blob: 2de3c464fe75b4e3bf1a57fd148dcc739d123761 [file] [log] [blame]
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001/*
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>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090030#include <linux/inet.h>
31#include <linux/netdevice.h>
32#include <linux/inetdevice.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090033#include <linux/proc_fs.h>
34#include <linux/seq_file.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090035#include <linux/init.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090036#include <net/protocol.h>
37#include <linux/skbuff.h>
38#include <net/sock.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090039#include <net/raw.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090040#include <linux/notifier.h>
41#include <linux/if_arp.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090042#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>
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +090048#include <linux/pim.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090049#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
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +090070static 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
Rami Rosen6ac7eb02008-04-10 12:40:10 +030077static struct mfc6_cache *mfc6_cache_array[MFC6_LINES]; /* Forwarding cache */
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090078
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);
Rami Rosen6ac7eb02008-04-10 12:40:10 +030096static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert);
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090097static int ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm);
98
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +090099#ifdef CONFIG_IPV6_PIMSM_V2
100static struct inet6_protocol pim6_protocol;
101#endif
102
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900103static 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,
200 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X\n",
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
300 seq_printf(seq,
301 NIP6_FMT " " NIP6_FMT " %-3d %8ld %8ld %8ld",
302 NIP6(mfc->mf6c_mcastgrp), NIP6(mfc->mf6c_origin),
303 mfc->mf6c_parent,
304 mfc->mfc_un.res.pkt,
305 mfc->mfc_un.res.bytes,
306 mfc->mfc_un.res.wrong_if);
307
308 if (it->cache != &mfc_unres_queue) {
309 for (n = mfc->mfc_un.res.minvif;
310 n < mfc->mfc_un.res.maxvif; n++) {
311 if (MIF_EXISTS(n) &&
312 mfc->mfc_un.res.ttls[n] < 255)
313 seq_printf(seq,
314 " %2d:%-3d",
315 n, mfc->mfc_un.res.ttls[n]);
316 }
317 }
318 seq_putc(seq, '\n');
319 }
320 return 0;
321}
322
323static struct seq_operations ipmr_mfc_seq_ops = {
324 .start = ipmr_mfc_seq_start,
325 .next = ipmr_mfc_seq_next,
326 .stop = ipmr_mfc_seq_stop,
327 .show = ipmr_mfc_seq_show,
328};
329
330static int ipmr_mfc_open(struct inode *inode, struct file *file)
331{
332 return seq_open_private(file, &ipmr_mfc_seq_ops,
333 sizeof(struct ipmr_mfc_iter));
334}
335
336static struct file_operations ip6mr_mfc_fops = {
337 .owner = THIS_MODULE,
338 .open = ipmr_mfc_open,
339 .read = seq_read,
340 .llseek = seq_lseek,
341 .release = seq_release,
342};
343#endif
344
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900345#ifdef CONFIG_IPV6_PIMSM_V2
346static int reg_vif_num = -1;
347
348static int pim6_rcv(struct sk_buff *skb)
349{
350 struct pimreghdr *pim;
351 struct ipv6hdr *encap;
352 struct net_device *reg_dev = NULL;
353
354 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
355 goto drop;
356
357 pim = (struct pimreghdr *)skb_transport_header(skb);
358 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
359 (pim->flags & PIM_NULL_REGISTER) ||
360 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
Al Viroec6b4862008-04-26 22:28:58 -0700361 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900362 goto drop;
363
364 /* check if the inner packet is destined to mcast group */
365 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
366 sizeof(*pim));
367
368 if (!ipv6_addr_is_multicast(&encap->daddr) ||
369 encap->payload_len == 0 ||
370 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
371 goto drop;
372
373 read_lock(&mrt_lock);
374 if (reg_vif_num >= 0)
375 reg_dev = vif6_table[reg_vif_num].dev;
376 if (reg_dev)
377 dev_hold(reg_dev);
378 read_unlock(&mrt_lock);
379
380 if (reg_dev == NULL)
381 goto drop;
382
383 skb->mac_header = skb->network_header;
384 skb_pull(skb, (u8 *)encap - skb->data);
385 skb_reset_network_header(skb);
386 skb->dev = reg_dev;
387 skb->protocol = htons(ETH_P_IP);
388 skb->ip_summed = 0;
389 skb->pkt_type = PACKET_HOST;
390 dst_release(skb->dst);
391 ((struct net_device_stats *)netdev_priv(reg_dev))->rx_bytes += skb->len;
392 ((struct net_device_stats *)netdev_priv(reg_dev))->rx_packets++;
393 skb->dst = NULL;
394 nf_reset(skb);
395 netif_rx(skb);
396 dev_put(reg_dev);
397 return 0;
398 drop:
399 kfree_skb(skb);
400 return 0;
401}
402
403static struct inet6_protocol pim6_protocol = {
404 .handler = pim6_rcv,
405};
406
407/* Service routines creating virtual interfaces: PIMREG */
408
409static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
410{
411 read_lock(&mrt_lock);
412 ((struct net_device_stats *)netdev_priv(dev))->tx_bytes += skb->len;
413 ((struct net_device_stats *)netdev_priv(dev))->tx_packets++;
414 ip6mr_cache_report(skb, reg_vif_num, MRT6MSG_WHOLEPKT);
415 read_unlock(&mrt_lock);
416 kfree_skb(skb);
417 return 0;
418}
419
420static struct net_device_stats *reg_vif_get_stats(struct net_device *dev)
421{
422 return (struct net_device_stats *)netdev_priv(dev);
423}
424
425static void reg_vif_setup(struct net_device *dev)
426{
427 dev->type = ARPHRD_PIMREG;
428 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
429 dev->flags = IFF_NOARP;
430 dev->hard_start_xmit = reg_vif_xmit;
431 dev->get_stats = reg_vif_get_stats;
432 dev->destructor = free_netdev;
433}
434
435static struct net_device *ip6mr_reg_vif(void)
436{
437 struct net_device *dev;
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900438
439 dev = alloc_netdev(sizeof(struct net_device_stats), "pim6reg",
440 reg_vif_setup);
441
442 if (dev == NULL)
443 return NULL;
444
445 if (register_netdevice(dev)) {
446 free_netdev(dev);
447 return NULL;
448 }
449 dev->iflink = 0;
450
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900451 if (dev_open(dev))
452 goto failure;
453
454 return dev;
455
456failure:
457 /* allow the register to be completed before unregistering. */
458 rtnl_unlock();
459 rtnl_lock();
460
461 unregister_netdevice(dev);
462 return NULL;
463}
464#endif
465
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900466/*
467 * Delete a VIF entry
468 */
469
470static int mif6_delete(int vifi)
471{
472 struct mif_device *v;
473 struct net_device *dev;
474 if (vifi < 0 || vifi >= maxvif)
475 return -EADDRNOTAVAIL;
476
477 v = &vif6_table[vifi];
478
479 write_lock_bh(&mrt_lock);
480 dev = v->dev;
481 v->dev = NULL;
482
483 if (!dev) {
484 write_unlock_bh(&mrt_lock);
485 return -EADDRNOTAVAIL;
486 }
487
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900488#ifdef CONFIG_IPV6_PIMSM_V2
489 if (vifi == reg_vif_num)
490 reg_vif_num = -1;
491#endif
492
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900493 if (vifi + 1 == maxvif) {
494 int tmp;
495 for (tmp = vifi - 1; tmp >= 0; tmp--) {
496 if (MIF_EXISTS(tmp))
497 break;
498 }
499 maxvif = tmp + 1;
500 }
501
502 write_unlock_bh(&mrt_lock);
503
504 dev_set_allmulti(dev, -1);
505
506 if (v->flags & MIFF_REGISTER)
507 unregister_netdevice(dev);
508
509 dev_put(dev);
510 return 0;
511}
512
513/* Destroy an unresolved cache entry, killing queued skbs
514 and reporting error to netlink readers.
515 */
516
517static void ip6mr_destroy_unres(struct mfc6_cache *c)
518{
519 struct sk_buff *skb;
520
521 atomic_dec(&cache_resolve_queue_len);
522
523 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
524 if (ipv6_hdr(skb)->version == 0) {
525 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
526 nlh->nlmsg_type = NLMSG_ERROR;
527 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
528 skb_trim(skb, nlh->nlmsg_len);
529 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT;
530 rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid);
531 } else
532 kfree_skb(skb);
533 }
534
535 kmem_cache_free(mrt_cachep, c);
536}
537
538
539/* Single timer process for all the unresolved queue. */
540
541static void ipmr_do_expire_process(unsigned long dummy)
542{
543 unsigned long now = jiffies;
544 unsigned long expires = 10 * HZ;
545 struct mfc6_cache *c, **cp;
546
547 cp = &mfc_unres_queue;
548
549 while ((c = *cp) != NULL) {
550 if (time_after(c->mfc_un.unres.expires, now)) {
551 /* not yet... */
552 unsigned long interval = c->mfc_un.unres.expires - now;
553 if (interval < expires)
554 expires = interval;
555 cp = &c->next;
556 continue;
557 }
558
559 *cp = c->next;
560 ip6mr_destroy_unres(c);
561 }
562
563 if (atomic_read(&cache_resolve_queue_len))
564 mod_timer(&ipmr_expire_timer, jiffies + expires);
565}
566
567static void ipmr_expire_process(unsigned long dummy)
568{
569 if (!spin_trylock(&mfc_unres_lock)) {
570 mod_timer(&ipmr_expire_timer, jiffies + 1);
571 return;
572 }
573
574 if (atomic_read(&cache_resolve_queue_len))
575 ipmr_do_expire_process(dummy);
576
577 spin_unlock(&mfc_unres_lock);
578}
579
580/* Fill oifs list. It is called under write locked mrt_lock. */
581
582static void ip6mr_update_thresholds(struct mfc6_cache *cache, unsigned char *ttls)
583{
584 int vifi;
585
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300586 cache->mfc_un.res.minvif = MAXMIFS;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900587 cache->mfc_un.res.maxvif = 0;
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300588 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900589
590 for (vifi = 0; vifi < maxvif; vifi++) {
591 if (MIF_EXISTS(vifi) && ttls[vifi] && ttls[vifi] < 255) {
592 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
593 if (cache->mfc_un.res.minvif > vifi)
594 cache->mfc_un.res.minvif = vifi;
595 if (cache->mfc_un.res.maxvif <= vifi)
596 cache->mfc_un.res.maxvif = vifi + 1;
597 }
598 }
599}
600
601static int mif6_add(struct mif6ctl *vifc, int mrtsock)
602{
603 int vifi = vifc->mif6c_mifi;
604 struct mif_device *v = &vif6_table[vifi];
605 struct net_device *dev;
606
607 /* Is vif busy ? */
608 if (MIF_EXISTS(vifi))
609 return -EADDRINUSE;
610
611 switch (vifc->mif6c_flags) {
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900612#ifdef CONFIG_IPV6_PIMSM_V2
613 case MIFF_REGISTER:
614 /*
615 * Special Purpose VIF in PIM
616 * All the packets will be sent to the daemon
617 */
618 if (reg_vif_num >= 0)
619 return -EADDRINUSE;
620 dev = ip6mr_reg_vif();
621 if (!dev)
622 return -ENOBUFS;
623 break;
624#endif
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900625 case 0:
626 dev = dev_get_by_index(&init_net, vifc->mif6c_pifi);
627 if (!dev)
628 return -EADDRNOTAVAIL;
629 dev_put(dev);
630 break;
631 default:
632 return -EINVAL;
633 }
634
635 dev_set_allmulti(dev, 1);
636
637 /*
638 * Fill in the VIF structures
639 */
640 v->rate_limit = vifc->vifc_rate_limit;
641 v->flags = vifc->mif6c_flags;
642 if (!mrtsock)
643 v->flags |= VIFF_STATIC;
644 v->threshold = vifc->vifc_threshold;
645 v->bytes_in = 0;
646 v->bytes_out = 0;
647 v->pkt_in = 0;
648 v->pkt_out = 0;
649 v->link = dev->ifindex;
650 if (v->flags & MIFF_REGISTER)
651 v->link = dev->iflink;
652
653 /* And finish update writing critical data */
654 write_lock_bh(&mrt_lock);
655 dev_hold(dev);
656 v->dev = dev;
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900657#ifdef CONFIG_IPV6_PIMSM_V2
658 if (v->flags & MIFF_REGISTER)
659 reg_vif_num = vifi;
660#endif
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900661 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));
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300689 c->mfc_un.res.minvif = MAXMIFS;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900690 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) {
YOSHIFUJI Hideaki549e0282008-04-05 22:17:39 +0900722 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900723 } 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
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300742static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert)
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900743{
744 struct sk_buff *skb;
745 struct mrt6msg *msg;
746 int ret;
747
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900748#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);
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900755
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
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900764#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 {
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900787 /*
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;
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300804 msg->im6_mif = mifi;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900805 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));
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900813 }
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900814
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
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300837ip6mr_cache_unresolved(mifi_t mifi, struct sk_buff *skb)
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900838{
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 */
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300872 if ((err = ip6mr_cache_report(skb, mifi, MRT6MSG_NOCACHE)) < 0) {
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900873 /* 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
937 if (dev_net(dev) != &init_net)
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
959void __init ip6_mr_init(void)
960{
961 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
962 sizeof(struct mfc6_cache),
963 0, SLAB_HWCACHE_ALIGN,
964 NULL);
965 if (!mrt_cachep)
966 panic("cannot allocate ip6_mrt_cache");
967
968 setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0);
969 register_netdevice_notifier(&ip6_mr_notifier);
970#ifdef CONFIG_PROC_FS
971 proc_net_fops_create(&init_net, "ip6_mr_vif", 0, &ip6mr_vif_fops);
972 proc_net_fops_create(&init_net, "ip6_mr_cache", 0, &ip6mr_mfc_fops);
973#endif
974}
975
976
977static int ip6mr_mfc_add(struct mf6cctl *mfc, int mrtsock)
978{
979 int line;
980 struct mfc6_cache *uc, *c, **cp;
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300981 unsigned char ttls[MAXMIFS];
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900982 int i;
983
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300984 memset(ttls, 255, MAXMIFS);
985 for (i = 0; i < MAXMIFS; i++) {
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900986 if (IF_ISSET(i, &mfc->mf6cc_ifset))
987 ttls[i] = 1;
988
989 }
990
991 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
992
993 for (cp = &mfc6_cache_array[line]; (c = *cp) != NULL; cp = &c->next) {
994 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
995 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr))
996 break;
997 }
998
999 if (c != NULL) {
1000 write_lock_bh(&mrt_lock);
1001 c->mf6c_parent = mfc->mf6cc_parent;
1002 ip6mr_update_thresholds(c, ttls);
1003 if (!mrtsock)
1004 c->mfc_flags |= MFC_STATIC;
1005 write_unlock_bh(&mrt_lock);
1006 return 0;
1007 }
1008
1009 if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1010 return -EINVAL;
1011
1012 c = ip6mr_cache_alloc();
1013 if (c == NULL)
1014 return -ENOMEM;
1015
1016 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1017 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1018 c->mf6c_parent = mfc->mf6cc_parent;
1019 ip6mr_update_thresholds(c, ttls);
1020 if (!mrtsock)
1021 c->mfc_flags |= MFC_STATIC;
1022
1023 write_lock_bh(&mrt_lock);
1024 c->next = mfc6_cache_array[line];
1025 mfc6_cache_array[line] = c;
1026 write_unlock_bh(&mrt_lock);
1027
1028 /*
1029 * Check to see if we resolved a queued list. If so we
1030 * need to send on the frames and tidy up.
1031 */
1032 spin_lock_bh(&mfc_unres_lock);
1033 for (cp = &mfc_unres_queue; (uc = *cp) != NULL;
1034 cp = &uc->next) {
1035 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1036 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1037 *cp = uc->next;
1038 if (atomic_dec_and_test(&cache_resolve_queue_len))
1039 del_timer(&ipmr_expire_timer);
1040 break;
1041 }
1042 }
1043 spin_unlock_bh(&mfc_unres_lock);
1044
1045 if (uc) {
1046 ip6mr_cache_resolve(uc, c);
1047 kmem_cache_free(mrt_cachep, uc);
1048 }
1049 return 0;
1050}
1051
1052/*
1053 * Close the multicast socket, and clear the vif tables etc
1054 */
1055
1056static void mroute_clean_tables(struct sock *sk)
1057{
1058 int i;
1059
1060 /*
1061 * Shut down all active vif entries
1062 */
1063 for (i = 0; i < maxvif; i++) {
1064 if (!(vif6_table[i].flags & VIFF_STATIC))
1065 mif6_delete(i);
1066 }
1067
1068 /*
1069 * Wipe the cache
1070 */
1071 for (i = 0; i < ARRAY_SIZE(mfc6_cache_array); i++) {
1072 struct mfc6_cache *c, **cp;
1073
1074 cp = &mfc6_cache_array[i];
1075 while ((c = *cp) != NULL) {
1076 if (c->mfc_flags & MFC_STATIC) {
1077 cp = &c->next;
1078 continue;
1079 }
1080 write_lock_bh(&mrt_lock);
1081 *cp = c->next;
1082 write_unlock_bh(&mrt_lock);
1083
1084 kmem_cache_free(mrt_cachep, c);
1085 }
1086 }
1087
1088 if (atomic_read(&cache_resolve_queue_len) != 0) {
1089 struct mfc6_cache *c;
1090
1091 spin_lock_bh(&mfc_unres_lock);
1092 while (mfc_unres_queue != NULL) {
1093 c = mfc_unres_queue;
1094 mfc_unres_queue = c->next;
1095 spin_unlock_bh(&mfc_unres_lock);
1096
1097 ip6mr_destroy_unres(c);
1098
1099 spin_lock_bh(&mfc_unres_lock);
1100 }
1101 spin_unlock_bh(&mfc_unres_lock);
1102 }
1103}
1104
1105static int ip6mr_sk_init(struct sock *sk)
1106{
1107 int err = 0;
1108
1109 rtnl_lock();
1110 write_lock_bh(&mrt_lock);
1111 if (likely(mroute6_socket == NULL))
1112 mroute6_socket = sk;
1113 else
1114 err = -EADDRINUSE;
1115 write_unlock_bh(&mrt_lock);
1116
1117 rtnl_unlock();
1118
1119 return err;
1120}
1121
1122int ip6mr_sk_done(struct sock *sk)
1123{
1124 int err = 0;
1125
1126 rtnl_lock();
1127 if (sk == mroute6_socket) {
1128 write_lock_bh(&mrt_lock);
1129 mroute6_socket = NULL;
1130 write_unlock_bh(&mrt_lock);
1131
1132 mroute_clean_tables(sk);
1133 } else
1134 err = -EACCES;
1135 rtnl_unlock();
1136
1137 return err;
1138}
1139
1140/*
1141 * Socket options and virtual interface manipulation. The whole
1142 * virtual interface system is a complete heap, but unfortunately
1143 * that's how BSD mrouted happens to think. Maybe one day with a proper
1144 * MOSPF/PIM router set up we can clean this up.
1145 */
1146
1147int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, int optlen)
1148{
1149 int ret;
1150 struct mif6ctl vif;
1151 struct mf6cctl mfc;
1152 mifi_t mifi;
1153
1154 if (optname != MRT6_INIT) {
1155 if (sk != mroute6_socket && !capable(CAP_NET_ADMIN))
1156 return -EACCES;
1157 }
1158
1159 switch (optname) {
1160 case MRT6_INIT:
1161 if (sk->sk_type != SOCK_RAW ||
1162 inet_sk(sk)->num != IPPROTO_ICMPV6)
1163 return -EOPNOTSUPP;
1164 if (optlen < sizeof(int))
1165 return -EINVAL;
1166
1167 return ip6mr_sk_init(sk);
1168
1169 case MRT6_DONE:
1170 return ip6mr_sk_done(sk);
1171
1172 case MRT6_ADD_MIF:
1173 if (optlen < sizeof(vif))
1174 return -EINVAL;
1175 if (copy_from_user(&vif, optval, sizeof(vif)))
1176 return -EFAULT;
Rami Rosen6ac7eb02008-04-10 12:40:10 +03001177 if (vif.mif6c_mifi >= MAXMIFS)
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001178 return -ENFILE;
1179 rtnl_lock();
1180 ret = mif6_add(&vif, sk == mroute6_socket);
1181 rtnl_unlock();
1182 return ret;
1183
1184 case MRT6_DEL_MIF:
1185 if (optlen < sizeof(mifi_t))
1186 return -EINVAL;
1187 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1188 return -EFAULT;
1189 rtnl_lock();
1190 ret = mif6_delete(mifi);
1191 rtnl_unlock();
1192 return ret;
1193
1194 /*
1195 * Manipulate the forwarding caches. These live
1196 * in a sort of kernel/user symbiosis.
1197 */
1198 case MRT6_ADD_MFC:
1199 case MRT6_DEL_MFC:
1200 if (optlen < sizeof(mfc))
1201 return -EINVAL;
1202 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1203 return -EFAULT;
1204 rtnl_lock();
1205 if (optname == MRT6_DEL_MFC)
1206 ret = ip6mr_mfc_delete(&mfc);
1207 else
1208 ret = ip6mr_mfc_add(&mfc, sk == mroute6_socket);
1209 rtnl_unlock();
1210 return ret;
1211
1212 /*
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001213 * Control PIM assert (to activate pim will activate assert)
1214 */
1215 case MRT6_ASSERT:
1216 {
1217 int v;
1218 if (get_user(v, (int __user *)optval))
1219 return -EFAULT;
1220 mroute_do_assert = !!v;
1221 return 0;
1222 }
1223
1224#ifdef CONFIG_IPV6_PIMSM_V2
1225 case MRT6_PIM:
1226 {
YOSHIFUJI Hideakia9f83bf2008-04-10 15:41:28 +09001227 int v;
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001228 if (get_user(v, (int __user *)optval))
1229 return -EFAULT;
1230 v = !!v;
1231 rtnl_lock();
1232 ret = 0;
1233 if (v != mroute_do_pim) {
1234 mroute_do_pim = v;
1235 mroute_do_assert = v;
1236 if (mroute_do_pim)
1237 ret = inet6_add_protocol(&pim6_protocol,
1238 IPPROTO_PIM);
1239 else
1240 ret = inet6_del_protocol(&pim6_protocol,
1241 IPPROTO_PIM);
1242 if (ret < 0)
1243 ret = -EAGAIN;
1244 }
1245 rtnl_unlock();
1246 return ret;
1247 }
1248
1249#endif
1250 /*
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001251 * Spurious command, or MRT_VERSION which you cannot
1252 * set.
1253 */
1254 default:
1255 return -ENOPROTOOPT;
1256 }
1257}
1258
1259/*
1260 * Getsock opt support for the multicast routing system.
1261 */
1262
1263int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1264 int __user *optlen)
1265{
1266 int olr;
1267 int val;
1268
1269 switch (optname) {
1270 case MRT6_VERSION:
1271 val = 0x0305;
1272 break;
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001273#ifdef CONFIG_IPV6_PIMSM_V2
1274 case MRT6_PIM:
1275 val = mroute_do_pim;
1276 break;
1277#endif
1278 case MRT6_ASSERT:
1279 val = mroute_do_assert;
1280 break;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001281 default:
1282 return -ENOPROTOOPT;
1283 }
1284
1285 if (get_user(olr, optlen))
1286 return -EFAULT;
1287
1288 olr = min_t(int, olr, sizeof(int));
1289 if (olr < 0)
1290 return -EINVAL;
1291
1292 if (put_user(olr, optlen))
1293 return -EFAULT;
1294 if (copy_to_user(optval, &val, olr))
1295 return -EFAULT;
1296 return 0;
1297}
1298
1299/*
1300 * The IP multicast ioctl support routines.
1301 */
1302
1303int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1304{
1305 struct sioc_sg_req6 sr;
1306 struct sioc_mif_req6 vr;
1307 struct mif_device *vif;
1308 struct mfc6_cache *c;
1309
1310 switch (cmd) {
1311 case SIOCGETMIFCNT_IN6:
1312 if (copy_from_user(&vr, arg, sizeof(vr)))
1313 return -EFAULT;
1314 if (vr.mifi >= maxvif)
1315 return -EINVAL;
1316 read_lock(&mrt_lock);
1317 vif = &vif6_table[vr.mifi];
1318 if (MIF_EXISTS(vr.mifi)) {
1319 vr.icount = vif->pkt_in;
1320 vr.ocount = vif->pkt_out;
1321 vr.ibytes = vif->bytes_in;
1322 vr.obytes = vif->bytes_out;
1323 read_unlock(&mrt_lock);
1324
1325 if (copy_to_user(arg, &vr, sizeof(vr)))
1326 return -EFAULT;
1327 return 0;
1328 }
1329 read_unlock(&mrt_lock);
1330 return -EADDRNOTAVAIL;
1331 case SIOCGETSGCNT_IN6:
1332 if (copy_from_user(&sr, arg, sizeof(sr)))
1333 return -EFAULT;
1334
1335 read_lock(&mrt_lock);
1336 c = ip6mr_cache_find(&sr.src.sin6_addr, &sr.grp.sin6_addr);
1337 if (c) {
1338 sr.pktcnt = c->mfc_un.res.pkt;
1339 sr.bytecnt = c->mfc_un.res.bytes;
1340 sr.wrong_if = c->mfc_un.res.wrong_if;
1341 read_unlock(&mrt_lock);
1342
1343 if (copy_to_user(arg, &sr, sizeof(sr)))
1344 return -EFAULT;
1345 return 0;
1346 }
1347 read_unlock(&mrt_lock);
1348 return -EADDRNOTAVAIL;
1349 default:
1350 return -ENOIOCTLCMD;
1351 }
1352}
1353
1354
1355static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1356{
Rami Rosen0912ea32008-04-13 23:59:13 -07001357 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001358 return dst_output(skb);
1359}
1360
1361/*
1362 * Processing handlers for ip6mr_forward
1363 */
1364
1365static int ip6mr_forward2(struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1366{
1367 struct ipv6hdr *ipv6h;
1368 struct mif_device *vif = &vif6_table[vifi];
1369 struct net_device *dev;
1370 struct dst_entry *dst;
1371 struct flowi fl;
1372
1373 if (vif->dev == NULL)
1374 goto out_free;
1375
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001376#ifdef CONFIG_IPV6_PIMSM_V2
1377 if (vif->flags & MIFF_REGISTER) {
1378 vif->pkt_out++;
1379 vif->bytes_out += skb->len;
1380 ((struct net_device_stats *)netdev_priv(vif->dev))->tx_bytes += skb->len;
1381 ((struct net_device_stats *)netdev_priv(vif->dev))->tx_packets++;
1382 ip6mr_cache_report(skb, vifi, MRT6MSG_WHOLEPKT);
1383 kfree_skb(skb);
1384 return 0;
1385 }
1386#endif
1387
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001388 ipv6h = ipv6_hdr(skb);
1389
1390 fl = (struct flowi) {
1391 .oif = vif->link,
1392 .nl_u = { .ip6_u =
1393 { .daddr = ipv6h->daddr, }
1394 }
1395 };
1396
1397 dst = ip6_route_output(&init_net, NULL, &fl);
1398 if (!dst)
1399 goto out_free;
1400
1401 dst_release(skb->dst);
1402 skb->dst = dst;
1403
1404 /*
1405 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1406 * not only before forwarding, but after forwarding on all output
1407 * interfaces. It is clear, if mrouter runs a multicasting
1408 * program, it should receive packets not depending to what interface
1409 * program is joined.
1410 * If we will not make it, the program will have to join on all
1411 * interfaces. On the other hand, multihoming host (or router, but
1412 * not mrouter) cannot join to more than one interface - it will
1413 * result in receiving multiple packets.
1414 */
1415 dev = vif->dev;
1416 skb->dev = dev;
1417 vif->pkt_out++;
1418 vif->bytes_out += skb->len;
1419
1420 /* We are about to write */
1421 /* XXX: extension headers? */
1422 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
1423 goto out_free;
1424
1425 ipv6h = ipv6_hdr(skb);
1426 ipv6h->hop_limit--;
1427
1428 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
1429
1430 return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dev,
1431 ip6mr_forward2_finish);
1432
1433out_free:
1434 kfree_skb(skb);
1435 return 0;
1436}
1437
1438static int ip6mr_find_vif(struct net_device *dev)
1439{
1440 int ct;
1441 for (ct = maxvif - 1; ct >= 0; ct--) {
1442 if (vif6_table[ct].dev == dev)
1443 break;
1444 }
1445 return ct;
1446}
1447
1448static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache)
1449{
1450 int psend = -1;
1451 int vif, ct;
1452
1453 vif = cache->mf6c_parent;
1454 cache->mfc_un.res.pkt++;
1455 cache->mfc_un.res.bytes += skb->len;
1456
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001457 /*
1458 * Wrong interface: drop packet and (maybe) send PIM assert.
1459 */
1460 if (vif6_table[vif].dev != skb->dev) {
1461 int true_vifi;
1462
1463 cache->mfc_un.res.wrong_if++;
1464 true_vifi = ip6mr_find_vif(skb->dev);
1465
1466 if (true_vifi >= 0 && mroute_do_assert &&
1467 /* pimsm uses asserts, when switching from RPT to SPT,
1468 so that we cannot check that packet arrived on an oif.
1469 It is bad, but otherwise we would need to move pretty
1470 large chunk of pimd to kernel. Ough... --ANK
1471 */
1472 (mroute_do_pim || cache->mfc_un.res.ttls[true_vifi] < 255) &&
1473 time_after(jiffies,
1474 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1475 cache->mfc_un.res.last_assert = jiffies;
1476 ip6mr_cache_report(skb, true_vifi, MRT6MSG_WRONGMIF);
1477 }
1478 goto dont_forward;
1479 }
1480
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001481 vif6_table[vif].pkt_in++;
1482 vif6_table[vif].bytes_in += skb->len;
1483
1484 /*
1485 * Forward the frame
1486 */
1487 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
1488 if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
1489 if (psend != -1) {
1490 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1491 if (skb2)
1492 ip6mr_forward2(skb2, cache, psend);
1493 }
1494 psend = ct;
1495 }
1496 }
1497 if (psend != -1) {
1498 ip6mr_forward2(skb, cache, psend);
1499 return 0;
1500 }
1501
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001502dont_forward:
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001503 kfree_skb(skb);
1504 return 0;
1505}
1506
1507
1508/*
1509 * Multicast packets for forwarding arrive here
1510 */
1511
1512int ip6_mr_input(struct sk_buff *skb)
1513{
1514 struct mfc6_cache *cache;
1515
1516 read_lock(&mrt_lock);
1517 cache = ip6mr_cache_find(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
1518
1519 /*
1520 * No usable cache entry
1521 */
1522 if (cache == NULL) {
1523 int vif;
1524
1525 vif = ip6mr_find_vif(skb->dev);
1526 if (vif >= 0) {
1527 int err = ip6mr_cache_unresolved(vif, skb);
1528 read_unlock(&mrt_lock);
1529
1530 return err;
1531 }
1532 read_unlock(&mrt_lock);
1533 kfree_skb(skb);
1534 return -ENODEV;
1535 }
1536
1537 ip6_mr_forward(skb, cache);
1538
1539 read_unlock(&mrt_lock);
1540
1541 return 0;
1542}
1543
1544
1545static int
1546ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm)
1547{
1548 int ct;
1549 struct rtnexthop *nhp;
1550 struct net_device *dev = vif6_table[c->mf6c_parent].dev;
YOSHIFUJI Hideaki549e0282008-04-05 22:17:39 +09001551 u8 *b = skb_tail_pointer(skb);
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001552 struct rtattr *mp_head;
1553
1554 if (dev)
1555 RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex);
1556
1557 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
1558
1559 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
1560 if (c->mfc_un.res.ttls[ct] < 255) {
1561 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
1562 goto rtattr_failure;
1563 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
1564 nhp->rtnh_flags = 0;
1565 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
1566 nhp->rtnh_ifindex = vif6_table[ct].dev->ifindex;
1567 nhp->rtnh_len = sizeof(*nhp);
1568 }
1569 }
1570 mp_head->rta_type = RTA_MULTIPATH;
YOSHIFUJI Hideaki549e0282008-04-05 22:17:39 +09001571 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001572 rtm->rtm_type = RTN_MULTICAST;
1573 return 1;
1574
1575rtattr_failure:
1576 nlmsg_trim(skb, b);
1577 return -EMSGSIZE;
1578}
1579
1580int ip6mr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait)
1581{
1582 int err;
1583 struct mfc6_cache *cache;
1584 struct rt6_info *rt = (struct rt6_info *)skb->dst;
1585
1586 read_lock(&mrt_lock);
1587 cache = ip6mr_cache_find(&rt->rt6i_src.addr, &rt->rt6i_dst.addr);
1588
1589 if (!cache) {
1590 struct sk_buff *skb2;
1591 struct ipv6hdr *iph;
1592 struct net_device *dev;
1593 int vif;
1594
1595 if (nowait) {
1596 read_unlock(&mrt_lock);
1597 return -EAGAIN;
1598 }
1599
1600 dev = skb->dev;
1601 if (dev == NULL || (vif = ip6mr_find_vif(dev)) < 0) {
1602 read_unlock(&mrt_lock);
1603 return -ENODEV;
1604 }
1605
1606 /* really correct? */
1607 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
1608 if (!skb2) {
1609 read_unlock(&mrt_lock);
1610 return -ENOMEM;
1611 }
1612
1613 skb_reset_transport_header(skb2);
1614
1615 skb_put(skb2, sizeof(struct ipv6hdr));
1616 skb_reset_network_header(skb2);
1617
1618 iph = ipv6_hdr(skb2);
1619 iph->version = 0;
1620 iph->priority = 0;
1621 iph->flow_lbl[0] = 0;
1622 iph->flow_lbl[1] = 0;
1623 iph->flow_lbl[2] = 0;
1624 iph->payload_len = 0;
1625 iph->nexthdr = IPPROTO_NONE;
1626 iph->hop_limit = 0;
1627 ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr);
1628 ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr);
1629
1630 err = ip6mr_cache_unresolved(vif, skb2);
1631 read_unlock(&mrt_lock);
1632
1633 return err;
1634 }
1635
1636 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
1637 cache->mfc_flags |= MFC_NOTIFY;
1638
1639 err = ip6mr_fill_mroute(skb, cache, rtm);
1640 read_unlock(&mrt_lock);
1641 return err;
1642}
1643