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 | ||
7bc570c8 YH |
71 | static struct mfc6_cache *mfc_unres_queue; /* Queue of unresolved entries */ |
72 | static atomic_t cache_resolve_queue_len; /* Size of unresolved */ | |
73 | ||
74 | /* Special spinlock for queue of unresolved entries */ | |
75 | static DEFINE_SPINLOCK(mfc_unres_lock); | |
76 | ||
77 | /* We return to original Alan's scheme. Hash table of resolved | |
78 | entries is changed only in process context and protected | |
79 | with weak lock mrt_lock. Queue of unresolved entries is protected | |
80 | with strong spinlock mfc_unres_lock. | |
81 | ||
82 | In this case data path is free of exclusive locks at all. | |
83 | */ | |
84 | ||
85 | static struct kmem_cache *mrt_cachep __read_mostly; | |
86 | ||
87 | static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache); | |
6ac7eb08 | 88 | static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert); |
7bc570c8 YH |
89 | static int ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm); |
90 | ||
14fb64e1 YH |
91 | #ifdef CONFIG_IPV6_PIMSM_V2 |
92 | static struct inet6_protocol pim6_protocol; | |
93 | #endif | |
94 | ||
7bc570c8 YH |
95 | static struct timer_list ipmr_expire_timer; |
96 | ||
97 | ||
98 | #ifdef CONFIG_PROC_FS | |
99 | ||
100 | struct ipmr_mfc_iter { | |
101 | struct mfc6_cache **cache; | |
102 | int ct; | |
103 | }; | |
104 | ||
105 | ||
106 | static struct mfc6_cache *ipmr_mfc_seq_idx(struct ipmr_mfc_iter *it, loff_t pos) | |
107 | { | |
108 | struct mfc6_cache *mfc; | |
109 | ||
4a6258a0 | 110 | it->cache = init_net.ipv6.mfc6_cache_array; |
7bc570c8 | 111 | read_lock(&mrt_lock); |
4a6258a0 BT |
112 | for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) |
113 | for (mfc = init_net.ipv6.mfc6_cache_array[it->ct]; | |
114 | mfc; mfc = mfc->next) | |
7bc570c8 YH |
115 | if (pos-- == 0) |
116 | return mfc; | |
117 | read_unlock(&mrt_lock); | |
118 | ||
119 | it->cache = &mfc_unres_queue; | |
120 | spin_lock_bh(&mfc_unres_lock); | |
121 | for (mfc = mfc_unres_queue; mfc; mfc = mfc->next) | |
122 | if (pos-- == 0) | |
123 | return mfc; | |
124 | spin_unlock_bh(&mfc_unres_lock); | |
125 | ||
126 | it->cache = NULL; | |
127 | return NULL; | |
128 | } | |
129 | ||
130 | ||
131 | ||
132 | ||
133 | /* | |
134 | * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif | |
135 | */ | |
136 | ||
137 | struct ipmr_vif_iter { | |
138 | int ct; | |
139 | }; | |
140 | ||
141 | static struct mif_device *ip6mr_vif_seq_idx(struct ipmr_vif_iter *iter, | |
142 | loff_t pos) | |
143 | { | |
4e16880c BT |
144 | for (iter->ct = 0; iter->ct < init_net.ipv6.maxvif; ++iter->ct) { |
145 | if (!MIF_EXISTS(&init_net, iter->ct)) | |
7bc570c8 YH |
146 | continue; |
147 | if (pos-- == 0) | |
4e16880c | 148 | return &init_net.ipv6.vif6_table[iter->ct]; |
7bc570c8 YH |
149 | } |
150 | return NULL; | |
151 | } | |
152 | ||
153 | static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos) | |
154 | __acquires(mrt_lock) | |
155 | { | |
156 | read_lock(&mrt_lock); | |
157 | return (*pos ? ip6mr_vif_seq_idx(seq->private, *pos - 1) | |
158 | : SEQ_START_TOKEN); | |
159 | } | |
160 | ||
161 | static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
162 | { | |
163 | struct ipmr_vif_iter *iter = seq->private; | |
164 | ||
165 | ++*pos; | |
166 | if (v == SEQ_START_TOKEN) | |
167 | return ip6mr_vif_seq_idx(iter, 0); | |
168 | ||
4e16880c BT |
169 | while (++iter->ct < init_net.ipv6.maxvif) { |
170 | if (!MIF_EXISTS(&init_net, iter->ct)) | |
7bc570c8 | 171 | continue; |
4e16880c | 172 | return &init_net.ipv6.vif6_table[iter->ct]; |
7bc570c8 YH |
173 | } |
174 | return NULL; | |
175 | } | |
176 | ||
177 | static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v) | |
178 | __releases(mrt_lock) | |
179 | { | |
180 | read_unlock(&mrt_lock); | |
181 | } | |
182 | ||
183 | static int ip6mr_vif_seq_show(struct seq_file *seq, void *v) | |
184 | { | |
185 | if (v == SEQ_START_TOKEN) { | |
186 | seq_puts(seq, | |
187 | "Interface BytesIn PktsIn BytesOut PktsOut Flags\n"); | |
188 | } else { | |
189 | const struct mif_device *vif = v; | |
190 | const char *name = vif->dev ? vif->dev->name : "none"; | |
191 | ||
192 | seq_printf(seq, | |
d430a227 | 193 | "%2td %-10s %8ld %7ld %8ld %7ld %05X\n", |
4e16880c | 194 | vif - init_net.ipv6.vif6_table, |
7bc570c8 YH |
195 | name, vif->bytes_in, vif->pkt_in, |
196 | vif->bytes_out, vif->pkt_out, | |
197 | vif->flags); | |
198 | } | |
199 | return 0; | |
200 | } | |
201 | ||
202 | static struct seq_operations ip6mr_vif_seq_ops = { | |
203 | .start = ip6mr_vif_seq_start, | |
204 | .next = ip6mr_vif_seq_next, | |
205 | .stop = ip6mr_vif_seq_stop, | |
206 | .show = ip6mr_vif_seq_show, | |
207 | }; | |
208 | ||
209 | static int ip6mr_vif_open(struct inode *inode, struct file *file) | |
210 | { | |
211 | return seq_open_private(file, &ip6mr_vif_seq_ops, | |
212 | sizeof(struct ipmr_vif_iter)); | |
213 | } | |
214 | ||
215 | static struct file_operations ip6mr_vif_fops = { | |
216 | .owner = THIS_MODULE, | |
217 | .open = ip6mr_vif_open, | |
218 | .read = seq_read, | |
219 | .llseek = seq_lseek, | |
eedd726e | 220 | .release = seq_release_private, |
7bc570c8 YH |
221 | }; |
222 | ||
223 | static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) | |
224 | { | |
225 | return (*pos ? ipmr_mfc_seq_idx(seq->private, *pos - 1) | |
226 | : SEQ_START_TOKEN); | |
227 | } | |
228 | ||
229 | static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
230 | { | |
231 | struct mfc6_cache *mfc = v; | |
232 | struct ipmr_mfc_iter *it = seq->private; | |
233 | ||
234 | ++*pos; | |
235 | ||
236 | if (v == SEQ_START_TOKEN) | |
237 | return ipmr_mfc_seq_idx(seq->private, 0); | |
238 | ||
239 | if (mfc->next) | |
240 | return mfc->next; | |
241 | ||
242 | if (it->cache == &mfc_unres_queue) | |
243 | goto end_of_list; | |
244 | ||
4a6258a0 | 245 | BUG_ON(it->cache != init_net.ipv6.mfc6_cache_array); |
7bc570c8 | 246 | |
4a6258a0 BT |
247 | while (++it->ct < MFC6_LINES) { |
248 | mfc = init_net.ipv6.mfc6_cache_array[it->ct]; | |
7bc570c8 YH |
249 | if (mfc) |
250 | return mfc; | |
251 | } | |
252 | ||
253 | /* exhausted cache_array, show unresolved */ | |
254 | read_unlock(&mrt_lock); | |
255 | it->cache = &mfc_unres_queue; | |
256 | it->ct = 0; | |
257 | ||
258 | spin_lock_bh(&mfc_unres_lock); | |
259 | mfc = mfc_unres_queue; | |
260 | if (mfc) | |
261 | return mfc; | |
262 | ||
263 | end_of_list: | |
264 | spin_unlock_bh(&mfc_unres_lock); | |
265 | it->cache = NULL; | |
266 | ||
267 | return NULL; | |
268 | } | |
269 | ||
270 | static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) | |
271 | { | |
272 | struct ipmr_mfc_iter *it = seq->private; | |
273 | ||
274 | if (it->cache == &mfc_unres_queue) | |
275 | spin_unlock_bh(&mfc_unres_lock); | |
4a6258a0 | 276 | else if (it->cache == init_net.ipv6.mfc6_cache_array) |
7bc570c8 YH |
277 | read_unlock(&mrt_lock); |
278 | } | |
279 | ||
280 | static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) | |
281 | { | |
282 | int n; | |
283 | ||
284 | if (v == SEQ_START_TOKEN) { | |
285 | seq_puts(seq, | |
286 | "Group " | |
287 | "Origin " | |
288 | "Iif Pkts Bytes Wrong Oifs\n"); | |
289 | } else { | |
290 | const struct mfc6_cache *mfc = v; | |
291 | const struct ipmr_mfc_iter *it = seq->private; | |
292 | ||
999890b2 | 293 | seq_printf(seq, "%pI6 %pI6 %-3hd", |
0c6ce78a | 294 | &mfc->mf6c_mcastgrp, &mfc->mf6c_origin, |
1ea472e2 | 295 | mfc->mf6c_parent); |
7bc570c8 YH |
296 | |
297 | if (it->cache != &mfc_unres_queue) { | |
1ea472e2 BT |
298 | seq_printf(seq, " %8lu %8lu %8lu", |
299 | mfc->mfc_un.res.pkt, | |
300 | mfc->mfc_un.res.bytes, | |
301 | mfc->mfc_un.res.wrong_if); | |
7bc570c8 YH |
302 | for (n = mfc->mfc_un.res.minvif; |
303 | n < mfc->mfc_un.res.maxvif; n++) { | |
4e16880c | 304 | if (MIF_EXISTS(&init_net, n) && |
7bc570c8 YH |
305 | mfc->mfc_un.res.ttls[n] < 255) |
306 | seq_printf(seq, | |
307 | " %2d:%-3d", | |
308 | n, mfc->mfc_un.res.ttls[n]); | |
309 | } | |
1ea472e2 BT |
310 | } else { |
311 | /* unresolved mfc_caches don't contain | |
312 | * pkt, bytes and wrong_if values | |
313 | */ | |
314 | seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); | |
7bc570c8 YH |
315 | } |
316 | seq_putc(seq, '\n'); | |
317 | } | |
318 | return 0; | |
319 | } | |
320 | ||
321 | static struct seq_operations ipmr_mfc_seq_ops = { | |
322 | .start = ipmr_mfc_seq_start, | |
323 | .next = ipmr_mfc_seq_next, | |
324 | .stop = ipmr_mfc_seq_stop, | |
325 | .show = ipmr_mfc_seq_show, | |
326 | }; | |
327 | ||
328 | static int ipmr_mfc_open(struct inode *inode, struct file *file) | |
329 | { | |
330 | return seq_open_private(file, &ipmr_mfc_seq_ops, | |
331 | sizeof(struct ipmr_mfc_iter)); | |
332 | } | |
333 | ||
334 | static struct file_operations ip6mr_mfc_fops = { | |
335 | .owner = THIS_MODULE, | |
336 | .open = ipmr_mfc_open, | |
337 | .read = seq_read, | |
338 | .llseek = seq_lseek, | |
eedd726e | 339 | .release = seq_release_private, |
7bc570c8 YH |
340 | }; |
341 | #endif | |
342 | ||
14fb64e1 YH |
343 | #ifdef CONFIG_IPV6_PIMSM_V2 |
344 | static int reg_vif_num = -1; | |
345 | ||
346 | static int pim6_rcv(struct sk_buff *skb) | |
347 | { | |
348 | struct pimreghdr *pim; | |
349 | struct ipv6hdr *encap; | |
350 | struct net_device *reg_dev = NULL; | |
351 | ||
352 | if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap))) | |
353 | goto drop; | |
354 | ||
355 | pim = (struct pimreghdr *)skb_transport_header(skb); | |
356 | if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) || | |
357 | (pim->flags & PIM_NULL_REGISTER) || | |
358 | (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && | |
ec6b486f | 359 | csum_fold(skb_checksum(skb, 0, skb->len, 0)))) |
14fb64e1 YH |
360 | goto drop; |
361 | ||
362 | /* check if the inner packet is destined to mcast group */ | |
363 | encap = (struct ipv6hdr *)(skb_transport_header(skb) + | |
364 | sizeof(*pim)); | |
365 | ||
366 | if (!ipv6_addr_is_multicast(&encap->daddr) || | |
367 | encap->payload_len == 0 || | |
368 | ntohs(encap->payload_len) + sizeof(*pim) > skb->len) | |
369 | goto drop; | |
370 | ||
371 | read_lock(&mrt_lock); | |
372 | if (reg_vif_num >= 0) | |
4e16880c | 373 | reg_dev = init_net.ipv6.vif6_table[reg_vif_num].dev; |
14fb64e1 YH |
374 | if (reg_dev) |
375 | dev_hold(reg_dev); | |
376 | read_unlock(&mrt_lock); | |
377 | ||
378 | if (reg_dev == NULL) | |
379 | goto drop; | |
380 | ||
381 | skb->mac_header = skb->network_header; | |
382 | skb_pull(skb, (u8 *)encap - skb->data); | |
383 | skb_reset_network_header(skb); | |
384 | skb->dev = reg_dev; | |
385 | skb->protocol = htons(ETH_P_IP); | |
386 | skb->ip_summed = 0; | |
387 | skb->pkt_type = PACKET_HOST; | |
388 | dst_release(skb->dst); | |
dc58c78c PE |
389 | reg_dev->stats.rx_bytes += skb->len; |
390 | reg_dev->stats.rx_packets++; | |
14fb64e1 YH |
391 | skb->dst = NULL; |
392 | nf_reset(skb); | |
393 | netif_rx(skb); | |
394 | dev_put(reg_dev); | |
395 | return 0; | |
396 | drop: | |
397 | kfree_skb(skb); | |
398 | return 0; | |
399 | } | |
400 | ||
401 | static struct inet6_protocol pim6_protocol = { | |
402 | .handler = pim6_rcv, | |
403 | }; | |
404 | ||
405 | /* Service routines creating virtual interfaces: PIMREG */ | |
406 | ||
407 | static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) | |
408 | { | |
409 | read_lock(&mrt_lock); | |
dc58c78c PE |
410 | dev->stats.tx_bytes += skb->len; |
411 | dev->stats.tx_packets++; | |
14fb64e1 YH |
412 | ip6mr_cache_report(skb, reg_vif_num, MRT6MSG_WHOLEPKT); |
413 | read_unlock(&mrt_lock); | |
414 | kfree_skb(skb); | |
415 | return 0; | |
416 | } | |
417 | ||
007c3838 SH |
418 | static const struct net_device_ops reg_vif_netdev_ops = { |
419 | .ndo_start_xmit = reg_vif_xmit, | |
420 | }; | |
421 | ||
14fb64e1 YH |
422 | static void reg_vif_setup(struct net_device *dev) |
423 | { | |
424 | dev->type = ARPHRD_PIMREG; | |
425 | dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8; | |
426 | dev->flags = IFF_NOARP; | |
007c3838 | 427 | dev->netdev_ops = ®_vif_netdev_ops; |
14fb64e1 YH |
428 | dev->destructor = free_netdev; |
429 | } | |
430 | ||
431 | static struct net_device *ip6mr_reg_vif(void) | |
432 | { | |
433 | struct net_device *dev; | |
14fb64e1 | 434 | |
dc58c78c | 435 | dev = alloc_netdev(0, "pim6reg", reg_vif_setup); |
14fb64e1 YH |
436 | if (dev == NULL) |
437 | return NULL; | |
438 | ||
439 | if (register_netdevice(dev)) { | |
440 | free_netdev(dev); | |
441 | return NULL; | |
442 | } | |
443 | dev->iflink = 0; | |
444 | ||
14fb64e1 YH |
445 | if (dev_open(dev)) |
446 | goto failure; | |
447 | ||
7af3db78 | 448 | dev_hold(dev); |
14fb64e1 YH |
449 | return dev; |
450 | ||
451 | failure: | |
452 | /* allow the register to be completed before unregistering. */ | |
453 | rtnl_unlock(); | |
454 | rtnl_lock(); | |
455 | ||
456 | unregister_netdevice(dev); | |
457 | return NULL; | |
458 | } | |
459 | #endif | |
460 | ||
7bc570c8 YH |
461 | /* |
462 | * Delete a VIF entry | |
463 | */ | |
464 | ||
465 | static int mif6_delete(int vifi) | |
466 | { | |
467 | struct mif_device *v; | |
468 | struct net_device *dev; | |
4e16880c | 469 | if (vifi < 0 || vifi >= init_net.ipv6.maxvif) |
7bc570c8 YH |
470 | return -EADDRNOTAVAIL; |
471 | ||
4e16880c | 472 | v = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 YH |
473 | |
474 | write_lock_bh(&mrt_lock); | |
475 | dev = v->dev; | |
476 | v->dev = NULL; | |
477 | ||
478 | if (!dev) { | |
479 | write_unlock_bh(&mrt_lock); | |
480 | return -EADDRNOTAVAIL; | |
481 | } | |
482 | ||
14fb64e1 YH |
483 | #ifdef CONFIG_IPV6_PIMSM_V2 |
484 | if (vifi == reg_vif_num) | |
485 | reg_vif_num = -1; | |
486 | #endif | |
487 | ||
4e16880c | 488 | if (vifi + 1 == init_net.ipv6.maxvif) { |
7bc570c8 YH |
489 | int tmp; |
490 | for (tmp = vifi - 1; tmp >= 0; tmp--) { | |
4e16880c | 491 | if (MIF_EXISTS(&init_net, tmp)) |
7bc570c8 YH |
492 | break; |
493 | } | |
4e16880c | 494 | init_net.ipv6.maxvif = tmp + 1; |
7bc570c8 YH |
495 | } |
496 | ||
497 | write_unlock_bh(&mrt_lock); | |
498 | ||
499 | dev_set_allmulti(dev, -1); | |
500 | ||
501 | if (v->flags & MIFF_REGISTER) | |
502 | unregister_netdevice(dev); | |
503 | ||
504 | dev_put(dev); | |
505 | return 0; | |
506 | } | |
507 | ||
58701ad4 BT |
508 | static inline void ip6mr_cache_free(struct mfc6_cache *c) |
509 | { | |
510 | release_net(mfc6_net(c)); | |
511 | kmem_cache_free(mrt_cachep, c); | |
512 | } | |
513 | ||
7bc570c8 YH |
514 | /* Destroy an unresolved cache entry, killing queued skbs |
515 | and reporting error to netlink readers. | |
516 | */ | |
517 | ||
518 | static void ip6mr_destroy_unres(struct mfc6_cache *c) | |
519 | { | |
520 | struct sk_buff *skb; | |
521 | ||
522 | atomic_dec(&cache_resolve_queue_len); | |
523 | ||
524 | while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) { | |
525 | if (ipv6_hdr(skb)->version == 0) { | |
526 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); | |
527 | nlh->nlmsg_type = NLMSG_ERROR; | |
528 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
529 | skb_trim(skb, nlh->nlmsg_len); | |
530 | ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT; | |
531 | rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); | |
532 | } else | |
533 | kfree_skb(skb); | |
534 | } | |
535 | ||
58701ad4 | 536 | ip6mr_cache_free(c); |
7bc570c8 YH |
537 | } |
538 | ||
539 | ||
540 | /* Single timer process for all the unresolved queue. */ | |
541 | ||
542 | static void ipmr_do_expire_process(unsigned long dummy) | |
543 | { | |
544 | unsigned long now = jiffies; | |
545 | unsigned long expires = 10 * HZ; | |
546 | struct mfc6_cache *c, **cp; | |
547 | ||
548 | cp = &mfc_unres_queue; | |
549 | ||
550 | while ((c = *cp) != NULL) { | |
551 | if (time_after(c->mfc_un.unres.expires, now)) { | |
552 | /* not yet... */ | |
553 | unsigned long interval = c->mfc_un.unres.expires - now; | |
554 | if (interval < expires) | |
555 | expires = interval; | |
556 | cp = &c->next; | |
557 | continue; | |
558 | } | |
559 | ||
560 | *cp = c->next; | |
561 | ip6mr_destroy_unres(c); | |
562 | } | |
563 | ||
564 | if (atomic_read(&cache_resolve_queue_len)) | |
565 | mod_timer(&ipmr_expire_timer, jiffies + expires); | |
566 | } | |
567 | ||
568 | static void ipmr_expire_process(unsigned long dummy) | |
569 | { | |
570 | if (!spin_trylock(&mfc_unres_lock)) { | |
571 | mod_timer(&ipmr_expire_timer, jiffies + 1); | |
572 | return; | |
573 | } | |
574 | ||
575 | if (atomic_read(&cache_resolve_queue_len)) | |
576 | ipmr_do_expire_process(dummy); | |
577 | ||
578 | spin_unlock(&mfc_unres_lock); | |
579 | } | |
580 | ||
581 | /* Fill oifs list. It is called under write locked mrt_lock. */ | |
582 | ||
583 | static void ip6mr_update_thresholds(struct mfc6_cache *cache, unsigned char *ttls) | |
584 | { | |
585 | int vifi; | |
586 | ||
6ac7eb08 | 587 | cache->mfc_un.res.minvif = MAXMIFS; |
7bc570c8 | 588 | cache->mfc_un.res.maxvif = 0; |
6ac7eb08 | 589 | memset(cache->mfc_un.res.ttls, 255, MAXMIFS); |
7bc570c8 | 590 | |
4e16880c BT |
591 | for (vifi = 0; vifi < init_net.ipv6.maxvif; vifi++) { |
592 | if (MIF_EXISTS(&init_net, vifi) && | |
593 | ttls[vifi] && ttls[vifi] < 255) { | |
7bc570c8 YH |
594 | cache->mfc_un.res.ttls[vifi] = ttls[vifi]; |
595 | if (cache->mfc_un.res.minvif > vifi) | |
596 | cache->mfc_un.res.minvif = vifi; | |
597 | if (cache->mfc_un.res.maxvif <= vifi) | |
598 | cache->mfc_un.res.maxvif = vifi + 1; | |
599 | } | |
600 | } | |
601 | } | |
602 | ||
603 | static int mif6_add(struct mif6ctl *vifc, int mrtsock) | |
604 | { | |
605 | int vifi = vifc->mif6c_mifi; | |
4e16880c | 606 | struct mif_device *v = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 | 607 | struct net_device *dev; |
5ae7b444 | 608 | int err; |
7bc570c8 YH |
609 | |
610 | /* Is vif busy ? */ | |
4e16880c | 611 | if (MIF_EXISTS(&init_net, vifi)) |
7bc570c8 YH |
612 | return -EADDRINUSE; |
613 | ||
614 | switch (vifc->mif6c_flags) { | |
14fb64e1 YH |
615 | #ifdef CONFIG_IPV6_PIMSM_V2 |
616 | case MIFF_REGISTER: | |
617 | /* | |
618 | * Special Purpose VIF in PIM | |
619 | * All the packets will be sent to the daemon | |
620 | */ | |
621 | if (reg_vif_num >= 0) | |
622 | return -EADDRINUSE; | |
623 | dev = ip6mr_reg_vif(); | |
624 | if (!dev) | |
625 | return -ENOBUFS; | |
5ae7b444 WC |
626 | err = dev_set_allmulti(dev, 1); |
627 | if (err) { | |
628 | unregister_netdevice(dev); | |
7af3db78 | 629 | dev_put(dev); |
5ae7b444 WC |
630 | return err; |
631 | } | |
14fb64e1 YH |
632 | break; |
633 | #endif | |
7bc570c8 YH |
634 | case 0: |
635 | dev = dev_get_by_index(&init_net, vifc->mif6c_pifi); | |
636 | if (!dev) | |
637 | return -EADDRNOTAVAIL; | |
5ae7b444 | 638 | err = dev_set_allmulti(dev, 1); |
7af3db78 WC |
639 | if (err) { |
640 | dev_put(dev); | |
5ae7b444 | 641 | return err; |
7af3db78 | 642 | } |
7bc570c8 YH |
643 | break; |
644 | default: | |
645 | return -EINVAL; | |
646 | } | |
647 | ||
7bc570c8 YH |
648 | /* |
649 | * Fill in the VIF structures | |
650 | */ | |
651 | v->rate_limit = vifc->vifc_rate_limit; | |
652 | v->flags = vifc->mif6c_flags; | |
653 | if (!mrtsock) | |
654 | v->flags |= VIFF_STATIC; | |
655 | v->threshold = vifc->vifc_threshold; | |
656 | v->bytes_in = 0; | |
657 | v->bytes_out = 0; | |
658 | v->pkt_in = 0; | |
659 | v->pkt_out = 0; | |
660 | v->link = dev->ifindex; | |
661 | if (v->flags & MIFF_REGISTER) | |
662 | v->link = dev->iflink; | |
663 | ||
664 | /* And finish update writing critical data */ | |
665 | write_lock_bh(&mrt_lock); | |
7bc570c8 | 666 | v->dev = dev; |
14fb64e1 YH |
667 | #ifdef CONFIG_IPV6_PIMSM_V2 |
668 | if (v->flags & MIFF_REGISTER) | |
669 | reg_vif_num = vifi; | |
670 | #endif | |
4e16880c BT |
671 | if (vifi + 1 > init_net.ipv6.maxvif) |
672 | init_net.ipv6.maxvif = vifi + 1; | |
7bc570c8 YH |
673 | write_unlock_bh(&mrt_lock); |
674 | return 0; | |
675 | } | |
676 | ||
677 | static struct mfc6_cache *ip6mr_cache_find(struct in6_addr *origin, struct in6_addr *mcastgrp) | |
678 | { | |
679 | int line = MFC6_HASH(mcastgrp, origin); | |
680 | struct mfc6_cache *c; | |
681 | ||
4a6258a0 | 682 | for (c = init_net.ipv6.mfc6_cache_array[line]; c; c = c->next) { |
7bc570c8 YH |
683 | if (ipv6_addr_equal(&c->mf6c_origin, origin) && |
684 | ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) | |
685 | break; | |
686 | } | |
687 | return c; | |
688 | } | |
689 | ||
690 | /* | |
691 | * Allocate a multicast cache entry | |
692 | */ | |
58701ad4 | 693 | static struct mfc6_cache *ip6mr_cache_alloc(struct net *net) |
7bc570c8 | 694 | { |
36cbac59 | 695 | struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); |
7bc570c8 YH |
696 | if (c == NULL) |
697 | return NULL; | |
6ac7eb08 | 698 | c->mfc_un.res.minvif = MAXMIFS; |
58701ad4 | 699 | mfc6_net_set(c, net); |
7bc570c8 YH |
700 | return c; |
701 | } | |
702 | ||
58701ad4 | 703 | static struct mfc6_cache *ip6mr_cache_alloc_unres(struct net *net) |
7bc570c8 | 704 | { |
36cbac59 | 705 | struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); |
7bc570c8 YH |
706 | if (c == NULL) |
707 | return NULL; | |
7bc570c8 YH |
708 | skb_queue_head_init(&c->mfc_un.unres.unresolved); |
709 | c->mfc_un.unres.expires = jiffies + 10 * HZ; | |
58701ad4 | 710 | mfc6_net_set(c, net); |
7bc570c8 YH |
711 | return c; |
712 | } | |
713 | ||
714 | /* | |
715 | * A cache entry has gone into a resolved state from queued | |
716 | */ | |
717 | ||
718 | static void ip6mr_cache_resolve(struct mfc6_cache *uc, struct mfc6_cache *c) | |
719 | { | |
720 | struct sk_buff *skb; | |
721 | ||
722 | /* | |
723 | * Play the pending entries through our router | |
724 | */ | |
725 | ||
726 | while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { | |
727 | if (ipv6_hdr(skb)->version == 0) { | |
728 | int err; | |
729 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); | |
730 | ||
731 | if (ip6mr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) { | |
549e028d | 732 | nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; |
7bc570c8 YH |
733 | } else { |
734 | nlh->nlmsg_type = NLMSG_ERROR; | |
735 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
736 | skb_trim(skb, nlh->nlmsg_len); | |
737 | ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE; | |
738 | } | |
739 | err = rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); | |
740 | } else | |
741 | ip6_mr_forward(skb, c); | |
742 | } | |
743 | } | |
744 | ||
745 | /* | |
746 | * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd | |
747 | * expects the following bizarre scheme. | |
748 | * | |
749 | * Called under mrt_lock. | |
750 | */ | |
751 | ||
6ac7eb08 | 752 | static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert) |
7bc570c8 YH |
753 | { |
754 | struct sk_buff *skb; | |
755 | struct mrt6msg *msg; | |
756 | int ret; | |
757 | ||
14fb64e1 YH |
758 | #ifdef CONFIG_IPV6_PIMSM_V2 |
759 | if (assert == MRT6MSG_WHOLEPKT) | |
760 | skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) | |
761 | +sizeof(*msg)); | |
762 | else | |
763 | #endif | |
764 | skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); | |
7bc570c8 YH |
765 | |
766 | if (!skb) | |
767 | return -ENOBUFS; | |
768 | ||
769 | /* I suppose that internal messages | |
770 | * do not require checksums */ | |
771 | ||
772 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
773 | ||
14fb64e1 YH |
774 | #ifdef CONFIG_IPV6_PIMSM_V2 |
775 | if (assert == MRT6MSG_WHOLEPKT) { | |
776 | /* Ugly, but we have no choice with this interface. | |
777 | Duplicate old header, fix length etc. | |
778 | And all this only to mangle msg->im6_msgtype and | |
779 | to set msg->im6_mbz to "mbz" :-) | |
780 | */ | |
781 | skb_push(skb, -skb_network_offset(pkt)); | |
782 | ||
783 | skb_push(skb, sizeof(*msg)); | |
784 | skb_reset_transport_header(skb); | |
785 | msg = (struct mrt6msg *)skb_transport_header(skb); | |
786 | msg->im6_mbz = 0; | |
787 | msg->im6_msgtype = MRT6MSG_WHOLEPKT; | |
788 | msg->im6_mif = reg_vif_num; | |
789 | msg->im6_pad = 0; | |
790 | ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr); | |
791 | ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr); | |
792 | ||
793 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
794 | } else | |
795 | #endif | |
796 | { | |
7bc570c8 YH |
797 | /* |
798 | * Copy the IP header | |
799 | */ | |
800 | ||
801 | skb_put(skb, sizeof(struct ipv6hdr)); | |
802 | skb_reset_network_header(skb); | |
803 | skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); | |
804 | ||
805 | /* | |
806 | * Add our header | |
807 | */ | |
808 | skb_put(skb, sizeof(*msg)); | |
809 | skb_reset_transport_header(skb); | |
810 | msg = (struct mrt6msg *)skb_transport_header(skb); | |
811 | ||
812 | msg->im6_mbz = 0; | |
813 | msg->im6_msgtype = assert; | |
6ac7eb08 | 814 | msg->im6_mif = mifi; |
7bc570c8 YH |
815 | msg->im6_pad = 0; |
816 | ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr); | |
817 | ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr); | |
818 | ||
819 | skb->dst = dst_clone(pkt->dst); | |
820 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
821 | ||
822 | skb_pull(skb, sizeof(struct ipv6hdr)); | |
14fb64e1 | 823 | } |
7bc570c8 | 824 | |
bd91b8bf | 825 | if (init_net.ipv6.mroute6_sk == NULL) { |
7bc570c8 YH |
826 | kfree_skb(skb); |
827 | return -EINVAL; | |
828 | } | |
829 | ||
830 | /* | |
831 | * Deliver to user space multicast routing algorithms | |
832 | */ | |
bd91b8bf BT |
833 | ret = sock_queue_rcv_skb(init_net.ipv6.mroute6_sk, skb); |
834 | if (ret < 0) { | |
7bc570c8 YH |
835 | if (net_ratelimit()) |
836 | printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n"); | |
837 | kfree_skb(skb); | |
838 | } | |
839 | ||
840 | return ret; | |
841 | } | |
842 | ||
843 | /* | |
844 | * Queue a packet for resolution. It gets locked cache entry! | |
845 | */ | |
846 | ||
847 | static int | |
6ac7eb08 | 848 | ip6mr_cache_unresolved(mifi_t mifi, struct sk_buff *skb) |
7bc570c8 YH |
849 | { |
850 | int err; | |
851 | struct mfc6_cache *c; | |
852 | ||
853 | spin_lock_bh(&mfc_unres_lock); | |
854 | for (c = mfc_unres_queue; c; c = c->next) { | |
855 | if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && | |
856 | ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) | |
857 | break; | |
858 | } | |
859 | ||
860 | if (c == NULL) { | |
861 | /* | |
862 | * Create a new entry if allowable | |
863 | */ | |
864 | ||
865 | if (atomic_read(&cache_resolve_queue_len) >= 10 || | |
58701ad4 | 866 | (c = ip6mr_cache_alloc_unres(&init_net)) == NULL) { |
7bc570c8 YH |
867 | spin_unlock_bh(&mfc_unres_lock); |
868 | ||
869 | kfree_skb(skb); | |
870 | return -ENOBUFS; | |
871 | } | |
872 | ||
873 | /* | |
874 | * Fill in the new cache entry | |
875 | */ | |
876 | c->mf6c_parent = -1; | |
877 | c->mf6c_origin = ipv6_hdr(skb)->saddr; | |
878 | c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; | |
879 | ||
880 | /* | |
881 | * Reflect first query at pim6sd | |
882 | */ | |
6ac7eb08 | 883 | if ((err = ip6mr_cache_report(skb, mifi, MRT6MSG_NOCACHE)) < 0) { |
7bc570c8 YH |
884 | /* If the report failed throw the cache entry |
885 | out - Brad Parker | |
886 | */ | |
887 | spin_unlock_bh(&mfc_unres_lock); | |
888 | ||
58701ad4 | 889 | ip6mr_cache_free(c); |
7bc570c8 YH |
890 | kfree_skb(skb); |
891 | return err; | |
892 | } | |
893 | ||
894 | atomic_inc(&cache_resolve_queue_len); | |
895 | c->next = mfc_unres_queue; | |
896 | mfc_unres_queue = c; | |
897 | ||
898 | ipmr_do_expire_process(1); | |
899 | } | |
900 | ||
901 | /* | |
902 | * See if we can append the packet | |
903 | */ | |
904 | if (c->mfc_un.unres.unresolved.qlen > 3) { | |
905 | kfree_skb(skb); | |
906 | err = -ENOBUFS; | |
907 | } else { | |
908 | skb_queue_tail(&c->mfc_un.unres.unresolved, skb); | |
909 | err = 0; | |
910 | } | |
911 | ||
912 | spin_unlock_bh(&mfc_unres_lock); | |
913 | return err; | |
914 | } | |
915 | ||
916 | /* | |
917 | * MFC6 cache manipulation by user space | |
918 | */ | |
919 | ||
920 | static int ip6mr_mfc_delete(struct mf6cctl *mfc) | |
921 | { | |
922 | int line; | |
923 | struct mfc6_cache *c, **cp; | |
924 | ||
925 | line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); | |
926 | ||
4a6258a0 BT |
927 | for (cp = &init_net.ipv6.mfc6_cache_array[line]; |
928 | (c = *cp) != NULL; cp = &c->next) { | |
7bc570c8 YH |
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 | } | |
4a6258a0 BT |
981 | |
982 | /* Forwarding cache */ | |
983 | net->ipv6.mfc6_cache_array = kcalloc(MFC6_LINES, | |
984 | sizeof(struct mfc6_cache *), | |
985 | GFP_KERNEL); | |
986 | if (!net->ipv6.mfc6_cache_array) { | |
987 | err = -ENOMEM; | |
988 | goto fail_mfc6_cache; | |
989 | } | |
990 | return 0; | |
991 | ||
992 | fail_mfc6_cache: | |
993 | kfree(net->ipv6.vif6_table); | |
4e16880c BT |
994 | fail: |
995 | return err; | |
996 | } | |
997 | ||
998 | static void __net_exit ip6mr_net_exit(struct net *net) | |
999 | { | |
4a6258a0 | 1000 | kfree(net->ipv6.mfc6_cache_array); |
4e16880c BT |
1001 | kfree(net->ipv6.vif6_table); |
1002 | } | |
1003 | ||
1004 | static struct pernet_operations ip6mr_net_ops = { | |
1005 | .init = ip6mr_net_init, | |
1006 | .exit = ip6mr_net_exit, | |
1007 | }; | |
1008 | ||
623d1a1a | 1009 | int __init ip6_mr_init(void) |
7bc570c8 | 1010 | { |
623d1a1a WC |
1011 | int err; |
1012 | ||
7bc570c8 YH |
1013 | mrt_cachep = kmem_cache_create("ip6_mrt_cache", |
1014 | sizeof(struct mfc6_cache), | |
1015 | 0, SLAB_HWCACHE_ALIGN, | |
1016 | NULL); | |
1017 | if (!mrt_cachep) | |
623d1a1a | 1018 | return -ENOMEM; |
7bc570c8 | 1019 | |
4e16880c BT |
1020 | err = register_pernet_subsys(&ip6mr_net_ops); |
1021 | if (err) | |
1022 | goto reg_pernet_fail; | |
1023 | ||
7bc570c8 | 1024 | setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0); |
623d1a1a WC |
1025 | err = register_netdevice_notifier(&ip6_mr_notifier); |
1026 | if (err) | |
1027 | goto reg_notif_fail; | |
1028 | #ifdef CONFIG_PROC_FS | |
1029 | err = -ENOMEM; | |
1030 | if (!proc_net_fops_create(&init_net, "ip6_mr_vif", 0, &ip6mr_vif_fops)) | |
1031 | goto proc_vif_fail; | |
1032 | if (!proc_net_fops_create(&init_net, "ip6_mr_cache", | |
1033 | 0, &ip6mr_mfc_fops)) | |
1034 | goto proc_cache_fail; | |
1035 | #endif | |
1036 | return 0; | |
7bc570c8 | 1037 | #ifdef CONFIG_PROC_FS |
623d1a1a WC |
1038 | proc_cache_fail: |
1039 | proc_net_remove(&init_net, "ip6_mr_vif"); | |
87b30a65 BT |
1040 | proc_vif_fail: |
1041 | unregister_netdevice_notifier(&ip6_mr_notifier); | |
7bc570c8 | 1042 | #endif |
87b30a65 BT |
1043 | reg_notif_fail: |
1044 | del_timer(&ipmr_expire_timer); | |
4e16880c BT |
1045 | unregister_pernet_subsys(&ip6mr_net_ops); |
1046 | reg_pernet_fail: | |
87b30a65 | 1047 | kmem_cache_destroy(mrt_cachep); |
623d1a1a | 1048 | return err; |
7bc570c8 YH |
1049 | } |
1050 | ||
623d1a1a WC |
1051 | void ip6_mr_cleanup(void) |
1052 | { | |
1053 | #ifdef CONFIG_PROC_FS | |
1054 | proc_net_remove(&init_net, "ip6_mr_cache"); | |
1055 | proc_net_remove(&init_net, "ip6_mr_vif"); | |
1056 | #endif | |
1057 | unregister_netdevice_notifier(&ip6_mr_notifier); | |
1058 | del_timer(&ipmr_expire_timer); | |
4e16880c | 1059 | unregister_pernet_subsys(&ip6mr_net_ops); |
623d1a1a WC |
1060 | kmem_cache_destroy(mrt_cachep); |
1061 | } | |
7bc570c8 YH |
1062 | |
1063 | static int ip6mr_mfc_add(struct mf6cctl *mfc, int mrtsock) | |
1064 | { | |
1065 | int line; | |
1066 | struct mfc6_cache *uc, *c, **cp; | |
6ac7eb08 | 1067 | unsigned char ttls[MAXMIFS]; |
7bc570c8 YH |
1068 | int i; |
1069 | ||
6ac7eb08 RR |
1070 | memset(ttls, 255, MAXMIFS); |
1071 | for (i = 0; i < MAXMIFS; i++) { | |
7bc570c8 YH |
1072 | if (IF_ISSET(i, &mfc->mf6cc_ifset)) |
1073 | ttls[i] = 1; | |
1074 | ||
1075 | } | |
1076 | ||
1077 | line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); | |
1078 | ||
4a6258a0 BT |
1079 | for (cp = &init_net.ipv6.mfc6_cache_array[line]; |
1080 | (c = *cp) != NULL; cp = &c->next) { | |
7bc570c8 YH |
1081 | if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && |
1082 | ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) | |
1083 | break; | |
1084 | } | |
1085 | ||
1086 | if (c != NULL) { | |
1087 | write_lock_bh(&mrt_lock); | |
1088 | c->mf6c_parent = mfc->mf6cc_parent; | |
1089 | ip6mr_update_thresholds(c, ttls); | |
1090 | if (!mrtsock) | |
1091 | c->mfc_flags |= MFC_STATIC; | |
1092 | write_unlock_bh(&mrt_lock); | |
1093 | return 0; | |
1094 | } | |
1095 | ||
1096 | if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) | |
1097 | return -EINVAL; | |
1098 | ||
58701ad4 | 1099 | c = ip6mr_cache_alloc(&init_net); |
7bc570c8 YH |
1100 | if (c == NULL) |
1101 | return -ENOMEM; | |
1102 | ||
1103 | c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; | |
1104 | c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; | |
1105 | c->mf6c_parent = mfc->mf6cc_parent; | |
1106 | ip6mr_update_thresholds(c, ttls); | |
1107 | if (!mrtsock) | |
1108 | c->mfc_flags |= MFC_STATIC; | |
1109 | ||
1110 | write_lock_bh(&mrt_lock); | |
4a6258a0 BT |
1111 | c->next = init_net.ipv6.mfc6_cache_array[line]; |
1112 | init_net.ipv6.mfc6_cache_array[line] = c; | |
7bc570c8 YH |
1113 | write_unlock_bh(&mrt_lock); |
1114 | ||
1115 | /* | |
1116 | * Check to see if we resolved a queued list. If so we | |
1117 | * need to send on the frames and tidy up. | |
1118 | */ | |
1119 | spin_lock_bh(&mfc_unres_lock); | |
1120 | for (cp = &mfc_unres_queue; (uc = *cp) != NULL; | |
1121 | cp = &uc->next) { | |
1122 | if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && | |
1123 | ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { | |
1124 | *cp = uc->next; | |
1125 | if (atomic_dec_and_test(&cache_resolve_queue_len)) | |
1126 | del_timer(&ipmr_expire_timer); | |
1127 | break; | |
1128 | } | |
1129 | } | |
1130 | spin_unlock_bh(&mfc_unres_lock); | |
1131 | ||
1132 | if (uc) { | |
1133 | ip6mr_cache_resolve(uc, c); | |
58701ad4 | 1134 | ip6mr_cache_free(uc); |
7bc570c8 YH |
1135 | } |
1136 | return 0; | |
1137 | } | |
1138 | ||
1139 | /* | |
1140 | * Close the multicast socket, and clear the vif tables etc | |
1141 | */ | |
1142 | ||
1143 | static void mroute_clean_tables(struct sock *sk) | |
1144 | { | |
1145 | int i; | |
1146 | ||
1147 | /* | |
1148 | * Shut down all active vif entries | |
1149 | */ | |
4e16880c BT |
1150 | for (i = 0; i < init_net.ipv6.maxvif; i++) { |
1151 | if (!(init_net.ipv6.vif6_table[i].flags & VIFF_STATIC)) | |
7bc570c8 YH |
1152 | mif6_delete(i); |
1153 | } | |
1154 | ||
1155 | /* | |
1156 | * Wipe the cache | |
1157 | */ | |
4a6258a0 | 1158 | for (i = 0; i < MFC6_LINES; i++) { |
7bc570c8 YH |
1159 | struct mfc6_cache *c, **cp; |
1160 | ||
4a6258a0 | 1161 | cp = &init_net.ipv6.mfc6_cache_array[i]; |
7bc570c8 YH |
1162 | while ((c = *cp) != NULL) { |
1163 | if (c->mfc_flags & MFC_STATIC) { | |
1164 | cp = &c->next; | |
1165 | continue; | |
1166 | } | |
1167 | write_lock_bh(&mrt_lock); | |
1168 | *cp = c->next; | |
1169 | write_unlock_bh(&mrt_lock); | |
1170 | ||
58701ad4 | 1171 | ip6mr_cache_free(c); |
7bc570c8 YH |
1172 | } |
1173 | } | |
1174 | ||
1175 | if (atomic_read(&cache_resolve_queue_len) != 0) { | |
1176 | struct mfc6_cache *c; | |
1177 | ||
1178 | spin_lock_bh(&mfc_unres_lock); | |
1179 | while (mfc_unres_queue != NULL) { | |
1180 | c = mfc_unres_queue; | |
1181 | mfc_unres_queue = c->next; | |
1182 | spin_unlock_bh(&mfc_unres_lock); | |
1183 | ||
1184 | ip6mr_destroy_unres(c); | |
1185 | ||
1186 | spin_lock_bh(&mfc_unres_lock); | |
1187 | } | |
1188 | spin_unlock_bh(&mfc_unres_lock); | |
1189 | } | |
1190 | } | |
1191 | ||
1192 | static int ip6mr_sk_init(struct sock *sk) | |
1193 | { | |
1194 | int err = 0; | |
1195 | ||
1196 | rtnl_lock(); | |
1197 | write_lock_bh(&mrt_lock); | |
bd91b8bf BT |
1198 | if (likely(init_net.ipv6.mroute6_sk == NULL)) |
1199 | init_net.ipv6.mroute6_sk = sk; | |
7bc570c8 YH |
1200 | else |
1201 | err = -EADDRINUSE; | |
1202 | write_unlock_bh(&mrt_lock); | |
1203 | ||
1204 | rtnl_unlock(); | |
1205 | ||
1206 | return err; | |
1207 | } | |
1208 | ||
1209 | int ip6mr_sk_done(struct sock *sk) | |
1210 | { | |
1211 | int err = 0; | |
1212 | ||
1213 | rtnl_lock(); | |
bd91b8bf | 1214 | if (sk == init_net.ipv6.mroute6_sk) { |
7bc570c8 | 1215 | write_lock_bh(&mrt_lock); |
bd91b8bf | 1216 | init_net.ipv6.mroute6_sk = NULL; |
7bc570c8 YH |
1217 | write_unlock_bh(&mrt_lock); |
1218 | ||
1219 | mroute_clean_tables(sk); | |
1220 | } else | |
1221 | err = -EACCES; | |
1222 | rtnl_unlock(); | |
1223 | ||
1224 | return err; | |
1225 | } | |
1226 | ||
1227 | /* | |
1228 | * Socket options and virtual interface manipulation. The whole | |
1229 | * virtual interface system is a complete heap, but unfortunately | |
1230 | * that's how BSD mrouted happens to think. Maybe one day with a proper | |
1231 | * MOSPF/PIM router set up we can clean this up. | |
1232 | */ | |
1233 | ||
1234 | int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, int optlen) | |
1235 | { | |
1236 | int ret; | |
1237 | struct mif6ctl vif; | |
1238 | struct mf6cctl mfc; | |
1239 | mifi_t mifi; | |
1240 | ||
1241 | if (optname != MRT6_INIT) { | |
bd91b8bf | 1242 | if (sk != init_net.ipv6.mroute6_sk && !capable(CAP_NET_ADMIN)) |
7bc570c8 YH |
1243 | return -EACCES; |
1244 | } | |
1245 | ||
1246 | switch (optname) { | |
1247 | case MRT6_INIT: | |
1248 | if (sk->sk_type != SOCK_RAW || | |
1249 | inet_sk(sk)->num != IPPROTO_ICMPV6) | |
1250 | return -EOPNOTSUPP; | |
1251 | if (optlen < sizeof(int)) | |
1252 | return -EINVAL; | |
1253 | ||
1254 | return ip6mr_sk_init(sk); | |
1255 | ||
1256 | case MRT6_DONE: | |
1257 | return ip6mr_sk_done(sk); | |
1258 | ||
1259 | case MRT6_ADD_MIF: | |
1260 | if (optlen < sizeof(vif)) | |
1261 | return -EINVAL; | |
1262 | if (copy_from_user(&vif, optval, sizeof(vif))) | |
1263 | return -EFAULT; | |
6ac7eb08 | 1264 | if (vif.mif6c_mifi >= MAXMIFS) |
7bc570c8 YH |
1265 | return -ENFILE; |
1266 | rtnl_lock(); | |
bd91b8bf | 1267 | ret = mif6_add(&vif, sk == init_net.ipv6.mroute6_sk); |
7bc570c8 YH |
1268 | rtnl_unlock(); |
1269 | return ret; | |
1270 | ||
1271 | case MRT6_DEL_MIF: | |
1272 | if (optlen < sizeof(mifi_t)) | |
1273 | return -EINVAL; | |
1274 | if (copy_from_user(&mifi, optval, sizeof(mifi_t))) | |
1275 | return -EFAULT; | |
1276 | rtnl_lock(); | |
1277 | ret = mif6_delete(mifi); | |
1278 | rtnl_unlock(); | |
1279 | return ret; | |
1280 | ||
1281 | /* | |
1282 | * Manipulate the forwarding caches. These live | |
1283 | * in a sort of kernel/user symbiosis. | |
1284 | */ | |
1285 | case MRT6_ADD_MFC: | |
1286 | case MRT6_DEL_MFC: | |
1287 | if (optlen < sizeof(mfc)) | |
1288 | return -EINVAL; | |
1289 | if (copy_from_user(&mfc, optval, sizeof(mfc))) | |
1290 | return -EFAULT; | |
1291 | rtnl_lock(); | |
1292 | if (optname == MRT6_DEL_MFC) | |
1293 | ret = ip6mr_mfc_delete(&mfc); | |
1294 | else | |
bd91b8bf | 1295 | ret = ip6mr_mfc_add(&mfc, sk == init_net.ipv6.mroute6_sk); |
7bc570c8 YH |
1296 | rtnl_unlock(); |
1297 | return ret; | |
1298 | ||
14fb64e1 YH |
1299 | /* |
1300 | * Control PIM assert (to activate pim will activate assert) | |
1301 | */ | |
1302 | case MRT6_ASSERT: | |
1303 | { | |
1304 | int v; | |
1305 | if (get_user(v, (int __user *)optval)) | |
1306 | return -EFAULT; | |
1307 | mroute_do_assert = !!v; | |
1308 | return 0; | |
1309 | } | |
1310 | ||
1311 | #ifdef CONFIG_IPV6_PIMSM_V2 | |
1312 | case MRT6_PIM: | |
1313 | { | |
a9f83bf3 | 1314 | int v; |
14fb64e1 YH |
1315 | if (get_user(v, (int __user *)optval)) |
1316 | return -EFAULT; | |
1317 | v = !!v; | |
1318 | rtnl_lock(); | |
1319 | ret = 0; | |
1320 | if (v != mroute_do_pim) { | |
1321 | mroute_do_pim = v; | |
1322 | mroute_do_assert = v; | |
1323 | if (mroute_do_pim) | |
1324 | ret = inet6_add_protocol(&pim6_protocol, | |
1325 | IPPROTO_PIM); | |
1326 | else | |
1327 | ret = inet6_del_protocol(&pim6_protocol, | |
1328 | IPPROTO_PIM); | |
1329 | if (ret < 0) | |
1330 | ret = -EAGAIN; | |
1331 | } | |
1332 | rtnl_unlock(); | |
1333 | return ret; | |
1334 | } | |
1335 | ||
1336 | #endif | |
7bc570c8 | 1337 | /* |
7d120c55 | 1338 | * Spurious command, or MRT6_VERSION which you cannot |
7bc570c8 YH |
1339 | * set. |
1340 | */ | |
1341 | default: | |
1342 | return -ENOPROTOOPT; | |
1343 | } | |
1344 | } | |
1345 | ||
1346 | /* | |
1347 | * Getsock opt support for the multicast routing system. | |
1348 | */ | |
1349 | ||
1350 | int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, | |
1351 | int __user *optlen) | |
1352 | { | |
1353 | int olr; | |
1354 | int val; | |
1355 | ||
1356 | switch (optname) { | |
1357 | case MRT6_VERSION: | |
1358 | val = 0x0305; | |
1359 | break; | |
14fb64e1 YH |
1360 | #ifdef CONFIG_IPV6_PIMSM_V2 |
1361 | case MRT6_PIM: | |
1362 | val = mroute_do_pim; | |
1363 | break; | |
1364 | #endif | |
1365 | case MRT6_ASSERT: | |
1366 | val = mroute_do_assert; | |
1367 | break; | |
7bc570c8 YH |
1368 | default: |
1369 | return -ENOPROTOOPT; | |
1370 | } | |
1371 | ||
1372 | if (get_user(olr, optlen)) | |
1373 | return -EFAULT; | |
1374 | ||
1375 | olr = min_t(int, olr, sizeof(int)); | |
1376 | if (olr < 0) | |
1377 | return -EINVAL; | |
1378 | ||
1379 | if (put_user(olr, optlen)) | |
1380 | return -EFAULT; | |
1381 | if (copy_to_user(optval, &val, olr)) | |
1382 | return -EFAULT; | |
1383 | return 0; | |
1384 | } | |
1385 | ||
1386 | /* | |
1387 | * The IP multicast ioctl support routines. | |
1388 | */ | |
1389 | ||
1390 | int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) | |
1391 | { | |
1392 | struct sioc_sg_req6 sr; | |
1393 | struct sioc_mif_req6 vr; | |
1394 | struct mif_device *vif; | |
1395 | struct mfc6_cache *c; | |
1396 | ||
1397 | switch (cmd) { | |
1398 | case SIOCGETMIFCNT_IN6: | |
1399 | if (copy_from_user(&vr, arg, sizeof(vr))) | |
1400 | return -EFAULT; | |
4e16880c | 1401 | if (vr.mifi >= init_net.ipv6.maxvif) |
7bc570c8 YH |
1402 | return -EINVAL; |
1403 | read_lock(&mrt_lock); | |
4e16880c BT |
1404 | vif = &init_net.ipv6.vif6_table[vr.mifi]; |
1405 | if (MIF_EXISTS(&init_net, vr.mifi)) { | |
7bc570c8 YH |
1406 | vr.icount = vif->pkt_in; |
1407 | vr.ocount = vif->pkt_out; | |
1408 | vr.ibytes = vif->bytes_in; | |
1409 | vr.obytes = vif->bytes_out; | |
1410 | read_unlock(&mrt_lock); | |
1411 | ||
1412 | if (copy_to_user(arg, &vr, sizeof(vr))) | |
1413 | return -EFAULT; | |
1414 | return 0; | |
1415 | } | |
1416 | read_unlock(&mrt_lock); | |
1417 | return -EADDRNOTAVAIL; | |
1418 | case SIOCGETSGCNT_IN6: | |
1419 | if (copy_from_user(&sr, arg, sizeof(sr))) | |
1420 | return -EFAULT; | |
1421 | ||
1422 | read_lock(&mrt_lock); | |
1423 | c = ip6mr_cache_find(&sr.src.sin6_addr, &sr.grp.sin6_addr); | |
1424 | if (c) { | |
1425 | sr.pktcnt = c->mfc_un.res.pkt; | |
1426 | sr.bytecnt = c->mfc_un.res.bytes; | |
1427 | sr.wrong_if = c->mfc_un.res.wrong_if; | |
1428 | read_unlock(&mrt_lock); | |
1429 | ||
1430 | if (copy_to_user(arg, &sr, sizeof(sr))) | |
1431 | return -EFAULT; | |
1432 | return 0; | |
1433 | } | |
1434 | read_unlock(&mrt_lock); | |
1435 | return -EADDRNOTAVAIL; | |
1436 | default: | |
1437 | return -ENOIOCTLCMD; | |
1438 | } | |
1439 | } | |
1440 | ||
1441 | ||
1442 | static inline int ip6mr_forward2_finish(struct sk_buff *skb) | |
1443 | { | |
483a47d2 DL |
1444 | IP6_INC_STATS_BH(dev_net(skb->dst->dev), ip6_dst_idev(skb->dst), |
1445 | IPSTATS_MIB_OUTFORWDATAGRAMS); | |
7bc570c8 YH |
1446 | return dst_output(skb); |
1447 | } | |
1448 | ||
1449 | /* | |
1450 | * Processing handlers for ip6mr_forward | |
1451 | */ | |
1452 | ||
1453 | static int ip6mr_forward2(struct sk_buff *skb, struct mfc6_cache *c, int vifi) | |
1454 | { | |
1455 | struct ipv6hdr *ipv6h; | |
4e16880c | 1456 | struct mif_device *vif = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 YH |
1457 | struct net_device *dev; |
1458 | struct dst_entry *dst; | |
1459 | struct flowi fl; | |
1460 | ||
1461 | if (vif->dev == NULL) | |
1462 | goto out_free; | |
1463 | ||
14fb64e1 YH |
1464 | #ifdef CONFIG_IPV6_PIMSM_V2 |
1465 | if (vif->flags & MIFF_REGISTER) { | |
1466 | vif->pkt_out++; | |
1467 | vif->bytes_out += skb->len; | |
dc58c78c PE |
1468 | vif->dev->stats.tx_bytes += skb->len; |
1469 | vif->dev->stats.tx_packets++; | |
14fb64e1 YH |
1470 | ip6mr_cache_report(skb, vifi, MRT6MSG_WHOLEPKT); |
1471 | kfree_skb(skb); | |
1472 | return 0; | |
1473 | } | |
1474 | #endif | |
1475 | ||
7bc570c8 YH |
1476 | ipv6h = ipv6_hdr(skb); |
1477 | ||
1478 | fl = (struct flowi) { | |
1479 | .oif = vif->link, | |
1480 | .nl_u = { .ip6_u = | |
1481 | { .daddr = ipv6h->daddr, } | |
1482 | } | |
1483 | }; | |
1484 | ||
1485 | dst = ip6_route_output(&init_net, NULL, &fl); | |
1486 | if (!dst) | |
1487 | goto out_free; | |
1488 | ||
1489 | dst_release(skb->dst); | |
1490 | skb->dst = dst; | |
1491 | ||
1492 | /* | |
1493 | * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally | |
1494 | * not only before forwarding, but after forwarding on all output | |
1495 | * interfaces. It is clear, if mrouter runs a multicasting | |
1496 | * program, it should receive packets not depending to what interface | |
1497 | * program is joined. | |
1498 | * If we will not make it, the program will have to join on all | |
1499 | * interfaces. On the other hand, multihoming host (or router, but | |
1500 | * not mrouter) cannot join to more than one interface - it will | |
1501 | * result in receiving multiple packets. | |
1502 | */ | |
1503 | dev = vif->dev; | |
1504 | skb->dev = dev; | |
1505 | vif->pkt_out++; | |
1506 | vif->bytes_out += skb->len; | |
1507 | ||
1508 | /* We are about to write */ | |
1509 | /* XXX: extension headers? */ | |
1510 | if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) | |
1511 | goto out_free; | |
1512 | ||
1513 | ipv6h = ipv6_hdr(skb); | |
1514 | ipv6h->hop_limit--; | |
1515 | ||
1516 | IP6CB(skb)->flags |= IP6SKB_FORWARDED; | |
1517 | ||
1518 | return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dev, | |
1519 | ip6mr_forward2_finish); | |
1520 | ||
1521 | out_free: | |
1522 | kfree_skb(skb); | |
1523 | return 0; | |
1524 | } | |
1525 | ||
1526 | static int ip6mr_find_vif(struct net_device *dev) | |
1527 | { | |
1528 | int ct; | |
4e16880c BT |
1529 | for (ct = init_net.ipv6.maxvif - 1; ct >= 0; ct--) { |
1530 | if (init_net.ipv6.vif6_table[ct].dev == dev) | |
7bc570c8 YH |
1531 | break; |
1532 | } | |
1533 | return ct; | |
1534 | } | |
1535 | ||
1536 | static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache) | |
1537 | { | |
1538 | int psend = -1; | |
1539 | int vif, ct; | |
1540 | ||
1541 | vif = cache->mf6c_parent; | |
1542 | cache->mfc_un.res.pkt++; | |
1543 | cache->mfc_un.res.bytes += skb->len; | |
1544 | ||
14fb64e1 YH |
1545 | /* |
1546 | * Wrong interface: drop packet and (maybe) send PIM assert. | |
1547 | */ | |
4e16880c | 1548 | if (init_net.ipv6.vif6_table[vif].dev != skb->dev) { |
14fb64e1 YH |
1549 | int true_vifi; |
1550 | ||
1551 | cache->mfc_un.res.wrong_if++; | |
1552 | true_vifi = ip6mr_find_vif(skb->dev); | |
1553 | ||
1554 | if (true_vifi >= 0 && mroute_do_assert && | |
1555 | /* pimsm uses asserts, when switching from RPT to SPT, | |
1556 | so that we cannot check that packet arrived on an oif. | |
1557 | It is bad, but otherwise we would need to move pretty | |
1558 | large chunk of pimd to kernel. Ough... --ANK | |
1559 | */ | |
1560 | (mroute_do_pim || cache->mfc_un.res.ttls[true_vifi] < 255) && | |
1561 | time_after(jiffies, | |
1562 | cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { | |
1563 | cache->mfc_un.res.last_assert = jiffies; | |
1564 | ip6mr_cache_report(skb, true_vifi, MRT6MSG_WRONGMIF); | |
1565 | } | |
1566 | goto dont_forward; | |
1567 | } | |
1568 | ||
4e16880c BT |
1569 | init_net.ipv6.vif6_table[vif].pkt_in++; |
1570 | init_net.ipv6.vif6_table[vif].bytes_in += skb->len; | |
7bc570c8 YH |
1571 | |
1572 | /* | |
1573 | * Forward the frame | |
1574 | */ | |
1575 | for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { | |
1576 | if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { | |
1577 | if (psend != -1) { | |
1578 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1579 | if (skb2) | |
1580 | ip6mr_forward2(skb2, cache, psend); | |
1581 | } | |
1582 | psend = ct; | |
1583 | } | |
1584 | } | |
1585 | if (psend != -1) { | |
1586 | ip6mr_forward2(skb, cache, psend); | |
1587 | return 0; | |
1588 | } | |
1589 | ||
14fb64e1 | 1590 | dont_forward: |
7bc570c8 YH |
1591 | kfree_skb(skb); |
1592 | return 0; | |
1593 | } | |
1594 | ||
1595 | ||
1596 | /* | |
1597 | * Multicast packets for forwarding arrive here | |
1598 | */ | |
1599 | ||
1600 | int ip6_mr_input(struct sk_buff *skb) | |
1601 | { | |
1602 | struct mfc6_cache *cache; | |
1603 | ||
1604 | read_lock(&mrt_lock); | |
1605 | cache = ip6mr_cache_find(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); | |
1606 | ||
1607 | /* | |
1608 | * No usable cache entry | |
1609 | */ | |
1610 | if (cache == NULL) { | |
1611 | int vif; | |
1612 | ||
1613 | vif = ip6mr_find_vif(skb->dev); | |
1614 | if (vif >= 0) { | |
1615 | int err = ip6mr_cache_unresolved(vif, skb); | |
1616 | read_unlock(&mrt_lock); | |
1617 | ||
1618 | return err; | |
1619 | } | |
1620 | read_unlock(&mrt_lock); | |
1621 | kfree_skb(skb); | |
1622 | return -ENODEV; | |
1623 | } | |
1624 | ||
1625 | ip6_mr_forward(skb, cache); | |
1626 | ||
1627 | read_unlock(&mrt_lock); | |
1628 | ||
1629 | return 0; | |
1630 | } | |
1631 | ||
1632 | ||
1633 | static int | |
1634 | ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm) | |
1635 | { | |
1636 | int ct; | |
1637 | struct rtnexthop *nhp; | |
4e16880c | 1638 | struct net_device *dev = init_net.ipv6.vif6_table[c->mf6c_parent].dev; |
549e028d | 1639 | u8 *b = skb_tail_pointer(skb); |
7bc570c8 YH |
1640 | struct rtattr *mp_head; |
1641 | ||
1642 | if (dev) | |
1643 | RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex); | |
1644 | ||
1645 | mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0)); | |
1646 | ||
1647 | for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { | |
1648 | if (c->mfc_un.res.ttls[ct] < 255) { | |
1649 | if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4)) | |
1650 | goto rtattr_failure; | |
1651 | nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp))); | |
1652 | nhp->rtnh_flags = 0; | |
1653 | nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; | |
4e16880c | 1654 | nhp->rtnh_ifindex = init_net.ipv6.vif6_table[ct].dev->ifindex; |
7bc570c8 YH |
1655 | nhp->rtnh_len = sizeof(*nhp); |
1656 | } | |
1657 | } | |
1658 | mp_head->rta_type = RTA_MULTIPATH; | |
549e028d | 1659 | mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head; |
7bc570c8 YH |
1660 | rtm->rtm_type = RTN_MULTICAST; |
1661 | return 1; | |
1662 | ||
1663 | rtattr_failure: | |
1664 | nlmsg_trim(skb, b); | |
1665 | return -EMSGSIZE; | |
1666 | } | |
1667 | ||
1668 | int ip6mr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait) | |
1669 | { | |
1670 | int err; | |
1671 | struct mfc6_cache *cache; | |
1672 | struct rt6_info *rt = (struct rt6_info *)skb->dst; | |
1673 | ||
1674 | read_lock(&mrt_lock); | |
1675 | cache = ip6mr_cache_find(&rt->rt6i_src.addr, &rt->rt6i_dst.addr); | |
1676 | ||
1677 | if (!cache) { | |
1678 | struct sk_buff *skb2; | |
1679 | struct ipv6hdr *iph; | |
1680 | struct net_device *dev; | |
1681 | int vif; | |
1682 | ||
1683 | if (nowait) { | |
1684 | read_unlock(&mrt_lock); | |
1685 | return -EAGAIN; | |
1686 | } | |
1687 | ||
1688 | dev = skb->dev; | |
1689 | if (dev == NULL || (vif = ip6mr_find_vif(dev)) < 0) { | |
1690 | read_unlock(&mrt_lock); | |
1691 | return -ENODEV; | |
1692 | } | |
1693 | ||
1694 | /* really correct? */ | |
1695 | skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); | |
1696 | if (!skb2) { | |
1697 | read_unlock(&mrt_lock); | |
1698 | return -ENOMEM; | |
1699 | } | |
1700 | ||
1701 | skb_reset_transport_header(skb2); | |
1702 | ||
1703 | skb_put(skb2, sizeof(struct ipv6hdr)); | |
1704 | skb_reset_network_header(skb2); | |
1705 | ||
1706 | iph = ipv6_hdr(skb2); | |
1707 | iph->version = 0; | |
1708 | iph->priority = 0; | |
1709 | iph->flow_lbl[0] = 0; | |
1710 | iph->flow_lbl[1] = 0; | |
1711 | iph->flow_lbl[2] = 0; | |
1712 | iph->payload_len = 0; | |
1713 | iph->nexthdr = IPPROTO_NONE; | |
1714 | iph->hop_limit = 0; | |
1715 | ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr); | |
1716 | ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr); | |
1717 | ||
1718 | err = ip6mr_cache_unresolved(vif, skb2); | |
1719 | read_unlock(&mrt_lock); | |
1720 | ||
1721 | return err; | |
1722 | } | |
1723 | ||
1724 | if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY)) | |
1725 | cache->mfc_flags |= MFC_NOTIFY; | |
1726 | ||
1727 | err = ip6mr_fill_mroute(skb, cache, rtm); | |
1728 | read_unlock(&mrt_lock); | |
1729 | return err; | |
1730 | } | |
1731 |