]>
Commit | Line | Data |
---|---|---|
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> | |
30 | #include <linux/in.h> | |
31 | #include <linux/inet.h> | |
32 | #include <linux/netdevice.h> | |
33 | #include <linux/inetdevice.h> | |
34 | #include <linux/igmp.h> | |
35 | #include <linux/proc_fs.h> | |
36 | #include <linux/seq_file.h> | |
37 | #include <linux/mroute.h> | |
38 | #include <linux/init.h> | |
39 | #include <net/ip.h> | |
40 | #include <net/protocol.h> | |
41 | #include <linux/skbuff.h> | |
42 | #include <net/sock.h> | |
43 | #include <net/icmp.h> | |
44 | #include <net/udp.h> | |
45 | #include <net/raw.h> | |
46 | #include <net/route.h> | |
47 | #include <linux/notifier.h> | |
48 | #include <linux/if_arp.h> | |
49 | #include <linux/netfilter_ipv4.h> | |
50 | #include <net/ipip.h> | |
51 | #include <net/checksum.h> | |
52 | #include <net/netlink.h> | |
53 | ||
54 | #include <net/ipv6.h> | |
55 | #include <net/ip6_route.h> | |
56 | #include <linux/mroute6.h> | |
14fb64e1 | 57 | #include <linux/pim.h> |
7bc570c8 YH |
58 | #include <net/addrconf.h> |
59 | #include <linux/netfilter_ipv6.h> | |
60 | ||
61 | struct sock *mroute6_socket; | |
62 | ||
63 | ||
64 | /* Big lock, protecting vif table, mrt cache and mroute socket state. | |
65 | Note that the changes are semaphored via rtnl_lock. | |
66 | */ | |
67 | ||
68 | static DEFINE_RWLOCK(mrt_lock); | |
69 | ||
70 | /* | |
71 | * Multicast router control variables | |
72 | */ | |
73 | ||
74 | static struct mif_device vif6_table[MAXMIFS]; /* Devices */ | |
75 | static int maxvif; | |
76 | ||
77 | #define MIF_EXISTS(idx) (vif6_table[idx].dev != NULL) | |
78 | ||
14fb64e1 YH |
79 | static int mroute_do_assert; /* Set in PIM assert */ |
80 | #ifdef CONFIG_IPV6_PIMSM_V2 | |
81 | static int mroute_do_pim; | |
82 | #else | |
83 | #define mroute_do_pim 0 | |
84 | #endif | |
85 | ||
7bc570c8 YH |
86 | static struct mfc6_cache *mfc6_cache_array[MFC_LINES]; /* Forwarding cache */ |
87 | ||
88 | static struct mfc6_cache *mfc_unres_queue; /* Queue of unresolved entries */ | |
89 | static atomic_t cache_resolve_queue_len; /* Size of unresolved */ | |
90 | ||
91 | /* Special spinlock for queue of unresolved entries */ | |
92 | static DEFINE_SPINLOCK(mfc_unres_lock); | |
93 | ||
94 | /* We return to original Alan's scheme. Hash table of resolved | |
95 | entries is changed only in process context and protected | |
96 | with weak lock mrt_lock. Queue of unresolved entries is protected | |
97 | with strong spinlock mfc_unres_lock. | |
98 | ||
99 | In this case data path is free of exclusive locks at all. | |
100 | */ | |
101 | ||
102 | static struct kmem_cache *mrt_cachep __read_mostly; | |
103 | ||
104 | static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache); | |
105 | static int ip6mr_cache_report(struct sk_buff *pkt, vifi_t vifi, int assert); | |
106 | static int ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm); | |
107 | ||
14fb64e1 YH |
108 | #ifdef CONFIG_IPV6_PIMSM_V2 |
109 | static struct inet6_protocol pim6_protocol; | |
110 | #endif | |
111 | ||
7bc570c8 YH |
112 | static struct timer_list ipmr_expire_timer; |
113 | ||
114 | ||
115 | #ifdef CONFIG_PROC_FS | |
116 | ||
117 | struct ipmr_mfc_iter { | |
118 | struct mfc6_cache **cache; | |
119 | int ct; | |
120 | }; | |
121 | ||
122 | ||
123 | static struct mfc6_cache *ipmr_mfc_seq_idx(struct ipmr_mfc_iter *it, loff_t pos) | |
124 | { | |
125 | struct mfc6_cache *mfc; | |
126 | ||
127 | it->cache = mfc6_cache_array; | |
128 | read_lock(&mrt_lock); | |
129 | for (it->ct = 0; it->ct < ARRAY_SIZE(mfc6_cache_array); it->ct++) | |
130 | for (mfc = mfc6_cache_array[it->ct]; mfc; mfc = mfc->next) | |
131 | if (pos-- == 0) | |
132 | return mfc; | |
133 | read_unlock(&mrt_lock); | |
134 | ||
135 | it->cache = &mfc_unres_queue; | |
136 | spin_lock_bh(&mfc_unres_lock); | |
137 | for (mfc = mfc_unres_queue; mfc; mfc = mfc->next) | |
138 | if (pos-- == 0) | |
139 | return mfc; | |
140 | spin_unlock_bh(&mfc_unres_lock); | |
141 | ||
142 | it->cache = NULL; | |
143 | return NULL; | |
144 | } | |
145 | ||
146 | ||
147 | ||
148 | ||
149 | /* | |
150 | * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif | |
151 | */ | |
152 | ||
153 | struct ipmr_vif_iter { | |
154 | int ct; | |
155 | }; | |
156 | ||
157 | static struct mif_device *ip6mr_vif_seq_idx(struct ipmr_vif_iter *iter, | |
158 | loff_t pos) | |
159 | { | |
160 | for (iter->ct = 0; iter->ct < maxvif; ++iter->ct) { | |
161 | if (!MIF_EXISTS(iter->ct)) | |
162 | continue; | |
163 | if (pos-- == 0) | |
164 | return &vif6_table[iter->ct]; | |
165 | } | |
166 | return NULL; | |
167 | } | |
168 | ||
169 | static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos) | |
170 | __acquires(mrt_lock) | |
171 | { | |
172 | read_lock(&mrt_lock); | |
173 | return (*pos ? ip6mr_vif_seq_idx(seq->private, *pos - 1) | |
174 | : SEQ_START_TOKEN); | |
175 | } | |
176 | ||
177 | static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
178 | { | |
179 | struct ipmr_vif_iter *iter = seq->private; | |
180 | ||
181 | ++*pos; | |
182 | if (v == SEQ_START_TOKEN) | |
183 | return ip6mr_vif_seq_idx(iter, 0); | |
184 | ||
185 | while (++iter->ct < maxvif) { | |
186 | if (!MIF_EXISTS(iter->ct)) | |
187 | continue; | |
188 | return &vif6_table[iter->ct]; | |
189 | } | |
190 | return NULL; | |
191 | } | |
192 | ||
193 | static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v) | |
194 | __releases(mrt_lock) | |
195 | { | |
196 | read_unlock(&mrt_lock); | |
197 | } | |
198 | ||
199 | static int ip6mr_vif_seq_show(struct seq_file *seq, void *v) | |
200 | { | |
201 | if (v == SEQ_START_TOKEN) { | |
202 | seq_puts(seq, | |
203 | "Interface BytesIn PktsIn BytesOut PktsOut Flags\n"); | |
204 | } else { | |
205 | const struct mif_device *vif = v; | |
206 | const char *name = vif->dev ? vif->dev->name : "none"; | |
207 | ||
208 | seq_printf(seq, | |
209 | "%2Zd %-10s %8ld %7ld %8ld %7ld %05X\n", | |
210 | vif - vif6_table, | |
211 | name, vif->bytes_in, vif->pkt_in, | |
212 | vif->bytes_out, vif->pkt_out, | |
213 | vif->flags); | |
214 | } | |
215 | return 0; | |
216 | } | |
217 | ||
218 | static struct seq_operations ip6mr_vif_seq_ops = { | |
219 | .start = ip6mr_vif_seq_start, | |
220 | .next = ip6mr_vif_seq_next, | |
221 | .stop = ip6mr_vif_seq_stop, | |
222 | .show = ip6mr_vif_seq_show, | |
223 | }; | |
224 | ||
225 | static int ip6mr_vif_open(struct inode *inode, struct file *file) | |
226 | { | |
227 | return seq_open_private(file, &ip6mr_vif_seq_ops, | |
228 | sizeof(struct ipmr_vif_iter)); | |
229 | } | |
230 | ||
231 | static struct file_operations ip6mr_vif_fops = { | |
232 | .owner = THIS_MODULE, | |
233 | .open = ip6mr_vif_open, | |
234 | .read = seq_read, | |
235 | .llseek = seq_lseek, | |
236 | .release = seq_release, | |
237 | }; | |
238 | ||
239 | static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) | |
240 | { | |
241 | return (*pos ? ipmr_mfc_seq_idx(seq->private, *pos - 1) | |
242 | : SEQ_START_TOKEN); | |
243 | } | |
244 | ||
245 | static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
246 | { | |
247 | struct mfc6_cache *mfc = v; | |
248 | struct ipmr_mfc_iter *it = seq->private; | |
249 | ||
250 | ++*pos; | |
251 | ||
252 | if (v == SEQ_START_TOKEN) | |
253 | return ipmr_mfc_seq_idx(seq->private, 0); | |
254 | ||
255 | if (mfc->next) | |
256 | return mfc->next; | |
257 | ||
258 | if (it->cache == &mfc_unres_queue) | |
259 | goto end_of_list; | |
260 | ||
261 | BUG_ON(it->cache != mfc6_cache_array); | |
262 | ||
263 | while (++it->ct < ARRAY_SIZE(mfc6_cache_array)) { | |
264 | mfc = mfc6_cache_array[it->ct]; | |
265 | if (mfc) | |
266 | return mfc; | |
267 | } | |
268 | ||
269 | /* exhausted cache_array, show unresolved */ | |
270 | read_unlock(&mrt_lock); | |
271 | it->cache = &mfc_unres_queue; | |
272 | it->ct = 0; | |
273 | ||
274 | spin_lock_bh(&mfc_unres_lock); | |
275 | mfc = mfc_unres_queue; | |
276 | if (mfc) | |
277 | return mfc; | |
278 | ||
279 | end_of_list: | |
280 | spin_unlock_bh(&mfc_unres_lock); | |
281 | it->cache = NULL; | |
282 | ||
283 | return NULL; | |
284 | } | |
285 | ||
286 | static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) | |
287 | { | |
288 | struct ipmr_mfc_iter *it = seq->private; | |
289 | ||
290 | if (it->cache == &mfc_unres_queue) | |
291 | spin_unlock_bh(&mfc_unres_lock); | |
292 | else if (it->cache == mfc6_cache_array) | |
293 | read_unlock(&mrt_lock); | |
294 | } | |
295 | ||
296 | static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) | |
297 | { | |
298 | int n; | |
299 | ||
300 | if (v == SEQ_START_TOKEN) { | |
301 | seq_puts(seq, | |
302 | "Group " | |
303 | "Origin " | |
304 | "Iif Pkts Bytes Wrong Oifs\n"); | |
305 | } else { | |
306 | const struct mfc6_cache *mfc = v; | |
307 | const struct ipmr_mfc_iter *it = seq->private; | |
308 | ||
309 | seq_printf(seq, | |
310 | NIP6_FMT " " NIP6_FMT " %-3d %8ld %8ld %8ld", | |
311 | NIP6(mfc->mf6c_mcastgrp), NIP6(mfc->mf6c_origin), | |
312 | mfc->mf6c_parent, | |
313 | mfc->mfc_un.res.pkt, | |
314 | mfc->mfc_un.res.bytes, | |
315 | mfc->mfc_un.res.wrong_if); | |
316 | ||
317 | if (it->cache != &mfc_unres_queue) { | |
318 | for (n = mfc->mfc_un.res.minvif; | |
319 | n < mfc->mfc_un.res.maxvif; n++) { | |
320 | if (MIF_EXISTS(n) && | |
321 | mfc->mfc_un.res.ttls[n] < 255) | |
322 | seq_printf(seq, | |
323 | " %2d:%-3d", | |
324 | n, mfc->mfc_un.res.ttls[n]); | |
325 | } | |
326 | } | |
327 | seq_putc(seq, '\n'); | |
328 | } | |
329 | return 0; | |
330 | } | |
331 | ||
332 | static struct seq_operations ipmr_mfc_seq_ops = { | |
333 | .start = ipmr_mfc_seq_start, | |
334 | .next = ipmr_mfc_seq_next, | |
335 | .stop = ipmr_mfc_seq_stop, | |
336 | .show = ipmr_mfc_seq_show, | |
337 | }; | |
338 | ||
339 | static int ipmr_mfc_open(struct inode *inode, struct file *file) | |
340 | { | |
341 | return seq_open_private(file, &ipmr_mfc_seq_ops, | |
342 | sizeof(struct ipmr_mfc_iter)); | |
343 | } | |
344 | ||
345 | static struct file_operations ip6mr_mfc_fops = { | |
346 | .owner = THIS_MODULE, | |
347 | .open = ipmr_mfc_open, | |
348 | .read = seq_read, | |
349 | .llseek = seq_lseek, | |
350 | .release = seq_release, | |
351 | }; | |
352 | #endif | |
353 | ||
14fb64e1 YH |
354 | #ifdef CONFIG_IPV6_PIMSM_V2 |
355 | static int reg_vif_num = -1; | |
356 | ||
357 | static int pim6_rcv(struct sk_buff *skb) | |
358 | { | |
359 | struct pimreghdr *pim; | |
360 | struct ipv6hdr *encap; | |
361 | struct net_device *reg_dev = NULL; | |
362 | ||
363 | if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap))) | |
364 | goto drop; | |
365 | ||
366 | pim = (struct pimreghdr *)skb_transport_header(skb); | |
367 | if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) || | |
368 | (pim->flags & PIM_NULL_REGISTER) || | |
369 | (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && | |
370 | (u16)csum_fold(skb_checksum(skb, 0, skb->len, 0)))) | |
371 | goto drop; | |
372 | ||
373 | /* check if the inner packet is destined to mcast group */ | |
374 | encap = (struct ipv6hdr *)(skb_transport_header(skb) + | |
375 | sizeof(*pim)); | |
376 | ||
377 | if (!ipv6_addr_is_multicast(&encap->daddr) || | |
378 | encap->payload_len == 0 || | |
379 | ntohs(encap->payload_len) + sizeof(*pim) > skb->len) | |
380 | goto drop; | |
381 | ||
382 | read_lock(&mrt_lock); | |
383 | if (reg_vif_num >= 0) | |
384 | reg_dev = vif6_table[reg_vif_num].dev; | |
385 | if (reg_dev) | |
386 | dev_hold(reg_dev); | |
387 | read_unlock(&mrt_lock); | |
388 | ||
389 | if (reg_dev == NULL) | |
390 | goto drop; | |
391 | ||
392 | skb->mac_header = skb->network_header; | |
393 | skb_pull(skb, (u8 *)encap - skb->data); | |
394 | skb_reset_network_header(skb); | |
395 | skb->dev = reg_dev; | |
396 | skb->protocol = htons(ETH_P_IP); | |
397 | skb->ip_summed = 0; | |
398 | skb->pkt_type = PACKET_HOST; | |
399 | dst_release(skb->dst); | |
400 | ((struct net_device_stats *)netdev_priv(reg_dev))->rx_bytes += skb->len; | |
401 | ((struct net_device_stats *)netdev_priv(reg_dev))->rx_packets++; | |
402 | skb->dst = NULL; | |
403 | nf_reset(skb); | |
404 | netif_rx(skb); | |
405 | dev_put(reg_dev); | |
406 | return 0; | |
407 | drop: | |
408 | kfree_skb(skb); | |
409 | return 0; | |
410 | } | |
411 | ||
412 | static struct inet6_protocol pim6_protocol = { | |
413 | .handler = pim6_rcv, | |
414 | }; | |
415 | ||
416 | /* Service routines creating virtual interfaces: PIMREG */ | |
417 | ||
418 | static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) | |
419 | { | |
420 | read_lock(&mrt_lock); | |
421 | ((struct net_device_stats *)netdev_priv(dev))->tx_bytes += skb->len; | |
422 | ((struct net_device_stats *)netdev_priv(dev))->tx_packets++; | |
423 | ip6mr_cache_report(skb, reg_vif_num, MRT6MSG_WHOLEPKT); | |
424 | read_unlock(&mrt_lock); | |
425 | kfree_skb(skb); | |
426 | return 0; | |
427 | } | |
428 | ||
429 | static struct net_device_stats *reg_vif_get_stats(struct net_device *dev) | |
430 | { | |
431 | return (struct net_device_stats *)netdev_priv(dev); | |
432 | } | |
433 | ||
434 | static void reg_vif_setup(struct net_device *dev) | |
435 | { | |
436 | dev->type = ARPHRD_PIMREG; | |
437 | dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8; | |
438 | dev->flags = IFF_NOARP; | |
439 | dev->hard_start_xmit = reg_vif_xmit; | |
440 | dev->get_stats = reg_vif_get_stats; | |
441 | dev->destructor = free_netdev; | |
442 | } | |
443 | ||
444 | static struct net_device *ip6mr_reg_vif(void) | |
445 | { | |
446 | struct net_device *dev; | |
447 | struct inet6_dev *in_dev; | |
448 | ||
449 | dev = alloc_netdev(sizeof(struct net_device_stats), "pim6reg", | |
450 | reg_vif_setup); | |
451 | ||
452 | if (dev == NULL) | |
453 | return NULL; | |
454 | ||
455 | if (register_netdevice(dev)) { | |
456 | free_netdev(dev); | |
457 | return NULL; | |
458 | } | |
459 | dev->iflink = 0; | |
460 | ||
461 | in_dev = ipv6_find_idev(dev); | |
462 | if (!in_dev) | |
463 | goto failure; | |
464 | ||
465 | if (dev_open(dev)) | |
466 | goto failure; | |
467 | ||
468 | return dev; | |
469 | ||
470 | failure: | |
471 | /* allow the register to be completed before unregistering. */ | |
472 | rtnl_unlock(); | |
473 | rtnl_lock(); | |
474 | ||
475 | unregister_netdevice(dev); | |
476 | return NULL; | |
477 | } | |
478 | #endif | |
479 | ||
7bc570c8 YH |
480 | /* |
481 | * Delete a VIF entry | |
482 | */ | |
483 | ||
484 | static int mif6_delete(int vifi) | |
485 | { | |
486 | struct mif_device *v; | |
487 | struct net_device *dev; | |
488 | if (vifi < 0 || vifi >= maxvif) | |
489 | return -EADDRNOTAVAIL; | |
490 | ||
491 | v = &vif6_table[vifi]; | |
492 | ||
493 | write_lock_bh(&mrt_lock); | |
494 | dev = v->dev; | |
495 | v->dev = NULL; | |
496 | ||
497 | if (!dev) { | |
498 | write_unlock_bh(&mrt_lock); | |
499 | return -EADDRNOTAVAIL; | |
500 | } | |
501 | ||
14fb64e1 YH |
502 | #ifdef CONFIG_IPV6_PIMSM_V2 |
503 | if (vifi == reg_vif_num) | |
504 | reg_vif_num = -1; | |
505 | #endif | |
506 | ||
7bc570c8 YH |
507 | if (vifi + 1 == maxvif) { |
508 | int tmp; | |
509 | for (tmp = vifi - 1; tmp >= 0; tmp--) { | |
510 | if (MIF_EXISTS(tmp)) | |
511 | break; | |
512 | } | |
513 | maxvif = tmp + 1; | |
514 | } | |
515 | ||
516 | write_unlock_bh(&mrt_lock); | |
517 | ||
518 | dev_set_allmulti(dev, -1); | |
519 | ||
520 | if (v->flags & MIFF_REGISTER) | |
521 | unregister_netdevice(dev); | |
522 | ||
523 | dev_put(dev); | |
524 | return 0; | |
525 | } | |
526 | ||
527 | /* Destroy an unresolved cache entry, killing queued skbs | |
528 | and reporting error to netlink readers. | |
529 | */ | |
530 | ||
531 | static void ip6mr_destroy_unres(struct mfc6_cache *c) | |
532 | { | |
533 | struct sk_buff *skb; | |
534 | ||
535 | atomic_dec(&cache_resolve_queue_len); | |
536 | ||
537 | while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) { | |
538 | if (ipv6_hdr(skb)->version == 0) { | |
539 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); | |
540 | nlh->nlmsg_type = NLMSG_ERROR; | |
541 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
542 | skb_trim(skb, nlh->nlmsg_len); | |
543 | ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT; | |
544 | rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); | |
545 | } else | |
546 | kfree_skb(skb); | |
547 | } | |
548 | ||
549 | kmem_cache_free(mrt_cachep, c); | |
550 | } | |
551 | ||
552 | ||
553 | /* Single timer process for all the unresolved queue. */ | |
554 | ||
555 | static void ipmr_do_expire_process(unsigned long dummy) | |
556 | { | |
557 | unsigned long now = jiffies; | |
558 | unsigned long expires = 10 * HZ; | |
559 | struct mfc6_cache *c, **cp; | |
560 | ||
561 | cp = &mfc_unres_queue; | |
562 | ||
563 | while ((c = *cp) != NULL) { | |
564 | if (time_after(c->mfc_un.unres.expires, now)) { | |
565 | /* not yet... */ | |
566 | unsigned long interval = c->mfc_un.unres.expires - now; | |
567 | if (interval < expires) | |
568 | expires = interval; | |
569 | cp = &c->next; | |
570 | continue; | |
571 | } | |
572 | ||
573 | *cp = c->next; | |
574 | ip6mr_destroy_unres(c); | |
575 | } | |
576 | ||
577 | if (atomic_read(&cache_resolve_queue_len)) | |
578 | mod_timer(&ipmr_expire_timer, jiffies + expires); | |
579 | } | |
580 | ||
581 | static void ipmr_expire_process(unsigned long dummy) | |
582 | { | |
583 | if (!spin_trylock(&mfc_unres_lock)) { | |
584 | mod_timer(&ipmr_expire_timer, jiffies + 1); | |
585 | return; | |
586 | } | |
587 | ||
588 | if (atomic_read(&cache_resolve_queue_len)) | |
589 | ipmr_do_expire_process(dummy); | |
590 | ||
591 | spin_unlock(&mfc_unres_lock); | |
592 | } | |
593 | ||
594 | /* Fill oifs list. It is called under write locked mrt_lock. */ | |
595 | ||
596 | static void ip6mr_update_thresholds(struct mfc6_cache *cache, unsigned char *ttls) | |
597 | { | |
598 | int vifi; | |
599 | ||
600 | cache->mfc_un.res.minvif = MAXVIFS; | |
601 | cache->mfc_un.res.maxvif = 0; | |
602 | memset(cache->mfc_un.res.ttls, 255, MAXVIFS); | |
603 | ||
604 | for (vifi = 0; vifi < maxvif; vifi++) { | |
605 | if (MIF_EXISTS(vifi) && ttls[vifi] && ttls[vifi] < 255) { | |
606 | cache->mfc_un.res.ttls[vifi] = ttls[vifi]; | |
607 | if (cache->mfc_un.res.minvif > vifi) | |
608 | cache->mfc_un.res.minvif = vifi; | |
609 | if (cache->mfc_un.res.maxvif <= vifi) | |
610 | cache->mfc_un.res.maxvif = vifi + 1; | |
611 | } | |
612 | } | |
613 | } | |
614 | ||
615 | static int mif6_add(struct mif6ctl *vifc, int mrtsock) | |
616 | { | |
617 | int vifi = vifc->mif6c_mifi; | |
618 | struct mif_device *v = &vif6_table[vifi]; | |
619 | struct net_device *dev; | |
620 | ||
621 | /* Is vif busy ? */ | |
622 | if (MIF_EXISTS(vifi)) | |
623 | return -EADDRINUSE; | |
624 | ||
625 | switch (vifc->mif6c_flags) { | |
14fb64e1 YH |
626 | #ifdef CONFIG_IPV6_PIMSM_V2 |
627 | case MIFF_REGISTER: | |
628 | /* | |
629 | * Special Purpose VIF in PIM | |
630 | * All the packets will be sent to the daemon | |
631 | */ | |
632 | if (reg_vif_num >= 0) | |
633 | return -EADDRINUSE; | |
634 | dev = ip6mr_reg_vif(); | |
635 | if (!dev) | |
636 | return -ENOBUFS; | |
637 | break; | |
638 | #endif | |
7bc570c8 YH |
639 | case 0: |
640 | dev = dev_get_by_index(&init_net, vifc->mif6c_pifi); | |
641 | if (!dev) | |
642 | return -EADDRNOTAVAIL; | |
643 | dev_put(dev); | |
644 | break; | |
645 | default: | |
646 | return -EINVAL; | |
647 | } | |
648 | ||
649 | dev_set_allmulti(dev, 1); | |
650 | ||
651 | /* | |
652 | * Fill in the VIF structures | |
653 | */ | |
654 | v->rate_limit = vifc->vifc_rate_limit; | |
655 | v->flags = vifc->mif6c_flags; | |
656 | if (!mrtsock) | |
657 | v->flags |= VIFF_STATIC; | |
658 | v->threshold = vifc->vifc_threshold; | |
659 | v->bytes_in = 0; | |
660 | v->bytes_out = 0; | |
661 | v->pkt_in = 0; | |
662 | v->pkt_out = 0; | |
663 | v->link = dev->ifindex; | |
664 | if (v->flags & MIFF_REGISTER) | |
665 | v->link = dev->iflink; | |
666 | ||
667 | /* And finish update writing critical data */ | |
668 | write_lock_bh(&mrt_lock); | |
669 | dev_hold(dev); | |
670 | v->dev = dev; | |
14fb64e1 YH |
671 | #ifdef CONFIG_IPV6_PIMSM_V2 |
672 | if (v->flags & MIFF_REGISTER) | |
673 | reg_vif_num = vifi; | |
674 | #endif | |
7bc570c8 YH |
675 | if (vifi + 1 > maxvif) |
676 | maxvif = vifi + 1; | |
677 | write_unlock_bh(&mrt_lock); | |
678 | return 0; | |
679 | } | |
680 | ||
681 | static struct mfc6_cache *ip6mr_cache_find(struct in6_addr *origin, struct in6_addr *mcastgrp) | |
682 | { | |
683 | int line = MFC6_HASH(mcastgrp, origin); | |
684 | struct mfc6_cache *c; | |
685 | ||
686 | for (c = mfc6_cache_array[line]; c; c = c->next) { | |
687 | if (ipv6_addr_equal(&c->mf6c_origin, origin) && | |
688 | ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) | |
689 | break; | |
690 | } | |
691 | return c; | |
692 | } | |
693 | ||
694 | /* | |
695 | * Allocate a multicast cache entry | |
696 | */ | |
697 | static struct mfc6_cache *ip6mr_cache_alloc(void) | |
698 | { | |
699 | struct mfc6_cache *c = kmem_cache_alloc(mrt_cachep, GFP_KERNEL); | |
700 | if (c == NULL) | |
701 | return NULL; | |
702 | memset(c, 0, sizeof(*c)); | |
703 | c->mfc_un.res.minvif = MAXVIFS; | |
704 | return c; | |
705 | } | |
706 | ||
707 | static struct mfc6_cache *ip6mr_cache_alloc_unres(void) | |
708 | { | |
709 | struct mfc6_cache *c = kmem_cache_alloc(mrt_cachep, GFP_ATOMIC); | |
710 | if (c == NULL) | |
711 | return NULL; | |
712 | memset(c, 0, sizeof(*c)); | |
713 | skb_queue_head_init(&c->mfc_un.unres.unresolved); | |
714 | c->mfc_un.unres.expires = jiffies + 10 * HZ; | |
715 | return c; | |
716 | } | |
717 | ||
718 | /* | |
719 | * A cache entry has gone into a resolved state from queued | |
720 | */ | |
721 | ||
722 | static void ip6mr_cache_resolve(struct mfc6_cache *uc, struct mfc6_cache *c) | |
723 | { | |
724 | struct sk_buff *skb; | |
725 | ||
726 | /* | |
727 | * Play the pending entries through our router | |
728 | */ | |
729 | ||
730 | while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { | |
731 | if (ipv6_hdr(skb)->version == 0) { | |
732 | int err; | |
733 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); | |
734 | ||
735 | if (ip6mr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) { | |
549e028d | 736 | nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; |
7bc570c8 YH |
737 | } else { |
738 | nlh->nlmsg_type = NLMSG_ERROR; | |
739 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
740 | skb_trim(skb, nlh->nlmsg_len); | |
741 | ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE; | |
742 | } | |
743 | err = rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); | |
744 | } else | |
745 | ip6_mr_forward(skb, c); | |
746 | } | |
747 | } | |
748 | ||
749 | /* | |
750 | * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd | |
751 | * expects the following bizarre scheme. | |
752 | * | |
753 | * Called under mrt_lock. | |
754 | */ | |
755 | ||
756 | static int ip6mr_cache_report(struct sk_buff *pkt, vifi_t vifi, int assert) | |
757 | { | |
758 | struct sk_buff *skb; | |
759 | struct mrt6msg *msg; | |
760 | int ret; | |
761 | ||
14fb64e1 YH |
762 | #ifdef CONFIG_IPV6_PIMSM_V2 |
763 | if (assert == MRT6MSG_WHOLEPKT) | |
764 | skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) | |
765 | +sizeof(*msg)); | |
766 | else | |
767 | #endif | |
768 | skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); | |
7bc570c8 YH |
769 | |
770 | if (!skb) | |
771 | return -ENOBUFS; | |
772 | ||
773 | /* I suppose that internal messages | |
774 | * do not require checksums */ | |
775 | ||
776 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
777 | ||
14fb64e1 YH |
778 | #ifdef CONFIG_IPV6_PIMSM_V2 |
779 | if (assert == MRT6MSG_WHOLEPKT) { | |
780 | /* Ugly, but we have no choice with this interface. | |
781 | Duplicate old header, fix length etc. | |
782 | And all this only to mangle msg->im6_msgtype and | |
783 | to set msg->im6_mbz to "mbz" :-) | |
784 | */ | |
785 | skb_push(skb, -skb_network_offset(pkt)); | |
786 | ||
787 | skb_push(skb, sizeof(*msg)); | |
788 | skb_reset_transport_header(skb); | |
789 | msg = (struct mrt6msg *)skb_transport_header(skb); | |
790 | msg->im6_mbz = 0; | |
791 | msg->im6_msgtype = MRT6MSG_WHOLEPKT; | |
792 | msg->im6_mif = reg_vif_num; | |
793 | msg->im6_pad = 0; | |
794 | ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr); | |
795 | ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr); | |
796 | ||
797 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
798 | } else | |
799 | #endif | |
800 | { | |
7bc570c8 YH |
801 | /* |
802 | * Copy the IP header | |
803 | */ | |
804 | ||
805 | skb_put(skb, sizeof(struct ipv6hdr)); | |
806 | skb_reset_network_header(skb); | |
807 | skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); | |
808 | ||
809 | /* | |
810 | * Add our header | |
811 | */ | |
812 | skb_put(skb, sizeof(*msg)); | |
813 | skb_reset_transport_header(skb); | |
814 | msg = (struct mrt6msg *)skb_transport_header(skb); | |
815 | ||
816 | msg->im6_mbz = 0; | |
817 | msg->im6_msgtype = assert; | |
818 | msg->im6_mif = vifi; | |
819 | msg->im6_pad = 0; | |
820 | ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr); | |
821 | ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr); | |
822 | ||
823 | skb->dst = dst_clone(pkt->dst); | |
824 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
825 | ||
826 | skb_pull(skb, sizeof(struct ipv6hdr)); | |
14fb64e1 | 827 | } |
7bc570c8 YH |
828 | |
829 | if (mroute6_socket == NULL) { | |
830 | kfree_skb(skb); | |
831 | return -EINVAL; | |
832 | } | |
833 | ||
834 | /* | |
835 | * Deliver to user space multicast routing algorithms | |
836 | */ | |
837 | if ((ret = sock_queue_rcv_skb(mroute6_socket, skb)) < 0) { | |
838 | if (net_ratelimit()) | |
839 | printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n"); | |
840 | kfree_skb(skb); | |
841 | } | |
842 | ||
843 | return ret; | |
844 | } | |
845 | ||
846 | /* | |
847 | * Queue a packet for resolution. It gets locked cache entry! | |
848 | */ | |
849 | ||
850 | static int | |
851 | ip6mr_cache_unresolved(vifi_t vifi, struct sk_buff *skb) | |
852 | { | |
853 | int err; | |
854 | struct mfc6_cache *c; | |
855 | ||
856 | spin_lock_bh(&mfc_unres_lock); | |
857 | for (c = mfc_unres_queue; c; c = c->next) { | |
858 | if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && | |
859 | ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) | |
860 | break; | |
861 | } | |
862 | ||
863 | if (c == NULL) { | |
864 | /* | |
865 | * Create a new entry if allowable | |
866 | */ | |
867 | ||
868 | if (atomic_read(&cache_resolve_queue_len) >= 10 || | |
869 | (c = ip6mr_cache_alloc_unres()) == NULL) { | |
870 | spin_unlock_bh(&mfc_unres_lock); | |
871 | ||
872 | kfree_skb(skb); | |
873 | return -ENOBUFS; | |
874 | } | |
875 | ||
876 | /* | |
877 | * Fill in the new cache entry | |
878 | */ | |
879 | c->mf6c_parent = -1; | |
880 | c->mf6c_origin = ipv6_hdr(skb)->saddr; | |
881 | c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; | |
882 | ||
883 | /* | |
884 | * Reflect first query at pim6sd | |
885 | */ | |
886 | if ((err = ip6mr_cache_report(skb, vifi, MRT6MSG_NOCACHE)) < 0) { | |
887 | /* If the report failed throw the cache entry | |
888 | out - Brad Parker | |
889 | */ | |
890 | spin_unlock_bh(&mfc_unres_lock); | |
891 | ||
892 | kmem_cache_free(mrt_cachep, c); | |
893 | kfree_skb(skb); | |
894 | return err; | |
895 | } | |
896 | ||
897 | atomic_inc(&cache_resolve_queue_len); | |
898 | c->next = mfc_unres_queue; | |
899 | mfc_unres_queue = c; | |
900 | ||
901 | ipmr_do_expire_process(1); | |
902 | } | |
903 | ||
904 | /* | |
905 | * See if we can append the packet | |
906 | */ | |
907 | if (c->mfc_un.unres.unresolved.qlen > 3) { | |
908 | kfree_skb(skb); | |
909 | err = -ENOBUFS; | |
910 | } else { | |
911 | skb_queue_tail(&c->mfc_un.unres.unresolved, skb); | |
912 | err = 0; | |
913 | } | |
914 | ||
915 | spin_unlock_bh(&mfc_unres_lock); | |
916 | return err; | |
917 | } | |
918 | ||
919 | /* | |
920 | * MFC6 cache manipulation by user space | |
921 | */ | |
922 | ||
923 | static int ip6mr_mfc_delete(struct mf6cctl *mfc) | |
924 | { | |
925 | int line; | |
926 | struct mfc6_cache *c, **cp; | |
927 | ||
928 | line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); | |
929 | ||
930 | for (cp = &mfc6_cache_array[line]; (c = *cp) != NULL; cp = &c->next) { | |
931 | if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && | |
932 | ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) { | |
933 | write_lock_bh(&mrt_lock); | |
934 | *cp = c->next; | |
935 | write_unlock_bh(&mrt_lock); | |
936 | ||
937 | kmem_cache_free(mrt_cachep, c); | |
938 | return 0; | |
939 | } | |
940 | } | |
941 | return -ENOENT; | |
942 | } | |
943 | ||
944 | static int ip6mr_device_event(struct notifier_block *this, | |
945 | unsigned long event, void *ptr) | |
946 | { | |
947 | struct net_device *dev = ptr; | |
948 | struct mif_device *v; | |
949 | int ct; | |
950 | ||
951 | if (dev_net(dev) != &init_net) | |
952 | return NOTIFY_DONE; | |
953 | ||
954 | if (event != NETDEV_UNREGISTER) | |
955 | return NOTIFY_DONE; | |
956 | ||
957 | v = &vif6_table[0]; | |
958 | for (ct = 0; ct < maxvif; ct++, v++) { | |
959 | if (v->dev == dev) | |
960 | mif6_delete(ct); | |
961 | } | |
962 | return NOTIFY_DONE; | |
963 | } | |
964 | ||
965 | static struct notifier_block ip6_mr_notifier = { | |
966 | .notifier_call = ip6mr_device_event | |
967 | }; | |
968 | ||
969 | /* | |
970 | * Setup for IP multicast routing | |
971 | */ | |
972 | ||
973 | void __init ip6_mr_init(void) | |
974 | { | |
975 | mrt_cachep = kmem_cache_create("ip6_mrt_cache", | |
976 | sizeof(struct mfc6_cache), | |
977 | 0, SLAB_HWCACHE_ALIGN, | |
978 | NULL); | |
979 | if (!mrt_cachep) | |
980 | panic("cannot allocate ip6_mrt_cache"); | |
981 | ||
982 | setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0); | |
983 | register_netdevice_notifier(&ip6_mr_notifier); | |
984 | #ifdef CONFIG_PROC_FS | |
985 | proc_net_fops_create(&init_net, "ip6_mr_vif", 0, &ip6mr_vif_fops); | |
986 | proc_net_fops_create(&init_net, "ip6_mr_cache", 0, &ip6mr_mfc_fops); | |
987 | #endif | |
988 | } | |
989 | ||
990 | ||
991 | static int ip6mr_mfc_add(struct mf6cctl *mfc, int mrtsock) | |
992 | { | |
993 | int line; | |
994 | struct mfc6_cache *uc, *c, **cp; | |
995 | unsigned char ttls[MAXVIFS]; | |
996 | int i; | |
997 | ||
998 | memset(ttls, 255, MAXVIFS); | |
999 | for (i = 0; i < MAXVIFS; i++) { | |
1000 | if (IF_ISSET(i, &mfc->mf6cc_ifset)) | |
1001 | ttls[i] = 1; | |
1002 | ||
1003 | } | |
1004 | ||
1005 | line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); | |
1006 | ||
1007 | for (cp = &mfc6_cache_array[line]; (c = *cp) != NULL; cp = &c->next) { | |
1008 | if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && | |
1009 | ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) | |
1010 | break; | |
1011 | } | |
1012 | ||
1013 | if (c != NULL) { | |
1014 | write_lock_bh(&mrt_lock); | |
1015 | c->mf6c_parent = mfc->mf6cc_parent; | |
1016 | ip6mr_update_thresholds(c, ttls); | |
1017 | if (!mrtsock) | |
1018 | c->mfc_flags |= MFC_STATIC; | |
1019 | write_unlock_bh(&mrt_lock); | |
1020 | return 0; | |
1021 | } | |
1022 | ||
1023 | if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) | |
1024 | return -EINVAL; | |
1025 | ||
1026 | c = ip6mr_cache_alloc(); | |
1027 | if (c == NULL) | |
1028 | return -ENOMEM; | |
1029 | ||
1030 | c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; | |
1031 | c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; | |
1032 | c->mf6c_parent = mfc->mf6cc_parent; | |
1033 | ip6mr_update_thresholds(c, ttls); | |
1034 | if (!mrtsock) | |
1035 | c->mfc_flags |= MFC_STATIC; | |
1036 | ||
1037 | write_lock_bh(&mrt_lock); | |
1038 | c->next = mfc6_cache_array[line]; | |
1039 | mfc6_cache_array[line] = c; | |
1040 | write_unlock_bh(&mrt_lock); | |
1041 | ||
1042 | /* | |
1043 | * Check to see if we resolved a queued list. If so we | |
1044 | * need to send on the frames and tidy up. | |
1045 | */ | |
1046 | spin_lock_bh(&mfc_unres_lock); | |
1047 | for (cp = &mfc_unres_queue; (uc = *cp) != NULL; | |
1048 | cp = &uc->next) { | |
1049 | if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && | |
1050 | ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { | |
1051 | *cp = uc->next; | |
1052 | if (atomic_dec_and_test(&cache_resolve_queue_len)) | |
1053 | del_timer(&ipmr_expire_timer); | |
1054 | break; | |
1055 | } | |
1056 | } | |
1057 | spin_unlock_bh(&mfc_unres_lock); | |
1058 | ||
1059 | if (uc) { | |
1060 | ip6mr_cache_resolve(uc, c); | |
1061 | kmem_cache_free(mrt_cachep, uc); | |
1062 | } | |
1063 | return 0; | |
1064 | } | |
1065 | ||
1066 | /* | |
1067 | * Close the multicast socket, and clear the vif tables etc | |
1068 | */ | |
1069 | ||
1070 | static void mroute_clean_tables(struct sock *sk) | |
1071 | { | |
1072 | int i; | |
1073 | ||
1074 | /* | |
1075 | * Shut down all active vif entries | |
1076 | */ | |
1077 | for (i = 0; i < maxvif; i++) { | |
1078 | if (!(vif6_table[i].flags & VIFF_STATIC)) | |
1079 | mif6_delete(i); | |
1080 | } | |
1081 | ||
1082 | /* | |
1083 | * Wipe the cache | |
1084 | */ | |
1085 | for (i = 0; i < ARRAY_SIZE(mfc6_cache_array); i++) { | |
1086 | struct mfc6_cache *c, **cp; | |
1087 | ||
1088 | cp = &mfc6_cache_array[i]; | |
1089 | while ((c = *cp) != NULL) { | |
1090 | if (c->mfc_flags & MFC_STATIC) { | |
1091 | cp = &c->next; | |
1092 | continue; | |
1093 | } | |
1094 | write_lock_bh(&mrt_lock); | |
1095 | *cp = c->next; | |
1096 | write_unlock_bh(&mrt_lock); | |
1097 | ||
1098 | kmem_cache_free(mrt_cachep, c); | |
1099 | } | |
1100 | } | |
1101 | ||
1102 | if (atomic_read(&cache_resolve_queue_len) != 0) { | |
1103 | struct mfc6_cache *c; | |
1104 | ||
1105 | spin_lock_bh(&mfc_unres_lock); | |
1106 | while (mfc_unres_queue != NULL) { | |
1107 | c = mfc_unres_queue; | |
1108 | mfc_unres_queue = c->next; | |
1109 | spin_unlock_bh(&mfc_unres_lock); | |
1110 | ||
1111 | ip6mr_destroy_unres(c); | |
1112 | ||
1113 | spin_lock_bh(&mfc_unres_lock); | |
1114 | } | |
1115 | spin_unlock_bh(&mfc_unres_lock); | |
1116 | } | |
1117 | } | |
1118 | ||
1119 | static int ip6mr_sk_init(struct sock *sk) | |
1120 | { | |
1121 | int err = 0; | |
1122 | ||
1123 | rtnl_lock(); | |
1124 | write_lock_bh(&mrt_lock); | |
1125 | if (likely(mroute6_socket == NULL)) | |
1126 | mroute6_socket = sk; | |
1127 | else | |
1128 | err = -EADDRINUSE; | |
1129 | write_unlock_bh(&mrt_lock); | |
1130 | ||
1131 | rtnl_unlock(); | |
1132 | ||
1133 | return err; | |
1134 | } | |
1135 | ||
1136 | int ip6mr_sk_done(struct sock *sk) | |
1137 | { | |
1138 | int err = 0; | |
1139 | ||
1140 | rtnl_lock(); | |
1141 | if (sk == mroute6_socket) { | |
1142 | write_lock_bh(&mrt_lock); | |
1143 | mroute6_socket = NULL; | |
1144 | write_unlock_bh(&mrt_lock); | |
1145 | ||
1146 | mroute_clean_tables(sk); | |
1147 | } else | |
1148 | err = -EACCES; | |
1149 | rtnl_unlock(); | |
1150 | ||
1151 | return err; | |
1152 | } | |
1153 | ||
1154 | /* | |
1155 | * Socket options and virtual interface manipulation. The whole | |
1156 | * virtual interface system is a complete heap, but unfortunately | |
1157 | * that's how BSD mrouted happens to think. Maybe one day with a proper | |
1158 | * MOSPF/PIM router set up we can clean this up. | |
1159 | */ | |
1160 | ||
1161 | int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, int optlen) | |
1162 | { | |
1163 | int ret; | |
1164 | struct mif6ctl vif; | |
1165 | struct mf6cctl mfc; | |
1166 | mifi_t mifi; | |
1167 | ||
1168 | if (optname != MRT6_INIT) { | |
1169 | if (sk != mroute6_socket && !capable(CAP_NET_ADMIN)) | |
1170 | return -EACCES; | |
1171 | } | |
1172 | ||
1173 | switch (optname) { | |
1174 | case MRT6_INIT: | |
1175 | if (sk->sk_type != SOCK_RAW || | |
1176 | inet_sk(sk)->num != IPPROTO_ICMPV6) | |
1177 | return -EOPNOTSUPP; | |
1178 | if (optlen < sizeof(int)) | |
1179 | return -EINVAL; | |
1180 | ||
1181 | return ip6mr_sk_init(sk); | |
1182 | ||
1183 | case MRT6_DONE: | |
1184 | return ip6mr_sk_done(sk); | |
1185 | ||
1186 | case MRT6_ADD_MIF: | |
1187 | if (optlen < sizeof(vif)) | |
1188 | return -EINVAL; | |
1189 | if (copy_from_user(&vif, optval, sizeof(vif))) | |
1190 | return -EFAULT; | |
1191 | if (vif.mif6c_mifi >= MAXVIFS) | |
1192 | return -ENFILE; | |
1193 | rtnl_lock(); | |
1194 | ret = mif6_add(&vif, sk == mroute6_socket); | |
1195 | rtnl_unlock(); | |
1196 | return ret; | |
1197 | ||
1198 | case MRT6_DEL_MIF: | |
1199 | if (optlen < sizeof(mifi_t)) | |
1200 | return -EINVAL; | |
1201 | if (copy_from_user(&mifi, optval, sizeof(mifi_t))) | |
1202 | return -EFAULT; | |
1203 | rtnl_lock(); | |
1204 | ret = mif6_delete(mifi); | |
1205 | rtnl_unlock(); | |
1206 | return ret; | |
1207 | ||
1208 | /* | |
1209 | * Manipulate the forwarding caches. These live | |
1210 | * in a sort of kernel/user symbiosis. | |
1211 | */ | |
1212 | case MRT6_ADD_MFC: | |
1213 | case MRT6_DEL_MFC: | |
1214 | if (optlen < sizeof(mfc)) | |
1215 | return -EINVAL; | |
1216 | if (copy_from_user(&mfc, optval, sizeof(mfc))) | |
1217 | return -EFAULT; | |
1218 | rtnl_lock(); | |
1219 | if (optname == MRT6_DEL_MFC) | |
1220 | ret = ip6mr_mfc_delete(&mfc); | |
1221 | else | |
1222 | ret = ip6mr_mfc_add(&mfc, sk == mroute6_socket); | |
1223 | rtnl_unlock(); | |
1224 | return ret; | |
1225 | ||
14fb64e1 YH |
1226 | /* |
1227 | * Control PIM assert (to activate pim will activate assert) | |
1228 | */ | |
1229 | case MRT6_ASSERT: | |
1230 | { | |
1231 | int v; | |
1232 | if (get_user(v, (int __user *)optval)) | |
1233 | return -EFAULT; | |
1234 | mroute_do_assert = !!v; | |
1235 | return 0; | |
1236 | } | |
1237 | ||
1238 | #ifdef CONFIG_IPV6_PIMSM_V2 | |
1239 | case MRT6_PIM: | |
1240 | { | |
a9f83bf3 | 1241 | int v; |
14fb64e1 YH |
1242 | if (get_user(v, (int __user *)optval)) |
1243 | return -EFAULT; | |
1244 | v = !!v; | |
1245 | rtnl_lock(); | |
1246 | ret = 0; | |
1247 | if (v != mroute_do_pim) { | |
1248 | mroute_do_pim = v; | |
1249 | mroute_do_assert = v; | |
1250 | if (mroute_do_pim) | |
1251 | ret = inet6_add_protocol(&pim6_protocol, | |
1252 | IPPROTO_PIM); | |
1253 | else | |
1254 | ret = inet6_del_protocol(&pim6_protocol, | |
1255 | IPPROTO_PIM); | |
1256 | if (ret < 0) | |
1257 | ret = -EAGAIN; | |
1258 | } | |
1259 | rtnl_unlock(); | |
1260 | return ret; | |
1261 | } | |
1262 | ||
1263 | #endif | |
7bc570c8 YH |
1264 | /* |
1265 | * Spurious command, or MRT_VERSION which you cannot | |
1266 | * set. | |
1267 | */ | |
1268 | default: | |
1269 | return -ENOPROTOOPT; | |
1270 | } | |
1271 | } | |
1272 | ||
1273 | /* | |
1274 | * Getsock opt support for the multicast routing system. | |
1275 | */ | |
1276 | ||
1277 | int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, | |
1278 | int __user *optlen) | |
1279 | { | |
1280 | int olr; | |
1281 | int val; | |
1282 | ||
1283 | switch (optname) { | |
1284 | case MRT6_VERSION: | |
1285 | val = 0x0305; | |
1286 | break; | |
14fb64e1 YH |
1287 | #ifdef CONFIG_IPV6_PIMSM_V2 |
1288 | case MRT6_PIM: | |
1289 | val = mroute_do_pim; | |
1290 | break; | |
1291 | #endif | |
1292 | case MRT6_ASSERT: | |
1293 | val = mroute_do_assert; | |
1294 | break; | |
7bc570c8 YH |
1295 | default: |
1296 | return -ENOPROTOOPT; | |
1297 | } | |
1298 | ||
1299 | if (get_user(olr, optlen)) | |
1300 | return -EFAULT; | |
1301 | ||
1302 | olr = min_t(int, olr, sizeof(int)); | |
1303 | if (olr < 0) | |
1304 | return -EINVAL; | |
1305 | ||
1306 | if (put_user(olr, optlen)) | |
1307 | return -EFAULT; | |
1308 | if (copy_to_user(optval, &val, olr)) | |
1309 | return -EFAULT; | |
1310 | return 0; | |
1311 | } | |
1312 | ||
1313 | /* | |
1314 | * The IP multicast ioctl support routines. | |
1315 | */ | |
1316 | ||
1317 | int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) | |
1318 | { | |
1319 | struct sioc_sg_req6 sr; | |
1320 | struct sioc_mif_req6 vr; | |
1321 | struct mif_device *vif; | |
1322 | struct mfc6_cache *c; | |
1323 | ||
1324 | switch (cmd) { | |
1325 | case SIOCGETMIFCNT_IN6: | |
1326 | if (copy_from_user(&vr, arg, sizeof(vr))) | |
1327 | return -EFAULT; | |
1328 | if (vr.mifi >= maxvif) | |
1329 | return -EINVAL; | |
1330 | read_lock(&mrt_lock); | |
1331 | vif = &vif6_table[vr.mifi]; | |
1332 | if (MIF_EXISTS(vr.mifi)) { | |
1333 | vr.icount = vif->pkt_in; | |
1334 | vr.ocount = vif->pkt_out; | |
1335 | vr.ibytes = vif->bytes_in; | |
1336 | vr.obytes = vif->bytes_out; | |
1337 | read_unlock(&mrt_lock); | |
1338 | ||
1339 | if (copy_to_user(arg, &vr, sizeof(vr))) | |
1340 | return -EFAULT; | |
1341 | return 0; | |
1342 | } | |
1343 | read_unlock(&mrt_lock); | |
1344 | return -EADDRNOTAVAIL; | |
1345 | case SIOCGETSGCNT_IN6: | |
1346 | if (copy_from_user(&sr, arg, sizeof(sr))) | |
1347 | return -EFAULT; | |
1348 | ||
1349 | read_lock(&mrt_lock); | |
1350 | c = ip6mr_cache_find(&sr.src.sin6_addr, &sr.grp.sin6_addr); | |
1351 | if (c) { | |
1352 | sr.pktcnt = c->mfc_un.res.pkt; | |
1353 | sr.bytecnt = c->mfc_un.res.bytes; | |
1354 | sr.wrong_if = c->mfc_un.res.wrong_if; | |
1355 | read_unlock(&mrt_lock); | |
1356 | ||
1357 | if (copy_to_user(arg, &sr, sizeof(sr))) | |
1358 | return -EFAULT; | |
1359 | return 0; | |
1360 | } | |
1361 | read_unlock(&mrt_lock); | |
1362 | return -EADDRNOTAVAIL; | |
1363 | default: | |
1364 | return -ENOIOCTLCMD; | |
1365 | } | |
1366 | } | |
1367 | ||
1368 | ||
1369 | static inline int ip6mr_forward2_finish(struct sk_buff *skb) | |
1370 | { | |
1371 | /* XXX stats */ | |
1372 | return dst_output(skb); | |
1373 | } | |
1374 | ||
1375 | /* | |
1376 | * Processing handlers for ip6mr_forward | |
1377 | */ | |
1378 | ||
1379 | static int ip6mr_forward2(struct sk_buff *skb, struct mfc6_cache *c, int vifi) | |
1380 | { | |
1381 | struct ipv6hdr *ipv6h; | |
1382 | struct mif_device *vif = &vif6_table[vifi]; | |
1383 | struct net_device *dev; | |
1384 | struct dst_entry *dst; | |
1385 | struct flowi fl; | |
1386 | ||
1387 | if (vif->dev == NULL) | |
1388 | goto out_free; | |
1389 | ||
14fb64e1 YH |
1390 | #ifdef CONFIG_IPV6_PIMSM_V2 |
1391 | if (vif->flags & MIFF_REGISTER) { | |
1392 | vif->pkt_out++; | |
1393 | vif->bytes_out += skb->len; | |
1394 | ((struct net_device_stats *)netdev_priv(vif->dev))->tx_bytes += skb->len; | |
1395 | ((struct net_device_stats *)netdev_priv(vif->dev))->tx_packets++; | |
1396 | ip6mr_cache_report(skb, vifi, MRT6MSG_WHOLEPKT); | |
1397 | kfree_skb(skb); | |
1398 | return 0; | |
1399 | } | |
1400 | #endif | |
1401 | ||
7bc570c8 YH |
1402 | ipv6h = ipv6_hdr(skb); |
1403 | ||
1404 | fl = (struct flowi) { | |
1405 | .oif = vif->link, | |
1406 | .nl_u = { .ip6_u = | |
1407 | { .daddr = ipv6h->daddr, } | |
1408 | } | |
1409 | }; | |
1410 | ||
1411 | dst = ip6_route_output(&init_net, NULL, &fl); | |
1412 | if (!dst) | |
1413 | goto out_free; | |
1414 | ||
1415 | dst_release(skb->dst); | |
1416 | skb->dst = dst; | |
1417 | ||
1418 | /* | |
1419 | * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally | |
1420 | * not only before forwarding, but after forwarding on all output | |
1421 | * interfaces. It is clear, if mrouter runs a multicasting | |
1422 | * program, it should receive packets not depending to what interface | |
1423 | * program is joined. | |
1424 | * If we will not make it, the program will have to join on all | |
1425 | * interfaces. On the other hand, multihoming host (or router, but | |
1426 | * not mrouter) cannot join to more than one interface - it will | |
1427 | * result in receiving multiple packets. | |
1428 | */ | |
1429 | dev = vif->dev; | |
1430 | skb->dev = dev; | |
1431 | vif->pkt_out++; | |
1432 | vif->bytes_out += skb->len; | |
1433 | ||
1434 | /* We are about to write */ | |
1435 | /* XXX: extension headers? */ | |
1436 | if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) | |
1437 | goto out_free; | |
1438 | ||
1439 | ipv6h = ipv6_hdr(skb); | |
1440 | ipv6h->hop_limit--; | |
1441 | ||
1442 | IP6CB(skb)->flags |= IP6SKB_FORWARDED; | |
1443 | ||
1444 | return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dev, | |
1445 | ip6mr_forward2_finish); | |
1446 | ||
1447 | out_free: | |
1448 | kfree_skb(skb); | |
1449 | return 0; | |
1450 | } | |
1451 | ||
1452 | static int ip6mr_find_vif(struct net_device *dev) | |
1453 | { | |
1454 | int ct; | |
1455 | for (ct = maxvif - 1; ct >= 0; ct--) { | |
1456 | if (vif6_table[ct].dev == dev) | |
1457 | break; | |
1458 | } | |
1459 | return ct; | |
1460 | } | |
1461 | ||
1462 | static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache) | |
1463 | { | |
1464 | int psend = -1; | |
1465 | int vif, ct; | |
1466 | ||
1467 | vif = cache->mf6c_parent; | |
1468 | cache->mfc_un.res.pkt++; | |
1469 | cache->mfc_un.res.bytes += skb->len; | |
1470 | ||
14fb64e1 YH |
1471 | /* |
1472 | * Wrong interface: drop packet and (maybe) send PIM assert. | |
1473 | */ | |
1474 | if (vif6_table[vif].dev != skb->dev) { | |
1475 | int true_vifi; | |
1476 | ||
1477 | cache->mfc_un.res.wrong_if++; | |
1478 | true_vifi = ip6mr_find_vif(skb->dev); | |
1479 | ||
1480 | if (true_vifi >= 0 && mroute_do_assert && | |
1481 | /* pimsm uses asserts, when switching from RPT to SPT, | |
1482 | so that we cannot check that packet arrived on an oif. | |
1483 | It is bad, but otherwise we would need to move pretty | |
1484 | large chunk of pimd to kernel. Ough... --ANK | |
1485 | */ | |
1486 | (mroute_do_pim || cache->mfc_un.res.ttls[true_vifi] < 255) && | |
1487 | time_after(jiffies, | |
1488 | cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { | |
1489 | cache->mfc_un.res.last_assert = jiffies; | |
1490 | ip6mr_cache_report(skb, true_vifi, MRT6MSG_WRONGMIF); | |
1491 | } | |
1492 | goto dont_forward; | |
1493 | } | |
1494 | ||
7bc570c8 YH |
1495 | vif6_table[vif].pkt_in++; |
1496 | vif6_table[vif].bytes_in += skb->len; | |
1497 | ||
1498 | /* | |
1499 | * Forward the frame | |
1500 | */ | |
1501 | for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { | |
1502 | if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { | |
1503 | if (psend != -1) { | |
1504 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1505 | if (skb2) | |
1506 | ip6mr_forward2(skb2, cache, psend); | |
1507 | } | |
1508 | psend = ct; | |
1509 | } | |
1510 | } | |
1511 | if (psend != -1) { | |
1512 | ip6mr_forward2(skb, cache, psend); | |
1513 | return 0; | |
1514 | } | |
1515 | ||
14fb64e1 | 1516 | dont_forward: |
7bc570c8 YH |
1517 | kfree_skb(skb); |
1518 | return 0; | |
1519 | } | |
1520 | ||
1521 | ||
1522 | /* | |
1523 | * Multicast packets for forwarding arrive here | |
1524 | */ | |
1525 | ||
1526 | int ip6_mr_input(struct sk_buff *skb) | |
1527 | { | |
1528 | struct mfc6_cache *cache; | |
1529 | ||
1530 | read_lock(&mrt_lock); | |
1531 | cache = ip6mr_cache_find(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); | |
1532 | ||
1533 | /* | |
1534 | * No usable cache entry | |
1535 | */ | |
1536 | if (cache == NULL) { | |
1537 | int vif; | |
1538 | ||
1539 | vif = ip6mr_find_vif(skb->dev); | |
1540 | if (vif >= 0) { | |
1541 | int err = ip6mr_cache_unresolved(vif, skb); | |
1542 | read_unlock(&mrt_lock); | |
1543 | ||
1544 | return err; | |
1545 | } | |
1546 | read_unlock(&mrt_lock); | |
1547 | kfree_skb(skb); | |
1548 | return -ENODEV; | |
1549 | } | |
1550 | ||
1551 | ip6_mr_forward(skb, cache); | |
1552 | ||
1553 | read_unlock(&mrt_lock); | |
1554 | ||
1555 | return 0; | |
1556 | } | |
1557 | ||
1558 | ||
1559 | static int | |
1560 | ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm) | |
1561 | { | |
1562 | int ct; | |
1563 | struct rtnexthop *nhp; | |
1564 | struct net_device *dev = vif6_table[c->mf6c_parent].dev; | |
549e028d | 1565 | u8 *b = skb_tail_pointer(skb); |
7bc570c8 YH |
1566 | struct rtattr *mp_head; |
1567 | ||
1568 | if (dev) | |
1569 | RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex); | |
1570 | ||
1571 | mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0)); | |
1572 | ||
1573 | for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { | |
1574 | if (c->mfc_un.res.ttls[ct] < 255) { | |
1575 | if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4)) | |
1576 | goto rtattr_failure; | |
1577 | nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp))); | |
1578 | nhp->rtnh_flags = 0; | |
1579 | nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; | |
1580 | nhp->rtnh_ifindex = vif6_table[ct].dev->ifindex; | |
1581 | nhp->rtnh_len = sizeof(*nhp); | |
1582 | } | |
1583 | } | |
1584 | mp_head->rta_type = RTA_MULTIPATH; | |
549e028d | 1585 | mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head; |
7bc570c8 YH |
1586 | rtm->rtm_type = RTN_MULTICAST; |
1587 | return 1; | |
1588 | ||
1589 | rtattr_failure: | |
1590 | nlmsg_trim(skb, b); | |
1591 | return -EMSGSIZE; | |
1592 | } | |
1593 | ||
1594 | int ip6mr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait) | |
1595 | { | |
1596 | int err; | |
1597 | struct mfc6_cache *cache; | |
1598 | struct rt6_info *rt = (struct rt6_info *)skb->dst; | |
1599 | ||
1600 | read_lock(&mrt_lock); | |
1601 | cache = ip6mr_cache_find(&rt->rt6i_src.addr, &rt->rt6i_dst.addr); | |
1602 | ||
1603 | if (!cache) { | |
1604 | struct sk_buff *skb2; | |
1605 | struct ipv6hdr *iph; | |
1606 | struct net_device *dev; | |
1607 | int vif; | |
1608 | ||
1609 | if (nowait) { | |
1610 | read_unlock(&mrt_lock); | |
1611 | return -EAGAIN; | |
1612 | } | |
1613 | ||
1614 | dev = skb->dev; | |
1615 | if (dev == NULL || (vif = ip6mr_find_vif(dev)) < 0) { | |
1616 | read_unlock(&mrt_lock); | |
1617 | return -ENODEV; | |
1618 | } | |
1619 | ||
1620 | /* really correct? */ | |
1621 | skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); | |
1622 | if (!skb2) { | |
1623 | read_unlock(&mrt_lock); | |
1624 | return -ENOMEM; | |
1625 | } | |
1626 | ||
1627 | skb_reset_transport_header(skb2); | |
1628 | ||
1629 | skb_put(skb2, sizeof(struct ipv6hdr)); | |
1630 | skb_reset_network_header(skb2); | |
1631 | ||
1632 | iph = ipv6_hdr(skb2); | |
1633 | iph->version = 0; | |
1634 | iph->priority = 0; | |
1635 | iph->flow_lbl[0] = 0; | |
1636 | iph->flow_lbl[1] = 0; | |
1637 | iph->flow_lbl[2] = 0; | |
1638 | iph->payload_len = 0; | |
1639 | iph->nexthdr = IPPROTO_NONE; | |
1640 | iph->hop_limit = 0; | |
1641 | ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr); | |
1642 | ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr); | |
1643 | ||
1644 | err = ip6mr_cache_unresolved(vif, skb2); | |
1645 | read_unlock(&mrt_lock); | |
1646 | ||
1647 | return err; | |
1648 | } | |
1649 | ||
1650 | if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY)) | |
1651 | cache->mfc_flags |= MFC_NOTIFY; | |
1652 | ||
1653 | err = ip6mr_fill_mroute(skb, cache, rtm); | |
1654 | read_unlock(&mrt_lock); | |
1655 | return err; | |
1656 | } | |
1657 |