blob: 14d99fbf102e12b0952b97183689a65d219f4b3d [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);
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000123 ret = dst_neigh_output(dst, neigh, skb);
124 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
771 *prevhdr = NEXTHDR_FRAGMENT;
Herbert Xua7ae1992011-11-18 02:20:04 +0000772 troom = rt->dst.dev->needed_tailroom;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
774 /*
775 * Keep copying data until we run out.
776 */
Ian Morris67ba4152014-08-24 21:53:10 +0100777 while (left > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 len = left;
779 /* IF: it doesn't fit, use 'mtu' - the data space left */
780 if (len > mtu)
781 len = mtu;
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300782 /* IF: we are not sending up to and including the packet end
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 then align the next start on an eight byte boundary */
784 if (len < left) {
785 len &= ~7;
786 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
Joe Perchescbffccc2014-11-05 14:39:21 -0800788 /* Allocate buffer */
789 frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) +
790 hroom + troom, GFP_ATOMIC);
791 if (!frag) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000792 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900793 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 err = -ENOMEM;
795 goto fail;
796 }
797
798 /*
799 * Set up data on packet
800 */
801
802 ip6_copy_metadata(frag, skb);
Herbert Xua7ae1992011-11-18 02:20:04 +0000803 skb_reserve(frag, hroom);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700805 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300806 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700807 frag->transport_header = (frag->network_header + hlen +
808 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809
810 /*
811 * Charge the memory for the fragment to any owner
812 * it might possess
813 */
814 if (skb->sk)
815 skb_set_owner_w(frag, skb->sk);
816
817 /*
818 * Copy the packet header into the new buffer.
819 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300820 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821
822 /*
823 * Build fragment header.
824 */
825 fh->nexthdr = nexthdr;
826 fh->reserved = 0;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700827 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828
829 /*
830 * Copy a block of the IP datagram.
831 */
Himangi Saraogie3f0b862014-07-12 01:57:17 +0530832 BUG_ON(skb_copy_bits(skb, ptr, skb_transport_header(frag),
833 len));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834 left -= len;
835
836 fh->frag_off = htons(offset);
837 if (left > 0)
838 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700839 ipv6_hdr(frag)->payload_len = htons(frag->len -
840 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841
842 ptr += len;
843 offset += len;
844
845 /*
846 * Put this fragment into the sending queue.
847 */
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500848 err = output(net, sk, frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 if (err)
850 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700851
Eric Dumazetadf30902009-06-02 05:19:30 +0000852 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700853 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000855 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900856 IPSTATS_MIB_FRAGOKS);
Eric Dumazet808db802012-04-24 10:17:59 +0000857 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 return err;
859
Florian Westphal485fca62015-05-22 00:44:16 +0200860fail_toobig:
861 if (skb->sk && dst_allfrag(skb_dst(skb)))
862 sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK);
863
864 skb->dev = skb_dst(skb)->dev;
865 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
866 err = -EMSGSIZE;
867
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868fail:
Eric Dumazetadf30902009-06-02 05:19:30 +0000869 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900870 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900871 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 return err;
873}
874
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000875static inline int ip6_rt_check(const struct rt6key *rt_key,
876 const struct in6_addr *fl_addr,
877 const struct in6_addr *addr_cache)
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700878{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000879 return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
Ian Morris63159f22015-03-29 14:00:04 +0100880 (!addr_cache || !ipv6_addr_equal(fl_addr, addr_cache));
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700881}
882
Herbert Xu497c6152006-07-30 20:19:33 -0700883static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
884 struct dst_entry *dst,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000885 const struct flowi6 *fl6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886{
Herbert Xu497c6152006-07-30 20:19:33 -0700887 struct ipv6_pinfo *np = inet6_sk(sk);
Eric Dumazeta963a372013-06-26 04:15:07 -0700888 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889
Herbert Xu497c6152006-07-30 20:19:33 -0700890 if (!dst)
891 goto out;
892
Eric Dumazeta963a372013-06-26 04:15:07 -0700893 if (dst->ops->family != AF_INET6) {
894 dst_release(dst);
895 return NULL;
896 }
897
898 rt = (struct rt6_info *)dst;
Herbert Xu497c6152006-07-30 20:19:33 -0700899 /* Yes, checking route validity in not connected
900 * case is not very simple. Take into account,
901 * that we do not support routing by source, TOS,
Ian Morris67ba4152014-08-24 21:53:10 +0100902 * and MSG_DONTROUTE --ANK (980726)
Herbert Xu497c6152006-07-30 20:19:33 -0700903 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700904 * 1. ip6_rt_check(): If route was host route,
905 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700906 * If it is network route, we still may
907 * check its validity using saved pointer
908 * to the last used address: daddr_cache.
909 * We do not want to save whole address now,
910 * (because main consumer of this service
911 * is tcp, which has not this problem),
912 * so that the last trick works only on connected
913 * sockets.
914 * 2. oif also should be the same.
915 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500916 if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700917#ifdef CONFIG_IPV6_SUBTREES
David S. Miller4c9483b2011-03-12 16:22:43 -0500918 ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700919#endif
David Ahernca254492015-10-12 11:47:10 -0700920 (!(fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF) &&
921 (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex))) {
Herbert Xu497c6152006-07-30 20:19:33 -0700922 dst_release(dst);
923 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 }
925
Herbert Xu497c6152006-07-30 20:19:33 -0700926out:
927 return dst;
928}
929
Eric Dumazet3aef9342015-09-25 07:39:12 -0700930static int ip6_dst_lookup_tail(struct net *net, const struct sock *sk,
David S. Miller4c9483b2011-03-12 16:22:43 -0500931 struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700932{
David S. Miller69cce1d2011-07-17 23:09:49 -0700933#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
934 struct neighbour *n;
David S. Miller97cac082012-07-02 22:43:47 -0700935 struct rt6_info *rt;
David S. Miller69cce1d2011-07-17 23:09:49 -0700936#endif
937 int err;
Paolo Abeni6f21c962016-01-29 12:30:19 +0100938 int flags = 0;
Herbert Xu497c6152006-07-30 20:19:33 -0700939
Markus Stenberge16e8882015-05-05 13:36:59 +0300940 /* The correct way to handle this would be to do
941 * ip6_route_get_saddr, and then ip6_route_output; however,
942 * the route-specific preferred source forces the
943 * ip6_route_output call _before_ ip6_route_get_saddr.
944 *
945 * In source specific routing (no src=any default route),
946 * ip6_route_output will fail given src=any saddr, though, so
947 * that's why we try it again later.
948 */
949 if (ipv6_addr_any(&fl6->saddr) && (!*dst || !(*dst)->error)) {
950 struct rt6_info *rt;
951 bool had_dst = *dst != NULL;
952
953 if (!had_dst)
954 *dst = ip6_route_output(net, sk, fl6);
955 rt = (*dst)->error ? NULL : (struct rt6_info *)*dst;
956 err = ip6_route_get_saddr(net, rt, &fl6->daddr,
957 sk ? inet6_sk(sk)->srcprefs : 0,
958 &fl6->saddr);
959 if (err)
960 goto out_err_release;
961
962 /* If we had an erroneous initial result, pretend it
963 * never existed and let the SA-enabled version take
964 * over.
965 */
966 if (!had_dst && (*dst)->error) {
967 dst_release(*dst);
968 *dst = NULL;
969 }
Paolo Abeni6f21c962016-01-29 12:30:19 +0100970
971 if (fl6->flowi6_oif)
972 flags |= RT6_LOOKUP_F_IFACE;
Markus Stenberge16e8882015-05-05 13:36:59 +0300973 }
974
Ian Morris63159f22015-03-29 14:00:04 +0100975 if (!*dst)
Paolo Abeni6f21c962016-01-29 12:30:19 +0100976 *dst = ip6_route_output_flags(net, sk, fl6, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977
Ian Morrise5d08d72014-11-23 21:28:43 +0000978 err = (*dst)->error;
979 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 goto out_err_release;
981
Neil Horman95c385b2007-04-25 17:08:10 -0700982#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700983 /*
984 * Here if the dst entry we've looked up
985 * has a neighbour entry that is in the INCOMPLETE
986 * state and the src address from the flow is
987 * marked as OPTIMISTIC, we release the found
988 * dst entry and replace it instead with the
989 * dst entry of the nexthop router
990 */
Eric Dumazetc56bf6f2012-07-06 09:19:05 +0200991 rt = (struct rt6_info *) *dst;
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000992 rcu_read_lock_bh();
Martin KaFai Lau2647a9b2015-05-22 20:55:58 -0700993 n = __ipv6_neigh_lookup_noref(rt->dst.dev,
994 rt6_nexthop(rt, &fl6->daddr));
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000995 err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0;
996 rcu_read_unlock_bh();
997
998 if (err) {
Neil Hormane550dfb2008-09-09 13:51:35 -0700999 struct inet6_ifaddr *ifp;
David S. Miller4c9483b2011-03-12 16:22:43 -05001000 struct flowi6 fl_gw6;
Neil Hormane550dfb2008-09-09 13:51:35 -07001001 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -07001002
David S. Miller4c9483b2011-03-12 16:22:43 -05001003 ifp = ipv6_get_ifaddr(net, &fl6->saddr,
Neil Hormane550dfb2008-09-09 13:51:35 -07001004 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -07001005
Neil Hormane550dfb2008-09-09 13:51:35 -07001006 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
1007 if (ifp)
1008 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -07001009
Neil Hormane550dfb2008-09-09 13:51:35 -07001010 if (redirect) {
1011 /*
1012 * We need to get the dst entry for the
1013 * default router instead
1014 */
1015 dst_release(*dst);
David S. Miller4c9483b2011-03-12 16:22:43 -05001016 memcpy(&fl_gw6, fl6, sizeof(struct flowi6));
1017 memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr));
1018 *dst = ip6_route_output(net, sk, &fl_gw6);
Ian Morrise5d08d72014-11-23 21:28:43 +00001019 err = (*dst)->error;
1020 if (err)
Neil Hormane550dfb2008-09-09 13:51:35 -07001021 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -07001022 }
Neil Hormane550dfb2008-09-09 13:51:35 -07001023 }
Neil Horman95c385b2007-04-25 17:08:10 -07001024#endif
1025
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 return 0;
1027
1028out_err_release:
1029 dst_release(*dst);
1030 *dst = NULL;
David Ahern8a966fc2016-09-10 12:09:59 -07001031
David Ahern0d240e72016-06-16 16:24:25 -07001032 if (err == -ENETUNREACH)
1033 IP6_INC_STATS(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 return err;
1035}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001036
Herbert Xu497c6152006-07-30 20:19:33 -07001037/**
1038 * ip6_dst_lookup - perform route lookup on flow
1039 * @sk: socket which provides route info
1040 * @dst: pointer to dst_entry * for result
David S. Miller4c9483b2011-03-12 16:22:43 -05001041 * @fl6: flow to lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001042 *
1043 * This function performs a route lookup on the given flow.
1044 *
1045 * It returns zero on success, or a standard errno code on error.
1046 */
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001047int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1048 struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -07001049{
1050 *dst = NULL;
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001051 return ip6_dst_lookup_tail(net, sk, dst, fl6);
Herbert Xu497c6152006-07-30 20:19:33 -07001052}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -08001053EXPORT_SYMBOL_GPL(ip6_dst_lookup);
1054
Herbert Xu497c6152006-07-30 20:19:33 -07001055/**
David S. Miller68d0c6d2011-03-01 13:19:07 -08001056 * ip6_dst_lookup_flow - perform route lookup on flow with ipsec
1057 * @sk: socket which provides route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001058 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001059 * @final_dst: final destination address for ipsec lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001060 *
1061 * This function performs a route lookup on the given flow.
1062 *
1063 * It returns a valid dst pointer on success, or a pointer encoded
1064 * error code.
1065 */
Eric Dumazet3aef9342015-09-25 07:39:12 -07001066struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001067 const struct in6_addr *final_dst)
David S. Miller68d0c6d2011-03-01 13:19:07 -08001068{
1069 struct dst_entry *dst = NULL;
1070 int err;
1071
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001072 err = ip6_dst_lookup_tail(sock_net(sk), sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001073 if (err)
1074 return ERR_PTR(err);
1075 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001076 fl6->daddr = *final_dst;
David S. Miller2774c132011-03-01 14:59:04 -08001077
Steffen Klassertf92ee612014-09-16 10:08:40 +02001078 return xfrm_lookup_route(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001079}
1080EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow);
1081
1082/**
1083 * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow
1084 * @sk: socket which provides the dst cache and route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001085 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001086 * @final_dst: final destination address for ipsec lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001087 *
1088 * This function performs a route lookup on the given flow with the
1089 * possibility of using the cached route in the socket if it is valid.
1090 * It will take the socket dst lock when operating on the dst cache.
1091 * As a result, this function can only be used in process context.
1092 *
David S. Miller68d0c6d2011-03-01 13:19:07 -08001093 * It returns a valid dst pointer on success, or a pointer encoded
1094 * error code.
Herbert Xu497c6152006-07-30 20:19:33 -07001095 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001096struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001097 const struct in6_addr *final_dst)
Herbert Xu497c6152006-07-30 20:19:33 -07001098{
David S. Miller68d0c6d2011-03-01 13:19:07 -08001099 struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
Herbert Xu497c6152006-07-30 20:19:33 -07001100
David S. Miller4c9483b2011-03-12 16:22:43 -05001101 dst = ip6_sk_dst_check(sk, dst, fl6);
Jakub Sitnicki00bc0ef2016-06-08 15:13:34 +02001102 if (!dst)
1103 dst = ip6_dst_lookup_flow(sk, fl6, final_dst);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001104
Jakub Sitnicki00bc0ef2016-06-08 15:13:34 +02001105 return dst;
Herbert Xu497c6152006-07-30 20:19:33 -07001106}
David S. Miller68d0c6d2011-03-01 13:19:07 -08001107EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
Herbert Xu497c6152006-07-30 20:19:33 -07001108
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001109static inline int ip6_ufo_append_data(struct sock *sk,
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001110 struct sk_buff_head *queue,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001111 int getfrag(void *from, char *to, int offset, int len,
1112 int odd, struct sk_buff *skb),
1113 void *from, int length, int hh_len, int fragheaderlen,
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001114 int exthdrlen, int transhdrlen, int mtu,
1115 unsigned int flags, const struct flowi6 *fl6)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001116
1117{
1118 struct sk_buff *skb;
1119 int err;
1120
1121 /* There is support for UDP large send offload by network
1122 * device, so create one single skb packet containing complete
1123 * udp datagram
1124 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001125 skb = skb_peek_tail(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001126 if (!skb) {
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001127 skb = sock_alloc_send_skb(sk,
1128 hh_len + fragheaderlen + transhdrlen + 20,
1129 (flags & MSG_DONTWAIT), &err);
Ian Morris63159f22015-03-29 14:00:04 +01001130 if (!skb)
Zheng Yan504744e2011-10-28 00:26:00 -04001131 return err;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001132
1133 /* reserve space for Hardware header */
1134 skb_reserve(skb, hh_len);
1135
1136 /* create space for UDP/IP header */
Ian Morris67ba4152014-08-24 21:53:10 +01001137 skb_put(skb, fragheaderlen + transhdrlen);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001138
1139 /* initialize network header pointer */
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001140 skb_set_network_header(skb, exthdrlen);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001141
1142 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001143 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001144
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001145 skb->protocol = htons(ETH_P_IPV6);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001146 skb->csum = 0;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001147
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001148 __skb_queue_tail(queue, skb);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001149 } else if (skb_is_gso(skb)) {
1150 goto append;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001151 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001152
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001153 skb->ip_summed = CHECKSUM_PARTIAL;
1154 /* Specify the length of each IPv6 datagram fragment.
1155 * It has to be a multiple of 8.
1156 */
1157 skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
1158 sizeof(struct frag_hdr)) & ~7;
1159 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001160 skb_shinfo(skb)->ip6_frag_id = ipv6_select_ident(sock_net(sk),
1161 &fl6->daddr,
1162 &fl6->saddr);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001163
1164append:
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001165 return skb_append_datato_frags(sk, skb, getfrag, from,
1166 (length - transhdrlen));
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001167}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168
Herbert Xu0178b692009-02-05 15:15:50 -08001169static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
1170 gfp_t gfp)
1171{
1172 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1173}
1174
1175static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
1176 gfp_t gfp)
1177{
1178 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1179}
1180
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001181static void ip6_append_data_mtu(unsigned int *mtu,
Gao feng0c183372012-05-26 01:30:53 +00001182 int *maxfraglen,
1183 unsigned int fragheaderlen,
1184 struct sk_buff *skb,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001185 struct rt6_info *rt,
luciene367c2d2014-03-17 12:51:01 +08001186 unsigned int orig_mtu)
Gao feng0c183372012-05-26 01:30:53 +00001187{
1188 if (!(rt->dst.flags & DST_XFRM_TUNNEL)) {
Ian Morris63159f22015-03-29 14:00:04 +01001189 if (!skb) {
Gao feng0c183372012-05-26 01:30:53 +00001190 /* first fragment, reserve header_len */
luciene367c2d2014-03-17 12:51:01 +08001191 *mtu = orig_mtu - rt->dst.header_len;
Gao feng0c183372012-05-26 01:30:53 +00001192
1193 } else {
1194 /*
1195 * this fragment is not first, the headers
1196 * space is regarded as data space.
1197 */
luciene367c2d2014-03-17 12:51:01 +08001198 *mtu = orig_mtu;
Gao feng0c183372012-05-26 01:30:53 +00001199 }
1200 *maxfraglen = ((*mtu - fragheaderlen) & ~7)
1201 + fragheaderlen - sizeof(struct frag_hdr);
1202 }
1203}
1204
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001205static int ip6_setup_cork(struct sock *sk, struct inet_cork_full *cork,
Wei Wang26879da2016-05-02 21:40:07 -07001206 struct inet6_cork *v6_cork, struct ipcm6_cookie *ipc6,
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001207 struct rt6_info *rt, struct flowi6 *fl6)
1208{
1209 struct ipv6_pinfo *np = inet6_sk(sk);
1210 unsigned int mtu;
Wei Wang26879da2016-05-02 21:40:07 -07001211 struct ipv6_txoptions *opt = ipc6->opt;
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001212
1213 /*
1214 * setup for corking
1215 */
1216 if (opt) {
1217 if (WARN_ON(v6_cork->opt))
1218 return -EINVAL;
1219
1220 v6_cork->opt = kzalloc(opt->tot_len, sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001221 if (unlikely(!v6_cork->opt))
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001222 return -ENOBUFS;
1223
1224 v6_cork->opt->tot_len = opt->tot_len;
1225 v6_cork->opt->opt_flen = opt->opt_flen;
1226 v6_cork->opt->opt_nflen = opt->opt_nflen;
1227
1228 v6_cork->opt->dst0opt = ip6_opt_dup(opt->dst0opt,
1229 sk->sk_allocation);
1230 if (opt->dst0opt && !v6_cork->opt->dst0opt)
1231 return -ENOBUFS;
1232
1233 v6_cork->opt->dst1opt = ip6_opt_dup(opt->dst1opt,
1234 sk->sk_allocation);
1235 if (opt->dst1opt && !v6_cork->opt->dst1opt)
1236 return -ENOBUFS;
1237
1238 v6_cork->opt->hopopt = ip6_opt_dup(opt->hopopt,
1239 sk->sk_allocation);
1240 if (opt->hopopt && !v6_cork->opt->hopopt)
1241 return -ENOBUFS;
1242
1243 v6_cork->opt->srcrt = ip6_rthdr_dup(opt->srcrt,
1244 sk->sk_allocation);
1245 if (opt->srcrt && !v6_cork->opt->srcrt)
1246 return -ENOBUFS;
1247
1248 /* need source address above miyazawa*/
1249 }
1250 dst_hold(&rt->dst);
1251 cork->base.dst = &rt->dst;
1252 cork->fl.u.ip6 = *fl6;
Wei Wang26879da2016-05-02 21:40:07 -07001253 v6_cork->hop_limit = ipc6->hlimit;
1254 v6_cork->tclass = ipc6->tclass;
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001255 if (rt->dst.flags & DST_XFRM_TUNNEL)
1256 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
1257 rt->dst.dev->mtu : dst_mtu(&rt->dst);
1258 else
1259 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
1260 rt->dst.dev->mtu : dst_mtu(rt->dst.path);
1261 if (np->frag_size < mtu) {
1262 if (np->frag_size)
1263 mtu = np->frag_size;
1264 }
1265 cork->base.fragsize = mtu;
1266 if (dst_allfrag(rt->dst.path))
1267 cork->base.flags |= IPCORK_ALLFRAG;
1268 cork->base.length = 0;
1269
1270 return 0;
1271}
1272
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001273static int __ip6_append_data(struct sock *sk,
1274 struct flowi6 *fl6,
1275 struct sk_buff_head *queue,
1276 struct inet_cork *cork,
1277 struct inet6_cork *v6_cork,
1278 struct page_frag *pfrag,
1279 int getfrag(void *from, char *to, int offset,
1280 int len, int odd, struct sk_buff *skb),
1281 void *from, int length, int transhdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001282 unsigned int flags, struct ipcm6_cookie *ipc6,
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001283 const struct sockcm_cookie *sockc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284{
Gao feng0c183372012-05-26 01:30:53 +00001285 struct sk_buff *skb, *skb_prev = NULL;
luciene367c2d2014-03-17 12:51:01 +08001286 unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001287 int exthdrlen = 0;
1288 int dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 int hh_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290 int copy;
1291 int err;
1292 int offset = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001293 __u8 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001294 u32 tskey = 0;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001295 struct rt6_info *rt = (struct rt6_info *)cork->dst;
1296 struct ipv6_txoptions *opt = v6_cork->opt;
Vlad Yasevich32dce962015-01-31 10:40:18 -05001297 int csummode = CHECKSUM_NONE;
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001298 unsigned int maxnonfragsize, headersize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001300 skb = skb_peek_tail(queue);
1301 if (!skb) {
1302 exthdrlen = opt ? opt->opt_flen : 0;
Romain KUNTZ7efdba5b2013-01-16 12:47:40 +00001303 dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 }
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001305
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001306 mtu = cork->fragsize;
luciene367c2d2014-03-17 12:51:01 +08001307 orig_mtu = mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308
Changli Gaod8d1f302010-06-10 23:31:35 -07001309 hh_len = LL_RESERVED_SPACE(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001311 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001312 (opt ? opt->opt_nflen : 0);
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001313 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen -
1314 sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001316 headersize = sizeof(struct ipv6hdr) +
1317 (opt ? opt->opt_flen + opt->opt_nflen : 0) +
1318 (dst_allfrag(&rt->dst) ?
1319 sizeof(struct frag_hdr) : 0) +
1320 rt->rt6i_nfheader_len;
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001321
Wei Wang26879da2016-05-02 21:40:07 -07001322 if (cork->length + length > mtu - headersize && ipc6->dontfrag &&
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001323 (sk->sk_protocol == IPPROTO_UDP ||
1324 sk->sk_protocol == IPPROTO_RAW)) {
1325 ipv6_local_rxpmtu(sk, fl6, mtu - headersize +
1326 sizeof(struct ipv6hdr));
1327 goto emsgsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 }
1329
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001330 if (ip6_sk_ignore_df(sk))
1331 maxnonfragsize = sizeof(struct ipv6hdr) + IPV6_MAXPLEN;
1332 else
1333 maxnonfragsize = mtu;
1334
1335 if (cork->length + length > maxnonfragsize - headersize) {
1336emsgsize:
1337 ipv6_local_error(sk, EMSGSIZE, fl6,
1338 mtu - headersize +
1339 sizeof(struct ipv6hdr));
1340 return -EMSGSIZE;
1341 }
1342
1343 /* CHECKSUM_PARTIAL only with no extension headers and when
1344 * we are not going to fragment
1345 */
1346 if (transhdrlen && sk->sk_protocol == IPPROTO_UDP &&
1347 headersize == sizeof(struct ipv6hdr) &&
Vlad Yasevich2b89ed62017-01-29 22:52:53 -05001348 length <= mtu - headersize &&
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001349 !(flags & MSG_MORE) &&
Tom Herbertc8cd0982015-12-14 11:19:44 -08001350 rt->dst.dev->features & (NETIF_F_IPV6_CSUM | NETIF_F_HW_CSUM))
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001351 csummode = CHECKSUM_PARTIAL;
1352
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001353 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_RAW) {
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001354 sock_tx_timestamp(sk, sockc->tsflags, &tx_flags);
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001355 if (tx_flags & SKBTX_ANY_SW_TSTAMP &&
1356 sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)
1357 tskey = sk->sk_tskey++;
1358 }
Anders Berggrena693e692011-02-28 12:32:11 -08001359
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 /*
1361 * Let's try using as much space as possible.
1362 * Use MTU if total length of the message fits into the MTU.
1363 * Otherwise, we need to reserve fragment header and
1364 * fragment alignment (= 8-15 octects, in total).
1365 *
1366 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001367 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 * the message.
1369 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001370 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 * at once if non-fragmentable extension headers
1372 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001373 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 */
1375
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001376 cork->length += length;
Zheng Lie4c5e132016-12-28 23:23:46 +08001377 if ((((length + fragheaderlen) > mtu) ||
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001378 (skb && skb_is_gso(skb))) &&
1379 (sk->sk_protocol == IPPROTO_UDP) &&
Jakub Sitnickif89c56c2016-10-26 11:21:14 +02001380 (rt->dst.dev->features & NETIF_F_UFO) && !rt->dst.header_len &&
Michal Kubeček40ba3302016-01-11 07:50:30 +01001381 (sk->sk_type == SOCK_DGRAM) && !udp_get_no_check6_tx(sk)) {
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001382 err = ip6_ufo_append_data(sk, queue, getfrag, from, length,
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001383 hh_len, fragheaderlen, exthdrlen,
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001384 transhdrlen, mtu, flags, fl6);
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001385 if (err)
1386 goto error;
1387 return 0;
1388 }
1389
1390 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 goto alloc_new_skb;
1392
1393 while (length > 0) {
1394 /* Check if the remaining data fits into current packet. */
David S. Millerbdc712b2011-05-06 15:02:07 -07001395 copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 if (copy < length)
1397 copy = maxfraglen - skb->len;
1398
1399 if (copy <= 0) {
1400 char *data;
1401 unsigned int datalen;
1402 unsigned int fraglen;
1403 unsigned int fraggap;
1404 unsigned int alloclen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405alloc_new_skb:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 /* There's no room in the current skb */
Gao feng0c183372012-05-26 01:30:53 +00001407 if (skb)
1408 fraggap = skb->len - maxfraglen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 else
1410 fraggap = 0;
Gao feng0c183372012-05-26 01:30:53 +00001411 /* update mtu and maxfraglen if necessary */
Ian Morris63159f22015-03-29 14:00:04 +01001412 if (!skb || !skb_prev)
Gao feng0c183372012-05-26 01:30:53 +00001413 ip6_append_data_mtu(&mtu, &maxfraglen,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001414 fragheaderlen, skb, rt,
luciene367c2d2014-03-17 12:51:01 +08001415 orig_mtu);
Gao feng0c183372012-05-26 01:30:53 +00001416
1417 skb_prev = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418
1419 /*
1420 * If remaining data exceeds the mtu,
1421 * we know we need more fragment(s).
1422 */
1423 datalen = length + fraggap;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424
Gao feng0c183372012-05-26 01:30:53 +00001425 if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1426 datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427 if ((flags & MSG_MORE) &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001428 !(rt->dst.dev->features&NETIF_F_SG))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 alloclen = mtu;
1430 else
1431 alloclen = datalen + fragheaderlen;
1432
Steffen Klassert299b0762011-10-11 01:43:33 +00001433 alloclen += dst_exthdrlen;
1434
Gao feng0c183372012-05-26 01:30:53 +00001435 if (datalen != length + fraggap) {
1436 /*
1437 * this is not the last fragment, the trailer
1438 * space is regarded as data space.
1439 */
1440 datalen += rt->dst.trailer_len;
1441 }
1442
1443 alloclen += rt->dst.trailer_len;
1444 fraglen = datalen + fragheaderlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445
1446 /*
1447 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001448 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 * (without MSG_MORE) fits into the MTU.
1450 */
1451 alloclen += sizeof(struct frag_hdr);
1452
1453 if (transhdrlen) {
1454 skb = sock_alloc_send_skb(sk,
1455 alloclen + hh_len,
1456 (flags & MSG_DONTWAIT), &err);
1457 } else {
1458 skb = NULL;
1459 if (atomic_read(&sk->sk_wmem_alloc) <=
1460 2 * sk->sk_sndbuf)
1461 skb = sock_wmalloc(sk,
1462 alloclen + hh_len, 1,
1463 sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001464 if (unlikely(!skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465 err = -ENOBUFS;
1466 }
Ian Morris63159f22015-03-29 14:00:04 +01001467 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 goto error;
1469 /*
1470 * Fill in the control structures
1471 */
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001472 skb->protocol = htons(ETH_P_IPV6);
Vlad Yasevich32dce962015-01-31 10:40:18 -05001473 skb->ip_summed = csummode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 skb->csum = 0;
Gao feng1f858512012-03-19 22:36:10 +00001475 /* reserve for fragmentation and ipsec header */
1476 skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
1477 dst_exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478
Willem de Bruijn11878b42014-07-14 17:55:06 -04001479 /* Only the initial fragment is time stamped */
1480 skb_shinfo(skb)->tx_flags = tx_flags;
1481 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001482 skb_shinfo(skb)->tskey = tskey;
1483 tskey = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001484
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 /*
1486 * Find where to start putting bytes
1487 */
Gao feng1f858512012-03-19 22:36:10 +00001488 data = skb_put(skb, fraglen);
1489 skb_set_network_header(skb, exthdrlen);
1490 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001491 skb->transport_header = (skb->network_header +
1492 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 if (fraggap) {
1494 skb->csum = skb_copy_and_csum_bits(
1495 skb_prev, maxfraglen,
1496 data + transhdrlen, fraggap, 0);
1497 skb_prev->csum = csum_sub(skb_prev->csum,
1498 skb->csum);
1499 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001500 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 }
1502 copy = datalen - transhdrlen - fraggap;
Steffen Klassert299b0762011-10-11 01:43:33 +00001503
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 if (copy < 0) {
1505 err = -EINVAL;
1506 kfree_skb(skb);
1507 goto error;
1508 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1509 err = -EFAULT;
1510 kfree_skb(skb);
1511 goto error;
1512 }
1513
1514 offset += copy;
1515 length -= datalen - fraggap;
1516 transhdrlen = 0;
1517 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001518 dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519
1520 /*
1521 * Put the packet on the pending queue
1522 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001523 __skb_queue_tail(queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 continue;
1525 }
1526
1527 if (copy > length)
1528 copy = length;
1529
Changli Gaod8d1f302010-06-10 23:31:35 -07001530 if (!(rt->dst.dev->features&NETIF_F_SG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 unsigned int off;
1532
1533 off = skb->len;
1534 if (getfrag(from, skb_put(skb, copy),
1535 offset, copy, off, skb) < 0) {
1536 __skb_trim(skb, off);
1537 err = -EFAULT;
1538 goto error;
1539 }
1540 } else {
1541 int i = skb_shinfo(skb)->nr_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542
Eric Dumazet5640f762012-09-23 23:04:42 +00001543 err = -ENOMEM;
1544 if (!sk_page_frag_refill(sk, pfrag))
1545 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546
Eric Dumazet5640f762012-09-23 23:04:42 +00001547 if (!skb_can_coalesce(skb, i, pfrag->page,
1548 pfrag->offset)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 err = -EMSGSIZE;
Eric Dumazet5640f762012-09-23 23:04:42 +00001550 if (i == MAX_SKB_FRAGS)
1551 goto error;
1552
1553 __skb_fill_page_desc(skb, i, pfrag->page,
1554 pfrag->offset, 0);
1555 skb_shinfo(skb)->nr_frags = ++i;
1556 get_page(pfrag->page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001558 copy = min_t(int, copy, pfrag->size - pfrag->offset);
Eric Dumazet9e903e02011-10-18 21:00:24 +00001559 if (getfrag(from,
Eric Dumazet5640f762012-09-23 23:04:42 +00001560 page_address(pfrag->page) + pfrag->offset,
1561 offset, copy, skb->len, skb) < 0)
1562 goto error_efault;
1563
1564 pfrag->offset += copy;
1565 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566 skb->len += copy;
1567 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001568 skb->truesize += copy;
1569 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570 }
1571 offset += copy;
1572 length -= copy;
1573 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001574
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575 return 0;
Eric Dumazet5640f762012-09-23 23:04:42 +00001576
1577error_efault:
1578 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579error:
David S. Millerbdc712b2011-05-06 15:02:07 -07001580 cork->length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001581 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 return err;
1583}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001584
1585int ip6_append_data(struct sock *sk,
1586 int getfrag(void *from, char *to, int offset, int len,
1587 int odd, struct sk_buff *skb),
Wei Wang26879da2016-05-02 21:40:07 -07001588 void *from, int length, int transhdrlen,
1589 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
1590 struct rt6_info *rt, unsigned int flags,
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001591 const struct sockcm_cookie *sockc)
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001592{
1593 struct inet_sock *inet = inet_sk(sk);
1594 struct ipv6_pinfo *np = inet6_sk(sk);
1595 int exthdrlen;
1596 int err;
1597
1598 if (flags&MSG_PROBE)
1599 return 0;
1600 if (skb_queue_empty(&sk->sk_write_queue)) {
1601 /*
1602 * setup for corking
1603 */
Wei Wang26879da2016-05-02 21:40:07 -07001604 err = ip6_setup_cork(sk, &inet->cork, &np->cork,
1605 ipc6, rt, fl6);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001606 if (err)
1607 return err;
1608
Wei Wang26879da2016-05-02 21:40:07 -07001609 exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001610 length += exthdrlen;
1611 transhdrlen += exthdrlen;
1612 } else {
1613 fl6 = &inet->cork.fl.u.ip6;
1614 transhdrlen = 0;
1615 }
1616
1617 return __ip6_append_data(sk, fl6, &sk->sk_write_queue, &inet->cork.base,
1618 &np->cork, sk_page_frag(sk), getfrag,
Wei Wang26879da2016-05-02 21:40:07 -07001619 from, length, transhdrlen, flags, ipc6, sockc);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001620}
Chris Elstona495f832012-04-29 21:48:53 +00001621EXPORT_SYMBOL_GPL(ip6_append_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001623static void ip6_cork_release(struct inet_cork_full *cork,
1624 struct inet6_cork *v6_cork)
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001625{
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001626 if (v6_cork->opt) {
1627 kfree(v6_cork->opt->dst0opt);
1628 kfree(v6_cork->opt->dst1opt);
1629 kfree(v6_cork->opt->hopopt);
1630 kfree(v6_cork->opt->srcrt);
1631 kfree(v6_cork->opt);
1632 v6_cork->opt = NULL;
Herbert Xu0178b692009-02-05 15:15:50 -08001633 }
1634
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001635 if (cork->base.dst) {
1636 dst_release(cork->base.dst);
1637 cork->base.dst = NULL;
1638 cork->base.flags &= ~IPCORK_ALLFRAG;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001639 }
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001640 memset(&cork->fl, 0, sizeof(cork->fl));
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001641}
1642
Vlad Yasevich64223982015-01-31 10:40:15 -05001643struct sk_buff *__ip6_make_skb(struct sock *sk,
1644 struct sk_buff_head *queue,
1645 struct inet_cork_full *cork,
1646 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647{
1648 struct sk_buff *skb, *tmp_skb;
1649 struct sk_buff **tail_skb;
1650 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001652 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 struct ipv6hdr *hdr;
Vlad Yasevich64223982015-01-31 10:40:15 -05001654 struct ipv6_txoptions *opt = v6_cork->opt;
1655 struct rt6_info *rt = (struct rt6_info *)cork->base.dst;
1656 struct flowi6 *fl6 = &cork->fl.u.ip6;
David S. Miller4c9483b2011-03-12 16:22:43 -05001657 unsigned char proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658
Vlad Yasevich64223982015-01-31 10:40:15 -05001659 skb = __skb_dequeue(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001660 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 goto out;
1662 tail_skb = &(skb_shinfo(skb)->frag_list);
1663
1664 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001665 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001666 __skb_pull(skb, skb_network_offset(skb));
Vlad Yasevich64223982015-01-31 10:40:15 -05001667 while ((tmp_skb = __skb_dequeue(queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001668 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 *tail_skb = tmp_skb;
1670 tail_skb = &(tmp_skb->next);
1671 skb->len += tmp_skb->len;
1672 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 skb->truesize += tmp_skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 tmp_skb->destructor = NULL;
1675 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676 }
1677
Herbert Xu28a894532008-02-12 18:07:27 -08001678 /* Allow local fragmentation. */
WANG Cong60ff7462014-05-04 16:39:18 -07001679 skb->ignore_df = ip6_sk_ignore_df(sk);
Herbert Xu28a894532008-02-12 18:07:27 -08001680
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001681 *final_dst = fl6->daddr;
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001682 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 if (opt && opt->opt_flen)
1684 ipv6_push_frag_opts(skb, opt, &proto);
1685 if (opt && opt->opt_nflen)
David Lebrun613fa3c2016-11-08 14:59:20 +01001686 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst, &fl6->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001688 skb_push(skb, sizeof(struct ipv6hdr));
1689 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001690 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001691
Vlad Yasevich64223982015-01-31 10:40:15 -05001692 ip6_flow_hdr(hdr, v6_cork->tclass,
Tom Herbertcb1ce2e2014-07-01 21:33:10 -07001693 ip6_make_flowlabel(net, skb, fl6->flowlabel,
Tom Herbert67800f92015-07-31 16:52:11 -07001694 np->autoflowlabel, fl6));
Vlad Yasevich64223982015-01-31 10:40:15 -05001695 hdr->hop_limit = v6_cork->hop_limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696 hdr->nexthdr = proto;
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001697 hdr->saddr = fl6->saddr;
1698 hdr->daddr = *final_dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699
Patrick McHardya2c20642006-01-08 22:37:26 -08001700 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001701 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001702
Changli Gaod8d1f302010-06-10 23:31:35 -07001703 skb_dst_set(skb, dst_clone(&rt->dst));
Neil Hormanedf391f2009-04-27 02:45:02 -07001704 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
David L Stevens14878f72007-09-16 16:52:35 -07001705 if (proto == IPPROTO_ICMPV6) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001706 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
David L Stevens14878f72007-09-16 16:52:35 -07001707
Hannes Frederic Sowa43a43b62014-03-31 20:14:10 +02001708 ICMP6MSGOUT_INC_STATS(net, idev, icmp6_hdr(skb)->icmp6_type);
1709 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001710 }
1711
Vlad Yasevich64223982015-01-31 10:40:15 -05001712 ip6_cork_release(cork, v6_cork);
1713out:
1714 return skb;
1715}
1716
1717int ip6_send_skb(struct sk_buff *skb)
1718{
1719 struct net *net = sock_net(skb->sk);
1720 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
1721 int err;
1722
Eric W. Biederman33224b12015-10-07 16:48:46 -05001723 err = ip6_local_out(net, skb->sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 if (err) {
1725 if (err > 0)
Eric Dumazet6ce9e7b2009-09-02 18:05:33 -07001726 err = net_xmit_errno(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 if (err)
Vlad Yasevich64223982015-01-31 10:40:15 -05001728 IP6_INC_STATS(net, rt->rt6i_idev,
1729 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 }
1731
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 return err;
Vlad Yasevich64223982015-01-31 10:40:15 -05001733}
1734
1735int ip6_push_pending_frames(struct sock *sk)
1736{
1737 struct sk_buff *skb;
1738
1739 skb = ip6_finish_skb(sk);
1740 if (!skb)
1741 return 0;
1742
1743 return ip6_send_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744}
Chris Elstona495f832012-04-29 21:48:53 +00001745EXPORT_SYMBOL_GPL(ip6_push_pending_frames);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001747static void __ip6_flush_pending_frames(struct sock *sk,
Vlad Yasevich64223982015-01-31 10:40:15 -05001748 struct sk_buff_head *queue,
1749 struct inet_cork_full *cork,
1750 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 struct sk_buff *skb;
1753
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001754 while ((skb = __skb_dequeue_tail(queue)) != NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001755 if (skb_dst(skb))
1756 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001757 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758 kfree_skb(skb);
1759 }
1760
Vlad Yasevich64223982015-01-31 10:40:15 -05001761 ip6_cork_release(cork, v6_cork);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001763
1764void ip6_flush_pending_frames(struct sock *sk)
1765{
Vlad Yasevich64223982015-01-31 10:40:15 -05001766 __ip6_flush_pending_frames(sk, &sk->sk_write_queue,
1767 &inet_sk(sk)->cork, &inet6_sk(sk)->cork);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001768}
Chris Elstona495f832012-04-29 21:48:53 +00001769EXPORT_SYMBOL_GPL(ip6_flush_pending_frames);
Vlad Yasevich64223982015-01-31 10:40:15 -05001770
1771struct sk_buff *ip6_make_skb(struct sock *sk,
1772 int getfrag(void *from, char *to, int offset,
1773 int len, int odd, struct sk_buff *skb),
1774 void *from, int length, int transhdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001775 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
Vlad Yasevich64223982015-01-31 10:40:15 -05001776 struct rt6_info *rt, unsigned int flags,
Wei Wang26879da2016-05-02 21:40:07 -07001777 const struct sockcm_cookie *sockc)
Vlad Yasevich64223982015-01-31 10:40:15 -05001778{
1779 struct inet_cork_full cork;
1780 struct inet6_cork v6_cork;
1781 struct sk_buff_head queue;
Wei Wang26879da2016-05-02 21:40:07 -07001782 int exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0);
Vlad Yasevich64223982015-01-31 10:40:15 -05001783 int err;
1784
1785 if (flags & MSG_PROBE)
1786 return NULL;
1787
1788 __skb_queue_head_init(&queue);
1789
1790 cork.base.flags = 0;
1791 cork.base.addr = 0;
1792 cork.base.opt = NULL;
1793 v6_cork.opt = NULL;
Wei Wang26879da2016-05-02 21:40:07 -07001794 err = ip6_setup_cork(sk, &cork, &v6_cork, ipc6, rt, fl6);
Vlad Yasevich64223982015-01-31 10:40:15 -05001795 if (err)
1796 return ERR_PTR(err);
1797
Wei Wang26879da2016-05-02 21:40:07 -07001798 if (ipc6->dontfrag < 0)
1799 ipc6->dontfrag = inet6_sk(sk)->dontfrag;
Vlad Yasevich64223982015-01-31 10:40:15 -05001800
1801 err = __ip6_append_data(sk, fl6, &queue, &cork.base, &v6_cork,
1802 &current->task_frag, getfrag, from,
1803 length + exthdrlen, transhdrlen + exthdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001804 flags, ipc6, sockc);
Vlad Yasevich64223982015-01-31 10:40:15 -05001805 if (err) {
1806 __ip6_flush_pending_frames(sk, &queue, &cork, &v6_cork);
1807 return ERR_PTR(err);
1808 }
1809
1810 return __ip6_make_skb(sk, &queue, &cork, &v6_cork);
1811}