blob: 59eb4ed99ce80c0b6fd2ece1a877e3c9a1d57bcc [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * IPv6 output functions
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09003 * Linux INET6 implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 *
5 * Authors:
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09006 * Pedro Roque <roque@di.fc.ul.pt>
Linus Torvalds1da177e2005-04-16 15:20:36 -07007 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * Based on linux/net/ipv4/ip_output.c
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
14 *
15 * Changes:
16 * A.N.Kuznetsov : airthmetics in fragmentation.
17 * extension headers are implemented.
18 * route changes now work.
19 * ip6_forward does not confuse sniffers.
20 * etc.
21 *
22 * H. von Brand : Added missing #include <linux/string.h>
Ian Morris67ba4152014-08-24 21:53:10 +010023 * Imran Patel : frag id should be in NBO
Linus Torvalds1da177e2005-04-16 15:20:36 -070024 * Kazunori MIYAZAWA @USAGI
25 * : add ip6_append_data and related functions
26 * for datagram xmit
27 */
28
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/errno.h>
Herbert Xuef76bc22008-01-11 19:15:08 -080030#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <linux/string.h>
32#include <linux/socket.h>
33#include <linux/net.h>
34#include <linux/netdevice.h>
35#include <linux/if_arp.h>
36#include <linux/in6.h>
37#include <linux/tcp.h>
38#include <linux/route.h>
Herbert Xub59f45d2006-05-27 23:05:54 -070039#include <linux/module.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090040#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
42#include <linux/netfilter.h>
43#include <linux/netfilter_ipv6.h>
44
45#include <net/sock.h>
46#include <net/snmp.h>
47
48#include <net/ipv6.h>
49#include <net/ndisc.h>
50#include <net/protocol.h>
51#include <net/ip6_route.h>
52#include <net/addrconf.h>
53#include <net/rawv6.h>
54#include <net/icmp.h>
55#include <net/xfrm.h>
56#include <net/checksum.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090057#include <linux/mroute6.h>
David Ahernca254492015-10-12 11:47:10 -070058#include <net/l3mdev.h>
Roopa Prabhu14972cb2016-08-24 20:10:43 -070059#include <net/lwtunnel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -050061static int ip6_finish_output2(struct net *net, struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -070062{
Eric Dumazetadf30902009-06-02 05:19:30 +000063 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070064 struct net_device *dev = dst->dev;
David S. Millerf6b72b62011-07-14 07:53:20 -070065 struct neighbour *neigh;
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +000066 struct in6_addr *nexthop;
67 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070068
69 skb->protocol = htons(ETH_P_IPV6);
70 skb->dev = dev;
71
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070072 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) {
Eric Dumazetadf30902009-06-02 05:19:30 +000073 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
David Miller7026b1d2015-04-05 22:19:04 -040075 if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(sk) &&
Eric W. Biederman78126c42015-09-15 20:04:09 -050076 ((mroute6_socket(net, skb) &&
Benjamin Therybd91b8b2008-12-10 16:07:08 -080077 !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) ||
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090078 ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
79 &ipv6_hdr(skb)->saddr))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
81
82 /* Do not check for IFF_ALLMULTI; multicast routing
83 is not supported in any case.
84 */
85 if (newskb)
Jan Engelhardtb2e0b382010-03-23 04:09:07 +010086 NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING,
Eric W. Biederman29a26a52015-09-15 20:04:16 -050087 net, sk, newskb, NULL, newskb->dev,
Michel Machado95603e22012-06-12 10:16:35 +000088 dev_loopback_xmit);
Linus Torvalds1da177e2005-04-16 15:20:36 -070089
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070090 if (ipv6_hdr(skb)->hop_limit == 0) {
Eric W. Biederman78126c42015-09-15 20:04:09 -050091 IP6_INC_STATS(net, idev,
Denis V. Lunev3bd653c2008-10-08 10:54:51 -070092 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -070093 kfree_skb(skb);
94 return 0;
95 }
96 }
97
Eric W. Biederman78126c42015-09-15 20:04:09 -050098 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUTMCAST, skb->len);
Hannes Frederic Sowadd408512013-02-10 02:33:35 +000099
100 if (IPV6_ADDR_MC_SCOPE(&ipv6_hdr(skb)->daddr) <=
101 IPV6_ADDR_SCOPE_NODELOCAL &&
102 !(dev->flags & IFF_LOOPBACK)) {
103 kfree_skb(skb);
104 return 0;
105 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 }
107
Roopa Prabhu14972cb2016-08-24 20:10:43 -0700108 if (lwtunnel_xmit_redirect(dst->lwtstate)) {
109 int res = lwtunnel_xmit(skb);
110
111 if (res < 0 || res == LWTUNNEL_XMIT_DONE)
112 return res;
113 }
114
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000115 rcu_read_lock_bh();
Martin KaFai Lau2647a9b2015-05-22 20:55:58 -0700116 nexthop = rt6_nexthop((struct rt6_info *)dst, &ipv6_hdr(skb)->daddr);
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000117 neigh = __ipv6_neigh_lookup_noref(dst->dev, nexthop);
118 if (unlikely(!neigh))
119 neigh = __neigh_create(&nd_tbl, nexthop, dst->dev, false);
120 if (!IS_ERR(neigh)) {
121 ret = dst_neigh_output(dst, neigh, skb);
122 rcu_read_unlock_bh();
123 return ret;
124 }
125 rcu_read_unlock_bh();
David S. Miller05e3aa02011-07-16 17:26:00 -0700126
Eric W. Biederman78126c42015-09-15 20:04:09 -0500127 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200128 kfree_skb(skb);
129 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130}
131
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -0500132static int ip6_finish_output(struct net *net, struct sock *sk, struct sk_buff *skb)
Jan Engelhardt9e508492010-04-13 15:28:11 +0200133{
134 if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
Jiri Pirko9037c352013-11-06 17:52:19 +0100135 dst_allfrag(skb_dst(skb)) ||
136 (IP6CB(skb)->frag_max_size && skb->len > IP6CB(skb)->frag_max_size))
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500137 return ip6_fragment(net, sk, skb, ip6_finish_output2);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200138 else
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500139 return ip6_finish_output2(net, sk, skb);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200140}
141
Eric W. Biedermanede20592015-10-07 16:48:47 -0500142int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143{
Jan Engelhardt9e508492010-04-13 15:28:11 +0200144 struct net_device *dev = skb_dst(skb)->dev;
Eric Dumazetadf30902009-06-02 05:19:30 +0000145 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Eric W. Biedermanbe10de02015-09-17 17:21:31 -0500146
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900147 if (unlikely(idev->cnf.disable_ipv6)) {
Eric W. Biederman19a06442015-09-15 20:04:10 -0500148 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900149 kfree_skb(skb);
150 return 0;
151 }
152
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500153 return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING,
154 net, sk, skb, NULL, dev,
Jan Engelhardt9c6eb282010-04-13 15:32:16 +0200155 ip6_finish_output,
156 !(IP6CB(skb)->flags & IP6SKB_REROUTED));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157}
158
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159/*
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700160 * xmit an sk_buff (used by TCP, SCTP and DCCP)
161 * Note : socket lock is not held for SYNACK packets, but might be modified
162 * by calls to skb_set_owner_w() and ipv6_local_error(),
163 * which are using proper atomic operations or spinlocks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164 */
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700165int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
Eric Dumazetb903d322011-10-27 00:44:35 -0400166 struct ipv6_txoptions *opt, int tclass)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167{
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700168 struct net *net = sock_net(sk);
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700169 const struct ipv6_pinfo *np = inet6_sk(sk);
David S. Miller4c9483b2011-03-12 16:22:43 -0500170 struct in6_addr *first_hop = &fl6->daddr;
Eric Dumazetadf30902009-06-02 05:19:30 +0000171 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 struct ipv6hdr *hdr;
David S. Miller4c9483b2011-03-12 16:22:43 -0500173 u8 proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 int seg_len = skb->len;
Gerrit Renkere651f032009-08-09 08:12:48 +0000175 int hlimit = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 u32 mtu;
177
178 if (opt) {
Chuck Leverc2636b42007-10-23 21:07:32 -0700179 unsigned int head_room;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180
181 /* First: exthdrs may take lots of space (~8K for now)
182 MAX_HEADER is not enough.
183 */
184 head_room = opt->opt_nflen + opt->opt_flen;
185 seg_len += head_room;
186 head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
187
188 if (skb_headroom(skb) < head_room) {
189 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
Ian Morris63159f22015-03-29 14:00:04 +0100190 if (!skb2) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000191 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900192 IPSTATS_MIB_OUTDISCARDS);
193 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 return -ENOBUFS;
195 }
Eric Dumazet808db802012-04-24 10:17:59 +0000196 consume_skb(skb);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900197 skb = skb2;
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700198 /* skb_set_owner_w() changes sk->sk_wmem_alloc atomically,
199 * it is safe to call in our context (socket lock not held)
200 */
201 skb_set_owner_w(skb, (struct sock *)sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 }
203 if (opt->opt_flen)
204 ipv6_push_frag_opts(skb, opt, &proto);
205 if (opt->opt_nflen)
206 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
207 }
208
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700209 skb_push(skb, sizeof(struct ipv6hdr));
210 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700211 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212
213 /*
214 * Fill in the IPv6 header
215 */
Eric Dumazetb903d322011-10-27 00:44:35 -0400216 if (np)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 hlimit = np->hop_limit;
218 if (hlimit < 0)
YOSHIFUJI Hideaki6b75d092008-03-10 06:00:30 -0400219 hlimit = ip6_dst_hoplimit(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220
Tom Herbertcb1ce2e2014-07-01 21:33:10 -0700221 ip6_flow_hdr(hdr, tclass, ip6_make_flowlabel(net, skb, fl6->flowlabel,
Tom Herbert67800f92015-07-31 16:52:11 -0700222 np->autoflowlabel, fl6));
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 hdr->payload_len = htons(seg_len);
225 hdr->nexthdr = proto;
226 hdr->hop_limit = hlimit;
227
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000228 hdr->saddr = fl6->saddr;
229 hdr->daddr = *first_hop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +0200231 skb->protocol = htons(ETH_P_IPV6);
Patrick McHardya2c20642006-01-08 22:37:26 -0800232 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800233 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -0800234
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 mtu = dst_mtu(dst);
WANG Cong60ff7462014-05-04 16:39:18 -0700236 if ((skb->len <= mtu) || skb->ignore_df || skb_is_gso(skb)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000237 IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)),
Neil Hormanedf391f2009-04-27 02:45:02 -0700238 IPSTATS_MIB_OUT, skb->len);
David Aherna8e3e1a2016-09-10 12:09:53 -0700239
240 /* if egress device is enslaved to an L3 master device pass the
241 * skb to its handler for processing
242 */
243 skb = l3mdev_ip6_out((struct sock *)sk, skb);
244 if (unlikely(!skb))
245 return 0;
246
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700247 /* hooks should never assume socket lock is held.
248 * we promote our socket to non const
249 */
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500250 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700251 net, (struct sock *)sk, skb, NULL, dst->dev,
Eric W. Biederman13206b62015-10-07 16:48:35 -0500252 dst_output);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253 }
254
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 skb->dev = dst->dev;
Eric Dumazet1c1e9d22015-09-25 07:39:20 -0700256 /* ipv6_local_error() does not require socket lock,
257 * we promote our socket to non const
258 */
259 ipv6_local_error((struct sock *)sk, EMSGSIZE, fl6, mtu);
260
Eric Dumazetadf30902009-06-02 05:19:30 +0000261 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 kfree_skb(skb);
263 return -EMSGSIZE;
264}
YOSHIFUJI Hideaki71590392007-02-22 22:05:40 +0900265EXPORT_SYMBOL(ip6_xmit);
266
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
268{
269 struct ip6_ra_chain *ra;
270 struct sock *last = NULL;
271
272 read_lock(&ip6_ra_lock);
273 for (ra = ip6_ra_chain; ra; ra = ra->next) {
274 struct sock *sk = ra->sk;
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700275 if (sk && ra->sel == sel &&
276 (!sk->sk_bound_dev_if ||
277 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 if (last) {
279 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
280 if (skb2)
281 rawv6_rcv(last, skb2);
282 }
283 last = sk;
284 }
285 }
286
287 if (last) {
288 rawv6_rcv(last, skb);
289 read_unlock(&ip6_ra_lock);
290 return 1;
291 }
292 read_unlock(&ip6_ra_lock);
293 return 0;
294}
295
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700296static int ip6_forward_proxy_check(struct sk_buff *skb)
297{
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700298 struct ipv6hdr *hdr = ipv6_hdr(skb);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700299 u8 nexthdr = hdr->nexthdr;
Jesse Gross75f28112011-11-30 17:05:51 -0800300 __be16 frag_off;
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700301 int offset;
302
303 if (ipv6_ext_hdr(nexthdr)) {
Jesse Gross75f28112011-11-30 17:05:51 -0800304 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700305 if (offset < 0)
306 return 0;
307 } else
308 offset = sizeof(struct ipv6hdr);
309
310 if (nexthdr == IPPROTO_ICMPV6) {
311 struct icmp6hdr *icmp6;
312
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700313 if (!pskb_may_pull(skb, (skb_network_header(skb) +
314 offset + 1 - skb->data)))
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700315 return 0;
316
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700317 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700318
319 switch (icmp6->icmp6_type) {
320 case NDISC_ROUTER_SOLICITATION:
321 case NDISC_ROUTER_ADVERTISEMENT:
322 case NDISC_NEIGHBOUR_SOLICITATION:
323 case NDISC_NEIGHBOUR_ADVERTISEMENT:
324 case NDISC_REDIRECT:
325 /* For reaction involving unicast neighbor discovery
326 * message destined to the proxied address, pass it to
327 * input function.
328 */
329 return 1;
330 default:
331 break;
332 }
333 }
334
Ville Nuorvala74553b02006-09-22 14:42:18 -0700335 /*
336 * The proxying router can't forward traffic sent to a link-local
337 * address, so signal the sender and discard the packet. This
338 * behavior is clarified by the MIPv6 specification.
339 */
340 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
341 dst_link_failure(skb);
342 return -1;
343 }
344
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700345 return 0;
346}
347
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -0500348static inline int ip6_forward_finish(struct net *net, struct sock *sk,
349 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350{
Eric W. Biederman13206b62015-10-07 16:48:35 -0500351 return dst_output(net, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352}
353
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100354static unsigned int ip6_dst_mtu_forward(const struct dst_entry *dst)
355{
356 unsigned int mtu;
357 struct inet6_dev *idev;
358
359 if (dst_metric_locked(dst, RTAX_MTU)) {
360 mtu = dst_metric_raw(dst, RTAX_MTU);
361 if (mtu)
362 return mtu;
363 }
364
365 mtu = IPV6_MIN_MTU;
366 rcu_read_lock();
367 idev = __in6_dev_get(dst->dev);
368 if (idev)
369 mtu = idev->cnf.mtu6;
370 rcu_read_unlock();
371
372 return mtu;
373}
374
Florian Westphalfe6cc552014-02-13 23:09:12 +0100375static bool ip6_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
376{
Florian Westphal418a3152014-05-05 00:03:34 +0200377 if (skb->len <= mtu)
Florian Westphalfe6cc552014-02-13 23:09:12 +0100378 return false;
379
WANG Cong60ff7462014-05-04 16:39:18 -0700380 /* ipv6 conntrack defrag sets max_frag_size + ignore_df */
Florian Westphalfe6cc552014-02-13 23:09:12 +0100381 if (IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu)
382 return true;
383
WANG Cong60ff7462014-05-04 16:39:18 -0700384 if (skb->ignore_df)
Florian Westphal418a3152014-05-05 00:03:34 +0200385 return false;
386
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -0300387 if (skb_is_gso(skb) && skb_gso_validate_mtu(skb, mtu))
Florian Westphalfe6cc552014-02-13 23:09:12 +0100388 return false;
389
390 return true;
391}
392
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393int ip6_forward(struct sk_buff *skb)
394{
Eric Dumazetadf30902009-06-02 05:19:30 +0000395 struct dst_entry *dst = skb_dst(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700396 struct ipv6hdr *hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 struct inet6_skb_parm *opt = IP6CB(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900398 struct net *net = dev_net(dst->dev);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800399 u32 mtu;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900400
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700401 if (net->ipv6.devconf_all->forwarding == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 goto error;
403
Li RongQing090f1162014-03-11 10:40:08 +0800404 if (skb->pkt_type != PACKET_HOST)
405 goto drop;
406
Hannes Frederic Sowa9ef2e962015-10-08 18:19:53 +0200407 if (unlikely(skb->sk))
408 goto drop;
409
Ben Hutchings4497b072008-06-19 16:22:28 -0700410 if (skb_warn_if_lro(skb))
411 goto drop;
412
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700414 __IP6_INC_STATS(net, ip6_dst_idev(dst),
415 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 goto drop;
417 }
418
Herbert Xu35fc92a2007-03-26 23:22:20 -0700419 skb_forward_csum(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420
421 /*
422 * We DO NOT make any processing on
423 * RA packets, pushing them to user level AS IS
424 * without ane WARRANTY that application will be able
425 * to interpret them. The reason is that we
426 * cannot make anything clever here.
427 *
428 * We are not end-node, so that if packet contains
429 * AH/ESP, we cannot make anything.
430 * Defragmentation also would be mistake, RA packets
431 * cannot be fragmented, because there is no warranty
432 * that different fragments will go along one path. --ANK
433 */
YOSHIFUJI Hideaki / 吉藤英明ab4eb352013-06-22 11:13:13 +0900434 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
435 if (ip6_call_ra_chain(skb, ntohs(opt->ra)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 return 0;
437 }
438
439 /*
440 * check and decrement ttl
441 */
442 if (hdr->hop_limit <= 1) {
443 /* Force OUTPUT device used as source address */
444 skb->dev = dst->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000445 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0);
Eric Dumazet1d015502016-04-27 16:44:40 -0700446 __IP6_INC_STATS(net, ip6_dst_idev(dst),
447 IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448
449 kfree_skb(skb);
450 return -ETIMEDOUT;
451 }
452
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -0700453 /* XXX: idev->cnf.proxy_ndp? */
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700454 if (net->ipv6.devconf_all->proxy_ndp &&
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800455 pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
Ville Nuorvala74553b02006-09-22 14:42:18 -0700456 int proxied = ip6_forward_proxy_check(skb);
457 if (proxied > 0)
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700458 return ip6_input(skb);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700459 else if (proxied < 0) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700460 __IP6_INC_STATS(net, ip6_dst_idev(dst),
461 IPSTATS_MIB_INDISCARDS);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700462 goto drop;
463 }
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700464 }
465
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 if (!xfrm6_route_forward(skb)) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700467 __IP6_INC_STATS(net, ip6_dst_idev(dst),
468 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 goto drop;
470 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000471 dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472
473 /* IPv6 specs say nothing about it, but it is clear that we cannot
474 send redirects to source routed frames.
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900475 We don't send redirects to frames decapsulated from IPsec.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 */
David S. Millerc45a3df2012-01-27 15:32:19 -0800477 if (skb->dev == dst->dev && opt->srcrt == 0 && !skb_sec_path(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 struct in6_addr *target = NULL;
David S. Millerfbfe95a2012-06-08 23:24:18 -0700479 struct inet_peer *peer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481
482 /*
483 * incoming and outgoing devices are the same
484 * send a redirect.
485 */
486
487 rt = (struct rt6_info *) dst;
David S. Millerc45a3df2012-01-27 15:32:19 -0800488 if (rt->rt6i_flags & RTF_GATEWAY)
489 target = &rt->rt6i_gateway;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 else
491 target = &hdr->daddr;
492
Martin KaFai Laufd0273d2015-05-22 20:55:57 -0700493 peer = inet_getpeer_v6(net->ipv6.peers, &hdr->daddr, 1);
David S. Miller92d86822011-02-04 15:55:25 -0800494
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 /* Limit redirects both by destination (here)
496 and by source (inside ndisc_send_redirect)
497 */
David S. Millerfbfe95a2012-06-08 23:24:18 -0700498 if (inet_peer_xrlim_allow(peer, 1*HZ))
David S. Miller49919692012-01-27 15:30:48 -0800499 ndisc_send_redirect(skb, target);
David S. Miller1d861aa2012-07-10 03:58:16 -0700500 if (peer)
501 inet_putpeer(peer);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700502 } else {
503 int addrtype = ipv6_addr_type(&hdr->saddr);
504
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 /* This check is security critical. */
YOSHIFUJI Hideakif81b2e72008-06-25 16:55:26 +0900506 if (addrtype == IPV6_ADDR_ANY ||
507 addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK))
David L Stevens5bb1ab02007-05-09 13:53:44 -0700508 goto error;
509 if (addrtype & IPV6_ADDR_LINKLOCAL) {
510 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000511 ICMPV6_NOT_NEIGHBOUR, 0);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700512 goto error;
513 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 }
515
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100516 mtu = ip6_dst_mtu_forward(dst);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800517 if (mtu < IPV6_MIN_MTU)
518 mtu = IPV6_MIN_MTU;
519
Florian Westphalfe6cc552014-02-13 23:09:12 +0100520 if (ip6_pkt_too_big(skb, mtu)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 /* Again, force OUTPUT device used as source address */
522 skb->dev = dst->dev;
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800523 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazet1d015502016-04-27 16:44:40 -0700524 __IP6_INC_STATS(net, ip6_dst_idev(dst),
525 IPSTATS_MIB_INTOOBIGERRORS);
526 __IP6_INC_STATS(net, ip6_dst_idev(dst),
527 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 kfree_skb(skb);
529 return -EMSGSIZE;
530 }
531
532 if (skb_cow(skb, dst->dev->hard_header_len)) {
Eric Dumazet1d015502016-04-27 16:44:40 -0700533 __IP6_INC_STATS(net, ip6_dst_idev(dst),
534 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 goto drop;
536 }
537
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700538 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
540 /* Mangling hops number delayed to point after skb COW */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900541
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 hdr->hop_limit--;
543
Eric Dumazet1d015502016-04-27 16:44:40 -0700544 __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
545 __IP6_ADD_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len);
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500546 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD,
547 net, NULL, skb, skb->dev, dst->dev,
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800548 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549
550error:
Eric Dumazet1d015502016-04-27 16:44:40 -0700551 __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552drop:
553 kfree_skb(skb);
554 return -EINVAL;
555}
556
557static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
558{
559 to->pkt_type = from->pkt_type;
560 to->priority = from->priority;
561 to->protocol = from->protocol;
Eric Dumazetadf30902009-06-02 05:19:30 +0000562 skb_dst_drop(to);
563 skb_dst_set(to, dst_clone(skb_dst(from)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 to->dev = from->dev;
Thomas Graf82e91ff2006-11-09 15:19:14 -0800565 to->mark = from->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566
567#ifdef CONFIG_NET_SCHED
568 to->tc_index = from->tc_index;
569#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -0700570 nf_copy(to, from);
James Morris984bc162006-06-09 00:29:17 -0700571 skb_copy_secmark(to, from);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572}
573
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500574int ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
575 int (*output)(struct net *, struct sock *, struct sk_buff *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 struct sk_buff *frag;
Ian Morris67ba4152014-08-24 21:53:10 +0100578 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +0200579 struct ipv6_pinfo *np = skb->sk && !dev_recursion_level() ?
580 inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 struct ipv6hdr *tmp_hdr;
582 struct frag_hdr *fh;
583 unsigned int mtu, hlen, left, len;
Herbert Xua7ae1992011-11-18 02:20:04 +0000584 int hroom, troom;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700585 __be32 frag_id;
Ian Morris67ba4152014-08-24 21:53:10 +0100586 int ptr, offset = 0, err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 u8 *prevhdr, nexthdr = 0;
588
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 hlen = ip6_find_1stfragopt(skb, &prevhdr);
590 nexthdr = *prevhdr;
591
John Heffner628a5c52007-04-20 15:53:27 -0700592 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700593
594 /* We must not fragment if the socket is set to force MTU discovery
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800595 * or if the skb it not generated by a local socket.
John Heffnerb881ef72007-04-20 15:52:39 -0700596 */
Florian Westphal485fca62015-05-22 00:44:16 +0200597 if (unlikely(!skb->ignore_df && skb->len > mtu))
598 goto fail_toobig;
Eric Dumazeta34a1012012-05-18 21:51:44 +0000599
Florian Westphal485fca62015-05-22 00:44:16 +0200600 if (IP6CB(skb)->frag_max_size) {
601 if (IP6CB(skb)->frag_max_size > mtu)
602 goto fail_toobig;
603
604 /* don't send fragments larger than what we received */
605 mtu = IP6CB(skb)->frag_max_size;
606 if (mtu < IPV6_MIN_MTU)
607 mtu = IPV6_MIN_MTU;
John Heffnerb881ef72007-04-20 15:52:39 -0700608 }
609
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800610 if (np && np->frag_size < mtu) {
611 if (np->frag_size)
612 mtu = np->frag_size;
613 }
Hannes Frederic Sowa89bc7842015-10-28 13:21:04 +0100614 if (mtu < hlen + sizeof(struct frag_hdr) + 8)
Hannes Frederic Sowab72a2b02015-10-16 11:32:43 +0200615 goto fail_toobig;
Hannes Frederic Sowa1e0d69a2015-10-28 13:21:03 +0100616 mtu -= hlen + sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617
Martin KaFai Laufd0273d2015-05-22 20:55:57 -0700618 frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
619 &ipv6_hdr(skb)->saddr);
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700620
Hannes Frederic Sowa405c92f2015-10-27 22:40:42 +0100621 if (skb->ip_summed == CHECKSUM_PARTIAL &&
622 (err = skb_checksum_help(skb)))
623 goto fail;
624
Florian Westphal1d325d22015-09-16 17:26:14 +0200625 hroom = LL_RESERVED_SPACE(rt->dst.dev);
David S. Miller21dc3302010-08-23 00:13:46 -0700626 if (skb_has_frag_list(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 int first_len = skb_pagelen(skb);
Eric Dumazet3d130082010-09-21 08:47:45 +0000628 struct sk_buff *frag2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629
630 if (first_len - hlen > mtu ||
631 ((first_len - hlen) & 7) ||
Florian Westphal1d325d22015-09-16 17:26:14 +0200632 skb_cloned(skb) ||
633 skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 goto slow_path;
635
David S. Miller4d9092b2009-06-09 00:20:05 -0700636 skb_walk_frags(skb, frag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 /* Correct geometry. */
638 if (frag->len > mtu ||
639 ((frag->len & 7) && frag->next) ||
Florian Westphal1d325d22015-09-16 17:26:14 +0200640 skb_headroom(frag) < (hlen + hroom + sizeof(struct frag_hdr)))
Eric Dumazet3d130082010-09-21 08:47:45 +0000641 goto slow_path_clean;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 /* Partially cloned skb? */
644 if (skb_shared(frag))
Eric Dumazet3d130082010-09-21 08:47:45 +0000645 goto slow_path_clean;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700646
647 BUG_ON(frag->sk);
648 if (skb->sk) {
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700649 frag->sk = skb->sk;
650 frag->destructor = sock_wfree;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700651 }
Eric Dumazet3d130082010-09-21 08:47:45 +0000652 skb->truesize -= frag->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 }
654
655 err = 0;
656 offset = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 /* BUILD HEADER */
658
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800659 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700660 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 if (!tmp_hdr) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000662 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700663 IPSTATS_MIB_FRAGFAILS);
Florian Westphal1d325d22015-09-16 17:26:14 +0200664 err = -ENOMEM;
665 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 }
Florian Westphal1d325d22015-09-16 17:26:14 +0200667 frag = skb_shinfo(skb)->frag_list;
668 skb_frag_list_init(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 __skb_pull(skb, hlen);
Ian Morris67ba4152014-08-24 21:53:10 +0100671 fh = (struct frag_hdr *)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700672 __skb_push(skb, hlen);
673 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700674 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 fh->nexthdr = nexthdr;
677 fh->reserved = 0;
678 fh->frag_off = htons(IP6_MF);
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700679 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680
681 first_len = skb_pagelen(skb);
682 skb->data_len = first_len - skb_headlen(skb);
683 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700684 ipv6_hdr(skb)->payload_len = htons(first_len -
685 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900686
Changli Gaod8d1f302010-06-10 23:31:35 -0700687 dst_hold(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688
689 for (;;) {
690 /* Prepare header of the next frame,
691 * before previous one went down. */
692 if (frag) {
693 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300694 skb_reset_transport_header(frag);
Ian Morris67ba4152014-08-24 21:53:10 +0100695 fh = (struct frag_hdr *)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700696 __skb_push(frag, hlen);
697 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700698 memcpy(skb_network_header(frag), tmp_hdr,
699 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 offset += skb->len - hlen - sizeof(struct frag_hdr);
701 fh->nexthdr = nexthdr;
702 fh->reserved = 0;
703 fh->frag_off = htons(offset);
Ian Morris53b24b82015-03-29 14:00:05 +0100704 if (frag->next)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 fh->frag_off |= htons(IP6_MF);
706 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700707 ipv6_hdr(frag)->payload_len =
708 htons(frag->len -
709 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 ip6_copy_metadata(frag, skb);
711 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900712
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500713 err = output(net, sk, skb);
Ian Morris67ba4152014-08-24 21:53:10 +0100714 if (!err)
Changli Gaod8d1f302010-06-10 23:31:35 -0700715 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700716 IPSTATS_MIB_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700717
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 if (err || !frag)
719 break;
720
721 skb = frag;
722 frag = skb->next;
723 skb->next = NULL;
724 }
725
Jesper Juhla51482b2005-11-08 09:41:34 -0800726 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727
728 if (err == 0) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700729 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700730 IPSTATS_MIB_FRAGOKS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000731 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 return 0;
733 }
734
Florian Westphal46cfd722014-09-10 01:08:46 +0200735 kfree_skb_list(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736
Changli Gaod8d1f302010-06-10 23:31:35 -0700737 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700738 IPSTATS_MIB_FRAGFAILS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000739 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 return err;
Eric Dumazet3d130082010-09-21 08:47:45 +0000741
742slow_path_clean:
743 skb_walk_frags(skb, frag2) {
744 if (frag2 == frag)
745 break;
746 frag2->sk = NULL;
747 frag2->destructor = NULL;
748 skb->truesize += frag2->truesize;
749 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 }
751
752slow_path:
753 left = skb->len - hlen; /* Space per frame */
754 ptr = hlen; /* Where to start from */
755
756 /*
757 * Fragment the datagram.
758 */
759
760 *prevhdr = NEXTHDR_FRAGMENT;
Herbert Xua7ae1992011-11-18 02:20:04 +0000761 troom = rt->dst.dev->needed_tailroom;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762
763 /*
764 * Keep copying data until we run out.
765 */
Ian Morris67ba4152014-08-24 21:53:10 +0100766 while (left > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 len = left;
768 /* IF: it doesn't fit, use 'mtu' - the data space left */
769 if (len > mtu)
770 len = mtu;
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300771 /* IF: we are not sending up to and including the packet end
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772 then align the next start on an eight byte boundary */
773 if (len < left) {
774 len &= ~7;
775 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
Joe Perchescbffccc2014-11-05 14:39:21 -0800777 /* Allocate buffer */
778 frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) +
779 hroom + troom, GFP_ATOMIC);
780 if (!frag) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000781 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900782 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 err = -ENOMEM;
784 goto fail;
785 }
786
787 /*
788 * Set up data on packet
789 */
790
791 ip6_copy_metadata(frag, skb);
Herbert Xua7ae1992011-11-18 02:20:04 +0000792 skb_reserve(frag, hroom);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700794 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300795 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700796 frag->transport_header = (frag->network_header + hlen +
797 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798
799 /*
800 * Charge the memory for the fragment to any owner
801 * it might possess
802 */
803 if (skb->sk)
804 skb_set_owner_w(frag, skb->sk);
805
806 /*
807 * Copy the packet header into the new buffer.
808 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300809 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810
811 /*
812 * Build fragment header.
813 */
814 fh->nexthdr = nexthdr;
815 fh->reserved = 0;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700816 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817
818 /*
819 * Copy a block of the IP datagram.
820 */
Himangi Saraogie3f0b862014-07-12 01:57:17 +0530821 BUG_ON(skb_copy_bits(skb, ptr, skb_transport_header(frag),
822 len));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 left -= len;
824
825 fh->frag_off = htons(offset);
826 if (left > 0)
827 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700828 ipv6_hdr(frag)->payload_len = htons(frag->len -
829 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830
831 ptr += len;
832 offset += len;
833
834 /*
835 * Put this fragment into the sending queue.
836 */
Eric W. Biederman7d8c6e32015-06-12 22:12:04 -0500837 err = output(net, sk, frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 if (err)
839 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700840
Eric Dumazetadf30902009-06-02 05:19:30 +0000841 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700842 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000844 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900845 IPSTATS_MIB_FRAGOKS);
Eric Dumazet808db802012-04-24 10:17:59 +0000846 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 return err;
848
Florian Westphal485fca62015-05-22 00:44:16 +0200849fail_toobig:
850 if (skb->sk && dst_allfrag(skb_dst(skb)))
851 sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK);
852
853 skb->dev = skb_dst(skb)->dev;
854 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
855 err = -EMSGSIZE;
856
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857fail:
Eric Dumazetadf30902009-06-02 05:19:30 +0000858 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900859 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900860 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 return err;
862}
863
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000864static inline int ip6_rt_check(const struct rt6key *rt_key,
865 const struct in6_addr *fl_addr,
866 const struct in6_addr *addr_cache)
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700867{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000868 return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
Ian Morris63159f22015-03-29 14:00:04 +0100869 (!addr_cache || !ipv6_addr_equal(fl_addr, addr_cache));
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700870}
871
Herbert Xu497c6152006-07-30 20:19:33 -0700872static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
873 struct dst_entry *dst,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000874 const struct flowi6 *fl6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875{
Herbert Xu497c6152006-07-30 20:19:33 -0700876 struct ipv6_pinfo *np = inet6_sk(sk);
Eric Dumazeta963a372013-06-26 04:15:07 -0700877 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
Herbert Xu497c6152006-07-30 20:19:33 -0700879 if (!dst)
880 goto out;
881
Eric Dumazeta963a372013-06-26 04:15:07 -0700882 if (dst->ops->family != AF_INET6) {
883 dst_release(dst);
884 return NULL;
885 }
886
887 rt = (struct rt6_info *)dst;
Herbert Xu497c6152006-07-30 20:19:33 -0700888 /* Yes, checking route validity in not connected
889 * case is not very simple. Take into account,
890 * that we do not support routing by source, TOS,
Ian Morris67ba4152014-08-24 21:53:10 +0100891 * and MSG_DONTROUTE --ANK (980726)
Herbert Xu497c6152006-07-30 20:19:33 -0700892 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700893 * 1. ip6_rt_check(): If route was host route,
894 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700895 * If it is network route, we still may
896 * check its validity using saved pointer
897 * to the last used address: daddr_cache.
898 * We do not want to save whole address now,
899 * (because main consumer of this service
900 * is tcp, which has not this problem),
901 * so that the last trick works only on connected
902 * sockets.
903 * 2. oif also should be the same.
904 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500905 if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700906#ifdef CONFIG_IPV6_SUBTREES
David S. Miller4c9483b2011-03-12 16:22:43 -0500907 ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700908#endif
David Ahernca254492015-10-12 11:47:10 -0700909 (!(fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF) &&
910 (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex))) {
Herbert Xu497c6152006-07-30 20:19:33 -0700911 dst_release(dst);
912 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913 }
914
Herbert Xu497c6152006-07-30 20:19:33 -0700915out:
916 return dst;
917}
918
Eric Dumazet3aef9342015-09-25 07:39:12 -0700919static int ip6_dst_lookup_tail(struct net *net, const struct sock *sk,
David S. Miller4c9483b2011-03-12 16:22:43 -0500920 struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700921{
David S. Miller69cce1d2011-07-17 23:09:49 -0700922#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
923 struct neighbour *n;
David S. Miller97cac082012-07-02 22:43:47 -0700924 struct rt6_info *rt;
David S. Miller69cce1d2011-07-17 23:09:49 -0700925#endif
926 int err;
Paolo Abeni6f21c962016-01-29 12:30:19 +0100927 int flags = 0;
Herbert Xu497c6152006-07-30 20:19:33 -0700928
Markus Stenberge16e8882015-05-05 13:36:59 +0300929 /* The correct way to handle this would be to do
930 * ip6_route_get_saddr, and then ip6_route_output; however,
931 * the route-specific preferred source forces the
932 * ip6_route_output call _before_ ip6_route_get_saddr.
933 *
934 * In source specific routing (no src=any default route),
935 * ip6_route_output will fail given src=any saddr, though, so
936 * that's why we try it again later.
937 */
938 if (ipv6_addr_any(&fl6->saddr) && (!*dst || !(*dst)->error)) {
939 struct rt6_info *rt;
940 bool had_dst = *dst != NULL;
941
942 if (!had_dst)
943 *dst = ip6_route_output(net, sk, fl6);
944 rt = (*dst)->error ? NULL : (struct rt6_info *)*dst;
945 err = ip6_route_get_saddr(net, rt, &fl6->daddr,
946 sk ? inet6_sk(sk)->srcprefs : 0,
947 &fl6->saddr);
948 if (err)
949 goto out_err_release;
950
951 /* If we had an erroneous initial result, pretend it
952 * never existed and let the SA-enabled version take
953 * over.
954 */
955 if (!had_dst && (*dst)->error) {
956 dst_release(*dst);
957 *dst = NULL;
958 }
Paolo Abeni6f21c962016-01-29 12:30:19 +0100959
960 if (fl6->flowi6_oif)
961 flags |= RT6_LOOKUP_F_IFACE;
Markus Stenberge16e8882015-05-05 13:36:59 +0300962 }
963
Ian Morris63159f22015-03-29 14:00:04 +0100964 if (!*dst)
Paolo Abeni6f21c962016-01-29 12:30:19 +0100965 *dst = ip6_route_output_flags(net, sk, fl6, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966
Ian Morrise5d08d72014-11-23 21:28:43 +0000967 err = (*dst)->error;
968 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 goto out_err_release;
970
Neil Horman95c385b2007-04-25 17:08:10 -0700971#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700972 /*
973 * Here if the dst entry we've looked up
974 * has a neighbour entry that is in the INCOMPLETE
975 * state and the src address from the flow is
976 * marked as OPTIMISTIC, we release the found
977 * dst entry and replace it instead with the
978 * dst entry of the nexthop router
979 */
Eric Dumazetc56bf6f2012-07-06 09:19:05 +0200980 rt = (struct rt6_info *) *dst;
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000981 rcu_read_lock_bh();
Martin KaFai Lau2647a9b2015-05-22 20:55:58 -0700982 n = __ipv6_neigh_lookup_noref(rt->dst.dev,
983 rt6_nexthop(rt, &fl6->daddr));
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000984 err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0;
985 rcu_read_unlock_bh();
986
987 if (err) {
Neil Hormane550dfb2008-09-09 13:51:35 -0700988 struct inet6_ifaddr *ifp;
David S. Miller4c9483b2011-03-12 16:22:43 -0500989 struct flowi6 fl_gw6;
Neil Hormane550dfb2008-09-09 13:51:35 -0700990 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -0700991
David S. Miller4c9483b2011-03-12 16:22:43 -0500992 ifp = ipv6_get_ifaddr(net, &fl6->saddr,
Neil Hormane550dfb2008-09-09 13:51:35 -0700993 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -0700994
Neil Hormane550dfb2008-09-09 13:51:35 -0700995 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
996 if (ifp)
997 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -0700998
Neil Hormane550dfb2008-09-09 13:51:35 -0700999 if (redirect) {
1000 /*
1001 * We need to get the dst entry for the
1002 * default router instead
1003 */
1004 dst_release(*dst);
David S. Miller4c9483b2011-03-12 16:22:43 -05001005 memcpy(&fl_gw6, fl6, sizeof(struct flowi6));
1006 memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr));
1007 *dst = ip6_route_output(net, sk, &fl_gw6);
Ian Morrise5d08d72014-11-23 21:28:43 +00001008 err = (*dst)->error;
1009 if (err)
Neil Hormane550dfb2008-09-09 13:51:35 -07001010 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -07001011 }
Neil Hormane550dfb2008-09-09 13:51:35 -07001012 }
Neil Horman95c385b2007-04-25 17:08:10 -07001013#endif
1014
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 return 0;
1016
1017out_err_release:
1018 dst_release(*dst);
1019 *dst = NULL;
David Ahern8a966fc2016-09-10 12:09:59 -07001020
David Ahern0d240e72016-06-16 16:24:25 -07001021 if (err == -ENETUNREACH)
1022 IP6_INC_STATS(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 return err;
1024}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001025
Herbert Xu497c6152006-07-30 20:19:33 -07001026/**
1027 * ip6_dst_lookup - perform route lookup on flow
1028 * @sk: socket which provides route info
1029 * @dst: pointer to dst_entry * for result
David S. Miller4c9483b2011-03-12 16:22:43 -05001030 * @fl6: flow to lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001031 *
1032 * This function performs a route lookup on the given flow.
1033 *
1034 * It returns zero on success, or a standard errno code on error.
1035 */
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001036int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1037 struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -07001038{
1039 *dst = NULL;
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001040 return ip6_dst_lookup_tail(net, sk, dst, fl6);
Herbert Xu497c6152006-07-30 20:19:33 -07001041}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -08001042EXPORT_SYMBOL_GPL(ip6_dst_lookup);
1043
Herbert Xu497c6152006-07-30 20:19:33 -07001044/**
David S. Miller68d0c6d2011-03-01 13:19:07 -08001045 * ip6_dst_lookup_flow - perform route lookup on flow with ipsec
1046 * @sk: socket which provides route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001047 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001048 * @final_dst: final destination address for ipsec lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001049 *
1050 * This function performs a route lookup on the given flow.
1051 *
1052 * It returns a valid dst pointer on success, or a pointer encoded
1053 * error code.
1054 */
Eric Dumazet3aef9342015-09-25 07:39:12 -07001055struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001056 const struct in6_addr *final_dst)
David S. Miller68d0c6d2011-03-01 13:19:07 -08001057{
1058 struct dst_entry *dst = NULL;
1059 int err;
1060
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001061 err = ip6_dst_lookup_tail(sock_net(sk), sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001062 if (err)
1063 return ERR_PTR(err);
1064 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001065 fl6->daddr = *final_dst;
David S. Miller2774c132011-03-01 14:59:04 -08001066
Steffen Klassertf92ee612014-09-16 10:08:40 +02001067 return xfrm_lookup_route(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001068}
1069EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow);
1070
1071/**
1072 * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow
1073 * @sk: socket which provides the dst cache and route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001074 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001075 * @final_dst: final destination address for ipsec lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001076 *
1077 * This function performs a route lookup on the given flow with the
1078 * possibility of using the cached route in the socket if it is valid.
1079 * It will take the socket dst lock when operating on the dst cache.
1080 * As a result, this function can only be used in process context.
1081 *
David S. Miller68d0c6d2011-03-01 13:19:07 -08001082 * It returns a valid dst pointer on success, or a pointer encoded
1083 * error code.
Herbert Xu497c6152006-07-30 20:19:33 -07001084 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001085struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001086 const struct in6_addr *final_dst)
Herbert Xu497c6152006-07-30 20:19:33 -07001087{
David S. Miller68d0c6d2011-03-01 13:19:07 -08001088 struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
Herbert Xu497c6152006-07-30 20:19:33 -07001089
David S. Miller4c9483b2011-03-12 16:22:43 -05001090 dst = ip6_sk_dst_check(sk, dst, fl6);
Jakub Sitnicki00bc0ef2016-06-08 15:13:34 +02001091 if (!dst)
1092 dst = ip6_dst_lookup_flow(sk, fl6, final_dst);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001093
Jakub Sitnicki00bc0ef2016-06-08 15:13:34 +02001094 return dst;
Herbert Xu497c6152006-07-30 20:19:33 -07001095}
David S. Miller68d0c6d2011-03-01 13:19:07 -08001096EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
Herbert Xu497c6152006-07-30 20:19:33 -07001097
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001098static inline int ip6_ufo_append_data(struct sock *sk,
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001099 struct sk_buff_head *queue,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001100 int getfrag(void *from, char *to, int offset, int len,
1101 int odd, struct sk_buff *skb),
1102 void *from, int length, int hh_len, int fragheaderlen,
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001103 int exthdrlen, int transhdrlen, int mtu,
1104 unsigned int flags, const struct flowi6 *fl6)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001105
1106{
1107 struct sk_buff *skb;
1108 int err;
1109
1110 /* There is support for UDP large send offload by network
1111 * device, so create one single skb packet containing complete
1112 * udp datagram
1113 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001114 skb = skb_peek_tail(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001115 if (!skb) {
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001116 skb = sock_alloc_send_skb(sk,
1117 hh_len + fragheaderlen + transhdrlen + 20,
1118 (flags & MSG_DONTWAIT), &err);
Ian Morris63159f22015-03-29 14:00:04 +01001119 if (!skb)
Zheng Yan504744e2011-10-28 00:26:00 -04001120 return err;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001121
1122 /* reserve space for Hardware header */
1123 skb_reserve(skb, hh_len);
1124
1125 /* create space for UDP/IP header */
Ian Morris67ba4152014-08-24 21:53:10 +01001126 skb_put(skb, fragheaderlen + transhdrlen);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001127
1128 /* initialize network header pointer */
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001129 skb_set_network_header(skb, exthdrlen);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001130
1131 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001132 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001133
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001134 skb->protocol = htons(ETH_P_IPV6);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001135 skb->csum = 0;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001136
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001137 __skb_queue_tail(queue, skb);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001138 } else if (skb_is_gso(skb)) {
1139 goto append;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001140 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001141
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001142 skb->ip_summed = CHECKSUM_PARTIAL;
1143 /* Specify the length of each IPv6 datagram fragment.
1144 * It has to be a multiple of 8.
1145 */
1146 skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
1147 sizeof(struct frag_hdr)) & ~7;
1148 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001149 skb_shinfo(skb)->ip6_frag_id = ipv6_select_ident(sock_net(sk),
1150 &fl6->daddr,
1151 &fl6->saddr);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001152
1153append:
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001154 return skb_append_datato_frags(sk, skb, getfrag, from,
1155 (length - transhdrlen));
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001156}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157
Herbert Xu0178b692009-02-05 15:15:50 -08001158static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
1159 gfp_t gfp)
1160{
1161 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1162}
1163
1164static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
1165 gfp_t gfp)
1166{
1167 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1168}
1169
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001170static void ip6_append_data_mtu(unsigned int *mtu,
Gao feng0c183372012-05-26 01:30:53 +00001171 int *maxfraglen,
1172 unsigned int fragheaderlen,
1173 struct sk_buff *skb,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001174 struct rt6_info *rt,
luciene367c2d2014-03-17 12:51:01 +08001175 unsigned int orig_mtu)
Gao feng0c183372012-05-26 01:30:53 +00001176{
1177 if (!(rt->dst.flags & DST_XFRM_TUNNEL)) {
Ian Morris63159f22015-03-29 14:00:04 +01001178 if (!skb) {
Gao feng0c183372012-05-26 01:30:53 +00001179 /* first fragment, reserve header_len */
luciene367c2d2014-03-17 12:51:01 +08001180 *mtu = orig_mtu - rt->dst.header_len;
Gao feng0c183372012-05-26 01:30:53 +00001181
1182 } else {
1183 /*
1184 * this fragment is not first, the headers
1185 * space is regarded as data space.
1186 */
luciene367c2d2014-03-17 12:51:01 +08001187 *mtu = orig_mtu;
Gao feng0c183372012-05-26 01:30:53 +00001188 }
1189 *maxfraglen = ((*mtu - fragheaderlen) & ~7)
1190 + fragheaderlen - sizeof(struct frag_hdr);
1191 }
1192}
1193
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001194static int ip6_setup_cork(struct sock *sk, struct inet_cork_full *cork,
Wei Wang26879da2016-05-02 21:40:07 -07001195 struct inet6_cork *v6_cork, struct ipcm6_cookie *ipc6,
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001196 struct rt6_info *rt, struct flowi6 *fl6)
1197{
1198 struct ipv6_pinfo *np = inet6_sk(sk);
1199 unsigned int mtu;
Wei Wang26879da2016-05-02 21:40:07 -07001200 struct ipv6_txoptions *opt = ipc6->opt;
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001201
1202 /*
1203 * setup for corking
1204 */
1205 if (opt) {
1206 if (WARN_ON(v6_cork->opt))
1207 return -EINVAL;
1208
1209 v6_cork->opt = kzalloc(opt->tot_len, sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001210 if (unlikely(!v6_cork->opt))
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001211 return -ENOBUFS;
1212
1213 v6_cork->opt->tot_len = opt->tot_len;
1214 v6_cork->opt->opt_flen = opt->opt_flen;
1215 v6_cork->opt->opt_nflen = opt->opt_nflen;
1216
1217 v6_cork->opt->dst0opt = ip6_opt_dup(opt->dst0opt,
1218 sk->sk_allocation);
1219 if (opt->dst0opt && !v6_cork->opt->dst0opt)
1220 return -ENOBUFS;
1221
1222 v6_cork->opt->dst1opt = ip6_opt_dup(opt->dst1opt,
1223 sk->sk_allocation);
1224 if (opt->dst1opt && !v6_cork->opt->dst1opt)
1225 return -ENOBUFS;
1226
1227 v6_cork->opt->hopopt = ip6_opt_dup(opt->hopopt,
1228 sk->sk_allocation);
1229 if (opt->hopopt && !v6_cork->opt->hopopt)
1230 return -ENOBUFS;
1231
1232 v6_cork->opt->srcrt = ip6_rthdr_dup(opt->srcrt,
1233 sk->sk_allocation);
1234 if (opt->srcrt && !v6_cork->opt->srcrt)
1235 return -ENOBUFS;
1236
1237 /* need source address above miyazawa*/
1238 }
1239 dst_hold(&rt->dst);
1240 cork->base.dst = &rt->dst;
1241 cork->fl.u.ip6 = *fl6;
Wei Wang26879da2016-05-02 21:40:07 -07001242 v6_cork->hop_limit = ipc6->hlimit;
1243 v6_cork->tclass = ipc6->tclass;
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001244 if (rt->dst.flags & DST_XFRM_TUNNEL)
1245 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
1246 rt->dst.dev->mtu : dst_mtu(&rt->dst);
1247 else
1248 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
1249 rt->dst.dev->mtu : dst_mtu(rt->dst.path);
1250 if (np->frag_size < mtu) {
1251 if (np->frag_size)
1252 mtu = np->frag_size;
1253 }
1254 cork->base.fragsize = mtu;
1255 if (dst_allfrag(rt->dst.path))
1256 cork->base.flags |= IPCORK_ALLFRAG;
1257 cork->base.length = 0;
1258
1259 return 0;
1260}
1261
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001262static int __ip6_append_data(struct sock *sk,
1263 struct flowi6 *fl6,
1264 struct sk_buff_head *queue,
1265 struct inet_cork *cork,
1266 struct inet6_cork *v6_cork,
1267 struct page_frag *pfrag,
1268 int getfrag(void *from, char *to, int offset,
1269 int len, int odd, struct sk_buff *skb),
1270 void *from, int length, int transhdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001271 unsigned int flags, struct ipcm6_cookie *ipc6,
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001272 const struct sockcm_cookie *sockc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273{
Gao feng0c183372012-05-26 01:30:53 +00001274 struct sk_buff *skb, *skb_prev = NULL;
luciene367c2d2014-03-17 12:51:01 +08001275 unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001276 int exthdrlen = 0;
1277 int dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 int hh_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 int copy;
1280 int err;
1281 int offset = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001282 __u8 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001283 u32 tskey = 0;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001284 struct rt6_info *rt = (struct rt6_info *)cork->dst;
1285 struct ipv6_txoptions *opt = v6_cork->opt;
Vlad Yasevich32dce962015-01-31 10:40:18 -05001286 int csummode = CHECKSUM_NONE;
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001287 unsigned int maxnonfragsize, headersize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001289 skb = skb_peek_tail(queue);
1290 if (!skb) {
1291 exthdrlen = opt ? opt->opt_flen : 0;
Romain KUNTZ7efdba52013-01-16 12:47:40 +00001292 dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 }
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001294
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001295 mtu = cork->fragsize;
luciene367c2d2014-03-17 12:51:01 +08001296 orig_mtu = mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297
Changli Gaod8d1f302010-06-10 23:31:35 -07001298 hh_len = LL_RESERVED_SPACE(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001300 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001301 (opt ? opt->opt_nflen : 0);
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001302 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen -
1303 sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001305 headersize = sizeof(struct ipv6hdr) +
1306 (opt ? opt->opt_flen + opt->opt_nflen : 0) +
1307 (dst_allfrag(&rt->dst) ?
1308 sizeof(struct frag_hdr) : 0) +
1309 rt->rt6i_nfheader_len;
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001310
Wei Wang26879da2016-05-02 21:40:07 -07001311 if (cork->length + length > mtu - headersize && ipc6->dontfrag &&
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001312 (sk->sk_protocol == IPPROTO_UDP ||
1313 sk->sk_protocol == IPPROTO_RAW)) {
1314 ipv6_local_rxpmtu(sk, fl6, mtu - headersize +
1315 sizeof(struct ipv6hdr));
1316 goto emsgsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317 }
1318
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001319 if (ip6_sk_ignore_df(sk))
1320 maxnonfragsize = sizeof(struct ipv6hdr) + IPV6_MAXPLEN;
1321 else
1322 maxnonfragsize = mtu;
1323
1324 if (cork->length + length > maxnonfragsize - headersize) {
1325emsgsize:
1326 ipv6_local_error(sk, EMSGSIZE, fl6,
1327 mtu - headersize +
1328 sizeof(struct ipv6hdr));
1329 return -EMSGSIZE;
1330 }
1331
1332 /* CHECKSUM_PARTIAL only with no extension headers and when
1333 * we are not going to fragment
1334 */
1335 if (transhdrlen && sk->sk_protocol == IPPROTO_UDP &&
1336 headersize == sizeof(struct ipv6hdr) &&
1337 length < mtu - headersize &&
1338 !(flags & MSG_MORE) &&
Tom Herbertc8cd0982015-12-14 11:19:44 -08001339 rt->dst.dev->features & (NETIF_F_IPV6_CSUM | NETIF_F_HW_CSUM))
Hannes Frederic Sowa682b1a92015-10-27 22:40:41 +01001340 csummode = CHECKSUM_PARTIAL;
1341
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001342 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_RAW) {
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001343 sock_tx_timestamp(sk, sockc->tsflags, &tx_flags);
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001344 if (tx_flags & SKBTX_ANY_SW_TSTAMP &&
1345 sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)
1346 tskey = sk->sk_tskey++;
1347 }
Anders Berggrena693e692011-02-28 12:32:11 -08001348
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349 /*
1350 * Let's try using as much space as possible.
1351 * Use MTU if total length of the message fits into the MTU.
1352 * Otherwise, we need to reserve fragment header and
1353 * fragment alignment (= 8-15 octects, in total).
1354 *
1355 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001356 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357 * the message.
1358 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001359 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 * at once if non-fragmentable extension headers
1361 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001362 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 */
1364
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001365 cork->length += length;
1366 if (((length > mtu) ||
1367 (skb && skb_is_gso(skb))) &&
1368 (sk->sk_protocol == IPPROTO_UDP) &&
Jakub Sitnickif89c56c2016-10-26 11:21:14 +02001369 (rt->dst.dev->features & NETIF_F_UFO) && !rt->dst.header_len &&
Michal Kubeček40ba3302016-01-11 07:50:30 +01001370 (sk->sk_type == SOCK_DGRAM) && !udp_get_no_check6_tx(sk)) {
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001371 err = ip6_ufo_append_data(sk, queue, getfrag, from, length,
Jakub Sitnicki3ba34582016-04-05 18:41:08 +02001372 hh_len, fragheaderlen, exthdrlen,
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001373 transhdrlen, mtu, flags, fl6);
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001374 if (err)
1375 goto error;
1376 return 0;
1377 }
1378
1379 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380 goto alloc_new_skb;
1381
1382 while (length > 0) {
1383 /* Check if the remaining data fits into current packet. */
David S. Millerbdc712b2011-05-06 15:02:07 -07001384 copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 if (copy < length)
1386 copy = maxfraglen - skb->len;
1387
1388 if (copy <= 0) {
1389 char *data;
1390 unsigned int datalen;
1391 unsigned int fraglen;
1392 unsigned int fraggap;
1393 unsigned int alloclen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394alloc_new_skb:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 /* There's no room in the current skb */
Gao feng0c183372012-05-26 01:30:53 +00001396 if (skb)
1397 fraggap = skb->len - maxfraglen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 else
1399 fraggap = 0;
Gao feng0c183372012-05-26 01:30:53 +00001400 /* update mtu and maxfraglen if necessary */
Ian Morris63159f22015-03-29 14:00:04 +01001401 if (!skb || !skb_prev)
Gao feng0c183372012-05-26 01:30:53 +00001402 ip6_append_data_mtu(&mtu, &maxfraglen,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001403 fragheaderlen, skb, rt,
luciene367c2d2014-03-17 12:51:01 +08001404 orig_mtu);
Gao feng0c183372012-05-26 01:30:53 +00001405
1406 skb_prev = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407
1408 /*
1409 * If remaining data exceeds the mtu,
1410 * we know we need more fragment(s).
1411 */
1412 datalen = length + fraggap;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413
Gao feng0c183372012-05-26 01:30:53 +00001414 if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1415 datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416 if ((flags & MSG_MORE) &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001417 !(rt->dst.dev->features&NETIF_F_SG))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418 alloclen = mtu;
1419 else
1420 alloclen = datalen + fragheaderlen;
1421
Steffen Klassert299b0762011-10-11 01:43:33 +00001422 alloclen += dst_exthdrlen;
1423
Gao feng0c183372012-05-26 01:30:53 +00001424 if (datalen != length + fraggap) {
1425 /*
1426 * this is not the last fragment, the trailer
1427 * space is regarded as data space.
1428 */
1429 datalen += rt->dst.trailer_len;
1430 }
1431
1432 alloclen += rt->dst.trailer_len;
1433 fraglen = datalen + fragheaderlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434
1435 /*
1436 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001437 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 * (without MSG_MORE) fits into the MTU.
1439 */
1440 alloclen += sizeof(struct frag_hdr);
1441
1442 if (transhdrlen) {
1443 skb = sock_alloc_send_skb(sk,
1444 alloclen + hh_len,
1445 (flags & MSG_DONTWAIT), &err);
1446 } else {
1447 skb = NULL;
1448 if (atomic_read(&sk->sk_wmem_alloc) <=
1449 2 * sk->sk_sndbuf)
1450 skb = sock_wmalloc(sk,
1451 alloclen + hh_len, 1,
1452 sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001453 if (unlikely(!skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 err = -ENOBUFS;
1455 }
Ian Morris63159f22015-03-29 14:00:04 +01001456 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 goto error;
1458 /*
1459 * Fill in the control structures
1460 */
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001461 skb->protocol = htons(ETH_P_IPV6);
Vlad Yasevich32dce962015-01-31 10:40:18 -05001462 skb->ip_summed = csummode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 skb->csum = 0;
Gao feng1f858512012-03-19 22:36:10 +00001464 /* reserve for fragmentation and ipsec header */
1465 skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
1466 dst_exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467
Willem de Bruijn11878b42014-07-14 17:55:06 -04001468 /* Only the initial fragment is time stamped */
1469 skb_shinfo(skb)->tx_flags = tx_flags;
1470 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001471 skb_shinfo(skb)->tskey = tskey;
1472 tskey = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001473
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 /*
1475 * Find where to start putting bytes
1476 */
Gao feng1f858512012-03-19 22:36:10 +00001477 data = skb_put(skb, fraglen);
1478 skb_set_network_header(skb, exthdrlen);
1479 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001480 skb->transport_header = (skb->network_header +
1481 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 if (fraggap) {
1483 skb->csum = skb_copy_and_csum_bits(
1484 skb_prev, maxfraglen,
1485 data + transhdrlen, fraggap, 0);
1486 skb_prev->csum = csum_sub(skb_prev->csum,
1487 skb->csum);
1488 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001489 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 }
1491 copy = datalen - transhdrlen - fraggap;
Steffen Klassert299b0762011-10-11 01:43:33 +00001492
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 if (copy < 0) {
1494 err = -EINVAL;
1495 kfree_skb(skb);
1496 goto error;
1497 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1498 err = -EFAULT;
1499 kfree_skb(skb);
1500 goto error;
1501 }
1502
1503 offset += copy;
1504 length -= datalen - fraggap;
1505 transhdrlen = 0;
1506 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001507 dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508
1509 /*
1510 * Put the packet on the pending queue
1511 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001512 __skb_queue_tail(queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 continue;
1514 }
1515
1516 if (copy > length)
1517 copy = length;
1518
Changli Gaod8d1f302010-06-10 23:31:35 -07001519 if (!(rt->dst.dev->features&NETIF_F_SG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 unsigned int off;
1521
1522 off = skb->len;
1523 if (getfrag(from, skb_put(skb, copy),
1524 offset, copy, off, skb) < 0) {
1525 __skb_trim(skb, off);
1526 err = -EFAULT;
1527 goto error;
1528 }
1529 } else {
1530 int i = skb_shinfo(skb)->nr_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531
Eric Dumazet5640f762012-09-23 23:04:42 +00001532 err = -ENOMEM;
1533 if (!sk_page_frag_refill(sk, pfrag))
1534 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535
Eric Dumazet5640f762012-09-23 23:04:42 +00001536 if (!skb_can_coalesce(skb, i, pfrag->page,
1537 pfrag->offset)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 err = -EMSGSIZE;
Eric Dumazet5640f762012-09-23 23:04:42 +00001539 if (i == MAX_SKB_FRAGS)
1540 goto error;
1541
1542 __skb_fill_page_desc(skb, i, pfrag->page,
1543 pfrag->offset, 0);
1544 skb_shinfo(skb)->nr_frags = ++i;
1545 get_page(pfrag->page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001547 copy = min_t(int, copy, pfrag->size - pfrag->offset);
Eric Dumazet9e903e02011-10-18 21:00:24 +00001548 if (getfrag(from,
Eric Dumazet5640f762012-09-23 23:04:42 +00001549 page_address(pfrag->page) + pfrag->offset,
1550 offset, copy, skb->len, skb) < 0)
1551 goto error_efault;
1552
1553 pfrag->offset += copy;
1554 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 skb->len += copy;
1556 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001557 skb->truesize += copy;
1558 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559 }
1560 offset += copy;
1561 length -= copy;
1562 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001563
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564 return 0;
Eric Dumazet5640f762012-09-23 23:04:42 +00001565
1566error_efault:
1567 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568error:
David S. Millerbdc712b2011-05-06 15:02:07 -07001569 cork->length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001570 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 return err;
1572}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001573
1574int ip6_append_data(struct sock *sk,
1575 int getfrag(void *from, char *to, int offset, int len,
1576 int odd, struct sk_buff *skb),
Wei Wang26879da2016-05-02 21:40:07 -07001577 void *from, int length, int transhdrlen,
1578 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
1579 struct rt6_info *rt, unsigned int flags,
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001580 const struct sockcm_cookie *sockc)
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001581{
1582 struct inet_sock *inet = inet_sk(sk);
1583 struct ipv6_pinfo *np = inet6_sk(sk);
1584 int exthdrlen;
1585 int err;
1586
1587 if (flags&MSG_PROBE)
1588 return 0;
1589 if (skb_queue_empty(&sk->sk_write_queue)) {
1590 /*
1591 * setup for corking
1592 */
Wei Wang26879da2016-05-02 21:40:07 -07001593 err = ip6_setup_cork(sk, &inet->cork, &np->cork,
1594 ipc6, rt, fl6);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001595 if (err)
1596 return err;
1597
Wei Wang26879da2016-05-02 21:40:07 -07001598 exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001599 length += exthdrlen;
1600 transhdrlen += exthdrlen;
1601 } else {
1602 fl6 = &inet->cork.fl.u.ip6;
1603 transhdrlen = 0;
1604 }
1605
1606 return __ip6_append_data(sk, fl6, &sk->sk_write_queue, &inet->cork.base,
1607 &np->cork, sk_page_frag(sk), getfrag,
Wei Wang26879da2016-05-02 21:40:07 -07001608 from, length, transhdrlen, flags, ipc6, sockc);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001609}
Chris Elstona495f832012-04-29 21:48:53 +00001610EXPORT_SYMBOL_GPL(ip6_append_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001612static void ip6_cork_release(struct inet_cork_full *cork,
1613 struct inet6_cork *v6_cork)
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001614{
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001615 if (v6_cork->opt) {
1616 kfree(v6_cork->opt->dst0opt);
1617 kfree(v6_cork->opt->dst1opt);
1618 kfree(v6_cork->opt->hopopt);
1619 kfree(v6_cork->opt->srcrt);
1620 kfree(v6_cork->opt);
1621 v6_cork->opt = NULL;
Herbert Xu0178b692009-02-05 15:15:50 -08001622 }
1623
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001624 if (cork->base.dst) {
1625 dst_release(cork->base.dst);
1626 cork->base.dst = NULL;
1627 cork->base.flags &= ~IPCORK_ALLFRAG;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001628 }
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001629 memset(&cork->fl, 0, sizeof(cork->fl));
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001630}
1631
Vlad Yasevich64223982015-01-31 10:40:15 -05001632struct sk_buff *__ip6_make_skb(struct sock *sk,
1633 struct sk_buff_head *queue,
1634 struct inet_cork_full *cork,
1635 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636{
1637 struct sk_buff *skb, *tmp_skb;
1638 struct sk_buff **tail_skb;
1639 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001641 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 struct ipv6hdr *hdr;
Vlad Yasevich64223982015-01-31 10:40:15 -05001643 struct ipv6_txoptions *opt = v6_cork->opt;
1644 struct rt6_info *rt = (struct rt6_info *)cork->base.dst;
1645 struct flowi6 *fl6 = &cork->fl.u.ip6;
David S. Miller4c9483b2011-03-12 16:22:43 -05001646 unsigned char proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647
Vlad Yasevich64223982015-01-31 10:40:15 -05001648 skb = __skb_dequeue(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001649 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 goto out;
1651 tail_skb = &(skb_shinfo(skb)->frag_list);
1652
1653 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001654 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001655 __skb_pull(skb, skb_network_offset(skb));
Vlad Yasevich64223982015-01-31 10:40:15 -05001656 while ((tmp_skb = __skb_dequeue(queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001657 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 *tail_skb = tmp_skb;
1659 tail_skb = &(tmp_skb->next);
1660 skb->len += tmp_skb->len;
1661 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 skb->truesize += tmp_skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 tmp_skb->destructor = NULL;
1664 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 }
1666
Herbert Xu28a894532008-02-12 18:07:27 -08001667 /* Allow local fragmentation. */
WANG Cong60ff7462014-05-04 16:39:18 -07001668 skb->ignore_df = ip6_sk_ignore_df(sk);
Herbert Xu28a894532008-02-12 18:07:27 -08001669
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001670 *final_dst = fl6->daddr;
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001671 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672 if (opt && opt->opt_flen)
1673 ipv6_push_frag_opts(skb, opt, &proto);
1674 if (opt && opt->opt_nflen)
1675 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1676
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001677 skb_push(skb, sizeof(struct ipv6hdr));
1678 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001679 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001680
Vlad Yasevich64223982015-01-31 10:40:15 -05001681 ip6_flow_hdr(hdr, v6_cork->tclass,
Tom Herbertcb1ce2e2014-07-01 21:33:10 -07001682 ip6_make_flowlabel(net, skb, fl6->flowlabel,
Tom Herbert67800f92015-07-31 16:52:11 -07001683 np->autoflowlabel, fl6));
Vlad Yasevich64223982015-01-31 10:40:15 -05001684 hdr->hop_limit = v6_cork->hop_limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 hdr->nexthdr = proto;
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001686 hdr->saddr = fl6->saddr;
1687 hdr->daddr = *final_dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688
Patrick McHardya2c20642006-01-08 22:37:26 -08001689 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001690 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001691
Changli Gaod8d1f302010-06-10 23:31:35 -07001692 skb_dst_set(skb, dst_clone(&rt->dst));
Neil Hormanedf391f2009-04-27 02:45:02 -07001693 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
David L Stevens14878f72007-09-16 16:52:35 -07001694 if (proto == IPPROTO_ICMPV6) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001695 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
David L Stevens14878f72007-09-16 16:52:35 -07001696
Hannes Frederic Sowa43a43b62014-03-31 20:14:10 +02001697 ICMP6MSGOUT_INC_STATS(net, idev, icmp6_hdr(skb)->icmp6_type);
1698 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001699 }
1700
Vlad Yasevich64223982015-01-31 10:40:15 -05001701 ip6_cork_release(cork, v6_cork);
1702out:
1703 return skb;
1704}
1705
1706int ip6_send_skb(struct sk_buff *skb)
1707{
1708 struct net *net = sock_net(skb->sk);
1709 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
1710 int err;
1711
Eric W. Biederman33224b12015-10-07 16:48:46 -05001712 err = ip6_local_out(net, skb->sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 if (err) {
1714 if (err > 0)
Eric Dumazet6ce9e7b2009-09-02 18:05:33 -07001715 err = net_xmit_errno(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 if (err)
Vlad Yasevich64223982015-01-31 10:40:15 -05001717 IP6_INC_STATS(net, rt->rt6i_idev,
1718 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 }
1720
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721 return err;
Vlad Yasevich64223982015-01-31 10:40:15 -05001722}
1723
1724int ip6_push_pending_frames(struct sock *sk)
1725{
1726 struct sk_buff *skb;
1727
1728 skb = ip6_finish_skb(sk);
1729 if (!skb)
1730 return 0;
1731
1732 return ip6_send_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733}
Chris Elstona495f832012-04-29 21:48:53 +00001734EXPORT_SYMBOL_GPL(ip6_push_pending_frames);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001736static void __ip6_flush_pending_frames(struct sock *sk,
Vlad Yasevich64223982015-01-31 10:40:15 -05001737 struct sk_buff_head *queue,
1738 struct inet_cork_full *cork,
1739 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 struct sk_buff *skb;
1742
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001743 while ((skb = __skb_dequeue_tail(queue)) != NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001744 if (skb_dst(skb))
1745 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001746 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 kfree_skb(skb);
1748 }
1749
Vlad Yasevich64223982015-01-31 10:40:15 -05001750 ip6_cork_release(cork, v6_cork);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001752
1753void ip6_flush_pending_frames(struct sock *sk)
1754{
Vlad Yasevich64223982015-01-31 10:40:15 -05001755 __ip6_flush_pending_frames(sk, &sk->sk_write_queue,
1756 &inet_sk(sk)->cork, &inet6_sk(sk)->cork);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001757}
Chris Elstona495f832012-04-29 21:48:53 +00001758EXPORT_SYMBOL_GPL(ip6_flush_pending_frames);
Vlad Yasevich64223982015-01-31 10:40:15 -05001759
1760struct sk_buff *ip6_make_skb(struct sock *sk,
1761 int getfrag(void *from, char *to, int offset,
1762 int len, int odd, struct sk_buff *skb),
1763 void *from, int length, int transhdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001764 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
Vlad Yasevich64223982015-01-31 10:40:15 -05001765 struct rt6_info *rt, unsigned int flags,
Wei Wang26879da2016-05-02 21:40:07 -07001766 const struct sockcm_cookie *sockc)
Vlad Yasevich64223982015-01-31 10:40:15 -05001767{
1768 struct inet_cork_full cork;
1769 struct inet6_cork v6_cork;
1770 struct sk_buff_head queue;
Wei Wang26879da2016-05-02 21:40:07 -07001771 int exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0);
Vlad Yasevich64223982015-01-31 10:40:15 -05001772 int err;
1773
1774 if (flags & MSG_PROBE)
1775 return NULL;
1776
1777 __skb_queue_head_init(&queue);
1778
1779 cork.base.flags = 0;
1780 cork.base.addr = 0;
1781 cork.base.opt = NULL;
1782 v6_cork.opt = NULL;
Wei Wang26879da2016-05-02 21:40:07 -07001783 err = ip6_setup_cork(sk, &cork, &v6_cork, ipc6, rt, fl6);
Vlad Yasevich64223982015-01-31 10:40:15 -05001784 if (err)
1785 return ERR_PTR(err);
1786
Wei Wang26879da2016-05-02 21:40:07 -07001787 if (ipc6->dontfrag < 0)
1788 ipc6->dontfrag = inet6_sk(sk)->dontfrag;
Vlad Yasevich64223982015-01-31 10:40:15 -05001789
1790 err = __ip6_append_data(sk, fl6, &queue, &cork.base, &v6_cork,
1791 &current->task_frag, getfrag, from,
1792 length + exthdrlen, transhdrlen + exthdrlen,
Wei Wang26879da2016-05-02 21:40:07 -07001793 flags, ipc6, sockc);
Vlad Yasevich64223982015-01-31 10:40:15 -05001794 if (err) {
1795 __ip6_flush_pending_frames(sk, &queue, &cork, &v6_cork);
1796 return ERR_PTR(err);
1797 }
1798
1799 return __ip6_make_skb(sk, &queue, &cork, &v6_cork);
1800}