blob: 8dddb45c433e53ad35d30cd4263a5a5080cd0da9 [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>
Hannes Frederic Sowab72a2b02015-10-16 11:32:43 +020031#include <linux/overflow-arith.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032#include <linux/string.h>
33#include <linux/socket.h>
34#include <linux/net.h>
35#include <linux/netdevice.h>
36#include <linux/if_arp.h>
37#include <linux/in6.h>
38#include <linux/tcp.h>
39#include <linux/route.h>
Herbert Xub59f45d2006-05-27 23:05:54 -070040#include <linux/module.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090041#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
43#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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070059
David Miller7026b1d2015-04-05 22:19:04 -040060static int ip6_finish_output2(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -070061{
Eric Dumazetadf30902009-06-02 05:19:30 +000062 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 struct net_device *dev = dst->dev;
David S. Millerf6b72b62011-07-14 07:53:20 -070064 struct neighbour *neigh;
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +000065 struct in6_addr *nexthop;
66 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070067
68 skb->protocol = htons(ETH_P_IPV6);
69 skb->dev = dev;
70
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070071 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) {
Eric Dumazetadf30902009-06-02 05:19:30 +000072 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
David Miller7026b1d2015-04-05 22:19:04 -040074 if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(sk) &&
Patrick McHardyd1db2752010-05-11 14:40:55 +020075 ((mroute6_socket(dev_net(dev), skb) &&
Benjamin Therybd91b8b2008-12-10 16:07:08 -080076 !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) ||
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090077 ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
78 &ipv6_hdr(skb)->saddr))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
80
81 /* Do not check for IFF_ALLMULTI; multicast routing
82 is not supported in any case.
83 */
84 if (newskb)
Jan Engelhardtb2e0b382010-03-23 04:09:07 +010085 NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING,
David Miller7026b1d2015-04-05 22:19:04 -040086 sk, newskb, NULL, newskb->dev,
Michel Machado95603e22012-06-12 10:16:35 +000087 dev_loopback_xmit);
Linus Torvalds1da177e2005-04-16 15:20:36 -070088
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -070089 if (ipv6_hdr(skb)->hop_limit == 0) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -070090 IP6_INC_STATS(dev_net(dev), idev,
91 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 kfree_skb(skb);
93 return 0;
94 }
95 }
96
Neil Hormanedf391f2009-04-27 02:45:02 -070097 IP6_UPD_PO_STATS(dev_net(dev), idev, IPSTATS_MIB_OUTMCAST,
98 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
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000108 rcu_read_lock_bh();
Martin KaFai Lau2647a9b2015-05-22 20:55:58 -0700109 nexthop = rt6_nexthop((struct rt6_info *)dst, &ipv6_hdr(skb)->daddr);
YOSHIFUJI Hideaki / 吉藤英明6fd6ce22013-01-17 12:54:00 +0000110 neigh = __ipv6_neigh_lookup_noref(dst->dev, nexthop);
111 if (unlikely(!neigh))
112 neigh = __neigh_create(&nd_tbl, nexthop, dst->dev, false);
113 if (!IS_ERR(neigh)) {
114 ret = dst_neigh_output(dst, neigh, skb);
115 rcu_read_unlock_bh();
116 return ret;
117 }
118 rcu_read_unlock_bh();
David S. Miller05e3aa02011-07-16 17:26:00 -0700119
Hannes Frederic Sowa7f88c6b2013-11-29 06:39:44 +0100120 IP6_INC_STATS(dev_net(dst->dev),
121 ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200122 kfree_skb(skb);
123 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124}
125
David Miller7026b1d2015-04-05 22:19:04 -0400126static int ip6_finish_output(struct sock *sk, struct sk_buff *skb)
Jan Engelhardt9e508492010-04-13 15:28:11 +0200127{
128 if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
Jiri Pirko9037c352013-11-06 17:52:19 +0100129 dst_allfrag(skb_dst(skb)) ||
130 (IP6CB(skb)->frag_max_size && skb->len > IP6CB(skb)->frag_max_size))
David Miller7026b1d2015-04-05 22:19:04 -0400131 return ip6_fragment(sk, skb, ip6_finish_output2);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200132 else
David Miller7026b1d2015-04-05 22:19:04 -0400133 return ip6_finish_output2(sk, skb);
Jan Engelhardt9e508492010-04-13 15:28:11 +0200134}
135
Eric Dumazetaad88722014-04-15 13:47:15 -0400136int ip6_output(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137{
Jan Engelhardt9e508492010-04-13 15:28:11 +0200138 struct net_device *dev = skb_dst(skb)->dev;
Eric Dumazetadf30902009-06-02 05:19:30 +0000139 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900140 if (unlikely(idev->cnf.disable_ipv6)) {
Jan Engelhardt9e508492010-04-13 15:28:11 +0200141 IP6_INC_STATS(dev_net(dev), idev,
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700142 IPSTATS_MIB_OUTDISCARDS);
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900143 kfree_skb(skb);
144 return 0;
145 }
146
David Miller7026b1d2015-04-05 22:19:04 -0400147 return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING, sk, skb,
148 NULL, dev,
Jan Engelhardt9c6eb282010-04-13 15:32:16 +0200149 ip6_finish_output,
150 !(IP6CB(skb)->flags & IP6SKB_REROUTED));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151}
152
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153/*
Shan Weib5d43992010-04-15 16:48:48 +0000154 * xmit an sk_buff (used by TCP, SCTP and DCCP)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 */
156
David S. Miller4c9483b2011-03-12 16:22:43 -0500157int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
Eric Dumazetb903d322011-10-27 00:44:35 -0400158 struct ipv6_txoptions *opt, int tclass)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159{
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700160 struct net *net = sock_net(sk);
Patrick McHardyb30bd282006-03-23 01:17:25 -0800161 struct ipv6_pinfo *np = inet6_sk(sk);
David S. Miller4c9483b2011-03-12 16:22:43 -0500162 struct in6_addr *first_hop = &fl6->daddr;
Eric Dumazetadf30902009-06-02 05:19:30 +0000163 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164 struct ipv6hdr *hdr;
David S. Miller4c9483b2011-03-12 16:22:43 -0500165 u8 proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 int seg_len = skb->len;
Gerrit Renkere651f032009-08-09 08:12:48 +0000167 int hlimit = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 u32 mtu;
169
170 if (opt) {
Chuck Leverc2636b42007-10-23 21:07:32 -0700171 unsigned int head_room;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172
173 /* First: exthdrs may take lots of space (~8K for now)
174 MAX_HEADER is not enough.
175 */
176 head_room = opt->opt_nflen + opt->opt_flen;
177 seg_len += head_room;
178 head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
179
180 if (skb_headroom(skb) < head_room) {
181 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
Ian Morris63159f22015-03-29 14:00:04 +0100182 if (!skb2) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000183 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900184 IPSTATS_MIB_OUTDISCARDS);
185 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 return -ENOBUFS;
187 }
Eric Dumazet808db802012-04-24 10:17:59 +0000188 consume_skb(skb);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900189 skb = skb2;
Dan Carpenter83d7eb22010-04-30 16:42:08 -0700190 skb_set_owner_w(skb, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 }
192 if (opt->opt_flen)
193 ipv6_push_frag_opts(skb, opt, &proto);
194 if (opt->opt_nflen)
195 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
196 }
197
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700198 skb_push(skb, sizeof(struct ipv6hdr));
199 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700200 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201
202 /*
203 * Fill in the IPv6 header
204 */
Eric Dumazetb903d322011-10-27 00:44:35 -0400205 if (np)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206 hlimit = np->hop_limit;
207 if (hlimit < 0)
YOSHIFUJI Hideaki6b75d092008-03-10 06:00:30 -0400208 hlimit = ip6_dst_hoplimit(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209
Tom Herbertcb1ce2e2014-07-01 21:33:10 -0700210 ip6_flow_hdr(hdr, tclass, ip6_make_flowlabel(net, skb, fl6->flowlabel,
Tom Herbert67800f92015-07-31 16:52:11 -0700211 np->autoflowlabel, fl6));
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900212
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 hdr->payload_len = htons(seg_len);
214 hdr->nexthdr = proto;
215 hdr->hop_limit = hlimit;
216
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000217 hdr->saddr = fl6->saddr;
218 hdr->daddr = *first_hop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +0200220 skb->protocol = htons(ETH_P_IPV6);
Patrick McHardya2c20642006-01-08 22:37:26 -0800221 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800222 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -0800223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 mtu = dst_mtu(dst);
WANG Cong60ff7462014-05-04 16:39:18 -0700225 if ((skb->len <= mtu) || skb->ignore_df || skb_is_gso(skb)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000226 IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)),
Neil Hormanedf391f2009-04-27 02:45:02 -0700227 IPSTATS_MIB_OUT, skb->len);
David Miller7026b1d2015-04-05 22:19:04 -0400228 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, sk, skb,
229 NULL, dst->dev, dst_output_sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 }
231
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 skb->dev = dst->dev;
Steffen Klassertf4e53e22013-02-05 01:48:05 +0000233 ipv6_local_error(sk, EMSGSIZE, fl6, mtu);
Eric Dumazetadf30902009-06-02 05:19:30 +0000234 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 kfree_skb(skb);
236 return -EMSGSIZE;
237}
YOSHIFUJI Hideaki71590392007-02-22 22:05:40 +0900238EXPORT_SYMBOL(ip6_xmit);
239
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
241{
242 struct ip6_ra_chain *ra;
243 struct sock *last = NULL;
244
245 read_lock(&ip6_ra_lock);
246 for (ra = ip6_ra_chain; ra; ra = ra->next) {
247 struct sock *sk = ra->sk;
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700248 if (sk && ra->sel == sel &&
249 (!sk->sk_bound_dev_if ||
250 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 if (last) {
252 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
253 if (skb2)
254 rawv6_rcv(last, skb2);
255 }
256 last = sk;
257 }
258 }
259
260 if (last) {
261 rawv6_rcv(last, skb);
262 read_unlock(&ip6_ra_lock);
263 return 1;
264 }
265 read_unlock(&ip6_ra_lock);
266 return 0;
267}
268
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700269static int ip6_forward_proxy_check(struct sk_buff *skb)
270{
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700271 struct ipv6hdr *hdr = ipv6_hdr(skb);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700272 u8 nexthdr = hdr->nexthdr;
Jesse Gross75f28112011-11-30 17:05:51 -0800273 __be16 frag_off;
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700274 int offset;
275
276 if (ipv6_ext_hdr(nexthdr)) {
Jesse Gross75f28112011-11-30 17:05:51 -0800277 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700278 if (offset < 0)
279 return 0;
280 } else
281 offset = sizeof(struct ipv6hdr);
282
283 if (nexthdr == IPPROTO_ICMPV6) {
284 struct icmp6hdr *icmp6;
285
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700286 if (!pskb_may_pull(skb, (skb_network_header(skb) +
287 offset + 1 - skb->data)))
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700288 return 0;
289
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700290 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700291
292 switch (icmp6->icmp6_type) {
293 case NDISC_ROUTER_SOLICITATION:
294 case NDISC_ROUTER_ADVERTISEMENT:
295 case NDISC_NEIGHBOUR_SOLICITATION:
296 case NDISC_NEIGHBOUR_ADVERTISEMENT:
297 case NDISC_REDIRECT:
298 /* For reaction involving unicast neighbor discovery
299 * message destined to the proxied address, pass it to
300 * input function.
301 */
302 return 1;
303 default:
304 break;
305 }
306 }
307
Ville Nuorvala74553b02006-09-22 14:42:18 -0700308 /*
309 * The proxying router can't forward traffic sent to a link-local
310 * address, so signal the sender and discard the packet. This
311 * behavior is clarified by the MIPv6 specification.
312 */
313 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
314 dst_link_failure(skb);
315 return -1;
316 }
317
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700318 return 0;
319}
320
David Miller7026b1d2015-04-05 22:19:04 -0400321static inline int ip6_forward_finish(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322{
Eric Dumazetc29390c2015-03-11 18:42:02 -0700323 skb_sender_cpu_clear(skb);
David Miller7026b1d2015-04-05 22:19:04 -0400324 return dst_output_sk(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325}
326
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100327static unsigned int ip6_dst_mtu_forward(const struct dst_entry *dst)
328{
329 unsigned int mtu;
330 struct inet6_dev *idev;
331
332 if (dst_metric_locked(dst, RTAX_MTU)) {
333 mtu = dst_metric_raw(dst, RTAX_MTU);
334 if (mtu)
335 return mtu;
336 }
337
338 mtu = IPV6_MIN_MTU;
339 rcu_read_lock();
340 idev = __in6_dev_get(dst->dev);
341 if (idev)
342 mtu = idev->cnf.mtu6;
343 rcu_read_unlock();
344
345 return mtu;
346}
347
Florian Westphalfe6cc552014-02-13 23:09:12 +0100348static bool ip6_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
349{
Florian Westphal418a3152014-05-05 00:03:34 +0200350 if (skb->len <= mtu)
Florian Westphalfe6cc552014-02-13 23:09:12 +0100351 return false;
352
WANG Cong60ff7462014-05-04 16:39:18 -0700353 /* ipv6 conntrack defrag sets max_frag_size + ignore_df */
Florian Westphalfe6cc552014-02-13 23:09:12 +0100354 if (IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu)
355 return true;
356
WANG Cong60ff7462014-05-04 16:39:18 -0700357 if (skb->ignore_df)
Florian Westphal418a3152014-05-05 00:03:34 +0200358 return false;
359
Florian Westphalfe6cc552014-02-13 23:09:12 +0100360 if (skb_is_gso(skb) && skb_gso_network_seglen(skb) <= mtu)
361 return false;
362
363 return true;
364}
365
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366int ip6_forward(struct sk_buff *skb)
367{
Eric Dumazetadf30902009-06-02 05:19:30 +0000368 struct dst_entry *dst = skb_dst(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700369 struct ipv6hdr *hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 struct inet6_skb_parm *opt = IP6CB(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900371 struct net *net = dev_net(dst->dev);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800372 u32 mtu;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900373
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700374 if (net->ipv6.devconf_all->forwarding == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375 goto error;
376
Li RongQing090f1162014-03-11 10:40:08 +0800377 if (skb->pkt_type != PACKET_HOST)
378 goto drop;
379
Hannes Frederic Sowa9ef2e962015-10-08 18:19:53 +0200380 if (unlikely(skb->sk))
381 goto drop;
382
Ben Hutchings4497b072008-06-19 16:22:28 -0700383 if (skb_warn_if_lro(skb))
384 goto drop;
385
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800387 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
388 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 goto drop;
390 }
391
Herbert Xu35fc92a2007-03-26 23:22:20 -0700392 skb_forward_csum(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393
394 /*
395 * We DO NOT make any processing on
396 * RA packets, pushing them to user level AS IS
397 * without ane WARRANTY that application will be able
398 * to interpret them. The reason is that we
399 * cannot make anything clever here.
400 *
401 * We are not end-node, so that if packet contains
402 * AH/ESP, we cannot make anything.
403 * Defragmentation also would be mistake, RA packets
404 * cannot be fragmented, because there is no warranty
405 * that different fragments will go along one path. --ANK
406 */
YOSHIFUJI Hideaki / 吉藤英明ab4eb352013-06-22 11:13:13 +0900407 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
408 if (ip6_call_ra_chain(skb, ntohs(opt->ra)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 return 0;
410 }
411
412 /*
413 * check and decrement ttl
414 */
415 if (hdr->hop_limit <= 1) {
416 /* Force OUTPUT device used as source address */
417 skb->dev = dst->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000418 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0);
Eric Dumazet15c77d82013-12-05 21:38:06 -0800419 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
420 IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421
422 kfree_skb(skb);
423 return -ETIMEDOUT;
424 }
425
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -0700426 /* XXX: idev->cnf.proxy_ndp? */
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700427 if (net->ipv6.devconf_all->proxy_ndp &&
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800428 pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
Ville Nuorvala74553b02006-09-22 14:42:18 -0700429 int proxied = ip6_forward_proxy_check(skb);
430 if (proxied > 0)
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700431 return ip6_input(skb);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700432 else if (proxied < 0) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800433 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
434 IPSTATS_MIB_INDISCARDS);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700435 goto drop;
436 }
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700437 }
438
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 if (!xfrm6_route_forward(skb)) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800440 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
441 IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 goto drop;
443 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000444 dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445
446 /* IPv6 specs say nothing about it, but it is clear that we cannot
447 send redirects to source routed frames.
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900448 We don't send redirects to frames decapsulated from IPsec.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 */
David S. Millerc45a3df2012-01-27 15:32:19 -0800450 if (skb->dev == dst->dev && opt->srcrt == 0 && !skb_sec_path(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 struct in6_addr *target = NULL;
David S. Millerfbfe95a2012-06-08 23:24:18 -0700452 struct inet_peer *peer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454
455 /*
456 * incoming and outgoing devices are the same
457 * send a redirect.
458 */
459
460 rt = (struct rt6_info *) dst;
David S. Millerc45a3df2012-01-27 15:32:19 -0800461 if (rt->rt6i_flags & RTF_GATEWAY)
462 target = &rt->rt6i_gateway;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 else
464 target = &hdr->daddr;
465
Martin KaFai Laufd0273d2015-05-22 20:55:57 -0700466 peer = inet_getpeer_v6(net->ipv6.peers, &hdr->daddr, 1);
David S. Miller92d86822011-02-04 15:55:25 -0800467
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 /* Limit redirects both by destination (here)
469 and by source (inside ndisc_send_redirect)
470 */
David S. Millerfbfe95a2012-06-08 23:24:18 -0700471 if (inet_peer_xrlim_allow(peer, 1*HZ))
David S. Miller49919692012-01-27 15:30:48 -0800472 ndisc_send_redirect(skb, target);
David S. Miller1d861aa2012-07-10 03:58:16 -0700473 if (peer)
474 inet_putpeer(peer);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700475 } else {
476 int addrtype = ipv6_addr_type(&hdr->saddr);
477
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 /* This check is security critical. */
YOSHIFUJI Hideakif81b2e72008-06-25 16:55:26 +0900479 if (addrtype == IPV6_ADDR_ANY ||
480 addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK))
David L Stevens5bb1ab02007-05-09 13:53:44 -0700481 goto error;
482 if (addrtype & IPV6_ADDR_LINKLOCAL) {
483 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000484 ICMPV6_NOT_NEIGHBOUR, 0);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700485 goto error;
486 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 }
488
Hannes Frederic Sowa0954cf92014-01-09 10:01:16 +0100489 mtu = ip6_dst_mtu_forward(dst);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800490 if (mtu < IPV6_MIN_MTU)
491 mtu = IPV6_MIN_MTU;
492
Florian Westphalfe6cc552014-02-13 23:09:12 +0100493 if (ip6_pkt_too_big(skb, mtu)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 /* Again, force OUTPUT device used as source address */
495 skb->dev = dst->dev;
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800496 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazet15c77d82013-12-05 21:38:06 -0800497 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
498 IPSTATS_MIB_INTOOBIGERRORS);
499 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
500 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 kfree_skb(skb);
502 return -EMSGSIZE;
503 }
504
505 if (skb_cow(skb, dst->dev->hard_header_len)) {
Eric Dumazet15c77d82013-12-05 21:38:06 -0800506 IP6_INC_STATS_BH(net, ip6_dst_idev(dst),
507 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 goto drop;
509 }
510
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700511 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512
513 /* Mangling hops number delayed to point after skb COW */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900514
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 hdr->hop_limit--;
516
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700517 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
Vincent Bernat2d8dbb02012-06-05 03:41:42 +0000518 IP6_ADD_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len);
David Miller7026b1d2015-04-05 22:19:04 -0400519 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, NULL, skb,
520 skb->dev, dst->dev,
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800521 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522
523error:
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700524 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525drop:
526 kfree_skb(skb);
527 return -EINVAL;
528}
529
530static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
531{
532 to->pkt_type = from->pkt_type;
533 to->priority = from->priority;
534 to->protocol = from->protocol;
Eric Dumazetadf30902009-06-02 05:19:30 +0000535 skb_dst_drop(to);
536 skb_dst_set(to, dst_clone(skb_dst(from)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 to->dev = from->dev;
Thomas Graf82e91ff2006-11-09 15:19:14 -0800538 to->mark = from->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
540#ifdef CONFIG_NET_SCHED
541 to->tc_index = from->tc_index;
542#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -0700543 nf_copy(to, from);
James Morris984bc162006-06-09 00:29:17 -0700544 skb_copy_secmark(to, from);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545}
546
David Miller7026b1d2015-04-05 22:19:04 -0400547int ip6_fragment(struct sock *sk, struct sk_buff *skb,
548 int (*output)(struct sock *, struct sk_buff *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 struct sk_buff *frag;
Ian Morris67ba4152014-08-24 21:53:10 +0100551 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +0200552 struct ipv6_pinfo *np = skb->sk && !dev_recursion_level() ?
553 inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 struct ipv6hdr *tmp_hdr;
555 struct frag_hdr *fh;
556 unsigned int mtu, hlen, left, len;
Herbert Xua7ae1992011-11-18 02:20:04 +0000557 int hroom, troom;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700558 __be32 frag_id;
Ian Morris67ba4152014-08-24 21:53:10 +0100559 int ptr, offset = 0, err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 u8 *prevhdr, nexthdr = 0;
Eric Dumazetadf30902009-06-02 05:19:30 +0000561 struct net *net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 hlen = ip6_find_1stfragopt(skb, &prevhdr);
564 nexthdr = *prevhdr;
565
John Heffner628a5c52007-04-20 15:53:27 -0700566 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700567
568 /* We must not fragment if the socket is set to force MTU discovery
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800569 * or if the skb it not generated by a local socket.
John Heffnerb881ef72007-04-20 15:52:39 -0700570 */
Florian Westphal485fca62015-05-22 00:44:16 +0200571 if (unlikely(!skb->ignore_df && skb->len > mtu))
572 goto fail_toobig;
Eric Dumazeta34a1012012-05-18 21:51:44 +0000573
Florian Westphal485fca62015-05-22 00:44:16 +0200574 if (IP6CB(skb)->frag_max_size) {
575 if (IP6CB(skb)->frag_max_size > mtu)
576 goto fail_toobig;
577
578 /* don't send fragments larger than what we received */
579 mtu = IP6CB(skb)->frag_max_size;
580 if (mtu < IPV6_MIN_MTU)
581 mtu = IPV6_MIN_MTU;
John Heffnerb881ef72007-04-20 15:52:39 -0700582 }
583
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800584 if (np && np->frag_size < mtu) {
585 if (np->frag_size)
586 mtu = np->frag_size;
587 }
Hannes Frederic Sowab72a2b02015-10-16 11:32:43 +0200588
589 if (overflow_usub(mtu, hlen + sizeof(struct frag_hdr), &mtu) ||
590 mtu <= 7)
591 goto fail_toobig;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592
Martin KaFai Laufd0273d2015-05-22 20:55:57 -0700593 frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
594 &ipv6_hdr(skb)->saddr);
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700595
Florian Westphal1d325d22015-09-16 17:26:14 +0200596 hroom = LL_RESERVED_SPACE(rt->dst.dev);
David S. Miller21dc3302010-08-23 00:13:46 -0700597 if (skb_has_frag_list(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 int first_len = skb_pagelen(skb);
Eric Dumazet3d130082010-09-21 08:47:45 +0000599 struct sk_buff *frag2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
601 if (first_len - hlen > mtu ||
602 ((first_len - hlen) & 7) ||
Florian Westphal1d325d22015-09-16 17:26:14 +0200603 skb_cloned(skb) ||
604 skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 goto slow_path;
606
David S. Miller4d9092b2009-06-09 00:20:05 -0700607 skb_walk_frags(skb, frag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 /* Correct geometry. */
609 if (frag->len > mtu ||
610 ((frag->len & 7) && frag->next) ||
Florian Westphal1d325d22015-09-16 17:26:14 +0200611 skb_headroom(frag) < (hlen + hroom + sizeof(struct frag_hdr)))
Eric Dumazet3d130082010-09-21 08:47:45 +0000612 goto slow_path_clean;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 /* Partially cloned skb? */
615 if (skb_shared(frag))
Eric Dumazet3d130082010-09-21 08:47:45 +0000616 goto slow_path_clean;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700617
618 BUG_ON(frag->sk);
619 if (skb->sk) {
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700620 frag->sk = skb->sk;
621 frag->destructor = sock_wfree;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700622 }
Eric Dumazet3d130082010-09-21 08:47:45 +0000623 skb->truesize -= frag->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 }
625
626 err = 0;
627 offset = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 /* BUILD HEADER */
629
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800630 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700631 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 if (!tmp_hdr) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000633 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700634 IPSTATS_MIB_FRAGFAILS);
Florian Westphal1d325d22015-09-16 17:26:14 +0200635 err = -ENOMEM;
636 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 }
Florian Westphal1d325d22015-09-16 17:26:14 +0200638 frag = skb_shinfo(skb)->frag_list;
639 skb_frag_list_init(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 __skb_pull(skb, hlen);
Ian Morris67ba4152014-08-24 21:53:10 +0100642 fh = (struct frag_hdr *)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700643 __skb_push(skb, hlen);
644 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700645 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 fh->nexthdr = nexthdr;
648 fh->reserved = 0;
649 fh->frag_off = htons(IP6_MF);
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700650 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651
652 first_len = skb_pagelen(skb);
653 skb->data_len = first_len - skb_headlen(skb);
654 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700655 ipv6_hdr(skb)->payload_len = htons(first_len -
656 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900657
Changli Gaod8d1f302010-06-10 23:31:35 -0700658 dst_hold(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659
660 for (;;) {
661 /* Prepare header of the next frame,
662 * before previous one went down. */
663 if (frag) {
664 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300665 skb_reset_transport_header(frag);
Ian Morris67ba4152014-08-24 21:53:10 +0100666 fh = (struct frag_hdr *)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700667 __skb_push(frag, hlen);
668 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700669 memcpy(skb_network_header(frag), tmp_hdr,
670 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 offset += skb->len - hlen - sizeof(struct frag_hdr);
672 fh->nexthdr = nexthdr;
673 fh->reserved = 0;
674 fh->frag_off = htons(offset);
Ian Morris53b24b82015-03-29 14:00:05 +0100675 if (frag->next)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 fh->frag_off |= htons(IP6_MF);
677 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700678 ipv6_hdr(frag)->payload_len =
679 htons(frag->len -
680 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 ip6_copy_metadata(frag, skb);
682 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900683
David Miller7026b1d2015-04-05 22:19:04 -0400684 err = output(sk, skb);
Ian Morris67ba4152014-08-24 21:53:10 +0100685 if (!err)
Changli Gaod8d1f302010-06-10 23:31:35 -0700686 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700687 IPSTATS_MIB_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700688
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 if (err || !frag)
690 break;
691
692 skb = frag;
693 frag = skb->next;
694 skb->next = NULL;
695 }
696
Jesper Juhla51482b2005-11-08 09:41:34 -0800697 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698
699 if (err == 0) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700700 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700701 IPSTATS_MIB_FRAGOKS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000702 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 return 0;
704 }
705
Florian Westphal46cfd722014-09-10 01:08:46 +0200706 kfree_skb_list(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707
Changli Gaod8d1f302010-06-10 23:31:35 -0700708 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700709 IPSTATS_MIB_FRAGFAILS);
Amerigo Wang94e187c2012-10-29 00:13:19 +0000710 ip6_rt_put(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 return err;
Eric Dumazet3d130082010-09-21 08:47:45 +0000712
713slow_path_clean:
714 skb_walk_frags(skb, frag2) {
715 if (frag2 == frag)
716 break;
717 frag2->sk = NULL;
718 frag2->destructor = NULL;
719 skb->truesize += frag2->truesize;
720 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 }
722
723slow_path:
Eric Dumazet72e843b2012-05-18 11:02:15 +0000724 if ((skb->ip_summed == CHECKSUM_PARTIAL) &&
725 skb_checksum_help(skb))
726 goto fail;
727
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 left = skb->len - hlen; /* Space per frame */
729 ptr = hlen; /* Where to start from */
730
731 /*
732 * Fragment the datagram.
733 */
734
735 *prevhdr = NEXTHDR_FRAGMENT;
Herbert Xua7ae1992011-11-18 02:20:04 +0000736 troom = rt->dst.dev->needed_tailroom;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737
738 /*
739 * Keep copying data until we run out.
740 */
Ian Morris67ba4152014-08-24 21:53:10 +0100741 while (left > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 len = left;
743 /* IF: it doesn't fit, use 'mtu' - the data space left */
744 if (len > mtu)
745 len = mtu;
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300746 /* IF: we are not sending up to and including the packet end
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 then align the next start on an eight byte boundary */
748 if (len < left) {
749 len &= ~7;
750 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751
Joe Perchescbffccc2014-11-05 14:39:21 -0800752 /* Allocate buffer */
753 frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) +
754 hroom + troom, GFP_ATOMIC);
755 if (!frag) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000756 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900757 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 err = -ENOMEM;
759 goto fail;
760 }
761
762 /*
763 * Set up data on packet
764 */
765
766 ip6_copy_metadata(frag, skb);
Herbert Xua7ae1992011-11-18 02:20:04 +0000767 skb_reserve(frag, hroom);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700769 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300770 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700771 frag->transport_header = (frag->network_header + hlen +
772 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
774 /*
775 * Charge the memory for the fragment to any owner
776 * it might possess
777 */
778 if (skb->sk)
779 skb_set_owner_w(frag, skb->sk);
780
781 /*
782 * Copy the packet header into the new buffer.
783 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300784 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785
786 /*
787 * Build fragment header.
788 */
789 fh->nexthdr = nexthdr;
790 fh->reserved = 0;
Martin KaFai Lau286c2342015-05-22 20:55:56 -0700791 fh->identification = frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
793 /*
794 * Copy a block of the IP datagram.
795 */
Himangi Saraogie3f0b862014-07-12 01:57:17 +0530796 BUG_ON(skb_copy_bits(skb, ptr, skb_transport_header(frag),
797 len));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 left -= len;
799
800 fh->frag_off = htons(offset);
801 if (left > 0)
802 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700803 ipv6_hdr(frag)->payload_len = htons(frag->len -
804 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805
806 ptr += len;
807 offset += len;
808
809 /*
810 * Put this fragment into the sending queue.
811 */
David Miller7026b1d2015-04-05 22:19:04 -0400812 err = output(sk, frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 if (err)
814 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700815
Eric Dumazetadf30902009-06-02 05:19:30 +0000816 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700817 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000819 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900820 IPSTATS_MIB_FRAGOKS);
Eric Dumazet808db802012-04-24 10:17:59 +0000821 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 return err;
823
Florian Westphal485fca62015-05-22 00:44:16 +0200824fail_toobig:
825 if (skb->sk && dst_allfrag(skb_dst(skb)))
826 sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK);
827
828 skb->dev = skb_dst(skb)->dev;
829 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
830 err = -EMSGSIZE;
831
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832fail:
Eric Dumazetadf30902009-06-02 05:19:30 +0000833 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900834 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900835 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 return err;
837}
838
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000839static inline int ip6_rt_check(const struct rt6key *rt_key,
840 const struct in6_addr *fl_addr,
841 const struct in6_addr *addr_cache)
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700842{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000843 return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
Ian Morris63159f22015-03-29 14:00:04 +0100844 (!addr_cache || !ipv6_addr_equal(fl_addr, addr_cache));
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700845}
846
Herbert Xu497c6152006-07-30 20:19:33 -0700847static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
848 struct dst_entry *dst,
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000849 const struct flowi6 *fl6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850{
Herbert Xu497c6152006-07-30 20:19:33 -0700851 struct ipv6_pinfo *np = inet6_sk(sk);
Eric Dumazeta963a372013-06-26 04:15:07 -0700852 struct rt6_info *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853
Herbert Xu497c6152006-07-30 20:19:33 -0700854 if (!dst)
855 goto out;
856
Eric Dumazeta963a372013-06-26 04:15:07 -0700857 if (dst->ops->family != AF_INET6) {
858 dst_release(dst);
859 return NULL;
860 }
861
862 rt = (struct rt6_info *)dst;
Herbert Xu497c6152006-07-30 20:19:33 -0700863 /* Yes, checking route validity in not connected
864 * case is not very simple. Take into account,
865 * that we do not support routing by source, TOS,
Ian Morris67ba4152014-08-24 21:53:10 +0100866 * and MSG_DONTROUTE --ANK (980726)
Herbert Xu497c6152006-07-30 20:19:33 -0700867 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700868 * 1. ip6_rt_check(): If route was host route,
869 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700870 * If it is network route, we still may
871 * check its validity using saved pointer
872 * to the last used address: daddr_cache.
873 * We do not want to save whole address now,
874 * (because main consumer of this service
875 * is tcp, which has not this problem),
876 * so that the last trick works only on connected
877 * sockets.
878 * 2. oif also should be the same.
879 */
David S. Miller4c9483b2011-03-12 16:22:43 -0500880 if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700881#ifdef CONFIG_IPV6_SUBTREES
David S. Miller4c9483b2011-03-12 16:22:43 -0500882 ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700883#endif
David Ahernf1900fb2015-10-19 08:26:05 -0700884 (!(fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF) &&
885 (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex))) {
Herbert Xu497c6152006-07-30 20:19:33 -0700886 dst_release(dst);
887 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 }
889
Herbert Xu497c6152006-07-30 20:19:33 -0700890out:
891 return dst;
892}
893
Roopa Prabhu343d60a2015-07-30 13:34:53 -0700894static int ip6_dst_lookup_tail(struct net *net, struct sock *sk,
David S. Miller4c9483b2011-03-12 16:22:43 -0500895 struct dst_entry **dst, struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -0700896{
David S. Miller69cce1d2011-07-17 23:09:49 -0700897#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
898 struct neighbour *n;
David S. Miller97cac082012-07-02 22:43:47 -0700899 struct rt6_info *rt;
David S. Miller69cce1d2011-07-17 23:09:49 -0700900#endif
901 int err;
Herbert Xu497c6152006-07-30 20:19:33 -0700902
Markus Stenberge16e8882015-05-05 13:36:59 +0300903 /* The correct way to handle this would be to do
904 * ip6_route_get_saddr, and then ip6_route_output; however,
905 * the route-specific preferred source forces the
906 * ip6_route_output call _before_ ip6_route_get_saddr.
907 *
908 * In source specific routing (no src=any default route),
909 * ip6_route_output will fail given src=any saddr, though, so
910 * that's why we try it again later.
911 */
912 if (ipv6_addr_any(&fl6->saddr) && (!*dst || !(*dst)->error)) {
913 struct rt6_info *rt;
914 bool had_dst = *dst != NULL;
915
916 if (!had_dst)
917 *dst = ip6_route_output(net, sk, fl6);
918 rt = (*dst)->error ? NULL : (struct rt6_info *)*dst;
919 err = ip6_route_get_saddr(net, rt, &fl6->daddr,
920 sk ? inet6_sk(sk)->srcprefs : 0,
921 &fl6->saddr);
922 if (err)
923 goto out_err_release;
924
925 /* If we had an erroneous initial result, pretend it
926 * never existed and let the SA-enabled version take
927 * over.
928 */
929 if (!had_dst && (*dst)->error) {
930 dst_release(*dst);
931 *dst = NULL;
932 }
933 }
934
Ian Morris63159f22015-03-29 14:00:04 +0100935 if (!*dst)
David S. Miller4c9483b2011-03-12 16:22:43 -0500936 *dst = ip6_route_output(net, sk, fl6);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937
Ian Morrise5d08d72014-11-23 21:28:43 +0000938 err = (*dst)->error;
939 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 goto out_err_release;
941
Neil Horman95c385b2007-04-25 17:08:10 -0700942#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700943 /*
944 * Here if the dst entry we've looked up
945 * has a neighbour entry that is in the INCOMPLETE
946 * state and the src address from the flow is
947 * marked as OPTIMISTIC, we release the found
948 * dst entry and replace it instead with the
949 * dst entry of the nexthop router
950 */
Eric Dumazetc56bf6f2012-07-06 09:19:05 +0200951 rt = (struct rt6_info *) *dst;
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000952 rcu_read_lock_bh();
Martin KaFai Lau2647a9b2015-05-22 20:55:58 -0700953 n = __ipv6_neigh_lookup_noref(rt->dst.dev,
954 rt6_nexthop(rt, &fl6->daddr));
YOSHIFUJI Hideaki / 吉藤英明707be1f2013-01-17 12:53:55 +0000955 err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0;
956 rcu_read_unlock_bh();
957
958 if (err) {
Neil Hormane550dfb2008-09-09 13:51:35 -0700959 struct inet6_ifaddr *ifp;
David S. Miller4c9483b2011-03-12 16:22:43 -0500960 struct flowi6 fl_gw6;
Neil Hormane550dfb2008-09-09 13:51:35 -0700961 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -0700962
David S. Miller4c9483b2011-03-12 16:22:43 -0500963 ifp = ipv6_get_ifaddr(net, &fl6->saddr,
Neil Hormane550dfb2008-09-09 13:51:35 -0700964 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -0700965
Neil Hormane550dfb2008-09-09 13:51:35 -0700966 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
967 if (ifp)
968 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -0700969
Neil Hormane550dfb2008-09-09 13:51:35 -0700970 if (redirect) {
971 /*
972 * We need to get the dst entry for the
973 * default router instead
974 */
975 dst_release(*dst);
David S. Miller4c9483b2011-03-12 16:22:43 -0500976 memcpy(&fl_gw6, fl6, sizeof(struct flowi6));
977 memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr));
978 *dst = ip6_route_output(net, sk, &fl_gw6);
Ian Morrise5d08d72014-11-23 21:28:43 +0000979 err = (*dst)->error;
980 if (err)
Neil Hormane550dfb2008-09-09 13:51:35 -0700981 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -0700982 }
Neil Hormane550dfb2008-09-09 13:51:35 -0700983 }
Neil Horman95c385b2007-04-25 17:08:10 -0700984#endif
985
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 return 0;
987
988out_err_release:
Mitsuru Chinenca46f9c2007-12-05 22:31:47 -0800989 if (err == -ENETUNREACH)
John Stultz5ac68e72013-10-07 15:52:01 -0700990 IP6_INC_STATS(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 dst_release(*dst);
992 *dst = NULL;
993 return err;
994}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -0800995
Herbert Xu497c6152006-07-30 20:19:33 -0700996/**
997 * ip6_dst_lookup - perform route lookup on flow
998 * @sk: socket which provides route info
999 * @dst: pointer to dst_entry * for result
David S. Miller4c9483b2011-03-12 16:22:43 -05001000 * @fl6: flow to lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001001 *
1002 * This function performs a route lookup on the given flow.
1003 *
1004 * It returns zero on success, or a standard errno code on error.
1005 */
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001006int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1007 struct flowi6 *fl6)
Herbert Xu497c6152006-07-30 20:19:33 -07001008{
1009 *dst = NULL;
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001010 return ip6_dst_lookup_tail(net, sk, dst, fl6);
Herbert Xu497c6152006-07-30 20:19:33 -07001011}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -08001012EXPORT_SYMBOL_GPL(ip6_dst_lookup);
1013
Herbert Xu497c6152006-07-30 20:19:33 -07001014/**
David S. Miller68d0c6d2011-03-01 13:19:07 -08001015 * ip6_dst_lookup_flow - perform route lookup on flow with ipsec
1016 * @sk: socket which provides route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001017 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001018 * @final_dst: final destination address for ipsec lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001019 *
1020 * This function performs a route lookup on the given flow.
1021 *
1022 * It returns a valid dst pointer on success, or a pointer encoded
1023 * error code.
1024 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001025struct dst_entry *ip6_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001026 const struct in6_addr *final_dst)
David S. Miller68d0c6d2011-03-01 13:19:07 -08001027{
1028 struct dst_entry *dst = NULL;
1029 int err;
1030
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001031 err = ip6_dst_lookup_tail(sock_net(sk), sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001032 if (err)
1033 return ERR_PTR(err);
1034 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001035 fl6->daddr = *final_dst;
Phil Suttera0a9f332015-07-15 21:56:26 +02001036 if (!fl6->flowi6_oif)
1037 fl6->flowi6_oif = dst->dev->ifindex;
David S. Miller2774c132011-03-01 14:59:04 -08001038
Steffen Klassertf92ee612014-09-16 10:08:40 +02001039 return xfrm_lookup_route(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001040}
1041EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow);
1042
1043/**
1044 * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow
1045 * @sk: socket which provides the dst cache and route info
David S. Miller4c9483b2011-03-12 16:22:43 -05001046 * @fl6: flow to lookup
David S. Miller68d0c6d2011-03-01 13:19:07 -08001047 * @final_dst: final destination address for ipsec lookup
Herbert Xu497c6152006-07-30 20:19:33 -07001048 *
1049 * This function performs a route lookup on the given flow with the
1050 * possibility of using the cached route in the socket if it is valid.
1051 * It will take the socket dst lock when operating on the dst cache.
1052 * As a result, this function can only be used in process context.
1053 *
David S. Miller68d0c6d2011-03-01 13:19:07 -08001054 * It returns a valid dst pointer on success, or a pointer encoded
1055 * error code.
Herbert Xu497c6152006-07-30 20:19:33 -07001056 */
David S. Miller4c9483b2011-03-12 16:22:43 -05001057struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001058 const struct in6_addr *final_dst)
Herbert Xu497c6152006-07-30 20:19:33 -07001059{
David S. Miller68d0c6d2011-03-01 13:19:07 -08001060 struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
1061 int err;
Herbert Xu497c6152006-07-30 20:19:33 -07001062
David S. Miller4c9483b2011-03-12 16:22:43 -05001063 dst = ip6_sk_dst_check(sk, dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001064
Roopa Prabhu343d60a2015-07-30 13:34:53 -07001065 err = ip6_dst_lookup_tail(sock_net(sk), sk, &dst, fl6);
David S. Miller68d0c6d2011-03-01 13:19:07 -08001066 if (err)
1067 return ERR_PTR(err);
1068 if (final_dst)
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001069 fl6->daddr = *final_dst;
David S. Miller2774c132011-03-01 14:59:04 -08001070
Steffen Klassertf92ee612014-09-16 10:08:40 +02001071 return xfrm_lookup_route(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
Herbert Xu497c6152006-07-30 20:19:33 -07001072}
David S. Miller68d0c6d2011-03-01 13:19:07 -08001073EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
Herbert Xu497c6152006-07-30 20:19:33 -07001074
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001075static inline int ip6_ufo_append_data(struct sock *sk,
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001076 struct sk_buff_head *queue,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001077 int getfrag(void *from, char *to, int offset, int len,
1078 int odd, struct sk_buff *skb),
1079 void *from, int length, int hh_len, int fragheaderlen,
Ian Morris67ba4152014-08-24 21:53:10 +01001080 int transhdrlen, int mtu, unsigned int flags,
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001081 const struct flowi6 *fl6)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001082
1083{
1084 struct sk_buff *skb;
1085 int err;
1086
1087 /* There is support for UDP large send offload by network
1088 * device, so create one single skb packet containing complete
1089 * udp datagram
1090 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001091 skb = skb_peek_tail(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001092 if (!skb) {
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001093 skb = sock_alloc_send_skb(sk,
1094 hh_len + fragheaderlen + transhdrlen + 20,
1095 (flags & MSG_DONTWAIT), &err);
Ian Morris63159f22015-03-29 14:00:04 +01001096 if (!skb)
Zheng Yan504744e2011-10-28 00:26:00 -04001097 return err;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001098
1099 /* reserve space for Hardware header */
1100 skb_reserve(skb, hh_len);
1101
1102 /* create space for UDP/IP header */
Ian Morris67ba4152014-08-24 21:53:10 +01001103 skb_put(skb, fragheaderlen + transhdrlen);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001104
1105 /* initialize network header pointer */
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001106 skb_reset_network_header(skb);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001107
1108 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001109 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001110
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001111 skb->protocol = htons(ETH_P_IPV6);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001112 skb->csum = 0;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001113
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001114 __skb_queue_tail(queue, skb);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001115 } else if (skb_is_gso(skb)) {
1116 goto append;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001117 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001118
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001119 skb->ip_summed = CHECKSUM_PARTIAL;
1120 /* Specify the length of each IPv6 datagram fragment.
1121 * It has to be a multiple of 8.
1122 */
1123 skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
1124 sizeof(struct frag_hdr)) & ~7;
1125 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001126 skb_shinfo(skb)->ip6_frag_id = ipv6_select_ident(sock_net(sk),
1127 &fl6->daddr,
1128 &fl6->saddr);
Jiri Pirkoc547dbf2013-10-19 12:29:16 +02001129
1130append:
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001131 return skb_append_datato_frags(sk, skb, getfrag, from,
1132 (length - transhdrlen));
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001133}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134
Herbert Xu0178b692009-02-05 15:15:50 -08001135static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
1136 gfp_t gfp)
1137{
1138 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1139}
1140
1141static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
1142 gfp_t gfp)
1143{
1144 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1145}
1146
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001147static void ip6_append_data_mtu(unsigned int *mtu,
Gao feng0c183372012-05-26 01:30:53 +00001148 int *maxfraglen,
1149 unsigned int fragheaderlen,
1150 struct sk_buff *skb,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001151 struct rt6_info *rt,
luciene367c2d2014-03-17 12:51:01 +08001152 unsigned int orig_mtu)
Gao feng0c183372012-05-26 01:30:53 +00001153{
1154 if (!(rt->dst.flags & DST_XFRM_TUNNEL)) {
Ian Morris63159f22015-03-29 14:00:04 +01001155 if (!skb) {
Gao feng0c183372012-05-26 01:30:53 +00001156 /* first fragment, reserve header_len */
luciene367c2d2014-03-17 12:51:01 +08001157 *mtu = orig_mtu - rt->dst.header_len;
Gao feng0c183372012-05-26 01:30:53 +00001158
1159 } else {
1160 /*
1161 * this fragment is not first, the headers
1162 * space is regarded as data space.
1163 */
luciene367c2d2014-03-17 12:51:01 +08001164 *mtu = orig_mtu;
Gao feng0c183372012-05-26 01:30:53 +00001165 }
1166 *maxfraglen = ((*mtu - fragheaderlen) & ~7)
1167 + fragheaderlen - sizeof(struct frag_hdr);
1168 }
1169}
1170
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001171static int ip6_setup_cork(struct sock *sk, struct inet_cork_full *cork,
1172 struct inet6_cork *v6_cork,
1173 int hlimit, int tclass, struct ipv6_txoptions *opt,
1174 struct rt6_info *rt, struct flowi6 *fl6)
1175{
1176 struct ipv6_pinfo *np = inet6_sk(sk);
1177 unsigned int mtu;
1178
1179 /*
1180 * setup for corking
1181 */
1182 if (opt) {
1183 if (WARN_ON(v6_cork->opt))
1184 return -EINVAL;
1185
1186 v6_cork->opt = kzalloc(opt->tot_len, sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001187 if (unlikely(!v6_cork->opt))
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001188 return -ENOBUFS;
1189
1190 v6_cork->opt->tot_len = opt->tot_len;
1191 v6_cork->opt->opt_flen = opt->opt_flen;
1192 v6_cork->opt->opt_nflen = opt->opt_nflen;
1193
1194 v6_cork->opt->dst0opt = ip6_opt_dup(opt->dst0opt,
1195 sk->sk_allocation);
1196 if (opt->dst0opt && !v6_cork->opt->dst0opt)
1197 return -ENOBUFS;
1198
1199 v6_cork->opt->dst1opt = ip6_opt_dup(opt->dst1opt,
1200 sk->sk_allocation);
1201 if (opt->dst1opt && !v6_cork->opt->dst1opt)
1202 return -ENOBUFS;
1203
1204 v6_cork->opt->hopopt = ip6_opt_dup(opt->hopopt,
1205 sk->sk_allocation);
1206 if (opt->hopopt && !v6_cork->opt->hopopt)
1207 return -ENOBUFS;
1208
1209 v6_cork->opt->srcrt = ip6_rthdr_dup(opt->srcrt,
1210 sk->sk_allocation);
1211 if (opt->srcrt && !v6_cork->opt->srcrt)
1212 return -ENOBUFS;
1213
1214 /* need source address above miyazawa*/
1215 }
1216 dst_hold(&rt->dst);
1217 cork->base.dst = &rt->dst;
1218 cork->fl.u.ip6 = *fl6;
1219 v6_cork->hop_limit = hlimit;
1220 v6_cork->tclass = tclass;
1221 if (rt->dst.flags & DST_XFRM_TUNNEL)
1222 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
1223 rt->dst.dev->mtu : dst_mtu(&rt->dst);
1224 else
1225 mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ?
1226 rt->dst.dev->mtu : dst_mtu(rt->dst.path);
1227 if (np->frag_size < mtu) {
1228 if (np->frag_size)
1229 mtu = np->frag_size;
1230 }
1231 cork->base.fragsize = mtu;
1232 if (dst_allfrag(rt->dst.path))
1233 cork->base.flags |= IPCORK_ALLFRAG;
1234 cork->base.length = 0;
1235
1236 return 0;
1237}
1238
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001239static int __ip6_append_data(struct sock *sk,
1240 struct flowi6 *fl6,
1241 struct sk_buff_head *queue,
1242 struct inet_cork *cork,
1243 struct inet6_cork *v6_cork,
1244 struct page_frag *pfrag,
1245 int getfrag(void *from, char *to, int offset,
1246 int len, int odd, struct sk_buff *skb),
1247 void *from, int length, int transhdrlen,
1248 unsigned int flags, int dontfrag)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249{
Gao feng0c183372012-05-26 01:30:53 +00001250 struct sk_buff *skb, *skb_prev = NULL;
luciene367c2d2014-03-17 12:51:01 +08001251 unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001252 int exthdrlen = 0;
1253 int dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 int hh_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 int copy;
1256 int err;
1257 int offset = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001258 __u8 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001259 u32 tskey = 0;
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001260 struct rt6_info *rt = (struct rt6_info *)cork->dst;
1261 struct ipv6_txoptions *opt = v6_cork->opt;
Vlad Yasevich32dce962015-01-31 10:40:18 -05001262 int csummode = CHECKSUM_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001264 skb = skb_peek_tail(queue);
1265 if (!skb) {
1266 exthdrlen = opt ? opt->opt_flen : 0;
Romain KUNTZ7efdba5b2013-01-16 12:47:40 +00001267 dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 }
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001269
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001270 mtu = cork->fragsize;
luciene367c2d2014-03-17 12:51:01 +08001271 orig_mtu = mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272
Changli Gaod8d1f302010-06-10 23:31:35 -07001273 hh_len = LL_RESERVED_SPACE(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001275 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001276 (opt ? opt->opt_nflen : 0);
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001277 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen -
1278 sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279
1280 if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001281 unsigned int maxnonfragsize, headersize;
1282
1283 headersize = sizeof(struct ipv6hdr) +
Hannes Frederic Sowa3a1cebe2014-05-11 23:01:13 +02001284 (opt ? opt->opt_flen + opt->opt_nflen : 0) +
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001285 (dst_allfrag(&rt->dst) ?
1286 sizeof(struct frag_hdr) : 0) +
1287 rt->rt6i_nfheader_len;
1288
WANG Cong60ff7462014-05-04 16:39:18 -07001289 if (ip6_sk_ignore_df(sk))
Hannes Frederic Sowa0b952272014-02-26 01:20:43 +01001290 maxnonfragsize = sizeof(struct ipv6hdr) + IPV6_MAXPLEN;
1291 else
1292 maxnonfragsize = mtu;
Hannes Frederic Sowa4df98e72013-12-16 12:36:44 +01001293
1294 /* dontfrag active */
1295 if ((cork->length + length > mtu - headersize) && dontfrag &&
1296 (sk->sk_protocol == IPPROTO_UDP ||
1297 sk->sk_protocol == IPPROTO_RAW)) {
1298 ipv6_local_rxpmtu(sk, fl6, mtu - headersize +
1299 sizeof(struct ipv6hdr));
1300 goto emsgsize;
1301 }
1302
1303 if (cork->length + length > maxnonfragsize - headersize) {
1304emsgsize:
1305 ipv6_local_error(sk, EMSGSIZE, fl6,
1306 mtu - headersize +
1307 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308 return -EMSGSIZE;
1309 }
1310 }
1311
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001312 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_RAW) {
Daniel Borkmannbf84a012013-04-14 08:08:13 +00001313 sock_tx_timestamp(sk, &tx_flags);
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001314 if (tx_flags & SKBTX_ANY_SW_TSTAMP &&
1315 sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)
1316 tskey = sk->sk_tskey++;
1317 }
Anders Berggrena693e692011-02-28 12:32:11 -08001318
Vlad Yasevich32dce962015-01-31 10:40:18 -05001319 /* If this is the first and only packet and device
1320 * supports checksum offloading, let's use it.
Vlad Yaseviche87a4682015-05-14 20:34:08 -04001321 * Use transhdrlen, same as IPv4, because partial
1322 * sums only work when transhdrlen is set.
Vlad Yasevich32dce962015-01-31 10:40:18 -05001323 */
Vlad Yaseviche87a4682015-05-14 20:34:08 -04001324 if (transhdrlen && sk->sk_protocol == IPPROTO_UDP &&
Vlad Yasevich32dce962015-01-31 10:40:18 -05001325 length + fragheaderlen < mtu &&
1326 rt->dst.dev->features & NETIF_F_V6_CSUM &&
1327 !exthdrlen)
1328 csummode = CHECKSUM_PARTIAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 /*
1330 * Let's try using as much space as possible.
1331 * Use MTU if total length of the message fits into the MTU.
1332 * Otherwise, we need to reserve fragment header and
1333 * fragment alignment (= 8-15 octects, in total).
1334 *
1335 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001336 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 * the message.
1338 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001339 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340 * at once if non-fragmentable extension headers
1341 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001342 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 */
1344
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001345 cork->length += length;
1346 if (((length > mtu) ||
1347 (skb && skb_is_gso(skb))) &&
1348 (sk->sk_protocol == IPPROTO_UDP) &&
Michal Kubečekacf8dd02015-03-02 18:27:11 +01001349 (rt->dst.dev->features & NETIF_F_UFO) &&
1350 (sk->sk_type == SOCK_DGRAM)) {
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001351 err = ip6_ufo_append_data(sk, queue, getfrag, from, length,
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001352 hh_len, fragheaderlen,
Martin KaFai Laufd0273d2015-05-22 20:55:57 -07001353 transhdrlen, mtu, flags, fl6);
Hannes Frederic Sowa2811eba2013-09-21 06:27:00 +02001354 if (err)
1355 goto error;
1356 return 0;
1357 }
1358
1359 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 goto alloc_new_skb;
1361
1362 while (length > 0) {
1363 /* Check if the remaining data fits into current packet. */
David S. Millerbdc712b2011-05-06 15:02:07 -07001364 copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 if (copy < length)
1366 copy = maxfraglen - skb->len;
1367
1368 if (copy <= 0) {
1369 char *data;
1370 unsigned int datalen;
1371 unsigned int fraglen;
1372 unsigned int fraggap;
1373 unsigned int alloclen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374alloc_new_skb:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 /* There's no room in the current skb */
Gao feng0c183372012-05-26 01:30:53 +00001376 if (skb)
1377 fraggap = skb->len - maxfraglen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 else
1379 fraggap = 0;
Gao feng0c183372012-05-26 01:30:53 +00001380 /* update mtu and maxfraglen if necessary */
Ian Morris63159f22015-03-29 14:00:04 +01001381 if (!skb || !skb_prev)
Gao feng0c183372012-05-26 01:30:53 +00001382 ip6_append_data_mtu(&mtu, &maxfraglen,
Hannes Frederic Sowa75a493e2013-07-02 08:04:05 +02001383 fragheaderlen, skb, rt,
luciene367c2d2014-03-17 12:51:01 +08001384 orig_mtu);
Gao feng0c183372012-05-26 01:30:53 +00001385
1386 skb_prev = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387
1388 /*
1389 * If remaining data exceeds the mtu,
1390 * we know we need more fragment(s).
1391 */
1392 datalen = length + fraggap;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393
Gao feng0c183372012-05-26 01:30:53 +00001394 if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1395 datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 if ((flags & MSG_MORE) &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001397 !(rt->dst.dev->features&NETIF_F_SG))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 alloclen = mtu;
1399 else
1400 alloclen = datalen + fragheaderlen;
1401
Steffen Klassert299b0762011-10-11 01:43:33 +00001402 alloclen += dst_exthdrlen;
1403
Gao feng0c183372012-05-26 01:30:53 +00001404 if (datalen != length + fraggap) {
1405 /*
1406 * this is not the last fragment, the trailer
1407 * space is regarded as data space.
1408 */
1409 datalen += rt->dst.trailer_len;
1410 }
1411
1412 alloclen += rt->dst.trailer_len;
1413 fraglen = datalen + fragheaderlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414
1415 /*
1416 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001417 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418 * (without MSG_MORE) fits into the MTU.
1419 */
1420 alloclen += sizeof(struct frag_hdr);
1421
1422 if (transhdrlen) {
1423 skb = sock_alloc_send_skb(sk,
1424 alloclen + hh_len,
1425 (flags & MSG_DONTWAIT), &err);
1426 } else {
1427 skb = NULL;
1428 if (atomic_read(&sk->sk_wmem_alloc) <=
1429 2 * sk->sk_sndbuf)
1430 skb = sock_wmalloc(sk,
1431 alloclen + hh_len, 1,
1432 sk->sk_allocation);
Ian Morris63159f22015-03-29 14:00:04 +01001433 if (unlikely(!skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434 err = -ENOBUFS;
1435 }
Ian Morris63159f22015-03-29 14:00:04 +01001436 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 goto error;
1438 /*
1439 * Fill in the control structures
1440 */
Hannes Frederic Sowa9c9c9ad2013-08-26 12:31:23 +02001441 skb->protocol = htons(ETH_P_IPV6);
Vlad Yasevich32dce962015-01-31 10:40:18 -05001442 skb->ip_summed = csummode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 skb->csum = 0;
Gao feng1f858512012-03-19 22:36:10 +00001444 /* reserve for fragmentation and ipsec header */
1445 skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
1446 dst_exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447
Willem de Bruijn11878b42014-07-14 17:55:06 -04001448 /* Only the initial fragment is time stamped */
1449 skb_shinfo(skb)->tx_flags = tx_flags;
1450 tx_flags = 0;
Willem de Bruijn09c2d252014-08-04 22:11:47 -04001451 skb_shinfo(skb)->tskey = tskey;
1452 tskey = 0;
Anders Berggrena693e692011-02-28 12:32:11 -08001453
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 /*
1455 * Find where to start putting bytes
1456 */
Gao feng1f858512012-03-19 22:36:10 +00001457 data = skb_put(skb, fraglen);
1458 skb_set_network_header(skb, exthdrlen);
1459 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001460 skb->transport_header = (skb->network_header +
1461 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 if (fraggap) {
1463 skb->csum = skb_copy_and_csum_bits(
1464 skb_prev, maxfraglen,
1465 data + transhdrlen, fraggap, 0);
1466 skb_prev->csum = csum_sub(skb_prev->csum,
1467 skb->csum);
1468 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001469 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 }
1471 copy = datalen - transhdrlen - fraggap;
Steffen Klassert299b0762011-10-11 01:43:33 +00001472
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 if (copy < 0) {
1474 err = -EINVAL;
1475 kfree_skb(skb);
1476 goto error;
1477 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1478 err = -EFAULT;
1479 kfree_skb(skb);
1480 goto error;
1481 }
1482
1483 offset += copy;
1484 length -= datalen - fraggap;
1485 transhdrlen = 0;
1486 exthdrlen = 0;
Steffen Klassert299b0762011-10-11 01:43:33 +00001487 dst_exthdrlen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488
1489 /*
1490 * Put the packet on the pending queue
1491 */
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001492 __skb_queue_tail(queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 continue;
1494 }
1495
1496 if (copy > length)
1497 copy = length;
1498
Changli Gaod8d1f302010-06-10 23:31:35 -07001499 if (!(rt->dst.dev->features&NETIF_F_SG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 unsigned int off;
1501
1502 off = skb->len;
1503 if (getfrag(from, skb_put(skb, copy),
1504 offset, copy, off, skb) < 0) {
1505 __skb_trim(skb, off);
1506 err = -EFAULT;
1507 goto error;
1508 }
1509 } else {
1510 int i = skb_shinfo(skb)->nr_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511
Eric Dumazet5640f762012-09-23 23:04:42 +00001512 err = -ENOMEM;
1513 if (!sk_page_frag_refill(sk, pfrag))
1514 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515
Eric Dumazet5640f762012-09-23 23:04:42 +00001516 if (!skb_can_coalesce(skb, i, pfrag->page,
1517 pfrag->offset)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 err = -EMSGSIZE;
Eric Dumazet5640f762012-09-23 23:04:42 +00001519 if (i == MAX_SKB_FRAGS)
1520 goto error;
1521
1522 __skb_fill_page_desc(skb, i, pfrag->page,
1523 pfrag->offset, 0);
1524 skb_shinfo(skb)->nr_frags = ++i;
1525 get_page(pfrag->page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001527 copy = min_t(int, copy, pfrag->size - pfrag->offset);
Eric Dumazet9e903e02011-10-18 21:00:24 +00001528 if (getfrag(from,
Eric Dumazet5640f762012-09-23 23:04:42 +00001529 page_address(pfrag->page) + pfrag->offset,
1530 offset, copy, skb->len, skb) < 0)
1531 goto error_efault;
1532
1533 pfrag->offset += copy;
1534 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 skb->len += copy;
1536 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001537 skb->truesize += copy;
1538 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 }
1540 offset += copy;
1541 length -= copy;
1542 }
Eric Dumazet5640f762012-09-23 23:04:42 +00001543
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 return 0;
Eric Dumazet5640f762012-09-23 23:04:42 +00001545
1546error_efault:
1547 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548error:
David S. Millerbdc712b2011-05-06 15:02:07 -07001549 cork->length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001550 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 return err;
1552}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001553
1554int ip6_append_data(struct sock *sk,
1555 int getfrag(void *from, char *to, int offset, int len,
1556 int odd, struct sk_buff *skb),
1557 void *from, int length, int transhdrlen, int hlimit,
1558 int tclass, struct ipv6_txoptions *opt, struct flowi6 *fl6,
1559 struct rt6_info *rt, unsigned int flags, int dontfrag)
1560{
1561 struct inet_sock *inet = inet_sk(sk);
1562 struct ipv6_pinfo *np = inet6_sk(sk);
1563 int exthdrlen;
1564 int err;
1565
1566 if (flags&MSG_PROBE)
1567 return 0;
1568 if (skb_queue_empty(&sk->sk_write_queue)) {
1569 /*
1570 * setup for corking
1571 */
1572 err = ip6_setup_cork(sk, &inet->cork, &np->cork, hlimit,
1573 tclass, opt, rt, fl6);
1574 if (err)
1575 return err;
1576
1577 exthdrlen = (opt ? opt->opt_flen : 0);
1578 length += exthdrlen;
1579 transhdrlen += exthdrlen;
1580 } else {
1581 fl6 = &inet->cork.fl.u.ip6;
1582 transhdrlen = 0;
1583 }
1584
1585 return __ip6_append_data(sk, fl6, &sk->sk_write_queue, &inet->cork.base,
1586 &np->cork, sk_page_frag(sk), getfrag,
1587 from, length, transhdrlen, flags, dontfrag);
1588}
Chris Elstona495f832012-04-29 21:48:53 +00001589EXPORT_SYMBOL_GPL(ip6_append_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001591static void ip6_cork_release(struct inet_cork_full *cork,
1592 struct inet6_cork *v6_cork)
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001593{
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001594 if (v6_cork->opt) {
1595 kfree(v6_cork->opt->dst0opt);
1596 kfree(v6_cork->opt->dst1opt);
1597 kfree(v6_cork->opt->hopopt);
1598 kfree(v6_cork->opt->srcrt);
1599 kfree(v6_cork->opt);
1600 v6_cork->opt = NULL;
Herbert Xu0178b692009-02-05 15:15:50 -08001601 }
1602
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001603 if (cork->base.dst) {
1604 dst_release(cork->base.dst);
1605 cork->base.dst = NULL;
1606 cork->base.flags &= ~IPCORK_ALLFRAG;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001607 }
Vlad Yasevich366e41d2015-01-31 10:40:13 -05001608 memset(&cork->fl, 0, sizeof(cork->fl));
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001609}
1610
Vlad Yasevich64223982015-01-31 10:40:15 -05001611struct sk_buff *__ip6_make_skb(struct sock *sk,
1612 struct sk_buff_head *queue,
1613 struct inet_cork_full *cork,
1614 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615{
1616 struct sk_buff *skb, *tmp_skb;
1617 struct sk_buff **tail_skb;
1618 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001620 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621 struct ipv6hdr *hdr;
Vlad Yasevich64223982015-01-31 10:40:15 -05001622 struct ipv6_txoptions *opt = v6_cork->opt;
1623 struct rt6_info *rt = (struct rt6_info *)cork->base.dst;
1624 struct flowi6 *fl6 = &cork->fl.u.ip6;
David S. Miller4c9483b2011-03-12 16:22:43 -05001625 unsigned char proto = fl6->flowi6_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626
Vlad Yasevich64223982015-01-31 10:40:15 -05001627 skb = __skb_dequeue(queue);
Ian Morris63159f22015-03-29 14:00:04 +01001628 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 goto out;
1630 tail_skb = &(skb_shinfo(skb)->frag_list);
1631
1632 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001633 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001634 __skb_pull(skb, skb_network_offset(skb));
Vlad Yasevich64223982015-01-31 10:40:15 -05001635 while ((tmp_skb = __skb_dequeue(queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001636 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 *tail_skb = tmp_skb;
1638 tail_skb = &(tmp_skb->next);
1639 skb->len += tmp_skb->len;
1640 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 skb->truesize += tmp_skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 tmp_skb->destructor = NULL;
1643 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644 }
1645
Herbert Xu28a894532008-02-12 18:07:27 -08001646 /* Allow local fragmentation. */
WANG Cong60ff7462014-05-04 16:39:18 -07001647 skb->ignore_df = ip6_sk_ignore_df(sk);
Herbert Xu28a894532008-02-12 18:07:27 -08001648
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001649 *final_dst = fl6->daddr;
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001650 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 if (opt && opt->opt_flen)
1652 ipv6_push_frag_opts(skb, opt, &proto);
1653 if (opt && opt->opt_nflen)
1654 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1655
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001656 skb_push(skb, sizeof(struct ipv6hdr));
1657 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001658 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001659
Vlad Yasevich64223982015-01-31 10:40:15 -05001660 ip6_flow_hdr(hdr, v6_cork->tclass,
Tom Herbertcb1ce2e2014-07-01 21:33:10 -07001661 ip6_make_flowlabel(net, skb, fl6->flowlabel,
Tom Herbert67800f92015-07-31 16:52:11 -07001662 np->autoflowlabel, fl6));
Vlad Yasevich64223982015-01-31 10:40:15 -05001663 hdr->hop_limit = v6_cork->hop_limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 hdr->nexthdr = proto;
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +00001665 hdr->saddr = fl6->saddr;
1666 hdr->daddr = *final_dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667
Patrick McHardya2c20642006-01-08 22:37:26 -08001668 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001669 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001670
Changli Gaod8d1f302010-06-10 23:31:35 -07001671 skb_dst_set(skb, dst_clone(&rt->dst));
Neil Hormanedf391f2009-04-27 02:45:02 -07001672 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
David L Stevens14878f72007-09-16 16:52:35 -07001673 if (proto == IPPROTO_ICMPV6) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001674 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
David L Stevens14878f72007-09-16 16:52:35 -07001675
Hannes Frederic Sowa43a43b62014-03-31 20:14:10 +02001676 ICMP6MSGOUT_INC_STATS(net, idev, icmp6_hdr(skb)->icmp6_type);
1677 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001678 }
1679
Vlad Yasevich64223982015-01-31 10:40:15 -05001680 ip6_cork_release(cork, v6_cork);
1681out:
1682 return skb;
1683}
1684
1685int ip6_send_skb(struct sk_buff *skb)
1686{
1687 struct net *net = sock_net(skb->sk);
1688 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
1689 int err;
1690
Herbert Xuef76bc22008-01-11 19:15:08 -08001691 err = ip6_local_out(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 if (err) {
1693 if (err > 0)
Eric Dumazet6ce9e7b2009-09-02 18:05:33 -07001694 err = net_xmit_errno(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 if (err)
Vlad Yasevich64223982015-01-31 10:40:15 -05001696 IP6_INC_STATS(net, rt->rt6i_idev,
1697 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 }
1699
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 return err;
Vlad Yasevich64223982015-01-31 10:40:15 -05001701}
1702
1703int ip6_push_pending_frames(struct sock *sk)
1704{
1705 struct sk_buff *skb;
1706
1707 skb = ip6_finish_skb(sk);
1708 if (!skb)
1709 return 0;
1710
1711 return ip6_send_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712}
Chris Elstona495f832012-04-29 21:48:53 +00001713EXPORT_SYMBOL_GPL(ip6_push_pending_frames);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001715static void __ip6_flush_pending_frames(struct sock *sk,
Vlad Yasevich64223982015-01-31 10:40:15 -05001716 struct sk_buff_head *queue,
1717 struct inet_cork_full *cork,
1718 struct inet6_cork *v6_cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 struct sk_buff *skb;
1721
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001722 while ((skb = __skb_dequeue_tail(queue)) != NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001723 if (skb_dst(skb))
1724 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001725 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726 kfree_skb(skb);
1727 }
1728
Vlad Yasevich64223982015-01-31 10:40:15 -05001729 ip6_cork_release(cork, v6_cork);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730}
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001731
1732void ip6_flush_pending_frames(struct sock *sk)
1733{
Vlad Yasevich64223982015-01-31 10:40:15 -05001734 __ip6_flush_pending_frames(sk, &sk->sk_write_queue,
1735 &inet_sk(sk)->cork, &inet6_sk(sk)->cork);
Vlad Yasevich0bbe84a2015-01-31 10:40:14 -05001736}
Chris Elstona495f832012-04-29 21:48:53 +00001737EXPORT_SYMBOL_GPL(ip6_flush_pending_frames);
Vlad Yasevich64223982015-01-31 10:40:15 -05001738
1739struct sk_buff *ip6_make_skb(struct sock *sk,
1740 int getfrag(void *from, char *to, int offset,
1741 int len, int odd, struct sk_buff *skb),
1742 void *from, int length, int transhdrlen,
1743 int hlimit, int tclass,
1744 struct ipv6_txoptions *opt, struct flowi6 *fl6,
1745 struct rt6_info *rt, unsigned int flags,
1746 int dontfrag)
1747{
1748 struct inet_cork_full cork;
1749 struct inet6_cork v6_cork;
1750 struct sk_buff_head queue;
1751 int exthdrlen = (opt ? opt->opt_flen : 0);
1752 int err;
1753
1754 if (flags & MSG_PROBE)
1755 return NULL;
1756
1757 __skb_queue_head_init(&queue);
1758
1759 cork.base.flags = 0;
1760 cork.base.addr = 0;
1761 cork.base.opt = NULL;
1762 v6_cork.opt = NULL;
1763 err = ip6_setup_cork(sk, &cork, &v6_cork, hlimit, tclass, opt, rt, fl6);
1764 if (err)
1765 return ERR_PTR(err);
1766
1767 if (dontfrag < 0)
1768 dontfrag = inet6_sk(sk)->dontfrag;
1769
1770 err = __ip6_append_data(sk, fl6, &queue, &cork.base, &v6_cork,
1771 &current->task_frag, getfrag, from,
1772 length + exthdrlen, transhdrlen + exthdrlen,
1773 flags, dontfrag);
1774 if (err) {
1775 __ip6_flush_pending_frames(sk, &queue, &cork, &v6_cork);
1776 return ERR_PTR(err);
1777 }
1778
1779 return __ip6_make_skb(sk, &queue, &cork, &v6_cork);
1780}