]> bbs.cooldavid.org Git - net-next-2.6.git/blame - net/ipv6/ip6mr.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: replace uses of NIP6_FMT with %p6
[net-next-2.6.git] / net / ipv6 / ip6mr.c
CommitLineData
7bc570c8
YH		 1/*
2 * Linux IPv6 multicast routing support for BSD pim6sd
3 * Based on net/ipv4/ipmr.c.
4 *
5 * (c) 2004 Mickael Hoerdt, <hoerdt@clarinet.u-strasbg.fr>
6 * LSIIT Laboratory, Strasbourg, France
7 * (c) 2004 Jean-Philippe Andriot, <jean-philippe.andriot@6WIND.com>
8 * 6WIND, Paris, France
9 * Copyright (C)2007,2008 USAGI/WIDE Project
10 * YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
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 */
18
19#include <asm/system.h>
20#include <asm/uaccess.h>
21#include <linux/types.h>
22#include <linux/sched.h>
23#include <linux/errno.h>
24#include <linux/timer.h>
25#include <linux/mm.h>
26#include <linux/kernel.h>
27#include <linux/fcntl.h>
28#include <linux/stat.h>
29#include <linux/socket.h>
7bc570c8
YH		 30#include <linux/inet.h>
31#include <linux/netdevice.h>
32#include <linux/inetdevice.h>
7bc570c8
YH		 33#include <linux/proc_fs.h>
34#include <linux/seq_file.h>
7bc570c8 35#include <linux/init.h>
7bc570c8
YH		 36#include <net/protocol.h>
37#include <linux/skbuff.h>
38#include <net/sock.h>
7bc570c8 39#include <net/raw.h>
7bc570c8
YH		 40#include <linux/notifier.h>
41#include <linux/if_arp.h>
7bc570c8
YH		 42#include <net/checksum.h>
43#include <net/netlink.h>
44
45#include <net/ipv6.h>
46#include <net/ip6_route.h>
47#include <linux/mroute6.h>
14fb64e1 48#include <linux/pim.h>
7bc570c8
YH		 49#include <net/addrconf.h>
50#include <linux/netfilter_ipv6.h>
51
52struct sock *mroute6_socket;
53
54
55/* Big lock, protecting vif table, mrt cache and mroute socket state.
56 Note that the changes are semaphored via rtnl_lock.
57 */
58
59static DEFINE_RWLOCK(mrt_lock);
60
61/*
62 * Multicast router control variables
63 */
64
65static struct mif_device vif6_table[MAXMIFS]; /* Devices */
66static int maxvif;
67
68#define MIF_EXISTS(idx) (vif6_table[idx].dev != NULL)
69
14fb64e1
YH		 70static int mroute_do_assert; /* Set in PIM assert */
71#ifdef CONFIG_IPV6_PIMSM_V2
72static int mroute_do_pim;
73#else
74#define mroute_do_pim 0
75#endif
76
6ac7eb08 77static struct mfc6_cache *mfc6_cache_array[MFC6_LINES]; /* Forwarding cache */
7bc570c8
YH		 78
79static struct mfc6_cache *mfc_unres_queue; /* Queue of unresolved entries */
80static atomic_t cache_resolve_queue_len; /* Size of unresolved */
81
82/* Special spinlock for queue of unresolved entries */
83static DEFINE_SPINLOCK(mfc_unres_lock);
84
85/* We return to original Alan's scheme. Hash table of resolved
86 entries is changed only in process context and protected
87 with weak lock mrt_lock. Queue of unresolved entries is protected
88 with strong spinlock mfc_unres_lock.
89
90 In this case data path is free of exclusive locks at all.
91 */
92
93static struct kmem_cache *mrt_cachep __read_mostly;
94
95static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache);
6ac7eb08 96static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert);
7bc570c8
YH		 97static int ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm);
98
14fb64e1
YH		 99#ifdef CONFIG_IPV6_PIMSM_V2
100static struct inet6_protocol pim6_protocol;
101#endif
102
7bc570c8
YH		 103static struct timer_list ipmr_expire_timer;
104
105
106#ifdef CONFIG_PROC_FS
107
108struct ipmr_mfc_iter {
109 struct mfc6_cache **cache;
110 int ct;
111};
112
113
114static struct mfc6_cache *ipmr_mfc_seq_idx(struct ipmr_mfc_iter *it, loff_t pos)
115{
116 struct mfc6_cache *mfc;
117
118 it->cache = mfc6_cache_array;
119 read_lock(&mrt_lock);
120 for (it->ct = 0; it->ct < ARRAY_SIZE(mfc6_cache_array); it->ct++)
121 for (mfc = mfc6_cache_array[it->ct]; mfc; mfc = mfc->next)
122 if (pos-- == 0)
123 return mfc;
124 read_unlock(&mrt_lock);
125
126 it->cache = &mfc_unres_queue;
127 spin_lock_bh(&mfc_unres_lock);
128 for (mfc = mfc_unres_queue; mfc; mfc = mfc->next)
129 if (pos-- == 0)
130 return mfc;
131 spin_unlock_bh(&mfc_unres_lock);
132
133 it->cache = NULL;
134 return NULL;
135}
136
137
138
139
140/*
141 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
142 */
143
144struct ipmr_vif_iter {
145 int ct;
146};
147
148static struct mif_device *ip6mr_vif_seq_idx(struct ipmr_vif_iter *iter,
149 loff_t pos)
150{
151 for (iter->ct = 0; iter->ct < maxvif; ++iter->ct) {
152 if (!MIF_EXISTS(iter->ct))
153 continue;
154 if (pos-- == 0)
155 return &vif6_table[iter->ct];
156 }
157 return NULL;
158}
159
160static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
161 __acquires(mrt_lock)
162{
163 read_lock(&mrt_lock);
164 return (*pos ? ip6mr_vif_seq_idx(seq->private, *pos - 1)
165 : SEQ_START_TOKEN);
166}
167
168static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
169{
170 struct ipmr_vif_iter *iter = seq->private;
171
172 ++*pos;
173 if (v == SEQ_START_TOKEN)
174 return ip6mr_vif_seq_idx(iter, 0);
175
176 while (++iter->ct < maxvif) {
177 if (!MIF_EXISTS(iter->ct))
178 continue;
179 return &vif6_table[iter->ct];
180 }
181 return NULL;
182}
183
184static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
185 __releases(mrt_lock)
186{
187 read_unlock(&mrt_lock);
188}
189
190static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
191{
192 if (v == SEQ_START_TOKEN) {
193 seq_puts(seq,
194 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
195 } else {
196 const struct mif_device *vif = v;
197 const char *name = vif->dev ? vif->dev->name : "none";
198
199 seq_printf(seq,
d430a227 200 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
7bc570c8
YH		 201 vif - vif6_table,
202 name, vif->bytes_in, vif->pkt_in,
203 vif->bytes_out, vif->pkt_out,
204 vif->flags);
205 }
206 return 0;
207}
208
209static struct seq_operations ip6mr_vif_seq_ops = {
210 .start = ip6mr_vif_seq_start,
211 .next = ip6mr_vif_seq_next,
212 .stop = ip6mr_vif_seq_stop,
213 .show = ip6mr_vif_seq_show,
214};
215
216static int ip6mr_vif_open(struct inode *inode, struct file *file)
217{
218 return seq_open_private(file, &ip6mr_vif_seq_ops,
219 sizeof(struct ipmr_vif_iter));
220}
221
222static struct file_operations ip6mr_vif_fops = {
223 .owner = THIS_MODULE,
224 .open = ip6mr_vif_open,
225 .read = seq_read,
226 .llseek = seq_lseek,
227 .release = seq_release,
228};
229
230static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
231{
232 return (*pos ? ipmr_mfc_seq_idx(seq->private, *pos - 1)
233 : SEQ_START_TOKEN);
234}
235
236static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
237{
238 struct mfc6_cache *mfc = v;
239 struct ipmr_mfc_iter *it = seq->private;
240
241 ++*pos;
242
243 if (v == SEQ_START_TOKEN)
244 return ipmr_mfc_seq_idx(seq->private, 0);
245
246 if (mfc->next)
247 return mfc->next;
248
249 if (it->cache == &mfc_unres_queue)
250 goto end_of_list;
251
252 BUG_ON(it->cache != mfc6_cache_array);
253
254 while (++it->ct < ARRAY_SIZE(mfc6_cache_array)) {
255 mfc = mfc6_cache_array[it->ct];
256 if (mfc)
257 return mfc;
258 }
259
260 /* exhausted cache_array, show unresolved */
261 read_unlock(&mrt_lock);
262 it->cache = &mfc_unres_queue;
263 it->ct = 0;
264
265 spin_lock_bh(&mfc_unres_lock);
266 mfc = mfc_unres_queue;
267 if (mfc)
268 return mfc;
269
270 end_of_list:
271 spin_unlock_bh(&mfc_unres_lock);
272 it->cache = NULL;
273
274 return NULL;
275}
276
277static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
278{
279 struct ipmr_mfc_iter *it = seq->private;
280
281 if (it->cache == &mfc_unres_queue)
282 spin_unlock_bh(&mfc_unres_lock);
283 else if (it->cache == mfc6_cache_array)
284 read_unlock(&mrt_lock);
285}
286
287static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
288{
289 int n;
290
291 if (v == SEQ_START_TOKEN) {
292 seq_puts(seq,
293 "Group "
294 "Origin "
295 "Iif Pkts Bytes Wrong Oifs\n");
296 } else {
297 const struct mfc6_cache *mfc = v;
298 const struct ipmr_mfc_iter *it = seq->private;
299
0c6ce78a
HH		 300 seq_printf(seq, "%p6 %p6 %-3d %8ld %8ld %8ld",
301 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
7bc570c8
YH		 302 mfc->mf6c_parent,
303 mfc->mfc_un.res.pkt,
304 mfc->mfc_un.res.bytes,
305 mfc->mfc_un.res.wrong_if);
306
307 if (it->cache != &mfc_unres_queue) {
308 for (n = mfc->mfc_un.res.minvif;
309 n < mfc->mfc_un.res.maxvif; n++) {
310 if (MIF_EXISTS(n) &&
311 mfc->mfc_un.res.ttls[n] < 255)
312 seq_printf(seq,
313 " %2d:%-3d",
314 n, mfc->mfc_un.res.ttls[n]);
315 }
316 }
317 seq_putc(seq, '\n');
318 }
319 return 0;
320}
321
322static struct seq_operations ipmr_mfc_seq_ops = {
323 .start = ipmr_mfc_seq_start,
324 .next = ipmr_mfc_seq_next,
325 .stop = ipmr_mfc_seq_stop,
326 .show = ipmr_mfc_seq_show,
327};
328
329static int ipmr_mfc_open(struct inode *inode, struct file *file)
330{
331 return seq_open_private(file, &ipmr_mfc_seq_ops,
332 sizeof(struct ipmr_mfc_iter));
333}
334
335static struct file_operations ip6mr_mfc_fops = {
336 .owner = THIS_MODULE,
337 .open = ipmr_mfc_open,
338 .read = seq_read,
339 .llseek = seq_lseek,
340 .release = seq_release,
341};
342#endif
343
14fb64e1
YH		 344#ifdef CONFIG_IPV6_PIMSM_V2
345static int reg_vif_num = -1;
346
347static int pim6_rcv(struct sk_buff *skb)
348{
349 struct pimreghdr *pim;
350 struct ipv6hdr *encap;
351 struct net_device *reg_dev = NULL;
352
353 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
354 goto drop;
355
356 pim = (struct pimreghdr *)skb_transport_header(skb);
357 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
358 (pim->flags & PIM_NULL_REGISTER) ||
359 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
ec6b486f 360 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
14fb64e1
YH		 361 goto drop;
362
363 /* check if the inner packet is destined to mcast group */
364 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
365 sizeof(*pim));
366
367 if (!ipv6_addr_is_multicast(&encap->daddr) ||
368 encap->payload_len == 0 ||
369 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
370 goto drop;
371
372 read_lock(&mrt_lock);
373 if (reg_vif_num >= 0)
374 reg_dev = vif6_table[reg_vif_num].dev;
375 if (reg_dev)
376 dev_hold(reg_dev);
377 read_unlock(&mrt_lock);
378
379 if (reg_dev == NULL)
380 goto drop;
381
382 skb->mac_header = skb->network_header;
383 skb_pull(skb, (u8 *)encap - skb->data);
384 skb_reset_network_header(skb);
385 skb->dev = reg_dev;
386 skb->protocol = htons(ETH_P_IP);
387 skb->ip_summed = 0;
388 skb->pkt_type = PACKET_HOST;
389 dst_release(skb->dst);
dc58c78c
PE		 390 reg_dev->stats.rx_bytes += skb->len;
391 reg_dev->stats.rx_packets++;
14fb64e1
YH		 392 skb->dst = NULL;
393 nf_reset(skb);
394 netif_rx(skb);
395 dev_put(reg_dev);
396 return 0;
397 drop:
398 kfree_skb(skb);
399 return 0;
400}
401
402static struct inet6_protocol pim6_protocol = {
403 .handler = pim6_rcv,
404};
405
406/* Service routines creating virtual interfaces: PIMREG */
407
408static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
409{
410 read_lock(&mrt_lock);
dc58c78c
PE		 411 dev->stats.tx_bytes += skb->len;
412 dev->stats.tx_packets++;
14fb64e1
YH		 413 ip6mr_cache_report(skb, reg_vif_num, MRT6MSG_WHOLEPKT);
414 read_unlock(&mrt_lock);
415 kfree_skb(skb);
416 return 0;
417}
418
14fb64e1
YH		 419static void reg_vif_setup(struct net_device *dev)
420{
421 dev->type = ARPHRD_PIMREG;
422 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
423 dev->flags = IFF_NOARP;
424 dev->hard_start_xmit = reg_vif_xmit;
14fb64e1
YH		 425 dev->destructor = free_netdev;
426}
427
428static struct net_device *ip6mr_reg_vif(void)
429{
430 struct net_device *dev;
14fb64e1 431
dc58c78c 432 dev = alloc_netdev(0, "pim6reg", reg_vif_setup);
14fb64e1
YH		 433 if (dev == NULL)
434 return NULL;
435
436 if (register_netdevice(dev)) {
437 free_netdev(dev);
438 return NULL;
439 }
440 dev->iflink = 0;
441
14fb64e1
YH		 442 if (dev_open(dev))
443 goto failure;
444
7af3db78 445 dev_hold(dev);
14fb64e1
YH		 446 return dev;
447
448failure:
449 /* allow the register to be completed before unregistering. */
450 rtnl_unlock();
451 rtnl_lock();
452
453 unregister_netdevice(dev);
454 return NULL;
455}
456#endif
457
7bc570c8
YH		 458/*
459 * Delete a VIF entry
460 */
461
462static int mif6_delete(int vifi)
463{
464 struct mif_device *v;
465 struct net_device *dev;
466 if (vifi < 0 || vifi >= maxvif)
467 return -EADDRNOTAVAIL;
468
469 v = &vif6_table[vifi];
470
471 write_lock_bh(&mrt_lock);
472 dev = v->dev;
473 v->dev = NULL;
474
475 if (!dev) {
476 write_unlock_bh(&mrt_lock);
477 return -EADDRNOTAVAIL;
478 }
479
14fb64e1
YH		 480#ifdef CONFIG_IPV6_PIMSM_V2
481 if (vifi == reg_vif_num)
482 reg_vif_num = -1;
483#endif
484
7bc570c8
YH		 485 if (vifi + 1 == maxvif) {
486 int tmp;
487 for (tmp = vifi - 1; tmp >= 0; tmp--) {
488 if (MIF_EXISTS(tmp))
489 break;
490 }
491 maxvif = tmp + 1;
492 }
493
494 write_unlock_bh(&mrt_lock);
495
496 dev_set_allmulti(dev, -1);
497
498 if (v->flags & MIFF_REGISTER)
499 unregister_netdevice(dev);
500
501 dev_put(dev);
502 return 0;
503}
504
505/* Destroy an unresolved cache entry, killing queued skbs
506 and reporting error to netlink readers.
507 */
508
509static void ip6mr_destroy_unres(struct mfc6_cache *c)
510{
511 struct sk_buff *skb;
512
513 atomic_dec(&cache_resolve_queue_len);
514
515 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
516 if (ipv6_hdr(skb)->version == 0) {
517 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
518 nlh->nlmsg_type = NLMSG_ERROR;
519 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
520 skb_trim(skb, nlh->nlmsg_len);
521 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT;
522 rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid);
523 } else
524 kfree_skb(skb);
525 }
526
527 kmem_cache_free(mrt_cachep, c);
528}
529
530
531/* Single timer process for all the unresolved queue. */
532
533static void ipmr_do_expire_process(unsigned long dummy)
534{
535 unsigned long now = jiffies;
536 unsigned long expires = 10 * HZ;
537 struct mfc6_cache *c, **cp;
538
539 cp = &mfc_unres_queue;
540
541 while ((c = *cp) != NULL) {
542 if (time_after(c->mfc_un.unres.expires, now)) {
543 /* not yet... */
544 unsigned long interval = c->mfc_un.unres.expires - now;
545 if (interval < expires)
546 expires = interval;
547 cp = &c->next;
548 continue;
549 }
550
551 *cp = c->next;
552 ip6mr_destroy_unres(c);
553 }
554
555 if (atomic_read(&cache_resolve_queue_len))
556 mod_timer(&ipmr_expire_timer, jiffies + expires);
557}
558
559static void ipmr_expire_process(unsigned long dummy)
560{
561 if (!spin_trylock(&mfc_unres_lock)) {
562 mod_timer(&ipmr_expire_timer, jiffies + 1);
563 return;
564 }
565
566 if (atomic_read(&cache_resolve_queue_len))
567 ipmr_do_expire_process(dummy);
568
569 spin_unlock(&mfc_unres_lock);
570}
571
572/* Fill oifs list. It is called under write locked mrt_lock. */
573
574static void ip6mr_update_thresholds(struct mfc6_cache *cache, unsigned char *ttls)
575{
576 int vifi;
577
6ac7eb08 578 cache->mfc_un.res.minvif = MAXMIFS;
7bc570c8 579 cache->mfc_un.res.maxvif = 0;
6ac7eb08 580 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
7bc570c8
YH		 581
582 for (vifi = 0; vifi < maxvif; vifi++) {
583 if (MIF_EXISTS(vifi) && ttls[vifi] && ttls[vifi] < 255) {
584 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
585 if (cache->mfc_un.res.minvif > vifi)
586 cache->mfc_un.res.minvif = vifi;
587 if (cache->mfc_un.res.maxvif <= vifi)
588 cache->mfc_un.res.maxvif = vifi + 1;
589 }
590 }
591}
592
593static int mif6_add(struct mif6ctl *vifc, int mrtsock)
594{
595 int vifi = vifc->mif6c_mifi;
596 struct mif_device *v = &vif6_table[vifi];
597 struct net_device *dev;
5ae7b444 598 int err;
7bc570c8
YH		 599
600 /* Is vif busy ? */
601 if (MIF_EXISTS(vifi))
602 return -EADDRINUSE;
603
604 switch (vifc->mif6c_flags) {
14fb64e1
YH		 605#ifdef CONFIG_IPV6_PIMSM_V2
606 case MIFF_REGISTER:
607 /*
608 * Special Purpose VIF in PIM
609 * All the packets will be sent to the daemon
610 */
611 if (reg_vif_num >= 0)
612 return -EADDRINUSE;
613 dev = ip6mr_reg_vif();
614 if (!dev)
615 return -ENOBUFS;
5ae7b444
WC		 616 err = dev_set_allmulti(dev, 1);
617 if (err) {
618 unregister_netdevice(dev);
7af3db78 619 dev_put(dev);
5ae7b444
WC		 620 return err;
621 }
14fb64e1
YH		 622 break;
623#endif
7bc570c8
YH		 624 case 0:
625 dev = dev_get_by_index(&init_net, vifc->mif6c_pifi);
626 if (!dev)
627 return -EADDRNOTAVAIL;
5ae7b444 628 err = dev_set_allmulti(dev, 1);
7af3db78
WC		 629 if (err) {
630 dev_put(dev);
5ae7b444 631 return err;
7af3db78 632 }
7bc570c8
YH		 633 break;
634 default:
635 return -EINVAL;
636 }
637
7bc570c8
YH		 638 /*
639 * Fill in the VIF structures
640 */
641 v->rate_limit = vifc->vifc_rate_limit;
642 v->flags = vifc->mif6c_flags;
643 if (!mrtsock)
644 v->flags |= VIFF_STATIC;
645 v->threshold = vifc->vifc_threshold;
646 v->bytes_in = 0;
647 v->bytes_out = 0;
648 v->pkt_in = 0;
649 v->pkt_out = 0;
650 v->link = dev->ifindex;
651 if (v->flags & MIFF_REGISTER)
652 v->link = dev->iflink;
653
654 /* And finish update writing critical data */
655 write_lock_bh(&mrt_lock);
7bc570c8 656 v->dev = dev;
14fb64e1
YH		 657#ifdef CONFIG_IPV6_PIMSM_V2
658 if (v->flags & MIFF_REGISTER)
659 reg_vif_num = vifi;
660#endif
7bc570c8
YH		 661 if (vifi + 1 > maxvif)
662 maxvif = vifi + 1;
663 write_unlock_bh(&mrt_lock);
664 return 0;
665}
666
667static struct mfc6_cache *ip6mr_cache_find(struct in6_addr *origin, struct in6_addr *mcastgrp)
668{
669 int line = MFC6_HASH(mcastgrp, origin);
670 struct mfc6_cache *c;
671
672 for (c = mfc6_cache_array[line]; c; c = c->next) {
673 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
674 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
675 break;
676 }
677 return c;
678}
679
680/*
681 * Allocate a multicast cache entry
682 */
683static struct mfc6_cache *ip6mr_cache_alloc(void)
684{
685 struct mfc6_cache *c = kmem_cache_alloc(mrt_cachep, GFP_KERNEL);
686 if (c == NULL)
687 return NULL;
688 memset(c, 0, sizeof(*c));
6ac7eb08 689 c->mfc_un.res.minvif = MAXMIFS;
7bc570c8
YH		 690 return c;
691}
692
693static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
694{
695 struct mfc6_cache *c = kmem_cache_alloc(mrt_cachep, GFP_ATOMIC);
696 if (c == NULL)
697 return NULL;
698 memset(c, 0, sizeof(*c));
699 skb_queue_head_init(&c->mfc_un.unres.unresolved);
700 c->mfc_un.unres.expires = jiffies + 10 * HZ;
701 return c;
702}
703
704/*
705 * A cache entry has gone into a resolved state from queued
706 */
707
708static void ip6mr_cache_resolve(struct mfc6_cache *uc, struct mfc6_cache *c)
709{
710 struct sk_buff *skb;
711
712 /*
713 * Play the pending entries through our router
714 */
715
716 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
717 if (ipv6_hdr(skb)->version == 0) {
718 int err;
719 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
720
721 if (ip6mr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) {
549e028d 722 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
7bc570c8
YH		 723 } else {
724 nlh->nlmsg_type = NLMSG_ERROR;
725 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
726 skb_trim(skb, nlh->nlmsg_len);
727 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE;
728 }
729 err = rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid);
730 } else
731 ip6_mr_forward(skb, c);
732 }
733}
734
735/*
736 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
737 * expects the following bizarre scheme.
738 *
739 * Called under mrt_lock.
740 */
741
6ac7eb08 742static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert)
7bc570c8
YH		 743{
744 struct sk_buff *skb;
745 struct mrt6msg *msg;
746 int ret;
747
14fb64e1
YH		 748#ifdef CONFIG_IPV6_PIMSM_V2
749 if (assert == MRT6MSG_WHOLEPKT)
750 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
751 +sizeof(*msg));
752 else
753#endif
754 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
7bc570c8
YH		 755
756 if (!skb)
757 return -ENOBUFS;
758
759 /* I suppose that internal messages
760 * do not require checksums */
761
762 skb->ip_summed = CHECKSUM_UNNECESSARY;
763
14fb64e1
YH		 764#ifdef CONFIG_IPV6_PIMSM_V2
765 if (assert == MRT6MSG_WHOLEPKT) {
766 /* Ugly, but we have no choice with this interface.
767 Duplicate old header, fix length etc.
768 And all this only to mangle msg->im6_msgtype and
769 to set msg->im6_mbz to "mbz" :-)
770 */
771 skb_push(skb, -skb_network_offset(pkt));
772
773 skb_push(skb, sizeof(*msg));
774 skb_reset_transport_header(skb);
775 msg = (struct mrt6msg *)skb_transport_header(skb);
776 msg->im6_mbz = 0;
777 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
778 msg->im6_mif = reg_vif_num;
779 msg->im6_pad = 0;
780 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
781 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
782
783 skb->ip_summed = CHECKSUM_UNNECESSARY;
784 } else
785#endif
786 {
7bc570c8
YH		 787 /*
788 * Copy the IP header
789 */
790
791 skb_put(skb, sizeof(struct ipv6hdr));
792 skb_reset_network_header(skb);
793 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
794
795 /*
796 * Add our header
797 */
798 skb_put(skb, sizeof(*msg));
799 skb_reset_transport_header(skb);
800 msg = (struct mrt6msg *)skb_transport_header(skb);
801
802 msg->im6_mbz = 0;
803 msg->im6_msgtype = assert;
6ac7eb08 804 msg->im6_mif = mifi;
7bc570c8
YH		 805 msg->im6_pad = 0;
806 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
807 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
808
809 skb->dst = dst_clone(pkt->dst);
810 skb->ip_summed = CHECKSUM_UNNECESSARY;
811
812 skb_pull(skb, sizeof(struct ipv6hdr));
14fb64e1 813 }
7bc570c8
YH		 814
815 if (mroute6_socket == NULL) {
816 kfree_skb(skb);
817 return -EINVAL;
818 }
819
820 /*
821 * Deliver to user space multicast routing algorithms
822 */
823 if ((ret = sock_queue_rcv_skb(mroute6_socket, skb)) < 0) {
824 if (net_ratelimit())
825 printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n");
826 kfree_skb(skb);
827 }
828
829 return ret;
830}
831
832/*
833 * Queue a packet for resolution. It gets locked cache entry!
834 */
835
836static int
6ac7eb08 837ip6mr_cache_unresolved(mifi_t mifi, struct sk_buff *skb)
7bc570c8
YH		 838{
839 int err;
840 struct mfc6_cache *c;
841
842 spin_lock_bh(&mfc_unres_lock);
843 for (c = mfc_unres_queue; c; c = c->next) {
844 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
845 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr))
846 break;
847 }
848
849 if (c == NULL) {
850 /*
851 * Create a new entry if allowable
852 */
853
854 if (atomic_read(&cache_resolve_queue_len) >= 10 ||
855 (c = ip6mr_cache_alloc_unres()) == NULL) {
856 spin_unlock_bh(&mfc_unres_lock);
857
858 kfree_skb(skb);
859 return -ENOBUFS;
860 }
861
862 /*
863 * Fill in the new cache entry
864 */
865 c->mf6c_parent = -1;
866 c->mf6c_origin = ipv6_hdr(skb)->saddr;
867 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
868
869 /*
870 * Reflect first query at pim6sd
871 */
6ac7eb08 872 if ((err = ip6mr_cache_report(skb, mifi, MRT6MSG_NOCACHE)) < 0) {
7bc570c8
YH		 873 /* If the report failed throw the cache entry
874 out - Brad Parker
875 */
876 spin_unlock_bh(&mfc_unres_lock);
877
878 kmem_cache_free(mrt_cachep, c);
879 kfree_skb(skb);
880 return err;
881 }
882
883 atomic_inc(&cache_resolve_queue_len);
884 c->next = mfc_unres_queue;
885 mfc_unres_queue = c;
886
887 ipmr_do_expire_process(1);
888 }
889
890 /*
891 * See if we can append the packet
892 */
893 if (c->mfc_un.unres.unresolved.qlen > 3) {
894 kfree_skb(skb);
895 err = -ENOBUFS;
896 } else {
897 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
898 err = 0;
899 }
900
901 spin_unlock_bh(&mfc_unres_lock);
902 return err;
903}
904
905/*
906 * MFC6 cache manipulation by user space
907 */
908
909static int ip6mr_mfc_delete(struct mf6cctl *mfc)
910{
911 int line;
912 struct mfc6_cache *c, **cp;
913
914 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
915
916 for (cp = &mfc6_cache_array[line]; (c = *cp) != NULL; cp = &c->next) {
917 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
918 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
919 write_lock_bh(&mrt_lock);
920 *cp = c->next;
921 write_unlock_bh(&mrt_lock);
922
923 kmem_cache_free(mrt_cachep, c);
924 return 0;
925 }
926 }
927 return -ENOENT;
928}
929
930static int ip6mr_device_event(struct notifier_block *this,
931 unsigned long event, void *ptr)
932{
933 struct net_device *dev = ptr;
934 struct mif_device *v;
935 int ct;
936
721499e8 937 if (!net_eq(dev_net(dev), &init_net))
7bc570c8
YH		 938 return NOTIFY_DONE;
939
940 if (event != NETDEV_UNREGISTER)
941 return NOTIFY_DONE;
942
943 v = &vif6_table[0];
944 for (ct = 0; ct < maxvif; ct++, v++) {
945 if (v->dev == dev)
946 mif6_delete(ct);
947 }
948 return NOTIFY_DONE;
949}
950
951static struct notifier_block ip6_mr_notifier = {
952 .notifier_call = ip6mr_device_event
953};
954
955/*
956 * Setup for IP multicast routing
957 */
958
623d1a1a 959int __init ip6_mr_init(void)
7bc570c8 960{
623d1a1a
WC		 961 int err;
962
7bc570c8
YH		 963 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
964 sizeof(struct mfc6_cache),
965 0, SLAB_HWCACHE_ALIGN,
966 NULL);
967 if (!mrt_cachep)
623d1a1a 968 return -ENOMEM;
7bc570c8
YH		 969
970 setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0);
623d1a1a
WC		 971 err = register_netdevice_notifier(&ip6_mr_notifier);
972 if (err)
973 goto reg_notif_fail;
974#ifdef CONFIG_PROC_FS
975 err = -ENOMEM;
976 if (!proc_net_fops_create(&init_net, "ip6_mr_vif", 0, &ip6mr_vif_fops))
977 goto proc_vif_fail;
978 if (!proc_net_fops_create(&init_net, "ip6_mr_cache",
979 0, &ip6mr_mfc_fops))
980 goto proc_cache_fail;
981#endif
982 return 0;
983reg_notif_fail:
984 kmem_cache_destroy(mrt_cachep);
7bc570c8 985#ifdef CONFIG_PROC_FS
623d1a1a
WC		 986proc_vif_fail:
987 unregister_netdevice_notifier(&ip6_mr_notifier);
988proc_cache_fail:
989 proc_net_remove(&init_net, "ip6_mr_vif");
7bc570c8 990#endif
623d1a1a 991 return err;
7bc570c8
YH		 992}
993
623d1a1a
WC		 994void ip6_mr_cleanup(void)
995{
996#ifdef CONFIG_PROC_FS
997 proc_net_remove(&init_net, "ip6_mr_cache");
998 proc_net_remove(&init_net, "ip6_mr_vif");
999#endif
1000 unregister_netdevice_notifier(&ip6_mr_notifier);
1001 del_timer(&ipmr_expire_timer);
1002 kmem_cache_destroy(mrt_cachep);
1003}
7bc570c8
YH		 1004
1005static int ip6mr_mfc_add(struct mf6cctl *mfc, int mrtsock)
1006{
1007 int line;
1008 struct mfc6_cache *uc, *c, **cp;
6ac7eb08 1009 unsigned char ttls[MAXMIFS];
7bc570c8
YH		 1010 int i;
1011
6ac7eb08
RR		 1012 memset(ttls, 255, MAXMIFS);
1013 for (i = 0; i < MAXMIFS; i++) {
7bc570c8
YH		 1014 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1015 ttls[i] = 1;
1016
1017 }
1018
1019 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1020
1021 for (cp = &mfc6_cache_array[line]; (c = *cp) != NULL; cp = &c->next) {
1022 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1023 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr))
1024 break;
1025 }
1026
1027 if (c != NULL) {
1028 write_lock_bh(&mrt_lock);
1029 c->mf6c_parent = mfc->mf6cc_parent;
1030 ip6mr_update_thresholds(c, ttls);
1031 if (!mrtsock)
1032 c->mfc_flags |= MFC_STATIC;
1033 write_unlock_bh(&mrt_lock);
1034 return 0;
1035 }
1036
1037 if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1038 return -EINVAL;
1039
1040 c = ip6mr_cache_alloc();
1041 if (c == NULL)
1042 return -ENOMEM;
1043
1044 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1045 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1046 c->mf6c_parent = mfc->mf6cc_parent;
1047 ip6mr_update_thresholds(c, ttls);
1048 if (!mrtsock)
1049 c->mfc_flags |= MFC_STATIC;
1050
1051 write_lock_bh(&mrt_lock);
1052 c->next = mfc6_cache_array[line];
1053 mfc6_cache_array[line] = c;
1054 write_unlock_bh(&mrt_lock);
1055
1056 /*
1057 * Check to see if we resolved a queued list. If so we
1058 * need to send on the frames and tidy up.
1059 */
1060 spin_lock_bh(&mfc_unres_lock);
1061 for (cp = &mfc_unres_queue; (uc = *cp) != NULL;
1062 cp = &uc->next) {
1063 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1064 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1065 *cp = uc->next;
1066 if (atomic_dec_and_test(&cache_resolve_queue_len))
1067 del_timer(&ipmr_expire_timer);
1068 break;
1069 }
1070 }
1071 spin_unlock_bh(&mfc_unres_lock);
1072
1073 if (uc) {
1074 ip6mr_cache_resolve(uc, c);
1075 kmem_cache_free(mrt_cachep, uc);
1076 }
1077 return 0;
1078}
1079
1080/*
1081 * Close the multicast socket, and clear the vif tables etc
1082 */
1083
1084static void mroute_clean_tables(struct sock *sk)
1085{
1086 int i;
1087
1088 /*
1089 * Shut down all active vif entries
1090 */
1091 for (i = 0; i < maxvif; i++) {
1092 if (!(vif6_table[i].flags & VIFF_STATIC))
1093 mif6_delete(i);
1094 }
1095
1096 /*
1097 * Wipe the cache
1098 */
1099 for (i = 0; i < ARRAY_SIZE(mfc6_cache_array); i++) {
1100 struct mfc6_cache *c, **cp;
1101
1102 cp = &mfc6_cache_array[i];
1103 while ((c = *cp) != NULL) {
1104 if (c->mfc_flags & MFC_STATIC) {
1105 cp = &c->next;
1106 continue;
1107 }
1108 write_lock_bh(&mrt_lock);
1109 *cp = c->next;
1110 write_unlock_bh(&mrt_lock);
1111
1112 kmem_cache_free(mrt_cachep, c);
1113 }
1114 }
1115
1116 if (atomic_read(&cache_resolve_queue_len) != 0) {
1117 struct mfc6_cache *c;
1118
1119 spin_lock_bh(&mfc_unres_lock);
1120 while (mfc_unres_queue != NULL) {
1121 c = mfc_unres_queue;
1122 mfc_unres_queue = c->next;
1123 spin_unlock_bh(&mfc_unres_lock);
1124
1125 ip6mr_destroy_unres(c);
1126
1127 spin_lock_bh(&mfc_unres_lock);
1128 }
1129 spin_unlock_bh(&mfc_unres_lock);
1130 }
1131}
1132
1133static int ip6mr_sk_init(struct sock *sk)
1134{
1135 int err = 0;
1136
1137 rtnl_lock();
1138 write_lock_bh(&mrt_lock);
1139 if (likely(mroute6_socket == NULL))
1140 mroute6_socket = sk;
1141 else
1142 err = -EADDRINUSE;
1143 write_unlock_bh(&mrt_lock);
1144
1145 rtnl_unlock();
1146
1147 return err;
1148}
1149
1150int ip6mr_sk_done(struct sock *sk)
1151{
1152 int err = 0;
1153
1154 rtnl_lock();
1155 if (sk == mroute6_socket) {
1156 write_lock_bh(&mrt_lock);
1157 mroute6_socket = NULL;
1158 write_unlock_bh(&mrt_lock);
1159
1160 mroute_clean_tables(sk);
1161 } else
1162 err = -EACCES;
1163 rtnl_unlock();
1164
1165 return err;
1166}
1167
1168/*
1169 * Socket options and virtual interface manipulation. The whole
1170 * virtual interface system is a complete heap, but unfortunately
1171 * that's how BSD mrouted happens to think. Maybe one day with a proper
1172 * MOSPF/PIM router set up we can clean this up.
1173 */
1174
1175int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, int optlen)
1176{
1177 int ret;
1178 struct mif6ctl vif;
1179 struct mf6cctl mfc;
1180 mifi_t mifi;
1181
1182 if (optname != MRT6_INIT) {
1183 if (sk != mroute6_socket && !capable(CAP_NET_ADMIN))
1184 return -EACCES;
1185 }
1186
1187 switch (optname) {
1188 case MRT6_INIT:
1189 if (sk->sk_type != SOCK_RAW ||
1190 inet_sk(sk)->num != IPPROTO_ICMPV6)
1191 return -EOPNOTSUPP;
1192 if (optlen < sizeof(int))
1193 return -EINVAL;
1194
1195 return ip6mr_sk_init(sk);
1196
1197 case MRT6_DONE:
1198 return ip6mr_sk_done(sk);
1199
1200 case MRT6_ADD_MIF:
1201 if (optlen < sizeof(vif))
1202 return -EINVAL;
1203 if (copy_from_user(&vif, optval, sizeof(vif)))
1204 return -EFAULT;
6ac7eb08 1205 if (vif.mif6c_mifi >= MAXMIFS)
7bc570c8
YH		 1206 return -ENFILE;
1207 rtnl_lock();
1208 ret = mif6_add(&vif, sk == mroute6_socket);
1209 rtnl_unlock();
1210 return ret;
1211
1212 case MRT6_DEL_MIF:
1213 if (optlen < sizeof(mifi_t))
1214 return -EINVAL;
1215 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1216 return -EFAULT;
1217 rtnl_lock();
1218 ret = mif6_delete(mifi);
1219 rtnl_unlock();
1220 return ret;
1221
1222 /*
1223 * Manipulate the forwarding caches. These live
1224 * in a sort of kernel/user symbiosis.
1225 */
1226 case MRT6_ADD_MFC:
1227 case MRT6_DEL_MFC:
1228 if (optlen < sizeof(mfc))
1229 return -EINVAL;
1230 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1231 return -EFAULT;
1232 rtnl_lock();
1233 if (optname == MRT6_DEL_MFC)
1234 ret = ip6mr_mfc_delete(&mfc);
1235 else
1236 ret = ip6mr_mfc_add(&mfc, sk == mroute6_socket);
1237 rtnl_unlock();
1238 return ret;
1239
14fb64e1
YH		 1240 /*
1241 * Control PIM assert (to activate pim will activate assert)
1242 */
1243 case MRT6_ASSERT:
1244 {
1245 int v;
1246 if (get_user(v, (int __user *)optval))
1247 return -EFAULT;
1248 mroute_do_assert = !!v;
1249 return 0;
1250 }
1251
1252#ifdef CONFIG_IPV6_PIMSM_V2
1253 case MRT6_PIM:
1254 {
a9f83bf3 1255 int v;
14fb64e1
YH		 1256 if (get_user(v, (int __user *)optval))
1257 return -EFAULT;
1258 v = !!v;
1259 rtnl_lock();
1260 ret = 0;
1261 if (v != mroute_do_pim) {
1262 mroute_do_pim = v;
1263 mroute_do_assert = v;
1264 if (mroute_do_pim)
1265 ret = inet6_add_protocol(&pim6_protocol,
1266 IPPROTO_PIM);
1267 else
1268 ret = inet6_del_protocol(&pim6_protocol,
1269 IPPROTO_PIM);
1270 if (ret < 0)
1271 ret = -EAGAIN;
1272 }
1273 rtnl_unlock();
1274 return ret;
1275 }
1276
1277#endif
7bc570c8 1278 /*
7d120c55 1279 * Spurious command, or MRT6_VERSION which you cannot
7bc570c8
YH		 1280 * set.
1281 */
1282 default:
1283 return -ENOPROTOOPT;
1284 }
1285}
1286
1287/*
1288 * Getsock opt support for the multicast routing system.
1289 */
1290
1291int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1292 int __user *optlen)
1293{
1294 int olr;
1295 int val;
1296
1297 switch (optname) {
1298 case MRT6_VERSION:
1299 val = 0x0305;
1300 break;
14fb64e1
YH		 1301#ifdef CONFIG_IPV6_PIMSM_V2
1302 case MRT6_PIM:
1303 val = mroute_do_pim;
1304 break;
1305#endif
1306 case MRT6_ASSERT:
1307 val = mroute_do_assert;
1308 break;
7bc570c8
YH		 1309 default:
1310 return -ENOPROTOOPT;
1311 }
1312
1313 if (get_user(olr, optlen))
1314 return -EFAULT;
1315
1316 olr = min_t(int, olr, sizeof(int));
1317 if (olr < 0)
1318 return -EINVAL;
1319
1320 if (put_user(olr, optlen))
1321 return -EFAULT;
1322 if (copy_to_user(optval, &val, olr))
1323 return -EFAULT;
1324 return 0;
1325}
1326
1327/*
1328 * The IP multicast ioctl support routines.
1329 */
1330
1331int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1332{
1333 struct sioc_sg_req6 sr;
1334 struct sioc_mif_req6 vr;
1335 struct mif_device *vif;
1336 struct mfc6_cache *c;
1337
1338 switch (cmd) {
1339 case SIOCGETMIFCNT_IN6:
1340 if (copy_from_user(&vr, arg, sizeof(vr)))
1341 return -EFAULT;
1342 if (vr.mifi >= maxvif)
1343 return -EINVAL;
1344 read_lock(&mrt_lock);
1345 vif = &vif6_table[vr.mifi];
1346 if (MIF_EXISTS(vr.mifi)) {
1347 vr.icount = vif->pkt_in;
1348 vr.ocount = vif->pkt_out;
1349 vr.ibytes = vif->bytes_in;
1350 vr.obytes = vif->bytes_out;
1351 read_unlock(&mrt_lock);
1352
1353 if (copy_to_user(arg, &vr, sizeof(vr)))
1354 return -EFAULT;
1355 return 0;
1356 }
1357 read_unlock(&mrt_lock);
1358 return -EADDRNOTAVAIL;
1359 case SIOCGETSGCNT_IN6:
1360 if (copy_from_user(&sr, arg, sizeof(sr)))
1361 return -EFAULT;
1362
1363 read_lock(&mrt_lock);
1364 c = ip6mr_cache_find(&sr.src.sin6_addr, &sr.grp.sin6_addr);
1365 if (c) {
1366 sr.pktcnt = c->mfc_un.res.pkt;
1367 sr.bytecnt = c->mfc_un.res.bytes;
1368 sr.wrong_if = c->mfc_un.res.wrong_if;
1369 read_unlock(&mrt_lock);
1370
1371 if (copy_to_user(arg, &sr, sizeof(sr)))
1372 return -EFAULT;
1373 return 0;
1374 }
1375 read_unlock(&mrt_lock);
1376 return -EADDRNOTAVAIL;
1377 default:
1378 return -ENOIOCTLCMD;
1379 }
1380}
1381
1382
1383static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1384{
483a47d2
DL		 1385 IP6_INC_STATS_BH(dev_net(skb->dst->dev), ip6_dst_idev(skb->dst),
1386 IPSTATS_MIB_OUTFORWDATAGRAMS);
7bc570c8
YH		 1387 return dst_output(skb);
1388}
1389
1390/*
1391 * Processing handlers for ip6mr_forward
1392 */
1393
1394static int ip6mr_forward2(struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1395{
1396 struct ipv6hdr *ipv6h;
1397 struct mif_device *vif = &vif6_table[vifi];
1398 struct net_device *dev;
1399 struct dst_entry *dst;
1400 struct flowi fl;
1401
1402 if (vif->dev == NULL)
1403 goto out_free;
1404
14fb64e1
YH		 1405#ifdef CONFIG_IPV6_PIMSM_V2
1406 if (vif->flags & MIFF_REGISTER) {
1407 vif->pkt_out++;
1408 vif->bytes_out += skb->len;
dc58c78c
PE		 1409 vif->dev->stats.tx_bytes += skb->len;
1410 vif->dev->stats.tx_packets++;
14fb64e1
YH		 1411 ip6mr_cache_report(skb, vifi, MRT6MSG_WHOLEPKT);
1412 kfree_skb(skb);
1413 return 0;
1414 }
1415#endif
1416
7bc570c8
YH		 1417 ipv6h = ipv6_hdr(skb);
1418
1419 fl = (struct flowi) {
1420 .oif = vif->link,
1421 .nl_u = { .ip6_u =
1422 { .daddr = ipv6h->daddr, }
1423 }
1424 };
1425
1426 dst = ip6_route_output(&init_net, NULL, &fl);
1427 if (!dst)
1428 goto out_free;
1429
1430 dst_release(skb->dst);
1431 skb->dst = dst;
1432
1433 /*
1434 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1435 * not only before forwarding, but after forwarding on all output
1436 * interfaces. It is clear, if mrouter runs a multicasting
1437 * program, it should receive packets not depending to what interface
1438 * program is joined.
1439 * If we will not make it, the program will have to join on all
1440 * interfaces. On the other hand, multihoming host (or router, but
1441 * not mrouter) cannot join to more than one interface - it will
1442 * result in receiving multiple packets.
1443 */
1444 dev = vif->dev;
1445 skb->dev = dev;
1446 vif->pkt_out++;
1447 vif->bytes_out += skb->len;
1448
1449 /* We are about to write */
1450 /* XXX: extension headers? */
1451 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
1452 goto out_free;
1453
1454 ipv6h = ipv6_hdr(skb);
1455 ipv6h->hop_limit--;
1456
1457 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
1458
1459 return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dev,
1460 ip6mr_forward2_finish);
1461
1462out_free:
1463 kfree_skb(skb);
1464 return 0;
1465}
1466
1467static int ip6mr_find_vif(struct net_device *dev)
1468{
1469 int ct;
1470 for (ct = maxvif - 1; ct >= 0; ct--) {
1471 if (vif6_table[ct].dev == dev)
1472 break;
1473 }
1474 return ct;
1475}
1476
1477static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache)
1478{
1479 int psend = -1;
1480 int vif, ct;
1481
1482 vif = cache->mf6c_parent;
1483 cache->mfc_un.res.pkt++;
1484 cache->mfc_un.res.bytes += skb->len;
1485
14fb64e1
YH		 1486 /*
1487 * Wrong interface: drop packet and (maybe) send PIM assert.
1488 */
1489 if (vif6_table[vif].dev != skb->dev) {
1490 int true_vifi;
1491
1492 cache->mfc_un.res.wrong_if++;
1493 true_vifi = ip6mr_find_vif(skb->dev);
1494
1495 if (true_vifi >= 0 && mroute_do_assert &&
1496 /* pimsm uses asserts, when switching from RPT to SPT,
1497 so that we cannot check that packet arrived on an oif.
1498 It is bad, but otherwise we would need to move pretty
1499 large chunk of pimd to kernel. Ough... --ANK
1500 */
1501 (mroute_do_pim || cache->mfc_un.res.ttls[true_vifi] < 255) &&
1502 time_after(jiffies,
1503 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1504 cache->mfc_un.res.last_assert = jiffies;
1505 ip6mr_cache_report(skb, true_vifi, MRT6MSG_WRONGMIF);
1506 }
1507 goto dont_forward;
1508 }
1509
7bc570c8
YH		 1510 vif6_table[vif].pkt_in++;
1511 vif6_table[vif].bytes_in += skb->len;
1512
1513 /*
1514 * Forward the frame
1515 */
1516 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
1517 if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
1518 if (psend != -1) {
1519 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1520 if (skb2)
1521 ip6mr_forward2(skb2, cache, psend);
1522 }
1523 psend = ct;
1524 }
1525 }
1526 if (psend != -1) {
1527 ip6mr_forward2(skb, cache, psend);
1528 return 0;
1529 }
1530
14fb64e1 1531dont_forward:
7bc570c8
YH		 1532 kfree_skb(skb);
1533 return 0;
1534}
1535
1536
1537/*
1538 * Multicast packets for forwarding arrive here
1539 */
1540
1541int ip6_mr_input(struct sk_buff *skb)
1542{
1543 struct mfc6_cache *cache;
1544
1545 read_lock(&mrt_lock);
1546 cache = ip6mr_cache_find(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
1547
1548 /*
1549 * No usable cache entry
1550 */
1551 if (cache == NULL) {
1552 int vif;
1553
1554 vif = ip6mr_find_vif(skb->dev);
1555 if (vif >= 0) {
1556 int err = ip6mr_cache_unresolved(vif, skb);
1557 read_unlock(&mrt_lock);
1558
1559 return err;
1560 }
1561 read_unlock(&mrt_lock);
1562 kfree_skb(skb);
1563 return -ENODEV;
1564 }
1565
1566 ip6_mr_forward(skb, cache);
1567
1568 read_unlock(&mrt_lock);
1569
1570 return 0;
1571}
1572
1573
1574static int
1575ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm)
1576{
1577 int ct;
1578 struct rtnexthop *nhp;
1579 struct net_device *dev = vif6_table[c->mf6c_parent].dev;
549e028d 1580 u8 *b = skb_tail_pointer(skb);
7bc570c8
YH		 1581 struct rtattr *mp_head;
1582
1583 if (dev)
1584 RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex);
1585
1586 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
1587
1588 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
1589 if (c->mfc_un.res.ttls[ct] < 255) {
1590 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
1591 goto rtattr_failure;
1592 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
1593 nhp->rtnh_flags = 0;
1594 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
1595 nhp->rtnh_ifindex = vif6_table[ct].dev->ifindex;
1596 nhp->rtnh_len = sizeof(*nhp);
1597 }
1598 }
1599 mp_head->rta_type = RTA_MULTIPATH;
549e028d 1600 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
7bc570c8
YH		 1601 rtm->rtm_type = RTN_MULTICAST;
1602 return 1;
1603
1604rtattr_failure:
1605 nlmsg_trim(skb, b);
1606 return -EMSGSIZE;
1607}
1608
1609int ip6mr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait)
1610{
1611 int err;
1612 struct mfc6_cache *cache;
1613 struct rt6_info *rt = (struct rt6_info *)skb->dst;
1614
1615 read_lock(&mrt_lock);
1616 cache = ip6mr_cache_find(&rt->rt6i_src.addr, &rt->rt6i_dst.addr);
1617
1618 if (!cache) {
1619 struct sk_buff *skb2;
1620 struct ipv6hdr *iph;
1621 struct net_device *dev;
1622 int vif;
1623
1624 if (nowait) {
1625 read_unlock(&mrt_lock);
1626 return -EAGAIN;
1627 }
1628
1629 dev = skb->dev;
1630 if (dev == NULL || (vif = ip6mr_find_vif(dev)) < 0) {
1631 read_unlock(&mrt_lock);
1632 return -ENODEV;
1633 }
1634
1635 /* really correct? */
1636 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
1637 if (!skb2) {
1638 read_unlock(&mrt_lock);
1639 return -ENOMEM;
1640 }
1641
1642 skb_reset_transport_header(skb2);
1643
1644 skb_put(skb2, sizeof(struct ipv6hdr));
1645 skb_reset_network_header(skb2);
1646
1647 iph = ipv6_hdr(skb2);
1648 iph->version = 0;
1649 iph->priority = 0;
1650 iph->flow_lbl[0] = 0;
1651 iph->flow_lbl[1] = 0;
1652 iph->flow_lbl[2] = 0;
1653 iph->payload_len = 0;
1654 iph->nexthdr = IPPROTO_NONE;
1655 iph->hop_limit = 0;
1656 ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr);
1657 ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr);
1658
1659 err = ip6mr_cache_unresolved(vif, skb2);
1660 read_unlock(&mrt_lock);
1661
1662 return err;
1663 }
1664
1665 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
1666 cache->mfc_flags |= MFC_NOTIFY;
1667
1668 err = ip6mr_fill_mroute(skb, cache, rtm);
1669 read_unlock(&mrt_lock);
1670 return err;
1671}
1672
Clone of davem's net-next-2.6 tree