projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
[IPV6]: Audit all ip6_dst_lookup/ip6_dst_store calls
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ipv6 / ip6_output.c
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
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
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 {
75
76 struct dst_entry *dst = skb->dst;
77 struct hh_cache *hh = dst->hh;
78
79 if (hh) {
80 int hh_alen;
81
82 read_lock_bh(&hh->hh_lock);
83 hh_alen = HH_DATA_ALIGN(hh->hh_len);
84 memcpy(skb->data - hh_alen, hh->hh_data, hh_alen);
85 read_unlock_bh(&hh->hh_lock);
86 skb_push(skb, hh->hh_len);
87 return hh->hh_output(skb);
88 } else if (dst->neighbour)
89 return dst->neighbour->output(skb);
90
91 IP6_INC_STATS_BH(IPSTATS_MIB_OUTNOROUTES);
92 kfree_skb(skb);
93 return -EINVAL;
94
95 }
96
97 /* dev_loopback_xmit for use with netfilter. */
98 static int ip6_dev_loopback_xmit(struct sk_buff *newskb)
99 {
100 newskb->mac.raw = newskb->data;
101 __skb_pull(newskb, newskb->nh.raw - newskb->data);
102 newskb->pkt_type = PACKET_LOOPBACK;
103 newskb->ip_summed = CHECKSUM_UNNECESSARY;
104 BUG_TRAP(newskb->dst);
105
106 netif_rx(newskb);
107 return 0;
108 }
109
110
111 static int ip6_output2(struct sk_buff *skb)
112 {
113 struct dst_entry *dst = skb->dst;
114 struct net_device *dev = dst->dev;
115
116 skb->protocol = htons(ETH_P_IPV6);
117 skb->dev = dev;
118
119 if (ipv6_addr_is_multicast(&skb->nh.ipv6h->daddr)) {
120 struct ipv6_pinfo* np = skb->sk ? inet6_sk(skb->sk) : NULL;
121
122 if (!(dev->flags & IFF_LOOPBACK) && (!np || np->mc_loop) &&
123 ipv6_chk_mcast_addr(dev, &skb->nh.ipv6h->daddr,
124 &skb->nh.ipv6h->saddr)) {
125 struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
126
127 /* Do not check for IFF_ALLMULTI; multicast routing
128 is not supported in any case.
129 */
130 if (newskb)
131 NF_HOOK(PF_INET6, NF_IP6_POST_ROUTING, newskb, NULL,
132 newskb->dev,
133 ip6_dev_loopback_xmit);
134
135 if (skb->nh.ipv6h->hop_limit == 0) {
136 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
137 kfree_skb(skb);
138 return 0;
139 }
140 }
141
142 IP6_INC_STATS(IPSTATS_MIB_OUTMCASTPKTS);
143 }
144
145 return NF_HOOK(PF_INET6, NF_IP6_POST_ROUTING, skb,NULL, skb->dev,ip6_output_finish);
146 }
147
148 int ip6_output(struct sk_buff *skb)
149 {
150 if ((skb->len > dst_mtu(skb->dst) && !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 * xmit an sk_buff (used by TCP)
159 */
160
161 int 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 int head_room;
175
176 /* First: exthdrs may take lots of space (~8K for now)
177 MAX_HEADER is not enough.
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 kfree_skb(skb);
186 skb = skb2;
187 if (skb == NULL) {
188 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
189 return -ENOBUFS;
190 }
191 if (sk)
192 skb_set_owner_w(skb, sk);
193 }
194 if (opt->opt_flen)
195 ipv6_push_frag_opts(skb, opt, &proto);
196 if (opt->opt_nflen)
197 ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
198 }
199
200 hdr = skb->nh.ipv6h = (struct ipv6hdr*)skb_push(skb, sizeof(struct ipv6hdr));
201
202 /*
203 * Fill in the IPv6 header
204 */
205
206 hlimit = -1;
207 if (np)
208 hlimit = np->hop_limit;
209 if (hlimit < 0)
210 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
211 if (hlimit < 0)
212 hlimit = ipv6_get_hoplimit(dst->dev);
213
214 tclass = -1;
215 if (np)
216 tclass = np->tclass;
217 if (tclass < 0)
218 tclass = 0;
219
220 *(u32 *)hdr = htonl(0x60000000 | (tclass << 20)) | fl->fl6_flowlabel;
221
222 hdr->payload_len = htons(seg_len);
223 hdr->nexthdr = proto;
224 hdr->hop_limit = hlimit;
225
226 ipv6_addr_copy(&hdr->saddr, &fl->fl6_src);
227 ipv6_addr_copy(&hdr->daddr, first_hop);
228
229 skb->priority = sk->sk_priority;
230
231 mtu = dst_mtu(dst);
232 if ((skb->len <= mtu) || ipfragok || skb_is_gso(skb)) {
233 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
234 return NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev,
235 dst_output);
236 }
237
238 if (net_ratelimit())
239 printk(KERN_DEBUG "IPv6: sending pkt_too_big to self\n");
240 skb->dev = dst->dev;
241 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, skb->dev);
242 IP6_INC_STATS(IPSTATS_MIB_FRAGFAILS);
243 kfree_skb(skb);
244 return -EMSGSIZE;
245 }
246
247 /*
248 * To avoid extra problems ND packets are send through this
249 * routine. It's code duplication but I really want to avoid
250 * extra checks since ipv6_build_header is used by TCP (which
251 * is for us performance critical)
252 */
253
254 int ip6_nd_hdr(struct sock *sk, struct sk_buff *skb, struct net_device *dev,
255 struct in6_addr *saddr, struct in6_addr *daddr,
256 int proto, int len)
257 {
258 struct ipv6_pinfo *np = inet6_sk(sk);
259 struct ipv6hdr *hdr;
260 int totlen;
261
262 skb->protocol = htons(ETH_P_IPV6);
263 skb->dev = dev;
264
265 totlen = len + sizeof(struct ipv6hdr);
266
267 hdr = (struct ipv6hdr *) skb_put(skb, sizeof(struct ipv6hdr));
268 skb->nh.ipv6h = hdr;
269
270 *(u32*)hdr = htonl(0x60000000);
271
272 hdr->payload_len = htons(len);
273 hdr->nexthdr = proto;
274 hdr->hop_limit = np->hop_limit;
275
276 ipv6_addr_copy(&hdr->saddr, saddr);
277 ipv6_addr_copy(&hdr->daddr, daddr);
278
279 return 0;
280 }
281
282 static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
283 {
284 struct ip6_ra_chain *ra;
285 struct sock *last = NULL;
286
287 read_lock(&ip6_ra_lock);
288 for (ra = ip6_ra_chain; ra; ra = ra->next) {
289 struct sock *sk = ra->sk;
290 if (sk && ra->sel == sel &&
291 (!sk->sk_bound_dev_if ||
292 sk->sk_bound_dev_if == skb->dev->ifindex)) {
293 if (last) {
294 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
295 if (skb2)
296 rawv6_rcv(last, skb2);
297 }
298 last = sk;
299 }
300 }
301
302 if (last) {
303 rawv6_rcv(last, skb);
304 read_unlock(&ip6_ra_lock);
305 return 1;
306 }
307 read_unlock(&ip6_ra_lock);
308 return 0;
309 }
310
311 static inline int ip6_forward_finish(struct sk_buff *skb)
312 {
313 return dst_output(skb);
314 }
315
316 int ip6_forward(struct sk_buff *skb)
317 {
318 struct dst_entry *dst = skb->dst;
319 struct ipv6hdr *hdr = skb->nh.ipv6h;
320 struct inet6_skb_parm *opt = IP6CB(skb);
321
322 if (ipv6_devconf.forwarding == 0)
323 goto error;
324
325 if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
326 IP6_INC_STATS(IPSTATS_MIB_INDISCARDS);
327 goto drop;
328 }
329
330 skb->ip_summed = CHECKSUM_NONE;
331
332 /*
333 * We DO NOT make any processing on
334 * RA packets, pushing them to user level AS IS
335 * without ane WARRANTY that application will be able
336 * to interpret them. The reason is that we
337 * cannot make anything clever here.
338 *
339 * We are not end-node, so that if packet contains
340 * AH/ESP, we cannot make anything.
341 * Defragmentation also would be mistake, RA packets
342 * cannot be fragmented, because there is no warranty
343 * that different fragments will go along one path. --ANK
344 */
345 if (opt->ra) {
346 u8 *ptr = skb->nh.raw + opt->ra;
347 if (ip6_call_ra_chain(skb, (ptr[2]<<8) + ptr[3]))
348 return 0;
349 }
350
351 /*
352 * check and decrement ttl
353 */
354 if (hdr->hop_limit <= 1) {
355 /* Force OUTPUT device used as source address */
356 skb->dev = dst->dev;
357 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
358 0, skb->dev);
359
360 kfree_skb(skb);
361 return -ETIMEDOUT;
362 }
363
364 if (!xfrm6_route_forward(skb)) {
365 IP6_INC_STATS(IPSTATS_MIB_INDISCARDS);
366 goto drop;
367 }
368 dst = skb->dst;
369
370 /* IPv6 specs say nothing about it, but it is clear that we cannot
371 send redirects to source routed frames.
372 */
373 if (skb->dev == dst->dev && dst->neighbour && opt->srcrt == 0) {
374 struct in6_addr *target = NULL;
375 struct rt6_info *rt;
376 struct neighbour *n = dst->neighbour;
377
378 /*
379 * incoming and outgoing devices are the same
380 * send a redirect.
381 */
382
383 rt = (struct rt6_info *) dst;
384 if ((rt->rt6i_flags & RTF_GATEWAY))
385 target = (struct in6_addr*)&n->primary_key;
386 else
387 target = &hdr->daddr;
388
389 /* Limit redirects both by destination (here)
390 and by source (inside ndisc_send_redirect)
391 */
392 if (xrlim_allow(dst, 1*HZ))
393 ndisc_send_redirect(skb, n, target);
394 } else if (ipv6_addr_type(&hdr->saddr)&(IPV6_ADDR_MULTICAST|IPV6_ADDR_LOOPBACK
395 |IPV6_ADDR_LINKLOCAL)) {
396 /* This check is security critical. */
397 goto error;
398 }
399
400 if (skb->len > dst_mtu(dst)) {
401 /* Again, force OUTPUT device used as source address */
402 skb->dev = dst->dev;
403 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, dst_mtu(dst), skb->dev);
404 IP6_INC_STATS_BH(IPSTATS_MIB_INTOOBIGERRORS);
405 IP6_INC_STATS_BH(IPSTATS_MIB_FRAGFAILS);
406 kfree_skb(skb);
407 return -EMSGSIZE;
408 }
409
410 if (skb_cow(skb, dst->dev->hard_header_len)) {
411 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
412 goto drop;
413 }
414
415 hdr = skb->nh.ipv6h;
416
417 /* Mangling hops number delayed to point after skb COW */
418
419 hdr->hop_limit--;
420
421 IP6_INC_STATS_BH(IPSTATS_MIB_OUTFORWDATAGRAMS);
422 return NF_HOOK(PF_INET6,NF_IP6_FORWARD, skb, skb->dev, dst->dev, ip6_forward_finish);
423
424 error:
425 IP6_INC_STATS_BH(IPSTATS_MIB_INADDRERRORS);
426 drop:
427 kfree_skb(skb);
428 return -EINVAL;
429 }
430
431 static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
432 {
433 to->pkt_type = from->pkt_type;
434 to->priority = from->priority;
435 to->protocol = from->protocol;
436 dst_release(to->dst);
437 to->dst = dst_clone(from->dst);
438 to->dev = from->dev;
439
440 #ifdef CONFIG_NET_SCHED
441 to->tc_index = from->tc_index;
442 #endif
443 #ifdef CONFIG_NETFILTER
444 to->nfmark = from->nfmark;
445 /* Connection association is same as pre-frag packet */
446 nf_conntrack_put(to->nfct);
447 to->nfct = from->nfct;
448 nf_conntrack_get(to->nfct);
449 to->nfctinfo = from->nfctinfo;
450 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
451 nf_conntrack_put_reasm(to->nfct_reasm);
452 to->nfct_reasm = from->nfct_reasm;
453 nf_conntrack_get_reasm(to->nfct_reasm);
454 #endif
455 #ifdef CONFIG_BRIDGE_NETFILTER
456 nf_bridge_put(to->nf_bridge);
457 to->nf_bridge = from->nf_bridge;
458 nf_bridge_get(to->nf_bridge);
459 #endif
460 #endif
461 skb_copy_secmark(to, from);
462 }
463
464 int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr)
465 {
466 u16 offset = sizeof(struct ipv6hdr);
467 struct ipv6_opt_hdr *exthdr = (struct ipv6_opt_hdr*)(skb->nh.ipv6h + 1);
468 unsigned int packet_len = skb->tail - skb->nh.raw;
469 int found_rhdr = 0;
470 *nexthdr = &skb->nh.ipv6h->nexthdr;
471
472 while (offset + 1 <= packet_len) {
473
474 switch (**nexthdr) {
475
476 case NEXTHDR_HOP:
477 case NEXTHDR_ROUTING:
478 case NEXTHDR_DEST:
479 if (**nexthdr == NEXTHDR_ROUTING) found_rhdr = 1;
480 if (**nexthdr == NEXTHDR_DEST && found_rhdr) return offset;
481 offset += ipv6_optlen(exthdr);
482 *nexthdr = &exthdr->nexthdr;
483 exthdr = (struct ipv6_opt_hdr*)(skb->nh.raw + offset);
484 break;
485 default :
486 return offset;
487 }
488 }
489
490 return offset;
491 }
492 EXPORT_SYMBOL_GPL(ip6_find_1stfragopt);
493
494 static int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
495 {
496 struct net_device *dev;
497 struct sk_buff *frag;
498 struct rt6_info *rt = (struct rt6_info*)skb->dst;
499 struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
500 struct ipv6hdr *tmp_hdr;
501 struct frag_hdr *fh;
502 unsigned int mtu, hlen, left, len;
503 u32 frag_id = 0;
504 int ptr, offset = 0, err=0;
505 u8 *prevhdr, nexthdr = 0;
506
507 dev = rt->u.dst.dev;
508 hlen = ip6_find_1stfragopt(skb, &prevhdr);
509 nexthdr = *prevhdr;
510
511 mtu = dst_mtu(&rt->u.dst);
512 if (np && np->frag_size < mtu) {
513 if (np->frag_size)
514 mtu = np->frag_size;
515 }
516 mtu -= hlen + sizeof(struct frag_hdr);
517
518 if (skb_shinfo(skb)->frag_list) {
519 int first_len = skb_pagelen(skb);
520
521 if (first_len - hlen > mtu ||
522 ((first_len - hlen) & 7) ||
523 skb_cloned(skb))
524 goto slow_path;
525
526 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next) {
527 /* Correct geometry. */
528 if (frag->len > mtu ||
529 ((frag->len & 7) && frag->next) ||
530 skb_headroom(frag) < hlen)
531 goto slow_path;
532
533 /* Partially cloned skb? */
534 if (skb_shared(frag))
535 goto slow_path;
536
537 BUG_ON(frag->sk);
538 if (skb->sk) {
539 sock_hold(skb->sk);
540 frag->sk = skb->sk;
541 frag->destructor = sock_wfree;
542 skb->truesize -= frag->truesize;
543 }
544 }
545
546 err = 0;
547 offset = 0;
548 frag = skb_shinfo(skb)->frag_list;
549 skb_shinfo(skb)->frag_list = NULL;
550 /* BUILD HEADER */
551
552 tmp_hdr = kmalloc(hlen, GFP_ATOMIC);
553 if (!tmp_hdr) {
554 IP6_INC_STATS(IPSTATS_MIB_FRAGFAILS);
555 return -ENOMEM;
556 }
557
558 *prevhdr = NEXTHDR_FRAGMENT;
559 memcpy(tmp_hdr, skb->nh.raw, hlen);
560 __skb_pull(skb, hlen);
561 fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr));
562 skb->nh.raw = __skb_push(skb, hlen);
563 memcpy(skb->nh.raw, tmp_hdr, hlen);
564
565 ipv6_select_ident(skb, fh);
566 fh->nexthdr = nexthdr;
567 fh->reserved = 0;
568 fh->frag_off = htons(IP6_MF);
569 frag_id = fh->identification;
570
571 first_len = skb_pagelen(skb);
572 skb->data_len = first_len - skb_headlen(skb);
573 skb->len = first_len;
574 skb->nh.ipv6h->payload_len = htons(first_len - sizeof(struct ipv6hdr));
575
576
577 for (;;) {
578 /* Prepare header of the next frame,
579 * before previous one went down. */
580 if (frag) {
581 frag->ip_summed = CHECKSUM_NONE;
582 frag->h.raw = frag->data;
583 fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr));
584 frag->nh.raw = __skb_push(frag, hlen);
585 memcpy(frag->nh.raw, tmp_hdr, hlen);
586 offset += skb->len - hlen - sizeof(struct frag_hdr);
587 fh->nexthdr = nexthdr;
588 fh->reserved = 0;
589 fh->frag_off = htons(offset);
590 if (frag->next != NULL)
591 fh->frag_off |= htons(IP6_MF);
592 fh->identification = frag_id;
593 frag->nh.ipv6h->payload_len = htons(frag->len - sizeof(struct ipv6hdr));
594 ip6_copy_metadata(frag, skb);
595 }
596
597 err = output(skb);
598 if (err || !frag)
599 break;
600
601 skb = frag;
602 frag = skb->next;
603 skb->next = NULL;
604 }
605
606 kfree(tmp_hdr);
607
608 if (err == 0) {
609 IP6_INC_STATS(IPSTATS_MIB_FRAGOKS);
610 return 0;
611 }
612
613 while (frag) {
614 skb = frag->next;
615 kfree_skb(frag);
616 frag = skb;
617 }
618
619 IP6_INC_STATS(IPSTATS_MIB_FRAGFAILS);
620 return err;
621 }
622
623 slow_path:
624 left = skb->len - hlen; /* Space per frame */
625 ptr = hlen; /* Where to start from */
626
627 /*
628 * Fragment the datagram.
629 */
630
631 *prevhdr = NEXTHDR_FRAGMENT;
632
633 /*
634 * Keep copying data until we run out.
635 */
636 while(left > 0) {
637 len = left;
638 /* IF: it doesn't fit, use 'mtu' - the data space left */
639 if (len > mtu)
640 len = mtu;
641 /* IF: we are not sending upto and including the packet end
642 then align the next start on an eight byte boundary */
643 if (len < left) {
644 len &= ~7;
645 }
646 /*
647 * Allocate buffer.
648 */
649
650 if ((frag = alloc_skb(len+hlen+sizeof(struct frag_hdr)+LL_RESERVED_SPACE(rt->u.dst.dev), GFP_ATOMIC)) == NULL) {
651 NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n");
652 IP6_INC_STATS(IPSTATS_MIB_FRAGFAILS);
653 err = -ENOMEM;
654 goto fail;
655 }
656
657 /*
658 * Set up data on packet
659 */
660
661 ip6_copy_metadata(frag, skb);
662 skb_reserve(frag, LL_RESERVED_SPACE(rt->u.dst.dev));
663 skb_put(frag, len + hlen + sizeof(struct frag_hdr));
664 frag->nh.raw = frag->data;
665 fh = (struct frag_hdr*)(frag->data + hlen);
666 frag->h.raw = frag->data + hlen + sizeof(struct frag_hdr);
667
668 /*
669 * Charge the memory for the fragment to any owner
670 * it might possess
671 */
672 if (skb->sk)
673 skb_set_owner_w(frag, skb->sk);
674
675 /*
676 * Copy the packet header into the new buffer.
677 */
678 memcpy(frag->nh.raw, skb->data, hlen);
679
680 /*
681 * Build fragment header.
682 */
683 fh->nexthdr = nexthdr;
684 fh->reserved = 0;
685 if (!frag_id) {
686 ipv6_select_ident(skb, fh);
687 frag_id = fh->identification;
688 } else
689 fh->identification = frag_id;
690
691 /*
692 * Copy a block of the IP datagram.
693 */
694 if (skb_copy_bits(skb, ptr, frag->h.raw, len))
695 BUG();
696 left -= len;
697
698 fh->frag_off = htons(offset);
699 if (left > 0)
700 fh->frag_off |= htons(IP6_MF);
701 frag->nh.ipv6h->payload_len = htons(frag->len - sizeof(struct ipv6hdr));
702
703 ptr += len;
704 offset += len;
705
706 /*
707 * Put this fragment into the sending queue.
708 */
709
710 IP6_INC_STATS(IPSTATS_MIB_FRAGCREATES);
711
712 err = output(frag);
713 if (err)
714 goto fail;
715 }
716 kfree_skb(skb);
717 IP6_INC_STATS(IPSTATS_MIB_FRAGOKS);
718 return err;
719
720 fail:
721 kfree_skb(skb);
722 IP6_INC_STATS(IPSTATS_MIB_FRAGFAILS);
723 return err;
724 }
725
726 static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
727 struct dst_entry *dst,
728 struct flowi *fl)
729 {
730 struct ipv6_pinfo *np = inet6_sk(sk);
731 struct rt6_info *rt = (struct rt6_info *)dst;
732
733 if (!dst)
734 goto out;
735
736 /* Yes, checking route validity in not connected
737 * case is not very simple. Take into account,
738 * that we do not support routing by source, TOS,
739 * and MSG_DONTROUTE --ANK (980726)
740 *
741 * 1. If route was host route, check that
742 * cached destination is current.
743 * If it is network route, we still may
744 * check its validity using saved pointer
745 * to the last used address: daddr_cache.
746 * We do not want to save whole address now,
747 * (because main consumer of this service
748 * is tcp, which has not this problem),
749 * so that the last trick works only on connected
750 * sockets.
751 * 2. oif also should be the same.
752 */
753 if (((rt->rt6i_dst.plen != 128 ||
754 !ipv6_addr_equal(&fl->fl6_dst, &rt->rt6i_dst.addr))
755 && (np->daddr_cache == NULL ||
756 !ipv6_addr_equal(&fl->fl6_dst, np->daddr_cache)))
757 || (fl->oif && fl->oif != dst->dev->ifindex)) {
758 dst_release(dst);
759 dst = NULL;
760 }
761
762 out:
763 return dst;
764 }
765
766 static int ip6_dst_lookup_tail(struct sock *sk,
767 struct dst_entry **dst, struct flowi *fl)
768 {
769 int err;
770
771 if (*dst == NULL)
772 *dst = ip6_route_output(sk, fl);
773
774 if ((err = (*dst)->error))
775 goto out_err_release;
776
777 if (ipv6_addr_any(&fl->fl6_src)) {
778 err = ipv6_get_saddr(*dst, &fl->fl6_dst, &fl->fl6_src);
779 if (err)
780 goto out_err_release;
781 }
782
783 return 0;
784
785 out_err_release:
786 dst_release(*dst);
787 *dst = NULL;
788 return err;
789 }
790
791 /**
792 * ip6_dst_lookup - perform route lookup on flow
793 * @sk: socket which provides route info
794 * @dst: pointer to dst_entry * for result
795 * @fl: flow to lookup
796 *
797 * This function performs a route lookup on the given flow.
798 *
799 * It returns zero on success, or a standard errno code on error.
800 */
801 int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl)
802 {
803 *dst = NULL;
804 return ip6_dst_lookup_tail(sk, dst, fl);
805 }
806 EXPORT_SYMBOL_GPL(ip6_dst_lookup);
807
808 /**
809 * ip6_sk_dst_lookup - perform socket cached route lookup on flow
810 * @sk: socket which provides the dst cache and route info
811 * @dst: pointer to dst_entry * for result
812 * @fl: flow to lookup
813 *
814 * This function performs a route lookup on the given flow with the
815 * possibility of using the cached route in the socket if it is valid.
816 * It will take the socket dst lock when operating on the dst cache.
817 * As a result, this function can only be used in process context.
818 *
819 * It returns zero on success, or a standard errno code on error.
820 */
821 int ip6_sk_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi *fl)
822 {
823 *dst = NULL;
824 if (sk) {
825 *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
826 *dst = ip6_sk_dst_check(sk, *dst, fl);
827 }
828
829 return ip6_dst_lookup_tail(sk, dst, fl);
830 }
831 EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup);
832
833 static inline int ip6_ufo_append_data(struct sock *sk,
834 int getfrag(void *from, char *to, int offset, int len,
835 int odd, struct sk_buff *skb),
836 void *from, int length, int hh_len, int fragheaderlen,
837 int transhdrlen, int mtu,unsigned int flags)
838
839 {
840 struct sk_buff *skb;
841 int err;
842
843 /* There is support for UDP large send offload by network
844 * device, so create one single skb packet containing complete
845 * udp datagram
846 */
847 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) {
848 skb = sock_alloc_send_skb(sk,
849 hh_len + fragheaderlen + transhdrlen + 20,
850 (flags & MSG_DONTWAIT), &err);
851 if (skb == NULL)
852 return -ENOMEM;
853
854 /* reserve space for Hardware header */
855 skb_reserve(skb, hh_len);
856
857 /* create space for UDP/IP header */
858 skb_put(skb,fragheaderlen + transhdrlen);
859
860 /* initialize network header pointer */
861 skb->nh.raw = skb->data;
862
863 /* initialize protocol header pointer */
864 skb->h.raw = skb->data + fragheaderlen;
865
866 skb->ip_summed = CHECKSUM_HW;
867 skb->csum = 0;
868 sk->sk_sndmsg_off = 0;
869 }
870
871 err = skb_append_datato_frags(sk,skb, getfrag, from,
872 (length - transhdrlen));
873 if (!err) {
874 struct frag_hdr fhdr;
875
876 /* specify the length of each IP datagram fragment*/
877 skb_shinfo(skb)->gso_size = mtu - fragheaderlen -
878 sizeof(struct frag_hdr);
879 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
880 ipv6_select_ident(skb, &fhdr);
881 skb_shinfo(skb)->ip6_frag_id = fhdr.identification;
882 __skb_queue_tail(&sk->sk_write_queue, skb);
883
884 return 0;
885 }
886 /* There is not enough support do UPD LSO,
887 * so follow normal path
888 */
889 kfree_skb(skb);
890
891 return err;
892 }
893
894 int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
895 int offset, int len, int odd, struct sk_buff *skb),
896 void *from, int length, int transhdrlen,
897 int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi *fl,
898 struct rt6_info *rt, unsigned int flags)
899 {
900 struct inet_sock *inet = inet_sk(sk);
901 struct ipv6_pinfo *np = inet6_sk(sk);
902 struct sk_buff *skb;
903 unsigned int maxfraglen, fragheaderlen;
904 int exthdrlen;
905 int hh_len;
906 int mtu;
907 int copy;
908 int err;
909 int offset = 0;
910 int csummode = CHECKSUM_NONE;
911
912 if (flags&MSG_PROBE)
913 return 0;
914 if (skb_queue_empty(&sk->sk_write_queue)) {
915 /*
916 * setup for corking
917 */
918 if (opt) {
919 if (np->cork.opt == NULL) {
920 np->cork.opt = kmalloc(opt->tot_len,
921 sk->sk_allocation);
922 if (unlikely(np->cork.opt == NULL))
923 return -ENOBUFS;
924 } else if (np->cork.opt->tot_len < opt->tot_len) {
925 printk(KERN_DEBUG "ip6_append_data: invalid option length\n");
926 return -EINVAL;
927 }
928 memcpy(np->cork.opt, opt, opt->tot_len);
929 inet->cork.flags |= IPCORK_OPT;
930 /* need source address above miyazawa*/
931 }
932 dst_hold(&rt->u.dst);
933 np->cork.rt = rt;
934 inet->cork.fl = *fl;
935 np->cork.hop_limit = hlimit;
936 np->cork.tclass = tclass;
937 mtu = dst_mtu(rt->u.dst.path);
938 if (np->frag_size < mtu) {
939 if (np->frag_size)
940 mtu = np->frag_size;
941 }
942 inet->cork.fragsize = mtu;
943 if (dst_allfrag(rt->u.dst.path))
944 inet->cork.flags |= IPCORK_ALLFRAG;
945 inet->cork.length = 0;
946 sk->sk_sndmsg_page = NULL;
947 sk->sk_sndmsg_off = 0;
948 exthdrlen = rt->u.dst.header_len + (opt ? opt->opt_flen : 0);
949 length += exthdrlen;
950 transhdrlen += exthdrlen;
951 } else {
952 rt = np->cork.rt;
953 fl = &inet->cork.fl;
954 if (inet->cork.flags & IPCORK_OPT)
955 opt = np->cork.opt;
956 transhdrlen = 0;
957 exthdrlen = 0;
958 mtu = inet->cork.fragsize;
959 }
960
961 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
962
963 fragheaderlen = sizeof(struct ipv6hdr) + (opt ? opt->opt_nflen : 0);
964 maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr);
965
966 if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
967 if (inet->cork.length + length > sizeof(struct ipv6hdr) + IPV6_MAXPLEN - fragheaderlen) {
968 ipv6_local_error(sk, EMSGSIZE, fl, mtu-exthdrlen);
969 return -EMSGSIZE;
970 }
971 }
972
973 /*
974 * Let's try using as much space as possible.
975 * Use MTU if total length of the message fits into the MTU.
976 * Otherwise, we need to reserve fragment header and
977 * fragment alignment (= 8-15 octects, in total).
978 *
979 * Note that we may need to "move" the data from the tail of
980 * of the buffer to the new fragment when we split
981 * the message.
982 *
983 * FIXME: It may be fragmented into multiple chunks
984 * at once if non-fragmentable extension headers
985 * are too large.
986 * --yoshfuji
987 */
988
989 inet->cork.length += length;
990 if (((length > mtu) && (sk->sk_protocol == IPPROTO_UDP)) &&
991 (rt->u.dst.dev->features & NETIF_F_UFO)) {
992
993 err = ip6_ufo_append_data(sk, getfrag, from, length, hh_len,
994 fragheaderlen, transhdrlen, mtu,
995 flags);
996 if (err)
997 goto error;
998 return 0;
999 }
1000
1001 if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL)
1002 goto alloc_new_skb;
1003
1004 while (length > 0) {
1005 /* Check if the remaining data fits into current packet. */
1006 copy = (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
1007 if (copy < length)
1008 copy = maxfraglen - skb->len;
1009
1010 if (copy <= 0) {
1011 char *data;
1012 unsigned int datalen;
1013 unsigned int fraglen;
1014 unsigned int fraggap;
1015 unsigned int alloclen;
1016 struct sk_buff *skb_prev;
1017 alloc_new_skb:
1018 skb_prev = skb;
1019
1020 /* There's no room in the current skb */
1021 if (skb_prev)
1022 fraggap = skb_prev->len - maxfraglen;
1023 else
1024 fraggap = 0;
1025
1026 /*
1027 * If remaining data exceeds the mtu,
1028 * we know we need more fragment(s).
1029 */
1030 datalen = length + fraggap;
1031 if (datalen > (inet->cork.length <= mtu && !(inet->cork.flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
1032 datalen = maxfraglen - fragheaderlen;
1033
1034 fraglen = datalen + fragheaderlen;
1035 if ((flags & MSG_MORE) &&
1036 !(rt->u.dst.dev->features&NETIF_F_SG))
1037 alloclen = mtu;
1038 else
1039 alloclen = datalen + fragheaderlen;
1040
1041 /*
1042 * The last fragment gets additional space at tail.
1043 * Note: we overallocate on fragments with MSG_MODE
1044 * because we have no idea if we're the last one.
1045 */
1046 if (datalen == length + fraggap)
1047 alloclen += rt->u.dst.trailer_len;
1048
1049 /*
1050 * We just reserve space for fragment header.
1051 * Note: this may be overallocation if the message
1052 * (without MSG_MORE) fits into the MTU.
1053 */
1054 alloclen += sizeof(struct frag_hdr);
1055
1056 if (transhdrlen) {
1057 skb = sock_alloc_send_skb(sk,
1058 alloclen + hh_len,
1059 (flags & MSG_DONTWAIT), &err);
1060 } else {
1061 skb = NULL;
1062 if (atomic_read(&sk->sk_wmem_alloc) <=
1063 2 * sk->sk_sndbuf)
1064 skb = sock_wmalloc(sk,
1065 alloclen + hh_len, 1,
1066 sk->sk_allocation);
1067 if (unlikely(skb == NULL))
1068 err = -ENOBUFS;
1069 }
1070 if (skb == NULL)
1071 goto error;
1072 /*
1073 * Fill in the control structures
1074 */
1075 skb->ip_summed = csummode;
1076 skb->csum = 0;
1077 /* reserve for fragmentation */
1078 skb_reserve(skb, hh_len+sizeof(struct frag_hdr));
1079
1080 /*
1081 * Find where to start putting bytes
1082 */
1083 data = skb_put(skb, fraglen);
1084 skb->nh.raw = data + exthdrlen;
1085 data += fragheaderlen;
1086 skb->h.raw = data + exthdrlen;
1087
1088 if (fraggap) {
1089 skb->csum = skb_copy_and_csum_bits(
1090 skb_prev, maxfraglen,
1091 data + transhdrlen, fraggap, 0);
1092 skb_prev->csum = csum_sub(skb_prev->csum,
1093 skb->csum);
1094 data += fraggap;
1095 skb_trim(skb_prev, maxfraglen);
1096 }
1097 copy = datalen - transhdrlen - fraggap;
1098 if (copy < 0) {
1099 err = -EINVAL;
1100 kfree_skb(skb);
1101 goto error;
1102 } else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
1103 err = -EFAULT;
1104 kfree_skb(skb);
1105 goto error;
1106 }
1107
1108 offset += copy;
1109 length -= datalen - fraggap;
1110 transhdrlen = 0;
1111 exthdrlen = 0;
1112 csummode = CHECKSUM_NONE;
1113
1114 /*
1115 * Put the packet on the pending queue
1116 */
1117 __skb_queue_tail(&sk->sk_write_queue, skb);
1118 continue;
1119 }
1120
1121 if (copy > length)
1122 copy = length;
1123
1124 if (!(rt->u.dst.dev->features&NETIF_F_SG)) {
1125 unsigned int off;
1126
1127 off = skb->len;
1128 if (getfrag(from, skb_put(skb, copy),
1129 offset, copy, off, skb) < 0) {
1130 __skb_trim(skb, off);
1131 err = -EFAULT;
1132 goto error;
1133 }
1134 } else {
1135 int i = skb_shinfo(skb)->nr_frags;
1136 skb_frag_t *frag = &skb_shinfo(skb)->frags[i-1];
1137 struct page *page = sk->sk_sndmsg_page;
1138 int off = sk->sk_sndmsg_off;
1139 unsigned int left;
1140
1141 if (page && (left = PAGE_SIZE - off) > 0) {
1142 if (copy >= left)
1143 copy = left;
1144 if (page != frag->page) {
1145 if (i == MAX_SKB_FRAGS) {
1146 err = -EMSGSIZE;
1147 goto error;
1148 }
1149 get_page(page);
1150 skb_fill_page_desc(skb, i, page, sk->sk_sndmsg_off, 0);
1151 frag = &skb_shinfo(skb)->frags[i];
1152 }
1153 } else if(i < MAX_SKB_FRAGS) {
1154 if (copy > PAGE_SIZE)
1155 copy = PAGE_SIZE;
1156 page = alloc_pages(sk->sk_allocation, 0);
1157 if (page == NULL) {
1158 err = -ENOMEM;
1159 goto error;
1160 }
1161 sk->sk_sndmsg_page = page;
1162 sk->sk_sndmsg_off = 0;
1163
1164 skb_fill_page_desc(skb, i, page, 0, 0);
1165 frag = &skb_shinfo(skb)->frags[i];
1166 skb->truesize += PAGE_SIZE;
1167 atomic_add(PAGE_SIZE, &sk->sk_wmem_alloc);
1168 } else {
1169 err = -EMSGSIZE;
1170 goto error;
1171 }
1172 if (getfrag(from, page_address(frag->page)+frag->page_offset+frag->size, offset, copy, skb->len, skb) < 0) {
1173 err = -EFAULT;
1174 goto error;
1175 }
1176 sk->sk_sndmsg_off += copy;
1177 frag->size += copy;
1178 skb->len += copy;
1179 skb->data_len += copy;
1180 }
1181 offset += copy;
1182 length -= copy;
1183 }
1184 return 0;
1185 error:
1186 inet->cork.length -= length;
1187 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
1188 return err;
1189 }
1190
1191 int ip6_push_pending_frames(struct sock *sk)
1192 {
1193 struct sk_buff *skb, *tmp_skb;
1194 struct sk_buff **tail_skb;
1195 struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
1196 struct inet_sock *inet = inet_sk(sk);
1197 struct ipv6_pinfo *np = inet6_sk(sk);
1198 struct ipv6hdr *hdr;
1199 struct ipv6_txoptions *opt = np->cork.opt;
1200 struct rt6_info *rt = np->cork.rt;
1201 struct flowi *fl = &inet->cork.fl;
1202 unsigned char proto = fl->proto;
1203 int err = 0;
1204
1205 if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL)
1206 goto out;
1207 tail_skb = &(skb_shinfo(skb)->frag_list);
1208
1209 /* move skb->data to ip header from ext header */
1210 if (skb->data < skb->nh.raw)
1211 __skb_pull(skb, skb->nh.raw - skb->data);
1212 while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) {
1213 __skb_pull(tmp_skb, skb->h.raw - skb->nh.raw);
1214 *tail_skb = tmp_skb;
1215 tail_skb = &(tmp_skb->next);
1216 skb->len += tmp_skb->len;
1217 skb->data_len += tmp_skb->len;
1218 skb->truesize += tmp_skb->truesize;
1219 __sock_put(tmp_skb->sk);
1220 tmp_skb->destructor = NULL;
1221 tmp_skb->sk = NULL;
1222 }
1223
1224 ipv6_addr_copy(final_dst, &fl->fl6_dst);
1225 __skb_pull(skb, skb->h.raw - skb->nh.raw);
1226 if (opt && opt->opt_flen)
1227 ipv6_push_frag_opts(skb, opt, &proto);
1228 if (opt && opt->opt_nflen)
1229 ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
1230
1231 skb->nh.ipv6h = hdr = (struct ipv6hdr*) skb_push(skb, sizeof(struct ipv6hdr));
1232
1233 *(u32*)hdr = fl->fl6_flowlabel |
1234 htonl(0x60000000 | ((int)np->cork.tclass << 20));
1235
1236 if (skb->len <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN)
1237 hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
1238 else
1239 hdr->payload_len = 0;
1240 hdr->hop_limit = np->cork.hop_limit;
1241 hdr->nexthdr = proto;
1242 ipv6_addr_copy(&hdr->saddr, &fl->fl6_src);
1243 ipv6_addr_copy(&hdr->daddr, final_dst);
1244
1245 skb->priority = sk->sk_priority;
1246
1247 skb->dst = dst_clone(&rt->u.dst);
1248 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
1249 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, skb->dst->dev, dst_output);
1250 if (err) {
1251 if (err > 0)
1252 err = np->recverr ? net_xmit_errno(err) : 0;
1253 if (err)
1254 goto error;
1255 }
1256
1257 out:
1258 inet->cork.flags &= ~IPCORK_OPT;
1259 kfree(np->cork.opt);
1260 np->cork.opt = NULL;
1261 if (np->cork.rt) {
1262 dst_release(&np->cork.rt->u.dst);
1263 np->cork.rt = NULL;
1264 inet->cork.flags &= ~IPCORK_ALLFRAG;
1265 }
1266 memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
1267 return err;
1268 error:
1269 goto out;
1270 }
1271
1272 void ip6_flush_pending_frames(struct sock *sk)
1273 {
1274 struct inet_sock *inet = inet_sk(sk);
1275 struct ipv6_pinfo *np = inet6_sk(sk);
1276 struct sk_buff *skb;
1277
1278 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) {
1279 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
1280 kfree_skb(skb);
1281 }
1282
1283 inet->cork.flags &= ~IPCORK_OPT;
1284
1285 kfree(np->cork.opt);
1286 np->cork.opt = NULL;
1287 if (np->cork.rt) {
1288 dst_release(&np->cork.rt->u.dst);
1289 np->cork.rt = NULL;
1290 inet->cork.flags &= ~IPCORK_ALLFRAG;
1291 }
1292 memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
1293 }
kernel source for telekom puls (alcatel/mediatek)