Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * IPv6 output functions | |
3 | * Linux INET6 implementation | |
4 | * | |
5 | * Authors: | |
6 | * Pedro Roque <roque@di.fc.ul.pt> | |
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 | { | |
91 | newskb->mac.raw = newskb->data; | |
92 | __skb_pull(newskb, newskb->nh.raw - newskb->data); | |
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 | ||
110 | if (ipv6_addr_is_multicast(&skb->nh.ipv6h->daddr)) { | |
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) && | |
115 | ipv6_chk_mcast_addr(dev, &skb->nh.ipv6h->daddr, | |
116 | &skb->nh.ipv6h->saddr)) { | |
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 | ||
127 | if (skb->nh.ipv6h->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 | ||
194 | hdr = skb->nh.ipv6h = (struct ipv6hdr*)skb_push(skb, sizeof(struct ipv6hdr)); | |
195 | ||
196 | /* | |
197 | * Fill in the IPv6 header | |
198 | */ | |
199 | ||
1da177e4 LT |
200 | hlimit = -1; |
201 | if (np) | |
202 | hlimit = np->hop_limit; | |
203 | if (hlimit < 0) | |
204 | hlimit = dst_metric(dst, RTAX_HOPLIMIT); | |
205 | if (hlimit < 0) | |
206 | hlimit = ipv6_get_hoplimit(dst->dev); | |
207 | ||
41a1f8ea YH |
208 | tclass = -1; |
209 | if (np) | |
210 | tclass = np->tclass; | |
211 | if (tclass < 0) | |
212 | tclass = 0; | |
213 | ||
90bcaf7b | 214 | *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | fl->fl6_flowlabel; |
41a1f8ea | 215 | |
1da177e4 LT |
216 | hdr->payload_len = htons(seg_len); |
217 | hdr->nexthdr = proto; | |
218 | hdr->hop_limit = hlimit; | |
219 | ||
220 | ipv6_addr_copy(&hdr->saddr, &fl->fl6_src); | |
221 | ipv6_addr_copy(&hdr->daddr, first_hop); | |
222 | ||
a2c2064f PM |
223 | skb->priority = sk->sk_priority; |
224 | ||
1da177e4 | 225 | mtu = dst_mtu(dst); |
89114afd | 226 | if ((skb->len <= mtu) || ipfragok || skb_is_gso(skb)) { |
a11d206d YH |
227 | IP6_INC_STATS(ip6_dst_idev(skb->dst), |
228 | IPSTATS_MIB_OUTREQUESTS); | |
6869c4d8 HW |
229 | return NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev, |
230 | dst_output); | |
1da177e4 LT |
231 | } |
232 | ||
233 | if (net_ratelimit()) | |
234 | printk(KERN_DEBUG "IPv6: sending pkt_too_big to self\n"); | |
235 | skb->dev = dst->dev; | |
236 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, skb->dev); | |
a11d206d | 237 | IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
238 | kfree_skb(skb); |
239 | return -EMSGSIZE; | |
240 | } | |
241 | ||
242 | /* | |
243 | * To avoid extra problems ND packets are send through this | |
244 | * routine. It's code duplication but I really want to avoid | |
245 | * extra checks since ipv6_build_header is used by TCP (which | |
246 | * is for us performance critical) | |
247 | */ | |
248 | ||
249 | int ip6_nd_hdr(struct sock *sk, struct sk_buff *skb, struct net_device *dev, | |
250 | struct in6_addr *saddr, struct in6_addr *daddr, | |
251 | int proto, int len) | |
252 | { | |
253 | struct ipv6_pinfo *np = inet6_sk(sk); | |
254 | struct ipv6hdr *hdr; | |
255 | int totlen; | |
256 | ||
257 | skb->protocol = htons(ETH_P_IPV6); | |
258 | skb->dev = dev; | |
259 | ||
260 | totlen = len + sizeof(struct ipv6hdr); | |
261 | ||
262 | hdr = (struct ipv6hdr *) skb_put(skb, sizeof(struct ipv6hdr)); | |
263 | skb->nh.ipv6h = hdr; | |
264 | ||
ae08e1f0 | 265 | *(__be32*)hdr = htonl(0x60000000); |
1da177e4 LT |
266 | |
267 | hdr->payload_len = htons(len); | |
268 | hdr->nexthdr = proto; | |
269 | hdr->hop_limit = np->hop_limit; | |
270 | ||
271 | ipv6_addr_copy(&hdr->saddr, saddr); | |
272 | ipv6_addr_copy(&hdr->daddr, daddr); | |
273 | ||
274 | return 0; | |
275 | } | |
276 | ||
277 | static int ip6_call_ra_chain(struct sk_buff *skb, int sel) | |
278 | { | |
279 | struct ip6_ra_chain *ra; | |
280 | struct sock *last = NULL; | |
281 | ||
282 | read_lock(&ip6_ra_lock); | |
283 | for (ra = ip6_ra_chain; ra; ra = ra->next) { | |
284 | struct sock *sk = ra->sk; | |
0bd1b59b AM |
285 | if (sk && ra->sel == sel && |
286 | (!sk->sk_bound_dev_if || | |
287 | sk->sk_bound_dev_if == skb->dev->ifindex)) { | |
1da177e4 LT |
288 | if (last) { |
289 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
290 | if (skb2) | |
291 | rawv6_rcv(last, skb2); | |
292 | } | |
293 | last = sk; | |
294 | } | |
295 | } | |
296 | ||
297 | if (last) { | |
298 | rawv6_rcv(last, skb); | |
299 | read_unlock(&ip6_ra_lock); | |
300 | return 1; | |
301 | } | |
302 | read_unlock(&ip6_ra_lock); | |
303 | return 0; | |
304 | } | |
305 | ||
e21e0b5f VN |
306 | static int ip6_forward_proxy_check(struct sk_buff *skb) |
307 | { | |
308 | struct ipv6hdr *hdr = skb->nh.ipv6h; | |
309 | u8 nexthdr = hdr->nexthdr; | |
310 | int offset; | |
311 | ||
312 | if (ipv6_ext_hdr(nexthdr)) { | |
313 | offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr); | |
314 | if (offset < 0) | |
315 | return 0; | |
316 | } else | |
317 | offset = sizeof(struct ipv6hdr); | |
318 | ||
319 | if (nexthdr == IPPROTO_ICMPV6) { | |
320 | struct icmp6hdr *icmp6; | |
321 | ||
322 | if (!pskb_may_pull(skb, skb->nh.raw + offset + 1 - skb->data)) | |
323 | return 0; | |
324 | ||
325 | icmp6 = (struct icmp6hdr *)(skb->nh.raw + offset); | |
326 | ||
327 | switch (icmp6->icmp6_type) { | |
328 | case NDISC_ROUTER_SOLICITATION: | |
329 | case NDISC_ROUTER_ADVERTISEMENT: | |
330 | case NDISC_NEIGHBOUR_SOLICITATION: | |
331 | case NDISC_NEIGHBOUR_ADVERTISEMENT: | |
332 | case NDISC_REDIRECT: | |
333 | /* For reaction involving unicast neighbor discovery | |
334 | * message destined to the proxied address, pass it to | |
335 | * input function. | |
336 | */ | |
337 | return 1; | |
338 | default: | |
339 | break; | |
340 | } | |
341 | } | |
342 | ||
74553b09 VN |
343 | /* |
344 | * The proxying router can't forward traffic sent to a link-local | |
345 | * address, so signal the sender and discard the packet. This | |
346 | * behavior is clarified by the MIPv6 specification. | |
347 | */ | |
348 | if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) { | |
349 | dst_link_failure(skb); | |
350 | return -1; | |
351 | } | |
352 | ||
e21e0b5f VN |
353 | return 0; |
354 | } | |
355 | ||
1da177e4 LT |
356 | static inline int ip6_forward_finish(struct sk_buff *skb) |
357 | { | |
358 | return dst_output(skb); | |
359 | } | |
360 | ||
361 | int ip6_forward(struct sk_buff *skb) | |
362 | { | |
363 | struct dst_entry *dst = skb->dst; | |
364 | struct ipv6hdr *hdr = skb->nh.ipv6h; | |
365 | struct inet6_skb_parm *opt = IP6CB(skb); | |
366 | ||
367 | if (ipv6_devconf.forwarding == 0) | |
368 | goto error; | |
369 | ||
370 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) { | |
a11d206d | 371 | IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS); |
1da177e4 LT |
372 | goto drop; |
373 | } | |
374 | ||
375 | skb->ip_summed = CHECKSUM_NONE; | |
376 | ||
377 | /* | |
378 | * We DO NOT make any processing on | |
379 | * RA packets, pushing them to user level AS IS | |
380 | * without ane WARRANTY that application will be able | |
381 | * to interpret them. The reason is that we | |
382 | * cannot make anything clever here. | |
383 | * | |
384 | * We are not end-node, so that if packet contains | |
385 | * AH/ESP, we cannot make anything. | |
386 | * Defragmentation also would be mistake, RA packets | |
387 | * cannot be fragmented, because there is no warranty | |
388 | * that different fragments will go along one path. --ANK | |
389 | */ | |
390 | if (opt->ra) { | |
391 | u8 *ptr = skb->nh.raw + opt->ra; | |
392 | if (ip6_call_ra_chain(skb, (ptr[2]<<8) + ptr[3])) | |
393 | return 0; | |
394 | } | |
395 | ||
396 | /* | |
397 | * check and decrement ttl | |
398 | */ | |
399 | if (hdr->hop_limit <= 1) { | |
400 | /* Force OUTPUT device used as source address */ | |
401 | skb->dev = dst->dev; | |
402 | icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, | |
403 | 0, skb->dev); | |
a11d206d | 404 | IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS); |
1da177e4 LT |
405 | |
406 | kfree_skb(skb); | |
407 | return -ETIMEDOUT; | |
408 | } | |
409 | ||
fbea49e1 YH |
410 | /* XXX: idev->cnf.proxy_ndp? */ |
411 | if (ipv6_devconf.proxy_ndp && | |
412 | pneigh_lookup(&nd_tbl, &hdr->daddr, skb->dev, 0)) { | |
74553b09 VN |
413 | int proxied = ip6_forward_proxy_check(skb); |
414 | if (proxied > 0) | |
e21e0b5f | 415 | return ip6_input(skb); |
74553b09 | 416 | else if (proxied < 0) { |
a11d206d | 417 | IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS); |
74553b09 VN |
418 | goto drop; |
419 | } | |
e21e0b5f VN |
420 | } |
421 | ||
1da177e4 | 422 | if (!xfrm6_route_forward(skb)) { |
a11d206d | 423 | IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS); |
1da177e4 LT |
424 | goto drop; |
425 | } | |
426 | dst = skb->dst; | |
427 | ||
428 | /* IPv6 specs say nothing about it, but it is clear that we cannot | |
429 | send redirects to source routed frames. | |
430 | */ | |
431 | if (skb->dev == dst->dev && dst->neighbour && opt->srcrt == 0) { | |
432 | struct in6_addr *target = NULL; | |
433 | struct rt6_info *rt; | |
434 | struct neighbour *n = dst->neighbour; | |
435 | ||
436 | /* | |
437 | * incoming and outgoing devices are the same | |
438 | * send a redirect. | |
439 | */ | |
440 | ||
441 | rt = (struct rt6_info *) dst; | |
442 | if ((rt->rt6i_flags & RTF_GATEWAY)) | |
443 | target = (struct in6_addr*)&n->primary_key; | |
444 | else | |
445 | target = &hdr->daddr; | |
446 | ||
447 | /* Limit redirects both by destination (here) | |
448 | and by source (inside ndisc_send_redirect) | |
449 | */ | |
450 | if (xrlim_allow(dst, 1*HZ)) | |
451 | ndisc_send_redirect(skb, n, target); | |
452 | } else if (ipv6_addr_type(&hdr->saddr)&(IPV6_ADDR_MULTICAST|IPV6_ADDR_LOOPBACK | |
453 | |IPV6_ADDR_LINKLOCAL)) { | |
454 | /* This check is security critical. */ | |
455 | goto error; | |
456 | } | |
457 | ||
458 | if (skb->len > dst_mtu(dst)) { | |
459 | /* Again, force OUTPUT device used as source address */ | |
460 | skb->dev = dst->dev; | |
461 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, dst_mtu(dst), skb->dev); | |
a11d206d YH |
462 | IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INTOOBIGERRORS); |
463 | IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_FRAGFAILS); | |
1da177e4 LT |
464 | kfree_skb(skb); |
465 | return -EMSGSIZE; | |
466 | } | |
467 | ||
468 | if (skb_cow(skb, dst->dev->hard_header_len)) { | |
a11d206d | 469 | IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
470 | goto drop; |
471 | } | |
472 | ||
473 | hdr = skb->nh.ipv6h; | |
474 | ||
475 | /* Mangling hops number delayed to point after skb COW */ | |
476 | ||
477 | hdr->hop_limit--; | |
478 | ||
a11d206d | 479 | IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS); |
1da177e4 LT |
480 | return NF_HOOK(PF_INET6,NF_IP6_FORWARD, skb, skb->dev, dst->dev, ip6_forward_finish); |
481 | ||
482 | error: | |
a11d206d | 483 | IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS); |
1da177e4 LT |
484 | drop: |
485 | kfree_skb(skb); | |
486 | return -EINVAL; | |
487 | } | |
488 | ||
489 | static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from) | |
490 | { | |
491 | to->pkt_type = from->pkt_type; | |
492 | to->priority = from->priority; | |
493 | to->protocol = from->protocol; | |
1da177e4 LT |
494 | dst_release(to->dst); |
495 | to->dst = dst_clone(from->dst); | |
496 | to->dev = from->dev; | |
82e91ffe | 497 | to->mark = from->mark; |
1da177e4 LT |
498 | |
499 | #ifdef CONFIG_NET_SCHED | |
500 | to->tc_index = from->tc_index; | |
501 | #endif | |
502 | #ifdef CONFIG_NETFILTER | |
1da177e4 | 503 | /* Connection association is same as pre-frag packet */ |
9fb9cbb1 | 504 | nf_conntrack_put(to->nfct); |
1da177e4 LT |
505 | to->nfct = from->nfct; |
506 | nf_conntrack_get(to->nfct); | |
507 | to->nfctinfo = from->nfctinfo; | |
9fb9cbb1 YK |
508 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
509 | nf_conntrack_put_reasm(to->nfct_reasm); | |
510 | to->nfct_reasm = from->nfct_reasm; | |
511 | nf_conntrack_get_reasm(to->nfct_reasm); | |
512 | #endif | |
1da177e4 LT |
513 | #ifdef CONFIG_BRIDGE_NETFILTER |
514 | nf_bridge_put(to->nf_bridge); | |
515 | to->nf_bridge = from->nf_bridge; | |
516 | nf_bridge_get(to->nf_bridge); | |
517 | #endif | |
1da177e4 | 518 | #endif |
984bc16c | 519 | skb_copy_secmark(to, from); |
1da177e4 LT |
520 | } |
521 | ||
522 | int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr) | |
523 | { | |
524 | u16 offset = sizeof(struct ipv6hdr); | |
525 | struct ipv6_opt_hdr *exthdr = (struct ipv6_opt_hdr*)(skb->nh.ipv6h + 1); | |
526 | unsigned int packet_len = skb->tail - skb->nh.raw; | |
527 | int found_rhdr = 0; | |
528 | *nexthdr = &skb->nh.ipv6h->nexthdr; | |
529 | ||
530 | while (offset + 1 <= packet_len) { | |
531 | ||
532 | switch (**nexthdr) { | |
533 | ||
534 | case NEXTHDR_HOP: | |
27637df9 | 535 | break; |
1da177e4 | 536 | case NEXTHDR_ROUTING: |
27637df9 MN |
537 | found_rhdr = 1; |
538 | break; | |
1da177e4 | 539 | case NEXTHDR_DEST: |
27637df9 MN |
540 | #ifdef CONFIG_IPV6_MIP6 |
541 | if (ipv6_find_tlv(skb, offset, IPV6_TLV_HAO) >= 0) | |
542 | break; | |
543 | #endif | |
544 | if (found_rhdr) | |
545 | return offset; | |
1da177e4 LT |
546 | break; |
547 | default : | |
548 | return offset; | |
549 | } | |
27637df9 MN |
550 | |
551 | offset += ipv6_optlen(exthdr); | |
552 | *nexthdr = &exthdr->nexthdr; | |
553 | exthdr = (struct ipv6_opt_hdr*)(skb->nh.raw + offset); | |
1da177e4 LT |
554 | } |
555 | ||
556 | return offset; | |
557 | } | |
b59f45d0 | 558 | EXPORT_SYMBOL_GPL(ip6_find_1stfragopt); |
1da177e4 LT |
559 | |
560 | static int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *)) | |
561 | { | |
562 | struct net_device *dev; | |
563 | struct sk_buff *frag; | |
564 | struct rt6_info *rt = (struct rt6_info*)skb->dst; | |
d91675f9 | 565 | struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL; |
1da177e4 LT |
566 | struct ipv6hdr *tmp_hdr; |
567 | struct frag_hdr *fh; | |
568 | unsigned int mtu, hlen, left, len; | |
ae08e1f0 | 569 | __be32 frag_id = 0; |
1da177e4 LT |
570 | int ptr, offset = 0, err=0; |
571 | u8 *prevhdr, nexthdr = 0; | |
572 | ||
573 | dev = rt->u.dst.dev; | |
574 | hlen = ip6_find_1stfragopt(skb, &prevhdr); | |
575 | nexthdr = *prevhdr; | |
576 | ||
d91675f9 YH |
577 | mtu = dst_mtu(&rt->u.dst); |
578 | if (np && np->frag_size < mtu) { | |
579 | if (np->frag_size) | |
580 | mtu = np->frag_size; | |
581 | } | |
582 | mtu -= hlen + sizeof(struct frag_hdr); | |
1da177e4 LT |
583 | |
584 | if (skb_shinfo(skb)->frag_list) { | |
585 | int first_len = skb_pagelen(skb); | |
586 | ||
587 | if (first_len - hlen > mtu || | |
588 | ((first_len - hlen) & 7) || | |
589 | skb_cloned(skb)) | |
590 | goto slow_path; | |
591 | ||
592 | for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next) { | |
593 | /* Correct geometry. */ | |
594 | if (frag->len > mtu || | |
595 | ((frag->len & 7) && frag->next) || | |
596 | skb_headroom(frag) < hlen) | |
597 | goto slow_path; | |
598 | ||
1da177e4 LT |
599 | /* Partially cloned skb? */ |
600 | if (skb_shared(frag)) | |
601 | goto slow_path; | |
2fdba6b0 HX |
602 | |
603 | BUG_ON(frag->sk); | |
604 | if (skb->sk) { | |
605 | sock_hold(skb->sk); | |
606 | frag->sk = skb->sk; | |
607 | frag->destructor = sock_wfree; | |
608 | skb->truesize -= frag->truesize; | |
609 | } | |
1da177e4 LT |
610 | } |
611 | ||
612 | err = 0; | |
613 | offset = 0; | |
614 | frag = skb_shinfo(skb)->frag_list; | |
615 | skb_shinfo(skb)->frag_list = NULL; | |
616 | /* BUILD HEADER */ | |
617 | ||
9a217a1c | 618 | *prevhdr = NEXTHDR_FRAGMENT; |
af879cc7 | 619 | tmp_hdr = kmemdup(skb->nh.raw, hlen, GFP_ATOMIC); |
1da177e4 | 620 | if (!tmp_hdr) { |
a11d206d | 621 | IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
622 | return -ENOMEM; |
623 | } | |
624 | ||
1da177e4 LT |
625 | __skb_pull(skb, hlen); |
626 | fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr)); | |
627 | skb->nh.raw = __skb_push(skb, hlen); | |
628 | memcpy(skb->nh.raw, tmp_hdr, hlen); | |
629 | ||
630 | ipv6_select_ident(skb, fh); | |
631 | fh->nexthdr = nexthdr; | |
632 | fh->reserved = 0; | |
633 | fh->frag_off = htons(IP6_MF); | |
634 | frag_id = fh->identification; | |
635 | ||
636 | first_len = skb_pagelen(skb); | |
637 | skb->data_len = first_len - skb_headlen(skb); | |
638 | skb->len = first_len; | |
639 | skb->nh.ipv6h->payload_len = htons(first_len - sizeof(struct ipv6hdr)); | |
a11d206d YH |
640 | |
641 | dst_hold(&rt->u.dst); | |
1da177e4 LT |
642 | |
643 | for (;;) { | |
644 | /* Prepare header of the next frame, | |
645 | * before previous one went down. */ | |
646 | if (frag) { | |
647 | frag->ip_summed = CHECKSUM_NONE; | |
648 | frag->h.raw = frag->data; | |
649 | fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr)); | |
650 | frag->nh.raw = __skb_push(frag, hlen); | |
651 | memcpy(frag->nh.raw, tmp_hdr, hlen); | |
652 | offset += skb->len - hlen - sizeof(struct frag_hdr); | |
653 | fh->nexthdr = nexthdr; | |
654 | fh->reserved = 0; | |
655 | fh->frag_off = htons(offset); | |
656 | if (frag->next != NULL) | |
657 | fh->frag_off |= htons(IP6_MF); | |
658 | fh->identification = frag_id; | |
659 | frag->nh.ipv6h->payload_len = htons(frag->len - sizeof(struct ipv6hdr)); | |
660 | ip6_copy_metadata(frag, skb); | |
661 | } | |
662 | ||
663 | err = output(skb); | |
dafee490 | 664 | if(!err) |
a11d206d | 665 | IP6_INC_STATS(ip6_dst_idev(&rt->u.dst), IPSTATS_MIB_FRAGCREATES); |
dafee490 | 666 | |
1da177e4 LT |
667 | if (err || !frag) |
668 | break; | |
669 | ||
670 | skb = frag; | |
671 | frag = skb->next; | |
672 | skb->next = NULL; | |
673 | } | |
674 | ||
a51482bd | 675 | kfree(tmp_hdr); |
1da177e4 LT |
676 | |
677 | if (err == 0) { | |
a11d206d YH |
678 | IP6_INC_STATS(ip6_dst_idev(&rt->u.dst), IPSTATS_MIB_FRAGOKS); |
679 | dst_release(&rt->u.dst); | |
1da177e4 LT |
680 | return 0; |
681 | } | |
682 | ||
683 | while (frag) { | |
684 | skb = frag->next; | |
685 | kfree_skb(frag); | |
686 | frag = skb; | |
687 | } | |
688 | ||
a11d206d YH |
689 | IP6_INC_STATS(ip6_dst_idev(&rt->u.dst), IPSTATS_MIB_FRAGFAILS); |
690 | dst_release(&rt->u.dst); | |
1da177e4 LT |
691 | return err; |
692 | } | |
693 | ||
694 | slow_path: | |
695 | left = skb->len - hlen; /* Space per frame */ | |
696 | ptr = hlen; /* Where to start from */ | |
697 | ||
698 | /* | |
699 | * Fragment the datagram. | |
700 | */ | |
701 | ||
702 | *prevhdr = NEXTHDR_FRAGMENT; | |
703 | ||
704 | /* | |
705 | * Keep copying data until we run out. | |
706 | */ | |
707 | while(left > 0) { | |
708 | len = left; | |
709 | /* IF: it doesn't fit, use 'mtu' - the data space left */ | |
710 | if (len > mtu) | |
711 | len = mtu; | |
712 | /* IF: we are not sending upto and including the packet end | |
713 | then align the next start on an eight byte boundary */ | |
714 | if (len < left) { | |
715 | len &= ~7; | |
716 | } | |
717 | /* | |
718 | * Allocate buffer. | |
719 | */ | |
720 | ||
721 | if ((frag = alloc_skb(len+hlen+sizeof(struct frag_hdr)+LL_RESERVED_SPACE(rt->u.dst.dev), GFP_ATOMIC)) == NULL) { | |
64ce2073 | 722 | NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n"); |
a11d206d YH |
723 | IP6_INC_STATS(ip6_dst_idev(skb->dst), |
724 | IPSTATS_MIB_FRAGFAILS); | |
1da177e4 LT |
725 | err = -ENOMEM; |
726 | goto fail; | |
727 | } | |
728 | ||
729 | /* | |
730 | * Set up data on packet | |
731 | */ | |
732 | ||
733 | ip6_copy_metadata(frag, skb); | |
734 | skb_reserve(frag, LL_RESERVED_SPACE(rt->u.dst.dev)); | |
735 | skb_put(frag, len + hlen + sizeof(struct frag_hdr)); | |
736 | frag->nh.raw = frag->data; | |
737 | fh = (struct frag_hdr*)(frag->data + hlen); | |
738 | frag->h.raw = frag->data + hlen + sizeof(struct frag_hdr); | |
739 | ||
740 | /* | |
741 | * Charge the memory for the fragment to any owner | |
742 | * it might possess | |
743 | */ | |
744 | if (skb->sk) | |
745 | skb_set_owner_w(frag, skb->sk); | |
746 | ||
747 | /* | |
748 | * Copy the packet header into the new buffer. | |
749 | */ | |
750 | memcpy(frag->nh.raw, skb->data, hlen); | |
751 | ||
752 | /* | |
753 | * Build fragment header. | |
754 | */ | |
755 | fh->nexthdr = nexthdr; | |
756 | fh->reserved = 0; | |
f36d6ab1 | 757 | if (!frag_id) { |
1da177e4 LT |
758 | ipv6_select_ident(skb, fh); |
759 | frag_id = fh->identification; | |
760 | } else | |
761 | fh->identification = frag_id; | |
762 | ||
763 | /* | |
764 | * Copy a block of the IP datagram. | |
765 | */ | |
766 | if (skb_copy_bits(skb, ptr, frag->h.raw, len)) | |
767 | BUG(); | |
768 | left -= len; | |
769 | ||
770 | fh->frag_off = htons(offset); | |
771 | if (left > 0) | |
772 | fh->frag_off |= htons(IP6_MF); | |
773 | frag->nh.ipv6h->payload_len = htons(frag->len - sizeof(struct ipv6hdr)); | |
774 | ||
775 | ptr += len; | |
776 | offset += len; | |
777 | ||
778 | /* | |
779 | * Put this fragment into the sending queue. | |
780 | */ | |
1da177e4 LT |
781 | err = output(frag); |
782 | if (err) | |
783 | goto fail; | |
dafee490 | 784 | |
a11d206d | 785 | IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGCREATES); |
1da177e4 | 786 | } |
a11d206d YH |
787 | IP6_INC_STATS(ip6_dst_idev(skb->dst), |
788 | IPSTATS_MIB_FRAGOKS); | |
1da177e4 | 789 | kfree_skb(skb); |
1da177e4 LT |
790 | return err; |
791 | ||
792 | fail: | |
a11d206d YH |
793 | IP6_INC_STATS(ip6_dst_idev(skb->dst), |
794 | IPSTATS_MIB_FRAGFAILS); | |
1da177e4 | 795 | kfree_skb(skb); |
1da177e4 LT |
796 | return err; |
797 | } | |
798 | ||
cf6b1982 YH |
799 | static inline int ip6_rt_check(struct rt6key *rt_key, |
800 | struct in6_addr *fl_addr, | |
801 | struct in6_addr *addr_cache) | |
802 | { | |
803 | return ((rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) && | |
804 | (addr_cache == NULL || !ipv6_addr_equal(fl_addr, addr_cache))); | |
805 | } | |
806 | ||
497c615a HX |
807 | static struct dst_entry *ip6_sk_dst_check(struct sock *sk, |
808 | struct dst_entry *dst, | |
809 | struct flowi *fl) | |
1da177e4 | 810 | { |
497c615a HX |
811 | struct ipv6_pinfo *np = inet6_sk(sk); |
812 | struct rt6_info *rt = (struct rt6_info *)dst; | |
1da177e4 | 813 | |
497c615a HX |
814 | if (!dst) |
815 | goto out; | |
816 | ||
817 | /* Yes, checking route validity in not connected | |
818 | * case is not very simple. Take into account, | |
819 | * that we do not support routing by source, TOS, | |
820 | * and MSG_DONTROUTE --ANK (980726) | |
821 | * | |
cf6b1982 YH |
822 | * 1. ip6_rt_check(): If route was host route, |
823 | * check that cached destination is current. | |
497c615a HX |
824 | * If it is network route, we still may |
825 | * check its validity using saved pointer | |
826 | * to the last used address: daddr_cache. | |
827 | * We do not want to save whole address now, | |
828 | * (because main consumer of this service | |
829 | * is tcp, which has not this problem), | |
830 | * so that the last trick works only on connected | |
831 | * sockets. | |
832 | * 2. oif also should be the same. | |
833 | */ | |
cf6b1982 | 834 | if (ip6_rt_check(&rt->rt6i_dst, &fl->fl6_dst, np->daddr_cache) || |
8e1ef0a9 YH |
835 | #ifdef CONFIG_IPV6_SUBTREES |
836 | ip6_rt_check(&rt->rt6i_src, &fl->fl6_src, np->saddr_cache) || | |
837 | #endif | |
cf6b1982 | 838 | (fl->oif && fl->oif != dst->dev->ifindex)) { |
497c615a HX |
839 | dst_release(dst); |
840 | dst = NULL; | |
1da177e4 LT |
841 | } |
842 | ||
497c615a HX |
843 | out: |
844 | return dst; | |
845 | } | |
846 | ||
847 | static int ip6_dst_lookup_tail(struct sock *sk, | |
848 | struct dst_entry **dst, struct flowi *fl) | |
849 | { | |
850 | int err; | |
851 | ||
1da177e4 LT |
852 | if (*dst == NULL) |
853 | *dst = ip6_route_output(sk, fl); | |
854 | ||
855 | if ((err = (*dst)->error)) | |
856 | goto out_err_release; | |
857 | ||
858 | if (ipv6_addr_any(&fl->fl6_src)) { | |
859 | err = ipv6_get_saddr(*dst, &fl->fl6_dst, &fl->fl6_src); | |
44456d37 | 860 | if (err) |
1da177e4 | 861 | goto out_err_release; |
1da177e4 LT |
862 | } |
863 | ||
864 | return 0; | |
865 | ||
866 | out_err_release: | |
867 | dst_release(*dst); | |
868 | *dst = NULL; | |
869 | return err; | |
870 | } | |
34a0b3cd | 871 | |
497c615a HX |
872 | /** |
873 | * ip6_dst_lookup - perform route lookup on flow | |
874 | * @sk: socket which provides route info | |
875 | * @dst: pointer to dst_entry * for result | |
876 | * @fl: flow to lookup | |
877 | * | |
878 | * This function performs a route lookup on the given flow. | |
879 | * | |
880 | * It returns zero on success, or a standard errno code on error. | |
881 | */ | |
882 | int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl) | |
883 | { | |
884 | *dst = NULL; | |
885 | return ip6_dst_lookup_tail(sk, dst, fl); | |
886 | } | |
3cf3dc6c ACM |
887 | EXPORT_SYMBOL_GPL(ip6_dst_lookup); |
888 | ||
497c615a HX |
889 | /** |
890 | * ip6_sk_dst_lookup - perform socket cached route lookup on flow | |
891 | * @sk: socket which provides the dst cache and route info | |
892 | * @dst: pointer to dst_entry * for result | |
893 | * @fl: flow to lookup | |
894 | * | |
895 | * This function performs a route lookup on the given flow with the | |
896 | * possibility of using the cached route in the socket if it is valid. | |
897 | * It will take the socket dst lock when operating on the dst cache. | |
898 | * As a result, this function can only be used in process context. | |
899 | * | |
900 | * It returns zero on success, or a standard errno code on error. | |
901 | */ | |
902 | int ip6_sk_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl) | |
903 | { | |
904 | *dst = NULL; | |
905 | if (sk) { | |
906 | *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie); | |
907 | *dst = ip6_sk_dst_check(sk, *dst, fl); | |
908 | } | |
909 | ||
910 | return ip6_dst_lookup_tail(sk, dst, fl); | |
911 | } | |
912 | EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup); | |
913 | ||
34a0b3cd | 914 | static inline int ip6_ufo_append_data(struct sock *sk, |
e89e9cf5 AR |
915 | int getfrag(void *from, char *to, int offset, int len, |
916 | int odd, struct sk_buff *skb), | |
917 | void *from, int length, int hh_len, int fragheaderlen, | |
918 | int transhdrlen, int mtu,unsigned int flags) | |
919 | ||
920 | { | |
921 | struct sk_buff *skb; | |
922 | int err; | |
923 | ||
924 | /* There is support for UDP large send offload by network | |
925 | * device, so create one single skb packet containing complete | |
926 | * udp datagram | |
927 | */ | |
928 | if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) { | |
929 | skb = sock_alloc_send_skb(sk, | |
930 | hh_len + fragheaderlen + transhdrlen + 20, | |
931 | (flags & MSG_DONTWAIT), &err); | |
932 | if (skb == NULL) | |
933 | return -ENOMEM; | |
934 | ||
935 | /* reserve space for Hardware header */ | |
936 | skb_reserve(skb, hh_len); | |
937 | ||
938 | /* create space for UDP/IP header */ | |
939 | skb_put(skb,fragheaderlen + transhdrlen); | |
940 | ||
941 | /* initialize network header pointer */ | |
942 | skb->nh.raw = skb->data; | |
943 | ||
944 | /* initialize protocol header pointer */ | |
945 | skb->h.raw = skb->data + fragheaderlen; | |
946 | ||
84fa7933 | 947 | skb->ip_summed = CHECKSUM_PARTIAL; |
e89e9cf5 AR |
948 | skb->csum = 0; |
949 | sk->sk_sndmsg_off = 0; | |
950 | } | |
951 | ||
952 | err = skb_append_datato_frags(sk,skb, getfrag, from, | |
953 | (length - transhdrlen)); | |
954 | if (!err) { | |
955 | struct frag_hdr fhdr; | |
956 | ||
957 | /* specify the length of each IP datagram fragment*/ | |
7967168c HX |
958 | skb_shinfo(skb)->gso_size = mtu - fragheaderlen - |
959 | sizeof(struct frag_hdr); | |
f83ef8c0 | 960 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP; |
e89e9cf5 AR |
961 | ipv6_select_ident(skb, &fhdr); |
962 | skb_shinfo(skb)->ip6_frag_id = fhdr.identification; | |
963 | __skb_queue_tail(&sk->sk_write_queue, skb); | |
964 | ||
965 | return 0; | |
966 | } | |
967 | /* There is not enough support do UPD LSO, | |
968 | * so follow normal path | |
969 | */ | |
970 | kfree_skb(skb); | |
971 | ||
972 | return err; | |
973 | } | |
1da177e4 | 974 | |
41a1f8ea YH |
975 | int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to, |
976 | int offset, int len, int odd, struct sk_buff *skb), | |
977 | void *from, int length, int transhdrlen, | |
978 | int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi *fl, | |
979 | struct rt6_info *rt, unsigned int flags) | |
1da177e4 LT |
980 | { |
981 | struct inet_sock *inet = inet_sk(sk); | |
982 | struct ipv6_pinfo *np = inet6_sk(sk); | |
983 | struct sk_buff *skb; | |
984 | unsigned int maxfraglen, fragheaderlen; | |
985 | int exthdrlen; | |
986 | int hh_len; | |
987 | int mtu; | |
988 | int copy; | |
989 | int err; | |
990 | int offset = 0; | |
991 | int csummode = CHECKSUM_NONE; | |
992 | ||
993 | if (flags&MSG_PROBE) | |
994 | return 0; | |
995 | if (skb_queue_empty(&sk->sk_write_queue)) { | |
996 | /* | |
997 | * setup for corking | |
998 | */ | |
999 | if (opt) { | |
1000 | if (np->cork.opt == NULL) { | |
1001 | np->cork.opt = kmalloc(opt->tot_len, | |
1002 | sk->sk_allocation); | |
1003 | if (unlikely(np->cork.opt == NULL)) | |
1004 | return -ENOBUFS; | |
1005 | } else if (np->cork.opt->tot_len < opt->tot_len) { | |
1006 | printk(KERN_DEBUG "ip6_append_data: invalid option length\n"); | |
1007 | return -EINVAL; | |
1008 | } | |
1009 | memcpy(np->cork.opt, opt, opt->tot_len); | |
1010 | inet->cork.flags |= IPCORK_OPT; | |
1011 | /* need source address above miyazawa*/ | |
1012 | } | |
1013 | dst_hold(&rt->u.dst); | |
1014 | np->cork.rt = rt; | |
1015 | inet->cork.fl = *fl; | |
1016 | np->cork.hop_limit = hlimit; | |
41a1f8ea | 1017 | np->cork.tclass = tclass; |
d91675f9 | 1018 | mtu = dst_mtu(rt->u.dst.path); |
c7503609 | 1019 | if (np->frag_size < mtu) { |
d91675f9 YH |
1020 | if (np->frag_size) |
1021 | mtu = np->frag_size; | |
1022 | } | |
1023 | inet->cork.fragsize = mtu; | |
1da177e4 LT |
1024 | if (dst_allfrag(rt->u.dst.path)) |
1025 | inet->cork.flags |= IPCORK_ALLFRAG; | |
1026 | inet->cork.length = 0; | |
1027 | sk->sk_sndmsg_page = NULL; | |
1028 | sk->sk_sndmsg_off = 0; | |
1029 | exthdrlen = rt->u.dst.header_len + (opt ? opt->opt_flen : 0); | |
1030 | length += exthdrlen; | |
1031 | transhdrlen += exthdrlen; | |
1032 | } else { | |
1033 | rt = np->cork.rt; | |
1034 | fl = &inet->cork.fl; | |
1035 | if (inet->cork.flags & IPCORK_OPT) | |
1036 | opt = np->cork.opt; | |
1037 | transhdrlen = 0; | |
1038 | exthdrlen = 0; | |
1039 | mtu = inet->cork.fragsize; | |
1040 | } | |
1041 | ||
1042 | hh_len = LL_RESERVED_SPACE(rt->u.dst.dev); | |
1043 | ||
1b5c2299 | 1044 | fragheaderlen = sizeof(struct ipv6hdr) + rt->u.dst.nfheader_len + (opt ? opt->opt_nflen : 0); |
1da177e4 LT |
1045 | maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr); |
1046 | ||
1047 | if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) { | |
1048 | if (inet->cork.length + length > sizeof(struct ipv6hdr) + IPV6_MAXPLEN - fragheaderlen) { | |
1049 | ipv6_local_error(sk, EMSGSIZE, fl, mtu-exthdrlen); | |
1050 | return -EMSGSIZE; | |
1051 | } | |
1052 | } | |
1053 | ||
1054 | /* | |
1055 | * Let's try using as much space as possible. | |
1056 | * Use MTU if total length of the message fits into the MTU. | |
1057 | * Otherwise, we need to reserve fragment header and | |
1058 | * fragment alignment (= 8-15 octects, in total). | |
1059 | * | |
1060 | * Note that we may need to "move" the data from the tail of | |
1061 | * of the buffer to the new fragment when we split | |
1062 | * the message. | |
1063 | * | |
1064 | * FIXME: It may be fragmented into multiple chunks | |
1065 | * at once if non-fragmentable extension headers | |
1066 | * are too large. | |
1067 | * --yoshfuji | |
1068 | */ | |
1069 | ||
1070 | inet->cork.length += length; | |
e89e9cf5 AR |
1071 | if (((length > mtu) && (sk->sk_protocol == IPPROTO_UDP)) && |
1072 | (rt->u.dst.dev->features & NETIF_F_UFO)) { | |
1073 | ||
baa829d8 PM |
1074 | err = ip6_ufo_append_data(sk, getfrag, from, length, hh_len, |
1075 | fragheaderlen, transhdrlen, mtu, | |
1076 | flags); | |
1077 | if (err) | |
e89e9cf5 | 1078 | goto error; |
e89e9cf5 AR |
1079 | return 0; |
1080 | } | |
1da177e4 LT |
1081 | |
1082 | if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) | |
1083 | goto alloc_new_skb; | |
1084 | ||
1085 | while (length > 0) { | |
1086 | /* Check if the remaining data fits into current packet. */ | |
1087 | copy = (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len; | |
1088 | if (copy < length) | |
1089 | copy = maxfraglen - skb->len; | |
1090 | ||
1091 | if (copy <= 0) { | |
1092 | char *data; | |
1093 | unsigned int datalen; | |
1094 | unsigned int fraglen; | |
1095 | unsigned int fraggap; | |
1096 | unsigned int alloclen; | |
1097 | struct sk_buff *skb_prev; | |
1098 | alloc_new_skb: | |
1099 | skb_prev = skb; | |
1100 | ||
1101 | /* There's no room in the current skb */ | |
1102 | if (skb_prev) | |
1103 | fraggap = skb_prev->len - maxfraglen; | |
1104 | else | |
1105 | fraggap = 0; | |
1106 | ||
1107 | /* | |
1108 | * If remaining data exceeds the mtu, | |
1109 | * we know we need more fragment(s). | |
1110 | */ | |
1111 | datalen = length + fraggap; | |
1112 | if (datalen > (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen) | |
1113 | datalen = maxfraglen - fragheaderlen; | |
1114 | ||
1115 | fraglen = datalen + fragheaderlen; | |
1116 | if ((flags & MSG_MORE) && | |
1117 | !(rt->u.dst.dev->features&NETIF_F_SG)) | |
1118 | alloclen = mtu; | |
1119 | else | |
1120 | alloclen = datalen + fragheaderlen; | |
1121 | ||
1122 | /* | |
1123 | * The last fragment gets additional space at tail. | |
1124 | * Note: we overallocate on fragments with MSG_MODE | |
1125 | * because we have no idea if we're the last one. | |
1126 | */ | |
1127 | if (datalen == length + fraggap) | |
1128 | alloclen += rt->u.dst.trailer_len; | |
1129 | ||
1130 | /* | |
1131 | * We just reserve space for fragment header. | |
1132 | * Note: this may be overallocation if the message | |
1133 | * (without MSG_MORE) fits into the MTU. | |
1134 | */ | |
1135 | alloclen += sizeof(struct frag_hdr); | |
1136 | ||
1137 | if (transhdrlen) { | |
1138 | skb = sock_alloc_send_skb(sk, | |
1139 | alloclen + hh_len, | |
1140 | (flags & MSG_DONTWAIT), &err); | |
1141 | } else { | |
1142 | skb = NULL; | |
1143 | if (atomic_read(&sk->sk_wmem_alloc) <= | |
1144 | 2 * sk->sk_sndbuf) | |
1145 | skb = sock_wmalloc(sk, | |
1146 | alloclen + hh_len, 1, | |
1147 | sk->sk_allocation); | |
1148 | if (unlikely(skb == NULL)) | |
1149 | err = -ENOBUFS; | |
1150 | } | |
1151 | if (skb == NULL) | |
1152 | goto error; | |
1153 | /* | |
1154 | * Fill in the control structures | |
1155 | */ | |
1156 | skb->ip_summed = csummode; | |
1157 | skb->csum = 0; | |
1158 | /* reserve for fragmentation */ | |
1159 | skb_reserve(skb, hh_len+sizeof(struct frag_hdr)); | |
1160 | ||
1161 | /* | |
1162 | * Find where to start putting bytes | |
1163 | */ | |
1164 | data = skb_put(skb, fraglen); | |
1165 | skb->nh.raw = data + exthdrlen; | |
1166 | data += fragheaderlen; | |
1167 | skb->h.raw = data + exthdrlen; | |
1168 | ||
1169 | if (fraggap) { | |
1170 | skb->csum = skb_copy_and_csum_bits( | |
1171 | skb_prev, maxfraglen, | |
1172 | data + transhdrlen, fraggap, 0); | |
1173 | skb_prev->csum = csum_sub(skb_prev->csum, | |
1174 | skb->csum); | |
1175 | data += fraggap; | |
e9fa4f7b | 1176 | pskb_trim_unique(skb_prev, maxfraglen); |
1da177e4 LT |
1177 | } |
1178 | copy = datalen - transhdrlen - fraggap; | |
1179 | if (copy < 0) { | |
1180 | err = -EINVAL; | |
1181 | kfree_skb(skb); | |
1182 | goto error; | |
1183 | } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) { | |
1184 | err = -EFAULT; | |
1185 | kfree_skb(skb); | |
1186 | goto error; | |
1187 | } | |
1188 | ||
1189 | offset += copy; | |
1190 | length -= datalen - fraggap; | |
1191 | transhdrlen = 0; | |
1192 | exthdrlen = 0; | |
1193 | csummode = CHECKSUM_NONE; | |
1194 | ||
1195 | /* | |
1196 | * Put the packet on the pending queue | |
1197 | */ | |
1198 | __skb_queue_tail(&sk->sk_write_queue, skb); | |
1199 | continue; | |
1200 | } | |
1201 | ||
1202 | if (copy > length) | |
1203 | copy = length; | |
1204 | ||
1205 | if (!(rt->u.dst.dev->features&NETIF_F_SG)) { | |
1206 | unsigned int off; | |
1207 | ||
1208 | off = skb->len; | |
1209 | if (getfrag(from, skb_put(skb, copy), | |
1210 | offset, copy, off, skb) < 0) { | |
1211 | __skb_trim(skb, off); | |
1212 | err = -EFAULT; | |
1213 | goto error; | |
1214 | } | |
1215 | } else { | |
1216 | int i = skb_shinfo(skb)->nr_frags; | |
1217 | skb_frag_t *frag = &skb_shinfo(skb)->frags[i-1]; | |
1218 | struct page *page = sk->sk_sndmsg_page; | |
1219 | int off = sk->sk_sndmsg_off; | |
1220 | unsigned int left; | |
1221 | ||
1222 | if (page && (left = PAGE_SIZE - off) > 0) { | |
1223 | if (copy >= left) | |
1224 | copy = left; | |
1225 | if (page != frag->page) { | |
1226 | if (i == MAX_SKB_FRAGS) { | |
1227 | err = -EMSGSIZE; | |
1228 | goto error; | |
1229 | } | |
1230 | get_page(page); | |
1231 | skb_fill_page_desc(skb, i, page, sk->sk_sndmsg_off, 0); | |
1232 | frag = &skb_shinfo(skb)->frags[i]; | |
1233 | } | |
1234 | } else if(i < MAX_SKB_FRAGS) { | |
1235 | if (copy > PAGE_SIZE) | |
1236 | copy = PAGE_SIZE; | |
1237 | page = alloc_pages(sk->sk_allocation, 0); | |
1238 | if (page == NULL) { | |
1239 | err = -ENOMEM; | |
1240 | goto error; | |
1241 | } | |
1242 | sk->sk_sndmsg_page = page; | |
1243 | sk->sk_sndmsg_off = 0; | |
1244 | ||
1245 | skb_fill_page_desc(skb, i, page, 0, 0); | |
1246 | frag = &skb_shinfo(skb)->frags[i]; | |
1247 | skb->truesize += PAGE_SIZE; | |
1248 | atomic_add(PAGE_SIZE, &sk->sk_wmem_alloc); | |
1249 | } else { | |
1250 | err = -EMSGSIZE; | |
1251 | goto error; | |
1252 | } | |
1253 | if (getfrag(from, page_address(frag->page)+frag->page_offset+frag->size, offset, copy, skb->len, skb) < 0) { | |
1254 | err = -EFAULT; | |
1255 | goto error; | |
1256 | } | |
1257 | sk->sk_sndmsg_off += copy; | |
1258 | frag->size += copy; | |
1259 | skb->len += copy; | |
1260 | skb->data_len += copy; | |
1261 | } | |
1262 | offset += copy; | |
1263 | length -= copy; | |
1264 | } | |
1265 | return 0; | |
1266 | error: | |
1267 | inet->cork.length -= length; | |
a11d206d | 1268 | IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); |
1da177e4 LT |
1269 | return err; |
1270 | } | |
1271 | ||
1272 | int ip6_push_pending_frames(struct sock *sk) | |
1273 | { | |
1274 | struct sk_buff *skb, *tmp_skb; | |
1275 | struct sk_buff **tail_skb; | |
1276 | struct in6_addr final_dst_buf, *final_dst = &final_dst_buf; | |
1277 | struct inet_sock *inet = inet_sk(sk); | |
1278 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1279 | struct ipv6hdr *hdr; | |
1280 | struct ipv6_txoptions *opt = np->cork.opt; | |
1281 | struct rt6_info *rt = np->cork.rt; | |
1282 | struct flowi *fl = &inet->cork.fl; | |
1283 | unsigned char proto = fl->proto; | |
1284 | int err = 0; | |
1285 | ||
1286 | if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL) | |
1287 | goto out; | |
1288 | tail_skb = &(skb_shinfo(skb)->frag_list); | |
1289 | ||
1290 | /* move skb->data to ip header from ext header */ | |
1291 | if (skb->data < skb->nh.raw) | |
1292 | __skb_pull(skb, skb->nh.raw - skb->data); | |
1293 | while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) { | |
1294 | __skb_pull(tmp_skb, skb->h.raw - skb->nh.raw); | |
1295 | *tail_skb = tmp_skb; | |
1296 | tail_skb = &(tmp_skb->next); | |
1297 | skb->len += tmp_skb->len; | |
1298 | skb->data_len += tmp_skb->len; | |
1da177e4 LT |
1299 | skb->truesize += tmp_skb->truesize; |
1300 | __sock_put(tmp_skb->sk); | |
1301 | tmp_skb->destructor = NULL; | |
1302 | tmp_skb->sk = NULL; | |
1da177e4 LT |
1303 | } |
1304 | ||
1305 | ipv6_addr_copy(final_dst, &fl->fl6_dst); | |
1306 | __skb_pull(skb, skb->h.raw - skb->nh.raw); | |
1307 | if (opt && opt->opt_flen) | |
1308 | ipv6_push_frag_opts(skb, opt, &proto); | |
1309 | if (opt && opt->opt_nflen) | |
1310 | ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst); | |
1311 | ||
1312 | skb->nh.ipv6h = hdr = (struct ipv6hdr*) skb_push(skb, sizeof(struct ipv6hdr)); | |
1313 | ||
90bcaf7b | 1314 | *(__be32*)hdr = fl->fl6_flowlabel | |
41a1f8ea | 1315 | htonl(0x60000000 | ((int)np->cork.tclass << 20)); |
1da177e4 LT |
1316 | |
1317 | if (skb->len <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) | |
1318 | hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr)); | |
1319 | else | |
1320 | hdr->payload_len = 0; | |
1321 | hdr->hop_limit = np->cork.hop_limit; | |
1322 | hdr->nexthdr = proto; | |
1323 | ipv6_addr_copy(&hdr->saddr, &fl->fl6_src); | |
1324 | ipv6_addr_copy(&hdr->daddr, final_dst); | |
1325 | ||
a2c2064f PM |
1326 | skb->priority = sk->sk_priority; |
1327 | ||
1da177e4 | 1328 | skb->dst = dst_clone(&rt->u.dst); |
a11d206d | 1329 | IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS); |
1da177e4 LT |
1330 | err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, skb->dst->dev, dst_output); |
1331 | if (err) { | |
1332 | if (err > 0) | |
3320da89 | 1333 | err = np->recverr ? net_xmit_errno(err) : 0; |
1da177e4 LT |
1334 | if (err) |
1335 | goto error; | |
1336 | } | |
1337 | ||
1338 | out: | |
1339 | inet->cork.flags &= ~IPCORK_OPT; | |
a51482bd JJ |
1340 | kfree(np->cork.opt); |
1341 | np->cork.opt = NULL; | |
1da177e4 LT |
1342 | if (np->cork.rt) { |
1343 | dst_release(&np->cork.rt->u.dst); | |
1344 | np->cork.rt = NULL; | |
1345 | inet->cork.flags &= ~IPCORK_ALLFRAG; | |
1346 | } | |
1347 | memset(&inet->cork.fl, 0, sizeof(inet->cork.fl)); | |
1348 | return err; | |
1349 | error: | |
1350 | goto out; | |
1351 | } | |
1352 | ||
1353 | void ip6_flush_pending_frames(struct sock *sk) | |
1354 | { | |
1355 | struct inet_sock *inet = inet_sk(sk); | |
1356 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1357 | struct sk_buff *skb; | |
1358 | ||
1359 | while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) { | |
a11d206d YH |
1360 | IP6_INC_STATS(ip6_dst_idev(skb->dst), |
1361 | IPSTATS_MIB_OUTDISCARDS); | |
1da177e4 LT |
1362 | kfree_skb(skb); |
1363 | } | |
1364 | ||
1365 | inet->cork.flags &= ~IPCORK_OPT; | |
1366 | ||
a51482bd JJ |
1367 | kfree(np->cork.opt); |
1368 | np->cork.opt = NULL; | |
1da177e4 LT |
1369 | if (np->cork.rt) { |
1370 | dst_release(&np->cork.rt->u.dst); | |
1371 | np->cork.rt = NULL; | |
1372 | inet->cork.flags &= ~IPCORK_ALLFRAG; | |
1373 | } | |
1374 | memset(&inet->cork.fl, 0, sizeof(inet->cork.fl)); | |
1375 | } |