]>
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> | |
7bc570c8 YH |
30 | #include <linux/inet.h> |
31 | #include <linux/netdevice.h> | |
32 | #include <linux/inetdevice.h> | |
7bc570c8 YH |
33 | #include <linux/proc_fs.h> |
34 | #include <linux/seq_file.h> | |
7bc570c8 | 35 | #include <linux/init.h> |
7bc570c8 YH |
36 | #include <net/protocol.h> |
37 | #include <linux/skbuff.h> | |
38 | #include <net/sock.h> | |
7bc570c8 | 39 | #include <net/raw.h> |
7bc570c8 YH |
40 | #include <linux/notifier.h> |
41 | #include <linux/if_arp.h> | |
7bc570c8 YH |
42 | #include <net/checksum.h> |
43 | #include <net/netlink.h> | |
44 | ||
45 | #include <net/ipv6.h> | |
46 | #include <net/ip6_route.h> | |
47 | #include <linux/mroute6.h> | |
14fb64e1 | 48 | #include <linux/pim.h> |
7bc570c8 YH |
49 | #include <net/addrconf.h> |
50 | #include <linux/netfilter_ipv6.h> | |
51 | ||
7bc570c8 YH |
52 | /* Big lock, protecting vif table, mrt cache and mroute socket state. |
53 | Note that the changes are semaphored via rtnl_lock. | |
54 | */ | |
55 | ||
56 | static DEFINE_RWLOCK(mrt_lock); | |
57 | ||
58 | /* | |
59 | * Multicast router control variables | |
60 | */ | |
61 | ||
4e16880c | 62 | #define MIF_EXISTS(_net, _idx) ((_net)->ipv6.vif6_table[_idx].dev != NULL) |
7bc570c8 | 63 | |
14fb64e1 YH |
64 | static int mroute_do_assert; /* Set in PIM assert */ |
65 | #ifdef CONFIG_IPV6_PIMSM_V2 | |
66 | static int mroute_do_pim; | |
67 | #else | |
68 | #define mroute_do_pim 0 | |
69 | #endif | |
70 | ||
6ac7eb08 | 71 | static struct mfc6_cache *mfc6_cache_array[MFC6_LINES]; /* Forwarding cache */ |
7bc570c8 YH |
72 | |
73 | static struct mfc6_cache *mfc_unres_queue; /* Queue of unresolved entries */ | |
74 | static atomic_t cache_resolve_queue_len; /* Size of unresolved */ | |
75 | ||
76 | /* Special spinlock for queue of unresolved entries */ | |
77 | static DEFINE_SPINLOCK(mfc_unres_lock); | |
78 | ||
79 | /* We return to original Alan's scheme. Hash table of resolved | |
80 | entries is changed only in process context and protected | |
81 | with weak lock mrt_lock. Queue of unresolved entries is protected | |
82 | with strong spinlock mfc_unres_lock. | |
83 | ||
84 | In this case data path is free of exclusive locks at all. | |
85 | */ | |
86 | ||
87 | static struct kmem_cache *mrt_cachep __read_mostly; | |
88 | ||
89 | static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache); | |
6ac7eb08 | 90 | static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert); |
7bc570c8 YH |
91 | static int ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm); |
92 | ||
14fb64e1 YH |
93 | #ifdef CONFIG_IPV6_PIMSM_V2 |
94 | static struct inet6_protocol pim6_protocol; | |
95 | #endif | |
96 | ||
7bc570c8 YH |
97 | static struct timer_list ipmr_expire_timer; |
98 | ||
99 | ||
100 | #ifdef CONFIG_PROC_FS | |
101 | ||
102 | struct ipmr_mfc_iter { | |
103 | struct mfc6_cache **cache; | |
104 | int ct; | |
105 | }; | |
106 | ||
107 | ||
108 | static struct mfc6_cache *ipmr_mfc_seq_idx(struct ipmr_mfc_iter *it, loff_t pos) | |
109 | { | |
110 | struct mfc6_cache *mfc; | |
111 | ||
112 | it->cache = mfc6_cache_array; | |
113 | read_lock(&mrt_lock); | |
114 | for (it->ct = 0; it->ct < ARRAY_SIZE(mfc6_cache_array); it->ct++) | |
115 | for (mfc = mfc6_cache_array[it->ct]; mfc; mfc = mfc->next) | |
116 | if (pos-- == 0) | |
117 | return mfc; | |
118 | read_unlock(&mrt_lock); | |
119 | ||
120 | it->cache = &mfc_unres_queue; | |
121 | spin_lock_bh(&mfc_unres_lock); | |
122 | for (mfc = mfc_unres_queue; mfc; mfc = mfc->next) | |
123 | if (pos-- == 0) | |
124 | return mfc; | |
125 | spin_unlock_bh(&mfc_unres_lock); | |
126 | ||
127 | it->cache = NULL; | |
128 | return NULL; | |
129 | } | |
130 | ||
131 | ||
132 | ||
133 | ||
134 | /* | |
135 | * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif | |
136 | */ | |
137 | ||
138 | struct ipmr_vif_iter { | |
139 | int ct; | |
140 | }; | |
141 | ||
142 | static struct mif_device *ip6mr_vif_seq_idx(struct ipmr_vif_iter *iter, | |
143 | loff_t pos) | |
144 | { | |
4e16880c BT |
145 | for (iter->ct = 0; iter->ct < init_net.ipv6.maxvif; ++iter->ct) { |
146 | if (!MIF_EXISTS(&init_net, iter->ct)) | |
7bc570c8 YH |
147 | continue; |
148 | if (pos-- == 0) | |
4e16880c | 149 | return &init_net.ipv6.vif6_table[iter->ct]; |
7bc570c8 YH |
150 | } |
151 | return NULL; | |
152 | } | |
153 | ||
154 | static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos) | |
155 | __acquires(mrt_lock) | |
156 | { | |
157 | read_lock(&mrt_lock); | |
158 | return (*pos ? ip6mr_vif_seq_idx(seq->private, *pos - 1) | |
159 | : SEQ_START_TOKEN); | |
160 | } | |
161 | ||
162 | static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
163 | { | |
164 | struct ipmr_vif_iter *iter = seq->private; | |
165 | ||
166 | ++*pos; | |
167 | if (v == SEQ_START_TOKEN) | |
168 | return ip6mr_vif_seq_idx(iter, 0); | |
169 | ||
4e16880c BT |
170 | while (++iter->ct < init_net.ipv6.maxvif) { |
171 | if (!MIF_EXISTS(&init_net, iter->ct)) | |
7bc570c8 | 172 | continue; |
4e16880c | 173 | return &init_net.ipv6.vif6_table[iter->ct]; |
7bc570c8 YH |
174 | } |
175 | return NULL; | |
176 | } | |
177 | ||
178 | static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v) | |
179 | __releases(mrt_lock) | |
180 | { | |
181 | read_unlock(&mrt_lock); | |
182 | } | |
183 | ||
184 | static int ip6mr_vif_seq_show(struct seq_file *seq, void *v) | |
185 | { | |
186 | if (v == SEQ_START_TOKEN) { | |
187 | seq_puts(seq, | |
188 | "Interface BytesIn PktsIn BytesOut PktsOut Flags\n"); | |
189 | } else { | |
190 | const struct mif_device *vif = v; | |
191 | const char *name = vif->dev ? vif->dev->name : "none"; | |
192 | ||
193 | seq_printf(seq, | |
d430a227 | 194 | "%2td %-10s %8ld %7ld %8ld %7ld %05X\n", |
4e16880c | 195 | vif - init_net.ipv6.vif6_table, |
7bc570c8 YH |
196 | name, vif->bytes_in, vif->pkt_in, |
197 | vif->bytes_out, vif->pkt_out, | |
198 | vif->flags); | |
199 | } | |
200 | return 0; | |
201 | } | |
202 | ||
203 | static struct seq_operations ip6mr_vif_seq_ops = { | |
204 | .start = ip6mr_vif_seq_start, | |
205 | .next = ip6mr_vif_seq_next, | |
206 | .stop = ip6mr_vif_seq_stop, | |
207 | .show = ip6mr_vif_seq_show, | |
208 | }; | |
209 | ||
210 | static int ip6mr_vif_open(struct inode *inode, struct file *file) | |
211 | { | |
212 | return seq_open_private(file, &ip6mr_vif_seq_ops, | |
213 | sizeof(struct ipmr_vif_iter)); | |
214 | } | |
215 | ||
216 | static struct file_operations ip6mr_vif_fops = { | |
217 | .owner = THIS_MODULE, | |
218 | .open = ip6mr_vif_open, | |
219 | .read = seq_read, | |
220 | .llseek = seq_lseek, | |
eedd726e | 221 | .release = seq_release_private, |
7bc570c8 YH |
222 | }; |
223 | ||
224 | static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) | |
225 | { | |
226 | return (*pos ? ipmr_mfc_seq_idx(seq->private, *pos - 1) | |
227 | : SEQ_START_TOKEN); | |
228 | } | |
229 | ||
230 | static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
231 | { | |
232 | struct mfc6_cache *mfc = v; | |
233 | struct ipmr_mfc_iter *it = seq->private; | |
234 | ||
235 | ++*pos; | |
236 | ||
237 | if (v == SEQ_START_TOKEN) | |
238 | return ipmr_mfc_seq_idx(seq->private, 0); | |
239 | ||
240 | if (mfc->next) | |
241 | return mfc->next; | |
242 | ||
243 | if (it->cache == &mfc_unres_queue) | |
244 | goto end_of_list; | |
245 | ||
246 | BUG_ON(it->cache != mfc6_cache_array); | |
247 | ||
248 | while (++it->ct < ARRAY_SIZE(mfc6_cache_array)) { | |
249 | mfc = mfc6_cache_array[it->ct]; | |
250 | if (mfc) | |
251 | return mfc; | |
252 | } | |
253 | ||
254 | /* exhausted cache_array, show unresolved */ | |
255 | read_unlock(&mrt_lock); | |
256 | it->cache = &mfc_unres_queue; | |
257 | it->ct = 0; | |
258 | ||
259 | spin_lock_bh(&mfc_unres_lock); | |
260 | mfc = mfc_unres_queue; | |
261 | if (mfc) | |
262 | return mfc; | |
263 | ||
264 | end_of_list: | |
265 | spin_unlock_bh(&mfc_unres_lock); | |
266 | it->cache = NULL; | |
267 | ||
268 | return NULL; | |
269 | } | |
270 | ||
271 | static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) | |
272 | { | |
273 | struct ipmr_mfc_iter *it = seq->private; | |
274 | ||
275 | if (it->cache == &mfc_unres_queue) | |
276 | spin_unlock_bh(&mfc_unres_lock); | |
277 | else if (it->cache == mfc6_cache_array) | |
278 | read_unlock(&mrt_lock); | |
279 | } | |
280 | ||
281 | static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) | |
282 | { | |
283 | int n; | |
284 | ||
285 | if (v == SEQ_START_TOKEN) { | |
286 | seq_puts(seq, | |
287 | "Group " | |
288 | "Origin " | |
289 | "Iif Pkts Bytes Wrong Oifs\n"); | |
290 | } else { | |
291 | const struct mfc6_cache *mfc = v; | |
292 | const struct ipmr_mfc_iter *it = seq->private; | |
293 | ||
999890b2 | 294 | seq_printf(seq, "%pI6 %pI6 %-3hd", |
0c6ce78a | 295 | &mfc->mf6c_mcastgrp, &mfc->mf6c_origin, |
1ea472e2 | 296 | mfc->mf6c_parent); |
7bc570c8 YH |
297 | |
298 | if (it->cache != &mfc_unres_queue) { | |
1ea472e2 BT |
299 | seq_printf(seq, " %8lu %8lu %8lu", |
300 | mfc->mfc_un.res.pkt, | |
301 | mfc->mfc_un.res.bytes, | |
302 | mfc->mfc_un.res.wrong_if); | |
7bc570c8 YH |
303 | for (n = mfc->mfc_un.res.minvif; |
304 | n < mfc->mfc_un.res.maxvif; n++) { | |
4e16880c | 305 | if (MIF_EXISTS(&init_net, n) && |
7bc570c8 YH |
306 | mfc->mfc_un.res.ttls[n] < 255) |
307 | seq_printf(seq, | |
308 | " %2d:%-3d", | |
309 | n, mfc->mfc_un.res.ttls[n]); | |
310 | } | |
1ea472e2 BT |
311 | } else { |
312 | /* unresolved mfc_caches don't contain | |
313 | * pkt, bytes and wrong_if values | |
314 | */ | |
315 | seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); | |
7bc570c8 YH |
316 | } |
317 | seq_putc(seq, '\n'); | |
318 | } | |
319 | return 0; | |
320 | } | |
321 | ||
322 | static struct seq_operations ipmr_mfc_seq_ops = { | |
323 | .start = ipmr_mfc_seq_start, | |
324 | .next = ipmr_mfc_seq_next, | |
325 | .stop = ipmr_mfc_seq_stop, | |
326 | .show = ipmr_mfc_seq_show, | |
327 | }; | |
328 | ||
329 | static int ipmr_mfc_open(struct inode *inode, struct file *file) | |
330 | { | |
331 | return seq_open_private(file, &ipmr_mfc_seq_ops, | |
332 | sizeof(struct ipmr_mfc_iter)); | |
333 | } | |
334 | ||
335 | static struct file_operations ip6mr_mfc_fops = { | |
336 | .owner = THIS_MODULE, | |
337 | .open = ipmr_mfc_open, | |
338 | .read = seq_read, | |
339 | .llseek = seq_lseek, | |
eedd726e | 340 | .release = seq_release_private, |
7bc570c8 YH |
341 | }; |
342 | #endif | |
343 | ||
14fb64e1 YH |
344 | #ifdef CONFIG_IPV6_PIMSM_V2 |
345 | static int reg_vif_num = -1; | |
346 | ||
347 | static int pim6_rcv(struct sk_buff *skb) | |
348 | { | |
349 | struct pimreghdr *pim; | |
350 | struct ipv6hdr *encap; | |
351 | struct net_device *reg_dev = NULL; | |
352 | ||
353 | if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap))) | |
354 | goto drop; | |
355 | ||
356 | pim = (struct pimreghdr *)skb_transport_header(skb); | |
357 | if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) || | |
358 | (pim->flags & PIM_NULL_REGISTER) || | |
359 | (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && | |
ec6b486f | 360 | csum_fold(skb_checksum(skb, 0, skb->len, 0)))) |
14fb64e1 YH |
361 | goto drop; |
362 | ||
363 | /* check if the inner packet is destined to mcast group */ | |
364 | encap = (struct ipv6hdr *)(skb_transport_header(skb) + | |
365 | sizeof(*pim)); | |
366 | ||
367 | if (!ipv6_addr_is_multicast(&encap->daddr) || | |
368 | encap->payload_len == 0 || | |
369 | ntohs(encap->payload_len) + sizeof(*pim) > skb->len) | |
370 | goto drop; | |
371 | ||
372 | read_lock(&mrt_lock); | |
373 | if (reg_vif_num >= 0) | |
4e16880c | 374 | reg_dev = init_net.ipv6.vif6_table[reg_vif_num].dev; |
14fb64e1 YH |
375 | if (reg_dev) |
376 | dev_hold(reg_dev); | |
377 | read_unlock(&mrt_lock); | |
378 | ||
379 | if (reg_dev == NULL) | |
380 | goto drop; | |
381 | ||
382 | skb->mac_header = skb->network_header; | |
383 | skb_pull(skb, (u8 *)encap - skb->data); | |
384 | skb_reset_network_header(skb); | |
385 | skb->dev = reg_dev; | |
386 | skb->protocol = htons(ETH_P_IP); | |
387 | skb->ip_summed = 0; | |
388 | skb->pkt_type = PACKET_HOST; | |
389 | dst_release(skb->dst); | |
dc58c78c PE |
390 | reg_dev->stats.rx_bytes += skb->len; |
391 | reg_dev->stats.rx_packets++; | |
14fb64e1 YH |
392 | skb->dst = NULL; |
393 | nf_reset(skb); | |
394 | netif_rx(skb); | |
395 | dev_put(reg_dev); | |
396 | return 0; | |
397 | drop: | |
398 | kfree_skb(skb); | |
399 | return 0; | |
400 | } | |
401 | ||
402 | static struct inet6_protocol pim6_protocol = { | |
403 | .handler = pim6_rcv, | |
404 | }; | |
405 | ||
406 | /* Service routines creating virtual interfaces: PIMREG */ | |
407 | ||
408 | static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) | |
409 | { | |
410 | read_lock(&mrt_lock); | |
dc58c78c PE |
411 | dev->stats.tx_bytes += skb->len; |
412 | dev->stats.tx_packets++; | |
14fb64e1 YH |
413 | ip6mr_cache_report(skb, reg_vif_num, MRT6MSG_WHOLEPKT); |
414 | read_unlock(&mrt_lock); | |
415 | kfree_skb(skb); | |
416 | return 0; | |
417 | } | |
418 | ||
007c3838 SH |
419 | static const struct net_device_ops reg_vif_netdev_ops = { |
420 | .ndo_start_xmit = reg_vif_xmit, | |
421 | }; | |
422 | ||
14fb64e1 YH |
423 | static void reg_vif_setup(struct net_device *dev) |
424 | { | |
425 | dev->type = ARPHRD_PIMREG; | |
426 | dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8; | |
427 | dev->flags = IFF_NOARP; | |
007c3838 | 428 | dev->netdev_ops = ®_vif_netdev_ops; |
14fb64e1 YH |
429 | dev->destructor = free_netdev; |
430 | } | |
431 | ||
432 | static struct net_device *ip6mr_reg_vif(void) | |
433 | { | |
434 | struct net_device *dev; | |
14fb64e1 | 435 | |
dc58c78c | 436 | dev = alloc_netdev(0, "pim6reg", reg_vif_setup); |
14fb64e1 YH |
437 | if (dev == NULL) |
438 | return NULL; | |
439 | ||
440 | if (register_netdevice(dev)) { | |
441 | free_netdev(dev); | |
442 | return NULL; | |
443 | } | |
444 | dev->iflink = 0; | |
445 | ||
14fb64e1 YH |
446 | if (dev_open(dev)) |
447 | goto failure; | |
448 | ||
7af3db78 | 449 | dev_hold(dev); |
14fb64e1 YH |
450 | return dev; |
451 | ||
452 | failure: | |
453 | /* allow the register to be completed before unregistering. */ | |
454 | rtnl_unlock(); | |
455 | rtnl_lock(); | |
456 | ||
457 | unregister_netdevice(dev); | |
458 | return NULL; | |
459 | } | |
460 | #endif | |
461 | ||
7bc570c8 YH |
462 | /* |
463 | * Delete a VIF entry | |
464 | */ | |
465 | ||
466 | static int mif6_delete(int vifi) | |
467 | { | |
468 | struct mif_device *v; | |
469 | struct net_device *dev; | |
4e16880c | 470 | if (vifi < 0 || vifi >= init_net.ipv6.maxvif) |
7bc570c8 YH |
471 | return -EADDRNOTAVAIL; |
472 | ||
4e16880c | 473 | v = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 YH |
474 | |
475 | write_lock_bh(&mrt_lock); | |
476 | dev = v->dev; | |
477 | v->dev = NULL; | |
478 | ||
479 | if (!dev) { | |
480 | write_unlock_bh(&mrt_lock); | |
481 | return -EADDRNOTAVAIL; | |
482 | } | |
483 | ||
14fb64e1 YH |
484 | #ifdef CONFIG_IPV6_PIMSM_V2 |
485 | if (vifi == reg_vif_num) | |
486 | reg_vif_num = -1; | |
487 | #endif | |
488 | ||
4e16880c | 489 | if (vifi + 1 == init_net.ipv6.maxvif) { |
7bc570c8 YH |
490 | int tmp; |
491 | for (tmp = vifi - 1; tmp >= 0; tmp--) { | |
4e16880c | 492 | if (MIF_EXISTS(&init_net, tmp)) |
7bc570c8 YH |
493 | break; |
494 | } | |
4e16880c | 495 | init_net.ipv6.maxvif = tmp + 1; |
7bc570c8 YH |
496 | } |
497 | ||
498 | write_unlock_bh(&mrt_lock); | |
499 | ||
500 | dev_set_allmulti(dev, -1); | |
501 | ||
502 | if (v->flags & MIFF_REGISTER) | |
503 | unregister_netdevice(dev); | |
504 | ||
505 | dev_put(dev); | |
506 | return 0; | |
507 | } | |
508 | ||
58701ad4 BT |
509 | static inline void ip6mr_cache_free(struct mfc6_cache *c) |
510 | { | |
511 | release_net(mfc6_net(c)); | |
512 | kmem_cache_free(mrt_cachep, c); | |
513 | } | |
514 | ||
7bc570c8 YH |
515 | /* Destroy an unresolved cache entry, killing queued skbs |
516 | and reporting error to netlink readers. | |
517 | */ | |
518 | ||
519 | static void ip6mr_destroy_unres(struct mfc6_cache *c) | |
520 | { | |
521 | struct sk_buff *skb; | |
522 | ||
523 | atomic_dec(&cache_resolve_queue_len); | |
524 | ||
525 | while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) { | |
526 | if (ipv6_hdr(skb)->version == 0) { | |
527 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); | |
528 | nlh->nlmsg_type = NLMSG_ERROR; | |
529 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
530 | skb_trim(skb, nlh->nlmsg_len); | |
531 | ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT; | |
532 | rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); | |
533 | } else | |
534 | kfree_skb(skb); | |
535 | } | |
536 | ||
58701ad4 | 537 | ip6mr_cache_free(c); |
7bc570c8 YH |
538 | } |
539 | ||
540 | ||
541 | /* Single timer process for all the unresolved queue. */ | |
542 | ||
543 | static void ipmr_do_expire_process(unsigned long dummy) | |
544 | { | |
545 | unsigned long now = jiffies; | |
546 | unsigned long expires = 10 * HZ; | |
547 | struct mfc6_cache *c, **cp; | |
548 | ||
549 | cp = &mfc_unres_queue; | |
550 | ||
551 | while ((c = *cp) != NULL) { | |
552 | if (time_after(c->mfc_un.unres.expires, now)) { | |
553 | /* not yet... */ | |
554 | unsigned long interval = c->mfc_un.unres.expires - now; | |
555 | if (interval < expires) | |
556 | expires = interval; | |
557 | cp = &c->next; | |
558 | continue; | |
559 | } | |
560 | ||
561 | *cp = c->next; | |
562 | ip6mr_destroy_unres(c); | |
563 | } | |
564 | ||
565 | if (atomic_read(&cache_resolve_queue_len)) | |
566 | mod_timer(&ipmr_expire_timer, jiffies + expires); | |
567 | } | |
568 | ||
569 | static void ipmr_expire_process(unsigned long dummy) | |
570 | { | |
571 | if (!spin_trylock(&mfc_unres_lock)) { | |
572 | mod_timer(&ipmr_expire_timer, jiffies + 1); | |
573 | return; | |
574 | } | |
575 | ||
576 | if (atomic_read(&cache_resolve_queue_len)) | |
577 | ipmr_do_expire_process(dummy); | |
578 | ||
579 | spin_unlock(&mfc_unres_lock); | |
580 | } | |
581 | ||
582 | /* Fill oifs list. It is called under write locked mrt_lock. */ | |
583 | ||
584 | static void ip6mr_update_thresholds(struct mfc6_cache *cache, unsigned char *ttls) | |
585 | { | |
586 | int vifi; | |
587 | ||
6ac7eb08 | 588 | cache->mfc_un.res.minvif = MAXMIFS; |
7bc570c8 | 589 | cache->mfc_un.res.maxvif = 0; |
6ac7eb08 | 590 | memset(cache->mfc_un.res.ttls, 255, MAXMIFS); |
7bc570c8 | 591 | |
4e16880c BT |
592 | for (vifi = 0; vifi < init_net.ipv6.maxvif; vifi++) { |
593 | if (MIF_EXISTS(&init_net, vifi) && | |
594 | ttls[vifi] && ttls[vifi] < 255) { | |
7bc570c8 YH |
595 | cache->mfc_un.res.ttls[vifi] = ttls[vifi]; |
596 | if (cache->mfc_un.res.minvif > vifi) | |
597 | cache->mfc_un.res.minvif = vifi; | |
598 | if (cache->mfc_un.res.maxvif <= vifi) | |
599 | cache->mfc_un.res.maxvif = vifi + 1; | |
600 | } | |
601 | } | |
602 | } | |
603 | ||
604 | static int mif6_add(struct mif6ctl *vifc, int mrtsock) | |
605 | { | |
606 | int vifi = vifc->mif6c_mifi; | |
4e16880c | 607 | struct mif_device *v = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 | 608 | struct net_device *dev; |
5ae7b444 | 609 | int err; |
7bc570c8 YH |
610 | |
611 | /* Is vif busy ? */ | |
4e16880c | 612 | if (MIF_EXISTS(&init_net, vifi)) |
7bc570c8 YH |
613 | return -EADDRINUSE; |
614 | ||
615 | switch (vifc->mif6c_flags) { | |
14fb64e1 YH |
616 | #ifdef CONFIG_IPV6_PIMSM_V2 |
617 | case MIFF_REGISTER: | |
618 | /* | |
619 | * Special Purpose VIF in PIM | |
620 | * All the packets will be sent to the daemon | |
621 | */ | |
622 | if (reg_vif_num >= 0) | |
623 | return -EADDRINUSE; | |
624 | dev = ip6mr_reg_vif(); | |
625 | if (!dev) | |
626 | return -ENOBUFS; | |
5ae7b444 WC |
627 | err = dev_set_allmulti(dev, 1); |
628 | if (err) { | |
629 | unregister_netdevice(dev); | |
7af3db78 | 630 | dev_put(dev); |
5ae7b444 WC |
631 | return err; |
632 | } | |
14fb64e1 YH |
633 | break; |
634 | #endif | |
7bc570c8 YH |
635 | case 0: |
636 | dev = dev_get_by_index(&init_net, vifc->mif6c_pifi); | |
637 | if (!dev) | |
638 | return -EADDRNOTAVAIL; | |
5ae7b444 | 639 | err = dev_set_allmulti(dev, 1); |
7af3db78 WC |
640 | if (err) { |
641 | dev_put(dev); | |
5ae7b444 | 642 | return err; |
7af3db78 | 643 | } |
7bc570c8 YH |
644 | break; |
645 | default: | |
646 | return -EINVAL; | |
647 | } | |
648 | ||
7bc570c8 YH |
649 | /* |
650 | * Fill in the VIF structures | |
651 | */ | |
652 | v->rate_limit = vifc->vifc_rate_limit; | |
653 | v->flags = vifc->mif6c_flags; | |
654 | if (!mrtsock) | |
655 | v->flags |= VIFF_STATIC; | |
656 | v->threshold = vifc->vifc_threshold; | |
657 | v->bytes_in = 0; | |
658 | v->bytes_out = 0; | |
659 | v->pkt_in = 0; | |
660 | v->pkt_out = 0; | |
661 | v->link = dev->ifindex; | |
662 | if (v->flags & MIFF_REGISTER) | |
663 | v->link = dev->iflink; | |
664 | ||
665 | /* And finish update writing critical data */ | |
666 | write_lock_bh(&mrt_lock); | |
7bc570c8 | 667 | v->dev = dev; |
14fb64e1 YH |
668 | #ifdef CONFIG_IPV6_PIMSM_V2 |
669 | if (v->flags & MIFF_REGISTER) | |
670 | reg_vif_num = vifi; | |
671 | #endif | |
4e16880c BT |
672 | if (vifi + 1 > init_net.ipv6.maxvif) |
673 | init_net.ipv6.maxvif = vifi + 1; | |
7bc570c8 YH |
674 | write_unlock_bh(&mrt_lock); |
675 | return 0; | |
676 | } | |
677 | ||
678 | static struct mfc6_cache *ip6mr_cache_find(struct in6_addr *origin, struct in6_addr *mcastgrp) | |
679 | { | |
680 | int line = MFC6_HASH(mcastgrp, origin); | |
681 | struct mfc6_cache *c; | |
682 | ||
683 | for (c = mfc6_cache_array[line]; c; c = c->next) { | |
684 | if (ipv6_addr_equal(&c->mf6c_origin, origin) && | |
685 | ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) | |
686 | break; | |
687 | } | |
688 | return c; | |
689 | } | |
690 | ||
691 | /* | |
692 | * Allocate a multicast cache entry | |
693 | */ | |
58701ad4 | 694 | static struct mfc6_cache *ip6mr_cache_alloc(struct net *net) |
7bc570c8 | 695 | { |
36cbac59 | 696 | struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); |
7bc570c8 YH |
697 | if (c == NULL) |
698 | return NULL; | |
6ac7eb08 | 699 | c->mfc_un.res.minvif = MAXMIFS; |
58701ad4 | 700 | mfc6_net_set(c, net); |
7bc570c8 YH |
701 | return c; |
702 | } | |
703 | ||
58701ad4 | 704 | static struct mfc6_cache *ip6mr_cache_alloc_unres(struct net *net) |
7bc570c8 | 705 | { |
36cbac59 | 706 | struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); |
7bc570c8 YH |
707 | if (c == NULL) |
708 | return NULL; | |
7bc570c8 YH |
709 | skb_queue_head_init(&c->mfc_un.unres.unresolved); |
710 | c->mfc_un.unres.expires = jiffies + 10 * HZ; | |
58701ad4 | 711 | mfc6_net_set(c, net); |
7bc570c8 YH |
712 | return c; |
713 | } | |
714 | ||
715 | /* | |
716 | * A cache entry has gone into a resolved state from queued | |
717 | */ | |
718 | ||
719 | static void ip6mr_cache_resolve(struct mfc6_cache *uc, struct mfc6_cache *c) | |
720 | { | |
721 | struct sk_buff *skb; | |
722 | ||
723 | /* | |
724 | * Play the pending entries through our router | |
725 | */ | |
726 | ||
727 | while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { | |
728 | if (ipv6_hdr(skb)->version == 0) { | |
729 | int err; | |
730 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); | |
731 | ||
732 | if (ip6mr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) { | |
549e028d | 733 | nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; |
7bc570c8 YH |
734 | } else { |
735 | nlh->nlmsg_type = NLMSG_ERROR; | |
736 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
737 | skb_trim(skb, nlh->nlmsg_len); | |
738 | ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE; | |
739 | } | |
740 | err = rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); | |
741 | } else | |
742 | ip6_mr_forward(skb, c); | |
743 | } | |
744 | } | |
745 | ||
746 | /* | |
747 | * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd | |
748 | * expects the following bizarre scheme. | |
749 | * | |
750 | * Called under mrt_lock. | |
751 | */ | |
752 | ||
6ac7eb08 | 753 | static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert) |
7bc570c8 YH |
754 | { |
755 | struct sk_buff *skb; | |
756 | struct mrt6msg *msg; | |
757 | int ret; | |
758 | ||
14fb64e1 YH |
759 | #ifdef CONFIG_IPV6_PIMSM_V2 |
760 | if (assert == MRT6MSG_WHOLEPKT) | |
761 | skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) | |
762 | +sizeof(*msg)); | |
763 | else | |
764 | #endif | |
765 | skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); | |
7bc570c8 YH |
766 | |
767 | if (!skb) | |
768 | return -ENOBUFS; | |
769 | ||
770 | /* I suppose that internal messages | |
771 | * do not require checksums */ | |
772 | ||
773 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
774 | ||
14fb64e1 YH |
775 | #ifdef CONFIG_IPV6_PIMSM_V2 |
776 | if (assert == MRT6MSG_WHOLEPKT) { | |
777 | /* Ugly, but we have no choice with this interface. | |
778 | Duplicate old header, fix length etc. | |
779 | And all this only to mangle msg->im6_msgtype and | |
780 | to set msg->im6_mbz to "mbz" :-) | |
781 | */ | |
782 | skb_push(skb, -skb_network_offset(pkt)); | |
783 | ||
784 | skb_push(skb, sizeof(*msg)); | |
785 | skb_reset_transport_header(skb); | |
786 | msg = (struct mrt6msg *)skb_transport_header(skb); | |
787 | msg->im6_mbz = 0; | |
788 | msg->im6_msgtype = MRT6MSG_WHOLEPKT; | |
789 | msg->im6_mif = reg_vif_num; | |
790 | msg->im6_pad = 0; | |
791 | ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr); | |
792 | ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr); | |
793 | ||
794 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
795 | } else | |
796 | #endif | |
797 | { | |
7bc570c8 YH |
798 | /* |
799 | * Copy the IP header | |
800 | */ | |
801 | ||
802 | skb_put(skb, sizeof(struct ipv6hdr)); | |
803 | skb_reset_network_header(skb); | |
804 | skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); | |
805 | ||
806 | /* | |
807 | * Add our header | |
808 | */ | |
809 | skb_put(skb, sizeof(*msg)); | |
810 | skb_reset_transport_header(skb); | |
811 | msg = (struct mrt6msg *)skb_transport_header(skb); | |
812 | ||
813 | msg->im6_mbz = 0; | |
814 | msg->im6_msgtype = assert; | |
6ac7eb08 | 815 | msg->im6_mif = mifi; |
7bc570c8 YH |
816 | msg->im6_pad = 0; |
817 | ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr); | |
818 | ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr); | |
819 | ||
820 | skb->dst = dst_clone(pkt->dst); | |
821 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
822 | ||
823 | skb_pull(skb, sizeof(struct ipv6hdr)); | |
14fb64e1 | 824 | } |
7bc570c8 | 825 | |
bd91b8bf | 826 | if (init_net.ipv6.mroute6_sk == NULL) { |
7bc570c8 YH |
827 | kfree_skb(skb); |
828 | return -EINVAL; | |
829 | } | |
830 | ||
831 | /* | |
832 | * Deliver to user space multicast routing algorithms | |
833 | */ | |
bd91b8bf BT |
834 | ret = sock_queue_rcv_skb(init_net.ipv6.mroute6_sk, skb); |
835 | if (ret < 0) { | |
7bc570c8 YH |
836 | if (net_ratelimit()) |
837 | printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n"); | |
838 | kfree_skb(skb); | |
839 | } | |
840 | ||
841 | return ret; | |
842 | } | |
843 | ||
844 | /* | |
845 | * Queue a packet for resolution. It gets locked cache entry! | |
846 | */ | |
847 | ||
848 | static int | |
6ac7eb08 | 849 | ip6mr_cache_unresolved(mifi_t mifi, struct sk_buff *skb) |
7bc570c8 YH |
850 | { |
851 | int err; | |
852 | struct mfc6_cache *c; | |
853 | ||
854 | spin_lock_bh(&mfc_unres_lock); | |
855 | for (c = mfc_unres_queue; c; c = c->next) { | |
856 | if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && | |
857 | ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) | |
858 | break; | |
859 | } | |
860 | ||
861 | if (c == NULL) { | |
862 | /* | |
863 | * Create a new entry if allowable | |
864 | */ | |
865 | ||
866 | if (atomic_read(&cache_resolve_queue_len) >= 10 || | |
58701ad4 | 867 | (c = ip6mr_cache_alloc_unres(&init_net)) == NULL) { |
7bc570c8 YH |
868 | spin_unlock_bh(&mfc_unres_lock); |
869 | ||
870 | kfree_skb(skb); | |
871 | return -ENOBUFS; | |
872 | } | |
873 | ||
874 | /* | |
875 | * Fill in the new cache entry | |
876 | */ | |
877 | c->mf6c_parent = -1; | |
878 | c->mf6c_origin = ipv6_hdr(skb)->saddr; | |
879 | c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; | |
880 | ||
881 | /* | |
882 | * Reflect first query at pim6sd | |
883 | */ | |
6ac7eb08 | 884 | if ((err = ip6mr_cache_report(skb, mifi, MRT6MSG_NOCACHE)) < 0) { |
7bc570c8 YH |
885 | /* If the report failed throw the cache entry |
886 | out - Brad Parker | |
887 | */ | |
888 | spin_unlock_bh(&mfc_unres_lock); | |
889 | ||
58701ad4 | 890 | ip6mr_cache_free(c); |
7bc570c8 YH |
891 | kfree_skb(skb); |
892 | return err; | |
893 | } | |
894 | ||
895 | atomic_inc(&cache_resolve_queue_len); | |
896 | c->next = mfc_unres_queue; | |
897 | mfc_unres_queue = c; | |
898 | ||
899 | ipmr_do_expire_process(1); | |
900 | } | |
901 | ||
902 | /* | |
903 | * See if we can append the packet | |
904 | */ | |
905 | if (c->mfc_un.unres.unresolved.qlen > 3) { | |
906 | kfree_skb(skb); | |
907 | err = -ENOBUFS; | |
908 | } else { | |
909 | skb_queue_tail(&c->mfc_un.unres.unresolved, skb); | |
910 | err = 0; | |
911 | } | |
912 | ||
913 | spin_unlock_bh(&mfc_unres_lock); | |
914 | return err; | |
915 | } | |
916 | ||
917 | /* | |
918 | * MFC6 cache manipulation by user space | |
919 | */ | |
920 | ||
921 | static int ip6mr_mfc_delete(struct mf6cctl *mfc) | |
922 | { | |
923 | int line; | |
924 | struct mfc6_cache *c, **cp; | |
925 | ||
926 | line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); | |
927 | ||
928 | for (cp = &mfc6_cache_array[line]; (c = *cp) != NULL; cp = &c->next) { | |
929 | if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && | |
930 | ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) { | |
931 | write_lock_bh(&mrt_lock); | |
932 | *cp = c->next; | |
933 | write_unlock_bh(&mrt_lock); | |
934 | ||
58701ad4 | 935 | ip6mr_cache_free(c); |
7bc570c8 YH |
936 | return 0; |
937 | } | |
938 | } | |
939 | return -ENOENT; | |
940 | } | |
941 | ||
942 | static int ip6mr_device_event(struct notifier_block *this, | |
943 | unsigned long event, void *ptr) | |
944 | { | |
945 | struct net_device *dev = ptr; | |
946 | struct mif_device *v; | |
947 | int ct; | |
948 | ||
721499e8 | 949 | if (!net_eq(dev_net(dev), &init_net)) |
7bc570c8 YH |
950 | return NOTIFY_DONE; |
951 | ||
952 | if (event != NETDEV_UNREGISTER) | |
953 | return NOTIFY_DONE; | |
954 | ||
4e16880c BT |
955 | v = &init_net.ipv6.vif6_table[0]; |
956 | for (ct = 0; ct < init_net.ipv6.maxvif; ct++, v++) { | |
7bc570c8 YH |
957 | if (v->dev == dev) |
958 | mif6_delete(ct); | |
959 | } | |
960 | return NOTIFY_DONE; | |
961 | } | |
962 | ||
963 | static struct notifier_block ip6_mr_notifier = { | |
964 | .notifier_call = ip6mr_device_event | |
965 | }; | |
966 | ||
967 | /* | |
968 | * Setup for IP multicast routing | |
969 | */ | |
970 | ||
4e16880c BT |
971 | static int __net_init ip6mr_net_init(struct net *net) |
972 | { | |
973 | int err = 0; | |
974 | ||
975 | net->ipv6.vif6_table = kcalloc(MAXMIFS, sizeof(struct mif_device), | |
976 | GFP_KERNEL); | |
977 | if (!net->ipv6.vif6_table) { | |
978 | err = -ENOMEM; | |
979 | goto fail; | |
980 | } | |
981 | fail: | |
982 | return err; | |
983 | } | |
984 | ||
985 | static void __net_exit ip6mr_net_exit(struct net *net) | |
986 | { | |
987 | kfree(net->ipv6.vif6_table); | |
988 | } | |
989 | ||
990 | static struct pernet_operations ip6mr_net_ops = { | |
991 | .init = ip6mr_net_init, | |
992 | .exit = ip6mr_net_exit, | |
993 | }; | |
994 | ||
623d1a1a | 995 | int __init ip6_mr_init(void) |
7bc570c8 | 996 | { |
623d1a1a WC |
997 | int err; |
998 | ||
7bc570c8 YH |
999 | mrt_cachep = kmem_cache_create("ip6_mrt_cache", |
1000 | sizeof(struct mfc6_cache), | |
1001 | 0, SLAB_HWCACHE_ALIGN, | |
1002 | NULL); | |
1003 | if (!mrt_cachep) | |
623d1a1a | 1004 | return -ENOMEM; |
7bc570c8 | 1005 | |
4e16880c BT |
1006 | err = register_pernet_subsys(&ip6mr_net_ops); |
1007 | if (err) | |
1008 | goto reg_pernet_fail; | |
1009 | ||
7bc570c8 | 1010 | setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0); |
623d1a1a WC |
1011 | err = register_netdevice_notifier(&ip6_mr_notifier); |
1012 | if (err) | |
1013 | goto reg_notif_fail; | |
1014 | #ifdef CONFIG_PROC_FS | |
1015 | err = -ENOMEM; | |
1016 | if (!proc_net_fops_create(&init_net, "ip6_mr_vif", 0, &ip6mr_vif_fops)) | |
1017 | goto proc_vif_fail; | |
1018 | if (!proc_net_fops_create(&init_net, "ip6_mr_cache", | |
1019 | 0, &ip6mr_mfc_fops)) | |
1020 | goto proc_cache_fail; | |
1021 | #endif | |
1022 | return 0; | |
7bc570c8 | 1023 | #ifdef CONFIG_PROC_FS |
623d1a1a WC |
1024 | proc_cache_fail: |
1025 | proc_net_remove(&init_net, "ip6_mr_vif"); | |
87b30a65 BT |
1026 | proc_vif_fail: |
1027 | unregister_netdevice_notifier(&ip6_mr_notifier); | |
7bc570c8 | 1028 | #endif |
87b30a65 BT |
1029 | reg_notif_fail: |
1030 | del_timer(&ipmr_expire_timer); | |
4e16880c BT |
1031 | unregister_pernet_subsys(&ip6mr_net_ops); |
1032 | reg_pernet_fail: | |
87b30a65 | 1033 | kmem_cache_destroy(mrt_cachep); |
623d1a1a | 1034 | return err; |
7bc570c8 YH |
1035 | } |
1036 | ||
623d1a1a WC |
1037 | void ip6_mr_cleanup(void) |
1038 | { | |
1039 | #ifdef CONFIG_PROC_FS | |
1040 | proc_net_remove(&init_net, "ip6_mr_cache"); | |
1041 | proc_net_remove(&init_net, "ip6_mr_vif"); | |
1042 | #endif | |
1043 | unregister_netdevice_notifier(&ip6_mr_notifier); | |
1044 | del_timer(&ipmr_expire_timer); | |
4e16880c | 1045 | unregister_pernet_subsys(&ip6mr_net_ops); |
623d1a1a WC |
1046 | kmem_cache_destroy(mrt_cachep); |
1047 | } | |
7bc570c8 YH |
1048 | |
1049 | static int ip6mr_mfc_add(struct mf6cctl *mfc, int mrtsock) | |
1050 | { | |
1051 | int line; | |
1052 | struct mfc6_cache *uc, *c, **cp; | |
6ac7eb08 | 1053 | unsigned char ttls[MAXMIFS]; |
7bc570c8 YH |
1054 | int i; |
1055 | ||
6ac7eb08 RR |
1056 | memset(ttls, 255, MAXMIFS); |
1057 | for (i = 0; i < MAXMIFS; i++) { | |
7bc570c8 YH |
1058 | if (IF_ISSET(i, &mfc->mf6cc_ifset)) |
1059 | ttls[i] = 1; | |
1060 | ||
1061 | } | |
1062 | ||
1063 | line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); | |
1064 | ||
1065 | for (cp = &mfc6_cache_array[line]; (c = *cp) != NULL; cp = &c->next) { | |
1066 | if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && | |
1067 | ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) | |
1068 | break; | |
1069 | } | |
1070 | ||
1071 | if (c != NULL) { | |
1072 | write_lock_bh(&mrt_lock); | |
1073 | c->mf6c_parent = mfc->mf6cc_parent; | |
1074 | ip6mr_update_thresholds(c, ttls); | |
1075 | if (!mrtsock) | |
1076 | c->mfc_flags |= MFC_STATIC; | |
1077 | write_unlock_bh(&mrt_lock); | |
1078 | return 0; | |
1079 | } | |
1080 | ||
1081 | if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) | |
1082 | return -EINVAL; | |
1083 | ||
58701ad4 | 1084 | c = ip6mr_cache_alloc(&init_net); |
7bc570c8 YH |
1085 | if (c == NULL) |
1086 | return -ENOMEM; | |
1087 | ||
1088 | c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; | |
1089 | c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; | |
1090 | c->mf6c_parent = mfc->mf6cc_parent; | |
1091 | ip6mr_update_thresholds(c, ttls); | |
1092 | if (!mrtsock) | |
1093 | c->mfc_flags |= MFC_STATIC; | |
1094 | ||
1095 | write_lock_bh(&mrt_lock); | |
1096 | c->next = mfc6_cache_array[line]; | |
1097 | mfc6_cache_array[line] = c; | |
1098 | write_unlock_bh(&mrt_lock); | |
1099 | ||
1100 | /* | |
1101 | * Check to see if we resolved a queued list. If so we | |
1102 | * need to send on the frames and tidy up. | |
1103 | */ | |
1104 | spin_lock_bh(&mfc_unres_lock); | |
1105 | for (cp = &mfc_unres_queue; (uc = *cp) != NULL; | |
1106 | cp = &uc->next) { | |
1107 | if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && | |
1108 | ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { | |
1109 | *cp = uc->next; | |
1110 | if (atomic_dec_and_test(&cache_resolve_queue_len)) | |
1111 | del_timer(&ipmr_expire_timer); | |
1112 | break; | |
1113 | } | |
1114 | } | |
1115 | spin_unlock_bh(&mfc_unres_lock); | |
1116 | ||
1117 | if (uc) { | |
1118 | ip6mr_cache_resolve(uc, c); | |
58701ad4 | 1119 | ip6mr_cache_free(uc); |
7bc570c8 YH |
1120 | } |
1121 | return 0; | |
1122 | } | |
1123 | ||
1124 | /* | |
1125 | * Close the multicast socket, and clear the vif tables etc | |
1126 | */ | |
1127 | ||
1128 | static void mroute_clean_tables(struct sock *sk) | |
1129 | { | |
1130 | int i; | |
1131 | ||
1132 | /* | |
1133 | * Shut down all active vif entries | |
1134 | */ | |
4e16880c BT |
1135 | for (i = 0; i < init_net.ipv6.maxvif; i++) { |
1136 | if (!(init_net.ipv6.vif6_table[i].flags & VIFF_STATIC)) | |
7bc570c8 YH |
1137 | mif6_delete(i); |
1138 | } | |
1139 | ||
1140 | /* | |
1141 | * Wipe the cache | |
1142 | */ | |
1143 | for (i = 0; i < ARRAY_SIZE(mfc6_cache_array); i++) { | |
1144 | struct mfc6_cache *c, **cp; | |
1145 | ||
1146 | cp = &mfc6_cache_array[i]; | |
1147 | while ((c = *cp) != NULL) { | |
1148 | if (c->mfc_flags & MFC_STATIC) { | |
1149 | cp = &c->next; | |
1150 | continue; | |
1151 | } | |
1152 | write_lock_bh(&mrt_lock); | |
1153 | *cp = c->next; | |
1154 | write_unlock_bh(&mrt_lock); | |
1155 | ||
58701ad4 | 1156 | ip6mr_cache_free(c); |
7bc570c8 YH |
1157 | } |
1158 | } | |
1159 | ||
1160 | if (atomic_read(&cache_resolve_queue_len) != 0) { | |
1161 | struct mfc6_cache *c; | |
1162 | ||
1163 | spin_lock_bh(&mfc_unres_lock); | |
1164 | while (mfc_unres_queue != NULL) { | |
1165 | c = mfc_unres_queue; | |
1166 | mfc_unres_queue = c->next; | |
1167 | spin_unlock_bh(&mfc_unres_lock); | |
1168 | ||
1169 | ip6mr_destroy_unres(c); | |
1170 | ||
1171 | spin_lock_bh(&mfc_unres_lock); | |
1172 | } | |
1173 | spin_unlock_bh(&mfc_unres_lock); | |
1174 | } | |
1175 | } | |
1176 | ||
1177 | static int ip6mr_sk_init(struct sock *sk) | |
1178 | { | |
1179 | int err = 0; | |
1180 | ||
1181 | rtnl_lock(); | |
1182 | write_lock_bh(&mrt_lock); | |
bd91b8bf BT |
1183 | if (likely(init_net.ipv6.mroute6_sk == NULL)) |
1184 | init_net.ipv6.mroute6_sk = sk; | |
7bc570c8 YH |
1185 | else |
1186 | err = -EADDRINUSE; | |
1187 | write_unlock_bh(&mrt_lock); | |
1188 | ||
1189 | rtnl_unlock(); | |
1190 | ||
1191 | return err; | |
1192 | } | |
1193 | ||
1194 | int ip6mr_sk_done(struct sock *sk) | |
1195 | { | |
1196 | int err = 0; | |
1197 | ||
1198 | rtnl_lock(); | |
bd91b8bf | 1199 | if (sk == init_net.ipv6.mroute6_sk) { |
7bc570c8 | 1200 | write_lock_bh(&mrt_lock); |
bd91b8bf | 1201 | init_net.ipv6.mroute6_sk = NULL; |
7bc570c8 YH |
1202 | write_unlock_bh(&mrt_lock); |
1203 | ||
1204 | mroute_clean_tables(sk); | |
1205 | } else | |
1206 | err = -EACCES; | |
1207 | rtnl_unlock(); | |
1208 | ||
1209 | return err; | |
1210 | } | |
1211 | ||
1212 | /* | |
1213 | * Socket options and virtual interface manipulation. The whole | |
1214 | * virtual interface system is a complete heap, but unfortunately | |
1215 | * that's how BSD mrouted happens to think. Maybe one day with a proper | |
1216 | * MOSPF/PIM router set up we can clean this up. | |
1217 | */ | |
1218 | ||
1219 | int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, int optlen) | |
1220 | { | |
1221 | int ret; | |
1222 | struct mif6ctl vif; | |
1223 | struct mf6cctl mfc; | |
1224 | mifi_t mifi; | |
1225 | ||
1226 | if (optname != MRT6_INIT) { | |
bd91b8bf | 1227 | if (sk != init_net.ipv6.mroute6_sk && !capable(CAP_NET_ADMIN)) |
7bc570c8 YH |
1228 | return -EACCES; |
1229 | } | |
1230 | ||
1231 | switch (optname) { | |
1232 | case MRT6_INIT: | |
1233 | if (sk->sk_type != SOCK_RAW || | |
1234 | inet_sk(sk)->num != IPPROTO_ICMPV6) | |
1235 | return -EOPNOTSUPP; | |
1236 | if (optlen < sizeof(int)) | |
1237 | return -EINVAL; | |
1238 | ||
1239 | return ip6mr_sk_init(sk); | |
1240 | ||
1241 | case MRT6_DONE: | |
1242 | return ip6mr_sk_done(sk); | |
1243 | ||
1244 | case MRT6_ADD_MIF: | |
1245 | if (optlen < sizeof(vif)) | |
1246 | return -EINVAL; | |
1247 | if (copy_from_user(&vif, optval, sizeof(vif))) | |
1248 | return -EFAULT; | |
6ac7eb08 | 1249 | if (vif.mif6c_mifi >= MAXMIFS) |
7bc570c8 YH |
1250 | return -ENFILE; |
1251 | rtnl_lock(); | |
bd91b8bf | 1252 | ret = mif6_add(&vif, sk == init_net.ipv6.mroute6_sk); |
7bc570c8 YH |
1253 | rtnl_unlock(); |
1254 | return ret; | |
1255 | ||
1256 | case MRT6_DEL_MIF: | |
1257 | if (optlen < sizeof(mifi_t)) | |
1258 | return -EINVAL; | |
1259 | if (copy_from_user(&mifi, optval, sizeof(mifi_t))) | |
1260 | return -EFAULT; | |
1261 | rtnl_lock(); | |
1262 | ret = mif6_delete(mifi); | |
1263 | rtnl_unlock(); | |
1264 | return ret; | |
1265 | ||
1266 | /* | |
1267 | * Manipulate the forwarding caches. These live | |
1268 | * in a sort of kernel/user symbiosis. | |
1269 | */ | |
1270 | case MRT6_ADD_MFC: | |
1271 | case MRT6_DEL_MFC: | |
1272 | if (optlen < sizeof(mfc)) | |
1273 | return -EINVAL; | |
1274 | if (copy_from_user(&mfc, optval, sizeof(mfc))) | |
1275 | return -EFAULT; | |
1276 | rtnl_lock(); | |
1277 | if (optname == MRT6_DEL_MFC) | |
1278 | ret = ip6mr_mfc_delete(&mfc); | |
1279 | else | |
bd91b8bf | 1280 | ret = ip6mr_mfc_add(&mfc, sk == init_net.ipv6.mroute6_sk); |
7bc570c8 YH |
1281 | rtnl_unlock(); |
1282 | return ret; | |
1283 | ||
14fb64e1 YH |
1284 | /* |
1285 | * Control PIM assert (to activate pim will activate assert) | |
1286 | */ | |
1287 | case MRT6_ASSERT: | |
1288 | { | |
1289 | int v; | |
1290 | if (get_user(v, (int __user *)optval)) | |
1291 | return -EFAULT; | |
1292 | mroute_do_assert = !!v; | |
1293 | return 0; | |
1294 | } | |
1295 | ||
1296 | #ifdef CONFIG_IPV6_PIMSM_V2 | |
1297 | case MRT6_PIM: | |
1298 | { | |
a9f83bf3 | 1299 | int v; |
14fb64e1 YH |
1300 | if (get_user(v, (int __user *)optval)) |
1301 | return -EFAULT; | |
1302 | v = !!v; | |
1303 | rtnl_lock(); | |
1304 | ret = 0; | |
1305 | if (v != mroute_do_pim) { | |
1306 | mroute_do_pim = v; | |
1307 | mroute_do_assert = v; | |
1308 | if (mroute_do_pim) | |
1309 | ret = inet6_add_protocol(&pim6_protocol, | |
1310 | IPPROTO_PIM); | |
1311 | else | |
1312 | ret = inet6_del_protocol(&pim6_protocol, | |
1313 | IPPROTO_PIM); | |
1314 | if (ret < 0) | |
1315 | ret = -EAGAIN; | |
1316 | } | |
1317 | rtnl_unlock(); | |
1318 | return ret; | |
1319 | } | |
1320 | ||
1321 | #endif | |
7bc570c8 | 1322 | /* |
7d120c55 | 1323 | * Spurious command, or MRT6_VERSION which you cannot |
7bc570c8 YH |
1324 | * set. |
1325 | */ | |
1326 | default: | |
1327 | return -ENOPROTOOPT; | |
1328 | } | |
1329 | } | |
1330 | ||
1331 | /* | |
1332 | * Getsock opt support for the multicast routing system. | |
1333 | */ | |
1334 | ||
1335 | int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, | |
1336 | int __user *optlen) | |
1337 | { | |
1338 | int olr; | |
1339 | int val; | |
1340 | ||
1341 | switch (optname) { | |
1342 | case MRT6_VERSION: | |
1343 | val = 0x0305; | |
1344 | break; | |
14fb64e1 YH |
1345 | #ifdef CONFIG_IPV6_PIMSM_V2 |
1346 | case MRT6_PIM: | |
1347 | val = mroute_do_pim; | |
1348 | break; | |
1349 | #endif | |
1350 | case MRT6_ASSERT: | |
1351 | val = mroute_do_assert; | |
1352 | break; | |
7bc570c8 YH |
1353 | default: |
1354 | return -ENOPROTOOPT; | |
1355 | } | |
1356 | ||
1357 | if (get_user(olr, optlen)) | |
1358 | return -EFAULT; | |
1359 | ||
1360 | olr = min_t(int, olr, sizeof(int)); | |
1361 | if (olr < 0) | |
1362 | return -EINVAL; | |
1363 | ||
1364 | if (put_user(olr, optlen)) | |
1365 | return -EFAULT; | |
1366 | if (copy_to_user(optval, &val, olr)) | |
1367 | return -EFAULT; | |
1368 | return 0; | |
1369 | } | |
1370 | ||
1371 | /* | |
1372 | * The IP multicast ioctl support routines. | |
1373 | */ | |
1374 | ||
1375 | int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) | |
1376 | { | |
1377 | struct sioc_sg_req6 sr; | |
1378 | struct sioc_mif_req6 vr; | |
1379 | struct mif_device *vif; | |
1380 | struct mfc6_cache *c; | |
1381 | ||
1382 | switch (cmd) { | |
1383 | case SIOCGETMIFCNT_IN6: | |
1384 | if (copy_from_user(&vr, arg, sizeof(vr))) | |
1385 | return -EFAULT; | |
4e16880c | 1386 | if (vr.mifi >= init_net.ipv6.maxvif) |
7bc570c8 YH |
1387 | return -EINVAL; |
1388 | read_lock(&mrt_lock); | |
4e16880c BT |
1389 | vif = &init_net.ipv6.vif6_table[vr.mifi]; |
1390 | if (MIF_EXISTS(&init_net, vr.mifi)) { | |
7bc570c8 YH |
1391 | vr.icount = vif->pkt_in; |
1392 | vr.ocount = vif->pkt_out; | |
1393 | vr.ibytes = vif->bytes_in; | |
1394 | vr.obytes = vif->bytes_out; | |
1395 | read_unlock(&mrt_lock); | |
1396 | ||
1397 | if (copy_to_user(arg, &vr, sizeof(vr))) | |
1398 | return -EFAULT; | |
1399 | return 0; | |
1400 | } | |
1401 | read_unlock(&mrt_lock); | |
1402 | return -EADDRNOTAVAIL; | |
1403 | case SIOCGETSGCNT_IN6: | |
1404 | if (copy_from_user(&sr, arg, sizeof(sr))) | |
1405 | return -EFAULT; | |
1406 | ||
1407 | read_lock(&mrt_lock); | |
1408 | c = ip6mr_cache_find(&sr.src.sin6_addr, &sr.grp.sin6_addr); | |
1409 | if (c) { | |
1410 | sr.pktcnt = c->mfc_un.res.pkt; | |
1411 | sr.bytecnt = c->mfc_un.res.bytes; | |
1412 | sr.wrong_if = c->mfc_un.res.wrong_if; | |
1413 | read_unlock(&mrt_lock); | |
1414 | ||
1415 | if (copy_to_user(arg, &sr, sizeof(sr))) | |
1416 | return -EFAULT; | |
1417 | return 0; | |
1418 | } | |
1419 | read_unlock(&mrt_lock); | |
1420 | return -EADDRNOTAVAIL; | |
1421 | default: | |
1422 | return -ENOIOCTLCMD; | |
1423 | } | |
1424 | } | |
1425 | ||
1426 | ||
1427 | static inline int ip6mr_forward2_finish(struct sk_buff *skb) | |
1428 | { | |
483a47d2 DL |
1429 | IP6_INC_STATS_BH(dev_net(skb->dst->dev), ip6_dst_idev(skb->dst), |
1430 | IPSTATS_MIB_OUTFORWDATAGRAMS); | |
7bc570c8 YH |
1431 | return dst_output(skb); |
1432 | } | |
1433 | ||
1434 | /* | |
1435 | * Processing handlers for ip6mr_forward | |
1436 | */ | |
1437 | ||
1438 | static int ip6mr_forward2(struct sk_buff *skb, struct mfc6_cache *c, int vifi) | |
1439 | { | |
1440 | struct ipv6hdr *ipv6h; | |
4e16880c | 1441 | struct mif_device *vif = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 YH |
1442 | struct net_device *dev; |
1443 | struct dst_entry *dst; | |
1444 | struct flowi fl; | |
1445 | ||
1446 | if (vif->dev == NULL) | |
1447 | goto out_free; | |
1448 | ||
14fb64e1 YH |
1449 | #ifdef CONFIG_IPV6_PIMSM_V2 |
1450 | if (vif->flags & MIFF_REGISTER) { | |
1451 | vif->pkt_out++; | |
1452 | vif->bytes_out += skb->len; | |
dc58c78c PE |
1453 | vif->dev->stats.tx_bytes += skb->len; |
1454 | vif->dev->stats.tx_packets++; | |
14fb64e1 YH |
1455 | ip6mr_cache_report(skb, vifi, MRT6MSG_WHOLEPKT); |
1456 | kfree_skb(skb); | |
1457 | return 0; | |
1458 | } | |
1459 | #endif | |
1460 | ||
7bc570c8 YH |
1461 | ipv6h = ipv6_hdr(skb); |
1462 | ||
1463 | fl = (struct flowi) { | |
1464 | .oif = vif->link, | |
1465 | .nl_u = { .ip6_u = | |
1466 | { .daddr = ipv6h->daddr, } | |
1467 | } | |
1468 | }; | |
1469 | ||
1470 | dst = ip6_route_output(&init_net, NULL, &fl); | |
1471 | if (!dst) | |
1472 | goto out_free; | |
1473 | ||
1474 | dst_release(skb->dst); | |
1475 | skb->dst = dst; | |
1476 | ||
1477 | /* | |
1478 | * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally | |
1479 | * not only before forwarding, but after forwarding on all output | |
1480 | * interfaces. It is clear, if mrouter runs a multicasting | |
1481 | * program, it should receive packets not depending to what interface | |
1482 | * program is joined. | |
1483 | * If we will not make it, the program will have to join on all | |
1484 | * interfaces. On the other hand, multihoming host (or router, but | |
1485 | * not mrouter) cannot join to more than one interface - it will | |
1486 | * result in receiving multiple packets. | |
1487 | */ | |
1488 | dev = vif->dev; | |
1489 | skb->dev = dev; | |
1490 | vif->pkt_out++; | |
1491 | vif->bytes_out += skb->len; | |
1492 | ||
1493 | /* We are about to write */ | |
1494 | /* XXX: extension headers? */ | |
1495 | if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) | |
1496 | goto out_free; | |
1497 | ||
1498 | ipv6h = ipv6_hdr(skb); | |
1499 | ipv6h->hop_limit--; | |
1500 | ||
1501 | IP6CB(skb)->flags |= IP6SKB_FORWARDED; | |
1502 | ||
1503 | return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dev, | |
1504 | ip6mr_forward2_finish); | |
1505 | ||
1506 | out_free: | |
1507 | kfree_skb(skb); | |
1508 | return 0; | |
1509 | } | |
1510 | ||
1511 | static int ip6mr_find_vif(struct net_device *dev) | |
1512 | { | |
1513 | int ct; | |
4e16880c BT |
1514 | for (ct = init_net.ipv6.maxvif - 1; ct >= 0; ct--) { |
1515 | if (init_net.ipv6.vif6_table[ct].dev == dev) | |
7bc570c8 YH |
1516 | break; |
1517 | } | |
1518 | return ct; | |
1519 | } | |
1520 | ||
1521 | static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache) | |
1522 | { | |
1523 | int psend = -1; | |
1524 | int vif, ct; | |
1525 | ||
1526 | vif = cache->mf6c_parent; | |
1527 | cache->mfc_un.res.pkt++; | |
1528 | cache->mfc_un.res.bytes += skb->len; | |
1529 | ||
14fb64e1 YH |
1530 | /* |
1531 | * Wrong interface: drop packet and (maybe) send PIM assert. | |
1532 | */ | |
4e16880c | 1533 | if (init_net.ipv6.vif6_table[vif].dev != skb->dev) { |
14fb64e1 YH |
1534 | int true_vifi; |
1535 | ||
1536 | cache->mfc_un.res.wrong_if++; | |
1537 | true_vifi = ip6mr_find_vif(skb->dev); | |
1538 | ||
1539 | if (true_vifi >= 0 && mroute_do_assert && | |
1540 | /* pimsm uses asserts, when switching from RPT to SPT, | |
1541 | so that we cannot check that packet arrived on an oif. | |
1542 | It is bad, but otherwise we would need to move pretty | |
1543 | large chunk of pimd to kernel. Ough... --ANK | |
1544 | */ | |
1545 | (mroute_do_pim || cache->mfc_un.res.ttls[true_vifi] < 255) && | |
1546 | time_after(jiffies, | |
1547 | cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { | |
1548 | cache->mfc_un.res.last_assert = jiffies; | |
1549 | ip6mr_cache_report(skb, true_vifi, MRT6MSG_WRONGMIF); | |
1550 | } | |
1551 | goto dont_forward; | |
1552 | } | |
1553 | ||
4e16880c BT |
1554 | init_net.ipv6.vif6_table[vif].pkt_in++; |
1555 | init_net.ipv6.vif6_table[vif].bytes_in += skb->len; | |
7bc570c8 YH |
1556 | |
1557 | /* | |
1558 | * Forward the frame | |
1559 | */ | |
1560 | for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { | |
1561 | if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { | |
1562 | if (psend != -1) { | |
1563 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1564 | if (skb2) | |
1565 | ip6mr_forward2(skb2, cache, psend); | |
1566 | } | |
1567 | psend = ct; | |
1568 | } | |
1569 | } | |
1570 | if (psend != -1) { | |
1571 | ip6mr_forward2(skb, cache, psend); | |
1572 | return 0; | |
1573 | } | |
1574 | ||
14fb64e1 | 1575 | dont_forward: |
7bc570c8 YH |
1576 | kfree_skb(skb); |
1577 | return 0; | |
1578 | } | |
1579 | ||
1580 | ||
1581 | /* | |
1582 | * Multicast packets for forwarding arrive here | |
1583 | */ | |
1584 | ||
1585 | int ip6_mr_input(struct sk_buff *skb) | |
1586 | { | |
1587 | struct mfc6_cache *cache; | |
1588 | ||
1589 | read_lock(&mrt_lock); | |
1590 | cache = ip6mr_cache_find(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); | |
1591 | ||
1592 | /* | |
1593 | * No usable cache entry | |
1594 | */ | |
1595 | if (cache == NULL) { | |
1596 | int vif; | |
1597 | ||
1598 | vif = ip6mr_find_vif(skb->dev); | |
1599 | if (vif >= 0) { | |
1600 | int err = ip6mr_cache_unresolved(vif, skb); | |
1601 | read_unlock(&mrt_lock); | |
1602 | ||
1603 | return err; | |
1604 | } | |
1605 | read_unlock(&mrt_lock); | |
1606 | kfree_skb(skb); | |
1607 | return -ENODEV; | |
1608 | } | |
1609 | ||
1610 | ip6_mr_forward(skb, cache); | |
1611 | ||
1612 | read_unlock(&mrt_lock); | |
1613 | ||
1614 | return 0; | |
1615 | } | |
1616 | ||
1617 | ||
1618 | static int | |
1619 | ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm) | |
1620 | { | |
1621 | int ct; | |
1622 | struct rtnexthop *nhp; | |
4e16880c | 1623 | struct net_device *dev = init_net.ipv6.vif6_table[c->mf6c_parent].dev; |
549e028d | 1624 | u8 *b = skb_tail_pointer(skb); |
7bc570c8 YH |
1625 | struct rtattr *mp_head; |
1626 | ||
1627 | if (dev) | |
1628 | RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex); | |
1629 | ||
1630 | mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0)); | |
1631 | ||
1632 | for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { | |
1633 | if (c->mfc_un.res.ttls[ct] < 255) { | |
1634 | if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4)) | |
1635 | goto rtattr_failure; | |
1636 | nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp))); | |
1637 | nhp->rtnh_flags = 0; | |
1638 | nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; | |
4e16880c | 1639 | nhp->rtnh_ifindex = init_net.ipv6.vif6_table[ct].dev->ifindex; |
7bc570c8 YH |
1640 | nhp->rtnh_len = sizeof(*nhp); |
1641 | } | |
1642 | } | |
1643 | mp_head->rta_type = RTA_MULTIPATH; | |
549e028d | 1644 | mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head; |
7bc570c8 YH |
1645 | rtm->rtm_type = RTN_MULTICAST; |
1646 | return 1; | |
1647 | ||
1648 | rtattr_failure: | |
1649 | nlmsg_trim(skb, b); | |
1650 | return -EMSGSIZE; | |
1651 | } | |
1652 | ||
1653 | int ip6mr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait) | |
1654 | { | |
1655 | int err; | |
1656 | struct mfc6_cache *cache; | |
1657 | struct rt6_info *rt = (struct rt6_info *)skb->dst; | |
1658 | ||
1659 | read_lock(&mrt_lock); | |
1660 | cache = ip6mr_cache_find(&rt->rt6i_src.addr, &rt->rt6i_dst.addr); | |
1661 | ||
1662 | if (!cache) { | |
1663 | struct sk_buff *skb2; | |
1664 | struct ipv6hdr *iph; | |
1665 | struct net_device *dev; | |
1666 | int vif; | |
1667 | ||
1668 | if (nowait) { | |
1669 | read_unlock(&mrt_lock); | |
1670 | return -EAGAIN; | |
1671 | } | |
1672 | ||
1673 | dev = skb->dev; | |
1674 | if (dev == NULL || (vif = ip6mr_find_vif(dev)) < 0) { | |
1675 | read_unlock(&mrt_lock); | |
1676 | return -ENODEV; | |
1677 | } | |
1678 | ||
1679 | /* really correct? */ | |
1680 | skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); | |
1681 | if (!skb2) { | |
1682 | read_unlock(&mrt_lock); | |
1683 | return -ENOMEM; | |
1684 | } | |
1685 | ||
1686 | skb_reset_transport_header(skb2); | |
1687 | ||
1688 | skb_put(skb2, sizeof(struct ipv6hdr)); | |
1689 | skb_reset_network_header(skb2); | |
1690 | ||
1691 | iph = ipv6_hdr(skb2); | |
1692 | iph->version = 0; | |
1693 | iph->priority = 0; | |
1694 | iph->flow_lbl[0] = 0; | |
1695 | iph->flow_lbl[1] = 0; | |
1696 | iph->flow_lbl[2] = 0; | |
1697 | iph->payload_len = 0; | |
1698 | iph->nexthdr = IPPROTO_NONE; | |
1699 | iph->hop_limit = 0; | |
1700 | ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr); | |
1701 | ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr); | |
1702 | ||
1703 | err = ip6mr_cache_unresolved(vif, skb2); | |
1704 | read_unlock(&mrt_lock); | |
1705 | ||
1706 | return err; | |
1707 | } | |
1708 | ||
1709 | if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY)) | |
1710 | cache->mfc_flags |= MFC_NOTIFY; | |
1711 | ||
1712 | err = ip6mr_fill_mroute(skb, cache, rtm); | |
1713 | read_unlock(&mrt_lock); | |
1714 | return err; | |
1715 | } | |
1716 |