1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
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>
41#include <linux/module.h>
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
59static int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *));
60
61static __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
73static int ip6_output_finish(struct sk_buff *skb)
74{
75 struct dst_entry *dst = skb->dst;
76
77 if (dst->hh)
78 return neigh_hh_output(dst->hh, skb);
79 else if (dst->neighbour)
80 return dst->neighbour->output(skb);
81
82 IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
83 kfree_skb(skb);
84 return -EINVAL;
85
86}
87
88
89static int ip6_dev_loopback_xmit(struct sk_buff *newskb)
90{
91 skb_reset_mac_header(newskb);
92 __skb_pull(newskb, skb_network_offset(newskb));
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
102static 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(&ipv6_hdr(skb)->daddr)) {
111 struct ipv6_pinfo* np = skb->sk ? inet6_sk(skb->sk) : NULL;
112 struct inet6_dev *idev = ip6_dst_idev(skb->dst);
113
114 if (!(dev->flags & IFF_LOOPBACK) && (!np || np->mc_loop) &&
115 ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
116 &ipv6_hdr(skb)->saddr)) {
117 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
118
119
120
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 (ipv6_hdr(skb)->hop_limit == 0) {
128 IP6_INC_STATS(idev, IPSTATS_MIB_OUTDISCARDS);
129 kfree_skb(skb);
130 return 0;
131 }
132 }
133
134 IP6_INC_STATS(idev, IPSTATS_MIB_OUTMCASTPKTS);
135 }
136
137 return NF_HOOK(PF_INET6, NF_IP6_POST_ROUTING, skb,NULL, skb->dev,ip6_output_finish);
138}
139
140static inline int ip6_skb_dst_mtu(struct sk_buff *skb)
141{
142 struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
143
144 return (np && np->pmtudisc == IPV6_PMTUDISC_PROBE) ?
145 skb->dst->dev->mtu : dst_mtu(skb->dst);
146}
147
148int ip6_output(struct sk_buff *skb)
149{
150 if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
151 dst_allfrag(skb->dst))
152 return ip6_fragment(skb, ip6_output2);
153 else
154 return ip6_output2(skb);
155}
156
157
158
159
160
161int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
162 struct ipv6_txoptions *opt, int ipfragok)
163{
164 struct ipv6_pinfo *np = inet6_sk(sk);
165 struct in6_addr *first_hop = &fl->fl6_dst;
166 struct dst_entry *dst = skb->dst;
167 struct ipv6hdr *hdr;
168 u8 proto = fl->proto;
169 int seg_len = skb->len;
170 int hlimit, tclass;
171 u32 mtu;
172
173 if (opt) {
174 unsigned int head_room;
175
176
177
178
179 head_room = opt->opt_nflen + opt->opt_flen;
180 seg_len += head_room;
181 head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
182
183 if (skb_headroom(skb) < head_room) {
184 struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
185 if (skb2 == NULL) {
186 IP6_INC_STATS(ip6_dst_idev(skb->dst),
187 IPSTATS_MIB_OUTDISCARDS);
188 kfree_skb(skb);
189 return -ENOBUFS;
190 }
191 kfree_skb(skb);
192 skb = skb2;
193 if (sk)
194 skb_set_owner_w(skb, sk);
195 }
196 if (opt->opt_flen)
197 ipv6_push_frag_opts(skb, opt, &proto);
198 if (opt->opt_nflen)
199 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
200 }
201
202 skb_push(skb, sizeof(struct ipv6hdr));
203 skb_reset_network_header(skb);
204 hdr = ipv6_hdr(skb);
205
206
207
208
209
210 hlimit = -1;
211 if (np)
212 hlimit = np->hop_limit;
213 if (hlimit < 0)
214 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
215 if (hlimit < 0)
216 hlimit = ipv6_get_hoplimit(dst->dev);
217
218 tclass = -1;
219 if (np)
220 tclass = np->tclass;
221 if (tclass < 0)
222 tclass = 0;
223
224 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | fl->fl6_flowlabel;
225
226 hdr->payload_len = htons(seg_len);
227 hdr->nexthdr = proto;
228 hdr->hop_limit = hlimit;
229
230 ipv6_addr_copy(&hdr->saddr, &fl->fl6_src);
231 ipv6_addr_copy(&hdr->daddr, first_hop);
232
233 skb->priority = sk->sk_priority;
234
235 mtu = dst_mtu(dst);
236 if ((skb->len <= mtu) || ipfragok || skb_is_gso(skb)) {
237 IP6_INC_STATS(ip6_dst_idev(skb->dst),
238 IPSTATS_MIB_OUTREQUESTS);
239 return NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev,
240 dst_output);
241 }
242
243 if (net_ratelimit())
244 printk(KERN_DEBUG "IPv6: sending pkt_too_big to self\n");
245 skb->dev = dst->dev;
246 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, skb->dev);
247 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGFAILS);
248 kfree_skb(skb);
249 return -EMSGSIZE;
250}
251
252EXPORT_SYMBOL(ip6_xmit);
253
254
255
256
257
258
259
260
261int ip6_nd_hdr(struct sock *sk, struct sk_buff *skb, struct net_device *dev,
262 struct in6_addr *saddr, struct in6_addr *daddr,
263 int proto, int len)
264{
265 struct ipv6_pinfo *np = inet6_sk(sk);
266 struct ipv6hdr *hdr;
267 int totlen;
268
269 skb->protocol = htons(ETH_P_IPV6);
270 skb->dev = dev;
271
272 totlen = len + sizeof(struct ipv6hdr);
273
274 skb_reset_network_header(skb);
275 skb_put(skb, sizeof(struct ipv6hdr));
276 hdr = ipv6_hdr(skb);
277
278 *(__be32*)hdr = htonl(0x60000000);
279
280 hdr->payload_len = htons(len);
281 hdr->nexthdr = proto;
282 hdr->hop_limit = np->hop_limit;
283
284 ipv6_addr_copy(&hdr->saddr, saddr);
285 ipv6_addr_copy(&hdr->daddr, daddr);
286
287 return 0;
288}
289
290static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
291{
292 struct ip6_ra_chain *ra;
293 struct sock *last = NULL;
294
295 read_lock(&ip6_ra_lock);
296 for (ra = ip6_ra_chain; ra; ra = ra->next) {
297 struct sock *sk = ra->sk;
298 if (sk && ra->sel == sel &&
299 (!sk->sk_bound_dev_if ||
300 sk->sk_bound_dev_if == skb->dev->ifindex)) {
301 if (last) {
302 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
303 if (skb2)
304 rawv6_rcv(last, skb2);
305 }
306 last = sk;
307 }
308 }
309
310 if (last) {
311 rawv6_rcv(last, skb);
312 read_unlock(&ip6_ra_lock);
313 return 1;
314 }
315 read_unlock(&ip6_ra_lock);
316 return 0;
317}
318
319static int ip6_forward_proxy_check(struct sk_buff *skb)
320{
321 struct ipv6hdr *hdr = ipv6_hdr(skb);
322 u8 nexthdr = hdr->nexthdr;
323 int offset;
324
325 if (ipv6_ext_hdr(nexthdr)) {
326 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr);
327 if (offset < 0)
328 return 0;
329 } else
330 offset = sizeof(struct ipv6hdr);
331
332 if (nexthdr == IPPROTO_ICMPV6) {
333 struct icmp6hdr *icmp6;
334
335 if (!pskb_may_pull(skb, (skb_network_header(skb) +
336 offset + 1 - skb->data)))
337 return 0;
338
339 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
340
341 switch (icmp6->icmp6_type) {
342 case NDISC_ROUTER_SOLICITATION:
343 case NDISC_ROUTER_ADVERTISEMENT:
344 case NDISC_NEIGHBOUR_SOLICITATION:
345 case NDISC_NEIGHBOUR_ADVERTISEMENT:
346 case NDISC_REDIRECT:
347
348
349
350
351 return 1;
352 default:
353 break;
354 }
355 }
356
357
358
359
360
361
362 if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
363 dst_link_failure(skb);
364 return -1;
365 }
366
367 return 0;
368}
369
370static inline int ip6_forward_finish(struct sk_buff *skb)
371{
372 return dst_output(skb);
373}
374
375int ip6_forward(struct sk_buff *skb)
376{
377 struct dst_entry *dst = skb->dst;
378 struct ipv6hdr *hdr = ipv6_hdr(skb);
379 struct inet6_skb_parm *opt = IP6CB(skb);
380
381 if (ipv6_devconf.forwarding == 0)
382 goto error;
383
384 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
385 IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
386 goto drop;
387 }
388
389 skb_forward_csum(skb);
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404 if (opt->ra) {
405 u8 *ptr = skb_network_header(skb) + opt->ra;
406 if (ip6_call_ra_chain(skb, (ptr[2]<<8) + ptr[3]))
407 return 0;
408 }
409
410
411
412
413 if (hdr->hop_limit <= 1) {
414
415 skb->dev = dst->dev;
416 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
417 0, skb->dev);
418 IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS);
419
420 kfree_skb(skb);
421 return -ETIMEDOUT;
422 }
423
424
425 if (ipv6_devconf.proxy_ndp &&
426 pneigh_lookup(&nd_tbl, &hdr->daddr, skb->dev, 0)) {
427 int proxied = ip6_forward_proxy_check(skb);
428 if (proxied > 0)
429 return ip6_input(skb);
430 else if (proxied < 0) {
431 IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
432 goto drop;
433 }
434 }
435
436 if (!xfrm6_route_forward(skb)) {
437 IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
438 goto drop;
439 }
440 dst = skb->dst;
441
442
443
444
445
446 if (skb->dev == dst->dev && dst->neighbour && opt->srcrt == 0 &&
447 !skb->sp) {
448 struct in6_addr *target = NULL;
449 struct rt6_info *rt;
450 struct neighbour *n = dst->neighbour;
451
452
453
454
455
456
457 rt = (struct rt6_info *) dst;
458 if ((rt->rt6i_flags & RTF_GATEWAY))
459 target = (struct in6_addr*)&n->primary_key;
460 else
461 target = &hdr->daddr;
462
463
464
465
466 if (xrlim_allow(dst, 1*HZ))
467 ndisc_send_redirect(skb, n, target);
468 } else {
469 int addrtype = ipv6_addr_type(&hdr->saddr);
470
471
472 if (addrtype & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LOOPBACK))
473 goto error;
474 if (addrtype & IPV6_ADDR_LINKLOCAL) {
475 icmpv6_send(skb, ICMPV6_DEST_UNREACH,
476 ICMPV6_NOT_NEIGHBOUR, 0, skb->dev);
477 goto error;
478 }
479 }
480
481 if (skb->len > dst_mtu(dst)) {
482
483 skb->dev = dst->dev;
484 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, dst_mtu(dst), skb->dev);
485 IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INTOOBIGERRORS);
486 IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_FRAGFAILS);
487 kfree_skb(skb);
488 return -EMSGSIZE;
489 }
490
491 if (skb_cow(skb, dst->dev->hard_header_len)) {
492 IP6_INC_STATS(ip6_dst_idev(dst), IPSTATS_MIB_OUTDISCARDS);
493 goto drop;
494 }
495
496 hdr = ipv6_hdr(skb);
497
498
499
500 hdr->hop_limit--;
501
502 IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
503 return NF_HOOK(PF_INET6,NF_IP6_FORWARD, skb, skb->dev, dst->dev, ip6_forward_finish);
504
505error:
506 IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
507drop:
508 kfree_skb(skb);
509 return -EINVAL;
510}
511
512static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
513{
514 to->pkt_type = from->pkt_type;
515 to->priority = from->priority;
516 to->protocol = from->protocol;
517 dst_release(to->dst);
518 to->dst = dst_clone(from->dst);
519 to->dev = from->dev;
520 to->mark = from->mark;
521
522#ifdef CONFIG_NET_SCHED
523 to->tc_index = from->tc_index;
524#endif
525 nf_copy(to, from);
526#if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
527 defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
528 to->nf_trace = from->nf_trace;
529#endif
530 skb_copy_secmark(to, from);
531}
532
533int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr)
534{
535 u16 offset = sizeof(struct ipv6hdr);
536 struct ipv6_opt_hdr *exthdr =
537 (struct ipv6_opt_hdr *)(ipv6_hdr(skb) + 1);
538 unsigned int packet_len = skb->tail - skb->network_header;
539 int found_rhdr = 0;
540 *nexthdr = &ipv6_hdr(skb)->nexthdr;
541
542 while (offset + 1 <= packet_len) {
543
544 switch (**nexthdr) {
545
546 case NEXTHDR_HOP:
547 break;
548 case NEXTHDR_ROUTING:
549 found_rhdr = 1;
550 break;
551 case NEXTHDR_DEST:
552#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
553 if (ipv6_find_tlv(skb, offset, IPV6_TLV_HAO) >= 0)
554 break;
555#endif
556 if (found_rhdr)
557 return offset;
558 break;
559 default :
560 return offset;
561 }
562
563 offset += ipv6_optlen(exthdr);
564 *nexthdr = &exthdr->nexthdr;
565 exthdr = (struct ipv6_opt_hdr *)(skb_network_header(skb) +
566 offset);
567 }
568
569 return offset;
570}
571EXPORT_SYMBOL_GPL(ip6_find_1stfragopt);
572
573static int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
574{
575 struct net_device *dev;
576 struct sk_buff *frag;
577 struct rt6_info *rt = (struct rt6_info*)skb->dst;
578 struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
579 struct ipv6hdr *tmp_hdr;
580 struct frag_hdr *fh;
581 unsigned int mtu, hlen, left, len;
582 __be32 frag_id = 0;
583 int ptr, offset = 0, err=0;
584 u8 *prevhdr, nexthdr = 0;
585
586 dev = rt->u.dst.dev;
587 hlen = ip6_find_1stfragopt(skb, &prevhdr);
588 nexthdr = *prevhdr;
589
590 mtu = ip6_skb_dst_mtu(skb);
591
592
593
594
595
596 if (!np || np->pmtudisc >= IPV6_PMTUDISC_DO) {
597 skb->dev = skb->dst->dev;
598 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, skb->dev);
599 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGFAILS);
600 kfree_skb(skb);
601 return -EMSGSIZE;
602 }
603
604 if (np && np->frag_size < mtu) {
605 if (np->frag_size)
606 mtu = np->frag_size;
607 }
608 mtu -= hlen + sizeof(struct frag_hdr);
609
610 if (skb_shinfo(skb)->frag_list) {
611 int first_len = skb_pagelen(skb);
612
613 if (first_len - hlen > mtu ||
614 ((first_len - hlen) & 7) ||
615 skb_cloned(skb))
616 goto slow_path;
617
618 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next) {
619
620 if (frag->len > mtu ||
621 ((frag->len & 7) && frag->next) ||
622 skb_headroom(frag) < hlen)
623 goto slow_path;
624
625
626 if (skb_shared(frag))
627 goto slow_path;
628
629 BUG_ON(frag->sk);
630 if (skb->sk) {
631 sock_hold(skb->sk);
632 frag->sk = skb->sk;
633 frag->destructor = sock_wfree;
634 skb->truesize -= frag->truesize;
635 }
636 }
637
638 err = 0;
639 offset = 0;
640 frag = skb_shinfo(skb)->frag_list;
641 skb_shinfo(skb)->frag_list = NULL;
642
643
644 *prevhdr = NEXTHDR_FRAGMENT;
645 tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
646 if (!tmp_hdr) {
647 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGFAILS);
648 return -ENOMEM;
649 }
650
651 __skb_pull(skb, hlen);
652 fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr));
653 __skb_push(skb, hlen);
654 skb_reset_network_header(skb);
655 memcpy(skb_network_header(skb), tmp_hdr, hlen);
656
657 ipv6_select_ident(skb, fh);
658 fh->nexthdr = nexthdr;
659 fh->reserved = 0;
660 fh->frag_off = htons(IP6_MF);
661 frag_id = fh->identification;
662
663 first_len = skb_pagelen(skb);
664 skb->data_len = first_len - skb_headlen(skb);
665 skb->len = first_len;
666 ipv6_hdr(skb)->payload_len = htons(first_len -
667 sizeof(struct ipv6hdr));
668
669 dst_hold(&rt->u.dst);
670
671 for (;;) {
672
673
674 if (frag) {
675 frag->ip_summed = CHECKSUM_NONE;
676 skb_reset_transport_header(frag);
677 fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr));
678 __skb_push(frag, hlen);
679 skb_reset_network_header(frag);
680 memcpy(skb_network_header(frag), tmp_hdr,
681 hlen);
682 offset += skb->len - hlen - sizeof(struct frag_hdr);
683 fh->nexthdr = nexthdr;
684 fh->reserved = 0;
685 fh->frag_off = htons(offset);
686 if (frag->next != NULL)
687 fh->frag_off |= htons(IP6_MF);
688 fh->identification = frag_id;
689 ipv6_hdr(frag)->payload_len =
690 htons(frag->len -
691 sizeof(struct ipv6hdr));
692 ip6_copy_metadata(frag, skb);
693 }
694
695 err = output(skb);
696 if(!err)
697 IP6_INC_STATS(ip6_dst_idev(&rt->u.dst), IPSTATS_MIB_FRAGCREATES);
698
699 if (err || !frag)
700 break;
701
702 skb = frag;
703 frag = skb->next;
704 skb->next = NULL;
705 }
706
707 kfree(tmp_hdr);
708
709 if (err == 0) {
710 IP6_INC_STATS(ip6_dst_idev(&rt->u.dst), IPSTATS_MIB_FRAGOKS);
711 dst_release(&rt->u.dst);
712 return 0;
713 }
714
715 while (frag) {
716 skb = frag->next;
717 kfree_skb(frag);
718 frag = skb;
719 }
720
721 IP6_INC_STATS(ip6_dst_idev(&rt->u.dst), IPSTATS_MIB_FRAGFAILS);
722 dst_release(&rt->u.dst);
723 return err;
724 }
725
726slow_path:
727 left = skb->len - hlen;
728 ptr = hlen;
729
730
731
732
733
734 *prevhdr = NEXTHDR_FRAGMENT;
735
736
737
738
739 while(left > 0) {
740 len = left;
741
742 if (len > mtu)
743 len = mtu;
744
745
746 if (len < left) {
747 len &= ~7;
748 }
749
750
751
752
753 if ((frag = alloc_skb(len+hlen+sizeof(struct frag_hdr)+LL_RESERVED_SPACE(rt->u.dst.dev), GFP_ATOMIC)) == NULL) {
754 NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n");
755 IP6_INC_STATS(ip6_dst_idev(skb->dst),
756 IPSTATS_MIB_FRAGFAILS);
757 err = -ENOMEM;
758 goto fail;
759 }
760
761
762
763
764
765 ip6_copy_metadata(frag, skb);
766 skb_reserve(frag, LL_RESERVED_SPACE(rt->u.dst.dev));
767 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
768 skb_reset_network_header(frag);
769 fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
770 frag->transport_header = (frag->network_header + hlen +
771 sizeof(struct frag_hdr));
772
773
774
775
776
777 if (skb->sk)
778 skb_set_owner_w(frag, skb->sk);
779
780
781
782
783 skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
784
785
786
787
788 fh->nexthdr = nexthdr;
789 fh->reserved = 0;
790 if (!frag_id) {
791 ipv6_select_ident(skb, fh);
792 frag_id = fh->identification;
793 } else
794 fh->identification = frag_id;
795
796
797
798
799 if (skb_copy_bits(skb, ptr, skb_transport_header(frag), len))
800 BUG();
801 left -= len;
802
803 fh->frag_off = htons(offset);
804 if (left > 0)
805 fh->frag_off |= htons(IP6_MF);
806 ipv6_hdr(frag)->payload_len = htons(frag->len -
807 sizeof(struct ipv6hdr));
808
809 ptr += len;
810 offset += len;
811
812
813
814
815 err = output(frag);
816 if (err)
817 goto fail;
818
819 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGCREATES);
820 }
821 IP6_INC_STATS(ip6_dst_idev(skb->dst),
822 IPSTATS_MIB_FRAGOKS);
823 kfree_skb(skb);
824 return err;
825
826fail:
827 IP6_INC_STATS(ip6_dst_idev(skb->dst),
828 IPSTATS_MIB_FRAGFAILS);
829 kfree_skb(skb);
830 return err;
831}
832
833static inline int ip6_rt_check(struct rt6key *rt_key,
834 struct in6_addr *fl_addr,
835 struct in6_addr *addr_cache)
836{
837 return ((rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
838 (addr_cache == NULL || !ipv6_addr_equal(fl_addr, addr_cache)));
839}
840
841static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
842 struct dst_entry *dst,
843 struct flowi *fl)
844{
845 struct ipv6_pinfo *np = inet6_sk(sk);
846 struct rt6_info *rt = (struct rt6_info *)dst;
847
848 if (!dst)
849 goto out;
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868 if (ip6_rt_check(&rt->rt6i_dst, &fl->fl6_dst, np->daddr_cache) ||
869#ifdef CONFIG_IPV6_SUBTREES
870 ip6_rt_check(&rt->rt6i_src, &fl->fl6_src, np->saddr_cache) ||
871#endif
872 (fl->oif && fl->oif != dst->dev->ifindex)) {
873 dst_release(dst);
874 dst = NULL;
875 }
876
877out:
878 return dst;
879}
880
881static int ip6_dst_lookup_tail(struct sock *sk,
882 struct dst_entry **dst, struct flowi *fl)
883{
884 int err;
885
886 if (*dst == NULL)
887 *dst = ip6_route_output(sk, fl);
888
889 if ((err = (*dst)->error))
890 goto out_err_release;
891
892 if (ipv6_addr_any(&fl->fl6_src)) {
893 err = ipv6_get_saddr(*dst, &fl->fl6_dst, &fl->fl6_src);
894 if (err)
895 goto out_err_release;
896 }
897
898#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
899
900
901
902
903
904
905
906
907 if (!((*dst)->neighbour->nud_state & NUD_VALID)) {
908 struct inet6_ifaddr *ifp;
909 struct flowi fl_gw;
910 int redirect;
911
912 ifp = ipv6_get_ifaddr(&fl->fl6_src, (*dst)->dev, 1);
913
914 redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
915 if (ifp)
916 in6_ifa_put(ifp);
917
918 if (redirect) {
919
920
921
922
923 dst_release(*dst);
924 memcpy(&fl_gw, fl, sizeof(struct flowi));
925 memset(&fl_gw.fl6_dst, 0, sizeof(struct in6_addr));
926 *dst = ip6_route_output(sk, &fl_gw);
927 if ((err = (*dst)->error))
928 goto out_err_release;
929 }
930 }
931#endif
932
933 return 0;
934
935out_err_release:
936 if (err == -ENETUNREACH)
937 IP6_INC_STATS_BH(NULL, IPSTATS_MIB_OUTNOROUTES);
938 dst_release(*dst);
939 *dst = NULL;
940 return err;
941}
942
943
944
945
946
947
948
949
950
951
952
953int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl)
954{
955 *dst = NULL;
956 return ip6_dst_lookup_tail(sk, dst, fl);
957}
958EXPORT_SYMBOL_GPL(ip6_dst_lookup);
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973int ip6_sk_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl)
974{
975 *dst = NULL;
976 if (sk) {
977 *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
978 *dst = ip6_sk_dst_check(sk, *dst, fl);
979 }
980
981 return ip6_dst_lookup_tail(sk, dst, fl);
982}
983EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup);
984
985static inline int ip6_ufo_append_data(struct sock *sk,
986 int getfrag(void *from, char *to, int offset, int len,
987 int odd, struct sk_buff *skb),
988 void *from, int length, int hh_len, int fragheaderlen,
989 int transhdrlen, int mtu,unsigned int flags)
990
991{
992 struct sk_buff *skb;
993 int err;
994
995
996
997
998
999 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) {
1000 skb = sock_alloc_send_skb(sk,
1001 hh_len + fragheaderlen + transhdrlen + 20,
1002 (flags & MSG_DONTWAIT), &err);
1003 if (skb == NULL)
1004 return -ENOMEM;
1005
1006
1007 skb_reserve(skb, hh_len);
1008
1009
1010 skb_put(skb,fragheaderlen + transhdrlen);
1011
1012
1013 skb_reset_network_header(skb);
1014
1015
1016 skb->transport_header = skb->network_header + fragheaderlen;
1017
1018 skb->ip_summed = CHECKSUM_PARTIAL;
1019 skb->csum = 0;
1020 sk->sk_sndmsg_off = 0;
1021 }
1022
1023 err = skb_append_datato_frags(sk,skb, getfrag, from,
1024 (length - transhdrlen));
1025 if (!err) {
1026 struct frag_hdr fhdr;
1027
1028
1029 skb_shinfo(skb)->gso_size = mtu - fragheaderlen -
1030 sizeof(struct frag_hdr);
1031 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
1032 ipv6_select_ident(skb, &fhdr);
1033 skb_shinfo(skb)->ip6_frag_id = fhdr.identification;
1034 __skb_queue_tail(&sk->sk_write_queue, skb);
1035
1036 return 0;
1037 }
1038
1039
1040
1041 kfree_skb(skb);
1042
1043 return err;
1044}
1045
1046int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
1047 int offset, int len, int odd, struct sk_buff *skb),
1048 void *from, int length, int transhdrlen,
1049 int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi *fl,
1050 struct rt6_info *rt, unsigned int flags)
1051{
1052 struct inet_sock *inet = inet_sk(sk);
1053 struct ipv6_pinfo *np = inet6_sk(sk);
1054 struct sk_buff *skb;
1055 unsigned int maxfraglen, fragheaderlen;
1056 int exthdrlen;
1057 int hh_len;
1058 int mtu;
1059 int copy;
1060 int err;
1061 int offset = 0;
1062 int csummode = CHECKSUM_NONE;
1063
1064 if (flags&MSG_PROBE)
1065 return 0;
1066 if (skb_queue_empty(&sk->sk_write_queue)) {
1067
1068
1069
1070 if (opt) {
1071 if (np->cork.opt == NULL) {
1072 np->cork.opt = kmalloc(opt->tot_len,
1073 sk->sk_allocation);
1074 if (unlikely(np->cork.opt == NULL))
1075 return -ENOBUFS;
1076 } else if (np->cork.opt->tot_len < opt->tot_len) {
1077 printk(KERN_DEBUG "ip6_append_data: invalid option length\n");
1078 return -EINVAL;
1079 }
1080 memcpy(np->cork.opt, opt, opt->tot_len);
1081 inet->cork.flags |= IPCORK_OPT;
1082
1083 }
1084 dst_hold(&rt->u.dst);
1085 np->cork.rt = rt;
1086 inet->cork.fl = *fl;
1087 np->cork.hop_limit = hlimit;
1088 np->cork.tclass = tclass;
1089 mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ?
1090 rt->u.dst.dev->mtu : dst_mtu(rt->u.dst.path);
1091 if (np->frag_size < mtu) {
1092 if (np->frag_size)
1093 mtu = np->frag_size;
1094 }
1095 inet->cork.fragsize = mtu;
1096 if (dst_allfrag(rt->u.dst.path))
1097 inet->cork.flags |= IPCORK_ALLFRAG;
1098 inet->cork.length = 0;
1099 sk->sk_sndmsg_page = NULL;
1100 sk->sk_sndmsg_off = 0;
1101 exthdrlen = rt->u.dst.header_len + (opt ? opt->opt_flen : 0);
1102 length += exthdrlen;
1103 transhdrlen += exthdrlen;
1104 } else {
1105 rt = np->cork.rt;
1106 fl = &inet->cork.fl;
1107 if (inet->cork.flags & IPCORK_OPT)
1108 opt = np->cork.opt;
1109 transhdrlen = 0;
1110 exthdrlen = 0;
1111 mtu = inet->cork.fragsize;
1112 }
1113
1114 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
1115
1116 fragheaderlen = sizeof(struct ipv6hdr) + rt->u.dst.nfheader_len + (opt ? opt->opt_nflen : 0);
1117 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr);
1118
1119 if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
1120 if (inet->cork.length + length > sizeof(struct ipv6hdr) + IPV6_MAXPLEN - fragheaderlen) {
1121 ipv6_local_error(sk, EMSGSIZE, fl, mtu-exthdrlen);
1122 return -EMSGSIZE;
1123 }
1124 }
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142 inet->cork.length += length;
1143 if (((length > mtu) && (sk->sk_protocol == IPPROTO_UDP)) &&
1144 (rt->u.dst.dev->features & NETIF_F_UFO)) {
1145
1146 err = ip6_ufo_append_data(sk, getfrag, from, length, hh_len,
1147 fragheaderlen, transhdrlen, mtu,
1148 flags);
1149 if (err)
1150 goto error;
1151 return 0;
1152 }
1153
1154 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL)
1155 goto alloc_new_skb;
1156
1157 while (length > 0) {
1158
1159 copy = (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
1160 if (copy < length)
1161 copy = maxfraglen - skb->len;
1162
1163 if (copy <= 0) {
1164 char *data;
1165 unsigned int datalen;
1166 unsigned int fraglen;
1167 unsigned int fraggap;
1168 unsigned int alloclen;
1169 struct sk_buff *skb_prev;
1170alloc_new_skb:
1171 skb_prev = skb;
1172
1173
1174 if (skb_prev)
1175 fraggap = skb_prev->len - maxfraglen;
1176 else
1177 fraggap = 0;
1178
1179
1180
1181
1182
1183 datalen = length + fraggap;
1184 if (datalen > (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1185 datalen = maxfraglen - fragheaderlen;
1186
1187 fraglen = datalen + fragheaderlen;
1188 if ((flags & MSG_MORE) &&
1189 !(rt->u.dst.dev->features&NETIF_F_SG))
1190 alloclen = mtu;
1191 else
1192 alloclen = datalen + fragheaderlen;
1193
1194
1195
1196
1197
1198
1199 if (datalen == length + fraggap)
1200 alloclen += rt->u.dst.trailer_len;
1201
1202
1203
1204
1205
1206
1207 alloclen += sizeof(struct frag_hdr);
1208
1209 if (transhdrlen) {
1210 skb = sock_alloc_send_skb(sk,
1211 alloclen + hh_len,
1212 (flags & MSG_DONTWAIT), &err);
1213 } else {
1214 skb = NULL;
1215 if (atomic_read(&sk->sk_wmem_alloc) <=
1216 2 * sk->sk_sndbuf)
1217 skb = sock_wmalloc(sk,
1218 alloclen + hh_len, 1,
1219 sk->sk_allocation);
1220 if (unlikely(skb == NULL))
1221 err = -ENOBUFS;
1222 }
1223 if (skb == NULL)
1224 goto error;
1225
1226
1227
1228 skb->ip_summed = csummode;
1229 skb->csum = 0;
1230
1231 skb_reserve(skb, hh_len+sizeof(struct frag_hdr));
1232
1233
1234
1235
1236 data = skb_put(skb, fraglen);
1237 skb_set_network_header(skb, exthdrlen);
1238 data += fragheaderlen;
1239 skb->transport_header = (skb->network_header +
1240 fragheaderlen);
1241 if (fraggap) {
1242 skb->csum = skb_copy_and_csum_bits(
1243 skb_prev, maxfraglen,
1244 data + transhdrlen, fraggap, 0);
1245 skb_prev->csum = csum_sub(skb_prev->csum,
1246 skb->csum);
1247 data += fraggap;
1248 pskb_trim_unique(skb_prev, maxfraglen);
1249 }
1250 copy = datalen - transhdrlen - fraggap;
1251 if (copy < 0) {
1252 err = -EINVAL;
1253 kfree_skb(skb);
1254 goto error;
1255 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1256 err = -EFAULT;
1257 kfree_skb(skb);
1258 goto error;
1259 }
1260
1261 offset += copy;
1262 length -= datalen - fraggap;
1263 transhdrlen = 0;
1264 exthdrlen = 0;
1265 csummode = CHECKSUM_NONE;
1266
1267
1268
1269
1270 __skb_queue_tail(&sk->sk_write_queue, skb);
1271 continue;
1272 }
1273
1274 if (copy > length)
1275 copy = length;
1276
1277 if (!(rt->u.dst.dev->features&NETIF_F_SG)) {
1278 unsigned int off;
1279
1280 off = skb->len;
1281 if (getfrag(from, skb_put(skb, copy),
1282 offset, copy, off, skb) < 0) {
1283 __skb_trim(skb, off);
1284 err = -EFAULT;
1285 goto error;
1286 }
1287 } else {
1288 int i = skb_shinfo(skb)->nr_frags;
1289 skb_frag_t *frag = &skb_shinfo(skb)->frags[i-1];
1290 struct page *page = sk->sk_sndmsg_page;
1291 int off = sk->sk_sndmsg_off;
1292 unsigned int left;
1293
1294 if (page && (left = PAGE_SIZE - off) > 0) {
1295 if (copy >= left)
1296 copy = left;
1297 if (page != frag->page) {
1298 if (i == MAX_SKB_FRAGS) {
1299 err = -EMSGSIZE;
1300 goto error;
1301 }
1302 get_page(page);
1303 skb_fill_page_desc(skb, i, page, sk->sk_sndmsg_off, 0);
1304 frag = &skb_shinfo(skb)->frags[i];
1305 }
1306 } else if(i < MAX_SKB_FRAGS) {
1307 if (copy > PAGE_SIZE)
1308 copy = PAGE_SIZE;
1309 page = alloc_pages(sk->sk_allocation, 0);
1310 if (page == NULL) {
1311 err = -ENOMEM;
1312 goto error;
1313 }
1314 sk->sk_sndmsg_page = page;
1315 sk->sk_sndmsg_off = 0;
1316
1317 skb_fill_page_desc(skb, i, page, 0, 0);
1318 frag = &skb_shinfo(skb)->frags[i];
1319 } else {
1320 err = -EMSGSIZE;
1321 goto error;
1322 }
1323 if (getfrag(from, page_address(frag->page)+frag->page_offset+frag->size, offset, copy, skb->len, skb) < 0) {
1324 err = -EFAULT;
1325 goto error;
1326 }
1327 sk->sk_sndmsg_off += copy;
1328 frag->size += copy;
1329 skb->len += copy;
1330 skb->data_len += copy;
1331 skb->truesize += copy;
1332 atomic_add(copy, &sk->sk_wmem_alloc);
1333 }
1334 offset += copy;
1335 length -= copy;
1336 }
1337 return 0;
1338error:
1339 inet->cork.length -= length;
1340 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
1341 return err;
1342}
1343
1344static void ip6_cork_release(struct inet_sock *inet, struct ipv6_pinfo *np)
1345{
1346 inet->cork.flags &= ~IPCORK_OPT;
1347 kfree(np->cork.opt);
1348 np->cork.opt = NULL;
1349 if (np->cork.rt) {
1350 dst_release(&np->cork.rt->u.dst);
1351 np->cork.rt = NULL;
1352 inet->cork.flags &= ~IPCORK_ALLFRAG;
1353 }
1354 memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
1355}
1356
1357int ip6_push_pending_frames(struct sock *sk)
1358{
1359 struct sk_buff *skb, *tmp_skb;
1360 struct sk_buff **tail_skb;
1361 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
1362 struct inet_sock *inet = inet_sk(sk);
1363 struct ipv6_pinfo *np = inet6_sk(sk);
1364 struct ipv6hdr *hdr;
1365 struct ipv6_txoptions *opt = np->cork.opt;
1366 struct rt6_info *rt = np->cork.rt;
1367 struct flowi *fl = &inet->cork.fl;
1368 unsigned char proto = fl->proto;
1369 int err = 0;
1370
1371 if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL)
1372 goto out;
1373 tail_skb = &(skb_shinfo(skb)->frag_list);
1374
1375
1376 if (skb->data < skb_network_header(skb))
1377 __skb_pull(skb, skb_network_offset(skb));
1378 while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) {
1379 __skb_pull(tmp_skb, skb_network_header_len(skb));
1380 *tail_skb = tmp_skb;
1381 tail_skb = &(tmp_skb->next);
1382 skb->len += tmp_skb->len;
1383 skb->data_len += tmp_skb->len;
1384 skb->truesize += tmp_skb->truesize;
1385 __sock_put(tmp_skb->sk);
1386 tmp_skb->destructor = NULL;
1387 tmp_skb->sk = NULL;
1388 }
1389
1390 ipv6_addr_copy(final_dst, &fl->fl6_dst);
1391 __skb_pull(skb, skb_network_header_len(skb));
1392 if (opt && opt->opt_flen)
1393 ipv6_push_frag_opts(skb, opt, &proto);
1394 if (opt && opt->opt_nflen)
1395 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1396
1397 skb_push(skb, sizeof(struct ipv6hdr));
1398 skb_reset_network_header(skb);
1399 hdr = ipv6_hdr(skb);
1400
1401 *(__be32*)hdr = fl->fl6_flowlabel |
1402 htonl(0x60000000 | ((int)np->cork.tclass << 20));
1403
1404 if (skb->len <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN)
1405 hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
1406 else
1407 hdr->payload_len = 0;
1408 hdr->hop_limit = np->cork.hop_limit;
1409 hdr->nexthdr = proto;
1410 ipv6_addr_copy(&hdr->saddr, &fl->fl6_src);
1411 ipv6_addr_copy(&hdr->daddr, final_dst);
1412
1413 skb->priority = sk->sk_priority;
1414
1415 skb->dst = dst_clone(&rt->u.dst);
1416 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
1417 if (proto == IPPROTO_ICMPV6) {
1418 struct inet6_dev *idev = ip6_dst_idev(skb->dst);
1419
1420 ICMP6MSGOUT_INC_STATS_BH(idev, icmp6_hdr(skb)->icmp6_type);
1421 ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTMSGS);
1422 }
1423
1424 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, skb->dst->dev, dst_output);
1425 if (err) {
1426 if (err > 0)
1427 err = np->recverr ? net_xmit_errno(err) : 0;
1428 if (err)
1429 goto error;
1430 }
1431
1432out:
1433 ip6_cork_release(inet, np);
1434 return err;
1435error:
1436 goto out;
1437}
1438
1439void ip6_flush_pending_frames(struct sock *sk)
1440{
1441 struct sk_buff *skb;
1442
1443 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) {
1444 if (skb->dst)
1445 IP6_INC_STATS(ip6_dst_idev(skb->dst),
1446 IPSTATS_MIB_OUTDISCARDS);
1447 kfree_skb(skb);
1448 }
1449
1450 ip6_cork_release(inet_sk(sk), inet6_sk(sk));
1451}
1452