Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * IPv6 output functions | |
1ab1457c | 3 | * Linux INET6 implementation |
1da177e4 LT |
4 | * |
5 | * Authors: | |
1ab1457c | 6 | * Pedro Roque <roque@di.fc.ul.pt> |
1da177e4 | 7 | * |
1da177e4 LT |
8 | * 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 | ||
1da177e4 | 29 | #include <linux/errno.h> |
ef76bc23 | 30 | #include <linux/kernel.h> |
1da177e4 LT |
31 | #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> | |
b59f45d0 | 39 | #include <linux/module.h> |
5a0e3ad6 | 40 | #include <linux/slab.h> |
1da177e4 LT |
41 | |
42 | #include <linux/netfilter.h> | |
43 | #include <linux/netfilter_ipv6.h> | |
44 | ||
45 | #include <net/sock.h> | |
46 | #include <net/snmp.h> | |
47 | ||
48 | #include <net/ipv6.h> | |
49 | #include <net/ndisc.h> | |
50 | #include <net/protocol.h> | |
51 | #include <net/ip6_route.h> | |
52 | #include <net/addrconf.h> | |
53 | #include <net/rawv6.h> | |
54 | #include <net/icmp.h> | |
55 | #include <net/xfrm.h> | |
56 | #include <net/checksum.h> | |
7bc570c8 | 57 | #include <linux/mroute6.h> |
1da177e4 | 58 | |
ad0081e4 | 59 | int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *)); |
1da177e4 | 60 | |
ef76bc23 HX |
61 | int __ip6_local_out(struct sk_buff *skb) |
62 | { | |
63 | int len; | |
64 | ||
65 | len = skb->len - sizeof(struct ipv6hdr); | |
66 | if (len > IPV6_MAXPLEN) | |
67 | len = 0; | |
68 | ipv6_hdr(skb)->payload_len = htons(len); | |
69 | ||
b2e0b385 JE |
70 | return nf_hook(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL, |
71 | skb_dst(skb)->dev, dst_output); | |
ef76bc23 HX |
72 | } |
73 | ||
74 | int ip6_local_out(struct sk_buff *skb) | |
75 | { | |
76 | int err; | |
77 | ||
78 | err = __ip6_local_out(skb); | |
79 | if (likely(err == 1)) | |
80 | err = dst_output(skb); | |
81 | ||
82 | return err; | |
83 | } | |
84 | EXPORT_SYMBOL_GPL(ip6_local_out); | |
85 | ||
9e508490 | 86 | static int ip6_finish_output2(struct sk_buff *skb) |
1da177e4 | 87 | { |
adf30907 | 88 | struct dst_entry *dst = skb_dst(skb); |
1da177e4 | 89 | struct net_device *dev = dst->dev; |
f6b72b62 | 90 | struct neighbour *neigh; |
97cac082 | 91 | struct rt6_info *rt; |
1da177e4 LT |
92 | |
93 | skb->protocol = htons(ETH_P_IPV6); | |
94 | skb->dev = dev; | |
95 | ||
0660e03f | 96 | if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) { |
adf30907 | 97 | struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb)); |
1da177e4 | 98 | |
7ad6848c | 99 | if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(skb->sk) && |
d1db275d | 100 | ((mroute6_socket(dev_net(dev), skb) && |
bd91b8bf | 101 | !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) || |
7bc570c8 YH |
102 | ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr, |
103 | &ipv6_hdr(skb)->saddr))) { | |
1da177e4 LT |
104 | struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC); |
105 | ||
106 | /* Do not check for IFF_ALLMULTI; multicast routing | |
107 | is not supported in any case. | |
108 | */ | |
109 | if (newskb) | |
b2e0b385 JE |
110 | NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING, |
111 | newskb, NULL, newskb->dev, | |
95603e22 | 112 | dev_loopback_xmit); |
1da177e4 | 113 | |
0660e03f | 114 | if (ipv6_hdr(skb)->hop_limit == 0) { |
3bd653c8 DL |
115 | IP6_INC_STATS(dev_net(dev), idev, |
116 | IPSTATS_MIB_OUTDISCARDS); | |
1da177e4 LT |
117 | kfree_skb(skb); |
118 | return 0; | |
119 | } | |
120 | } | |
121 | ||
edf391ff NH |
122 | IP6_UPD_PO_STATS(dev_net(dev), idev, IPSTATS_MIB_OUTMCAST, |
123 | skb->len); | |
1da177e4 LT |
124 | } |
125 | ||
97cac082 DM |
126 | rt = (struct rt6_info *) dst; |
127 | neigh = rt->n; | |
fdd6681d AW |
128 | if (neigh) |
129 | return dst_neigh_output(dst, neigh, skb); | |
05e3aa09 | 130 | |
9e508490 JE |
131 | IP6_INC_STATS_BH(dev_net(dst->dev), |
132 | ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES); | |
133 | kfree_skb(skb); | |
134 | return -EINVAL; | |
1da177e4 LT |
135 | } |
136 | ||
9e508490 JE |
137 | static int ip6_finish_output(struct sk_buff *skb) |
138 | { | |
139 | if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) || | |
140 | dst_allfrag(skb_dst(skb))) | |
141 | return ip6_fragment(skb, ip6_finish_output2); | |
142 | else | |
143 | return ip6_finish_output2(skb); | |
144 | } | |
145 | ||
1da177e4 LT |
146 | int ip6_output(struct sk_buff *skb) |
147 | { | |
9e508490 | 148 | struct net_device *dev = skb_dst(skb)->dev; |
adf30907 | 149 | struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb)); |
778d80be | 150 | if (unlikely(idev->cnf.disable_ipv6)) { |
9e508490 | 151 | IP6_INC_STATS(dev_net(dev), idev, |
3bd653c8 | 152 | IPSTATS_MIB_OUTDISCARDS); |
778d80be YH |
153 | kfree_skb(skb); |
154 | return 0; | |
155 | } | |
156 | ||
9c6eb28a JE |
157 | return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING, skb, NULL, dev, |
158 | ip6_finish_output, | |
159 | !(IP6CB(skb)->flags & IP6SKB_REROUTED)); | |
1da177e4 LT |
160 | } |
161 | ||
1da177e4 | 162 | /* |
b5d43998 | 163 | * xmit an sk_buff (used by TCP, SCTP and DCCP) |
1da177e4 LT |
164 | */ |
165 | ||
4c9483b2 | 166 | int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6, |
b903d324 | 167 | struct ipv6_txoptions *opt, int tclass) |
1da177e4 | 168 | { |
3bd653c8 | 169 | struct net *net = sock_net(sk); |
b30bd282 | 170 | struct ipv6_pinfo *np = inet6_sk(sk); |
4c9483b2 | 171 | struct in6_addr *first_hop = &fl6->daddr; |
adf30907 | 172 | struct dst_entry *dst = skb_dst(skb); |
1da177e4 | 173 | struct ipv6hdr *hdr; |
4c9483b2 | 174 | u8 proto = fl6->flowi6_proto; |
1da177e4 | 175 | int seg_len = skb->len; |
e651f03a | 176 | int hlimit = -1; |
1da177e4 LT |
177 | u32 mtu; |
178 | ||
179 | if (opt) { | |
c2636b4d | 180 | unsigned int head_room; |
1da177e4 LT |
181 | |
182 | /* First: exthdrs may take lots of space (~8K for now) | |
183 | MAX_HEADER is not enough. | |
184 | */ | |
185 | head_room = opt->opt_nflen + opt->opt_flen; | |
186 | seg_len += head_room; | |
187 | head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev); | |
188 | ||
189 | if (skb_headroom(skb) < head_room) { | |
190 | struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room); | |
a11d206d | 191 | if (skb2 == NULL) { |
adf30907 | 192 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d YH |
193 | IPSTATS_MIB_OUTDISCARDS); |
194 | kfree_skb(skb); | |
1da177e4 LT |
195 | return -ENOBUFS; |
196 | } | |
808db80a | 197 | consume_skb(skb); |
a11d206d | 198 | skb = skb2; |
83d7eb29 | 199 | skb_set_owner_w(skb, sk); |
1da177e4 LT |
200 | } |
201 | if (opt->opt_flen) | |
202 | ipv6_push_frag_opts(skb, opt, &proto); | |
203 | if (opt->opt_nflen) | |
204 | ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop); | |
205 | } | |
206 | ||
e2d1bca7 ACM |
207 | skb_push(skb, sizeof(struct ipv6hdr)); |
208 | skb_reset_network_header(skb); | |
0660e03f | 209 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
210 | |
211 | /* | |
212 | * Fill in the IPv6 header | |
213 | */ | |
b903d324 | 214 | if (np) |
1da177e4 LT |
215 | hlimit = np->hop_limit; |
216 | if (hlimit < 0) | |
6b75d090 | 217 | hlimit = ip6_dst_hoplimit(dst); |
1da177e4 | 218 | |
4c9483b2 | 219 | *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | fl6->flowlabel; |
41a1f8ea | 220 | |
1da177e4 LT |
221 | hdr->payload_len = htons(seg_len); |
222 | hdr->nexthdr = proto; | |
223 | hdr->hop_limit = hlimit; | |
224 | ||
4e3fd7a0 AD |
225 | hdr->saddr = fl6->saddr; |
226 | hdr->daddr = *first_hop; | |
1da177e4 | 227 | |
a2c2064f | 228 | skb->priority = sk->sk_priority; |
4a19ec58 | 229 | skb->mark = sk->sk_mark; |
a2c2064f | 230 | |
1da177e4 | 231 | mtu = dst_mtu(dst); |
283d07ac | 232 | if ((skb->len <= mtu) || skb->local_df || skb_is_gso(skb)) { |
adf30907 | 233 | IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)), |
edf391ff | 234 | IPSTATS_MIB_OUT, skb->len); |
b2e0b385 JE |
235 | return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL, |
236 | dst->dev, dst_output); | |
1da177e4 LT |
237 | } |
238 | ||
e87cc472 | 239 | net_dbg_ratelimited("IPv6: sending pkt_too_big to self\n"); |
1da177e4 | 240 | skb->dev = dst->dev; |
3ffe533c | 241 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); |
adf30907 | 242 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
243 | kfree_skb(skb); |
244 | return -EMSGSIZE; | |
245 | } | |
246 | ||
7159039a YH |
247 | EXPORT_SYMBOL(ip6_xmit); |
248 | ||
1da177e4 LT |
249 | /* |
250 | * To avoid extra problems ND packets are send through this | |
251 | * routine. It's code duplication but I really want to avoid | |
252 | * extra checks since ipv6_build_header is used by TCP (which | |
253 | * is for us performance critical) | |
254 | */ | |
255 | ||
256 | int ip6_nd_hdr(struct sock *sk, struct sk_buff *skb, struct net_device *dev, | |
9acd9f3a | 257 | const struct in6_addr *saddr, const struct in6_addr *daddr, |
1da177e4 LT |
258 | int proto, int len) |
259 | { | |
260 | struct ipv6_pinfo *np = inet6_sk(sk); | |
261 | struct ipv6hdr *hdr; | |
1da177e4 LT |
262 | |
263 | skb->protocol = htons(ETH_P_IPV6); | |
264 | skb->dev = dev; | |
265 | ||
55f79cc0 ACM |
266 | skb_reset_network_header(skb); |
267 | skb_put(skb, sizeof(struct ipv6hdr)); | |
0660e03f | 268 | hdr = ipv6_hdr(skb); |
1da177e4 | 269 | |
ae08e1f0 | 270 | *(__be32*)hdr = htonl(0x60000000); |
1da177e4 LT |
271 | |
272 | hdr->payload_len = htons(len); | |
273 | hdr->nexthdr = proto; | |
274 | hdr->hop_limit = np->hop_limit; | |
275 | ||
4e3fd7a0 AD |
276 | hdr->saddr = *saddr; |
277 | hdr->daddr = *daddr; | |
1da177e4 LT |
278 | |
279 | return 0; | |
280 | } | |
281 | ||
282 | static int ip6_call_ra_chain(struct sk_buff *skb, int sel) | |
283 | { | |
284 | struct ip6_ra_chain *ra; | |
285 | struct sock *last = NULL; | |
286 | ||
287 | read_lock(&ip6_ra_lock); | |
288 | for (ra = ip6_ra_chain; ra; ra = ra->next) { | |
289 | struct sock *sk = ra->sk; | |
0bd1b59b AM |
290 | if (sk && ra->sel == sel && |
291 | (!sk->sk_bound_dev_if || | |
292 | sk->sk_bound_dev_if == skb->dev->ifindex)) { | |
1da177e4 LT |
293 | if (last) { |
294 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
295 | if (skb2) | |
296 | rawv6_rcv(last, skb2); | |
297 | } | |
298 | last = sk; | |
299 | } | |
300 | } | |
301 | ||
302 | if (last) { | |
303 | rawv6_rcv(last, skb); | |
304 | read_unlock(&ip6_ra_lock); | |
305 | return 1; | |
306 | } | |
307 | read_unlock(&ip6_ra_lock); | |
308 | return 0; | |
309 | } | |
310 | ||
e21e0b5f VN |
311 | static int ip6_forward_proxy_check(struct sk_buff *skb) |
312 | { | |
0660e03f | 313 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
e21e0b5f | 314 | u8 nexthdr = hdr->nexthdr; |
75f2811c | 315 | __be16 frag_off; |
e21e0b5f VN |
316 | int offset; |
317 | ||
318 | if (ipv6_ext_hdr(nexthdr)) { | |
75f2811c | 319 | offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off); |
e21e0b5f VN |
320 | if (offset < 0) |
321 | return 0; | |
322 | } else | |
323 | offset = sizeof(struct ipv6hdr); | |
324 | ||
325 | if (nexthdr == IPPROTO_ICMPV6) { | |
326 | struct icmp6hdr *icmp6; | |
327 | ||
d56f90a7 ACM |
328 | if (!pskb_may_pull(skb, (skb_network_header(skb) + |
329 | offset + 1 - skb->data))) | |
e21e0b5f VN |
330 | return 0; |
331 | ||
d56f90a7 | 332 | icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset); |
e21e0b5f VN |
333 | |
334 | switch (icmp6->icmp6_type) { | |
335 | case NDISC_ROUTER_SOLICITATION: | |
336 | case NDISC_ROUTER_ADVERTISEMENT: | |
337 | case NDISC_NEIGHBOUR_SOLICITATION: | |
338 | case NDISC_NEIGHBOUR_ADVERTISEMENT: | |
339 | case NDISC_REDIRECT: | |
340 | /* For reaction involving unicast neighbor discovery | |
341 | * message destined to the proxied address, pass it to | |
342 | * input function. | |
343 | */ | |
344 | return 1; | |
345 | default: | |
346 | break; | |
347 | } | |
348 | } | |
349 | ||
74553b09 VN |
350 | /* |
351 | * The proxying router can't forward traffic sent to a link-local | |
352 | * address, so signal the sender and discard the packet. This | |
353 | * behavior is clarified by the MIPv6 specification. | |
354 | */ | |
355 | if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) { | |
356 | dst_link_failure(skb); | |
357 | return -1; | |
358 | } | |
359 | ||
e21e0b5f VN |
360 | return 0; |
361 | } | |
362 | ||
1da177e4 LT |
363 | static inline int ip6_forward_finish(struct sk_buff *skb) |
364 | { | |
365 | return dst_output(skb); | |
366 | } | |
367 | ||
368 | int ip6_forward(struct sk_buff *skb) | |
369 | { | |
adf30907 | 370 | struct dst_entry *dst = skb_dst(skb); |
0660e03f | 371 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
1da177e4 | 372 | struct inet6_skb_parm *opt = IP6CB(skb); |
c346dca1 | 373 | struct net *net = dev_net(dst->dev); |
14f3ad6f | 374 | u32 mtu; |
1ab1457c | 375 | |
53b7997f | 376 | if (net->ipv6.devconf_all->forwarding == 0) |
1da177e4 LT |
377 | goto error; |
378 | ||
4497b076 BH |
379 | if (skb_warn_if_lro(skb)) |
380 | goto drop; | |
381 | ||
1da177e4 | 382 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) { |
3bd653c8 | 383 | IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS); |
1da177e4 LT |
384 | goto drop; |
385 | } | |
386 | ||
72b43d08 AK |
387 | if (skb->pkt_type != PACKET_HOST) |
388 | goto drop; | |
389 | ||
35fc92a9 | 390 | skb_forward_csum(skb); |
1da177e4 LT |
391 | |
392 | /* | |
393 | * We DO NOT make any processing on | |
394 | * RA packets, pushing them to user level AS IS | |
395 | * without ane WARRANTY that application will be able | |
396 | * to interpret them. The reason is that we | |
397 | * cannot make anything clever here. | |
398 | * | |
399 | * We are not end-node, so that if packet contains | |
400 | * AH/ESP, we cannot make anything. | |
401 | * Defragmentation also would be mistake, RA packets | |
402 | * cannot be fragmented, because there is no warranty | |
403 | * that different fragments will go along one path. --ANK | |
404 | */ | |
405 | if (opt->ra) { | |
d56f90a7 | 406 | u8 *ptr = skb_network_header(skb) + opt->ra; |
1da177e4 LT |
407 | if (ip6_call_ra_chain(skb, (ptr[2]<<8) + ptr[3])) |
408 | return 0; | |
409 | } | |
410 | ||
411 | /* | |
412 | * check and decrement ttl | |
413 | */ | |
414 | if (hdr->hop_limit <= 1) { | |
415 | /* Force OUTPUT device used as source address */ | |
416 | skb->dev = dst->dev; | |
3ffe533c | 417 | icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0); |
483a47d2 DL |
418 | IP6_INC_STATS_BH(net, |
419 | ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS); | |
1da177e4 LT |
420 | |
421 | kfree_skb(skb); | |
422 | return -ETIMEDOUT; | |
423 | } | |
424 | ||
fbea49e1 | 425 | /* XXX: idev->cnf.proxy_ndp? */ |
53b7997f | 426 | if (net->ipv6.devconf_all->proxy_ndp && |
8a3edd80 | 427 | pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) { |
74553b09 VN |
428 | int proxied = ip6_forward_proxy_check(skb); |
429 | if (proxied > 0) | |
e21e0b5f | 430 | return ip6_input(skb); |
74553b09 | 431 | else if (proxied < 0) { |
3bd653c8 DL |
432 | IP6_INC_STATS(net, ip6_dst_idev(dst), |
433 | IPSTATS_MIB_INDISCARDS); | |
74553b09 VN |
434 | goto drop; |
435 | } | |
e21e0b5f VN |
436 | } |
437 | ||
1da177e4 | 438 | if (!xfrm6_route_forward(skb)) { |
3bd653c8 | 439 | IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS); |
1da177e4 LT |
440 | goto drop; |
441 | } | |
adf30907 | 442 | dst = skb_dst(skb); |
1da177e4 LT |
443 | |
444 | /* IPv6 specs say nothing about it, but it is clear that we cannot | |
445 | send redirects to source routed frames. | |
1e5dc146 | 446 | We don't send redirects to frames decapsulated from IPsec. |
1da177e4 | 447 | */ |
c45a3dfb | 448 | if (skb->dev == dst->dev && opt->srcrt == 0 && !skb_sec_path(skb)) { |
1da177e4 | 449 | struct in6_addr *target = NULL; |
fbfe95a4 | 450 | struct inet_peer *peer; |
1da177e4 | 451 | struct rt6_info *rt; |
1da177e4 LT |
452 | |
453 | /* | |
454 | * incoming and outgoing devices are the same | |
455 | * send a redirect. | |
456 | */ | |
457 | ||
458 | rt = (struct rt6_info *) dst; | |
c45a3dfb DM |
459 | if (rt->rt6i_flags & RTF_GATEWAY) |
460 | target = &rt->rt6i_gateway; | |
1da177e4 LT |
461 | else |
462 | target = &hdr->daddr; | |
463 | ||
1d861aa4 | 464 | peer = inet_getpeer_v6(net->ipv6.peers, &rt->rt6i_dst.addr, 1); |
92d86829 | 465 | |
1da177e4 LT |
466 | /* Limit redirects both by destination (here) |
467 | and by source (inside ndisc_send_redirect) | |
468 | */ | |
fbfe95a4 | 469 | if (inet_peer_xrlim_allow(peer, 1*HZ)) |
4991969a | 470 | ndisc_send_redirect(skb, target); |
1d861aa4 DM |
471 | if (peer) |
472 | inet_putpeer(peer); | |
5bb1ab09 DS |
473 | } else { |
474 | int addrtype = ipv6_addr_type(&hdr->saddr); | |
475 | ||
1da177e4 | 476 | /* This check is security critical. */ |
f81b2e7d YH |
477 | if (addrtype == IPV6_ADDR_ANY || |
478 | addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK)) | |
5bb1ab09 DS |
479 | goto error; |
480 | if (addrtype & IPV6_ADDR_LINKLOCAL) { | |
481 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, | |
3ffe533c | 482 | ICMPV6_NOT_NEIGHBOUR, 0); |
5bb1ab09 DS |
483 | goto error; |
484 | } | |
1da177e4 LT |
485 | } |
486 | ||
14f3ad6f UW |
487 | mtu = dst_mtu(dst); |
488 | if (mtu < IPV6_MIN_MTU) | |
489 | mtu = IPV6_MIN_MTU; | |
490 | ||
4cdd3408 PM |
491 | if ((!skb->local_df && skb->len > mtu && !skb_is_gso(skb)) || |
492 | (IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu)) { | |
1da177e4 LT |
493 | /* Again, force OUTPUT device used as source address */ |
494 | skb->dev = dst->dev; | |
14f3ad6f | 495 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); |
483a47d2 DL |
496 | IP6_INC_STATS_BH(net, |
497 | ip6_dst_idev(dst), IPSTATS_MIB_INTOOBIGERRORS); | |
498 | IP6_INC_STATS_BH(net, | |
499 | ip6_dst_idev(dst), IPSTATS_MIB_FRAGFAILS); | |
1da177e4 LT |
500 | kfree_skb(skb); |
501 | return -EMSGSIZE; | |
502 | } | |
503 | ||
504 | if (skb_cow(skb, dst->dev->hard_header_len)) { | |
3bd653c8 | 505 | IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
506 | goto drop; |
507 | } | |
508 | ||
0660e03f | 509 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
510 | |
511 | /* Mangling hops number delayed to point after skb COW */ | |
1ab1457c | 512 | |
1da177e4 LT |
513 | hdr->hop_limit--; |
514 | ||
483a47d2 | 515 | IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS); |
2d8dbb04 | 516 | IP6_ADD_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len); |
b2e0b385 | 517 | return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dst->dev, |
6e23ae2a | 518 | ip6_forward_finish); |
1da177e4 LT |
519 | |
520 | error: | |
483a47d2 | 521 | IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS); |
1da177e4 LT |
522 | drop: |
523 | kfree_skb(skb); | |
524 | return -EINVAL; | |
525 | } | |
526 | ||
527 | static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from) | |
528 | { | |
529 | to->pkt_type = from->pkt_type; | |
530 | to->priority = from->priority; | |
531 | to->protocol = from->protocol; | |
adf30907 ED |
532 | skb_dst_drop(to); |
533 | skb_dst_set(to, dst_clone(skb_dst(from))); | |
1da177e4 | 534 | to->dev = from->dev; |
82e91ffe | 535 | to->mark = from->mark; |
1da177e4 LT |
536 | |
537 | #ifdef CONFIG_NET_SCHED | |
538 | to->tc_index = from->tc_index; | |
539 | #endif | |
e7ac05f3 | 540 | nf_copy(to, from); |
ba9dda3a JK |
541 | #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \ |
542 | defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE) | |
543 | to->nf_trace = from->nf_trace; | |
544 | #endif | |
984bc16c | 545 | skb_copy_secmark(to, from); |
1da177e4 LT |
546 | } |
547 | ||
548 | int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr) | |
549 | { | |
550 | u16 offset = sizeof(struct ipv6hdr); | |
0660e03f ACM |
551 | struct ipv6_opt_hdr *exthdr = |
552 | (struct ipv6_opt_hdr *)(ipv6_hdr(skb) + 1); | |
27a884dc | 553 | unsigned int packet_len = skb->tail - skb->network_header; |
1da177e4 | 554 | int found_rhdr = 0; |
0660e03f | 555 | *nexthdr = &ipv6_hdr(skb)->nexthdr; |
1da177e4 LT |
556 | |
557 | while (offset + 1 <= packet_len) { | |
558 | ||
559 | switch (**nexthdr) { | |
560 | ||
561 | case NEXTHDR_HOP: | |
27637df9 | 562 | break; |
1da177e4 | 563 | case NEXTHDR_ROUTING: |
27637df9 MN |
564 | found_rhdr = 1; |
565 | break; | |
1da177e4 | 566 | case NEXTHDR_DEST: |
59fbb3a6 | 567 | #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) |
27637df9 MN |
568 | if (ipv6_find_tlv(skb, offset, IPV6_TLV_HAO) >= 0) |
569 | break; | |
570 | #endif | |
571 | if (found_rhdr) | |
572 | return offset; | |
1da177e4 LT |
573 | break; |
574 | default : | |
575 | return offset; | |
576 | } | |
27637df9 MN |
577 | |
578 | offset += ipv6_optlen(exthdr); | |
579 | *nexthdr = &exthdr->nexthdr; | |
d56f90a7 ACM |
580 | exthdr = (struct ipv6_opt_hdr *)(skb_network_header(skb) + |
581 | offset); | |
1da177e4 LT |
582 | } |
583 | ||
584 | return offset; | |
585 | } | |
586 | ||
87c48fa3 ED |
587 | void ipv6_select_ident(struct frag_hdr *fhdr, struct rt6_info *rt) |
588 | { | |
589 | static atomic_t ipv6_fragmentation_id; | |
590 | int old, new; | |
591 | ||
e688a604 | 592 | if (rt && !(rt->dst.flags & DST_NOPEER)) { |
1d861aa4 DM |
593 | struct inet_peer *peer; |
594 | struct net *net; | |
87c48fa3 | 595 | |
1d861aa4 DM |
596 | net = dev_net(rt->dst.dev); |
597 | peer = inet_getpeer_v6(net->ipv6.peers, &rt->rt6i_dst.addr, 1); | |
87c48fa3 ED |
598 | if (peer) { |
599 | fhdr->identification = htonl(inet_getid(peer, 0)); | |
1d861aa4 | 600 | inet_putpeer(peer); |
87c48fa3 ED |
601 | return; |
602 | } | |
603 | } | |
604 | do { | |
605 | old = atomic_read(&ipv6_fragmentation_id); | |
606 | new = old + 1; | |
607 | if (!new) | |
608 | new = 1; | |
609 | } while (atomic_cmpxchg(&ipv6_fragmentation_id, old, new) != old); | |
610 | fhdr->identification = htonl(new); | |
611 | } | |
612 | ||
ad0081e4 | 613 | int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *)) |
1da177e4 | 614 | { |
1da177e4 | 615 | struct sk_buff *frag; |
adf30907 | 616 | struct rt6_info *rt = (struct rt6_info*)skb_dst(skb); |
d91675f9 | 617 | struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL; |
1da177e4 LT |
618 | struct ipv6hdr *tmp_hdr; |
619 | struct frag_hdr *fh; | |
620 | unsigned int mtu, hlen, left, len; | |
a7ae1992 | 621 | int hroom, troom; |
ae08e1f0 | 622 | __be32 frag_id = 0; |
1da177e4 LT |
623 | int ptr, offset = 0, err=0; |
624 | u8 *prevhdr, nexthdr = 0; | |
adf30907 | 625 | struct net *net = dev_net(skb_dst(skb)->dev); |
1da177e4 | 626 | |
1da177e4 LT |
627 | hlen = ip6_find_1stfragopt(skb, &prevhdr); |
628 | nexthdr = *prevhdr; | |
629 | ||
628a5c56 | 630 | mtu = ip6_skb_dst_mtu(skb); |
b881ef76 JH |
631 | |
632 | /* We must not fragment if the socket is set to force MTU discovery | |
14f3ad6f | 633 | * or if the skb it not generated by a local socket. |
b881ef76 | 634 | */ |
4cdd3408 PM |
635 | if (unlikely(!skb->local_df && skb->len > mtu) || |
636 | (IP6CB(skb)->frag_max_size && | |
637 | IP6CB(skb)->frag_max_size > mtu)) { | |
a34a101e ED |
638 | if (skb->sk && dst_allfrag(skb_dst(skb))) |
639 | sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK); | |
640 | ||
adf30907 | 641 | skb->dev = skb_dst(skb)->dev; |
3ffe533c | 642 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); |
adf30907 | 643 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
3bd653c8 | 644 | IPSTATS_MIB_FRAGFAILS); |
b881ef76 JH |
645 | kfree_skb(skb); |
646 | return -EMSGSIZE; | |
647 | } | |
648 | ||
d91675f9 YH |
649 | if (np && np->frag_size < mtu) { |
650 | if (np->frag_size) | |
651 | mtu = np->frag_size; | |
652 | } | |
653 | mtu -= hlen + sizeof(struct frag_hdr); | |
1da177e4 | 654 | |
21dc3301 | 655 | if (skb_has_frag_list(skb)) { |
1da177e4 | 656 | int first_len = skb_pagelen(skb); |
3d13008e | 657 | struct sk_buff *frag2; |
1da177e4 LT |
658 | |
659 | if (first_len - hlen > mtu || | |
660 | ((first_len - hlen) & 7) || | |
661 | skb_cloned(skb)) | |
662 | goto slow_path; | |
663 | ||
4d9092bb | 664 | skb_walk_frags(skb, frag) { |
1da177e4 LT |
665 | /* Correct geometry. */ |
666 | if (frag->len > mtu || | |
667 | ((frag->len & 7) && frag->next) || | |
668 | skb_headroom(frag) < hlen) | |
3d13008e | 669 | goto slow_path_clean; |
1da177e4 | 670 | |
1da177e4 LT |
671 | /* Partially cloned skb? */ |
672 | if (skb_shared(frag)) | |
3d13008e | 673 | goto slow_path_clean; |
2fdba6b0 HX |
674 | |
675 | BUG_ON(frag->sk); | |
676 | if (skb->sk) { | |
2fdba6b0 HX |
677 | frag->sk = skb->sk; |
678 | frag->destructor = sock_wfree; | |
2fdba6b0 | 679 | } |
3d13008e | 680 | skb->truesize -= frag->truesize; |
1da177e4 LT |
681 | } |
682 | ||
683 | err = 0; | |
684 | offset = 0; | |
685 | frag = skb_shinfo(skb)->frag_list; | |
4d9092bb | 686 | skb_frag_list_init(skb); |
1da177e4 LT |
687 | /* BUILD HEADER */ |
688 | ||
9a217a1c | 689 | *prevhdr = NEXTHDR_FRAGMENT; |
d56f90a7 | 690 | tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC); |
1da177e4 | 691 | if (!tmp_hdr) { |
adf30907 | 692 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
3bd653c8 | 693 | IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
694 | return -ENOMEM; |
695 | } | |
696 | ||
1da177e4 LT |
697 | __skb_pull(skb, hlen); |
698 | fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr)); | |
e2d1bca7 ACM |
699 | __skb_push(skb, hlen); |
700 | skb_reset_network_header(skb); | |
d56f90a7 | 701 | memcpy(skb_network_header(skb), tmp_hdr, hlen); |
1da177e4 | 702 | |
87c48fa3 | 703 | ipv6_select_ident(fh, rt); |
1da177e4 LT |
704 | fh->nexthdr = nexthdr; |
705 | fh->reserved = 0; | |
706 | fh->frag_off = htons(IP6_MF); | |
707 | frag_id = fh->identification; | |
708 | ||
709 | first_len = skb_pagelen(skb); | |
710 | skb->data_len = first_len - skb_headlen(skb); | |
711 | skb->len = first_len; | |
0660e03f ACM |
712 | ipv6_hdr(skb)->payload_len = htons(first_len - |
713 | sizeof(struct ipv6hdr)); | |
a11d206d | 714 | |
d8d1f30b | 715 | dst_hold(&rt->dst); |
1da177e4 LT |
716 | |
717 | for (;;) { | |
718 | /* Prepare header of the next frame, | |
719 | * before previous one went down. */ | |
720 | if (frag) { | |
721 | frag->ip_summed = CHECKSUM_NONE; | |
badff6d0 | 722 | skb_reset_transport_header(frag); |
1da177e4 | 723 | fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr)); |
e2d1bca7 ACM |
724 | __skb_push(frag, hlen); |
725 | skb_reset_network_header(frag); | |
d56f90a7 ACM |
726 | memcpy(skb_network_header(frag), tmp_hdr, |
727 | hlen); | |
1da177e4 LT |
728 | offset += skb->len - hlen - sizeof(struct frag_hdr); |
729 | fh->nexthdr = nexthdr; | |
730 | fh->reserved = 0; | |
731 | fh->frag_off = htons(offset); | |
732 | if (frag->next != NULL) | |
733 | fh->frag_off |= htons(IP6_MF); | |
734 | fh->identification = frag_id; | |
0660e03f ACM |
735 | ipv6_hdr(frag)->payload_len = |
736 | htons(frag->len - | |
737 | sizeof(struct ipv6hdr)); | |
1da177e4 LT |
738 | ip6_copy_metadata(frag, skb); |
739 | } | |
1ab1457c | 740 | |
1da177e4 | 741 | err = output(skb); |
dafee490 | 742 | if(!err) |
d8d1f30b | 743 | IP6_INC_STATS(net, ip6_dst_idev(&rt->dst), |
3bd653c8 | 744 | IPSTATS_MIB_FRAGCREATES); |
dafee490 | 745 | |
1da177e4 LT |
746 | if (err || !frag) |
747 | break; | |
748 | ||
749 | skb = frag; | |
750 | frag = skb->next; | |
751 | skb->next = NULL; | |
752 | } | |
753 | ||
a51482bd | 754 | kfree(tmp_hdr); |
1da177e4 LT |
755 | |
756 | if (err == 0) { | |
d8d1f30b | 757 | IP6_INC_STATS(net, ip6_dst_idev(&rt->dst), |
3bd653c8 | 758 | IPSTATS_MIB_FRAGOKS); |
d8d1f30b | 759 | dst_release(&rt->dst); |
1da177e4 LT |
760 | return 0; |
761 | } | |
762 | ||
763 | while (frag) { | |
764 | skb = frag->next; | |
765 | kfree_skb(frag); | |
766 | frag = skb; | |
767 | } | |
768 | ||
d8d1f30b | 769 | IP6_INC_STATS(net, ip6_dst_idev(&rt->dst), |
3bd653c8 | 770 | IPSTATS_MIB_FRAGFAILS); |
d8d1f30b | 771 | dst_release(&rt->dst); |
1da177e4 | 772 | return err; |
3d13008e ED |
773 | |
774 | slow_path_clean: | |
775 | skb_walk_frags(skb, frag2) { | |
776 | if (frag2 == frag) | |
777 | break; | |
778 | frag2->sk = NULL; | |
779 | frag2->destructor = NULL; | |
780 | skb->truesize += frag2->truesize; | |
781 | } | |
1da177e4 LT |
782 | } |
783 | ||
784 | slow_path: | |
72e843bb ED |
785 | if ((skb->ip_summed == CHECKSUM_PARTIAL) && |
786 | skb_checksum_help(skb)) | |
787 | goto fail; | |
788 | ||
1da177e4 LT |
789 | left = skb->len - hlen; /* Space per frame */ |
790 | ptr = hlen; /* Where to start from */ | |
791 | ||
792 | /* | |
793 | * Fragment the datagram. | |
794 | */ | |
795 | ||
796 | *prevhdr = NEXTHDR_FRAGMENT; | |
a7ae1992 HX |
797 | hroom = LL_RESERVED_SPACE(rt->dst.dev); |
798 | troom = rt->dst.dev->needed_tailroom; | |
1da177e4 LT |
799 | |
800 | /* | |
801 | * Keep copying data until we run out. | |
802 | */ | |
803 | while(left > 0) { | |
804 | len = left; | |
805 | /* IF: it doesn't fit, use 'mtu' - the data space left */ | |
806 | if (len > mtu) | |
807 | len = mtu; | |
25985edc | 808 | /* IF: we are not sending up to and including the packet end |
1da177e4 LT |
809 | then align the next start on an eight byte boundary */ |
810 | if (len < left) { | |
811 | len &= ~7; | |
812 | } | |
813 | /* | |
814 | * Allocate buffer. | |
815 | */ | |
816 | ||
a7ae1992 HX |
817 | if ((frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) + |
818 | hroom + troom, GFP_ATOMIC)) == NULL) { | |
64ce2073 | 819 | NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n"); |
adf30907 | 820 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 821 | IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
822 | err = -ENOMEM; |
823 | goto fail; | |
824 | } | |
825 | ||
826 | /* | |
827 | * Set up data on packet | |
828 | */ | |
829 | ||
830 | ip6_copy_metadata(frag, skb); | |
a7ae1992 | 831 | skb_reserve(frag, hroom); |
1da177e4 | 832 | skb_put(frag, len + hlen + sizeof(struct frag_hdr)); |
c1d2bbe1 | 833 | skb_reset_network_header(frag); |
badff6d0 | 834 | fh = (struct frag_hdr *)(skb_network_header(frag) + hlen); |
b0e380b1 ACM |
835 | frag->transport_header = (frag->network_header + hlen + |
836 | sizeof(struct frag_hdr)); | |
1da177e4 LT |
837 | |
838 | /* | |
839 | * Charge the memory for the fragment to any owner | |
840 | * it might possess | |
841 | */ | |
842 | if (skb->sk) | |
843 | skb_set_owner_w(frag, skb->sk); | |
844 | ||
845 | /* | |
846 | * Copy the packet header into the new buffer. | |
847 | */ | |
d626f62b | 848 | skb_copy_from_linear_data(skb, skb_network_header(frag), hlen); |
1da177e4 LT |
849 | |
850 | /* | |
851 | * Build fragment header. | |
852 | */ | |
853 | fh->nexthdr = nexthdr; | |
854 | fh->reserved = 0; | |
f36d6ab1 | 855 | if (!frag_id) { |
87c48fa3 | 856 | ipv6_select_ident(fh, rt); |
1da177e4 LT |
857 | frag_id = fh->identification; |
858 | } else | |
859 | fh->identification = frag_id; | |
860 | ||
861 | /* | |
862 | * Copy a block of the IP datagram. | |
863 | */ | |
8984e41d | 864 | if (skb_copy_bits(skb, ptr, skb_transport_header(frag), len)) |
1da177e4 LT |
865 | BUG(); |
866 | left -= len; | |
867 | ||
868 | fh->frag_off = htons(offset); | |
869 | if (left > 0) | |
870 | fh->frag_off |= htons(IP6_MF); | |
0660e03f ACM |
871 | ipv6_hdr(frag)->payload_len = htons(frag->len - |
872 | sizeof(struct ipv6hdr)); | |
1da177e4 LT |
873 | |
874 | ptr += len; | |
875 | offset += len; | |
876 | ||
877 | /* | |
878 | * Put this fragment into the sending queue. | |
879 | */ | |
1da177e4 LT |
880 | err = output(frag); |
881 | if (err) | |
882 | goto fail; | |
dafee490 | 883 | |
adf30907 | 884 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
3bd653c8 | 885 | IPSTATS_MIB_FRAGCREATES); |
1da177e4 | 886 | } |
adf30907 | 887 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 888 | IPSTATS_MIB_FRAGOKS); |
808db80a | 889 | consume_skb(skb); |
1da177e4 LT |
890 | return err; |
891 | ||
892 | fail: | |
adf30907 | 893 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 894 | IPSTATS_MIB_FRAGFAILS); |
1ab1457c | 895 | kfree_skb(skb); |
1da177e4 LT |
896 | return err; |
897 | } | |
898 | ||
b71d1d42 ED |
899 | static inline int ip6_rt_check(const struct rt6key *rt_key, |
900 | const struct in6_addr *fl_addr, | |
901 | const struct in6_addr *addr_cache) | |
cf6b1982 | 902 | { |
a02cec21 ED |
903 | return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) && |
904 | (addr_cache == NULL || !ipv6_addr_equal(fl_addr, addr_cache)); | |
cf6b1982 YH |
905 | } |
906 | ||
497c615a HX |
907 | static struct dst_entry *ip6_sk_dst_check(struct sock *sk, |
908 | struct dst_entry *dst, | |
b71d1d42 | 909 | const struct flowi6 *fl6) |
1da177e4 | 910 | { |
497c615a HX |
911 | struct ipv6_pinfo *np = inet6_sk(sk); |
912 | struct rt6_info *rt = (struct rt6_info *)dst; | |
1da177e4 | 913 | |
497c615a HX |
914 | if (!dst) |
915 | goto out; | |
916 | ||
917 | /* Yes, checking route validity in not connected | |
918 | * case is not very simple. Take into account, | |
919 | * that we do not support routing by source, TOS, | |
920 | * and MSG_DONTROUTE --ANK (980726) | |
921 | * | |
cf6b1982 YH |
922 | * 1. ip6_rt_check(): If route was host route, |
923 | * check that cached destination is current. | |
497c615a HX |
924 | * If it is network route, we still may |
925 | * check its validity using saved pointer | |
926 | * to the last used address: daddr_cache. | |
927 | * We do not want to save whole address now, | |
928 | * (because main consumer of this service | |
929 | * is tcp, which has not this problem), | |
930 | * so that the last trick works only on connected | |
931 | * sockets. | |
932 | * 2. oif also should be the same. | |
933 | */ | |
4c9483b2 | 934 | if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) || |
8e1ef0a9 | 935 | #ifdef CONFIG_IPV6_SUBTREES |
4c9483b2 | 936 | ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) || |
8e1ef0a9 | 937 | #endif |
4c9483b2 | 938 | (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex)) { |
497c615a HX |
939 | dst_release(dst); |
940 | dst = NULL; | |
1da177e4 LT |
941 | } |
942 | ||
497c615a HX |
943 | out: |
944 | return dst; | |
945 | } | |
946 | ||
947 | static int ip6_dst_lookup_tail(struct sock *sk, | |
4c9483b2 | 948 | struct dst_entry **dst, struct flowi6 *fl6) |
497c615a | 949 | { |
3b1e0a65 | 950 | struct net *net = sock_net(sk); |
69cce1d1 DM |
951 | #ifdef CONFIG_IPV6_OPTIMISTIC_DAD |
952 | struct neighbour *n; | |
97cac082 | 953 | struct rt6_info *rt; |
69cce1d1 DM |
954 | #endif |
955 | int err; | |
497c615a | 956 | |
1da177e4 | 957 | if (*dst == NULL) |
4c9483b2 | 958 | *dst = ip6_route_output(net, sk, fl6); |
1da177e4 LT |
959 | |
960 | if ((err = (*dst)->error)) | |
961 | goto out_err_release; | |
962 | ||
4c9483b2 | 963 | if (ipv6_addr_any(&fl6->saddr)) { |
c3968a85 DW |
964 | struct rt6_info *rt = (struct rt6_info *) *dst; |
965 | err = ip6_route_get_saddr(net, rt, &fl6->daddr, | |
966 | sk ? inet6_sk(sk)->srcprefs : 0, | |
967 | &fl6->saddr); | |
44456d37 | 968 | if (err) |
1da177e4 | 969 | goto out_err_release; |
1da177e4 LT |
970 | } |
971 | ||
95c385b4 | 972 | #ifdef CONFIG_IPV6_OPTIMISTIC_DAD |
e550dfb0 NH |
973 | /* |
974 | * Here if the dst entry we've looked up | |
975 | * has a neighbour entry that is in the INCOMPLETE | |
976 | * state and the src address from the flow is | |
977 | * marked as OPTIMISTIC, we release the found | |
978 | * dst entry and replace it instead with the | |
979 | * dst entry of the nexthop router | |
980 | */ | |
c56bf6fe | 981 | rt = (struct rt6_info *) *dst; |
97cac082 | 982 | n = rt->n; |
69cce1d1 | 983 | if (n && !(n->nud_state & NUD_VALID)) { |
e550dfb0 | 984 | struct inet6_ifaddr *ifp; |
4c9483b2 | 985 | struct flowi6 fl_gw6; |
e550dfb0 NH |
986 | int redirect; |
987 | ||
4c9483b2 | 988 | ifp = ipv6_get_ifaddr(net, &fl6->saddr, |
e550dfb0 NH |
989 | (*dst)->dev, 1); |
990 | ||
991 | redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC); | |
992 | if (ifp) | |
993 | in6_ifa_put(ifp); | |
994 | ||
995 | if (redirect) { | |
996 | /* | |
997 | * We need to get the dst entry for the | |
998 | * default router instead | |
999 | */ | |
1000 | dst_release(*dst); | |
4c9483b2 DM |
1001 | memcpy(&fl_gw6, fl6, sizeof(struct flowi6)); |
1002 | memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr)); | |
1003 | *dst = ip6_route_output(net, sk, &fl_gw6); | |
e550dfb0 NH |
1004 | if ((err = (*dst)->error)) |
1005 | goto out_err_release; | |
95c385b4 | 1006 | } |
e550dfb0 | 1007 | } |
95c385b4 NH |
1008 | #endif |
1009 | ||
1da177e4 LT |
1010 | return 0; |
1011 | ||
1012 | out_err_release: | |
ca46f9c8 | 1013 | if (err == -ENETUNREACH) |
483a47d2 | 1014 | IP6_INC_STATS_BH(net, NULL, IPSTATS_MIB_OUTNOROUTES); |
1da177e4 LT |
1015 | dst_release(*dst); |
1016 | *dst = NULL; | |
1017 | return err; | |
1018 | } | |
34a0b3cd | 1019 | |
497c615a HX |
1020 | /** |
1021 | * ip6_dst_lookup - perform route lookup on flow | |
1022 | * @sk: socket which provides route info | |
1023 | * @dst: pointer to dst_entry * for result | |
4c9483b2 | 1024 | * @fl6: flow to lookup |
497c615a HX |
1025 | * |
1026 | * This function performs a route lookup on the given flow. | |
1027 | * | |
1028 | * It returns zero on success, or a standard errno code on error. | |
1029 | */ | |
4c9483b2 | 1030 | int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi6 *fl6) |
497c615a HX |
1031 | { |
1032 | *dst = NULL; | |
4c9483b2 | 1033 | return ip6_dst_lookup_tail(sk, dst, fl6); |
497c615a | 1034 | } |
3cf3dc6c ACM |
1035 | EXPORT_SYMBOL_GPL(ip6_dst_lookup); |
1036 | ||
497c615a | 1037 | /** |
68d0c6d3 DM |
1038 | * ip6_dst_lookup_flow - perform route lookup on flow with ipsec |
1039 | * @sk: socket which provides route info | |
4c9483b2 | 1040 | * @fl6: flow to lookup |
68d0c6d3 | 1041 | * @final_dst: final destination address for ipsec lookup |
a1414715 | 1042 | * @can_sleep: we are in a sleepable context |
68d0c6d3 DM |
1043 | * |
1044 | * This function performs a route lookup on the given flow. | |
1045 | * | |
1046 | * It returns a valid dst pointer on success, or a pointer encoded | |
1047 | * error code. | |
1048 | */ | |
4c9483b2 | 1049 | struct dst_entry *ip6_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6, |
68d0c6d3 | 1050 | const struct in6_addr *final_dst, |
a1414715 | 1051 | bool can_sleep) |
68d0c6d3 DM |
1052 | { |
1053 | struct dst_entry *dst = NULL; | |
1054 | int err; | |
1055 | ||
4c9483b2 | 1056 | err = ip6_dst_lookup_tail(sk, &dst, fl6); |
68d0c6d3 DM |
1057 | if (err) |
1058 | return ERR_PTR(err); | |
1059 | if (final_dst) | |
4e3fd7a0 | 1060 | fl6->daddr = *final_dst; |
2774c131 | 1061 | if (can_sleep) |
4c9483b2 | 1062 | fl6->flowi6_flags |= FLOWI_FLAG_CAN_SLEEP; |
2774c131 | 1063 | |
4c9483b2 | 1064 | return xfrm_lookup(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0); |
68d0c6d3 DM |
1065 | } |
1066 | EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow); | |
1067 | ||
1068 | /** | |
1069 | * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow | |
497c615a | 1070 | * @sk: socket which provides the dst cache and route info |
4c9483b2 | 1071 | * @fl6: flow to lookup |
68d0c6d3 | 1072 | * @final_dst: final destination address for ipsec lookup |
a1414715 | 1073 | * @can_sleep: we are in a sleepable context |
497c615a HX |
1074 | * |
1075 | * This function performs a route lookup on the given flow with the | |
1076 | * possibility of using the cached route in the socket if it is valid. | |
1077 | * It will take the socket dst lock when operating on the dst cache. | |
1078 | * As a result, this function can only be used in process context. | |
1079 | * | |
68d0c6d3 DM |
1080 | * It returns a valid dst pointer on success, or a pointer encoded |
1081 | * error code. | |
497c615a | 1082 | */ |
4c9483b2 | 1083 | struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6, |
68d0c6d3 | 1084 | const struct in6_addr *final_dst, |
a1414715 | 1085 | bool can_sleep) |
497c615a | 1086 | { |
68d0c6d3 DM |
1087 | struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie); |
1088 | int err; | |
497c615a | 1089 | |
4c9483b2 | 1090 | dst = ip6_sk_dst_check(sk, dst, fl6); |
68d0c6d3 | 1091 | |
4c9483b2 | 1092 | err = ip6_dst_lookup_tail(sk, &dst, fl6); |
68d0c6d3 DM |
1093 | if (err) |
1094 | return ERR_PTR(err); | |
1095 | if (final_dst) | |
4e3fd7a0 | 1096 | fl6->daddr = *final_dst; |
2774c131 | 1097 | if (can_sleep) |
4c9483b2 | 1098 | fl6->flowi6_flags |= FLOWI_FLAG_CAN_SLEEP; |
2774c131 | 1099 | |
4c9483b2 | 1100 | return xfrm_lookup(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0); |
497c615a | 1101 | } |
68d0c6d3 | 1102 | EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow); |
497c615a | 1103 | |
34a0b3cd | 1104 | static inline int ip6_ufo_append_data(struct sock *sk, |
e89e9cf5 AR |
1105 | int getfrag(void *from, char *to, int offset, int len, |
1106 | int odd, struct sk_buff *skb), | |
1107 | void *from, int length, int hh_len, int fragheaderlen, | |
87c48fa3 ED |
1108 | int transhdrlen, int mtu,unsigned int flags, |
1109 | struct rt6_info *rt) | |
e89e9cf5 AR |
1110 | |
1111 | { | |
1112 | struct sk_buff *skb; | |
1113 | int err; | |
1114 | ||
1115 | /* There is support for UDP large send offload by network | |
1116 | * device, so create one single skb packet containing complete | |
1117 | * udp datagram | |
1118 | */ | |
1119 | if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) { | |
1120 | skb = sock_alloc_send_skb(sk, | |
1121 | hh_len + fragheaderlen + transhdrlen + 20, | |
1122 | (flags & MSG_DONTWAIT), &err); | |
1123 | if (skb == NULL) | |
504744e4 | 1124 | return err; |
e89e9cf5 AR |
1125 | |
1126 | /* reserve space for Hardware header */ | |
1127 | skb_reserve(skb, hh_len); | |
1128 | ||
1129 | /* create space for UDP/IP header */ | |
1130 | skb_put(skb,fragheaderlen + transhdrlen); | |
1131 | ||
1132 | /* initialize network header pointer */ | |
c1d2bbe1 | 1133 | skb_reset_network_header(skb); |
e89e9cf5 AR |
1134 | |
1135 | /* initialize protocol header pointer */ | |
b0e380b1 | 1136 | skb->transport_header = skb->network_header + fragheaderlen; |
e89e9cf5 | 1137 | |
84fa7933 | 1138 | skb->ip_summed = CHECKSUM_PARTIAL; |
e89e9cf5 | 1139 | skb->csum = 0; |
e89e9cf5 AR |
1140 | } |
1141 | ||
1142 | err = skb_append_datato_frags(sk,skb, getfrag, from, | |
1143 | (length - transhdrlen)); | |
1144 | if (!err) { | |
1145 | struct frag_hdr fhdr; | |
1146 | ||
c31d5326 SS |
1147 | /* Specify the length of each IPv6 datagram fragment. |
1148 | * It has to be a multiple of 8. | |
1149 | */ | |
1150 | skb_shinfo(skb)->gso_size = (mtu - fragheaderlen - | |
1151 | sizeof(struct frag_hdr)) & ~7; | |
f83ef8c0 | 1152 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP; |
87c48fa3 | 1153 | ipv6_select_ident(&fhdr, rt); |
e89e9cf5 AR |
1154 | skb_shinfo(skb)->ip6_frag_id = fhdr.identification; |
1155 | __skb_queue_tail(&sk->sk_write_queue, skb); | |
1156 | ||
1157 | return 0; | |
1158 | } | |
1159 | /* There is not enough support do UPD LSO, | |
1160 | * so follow normal path | |
1161 | */ | |
1162 | kfree_skb(skb); | |
1163 | ||
1164 | return err; | |
1165 | } | |
1da177e4 | 1166 | |
0178b695 HX |
1167 | static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src, |
1168 | gfp_t gfp) | |
1169 | { | |
1170 | return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL; | |
1171 | } | |
1172 | ||
1173 | static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src, | |
1174 | gfp_t gfp) | |
1175 | { | |
1176 | return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL; | |
1177 | } | |
1178 | ||
0c183379 G |
1179 | static void ip6_append_data_mtu(int *mtu, |
1180 | int *maxfraglen, | |
1181 | unsigned int fragheaderlen, | |
1182 | struct sk_buff *skb, | |
1183 | struct rt6_info *rt) | |
1184 | { | |
1185 | if (!(rt->dst.flags & DST_XFRM_TUNNEL)) { | |
1186 | if (skb == NULL) { | |
1187 | /* first fragment, reserve header_len */ | |
1188 | *mtu = *mtu - rt->dst.header_len; | |
1189 | ||
1190 | } else { | |
1191 | /* | |
1192 | * this fragment is not first, the headers | |
1193 | * space is regarded as data space. | |
1194 | */ | |
1195 | *mtu = dst_mtu(rt->dst.path); | |
1196 | } | |
1197 | *maxfraglen = ((*mtu - fragheaderlen) & ~7) | |
1198 | + fragheaderlen - sizeof(struct frag_hdr); | |
1199 | } | |
1200 | } | |
1201 | ||
41a1f8ea YH |
1202 | int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to, |
1203 | int offset, int len, int odd, struct sk_buff *skb), | |
1204 | void *from, int length, int transhdrlen, | |
4c9483b2 | 1205 | int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi6 *fl6, |
13b52cd4 | 1206 | struct rt6_info *rt, unsigned int flags, int dontfrag) |
1da177e4 LT |
1207 | { |
1208 | struct inet_sock *inet = inet_sk(sk); | |
1209 | struct ipv6_pinfo *np = inet6_sk(sk); | |
bdc712b4 | 1210 | struct inet_cork *cork; |
0c183379 | 1211 | struct sk_buff *skb, *skb_prev = NULL; |
1da177e4 LT |
1212 | unsigned int maxfraglen, fragheaderlen; |
1213 | int exthdrlen; | |
299b0767 | 1214 | int dst_exthdrlen; |
1da177e4 LT |
1215 | int hh_len; |
1216 | int mtu; | |
1217 | int copy; | |
1218 | int err; | |
1219 | int offset = 0; | |
a693e698 | 1220 | __u8 tx_flags = 0; |
1da177e4 LT |
1221 | |
1222 | if (flags&MSG_PROBE) | |
1223 | return 0; | |
bdc712b4 | 1224 | cork = &inet->cork.base; |
1da177e4 LT |
1225 | if (skb_queue_empty(&sk->sk_write_queue)) { |
1226 | /* | |
1227 | * setup for corking | |
1228 | */ | |
1229 | if (opt) { | |
0178b695 | 1230 | if (WARN_ON(np->cork.opt)) |
1da177e4 | 1231 | return -EINVAL; |
0178b695 HX |
1232 | |
1233 | np->cork.opt = kmalloc(opt->tot_len, sk->sk_allocation); | |
1234 | if (unlikely(np->cork.opt == NULL)) | |
1235 | return -ENOBUFS; | |
1236 | ||
1237 | np->cork.opt->tot_len = opt->tot_len; | |
1238 | np->cork.opt->opt_flen = opt->opt_flen; | |
1239 | np->cork.opt->opt_nflen = opt->opt_nflen; | |
1240 | ||
1241 | np->cork.opt->dst0opt = ip6_opt_dup(opt->dst0opt, | |
1242 | sk->sk_allocation); | |
1243 | if (opt->dst0opt && !np->cork.opt->dst0opt) | |
1244 | return -ENOBUFS; | |
1245 | ||
1246 | np->cork.opt->dst1opt = ip6_opt_dup(opt->dst1opt, | |
1247 | sk->sk_allocation); | |
1248 | if (opt->dst1opt && !np->cork.opt->dst1opt) | |
1249 | return -ENOBUFS; | |
1250 | ||
1251 | np->cork.opt->hopopt = ip6_opt_dup(opt->hopopt, | |
1252 | sk->sk_allocation); | |
1253 | if (opt->hopopt && !np->cork.opt->hopopt) | |
1254 | return -ENOBUFS; | |
1255 | ||
1256 | np->cork.opt->srcrt = ip6_rthdr_dup(opt->srcrt, | |
1257 | sk->sk_allocation); | |
1258 | if (opt->srcrt && !np->cork.opt->srcrt) | |
1259 | return -ENOBUFS; | |
1260 | ||
1da177e4 LT |
1261 | /* need source address above miyazawa*/ |
1262 | } | |
d8d1f30b | 1263 | dst_hold(&rt->dst); |
bdc712b4 | 1264 | cork->dst = &rt->dst; |
4c9483b2 | 1265 | inet->cork.fl.u.ip6 = *fl6; |
1da177e4 | 1266 | np->cork.hop_limit = hlimit; |
41a1f8ea | 1267 | np->cork.tclass = tclass; |
0c183379 G |
1268 | if (rt->dst.flags & DST_XFRM_TUNNEL) |
1269 | mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ? | |
1270 | rt->dst.dev->mtu : dst_mtu(&rt->dst); | |
1271 | else | |
1272 | mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ? | |
1273 | rt->dst.dev->mtu : dst_mtu(rt->dst.path); | |
c7503609 | 1274 | if (np->frag_size < mtu) { |
d91675f9 YH |
1275 | if (np->frag_size) |
1276 | mtu = np->frag_size; | |
1277 | } | |
bdc712b4 | 1278 | cork->fragsize = mtu; |
d8d1f30b | 1279 | if (dst_allfrag(rt->dst.path)) |
bdc712b4 DM |
1280 | cork->flags |= IPCORK_ALLFRAG; |
1281 | cork->length = 0; | |
299b0767 | 1282 | exthdrlen = (opt ? opt->opt_flen : 0) - rt->rt6i_nfheader_len; |
1da177e4 LT |
1283 | length += exthdrlen; |
1284 | transhdrlen += exthdrlen; | |
299b0767 | 1285 | dst_exthdrlen = rt->dst.header_len; |
1da177e4 | 1286 | } else { |
bdc712b4 | 1287 | rt = (struct rt6_info *)cork->dst; |
4c9483b2 | 1288 | fl6 = &inet->cork.fl.u.ip6; |
0178b695 | 1289 | opt = np->cork.opt; |
1da177e4 LT |
1290 | transhdrlen = 0; |
1291 | exthdrlen = 0; | |
299b0767 | 1292 | dst_exthdrlen = 0; |
bdc712b4 | 1293 | mtu = cork->fragsize; |
1da177e4 LT |
1294 | } |
1295 | ||
d8d1f30b | 1296 | hh_len = LL_RESERVED_SPACE(rt->dst.dev); |
1da177e4 | 1297 | |
a1b05140 | 1298 | fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len + |
b4ce9277 | 1299 | (opt ? opt->opt_nflen : 0); |
1da177e4 LT |
1300 | maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr); |
1301 | ||
1302 | if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) { | |
bdc712b4 | 1303 | if (cork->length + length > sizeof(struct ipv6hdr) + IPV6_MAXPLEN - fragheaderlen) { |
4c9483b2 | 1304 | ipv6_local_error(sk, EMSGSIZE, fl6, mtu-exthdrlen); |
1da177e4 LT |
1305 | return -EMSGSIZE; |
1306 | } | |
1307 | } | |
1308 | ||
a693e698 AB |
1309 | /* For UDP, check if TX timestamp is enabled */ |
1310 | if (sk->sk_type == SOCK_DGRAM) { | |
1311 | err = sock_tx_timestamp(sk, &tx_flags); | |
1312 | if (err) | |
1313 | goto error; | |
1314 | } | |
1315 | ||
1da177e4 LT |
1316 | /* |
1317 | * Let's try using as much space as possible. | |
1318 | * Use MTU if total length of the message fits into the MTU. | |
1319 | * Otherwise, we need to reserve fragment header and | |
1320 | * fragment alignment (= 8-15 octects, in total). | |
1321 | * | |
1322 | * Note that we may need to "move" the data from the tail of | |
1ab1457c | 1323 | * of the buffer to the new fragment when we split |
1da177e4 LT |
1324 | * the message. |
1325 | * | |
1ab1457c | 1326 | * FIXME: It may be fragmented into multiple chunks |
1da177e4 LT |
1327 | * at once if non-fragmentable extension headers |
1328 | * are too large. | |
1ab1457c | 1329 | * --yoshfuji |
1da177e4 LT |
1330 | */ |
1331 | ||
bdc712b4 | 1332 | cork->length += length; |
4b340ae2 BH |
1333 | if (length > mtu) { |
1334 | int proto = sk->sk_protocol; | |
1335 | if (dontfrag && (proto == IPPROTO_UDP || proto == IPPROTO_RAW)){ | |
4c9483b2 | 1336 | ipv6_local_rxpmtu(sk, fl6, mtu-exthdrlen); |
4b340ae2 BH |
1337 | return -EMSGSIZE; |
1338 | } | |
e89e9cf5 | 1339 | |
4b340ae2 | 1340 | if (proto == IPPROTO_UDP && |
d8d1f30b | 1341 | (rt->dst.dev->features & NETIF_F_UFO)) { |
4b340ae2 BH |
1342 | |
1343 | err = ip6_ufo_append_data(sk, getfrag, from, length, | |
1344 | hh_len, fragheaderlen, | |
87c48fa3 | 1345 | transhdrlen, mtu, flags, rt); |
4b340ae2 BH |
1346 | if (err) |
1347 | goto error; | |
1348 | return 0; | |
1349 | } | |
e89e9cf5 | 1350 | } |
1da177e4 LT |
1351 | |
1352 | if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) | |
1353 | goto alloc_new_skb; | |
1354 | ||
1355 | while (length > 0) { | |
1356 | /* Check if the remaining data fits into current packet. */ | |
bdc712b4 | 1357 | copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len; |
1da177e4 LT |
1358 | if (copy < length) |
1359 | copy = maxfraglen - skb->len; | |
1360 | ||
1361 | if (copy <= 0) { | |
1362 | char *data; | |
1363 | unsigned int datalen; | |
1364 | unsigned int fraglen; | |
1365 | unsigned int fraggap; | |
1366 | unsigned int alloclen; | |
1da177e4 | 1367 | alloc_new_skb: |
1da177e4 | 1368 | /* There's no room in the current skb */ |
0c183379 G |
1369 | if (skb) |
1370 | fraggap = skb->len - maxfraglen; | |
1da177e4 LT |
1371 | else |
1372 | fraggap = 0; | |
0c183379 G |
1373 | /* update mtu and maxfraglen if necessary */ |
1374 | if (skb == NULL || skb_prev == NULL) | |
1375 | ip6_append_data_mtu(&mtu, &maxfraglen, | |
1376 | fragheaderlen, skb, rt); | |
1377 | ||
1378 | skb_prev = skb; | |
1da177e4 LT |
1379 | |
1380 | /* | |
1381 | * If remaining data exceeds the mtu, | |
1382 | * we know we need more fragment(s). | |
1383 | */ | |
1384 | datalen = length + fraggap; | |
1da177e4 | 1385 | |
0c183379 G |
1386 | if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen) |
1387 | datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len; | |
1da177e4 | 1388 | if ((flags & MSG_MORE) && |
d8d1f30b | 1389 | !(rt->dst.dev->features&NETIF_F_SG)) |
1da177e4 LT |
1390 | alloclen = mtu; |
1391 | else | |
1392 | alloclen = datalen + fragheaderlen; | |
1393 | ||
299b0767 SK |
1394 | alloclen += dst_exthdrlen; |
1395 | ||
0c183379 G |
1396 | if (datalen != length + fraggap) { |
1397 | /* | |
1398 | * this is not the last fragment, the trailer | |
1399 | * space is regarded as data space. | |
1400 | */ | |
1401 | datalen += rt->dst.trailer_len; | |
1402 | } | |
1403 | ||
1404 | alloclen += rt->dst.trailer_len; | |
1405 | fraglen = datalen + fragheaderlen; | |
1da177e4 LT |
1406 | |
1407 | /* | |
1408 | * We just reserve space for fragment header. | |
1ab1457c | 1409 | * Note: this may be overallocation if the message |
1da177e4 LT |
1410 | * (without MSG_MORE) fits into the MTU. |
1411 | */ | |
1412 | alloclen += sizeof(struct frag_hdr); | |
1413 | ||
1414 | if (transhdrlen) { | |
1415 | skb = sock_alloc_send_skb(sk, | |
1416 | alloclen + hh_len, | |
1417 | (flags & MSG_DONTWAIT), &err); | |
1418 | } else { | |
1419 | skb = NULL; | |
1420 | if (atomic_read(&sk->sk_wmem_alloc) <= | |
1421 | 2 * sk->sk_sndbuf) | |
1422 | skb = sock_wmalloc(sk, | |
1423 | alloclen + hh_len, 1, | |
1424 | sk->sk_allocation); | |
1425 | if (unlikely(skb == NULL)) | |
1426 | err = -ENOBUFS; | |
a693e698 AB |
1427 | else { |
1428 | /* Only the initial fragment | |
1429 | * is time stamped. | |
1430 | */ | |
1431 | tx_flags = 0; | |
1432 | } | |
1da177e4 LT |
1433 | } |
1434 | if (skb == NULL) | |
1435 | goto error; | |
1436 | /* | |
1437 | * Fill in the control structures | |
1438 | */ | |
d7f7c0ac | 1439 | skb->ip_summed = CHECKSUM_NONE; |
1da177e4 | 1440 | skb->csum = 0; |
1f85851e G |
1441 | /* reserve for fragmentation and ipsec header */ |
1442 | skb_reserve(skb, hh_len + sizeof(struct frag_hdr) + | |
1443 | dst_exthdrlen); | |
1da177e4 | 1444 | |
a693e698 AB |
1445 | if (sk->sk_type == SOCK_DGRAM) |
1446 | skb_shinfo(skb)->tx_flags = tx_flags; | |
1447 | ||
1da177e4 LT |
1448 | /* |
1449 | * Find where to start putting bytes | |
1450 | */ | |
1f85851e G |
1451 | data = skb_put(skb, fraglen); |
1452 | skb_set_network_header(skb, exthdrlen); | |
1453 | data += fragheaderlen; | |
b0e380b1 ACM |
1454 | skb->transport_header = (skb->network_header + |
1455 | fragheaderlen); | |
1da177e4 LT |
1456 | if (fraggap) { |
1457 | skb->csum = skb_copy_and_csum_bits( | |
1458 | skb_prev, maxfraglen, | |
1459 | data + transhdrlen, fraggap, 0); | |
1460 | skb_prev->csum = csum_sub(skb_prev->csum, | |
1461 | skb->csum); | |
1462 | data += fraggap; | |
e9fa4f7b | 1463 | pskb_trim_unique(skb_prev, maxfraglen); |
1da177e4 LT |
1464 | } |
1465 | copy = datalen - transhdrlen - fraggap; | |
299b0767 | 1466 | |
1da177e4 LT |
1467 | if (copy < 0) { |
1468 | err = -EINVAL; | |
1469 | kfree_skb(skb); | |
1470 | goto error; | |
1471 | } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) { | |
1472 | err = -EFAULT; | |
1473 | kfree_skb(skb); | |
1474 | goto error; | |
1475 | } | |
1476 | ||
1477 | offset += copy; | |
1478 | length -= datalen - fraggap; | |
1479 | transhdrlen = 0; | |
1480 | exthdrlen = 0; | |
299b0767 | 1481 | dst_exthdrlen = 0; |
1da177e4 LT |
1482 | |
1483 | /* | |
1484 | * Put the packet on the pending queue | |
1485 | */ | |
1486 | __skb_queue_tail(&sk->sk_write_queue, skb); | |
1487 | continue; | |
1488 | } | |
1489 | ||
1490 | if (copy > length) | |
1491 | copy = length; | |
1492 | ||
d8d1f30b | 1493 | if (!(rt->dst.dev->features&NETIF_F_SG)) { |
1da177e4 LT |
1494 | unsigned int off; |
1495 | ||
1496 | off = skb->len; | |
1497 | if (getfrag(from, skb_put(skb, copy), | |
1498 | offset, copy, off, skb) < 0) { | |
1499 | __skb_trim(skb, off); | |
1500 | err = -EFAULT; | |
1501 | goto error; | |
1502 | } | |
1503 | } else { | |
1504 | int i = skb_shinfo(skb)->nr_frags; | |
5640f768 | 1505 | struct page_frag *pfrag = sk_page_frag(sk); |
1da177e4 | 1506 | |
5640f768 ED |
1507 | err = -ENOMEM; |
1508 | if (!sk_page_frag_refill(sk, pfrag)) | |
1da177e4 | 1509 | goto error; |
5640f768 ED |
1510 | |
1511 | if (!skb_can_coalesce(skb, i, pfrag->page, | |
1512 | pfrag->offset)) { | |
1513 | err = -EMSGSIZE; | |
1514 | if (i == MAX_SKB_FRAGS) | |
1515 | goto error; | |
1516 | ||
1517 | __skb_fill_page_desc(skb, i, pfrag->page, | |
1518 | pfrag->offset, 0); | |
1519 | skb_shinfo(skb)->nr_frags = ++i; | |
1520 | get_page(pfrag->page); | |
1da177e4 | 1521 | } |
5640f768 | 1522 | copy = min_t(int, copy, pfrag->size - pfrag->offset); |
9e903e08 | 1523 | if (getfrag(from, |
5640f768 ED |
1524 | page_address(pfrag->page) + pfrag->offset, |
1525 | offset, copy, skb->len, skb) < 0) | |
1526 | goto error_efault; | |
1527 | ||
1528 | pfrag->offset += copy; | |
1529 | skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy); | |
1da177e4 LT |
1530 | skb->len += copy; |
1531 | skb->data_len += copy; | |
f945fa7a HX |
1532 | skb->truesize += copy; |
1533 | atomic_add(copy, &sk->sk_wmem_alloc); | |
1da177e4 LT |
1534 | } |
1535 | offset += copy; | |
1536 | length -= copy; | |
1537 | } | |
5640f768 | 1538 | |
1da177e4 | 1539 | return 0; |
5640f768 ED |
1540 | |
1541 | error_efault: | |
1542 | err = -EFAULT; | |
1da177e4 | 1543 | error: |
bdc712b4 | 1544 | cork->length -= length; |
3bd653c8 | 1545 | IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
1546 | return err; |
1547 | } | |
a495f836 | 1548 | EXPORT_SYMBOL_GPL(ip6_append_data); |
1da177e4 | 1549 | |
bf138862 PE |
1550 | static void ip6_cork_release(struct inet_sock *inet, struct ipv6_pinfo *np) |
1551 | { | |
0178b695 HX |
1552 | if (np->cork.opt) { |
1553 | kfree(np->cork.opt->dst0opt); | |
1554 | kfree(np->cork.opt->dst1opt); | |
1555 | kfree(np->cork.opt->hopopt); | |
1556 | kfree(np->cork.opt->srcrt); | |
1557 | kfree(np->cork.opt); | |
1558 | np->cork.opt = NULL; | |
1559 | } | |
1560 | ||
bdc712b4 DM |
1561 | if (inet->cork.base.dst) { |
1562 | dst_release(inet->cork.base.dst); | |
1563 | inet->cork.base.dst = NULL; | |
1564 | inet->cork.base.flags &= ~IPCORK_ALLFRAG; | |
bf138862 PE |
1565 | } |
1566 | memset(&inet->cork.fl, 0, sizeof(inet->cork.fl)); | |
1567 | } | |
1568 | ||
1da177e4 LT |
1569 | int ip6_push_pending_frames(struct sock *sk) |
1570 | { | |
1571 | struct sk_buff *skb, *tmp_skb; | |
1572 | struct sk_buff **tail_skb; | |
1573 | struct in6_addr final_dst_buf, *final_dst = &final_dst_buf; | |
1574 | struct inet_sock *inet = inet_sk(sk); | |
1575 | struct ipv6_pinfo *np = inet6_sk(sk); | |
3bd653c8 | 1576 | struct net *net = sock_net(sk); |
1da177e4 LT |
1577 | struct ipv6hdr *hdr; |
1578 | struct ipv6_txoptions *opt = np->cork.opt; | |
bdc712b4 | 1579 | struct rt6_info *rt = (struct rt6_info *)inet->cork.base.dst; |
4c9483b2 DM |
1580 | struct flowi6 *fl6 = &inet->cork.fl.u.ip6; |
1581 | unsigned char proto = fl6->flowi6_proto; | |
1da177e4 LT |
1582 | int err = 0; |
1583 | ||
1584 | if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL) | |
1585 | goto out; | |
1586 | tail_skb = &(skb_shinfo(skb)->frag_list); | |
1587 | ||
1588 | /* move skb->data to ip header from ext header */ | |
d56f90a7 | 1589 | if (skb->data < skb_network_header(skb)) |
bbe735e4 | 1590 | __skb_pull(skb, skb_network_offset(skb)); |
1da177e4 | 1591 | while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) { |
cfe1fc77 | 1592 | __skb_pull(tmp_skb, skb_network_header_len(skb)); |
1da177e4 LT |
1593 | *tail_skb = tmp_skb; |
1594 | tail_skb = &(tmp_skb->next); | |
1595 | skb->len += tmp_skb->len; | |
1596 | skb->data_len += tmp_skb->len; | |
1da177e4 | 1597 | skb->truesize += tmp_skb->truesize; |
1da177e4 LT |
1598 | tmp_skb->destructor = NULL; |
1599 | tmp_skb->sk = NULL; | |
1da177e4 LT |
1600 | } |
1601 | ||
28a89453 | 1602 | /* Allow local fragmentation. */ |
b5c15fc0 | 1603 | if (np->pmtudisc < IPV6_PMTUDISC_DO) |
28a89453 HX |
1604 | skb->local_df = 1; |
1605 | ||
4e3fd7a0 | 1606 | *final_dst = fl6->daddr; |
cfe1fc77 | 1607 | __skb_pull(skb, skb_network_header_len(skb)); |
1da177e4 LT |
1608 | if (opt && opt->opt_flen) |
1609 | ipv6_push_frag_opts(skb, opt, &proto); | |
1610 | if (opt && opt->opt_nflen) | |
1611 | ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst); | |
1612 | ||
e2d1bca7 ACM |
1613 | skb_push(skb, sizeof(struct ipv6hdr)); |
1614 | skb_reset_network_header(skb); | |
0660e03f | 1615 | hdr = ipv6_hdr(skb); |
1ab1457c | 1616 | |
4c9483b2 | 1617 | *(__be32*)hdr = fl6->flowlabel | |
41a1f8ea | 1618 | htonl(0x60000000 | ((int)np->cork.tclass << 20)); |
1da177e4 | 1619 | |
1da177e4 LT |
1620 | hdr->hop_limit = np->cork.hop_limit; |
1621 | hdr->nexthdr = proto; | |
4e3fd7a0 AD |
1622 | hdr->saddr = fl6->saddr; |
1623 | hdr->daddr = *final_dst; | |
1da177e4 | 1624 | |
a2c2064f | 1625 | skb->priority = sk->sk_priority; |
4a19ec58 | 1626 | skb->mark = sk->sk_mark; |
a2c2064f | 1627 | |
d8d1f30b | 1628 | skb_dst_set(skb, dst_clone(&rt->dst)); |
edf391ff | 1629 | IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); |
14878f75 | 1630 | if (proto == IPPROTO_ICMPV6) { |
adf30907 | 1631 | struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb)); |
14878f75 | 1632 | |
5a57d4c7 | 1633 | ICMP6MSGOUT_INC_STATS_BH(net, idev, icmp6_hdr(skb)->icmp6_type); |
e41b5368 | 1634 | ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTMSGS); |
14878f75 DS |
1635 | } |
1636 | ||
ef76bc23 | 1637 | err = ip6_local_out(skb); |
1da177e4 LT |
1638 | if (err) { |
1639 | if (err > 0) | |
6ce9e7b5 | 1640 | err = net_xmit_errno(err); |
1da177e4 LT |
1641 | if (err) |
1642 | goto error; | |
1643 | } | |
1644 | ||
1645 | out: | |
bf138862 | 1646 | ip6_cork_release(inet, np); |
1da177e4 LT |
1647 | return err; |
1648 | error: | |
06254914 | 1649 | IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
1650 | goto out; |
1651 | } | |
a495f836 | 1652 | EXPORT_SYMBOL_GPL(ip6_push_pending_frames); |
1da177e4 LT |
1653 | |
1654 | void ip6_flush_pending_frames(struct sock *sk) | |
1655 | { | |
1da177e4 LT |
1656 | struct sk_buff *skb; |
1657 | ||
1658 | while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) { | |
adf30907 ED |
1659 | if (skb_dst(skb)) |
1660 | IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)), | |
e1f52208 | 1661 | IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
1662 | kfree_skb(skb); |
1663 | } | |
1664 | ||
bf138862 | 1665 | ip6_cork_release(inet_sk(sk), inet6_sk(sk)); |
1da177e4 | 1666 | } |
a495f836 | 1667 | EXPORT_SYMBOL_GPL(ip6_flush_pending_frames); |