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 | 71 | static struct mfc6_cache *mfc_unres_queue; /* Queue of unresolved entries */ |
7bc570c8 YH |
72 | |
73 | /* Special spinlock for queue of unresolved entries */ | |
74 | static DEFINE_SPINLOCK(mfc_unres_lock); | |
75 | ||
76 | /* We return to original Alan's scheme. Hash table of resolved | |
77 | entries is changed only in process context and protected | |
78 | with weak lock mrt_lock. Queue of unresolved entries is protected | |
79 | with strong spinlock mfc_unres_lock. | |
80 | ||
81 | In this case data path is free of exclusive locks at all. | |
82 | */ | |
83 | ||
84 | static struct kmem_cache *mrt_cachep __read_mostly; | |
85 | ||
86 | static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache); | |
6ac7eb08 | 87 | static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert); |
7bc570c8 YH |
88 | static int ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm); |
89 | ||
14fb64e1 YH |
90 | #ifdef CONFIG_IPV6_PIMSM_V2 |
91 | static struct inet6_protocol pim6_protocol; | |
92 | #endif | |
93 | ||
7bc570c8 YH |
94 | static struct timer_list ipmr_expire_timer; |
95 | ||
96 | ||
97 | #ifdef CONFIG_PROC_FS | |
98 | ||
99 | struct ipmr_mfc_iter { | |
100 | struct mfc6_cache **cache; | |
101 | int ct; | |
102 | }; | |
103 | ||
104 | ||
105 | static struct mfc6_cache *ipmr_mfc_seq_idx(struct ipmr_mfc_iter *it, loff_t pos) | |
106 | { | |
107 | struct mfc6_cache *mfc; | |
108 | ||
4a6258a0 | 109 | it->cache = init_net.ipv6.mfc6_cache_array; |
7bc570c8 | 110 | read_lock(&mrt_lock); |
4a6258a0 BT |
111 | for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) |
112 | for (mfc = init_net.ipv6.mfc6_cache_array[it->ct]; | |
113 | mfc; mfc = mfc->next) | |
7bc570c8 YH |
114 | if (pos-- == 0) |
115 | return mfc; | |
116 | read_unlock(&mrt_lock); | |
117 | ||
118 | it->cache = &mfc_unres_queue; | |
119 | spin_lock_bh(&mfc_unres_lock); | |
120 | for (mfc = mfc_unres_queue; mfc; mfc = mfc->next) | |
121 | if (pos-- == 0) | |
122 | return mfc; | |
123 | spin_unlock_bh(&mfc_unres_lock); | |
124 | ||
125 | it->cache = NULL; | |
126 | return NULL; | |
127 | } | |
128 | ||
129 | ||
130 | ||
131 | ||
132 | /* | |
133 | * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif | |
134 | */ | |
135 | ||
136 | struct ipmr_vif_iter { | |
137 | int ct; | |
138 | }; | |
139 | ||
140 | static struct mif_device *ip6mr_vif_seq_idx(struct ipmr_vif_iter *iter, | |
141 | loff_t pos) | |
142 | { | |
4e16880c BT |
143 | for (iter->ct = 0; iter->ct < init_net.ipv6.maxvif; ++iter->ct) { |
144 | if (!MIF_EXISTS(&init_net, iter->ct)) | |
7bc570c8 YH |
145 | continue; |
146 | if (pos-- == 0) | |
4e16880c | 147 | return &init_net.ipv6.vif6_table[iter->ct]; |
7bc570c8 YH |
148 | } |
149 | return NULL; | |
150 | } | |
151 | ||
152 | static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos) | |
153 | __acquires(mrt_lock) | |
154 | { | |
155 | read_lock(&mrt_lock); | |
156 | return (*pos ? ip6mr_vif_seq_idx(seq->private, *pos - 1) | |
157 | : SEQ_START_TOKEN); | |
158 | } | |
159 | ||
160 | static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
161 | { | |
162 | struct ipmr_vif_iter *iter = seq->private; | |
163 | ||
164 | ++*pos; | |
165 | if (v == SEQ_START_TOKEN) | |
166 | return ip6mr_vif_seq_idx(iter, 0); | |
167 | ||
4e16880c BT |
168 | while (++iter->ct < init_net.ipv6.maxvif) { |
169 | if (!MIF_EXISTS(&init_net, iter->ct)) | |
7bc570c8 | 170 | continue; |
4e16880c | 171 | return &init_net.ipv6.vif6_table[iter->ct]; |
7bc570c8 YH |
172 | } |
173 | return NULL; | |
174 | } | |
175 | ||
176 | static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v) | |
177 | __releases(mrt_lock) | |
178 | { | |
179 | read_unlock(&mrt_lock); | |
180 | } | |
181 | ||
182 | static int ip6mr_vif_seq_show(struct seq_file *seq, void *v) | |
183 | { | |
184 | if (v == SEQ_START_TOKEN) { | |
185 | seq_puts(seq, | |
186 | "Interface BytesIn PktsIn BytesOut PktsOut Flags\n"); | |
187 | } else { | |
188 | const struct mif_device *vif = v; | |
189 | const char *name = vif->dev ? vif->dev->name : "none"; | |
190 | ||
191 | seq_printf(seq, | |
d430a227 | 192 | "%2td %-10s %8ld %7ld %8ld %7ld %05X\n", |
4e16880c | 193 | vif - init_net.ipv6.vif6_table, |
7bc570c8 YH |
194 | name, vif->bytes_in, vif->pkt_in, |
195 | vif->bytes_out, vif->pkt_out, | |
196 | vif->flags); | |
197 | } | |
198 | return 0; | |
199 | } | |
200 | ||
201 | static struct seq_operations ip6mr_vif_seq_ops = { | |
202 | .start = ip6mr_vif_seq_start, | |
203 | .next = ip6mr_vif_seq_next, | |
204 | .stop = ip6mr_vif_seq_stop, | |
205 | .show = ip6mr_vif_seq_show, | |
206 | }; | |
207 | ||
208 | static int ip6mr_vif_open(struct inode *inode, struct file *file) | |
209 | { | |
210 | return seq_open_private(file, &ip6mr_vif_seq_ops, | |
211 | sizeof(struct ipmr_vif_iter)); | |
212 | } | |
213 | ||
214 | static struct file_operations ip6mr_vif_fops = { | |
215 | .owner = THIS_MODULE, | |
216 | .open = ip6mr_vif_open, | |
217 | .read = seq_read, | |
218 | .llseek = seq_lseek, | |
eedd726e | 219 | .release = seq_release_private, |
7bc570c8 YH |
220 | }; |
221 | ||
222 | static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) | |
223 | { | |
224 | return (*pos ? ipmr_mfc_seq_idx(seq->private, *pos - 1) | |
225 | : SEQ_START_TOKEN); | |
226 | } | |
227 | ||
228 | static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
229 | { | |
230 | struct mfc6_cache *mfc = v; | |
231 | struct ipmr_mfc_iter *it = seq->private; | |
232 | ||
233 | ++*pos; | |
234 | ||
235 | if (v == SEQ_START_TOKEN) | |
236 | return ipmr_mfc_seq_idx(seq->private, 0); | |
237 | ||
238 | if (mfc->next) | |
239 | return mfc->next; | |
240 | ||
241 | if (it->cache == &mfc_unres_queue) | |
242 | goto end_of_list; | |
243 | ||
4a6258a0 | 244 | BUG_ON(it->cache != init_net.ipv6.mfc6_cache_array); |
7bc570c8 | 245 | |
4a6258a0 BT |
246 | while (++it->ct < MFC6_LINES) { |
247 | mfc = init_net.ipv6.mfc6_cache_array[it->ct]; | |
7bc570c8 YH |
248 | if (mfc) |
249 | return mfc; | |
250 | } | |
251 | ||
252 | /* exhausted cache_array, show unresolved */ | |
253 | read_unlock(&mrt_lock); | |
254 | it->cache = &mfc_unres_queue; | |
255 | it->ct = 0; | |
256 | ||
257 | spin_lock_bh(&mfc_unres_lock); | |
258 | mfc = mfc_unres_queue; | |
259 | if (mfc) | |
260 | return mfc; | |
261 | ||
262 | end_of_list: | |
263 | spin_unlock_bh(&mfc_unres_lock); | |
264 | it->cache = NULL; | |
265 | ||
266 | return NULL; | |
267 | } | |
268 | ||
269 | static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) | |
270 | { | |
271 | struct ipmr_mfc_iter *it = seq->private; | |
272 | ||
273 | if (it->cache == &mfc_unres_queue) | |
274 | spin_unlock_bh(&mfc_unres_lock); | |
4a6258a0 | 275 | else if (it->cache == init_net.ipv6.mfc6_cache_array) |
7bc570c8 YH |
276 | read_unlock(&mrt_lock); |
277 | } | |
278 | ||
279 | static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) | |
280 | { | |
281 | int n; | |
282 | ||
283 | if (v == SEQ_START_TOKEN) { | |
284 | seq_puts(seq, | |
285 | "Group " | |
286 | "Origin " | |
287 | "Iif Pkts Bytes Wrong Oifs\n"); | |
288 | } else { | |
289 | const struct mfc6_cache *mfc = v; | |
290 | const struct ipmr_mfc_iter *it = seq->private; | |
291 | ||
999890b2 | 292 | seq_printf(seq, "%pI6 %pI6 %-3hd", |
0c6ce78a | 293 | &mfc->mf6c_mcastgrp, &mfc->mf6c_origin, |
1ea472e2 | 294 | mfc->mf6c_parent); |
7bc570c8 YH |
295 | |
296 | if (it->cache != &mfc_unres_queue) { | |
1ea472e2 BT |
297 | seq_printf(seq, " %8lu %8lu %8lu", |
298 | mfc->mfc_un.res.pkt, | |
299 | mfc->mfc_un.res.bytes, | |
300 | mfc->mfc_un.res.wrong_if); | |
7bc570c8 YH |
301 | for (n = mfc->mfc_un.res.minvif; |
302 | n < mfc->mfc_un.res.maxvif; n++) { | |
4e16880c | 303 | if (MIF_EXISTS(&init_net, n) && |
7bc570c8 YH |
304 | mfc->mfc_un.res.ttls[n] < 255) |
305 | seq_printf(seq, | |
306 | " %2d:%-3d", | |
307 | n, mfc->mfc_un.res.ttls[n]); | |
308 | } | |
1ea472e2 BT |
309 | } else { |
310 | /* unresolved mfc_caches don't contain | |
311 | * pkt, bytes and wrong_if values | |
312 | */ | |
313 | seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); | |
7bc570c8 YH |
314 | } |
315 | seq_putc(seq, '\n'); | |
316 | } | |
317 | return 0; | |
318 | } | |
319 | ||
320 | static struct seq_operations ipmr_mfc_seq_ops = { | |
321 | .start = ipmr_mfc_seq_start, | |
322 | .next = ipmr_mfc_seq_next, | |
323 | .stop = ipmr_mfc_seq_stop, | |
324 | .show = ipmr_mfc_seq_show, | |
325 | }; | |
326 | ||
327 | static int ipmr_mfc_open(struct inode *inode, struct file *file) | |
328 | { | |
329 | return seq_open_private(file, &ipmr_mfc_seq_ops, | |
330 | sizeof(struct ipmr_mfc_iter)); | |
331 | } | |
332 | ||
333 | static struct file_operations ip6mr_mfc_fops = { | |
334 | .owner = THIS_MODULE, | |
335 | .open = ipmr_mfc_open, | |
336 | .read = seq_read, | |
337 | .llseek = seq_lseek, | |
eedd726e | 338 | .release = seq_release_private, |
7bc570c8 YH |
339 | }; |
340 | #endif | |
341 | ||
14fb64e1 YH |
342 | #ifdef CONFIG_IPV6_PIMSM_V2 |
343 | static int reg_vif_num = -1; | |
344 | ||
345 | static int pim6_rcv(struct sk_buff *skb) | |
346 | { | |
347 | struct pimreghdr *pim; | |
348 | struct ipv6hdr *encap; | |
349 | struct net_device *reg_dev = NULL; | |
350 | ||
351 | if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap))) | |
352 | goto drop; | |
353 | ||
354 | pim = (struct pimreghdr *)skb_transport_header(skb); | |
355 | if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) || | |
356 | (pim->flags & PIM_NULL_REGISTER) || | |
357 | (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && | |
ec6b486f | 358 | csum_fold(skb_checksum(skb, 0, skb->len, 0)))) |
14fb64e1 YH |
359 | goto drop; |
360 | ||
361 | /* check if the inner packet is destined to mcast group */ | |
362 | encap = (struct ipv6hdr *)(skb_transport_header(skb) + | |
363 | sizeof(*pim)); | |
364 | ||
365 | if (!ipv6_addr_is_multicast(&encap->daddr) || | |
366 | encap->payload_len == 0 || | |
367 | ntohs(encap->payload_len) + sizeof(*pim) > skb->len) | |
368 | goto drop; | |
369 | ||
370 | read_lock(&mrt_lock); | |
371 | if (reg_vif_num >= 0) | |
4e16880c | 372 | reg_dev = init_net.ipv6.vif6_table[reg_vif_num].dev; |
14fb64e1 YH |
373 | if (reg_dev) |
374 | dev_hold(reg_dev); | |
375 | read_unlock(&mrt_lock); | |
376 | ||
377 | if (reg_dev == NULL) | |
378 | goto drop; | |
379 | ||
380 | skb->mac_header = skb->network_header; | |
381 | skb_pull(skb, (u8 *)encap - skb->data); | |
382 | skb_reset_network_header(skb); | |
383 | skb->dev = reg_dev; | |
384 | skb->protocol = htons(ETH_P_IP); | |
385 | skb->ip_summed = 0; | |
386 | skb->pkt_type = PACKET_HOST; | |
387 | dst_release(skb->dst); | |
dc58c78c PE |
388 | reg_dev->stats.rx_bytes += skb->len; |
389 | reg_dev->stats.rx_packets++; | |
14fb64e1 YH |
390 | skb->dst = NULL; |
391 | nf_reset(skb); | |
392 | netif_rx(skb); | |
393 | dev_put(reg_dev); | |
394 | return 0; | |
395 | drop: | |
396 | kfree_skb(skb); | |
397 | return 0; | |
398 | } | |
399 | ||
400 | static struct inet6_protocol pim6_protocol = { | |
401 | .handler = pim6_rcv, | |
402 | }; | |
403 | ||
404 | /* Service routines creating virtual interfaces: PIMREG */ | |
405 | ||
406 | static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) | |
407 | { | |
408 | read_lock(&mrt_lock); | |
dc58c78c PE |
409 | dev->stats.tx_bytes += skb->len; |
410 | dev->stats.tx_packets++; | |
14fb64e1 YH |
411 | ip6mr_cache_report(skb, reg_vif_num, MRT6MSG_WHOLEPKT); |
412 | read_unlock(&mrt_lock); | |
413 | kfree_skb(skb); | |
414 | return 0; | |
415 | } | |
416 | ||
007c3838 SH |
417 | static const struct net_device_ops reg_vif_netdev_ops = { |
418 | .ndo_start_xmit = reg_vif_xmit, | |
419 | }; | |
420 | ||
14fb64e1 YH |
421 | static void reg_vif_setup(struct net_device *dev) |
422 | { | |
423 | dev->type = ARPHRD_PIMREG; | |
424 | dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8; | |
425 | dev->flags = IFF_NOARP; | |
007c3838 | 426 | dev->netdev_ops = ®_vif_netdev_ops; |
14fb64e1 YH |
427 | dev->destructor = free_netdev; |
428 | } | |
429 | ||
430 | static struct net_device *ip6mr_reg_vif(void) | |
431 | { | |
432 | struct net_device *dev; | |
14fb64e1 | 433 | |
dc58c78c | 434 | dev = alloc_netdev(0, "pim6reg", reg_vif_setup); |
14fb64e1 YH |
435 | if (dev == NULL) |
436 | return NULL; | |
437 | ||
438 | if (register_netdevice(dev)) { | |
439 | free_netdev(dev); | |
440 | return NULL; | |
441 | } | |
442 | dev->iflink = 0; | |
443 | ||
14fb64e1 YH |
444 | if (dev_open(dev)) |
445 | goto failure; | |
446 | ||
7af3db78 | 447 | dev_hold(dev); |
14fb64e1 YH |
448 | return dev; |
449 | ||
450 | failure: | |
451 | /* allow the register to be completed before unregistering. */ | |
452 | rtnl_unlock(); | |
453 | rtnl_lock(); | |
454 | ||
455 | unregister_netdevice(dev); | |
456 | return NULL; | |
457 | } | |
458 | #endif | |
459 | ||
7bc570c8 YH |
460 | /* |
461 | * Delete a VIF entry | |
462 | */ | |
463 | ||
464 | static int mif6_delete(int vifi) | |
465 | { | |
466 | struct mif_device *v; | |
467 | struct net_device *dev; | |
4e16880c | 468 | if (vifi < 0 || vifi >= init_net.ipv6.maxvif) |
7bc570c8 YH |
469 | return -EADDRNOTAVAIL; |
470 | ||
4e16880c | 471 | v = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 YH |
472 | |
473 | write_lock_bh(&mrt_lock); | |
474 | dev = v->dev; | |
475 | v->dev = NULL; | |
476 | ||
477 | if (!dev) { | |
478 | write_unlock_bh(&mrt_lock); | |
479 | return -EADDRNOTAVAIL; | |
480 | } | |
481 | ||
14fb64e1 YH |
482 | #ifdef CONFIG_IPV6_PIMSM_V2 |
483 | if (vifi == reg_vif_num) | |
484 | reg_vif_num = -1; | |
485 | #endif | |
486 | ||
4e16880c | 487 | if (vifi + 1 == init_net.ipv6.maxvif) { |
7bc570c8 YH |
488 | int tmp; |
489 | for (tmp = vifi - 1; tmp >= 0; tmp--) { | |
4e16880c | 490 | if (MIF_EXISTS(&init_net, tmp)) |
7bc570c8 YH |
491 | break; |
492 | } | |
4e16880c | 493 | init_net.ipv6.maxvif = tmp + 1; |
7bc570c8 YH |
494 | } |
495 | ||
496 | write_unlock_bh(&mrt_lock); | |
497 | ||
498 | dev_set_allmulti(dev, -1); | |
499 | ||
500 | if (v->flags & MIFF_REGISTER) | |
501 | unregister_netdevice(dev); | |
502 | ||
503 | dev_put(dev); | |
504 | return 0; | |
505 | } | |
506 | ||
58701ad4 BT |
507 | static inline void ip6mr_cache_free(struct mfc6_cache *c) |
508 | { | |
509 | release_net(mfc6_net(c)); | |
510 | kmem_cache_free(mrt_cachep, c); | |
511 | } | |
512 | ||
7bc570c8 YH |
513 | /* Destroy an unresolved cache entry, killing queued skbs |
514 | and reporting error to netlink readers. | |
515 | */ | |
516 | ||
517 | static void ip6mr_destroy_unres(struct mfc6_cache *c) | |
518 | { | |
519 | struct sk_buff *skb; | |
520 | ||
4045e57c | 521 | atomic_dec(&init_net.ipv6.cache_resolve_queue_len); |
7bc570c8 YH |
522 | |
523 | while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) { | |
524 | if (ipv6_hdr(skb)->version == 0) { | |
525 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); | |
526 | nlh->nlmsg_type = NLMSG_ERROR; | |
527 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
528 | skb_trim(skb, nlh->nlmsg_len); | |
529 | ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT; | |
530 | rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); | |
531 | } else | |
532 | kfree_skb(skb); | |
533 | } | |
534 | ||
58701ad4 | 535 | ip6mr_cache_free(c); |
7bc570c8 YH |
536 | } |
537 | ||
538 | ||
539 | /* Single timer process for all the unresolved queue. */ | |
540 | ||
541 | static void ipmr_do_expire_process(unsigned long dummy) | |
542 | { | |
543 | unsigned long now = jiffies; | |
544 | unsigned long expires = 10 * HZ; | |
545 | struct mfc6_cache *c, **cp; | |
546 | ||
547 | cp = &mfc_unres_queue; | |
548 | ||
549 | while ((c = *cp) != NULL) { | |
550 | if (time_after(c->mfc_un.unres.expires, now)) { | |
551 | /* not yet... */ | |
552 | unsigned long interval = c->mfc_un.unres.expires - now; | |
553 | if (interval < expires) | |
554 | expires = interval; | |
555 | cp = &c->next; | |
556 | continue; | |
557 | } | |
558 | ||
559 | *cp = c->next; | |
560 | ip6mr_destroy_unres(c); | |
561 | } | |
562 | ||
4045e57c | 563 | if (mfc_unres_queue != NULL) |
7bc570c8 YH |
564 | mod_timer(&ipmr_expire_timer, jiffies + expires); |
565 | } | |
566 | ||
567 | static void ipmr_expire_process(unsigned long dummy) | |
568 | { | |
569 | if (!spin_trylock(&mfc_unres_lock)) { | |
570 | mod_timer(&ipmr_expire_timer, jiffies + 1); | |
571 | return; | |
572 | } | |
573 | ||
4045e57c | 574 | if (mfc_unres_queue != NULL) |
7bc570c8 YH |
575 | ipmr_do_expire_process(dummy); |
576 | ||
577 | spin_unlock(&mfc_unres_lock); | |
578 | } | |
579 | ||
580 | /* Fill oifs list. It is called under write locked mrt_lock. */ | |
581 | ||
582 | static void ip6mr_update_thresholds(struct mfc6_cache *cache, unsigned char *ttls) | |
583 | { | |
584 | int vifi; | |
585 | ||
6ac7eb08 | 586 | cache->mfc_un.res.minvif = MAXMIFS; |
7bc570c8 | 587 | cache->mfc_un.res.maxvif = 0; |
6ac7eb08 | 588 | memset(cache->mfc_un.res.ttls, 255, MAXMIFS); |
7bc570c8 | 589 | |
4e16880c BT |
590 | for (vifi = 0; vifi < init_net.ipv6.maxvif; vifi++) { |
591 | if (MIF_EXISTS(&init_net, vifi) && | |
592 | ttls[vifi] && ttls[vifi] < 255) { | |
7bc570c8 YH |
593 | cache->mfc_un.res.ttls[vifi] = ttls[vifi]; |
594 | if (cache->mfc_un.res.minvif > vifi) | |
595 | cache->mfc_un.res.minvif = vifi; | |
596 | if (cache->mfc_un.res.maxvif <= vifi) | |
597 | cache->mfc_un.res.maxvif = vifi + 1; | |
598 | } | |
599 | } | |
600 | } | |
601 | ||
602 | static int mif6_add(struct mif6ctl *vifc, int mrtsock) | |
603 | { | |
604 | int vifi = vifc->mif6c_mifi; | |
4e16880c | 605 | struct mif_device *v = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 | 606 | struct net_device *dev; |
5ae7b444 | 607 | int err; |
7bc570c8 YH |
608 | |
609 | /* Is vif busy ? */ | |
4e16880c | 610 | if (MIF_EXISTS(&init_net, vifi)) |
7bc570c8 YH |
611 | return -EADDRINUSE; |
612 | ||
613 | switch (vifc->mif6c_flags) { | |
14fb64e1 YH |
614 | #ifdef CONFIG_IPV6_PIMSM_V2 |
615 | case MIFF_REGISTER: | |
616 | /* | |
617 | * Special Purpose VIF in PIM | |
618 | * All the packets will be sent to the daemon | |
619 | */ | |
620 | if (reg_vif_num >= 0) | |
621 | return -EADDRINUSE; | |
622 | dev = ip6mr_reg_vif(); | |
623 | if (!dev) | |
624 | return -ENOBUFS; | |
5ae7b444 WC |
625 | err = dev_set_allmulti(dev, 1); |
626 | if (err) { | |
627 | unregister_netdevice(dev); | |
7af3db78 | 628 | dev_put(dev); |
5ae7b444 WC |
629 | return err; |
630 | } | |
14fb64e1 YH |
631 | break; |
632 | #endif | |
7bc570c8 YH |
633 | case 0: |
634 | dev = dev_get_by_index(&init_net, vifc->mif6c_pifi); | |
635 | if (!dev) | |
636 | return -EADDRNOTAVAIL; | |
5ae7b444 | 637 | err = dev_set_allmulti(dev, 1); |
7af3db78 WC |
638 | if (err) { |
639 | dev_put(dev); | |
5ae7b444 | 640 | return err; |
7af3db78 | 641 | } |
7bc570c8 YH |
642 | break; |
643 | default: | |
644 | return -EINVAL; | |
645 | } | |
646 | ||
7bc570c8 YH |
647 | /* |
648 | * Fill in the VIF structures | |
649 | */ | |
650 | v->rate_limit = vifc->vifc_rate_limit; | |
651 | v->flags = vifc->mif6c_flags; | |
652 | if (!mrtsock) | |
653 | v->flags |= VIFF_STATIC; | |
654 | v->threshold = vifc->vifc_threshold; | |
655 | v->bytes_in = 0; | |
656 | v->bytes_out = 0; | |
657 | v->pkt_in = 0; | |
658 | v->pkt_out = 0; | |
659 | v->link = dev->ifindex; | |
660 | if (v->flags & MIFF_REGISTER) | |
661 | v->link = dev->iflink; | |
662 | ||
663 | /* And finish update writing critical data */ | |
664 | write_lock_bh(&mrt_lock); | |
7bc570c8 | 665 | v->dev = dev; |
14fb64e1 YH |
666 | #ifdef CONFIG_IPV6_PIMSM_V2 |
667 | if (v->flags & MIFF_REGISTER) | |
668 | reg_vif_num = vifi; | |
669 | #endif | |
4e16880c BT |
670 | if (vifi + 1 > init_net.ipv6.maxvif) |
671 | init_net.ipv6.maxvif = vifi + 1; | |
7bc570c8 YH |
672 | write_unlock_bh(&mrt_lock); |
673 | return 0; | |
674 | } | |
675 | ||
676 | static struct mfc6_cache *ip6mr_cache_find(struct in6_addr *origin, struct in6_addr *mcastgrp) | |
677 | { | |
678 | int line = MFC6_HASH(mcastgrp, origin); | |
679 | struct mfc6_cache *c; | |
680 | ||
4a6258a0 | 681 | for (c = init_net.ipv6.mfc6_cache_array[line]; c; c = c->next) { |
7bc570c8 YH |
682 | if (ipv6_addr_equal(&c->mf6c_origin, origin) && |
683 | ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) | |
684 | break; | |
685 | } | |
686 | return c; | |
687 | } | |
688 | ||
689 | /* | |
690 | * Allocate a multicast cache entry | |
691 | */ | |
58701ad4 | 692 | static struct mfc6_cache *ip6mr_cache_alloc(struct net *net) |
7bc570c8 | 693 | { |
36cbac59 | 694 | struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); |
7bc570c8 YH |
695 | if (c == NULL) |
696 | return NULL; | |
6ac7eb08 | 697 | c->mfc_un.res.minvif = MAXMIFS; |
58701ad4 | 698 | mfc6_net_set(c, net); |
7bc570c8 YH |
699 | return c; |
700 | } | |
701 | ||
58701ad4 | 702 | static struct mfc6_cache *ip6mr_cache_alloc_unres(struct net *net) |
7bc570c8 | 703 | { |
36cbac59 | 704 | struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); |
7bc570c8 YH |
705 | if (c == NULL) |
706 | return NULL; | |
7bc570c8 YH |
707 | skb_queue_head_init(&c->mfc_un.unres.unresolved); |
708 | c->mfc_un.unres.expires = jiffies + 10 * HZ; | |
58701ad4 | 709 | mfc6_net_set(c, net); |
7bc570c8 YH |
710 | return c; |
711 | } | |
712 | ||
713 | /* | |
714 | * A cache entry has gone into a resolved state from queued | |
715 | */ | |
716 | ||
717 | static void ip6mr_cache_resolve(struct mfc6_cache *uc, struct mfc6_cache *c) | |
718 | { | |
719 | struct sk_buff *skb; | |
720 | ||
721 | /* | |
722 | * Play the pending entries through our router | |
723 | */ | |
724 | ||
725 | while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { | |
726 | if (ipv6_hdr(skb)->version == 0) { | |
727 | int err; | |
728 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); | |
729 | ||
730 | if (ip6mr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) { | |
549e028d | 731 | nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; |
7bc570c8 YH |
732 | } else { |
733 | nlh->nlmsg_type = NLMSG_ERROR; | |
734 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
735 | skb_trim(skb, nlh->nlmsg_len); | |
736 | ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE; | |
737 | } | |
738 | err = rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); | |
739 | } else | |
740 | ip6_mr_forward(skb, c); | |
741 | } | |
742 | } | |
743 | ||
744 | /* | |
745 | * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd | |
746 | * expects the following bizarre scheme. | |
747 | * | |
748 | * Called under mrt_lock. | |
749 | */ | |
750 | ||
6ac7eb08 | 751 | static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert) |
7bc570c8 YH |
752 | { |
753 | struct sk_buff *skb; | |
754 | struct mrt6msg *msg; | |
755 | int ret; | |
756 | ||
14fb64e1 YH |
757 | #ifdef CONFIG_IPV6_PIMSM_V2 |
758 | if (assert == MRT6MSG_WHOLEPKT) | |
759 | skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) | |
760 | +sizeof(*msg)); | |
761 | else | |
762 | #endif | |
763 | skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); | |
7bc570c8 YH |
764 | |
765 | if (!skb) | |
766 | return -ENOBUFS; | |
767 | ||
768 | /* I suppose that internal messages | |
769 | * do not require checksums */ | |
770 | ||
771 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
772 | ||
14fb64e1 YH |
773 | #ifdef CONFIG_IPV6_PIMSM_V2 |
774 | if (assert == MRT6MSG_WHOLEPKT) { | |
775 | /* Ugly, but we have no choice with this interface. | |
776 | Duplicate old header, fix length etc. | |
777 | And all this only to mangle msg->im6_msgtype and | |
778 | to set msg->im6_mbz to "mbz" :-) | |
779 | */ | |
780 | skb_push(skb, -skb_network_offset(pkt)); | |
781 | ||
782 | skb_push(skb, sizeof(*msg)); | |
783 | skb_reset_transport_header(skb); | |
784 | msg = (struct mrt6msg *)skb_transport_header(skb); | |
785 | msg->im6_mbz = 0; | |
786 | msg->im6_msgtype = MRT6MSG_WHOLEPKT; | |
787 | msg->im6_mif = reg_vif_num; | |
788 | msg->im6_pad = 0; | |
789 | ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr); | |
790 | ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr); | |
791 | ||
792 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
793 | } else | |
794 | #endif | |
795 | { | |
7bc570c8 YH |
796 | /* |
797 | * Copy the IP header | |
798 | */ | |
799 | ||
800 | skb_put(skb, sizeof(struct ipv6hdr)); | |
801 | skb_reset_network_header(skb); | |
802 | skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); | |
803 | ||
804 | /* | |
805 | * Add our header | |
806 | */ | |
807 | skb_put(skb, sizeof(*msg)); | |
808 | skb_reset_transport_header(skb); | |
809 | msg = (struct mrt6msg *)skb_transport_header(skb); | |
810 | ||
811 | msg->im6_mbz = 0; | |
812 | msg->im6_msgtype = assert; | |
6ac7eb08 | 813 | msg->im6_mif = mifi; |
7bc570c8 YH |
814 | msg->im6_pad = 0; |
815 | ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr); | |
816 | ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr); | |
817 | ||
818 | skb->dst = dst_clone(pkt->dst); | |
819 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
820 | ||
821 | skb_pull(skb, sizeof(struct ipv6hdr)); | |
14fb64e1 | 822 | } |
7bc570c8 | 823 | |
bd91b8bf | 824 | if (init_net.ipv6.mroute6_sk == NULL) { |
7bc570c8 YH |
825 | kfree_skb(skb); |
826 | return -EINVAL; | |
827 | } | |
828 | ||
829 | /* | |
830 | * Deliver to user space multicast routing algorithms | |
831 | */ | |
bd91b8bf BT |
832 | ret = sock_queue_rcv_skb(init_net.ipv6.mroute6_sk, skb); |
833 | if (ret < 0) { | |
7bc570c8 YH |
834 | if (net_ratelimit()) |
835 | printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n"); | |
836 | kfree_skb(skb); | |
837 | } | |
838 | ||
839 | return ret; | |
840 | } | |
841 | ||
842 | /* | |
843 | * Queue a packet for resolution. It gets locked cache entry! | |
844 | */ | |
845 | ||
846 | static int | |
6ac7eb08 | 847 | ip6mr_cache_unresolved(mifi_t mifi, struct sk_buff *skb) |
7bc570c8 YH |
848 | { |
849 | int err; | |
850 | struct mfc6_cache *c; | |
851 | ||
852 | spin_lock_bh(&mfc_unres_lock); | |
853 | for (c = mfc_unres_queue; c; c = c->next) { | |
4045e57c BT |
854 | if (net_eq(mfc6_net(c), &init_net) && |
855 | ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && | |
7bc570c8 YH |
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 | ||
4045e57c | 865 | if (atomic_read(&init_net.ipv6.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 | ||
4045e57c | 894 | atomic_inc(&init_net.ipv6.cache_resolve_queue_len); |
7bc570c8 YH |
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) { | |
4045e57c BT |
1122 | if (net_eq(mfc6_net(uc), &init_net) && |
1123 | ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && | |
7bc570c8 YH |
1124 | ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { |
1125 | *cp = uc->next; | |
4045e57c | 1126 | atomic_dec(&init_net.ipv6.cache_resolve_queue_len); |
7bc570c8 YH |
1127 | break; |
1128 | } | |
1129 | } | |
4045e57c BT |
1130 | if (mfc_unres_queue == NULL) |
1131 | del_timer(&ipmr_expire_timer); | |
7bc570c8 YH |
1132 | spin_unlock_bh(&mfc_unres_lock); |
1133 | ||
1134 | if (uc) { | |
1135 | ip6mr_cache_resolve(uc, c); | |
58701ad4 | 1136 | ip6mr_cache_free(uc); |
7bc570c8 YH |
1137 | } |
1138 | return 0; | |
1139 | } | |
1140 | ||
1141 | /* | |
1142 | * Close the multicast socket, and clear the vif tables etc | |
1143 | */ | |
1144 | ||
1145 | static void mroute_clean_tables(struct sock *sk) | |
1146 | { | |
1147 | int i; | |
1148 | ||
1149 | /* | |
1150 | * Shut down all active vif entries | |
1151 | */ | |
4e16880c BT |
1152 | for (i = 0; i < init_net.ipv6.maxvif; i++) { |
1153 | if (!(init_net.ipv6.vif6_table[i].flags & VIFF_STATIC)) | |
7bc570c8 YH |
1154 | mif6_delete(i); |
1155 | } | |
1156 | ||
1157 | /* | |
1158 | * Wipe the cache | |
1159 | */ | |
4a6258a0 | 1160 | for (i = 0; i < MFC6_LINES; i++) { |
7bc570c8 YH |
1161 | struct mfc6_cache *c, **cp; |
1162 | ||
4a6258a0 | 1163 | cp = &init_net.ipv6.mfc6_cache_array[i]; |
7bc570c8 YH |
1164 | while ((c = *cp) != NULL) { |
1165 | if (c->mfc_flags & MFC_STATIC) { | |
1166 | cp = &c->next; | |
1167 | continue; | |
1168 | } | |
1169 | write_lock_bh(&mrt_lock); | |
1170 | *cp = c->next; | |
1171 | write_unlock_bh(&mrt_lock); | |
1172 | ||
58701ad4 | 1173 | ip6mr_cache_free(c); |
7bc570c8 YH |
1174 | } |
1175 | } | |
1176 | ||
4045e57c BT |
1177 | if (atomic_read(&init_net.ipv6.cache_resolve_queue_len) != 0) { |
1178 | struct mfc6_cache *c, **cp; | |
7bc570c8 YH |
1179 | |
1180 | spin_lock_bh(&mfc_unres_lock); | |
4045e57c BT |
1181 | cp = &mfc_unres_queue; |
1182 | while ((c = *cp) != NULL) { | |
1183 | if (!net_eq(mfc6_net(c), &init_net)) { | |
1184 | cp = &c->next; | |
1185 | continue; | |
1186 | } | |
1187 | *cp = c->next; | |
7bc570c8 | 1188 | ip6mr_destroy_unres(c); |
7bc570c8 YH |
1189 | } |
1190 | spin_unlock_bh(&mfc_unres_lock); | |
1191 | } | |
1192 | } | |
1193 | ||
1194 | static int ip6mr_sk_init(struct sock *sk) | |
1195 | { | |
1196 | int err = 0; | |
1197 | ||
1198 | rtnl_lock(); | |
1199 | write_lock_bh(&mrt_lock); | |
bd91b8bf BT |
1200 | if (likely(init_net.ipv6.mroute6_sk == NULL)) |
1201 | init_net.ipv6.mroute6_sk = sk; | |
7bc570c8 YH |
1202 | else |
1203 | err = -EADDRINUSE; | |
1204 | write_unlock_bh(&mrt_lock); | |
1205 | ||
1206 | rtnl_unlock(); | |
1207 | ||
1208 | return err; | |
1209 | } | |
1210 | ||
1211 | int ip6mr_sk_done(struct sock *sk) | |
1212 | { | |
1213 | int err = 0; | |
1214 | ||
1215 | rtnl_lock(); | |
bd91b8bf | 1216 | if (sk == init_net.ipv6.mroute6_sk) { |
7bc570c8 | 1217 | write_lock_bh(&mrt_lock); |
bd91b8bf | 1218 | init_net.ipv6.mroute6_sk = NULL; |
7bc570c8 YH |
1219 | write_unlock_bh(&mrt_lock); |
1220 | ||
1221 | mroute_clean_tables(sk); | |
1222 | } else | |
1223 | err = -EACCES; | |
1224 | rtnl_unlock(); | |
1225 | ||
1226 | return err; | |
1227 | } | |
1228 | ||
1229 | /* | |
1230 | * Socket options and virtual interface manipulation. The whole | |
1231 | * virtual interface system is a complete heap, but unfortunately | |
1232 | * that's how BSD mrouted happens to think. Maybe one day with a proper | |
1233 | * MOSPF/PIM router set up we can clean this up. | |
1234 | */ | |
1235 | ||
1236 | int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, int optlen) | |
1237 | { | |
1238 | int ret; | |
1239 | struct mif6ctl vif; | |
1240 | struct mf6cctl mfc; | |
1241 | mifi_t mifi; | |
1242 | ||
1243 | if (optname != MRT6_INIT) { | |
bd91b8bf | 1244 | if (sk != init_net.ipv6.mroute6_sk && !capable(CAP_NET_ADMIN)) |
7bc570c8 YH |
1245 | return -EACCES; |
1246 | } | |
1247 | ||
1248 | switch (optname) { | |
1249 | case MRT6_INIT: | |
1250 | if (sk->sk_type != SOCK_RAW || | |
1251 | inet_sk(sk)->num != IPPROTO_ICMPV6) | |
1252 | return -EOPNOTSUPP; | |
1253 | if (optlen < sizeof(int)) | |
1254 | return -EINVAL; | |
1255 | ||
1256 | return ip6mr_sk_init(sk); | |
1257 | ||
1258 | case MRT6_DONE: | |
1259 | return ip6mr_sk_done(sk); | |
1260 | ||
1261 | case MRT6_ADD_MIF: | |
1262 | if (optlen < sizeof(vif)) | |
1263 | return -EINVAL; | |
1264 | if (copy_from_user(&vif, optval, sizeof(vif))) | |
1265 | return -EFAULT; | |
6ac7eb08 | 1266 | if (vif.mif6c_mifi >= MAXMIFS) |
7bc570c8 YH |
1267 | return -ENFILE; |
1268 | rtnl_lock(); | |
bd91b8bf | 1269 | ret = mif6_add(&vif, sk == init_net.ipv6.mroute6_sk); |
7bc570c8 YH |
1270 | rtnl_unlock(); |
1271 | return ret; | |
1272 | ||
1273 | case MRT6_DEL_MIF: | |
1274 | if (optlen < sizeof(mifi_t)) | |
1275 | return -EINVAL; | |
1276 | if (copy_from_user(&mifi, optval, sizeof(mifi_t))) | |
1277 | return -EFAULT; | |
1278 | rtnl_lock(); | |
1279 | ret = mif6_delete(mifi); | |
1280 | rtnl_unlock(); | |
1281 | return ret; | |
1282 | ||
1283 | /* | |
1284 | * Manipulate the forwarding caches. These live | |
1285 | * in a sort of kernel/user symbiosis. | |
1286 | */ | |
1287 | case MRT6_ADD_MFC: | |
1288 | case MRT6_DEL_MFC: | |
1289 | if (optlen < sizeof(mfc)) | |
1290 | return -EINVAL; | |
1291 | if (copy_from_user(&mfc, optval, sizeof(mfc))) | |
1292 | return -EFAULT; | |
1293 | rtnl_lock(); | |
1294 | if (optname == MRT6_DEL_MFC) | |
1295 | ret = ip6mr_mfc_delete(&mfc); | |
1296 | else | |
bd91b8bf | 1297 | ret = ip6mr_mfc_add(&mfc, sk == init_net.ipv6.mroute6_sk); |
7bc570c8 YH |
1298 | rtnl_unlock(); |
1299 | return ret; | |
1300 | ||
14fb64e1 YH |
1301 | /* |
1302 | * Control PIM assert (to activate pim will activate assert) | |
1303 | */ | |
1304 | case MRT6_ASSERT: | |
1305 | { | |
1306 | int v; | |
1307 | if (get_user(v, (int __user *)optval)) | |
1308 | return -EFAULT; | |
1309 | mroute_do_assert = !!v; | |
1310 | return 0; | |
1311 | } | |
1312 | ||
1313 | #ifdef CONFIG_IPV6_PIMSM_V2 | |
1314 | case MRT6_PIM: | |
1315 | { | |
a9f83bf3 | 1316 | int v; |
14fb64e1 YH |
1317 | if (get_user(v, (int __user *)optval)) |
1318 | return -EFAULT; | |
1319 | v = !!v; | |
1320 | rtnl_lock(); | |
1321 | ret = 0; | |
1322 | if (v != mroute_do_pim) { | |
1323 | mroute_do_pim = v; | |
1324 | mroute_do_assert = v; | |
1325 | if (mroute_do_pim) | |
1326 | ret = inet6_add_protocol(&pim6_protocol, | |
1327 | IPPROTO_PIM); | |
1328 | else | |
1329 | ret = inet6_del_protocol(&pim6_protocol, | |
1330 | IPPROTO_PIM); | |
1331 | if (ret < 0) | |
1332 | ret = -EAGAIN; | |
1333 | } | |
1334 | rtnl_unlock(); | |
1335 | return ret; | |
1336 | } | |
1337 | ||
1338 | #endif | |
7bc570c8 | 1339 | /* |
7d120c55 | 1340 | * Spurious command, or MRT6_VERSION which you cannot |
7bc570c8 YH |
1341 | * set. |
1342 | */ | |
1343 | default: | |
1344 | return -ENOPROTOOPT; | |
1345 | } | |
1346 | } | |
1347 | ||
1348 | /* | |
1349 | * Getsock opt support for the multicast routing system. | |
1350 | */ | |
1351 | ||
1352 | int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, | |
1353 | int __user *optlen) | |
1354 | { | |
1355 | int olr; | |
1356 | int val; | |
1357 | ||
1358 | switch (optname) { | |
1359 | case MRT6_VERSION: | |
1360 | val = 0x0305; | |
1361 | break; | |
14fb64e1 YH |
1362 | #ifdef CONFIG_IPV6_PIMSM_V2 |
1363 | case MRT6_PIM: | |
1364 | val = mroute_do_pim; | |
1365 | break; | |
1366 | #endif | |
1367 | case MRT6_ASSERT: | |
1368 | val = mroute_do_assert; | |
1369 | break; | |
7bc570c8 YH |
1370 | default: |
1371 | return -ENOPROTOOPT; | |
1372 | } | |
1373 | ||
1374 | if (get_user(olr, optlen)) | |
1375 | return -EFAULT; | |
1376 | ||
1377 | olr = min_t(int, olr, sizeof(int)); | |
1378 | if (olr < 0) | |
1379 | return -EINVAL; | |
1380 | ||
1381 | if (put_user(olr, optlen)) | |
1382 | return -EFAULT; | |
1383 | if (copy_to_user(optval, &val, olr)) | |
1384 | return -EFAULT; | |
1385 | return 0; | |
1386 | } | |
1387 | ||
1388 | /* | |
1389 | * The IP multicast ioctl support routines. | |
1390 | */ | |
1391 | ||
1392 | int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) | |
1393 | { | |
1394 | struct sioc_sg_req6 sr; | |
1395 | struct sioc_mif_req6 vr; | |
1396 | struct mif_device *vif; | |
1397 | struct mfc6_cache *c; | |
1398 | ||
1399 | switch (cmd) { | |
1400 | case SIOCGETMIFCNT_IN6: | |
1401 | if (copy_from_user(&vr, arg, sizeof(vr))) | |
1402 | return -EFAULT; | |
4e16880c | 1403 | if (vr.mifi >= init_net.ipv6.maxvif) |
7bc570c8 YH |
1404 | return -EINVAL; |
1405 | read_lock(&mrt_lock); | |
4e16880c BT |
1406 | vif = &init_net.ipv6.vif6_table[vr.mifi]; |
1407 | if (MIF_EXISTS(&init_net, vr.mifi)) { | |
7bc570c8 YH |
1408 | vr.icount = vif->pkt_in; |
1409 | vr.ocount = vif->pkt_out; | |
1410 | vr.ibytes = vif->bytes_in; | |
1411 | vr.obytes = vif->bytes_out; | |
1412 | read_unlock(&mrt_lock); | |
1413 | ||
1414 | if (copy_to_user(arg, &vr, sizeof(vr))) | |
1415 | return -EFAULT; | |
1416 | return 0; | |
1417 | } | |
1418 | read_unlock(&mrt_lock); | |
1419 | return -EADDRNOTAVAIL; | |
1420 | case SIOCGETSGCNT_IN6: | |
1421 | if (copy_from_user(&sr, arg, sizeof(sr))) | |
1422 | return -EFAULT; | |
1423 | ||
1424 | read_lock(&mrt_lock); | |
1425 | c = ip6mr_cache_find(&sr.src.sin6_addr, &sr.grp.sin6_addr); | |
1426 | if (c) { | |
1427 | sr.pktcnt = c->mfc_un.res.pkt; | |
1428 | sr.bytecnt = c->mfc_un.res.bytes; | |
1429 | sr.wrong_if = c->mfc_un.res.wrong_if; | |
1430 | read_unlock(&mrt_lock); | |
1431 | ||
1432 | if (copy_to_user(arg, &sr, sizeof(sr))) | |
1433 | return -EFAULT; | |
1434 | return 0; | |
1435 | } | |
1436 | read_unlock(&mrt_lock); | |
1437 | return -EADDRNOTAVAIL; | |
1438 | default: | |
1439 | return -ENOIOCTLCMD; | |
1440 | } | |
1441 | } | |
1442 | ||
1443 | ||
1444 | static inline int ip6mr_forward2_finish(struct sk_buff *skb) | |
1445 | { | |
483a47d2 DL |
1446 | IP6_INC_STATS_BH(dev_net(skb->dst->dev), ip6_dst_idev(skb->dst), |
1447 | IPSTATS_MIB_OUTFORWDATAGRAMS); | |
7bc570c8 YH |
1448 | return dst_output(skb); |
1449 | } | |
1450 | ||
1451 | /* | |
1452 | * Processing handlers for ip6mr_forward | |
1453 | */ | |
1454 | ||
1455 | static int ip6mr_forward2(struct sk_buff *skb, struct mfc6_cache *c, int vifi) | |
1456 | { | |
1457 | struct ipv6hdr *ipv6h; | |
4e16880c | 1458 | struct mif_device *vif = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 YH |
1459 | struct net_device *dev; |
1460 | struct dst_entry *dst; | |
1461 | struct flowi fl; | |
1462 | ||
1463 | if (vif->dev == NULL) | |
1464 | goto out_free; | |
1465 | ||
14fb64e1 YH |
1466 | #ifdef CONFIG_IPV6_PIMSM_V2 |
1467 | if (vif->flags & MIFF_REGISTER) { | |
1468 | vif->pkt_out++; | |
1469 | vif->bytes_out += skb->len; | |
dc58c78c PE |
1470 | vif->dev->stats.tx_bytes += skb->len; |
1471 | vif->dev->stats.tx_packets++; | |
14fb64e1 YH |
1472 | ip6mr_cache_report(skb, vifi, MRT6MSG_WHOLEPKT); |
1473 | kfree_skb(skb); | |
1474 | return 0; | |
1475 | } | |
1476 | #endif | |
1477 | ||
7bc570c8 YH |
1478 | ipv6h = ipv6_hdr(skb); |
1479 | ||
1480 | fl = (struct flowi) { | |
1481 | .oif = vif->link, | |
1482 | .nl_u = { .ip6_u = | |
1483 | { .daddr = ipv6h->daddr, } | |
1484 | } | |
1485 | }; | |
1486 | ||
1487 | dst = ip6_route_output(&init_net, NULL, &fl); | |
1488 | if (!dst) | |
1489 | goto out_free; | |
1490 | ||
1491 | dst_release(skb->dst); | |
1492 | skb->dst = dst; | |
1493 | ||
1494 | /* | |
1495 | * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally | |
1496 | * not only before forwarding, but after forwarding on all output | |
1497 | * interfaces. It is clear, if mrouter runs a multicasting | |
1498 | * program, it should receive packets not depending to what interface | |
1499 | * program is joined. | |
1500 | * If we will not make it, the program will have to join on all | |
1501 | * interfaces. On the other hand, multihoming host (or router, but | |
1502 | * not mrouter) cannot join to more than one interface - it will | |
1503 | * result in receiving multiple packets. | |
1504 | */ | |
1505 | dev = vif->dev; | |
1506 | skb->dev = dev; | |
1507 | vif->pkt_out++; | |
1508 | vif->bytes_out += skb->len; | |
1509 | ||
1510 | /* We are about to write */ | |
1511 | /* XXX: extension headers? */ | |
1512 | if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) | |
1513 | goto out_free; | |
1514 | ||
1515 | ipv6h = ipv6_hdr(skb); | |
1516 | ipv6h->hop_limit--; | |
1517 | ||
1518 | IP6CB(skb)->flags |= IP6SKB_FORWARDED; | |
1519 | ||
1520 | return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dev, | |
1521 | ip6mr_forward2_finish); | |
1522 | ||
1523 | out_free: | |
1524 | kfree_skb(skb); | |
1525 | return 0; | |
1526 | } | |
1527 | ||
1528 | static int ip6mr_find_vif(struct net_device *dev) | |
1529 | { | |
1530 | int ct; | |
4e16880c BT |
1531 | for (ct = init_net.ipv6.maxvif - 1; ct >= 0; ct--) { |
1532 | if (init_net.ipv6.vif6_table[ct].dev == dev) | |
7bc570c8 YH |
1533 | break; |
1534 | } | |
1535 | return ct; | |
1536 | } | |
1537 | ||
1538 | static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache) | |
1539 | { | |
1540 | int psend = -1; | |
1541 | int vif, ct; | |
1542 | ||
1543 | vif = cache->mf6c_parent; | |
1544 | cache->mfc_un.res.pkt++; | |
1545 | cache->mfc_un.res.bytes += skb->len; | |
1546 | ||
14fb64e1 YH |
1547 | /* |
1548 | * Wrong interface: drop packet and (maybe) send PIM assert. | |
1549 | */ | |
4e16880c | 1550 | if (init_net.ipv6.vif6_table[vif].dev != skb->dev) { |
14fb64e1 YH |
1551 | int true_vifi; |
1552 | ||
1553 | cache->mfc_un.res.wrong_if++; | |
1554 | true_vifi = ip6mr_find_vif(skb->dev); | |
1555 | ||
1556 | if (true_vifi >= 0 && mroute_do_assert && | |
1557 | /* pimsm uses asserts, when switching from RPT to SPT, | |
1558 | so that we cannot check that packet arrived on an oif. | |
1559 | It is bad, but otherwise we would need to move pretty | |
1560 | large chunk of pimd to kernel. Ough... --ANK | |
1561 | */ | |
1562 | (mroute_do_pim || cache->mfc_un.res.ttls[true_vifi] < 255) && | |
1563 | time_after(jiffies, | |
1564 | cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { | |
1565 | cache->mfc_un.res.last_assert = jiffies; | |
1566 | ip6mr_cache_report(skb, true_vifi, MRT6MSG_WRONGMIF); | |
1567 | } | |
1568 | goto dont_forward; | |
1569 | } | |
1570 | ||
4e16880c BT |
1571 | init_net.ipv6.vif6_table[vif].pkt_in++; |
1572 | init_net.ipv6.vif6_table[vif].bytes_in += skb->len; | |
7bc570c8 YH |
1573 | |
1574 | /* | |
1575 | * Forward the frame | |
1576 | */ | |
1577 | for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { | |
1578 | if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { | |
1579 | if (psend != -1) { | |
1580 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1581 | if (skb2) | |
1582 | ip6mr_forward2(skb2, cache, psend); | |
1583 | } | |
1584 | psend = ct; | |
1585 | } | |
1586 | } | |
1587 | if (psend != -1) { | |
1588 | ip6mr_forward2(skb, cache, psend); | |
1589 | return 0; | |
1590 | } | |
1591 | ||
14fb64e1 | 1592 | dont_forward: |
7bc570c8 YH |
1593 | kfree_skb(skb); |
1594 | return 0; | |
1595 | } | |
1596 | ||
1597 | ||
1598 | /* | |
1599 | * Multicast packets for forwarding arrive here | |
1600 | */ | |
1601 | ||
1602 | int ip6_mr_input(struct sk_buff *skb) | |
1603 | { | |
1604 | struct mfc6_cache *cache; | |
1605 | ||
1606 | read_lock(&mrt_lock); | |
1607 | cache = ip6mr_cache_find(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); | |
1608 | ||
1609 | /* | |
1610 | * No usable cache entry | |
1611 | */ | |
1612 | if (cache == NULL) { | |
1613 | int vif; | |
1614 | ||
1615 | vif = ip6mr_find_vif(skb->dev); | |
1616 | if (vif >= 0) { | |
1617 | int err = ip6mr_cache_unresolved(vif, skb); | |
1618 | read_unlock(&mrt_lock); | |
1619 | ||
1620 | return err; | |
1621 | } | |
1622 | read_unlock(&mrt_lock); | |
1623 | kfree_skb(skb); | |
1624 | return -ENODEV; | |
1625 | } | |
1626 | ||
1627 | ip6_mr_forward(skb, cache); | |
1628 | ||
1629 | read_unlock(&mrt_lock); | |
1630 | ||
1631 | return 0; | |
1632 | } | |
1633 | ||
1634 | ||
1635 | static int | |
1636 | ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm) | |
1637 | { | |
1638 | int ct; | |
1639 | struct rtnexthop *nhp; | |
4e16880c | 1640 | struct net_device *dev = init_net.ipv6.vif6_table[c->mf6c_parent].dev; |
549e028d | 1641 | u8 *b = skb_tail_pointer(skb); |
7bc570c8 YH |
1642 | struct rtattr *mp_head; |
1643 | ||
1644 | if (dev) | |
1645 | RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex); | |
1646 | ||
1647 | mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0)); | |
1648 | ||
1649 | for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { | |
1650 | if (c->mfc_un.res.ttls[ct] < 255) { | |
1651 | if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4)) | |
1652 | goto rtattr_failure; | |
1653 | nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp))); | |
1654 | nhp->rtnh_flags = 0; | |
1655 | nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; | |
4e16880c | 1656 | nhp->rtnh_ifindex = init_net.ipv6.vif6_table[ct].dev->ifindex; |
7bc570c8 YH |
1657 | nhp->rtnh_len = sizeof(*nhp); |
1658 | } | |
1659 | } | |
1660 | mp_head->rta_type = RTA_MULTIPATH; | |
549e028d | 1661 | mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head; |
7bc570c8 YH |
1662 | rtm->rtm_type = RTN_MULTICAST; |
1663 | return 1; | |
1664 | ||
1665 | rtattr_failure: | |
1666 | nlmsg_trim(skb, b); | |
1667 | return -EMSGSIZE; | |
1668 | } | |
1669 | ||
1670 | int ip6mr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait) | |
1671 | { | |
1672 | int err; | |
1673 | struct mfc6_cache *cache; | |
1674 | struct rt6_info *rt = (struct rt6_info *)skb->dst; | |
1675 | ||
1676 | read_lock(&mrt_lock); | |
1677 | cache = ip6mr_cache_find(&rt->rt6i_src.addr, &rt->rt6i_dst.addr); | |
1678 | ||
1679 | if (!cache) { | |
1680 | struct sk_buff *skb2; | |
1681 | struct ipv6hdr *iph; | |
1682 | struct net_device *dev; | |
1683 | int vif; | |
1684 | ||
1685 | if (nowait) { | |
1686 | read_unlock(&mrt_lock); | |
1687 | return -EAGAIN; | |
1688 | } | |
1689 | ||
1690 | dev = skb->dev; | |
1691 | if (dev == NULL || (vif = ip6mr_find_vif(dev)) < 0) { | |
1692 | read_unlock(&mrt_lock); | |
1693 | return -ENODEV; | |
1694 | } | |
1695 | ||
1696 | /* really correct? */ | |
1697 | skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); | |
1698 | if (!skb2) { | |
1699 | read_unlock(&mrt_lock); | |
1700 | return -ENOMEM; | |
1701 | } | |
1702 | ||
1703 | skb_reset_transport_header(skb2); | |
1704 | ||
1705 | skb_put(skb2, sizeof(struct ipv6hdr)); | |
1706 | skb_reset_network_header(skb2); | |
1707 | ||
1708 | iph = ipv6_hdr(skb2); | |
1709 | iph->version = 0; | |
1710 | iph->priority = 0; | |
1711 | iph->flow_lbl[0] = 0; | |
1712 | iph->flow_lbl[1] = 0; | |
1713 | iph->flow_lbl[2] = 0; | |
1714 | iph->payload_len = 0; | |
1715 | iph->nexthdr = IPPROTO_NONE; | |
1716 | iph->hop_limit = 0; | |
1717 | ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr); | |
1718 | ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr); | |
1719 | ||
1720 | err = ip6mr_cache_unresolved(vif, skb2); | |
1721 | read_unlock(&mrt_lock); | |
1722 | ||
1723 | return err; | |
1724 | } | |
1725 | ||
1726 | if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY)) | |
1727 | cache->mfc_flags |= MFC_NOTIFY; | |
1728 | ||
1729 | err = ip6mr_fill_mroute(skb, cache, rtm); | |
1730 | read_unlock(&mrt_lock); | |
1731 | return err; | |
1732 | } | |
1733 |