blob: 5f8d242be3f3016592678d09e9cf4654be9234a7 [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>
23 * Imran Patel : frag id should be in NBO
24 * 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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070058
David Stevensad0081e2010-12-17 11:42:42 +000059int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *));
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
Herbert Xuef76bc22008-01-11 19:15:08 -080061int __ip6_local_out(struct sk_buff *skb)
62{
63 int len;
64
65 len = skb->len - sizeof(struct ipv6hdr);
66 if (len > IPV6_MAXPLEN)
67 len = 0;
68 ipv6_hdr(skb)->payload_len = htons(len);
69
Jan Engelhardtb2e0b382010-03-23 04:09:07 +010070 return nf_hook(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
71 skb_dst(skb)->dev, dst_output);
Herbert Xuef76bc22008-01-11 19:15:08 -080072}
73
74int ip6_local_out(struct sk_buff *skb)
75{
76 int err;
77
78 err = __ip6_local_out(skb);
79 if (likely(err == 1))
80 err = dst_output(skb);
81
82 return err;
83}
84EXPORT_SYMBOL_GPL(ip6_local_out);
85
Linus Torvalds1da177e2005-04-16 15:20:36 -070086/* dev_loopback_xmit for use with netfilter. */
87static int ip6_dev_loopback_xmit(struct sk_buff *newskb)
88{
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -070089 skb_reset_mac_header(newskb);
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -030090 __skb_pull(newskb, skb_network_offset(newskb));
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 newskb->pkt_type = PACKET_LOOPBACK;
92 newskb->ip_summed = CHECKSUM_UNNECESSARY;
Eric Dumazetadf30902009-06-02 05:19:30 +000093 WARN_ON(!skb_dst(newskb));
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
Eric Dumazete30b38c2010-04-15 09:13:03 +000095 netif_rx_ni(newskb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070096 return 0;
97}
98
Jan Engelhardt9e508492010-04-13 15:28:11 +020099static int ip6_finish_output2(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100{
Eric Dumazetadf30902009-06-02 05:19:30 +0000101 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102 struct net_device *dev = dst->dev;
103
104 skb->protocol = htons(ETH_P_IPV6);
105 skb->dev = dev;
106
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700107 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000108 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109
Octavian Purdila7ad68482010-01-06 20:37:01 -0800110 if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(skb->sk) &&
Patrick McHardyd1db2752010-05-11 14:40:55 +0200111 ((mroute6_socket(dev_net(dev), skb) &&
Benjamin Therybd91b8b2008-12-10 16:07:08 -0800112 !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) ||
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900113 ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
114 &ipv6_hdr(skb)->saddr))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
116
117 /* Do not check for IFF_ALLMULTI; multicast routing
118 is not supported in any case.
119 */
120 if (newskb)
Jan Engelhardtb2e0b382010-03-23 04:09:07 +0100121 NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING,
122 newskb, NULL, newskb->dev,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 ip6_dev_loopback_xmit);
124
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700125 if (ipv6_hdr(skb)->hop_limit == 0) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700126 IP6_INC_STATS(dev_net(dev), idev,
127 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128 kfree_skb(skb);
129 return 0;
130 }
131 }
132
Neil Hormanedf391f2009-04-27 02:45:02 -0700133 IP6_UPD_PO_STATS(dev_net(dev), idev, IPSTATS_MIB_OUTMCAST,
134 skb->len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135 }
136
Jan Engelhardt9e508492010-04-13 15:28:11 +0200137 if (dst->hh)
138 return neigh_hh_output(dst->hh, skb);
139 else if (dst->neighbour)
140 return dst->neighbour->output(skb);
141
142 IP6_INC_STATS_BH(dev_net(dst->dev),
143 ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
144 kfree_skb(skb);
145 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146}
147
Jan Engelhardt9e508492010-04-13 15:28:11 +0200148static int ip6_finish_output(struct sk_buff *skb)
149{
150 if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
151 dst_allfrag(skb_dst(skb)))
152 return ip6_fragment(skb, ip6_finish_output2);
153 else
154 return ip6_finish_output2(skb);
155}
156
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157int ip6_output(struct sk_buff *skb)
158{
Jan Engelhardt9e508492010-04-13 15:28:11 +0200159 struct net_device *dev = skb_dst(skb)->dev;
Eric Dumazetadf30902009-06-02 05:19:30 +0000160 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900161 if (unlikely(idev->cnf.disable_ipv6)) {
Jan Engelhardt9e508492010-04-13 15:28:11 +0200162 IP6_INC_STATS(dev_net(dev), idev,
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700163 IPSTATS_MIB_OUTDISCARDS);
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900164 kfree_skb(skb);
165 return 0;
166 }
167
Jan Engelhardt9c6eb282010-04-13 15:32:16 +0200168 return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING, skb, NULL, dev,
169 ip6_finish_output,
170 !(IP6CB(skb)->flags & IP6SKB_REROUTED));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171}
172
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173/*
Shan Weib5d43992010-04-15 16:48:48 +0000174 * xmit an sk_buff (used by TCP, SCTP and DCCP)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 */
176
177int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
Shan Wei4e15ed42010-04-15 16:43:08 +0000178 struct ipv6_txoptions *opt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179{
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700180 struct net *net = sock_net(sk);
Patrick McHardyb30bd282006-03-23 01:17:25 -0800181 struct ipv6_pinfo *np = inet6_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 struct in6_addr *first_hop = &fl->fl6_dst;
Eric Dumazetadf30902009-06-02 05:19:30 +0000183 struct dst_entry *dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 struct ipv6hdr *hdr;
185 u8 proto = fl->proto;
186 int seg_len = skb->len;
Gerrit Renkere651f032009-08-09 08:12:48 +0000187 int hlimit = -1;
188 int tclass = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 u32 mtu;
190
191 if (opt) {
Chuck Leverc2636b42007-10-23 21:07:32 -0700192 unsigned int head_room;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193
194 /* First: exthdrs may take lots of space (~8K for now)
195 MAX_HEADER is not enough.
196 */
197 head_room = opt->opt_nflen + opt->opt_flen;
198 seg_len += head_room;
199 head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
200
201 if (skb_headroom(skb) < head_room) {
202 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900203 if (skb2 == NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000204 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900205 IPSTATS_MIB_OUTDISCARDS);
206 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 return -ENOBUFS;
208 }
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900209 kfree_skb(skb);
210 skb = skb2;
Dan Carpenter83d7eb22010-04-30 16:42:08 -0700211 skb_set_owner_w(skb, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212 }
213 if (opt->opt_flen)
214 ipv6_push_frag_opts(skb, opt, &proto);
215 if (opt->opt_nflen)
216 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
217 }
218
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700219 skb_push(skb, sizeof(struct ipv6hdr));
220 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700221 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222
223 /*
224 * Fill in the IPv6 header
225 */
Gerrit Renkere651f032009-08-09 08:12:48 +0000226 if (np) {
227 tclass = np->tclass;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 hlimit = np->hop_limit;
Gerrit Renkere651f032009-08-09 08:12:48 +0000229 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 if (hlimit < 0)
YOSHIFUJI Hideaki6b75d092008-03-10 06:00:30 -0400231 hlimit = ip6_dst_hoplimit(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232
Al Viro90bcaf72006-11-08 00:25:17 -0800233 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | fl->fl6_flowlabel;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900234
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 hdr->payload_len = htons(seg_len);
236 hdr->nexthdr = proto;
237 hdr->hop_limit = hlimit;
238
239 ipv6_addr_copy(&hdr->saddr, &fl->fl6_src);
240 ipv6_addr_copy(&hdr->daddr, first_hop);
241
Patrick McHardya2c20642006-01-08 22:37:26 -0800242 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800243 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -0800244
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 mtu = dst_mtu(dst);
Wei Yongjun283d07ac2008-08-03 21:15:59 -0700246 if ((skb->len <= mtu) || skb->local_df || skb_is_gso(skb)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000247 IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)),
Neil Hormanedf391f2009-04-27 02:45:02 -0700248 IPSTATS_MIB_OUT, skb->len);
Jan Engelhardtb2e0b382010-03-23 04:09:07 +0100249 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
250 dst->dev, dst_output);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 }
252
253 if (net_ratelimit())
254 printk(KERN_DEBUG "IPv6: sending pkt_too_big to self\n");
255 skb->dev = dst->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000256 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazetadf30902009-06-02 05:19:30 +0000257 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258 kfree_skb(skb);
259 return -EMSGSIZE;
260}
261
YOSHIFUJI Hideaki71590392007-02-22 22:05:40 +0900262EXPORT_SYMBOL(ip6_xmit);
263
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264/*
265 * To avoid extra problems ND packets are send through this
266 * routine. It's code duplication but I really want to avoid
267 * extra checks since ipv6_build_header is used by TCP (which
268 * is for us performance critical)
269 */
270
271int ip6_nd_hdr(struct sock *sk, struct sk_buff *skb, struct net_device *dev,
YOSHIFUJI Hideaki9acd9f32008-04-10 15:42:10 +0900272 const struct in6_addr *saddr, const struct in6_addr *daddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 int proto, int len)
274{
275 struct ipv6_pinfo *np = inet6_sk(sk);
276 struct ipv6hdr *hdr;
277 int totlen;
278
279 skb->protocol = htons(ETH_P_IPV6);
280 skb->dev = dev;
281
282 totlen = len + sizeof(struct ipv6hdr);
283
Arnaldo Carvalho de Melo55f79cc2007-03-14 21:05:03 -0300284 skb_reset_network_header(skb);
285 skb_put(skb, sizeof(struct ipv6hdr));
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700286 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287
Al Viroae08e1f2006-11-08 00:27:11 -0800288 *(__be32*)hdr = htonl(0x60000000);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289
290 hdr->payload_len = htons(len);
291 hdr->nexthdr = proto;
292 hdr->hop_limit = np->hop_limit;
293
294 ipv6_addr_copy(&hdr->saddr, saddr);
295 ipv6_addr_copy(&hdr->daddr, daddr);
296
297 return 0;
298}
299
300static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
301{
302 struct ip6_ra_chain *ra;
303 struct sock *last = NULL;
304
305 read_lock(&ip6_ra_lock);
306 for (ra = ip6_ra_chain; ra; ra = ra->next) {
307 struct sock *sk = ra->sk;
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700308 if (sk && ra->sel == sel &&
309 (!sk->sk_bound_dev_if ||
310 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 if (last) {
312 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
313 if (skb2)
314 rawv6_rcv(last, skb2);
315 }
316 last = sk;
317 }
318 }
319
320 if (last) {
321 rawv6_rcv(last, skb);
322 read_unlock(&ip6_ra_lock);
323 return 1;
324 }
325 read_unlock(&ip6_ra_lock);
326 return 0;
327}
328
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700329static int ip6_forward_proxy_check(struct sk_buff *skb)
330{
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700331 struct ipv6hdr *hdr = ipv6_hdr(skb);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700332 u8 nexthdr = hdr->nexthdr;
333 int offset;
334
335 if (ipv6_ext_hdr(nexthdr)) {
336 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr);
337 if (offset < 0)
338 return 0;
339 } else
340 offset = sizeof(struct ipv6hdr);
341
342 if (nexthdr == IPPROTO_ICMPV6) {
343 struct icmp6hdr *icmp6;
344
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700345 if (!pskb_may_pull(skb, (skb_network_header(skb) +
346 offset + 1 - skb->data)))
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700347 return 0;
348
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700349 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700350
351 switch (icmp6->icmp6_type) {
352 case NDISC_ROUTER_SOLICITATION:
353 case NDISC_ROUTER_ADVERTISEMENT:
354 case NDISC_NEIGHBOUR_SOLICITATION:
355 case NDISC_NEIGHBOUR_ADVERTISEMENT:
356 case NDISC_REDIRECT:
357 /* For reaction involving unicast neighbor discovery
358 * message destined to the proxied address, pass it to
359 * input function.
360 */
361 return 1;
362 default:
363 break;
364 }
365 }
366
Ville Nuorvala74553b02006-09-22 14:42:18 -0700367 /*
368 * The proxying router can't forward traffic sent to a link-local
369 * address, so signal the sender and discard the packet. This
370 * behavior is clarified by the MIPv6 specification.
371 */
372 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
373 dst_link_failure(skb);
374 return -1;
375 }
376
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700377 return 0;
378}
379
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380static inline int ip6_forward_finish(struct sk_buff *skb)
381{
382 return dst_output(skb);
383}
384
385int ip6_forward(struct sk_buff *skb)
386{
Eric Dumazetadf30902009-06-02 05:19:30 +0000387 struct dst_entry *dst = skb_dst(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700388 struct ipv6hdr *hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 struct inet6_skb_parm *opt = IP6CB(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900390 struct net *net = dev_net(dst->dev);
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800391 u32 mtu;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900392
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700393 if (net->ipv6.devconf_all->forwarding == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 goto error;
395
Ben Hutchings4497b072008-06-19 16:22:28 -0700396 if (skb_warn_if_lro(skb))
397 goto drop;
398
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700400 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 goto drop;
402 }
403
Alexey Kuznetsov72b43d02011-01-12 08:34:08 +0000404 if (skb->pkt_type != PACKET_HOST)
405 goto drop;
406
Herbert Xu35fc92a2007-03-26 23:22:20 -0700407 skb_forward_csum(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408
409 /*
410 * We DO NOT make any processing on
411 * RA packets, pushing them to user level AS IS
412 * without ane WARRANTY that application will be able
413 * to interpret them. The reason is that we
414 * cannot make anything clever here.
415 *
416 * We are not end-node, so that if packet contains
417 * AH/ESP, we cannot make anything.
418 * Defragmentation also would be mistake, RA packets
419 * cannot be fragmented, because there is no warranty
420 * that different fragments will go along one path. --ANK
421 */
422 if (opt->ra) {
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700423 u8 *ptr = skb_network_header(skb) + opt->ra;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 if (ip6_call_ra_chain(skb, (ptr[2]<<8) + ptr[3]))
425 return 0;
426 }
427
428 /*
429 * check and decrement ttl
430 */
431 if (hdr->hop_limit <= 1) {
432 /* Force OUTPUT device used as source address */
433 skb->dev = dst->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000434 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0);
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700435 IP6_INC_STATS_BH(net,
436 ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
438 kfree_skb(skb);
439 return -ETIMEDOUT;
440 }
441
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -0700442 /* XXX: idev->cnf.proxy_ndp? */
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700443 if (net->ipv6.devconf_all->proxy_ndp &&
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800444 pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
Ville Nuorvala74553b02006-09-22 14:42:18 -0700445 int proxied = ip6_forward_proxy_check(skb);
446 if (proxied > 0)
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700447 return ip6_input(skb);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700448 else if (proxied < 0) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700449 IP6_INC_STATS(net, ip6_dst_idev(dst),
450 IPSTATS_MIB_INDISCARDS);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700451 goto drop;
452 }
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700453 }
454
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 if (!xfrm6_route_forward(skb)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700456 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 goto drop;
458 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000459 dst = skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460
461 /* IPv6 specs say nothing about it, but it is clear that we cannot
462 send redirects to source routed frames.
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900463 We don't send redirects to frames decapsulated from IPsec.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 */
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900465 if (skb->dev == dst->dev && dst->neighbour && opt->srcrt == 0 &&
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700466 !skb_sec_path(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 struct in6_addr *target = NULL;
468 struct rt6_info *rt;
469 struct neighbour *n = dst->neighbour;
470
471 /*
472 * incoming and outgoing devices are the same
473 * send a redirect.
474 */
475
476 rt = (struct rt6_info *) dst;
477 if ((rt->rt6i_flags & RTF_GATEWAY))
478 target = (struct in6_addr*)&n->primary_key;
479 else
480 target = &hdr->daddr;
481
482 /* Limit redirects both by destination (here)
483 and by source (inside ndisc_send_redirect)
484 */
485 if (xrlim_allow(dst, 1*HZ))
486 ndisc_send_redirect(skb, n, target);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700487 } else {
488 int addrtype = ipv6_addr_type(&hdr->saddr);
489
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 /* This check is security critical. */
YOSHIFUJI Hideakif81b2e72008-06-25 16:55:26 +0900491 if (addrtype == IPV6_ADDR_ANY ||
492 addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK))
David L Stevens5bb1ab02007-05-09 13:53:44 -0700493 goto error;
494 if (addrtype & IPV6_ADDR_LINKLOCAL) {
495 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000496 ICMPV6_NOT_NEIGHBOUR, 0);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700497 goto error;
498 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 }
500
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800501 mtu = dst_mtu(dst);
502 if (mtu < IPV6_MIN_MTU)
503 mtu = IPV6_MIN_MTU;
504
Herbert Xu0aa68272010-05-27 16:14:30 -0700505 if (skb->len > mtu && !skb_is_gso(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 /* Again, force OUTPUT device used as source address */
507 skb->dev = dst->dev;
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800508 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700509 IP6_INC_STATS_BH(net,
510 ip6_dst_idev(dst), IPSTATS_MIB_INTOOBIGERRORS);
511 IP6_INC_STATS_BH(net,
512 ip6_dst_idev(dst), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 kfree_skb(skb);
514 return -EMSGSIZE;
515 }
516
517 if (skb_cow(skb, dst->dev->hard_header_len)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700518 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 goto drop;
520 }
521
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700522 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523
524 /* Mangling hops number delayed to point after skb COW */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900525
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 hdr->hop_limit--;
527
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700528 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
Jan Engelhardtb2e0b382010-03-23 04:09:07 +0100529 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dst->dev,
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800530 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531
532error:
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700533 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534drop:
535 kfree_skb(skb);
536 return -EINVAL;
537}
538
539static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
540{
541 to->pkt_type = from->pkt_type;
542 to->priority = from->priority;
543 to->protocol = from->protocol;
Eric Dumazetadf30902009-06-02 05:19:30 +0000544 skb_dst_drop(to);
545 skb_dst_set(to, dst_clone(skb_dst(from)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 to->dev = from->dev;
Thomas Graf82e91ff2006-11-09 15:19:14 -0800547 to->mark = from->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548
549#ifdef CONFIG_NET_SCHED
550 to->tc_index = from->tc_index;
551#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -0700552 nf_copy(to, from);
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700553#if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
554 defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
555 to->nf_trace = from->nf_trace;
556#endif
James Morris984bc162006-06-09 00:29:17 -0700557 skb_copy_secmark(to, from);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558}
559
560int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr)
561{
562 u16 offset = sizeof(struct ipv6hdr);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700563 struct ipv6_opt_hdr *exthdr =
564 (struct ipv6_opt_hdr *)(ipv6_hdr(skb) + 1);
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700565 unsigned int packet_len = skb->tail - skb->network_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 int found_rhdr = 0;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700567 *nexthdr = &ipv6_hdr(skb)->nexthdr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568
569 while (offset + 1 <= packet_len) {
570
571 switch (**nexthdr) {
572
573 case NEXTHDR_HOP:
Masahide NAKAMURA27637df2006-08-23 19:29:47 -0700574 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 case NEXTHDR_ROUTING:
Masahide NAKAMURA27637df2006-08-23 19:29:47 -0700576 found_rhdr = 1;
577 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 case NEXTHDR_DEST:
Masahide NAKAMURA59fbb3a2007-06-26 23:56:32 -0700579#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
Masahide NAKAMURA27637df2006-08-23 19:29:47 -0700580 if (ipv6_find_tlv(skb, offset, IPV6_TLV_HAO) >= 0)
581 break;
582#endif
583 if (found_rhdr)
584 return offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 break;
586 default :
587 return offset;
588 }
Masahide NAKAMURA27637df2006-08-23 19:29:47 -0700589
590 offset += ipv6_optlen(exthdr);
591 *nexthdr = &exthdr->nexthdr;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700592 exthdr = (struct ipv6_opt_hdr *)(skb_network_header(skb) +
593 offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 }
595
596 return offset;
597}
598
David Stevensad0081e2010-12-17 11:42:42 +0000599int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 struct sk_buff *frag;
Eric Dumazetadf30902009-06-02 05:19:30 +0000602 struct rt6_info *rt = (struct rt6_info*)skb_dst(skb);
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800603 struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 struct ipv6hdr *tmp_hdr;
605 struct frag_hdr *fh;
606 unsigned int mtu, hlen, left, len;
Al Viroae08e1f2006-11-08 00:27:11 -0800607 __be32 frag_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 int ptr, offset = 0, err=0;
609 u8 *prevhdr, nexthdr = 0;
Eric Dumazetadf30902009-06-02 05:19:30 +0000610 struct net *net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 hlen = ip6_find_1stfragopt(skb, &prevhdr);
613 nexthdr = *prevhdr;
614
John Heffner628a5c52007-04-20 15:53:27 -0700615 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700616
617 /* We must not fragment if the socket is set to force MTU discovery
Ulrich Weber14f3ad62010-02-26 04:34:49 -0800618 * or if the skb it not generated by a local socket.
John Heffnerb881ef72007-04-20 15:52:39 -0700619 */
Shan Weif2228f72010-04-18 16:58:22 +0000620 if (!skb->local_df && skb->len > mtu) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000621 skb->dev = skb_dst(skb)->dev;
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000622 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Eric Dumazetadf30902009-06-02 05:19:30 +0000623 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700624 IPSTATS_MIB_FRAGFAILS);
John Heffnerb881ef72007-04-20 15:52:39 -0700625 kfree_skb(skb);
626 return -EMSGSIZE;
627 }
628
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800629 if (np && np->frag_size < mtu) {
630 if (np->frag_size)
631 mtu = np->frag_size;
632 }
633 mtu -= hlen + sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634
David S. Miller21dc3302010-08-23 00:13:46 -0700635 if (skb_has_frag_list(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 int first_len = skb_pagelen(skb);
Eric Dumazet3d130082010-09-21 08:47:45 +0000637 struct sk_buff *frag2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638
639 if (first_len - hlen > mtu ||
640 ((first_len - hlen) & 7) ||
641 skb_cloned(skb))
642 goto slow_path;
643
David S. Miller4d9092b2009-06-09 00:20:05 -0700644 skb_walk_frags(skb, frag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645 /* Correct geometry. */
646 if (frag->len > mtu ||
647 ((frag->len & 7) && frag->next) ||
648 skb_headroom(frag) < hlen)
Eric Dumazet3d130082010-09-21 08:47:45 +0000649 goto slow_path_clean;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 /* Partially cloned skb? */
652 if (skb_shared(frag))
Eric Dumazet3d130082010-09-21 08:47:45 +0000653 goto slow_path_clean;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700654
655 BUG_ON(frag->sk);
656 if (skb->sk) {
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700657 frag->sk = skb->sk;
658 frag->destructor = sock_wfree;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700659 }
Eric Dumazet3d130082010-09-21 08:47:45 +0000660 skb->truesize -= frag->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 }
662
663 err = 0;
664 offset = 0;
665 frag = skb_shinfo(skb)->frag_list;
David S. Miller4d9092b2009-06-09 00:20:05 -0700666 skb_frag_list_init(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 /* BUILD HEADER */
668
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800669 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700670 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 if (!tmp_hdr) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000672 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700673 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 return -ENOMEM;
675 }
676
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 __skb_pull(skb, hlen);
678 fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700679 __skb_push(skb, hlen);
680 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700681 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682
Sridhar Samudrala7ea2f2c2009-07-09 08:10:01 +0000683 ipv6_select_ident(fh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 fh->nexthdr = nexthdr;
685 fh->reserved = 0;
686 fh->frag_off = htons(IP6_MF);
687 frag_id = fh->identification;
688
689 first_len = skb_pagelen(skb);
690 skb->data_len = first_len - skb_headlen(skb);
691 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700692 ipv6_hdr(skb)->payload_len = htons(first_len -
693 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900694
Changli Gaod8d1f302010-06-10 23:31:35 -0700695 dst_hold(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696
697 for (;;) {
698 /* Prepare header of the next frame,
699 * before previous one went down. */
700 if (frag) {
701 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300702 skb_reset_transport_header(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700704 __skb_push(frag, hlen);
705 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700706 memcpy(skb_network_header(frag), tmp_hdr,
707 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 offset += skb->len - hlen - sizeof(struct frag_hdr);
709 fh->nexthdr = nexthdr;
710 fh->reserved = 0;
711 fh->frag_off = htons(offset);
712 if (frag->next != NULL)
713 fh->frag_off |= htons(IP6_MF);
714 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700715 ipv6_hdr(frag)->payload_len =
716 htons(frag->len -
717 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 ip6_copy_metadata(frag, skb);
719 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900720
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 err = output(skb);
Wei Dongdafee492006-08-02 13:41:21 -0700722 if(!err)
Changli Gaod8d1f302010-06-10 23:31:35 -0700723 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700724 IPSTATS_MIB_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700725
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 if (err || !frag)
727 break;
728
729 skb = frag;
730 frag = skb->next;
731 skb->next = NULL;
732 }
733
Jesper Juhla51482b2005-11-08 09:41:34 -0800734 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735
736 if (err == 0) {
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_FRAGOKS);
Changli Gaod8d1f302010-06-10 23:31:35 -0700739 dst_release(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 return 0;
741 }
742
743 while (frag) {
744 skb = frag->next;
745 kfree_skb(frag);
746 frag = skb;
747 }
748
Changli Gaod8d1f302010-06-10 23:31:35 -0700749 IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700750 IPSTATS_MIB_FRAGFAILS);
Changli Gaod8d1f302010-06-10 23:31:35 -0700751 dst_release(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 return err;
Eric Dumazet3d130082010-09-21 08:47:45 +0000753
754slow_path_clean:
755 skb_walk_frags(skb, frag2) {
756 if (frag2 == frag)
757 break;
758 frag2->sk = NULL;
759 frag2->destructor = NULL;
760 skb->truesize += frag2->truesize;
761 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 }
763
764slow_path:
765 left = skb->len - hlen; /* Space per frame */
766 ptr = hlen; /* Where to start from */
767
768 /*
769 * Fragment the datagram.
770 */
771
772 *prevhdr = NEXTHDR_FRAGMENT;
773
774 /*
775 * Keep copying data until we run out.
776 */
777 while(left > 0) {
778 len = left;
779 /* IF: it doesn't fit, use 'mtu' - the data space left */
780 if (len > mtu)
781 len = mtu;
782 /* IF: we are not sending upto and including the packet end
783 then align the next start on an eight byte boundary */
784 if (len < left) {
785 len &= ~7;
786 }
787 /*
788 * Allocate buffer.
789 */
790
Changli Gaod8d1f302010-06-10 23:31:35 -0700791 if ((frag = alloc_skb(len+hlen+sizeof(struct frag_hdr)+LL_ALLOCATED_SPACE(rt->dst.dev), GFP_ATOMIC)) == NULL) {
Patrick McHardy64ce2072005-08-09 20:50:53 -0700792 NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n");
Eric Dumazetadf30902009-06-02 05:19:30 +0000793 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900794 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 err = -ENOMEM;
796 goto fail;
797 }
798
799 /*
800 * Set up data on packet
801 */
802
803 ip6_copy_metadata(frag, skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700804 skb_reserve(frag, LL_RESERVED_SPACE(rt->dst.dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700806 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300807 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700808 frag->transport_header = (frag->network_header + hlen +
809 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810
811 /*
812 * Charge the memory for the fragment to any owner
813 * it might possess
814 */
815 if (skb->sk)
816 skb_set_owner_w(frag, skb->sk);
817
818 /*
819 * Copy the packet header into the new buffer.
820 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300821 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822
823 /*
824 * Build fragment header.
825 */
826 fh->nexthdr = nexthdr;
827 fh->reserved = 0;
Yan Zhengf36d6ab2005-10-03 14:19:15 -0700828 if (!frag_id) {
Sridhar Samudrala7ea2f2c2009-07-09 08:10:01 +0000829 ipv6_select_ident(fh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 frag_id = fh->identification;
831 } else
832 fh->identification = frag_id;
833
834 /*
835 * Copy a block of the IP datagram.
836 */
Wei Yongjun8984e412007-08-21 20:59:08 -0700837 if (skb_copy_bits(skb, ptr, skb_transport_header(frag), len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 BUG();
839 left -= len;
840
841 fh->frag_off = htons(offset);
842 if (left > 0)
843 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700844 ipv6_hdr(frag)->payload_len = htons(frag->len -
845 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846
847 ptr += len;
848 offset += len;
849
850 /*
851 * Put this fragment into the sending queue.
852 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 err = output(frag);
854 if (err)
855 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700856
Eric Dumazetadf30902009-06-02 05:19:30 +0000857 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700858 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 }
Eric Dumazetadf30902009-06-02 05:19:30 +0000860 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900861 IPSTATS_MIB_FRAGOKS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 return err;
864
865fail:
Eric Dumazetadf30902009-06-02 05:19:30 +0000866 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900867 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900868 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 return err;
870}
871
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700872static inline int ip6_rt_check(struct rt6key *rt_key,
873 struct in6_addr *fl_addr,
874 struct in6_addr *addr_cache)
875{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000876 return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
877 (addr_cache == NULL || !ipv6_addr_equal(fl_addr, addr_cache));
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700878}
879
Herbert Xu497c6152006-07-30 20:19:33 -0700880static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
881 struct dst_entry *dst,
882 struct flowi *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883{
Herbert Xu497c6152006-07-30 20:19:33 -0700884 struct ipv6_pinfo *np = inet6_sk(sk);
885 struct rt6_info *rt = (struct rt6_info *)dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886
Herbert Xu497c6152006-07-30 20:19:33 -0700887 if (!dst)
888 goto out;
889
890 /* Yes, checking route validity in not connected
891 * case is not very simple. Take into account,
892 * that we do not support routing by source, TOS,
893 * and MSG_DONTROUTE --ANK (980726)
894 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700895 * 1. ip6_rt_check(): If route was host route,
896 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700897 * If it is network route, we still may
898 * check its validity using saved pointer
899 * to the last used address: daddr_cache.
900 * We do not want to save whole address now,
901 * (because main consumer of this service
902 * is tcp, which has not this problem),
903 * so that the last trick works only on connected
904 * sockets.
905 * 2. oif also should be the same.
906 */
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700907 if (ip6_rt_check(&rt->rt6i_dst, &fl->fl6_dst, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700908#ifdef CONFIG_IPV6_SUBTREES
909 ip6_rt_check(&rt->rt6i_src, &fl->fl6_src, np->saddr_cache) ||
910#endif
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700911 (fl->oif && fl->oif != dst->dev->ifindex)) {
Herbert Xu497c6152006-07-30 20:19:33 -0700912 dst_release(dst);
913 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 }
915
Herbert Xu497c6152006-07-30 20:19:33 -0700916out:
917 return dst;
918}
919
920static int ip6_dst_lookup_tail(struct sock *sk,
921 struct dst_entry **dst, struct flowi *fl)
922{
923 int err;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900924 struct net *net = sock_net(sk);
Herbert Xu497c6152006-07-30 20:19:33 -0700925
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 if (*dst == NULL)
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800927 *dst = ip6_route_output(net, sk, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928
929 if ((err = (*dst)->error))
930 goto out_err_release;
931
932 if (ipv6_addr_any(&fl->fl6_src)) {
Brian Haley191cd582008-08-14 15:33:21 -0700933 err = ipv6_dev_get_saddr(net, ip6_dst_idev(*dst)->dev,
YOSHIFUJI Hideaki7cbca672008-03-25 09:37:42 +0900934 &fl->fl6_dst,
935 sk ? inet6_sk(sk)->srcprefs : 0,
936 &fl->fl6_src);
Olaf Hering44456d32005-07-27 11:45:17 -0700937 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 goto out_err_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 }
940
Neil Horman95c385b2007-04-25 17:08:10 -0700941#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700942 /*
943 * Here if the dst entry we've looked up
944 * has a neighbour entry that is in the INCOMPLETE
945 * state and the src address from the flow is
946 * marked as OPTIMISTIC, we release the found
947 * dst entry and replace it instead with the
948 * dst entry of the nexthop router
949 */
950 if ((*dst)->neighbour && !((*dst)->neighbour->nud_state & NUD_VALID)) {
951 struct inet6_ifaddr *ifp;
952 struct flowi fl_gw;
953 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -0700954
Neil Hormane550dfb2008-09-09 13:51:35 -0700955 ifp = ipv6_get_ifaddr(net, &fl->fl6_src,
956 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -0700957
Neil Hormane550dfb2008-09-09 13:51:35 -0700958 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
959 if (ifp)
960 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -0700961
Neil Hormane550dfb2008-09-09 13:51:35 -0700962 if (redirect) {
963 /*
964 * We need to get the dst entry for the
965 * default router instead
966 */
967 dst_release(*dst);
968 memcpy(&fl_gw, fl, sizeof(struct flowi));
969 memset(&fl_gw.fl6_dst, 0, sizeof(struct in6_addr));
970 *dst = ip6_route_output(net, sk, &fl_gw);
971 if ((err = (*dst)->error))
972 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -0700973 }
Neil Hormane550dfb2008-09-09 13:51:35 -0700974 }
Neil Horman95c385b2007-04-25 17:08:10 -0700975#endif
976
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 return 0;
978
979out_err_release:
Mitsuru Chinenca46f9c2007-12-05 22:31:47 -0800980 if (err == -ENETUNREACH)
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700981 IP6_INC_STATS_BH(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 dst_release(*dst);
983 *dst = NULL;
984 return err;
985}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -0800986
Herbert Xu497c6152006-07-30 20:19:33 -0700987/**
988 * ip6_dst_lookup - perform route lookup on flow
989 * @sk: socket which provides route info
990 * @dst: pointer to dst_entry * for result
991 * @fl: flow to lookup
992 *
993 * This function performs a route lookup on the given flow.
994 *
995 * It returns zero on success, or a standard errno code on error.
996 */
997int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl)
998{
999 *dst = NULL;
1000 return ip6_dst_lookup_tail(sk, dst, fl);
1001}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -08001002EXPORT_SYMBOL_GPL(ip6_dst_lookup);
1003
Herbert Xu497c6152006-07-30 20:19:33 -07001004/**
1005 * ip6_sk_dst_lookup - perform socket cached route lookup on flow
1006 * @sk: socket which provides the dst cache and route info
1007 * @dst: pointer to dst_entry * for result
1008 * @fl: flow to lookup
1009 *
1010 * This function performs a route lookup on the given flow with the
1011 * possibility of using the cached route in the socket if it is valid.
1012 * It will take the socket dst lock when operating on the dst cache.
1013 * As a result, this function can only be used in process context.
1014 *
1015 * It returns zero on success, or a standard errno code on error.
1016 */
1017int ip6_sk_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl)
1018{
1019 *dst = NULL;
1020 if (sk) {
1021 *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
1022 *dst = ip6_sk_dst_check(sk, *dst, fl);
1023 }
1024
1025 return ip6_dst_lookup_tail(sk, dst, fl);
1026}
1027EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup);
1028
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001029static inline int ip6_ufo_append_data(struct sock *sk,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001030 int getfrag(void *from, char *to, int offset, int len,
1031 int odd, struct sk_buff *skb),
1032 void *from, int length, int hh_len, int fragheaderlen,
1033 int transhdrlen, int mtu,unsigned int flags)
1034
1035{
1036 struct sk_buff *skb;
1037 int err;
1038
1039 /* There is support for UDP large send offload by network
1040 * device, so create one single skb packet containing complete
1041 * udp datagram
1042 */
1043 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) {
1044 skb = sock_alloc_send_skb(sk,
1045 hh_len + fragheaderlen + transhdrlen + 20,
1046 (flags & MSG_DONTWAIT), &err);
1047 if (skb == NULL)
1048 return -ENOMEM;
1049
1050 /* reserve space for Hardware header */
1051 skb_reserve(skb, hh_len);
1052
1053 /* create space for UDP/IP header */
1054 skb_put(skb,fragheaderlen + transhdrlen);
1055
1056 /* initialize network header pointer */
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001057 skb_reset_network_header(skb);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001058
1059 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001060 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001061
Patrick McHardy84fa7932006-08-29 16:44:56 -07001062 skb->ip_summed = CHECKSUM_PARTIAL;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001063 skb->csum = 0;
1064 sk->sk_sndmsg_off = 0;
1065 }
1066
1067 err = skb_append_datato_frags(sk,skb, getfrag, from,
1068 (length - transhdrlen));
1069 if (!err) {
1070 struct frag_hdr fhdr;
1071
Sridhar Samudralac31d5322009-07-09 08:09:58 +00001072 /* Specify the length of each IPv6 datagram fragment.
1073 * It has to be a multiple of 8.
1074 */
1075 skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
1076 sizeof(struct frag_hdr)) & ~7;
Herbert Xuf83ef8c2006-06-30 13:37:03 -07001077 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Sridhar Samudrala7ea2f2c2009-07-09 08:10:01 +00001078 ipv6_select_ident(&fhdr);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001079 skb_shinfo(skb)->ip6_frag_id = fhdr.identification;
1080 __skb_queue_tail(&sk->sk_write_queue, skb);
1081
1082 return 0;
1083 }
1084 /* There is not enough support do UPD LSO,
1085 * so follow normal path
1086 */
1087 kfree_skb(skb);
1088
1089 return err;
1090}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091
Herbert Xu0178b692009-02-05 15:15:50 -08001092static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
1093 gfp_t gfp)
1094{
1095 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1096}
1097
1098static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
1099 gfp_t gfp)
1100{
1101 return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
1102}
1103
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001104int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
1105 int offset, int len, int odd, struct sk_buff *skb),
1106 void *from, int length, int transhdrlen,
1107 int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi *fl,
Brian Haley13b52cd2010-04-23 11:26:08 +00001108 struct rt6_info *rt, unsigned int flags, int dontfrag)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109{
1110 struct inet_sock *inet = inet_sk(sk);
1111 struct ipv6_pinfo *np = inet6_sk(sk);
1112 struct sk_buff *skb;
1113 unsigned int maxfraglen, fragheaderlen;
1114 int exthdrlen;
1115 int hh_len;
1116 int mtu;
1117 int copy;
1118 int err;
1119 int offset = 0;
1120 int csummode = CHECKSUM_NONE;
1121
1122 if (flags&MSG_PROBE)
1123 return 0;
1124 if (skb_queue_empty(&sk->sk_write_queue)) {
1125 /*
1126 * setup for corking
1127 */
1128 if (opt) {
Herbert Xu0178b692009-02-05 15:15:50 -08001129 if (WARN_ON(np->cork.opt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 return -EINVAL;
Herbert Xu0178b692009-02-05 15:15:50 -08001131
1132 np->cork.opt = kmalloc(opt->tot_len, sk->sk_allocation);
1133 if (unlikely(np->cork.opt == NULL))
1134 return -ENOBUFS;
1135
1136 np->cork.opt->tot_len = opt->tot_len;
1137 np->cork.opt->opt_flen = opt->opt_flen;
1138 np->cork.opt->opt_nflen = opt->opt_nflen;
1139
1140 np->cork.opt->dst0opt = ip6_opt_dup(opt->dst0opt,
1141 sk->sk_allocation);
1142 if (opt->dst0opt && !np->cork.opt->dst0opt)
1143 return -ENOBUFS;
1144
1145 np->cork.opt->dst1opt = ip6_opt_dup(opt->dst1opt,
1146 sk->sk_allocation);
1147 if (opt->dst1opt && !np->cork.opt->dst1opt)
1148 return -ENOBUFS;
1149
1150 np->cork.opt->hopopt = ip6_opt_dup(opt->hopopt,
1151 sk->sk_allocation);
1152 if (opt->hopopt && !np->cork.opt->hopopt)
1153 return -ENOBUFS;
1154
1155 np->cork.opt->srcrt = ip6_rthdr_dup(opt->srcrt,
1156 sk->sk_allocation);
1157 if (opt->srcrt && !np->cork.opt->srcrt)
1158 return -ENOBUFS;
1159
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 /* need source address above miyazawa*/
1161 }
Changli Gaod8d1f302010-06-10 23:31:35 -07001162 dst_hold(&rt->dst);
1163 inet->cork.dst = &rt->dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 inet->cork.fl = *fl;
1165 np->cork.hop_limit = hlimit;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001166 np->cork.tclass = tclass;
John Heffner628a5c52007-04-20 15:53:27 -07001167 mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ?
Changli Gaod8d1f302010-06-10 23:31:35 -07001168 rt->dst.dev->mtu : dst_mtu(rt->dst.path);
Dave Jonesc7503602006-03-20 22:44:52 -08001169 if (np->frag_size < mtu) {
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -08001170 if (np->frag_size)
1171 mtu = np->frag_size;
1172 }
1173 inet->cork.fragsize = mtu;
Changli Gaod8d1f302010-06-10 23:31:35 -07001174 if (dst_allfrag(rt->dst.path))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 inet->cork.flags |= IPCORK_ALLFRAG;
1176 inet->cork.length = 0;
1177 sk->sk_sndmsg_page = NULL;
1178 sk->sk_sndmsg_off = 0;
Changli Gaod8d1f302010-06-10 23:31:35 -07001179 exthdrlen = rt->dst.header_len + (opt ? opt->opt_flen : 0) -
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001180 rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181 length += exthdrlen;
1182 transhdrlen += exthdrlen;
1183 } else {
YOSHIFUJI Hideakic8cdaf92008-03-10 04:30:37 -04001184 rt = (struct rt6_info *)inet->cork.dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 fl = &inet->cork.fl;
Herbert Xu0178b692009-02-05 15:15:50 -08001186 opt = np->cork.opt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 transhdrlen = 0;
1188 exthdrlen = 0;
1189 mtu = inet->cork.fragsize;
1190 }
1191
Changli Gaod8d1f302010-06-10 23:31:35 -07001192 hh_len = LL_RESERVED_SPACE(rt->dst.dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001194 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001195 (opt ? opt->opt_nflen : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr);
1197
1198 if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
1199 if (inet->cork.length + length > sizeof(struct ipv6hdr) + IPV6_MAXPLEN - fragheaderlen) {
1200 ipv6_local_error(sk, EMSGSIZE, fl, mtu-exthdrlen);
1201 return -EMSGSIZE;
1202 }
1203 }
1204
1205 /*
1206 * Let's try using as much space as possible.
1207 * Use MTU if total length of the message fits into the MTU.
1208 * Otherwise, we need to reserve fragment header and
1209 * fragment alignment (= 8-15 octects, in total).
1210 *
1211 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001212 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 * the message.
1214 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001215 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 * at once if non-fragmentable extension headers
1217 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001218 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 */
1220
1221 inet->cork.length += length;
Brian Haley4b340ae2010-04-23 11:26:09 +00001222 if (length > mtu) {
1223 int proto = sk->sk_protocol;
1224 if (dontfrag && (proto == IPPROTO_UDP || proto == IPPROTO_RAW)){
1225 ipv6_local_rxpmtu(sk, fl, mtu-exthdrlen);
1226 return -EMSGSIZE;
1227 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001228
Brian Haley4b340ae2010-04-23 11:26:09 +00001229 if (proto == IPPROTO_UDP &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001230 (rt->dst.dev->features & NETIF_F_UFO)) {
Brian Haley4b340ae2010-04-23 11:26:09 +00001231
1232 err = ip6_ufo_append_data(sk, getfrag, from, length,
1233 hh_len, fragheaderlen,
1234 transhdrlen, mtu, flags);
1235 if (err)
1236 goto error;
1237 return 0;
1238 }
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001239 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240
1241 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL)
1242 goto alloc_new_skb;
1243
1244 while (length > 0) {
1245 /* Check if the remaining data fits into current packet. */
1246 copy = (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
1247 if (copy < length)
1248 copy = maxfraglen - skb->len;
1249
1250 if (copy <= 0) {
1251 char *data;
1252 unsigned int datalen;
1253 unsigned int fraglen;
1254 unsigned int fraggap;
1255 unsigned int alloclen;
1256 struct sk_buff *skb_prev;
1257alloc_new_skb:
1258 skb_prev = skb;
1259
1260 /* There's no room in the current skb */
1261 if (skb_prev)
1262 fraggap = skb_prev->len - maxfraglen;
1263 else
1264 fraggap = 0;
1265
1266 /*
1267 * If remaining data exceeds the mtu,
1268 * we know we need more fragment(s).
1269 */
1270 datalen = length + fraggap;
1271 if (datalen > (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1272 datalen = maxfraglen - fragheaderlen;
1273
1274 fraglen = datalen + fragheaderlen;
1275 if ((flags & MSG_MORE) &&
Changli Gaod8d1f302010-06-10 23:31:35 -07001276 !(rt->dst.dev->features&NETIF_F_SG))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 alloclen = mtu;
1278 else
1279 alloclen = datalen + fragheaderlen;
1280
1281 /*
1282 * The last fragment gets additional space at tail.
1283 * Note: we overallocate on fragments with MSG_MODE
1284 * because we have no idea if we're the last one.
1285 */
1286 if (datalen == length + fraggap)
Changli Gaod8d1f302010-06-10 23:31:35 -07001287 alloclen += rt->dst.trailer_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288
1289 /*
1290 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001291 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 * (without MSG_MORE) fits into the MTU.
1293 */
1294 alloclen += sizeof(struct frag_hdr);
1295
1296 if (transhdrlen) {
1297 skb = sock_alloc_send_skb(sk,
1298 alloclen + hh_len,
1299 (flags & MSG_DONTWAIT), &err);
1300 } else {
1301 skb = NULL;
1302 if (atomic_read(&sk->sk_wmem_alloc) <=
1303 2 * sk->sk_sndbuf)
1304 skb = sock_wmalloc(sk,
1305 alloclen + hh_len, 1,
1306 sk->sk_allocation);
1307 if (unlikely(skb == NULL))
1308 err = -ENOBUFS;
1309 }
1310 if (skb == NULL)
1311 goto error;
1312 /*
1313 * Fill in the control structures
1314 */
1315 skb->ip_summed = csummode;
1316 skb->csum = 0;
1317 /* reserve for fragmentation */
1318 skb_reserve(skb, hh_len+sizeof(struct frag_hdr));
1319
1320 /*
1321 * Find where to start putting bytes
1322 */
1323 data = skb_put(skb, fraglen);
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001324 skb_set_network_header(skb, exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001326 skb->transport_header = (skb->network_header +
1327 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 if (fraggap) {
1329 skb->csum = skb_copy_and_csum_bits(
1330 skb_prev, maxfraglen,
1331 data + transhdrlen, fraggap, 0);
1332 skb_prev->csum = csum_sub(skb_prev->csum,
1333 skb->csum);
1334 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001335 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 }
1337 copy = datalen - transhdrlen - fraggap;
1338 if (copy < 0) {
1339 err = -EINVAL;
1340 kfree_skb(skb);
1341 goto error;
1342 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1343 err = -EFAULT;
1344 kfree_skb(skb);
1345 goto error;
1346 }
1347
1348 offset += copy;
1349 length -= datalen - fraggap;
1350 transhdrlen = 0;
1351 exthdrlen = 0;
1352 csummode = CHECKSUM_NONE;
1353
1354 /*
1355 * Put the packet on the pending queue
1356 */
1357 __skb_queue_tail(&sk->sk_write_queue, skb);
1358 continue;
1359 }
1360
1361 if (copy > length)
1362 copy = length;
1363
Changli Gaod8d1f302010-06-10 23:31:35 -07001364 if (!(rt->dst.dev->features&NETIF_F_SG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 unsigned int off;
1366
1367 off = skb->len;
1368 if (getfrag(from, skb_put(skb, copy),
1369 offset, copy, off, skb) < 0) {
1370 __skb_trim(skb, off);
1371 err = -EFAULT;
1372 goto error;
1373 }
1374 } else {
1375 int i = skb_shinfo(skb)->nr_frags;
1376 skb_frag_t *frag = &skb_shinfo(skb)->frags[i-1];
1377 struct page *page = sk->sk_sndmsg_page;
1378 int off = sk->sk_sndmsg_off;
1379 unsigned int left;
1380
1381 if (page && (left = PAGE_SIZE - off) > 0) {
1382 if (copy >= left)
1383 copy = left;
1384 if (page != frag->page) {
1385 if (i == MAX_SKB_FRAGS) {
1386 err = -EMSGSIZE;
1387 goto error;
1388 }
1389 get_page(page);
1390 skb_fill_page_desc(skb, i, page, sk->sk_sndmsg_off, 0);
1391 frag = &skb_shinfo(skb)->frags[i];
1392 }
1393 } else if(i < MAX_SKB_FRAGS) {
1394 if (copy > PAGE_SIZE)
1395 copy = PAGE_SIZE;
1396 page = alloc_pages(sk->sk_allocation, 0);
1397 if (page == NULL) {
1398 err = -ENOMEM;
1399 goto error;
1400 }
1401 sk->sk_sndmsg_page = page;
1402 sk->sk_sndmsg_off = 0;
1403
1404 skb_fill_page_desc(skb, i, page, 0, 0);
1405 frag = &skb_shinfo(skb)->frags[i];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 } else {
1407 err = -EMSGSIZE;
1408 goto error;
1409 }
1410 if (getfrag(from, page_address(frag->page)+frag->page_offset+frag->size, offset, copy, skb->len, skb) < 0) {
1411 err = -EFAULT;
1412 goto error;
1413 }
1414 sk->sk_sndmsg_off += copy;
1415 frag->size += copy;
1416 skb->len += copy;
1417 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001418 skb->truesize += copy;
1419 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 }
1421 offset += copy;
1422 length -= copy;
1423 }
1424 return 0;
1425error:
1426 inet->cork.length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001427 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 return err;
1429}
1430
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001431static void ip6_cork_release(struct inet_sock *inet, struct ipv6_pinfo *np)
1432{
Herbert Xu0178b692009-02-05 15:15:50 -08001433 if (np->cork.opt) {
1434 kfree(np->cork.opt->dst0opt);
1435 kfree(np->cork.opt->dst1opt);
1436 kfree(np->cork.opt->hopopt);
1437 kfree(np->cork.opt->srcrt);
1438 kfree(np->cork.opt);
1439 np->cork.opt = NULL;
1440 }
1441
YOSHIFUJI Hideakic8cdaf92008-03-10 04:30:37 -04001442 if (inet->cork.dst) {
1443 dst_release(inet->cork.dst);
1444 inet->cork.dst = NULL;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001445 inet->cork.flags &= ~IPCORK_ALLFRAG;
1446 }
1447 memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
1448}
1449
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450int ip6_push_pending_frames(struct sock *sk)
1451{
1452 struct sk_buff *skb, *tmp_skb;
1453 struct sk_buff **tail_skb;
1454 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
1455 struct inet_sock *inet = inet_sk(sk);
1456 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001457 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458 struct ipv6hdr *hdr;
1459 struct ipv6_txoptions *opt = np->cork.opt;
YOSHIFUJI Hideakic8cdaf92008-03-10 04:30:37 -04001460 struct rt6_info *rt = (struct rt6_info *)inet->cork.dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 struct flowi *fl = &inet->cork.fl;
1462 unsigned char proto = fl->proto;
1463 int err = 0;
1464
1465 if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL)
1466 goto out;
1467 tail_skb = &(skb_shinfo(skb)->frag_list);
1468
1469 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001470 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001471 __skb_pull(skb, skb_network_offset(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001473 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 *tail_skb = tmp_skb;
1475 tail_skb = &(tmp_skb->next);
1476 skb->len += tmp_skb->len;
1477 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 skb->truesize += tmp_skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 tmp_skb->destructor = NULL;
1480 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 }
1482
Herbert Xu28a89452008-02-12 18:07:27 -08001483 /* Allow local fragmentation. */
Herbert Xub5c15fc2008-02-14 23:49:37 -08001484 if (np->pmtudisc < IPV6_PMTUDISC_DO)
Herbert Xu28a89452008-02-12 18:07:27 -08001485 skb->local_df = 1;
1486
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 ipv6_addr_copy(final_dst, &fl->fl6_dst);
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001488 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489 if (opt && opt->opt_flen)
1490 ipv6_push_frag_opts(skb, opt, &proto);
1491 if (opt && opt->opt_nflen)
1492 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1493
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001494 skb_push(skb, sizeof(struct ipv6hdr));
1495 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001496 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001497
Al Viro90bcaf72006-11-08 00:25:17 -08001498 *(__be32*)hdr = fl->fl6_flowlabel |
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001499 htonl(0x60000000 | ((int)np->cork.tclass << 20));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 hdr->hop_limit = np->cork.hop_limit;
1502 hdr->nexthdr = proto;
1503 ipv6_addr_copy(&hdr->saddr, &fl->fl6_src);
1504 ipv6_addr_copy(&hdr->daddr, final_dst);
1505
Patrick McHardya2c20642006-01-08 22:37:26 -08001506 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001507 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001508
Changli Gaod8d1f302010-06-10 23:31:35 -07001509 skb_dst_set(skb, dst_clone(&rt->dst));
Neil Hormanedf391f2009-04-27 02:45:02 -07001510 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
David L Stevens14878f72007-09-16 16:52:35 -07001511 if (proto == IPPROTO_ICMPV6) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001512 struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
David L Stevens14878f72007-09-16 16:52:35 -07001513
Denis V. Lunev5a57d4c2008-10-08 10:34:14 -07001514 ICMP6MSGOUT_INC_STATS_BH(net, idev, icmp6_hdr(skb)->icmp6_type);
Denis V. Luneve41b5362008-10-08 10:33:26 -07001515 ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001516 }
1517
Herbert Xuef76bc22008-01-11 19:15:08 -08001518 err = ip6_local_out(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 if (err) {
1520 if (err > 0)
Eric Dumazet6ce9e7b2009-09-02 18:05:33 -07001521 err = net_xmit_errno(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522 if (err)
1523 goto error;
1524 }
1525
1526out:
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001527 ip6_cork_release(inet, np);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 return err;
1529error:
Eric Dumazet06254912009-09-01 18:37:16 -07001530 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 goto out;
1532}
1533
1534void ip6_flush_pending_frames(struct sock *sk)
1535{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 struct sk_buff *skb;
1537
1538 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) {
Eric Dumazetadf30902009-06-02 05:19:30 +00001539 if (skb_dst(skb))
1540 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001541 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542 kfree_skb(skb);
1543 }
1544
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001545 ip6_cork_release(inet_sk(sk), inet6_sk(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546}