blob: 58f6288e9ba53e6964b74d71dde7615ead695c06 [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
Daniel Mack33b48672016-11-23 16:52:29 +010042#include <linux/bpf-cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include <linux/netfilter.h>
44#include <linux/netfilter_ipv6.h>
45
46#include <net/sock.h>
47#include <net/snmp.h>
48
49#include <net/ipv6.h>
50#include <net/ndisc.h>
51#include <net/protocol.h>
52#include <net/ip6_route.h>
53#include <net/addrconf.h>
54#include <net/rawv6.h>
55#include <net/icmp.h>
56#include <net/xfrm.h>
57#include <net/checksum.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090058#include <linux/mroute6.h>
David Ahernca254492015-10-12 11:47:10 -070059#include <net/l3mdev.h>
Roopa Prabhu14972cb2016-08-24 20:10:43 -070060#include <net/lwtunnel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -050062static int ip6_finish_output2(struct net *net, struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -070063{
Eric Dumazetadf30902009-06-02 05:19:30 +000064 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 struct net_device *dev = dst->dev;
David S. Millerf6b72b62011-07-14 07:53:20 -070066 struct neighbour *neigh;
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +000067 struct in6_addr *nexthop;
68 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070069
70 skb->protocol = htons(ETH_P_IPV6);
71 skb->dev = dev;
72
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070073 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) {
Eric Dumazetadf30902009-06-02 05:19:30 +000074 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -070075
David Miller7026b1d2015-04-05 22:19:04 -040076 if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(sk) &&
Eric W. Biederman78126c42015-09-15 20:04:09 -050077 ((mroute6_socket(net, skb) &&
Benjamin Therybd91b8b2008-12-10 16:07:08 -080078 !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) ||
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090079 ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
80 &ipv6_hdr(skb)->saddr))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070081 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
82
83 /* Do not check for IFF_ALLMULTI; multicast routing
84 is not supported in any case.
85 */
86 if (newskb)
Jan Engelhardtb2e0b382010-03-23 04:09:07 +010087 NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING,
Eric W. Biederman29a26a52015-09-15 20:04:16 -050088 net, sk, newskb, NULL, newskb->dev,
Michel Machado95603e22012-06-12 10:16:35 +000089 dev_loopback_xmit);
Linus Torvalds1da177e2005-04-16 15:20:36 -070090
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070091 if (ipv6_hdr(skb)->hop_limit == 0) {
Eric W. Biederman78126c42015-09-15 20:04:09 -050092 IP6_INC_STATS(net, idev,
Denis V. Lunev3bd653c2008-10-08 10:54:51 -070093 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -070094 kfree_skb(skb);
95 return 0;
96 }
97 }
98
Eric W. Biederman78126c42015-09-15 20:04:09 -050099 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUTMCAST, skb->len);
Hannes Frederic Sowadd408512013-02-10 02:33:35 +0000100
101 if (IPV6_ADDR_MC_SCOPE(&ipv6_hdr(skb)->daddr) <=
102 IPV6_ADDR_SCOPE_NODELOCAL &&
103 !(dev->flags & IFF_LOOPBACK)) {
104 kfree_skb(skb);
105 return 0;
106 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 }
108
Roopa Prabhu14972cb2016-08-24 20:10:43 -0700109 if (lwtunnel_xmit_redirect(dst->lwtstate)) {
110 int res = lwtunnel_xmit(skb);
111
112 if (res < 0 || res == LWTUNNEL_XMIT_DONE)
113 return res;
114 }
115
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000116 rcu_read_lock_bh();
Martin KaFai Lau2647a9b2015-05-22 20:55:58 -0700117 nexthop = rt6_nexthop((struct rt6_info *)dst, &ipv6_hdr(skb)->daddr);
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000118 neigh = __ipv6_neigh_lookup_noref(dst->dev, nexthop);
119 if (unlikely(!neigh))
120 neigh = __neigh_create(&nd_tbl, nexthop, dst->dev, false);
121 if (!IS_ERR(neigh)) {
Julian Anastasov4ff06202017-02-06 23:14:12 +0200122 sock_confirm_neigh(skb, neigh);
Julian Anastasovc16ec1852017-02-11 13:49:20 +0200123 ret = neigh_output(neigh, skb);
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000124 rcu_read_unlock_bh();
125 return ret;
126 }
127 rcu_read_unlock_bh();
David S. Miller05e3aa02011-07-16 17:26:00 -0700128
Eric W. Biederman78126c42015-09-15 20:04:09 -0500129 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200130 kfree_skb(skb);
131 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132}
133
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -0500134static int ip6_finish_output(struct net *net, struct sock *sk, struct sk_buff *skb)
Jan Engelhardt9e508492010-04-13 15:28:11 +0200135{
Daniel Mack33b48672016-11-23 16:52:29 +0100136 int ret;
137
138 ret = BPF_CGROUP_RUN_PROG_INET_EGRESS(sk, skb);
139 if (ret) {
140 kfree_skb(skb);
141 return ret;
142 }
143
Jan Engelhardt9e508492010-04-13 15:28:11 +0200144 if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
Jiri Pirko9037c352013-11-06 17:52:19 +0100145 dst_allfrag(skb_dst(skb)) ||
146 (IP6CB(skb)->frag_max_size && skb->len > IP6CB(skb)->frag_max_size))
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500147 return ip6_fragment(net, sk, skb, ip6_finish_output2);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200148 else
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500149 return ip6_finish_output2(net, sk, skb);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200150}
151
Eric W. Biedermanede20592015-10-07 16:48:47 -0500152int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153{
Jan Engelhardt9e508492010-04-13 15:28:11 +0200154 struct net_device *dev = skb_dst(skb)->dev;
Eric Dumazetadf30902009-06-02 05:19:30 +0000155 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Eric W. Biedermanbe10de02015-09-17 17:21:31 -0500156
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900157 if (unlikely(idev->cnf.disable_ipv6)) {
Eric W. Biederman19a06442015-09-15 20:04:10 -0500158 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900159 kfree_skb(skb);
160 return 0;
161 }
162
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500163 return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING,
164 net, sk, skb, NULL, dev,
Jan Engelhardt9c6eb282010-04-13 15:32:16 +0200165 ip6_finish_output,
166 !(IP6CB(skb)->flags & IP6SKB_REROUTED));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167}
168
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169/*
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700170 * xmit an sk_buff (used by TCP, SCTP and DCCP)
171 * Note : socket lock is not held for SYNACK packets, but might be modified
172 * by calls to skb_set_owner_w() and ipv6_local_error(),
173 * which are using proper atomic operations or spinlocks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 */
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700175int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
Pablo Neira92e55f42017-01-26 22:56:21 +0100176 __u32 mark, struct ipv6_txoptions *opt, int tclass)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177{
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700178 struct net *net = sock_net(sk);
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700179 const struct ipv6_pinfo *np = inet6_sk(sk);
David S. Miller4c9483b2011-03-12 16:22:43 -0500180 struct in6_addr *first_hop = &fl6->daddr;
Eric Dumazetadf30902009-06-02 05:19:30 +0000181 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 struct ipv6hdr *hdr;
David S. Miller4c9483b2011-03-12 16:22:43 -0500183 u8 proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 int seg_len = skb->len;
Gerrit Renkere651f032009-08-09 08:12:48 +0000185 int hlimit = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 u32 mtu;
187
188 if (opt) {
Chuck Leverc2636b42007-10-23 21:07:32 -0700189 unsigned int head_room;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190
191 /* First: exthdrs may take lots of space (~8K for now)
192 MAX_HEADER is not enough.
193 */
194 head_room = opt->opt_nflen + opt->opt_flen;
195 seg_len += head_room;
196 head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
197
198 if (skb_headroom(skb) < head_room) {
199 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
Ian Morris63159f22015-03-29 14:00:04 +0100200 if (!skb2) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000201 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900202 IPSTATS_MIB_OUTDISCARDS);
203 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 return -ENOBUFS;
205 }
Eric Dumazet808db802012-04-24 10:17:59 +0000206 consume_skb(skb);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900207 skb = skb2;
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700208 /* skb_set_owner_w() changes sk->sk_wmem_alloc atomically,
209 * it is safe to call in our context (socket lock not held)
210 */
211 skb_set_owner_w(skb, (struct sock *)sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212 }
213 if (opt->opt_flen)
214 ipv6_push_frag_opts(skb, opt, &proto);
215 if (opt->opt_nflen)
David Lebrun613fa3c2016-11-08 14:59:20 +0100216 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop,
217 &fl6->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 }
219
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700220 skb_push(skb, sizeof(struct ipv6hdr));
221 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700222 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223
224 /*
225 * Fill in the IPv6 header
226 */
Eric Dumazetb903d322011-10-27 00:44:35 -0400227 if (np)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 hlimit = np->hop_limit;
229 if (hlimit < 0)
YOSHIFUJI Hideaki6b75d092008-03-10 06:00:30 -0400230 hlimit = ip6_dst_hoplimit(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231
Tom Herbertcb1ce2e2014-07-01 21:33:10 -0700232 ip6_flow_hdr(hdr, tclass, ip6_make_flowlabel(net, skb, fl6->flowlabel,
Tom Herbert67800f92015-07-31 16:52:11 -0700233 np->autoflowlabel, fl6));
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900234
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 hdr->payload_len = htons(seg_len);
236 hdr->nexthdr = proto;
237 hdr->hop_limit = hlimit;
238
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000239 hdr->saddr = fl6->saddr;
240 hdr->daddr = *first_hop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +0200242 skb->protocol = htons(ETH_P_IPV6);
Patrick McHardya2c20642006-01-08 22:37:26 -0800243 skb->priority = sk->sk_priority;
Pablo Neira92e55f42017-01-26 22:56:21 +0100244 skb->mark = mark;
Patrick McHardya2c20642006-01-08 22:37:26 -0800245
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 mtu = dst_mtu(dst);
WANG Cong60ff7462014-05-04 16:39:18 -0700247 if ((skb->len <= mtu) || skb->ignore_df || skb_is_gso(skb)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000248 IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)),
Neil Hormanedf391f2009-04-27 02:45:02 -0700249 IPSTATS_MIB_OUT, skb->len);
David Aherna8e3e1a2016-09-10 12:09:53 -0700250
251 /* if egress device is enslaved to an L3 master device pass the
252 * skb to its handler for processing
253 */
254 skb = l3mdev_ip6_out((struct sock *)sk, skb);
255 if (unlikely(!skb))
256 return 0;
257
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700258 /* hooks should never assume socket lock is held.
259 * we promote our socket to non const
260 */
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500261 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700262 net, (struct sock *)sk, skb, NULL, dst->dev,
Eric W. Biederman13206b62015-10-07 16:48:35 -0500263 dst_output);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 }
265
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 skb->dev = dst->dev;
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700267 /* ipv6_local_error() does not require socket lock,
268 * we promote our socket to non const
269 */
270 ipv6_local_error((struct sock *)sk, EMSGSIZE, fl6, mtu);
271
Eric Dumazetadf30902009-06-02 05:19:30 +0000272 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 kfree_skb(skb);
274 return -EMSGSIZE;
275}
YOSHIFUJI Hideaki71590392007-02-22 22:05:40 +0900276EXPORT_SYMBOL(ip6_xmit);
277
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
279{
280 struct ip6_ra_chain *ra;
281 struct sock *last = NULL;
282
283 read_lock(&ip6_ra_lock);
284 for (ra = ip6_ra_chain; ra; ra = ra->next) {
285 struct sock *sk = ra->sk;
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700286 if (sk && ra->sel == sel &&
287 (!sk->sk_bound_dev_if ||
288 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 if (last) {
290 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
291 if (skb2)
292 rawv6_rcv(last, skb2);
293 }
294 last = sk;
295 }
296 }
297
298 if (last) {
299 rawv6_rcv(last, skb);
300 read_unlock(&ip6_ra_lock);
301 return 1;
302 }
303 read_unlock(&ip6_ra_lock);
304 return 0;
305}
306
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700307static int ip6_forward_proxy_check(struct sk_buff *skb)
308{
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700309 struct ipv6hdr *hdr = ipv6_hdr(skb);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700310 u8 nexthdr = hdr->nexthdr;
Jesse Gross75f28112011-11-30 17:05:51 -0800311 __be16 frag_off;
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700312 int offset;
313
314 if (ipv6_ext_hdr(nexthdr)) {
Jesse Gross75f28112011-11-30 17:05:51 -0800315 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700316 if (offset < 0)
317 return 0;
318 } else
319 offset = sizeof(struct ipv6hdr);
320
321 if (nexthdr == IPPROTO_ICMPV6) {
322 struct icmp6hdr *icmp6;
323
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700324 if (!pskb_may_pull(skb, (skb_network_header(skb) +
325 offset + 1 - skb->data)))
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700326 return 0;
327
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700328 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700329
330 switch (icmp6->icmp6_type) {
331 case NDISC_ROUTER_SOLICITATION:
332 case NDISC_ROUTER_ADVERTISEMENT:
333 case NDISC_NEIGHBOUR_SOLICITATION:
334 case NDISC_NEIGHBOUR_ADVERTISEMENT:
335 case NDISC_REDIRECT:
336 /* For reaction involving unicast neighbor discovery
337 * message destined to the proxied address, pass it to
338 * input function.
339 */
340 return 1;
341 default:
342 break;
343 }
344 }
345
Ville Nuorvala74553b02006-09-22 14:42:18 -0700346 /*
347 * The proxying router can't forward traffic sent to a link-local
348 * address, so signal the sender and discard the packet. This
349 * behavior is clarified by the MIPv6 specification.
350 */
351 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
352 dst_link_failure(skb);
353 return -1;
354 }
355
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700356 return 0;
357}
358
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -0500359static inline int ip6_forward_finish(struct net *net, struct sock *sk,
360 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361{
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 */
David S. Millerc45a3df2012-01-27 15:32:19 -0800488 if (skb->dev == dst->dev && opt->srcrt == 0 && !skb_sec_path(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 struct in6_addr *target = NULL;
David S. Millerfbfe95a2012-06-08 23:24:18 -0700490 struct inet_peer *peer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492
493 /*
494 * incoming and outgoing devices are the same
495 * send a redirect.
496 */
497
498 rt = (struct rt6_info *) dst;
David S. Millerc45a3df2012-01-27 15:32:19 -0800499 if (rt->rt6i_flags & RTF_GATEWAY)
500 target = &rt->rt6i_gateway;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 else
502 target = &hdr->daddr;
503
Martin KaFai Laufd0273d2015-05-22 20:55:57 -0700504 peer = inet_getpeer_v6(net->ipv6.peers, &hdr->daddr, 1);
David S. Miller92d86822011-02-04 15:55:25 -0800505
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 /* Limit redirects both by destination (here)
507 and by source (inside ndisc_send_redirect)
508 */
David S. Millerfbfe95a2012-06-08 23:24:18 -0700509 if (inet_peer_xrlim_allow(peer, 1*HZ))
David S. Miller49919692012-01-27 15:30:48 -0800510 ndisc_send_redirect(skb, target);
David S. Miller1d861aa2012-07-10 03:58:16 -0700511 if (peer)
512 inet_putpeer(peer);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700513 } else {
514 int addrtype = ipv6_addr_type(&hdr->saddr);
515
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 /* This check is security critical. */
YOSHIFUJI Hideakif81b2e72008-06-25 16:55:26 +0900517 if (addrtype == IPV6_ADDR_ANY ||
518 addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK))
David L Stevens5bb1ab02007-05-09 13:53:44 -0700519 goto error;
520 if (addrtype & IPV6_ADDR_LINKLOCAL) {
521 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000522 ICMPV6_NOT_NEIGHBOUR, 0);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700523 goto error;
524 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 }
526
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100527 mtu = ip6_dst_mtu_forward(dst);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800528 if (mtu < IPV6_MIN_MTU)
529 mtu = IPV6_MIN_MTU;
530
Florian Westphalfe6cc552014-02-13 23:09:12 +0100531 if (ip6_pkt_too_big(skb, mtu)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 /* Again, force OUTPUT device used as source address */
533 skb->dev = dst->dev;
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800534 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazet1d015502016-04-27 16:44:40 -0700535 __IP6_INC_STATS(net, ip6_dst_idev(dst),
536 IPSTATS_MIB_INTOOBIGERRORS);
537 __IP6_INC_STATS(net, ip6_dst_idev(dst),
538 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 kfree_skb(skb);
540 return -EMSGSIZE;
541 }
542
543 if (skb_cow(skb, dst->dev->hard_header_len)) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700544 __IP6_INC_STATS(net, ip6_dst_idev(dst),
545 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 goto drop;
547 }
548
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700549 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550
551 /* Mangling hops number delayed to point after skb COW */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900552
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 hdr->hop_limit--;
554
Eric Dumazet1d015502016-04-27 16:44:40 -0700555 __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
556 __IP6_ADD_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len);
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500557 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
558 net, NULL, skb, skb->dev, dst->dev,
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800559 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560
561error:
Eric Dumazet1d015502016-04-27 16:44:40 -0700562 __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563drop:
564 kfree_skb(skb);
565 return -EINVAL;
566}
567
568static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
569{
570 to->pkt_type = from->pkt_type;
571 to->priority = from->priority;
572 to->protocol = from->protocol;
Eric Dumazetadf30902009-06-02 05:19:30 +0000573 skb_dst_drop(to);
574 skb_dst_set(to, dst_clone(skb_dst(from)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 to->dev = from->dev;
Thomas Graf82e91ff2006-11-09 15:19:14 -0800576 to->mark = from->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577
578#ifdef CONFIG_NET_SCHED
579 to->tc_index = from->tc_index;
580#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -0700581 nf_copy(to, from);
James Morris984bc162006-06-09 00:29:17 -0700582 skb_copy_secmark(to, from);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583}
584
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500585int ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
586 int (*output)(struct net *, struct sock *, struct sk_buff *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 struct sk_buff *frag;
Ian Morris67ba4152014-08-24 21:53:10 +0100589 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +0200590 struct ipv6_pinfo *np = skb->sk && !dev_recursion_level() ?
591 inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 struct ipv6hdr *tmp_hdr;
593 struct frag_hdr *fh;
594 unsigned int mtu, hlen, left, len;
Herbert Xua7ae1992011-11-18 02:20:04 +0000595 int hroom, troom;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700596 __be32 frag_id;
Ian Morris67ba4152014-08-24 21:53:10 +0100597 int ptr, offset = 0, err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 u8 *prevhdr, nexthdr = 0;
599
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 hlen = ip6_find_1stfragopt(skb, &prevhdr);
601 nexthdr = *prevhdr;
602
John Heffner628a5c52007-04-20 15:53:27 -0700603 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700604
605 /* We must not fragment if the socket is set to force MTU discovery
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800606 * or if the skb it not generated by a local socket.
John Heffnerb881ef72007-04-20 15:52:39 -0700607 */
Florian Westphal485fca62015-05-22 00:44:16 +0200608 if (unlikely(!skb->ignore_df && skb->len > mtu))
609 goto fail_toobig;
Eric Dumazeta34a1012012-05-18 21:51:44 +0000610
Florian Westphal485fca62015-05-22 00:44:16 +0200611 if (IP6CB(skb)->frag_max_size) {
612 if (IP6CB(skb)->frag_max_size > mtu)
613 goto fail_toobig;
614
615 /* don't send fragments larger than what we received */
616 mtu = IP6CB(skb)->frag_max_size;
617 if (mtu < IPV6_MIN_MTU)
618 mtu = IPV6_MIN_MTU;
John Heffnerb881ef72007-04-20 15:52:39 -0700619 }
620
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800621 if (np && np->frag_size < mtu) {
622 if (np->frag_size)
623 mtu = np->frag_size;
624 }
Hannes Frederic Sowa89bc7842015-10-28 13:21:04 +0100625 if (mtu < hlen + sizeof(struct frag_hdr) + 8)
Hannes Frederic Sowab72a2b02015-10-16 11:32:43 +0200626 goto fail_toobig;
Hannes Frederic Sowa1e0d69a2015-10-28 13:21:03 +0100627 mtu -= hlen + sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628
Martin KaFai Laufd0273d2015-05-22 20:55:57 -0700629 frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
630 &ipv6_hdr(skb)->saddr);
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700631
Hannes Frederic Sowa405c92f2015-10-27 22:40:42 +0100632 if (skb->ip_summed == CHECKSUM_PARTIAL &&
633 (err = skb_checksum_help(skb)))
634 goto fail;
635
Florian Westphal1d325d22015-09-16 17:26:14 +0200636 hroom = LL_RESERVED_SPACE(rt->dst.dev);
David S. Miller21dc3302010-08-23 00:13:46 -0700637 if (skb_has_frag_list(skb)) {
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +0300638 unsigned int first_len = skb_pagelen(skb);
Eric Dumazet3d130082010-09-21 08:47:45 +0000639 struct sk_buff *frag2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640
641 if (first_len - hlen > mtu ||
642 ((first_len - hlen) & 7) ||
Florian Westphal1d325d22015-09-16 17:26:14 +0200643 skb_cloned(skb) ||
644 skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645 goto slow_path;
646
David S. Miller4d9092b2009-06-09 00:20:05 -0700647 skb_walk_frags(skb, frag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 /* Correct geometry. */
649 if (frag->len > mtu ||
650 ((frag->len & 7) && frag->next) ||
Florian Westphal1d325d22015-09-16 17:26:14 +0200651 skb_headroom(frag) < (hlen + hroom + sizeof(struct frag_hdr)))
Eric Dumazet3d130082010-09-21 08:47:45 +0000652 goto slow_path_clean;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 /* Partially cloned skb? */
655 if (skb_shared(frag))
Eric Dumazet3d130082010-09-21 08:47:45 +0000656 goto slow_path_clean;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700657
658 BUG_ON(frag->sk);
659 if (skb->sk) {
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700660 frag->sk = skb->sk;
661 frag->destructor = sock_wfree;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700662 }
Eric Dumazet3d130082010-09-21 08:47:45 +0000663 skb->truesize -= frag->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 }
665
666 err = 0;
667 offset = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 /* BUILD HEADER */
669
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800670 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700671 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 if (!tmp_hdr) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000673 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700674 IPSTATS_MIB_FRAGFAILS);
Florian Westphal1d325d22015-09-16 17:26:14 +0200675 err = -ENOMEM;
676 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 }
Florian Westphal1d325d22015-09-16 17:26:14 +0200678 frag = skb_shinfo(skb)->frag_list;
679 skb_frag_list_init(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 __skb_pull(skb, hlen);
Ian Morris67ba4152014-08-24 21:53:10 +0100682 fh = (struct frag_hdr *)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700683 __skb_push(skb, hlen);
684 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700685 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 fh->nexthdr = nexthdr;
688 fh->reserved = 0;
689 fh->frag_off = htons(IP6_MF);
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700690 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691
692 first_len = skb_pagelen(skb);
693 skb->data_len = first_len - skb_headlen(skb);
694 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700695 ipv6_hdr(skb)->payload_len = htons(first_len -
696 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900697
Changli Gaod8d1f302010-06-10 23:31:35 -0700698 dst_hold(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699
700 for (;;) {
701 /* Prepare header of the next frame,
702 * before previous one went down. */
703 if (frag) {
704 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300705 skb_reset_transport_header(frag);
Ian Morris67ba4152014-08-24 21:53:10 +0100706 fh = (struct frag_hdr *)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700707 __skb_push(frag, hlen);
708 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700709 memcpy(skb_network_header(frag), tmp_hdr,
710 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 offset += skb->len - hlen - sizeof(struct frag_hdr);
712 fh->nexthdr = nexthdr;
713 fh->reserved = 0;
714 fh->frag_off = htons(offset);
Ian Morris53b24b82015-03-29 14:00:05 +0100715 if (frag->next)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 fh->frag_off |= htons(IP6_MF);
717 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700718 ipv6_hdr(frag)->payload_len =
719 htons(frag->len -
720 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 ip6_copy_metadata(frag, skb);
722 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900723
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500724 err = output(net, sk, skb);
Ian Morris67ba4152014-08-24 21:53:10 +0100725 if (!err)
Changli Gaod8d1f302010-06-10 23:31:35 -0700726 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700727 IPSTATS_MIB_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700728
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 if (err || !frag)
730 break;
731
732 skb = frag;
733 frag = skb->next;
734 skb->next = NULL;
735 }
736
Jesper Juhla51482b2005-11-08 09:41:34 -0800737 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738
739 if (err == 0) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700740 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700741 IPSTATS_MIB_FRAGOKS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000742 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 return 0;
744 }
745
Florian Westphal46cfd722014-09-10 01:08:46 +0200746 kfree_skb_list(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747
Changli Gaod8d1f302010-06-10 23:31:35 -0700748 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700749 IPSTATS_MIB_FRAGFAILS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000750 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 return err;
Eric Dumazet3d130082010-09-21 08:47:45 +0000752
753slow_path_clean:
754 skb_walk_frags(skb, frag2) {
755 if (frag2 == frag)
756 break;
757 frag2->sk = NULL;
758 frag2->destructor = NULL;
759 skb->truesize += frag2->truesize;
760 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 }
762
763slow_path:
764 left = skb->len - hlen; /* Space per frame */
765 ptr = hlen; /* Where to start from */
766
767 /*
768 * Fragment the datagram.
769 */
770
Herbert Xua7ae1992011-11-18 02:20:04 +0000771 troom = rt->dst.dev->needed_tailroom;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772
773 /*
774 * Keep copying data until we run out.
775 */
Ian Morris67ba4152014-08-24 21:53:10 +0100776 while (left > 0) {
Florian Westphal79e49502017-03-13 16:24:28 +0100777 u8 *fragnexthdr_offset;
778
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 len = left;
780 /* IF: it doesn't fit, use 'mtu' - the data space left */
781 if (len > mtu)
782 len = mtu;
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300783 /* IF: we are not sending up to and including the packet end
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 then align the next start on an eight byte boundary */
785 if (len < left) {
786 len &= ~7;
787 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
Joe Perchescbffccc2014-11-05 14:39:21 -0800789 /* Allocate buffer */
790 frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) +
791 hroom + troom, GFP_ATOMIC);
792 if (!frag) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000793 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900794 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 err = -ENOMEM;
796 goto fail;
797 }
798
799 /*
800 * Set up data on packet
801 */
802
803 ip6_copy_metadata(frag, skb);
Herbert Xua7ae1992011-11-18 02:20:04 +0000804 skb_reserve(frag, hroom);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700806 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300807 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700808 frag->transport_header = (frag->network_header + hlen +
809 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810
811 /*
812 * Charge the memory for the fragment to any owner
813 * it might possess
814 */
815 if (skb->sk)
816 skb_set_owner_w(frag, skb->sk);
817
818 /*
819 * Copy the packet header into the new buffer.
820 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300821 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822
Florian Westphal79e49502017-03-13 16:24:28 +0100823 fragnexthdr_offset = skb_network_header(frag);
824 fragnexthdr_offset += prevhdr - skb_network_header(skb);
825 *fragnexthdr_offset = NEXTHDR_FRAGMENT;
826
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 /*
828 * Build fragment header.
829 */
830 fh->nexthdr = nexthdr;
831 fh->reserved = 0;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700832 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833
834 /*
835 * Copy a block of the IP datagram.
836 */
Himangi Saraogie3f0b862014-07-12 01:57:17 +0530837 BUG_ON(skb_copy_bits(skb, ptr, skb_transport_header(frag),
838 len));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 left -= len;
840
841 fh->frag_off = htons(offset);
842 if (left > 0)
843 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700844 ipv6_hdr(frag)->payload_len = htons(frag->len -
845 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846
847 ptr += len;
848 offset += len;
849
850 /*
851 * Put this fragment into the sending queue.
852 */
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500853 err = output(net, sk, frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 if (err)
855 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700856
Eric Dumazetadf30902009-06-02 05:19:30 +0000857 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700858 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000860 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900861 IPSTATS_MIB_FRAGOKS);
Eric Dumazet808db802012-04-24 10:17:59 +0000862 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 return err;
864
Florian Westphal485fca62015-05-22 00:44:16 +0200865fail_toobig:
866 if (skb->sk && dst_allfrag(skb_dst(skb)))
867 sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK);
868
869 skb->dev = skb_dst(skb)->dev;
870 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
871 err = -EMSGSIZE;
872
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873fail:
Eric Dumazetadf30902009-06-02 05:19:30 +0000874 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900875 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900876 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 return err;
878}
879
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000880static inline int ip6_rt_check(const struct rt6key *rt_key,
881 const struct in6_addr *fl_addr,
882 const struct in6_addr *addr_cache)
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700883{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000884 return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
Ian Morris63159f22015-03-29 14:00:04 +0100885 (!addr_cache || !ipv6_addr_equal(fl_addr, addr_cache));
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700886}
887
Herbert Xu497c6152006-07-30 20:19:33 -0700888static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
889 struct dst_entry *dst,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000890 const struct flowi6 *fl6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891{
Herbert Xu497c6152006-07-30 20:19:33 -0700892 struct ipv6_pinfo *np = inet6_sk(sk);
Eric Dumazeta963a372013-06-26 04:15:07 -0700893 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894
Herbert Xu497c6152006-07-30 20:19:33 -0700895 if (!dst)
896 goto out;
897
Eric Dumazeta963a372013-06-26 04:15:07 -0700898 if (dst->ops->family != AF_INET6) {
899 dst_release(dst);
900 return NULL;
901 }
902
903 rt = (struct rt6_info *)dst;
Herbert Xu497c6152006-07-30 20:19:33 -0700904 /* Yes, checking route validity in not connected
905 * case is not very simple. Take into account,
906 * that we do not support routing by source, TOS,
Ian Morris67ba4152014-08-24 21:53:10 +0100907 * and MSG_DONTROUTE --ANK (980726)
Herbert Xu497c6152006-07-30 20:19:33 -0700908 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700909 * 1. ip6_rt_check(): If route was host route,
910 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700911 * If it is network route, we still may
912 * check its validity using saved pointer
913 * to the last used address: daddr_cache.
914 * We do not want to save whole address now,
915 * (because main consumer of this service
916 * is tcp, which has not this problem),
917 * so that the last trick works only on connected
918 * sockets.
919 * 2. oif also should be the same.
920 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500921 if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700922#ifdef CONFIG_IPV6_SUBTREES
David S. Miller4c9483b2011-03-12 16:22:43 -0500923 ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700924#endif
David Ahernca254492015-10-12 11:47:10 -0700925 (!(fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF) &&
926 (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex))) {
Herbert Xu497c6152006-07-30 20:19:33 -0700927 dst_release(dst);
928 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 }
930
Herbert Xu497c6152006-07-30 20:19:33 -0700931out:
932 return dst;
933}
934
Eric Dumazet3aef9342015-09-25 07:39:12 -0700935static int ip6_dst_lookup_tail(struct net *net, const struct sock *sk,
David S. Miller4c9483b2011-03-12 16:22:43 -0500936 struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700937{
David S. Miller69cce1d2011-07-17 23:09:49 -0700938#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
939 struct neighbour *n;
David S. Miller97cac082012-07-02 22:43:47 -0700940 struct rt6_info *rt;
David S. Miller69cce1d2011-07-17 23:09:49 -0700941#endif
942 int err;
Paolo Abeni6f21c962016-01-29 12:30:19 +0100943 int flags = 0;
Herbert Xu497c6152006-07-30 20:19:33 -0700944
Markus Stenberge16e8882015-05-05 13:36:59 +0300945 /* The correct way to handle this would be to do
946 * ip6_route_get_saddr, and then ip6_route_output; however,
947 * the route-specific preferred source forces the
948 * ip6_route_output call _before_ ip6_route_get_saddr.
949 *
950 * In source specific routing (no src=any default route),
951 * ip6_route_output will fail given src=any saddr, though, so
952 * that's why we try it again later.
953 */
954 if (ipv6_addr_any(&fl6->saddr) && (!*dst || !(*dst)->error)) {
955 struct rt6_info *rt;
956 bool had_dst = *dst != NULL;
957
958 if (!had_dst)
959 *dst = ip6_route_output(net, sk, fl6);
960 rt = (*dst)->error ? NULL : (struct rt6_info *)*dst;
961 err = ip6_route_get_saddr(net, rt, &fl6->daddr,
962 sk ? inet6_sk(sk)->srcprefs : 0,
963 &fl6->saddr);
964 if (err)
965 goto out_err_release;
966
967 /* If we had an erroneous initial result, pretend it
968 * never existed and let the SA-enabled version take
969 * over.
970 */
971 if (!had_dst && (*dst)->error) {
972 dst_release(*dst);
973 *dst = NULL;
974 }
Paolo Abeni6f21c962016-01-29 12:30:19 +0100975
976 if (fl6->flowi6_oif)
977 flags |= RT6_LOOKUP_F_IFACE;
Markus Stenberge16e8882015-05-05 13:36:59 +0300978 }
979
Ian Morris63159f22015-03-29 14:00:04 +0100980 if (!*dst)
Paolo Abeni6f21c962016-01-29 12:30:19 +0100981 *dst = ip6_route_output_flags(net, sk, fl6, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
Ian Morrise5d08d72014-11-23 21:28:43 +0000983 err = (*dst)->error;
984 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 goto out_err_release;
986
Neil Horman95c385b2007-04-25 17:08:10 -0700987#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700988 /*
989 * Here if the dst entry we've looked up
990 * has a neighbour entry that is in the INCOMPLETE
991 * state and the src address from the flow is
992 * marked as OPTIMISTIC, we release the found
993 * dst entry and replace it instead with the
994 * dst entry of the nexthop router
995 */
Eric Dumazetc56bf6f2012-07-06 09:19:05 +0200996 rt = (struct rt6_info *) *dst;
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000997 rcu_read_lock_bh();
Martin KaFai Lau2647a9b2015-05-22 20:55:58 -0700998 n = __ipv6_neigh_lookup_noref(rt->dst.dev,
999 rt6_nexthop(rt, &fl6->daddr));
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +00001000 err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0;
1001 rcu_read_unlock_bh();
1002
1003 if (err) {
Neil Hormane550dfb2008-09-09 13:51:35 -07001004 struct inet6_ifaddr *ifp;
David S. Miller4c9483b2011-03-12 16:22:43 -05001005 struct flowi6 fl_gw6;
Neil Hormane550dfb2008-09-09 13:51:35 -07001006 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -07001007
David S. Miller4c9483b2011-03-12 16:22:43 -05001008 ifp = ipv6_get_ifaddr(net, &fl6->saddr,
Neil Hormane550dfb2008-09-09 13:51:35 -07001009 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -07001010
Neil Hormane550dfb2008-09-09 13:51:35 -07001011 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
1012 if (ifp)
1013 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -07001014
Neil Hormane550dfb2008-09-09 13:51:35 -07001015 if (redirect) {
1016 /*
1017 * We need to get the dst entry for the
1018 * default router instead
1019 */
1020 dst_release(*dst);
David S. Miller4c9483b2011-03-12 16:22:43 -05001021 memcpy(&fl_gw6, fl6, sizeof(struct flowi6));
1022 memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr));
1023 *dst = ip6_route_output(net, sk, &fl_gw6);
Ian Morrise5d08d72014-11-23 21:28:43 +00001024 err = (*dst)->error;
1025 if (err)
Neil Hormane550dfb2008-09-09 13:51:35 -07001026 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -07001027 }
Neil Hormane550dfb2008-09-09 13:51:35 -07001028 }
Neil Horman95c385b2007-04-25 17:08:10 -07001029#endif
Jonathan T. Leightonec5e3b02017-02-12 17:26:06 -05001030 if (ipv6_addr_v4mapped(&fl6->saddr) &&
Willem de Bruijn00ea1ce2017-02-18 19:00:45 -05001031 !(ipv6_addr_v4mapped(&fl6->daddr) || ipv6_addr_any(&fl6->daddr))) {
1032 err = -EAFNOSUPPORT;
1033 goto out_err_release;
1034 }
Neil Horman95c385b2007-04-25 17:08:10 -07001035
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 return 0;
1037
1038out_err_release:
1039 dst_release(*dst);
1040 *dst = NULL;
David Ahern8a966fc2016-09-10 12:09:59 -07001041
David Ahern0d240e72016-06-16 16:24:25 -07001042 if (err == -ENETUNREACH)
1043 IP6_INC_STATS(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 return err;
1045}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001046
Herbert Xu497c6152006-07-30 20:19:33 -07001047/**
1048 * ip6_dst_lookup - perform route lookup on flow
1049 * @sk: socket which provides route info
1050 * @dst: pointer to dst_entry * for result
David S. Miller4c9483b2011-03-12 16:22:43 -05001051 * @fl6: flow to lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001052 *
1053 * This function performs a route lookup on the given flow.
1054 *
1055 * It returns zero on success, or a standard errno code on error.
1056 */
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001057int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1058 struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -07001059{
1060 *dst = NULL;
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001061 return ip6_dst_lookup_tail(net, sk, dst, fl6);
Herbert Xu497c6152006-07-30 20:19:33 -07001062}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -08001063EXPORT_SYMBOL_GPL(ip6_dst_lookup);
1064
Herbert Xu497c6152006-07-30 20:19:33 -07001065/**
David S. Miller68d0c6d2011-03-01 13:19:07 -08001066 * ip6_dst_lookup_flow - perform route lookup on flow with ipsec
1067 * @sk: socket which provides route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001068 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001069 * @final_dst: final destination address for ipsec lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001070 *
1071 * This function performs a route lookup on the given flow.
1072 *
1073 * It returns a valid dst pointer on success, or a pointer encoded
1074 * error code.
1075 */
Eric Dumazet3aef9342015-09-25 07:39:12 -07001076struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001077 const struct in6_addr *final_dst)
David S. Miller68d0c6d2011-03-01 13:19:07 -08001078{
1079 struct dst_entry *dst = NULL;
1080 int err;
1081
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001082 err = ip6_dst_lookup_tail(sock_net(sk), sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001083 if (err)
1084 return ERR_PTR(err);
1085 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001086 fl6->daddr = *final_dst;
David S. Miller2774c132011-03-01 14:59:04 -08001087
Steffen Klassertf92ee612014-09-16 10:08:40 +02001088 return xfrm_lookup_route(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001089}
1090EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow);
1091
1092/**
1093 * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow
1094 * @sk: socket which provides the dst cache and route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001095 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001096 * @final_dst: final destination address for ipsec lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001097 *
1098 * This function performs a route lookup on the given flow with the
1099 * possibility of using the cached route in the socket if it is valid.
1100 * It will take the socket dst lock when operating on the dst cache.
1101 * As a result, this function can only be used in process context.
1102 *
David S. Miller68d0c6d2011-03-01 13:19:07 -08001103 * It returns a valid dst pointer on success, or a pointer encoded
1104 * error code.
Herbert Xu497c6152006-07-30 20:19:33 -07001105 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001106struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001107 const struct in6_addr *final_dst)
Herbert Xu497c6152006-07-30 20:19:33 -07001108{
David S. Miller68d0c6d2011-03-01 13:19:07 -08001109 struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
Herbert Xu497c6152006-07-30 20:19:33 -07001110
David S. Miller4c9483b2011-03-12 16:22:43 -05001111 dst = ip6_sk_dst_check(sk, dst, fl6);
Jakub Sitnicki00bc0ef2016-06-08 15:13:34 +02001112 if (!dst)
1113 dst = ip6_dst_lookup_flow(sk, fl6, final_dst);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001114
Jakub Sitnicki00bc0ef2016-06-08 15:13:34 +02001115 return dst;
Herbert Xu497c6152006-07-30 20:19:33 -07001116}
David S. Miller68d0c6d2011-03-01 13:19:07 -08001117EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
Herbert Xu497c6152006-07-30 20:19:33 -07001118
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001119static inline int ip6_ufo_append_data(struct sock *sk,
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001120 struct sk_buff_head *queue,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001121 int getfrag(void *from, char *to, int offset, int len,
1122 int odd, struct sk_buff *skb),
1123 void *from, int length, int hh_len, int fragheaderlen,
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001124 int exthdrlen, int transhdrlen, int mtu,
1125 unsigned int flags, const struct flowi6 *fl6)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001126
1127{
1128 struct sk_buff *skb;
1129 int err;
1130
1131 /* There is support for UDP large send offload by network
1132 * device, so create one single skb packet containing complete
1133 * udp datagram
1134 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001135 skb = skb_peek_tail(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001136 if (!skb) {
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001137 skb = sock_alloc_send_skb(sk,
1138 hh_len + fragheaderlen + transhdrlen + 20,
1139 (flags & MSG_DONTWAIT), &err);
Ian Morris63159f22015-03-29 14:00:04 +01001140 if (!skb)
Zheng Yan504744e2011-10-28 00:26:00 -04001141 return err;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001142
1143 /* reserve space for Hardware header */
1144 skb_reserve(skb, hh_len);
1145
1146 /* create space for UDP/IP header */
Ian Morris67ba4152014-08-24 21:53:10 +01001147 skb_put(skb, fragheaderlen + transhdrlen);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001148
1149 /* initialize network header pointer */
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001150 skb_set_network_header(skb, exthdrlen);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001151
1152 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001153 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001154
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001155 skb->protocol = htons(ETH_P_IPV6);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001156 skb->csum = 0;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001157
Julian Anastasov0dec8792017-02-06 23:14:16 +02001158 if (flags & MSG_CONFIRM)
1159 skb_set_dst_pending_confirm(skb, 1);
1160
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001161 __skb_queue_tail(queue, skb);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001162 } else if (skb_is_gso(skb)) {
1163 goto append;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001164 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001165
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001166 skb->ip_summed = CHECKSUM_PARTIAL;
1167 /* Specify the length of each IPv6 datagram fragment.
1168 * It has to be a multiple of 8.
1169 */
1170 skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
1171 sizeof(struct frag_hdr)) & ~7;
1172 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001173 skb_shinfo(skb)->ip6_frag_id = ipv6_select_ident(sock_net(sk),
1174 &fl6->daddr,
1175 &fl6->saddr);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001176
1177append:
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001178 return skb_append_datato_frags(sk, skb, getfrag, from,
1179 (length - transhdrlen));
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001180}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181
Herbert Xu0178b692009-02-05 15:15:50 -08001182static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
1183 gfp_t gfp)
1184{
1185 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1186}
1187
1188static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
1189 gfp_t gfp)
1190{
1191 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1192}
1193
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001194static void ip6_append_data_mtu(unsigned int *mtu,
Gao feng0c183372012-05-26 01:30:53 +00001195 int *maxfraglen,
1196 unsigned int fragheaderlen,
1197 struct sk_buff *skb,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001198 struct rt6_info *rt,
luciene367c2d2014-03-17 12:51:01 +08001199 unsigned int orig_mtu)
Gao feng0c183372012-05-26 01:30:53 +00001200{
1201 if (!(rt->dst.flags & DST_XFRM_TUNNEL)) {
Ian Morris63159f22015-03-29 14:00:04 +01001202 if (!skb) {
Gao feng0c183372012-05-26 01:30:53 +00001203 /* first fragment, reserve header_len */
luciene367c2d2014-03-17 12:51:01 +08001204 *mtu = orig_mtu - rt->dst.header_len;
Gao feng0c183372012-05-26 01:30:53 +00001205
1206 } else {
1207 /*
1208 * this fragment is not first, the headers
1209 * space is regarded as data space.
1210 */
luciene367c2d2014-03-17 12:51:01 +08001211 *mtu = orig_mtu;
Gao feng0c183372012-05-26 01:30:53 +00001212 }
1213 *maxfraglen = ((*mtu - fragheaderlen) & ~7)
1214 + fragheaderlen - sizeof(struct frag_hdr);
1215 }
1216}
1217
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001218static int ip6_setup_cork(struct sock *sk, struct inet_cork_full *cork,
Wei Wang26879da2016-05-02 21:40:07 -07001219 struct inet6_cork *v6_cork, struct ipcm6_cookie *ipc6,
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001220 struct rt6_info *rt, struct flowi6 *fl6)
1221{
1222 struct ipv6_pinfo *np = inet6_sk(sk);
1223 unsigned int mtu;
Wei Wang26879da2016-05-02 21:40:07 -07001224 struct ipv6_txoptions *opt = ipc6->opt;
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001225
1226 /*
1227 * setup for corking
1228 */
1229 if (opt) {
1230 if (WARN_ON(v6_cork->opt))
1231 return -EINVAL;
1232
1233 v6_cork->opt = kzalloc(opt->tot_len, sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001234 if (unlikely(!v6_cork->opt))
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001235 return -ENOBUFS;
1236
1237 v6_cork->opt->tot_len = opt->tot_len;
1238 v6_cork->opt->opt_flen = opt->opt_flen;
1239 v6_cork->opt->opt_nflen = opt->opt_nflen;
1240
1241 v6_cork->opt->dst0opt = ip6_opt_dup(opt->dst0opt,
1242 sk->sk_allocation);
1243 if (opt->dst0opt && !v6_cork->opt->dst0opt)
1244 return -ENOBUFS;
1245
1246 v6_cork->opt->dst1opt = ip6_opt_dup(opt->dst1opt,
1247 sk->sk_allocation);
1248 if (opt->dst1opt && !v6_cork->opt->dst1opt)
1249 return -ENOBUFS;
1250
1251 v6_cork->opt->hopopt = ip6_opt_dup(opt->hopopt,
1252 sk->sk_allocation);
1253 if (opt->hopopt && !v6_cork->opt->hopopt)
1254 return -ENOBUFS;
1255
1256 v6_cork->opt->srcrt = ip6_rthdr_dup(opt->srcrt,
1257 sk->sk_allocation);
1258 if (opt->srcrt && !v6_cork->opt->srcrt)
1259 return -ENOBUFS;
1260
1261 /* need source address above miyazawa*/
1262 }
1263 dst_hold(&rt->dst);
1264 cork->base.dst = &rt->dst;
1265 cork->fl.u.ip6 = *fl6;
Wei Wang26879da2016-05-02 21:40:07 -07001266 v6_cork->hop_limit = ipc6->hlimit;
1267 v6_cork->tclass = ipc6->tclass;
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001268 if (rt->dst.flags & DST_XFRM_TUNNEL)
1269 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
1270 rt->dst.dev->mtu : dst_mtu(&rt->dst);
1271 else
1272 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
1273 rt->dst.dev->mtu : dst_mtu(rt->dst.path);
1274 if (np->frag_size < mtu) {
1275 if (np->frag_size)
1276 mtu = np->frag_size;
1277 }
1278 cork->base.fragsize = mtu;
1279 if (dst_allfrag(rt->dst.path))
1280 cork->base.flags |= IPCORK_ALLFRAG;
1281 cork->base.length = 0;
1282
1283 return 0;
1284}
1285
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001286static int __ip6_append_data(struct sock *sk,
1287 struct flowi6 *fl6,
1288 struct sk_buff_head *queue,
1289 struct inet_cork *cork,
1290 struct inet6_cork *v6_cork,
1291 struct page_frag *pfrag,
1292 int getfrag(void *from, char *to, int offset,
1293 int len, int odd, struct sk_buff *skb),
1294 void *from, int length, int transhdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001295 unsigned int flags, struct ipcm6_cookie *ipc6,
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001296 const struct sockcm_cookie *sockc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297{
Gao feng0c183372012-05-26 01:30:53 +00001298 struct sk_buff *skb, *skb_prev = NULL;
luciene367c2d2014-03-17 12:51:01 +08001299 unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001300 int exthdrlen = 0;
1301 int dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302 int hh_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303 int copy;
1304 int err;
1305 int offset = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001306 __u8 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001307 u32 tskey = 0;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001308 struct rt6_info *rt = (struct rt6_info *)cork->dst;
1309 struct ipv6_txoptions *opt = v6_cork->opt;
Vlad Yasevich32dce962015-01-31 10:40:18 -05001310 int csummode = CHECKSUM_NONE;
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001311 unsigned int maxnonfragsize, headersize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001313 skb = skb_peek_tail(queue);
1314 if (!skb) {
1315 exthdrlen = opt ? opt->opt_flen : 0;
Romain KUNTZ7efdba52013-01-16 12:47:40 +00001316 dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317 }
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001318
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001319 mtu = cork->fragsize;
luciene367c2d2014-03-17 12:51:01 +08001320 orig_mtu = mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321
Changli Gaod8d1f302010-06-10 23:31:35 -07001322 hh_len = LL_RESERVED_SPACE(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001324 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001325 (opt ? opt->opt_nflen : 0);
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001326 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen -
1327 sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001329 headersize = sizeof(struct ipv6hdr) +
1330 (opt ? opt->opt_flen + opt->opt_nflen : 0) +
1331 (dst_allfrag(&rt->dst) ?
1332 sizeof(struct frag_hdr) : 0) +
1333 rt->rt6i_nfheader_len;
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001334
Wei Wang26879da2016-05-02 21:40:07 -07001335 if (cork->length + length > mtu - headersize && ipc6->dontfrag &&
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001336 (sk->sk_protocol == IPPROTO_UDP ||
1337 sk->sk_protocol == IPPROTO_RAW)) {
1338 ipv6_local_rxpmtu(sk, fl6, mtu - headersize +
1339 sizeof(struct ipv6hdr));
1340 goto emsgsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341 }
1342
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001343 if (ip6_sk_ignore_df(sk))
1344 maxnonfragsize = sizeof(struct ipv6hdr) + IPV6_MAXPLEN;
1345 else
1346 maxnonfragsize = mtu;
1347
1348 if (cork->length + length > maxnonfragsize - headersize) {
1349emsgsize:
1350 ipv6_local_error(sk, EMSGSIZE, fl6,
1351 mtu - headersize +
1352 sizeof(struct ipv6hdr));
1353 return -EMSGSIZE;
1354 }
1355
1356 /* CHECKSUM_PARTIAL only with no extension headers and when
1357 * we are not going to fragment
1358 */
1359 if (transhdrlen && sk->sk_protocol == IPPROTO_UDP &&
1360 headersize == sizeof(struct ipv6hdr) &&
Vlad Yasevich2b89ed62017-01-29 22:52:53 -05001361 length <= mtu - headersize &&
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001362 !(flags & MSG_MORE) &&
Tom Herbertc8cd0982015-12-14 11:19:44 -08001363 rt->dst.dev->features & (NETIF_F_IPV6_CSUM | NETIF_F_HW_CSUM))
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001364 csummode = CHECKSUM_PARTIAL;
1365
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001366 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_RAW) {
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001367 sock_tx_timestamp(sk, sockc->tsflags, &tx_flags);
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001368 if (tx_flags & SKBTX_ANY_SW_TSTAMP &&
1369 sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)
1370 tskey = sk->sk_tskey++;
1371 }
Anders Berggrena693e692011-02-28 12:32:11 -08001372
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 /*
1374 * Let's try using as much space as possible.
1375 * Use MTU if total length of the message fits into the MTU.
1376 * Otherwise, we need to reserve fragment header and
1377 * fragment alignment (= 8-15 octects, in total).
1378 *
1379 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001380 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381 * the message.
1382 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001383 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 * at once if non-fragmentable extension headers
1385 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001386 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 */
1388
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001389 cork->length += length;
Zheng Lie4c5e132016-12-28 23:23:46 +08001390 if ((((length + fragheaderlen) > mtu) ||
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001391 (skb && skb_is_gso(skb))) &&
1392 (sk->sk_protocol == IPPROTO_UDP) &&
Alexey Kodanev4b3b45e2017-03-09 13:56:46 +03001393 (rt->dst.dev->features & NETIF_F_UFO) && !dst_xfrm(&rt->dst) &&
Michal Kubeček40ba3302016-01-11 07:50:30 +01001394 (sk->sk_type == SOCK_DGRAM) && !udp_get_no_check6_tx(sk)) {
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001395 err = ip6_ufo_append_data(sk, queue, getfrag, from, length,
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001396 hh_len, fragheaderlen, exthdrlen,
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001397 transhdrlen, mtu, flags, fl6);
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001398 if (err)
1399 goto error;
1400 return 0;
1401 }
1402
1403 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 goto alloc_new_skb;
1405
1406 while (length > 0) {
1407 /* Check if the remaining data fits into current packet. */
David S. Millerbdc712b2011-05-06 15:02:07 -07001408 copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 if (copy < length)
1410 copy = maxfraglen - skb->len;
1411
1412 if (copy <= 0) {
1413 char *data;
1414 unsigned int datalen;
1415 unsigned int fraglen;
1416 unsigned int fraggap;
1417 unsigned int alloclen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418alloc_new_skb:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 /* There's no room in the current skb */
Gao feng0c183372012-05-26 01:30:53 +00001420 if (skb)
1421 fraggap = skb->len - maxfraglen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 else
1423 fraggap = 0;
Gao feng0c183372012-05-26 01:30:53 +00001424 /* update mtu and maxfraglen if necessary */
Ian Morris63159f22015-03-29 14:00:04 +01001425 if (!skb || !skb_prev)
Gao feng0c183372012-05-26 01:30:53 +00001426 ip6_append_data_mtu(&mtu, &maxfraglen,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001427 fragheaderlen, skb, rt,
luciene367c2d2014-03-17 12:51:01 +08001428 orig_mtu);
Gao feng0c183372012-05-26 01:30:53 +00001429
1430 skb_prev = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431
1432 /*
1433 * If remaining data exceeds the mtu,
1434 * we know we need more fragment(s).
1435 */
1436 datalen = length + fraggap;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437
Gao feng0c183372012-05-26 01:30:53 +00001438 if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1439 datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440 if ((flags & MSG_MORE) &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001441 !(rt->dst.dev->features&NETIF_F_SG))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 alloclen = mtu;
1443 else
1444 alloclen = datalen + fragheaderlen;
1445
Steffen Klassert299b0762011-10-11 01:43:33 +00001446 alloclen += dst_exthdrlen;
1447
Gao feng0c183372012-05-26 01:30:53 +00001448 if (datalen != length + fraggap) {
1449 /*
1450 * this is not the last fragment, the trailer
1451 * space is regarded as data space.
1452 */
1453 datalen += rt->dst.trailer_len;
1454 }
1455
1456 alloclen += rt->dst.trailer_len;
1457 fraglen = datalen + fragheaderlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458
1459 /*
1460 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001461 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 * (without MSG_MORE) fits into the MTU.
1463 */
1464 alloclen += sizeof(struct frag_hdr);
1465
1466 if (transhdrlen) {
1467 skb = sock_alloc_send_skb(sk,
1468 alloclen + hh_len,
1469 (flags & MSG_DONTWAIT), &err);
1470 } else {
1471 skb = NULL;
1472 if (atomic_read(&sk->sk_wmem_alloc) <=
1473 2 * sk->sk_sndbuf)
1474 skb = sock_wmalloc(sk,
1475 alloclen + hh_len, 1,
1476 sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001477 if (unlikely(!skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 err = -ENOBUFS;
1479 }
Ian Morris63159f22015-03-29 14:00:04 +01001480 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 goto error;
1482 /*
1483 * Fill in the control structures
1484 */
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001485 skb->protocol = htons(ETH_P_IPV6);
Vlad Yasevich32dce962015-01-31 10:40:18 -05001486 skb->ip_summed = csummode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 skb->csum = 0;
Gao feng1f858512012-03-19 22:36:10 +00001488 /* reserve for fragmentation and ipsec header */
1489 skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
1490 dst_exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491
Willem de Bruijn11878b42014-07-14 17:55:06 -04001492 /* Only the initial fragment is time stamped */
1493 skb_shinfo(skb)->tx_flags = tx_flags;
1494 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001495 skb_shinfo(skb)->tskey = tskey;
1496 tskey = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001497
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498 /*
1499 * Find where to start putting bytes
1500 */
Gao feng1f858512012-03-19 22:36:10 +00001501 data = skb_put(skb, fraglen);
1502 skb_set_network_header(skb, exthdrlen);
1503 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001504 skb->transport_header = (skb->network_header +
1505 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 if (fraggap) {
1507 skb->csum = skb_copy_and_csum_bits(
1508 skb_prev, maxfraglen,
1509 data + transhdrlen, fraggap, 0);
1510 skb_prev->csum = csum_sub(skb_prev->csum,
1511 skb->csum);
1512 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001513 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 }
1515 copy = datalen - transhdrlen - fraggap;
Steffen Klassert299b0762011-10-11 01:43:33 +00001516
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517 if (copy < 0) {
1518 err = -EINVAL;
1519 kfree_skb(skb);
1520 goto error;
1521 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1522 err = -EFAULT;
1523 kfree_skb(skb);
1524 goto error;
1525 }
1526
1527 offset += copy;
1528 length -= datalen - fraggap;
1529 transhdrlen = 0;
1530 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001531 dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532
Julian Anastasov0dec8792017-02-06 23:14:16 +02001533 if ((flags & MSG_CONFIRM) && !skb_prev)
1534 skb_set_dst_pending_confirm(skb, 1);
1535
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 /*
1537 * Put the packet on the pending queue
1538 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001539 __skb_queue_tail(queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 continue;
1541 }
1542
1543 if (copy > length)
1544 copy = length;
1545
Changli Gaod8d1f302010-06-10 23:31:35 -07001546 if (!(rt->dst.dev->features&NETIF_F_SG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547 unsigned int off;
1548
1549 off = skb->len;
1550 if (getfrag(from, skb_put(skb, copy),
1551 offset, copy, off, skb) < 0) {
1552 __skb_trim(skb, off);
1553 err = -EFAULT;
1554 goto error;
1555 }
1556 } else {
1557 int i = skb_shinfo(skb)->nr_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558
Eric Dumazet5640f762012-09-23 23:04:42 +00001559 err = -ENOMEM;
1560 if (!sk_page_frag_refill(sk, pfrag))
1561 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562
Eric Dumazet5640f762012-09-23 23:04:42 +00001563 if (!skb_can_coalesce(skb, i, pfrag->page,
1564 pfrag->offset)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 err = -EMSGSIZE;
Eric Dumazet5640f762012-09-23 23:04:42 +00001566 if (i == MAX_SKB_FRAGS)
1567 goto error;
1568
1569 __skb_fill_page_desc(skb, i, pfrag->page,
1570 pfrag->offset, 0);
1571 skb_shinfo(skb)->nr_frags = ++i;
1572 get_page(pfrag->page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001574 copy = min_t(int, copy, pfrag->size - pfrag->offset);
Eric Dumazet9e903e02011-10-18 21:00:24 +00001575 if (getfrag(from,
Eric Dumazet5640f762012-09-23 23:04:42 +00001576 page_address(pfrag->page) + pfrag->offset,
1577 offset, copy, skb->len, skb) < 0)
1578 goto error_efault;
1579
1580 pfrag->offset += copy;
1581 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 skb->len += copy;
1583 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001584 skb->truesize += copy;
1585 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586 }
1587 offset += copy;
1588 length -= copy;
1589 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001590
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591 return 0;
Eric Dumazet5640f762012-09-23 23:04:42 +00001592
1593error_efault:
1594 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595error:
David S. Millerbdc712b2011-05-06 15:02:07 -07001596 cork->length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001597 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 return err;
1599}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001600
1601int ip6_append_data(struct sock *sk,
1602 int getfrag(void *from, char *to, int offset, int len,
1603 int odd, struct sk_buff *skb),
Wei Wang26879da2016-05-02 21:40:07 -07001604 void *from, int length, int transhdrlen,
1605 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
1606 struct rt6_info *rt, unsigned int flags,
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001607 const struct sockcm_cookie *sockc)
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001608{
1609 struct inet_sock *inet = inet_sk(sk);
1610 struct ipv6_pinfo *np = inet6_sk(sk);
1611 int exthdrlen;
1612 int err;
1613
1614 if (flags&MSG_PROBE)
1615 return 0;
1616 if (skb_queue_empty(&sk->sk_write_queue)) {
1617 /*
1618 * setup for corking
1619 */
Wei Wang26879da2016-05-02 21:40:07 -07001620 err = ip6_setup_cork(sk, &inet->cork, &np->cork,
1621 ipc6, rt, fl6);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001622 if (err)
1623 return err;
1624
Wei Wang26879da2016-05-02 21:40:07 -07001625 exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001626 length += exthdrlen;
1627 transhdrlen += exthdrlen;
1628 } else {
1629 fl6 = &inet->cork.fl.u.ip6;
1630 transhdrlen = 0;
1631 }
1632
1633 return __ip6_append_data(sk, fl6, &sk->sk_write_queue, &inet->cork.base,
1634 &np->cork, sk_page_frag(sk), getfrag,
Wei Wang26879da2016-05-02 21:40:07 -07001635 from, length, transhdrlen, flags, ipc6, sockc);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001636}
Chris Elstona495f832012-04-29 21:48:53 +00001637EXPORT_SYMBOL_GPL(ip6_append_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001639static void ip6_cork_release(struct inet_cork_full *cork,
1640 struct inet6_cork *v6_cork)
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001641{
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001642 if (v6_cork->opt) {
1643 kfree(v6_cork->opt->dst0opt);
1644 kfree(v6_cork->opt->dst1opt);
1645 kfree(v6_cork->opt->hopopt);
1646 kfree(v6_cork->opt->srcrt);
1647 kfree(v6_cork->opt);
1648 v6_cork->opt = NULL;
Herbert Xu0178b692009-02-05 15:15:50 -08001649 }
1650
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001651 if (cork->base.dst) {
1652 dst_release(cork->base.dst);
1653 cork->base.dst = NULL;
1654 cork->base.flags &= ~IPCORK_ALLFRAG;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001655 }
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001656 memset(&cork->fl, 0, sizeof(cork->fl));
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001657}
1658
Vlad Yasevich64223982015-01-31 10:40:15 -05001659struct sk_buff *__ip6_make_skb(struct sock *sk,
1660 struct sk_buff_head *queue,
1661 struct inet_cork_full *cork,
1662 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663{
1664 struct sk_buff *skb, *tmp_skb;
1665 struct sk_buff **tail_skb;
1666 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001668 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 struct ipv6hdr *hdr;
Vlad Yasevich64223982015-01-31 10:40:15 -05001670 struct ipv6_txoptions *opt = v6_cork->opt;
1671 struct rt6_info *rt = (struct rt6_info *)cork->base.dst;
1672 struct flowi6 *fl6 = &cork->fl.u.ip6;
David S. Miller4c9483b2011-03-12 16:22:43 -05001673 unsigned char proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674
Vlad Yasevich64223982015-01-31 10:40:15 -05001675 skb = __skb_dequeue(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001676 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 goto out;
1678 tail_skb = &(skb_shinfo(skb)->frag_list);
1679
1680 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001681 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001682 __skb_pull(skb, skb_network_offset(skb));
Vlad Yasevich64223982015-01-31 10:40:15 -05001683 while ((tmp_skb = __skb_dequeue(queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001684 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 *tail_skb = tmp_skb;
1686 tail_skb = &(tmp_skb->next);
1687 skb->len += tmp_skb->len;
1688 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 skb->truesize += tmp_skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690 tmp_skb->destructor = NULL;
1691 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 }
1693
Herbert Xu28a894532008-02-12 18:07:27 -08001694 /* Allow local fragmentation. */
WANG Cong60ff7462014-05-04 16:39:18 -07001695 skb->ignore_df = ip6_sk_ignore_df(sk);
Herbert Xu28a894532008-02-12 18:07:27 -08001696
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001697 *final_dst = fl6->daddr;
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001698 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 if (opt && opt->opt_flen)
1700 ipv6_push_frag_opts(skb, opt, &proto);
1701 if (opt && opt->opt_nflen)
David Lebrun613fa3c2016-11-08 14:59:20 +01001702 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst, &fl6->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001704 skb_push(skb, sizeof(struct ipv6hdr));
1705 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001706 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001707
Vlad Yasevich64223982015-01-31 10:40:15 -05001708 ip6_flow_hdr(hdr, v6_cork->tclass,
Tom Herbertcb1ce2e2014-07-01 21:33:10 -07001709 ip6_make_flowlabel(net, skb, fl6->flowlabel,
Tom Herbert67800f92015-07-31 16:52:11 -07001710 np->autoflowlabel, fl6));
Vlad Yasevich64223982015-01-31 10:40:15 -05001711 hdr->hop_limit = v6_cork->hop_limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 hdr->nexthdr = proto;
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001713 hdr->saddr = fl6->saddr;
1714 hdr->daddr = *final_dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715
Patrick McHardya2c20642006-01-08 22:37:26 -08001716 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001717 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001718
Changli Gaod8d1f302010-06-10 23:31:35 -07001719 skb_dst_set(skb, dst_clone(&rt->dst));
Neil Hormanedf391f2009-04-27 02:45:02 -07001720 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
David L Stevens14878f72007-09-16 16:52:35 -07001721 if (proto == IPPROTO_ICMPV6) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001722 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
David L Stevens14878f72007-09-16 16:52:35 -07001723
Hannes Frederic Sowa43a43b62014-03-31 20:14:10 +02001724 ICMP6MSGOUT_INC_STATS(net, idev, icmp6_hdr(skb)->icmp6_type);
1725 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001726 }
1727
Vlad Yasevich64223982015-01-31 10:40:15 -05001728 ip6_cork_release(cork, v6_cork);
1729out:
1730 return skb;
1731}
1732
1733int ip6_send_skb(struct sk_buff *skb)
1734{
1735 struct net *net = sock_net(skb->sk);
1736 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
1737 int err;
1738
Eric W. Biederman33224b12015-10-07 16:48:46 -05001739 err = ip6_local_out(net, skb->sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740 if (err) {
1741 if (err > 0)
Eric Dumazet6ce9e7b2009-09-02 18:05:33 -07001742 err = net_xmit_errno(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 if (err)
Vlad Yasevich64223982015-01-31 10:40:15 -05001744 IP6_INC_STATS(net, rt->rt6i_idev,
1745 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 }
1747
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 return err;
Vlad Yasevich64223982015-01-31 10:40:15 -05001749}
1750
1751int ip6_push_pending_frames(struct sock *sk)
1752{
1753 struct sk_buff *skb;
1754
1755 skb = ip6_finish_skb(sk);
1756 if (!skb)
1757 return 0;
1758
1759 return ip6_send_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760}
Chris Elstona495f832012-04-29 21:48:53 +00001761EXPORT_SYMBOL_GPL(ip6_push_pending_frames);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001763static void __ip6_flush_pending_frames(struct sock *sk,
Vlad Yasevich64223982015-01-31 10:40:15 -05001764 struct sk_buff_head *queue,
1765 struct inet_cork_full *cork,
1766 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 struct sk_buff *skb;
1769
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001770 while ((skb = __skb_dequeue_tail(queue)) != NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001771 if (skb_dst(skb))
1772 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001773 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774 kfree_skb(skb);
1775 }
1776
Vlad Yasevich64223982015-01-31 10:40:15 -05001777 ip6_cork_release(cork, v6_cork);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001779
1780void ip6_flush_pending_frames(struct sock *sk)
1781{
Vlad Yasevich64223982015-01-31 10:40:15 -05001782 __ip6_flush_pending_frames(sk, &sk->sk_write_queue,
1783 &inet_sk(sk)->cork, &inet6_sk(sk)->cork);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001784}
Chris Elstona495f832012-04-29 21:48:53 +00001785EXPORT_SYMBOL_GPL(ip6_flush_pending_frames);
Vlad Yasevich64223982015-01-31 10:40:15 -05001786
1787struct sk_buff *ip6_make_skb(struct sock *sk,
1788 int getfrag(void *from, char *to, int offset,
1789 int len, int odd, struct sk_buff *skb),
1790 void *from, int length, int transhdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001791 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
Vlad Yasevich64223982015-01-31 10:40:15 -05001792 struct rt6_info *rt, unsigned int flags,
Wei Wang26879da2016-05-02 21:40:07 -07001793 const struct sockcm_cookie *sockc)
Vlad Yasevich64223982015-01-31 10:40:15 -05001794{
1795 struct inet_cork_full cork;
1796 struct inet6_cork v6_cork;
1797 struct sk_buff_head queue;
Wei Wang26879da2016-05-02 21:40:07 -07001798 int exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0);
Vlad Yasevich64223982015-01-31 10:40:15 -05001799 int err;
1800
1801 if (flags & MSG_PROBE)
1802 return NULL;
1803
1804 __skb_queue_head_init(&queue);
1805
1806 cork.base.flags = 0;
1807 cork.base.addr = 0;
1808 cork.base.opt = NULL;
1809 v6_cork.opt = NULL;
Wei Wang26879da2016-05-02 21:40:07 -07001810 err = ip6_setup_cork(sk, &cork, &v6_cork, ipc6, rt, fl6);
Vlad Yasevich64223982015-01-31 10:40:15 -05001811 if (err)
1812 return ERR_PTR(err);
1813
Wei Wang26879da2016-05-02 21:40:07 -07001814 if (ipc6->dontfrag < 0)
1815 ipc6->dontfrag = inet6_sk(sk)->dontfrag;
Vlad Yasevich64223982015-01-31 10:40:15 -05001816
1817 err = __ip6_append_data(sk, fl6, &queue, &cork.base, &v6_cork,
1818 &current->task_frag, getfrag, from,
1819 length + exthdrlen, transhdrlen + exthdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001820 flags, ipc6, sockc);
Vlad Yasevich64223982015-01-31 10:40:15 -05001821 if (err) {
1822 __ip6_flush_pending_frames(sk, &queue, &cork, &v6_cork);
1823 return ERR_PTR(err);
1824 }
1825
1826 return __ip6_make_skb(sk, &queue, &cork, &v6_cork);
1827}