Commit | Line | Data |
---|---|---|
2874c5fd | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
1da177e4 LT |
2 | /* |
3 | * IPv6 output functions | |
1ab1457c | 4 | * Linux INET6 implementation |
1da177e4 LT |
5 | * |
6 | * Authors: | |
1ab1457c | 7 | * Pedro Roque <roque@di.fc.ul.pt> |
1da177e4 | 8 | * |
1da177e4 LT |
9 | * Based on linux/net/ipv4/ip_output.c |
10 | * | |
1da177e4 LT |
11 | * Changes: |
12 | * A.N.Kuznetsov : airthmetics in fragmentation. | |
13 | * extension headers are implemented. | |
14 | * route changes now work. | |
15 | * ip6_forward does not confuse sniffers. | |
16 | * etc. | |
17 | * | |
18 | * H. von Brand : Added missing #include <linux/string.h> | |
67ba4152 | 19 | * Imran Patel : frag id should be in NBO |
1da177e4 LT |
20 | * Kazunori MIYAZAWA @USAGI |
21 | * : add ip6_append_data and related functions | |
22 | * for datagram xmit | |
23 | */ | |
24 | ||
1da177e4 | 25 | #include <linux/errno.h> |
ef76bc23 | 26 | #include <linux/kernel.h> |
1da177e4 LT |
27 | #include <linux/string.h> |
28 | #include <linux/socket.h> | |
29 | #include <linux/net.h> | |
30 | #include <linux/netdevice.h> | |
31 | #include <linux/if_arp.h> | |
32 | #include <linux/in6.h> | |
33 | #include <linux/tcp.h> | |
34 | #include <linux/route.h> | |
b59f45d0 | 35 | #include <linux/module.h> |
5a0e3ad6 | 36 | #include <linux/slab.h> |
1da177e4 | 37 | |
33b48679 | 38 | #include <linux/bpf-cgroup.h> |
1da177e4 LT |
39 | #include <linux/netfilter.h> |
40 | #include <linux/netfilter_ipv6.h> | |
41 | ||
42 | #include <net/sock.h> | |
43 | #include <net/snmp.h> | |
44 | ||
45 | #include <net/ipv6.h> | |
46 | #include <net/ndisc.h> | |
47 | #include <net/protocol.h> | |
48 | #include <net/ip6_route.h> | |
49 | #include <net/addrconf.h> | |
50 | #include <net/rawv6.h> | |
51 | #include <net/icmp.h> | |
52 | #include <net/xfrm.h> | |
53 | #include <net/checksum.h> | |
7bc570c8 | 54 | #include <linux/mroute6.h> |
ca254490 | 55 | #include <net/l3mdev.h> |
14972cbd | 56 | #include <net/lwtunnel.h> |
571912c6 | 57 | #include <net/ip_tunnels.h> |
1da177e4 | 58 | |
7d8c6e39 | 59 | static int ip6_finish_output2(struct net *net, struct sock *sk, struct sk_buff *skb) |
1da177e4 | 60 | { |
adf30907 | 61 | struct dst_entry *dst = skb_dst(skb); |
1da177e4 | 62 | struct net_device *dev = dst->dev; |
e415ed3a | 63 | struct inet6_dev *idev = ip6_dst_idev(dst); |
5796015f | 64 | unsigned int hh_len = LL_RESERVED_SPACE(dev); |
e415ed3a VA |
65 | const struct in6_addr *daddr, *nexthop; |
66 | struct ipv6hdr *hdr; | |
f6b72b62 | 67 | struct neighbour *neigh; |
6fd6ce20 | 68 | int ret; |
1da177e4 | 69 | |
5796015f | 70 | /* Be paranoid, rather than too clever. */ |
e415ed3a VA |
71 | if (unlikely(hh_len > skb_headroom(skb)) && dev->header_ops) { |
72 | skb = skb_expand_head(skb, hh_len); | |
5796015f | 73 | if (!skb) { |
e415ed3a | 74 | IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS); |
5796015f VA |
75 | return -ENOMEM; |
76 | } | |
77 | } | |
78 | ||
e415ed3a VA |
79 | hdr = ipv6_hdr(skb); |
80 | daddr = &hdr->daddr; | |
81 | if (ipv6_addr_is_multicast(daddr)) { | |
7026b1dd | 82 | if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(sk) && |
8571ab47 | 83 | ((mroute6_is_socket(net, skb) && |
bd91b8bf | 84 | !(IP6CB(skb)->flags & IP6SKB_FORWARDED)) || |
e415ed3a | 85 | ipv6_chk_mcast_addr(dev, daddr, &hdr->saddr))) { |
1da177e4 LT |
86 | struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC); |
87 | ||
88 | /* Do not check for IFF_ALLMULTI; multicast routing | |
89 | is not supported in any case. | |
90 | */ | |
91 | if (newskb) | |
b2e0b385 | 92 | NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING, |
29a26a56 | 93 | net, sk, newskb, NULL, newskb->dev, |
95603e22 | 94 | dev_loopback_xmit); |
1da177e4 | 95 | |
e415ed3a | 96 | if (hdr->hop_limit == 0) { |
78126c41 | 97 | IP6_INC_STATS(net, idev, |
3bd653c8 | 98 | IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
99 | kfree_skb(skb); |
100 | return 0; | |
101 | } | |
102 | } | |
103 | ||
78126c41 | 104 | IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUTMCAST, skb->len); |
e415ed3a | 105 | if (IPV6_ADDR_MC_SCOPE(daddr) <= IPV6_ADDR_SCOPE_NODELOCAL && |
dd408515 HFS |
106 | !(dev->flags & IFF_LOOPBACK)) { |
107 | kfree_skb(skb); | |
108 | return 0; | |
109 | } | |
1da177e4 LT |
110 | } |
111 | ||
14972cbd RP |
112 | if (lwtunnel_xmit_redirect(dst->lwtstate)) { |
113 | int res = lwtunnel_xmit(skb); | |
114 | ||
115 | if (res < 0 || res == LWTUNNEL_XMIT_DONE) | |
116 | return res; | |
117 | } | |
118 | ||
6fd6ce20 | 119 | rcu_read_lock_bh(); |
e415ed3a VA |
120 | nexthop = rt6_nexthop((struct rt6_info *)dst, daddr); |
121 | neigh = __ipv6_neigh_lookup_noref(dev, nexthop); | |
6fd6ce20 | 122 | if (unlikely(!neigh)) |
e415ed3a | 123 | neigh = __neigh_create(&nd_tbl, nexthop, dev, false); |
6fd6ce20 | 124 | if (!IS_ERR(neigh)) { |
4ff06203 | 125 | sock_confirm_neigh(skb, neigh); |
0353f282 | 126 | ret = neigh_output(neigh, skb, false); |
6fd6ce20 YH |
127 | rcu_read_unlock_bh(); |
128 | return ret; | |
129 | } | |
130 | rcu_read_unlock_bh(); | |
05e3aa09 | 131 | |
e415ed3a | 132 | IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTNOROUTES); |
9e508490 JE |
133 | kfree_skb(skb); |
134 | return -EINVAL; | |
1da177e4 LT |
135 | } |
136 | ||
b210de4f AL |
137 | static int |
138 | ip6_finish_output_gso_slowpath_drop(struct net *net, struct sock *sk, | |
139 | struct sk_buff *skb, unsigned int mtu) | |
140 | { | |
141 | struct sk_buff *segs, *nskb; | |
142 | netdev_features_t features; | |
143 | int ret = 0; | |
144 | ||
145 | /* Please see corresponding comment in ip_finish_output_gso | |
146 | * describing the cases where GSO segment length exceeds the | |
147 | * egress MTU. | |
148 | */ | |
149 | features = netif_skb_features(skb); | |
150 | segs = skb_gso_segment(skb, features & ~NETIF_F_GSO_MASK); | |
151 | if (IS_ERR_OR_NULL(segs)) { | |
152 | kfree_skb(skb); | |
153 | return -ENOMEM; | |
154 | } | |
155 | ||
156 | consume_skb(skb); | |
157 | ||
158 | skb_list_walk_safe(segs, segs, nskb) { | |
159 | int err; | |
160 | ||
161 | skb_mark_not_on_list(segs); | |
162 | err = ip6_fragment(net, sk, segs, ip6_finish_output2); | |
163 | if (err && ret == 0) | |
164 | ret = err; | |
165 | } | |
166 | ||
167 | return ret; | |
168 | } | |
169 | ||
956fe219 | 170 | static int __ip6_finish_output(struct net *net, struct sock *sk, struct sk_buff *skb) |
9e508490 | 171 | { |
b210de4f AL |
172 | unsigned int mtu; |
173 | ||
09ee9dba TB |
174 | #if defined(CONFIG_NETFILTER) && defined(CONFIG_XFRM) |
175 | /* Policy lookup after SNAT yielded a new policy */ | |
176 | if (skb_dst(skb)->xfrm) { | |
19d36c5f | 177 | IP6CB(skb)->flags |= IP6SKB_REROUTED; |
09ee9dba TB |
178 | return dst_output(net, sk, skb); |
179 | } | |
180 | #endif | |
181 | ||
b210de4f AL |
182 | mtu = ip6_skb_dst_mtu(skb); |
183 | if (skb_is_gso(skb) && !skb_gso_validate_network_len(skb, mtu)) | |
184 | return ip6_finish_output_gso_slowpath_drop(net, sk, skb, mtu); | |
185 | ||
186 | if ((skb->len > mtu && !skb_is_gso(skb)) || | |
9037c357 JP |
187 | dst_allfrag(skb_dst(skb)) || |
188 | (IP6CB(skb)->frag_max_size && skb->len > IP6CB(skb)->frag_max_size)) | |
7d8c6e39 | 189 | return ip6_fragment(net, sk, skb, ip6_finish_output2); |
9e508490 | 190 | else |
7d8c6e39 | 191 | return ip6_finish_output2(net, sk, skb); |
9e508490 JE |
192 | } |
193 | ||
956fe219 | 194 | static int ip6_finish_output(struct net *net, struct sock *sk, struct sk_buff *skb) |
195 | { | |
196 | int ret; | |
197 | ||
198 | ret = BPF_CGROUP_RUN_PROG_INET_EGRESS(sk, skb); | |
199 | switch (ret) { | |
200 | case NET_XMIT_SUCCESS: | |
201 | return __ip6_finish_output(net, sk, skb); | |
202 | case NET_XMIT_CN: | |
203 | return __ip6_finish_output(net, sk, skb) ? : ret; | |
204 | default: | |
205 | kfree_skb(skb); | |
206 | return ret; | |
207 | } | |
208 | } | |
209 | ||
ede2059d | 210 | int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb) |
1da177e4 | 211 | { |
28f8bfd1 | 212 | struct net_device *dev = skb_dst(skb)->dev, *indev = skb->dev; |
adf30907 | 213 | struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb)); |
be10de0a | 214 | |
97a7a37a CF |
215 | skb->protocol = htons(ETH_P_IPV6); |
216 | skb->dev = dev; | |
217 | ||
778d80be | 218 | if (unlikely(idev->cnf.disable_ipv6)) { |
19a0644c | 219 | IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS); |
778d80be YH |
220 | kfree_skb(skb); |
221 | return 0; | |
222 | } | |
223 | ||
29a26a56 | 224 | return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING, |
28f8bfd1 | 225 | net, sk, skb, indev, dev, |
9c6eb28a JE |
226 | ip6_finish_output, |
227 | !(IP6CB(skb)->flags & IP6SKB_REROUTED)); | |
1da177e4 | 228 | } |
6585d7dc | 229 | EXPORT_SYMBOL(ip6_output); |
1da177e4 | 230 | |
e9191ffb | 231 | bool ip6_autoflowlabel(struct net *net, const struct ipv6_pinfo *np) |
513674b5 SL |
232 | { |
233 | if (!np->autoflowlabel_set) | |
234 | return ip6_default_np_autolabel(net); | |
235 | else | |
236 | return np->autoflowlabel; | |
237 | } | |
238 | ||
1da177e4 | 239 | /* |
1c1e9d2b ED |
240 | * xmit an sk_buff (used by TCP, SCTP and DCCP) |
241 | * Note : socket lock is not held for SYNACK packets, but might be modified | |
242 | * by calls to skb_set_owner_w() and ipv6_local_error(), | |
243 | * which are using proper atomic operations or spinlocks. | |
1da177e4 | 244 | */ |
1c1e9d2b | 245 | int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6, |
4f6570d7 | 246 | __u32 mark, struct ipv6_txoptions *opt, int tclass, u32 priority) |
1da177e4 | 247 | { |
3bd653c8 | 248 | struct net *net = sock_net(sk); |
1c1e9d2b | 249 | const struct ipv6_pinfo *np = inet6_sk(sk); |
4c9483b2 | 250 | struct in6_addr *first_hop = &fl6->daddr; |
adf30907 | 251 | struct dst_entry *dst = skb_dst(skb); |
0c9f227b VA |
252 | struct net_device *dev = dst->dev; |
253 | struct inet6_dev *idev = ip6_dst_idev(dst); | |
66033f47 | 254 | unsigned int head_room; |
1da177e4 | 255 | struct ipv6hdr *hdr; |
4c9483b2 | 256 | u8 proto = fl6->flowi6_proto; |
1da177e4 | 257 | int seg_len = skb->len; |
e651f03a | 258 | int hlimit = -1; |
1da177e4 LT |
259 | u32 mtu; |
260 | ||
0c9f227b | 261 | head_room = sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dev); |
66033f47 SB |
262 | if (opt) |
263 | head_room += opt->opt_nflen + opt->opt_flen; | |
264 | ||
0c9f227b VA |
265 | if (unlikely(head_room > skb_headroom(skb))) { |
266 | skb = skb_expand_head(skb, head_room); | |
267 | if (!skb) { | |
268 | IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS); | |
66033f47 | 269 | return -ENOBUFS; |
1da177e4 | 270 | } |
66033f47 SB |
271 | } |
272 | ||
273 | if (opt) { | |
274 | seg_len += opt->opt_nflen + opt->opt_flen; | |
275 | ||
1da177e4 LT |
276 | if (opt->opt_flen) |
277 | ipv6_push_frag_opts(skb, opt, &proto); | |
66033f47 | 278 | |
1da177e4 | 279 | if (opt->opt_nflen) |
613fa3ca DL |
280 | ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop, |
281 | &fl6->saddr); | |
1da177e4 LT |
282 | } |
283 | ||
e2d1bca7 ACM |
284 | skb_push(skb, sizeof(struct ipv6hdr)); |
285 | skb_reset_network_header(skb); | |
0660e03f | 286 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
287 | |
288 | /* | |
289 | * Fill in the IPv6 header | |
290 | */ | |
b903d324 | 291 | if (np) |
1da177e4 LT |
292 | hlimit = np->hop_limit; |
293 | if (hlimit < 0) | |
6b75d090 | 294 | hlimit = ip6_dst_hoplimit(dst); |
1da177e4 | 295 | |
cb1ce2ef | 296 | ip6_flow_hdr(hdr, tclass, ip6_make_flowlabel(net, skb, fl6->flowlabel, |
513674b5 | 297 | ip6_autoflowlabel(net, np), fl6)); |
41a1f8ea | 298 | |
1da177e4 LT |
299 | hdr->payload_len = htons(seg_len); |
300 | hdr->nexthdr = proto; | |
301 | hdr->hop_limit = hlimit; | |
302 | ||
4e3fd7a0 AD |
303 | hdr->saddr = fl6->saddr; |
304 | hdr->daddr = *first_hop; | |
1da177e4 | 305 | |
9c9c9ad5 | 306 | skb->protocol = htons(ETH_P_IPV6); |
4f6570d7 | 307 | skb->priority = priority; |
92e55f41 | 308 | skb->mark = mark; |
a2c2064f | 309 | |
1da177e4 | 310 | mtu = dst_mtu(dst); |
60ff7467 | 311 | if ((skb->len <= mtu) || skb->ignore_df || skb_is_gso(skb)) { |
0c9f227b | 312 | IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len); |
a8e3e1a9 DA |
313 | |
314 | /* if egress device is enslaved to an L3 master device pass the | |
315 | * skb to its handler for processing | |
316 | */ | |
317 | skb = l3mdev_ip6_out((struct sock *)sk, skb); | |
318 | if (unlikely(!skb)) | |
319 | return 0; | |
320 | ||
1c1e9d2b ED |
321 | /* hooks should never assume socket lock is held. |
322 | * we promote our socket to non const | |
323 | */ | |
29a26a56 | 324 | return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, |
0c9f227b | 325 | net, (struct sock *)sk, skb, NULL, dev, |
13206b6b | 326 | dst_output); |
1da177e4 LT |
327 | } |
328 | ||
0c9f227b | 329 | skb->dev = dev; |
1c1e9d2b ED |
330 | /* ipv6_local_error() does not require socket lock, |
331 | * we promote our socket to non const | |
332 | */ | |
333 | ipv6_local_error((struct sock *)sk, EMSGSIZE, fl6, mtu); | |
334 | ||
0c9f227b | 335 | IP6_INC_STATS(net, idev, IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
336 | kfree_skb(skb); |
337 | return -EMSGSIZE; | |
338 | } | |
7159039a YH |
339 | EXPORT_SYMBOL(ip6_xmit); |
340 | ||
1da177e4 LT |
341 | static int ip6_call_ra_chain(struct sk_buff *skb, int sel) |
342 | { | |
343 | struct ip6_ra_chain *ra; | |
344 | struct sock *last = NULL; | |
345 | ||
346 | read_lock(&ip6_ra_lock); | |
347 | for (ra = ip6_ra_chain; ra; ra = ra->next) { | |
348 | struct sock *sk = ra->sk; | |
0bd1b59b AM |
349 | if (sk && ra->sel == sel && |
350 | (!sk->sk_bound_dev_if || | |
351 | sk->sk_bound_dev_if == skb->dev->ifindex)) { | |
9036b2fe FR |
352 | struct ipv6_pinfo *np = inet6_sk(sk); |
353 | ||
354 | if (np && np->rtalert_isolate && | |
355 | !net_eq(sock_net(sk), dev_net(skb->dev))) { | |
356 | continue; | |
357 | } | |
1da177e4 LT |
358 | if (last) { |
359 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
360 | if (skb2) | |
361 | rawv6_rcv(last, skb2); | |
362 | } | |
363 | last = sk; | |
364 | } | |
365 | } | |
366 | ||
367 | if (last) { | |
368 | rawv6_rcv(last, skb); | |
369 | read_unlock(&ip6_ra_lock); | |
370 | return 1; | |
371 | } | |
372 | read_unlock(&ip6_ra_lock); | |
373 | return 0; | |
374 | } | |
375 | ||
e21e0b5f VN |
376 | static int ip6_forward_proxy_check(struct sk_buff *skb) |
377 | { | |
0660e03f | 378 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
e21e0b5f | 379 | u8 nexthdr = hdr->nexthdr; |
75f2811c | 380 | __be16 frag_off; |
e21e0b5f VN |
381 | int offset; |
382 | ||
383 | if (ipv6_ext_hdr(nexthdr)) { | |
75f2811c | 384 | offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off); |
e21e0b5f VN |
385 | if (offset < 0) |
386 | return 0; | |
387 | } else | |
388 | offset = sizeof(struct ipv6hdr); | |
389 | ||
390 | if (nexthdr == IPPROTO_ICMPV6) { | |
391 | struct icmp6hdr *icmp6; | |
392 | ||
d56f90a7 ACM |
393 | if (!pskb_may_pull(skb, (skb_network_header(skb) + |
394 | offset + 1 - skb->data))) | |
e21e0b5f VN |
395 | return 0; |
396 | ||
d56f90a7 | 397 | icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset); |
e21e0b5f VN |
398 | |
399 | switch (icmp6->icmp6_type) { | |
400 | case NDISC_ROUTER_SOLICITATION: | |
401 | case NDISC_ROUTER_ADVERTISEMENT: | |
402 | case NDISC_NEIGHBOUR_SOLICITATION: | |
403 | case NDISC_NEIGHBOUR_ADVERTISEMENT: | |
404 | case NDISC_REDIRECT: | |
405 | /* For reaction involving unicast neighbor discovery | |
406 | * message destined to the proxied address, pass it to | |
407 | * input function. | |
408 | */ | |
409 | return 1; | |
410 | default: | |
411 | break; | |
412 | } | |
413 | } | |
414 | ||
74553b09 VN |
415 | /* |
416 | * The proxying router can't forward traffic sent to a link-local | |
417 | * address, so signal the sender and discard the packet. This | |
418 | * behavior is clarified by the MIPv6 specification. | |
419 | */ | |
420 | if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) { | |
421 | dst_link_failure(skb); | |
422 | return -1; | |
423 | } | |
424 | ||
e21e0b5f VN |
425 | return 0; |
426 | } | |
427 | ||
0c4b51f0 EB |
428 | static inline int ip6_forward_finish(struct net *net, struct sock *sk, |
429 | struct sk_buff *skb) | |
1da177e4 | 430 | { |
71a1c915 JB |
431 | struct dst_entry *dst = skb_dst(skb); |
432 | ||
433 | __IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS); | |
434 | __IP6_ADD_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len); | |
435 | ||
f839a6c9 IS |
436 | #ifdef CONFIG_NET_SWITCHDEV |
437 | if (skb->offload_l3_fwd_mark) { | |
438 | consume_skb(skb); | |
439 | return 0; | |
440 | } | |
441 | #endif | |
442 | ||
8203e2d8 | 443 | skb->tstamp = 0; |
13206b6b | 444 | return dst_output(net, sk, skb); |
1da177e4 LT |
445 | } |
446 | ||
fe6cc55f FW |
447 | static bool ip6_pkt_too_big(const struct sk_buff *skb, unsigned int mtu) |
448 | { | |
418a3156 | 449 | if (skb->len <= mtu) |
fe6cc55f FW |
450 | return false; |
451 | ||
60ff7467 | 452 | /* ipv6 conntrack defrag sets max_frag_size + ignore_df */ |
fe6cc55f FW |
453 | if (IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu) |
454 | return true; | |
455 | ||
60ff7467 | 456 | if (skb->ignore_df) |
418a3156 FW |
457 | return false; |
458 | ||
779b7931 | 459 | if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu)) |
fe6cc55f FW |
460 | return false; |
461 | ||
462 | return true; | |
463 | } | |
464 | ||
1da177e4 LT |
465 | int ip6_forward(struct sk_buff *skb) |
466 | { | |
adf30907 | 467 | struct dst_entry *dst = skb_dst(skb); |
0660e03f | 468 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
1da177e4 | 469 | struct inet6_skb_parm *opt = IP6CB(skb); |
c346dca1 | 470 | struct net *net = dev_net(dst->dev); |
0857d6f8 | 471 | struct inet6_dev *idev; |
14f3ad6f | 472 | u32 mtu; |
1ab1457c | 473 | |
0857d6f8 | 474 | idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif)); |
53b7997f | 475 | if (net->ipv6.devconf_all->forwarding == 0) |
1da177e4 LT |
476 | goto error; |
477 | ||
090f1166 LR |
478 | if (skb->pkt_type != PACKET_HOST) |
479 | goto drop; | |
480 | ||
9ef2e965 HFS |
481 | if (unlikely(skb->sk)) |
482 | goto drop; | |
483 | ||
4497b076 BH |
484 | if (skb_warn_if_lro(skb)) |
485 | goto drop; | |
486 | ||
ccd27f05 ND |
487 | if (!net->ipv6.devconf_all->disable_policy && |
488 | !idev->cnf.disable_policy && | |
489 | !xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) { | |
bdb7cc64 | 490 | __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS); |
1da177e4 LT |
491 | goto drop; |
492 | } | |
493 | ||
35fc92a9 | 494 | skb_forward_csum(skb); |
1da177e4 LT |
495 | |
496 | /* | |
497 | * We DO NOT make any processing on | |
498 | * RA packets, pushing them to user level AS IS | |
499 | * without ane WARRANTY that application will be able | |
500 | * to interpret them. The reason is that we | |
501 | * cannot make anything clever here. | |
502 | * | |
503 | * We are not end-node, so that if packet contains | |
504 | * AH/ESP, we cannot make anything. | |
505 | * Defragmentation also would be mistake, RA packets | |
506 | * cannot be fragmented, because there is no warranty | |
507 | * that different fragments will go along one path. --ANK | |
508 | */ | |
ab4eb353 YH |
509 | if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) { |
510 | if (ip6_call_ra_chain(skb, ntohs(opt->ra))) | |
1da177e4 LT |
511 | return 0; |
512 | } | |
513 | ||
514 | /* | |
515 | * check and decrement ttl | |
516 | */ | |
517 | if (hdr->hop_limit <= 1) { | |
3ffe533c | 518 | icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0); |
bdb7cc64 | 519 | __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS); |
1da177e4 LT |
520 | |
521 | kfree_skb(skb); | |
522 | return -ETIMEDOUT; | |
523 | } | |
524 | ||
fbea49e1 | 525 | /* XXX: idev->cnf.proxy_ndp? */ |
53b7997f | 526 | if (net->ipv6.devconf_all->proxy_ndp && |
8a3edd80 | 527 | pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) { |
74553b09 | 528 | int proxied = ip6_forward_proxy_check(skb); |
46c7655f KP |
529 | if (proxied > 0) { |
530 | hdr->hop_limit--; | |
e21e0b5f | 531 | return ip6_input(skb); |
46c7655f | 532 | } else if (proxied < 0) { |
bdb7cc64 | 533 | __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS); |
74553b09 VN |
534 | goto drop; |
535 | } | |
e21e0b5f VN |
536 | } |
537 | ||
1da177e4 | 538 | if (!xfrm6_route_forward(skb)) { |
bdb7cc64 | 539 | __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS); |
1da177e4 LT |
540 | goto drop; |
541 | } | |
adf30907 | 542 | dst = skb_dst(skb); |
1da177e4 LT |
543 | |
544 | /* IPv6 specs say nothing about it, but it is clear that we cannot | |
545 | send redirects to source routed frames. | |
1e5dc146 | 546 | We don't send redirects to frames decapsulated from IPsec. |
1da177e4 | 547 | */ |
2f17becf SS |
548 | if (IP6CB(skb)->iif == dst->dev->ifindex && |
549 | opt->srcrt == 0 && !skb_sec_path(skb)) { | |
1da177e4 | 550 | struct in6_addr *target = NULL; |
fbfe95a4 | 551 | struct inet_peer *peer; |
1da177e4 | 552 | struct rt6_info *rt; |
1da177e4 LT |
553 | |
554 | /* | |
555 | * incoming and outgoing devices are the same | |
556 | * send a redirect. | |
557 | */ | |
558 | ||
559 | rt = (struct rt6_info *) dst; | |
c45a3dfb DM |
560 | if (rt->rt6i_flags & RTF_GATEWAY) |
561 | target = &rt->rt6i_gateway; | |
1da177e4 LT |
562 | else |
563 | target = &hdr->daddr; | |
564 | ||
fd0273d7 | 565 | peer = inet_getpeer_v6(net->ipv6.peers, &hdr->daddr, 1); |
92d86829 | 566 | |
1da177e4 LT |
567 | /* Limit redirects both by destination (here) |
568 | and by source (inside ndisc_send_redirect) | |
569 | */ | |
fbfe95a4 | 570 | if (inet_peer_xrlim_allow(peer, 1*HZ)) |
4991969a | 571 | ndisc_send_redirect(skb, target); |
1d861aa4 DM |
572 | if (peer) |
573 | inet_putpeer(peer); | |
5bb1ab09 DS |
574 | } else { |
575 | int addrtype = ipv6_addr_type(&hdr->saddr); | |
576 | ||
1da177e4 | 577 | /* This check is security critical. */ |
f81b2e7d YH |
578 | if (addrtype == IPV6_ADDR_ANY || |
579 | addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK)) | |
5bb1ab09 DS |
580 | goto error; |
581 | if (addrtype & IPV6_ADDR_LINKLOCAL) { | |
582 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, | |
3ffe533c | 583 | ICMPV6_NOT_NEIGHBOUR, 0); |
5bb1ab09 DS |
584 | goto error; |
585 | } | |
1da177e4 LT |
586 | } |
587 | ||
427faee1 | 588 | mtu = ip6_dst_mtu_maybe_forward(dst, true); |
14f3ad6f UW |
589 | if (mtu < IPV6_MIN_MTU) |
590 | mtu = IPV6_MIN_MTU; | |
591 | ||
fe6cc55f | 592 | if (ip6_pkt_too_big(skb, mtu)) { |
1da177e4 LT |
593 | /* Again, force OUTPUT device used as source address */ |
594 | skb->dev = dst->dev; | |
14f3ad6f | 595 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); |
bdb7cc64 | 596 | __IP6_INC_STATS(net, idev, IPSTATS_MIB_INTOOBIGERRORS); |
1d015503 ED |
597 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
598 | IPSTATS_MIB_FRAGFAILS); | |
1da177e4 LT |
599 | kfree_skb(skb); |
600 | return -EMSGSIZE; | |
601 | } | |
602 | ||
603 | if (skb_cow(skb, dst->dev->hard_header_len)) { | |
1d015503 ED |
604 | __IP6_INC_STATS(net, ip6_dst_idev(dst), |
605 | IPSTATS_MIB_OUTDISCARDS); | |
1da177e4 LT |
606 | goto drop; |
607 | } | |
608 | ||
0660e03f | 609 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
610 | |
611 | /* Mangling hops number delayed to point after skb COW */ | |
1ab1457c | 612 | |
1da177e4 LT |
613 | hdr->hop_limit--; |
614 | ||
29a26a56 EB |
615 | return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, |
616 | net, NULL, skb, skb->dev, dst->dev, | |
6e23ae2a | 617 | ip6_forward_finish); |
1da177e4 LT |
618 | |
619 | error: | |
bdb7cc64 | 620 | __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS); |
1da177e4 LT |
621 | drop: |
622 | kfree_skb(skb); | |
623 | return -EINVAL; | |
624 | } | |
625 | ||
626 | static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from) | |
627 | { | |
628 | to->pkt_type = from->pkt_type; | |
629 | to->priority = from->priority; | |
630 | to->protocol = from->protocol; | |
adf30907 ED |
631 | skb_dst_drop(to); |
632 | skb_dst_set(to, dst_clone(skb_dst(from))); | |
1da177e4 | 633 | to->dev = from->dev; |
82e91ffe | 634 | to->mark = from->mark; |
1da177e4 | 635 | |
3dd1c9a1 PA |
636 | skb_copy_hash(to, from); |
637 | ||
1da177e4 LT |
638 | #ifdef CONFIG_NET_SCHED |
639 | to->tc_index = from->tc_index; | |
640 | #endif | |
e7ac05f3 | 641 | nf_copy(to, from); |
df5042f4 | 642 | skb_ext_copy(to, from); |
984bc16c | 643 | skb_copy_secmark(to, from); |
1da177e4 LT |
644 | } |
645 | ||
0feca619 PNA |
646 | int ip6_fraglist_init(struct sk_buff *skb, unsigned int hlen, u8 *prevhdr, |
647 | u8 nexthdr, __be32 frag_id, | |
648 | struct ip6_fraglist_iter *iter) | |
649 | { | |
650 | unsigned int first_len; | |
651 | struct frag_hdr *fh; | |
652 | ||
653 | /* BUILD HEADER */ | |
654 | *prevhdr = NEXTHDR_FRAGMENT; | |
655 | iter->tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC); | |
656 | if (!iter->tmp_hdr) | |
657 | return -ENOMEM; | |
658 | ||
b7034146 | 659 | iter->frag = skb_shinfo(skb)->frag_list; |
0feca619 PNA |
660 | skb_frag_list_init(skb); |
661 | ||
662 | iter->offset = 0; | |
663 | iter->hlen = hlen; | |
664 | iter->frag_id = frag_id; | |
665 | iter->nexthdr = nexthdr; | |
666 | ||
667 | __skb_pull(skb, hlen); | |
668 | fh = __skb_push(skb, sizeof(struct frag_hdr)); | |
669 | __skb_push(skb, hlen); | |
670 | skb_reset_network_header(skb); | |
671 | memcpy(skb_network_header(skb), iter->tmp_hdr, hlen); | |
672 | ||
673 | fh->nexthdr = nexthdr; | |
674 | fh->reserved = 0; | |
675 | fh->frag_off = htons(IP6_MF); | |
676 | fh->identification = frag_id; | |
677 | ||
678 | first_len = skb_pagelen(skb); | |
679 | skb->data_len = first_len - skb_headlen(skb); | |
680 | skb->len = first_len; | |
681 | ipv6_hdr(skb)->payload_len = htons(first_len - sizeof(struct ipv6hdr)); | |
682 | ||
683 | return 0; | |
684 | } | |
685 | EXPORT_SYMBOL(ip6_fraglist_init); | |
686 | ||
687 | void ip6_fraglist_prepare(struct sk_buff *skb, | |
688 | struct ip6_fraglist_iter *iter) | |
689 | { | |
690 | struct sk_buff *frag = iter->frag; | |
691 | unsigned int hlen = iter->hlen; | |
692 | struct frag_hdr *fh; | |
693 | ||
694 | frag->ip_summed = CHECKSUM_NONE; | |
695 | skb_reset_transport_header(frag); | |
696 | fh = __skb_push(frag, sizeof(struct frag_hdr)); | |
697 | __skb_push(frag, hlen); | |
698 | skb_reset_network_header(frag); | |
699 | memcpy(skb_network_header(frag), iter->tmp_hdr, hlen); | |
700 | iter->offset += skb->len - hlen - sizeof(struct frag_hdr); | |
701 | fh->nexthdr = iter->nexthdr; | |
702 | fh->reserved = 0; | |
703 | fh->frag_off = htons(iter->offset); | |
704 | if (frag->next) | |
705 | fh->frag_off |= htons(IP6_MF); | |
706 | fh->identification = iter->frag_id; | |
707 | ipv6_hdr(frag)->payload_len = htons(frag->len - sizeof(struct ipv6hdr)); | |
708 | ip6_copy_metadata(frag, skb); | |
709 | } | |
710 | EXPORT_SYMBOL(ip6_fraglist_prepare); | |
711 | ||
8a6a1f17 PNA |
712 | void ip6_frag_init(struct sk_buff *skb, unsigned int hlen, unsigned int mtu, |
713 | unsigned short needed_tailroom, int hdr_room, u8 *prevhdr, | |
714 | u8 nexthdr, __be32 frag_id, struct ip6_frag_state *state) | |
715 | { | |
716 | state->prevhdr = prevhdr; | |
717 | state->nexthdr = nexthdr; | |
718 | state->frag_id = frag_id; | |
719 | ||
720 | state->hlen = hlen; | |
721 | state->mtu = mtu; | |
722 | ||
723 | state->left = skb->len - hlen; /* Space per frame */ | |
724 | state->ptr = hlen; /* Where to start from */ | |
725 | ||
726 | state->hroom = hdr_room; | |
727 | state->troom = needed_tailroom; | |
728 | ||
729 | state->offset = 0; | |
730 | } | |
731 | EXPORT_SYMBOL(ip6_frag_init); | |
732 | ||
733 | struct sk_buff *ip6_frag_next(struct sk_buff *skb, struct ip6_frag_state *state) | |
734 | { | |
735 | u8 *prevhdr = state->prevhdr, *fragnexthdr_offset; | |
736 | struct sk_buff *frag; | |
737 | struct frag_hdr *fh; | |
738 | unsigned int len; | |
739 | ||
740 | len = state->left; | |
741 | /* IF: it doesn't fit, use 'mtu' - the data space left */ | |
742 | if (len > state->mtu) | |
743 | len = state->mtu; | |
744 | /* IF: we are not sending up to and including the packet end | |
745 | then align the next start on an eight byte boundary */ | |
746 | if (len < state->left) | |
747 | len &= ~7; | |
748 | ||
749 | /* Allocate buffer */ | |
750 | frag = alloc_skb(len + state->hlen + sizeof(struct frag_hdr) + | |
751 | state->hroom + state->troom, GFP_ATOMIC); | |
752 | if (!frag) | |
753 | return ERR_PTR(-ENOMEM); | |
754 | ||
755 | /* | |
756 | * Set up data on packet | |
757 | */ | |
758 | ||
759 | ip6_copy_metadata(frag, skb); | |
760 | skb_reserve(frag, state->hroom); | |
761 | skb_put(frag, len + state->hlen + sizeof(struct frag_hdr)); | |
762 | skb_reset_network_header(frag); | |
763 | fh = (struct frag_hdr *)(skb_network_header(frag) + state->hlen); | |
764 | frag->transport_header = (frag->network_header + state->hlen + | |
765 | sizeof(struct frag_hdr)); | |
766 | ||
767 | /* | |
768 | * Charge the memory for the fragment to any owner | |
769 | * it might possess | |
770 | */ | |
771 | if (skb->sk) | |
772 | skb_set_owner_w(frag, skb->sk); | |
773 | ||
774 | /* | |
775 | * Copy the packet header into the new buffer. | |
776 | */ | |
777 | skb_copy_from_linear_data(skb, skb_network_header(frag), state->hlen); | |
778 | ||
779 | fragnexthdr_offset = skb_network_header(frag); | |
780 | fragnexthdr_offset += prevhdr - skb_network_header(skb); | |
781 | *fragnexthdr_offset = NEXTHDR_FRAGMENT; | |
782 | ||
783 | /* | |
784 | * Build fragment header. | |
785 | */ | |
786 | fh->nexthdr = state->nexthdr; | |
787 | fh->reserved = 0; | |
788 | fh->identification = state->frag_id; | |
789 | ||
790 | /* | |
791 | * Copy a block of the IP datagram. | |
792 | */ | |
793 | BUG_ON(skb_copy_bits(skb, state->ptr, skb_transport_header(frag), | |
794 | len)); | |
795 | state->left -= len; | |
796 | ||
797 | fh->frag_off = htons(state->offset); | |
798 | if (state->left > 0) | |
799 | fh->frag_off |= htons(IP6_MF); | |
800 | ipv6_hdr(frag)->payload_len = htons(frag->len - sizeof(struct ipv6hdr)); | |
801 | ||
802 | state->ptr += len; | |
803 | state->offset += len; | |
804 | ||
805 | return frag; | |
806 | } | |
807 | EXPORT_SYMBOL(ip6_frag_next); | |
808 | ||
7d8c6e39 EB |
809 | int ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb, |
810 | int (*output)(struct net *, struct sock *, struct sk_buff *)) | |
1da177e4 | 811 | { |
1da177e4 | 812 | struct sk_buff *frag; |
67ba4152 | 813 | struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); |
f60e5990 | 814 | struct ipv6_pinfo *np = skb->sk && !dev_recursion_level() ? |
815 | inet6_sk(skb->sk) : NULL; | |
8a6a1f17 PNA |
816 | struct ip6_frag_state state; |
817 | unsigned int mtu, hlen, nexthdr_offset; | |
9669fffc | 818 | ktime_t tstamp = skb->tstamp; |
8a6a1f17 | 819 | int hroom, err = 0; |
286c2349 | 820 | __be32 frag_id; |
1da177e4 LT |
821 | u8 *prevhdr, nexthdr = 0; |
822 | ||
7dd7eb95 DM |
823 | err = ip6_find_1stfragopt(skb, &prevhdr); |
824 | if (err < 0) | |
2423496a | 825 | goto fail; |
7dd7eb95 | 826 | hlen = err; |
1da177e4 | 827 | nexthdr = *prevhdr; |
ef0efcd3 | 828 | nexthdr_offset = prevhdr - skb_network_header(skb); |
1da177e4 | 829 | |
628a5c56 | 830 | mtu = ip6_skb_dst_mtu(skb); |
b881ef76 JH |
831 | |
832 | /* We must not fragment if the socket is set to force MTU discovery | |
14f3ad6f | 833 | * or if the skb it not generated by a local socket. |
b881ef76 | 834 | */ |
485fca66 FW |
835 | if (unlikely(!skb->ignore_df && skb->len > mtu)) |
836 | goto fail_toobig; | |
a34a101e | 837 | |
485fca66 FW |
838 | if (IP6CB(skb)->frag_max_size) { |
839 | if (IP6CB(skb)->frag_max_size > mtu) | |
840 | goto fail_toobig; | |
841 | ||
842 | /* don't send fragments larger than what we received */ | |
843 | mtu = IP6CB(skb)->frag_max_size; | |
844 | if (mtu < IPV6_MIN_MTU) | |
845 | mtu = IPV6_MIN_MTU; | |
b881ef76 JH |
846 | } |
847 | ||
d91675f9 YH |
848 | if (np && np->frag_size < mtu) { |
849 | if (np->frag_size) | |
850 | mtu = np->frag_size; | |
851 | } | |
89bc7848 | 852 | if (mtu < hlen + sizeof(struct frag_hdr) + 8) |
b72a2b01 | 853 | goto fail_toobig; |
1e0d69a9 | 854 | mtu -= hlen + sizeof(struct frag_hdr); |
1da177e4 | 855 | |
fd0273d7 MKL |
856 | frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr, |
857 | &ipv6_hdr(skb)->saddr); | |
286c2349 | 858 | |
405c92f7 HFS |
859 | if (skb->ip_summed == CHECKSUM_PARTIAL && |
860 | (err = skb_checksum_help(skb))) | |
861 | goto fail; | |
862 | ||
ef0efcd3 | 863 | prevhdr = skb_network_header(skb) + nexthdr_offset; |
1d325d21 | 864 | hroom = LL_RESERVED_SPACE(rt->dst.dev); |
21dc3301 | 865 | if (skb_has_frag_list(skb)) { |
c72d8cda | 866 | unsigned int first_len = skb_pagelen(skb); |
0feca619 | 867 | struct ip6_fraglist_iter iter; |
3d13008e | 868 | struct sk_buff *frag2; |
1da177e4 LT |
869 | |
870 | if (first_len - hlen > mtu || | |
871 | ((first_len - hlen) & 7) || | |
1d325d21 FW |
872 | skb_cloned(skb) || |
873 | skb_headroom(skb) < (hroom + sizeof(struct frag_hdr))) | |
1da177e4 LT |
874 | goto slow_path; |
875 | ||
4d9092bb | 876 | skb_walk_frags(skb, frag) { |
1da177e4 LT |
877 | /* Correct geometry. */ |
878 | if (frag->len > mtu || | |
879 | ((frag->len & 7) && frag->next) || | |
1d325d21 | 880 | skb_headroom(frag) < (hlen + hroom + sizeof(struct frag_hdr))) |
3d13008e | 881 | goto slow_path_clean; |
1da177e4 | 882 | |
1da177e4 LT |
883 | /* Partially cloned skb? */ |
884 | if (skb_shared(frag)) | |
3d13008e | 885 | goto slow_path_clean; |
2fdba6b0 HX |
886 | |
887 | BUG_ON(frag->sk); | |
888 | if (skb->sk) { | |
2fdba6b0 HX |
889 | frag->sk = skb->sk; |
890 | frag->destructor = sock_wfree; | |
2fdba6b0 | 891 | } |
3d13008e | 892 | skb->truesize -= frag->truesize; |
1da177e4 LT |
893 | } |
894 | ||
0feca619 PNA |
895 | err = ip6_fraglist_init(skb, hlen, prevhdr, nexthdr, frag_id, |
896 | &iter); | |
897 | if (err < 0) | |
1d325d21 | 898 | goto fail; |
a11d206d | 899 | |
1da177e4 LT |
900 | for (;;) { |
901 | /* Prepare header of the next frame, | |
902 | * before previous one went down. */ | |
0feca619 PNA |
903 | if (iter.frag) |
904 | ip6_fraglist_prepare(skb, &iter); | |
1ab1457c | 905 | |
9669fffc | 906 | skb->tstamp = tstamp; |
7d8c6e39 | 907 | err = output(net, sk, skb); |
67ba4152 | 908 | if (!err) |
d8d1f30b | 909 | IP6_INC_STATS(net, ip6_dst_idev(&rt->dst), |
3bd653c8 | 910 | IPSTATS_MIB_FRAGCREATES); |
dafee490 | 911 | |
0feca619 | 912 | if (err || !iter.frag) |
1da177e4 LT |
913 | break; |
914 | ||
0feca619 | 915 | skb = ip6_fraglist_next(&iter); |
1da177e4 LT |
916 | } |
917 | ||
0feca619 | 918 | kfree(iter.tmp_hdr); |
1da177e4 LT |
919 | |
920 | if (err == 0) { | |
d8d1f30b | 921 | IP6_INC_STATS(net, ip6_dst_idev(&rt->dst), |
3bd653c8 | 922 | IPSTATS_MIB_FRAGOKS); |
1da177e4 LT |
923 | return 0; |
924 | } | |
925 | ||
b7034146 | 926 | kfree_skb_list(iter.frag); |
1da177e4 | 927 | |
d8d1f30b | 928 | IP6_INC_STATS(net, ip6_dst_idev(&rt->dst), |
3bd653c8 | 929 | IPSTATS_MIB_FRAGFAILS); |
1da177e4 | 930 | return err; |
3d13008e ED |
931 | |
932 | slow_path_clean: | |
933 | skb_walk_frags(skb, frag2) { | |
934 | if (frag2 == frag) | |
935 | break; | |
936 | frag2->sk = NULL; | |
937 | frag2->destructor = NULL; | |
938 | skb->truesize += frag2->truesize; | |
939 | } | |
1da177e4 LT |
940 | } |
941 | ||
942 | slow_path: | |
1da177e4 LT |
943 | /* |
944 | * Fragment the datagram. | |
945 | */ | |
946 | ||
8a6a1f17 PNA |
947 | ip6_frag_init(skb, hlen, mtu, rt->dst.dev->needed_tailroom, |
948 | LL_RESERVED_SPACE(rt->dst.dev), prevhdr, nexthdr, frag_id, | |
949 | &state); | |
1da177e4 LT |
950 | |
951 | /* | |
952 | * Keep copying data until we run out. | |
953 | */ | |
1da177e4 | 954 | |
8a6a1f17 PNA |
955 | while (state.left > 0) { |
956 | frag = ip6_frag_next(skb, &state); | |
957 | if (IS_ERR(frag)) { | |
958 | err = PTR_ERR(frag); | |
1da177e4 LT |
959 | goto fail; |
960 | } | |
961 | ||
1da177e4 LT |
962 | /* |
963 | * Put this fragment into the sending queue. | |
964 | */ | |
9669fffc | 965 | frag->tstamp = tstamp; |
7d8c6e39 | 966 | err = output(net, sk, frag); |
1da177e4 LT |
967 | if (err) |
968 | goto fail; | |
dafee490 | 969 | |
adf30907 | 970 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
3bd653c8 | 971 | IPSTATS_MIB_FRAGCREATES); |
1da177e4 | 972 | } |
adf30907 | 973 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 974 | IPSTATS_MIB_FRAGOKS); |
808db80a | 975 | consume_skb(skb); |
1da177e4 LT |
976 | return err; |
977 | ||
485fca66 FW |
978 | fail_toobig: |
979 | if (skb->sk && dst_allfrag(skb_dst(skb))) | |
aba54656 | 980 | sk_gso_disable(skb->sk); |
485fca66 | 981 | |
485fca66 FW |
982 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); |
983 | err = -EMSGSIZE; | |
984 | ||
1da177e4 | 985 | fail: |
adf30907 | 986 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 987 | IPSTATS_MIB_FRAGFAILS); |
1ab1457c | 988 | kfree_skb(skb); |
1da177e4 LT |
989 | return err; |
990 | } | |
991 | ||
b71d1d42 ED |
992 | static inline int ip6_rt_check(const struct rt6key *rt_key, |
993 | const struct in6_addr *fl_addr, | |
994 | const struct in6_addr *addr_cache) | |
cf6b1982 | 995 | { |
a02cec21 | 996 | return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) && |
63159f29 | 997 | (!addr_cache || !ipv6_addr_equal(fl_addr, addr_cache)); |
cf6b1982 YH |
998 | } |
999 | ||
497c615a HX |
1000 | static struct dst_entry *ip6_sk_dst_check(struct sock *sk, |
1001 | struct dst_entry *dst, | |
b71d1d42 | 1002 | const struct flowi6 *fl6) |
1da177e4 | 1003 | { |
497c615a | 1004 | struct ipv6_pinfo *np = inet6_sk(sk); |
a963a37d | 1005 | struct rt6_info *rt; |
1da177e4 | 1006 | |
497c615a HX |
1007 | if (!dst) |
1008 | goto out; | |
1009 | ||
a963a37d ED |
1010 | if (dst->ops->family != AF_INET6) { |
1011 | dst_release(dst); | |
1012 | return NULL; | |
1013 | } | |
1014 | ||
1015 | rt = (struct rt6_info *)dst; | |
497c615a HX |
1016 | /* Yes, checking route validity in not connected |
1017 | * case is not very simple. Take into account, | |
1018 | * that we do not support routing by source, TOS, | |
67ba4152 | 1019 | * and MSG_DONTROUTE --ANK (980726) |
497c615a | 1020 | * |
cf6b1982 YH |
1021 | * 1. ip6_rt_check(): If route was host route, |
1022 | * check that cached destination is current. | |
497c615a HX |
1023 | * If it is network route, we still may |
1024 | * check its validity using saved pointer | |
1025 | * to the last used address: daddr_cache. | |
1026 | * We do not want to save whole address now, | |
1027 | * (because main consumer of this service | |
1028 | * is tcp, which has not this problem), | |
1029 | * so that the last trick works only on connected | |
1030 | * sockets. | |
1031 | * 2. oif also should be the same. | |
1032 | */ | |
4c9483b2 | 1033 | if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) || |
8e1ef0a9 | 1034 | #ifdef CONFIG_IPV6_SUBTREES |
4c9483b2 | 1035 | ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) || |
8e1ef0a9 | 1036 | #endif |
ca254490 DA |
1037 | (!(fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF) && |
1038 | (fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex))) { | |
497c615a HX |
1039 | dst_release(dst); |
1040 | dst = NULL; | |
1da177e4 LT |
1041 | } |
1042 | ||
497c615a HX |
1043 | out: |
1044 | return dst; | |
1045 | } | |
1046 | ||
3aef934f | 1047 | static int ip6_dst_lookup_tail(struct net *net, const struct sock *sk, |
4c9483b2 | 1048 | struct dst_entry **dst, struct flowi6 *fl6) |
497c615a | 1049 | { |
69cce1d1 DM |
1050 | #ifdef CONFIG_IPV6_OPTIMISTIC_DAD |
1051 | struct neighbour *n; | |
97cac082 | 1052 | struct rt6_info *rt; |
69cce1d1 DM |
1053 | #endif |
1054 | int err; | |
6f21c96a | 1055 | int flags = 0; |
497c615a | 1056 | |
e16e888b MS |
1057 | /* The correct way to handle this would be to do |
1058 | * ip6_route_get_saddr, and then ip6_route_output; however, | |
1059 | * the route-specific preferred source forces the | |
1060 | * ip6_route_output call _before_ ip6_route_get_saddr. | |
1061 | * | |
1062 | * In source specific routing (no src=any default route), | |
1063 | * ip6_route_output will fail given src=any saddr, though, so | |
1064 | * that's why we try it again later. | |
1065 | */ | |
c305b9e6 | 1066 | if (ipv6_addr_any(&fl6->saddr)) { |
a68886a6 | 1067 | struct fib6_info *from; |
e16e888b | 1068 | struct rt6_info *rt; |
1da177e4 | 1069 | |
c305b9e6 | 1070 | *dst = ip6_route_output(net, sk, fl6); |
e16e888b | 1071 | rt = (*dst)->error ? NULL : (struct rt6_info *)*dst; |
a68886a6 DA |
1072 | |
1073 | rcu_read_lock(); | |
1074 | from = rt ? rcu_dereference(rt->from) : NULL; | |
1075 | err = ip6_route_get_saddr(net, from, &fl6->daddr, | |
c3968a85 DW |
1076 | sk ? inet6_sk(sk)->srcprefs : 0, |
1077 | &fl6->saddr); | |
a68886a6 DA |
1078 | rcu_read_unlock(); |
1079 | ||
44456d37 | 1080 | if (err) |
1da177e4 | 1081 | goto out_err_release; |
e16e888b MS |
1082 | |
1083 | /* If we had an erroneous initial result, pretend it | |
1084 | * never existed and let the SA-enabled version take | |
1085 | * over. | |
1086 | */ | |
c305b9e6 | 1087 | if ((*dst)->error) { |
e16e888b MS |
1088 | dst_release(*dst); |
1089 | *dst = NULL; | |
1090 | } | |
6f21c96a PA |
1091 | |
1092 | if (fl6->flowi6_oif) | |
1093 | flags |= RT6_LOOKUP_F_IFACE; | |
1da177e4 LT |
1094 | } |
1095 | ||
e16e888b | 1096 | if (!*dst) |
6f21c96a | 1097 | *dst = ip6_route_output_flags(net, sk, fl6, flags); |
e16e888b MS |
1098 | |
1099 | err = (*dst)->error; | |
1100 | if (err) | |
1101 | goto out_err_release; | |
1102 | ||
95c385b4 | 1103 | #ifdef CONFIG_IPV6_OPTIMISTIC_DAD |
e550dfb0 NH |
1104 | /* |
1105 | * Here if the dst entry we've looked up | |
1106 | * has a neighbour entry that is in the INCOMPLETE | |
1107 | * state and the src address from the flow is | |
1108 | * marked as OPTIMISTIC, we release the found | |
1109 | * dst entry and replace it instead with the | |
1110 | * dst entry of the nexthop router | |
1111 | */ | |
c56bf6fe | 1112 | rt = (struct rt6_info *) *dst; |
707be1ff | 1113 | rcu_read_lock_bh(); |
2647a9b0 MKL |
1114 | n = __ipv6_neigh_lookup_noref(rt->dst.dev, |
1115 | rt6_nexthop(rt, &fl6->daddr)); | |
707be1ff YH |
1116 | err = n && !(n->nud_state & NUD_VALID) ? -EINVAL : 0; |
1117 | rcu_read_unlock_bh(); | |
1118 | ||
1119 | if (err) { | |
e550dfb0 | 1120 | struct inet6_ifaddr *ifp; |
4c9483b2 | 1121 | struct flowi6 fl_gw6; |
e550dfb0 NH |
1122 | int redirect; |
1123 | ||
4c9483b2 | 1124 | ifp = ipv6_get_ifaddr(net, &fl6->saddr, |
e550dfb0 NH |
1125 | (*dst)->dev, 1); |
1126 | ||
1127 | redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC); | |
1128 | if (ifp) | |
1129 | in6_ifa_put(ifp); | |
1130 | ||
1131 | if (redirect) { | |
1132 | /* | |
1133 | * We need to get the dst entry for the | |
1134 | * default router instead | |
1135 | */ | |
1136 | dst_release(*dst); | |
4c9483b2 DM |
1137 | memcpy(&fl_gw6, fl6, sizeof(struct flowi6)); |
1138 | memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr)); | |
1139 | *dst = ip6_route_output(net, sk, &fl_gw6); | |
e5d08d71 IM |
1140 | err = (*dst)->error; |
1141 | if (err) | |
e550dfb0 | 1142 | goto out_err_release; |
95c385b4 | 1143 | } |
e550dfb0 | 1144 | } |
95c385b4 | 1145 | #endif |
ec5e3b0a | 1146 | if (ipv6_addr_v4mapped(&fl6->saddr) && |
00ea1cee WB |
1147 | !(ipv6_addr_v4mapped(&fl6->daddr) || ipv6_addr_any(&fl6->daddr))) { |
1148 | err = -EAFNOSUPPORT; | |
1149 | goto out_err_release; | |
1150 | } | |
95c385b4 | 1151 | |
1da177e4 LT |
1152 | return 0; |
1153 | ||
1154 | out_err_release: | |
1155 | dst_release(*dst); | |
1156 | *dst = NULL; | |
8a966fc0 | 1157 | |
0d240e78 DA |
1158 | if (err == -ENETUNREACH) |
1159 | IP6_INC_STATS(net, NULL, IPSTATS_MIB_OUTNOROUTES); | |
1da177e4 LT |
1160 | return err; |
1161 | } | |
34a0b3cd | 1162 | |
497c615a HX |
1163 | /** |
1164 | * ip6_dst_lookup - perform route lookup on flow | |
b51cd7c8 | 1165 | * @net: Network namespace to perform lookup in |
497c615a HX |
1166 | * @sk: socket which provides route info |
1167 | * @dst: pointer to dst_entry * for result | |
4c9483b2 | 1168 | * @fl6: flow to lookup |
497c615a HX |
1169 | * |
1170 | * This function performs a route lookup on the given flow. | |
1171 | * | |
1172 | * It returns zero on success, or a standard errno code on error. | |
1173 | */ | |
343d60aa RP |
1174 | int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst, |
1175 | struct flowi6 *fl6) | |
497c615a HX |
1176 | { |
1177 | *dst = NULL; | |
343d60aa | 1178 | return ip6_dst_lookup_tail(net, sk, dst, fl6); |
497c615a | 1179 | } |
3cf3dc6c ACM |
1180 | EXPORT_SYMBOL_GPL(ip6_dst_lookup); |
1181 | ||
497c615a | 1182 | /** |
68d0c6d3 | 1183 | * ip6_dst_lookup_flow - perform route lookup on flow with ipsec |
b51cd7c8 | 1184 | * @net: Network namespace to perform lookup in |
68d0c6d3 | 1185 | * @sk: socket which provides route info |
4c9483b2 | 1186 | * @fl6: flow to lookup |
68d0c6d3 | 1187 | * @final_dst: final destination address for ipsec lookup |
68d0c6d3 DM |
1188 | * |
1189 | * This function performs a route lookup on the given flow. | |
1190 | * | |
1191 | * It returns a valid dst pointer on success, or a pointer encoded | |
1192 | * error code. | |
1193 | */ | |
c4e85f73 | 1194 | struct dst_entry *ip6_dst_lookup_flow(struct net *net, const struct sock *sk, struct flowi6 *fl6, |
0e0d44ab | 1195 | const struct in6_addr *final_dst) |
68d0c6d3 DM |
1196 | { |
1197 | struct dst_entry *dst = NULL; | |
1198 | int err; | |
1199 | ||
c4e85f73 | 1200 | err = ip6_dst_lookup_tail(net, sk, &dst, fl6); |
68d0c6d3 DM |
1201 | if (err) |
1202 | return ERR_PTR(err); | |
1203 | if (final_dst) | |
4e3fd7a0 | 1204 | fl6->daddr = *final_dst; |
2774c131 | 1205 | |
c4e85f73 | 1206 | return xfrm_lookup_route(net, dst, flowi6_to_flowi(fl6), sk, 0); |
68d0c6d3 DM |
1207 | } |
1208 | EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow); | |
1209 | ||
1210 | /** | |
1211 | * ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow | |
497c615a | 1212 | * @sk: socket which provides the dst cache and route info |
4c9483b2 | 1213 | * @fl6: flow to lookup |
68d0c6d3 | 1214 | * @final_dst: final destination address for ipsec lookup |
96818159 | 1215 | * @connected: whether @sk is connected or not |
497c615a HX |
1216 | * |
1217 | * This function performs a route lookup on the given flow with the | |
1218 | * possibility of using the cached route in the socket if it is valid. | |
1219 | * It will take the socket dst lock when operating on the dst cache. | |
1220 | * As a result, this function can only be used in process context. | |
1221 | * | |
96818159 AK |
1222 | * In addition, for a connected socket, cache the dst in the socket |
1223 | * if the current cache is not valid. | |
1224 | * | |
68d0c6d3 DM |
1225 | * It returns a valid dst pointer on success, or a pointer encoded |
1226 | * error code. | |
497c615a | 1227 | */ |
4c9483b2 | 1228 | struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6, |
96818159 AK |
1229 | const struct in6_addr *final_dst, |
1230 | bool connected) | |
497c615a | 1231 | { |
68d0c6d3 | 1232 | struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie); |
497c615a | 1233 | |
4c9483b2 | 1234 | dst = ip6_sk_dst_check(sk, dst, fl6); |
96818159 AK |
1235 | if (dst) |
1236 | return dst; | |
1237 | ||
c4e85f73 | 1238 | dst = ip6_dst_lookup_flow(sock_net(sk), sk, fl6, final_dst); |
96818159 AK |
1239 | if (connected && !IS_ERR(dst)) |
1240 | ip6_sk_dst_store_flow(sk, dst_clone(dst), fl6); | |
68d0c6d3 | 1241 | |
00bc0ef5 | 1242 | return dst; |
497c615a | 1243 | } |
68d0c6d3 | 1244 | EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow); |
497c615a | 1245 | |
571912c6 MV |
1246 | /** |
1247 | * ip6_dst_lookup_tunnel - perform route lookup on tunnel | |
1248 | * @skb: Packet for which lookup is done | |
1249 | * @dev: Tunnel device | |
1250 | * @net: Network namespace of tunnel device | |
b51cd7c8 | 1251 | * @sock: Socket which provides route info |
571912c6 MV |
1252 | * @saddr: Memory to store the src ip address |
1253 | * @info: Tunnel information | |
1254 | * @protocol: IP protocol | |
b51cd7c8 | 1255 | * @use_cache: Flag to enable cache usage |
571912c6 MV |
1256 | * This function performs a route lookup on a tunnel |
1257 | * | |
1258 | * It returns a valid dst pointer and stores src address to be used in | |
1259 | * tunnel in param saddr on success, else a pointer encoded error code. | |
1260 | */ | |
1261 | ||
1262 | struct dst_entry *ip6_dst_lookup_tunnel(struct sk_buff *skb, | |
1263 | struct net_device *dev, | |
1264 | struct net *net, | |
1265 | struct socket *sock, | |
1266 | struct in6_addr *saddr, | |
1267 | const struct ip_tunnel_info *info, | |
1268 | u8 protocol, | |
1269 | bool use_cache) | |
1270 | { | |
1271 | struct dst_entry *dst = NULL; | |
1272 | #ifdef CONFIG_DST_CACHE | |
1273 | struct dst_cache *dst_cache; | |
1274 | #endif | |
1275 | struct flowi6 fl6; | |
1276 | __u8 prio; | |
1277 | ||
1278 | #ifdef CONFIG_DST_CACHE | |
1279 | dst_cache = (struct dst_cache *)&info->dst_cache; | |
1280 | if (use_cache) { | |
1281 | dst = dst_cache_get_ip6(dst_cache, saddr); | |
1282 | if (dst) | |
1283 | return dst; | |
1284 | } | |
1285 | #endif | |
1286 | memset(&fl6, 0, sizeof(fl6)); | |
1287 | fl6.flowi6_mark = skb->mark; | |
1288 | fl6.flowi6_proto = protocol; | |
1289 | fl6.daddr = info->key.u.ipv6.dst; | |
1290 | fl6.saddr = info->key.u.ipv6.src; | |
1291 | prio = info->key.tos; | |
1292 | fl6.flowlabel = ip6_make_flowinfo(RT_TOS(prio), | |
1293 | info->key.label); | |
1294 | ||
1295 | dst = ipv6_stub->ipv6_dst_lookup_flow(net, sock->sk, &fl6, | |
1296 | NULL); | |
1297 | if (IS_ERR(dst)) { | |
1298 | netdev_dbg(dev, "no route to %pI6\n", &fl6.daddr); | |
1299 | return ERR_PTR(-ENETUNREACH); | |
1300 | } | |
1301 | if (dst->dev == dev) { /* is this necessary? */ | |
1302 | netdev_dbg(dev, "circular route to %pI6\n", &fl6.daddr); | |
1303 | dst_release(dst); | |
1304 | return ERR_PTR(-ELOOP); | |
1305 | } | |
1306 | #ifdef CONFIG_DST_CACHE | |
1307 | if (use_cache) | |
1308 | dst_cache_set_ip6(dst_cache, dst, &fl6.saddr); | |
1309 | #endif | |
1310 | *saddr = fl6.saddr; | |
1311 | return dst; | |
1312 | } | |
1313 | EXPORT_SYMBOL_GPL(ip6_dst_lookup_tunnel); | |
1314 | ||
0178b695 HX |
1315 | static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src, |
1316 | gfp_t gfp) | |
1317 | { | |
1318 | return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL; | |
1319 | } | |
1320 | ||
1321 | static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src, | |
1322 | gfp_t gfp) | |
1323 | { | |
1324 | return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL; | |
1325 | } | |
1326 | ||
75a493e6 | 1327 | static void ip6_append_data_mtu(unsigned int *mtu, |
0c183379 G |
1328 | int *maxfraglen, |
1329 | unsigned int fragheaderlen, | |
1330 | struct sk_buff *skb, | |
75a493e6 | 1331 | struct rt6_info *rt, |
e367c2d0 | 1332 | unsigned int orig_mtu) |
0c183379 G |
1333 | { |
1334 | if (!(rt->dst.flags & DST_XFRM_TUNNEL)) { | |
63159f29 | 1335 | if (!skb) { |
0c183379 | 1336 | /* first fragment, reserve header_len */ |
e367c2d0 | 1337 | *mtu = orig_mtu - rt->dst.header_len; |
0c183379 G |
1338 | |
1339 | } else { | |
1340 | /* | |
1341 | * this fragment is not first, the headers | |
1342 | * space is regarded as data space. | |
1343 | */ | |
e367c2d0 | 1344 | *mtu = orig_mtu; |
0c183379 G |
1345 | } |
1346 | *maxfraglen = ((*mtu - fragheaderlen) & ~7) | |
1347 | + fragheaderlen - sizeof(struct frag_hdr); | |
1348 | } | |
1349 | } | |
1350 | ||
366e41d9 | 1351 | static int ip6_setup_cork(struct sock *sk, struct inet_cork_full *cork, |
26879da5 | 1352 | struct inet6_cork *v6_cork, struct ipcm6_cookie *ipc6, |
5fdaa88d | 1353 | struct rt6_info *rt, struct flowi6 *fl6) |
366e41d9 VY |
1354 | { |
1355 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1356 | unsigned int mtu; | |
26879da5 | 1357 | struct ipv6_txoptions *opt = ipc6->opt; |
366e41d9 VY |
1358 | |
1359 | /* | |
1360 | * setup for corking | |
1361 | */ | |
1362 | if (opt) { | |
1363 | if (WARN_ON(v6_cork->opt)) | |
1364 | return -EINVAL; | |
1365 | ||
864e2a1f | 1366 | v6_cork->opt = kzalloc(sizeof(*opt), sk->sk_allocation); |
63159f29 | 1367 | if (unlikely(!v6_cork->opt)) |
366e41d9 VY |
1368 | return -ENOBUFS; |
1369 | ||
864e2a1f | 1370 | v6_cork->opt->tot_len = sizeof(*opt); |
366e41d9 VY |
1371 | v6_cork->opt->opt_flen = opt->opt_flen; |
1372 | v6_cork->opt->opt_nflen = opt->opt_nflen; | |
1373 | ||
1374 | v6_cork->opt->dst0opt = ip6_opt_dup(opt->dst0opt, | |
1375 | sk->sk_allocation); | |
1376 | if (opt->dst0opt && !v6_cork->opt->dst0opt) | |
1377 | return -ENOBUFS; | |
1378 | ||
1379 | v6_cork->opt->dst1opt = ip6_opt_dup(opt->dst1opt, | |
1380 | sk->sk_allocation); | |
1381 | if (opt->dst1opt && !v6_cork->opt->dst1opt) | |
1382 | return -ENOBUFS; | |
1383 | ||
1384 | v6_cork->opt->hopopt = ip6_opt_dup(opt->hopopt, | |
1385 | sk->sk_allocation); | |
1386 | if (opt->hopopt && !v6_cork->opt->hopopt) | |
1387 | return -ENOBUFS; | |
1388 | ||
1389 | v6_cork->opt->srcrt = ip6_rthdr_dup(opt->srcrt, | |
1390 | sk->sk_allocation); | |
1391 | if (opt->srcrt && !v6_cork->opt->srcrt) | |
1392 | return -ENOBUFS; | |
1393 | ||
1394 | /* need source address above miyazawa*/ | |
1395 | } | |
1396 | dst_hold(&rt->dst); | |
1397 | cork->base.dst = &rt->dst; | |
1398 | cork->fl.u.ip6 = *fl6; | |
26879da5 WW |
1399 | v6_cork->hop_limit = ipc6->hlimit; |
1400 | v6_cork->tclass = ipc6->tclass; | |
366e41d9 VY |
1401 | if (rt->dst.flags & DST_XFRM_TUNNEL) |
1402 | mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ? | |
749439bf | 1403 | READ_ONCE(rt->dst.dev->mtu) : dst_mtu(&rt->dst); |
366e41d9 VY |
1404 | else |
1405 | mtu = np->pmtudisc >= IPV6_PMTUDISC_PROBE ? | |
c02b3741 | 1406 | READ_ONCE(rt->dst.dev->mtu) : dst_mtu(xfrm_dst_path(&rt->dst)); |
366e41d9 VY |
1407 | if (np->frag_size < mtu) { |
1408 | if (np->frag_size) | |
1409 | mtu = np->frag_size; | |
1410 | } | |
749439bf MM |
1411 | if (mtu < IPV6_MIN_MTU) |
1412 | return -EINVAL; | |
366e41d9 | 1413 | cork->base.fragsize = mtu; |
fbf47813 | 1414 | cork->base.gso_size = ipc6->gso_size; |
678ca42d | 1415 | cork->base.tx_flags = 0; |
c6af0c22 | 1416 | cork->base.mark = ipc6->sockc.mark; |
678ca42d | 1417 | sock_tx_timestamp(sk, ipc6->sockc.tsflags, &cork->base.tx_flags); |
bec1f6f6 | 1418 | |
0f6c480f | 1419 | if (dst_allfrag(xfrm_dst_path(&rt->dst))) |
366e41d9 VY |
1420 | cork->base.flags |= IPCORK_ALLFRAG; |
1421 | cork->base.length = 0; | |
1422 | ||
5fdaa88d | 1423 | cork->base.transmit_time = ipc6->sockc.transmit_time; |
a818f75e | 1424 | |
366e41d9 VY |
1425 | return 0; |
1426 | } | |
1427 | ||
0bbe84a6 VY |
1428 | static int __ip6_append_data(struct sock *sk, |
1429 | struct flowi6 *fl6, | |
1430 | struct sk_buff_head *queue, | |
1431 | struct inet_cork *cork, | |
1432 | struct inet6_cork *v6_cork, | |
1433 | struct page_frag *pfrag, | |
1434 | int getfrag(void *from, char *to, int offset, | |
1435 | int len, int odd, struct sk_buff *skb), | |
1436 | void *from, int length, int transhdrlen, | |
5fdaa88d | 1437 | unsigned int flags, struct ipcm6_cookie *ipc6) |
1da177e4 | 1438 | { |
0c183379 | 1439 | struct sk_buff *skb, *skb_prev = NULL; |
10b8a3de | 1440 | unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu, pmtu; |
b5947e5d | 1441 | struct ubuf_info *uarg = NULL; |
0bbe84a6 VY |
1442 | int exthdrlen = 0; |
1443 | int dst_exthdrlen = 0; | |
1da177e4 | 1444 | int hh_len; |
1da177e4 LT |
1445 | int copy; |
1446 | int err; | |
1447 | int offset = 0; | |
09c2d251 | 1448 | u32 tskey = 0; |
0bbe84a6 VY |
1449 | struct rt6_info *rt = (struct rt6_info *)cork->dst; |
1450 | struct ipv6_txoptions *opt = v6_cork->opt; | |
32dce968 | 1451 | int csummode = CHECKSUM_NONE; |
682b1a9d | 1452 | unsigned int maxnonfragsize, headersize; |
1f4c6eb2 | 1453 | unsigned int wmem_alloc_delta = 0; |
100f6d8e | 1454 | bool paged, extra_uref = false; |
1da177e4 | 1455 | |
0bbe84a6 VY |
1456 | skb = skb_peek_tail(queue); |
1457 | if (!skb) { | |
1458 | exthdrlen = opt ? opt->opt_flen : 0; | |
7efdba5b | 1459 | dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len; |
1da177e4 | 1460 | } |
0bbe84a6 | 1461 | |
15e36f5b | 1462 | paged = !!cork->gso_size; |
bec1f6f6 | 1463 | mtu = cork->gso_size ? IP6_MAX_MTU : cork->fragsize; |
e367c2d0 | 1464 | orig_mtu = mtu; |
1da177e4 | 1465 | |
678ca42d WB |
1466 | if (cork->tx_flags & SKBTX_ANY_SW_TSTAMP && |
1467 | sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID) | |
1468 | tskey = sk->sk_tskey++; | |
1469 | ||
d8d1f30b | 1470 | hh_len = LL_RESERVED_SPACE(rt->dst.dev); |
1da177e4 | 1471 | |
a1b05140 | 1472 | fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len + |
b4ce9277 | 1473 | (opt ? opt->opt_nflen : 0); |
4df98e76 HFS |
1474 | maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - |
1475 | sizeof(struct frag_hdr); | |
1da177e4 | 1476 | |
682b1a9d HFS |
1477 | headersize = sizeof(struct ipv6hdr) + |
1478 | (opt ? opt->opt_flen + opt->opt_nflen : 0) + | |
1479 | (dst_allfrag(&rt->dst) ? | |
1480 | sizeof(struct frag_hdr) : 0) + | |
1481 | rt->rt6i_nfheader_len; | |
1482 | ||
10b8a3de PA |
1483 | /* as per RFC 7112 section 5, the entire IPv6 Header Chain must fit |
1484 | * the first fragment | |
1485 | */ | |
1486 | if (headersize + transhdrlen > mtu) | |
1487 | goto emsgsize; | |
1488 | ||
26879da5 | 1489 | if (cork->length + length > mtu - headersize && ipc6->dontfrag && |
682b1a9d HFS |
1490 | (sk->sk_protocol == IPPROTO_UDP || |
1491 | sk->sk_protocol == IPPROTO_RAW)) { | |
1492 | ipv6_local_rxpmtu(sk, fl6, mtu - headersize + | |
1493 | sizeof(struct ipv6hdr)); | |
1494 | goto emsgsize; | |
1495 | } | |
4df98e76 | 1496 | |
682b1a9d HFS |
1497 | if (ip6_sk_ignore_df(sk)) |
1498 | maxnonfragsize = sizeof(struct ipv6hdr) + IPV6_MAXPLEN; | |
1499 | else | |
1500 | maxnonfragsize = mtu; | |
4df98e76 | 1501 | |
682b1a9d | 1502 | if (cork->length + length > maxnonfragsize - headersize) { |
4df98e76 | 1503 | emsgsize: |
10b8a3de PA |
1504 | pmtu = max_t(int, mtu - headersize + sizeof(struct ipv6hdr), 0); |
1505 | ipv6_local_error(sk, EMSGSIZE, fl6, pmtu); | |
682b1a9d | 1506 | return -EMSGSIZE; |
1da177e4 LT |
1507 | } |
1508 | ||
682b1a9d HFS |
1509 | /* CHECKSUM_PARTIAL only with no extension headers and when |
1510 | * we are not going to fragment | |
1511 | */ | |
1512 | if (transhdrlen && sk->sk_protocol == IPPROTO_UDP && | |
1513 | headersize == sizeof(struct ipv6hdr) && | |
2b89ed65 | 1514 | length <= mtu - headersize && |
bec1f6f6 | 1515 | (!(flags & MSG_MORE) || cork->gso_size) && |
c8cd0989 | 1516 | rt->dst.dev->features & (NETIF_F_IPV6_CSUM | NETIF_F_HW_CSUM)) |
682b1a9d HFS |
1517 | csummode = CHECKSUM_PARTIAL; |
1518 | ||
b5947e5d | 1519 | if (flags & MSG_ZEROCOPY && length && sock_flag(sk, SOCK_ZEROCOPY)) { |
8c793822 | 1520 | uarg = msg_zerocopy_realloc(sk, length, skb_zcopy(skb)); |
b5947e5d WB |
1521 | if (!uarg) |
1522 | return -ENOBUFS; | |
522924b5 | 1523 | extra_uref = !skb_zcopy(skb); /* only ref on new uarg */ |
b5947e5d WB |
1524 | if (rt->dst.dev->features & NETIF_F_SG && |
1525 | csummode == CHECKSUM_PARTIAL) { | |
1526 | paged = true; | |
1527 | } else { | |
1528 | uarg->zerocopy = 0; | |
52900d22 | 1529 | skb_zcopy_set(skb, uarg, &extra_uref); |
b5947e5d WB |
1530 | } |
1531 | } | |
1532 | ||
1da177e4 LT |
1533 | /* |
1534 | * Let's try using as much space as possible. | |
1535 | * Use MTU if total length of the message fits into the MTU. | |
1536 | * Otherwise, we need to reserve fragment header and | |
1537 | * fragment alignment (= 8-15 octects, in total). | |
1538 | * | |
634a63e7 | 1539 | * Note that we may need to "move" the data from the tail |
1ab1457c | 1540 | * of the buffer to the new fragment when we split |
1da177e4 LT |
1541 | * the message. |
1542 | * | |
1ab1457c | 1543 | * FIXME: It may be fragmented into multiple chunks |
1da177e4 LT |
1544 | * at once if non-fragmentable extension headers |
1545 | * are too large. | |
1ab1457c | 1546 | * --yoshfuji |
1da177e4 LT |
1547 | */ |
1548 | ||
2811ebac | 1549 | cork->length += length; |
2811ebac | 1550 | if (!skb) |
1da177e4 LT |
1551 | goto alloc_new_skb; |
1552 | ||
1553 | while (length > 0) { | |
1554 | /* Check if the remaining data fits into current packet. */ | |
bdc712b4 | 1555 | copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len; |
1da177e4 LT |
1556 | if (copy < length) |
1557 | copy = maxfraglen - skb->len; | |
1558 | ||
1559 | if (copy <= 0) { | |
1560 | char *data; | |
1561 | unsigned int datalen; | |
1562 | unsigned int fraglen; | |
1563 | unsigned int fraggap; | |
6d123b81 | 1564 | unsigned int alloclen, alloc_extra; |
aba36930 | 1565 | unsigned int pagedlen; |
1da177e4 | 1566 | alloc_new_skb: |
1da177e4 | 1567 | /* There's no room in the current skb */ |
0c183379 G |
1568 | if (skb) |
1569 | fraggap = skb->len - maxfraglen; | |
1da177e4 LT |
1570 | else |
1571 | fraggap = 0; | |
0c183379 | 1572 | /* update mtu and maxfraglen if necessary */ |
63159f29 | 1573 | if (!skb || !skb_prev) |
0c183379 | 1574 | ip6_append_data_mtu(&mtu, &maxfraglen, |
75a493e6 | 1575 | fragheaderlen, skb, rt, |
e367c2d0 | 1576 | orig_mtu); |
0c183379 G |
1577 | |
1578 | skb_prev = skb; | |
1da177e4 LT |
1579 | |
1580 | /* | |
1581 | * If remaining data exceeds the mtu, | |
1582 | * we know we need more fragment(s). | |
1583 | */ | |
1584 | datalen = length + fraggap; | |
1da177e4 | 1585 | |
0c183379 G |
1586 | if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen) |
1587 | datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len; | |
15e36f5b | 1588 | fraglen = datalen + fragheaderlen; |
aba36930 | 1589 | pagedlen = 0; |
15e36f5b | 1590 | |
6d123b81 JK |
1591 | alloc_extra = hh_len; |
1592 | alloc_extra += dst_exthdrlen; | |
1593 | alloc_extra += rt->dst.trailer_len; | |
1594 | ||
1595 | /* We just reserve space for fragment header. | |
1596 | * Note: this may be overallocation if the message | |
1597 | * (without MSG_MORE) fits into the MTU. | |
1598 | */ | |
1599 | alloc_extra += sizeof(struct frag_hdr); | |
1600 | ||
1da177e4 | 1601 | if ((flags & MSG_MORE) && |
d8d1f30b | 1602 | !(rt->dst.dev->features&NETIF_F_SG)) |
1da177e4 | 1603 | alloclen = mtu; |
6d123b81 JK |
1604 | else if (!paged && |
1605 | (fraglen + alloc_extra < SKB_MAX_ALLOC || | |
1606 | !(rt->dst.dev->features & NETIF_F_SG))) | |
15e36f5b WB |
1607 | alloclen = fraglen; |
1608 | else { | |
1609 | alloclen = min_t(int, fraglen, MAX_HEADER); | |
1610 | pagedlen = fraglen - alloclen; | |
1611 | } | |
6d123b81 | 1612 | alloclen += alloc_extra; |
299b0767 | 1613 | |
0c183379 G |
1614 | if (datalen != length + fraggap) { |
1615 | /* | |
1616 | * this is not the last fragment, the trailer | |
1617 | * space is regarded as data space. | |
1618 | */ | |
1619 | datalen += rt->dst.trailer_len; | |
1620 | } | |
1621 | ||
0c183379 | 1622 | fraglen = datalen + fragheaderlen; |
1da177e4 | 1623 | |
15e36f5b | 1624 | copy = datalen - transhdrlen - fraggap - pagedlen; |
232cd35d ED |
1625 | if (copy < 0) { |
1626 | err = -EINVAL; | |
1627 | goto error; | |
1628 | } | |
1da177e4 | 1629 | if (transhdrlen) { |
6d123b81 | 1630 | skb = sock_alloc_send_skb(sk, alloclen, |
1da177e4 LT |
1631 | (flags & MSG_DONTWAIT), &err); |
1632 | } else { | |
1633 | skb = NULL; | |
1f4c6eb2 | 1634 | if (refcount_read(&sk->sk_wmem_alloc) + wmem_alloc_delta <= |
1da177e4 | 1635 | 2 * sk->sk_sndbuf) |
6d123b81 | 1636 | skb = alloc_skb(alloclen, |
1f4c6eb2 | 1637 | sk->sk_allocation); |
63159f29 | 1638 | if (unlikely(!skb)) |
1da177e4 LT |
1639 | err = -ENOBUFS; |
1640 | } | |
63159f29 | 1641 | if (!skb) |
1da177e4 LT |
1642 | goto error; |
1643 | /* | |
1644 | * Fill in the control structures | |
1645 | */ | |
9c9c9ad5 | 1646 | skb->protocol = htons(ETH_P_IPV6); |
32dce968 | 1647 | skb->ip_summed = csummode; |
1da177e4 | 1648 | skb->csum = 0; |
1f85851e G |
1649 | /* reserve for fragmentation and ipsec header */ |
1650 | skb_reserve(skb, hh_len + sizeof(struct frag_hdr) + | |
1651 | dst_exthdrlen); | |
1da177e4 LT |
1652 | |
1653 | /* | |
1654 | * Find where to start putting bytes | |
1655 | */ | |
15e36f5b | 1656 | data = skb_put(skb, fraglen - pagedlen); |
1f85851e G |
1657 | skb_set_network_header(skb, exthdrlen); |
1658 | data += fragheaderlen; | |
b0e380b1 ACM |
1659 | skb->transport_header = (skb->network_header + |
1660 | fragheaderlen); | |
1da177e4 LT |
1661 | if (fraggap) { |
1662 | skb->csum = skb_copy_and_csum_bits( | |
1663 | skb_prev, maxfraglen, | |
8d5930df | 1664 | data + transhdrlen, fraggap); |
1da177e4 LT |
1665 | skb_prev->csum = csum_sub(skb_prev->csum, |
1666 | skb->csum); | |
1667 | data += fraggap; | |
e9fa4f7b | 1668 | pskb_trim_unique(skb_prev, maxfraglen); |
1da177e4 | 1669 | } |
232cd35d ED |
1670 | if (copy > 0 && |
1671 | getfrag(from, data + transhdrlen, offset, | |
1672 | copy, fraggap, skb) < 0) { | |
1da177e4 LT |
1673 | err = -EFAULT; |
1674 | kfree_skb(skb); | |
1675 | goto error; | |
1676 | } | |
1677 | ||
1678 | offset += copy; | |
15e36f5b | 1679 | length -= copy + transhdrlen; |
1da177e4 LT |
1680 | transhdrlen = 0; |
1681 | exthdrlen = 0; | |
299b0767 | 1682 | dst_exthdrlen = 0; |
1da177e4 | 1683 | |
52900d22 WB |
1684 | /* Only the initial fragment is time stamped */ |
1685 | skb_shinfo(skb)->tx_flags = cork->tx_flags; | |
1686 | cork->tx_flags = 0; | |
1687 | skb_shinfo(skb)->tskey = tskey; | |
1688 | tskey = 0; | |
1689 | skb_zcopy_set(skb, uarg, &extra_uref); | |
1690 | ||
0dec879f JA |
1691 | if ((flags & MSG_CONFIRM) && !skb_prev) |
1692 | skb_set_dst_pending_confirm(skb, 1); | |
1693 | ||
1da177e4 LT |
1694 | /* |
1695 | * Put the packet on the pending queue | |
1696 | */ | |
1f4c6eb2 ED |
1697 | if (!skb->destructor) { |
1698 | skb->destructor = sock_wfree; | |
1699 | skb->sk = sk; | |
1700 | wmem_alloc_delta += skb->truesize; | |
1701 | } | |
0bbe84a6 | 1702 | __skb_queue_tail(queue, skb); |
1da177e4 LT |
1703 | continue; |
1704 | } | |
1705 | ||
1706 | if (copy > length) | |
1707 | copy = length; | |
1708 | ||
113f99c3 WB |
1709 | if (!(rt->dst.dev->features&NETIF_F_SG) && |
1710 | skb_tailroom(skb) >= copy) { | |
1da177e4 LT |
1711 | unsigned int off; |
1712 | ||
1713 | off = skb->len; | |
1714 | if (getfrag(from, skb_put(skb, copy), | |
1715 | offset, copy, off, skb) < 0) { | |
1716 | __skb_trim(skb, off); | |
1717 | err = -EFAULT; | |
1718 | goto error; | |
1719 | } | |
b5947e5d | 1720 | } else if (!uarg || !uarg->zerocopy) { |
1da177e4 | 1721 | int i = skb_shinfo(skb)->nr_frags; |
1da177e4 | 1722 | |
5640f768 ED |
1723 | err = -ENOMEM; |
1724 | if (!sk_page_frag_refill(sk, pfrag)) | |
1da177e4 | 1725 | goto error; |
5640f768 ED |
1726 | |
1727 | if (!skb_can_coalesce(skb, i, pfrag->page, | |
1728 | pfrag->offset)) { | |
1729 | err = -EMSGSIZE; | |
1730 | if (i == MAX_SKB_FRAGS) | |
1731 | goto error; | |
1732 | ||
1733 | __skb_fill_page_desc(skb, i, pfrag->page, | |
1734 | pfrag->offset, 0); | |
1735 | skb_shinfo(skb)->nr_frags = ++i; | |
1736 | get_page(pfrag->page); | |
1da177e4 | 1737 | } |
5640f768 | 1738 | copy = min_t(int, copy, pfrag->size - pfrag->offset); |
9e903e08 | 1739 | if (getfrag(from, |
5640f768 ED |
1740 | page_address(pfrag->page) + pfrag->offset, |
1741 | offset, copy, skb->len, skb) < 0) | |
1742 | goto error_efault; | |
1743 | ||
1744 | pfrag->offset += copy; | |
1745 | skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy); | |
1da177e4 LT |
1746 | skb->len += copy; |
1747 | skb->data_len += copy; | |
f945fa7a | 1748 | skb->truesize += copy; |
1f4c6eb2 | 1749 | wmem_alloc_delta += copy; |
b5947e5d WB |
1750 | } else { |
1751 | err = skb_zerocopy_iter_dgram(skb, from, copy); | |
1752 | if (err < 0) | |
1753 | goto error; | |
1da177e4 LT |
1754 | } |
1755 | offset += copy; | |
1756 | length -= copy; | |
1757 | } | |
5640f768 | 1758 | |
9e8445a5 PA |
1759 | if (wmem_alloc_delta) |
1760 | refcount_add(wmem_alloc_delta, &sk->sk_wmem_alloc); | |
1da177e4 | 1761 | return 0; |
5640f768 ED |
1762 | |
1763 | error_efault: | |
1764 | err = -EFAULT; | |
1da177e4 | 1765 | error: |
8e044917 | 1766 | net_zcopy_put_abort(uarg, extra_uref); |
bdc712b4 | 1767 | cork->length -= length; |
3bd653c8 | 1768 | IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); |
1f4c6eb2 | 1769 | refcount_add(wmem_alloc_delta, &sk->sk_wmem_alloc); |
1da177e4 LT |
1770 | return err; |
1771 | } | |
0bbe84a6 VY |
1772 | |
1773 | int ip6_append_data(struct sock *sk, | |
1774 | int getfrag(void *from, char *to, int offset, int len, | |
1775 | int odd, struct sk_buff *skb), | |
26879da5 WW |
1776 | void *from, int length, int transhdrlen, |
1777 | struct ipcm6_cookie *ipc6, struct flowi6 *fl6, | |
5fdaa88d | 1778 | struct rt6_info *rt, unsigned int flags) |
0bbe84a6 VY |
1779 | { |
1780 | struct inet_sock *inet = inet_sk(sk); | |
1781 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1782 | int exthdrlen; | |
1783 | int err; | |
1784 | ||
1785 | if (flags&MSG_PROBE) | |
1786 | return 0; | |
1787 | if (skb_queue_empty(&sk->sk_write_queue)) { | |
1788 | /* | |
1789 | * setup for corking | |
1790 | */ | |
26879da5 | 1791 | err = ip6_setup_cork(sk, &inet->cork, &np->cork, |
5fdaa88d | 1792 | ipc6, rt, fl6); |
0bbe84a6 VY |
1793 | if (err) |
1794 | return err; | |
1795 | ||
26879da5 | 1796 | exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0); |
0bbe84a6 VY |
1797 | length += exthdrlen; |
1798 | transhdrlen += exthdrlen; | |
1799 | } else { | |
1800 | fl6 = &inet->cork.fl.u.ip6; | |
1801 | transhdrlen = 0; | |
1802 | } | |
1803 | ||
1804 | return __ip6_append_data(sk, fl6, &sk->sk_write_queue, &inet->cork.base, | |
1805 | &np->cork, sk_page_frag(sk), getfrag, | |
5fdaa88d | 1806 | from, length, transhdrlen, flags, ipc6); |
0bbe84a6 | 1807 | } |
a495f836 | 1808 | EXPORT_SYMBOL_GPL(ip6_append_data); |
1da177e4 | 1809 | |
cd3c7480 PB |
1810 | static void ip6_cork_steal_dst(struct sk_buff *skb, struct inet_cork_full *cork) |
1811 | { | |
1812 | struct dst_entry *dst = cork->base.dst; | |
1813 | ||
1814 | cork->base.dst = NULL; | |
1815 | cork->base.flags &= ~IPCORK_ALLFRAG; | |
1816 | skb_dst_set(skb, dst); | |
1817 | } | |
1818 | ||
366e41d9 VY |
1819 | static void ip6_cork_release(struct inet_cork_full *cork, |
1820 | struct inet6_cork *v6_cork) | |
bf138862 | 1821 | { |
366e41d9 VY |
1822 | if (v6_cork->opt) { |
1823 | kfree(v6_cork->opt->dst0opt); | |
1824 | kfree(v6_cork->opt->dst1opt); | |
1825 | kfree(v6_cork->opt->hopopt); | |
1826 | kfree(v6_cork->opt->srcrt); | |
1827 | kfree(v6_cork->opt); | |
1828 | v6_cork->opt = NULL; | |
0178b695 HX |
1829 | } |
1830 | ||
366e41d9 VY |
1831 | if (cork->base.dst) { |
1832 | dst_release(cork->base.dst); | |
1833 | cork->base.dst = NULL; | |
1834 | cork->base.flags &= ~IPCORK_ALLFRAG; | |
bf138862 | 1835 | } |
366e41d9 | 1836 | memset(&cork->fl, 0, sizeof(cork->fl)); |
bf138862 PE |
1837 | } |
1838 | ||
6422398c VY |
1839 | struct sk_buff *__ip6_make_skb(struct sock *sk, |
1840 | struct sk_buff_head *queue, | |
1841 | struct inet_cork_full *cork, | |
1842 | struct inet6_cork *v6_cork) | |
1da177e4 LT |
1843 | { |
1844 | struct sk_buff *skb, *tmp_skb; | |
1845 | struct sk_buff **tail_skb; | |
b60d4e58 | 1846 | struct in6_addr *final_dst; |
1da177e4 | 1847 | struct ipv6_pinfo *np = inet6_sk(sk); |
3bd653c8 | 1848 | struct net *net = sock_net(sk); |
1da177e4 | 1849 | struct ipv6hdr *hdr; |
6422398c VY |
1850 | struct ipv6_txoptions *opt = v6_cork->opt; |
1851 | struct rt6_info *rt = (struct rt6_info *)cork->base.dst; | |
1852 | struct flowi6 *fl6 = &cork->fl.u.ip6; | |
4c9483b2 | 1853 | unsigned char proto = fl6->flowi6_proto; |
1da177e4 | 1854 | |
6422398c | 1855 | skb = __skb_dequeue(queue); |
63159f29 | 1856 | if (!skb) |
1da177e4 LT |
1857 | goto out; |
1858 | tail_skb = &(skb_shinfo(skb)->frag_list); | |
1859 | ||
1860 | /* move skb->data to ip header from ext header */ | |
d56f90a7 | 1861 | if (skb->data < skb_network_header(skb)) |
bbe735e4 | 1862 | __skb_pull(skb, skb_network_offset(skb)); |
6422398c | 1863 | while ((tmp_skb = __skb_dequeue(queue)) != NULL) { |
cfe1fc77 | 1864 | __skb_pull(tmp_skb, skb_network_header_len(skb)); |
1da177e4 LT |
1865 | *tail_skb = tmp_skb; |
1866 | tail_skb = &(tmp_skb->next); | |
1867 | skb->len += tmp_skb->len; | |
1868 | skb->data_len += tmp_skb->len; | |
1da177e4 | 1869 | skb->truesize += tmp_skb->truesize; |
1da177e4 LT |
1870 | tmp_skb->destructor = NULL; |
1871 | tmp_skb->sk = NULL; | |
1da177e4 LT |
1872 | } |
1873 | ||
28a89453 | 1874 | /* Allow local fragmentation. */ |
60ff7467 | 1875 | skb->ignore_df = ip6_sk_ignore_df(sk); |
cfe1fc77 | 1876 | __skb_pull(skb, skb_network_header_len(skb)); |
b60d4e58 PB |
1877 | |
1878 | final_dst = &fl6->daddr; | |
1da177e4 LT |
1879 | if (opt && opt->opt_flen) |
1880 | ipv6_push_frag_opts(skb, opt, &proto); | |
1881 | if (opt && opt->opt_nflen) | |
613fa3ca | 1882 | ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst, &fl6->saddr); |
1da177e4 | 1883 | |
e2d1bca7 ACM |
1884 | skb_push(skb, sizeof(struct ipv6hdr)); |
1885 | skb_reset_network_header(skb); | |
0660e03f | 1886 | hdr = ipv6_hdr(skb); |
1ab1457c | 1887 | |
6422398c | 1888 | ip6_flow_hdr(hdr, v6_cork->tclass, |
cb1ce2ef | 1889 | ip6_make_flowlabel(net, skb, fl6->flowlabel, |
513674b5 | 1890 | ip6_autoflowlabel(net, np), fl6)); |
6422398c | 1891 | hdr->hop_limit = v6_cork->hop_limit; |
1da177e4 | 1892 | hdr->nexthdr = proto; |
4e3fd7a0 AD |
1893 | hdr->saddr = fl6->saddr; |
1894 | hdr->daddr = *final_dst; | |
1da177e4 | 1895 | |
a2c2064f | 1896 | skb->priority = sk->sk_priority; |
c6af0c22 | 1897 | skb->mark = cork->base.mark; |
a818f75e JSP |
1898 | skb->tstamp = cork->base.transmit_time; |
1899 | ||
cd3c7480 | 1900 | ip6_cork_steal_dst(skb, cork); |
edf391ff | 1901 | IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); |
14878f75 | 1902 | if (proto == IPPROTO_ICMPV6) { |
adf30907 | 1903 | struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb)); |
14878f75 | 1904 | |
43a43b60 HFS |
1905 | ICMP6MSGOUT_INC_STATS(net, idev, icmp6_hdr(skb)->icmp6_type); |
1906 | ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS); | |
14878f75 DS |
1907 | } |
1908 | ||
6422398c VY |
1909 | ip6_cork_release(cork, v6_cork); |
1910 | out: | |
1911 | return skb; | |
1912 | } | |
1913 | ||
1914 | int ip6_send_skb(struct sk_buff *skb) | |
1915 | { | |
1916 | struct net *net = sock_net(skb->sk); | |
1917 | struct rt6_info *rt = (struct rt6_info *)skb_dst(skb); | |
1918 | int err; | |
1919 | ||
33224b16 | 1920 | err = ip6_local_out(net, skb->sk, skb); |
1da177e4 LT |
1921 | if (err) { |
1922 | if (err > 0) | |
6ce9e7b5 | 1923 | err = net_xmit_errno(err); |
1da177e4 | 1924 | if (err) |
6422398c VY |
1925 | IP6_INC_STATS(net, rt->rt6i_idev, |
1926 | IPSTATS_MIB_OUTDISCARDS); | |
1da177e4 LT |
1927 | } |
1928 | ||
1da177e4 | 1929 | return err; |
6422398c VY |
1930 | } |
1931 | ||
1932 | int ip6_push_pending_frames(struct sock *sk) | |
1933 | { | |
1934 | struct sk_buff *skb; | |
1935 | ||
1936 | skb = ip6_finish_skb(sk); | |
1937 | if (!skb) | |
1938 | return 0; | |
1939 | ||
1940 | return ip6_send_skb(skb); | |
1da177e4 | 1941 | } |
a495f836 | 1942 | EXPORT_SYMBOL_GPL(ip6_push_pending_frames); |
1da177e4 | 1943 | |
0bbe84a6 | 1944 | static void __ip6_flush_pending_frames(struct sock *sk, |
6422398c VY |
1945 | struct sk_buff_head *queue, |
1946 | struct inet_cork_full *cork, | |
1947 | struct inet6_cork *v6_cork) | |
1da177e4 | 1948 | { |
1da177e4 LT |
1949 | struct sk_buff *skb; |
1950 | ||
0bbe84a6 | 1951 | while ((skb = __skb_dequeue_tail(queue)) != NULL) { |
adf30907 ED |
1952 | if (skb_dst(skb)) |
1953 | IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)), | |
e1f52208 | 1954 | IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
1955 | kfree_skb(skb); |
1956 | } | |
1957 | ||
6422398c | 1958 | ip6_cork_release(cork, v6_cork); |
1da177e4 | 1959 | } |
0bbe84a6 VY |
1960 | |
1961 | void ip6_flush_pending_frames(struct sock *sk) | |
1962 | { | |
6422398c VY |
1963 | __ip6_flush_pending_frames(sk, &sk->sk_write_queue, |
1964 | &inet_sk(sk)->cork, &inet6_sk(sk)->cork); | |
0bbe84a6 | 1965 | } |
a495f836 | 1966 | EXPORT_SYMBOL_GPL(ip6_flush_pending_frames); |
6422398c VY |
1967 | |
1968 | struct sk_buff *ip6_make_skb(struct sock *sk, | |
1969 | int getfrag(void *from, char *to, int offset, | |
1970 | int len, int odd, struct sk_buff *skb), | |
1971 | void *from, int length, int transhdrlen, | |
26879da5 | 1972 | struct ipcm6_cookie *ipc6, struct flowi6 *fl6, |
6422398c | 1973 | struct rt6_info *rt, unsigned int flags, |
5fdaa88d | 1974 | struct inet_cork_full *cork) |
6422398c | 1975 | { |
6422398c VY |
1976 | struct inet6_cork v6_cork; |
1977 | struct sk_buff_head queue; | |
26879da5 | 1978 | int exthdrlen = (ipc6->opt ? ipc6->opt->opt_flen : 0); |
6422398c VY |
1979 | int err; |
1980 | ||
1981 | if (flags & MSG_PROBE) | |
1982 | return NULL; | |
1983 | ||
1984 | __skb_queue_head_init(&queue); | |
1985 | ||
1cd7884d WB |
1986 | cork->base.flags = 0; |
1987 | cork->base.addr = 0; | |
1988 | cork->base.opt = NULL; | |
1989 | cork->base.dst = NULL; | |
6422398c | 1990 | v6_cork.opt = NULL; |
5fdaa88d | 1991 | err = ip6_setup_cork(sk, cork, &v6_cork, ipc6, rt, fl6); |
862c03ee | 1992 | if (err) { |
1cd7884d | 1993 | ip6_cork_release(cork, &v6_cork); |
6422398c | 1994 | return ERR_PTR(err); |
862c03ee | 1995 | } |
26879da5 WW |
1996 | if (ipc6->dontfrag < 0) |
1997 | ipc6->dontfrag = inet6_sk(sk)->dontfrag; | |
6422398c | 1998 | |
1cd7884d | 1999 | err = __ip6_append_data(sk, fl6, &queue, &cork->base, &v6_cork, |
6422398c VY |
2000 | ¤t->task_frag, getfrag, from, |
2001 | length + exthdrlen, transhdrlen + exthdrlen, | |
5fdaa88d | 2002 | flags, ipc6); |
6422398c | 2003 | if (err) { |
1cd7884d | 2004 | __ip6_flush_pending_frames(sk, &queue, cork, &v6_cork); |
6422398c VY |
2005 | return ERR_PTR(err); |
2006 | } | |
2007 | ||
1cd7884d | 2008 | return __ip6_make_skb(sk, &queue, cork, &v6_cork); |
6422398c | 2009 | } |