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 | |
7bc570c8 | 64 | static struct mfc6_cache *mfc_unres_queue; /* Queue of unresolved entries */ |
7bc570c8 YH |
65 | |
66 | /* Special spinlock for queue of unresolved entries */ | |
67 | static DEFINE_SPINLOCK(mfc_unres_lock); | |
68 | ||
69 | /* We return to original Alan's scheme. Hash table of resolved | |
70 | entries is changed only in process context and protected | |
71 | with weak lock mrt_lock. Queue of unresolved entries is protected | |
72 | with strong spinlock mfc_unres_lock. | |
73 | ||
74 | In this case data path is free of exclusive locks at all. | |
75 | */ | |
76 | ||
77 | static struct kmem_cache *mrt_cachep __read_mostly; | |
78 | ||
79 | static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache); | |
6ac7eb08 | 80 | static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert); |
7bc570c8 YH |
81 | static int ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm); |
82 | ||
14fb64e1 YH |
83 | #ifdef CONFIG_IPV6_PIMSM_V2 |
84 | static struct inet6_protocol pim6_protocol; | |
85 | #endif | |
86 | ||
7bc570c8 YH |
87 | static struct timer_list ipmr_expire_timer; |
88 | ||
89 | ||
90 | #ifdef CONFIG_PROC_FS | |
91 | ||
92 | struct ipmr_mfc_iter { | |
8b90fc7e | 93 | struct seq_net_private p; |
7bc570c8 YH |
94 | struct mfc6_cache **cache; |
95 | int ct; | |
96 | }; | |
97 | ||
98 | ||
8b90fc7e BT |
99 | static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net, |
100 | struct ipmr_mfc_iter *it, loff_t pos) | |
7bc570c8 YH |
101 | { |
102 | struct mfc6_cache *mfc; | |
103 | ||
8b90fc7e | 104 | it->cache = net->ipv6.mfc6_cache_array; |
7bc570c8 | 105 | read_lock(&mrt_lock); |
4a6258a0 | 106 | for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) |
8b90fc7e | 107 | for (mfc = net->ipv6.mfc6_cache_array[it->ct]; |
4a6258a0 | 108 | mfc; mfc = mfc->next) |
7bc570c8 YH |
109 | if (pos-- == 0) |
110 | return mfc; | |
111 | read_unlock(&mrt_lock); | |
112 | ||
113 | it->cache = &mfc_unres_queue; | |
114 | spin_lock_bh(&mfc_unres_lock); | |
115 | for (mfc = mfc_unres_queue; mfc; mfc = mfc->next) | |
8b90fc7e BT |
116 | if (net_eq(mfc6_net(mfc), net) && |
117 | pos-- == 0) | |
7bc570c8 YH |
118 | return mfc; |
119 | spin_unlock_bh(&mfc_unres_lock); | |
120 | ||
121 | it->cache = NULL; | |
122 | return NULL; | |
123 | } | |
124 | ||
125 | ||
126 | ||
127 | ||
128 | /* | |
129 | * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif | |
130 | */ | |
131 | ||
132 | struct ipmr_vif_iter { | |
8b90fc7e | 133 | struct seq_net_private p; |
7bc570c8 YH |
134 | int ct; |
135 | }; | |
136 | ||
8b90fc7e BT |
137 | static struct mif_device *ip6mr_vif_seq_idx(struct net *net, |
138 | struct ipmr_vif_iter *iter, | |
7bc570c8 YH |
139 | loff_t pos) |
140 | { | |
8b90fc7e BT |
141 | for (iter->ct = 0; iter->ct < net->ipv6.maxvif; ++iter->ct) { |
142 | if (!MIF_EXISTS(net, iter->ct)) | |
7bc570c8 YH |
143 | continue; |
144 | if (pos-- == 0) | |
8b90fc7e | 145 | return &net->ipv6.vif6_table[iter->ct]; |
7bc570c8 YH |
146 | } |
147 | return NULL; | |
148 | } | |
149 | ||
150 | static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos) | |
151 | __acquires(mrt_lock) | |
152 | { | |
8b90fc7e BT |
153 | struct net *net = seq_file_net(seq); |
154 | ||
7bc570c8 | 155 | read_lock(&mrt_lock); |
8b90fc7e BT |
156 | return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1) |
157 | : SEQ_START_TOKEN; | |
7bc570c8 YH |
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; | |
8b90fc7e | 163 | struct net *net = seq_file_net(seq); |
7bc570c8 YH |
164 | |
165 | ++*pos; | |
166 | if (v == SEQ_START_TOKEN) | |
8b90fc7e | 167 | return ip6mr_vif_seq_idx(net, iter, 0); |
7bc570c8 | 168 | |
8b90fc7e BT |
169 | while (++iter->ct < net->ipv6.maxvif) { |
170 | if (!MIF_EXISTS(net, iter->ct)) | |
7bc570c8 | 171 | continue; |
8b90fc7e | 172 | return &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 | { | |
8b90fc7e BT |
185 | struct net *net = seq_file_net(seq); |
186 | ||
7bc570c8 YH |
187 | if (v == SEQ_START_TOKEN) { |
188 | seq_puts(seq, | |
189 | "Interface BytesIn PktsIn BytesOut PktsOut Flags\n"); | |
190 | } else { | |
191 | const struct mif_device *vif = v; | |
192 | const char *name = vif->dev ? vif->dev->name : "none"; | |
193 | ||
194 | seq_printf(seq, | |
d430a227 | 195 | "%2td %-10s %8ld %7ld %8ld %7ld %05X\n", |
8b90fc7e | 196 | vif - net->ipv6.vif6_table, |
7bc570c8 YH |
197 | name, vif->bytes_in, vif->pkt_in, |
198 | vif->bytes_out, vif->pkt_out, | |
199 | vif->flags); | |
200 | } | |
201 | return 0; | |
202 | } | |
203 | ||
204 | static struct seq_operations ip6mr_vif_seq_ops = { | |
205 | .start = ip6mr_vif_seq_start, | |
206 | .next = ip6mr_vif_seq_next, | |
207 | .stop = ip6mr_vif_seq_stop, | |
208 | .show = ip6mr_vif_seq_show, | |
209 | }; | |
210 | ||
211 | static int ip6mr_vif_open(struct inode *inode, struct file *file) | |
212 | { | |
8b90fc7e BT |
213 | return seq_open_net(inode, file, &ip6mr_vif_seq_ops, |
214 | sizeof(struct ipmr_vif_iter)); | |
7bc570c8 YH |
215 | } |
216 | ||
217 | static struct file_operations ip6mr_vif_fops = { | |
218 | .owner = THIS_MODULE, | |
219 | .open = ip6mr_vif_open, | |
220 | .read = seq_read, | |
221 | .llseek = seq_lseek, | |
8b90fc7e | 222 | .release = seq_release_net, |
7bc570c8 YH |
223 | }; |
224 | ||
225 | static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) | |
226 | { | |
8b90fc7e BT |
227 | struct net *net = seq_file_net(seq); |
228 | ||
229 | return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1) | |
230 | : SEQ_START_TOKEN; | |
7bc570c8 YH |
231 | } |
232 | ||
233 | static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
234 | { | |
235 | struct mfc6_cache *mfc = v; | |
236 | struct ipmr_mfc_iter *it = seq->private; | |
8b90fc7e | 237 | struct net *net = seq_file_net(seq); |
7bc570c8 YH |
238 | |
239 | ++*pos; | |
240 | ||
241 | if (v == SEQ_START_TOKEN) | |
8b90fc7e | 242 | return ipmr_mfc_seq_idx(net, seq->private, 0); |
7bc570c8 YH |
243 | |
244 | if (mfc->next) | |
245 | return mfc->next; | |
246 | ||
247 | if (it->cache == &mfc_unres_queue) | |
248 | goto end_of_list; | |
249 | ||
8b90fc7e | 250 | BUG_ON(it->cache != net->ipv6.mfc6_cache_array); |
7bc570c8 | 251 | |
4a6258a0 | 252 | while (++it->ct < MFC6_LINES) { |
8b90fc7e | 253 | mfc = net->ipv6.mfc6_cache_array[it->ct]; |
7bc570c8 YH |
254 | if (mfc) |
255 | return mfc; | |
256 | } | |
257 | ||
258 | /* exhausted cache_array, show unresolved */ | |
259 | read_unlock(&mrt_lock); | |
260 | it->cache = &mfc_unres_queue; | |
261 | it->ct = 0; | |
262 | ||
263 | spin_lock_bh(&mfc_unres_lock); | |
264 | mfc = mfc_unres_queue; | |
265 | if (mfc) | |
266 | return mfc; | |
267 | ||
268 | end_of_list: | |
269 | spin_unlock_bh(&mfc_unres_lock); | |
270 | it->cache = NULL; | |
271 | ||
272 | return NULL; | |
273 | } | |
274 | ||
275 | static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) | |
276 | { | |
277 | struct ipmr_mfc_iter *it = seq->private; | |
8b90fc7e | 278 | struct net *net = seq_file_net(seq); |
7bc570c8 YH |
279 | |
280 | if (it->cache == &mfc_unres_queue) | |
281 | spin_unlock_bh(&mfc_unres_lock); | |
8b90fc7e | 282 | else if (it->cache == net->ipv6.mfc6_cache_array) |
7bc570c8 YH |
283 | read_unlock(&mrt_lock); |
284 | } | |
285 | ||
286 | static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) | |
287 | { | |
288 | int n; | |
8b90fc7e | 289 | struct net *net = seq_file_net(seq); |
7bc570c8 YH |
290 | |
291 | if (v == SEQ_START_TOKEN) { | |
292 | seq_puts(seq, | |
293 | "Group " | |
294 | "Origin " | |
295 | "Iif Pkts Bytes Wrong Oifs\n"); | |
296 | } else { | |
297 | const struct mfc6_cache *mfc = v; | |
298 | const struct ipmr_mfc_iter *it = seq->private; | |
299 | ||
999890b2 | 300 | seq_printf(seq, "%pI6 %pI6 %-3hd", |
0c6ce78a | 301 | &mfc->mf6c_mcastgrp, &mfc->mf6c_origin, |
1ea472e2 | 302 | mfc->mf6c_parent); |
7bc570c8 YH |
303 | |
304 | if (it->cache != &mfc_unres_queue) { | |
1ea472e2 BT |
305 | seq_printf(seq, " %8lu %8lu %8lu", |
306 | mfc->mfc_un.res.pkt, | |
307 | mfc->mfc_un.res.bytes, | |
308 | mfc->mfc_un.res.wrong_if); | |
7bc570c8 YH |
309 | for (n = mfc->mfc_un.res.minvif; |
310 | n < mfc->mfc_un.res.maxvif; n++) { | |
8b90fc7e | 311 | if (MIF_EXISTS(net, n) && |
7bc570c8 YH |
312 | mfc->mfc_un.res.ttls[n] < 255) |
313 | seq_printf(seq, | |
314 | " %2d:%-3d", | |
315 | n, mfc->mfc_un.res.ttls[n]); | |
316 | } | |
1ea472e2 BT |
317 | } else { |
318 | /* unresolved mfc_caches don't contain | |
319 | * pkt, bytes and wrong_if values | |
320 | */ | |
321 | seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); | |
7bc570c8 YH |
322 | } |
323 | seq_putc(seq, '\n'); | |
324 | } | |
325 | return 0; | |
326 | } | |
327 | ||
328 | static struct seq_operations ipmr_mfc_seq_ops = { | |
329 | .start = ipmr_mfc_seq_start, | |
330 | .next = ipmr_mfc_seq_next, | |
331 | .stop = ipmr_mfc_seq_stop, | |
332 | .show = ipmr_mfc_seq_show, | |
333 | }; | |
334 | ||
335 | static int ipmr_mfc_open(struct inode *inode, struct file *file) | |
336 | { | |
8b90fc7e BT |
337 | return seq_open_net(inode, file, &ipmr_mfc_seq_ops, |
338 | sizeof(struct ipmr_mfc_iter)); | |
7bc570c8 YH |
339 | } |
340 | ||
341 | static struct file_operations ip6mr_mfc_fops = { | |
342 | .owner = THIS_MODULE, | |
343 | .open = ipmr_mfc_open, | |
344 | .read = seq_read, | |
345 | .llseek = seq_lseek, | |
8b90fc7e | 346 | .release = seq_release_net, |
7bc570c8 YH |
347 | }; |
348 | #endif | |
349 | ||
14fb64e1 | 350 | #ifdef CONFIG_IPV6_PIMSM_V2 |
14fb64e1 YH |
351 | |
352 | static int pim6_rcv(struct sk_buff *skb) | |
353 | { | |
354 | struct pimreghdr *pim; | |
355 | struct ipv6hdr *encap; | |
356 | struct net_device *reg_dev = NULL; | |
950d5704 | 357 | int reg_vif_num = init_net.ipv6.mroute_reg_vif_num; |
14fb64e1 YH |
358 | |
359 | if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap))) | |
360 | goto drop; | |
361 | ||
362 | pim = (struct pimreghdr *)skb_transport_header(skb); | |
363 | if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) || | |
364 | (pim->flags & PIM_NULL_REGISTER) || | |
365 | (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && | |
ec6b486f | 366 | csum_fold(skb_checksum(skb, 0, skb->len, 0)))) |
14fb64e1 YH |
367 | goto drop; |
368 | ||
369 | /* check if the inner packet is destined to mcast group */ | |
370 | encap = (struct ipv6hdr *)(skb_transport_header(skb) + | |
371 | sizeof(*pim)); | |
372 | ||
373 | if (!ipv6_addr_is_multicast(&encap->daddr) || | |
374 | encap->payload_len == 0 || | |
375 | ntohs(encap->payload_len) + sizeof(*pim) > skb->len) | |
376 | goto drop; | |
377 | ||
378 | read_lock(&mrt_lock); | |
379 | if (reg_vif_num >= 0) | |
4e16880c | 380 | reg_dev = init_net.ipv6.vif6_table[reg_vif_num].dev; |
14fb64e1 YH |
381 | if (reg_dev) |
382 | dev_hold(reg_dev); | |
383 | read_unlock(&mrt_lock); | |
384 | ||
385 | if (reg_dev == NULL) | |
386 | goto drop; | |
387 | ||
388 | skb->mac_header = skb->network_header; | |
389 | skb_pull(skb, (u8 *)encap - skb->data); | |
390 | skb_reset_network_header(skb); | |
391 | skb->dev = reg_dev; | |
392 | skb->protocol = htons(ETH_P_IP); | |
393 | skb->ip_summed = 0; | |
394 | skb->pkt_type = PACKET_HOST; | |
395 | dst_release(skb->dst); | |
dc58c78c PE |
396 | reg_dev->stats.rx_bytes += skb->len; |
397 | reg_dev->stats.rx_packets++; | |
14fb64e1 YH |
398 | skb->dst = NULL; |
399 | nf_reset(skb); | |
400 | netif_rx(skb); | |
401 | dev_put(reg_dev); | |
402 | return 0; | |
403 | drop: | |
404 | kfree_skb(skb); | |
405 | return 0; | |
406 | } | |
407 | ||
408 | static struct inet6_protocol pim6_protocol = { | |
409 | .handler = pim6_rcv, | |
410 | }; | |
411 | ||
412 | /* Service routines creating virtual interfaces: PIMREG */ | |
413 | ||
414 | static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) | |
415 | { | |
416 | read_lock(&mrt_lock); | |
dc58c78c PE |
417 | dev->stats.tx_bytes += skb->len; |
418 | dev->stats.tx_packets++; | |
950d5704 | 419 | ip6mr_cache_report(skb, init_net.ipv6.mroute_reg_vif_num, MRT6MSG_WHOLEPKT); |
14fb64e1 YH |
420 | read_unlock(&mrt_lock); |
421 | kfree_skb(skb); | |
422 | return 0; | |
423 | } | |
424 | ||
007c3838 SH |
425 | static const struct net_device_ops reg_vif_netdev_ops = { |
426 | .ndo_start_xmit = reg_vif_xmit, | |
427 | }; | |
428 | ||
14fb64e1 YH |
429 | static void reg_vif_setup(struct net_device *dev) |
430 | { | |
431 | dev->type = ARPHRD_PIMREG; | |
432 | dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8; | |
433 | dev->flags = IFF_NOARP; | |
007c3838 | 434 | dev->netdev_ops = ®_vif_netdev_ops; |
14fb64e1 YH |
435 | dev->destructor = free_netdev; |
436 | } | |
437 | ||
438 | static struct net_device *ip6mr_reg_vif(void) | |
439 | { | |
440 | struct net_device *dev; | |
14fb64e1 | 441 | |
dc58c78c | 442 | dev = alloc_netdev(0, "pim6reg", reg_vif_setup); |
14fb64e1 YH |
443 | if (dev == NULL) |
444 | return NULL; | |
445 | ||
446 | if (register_netdevice(dev)) { | |
447 | free_netdev(dev); | |
448 | return NULL; | |
449 | } | |
450 | dev->iflink = 0; | |
451 | ||
14fb64e1 YH |
452 | if (dev_open(dev)) |
453 | goto failure; | |
454 | ||
7af3db78 | 455 | dev_hold(dev); |
14fb64e1 YH |
456 | return dev; |
457 | ||
458 | failure: | |
459 | /* allow the register to be completed before unregistering. */ | |
460 | rtnl_unlock(); | |
461 | rtnl_lock(); | |
462 | ||
463 | unregister_netdevice(dev); | |
464 | return NULL; | |
465 | } | |
466 | #endif | |
467 | ||
7bc570c8 YH |
468 | /* |
469 | * Delete a VIF entry | |
470 | */ | |
471 | ||
472 | static int mif6_delete(int vifi) | |
473 | { | |
474 | struct mif_device *v; | |
475 | struct net_device *dev; | |
4e16880c | 476 | if (vifi < 0 || vifi >= init_net.ipv6.maxvif) |
7bc570c8 YH |
477 | return -EADDRNOTAVAIL; |
478 | ||
4e16880c | 479 | v = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 YH |
480 | |
481 | write_lock_bh(&mrt_lock); | |
482 | dev = v->dev; | |
483 | v->dev = NULL; | |
484 | ||
485 | if (!dev) { | |
486 | write_unlock_bh(&mrt_lock); | |
487 | return -EADDRNOTAVAIL; | |
488 | } | |
489 | ||
14fb64e1 | 490 | #ifdef CONFIG_IPV6_PIMSM_V2 |
950d5704 BT |
491 | if (vifi == init_net.ipv6.mroute_reg_vif_num) |
492 | init_net.ipv6.mroute_reg_vif_num = -1; | |
14fb64e1 YH |
493 | #endif |
494 | ||
4e16880c | 495 | if (vifi + 1 == init_net.ipv6.maxvif) { |
7bc570c8 YH |
496 | int tmp; |
497 | for (tmp = vifi - 1; tmp >= 0; tmp--) { | |
4e16880c | 498 | if (MIF_EXISTS(&init_net, tmp)) |
7bc570c8 YH |
499 | break; |
500 | } | |
4e16880c | 501 | init_net.ipv6.maxvif = tmp + 1; |
7bc570c8 YH |
502 | } |
503 | ||
504 | write_unlock_bh(&mrt_lock); | |
505 | ||
506 | dev_set_allmulti(dev, -1); | |
507 | ||
508 | if (v->flags & MIFF_REGISTER) | |
509 | unregister_netdevice(dev); | |
510 | ||
511 | dev_put(dev); | |
512 | return 0; | |
513 | } | |
514 | ||
58701ad4 BT |
515 | static inline void ip6mr_cache_free(struct mfc6_cache *c) |
516 | { | |
517 | release_net(mfc6_net(c)); | |
518 | kmem_cache_free(mrt_cachep, c); | |
519 | } | |
520 | ||
7bc570c8 YH |
521 | /* Destroy an unresolved cache entry, killing queued skbs |
522 | and reporting error to netlink readers. | |
523 | */ | |
524 | ||
525 | static void ip6mr_destroy_unres(struct mfc6_cache *c) | |
526 | { | |
527 | struct sk_buff *skb; | |
528 | ||
4045e57c | 529 | atomic_dec(&init_net.ipv6.cache_resolve_queue_len); |
7bc570c8 YH |
530 | |
531 | while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) { | |
532 | if (ipv6_hdr(skb)->version == 0) { | |
533 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); | |
534 | nlh->nlmsg_type = NLMSG_ERROR; | |
535 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
536 | skb_trim(skb, nlh->nlmsg_len); | |
537 | ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT; | |
538 | rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); | |
539 | } else | |
540 | kfree_skb(skb); | |
541 | } | |
542 | ||
58701ad4 | 543 | ip6mr_cache_free(c); |
7bc570c8 YH |
544 | } |
545 | ||
546 | ||
547 | /* Single timer process for all the unresolved queue. */ | |
548 | ||
549 | static void ipmr_do_expire_process(unsigned long dummy) | |
550 | { | |
551 | unsigned long now = jiffies; | |
552 | unsigned long expires = 10 * HZ; | |
553 | struct mfc6_cache *c, **cp; | |
554 | ||
555 | cp = &mfc_unres_queue; | |
556 | ||
557 | while ((c = *cp) != NULL) { | |
558 | if (time_after(c->mfc_un.unres.expires, now)) { | |
559 | /* not yet... */ | |
560 | unsigned long interval = c->mfc_un.unres.expires - now; | |
561 | if (interval < expires) | |
562 | expires = interval; | |
563 | cp = &c->next; | |
564 | continue; | |
565 | } | |
566 | ||
567 | *cp = c->next; | |
568 | ip6mr_destroy_unres(c); | |
569 | } | |
570 | ||
4045e57c | 571 | if (mfc_unres_queue != NULL) |
7bc570c8 YH |
572 | mod_timer(&ipmr_expire_timer, jiffies + expires); |
573 | } | |
574 | ||
575 | static void ipmr_expire_process(unsigned long dummy) | |
576 | { | |
577 | if (!spin_trylock(&mfc_unres_lock)) { | |
578 | mod_timer(&ipmr_expire_timer, jiffies + 1); | |
579 | return; | |
580 | } | |
581 | ||
4045e57c | 582 | if (mfc_unres_queue != NULL) |
7bc570c8 YH |
583 | ipmr_do_expire_process(dummy); |
584 | ||
585 | spin_unlock(&mfc_unres_lock); | |
586 | } | |
587 | ||
588 | /* Fill oifs list. It is called under write locked mrt_lock. */ | |
589 | ||
590 | static void ip6mr_update_thresholds(struct mfc6_cache *cache, unsigned char *ttls) | |
591 | { | |
592 | int vifi; | |
593 | ||
6ac7eb08 | 594 | cache->mfc_un.res.minvif = MAXMIFS; |
7bc570c8 | 595 | cache->mfc_un.res.maxvif = 0; |
6ac7eb08 | 596 | memset(cache->mfc_un.res.ttls, 255, MAXMIFS); |
7bc570c8 | 597 | |
4e16880c BT |
598 | for (vifi = 0; vifi < init_net.ipv6.maxvif; vifi++) { |
599 | if (MIF_EXISTS(&init_net, vifi) && | |
600 | ttls[vifi] && ttls[vifi] < 255) { | |
7bc570c8 YH |
601 | cache->mfc_un.res.ttls[vifi] = ttls[vifi]; |
602 | if (cache->mfc_un.res.minvif > vifi) | |
603 | cache->mfc_un.res.minvif = vifi; | |
604 | if (cache->mfc_un.res.maxvif <= vifi) | |
605 | cache->mfc_un.res.maxvif = vifi + 1; | |
606 | } | |
607 | } | |
608 | } | |
609 | ||
610 | static int mif6_add(struct mif6ctl *vifc, int mrtsock) | |
611 | { | |
612 | int vifi = vifc->mif6c_mifi; | |
4e16880c | 613 | struct mif_device *v = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 | 614 | struct net_device *dev; |
5ae7b444 | 615 | int err; |
7bc570c8 YH |
616 | |
617 | /* Is vif busy ? */ | |
4e16880c | 618 | if (MIF_EXISTS(&init_net, vifi)) |
7bc570c8 YH |
619 | return -EADDRINUSE; |
620 | ||
621 | switch (vifc->mif6c_flags) { | |
14fb64e1 YH |
622 | #ifdef CONFIG_IPV6_PIMSM_V2 |
623 | case MIFF_REGISTER: | |
624 | /* | |
625 | * Special Purpose VIF in PIM | |
626 | * All the packets will be sent to the daemon | |
627 | */ | |
950d5704 | 628 | if (init_net.ipv6.mroute_reg_vif_num >= 0) |
14fb64e1 YH |
629 | return -EADDRINUSE; |
630 | dev = ip6mr_reg_vif(); | |
631 | if (!dev) | |
632 | return -ENOBUFS; | |
5ae7b444 WC |
633 | err = dev_set_allmulti(dev, 1); |
634 | if (err) { | |
635 | unregister_netdevice(dev); | |
7af3db78 | 636 | dev_put(dev); |
5ae7b444 WC |
637 | return err; |
638 | } | |
14fb64e1 YH |
639 | break; |
640 | #endif | |
7bc570c8 YH |
641 | case 0: |
642 | dev = dev_get_by_index(&init_net, vifc->mif6c_pifi); | |
643 | if (!dev) | |
644 | return -EADDRNOTAVAIL; | |
5ae7b444 | 645 | err = dev_set_allmulti(dev, 1); |
7af3db78 WC |
646 | if (err) { |
647 | dev_put(dev); | |
5ae7b444 | 648 | return err; |
7af3db78 | 649 | } |
7bc570c8 YH |
650 | break; |
651 | default: | |
652 | return -EINVAL; | |
653 | } | |
654 | ||
7bc570c8 YH |
655 | /* |
656 | * Fill in the VIF structures | |
657 | */ | |
658 | v->rate_limit = vifc->vifc_rate_limit; | |
659 | v->flags = vifc->mif6c_flags; | |
660 | if (!mrtsock) | |
661 | v->flags |= VIFF_STATIC; | |
662 | v->threshold = vifc->vifc_threshold; | |
663 | v->bytes_in = 0; | |
664 | v->bytes_out = 0; | |
665 | v->pkt_in = 0; | |
666 | v->pkt_out = 0; | |
667 | v->link = dev->ifindex; | |
668 | if (v->flags & MIFF_REGISTER) | |
669 | v->link = dev->iflink; | |
670 | ||
671 | /* And finish update writing critical data */ | |
672 | write_lock_bh(&mrt_lock); | |
7bc570c8 | 673 | v->dev = dev; |
14fb64e1 YH |
674 | #ifdef CONFIG_IPV6_PIMSM_V2 |
675 | if (v->flags & MIFF_REGISTER) | |
950d5704 | 676 | init_net.ipv6.mroute_reg_vif_num = vifi; |
14fb64e1 | 677 | #endif |
4e16880c BT |
678 | if (vifi + 1 > init_net.ipv6.maxvif) |
679 | init_net.ipv6.maxvif = vifi + 1; | |
7bc570c8 YH |
680 | write_unlock_bh(&mrt_lock); |
681 | return 0; | |
682 | } | |
683 | ||
684 | static struct mfc6_cache *ip6mr_cache_find(struct in6_addr *origin, struct in6_addr *mcastgrp) | |
685 | { | |
686 | int line = MFC6_HASH(mcastgrp, origin); | |
687 | struct mfc6_cache *c; | |
688 | ||
4a6258a0 | 689 | for (c = init_net.ipv6.mfc6_cache_array[line]; c; c = c->next) { |
7bc570c8 YH |
690 | if (ipv6_addr_equal(&c->mf6c_origin, origin) && |
691 | ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) | |
692 | break; | |
693 | } | |
694 | return c; | |
695 | } | |
696 | ||
697 | /* | |
698 | * Allocate a multicast cache entry | |
699 | */ | |
58701ad4 | 700 | static struct mfc6_cache *ip6mr_cache_alloc(struct net *net) |
7bc570c8 | 701 | { |
36cbac59 | 702 | struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); |
7bc570c8 YH |
703 | if (c == NULL) |
704 | return NULL; | |
6ac7eb08 | 705 | c->mfc_un.res.minvif = MAXMIFS; |
58701ad4 | 706 | mfc6_net_set(c, net); |
7bc570c8 YH |
707 | return c; |
708 | } | |
709 | ||
58701ad4 | 710 | static struct mfc6_cache *ip6mr_cache_alloc_unres(struct net *net) |
7bc570c8 | 711 | { |
36cbac59 | 712 | struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); |
7bc570c8 YH |
713 | if (c == NULL) |
714 | return NULL; | |
7bc570c8 YH |
715 | skb_queue_head_init(&c->mfc_un.unres.unresolved); |
716 | c->mfc_un.unres.expires = jiffies + 10 * HZ; | |
58701ad4 | 717 | mfc6_net_set(c, net); |
7bc570c8 YH |
718 | return c; |
719 | } | |
720 | ||
721 | /* | |
722 | * A cache entry has gone into a resolved state from queued | |
723 | */ | |
724 | ||
725 | static void ip6mr_cache_resolve(struct mfc6_cache *uc, struct mfc6_cache *c) | |
726 | { | |
727 | struct sk_buff *skb; | |
728 | ||
729 | /* | |
730 | * Play the pending entries through our router | |
731 | */ | |
732 | ||
733 | while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { | |
734 | if (ipv6_hdr(skb)->version == 0) { | |
735 | int err; | |
736 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); | |
737 | ||
738 | if (ip6mr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) { | |
549e028d | 739 | nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; |
7bc570c8 YH |
740 | } else { |
741 | nlh->nlmsg_type = NLMSG_ERROR; | |
742 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
743 | skb_trim(skb, nlh->nlmsg_len); | |
744 | ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE; | |
745 | } | |
746 | err = rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); | |
747 | } else | |
748 | ip6_mr_forward(skb, c); | |
749 | } | |
750 | } | |
751 | ||
752 | /* | |
753 | * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd | |
754 | * expects the following bizarre scheme. | |
755 | * | |
756 | * Called under mrt_lock. | |
757 | */ | |
758 | ||
6ac7eb08 | 759 | static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert) |
7bc570c8 YH |
760 | { |
761 | struct sk_buff *skb; | |
762 | struct mrt6msg *msg; | |
763 | int ret; | |
764 | ||
14fb64e1 YH |
765 | #ifdef CONFIG_IPV6_PIMSM_V2 |
766 | if (assert == MRT6MSG_WHOLEPKT) | |
767 | skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) | |
768 | +sizeof(*msg)); | |
769 | else | |
770 | #endif | |
771 | skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); | |
7bc570c8 YH |
772 | |
773 | if (!skb) | |
774 | return -ENOBUFS; | |
775 | ||
776 | /* I suppose that internal messages | |
777 | * do not require checksums */ | |
778 | ||
779 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
780 | ||
14fb64e1 YH |
781 | #ifdef CONFIG_IPV6_PIMSM_V2 |
782 | if (assert == MRT6MSG_WHOLEPKT) { | |
783 | /* Ugly, but we have no choice with this interface. | |
784 | Duplicate old header, fix length etc. | |
785 | And all this only to mangle msg->im6_msgtype and | |
786 | to set msg->im6_mbz to "mbz" :-) | |
787 | */ | |
788 | skb_push(skb, -skb_network_offset(pkt)); | |
789 | ||
790 | skb_push(skb, sizeof(*msg)); | |
791 | skb_reset_transport_header(skb); | |
792 | msg = (struct mrt6msg *)skb_transport_header(skb); | |
793 | msg->im6_mbz = 0; | |
794 | msg->im6_msgtype = MRT6MSG_WHOLEPKT; | |
950d5704 | 795 | msg->im6_mif = init_net.ipv6.mroute_reg_vif_num; |
14fb64e1 YH |
796 | msg->im6_pad = 0; |
797 | ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr); | |
798 | ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr); | |
799 | ||
800 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
801 | } else | |
802 | #endif | |
803 | { | |
7bc570c8 YH |
804 | /* |
805 | * Copy the IP header | |
806 | */ | |
807 | ||
808 | skb_put(skb, sizeof(struct ipv6hdr)); | |
809 | skb_reset_network_header(skb); | |
810 | skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); | |
811 | ||
812 | /* | |
813 | * Add our header | |
814 | */ | |
815 | skb_put(skb, sizeof(*msg)); | |
816 | skb_reset_transport_header(skb); | |
817 | msg = (struct mrt6msg *)skb_transport_header(skb); | |
818 | ||
819 | msg->im6_mbz = 0; | |
820 | msg->im6_msgtype = assert; | |
6ac7eb08 | 821 | msg->im6_mif = mifi; |
7bc570c8 YH |
822 | msg->im6_pad = 0; |
823 | ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr); | |
824 | ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr); | |
825 | ||
826 | skb->dst = dst_clone(pkt->dst); | |
827 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
828 | ||
829 | skb_pull(skb, sizeof(struct ipv6hdr)); | |
14fb64e1 | 830 | } |
7bc570c8 | 831 | |
bd91b8bf | 832 | if (init_net.ipv6.mroute6_sk == NULL) { |
7bc570c8 YH |
833 | kfree_skb(skb); |
834 | return -EINVAL; | |
835 | } | |
836 | ||
837 | /* | |
838 | * Deliver to user space multicast routing algorithms | |
839 | */ | |
bd91b8bf BT |
840 | ret = sock_queue_rcv_skb(init_net.ipv6.mroute6_sk, skb); |
841 | if (ret < 0) { | |
7bc570c8 YH |
842 | if (net_ratelimit()) |
843 | printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n"); | |
844 | kfree_skb(skb); | |
845 | } | |
846 | ||
847 | return ret; | |
848 | } | |
849 | ||
850 | /* | |
851 | * Queue a packet for resolution. It gets locked cache entry! | |
852 | */ | |
853 | ||
854 | static int | |
6ac7eb08 | 855 | ip6mr_cache_unresolved(mifi_t mifi, struct sk_buff *skb) |
7bc570c8 YH |
856 | { |
857 | int err; | |
858 | struct mfc6_cache *c; | |
859 | ||
860 | spin_lock_bh(&mfc_unres_lock); | |
861 | for (c = mfc_unres_queue; c; c = c->next) { | |
4045e57c BT |
862 | if (net_eq(mfc6_net(c), &init_net) && |
863 | ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && | |
7bc570c8 YH |
864 | ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) |
865 | break; | |
866 | } | |
867 | ||
868 | if (c == NULL) { | |
869 | /* | |
870 | * Create a new entry if allowable | |
871 | */ | |
872 | ||
4045e57c | 873 | if (atomic_read(&init_net.ipv6.cache_resolve_queue_len) >= 10 || |
58701ad4 | 874 | (c = ip6mr_cache_alloc_unres(&init_net)) == NULL) { |
7bc570c8 YH |
875 | spin_unlock_bh(&mfc_unres_lock); |
876 | ||
877 | kfree_skb(skb); | |
878 | return -ENOBUFS; | |
879 | } | |
880 | ||
881 | /* | |
882 | * Fill in the new cache entry | |
883 | */ | |
884 | c->mf6c_parent = -1; | |
885 | c->mf6c_origin = ipv6_hdr(skb)->saddr; | |
886 | c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; | |
887 | ||
888 | /* | |
889 | * Reflect first query at pim6sd | |
890 | */ | |
6ac7eb08 | 891 | if ((err = ip6mr_cache_report(skb, mifi, MRT6MSG_NOCACHE)) < 0) { |
7bc570c8 YH |
892 | /* If the report failed throw the cache entry |
893 | out - Brad Parker | |
894 | */ | |
895 | spin_unlock_bh(&mfc_unres_lock); | |
896 | ||
58701ad4 | 897 | ip6mr_cache_free(c); |
7bc570c8 YH |
898 | kfree_skb(skb); |
899 | return err; | |
900 | } | |
901 | ||
4045e57c | 902 | atomic_inc(&init_net.ipv6.cache_resolve_queue_len); |
7bc570c8 YH |
903 | c->next = mfc_unres_queue; |
904 | mfc_unres_queue = c; | |
905 | ||
906 | ipmr_do_expire_process(1); | |
907 | } | |
908 | ||
909 | /* | |
910 | * See if we can append the packet | |
911 | */ | |
912 | if (c->mfc_un.unres.unresolved.qlen > 3) { | |
913 | kfree_skb(skb); | |
914 | err = -ENOBUFS; | |
915 | } else { | |
916 | skb_queue_tail(&c->mfc_un.unres.unresolved, skb); | |
917 | err = 0; | |
918 | } | |
919 | ||
920 | spin_unlock_bh(&mfc_unres_lock); | |
921 | return err; | |
922 | } | |
923 | ||
924 | /* | |
925 | * MFC6 cache manipulation by user space | |
926 | */ | |
927 | ||
928 | static int ip6mr_mfc_delete(struct mf6cctl *mfc) | |
929 | { | |
930 | int line; | |
931 | struct mfc6_cache *c, **cp; | |
932 | ||
933 | line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); | |
934 | ||
4a6258a0 BT |
935 | for (cp = &init_net.ipv6.mfc6_cache_array[line]; |
936 | (c = *cp) != NULL; cp = &c->next) { | |
7bc570c8 YH |
937 | if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && |
938 | ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) { | |
939 | write_lock_bh(&mrt_lock); | |
940 | *cp = c->next; | |
941 | write_unlock_bh(&mrt_lock); | |
942 | ||
58701ad4 | 943 | ip6mr_cache_free(c); |
7bc570c8 YH |
944 | return 0; |
945 | } | |
946 | } | |
947 | return -ENOENT; | |
948 | } | |
949 | ||
950 | static int ip6mr_device_event(struct notifier_block *this, | |
951 | unsigned long event, void *ptr) | |
952 | { | |
953 | struct net_device *dev = ptr; | |
954 | struct mif_device *v; | |
955 | int ct; | |
956 | ||
721499e8 | 957 | if (!net_eq(dev_net(dev), &init_net)) |
7bc570c8 YH |
958 | return NOTIFY_DONE; |
959 | ||
960 | if (event != NETDEV_UNREGISTER) | |
961 | return NOTIFY_DONE; | |
962 | ||
4e16880c BT |
963 | v = &init_net.ipv6.vif6_table[0]; |
964 | for (ct = 0; ct < init_net.ipv6.maxvif; ct++, v++) { | |
7bc570c8 YH |
965 | if (v->dev == dev) |
966 | mif6_delete(ct); | |
967 | } | |
968 | return NOTIFY_DONE; | |
969 | } | |
970 | ||
971 | static struct notifier_block ip6_mr_notifier = { | |
972 | .notifier_call = ip6mr_device_event | |
973 | }; | |
974 | ||
975 | /* | |
976 | * Setup for IP multicast routing | |
977 | */ | |
978 | ||
4e16880c BT |
979 | static int __net_init ip6mr_net_init(struct net *net) |
980 | { | |
981 | int err = 0; | |
4e16880c BT |
982 | net->ipv6.vif6_table = kcalloc(MAXMIFS, sizeof(struct mif_device), |
983 | GFP_KERNEL); | |
984 | if (!net->ipv6.vif6_table) { | |
985 | err = -ENOMEM; | |
986 | goto fail; | |
987 | } | |
4a6258a0 BT |
988 | |
989 | /* Forwarding cache */ | |
990 | net->ipv6.mfc6_cache_array = kcalloc(MFC6_LINES, | |
991 | sizeof(struct mfc6_cache *), | |
992 | GFP_KERNEL); | |
993 | if (!net->ipv6.mfc6_cache_array) { | |
994 | err = -ENOMEM; | |
995 | goto fail_mfc6_cache; | |
996 | } | |
950d5704 BT |
997 | |
998 | #ifdef CONFIG_IPV6_PIMSM_V2 | |
999 | net->ipv6.mroute_reg_vif_num = -1; | |
1000 | #endif | |
8b90fc7e BT |
1001 | |
1002 | #ifdef CONFIG_PROC_FS | |
1003 | err = -ENOMEM; | |
1004 | if (!proc_net_fops_create(net, "ip6_mr_vif", 0, &ip6mr_vif_fops)) | |
1005 | goto proc_vif_fail; | |
1006 | if (!proc_net_fops_create(net, "ip6_mr_cache", 0, &ip6mr_mfc_fops)) | |
1007 | goto proc_cache_fail; | |
1008 | #endif | |
4a6258a0 BT |
1009 | return 0; |
1010 | ||
8b90fc7e BT |
1011 | #ifdef CONFIG_PROC_FS |
1012 | proc_cache_fail: | |
1013 | proc_net_remove(net, "ip6_mr_vif"); | |
1014 | proc_vif_fail: | |
1015 | kfree(net->ipv6.mfc6_cache_array); | |
1016 | #endif | |
4a6258a0 BT |
1017 | fail_mfc6_cache: |
1018 | kfree(net->ipv6.vif6_table); | |
4e16880c BT |
1019 | fail: |
1020 | return err; | |
1021 | } | |
1022 | ||
1023 | static void __net_exit ip6mr_net_exit(struct net *net) | |
1024 | { | |
8b90fc7e BT |
1025 | #ifdef CONFIG_PROC_FS |
1026 | proc_net_remove(net, "ip6_mr_cache"); | |
1027 | proc_net_remove(net, "ip6_mr_vif"); | |
1028 | #endif | |
4a6258a0 | 1029 | kfree(net->ipv6.mfc6_cache_array); |
4e16880c BT |
1030 | kfree(net->ipv6.vif6_table); |
1031 | } | |
1032 | ||
1033 | static struct pernet_operations ip6mr_net_ops = { | |
1034 | .init = ip6mr_net_init, | |
1035 | .exit = ip6mr_net_exit, | |
1036 | }; | |
1037 | ||
623d1a1a | 1038 | int __init ip6_mr_init(void) |
7bc570c8 | 1039 | { |
623d1a1a WC |
1040 | int err; |
1041 | ||
7bc570c8 YH |
1042 | mrt_cachep = kmem_cache_create("ip6_mrt_cache", |
1043 | sizeof(struct mfc6_cache), | |
1044 | 0, SLAB_HWCACHE_ALIGN, | |
1045 | NULL); | |
1046 | if (!mrt_cachep) | |
623d1a1a | 1047 | return -ENOMEM; |
7bc570c8 | 1048 | |
4e16880c BT |
1049 | err = register_pernet_subsys(&ip6mr_net_ops); |
1050 | if (err) | |
1051 | goto reg_pernet_fail; | |
1052 | ||
7bc570c8 | 1053 | setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0); |
623d1a1a WC |
1054 | err = register_netdevice_notifier(&ip6_mr_notifier); |
1055 | if (err) | |
1056 | goto reg_notif_fail; | |
623d1a1a | 1057 | return 0; |
87b30a65 BT |
1058 | reg_notif_fail: |
1059 | del_timer(&ipmr_expire_timer); | |
4e16880c BT |
1060 | unregister_pernet_subsys(&ip6mr_net_ops); |
1061 | reg_pernet_fail: | |
87b30a65 | 1062 | kmem_cache_destroy(mrt_cachep); |
623d1a1a | 1063 | return err; |
7bc570c8 YH |
1064 | } |
1065 | ||
623d1a1a WC |
1066 | void ip6_mr_cleanup(void) |
1067 | { | |
623d1a1a WC |
1068 | unregister_netdevice_notifier(&ip6_mr_notifier); |
1069 | del_timer(&ipmr_expire_timer); | |
4e16880c | 1070 | unregister_pernet_subsys(&ip6mr_net_ops); |
623d1a1a WC |
1071 | kmem_cache_destroy(mrt_cachep); |
1072 | } | |
7bc570c8 YH |
1073 | |
1074 | static int ip6mr_mfc_add(struct mf6cctl *mfc, int mrtsock) | |
1075 | { | |
1076 | int line; | |
1077 | struct mfc6_cache *uc, *c, **cp; | |
6ac7eb08 | 1078 | unsigned char ttls[MAXMIFS]; |
7bc570c8 YH |
1079 | int i; |
1080 | ||
6ac7eb08 RR |
1081 | memset(ttls, 255, MAXMIFS); |
1082 | for (i = 0; i < MAXMIFS; i++) { | |
7bc570c8 YH |
1083 | if (IF_ISSET(i, &mfc->mf6cc_ifset)) |
1084 | ttls[i] = 1; | |
1085 | ||
1086 | } | |
1087 | ||
1088 | line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); | |
1089 | ||
4a6258a0 BT |
1090 | for (cp = &init_net.ipv6.mfc6_cache_array[line]; |
1091 | (c = *cp) != NULL; cp = &c->next) { | |
7bc570c8 YH |
1092 | if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && |
1093 | ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) | |
1094 | break; | |
1095 | } | |
1096 | ||
1097 | if (c != NULL) { | |
1098 | write_lock_bh(&mrt_lock); | |
1099 | c->mf6c_parent = mfc->mf6cc_parent; | |
1100 | ip6mr_update_thresholds(c, ttls); | |
1101 | if (!mrtsock) | |
1102 | c->mfc_flags |= MFC_STATIC; | |
1103 | write_unlock_bh(&mrt_lock); | |
1104 | return 0; | |
1105 | } | |
1106 | ||
1107 | if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) | |
1108 | return -EINVAL; | |
1109 | ||
58701ad4 | 1110 | c = ip6mr_cache_alloc(&init_net); |
7bc570c8 YH |
1111 | if (c == NULL) |
1112 | return -ENOMEM; | |
1113 | ||
1114 | c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; | |
1115 | c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; | |
1116 | c->mf6c_parent = mfc->mf6cc_parent; | |
1117 | ip6mr_update_thresholds(c, ttls); | |
1118 | if (!mrtsock) | |
1119 | c->mfc_flags |= MFC_STATIC; | |
1120 | ||
1121 | write_lock_bh(&mrt_lock); | |
4a6258a0 BT |
1122 | c->next = init_net.ipv6.mfc6_cache_array[line]; |
1123 | init_net.ipv6.mfc6_cache_array[line] = c; | |
7bc570c8 YH |
1124 | write_unlock_bh(&mrt_lock); |
1125 | ||
1126 | /* | |
1127 | * Check to see if we resolved a queued list. If so we | |
1128 | * need to send on the frames and tidy up. | |
1129 | */ | |
1130 | spin_lock_bh(&mfc_unres_lock); | |
1131 | for (cp = &mfc_unres_queue; (uc = *cp) != NULL; | |
1132 | cp = &uc->next) { | |
4045e57c BT |
1133 | if (net_eq(mfc6_net(uc), &init_net) && |
1134 | ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && | |
7bc570c8 YH |
1135 | ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { |
1136 | *cp = uc->next; | |
4045e57c | 1137 | atomic_dec(&init_net.ipv6.cache_resolve_queue_len); |
7bc570c8 YH |
1138 | break; |
1139 | } | |
1140 | } | |
4045e57c BT |
1141 | if (mfc_unres_queue == NULL) |
1142 | del_timer(&ipmr_expire_timer); | |
7bc570c8 YH |
1143 | spin_unlock_bh(&mfc_unres_lock); |
1144 | ||
1145 | if (uc) { | |
1146 | ip6mr_cache_resolve(uc, c); | |
58701ad4 | 1147 | ip6mr_cache_free(uc); |
7bc570c8 YH |
1148 | } |
1149 | return 0; | |
1150 | } | |
1151 | ||
1152 | /* | |
1153 | * Close the multicast socket, and clear the vif tables etc | |
1154 | */ | |
1155 | ||
1156 | static void mroute_clean_tables(struct sock *sk) | |
1157 | { | |
1158 | int i; | |
1159 | ||
1160 | /* | |
1161 | * Shut down all active vif entries | |
1162 | */ | |
4e16880c BT |
1163 | for (i = 0; i < init_net.ipv6.maxvif; i++) { |
1164 | if (!(init_net.ipv6.vif6_table[i].flags & VIFF_STATIC)) | |
7bc570c8 YH |
1165 | mif6_delete(i); |
1166 | } | |
1167 | ||
1168 | /* | |
1169 | * Wipe the cache | |
1170 | */ | |
4a6258a0 | 1171 | for (i = 0; i < MFC6_LINES; i++) { |
7bc570c8 YH |
1172 | struct mfc6_cache *c, **cp; |
1173 | ||
4a6258a0 | 1174 | cp = &init_net.ipv6.mfc6_cache_array[i]; |
7bc570c8 YH |
1175 | while ((c = *cp) != NULL) { |
1176 | if (c->mfc_flags & MFC_STATIC) { | |
1177 | cp = &c->next; | |
1178 | continue; | |
1179 | } | |
1180 | write_lock_bh(&mrt_lock); | |
1181 | *cp = c->next; | |
1182 | write_unlock_bh(&mrt_lock); | |
1183 | ||
58701ad4 | 1184 | ip6mr_cache_free(c); |
7bc570c8 YH |
1185 | } |
1186 | } | |
1187 | ||
4045e57c BT |
1188 | if (atomic_read(&init_net.ipv6.cache_resolve_queue_len) != 0) { |
1189 | struct mfc6_cache *c, **cp; | |
7bc570c8 YH |
1190 | |
1191 | spin_lock_bh(&mfc_unres_lock); | |
4045e57c BT |
1192 | cp = &mfc_unres_queue; |
1193 | while ((c = *cp) != NULL) { | |
1194 | if (!net_eq(mfc6_net(c), &init_net)) { | |
1195 | cp = &c->next; | |
1196 | continue; | |
1197 | } | |
1198 | *cp = c->next; | |
7bc570c8 | 1199 | ip6mr_destroy_unres(c); |
7bc570c8 YH |
1200 | } |
1201 | spin_unlock_bh(&mfc_unres_lock); | |
1202 | } | |
1203 | } | |
1204 | ||
1205 | static int ip6mr_sk_init(struct sock *sk) | |
1206 | { | |
1207 | int err = 0; | |
1208 | ||
1209 | rtnl_lock(); | |
1210 | write_lock_bh(&mrt_lock); | |
bd91b8bf BT |
1211 | if (likely(init_net.ipv6.mroute6_sk == NULL)) |
1212 | init_net.ipv6.mroute6_sk = sk; | |
7bc570c8 YH |
1213 | else |
1214 | err = -EADDRINUSE; | |
1215 | write_unlock_bh(&mrt_lock); | |
1216 | ||
1217 | rtnl_unlock(); | |
1218 | ||
1219 | return err; | |
1220 | } | |
1221 | ||
1222 | int ip6mr_sk_done(struct sock *sk) | |
1223 | { | |
1224 | int err = 0; | |
1225 | ||
1226 | rtnl_lock(); | |
bd91b8bf | 1227 | if (sk == init_net.ipv6.mroute6_sk) { |
7bc570c8 | 1228 | write_lock_bh(&mrt_lock); |
bd91b8bf | 1229 | init_net.ipv6.mroute6_sk = NULL; |
7bc570c8 YH |
1230 | write_unlock_bh(&mrt_lock); |
1231 | ||
1232 | mroute_clean_tables(sk); | |
1233 | } else | |
1234 | err = -EACCES; | |
1235 | rtnl_unlock(); | |
1236 | ||
1237 | return err; | |
1238 | } | |
1239 | ||
1240 | /* | |
1241 | * Socket options and virtual interface manipulation. The whole | |
1242 | * virtual interface system is a complete heap, but unfortunately | |
1243 | * that's how BSD mrouted happens to think. Maybe one day with a proper | |
1244 | * MOSPF/PIM router set up we can clean this up. | |
1245 | */ | |
1246 | ||
1247 | int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, int optlen) | |
1248 | { | |
1249 | int ret; | |
1250 | struct mif6ctl vif; | |
1251 | struct mf6cctl mfc; | |
1252 | mifi_t mifi; | |
1253 | ||
1254 | if (optname != MRT6_INIT) { | |
bd91b8bf | 1255 | if (sk != init_net.ipv6.mroute6_sk && !capable(CAP_NET_ADMIN)) |
7bc570c8 YH |
1256 | return -EACCES; |
1257 | } | |
1258 | ||
1259 | switch (optname) { | |
1260 | case MRT6_INIT: | |
1261 | if (sk->sk_type != SOCK_RAW || | |
1262 | inet_sk(sk)->num != IPPROTO_ICMPV6) | |
1263 | return -EOPNOTSUPP; | |
1264 | if (optlen < sizeof(int)) | |
1265 | return -EINVAL; | |
1266 | ||
1267 | return ip6mr_sk_init(sk); | |
1268 | ||
1269 | case MRT6_DONE: | |
1270 | return ip6mr_sk_done(sk); | |
1271 | ||
1272 | case MRT6_ADD_MIF: | |
1273 | if (optlen < sizeof(vif)) | |
1274 | return -EINVAL; | |
1275 | if (copy_from_user(&vif, optval, sizeof(vif))) | |
1276 | return -EFAULT; | |
6ac7eb08 | 1277 | if (vif.mif6c_mifi >= MAXMIFS) |
7bc570c8 YH |
1278 | return -ENFILE; |
1279 | rtnl_lock(); | |
bd91b8bf | 1280 | ret = mif6_add(&vif, sk == init_net.ipv6.mroute6_sk); |
7bc570c8 YH |
1281 | rtnl_unlock(); |
1282 | return ret; | |
1283 | ||
1284 | case MRT6_DEL_MIF: | |
1285 | if (optlen < sizeof(mifi_t)) | |
1286 | return -EINVAL; | |
1287 | if (copy_from_user(&mifi, optval, sizeof(mifi_t))) | |
1288 | return -EFAULT; | |
1289 | rtnl_lock(); | |
1290 | ret = mif6_delete(mifi); | |
1291 | rtnl_unlock(); | |
1292 | return ret; | |
1293 | ||
1294 | /* | |
1295 | * Manipulate the forwarding caches. These live | |
1296 | * in a sort of kernel/user symbiosis. | |
1297 | */ | |
1298 | case MRT6_ADD_MFC: | |
1299 | case MRT6_DEL_MFC: | |
1300 | if (optlen < sizeof(mfc)) | |
1301 | return -EINVAL; | |
1302 | if (copy_from_user(&mfc, optval, sizeof(mfc))) | |
1303 | return -EFAULT; | |
1304 | rtnl_lock(); | |
1305 | if (optname == MRT6_DEL_MFC) | |
1306 | ret = ip6mr_mfc_delete(&mfc); | |
1307 | else | |
bd91b8bf | 1308 | ret = ip6mr_mfc_add(&mfc, sk == init_net.ipv6.mroute6_sk); |
7bc570c8 YH |
1309 | rtnl_unlock(); |
1310 | return ret; | |
1311 | ||
14fb64e1 YH |
1312 | /* |
1313 | * Control PIM assert (to activate pim will activate assert) | |
1314 | */ | |
1315 | case MRT6_ASSERT: | |
1316 | { | |
1317 | int v; | |
1318 | if (get_user(v, (int __user *)optval)) | |
1319 | return -EFAULT; | |
a21f3f99 | 1320 | init_net.ipv6.mroute_do_assert = !!v; |
14fb64e1 YH |
1321 | return 0; |
1322 | } | |
1323 | ||
1324 | #ifdef CONFIG_IPV6_PIMSM_V2 | |
1325 | case MRT6_PIM: | |
1326 | { | |
a9f83bf3 | 1327 | int v; |
14fb64e1 YH |
1328 | if (get_user(v, (int __user *)optval)) |
1329 | return -EFAULT; | |
1330 | v = !!v; | |
1331 | rtnl_lock(); | |
1332 | ret = 0; | |
a21f3f99 BT |
1333 | if (v != init_net.ipv6.mroute_do_pim) { |
1334 | init_net.ipv6.mroute_do_pim = v; | |
1335 | init_net.ipv6.mroute_do_assert = v; | |
1336 | if (init_net.ipv6.mroute_do_pim) | |
14fb64e1 YH |
1337 | ret = inet6_add_protocol(&pim6_protocol, |
1338 | IPPROTO_PIM); | |
1339 | else | |
1340 | ret = inet6_del_protocol(&pim6_protocol, | |
1341 | IPPROTO_PIM); | |
1342 | if (ret < 0) | |
1343 | ret = -EAGAIN; | |
1344 | } | |
1345 | rtnl_unlock(); | |
1346 | return ret; | |
1347 | } | |
1348 | ||
1349 | #endif | |
7bc570c8 | 1350 | /* |
7d120c55 | 1351 | * Spurious command, or MRT6_VERSION which you cannot |
7bc570c8 YH |
1352 | * set. |
1353 | */ | |
1354 | default: | |
1355 | return -ENOPROTOOPT; | |
1356 | } | |
1357 | } | |
1358 | ||
1359 | /* | |
1360 | * Getsock opt support for the multicast routing system. | |
1361 | */ | |
1362 | ||
1363 | int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, | |
1364 | int __user *optlen) | |
1365 | { | |
1366 | int olr; | |
1367 | int val; | |
1368 | ||
1369 | switch (optname) { | |
1370 | case MRT6_VERSION: | |
1371 | val = 0x0305; | |
1372 | break; | |
14fb64e1 YH |
1373 | #ifdef CONFIG_IPV6_PIMSM_V2 |
1374 | case MRT6_PIM: | |
a21f3f99 | 1375 | val = init_net.ipv6.mroute_do_pim; |
14fb64e1 YH |
1376 | break; |
1377 | #endif | |
1378 | case MRT6_ASSERT: | |
a21f3f99 | 1379 | val = init_net.ipv6.mroute_do_assert; |
14fb64e1 | 1380 | break; |
7bc570c8 YH |
1381 | default: |
1382 | return -ENOPROTOOPT; | |
1383 | } | |
1384 | ||
1385 | if (get_user(olr, optlen)) | |
1386 | return -EFAULT; | |
1387 | ||
1388 | olr = min_t(int, olr, sizeof(int)); | |
1389 | if (olr < 0) | |
1390 | return -EINVAL; | |
1391 | ||
1392 | if (put_user(olr, optlen)) | |
1393 | return -EFAULT; | |
1394 | if (copy_to_user(optval, &val, olr)) | |
1395 | return -EFAULT; | |
1396 | return 0; | |
1397 | } | |
1398 | ||
1399 | /* | |
1400 | * The IP multicast ioctl support routines. | |
1401 | */ | |
1402 | ||
1403 | int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) | |
1404 | { | |
1405 | struct sioc_sg_req6 sr; | |
1406 | struct sioc_mif_req6 vr; | |
1407 | struct mif_device *vif; | |
1408 | struct mfc6_cache *c; | |
1409 | ||
1410 | switch (cmd) { | |
1411 | case SIOCGETMIFCNT_IN6: | |
1412 | if (copy_from_user(&vr, arg, sizeof(vr))) | |
1413 | return -EFAULT; | |
4e16880c | 1414 | if (vr.mifi >= init_net.ipv6.maxvif) |
7bc570c8 YH |
1415 | return -EINVAL; |
1416 | read_lock(&mrt_lock); | |
4e16880c BT |
1417 | vif = &init_net.ipv6.vif6_table[vr.mifi]; |
1418 | if (MIF_EXISTS(&init_net, vr.mifi)) { | |
7bc570c8 YH |
1419 | vr.icount = vif->pkt_in; |
1420 | vr.ocount = vif->pkt_out; | |
1421 | vr.ibytes = vif->bytes_in; | |
1422 | vr.obytes = vif->bytes_out; | |
1423 | read_unlock(&mrt_lock); | |
1424 | ||
1425 | if (copy_to_user(arg, &vr, sizeof(vr))) | |
1426 | return -EFAULT; | |
1427 | return 0; | |
1428 | } | |
1429 | read_unlock(&mrt_lock); | |
1430 | return -EADDRNOTAVAIL; | |
1431 | case SIOCGETSGCNT_IN6: | |
1432 | if (copy_from_user(&sr, arg, sizeof(sr))) | |
1433 | return -EFAULT; | |
1434 | ||
1435 | read_lock(&mrt_lock); | |
1436 | c = ip6mr_cache_find(&sr.src.sin6_addr, &sr.grp.sin6_addr); | |
1437 | if (c) { | |
1438 | sr.pktcnt = c->mfc_un.res.pkt; | |
1439 | sr.bytecnt = c->mfc_un.res.bytes; | |
1440 | sr.wrong_if = c->mfc_un.res.wrong_if; | |
1441 | read_unlock(&mrt_lock); | |
1442 | ||
1443 | if (copy_to_user(arg, &sr, sizeof(sr))) | |
1444 | return -EFAULT; | |
1445 | return 0; | |
1446 | } | |
1447 | read_unlock(&mrt_lock); | |
1448 | return -EADDRNOTAVAIL; | |
1449 | default: | |
1450 | return -ENOIOCTLCMD; | |
1451 | } | |
1452 | } | |
1453 | ||
1454 | ||
1455 | static inline int ip6mr_forward2_finish(struct sk_buff *skb) | |
1456 | { | |
483a47d2 DL |
1457 | IP6_INC_STATS_BH(dev_net(skb->dst->dev), ip6_dst_idev(skb->dst), |
1458 | IPSTATS_MIB_OUTFORWDATAGRAMS); | |
7bc570c8 YH |
1459 | return dst_output(skb); |
1460 | } | |
1461 | ||
1462 | /* | |
1463 | * Processing handlers for ip6mr_forward | |
1464 | */ | |
1465 | ||
1466 | static int ip6mr_forward2(struct sk_buff *skb, struct mfc6_cache *c, int vifi) | |
1467 | { | |
1468 | struct ipv6hdr *ipv6h; | |
4e16880c | 1469 | struct mif_device *vif = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 YH |
1470 | struct net_device *dev; |
1471 | struct dst_entry *dst; | |
1472 | struct flowi fl; | |
1473 | ||
1474 | if (vif->dev == NULL) | |
1475 | goto out_free; | |
1476 | ||
14fb64e1 YH |
1477 | #ifdef CONFIG_IPV6_PIMSM_V2 |
1478 | if (vif->flags & MIFF_REGISTER) { | |
1479 | vif->pkt_out++; | |
1480 | vif->bytes_out += skb->len; | |
dc58c78c PE |
1481 | vif->dev->stats.tx_bytes += skb->len; |
1482 | vif->dev->stats.tx_packets++; | |
14fb64e1 YH |
1483 | ip6mr_cache_report(skb, vifi, MRT6MSG_WHOLEPKT); |
1484 | kfree_skb(skb); | |
1485 | return 0; | |
1486 | } | |
1487 | #endif | |
1488 | ||
7bc570c8 YH |
1489 | ipv6h = ipv6_hdr(skb); |
1490 | ||
1491 | fl = (struct flowi) { | |
1492 | .oif = vif->link, | |
1493 | .nl_u = { .ip6_u = | |
1494 | { .daddr = ipv6h->daddr, } | |
1495 | } | |
1496 | }; | |
1497 | ||
1498 | dst = ip6_route_output(&init_net, NULL, &fl); | |
1499 | if (!dst) | |
1500 | goto out_free; | |
1501 | ||
1502 | dst_release(skb->dst); | |
1503 | skb->dst = dst; | |
1504 | ||
1505 | /* | |
1506 | * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally | |
1507 | * not only before forwarding, but after forwarding on all output | |
1508 | * interfaces. It is clear, if mrouter runs a multicasting | |
1509 | * program, it should receive packets not depending to what interface | |
1510 | * program is joined. | |
1511 | * If we will not make it, the program will have to join on all | |
1512 | * interfaces. On the other hand, multihoming host (or router, but | |
1513 | * not mrouter) cannot join to more than one interface - it will | |
1514 | * result in receiving multiple packets. | |
1515 | */ | |
1516 | dev = vif->dev; | |
1517 | skb->dev = dev; | |
1518 | vif->pkt_out++; | |
1519 | vif->bytes_out += skb->len; | |
1520 | ||
1521 | /* We are about to write */ | |
1522 | /* XXX: extension headers? */ | |
1523 | if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) | |
1524 | goto out_free; | |
1525 | ||
1526 | ipv6h = ipv6_hdr(skb); | |
1527 | ipv6h->hop_limit--; | |
1528 | ||
1529 | IP6CB(skb)->flags |= IP6SKB_FORWARDED; | |
1530 | ||
1531 | return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dev, | |
1532 | ip6mr_forward2_finish); | |
1533 | ||
1534 | out_free: | |
1535 | kfree_skb(skb); | |
1536 | return 0; | |
1537 | } | |
1538 | ||
1539 | static int ip6mr_find_vif(struct net_device *dev) | |
1540 | { | |
1541 | int ct; | |
4e16880c BT |
1542 | for (ct = init_net.ipv6.maxvif - 1; ct >= 0; ct--) { |
1543 | if (init_net.ipv6.vif6_table[ct].dev == dev) | |
7bc570c8 YH |
1544 | break; |
1545 | } | |
1546 | return ct; | |
1547 | } | |
1548 | ||
1549 | static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache) | |
1550 | { | |
1551 | int psend = -1; | |
1552 | int vif, ct; | |
1553 | ||
1554 | vif = cache->mf6c_parent; | |
1555 | cache->mfc_un.res.pkt++; | |
1556 | cache->mfc_un.res.bytes += skb->len; | |
1557 | ||
14fb64e1 YH |
1558 | /* |
1559 | * Wrong interface: drop packet and (maybe) send PIM assert. | |
1560 | */ | |
4e16880c | 1561 | if (init_net.ipv6.vif6_table[vif].dev != skb->dev) { |
14fb64e1 YH |
1562 | int true_vifi; |
1563 | ||
1564 | cache->mfc_un.res.wrong_if++; | |
1565 | true_vifi = ip6mr_find_vif(skb->dev); | |
1566 | ||
a21f3f99 | 1567 | if (true_vifi >= 0 && init_net.ipv6.mroute_do_assert && |
14fb64e1 YH |
1568 | /* pimsm uses asserts, when switching from RPT to SPT, |
1569 | so that we cannot check that packet arrived on an oif. | |
1570 | It is bad, but otherwise we would need to move pretty | |
1571 | large chunk of pimd to kernel. Ough... --ANK | |
1572 | */ | |
a21f3f99 BT |
1573 | (init_net.ipv6.mroute_do_pim || |
1574 | cache->mfc_un.res.ttls[true_vifi] < 255) && | |
14fb64e1 YH |
1575 | time_after(jiffies, |
1576 | cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { | |
1577 | cache->mfc_un.res.last_assert = jiffies; | |
1578 | ip6mr_cache_report(skb, true_vifi, MRT6MSG_WRONGMIF); | |
1579 | } | |
1580 | goto dont_forward; | |
1581 | } | |
1582 | ||
4e16880c BT |
1583 | init_net.ipv6.vif6_table[vif].pkt_in++; |
1584 | init_net.ipv6.vif6_table[vif].bytes_in += skb->len; | |
7bc570c8 YH |
1585 | |
1586 | /* | |
1587 | * Forward the frame | |
1588 | */ | |
1589 | for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { | |
1590 | if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { | |
1591 | if (psend != -1) { | |
1592 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1593 | if (skb2) | |
1594 | ip6mr_forward2(skb2, cache, psend); | |
1595 | } | |
1596 | psend = ct; | |
1597 | } | |
1598 | } | |
1599 | if (psend != -1) { | |
1600 | ip6mr_forward2(skb, cache, psend); | |
1601 | return 0; | |
1602 | } | |
1603 | ||
14fb64e1 | 1604 | dont_forward: |
7bc570c8 YH |
1605 | kfree_skb(skb); |
1606 | return 0; | |
1607 | } | |
1608 | ||
1609 | ||
1610 | /* | |
1611 | * Multicast packets for forwarding arrive here | |
1612 | */ | |
1613 | ||
1614 | int ip6_mr_input(struct sk_buff *skb) | |
1615 | { | |
1616 | struct mfc6_cache *cache; | |
1617 | ||
1618 | read_lock(&mrt_lock); | |
1619 | cache = ip6mr_cache_find(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); | |
1620 | ||
1621 | /* | |
1622 | * No usable cache entry | |
1623 | */ | |
1624 | if (cache == NULL) { | |
1625 | int vif; | |
1626 | ||
1627 | vif = ip6mr_find_vif(skb->dev); | |
1628 | if (vif >= 0) { | |
1629 | int err = ip6mr_cache_unresolved(vif, skb); | |
1630 | read_unlock(&mrt_lock); | |
1631 | ||
1632 | return err; | |
1633 | } | |
1634 | read_unlock(&mrt_lock); | |
1635 | kfree_skb(skb); | |
1636 | return -ENODEV; | |
1637 | } | |
1638 | ||
1639 | ip6_mr_forward(skb, cache); | |
1640 | ||
1641 | read_unlock(&mrt_lock); | |
1642 | ||
1643 | return 0; | |
1644 | } | |
1645 | ||
1646 | ||
1647 | static int | |
1648 | ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm) | |
1649 | { | |
1650 | int ct; | |
1651 | struct rtnexthop *nhp; | |
4e16880c | 1652 | struct net_device *dev = init_net.ipv6.vif6_table[c->mf6c_parent].dev; |
549e028d | 1653 | u8 *b = skb_tail_pointer(skb); |
7bc570c8 YH |
1654 | struct rtattr *mp_head; |
1655 | ||
1656 | if (dev) | |
1657 | RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex); | |
1658 | ||
1659 | mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0)); | |
1660 | ||
1661 | for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { | |
1662 | if (c->mfc_un.res.ttls[ct] < 255) { | |
1663 | if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4)) | |
1664 | goto rtattr_failure; | |
1665 | nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp))); | |
1666 | nhp->rtnh_flags = 0; | |
1667 | nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; | |
4e16880c | 1668 | nhp->rtnh_ifindex = init_net.ipv6.vif6_table[ct].dev->ifindex; |
7bc570c8 YH |
1669 | nhp->rtnh_len = sizeof(*nhp); |
1670 | } | |
1671 | } | |
1672 | mp_head->rta_type = RTA_MULTIPATH; | |
549e028d | 1673 | mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head; |
7bc570c8 YH |
1674 | rtm->rtm_type = RTN_MULTICAST; |
1675 | return 1; | |
1676 | ||
1677 | rtattr_failure: | |
1678 | nlmsg_trim(skb, b); | |
1679 | return -EMSGSIZE; | |
1680 | } | |
1681 | ||
1682 | int ip6mr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait) | |
1683 | { | |
1684 | int err; | |
1685 | struct mfc6_cache *cache; | |
1686 | struct rt6_info *rt = (struct rt6_info *)skb->dst; | |
1687 | ||
1688 | read_lock(&mrt_lock); | |
1689 | cache = ip6mr_cache_find(&rt->rt6i_src.addr, &rt->rt6i_dst.addr); | |
1690 | ||
1691 | if (!cache) { | |
1692 | struct sk_buff *skb2; | |
1693 | struct ipv6hdr *iph; | |
1694 | struct net_device *dev; | |
1695 | int vif; | |
1696 | ||
1697 | if (nowait) { | |
1698 | read_unlock(&mrt_lock); | |
1699 | return -EAGAIN; | |
1700 | } | |
1701 | ||
1702 | dev = skb->dev; | |
1703 | if (dev == NULL || (vif = ip6mr_find_vif(dev)) < 0) { | |
1704 | read_unlock(&mrt_lock); | |
1705 | return -ENODEV; | |
1706 | } | |
1707 | ||
1708 | /* really correct? */ | |
1709 | skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); | |
1710 | if (!skb2) { | |
1711 | read_unlock(&mrt_lock); | |
1712 | return -ENOMEM; | |
1713 | } | |
1714 | ||
1715 | skb_reset_transport_header(skb2); | |
1716 | ||
1717 | skb_put(skb2, sizeof(struct ipv6hdr)); | |
1718 | skb_reset_network_header(skb2); | |
1719 | ||
1720 | iph = ipv6_hdr(skb2); | |
1721 | iph->version = 0; | |
1722 | iph->priority = 0; | |
1723 | iph->flow_lbl[0] = 0; | |
1724 | iph->flow_lbl[1] = 0; | |
1725 | iph->flow_lbl[2] = 0; | |
1726 | iph->payload_len = 0; | |
1727 | iph->nexthdr = IPPROTO_NONE; | |
1728 | iph->hop_limit = 0; | |
1729 | ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr); | |
1730 | ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr); | |
1731 | ||
1732 | err = ip6mr_cache_unresolved(vif, skb2); | |
1733 | read_unlock(&mrt_lock); | |
1734 | ||
1735 | return err; | |
1736 | } | |
1737 | ||
1738 | if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY)) | |
1739 | cache->mfc_flags |= MFC_NOTIFY; | |
1740 | ||
1741 | err = ip6mr_fill_mroute(skb, cache, rtm); | |
1742 | read_unlock(&mrt_lock); | |
1743 | return err; | |
1744 | } | |
1745 |