blob: 40a2813a63d17ca6987980c3ac31fb0301cb0d11 [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
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900106 IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 kfree_skb(skb);
108 return -EINVAL;
109
110}
111
112/* dev_loopback_xmit for use with netfilter. */
113static int ip6_dev_loopback_xmit(struct sk_buff *newskb)
114{
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -0700115 skb_reset_mac_header(newskb);
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -0300116 __skb_pull(newskb, skb_network_offset(newskb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 newskb->pkt_type = PACKET_LOOPBACK;
118 newskb->ip_summed = CHECKSUM_UNNECESSARY;
119 BUG_TRAP(newskb->dst);
120
121 netif_rx(newskb);
122 return 0;
123}
124
125
126static int ip6_output2(struct sk_buff *skb)
127{
128 struct dst_entry *dst = skb->dst;
129 struct net_device *dev = dst->dev;
130
131 skb->protocol = htons(ETH_P_IPV6);
132 skb->dev = dev;
133
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700134 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135 struct ipv6_pinfo* np = skb->sk ? inet6_sk(skb->sk) : NULL;
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900136 struct inet6_dev *idev = ip6_dst_idev(skb->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137
138 if (!(dev->flags & IFF_LOOPBACK) && (!np || np->mc_loop) &&
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900139 ((mroute6_socket && !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) ||
140 ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
141 &ipv6_hdr(skb)->saddr))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
143
144 /* Do not check for IFF_ALLMULTI; multicast routing
145 is not supported in any case.
146 */
147 if (newskb)
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800148 NF_HOOK(PF_INET6, NF_INET_POST_ROUTING, newskb,
149 NULL, newskb->dev,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 ip6_dev_loopback_xmit);
151
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700152 if (ipv6_hdr(skb)->hop_limit == 0) {
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900153 IP6_INC_STATS(idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154 kfree_skb(skb);
155 return 0;
156 }
157 }
158
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900159 IP6_INC_STATS(idev, IPSTATS_MIB_OUTMCASTPKTS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 }
161
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800162 return NF_HOOK(PF_INET6, NF_INET_POST_ROUTING, skb, NULL, skb->dev,
163 ip6_output_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164}
165
John Heffner628a5c52007-04-20 15:53:27 -0700166static inline int ip6_skb_dst_mtu(struct sk_buff *skb)
167{
168 struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
169
170 return (np && np->pmtudisc == IPV6_PMTUDISC_PROBE) ?
171 skb->dst->dev->mtu : dst_mtu(skb->dst);
172}
173
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174int ip6_output(struct sk_buff *skb)
175{
John Heffner628a5c52007-04-20 15:53:27 -0700176 if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700177 dst_allfrag(skb->dst))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 return ip6_fragment(skb, ip6_output2);
179 else
180 return ip6_output2(skb);
181}
182
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183/*
184 * xmit an sk_buff (used by TCP)
185 */
186
187int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
188 struct ipv6_txoptions *opt, int ipfragok)
189{
Patrick McHardyb30bd282006-03-23 01:17:25 -0800190 struct ipv6_pinfo *np = inet6_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 struct in6_addr *first_hop = &fl->fl6_dst;
192 struct dst_entry *dst = skb->dst;
193 struct ipv6hdr *hdr;
194 u8 proto = fl->proto;
195 int seg_len = skb->len;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900196 int hlimit, tclass;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 u32 mtu;
198
199 if (opt) {
Chuck Leverc2636b42007-10-23 21:07:32 -0700200 unsigned int head_room;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201
202 /* First: exthdrs may take lots of space (~8K for now)
203 MAX_HEADER is not enough.
204 */
205 head_room = opt->opt_nflen + opt->opt_flen;
206 seg_len += head_room;
207 head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
208
209 if (skb_headroom(skb) < head_room) {
210 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900211 if (skb2 == NULL) {
212 IP6_INC_STATS(ip6_dst_idev(skb->dst),
213 IPSTATS_MIB_OUTDISCARDS);
214 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215 return -ENOBUFS;
216 }
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900217 kfree_skb(skb);
218 skb = skb2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 if (sk)
220 skb_set_owner_w(skb, sk);
221 }
222 if (opt->opt_flen)
223 ipv6_push_frag_opts(skb, opt, &proto);
224 if (opt->opt_nflen)
225 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
226 }
227
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700228 skb_push(skb, sizeof(struct ipv6hdr));
229 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700230 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231
232 /*
233 * Fill in the IPv6 header
234 */
235
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 hlimit = -1;
237 if (np)
238 hlimit = np->hop_limit;
239 if (hlimit < 0)
YOSHIFUJI Hideaki6b75d092008-03-10 06:00:30 -0400240 hlimit = ip6_dst_hoplimit(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900242 tclass = -1;
243 if (np)
244 tclass = np->tclass;
245 if (tclass < 0)
246 tclass = 0;
247
Al Viro90bcaf72006-11-08 00:25:17 -0800248 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | fl->fl6_flowlabel;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900249
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 hdr->payload_len = htons(seg_len);
251 hdr->nexthdr = proto;
252 hdr->hop_limit = hlimit;
253
254 ipv6_addr_copy(&hdr->saddr, &fl->fl6_src);
255 ipv6_addr_copy(&hdr->daddr, first_hop);
256
Patrick McHardya2c20642006-01-08 22:37:26 -0800257 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800258 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -0800259
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 mtu = dst_mtu(dst);
Herbert Xu89114af2006-07-08 13:34:32 -0700261 if ((skb->len <= mtu) || ipfragok || skb_is_gso(skb)) {
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900262 IP6_INC_STATS(ip6_dst_idev(skb->dst),
263 IPSTATS_MIB_OUTREQUESTS);
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800264 return NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, dst->dev,
Harald Welte6869c4d2005-08-09 19:24:19 -0700265 dst_output);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 }
267
268 if (net_ratelimit())
269 printk(KERN_DEBUG "IPv6: sending pkt_too_big to self\n");
270 skb->dev = dst->dev;
271 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, skb->dev);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900272 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 kfree_skb(skb);
274 return -EMSGSIZE;
275}
276
YOSHIFUJI Hideaki71590392007-02-22 22:05:40 +0900277EXPORT_SYMBOL(ip6_xmit);
278
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279/*
280 * To avoid extra problems ND packets are send through this
281 * routine. It's code duplication but I really want to avoid
282 * extra checks since ipv6_build_header is used by TCP (which
283 * is for us performance critical)
284 */
285
286int ip6_nd_hdr(struct sock *sk, struct sk_buff *skb, struct net_device *dev,
YOSHIFUJI Hideaki9acd9f32008-04-10 15:42:10 +0900287 const struct in6_addr *saddr, const struct in6_addr *daddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 int proto, int len)
289{
290 struct ipv6_pinfo *np = inet6_sk(sk);
291 struct ipv6hdr *hdr;
292 int totlen;
293
294 skb->protocol = htons(ETH_P_IPV6);
295 skb->dev = dev;
296
297 totlen = len + sizeof(struct ipv6hdr);
298
Arnaldo Carvalho de Melo55f79cc2007-03-14 21:05:03 -0300299 skb_reset_network_header(skb);
300 skb_put(skb, sizeof(struct ipv6hdr));
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700301 hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302
Al Viroae08e1f2006-11-08 00:27:11 -0800303 *(__be32*)hdr = htonl(0x60000000);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304
305 hdr->payload_len = htons(len);
306 hdr->nexthdr = proto;
307 hdr->hop_limit = np->hop_limit;
308
309 ipv6_addr_copy(&hdr->saddr, saddr);
310 ipv6_addr_copy(&hdr->daddr, daddr);
311
312 return 0;
313}
314
315static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
316{
317 struct ip6_ra_chain *ra;
318 struct sock *last = NULL;
319
320 read_lock(&ip6_ra_lock);
321 for (ra = ip6_ra_chain; ra; ra = ra->next) {
322 struct sock *sk = ra->sk;
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700323 if (sk && ra->sel == sel &&
324 (!sk->sk_bound_dev_if ||
325 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326 if (last) {
327 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
328 if (skb2)
329 rawv6_rcv(last, skb2);
330 }
331 last = sk;
332 }
333 }
334
335 if (last) {
336 rawv6_rcv(last, skb);
337 read_unlock(&ip6_ra_lock);
338 return 1;
339 }
340 read_unlock(&ip6_ra_lock);
341 return 0;
342}
343
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700344static int ip6_forward_proxy_check(struct sk_buff *skb)
345{
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700346 struct ipv6hdr *hdr = ipv6_hdr(skb);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700347 u8 nexthdr = hdr->nexthdr;
348 int offset;
349
350 if (ipv6_ext_hdr(nexthdr)) {
351 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr);
352 if (offset < 0)
353 return 0;
354 } else
355 offset = sizeof(struct ipv6hdr);
356
357 if (nexthdr == IPPROTO_ICMPV6) {
358 struct icmp6hdr *icmp6;
359
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700360 if (!pskb_may_pull(skb, (skb_network_header(skb) +
361 offset + 1 - skb->data)))
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700362 return 0;
363
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700364 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700365
366 switch (icmp6->icmp6_type) {
367 case NDISC_ROUTER_SOLICITATION:
368 case NDISC_ROUTER_ADVERTISEMENT:
369 case NDISC_NEIGHBOUR_SOLICITATION:
370 case NDISC_NEIGHBOUR_ADVERTISEMENT:
371 case NDISC_REDIRECT:
372 /* For reaction involving unicast neighbor discovery
373 * message destined to the proxied address, pass it to
374 * input function.
375 */
376 return 1;
377 default:
378 break;
379 }
380 }
381
Ville Nuorvala74553b02006-09-22 14:42:18 -0700382 /*
383 * The proxying router can't forward traffic sent to a link-local
384 * address, so signal the sender and discard the packet. This
385 * behavior is clarified by the MIPv6 specification.
386 */
387 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
388 dst_link_failure(skb);
389 return -1;
390 }
391
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700392 return 0;
393}
394
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395static inline int ip6_forward_finish(struct sk_buff *skb)
396{
397 return dst_output(skb);
398}
399
400int ip6_forward(struct sk_buff *skb)
401{
402 struct dst_entry *dst = skb->dst;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700403 struct ipv6hdr *hdr = ipv6_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 struct inet6_skb_parm *opt = IP6CB(skb);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900405 struct net *net = dev_net(dst->dev);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900406
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 if (ipv6_devconf.forwarding == 0)
408 goto error;
409
410 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900411 IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 goto drop;
413 }
414
Herbert Xu35fc92a2007-03-26 23:22:20 -0700415 skb_forward_csum(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416
417 /*
418 * We DO NOT make any processing on
419 * RA packets, pushing them to user level AS IS
420 * without ane WARRANTY that application will be able
421 * to interpret them. The reason is that we
422 * cannot make anything clever here.
423 *
424 * We are not end-node, so that if packet contains
425 * AH/ESP, we cannot make anything.
426 * Defragmentation also would be mistake, RA packets
427 * cannot be fragmented, because there is no warranty
428 * that different fragments will go along one path. --ANK
429 */
430 if (opt->ra) {
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700431 u8 *ptr = skb_network_header(skb) + opt->ra;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 if (ip6_call_ra_chain(skb, (ptr[2]<<8) + ptr[3]))
433 return 0;
434 }
435
436 /*
437 * check and decrement ttl
438 */
439 if (hdr->hop_limit <= 1) {
440 /* Force OUTPUT device used as source address */
441 skb->dev = dst->dev;
442 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
443 0, skb->dev);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900444 IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445
446 kfree_skb(skb);
447 return -ETIMEDOUT;
448 }
449
YOSHIFUJI Hideakifbea49e2006-09-22 14:43:49 -0700450 /* XXX: idev->cnf.proxy_ndp? */
451 if (ipv6_devconf.proxy_ndp &&
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800452 pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
Ville Nuorvala74553b02006-09-22 14:42:18 -0700453 int proxied = ip6_forward_proxy_check(skb);
454 if (proxied > 0)
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700455 return ip6_input(skb);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700456 else if (proxied < 0) {
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900457 IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
Ville Nuorvala74553b02006-09-22 14:42:18 -0700458 goto drop;
459 }
Ville Nuorvalae21e0b52006-09-22 14:41:44 -0700460 }
461
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 if (!xfrm6_route_forward(skb)) {
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900463 IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 goto drop;
465 }
466 dst = skb->dst;
467
468 /* IPv6 specs say nothing about it, but it is clear that we cannot
469 send redirects to source routed frames.
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900470 We don't send redirects to frames decapsulated from IPsec.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 */
Masahide NAKAMURA1e5dc142007-08-24 19:08:55 +0900472 if (skb->dev == dst->dev && dst->neighbour && opt->srcrt == 0 &&
473 !skb->sp) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 struct in6_addr *target = NULL;
475 struct rt6_info *rt;
476 struct neighbour *n = dst->neighbour;
477
478 /*
479 * incoming and outgoing devices are the same
480 * send a redirect.
481 */
482
483 rt = (struct rt6_info *) dst;
484 if ((rt->rt6i_flags & RTF_GATEWAY))
485 target = (struct in6_addr*)&n->primary_key;
486 else
487 target = &hdr->daddr;
488
489 /* Limit redirects both by destination (here)
490 and by source (inside ndisc_send_redirect)
491 */
492 if (xrlim_allow(dst, 1*HZ))
493 ndisc_send_redirect(skb, n, target);
David L Stevens5bb1ab02007-05-09 13:53:44 -0700494 } else {
495 int addrtype = ipv6_addr_type(&hdr->saddr);
496
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 /* This check is security critical. */
David L Stevens5bb1ab02007-05-09 13:53:44 -0700498 if (addrtype & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LOOPBACK))
499 goto error;
500 if (addrtype & IPV6_ADDR_LINKLOCAL) {
501 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
502 ICMPV6_NOT_NEIGHBOUR, 0, skb->dev);
503 goto error;
504 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 }
506
507 if (skb->len > dst_mtu(dst)) {
508 /* Again, force OUTPUT device used as source address */
509 skb->dev = dst->dev;
510 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, dst_mtu(dst), skb->dev);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900511 IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INTOOBIGERRORS);
512 IP6_INC_STATS_BH(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)) {
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900518 IP6_INC_STATS(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
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900528 IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800529 return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dst->dev,
530 ip6_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531
532error:
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900533 IP6_INC_STATS_BH(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;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 dst_release(to->dst);
545 to->dst = dst_clone(from->dst);
546 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
599static int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
600{
601 struct net_device *dev;
602 struct sk_buff *frag;
603 struct rt6_info *rt = (struct rt6_info*)skb->dst;
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800604 struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 struct ipv6hdr *tmp_hdr;
606 struct frag_hdr *fh;
607 unsigned int mtu, hlen, left, len;
Al Viroae08e1f2006-11-08 00:27:11 -0800608 __be32 frag_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 int ptr, offset = 0, err=0;
610 u8 *prevhdr, nexthdr = 0;
611
612 dev = rt->u.dst.dev;
613 hlen = ip6_find_1stfragopt(skb, &prevhdr);
614 nexthdr = *prevhdr;
615
John Heffner628a5c52007-04-20 15:53:27 -0700616 mtu = ip6_skb_dst_mtu(skb);
John Heffnerb881ef72007-04-20 15:52:39 -0700617
618 /* We must not fragment if the socket is set to force MTU discovery
619 * or if the skb it not generated by a local socket. (This last
620 * check should be redundant, but it's free.)
621 */
Herbert Xub5c15fc2008-02-14 23:49:37 -0800622 if (!skb->local_df) {
John Heffnerb881ef72007-04-20 15:52:39 -0700623 skb->dev = skb->dst->dev;
624 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, skb->dev);
625 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGFAILS);
626 kfree_skb(skb);
627 return -EMSGSIZE;
628 }
629
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -0800630 if (np && np->frag_size < mtu) {
631 if (np->frag_size)
632 mtu = np->frag_size;
633 }
634 mtu -= hlen + sizeof(struct frag_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635
636 if (skb_shinfo(skb)->frag_list) {
637 int first_len = skb_pagelen(skb);
Herbert Xu29ffe1a2008-01-28 20:45:20 -0800638 int truesizes = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639
640 if (first_len - hlen > mtu ||
641 ((first_len - hlen) & 7) ||
642 skb_cloned(skb))
643 goto slow_path;
644
645 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next) {
646 /* Correct geometry. */
647 if (frag->len > mtu ||
648 ((frag->len & 7) && frag->next) ||
649 skb_headroom(frag) < hlen)
650 goto slow_path;
651
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 /* Partially cloned skb? */
653 if (skb_shared(frag))
654 goto slow_path;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700655
656 BUG_ON(frag->sk);
657 if (skb->sk) {
658 sock_hold(skb->sk);
659 frag->sk = skb->sk;
660 frag->destructor = sock_wfree;
Herbert Xu29ffe1a2008-01-28 20:45:20 -0800661 truesizes += frag->truesize;
Herbert Xu2fdba6b2005-05-18 22:52:33 -0700662 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 }
664
665 err = 0;
666 offset = 0;
667 frag = skb_shinfo(skb)->frag_list;
668 skb_shinfo(skb)->frag_list = NULL;
669 /* BUILD HEADER */
670
YOSHIFUJI Hideaki9a217a12006-12-05 13:47:21 -0800671 *prevhdr = NEXTHDR_FRAGMENT;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700672 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 if (!tmp_hdr) {
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900674 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 return -ENOMEM;
676 }
677
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 __skb_pull(skb, hlen);
679 fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700680 __skb_push(skb, hlen);
681 skb_reset_network_header(skb);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700682 memcpy(skb_network_header(skb), tmp_hdr, hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683
684 ipv6_select_ident(skb, fh);
685 fh->nexthdr = nexthdr;
686 fh->reserved = 0;
687 fh->frag_off = htons(IP6_MF);
688 frag_id = fh->identification;
689
690 first_len = skb_pagelen(skb);
691 skb->data_len = first_len - skb_headlen(skb);
Herbert Xu29ffe1a2008-01-28 20:45:20 -0800692 skb->truesize -= truesizes;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 skb->len = first_len;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700694 ipv6_hdr(skb)->payload_len = htons(first_len -
695 sizeof(struct ipv6hdr));
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900696
697 dst_hold(&rt->u.dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698
699 for (;;) {
700 /* Prepare header of the next frame,
701 * before previous one went down. */
702 if (frag) {
703 frag->ip_summed = CHECKSUM_NONE;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300704 skb_reset_transport_header(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700706 __skb_push(frag, hlen);
707 skb_reset_network_header(frag);
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700708 memcpy(skb_network_header(frag), tmp_hdr,
709 hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 offset += skb->len - hlen - sizeof(struct frag_hdr);
711 fh->nexthdr = nexthdr;
712 fh->reserved = 0;
713 fh->frag_off = htons(offset);
714 if (frag->next != NULL)
715 fh->frag_off |= htons(IP6_MF);
716 fh->identification = frag_id;
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700717 ipv6_hdr(frag)->payload_len =
718 htons(frag->len -
719 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 ip6_copy_metadata(frag, skb);
721 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900722
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 err = output(skb);
Wei Dongdafee492006-08-02 13:41:21 -0700724 if(!err)
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900725 IP6_INC_STATS(ip6_dst_idev(&rt->u.dst), IPSTATS_MIB_FRAGCREATES);
Wei Dongdafee492006-08-02 13:41:21 -0700726
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 if (err || !frag)
728 break;
729
730 skb = frag;
731 frag = skb->next;
732 skb->next = NULL;
733 }
734
Jesper Juhla51482b2005-11-08 09:41:34 -0800735 kfree(tmp_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736
737 if (err == 0) {
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900738 IP6_INC_STATS(ip6_dst_idev(&rt->u.dst), IPSTATS_MIB_FRAGOKS);
739 dst_release(&rt->u.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
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900749 IP6_INC_STATS(ip6_dst_idev(&rt->u.dst), IPSTATS_MIB_FRAGFAILS);
750 dst_release(&rt->u.dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 return err;
752 }
753
754slow_path:
755 left = skb->len - hlen; /* Space per frame */
756 ptr = hlen; /* Where to start from */
757
758 /*
759 * Fragment the datagram.
760 */
761
762 *prevhdr = NEXTHDR_FRAGMENT;
763
764 /*
765 * Keep copying data until we run out.
766 */
767 while(left > 0) {
768 len = left;
769 /* IF: it doesn't fit, use 'mtu' - the data space left */
770 if (len > mtu)
771 len = mtu;
772 /* IF: we are not sending upto and including the packet end
773 then align the next start on an eight byte boundary */
774 if (len < left) {
775 len &= ~7;
776 }
777 /*
778 * Allocate buffer.
779 */
780
Johannes Bergf5184d22008-05-12 20:48:31 -0700781 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 -0700782 NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n");
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900783 IP6_INC_STATS(ip6_dst_idev(skb->dst),
784 IPSTATS_MIB_FRAGFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 err = -ENOMEM;
786 goto fail;
787 }
788
789 /*
790 * Set up data on packet
791 */
792
793 ip6_copy_metadata(frag, skb);
794 skb_reserve(frag, LL_RESERVED_SPACE(rt->u.dst.dev));
795 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700796 skb_reset_network_header(frag);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300797 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700798 frag->transport_header = (frag->network_header + hlen +
799 sizeof(struct frag_hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800
801 /*
802 * Charge the memory for the fragment to any owner
803 * it might possess
804 */
805 if (skb->sk)
806 skb_set_owner_w(frag, skb->sk);
807
808 /*
809 * Copy the packet header into the new buffer.
810 */
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -0300811 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812
813 /*
814 * Build fragment header.
815 */
816 fh->nexthdr = nexthdr;
817 fh->reserved = 0;
Yan Zhengf36d6ab2005-10-03 14:19:15 -0700818 if (!frag_id) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 ipv6_select_ident(skb, fh);
820 frag_id = fh->identification;
821 } else
822 fh->identification = frag_id;
823
824 /*
825 * Copy a block of the IP datagram.
826 */
Wei Yongjun8984e412007-08-21 20:59:08 -0700827 if (skb_copy_bits(skb, ptr, skb_transport_header(frag), len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 BUG();
829 left -= len;
830
831 fh->frag_off = htons(offset);
832 if (left > 0)
833 fh->frag_off |= htons(IP6_MF);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700834 ipv6_hdr(frag)->payload_len = htons(frag->len -
835 sizeof(struct ipv6hdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836
837 ptr += len;
838 offset += len;
839
840 /*
841 * Put this fragment into the sending queue.
842 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 err = output(frag);
844 if (err)
845 goto fail;
Wei Dongdafee492006-08-02 13:41:21 -0700846
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900847 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGCREATES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 }
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900849 IP6_INC_STATS(ip6_dst_idev(skb->dst),
850 IPSTATS_MIB_FRAGOKS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 return err;
853
854fail:
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +0900855 IP6_INC_STATS(ip6_dst_idev(skb->dst),
856 IPSTATS_MIB_FRAGFAILS);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900857 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 return err;
859}
860
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700861static inline int ip6_rt_check(struct rt6key *rt_key,
862 struct in6_addr *fl_addr,
863 struct in6_addr *addr_cache)
864{
865 return ((rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
866 (addr_cache == NULL || !ipv6_addr_equal(fl_addr, addr_cache)));
867}
868
Herbert Xu497c6152006-07-30 20:19:33 -0700869static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
870 struct dst_entry *dst,
871 struct flowi *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872{
Herbert Xu497c6152006-07-30 20:19:33 -0700873 struct ipv6_pinfo *np = inet6_sk(sk);
874 struct rt6_info *rt = (struct rt6_info *)dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875
Herbert Xu497c6152006-07-30 20:19:33 -0700876 if (!dst)
877 goto out;
878
879 /* Yes, checking route validity in not connected
880 * case is not very simple. Take into account,
881 * that we do not support routing by source, TOS,
882 * and MSG_DONTROUTE --ANK (980726)
883 *
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700884 * 1. ip6_rt_check(): If route was host route,
885 * check that cached destination is current.
Herbert Xu497c6152006-07-30 20:19:33 -0700886 * If it is network route, we still may
887 * check its validity using saved pointer
888 * to the last used address: daddr_cache.
889 * We do not want to save whole address now,
890 * (because main consumer of this service
891 * is tcp, which has not this problem),
892 * so that the last trick works only on connected
893 * sockets.
894 * 2. oif also should be the same.
895 */
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700896 if (ip6_rt_check(&rt->rt6i_dst, &fl->fl6_dst, np->daddr_cache) ||
YOSHIFUJI Hideaki8e1ef0a2006-08-29 17:15:09 -0700897#ifdef CONFIG_IPV6_SUBTREES
898 ip6_rt_check(&rt->rt6i_src, &fl->fl6_src, np->saddr_cache) ||
899#endif
YOSHIFUJI Hideakicf6b1982006-08-23 17:19:18 -0700900 (fl->oif && fl->oif != dst->dev->ifindex)) {
Herbert Xu497c6152006-07-30 20:19:33 -0700901 dst_release(dst);
902 dst = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 }
904
Herbert Xu497c6152006-07-30 20:19:33 -0700905out:
906 return dst;
907}
908
909static int ip6_dst_lookup_tail(struct sock *sk,
910 struct dst_entry **dst, struct flowi *fl)
911{
912 int err;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900913 struct net *net = sock_net(sk);
Herbert Xu497c6152006-07-30 20:19:33 -0700914
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 if (*dst == NULL)
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800916 *dst = ip6_route_output(net, sk, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917
918 if ((err = (*dst)->error))
919 goto out_err_release;
920
921 if (ipv6_addr_any(&fl->fl6_src)) {
YOSHIFUJI Hideaki5e5f3f02008-03-03 21:44:34 +0900922 err = ipv6_dev_get_saddr(ip6_dst_idev(*dst)->dev,
YOSHIFUJI Hideaki7cbca672008-03-25 09:37:42 +0900923 &fl->fl6_dst,
924 sk ? inet6_sk(sk)->srcprefs : 0,
925 &fl->fl6_src);
Olaf Hering44456d32005-07-27 11:45:17 -0700926 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 goto out_err_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928 }
929
Neil Horman95c385b2007-04-25 17:08:10 -0700930#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
931 /*
932 * Here if the dst entry we've looked up
933 * has a neighbour entry that is in the INCOMPLETE
934 * state and the src address from the flow is
935 * marked as OPTIMISTIC, we release the found
936 * dst entry and replace it instead with the
937 * dst entry of the nexthop router
938 */
939 if (!((*dst)->neighbour->nud_state & NUD_VALID)) {
940 struct inet6_ifaddr *ifp;
941 struct flowi fl_gw;
942 int redirect;
943
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800944 ifp = ipv6_get_ifaddr(net, &fl->fl6_src,
Daniel Lezcano1cab3da2008-01-10 22:44:09 -0800945 (*dst)->dev, 1);
Neil Horman95c385b2007-04-25 17:08:10 -0700946
947 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
948 if (ifp)
949 in6_ifa_put(ifp);
950
951 if (redirect) {
952 /*
953 * We need to get the dst entry for the
954 * default router instead
955 */
956 dst_release(*dst);
957 memcpy(&fl_gw, fl, sizeof(struct flowi));
958 memset(&fl_gw.fl6_dst, 0, sizeof(struct in6_addr));
Daniel Lezcano8a3edd82008-03-07 11:14:16 -0800959 *dst = ip6_route_output(net, sk, &fl_gw);
Neil Horman95c385b2007-04-25 17:08:10 -0700960 if ((err = (*dst)->error))
961 goto out_err_release;
962 }
963 }
964#endif
965
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 return 0;
967
968out_err_release:
Mitsuru Chinenca46f9c2007-12-05 22:31:47 -0800969 if (err == -ENETUNREACH)
970 IP6_INC_STATS_BH(NULL, IPSTATS_MIB_OUTNOROUTES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 dst_release(*dst);
972 *dst = NULL;
973 return err;
974}
Adrian Bunk34a0b3c2005-11-29 16:28:56 -0800975
Herbert Xu497c6152006-07-30 20:19:33 -0700976/**
977 * ip6_dst_lookup - perform route lookup on flow
978 * @sk: socket which provides route info
979 * @dst: pointer to dst_entry * for result
980 * @fl: flow to lookup
981 *
982 * This function performs a route lookup on the given flow.
983 *
984 * It returns zero on success, or a standard errno code on error.
985 */
986int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl)
987{
988 *dst = NULL;
989 return ip6_dst_lookup_tail(sk, dst, fl);
990}
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -0800991EXPORT_SYMBOL_GPL(ip6_dst_lookup);
992
Herbert Xu497c6152006-07-30 20:19:33 -0700993/**
994 * ip6_sk_dst_lookup - perform socket cached route lookup on flow
995 * @sk: socket which provides the dst cache and route info
996 * @dst: pointer to dst_entry * for result
997 * @fl: flow to lookup
998 *
999 * This function performs a route lookup on the given flow with the
1000 * possibility of using the cached route in the socket if it is valid.
1001 * It will take the socket dst lock when operating on the dst cache.
1002 * As a result, this function can only be used in process context.
1003 *
1004 * It returns zero on success, or a standard errno code on error.
1005 */
1006int ip6_sk_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl)
1007{
1008 *dst = NULL;
1009 if (sk) {
1010 *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
1011 *dst = ip6_sk_dst_check(sk, *dst, fl);
1012 }
1013
1014 return ip6_dst_lookup_tail(sk, dst, fl);
1015}
1016EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup);
1017
Adrian Bunk34a0b3c2005-11-29 16:28:56 -08001018static inline int ip6_ufo_append_data(struct sock *sk,
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001019 int getfrag(void *from, char *to, int offset, int len,
1020 int odd, struct sk_buff *skb),
1021 void *from, int length, int hh_len, int fragheaderlen,
1022 int transhdrlen, int mtu,unsigned int flags)
1023
1024{
1025 struct sk_buff *skb;
1026 int err;
1027
1028 /* There is support for UDP large send offload by network
1029 * device, so create one single skb packet containing complete
1030 * udp datagram
1031 */
1032 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) {
1033 skb = sock_alloc_send_skb(sk,
1034 hh_len + fragheaderlen + transhdrlen + 20,
1035 (flags & MSG_DONTWAIT), &err);
1036 if (skb == NULL)
1037 return -ENOMEM;
1038
1039 /* reserve space for Hardware header */
1040 skb_reserve(skb, hh_len);
1041
1042 /* create space for UDP/IP header */
1043 skb_put(skb,fragheaderlen + transhdrlen);
1044
1045 /* initialize network header pointer */
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001046 skb_reset_network_header(skb);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001047
1048 /* initialize protocol header pointer */
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001049 skb->transport_header = skb->network_header + fragheaderlen;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001050
Patrick McHardy84fa7932006-08-29 16:44:56 -07001051 skb->ip_summed = CHECKSUM_PARTIAL;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001052 skb->csum = 0;
1053 sk->sk_sndmsg_off = 0;
1054 }
1055
1056 err = skb_append_datato_frags(sk,skb, getfrag, from,
1057 (length - transhdrlen));
1058 if (!err) {
1059 struct frag_hdr fhdr;
1060
1061 /* specify the length of each IP datagram fragment*/
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001062 skb_shinfo(skb)->gso_size = mtu - fragheaderlen -
Herbert Xu79671682006-06-22 02:40:14 -07001063 sizeof(struct frag_hdr);
Herbert Xuf83ef8c2006-06-30 13:37:03 -07001064 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001065 ipv6_select_ident(skb, &fhdr);
1066 skb_shinfo(skb)->ip6_frag_id = fhdr.identification;
1067 __skb_queue_tail(&sk->sk_write_queue, skb);
1068
1069 return 0;
1070 }
1071 /* There is not enough support do UPD LSO,
1072 * so follow normal path
1073 */
1074 kfree_skb(skb);
1075
1076 return err;
1077}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001079int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
1080 int offset, int len, int odd, struct sk_buff *skb),
1081 void *from, int length, int transhdrlen,
1082 int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi *fl,
1083 struct rt6_info *rt, unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084{
1085 struct inet_sock *inet = inet_sk(sk);
1086 struct ipv6_pinfo *np = inet6_sk(sk);
1087 struct sk_buff *skb;
1088 unsigned int maxfraglen, fragheaderlen;
1089 int exthdrlen;
1090 int hh_len;
1091 int mtu;
1092 int copy;
1093 int err;
1094 int offset = 0;
1095 int csummode = CHECKSUM_NONE;
1096
1097 if (flags&MSG_PROBE)
1098 return 0;
1099 if (skb_queue_empty(&sk->sk_write_queue)) {
1100 /*
1101 * setup for corking
1102 */
1103 if (opt) {
1104 if (np->cork.opt == NULL) {
1105 np->cork.opt = kmalloc(opt->tot_len,
1106 sk->sk_allocation);
1107 if (unlikely(np->cork.opt == NULL))
1108 return -ENOBUFS;
1109 } else if (np->cork.opt->tot_len < opt->tot_len) {
1110 printk(KERN_DEBUG "ip6_append_data: invalid option length\n");
1111 return -EINVAL;
1112 }
1113 memcpy(np->cork.opt, opt, opt->tot_len);
1114 inet->cork.flags |= IPCORK_OPT;
1115 /* need source address above miyazawa*/
1116 }
1117 dst_hold(&rt->u.dst);
YOSHIFUJI Hideakic8cdaf92008-03-10 04:30:37 -04001118 inet->cork.dst = &rt->u.dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 inet->cork.fl = *fl;
1120 np->cork.hop_limit = hlimit;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001121 np->cork.tclass = tclass;
John Heffner628a5c52007-04-20 15:53:27 -07001122 mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ?
1123 rt->u.dst.dev->mtu : dst_mtu(rt->u.dst.path);
Dave Jonesc7503602006-03-20 22:44:52 -08001124 if (np->frag_size < mtu) {
YOSHIFUJI Hideakid91675f2006-02-24 13:18:33 -08001125 if (np->frag_size)
1126 mtu = np->frag_size;
1127 }
1128 inet->cork.fragsize = mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 if (dst_allfrag(rt->u.dst.path))
1130 inet->cork.flags |= IPCORK_ALLFRAG;
1131 inet->cork.length = 0;
1132 sk->sk_sndmsg_page = NULL;
1133 sk->sk_sndmsg_off = 0;
Herbert Xu014889422007-11-13 21:32:26 -08001134 exthdrlen = rt->u.dst.header_len + (opt ? opt->opt_flen : 0) -
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001135 rt->rt6i_nfheader_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136 length += exthdrlen;
1137 transhdrlen += exthdrlen;
1138 } else {
YOSHIFUJI Hideakic8cdaf92008-03-10 04:30:37 -04001139 rt = (struct rt6_info *)inet->cork.dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 fl = &inet->cork.fl;
1141 if (inet->cork.flags & IPCORK_OPT)
1142 opt = np->cork.opt;
1143 transhdrlen = 0;
1144 exthdrlen = 0;
1145 mtu = inet->cork.fragsize;
1146 }
1147
1148 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
1149
Masahide NAKAMURAa1b05142007-12-20 20:41:12 -08001150 fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
Herbert Xub4ce9272007-11-13 21:33:32 -08001151 (opt ? opt->opt_nflen : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr);
1153
1154 if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
1155 if (inet->cork.length + length > sizeof(struct ipv6hdr) + IPV6_MAXPLEN - fragheaderlen) {
1156 ipv6_local_error(sk, EMSGSIZE, fl, mtu-exthdrlen);
1157 return -EMSGSIZE;
1158 }
1159 }
1160
1161 /*
1162 * Let's try using as much space as possible.
1163 * Use MTU if total length of the message fits into the MTU.
1164 * Otherwise, we need to reserve fragment header and
1165 * fragment alignment (= 8-15 octects, in total).
1166 *
1167 * Note that we may need to "move" the data from the tail of
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001168 * of the buffer to the new fragment when we split
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 * the message.
1170 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001171 * FIXME: It may be fragmented into multiple chunks
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 * at once if non-fragmentable extension headers
1173 * are too large.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001174 * --yoshfuji
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 */
1176
1177 inet->cork.length += length;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001178 if (((length > mtu) && (sk->sk_protocol == IPPROTO_UDP)) &&
1179 (rt->u.dst.dev->features & NETIF_F_UFO)) {
1180
Patrick McHardybaa829d2006-03-12 20:35:12 -08001181 err = ip6_ufo_append_data(sk, getfrag, from, length, hh_len,
1182 fragheaderlen, transhdrlen, mtu,
1183 flags);
1184 if (err)
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001185 goto error;
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001186 return 0;
1187 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188
1189 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL)
1190 goto alloc_new_skb;
1191
1192 while (length > 0) {
1193 /* Check if the remaining data fits into current packet. */
1194 copy = (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
1195 if (copy < length)
1196 copy = maxfraglen - skb->len;
1197
1198 if (copy <= 0) {
1199 char *data;
1200 unsigned int datalen;
1201 unsigned int fraglen;
1202 unsigned int fraggap;
1203 unsigned int alloclen;
1204 struct sk_buff *skb_prev;
1205alloc_new_skb:
1206 skb_prev = skb;
1207
1208 /* There's no room in the current skb */
1209 if (skb_prev)
1210 fraggap = skb_prev->len - maxfraglen;
1211 else
1212 fraggap = 0;
1213
1214 /*
1215 * If remaining data exceeds the mtu,
1216 * we know we need more fragment(s).
1217 */
1218 datalen = length + fraggap;
1219 if (datalen > (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1220 datalen = maxfraglen - fragheaderlen;
1221
1222 fraglen = datalen + fragheaderlen;
1223 if ((flags & MSG_MORE) &&
1224 !(rt->u.dst.dev->features&NETIF_F_SG))
1225 alloclen = mtu;
1226 else
1227 alloclen = datalen + fragheaderlen;
1228
1229 /*
1230 * The last fragment gets additional space at tail.
1231 * Note: we overallocate on fragments with MSG_MODE
1232 * because we have no idea if we're the last one.
1233 */
1234 if (datalen == length + fraggap)
1235 alloclen += rt->u.dst.trailer_len;
1236
1237 /*
1238 * We just reserve space for fragment header.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001239 * Note: this may be overallocation if the message
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 * (without MSG_MORE) fits into the MTU.
1241 */
1242 alloclen += sizeof(struct frag_hdr);
1243
1244 if (transhdrlen) {
1245 skb = sock_alloc_send_skb(sk,
1246 alloclen + hh_len,
1247 (flags & MSG_DONTWAIT), &err);
1248 } else {
1249 skb = NULL;
1250 if (atomic_read(&sk->sk_wmem_alloc) <=
1251 2 * sk->sk_sndbuf)
1252 skb = sock_wmalloc(sk,
1253 alloclen + hh_len, 1,
1254 sk->sk_allocation);
1255 if (unlikely(skb == NULL))
1256 err = -ENOBUFS;
1257 }
1258 if (skb == NULL)
1259 goto error;
1260 /*
1261 * Fill in the control structures
1262 */
1263 skb->ip_summed = csummode;
1264 skb->csum = 0;
1265 /* reserve for fragmentation */
1266 skb_reserve(skb, hh_len+sizeof(struct frag_hdr));
1267
1268 /*
1269 * Find where to start putting bytes
1270 */
1271 data = skb_put(skb, fraglen);
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001272 skb_set_network_header(skb, exthdrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 data += fragheaderlen;
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001274 skb->transport_header = (skb->network_header +
1275 fragheaderlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 if (fraggap) {
1277 skb->csum = skb_copy_and_csum_bits(
1278 skb_prev, maxfraglen,
1279 data + transhdrlen, fraggap, 0);
1280 skb_prev->csum = csum_sub(skb_prev->csum,
1281 skb->csum);
1282 data += fraggap;
Herbert Xue9fa4f72006-08-13 20:12:58 -07001283 pskb_trim_unique(skb_prev, maxfraglen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 }
1285 copy = datalen - transhdrlen - fraggap;
1286 if (copy < 0) {
1287 err = -EINVAL;
1288 kfree_skb(skb);
1289 goto error;
1290 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1291 err = -EFAULT;
1292 kfree_skb(skb);
1293 goto error;
1294 }
1295
1296 offset += copy;
1297 length -= datalen - fraggap;
1298 transhdrlen = 0;
1299 exthdrlen = 0;
1300 csummode = CHECKSUM_NONE;
1301
1302 /*
1303 * Put the packet on the pending queue
1304 */
1305 __skb_queue_tail(&sk->sk_write_queue, skb);
1306 continue;
1307 }
1308
1309 if (copy > length)
1310 copy = length;
1311
1312 if (!(rt->u.dst.dev->features&NETIF_F_SG)) {
1313 unsigned int off;
1314
1315 off = skb->len;
1316 if (getfrag(from, skb_put(skb, copy),
1317 offset, copy, off, skb) < 0) {
1318 __skb_trim(skb, off);
1319 err = -EFAULT;
1320 goto error;
1321 }
1322 } else {
1323 int i = skb_shinfo(skb)->nr_frags;
1324 skb_frag_t *frag = &skb_shinfo(skb)->frags[i-1];
1325 struct page *page = sk->sk_sndmsg_page;
1326 int off = sk->sk_sndmsg_off;
1327 unsigned int left;
1328
1329 if (page && (left = PAGE_SIZE - off) > 0) {
1330 if (copy >= left)
1331 copy = left;
1332 if (page != frag->page) {
1333 if (i == MAX_SKB_FRAGS) {
1334 err = -EMSGSIZE;
1335 goto error;
1336 }
1337 get_page(page);
1338 skb_fill_page_desc(skb, i, page, sk->sk_sndmsg_off, 0);
1339 frag = &skb_shinfo(skb)->frags[i];
1340 }
1341 } else if(i < MAX_SKB_FRAGS) {
1342 if (copy > PAGE_SIZE)
1343 copy = PAGE_SIZE;
1344 page = alloc_pages(sk->sk_allocation, 0);
1345 if (page == NULL) {
1346 err = -ENOMEM;
1347 goto error;
1348 }
1349 sk->sk_sndmsg_page = page;
1350 sk->sk_sndmsg_off = 0;
1351
1352 skb_fill_page_desc(skb, i, page, 0, 0);
1353 frag = &skb_shinfo(skb)->frags[i];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 } else {
1355 err = -EMSGSIZE;
1356 goto error;
1357 }
1358 if (getfrag(from, page_address(frag->page)+frag->page_offset+frag->size, offset, copy, skb->len, skb) < 0) {
1359 err = -EFAULT;
1360 goto error;
1361 }
1362 sk->sk_sndmsg_off += copy;
1363 frag->size += copy;
1364 skb->len += copy;
1365 skb->data_len += copy;
Herbert Xuf945fa72008-01-22 22:39:26 -08001366 skb->truesize += copy;
1367 atomic_add(copy, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 }
1369 offset += copy;
1370 length -= copy;
1371 }
1372 return 0;
1373error:
1374 inet->cork.length -= length;
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +09001375 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 return err;
1377}
1378
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001379static void ip6_cork_release(struct inet_sock *inet, struct ipv6_pinfo *np)
1380{
1381 inet->cork.flags &= ~IPCORK_OPT;
1382 kfree(np->cork.opt);
1383 np->cork.opt = NULL;
YOSHIFUJI Hideakic8cdaf92008-03-10 04:30:37 -04001384 if (inet->cork.dst) {
1385 dst_release(inet->cork.dst);
1386 inet->cork.dst = NULL;
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001387 inet->cork.flags &= ~IPCORK_ALLFRAG;
1388 }
1389 memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
1390}
1391
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392int ip6_push_pending_frames(struct sock *sk)
1393{
1394 struct sk_buff *skb, *tmp_skb;
1395 struct sk_buff **tail_skb;
1396 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
1397 struct inet_sock *inet = inet_sk(sk);
1398 struct ipv6_pinfo *np = inet6_sk(sk);
1399 struct ipv6hdr *hdr;
1400 struct ipv6_txoptions *opt = np->cork.opt;
YOSHIFUJI Hideakic8cdaf92008-03-10 04:30:37 -04001401 struct rt6_info *rt = (struct rt6_info *)inet->cork.dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 struct flowi *fl = &inet->cork.fl;
1403 unsigned char proto = fl->proto;
1404 int err = 0;
1405
1406 if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL)
1407 goto out;
1408 tail_skb = &(skb_shinfo(skb)->frag_list);
1409
1410 /* move skb->data to ip header from ext header */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001411 if (skb->data < skb_network_header(skb))
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001412 __skb_pull(skb, skb_network_offset(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413 while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) {
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001414 __skb_pull(tmp_skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415 *tail_skb = tmp_skb;
1416 tail_skb = &(tmp_skb->next);
1417 skb->len += tmp_skb->len;
1418 skb->data_len += tmp_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 skb->truesize += tmp_skb->truesize;
1420 __sock_put(tmp_skb->sk);
1421 tmp_skb->destructor = NULL;
1422 tmp_skb->sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423 }
1424
Herbert Xu28a89452008-02-12 18:07:27 -08001425 /* Allow local fragmentation. */
Herbert Xub5c15fc2008-02-14 23:49:37 -08001426 if (np->pmtudisc < IPV6_PMTUDISC_DO)
Herbert Xu28a89452008-02-12 18:07:27 -08001427 skb->local_df = 1;
1428
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 ipv6_addr_copy(final_dst, &fl->fl6_dst);
Arnaldo Carvalho de Melocfe1fc72007-03-16 17:26:39 -03001430 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431 if (opt && opt->opt_flen)
1432 ipv6_push_frag_opts(skb, opt, &proto);
1433 if (opt && opt->opt_nflen)
1434 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1435
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -07001436 skb_push(skb, sizeof(struct ipv6hdr));
1437 skb_reset_network_header(skb);
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -07001438 hdr = ipv6_hdr(skb);
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001439
Al Viro90bcaf72006-11-08 00:25:17 -08001440 *(__be32*)hdr = fl->fl6_flowlabel |
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +09001441 htonl(0x60000000 | ((int)np->cork.tclass << 20));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 hdr->hop_limit = np->cork.hop_limit;
1444 hdr->nexthdr = proto;
1445 ipv6_addr_copy(&hdr->saddr, &fl->fl6_src);
1446 ipv6_addr_copy(&hdr->daddr, final_dst);
1447
Patrick McHardya2c20642006-01-08 22:37:26 -08001448 skb->priority = sk->sk_priority;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001449 skb->mark = sk->sk_mark;
Patrick McHardya2c20642006-01-08 22:37:26 -08001450
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451 skb->dst = dst_clone(&rt->u.dst);
YOSHIFUJI Hideakia11d2062006-11-04 20:11:37 +09001452 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
David L Stevens14878f72007-09-16 16:52:35 -07001453 if (proto == IPPROTO_ICMPV6) {
1454 struct inet6_dev *idev = ip6_dst_idev(skb->dst);
1455
1456 ICMP6MSGOUT_INC_STATS_BH(idev, icmp6_hdr(skb)->icmp6_type);
1457 ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTMSGS);
1458 }
1459
Herbert Xuef76bc22008-01-11 19:15:08 -08001460 err = ip6_local_out(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 if (err) {
1462 if (err > 0)
Herbert Xu3320da82005-04-19 22:32:22 -07001463 err = np->recverr ? net_xmit_errno(err) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 if (err)
1465 goto error;
1466 }
1467
1468out:
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001469 ip6_cork_release(inet, np);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 return err;
1471error:
1472 goto out;
1473}
1474
1475void ip6_flush_pending_frames(struct sock *sk)
1476{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477 struct sk_buff *skb;
1478
1479 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) {
YOSHIFUJI Hideakie1f522082007-09-11 11:31:43 +02001480 if (skb->dst)
1481 IP6_INC_STATS(ip6_dst_idev(skb->dst),
1482 IPSTATS_MIB_OUTDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 kfree_skb(skb);
1484 }
1485
Pavel Emelyanovbf138862007-11-05 21:04:31 -08001486 ip6_cork_release(inet_sk(sk), inet6_sk(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487}