]>
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 { | |
93 | struct mfc6_cache **cache; | |
94 | int ct; | |
95 | }; | |
96 | ||
97 | ||
98 | static struct mfc6_cache *ipmr_mfc_seq_idx(struct ipmr_mfc_iter *it, loff_t pos) | |
99 | { | |
100 | struct mfc6_cache *mfc; | |
101 | ||
4a6258a0 | 102 | it->cache = init_net.ipv6.mfc6_cache_array; |
7bc570c8 | 103 | read_lock(&mrt_lock); |
4a6258a0 BT |
104 | for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) |
105 | for (mfc = init_net.ipv6.mfc6_cache_array[it->ct]; | |
106 | mfc; mfc = mfc->next) | |
7bc570c8 YH |
107 | if (pos-- == 0) |
108 | return mfc; | |
109 | read_unlock(&mrt_lock); | |
110 | ||
111 | it->cache = &mfc_unres_queue; | |
112 | spin_lock_bh(&mfc_unres_lock); | |
113 | for (mfc = mfc_unres_queue; mfc; mfc = mfc->next) | |
114 | if (pos-- == 0) | |
115 | return mfc; | |
116 | spin_unlock_bh(&mfc_unres_lock); | |
117 | ||
118 | it->cache = NULL; | |
119 | return NULL; | |
120 | } | |
121 | ||
122 | ||
123 | ||
124 | ||
125 | /* | |
126 | * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif | |
127 | */ | |
128 | ||
129 | struct ipmr_vif_iter { | |
130 | int ct; | |
131 | }; | |
132 | ||
133 | static struct mif_device *ip6mr_vif_seq_idx(struct ipmr_vif_iter *iter, | |
134 | loff_t pos) | |
135 | { | |
4e16880c BT |
136 | for (iter->ct = 0; iter->ct < init_net.ipv6.maxvif; ++iter->ct) { |
137 | if (!MIF_EXISTS(&init_net, iter->ct)) | |
7bc570c8 YH |
138 | continue; |
139 | if (pos-- == 0) | |
4e16880c | 140 | return &init_net.ipv6.vif6_table[iter->ct]; |
7bc570c8 YH |
141 | } |
142 | return NULL; | |
143 | } | |
144 | ||
145 | static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos) | |
146 | __acquires(mrt_lock) | |
147 | { | |
148 | read_lock(&mrt_lock); | |
149 | return (*pos ? ip6mr_vif_seq_idx(seq->private, *pos - 1) | |
150 | : SEQ_START_TOKEN); | |
151 | } | |
152 | ||
153 | static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
154 | { | |
155 | struct ipmr_vif_iter *iter = seq->private; | |
156 | ||
157 | ++*pos; | |
158 | if (v == SEQ_START_TOKEN) | |
159 | return ip6mr_vif_seq_idx(iter, 0); | |
160 | ||
4e16880c BT |
161 | while (++iter->ct < init_net.ipv6.maxvif) { |
162 | if (!MIF_EXISTS(&init_net, iter->ct)) | |
7bc570c8 | 163 | continue; |
4e16880c | 164 | return &init_net.ipv6.vif6_table[iter->ct]; |
7bc570c8 YH |
165 | } |
166 | return NULL; | |
167 | } | |
168 | ||
169 | static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v) | |
170 | __releases(mrt_lock) | |
171 | { | |
172 | read_unlock(&mrt_lock); | |
173 | } | |
174 | ||
175 | static int ip6mr_vif_seq_show(struct seq_file *seq, void *v) | |
176 | { | |
177 | if (v == SEQ_START_TOKEN) { | |
178 | seq_puts(seq, | |
179 | "Interface BytesIn PktsIn BytesOut PktsOut Flags\n"); | |
180 | } else { | |
181 | const struct mif_device *vif = v; | |
182 | const char *name = vif->dev ? vif->dev->name : "none"; | |
183 | ||
184 | seq_printf(seq, | |
d430a227 | 185 | "%2td %-10s %8ld %7ld %8ld %7ld %05X\n", |
4e16880c | 186 | vif - init_net.ipv6.vif6_table, |
7bc570c8 YH |
187 | name, vif->bytes_in, vif->pkt_in, |
188 | vif->bytes_out, vif->pkt_out, | |
189 | vif->flags); | |
190 | } | |
191 | return 0; | |
192 | } | |
193 | ||
194 | static struct seq_operations ip6mr_vif_seq_ops = { | |
195 | .start = ip6mr_vif_seq_start, | |
196 | .next = ip6mr_vif_seq_next, | |
197 | .stop = ip6mr_vif_seq_stop, | |
198 | .show = ip6mr_vif_seq_show, | |
199 | }; | |
200 | ||
201 | static int ip6mr_vif_open(struct inode *inode, struct file *file) | |
202 | { | |
203 | return seq_open_private(file, &ip6mr_vif_seq_ops, | |
204 | sizeof(struct ipmr_vif_iter)); | |
205 | } | |
206 | ||
207 | static struct file_operations ip6mr_vif_fops = { | |
208 | .owner = THIS_MODULE, | |
209 | .open = ip6mr_vif_open, | |
210 | .read = seq_read, | |
211 | .llseek = seq_lseek, | |
eedd726e | 212 | .release = seq_release_private, |
7bc570c8 YH |
213 | }; |
214 | ||
215 | static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) | |
216 | { | |
217 | return (*pos ? ipmr_mfc_seq_idx(seq->private, *pos - 1) | |
218 | : SEQ_START_TOKEN); | |
219 | } | |
220 | ||
221 | static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
222 | { | |
223 | struct mfc6_cache *mfc = v; | |
224 | struct ipmr_mfc_iter *it = seq->private; | |
225 | ||
226 | ++*pos; | |
227 | ||
228 | if (v == SEQ_START_TOKEN) | |
229 | return ipmr_mfc_seq_idx(seq->private, 0); | |
230 | ||
231 | if (mfc->next) | |
232 | return mfc->next; | |
233 | ||
234 | if (it->cache == &mfc_unres_queue) | |
235 | goto end_of_list; | |
236 | ||
4a6258a0 | 237 | BUG_ON(it->cache != init_net.ipv6.mfc6_cache_array); |
7bc570c8 | 238 | |
4a6258a0 BT |
239 | while (++it->ct < MFC6_LINES) { |
240 | mfc = init_net.ipv6.mfc6_cache_array[it->ct]; | |
7bc570c8 YH |
241 | if (mfc) |
242 | return mfc; | |
243 | } | |
244 | ||
245 | /* exhausted cache_array, show unresolved */ | |
246 | read_unlock(&mrt_lock); | |
247 | it->cache = &mfc_unres_queue; | |
248 | it->ct = 0; | |
249 | ||
250 | spin_lock_bh(&mfc_unres_lock); | |
251 | mfc = mfc_unres_queue; | |
252 | if (mfc) | |
253 | return mfc; | |
254 | ||
255 | end_of_list: | |
256 | spin_unlock_bh(&mfc_unres_lock); | |
257 | it->cache = NULL; | |
258 | ||
259 | return NULL; | |
260 | } | |
261 | ||
262 | static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) | |
263 | { | |
264 | struct ipmr_mfc_iter *it = seq->private; | |
265 | ||
266 | if (it->cache == &mfc_unres_queue) | |
267 | spin_unlock_bh(&mfc_unres_lock); | |
4a6258a0 | 268 | else if (it->cache == init_net.ipv6.mfc6_cache_array) |
7bc570c8 YH |
269 | read_unlock(&mrt_lock); |
270 | } | |
271 | ||
272 | static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) | |
273 | { | |
274 | int n; | |
275 | ||
276 | if (v == SEQ_START_TOKEN) { | |
277 | seq_puts(seq, | |
278 | "Group " | |
279 | "Origin " | |
280 | "Iif Pkts Bytes Wrong Oifs\n"); | |
281 | } else { | |
282 | const struct mfc6_cache *mfc = v; | |
283 | const struct ipmr_mfc_iter *it = seq->private; | |
284 | ||
999890b2 | 285 | seq_printf(seq, "%pI6 %pI6 %-3hd", |
0c6ce78a | 286 | &mfc->mf6c_mcastgrp, &mfc->mf6c_origin, |
1ea472e2 | 287 | mfc->mf6c_parent); |
7bc570c8 YH |
288 | |
289 | if (it->cache != &mfc_unres_queue) { | |
1ea472e2 BT |
290 | seq_printf(seq, " %8lu %8lu %8lu", |
291 | mfc->mfc_un.res.pkt, | |
292 | mfc->mfc_un.res.bytes, | |
293 | mfc->mfc_un.res.wrong_if); | |
7bc570c8 YH |
294 | for (n = mfc->mfc_un.res.minvif; |
295 | n < mfc->mfc_un.res.maxvif; n++) { | |
4e16880c | 296 | if (MIF_EXISTS(&init_net, n) && |
7bc570c8 YH |
297 | mfc->mfc_un.res.ttls[n] < 255) |
298 | seq_printf(seq, | |
299 | " %2d:%-3d", | |
300 | n, mfc->mfc_un.res.ttls[n]); | |
301 | } | |
1ea472e2 BT |
302 | } else { |
303 | /* unresolved mfc_caches don't contain | |
304 | * pkt, bytes and wrong_if values | |
305 | */ | |
306 | seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); | |
7bc570c8 YH |
307 | } |
308 | seq_putc(seq, '\n'); | |
309 | } | |
310 | return 0; | |
311 | } | |
312 | ||
313 | static struct seq_operations ipmr_mfc_seq_ops = { | |
314 | .start = ipmr_mfc_seq_start, | |
315 | .next = ipmr_mfc_seq_next, | |
316 | .stop = ipmr_mfc_seq_stop, | |
317 | .show = ipmr_mfc_seq_show, | |
318 | }; | |
319 | ||
320 | static int ipmr_mfc_open(struct inode *inode, struct file *file) | |
321 | { | |
322 | return seq_open_private(file, &ipmr_mfc_seq_ops, | |
323 | sizeof(struct ipmr_mfc_iter)); | |
324 | } | |
325 | ||
326 | static struct file_operations ip6mr_mfc_fops = { | |
327 | .owner = THIS_MODULE, | |
328 | .open = ipmr_mfc_open, | |
329 | .read = seq_read, | |
330 | .llseek = seq_lseek, | |
eedd726e | 331 | .release = seq_release_private, |
7bc570c8 YH |
332 | }; |
333 | #endif | |
334 | ||
14fb64e1 | 335 | #ifdef CONFIG_IPV6_PIMSM_V2 |
14fb64e1 YH |
336 | |
337 | static int pim6_rcv(struct sk_buff *skb) | |
338 | { | |
339 | struct pimreghdr *pim; | |
340 | struct ipv6hdr *encap; | |
341 | struct net_device *reg_dev = NULL; | |
950d5704 | 342 | int reg_vif_num = init_net.ipv6.mroute_reg_vif_num; |
14fb64e1 YH |
343 | |
344 | if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap))) | |
345 | goto drop; | |
346 | ||
347 | pim = (struct pimreghdr *)skb_transport_header(skb); | |
348 | if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) || | |
349 | (pim->flags & PIM_NULL_REGISTER) || | |
350 | (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && | |
ec6b486f | 351 | csum_fold(skb_checksum(skb, 0, skb->len, 0)))) |
14fb64e1 YH |
352 | goto drop; |
353 | ||
354 | /* check if the inner packet is destined to mcast group */ | |
355 | encap = (struct ipv6hdr *)(skb_transport_header(skb) + | |
356 | sizeof(*pim)); | |
357 | ||
358 | if (!ipv6_addr_is_multicast(&encap->daddr) || | |
359 | encap->payload_len == 0 || | |
360 | ntohs(encap->payload_len) + sizeof(*pim) > skb->len) | |
361 | goto drop; | |
362 | ||
363 | read_lock(&mrt_lock); | |
364 | if (reg_vif_num >= 0) | |
4e16880c | 365 | reg_dev = init_net.ipv6.vif6_table[reg_vif_num].dev; |
14fb64e1 YH |
366 | if (reg_dev) |
367 | dev_hold(reg_dev); | |
368 | read_unlock(&mrt_lock); | |
369 | ||
370 | if (reg_dev == NULL) | |
371 | goto drop; | |
372 | ||
373 | skb->mac_header = skb->network_header; | |
374 | skb_pull(skb, (u8 *)encap - skb->data); | |
375 | skb_reset_network_header(skb); | |
376 | skb->dev = reg_dev; | |
377 | skb->protocol = htons(ETH_P_IP); | |
378 | skb->ip_summed = 0; | |
379 | skb->pkt_type = PACKET_HOST; | |
380 | dst_release(skb->dst); | |
dc58c78c PE |
381 | reg_dev->stats.rx_bytes += skb->len; |
382 | reg_dev->stats.rx_packets++; | |
14fb64e1 YH |
383 | skb->dst = NULL; |
384 | nf_reset(skb); | |
385 | netif_rx(skb); | |
386 | dev_put(reg_dev); | |
387 | return 0; | |
388 | drop: | |
389 | kfree_skb(skb); | |
390 | return 0; | |
391 | } | |
392 | ||
393 | static struct inet6_protocol pim6_protocol = { | |
394 | .handler = pim6_rcv, | |
395 | }; | |
396 | ||
397 | /* Service routines creating virtual interfaces: PIMREG */ | |
398 | ||
399 | static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) | |
400 | { | |
401 | read_lock(&mrt_lock); | |
dc58c78c PE |
402 | dev->stats.tx_bytes += skb->len; |
403 | dev->stats.tx_packets++; | |
950d5704 | 404 | ip6mr_cache_report(skb, init_net.ipv6.mroute_reg_vif_num, MRT6MSG_WHOLEPKT); |
14fb64e1 YH |
405 | read_unlock(&mrt_lock); |
406 | kfree_skb(skb); | |
407 | return 0; | |
408 | } | |
409 | ||
007c3838 SH |
410 | static const struct net_device_ops reg_vif_netdev_ops = { |
411 | .ndo_start_xmit = reg_vif_xmit, | |
412 | }; | |
413 | ||
14fb64e1 YH |
414 | static void reg_vif_setup(struct net_device *dev) |
415 | { | |
416 | dev->type = ARPHRD_PIMREG; | |
417 | dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8; | |
418 | dev->flags = IFF_NOARP; | |
007c3838 | 419 | dev->netdev_ops = ®_vif_netdev_ops; |
14fb64e1 YH |
420 | dev->destructor = free_netdev; |
421 | } | |
422 | ||
423 | static struct net_device *ip6mr_reg_vif(void) | |
424 | { | |
425 | struct net_device *dev; | |
14fb64e1 | 426 | |
dc58c78c | 427 | dev = alloc_netdev(0, "pim6reg", reg_vif_setup); |
14fb64e1 YH |
428 | if (dev == NULL) |
429 | return NULL; | |
430 | ||
431 | if (register_netdevice(dev)) { | |
432 | free_netdev(dev); | |
433 | return NULL; | |
434 | } | |
435 | dev->iflink = 0; | |
436 | ||
14fb64e1 YH |
437 | if (dev_open(dev)) |
438 | goto failure; | |
439 | ||
7af3db78 | 440 | dev_hold(dev); |
14fb64e1 YH |
441 | return dev; |
442 | ||
443 | failure: | |
444 | /* allow the register to be completed before unregistering. */ | |
445 | rtnl_unlock(); | |
446 | rtnl_lock(); | |
447 | ||
448 | unregister_netdevice(dev); | |
449 | return NULL; | |
450 | } | |
451 | #endif | |
452 | ||
7bc570c8 YH |
453 | /* |
454 | * Delete a VIF entry | |
455 | */ | |
456 | ||
457 | static int mif6_delete(int vifi) | |
458 | { | |
459 | struct mif_device *v; | |
460 | struct net_device *dev; | |
4e16880c | 461 | if (vifi < 0 || vifi >= init_net.ipv6.maxvif) |
7bc570c8 YH |
462 | return -EADDRNOTAVAIL; |
463 | ||
4e16880c | 464 | v = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 YH |
465 | |
466 | write_lock_bh(&mrt_lock); | |
467 | dev = v->dev; | |
468 | v->dev = NULL; | |
469 | ||
470 | if (!dev) { | |
471 | write_unlock_bh(&mrt_lock); | |
472 | return -EADDRNOTAVAIL; | |
473 | } | |
474 | ||
14fb64e1 | 475 | #ifdef CONFIG_IPV6_PIMSM_V2 |
950d5704 BT |
476 | if (vifi == init_net.ipv6.mroute_reg_vif_num) |
477 | init_net.ipv6.mroute_reg_vif_num = -1; | |
14fb64e1 YH |
478 | #endif |
479 | ||
4e16880c | 480 | if (vifi + 1 == init_net.ipv6.maxvif) { |
7bc570c8 YH |
481 | int tmp; |
482 | for (tmp = vifi - 1; tmp >= 0; tmp--) { | |
4e16880c | 483 | if (MIF_EXISTS(&init_net, tmp)) |
7bc570c8 YH |
484 | break; |
485 | } | |
4e16880c | 486 | init_net.ipv6.maxvif = tmp + 1; |
7bc570c8 YH |
487 | } |
488 | ||
489 | write_unlock_bh(&mrt_lock); | |
490 | ||
491 | dev_set_allmulti(dev, -1); | |
492 | ||
493 | if (v->flags & MIFF_REGISTER) | |
494 | unregister_netdevice(dev); | |
495 | ||
496 | dev_put(dev); | |
497 | return 0; | |
498 | } | |
499 | ||
58701ad4 BT |
500 | static inline void ip6mr_cache_free(struct mfc6_cache *c) |
501 | { | |
502 | release_net(mfc6_net(c)); | |
503 | kmem_cache_free(mrt_cachep, c); | |
504 | } | |
505 | ||
7bc570c8 YH |
506 | /* Destroy an unresolved cache entry, killing queued skbs |
507 | and reporting error to netlink readers. | |
508 | */ | |
509 | ||
510 | static void ip6mr_destroy_unres(struct mfc6_cache *c) | |
511 | { | |
512 | struct sk_buff *skb; | |
513 | ||
4045e57c | 514 | atomic_dec(&init_net.ipv6.cache_resolve_queue_len); |
7bc570c8 YH |
515 | |
516 | while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) { | |
517 | if (ipv6_hdr(skb)->version == 0) { | |
518 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); | |
519 | nlh->nlmsg_type = NLMSG_ERROR; | |
520 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
521 | skb_trim(skb, nlh->nlmsg_len); | |
522 | ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT; | |
523 | rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); | |
524 | } else | |
525 | kfree_skb(skb); | |
526 | } | |
527 | ||
58701ad4 | 528 | ip6mr_cache_free(c); |
7bc570c8 YH |
529 | } |
530 | ||
531 | ||
532 | /* Single timer process for all the unresolved queue. */ | |
533 | ||
534 | static void ipmr_do_expire_process(unsigned long dummy) | |
535 | { | |
536 | unsigned long now = jiffies; | |
537 | unsigned long expires = 10 * HZ; | |
538 | struct mfc6_cache *c, **cp; | |
539 | ||
540 | cp = &mfc_unres_queue; | |
541 | ||
542 | while ((c = *cp) != NULL) { | |
543 | if (time_after(c->mfc_un.unres.expires, now)) { | |
544 | /* not yet... */ | |
545 | unsigned long interval = c->mfc_un.unres.expires - now; | |
546 | if (interval < expires) | |
547 | expires = interval; | |
548 | cp = &c->next; | |
549 | continue; | |
550 | } | |
551 | ||
552 | *cp = c->next; | |
553 | ip6mr_destroy_unres(c); | |
554 | } | |
555 | ||
4045e57c | 556 | if (mfc_unres_queue != NULL) |
7bc570c8 YH |
557 | mod_timer(&ipmr_expire_timer, jiffies + expires); |
558 | } | |
559 | ||
560 | static void ipmr_expire_process(unsigned long dummy) | |
561 | { | |
562 | if (!spin_trylock(&mfc_unres_lock)) { | |
563 | mod_timer(&ipmr_expire_timer, jiffies + 1); | |
564 | return; | |
565 | } | |
566 | ||
4045e57c | 567 | if (mfc_unres_queue != NULL) |
7bc570c8 YH |
568 | ipmr_do_expire_process(dummy); |
569 | ||
570 | spin_unlock(&mfc_unres_lock); | |
571 | } | |
572 | ||
573 | /* Fill oifs list. It is called under write locked mrt_lock. */ | |
574 | ||
575 | static void ip6mr_update_thresholds(struct mfc6_cache *cache, unsigned char *ttls) | |
576 | { | |
577 | int vifi; | |
578 | ||
6ac7eb08 | 579 | cache->mfc_un.res.minvif = MAXMIFS; |
7bc570c8 | 580 | cache->mfc_un.res.maxvif = 0; |
6ac7eb08 | 581 | memset(cache->mfc_un.res.ttls, 255, MAXMIFS); |
7bc570c8 | 582 | |
4e16880c BT |
583 | for (vifi = 0; vifi < init_net.ipv6.maxvif; vifi++) { |
584 | if (MIF_EXISTS(&init_net, vifi) && | |
585 | ttls[vifi] && ttls[vifi] < 255) { | |
7bc570c8 YH |
586 | cache->mfc_un.res.ttls[vifi] = ttls[vifi]; |
587 | if (cache->mfc_un.res.minvif > vifi) | |
588 | cache->mfc_un.res.minvif = vifi; | |
589 | if (cache->mfc_un.res.maxvif <= vifi) | |
590 | cache->mfc_un.res.maxvif = vifi + 1; | |
591 | } | |
592 | } | |
593 | } | |
594 | ||
595 | static int mif6_add(struct mif6ctl *vifc, int mrtsock) | |
596 | { | |
597 | int vifi = vifc->mif6c_mifi; | |
4e16880c | 598 | struct mif_device *v = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 | 599 | struct net_device *dev; |
5ae7b444 | 600 | int err; |
7bc570c8 YH |
601 | |
602 | /* Is vif busy ? */ | |
4e16880c | 603 | if (MIF_EXISTS(&init_net, vifi)) |
7bc570c8 YH |
604 | return -EADDRINUSE; |
605 | ||
606 | switch (vifc->mif6c_flags) { | |
14fb64e1 YH |
607 | #ifdef CONFIG_IPV6_PIMSM_V2 |
608 | case MIFF_REGISTER: | |
609 | /* | |
610 | * Special Purpose VIF in PIM | |
611 | * All the packets will be sent to the daemon | |
612 | */ | |
950d5704 | 613 | if (init_net.ipv6.mroute_reg_vif_num >= 0) |
14fb64e1 YH |
614 | return -EADDRINUSE; |
615 | dev = ip6mr_reg_vif(); | |
616 | if (!dev) | |
617 | return -ENOBUFS; | |
5ae7b444 WC |
618 | err = dev_set_allmulti(dev, 1); |
619 | if (err) { | |
620 | unregister_netdevice(dev); | |
7af3db78 | 621 | dev_put(dev); |
5ae7b444 WC |
622 | return err; |
623 | } | |
14fb64e1 YH |
624 | break; |
625 | #endif | |
7bc570c8 YH |
626 | case 0: |
627 | dev = dev_get_by_index(&init_net, vifc->mif6c_pifi); | |
628 | if (!dev) | |
629 | return -EADDRNOTAVAIL; | |
5ae7b444 | 630 | err = dev_set_allmulti(dev, 1); |
7af3db78 WC |
631 | if (err) { |
632 | dev_put(dev); | |
5ae7b444 | 633 | return err; |
7af3db78 | 634 | } |
7bc570c8 YH |
635 | break; |
636 | default: | |
637 | return -EINVAL; | |
638 | } | |
639 | ||
7bc570c8 YH |
640 | /* |
641 | * Fill in the VIF structures | |
642 | */ | |
643 | v->rate_limit = vifc->vifc_rate_limit; | |
644 | v->flags = vifc->mif6c_flags; | |
645 | if (!mrtsock) | |
646 | v->flags |= VIFF_STATIC; | |
647 | v->threshold = vifc->vifc_threshold; | |
648 | v->bytes_in = 0; | |
649 | v->bytes_out = 0; | |
650 | v->pkt_in = 0; | |
651 | v->pkt_out = 0; | |
652 | v->link = dev->ifindex; | |
653 | if (v->flags & MIFF_REGISTER) | |
654 | v->link = dev->iflink; | |
655 | ||
656 | /* And finish update writing critical data */ | |
657 | write_lock_bh(&mrt_lock); | |
7bc570c8 | 658 | v->dev = dev; |
14fb64e1 YH |
659 | #ifdef CONFIG_IPV6_PIMSM_V2 |
660 | if (v->flags & MIFF_REGISTER) | |
950d5704 | 661 | init_net.ipv6.mroute_reg_vif_num = vifi; |
14fb64e1 | 662 | #endif |
4e16880c BT |
663 | if (vifi + 1 > init_net.ipv6.maxvif) |
664 | init_net.ipv6.maxvif = vifi + 1; | |
7bc570c8 YH |
665 | write_unlock_bh(&mrt_lock); |
666 | return 0; | |
667 | } | |
668 | ||
669 | static struct mfc6_cache *ip6mr_cache_find(struct in6_addr *origin, struct in6_addr *mcastgrp) | |
670 | { | |
671 | int line = MFC6_HASH(mcastgrp, origin); | |
672 | struct mfc6_cache *c; | |
673 | ||
4a6258a0 | 674 | for (c = init_net.ipv6.mfc6_cache_array[line]; c; c = c->next) { |
7bc570c8 YH |
675 | if (ipv6_addr_equal(&c->mf6c_origin, origin) && |
676 | ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp)) | |
677 | break; | |
678 | } | |
679 | return c; | |
680 | } | |
681 | ||
682 | /* | |
683 | * Allocate a multicast cache entry | |
684 | */ | |
58701ad4 | 685 | static struct mfc6_cache *ip6mr_cache_alloc(struct net *net) |
7bc570c8 | 686 | { |
36cbac59 | 687 | struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); |
7bc570c8 YH |
688 | if (c == NULL) |
689 | return NULL; | |
6ac7eb08 | 690 | c->mfc_un.res.minvif = MAXMIFS; |
58701ad4 | 691 | mfc6_net_set(c, net); |
7bc570c8 YH |
692 | return c; |
693 | } | |
694 | ||
58701ad4 | 695 | static struct mfc6_cache *ip6mr_cache_alloc_unres(struct net *net) |
7bc570c8 | 696 | { |
36cbac59 | 697 | struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); |
7bc570c8 YH |
698 | if (c == NULL) |
699 | return NULL; | |
7bc570c8 YH |
700 | skb_queue_head_init(&c->mfc_un.unres.unresolved); |
701 | c->mfc_un.unres.expires = jiffies + 10 * HZ; | |
58701ad4 | 702 | mfc6_net_set(c, net); |
7bc570c8 YH |
703 | return c; |
704 | } | |
705 | ||
706 | /* | |
707 | * A cache entry has gone into a resolved state from queued | |
708 | */ | |
709 | ||
710 | static void ip6mr_cache_resolve(struct mfc6_cache *uc, struct mfc6_cache *c) | |
711 | { | |
712 | struct sk_buff *skb; | |
713 | ||
714 | /* | |
715 | * Play the pending entries through our router | |
716 | */ | |
717 | ||
718 | while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { | |
719 | if (ipv6_hdr(skb)->version == 0) { | |
720 | int err; | |
721 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr)); | |
722 | ||
723 | if (ip6mr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) { | |
549e028d | 724 | nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh; |
7bc570c8 YH |
725 | } else { |
726 | nlh->nlmsg_type = NLMSG_ERROR; | |
727 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
728 | skb_trim(skb, nlh->nlmsg_len); | |
729 | ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE; | |
730 | } | |
731 | err = rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); | |
732 | } else | |
733 | ip6_mr_forward(skb, c); | |
734 | } | |
735 | } | |
736 | ||
737 | /* | |
738 | * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd | |
739 | * expects the following bizarre scheme. | |
740 | * | |
741 | * Called under mrt_lock. | |
742 | */ | |
743 | ||
6ac7eb08 | 744 | static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert) |
7bc570c8 YH |
745 | { |
746 | struct sk_buff *skb; | |
747 | struct mrt6msg *msg; | |
748 | int ret; | |
749 | ||
14fb64e1 YH |
750 | #ifdef CONFIG_IPV6_PIMSM_V2 |
751 | if (assert == MRT6MSG_WHOLEPKT) | |
752 | skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt) | |
753 | +sizeof(*msg)); | |
754 | else | |
755 | #endif | |
756 | skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC); | |
7bc570c8 YH |
757 | |
758 | if (!skb) | |
759 | return -ENOBUFS; | |
760 | ||
761 | /* I suppose that internal messages | |
762 | * do not require checksums */ | |
763 | ||
764 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
765 | ||
14fb64e1 YH |
766 | #ifdef CONFIG_IPV6_PIMSM_V2 |
767 | if (assert == MRT6MSG_WHOLEPKT) { | |
768 | /* Ugly, but we have no choice with this interface. | |
769 | Duplicate old header, fix length etc. | |
770 | And all this only to mangle msg->im6_msgtype and | |
771 | to set msg->im6_mbz to "mbz" :-) | |
772 | */ | |
773 | skb_push(skb, -skb_network_offset(pkt)); | |
774 | ||
775 | skb_push(skb, sizeof(*msg)); | |
776 | skb_reset_transport_header(skb); | |
777 | msg = (struct mrt6msg *)skb_transport_header(skb); | |
778 | msg->im6_mbz = 0; | |
779 | msg->im6_msgtype = MRT6MSG_WHOLEPKT; | |
950d5704 | 780 | msg->im6_mif = init_net.ipv6.mroute_reg_vif_num; |
14fb64e1 YH |
781 | msg->im6_pad = 0; |
782 | ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr); | |
783 | ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr); | |
784 | ||
785 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
786 | } else | |
787 | #endif | |
788 | { | |
7bc570c8 YH |
789 | /* |
790 | * Copy the IP header | |
791 | */ | |
792 | ||
793 | skb_put(skb, sizeof(struct ipv6hdr)); | |
794 | skb_reset_network_header(skb); | |
795 | skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr)); | |
796 | ||
797 | /* | |
798 | * Add our header | |
799 | */ | |
800 | skb_put(skb, sizeof(*msg)); | |
801 | skb_reset_transport_header(skb); | |
802 | msg = (struct mrt6msg *)skb_transport_header(skb); | |
803 | ||
804 | msg->im6_mbz = 0; | |
805 | msg->im6_msgtype = assert; | |
6ac7eb08 | 806 | msg->im6_mif = mifi; |
7bc570c8 YH |
807 | msg->im6_pad = 0; |
808 | ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr); | |
809 | ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr); | |
810 | ||
811 | skb->dst = dst_clone(pkt->dst); | |
812 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
813 | ||
814 | skb_pull(skb, sizeof(struct ipv6hdr)); | |
14fb64e1 | 815 | } |
7bc570c8 | 816 | |
bd91b8bf | 817 | if (init_net.ipv6.mroute6_sk == NULL) { |
7bc570c8 YH |
818 | kfree_skb(skb); |
819 | return -EINVAL; | |
820 | } | |
821 | ||
822 | /* | |
823 | * Deliver to user space multicast routing algorithms | |
824 | */ | |
bd91b8bf BT |
825 | ret = sock_queue_rcv_skb(init_net.ipv6.mroute6_sk, skb); |
826 | if (ret < 0) { | |
7bc570c8 YH |
827 | if (net_ratelimit()) |
828 | printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n"); | |
829 | kfree_skb(skb); | |
830 | } | |
831 | ||
832 | return ret; | |
833 | } | |
834 | ||
835 | /* | |
836 | * Queue a packet for resolution. It gets locked cache entry! | |
837 | */ | |
838 | ||
839 | static int | |
6ac7eb08 | 840 | ip6mr_cache_unresolved(mifi_t mifi, struct sk_buff *skb) |
7bc570c8 YH |
841 | { |
842 | int err; | |
843 | struct mfc6_cache *c; | |
844 | ||
845 | spin_lock_bh(&mfc_unres_lock); | |
846 | for (c = mfc_unres_queue; c; c = c->next) { | |
4045e57c BT |
847 | if (net_eq(mfc6_net(c), &init_net) && |
848 | ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) && | |
7bc570c8 YH |
849 | ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) |
850 | break; | |
851 | } | |
852 | ||
853 | if (c == NULL) { | |
854 | /* | |
855 | * Create a new entry if allowable | |
856 | */ | |
857 | ||
4045e57c | 858 | if (atomic_read(&init_net.ipv6.cache_resolve_queue_len) >= 10 || |
58701ad4 | 859 | (c = ip6mr_cache_alloc_unres(&init_net)) == NULL) { |
7bc570c8 YH |
860 | spin_unlock_bh(&mfc_unres_lock); |
861 | ||
862 | kfree_skb(skb); | |
863 | return -ENOBUFS; | |
864 | } | |
865 | ||
866 | /* | |
867 | * Fill in the new cache entry | |
868 | */ | |
869 | c->mf6c_parent = -1; | |
870 | c->mf6c_origin = ipv6_hdr(skb)->saddr; | |
871 | c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr; | |
872 | ||
873 | /* | |
874 | * Reflect first query at pim6sd | |
875 | */ | |
6ac7eb08 | 876 | if ((err = ip6mr_cache_report(skb, mifi, MRT6MSG_NOCACHE)) < 0) { |
7bc570c8 YH |
877 | /* If the report failed throw the cache entry |
878 | out - Brad Parker | |
879 | */ | |
880 | spin_unlock_bh(&mfc_unres_lock); | |
881 | ||
58701ad4 | 882 | ip6mr_cache_free(c); |
7bc570c8 YH |
883 | kfree_skb(skb); |
884 | return err; | |
885 | } | |
886 | ||
4045e57c | 887 | atomic_inc(&init_net.ipv6.cache_resolve_queue_len); |
7bc570c8 YH |
888 | c->next = mfc_unres_queue; |
889 | mfc_unres_queue = c; | |
890 | ||
891 | ipmr_do_expire_process(1); | |
892 | } | |
893 | ||
894 | /* | |
895 | * See if we can append the packet | |
896 | */ | |
897 | if (c->mfc_un.unres.unresolved.qlen > 3) { | |
898 | kfree_skb(skb); | |
899 | err = -ENOBUFS; | |
900 | } else { | |
901 | skb_queue_tail(&c->mfc_un.unres.unresolved, skb); | |
902 | err = 0; | |
903 | } | |
904 | ||
905 | spin_unlock_bh(&mfc_unres_lock); | |
906 | return err; | |
907 | } | |
908 | ||
909 | /* | |
910 | * MFC6 cache manipulation by user space | |
911 | */ | |
912 | ||
913 | static int ip6mr_mfc_delete(struct mf6cctl *mfc) | |
914 | { | |
915 | int line; | |
916 | struct mfc6_cache *c, **cp; | |
917 | ||
918 | line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); | |
919 | ||
4a6258a0 BT |
920 | for (cp = &init_net.ipv6.mfc6_cache_array[line]; |
921 | (c = *cp) != NULL; cp = &c->next) { | |
7bc570c8 YH |
922 | if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && |
923 | ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) { | |
924 | write_lock_bh(&mrt_lock); | |
925 | *cp = c->next; | |
926 | write_unlock_bh(&mrt_lock); | |
927 | ||
58701ad4 | 928 | ip6mr_cache_free(c); |
7bc570c8 YH |
929 | return 0; |
930 | } | |
931 | } | |
932 | return -ENOENT; | |
933 | } | |
934 | ||
935 | static int ip6mr_device_event(struct notifier_block *this, | |
936 | unsigned long event, void *ptr) | |
937 | { | |
938 | struct net_device *dev = ptr; | |
939 | struct mif_device *v; | |
940 | int ct; | |
941 | ||
721499e8 | 942 | if (!net_eq(dev_net(dev), &init_net)) |
7bc570c8 YH |
943 | return NOTIFY_DONE; |
944 | ||
945 | if (event != NETDEV_UNREGISTER) | |
946 | return NOTIFY_DONE; | |
947 | ||
4e16880c BT |
948 | v = &init_net.ipv6.vif6_table[0]; |
949 | for (ct = 0; ct < init_net.ipv6.maxvif; ct++, v++) { | |
7bc570c8 YH |
950 | if (v->dev == dev) |
951 | mif6_delete(ct); | |
952 | } | |
953 | return NOTIFY_DONE; | |
954 | } | |
955 | ||
956 | static struct notifier_block ip6_mr_notifier = { | |
957 | .notifier_call = ip6mr_device_event | |
958 | }; | |
959 | ||
960 | /* | |
961 | * Setup for IP multicast routing | |
962 | */ | |
963 | ||
4e16880c BT |
964 | static int __net_init ip6mr_net_init(struct net *net) |
965 | { | |
966 | int err = 0; | |
4e16880c BT |
967 | net->ipv6.vif6_table = kcalloc(MAXMIFS, sizeof(struct mif_device), |
968 | GFP_KERNEL); | |
969 | if (!net->ipv6.vif6_table) { | |
970 | err = -ENOMEM; | |
971 | goto fail; | |
972 | } | |
4a6258a0 BT |
973 | |
974 | /* Forwarding cache */ | |
975 | net->ipv6.mfc6_cache_array = kcalloc(MFC6_LINES, | |
976 | sizeof(struct mfc6_cache *), | |
977 | GFP_KERNEL); | |
978 | if (!net->ipv6.mfc6_cache_array) { | |
979 | err = -ENOMEM; | |
980 | goto fail_mfc6_cache; | |
981 | } | |
950d5704 BT |
982 | |
983 | #ifdef CONFIG_IPV6_PIMSM_V2 | |
984 | net->ipv6.mroute_reg_vif_num = -1; | |
985 | #endif | |
4a6258a0 BT |
986 | return 0; |
987 | ||
988 | fail_mfc6_cache: | |
989 | kfree(net->ipv6.vif6_table); | |
4e16880c BT |
990 | fail: |
991 | return err; | |
992 | } | |
993 | ||
994 | static void __net_exit ip6mr_net_exit(struct net *net) | |
995 | { | |
4a6258a0 | 996 | kfree(net->ipv6.mfc6_cache_array); |
4e16880c BT |
997 | kfree(net->ipv6.vif6_table); |
998 | } | |
999 | ||
1000 | static struct pernet_operations ip6mr_net_ops = { | |
1001 | .init = ip6mr_net_init, | |
1002 | .exit = ip6mr_net_exit, | |
1003 | }; | |
1004 | ||
623d1a1a | 1005 | int __init ip6_mr_init(void) |
7bc570c8 | 1006 | { |
623d1a1a WC |
1007 | int err; |
1008 | ||
7bc570c8 YH |
1009 | mrt_cachep = kmem_cache_create("ip6_mrt_cache", |
1010 | sizeof(struct mfc6_cache), | |
1011 | 0, SLAB_HWCACHE_ALIGN, | |
1012 | NULL); | |
1013 | if (!mrt_cachep) | |
623d1a1a | 1014 | return -ENOMEM; |
7bc570c8 | 1015 | |
4e16880c BT |
1016 | err = register_pernet_subsys(&ip6mr_net_ops); |
1017 | if (err) | |
1018 | goto reg_pernet_fail; | |
1019 | ||
7bc570c8 | 1020 | setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0); |
623d1a1a WC |
1021 | err = register_netdevice_notifier(&ip6_mr_notifier); |
1022 | if (err) | |
1023 | goto reg_notif_fail; | |
1024 | #ifdef CONFIG_PROC_FS | |
1025 | err = -ENOMEM; | |
1026 | if (!proc_net_fops_create(&init_net, "ip6_mr_vif", 0, &ip6mr_vif_fops)) | |
1027 | goto proc_vif_fail; | |
1028 | if (!proc_net_fops_create(&init_net, "ip6_mr_cache", | |
1029 | 0, &ip6mr_mfc_fops)) | |
1030 | goto proc_cache_fail; | |
1031 | #endif | |
1032 | return 0; | |
7bc570c8 | 1033 | #ifdef CONFIG_PROC_FS |
623d1a1a WC |
1034 | proc_cache_fail: |
1035 | proc_net_remove(&init_net, "ip6_mr_vif"); | |
87b30a65 BT |
1036 | proc_vif_fail: |
1037 | unregister_netdevice_notifier(&ip6_mr_notifier); | |
7bc570c8 | 1038 | #endif |
87b30a65 BT |
1039 | reg_notif_fail: |
1040 | del_timer(&ipmr_expire_timer); | |
4e16880c BT |
1041 | unregister_pernet_subsys(&ip6mr_net_ops); |
1042 | reg_pernet_fail: | |
87b30a65 | 1043 | kmem_cache_destroy(mrt_cachep); |
623d1a1a | 1044 | return err; |
7bc570c8 YH |
1045 | } |
1046 | ||
623d1a1a WC |
1047 | void ip6_mr_cleanup(void) |
1048 | { | |
1049 | #ifdef CONFIG_PROC_FS | |
1050 | proc_net_remove(&init_net, "ip6_mr_cache"); | |
1051 | proc_net_remove(&init_net, "ip6_mr_vif"); | |
1052 | #endif | |
1053 | unregister_netdevice_notifier(&ip6_mr_notifier); | |
1054 | del_timer(&ipmr_expire_timer); | |
4e16880c | 1055 | unregister_pernet_subsys(&ip6mr_net_ops); |
623d1a1a WC |
1056 | kmem_cache_destroy(mrt_cachep); |
1057 | } | |
7bc570c8 YH |
1058 | |
1059 | static int ip6mr_mfc_add(struct mf6cctl *mfc, int mrtsock) | |
1060 | { | |
1061 | int line; | |
1062 | struct mfc6_cache *uc, *c, **cp; | |
6ac7eb08 | 1063 | unsigned char ttls[MAXMIFS]; |
7bc570c8 YH |
1064 | int i; |
1065 | ||
6ac7eb08 RR |
1066 | memset(ttls, 255, MAXMIFS); |
1067 | for (i = 0; i < MAXMIFS; i++) { | |
7bc570c8 YH |
1068 | if (IF_ISSET(i, &mfc->mf6cc_ifset)) |
1069 | ttls[i] = 1; | |
1070 | ||
1071 | } | |
1072 | ||
1073 | line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr); | |
1074 | ||
4a6258a0 BT |
1075 | for (cp = &init_net.ipv6.mfc6_cache_array[line]; |
1076 | (c = *cp) != NULL; cp = &c->next) { | |
7bc570c8 YH |
1077 | if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) && |
1078 | ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) | |
1079 | break; | |
1080 | } | |
1081 | ||
1082 | if (c != NULL) { | |
1083 | write_lock_bh(&mrt_lock); | |
1084 | c->mf6c_parent = mfc->mf6cc_parent; | |
1085 | ip6mr_update_thresholds(c, ttls); | |
1086 | if (!mrtsock) | |
1087 | c->mfc_flags |= MFC_STATIC; | |
1088 | write_unlock_bh(&mrt_lock); | |
1089 | return 0; | |
1090 | } | |
1091 | ||
1092 | if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr)) | |
1093 | return -EINVAL; | |
1094 | ||
58701ad4 | 1095 | c = ip6mr_cache_alloc(&init_net); |
7bc570c8 YH |
1096 | if (c == NULL) |
1097 | return -ENOMEM; | |
1098 | ||
1099 | c->mf6c_origin = mfc->mf6cc_origin.sin6_addr; | |
1100 | c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr; | |
1101 | c->mf6c_parent = mfc->mf6cc_parent; | |
1102 | ip6mr_update_thresholds(c, ttls); | |
1103 | if (!mrtsock) | |
1104 | c->mfc_flags |= MFC_STATIC; | |
1105 | ||
1106 | write_lock_bh(&mrt_lock); | |
4a6258a0 BT |
1107 | c->next = init_net.ipv6.mfc6_cache_array[line]; |
1108 | init_net.ipv6.mfc6_cache_array[line] = c; | |
7bc570c8 YH |
1109 | write_unlock_bh(&mrt_lock); |
1110 | ||
1111 | /* | |
1112 | * Check to see if we resolved a queued list. If so we | |
1113 | * need to send on the frames and tidy up. | |
1114 | */ | |
1115 | spin_lock_bh(&mfc_unres_lock); | |
1116 | for (cp = &mfc_unres_queue; (uc = *cp) != NULL; | |
1117 | cp = &uc->next) { | |
4045e57c BT |
1118 | if (net_eq(mfc6_net(uc), &init_net) && |
1119 | ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) && | |
7bc570c8 YH |
1120 | ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) { |
1121 | *cp = uc->next; | |
4045e57c | 1122 | atomic_dec(&init_net.ipv6.cache_resolve_queue_len); |
7bc570c8 YH |
1123 | break; |
1124 | } | |
1125 | } | |
4045e57c BT |
1126 | if (mfc_unres_queue == NULL) |
1127 | del_timer(&ipmr_expire_timer); | |
7bc570c8 YH |
1128 | spin_unlock_bh(&mfc_unres_lock); |
1129 | ||
1130 | if (uc) { | |
1131 | ip6mr_cache_resolve(uc, c); | |
58701ad4 | 1132 | ip6mr_cache_free(uc); |
7bc570c8 YH |
1133 | } |
1134 | return 0; | |
1135 | } | |
1136 | ||
1137 | /* | |
1138 | * Close the multicast socket, and clear the vif tables etc | |
1139 | */ | |
1140 | ||
1141 | static void mroute_clean_tables(struct sock *sk) | |
1142 | { | |
1143 | int i; | |
1144 | ||
1145 | /* | |
1146 | * Shut down all active vif entries | |
1147 | */ | |
4e16880c BT |
1148 | for (i = 0; i < init_net.ipv6.maxvif; i++) { |
1149 | if (!(init_net.ipv6.vif6_table[i].flags & VIFF_STATIC)) | |
7bc570c8 YH |
1150 | mif6_delete(i); |
1151 | } | |
1152 | ||
1153 | /* | |
1154 | * Wipe the cache | |
1155 | */ | |
4a6258a0 | 1156 | for (i = 0; i < MFC6_LINES; i++) { |
7bc570c8 YH |
1157 | struct mfc6_cache *c, **cp; |
1158 | ||
4a6258a0 | 1159 | cp = &init_net.ipv6.mfc6_cache_array[i]; |
7bc570c8 YH |
1160 | while ((c = *cp) != NULL) { |
1161 | if (c->mfc_flags & MFC_STATIC) { | |
1162 | cp = &c->next; | |
1163 | continue; | |
1164 | } | |
1165 | write_lock_bh(&mrt_lock); | |
1166 | *cp = c->next; | |
1167 | write_unlock_bh(&mrt_lock); | |
1168 | ||
58701ad4 | 1169 | ip6mr_cache_free(c); |
7bc570c8 YH |
1170 | } |
1171 | } | |
1172 | ||
4045e57c BT |
1173 | if (atomic_read(&init_net.ipv6.cache_resolve_queue_len) != 0) { |
1174 | struct mfc6_cache *c, **cp; | |
7bc570c8 YH |
1175 | |
1176 | spin_lock_bh(&mfc_unres_lock); | |
4045e57c BT |
1177 | cp = &mfc_unres_queue; |
1178 | while ((c = *cp) != NULL) { | |
1179 | if (!net_eq(mfc6_net(c), &init_net)) { | |
1180 | cp = &c->next; | |
1181 | continue; | |
1182 | } | |
1183 | *cp = c->next; | |
7bc570c8 | 1184 | ip6mr_destroy_unres(c); |
7bc570c8 YH |
1185 | } |
1186 | spin_unlock_bh(&mfc_unres_lock); | |
1187 | } | |
1188 | } | |
1189 | ||
1190 | static int ip6mr_sk_init(struct sock *sk) | |
1191 | { | |
1192 | int err = 0; | |
1193 | ||
1194 | rtnl_lock(); | |
1195 | write_lock_bh(&mrt_lock); | |
bd91b8bf BT |
1196 | if (likely(init_net.ipv6.mroute6_sk == NULL)) |
1197 | init_net.ipv6.mroute6_sk = sk; | |
7bc570c8 YH |
1198 | else |
1199 | err = -EADDRINUSE; | |
1200 | write_unlock_bh(&mrt_lock); | |
1201 | ||
1202 | rtnl_unlock(); | |
1203 | ||
1204 | return err; | |
1205 | } | |
1206 | ||
1207 | int ip6mr_sk_done(struct sock *sk) | |
1208 | { | |
1209 | int err = 0; | |
1210 | ||
1211 | rtnl_lock(); | |
bd91b8bf | 1212 | if (sk == init_net.ipv6.mroute6_sk) { |
7bc570c8 | 1213 | write_lock_bh(&mrt_lock); |
bd91b8bf | 1214 | init_net.ipv6.mroute6_sk = NULL; |
7bc570c8 YH |
1215 | write_unlock_bh(&mrt_lock); |
1216 | ||
1217 | mroute_clean_tables(sk); | |
1218 | } else | |
1219 | err = -EACCES; | |
1220 | rtnl_unlock(); | |
1221 | ||
1222 | return err; | |
1223 | } | |
1224 | ||
1225 | /* | |
1226 | * Socket options and virtual interface manipulation. The whole | |
1227 | * virtual interface system is a complete heap, but unfortunately | |
1228 | * that's how BSD mrouted happens to think. Maybe one day with a proper | |
1229 | * MOSPF/PIM router set up we can clean this up. | |
1230 | */ | |
1231 | ||
1232 | int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, int optlen) | |
1233 | { | |
1234 | int ret; | |
1235 | struct mif6ctl vif; | |
1236 | struct mf6cctl mfc; | |
1237 | mifi_t mifi; | |
1238 | ||
1239 | if (optname != MRT6_INIT) { | |
bd91b8bf | 1240 | if (sk != init_net.ipv6.mroute6_sk && !capable(CAP_NET_ADMIN)) |
7bc570c8 YH |
1241 | return -EACCES; |
1242 | } | |
1243 | ||
1244 | switch (optname) { | |
1245 | case MRT6_INIT: | |
1246 | if (sk->sk_type != SOCK_RAW || | |
1247 | inet_sk(sk)->num != IPPROTO_ICMPV6) | |
1248 | return -EOPNOTSUPP; | |
1249 | if (optlen < sizeof(int)) | |
1250 | return -EINVAL; | |
1251 | ||
1252 | return ip6mr_sk_init(sk); | |
1253 | ||
1254 | case MRT6_DONE: | |
1255 | return ip6mr_sk_done(sk); | |
1256 | ||
1257 | case MRT6_ADD_MIF: | |
1258 | if (optlen < sizeof(vif)) | |
1259 | return -EINVAL; | |
1260 | if (copy_from_user(&vif, optval, sizeof(vif))) | |
1261 | return -EFAULT; | |
6ac7eb08 | 1262 | if (vif.mif6c_mifi >= MAXMIFS) |
7bc570c8 YH |
1263 | return -ENFILE; |
1264 | rtnl_lock(); | |
bd91b8bf | 1265 | ret = mif6_add(&vif, sk == init_net.ipv6.mroute6_sk); |
7bc570c8 YH |
1266 | rtnl_unlock(); |
1267 | return ret; | |
1268 | ||
1269 | case MRT6_DEL_MIF: | |
1270 | if (optlen < sizeof(mifi_t)) | |
1271 | return -EINVAL; | |
1272 | if (copy_from_user(&mifi, optval, sizeof(mifi_t))) | |
1273 | return -EFAULT; | |
1274 | rtnl_lock(); | |
1275 | ret = mif6_delete(mifi); | |
1276 | rtnl_unlock(); | |
1277 | return ret; | |
1278 | ||
1279 | /* | |
1280 | * Manipulate the forwarding caches. These live | |
1281 | * in a sort of kernel/user symbiosis. | |
1282 | */ | |
1283 | case MRT6_ADD_MFC: | |
1284 | case MRT6_DEL_MFC: | |
1285 | if (optlen < sizeof(mfc)) | |
1286 | return -EINVAL; | |
1287 | if (copy_from_user(&mfc, optval, sizeof(mfc))) | |
1288 | return -EFAULT; | |
1289 | rtnl_lock(); | |
1290 | if (optname == MRT6_DEL_MFC) | |
1291 | ret = ip6mr_mfc_delete(&mfc); | |
1292 | else | |
bd91b8bf | 1293 | ret = ip6mr_mfc_add(&mfc, sk == init_net.ipv6.mroute6_sk); |
7bc570c8 YH |
1294 | rtnl_unlock(); |
1295 | return ret; | |
1296 | ||
14fb64e1 YH |
1297 | /* |
1298 | * Control PIM assert (to activate pim will activate assert) | |
1299 | */ | |
1300 | case MRT6_ASSERT: | |
1301 | { | |
1302 | int v; | |
1303 | if (get_user(v, (int __user *)optval)) | |
1304 | return -EFAULT; | |
a21f3f99 | 1305 | init_net.ipv6.mroute_do_assert = !!v; |
14fb64e1 YH |
1306 | return 0; |
1307 | } | |
1308 | ||
1309 | #ifdef CONFIG_IPV6_PIMSM_V2 | |
1310 | case MRT6_PIM: | |
1311 | { | |
a9f83bf3 | 1312 | int v; |
14fb64e1 YH |
1313 | if (get_user(v, (int __user *)optval)) |
1314 | return -EFAULT; | |
1315 | v = !!v; | |
1316 | rtnl_lock(); | |
1317 | ret = 0; | |
a21f3f99 BT |
1318 | if (v != init_net.ipv6.mroute_do_pim) { |
1319 | init_net.ipv6.mroute_do_pim = v; | |
1320 | init_net.ipv6.mroute_do_assert = v; | |
1321 | if (init_net.ipv6.mroute_do_pim) | |
14fb64e1 YH |
1322 | ret = inet6_add_protocol(&pim6_protocol, |
1323 | IPPROTO_PIM); | |
1324 | else | |
1325 | ret = inet6_del_protocol(&pim6_protocol, | |
1326 | IPPROTO_PIM); | |
1327 | if (ret < 0) | |
1328 | ret = -EAGAIN; | |
1329 | } | |
1330 | rtnl_unlock(); | |
1331 | return ret; | |
1332 | } | |
1333 | ||
1334 | #endif | |
7bc570c8 | 1335 | /* |
7d120c55 | 1336 | * Spurious command, or MRT6_VERSION which you cannot |
7bc570c8 YH |
1337 | * set. |
1338 | */ | |
1339 | default: | |
1340 | return -ENOPROTOOPT; | |
1341 | } | |
1342 | } | |
1343 | ||
1344 | /* | |
1345 | * Getsock opt support for the multicast routing system. | |
1346 | */ | |
1347 | ||
1348 | int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, | |
1349 | int __user *optlen) | |
1350 | { | |
1351 | int olr; | |
1352 | int val; | |
1353 | ||
1354 | switch (optname) { | |
1355 | case MRT6_VERSION: | |
1356 | val = 0x0305; | |
1357 | break; | |
14fb64e1 YH |
1358 | #ifdef CONFIG_IPV6_PIMSM_V2 |
1359 | case MRT6_PIM: | |
a21f3f99 | 1360 | val = init_net.ipv6.mroute_do_pim; |
14fb64e1 YH |
1361 | break; |
1362 | #endif | |
1363 | case MRT6_ASSERT: | |
a21f3f99 | 1364 | val = init_net.ipv6.mroute_do_assert; |
14fb64e1 | 1365 | break; |
7bc570c8 YH |
1366 | default: |
1367 | return -ENOPROTOOPT; | |
1368 | } | |
1369 | ||
1370 | if (get_user(olr, optlen)) | |
1371 | return -EFAULT; | |
1372 | ||
1373 | olr = min_t(int, olr, sizeof(int)); | |
1374 | if (olr < 0) | |
1375 | return -EINVAL; | |
1376 | ||
1377 | if (put_user(olr, optlen)) | |
1378 | return -EFAULT; | |
1379 | if (copy_to_user(optval, &val, olr)) | |
1380 | return -EFAULT; | |
1381 | return 0; | |
1382 | } | |
1383 | ||
1384 | /* | |
1385 | * The IP multicast ioctl support routines. | |
1386 | */ | |
1387 | ||
1388 | int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg) | |
1389 | { | |
1390 | struct sioc_sg_req6 sr; | |
1391 | struct sioc_mif_req6 vr; | |
1392 | struct mif_device *vif; | |
1393 | struct mfc6_cache *c; | |
1394 | ||
1395 | switch (cmd) { | |
1396 | case SIOCGETMIFCNT_IN6: | |
1397 | if (copy_from_user(&vr, arg, sizeof(vr))) | |
1398 | return -EFAULT; | |
4e16880c | 1399 | if (vr.mifi >= init_net.ipv6.maxvif) |
7bc570c8 YH |
1400 | return -EINVAL; |
1401 | read_lock(&mrt_lock); | |
4e16880c BT |
1402 | vif = &init_net.ipv6.vif6_table[vr.mifi]; |
1403 | if (MIF_EXISTS(&init_net, vr.mifi)) { | |
7bc570c8 YH |
1404 | vr.icount = vif->pkt_in; |
1405 | vr.ocount = vif->pkt_out; | |
1406 | vr.ibytes = vif->bytes_in; | |
1407 | vr.obytes = vif->bytes_out; | |
1408 | read_unlock(&mrt_lock); | |
1409 | ||
1410 | if (copy_to_user(arg, &vr, sizeof(vr))) | |
1411 | return -EFAULT; | |
1412 | return 0; | |
1413 | } | |
1414 | read_unlock(&mrt_lock); | |
1415 | return -EADDRNOTAVAIL; | |
1416 | case SIOCGETSGCNT_IN6: | |
1417 | if (copy_from_user(&sr, arg, sizeof(sr))) | |
1418 | return -EFAULT; | |
1419 | ||
1420 | read_lock(&mrt_lock); | |
1421 | c = ip6mr_cache_find(&sr.src.sin6_addr, &sr.grp.sin6_addr); | |
1422 | if (c) { | |
1423 | sr.pktcnt = c->mfc_un.res.pkt; | |
1424 | sr.bytecnt = c->mfc_un.res.bytes; | |
1425 | sr.wrong_if = c->mfc_un.res.wrong_if; | |
1426 | read_unlock(&mrt_lock); | |
1427 | ||
1428 | if (copy_to_user(arg, &sr, sizeof(sr))) | |
1429 | return -EFAULT; | |
1430 | return 0; | |
1431 | } | |
1432 | read_unlock(&mrt_lock); | |
1433 | return -EADDRNOTAVAIL; | |
1434 | default: | |
1435 | return -ENOIOCTLCMD; | |
1436 | } | |
1437 | } | |
1438 | ||
1439 | ||
1440 | static inline int ip6mr_forward2_finish(struct sk_buff *skb) | |
1441 | { | |
483a47d2 DL |
1442 | IP6_INC_STATS_BH(dev_net(skb->dst->dev), ip6_dst_idev(skb->dst), |
1443 | IPSTATS_MIB_OUTFORWDATAGRAMS); | |
7bc570c8 YH |
1444 | return dst_output(skb); |
1445 | } | |
1446 | ||
1447 | /* | |
1448 | * Processing handlers for ip6mr_forward | |
1449 | */ | |
1450 | ||
1451 | static int ip6mr_forward2(struct sk_buff *skb, struct mfc6_cache *c, int vifi) | |
1452 | { | |
1453 | struct ipv6hdr *ipv6h; | |
4e16880c | 1454 | struct mif_device *vif = &init_net.ipv6.vif6_table[vifi]; |
7bc570c8 YH |
1455 | struct net_device *dev; |
1456 | struct dst_entry *dst; | |
1457 | struct flowi fl; | |
1458 | ||
1459 | if (vif->dev == NULL) | |
1460 | goto out_free; | |
1461 | ||
14fb64e1 YH |
1462 | #ifdef CONFIG_IPV6_PIMSM_V2 |
1463 | if (vif->flags & MIFF_REGISTER) { | |
1464 | vif->pkt_out++; | |
1465 | vif->bytes_out += skb->len; | |
dc58c78c PE |
1466 | vif->dev->stats.tx_bytes += skb->len; |
1467 | vif->dev->stats.tx_packets++; | |
14fb64e1 YH |
1468 | ip6mr_cache_report(skb, vifi, MRT6MSG_WHOLEPKT); |
1469 | kfree_skb(skb); | |
1470 | return 0; | |
1471 | } | |
1472 | #endif | |
1473 | ||
7bc570c8 YH |
1474 | ipv6h = ipv6_hdr(skb); |
1475 | ||
1476 | fl = (struct flowi) { | |
1477 | .oif = vif->link, | |
1478 | .nl_u = { .ip6_u = | |
1479 | { .daddr = ipv6h->daddr, } | |
1480 | } | |
1481 | }; | |
1482 | ||
1483 | dst = ip6_route_output(&init_net, NULL, &fl); | |
1484 | if (!dst) | |
1485 | goto out_free; | |
1486 | ||
1487 | dst_release(skb->dst); | |
1488 | skb->dst = dst; | |
1489 | ||
1490 | /* | |
1491 | * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally | |
1492 | * not only before forwarding, but after forwarding on all output | |
1493 | * interfaces. It is clear, if mrouter runs a multicasting | |
1494 | * program, it should receive packets not depending to what interface | |
1495 | * program is joined. | |
1496 | * If we will not make it, the program will have to join on all | |
1497 | * interfaces. On the other hand, multihoming host (or router, but | |
1498 | * not mrouter) cannot join to more than one interface - it will | |
1499 | * result in receiving multiple packets. | |
1500 | */ | |
1501 | dev = vif->dev; | |
1502 | skb->dev = dev; | |
1503 | vif->pkt_out++; | |
1504 | vif->bytes_out += skb->len; | |
1505 | ||
1506 | /* We are about to write */ | |
1507 | /* XXX: extension headers? */ | |
1508 | if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev))) | |
1509 | goto out_free; | |
1510 | ||
1511 | ipv6h = ipv6_hdr(skb); | |
1512 | ipv6h->hop_limit--; | |
1513 | ||
1514 | IP6CB(skb)->flags |= IP6SKB_FORWARDED; | |
1515 | ||
1516 | return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dev, | |
1517 | ip6mr_forward2_finish); | |
1518 | ||
1519 | out_free: | |
1520 | kfree_skb(skb); | |
1521 | return 0; | |
1522 | } | |
1523 | ||
1524 | static int ip6mr_find_vif(struct net_device *dev) | |
1525 | { | |
1526 | int ct; | |
4e16880c BT |
1527 | for (ct = init_net.ipv6.maxvif - 1; ct >= 0; ct--) { |
1528 | if (init_net.ipv6.vif6_table[ct].dev == dev) | |
7bc570c8 YH |
1529 | break; |
1530 | } | |
1531 | return ct; | |
1532 | } | |
1533 | ||
1534 | static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache) | |
1535 | { | |
1536 | int psend = -1; | |
1537 | int vif, ct; | |
1538 | ||
1539 | vif = cache->mf6c_parent; | |
1540 | cache->mfc_un.res.pkt++; | |
1541 | cache->mfc_un.res.bytes += skb->len; | |
1542 | ||
14fb64e1 YH |
1543 | /* |
1544 | * Wrong interface: drop packet and (maybe) send PIM assert. | |
1545 | */ | |
4e16880c | 1546 | if (init_net.ipv6.vif6_table[vif].dev != skb->dev) { |
14fb64e1 YH |
1547 | int true_vifi; |
1548 | ||
1549 | cache->mfc_un.res.wrong_if++; | |
1550 | true_vifi = ip6mr_find_vif(skb->dev); | |
1551 | ||
a21f3f99 | 1552 | if (true_vifi >= 0 && init_net.ipv6.mroute_do_assert && |
14fb64e1 YH |
1553 | /* pimsm uses asserts, when switching from RPT to SPT, |
1554 | so that we cannot check that packet arrived on an oif. | |
1555 | It is bad, but otherwise we would need to move pretty | |
1556 | large chunk of pimd to kernel. Ough... --ANK | |
1557 | */ | |
a21f3f99 BT |
1558 | (init_net.ipv6.mroute_do_pim || |
1559 | cache->mfc_un.res.ttls[true_vifi] < 255) && | |
14fb64e1 YH |
1560 | time_after(jiffies, |
1561 | cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { | |
1562 | cache->mfc_un.res.last_assert = jiffies; | |
1563 | ip6mr_cache_report(skb, true_vifi, MRT6MSG_WRONGMIF); | |
1564 | } | |
1565 | goto dont_forward; | |
1566 | } | |
1567 | ||
4e16880c BT |
1568 | init_net.ipv6.vif6_table[vif].pkt_in++; |
1569 | init_net.ipv6.vif6_table[vif].bytes_in += skb->len; | |
7bc570c8 YH |
1570 | |
1571 | /* | |
1572 | * Forward the frame | |
1573 | */ | |
1574 | for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) { | |
1575 | if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) { | |
1576 | if (psend != -1) { | |
1577 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1578 | if (skb2) | |
1579 | ip6mr_forward2(skb2, cache, psend); | |
1580 | } | |
1581 | psend = ct; | |
1582 | } | |
1583 | } | |
1584 | if (psend != -1) { | |
1585 | ip6mr_forward2(skb, cache, psend); | |
1586 | return 0; | |
1587 | } | |
1588 | ||
14fb64e1 | 1589 | dont_forward: |
7bc570c8 YH |
1590 | kfree_skb(skb); |
1591 | return 0; | |
1592 | } | |
1593 | ||
1594 | ||
1595 | /* | |
1596 | * Multicast packets for forwarding arrive here | |
1597 | */ | |
1598 | ||
1599 | int ip6_mr_input(struct sk_buff *skb) | |
1600 | { | |
1601 | struct mfc6_cache *cache; | |
1602 | ||
1603 | read_lock(&mrt_lock); | |
1604 | cache = ip6mr_cache_find(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr); | |
1605 | ||
1606 | /* | |
1607 | * No usable cache entry | |
1608 | */ | |
1609 | if (cache == NULL) { | |
1610 | int vif; | |
1611 | ||
1612 | vif = ip6mr_find_vif(skb->dev); | |
1613 | if (vif >= 0) { | |
1614 | int err = ip6mr_cache_unresolved(vif, skb); | |
1615 | read_unlock(&mrt_lock); | |
1616 | ||
1617 | return err; | |
1618 | } | |
1619 | read_unlock(&mrt_lock); | |
1620 | kfree_skb(skb); | |
1621 | return -ENODEV; | |
1622 | } | |
1623 | ||
1624 | ip6_mr_forward(skb, cache); | |
1625 | ||
1626 | read_unlock(&mrt_lock); | |
1627 | ||
1628 | return 0; | |
1629 | } | |
1630 | ||
1631 | ||
1632 | static int | |
1633 | ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm) | |
1634 | { | |
1635 | int ct; | |
1636 | struct rtnexthop *nhp; | |
4e16880c | 1637 | struct net_device *dev = init_net.ipv6.vif6_table[c->mf6c_parent].dev; |
549e028d | 1638 | u8 *b = skb_tail_pointer(skb); |
7bc570c8 YH |
1639 | struct rtattr *mp_head; |
1640 | ||
1641 | if (dev) | |
1642 | RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex); | |
1643 | ||
1644 | mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0)); | |
1645 | ||
1646 | for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { | |
1647 | if (c->mfc_un.res.ttls[ct] < 255) { | |
1648 | if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4)) | |
1649 | goto rtattr_failure; | |
1650 | nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp))); | |
1651 | nhp->rtnh_flags = 0; | |
1652 | nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; | |
4e16880c | 1653 | nhp->rtnh_ifindex = init_net.ipv6.vif6_table[ct].dev->ifindex; |
7bc570c8 YH |
1654 | nhp->rtnh_len = sizeof(*nhp); |
1655 | } | |
1656 | } | |
1657 | mp_head->rta_type = RTA_MULTIPATH; | |
549e028d | 1658 | mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head; |
7bc570c8 YH |
1659 | rtm->rtm_type = RTN_MULTICAST; |
1660 | return 1; | |
1661 | ||
1662 | rtattr_failure: | |
1663 | nlmsg_trim(skb, b); | |
1664 | return -EMSGSIZE; | |
1665 | } | |
1666 | ||
1667 | int ip6mr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait) | |
1668 | { | |
1669 | int err; | |
1670 | struct mfc6_cache *cache; | |
1671 | struct rt6_info *rt = (struct rt6_info *)skb->dst; | |
1672 | ||
1673 | read_lock(&mrt_lock); | |
1674 | cache = ip6mr_cache_find(&rt->rt6i_src.addr, &rt->rt6i_dst.addr); | |
1675 | ||
1676 | if (!cache) { | |
1677 | struct sk_buff *skb2; | |
1678 | struct ipv6hdr *iph; | |
1679 | struct net_device *dev; | |
1680 | int vif; | |
1681 | ||
1682 | if (nowait) { | |
1683 | read_unlock(&mrt_lock); | |
1684 | return -EAGAIN; | |
1685 | } | |
1686 | ||
1687 | dev = skb->dev; | |
1688 | if (dev == NULL || (vif = ip6mr_find_vif(dev)) < 0) { | |
1689 | read_unlock(&mrt_lock); | |
1690 | return -ENODEV; | |
1691 | } | |
1692 | ||
1693 | /* really correct? */ | |
1694 | skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); | |
1695 | if (!skb2) { | |
1696 | read_unlock(&mrt_lock); | |
1697 | return -ENOMEM; | |
1698 | } | |
1699 | ||
1700 | skb_reset_transport_header(skb2); | |
1701 | ||
1702 | skb_put(skb2, sizeof(struct ipv6hdr)); | |
1703 | skb_reset_network_header(skb2); | |
1704 | ||
1705 | iph = ipv6_hdr(skb2); | |
1706 | iph->version = 0; | |
1707 | iph->priority = 0; | |
1708 | iph->flow_lbl[0] = 0; | |
1709 | iph->flow_lbl[1] = 0; | |
1710 | iph->flow_lbl[2] = 0; | |
1711 | iph->payload_len = 0; | |
1712 | iph->nexthdr = IPPROTO_NONE; | |
1713 | iph->hop_limit = 0; | |
1714 | ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr); | |
1715 | ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr); | |
1716 | ||
1717 | err = ip6mr_cache_unresolved(vif, skb2); | |
1718 | read_unlock(&mrt_lock); | |
1719 | ||
1720 | return err; | |
1721 | } | |
1722 | ||
1723 | if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY)) | |
1724 | cache->mfc_flags |= MFC_NOTIFY; | |
1725 | ||
1726 | err = ip6mr_fill_mroute(skb, cache, rtm); | |
1727 | read_unlock(&mrt_lock); | |
1728 | return err; | |
1729 | } | |
1730 |