]>
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 LT |
7 | * |
8 | * $Id: ip6_output.c,v 1.34 2002/02/01 22:01:04 davem Exp $ | |
9 | * | |
10 | * Based on linux/net/ipv4/ip_output.c | |
11 | * | |
12 | * This program is free software; you can redistribute it and/or | |
13 | * modify it under the terms of the GNU General Public License | |
14 | * as published by the Free Software Foundation; either version | |
15 | * 2 of the License, or (at your option) any later version. | |
16 | * | |
17 | * Changes: | |
18 | * A.N.Kuznetsov : airthmetics in fragmentation. | |
19 | * extension headers are implemented. | |
20 | * route changes now work. | |
21 | * ip6_forward does not confuse sniffers. | |
22 | * etc. | |
23 | * | |
24 | * H. von Brand : Added missing #include <linux/string.h> | |
25 | * Imran Patel : frag id should be in NBO | |
26 | * Kazunori MIYAZAWA @USAGI | |
27 | * : add ip6_append_data and related functions | |
28 | * for datagram xmit | |
29 | */ | |
30 | ||
1da177e4 LT |
31 | #include <linux/errno.h> |
32 | #include <linux/types.h> | |
33 | #include <linux/string.h> | |
34 | #include <linux/socket.h> | |
35 | #include <linux/net.h> | |
36 | #include <linux/netdevice.h> | |
37 | #include <linux/if_arp.h> | |
38 | #include <linux/in6.h> | |
39 | #include <linux/tcp.h> | |
40 | #include <linux/route.h> | |
b59f45d0 | 41 | #include <linux/module.h> |
1da177e4 LT |
42 | |
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> | |
58 | ||
59 | static int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *)); | |
60 | ||
61 | static __inline__ void ipv6_select_ident(struct sk_buff *skb, struct frag_hdr *fhdr) | |
62 | { | |
63 | static u32 ipv6_fragmentation_id = 1; | |
64 | static DEFINE_SPINLOCK(ip6_id_lock); | |
65 | ||
66 | spin_lock_bh(&ip6_id_lock); | |
67 | fhdr->identification = htonl(ipv6_fragmentation_id); | |
68 | if (++ipv6_fragmentation_id == 0) | |
69 | ipv6_fragmentation_id = 1; | |
70 | spin_unlock_bh(&ip6_id_lock); | |
71 | } | |
72 | ||
73 | static inline int ip6_output_finish(struct sk_buff *skb) | |
74 | { | |
1da177e4 | 75 | struct dst_entry *dst = skb->dst; |
1da177e4 | 76 | |
3644f0ce SH |
77 | if (dst->hh) |
78 | return neigh_hh_output(dst->hh, skb); | |
79 | else if (dst->neighbour) | |
1da177e4 LT |
80 | return dst->neighbour->output(skb); |
81 | ||
a11d206d | 82 | IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES); |
1da177e4 LT |
83 | kfree_skb(skb); |
84 | return -EINVAL; | |
85 | ||
86 | } | |
87 | ||
88 | /* dev_loopback_xmit for use with netfilter. */ | |
89 | static int ip6_dev_loopback_xmit(struct sk_buff *newskb) | |
90 | { | |
459a98ed | 91 | skb_reset_mac_header(newskb); |
bbe735e4 | 92 | __skb_pull(newskb, skb_network_offset(newskb)); |
1da177e4 LT |
93 | newskb->pkt_type = PACKET_LOOPBACK; |
94 | newskb->ip_summed = CHECKSUM_UNNECESSARY; | |
95 | BUG_TRAP(newskb->dst); | |
96 | ||
97 | netif_rx(newskb); | |
98 | return 0; | |
99 | } | |
100 | ||
101 | ||
102 | static int ip6_output2(struct sk_buff *skb) | |
103 | { | |
104 | struct dst_entry *dst = skb->dst; | |
105 | struct net_device *dev = dst->dev; | |
106 | ||
107 | skb->protocol = htons(ETH_P_IPV6); | |
108 | skb->dev = dev; | |
109 | ||
0660e03f | 110 | if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) { |
1da177e4 | 111 | struct ipv6_pinfo* np = skb->sk ? inet6_sk(skb->sk) : NULL; |
a11d206d | 112 | struct inet6_dev *idev = ip6_dst_idev(skb->dst); |
1da177e4 LT |
113 | |
114 | if (!(dev->flags & IFF_LOOPBACK) && (!np || np->mc_loop) && | |
0660e03f ACM |
115 | ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr, |
116 | &ipv6_hdr(skb)->saddr)) { | |
1da177e4 LT |
117 | struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC); |
118 | ||
119 | /* Do not check for IFF_ALLMULTI; multicast routing | |
120 | is not supported in any case. | |
121 | */ | |
122 | if (newskb) | |
123 | NF_HOOK(PF_INET6, NF_IP6_POST_ROUTING, newskb, NULL, | |
124 | newskb->dev, | |
125 | ip6_dev_loopback_xmit); | |
126 | ||
0660e03f | 127 | if (ipv6_hdr(skb)->hop_limit == 0) { |
a11d206d | 128 | IP6_INC_STATS(idev, IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
129 | kfree_skb(skb); |
130 | return 0; | |
131 | } | |
132 | } | |
133 | ||
a11d206d | 134 | IP6_INC_STATS(idev, IPSTATS_MIB_OUTMCASTPKTS); |
1da177e4 LT |
135 | } |
136 | ||
137 | return NF_HOOK(PF_INET6, NF_IP6_POST_ROUTING, skb,NULL, skb->dev,ip6_output_finish); | |
138 | } | |
139 | ||
140 | int ip6_output(struct sk_buff *skb) | |
141 | { | |
89114afd | 142 | if ((skb->len > dst_mtu(skb->dst) && !skb_is_gso(skb)) || |
e89e9cf5 | 143 | dst_allfrag(skb->dst)) |
1da177e4 LT |
144 | return ip6_fragment(skb, ip6_output2); |
145 | else | |
146 | return ip6_output2(skb); | |
147 | } | |
148 | ||
1da177e4 LT |
149 | /* |
150 | * xmit an sk_buff (used by TCP) | |
151 | */ | |
152 | ||
153 | int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl, | |
154 | struct ipv6_txoptions *opt, int ipfragok) | |
155 | { | |
b30bd282 | 156 | struct ipv6_pinfo *np = inet6_sk(sk); |
1da177e4 LT |
157 | struct in6_addr *first_hop = &fl->fl6_dst; |
158 | struct dst_entry *dst = skb->dst; | |
159 | struct ipv6hdr *hdr; | |
160 | u8 proto = fl->proto; | |
161 | int seg_len = skb->len; | |
41a1f8ea | 162 | int hlimit, tclass; |
1da177e4 LT |
163 | u32 mtu; |
164 | ||
165 | if (opt) { | |
166 | int head_room; | |
167 | ||
168 | /* First: exthdrs may take lots of space (~8K for now) | |
169 | MAX_HEADER is not enough. | |
170 | */ | |
171 | head_room = opt->opt_nflen + opt->opt_flen; | |
172 | seg_len += head_room; | |
173 | head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev); | |
174 | ||
175 | if (skb_headroom(skb) < head_room) { | |
176 | struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room); | |
a11d206d YH |
177 | if (skb2 == NULL) { |
178 | IP6_INC_STATS(ip6_dst_idev(skb->dst), | |
179 | IPSTATS_MIB_OUTDISCARDS); | |
180 | kfree_skb(skb); | |
1da177e4 LT |
181 | return -ENOBUFS; |
182 | } | |
a11d206d YH |
183 | kfree_skb(skb); |
184 | skb = skb2; | |
1da177e4 LT |
185 | if (sk) |
186 | skb_set_owner_w(skb, sk); | |
187 | } | |
188 | if (opt->opt_flen) | |
189 | ipv6_push_frag_opts(skb, opt, &proto); | |
190 | if (opt->opt_nflen) | |
191 | ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop); | |
192 | } | |
193 | ||
e2d1bca7 ACM |
194 | skb_push(skb, sizeof(struct ipv6hdr)); |
195 | skb_reset_network_header(skb); | |
0660e03f | 196 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
197 | |
198 | /* | |
199 | * Fill in the IPv6 header | |
200 | */ | |
201 | ||
1da177e4 LT |
202 | hlimit = -1; |
203 | if (np) | |
204 | hlimit = np->hop_limit; | |
205 | if (hlimit < 0) | |
206 | hlimit = dst_metric(dst, RTAX_HOPLIMIT); | |
207 | if (hlimit < 0) | |
208 | hlimit = ipv6_get_hoplimit(dst->dev); | |
209 | ||
41a1f8ea YH |
210 | tclass = -1; |
211 | if (np) | |
212 | tclass = np->tclass; | |
213 | if (tclass < 0) | |
214 | tclass = 0; | |
215 | ||
90bcaf7b | 216 | *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | fl->fl6_flowlabel; |
41a1f8ea | 217 | |
1da177e4 LT |
218 | hdr->payload_len = htons(seg_len); |
219 | hdr->nexthdr = proto; | |
220 | hdr->hop_limit = hlimit; | |
221 | ||
222 | ipv6_addr_copy(&hdr->saddr, &fl->fl6_src); | |
223 | ipv6_addr_copy(&hdr->daddr, first_hop); | |
224 | ||
a2c2064f PM |
225 | skb->priority = sk->sk_priority; |
226 | ||
1da177e4 | 227 | mtu = dst_mtu(dst); |
89114afd | 228 | if ((skb->len <= mtu) || ipfragok || skb_is_gso(skb)) { |
a11d206d YH |
229 | IP6_INC_STATS(ip6_dst_idev(skb->dst), |
230 | IPSTATS_MIB_OUTREQUESTS); | |
6869c4d8 HW |
231 | return NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev, |
232 | dst_output); | |
1da177e4 LT |
233 | } |
234 | ||
235 | if (net_ratelimit()) | |
236 | printk(KERN_DEBUG "IPv6: sending pkt_too_big to self\n"); | |
237 | skb->dev = dst->dev; | |
238 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, skb->dev); | |
a11d206d | 239 | IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
240 | kfree_skb(skb); |
241 | return -EMSGSIZE; | |
242 | } | |
243 | ||
7159039a YH |
244 | EXPORT_SYMBOL(ip6_xmit); |
245 | ||
1da177e4 LT |
246 | /* |
247 | * To avoid extra problems ND packets are send through this | |
248 | * routine. It's code duplication but I really want to avoid | |
249 | * extra checks since ipv6_build_header is used by TCP (which | |
250 | * is for us performance critical) | |
251 | */ | |
252 | ||
253 | int ip6_nd_hdr(struct sock *sk, struct sk_buff *skb, struct net_device *dev, | |
254 | struct in6_addr *saddr, struct in6_addr *daddr, | |
255 | int proto, int len) | |
256 | { | |
257 | struct ipv6_pinfo *np = inet6_sk(sk); | |
258 | struct ipv6hdr *hdr; | |
259 | int totlen; | |
260 | ||
261 | skb->protocol = htons(ETH_P_IPV6); | |
262 | skb->dev = dev; | |
263 | ||
264 | totlen = len + sizeof(struct ipv6hdr); | |
265 | ||
0660e03f ACM |
266 | skb->nh.raw = skb_put(skb, sizeof(struct ipv6hdr)); |
267 | hdr = ipv6_hdr(skb); | |
1da177e4 | 268 | |
ae08e1f0 | 269 | *(__be32*)hdr = htonl(0x60000000); |
1da177e4 LT |
270 | |
271 | hdr->payload_len = htons(len); | |
272 | hdr->nexthdr = proto; | |
273 | hdr->hop_limit = np->hop_limit; | |
274 | ||
275 | ipv6_addr_copy(&hdr->saddr, saddr); | |
276 | ipv6_addr_copy(&hdr->daddr, daddr); | |
277 | ||
278 | return 0; | |
279 | } | |
280 | ||
281 | static int ip6_call_ra_chain(struct sk_buff *skb, int sel) | |
282 | { | |
283 | struct ip6_ra_chain *ra; | |
284 | struct sock *last = NULL; | |
285 | ||
286 | read_lock(&ip6_ra_lock); | |
287 | for (ra = ip6_ra_chain; ra; ra = ra->next) { | |
288 | struct sock *sk = ra->sk; | |
0bd1b59b AM |
289 | if (sk && ra->sel == sel && |
290 | (!sk->sk_bound_dev_if || | |
291 | sk->sk_bound_dev_if == skb->dev->ifindex)) { | |
1da177e4 LT |
292 | if (last) { |
293 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
294 | if (skb2) | |
295 | rawv6_rcv(last, skb2); | |
296 | } | |
297 | last = sk; | |
298 | } | |
299 | } | |
300 | ||
301 | if (last) { | |
302 | rawv6_rcv(last, skb); | |
303 | read_unlock(&ip6_ra_lock); | |
304 | return 1; | |
305 | } | |
306 | read_unlock(&ip6_ra_lock); | |
307 | return 0; | |
308 | } | |
309 | ||
e21e0b5f VN |
310 | static int ip6_forward_proxy_check(struct sk_buff *skb) |
311 | { | |
0660e03f | 312 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
e21e0b5f VN |
313 | u8 nexthdr = hdr->nexthdr; |
314 | int offset; | |
315 | ||
316 | if (ipv6_ext_hdr(nexthdr)) { | |
317 | offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr); | |
318 | if (offset < 0) | |
319 | return 0; | |
320 | } else | |
321 | offset = sizeof(struct ipv6hdr); | |
322 | ||
323 | if (nexthdr == IPPROTO_ICMPV6) { | |
324 | struct icmp6hdr *icmp6; | |
325 | ||
d56f90a7 ACM |
326 | if (!pskb_may_pull(skb, (skb_network_header(skb) + |
327 | offset + 1 - skb->data))) | |
e21e0b5f VN |
328 | return 0; |
329 | ||
d56f90a7 | 330 | icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset); |
e21e0b5f VN |
331 | |
332 | switch (icmp6->icmp6_type) { | |
333 | case NDISC_ROUTER_SOLICITATION: | |
334 | case NDISC_ROUTER_ADVERTISEMENT: | |
335 | case NDISC_NEIGHBOUR_SOLICITATION: | |
336 | case NDISC_NEIGHBOUR_ADVERTISEMENT: | |
337 | case NDISC_REDIRECT: | |
338 | /* For reaction involving unicast neighbor discovery | |
339 | * message destined to the proxied address, pass it to | |
340 | * input function. | |
341 | */ | |
342 | return 1; | |
343 | default: | |
344 | break; | |
345 | } | |
346 | } | |
347 | ||
74553b09 VN |
348 | /* |
349 | * The proxying router can't forward traffic sent to a link-local | |
350 | * address, so signal the sender and discard the packet. This | |
351 | * behavior is clarified by the MIPv6 specification. | |
352 | */ | |
353 | if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) { | |
354 | dst_link_failure(skb); | |
355 | return -1; | |
356 | } | |
357 | ||
e21e0b5f VN |
358 | return 0; |
359 | } | |
360 | ||
1da177e4 LT |
361 | static inline int ip6_forward_finish(struct sk_buff *skb) |
362 | { | |
363 | return dst_output(skb); | |
364 | } | |
365 | ||
366 | int ip6_forward(struct sk_buff *skb) | |
367 | { | |
368 | struct dst_entry *dst = skb->dst; | |
0660e03f | 369 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
1da177e4 | 370 | struct inet6_skb_parm *opt = IP6CB(skb); |
1ab1457c | 371 | |
1da177e4 LT |
372 | if (ipv6_devconf.forwarding == 0) |
373 | goto error; | |
374 | ||
375 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) { | |
a11d206d | 376 | IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS); |
1da177e4 LT |
377 | goto drop; |
378 | } | |
379 | ||
380 | skb->ip_summed = CHECKSUM_NONE; | |
381 | ||
382 | /* | |
383 | * We DO NOT make any processing on | |
384 | * RA packets, pushing them to user level AS IS | |
385 | * without ane WARRANTY that application will be able | |
386 | * to interpret them. The reason is that we | |
387 | * cannot make anything clever here. | |
388 | * | |
389 | * We are not end-node, so that if packet contains | |
390 | * AH/ESP, we cannot make anything. | |
391 | * Defragmentation also would be mistake, RA packets | |
392 | * cannot be fragmented, because there is no warranty | |
393 | * that different fragments will go along one path. --ANK | |
394 | */ | |
395 | if (opt->ra) { | |
d56f90a7 | 396 | u8 *ptr = skb_network_header(skb) + opt->ra; |
1da177e4 LT |
397 | if (ip6_call_ra_chain(skb, (ptr[2]<<8) + ptr[3])) |
398 | return 0; | |
399 | } | |
400 | ||
401 | /* | |
402 | * check and decrement ttl | |
403 | */ | |
404 | if (hdr->hop_limit <= 1) { | |
405 | /* Force OUTPUT device used as source address */ | |
406 | skb->dev = dst->dev; | |
407 | icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, | |
408 | 0, skb->dev); | |
a11d206d | 409 | IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS); |
1da177e4 LT |
410 | |
411 | kfree_skb(skb); | |
412 | return -ETIMEDOUT; | |
413 | } | |
414 | ||
fbea49e1 YH |
415 | /* XXX: idev->cnf.proxy_ndp? */ |
416 | if (ipv6_devconf.proxy_ndp && | |
417 | pneigh_lookup(&nd_tbl, &hdr->daddr, skb->dev, 0)) { | |
74553b09 VN |
418 | int proxied = ip6_forward_proxy_check(skb); |
419 | if (proxied > 0) | |
e21e0b5f | 420 | return ip6_input(skb); |
74553b09 | 421 | else if (proxied < 0) { |
a11d206d | 422 | IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS); |
74553b09 VN |
423 | goto drop; |
424 | } | |
e21e0b5f VN |
425 | } |
426 | ||
1da177e4 | 427 | if (!xfrm6_route_forward(skb)) { |
a11d206d | 428 | IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS); |
1da177e4 LT |
429 | goto drop; |
430 | } | |
431 | dst = skb->dst; | |
432 | ||
433 | /* IPv6 specs say nothing about it, but it is clear that we cannot | |
434 | send redirects to source routed frames. | |
435 | */ | |
436 | if (skb->dev == dst->dev && dst->neighbour && opt->srcrt == 0) { | |
437 | struct in6_addr *target = NULL; | |
438 | struct rt6_info *rt; | |
439 | struct neighbour *n = dst->neighbour; | |
440 | ||
441 | /* | |
442 | * incoming and outgoing devices are the same | |
443 | * send a redirect. | |
444 | */ | |
445 | ||
446 | rt = (struct rt6_info *) dst; | |
447 | if ((rt->rt6i_flags & RTF_GATEWAY)) | |
448 | target = (struct in6_addr*)&n->primary_key; | |
449 | else | |
450 | target = &hdr->daddr; | |
451 | ||
452 | /* Limit redirects both by destination (here) | |
453 | and by source (inside ndisc_send_redirect) | |
454 | */ | |
455 | if (xrlim_allow(dst, 1*HZ)) | |
456 | ndisc_send_redirect(skb, n, target); | |
457 | } else if (ipv6_addr_type(&hdr->saddr)&(IPV6_ADDR_MULTICAST|IPV6_ADDR_LOOPBACK | |
458 | |IPV6_ADDR_LINKLOCAL)) { | |
459 | /* This check is security critical. */ | |
460 | goto error; | |
461 | } | |
462 | ||
463 | if (skb->len > dst_mtu(dst)) { | |
464 | /* Again, force OUTPUT device used as source address */ | |
465 | skb->dev = dst->dev; | |
466 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, dst_mtu(dst), skb->dev); | |
a11d206d YH |
467 | IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INTOOBIGERRORS); |
468 | IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_FRAGFAILS); | |
1da177e4 LT |
469 | kfree_skb(skb); |
470 | return -EMSGSIZE; | |
471 | } | |
472 | ||
473 | if (skb_cow(skb, dst->dev->hard_header_len)) { | |
a11d206d | 474 | IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
475 | goto drop; |
476 | } | |
477 | ||
0660e03f | 478 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
479 | |
480 | /* Mangling hops number delayed to point after skb COW */ | |
1ab1457c | 481 | |
1da177e4 LT |
482 | hdr->hop_limit--; |
483 | ||
a11d206d | 484 | IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS); |
1da177e4 LT |
485 | return NF_HOOK(PF_INET6,NF_IP6_FORWARD, skb, skb->dev, dst->dev, ip6_forward_finish); |
486 | ||
487 | error: | |
a11d206d | 488 | IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS); |
1da177e4 LT |
489 | drop: |
490 | kfree_skb(skb); | |
491 | return -EINVAL; | |
492 | } | |
493 | ||
494 | static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from) | |
495 | { | |
496 | to->pkt_type = from->pkt_type; | |
497 | to->priority = from->priority; | |
498 | to->protocol = from->protocol; | |
1da177e4 LT |
499 | dst_release(to->dst); |
500 | to->dst = dst_clone(from->dst); | |
501 | to->dev = from->dev; | |
82e91ffe | 502 | to->mark = from->mark; |
1da177e4 LT |
503 | |
504 | #ifdef CONFIG_NET_SCHED | |
505 | to->tc_index = from->tc_index; | |
506 | #endif | |
507 | #ifdef CONFIG_NETFILTER | |
1da177e4 | 508 | /* Connection association is same as pre-frag packet */ |
9fb9cbb1 | 509 | nf_conntrack_put(to->nfct); |
1da177e4 LT |
510 | to->nfct = from->nfct; |
511 | nf_conntrack_get(to->nfct); | |
512 | to->nfctinfo = from->nfctinfo; | |
9fb9cbb1 YK |
513 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
514 | nf_conntrack_put_reasm(to->nfct_reasm); | |
515 | to->nfct_reasm = from->nfct_reasm; | |
516 | nf_conntrack_get_reasm(to->nfct_reasm); | |
517 | #endif | |
1da177e4 LT |
518 | #ifdef CONFIG_BRIDGE_NETFILTER |
519 | nf_bridge_put(to->nf_bridge); | |
520 | to->nf_bridge = from->nf_bridge; | |
521 | nf_bridge_get(to->nf_bridge); | |
522 | #endif | |
1da177e4 | 523 | #endif |
984bc16c | 524 | skb_copy_secmark(to, from); |
1da177e4 LT |
525 | } |
526 | ||
527 | int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr) | |
528 | { | |
529 | u16 offset = sizeof(struct ipv6hdr); | |
0660e03f ACM |
530 | struct ipv6_opt_hdr *exthdr = |
531 | (struct ipv6_opt_hdr *)(ipv6_hdr(skb) + 1); | |
d56f90a7 | 532 | unsigned int packet_len = skb->tail - skb_network_header(skb); |
1da177e4 | 533 | int found_rhdr = 0; |
0660e03f | 534 | *nexthdr = &ipv6_hdr(skb)->nexthdr; |
1da177e4 LT |
535 | |
536 | while (offset + 1 <= packet_len) { | |
537 | ||
538 | switch (**nexthdr) { | |
539 | ||
540 | case NEXTHDR_HOP: | |
27637df9 | 541 | break; |
1da177e4 | 542 | case NEXTHDR_ROUTING: |
27637df9 MN |
543 | found_rhdr = 1; |
544 | break; | |
1da177e4 | 545 | case NEXTHDR_DEST: |
27637df9 MN |
546 | #ifdef CONFIG_IPV6_MIP6 |
547 | if (ipv6_find_tlv(skb, offset, IPV6_TLV_HAO) >= 0) | |
548 | break; | |
549 | #endif | |
550 | if (found_rhdr) | |
551 | return offset; | |
1da177e4 LT |
552 | break; |
553 | default : | |
554 | return offset; | |
555 | } | |
27637df9 MN |
556 | |
557 | offset += ipv6_optlen(exthdr); | |
558 | *nexthdr = &exthdr->nexthdr; | |
d56f90a7 ACM |
559 | exthdr = (struct ipv6_opt_hdr *)(skb_network_header(skb) + |
560 | offset); | |
1da177e4 LT |
561 | } |
562 | ||
563 | return offset; | |
564 | } | |
b59f45d0 | 565 | EXPORT_SYMBOL_GPL(ip6_find_1stfragopt); |
1da177e4 LT |
566 | |
567 | static int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *)) | |
568 | { | |
569 | struct net_device *dev; | |
570 | struct sk_buff *frag; | |
571 | struct rt6_info *rt = (struct rt6_info*)skb->dst; | |
d91675f9 | 572 | struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL; |
1da177e4 LT |
573 | struct ipv6hdr *tmp_hdr; |
574 | struct frag_hdr *fh; | |
575 | unsigned int mtu, hlen, left, len; | |
ae08e1f0 | 576 | __be32 frag_id = 0; |
1da177e4 LT |
577 | int ptr, offset = 0, err=0; |
578 | u8 *prevhdr, nexthdr = 0; | |
579 | ||
580 | dev = rt->u.dst.dev; | |
581 | hlen = ip6_find_1stfragopt(skb, &prevhdr); | |
582 | nexthdr = *prevhdr; | |
583 | ||
d91675f9 YH |
584 | mtu = dst_mtu(&rt->u.dst); |
585 | if (np && np->frag_size < mtu) { | |
586 | if (np->frag_size) | |
587 | mtu = np->frag_size; | |
588 | } | |
589 | mtu -= hlen + sizeof(struct frag_hdr); | |
1da177e4 LT |
590 | |
591 | if (skb_shinfo(skb)->frag_list) { | |
592 | int first_len = skb_pagelen(skb); | |
593 | ||
594 | if (first_len - hlen > mtu || | |
595 | ((first_len - hlen) & 7) || | |
596 | skb_cloned(skb)) | |
597 | goto slow_path; | |
598 | ||
599 | for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next) { | |
600 | /* Correct geometry. */ | |
601 | if (frag->len > mtu || | |
602 | ((frag->len & 7) && frag->next) || | |
603 | skb_headroom(frag) < hlen) | |
604 | goto slow_path; | |
605 | ||
1da177e4 LT |
606 | /* Partially cloned skb? */ |
607 | if (skb_shared(frag)) | |
608 | goto slow_path; | |
2fdba6b0 HX |
609 | |
610 | BUG_ON(frag->sk); | |
611 | if (skb->sk) { | |
612 | sock_hold(skb->sk); | |
613 | frag->sk = skb->sk; | |
614 | frag->destructor = sock_wfree; | |
615 | skb->truesize -= frag->truesize; | |
616 | } | |
1da177e4 LT |
617 | } |
618 | ||
619 | err = 0; | |
620 | offset = 0; | |
621 | frag = skb_shinfo(skb)->frag_list; | |
622 | skb_shinfo(skb)->frag_list = NULL; | |
623 | /* BUILD HEADER */ | |
624 | ||
9a217a1c | 625 | *prevhdr = NEXTHDR_FRAGMENT; |
d56f90a7 | 626 | tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC); |
1da177e4 | 627 | if (!tmp_hdr) { |
a11d206d | 628 | IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
629 | return -ENOMEM; |
630 | } | |
631 | ||
1da177e4 LT |
632 | __skb_pull(skb, hlen); |
633 | fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr)); | |
e2d1bca7 ACM |
634 | __skb_push(skb, hlen); |
635 | skb_reset_network_header(skb); | |
d56f90a7 | 636 | memcpy(skb_network_header(skb), tmp_hdr, hlen); |
1da177e4 LT |
637 | |
638 | ipv6_select_ident(skb, fh); | |
639 | fh->nexthdr = nexthdr; | |
640 | fh->reserved = 0; | |
641 | fh->frag_off = htons(IP6_MF); | |
642 | frag_id = fh->identification; | |
643 | ||
644 | first_len = skb_pagelen(skb); | |
645 | skb->data_len = first_len - skb_headlen(skb); | |
646 | skb->len = first_len; | |
0660e03f ACM |
647 | ipv6_hdr(skb)->payload_len = htons(first_len - |
648 | sizeof(struct ipv6hdr)); | |
a11d206d YH |
649 | |
650 | dst_hold(&rt->u.dst); | |
1da177e4 LT |
651 | |
652 | for (;;) { | |
653 | /* Prepare header of the next frame, | |
654 | * before previous one went down. */ | |
655 | if (frag) { | |
656 | frag->ip_summed = CHECKSUM_NONE; | |
657 | frag->h.raw = frag->data; | |
658 | fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr)); | |
e2d1bca7 ACM |
659 | __skb_push(frag, hlen); |
660 | skb_reset_network_header(frag); | |
d56f90a7 ACM |
661 | memcpy(skb_network_header(frag), tmp_hdr, |
662 | hlen); | |
1da177e4 LT |
663 | offset += skb->len - hlen - sizeof(struct frag_hdr); |
664 | fh->nexthdr = nexthdr; | |
665 | fh->reserved = 0; | |
666 | fh->frag_off = htons(offset); | |
667 | if (frag->next != NULL) | |
668 | fh->frag_off |= htons(IP6_MF); | |
669 | fh->identification = frag_id; | |
0660e03f ACM |
670 | ipv6_hdr(frag)->payload_len = |
671 | htons(frag->len - | |
672 | sizeof(struct ipv6hdr)); | |
1da177e4 LT |
673 | ip6_copy_metadata(frag, skb); |
674 | } | |
1ab1457c | 675 | |
1da177e4 | 676 | err = output(skb); |
dafee490 | 677 | if(!err) |
a11d206d | 678 | IP6_INC_STATS(ip6_dst_idev(&rt->u.dst), IPSTATS_MIB_FRAGCREATES); |
dafee490 | 679 | |
1da177e4 LT |
680 | if (err || !frag) |
681 | break; | |
682 | ||
683 | skb = frag; | |
684 | frag = skb->next; | |
685 | skb->next = NULL; | |
686 | } | |
687 | ||
a51482bd | 688 | kfree(tmp_hdr); |
1da177e4 LT |
689 | |
690 | if (err == 0) { | |
a11d206d YH |
691 | IP6_INC_STATS(ip6_dst_idev(&rt->u.dst), IPSTATS_MIB_FRAGOKS); |
692 | dst_release(&rt->u.dst); | |
1da177e4 LT |
693 | return 0; |
694 | } | |
695 | ||
696 | while (frag) { | |
697 | skb = frag->next; | |
698 | kfree_skb(frag); | |
699 | frag = skb; | |
700 | } | |
701 | ||
a11d206d YH |
702 | IP6_INC_STATS(ip6_dst_idev(&rt->u.dst), IPSTATS_MIB_FRAGFAILS); |
703 | dst_release(&rt->u.dst); | |
1da177e4 LT |
704 | return err; |
705 | } | |
706 | ||
707 | slow_path: | |
708 | left = skb->len - hlen; /* Space per frame */ | |
709 | ptr = hlen; /* Where to start from */ | |
710 | ||
711 | /* | |
712 | * Fragment the datagram. | |
713 | */ | |
714 | ||
715 | *prevhdr = NEXTHDR_FRAGMENT; | |
716 | ||
717 | /* | |
718 | * Keep copying data until we run out. | |
719 | */ | |
720 | while(left > 0) { | |
721 | len = left; | |
722 | /* IF: it doesn't fit, use 'mtu' - the data space left */ | |
723 | if (len > mtu) | |
724 | len = mtu; | |
725 | /* IF: we are not sending upto and including the packet end | |
726 | then align the next start on an eight byte boundary */ | |
727 | if (len < left) { | |
728 | len &= ~7; | |
729 | } | |
730 | /* | |
731 | * Allocate buffer. | |
732 | */ | |
733 | ||
734 | if ((frag = alloc_skb(len+hlen+sizeof(struct frag_hdr)+LL_RESERVED_SPACE(rt->u.dst.dev), GFP_ATOMIC)) == NULL) { | |
64ce2073 | 735 | NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n"); |
a11d206d YH |
736 | IP6_INC_STATS(ip6_dst_idev(skb->dst), |
737 | IPSTATS_MIB_FRAGFAILS); | |
1da177e4 LT |
738 | err = -ENOMEM; |
739 | goto fail; | |
740 | } | |
741 | ||
742 | /* | |
743 | * Set up data on packet | |
744 | */ | |
745 | ||
746 | ip6_copy_metadata(frag, skb); | |
747 | skb_reserve(frag, LL_RESERVED_SPACE(rt->u.dst.dev)); | |
748 | skb_put(frag, len + hlen + sizeof(struct frag_hdr)); | |
c1d2bbe1 | 749 | skb_reset_network_header(frag); |
1da177e4 LT |
750 | fh = (struct frag_hdr*)(frag->data + hlen); |
751 | frag->h.raw = frag->data + hlen + sizeof(struct frag_hdr); | |
752 | ||
753 | /* | |
754 | * Charge the memory for the fragment to any owner | |
755 | * it might possess | |
756 | */ | |
757 | if (skb->sk) | |
758 | skb_set_owner_w(frag, skb->sk); | |
759 | ||
760 | /* | |
761 | * Copy the packet header into the new buffer. | |
762 | */ | |
d56f90a7 | 763 | memcpy(skb_network_header(frag), skb->data, hlen); |
1da177e4 LT |
764 | |
765 | /* | |
766 | * Build fragment header. | |
767 | */ | |
768 | fh->nexthdr = nexthdr; | |
769 | fh->reserved = 0; | |
f36d6ab1 | 770 | if (!frag_id) { |
1da177e4 LT |
771 | ipv6_select_ident(skb, fh); |
772 | frag_id = fh->identification; | |
773 | } else | |
774 | fh->identification = frag_id; | |
775 | ||
776 | /* | |
777 | * Copy a block of the IP datagram. | |
778 | */ | |
779 | if (skb_copy_bits(skb, ptr, frag->h.raw, len)) | |
780 | BUG(); | |
781 | left -= len; | |
782 | ||
783 | fh->frag_off = htons(offset); | |
784 | if (left > 0) | |
785 | fh->frag_off |= htons(IP6_MF); | |
0660e03f ACM |
786 | ipv6_hdr(frag)->payload_len = htons(frag->len - |
787 | sizeof(struct ipv6hdr)); | |
1da177e4 LT |
788 | |
789 | ptr += len; | |
790 | offset += len; | |
791 | ||
792 | /* | |
793 | * Put this fragment into the sending queue. | |
794 | */ | |
1da177e4 LT |
795 | err = output(frag); |
796 | if (err) | |
797 | goto fail; | |
dafee490 | 798 | |
a11d206d | 799 | IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGCREATES); |
1da177e4 | 800 | } |
a11d206d YH |
801 | IP6_INC_STATS(ip6_dst_idev(skb->dst), |
802 | IPSTATS_MIB_FRAGOKS); | |
1da177e4 | 803 | kfree_skb(skb); |
1da177e4 LT |
804 | return err; |
805 | ||
806 | fail: | |
a11d206d YH |
807 | IP6_INC_STATS(ip6_dst_idev(skb->dst), |
808 | IPSTATS_MIB_FRAGFAILS); | |
1ab1457c | 809 | kfree_skb(skb); |
1da177e4 LT |
810 | return err; |
811 | } | |
812 | ||
cf6b1982 YH |
813 | static inline int ip6_rt_check(struct rt6key *rt_key, |
814 | struct in6_addr *fl_addr, | |
815 | struct in6_addr *addr_cache) | |
816 | { | |
817 | return ((rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) && | |
818 | (addr_cache == NULL || !ipv6_addr_equal(fl_addr, addr_cache))); | |
819 | } | |
820 | ||
497c615a HX |
821 | static struct dst_entry *ip6_sk_dst_check(struct sock *sk, |
822 | struct dst_entry *dst, | |
823 | struct flowi *fl) | |
1da177e4 | 824 | { |
497c615a HX |
825 | struct ipv6_pinfo *np = inet6_sk(sk); |
826 | struct rt6_info *rt = (struct rt6_info *)dst; | |
1da177e4 | 827 | |
497c615a HX |
828 | if (!dst) |
829 | goto out; | |
830 | ||
831 | /* Yes, checking route validity in not connected | |
832 | * case is not very simple. Take into account, | |
833 | * that we do not support routing by source, TOS, | |
834 | * and MSG_DONTROUTE --ANK (980726) | |
835 | * | |
cf6b1982 YH |
836 | * 1. ip6_rt_check(): If route was host route, |
837 | * check that cached destination is current. | |
497c615a HX |
838 | * If it is network route, we still may |
839 | * check its validity using saved pointer | |
840 | * to the last used address: daddr_cache. | |
841 | * We do not want to save whole address now, | |
842 | * (because main consumer of this service | |
843 | * is tcp, which has not this problem), | |
844 | * so that the last trick works only on connected | |
845 | * sockets. | |
846 | * 2. oif also should be the same. | |
847 | */ | |
cf6b1982 | 848 | if (ip6_rt_check(&rt->rt6i_dst, &fl->fl6_dst, np->daddr_cache) || |
8e1ef0a9 YH |
849 | #ifdef CONFIG_IPV6_SUBTREES |
850 | ip6_rt_check(&rt->rt6i_src, &fl->fl6_src, np->saddr_cache) || | |
851 | #endif | |
cf6b1982 | 852 | (fl->oif && fl->oif != dst->dev->ifindex)) { |
497c615a HX |
853 | dst_release(dst); |
854 | dst = NULL; | |
1da177e4 LT |
855 | } |
856 | ||
497c615a HX |
857 | out: |
858 | return dst; | |
859 | } | |
860 | ||
861 | static int ip6_dst_lookup_tail(struct sock *sk, | |
862 | struct dst_entry **dst, struct flowi *fl) | |
863 | { | |
864 | int err; | |
865 | ||
1da177e4 LT |
866 | if (*dst == NULL) |
867 | *dst = ip6_route_output(sk, fl); | |
868 | ||
869 | if ((err = (*dst)->error)) | |
870 | goto out_err_release; | |
871 | ||
872 | if (ipv6_addr_any(&fl->fl6_src)) { | |
873 | err = ipv6_get_saddr(*dst, &fl->fl6_dst, &fl->fl6_src); | |
44456d37 | 874 | if (err) |
1da177e4 | 875 | goto out_err_release; |
1da177e4 LT |
876 | } |
877 | ||
95c385b4 NH |
878 | #ifdef CONFIG_IPV6_OPTIMISTIC_DAD |
879 | /* | |
880 | * Here if the dst entry we've looked up | |
881 | * has a neighbour entry that is in the INCOMPLETE | |
882 | * state and the src address from the flow is | |
883 | * marked as OPTIMISTIC, we release the found | |
884 | * dst entry and replace it instead with the | |
885 | * dst entry of the nexthop router | |
886 | */ | |
887 | if (!((*dst)->neighbour->nud_state & NUD_VALID)) { | |
888 | struct inet6_ifaddr *ifp; | |
889 | struct flowi fl_gw; | |
890 | int redirect; | |
891 | ||
892 | ifp = ipv6_get_ifaddr(&fl->fl6_src, (*dst)->dev, 1); | |
893 | ||
894 | redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC); | |
895 | if (ifp) | |
896 | in6_ifa_put(ifp); | |
897 | ||
898 | if (redirect) { | |
899 | /* | |
900 | * We need to get the dst entry for the | |
901 | * default router instead | |
902 | */ | |
903 | dst_release(*dst); | |
904 | memcpy(&fl_gw, fl, sizeof(struct flowi)); | |
905 | memset(&fl_gw.fl6_dst, 0, sizeof(struct in6_addr)); | |
906 | *dst = ip6_route_output(sk, &fl_gw); | |
907 | if ((err = (*dst)->error)) | |
908 | goto out_err_release; | |
909 | } | |
910 | } | |
911 | #endif | |
912 | ||
1da177e4 LT |
913 | return 0; |
914 | ||
915 | out_err_release: | |
916 | dst_release(*dst); | |
917 | *dst = NULL; | |
918 | return err; | |
919 | } | |
34a0b3cd | 920 | |
497c615a HX |
921 | /** |
922 | * ip6_dst_lookup - perform route lookup on flow | |
923 | * @sk: socket which provides route info | |
924 | * @dst: pointer to dst_entry * for result | |
925 | * @fl: flow to lookup | |
926 | * | |
927 | * This function performs a route lookup on the given flow. | |
928 | * | |
929 | * It returns zero on success, or a standard errno code on error. | |
930 | */ | |
931 | int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl) | |
932 | { | |
933 | *dst = NULL; | |
934 | return ip6_dst_lookup_tail(sk, dst, fl); | |
935 | } | |
3cf3dc6c ACM |
936 | EXPORT_SYMBOL_GPL(ip6_dst_lookup); |
937 | ||
497c615a HX |
938 | /** |
939 | * ip6_sk_dst_lookup - perform socket cached route lookup on flow | |
940 | * @sk: socket which provides the dst cache and route info | |
941 | * @dst: pointer to dst_entry * for result | |
942 | * @fl: flow to lookup | |
943 | * | |
944 | * This function performs a route lookup on the given flow with the | |
945 | * possibility of using the cached route in the socket if it is valid. | |
946 | * It will take the socket dst lock when operating on the dst cache. | |
947 | * As a result, this function can only be used in process context. | |
948 | * | |
949 | * It returns zero on success, or a standard errno code on error. | |
950 | */ | |
951 | int ip6_sk_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl) | |
952 | { | |
953 | *dst = NULL; | |
954 | if (sk) { | |
955 | *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie); | |
956 | *dst = ip6_sk_dst_check(sk, *dst, fl); | |
957 | } | |
958 | ||
959 | return ip6_dst_lookup_tail(sk, dst, fl); | |
960 | } | |
961 | EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup); | |
962 | ||
34a0b3cd | 963 | static inline int ip6_ufo_append_data(struct sock *sk, |
e89e9cf5 AR |
964 | int getfrag(void *from, char *to, int offset, int len, |
965 | int odd, struct sk_buff *skb), | |
966 | void *from, int length, int hh_len, int fragheaderlen, | |
967 | int transhdrlen, int mtu,unsigned int flags) | |
968 | ||
969 | { | |
970 | struct sk_buff *skb; | |
971 | int err; | |
972 | ||
973 | /* There is support for UDP large send offload by network | |
974 | * device, so create one single skb packet containing complete | |
975 | * udp datagram | |
976 | */ | |
977 | if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) { | |
978 | skb = sock_alloc_send_skb(sk, | |
979 | hh_len + fragheaderlen + transhdrlen + 20, | |
980 | (flags & MSG_DONTWAIT), &err); | |
981 | if (skb == NULL) | |
982 | return -ENOMEM; | |
983 | ||
984 | /* reserve space for Hardware header */ | |
985 | skb_reserve(skb, hh_len); | |
986 | ||
987 | /* create space for UDP/IP header */ | |
988 | skb_put(skb,fragheaderlen + transhdrlen); | |
989 | ||
990 | /* initialize network header pointer */ | |
c1d2bbe1 | 991 | skb_reset_network_header(skb); |
e89e9cf5 AR |
992 | |
993 | /* initialize protocol header pointer */ | |
994 | skb->h.raw = skb->data + fragheaderlen; | |
995 | ||
84fa7933 | 996 | skb->ip_summed = CHECKSUM_PARTIAL; |
e89e9cf5 AR |
997 | skb->csum = 0; |
998 | sk->sk_sndmsg_off = 0; | |
999 | } | |
1000 | ||
1001 | err = skb_append_datato_frags(sk,skb, getfrag, from, | |
1002 | (length - transhdrlen)); | |
1003 | if (!err) { | |
1004 | struct frag_hdr fhdr; | |
1005 | ||
1006 | /* specify the length of each IP datagram fragment*/ | |
1ab1457c | 1007 | skb_shinfo(skb)->gso_size = mtu - fragheaderlen - |
7967168c | 1008 | sizeof(struct frag_hdr); |
f83ef8c0 | 1009 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP; |
e89e9cf5 AR |
1010 | ipv6_select_ident(skb, &fhdr); |
1011 | skb_shinfo(skb)->ip6_frag_id = fhdr.identification; | |
1012 | __skb_queue_tail(&sk->sk_write_queue, skb); | |
1013 | ||
1014 | return 0; | |
1015 | } | |
1016 | /* There is not enough support do UPD LSO, | |
1017 | * so follow normal path | |
1018 | */ | |
1019 | kfree_skb(skb); | |
1020 | ||
1021 | return err; | |
1022 | } | |
1da177e4 | 1023 | |
41a1f8ea YH |
1024 | int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to, |
1025 | int offset, int len, int odd, struct sk_buff *skb), | |
1026 | void *from, int length, int transhdrlen, | |
1027 | int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi *fl, | |
1028 | struct rt6_info *rt, unsigned int flags) | |
1da177e4 LT |
1029 | { |
1030 | struct inet_sock *inet = inet_sk(sk); | |
1031 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1032 | struct sk_buff *skb; | |
1033 | unsigned int maxfraglen, fragheaderlen; | |
1034 | int exthdrlen; | |
1035 | int hh_len; | |
1036 | int mtu; | |
1037 | int copy; | |
1038 | int err; | |
1039 | int offset = 0; | |
1040 | int csummode = CHECKSUM_NONE; | |
1041 | ||
1042 | if (flags&MSG_PROBE) | |
1043 | return 0; | |
1044 | if (skb_queue_empty(&sk->sk_write_queue)) { | |
1045 | /* | |
1046 | * setup for corking | |
1047 | */ | |
1048 | if (opt) { | |
1049 | if (np->cork.opt == NULL) { | |
1050 | np->cork.opt = kmalloc(opt->tot_len, | |
1051 | sk->sk_allocation); | |
1052 | if (unlikely(np->cork.opt == NULL)) | |
1053 | return -ENOBUFS; | |
1054 | } else if (np->cork.opt->tot_len < opt->tot_len) { | |
1055 | printk(KERN_DEBUG "ip6_append_data: invalid option length\n"); | |
1056 | return -EINVAL; | |
1057 | } | |
1058 | memcpy(np->cork.opt, opt, opt->tot_len); | |
1059 | inet->cork.flags |= IPCORK_OPT; | |
1060 | /* need source address above miyazawa*/ | |
1061 | } | |
1062 | dst_hold(&rt->u.dst); | |
1063 | np->cork.rt = rt; | |
1064 | inet->cork.fl = *fl; | |
1065 | np->cork.hop_limit = hlimit; | |
41a1f8ea | 1066 | np->cork.tclass = tclass; |
d91675f9 | 1067 | mtu = dst_mtu(rt->u.dst.path); |
c7503609 | 1068 | if (np->frag_size < mtu) { |
d91675f9 YH |
1069 | if (np->frag_size) |
1070 | mtu = np->frag_size; | |
1071 | } | |
1072 | inet->cork.fragsize = mtu; | |
1da177e4 LT |
1073 | if (dst_allfrag(rt->u.dst.path)) |
1074 | inet->cork.flags |= IPCORK_ALLFRAG; | |
1075 | inet->cork.length = 0; | |
1076 | sk->sk_sndmsg_page = NULL; | |
1077 | sk->sk_sndmsg_off = 0; | |
1078 | exthdrlen = rt->u.dst.header_len + (opt ? opt->opt_flen : 0); | |
1079 | length += exthdrlen; | |
1080 | transhdrlen += exthdrlen; | |
1081 | } else { | |
1082 | rt = np->cork.rt; | |
1083 | fl = &inet->cork.fl; | |
1084 | if (inet->cork.flags & IPCORK_OPT) | |
1085 | opt = np->cork.opt; | |
1086 | transhdrlen = 0; | |
1087 | exthdrlen = 0; | |
1088 | mtu = inet->cork.fragsize; | |
1089 | } | |
1090 | ||
1091 | hh_len = LL_RESERVED_SPACE(rt->u.dst.dev); | |
1092 | ||
1b5c2299 | 1093 | fragheaderlen = sizeof(struct ipv6hdr) + rt->u.dst.nfheader_len + (opt ? opt->opt_nflen : 0); |
1da177e4 LT |
1094 | maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr); |
1095 | ||
1096 | if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) { | |
1097 | if (inet->cork.length + length > sizeof(struct ipv6hdr) + IPV6_MAXPLEN - fragheaderlen) { | |
1098 | ipv6_local_error(sk, EMSGSIZE, fl, mtu-exthdrlen); | |
1099 | return -EMSGSIZE; | |
1100 | } | |
1101 | } | |
1102 | ||
1103 | /* | |
1104 | * Let's try using as much space as possible. | |
1105 | * Use MTU if total length of the message fits into the MTU. | |
1106 | * Otherwise, we need to reserve fragment header and | |
1107 | * fragment alignment (= 8-15 octects, in total). | |
1108 | * | |
1109 | * Note that we may need to "move" the data from the tail of | |
1ab1457c | 1110 | * of the buffer to the new fragment when we split |
1da177e4 LT |
1111 | * the message. |
1112 | * | |
1ab1457c | 1113 | * FIXME: It may be fragmented into multiple chunks |
1da177e4 LT |
1114 | * at once if non-fragmentable extension headers |
1115 | * are too large. | |
1ab1457c | 1116 | * --yoshfuji |
1da177e4 LT |
1117 | */ |
1118 | ||
1119 | inet->cork.length += length; | |
e89e9cf5 AR |
1120 | if (((length > mtu) && (sk->sk_protocol == IPPROTO_UDP)) && |
1121 | (rt->u.dst.dev->features & NETIF_F_UFO)) { | |
1122 | ||
baa829d8 PM |
1123 | err = ip6_ufo_append_data(sk, getfrag, from, length, hh_len, |
1124 | fragheaderlen, transhdrlen, mtu, | |
1125 | flags); | |
1126 | if (err) | |
e89e9cf5 | 1127 | goto error; |
e89e9cf5 AR |
1128 | return 0; |
1129 | } | |
1da177e4 LT |
1130 | |
1131 | if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) | |
1132 | goto alloc_new_skb; | |
1133 | ||
1134 | while (length > 0) { | |
1135 | /* Check if the remaining data fits into current packet. */ | |
1136 | copy = (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len; | |
1137 | if (copy < length) | |
1138 | copy = maxfraglen - skb->len; | |
1139 | ||
1140 | if (copy <= 0) { | |
1141 | char *data; | |
1142 | unsigned int datalen; | |
1143 | unsigned int fraglen; | |
1144 | unsigned int fraggap; | |
1145 | unsigned int alloclen; | |
1146 | struct sk_buff *skb_prev; | |
1147 | alloc_new_skb: | |
1148 | skb_prev = skb; | |
1149 | ||
1150 | /* There's no room in the current skb */ | |
1151 | if (skb_prev) | |
1152 | fraggap = skb_prev->len - maxfraglen; | |
1153 | else | |
1154 | fraggap = 0; | |
1155 | ||
1156 | /* | |
1157 | * If remaining data exceeds the mtu, | |
1158 | * we know we need more fragment(s). | |
1159 | */ | |
1160 | datalen = length + fraggap; | |
1161 | if (datalen > (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen) | |
1162 | datalen = maxfraglen - fragheaderlen; | |
1163 | ||
1164 | fraglen = datalen + fragheaderlen; | |
1165 | if ((flags & MSG_MORE) && | |
1166 | !(rt->u.dst.dev->features&NETIF_F_SG)) | |
1167 | alloclen = mtu; | |
1168 | else | |
1169 | alloclen = datalen + fragheaderlen; | |
1170 | ||
1171 | /* | |
1172 | * The last fragment gets additional space at tail. | |
1173 | * Note: we overallocate on fragments with MSG_MODE | |
1174 | * because we have no idea if we're the last one. | |
1175 | */ | |
1176 | if (datalen == length + fraggap) | |
1177 | alloclen += rt->u.dst.trailer_len; | |
1178 | ||
1179 | /* | |
1180 | * We just reserve space for fragment header. | |
1ab1457c | 1181 | * Note: this may be overallocation if the message |
1da177e4 LT |
1182 | * (without MSG_MORE) fits into the MTU. |
1183 | */ | |
1184 | alloclen += sizeof(struct frag_hdr); | |
1185 | ||
1186 | if (transhdrlen) { | |
1187 | skb = sock_alloc_send_skb(sk, | |
1188 | alloclen + hh_len, | |
1189 | (flags & MSG_DONTWAIT), &err); | |
1190 | } else { | |
1191 | skb = NULL; | |
1192 | if (atomic_read(&sk->sk_wmem_alloc) <= | |
1193 | 2 * sk->sk_sndbuf) | |
1194 | skb = sock_wmalloc(sk, | |
1195 | alloclen + hh_len, 1, | |
1196 | sk->sk_allocation); | |
1197 | if (unlikely(skb == NULL)) | |
1198 | err = -ENOBUFS; | |
1199 | } | |
1200 | if (skb == NULL) | |
1201 | goto error; | |
1202 | /* | |
1203 | * Fill in the control structures | |
1204 | */ | |
1205 | skb->ip_summed = csummode; | |
1206 | skb->csum = 0; | |
1207 | /* reserve for fragmentation */ | |
1208 | skb_reserve(skb, hh_len+sizeof(struct frag_hdr)); | |
1209 | ||
1210 | /* | |
1211 | * Find where to start putting bytes | |
1212 | */ | |
1213 | data = skb_put(skb, fraglen); | |
c14d2450 | 1214 | skb_set_network_header(skb, exthdrlen); |
1da177e4 | 1215 | data += fragheaderlen; |
c14d2450 | 1216 | skb->h.raw = skb->nh.raw + fragheaderlen; |
1da177e4 LT |
1217 | |
1218 | if (fraggap) { | |
1219 | skb->csum = skb_copy_and_csum_bits( | |
1220 | skb_prev, maxfraglen, | |
1221 | data + transhdrlen, fraggap, 0); | |
1222 | skb_prev->csum = csum_sub(skb_prev->csum, | |
1223 | skb->csum); | |
1224 | data += fraggap; | |
e9fa4f7b | 1225 | pskb_trim_unique(skb_prev, maxfraglen); |
1da177e4 LT |
1226 | } |
1227 | copy = datalen - transhdrlen - fraggap; | |
1228 | if (copy < 0) { | |
1229 | err = -EINVAL; | |
1230 | kfree_skb(skb); | |
1231 | goto error; | |
1232 | } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) { | |
1233 | err = -EFAULT; | |
1234 | kfree_skb(skb); | |
1235 | goto error; | |
1236 | } | |
1237 | ||
1238 | offset += copy; | |
1239 | length -= datalen - fraggap; | |
1240 | transhdrlen = 0; | |
1241 | exthdrlen = 0; | |
1242 | csummode = CHECKSUM_NONE; | |
1243 | ||
1244 | /* | |
1245 | * Put the packet on the pending queue | |
1246 | */ | |
1247 | __skb_queue_tail(&sk->sk_write_queue, skb); | |
1248 | continue; | |
1249 | } | |
1250 | ||
1251 | if (copy > length) | |
1252 | copy = length; | |
1253 | ||
1254 | if (!(rt->u.dst.dev->features&NETIF_F_SG)) { | |
1255 | unsigned int off; | |
1256 | ||
1257 | off = skb->len; | |
1258 | if (getfrag(from, skb_put(skb, copy), | |
1259 | offset, copy, off, skb) < 0) { | |
1260 | __skb_trim(skb, off); | |
1261 | err = -EFAULT; | |
1262 | goto error; | |
1263 | } | |
1264 | } else { | |
1265 | int i = skb_shinfo(skb)->nr_frags; | |
1266 | skb_frag_t *frag = &skb_shinfo(skb)->frags[i-1]; | |
1267 | struct page *page = sk->sk_sndmsg_page; | |
1268 | int off = sk->sk_sndmsg_off; | |
1269 | unsigned int left; | |
1270 | ||
1271 | if (page && (left = PAGE_SIZE - off) > 0) { | |
1272 | if (copy >= left) | |
1273 | copy = left; | |
1274 | if (page != frag->page) { | |
1275 | if (i == MAX_SKB_FRAGS) { | |
1276 | err = -EMSGSIZE; | |
1277 | goto error; | |
1278 | } | |
1279 | get_page(page); | |
1280 | skb_fill_page_desc(skb, i, page, sk->sk_sndmsg_off, 0); | |
1281 | frag = &skb_shinfo(skb)->frags[i]; | |
1282 | } | |
1283 | } else if(i < MAX_SKB_FRAGS) { | |
1284 | if (copy > PAGE_SIZE) | |
1285 | copy = PAGE_SIZE; | |
1286 | page = alloc_pages(sk->sk_allocation, 0); | |
1287 | if (page == NULL) { | |
1288 | err = -ENOMEM; | |
1289 | goto error; | |
1290 | } | |
1291 | sk->sk_sndmsg_page = page; | |
1292 | sk->sk_sndmsg_off = 0; | |
1293 | ||
1294 | skb_fill_page_desc(skb, i, page, 0, 0); | |
1295 | frag = &skb_shinfo(skb)->frags[i]; | |
1296 | skb->truesize += PAGE_SIZE; | |
1297 | atomic_add(PAGE_SIZE, &sk->sk_wmem_alloc); | |
1298 | } else { | |
1299 | err = -EMSGSIZE; | |
1300 | goto error; | |
1301 | } | |
1302 | if (getfrag(from, page_address(frag->page)+frag->page_offset+frag->size, offset, copy, skb->len, skb) < 0) { | |
1303 | err = -EFAULT; | |
1304 | goto error; | |
1305 | } | |
1306 | sk->sk_sndmsg_off += copy; | |
1307 | frag->size += copy; | |
1308 | skb->len += copy; | |
1309 | skb->data_len += copy; | |
1310 | } | |
1311 | offset += copy; | |
1312 | length -= copy; | |
1313 | } | |
1314 | return 0; | |
1315 | error: | |
1316 | inet->cork.length -= length; | |
a11d206d | 1317 | IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
1318 | return err; |
1319 | } | |
1320 | ||
1321 | int ip6_push_pending_frames(struct sock *sk) | |
1322 | { | |
1323 | struct sk_buff *skb, *tmp_skb; | |
1324 | struct sk_buff **tail_skb; | |
1325 | struct in6_addr final_dst_buf, *final_dst = &final_dst_buf; | |
1326 | struct inet_sock *inet = inet_sk(sk); | |
1327 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1328 | struct ipv6hdr *hdr; | |
1329 | struct ipv6_txoptions *opt = np->cork.opt; | |
1330 | struct rt6_info *rt = np->cork.rt; | |
1331 | struct flowi *fl = &inet->cork.fl; | |
1332 | unsigned char proto = fl->proto; | |
1333 | int err = 0; | |
1334 | ||
1335 | if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL) | |
1336 | goto out; | |
1337 | tail_skb = &(skb_shinfo(skb)->frag_list); | |
1338 | ||
1339 | /* move skb->data to ip header from ext header */ | |
d56f90a7 | 1340 | if (skb->data < skb_network_header(skb)) |
bbe735e4 | 1341 | __skb_pull(skb, skb_network_offset(skb)); |
1da177e4 LT |
1342 | while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) { |
1343 | __skb_pull(tmp_skb, skb->h.raw - skb->nh.raw); | |
1344 | *tail_skb = tmp_skb; | |
1345 | tail_skb = &(tmp_skb->next); | |
1346 | skb->len += tmp_skb->len; | |
1347 | skb->data_len += tmp_skb->len; | |
1da177e4 LT |
1348 | skb->truesize += tmp_skb->truesize; |
1349 | __sock_put(tmp_skb->sk); | |
1350 | tmp_skb->destructor = NULL; | |
1351 | tmp_skb->sk = NULL; | |
1da177e4 LT |
1352 | } |
1353 | ||
1354 | ipv6_addr_copy(final_dst, &fl->fl6_dst); | |
1355 | __skb_pull(skb, skb->h.raw - skb->nh.raw); | |
1356 | if (opt && opt->opt_flen) | |
1357 | ipv6_push_frag_opts(skb, opt, &proto); | |
1358 | if (opt && opt->opt_nflen) | |
1359 | ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst); | |
1360 | ||
e2d1bca7 ACM |
1361 | skb_push(skb, sizeof(struct ipv6hdr)); |
1362 | skb_reset_network_header(skb); | |
0660e03f | 1363 | hdr = ipv6_hdr(skb); |
1ab1457c | 1364 | |
90bcaf7b | 1365 | *(__be32*)hdr = fl->fl6_flowlabel | |
41a1f8ea | 1366 | htonl(0x60000000 | ((int)np->cork.tclass << 20)); |
1da177e4 LT |
1367 | |
1368 | if (skb->len <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) | |
1369 | hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr)); | |
1370 | else | |
1371 | hdr->payload_len = 0; | |
1372 | hdr->hop_limit = np->cork.hop_limit; | |
1373 | hdr->nexthdr = proto; | |
1374 | ipv6_addr_copy(&hdr->saddr, &fl->fl6_src); | |
1375 | ipv6_addr_copy(&hdr->daddr, final_dst); | |
1376 | ||
a2c2064f PM |
1377 | skb->priority = sk->sk_priority; |
1378 | ||
1da177e4 | 1379 | skb->dst = dst_clone(&rt->u.dst); |
a11d206d | 1380 | IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS); |
1da177e4 LT |
1381 | err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, skb->dst->dev, dst_output); |
1382 | if (err) { | |
1383 | if (err > 0) | |
3320da89 | 1384 | err = np->recverr ? net_xmit_errno(err) : 0; |
1da177e4 LT |
1385 | if (err) |
1386 | goto error; | |
1387 | } | |
1388 | ||
1389 | out: | |
1390 | inet->cork.flags &= ~IPCORK_OPT; | |
a51482bd JJ |
1391 | kfree(np->cork.opt); |
1392 | np->cork.opt = NULL; | |
1da177e4 LT |
1393 | if (np->cork.rt) { |
1394 | dst_release(&np->cork.rt->u.dst); | |
1395 | np->cork.rt = NULL; | |
1396 | inet->cork.flags &= ~IPCORK_ALLFRAG; | |
1397 | } | |
1398 | memset(&inet->cork.fl, 0, sizeof(inet->cork.fl)); | |
1399 | return err; | |
1400 | error: | |
1401 | goto out; | |
1402 | } | |
1403 | ||
1404 | void ip6_flush_pending_frames(struct sock *sk) | |
1405 | { | |
1406 | struct inet_sock *inet = inet_sk(sk); | |
1407 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1408 | struct sk_buff *skb; | |
1409 | ||
1410 | while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) { | |
a11d206d YH |
1411 | IP6_INC_STATS(ip6_dst_idev(skb->dst), |
1412 | IPSTATS_MIB_OUTDISCARDS); | |
1da177e4 LT |
1413 | kfree_skb(skb); |
1414 | } | |
1415 | ||
1416 | inet->cork.flags &= ~IPCORK_OPT; | |
1417 | ||
a51482bd JJ |
1418 | kfree(np->cork.opt); |
1419 | np->cork.opt = NULL; | |
1da177e4 LT |
1420 | if (np->cork.rt) { |
1421 | dst_release(&np->cork.rt->u.dst); | |
1422 | np->cork.rt = NULL; | |
1423 | inet->cork.flags &= ~IPCORK_ALLFRAG; | |
1424 | } | |
1425 | memset(&inet->cork.fl, 0, sizeof(inet->cork.fl)); | |
1426 | } |