blob: 11407dd6bc7c09996c8e1675857bd4c966403043 [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
Ben Hutchings8b0d3e82018-01-22 20:06:42 +0000159bool ip6_autoflowlabel(struct net *net, const struct ipv6_pinfo *np)
Shaohua Lib3b56032017-12-20 12:10:21 -0800160{
161 if (!np->autoflowlabel_set)
162 return ip6_default_np_autolabel(net);
163 else
164 return np->autoflowlabel;
165}
166
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167/*
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700168 * xmit an sk_buff (used by TCP, SCTP and DCCP)
169 * Note : socket lock is not held for SYNACK packets, but might be modified
170 * by calls to skb_set_owner_w() and ipv6_local_error(),
171 * which are using proper atomic operations or spinlocks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 */
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700173int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
Pablo Neira0d4c19e2017-01-26 22:56:21 +0100174 __u32 mark, struct ipv6_txoptions *opt, int tclass)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175{
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700176 struct net *net = sock_net(sk);
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700177 const struct ipv6_pinfo *np = inet6_sk(sk);
David S. Miller4c9483b2011-03-12 16:22:43 -0500178 struct in6_addr *first_hop = &fl6->daddr;
Eric Dumazetadf30902009-06-02 05:19:30 +0000179 struct dst_entry *dst = skb_dst(skb);
Stefano Brivio901936a2018-12-06 19:30:36 +0100180 unsigned int head_room;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 struct ipv6hdr *hdr;
David S. Miller4c9483b2011-03-12 16:22:43 -0500182 u8 proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 int seg_len = skb->len;
Gerrit Renkere651f032009-08-09 08:12:48 +0000184 int hlimit = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 u32 mtu;
186
Stefano Brivio901936a2018-12-06 19:30:36 +0100187 head_room = sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
188 if (opt)
189 head_room += opt->opt_nflen + opt->opt_flen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190
Stefano Brivio901936a2018-12-06 19:30:36 +0100191 if (unlikely(skb_headroom(skb) < head_room)) {
192 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
193 if (!skb2) {
194 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
195 IPSTATS_MIB_OUTDISCARDS);
196 kfree_skb(skb);
197 return -ENOBUFS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 }
Stefano Brivio901936a2018-12-06 19:30:36 +0100199 if (skb->sk)
200 skb_set_owner_w(skb2, skb->sk);
201 consume_skb(skb);
202 skb = skb2;
203 }
204
205 if (opt) {
206 seg_len += opt->opt_nflen + opt->opt_flen;
207
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 if (opt->opt_flen)
209 ipv6_push_frag_opts(skb, opt, &proto);
Stefano Brivio901936a2018-12-06 19:30:36 +0100210
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 if (opt->opt_nflen)
212 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
213 }
214
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700215 skb_push(skb, sizeof(struct ipv6hdr));
216 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700217 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218
219 /*
220 * Fill in the IPv6 header
221 */
Eric Dumazetb903d322011-10-27 00:44:35 -0400222 if (np)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 hlimit = np->hop_limit;
224 if (hlimit < 0)
YOSHIFUJI Hideaki6b75d092008-03-10 06:00:30 -0400225 hlimit = ip6_dst_hoplimit(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226
Tom Herbertcb1ce2e2014-07-01 21:33:10 -0700227 ip6_flow_hdr(hdr, tclass, ip6_make_flowlabel(net, skb, fl6->flowlabel,
Shaohua Lib3b56032017-12-20 12:10:21 -0800228 ip6_autoflowlabel(net, np), fl6));
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900229
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 hdr->payload_len = htons(seg_len);
231 hdr->nexthdr = proto;
232 hdr->hop_limit = hlimit;
233
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000234 hdr->saddr = fl6->saddr;
235 hdr->daddr = *first_hop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +0200237 skb->protocol = htons(ETH_P_IPV6);
Patrick McHardya2c20642006-01-08 22:37:26 -0800238 skb->priority = sk->sk_priority;
Pablo Neira0d4c19e2017-01-26 22:56:21 +0100239 skb->mark = mark;
Patrick McHardya2c20642006-01-08 22:37:26 -0800240
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 mtu = dst_mtu(dst);
WANG Cong60ff7462014-05-04 16:39:18 -0700242 if ((skb->len <= mtu) || skb->ignore_df || skb_is_gso(skb)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000243 IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)),
Neil Hormanedf391f2009-04-27 02:45:02 -0700244 IPSTATS_MIB_OUT, skb->len);
David Aherna8e3e1a2016-09-10 12:09:53 -0700245
246 /* if egress device is enslaved to an L3 master device pass the
247 * skb to its handler for processing
248 */
249 skb = l3mdev_ip6_out((struct sock *)sk, skb);
250 if (unlikely(!skb))
251 return 0;
252
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700253 /* hooks should never assume socket lock is held.
254 * we promote our socket to non const
255 */
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500256 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700257 net, (struct sock *)sk, skb, NULL, dst->dev,
Eric W. Biederman13206b62015-10-07 16:48:35 -0500258 dst_output);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 }
260
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 skb->dev = dst->dev;
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700262 /* ipv6_local_error() does not require socket lock,
263 * we promote our socket to non const
264 */
265 ipv6_local_error((struct sock *)sk, EMSGSIZE, fl6, mtu);
266
Eric Dumazetadf30902009-06-02 05:19:30 +0000267 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 kfree_skb(skb);
269 return -EMSGSIZE;
270}
YOSHIFUJI Hideaki71590392007-02-22 22:05:40 +0900271EXPORT_SYMBOL(ip6_xmit);
272
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
274{
275 struct ip6_ra_chain *ra;
276 struct sock *last = NULL;
277
278 read_lock(&ip6_ra_lock);
279 for (ra = ip6_ra_chain; ra; ra = ra->next) {
280 struct sock *sk = ra->sk;
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700281 if (sk && ra->sel == sel &&
282 (!sk->sk_bound_dev_if ||
283 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 if (last) {
285 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
286 if (skb2)
287 rawv6_rcv(last, skb2);
288 }
289 last = sk;
290 }
291 }
292
293 if (last) {
294 rawv6_rcv(last, skb);
295 read_unlock(&ip6_ra_lock);
296 return 1;
297 }
298 read_unlock(&ip6_ra_lock);
299 return 0;
300}
301
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700302static int ip6_forward_proxy_check(struct sk_buff *skb)
303{
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700304 struct ipv6hdr *hdr = ipv6_hdr(skb);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700305 u8 nexthdr = hdr->nexthdr;
Jesse Gross75f28112011-11-30 17:05:51 -0800306 __be16 frag_off;
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700307 int offset;
308
309 if (ipv6_ext_hdr(nexthdr)) {
Jesse Gross75f28112011-11-30 17:05:51 -0800310 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700311 if (offset < 0)
312 return 0;
313 } else
314 offset = sizeof(struct ipv6hdr);
315
316 if (nexthdr == IPPROTO_ICMPV6) {
317 struct icmp6hdr *icmp6;
318
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700319 if (!pskb_may_pull(skb, (skb_network_header(skb) +
320 offset + 1 - skb->data)))
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700321 return 0;
322
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700323 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700324
325 switch (icmp6->icmp6_type) {
326 case NDISC_ROUTER_SOLICITATION:
327 case NDISC_ROUTER_ADVERTISEMENT:
328 case NDISC_NEIGHBOUR_SOLICITATION:
329 case NDISC_NEIGHBOUR_ADVERTISEMENT:
330 case NDISC_REDIRECT:
331 /* For reaction involving unicast neighbor discovery
332 * message destined to the proxied address, pass it to
333 * input function.
334 */
335 return 1;
336 default:
337 break;
338 }
339 }
340
Ville Nuorvala74553b02006-09-22 14:42:18 -0700341 /*
342 * The proxying router can't forward traffic sent to a link-local
343 * address, so signal the sender and discard the packet. This
344 * behavior is clarified by the MIPv6 specification.
345 */
346 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
347 dst_link_failure(skb);
348 return -1;
349 }
350
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700351 return 0;
352}
353
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -0500354static inline int ip6_forward_finish(struct net *net, struct sock *sk,
355 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356{
Jeff Barnhill584d5412018-04-05 21:29:47 +0000357 struct dst_entry *dst = skb_dst(skb);
358
359 __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
360 __IP6_ADD_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len);
361
Eric W. Biederman13206b62015-10-07 16:48:35 -0500362 return dst_output(net, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363}
364
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100365static unsigned int ip6_dst_mtu_forward(const struct dst_entry *dst)
366{
367 unsigned int mtu;
368 struct inet6_dev *idev;
369
370 if (dst_metric_locked(dst, RTAX_MTU)) {
371 mtu = dst_metric_raw(dst, RTAX_MTU);
372 if (mtu)
373 return mtu;
374 }
375
376 mtu = IPV6_MIN_MTU;
377 rcu_read_lock();
378 idev = __in6_dev_get(dst->dev);
379 if (idev)
380 mtu = idev->cnf.mtu6;
381 rcu_read_unlock();
382
383 return mtu;
384}
385
Florian Westphalfe6cc552014-02-13 23:09:12 +0100386static bool ip6_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
387{
Florian Westphal418a3152014-05-05 00:03:34 +0200388 if (skb->len <= mtu)
Florian Westphalfe6cc552014-02-13 23:09:12 +0100389 return false;
390
WANG Cong60ff7462014-05-04 16:39:18 -0700391 /* ipv6 conntrack defrag sets max_frag_size + ignore_df */
Florian Westphalfe6cc552014-02-13 23:09:12 +0100392 if (IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu)
393 return true;
394
WANG Cong60ff7462014-05-04 16:39:18 -0700395 if (skb->ignore_df)
Florian Westphal418a3152014-05-05 00:03:34 +0200396 return false;
397
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -0300398 if (skb_is_gso(skb) && skb_gso_validate_mtu(skb, mtu))
Florian Westphalfe6cc552014-02-13 23:09:12 +0100399 return false;
400
401 return true;
402}
403
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404int ip6_forward(struct sk_buff *skb)
405{
Eric Dumazetadf30902009-06-02 05:19:30 +0000406 struct dst_entry *dst = skb_dst(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700407 struct ipv6hdr *hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 struct inet6_skb_parm *opt = IP6CB(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900409 struct net *net = dev_net(dst->dev);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800410 u32 mtu;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900411
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700412 if (net->ipv6.devconf_all->forwarding == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 goto error;
414
Li RongQing090f1162014-03-11 10:40:08 +0800415 if (skb->pkt_type != PACKET_HOST)
416 goto drop;
417
Hannes Frederic Sowa9ef2e962015-10-08 18:19:53 +0200418 if (unlikely(skb->sk))
419 goto drop;
420
Ben Hutchings4497b072008-06-19 16:22:28 -0700421 if (skb_warn_if_lro(skb))
422 goto drop;
423
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700425 __IP6_INC_STATS(net, ip6_dst_idev(dst),
426 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 goto drop;
428 }
429
Herbert Xu35fc92a2007-03-26 23:22:20 -0700430 skb_forward_csum(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431
432 /*
433 * We DO NOT make any processing on
434 * RA packets, pushing them to user level AS IS
435 * without ane WARRANTY that application will be able
436 * to interpret them. The reason is that we
437 * cannot make anything clever here.
438 *
439 * We are not end-node, so that if packet contains
440 * AH/ESP, we cannot make anything.
441 * Defragmentation also would be mistake, RA packets
442 * cannot be fragmented, because there is no warranty
443 * that different fragments will go along one path. --ANK
444 */
YOSHIFUJI Hideaki / 吉藤英明ab4eb352013-06-22 11:13:13 +0900445 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
446 if (ip6_call_ra_chain(skb, ntohs(opt->ra)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 return 0;
448 }
449
450 /*
451 * check and decrement ttl
452 */
453 if (hdr->hop_limit <= 1) {
454 /* Force OUTPUT device used as source address */
455 skb->dev = dst->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000456 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0);
Eric Dumazet1d015502016-04-27 16:44:40 -0700457 __IP6_INC_STATS(net, ip6_dst_idev(dst),
458 IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459
460 kfree_skb(skb);
461 return -ETIMEDOUT;
462 }
463
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -0700464 /* XXX: idev->cnf.proxy_ndp? */
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700465 if (net->ipv6.devconf_all->proxy_ndp &&
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800466 pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
Ville Nuorvala74553b02006-09-22 14:42:18 -0700467 int proxied = ip6_forward_proxy_check(skb);
468 if (proxied > 0)
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700469 return ip6_input(skb);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700470 else if (proxied < 0) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700471 __IP6_INC_STATS(net, ip6_dst_idev(dst),
472 IPSTATS_MIB_INDISCARDS);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700473 goto drop;
474 }
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700475 }
476
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 if (!xfrm6_route_forward(skb)) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700478 __IP6_INC_STATS(net, ip6_dst_idev(dst),
479 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 goto drop;
481 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000482 dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483
484 /* IPv6 specs say nothing about it, but it is clear that we cannot
485 send redirects to source routed frames.
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900486 We don't send redirects to frames decapsulated from IPsec.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 */
Stephen Suryaputra869584e2018-06-01 00:05:21 -0400488 if (IP6CB(skb)->iif == dst->dev->ifindex &&
489 opt->srcrt == 0 && !skb_sec_path(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 struct in6_addr *target = NULL;
David S. Millerfbfe95a2012-06-08 23:24:18 -0700491 struct inet_peer *peer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493
494 /*
495 * incoming and outgoing devices are the same
496 * send a redirect.
497 */
498
499 rt = (struct rt6_info *) dst;
David S. Millerc45a3df2012-01-27 15:32:19 -0800500 if (rt->rt6i_flags & RTF_GATEWAY)
501 target = &rt->rt6i_gateway;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 else
503 target = &hdr->daddr;
504
Martin KaFai Laufd0273d2015-05-22 20:55:57 -0700505 peer = inet_getpeer_v6(net->ipv6.peers, &hdr->daddr, 1);
David S. Miller92d86822011-02-04 15:55:25 -0800506
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 /* Limit redirects both by destination (here)
508 and by source (inside ndisc_send_redirect)
509 */
David S. Millerfbfe95a2012-06-08 23:24:18 -0700510 if (inet_peer_xrlim_allow(peer, 1*HZ))
David S. Miller49919692012-01-27 15:30:48 -0800511 ndisc_send_redirect(skb, target);
David S. Miller1d861aa2012-07-10 03:58:16 -0700512 if (peer)
513 inet_putpeer(peer);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700514 } else {
515 int addrtype = ipv6_addr_type(&hdr->saddr);
516
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 /* This check is security critical. */
YOSHIFUJI Hideakif81b2e72008-06-25 16:55:26 +0900518 if (addrtype == IPV6_ADDR_ANY ||
519 addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK))
David L Stevens5bb1ab02007-05-09 13:53:44 -0700520 goto error;
521 if (addrtype & IPV6_ADDR_LINKLOCAL) {
522 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000523 ICMPV6_NOT_NEIGHBOUR, 0);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700524 goto error;
525 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 }
527
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100528 mtu = ip6_dst_mtu_forward(dst);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800529 if (mtu < IPV6_MIN_MTU)
530 mtu = IPV6_MIN_MTU;
531
Florian Westphalfe6cc552014-02-13 23:09:12 +0100532 if (ip6_pkt_too_big(skb, mtu)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 /* Again, force OUTPUT device used as source address */
534 skb->dev = dst->dev;
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800535 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazet1d015502016-04-27 16:44:40 -0700536 __IP6_INC_STATS(net, ip6_dst_idev(dst),
537 IPSTATS_MIB_INTOOBIGERRORS);
538 __IP6_INC_STATS(net, ip6_dst_idev(dst),
539 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 kfree_skb(skb);
541 return -EMSGSIZE;
542 }
543
544 if (skb_cow(skb, dst->dev->hard_header_len)) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700545 __IP6_INC_STATS(net, ip6_dst_idev(dst),
546 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 goto drop;
548 }
549
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700550 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
552 /* Mangling hops number delayed to point after skb COW */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900553
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 hdr->hop_limit--;
555
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500556 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
557 net, NULL, skb, skb->dev, dst->dev,
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800558 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559
560error:
Eric Dumazet1d015502016-04-27 16:44:40 -0700561 __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562drop:
563 kfree_skb(skb);
564 return -EINVAL;
565}
566
567static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
568{
569 to->pkt_type = from->pkt_type;
570 to->priority = from->priority;
571 to->protocol = from->protocol;
Eric Dumazetadf30902009-06-02 05:19:30 +0000572 skb_dst_drop(to);
573 skb_dst_set(to, dst_clone(skb_dst(from)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 to->dev = from->dev;
Thomas Graf82e91ff2006-11-09 15:19:14 -0800575 to->mark = from->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576
Paolo Abeni93d94fe2018-07-23 16:50:48 +0200577 skb_copy_hash(to, from);
578
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579#ifdef CONFIG_NET_SCHED
580 to->tc_index = from->tc_index;
581#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -0700582 nf_copy(to, from);
James Morris984bc162006-06-09 00:29:17 -0700583 skb_copy_secmark(to, from);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584}
585
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500586int ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
587 int (*output)(struct net *, struct sock *, struct sk_buff *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 struct sk_buff *frag;
Ian Morris67ba4152014-08-24 21:53:10 +0100590 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +0200591 struct ipv6_pinfo *np = skb->sk && !dev_recursion_level() ?
592 inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 struct ipv6hdr *tmp_hdr;
594 struct frag_hdr *fh;
Junwei Hue33684f2019-04-02 19:38:04 +0800595 unsigned int mtu, hlen, left, len, nexthdr_offset;
Herbert Xua7ae1992011-11-18 02:20:04 +0000596 int hroom, troom;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700597 __be32 frag_id;
Ian Morris67ba4152014-08-24 21:53:10 +0100598 int ptr, offset = 0, err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 u8 *prevhdr, nexthdr = 0;
600
David S. Miller3fa202e2017-05-17 22:54:11 -0400601 err = ip6_find_1stfragopt(skb, &prevhdr);
602 if (err < 0)
Craig Galleka2c845e2017-05-16 14:36:23 -0400603 goto fail;
David S. Miller3fa202e2017-05-17 22:54:11 -0400604 hlen = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 nexthdr = *prevhdr;
Junwei Hue33684f2019-04-02 19:38:04 +0800606 nexthdr_offset = prevhdr - skb_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607
John Heffner628a5c52007-04-20 15:53:27 -0700608 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700609
610 /* We must not fragment if the socket is set to force MTU discovery
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800611 * or if the skb it not generated by a local socket.
John Heffnerb881ef72007-04-20 15:52:39 -0700612 */
Florian Westphal485fca62015-05-22 00:44:16 +0200613 if (unlikely(!skb->ignore_df && skb->len > mtu))
614 goto fail_toobig;
Eric Dumazeta34a1012012-05-18 21:51:44 +0000615
Florian Westphal485fca62015-05-22 00:44:16 +0200616 if (IP6CB(skb)->frag_max_size) {
617 if (IP6CB(skb)->frag_max_size > mtu)
618 goto fail_toobig;
619
620 /* don't send fragments larger than what we received */
621 mtu = IP6CB(skb)->frag_max_size;
622 if (mtu < IPV6_MIN_MTU)
623 mtu = IPV6_MIN_MTU;
John Heffnerb881ef72007-04-20 15:52:39 -0700624 }
625
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800626 if (np && np->frag_size < mtu) {
627 if (np->frag_size)
628 mtu = np->frag_size;
629 }
Hannes Frederic Sowa89bc7842015-10-28 13:21:04 +0100630 if (mtu < hlen + sizeof(struct frag_hdr) + 8)
Hannes Frederic Sowab72a2b02015-10-16 11:32:43 +0200631 goto fail_toobig;
Hannes Frederic Sowa1e0d69a2015-10-28 13:21:03 +0100632 mtu -= hlen + sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633
Martin KaFai Laufd0273d2015-05-22 20:55:57 -0700634 frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
635 &ipv6_hdr(skb)->saddr);
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700636
Hannes Frederic Sowa405c92f2015-10-27 22:40:42 +0100637 if (skb->ip_summed == CHECKSUM_PARTIAL &&
638 (err = skb_checksum_help(skb)))
639 goto fail;
640
Junwei Hue33684f2019-04-02 19:38:04 +0800641 prevhdr = skb_network_header(skb) + nexthdr_offset;
Florian Westphal1d325d22015-09-16 17:26:14 +0200642 hroom = LL_RESERVED_SPACE(rt->dst.dev);
David S. Miller21dc3302010-08-23 00:13:46 -0700643 if (skb_has_frag_list(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 int first_len = skb_pagelen(skb);
Eric Dumazet3d130082010-09-21 08:47:45 +0000645 struct sk_buff *frag2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646
647 if (first_len - hlen > mtu ||
648 ((first_len - hlen) & 7) ||
Florian Westphal1d325d22015-09-16 17:26:14 +0200649 skb_cloned(skb) ||
650 skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 goto slow_path;
652
David S. Miller4d9092b2009-06-09 00:20:05 -0700653 skb_walk_frags(skb, frag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 /* Correct geometry. */
655 if (frag->len > mtu ||
656 ((frag->len & 7) && frag->next) ||
Florian Westphal1d325d22015-09-16 17:26:14 +0200657 skb_headroom(frag) < (hlen + hroom + sizeof(struct frag_hdr)))
Eric Dumazet3d130082010-09-21 08:47:45 +0000658 goto slow_path_clean;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 /* Partially cloned skb? */
661 if (skb_shared(frag))
Eric Dumazet3d130082010-09-21 08:47:45 +0000662 goto slow_path_clean;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700663
664 BUG_ON(frag->sk);
665 if (skb->sk) {
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700666 frag->sk = skb->sk;
667 frag->destructor = sock_wfree;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700668 }
Eric Dumazet3d130082010-09-21 08:47:45 +0000669 skb->truesize -= frag->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 }
671
672 err = 0;
673 offset = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 /* BUILD HEADER */
675
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800676 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700677 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 if (!tmp_hdr) {
Florian Westphal1d325d22015-09-16 17:26:14 +0200679 err = -ENOMEM;
680 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 }
Florian Westphal1d325d22015-09-16 17:26:14 +0200682 frag = skb_shinfo(skb)->frag_list;
683 skb_frag_list_init(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 __skb_pull(skb, hlen);
Ian Morris67ba4152014-08-24 21:53:10 +0100686 fh = (struct frag_hdr *)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700687 __skb_push(skb, hlen);
688 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700689 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 fh->nexthdr = nexthdr;
692 fh->reserved = 0;
693 fh->frag_off = htons(IP6_MF);
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700694 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695
696 first_len = skb_pagelen(skb);
697 skb->data_len = first_len - skb_headlen(skb);
698 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700699 ipv6_hdr(skb)->payload_len = htons(first_len -
700 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900701
Changli Gaod8d1f302010-06-10 23:31:35 -0700702 dst_hold(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703
704 for (;;) {
705 /* Prepare header of the next frame,
706 * before previous one went down. */
707 if (frag) {
708 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300709 skb_reset_transport_header(frag);
Ian Morris67ba4152014-08-24 21:53:10 +0100710 fh = (struct frag_hdr *)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700711 __skb_push(frag, hlen);
712 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700713 memcpy(skb_network_header(frag), tmp_hdr,
714 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 offset += skb->len - hlen - sizeof(struct frag_hdr);
716 fh->nexthdr = nexthdr;
717 fh->reserved = 0;
718 fh->frag_off = htons(offset);
Ian Morris53b24b82015-03-29 14:00:05 +0100719 if (frag->next)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 fh->frag_off |= htons(IP6_MF);
721 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700722 ipv6_hdr(frag)->payload_len =
723 htons(frag->len -
724 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 ip6_copy_metadata(frag, skb);
726 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900727
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500728 err = output(net, sk, skb);
Ian Morris67ba4152014-08-24 21:53:10 +0100729 if (!err)
Changli Gaod8d1f302010-06-10 23:31:35 -0700730 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700731 IPSTATS_MIB_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700732
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 if (err || !frag)
734 break;
735
736 skb = frag;
737 frag = skb->next;
738 skb->next = NULL;
739 }
740
Jesper Juhla51482b2005-11-08 09:41:34 -0800741 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742
743 if (err == 0) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700744 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700745 IPSTATS_MIB_FRAGOKS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000746 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 return 0;
748 }
749
Florian Westphal46cfd722014-09-10 01:08:46 +0200750 kfree_skb_list(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751
Changli Gaod8d1f302010-06-10 23:31:35 -0700752 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700753 IPSTATS_MIB_FRAGFAILS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000754 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 return err;
Eric Dumazet3d130082010-09-21 08:47:45 +0000756
757slow_path_clean:
758 skb_walk_frags(skb, frag2) {
759 if (frag2 == frag)
760 break;
761 frag2->sk = NULL;
762 frag2->destructor = NULL;
763 skb->truesize += frag2->truesize;
764 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 }
766
767slow_path:
768 left = skb->len - hlen; /* Space per frame */
769 ptr = hlen; /* Where to start from */
770
771 /*
772 * Fragment the datagram.
773 */
774
Herbert Xua7ae1992011-11-18 02:20:04 +0000775 troom = rt->dst.dev->needed_tailroom;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
777 /*
778 * Keep copying data until we run out.
779 */
Ian Morris67ba4152014-08-24 21:53:10 +0100780 while (left > 0) {
Florian Westphal683100e2017-03-13 16:24:28 +0100781 u8 *fragnexthdr_offset;
782
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 len = left;
784 /* IF: it doesn't fit, use 'mtu' - the data space left */
785 if (len > mtu)
786 len = mtu;
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300787 /* IF: we are not sending up to and including the packet end
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 then align the next start on an eight byte boundary */
789 if (len < left) {
790 len &= ~7;
791 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
Joe Perchescbffccc2014-11-05 14:39:21 -0800793 /* Allocate buffer */
794 frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) +
795 hroom + troom, GFP_ATOMIC);
796 if (!frag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 err = -ENOMEM;
798 goto fail;
799 }
800
801 /*
802 * Set up data on packet
803 */
804
805 ip6_copy_metadata(frag, skb);
Herbert Xua7ae1992011-11-18 02:20:04 +0000806 skb_reserve(frag, hroom);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700808 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300809 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700810 frag->transport_header = (frag->network_header + hlen +
811 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812
813 /*
814 * Charge the memory for the fragment to any owner
815 * it might possess
816 */
817 if (skb->sk)
818 skb_set_owner_w(frag, skb->sk);
819
820 /*
821 * Copy the packet header into the new buffer.
822 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300823 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824
Florian Westphal683100e2017-03-13 16:24:28 +0100825 fragnexthdr_offset = skb_network_header(frag);
826 fragnexthdr_offset += prevhdr - skb_network_header(skb);
827 *fragnexthdr_offset = NEXTHDR_FRAGMENT;
828
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 /*
830 * Build fragment header.
831 */
832 fh->nexthdr = nexthdr;
833 fh->reserved = 0;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700834 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835
836 /*
837 * Copy a block of the IP datagram.
838 */
Himangi Saraogie3f0b862014-07-12 01:57:17 +0530839 BUG_ON(skb_copy_bits(skb, ptr, skb_transport_header(frag),
840 len));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 left -= len;
842
843 fh->frag_off = htons(offset);
844 if (left > 0)
845 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700846 ipv6_hdr(frag)->payload_len = htons(frag->len -
847 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848
849 ptr += len;
850 offset += len;
851
852 /*
853 * Put this fragment into the sending queue.
854 */
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500855 err = output(net, sk, frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 if (err)
857 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700858
Eric Dumazetadf30902009-06-02 05:19:30 +0000859 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700860 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000862 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900863 IPSTATS_MIB_FRAGOKS);
Eric Dumazet808db802012-04-24 10:17:59 +0000864 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 return err;
866
Florian Westphal485fca62015-05-22 00:44:16 +0200867fail_toobig:
868 if (skb->sk && dst_allfrag(skb_dst(skb)))
869 sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK);
870
871 skb->dev = skb_dst(skb)->dev;
872 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
873 err = -EMSGSIZE;
874
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875fail:
Eric Dumazetadf30902009-06-02 05:19:30 +0000876 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900877 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900878 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 return err;
880}
881
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000882static inline int ip6_rt_check(const struct rt6key *rt_key,
883 const struct in6_addr *fl_addr,
884 const struct in6_addr *addr_cache)
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700885{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000886 return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
Ian Morris63159f22015-03-29 14:00:04 +0100887 (!addr_cache || !ipv6_addr_equal(fl_addr, addr_cache));
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700888}
889
Herbert Xu497c6152006-07-30 20:19:33 -0700890static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
891 struct dst_entry *dst,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000892 const struct flowi6 *fl6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893{
Herbert Xu497c6152006-07-30 20:19:33 -0700894 struct ipv6_pinfo *np = inet6_sk(sk);
Eric Dumazeta963a372013-06-26 04:15:07 -0700895 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896
Herbert Xu497c6152006-07-30 20:19:33 -0700897 if (!dst)
898 goto out;
899
Eric Dumazeta963a372013-06-26 04:15:07 -0700900 if (dst->ops->family != AF_INET6) {
901 dst_release(dst);
902 return NULL;
903 }
904
905 rt = (struct rt6_info *)dst;
Herbert Xu497c6152006-07-30 20:19:33 -0700906 /* Yes, checking route validity in not connected
907 * case is not very simple. Take into account,
908 * that we do not support routing by source, TOS,
Ian Morris67ba4152014-08-24 21:53:10 +0100909 * and MSG_DONTROUTE --ANK (980726)
Herbert Xu497c6152006-07-30 20:19:33 -0700910 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700911 * 1. ip6_rt_check(): If route was host route,
912 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700913 * If it is network route, we still may
914 * check its validity using saved pointer
915 * to the last used address: daddr_cache.
916 * We do not want to save whole address now,
917 * (because main consumer of this service
918 * is tcp, which has not this problem),
919 * so that the last trick works only on connected
920 * sockets.
921 * 2. oif also should be the same.
922 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500923 if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700924#ifdef CONFIG_IPV6_SUBTREES
David S. Miller4c9483b2011-03-12 16:22:43 -0500925 ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700926#endif
David Ahernca254492015-10-12 11:47:10 -0700927 (!(fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF) &&
928 (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex))) {
Herbert Xu497c6152006-07-30 20:19:33 -0700929 dst_release(dst);
930 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 }
932
Herbert Xu497c6152006-07-30 20:19:33 -0700933out:
934 return dst;
935}
936
Eric Dumazet3aef9342015-09-25 07:39:12 -0700937static int ip6_dst_lookup_tail(struct net *net, const struct sock *sk,
David S. Miller4c9483b2011-03-12 16:22:43 -0500938 struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700939{
David S. Miller69cce1d2011-07-17 23:09:49 -0700940#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
941 struct neighbour *n;
David S. Miller97cac082012-07-02 22:43:47 -0700942 struct rt6_info *rt;
David S. Miller69cce1d2011-07-17 23:09:49 -0700943#endif
944 int err;
Paolo Abeni6f21c962016-01-29 12:30:19 +0100945 int flags = 0;
Herbert Xu497c6152006-07-30 20:19:33 -0700946
Markus Stenberge16e8882015-05-05 13:36:59 +0300947 /* The correct way to handle this would be to do
948 * ip6_route_get_saddr, and then ip6_route_output; however,
949 * the route-specific preferred source forces the
950 * ip6_route_output call _before_ ip6_route_get_saddr.
951 *
952 * In source specific routing (no src=any default route),
953 * ip6_route_output will fail given src=any saddr, though, so
954 * that's why we try it again later.
955 */
956 if (ipv6_addr_any(&fl6->saddr) && (!*dst || !(*dst)->error)) {
957 struct rt6_info *rt;
958 bool had_dst = *dst != NULL;
959
960 if (!had_dst)
961 *dst = ip6_route_output(net, sk, fl6);
962 rt = (*dst)->error ? NULL : (struct rt6_info *)*dst;
963 err = ip6_route_get_saddr(net, rt, &fl6->daddr,
964 sk ? inet6_sk(sk)->srcprefs : 0,
965 &fl6->saddr);
966 if (err)
967 goto out_err_release;
968
969 /* If we had an erroneous initial result, pretend it
970 * never existed and let the SA-enabled version take
971 * over.
972 */
973 if (!had_dst && (*dst)->error) {
974 dst_release(*dst);
975 *dst = NULL;
976 }
Paolo Abeni6f21c962016-01-29 12:30:19 +0100977
978 if (fl6->flowi6_oif)
979 flags |= RT6_LOOKUP_F_IFACE;
Markus Stenberge16e8882015-05-05 13:36:59 +0300980 }
981
Ian Morris63159f22015-03-29 14:00:04 +0100982 if (!*dst)
Paolo Abeni6f21c962016-01-29 12:30:19 +0100983 *dst = ip6_route_output_flags(net, sk, fl6, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984
Ian Morrise5d08d72014-11-23 21:28:43 +0000985 err = (*dst)->error;
986 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 goto out_err_release;
988
Neil Horman95c385b2007-04-25 17:08:10 -0700989#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700990 /*
991 * Here if the dst entry we've looked up
992 * has a neighbour entry that is in the INCOMPLETE
993 * state and the src address from the flow is
994 * marked as OPTIMISTIC, we release the found
995 * dst entry and replace it instead with the
996 * dst entry of the nexthop router
997 */
Eric Dumazetc56bf6f2012-07-06 09:19:05 +0200998 rt = (struct rt6_info *) *dst;
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000999 rcu_read_lock_bh();
Martin KaFai Lau2647a9b2015-05-22 20:55:58 -07001000 n = __ipv6_neigh_lookup_noref(rt->dst.dev,
1001 rt6_nexthop(rt, &fl6->daddr));
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +00001002 err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0;
1003 rcu_read_unlock_bh();
1004
1005 if (err) {
Neil Hormane550dfb2008-09-09 13:51:35 -07001006 struct inet6_ifaddr *ifp;
David S. Miller4c9483b2011-03-12 16:22:43 -05001007 struct flowi6 fl_gw6;
Neil Hormane550dfb2008-09-09 13:51:35 -07001008 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -07001009
David S. Miller4c9483b2011-03-12 16:22:43 -05001010 ifp = ipv6_get_ifaddr(net, &fl6->saddr,
Neil Hormane550dfb2008-09-09 13:51:35 -07001011 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -07001012
Neil Hormane550dfb2008-09-09 13:51:35 -07001013 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
1014 if (ifp)
1015 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -07001016
Neil Hormane550dfb2008-09-09 13:51:35 -07001017 if (redirect) {
1018 /*
1019 * We need to get the dst entry for the
1020 * default router instead
1021 */
1022 dst_release(*dst);
David S. Miller4c9483b2011-03-12 16:22:43 -05001023 memcpy(&fl_gw6, fl6, sizeof(struct flowi6));
1024 memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr));
1025 *dst = ip6_route_output(net, sk, &fl_gw6);
Ian Morrise5d08d72014-11-23 21:28:43 +00001026 err = (*dst)->error;
1027 if (err)
Neil Hormane550dfb2008-09-09 13:51:35 -07001028 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -07001029 }
Neil Hormane550dfb2008-09-09 13:51:35 -07001030 }
Neil Horman95c385b2007-04-25 17:08:10 -07001031#endif
Jonathan T. Leightoncb351da2017-05-23 21:53:34 -04001032 if (ipv6_addr_v4mapped(&fl6->saddr) &&
Willem de Bruijn98184bb2017-02-18 19:00:45 -05001033 !(ipv6_addr_v4mapped(&fl6->daddr) || ipv6_addr_any(&fl6->daddr))) {
1034 err = -EAFNOSUPPORT;
1035 goto out_err_release;
1036 }
Neil Horman95c385b2007-04-25 17:08:10 -07001037
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 return 0;
1039
1040out_err_release:
1041 dst_release(*dst);
1042 *dst = NULL;
David Ahern8a966fc2016-09-10 12:09:59 -07001043
David Ahern0d240e72016-06-16 16:24:25 -07001044 if (err == -ENETUNREACH)
1045 IP6_INC_STATS(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046 return err;
1047}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001048
Herbert Xu497c6152006-07-30 20:19:33 -07001049/**
1050 * ip6_dst_lookup - perform route lookup on flow
1051 * @sk: socket which provides route info
1052 * @dst: pointer to dst_entry * for result
David S. Miller4c9483b2011-03-12 16:22:43 -05001053 * @fl6: flow to lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001054 *
1055 * This function performs a route lookup on the given flow.
1056 *
1057 * It returns zero on success, or a standard errno code on error.
1058 */
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001059int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1060 struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -07001061{
1062 *dst = NULL;
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001063 return ip6_dst_lookup_tail(net, sk, dst, fl6);
Herbert Xu497c6152006-07-30 20:19:33 -07001064}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -08001065EXPORT_SYMBOL_GPL(ip6_dst_lookup);
1066
Herbert Xu497c6152006-07-30 20:19:33 -07001067/**
David S. Miller68d0c6d2011-03-01 13:19:07 -08001068 * ip6_dst_lookup_flow - perform route lookup on flow with ipsec
1069 * @sk: socket which provides route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001070 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001071 * @final_dst: final destination address for ipsec lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001072 *
1073 * This function performs a route lookup on the given flow.
1074 *
1075 * It returns a valid dst pointer on success, or a pointer encoded
1076 * error code.
1077 */
Eric Dumazet3aef9342015-09-25 07:39:12 -07001078struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001079 const struct in6_addr *final_dst)
David S. Miller68d0c6d2011-03-01 13:19:07 -08001080{
1081 struct dst_entry *dst = NULL;
1082 int err;
1083
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001084 err = ip6_dst_lookup_tail(sock_net(sk), sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001085 if (err)
1086 return ERR_PTR(err);
1087 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001088 fl6->daddr = *final_dst;
David S. Miller2774c132011-03-01 14:59:04 -08001089
Steffen Klassertf92ee612014-09-16 10:08:40 +02001090 return xfrm_lookup_route(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001091}
1092EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow);
1093
1094/**
1095 * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow
1096 * @sk: socket which provides the dst cache and route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001097 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001098 * @final_dst: final destination address for ipsec lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001099 *
1100 * This function performs a route lookup on the given flow with the
1101 * possibility of using the cached route in the socket if it is valid.
1102 * It will take the socket dst lock when operating on the dst cache.
1103 * As a result, this function can only be used in process context.
1104 *
David S. Miller68d0c6d2011-03-01 13:19:07 -08001105 * It returns a valid dst pointer on success, or a pointer encoded
1106 * error code.
Herbert Xu497c6152006-07-30 20:19:33 -07001107 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001108struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001109 const struct in6_addr *final_dst)
Herbert Xu497c6152006-07-30 20:19:33 -07001110{
David S. Miller68d0c6d2011-03-01 13:19:07 -08001111 struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
Herbert Xu497c6152006-07-30 20:19:33 -07001112
David S. Miller4c9483b2011-03-12 16:22:43 -05001113 dst = ip6_sk_dst_check(sk, dst, fl6);
Jakub Sitnicki00bc0ef2016-06-08 15:13:34 +02001114 if (!dst)
1115 dst = ip6_dst_lookup_flow(sk, fl6, final_dst);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001116
Jakub Sitnicki00bc0ef2016-06-08 15:13:34 +02001117 return dst;
Herbert Xu497c6152006-07-30 20:19:33 -07001118}
David S. Miller68d0c6d2011-03-01 13:19:07 -08001119EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
Herbert Xu497c6152006-07-30 20:19:33 -07001120
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001121static inline int ip6_ufo_append_data(struct sock *sk,
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001122 struct sk_buff_head *queue,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001123 int getfrag(void *from, char *to, int offset, int len,
1124 int odd, struct sk_buff *skb),
1125 void *from, int length, int hh_len, int fragheaderlen,
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001126 int exthdrlen, int transhdrlen, int mtu,
1127 unsigned int flags, const struct flowi6 *fl6)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001128
1129{
1130 struct sk_buff *skb;
1131 int err;
1132
1133 /* There is support for UDP large send offload by network
1134 * device, so create one single skb packet containing complete
1135 * udp datagram
1136 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001137 skb = skb_peek_tail(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001138 if (!skb) {
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001139 skb = sock_alloc_send_skb(sk,
1140 hh_len + fragheaderlen + transhdrlen + 20,
1141 (flags & MSG_DONTWAIT), &err);
Ian Morris63159f22015-03-29 14:00:04 +01001142 if (!skb)
Zheng Yan504744e2011-10-28 00:26:00 -04001143 return err;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001144
1145 /* reserve space for Hardware header */
1146 skb_reserve(skb, hh_len);
1147
1148 /* create space for UDP/IP header */
Ian Morris67ba4152014-08-24 21:53:10 +01001149 skb_put(skb, fragheaderlen + transhdrlen);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001150
1151 /* initialize network header pointer */
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001152 skb_set_network_header(skb, exthdrlen);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001153
1154 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001155 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001156
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001157 skb->protocol = htons(ETH_P_IPV6);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001158 skb->csum = 0;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001159
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001160 __skb_queue_tail(queue, skb);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001161 } else if (skb_is_gso(skb)) {
1162 goto append;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001163 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001164
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001165 skb->ip_summed = CHECKSUM_PARTIAL;
1166 /* Specify the length of each IPv6 datagram fragment.
1167 * It has to be a multiple of 8.
1168 */
1169 skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
1170 sizeof(struct frag_hdr)) & ~7;
1171 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001172 skb_shinfo(skb)->ip6_frag_id = ipv6_select_ident(sock_net(sk),
1173 &fl6->daddr,
1174 &fl6->saddr);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001175
1176append:
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001177 return skb_append_datato_frags(sk, skb, getfrag, from,
1178 (length - transhdrlen));
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001179}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180
Herbert Xu0178b692009-02-05 15:15:50 -08001181static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
1182 gfp_t gfp)
1183{
1184 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1185}
1186
1187static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
1188 gfp_t gfp)
1189{
1190 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1191}
1192
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001193static void ip6_append_data_mtu(unsigned int *mtu,
Gao feng0c183372012-05-26 01:30:53 +00001194 int *maxfraglen,
1195 unsigned int fragheaderlen,
1196 struct sk_buff *skb,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001197 struct rt6_info *rt,
luciene367c2d2014-03-17 12:51:01 +08001198 unsigned int orig_mtu)
Gao feng0c183372012-05-26 01:30:53 +00001199{
1200 if (!(rt->dst.flags & DST_XFRM_TUNNEL)) {
Ian Morris63159f22015-03-29 14:00:04 +01001201 if (!skb) {
Gao feng0c183372012-05-26 01:30:53 +00001202 /* first fragment, reserve header_len */
luciene367c2d2014-03-17 12:51:01 +08001203 *mtu = orig_mtu - rt->dst.header_len;
Gao feng0c183372012-05-26 01:30:53 +00001204
1205 } else {
1206 /*
1207 * this fragment is not first, the headers
1208 * space is regarded as data space.
1209 */
luciene367c2d2014-03-17 12:51:01 +08001210 *mtu = orig_mtu;
Gao feng0c183372012-05-26 01:30:53 +00001211 }
1212 *maxfraglen = ((*mtu - fragheaderlen) & ~7)
1213 + fragheaderlen - sizeof(struct frag_hdr);
1214 }
1215}
1216
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001217static int ip6_setup_cork(struct sock *sk, struct inet_cork_full *cork,
Wei Wang26879da2016-05-02 21:40:07 -07001218 struct inet6_cork *v6_cork, struct ipcm6_cookie *ipc6,
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001219 struct rt6_info *rt, struct flowi6 *fl6)
1220{
1221 struct ipv6_pinfo *np = inet6_sk(sk);
1222 unsigned int mtu;
Wei Wang26879da2016-05-02 21:40:07 -07001223 struct ipv6_txoptions *opt = ipc6->opt;
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001224
1225 /*
1226 * setup for corking
1227 */
1228 if (opt) {
1229 if (WARN_ON(v6_cork->opt))
1230 return -EINVAL;
1231
Eric Dumazet28fa5832017-10-21 12:26:23 -07001232 v6_cork->opt = kzalloc(sizeof(*opt), sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001233 if (unlikely(!v6_cork->opt))
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001234 return -ENOBUFS;
1235
Eric Dumazet28fa5832017-10-21 12:26:23 -07001236 v6_cork->opt->tot_len = sizeof(*opt);
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001237 v6_cork->opt->opt_flen = opt->opt_flen;
1238 v6_cork->opt->opt_nflen = opt->opt_nflen;
1239
1240 v6_cork->opt->dst0opt = ip6_opt_dup(opt->dst0opt,
1241 sk->sk_allocation);
1242 if (opt->dst0opt && !v6_cork->opt->dst0opt)
1243 return -ENOBUFS;
1244
1245 v6_cork->opt->dst1opt = ip6_opt_dup(opt->dst1opt,
1246 sk->sk_allocation);
1247 if (opt->dst1opt && !v6_cork->opt->dst1opt)
1248 return -ENOBUFS;
1249
1250 v6_cork->opt->hopopt = ip6_opt_dup(opt->hopopt,
1251 sk->sk_allocation);
1252 if (opt->hopopt && !v6_cork->opt->hopopt)
1253 return -ENOBUFS;
1254
1255 v6_cork->opt->srcrt = ip6_rthdr_dup(opt->srcrt,
1256 sk->sk_allocation);
1257 if (opt->srcrt && !v6_cork->opt->srcrt)
1258 return -ENOBUFS;
1259
1260 /* need source address above miyazawa*/
1261 }
1262 dst_hold(&rt->dst);
1263 cork->base.dst = &rt->dst;
1264 cork->fl.u.ip6 = *fl6;
Wei Wang26879da2016-05-02 21:40:07 -07001265 v6_cork->hop_limit = ipc6->hlimit;
1266 v6_cork->tclass = ipc6->tclass;
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001267 if (rt->dst.flags & DST_XFRM_TUNNEL)
1268 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
Mike Maloneyfb50d8c2018-01-10 12:45:10 -05001269 READ_ONCE(rt->dst.dev->mtu) : dst_mtu(&rt->dst);
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001270 else
1271 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
Mike Maloneyfb50d8c2018-01-10 12:45:10 -05001272 READ_ONCE(rt->dst.dev->mtu) : dst_mtu(rt->dst.path);
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001273 if (np->frag_size < mtu) {
1274 if (np->frag_size)
1275 mtu = np->frag_size;
1276 }
Mike Maloneyfb50d8c2018-01-10 12:45:10 -05001277 if (mtu < IPV6_MIN_MTU)
1278 return -EINVAL;
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001279 cork->base.fragsize = mtu;
1280 if (dst_allfrag(rt->dst.path))
1281 cork->base.flags |= IPCORK_ALLFRAG;
1282 cork->base.length = 0;
1283
1284 return 0;
1285}
1286
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001287static int __ip6_append_data(struct sock *sk,
1288 struct flowi6 *fl6,
1289 struct sk_buff_head *queue,
1290 struct inet_cork *cork,
1291 struct inet6_cork *v6_cork,
1292 struct page_frag *pfrag,
1293 int getfrag(void *from, char *to, int offset,
1294 int len, int odd, struct sk_buff *skb),
1295 void *from, int length, int transhdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001296 unsigned int flags, struct ipcm6_cookie *ipc6,
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001297 const struct sockcm_cookie *sockc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298{
Gao feng0c183372012-05-26 01:30:53 +00001299 struct sk_buff *skb, *skb_prev = NULL;
Paolo Abeni81eb03e2018-03-23 14:47:30 +01001300 unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu, pmtu;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001301 int exthdrlen = 0;
1302 int dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303 int hh_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 int copy;
1305 int err;
1306 int offset = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001307 __u8 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001308 u32 tskey = 0;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001309 struct rt6_info *rt = (struct rt6_info *)cork->dst;
1310 struct ipv6_txoptions *opt = v6_cork->opt;
Vlad Yasevich32dce962015-01-31 10:40:18 -05001311 int csummode = CHECKSUM_NONE;
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001312 unsigned int maxnonfragsize, headersize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001314 skb = skb_peek_tail(queue);
1315 if (!skb) {
1316 exthdrlen = opt ? opt->opt_flen : 0;
Romain KUNTZ7efdba52013-01-16 12:47:40 +00001317 dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 }
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001319
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001320 mtu = cork->fragsize;
luciene367c2d2014-03-17 12:51:01 +08001321 orig_mtu = mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322
Changli Gaod8d1f302010-06-10 23:31:35 -07001323 hh_len = LL_RESERVED_SPACE(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001325 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001326 (opt ? opt->opt_nflen : 0);
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001327 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen -
1328 sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001330 headersize = sizeof(struct ipv6hdr) +
1331 (opt ? opt->opt_flen + opt->opt_nflen : 0) +
1332 (dst_allfrag(&rt->dst) ?
1333 sizeof(struct frag_hdr) : 0) +
1334 rt->rt6i_nfheader_len;
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001335
Paolo Abeni81eb03e2018-03-23 14:47:30 +01001336 /* as per RFC 7112 section 5, the entire IPv6 Header Chain must fit
1337 * the first fragment
1338 */
1339 if (headersize + transhdrlen > mtu)
1340 goto emsgsize;
1341
Wei Wang26879da2016-05-02 21:40:07 -07001342 if (cork->length + length > mtu - headersize && ipc6->dontfrag &&
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001343 (sk->sk_protocol == IPPROTO_UDP ||
1344 sk->sk_protocol == IPPROTO_RAW)) {
1345 ipv6_local_rxpmtu(sk, fl6, mtu - headersize +
1346 sizeof(struct ipv6hdr));
1347 goto emsgsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348 }
1349
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001350 if (ip6_sk_ignore_df(sk))
1351 maxnonfragsize = sizeof(struct ipv6hdr) + IPV6_MAXPLEN;
1352 else
1353 maxnonfragsize = mtu;
1354
1355 if (cork->length + length > maxnonfragsize - headersize) {
1356emsgsize:
Paolo Abeni81eb03e2018-03-23 14:47:30 +01001357 pmtu = max_t(int, mtu - headersize + sizeof(struct ipv6hdr), 0);
1358 ipv6_local_error(sk, EMSGSIZE, fl6, pmtu);
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001359 return -EMSGSIZE;
1360 }
1361
1362 /* CHECKSUM_PARTIAL only with no extension headers and when
1363 * we are not going to fragment
1364 */
1365 if (transhdrlen && sk->sk_protocol == IPPROTO_UDP &&
1366 headersize == sizeof(struct ipv6hdr) &&
1367 length < mtu - headersize &&
1368 !(flags & MSG_MORE) &&
Tom Herbertc8cd0982015-12-14 11:19:44 -08001369 rt->dst.dev->features & (NETIF_F_IPV6_CSUM | NETIF_F_HW_CSUM))
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001370 csummode = CHECKSUM_PARTIAL;
1371
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001372 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_RAW) {
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001373 sock_tx_timestamp(sk, sockc->tsflags, &tx_flags);
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001374 if (tx_flags & SKBTX_ANY_SW_TSTAMP &&
1375 sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)
1376 tskey = sk->sk_tskey++;
1377 }
Anders Berggrena693e692011-02-28 12:32:11 -08001378
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 /*
1380 * Let's try using as much space as possible.
1381 * Use MTU if total length of the message fits into the MTU.
1382 * Otherwise, we need to reserve fragment header and
1383 * fragment alignment (= 8-15 octects, in total).
1384 *
1385 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001386 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 * the message.
1388 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001389 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 * at once if non-fragmentable extension headers
1391 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001392 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393 */
1394
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001395 cork->length += length;
Willem de Bruijn33dc6a62017-08-10 12:29:19 -04001396 if ((skb && skb_is_gso(skb)) ||
1397 (((length + fragheaderlen) > mtu) &&
1398 (skb_queue_len(queue) <= 1) &&
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001399 (sk->sk_protocol == IPPROTO_UDP) &&
Jakub Sitnickif89c56c2016-10-26 11:21:14 +02001400 (rt->dst.dev->features & NETIF_F_UFO) && !rt->dst.header_len &&
Willem de Bruijn33dc6a62017-08-10 12:29:19 -04001401 (sk->sk_type == SOCK_DGRAM) && !udp_get_no_check6_tx(sk))) {
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001402 err = ip6_ufo_append_data(sk, queue, getfrag, from, length,
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001403 hh_len, fragheaderlen, exthdrlen,
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001404 transhdrlen, mtu, flags, fl6);
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001405 if (err)
1406 goto error;
1407 return 0;
1408 }
1409
1410 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411 goto alloc_new_skb;
1412
1413 while (length > 0) {
1414 /* Check if the remaining data fits into current packet. */
David S. Millerbdc712b2011-05-06 15:02:07 -07001415 copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416 if (copy < length)
1417 copy = maxfraglen - skb->len;
1418
1419 if (copy <= 0) {
1420 char *data;
1421 unsigned int datalen;
1422 unsigned int fraglen;
1423 unsigned int fraggap;
1424 unsigned int alloclen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425alloc_new_skb:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 /* There's no room in the current skb */
Gao feng0c183372012-05-26 01:30:53 +00001427 if (skb)
1428 fraggap = skb->len - maxfraglen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 else
1430 fraggap = 0;
Gao feng0c183372012-05-26 01:30:53 +00001431 /* update mtu and maxfraglen if necessary */
Ian Morris63159f22015-03-29 14:00:04 +01001432 if (!skb || !skb_prev)
Gao feng0c183372012-05-26 01:30:53 +00001433 ip6_append_data_mtu(&mtu, &maxfraglen,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001434 fragheaderlen, skb, rt,
luciene367c2d2014-03-17 12:51:01 +08001435 orig_mtu);
Gao feng0c183372012-05-26 01:30:53 +00001436
1437 skb_prev = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438
1439 /*
1440 * If remaining data exceeds the mtu,
1441 * we know we need more fragment(s).
1442 */
1443 datalen = length + fraggap;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444
Gao feng0c183372012-05-26 01:30:53 +00001445 if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1446 datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447 if ((flags & MSG_MORE) &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001448 !(rt->dst.dev->features&NETIF_F_SG))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 alloclen = mtu;
1450 else
1451 alloclen = datalen + fragheaderlen;
1452
Steffen Klassert299b0762011-10-11 01:43:33 +00001453 alloclen += dst_exthdrlen;
1454
Gao feng0c183372012-05-26 01:30:53 +00001455 if (datalen != length + fraggap) {
1456 /*
1457 * this is not the last fragment, the trailer
1458 * space is regarded as data space.
1459 */
1460 datalen += rt->dst.trailer_len;
1461 }
1462
1463 alloclen += rt->dst.trailer_len;
1464 fraglen = datalen + fragheaderlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465
1466 /*
1467 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001468 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 * (without MSG_MORE) fits into the MTU.
1470 */
1471 alloclen += sizeof(struct frag_hdr);
1472
Eric Dumazet304b4102017-05-19 14:17:48 -07001473 copy = datalen - transhdrlen - fraggap;
1474 if (copy < 0) {
1475 err = -EINVAL;
1476 goto error;
1477 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 if (transhdrlen) {
1479 skb = sock_alloc_send_skb(sk,
1480 alloclen + hh_len,
1481 (flags & MSG_DONTWAIT), &err);
1482 } else {
1483 skb = NULL;
1484 if (atomic_read(&sk->sk_wmem_alloc) <=
1485 2 * sk->sk_sndbuf)
1486 skb = sock_wmalloc(sk,
1487 alloclen + hh_len, 1,
1488 sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001489 if (unlikely(!skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 err = -ENOBUFS;
1491 }
Ian Morris63159f22015-03-29 14:00:04 +01001492 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 goto error;
1494 /*
1495 * Fill in the control structures
1496 */
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001497 skb->protocol = htons(ETH_P_IPV6);
Vlad Yasevich32dce962015-01-31 10:40:18 -05001498 skb->ip_summed = csummode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 skb->csum = 0;
Gao feng1f858512012-03-19 22:36:10 +00001500 /* reserve for fragmentation and ipsec header */
1501 skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
1502 dst_exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503
Willem de Bruijn11878b42014-07-14 17:55:06 -04001504 /* Only the initial fragment is time stamped */
1505 skb_shinfo(skb)->tx_flags = tx_flags;
1506 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001507 skb_shinfo(skb)->tskey = tskey;
1508 tskey = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001509
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 /*
1511 * Find where to start putting bytes
1512 */
Gao feng1f858512012-03-19 22:36:10 +00001513 data = skb_put(skb, fraglen);
1514 skb_set_network_header(skb, exthdrlen);
1515 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001516 skb->transport_header = (skb->network_header +
1517 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 if (fraggap) {
1519 skb->csum = skb_copy_and_csum_bits(
1520 skb_prev, maxfraglen,
1521 data + transhdrlen, fraggap, 0);
1522 skb_prev->csum = csum_sub(skb_prev->csum,
1523 skb->csum);
1524 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001525 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 }
Eric Dumazet304b4102017-05-19 14:17:48 -07001527 if (copy > 0 &&
1528 getfrag(from, data + transhdrlen, offset,
1529 copy, fraggap, skb) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 err = -EFAULT;
1531 kfree_skb(skb);
1532 goto error;
1533 }
1534
1535 offset += copy;
1536 length -= datalen - fraggap;
1537 transhdrlen = 0;
1538 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001539 dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540
1541 /*
1542 * Put the packet on the pending queue
1543 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001544 __skb_queue_tail(queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545 continue;
1546 }
1547
1548 if (copy > length)
1549 copy = length;
1550
Willem de Bruijn2ef22bd2018-05-17 13:13:29 -04001551 if (!(rt->dst.dev->features&NETIF_F_SG) &&
1552 skb_tailroom(skb) >= copy) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553 unsigned int off;
1554
1555 off = skb->len;
1556 if (getfrag(from, skb_put(skb, copy),
1557 offset, copy, off, skb) < 0) {
1558 __skb_trim(skb, off);
1559 err = -EFAULT;
1560 goto error;
1561 }
1562 } else {
1563 int i = skb_shinfo(skb)->nr_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564
Eric Dumazet5640f762012-09-23 23:04:42 +00001565 err = -ENOMEM;
1566 if (!sk_page_frag_refill(sk, pfrag))
1567 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568
Eric Dumazet5640f762012-09-23 23:04:42 +00001569 if (!skb_can_coalesce(skb, i, pfrag->page,
1570 pfrag->offset)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 err = -EMSGSIZE;
Eric Dumazet5640f762012-09-23 23:04:42 +00001572 if (i == MAX_SKB_FRAGS)
1573 goto error;
1574
1575 __skb_fill_page_desc(skb, i, pfrag->page,
1576 pfrag->offset, 0);
1577 skb_shinfo(skb)->nr_frags = ++i;
1578 get_page(pfrag->page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001580 copy = min_t(int, copy, pfrag->size - pfrag->offset);
Eric Dumazet9e903e02011-10-18 21:00:24 +00001581 if (getfrag(from,
Eric Dumazet5640f762012-09-23 23:04:42 +00001582 page_address(pfrag->page) + pfrag->offset,
1583 offset, copy, skb->len, skb) < 0)
1584 goto error_efault;
1585
1586 pfrag->offset += copy;
1587 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588 skb->len += copy;
1589 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001590 skb->truesize += copy;
1591 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 }
1593 offset += copy;
1594 length -= copy;
1595 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001596
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 return 0;
Eric Dumazet5640f762012-09-23 23:04:42 +00001598
1599error_efault:
1600 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601error:
David S. Millerbdc712b2011-05-06 15:02:07 -07001602 cork->length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001603 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 return err;
1605}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001606
1607int ip6_append_data(struct sock *sk,
1608 int getfrag(void *from, char *to, int offset, int len,
1609 int odd, struct sk_buff *skb),
Wei Wang26879da2016-05-02 21:40:07 -07001610 void *from, int length, int transhdrlen,
1611 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
1612 struct rt6_info *rt, unsigned int flags,
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001613 const struct sockcm_cookie *sockc)
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001614{
1615 struct inet_sock *inet = inet_sk(sk);
1616 struct ipv6_pinfo *np = inet6_sk(sk);
1617 int exthdrlen;
1618 int err;
1619
1620 if (flags&MSG_PROBE)
1621 return 0;
1622 if (skb_queue_empty(&sk->sk_write_queue)) {
1623 /*
1624 * setup for corking
1625 */
Wei Wang26879da2016-05-02 21:40:07 -07001626 err = ip6_setup_cork(sk, &inet->cork, &np->cork,
1627 ipc6, rt, fl6);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001628 if (err)
1629 return err;
1630
Wei Wang26879da2016-05-02 21:40:07 -07001631 exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001632 length += exthdrlen;
1633 transhdrlen += exthdrlen;
1634 } else {
1635 fl6 = &inet->cork.fl.u.ip6;
1636 transhdrlen = 0;
1637 }
1638
1639 return __ip6_append_data(sk, fl6, &sk->sk_write_queue, &inet->cork.base,
1640 &np->cork, sk_page_frag(sk), getfrag,
Wei Wang26879da2016-05-02 21:40:07 -07001641 from, length, transhdrlen, flags, ipc6, sockc);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001642}
Chris Elstona495f832012-04-29 21:48:53 +00001643EXPORT_SYMBOL_GPL(ip6_append_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001645static void ip6_cork_release(struct inet_cork_full *cork,
1646 struct inet6_cork *v6_cork)
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001647{
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001648 if (v6_cork->opt) {
1649 kfree(v6_cork->opt->dst0opt);
1650 kfree(v6_cork->opt->dst1opt);
1651 kfree(v6_cork->opt->hopopt);
1652 kfree(v6_cork->opt->srcrt);
1653 kfree(v6_cork->opt);
1654 v6_cork->opt = NULL;
Herbert Xu0178b692009-02-05 15:15:50 -08001655 }
1656
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001657 if (cork->base.dst) {
1658 dst_release(cork->base.dst);
1659 cork->base.dst = NULL;
1660 cork->base.flags &= ~IPCORK_ALLFRAG;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001661 }
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001662 memset(&cork->fl, 0, sizeof(cork->fl));
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001663}
1664
Vlad Yasevich64223982015-01-31 10:40:15 -05001665struct sk_buff *__ip6_make_skb(struct sock *sk,
1666 struct sk_buff_head *queue,
1667 struct inet_cork_full *cork,
1668 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669{
1670 struct sk_buff *skb, *tmp_skb;
1671 struct sk_buff **tail_skb;
1672 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001674 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675 struct ipv6hdr *hdr;
Vlad Yasevich64223982015-01-31 10:40:15 -05001676 struct ipv6_txoptions *opt = v6_cork->opt;
1677 struct rt6_info *rt = (struct rt6_info *)cork->base.dst;
1678 struct flowi6 *fl6 = &cork->fl.u.ip6;
David S. Miller4c9483b2011-03-12 16:22:43 -05001679 unsigned char proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680
Vlad Yasevich64223982015-01-31 10:40:15 -05001681 skb = __skb_dequeue(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001682 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 goto out;
1684 tail_skb = &(skb_shinfo(skb)->frag_list);
1685
1686 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001687 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001688 __skb_pull(skb, skb_network_offset(skb));
Vlad Yasevich64223982015-01-31 10:40:15 -05001689 while ((tmp_skb = __skb_dequeue(queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001690 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 *tail_skb = tmp_skb;
1692 tail_skb = &(tmp_skb->next);
1693 skb->len += tmp_skb->len;
1694 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 skb->truesize += tmp_skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696 tmp_skb->destructor = NULL;
1697 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 }
1699
Herbert Xu28a894532008-02-12 18:07:27 -08001700 /* Allow local fragmentation. */
WANG Cong60ff7462014-05-04 16:39:18 -07001701 skb->ignore_df = ip6_sk_ignore_df(sk);
Herbert Xu28a894532008-02-12 18:07:27 -08001702
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001703 *final_dst = fl6->daddr;
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001704 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705 if (opt && opt->opt_flen)
1706 ipv6_push_frag_opts(skb, opt, &proto);
1707 if (opt && opt->opt_nflen)
1708 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1709
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001710 skb_push(skb, sizeof(struct ipv6hdr));
1711 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001712 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001713
Vlad Yasevich64223982015-01-31 10:40:15 -05001714 ip6_flow_hdr(hdr, v6_cork->tclass,
Tom Herbertcb1ce2e2014-07-01 21:33:10 -07001715 ip6_make_flowlabel(net, skb, fl6->flowlabel,
Shaohua Lib3b56032017-12-20 12:10:21 -08001716 ip6_autoflowlabel(net, np), fl6));
Vlad Yasevich64223982015-01-31 10:40:15 -05001717 hdr->hop_limit = v6_cork->hop_limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718 hdr->nexthdr = proto;
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001719 hdr->saddr = fl6->saddr;
1720 hdr->daddr = *final_dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721
Patrick McHardya2c20642006-01-08 22:37:26 -08001722 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001723 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001724
Changli Gaod8d1f302010-06-10 23:31:35 -07001725 skb_dst_set(skb, dst_clone(&rt->dst));
Neil Hormanedf391f2009-04-27 02:45:02 -07001726 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
David L Stevens14878f72007-09-16 16:52:35 -07001727 if (proto == IPPROTO_ICMPV6) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001728 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
David L Stevens14878f72007-09-16 16:52:35 -07001729
Hannes Frederic Sowa43a43b62014-03-31 20:14:10 +02001730 ICMP6MSGOUT_INC_STATS(net, idev, icmp6_hdr(skb)->icmp6_type);
1731 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001732 }
1733
Vlad Yasevich64223982015-01-31 10:40:15 -05001734 ip6_cork_release(cork, v6_cork);
1735out:
1736 return skb;
1737}
1738
1739int ip6_send_skb(struct sk_buff *skb)
1740{
1741 struct net *net = sock_net(skb->sk);
1742 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
1743 int err;
1744
Eric W. Biederman33224b12015-10-07 16:48:46 -05001745 err = ip6_local_out(net, skb->sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 if (err) {
1747 if (err > 0)
Eric Dumazet6ce9e7b2009-09-02 18:05:33 -07001748 err = net_xmit_errno(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749 if (err)
Vlad Yasevich64223982015-01-31 10:40:15 -05001750 IP6_INC_STATS(net, rt->rt6i_idev,
1751 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 }
1753
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 return err;
Vlad Yasevich64223982015-01-31 10:40:15 -05001755}
1756
1757int ip6_push_pending_frames(struct sock *sk)
1758{
1759 struct sk_buff *skb;
1760
1761 skb = ip6_finish_skb(sk);
1762 if (!skb)
1763 return 0;
1764
1765 return ip6_send_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766}
Chris Elstona495f832012-04-29 21:48:53 +00001767EXPORT_SYMBOL_GPL(ip6_push_pending_frames);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001769static void __ip6_flush_pending_frames(struct sock *sk,
Vlad Yasevich64223982015-01-31 10:40:15 -05001770 struct sk_buff_head *queue,
1771 struct inet_cork_full *cork,
1772 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774 struct sk_buff *skb;
1775
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001776 while ((skb = __skb_dequeue_tail(queue)) != NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001777 if (skb_dst(skb))
1778 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001779 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 kfree_skb(skb);
1781 }
1782
Vlad Yasevich64223982015-01-31 10:40:15 -05001783 ip6_cork_release(cork, v6_cork);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001785
1786void ip6_flush_pending_frames(struct sock *sk)
1787{
Vlad Yasevich64223982015-01-31 10:40:15 -05001788 __ip6_flush_pending_frames(sk, &sk->sk_write_queue,
1789 &inet_sk(sk)->cork, &inet6_sk(sk)->cork);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001790}
Chris Elstona495f832012-04-29 21:48:53 +00001791EXPORT_SYMBOL_GPL(ip6_flush_pending_frames);
Vlad Yasevich64223982015-01-31 10:40:15 -05001792
1793struct sk_buff *ip6_make_skb(struct sock *sk,
1794 int getfrag(void *from, char *to, int offset,
1795 int len, int odd, struct sk_buff *skb),
1796 void *from, int length, int transhdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001797 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
Vlad Yasevich64223982015-01-31 10:40:15 -05001798 struct rt6_info *rt, unsigned int flags,
Wei Wang26879da2016-05-02 21:40:07 -07001799 const struct sockcm_cookie *sockc)
Vlad Yasevich64223982015-01-31 10:40:15 -05001800{
1801 struct inet_cork_full cork;
1802 struct inet6_cork v6_cork;
1803 struct sk_buff_head queue;
Wei Wang26879da2016-05-02 21:40:07 -07001804 int exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0);
Vlad Yasevich64223982015-01-31 10:40:15 -05001805 int err;
1806
1807 if (flags & MSG_PROBE)
1808 return NULL;
1809
1810 __skb_queue_head_init(&queue);
1811
1812 cork.base.flags = 0;
1813 cork.base.addr = 0;
1814 cork.base.opt = NULL;
Eric Dumazetc2ceff12018-01-11 22:31:18 -08001815 cork.base.dst = NULL;
Vlad Yasevich64223982015-01-31 10:40:15 -05001816 v6_cork.opt = NULL;
Wei Wang26879da2016-05-02 21:40:07 -07001817 err = ip6_setup_cork(sk, &cork, &v6_cork, ipc6, rt, fl6);
Eric Dumazetdde00c92018-01-10 03:45:49 -08001818 if (err) {
1819 ip6_cork_release(&cork, &v6_cork);
Vlad Yasevich64223982015-01-31 10:40:15 -05001820 return ERR_PTR(err);
Eric Dumazetdde00c92018-01-10 03:45:49 -08001821 }
Wei Wang26879da2016-05-02 21:40:07 -07001822 if (ipc6->dontfrag < 0)
1823 ipc6->dontfrag = inet6_sk(sk)->dontfrag;
Vlad Yasevich64223982015-01-31 10:40:15 -05001824
1825 err = __ip6_append_data(sk, fl6, &queue, &cork.base, &v6_cork,
1826 &current->task_frag, getfrag, from,
1827 length + exthdrlen, transhdrlen + exthdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001828 flags, ipc6, sockc);
Vlad Yasevich64223982015-01-31 10:40:15 -05001829 if (err) {
1830 __ip6_flush_pending_frames(sk, &queue, &cork, &v6_cork);
1831 return ERR_PTR(err);
1832 }
1833
1834 return __ip6_make_skb(sk, &queue, &cork, &v6_cork);
1835}