]>
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 LT |
69 | |
70 | skb->protocol = htons(ETH_P_IPV6); | |
71 | skb->dev = dev; | |
72 | ||
0660e03f | 73 | if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) { |
adf30907 | 74 | struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb)); |
1da177e4 | 75 | |
7026b1dd | 76 | if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(sk) && |
78126c41 | 77 | ((mroute6_socket(net, skb) && |
bd91b8bf | 78 | !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) || |
7bc570c8 YH |
79 | ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr, |
80 | &ipv6_hdr(skb)->saddr))) { | |
1da177e4 LT |
81 | struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC); |
82 | ||
83 | /* Do not check for IFF_ALLMULTI; multicast routing | |
84 | is not supported in any case. | |
85 | */ | |
86 | if (newskb) | |
b2e0b385 | 87 | NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING, |
29a26a56 | 88 | net, sk, newskb, NULL, newskb->dev, |
95603e22 | 89 | dev_loopback_xmit); |
1da177e4 | 90 | |
0660e03f | 91 | if (ipv6_hdr(skb)->hop_limit == 0) { |
78126c41 | 92 | IP6_INC_STATS(net, idev, |
3bd653c8 | 93 | IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
94 | kfree_skb(skb); |
95 | return 0; | |
96 | } | |
97 | } | |
98 | ||
78126c41 | 99 | IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUTMCAST, skb->len); |
dd408515 HFS |
100 | |
101 | if (IPV6_ADDR_MC_SCOPE(&ipv6_hdr(skb)->daddr) <= | |
102 | IPV6_ADDR_SCOPE_NODELOCAL && | |
103 | !(dev->flags & IFF_LOOPBACK)) { | |
104 | kfree_skb(skb); | |
105 | return 0; | |
106 | } | |
1da177e4 LT |
107 | } |
108 | ||
14972cbd RP |
109 | if (lwtunnel_xmit_redirect(dst->lwtstate)) { |
110 | int res = lwtunnel_xmit(skb); | |
111 | ||
112 | if (res < 0 || res == LWTUNNEL_XMIT_DONE) | |
113 | return res; | |
114 | } | |
115 | ||
6fd6ce20 | 116 | rcu_read_lock_bh(); |
2647a9b0 | 117 | nexthop = rt6_nexthop((struct rt6_info *)dst, &ipv6_hdr(skb)->daddr); |
6fd6ce20 YH |
118 | neigh = __ipv6_neigh_lookup_noref(dst->dev, nexthop); |
119 | if (unlikely(!neigh)) | |
120 | neigh = __neigh_create(&nd_tbl, nexthop, dst->dev, false); | |
121 | if (!IS_ERR(neigh)) { | |
4ff06203 | 122 | sock_confirm_neigh(skb, neigh); |
c16ec185 | 123 | ret = neigh_output(neigh, skb); |
6fd6ce20 YH |
124 | rcu_read_unlock_bh(); |
125 | return ret; | |
126 | } | |
127 | rcu_read_unlock_bh(); | |
05e3aa09 | 128 | |
78126c41 | 129 | IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES); |
9e508490 JE |
130 | kfree_skb(skb); |
131 | return -EINVAL; | |
1da177e4 LT |
132 | } |
133 | ||
0c4b51f0 | 134 | static int ip6_finish_output(struct net *net, struct sock *sk, struct sk_buff *skb) |
9e508490 | 135 | { |
33b48679 DM |
136 | int ret; |
137 | ||
138 | ret = BPF_CGROUP_RUN_PROG_INET_EGRESS(sk, skb); | |
139 | if (ret) { | |
140 | kfree_skb(skb); | |
141 | return ret; | |
142 | } | |
143 | ||
9e508490 | 144 | if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) || |
9037c357 JP |
145 | dst_allfrag(skb_dst(skb)) || |
146 | (IP6CB(skb)->frag_max_size && skb->len > IP6CB(skb)->frag_max_size)) | |
7d8c6e39 | 147 | return ip6_fragment(net, sk, skb, ip6_finish_output2); |
9e508490 | 148 | else |
7d8c6e39 | 149 | return ip6_finish_output2(net, sk, skb); |
9e508490 JE |
150 | } |
151 | ||
ede2059d | 152 | int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb) |
1da177e4 | 153 | { |
9e508490 | 154 | struct net_device *dev = skb_dst(skb)->dev; |
adf30907 | 155 | struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb)); |
be10de0a | 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 | ||
1da177e4 | 169 | /* |
1c1e9d2b ED |
170 | * xmit an sk_buff (used by TCP, SCTP and DCCP) |
171 | * Note : socket lock is not held for SYNACK packets, but might be modified | |
172 | * by calls to skb_set_owner_w() and ipv6_local_error(), | |
173 | * which are using proper atomic operations or spinlocks. | |
1da177e4 | 174 | */ |
1c1e9d2b | 175 | int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6, |
92e55f41 | 176 | __u32 mark, struct ipv6_txoptions *opt, int tclass) |
1da177e4 | 177 | { |
3bd653c8 | 178 | struct net *net = sock_net(sk); |
1c1e9d2b | 179 | const struct ipv6_pinfo *np = inet6_sk(sk); |
4c9483b2 | 180 | struct in6_addr *first_hop = &fl6->daddr; |
adf30907 | 181 | struct dst_entry *dst = skb_dst(skb); |
1da177e4 | 182 | struct ipv6hdr *hdr; |
4c9483b2 | 183 | u8 proto = fl6->flowi6_proto; |
1da177e4 | 184 | int seg_len = skb->len; |
e651f03a | 185 | int hlimit = -1; |
1da177e4 LT |
186 | u32 mtu; |
187 | ||
188 | if (opt) { | |
c2636b4d | 189 | unsigned int head_room; |
1da177e4 LT |
190 | |
191 | /* First: exthdrs may take lots of space (~8K for now) | |
192 | MAX_HEADER is not enough. | |
193 | */ | |
194 | head_room = opt->opt_nflen + opt->opt_flen; | |
195 | seg_len += head_room; | |
196 | head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev); | |
197 | ||
198 | if (skb_headroom(skb) < head_room) { | |
199 | struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room); | |
63159f29 | 200 | if (!skb2) { |
adf30907 | 201 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d YH |
202 | IPSTATS_MIB_OUTDISCARDS); |
203 | kfree_skb(skb); | |
1da177e4 LT |
204 | return -ENOBUFS; |
205 | } | |
808db80a | 206 | consume_skb(skb); |
a11d206d | 207 | skb = skb2; |
1c1e9d2b ED |
208 | /* skb_set_owner_w() changes sk->sk_wmem_alloc atomically, |
209 | * it is safe to call in our context (socket lock not held) | |
210 | */ | |
211 | skb_set_owner_w(skb, (struct sock *)sk); | |
1da177e4 LT |
212 | } |
213 | if (opt->opt_flen) | |
214 | ipv6_push_frag_opts(skb, opt, &proto); | |
215 | if (opt->opt_nflen) | |
613fa3ca DL |
216 | ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop, |
217 | &fl6->saddr); | |
1da177e4 LT |
218 | } |
219 | ||
e2d1bca7 ACM |
220 | skb_push(skb, sizeof(struct ipv6hdr)); |
221 | skb_reset_network_header(skb); | |
0660e03f | 222 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
223 | |
224 | /* | |
225 | * Fill in the IPv6 header | |
226 | */ | |
b903d324 | 227 | if (np) |
1da177e4 LT |
228 | hlimit = np->hop_limit; |
229 | if (hlimit < 0) | |
6b75d090 | 230 | hlimit = ip6_dst_hoplimit(dst); |
1da177e4 | 231 | |
cb1ce2ef | 232 | ip6_flow_hdr(hdr, tclass, ip6_make_flowlabel(net, skb, fl6->flowlabel, |
67800f9b | 233 | np->autoflowlabel, fl6)); |
41a1f8ea | 234 | |
1da177e4 LT |
235 | hdr->payload_len = htons(seg_len); |
236 | hdr->nexthdr = proto; | |
237 | hdr->hop_limit = hlimit; | |
238 | ||
4e3fd7a0 AD |
239 | hdr->saddr = fl6->saddr; |
240 | hdr->daddr = *first_hop; | |
1da177e4 | 241 | |
9c9c9ad5 | 242 | skb->protocol = htons(ETH_P_IPV6); |
a2c2064f | 243 | skb->priority = sk->sk_priority; |
92e55f41 | 244 | skb->mark = mark; |
a2c2064f | 245 | |
1da177e4 | 246 | mtu = dst_mtu(dst); |
60ff7467 | 247 | if ((skb->len <= mtu) || skb->ignore_df || skb_is_gso(skb)) { |
adf30907 | 248 | IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)), |
edf391ff | 249 | IPSTATS_MIB_OUT, skb->len); |
a8e3e1a9 DA |
250 | |
251 | /* if egress device is enslaved to an L3 master device pass the | |
252 | * skb to its handler for processing | |
253 | */ | |
254 | skb = l3mdev_ip6_out((struct sock *)sk, skb); | |
255 | if (unlikely(!skb)) | |
256 | return 0; | |
257 | ||
1c1e9d2b ED |
258 | /* hooks should never assume socket lock is held. |
259 | * we promote our socket to non const | |
260 | */ | |
29a26a56 | 261 | return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, |
1c1e9d2b | 262 | net, (struct sock *)sk, skb, NULL, dst->dev, |
13206b6b | 263 | dst_output); |
1da177e4 LT |
264 | } |
265 | ||
1da177e4 | 266 | skb->dev = dst->dev; |
1c1e9d2b ED |
267 | /* ipv6_local_error() does not require socket lock, |
268 | * we promote our socket to non const | |
269 | */ | |
270 | ipv6_local_error((struct sock *)sk, EMSGSIZE, fl6, mtu); | |
271 | ||
adf30907 | 272 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
273 | kfree_skb(skb); |
274 | return -EMSGSIZE; | |
275 | } | |
7159039a YH |
276 | EXPORT_SYMBOL(ip6_xmit); |
277 | ||
1da177e4 LT |
278 | static int ip6_call_ra_chain(struct sk_buff *skb, int sel) |
279 | { | |
280 | struct ip6_ra_chain *ra; | |
281 | struct sock *last = NULL; | |
282 | ||
283 | read_lock(&ip6_ra_lock); | |
284 | for (ra = ip6_ra_chain; ra; ra = ra->next) { | |
285 | struct sock *sk = ra->sk; | |
0bd1b59b AM |
286 | if (sk && ra->sel == sel && |
287 | (!sk->sk_bound_dev_if || | |
288 | sk->sk_bound_dev_if == skb->dev->ifindex)) { | |
1da177e4 LT |
289 | if (last) { |
290 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
291 | if (skb2) | |
292 | rawv6_rcv(last, skb2); | |
293 | } | |
294 | last = sk; | |
295 | } | |
296 | } | |
297 | ||
298 | if (last) { | |
299 | rawv6_rcv(last, skb); | |
300 | read_unlock(&ip6_ra_lock); | |
301 | return 1; | |
302 | } | |
303 | read_unlock(&ip6_ra_lock); | |
304 | return 0; | |
305 | } | |
306 | ||
e21e0b5f VN |
307 | static int ip6_forward_proxy_check(struct sk_buff *skb) |
308 | { | |
0660e03f | 309 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
e21e0b5f | 310 | u8 nexthdr = hdr->nexthdr; |
75f2811c | 311 | __be16 frag_off; |
e21e0b5f VN |
312 | int offset; |
313 | ||
314 | if (ipv6_ext_hdr(nexthdr)) { | |
75f2811c | 315 | offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off); |
e21e0b5f VN |
316 | if (offset < 0) |
317 | return 0; | |
318 | } else | |
319 | offset = sizeof(struct ipv6hdr); | |
320 | ||
321 | if (nexthdr == IPPROTO_ICMPV6) { | |
322 | struct icmp6hdr *icmp6; | |
323 | ||
d56f90a7 ACM |
324 | if (!pskb_may_pull(skb, (skb_network_header(skb) + |
325 | offset + 1 - skb->data))) | |
e21e0b5f VN |
326 | return 0; |
327 | ||
d56f90a7 | 328 | icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset); |
e21e0b5f VN |
329 | |
330 | switch (icmp6->icmp6_type) { | |
331 | case NDISC_ROUTER_SOLICITATION: | |
332 | case NDISC_ROUTER_ADVERTISEMENT: | |
333 | case NDISC_NEIGHBOUR_SOLICITATION: | |
334 | case NDISC_NEIGHBOUR_ADVERTISEMENT: | |
335 | case NDISC_REDIRECT: | |
336 | /* For reaction involving unicast neighbor discovery | |
337 | * message destined to the proxied address, pass it to | |
338 | * input function. | |
339 | */ | |
340 | return 1; | |
341 | default: | |
342 | break; | |
343 | } | |
344 | } | |
345 | ||
74553b09 VN |
346 | /* |
347 | * The proxying router can't forward traffic sent to a link-local | |
348 | * address, so signal the sender and discard the packet. This | |
349 | * behavior is clarified by the MIPv6 specification. | |
350 | */ | |
351 | if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) { | |
352 | dst_link_failure(skb); | |
353 | return -1; | |
354 | } | |
355 | ||
e21e0b5f VN |
356 | return 0; |
357 | } | |
358 | ||
0c4b51f0 EB |
359 | static inline int ip6_forward_finish(struct net *net, struct sock *sk, |
360 | struct sk_buff *skb) | |
1da177e4 | 361 | { |
13206b6b | 362 | return dst_output(net, sk, skb); |
1da177e4 LT |
363 | } |
364 | ||
0954cf9c HFS |
365 | static unsigned int ip6_dst_mtu_forward(const struct dst_entry *dst) |
366 | { | |
367 | unsigned int mtu; | |
368 | struct inet6_dev *idev; | |
369 | ||
370 | if (dst_metric_locked(dst, RTAX_MTU)) { | |
371 | mtu = dst_metric_raw(dst, RTAX_MTU); | |
372 | if (mtu) | |
373 | return mtu; | |
374 | } | |
375 | ||
376 | mtu = IPV6_MIN_MTU; | |
377 | rcu_read_lock(); | |
378 | idev = __in6_dev_get(dst->dev); | |
379 | if (idev) | |
380 | mtu = idev->cnf.mtu6; | |
381 | rcu_read_unlock(); | |
382 | ||
383 | return mtu; | |
384 | } | |
385 | ||
fe6cc55f FW |
386 | static bool ip6_pkt_too_big(const struct sk_buff *skb, unsigned int mtu) |
387 | { | |
418a3156 | 388 | if (skb->len <= mtu) |
fe6cc55f FW |
389 | return false; |
390 | ||
60ff7467 | 391 | /* ipv6 conntrack defrag sets max_frag_size + ignore_df */ |
fe6cc55f FW |
392 | if (IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu) |
393 | return true; | |
394 | ||
60ff7467 | 395 | if (skb->ignore_df) |
418a3156 FW |
396 | return false; |
397 | ||
ae7ef81e | 398 | if (skb_is_gso(skb) && skb_gso_validate_mtu(skb, mtu)) |
fe6cc55f FW |
399 | return false; |
400 | ||
401 | return true; | |
402 | } | |
403 | ||
1da177e4 LT |
404 | int ip6_forward(struct sk_buff *skb) |
405 | { | |
adf30907 | 406 | struct dst_entry *dst = skb_dst(skb); |
0660e03f | 407 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
1da177e4 | 408 | struct inet6_skb_parm *opt = IP6CB(skb); |
c346dca1 | 409 | struct net *net = dev_net(dst->dev); |
14f3ad6f | 410 | u32 mtu; |
1ab1457c | 411 | |
53b7997f | 412 | if (net->ipv6.devconf_all->forwarding == 0) |
1da177e4 LT |
413 | goto error; |
414 | ||
090f1166 LR |
415 | if (skb->pkt_type != PACKET_HOST) |
416 | goto drop; | |
417 | ||
9ef2e965 HFS |
418 | if (unlikely(skb->sk)) |
419 | goto drop; | |
420 | ||
4497b076 BH |
421 | if (skb_warn_if_lro(skb)) |
422 | goto drop; | |
423 | ||
1da177e4 | 424 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) { |
1d015503 ED |
425 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
426 | IPSTATS_MIB_INDISCARDS); | |
1da177e4 LT |
427 | goto drop; |
428 | } | |
429 | ||
35fc92a9 | 430 | skb_forward_csum(skb); |
1da177e4 LT |
431 | |
432 | /* | |
433 | * We DO NOT make any processing on | |
434 | * RA packets, pushing them to user level AS IS | |
435 | * without ane WARRANTY that application will be able | |
436 | * to interpret them. The reason is that we | |
437 | * cannot make anything clever here. | |
438 | * | |
439 | * We are not end-node, so that if packet contains | |
440 | * AH/ESP, we cannot make anything. | |
441 | * Defragmentation also would be mistake, RA packets | |
442 | * cannot be fragmented, because there is no warranty | |
443 | * that different fragments will go along one path. --ANK | |
444 | */ | |
ab4eb353 YH |
445 | if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) { |
446 | if (ip6_call_ra_chain(skb, ntohs(opt->ra))) | |
1da177e4 LT |
447 | return 0; |
448 | } | |
449 | ||
450 | /* | |
451 | * check and decrement ttl | |
452 | */ | |
453 | if (hdr->hop_limit <= 1) { | |
454 | /* Force OUTPUT device used as source address */ | |
455 | skb->dev = dst->dev; | |
3ffe533c | 456 | icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0); |
1d015503 ED |
457 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
458 | IPSTATS_MIB_INHDRERRORS); | |
1da177e4 LT |
459 | |
460 | kfree_skb(skb); | |
461 | return -ETIMEDOUT; | |
462 | } | |
463 | ||
fbea49e1 | 464 | /* XXX: idev->cnf.proxy_ndp? */ |
53b7997f | 465 | if (net->ipv6.devconf_all->proxy_ndp && |
8a3edd80 | 466 | pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) { |
74553b09 VN |
467 | int proxied = ip6_forward_proxy_check(skb); |
468 | if (proxied > 0) | |
e21e0b5f | 469 | return ip6_input(skb); |
74553b09 | 470 | else if (proxied < 0) { |
1d015503 ED |
471 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
472 | IPSTATS_MIB_INDISCARDS); | |
74553b09 VN |
473 | goto drop; |
474 | } | |
e21e0b5f VN |
475 | } |
476 | ||
1da177e4 | 477 | if (!xfrm6_route_forward(skb)) { |
1d015503 ED |
478 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
479 | IPSTATS_MIB_INDISCARDS); | |
1da177e4 LT |
480 | goto drop; |
481 | } | |
adf30907 | 482 | dst = skb_dst(skb); |
1da177e4 LT |
483 | |
484 | /* IPv6 specs say nothing about it, but it is clear that we cannot | |
485 | send redirects to source routed frames. | |
1e5dc146 | 486 | We don't send redirects to frames decapsulated from IPsec. |
1da177e4 | 487 | */ |
c45a3dfb | 488 | if (skb->dev == dst->dev && opt->srcrt == 0 && !skb_sec_path(skb)) { |
1da177e4 | 489 | struct in6_addr *target = NULL; |
fbfe95a4 | 490 | struct inet_peer *peer; |
1da177e4 | 491 | struct rt6_info *rt; |
1da177e4 LT |
492 | |
493 | /* | |
494 | * incoming and outgoing devices are the same | |
495 | * send a redirect. | |
496 | */ | |
497 | ||
498 | rt = (struct rt6_info *) dst; | |
c45a3dfb DM |
499 | if (rt->rt6i_flags & RTF_GATEWAY) |
500 | target = &rt->rt6i_gateway; | |
1da177e4 LT |
501 | else |
502 | target = &hdr->daddr; | |
503 | ||
fd0273d7 | 504 | peer = inet_getpeer_v6(net->ipv6.peers, &hdr->daddr, 1); |
92d86829 | 505 | |
1da177e4 LT |
506 | /* Limit redirects both by destination (here) |
507 | and by source (inside ndisc_send_redirect) | |
508 | */ | |
fbfe95a4 | 509 | if (inet_peer_xrlim_allow(peer, 1*HZ)) |
4991969a | 510 | ndisc_send_redirect(skb, target); |
1d861aa4 DM |
511 | if (peer) |
512 | inet_putpeer(peer); | |
5bb1ab09 DS |
513 | } else { |
514 | int addrtype = ipv6_addr_type(&hdr->saddr); | |
515 | ||
1da177e4 | 516 | /* This check is security critical. */ |
f81b2e7d YH |
517 | if (addrtype == IPV6_ADDR_ANY || |
518 | addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK)) | |
5bb1ab09 DS |
519 | goto error; |
520 | if (addrtype & IPV6_ADDR_LINKLOCAL) { | |
521 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, | |
3ffe533c | 522 | ICMPV6_NOT_NEIGHBOUR, 0); |
5bb1ab09 DS |
523 | goto error; |
524 | } | |
1da177e4 LT |
525 | } |
526 | ||
0954cf9c | 527 | mtu = ip6_dst_mtu_forward(dst); |
14f3ad6f UW |
528 | if (mtu < IPV6_MIN_MTU) |
529 | mtu = IPV6_MIN_MTU; | |
530 | ||
fe6cc55f | 531 | if (ip6_pkt_too_big(skb, mtu)) { |
1da177e4 LT |
532 | /* Again, force OUTPUT device used as source address */ |
533 | skb->dev = dst->dev; | |
14f3ad6f | 534 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); |
1d015503 ED |
535 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
536 | IPSTATS_MIB_INTOOBIGERRORS); | |
537 | __IP6_INC_STATS(net, ip6_dst_idev(dst), | |
538 | IPSTATS_MIB_FRAGFAILS); | |
1da177e4 LT |
539 | kfree_skb(skb); |
540 | return -EMSGSIZE; | |
541 | } | |
542 | ||
543 | if (skb_cow(skb, dst->dev->hard_header_len)) { | |
1d015503 ED |
544 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
545 | IPSTATS_MIB_OUTDISCARDS); | |
1da177e4 LT |
546 | goto drop; |
547 | } | |
548 | ||
0660e03f | 549 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
550 | |
551 | /* Mangling hops number delayed to point after skb COW */ | |
1ab1457c | 552 | |
1da177e4 LT |
553 | hdr->hop_limit--; |
554 | ||
1d015503 ED |
555 | __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS); |
556 | __IP6_ADD_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len); | |
29a26a56 EB |
557 | return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, |
558 | net, NULL, skb, skb->dev, dst->dev, | |
6e23ae2a | 559 | ip6_forward_finish); |
1da177e4 LT |
560 | |
561 | error: | |
1d015503 | 562 | __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS); |
1da177e4 LT |
563 | drop: |
564 | kfree_skb(skb); | |
565 | return -EINVAL; | |
566 | } | |
567 | ||
568 | static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from) | |
569 | { | |
570 | to->pkt_type = from->pkt_type; | |
571 | to->priority = from->priority; | |
572 | to->protocol = from->protocol; | |
adf30907 ED |
573 | skb_dst_drop(to); |
574 | skb_dst_set(to, dst_clone(skb_dst(from))); | |
1da177e4 | 575 | to->dev = from->dev; |
82e91ffe | 576 | to->mark = from->mark; |
1da177e4 LT |
577 | |
578 | #ifdef CONFIG_NET_SCHED | |
579 | to->tc_index = from->tc_index; | |
580 | #endif | |
e7ac05f3 | 581 | nf_copy(to, from); |
984bc16c | 582 | skb_copy_secmark(to, from); |
1da177e4 LT |
583 | } |
584 | ||
7d8c6e39 EB |
585 | int ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb, |
586 | int (*output)(struct net *, struct sock *, struct sk_buff *)) | |
1da177e4 | 587 | { |
1da177e4 | 588 | struct sk_buff *frag; |
67ba4152 | 589 | struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); |
f60e5990 | 590 | struct ipv6_pinfo *np = skb->sk && !dev_recursion_level() ? |
591 | inet6_sk(skb->sk) : NULL; | |
1da177e4 LT |
592 | struct ipv6hdr *tmp_hdr; |
593 | struct frag_hdr *fh; | |
594 | unsigned int mtu, hlen, left, len; | |
a7ae1992 | 595 | int hroom, troom; |
286c2349 | 596 | __be32 frag_id; |
67ba4152 | 597 | int ptr, offset = 0, err = 0; |
1da177e4 LT |
598 | u8 *prevhdr, nexthdr = 0; |
599 | ||
1da177e4 LT |
600 | hlen = ip6_find_1stfragopt(skb, &prevhdr); |
601 | nexthdr = *prevhdr; | |
602 | ||
628a5c56 | 603 | mtu = ip6_skb_dst_mtu(skb); |
b881ef76 JH |
604 | |
605 | /* We must not fragment if the socket is set to force MTU discovery | |
14f3ad6f | 606 | * or if the skb it not generated by a local socket. |
b881ef76 | 607 | */ |
485fca66 FW |
608 | if (unlikely(!skb->ignore_df && skb->len > mtu)) |
609 | goto fail_toobig; | |
a34a101e | 610 | |
485fca66 FW |
611 | if (IP6CB(skb)->frag_max_size) { |
612 | if (IP6CB(skb)->frag_max_size > mtu) | |
613 | goto fail_toobig; | |
614 | ||
615 | /* don't send fragments larger than what we received */ | |
616 | mtu = IP6CB(skb)->frag_max_size; | |
617 | if (mtu < IPV6_MIN_MTU) | |
618 | mtu = IPV6_MIN_MTU; | |
b881ef76 JH |
619 | } |
620 | ||
d91675f9 YH |
621 | if (np && np->frag_size < mtu) { |
622 | if (np->frag_size) | |
623 | mtu = np->frag_size; | |
624 | } | |
89bc7848 | 625 | if (mtu < hlen + sizeof(struct frag_hdr) + 8) |
b72a2b01 | 626 | goto fail_toobig; |
1e0d69a9 | 627 | mtu -= hlen + sizeof(struct frag_hdr); |
1da177e4 | 628 | |
fd0273d7 MKL |
629 | frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr, |
630 | &ipv6_hdr(skb)->saddr); | |
286c2349 | 631 | |
405c92f7 HFS |
632 | if (skb->ip_summed == CHECKSUM_PARTIAL && |
633 | (err = skb_checksum_help(skb))) | |
634 | goto fail; | |
635 | ||
1d325d21 | 636 | hroom = LL_RESERVED_SPACE(rt->dst.dev); |
21dc3301 | 637 | if (skb_has_frag_list(skb)) { |
c72d8cda | 638 | unsigned int first_len = skb_pagelen(skb); |
3d13008e | 639 | struct sk_buff *frag2; |
1da177e4 LT |
640 | |
641 | if (first_len - hlen > mtu || | |
642 | ((first_len - hlen) & 7) || | |
1d325d21 FW |
643 | skb_cloned(skb) || |
644 | skb_headroom(skb) < (hroom + sizeof(struct frag_hdr))) | |
1da177e4 LT |
645 | goto slow_path; |
646 | ||
4d9092bb | 647 | skb_walk_frags(skb, frag) { |
1da177e4 LT |
648 | /* Correct geometry. */ |
649 | if (frag->len > mtu || | |
650 | ((frag->len & 7) && frag->next) || | |
1d325d21 | 651 | skb_headroom(frag) < (hlen + hroom + sizeof(struct frag_hdr))) |
3d13008e | 652 | goto slow_path_clean; |
1da177e4 | 653 | |
1da177e4 LT |
654 | /* Partially cloned skb? */ |
655 | if (skb_shared(frag)) | |
3d13008e | 656 | goto slow_path_clean; |
2fdba6b0 HX |
657 | |
658 | BUG_ON(frag->sk); | |
659 | if (skb->sk) { | |
2fdba6b0 HX |
660 | frag->sk = skb->sk; |
661 | frag->destructor = sock_wfree; | |
2fdba6b0 | 662 | } |
3d13008e | 663 | skb->truesize -= frag->truesize; |
1da177e4 LT |
664 | } |
665 | ||
666 | err = 0; | |
667 | offset = 0; | |
1da177e4 LT |
668 | /* BUILD HEADER */ |
669 | ||
9a217a1c | 670 | *prevhdr = NEXTHDR_FRAGMENT; |
d56f90a7 | 671 | tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC); |
1da177e4 | 672 | if (!tmp_hdr) { |
adf30907 | 673 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
3bd653c8 | 674 | IPSTATS_MIB_FRAGFAILS); |
1d325d21 FW |
675 | err = -ENOMEM; |
676 | goto fail; | |
1da177e4 | 677 | } |
1d325d21 FW |
678 | frag = skb_shinfo(skb)->frag_list; |
679 | skb_frag_list_init(skb); | |
1da177e4 | 680 | |
1da177e4 | 681 | __skb_pull(skb, hlen); |
67ba4152 | 682 | fh = (struct frag_hdr *)__skb_push(skb, sizeof(struct frag_hdr)); |
e2d1bca7 ACM |
683 | __skb_push(skb, hlen); |
684 | skb_reset_network_header(skb); | |
d56f90a7 | 685 | memcpy(skb_network_header(skb), tmp_hdr, hlen); |
1da177e4 | 686 | |
1da177e4 LT |
687 | fh->nexthdr = nexthdr; |
688 | fh->reserved = 0; | |
689 | fh->frag_off = htons(IP6_MF); | |
286c2349 | 690 | fh->identification = frag_id; |
1da177e4 LT |
691 | |
692 | first_len = skb_pagelen(skb); | |
693 | skb->data_len = first_len - skb_headlen(skb); | |
694 | skb->len = first_len; | |
0660e03f ACM |
695 | ipv6_hdr(skb)->payload_len = htons(first_len - |
696 | sizeof(struct ipv6hdr)); | |
a11d206d | 697 | |
d8d1f30b | 698 | dst_hold(&rt->dst); |
1da177e4 LT |
699 | |
700 | for (;;) { | |
701 | /* Prepare header of the next frame, | |
702 | * before previous one went down. */ | |
703 | if (frag) { | |
704 | frag->ip_summed = CHECKSUM_NONE; | |
badff6d0 | 705 | skb_reset_transport_header(frag); |
67ba4152 | 706 | fh = (struct frag_hdr *)__skb_push(frag, sizeof(struct frag_hdr)); |
e2d1bca7 ACM |
707 | __skb_push(frag, hlen); |
708 | skb_reset_network_header(frag); | |
d56f90a7 ACM |
709 | memcpy(skb_network_header(frag), tmp_hdr, |
710 | hlen); | |
1da177e4 LT |
711 | offset += skb->len - hlen - sizeof(struct frag_hdr); |
712 | fh->nexthdr = nexthdr; | |
713 | fh->reserved = 0; | |
714 | fh->frag_off = htons(offset); | |
53b24b8f | 715 | if (frag->next) |
1da177e4 LT |
716 | fh->frag_off |= htons(IP6_MF); |
717 | fh->identification = frag_id; | |
0660e03f ACM |
718 | ipv6_hdr(frag)->payload_len = |
719 | htons(frag->len - | |
720 | sizeof(struct ipv6hdr)); | |
1da177e4 LT |
721 | ip6_copy_metadata(frag, skb); |
722 | } | |
1ab1457c | 723 | |
7d8c6e39 | 724 | err = output(net, sk, skb); |
67ba4152 | 725 | if (!err) |
d8d1f30b | 726 | IP6_INC_STATS(net, ip6_dst_idev(&rt->dst), |
3bd653c8 | 727 | IPSTATS_MIB_FRAGCREATES); |
dafee490 | 728 | |
1da177e4 LT |
729 | if (err || !frag) |
730 | break; | |
731 | ||
732 | skb = frag; | |
733 | frag = skb->next; | |
734 | skb->next = NULL; | |
735 | } | |
736 | ||
a51482bd | 737 | kfree(tmp_hdr); |
1da177e4 LT |
738 | |
739 | if (err == 0) { | |
d8d1f30b | 740 | IP6_INC_STATS(net, ip6_dst_idev(&rt->dst), |
3bd653c8 | 741 | IPSTATS_MIB_FRAGOKS); |
94e187c0 | 742 | ip6_rt_put(rt); |
1da177e4 LT |
743 | return 0; |
744 | } | |
745 | ||
46cfd725 | 746 | kfree_skb_list(frag); |
1da177e4 | 747 | |
d8d1f30b | 748 | IP6_INC_STATS(net, ip6_dst_idev(&rt->dst), |
3bd653c8 | 749 | IPSTATS_MIB_FRAGFAILS); |
94e187c0 | 750 | ip6_rt_put(rt); |
1da177e4 | 751 | return err; |
3d13008e ED |
752 | |
753 | slow_path_clean: | |
754 | skb_walk_frags(skb, frag2) { | |
755 | if (frag2 == frag) | |
756 | break; | |
757 | frag2->sk = NULL; | |
758 | frag2->destructor = NULL; | |
759 | skb->truesize += frag2->truesize; | |
760 | } | |
1da177e4 LT |
761 | } |
762 | ||
763 | slow_path: | |
764 | left = skb->len - hlen; /* Space per frame */ | |
765 | ptr = hlen; /* Where to start from */ | |
766 | ||
767 | /* | |
768 | * Fragment the datagram. | |
769 | */ | |
770 | ||
771 | *prevhdr = NEXTHDR_FRAGMENT; | |
a7ae1992 | 772 | troom = rt->dst.dev->needed_tailroom; |
1da177e4 LT |
773 | |
774 | /* | |
775 | * Keep copying data until we run out. | |
776 | */ | |
67ba4152 | 777 | while (left > 0) { |
1da177e4 LT |
778 | len = left; |
779 | /* IF: it doesn't fit, use 'mtu' - the data space left */ | |
780 | if (len > mtu) | |
781 | len = mtu; | |
25985edc | 782 | /* IF: we are not sending up to and including the packet end |
1da177e4 LT |
783 | then align the next start on an eight byte boundary */ |
784 | if (len < left) { | |
785 | len &= ~7; | |
786 | } | |
1da177e4 | 787 | |
cbffccc9 JP |
788 | /* Allocate buffer */ |
789 | frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) + | |
790 | hroom + troom, GFP_ATOMIC); | |
791 | if (!frag) { | |
adf30907 | 792 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 793 | IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
794 | err = -ENOMEM; |
795 | goto fail; | |
796 | } | |
797 | ||
798 | /* | |
799 | * Set up data on packet | |
800 | */ | |
801 | ||
802 | ip6_copy_metadata(frag, skb); | |
a7ae1992 | 803 | skb_reserve(frag, hroom); |
1da177e4 | 804 | skb_put(frag, len + hlen + sizeof(struct frag_hdr)); |
c1d2bbe1 | 805 | skb_reset_network_header(frag); |
badff6d0 | 806 | fh = (struct frag_hdr *)(skb_network_header(frag) + hlen); |
b0e380b1 ACM |
807 | frag->transport_header = (frag->network_header + hlen + |
808 | sizeof(struct frag_hdr)); | |
1da177e4 LT |
809 | |
810 | /* | |
811 | * Charge the memory for the fragment to any owner | |
812 | * it might possess | |
813 | */ | |
814 | if (skb->sk) | |
815 | skb_set_owner_w(frag, skb->sk); | |
816 | ||
817 | /* | |
818 | * Copy the packet header into the new buffer. | |
819 | */ | |
d626f62b | 820 | skb_copy_from_linear_data(skb, skb_network_header(frag), hlen); |
1da177e4 LT |
821 | |
822 | /* | |
823 | * Build fragment header. | |
824 | */ | |
825 | fh->nexthdr = nexthdr; | |
826 | fh->reserved = 0; | |
286c2349 | 827 | fh->identification = frag_id; |
1da177e4 LT |
828 | |
829 | /* | |
830 | * Copy a block of the IP datagram. | |
831 | */ | |
e3f0b86b HS |
832 | BUG_ON(skb_copy_bits(skb, ptr, skb_transport_header(frag), |
833 | len)); | |
1da177e4 LT |
834 | left -= len; |
835 | ||
836 | fh->frag_off = htons(offset); | |
837 | if (left > 0) | |
838 | fh->frag_off |= htons(IP6_MF); | |
0660e03f ACM |
839 | ipv6_hdr(frag)->payload_len = htons(frag->len - |
840 | sizeof(struct ipv6hdr)); | |
1da177e4 LT |
841 | |
842 | ptr += len; | |
843 | offset += len; | |
844 | ||
845 | /* | |
846 | * Put this fragment into the sending queue. | |
847 | */ | |
7d8c6e39 | 848 | err = output(net, sk, frag); |
1da177e4 LT |
849 | if (err) |
850 | goto fail; | |
dafee490 | 851 | |
adf30907 | 852 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
3bd653c8 | 853 | IPSTATS_MIB_FRAGCREATES); |
1da177e4 | 854 | } |
adf30907 | 855 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 856 | IPSTATS_MIB_FRAGOKS); |
808db80a | 857 | consume_skb(skb); |
1da177e4 LT |
858 | return err; |
859 | ||
485fca66 FW |
860 | fail_toobig: |
861 | if (skb->sk && dst_allfrag(skb_dst(skb))) | |
862 | sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK); | |
863 | ||
864 | skb->dev = skb_dst(skb)->dev; | |
865 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); | |
866 | err = -EMSGSIZE; | |
867 | ||
1da177e4 | 868 | fail: |
adf30907 | 869 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 870 | IPSTATS_MIB_FRAGFAILS); |
1ab1457c | 871 | kfree_skb(skb); |
1da177e4 LT |
872 | return err; |
873 | } | |
874 | ||
b71d1d42 ED |
875 | static inline int ip6_rt_check(const struct rt6key *rt_key, |
876 | const struct in6_addr *fl_addr, | |
877 | const struct in6_addr *addr_cache) | |
cf6b1982 | 878 | { |
a02cec21 | 879 | return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) && |
63159f29 | 880 | (!addr_cache || !ipv6_addr_equal(fl_addr, addr_cache)); |
cf6b1982 YH |
881 | } |
882 | ||
497c615a HX |
883 | static struct dst_entry *ip6_sk_dst_check(struct sock *sk, |
884 | struct dst_entry *dst, | |
b71d1d42 | 885 | const struct flowi6 *fl6) |
1da177e4 | 886 | { |
497c615a | 887 | struct ipv6_pinfo *np = inet6_sk(sk); |
a963a37d | 888 | struct rt6_info *rt; |
1da177e4 | 889 | |
497c615a HX |
890 | if (!dst) |
891 | goto out; | |
892 | ||
a963a37d ED |
893 | if (dst->ops->family != AF_INET6) { |
894 | dst_release(dst); | |
895 | return NULL; | |
896 | } | |
897 | ||
898 | rt = (struct rt6_info *)dst; | |
497c615a HX |
899 | /* Yes, checking route validity in not connected |
900 | * case is not very simple. Take into account, | |
901 | * that we do not support routing by source, TOS, | |
67ba4152 | 902 | * and MSG_DONTROUTE --ANK (980726) |
497c615a | 903 | * |
cf6b1982 YH |
904 | * 1. ip6_rt_check(): If route was host route, |
905 | * check that cached destination is current. | |
497c615a HX |
906 | * If it is network route, we still may |
907 | * check its validity using saved pointer | |
908 | * to the last used address: daddr_cache. | |
909 | * We do not want to save whole address now, | |
910 | * (because main consumer of this service | |
911 | * is tcp, which has not this problem), | |
912 | * so that the last trick works only on connected | |
913 | * sockets. | |
914 | * 2. oif also should be the same. | |
915 | */ | |
4c9483b2 | 916 | if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) || |
8e1ef0a9 | 917 | #ifdef CONFIG_IPV6_SUBTREES |
4c9483b2 | 918 | ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) || |
8e1ef0a9 | 919 | #endif |
ca254490 DA |
920 | (!(fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF) && |
921 | (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex))) { | |
497c615a HX |
922 | dst_release(dst); |
923 | dst = NULL; | |
1da177e4 LT |
924 | } |
925 | ||
497c615a HX |
926 | out: |
927 | return dst; | |
928 | } | |
929 | ||
3aef934f | 930 | static int ip6_dst_lookup_tail(struct net *net, const struct sock *sk, |
4c9483b2 | 931 | struct dst_entry **dst, struct flowi6 *fl6) |
497c615a | 932 | { |
69cce1d1 DM |
933 | #ifdef CONFIG_IPV6_OPTIMISTIC_DAD |
934 | struct neighbour *n; | |
97cac082 | 935 | struct rt6_info *rt; |
69cce1d1 DM |
936 | #endif |
937 | int err; | |
6f21c96a | 938 | int flags = 0; |
497c615a | 939 | |
e16e888b MS |
940 | /* The correct way to handle this would be to do |
941 | * ip6_route_get_saddr, and then ip6_route_output; however, | |
942 | * the route-specific preferred source forces the | |
943 | * ip6_route_output call _before_ ip6_route_get_saddr. | |
944 | * | |
945 | * In source specific routing (no src=any default route), | |
946 | * ip6_route_output will fail given src=any saddr, though, so | |
947 | * that's why we try it again later. | |
948 | */ | |
949 | if (ipv6_addr_any(&fl6->saddr) && (!*dst || !(*dst)->error)) { | |
950 | struct rt6_info *rt; | |
951 | bool had_dst = *dst != NULL; | |
1da177e4 | 952 | |
e16e888b MS |
953 | if (!had_dst) |
954 | *dst = ip6_route_output(net, sk, fl6); | |
955 | rt = (*dst)->error ? NULL : (struct rt6_info *)*dst; | |
c3968a85 DW |
956 | err = ip6_route_get_saddr(net, rt, &fl6->daddr, |
957 | sk ? inet6_sk(sk)->srcprefs : 0, | |
958 | &fl6->saddr); | |
44456d37 | 959 | if (err) |
1da177e4 | 960 | goto out_err_release; |
e16e888b MS |
961 | |
962 | /* If we had an erroneous initial result, pretend it | |
963 | * never existed and let the SA-enabled version take | |
964 | * over. | |
965 | */ | |
966 | if (!had_dst && (*dst)->error) { | |
967 | dst_release(*dst); | |
968 | *dst = NULL; | |
969 | } | |
6f21c96a PA |
970 | |
971 | if (fl6->flowi6_oif) | |
972 | flags |= RT6_LOOKUP_F_IFACE; | |
1da177e4 LT |
973 | } |
974 | ||
e16e888b | 975 | if (!*dst) |
6f21c96a | 976 | *dst = ip6_route_output_flags(net, sk, fl6, flags); |
e16e888b MS |
977 | |
978 | err = (*dst)->error; | |
979 | if (err) | |
980 | goto out_err_release; | |
981 | ||
95c385b4 | 982 | #ifdef CONFIG_IPV6_OPTIMISTIC_DAD |
e550dfb0 NH |
983 | /* |
984 | * Here if the dst entry we've looked up | |
985 | * has a neighbour entry that is in the INCOMPLETE | |
986 | * state and the src address from the flow is | |
987 | * marked as OPTIMISTIC, we release the found | |
988 | * dst entry and replace it instead with the | |
989 | * dst entry of the nexthop router | |
990 | */ | |
c56bf6fe | 991 | rt = (struct rt6_info *) *dst; |
707be1ff | 992 | rcu_read_lock_bh(); |
2647a9b0 MKL |
993 | n = __ipv6_neigh_lookup_noref(rt->dst.dev, |
994 | rt6_nexthop(rt, &fl6->daddr)); | |
707be1ff YH |
995 | err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0; |
996 | rcu_read_unlock_bh(); | |
997 | ||
998 | if (err) { | |
e550dfb0 | 999 | struct inet6_ifaddr *ifp; |
4c9483b2 | 1000 | struct flowi6 fl_gw6; |
e550dfb0 NH |
1001 | int redirect; |
1002 | ||
4c9483b2 | 1003 | ifp = ipv6_get_ifaddr(net, &fl6->saddr, |
e550dfb0 NH |
1004 | (*dst)->dev, 1); |
1005 | ||
1006 | redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC); | |
1007 | if (ifp) | |
1008 | in6_ifa_put(ifp); | |
1009 | ||
1010 | if (redirect) { | |
1011 | /* | |
1012 | * We need to get the dst entry for the | |
1013 | * default router instead | |
1014 | */ | |
1015 | dst_release(*dst); | |
4c9483b2 DM |
1016 | memcpy(&fl_gw6, fl6, sizeof(struct flowi6)); |
1017 | memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr)); | |
1018 | *dst = ip6_route_output(net, sk, &fl_gw6); | |
e5d08d71 IM |
1019 | err = (*dst)->error; |
1020 | if (err) | |
e550dfb0 | 1021 | goto out_err_release; |
95c385b4 | 1022 | } |
e550dfb0 | 1023 | } |
95c385b4 | 1024 | #endif |
ec5e3b0a | 1025 | if (ipv6_addr_v4mapped(&fl6->saddr) && |
00ea1cee WB |
1026 | !(ipv6_addr_v4mapped(&fl6->daddr) || ipv6_addr_any(&fl6->daddr))) { |
1027 | err = -EAFNOSUPPORT; | |
1028 | goto out_err_release; | |
1029 | } | |
95c385b4 | 1030 | |
1da177e4 LT |
1031 | return 0; |
1032 | ||
1033 | out_err_release: | |
1034 | dst_release(*dst); | |
1035 | *dst = NULL; | |
8a966fc0 | 1036 | |
0d240e78 DA |
1037 | if (err == -ENETUNREACH) |
1038 | IP6_INC_STATS(net, NULL, IPSTATS_MIB_OUTNOROUTES); | |
1da177e4 LT |
1039 | return err; |
1040 | } | |
34a0b3cd | 1041 | |
497c615a HX |
1042 | /** |
1043 | * ip6_dst_lookup - perform route lookup on flow | |
1044 | * @sk: socket which provides route info | |
1045 | * @dst: pointer to dst_entry * for result | |
4c9483b2 | 1046 | * @fl6: flow to lookup |
497c615a HX |
1047 | * |
1048 | * This function performs a route lookup on the given flow. | |
1049 | * | |
1050 | * It returns zero on success, or a standard errno code on error. | |
1051 | */ | |
343d60aa RP |
1052 | int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst, |
1053 | struct flowi6 *fl6) | |
497c615a HX |
1054 | { |
1055 | *dst = NULL; | |
343d60aa | 1056 | return ip6_dst_lookup_tail(net, sk, dst, fl6); |
497c615a | 1057 | } |
3cf3dc6c ACM |
1058 | EXPORT_SYMBOL_GPL(ip6_dst_lookup); |
1059 | ||
497c615a | 1060 | /** |
68d0c6d3 DM |
1061 | * ip6_dst_lookup_flow - perform route lookup on flow with ipsec |
1062 | * @sk: socket which provides route info | |
4c9483b2 | 1063 | * @fl6: flow to lookup |
68d0c6d3 | 1064 | * @final_dst: final destination address for ipsec lookup |
68d0c6d3 DM |
1065 | * |
1066 | * This function performs a route lookup on the given flow. | |
1067 | * | |
1068 | * It returns a valid dst pointer on success, or a pointer encoded | |
1069 | * error code. | |
1070 | */ | |
3aef934f | 1071 | struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6, |
0e0d44ab | 1072 | const struct in6_addr *final_dst) |
68d0c6d3 DM |
1073 | { |
1074 | struct dst_entry *dst = NULL; | |
1075 | int err; | |
1076 | ||
343d60aa | 1077 | err = ip6_dst_lookup_tail(sock_net(sk), sk, &dst, fl6); |
68d0c6d3 DM |
1078 | if (err) |
1079 | return ERR_PTR(err); | |
1080 | if (final_dst) | |
4e3fd7a0 | 1081 | fl6->daddr = *final_dst; |
2774c131 | 1082 | |
f92ee619 | 1083 | return xfrm_lookup_route(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0); |
68d0c6d3 DM |
1084 | } |
1085 | EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow); | |
1086 | ||
1087 | /** | |
1088 | * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow | |
497c615a | 1089 | * @sk: socket which provides the dst cache and route info |
4c9483b2 | 1090 | * @fl6: flow to lookup |
68d0c6d3 | 1091 | * @final_dst: final destination address for ipsec lookup |
497c615a HX |
1092 | * |
1093 | * This function performs a route lookup on the given flow with the | |
1094 | * possibility of using the cached route in the socket if it is valid. | |
1095 | * It will take the socket dst lock when operating on the dst cache. | |
1096 | * As a result, this function can only be used in process context. | |
1097 | * | |
68d0c6d3 DM |
1098 | * It returns a valid dst pointer on success, or a pointer encoded |
1099 | * error code. | |
497c615a | 1100 | */ |
4c9483b2 | 1101 | struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6, |
0e0d44ab | 1102 | const struct in6_addr *final_dst) |
497c615a | 1103 | { |
68d0c6d3 | 1104 | struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie); |
497c615a | 1105 | |
4c9483b2 | 1106 | dst = ip6_sk_dst_check(sk, dst, fl6); |
00bc0ef5 JS |
1107 | if (!dst) |
1108 | dst = ip6_dst_lookup_flow(sk, fl6, final_dst); | |
68d0c6d3 | 1109 | |
00bc0ef5 | 1110 | return dst; |
497c615a | 1111 | } |
68d0c6d3 | 1112 | EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow); |
497c615a | 1113 | |
34a0b3cd | 1114 | static inline int ip6_ufo_append_data(struct sock *sk, |
0bbe84a6 | 1115 | struct sk_buff_head *queue, |
e89e9cf5 AR |
1116 | int getfrag(void *from, char *to, int offset, int len, |
1117 | int odd, struct sk_buff *skb), | |
1118 | void *from, int length, int hh_len, int fragheaderlen, | |
3ba3458f JS |
1119 | int exthdrlen, int transhdrlen, int mtu, |
1120 | unsigned int flags, const struct flowi6 *fl6) | |
e89e9cf5 AR |
1121 | |
1122 | { | |
1123 | struct sk_buff *skb; | |
1124 | int err; | |
1125 | ||
1126 | /* There is support for UDP large send offload by network | |
1127 | * device, so create one single skb packet containing complete | |
1128 | * udp datagram | |
1129 | */ | |
0bbe84a6 | 1130 | skb = skb_peek_tail(queue); |
63159f29 | 1131 | if (!skb) { |
e89e9cf5 AR |
1132 | skb = sock_alloc_send_skb(sk, |
1133 | hh_len + fragheaderlen + transhdrlen + 20, | |
1134 | (flags & MSG_DONTWAIT), &err); | |
63159f29 | 1135 | if (!skb) |
504744e4 | 1136 | return err; |
e89e9cf5 AR |
1137 | |
1138 | /* reserve space for Hardware header */ | |
1139 | skb_reserve(skb, hh_len); | |
1140 | ||
1141 | /* create space for UDP/IP header */ | |
67ba4152 | 1142 | skb_put(skb, fragheaderlen + transhdrlen); |
e89e9cf5 AR |
1143 | |
1144 | /* initialize network header pointer */ | |
3ba3458f | 1145 | skb_set_network_header(skb, exthdrlen); |
e89e9cf5 AR |
1146 | |
1147 | /* initialize protocol header pointer */ | |
b0e380b1 | 1148 | skb->transport_header = skb->network_header + fragheaderlen; |
e89e9cf5 | 1149 | |
9c9c9ad5 | 1150 | skb->protocol = htons(ETH_P_IPV6); |
e89e9cf5 | 1151 | skb->csum = 0; |
e89e9cf5 | 1152 | |
0dec879f JA |
1153 | if (flags & MSG_CONFIRM) |
1154 | skb_set_dst_pending_confirm(skb, 1); | |
1155 | ||
0bbe84a6 | 1156 | __skb_queue_tail(queue, skb); |
c547dbf5 JP |
1157 | } else if (skb_is_gso(skb)) { |
1158 | goto append; | |
e89e9cf5 | 1159 | } |
e89e9cf5 | 1160 | |
c547dbf5 JP |
1161 | skb->ip_summed = CHECKSUM_PARTIAL; |
1162 | /* Specify the length of each IPv6 datagram fragment. | |
1163 | * It has to be a multiple of 8. | |
1164 | */ | |
1165 | skb_shinfo(skb)->gso_size = (mtu - fragheaderlen - | |
1166 | sizeof(struct frag_hdr)) & ~7; | |
1167 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP; | |
fd0273d7 MKL |
1168 | skb_shinfo(skb)->ip6_frag_id = ipv6_select_ident(sock_net(sk), |
1169 | &fl6->daddr, | |
1170 | &fl6->saddr); | |
c547dbf5 JP |
1171 | |
1172 | append: | |
2811ebac HFS |
1173 | return skb_append_datato_frags(sk, skb, getfrag, from, |
1174 | (length - transhdrlen)); | |
e89e9cf5 | 1175 | } |
1da177e4 | 1176 | |
0178b695 HX |
1177 | static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src, |
1178 | gfp_t gfp) | |
1179 | { | |
1180 | return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL; | |
1181 | } | |
1182 | ||
1183 | static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src, | |
1184 | gfp_t gfp) | |
1185 | { | |
1186 | return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL; | |
1187 | } | |
1188 | ||
75a493e6 | 1189 | static void ip6_append_data_mtu(unsigned int *mtu, |
0c183379 G |
1190 | int *maxfraglen, |
1191 | unsigned int fragheaderlen, | |
1192 | struct sk_buff *skb, | |
75a493e6 | 1193 | struct rt6_info *rt, |
e367c2d0 | 1194 | unsigned int orig_mtu) |
0c183379 G |
1195 | { |
1196 | if (!(rt->dst.flags & DST_XFRM_TUNNEL)) { | |
63159f29 | 1197 | if (!skb) { |
0c183379 | 1198 | /* first fragment, reserve header_len */ |
e367c2d0 | 1199 | *mtu = orig_mtu - rt->dst.header_len; |
0c183379 G |
1200 | |
1201 | } else { | |
1202 | /* | |
1203 | * this fragment is not first, the headers | |
1204 | * space is regarded as data space. | |
1205 | */ | |
e367c2d0 | 1206 | *mtu = orig_mtu; |
0c183379 G |
1207 | } |
1208 | *maxfraglen = ((*mtu - fragheaderlen) & ~7) | |
1209 | + fragheaderlen - sizeof(struct frag_hdr); | |
1210 | } | |
1211 | } | |
1212 | ||
366e41d9 | 1213 | static int ip6_setup_cork(struct sock *sk, struct inet_cork_full *cork, |
26879da5 | 1214 | struct inet6_cork *v6_cork, struct ipcm6_cookie *ipc6, |
366e41d9 VY |
1215 | struct rt6_info *rt, struct flowi6 *fl6) |
1216 | { | |
1217 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1218 | unsigned int mtu; | |
26879da5 | 1219 | struct ipv6_txoptions *opt = ipc6->opt; |
366e41d9 VY |
1220 | |
1221 | /* | |
1222 | * setup for corking | |
1223 | */ | |
1224 | if (opt) { | |
1225 | if (WARN_ON(v6_cork->opt)) | |
1226 | return -EINVAL; | |
1227 | ||
1228 | v6_cork->opt = kzalloc(opt->tot_len, sk->sk_allocation); | |
63159f29 | 1229 | if (unlikely(!v6_cork->opt)) |
366e41d9 VY |
1230 | return -ENOBUFS; |
1231 | ||
1232 | v6_cork->opt->tot_len = opt->tot_len; | |
1233 | v6_cork->opt->opt_flen = opt->opt_flen; | |
1234 | v6_cork->opt->opt_nflen = opt->opt_nflen; | |
1235 | ||
1236 | v6_cork->opt->dst0opt = ip6_opt_dup(opt->dst0opt, | |
1237 | sk->sk_allocation); | |
1238 | if (opt->dst0opt && !v6_cork->opt->dst0opt) | |
1239 | return -ENOBUFS; | |
1240 | ||
1241 | v6_cork->opt->dst1opt = ip6_opt_dup(opt->dst1opt, | |
1242 | sk->sk_allocation); | |
1243 | if (opt->dst1opt && !v6_cork->opt->dst1opt) | |
1244 | return -ENOBUFS; | |
1245 | ||
1246 | v6_cork->opt->hopopt = ip6_opt_dup(opt->hopopt, | |
1247 | sk->sk_allocation); | |
1248 | if (opt->hopopt && !v6_cork->opt->hopopt) | |
1249 | return -ENOBUFS; | |
1250 | ||
1251 | v6_cork->opt->srcrt = ip6_rthdr_dup(opt->srcrt, | |
1252 | sk->sk_allocation); | |
1253 | if (opt->srcrt && !v6_cork->opt->srcrt) | |
1254 | return -ENOBUFS; | |
1255 | ||
1256 | /* need source address above miyazawa*/ | |
1257 | } | |
1258 | dst_hold(&rt->dst); | |
1259 | cork->base.dst = &rt->dst; | |
1260 | cork->fl.u.ip6 = *fl6; | |
26879da5 WW |
1261 | v6_cork->hop_limit = ipc6->hlimit; |
1262 | v6_cork->tclass = ipc6->tclass; | |
366e41d9 VY |
1263 | if (rt->dst.flags & DST_XFRM_TUNNEL) |
1264 | mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ? | |
1265 | rt->dst.dev->mtu : dst_mtu(&rt->dst); | |
1266 | else | |
1267 | mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ? | |
1268 | rt->dst.dev->mtu : dst_mtu(rt->dst.path); | |
1269 | if (np->frag_size < mtu) { | |
1270 | if (np->frag_size) | |
1271 | mtu = np->frag_size; | |
1272 | } | |
1273 | cork->base.fragsize = mtu; | |
1274 | if (dst_allfrag(rt->dst.path)) | |
1275 | cork->base.flags |= IPCORK_ALLFRAG; | |
1276 | cork->base.length = 0; | |
1277 | ||
1278 | return 0; | |
1279 | } | |
1280 | ||
0bbe84a6 VY |
1281 | static int __ip6_append_data(struct sock *sk, |
1282 | struct flowi6 *fl6, | |
1283 | struct sk_buff_head *queue, | |
1284 | struct inet_cork *cork, | |
1285 | struct inet6_cork *v6_cork, | |
1286 | struct page_frag *pfrag, | |
1287 | int getfrag(void *from, char *to, int offset, | |
1288 | int len, int odd, struct sk_buff *skb), | |
1289 | void *from, int length, int transhdrlen, | |
26879da5 | 1290 | unsigned int flags, struct ipcm6_cookie *ipc6, |
c14ac945 | 1291 | const struct sockcm_cookie *sockc) |
1da177e4 | 1292 | { |
0c183379 | 1293 | struct sk_buff *skb, *skb_prev = NULL; |
e367c2d0 | 1294 | unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu; |
0bbe84a6 VY |
1295 | int exthdrlen = 0; |
1296 | int dst_exthdrlen = 0; | |
1da177e4 | 1297 | int hh_len; |
1da177e4 LT |
1298 | int copy; |
1299 | int err; | |
1300 | int offset = 0; | |
a693e698 | 1301 | __u8 tx_flags = 0; |
09c2d251 | 1302 | u32 tskey = 0; |
0bbe84a6 VY |
1303 | struct rt6_info *rt = (struct rt6_info *)cork->dst; |
1304 | struct ipv6_txoptions *opt = v6_cork->opt; | |
32dce968 | 1305 | int csummode = CHECKSUM_NONE; |
682b1a9d | 1306 | unsigned int maxnonfragsize, headersize; |
1da177e4 | 1307 | |
0bbe84a6 VY |
1308 | skb = skb_peek_tail(queue); |
1309 | if (!skb) { | |
1310 | exthdrlen = opt ? opt->opt_flen : 0; | |
7efdba5b | 1311 | dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len; |
1da177e4 | 1312 | } |
0bbe84a6 | 1313 | |
366e41d9 | 1314 | mtu = cork->fragsize; |
e367c2d0 | 1315 | orig_mtu = mtu; |
1da177e4 | 1316 | |
d8d1f30b | 1317 | hh_len = LL_RESERVED_SPACE(rt->dst.dev); |
1da177e4 | 1318 | |
a1b05140 | 1319 | fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len + |
b4ce9277 | 1320 | (opt ? opt->opt_nflen : 0); |
4df98e76 HFS |
1321 | maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - |
1322 | sizeof(struct frag_hdr); | |
1da177e4 | 1323 | |
682b1a9d HFS |
1324 | headersize = sizeof(struct ipv6hdr) + |
1325 | (opt ? opt->opt_flen + opt->opt_nflen : 0) + | |
1326 | (dst_allfrag(&rt->dst) ? | |
1327 | sizeof(struct frag_hdr) : 0) + | |
1328 | rt->rt6i_nfheader_len; | |
1329 | ||
26879da5 | 1330 | if (cork->length + length > mtu - headersize && ipc6->dontfrag && |
682b1a9d HFS |
1331 | (sk->sk_protocol == IPPROTO_UDP || |
1332 | sk->sk_protocol == IPPROTO_RAW)) { | |
1333 | ipv6_local_rxpmtu(sk, fl6, mtu - headersize + | |
1334 | sizeof(struct ipv6hdr)); | |
1335 | goto emsgsize; | |
1336 | } | |
4df98e76 | 1337 | |
682b1a9d HFS |
1338 | if (ip6_sk_ignore_df(sk)) |
1339 | maxnonfragsize = sizeof(struct ipv6hdr) + IPV6_MAXPLEN; | |
1340 | else | |
1341 | maxnonfragsize = mtu; | |
4df98e76 | 1342 | |
682b1a9d | 1343 | if (cork->length + length > maxnonfragsize - headersize) { |
4df98e76 | 1344 | emsgsize: |
682b1a9d HFS |
1345 | ipv6_local_error(sk, EMSGSIZE, fl6, |
1346 | mtu - headersize + | |
1347 | sizeof(struct ipv6hdr)); | |
1348 | return -EMSGSIZE; | |
1da177e4 LT |
1349 | } |
1350 | ||
682b1a9d HFS |
1351 | /* CHECKSUM_PARTIAL only with no extension headers and when |
1352 | * we are not going to fragment | |
1353 | */ | |
1354 | if (transhdrlen && sk->sk_protocol == IPPROTO_UDP && | |
1355 | headersize == sizeof(struct ipv6hdr) && | |
2b89ed65 | 1356 | length <= mtu - headersize && |
682b1a9d | 1357 | !(flags & MSG_MORE) && |
c8cd0989 | 1358 | rt->dst.dev->features & (NETIF_F_IPV6_CSUM | NETIF_F_HW_CSUM)) |
682b1a9d HFS |
1359 | csummode = CHECKSUM_PARTIAL; |
1360 | ||
09c2d251 | 1361 | if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_RAW) { |
c14ac945 | 1362 | sock_tx_timestamp(sk, sockc->tsflags, &tx_flags); |
09c2d251 WB |
1363 | if (tx_flags & SKBTX_ANY_SW_TSTAMP && |
1364 | sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID) | |
1365 | tskey = sk->sk_tskey++; | |
1366 | } | |
a693e698 | 1367 | |
1da177e4 LT |
1368 | /* |
1369 | * Let's try using as much space as possible. | |
1370 | * Use MTU if total length of the message fits into the MTU. | |
1371 | * Otherwise, we need to reserve fragment header and | |
1372 | * fragment alignment (= 8-15 octects, in total). | |
1373 | * | |
1374 | * Note that we may need to "move" the data from the tail of | |
1ab1457c | 1375 | * of the buffer to the new fragment when we split |
1da177e4 LT |
1376 | * the message. |
1377 | * | |
1ab1457c | 1378 | * FIXME: It may be fragmented into multiple chunks |
1da177e4 LT |
1379 | * at once if non-fragmentable extension headers |
1380 | * are too large. | |
1ab1457c | 1381 | * --yoshfuji |
1da177e4 LT |
1382 | */ |
1383 | ||
2811ebac | 1384 | cork->length += length; |
e4c5e13a | 1385 | if ((((length + fragheaderlen) > mtu) || |
2811ebac HFS |
1386 | (skb && skb_is_gso(skb))) && |
1387 | (sk->sk_protocol == IPPROTO_UDP) && | |
4b3b45ed | 1388 | (rt->dst.dev->features & NETIF_F_UFO) && !dst_xfrm(&rt->dst) && |
40ba3302 | 1389 | (sk->sk_type == SOCK_DGRAM) && !udp_get_no_check6_tx(sk)) { |
0bbe84a6 | 1390 | err = ip6_ufo_append_data(sk, queue, getfrag, from, length, |
3ba3458f | 1391 | hh_len, fragheaderlen, exthdrlen, |
fd0273d7 | 1392 | transhdrlen, mtu, flags, fl6); |
2811ebac HFS |
1393 | if (err) |
1394 | goto error; | |
1395 | return 0; | |
e89e9cf5 | 1396 | } |
1da177e4 | 1397 | |
2811ebac | 1398 | if (!skb) |
1da177e4 LT |
1399 | goto alloc_new_skb; |
1400 | ||
1401 | while (length > 0) { | |
1402 | /* Check if the remaining data fits into current packet. */ | |
bdc712b4 | 1403 | copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len; |
1da177e4 LT |
1404 | if (copy < length) |
1405 | copy = maxfraglen - skb->len; | |
1406 | ||
1407 | if (copy <= 0) { | |
1408 | char *data; | |
1409 | unsigned int datalen; | |
1410 | unsigned int fraglen; | |
1411 | unsigned int fraggap; | |
1412 | unsigned int alloclen; | |
1da177e4 | 1413 | alloc_new_skb: |
1da177e4 | 1414 | /* There's no room in the current skb */ |
0c183379 G |
1415 | if (skb) |
1416 | fraggap = skb->len - maxfraglen; | |
1da177e4 LT |
1417 | else |
1418 | fraggap = 0; | |
0c183379 | 1419 | /* update mtu and maxfraglen if necessary */ |
63159f29 | 1420 | if (!skb || !skb_prev) |
0c183379 | 1421 | ip6_append_data_mtu(&mtu, &maxfraglen, |
75a493e6 | 1422 | fragheaderlen, skb, rt, |
e367c2d0 | 1423 | orig_mtu); |
0c183379 G |
1424 | |
1425 | skb_prev = skb; | |
1da177e4 LT |
1426 | |
1427 | /* | |
1428 | * If remaining data exceeds the mtu, | |
1429 | * we know we need more fragment(s). | |
1430 | */ | |
1431 | datalen = length + fraggap; | |
1da177e4 | 1432 | |
0c183379 G |
1433 | if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen) |
1434 | datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len; | |
1da177e4 | 1435 | if ((flags & MSG_MORE) && |
d8d1f30b | 1436 | !(rt->dst.dev->features&NETIF_F_SG)) |
1da177e4 LT |
1437 | alloclen = mtu; |
1438 | else | |
1439 | alloclen = datalen + fragheaderlen; | |
1440 | ||
299b0767 SK |
1441 | alloclen += dst_exthdrlen; |
1442 | ||
0c183379 G |
1443 | if (datalen != length + fraggap) { |
1444 | /* | |
1445 | * this is not the last fragment, the trailer | |
1446 | * space is regarded as data space. | |
1447 | */ | |
1448 | datalen += rt->dst.trailer_len; | |
1449 | } | |
1450 | ||
1451 | alloclen += rt->dst.trailer_len; | |
1452 | fraglen = datalen + fragheaderlen; | |
1da177e4 LT |
1453 | |
1454 | /* | |
1455 | * We just reserve space for fragment header. | |
1ab1457c | 1456 | * Note: this may be overallocation if the message |
1da177e4 LT |
1457 | * (without MSG_MORE) fits into the MTU. |
1458 | */ | |
1459 | alloclen += sizeof(struct frag_hdr); | |
1460 | ||
1461 | if (transhdrlen) { | |
1462 | skb = sock_alloc_send_skb(sk, | |
1463 | alloclen + hh_len, | |
1464 | (flags & MSG_DONTWAIT), &err); | |
1465 | } else { | |
1466 | skb = NULL; | |
1467 | if (atomic_read(&sk->sk_wmem_alloc) <= | |
1468 | 2 * sk->sk_sndbuf) | |
1469 | skb = sock_wmalloc(sk, | |
1470 | alloclen + hh_len, 1, | |
1471 | sk->sk_allocation); | |
63159f29 | 1472 | if (unlikely(!skb)) |
1da177e4 LT |
1473 | err = -ENOBUFS; |
1474 | } | |
63159f29 | 1475 | if (!skb) |
1da177e4 LT |
1476 | goto error; |
1477 | /* | |
1478 | * Fill in the control structures | |
1479 | */ | |
9c9c9ad5 | 1480 | skb->protocol = htons(ETH_P_IPV6); |
32dce968 | 1481 | skb->ip_summed = csummode; |
1da177e4 | 1482 | skb->csum = 0; |
1f85851e G |
1483 | /* reserve for fragmentation and ipsec header */ |
1484 | skb_reserve(skb, hh_len + sizeof(struct frag_hdr) + | |
1485 | dst_exthdrlen); | |
1da177e4 | 1486 | |
11878b40 WB |
1487 | /* Only the initial fragment is time stamped */ |
1488 | skb_shinfo(skb)->tx_flags = tx_flags; | |
1489 | tx_flags = 0; | |
09c2d251 WB |
1490 | skb_shinfo(skb)->tskey = tskey; |
1491 | tskey = 0; | |
a693e698 | 1492 | |
1da177e4 LT |
1493 | /* |
1494 | * Find where to start putting bytes | |
1495 | */ | |
1f85851e G |
1496 | data = skb_put(skb, fraglen); |
1497 | skb_set_network_header(skb, exthdrlen); | |
1498 | data += fragheaderlen; | |
b0e380b1 ACM |
1499 | skb->transport_header = (skb->network_header + |
1500 | fragheaderlen); | |
1da177e4 LT |
1501 | if (fraggap) { |
1502 | skb->csum = skb_copy_and_csum_bits( | |
1503 | skb_prev, maxfraglen, | |
1504 | data + transhdrlen, fraggap, 0); | |
1505 | skb_prev->csum = csum_sub(skb_prev->csum, | |
1506 | skb->csum); | |
1507 | data += fraggap; | |
e9fa4f7b | 1508 | pskb_trim_unique(skb_prev, maxfraglen); |
1da177e4 LT |
1509 | } |
1510 | copy = datalen - transhdrlen - fraggap; | |
299b0767 | 1511 | |
1da177e4 LT |
1512 | if (copy < 0) { |
1513 | err = -EINVAL; | |
1514 | kfree_skb(skb); | |
1515 | goto error; | |
1516 | } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) { | |
1517 | err = -EFAULT; | |
1518 | kfree_skb(skb); | |
1519 | goto error; | |
1520 | } | |
1521 | ||
1522 | offset += copy; | |
1523 | length -= datalen - fraggap; | |
1524 | transhdrlen = 0; | |
1525 | exthdrlen = 0; | |
299b0767 | 1526 | dst_exthdrlen = 0; |
1da177e4 | 1527 | |
0dec879f JA |
1528 | if ((flags & MSG_CONFIRM) && !skb_prev) |
1529 | skb_set_dst_pending_confirm(skb, 1); | |
1530 | ||
1da177e4 LT |
1531 | /* |
1532 | * Put the packet on the pending queue | |
1533 | */ | |
0bbe84a6 | 1534 | __skb_queue_tail(queue, skb); |
1da177e4 LT |
1535 | continue; |
1536 | } | |
1537 | ||
1538 | if (copy > length) | |
1539 | copy = length; | |
1540 | ||
d8d1f30b | 1541 | if (!(rt->dst.dev->features&NETIF_F_SG)) { |
1da177e4 LT |
1542 | unsigned int off; |
1543 | ||
1544 | off = skb->len; | |
1545 | if (getfrag(from, skb_put(skb, copy), | |
1546 | offset, copy, off, skb) < 0) { | |
1547 | __skb_trim(skb, off); | |
1548 | err = -EFAULT; | |
1549 | goto error; | |
1550 | } | |
1551 | } else { | |
1552 | int i = skb_shinfo(skb)->nr_frags; | |
1da177e4 | 1553 | |
5640f768 ED |
1554 | err = -ENOMEM; |
1555 | if (!sk_page_frag_refill(sk, pfrag)) | |
1da177e4 | 1556 | goto error; |
5640f768 ED |
1557 | |
1558 | if (!skb_can_coalesce(skb, i, pfrag->page, | |
1559 | pfrag->offset)) { | |
1560 | err = -EMSGSIZE; | |
1561 | if (i == MAX_SKB_FRAGS) | |
1562 | goto error; | |
1563 | ||
1564 | __skb_fill_page_desc(skb, i, pfrag->page, | |
1565 | pfrag->offset, 0); | |
1566 | skb_shinfo(skb)->nr_frags = ++i; | |
1567 | get_page(pfrag->page); | |
1da177e4 | 1568 | } |
5640f768 | 1569 | copy = min_t(int, copy, pfrag->size - pfrag->offset); |
9e903e08 | 1570 | if (getfrag(from, |
5640f768 ED |
1571 | page_address(pfrag->page) + pfrag->offset, |
1572 | offset, copy, skb->len, skb) < 0) | |
1573 | goto error_efault; | |
1574 | ||
1575 | pfrag->offset += copy; | |
1576 | skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy); | |
1da177e4 LT |
1577 | skb->len += copy; |
1578 | skb->data_len += copy; | |
f945fa7a HX |
1579 | skb->truesize += copy; |
1580 | atomic_add(copy, &sk->sk_wmem_alloc); | |
1da177e4 LT |
1581 | } |
1582 | offset += copy; | |
1583 | length -= copy; | |
1584 | } | |
5640f768 | 1585 | |
1da177e4 | 1586 | return 0; |
5640f768 ED |
1587 | |
1588 | error_efault: | |
1589 | err = -EFAULT; | |
1da177e4 | 1590 | error: |
bdc712b4 | 1591 | cork->length -= length; |
3bd653c8 | 1592 | IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
1593 | return err; |
1594 | } | |
0bbe84a6 VY |
1595 | |
1596 | int ip6_append_data(struct sock *sk, | |
1597 | int getfrag(void *from, char *to, int offset, int len, | |
1598 | int odd, struct sk_buff *skb), | |
26879da5 WW |
1599 | void *from, int length, int transhdrlen, |
1600 | struct ipcm6_cookie *ipc6, struct flowi6 *fl6, | |
1601 | struct rt6_info *rt, unsigned int flags, | |
c14ac945 | 1602 | const struct sockcm_cookie *sockc) |
0bbe84a6 VY |
1603 | { |
1604 | struct inet_sock *inet = inet_sk(sk); | |
1605 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1606 | int exthdrlen; | |
1607 | int err; | |
1608 | ||
1609 | if (flags&MSG_PROBE) | |
1610 | return 0; | |
1611 | if (skb_queue_empty(&sk->sk_write_queue)) { | |
1612 | /* | |
1613 | * setup for corking | |
1614 | */ | |
26879da5 WW |
1615 | err = ip6_setup_cork(sk, &inet->cork, &np->cork, |
1616 | ipc6, rt, fl6); | |
0bbe84a6 VY |
1617 | if (err) |
1618 | return err; | |
1619 | ||
26879da5 | 1620 | exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0); |
0bbe84a6 VY |
1621 | length += exthdrlen; |
1622 | transhdrlen += exthdrlen; | |
1623 | } else { | |
1624 | fl6 = &inet->cork.fl.u.ip6; | |
1625 | transhdrlen = 0; | |
1626 | } | |
1627 | ||
1628 | return __ip6_append_data(sk, fl6, &sk->sk_write_queue, &inet->cork.base, | |
1629 | &np->cork, sk_page_frag(sk), getfrag, | |
26879da5 | 1630 | from, length, transhdrlen, flags, ipc6, sockc); |
0bbe84a6 | 1631 | } |
a495f836 | 1632 | EXPORT_SYMBOL_GPL(ip6_append_data); |
1da177e4 | 1633 | |
366e41d9 VY |
1634 | static void ip6_cork_release(struct inet_cork_full *cork, |
1635 | struct inet6_cork *v6_cork) | |
bf138862 | 1636 | { |
366e41d9 VY |
1637 | if (v6_cork->opt) { |
1638 | kfree(v6_cork->opt->dst0opt); | |
1639 | kfree(v6_cork->opt->dst1opt); | |
1640 | kfree(v6_cork->opt->hopopt); | |
1641 | kfree(v6_cork->opt->srcrt); | |
1642 | kfree(v6_cork->opt); | |
1643 | v6_cork->opt = NULL; | |
0178b695 HX |
1644 | } |
1645 | ||
366e41d9 VY |
1646 | if (cork->base.dst) { |
1647 | dst_release(cork->base.dst); | |
1648 | cork->base.dst = NULL; | |
1649 | cork->base.flags &= ~IPCORK_ALLFRAG; | |
bf138862 | 1650 | } |
366e41d9 | 1651 | memset(&cork->fl, 0, sizeof(cork->fl)); |
bf138862 PE |
1652 | } |
1653 | ||
6422398c VY |
1654 | struct sk_buff *__ip6_make_skb(struct sock *sk, |
1655 | struct sk_buff_head *queue, | |
1656 | struct inet_cork_full *cork, | |
1657 | struct inet6_cork *v6_cork) | |
1da177e4 LT |
1658 | { |
1659 | struct sk_buff *skb, *tmp_skb; | |
1660 | struct sk_buff **tail_skb; | |
1661 | struct in6_addr final_dst_buf, *final_dst = &final_dst_buf; | |
1da177e4 | 1662 | struct ipv6_pinfo *np = inet6_sk(sk); |
3bd653c8 | 1663 | struct net *net = sock_net(sk); |
1da177e4 | 1664 | struct ipv6hdr *hdr; |
6422398c VY |
1665 | struct ipv6_txoptions *opt = v6_cork->opt; |
1666 | struct rt6_info *rt = (struct rt6_info *)cork->base.dst; | |
1667 | struct flowi6 *fl6 = &cork->fl.u.ip6; | |
4c9483b2 | 1668 | unsigned char proto = fl6->flowi6_proto; |
1da177e4 | 1669 | |
6422398c | 1670 | skb = __skb_dequeue(queue); |
63159f29 | 1671 | if (!skb) |
1da177e4 LT |
1672 | goto out; |
1673 | tail_skb = &(skb_shinfo(skb)->frag_list); | |
1674 | ||
1675 | /* move skb->data to ip header from ext header */ | |
d56f90a7 | 1676 | if (skb->data < skb_network_header(skb)) |
bbe735e4 | 1677 | __skb_pull(skb, skb_network_offset(skb)); |
6422398c | 1678 | while ((tmp_skb = __skb_dequeue(queue)) != NULL) { |
cfe1fc77 | 1679 | __skb_pull(tmp_skb, skb_network_header_len(skb)); |
1da177e4 LT |
1680 | *tail_skb = tmp_skb; |
1681 | tail_skb = &(tmp_skb->next); | |
1682 | skb->len += tmp_skb->len; | |
1683 | skb->data_len += tmp_skb->len; | |
1da177e4 | 1684 | skb->truesize += tmp_skb->truesize; |
1da177e4 LT |
1685 | tmp_skb->destructor = NULL; |
1686 | tmp_skb->sk = NULL; | |
1da177e4 LT |
1687 | } |
1688 | ||
28a89453 | 1689 | /* Allow local fragmentation. */ |
60ff7467 | 1690 | skb->ignore_df = ip6_sk_ignore_df(sk); |
28a89453 | 1691 | |
4e3fd7a0 | 1692 | *final_dst = fl6->daddr; |
cfe1fc77 | 1693 | __skb_pull(skb, skb_network_header_len(skb)); |
1da177e4 LT |
1694 | if (opt && opt->opt_flen) |
1695 | ipv6_push_frag_opts(skb, opt, &proto); | |
1696 | if (opt && opt->opt_nflen) | |
613fa3ca | 1697 | ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst, &fl6->saddr); |
1da177e4 | 1698 | |
e2d1bca7 ACM |
1699 | skb_push(skb, sizeof(struct ipv6hdr)); |
1700 | skb_reset_network_header(skb); | |
0660e03f | 1701 | hdr = ipv6_hdr(skb); |
1ab1457c | 1702 | |
6422398c | 1703 | ip6_flow_hdr(hdr, v6_cork->tclass, |
cb1ce2ef | 1704 | ip6_make_flowlabel(net, skb, fl6->flowlabel, |
67800f9b | 1705 | np->autoflowlabel, fl6)); |
6422398c | 1706 | hdr->hop_limit = v6_cork->hop_limit; |
1da177e4 | 1707 | hdr->nexthdr = proto; |
4e3fd7a0 AD |
1708 | hdr->saddr = fl6->saddr; |
1709 | hdr->daddr = *final_dst; | |
1da177e4 | 1710 | |
a2c2064f | 1711 | skb->priority = sk->sk_priority; |
4a19ec58 | 1712 | skb->mark = sk->sk_mark; |
a2c2064f | 1713 | |
d8d1f30b | 1714 | skb_dst_set(skb, dst_clone(&rt->dst)); |
edf391ff | 1715 | IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); |
14878f75 | 1716 | if (proto == IPPROTO_ICMPV6) { |
adf30907 | 1717 | struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb)); |
14878f75 | 1718 | |
43a43b60 HFS |
1719 | ICMP6MSGOUT_INC_STATS(net, idev, icmp6_hdr(skb)->icmp6_type); |
1720 | ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS); | |
14878f75 DS |
1721 | } |
1722 | ||
6422398c VY |
1723 | ip6_cork_release(cork, v6_cork); |
1724 | out: | |
1725 | return skb; | |
1726 | } | |
1727 | ||
1728 | int ip6_send_skb(struct sk_buff *skb) | |
1729 | { | |
1730 | struct net *net = sock_net(skb->sk); | |
1731 | struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); | |
1732 | int err; | |
1733 | ||
33224b16 | 1734 | err = ip6_local_out(net, skb->sk, skb); |
1da177e4 LT |
1735 | if (err) { |
1736 | if (err > 0) | |
6ce9e7b5 | 1737 | err = net_xmit_errno(err); |
1da177e4 | 1738 | if (err) |
6422398c VY |
1739 | IP6_INC_STATS(net, rt->rt6i_idev, |
1740 | IPSTATS_MIB_OUTDISCARDS); | |
1da177e4 LT |
1741 | } |
1742 | ||
1da177e4 | 1743 | return err; |
6422398c VY |
1744 | } |
1745 | ||
1746 | int ip6_push_pending_frames(struct sock *sk) | |
1747 | { | |
1748 | struct sk_buff *skb; | |
1749 | ||
1750 | skb = ip6_finish_skb(sk); | |
1751 | if (!skb) | |
1752 | return 0; | |
1753 | ||
1754 | return ip6_send_skb(skb); | |
1da177e4 | 1755 | } |
a495f836 | 1756 | EXPORT_SYMBOL_GPL(ip6_push_pending_frames); |
1da177e4 | 1757 | |
0bbe84a6 | 1758 | static void __ip6_flush_pending_frames(struct sock *sk, |
6422398c VY |
1759 | struct sk_buff_head *queue, |
1760 | struct inet_cork_full *cork, | |
1761 | struct inet6_cork *v6_cork) | |
1da177e4 | 1762 | { |
1da177e4 LT |
1763 | struct sk_buff *skb; |
1764 | ||
0bbe84a6 | 1765 | while ((skb = __skb_dequeue_tail(queue)) != NULL) { |
adf30907 ED |
1766 | if (skb_dst(skb)) |
1767 | IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)), | |
e1f52208 | 1768 | IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
1769 | kfree_skb(skb); |
1770 | } | |
1771 | ||
6422398c | 1772 | ip6_cork_release(cork, v6_cork); |
1da177e4 | 1773 | } |
0bbe84a6 VY |
1774 | |
1775 | void ip6_flush_pending_frames(struct sock *sk) | |
1776 | { | |
6422398c VY |
1777 | __ip6_flush_pending_frames(sk, &sk->sk_write_queue, |
1778 | &inet_sk(sk)->cork, &inet6_sk(sk)->cork); | |
0bbe84a6 | 1779 | } |
a495f836 | 1780 | EXPORT_SYMBOL_GPL(ip6_flush_pending_frames); |
6422398c VY |
1781 | |
1782 | struct sk_buff *ip6_make_skb(struct sock *sk, | |
1783 | int getfrag(void *from, char *to, int offset, | |
1784 | int len, int odd, struct sk_buff *skb), | |
1785 | void *from, int length, int transhdrlen, | |
26879da5 | 1786 | struct ipcm6_cookie *ipc6, struct flowi6 *fl6, |
6422398c | 1787 | struct rt6_info *rt, unsigned int flags, |
26879da5 | 1788 | const struct sockcm_cookie *sockc) |
6422398c VY |
1789 | { |
1790 | struct inet_cork_full cork; | |
1791 | struct inet6_cork v6_cork; | |
1792 | struct sk_buff_head queue; | |
26879da5 | 1793 | int exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0); |
6422398c VY |
1794 | int err; |
1795 | ||
1796 | if (flags & MSG_PROBE) | |
1797 | return NULL; | |
1798 | ||
1799 | __skb_queue_head_init(&queue); | |
1800 | ||
1801 | cork.base.flags = 0; | |
1802 | cork.base.addr = 0; | |
1803 | cork.base.opt = NULL; | |
1804 | v6_cork.opt = NULL; | |
26879da5 | 1805 | err = ip6_setup_cork(sk, &cork, &v6_cork, ipc6, rt, fl6); |
6422398c VY |
1806 | if (err) |
1807 | return ERR_PTR(err); | |
1808 | ||
26879da5 WW |
1809 | if (ipc6->dontfrag < 0) |
1810 | ipc6->dontfrag = inet6_sk(sk)->dontfrag; | |
6422398c VY |
1811 | |
1812 | err = __ip6_append_data(sk, fl6, &queue, &cork.base, &v6_cork, | |
1813 | ¤t->task_frag, getfrag, from, | |
1814 | length + exthdrlen, transhdrlen + exthdrlen, | |
26879da5 | 1815 | flags, ipc6, sockc); |
6422398c VY |
1816 | if (err) { |
1817 | __ip6_flush_pending_frames(sk, &queue, &cork, &v6_cork); | |
1818 | return ERR_PTR(err); | |
1819 | } | |
1820 | ||
1821 | return __ip6_make_skb(sk, &queue, &cork, &v6_cork); | |
1822 | } |