]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * IP multicast routing support for mrouted 3.6/3.8 | |
3 | * | |
113aa838 | 4 | * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk> |
1da177e4 LT |
5 | * Linux Consultancy and Custom Driver Development |
6 | * | |
7 | * This program is free software; you can redistribute it and/or | |
8 | * modify it under the terms of the GNU General Public License | |
9 | * as published by the Free Software Foundation; either version | |
10 | * 2 of the License, or (at your option) any later version. | |
11 | * | |
1da177e4 LT |
12 | * Fixes: |
13 | * Michael Chastain : Incorrect size of copying. | |
14 | * Alan Cox : Added the cache manager code | |
15 | * Alan Cox : Fixed the clone/copy bug and device race. | |
16 | * Mike McLagan : Routing by source | |
17 | * Malcolm Beattie : Buffer handling fixes. | |
18 | * Alexey Kuznetsov : Double buffer free and other fixes. | |
19 | * SVR Anand : Fixed several multicast bugs and problems. | |
20 | * Alexey Kuznetsov : Status, optimisations and more. | |
21 | * Brad Parker : Better behaviour on mrouted upcall | |
22 | * overflow. | |
23 | * Carlos Picoto : PIMv1 Support | |
24 | * Pavlin Ivanov Radoslavov: PIMv2 Registers must checksum only PIM header | |
25 | * Relax this requrement to work with older peers. | |
26 | * | |
27 | */ | |
28 | ||
1da177e4 LT |
29 | #include <asm/system.h> |
30 | #include <asm/uaccess.h> | |
31 | #include <linux/types.h> | |
4fc268d2 | 32 | #include <linux/capability.h> |
1da177e4 LT |
33 | #include <linux/errno.h> |
34 | #include <linux/timer.h> | |
35 | #include <linux/mm.h> | |
36 | #include <linux/kernel.h> | |
37 | #include <linux/fcntl.h> | |
38 | #include <linux/stat.h> | |
39 | #include <linux/socket.h> | |
40 | #include <linux/in.h> | |
41 | #include <linux/inet.h> | |
42 | #include <linux/netdevice.h> | |
43 | #include <linux/inetdevice.h> | |
44 | #include <linux/igmp.h> | |
45 | #include <linux/proc_fs.h> | |
46 | #include <linux/seq_file.h> | |
47 | #include <linux/mroute.h> | |
48 | #include <linux/init.h> | |
46f25dff | 49 | #include <linux/if_ether.h> |
457c4cbc | 50 | #include <net/net_namespace.h> |
1da177e4 LT |
51 | #include <net/ip.h> |
52 | #include <net/protocol.h> | |
53 | #include <linux/skbuff.h> | |
14c85021 | 54 | #include <net/route.h> |
1da177e4 LT |
55 | #include <net/sock.h> |
56 | #include <net/icmp.h> | |
57 | #include <net/udp.h> | |
58 | #include <net/raw.h> | |
59 | #include <linux/notifier.h> | |
60 | #include <linux/if_arp.h> | |
61 | #include <linux/netfilter_ipv4.h> | |
62 | #include <net/ipip.h> | |
63 | #include <net/checksum.h> | |
dc5fc579 | 64 | #include <net/netlink.h> |
1da177e4 LT |
65 | |
66 | #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2) | |
67 | #define CONFIG_IP_PIMSM 1 | |
68 | #endif | |
69 | ||
1da177e4 LT |
70 | /* Big lock, protecting vif table, mrt cache and mroute socket state. |
71 | Note that the changes are semaphored via rtnl_lock. | |
72 | */ | |
73 | ||
74 | static DEFINE_RWLOCK(mrt_lock); | |
75 | ||
76 | /* | |
77 | * Multicast router control variables | |
78 | */ | |
79 | ||
cf958ae3 | 80 | #define VIF_EXISTS(_net, _idx) ((_net)->ipv4.vif_table[_idx].dev != NULL) |
1da177e4 LT |
81 | |
82 | static int mroute_do_assert; /* Set in PIM assert */ | |
83 | static int mroute_do_pim; | |
84 | ||
85 | static struct mfc_cache *mfc_cache_array[MFC_LINES]; /* Forwarding cache */ | |
86 | ||
87 | static struct mfc_cache *mfc_unres_queue; /* Queue of unresolved entries */ | |
88 | static atomic_t cache_resolve_queue_len; /* Size of unresolved */ | |
89 | ||
90 | /* Special spinlock for queue of unresolved entries */ | |
91 | static DEFINE_SPINLOCK(mfc_unres_lock); | |
92 | ||
93 | /* We return to original Alan's scheme. Hash table of resolved | |
94 | entries is changed only in process context and protected | |
95 | with weak lock mrt_lock. Queue of unresolved entries is protected | |
96 | with strong spinlock mfc_unres_lock. | |
97 | ||
98 | In this case data path is free of exclusive locks at all. | |
99 | */ | |
100 | ||
e18b890b | 101 | static struct kmem_cache *mrt_cachep __read_mostly; |
1da177e4 LT |
102 | |
103 | static int ip_mr_forward(struct sk_buff *skb, struct mfc_cache *cache, int local); | |
104 | static int ipmr_cache_report(struct sk_buff *pkt, vifi_t vifi, int assert); | |
105 | static int ipmr_fill_mroute(struct sk_buff *skb, struct mfc_cache *c, struct rtmsg *rtm); | |
106 | ||
107 | #ifdef CONFIG_IP_PIMSM_V2 | |
108 | static struct net_protocol pim_protocol; | |
109 | #endif | |
110 | ||
111 | static struct timer_list ipmr_expire_timer; | |
112 | ||
113 | /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */ | |
114 | ||
d607032d WC |
115 | static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v) |
116 | { | |
117 | dev_close(dev); | |
118 | ||
119 | dev = __dev_get_by_name(&init_net, "tunl0"); | |
120 | if (dev) { | |
5bc3eb7e | 121 | const struct net_device_ops *ops = dev->netdev_ops; |
d607032d | 122 | struct ifreq ifr; |
d607032d WC |
123 | struct ip_tunnel_parm p; |
124 | ||
125 | memset(&p, 0, sizeof(p)); | |
126 | p.iph.daddr = v->vifc_rmt_addr.s_addr; | |
127 | p.iph.saddr = v->vifc_lcl_addr.s_addr; | |
128 | p.iph.version = 4; | |
129 | p.iph.ihl = 5; | |
130 | p.iph.protocol = IPPROTO_IPIP; | |
131 | sprintf(p.name, "dvmrp%d", v->vifc_vifi); | |
132 | ifr.ifr_ifru.ifru_data = (__force void __user *)&p; | |
133 | ||
5bc3eb7e SH |
134 | if (ops->ndo_do_ioctl) { |
135 | mm_segment_t oldfs = get_fs(); | |
136 | ||
137 | set_fs(KERNEL_DS); | |
138 | ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL); | |
139 | set_fs(oldfs); | |
140 | } | |
d607032d WC |
141 | } |
142 | } | |
143 | ||
1da177e4 LT |
144 | static |
145 | struct net_device *ipmr_new_tunnel(struct vifctl *v) | |
146 | { | |
147 | struct net_device *dev; | |
148 | ||
881d966b | 149 | dev = __dev_get_by_name(&init_net, "tunl0"); |
1da177e4 LT |
150 | |
151 | if (dev) { | |
5bc3eb7e | 152 | const struct net_device_ops *ops = dev->netdev_ops; |
1da177e4 LT |
153 | int err; |
154 | struct ifreq ifr; | |
1da177e4 LT |
155 | struct ip_tunnel_parm p; |
156 | struct in_device *in_dev; | |
157 | ||
158 | memset(&p, 0, sizeof(p)); | |
159 | p.iph.daddr = v->vifc_rmt_addr.s_addr; | |
160 | p.iph.saddr = v->vifc_lcl_addr.s_addr; | |
161 | p.iph.version = 4; | |
162 | p.iph.ihl = 5; | |
163 | p.iph.protocol = IPPROTO_IPIP; | |
164 | sprintf(p.name, "dvmrp%d", v->vifc_vifi); | |
ba93ef74 | 165 | ifr.ifr_ifru.ifru_data = (__force void __user *)&p; |
1da177e4 | 166 | |
5bc3eb7e SH |
167 | if (ops->ndo_do_ioctl) { |
168 | mm_segment_t oldfs = get_fs(); | |
169 | ||
170 | set_fs(KERNEL_DS); | |
171 | err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL); | |
172 | set_fs(oldfs); | |
173 | } else | |
174 | err = -EOPNOTSUPP; | |
1da177e4 LT |
175 | |
176 | dev = NULL; | |
177 | ||
881d966b | 178 | if (err == 0 && (dev = __dev_get_by_name(&init_net, p.name)) != NULL) { |
1da177e4 LT |
179 | dev->flags |= IFF_MULTICAST; |
180 | ||
e5ed6399 | 181 | in_dev = __in_dev_get_rtnl(dev); |
71e27da9 | 182 | if (in_dev == NULL) |
1da177e4 | 183 | goto failure; |
71e27da9 HX |
184 | |
185 | ipv4_devconf_setall(in_dev); | |
186 | IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0; | |
1da177e4 LT |
187 | |
188 | if (dev_open(dev)) | |
189 | goto failure; | |
7dc00c82 | 190 | dev_hold(dev); |
1da177e4 LT |
191 | } |
192 | } | |
193 | return dev; | |
194 | ||
195 | failure: | |
196 | /* allow the register to be completed before unregistering. */ | |
197 | rtnl_unlock(); | |
198 | rtnl_lock(); | |
199 | ||
200 | unregister_netdevice(dev); | |
201 | return NULL; | |
202 | } | |
203 | ||
204 | #ifdef CONFIG_IP_PIMSM | |
205 | ||
206 | static int reg_vif_num = -1; | |
207 | ||
208 | static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) | |
209 | { | |
210 | read_lock(&mrt_lock); | |
cf3677ae PE |
211 | dev->stats.tx_bytes += skb->len; |
212 | dev->stats.tx_packets++; | |
1da177e4 LT |
213 | ipmr_cache_report(skb, reg_vif_num, IGMPMSG_WHOLEPKT); |
214 | read_unlock(&mrt_lock); | |
215 | kfree_skb(skb); | |
216 | return 0; | |
217 | } | |
218 | ||
007c3838 SH |
219 | static const struct net_device_ops reg_vif_netdev_ops = { |
220 | .ndo_start_xmit = reg_vif_xmit, | |
221 | }; | |
222 | ||
1da177e4 LT |
223 | static void reg_vif_setup(struct net_device *dev) |
224 | { | |
225 | dev->type = ARPHRD_PIMREG; | |
46f25dff | 226 | dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8; |
1da177e4 | 227 | dev->flags = IFF_NOARP; |
007c3838 | 228 | dev->netdev_ops = ®_vif_netdev_ops, |
1da177e4 LT |
229 | dev->destructor = free_netdev; |
230 | } | |
231 | ||
232 | static struct net_device *ipmr_reg_vif(void) | |
233 | { | |
234 | struct net_device *dev; | |
235 | struct in_device *in_dev; | |
236 | ||
cf3677ae | 237 | dev = alloc_netdev(0, "pimreg", reg_vif_setup); |
1da177e4 LT |
238 | |
239 | if (dev == NULL) | |
240 | return NULL; | |
241 | ||
242 | if (register_netdevice(dev)) { | |
243 | free_netdev(dev); | |
244 | return NULL; | |
245 | } | |
246 | dev->iflink = 0; | |
247 | ||
71e27da9 HX |
248 | rcu_read_lock(); |
249 | if ((in_dev = __in_dev_get_rcu(dev)) == NULL) { | |
250 | rcu_read_unlock(); | |
1da177e4 | 251 | goto failure; |
71e27da9 | 252 | } |
1da177e4 | 253 | |
71e27da9 HX |
254 | ipv4_devconf_setall(in_dev); |
255 | IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0; | |
256 | rcu_read_unlock(); | |
1da177e4 LT |
257 | |
258 | if (dev_open(dev)) | |
259 | goto failure; | |
260 | ||
7dc00c82 WC |
261 | dev_hold(dev); |
262 | ||
1da177e4 LT |
263 | return dev; |
264 | ||
265 | failure: | |
266 | /* allow the register to be completed before unregistering. */ | |
267 | rtnl_unlock(); | |
268 | rtnl_lock(); | |
269 | ||
270 | unregister_netdevice(dev); | |
271 | return NULL; | |
272 | } | |
273 | #endif | |
274 | ||
275 | /* | |
276 | * Delete a VIF entry | |
7dc00c82 | 277 | * @notify: Set to 1, if the caller is a notifier_call |
1da177e4 | 278 | */ |
e905a9ed | 279 | |
7dc00c82 | 280 | static int vif_delete(int vifi, int notify) |
1da177e4 LT |
281 | { |
282 | struct vif_device *v; | |
283 | struct net_device *dev; | |
284 | struct in_device *in_dev; | |
285 | ||
cf958ae3 | 286 | if (vifi < 0 || vifi >= init_net.ipv4.maxvif) |
1da177e4 LT |
287 | return -EADDRNOTAVAIL; |
288 | ||
cf958ae3 | 289 | v = &init_net.ipv4.vif_table[vifi]; |
1da177e4 LT |
290 | |
291 | write_lock_bh(&mrt_lock); | |
292 | dev = v->dev; | |
293 | v->dev = NULL; | |
294 | ||
295 | if (!dev) { | |
296 | write_unlock_bh(&mrt_lock); | |
297 | return -EADDRNOTAVAIL; | |
298 | } | |
299 | ||
300 | #ifdef CONFIG_IP_PIMSM | |
301 | if (vifi == reg_vif_num) | |
302 | reg_vif_num = -1; | |
303 | #endif | |
304 | ||
cf958ae3 | 305 | if (vifi+1 == init_net.ipv4.maxvif) { |
1da177e4 LT |
306 | int tmp; |
307 | for (tmp=vifi-1; tmp>=0; tmp--) { | |
cf958ae3 | 308 | if (VIF_EXISTS(&init_net, tmp)) |
1da177e4 LT |
309 | break; |
310 | } | |
cf958ae3 | 311 | init_net.ipv4.maxvif = tmp+1; |
1da177e4 LT |
312 | } |
313 | ||
314 | write_unlock_bh(&mrt_lock); | |
315 | ||
316 | dev_set_allmulti(dev, -1); | |
317 | ||
e5ed6399 | 318 | if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) { |
42f811b8 | 319 | IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--; |
1da177e4 LT |
320 | ip_rt_multicast_event(in_dev); |
321 | } | |
322 | ||
7dc00c82 | 323 | if (v->flags&(VIFF_TUNNEL|VIFF_REGISTER) && !notify) |
1da177e4 LT |
324 | unregister_netdevice(dev); |
325 | ||
326 | dev_put(dev); | |
327 | return 0; | |
328 | } | |
329 | ||
330 | /* Destroy an unresolved cache entry, killing queued skbs | |
331 | and reporting error to netlink readers. | |
332 | */ | |
333 | ||
334 | static void ipmr_destroy_unres(struct mfc_cache *c) | |
335 | { | |
336 | struct sk_buff *skb; | |
9ef1d4c7 | 337 | struct nlmsgerr *e; |
1da177e4 LT |
338 | |
339 | atomic_dec(&cache_resolve_queue_len); | |
340 | ||
c354e124 | 341 | while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) { |
eddc9ec5 | 342 | if (ip_hdr(skb)->version == 0) { |
1da177e4 LT |
343 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr)); |
344 | nlh->nlmsg_type = NLMSG_ERROR; | |
345 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
346 | skb_trim(skb, nlh->nlmsg_len); | |
9ef1d4c7 PM |
347 | e = NLMSG_DATA(nlh); |
348 | e->error = -ETIMEDOUT; | |
349 | memset(&e->msg, 0, sizeof(e->msg)); | |
2942e900 | 350 | |
97c53cac | 351 | rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); |
1da177e4 LT |
352 | } else |
353 | kfree_skb(skb); | |
354 | } | |
355 | ||
356 | kmem_cache_free(mrt_cachep, c); | |
357 | } | |
358 | ||
359 | ||
360 | /* Single timer process for all the unresolved queue. */ | |
361 | ||
362 | static void ipmr_expire_process(unsigned long dummy) | |
363 | { | |
364 | unsigned long now; | |
365 | unsigned long expires; | |
366 | struct mfc_cache *c, **cp; | |
367 | ||
368 | if (!spin_trylock(&mfc_unres_lock)) { | |
369 | mod_timer(&ipmr_expire_timer, jiffies+HZ/10); | |
370 | return; | |
371 | } | |
372 | ||
373 | if (atomic_read(&cache_resolve_queue_len) == 0) | |
374 | goto out; | |
375 | ||
376 | now = jiffies; | |
377 | expires = 10*HZ; | |
378 | cp = &mfc_unres_queue; | |
379 | ||
380 | while ((c=*cp) != NULL) { | |
381 | if (time_after(c->mfc_un.unres.expires, now)) { | |
382 | unsigned long interval = c->mfc_un.unres.expires - now; | |
383 | if (interval < expires) | |
384 | expires = interval; | |
385 | cp = &c->next; | |
386 | continue; | |
387 | } | |
388 | ||
389 | *cp = c->next; | |
390 | ||
391 | ipmr_destroy_unres(c); | |
392 | } | |
393 | ||
394 | if (atomic_read(&cache_resolve_queue_len)) | |
395 | mod_timer(&ipmr_expire_timer, jiffies + expires); | |
396 | ||
397 | out: | |
398 | spin_unlock(&mfc_unres_lock); | |
399 | } | |
400 | ||
401 | /* Fill oifs list. It is called under write locked mrt_lock. */ | |
402 | ||
d1b04c08 | 403 | static void ipmr_update_thresholds(struct mfc_cache *cache, unsigned char *ttls) |
1da177e4 LT |
404 | { |
405 | int vifi; | |
406 | ||
407 | cache->mfc_un.res.minvif = MAXVIFS; | |
408 | cache->mfc_un.res.maxvif = 0; | |
409 | memset(cache->mfc_un.res.ttls, 255, MAXVIFS); | |
410 | ||
cf958ae3 BT |
411 | for (vifi = 0; vifi < init_net.ipv4.maxvif; vifi++) { |
412 | if (VIF_EXISTS(&init_net, vifi) && | |
413 | ttls[vifi] && ttls[vifi] < 255) { | |
1da177e4 LT |
414 | cache->mfc_un.res.ttls[vifi] = ttls[vifi]; |
415 | if (cache->mfc_un.res.minvif > vifi) | |
416 | cache->mfc_un.res.minvif = vifi; | |
417 | if (cache->mfc_un.res.maxvif <= vifi) | |
418 | cache->mfc_un.res.maxvif = vifi + 1; | |
419 | } | |
420 | } | |
421 | } | |
422 | ||
423 | static int vif_add(struct vifctl *vifc, int mrtsock) | |
424 | { | |
425 | int vifi = vifc->vifc_vifi; | |
cf958ae3 | 426 | struct vif_device *v = &init_net.ipv4.vif_table[vifi]; |
1da177e4 LT |
427 | struct net_device *dev; |
428 | struct in_device *in_dev; | |
d607032d | 429 | int err; |
1da177e4 LT |
430 | |
431 | /* Is vif busy ? */ | |
cf958ae3 | 432 | if (VIF_EXISTS(&init_net, vifi)) |
1da177e4 LT |
433 | return -EADDRINUSE; |
434 | ||
435 | switch (vifc->vifc_flags) { | |
436 | #ifdef CONFIG_IP_PIMSM | |
437 | case VIFF_REGISTER: | |
438 | /* | |
439 | * Special Purpose VIF in PIM | |
440 | * All the packets will be sent to the daemon | |
441 | */ | |
442 | if (reg_vif_num >= 0) | |
443 | return -EADDRINUSE; | |
444 | dev = ipmr_reg_vif(); | |
445 | if (!dev) | |
446 | return -ENOBUFS; | |
d607032d WC |
447 | err = dev_set_allmulti(dev, 1); |
448 | if (err) { | |
449 | unregister_netdevice(dev); | |
7dc00c82 | 450 | dev_put(dev); |
d607032d WC |
451 | return err; |
452 | } | |
1da177e4 LT |
453 | break; |
454 | #endif | |
e905a9ed | 455 | case VIFF_TUNNEL: |
1da177e4 LT |
456 | dev = ipmr_new_tunnel(vifc); |
457 | if (!dev) | |
458 | return -ENOBUFS; | |
d607032d WC |
459 | err = dev_set_allmulti(dev, 1); |
460 | if (err) { | |
461 | ipmr_del_tunnel(dev, vifc); | |
7dc00c82 | 462 | dev_put(dev); |
d607032d WC |
463 | return err; |
464 | } | |
1da177e4 LT |
465 | break; |
466 | case 0: | |
1ab35276 | 467 | dev = ip_dev_find(&init_net, vifc->vifc_lcl_addr.s_addr); |
1da177e4 LT |
468 | if (!dev) |
469 | return -EADDRNOTAVAIL; | |
d607032d | 470 | err = dev_set_allmulti(dev, 1); |
7dc00c82 WC |
471 | if (err) { |
472 | dev_put(dev); | |
d607032d | 473 | return err; |
7dc00c82 | 474 | } |
1da177e4 LT |
475 | break; |
476 | default: | |
477 | return -EINVAL; | |
478 | } | |
479 | ||
e5ed6399 | 480 | if ((in_dev = __in_dev_get_rtnl(dev)) == NULL) |
1da177e4 | 481 | return -EADDRNOTAVAIL; |
42f811b8 | 482 | IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++; |
1da177e4 LT |
483 | ip_rt_multicast_event(in_dev); |
484 | ||
485 | /* | |
486 | * Fill in the VIF structures | |
487 | */ | |
c354e124 JK |
488 | v->rate_limit = vifc->vifc_rate_limit; |
489 | v->local = vifc->vifc_lcl_addr.s_addr; | |
490 | v->remote = vifc->vifc_rmt_addr.s_addr; | |
491 | v->flags = vifc->vifc_flags; | |
1da177e4 LT |
492 | if (!mrtsock) |
493 | v->flags |= VIFF_STATIC; | |
c354e124 | 494 | v->threshold = vifc->vifc_threshold; |
1da177e4 LT |
495 | v->bytes_in = 0; |
496 | v->bytes_out = 0; | |
497 | v->pkt_in = 0; | |
498 | v->pkt_out = 0; | |
499 | v->link = dev->ifindex; | |
500 | if (v->flags&(VIFF_TUNNEL|VIFF_REGISTER)) | |
501 | v->link = dev->iflink; | |
502 | ||
503 | /* And finish update writing critical data */ | |
504 | write_lock_bh(&mrt_lock); | |
c354e124 | 505 | v->dev = dev; |
1da177e4 LT |
506 | #ifdef CONFIG_IP_PIMSM |
507 | if (v->flags&VIFF_REGISTER) | |
508 | reg_vif_num = vifi; | |
509 | #endif | |
cf958ae3 BT |
510 | if (vifi+1 > init_net.ipv4.maxvif) |
511 | init_net.ipv4.maxvif = vifi+1; | |
1da177e4 LT |
512 | write_unlock_bh(&mrt_lock); |
513 | return 0; | |
514 | } | |
515 | ||
114c7844 | 516 | static struct mfc_cache *ipmr_cache_find(__be32 origin, __be32 mcastgrp) |
1da177e4 | 517 | { |
c354e124 | 518 | int line = MFC_HASH(mcastgrp, origin); |
1da177e4 LT |
519 | struct mfc_cache *c; |
520 | ||
521 | for (c=mfc_cache_array[line]; c; c = c->next) { | |
522 | if (c->mfc_origin==origin && c->mfc_mcastgrp==mcastgrp) | |
523 | break; | |
524 | } | |
525 | return c; | |
526 | } | |
527 | ||
528 | /* | |
529 | * Allocate a multicast cache entry | |
530 | */ | |
531 | static struct mfc_cache *ipmr_cache_alloc(void) | |
532 | { | |
c354e124 JK |
533 | struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); |
534 | if (c == NULL) | |
1da177e4 | 535 | return NULL; |
1da177e4 LT |
536 | c->mfc_un.res.minvif = MAXVIFS; |
537 | return c; | |
538 | } | |
539 | ||
540 | static struct mfc_cache *ipmr_cache_alloc_unres(void) | |
541 | { | |
c354e124 JK |
542 | struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); |
543 | if (c == NULL) | |
1da177e4 | 544 | return NULL; |
1da177e4 LT |
545 | skb_queue_head_init(&c->mfc_un.unres.unresolved); |
546 | c->mfc_un.unres.expires = jiffies + 10*HZ; | |
547 | return c; | |
548 | } | |
549 | ||
550 | /* | |
551 | * A cache entry has gone into a resolved state from queued | |
552 | */ | |
e905a9ed | 553 | |
1da177e4 LT |
554 | static void ipmr_cache_resolve(struct mfc_cache *uc, struct mfc_cache *c) |
555 | { | |
556 | struct sk_buff *skb; | |
9ef1d4c7 | 557 | struct nlmsgerr *e; |
1da177e4 LT |
558 | |
559 | /* | |
560 | * Play the pending entries through our router | |
561 | */ | |
562 | ||
c354e124 | 563 | while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { |
eddc9ec5 | 564 | if (ip_hdr(skb)->version == 0) { |
1da177e4 LT |
565 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr)); |
566 | ||
567 | if (ipmr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) { | |
27a884dc ACM |
568 | nlh->nlmsg_len = (skb_tail_pointer(skb) - |
569 | (u8 *)nlh); | |
1da177e4 LT |
570 | } else { |
571 | nlh->nlmsg_type = NLMSG_ERROR; | |
572 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
573 | skb_trim(skb, nlh->nlmsg_len); | |
9ef1d4c7 PM |
574 | e = NLMSG_DATA(nlh); |
575 | e->error = -EMSGSIZE; | |
576 | memset(&e->msg, 0, sizeof(e->msg)); | |
1da177e4 | 577 | } |
2942e900 | 578 | |
97c53cac | 579 | rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); |
1da177e4 LT |
580 | } else |
581 | ip_mr_forward(skb, c, 0); | |
582 | } | |
583 | } | |
584 | ||
585 | /* | |
586 | * Bounce a cache query up to mrouted. We could use netlink for this but mrouted | |
587 | * expects the following bizarre scheme. | |
588 | * | |
589 | * Called under mrt_lock. | |
590 | */ | |
e905a9ed | 591 | |
1da177e4 LT |
592 | static int ipmr_cache_report(struct sk_buff *pkt, vifi_t vifi, int assert) |
593 | { | |
594 | struct sk_buff *skb; | |
c9bdd4b5 | 595 | const int ihl = ip_hdrlen(pkt); |
1da177e4 LT |
596 | struct igmphdr *igmp; |
597 | struct igmpmsg *msg; | |
598 | int ret; | |
599 | ||
600 | #ifdef CONFIG_IP_PIMSM | |
601 | if (assert == IGMPMSG_WHOLEPKT) | |
602 | skb = skb_realloc_headroom(pkt, sizeof(struct iphdr)); | |
603 | else | |
604 | #endif | |
605 | skb = alloc_skb(128, GFP_ATOMIC); | |
606 | ||
132adf54 | 607 | if (!skb) |
1da177e4 LT |
608 | return -ENOBUFS; |
609 | ||
610 | #ifdef CONFIG_IP_PIMSM | |
611 | if (assert == IGMPMSG_WHOLEPKT) { | |
612 | /* Ugly, but we have no choice with this interface. | |
613 | Duplicate old header, fix ihl, length etc. | |
614 | And all this only to mangle msg->im_msgtype and | |
615 | to set msg->im_mbz to "mbz" :-) | |
616 | */ | |
878c8145 ACM |
617 | skb_push(skb, sizeof(struct iphdr)); |
618 | skb_reset_network_header(skb); | |
badff6d0 | 619 | skb_reset_transport_header(skb); |
0272ffc4 | 620 | msg = (struct igmpmsg *)skb_network_header(skb); |
d56f90a7 | 621 | memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr)); |
1da177e4 LT |
622 | msg->im_msgtype = IGMPMSG_WHOLEPKT; |
623 | msg->im_mbz = 0; | |
e905a9ed | 624 | msg->im_vif = reg_vif_num; |
eddc9ec5 ACM |
625 | ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2; |
626 | ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) + | |
627 | sizeof(struct iphdr)); | |
e905a9ed | 628 | } else |
1da177e4 | 629 | #endif |
e905a9ed YH |
630 | { |
631 | ||
1da177e4 LT |
632 | /* |
633 | * Copy the IP header | |
634 | */ | |
635 | ||
27a884dc | 636 | skb->network_header = skb->tail; |
ddc7b8e3 | 637 | skb_put(skb, ihl); |
27d7ff46 | 638 | skb_copy_to_linear_data(skb, pkt->data, ihl); |
eddc9ec5 ACM |
639 | ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */ |
640 | msg = (struct igmpmsg *)skb_network_header(skb); | |
1da177e4 LT |
641 | msg->im_vif = vifi; |
642 | skb->dst = dst_clone(pkt->dst); | |
643 | ||
644 | /* | |
645 | * Add our header | |
646 | */ | |
647 | ||
c354e124 | 648 | igmp=(struct igmphdr *)skb_put(skb, sizeof(struct igmphdr)); |
1da177e4 LT |
649 | igmp->type = |
650 | msg->im_msgtype = assert; | |
651 | igmp->code = 0; | |
eddc9ec5 | 652 | ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */ |
b0e380b1 | 653 | skb->transport_header = skb->network_header; |
e905a9ed | 654 | } |
1da177e4 | 655 | |
70a269e6 | 656 | if (init_net.ipv4.mroute_sk == NULL) { |
1da177e4 LT |
657 | kfree_skb(skb); |
658 | return -EINVAL; | |
659 | } | |
660 | ||
661 | /* | |
662 | * Deliver to mrouted | |
663 | */ | |
70a269e6 BT |
664 | ret = sock_queue_rcv_skb(init_net.ipv4.mroute_sk, skb); |
665 | if (ret < 0) { | |
1da177e4 LT |
666 | if (net_ratelimit()) |
667 | printk(KERN_WARNING "mroute: pending queue full, dropping entries.\n"); | |
668 | kfree_skb(skb); | |
669 | } | |
670 | ||
671 | return ret; | |
672 | } | |
673 | ||
674 | /* | |
675 | * Queue a packet for resolution. It gets locked cache entry! | |
676 | */ | |
e905a9ed | 677 | |
1da177e4 LT |
678 | static int |
679 | ipmr_cache_unresolved(vifi_t vifi, struct sk_buff *skb) | |
680 | { | |
681 | int err; | |
682 | struct mfc_cache *c; | |
eddc9ec5 | 683 | const struct iphdr *iph = ip_hdr(skb); |
1da177e4 LT |
684 | |
685 | spin_lock_bh(&mfc_unres_lock); | |
686 | for (c=mfc_unres_queue; c; c=c->next) { | |
eddc9ec5 ACM |
687 | if (c->mfc_mcastgrp == iph->daddr && |
688 | c->mfc_origin == iph->saddr) | |
1da177e4 LT |
689 | break; |
690 | } | |
691 | ||
692 | if (c == NULL) { | |
693 | /* | |
694 | * Create a new entry if allowable | |
695 | */ | |
696 | ||
c354e124 | 697 | if (atomic_read(&cache_resolve_queue_len) >= 10 || |
1da177e4 LT |
698 | (c=ipmr_cache_alloc_unres())==NULL) { |
699 | spin_unlock_bh(&mfc_unres_lock); | |
700 | ||
701 | kfree_skb(skb); | |
702 | return -ENOBUFS; | |
703 | } | |
704 | ||
705 | /* | |
706 | * Fill in the new cache entry | |
707 | */ | |
eddc9ec5 ACM |
708 | c->mfc_parent = -1; |
709 | c->mfc_origin = iph->saddr; | |
710 | c->mfc_mcastgrp = iph->daddr; | |
1da177e4 LT |
711 | |
712 | /* | |
713 | * Reflect first query at mrouted. | |
714 | */ | |
715 | if ((err = ipmr_cache_report(skb, vifi, IGMPMSG_NOCACHE))<0) { | |
e905a9ed | 716 | /* If the report failed throw the cache entry |
1da177e4 LT |
717 | out - Brad Parker |
718 | */ | |
719 | spin_unlock_bh(&mfc_unres_lock); | |
720 | ||
721 | kmem_cache_free(mrt_cachep, c); | |
722 | kfree_skb(skb); | |
723 | return err; | |
724 | } | |
725 | ||
726 | atomic_inc(&cache_resolve_queue_len); | |
727 | c->next = mfc_unres_queue; | |
728 | mfc_unres_queue = c; | |
729 | ||
730 | mod_timer(&ipmr_expire_timer, c->mfc_un.unres.expires); | |
731 | } | |
732 | ||
733 | /* | |
734 | * See if we can append the packet | |
735 | */ | |
736 | if (c->mfc_un.unres.unresolved.qlen>3) { | |
737 | kfree_skb(skb); | |
738 | err = -ENOBUFS; | |
739 | } else { | |
c354e124 | 740 | skb_queue_tail(&c->mfc_un.unres.unresolved, skb); |
1da177e4 LT |
741 | err = 0; |
742 | } | |
743 | ||
744 | spin_unlock_bh(&mfc_unres_lock); | |
745 | return err; | |
746 | } | |
747 | ||
748 | /* | |
749 | * MFC cache manipulation by user space mroute daemon | |
750 | */ | |
751 | ||
752 | static int ipmr_mfc_delete(struct mfcctl *mfc) | |
753 | { | |
754 | int line; | |
755 | struct mfc_cache *c, **cp; | |
756 | ||
c354e124 | 757 | line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr); |
1da177e4 LT |
758 | |
759 | for (cp=&mfc_cache_array[line]; (c=*cp) != NULL; cp = &c->next) { | |
760 | if (c->mfc_origin == mfc->mfcc_origin.s_addr && | |
761 | c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr) { | |
762 | write_lock_bh(&mrt_lock); | |
763 | *cp = c->next; | |
764 | write_unlock_bh(&mrt_lock); | |
765 | ||
766 | kmem_cache_free(mrt_cachep, c); | |
767 | return 0; | |
768 | } | |
769 | } | |
770 | return -ENOENT; | |
771 | } | |
772 | ||
773 | static int ipmr_mfc_add(struct mfcctl *mfc, int mrtsock) | |
774 | { | |
775 | int line; | |
776 | struct mfc_cache *uc, *c, **cp; | |
777 | ||
c354e124 | 778 | line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr); |
1da177e4 LT |
779 | |
780 | for (cp=&mfc_cache_array[line]; (c=*cp) != NULL; cp = &c->next) { | |
781 | if (c->mfc_origin == mfc->mfcc_origin.s_addr && | |
782 | c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr) | |
783 | break; | |
784 | } | |
785 | ||
786 | if (c != NULL) { | |
787 | write_lock_bh(&mrt_lock); | |
788 | c->mfc_parent = mfc->mfcc_parent; | |
d1b04c08 | 789 | ipmr_update_thresholds(c, mfc->mfcc_ttls); |
1da177e4 LT |
790 | if (!mrtsock) |
791 | c->mfc_flags |= MFC_STATIC; | |
792 | write_unlock_bh(&mrt_lock); | |
793 | return 0; | |
794 | } | |
795 | ||
f97c1e0c | 796 | if (!ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr)) |
1da177e4 LT |
797 | return -EINVAL; |
798 | ||
c354e124 JK |
799 | c = ipmr_cache_alloc(); |
800 | if (c == NULL) | |
1da177e4 LT |
801 | return -ENOMEM; |
802 | ||
c354e124 JK |
803 | c->mfc_origin = mfc->mfcc_origin.s_addr; |
804 | c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr; | |
805 | c->mfc_parent = mfc->mfcc_parent; | |
d1b04c08 | 806 | ipmr_update_thresholds(c, mfc->mfcc_ttls); |
1da177e4 LT |
807 | if (!mrtsock) |
808 | c->mfc_flags |= MFC_STATIC; | |
809 | ||
810 | write_lock_bh(&mrt_lock); | |
811 | c->next = mfc_cache_array[line]; | |
812 | mfc_cache_array[line] = c; | |
813 | write_unlock_bh(&mrt_lock); | |
814 | ||
815 | /* | |
816 | * Check to see if we resolved a queued list. If so we | |
817 | * need to send on the frames and tidy up. | |
818 | */ | |
819 | spin_lock_bh(&mfc_unres_lock); | |
820 | for (cp = &mfc_unres_queue; (uc=*cp) != NULL; | |
821 | cp = &uc->next) { | |
822 | if (uc->mfc_origin == c->mfc_origin && | |
823 | uc->mfc_mcastgrp == c->mfc_mcastgrp) { | |
824 | *cp = uc->next; | |
825 | if (atomic_dec_and_test(&cache_resolve_queue_len)) | |
826 | del_timer(&ipmr_expire_timer); | |
827 | break; | |
828 | } | |
829 | } | |
830 | spin_unlock_bh(&mfc_unres_lock); | |
831 | ||
832 | if (uc) { | |
833 | ipmr_cache_resolve(uc, c); | |
834 | kmem_cache_free(mrt_cachep, uc); | |
835 | } | |
836 | return 0; | |
837 | } | |
838 | ||
839 | /* | |
840 | * Close the multicast socket, and clear the vif tables etc | |
841 | */ | |
e905a9ed | 842 | |
1da177e4 LT |
843 | static void mroute_clean_tables(struct sock *sk) |
844 | { | |
845 | int i; | |
e905a9ed | 846 | |
1da177e4 LT |
847 | /* |
848 | * Shut down all active vif entries | |
849 | */ | |
cf958ae3 BT |
850 | for (i = 0; i < init_net.ipv4.maxvif; i++) { |
851 | if (!(init_net.ipv4.vif_table[i].flags&VIFF_STATIC)) | |
7dc00c82 | 852 | vif_delete(i, 0); |
1da177e4 LT |
853 | } |
854 | ||
855 | /* | |
856 | * Wipe the cache | |
857 | */ | |
c354e124 | 858 | for (i=0; i<MFC_LINES; i++) { |
1da177e4 LT |
859 | struct mfc_cache *c, **cp; |
860 | ||
861 | cp = &mfc_cache_array[i]; | |
862 | while ((c = *cp) != NULL) { | |
863 | if (c->mfc_flags&MFC_STATIC) { | |
864 | cp = &c->next; | |
865 | continue; | |
866 | } | |
867 | write_lock_bh(&mrt_lock); | |
868 | *cp = c->next; | |
869 | write_unlock_bh(&mrt_lock); | |
870 | ||
871 | kmem_cache_free(mrt_cachep, c); | |
872 | } | |
873 | } | |
874 | ||
875 | if (atomic_read(&cache_resolve_queue_len) != 0) { | |
876 | struct mfc_cache *c; | |
877 | ||
878 | spin_lock_bh(&mfc_unres_lock); | |
879 | while (mfc_unres_queue != NULL) { | |
880 | c = mfc_unres_queue; | |
881 | mfc_unres_queue = c->next; | |
882 | spin_unlock_bh(&mfc_unres_lock); | |
883 | ||
884 | ipmr_destroy_unres(c); | |
885 | ||
886 | spin_lock_bh(&mfc_unres_lock); | |
887 | } | |
888 | spin_unlock_bh(&mfc_unres_lock); | |
889 | } | |
890 | } | |
891 | ||
892 | static void mrtsock_destruct(struct sock *sk) | |
893 | { | |
894 | rtnl_lock(); | |
70a269e6 | 895 | if (sk == init_net.ipv4.mroute_sk) { |
3b1e0a65 | 896 | IPV4_DEVCONF_ALL(sock_net(sk), MC_FORWARDING)--; |
1da177e4 LT |
897 | |
898 | write_lock_bh(&mrt_lock); | |
70a269e6 | 899 | init_net.ipv4.mroute_sk = NULL; |
1da177e4 LT |
900 | write_unlock_bh(&mrt_lock); |
901 | ||
902 | mroute_clean_tables(sk); | |
903 | } | |
904 | rtnl_unlock(); | |
905 | } | |
906 | ||
907 | /* | |
908 | * Socket options and virtual interface manipulation. The whole | |
909 | * virtual interface system is a complete heap, but unfortunately | |
910 | * that's how BSD mrouted happens to think. Maybe one day with a proper | |
911 | * MOSPF/PIM router set up we can clean this up. | |
912 | */ | |
e905a9ed | 913 | |
c354e124 | 914 | int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, int optlen) |
1da177e4 LT |
915 | { |
916 | int ret; | |
917 | struct vifctl vif; | |
918 | struct mfcctl mfc; | |
e905a9ed | 919 | |
132adf54 | 920 | if (optname != MRT_INIT) { |
70a269e6 | 921 | if (sk != init_net.ipv4.mroute_sk && !capable(CAP_NET_ADMIN)) |
1da177e4 LT |
922 | return -EACCES; |
923 | } | |
924 | ||
132adf54 SH |
925 | switch (optname) { |
926 | case MRT_INIT: | |
927 | if (sk->sk_type != SOCK_RAW || | |
928 | inet_sk(sk)->num != IPPROTO_IGMP) | |
929 | return -EOPNOTSUPP; | |
c354e124 | 930 | if (optlen != sizeof(int)) |
132adf54 | 931 | return -ENOPROTOOPT; |
1da177e4 | 932 | |
132adf54 | 933 | rtnl_lock(); |
70a269e6 | 934 | if (init_net.ipv4.mroute_sk) { |
1da177e4 | 935 | rtnl_unlock(); |
132adf54 SH |
936 | return -EADDRINUSE; |
937 | } | |
938 | ||
939 | ret = ip_ra_control(sk, 1, mrtsock_destruct); | |
940 | if (ret == 0) { | |
941 | write_lock_bh(&mrt_lock); | |
70a269e6 | 942 | init_net.ipv4.mroute_sk = sk; |
132adf54 SH |
943 | write_unlock_bh(&mrt_lock); |
944 | ||
3b1e0a65 | 945 | IPV4_DEVCONF_ALL(sock_net(sk), MC_FORWARDING)++; |
132adf54 SH |
946 | } |
947 | rtnl_unlock(); | |
948 | return ret; | |
949 | case MRT_DONE: | |
70a269e6 | 950 | if (sk != init_net.ipv4.mroute_sk) |
132adf54 SH |
951 | return -EACCES; |
952 | return ip_ra_control(sk, 0, NULL); | |
953 | case MRT_ADD_VIF: | |
954 | case MRT_DEL_VIF: | |
c354e124 | 955 | if (optlen != sizeof(vif)) |
132adf54 | 956 | return -EINVAL; |
c354e124 | 957 | if (copy_from_user(&vif, optval, sizeof(vif))) |
132adf54 SH |
958 | return -EFAULT; |
959 | if (vif.vifc_vifi >= MAXVIFS) | |
960 | return -ENFILE; | |
961 | rtnl_lock(); | |
c354e124 | 962 | if (optname == MRT_ADD_VIF) { |
70a269e6 | 963 | ret = vif_add(&vif, sk == init_net.ipv4.mroute_sk); |
132adf54 | 964 | } else { |
7dc00c82 | 965 | ret = vif_delete(vif.vifc_vifi, 0); |
132adf54 SH |
966 | } |
967 | rtnl_unlock(); | |
968 | return ret; | |
1da177e4 LT |
969 | |
970 | /* | |
971 | * Manipulate the forwarding caches. These live | |
972 | * in a sort of kernel/user symbiosis. | |
973 | */ | |
132adf54 SH |
974 | case MRT_ADD_MFC: |
975 | case MRT_DEL_MFC: | |
c354e124 | 976 | if (optlen != sizeof(mfc)) |
132adf54 | 977 | return -EINVAL; |
c354e124 | 978 | if (copy_from_user(&mfc, optval, sizeof(mfc))) |
132adf54 SH |
979 | return -EFAULT; |
980 | rtnl_lock(); | |
c354e124 | 981 | if (optname == MRT_DEL_MFC) |
132adf54 SH |
982 | ret = ipmr_mfc_delete(&mfc); |
983 | else | |
70a269e6 | 984 | ret = ipmr_mfc_add(&mfc, sk == init_net.ipv4.mroute_sk); |
132adf54 SH |
985 | rtnl_unlock(); |
986 | return ret; | |
1da177e4 LT |
987 | /* |
988 | * Control PIM assert. | |
989 | */ | |
132adf54 SH |
990 | case MRT_ASSERT: |
991 | { | |
992 | int v; | |
993 | if (get_user(v,(int __user *)optval)) | |
994 | return -EFAULT; | |
995 | mroute_do_assert=(v)?1:0; | |
996 | return 0; | |
997 | } | |
1da177e4 | 998 | #ifdef CONFIG_IP_PIMSM |
132adf54 SH |
999 | case MRT_PIM: |
1000 | { | |
ba93ef74 SH |
1001 | int v; |
1002 | ||
132adf54 SH |
1003 | if (get_user(v,(int __user *)optval)) |
1004 | return -EFAULT; | |
ba93ef74 SH |
1005 | v = (v) ? 1 : 0; |
1006 | ||
132adf54 SH |
1007 | rtnl_lock(); |
1008 | ret = 0; | |
1009 | if (v != mroute_do_pim) { | |
1010 | mroute_do_pim = v; | |
1011 | mroute_do_assert = v; | |
1da177e4 | 1012 | #ifdef CONFIG_IP_PIMSM_V2 |
132adf54 SH |
1013 | if (mroute_do_pim) |
1014 | ret = inet_add_protocol(&pim_protocol, | |
1015 | IPPROTO_PIM); | |
1016 | else | |
1017 | ret = inet_del_protocol(&pim_protocol, | |
1018 | IPPROTO_PIM); | |
1019 | if (ret < 0) | |
1020 | ret = -EAGAIN; | |
1da177e4 | 1021 | #endif |
1da177e4 | 1022 | } |
132adf54 SH |
1023 | rtnl_unlock(); |
1024 | return ret; | |
1025 | } | |
1da177e4 | 1026 | #endif |
132adf54 SH |
1027 | /* |
1028 | * Spurious command, or MRT_VERSION which you cannot | |
1029 | * set. | |
1030 | */ | |
1031 | default: | |
1032 | return -ENOPROTOOPT; | |
1da177e4 LT |
1033 | } |
1034 | } | |
1035 | ||
1036 | /* | |
1037 | * Getsock opt support for the multicast routing system. | |
1038 | */ | |
e905a9ed | 1039 | |
c354e124 | 1040 | int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen) |
1da177e4 LT |
1041 | { |
1042 | int olr; | |
1043 | int val; | |
1044 | ||
c354e124 | 1045 | if (optname != MRT_VERSION && |
1da177e4 LT |
1046 | #ifdef CONFIG_IP_PIMSM |
1047 | optname!=MRT_PIM && | |
1048 | #endif | |
1049 | optname!=MRT_ASSERT) | |
1050 | return -ENOPROTOOPT; | |
1051 | ||
1052 | if (get_user(olr, optlen)) | |
1053 | return -EFAULT; | |
1054 | ||
1055 | olr = min_t(unsigned int, olr, sizeof(int)); | |
1056 | if (olr < 0) | |
1057 | return -EINVAL; | |
e905a9ed | 1058 | |
c354e124 | 1059 | if (put_user(olr, optlen)) |
1da177e4 | 1060 | return -EFAULT; |
c354e124 JK |
1061 | if (optname == MRT_VERSION) |
1062 | val = 0x0305; | |
1da177e4 | 1063 | #ifdef CONFIG_IP_PIMSM |
c354e124 JK |
1064 | else if (optname == MRT_PIM) |
1065 | val = mroute_do_pim; | |
1da177e4 LT |
1066 | #endif |
1067 | else | |
c354e124 JK |
1068 | val = mroute_do_assert; |
1069 | if (copy_to_user(optval, &val, olr)) | |
1da177e4 LT |
1070 | return -EFAULT; |
1071 | return 0; | |
1072 | } | |
1073 | ||
1074 | /* | |
1075 | * The IP multicast ioctl support routines. | |
1076 | */ | |
e905a9ed | 1077 | |
1da177e4 LT |
1078 | int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg) |
1079 | { | |
1080 | struct sioc_sg_req sr; | |
1081 | struct sioc_vif_req vr; | |
1082 | struct vif_device *vif; | |
1083 | struct mfc_cache *c; | |
e905a9ed | 1084 | |
132adf54 SH |
1085 | switch (cmd) { |
1086 | case SIOCGETVIFCNT: | |
c354e124 | 1087 | if (copy_from_user(&vr, arg, sizeof(vr))) |
132adf54 | 1088 | return -EFAULT; |
cf958ae3 | 1089 | if (vr.vifi >= init_net.ipv4.maxvif) |
132adf54 SH |
1090 | return -EINVAL; |
1091 | read_lock(&mrt_lock); | |
cf958ae3 BT |
1092 | vif = &init_net.ipv4.vif_table[vr.vifi]; |
1093 | if (VIF_EXISTS(&init_net, vr.vifi)) { | |
c354e124 JK |
1094 | vr.icount = vif->pkt_in; |
1095 | vr.ocount = vif->pkt_out; | |
1096 | vr.ibytes = vif->bytes_in; | |
1097 | vr.obytes = vif->bytes_out; | |
1da177e4 | 1098 | read_unlock(&mrt_lock); |
1da177e4 | 1099 | |
c354e124 | 1100 | if (copy_to_user(arg, &vr, sizeof(vr))) |
132adf54 SH |
1101 | return -EFAULT; |
1102 | return 0; | |
1103 | } | |
1104 | read_unlock(&mrt_lock); | |
1105 | return -EADDRNOTAVAIL; | |
1106 | case SIOCGETSGCNT: | |
c354e124 | 1107 | if (copy_from_user(&sr, arg, sizeof(sr))) |
132adf54 SH |
1108 | return -EFAULT; |
1109 | ||
1110 | read_lock(&mrt_lock); | |
1111 | c = ipmr_cache_find(sr.src.s_addr, sr.grp.s_addr); | |
1112 | if (c) { | |
1113 | sr.pktcnt = c->mfc_un.res.pkt; | |
1114 | sr.bytecnt = c->mfc_un.res.bytes; | |
1115 | sr.wrong_if = c->mfc_un.res.wrong_if; | |
1da177e4 | 1116 | read_unlock(&mrt_lock); |
132adf54 | 1117 | |
c354e124 | 1118 | if (copy_to_user(arg, &sr, sizeof(sr))) |
132adf54 SH |
1119 | return -EFAULT; |
1120 | return 0; | |
1121 | } | |
1122 | read_unlock(&mrt_lock); | |
1123 | return -EADDRNOTAVAIL; | |
1124 | default: | |
1125 | return -ENOIOCTLCMD; | |
1da177e4 LT |
1126 | } |
1127 | } | |
1128 | ||
1129 | ||
1130 | static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr) | |
1131 | { | |
e9dc8653 | 1132 | struct net_device *dev = ptr; |
1da177e4 LT |
1133 | struct vif_device *v; |
1134 | int ct; | |
e9dc8653 | 1135 | |
721499e8 | 1136 | if (!net_eq(dev_net(dev), &init_net)) |
e9dc8653 EB |
1137 | return NOTIFY_DONE; |
1138 | ||
1da177e4 LT |
1139 | if (event != NETDEV_UNREGISTER) |
1140 | return NOTIFY_DONE; | |
cf958ae3 BT |
1141 | v = &init_net.ipv4.vif_table[0]; |
1142 | for (ct = 0; ct < init_net.ipv4.maxvif; ct++, v++) { | |
c354e124 | 1143 | if (v->dev == dev) |
7dc00c82 | 1144 | vif_delete(ct, 1); |
1da177e4 LT |
1145 | } |
1146 | return NOTIFY_DONE; | |
1147 | } | |
1148 | ||
1149 | ||
c354e124 | 1150 | static struct notifier_block ip_mr_notifier = { |
1da177e4 LT |
1151 | .notifier_call = ipmr_device_event, |
1152 | }; | |
1153 | ||
1154 | /* | |
1155 | * Encapsulate a packet by attaching a valid IPIP header to it. | |
1156 | * This avoids tunnel drivers and other mess and gives us the speed so | |
1157 | * important for multicast video. | |
1158 | */ | |
e905a9ed | 1159 | |
114c7844 | 1160 | static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr) |
1da177e4 | 1161 | { |
8856dfa3 | 1162 | struct iphdr *iph; |
eddc9ec5 | 1163 | struct iphdr *old_iph = ip_hdr(skb); |
8856dfa3 ACM |
1164 | |
1165 | skb_push(skb, sizeof(struct iphdr)); | |
b0e380b1 | 1166 | skb->transport_header = skb->network_header; |
8856dfa3 | 1167 | skb_reset_network_header(skb); |
eddc9ec5 | 1168 | iph = ip_hdr(skb); |
1da177e4 LT |
1169 | |
1170 | iph->version = 4; | |
e023dd64 ACM |
1171 | iph->tos = old_iph->tos; |
1172 | iph->ttl = old_iph->ttl; | |
1da177e4 LT |
1173 | iph->frag_off = 0; |
1174 | iph->daddr = daddr; | |
1175 | iph->saddr = saddr; | |
1176 | iph->protocol = IPPROTO_IPIP; | |
1177 | iph->ihl = 5; | |
1178 | iph->tot_len = htons(skb->len); | |
1179 | ip_select_ident(iph, skb->dst, NULL); | |
1180 | ip_send_check(iph); | |
1181 | ||
1da177e4 LT |
1182 | memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); |
1183 | nf_reset(skb); | |
1184 | } | |
1185 | ||
1186 | static inline int ipmr_forward_finish(struct sk_buff *skb) | |
1187 | { | |
1188 | struct ip_options * opt = &(IPCB(skb)->opt); | |
1189 | ||
7c73a6fa | 1190 | IP_INC_STATS_BH(dev_net(skb->dst->dev), IPSTATS_MIB_OUTFORWDATAGRAMS); |
1da177e4 LT |
1191 | |
1192 | if (unlikely(opt->optlen)) | |
1193 | ip_forward_options(skb); | |
1194 | ||
1195 | return dst_output(skb); | |
1196 | } | |
1197 | ||
1198 | /* | |
1199 | * Processing handlers for ipmr_forward | |
1200 | */ | |
1201 | ||
1202 | static void ipmr_queue_xmit(struct sk_buff *skb, struct mfc_cache *c, int vifi) | |
1203 | { | |
eddc9ec5 | 1204 | const struct iphdr *iph = ip_hdr(skb); |
cf958ae3 | 1205 | struct vif_device *vif = &init_net.ipv4.vif_table[vifi]; |
1da177e4 LT |
1206 | struct net_device *dev; |
1207 | struct rtable *rt; | |
1208 | int encap = 0; | |
1209 | ||
1210 | if (vif->dev == NULL) | |
1211 | goto out_free; | |
1212 | ||
1213 | #ifdef CONFIG_IP_PIMSM | |
1214 | if (vif->flags & VIFF_REGISTER) { | |
1215 | vif->pkt_out++; | |
c354e124 | 1216 | vif->bytes_out += skb->len; |
cf3677ae PE |
1217 | vif->dev->stats.tx_bytes += skb->len; |
1218 | vif->dev->stats.tx_packets++; | |
1da177e4 LT |
1219 | ipmr_cache_report(skb, vifi, IGMPMSG_WHOLEPKT); |
1220 | kfree_skb(skb); | |
1221 | return; | |
1222 | } | |
1223 | #endif | |
1224 | ||
1225 | if (vif->flags&VIFF_TUNNEL) { | |
1226 | struct flowi fl = { .oif = vif->link, | |
1227 | .nl_u = { .ip4_u = | |
1228 | { .daddr = vif->remote, | |
1229 | .saddr = vif->local, | |
1230 | .tos = RT_TOS(iph->tos) } }, | |
1231 | .proto = IPPROTO_IPIP }; | |
f206351a | 1232 | if (ip_route_output_key(&init_net, &rt, &fl)) |
1da177e4 LT |
1233 | goto out_free; |
1234 | encap = sizeof(struct iphdr); | |
1235 | } else { | |
1236 | struct flowi fl = { .oif = vif->link, | |
1237 | .nl_u = { .ip4_u = | |
1238 | { .daddr = iph->daddr, | |
1239 | .tos = RT_TOS(iph->tos) } }, | |
1240 | .proto = IPPROTO_IPIP }; | |
f206351a | 1241 | if (ip_route_output_key(&init_net, &rt, &fl)) |
1da177e4 LT |
1242 | goto out_free; |
1243 | } | |
1244 | ||
1245 | dev = rt->u.dst.dev; | |
1246 | ||
1247 | if (skb->len+encap > dst_mtu(&rt->u.dst) && (ntohs(iph->frag_off) & IP_DF)) { | |
1248 | /* Do not fragment multicasts. Alas, IPv4 does not | |
1249 | allow to send ICMP, so that packets will disappear | |
1250 | to blackhole. | |
1251 | */ | |
1252 | ||
7c73a6fa | 1253 | IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
1254 | ip_rt_put(rt); |
1255 | goto out_free; | |
1256 | } | |
1257 | ||
1258 | encap += LL_RESERVED_SPACE(dev) + rt->u.dst.header_len; | |
1259 | ||
1260 | if (skb_cow(skb, encap)) { | |
e905a9ed | 1261 | ip_rt_put(rt); |
1da177e4 LT |
1262 | goto out_free; |
1263 | } | |
1264 | ||
1265 | vif->pkt_out++; | |
c354e124 | 1266 | vif->bytes_out += skb->len; |
1da177e4 LT |
1267 | |
1268 | dst_release(skb->dst); | |
1269 | skb->dst = &rt->u.dst; | |
eddc9ec5 | 1270 | ip_decrease_ttl(ip_hdr(skb)); |
1da177e4 LT |
1271 | |
1272 | /* FIXME: forward and output firewalls used to be called here. | |
1273 | * What do we do with netfilter? -- RR */ | |
1274 | if (vif->flags & VIFF_TUNNEL) { | |
1275 | ip_encap(skb, vif->local, vif->remote); | |
1276 | /* FIXME: extra output firewall step used to be here. --RR */ | |
2f4c02d4 PE |
1277 | vif->dev->stats.tx_packets++; |
1278 | vif->dev->stats.tx_bytes += skb->len; | |
1da177e4 LT |
1279 | } |
1280 | ||
1281 | IPCB(skb)->flags |= IPSKB_FORWARDED; | |
1282 | ||
1283 | /* | |
1284 | * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally | |
1285 | * not only before forwarding, but after forwarding on all output | |
1286 | * interfaces. It is clear, if mrouter runs a multicasting | |
1287 | * program, it should receive packets not depending to what interface | |
1288 | * program is joined. | |
1289 | * If we will not make it, the program will have to join on all | |
1290 | * interfaces. On the other hand, multihoming host (or router, but | |
1291 | * not mrouter) cannot join to more than one interface - it will | |
1292 | * result in receiving multiple packets. | |
1293 | */ | |
6e23ae2a | 1294 | NF_HOOK(PF_INET, NF_INET_FORWARD, skb, skb->dev, dev, |
1da177e4 LT |
1295 | ipmr_forward_finish); |
1296 | return; | |
1297 | ||
1298 | out_free: | |
1299 | kfree_skb(skb); | |
1300 | return; | |
1301 | } | |
1302 | ||
1303 | static int ipmr_find_vif(struct net_device *dev) | |
1304 | { | |
1305 | int ct; | |
cf958ae3 BT |
1306 | for (ct = init_net.ipv4.maxvif-1; ct >= 0; ct--) { |
1307 | if (init_net.ipv4.vif_table[ct].dev == dev) | |
1da177e4 LT |
1308 | break; |
1309 | } | |
1310 | return ct; | |
1311 | } | |
1312 | ||
1313 | /* "local" means that we should preserve one skb (for local delivery) */ | |
1314 | ||
1315 | static int ip_mr_forward(struct sk_buff *skb, struct mfc_cache *cache, int local) | |
1316 | { | |
1317 | int psend = -1; | |
1318 | int vif, ct; | |
1319 | ||
1320 | vif = cache->mfc_parent; | |
1321 | cache->mfc_un.res.pkt++; | |
1322 | cache->mfc_un.res.bytes += skb->len; | |
1323 | ||
1324 | /* | |
1325 | * Wrong interface: drop packet and (maybe) send PIM assert. | |
1326 | */ | |
cf958ae3 | 1327 | if (init_net.ipv4.vif_table[vif].dev != skb->dev) { |
1da177e4 LT |
1328 | int true_vifi; |
1329 | ||
ee6b9673 | 1330 | if (skb->rtable->fl.iif == 0) { |
1da177e4 LT |
1331 | /* It is our own packet, looped back. |
1332 | Very complicated situation... | |
1333 | ||
1334 | The best workaround until routing daemons will be | |
1335 | fixed is not to redistribute packet, if it was | |
1336 | send through wrong interface. It means, that | |
1337 | multicast applications WILL NOT work for | |
1338 | (S,G), which have default multicast route pointing | |
1339 | to wrong oif. In any case, it is not a good | |
1340 | idea to use multicasting applications on router. | |
1341 | */ | |
1342 | goto dont_forward; | |
1343 | } | |
1344 | ||
1345 | cache->mfc_un.res.wrong_if++; | |
1346 | true_vifi = ipmr_find_vif(skb->dev); | |
1347 | ||
1348 | if (true_vifi >= 0 && mroute_do_assert && | |
1349 | /* pimsm uses asserts, when switching from RPT to SPT, | |
1350 | so that we cannot check that packet arrived on an oif. | |
1351 | It is bad, but otherwise we would need to move pretty | |
1352 | large chunk of pimd to kernel. Ough... --ANK | |
1353 | */ | |
1354 | (mroute_do_pim || cache->mfc_un.res.ttls[true_vifi] < 255) && | |
e905a9ed | 1355 | time_after(jiffies, |
1da177e4 LT |
1356 | cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { |
1357 | cache->mfc_un.res.last_assert = jiffies; | |
1358 | ipmr_cache_report(skb, true_vifi, IGMPMSG_WRONGVIF); | |
1359 | } | |
1360 | goto dont_forward; | |
1361 | } | |
1362 | ||
cf958ae3 BT |
1363 | init_net.ipv4.vif_table[vif].pkt_in++; |
1364 | init_net.ipv4.vif_table[vif].bytes_in += skb->len; | |
1da177e4 LT |
1365 | |
1366 | /* | |
1367 | * Forward the frame | |
1368 | */ | |
1369 | for (ct = cache->mfc_un.res.maxvif-1; ct >= cache->mfc_un.res.minvif; ct--) { | |
eddc9ec5 | 1370 | if (ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) { |
1da177e4 LT |
1371 | if (psend != -1) { |
1372 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1373 | if (skb2) | |
1374 | ipmr_queue_xmit(skb2, cache, psend); | |
1375 | } | |
c354e124 | 1376 | psend = ct; |
1da177e4 LT |
1377 | } |
1378 | } | |
1379 | if (psend != -1) { | |
1380 | if (local) { | |
1381 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1382 | if (skb2) | |
1383 | ipmr_queue_xmit(skb2, cache, psend); | |
1384 | } else { | |
1385 | ipmr_queue_xmit(skb, cache, psend); | |
1386 | return 0; | |
1387 | } | |
1388 | } | |
1389 | ||
1390 | dont_forward: | |
1391 | if (!local) | |
1392 | kfree_skb(skb); | |
1393 | return 0; | |
1394 | } | |
1395 | ||
1396 | ||
1397 | /* | |
1398 | * Multicast packets for forwarding arrive here | |
1399 | */ | |
1400 | ||
1401 | int ip_mr_input(struct sk_buff *skb) | |
1402 | { | |
1403 | struct mfc_cache *cache; | |
ee6b9673 | 1404 | int local = skb->rtable->rt_flags&RTCF_LOCAL; |
1da177e4 LT |
1405 | |
1406 | /* Packet is looped back after forward, it should not be | |
1407 | forwarded second time, but still can be delivered locally. | |
1408 | */ | |
1409 | if (IPCB(skb)->flags&IPSKB_FORWARDED) | |
1410 | goto dont_forward; | |
1411 | ||
1412 | if (!local) { | |
1413 | if (IPCB(skb)->opt.router_alert) { | |
1414 | if (ip_call_ra_chain(skb)) | |
1415 | return 0; | |
eddc9ec5 | 1416 | } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP){ |
1da177e4 LT |
1417 | /* IGMPv1 (and broken IGMPv2 implementations sort of |
1418 | Cisco IOS <= 11.2(8)) do not put router alert | |
1419 | option to IGMP packets destined to routable | |
1420 | groups. It is very bad, because it means | |
1421 | that we can forward NO IGMP messages. | |
1422 | */ | |
1423 | read_lock(&mrt_lock); | |
70a269e6 | 1424 | if (init_net.ipv4.mroute_sk) { |
2715bcf9 | 1425 | nf_reset(skb); |
70a269e6 | 1426 | raw_rcv(init_net.ipv4.mroute_sk, skb); |
1da177e4 LT |
1427 | read_unlock(&mrt_lock); |
1428 | return 0; | |
1429 | } | |
1430 | read_unlock(&mrt_lock); | |
1431 | } | |
1432 | } | |
1433 | ||
1434 | read_lock(&mrt_lock); | |
eddc9ec5 | 1435 | cache = ipmr_cache_find(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr); |
1da177e4 LT |
1436 | |
1437 | /* | |
1438 | * No usable cache entry | |
1439 | */ | |
c354e124 | 1440 | if (cache == NULL) { |
1da177e4 LT |
1441 | int vif; |
1442 | ||
1443 | if (local) { | |
1444 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1445 | ip_local_deliver(skb); | |
1446 | if (skb2 == NULL) { | |
1447 | read_unlock(&mrt_lock); | |
1448 | return -ENOBUFS; | |
1449 | } | |
1450 | skb = skb2; | |
1451 | } | |
1452 | ||
1453 | vif = ipmr_find_vif(skb->dev); | |
1454 | if (vif >= 0) { | |
1455 | int err = ipmr_cache_unresolved(vif, skb); | |
1456 | read_unlock(&mrt_lock); | |
1457 | ||
1458 | return err; | |
1459 | } | |
1460 | read_unlock(&mrt_lock); | |
1461 | kfree_skb(skb); | |
1462 | return -ENODEV; | |
1463 | } | |
1464 | ||
1465 | ip_mr_forward(skb, cache, local); | |
1466 | ||
1467 | read_unlock(&mrt_lock); | |
1468 | ||
1469 | if (local) | |
1470 | return ip_local_deliver(skb); | |
1471 | ||
1472 | return 0; | |
1473 | ||
1474 | dont_forward: | |
1475 | if (local) | |
1476 | return ip_local_deliver(skb); | |
1477 | kfree_skb(skb); | |
1478 | return 0; | |
1479 | } | |
1480 | ||
b1879204 IJ |
1481 | #ifdef CONFIG_IP_PIMSM |
1482 | static int __pim_rcv(struct sk_buff *skb, unsigned int pimlen) | |
1da177e4 | 1483 | { |
b1879204 IJ |
1484 | struct net_device *reg_dev = NULL; |
1485 | struct iphdr *encap; | |
1da177e4 | 1486 | |
b1879204 | 1487 | encap = (struct iphdr *)(skb_transport_header(skb) + pimlen); |
1da177e4 LT |
1488 | /* |
1489 | Check that: | |
1490 | a. packet is really destinted to a multicast group | |
1491 | b. packet is not a NULL-REGISTER | |
1492 | c. packet is not truncated | |
1493 | */ | |
f97c1e0c | 1494 | if (!ipv4_is_multicast(encap->daddr) || |
1da177e4 | 1495 | encap->tot_len == 0 || |
b1879204 IJ |
1496 | ntohs(encap->tot_len) + pimlen > skb->len) |
1497 | return 1; | |
1da177e4 LT |
1498 | |
1499 | read_lock(&mrt_lock); | |
1500 | if (reg_vif_num >= 0) | |
cf958ae3 | 1501 | reg_dev = init_net.ipv4.vif_table[reg_vif_num].dev; |
1da177e4 LT |
1502 | if (reg_dev) |
1503 | dev_hold(reg_dev); | |
1504 | read_unlock(&mrt_lock); | |
1505 | ||
e905a9ed | 1506 | if (reg_dev == NULL) |
b1879204 | 1507 | return 1; |
1da177e4 | 1508 | |
b0e380b1 | 1509 | skb->mac_header = skb->network_header; |
1da177e4 | 1510 | skb_pull(skb, (u8*)encap - skb->data); |
31c7711b | 1511 | skb_reset_network_header(skb); |
1da177e4 | 1512 | skb->dev = reg_dev; |
1da177e4 LT |
1513 | skb->protocol = htons(ETH_P_IP); |
1514 | skb->ip_summed = 0; | |
1515 | skb->pkt_type = PACKET_HOST; | |
1516 | dst_release(skb->dst); | |
1517 | skb->dst = NULL; | |
cf3677ae PE |
1518 | reg_dev->stats.rx_bytes += skb->len; |
1519 | reg_dev->stats.rx_packets++; | |
1da177e4 LT |
1520 | nf_reset(skb); |
1521 | netif_rx(skb); | |
1522 | dev_put(reg_dev); | |
b1879204 | 1523 | |
1da177e4 | 1524 | return 0; |
b1879204 IJ |
1525 | } |
1526 | #endif | |
1527 | ||
1528 | #ifdef CONFIG_IP_PIMSM_V1 | |
1529 | /* | |
1530 | * Handle IGMP messages of PIMv1 | |
1531 | */ | |
1532 | ||
1533 | int pim_rcv_v1(struct sk_buff * skb) | |
1534 | { | |
1535 | struct igmphdr *pim; | |
1536 | ||
1537 | if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) | |
1538 | goto drop; | |
1539 | ||
1540 | pim = igmp_hdr(skb); | |
1541 | ||
1542 | if (!mroute_do_pim || | |
1543 | pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER) | |
1544 | goto drop; | |
1545 | ||
1546 | if (__pim_rcv(skb, sizeof(*pim))) { | |
1547 | drop: | |
1548 | kfree_skb(skb); | |
1549 | } | |
1da177e4 LT |
1550 | return 0; |
1551 | } | |
1552 | #endif | |
1553 | ||
1554 | #ifdef CONFIG_IP_PIMSM_V2 | |
1555 | static int pim_rcv(struct sk_buff * skb) | |
1556 | { | |
1557 | struct pimreghdr *pim; | |
1da177e4 | 1558 | |
b1879204 | 1559 | if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) |
1da177e4 LT |
1560 | goto drop; |
1561 | ||
9c70220b | 1562 | pim = (struct pimreghdr *)skb_transport_header(skb); |
e905a9ed | 1563 | if (pim->type != ((PIM_VERSION<<4)|(PIM_REGISTER)) || |
1da177e4 | 1564 | (pim->flags&PIM_NULL_REGISTER) || |
e905a9ed | 1565 | (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && |
d3bc23e7 | 1566 | csum_fold(skb_checksum(skb, 0, skb->len, 0)))) |
1da177e4 LT |
1567 | goto drop; |
1568 | ||
b1879204 IJ |
1569 | if (__pim_rcv(skb, sizeof(*pim))) { |
1570 | drop: | |
1571 | kfree_skb(skb); | |
1572 | } | |
1da177e4 LT |
1573 | return 0; |
1574 | } | |
1575 | #endif | |
1576 | ||
1577 | static int | |
1578 | ipmr_fill_mroute(struct sk_buff *skb, struct mfc_cache *c, struct rtmsg *rtm) | |
1579 | { | |
1580 | int ct; | |
1581 | struct rtnexthop *nhp; | |
cf958ae3 | 1582 | struct net_device *dev = init_net.ipv4.vif_table[c->mfc_parent].dev; |
27a884dc | 1583 | u8 *b = skb_tail_pointer(skb); |
1da177e4 LT |
1584 | struct rtattr *mp_head; |
1585 | ||
1586 | if (dev) | |
1587 | RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex); | |
1588 | ||
c354e124 | 1589 | mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0)); |
1da177e4 LT |
1590 | |
1591 | for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { | |
1592 | if (c->mfc_un.res.ttls[ct] < 255) { | |
1593 | if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4)) | |
1594 | goto rtattr_failure; | |
c354e124 | 1595 | nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp))); |
1da177e4 LT |
1596 | nhp->rtnh_flags = 0; |
1597 | nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; | |
cf958ae3 | 1598 | nhp->rtnh_ifindex = init_net.ipv4.vif_table[ct].dev->ifindex; |
1da177e4 LT |
1599 | nhp->rtnh_len = sizeof(*nhp); |
1600 | } | |
1601 | } | |
1602 | mp_head->rta_type = RTA_MULTIPATH; | |
27a884dc | 1603 | mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head; |
1da177e4 LT |
1604 | rtm->rtm_type = RTN_MULTICAST; |
1605 | return 1; | |
1606 | ||
1607 | rtattr_failure: | |
dc5fc579 | 1608 | nlmsg_trim(skb, b); |
1da177e4 LT |
1609 | return -EMSGSIZE; |
1610 | } | |
1611 | ||
1612 | int ipmr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait) | |
1613 | { | |
1614 | int err; | |
1615 | struct mfc_cache *cache; | |
ee6b9673 | 1616 | struct rtable *rt = skb->rtable; |
1da177e4 LT |
1617 | |
1618 | read_lock(&mrt_lock); | |
1619 | cache = ipmr_cache_find(rt->rt_src, rt->rt_dst); | |
1620 | ||
c354e124 | 1621 | if (cache == NULL) { |
72287490 | 1622 | struct sk_buff *skb2; |
eddc9ec5 | 1623 | struct iphdr *iph; |
1da177e4 LT |
1624 | struct net_device *dev; |
1625 | int vif; | |
1626 | ||
1627 | if (nowait) { | |
1628 | read_unlock(&mrt_lock); | |
1629 | return -EAGAIN; | |
1630 | } | |
1631 | ||
1632 | dev = skb->dev; | |
1633 | if (dev == NULL || (vif = ipmr_find_vif(dev)) < 0) { | |
1634 | read_unlock(&mrt_lock); | |
1635 | return -ENODEV; | |
1636 | } | |
72287490 AK |
1637 | skb2 = skb_clone(skb, GFP_ATOMIC); |
1638 | if (!skb2) { | |
1639 | read_unlock(&mrt_lock); | |
1640 | return -ENOMEM; | |
1641 | } | |
1642 | ||
e2d1bca7 ACM |
1643 | skb_push(skb2, sizeof(struct iphdr)); |
1644 | skb_reset_network_header(skb2); | |
eddc9ec5 ACM |
1645 | iph = ip_hdr(skb2); |
1646 | iph->ihl = sizeof(struct iphdr) >> 2; | |
1647 | iph->saddr = rt->rt_src; | |
1648 | iph->daddr = rt->rt_dst; | |
1649 | iph->version = 0; | |
72287490 | 1650 | err = ipmr_cache_unresolved(vif, skb2); |
1da177e4 LT |
1651 | read_unlock(&mrt_lock); |
1652 | return err; | |
1653 | } | |
1654 | ||
1655 | if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY)) | |
1656 | cache->mfc_flags |= MFC_NOTIFY; | |
1657 | err = ipmr_fill_mroute(skb, cache, rtm); | |
1658 | read_unlock(&mrt_lock); | |
1659 | return err; | |
1660 | } | |
1661 | ||
e905a9ed | 1662 | #ifdef CONFIG_PROC_FS |
1da177e4 LT |
1663 | /* |
1664 | * The /proc interfaces to multicast routing /proc/ip_mr_cache /proc/ip_mr_vif | |
1665 | */ | |
1666 | struct ipmr_vif_iter { | |
1667 | int ct; | |
1668 | }; | |
1669 | ||
1670 | static struct vif_device *ipmr_vif_seq_idx(struct ipmr_vif_iter *iter, | |
1671 | loff_t pos) | |
1672 | { | |
cf958ae3 BT |
1673 | for (iter->ct = 0; iter->ct < init_net.ipv4.maxvif; ++iter->ct) { |
1674 | if (!VIF_EXISTS(&init_net, iter->ct)) | |
1da177e4 | 1675 | continue; |
e905a9ed | 1676 | if (pos-- == 0) |
cf958ae3 | 1677 | return &init_net.ipv4.vif_table[iter->ct]; |
1da177e4 LT |
1678 | } |
1679 | return NULL; | |
1680 | } | |
1681 | ||
1682 | static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos) | |
ba93ef74 | 1683 | __acquires(mrt_lock) |
1da177e4 LT |
1684 | { |
1685 | read_lock(&mrt_lock); | |
e905a9ed | 1686 | return *pos ? ipmr_vif_seq_idx(seq->private, *pos - 1) |
1da177e4 LT |
1687 | : SEQ_START_TOKEN; |
1688 | } | |
1689 | ||
1690 | static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
1691 | { | |
1692 | struct ipmr_vif_iter *iter = seq->private; | |
1693 | ||
1694 | ++*pos; | |
1695 | if (v == SEQ_START_TOKEN) | |
1696 | return ipmr_vif_seq_idx(iter, 0); | |
e905a9ed | 1697 | |
cf958ae3 BT |
1698 | while (++iter->ct < init_net.ipv4.maxvif) { |
1699 | if (!VIF_EXISTS(&init_net, iter->ct)) | |
1da177e4 | 1700 | continue; |
cf958ae3 | 1701 | return &init_net.ipv4.vif_table[iter->ct]; |
1da177e4 LT |
1702 | } |
1703 | return NULL; | |
1704 | } | |
1705 | ||
1706 | static void ipmr_vif_seq_stop(struct seq_file *seq, void *v) | |
ba93ef74 | 1707 | __releases(mrt_lock) |
1da177e4 LT |
1708 | { |
1709 | read_unlock(&mrt_lock); | |
1710 | } | |
1711 | ||
1712 | static int ipmr_vif_seq_show(struct seq_file *seq, void *v) | |
1713 | { | |
1714 | if (v == SEQ_START_TOKEN) { | |
e905a9ed | 1715 | seq_puts(seq, |
1da177e4 LT |
1716 | "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n"); |
1717 | } else { | |
1718 | const struct vif_device *vif = v; | |
1719 | const char *name = vif->dev ? vif->dev->name : "none"; | |
1720 | ||
1721 | seq_printf(seq, | |
1722 | "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n", | |
cf958ae3 | 1723 | vif - init_net.ipv4.vif_table, |
e905a9ed | 1724 | name, vif->bytes_in, vif->pkt_in, |
1da177e4 LT |
1725 | vif->bytes_out, vif->pkt_out, |
1726 | vif->flags, vif->local, vif->remote); | |
1727 | } | |
1728 | return 0; | |
1729 | } | |
1730 | ||
f690808e | 1731 | static const struct seq_operations ipmr_vif_seq_ops = { |
1da177e4 LT |
1732 | .start = ipmr_vif_seq_start, |
1733 | .next = ipmr_vif_seq_next, | |
1734 | .stop = ipmr_vif_seq_stop, | |
1735 | .show = ipmr_vif_seq_show, | |
1736 | }; | |
1737 | ||
1738 | static int ipmr_vif_open(struct inode *inode, struct file *file) | |
1739 | { | |
cf7732e4 PE |
1740 | return seq_open_private(file, &ipmr_vif_seq_ops, |
1741 | sizeof(struct ipmr_vif_iter)); | |
1da177e4 LT |
1742 | } |
1743 | ||
9a32144e | 1744 | static const struct file_operations ipmr_vif_fops = { |
1da177e4 LT |
1745 | .owner = THIS_MODULE, |
1746 | .open = ipmr_vif_open, | |
1747 | .read = seq_read, | |
1748 | .llseek = seq_lseek, | |
1749 | .release = seq_release_private, | |
1750 | }; | |
1751 | ||
1752 | struct ipmr_mfc_iter { | |
1753 | struct mfc_cache **cache; | |
1754 | int ct; | |
1755 | }; | |
1756 | ||
1757 | ||
1758 | static struct mfc_cache *ipmr_mfc_seq_idx(struct ipmr_mfc_iter *it, loff_t pos) | |
1759 | { | |
1760 | struct mfc_cache *mfc; | |
1761 | ||
1762 | it->cache = mfc_cache_array; | |
1763 | read_lock(&mrt_lock); | |
e905a9ed | 1764 | for (it->ct = 0; it->ct < MFC_LINES; it->ct++) |
132adf54 | 1765 | for (mfc = mfc_cache_array[it->ct]; mfc; mfc = mfc->next) |
e905a9ed | 1766 | if (pos-- == 0) |
1da177e4 LT |
1767 | return mfc; |
1768 | read_unlock(&mrt_lock); | |
1769 | ||
1770 | it->cache = &mfc_unres_queue; | |
1771 | spin_lock_bh(&mfc_unres_lock); | |
132adf54 | 1772 | for (mfc = mfc_unres_queue; mfc; mfc = mfc->next) |
1da177e4 LT |
1773 | if (pos-- == 0) |
1774 | return mfc; | |
1775 | spin_unlock_bh(&mfc_unres_lock); | |
1776 | ||
1777 | it->cache = NULL; | |
1778 | return NULL; | |
1779 | } | |
1780 | ||
1781 | ||
1782 | static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) | |
1783 | { | |
1784 | struct ipmr_mfc_iter *it = seq->private; | |
1785 | it->cache = NULL; | |
1786 | it->ct = 0; | |
e905a9ed | 1787 | return *pos ? ipmr_mfc_seq_idx(seq->private, *pos - 1) |
1da177e4 LT |
1788 | : SEQ_START_TOKEN; |
1789 | } | |
1790 | ||
1791 | static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
1792 | { | |
1793 | struct mfc_cache *mfc = v; | |
1794 | struct ipmr_mfc_iter *it = seq->private; | |
1795 | ||
1796 | ++*pos; | |
1797 | ||
1798 | if (v == SEQ_START_TOKEN) | |
1799 | return ipmr_mfc_seq_idx(seq->private, 0); | |
1800 | ||
1801 | if (mfc->next) | |
1802 | return mfc->next; | |
e905a9ed YH |
1803 | |
1804 | if (it->cache == &mfc_unres_queue) | |
1da177e4 LT |
1805 | goto end_of_list; |
1806 | ||
1807 | BUG_ON(it->cache != mfc_cache_array); | |
1808 | ||
1809 | while (++it->ct < MFC_LINES) { | |
1810 | mfc = mfc_cache_array[it->ct]; | |
1811 | if (mfc) | |
1812 | return mfc; | |
1813 | } | |
1814 | ||
1815 | /* exhausted cache_array, show unresolved */ | |
1816 | read_unlock(&mrt_lock); | |
1817 | it->cache = &mfc_unres_queue; | |
1818 | it->ct = 0; | |
e905a9ed | 1819 | |
1da177e4 LT |
1820 | spin_lock_bh(&mfc_unres_lock); |
1821 | mfc = mfc_unres_queue; | |
e905a9ed | 1822 | if (mfc) |
1da177e4 LT |
1823 | return mfc; |
1824 | ||
1825 | end_of_list: | |
1826 | spin_unlock_bh(&mfc_unres_lock); | |
1827 | it->cache = NULL; | |
1828 | ||
1829 | return NULL; | |
1830 | } | |
1831 | ||
1832 | static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) | |
1833 | { | |
1834 | struct ipmr_mfc_iter *it = seq->private; | |
1835 | ||
1836 | if (it->cache == &mfc_unres_queue) | |
1837 | spin_unlock_bh(&mfc_unres_lock); | |
1838 | else if (it->cache == mfc_cache_array) | |
1839 | read_unlock(&mrt_lock); | |
1840 | } | |
1841 | ||
1842 | static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) | |
1843 | { | |
1844 | int n; | |
1845 | ||
1846 | if (v == SEQ_START_TOKEN) { | |
e905a9ed | 1847 | seq_puts(seq, |
1da177e4 LT |
1848 | "Group Origin Iif Pkts Bytes Wrong Oifs\n"); |
1849 | } else { | |
1850 | const struct mfc_cache *mfc = v; | |
1851 | const struct ipmr_mfc_iter *it = seq->private; | |
e905a9ed | 1852 | |
999890b2 | 1853 | seq_printf(seq, "%08lX %08lX %-3hd", |
1da177e4 LT |
1854 | (unsigned long) mfc->mfc_mcastgrp, |
1855 | (unsigned long) mfc->mfc_origin, | |
1ea472e2 | 1856 | mfc->mfc_parent); |
1da177e4 LT |
1857 | |
1858 | if (it->cache != &mfc_unres_queue) { | |
1ea472e2 BT |
1859 | seq_printf(seq, " %8lu %8lu %8lu", |
1860 | mfc->mfc_un.res.pkt, | |
1861 | mfc->mfc_un.res.bytes, | |
1862 | mfc->mfc_un.res.wrong_if); | |
132adf54 SH |
1863 | for (n = mfc->mfc_un.res.minvif; |
1864 | n < mfc->mfc_un.res.maxvif; n++ ) { | |
cf958ae3 BT |
1865 | if (VIF_EXISTS(&init_net, n) && |
1866 | mfc->mfc_un.res.ttls[n] < 255) | |
1867 | seq_printf(seq, | |
e905a9ed | 1868 | " %2d:%-3d", |
1da177e4 LT |
1869 | n, mfc->mfc_un.res.ttls[n]); |
1870 | } | |
1ea472e2 BT |
1871 | } else { |
1872 | /* unresolved mfc_caches don't contain | |
1873 | * pkt, bytes and wrong_if values | |
1874 | */ | |
1875 | seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); | |
1da177e4 LT |
1876 | } |
1877 | seq_putc(seq, '\n'); | |
1878 | } | |
1879 | return 0; | |
1880 | } | |
1881 | ||
f690808e | 1882 | static const struct seq_operations ipmr_mfc_seq_ops = { |
1da177e4 LT |
1883 | .start = ipmr_mfc_seq_start, |
1884 | .next = ipmr_mfc_seq_next, | |
1885 | .stop = ipmr_mfc_seq_stop, | |
1886 | .show = ipmr_mfc_seq_show, | |
1887 | }; | |
1888 | ||
1889 | static int ipmr_mfc_open(struct inode *inode, struct file *file) | |
1890 | { | |
cf7732e4 PE |
1891 | return seq_open_private(file, &ipmr_mfc_seq_ops, |
1892 | sizeof(struct ipmr_mfc_iter)); | |
1da177e4 LT |
1893 | } |
1894 | ||
9a32144e | 1895 | static const struct file_operations ipmr_mfc_fops = { |
1da177e4 LT |
1896 | .owner = THIS_MODULE, |
1897 | .open = ipmr_mfc_open, | |
1898 | .read = seq_read, | |
1899 | .llseek = seq_lseek, | |
1900 | .release = seq_release_private, | |
1901 | }; | |
e905a9ed | 1902 | #endif |
1da177e4 LT |
1903 | |
1904 | #ifdef CONFIG_IP_PIMSM_V2 | |
1905 | static struct net_protocol pim_protocol = { | |
1906 | .handler = pim_rcv, | |
1907 | }; | |
1908 | #endif | |
1909 | ||
1910 | ||
1911 | /* | |
1912 | * Setup for IP multicast routing | |
1913 | */ | |
cf958ae3 BT |
1914 | static int __net_init ipmr_net_init(struct net *net) |
1915 | { | |
1916 | int err = 0; | |
1917 | ||
1918 | net->ipv4.vif_table = kcalloc(MAXVIFS, sizeof(struct vif_device), | |
1919 | GFP_KERNEL); | |
1920 | if (!net->ipv4.vif_table) { | |
1921 | err = -ENOMEM; | |
1922 | goto fail; | |
1923 | } | |
1924 | fail: | |
1925 | return err; | |
1926 | } | |
1927 | ||
1928 | static void __net_exit ipmr_net_exit(struct net *net) | |
1929 | { | |
1930 | kfree(net->ipv4.vif_table); | |
1931 | } | |
1932 | ||
1933 | static struct pernet_operations ipmr_net_ops = { | |
1934 | .init = ipmr_net_init, | |
1935 | .exit = ipmr_net_exit, | |
1936 | }; | |
e905a9ed | 1937 | |
03d2f897 | 1938 | int __init ip_mr_init(void) |
1da177e4 | 1939 | { |
03d2f897 WC |
1940 | int err; |
1941 | ||
1da177e4 LT |
1942 | mrt_cachep = kmem_cache_create("ip_mrt_cache", |
1943 | sizeof(struct mfc_cache), | |
e5d679f3 | 1944 | 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, |
20c2df83 | 1945 | NULL); |
03d2f897 WC |
1946 | if (!mrt_cachep) |
1947 | return -ENOMEM; | |
1948 | ||
cf958ae3 BT |
1949 | err = register_pernet_subsys(&ipmr_net_ops); |
1950 | if (err) | |
1951 | goto reg_pernet_fail; | |
1952 | ||
b24b8a24 | 1953 | setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0); |
03d2f897 WC |
1954 | err = register_netdevice_notifier(&ip_mr_notifier); |
1955 | if (err) | |
1956 | goto reg_notif_fail; | |
e905a9ed | 1957 | #ifdef CONFIG_PROC_FS |
03d2f897 WC |
1958 | err = -ENOMEM; |
1959 | if (!proc_net_fops_create(&init_net, "ip_mr_vif", 0, &ipmr_vif_fops)) | |
1960 | goto proc_vif_fail; | |
1961 | if (!proc_net_fops_create(&init_net, "ip_mr_cache", 0, &ipmr_mfc_fops)) | |
1962 | goto proc_cache_fail; | |
e905a9ed | 1963 | #endif |
03d2f897 | 1964 | return 0; |
03d2f897 | 1965 | #ifdef CONFIG_PROC_FS |
03d2f897 WC |
1966 | proc_cache_fail: |
1967 | proc_net_remove(&init_net, "ip_mr_vif"); | |
c3e38896 BT |
1968 | proc_vif_fail: |
1969 | unregister_netdevice_notifier(&ip_mr_notifier); | |
03d2f897 | 1970 | #endif |
c3e38896 BT |
1971 | reg_notif_fail: |
1972 | del_timer(&ipmr_expire_timer); | |
cf958ae3 BT |
1973 | unregister_pernet_subsys(&ipmr_net_ops); |
1974 | reg_pernet_fail: | |
c3e38896 | 1975 | kmem_cache_destroy(mrt_cachep); |
03d2f897 | 1976 | return err; |
1da177e4 | 1977 | } |