blob: c77db0b95e263cd24a2571c0f76c69dfcc89a0bd [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070040
41#include <linux/netfilter.h>
42#include <linux/netfilter_ipv6.h>
43
44#include <net/sock.h>
45#include <net/snmp.h>
46
47#include <net/ipv6.h>
48#include <net/ndisc.h>
49#include <net/protocol.h>
50#include <net/ip6_route.h>
51#include <net/addrconf.h>
52#include <net/rawv6.h>
53#include <net/icmp.h>
54#include <net/xfrm.h>
55#include <net/checksum.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090056#include <linux/mroute6.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070057
58static int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *));
59
60static __inline__ void ipv6_select_ident(struct sk_buff *skb, struct frag_hdr *fhdr)
61{
62 static u32 ipv6_fragmentation_id = 1;
63 static DEFINE_SPINLOCK(ip6_id_lock);
64
65 spin_lock_bh(&ip6_id_lock);
66 fhdr->identification = htonl(ipv6_fragmentation_id);
67 if (++ipv6_fragmentation_id == 0)
68 ipv6_fragmentation_id = 1;
69 spin_unlock_bh(&ip6_id_lock);
70}
71
Herbert Xuef76bc22008-01-11 19:15:08 -080072int __ip6_local_out(struct sk_buff *skb)
73{
74 int len;
75
76 len = skb->len - sizeof(struct ipv6hdr);
77 if (len > IPV6_MAXPLEN)
78 len = 0;
79 ipv6_hdr(skb)->payload_len = htons(len);
80
Patrick McHardy6e23ae22007-11-19 18:53:30 -080081 return nf_hook(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, skb->dst->dev,
Herbert Xuef76bc22008-01-11 19:15:08 -080082 dst_output);
83}
84
85int ip6_local_out(struct sk_buff *skb)
86{
87 int err;
88
89 err = __ip6_local_out(skb);
90 if (likely(err == 1))
91 err = dst_output(skb);
92
93 return err;
94}
95EXPORT_SYMBOL_GPL(ip6_local_out);
96
Patrick McHardyad643a72007-10-15 01:51:38 -070097static int ip6_output_finish(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -070098{
Linus Torvalds1da177e2005-04-16 15:20:36 -070099 struct dst_entry *dst = skb->dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
Stephen Hemminger3644f0c2006-12-07 15:08:17 -0800101 if (dst->hh)
102 return neigh_hh_output(dst->hh, skb);
103 else if (dst->neighbour)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104 return dst->neighbour->output(skb);
105
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700106 IP6_INC_STATS_BH(dev_net(dst->dev),
107 ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108 kfree_skb(skb);
109 return -EINVAL;
110
111}
112
113/* dev_loopback_xmit for use with netfilter. */
114static int ip6_dev_loopback_xmit(struct sk_buff *newskb)
115{
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -0700116 skb_reset_mac_header(newskb);
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -0300117 __skb_pull(newskb, skb_network_offset(newskb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118 newskb->pkt_type = PACKET_LOOPBACK;
119 newskb->ip_summed = CHECKSUM_UNNECESSARY;
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700120 WARN_ON(!newskb->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121
122 netif_rx(newskb);
123 return 0;
124}
125
126
127static int ip6_output2(struct sk_buff *skb)
128{
129 struct dst_entry *dst = skb->dst;
130 struct net_device *dev = dst->dev;
131
132 skb->protocol = htons(ETH_P_IPV6);
133 skb->dev = dev;
134
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700135 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 struct ipv6_pinfo* np = skb->sk ? inet6_sk(skb->sk) : NULL;
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900137 struct inet6_dev *idev = ip6_dst_idev(skb->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138
139 if (!(dev->flags & IFF_LOOPBACK) && (!np || np->mc_loop) &&
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900140 ((mroute6_socket && !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) ||
141 ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
142 &ipv6_hdr(skb)->saddr))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
144
145 /* Do not check for IFF_ALLMULTI; multicast routing
146 is not supported in any case.
147 */
148 if (newskb)
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800149 NF_HOOK(PF_INET6, NF_INET_POST_ROUTING, newskb,
150 NULL, newskb->dev,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 ip6_dev_loopback_xmit);
152
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700153 if (ipv6_hdr(skb)->hop_limit == 0) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700154 IP6_INC_STATS(dev_net(dev), idev,
155 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 kfree_skb(skb);
157 return 0;
158 }
159 }
160
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700161 IP6_INC_STATS(dev_net(dev), idev, IPSTATS_MIB_OUTMCASTPKTS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 }
163
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800164 return NF_HOOK(PF_INET6, NF_INET_POST_ROUTING, skb, NULL, skb->dev,
165 ip6_output_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166}
167
John Heffner628a5c52007-04-20 15:53:27 -0700168static inline int ip6_skb_dst_mtu(struct sk_buff *skb)
169{
170 struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
171
172 return (np && np->pmtudisc == IPV6_PMTUDISC_PROBE) ?
173 skb->dst->dev->mtu : dst_mtu(skb->dst);
174}
175
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176int ip6_output(struct sk_buff *skb)
177{
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900178 struct inet6_dev *idev = ip6_dst_idev(skb->dst);
179 if (unlikely(idev->cnf.disable_ipv6)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700180 IP6_INC_STATS(dev_net(skb->dst->dev), idev,
181 IPSTATS_MIB_OUTDISCARDS);
YOSHIFUJI Hideaki778d80b2008-06-28 14:17:11 +0900182 kfree_skb(skb);
183 return 0;
184 }
185
John Heffner628a5c52007-04-20 15:53:27 -0700186 if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700187 dst_allfrag(skb->dst))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 return ip6_fragment(skb, ip6_output2);
189 else
190 return ip6_output2(skb);
191}
192
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193/*
194 * xmit an sk_buff (used by TCP)
195 */
196
197int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
198 struct ipv6_txoptions *opt, int ipfragok)
199{
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700200 struct net *net = sock_net(sk);
Patrick McHardyb30bd282006-03-23 01:17:25 -0800201 struct ipv6_pinfo *np = inet6_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 struct in6_addr *first_hop = &fl->fl6_dst;
203 struct dst_entry *dst = skb->dst;
204 struct ipv6hdr *hdr;
205 u8 proto = fl->proto;
206 int seg_len = skb->len;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900207 int hlimit, tclass;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 u32 mtu;
209
210 if (opt) {
Chuck Leverc2636b42007-10-23 21:07:32 -0700211 unsigned int head_room;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212
213 /* First: exthdrs may take lots of space (~8K for now)
214 MAX_HEADER is not enough.
215 */
216 head_room = opt->opt_nflen + opt->opt_flen;
217 seg_len += head_room;
218 head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
219
220 if (skb_headroom(skb) < head_room) {
221 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900222 if (skb2 == NULL) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700223 IP6_INC_STATS(net, ip6_dst_idev(skb->dst),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900224 IPSTATS_MIB_OUTDISCARDS);
225 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 return -ENOBUFS;
227 }
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900228 kfree_skb(skb);
229 skb = skb2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 if (sk)
231 skb_set_owner_w(skb, sk);
232 }
233 if (opt->opt_flen)
234 ipv6_push_frag_opts(skb, opt, &proto);
235 if (opt->opt_nflen)
236 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
237 }
238
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700239 skb_push(skb, sizeof(struct ipv6hdr));
240 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700241 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242
Wei Yongjun77e2f142008-07-31 20:46:47 -0700243 /* Allow local fragmentation. */
244 if (ipfragok)
245 skb->local_df = 1;
246
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 /*
248 * Fill in the IPv6 header
249 */
250
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 hlimit = -1;
252 if (np)
253 hlimit = np->hop_limit;
254 if (hlimit < 0)
YOSHIFUJI Hideaki6b75d092008-03-10 06:00:30 -0400255 hlimit = ip6_dst_hoplimit(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900257 tclass = -1;
258 if (np)
259 tclass = np->tclass;
260 if (tclass < 0)
261 tclass = 0;
262
Al Viro90bcaf72006-11-08 00:25:17 -0800263 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | fl->fl6_flowlabel;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900264
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 hdr->payload_len = htons(seg_len);
266 hdr->nexthdr = proto;
267 hdr->hop_limit = hlimit;
268
269 ipv6_addr_copy(&hdr->saddr, &fl->fl6_src);
270 ipv6_addr_copy(&hdr->daddr, first_hop);
271
Patrick McHardya2c20642006-01-08 22:37:26 -0800272 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800273 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -0800274
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 mtu = dst_mtu(dst);
Wei Yongjun283d07ac2008-08-03 21:15:59 -0700276 if ((skb->len <= mtu) || skb->local_df || skb_is_gso(skb)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700277 IP6_INC_STATS(net, ip6_dst_idev(skb->dst),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900278 IPSTATS_MIB_OUTREQUESTS);
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800279 return NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, dst->dev,
Harald Welte6869c4d2005-08-09 19:24:19 -0700280 dst_output);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 }
282
283 if (net_ratelimit())
284 printk(KERN_DEBUG "IPv6: sending pkt_too_big to self\n");
285 skb->dev = dst->dev;
286 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, skb->dev);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700287 IP6_INC_STATS(net, ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 kfree_skb(skb);
289 return -EMSGSIZE;
290}
291
YOSHIFUJI Hideaki71590392007-02-22 22:05:40 +0900292EXPORT_SYMBOL(ip6_xmit);
293
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294/*
295 * To avoid extra problems ND packets are send through this
296 * routine. It's code duplication but I really want to avoid
297 * extra checks since ipv6_build_header is used by TCP (which
298 * is for us performance critical)
299 */
300
301int ip6_nd_hdr(struct sock *sk, struct sk_buff *skb, struct net_device *dev,
YOSHIFUJI Hideaki9acd9f32008-04-10 15:42:10 +0900302 const struct in6_addr *saddr, const struct in6_addr *daddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 int proto, int len)
304{
305 struct ipv6_pinfo *np = inet6_sk(sk);
306 struct ipv6hdr *hdr;
307 int totlen;
308
309 skb->protocol = htons(ETH_P_IPV6);
310 skb->dev = dev;
311
312 totlen = len + sizeof(struct ipv6hdr);
313
Arnaldo Carvalho de Melo55f79cc2007-03-14 21:05:03 -0300314 skb_reset_network_header(skb);
315 skb_put(skb, sizeof(struct ipv6hdr));
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700316 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317
Al Viroae08e1f2006-11-08 00:27:11 -0800318 *(__be32*)hdr = htonl(0x60000000);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319
320 hdr->payload_len = htons(len);
321 hdr->nexthdr = proto;
322 hdr->hop_limit = np->hop_limit;
323
324 ipv6_addr_copy(&hdr->saddr, saddr);
325 ipv6_addr_copy(&hdr->daddr, daddr);
326
327 return 0;
328}
329
330static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
331{
332 struct ip6_ra_chain *ra;
333 struct sock *last = NULL;
334
335 read_lock(&ip6_ra_lock);
336 for (ra = ip6_ra_chain; ra; ra = ra->next) {
337 struct sock *sk = ra->sk;
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700338 if (sk && ra->sel == sel &&
339 (!sk->sk_bound_dev_if ||
340 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 if (last) {
342 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
343 if (skb2)
344 rawv6_rcv(last, skb2);
345 }
346 last = sk;
347 }
348 }
349
350 if (last) {
351 rawv6_rcv(last, skb);
352 read_unlock(&ip6_ra_lock);
353 return 1;
354 }
355 read_unlock(&ip6_ra_lock);
356 return 0;
357}
358
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700359static int ip6_forward_proxy_check(struct sk_buff *skb)
360{
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700361 struct ipv6hdr *hdr = ipv6_hdr(skb);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700362 u8 nexthdr = hdr->nexthdr;
363 int offset;
364
365 if (ipv6_ext_hdr(nexthdr)) {
366 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr);
367 if (offset < 0)
368 return 0;
369 } else
370 offset = sizeof(struct ipv6hdr);
371
372 if (nexthdr == IPPROTO_ICMPV6) {
373 struct icmp6hdr *icmp6;
374
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700375 if (!pskb_may_pull(skb, (skb_network_header(skb) +
376 offset + 1 - skb->data)))
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700377 return 0;
378
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700379 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700380
381 switch (icmp6->icmp6_type) {
382 case NDISC_ROUTER_SOLICITATION:
383 case NDISC_ROUTER_ADVERTISEMENT:
384 case NDISC_NEIGHBOUR_SOLICITATION:
385 case NDISC_NEIGHBOUR_ADVERTISEMENT:
386 case NDISC_REDIRECT:
387 /* For reaction involving unicast neighbor discovery
388 * message destined to the proxied address, pass it to
389 * input function.
390 */
391 return 1;
392 default:
393 break;
394 }
395 }
396
Ville Nuorvala74553b02006-09-22 14:42:18 -0700397 /*
398 * The proxying router can't forward traffic sent to a link-local
399 * address, so signal the sender and discard the packet. This
400 * behavior is clarified by the MIPv6 specification.
401 */
402 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
403 dst_link_failure(skb);
404 return -1;
405 }
406
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700407 return 0;
408}
409
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410static inline int ip6_forward_finish(struct sk_buff *skb)
411{
412 return dst_output(skb);
413}
414
415int ip6_forward(struct sk_buff *skb)
416{
417 struct dst_entry *dst = skb->dst;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700418 struct ipv6hdr *hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 struct inet6_skb_parm *opt = IP6CB(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900420 struct net *net = dev_net(dst->dev);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900421
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700422 if (net->ipv6.devconf_all->forwarding == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 goto error;
424
Ben Hutchings4497b072008-06-19 16:22:28 -0700425 if (skb_warn_if_lro(skb))
426 goto drop;
427
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700429 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 goto drop;
431 }
432
Herbert Xu35fc92a2007-03-26 23:22:20 -0700433 skb_forward_csum(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434
435 /*
436 * We DO NOT make any processing on
437 * RA packets, pushing them to user level AS IS
438 * without ane WARRANTY that application will be able
439 * to interpret them. The reason is that we
440 * cannot make anything clever here.
441 *
442 * We are not end-node, so that if packet contains
443 * AH/ESP, we cannot make anything.
444 * Defragmentation also would be mistake, RA packets
445 * cannot be fragmented, because there is no warranty
446 * that different fragments will go along one path. --ANK
447 */
448 if (opt->ra) {
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700449 u8 *ptr = skb_network_header(skb) + opt->ra;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 if (ip6_call_ra_chain(skb, (ptr[2]<<8) + ptr[3]))
451 return 0;
452 }
453
454 /*
455 * check and decrement ttl
456 */
457 if (hdr->hop_limit <= 1) {
458 /* Force OUTPUT device used as source address */
459 skb->dev = dst->dev;
460 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
461 0, skb->dev);
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700462 IP6_INC_STATS_BH(net,
463 ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464
465 kfree_skb(skb);
466 return -ETIMEDOUT;
467 }
468
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -0700469 /* XXX: idev->cnf.proxy_ndp? */
YOSHIFUJI Hideaki53b79972008-07-19 22:35:03 -0700470 if (net->ipv6.devconf_all->proxy_ndp &&
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800471 pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
Ville Nuorvala74553b02006-09-22 14:42:18 -0700472 int proxied = ip6_forward_proxy_check(skb);
473 if (proxied > 0)
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700474 return ip6_input(skb);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700475 else if (proxied < 0) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700476 IP6_INC_STATS(net, ip6_dst_idev(dst),
477 IPSTATS_MIB_INDISCARDS);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700478 goto drop;
479 }
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700480 }
481
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 if (!xfrm6_route_forward(skb)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700483 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 goto drop;
485 }
486 dst = skb->dst;
487
488 /* IPv6 specs say nothing about it, but it is clear that we cannot
489 send redirects to source routed frames.
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900490 We don't send redirects to frames decapsulated from IPsec.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 */
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900492 if (skb->dev == dst->dev && dst->neighbour && opt->srcrt == 0 &&
493 !skb->sp) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 struct in6_addr *target = NULL;
495 struct rt6_info *rt;
496 struct neighbour *n = dst->neighbour;
497
498 /*
499 * incoming and outgoing devices are the same
500 * send a redirect.
501 */
502
503 rt = (struct rt6_info *) dst;
504 if ((rt->rt6i_flags & RTF_GATEWAY))
505 target = (struct in6_addr*)&n->primary_key;
506 else
507 target = &hdr->daddr;
508
509 /* Limit redirects both by destination (here)
510 and by source (inside ndisc_send_redirect)
511 */
512 if (xrlim_allow(dst, 1*HZ))
513 ndisc_send_redirect(skb, n, target);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700514 } else {
515 int addrtype = ipv6_addr_type(&hdr->saddr);
516
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 /* This check is security critical. */
YOSHIFUJI Hideakif81b2e72008-06-25 16:55:26 +0900518 if (addrtype == IPV6_ADDR_ANY ||
519 addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK))
David L Stevens5bb1ab02007-05-09 13:53:44 -0700520 goto error;
521 if (addrtype & IPV6_ADDR_LINKLOCAL) {
522 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
523 ICMPV6_NOT_NEIGHBOUR, 0, skb->dev);
524 goto error;
525 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 }
527
528 if (skb->len > dst_mtu(dst)) {
529 /* Again, force OUTPUT device used as source address */
530 skb->dev = dst->dev;
531 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, dst_mtu(dst), skb->dev);
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700532 IP6_INC_STATS_BH(net,
533 ip6_dst_idev(dst), IPSTATS_MIB_INTOOBIGERRORS);
534 IP6_INC_STATS_BH(net,
535 ip6_dst_idev(dst), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 kfree_skb(skb);
537 return -EMSGSIZE;
538 }
539
540 if (skb_cow(skb, dst->dev->hard_header_len)) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700541 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 goto drop;
543 }
544
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700545 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546
547 /* Mangling hops number delayed to point after skb COW */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900548
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 hdr->hop_limit--;
550
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700551 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800552 return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dst->dev,
553 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554
555error:
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700556 IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557drop:
558 kfree_skb(skb);
559 return -EINVAL;
560}
561
562static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
563{
564 to->pkt_type = from->pkt_type;
565 to->priority = from->priority;
566 to->protocol = from->protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 dst_release(to->dst);
568 to->dst = dst_clone(from->dst);
569 to->dev = from->dev;
Thomas Graf82e91ff2006-11-09 15:19:14 -0800570 to->mark = from->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571
572#ifdef CONFIG_NET_SCHED
573 to->tc_index = from->tc_index;
574#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -0700575 nf_copy(to, from);
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700576#if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
577 defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
578 to->nf_trace = from->nf_trace;
579#endif
James Morris984bc162006-06-09 00:29:17 -0700580 skb_copy_secmark(to, from);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581}
582
583int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr)
584{
585 u16 offset = sizeof(struct ipv6hdr);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700586 struct ipv6_opt_hdr *exthdr =
587 (struct ipv6_opt_hdr *)(ipv6_hdr(skb) + 1);
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700588 unsigned int packet_len = skb->tail - skb->network_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 int found_rhdr = 0;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700590 *nexthdr = &ipv6_hdr(skb)->nexthdr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591
592 while (offset + 1 <= packet_len) {
593
594 switch (**nexthdr) {
595
596 case NEXTHDR_HOP:
Masahide NAKAMURA27637df2006-08-23 19:29:47 -0700597 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 case NEXTHDR_ROUTING:
Masahide NAKAMURA27637df2006-08-23 19:29:47 -0700599 found_rhdr = 1;
600 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 case NEXTHDR_DEST:
Masahide NAKAMURA59fbb3a62007-06-26 23:56:32 -0700602#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
Masahide NAKAMURA27637df2006-08-23 19:29:47 -0700603 if (ipv6_find_tlv(skb, offset, IPV6_TLV_HAO) >= 0)
604 break;
605#endif
606 if (found_rhdr)
607 return offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 break;
609 default :
610 return offset;
611 }
Masahide NAKAMURA27637df2006-08-23 19:29:47 -0700612
613 offset += ipv6_optlen(exthdr);
614 *nexthdr = &exthdr->nexthdr;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700615 exthdr = (struct ipv6_opt_hdr *)(skb_network_header(skb) +
616 offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 }
618
619 return offset;
620}
621
622static int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
623{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 struct sk_buff *frag;
625 struct rt6_info *rt = (struct rt6_info*)skb->dst;
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800626 struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 struct ipv6hdr *tmp_hdr;
628 struct frag_hdr *fh;
629 unsigned int mtu, hlen, left, len;
Al Viroae08e1f2006-11-08 00:27:11 -0800630 __be32 frag_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 int ptr, offset = 0, err=0;
632 u8 *prevhdr, nexthdr = 0;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700633 struct net *net = dev_net(skb->dst->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 hlen = ip6_find_1stfragopt(skb, &prevhdr);
636 nexthdr = *prevhdr;
637
John Heffner628a5c52007-04-20 15:53:27 -0700638 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700639
640 /* We must not fragment if the socket is set to force MTU discovery
641 * or if the skb it not generated by a local socket. (This last
642 * check should be redundant, but it's free.)
643 */
Herbert Xub5c15fc2008-02-14 23:49:37 -0800644 if (!skb->local_df) {
John Heffnerb881ef72007-04-20 15:52:39 -0700645 skb->dev = skb->dst->dev;
646 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, skb->dev);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700647 IP6_INC_STATS(net, ip6_dst_idev(skb->dst),
648 IPSTATS_MIB_FRAGFAILS);
John Heffnerb881ef72007-04-20 15:52:39 -0700649 kfree_skb(skb);
650 return -EMSGSIZE;
651 }
652
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800653 if (np && np->frag_size < mtu) {
654 if (np->frag_size)
655 mtu = np->frag_size;
656 }
657 mtu -= hlen + sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
659 if (skb_shinfo(skb)->frag_list) {
660 int first_len = skb_pagelen(skb);
Herbert Xu29ffe1a2008-01-28 20:45:20 -0800661 int truesizes = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662
663 if (first_len - hlen > mtu ||
664 ((first_len - hlen) & 7) ||
665 skb_cloned(skb))
666 goto slow_path;
667
668 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next) {
669 /* Correct geometry. */
670 if (frag->len > mtu ||
671 ((frag->len & 7) && frag->next) ||
672 skb_headroom(frag) < hlen)
673 goto slow_path;
674
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 /* Partially cloned skb? */
676 if (skb_shared(frag))
677 goto slow_path;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700678
679 BUG_ON(frag->sk);
680 if (skb->sk) {
681 sock_hold(skb->sk);
682 frag->sk = skb->sk;
683 frag->destructor = sock_wfree;
Herbert Xu29ffe1a2008-01-28 20:45:20 -0800684 truesizes += frag->truesize;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700685 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 }
687
688 err = 0;
689 offset = 0;
690 frag = skb_shinfo(skb)->frag_list;
691 skb_shinfo(skb)->frag_list = NULL;
692 /* BUILD HEADER */
693
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800694 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700695 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 if (!tmp_hdr) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700697 IP6_INC_STATS(net, ip6_dst_idev(skb->dst),
698 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 return -ENOMEM;
700 }
701
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 __skb_pull(skb, hlen);
703 fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700704 __skb_push(skb, hlen);
705 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700706 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707
708 ipv6_select_ident(skb, fh);
709 fh->nexthdr = nexthdr;
710 fh->reserved = 0;
711 fh->frag_off = htons(IP6_MF);
712 frag_id = fh->identification;
713
714 first_len = skb_pagelen(skb);
715 skb->data_len = first_len - skb_headlen(skb);
Herbert Xu29ffe1a2008-01-28 20:45:20 -0800716 skb->truesize -= truesizes;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700718 ipv6_hdr(skb)->payload_len = htons(first_len -
719 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900720
721 dst_hold(&rt->u.dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722
723 for (;;) {
724 /* Prepare header of the next frame,
725 * before previous one went down. */
726 if (frag) {
727 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300728 skb_reset_transport_header(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700730 __skb_push(frag, hlen);
731 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700732 memcpy(skb_network_header(frag), tmp_hdr,
733 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 offset += skb->len - hlen - sizeof(struct frag_hdr);
735 fh->nexthdr = nexthdr;
736 fh->reserved = 0;
737 fh->frag_off = htons(offset);
738 if (frag->next != NULL)
739 fh->frag_off |= htons(IP6_MF);
740 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700741 ipv6_hdr(frag)->payload_len =
742 htons(frag->len -
743 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 ip6_copy_metadata(frag, skb);
745 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900746
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 err = output(skb);
Wei Dongdafee492006-08-02 13:41:21 -0700748 if(!err)
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700749 IP6_INC_STATS(net, ip6_dst_idev(&rt->u.dst),
750 IPSTATS_MIB_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700751
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 if (err || !frag)
753 break;
754
755 skb = frag;
756 frag = skb->next;
757 skb->next = NULL;
758 }
759
Jesper Juhla51482b2005-11-08 09:41:34 -0800760 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
762 if (err == 0) {
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700763 IP6_INC_STATS(net, ip6_dst_idev(&rt->u.dst),
764 IPSTATS_MIB_FRAGOKS);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900765 dst_release(&rt->u.dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 return 0;
767 }
768
769 while (frag) {
770 skb = frag->next;
771 kfree_skb(frag);
772 frag = skb;
773 }
774
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700775 IP6_INC_STATS(net, ip6_dst_idev(&rt->u.dst),
776 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900777 dst_release(&rt->u.dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 return err;
779 }
780
781slow_path:
782 left = skb->len - hlen; /* Space per frame */
783 ptr = hlen; /* Where to start from */
784
785 /*
786 * Fragment the datagram.
787 */
788
789 *prevhdr = NEXTHDR_FRAGMENT;
790
791 /*
792 * Keep copying data until we run out.
793 */
794 while(left > 0) {
795 len = left;
796 /* IF: it doesn't fit, use 'mtu' - the data space left */
797 if (len > mtu)
798 len = mtu;
799 /* IF: we are not sending upto and including the packet end
800 then align the next start on an eight byte boundary */
801 if (len < left) {
802 len &= ~7;
803 }
804 /*
805 * Allocate buffer.
806 */
807
Johannes Bergf5184d22008-05-12 20:48:31 -0700808 if ((frag = alloc_skb(len+hlen+sizeof(struct frag_hdr)+LL_ALLOCATED_SPACE(rt->u.dst.dev), GFP_ATOMIC)) == NULL) {
Patrick McHardy64ce2072005-08-09 20:50:53 -0700809 NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n");
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700810 IP6_INC_STATS(net, ip6_dst_idev(skb->dst),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900811 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 err = -ENOMEM;
813 goto fail;
814 }
815
816 /*
817 * Set up data on packet
818 */
819
820 ip6_copy_metadata(frag, skb);
821 skb_reserve(frag, LL_RESERVED_SPACE(rt->u.dst.dev));
822 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700823 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300824 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700825 frag->transport_header = (frag->network_header + hlen +
826 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827
828 /*
829 * Charge the memory for the fragment to any owner
830 * it might possess
831 */
832 if (skb->sk)
833 skb_set_owner_w(frag, skb->sk);
834
835 /*
836 * Copy the packet header into the new buffer.
837 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300838 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839
840 /*
841 * Build fragment header.
842 */
843 fh->nexthdr = nexthdr;
844 fh->reserved = 0;
Yan Zhengf36d6ab2005-10-03 14:19:15 -0700845 if (!frag_id) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846 ipv6_select_ident(skb, fh);
847 frag_id = fh->identification;
848 } else
849 fh->identification = frag_id;
850
851 /*
852 * Copy a block of the IP datagram.
853 */
Wei Yongjun8984e412007-08-21 20:59:08 -0700854 if (skb_copy_bits(skb, ptr, skb_transport_header(frag), len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 BUG();
856 left -= len;
857
858 fh->frag_off = htons(offset);
859 if (left > 0)
860 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700861 ipv6_hdr(frag)->payload_len = htons(frag->len -
862 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863
864 ptr += len;
865 offset += len;
866
867 /*
868 * Put this fragment into the sending queue.
869 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870 err = output(frag);
871 if (err)
872 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700873
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700874 IP6_INC_STATS(net, ip6_dst_idev(skb->dst),
875 IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 }
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700877 IP6_INC_STATS(net, ip6_dst_idev(skb->dst),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900878 IPSTATS_MIB_FRAGOKS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 return err;
881
882fail:
Denis V. Lunev3bd653c2008-10-08 10:54:51 -0700883 IP6_INC_STATS(net, ip6_dst_idev(skb->dst),
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900884 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900885 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 return err;
887}
888
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700889static inline int ip6_rt_check(struct rt6key *rt_key,
890 struct in6_addr *fl_addr,
891 struct in6_addr *addr_cache)
892{
893 return ((rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
894 (addr_cache == NULL || !ipv6_addr_equal(fl_addr, addr_cache)));
895}
896
Herbert Xu497c6152006-07-30 20:19:33 -0700897static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
898 struct dst_entry *dst,
899 struct flowi *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900{
Herbert Xu497c6152006-07-30 20:19:33 -0700901 struct ipv6_pinfo *np = inet6_sk(sk);
902 struct rt6_info *rt = (struct rt6_info *)dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903
Herbert Xu497c6152006-07-30 20:19:33 -0700904 if (!dst)
905 goto out;
906
907 /* Yes, checking route validity in not connected
908 * case is not very simple. Take into account,
909 * that we do not support routing by source, TOS,
910 * and MSG_DONTROUTE --ANK (980726)
911 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700912 * 1. ip6_rt_check(): If route was host route,
913 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700914 * If it is network route, we still may
915 * check its validity using saved pointer
916 * to the last used address: daddr_cache.
917 * We do not want to save whole address now,
918 * (because main consumer of this service
919 * is tcp, which has not this problem),
920 * so that the last trick works only on connected
921 * sockets.
922 * 2. oif also should be the same.
923 */
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700924 if (ip6_rt_check(&rt->rt6i_dst, &fl->fl6_dst, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700925#ifdef CONFIG_IPV6_SUBTREES
926 ip6_rt_check(&rt->rt6i_src, &fl->fl6_src, np->saddr_cache) ||
927#endif
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700928 (fl->oif && fl->oif != dst->dev->ifindex)) {
Herbert Xu497c6152006-07-30 20:19:33 -0700929 dst_release(dst);
930 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 }
932
Herbert Xu497c6152006-07-30 20:19:33 -0700933out:
934 return dst;
935}
936
937static int ip6_dst_lookup_tail(struct sock *sk,
938 struct dst_entry **dst, struct flowi *fl)
939{
940 int err;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900941 struct net *net = sock_net(sk);
Herbert Xu497c6152006-07-30 20:19:33 -0700942
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 if (*dst == NULL)
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800944 *dst = ip6_route_output(net, sk, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945
946 if ((err = (*dst)->error))
947 goto out_err_release;
948
949 if (ipv6_addr_any(&fl->fl6_src)) {
Brian Haley191cd582008-08-14 15:33:21 -0700950 err = ipv6_dev_get_saddr(net, ip6_dst_idev(*dst)->dev,
YOSHIFUJI Hideaki7cbca672008-03-25 09:37:42 +0900951 &fl->fl6_dst,
952 sk ? inet6_sk(sk)->srcprefs : 0,
953 &fl->fl6_src);
Olaf Hering44456d32005-07-27 11:45:17 -0700954 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 goto out_err_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 }
957
Neil Horman95c385b2007-04-25 17:08:10 -0700958#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
Neil Hormane550dfb2008-09-09 13:51:35 -0700959 /*
960 * Here if the dst entry we've looked up
961 * has a neighbour entry that is in the INCOMPLETE
962 * state and the src address from the flow is
963 * marked as OPTIMISTIC, we release the found
964 * dst entry and replace it instead with the
965 * dst entry of the nexthop router
966 */
967 if ((*dst)->neighbour && !((*dst)->neighbour->nud_state & NUD_VALID)) {
968 struct inet6_ifaddr *ifp;
969 struct flowi fl_gw;
970 int redirect;
Neil Horman95c385b2007-04-25 17:08:10 -0700971
Neil Hormane550dfb2008-09-09 13:51:35 -0700972 ifp = ipv6_get_ifaddr(net, &fl->fl6_src,
973 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -0700974
Neil Hormane550dfb2008-09-09 13:51:35 -0700975 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
976 if (ifp)
977 in6_ifa_put(ifp);
Neil Horman95c385b2007-04-25 17:08:10 -0700978
Neil Hormane550dfb2008-09-09 13:51:35 -0700979 if (redirect) {
980 /*
981 * We need to get the dst entry for the
982 * default router instead
983 */
984 dst_release(*dst);
985 memcpy(&fl_gw, fl, sizeof(struct flowi));
986 memset(&fl_gw.fl6_dst, 0, sizeof(struct in6_addr));
987 *dst = ip6_route_output(net, sk, &fl_gw);
988 if ((err = (*dst)->error))
989 goto out_err_release;
Neil Horman95c385b2007-04-25 17:08:10 -0700990 }
Neil Hormane550dfb2008-09-09 13:51:35 -0700991 }
Neil Horman95c385b2007-04-25 17:08:10 -0700992#endif
993
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 return 0;
995
996out_err_release:
Mitsuru Chinenca46f9c2007-12-05 22:31:47 -0800997 if (err == -ENETUNREACH)
Denis V. Lunev483a47d2008-10-08 11:09:27 -0700998 IP6_INC_STATS_BH(net, NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 dst_release(*dst);
1000 *dst = NULL;
1001 return err;
1002}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001003
Herbert Xu497c6152006-07-30 20:19:33 -07001004/**
1005 * ip6_dst_lookup - perform route lookup on flow
1006 * @sk: socket which provides 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.
1011 *
1012 * It returns zero on success, or a standard errno code on error.
1013 */
1014int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl)
1015{
1016 *dst = NULL;
1017 return ip6_dst_lookup_tail(sk, dst, fl);
1018}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -08001019EXPORT_SYMBOL_GPL(ip6_dst_lookup);
1020
Herbert Xu497c6152006-07-30 20:19:33 -07001021/**
1022 * ip6_sk_dst_lookup - perform socket cached route lookup on flow
1023 * @sk: socket which provides the dst cache and route info
1024 * @dst: pointer to dst_entry * for result
1025 * @fl: flow to lookup
1026 *
1027 * This function performs a route lookup on the given flow with the
1028 * possibility of using the cached route in the socket if it is valid.
1029 * It will take the socket dst lock when operating on the dst cache.
1030 * As a result, this function can only be used in process context.
1031 *
1032 * It returns zero on success, or a standard errno code on error.
1033 */
1034int ip6_sk_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl)
1035{
1036 *dst = NULL;
1037 if (sk) {
1038 *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
1039 *dst = ip6_sk_dst_check(sk, *dst, fl);
1040 }
1041
1042 return ip6_dst_lookup_tail(sk, dst, fl);
1043}
1044EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup);
1045
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001046static inline int ip6_ufo_append_data(struct sock *sk,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001047 int getfrag(void *from, char *to, int offset, int len,
1048 int odd, struct sk_buff *skb),
1049 void *from, int length, int hh_len, int fragheaderlen,
1050 int transhdrlen, int mtu,unsigned int flags)
1051
1052{
1053 struct sk_buff *skb;
1054 int err;
1055
1056 /* There is support for UDP large send offload by network
1057 * device, so create one single skb packet containing complete
1058 * udp datagram
1059 */
1060 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) {
1061 skb = sock_alloc_send_skb(sk,
1062 hh_len + fragheaderlen + transhdrlen + 20,
1063 (flags & MSG_DONTWAIT), &err);
1064 if (skb == NULL)
1065 return -ENOMEM;
1066
1067 /* reserve space for Hardware header */
1068 skb_reserve(skb, hh_len);
1069
1070 /* create space for UDP/IP header */
1071 skb_put(skb,fragheaderlen + transhdrlen);
1072
1073 /* initialize network header pointer */
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001074 skb_reset_network_header(skb);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001075
1076 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001077 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001078
Patrick McHardy84fa7932006-08-29 16:44:56 -07001079 skb->ip_summed = CHECKSUM_PARTIAL;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001080 skb->csum = 0;
1081 sk->sk_sndmsg_off = 0;
1082 }
1083
1084 err = skb_append_datato_frags(sk,skb, getfrag, from,
1085 (length - transhdrlen));
1086 if (!err) {
1087 struct frag_hdr fhdr;
1088
1089 /* specify the length of each IP datagram fragment*/
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001090 skb_shinfo(skb)->gso_size = mtu - fragheaderlen -
Herbert Xu79671682006-06-22 02:40:14 -07001091 sizeof(struct frag_hdr);
Herbert Xuf83ef8c2006-06-30 13:37:03 -07001092 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001093 ipv6_select_ident(skb, &fhdr);
1094 skb_shinfo(skb)->ip6_frag_id = fhdr.identification;
1095 __skb_queue_tail(&sk->sk_write_queue, skb);
1096
1097 return 0;
1098 }
1099 /* There is not enough support do UPD LSO,
1100 * so follow normal path
1101 */
1102 kfree_skb(skb);
1103
1104 return err;
1105}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001107int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
1108 int offset, int len, int odd, struct sk_buff *skb),
1109 void *from, int length, int transhdrlen,
1110 int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi *fl,
1111 struct rt6_info *rt, unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112{
1113 struct inet_sock *inet = inet_sk(sk);
1114 struct ipv6_pinfo *np = inet6_sk(sk);
1115 struct sk_buff *skb;
1116 unsigned int maxfraglen, fragheaderlen;
1117 int exthdrlen;
1118 int hh_len;
1119 int mtu;
1120 int copy;
1121 int err;
1122 int offset = 0;
1123 int csummode = CHECKSUM_NONE;
1124
1125 if (flags&MSG_PROBE)
1126 return 0;
1127 if (skb_queue_empty(&sk->sk_write_queue)) {
1128 /*
1129 * setup for corking
1130 */
1131 if (opt) {
1132 if (np->cork.opt == NULL) {
1133 np->cork.opt = kmalloc(opt->tot_len,
1134 sk->sk_allocation);
1135 if (unlikely(np->cork.opt == NULL))
1136 return -ENOBUFS;
1137 } else if (np->cork.opt->tot_len < opt->tot_len) {
1138 printk(KERN_DEBUG "ip6_append_data: invalid option length\n");
1139 return -EINVAL;
1140 }
1141 memcpy(np->cork.opt, opt, opt->tot_len);
1142 inet->cork.flags |= IPCORK_OPT;
1143 /* need source address above miyazawa*/
1144 }
1145 dst_hold(&rt->u.dst);
YOSHIFUJI Hideakic8cdaf92008-03-10 04:30:37 -04001146 inet->cork.dst = &rt->u.dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 inet->cork.fl = *fl;
1148 np->cork.hop_limit = hlimit;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001149 np->cork.tclass = tclass;
John Heffner628a5c52007-04-20 15:53:27 -07001150 mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ?
1151 rt->u.dst.dev->mtu : dst_mtu(rt->u.dst.path);
Dave Jonesc7503602006-03-20 22:44:52 -08001152 if (np->frag_size < mtu) {
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -08001153 if (np->frag_size)
1154 mtu = np->frag_size;
1155 }
1156 inet->cork.fragsize = mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 if (dst_allfrag(rt->u.dst.path))
1158 inet->cork.flags |= IPCORK_ALLFRAG;
1159 inet->cork.length = 0;
1160 sk->sk_sndmsg_page = NULL;
1161 sk->sk_sndmsg_off = 0;
Herbert Xu014889422007-11-13 21:32:26 -08001162 exthdrlen = rt->u.dst.header_len + (opt ? opt->opt_flen : 0) -
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001163 rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 length += exthdrlen;
1165 transhdrlen += exthdrlen;
1166 } else {
YOSHIFUJI Hideakic8cdaf92008-03-10 04:30:37 -04001167 rt = (struct rt6_info *)inet->cork.dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 fl = &inet->cork.fl;
1169 if (inet->cork.flags & IPCORK_OPT)
1170 opt = np->cork.opt;
1171 transhdrlen = 0;
1172 exthdrlen = 0;
1173 mtu = inet->cork.fragsize;
1174 }
1175
1176 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
1177
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001178 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001179 (opt ? opt->opt_nflen : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr);
1181
1182 if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
1183 if (inet->cork.length + length > sizeof(struct ipv6hdr) + IPV6_MAXPLEN - fragheaderlen) {
1184 ipv6_local_error(sk, EMSGSIZE, fl, mtu-exthdrlen);
1185 return -EMSGSIZE;
1186 }
1187 }
1188
1189 /*
1190 * Let's try using as much space as possible.
1191 * Use MTU if total length of the message fits into the MTU.
1192 * Otherwise, we need to reserve fragment header and
1193 * fragment alignment (= 8-15 octects, in total).
1194 *
1195 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001196 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197 * the message.
1198 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001199 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200 * at once if non-fragmentable extension headers
1201 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001202 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203 */
1204
1205 inet->cork.length += length;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001206 if (((length > mtu) && (sk->sk_protocol == IPPROTO_UDP)) &&
1207 (rt->u.dst.dev->features & NETIF_F_UFO)) {
1208
Patrick McHardybaa829d2006-03-12 20:35:12 -08001209 err = ip6_ufo_append_data(sk, getfrag, from, length, hh_len,
1210 fragheaderlen, transhdrlen, mtu,
1211 flags);
1212 if (err)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001213 goto error;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001214 return 0;
1215 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216
1217 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL)
1218 goto alloc_new_skb;
1219
1220 while (length > 0) {
1221 /* Check if the remaining data fits into current packet. */
1222 copy = (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
1223 if (copy < length)
1224 copy = maxfraglen - skb->len;
1225
1226 if (copy <= 0) {
1227 char *data;
1228 unsigned int datalen;
1229 unsigned int fraglen;
1230 unsigned int fraggap;
1231 unsigned int alloclen;
1232 struct sk_buff *skb_prev;
1233alloc_new_skb:
1234 skb_prev = skb;
1235
1236 /* There's no room in the current skb */
1237 if (skb_prev)
1238 fraggap = skb_prev->len - maxfraglen;
1239 else
1240 fraggap = 0;
1241
1242 /*
1243 * If remaining data exceeds the mtu,
1244 * we know we need more fragment(s).
1245 */
1246 datalen = length + fraggap;
1247 if (datalen > (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1248 datalen = maxfraglen - fragheaderlen;
1249
1250 fraglen = datalen + fragheaderlen;
1251 if ((flags & MSG_MORE) &&
1252 !(rt->u.dst.dev->features&NETIF_F_SG))
1253 alloclen = mtu;
1254 else
1255 alloclen = datalen + fragheaderlen;
1256
1257 /*
1258 * The last fragment gets additional space at tail.
1259 * Note: we overallocate on fragments with MSG_MODE
1260 * because we have no idea if we're the last one.
1261 */
1262 if (datalen == length + fraggap)
1263 alloclen += rt->u.dst.trailer_len;
1264
1265 /*
1266 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001267 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 * (without MSG_MORE) fits into the MTU.
1269 */
1270 alloclen += sizeof(struct frag_hdr);
1271
1272 if (transhdrlen) {
1273 skb = sock_alloc_send_skb(sk,
1274 alloclen + hh_len,
1275 (flags & MSG_DONTWAIT), &err);
1276 } else {
1277 skb = NULL;
1278 if (atomic_read(&sk->sk_wmem_alloc) <=
1279 2 * sk->sk_sndbuf)
1280 skb = sock_wmalloc(sk,
1281 alloclen + hh_len, 1,
1282 sk->sk_allocation);
1283 if (unlikely(skb == NULL))
1284 err = -ENOBUFS;
1285 }
1286 if (skb == NULL)
1287 goto error;
1288 /*
1289 * Fill in the control structures
1290 */
1291 skb->ip_summed = csummode;
1292 skb->csum = 0;
1293 /* reserve for fragmentation */
1294 skb_reserve(skb, hh_len+sizeof(struct frag_hdr));
1295
1296 /*
1297 * Find where to start putting bytes
1298 */
1299 data = skb_put(skb, fraglen);
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001300 skb_set_network_header(skb, exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001302 skb->transport_header = (skb->network_header +
1303 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 if (fraggap) {
1305 skb->csum = skb_copy_and_csum_bits(
1306 skb_prev, maxfraglen,
1307 data + transhdrlen, fraggap, 0);
1308 skb_prev->csum = csum_sub(skb_prev->csum,
1309 skb->csum);
1310 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001311 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 }
1313 copy = datalen - transhdrlen - fraggap;
1314 if (copy < 0) {
1315 err = -EINVAL;
1316 kfree_skb(skb);
1317 goto error;
1318 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1319 err = -EFAULT;
1320 kfree_skb(skb);
1321 goto error;
1322 }
1323
1324 offset += copy;
1325 length -= datalen - fraggap;
1326 transhdrlen = 0;
1327 exthdrlen = 0;
1328 csummode = CHECKSUM_NONE;
1329
1330 /*
1331 * Put the packet on the pending queue
1332 */
1333 __skb_queue_tail(&sk->sk_write_queue, skb);
1334 continue;
1335 }
1336
1337 if (copy > length)
1338 copy = length;
1339
1340 if (!(rt->u.dst.dev->features&NETIF_F_SG)) {
1341 unsigned int off;
1342
1343 off = skb->len;
1344 if (getfrag(from, skb_put(skb, copy),
1345 offset, copy, off, skb) < 0) {
1346 __skb_trim(skb, off);
1347 err = -EFAULT;
1348 goto error;
1349 }
1350 } else {
1351 int i = skb_shinfo(skb)->nr_frags;
1352 skb_frag_t *frag = &skb_shinfo(skb)->frags[i-1];
1353 struct page *page = sk->sk_sndmsg_page;
1354 int off = sk->sk_sndmsg_off;
1355 unsigned int left;
1356
1357 if (page && (left = PAGE_SIZE - off) > 0) {
1358 if (copy >= left)
1359 copy = left;
1360 if (page != frag->page) {
1361 if (i == MAX_SKB_FRAGS) {
1362 err = -EMSGSIZE;
1363 goto error;
1364 }
1365 get_page(page);
1366 skb_fill_page_desc(skb, i, page, sk->sk_sndmsg_off, 0);
1367 frag = &skb_shinfo(skb)->frags[i];
1368 }
1369 } else if(i < MAX_SKB_FRAGS) {
1370 if (copy > PAGE_SIZE)
1371 copy = PAGE_SIZE;
1372 page = alloc_pages(sk->sk_allocation, 0);
1373 if (page == NULL) {
1374 err = -ENOMEM;
1375 goto error;
1376 }
1377 sk->sk_sndmsg_page = page;
1378 sk->sk_sndmsg_off = 0;
1379
1380 skb_fill_page_desc(skb, i, page, 0, 0);
1381 frag = &skb_shinfo(skb)->frags[i];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 } else {
1383 err = -EMSGSIZE;
1384 goto error;
1385 }
1386 if (getfrag(from, page_address(frag->page)+frag->page_offset+frag->size, offset, copy, skb->len, skb) < 0) {
1387 err = -EFAULT;
1388 goto error;
1389 }
1390 sk->sk_sndmsg_off += copy;
1391 frag->size += copy;
1392 skb->len += copy;
1393 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001394 skb->truesize += copy;
1395 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 }
1397 offset += copy;
1398 length -= copy;
1399 }
1400 return 0;
1401error:
1402 inet->cork.length -= length;
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001403 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 return err;
1405}
1406
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001407static void ip6_cork_release(struct inet_sock *inet, struct ipv6_pinfo *np)
1408{
1409 inet->cork.flags &= ~IPCORK_OPT;
1410 kfree(np->cork.opt);
1411 np->cork.opt = NULL;
YOSHIFUJI Hideakic8cdaf92008-03-10 04:30:37 -04001412 if (inet->cork.dst) {
1413 dst_release(inet->cork.dst);
1414 inet->cork.dst = NULL;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001415 inet->cork.flags &= ~IPCORK_ALLFRAG;
1416 }
1417 memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
1418}
1419
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420int ip6_push_pending_frames(struct sock *sk)
1421{
1422 struct sk_buff *skb, *tmp_skb;
1423 struct sk_buff **tail_skb;
1424 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
1425 struct inet_sock *inet = inet_sk(sk);
1426 struct ipv6_pinfo *np = inet6_sk(sk);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001427 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 struct ipv6hdr *hdr;
1429 struct ipv6_txoptions *opt = np->cork.opt;
YOSHIFUJI Hideakic8cdaf92008-03-10 04:30:37 -04001430 struct rt6_info *rt = (struct rt6_info *)inet->cork.dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431 struct flowi *fl = &inet->cork.fl;
1432 unsigned char proto = fl->proto;
1433 int err = 0;
1434
1435 if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL)
1436 goto out;
1437 tail_skb = &(skb_shinfo(skb)->frag_list);
1438
1439 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001440 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001441 __skb_pull(skb, skb_network_offset(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001443 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 *tail_skb = tmp_skb;
1445 tail_skb = &(tmp_skb->next);
1446 skb->len += tmp_skb->len;
1447 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 skb->truesize += tmp_skb->truesize;
1449 __sock_put(tmp_skb->sk);
1450 tmp_skb->destructor = NULL;
1451 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452 }
1453
Herbert Xu28a89452008-02-12 18:07:27 -08001454 /* Allow local fragmentation. */
Herbert Xub5c15fc2008-02-14 23:49:37 -08001455 if (np->pmtudisc < IPV6_PMTUDISC_DO)
Herbert Xu28a89452008-02-12 18:07:27 -08001456 skb->local_df = 1;
1457
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458 ipv6_addr_copy(final_dst, &fl->fl6_dst);
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001459 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460 if (opt && opt->opt_flen)
1461 ipv6_push_frag_opts(skb, opt, &proto);
1462 if (opt && opt->opt_nflen)
1463 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1464
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001465 skb_push(skb, sizeof(struct ipv6hdr));
1466 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001467 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001468
Al Viro90bcaf72006-11-08 00:25:17 -08001469 *(__be32*)hdr = fl->fl6_flowlabel |
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001470 htonl(0x60000000 | ((int)np->cork.tclass << 20));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 hdr->hop_limit = np->cork.hop_limit;
1473 hdr->nexthdr = proto;
1474 ipv6_addr_copy(&hdr->saddr, &fl->fl6_src);
1475 ipv6_addr_copy(&hdr->daddr, final_dst);
1476
Patrick McHardya2c20642006-01-08 22:37:26 -08001477 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001478 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001479
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 skb->dst = dst_clone(&rt->u.dst);
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001481 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
David L Stevens14878f72007-09-16 16:52:35 -07001482 if (proto == IPPROTO_ICMPV6) {
1483 struct inet6_dev *idev = ip6_dst_idev(skb->dst);
1484
Denis V. Lunev5a57d4c2008-10-08 10:34:14 -07001485 ICMP6MSGOUT_INC_STATS_BH(net, idev, icmp6_hdr(skb)->icmp6_type);
Denis V. Luneve41b5362008-10-08 10:33:26 -07001486 ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTMSGS);
David L Stevens14878f72007-09-16 16:52:35 -07001487 }
1488
Herbert Xuef76bc22008-01-11 19:15:08 -08001489 err = ip6_local_out(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 if (err) {
1491 if (err > 0)
Herbert Xu3320da82005-04-19 22:32:22 -07001492 err = np->recverr ? net_xmit_errno(err) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 if (err)
1494 goto error;
1495 }
1496
1497out:
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001498 ip6_cork_release(inet, np);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 return err;
1500error:
1501 goto out;
1502}
1503
1504void ip6_flush_pending_frames(struct sock *sk)
1505{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 struct sk_buff *skb;
1507
1508 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) {
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001509 if (skb->dst)
Denis V. Lunev3bd653c2008-10-08 10:54:51 -07001510 IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb->dst),
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001511 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 kfree_skb(skb);
1513 }
1514
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001515 ip6_cork_release(inet_sk(sk), inet6_sk(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516}