]>
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> | |
67ba4152 | 23 | * Imran Patel : frag id should be in NBO |
1da177e4 LT |
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 | 41 | |
33b48679 | 42 | #include <linux/bpf-cgroup.h> |
1da177e4 LT |
43 | #include <linux/netfilter.h> |
44 | #include <linux/netfilter_ipv6.h> | |
45 | ||
46 | #include <net/sock.h> | |
47 | #include <net/snmp.h> | |
48 | ||
49 | #include <net/ipv6.h> | |
50 | #include <net/ndisc.h> | |
51 | #include <net/protocol.h> | |
52 | #include <net/ip6_route.h> | |
53 | #include <net/addrconf.h> | |
54 | #include <net/rawv6.h> | |
55 | #include <net/icmp.h> | |
56 | #include <net/xfrm.h> | |
57 | #include <net/checksum.h> | |
7bc570c8 | 58 | #include <linux/mroute6.h> |
ca254490 | 59 | #include <net/l3mdev.h> |
14972cbd | 60 | #include <net/lwtunnel.h> |
1da177e4 | 61 | |
7d8c6e39 | 62 | static int ip6_finish_output2(struct net *net, struct sock *sk, struct sk_buff *skb) |
1da177e4 | 63 | { |
adf30907 | 64 | struct dst_entry *dst = skb_dst(skb); |
1da177e4 | 65 | struct net_device *dev = dst->dev; |
f6b72b62 | 66 | struct neighbour *neigh; |
6fd6ce20 YH |
67 | struct in6_addr *nexthop; |
68 | int ret; | |
1da177e4 | 69 | |
0660e03f | 70 | if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) { |
adf30907 | 71 | struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb)); |
1da177e4 | 72 | |
7026b1dd | 73 | if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(sk) && |
78126c41 | 74 | ((mroute6_socket(net, skb) && |
bd91b8bf | 75 | !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) || |
7bc570c8 YH |
76 | ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr, |
77 | &ipv6_hdr(skb)->saddr))) { | |
1da177e4 LT |
78 | struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC); |
79 | ||
80 | /* Do not check for IFF_ALLMULTI; multicast routing | |
81 | is not supported in any case. | |
82 | */ | |
83 | if (newskb) | |
b2e0b385 | 84 | NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING, |
29a26a56 | 85 | net, sk, newskb, NULL, newskb->dev, |
95603e22 | 86 | dev_loopback_xmit); |
1da177e4 | 87 | |
0660e03f | 88 | if (ipv6_hdr(skb)->hop_limit == 0) { |
78126c41 | 89 | IP6_INC_STATS(net, idev, |
3bd653c8 | 90 | IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
91 | kfree_skb(skb); |
92 | return 0; | |
93 | } | |
94 | } | |
95 | ||
78126c41 | 96 | IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUTMCAST, skb->len); |
dd408515 HFS |
97 | |
98 | if (IPV6_ADDR_MC_SCOPE(&ipv6_hdr(skb)->daddr) <= | |
99 | IPV6_ADDR_SCOPE_NODELOCAL && | |
100 | !(dev->flags & IFF_LOOPBACK)) { | |
101 | kfree_skb(skb); | |
102 | return 0; | |
103 | } | |
1da177e4 LT |
104 | } |
105 | ||
14972cbd RP |
106 | if (lwtunnel_xmit_redirect(dst->lwtstate)) { |
107 | int res = lwtunnel_xmit(skb); | |
108 | ||
109 | if (res < 0 || res == LWTUNNEL_XMIT_DONE) | |
110 | return res; | |
111 | } | |
112 | ||
6fd6ce20 | 113 | rcu_read_lock_bh(); |
2647a9b0 | 114 | nexthop = rt6_nexthop((struct rt6_info *)dst, &ipv6_hdr(skb)->daddr); |
6fd6ce20 YH |
115 | neigh = __ipv6_neigh_lookup_noref(dst->dev, nexthop); |
116 | if (unlikely(!neigh)) | |
117 | neigh = __neigh_create(&nd_tbl, nexthop, dst->dev, false); | |
118 | if (!IS_ERR(neigh)) { | |
4ff06203 | 119 | sock_confirm_neigh(skb, neigh); |
c16ec185 | 120 | ret = neigh_output(neigh, skb); |
6fd6ce20 YH |
121 | rcu_read_unlock_bh(); |
122 | return ret; | |
123 | } | |
124 | rcu_read_unlock_bh(); | |
05e3aa09 | 125 | |
78126c41 | 126 | IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES); |
9e508490 JE |
127 | kfree_skb(skb); |
128 | return -EINVAL; | |
1da177e4 LT |
129 | } |
130 | ||
0c4b51f0 | 131 | static int ip6_finish_output(struct net *net, struct sock *sk, struct sk_buff *skb) |
9e508490 | 132 | { |
33b48679 DM |
133 | int ret; |
134 | ||
135 | ret = BPF_CGROUP_RUN_PROG_INET_EGRESS(sk, skb); | |
136 | if (ret) { | |
137 | kfree_skb(skb); | |
138 | return ret; | |
139 | } | |
140 | ||
9e508490 | 141 | if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) || |
9037c357 JP |
142 | dst_allfrag(skb_dst(skb)) || |
143 | (IP6CB(skb)->frag_max_size && skb->len > IP6CB(skb)->frag_max_size)) | |
7d8c6e39 | 144 | return ip6_fragment(net, sk, skb, ip6_finish_output2); |
9e508490 | 145 | else |
7d8c6e39 | 146 | return ip6_finish_output2(net, sk, skb); |
9e508490 JE |
147 | } |
148 | ||
ede2059d | 149 | int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb) |
1da177e4 | 150 | { |
9e508490 | 151 | struct net_device *dev = skb_dst(skb)->dev; |
adf30907 | 152 | struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb)); |
be10de0a | 153 | |
97a7a37a CF |
154 | skb->protocol = htons(ETH_P_IPV6); |
155 | skb->dev = dev; | |
156 | ||
778d80be | 157 | if (unlikely(idev->cnf.disable_ipv6)) { |
19a0644c | 158 | IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS); |
778d80be YH |
159 | kfree_skb(skb); |
160 | return 0; | |
161 | } | |
162 | ||
29a26a56 EB |
163 | return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING, |
164 | net, sk, skb, NULL, dev, | |
9c6eb28a JE |
165 | ip6_finish_output, |
166 | !(IP6CB(skb)->flags & IP6SKB_REROUTED)); | |
1da177e4 LT |
167 | } |
168 | ||
513674b5 SL |
169 | static bool ip6_autoflowlabel(struct net *net, const struct ipv6_pinfo *np) |
170 | { | |
171 | if (!np->autoflowlabel_set) | |
172 | return ip6_default_np_autolabel(net); | |
173 | else | |
174 | return np->autoflowlabel; | |
175 | } | |
176 | ||
1da177e4 | 177 | /* |
1c1e9d2b ED |
178 | * xmit an sk_buff (used by TCP, SCTP and DCCP) |
179 | * Note : socket lock is not held for SYNACK packets, but might be modified | |
180 | * by calls to skb_set_owner_w() and ipv6_local_error(), | |
181 | * which are using proper atomic operations or spinlocks. | |
1da177e4 | 182 | */ |
1c1e9d2b | 183 | int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6, |
92e55f41 | 184 | __u32 mark, struct ipv6_txoptions *opt, int tclass) |
1da177e4 | 185 | { |
3bd653c8 | 186 | struct net *net = sock_net(sk); |
1c1e9d2b | 187 | const struct ipv6_pinfo *np = inet6_sk(sk); |
4c9483b2 | 188 | struct in6_addr *first_hop = &fl6->daddr; |
adf30907 | 189 | struct dst_entry *dst = skb_dst(skb); |
1da177e4 | 190 | struct ipv6hdr *hdr; |
4c9483b2 | 191 | u8 proto = fl6->flowi6_proto; |
1da177e4 | 192 | int seg_len = skb->len; |
e651f03a | 193 | int hlimit = -1; |
1da177e4 LT |
194 | u32 mtu; |
195 | ||
196 | if (opt) { | |
c2636b4d | 197 | unsigned int head_room; |
1da177e4 LT |
198 | |
199 | /* First: exthdrs may take lots of space (~8K for now) | |
200 | MAX_HEADER is not enough. | |
201 | */ | |
202 | head_room = opt->opt_nflen + opt->opt_flen; | |
203 | seg_len += head_room; | |
204 | head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev); | |
205 | ||
206 | if (skb_headroom(skb) < head_room) { | |
207 | struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room); | |
63159f29 | 208 | if (!skb2) { |
adf30907 | 209 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d YH |
210 | IPSTATS_MIB_OUTDISCARDS); |
211 | kfree_skb(skb); | |
1da177e4 LT |
212 | return -ENOBUFS; |
213 | } | |
808db80a | 214 | consume_skb(skb); |
a11d206d | 215 | skb = skb2; |
1c1e9d2b ED |
216 | /* skb_set_owner_w() changes sk->sk_wmem_alloc atomically, |
217 | * it is safe to call in our context (socket lock not held) | |
218 | */ | |
219 | skb_set_owner_w(skb, (struct sock *)sk); | |
1da177e4 LT |
220 | } |
221 | if (opt->opt_flen) | |
222 | ipv6_push_frag_opts(skb, opt, &proto); | |
223 | if (opt->opt_nflen) | |
613fa3ca DL |
224 | ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop, |
225 | &fl6->saddr); | |
1da177e4 LT |
226 | } |
227 | ||
e2d1bca7 ACM |
228 | skb_push(skb, sizeof(struct ipv6hdr)); |
229 | skb_reset_network_header(skb); | |
0660e03f | 230 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
231 | |
232 | /* | |
233 | * Fill in the IPv6 header | |
234 | */ | |
b903d324 | 235 | if (np) |
1da177e4 LT |
236 | hlimit = np->hop_limit; |
237 | if (hlimit < 0) | |
6b75d090 | 238 | hlimit = ip6_dst_hoplimit(dst); |
1da177e4 | 239 | |
cb1ce2ef | 240 | ip6_flow_hdr(hdr, tclass, ip6_make_flowlabel(net, skb, fl6->flowlabel, |
513674b5 | 241 | ip6_autoflowlabel(net, np), fl6)); |
41a1f8ea | 242 | |
1da177e4 LT |
243 | hdr->payload_len = htons(seg_len); |
244 | hdr->nexthdr = proto; | |
245 | hdr->hop_limit = hlimit; | |
246 | ||
4e3fd7a0 AD |
247 | hdr->saddr = fl6->saddr; |
248 | hdr->daddr = *first_hop; | |
1da177e4 | 249 | |
9c9c9ad5 | 250 | skb->protocol = htons(ETH_P_IPV6); |
a2c2064f | 251 | skb->priority = sk->sk_priority; |
92e55f41 | 252 | skb->mark = mark; |
a2c2064f | 253 | |
1da177e4 | 254 | mtu = dst_mtu(dst); |
60ff7467 | 255 | if ((skb->len <= mtu) || skb->ignore_df || skb_is_gso(skb)) { |
adf30907 | 256 | IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)), |
edf391ff | 257 | IPSTATS_MIB_OUT, skb->len); |
a8e3e1a9 DA |
258 | |
259 | /* if egress device is enslaved to an L3 master device pass the | |
260 | * skb to its handler for processing | |
261 | */ | |
262 | skb = l3mdev_ip6_out((struct sock *)sk, skb); | |
263 | if (unlikely(!skb)) | |
264 | return 0; | |
265 | ||
1c1e9d2b ED |
266 | /* hooks should never assume socket lock is held. |
267 | * we promote our socket to non const | |
268 | */ | |
29a26a56 | 269 | return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, |
1c1e9d2b | 270 | net, (struct sock *)sk, skb, NULL, dst->dev, |
13206b6b | 271 | dst_output); |
1da177e4 LT |
272 | } |
273 | ||
1da177e4 | 274 | skb->dev = dst->dev; |
1c1e9d2b ED |
275 | /* ipv6_local_error() does not require socket lock, |
276 | * we promote our socket to non const | |
277 | */ | |
278 | ipv6_local_error((struct sock *)sk, EMSGSIZE, fl6, mtu); | |
279 | ||
adf30907 | 280 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
281 | kfree_skb(skb); |
282 | return -EMSGSIZE; | |
283 | } | |
7159039a YH |
284 | EXPORT_SYMBOL(ip6_xmit); |
285 | ||
1da177e4 LT |
286 | static int ip6_call_ra_chain(struct sk_buff *skb, int sel) |
287 | { | |
288 | struct ip6_ra_chain *ra; | |
289 | struct sock *last = NULL; | |
290 | ||
291 | read_lock(&ip6_ra_lock); | |
292 | for (ra = ip6_ra_chain; ra; ra = ra->next) { | |
293 | struct sock *sk = ra->sk; | |
0bd1b59b AM |
294 | if (sk && ra->sel == sel && |
295 | (!sk->sk_bound_dev_if || | |
296 | sk->sk_bound_dev_if == skb->dev->ifindex)) { | |
1da177e4 LT |
297 | if (last) { |
298 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
299 | if (skb2) | |
300 | rawv6_rcv(last, skb2); | |
301 | } | |
302 | last = sk; | |
303 | } | |
304 | } | |
305 | ||
306 | if (last) { | |
307 | rawv6_rcv(last, skb); | |
308 | read_unlock(&ip6_ra_lock); | |
309 | return 1; | |
310 | } | |
311 | read_unlock(&ip6_ra_lock); | |
312 | return 0; | |
313 | } | |
314 | ||
e21e0b5f VN |
315 | static int ip6_forward_proxy_check(struct sk_buff *skb) |
316 | { | |
0660e03f | 317 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
e21e0b5f | 318 | u8 nexthdr = hdr->nexthdr; |
75f2811c | 319 | __be16 frag_off; |
e21e0b5f VN |
320 | int offset; |
321 | ||
322 | if (ipv6_ext_hdr(nexthdr)) { | |
75f2811c | 323 | offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off); |
e21e0b5f VN |
324 | if (offset < 0) |
325 | return 0; | |
326 | } else | |
327 | offset = sizeof(struct ipv6hdr); | |
328 | ||
329 | if (nexthdr == IPPROTO_ICMPV6) { | |
330 | struct icmp6hdr *icmp6; | |
331 | ||
d56f90a7 ACM |
332 | if (!pskb_may_pull(skb, (skb_network_header(skb) + |
333 | offset + 1 - skb->data))) | |
e21e0b5f VN |
334 | return 0; |
335 | ||
d56f90a7 | 336 | icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset); |
e21e0b5f VN |
337 | |
338 | switch (icmp6->icmp6_type) { | |
339 | case NDISC_ROUTER_SOLICITATION: | |
340 | case NDISC_ROUTER_ADVERTISEMENT: | |
341 | case NDISC_NEIGHBOUR_SOLICITATION: | |
342 | case NDISC_NEIGHBOUR_ADVERTISEMENT: | |
343 | case NDISC_REDIRECT: | |
344 | /* For reaction involving unicast neighbor discovery | |
345 | * message destined to the proxied address, pass it to | |
346 | * input function. | |
347 | */ | |
348 | return 1; | |
349 | default: | |
350 | break; | |
351 | } | |
352 | } | |
353 | ||
74553b09 VN |
354 | /* |
355 | * The proxying router can't forward traffic sent to a link-local | |
356 | * address, so signal the sender and discard the packet. This | |
357 | * behavior is clarified by the MIPv6 specification. | |
358 | */ | |
359 | if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) { | |
360 | dst_link_failure(skb); | |
361 | return -1; | |
362 | } | |
363 | ||
e21e0b5f VN |
364 | return 0; |
365 | } | |
366 | ||
0c4b51f0 EB |
367 | static inline int ip6_forward_finish(struct net *net, struct sock *sk, |
368 | struct sk_buff *skb) | |
1da177e4 | 369 | { |
13206b6b | 370 | return dst_output(net, sk, skb); |
1da177e4 LT |
371 | } |
372 | ||
0954cf9c HFS |
373 | static unsigned int ip6_dst_mtu_forward(const struct dst_entry *dst) |
374 | { | |
375 | unsigned int mtu; | |
376 | struct inet6_dev *idev; | |
377 | ||
378 | if (dst_metric_locked(dst, RTAX_MTU)) { | |
379 | mtu = dst_metric_raw(dst, RTAX_MTU); | |
380 | if (mtu) | |
381 | return mtu; | |
382 | } | |
383 | ||
384 | mtu = IPV6_MIN_MTU; | |
385 | rcu_read_lock(); | |
386 | idev = __in6_dev_get(dst->dev); | |
387 | if (idev) | |
388 | mtu = idev->cnf.mtu6; | |
389 | rcu_read_unlock(); | |
390 | ||
391 | return mtu; | |
392 | } | |
393 | ||
fe6cc55f FW |
394 | static bool ip6_pkt_too_big(const struct sk_buff *skb, unsigned int mtu) |
395 | { | |
418a3156 | 396 | if (skb->len <= mtu) |
fe6cc55f FW |
397 | return false; |
398 | ||
60ff7467 | 399 | /* ipv6 conntrack defrag sets max_frag_size + ignore_df */ |
fe6cc55f FW |
400 | if (IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu) |
401 | return true; | |
402 | ||
60ff7467 | 403 | if (skb->ignore_df) |
418a3156 FW |
404 | return false; |
405 | ||
ae7ef81e | 406 | if (skb_is_gso(skb) && skb_gso_validate_mtu(skb, mtu)) |
fe6cc55f FW |
407 | return false; |
408 | ||
409 | return true; | |
410 | } | |
411 | ||
1da177e4 LT |
412 | int ip6_forward(struct sk_buff *skb) |
413 | { | |
adf30907 | 414 | struct dst_entry *dst = skb_dst(skb); |
0660e03f | 415 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
1da177e4 | 416 | struct inet6_skb_parm *opt = IP6CB(skb); |
c346dca1 | 417 | struct net *net = dev_net(dst->dev); |
14f3ad6f | 418 | u32 mtu; |
1ab1457c | 419 | |
53b7997f | 420 | if (net->ipv6.devconf_all->forwarding == 0) |
1da177e4 LT |
421 | goto error; |
422 | ||
090f1166 LR |
423 | if (skb->pkt_type != PACKET_HOST) |
424 | goto drop; | |
425 | ||
9ef2e965 HFS |
426 | if (unlikely(skb->sk)) |
427 | goto drop; | |
428 | ||
4497b076 BH |
429 | if (skb_warn_if_lro(skb)) |
430 | goto drop; | |
431 | ||
1da177e4 | 432 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) { |
1d015503 ED |
433 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
434 | IPSTATS_MIB_INDISCARDS); | |
1da177e4 LT |
435 | goto drop; |
436 | } | |
437 | ||
35fc92a9 | 438 | skb_forward_csum(skb); |
1da177e4 LT |
439 | |
440 | /* | |
441 | * We DO NOT make any processing on | |
442 | * RA packets, pushing them to user level AS IS | |
443 | * without ane WARRANTY that application will be able | |
444 | * to interpret them. The reason is that we | |
445 | * cannot make anything clever here. | |
446 | * | |
447 | * We are not end-node, so that if packet contains | |
448 | * AH/ESP, we cannot make anything. | |
449 | * Defragmentation also would be mistake, RA packets | |
450 | * cannot be fragmented, because there is no warranty | |
451 | * that different fragments will go along one path. --ANK | |
452 | */ | |
ab4eb353 YH |
453 | if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) { |
454 | if (ip6_call_ra_chain(skb, ntohs(opt->ra))) | |
1da177e4 LT |
455 | return 0; |
456 | } | |
457 | ||
458 | /* | |
459 | * check and decrement ttl | |
460 | */ | |
461 | if (hdr->hop_limit <= 1) { | |
462 | /* Force OUTPUT device used as source address */ | |
463 | skb->dev = dst->dev; | |
3ffe533c | 464 | icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0); |
1d015503 ED |
465 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
466 | IPSTATS_MIB_INHDRERRORS); | |
1da177e4 LT |
467 | |
468 | kfree_skb(skb); | |
469 | return -ETIMEDOUT; | |
470 | } | |
471 | ||
fbea49e1 | 472 | /* XXX: idev->cnf.proxy_ndp? */ |
53b7997f | 473 | if (net->ipv6.devconf_all->proxy_ndp && |
8a3edd80 | 474 | pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) { |
74553b09 VN |
475 | int proxied = ip6_forward_proxy_check(skb); |
476 | if (proxied > 0) | |
e21e0b5f | 477 | return ip6_input(skb); |
74553b09 | 478 | else if (proxied < 0) { |
1d015503 ED |
479 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
480 | IPSTATS_MIB_INDISCARDS); | |
74553b09 VN |
481 | goto drop; |
482 | } | |
e21e0b5f VN |
483 | } |
484 | ||
1da177e4 | 485 | if (!xfrm6_route_forward(skb)) { |
1d015503 ED |
486 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
487 | IPSTATS_MIB_INDISCARDS); | |
1da177e4 LT |
488 | goto drop; |
489 | } | |
adf30907 | 490 | dst = skb_dst(skb); |
1da177e4 LT |
491 | |
492 | /* IPv6 specs say nothing about it, but it is clear that we cannot | |
493 | send redirects to source routed frames. | |
1e5dc146 | 494 | We don't send redirects to frames decapsulated from IPsec. |
1da177e4 | 495 | */ |
c45a3dfb | 496 | if (skb->dev == dst->dev && opt->srcrt == 0 && !skb_sec_path(skb)) { |
1da177e4 | 497 | struct in6_addr *target = NULL; |
fbfe95a4 | 498 | struct inet_peer *peer; |
1da177e4 | 499 | struct rt6_info *rt; |
1da177e4 LT |
500 | |
501 | /* | |
502 | * incoming and outgoing devices are the same | |
503 | * send a redirect. | |
504 | */ | |
505 | ||
506 | rt = (struct rt6_info *) dst; | |
c45a3dfb DM |
507 | if (rt->rt6i_flags & RTF_GATEWAY) |
508 | target = &rt->rt6i_gateway; | |
1da177e4 LT |
509 | else |
510 | target = &hdr->daddr; | |
511 | ||
fd0273d7 | 512 | peer = inet_getpeer_v6(net->ipv6.peers, &hdr->daddr, 1); |
92d86829 | 513 | |
1da177e4 LT |
514 | /* Limit redirects both by destination (here) |
515 | and by source (inside ndisc_send_redirect) | |
516 | */ | |
fbfe95a4 | 517 | if (inet_peer_xrlim_allow(peer, 1*HZ)) |
4991969a | 518 | ndisc_send_redirect(skb, target); |
1d861aa4 DM |
519 | if (peer) |
520 | inet_putpeer(peer); | |
5bb1ab09 DS |
521 | } else { |
522 | int addrtype = ipv6_addr_type(&hdr->saddr); | |
523 | ||
1da177e4 | 524 | /* This check is security critical. */ |
f81b2e7d YH |
525 | if (addrtype == IPV6_ADDR_ANY || |
526 | addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK)) | |
5bb1ab09 DS |
527 | goto error; |
528 | if (addrtype & IPV6_ADDR_LINKLOCAL) { | |
529 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, | |
3ffe533c | 530 | ICMPV6_NOT_NEIGHBOUR, 0); |
5bb1ab09 DS |
531 | goto error; |
532 | } | |
1da177e4 LT |
533 | } |
534 | ||
0954cf9c | 535 | mtu = ip6_dst_mtu_forward(dst); |
14f3ad6f UW |
536 | if (mtu < IPV6_MIN_MTU) |
537 | mtu = IPV6_MIN_MTU; | |
538 | ||
fe6cc55f | 539 | if (ip6_pkt_too_big(skb, mtu)) { |
1da177e4 LT |
540 | /* Again, force OUTPUT device used as source address */ |
541 | skb->dev = dst->dev; | |
14f3ad6f | 542 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); |
1d015503 ED |
543 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
544 | IPSTATS_MIB_INTOOBIGERRORS); | |
545 | __IP6_INC_STATS(net, ip6_dst_idev(dst), | |
546 | IPSTATS_MIB_FRAGFAILS); | |
1da177e4 LT |
547 | kfree_skb(skb); |
548 | return -EMSGSIZE; | |
549 | } | |
550 | ||
551 | if (skb_cow(skb, dst->dev->hard_header_len)) { | |
1d015503 ED |
552 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
553 | IPSTATS_MIB_OUTDISCARDS); | |
1da177e4 LT |
554 | goto drop; |
555 | } | |
556 | ||
0660e03f | 557 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
558 | |
559 | /* Mangling hops number delayed to point after skb COW */ | |
1ab1457c | 560 | |
1da177e4 LT |
561 | hdr->hop_limit--; |
562 | ||
1d015503 ED |
563 | __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS); |
564 | __IP6_ADD_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len); | |
29a26a56 EB |
565 | return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, |
566 | net, NULL, skb, skb->dev, dst->dev, | |
6e23ae2a | 567 | ip6_forward_finish); |
1da177e4 LT |
568 | |
569 | error: | |
1d015503 | 570 | __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS); |
1da177e4 LT |
571 | drop: |
572 | kfree_skb(skb); | |
573 | return -EINVAL; | |
574 | } | |
575 | ||
576 | static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from) | |
577 | { | |
578 | to->pkt_type = from->pkt_type; | |
579 | to->priority = from->priority; | |
580 | to->protocol = from->protocol; | |
adf30907 ED |
581 | skb_dst_drop(to); |
582 | skb_dst_set(to, dst_clone(skb_dst(from))); | |
1da177e4 | 583 | to->dev = from->dev; |
82e91ffe | 584 | to->mark = from->mark; |
1da177e4 LT |
585 | |
586 | #ifdef CONFIG_NET_SCHED | |
587 | to->tc_index = from->tc_index; | |
588 | #endif | |
e7ac05f3 | 589 | nf_copy(to, from); |
984bc16c | 590 | skb_copy_secmark(to, from); |
1da177e4 LT |
591 | } |
592 | ||
7d8c6e39 EB |
593 | int ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb, |
594 | int (*output)(struct net *, struct sock *, struct sk_buff *)) | |
1da177e4 | 595 | { |
1da177e4 | 596 | struct sk_buff *frag; |
67ba4152 | 597 | struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); |
f60e5990 | 598 | struct ipv6_pinfo *np = skb->sk && !dev_recursion_level() ? |
599 | inet6_sk(skb->sk) : NULL; | |
1da177e4 LT |
600 | struct ipv6hdr *tmp_hdr; |
601 | struct frag_hdr *fh; | |
602 | unsigned int mtu, hlen, left, len; | |
a7ae1992 | 603 | int hroom, troom; |
286c2349 | 604 | __be32 frag_id; |
67ba4152 | 605 | int ptr, offset = 0, err = 0; |
1da177e4 LT |
606 | u8 *prevhdr, nexthdr = 0; |
607 | ||
7dd7eb95 DM |
608 | err = ip6_find_1stfragopt(skb, &prevhdr); |
609 | if (err < 0) | |
2423496a | 610 | goto fail; |
7dd7eb95 | 611 | hlen = err; |
1da177e4 LT |
612 | nexthdr = *prevhdr; |
613 | ||
628a5c56 | 614 | mtu = ip6_skb_dst_mtu(skb); |
b881ef76 JH |
615 | |
616 | /* We must not fragment if the socket is set to force MTU discovery | |
14f3ad6f | 617 | * or if the skb it not generated by a local socket. |
b881ef76 | 618 | */ |
485fca66 FW |
619 | if (unlikely(!skb->ignore_df && skb->len > mtu)) |
620 | goto fail_toobig; | |
a34a101e | 621 | |
485fca66 FW |
622 | if (IP6CB(skb)->frag_max_size) { |
623 | if (IP6CB(skb)->frag_max_size > mtu) | |
624 | goto fail_toobig; | |
625 | ||
626 | /* don't send fragments larger than what we received */ | |
627 | mtu = IP6CB(skb)->frag_max_size; | |
628 | if (mtu < IPV6_MIN_MTU) | |
629 | mtu = IPV6_MIN_MTU; | |
b881ef76 JH |
630 | } |
631 | ||
d91675f9 YH |
632 | if (np && np->frag_size < mtu) { |
633 | if (np->frag_size) | |
634 | mtu = np->frag_size; | |
635 | } | |
89bc7848 | 636 | if (mtu < hlen + sizeof(struct frag_hdr) + 8) |
b72a2b01 | 637 | goto fail_toobig; |
1e0d69a9 | 638 | mtu -= hlen + sizeof(struct frag_hdr); |
1da177e4 | 639 | |
fd0273d7 MKL |
640 | frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr, |
641 | &ipv6_hdr(skb)->saddr); | |
286c2349 | 642 | |
405c92f7 HFS |
643 | if (skb->ip_summed == CHECKSUM_PARTIAL && |
644 | (err = skb_checksum_help(skb))) | |
645 | goto fail; | |
646 | ||
1d325d21 | 647 | hroom = LL_RESERVED_SPACE(rt->dst.dev); |
21dc3301 | 648 | if (skb_has_frag_list(skb)) { |
c72d8cda | 649 | unsigned int first_len = skb_pagelen(skb); |
3d13008e | 650 | struct sk_buff *frag2; |
1da177e4 LT |
651 | |
652 | if (first_len - hlen > mtu || | |
653 | ((first_len - hlen) & 7) || | |
1d325d21 FW |
654 | skb_cloned(skb) || |
655 | skb_headroom(skb) < (hroom + sizeof(struct frag_hdr))) | |
1da177e4 LT |
656 | goto slow_path; |
657 | ||
4d9092bb | 658 | skb_walk_frags(skb, frag) { |
1da177e4 LT |
659 | /* Correct geometry. */ |
660 | if (frag->len > mtu || | |
661 | ((frag->len & 7) && frag->next) || | |
1d325d21 | 662 | skb_headroom(frag) < (hlen + hroom + sizeof(struct frag_hdr))) |
3d13008e | 663 | goto slow_path_clean; |
1da177e4 | 664 | |
1da177e4 LT |
665 | /* Partially cloned skb? */ |
666 | if (skb_shared(frag)) | |
3d13008e | 667 | goto slow_path_clean; |
2fdba6b0 HX |
668 | |
669 | BUG_ON(frag->sk); | |
670 | if (skb->sk) { | |
2fdba6b0 HX |
671 | frag->sk = skb->sk; |
672 | frag->destructor = sock_wfree; | |
2fdba6b0 | 673 | } |
3d13008e | 674 | skb->truesize -= frag->truesize; |
1da177e4 LT |
675 | } |
676 | ||
677 | err = 0; | |
678 | offset = 0; | |
1da177e4 LT |
679 | /* BUILD HEADER */ |
680 | ||
9a217a1c | 681 | *prevhdr = NEXTHDR_FRAGMENT; |
d56f90a7 | 682 | tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC); |
1da177e4 | 683 | if (!tmp_hdr) { |
1d325d21 FW |
684 | err = -ENOMEM; |
685 | goto fail; | |
1da177e4 | 686 | } |
1d325d21 FW |
687 | frag = skb_shinfo(skb)->frag_list; |
688 | skb_frag_list_init(skb); | |
1da177e4 | 689 | |
1da177e4 | 690 | __skb_pull(skb, hlen); |
d58ff351 | 691 | fh = __skb_push(skb, sizeof(struct frag_hdr)); |
e2d1bca7 ACM |
692 | __skb_push(skb, hlen); |
693 | skb_reset_network_header(skb); | |
d56f90a7 | 694 | memcpy(skb_network_header(skb), tmp_hdr, hlen); |
1da177e4 | 695 | |
1da177e4 LT |
696 | fh->nexthdr = nexthdr; |
697 | fh->reserved = 0; | |
698 | fh->frag_off = htons(IP6_MF); | |
286c2349 | 699 | fh->identification = frag_id; |
1da177e4 LT |
700 | |
701 | first_len = skb_pagelen(skb); | |
702 | skb->data_len = first_len - skb_headlen(skb); | |
703 | skb->len = first_len; | |
0660e03f ACM |
704 | ipv6_hdr(skb)->payload_len = htons(first_len - |
705 | sizeof(struct ipv6hdr)); | |
a11d206d | 706 | |
1da177e4 LT |
707 | for (;;) { |
708 | /* Prepare header of the next frame, | |
709 | * before previous one went down. */ | |
710 | if (frag) { | |
711 | frag->ip_summed = CHECKSUM_NONE; | |
badff6d0 | 712 | skb_reset_transport_header(frag); |
d58ff351 | 713 | fh = __skb_push(frag, sizeof(struct frag_hdr)); |
e2d1bca7 ACM |
714 | __skb_push(frag, hlen); |
715 | skb_reset_network_header(frag); | |
d56f90a7 ACM |
716 | memcpy(skb_network_header(frag), tmp_hdr, |
717 | hlen); | |
1da177e4 LT |
718 | offset += skb->len - hlen - sizeof(struct frag_hdr); |
719 | fh->nexthdr = nexthdr; | |
720 | fh->reserved = 0; | |
721 | fh->frag_off = htons(offset); | |
53b24b8f | 722 | if (frag->next) |
1da177e4 LT |
723 | fh->frag_off |= htons(IP6_MF); |
724 | fh->identification = frag_id; | |
0660e03f ACM |
725 | ipv6_hdr(frag)->payload_len = |
726 | htons(frag->len - | |
727 | sizeof(struct ipv6hdr)); | |
1da177e4 LT |
728 | ip6_copy_metadata(frag, skb); |
729 | } | |
1ab1457c | 730 | |
7d8c6e39 | 731 | err = output(net, sk, skb); |
67ba4152 | 732 | if (!err) |
d8d1f30b | 733 | IP6_INC_STATS(net, ip6_dst_idev(&rt->dst), |
3bd653c8 | 734 | IPSTATS_MIB_FRAGCREATES); |
dafee490 | 735 | |
1da177e4 LT |
736 | if (err || !frag) |
737 | break; | |
738 | ||
739 | skb = frag; | |
740 | frag = skb->next; | |
741 | skb->next = NULL; | |
742 | } | |
743 | ||
a51482bd | 744 | kfree(tmp_hdr); |
1da177e4 LT |
745 | |
746 | if (err == 0) { | |
d8d1f30b | 747 | IP6_INC_STATS(net, ip6_dst_idev(&rt->dst), |
3bd653c8 | 748 | IPSTATS_MIB_FRAGOKS); |
1da177e4 LT |
749 | return 0; |
750 | } | |
751 | ||
46cfd725 | 752 | kfree_skb_list(frag); |
1da177e4 | 753 | |
d8d1f30b | 754 | IP6_INC_STATS(net, ip6_dst_idev(&rt->dst), |
3bd653c8 | 755 | IPSTATS_MIB_FRAGFAILS); |
1da177e4 | 756 | return err; |
3d13008e ED |
757 | |
758 | slow_path_clean: | |
759 | skb_walk_frags(skb, frag2) { | |
760 | if (frag2 == frag) | |
761 | break; | |
762 | frag2->sk = NULL; | |
763 | frag2->destructor = NULL; | |
764 | skb->truesize += frag2->truesize; | |
765 | } | |
1da177e4 LT |
766 | } |
767 | ||
768 | slow_path: | |
769 | left = skb->len - hlen; /* Space per frame */ | |
770 | ptr = hlen; /* Where to start from */ | |
771 | ||
772 | /* | |
773 | * Fragment the datagram. | |
774 | */ | |
775 | ||
a7ae1992 | 776 | troom = rt->dst.dev->needed_tailroom; |
1da177e4 LT |
777 | |
778 | /* | |
779 | * Keep copying data until we run out. | |
780 | */ | |
67ba4152 | 781 | while (left > 0) { |
79e49503 FW |
782 | u8 *fragnexthdr_offset; |
783 | ||
1da177e4 LT |
784 | len = left; |
785 | /* IF: it doesn't fit, use 'mtu' - the data space left */ | |
786 | if (len > mtu) | |
787 | len = mtu; | |
25985edc | 788 | /* IF: we are not sending up to and including the packet end |
1da177e4 LT |
789 | then align the next start on an eight byte boundary */ |
790 | if (len < left) { | |
791 | len &= ~7; | |
792 | } | |
1da177e4 | 793 | |
cbffccc9 JP |
794 | /* Allocate buffer */ |
795 | frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) + | |
796 | hroom + troom, GFP_ATOMIC); | |
797 | if (!frag) { | |
1da177e4 LT |
798 | err = -ENOMEM; |
799 | goto fail; | |
800 | } | |
801 | ||
802 | /* | |
803 | * Set up data on packet | |
804 | */ | |
805 | ||
806 | ip6_copy_metadata(frag, skb); | |
a7ae1992 | 807 | skb_reserve(frag, hroom); |
1da177e4 | 808 | skb_put(frag, len + hlen + sizeof(struct frag_hdr)); |
c1d2bbe1 | 809 | skb_reset_network_header(frag); |
badff6d0 | 810 | fh = (struct frag_hdr *)(skb_network_header(frag) + hlen); |
b0e380b1 ACM |
811 | frag->transport_header = (frag->network_header + hlen + |
812 | sizeof(struct frag_hdr)); | |
1da177e4 LT |
813 | |
814 | /* | |
815 | * Charge the memory for the fragment to any owner | |
816 | * it might possess | |
817 | */ | |
818 | if (skb->sk) | |
819 | skb_set_owner_w(frag, skb->sk); | |
820 | ||
821 | /* | |
822 | * Copy the packet header into the new buffer. | |
823 | */ | |
d626f62b | 824 | skb_copy_from_linear_data(skb, skb_network_header(frag), hlen); |
1da177e4 | 825 | |
79e49503 FW |
826 | fragnexthdr_offset = skb_network_header(frag); |
827 | fragnexthdr_offset += prevhdr - skb_network_header(skb); | |
828 | *fragnexthdr_offset = NEXTHDR_FRAGMENT; | |
829 | ||
1da177e4 LT |
830 | /* |
831 | * Build fragment header. | |
832 | */ | |
833 | fh->nexthdr = nexthdr; | |
834 | fh->reserved = 0; | |
286c2349 | 835 | fh->identification = frag_id; |
1da177e4 LT |
836 | |
837 | /* | |
838 | * Copy a block of the IP datagram. | |
839 | */ | |
e3f0b86b HS |
840 | BUG_ON(skb_copy_bits(skb, ptr, skb_transport_header(frag), |
841 | len)); | |
1da177e4 LT |
842 | left -= len; |
843 | ||
844 | fh->frag_off = htons(offset); | |
845 | if (left > 0) | |
846 | fh->frag_off |= htons(IP6_MF); | |
0660e03f ACM |
847 | ipv6_hdr(frag)->payload_len = htons(frag->len - |
848 | sizeof(struct ipv6hdr)); | |
1da177e4 LT |
849 | |
850 | ptr += len; | |
851 | offset += len; | |
852 | ||
853 | /* | |
854 | * Put this fragment into the sending queue. | |
855 | */ | |
7d8c6e39 | 856 | err = output(net, sk, frag); |
1da177e4 LT |
857 | if (err) |
858 | goto fail; | |
dafee490 | 859 | |
adf30907 | 860 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
3bd653c8 | 861 | IPSTATS_MIB_FRAGCREATES); |
1da177e4 | 862 | } |
adf30907 | 863 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 864 | IPSTATS_MIB_FRAGOKS); |
808db80a | 865 | consume_skb(skb); |
1da177e4 LT |
866 | return err; |
867 | ||
485fca66 FW |
868 | fail_toobig: |
869 | if (skb->sk && dst_allfrag(skb_dst(skb))) | |
870 | sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK); | |
871 | ||
485fca66 FW |
872 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); |
873 | err = -EMSGSIZE; | |
874 | ||
1da177e4 | 875 | fail: |
adf30907 | 876 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 877 | IPSTATS_MIB_FRAGFAILS); |
1ab1457c | 878 | kfree_skb(skb); |
1da177e4 LT |
879 | return err; |
880 | } | |
881 | ||
b71d1d42 ED |
882 | static inline int ip6_rt_check(const struct rt6key *rt_key, |
883 | const struct in6_addr *fl_addr, | |
884 | const struct in6_addr *addr_cache) | |
cf6b1982 | 885 | { |
a02cec21 | 886 | return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) && |
63159f29 | 887 | (!addr_cache || !ipv6_addr_equal(fl_addr, addr_cache)); |
cf6b1982 YH |
888 | } |
889 | ||
497c615a HX |
890 | static struct dst_entry *ip6_sk_dst_check(struct sock *sk, |
891 | struct dst_entry *dst, | |
b71d1d42 | 892 | const struct flowi6 *fl6) |
1da177e4 | 893 | { |
497c615a | 894 | struct ipv6_pinfo *np = inet6_sk(sk); |
a963a37d | 895 | struct rt6_info *rt; |
1da177e4 | 896 | |
497c615a HX |
897 | if (!dst) |
898 | goto out; | |
899 | ||
a963a37d ED |
900 | if (dst->ops->family != AF_INET6) { |
901 | dst_release(dst); | |
902 | return NULL; | |
903 | } | |
904 | ||
905 | rt = (struct rt6_info *)dst; | |
497c615a HX |
906 | /* Yes, checking route validity in not connected |
907 | * case is not very simple. Take into account, | |
908 | * that we do not support routing by source, TOS, | |
67ba4152 | 909 | * and MSG_DONTROUTE --ANK (980726) |
497c615a | 910 | * |
cf6b1982 YH |
911 | * 1. ip6_rt_check(): If route was host route, |
912 | * check that cached destination is current. | |
497c615a HX |
913 | * If it is network route, we still may |
914 | * check its validity using saved pointer | |
915 | * to the last used address: daddr_cache. | |
916 | * We do not want to save whole address now, | |
917 | * (because main consumer of this service | |
918 | * is tcp, which has not this problem), | |
919 | * so that the last trick works only on connected | |
920 | * sockets. | |
921 | * 2. oif also should be the same. | |
922 | */ | |
4c9483b2 | 923 | if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) || |
8e1ef0a9 | 924 | #ifdef CONFIG_IPV6_SUBTREES |
4c9483b2 | 925 | ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) || |
8e1ef0a9 | 926 | #endif |
ca254490 DA |
927 | (!(fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF) && |
928 | (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex))) { | |
497c615a HX |
929 | dst_release(dst); |
930 | dst = NULL; | |
1da177e4 LT |
931 | } |
932 | ||
497c615a HX |
933 | out: |
934 | return dst; | |
935 | } | |
936 | ||
3aef934f | 937 | static int ip6_dst_lookup_tail(struct net *net, const struct sock *sk, |
4c9483b2 | 938 | struct dst_entry **dst, struct flowi6 *fl6) |
497c615a | 939 | { |
69cce1d1 DM |
940 | #ifdef CONFIG_IPV6_OPTIMISTIC_DAD |
941 | struct neighbour *n; | |
97cac082 | 942 | struct rt6_info *rt; |
69cce1d1 DM |
943 | #endif |
944 | int err; | |
6f21c96a | 945 | int flags = 0; |
497c615a | 946 | |
e16e888b MS |
947 | /* The correct way to handle this would be to do |
948 | * ip6_route_get_saddr, and then ip6_route_output; however, | |
949 | * the route-specific preferred source forces the | |
950 | * ip6_route_output call _before_ ip6_route_get_saddr. | |
951 | * | |
952 | * In source specific routing (no src=any default route), | |
953 | * ip6_route_output will fail given src=any saddr, though, so | |
954 | * that's why we try it again later. | |
955 | */ | |
956 | if (ipv6_addr_any(&fl6->saddr) && (!*dst || !(*dst)->error)) { | |
957 | struct rt6_info *rt; | |
958 | bool had_dst = *dst != NULL; | |
1da177e4 | 959 | |
e16e888b MS |
960 | if (!had_dst) |
961 | *dst = ip6_route_output(net, sk, fl6); | |
962 | rt = (*dst)->error ? NULL : (struct rt6_info *)*dst; | |
c3968a85 DW |
963 | err = ip6_route_get_saddr(net, rt, &fl6->daddr, |
964 | sk ? inet6_sk(sk)->srcprefs : 0, | |
965 | &fl6->saddr); | |
44456d37 | 966 | if (err) |
1da177e4 | 967 | goto out_err_release; |
e16e888b MS |
968 | |
969 | /* If we had an erroneous initial result, pretend it | |
970 | * never existed and let the SA-enabled version take | |
971 | * over. | |
972 | */ | |
973 | if (!had_dst && (*dst)->error) { | |
974 | dst_release(*dst); | |
975 | *dst = NULL; | |
976 | } | |
6f21c96a PA |
977 | |
978 | if (fl6->flowi6_oif) | |
979 | flags |= RT6_LOOKUP_F_IFACE; | |
1da177e4 LT |
980 | } |
981 | ||
e16e888b | 982 | if (!*dst) |
6f21c96a | 983 | *dst = ip6_route_output_flags(net, sk, fl6, flags); |
e16e888b MS |
984 | |
985 | err = (*dst)->error; | |
986 | if (err) | |
987 | goto out_err_release; | |
988 | ||
95c385b4 | 989 | #ifdef CONFIG_IPV6_OPTIMISTIC_DAD |
e550dfb0 NH |
990 | /* |
991 | * Here if the dst entry we've looked up | |
992 | * has a neighbour entry that is in the INCOMPLETE | |
993 | * state and the src address from the flow is | |
994 | * marked as OPTIMISTIC, we release the found | |
995 | * dst entry and replace it instead with the | |
996 | * dst entry of the nexthop router | |
997 | */ | |
c56bf6fe | 998 | rt = (struct rt6_info *) *dst; |
707be1ff | 999 | rcu_read_lock_bh(); |
2647a9b0 MKL |
1000 | n = __ipv6_neigh_lookup_noref(rt->dst.dev, |
1001 | rt6_nexthop(rt, &fl6->daddr)); | |
707be1ff YH |
1002 | err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0; |
1003 | rcu_read_unlock_bh(); | |
1004 | ||
1005 | if (err) { | |
e550dfb0 | 1006 | struct inet6_ifaddr *ifp; |
4c9483b2 | 1007 | struct flowi6 fl_gw6; |
e550dfb0 NH |
1008 | int redirect; |
1009 | ||
4c9483b2 | 1010 | ifp = ipv6_get_ifaddr(net, &fl6->saddr, |
e550dfb0 NH |
1011 | (*dst)->dev, 1); |
1012 | ||
1013 | redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC); | |
1014 | if (ifp) | |
1015 | in6_ifa_put(ifp); | |
1016 | ||
1017 | if (redirect) { | |
1018 | /* | |
1019 | * We need to get the dst entry for the | |
1020 | * default router instead | |
1021 | */ | |
1022 | dst_release(*dst); | |
4c9483b2 DM |
1023 | memcpy(&fl_gw6, fl6, sizeof(struct flowi6)); |
1024 | memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr)); | |
1025 | *dst = ip6_route_output(net, sk, &fl_gw6); | |
e5d08d71 IM |
1026 | err = (*dst)->error; |
1027 | if (err) | |
e550dfb0 | 1028 | goto out_err_release; |
95c385b4 | 1029 | } |
e550dfb0 | 1030 | } |
95c385b4 | 1031 | #endif |
ec5e3b0a | 1032 | if (ipv6_addr_v4mapped(&fl6->saddr) && |
00ea1cee WB |
1033 | !(ipv6_addr_v4mapped(&fl6->daddr) || ipv6_addr_any(&fl6->daddr))) { |
1034 | err = -EAFNOSUPPORT; | |
1035 | goto out_err_release; | |
1036 | } | |
95c385b4 | 1037 | |
1da177e4 LT |
1038 | return 0; |
1039 | ||
1040 | out_err_release: | |
1041 | dst_release(*dst); | |
1042 | *dst = NULL; | |
8a966fc0 | 1043 | |
0d240e78 DA |
1044 | if (err == -ENETUNREACH) |
1045 | IP6_INC_STATS(net, NULL, IPSTATS_MIB_OUTNOROUTES); | |
1da177e4 LT |
1046 | return err; |
1047 | } | |
34a0b3cd | 1048 | |
497c615a HX |
1049 | /** |
1050 | * ip6_dst_lookup - perform route lookup on flow | |
1051 | * @sk: socket which provides route info | |
1052 | * @dst: pointer to dst_entry * for result | |
4c9483b2 | 1053 | * @fl6: flow to lookup |
497c615a HX |
1054 | * |
1055 | * This function performs a route lookup on the given flow. | |
1056 | * | |
1057 | * It returns zero on success, or a standard errno code on error. | |
1058 | */ | |
343d60aa RP |
1059 | int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst, |
1060 | struct flowi6 *fl6) | |
497c615a HX |
1061 | { |
1062 | *dst = NULL; | |
343d60aa | 1063 | return ip6_dst_lookup_tail(net, sk, dst, fl6); |
497c615a | 1064 | } |
3cf3dc6c ACM |
1065 | EXPORT_SYMBOL_GPL(ip6_dst_lookup); |
1066 | ||
497c615a | 1067 | /** |
68d0c6d3 DM |
1068 | * ip6_dst_lookup_flow - perform route lookup on flow with ipsec |
1069 | * @sk: socket which provides route info | |
4c9483b2 | 1070 | * @fl6: flow to lookup |
68d0c6d3 | 1071 | * @final_dst: final destination address for ipsec lookup |
68d0c6d3 DM |
1072 | * |
1073 | * This function performs a route lookup on the given flow. | |
1074 | * | |
1075 | * It returns a valid dst pointer on success, or a pointer encoded | |
1076 | * error code. | |
1077 | */ | |
3aef934f | 1078 | struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6, |
0e0d44ab | 1079 | const struct in6_addr *final_dst) |
68d0c6d3 DM |
1080 | { |
1081 | struct dst_entry *dst = NULL; | |
1082 | int err; | |
1083 | ||
343d60aa | 1084 | err = ip6_dst_lookup_tail(sock_net(sk), sk, &dst, fl6); |
68d0c6d3 DM |
1085 | if (err) |
1086 | return ERR_PTR(err); | |
1087 | if (final_dst) | |
4e3fd7a0 | 1088 | fl6->daddr = *final_dst; |
2774c131 | 1089 | |
f92ee619 | 1090 | return xfrm_lookup_route(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0); |
68d0c6d3 DM |
1091 | } |
1092 | EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow); | |
1093 | ||
1094 | /** | |
1095 | * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow | |
497c615a | 1096 | * @sk: socket which provides the dst cache and route info |
4c9483b2 | 1097 | * @fl6: flow to lookup |
68d0c6d3 | 1098 | * @final_dst: final destination address for ipsec lookup |
497c615a HX |
1099 | * |
1100 | * This function performs a route lookup on the given flow with the | |
1101 | * possibility of using the cached route in the socket if it is valid. | |
1102 | * It will take the socket dst lock when operating on the dst cache. | |
1103 | * As a result, this function can only be used in process context. | |
1104 | * | |
68d0c6d3 DM |
1105 | * It returns a valid dst pointer on success, or a pointer encoded |
1106 | * error code. | |
497c615a | 1107 | */ |
4c9483b2 | 1108 | struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6, |
0e0d44ab | 1109 | const struct in6_addr *final_dst) |
497c615a | 1110 | { |
68d0c6d3 | 1111 | struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie); |
497c615a | 1112 | |
4c9483b2 | 1113 | dst = ip6_sk_dst_check(sk, dst, fl6); |
00bc0ef5 JS |
1114 | if (!dst) |
1115 | dst = ip6_dst_lookup_flow(sk, fl6, final_dst); | |
68d0c6d3 | 1116 | |
00bc0ef5 | 1117 | return dst; |
497c615a | 1118 | } |
68d0c6d3 | 1119 | EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow); |
497c615a | 1120 | |
0178b695 HX |
1121 | static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src, |
1122 | gfp_t gfp) | |
1123 | { | |
1124 | return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL; | |
1125 | } | |
1126 | ||
1127 | static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src, | |
1128 | gfp_t gfp) | |
1129 | { | |
1130 | return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL; | |
1131 | } | |
1132 | ||
75a493e6 | 1133 | static void ip6_append_data_mtu(unsigned int *mtu, |
0c183379 G |
1134 | int *maxfraglen, |
1135 | unsigned int fragheaderlen, | |
1136 | struct sk_buff *skb, | |
75a493e6 | 1137 | struct rt6_info *rt, |
e367c2d0 | 1138 | unsigned int orig_mtu) |
0c183379 G |
1139 | { |
1140 | if (!(rt->dst.flags & DST_XFRM_TUNNEL)) { | |
63159f29 | 1141 | if (!skb) { |
0c183379 | 1142 | /* first fragment, reserve header_len */ |
e367c2d0 | 1143 | *mtu = orig_mtu - rt->dst.header_len; |
0c183379 G |
1144 | |
1145 | } else { | |
1146 | /* | |
1147 | * this fragment is not first, the headers | |
1148 | * space is regarded as data space. | |
1149 | */ | |
e367c2d0 | 1150 | *mtu = orig_mtu; |
0c183379 G |
1151 | } |
1152 | *maxfraglen = ((*mtu - fragheaderlen) & ~7) | |
1153 | + fragheaderlen - sizeof(struct frag_hdr); | |
1154 | } | |
1155 | } | |
1156 | ||
366e41d9 | 1157 | static int ip6_setup_cork(struct sock *sk, struct inet_cork_full *cork, |
26879da5 | 1158 | struct inet6_cork *v6_cork, struct ipcm6_cookie *ipc6, |
366e41d9 VY |
1159 | struct rt6_info *rt, struct flowi6 *fl6) |
1160 | { | |
1161 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1162 | unsigned int mtu; | |
26879da5 | 1163 | struct ipv6_txoptions *opt = ipc6->opt; |
366e41d9 VY |
1164 | |
1165 | /* | |
1166 | * setup for corking | |
1167 | */ | |
1168 | if (opt) { | |
1169 | if (WARN_ON(v6_cork->opt)) | |
1170 | return -EINVAL; | |
1171 | ||
864e2a1f | 1172 | v6_cork->opt = kzalloc(sizeof(*opt), sk->sk_allocation); |
63159f29 | 1173 | if (unlikely(!v6_cork->opt)) |
366e41d9 VY |
1174 | return -ENOBUFS; |
1175 | ||
864e2a1f | 1176 | v6_cork->opt->tot_len = sizeof(*opt); |
366e41d9 VY |
1177 | v6_cork->opt->opt_flen = opt->opt_flen; |
1178 | v6_cork->opt->opt_nflen = opt->opt_nflen; | |
1179 | ||
1180 | v6_cork->opt->dst0opt = ip6_opt_dup(opt->dst0opt, | |
1181 | sk->sk_allocation); | |
1182 | if (opt->dst0opt && !v6_cork->opt->dst0opt) | |
1183 | return -ENOBUFS; | |
1184 | ||
1185 | v6_cork->opt->dst1opt = ip6_opt_dup(opt->dst1opt, | |
1186 | sk->sk_allocation); | |
1187 | if (opt->dst1opt && !v6_cork->opt->dst1opt) | |
1188 | return -ENOBUFS; | |
1189 | ||
1190 | v6_cork->opt->hopopt = ip6_opt_dup(opt->hopopt, | |
1191 | sk->sk_allocation); | |
1192 | if (opt->hopopt && !v6_cork->opt->hopopt) | |
1193 | return -ENOBUFS; | |
1194 | ||
1195 | v6_cork->opt->srcrt = ip6_rthdr_dup(opt->srcrt, | |
1196 | sk->sk_allocation); | |
1197 | if (opt->srcrt && !v6_cork->opt->srcrt) | |
1198 | return -ENOBUFS; | |
1199 | ||
1200 | /* need source address above miyazawa*/ | |
1201 | } | |
1202 | dst_hold(&rt->dst); | |
1203 | cork->base.dst = &rt->dst; | |
1204 | cork->fl.u.ip6 = *fl6; | |
26879da5 WW |
1205 | v6_cork->hop_limit = ipc6->hlimit; |
1206 | v6_cork->tclass = ipc6->tclass; | |
366e41d9 VY |
1207 | if (rt->dst.flags & DST_XFRM_TUNNEL) |
1208 | mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ? | |
1209 | rt->dst.dev->mtu : dst_mtu(&rt->dst); | |
1210 | else | |
1211 | mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ? | |
1212 | rt->dst.dev->mtu : dst_mtu(rt->dst.path); | |
1213 | if (np->frag_size < mtu) { | |
1214 | if (np->frag_size) | |
1215 | mtu = np->frag_size; | |
1216 | } | |
1217 | cork->base.fragsize = mtu; | |
1218 | if (dst_allfrag(rt->dst.path)) | |
1219 | cork->base.flags |= IPCORK_ALLFRAG; | |
1220 | cork->base.length = 0; | |
1221 | ||
1222 | return 0; | |
1223 | } | |
1224 | ||
0bbe84a6 VY |
1225 | static int __ip6_append_data(struct sock *sk, |
1226 | struct flowi6 *fl6, | |
1227 | struct sk_buff_head *queue, | |
1228 | struct inet_cork *cork, | |
1229 | struct inet6_cork *v6_cork, | |
1230 | struct page_frag *pfrag, | |
1231 | int getfrag(void *from, char *to, int offset, | |
1232 | int len, int odd, struct sk_buff *skb), | |
1233 | void *from, int length, int transhdrlen, | |
26879da5 | 1234 | unsigned int flags, struct ipcm6_cookie *ipc6, |
c14ac945 | 1235 | const struct sockcm_cookie *sockc) |
1da177e4 | 1236 | { |
0c183379 | 1237 | struct sk_buff *skb, *skb_prev = NULL; |
e367c2d0 | 1238 | unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu; |
0bbe84a6 VY |
1239 | int exthdrlen = 0; |
1240 | int dst_exthdrlen = 0; | |
1da177e4 | 1241 | int hh_len; |
1da177e4 LT |
1242 | int copy; |
1243 | int err; | |
1244 | int offset = 0; | |
a693e698 | 1245 | __u8 tx_flags = 0; |
09c2d251 | 1246 | u32 tskey = 0; |
0bbe84a6 VY |
1247 | struct rt6_info *rt = (struct rt6_info *)cork->dst; |
1248 | struct ipv6_txoptions *opt = v6_cork->opt; | |
32dce968 | 1249 | int csummode = CHECKSUM_NONE; |
682b1a9d | 1250 | unsigned int maxnonfragsize, headersize; |
1da177e4 | 1251 | |
0bbe84a6 VY |
1252 | skb = skb_peek_tail(queue); |
1253 | if (!skb) { | |
1254 | exthdrlen = opt ? opt->opt_flen : 0; | |
7efdba5b | 1255 | dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len; |
1da177e4 | 1256 | } |
0bbe84a6 | 1257 | |
366e41d9 | 1258 | mtu = cork->fragsize; |
e367c2d0 | 1259 | orig_mtu = mtu; |
1da177e4 | 1260 | |
d8d1f30b | 1261 | hh_len = LL_RESERVED_SPACE(rt->dst.dev); |
1da177e4 | 1262 | |
a1b05140 | 1263 | fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len + |
b4ce9277 | 1264 | (opt ? opt->opt_nflen : 0); |
4df98e76 HFS |
1265 | maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - |
1266 | sizeof(struct frag_hdr); | |
1da177e4 | 1267 | |
682b1a9d HFS |
1268 | headersize = sizeof(struct ipv6hdr) + |
1269 | (opt ? opt->opt_flen + opt->opt_nflen : 0) + | |
1270 | (dst_allfrag(&rt->dst) ? | |
1271 | sizeof(struct frag_hdr) : 0) + | |
1272 | rt->rt6i_nfheader_len; | |
1273 | ||
26879da5 | 1274 | if (cork->length + length > mtu - headersize && ipc6->dontfrag && |
682b1a9d HFS |
1275 | (sk->sk_protocol == IPPROTO_UDP || |
1276 | sk->sk_protocol == IPPROTO_RAW)) { | |
1277 | ipv6_local_rxpmtu(sk, fl6, mtu - headersize + | |
1278 | sizeof(struct ipv6hdr)); | |
1279 | goto emsgsize; | |
1280 | } | |
4df98e76 | 1281 | |
682b1a9d HFS |
1282 | if (ip6_sk_ignore_df(sk)) |
1283 | maxnonfragsize = sizeof(struct ipv6hdr) + IPV6_MAXPLEN; | |
1284 | else | |
1285 | maxnonfragsize = mtu; | |
4df98e76 | 1286 | |
682b1a9d | 1287 | if (cork->length + length > maxnonfragsize - headersize) { |
4df98e76 | 1288 | emsgsize: |
682b1a9d HFS |
1289 | ipv6_local_error(sk, EMSGSIZE, fl6, |
1290 | mtu - headersize + | |
1291 | sizeof(struct ipv6hdr)); | |
1292 | return -EMSGSIZE; | |
1da177e4 LT |
1293 | } |
1294 | ||
682b1a9d HFS |
1295 | /* CHECKSUM_PARTIAL only with no extension headers and when |
1296 | * we are not going to fragment | |
1297 | */ | |
1298 | if (transhdrlen && sk->sk_protocol == IPPROTO_UDP && | |
1299 | headersize == sizeof(struct ipv6hdr) && | |
2b89ed65 | 1300 | length <= mtu - headersize && |
682b1a9d | 1301 | !(flags & MSG_MORE) && |
c8cd0989 | 1302 | rt->dst.dev->features & (NETIF_F_IPV6_CSUM | NETIF_F_HW_CSUM)) |
682b1a9d HFS |
1303 | csummode = CHECKSUM_PARTIAL; |
1304 | ||
09c2d251 | 1305 | if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_RAW) { |
c14ac945 | 1306 | sock_tx_timestamp(sk, sockc->tsflags, &tx_flags); |
09c2d251 WB |
1307 | if (tx_flags & SKBTX_ANY_SW_TSTAMP && |
1308 | sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID) | |
1309 | tskey = sk->sk_tskey++; | |
1310 | } | |
a693e698 | 1311 | |
1da177e4 LT |
1312 | /* |
1313 | * Let's try using as much space as possible. | |
1314 | * Use MTU if total length of the message fits into the MTU. | |
1315 | * Otherwise, we need to reserve fragment header and | |
1316 | * fragment alignment (= 8-15 octects, in total). | |
1317 | * | |
1318 | * Note that we may need to "move" the data from the tail of | |
1ab1457c | 1319 | * of the buffer to the new fragment when we split |
1da177e4 LT |
1320 | * the message. |
1321 | * | |
1ab1457c | 1322 | * FIXME: It may be fragmented into multiple chunks |
1da177e4 LT |
1323 | * at once if non-fragmentable extension headers |
1324 | * are too large. | |
1ab1457c | 1325 | * --yoshfuji |
1da177e4 LT |
1326 | */ |
1327 | ||
2811ebac | 1328 | cork->length += length; |
2811ebac | 1329 | if (!skb) |
1da177e4 LT |
1330 | goto alloc_new_skb; |
1331 | ||
1332 | while (length > 0) { | |
1333 | /* Check if the remaining data fits into current packet. */ | |
bdc712b4 | 1334 | copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len; |
1da177e4 LT |
1335 | if (copy < length) |
1336 | copy = maxfraglen - skb->len; | |
1337 | ||
1338 | if (copy <= 0) { | |
1339 | char *data; | |
1340 | unsigned int datalen; | |
1341 | unsigned int fraglen; | |
1342 | unsigned int fraggap; | |
1343 | unsigned int alloclen; | |
1da177e4 | 1344 | alloc_new_skb: |
1da177e4 | 1345 | /* There's no room in the current skb */ |
0c183379 G |
1346 | if (skb) |
1347 | fraggap = skb->len - maxfraglen; | |
1da177e4 LT |
1348 | else |
1349 | fraggap = 0; | |
0c183379 | 1350 | /* update mtu and maxfraglen if necessary */ |
63159f29 | 1351 | if (!skb || !skb_prev) |
0c183379 | 1352 | ip6_append_data_mtu(&mtu, &maxfraglen, |
75a493e6 | 1353 | fragheaderlen, skb, rt, |
e367c2d0 | 1354 | orig_mtu); |
0c183379 G |
1355 | |
1356 | skb_prev = skb; | |
1da177e4 LT |
1357 | |
1358 | /* | |
1359 | * If remaining data exceeds the mtu, | |
1360 | * we know we need more fragment(s). | |
1361 | */ | |
1362 | datalen = length + fraggap; | |
1da177e4 | 1363 | |
0c183379 G |
1364 | if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen) |
1365 | datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len; | |
1da177e4 | 1366 | if ((flags & MSG_MORE) && |
d8d1f30b | 1367 | !(rt->dst.dev->features&NETIF_F_SG)) |
1da177e4 LT |
1368 | alloclen = mtu; |
1369 | else | |
1370 | alloclen = datalen + fragheaderlen; | |
1371 | ||
299b0767 SK |
1372 | alloclen += dst_exthdrlen; |
1373 | ||
0c183379 G |
1374 | if (datalen != length + fraggap) { |
1375 | /* | |
1376 | * this is not the last fragment, the trailer | |
1377 | * space is regarded as data space. | |
1378 | */ | |
1379 | datalen += rt->dst.trailer_len; | |
1380 | } | |
1381 | ||
1382 | alloclen += rt->dst.trailer_len; | |
1383 | fraglen = datalen + fragheaderlen; | |
1da177e4 LT |
1384 | |
1385 | /* | |
1386 | * We just reserve space for fragment header. | |
1ab1457c | 1387 | * Note: this may be overallocation if the message |
1da177e4 LT |
1388 | * (without MSG_MORE) fits into the MTU. |
1389 | */ | |
1390 | alloclen += sizeof(struct frag_hdr); | |
1391 | ||
232cd35d ED |
1392 | copy = datalen - transhdrlen - fraggap; |
1393 | if (copy < 0) { | |
1394 | err = -EINVAL; | |
1395 | goto error; | |
1396 | } | |
1da177e4 LT |
1397 | if (transhdrlen) { |
1398 | skb = sock_alloc_send_skb(sk, | |
1399 | alloclen + hh_len, | |
1400 | (flags & MSG_DONTWAIT), &err); | |
1401 | } else { | |
1402 | skb = NULL; | |
14afee4b | 1403 | if (refcount_read(&sk->sk_wmem_alloc) <= |
1da177e4 LT |
1404 | 2 * sk->sk_sndbuf) |
1405 | skb = sock_wmalloc(sk, | |
1406 | alloclen + hh_len, 1, | |
1407 | sk->sk_allocation); | |
63159f29 | 1408 | if (unlikely(!skb)) |
1da177e4 LT |
1409 | err = -ENOBUFS; |
1410 | } | |
63159f29 | 1411 | if (!skb) |
1da177e4 LT |
1412 | goto error; |
1413 | /* | |
1414 | * Fill in the control structures | |
1415 | */ | |
9c9c9ad5 | 1416 | skb->protocol = htons(ETH_P_IPV6); |
32dce968 | 1417 | skb->ip_summed = csummode; |
1da177e4 | 1418 | skb->csum = 0; |
1f85851e G |
1419 | /* reserve for fragmentation and ipsec header */ |
1420 | skb_reserve(skb, hh_len + sizeof(struct frag_hdr) + | |
1421 | dst_exthdrlen); | |
1da177e4 | 1422 | |
11878b40 WB |
1423 | /* Only the initial fragment is time stamped */ |
1424 | skb_shinfo(skb)->tx_flags = tx_flags; | |
1425 | tx_flags = 0; | |
09c2d251 WB |
1426 | skb_shinfo(skb)->tskey = tskey; |
1427 | tskey = 0; | |
a693e698 | 1428 | |
1da177e4 LT |
1429 | /* |
1430 | * Find where to start putting bytes | |
1431 | */ | |
1f85851e G |
1432 | data = skb_put(skb, fraglen); |
1433 | skb_set_network_header(skb, exthdrlen); | |
1434 | data += fragheaderlen; | |
b0e380b1 ACM |
1435 | skb->transport_header = (skb->network_header + |
1436 | fragheaderlen); | |
1da177e4 LT |
1437 | if (fraggap) { |
1438 | skb->csum = skb_copy_and_csum_bits( | |
1439 | skb_prev, maxfraglen, | |
1440 | data + transhdrlen, fraggap, 0); | |
1441 | skb_prev->csum = csum_sub(skb_prev->csum, | |
1442 | skb->csum); | |
1443 | data += fraggap; | |
e9fa4f7b | 1444 | pskb_trim_unique(skb_prev, maxfraglen); |
1da177e4 | 1445 | } |
232cd35d ED |
1446 | if (copy > 0 && |
1447 | getfrag(from, data + transhdrlen, offset, | |
1448 | copy, fraggap, skb) < 0) { | |
1da177e4 LT |
1449 | err = -EFAULT; |
1450 | kfree_skb(skb); | |
1451 | goto error; | |
1452 | } | |
1453 | ||
1454 | offset += copy; | |
1455 | length -= datalen - fraggap; | |
1456 | transhdrlen = 0; | |
1457 | exthdrlen = 0; | |
299b0767 | 1458 | dst_exthdrlen = 0; |
1da177e4 | 1459 | |
0dec879f JA |
1460 | if ((flags & MSG_CONFIRM) && !skb_prev) |
1461 | skb_set_dst_pending_confirm(skb, 1); | |
1462 | ||
1da177e4 LT |
1463 | /* |
1464 | * Put the packet on the pending queue | |
1465 | */ | |
0bbe84a6 | 1466 | __skb_queue_tail(queue, skb); |
1da177e4 LT |
1467 | continue; |
1468 | } | |
1469 | ||
1470 | if (copy > length) | |
1471 | copy = length; | |
1472 | ||
d8d1f30b | 1473 | if (!(rt->dst.dev->features&NETIF_F_SG)) { |
1da177e4 LT |
1474 | unsigned int off; |
1475 | ||
1476 | off = skb->len; | |
1477 | if (getfrag(from, skb_put(skb, copy), | |
1478 | offset, copy, off, skb) < 0) { | |
1479 | __skb_trim(skb, off); | |
1480 | err = -EFAULT; | |
1481 | goto error; | |
1482 | } | |
1483 | } else { | |
1484 | int i = skb_shinfo(skb)->nr_frags; | |
1da177e4 | 1485 | |
5640f768 ED |
1486 | err = -ENOMEM; |
1487 | if (!sk_page_frag_refill(sk, pfrag)) | |
1da177e4 | 1488 | goto error; |
5640f768 ED |
1489 | |
1490 | if (!skb_can_coalesce(skb, i, pfrag->page, | |
1491 | pfrag->offset)) { | |
1492 | err = -EMSGSIZE; | |
1493 | if (i == MAX_SKB_FRAGS) | |
1494 | goto error; | |
1495 | ||
1496 | __skb_fill_page_desc(skb, i, pfrag->page, | |
1497 | pfrag->offset, 0); | |
1498 | skb_shinfo(skb)->nr_frags = ++i; | |
1499 | get_page(pfrag->page); | |
1da177e4 | 1500 | } |
5640f768 | 1501 | copy = min_t(int, copy, pfrag->size - pfrag->offset); |
9e903e08 | 1502 | if (getfrag(from, |
5640f768 ED |
1503 | page_address(pfrag->page) + pfrag->offset, |
1504 | offset, copy, skb->len, skb) < 0) | |
1505 | goto error_efault; | |
1506 | ||
1507 | pfrag->offset += copy; | |
1508 | skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy); | |
1da177e4 LT |
1509 | skb->len += copy; |
1510 | skb->data_len += copy; | |
f945fa7a | 1511 | skb->truesize += copy; |
14afee4b | 1512 | refcount_add(copy, &sk->sk_wmem_alloc); |
1da177e4 LT |
1513 | } |
1514 | offset += copy; | |
1515 | length -= copy; | |
1516 | } | |
5640f768 | 1517 | |
1da177e4 | 1518 | return 0; |
5640f768 ED |
1519 | |
1520 | error_efault: | |
1521 | err = -EFAULT; | |
1da177e4 | 1522 | error: |
bdc712b4 | 1523 | cork->length -= length; |
3bd653c8 | 1524 | IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
1525 | return err; |
1526 | } | |
0bbe84a6 VY |
1527 | |
1528 | int ip6_append_data(struct sock *sk, | |
1529 | int getfrag(void *from, char *to, int offset, int len, | |
1530 | int odd, struct sk_buff *skb), | |
26879da5 WW |
1531 | void *from, int length, int transhdrlen, |
1532 | struct ipcm6_cookie *ipc6, struct flowi6 *fl6, | |
1533 | struct rt6_info *rt, unsigned int flags, | |
c14ac945 | 1534 | const struct sockcm_cookie *sockc) |
0bbe84a6 VY |
1535 | { |
1536 | struct inet_sock *inet = inet_sk(sk); | |
1537 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1538 | int exthdrlen; | |
1539 | int err; | |
1540 | ||
1541 | if (flags&MSG_PROBE) | |
1542 | return 0; | |
1543 | if (skb_queue_empty(&sk->sk_write_queue)) { | |
1544 | /* | |
1545 | * setup for corking | |
1546 | */ | |
26879da5 WW |
1547 | err = ip6_setup_cork(sk, &inet->cork, &np->cork, |
1548 | ipc6, rt, fl6); | |
0bbe84a6 VY |
1549 | if (err) |
1550 | return err; | |
1551 | ||
26879da5 | 1552 | exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0); |
0bbe84a6 VY |
1553 | length += exthdrlen; |
1554 | transhdrlen += exthdrlen; | |
1555 | } else { | |
1556 | fl6 = &inet->cork.fl.u.ip6; | |
1557 | transhdrlen = 0; | |
1558 | } | |
1559 | ||
1560 | return __ip6_append_data(sk, fl6, &sk->sk_write_queue, &inet->cork.base, | |
1561 | &np->cork, sk_page_frag(sk), getfrag, | |
26879da5 | 1562 | from, length, transhdrlen, flags, ipc6, sockc); |
0bbe84a6 | 1563 | } |
a495f836 | 1564 | EXPORT_SYMBOL_GPL(ip6_append_data); |
1da177e4 | 1565 | |
366e41d9 VY |
1566 | static void ip6_cork_release(struct inet_cork_full *cork, |
1567 | struct inet6_cork *v6_cork) | |
bf138862 | 1568 | { |
366e41d9 VY |
1569 | if (v6_cork->opt) { |
1570 | kfree(v6_cork->opt->dst0opt); | |
1571 | kfree(v6_cork->opt->dst1opt); | |
1572 | kfree(v6_cork->opt->hopopt); | |
1573 | kfree(v6_cork->opt->srcrt); | |
1574 | kfree(v6_cork->opt); | |
1575 | v6_cork->opt = NULL; | |
0178b695 HX |
1576 | } |
1577 | ||
366e41d9 VY |
1578 | if (cork->base.dst) { |
1579 | dst_release(cork->base.dst); | |
1580 | cork->base.dst = NULL; | |
1581 | cork->base.flags &= ~IPCORK_ALLFRAG; | |
bf138862 | 1582 | } |
366e41d9 | 1583 | memset(&cork->fl, 0, sizeof(cork->fl)); |
bf138862 PE |
1584 | } |
1585 | ||
6422398c VY |
1586 | struct sk_buff *__ip6_make_skb(struct sock *sk, |
1587 | struct sk_buff_head *queue, | |
1588 | struct inet_cork_full *cork, | |
1589 | struct inet6_cork *v6_cork) | |
1da177e4 LT |
1590 | { |
1591 | struct sk_buff *skb, *tmp_skb; | |
1592 | struct sk_buff **tail_skb; | |
1593 | struct in6_addr final_dst_buf, *final_dst = &final_dst_buf; | |
1da177e4 | 1594 | struct ipv6_pinfo *np = inet6_sk(sk); |
3bd653c8 | 1595 | struct net *net = sock_net(sk); |
1da177e4 | 1596 | struct ipv6hdr *hdr; |
6422398c VY |
1597 | struct ipv6_txoptions *opt = v6_cork->opt; |
1598 | struct rt6_info *rt = (struct rt6_info *)cork->base.dst; | |
1599 | struct flowi6 *fl6 = &cork->fl.u.ip6; | |
4c9483b2 | 1600 | unsigned char proto = fl6->flowi6_proto; |
1da177e4 | 1601 | |
6422398c | 1602 | skb = __skb_dequeue(queue); |
63159f29 | 1603 | if (!skb) |
1da177e4 LT |
1604 | goto out; |
1605 | tail_skb = &(skb_shinfo(skb)->frag_list); | |
1606 | ||
1607 | /* move skb->data to ip header from ext header */ | |
d56f90a7 | 1608 | if (skb->data < skb_network_header(skb)) |
bbe735e4 | 1609 | __skb_pull(skb, skb_network_offset(skb)); |
6422398c | 1610 | while ((tmp_skb = __skb_dequeue(queue)) != NULL) { |
cfe1fc77 | 1611 | __skb_pull(tmp_skb, skb_network_header_len(skb)); |
1da177e4 LT |
1612 | *tail_skb = tmp_skb; |
1613 | tail_skb = &(tmp_skb->next); | |
1614 | skb->len += tmp_skb->len; | |
1615 | skb->data_len += tmp_skb->len; | |
1da177e4 | 1616 | skb->truesize += tmp_skb->truesize; |
1da177e4 LT |
1617 | tmp_skb->destructor = NULL; |
1618 | tmp_skb->sk = NULL; | |
1da177e4 LT |
1619 | } |
1620 | ||
28a89453 | 1621 | /* Allow local fragmentation. */ |
60ff7467 | 1622 | skb->ignore_df = ip6_sk_ignore_df(sk); |
28a89453 | 1623 | |
4e3fd7a0 | 1624 | *final_dst = fl6->daddr; |
cfe1fc77 | 1625 | __skb_pull(skb, skb_network_header_len(skb)); |
1da177e4 LT |
1626 | if (opt && opt->opt_flen) |
1627 | ipv6_push_frag_opts(skb, opt, &proto); | |
1628 | if (opt && opt->opt_nflen) | |
613fa3ca | 1629 | ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst, &fl6->saddr); |
1da177e4 | 1630 | |
e2d1bca7 ACM |
1631 | skb_push(skb, sizeof(struct ipv6hdr)); |
1632 | skb_reset_network_header(skb); | |
0660e03f | 1633 | hdr = ipv6_hdr(skb); |
1ab1457c | 1634 | |
6422398c | 1635 | ip6_flow_hdr(hdr, v6_cork->tclass, |
cb1ce2ef | 1636 | ip6_make_flowlabel(net, skb, fl6->flowlabel, |
513674b5 | 1637 | ip6_autoflowlabel(net, np), fl6)); |
6422398c | 1638 | hdr->hop_limit = v6_cork->hop_limit; |
1da177e4 | 1639 | hdr->nexthdr = proto; |
4e3fd7a0 AD |
1640 | hdr->saddr = fl6->saddr; |
1641 | hdr->daddr = *final_dst; | |
1da177e4 | 1642 | |
a2c2064f | 1643 | skb->priority = sk->sk_priority; |
4a19ec58 | 1644 | skb->mark = sk->sk_mark; |
a2c2064f | 1645 | |
d8d1f30b | 1646 | skb_dst_set(skb, dst_clone(&rt->dst)); |
edf391ff | 1647 | IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); |
14878f75 | 1648 | if (proto == IPPROTO_ICMPV6) { |
adf30907 | 1649 | struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb)); |
14878f75 | 1650 | |
43a43b60 HFS |
1651 | ICMP6MSGOUT_INC_STATS(net, idev, icmp6_hdr(skb)->icmp6_type); |
1652 | ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS); | |
14878f75 DS |
1653 | } |
1654 | ||
6422398c VY |
1655 | ip6_cork_release(cork, v6_cork); |
1656 | out: | |
1657 | return skb; | |
1658 | } | |
1659 | ||
1660 | int ip6_send_skb(struct sk_buff *skb) | |
1661 | { | |
1662 | struct net *net = sock_net(skb->sk); | |
1663 | struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); | |
1664 | int err; | |
1665 | ||
33224b16 | 1666 | err = ip6_local_out(net, skb->sk, skb); |
1da177e4 LT |
1667 | if (err) { |
1668 | if (err > 0) | |
6ce9e7b5 | 1669 | err = net_xmit_errno(err); |
1da177e4 | 1670 | if (err) |
6422398c VY |
1671 | IP6_INC_STATS(net, rt->rt6i_idev, |
1672 | IPSTATS_MIB_OUTDISCARDS); | |
1da177e4 LT |
1673 | } |
1674 | ||
1da177e4 | 1675 | return err; |
6422398c VY |
1676 | } |
1677 | ||
1678 | int ip6_push_pending_frames(struct sock *sk) | |
1679 | { | |
1680 | struct sk_buff *skb; | |
1681 | ||
1682 | skb = ip6_finish_skb(sk); | |
1683 | if (!skb) | |
1684 | return 0; | |
1685 | ||
1686 | return ip6_send_skb(skb); | |
1da177e4 | 1687 | } |
a495f836 | 1688 | EXPORT_SYMBOL_GPL(ip6_push_pending_frames); |
1da177e4 | 1689 | |
0bbe84a6 | 1690 | static void __ip6_flush_pending_frames(struct sock *sk, |
6422398c VY |
1691 | struct sk_buff_head *queue, |
1692 | struct inet_cork_full *cork, | |
1693 | struct inet6_cork *v6_cork) | |
1da177e4 | 1694 | { |
1da177e4 LT |
1695 | struct sk_buff *skb; |
1696 | ||
0bbe84a6 | 1697 | while ((skb = __skb_dequeue_tail(queue)) != NULL) { |
adf30907 ED |
1698 | if (skb_dst(skb)) |
1699 | IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)), | |
e1f52208 | 1700 | IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
1701 | kfree_skb(skb); |
1702 | } | |
1703 | ||
6422398c | 1704 | ip6_cork_release(cork, v6_cork); |
1da177e4 | 1705 | } |
0bbe84a6 VY |
1706 | |
1707 | void ip6_flush_pending_frames(struct sock *sk) | |
1708 | { | |
6422398c VY |
1709 | __ip6_flush_pending_frames(sk, &sk->sk_write_queue, |
1710 | &inet_sk(sk)->cork, &inet6_sk(sk)->cork); | |
0bbe84a6 | 1711 | } |
a495f836 | 1712 | EXPORT_SYMBOL_GPL(ip6_flush_pending_frames); |
6422398c VY |
1713 | |
1714 | struct sk_buff *ip6_make_skb(struct sock *sk, | |
1715 | int getfrag(void *from, char *to, int offset, | |
1716 | int len, int odd, struct sk_buff *skb), | |
1717 | void *from, int length, int transhdrlen, | |
26879da5 | 1718 | struct ipcm6_cookie *ipc6, struct flowi6 *fl6, |
6422398c | 1719 | struct rt6_info *rt, unsigned int flags, |
26879da5 | 1720 | const struct sockcm_cookie *sockc) |
6422398c VY |
1721 | { |
1722 | struct inet_cork_full cork; | |
1723 | struct inet6_cork v6_cork; | |
1724 | struct sk_buff_head queue; | |
26879da5 | 1725 | int exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0); |
6422398c VY |
1726 | int err; |
1727 | ||
1728 | if (flags & MSG_PROBE) | |
1729 | return NULL; | |
1730 | ||
1731 | __skb_queue_head_init(&queue); | |
1732 | ||
1733 | cork.base.flags = 0; | |
1734 | cork.base.addr = 0; | |
1735 | cork.base.opt = NULL; | |
1736 | v6_cork.opt = NULL; | |
26879da5 | 1737 | err = ip6_setup_cork(sk, &cork, &v6_cork, ipc6, rt, fl6); |
6422398c VY |
1738 | if (err) |
1739 | return ERR_PTR(err); | |
1740 | ||
26879da5 WW |
1741 | if (ipc6->dontfrag < 0) |
1742 | ipc6->dontfrag = inet6_sk(sk)->dontfrag; | |
6422398c VY |
1743 | |
1744 | err = __ip6_append_data(sk, fl6, &queue, &cork.base, &v6_cork, | |
1745 | ¤t->task_frag, getfrag, from, | |
1746 | length + exthdrlen, transhdrlen + exthdrlen, | |
26879da5 | 1747 | flags, ipc6, sockc); |
6422398c VY |
1748 | if (err) { |
1749 | __ip6_flush_pending_frames(sk, &queue, &cork, &v6_cork); | |
1750 | return ERR_PTR(err); | |
1751 | } | |
1752 | ||
1753 | return __ip6_make_skb(sk, &queue, &cork, &v6_cork); | |
1754 | } |