dst;
944 err = ip6_route_get_saddr(net, rt, &fl6->daddr,
945 sk ? inet6_sk(sk)->srcprefs : 0,
946 &fl6->saddr);
947 if (err)
948 goto out_err_release;
949
950 /* If we had an erroneous initial result, pretend it
951 * never existed and let the SA-enabled version take
952 * over.
953 */
954 if (!had_dst && (*dst)->error) {
955 dst_release(*dst);
956 *dst = NULL;
957 }
Paolo Abeni6f21c962016-01-29 12:30:19 +0100958
959 if (fl6->flowi6_oif)
960 flags |= RT6_LOOKUP_F_IFACE;
Markus Stenberge16e8882015-05-05 13:36:59 +0300961 }
962
Ian Morris63159f22015-03-29 14:00:04 +0100963 if (!*dst)
Paolo Abeni6f21c962016-01-29 12:30:19 +0100964 *dst = ip6_route_output_flags(net, sk, fl6, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965
Ian Morrise5d08d72014-11-23 21:28:43 +0000966 err = (*dst)->error;
967 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 goto out_err_release;
969
Neil Horman95c385b2007-04-25 17:08:10 -0700970#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700971 /*
972 * Here if the dst entry we've looked up
973 * has a neighbour entry that is in the INCOMPLETE
974 * state and the src address from the flow is
975 * marked as OPTIMISTIC, we release the found
976 * dst entry and replace it instead with the
977 * dst entry of the nexthop router
978 */
Eric Dumazetc56bf6f2012-07-06 09:19:05 +0200979 rt = (struct rt6_info *) *dst;
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000980 rcu_read_lock_bh();
Martin KaFai Lau2647a9b2015-05-22 20:55:58 -0700981 n = __ipv6_neigh_lookup_noref(rt->dst.dev,
982 rt6_nexthop(rt, &fl6->daddr));
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000983 err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0;
984 rcu_read_unlock_bh();
985
986 if (err) {
Neil Hormane550dfb2008-09-09 13:51:35 -0700987 struct inet6_ifaddr *ifp;
David S. Miller4c9483b2011-03-12 16:22:43 -0500988 struct flowi6 fl_gw6;
Neil Hormane550dfb2008-09-09 13:51:35 -0700989 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -0700990
David S. Miller4c9483b2011-03-12 16:22:43 -0500991 ifp = ipv6_get_ifaddr(net, &fl6->saddr,
Neil Hormane550dfb2008-09-09 13:51:35 -0700992 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -0700993
Neil Hormane550dfb2008-09-09 13:51:35 -0700994 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
995 if (ifp)
996 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -0700997
Neil Hormane550dfb2008-09-09 13:51:35 -0700998 if (redirect) {
999 /*
1000 * We need to get the dst entry for the
1001 * default router instead
1002 */
1003 dst_release(*dst);
David S. Miller4c9483b2011-03-12 16:22:43 -05001004 memcpy(&fl_gw6, fl6, sizeof(struct flowi6));
1005 memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr));
1006 *dst = ip6_route_output(net, sk, &fl_gw6);
Ian Morrise5d08d72014-11-23 21:28:43 +00001007 err = (*dst)->error;
1008 if (err)
Neil Hormane550dfb2008-09-09 13:51:35 -07001009 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -07001010 }
Neil Hormane550dfb2008-09-09 13:51:35 -07001011 }
Neil Horman95c385b2007-04-25 17:08:10 -07001012#endif
1013
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 return 0;
1015
1016out_err_release:
1017 dst_release(*dst);
1018 *dst = NULL;
David Ahern0d240e72016-06-16 16:24:25 -07001019out_err:
1020 if (err == -ENETUNREACH)
1021 IP6_INC_STATS(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 return err;
1023}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001024
Herbert Xu497c6152006-07-30 20:19:33 -07001025/**
1026 * ip6_dst_lookup - perform route lookup on flow
1027 * @sk: socket which provides route info
1028 * @dst: pointer to dst_entry * for result
David S. Miller4c9483b2011-03-12 16:22:43 -05001029 * @fl6: flow to lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001030 *
1031 * This function performs a route lookup on the given flow.
1032 *
1033 * It returns zero on success, or a standard errno code on error.
1034 */
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001035int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1036 struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -07001037{
1038 *dst = NULL;
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001039 return ip6_dst_lookup_tail(net, sk, dst, fl6);
Herbert Xu497c6152006-07-30 20:19:33 -07001040}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -08001041EXPORT_SYMBOL_GPL(ip6_dst_lookup);
1042
Herbert Xu497c6152006-07-30 20:19:33 -07001043/**
David S. Miller68d0c6d2011-03-01 13:19:07 -08001044 * ip6_dst_lookup_flow - perform route lookup on flow with ipsec
1045 * @sk: socket which provides route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001046 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001047 * @final_dst: final destination address for ipsec lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001048 *
1049 * This function performs a route lookup on the given flow.
1050 *
1051 * It returns a valid dst pointer on success, or a pointer encoded
1052 * error code.
1053 */
Eric Dumazet3aef9342015-09-25 07:39:12 -07001054struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001055 const struct in6_addr *final_dst)
David S. Miller68d0c6d2011-03-01 13:19:07 -08001056{
1057 struct dst_entry *dst = NULL;
1058 int err;
1059
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001060 err = ip6_dst_lookup_tail(sock_net(sk), sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001061 if (err)
1062 return ERR_PTR(err);
1063 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001064 fl6->daddr = *final_dst;
Phil Suttera0a9f332015-07-15 21:56:26 +02001065 if (!fl6->flowi6_oif)
David Ahernca254492015-10-12 11:47:10 -07001066 fl6->flowi6_oif = l3mdev_fib_oif(dst->dev);
David S. Miller2774c132011-03-01 14:59:04 -08001067
Steffen Klassertf92ee612014-09-16 10:08:40 +02001068 return xfrm_lookup_route(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001069}
1070EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow);
1071
1072/**
1073 * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow
1074 * @sk: socket which provides the dst cache and route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001075 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001076 * @final_dst: final destination address for ipsec lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001077 *
1078 * This function performs a route lookup on the given flow with the
1079 * possibility of using the cached route in the socket if it is valid.
1080 * It will take the socket dst lock when operating on the dst cache.
1081 * As a result, this function can only be used in process context.
1082 *
David S. Miller68d0c6d2011-03-01 13:19:07 -08001083 * It returns a valid dst pointer on success, or a pointer encoded
1084 * error code.
Herbert Xu497c6152006-07-30 20:19:33 -07001085 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001086struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001087 const struct in6_addr *final_dst)
Herbert Xu497c6152006-07-30 20:19:33 -07001088{
David S. Miller68d0c6d2011-03-01 13:19:07 -08001089 struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
Herbert Xu497c6152006-07-30 20:19:33 -07001090
David S. Miller4c9483b2011-03-12 16:22:43 -05001091 dst = ip6_sk_dst_check(sk, dst, fl6);
Jakub Sitnicki00bc0ef2016-06-08 15:13:34 +02001092 if (!dst)
1093 dst = ip6_dst_lookup_flow(sk, fl6, final_dst);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001094
Jakub Sitnicki00bc0ef2016-06-08 15:13:34 +02001095 return dst;
Herbert Xu497c6152006-07-30 20:19:33 -07001096}
David S. Miller68d0c6d2011-03-01 13:19:07 -08001097EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
Herbert Xu497c6152006-07-30 20:19:33 -07001098
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001099static inline int ip6_ufo_append_data(struct sock *sk,
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001100 struct sk_buff_head *queue,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001101 int getfrag(void *from, char *to, int offset, int len,
1102 int odd, struct sk_buff *skb),
1103 void *from, int length, int hh_len, int fragheaderlen,
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001104 int exthdrlen, int transhdrlen, int mtu,
1105 unsigned int flags, const struct flowi6 *fl6)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001106
1107{
1108 struct sk_buff *skb;
1109 int err;
1110
1111 /* There is support for UDP large send offload by network
1112 * device, so create one single skb packet containing complete
1113 * udp datagram
1114 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001115 skb = skb_peek_tail(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001116 if (!skb) {
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001117 skb = sock_alloc_send_skb(sk,
1118 hh_len + fragheaderlen + transhdrlen + 20,
1119 (flags & MSG_DONTWAIT), &err);
Ian Morris63159f22015-03-29 14:00:04 +01001120 if (!skb)
Zheng Yan504744e2011-10-28 00:26:00 -04001121 return err;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001122
1123 /* reserve space for Hardware header */
1124 skb_reserve(skb, hh_len);
1125
1126 /* create space for UDP/IP header */
Ian Morris67ba4152014-08-24 21:53:10 +01001127 skb_put(skb, fragheaderlen + transhdrlen);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001128
1129 /* initialize network header pointer */
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001130 skb_set_network_header(skb, exthdrlen);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001131
1132 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001133 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001134
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001135 skb->protocol = htons(ETH_P_IPV6);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001136 skb->csum = 0;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001137
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001138 __skb_queue_tail(queue, skb);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001139 } else if (skb_is_gso(skb)) {
1140 goto append;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001141 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001142
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001143 skb->ip_summed = CHECKSUM_PARTIAL;
1144 /* Specify the length of each IPv6 datagram fragment.
1145 * It has to be a multiple of 8.
1146 */
1147 skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
1148 sizeof(struct frag_hdr)) & ~7;
1149 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001150 skb_shinfo(skb)->ip6_frag_id = ipv6_select_ident(sock_net(sk),
1151 &fl6->daddr,
1152 &fl6->saddr);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001153
1154append:
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001155 return skb_append_datato_frags(sk, skb, getfrag, from,
1156 (length - transhdrlen));
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001157}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158
Herbert Xu0178b692009-02-05 15:15:50 -08001159static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
1160 gfp_t gfp)
1161{
1162 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1163}
1164
1165static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
1166 gfp_t gfp)
1167{
1168 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1169}
1170
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001171static void ip6_append_data_mtu(unsigned int *mtu,
Gao feng0c183372012-05-26 01:30:53 +00001172 int *maxfraglen,
1173 unsigned int fragheaderlen,
1174 struct sk_buff *skb,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001175 struct rt6_info *rt,
luciene367c2d2014-03-17 12:51:01 +08001176 unsigned int orig_mtu)
Gao feng0c183372012-05-26 01:30:53 +00001177{
1178 if (!(rt->dst.flags & DST_XFRM_TUNNEL)) {
Ian Morris63159f22015-03-29 14:00:04 +01001179 if (!skb) {
Gao feng0c183372012-05-26 01:30:53 +00001180 /* first fragment, reserve header_len */
luciene367c2d2014-03-17 12:51:01 +08001181 *mtu = orig_mtu - rt->dst.header_len;
Gao feng0c183372012-05-26 01:30:53 +00001182
1183 } else {
1184 /*
1185 * this fragment is not first, the headers
1186 * space is regarded as data space.
1187 */
luciene367c2d2014-03-17 12:51:01 +08001188 *mtu = orig_mtu;
Gao feng0c183372012-05-26 01:30:53 +00001189 }
1190 *maxfraglen = ((*mtu - fragheaderlen) & ~7)
1191 + fragheaderlen - sizeof(struct frag_hdr);
1192 }
1193}
1194
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001195static int ip6_setup_cork(struct sock *sk, struct inet_cork_full *cork,
Wei Wang26879da2016-05-02 21:40:07 -07001196 struct inet6_cork *v6_cork, struct ipcm6_cookie *ipc6,
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001197 struct rt6_info *rt, struct flowi6 *fl6)
1198{
1199 struct ipv6_pinfo *np = inet6_sk(sk);
1200 unsigned int mtu;
Wei Wang26879da2016-05-02 21:40:07 -07001201 struct ipv6_txoptions *opt = ipc6->opt;
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001202
1203 /*
1204 * setup for corking
1205 */
1206 if (opt) {
1207 if (WARN_ON(v6_cork->opt))
1208 return -EINVAL;
1209
1210 v6_cork->opt = kzalloc(opt->tot_len, sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001211 if (unlikely(!v6_cork->opt))
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001212 return -ENOBUFS;
1213
1214 v6_cork->opt->tot_len = opt->tot_len;
1215 v6_cork->opt->opt_flen = opt->opt_flen;
1216 v6_cork->opt->opt_nflen = opt->opt_nflen;
1217
1218 v6_cork->opt->dst0opt = ip6_opt_dup(opt->dst0opt,
1219 sk->sk_allocation);
1220 if (opt->dst0opt && !v6_cork->opt->dst0opt)
1221 return -ENOBUFS;
1222
1223 v6_cork->opt->dst1opt = ip6_opt_dup(opt->dst1opt,
1224 sk->sk_allocation);
1225 if (opt->dst1opt && !v6_cork->opt->dst1opt)
1226 return -ENOBUFS;
1227
1228 v6_cork->opt->hopopt = ip6_opt_dup(opt->hopopt,
1229 sk->sk_allocation);
1230 if (opt->hopopt && !v6_cork->opt->hopopt)
1231 return -ENOBUFS;
1232
1233 v6_cork->opt->srcrt = ip6_rthdr_dup(opt->srcrt,
1234 sk->sk_allocation);
1235 if (opt->srcrt && !v6_cork->opt->srcrt)
1236 return -ENOBUFS;
1237
1238 /* need source address above miyazawa*/
1239 }
1240 dst_hold(&rt->dst);
1241 cork->base.dst = &rt->dst;
1242 cork->fl.u.ip6 = *fl6;
Wei Wang26879da2016-05-02 21:40:07 -07001243 v6_cork->hop_limit = ipc6->hlimit;
1244 v6_cork->tclass = ipc6->tclass;
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001245 if (rt->dst.flags & DST_XFRM_TUNNEL)
1246 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
1247 rt->dst.dev->mtu : dst_mtu(&rt->dst);
1248 else
1249 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
1250 rt->dst.dev->mtu : dst_mtu(rt->dst.path);
1251 if (np->frag_size < mtu) {
1252 if (np->frag_size)
1253 mtu = np->frag_size;
1254 }
1255 cork->base.fragsize = mtu;
1256 if (dst_allfrag(rt->dst.path))
1257 cork->base.flags |= IPCORK_ALLFRAG;
1258 cork->base.length = 0;
1259
1260 return 0;
1261}
1262
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001263static int __ip6_append_data(struct sock *sk,
1264 struct flowi6 *fl6,
1265 struct sk_buff_head *queue,
1266 struct inet_cork *cork,
1267 struct inet6_cork *v6_cork,
1268 struct page_frag *pfrag,
1269 int getfrag(void *from, char *to, int offset,
1270 int len, int odd, struct sk_buff *skb),
1271 void *from, int length, int transhdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001272 unsigned int flags, struct ipcm6_cookie *ipc6,
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001273 const struct sockcm_cookie *sockc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274{
Gao feng0c183372012-05-26 01:30:53 +00001275 struct sk_buff *skb, *skb_prev = NULL;
luciene367c2d2014-03-17 12:51:01 +08001276 unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001277 int exthdrlen = 0;
1278 int dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 int hh_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280 int copy;
1281 int err;
1282 int offset = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001283 __u8 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001284 u32 tskey = 0;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001285 struct rt6_info *rt = (struct rt6_info *)cork->dst;
1286 struct ipv6_txoptions *opt = v6_cork->opt;
Vlad Yasevich32dce962015-01-31 10:40:18 -05001287 int csummode = CHECKSUM_NONE;
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001288 unsigned int maxnonfragsize, headersize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001290 skb = skb_peek_tail(queue);
1291 if (!skb) {
1292 exthdrlen = opt ? opt->opt_flen : 0;
Romain KUNTZ7efdba5b2013-01-16 12:47:40 +00001293 dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294 }
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001295
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001296 mtu = cork->fragsize;
luciene367c2d2014-03-17 12:51:01 +08001297 orig_mtu = mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298
Changli Gaod8d1f302010-06-10 23:31:35 -07001299 hh_len = LL_RESERVED_SPACE(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001301 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001302 (opt ? opt->opt_nflen : 0);
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001303 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen -
1304 sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001306 headersize = sizeof(struct ipv6hdr) +
1307 (opt ? opt->opt_flen + opt->opt_nflen : 0) +
1308 (dst_allfrag(&rt->dst) ?
1309 sizeof(struct frag_hdr) : 0) +
1310 rt->rt6i_nfheader_len;
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001311
Wei Wang26879da2016-05-02 21:40:07 -07001312 if (cork->length + length > mtu - headersize && ipc6->dontfrag &&
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001313 (sk->sk_protocol == IPPROTO_UDP ||
1314 sk->sk_protocol == IPPROTO_RAW)) {
1315 ipv6_local_rxpmtu(sk, fl6, mtu - headersize +
1316 sizeof(struct ipv6hdr));
1317 goto emsgsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 }
1319
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001320 if (ip6_sk_ignore_df(sk))
1321 maxnonfragsize = sizeof(struct ipv6hdr) + IPV6_MAXPLEN;
1322 else
1323 maxnonfragsize = mtu;
1324
1325 if (cork->length + length > maxnonfragsize - headersize) {
1326emsgsize:
1327 ipv6_local_error(sk, EMSGSIZE, fl6,
1328 mtu - headersize +
1329 sizeof(struct ipv6hdr));
1330 return -EMSGSIZE;
1331 }
1332
1333 /* CHECKSUM_PARTIAL only with no extension headers and when
1334 * we are not going to fragment
1335 */
1336 if (transhdrlen && sk->sk_protocol == IPPROTO_UDP &&
1337 headersize == sizeof(struct ipv6hdr) &&
1338 length < mtu - headersize &&
1339 !(flags & MSG_MORE) &&
Tom Herbertc8cd0982015-12-14 11:19:44 -08001340 rt->dst.dev->features & (NETIF_F_IPV6_CSUM | NETIF_F_HW_CSUM))
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001341 csummode = CHECKSUM_PARTIAL;
1342
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001343 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_RAW) {
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001344 sock_tx_timestamp(sk, sockc->tsflags, &tx_flags);
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001345 if (tx_flags & SKBTX_ANY_SW_TSTAMP &&
1346 sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)
1347 tskey = sk->sk_tskey++;
1348 }
Anders Berggrena693e692011-02-28 12:32:11 -08001349
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 /*
1351 * Let's try using as much space as possible.
1352 * Use MTU if total length of the message fits into the MTU.
1353 * Otherwise, we need to reserve fragment header and
1354 * fragment alignment (= 8-15 octects, in total).
1355 *
1356 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001357 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 * the message.
1359 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001360 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 * at once if non-fragmentable extension headers
1362 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001363 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 */
1365
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001366 cork->length += length;
1367 if (((length > mtu) ||
1368 (skb && skb_is_gso(skb))) &&
1369 (sk->sk_protocol == IPPROTO_UDP) &&
Michal Kubečekacf8dd02015-03-02 18:27:11 +01001370 (rt->dst.dev->features & NETIF_F_UFO) &&
Michal Kubeček40ba3302016-01-11 07:50:30 +01001371 (sk->sk_type == SOCK_DGRAM) && !udp_get_no_check6_tx(sk)) {
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001372 err = ip6_ufo_append_data(sk, queue, getfrag, from, length,
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001373 hh_len, fragheaderlen, exthdrlen,
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001374 transhdrlen, mtu, flags, fl6);
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001375 if (err)
1376 goto error;
1377 return 0;
1378 }
1379
1380 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381 goto alloc_new_skb;
1382
1383 while (length > 0) {
1384 /* Check if the remaining data fits into current packet. */
David S. Millerbdc712b2011-05-06 15:02:07 -07001385 copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 if (copy < length)
1387 copy = maxfraglen - skb->len;
1388
1389 if (copy <= 0) {
1390 char *data;
1391 unsigned int datalen;
1392 unsigned int fraglen;
1393 unsigned int fraggap;
1394 unsigned int alloclen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395alloc_new_skb:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 /* There's no room in the current skb */
Gao feng0c183372012-05-26 01:30:53 +00001397 if (skb)
1398 fraggap = skb->len - maxfraglen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 else
1400 fraggap = 0;
Gao feng0c183372012-05-26 01:30:53 +00001401 /* update mtu and maxfraglen if necessary */
Ian Morris63159f22015-03-29 14:00:04 +01001402 if (!skb || !skb_prev)
Gao feng0c183372012-05-26 01:30:53 +00001403 ip6_append_data_mtu(&mtu, &maxfraglen,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001404 fragheaderlen, skb, rt,
luciene367c2d2014-03-17 12:51:01 +08001405 orig_mtu);
Gao feng0c183372012-05-26 01:30:53 +00001406
1407 skb_prev = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408
1409 /*
1410 * If remaining data exceeds the mtu,
1411 * we know we need more fragment(s).
1412 */
1413 datalen = length + fraggap;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414
Gao feng0c183372012-05-26 01:30:53 +00001415 if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1416 datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417 if ((flags & MSG_MORE) &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001418 !(rt->dst.dev->features&NETIF_F_SG))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 alloclen = mtu;
1420 else
1421 alloclen = datalen + fragheaderlen;
1422
Steffen Klassert299b0762011-10-11 01:43:33 +00001423 alloclen += dst_exthdrlen;
1424
Gao feng0c183372012-05-26 01:30:53 +00001425 if (datalen != length + fraggap) {
1426 /*
1427 * this is not the last fragment, the trailer
1428 * space is regarded as data space.
1429 */
1430 datalen += rt->dst.trailer_len;
1431 }
1432
1433 alloclen += rt->dst.trailer_len;
1434 fraglen = datalen + fragheaderlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435
1436 /*
1437 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001438 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 * (without MSG_MORE) fits into the MTU.
1440 */
1441 alloclen += sizeof(struct frag_hdr);
1442
1443 if (transhdrlen) {
1444 skb = sock_alloc_send_skb(sk,
1445 alloclen + hh_len,
1446 (flags & MSG_DONTWAIT), &err);
1447 } else {
1448 skb = NULL;
1449 if (atomic_read(&sk->sk_wmem_alloc) <=
1450 2 * sk->sk_sndbuf)
1451 skb = sock_wmalloc(sk,
1452 alloclen + hh_len, 1,
1453 sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001454 if (unlikely(!skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 err = -ENOBUFS;
1456 }
Ian Morris63159f22015-03-29 14:00:04 +01001457 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458 goto error;
1459 /*
1460 * Fill in the control structures
1461 */
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001462 skb->protocol = htons(ETH_P_IPV6);
Vlad Yasevich32dce962015-01-31 10:40:18 -05001463 skb->ip_summed = csummode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 skb->csum = 0;
Gao feng1f858512012-03-19 22:36:10 +00001465 /* reserve for fragmentation and ipsec header */
1466 skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
1467 dst_exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468
Willem de Bruijn11878b42014-07-14 17:55:06 -04001469 /* Only the initial fragment is time stamped */
1470 skb_shinfo(skb)->tx_flags = tx_flags;
1471 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001472 skb_shinfo(skb)->tskey = tskey;
1473 tskey = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001474
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 /*
1476 * Find where to start putting bytes
1477 */
Gao feng1f858512012-03-19 22:36:10 +00001478 data = skb_put(skb, fraglen);
1479 skb_set_network_header(skb, exthdrlen);
1480 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001481 skb->transport_header = (skb->network_header +
1482 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 if (fraggap) {
1484 skb->csum = skb_copy_and_csum_bits(
1485 skb_prev, maxfraglen,
1486 data + transhdrlen, fraggap, 0);
1487 skb_prev->csum = csum_sub(skb_prev->csum,
1488 skb->csum);
1489 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001490 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 }
1492 copy = datalen - transhdrlen - fraggap;
Steffen Klassert299b0762011-10-11 01:43:33 +00001493
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494 if (copy < 0) {
1495 err = -EINVAL;
1496 kfree_skb(skb);
1497 goto error;
1498 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1499 err = -EFAULT;
1500 kfree_skb(skb);
1501 goto error;
1502 }
1503
1504 offset += copy;
1505 length -= datalen - fraggap;
1506 transhdrlen = 0;
1507 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001508 dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509
1510 /*
1511 * Put the packet on the pending queue
1512 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001513 __skb_queue_tail(queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 continue;
1515 }
1516
1517 if (copy > length)
1518 copy = length;
1519
Changli Gaod8d1f302010-06-10 23:31:35 -07001520 if (!(rt->dst.dev->features&NETIF_F_SG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 unsigned int off;
1522
1523 off = skb->len;
1524 if (getfrag(from, skb_put(skb, copy),
1525 offset, copy, off, skb) < 0) {
1526 __skb_trim(skb, off);
1527 err = -EFAULT;
1528 goto error;
1529 }
1530 } else {
1531 int i = skb_shinfo(skb)->nr_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532
Eric Dumazet5640f762012-09-23 23:04:42 +00001533 err = -ENOMEM;
1534 if (!sk_page_frag_refill(sk, pfrag))
1535 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536
Eric Dumazet5640f762012-09-23 23:04:42 +00001537 if (!skb_can_coalesce(skb, i, pfrag->page,
1538 pfrag->offset)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 err = -EMSGSIZE;
Eric Dumazet5640f762012-09-23 23:04:42 +00001540 if (i == MAX_SKB_FRAGS)
1541 goto error;
1542
1543 __skb_fill_page_desc(skb, i, pfrag->page,
1544 pfrag->offset, 0);
1545 skb_shinfo(skb)->nr_frags = ++i;
1546 get_page(pfrag->page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001548 copy = min_t(int, copy, pfrag->size - pfrag->offset);
Eric Dumazet9e903e02011-10-18 21:00:24 +00001549 if (getfrag(from,
Eric Dumazet5640f762012-09-23 23:04:42 +00001550 page_address(pfrag->page) + pfrag->offset,
1551 offset, copy, skb->len, skb) < 0)
1552 goto error_efault;
1553
1554 pfrag->offset += copy;
1555 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556 skb->len += copy;
1557 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001558 skb->truesize += copy;
1559 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 }
1561 offset += copy;
1562 length -= copy;
1563 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001564
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 return 0;
Eric Dumazet5640f762012-09-23 23:04:42 +00001566
1567error_efault:
1568 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569error:
David S. Millerbdc712b2011-05-06 15:02:07 -07001570 cork->length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001571 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572 return err;
1573}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001574
1575int ip6_append_data(struct sock *sk,
1576 int getfrag(void *from, char *to, int offset, int len,
1577 int odd, struct sk_buff *skb),
Wei Wang26879da2016-05-02 21:40:07 -07001578 void *from, int length, int transhdrlen,
1579 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
1580 struct rt6_info *rt, unsigned int flags,
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001581 const struct sockcm_cookie *sockc)
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001582{
1583 struct inet_sock *inet = inet_sk(sk);
1584 struct ipv6_pinfo *np = inet6_sk(sk);
1585 int exthdrlen;
1586 int err;
1587
1588 if (flags&MSG_PROBE)
1589 return 0;
1590 if (skb_queue_empty(&sk->sk_write_queue)) {
1591 /*
1592 * setup for corking
1593 */
Wei Wang26879da2016-05-02 21:40:07 -07001594 err = ip6_setup_cork(sk, &inet->cork, &np->cork,
1595 ipc6, rt, fl6);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001596 if (err)
1597 return err;
1598
Wei Wang26879da2016-05-02 21:40:07 -07001599 exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001600 length += exthdrlen;
1601 transhdrlen += exthdrlen;
1602 } else {
1603 fl6 = &inet->cork.fl.u.ip6;
1604 transhdrlen = 0;
1605 }
1606
1607 return __ip6_append_data(sk, fl6, &sk->sk_write_queue, &inet->cork.base,
1608 &np->cork, sk_page_frag(sk), getfrag,
Wei Wang26879da2016-05-02 21:40:07 -07001609 from, length, transhdrlen, flags, ipc6, sockc);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001610}
Chris Elstona495f832012-04-29 21:48:53 +00001611EXPORT_SYMBOL_GPL(ip6_append_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001613static void ip6_cork_release(struct inet_cork_full *cork,
1614 struct inet6_cork *v6_cork)
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001615{
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001616 if (v6_cork->opt) {
1617 kfree(v6_cork->opt->dst0opt);
1618 kfree(v6_cork->opt->dst1opt);
1619 kfree(v6_cork->opt->hopopt);
1620 kfree(v6_cork->opt->srcrt);
1621 kfree(v6_cork->opt);
1622 v6_cork->opt = NULL;
Herbert Xu0178b692009-02-05 15:15:50 -08001623 }
1624
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001625 if (cork->base.dst) {
1626 dst_release(cork->base.dst);
1627 cork->base.dst = NULL;
1628 cork->base.flags &= ~IPCORK_ALLFRAG;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001629 }
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001630 memset(&cork->fl, 0, sizeof(cork->fl));
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001631}
1632
Vlad Yasevich64223982015-01-31 10:40:15 -05001633struct sk_buff *__ip6_make_skb(struct sock *sk,
1634 struct sk_buff_head *queue,
1635 struct inet_cork_full *cork,
1636 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637{
1638 struct sk_buff *skb, *tmp_skb;
1639 struct sk_buff **tail_skb;
1640 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001642 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 struct ipv6hdr *hdr;
Vlad Yasevich64223982015-01-31 10:40:15 -05001644 struct ipv6_txoptions *opt = v6_cork->opt;
1645 struct rt6_info *rt = (struct rt6_info *)cork->base.dst;
1646 struct flowi6 *fl6 = &cork->fl.u.ip6;
David S. Miller4c9483b2011-03-12 16:22:43 -05001647 unsigned char proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648
Vlad Yasevich64223982015-01-31 10:40:15 -05001649 skb = __skb_dequeue(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001650 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 goto out;
1652 tail_skb = &(skb_shinfo(skb)->frag_list);
1653
1654 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001655 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001656 __skb_pull(skb, skb_network_offset(skb));
Vlad Yasevich64223982015-01-31 10:40:15 -05001657 while ((tmp_skb = __skb_dequeue(queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001658 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 *tail_skb = tmp_skb;
1660 tail_skb = &(tmp_skb->next);
1661 skb->len += tmp_skb->len;
1662 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 skb->truesize += tmp_skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 tmp_skb->destructor = NULL;
1665 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 }
1667
Herbert Xu28a894532008-02-12 18:07:27 -08001668 /* Allow local fragmentation. */
WANG Cong60ff7462014-05-04 16:39:18 -07001669 skb->ignore_df = ip6_sk_ignore_df(sk);
Herbert Xu28a894532008-02-12 18:07:27 -08001670
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001671 *final_dst = fl6->daddr;
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001672 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 if (opt && opt->opt_flen)
1674 ipv6_push_frag_opts(skb, opt, &proto);
1675 if (opt && opt->opt_nflen)
1676 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1677
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001678 skb_push(skb, sizeof(struct ipv6hdr));
1679 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001680 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001681
Vlad Yasevich64223982015-01-31 10:40:15 -05001682 ip6_flow_hdr(hdr, v6_cork->tclass,
Tom Herbertcb1ce2e2014-07-01 21:33:10 -07001683 ip6_make_flowlabel(net, skb, fl6->flowlabel,
Tom Herbert67800f92015-07-31 16:52:11 -07001684 np->autoflowlabel, fl6));
Vlad Yasevich64223982015-01-31 10:40:15 -05001685 hdr->hop_limit = v6_cork->hop_limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686 hdr->nexthdr = proto;
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001687 hdr->saddr = fl6->saddr;
1688 hdr->daddr = *final_dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689
Patrick McHardya2c20642006-01-08 22:37:26 -08001690 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001691 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001692
Changli Gaod8d1f302010-06-10 23:31:35 -07001693 skb_dst_set(skb, dst_clone(&rt->dst));
Neil Hormanedf391f2009-04-27 02:45:02 -07001694 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
David L Stevens14878f72007-09-16 16:52:35 -07001695 if (proto == IPPROTO_ICMPV6) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001696 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
David L Stevens14878f72007-09-16 16:52:35 -07001697
Hannes Frederic Sowa43a43b62014-03-31 20:14:10 +02001698 ICMP6MSGOUT_INC_STATS(net, idev, icmp6_hdr(skb)->icmp6_type);
1699 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001700 }
1701
Vlad Yasevich64223982015-01-31 10:40:15 -05001702 ip6_cork_release(cork, v6_cork);
1703out:
1704 return skb;
1705}
1706
1707int ip6_send_skb(struct sk_buff *skb)
1708{
1709 struct net *net = sock_net(skb->sk);
1710 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
1711 int err;
1712
Eric W. Biederman33224b12015-10-07 16:48:46 -05001713 err = ip6_local_out(net, skb->sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 if (err) {
1715 if (err > 0)
Eric Dumazet6ce9e7b2009-09-02 18:05:33 -07001716 err = net_xmit_errno(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 if (err)
Vlad Yasevich64223982015-01-31 10:40:15 -05001718 IP6_INC_STATS(net, rt->rt6i_idev,
1719 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 }
1721
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 return err;
Vlad Yasevich64223982015-01-31 10:40:15 -05001723}
1724
1725int ip6_push_pending_frames(struct sock *sk)
1726{
1727 struct sk_buff *skb;
1728
1729 skb = ip6_finish_skb(sk);
1730 if (!skb)
1731 return 0;
1732
1733 return ip6_send_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734}
Chris Elstona495f832012-04-29 21:48:53 +00001735EXPORT_SYMBOL_GPL(ip6_push_pending_frames);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001737static void __ip6_flush_pending_frames(struct sock *sk,
Vlad Yasevich64223982015-01-31 10:40:15 -05001738 struct sk_buff_head *queue,
1739 struct inet_cork_full *cork,
1740 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742 struct sk_buff *skb;
1743
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001744 while ((skb = __skb_dequeue_tail(queue)) != NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001745 if (skb_dst(skb))
1746 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001747 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 kfree_skb(skb);
1749 }
1750
Vlad Yasevich64223982015-01-31 10:40:15 -05001751 ip6_cork_release(cork, v6_cork);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001753
1754void ip6_flush_pending_frames(struct sock *sk)
1755{
Vlad Yasevich64223982015-01-31 10:40:15 -05001756 __ip6_flush_pending_frames(sk, &sk->sk_write_queue,
1757 &inet_sk(sk)->cork, &inet6_sk(sk)->cork);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001758}
Chris Elstona495f832012-04-29 21:48:53 +00001759EXPORT_SYMBOL_GPL(ip6_flush_pending_frames);
Vlad Yasevich64223982015-01-31 10:40:15 -05001760
1761struct sk_buff *ip6_make_skb(struct sock *sk,
1762 int getfrag(void *from, char *to, int offset,
1763 int len, int odd, struct sk_buff *skb),
1764 void *from, int length, int transhdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001765 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
Vlad Yasevich64223982015-01-31 10:40:15 -05001766 struct rt6_info *rt, unsigned int flags,
Wei Wang26879da2016-05-02 21:40:07 -07001767 const struct sockcm_cookie *sockc)
Vlad Yasevich64223982015-01-31 10:40:15 -05001768{
1769 struct inet_cork_full cork;
1770 struct inet6_cork v6_cork;
1771 struct sk_buff_head queue;
Wei Wang26879da2016-05-02 21:40:07 -07001772 int exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0);
Vlad Yasevich64223982015-01-31 10:40:15 -05001773 int err;
1774
1775 if (flags & MSG_PROBE)
1776 return NULL;
1777
1778 __skb_queue_head_init(&queue);
1779
1780 cork.base.flags = 0;
1781 cork.base.addr = 0;
1782 cork.base.opt = NULL;
1783 v6_cork.opt = NULL;
Wei Wang26879da2016-05-02 21:40:07 -07001784 err = ip6_setup_cork(sk, &cork, &v6_cork, ipc6, rt, fl6);
Vlad Yasevich64223982015-01-31 10:40:15 -05001785 if (err)
1786 return ERR_PTR(err);
1787
Wei Wang26879da2016-05-02 21:40:07 -07001788 if (ipc6->dontfrag < 0)
1789 ipc6->dontfrag = inet6_sk(sk)->dontfrag;
Vlad Yasevich64223982015-01-31 10:40:15 -05001790
1791 err = __ip6_append_data(sk, fl6, &queue, &cork.base, &v6_cork,
1792 &current->task_frag, getfrag, from,
1793 length + exthdrlen, transhdrlen + exthdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001794 flags, ipc6, sockc);
Vlad Yasevich64223982015-01-31 10:40:15 -05001795 if (err) {
1796 __ip6_flush_pending_frames(sk, &queue, &cork, &v6_cork);
1797 return ERR_PTR(err);
1798 }
1799
1800 return __ip6_make_skb(sk, &queue, &cork, &v6_cork);
1801}