]>
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 | ||
1da177e4 | 85 | static struct mfc_cache *mfc_unres_queue; /* Queue of unresolved entries */ |
1da177e4 LT |
86 | |
87 | /* Special spinlock for queue of unresolved entries */ | |
88 | static DEFINE_SPINLOCK(mfc_unres_lock); | |
89 | ||
90 | /* We return to original Alan's scheme. Hash table of resolved | |
91 | entries is changed only in process context and protected | |
92 | with weak lock mrt_lock. Queue of unresolved entries is protected | |
93 | with strong spinlock mfc_unres_lock. | |
94 | ||
95 | In this case data path is free of exclusive locks at all. | |
96 | */ | |
97 | ||
e18b890b | 98 | static struct kmem_cache *mrt_cachep __read_mostly; |
1da177e4 LT |
99 | |
100 | static int ip_mr_forward(struct sk_buff *skb, struct mfc_cache *cache, int local); | |
101 | static int ipmr_cache_report(struct sk_buff *pkt, vifi_t vifi, int assert); | |
102 | static int ipmr_fill_mroute(struct sk_buff *skb, struct mfc_cache *c, struct rtmsg *rtm); | |
103 | ||
104 | #ifdef CONFIG_IP_PIMSM_V2 | |
105 | static struct net_protocol pim_protocol; | |
106 | #endif | |
107 | ||
108 | static struct timer_list ipmr_expire_timer; | |
109 | ||
110 | /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */ | |
111 | ||
d607032d WC |
112 | static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v) |
113 | { | |
114 | dev_close(dev); | |
115 | ||
116 | dev = __dev_get_by_name(&init_net, "tunl0"); | |
117 | if (dev) { | |
5bc3eb7e | 118 | const struct net_device_ops *ops = dev->netdev_ops; |
d607032d | 119 | struct ifreq ifr; |
d607032d WC |
120 | struct ip_tunnel_parm p; |
121 | ||
122 | memset(&p, 0, sizeof(p)); | |
123 | p.iph.daddr = v->vifc_rmt_addr.s_addr; | |
124 | p.iph.saddr = v->vifc_lcl_addr.s_addr; | |
125 | p.iph.version = 4; | |
126 | p.iph.ihl = 5; | |
127 | p.iph.protocol = IPPROTO_IPIP; | |
128 | sprintf(p.name, "dvmrp%d", v->vifc_vifi); | |
129 | ifr.ifr_ifru.ifru_data = (__force void __user *)&p; | |
130 | ||
5bc3eb7e SH |
131 | if (ops->ndo_do_ioctl) { |
132 | mm_segment_t oldfs = get_fs(); | |
133 | ||
134 | set_fs(KERNEL_DS); | |
135 | ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL); | |
136 | set_fs(oldfs); | |
137 | } | |
d607032d WC |
138 | } |
139 | } | |
140 | ||
1da177e4 LT |
141 | static |
142 | struct net_device *ipmr_new_tunnel(struct vifctl *v) | |
143 | { | |
144 | struct net_device *dev; | |
145 | ||
881d966b | 146 | dev = __dev_get_by_name(&init_net, "tunl0"); |
1da177e4 LT |
147 | |
148 | if (dev) { | |
5bc3eb7e | 149 | const struct net_device_ops *ops = dev->netdev_ops; |
1da177e4 LT |
150 | int err; |
151 | struct ifreq ifr; | |
1da177e4 LT |
152 | struct ip_tunnel_parm p; |
153 | struct in_device *in_dev; | |
154 | ||
155 | memset(&p, 0, sizeof(p)); | |
156 | p.iph.daddr = v->vifc_rmt_addr.s_addr; | |
157 | p.iph.saddr = v->vifc_lcl_addr.s_addr; | |
158 | p.iph.version = 4; | |
159 | p.iph.ihl = 5; | |
160 | p.iph.protocol = IPPROTO_IPIP; | |
161 | sprintf(p.name, "dvmrp%d", v->vifc_vifi); | |
ba93ef74 | 162 | ifr.ifr_ifru.ifru_data = (__force void __user *)&p; |
1da177e4 | 163 | |
5bc3eb7e SH |
164 | if (ops->ndo_do_ioctl) { |
165 | mm_segment_t oldfs = get_fs(); | |
166 | ||
167 | set_fs(KERNEL_DS); | |
168 | err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL); | |
169 | set_fs(oldfs); | |
170 | } else | |
171 | err = -EOPNOTSUPP; | |
1da177e4 LT |
172 | |
173 | dev = NULL; | |
174 | ||
881d966b | 175 | if (err == 0 && (dev = __dev_get_by_name(&init_net, p.name)) != NULL) { |
1da177e4 LT |
176 | dev->flags |= IFF_MULTICAST; |
177 | ||
e5ed6399 | 178 | in_dev = __in_dev_get_rtnl(dev); |
71e27da9 | 179 | if (in_dev == NULL) |
1da177e4 | 180 | goto failure; |
71e27da9 HX |
181 | |
182 | ipv4_devconf_setall(in_dev); | |
183 | IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0; | |
1da177e4 LT |
184 | |
185 | if (dev_open(dev)) | |
186 | goto failure; | |
7dc00c82 | 187 | dev_hold(dev); |
1da177e4 LT |
188 | } |
189 | } | |
190 | return dev; | |
191 | ||
192 | failure: | |
193 | /* allow the register to be completed before unregistering. */ | |
194 | rtnl_unlock(); | |
195 | rtnl_lock(); | |
196 | ||
197 | unregister_netdevice(dev); | |
198 | return NULL; | |
199 | } | |
200 | ||
201 | #ifdef CONFIG_IP_PIMSM | |
202 | ||
203 | static int reg_vif_num = -1; | |
204 | ||
205 | static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev) | |
206 | { | |
207 | read_lock(&mrt_lock); | |
cf3677ae PE |
208 | dev->stats.tx_bytes += skb->len; |
209 | dev->stats.tx_packets++; | |
1da177e4 LT |
210 | ipmr_cache_report(skb, reg_vif_num, IGMPMSG_WHOLEPKT); |
211 | read_unlock(&mrt_lock); | |
212 | kfree_skb(skb); | |
213 | return 0; | |
214 | } | |
215 | ||
007c3838 SH |
216 | static const struct net_device_ops reg_vif_netdev_ops = { |
217 | .ndo_start_xmit = reg_vif_xmit, | |
218 | }; | |
219 | ||
1da177e4 LT |
220 | static void reg_vif_setup(struct net_device *dev) |
221 | { | |
222 | dev->type = ARPHRD_PIMREG; | |
46f25dff | 223 | dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8; |
1da177e4 | 224 | dev->flags = IFF_NOARP; |
007c3838 | 225 | dev->netdev_ops = ®_vif_netdev_ops, |
1da177e4 LT |
226 | dev->destructor = free_netdev; |
227 | } | |
228 | ||
229 | static struct net_device *ipmr_reg_vif(void) | |
230 | { | |
231 | struct net_device *dev; | |
232 | struct in_device *in_dev; | |
233 | ||
cf3677ae | 234 | dev = alloc_netdev(0, "pimreg", reg_vif_setup); |
1da177e4 LT |
235 | |
236 | if (dev == NULL) | |
237 | return NULL; | |
238 | ||
239 | if (register_netdevice(dev)) { | |
240 | free_netdev(dev); | |
241 | return NULL; | |
242 | } | |
243 | dev->iflink = 0; | |
244 | ||
71e27da9 HX |
245 | rcu_read_lock(); |
246 | if ((in_dev = __in_dev_get_rcu(dev)) == NULL) { | |
247 | rcu_read_unlock(); | |
1da177e4 | 248 | goto failure; |
71e27da9 | 249 | } |
1da177e4 | 250 | |
71e27da9 HX |
251 | ipv4_devconf_setall(in_dev); |
252 | IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0; | |
253 | rcu_read_unlock(); | |
1da177e4 LT |
254 | |
255 | if (dev_open(dev)) | |
256 | goto failure; | |
257 | ||
7dc00c82 WC |
258 | dev_hold(dev); |
259 | ||
1da177e4 LT |
260 | return dev; |
261 | ||
262 | failure: | |
263 | /* allow the register to be completed before unregistering. */ | |
264 | rtnl_unlock(); | |
265 | rtnl_lock(); | |
266 | ||
267 | unregister_netdevice(dev); | |
268 | return NULL; | |
269 | } | |
270 | #endif | |
271 | ||
272 | /* | |
273 | * Delete a VIF entry | |
7dc00c82 | 274 | * @notify: Set to 1, if the caller is a notifier_call |
1da177e4 | 275 | */ |
e905a9ed | 276 | |
7dc00c82 | 277 | static int vif_delete(int vifi, int notify) |
1da177e4 LT |
278 | { |
279 | struct vif_device *v; | |
280 | struct net_device *dev; | |
281 | struct in_device *in_dev; | |
282 | ||
cf958ae3 | 283 | if (vifi < 0 || vifi >= init_net.ipv4.maxvif) |
1da177e4 LT |
284 | return -EADDRNOTAVAIL; |
285 | ||
cf958ae3 | 286 | v = &init_net.ipv4.vif_table[vifi]; |
1da177e4 LT |
287 | |
288 | write_lock_bh(&mrt_lock); | |
289 | dev = v->dev; | |
290 | v->dev = NULL; | |
291 | ||
292 | if (!dev) { | |
293 | write_unlock_bh(&mrt_lock); | |
294 | return -EADDRNOTAVAIL; | |
295 | } | |
296 | ||
297 | #ifdef CONFIG_IP_PIMSM | |
298 | if (vifi == reg_vif_num) | |
299 | reg_vif_num = -1; | |
300 | #endif | |
301 | ||
cf958ae3 | 302 | if (vifi+1 == init_net.ipv4.maxvif) { |
1da177e4 LT |
303 | int tmp; |
304 | for (tmp=vifi-1; tmp>=0; tmp--) { | |
cf958ae3 | 305 | if (VIF_EXISTS(&init_net, tmp)) |
1da177e4 LT |
306 | break; |
307 | } | |
cf958ae3 | 308 | init_net.ipv4.maxvif = tmp+1; |
1da177e4 LT |
309 | } |
310 | ||
311 | write_unlock_bh(&mrt_lock); | |
312 | ||
313 | dev_set_allmulti(dev, -1); | |
314 | ||
e5ed6399 | 315 | if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) { |
42f811b8 | 316 | IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--; |
1da177e4 LT |
317 | ip_rt_multicast_event(in_dev); |
318 | } | |
319 | ||
7dc00c82 | 320 | if (v->flags&(VIFF_TUNNEL|VIFF_REGISTER) && !notify) |
1da177e4 LT |
321 | unregister_netdevice(dev); |
322 | ||
323 | dev_put(dev); | |
324 | return 0; | |
325 | } | |
326 | ||
5c0a66f5 BT |
327 | static inline void ipmr_cache_free(struct mfc_cache *c) |
328 | { | |
329 | release_net(mfc_net(c)); | |
330 | kmem_cache_free(mrt_cachep, c); | |
331 | } | |
332 | ||
1da177e4 LT |
333 | /* Destroy an unresolved cache entry, killing queued skbs |
334 | and reporting error to netlink readers. | |
335 | */ | |
336 | ||
337 | static void ipmr_destroy_unres(struct mfc_cache *c) | |
338 | { | |
339 | struct sk_buff *skb; | |
9ef1d4c7 | 340 | struct nlmsgerr *e; |
1da177e4 | 341 | |
1e8fb3b6 | 342 | atomic_dec(&init_net.ipv4.cache_resolve_queue_len); |
1da177e4 | 343 | |
c354e124 | 344 | while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) { |
eddc9ec5 | 345 | if (ip_hdr(skb)->version == 0) { |
1da177e4 LT |
346 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr)); |
347 | nlh->nlmsg_type = NLMSG_ERROR; | |
348 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
349 | skb_trim(skb, nlh->nlmsg_len); | |
9ef1d4c7 PM |
350 | e = NLMSG_DATA(nlh); |
351 | e->error = -ETIMEDOUT; | |
352 | memset(&e->msg, 0, sizeof(e->msg)); | |
2942e900 | 353 | |
97c53cac | 354 | rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); |
1da177e4 LT |
355 | } else |
356 | kfree_skb(skb); | |
357 | } | |
358 | ||
5c0a66f5 | 359 | ipmr_cache_free(c); |
1da177e4 LT |
360 | } |
361 | ||
362 | ||
363 | /* Single timer process for all the unresolved queue. */ | |
364 | ||
365 | static void ipmr_expire_process(unsigned long dummy) | |
366 | { | |
367 | unsigned long now; | |
368 | unsigned long expires; | |
369 | struct mfc_cache *c, **cp; | |
370 | ||
371 | if (!spin_trylock(&mfc_unres_lock)) { | |
372 | mod_timer(&ipmr_expire_timer, jiffies+HZ/10); | |
373 | return; | |
374 | } | |
375 | ||
1e8fb3b6 | 376 | if (mfc_unres_queue == NULL) |
1da177e4 LT |
377 | goto out; |
378 | ||
379 | now = jiffies; | |
380 | expires = 10*HZ; | |
381 | cp = &mfc_unres_queue; | |
382 | ||
383 | while ((c=*cp) != NULL) { | |
384 | if (time_after(c->mfc_un.unres.expires, now)) { | |
385 | unsigned long interval = c->mfc_un.unres.expires - now; | |
386 | if (interval < expires) | |
387 | expires = interval; | |
388 | cp = &c->next; | |
389 | continue; | |
390 | } | |
391 | ||
392 | *cp = c->next; | |
393 | ||
394 | ipmr_destroy_unres(c); | |
395 | } | |
396 | ||
1e8fb3b6 | 397 | if (mfc_unres_queue != NULL) |
1da177e4 LT |
398 | mod_timer(&ipmr_expire_timer, jiffies + expires); |
399 | ||
400 | out: | |
401 | spin_unlock(&mfc_unres_lock); | |
402 | } | |
403 | ||
404 | /* Fill oifs list. It is called under write locked mrt_lock. */ | |
405 | ||
d1b04c08 | 406 | static void ipmr_update_thresholds(struct mfc_cache *cache, unsigned char *ttls) |
1da177e4 LT |
407 | { |
408 | int vifi; | |
409 | ||
410 | cache->mfc_un.res.minvif = MAXVIFS; | |
411 | cache->mfc_un.res.maxvif = 0; | |
412 | memset(cache->mfc_un.res.ttls, 255, MAXVIFS); | |
413 | ||
cf958ae3 BT |
414 | for (vifi = 0; vifi < init_net.ipv4.maxvif; vifi++) { |
415 | if (VIF_EXISTS(&init_net, vifi) && | |
416 | ttls[vifi] && ttls[vifi] < 255) { | |
1da177e4 LT |
417 | cache->mfc_un.res.ttls[vifi] = ttls[vifi]; |
418 | if (cache->mfc_un.res.minvif > vifi) | |
419 | cache->mfc_un.res.minvif = vifi; | |
420 | if (cache->mfc_un.res.maxvif <= vifi) | |
421 | cache->mfc_un.res.maxvif = vifi + 1; | |
422 | } | |
423 | } | |
424 | } | |
425 | ||
426 | static int vif_add(struct vifctl *vifc, int mrtsock) | |
427 | { | |
428 | int vifi = vifc->vifc_vifi; | |
cf958ae3 | 429 | struct vif_device *v = &init_net.ipv4.vif_table[vifi]; |
1da177e4 LT |
430 | struct net_device *dev; |
431 | struct in_device *in_dev; | |
d607032d | 432 | int err; |
1da177e4 LT |
433 | |
434 | /* Is vif busy ? */ | |
cf958ae3 | 435 | if (VIF_EXISTS(&init_net, vifi)) |
1da177e4 LT |
436 | return -EADDRINUSE; |
437 | ||
438 | switch (vifc->vifc_flags) { | |
439 | #ifdef CONFIG_IP_PIMSM | |
440 | case VIFF_REGISTER: | |
441 | /* | |
442 | * Special Purpose VIF in PIM | |
443 | * All the packets will be sent to the daemon | |
444 | */ | |
445 | if (reg_vif_num >= 0) | |
446 | return -EADDRINUSE; | |
447 | dev = ipmr_reg_vif(); | |
448 | if (!dev) | |
449 | return -ENOBUFS; | |
d607032d WC |
450 | err = dev_set_allmulti(dev, 1); |
451 | if (err) { | |
452 | unregister_netdevice(dev); | |
7dc00c82 | 453 | dev_put(dev); |
d607032d WC |
454 | return err; |
455 | } | |
1da177e4 LT |
456 | break; |
457 | #endif | |
e905a9ed | 458 | case VIFF_TUNNEL: |
1da177e4 LT |
459 | dev = ipmr_new_tunnel(vifc); |
460 | if (!dev) | |
461 | return -ENOBUFS; | |
d607032d WC |
462 | err = dev_set_allmulti(dev, 1); |
463 | if (err) { | |
464 | ipmr_del_tunnel(dev, vifc); | |
7dc00c82 | 465 | dev_put(dev); |
d607032d WC |
466 | return err; |
467 | } | |
1da177e4 LT |
468 | break; |
469 | case 0: | |
1ab35276 | 470 | dev = ip_dev_find(&init_net, vifc->vifc_lcl_addr.s_addr); |
1da177e4 LT |
471 | if (!dev) |
472 | return -EADDRNOTAVAIL; | |
d607032d | 473 | err = dev_set_allmulti(dev, 1); |
7dc00c82 WC |
474 | if (err) { |
475 | dev_put(dev); | |
d607032d | 476 | return err; |
7dc00c82 | 477 | } |
1da177e4 LT |
478 | break; |
479 | default: | |
480 | return -EINVAL; | |
481 | } | |
482 | ||
e5ed6399 | 483 | if ((in_dev = __in_dev_get_rtnl(dev)) == NULL) |
1da177e4 | 484 | return -EADDRNOTAVAIL; |
42f811b8 | 485 | IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++; |
1da177e4 LT |
486 | ip_rt_multicast_event(in_dev); |
487 | ||
488 | /* | |
489 | * Fill in the VIF structures | |
490 | */ | |
c354e124 JK |
491 | v->rate_limit = vifc->vifc_rate_limit; |
492 | v->local = vifc->vifc_lcl_addr.s_addr; | |
493 | v->remote = vifc->vifc_rmt_addr.s_addr; | |
494 | v->flags = vifc->vifc_flags; | |
1da177e4 LT |
495 | if (!mrtsock) |
496 | v->flags |= VIFF_STATIC; | |
c354e124 | 497 | v->threshold = vifc->vifc_threshold; |
1da177e4 LT |
498 | v->bytes_in = 0; |
499 | v->bytes_out = 0; | |
500 | v->pkt_in = 0; | |
501 | v->pkt_out = 0; | |
502 | v->link = dev->ifindex; | |
503 | if (v->flags&(VIFF_TUNNEL|VIFF_REGISTER)) | |
504 | v->link = dev->iflink; | |
505 | ||
506 | /* And finish update writing critical data */ | |
507 | write_lock_bh(&mrt_lock); | |
c354e124 | 508 | v->dev = dev; |
1da177e4 LT |
509 | #ifdef CONFIG_IP_PIMSM |
510 | if (v->flags&VIFF_REGISTER) | |
511 | reg_vif_num = vifi; | |
512 | #endif | |
cf958ae3 BT |
513 | if (vifi+1 > init_net.ipv4.maxvif) |
514 | init_net.ipv4.maxvif = vifi+1; | |
1da177e4 LT |
515 | write_unlock_bh(&mrt_lock); |
516 | return 0; | |
517 | } | |
518 | ||
114c7844 | 519 | static struct mfc_cache *ipmr_cache_find(__be32 origin, __be32 mcastgrp) |
1da177e4 | 520 | { |
c354e124 | 521 | int line = MFC_HASH(mcastgrp, origin); |
1da177e4 LT |
522 | struct mfc_cache *c; |
523 | ||
2bb8b26c | 524 | for (c = init_net.ipv4.mfc_cache_array[line]; c; c = c->next) { |
1da177e4 LT |
525 | if (c->mfc_origin==origin && c->mfc_mcastgrp==mcastgrp) |
526 | break; | |
527 | } | |
528 | return c; | |
529 | } | |
530 | ||
531 | /* | |
532 | * Allocate a multicast cache entry | |
533 | */ | |
5c0a66f5 | 534 | static struct mfc_cache *ipmr_cache_alloc(struct net *net) |
1da177e4 | 535 | { |
c354e124 JK |
536 | struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL); |
537 | if (c == NULL) | |
1da177e4 | 538 | return NULL; |
1da177e4 | 539 | c->mfc_un.res.minvif = MAXVIFS; |
5c0a66f5 | 540 | mfc_net_set(c, net); |
1da177e4 LT |
541 | return c; |
542 | } | |
543 | ||
5c0a66f5 | 544 | static struct mfc_cache *ipmr_cache_alloc_unres(struct net *net) |
1da177e4 | 545 | { |
c354e124 JK |
546 | struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC); |
547 | if (c == NULL) | |
1da177e4 | 548 | return NULL; |
1da177e4 LT |
549 | skb_queue_head_init(&c->mfc_un.unres.unresolved); |
550 | c->mfc_un.unres.expires = jiffies + 10*HZ; | |
5c0a66f5 | 551 | mfc_net_set(c, net); |
1da177e4 LT |
552 | return c; |
553 | } | |
554 | ||
555 | /* | |
556 | * A cache entry has gone into a resolved state from queued | |
557 | */ | |
e905a9ed | 558 | |
1da177e4 LT |
559 | static void ipmr_cache_resolve(struct mfc_cache *uc, struct mfc_cache *c) |
560 | { | |
561 | struct sk_buff *skb; | |
9ef1d4c7 | 562 | struct nlmsgerr *e; |
1da177e4 LT |
563 | |
564 | /* | |
565 | * Play the pending entries through our router | |
566 | */ | |
567 | ||
c354e124 | 568 | while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) { |
eddc9ec5 | 569 | if (ip_hdr(skb)->version == 0) { |
1da177e4 LT |
570 | struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr)); |
571 | ||
572 | if (ipmr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) { | |
27a884dc ACM |
573 | nlh->nlmsg_len = (skb_tail_pointer(skb) - |
574 | (u8 *)nlh); | |
1da177e4 LT |
575 | } else { |
576 | nlh->nlmsg_type = NLMSG_ERROR; | |
577 | nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); | |
578 | skb_trim(skb, nlh->nlmsg_len); | |
9ef1d4c7 PM |
579 | e = NLMSG_DATA(nlh); |
580 | e->error = -EMSGSIZE; | |
581 | memset(&e->msg, 0, sizeof(e->msg)); | |
1da177e4 | 582 | } |
2942e900 | 583 | |
97c53cac | 584 | rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid); |
1da177e4 LT |
585 | } else |
586 | ip_mr_forward(skb, c, 0); | |
587 | } | |
588 | } | |
589 | ||
590 | /* | |
591 | * Bounce a cache query up to mrouted. We could use netlink for this but mrouted | |
592 | * expects the following bizarre scheme. | |
593 | * | |
594 | * Called under mrt_lock. | |
595 | */ | |
e905a9ed | 596 | |
1da177e4 LT |
597 | static int ipmr_cache_report(struct sk_buff *pkt, vifi_t vifi, int assert) |
598 | { | |
599 | struct sk_buff *skb; | |
c9bdd4b5 | 600 | const int ihl = ip_hdrlen(pkt); |
1da177e4 LT |
601 | struct igmphdr *igmp; |
602 | struct igmpmsg *msg; | |
603 | int ret; | |
604 | ||
605 | #ifdef CONFIG_IP_PIMSM | |
606 | if (assert == IGMPMSG_WHOLEPKT) | |
607 | skb = skb_realloc_headroom(pkt, sizeof(struct iphdr)); | |
608 | else | |
609 | #endif | |
610 | skb = alloc_skb(128, GFP_ATOMIC); | |
611 | ||
132adf54 | 612 | if (!skb) |
1da177e4 LT |
613 | return -ENOBUFS; |
614 | ||
615 | #ifdef CONFIG_IP_PIMSM | |
616 | if (assert == IGMPMSG_WHOLEPKT) { | |
617 | /* Ugly, but we have no choice with this interface. | |
618 | Duplicate old header, fix ihl, length etc. | |
619 | And all this only to mangle msg->im_msgtype and | |
620 | to set msg->im_mbz to "mbz" :-) | |
621 | */ | |
878c8145 ACM |
622 | skb_push(skb, sizeof(struct iphdr)); |
623 | skb_reset_network_header(skb); | |
badff6d0 | 624 | skb_reset_transport_header(skb); |
0272ffc4 | 625 | msg = (struct igmpmsg *)skb_network_header(skb); |
d56f90a7 | 626 | memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr)); |
1da177e4 LT |
627 | msg->im_msgtype = IGMPMSG_WHOLEPKT; |
628 | msg->im_mbz = 0; | |
e905a9ed | 629 | msg->im_vif = reg_vif_num; |
eddc9ec5 ACM |
630 | ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2; |
631 | ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) + | |
632 | sizeof(struct iphdr)); | |
e905a9ed | 633 | } else |
1da177e4 | 634 | #endif |
e905a9ed YH |
635 | { |
636 | ||
1da177e4 LT |
637 | /* |
638 | * Copy the IP header | |
639 | */ | |
640 | ||
27a884dc | 641 | skb->network_header = skb->tail; |
ddc7b8e3 | 642 | skb_put(skb, ihl); |
27d7ff46 | 643 | skb_copy_to_linear_data(skb, pkt->data, ihl); |
eddc9ec5 ACM |
644 | ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */ |
645 | msg = (struct igmpmsg *)skb_network_header(skb); | |
1da177e4 LT |
646 | msg->im_vif = vifi; |
647 | skb->dst = dst_clone(pkt->dst); | |
648 | ||
649 | /* | |
650 | * Add our header | |
651 | */ | |
652 | ||
c354e124 | 653 | igmp=(struct igmphdr *)skb_put(skb, sizeof(struct igmphdr)); |
1da177e4 LT |
654 | igmp->type = |
655 | msg->im_msgtype = assert; | |
656 | igmp->code = 0; | |
eddc9ec5 | 657 | ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */ |
b0e380b1 | 658 | skb->transport_header = skb->network_header; |
e905a9ed | 659 | } |
1da177e4 | 660 | |
70a269e6 | 661 | if (init_net.ipv4.mroute_sk == NULL) { |
1da177e4 LT |
662 | kfree_skb(skb); |
663 | return -EINVAL; | |
664 | } | |
665 | ||
666 | /* | |
667 | * Deliver to mrouted | |
668 | */ | |
70a269e6 BT |
669 | ret = sock_queue_rcv_skb(init_net.ipv4.mroute_sk, skb); |
670 | if (ret < 0) { | |
1da177e4 LT |
671 | if (net_ratelimit()) |
672 | printk(KERN_WARNING "mroute: pending queue full, dropping entries.\n"); | |
673 | kfree_skb(skb); | |
674 | } | |
675 | ||
676 | return ret; | |
677 | } | |
678 | ||
679 | /* | |
680 | * Queue a packet for resolution. It gets locked cache entry! | |
681 | */ | |
e905a9ed | 682 | |
1da177e4 LT |
683 | static int |
684 | ipmr_cache_unresolved(vifi_t vifi, struct sk_buff *skb) | |
685 | { | |
686 | int err; | |
687 | struct mfc_cache *c; | |
eddc9ec5 | 688 | const struct iphdr *iph = ip_hdr(skb); |
1da177e4 LT |
689 | |
690 | spin_lock_bh(&mfc_unres_lock); | |
691 | for (c=mfc_unres_queue; c; c=c->next) { | |
1e8fb3b6 BT |
692 | if (net_eq(mfc_net(c), &init_net) && |
693 | c->mfc_mcastgrp == iph->daddr && | |
eddc9ec5 | 694 | c->mfc_origin == iph->saddr) |
1da177e4 LT |
695 | break; |
696 | } | |
697 | ||
698 | if (c == NULL) { | |
699 | /* | |
700 | * Create a new entry if allowable | |
701 | */ | |
702 | ||
1e8fb3b6 | 703 | if (atomic_read(&init_net.ipv4.cache_resolve_queue_len) >= 10 || |
5c0a66f5 | 704 | (c = ipmr_cache_alloc_unres(&init_net)) == NULL) { |
1da177e4 LT |
705 | spin_unlock_bh(&mfc_unres_lock); |
706 | ||
707 | kfree_skb(skb); | |
708 | return -ENOBUFS; | |
709 | } | |
710 | ||
711 | /* | |
712 | * Fill in the new cache entry | |
713 | */ | |
eddc9ec5 ACM |
714 | c->mfc_parent = -1; |
715 | c->mfc_origin = iph->saddr; | |
716 | c->mfc_mcastgrp = iph->daddr; | |
1da177e4 LT |
717 | |
718 | /* | |
719 | * Reflect first query at mrouted. | |
720 | */ | |
721 | if ((err = ipmr_cache_report(skb, vifi, IGMPMSG_NOCACHE))<0) { | |
e905a9ed | 722 | /* If the report failed throw the cache entry |
1da177e4 LT |
723 | out - Brad Parker |
724 | */ | |
725 | spin_unlock_bh(&mfc_unres_lock); | |
726 | ||
5c0a66f5 | 727 | ipmr_cache_free(c); |
1da177e4 LT |
728 | kfree_skb(skb); |
729 | return err; | |
730 | } | |
731 | ||
1e8fb3b6 | 732 | atomic_inc(&init_net.ipv4.cache_resolve_queue_len); |
1da177e4 LT |
733 | c->next = mfc_unres_queue; |
734 | mfc_unres_queue = c; | |
735 | ||
736 | mod_timer(&ipmr_expire_timer, c->mfc_un.unres.expires); | |
737 | } | |
738 | ||
739 | /* | |
740 | * See if we can append the packet | |
741 | */ | |
742 | if (c->mfc_un.unres.unresolved.qlen>3) { | |
743 | kfree_skb(skb); | |
744 | err = -ENOBUFS; | |
745 | } else { | |
c354e124 | 746 | skb_queue_tail(&c->mfc_un.unres.unresolved, skb); |
1da177e4 LT |
747 | err = 0; |
748 | } | |
749 | ||
750 | spin_unlock_bh(&mfc_unres_lock); | |
751 | return err; | |
752 | } | |
753 | ||
754 | /* | |
755 | * MFC cache manipulation by user space mroute daemon | |
756 | */ | |
757 | ||
758 | static int ipmr_mfc_delete(struct mfcctl *mfc) | |
759 | { | |
760 | int line; | |
761 | struct mfc_cache *c, **cp; | |
762 | ||
c354e124 | 763 | line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr); |
1da177e4 | 764 | |
2bb8b26c BT |
765 | for (cp = &init_net.ipv4.mfc_cache_array[line]; |
766 | (c = *cp) != NULL; cp = &c->next) { | |
1da177e4 LT |
767 | if (c->mfc_origin == mfc->mfcc_origin.s_addr && |
768 | c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr) { | |
769 | write_lock_bh(&mrt_lock); | |
770 | *cp = c->next; | |
771 | write_unlock_bh(&mrt_lock); | |
772 | ||
5c0a66f5 | 773 | ipmr_cache_free(c); |
1da177e4 LT |
774 | return 0; |
775 | } | |
776 | } | |
777 | return -ENOENT; | |
778 | } | |
779 | ||
780 | static int ipmr_mfc_add(struct mfcctl *mfc, int mrtsock) | |
781 | { | |
782 | int line; | |
783 | struct mfc_cache *uc, *c, **cp; | |
784 | ||
c354e124 | 785 | line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr); |
1da177e4 | 786 | |
2bb8b26c BT |
787 | for (cp = &init_net.ipv4.mfc_cache_array[line]; |
788 | (c = *cp) != NULL; cp = &c->next) { | |
1da177e4 LT |
789 | if (c->mfc_origin == mfc->mfcc_origin.s_addr && |
790 | c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr) | |
791 | break; | |
792 | } | |
793 | ||
794 | if (c != NULL) { | |
795 | write_lock_bh(&mrt_lock); | |
796 | c->mfc_parent = mfc->mfcc_parent; | |
d1b04c08 | 797 | ipmr_update_thresholds(c, mfc->mfcc_ttls); |
1da177e4 LT |
798 | if (!mrtsock) |
799 | c->mfc_flags |= MFC_STATIC; | |
800 | write_unlock_bh(&mrt_lock); | |
801 | return 0; | |
802 | } | |
803 | ||
f97c1e0c | 804 | if (!ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr)) |
1da177e4 LT |
805 | return -EINVAL; |
806 | ||
5c0a66f5 | 807 | c = ipmr_cache_alloc(&init_net); |
c354e124 | 808 | if (c == NULL) |
1da177e4 LT |
809 | return -ENOMEM; |
810 | ||
c354e124 JK |
811 | c->mfc_origin = mfc->mfcc_origin.s_addr; |
812 | c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr; | |
813 | c->mfc_parent = mfc->mfcc_parent; | |
d1b04c08 | 814 | ipmr_update_thresholds(c, mfc->mfcc_ttls); |
1da177e4 LT |
815 | if (!mrtsock) |
816 | c->mfc_flags |= MFC_STATIC; | |
817 | ||
818 | write_lock_bh(&mrt_lock); | |
2bb8b26c BT |
819 | c->next = init_net.ipv4.mfc_cache_array[line]; |
820 | init_net.ipv4.mfc_cache_array[line] = c; | |
1da177e4 LT |
821 | write_unlock_bh(&mrt_lock); |
822 | ||
823 | /* | |
824 | * Check to see if we resolved a queued list. If so we | |
825 | * need to send on the frames and tidy up. | |
826 | */ | |
827 | spin_lock_bh(&mfc_unres_lock); | |
828 | for (cp = &mfc_unres_queue; (uc=*cp) != NULL; | |
829 | cp = &uc->next) { | |
1e8fb3b6 BT |
830 | if (net_eq(mfc_net(uc), &init_net) && |
831 | uc->mfc_origin == c->mfc_origin && | |
1da177e4 LT |
832 | uc->mfc_mcastgrp == c->mfc_mcastgrp) { |
833 | *cp = uc->next; | |
1e8fb3b6 | 834 | atomic_dec(&init_net.ipv4.cache_resolve_queue_len); |
1da177e4 LT |
835 | break; |
836 | } | |
837 | } | |
1e8fb3b6 BT |
838 | if (mfc_unres_queue == NULL) |
839 | del_timer(&ipmr_expire_timer); | |
1da177e4 LT |
840 | spin_unlock_bh(&mfc_unres_lock); |
841 | ||
842 | if (uc) { | |
843 | ipmr_cache_resolve(uc, c); | |
5c0a66f5 | 844 | ipmr_cache_free(uc); |
1da177e4 LT |
845 | } |
846 | return 0; | |
847 | } | |
848 | ||
849 | /* | |
850 | * Close the multicast socket, and clear the vif tables etc | |
851 | */ | |
e905a9ed | 852 | |
1da177e4 LT |
853 | static void mroute_clean_tables(struct sock *sk) |
854 | { | |
855 | int i; | |
e905a9ed | 856 | |
1da177e4 LT |
857 | /* |
858 | * Shut down all active vif entries | |
859 | */ | |
cf958ae3 BT |
860 | for (i = 0; i < init_net.ipv4.maxvif; i++) { |
861 | if (!(init_net.ipv4.vif_table[i].flags&VIFF_STATIC)) | |
7dc00c82 | 862 | vif_delete(i, 0); |
1da177e4 LT |
863 | } |
864 | ||
865 | /* | |
866 | * Wipe the cache | |
867 | */ | |
c354e124 | 868 | for (i=0; i<MFC_LINES; i++) { |
1da177e4 LT |
869 | struct mfc_cache *c, **cp; |
870 | ||
2bb8b26c | 871 | cp = &init_net.ipv4.mfc_cache_array[i]; |
1da177e4 LT |
872 | while ((c = *cp) != NULL) { |
873 | if (c->mfc_flags&MFC_STATIC) { | |
874 | cp = &c->next; | |
875 | continue; | |
876 | } | |
877 | write_lock_bh(&mrt_lock); | |
878 | *cp = c->next; | |
879 | write_unlock_bh(&mrt_lock); | |
880 | ||
5c0a66f5 | 881 | ipmr_cache_free(c); |
1da177e4 LT |
882 | } |
883 | } | |
884 | ||
1e8fb3b6 BT |
885 | if (atomic_read(&init_net.ipv4.cache_resolve_queue_len) != 0) { |
886 | struct mfc_cache *c, **cp; | |
1da177e4 LT |
887 | |
888 | spin_lock_bh(&mfc_unres_lock); | |
1e8fb3b6 BT |
889 | cp = &mfc_unres_queue; |
890 | while ((c = *cp) != NULL) { | |
891 | if (!net_eq(mfc_net(c), &init_net)) { | |
892 | cp = &c->next; | |
893 | continue; | |
894 | } | |
895 | *cp = c->next; | |
1da177e4 LT |
896 | |
897 | ipmr_destroy_unres(c); | |
1da177e4 LT |
898 | } |
899 | spin_unlock_bh(&mfc_unres_lock); | |
900 | } | |
901 | } | |
902 | ||
903 | static void mrtsock_destruct(struct sock *sk) | |
904 | { | |
905 | rtnl_lock(); | |
70a269e6 | 906 | if (sk == init_net.ipv4.mroute_sk) { |
3b1e0a65 | 907 | IPV4_DEVCONF_ALL(sock_net(sk), MC_FORWARDING)--; |
1da177e4 LT |
908 | |
909 | write_lock_bh(&mrt_lock); | |
70a269e6 | 910 | init_net.ipv4.mroute_sk = NULL; |
1da177e4 LT |
911 | write_unlock_bh(&mrt_lock); |
912 | ||
913 | mroute_clean_tables(sk); | |
914 | } | |
915 | rtnl_unlock(); | |
916 | } | |
917 | ||
918 | /* | |
919 | * Socket options and virtual interface manipulation. The whole | |
920 | * virtual interface system is a complete heap, but unfortunately | |
921 | * that's how BSD mrouted happens to think. Maybe one day with a proper | |
922 | * MOSPF/PIM router set up we can clean this up. | |
923 | */ | |
e905a9ed | 924 | |
c354e124 | 925 | int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, int optlen) |
1da177e4 LT |
926 | { |
927 | int ret; | |
928 | struct vifctl vif; | |
929 | struct mfcctl mfc; | |
e905a9ed | 930 | |
132adf54 | 931 | if (optname != MRT_INIT) { |
70a269e6 | 932 | if (sk != init_net.ipv4.mroute_sk && !capable(CAP_NET_ADMIN)) |
1da177e4 LT |
933 | return -EACCES; |
934 | } | |
935 | ||
132adf54 SH |
936 | switch (optname) { |
937 | case MRT_INIT: | |
938 | if (sk->sk_type != SOCK_RAW || | |
939 | inet_sk(sk)->num != IPPROTO_IGMP) | |
940 | return -EOPNOTSUPP; | |
c354e124 | 941 | if (optlen != sizeof(int)) |
132adf54 | 942 | return -ENOPROTOOPT; |
1da177e4 | 943 | |
132adf54 | 944 | rtnl_lock(); |
70a269e6 | 945 | if (init_net.ipv4.mroute_sk) { |
1da177e4 | 946 | rtnl_unlock(); |
132adf54 SH |
947 | return -EADDRINUSE; |
948 | } | |
949 | ||
950 | ret = ip_ra_control(sk, 1, mrtsock_destruct); | |
951 | if (ret == 0) { | |
952 | write_lock_bh(&mrt_lock); | |
70a269e6 | 953 | init_net.ipv4.mroute_sk = sk; |
132adf54 SH |
954 | write_unlock_bh(&mrt_lock); |
955 | ||
3b1e0a65 | 956 | IPV4_DEVCONF_ALL(sock_net(sk), MC_FORWARDING)++; |
132adf54 SH |
957 | } |
958 | rtnl_unlock(); | |
959 | return ret; | |
960 | case MRT_DONE: | |
70a269e6 | 961 | if (sk != init_net.ipv4.mroute_sk) |
132adf54 SH |
962 | return -EACCES; |
963 | return ip_ra_control(sk, 0, NULL); | |
964 | case MRT_ADD_VIF: | |
965 | case MRT_DEL_VIF: | |
c354e124 | 966 | if (optlen != sizeof(vif)) |
132adf54 | 967 | return -EINVAL; |
c354e124 | 968 | if (copy_from_user(&vif, optval, sizeof(vif))) |
132adf54 SH |
969 | return -EFAULT; |
970 | if (vif.vifc_vifi >= MAXVIFS) | |
971 | return -ENFILE; | |
972 | rtnl_lock(); | |
c354e124 | 973 | if (optname == MRT_ADD_VIF) { |
70a269e6 | 974 | ret = vif_add(&vif, sk == init_net.ipv4.mroute_sk); |
132adf54 | 975 | } else { |
7dc00c82 | 976 | ret = vif_delete(vif.vifc_vifi, 0); |
132adf54 SH |
977 | } |
978 | rtnl_unlock(); | |
979 | return ret; | |
1da177e4 LT |
980 | |
981 | /* | |
982 | * Manipulate the forwarding caches. These live | |
983 | * in a sort of kernel/user symbiosis. | |
984 | */ | |
132adf54 SH |
985 | case MRT_ADD_MFC: |
986 | case MRT_DEL_MFC: | |
c354e124 | 987 | if (optlen != sizeof(mfc)) |
132adf54 | 988 | return -EINVAL; |
c354e124 | 989 | if (copy_from_user(&mfc, optval, sizeof(mfc))) |
132adf54 SH |
990 | return -EFAULT; |
991 | rtnl_lock(); | |
c354e124 | 992 | if (optname == MRT_DEL_MFC) |
132adf54 SH |
993 | ret = ipmr_mfc_delete(&mfc); |
994 | else | |
70a269e6 | 995 | ret = ipmr_mfc_add(&mfc, sk == init_net.ipv4.mroute_sk); |
132adf54 SH |
996 | rtnl_unlock(); |
997 | return ret; | |
1da177e4 LT |
998 | /* |
999 | * Control PIM assert. | |
1000 | */ | |
132adf54 SH |
1001 | case MRT_ASSERT: |
1002 | { | |
1003 | int v; | |
1004 | if (get_user(v,(int __user *)optval)) | |
1005 | return -EFAULT; | |
1006 | mroute_do_assert=(v)?1:0; | |
1007 | return 0; | |
1008 | } | |
1da177e4 | 1009 | #ifdef CONFIG_IP_PIMSM |
132adf54 SH |
1010 | case MRT_PIM: |
1011 | { | |
ba93ef74 SH |
1012 | int v; |
1013 | ||
132adf54 SH |
1014 | if (get_user(v,(int __user *)optval)) |
1015 | return -EFAULT; | |
ba93ef74 SH |
1016 | v = (v) ? 1 : 0; |
1017 | ||
132adf54 SH |
1018 | rtnl_lock(); |
1019 | ret = 0; | |
1020 | if (v != mroute_do_pim) { | |
1021 | mroute_do_pim = v; | |
1022 | mroute_do_assert = v; | |
1da177e4 | 1023 | #ifdef CONFIG_IP_PIMSM_V2 |
132adf54 SH |
1024 | if (mroute_do_pim) |
1025 | ret = inet_add_protocol(&pim_protocol, | |
1026 | IPPROTO_PIM); | |
1027 | else | |
1028 | ret = inet_del_protocol(&pim_protocol, | |
1029 | IPPROTO_PIM); | |
1030 | if (ret < 0) | |
1031 | ret = -EAGAIN; | |
1da177e4 | 1032 | #endif |
1da177e4 | 1033 | } |
132adf54 SH |
1034 | rtnl_unlock(); |
1035 | return ret; | |
1036 | } | |
1da177e4 | 1037 | #endif |
132adf54 SH |
1038 | /* |
1039 | * Spurious command, or MRT_VERSION which you cannot | |
1040 | * set. | |
1041 | */ | |
1042 | default: | |
1043 | return -ENOPROTOOPT; | |
1da177e4 LT |
1044 | } |
1045 | } | |
1046 | ||
1047 | /* | |
1048 | * Getsock opt support for the multicast routing system. | |
1049 | */ | |
e905a9ed | 1050 | |
c354e124 | 1051 | int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen) |
1da177e4 LT |
1052 | { |
1053 | int olr; | |
1054 | int val; | |
1055 | ||
c354e124 | 1056 | if (optname != MRT_VERSION && |
1da177e4 LT |
1057 | #ifdef CONFIG_IP_PIMSM |
1058 | optname!=MRT_PIM && | |
1059 | #endif | |
1060 | optname!=MRT_ASSERT) | |
1061 | return -ENOPROTOOPT; | |
1062 | ||
1063 | if (get_user(olr, optlen)) | |
1064 | return -EFAULT; | |
1065 | ||
1066 | olr = min_t(unsigned int, olr, sizeof(int)); | |
1067 | if (olr < 0) | |
1068 | return -EINVAL; | |
e905a9ed | 1069 | |
c354e124 | 1070 | if (put_user(olr, optlen)) |
1da177e4 | 1071 | return -EFAULT; |
c354e124 JK |
1072 | if (optname == MRT_VERSION) |
1073 | val = 0x0305; | |
1da177e4 | 1074 | #ifdef CONFIG_IP_PIMSM |
c354e124 JK |
1075 | else if (optname == MRT_PIM) |
1076 | val = mroute_do_pim; | |
1da177e4 LT |
1077 | #endif |
1078 | else | |
c354e124 JK |
1079 | val = mroute_do_assert; |
1080 | if (copy_to_user(optval, &val, olr)) | |
1da177e4 LT |
1081 | return -EFAULT; |
1082 | return 0; | |
1083 | } | |
1084 | ||
1085 | /* | |
1086 | * The IP multicast ioctl support routines. | |
1087 | */ | |
e905a9ed | 1088 | |
1da177e4 LT |
1089 | int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg) |
1090 | { | |
1091 | struct sioc_sg_req sr; | |
1092 | struct sioc_vif_req vr; | |
1093 | struct vif_device *vif; | |
1094 | struct mfc_cache *c; | |
e905a9ed | 1095 | |
132adf54 SH |
1096 | switch (cmd) { |
1097 | case SIOCGETVIFCNT: | |
c354e124 | 1098 | if (copy_from_user(&vr, arg, sizeof(vr))) |
132adf54 | 1099 | return -EFAULT; |
cf958ae3 | 1100 | if (vr.vifi >= init_net.ipv4.maxvif) |
132adf54 SH |
1101 | return -EINVAL; |
1102 | read_lock(&mrt_lock); | |
cf958ae3 BT |
1103 | vif = &init_net.ipv4.vif_table[vr.vifi]; |
1104 | if (VIF_EXISTS(&init_net, vr.vifi)) { | |
c354e124 JK |
1105 | vr.icount = vif->pkt_in; |
1106 | vr.ocount = vif->pkt_out; | |
1107 | vr.ibytes = vif->bytes_in; | |
1108 | vr.obytes = vif->bytes_out; | |
1da177e4 | 1109 | read_unlock(&mrt_lock); |
1da177e4 | 1110 | |
c354e124 | 1111 | if (copy_to_user(arg, &vr, sizeof(vr))) |
132adf54 SH |
1112 | return -EFAULT; |
1113 | return 0; | |
1114 | } | |
1115 | read_unlock(&mrt_lock); | |
1116 | return -EADDRNOTAVAIL; | |
1117 | case SIOCGETSGCNT: | |
c354e124 | 1118 | if (copy_from_user(&sr, arg, sizeof(sr))) |
132adf54 SH |
1119 | return -EFAULT; |
1120 | ||
1121 | read_lock(&mrt_lock); | |
1122 | c = ipmr_cache_find(sr.src.s_addr, sr.grp.s_addr); | |
1123 | if (c) { | |
1124 | sr.pktcnt = c->mfc_un.res.pkt; | |
1125 | sr.bytecnt = c->mfc_un.res.bytes; | |
1126 | sr.wrong_if = c->mfc_un.res.wrong_if; | |
1da177e4 | 1127 | read_unlock(&mrt_lock); |
132adf54 | 1128 | |
c354e124 | 1129 | if (copy_to_user(arg, &sr, sizeof(sr))) |
132adf54 SH |
1130 | return -EFAULT; |
1131 | return 0; | |
1132 | } | |
1133 | read_unlock(&mrt_lock); | |
1134 | return -EADDRNOTAVAIL; | |
1135 | default: | |
1136 | return -ENOIOCTLCMD; | |
1da177e4 LT |
1137 | } |
1138 | } | |
1139 | ||
1140 | ||
1141 | static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr) | |
1142 | { | |
e9dc8653 | 1143 | struct net_device *dev = ptr; |
1da177e4 LT |
1144 | struct vif_device *v; |
1145 | int ct; | |
e9dc8653 | 1146 | |
721499e8 | 1147 | if (!net_eq(dev_net(dev), &init_net)) |
e9dc8653 EB |
1148 | return NOTIFY_DONE; |
1149 | ||
1da177e4 LT |
1150 | if (event != NETDEV_UNREGISTER) |
1151 | return NOTIFY_DONE; | |
cf958ae3 BT |
1152 | v = &init_net.ipv4.vif_table[0]; |
1153 | for (ct = 0; ct < init_net.ipv4.maxvif; ct++, v++) { | |
c354e124 | 1154 | if (v->dev == dev) |
7dc00c82 | 1155 | vif_delete(ct, 1); |
1da177e4 LT |
1156 | } |
1157 | return NOTIFY_DONE; | |
1158 | } | |
1159 | ||
1160 | ||
c354e124 | 1161 | static struct notifier_block ip_mr_notifier = { |
1da177e4 LT |
1162 | .notifier_call = ipmr_device_event, |
1163 | }; | |
1164 | ||
1165 | /* | |
1166 | * Encapsulate a packet by attaching a valid IPIP header to it. | |
1167 | * This avoids tunnel drivers and other mess and gives us the speed so | |
1168 | * important for multicast video. | |
1169 | */ | |
e905a9ed | 1170 | |
114c7844 | 1171 | static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr) |
1da177e4 | 1172 | { |
8856dfa3 | 1173 | struct iphdr *iph; |
eddc9ec5 | 1174 | struct iphdr *old_iph = ip_hdr(skb); |
8856dfa3 ACM |
1175 | |
1176 | skb_push(skb, sizeof(struct iphdr)); | |
b0e380b1 | 1177 | skb->transport_header = skb->network_header; |
8856dfa3 | 1178 | skb_reset_network_header(skb); |
eddc9ec5 | 1179 | iph = ip_hdr(skb); |
1da177e4 LT |
1180 | |
1181 | iph->version = 4; | |
e023dd64 ACM |
1182 | iph->tos = old_iph->tos; |
1183 | iph->ttl = old_iph->ttl; | |
1da177e4 LT |
1184 | iph->frag_off = 0; |
1185 | iph->daddr = daddr; | |
1186 | iph->saddr = saddr; | |
1187 | iph->protocol = IPPROTO_IPIP; | |
1188 | iph->ihl = 5; | |
1189 | iph->tot_len = htons(skb->len); | |
1190 | ip_select_ident(iph, skb->dst, NULL); | |
1191 | ip_send_check(iph); | |
1192 | ||
1da177e4 LT |
1193 | memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); |
1194 | nf_reset(skb); | |
1195 | } | |
1196 | ||
1197 | static inline int ipmr_forward_finish(struct sk_buff *skb) | |
1198 | { | |
1199 | struct ip_options * opt = &(IPCB(skb)->opt); | |
1200 | ||
7c73a6fa | 1201 | IP_INC_STATS_BH(dev_net(skb->dst->dev), IPSTATS_MIB_OUTFORWDATAGRAMS); |
1da177e4 LT |
1202 | |
1203 | if (unlikely(opt->optlen)) | |
1204 | ip_forward_options(skb); | |
1205 | ||
1206 | return dst_output(skb); | |
1207 | } | |
1208 | ||
1209 | /* | |
1210 | * Processing handlers for ipmr_forward | |
1211 | */ | |
1212 | ||
1213 | static void ipmr_queue_xmit(struct sk_buff *skb, struct mfc_cache *c, int vifi) | |
1214 | { | |
eddc9ec5 | 1215 | const struct iphdr *iph = ip_hdr(skb); |
cf958ae3 | 1216 | struct vif_device *vif = &init_net.ipv4.vif_table[vifi]; |
1da177e4 LT |
1217 | struct net_device *dev; |
1218 | struct rtable *rt; | |
1219 | int encap = 0; | |
1220 | ||
1221 | if (vif->dev == NULL) | |
1222 | goto out_free; | |
1223 | ||
1224 | #ifdef CONFIG_IP_PIMSM | |
1225 | if (vif->flags & VIFF_REGISTER) { | |
1226 | vif->pkt_out++; | |
c354e124 | 1227 | vif->bytes_out += skb->len; |
cf3677ae PE |
1228 | vif->dev->stats.tx_bytes += skb->len; |
1229 | vif->dev->stats.tx_packets++; | |
1da177e4 LT |
1230 | ipmr_cache_report(skb, vifi, IGMPMSG_WHOLEPKT); |
1231 | kfree_skb(skb); | |
1232 | return; | |
1233 | } | |
1234 | #endif | |
1235 | ||
1236 | if (vif->flags&VIFF_TUNNEL) { | |
1237 | struct flowi fl = { .oif = vif->link, | |
1238 | .nl_u = { .ip4_u = | |
1239 | { .daddr = vif->remote, | |
1240 | .saddr = vif->local, | |
1241 | .tos = RT_TOS(iph->tos) } }, | |
1242 | .proto = IPPROTO_IPIP }; | |
f206351a | 1243 | if (ip_route_output_key(&init_net, &rt, &fl)) |
1da177e4 LT |
1244 | goto out_free; |
1245 | encap = sizeof(struct iphdr); | |
1246 | } else { | |
1247 | struct flowi fl = { .oif = vif->link, | |
1248 | .nl_u = { .ip4_u = | |
1249 | { .daddr = iph->daddr, | |
1250 | .tos = RT_TOS(iph->tos) } }, | |
1251 | .proto = IPPROTO_IPIP }; | |
f206351a | 1252 | if (ip_route_output_key(&init_net, &rt, &fl)) |
1da177e4 LT |
1253 | goto out_free; |
1254 | } | |
1255 | ||
1256 | dev = rt->u.dst.dev; | |
1257 | ||
1258 | if (skb->len+encap > dst_mtu(&rt->u.dst) && (ntohs(iph->frag_off) & IP_DF)) { | |
1259 | /* Do not fragment multicasts. Alas, IPv4 does not | |
1260 | allow to send ICMP, so that packets will disappear | |
1261 | to blackhole. | |
1262 | */ | |
1263 | ||
7c73a6fa | 1264 | IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS); |
1da177e4 LT |
1265 | ip_rt_put(rt); |
1266 | goto out_free; | |
1267 | } | |
1268 | ||
1269 | encap += LL_RESERVED_SPACE(dev) + rt->u.dst.header_len; | |
1270 | ||
1271 | if (skb_cow(skb, encap)) { | |
e905a9ed | 1272 | ip_rt_put(rt); |
1da177e4 LT |
1273 | goto out_free; |
1274 | } | |
1275 | ||
1276 | vif->pkt_out++; | |
c354e124 | 1277 | vif->bytes_out += skb->len; |
1da177e4 LT |
1278 | |
1279 | dst_release(skb->dst); | |
1280 | skb->dst = &rt->u.dst; | |
eddc9ec5 | 1281 | ip_decrease_ttl(ip_hdr(skb)); |
1da177e4 LT |
1282 | |
1283 | /* FIXME: forward and output firewalls used to be called here. | |
1284 | * What do we do with netfilter? -- RR */ | |
1285 | if (vif->flags & VIFF_TUNNEL) { | |
1286 | ip_encap(skb, vif->local, vif->remote); | |
1287 | /* FIXME: extra output firewall step used to be here. --RR */ | |
2f4c02d4 PE |
1288 | vif->dev->stats.tx_packets++; |
1289 | vif->dev->stats.tx_bytes += skb->len; | |
1da177e4 LT |
1290 | } |
1291 | ||
1292 | IPCB(skb)->flags |= IPSKB_FORWARDED; | |
1293 | ||
1294 | /* | |
1295 | * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally | |
1296 | * not only before forwarding, but after forwarding on all output | |
1297 | * interfaces. It is clear, if mrouter runs a multicasting | |
1298 | * program, it should receive packets not depending to what interface | |
1299 | * program is joined. | |
1300 | * If we will not make it, the program will have to join on all | |
1301 | * interfaces. On the other hand, multihoming host (or router, but | |
1302 | * not mrouter) cannot join to more than one interface - it will | |
1303 | * result in receiving multiple packets. | |
1304 | */ | |
6e23ae2a | 1305 | NF_HOOK(PF_INET, NF_INET_FORWARD, skb, skb->dev, dev, |
1da177e4 LT |
1306 | ipmr_forward_finish); |
1307 | return; | |
1308 | ||
1309 | out_free: | |
1310 | kfree_skb(skb); | |
1311 | return; | |
1312 | } | |
1313 | ||
1314 | static int ipmr_find_vif(struct net_device *dev) | |
1315 | { | |
1316 | int ct; | |
cf958ae3 BT |
1317 | for (ct = init_net.ipv4.maxvif-1; ct >= 0; ct--) { |
1318 | if (init_net.ipv4.vif_table[ct].dev == dev) | |
1da177e4 LT |
1319 | break; |
1320 | } | |
1321 | return ct; | |
1322 | } | |
1323 | ||
1324 | /* "local" means that we should preserve one skb (for local delivery) */ | |
1325 | ||
1326 | static int ip_mr_forward(struct sk_buff *skb, struct mfc_cache *cache, int local) | |
1327 | { | |
1328 | int psend = -1; | |
1329 | int vif, ct; | |
1330 | ||
1331 | vif = cache->mfc_parent; | |
1332 | cache->mfc_un.res.pkt++; | |
1333 | cache->mfc_un.res.bytes += skb->len; | |
1334 | ||
1335 | /* | |
1336 | * Wrong interface: drop packet and (maybe) send PIM assert. | |
1337 | */ | |
cf958ae3 | 1338 | if (init_net.ipv4.vif_table[vif].dev != skb->dev) { |
1da177e4 LT |
1339 | int true_vifi; |
1340 | ||
ee6b9673 | 1341 | if (skb->rtable->fl.iif == 0) { |
1da177e4 LT |
1342 | /* It is our own packet, looped back. |
1343 | Very complicated situation... | |
1344 | ||
1345 | The best workaround until routing daemons will be | |
1346 | fixed is not to redistribute packet, if it was | |
1347 | send through wrong interface. It means, that | |
1348 | multicast applications WILL NOT work for | |
1349 | (S,G), which have default multicast route pointing | |
1350 | to wrong oif. In any case, it is not a good | |
1351 | idea to use multicasting applications on router. | |
1352 | */ | |
1353 | goto dont_forward; | |
1354 | } | |
1355 | ||
1356 | cache->mfc_un.res.wrong_if++; | |
1357 | true_vifi = ipmr_find_vif(skb->dev); | |
1358 | ||
1359 | if (true_vifi >= 0 && mroute_do_assert && | |
1360 | /* pimsm uses asserts, when switching from RPT to SPT, | |
1361 | so that we cannot check that packet arrived on an oif. | |
1362 | It is bad, but otherwise we would need to move pretty | |
1363 | large chunk of pimd to kernel. Ough... --ANK | |
1364 | */ | |
1365 | (mroute_do_pim || cache->mfc_un.res.ttls[true_vifi] < 255) && | |
e905a9ed | 1366 | time_after(jiffies, |
1da177e4 LT |
1367 | cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) { |
1368 | cache->mfc_un.res.last_assert = jiffies; | |
1369 | ipmr_cache_report(skb, true_vifi, IGMPMSG_WRONGVIF); | |
1370 | } | |
1371 | goto dont_forward; | |
1372 | } | |
1373 | ||
cf958ae3 BT |
1374 | init_net.ipv4.vif_table[vif].pkt_in++; |
1375 | init_net.ipv4.vif_table[vif].bytes_in += skb->len; | |
1da177e4 LT |
1376 | |
1377 | /* | |
1378 | * Forward the frame | |
1379 | */ | |
1380 | for (ct = cache->mfc_un.res.maxvif-1; ct >= cache->mfc_un.res.minvif; ct--) { | |
eddc9ec5 | 1381 | if (ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) { |
1da177e4 LT |
1382 | if (psend != -1) { |
1383 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1384 | if (skb2) | |
1385 | ipmr_queue_xmit(skb2, cache, psend); | |
1386 | } | |
c354e124 | 1387 | psend = ct; |
1da177e4 LT |
1388 | } |
1389 | } | |
1390 | if (psend != -1) { | |
1391 | if (local) { | |
1392 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1393 | if (skb2) | |
1394 | ipmr_queue_xmit(skb2, cache, psend); | |
1395 | } else { | |
1396 | ipmr_queue_xmit(skb, cache, psend); | |
1397 | return 0; | |
1398 | } | |
1399 | } | |
1400 | ||
1401 | dont_forward: | |
1402 | if (!local) | |
1403 | kfree_skb(skb); | |
1404 | return 0; | |
1405 | } | |
1406 | ||
1407 | ||
1408 | /* | |
1409 | * Multicast packets for forwarding arrive here | |
1410 | */ | |
1411 | ||
1412 | int ip_mr_input(struct sk_buff *skb) | |
1413 | { | |
1414 | struct mfc_cache *cache; | |
ee6b9673 | 1415 | int local = skb->rtable->rt_flags&RTCF_LOCAL; |
1da177e4 LT |
1416 | |
1417 | /* Packet is looped back after forward, it should not be | |
1418 | forwarded second time, but still can be delivered locally. | |
1419 | */ | |
1420 | if (IPCB(skb)->flags&IPSKB_FORWARDED) | |
1421 | goto dont_forward; | |
1422 | ||
1423 | if (!local) { | |
1424 | if (IPCB(skb)->opt.router_alert) { | |
1425 | if (ip_call_ra_chain(skb)) | |
1426 | return 0; | |
eddc9ec5 | 1427 | } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP){ |
1da177e4 LT |
1428 | /* IGMPv1 (and broken IGMPv2 implementations sort of |
1429 | Cisco IOS <= 11.2(8)) do not put router alert | |
1430 | option to IGMP packets destined to routable | |
1431 | groups. It is very bad, because it means | |
1432 | that we can forward NO IGMP messages. | |
1433 | */ | |
1434 | read_lock(&mrt_lock); | |
70a269e6 | 1435 | if (init_net.ipv4.mroute_sk) { |
2715bcf9 | 1436 | nf_reset(skb); |
70a269e6 | 1437 | raw_rcv(init_net.ipv4.mroute_sk, skb); |
1da177e4 LT |
1438 | read_unlock(&mrt_lock); |
1439 | return 0; | |
1440 | } | |
1441 | read_unlock(&mrt_lock); | |
1442 | } | |
1443 | } | |
1444 | ||
1445 | read_lock(&mrt_lock); | |
eddc9ec5 | 1446 | cache = ipmr_cache_find(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr); |
1da177e4 LT |
1447 | |
1448 | /* | |
1449 | * No usable cache entry | |
1450 | */ | |
c354e124 | 1451 | if (cache == NULL) { |
1da177e4 LT |
1452 | int vif; |
1453 | ||
1454 | if (local) { | |
1455 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1456 | ip_local_deliver(skb); | |
1457 | if (skb2 == NULL) { | |
1458 | read_unlock(&mrt_lock); | |
1459 | return -ENOBUFS; | |
1460 | } | |
1461 | skb = skb2; | |
1462 | } | |
1463 | ||
1464 | vif = ipmr_find_vif(skb->dev); | |
1465 | if (vif >= 0) { | |
1466 | int err = ipmr_cache_unresolved(vif, skb); | |
1467 | read_unlock(&mrt_lock); | |
1468 | ||
1469 | return err; | |
1470 | } | |
1471 | read_unlock(&mrt_lock); | |
1472 | kfree_skb(skb); | |
1473 | return -ENODEV; | |
1474 | } | |
1475 | ||
1476 | ip_mr_forward(skb, cache, local); | |
1477 | ||
1478 | read_unlock(&mrt_lock); | |
1479 | ||
1480 | if (local) | |
1481 | return ip_local_deliver(skb); | |
1482 | ||
1483 | return 0; | |
1484 | ||
1485 | dont_forward: | |
1486 | if (local) | |
1487 | return ip_local_deliver(skb); | |
1488 | kfree_skb(skb); | |
1489 | return 0; | |
1490 | } | |
1491 | ||
b1879204 IJ |
1492 | #ifdef CONFIG_IP_PIMSM |
1493 | static int __pim_rcv(struct sk_buff *skb, unsigned int pimlen) | |
1da177e4 | 1494 | { |
b1879204 IJ |
1495 | struct net_device *reg_dev = NULL; |
1496 | struct iphdr *encap; | |
1da177e4 | 1497 | |
b1879204 | 1498 | encap = (struct iphdr *)(skb_transport_header(skb) + pimlen); |
1da177e4 LT |
1499 | /* |
1500 | Check that: | |
1501 | a. packet is really destinted to a multicast group | |
1502 | b. packet is not a NULL-REGISTER | |
1503 | c. packet is not truncated | |
1504 | */ | |
f97c1e0c | 1505 | if (!ipv4_is_multicast(encap->daddr) || |
1da177e4 | 1506 | encap->tot_len == 0 || |
b1879204 IJ |
1507 | ntohs(encap->tot_len) + pimlen > skb->len) |
1508 | return 1; | |
1da177e4 LT |
1509 | |
1510 | read_lock(&mrt_lock); | |
1511 | if (reg_vif_num >= 0) | |
cf958ae3 | 1512 | reg_dev = init_net.ipv4.vif_table[reg_vif_num].dev; |
1da177e4 LT |
1513 | if (reg_dev) |
1514 | dev_hold(reg_dev); | |
1515 | read_unlock(&mrt_lock); | |
1516 | ||
e905a9ed | 1517 | if (reg_dev == NULL) |
b1879204 | 1518 | return 1; |
1da177e4 | 1519 | |
b0e380b1 | 1520 | skb->mac_header = skb->network_header; |
1da177e4 | 1521 | skb_pull(skb, (u8*)encap - skb->data); |
31c7711b | 1522 | skb_reset_network_header(skb); |
1da177e4 | 1523 | skb->dev = reg_dev; |
1da177e4 LT |
1524 | skb->protocol = htons(ETH_P_IP); |
1525 | skb->ip_summed = 0; | |
1526 | skb->pkt_type = PACKET_HOST; | |
1527 | dst_release(skb->dst); | |
1528 | skb->dst = NULL; | |
cf3677ae PE |
1529 | reg_dev->stats.rx_bytes += skb->len; |
1530 | reg_dev->stats.rx_packets++; | |
1da177e4 LT |
1531 | nf_reset(skb); |
1532 | netif_rx(skb); | |
1533 | dev_put(reg_dev); | |
b1879204 | 1534 | |
1da177e4 | 1535 | return 0; |
b1879204 IJ |
1536 | } |
1537 | #endif | |
1538 | ||
1539 | #ifdef CONFIG_IP_PIMSM_V1 | |
1540 | /* | |
1541 | * Handle IGMP messages of PIMv1 | |
1542 | */ | |
1543 | ||
1544 | int pim_rcv_v1(struct sk_buff * skb) | |
1545 | { | |
1546 | struct igmphdr *pim; | |
1547 | ||
1548 | if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) | |
1549 | goto drop; | |
1550 | ||
1551 | pim = igmp_hdr(skb); | |
1552 | ||
1553 | if (!mroute_do_pim || | |
1554 | pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER) | |
1555 | goto drop; | |
1556 | ||
1557 | if (__pim_rcv(skb, sizeof(*pim))) { | |
1558 | drop: | |
1559 | kfree_skb(skb); | |
1560 | } | |
1da177e4 LT |
1561 | return 0; |
1562 | } | |
1563 | #endif | |
1564 | ||
1565 | #ifdef CONFIG_IP_PIMSM_V2 | |
1566 | static int pim_rcv(struct sk_buff * skb) | |
1567 | { | |
1568 | struct pimreghdr *pim; | |
1da177e4 | 1569 | |
b1879204 | 1570 | if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr))) |
1da177e4 LT |
1571 | goto drop; |
1572 | ||
9c70220b | 1573 | pim = (struct pimreghdr *)skb_transport_header(skb); |
e905a9ed | 1574 | if (pim->type != ((PIM_VERSION<<4)|(PIM_REGISTER)) || |
1da177e4 | 1575 | (pim->flags&PIM_NULL_REGISTER) || |
e905a9ed | 1576 | (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 && |
d3bc23e7 | 1577 | csum_fold(skb_checksum(skb, 0, skb->len, 0)))) |
1da177e4 LT |
1578 | goto drop; |
1579 | ||
b1879204 IJ |
1580 | if (__pim_rcv(skb, sizeof(*pim))) { |
1581 | drop: | |
1582 | kfree_skb(skb); | |
1583 | } | |
1da177e4 LT |
1584 | return 0; |
1585 | } | |
1586 | #endif | |
1587 | ||
1588 | static int | |
1589 | ipmr_fill_mroute(struct sk_buff *skb, struct mfc_cache *c, struct rtmsg *rtm) | |
1590 | { | |
1591 | int ct; | |
1592 | struct rtnexthop *nhp; | |
cf958ae3 | 1593 | struct net_device *dev = init_net.ipv4.vif_table[c->mfc_parent].dev; |
27a884dc | 1594 | u8 *b = skb_tail_pointer(skb); |
1da177e4 LT |
1595 | struct rtattr *mp_head; |
1596 | ||
1597 | if (dev) | |
1598 | RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex); | |
1599 | ||
c354e124 | 1600 | mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0)); |
1da177e4 LT |
1601 | |
1602 | for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) { | |
1603 | if (c->mfc_un.res.ttls[ct] < 255) { | |
1604 | if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4)) | |
1605 | goto rtattr_failure; | |
c354e124 | 1606 | nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp))); |
1da177e4 LT |
1607 | nhp->rtnh_flags = 0; |
1608 | nhp->rtnh_hops = c->mfc_un.res.ttls[ct]; | |
cf958ae3 | 1609 | nhp->rtnh_ifindex = init_net.ipv4.vif_table[ct].dev->ifindex; |
1da177e4 LT |
1610 | nhp->rtnh_len = sizeof(*nhp); |
1611 | } | |
1612 | } | |
1613 | mp_head->rta_type = RTA_MULTIPATH; | |
27a884dc | 1614 | mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head; |
1da177e4 LT |
1615 | rtm->rtm_type = RTN_MULTICAST; |
1616 | return 1; | |
1617 | ||
1618 | rtattr_failure: | |
dc5fc579 | 1619 | nlmsg_trim(skb, b); |
1da177e4 LT |
1620 | return -EMSGSIZE; |
1621 | } | |
1622 | ||
1623 | int ipmr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait) | |
1624 | { | |
1625 | int err; | |
1626 | struct mfc_cache *cache; | |
ee6b9673 | 1627 | struct rtable *rt = skb->rtable; |
1da177e4 LT |
1628 | |
1629 | read_lock(&mrt_lock); | |
1630 | cache = ipmr_cache_find(rt->rt_src, rt->rt_dst); | |
1631 | ||
c354e124 | 1632 | if (cache == NULL) { |
72287490 | 1633 | struct sk_buff *skb2; |
eddc9ec5 | 1634 | struct iphdr *iph; |
1da177e4 LT |
1635 | struct net_device *dev; |
1636 | int vif; | |
1637 | ||
1638 | if (nowait) { | |
1639 | read_unlock(&mrt_lock); | |
1640 | return -EAGAIN; | |
1641 | } | |
1642 | ||
1643 | dev = skb->dev; | |
1644 | if (dev == NULL || (vif = ipmr_find_vif(dev)) < 0) { | |
1645 | read_unlock(&mrt_lock); | |
1646 | return -ENODEV; | |
1647 | } | |
72287490 AK |
1648 | skb2 = skb_clone(skb, GFP_ATOMIC); |
1649 | if (!skb2) { | |
1650 | read_unlock(&mrt_lock); | |
1651 | return -ENOMEM; | |
1652 | } | |
1653 | ||
e2d1bca7 ACM |
1654 | skb_push(skb2, sizeof(struct iphdr)); |
1655 | skb_reset_network_header(skb2); | |
eddc9ec5 ACM |
1656 | iph = ip_hdr(skb2); |
1657 | iph->ihl = sizeof(struct iphdr) >> 2; | |
1658 | iph->saddr = rt->rt_src; | |
1659 | iph->daddr = rt->rt_dst; | |
1660 | iph->version = 0; | |
72287490 | 1661 | err = ipmr_cache_unresolved(vif, skb2); |
1da177e4 LT |
1662 | read_unlock(&mrt_lock); |
1663 | return err; | |
1664 | } | |
1665 | ||
1666 | if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY)) | |
1667 | cache->mfc_flags |= MFC_NOTIFY; | |
1668 | err = ipmr_fill_mroute(skb, cache, rtm); | |
1669 | read_unlock(&mrt_lock); | |
1670 | return err; | |
1671 | } | |
1672 | ||
e905a9ed | 1673 | #ifdef CONFIG_PROC_FS |
1da177e4 LT |
1674 | /* |
1675 | * The /proc interfaces to multicast routing /proc/ip_mr_cache /proc/ip_mr_vif | |
1676 | */ | |
1677 | struct ipmr_vif_iter { | |
1678 | int ct; | |
1679 | }; | |
1680 | ||
1681 | static struct vif_device *ipmr_vif_seq_idx(struct ipmr_vif_iter *iter, | |
1682 | loff_t pos) | |
1683 | { | |
cf958ae3 BT |
1684 | for (iter->ct = 0; iter->ct < init_net.ipv4.maxvif; ++iter->ct) { |
1685 | if (!VIF_EXISTS(&init_net, iter->ct)) | |
1da177e4 | 1686 | continue; |
e905a9ed | 1687 | if (pos-- == 0) |
cf958ae3 | 1688 | return &init_net.ipv4.vif_table[iter->ct]; |
1da177e4 LT |
1689 | } |
1690 | return NULL; | |
1691 | } | |
1692 | ||
1693 | static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos) | |
ba93ef74 | 1694 | __acquires(mrt_lock) |
1da177e4 LT |
1695 | { |
1696 | read_lock(&mrt_lock); | |
e905a9ed | 1697 | return *pos ? ipmr_vif_seq_idx(seq->private, *pos - 1) |
1da177e4 LT |
1698 | : SEQ_START_TOKEN; |
1699 | } | |
1700 | ||
1701 | static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
1702 | { | |
1703 | struct ipmr_vif_iter *iter = seq->private; | |
1704 | ||
1705 | ++*pos; | |
1706 | if (v == SEQ_START_TOKEN) | |
1707 | return ipmr_vif_seq_idx(iter, 0); | |
e905a9ed | 1708 | |
cf958ae3 BT |
1709 | while (++iter->ct < init_net.ipv4.maxvif) { |
1710 | if (!VIF_EXISTS(&init_net, iter->ct)) | |
1da177e4 | 1711 | continue; |
cf958ae3 | 1712 | return &init_net.ipv4.vif_table[iter->ct]; |
1da177e4 LT |
1713 | } |
1714 | return NULL; | |
1715 | } | |
1716 | ||
1717 | static void ipmr_vif_seq_stop(struct seq_file *seq, void *v) | |
ba93ef74 | 1718 | __releases(mrt_lock) |
1da177e4 LT |
1719 | { |
1720 | read_unlock(&mrt_lock); | |
1721 | } | |
1722 | ||
1723 | static int ipmr_vif_seq_show(struct seq_file *seq, void *v) | |
1724 | { | |
1725 | if (v == SEQ_START_TOKEN) { | |
e905a9ed | 1726 | seq_puts(seq, |
1da177e4 LT |
1727 | "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n"); |
1728 | } else { | |
1729 | const struct vif_device *vif = v; | |
1730 | const char *name = vif->dev ? vif->dev->name : "none"; | |
1731 | ||
1732 | seq_printf(seq, | |
1733 | "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n", | |
cf958ae3 | 1734 | vif - init_net.ipv4.vif_table, |
e905a9ed | 1735 | name, vif->bytes_in, vif->pkt_in, |
1da177e4 LT |
1736 | vif->bytes_out, vif->pkt_out, |
1737 | vif->flags, vif->local, vif->remote); | |
1738 | } | |
1739 | return 0; | |
1740 | } | |
1741 | ||
f690808e | 1742 | static const struct seq_operations ipmr_vif_seq_ops = { |
1da177e4 LT |
1743 | .start = ipmr_vif_seq_start, |
1744 | .next = ipmr_vif_seq_next, | |
1745 | .stop = ipmr_vif_seq_stop, | |
1746 | .show = ipmr_vif_seq_show, | |
1747 | }; | |
1748 | ||
1749 | static int ipmr_vif_open(struct inode *inode, struct file *file) | |
1750 | { | |
cf7732e4 PE |
1751 | return seq_open_private(file, &ipmr_vif_seq_ops, |
1752 | sizeof(struct ipmr_vif_iter)); | |
1da177e4 LT |
1753 | } |
1754 | ||
9a32144e | 1755 | static const struct file_operations ipmr_vif_fops = { |
1da177e4 LT |
1756 | .owner = THIS_MODULE, |
1757 | .open = ipmr_vif_open, | |
1758 | .read = seq_read, | |
1759 | .llseek = seq_lseek, | |
1760 | .release = seq_release_private, | |
1761 | }; | |
1762 | ||
1763 | struct ipmr_mfc_iter { | |
1764 | struct mfc_cache **cache; | |
1765 | int ct; | |
1766 | }; | |
1767 | ||
1768 | ||
1769 | static struct mfc_cache *ipmr_mfc_seq_idx(struct ipmr_mfc_iter *it, loff_t pos) | |
1770 | { | |
1771 | struct mfc_cache *mfc; | |
1772 | ||
2bb8b26c | 1773 | it->cache = init_net.ipv4.mfc_cache_array; |
1da177e4 | 1774 | read_lock(&mrt_lock); |
e905a9ed | 1775 | for (it->ct = 0; it->ct < MFC_LINES; it->ct++) |
2bb8b26c BT |
1776 | for (mfc = init_net.ipv4.mfc_cache_array[it->ct]; |
1777 | mfc; mfc = mfc->next) | |
e905a9ed | 1778 | if (pos-- == 0) |
1da177e4 LT |
1779 | return mfc; |
1780 | read_unlock(&mrt_lock); | |
1781 | ||
1782 | it->cache = &mfc_unres_queue; | |
1783 | spin_lock_bh(&mfc_unres_lock); | |
132adf54 | 1784 | for (mfc = mfc_unres_queue; mfc; mfc = mfc->next) |
1da177e4 LT |
1785 | if (pos-- == 0) |
1786 | return mfc; | |
1787 | spin_unlock_bh(&mfc_unres_lock); | |
1788 | ||
1789 | it->cache = NULL; | |
1790 | return NULL; | |
1791 | } | |
1792 | ||
1793 | ||
1794 | static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos) | |
1795 | { | |
1796 | struct ipmr_mfc_iter *it = seq->private; | |
1797 | it->cache = NULL; | |
1798 | it->ct = 0; | |
e905a9ed | 1799 | return *pos ? ipmr_mfc_seq_idx(seq->private, *pos - 1) |
1da177e4 LT |
1800 | : SEQ_START_TOKEN; |
1801 | } | |
1802 | ||
1803 | static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
1804 | { | |
1805 | struct mfc_cache *mfc = v; | |
1806 | struct ipmr_mfc_iter *it = seq->private; | |
1807 | ||
1808 | ++*pos; | |
1809 | ||
1810 | if (v == SEQ_START_TOKEN) | |
1811 | return ipmr_mfc_seq_idx(seq->private, 0); | |
1812 | ||
1813 | if (mfc->next) | |
1814 | return mfc->next; | |
e905a9ed YH |
1815 | |
1816 | if (it->cache == &mfc_unres_queue) | |
1da177e4 LT |
1817 | goto end_of_list; |
1818 | ||
2bb8b26c | 1819 | BUG_ON(it->cache != init_net.ipv4.mfc_cache_array); |
1da177e4 LT |
1820 | |
1821 | while (++it->ct < MFC_LINES) { | |
2bb8b26c | 1822 | mfc = init_net.ipv4.mfc_cache_array[it->ct]; |
1da177e4 LT |
1823 | if (mfc) |
1824 | return mfc; | |
1825 | } | |
1826 | ||
1827 | /* exhausted cache_array, show unresolved */ | |
1828 | read_unlock(&mrt_lock); | |
1829 | it->cache = &mfc_unres_queue; | |
1830 | it->ct = 0; | |
e905a9ed | 1831 | |
1da177e4 LT |
1832 | spin_lock_bh(&mfc_unres_lock); |
1833 | mfc = mfc_unres_queue; | |
e905a9ed | 1834 | if (mfc) |
1da177e4 LT |
1835 | return mfc; |
1836 | ||
1837 | end_of_list: | |
1838 | spin_unlock_bh(&mfc_unres_lock); | |
1839 | it->cache = NULL; | |
1840 | ||
1841 | return NULL; | |
1842 | } | |
1843 | ||
1844 | static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v) | |
1845 | { | |
1846 | struct ipmr_mfc_iter *it = seq->private; | |
1847 | ||
1848 | if (it->cache == &mfc_unres_queue) | |
1849 | spin_unlock_bh(&mfc_unres_lock); | |
2bb8b26c | 1850 | else if (it->cache == init_net.ipv4.mfc_cache_array) |
1da177e4 LT |
1851 | read_unlock(&mrt_lock); |
1852 | } | |
1853 | ||
1854 | static int ipmr_mfc_seq_show(struct seq_file *seq, void *v) | |
1855 | { | |
1856 | int n; | |
1857 | ||
1858 | if (v == SEQ_START_TOKEN) { | |
e905a9ed | 1859 | seq_puts(seq, |
1da177e4 LT |
1860 | "Group Origin Iif Pkts Bytes Wrong Oifs\n"); |
1861 | } else { | |
1862 | const struct mfc_cache *mfc = v; | |
1863 | const struct ipmr_mfc_iter *it = seq->private; | |
e905a9ed | 1864 | |
999890b2 | 1865 | seq_printf(seq, "%08lX %08lX %-3hd", |
1da177e4 LT |
1866 | (unsigned long) mfc->mfc_mcastgrp, |
1867 | (unsigned long) mfc->mfc_origin, | |
1ea472e2 | 1868 | mfc->mfc_parent); |
1da177e4 LT |
1869 | |
1870 | if (it->cache != &mfc_unres_queue) { | |
1ea472e2 BT |
1871 | seq_printf(seq, " %8lu %8lu %8lu", |
1872 | mfc->mfc_un.res.pkt, | |
1873 | mfc->mfc_un.res.bytes, | |
1874 | mfc->mfc_un.res.wrong_if); | |
132adf54 SH |
1875 | for (n = mfc->mfc_un.res.minvif; |
1876 | n < mfc->mfc_un.res.maxvif; n++ ) { | |
cf958ae3 BT |
1877 | if (VIF_EXISTS(&init_net, n) && |
1878 | mfc->mfc_un.res.ttls[n] < 255) | |
1879 | seq_printf(seq, | |
e905a9ed | 1880 | " %2d:%-3d", |
1da177e4 LT |
1881 | n, mfc->mfc_un.res.ttls[n]); |
1882 | } | |
1ea472e2 BT |
1883 | } else { |
1884 | /* unresolved mfc_caches don't contain | |
1885 | * pkt, bytes and wrong_if values | |
1886 | */ | |
1887 | seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul); | |
1da177e4 LT |
1888 | } |
1889 | seq_putc(seq, '\n'); | |
1890 | } | |
1891 | return 0; | |
1892 | } | |
1893 | ||
f690808e | 1894 | static const struct seq_operations ipmr_mfc_seq_ops = { |
1da177e4 LT |
1895 | .start = ipmr_mfc_seq_start, |
1896 | .next = ipmr_mfc_seq_next, | |
1897 | .stop = ipmr_mfc_seq_stop, | |
1898 | .show = ipmr_mfc_seq_show, | |
1899 | }; | |
1900 | ||
1901 | static int ipmr_mfc_open(struct inode *inode, struct file *file) | |
1902 | { | |
cf7732e4 PE |
1903 | return seq_open_private(file, &ipmr_mfc_seq_ops, |
1904 | sizeof(struct ipmr_mfc_iter)); | |
1da177e4 LT |
1905 | } |
1906 | ||
9a32144e | 1907 | static const struct file_operations ipmr_mfc_fops = { |
1da177e4 LT |
1908 | .owner = THIS_MODULE, |
1909 | .open = ipmr_mfc_open, | |
1910 | .read = seq_read, | |
1911 | .llseek = seq_lseek, | |
1912 | .release = seq_release_private, | |
1913 | }; | |
e905a9ed | 1914 | #endif |
1da177e4 LT |
1915 | |
1916 | #ifdef CONFIG_IP_PIMSM_V2 | |
1917 | static struct net_protocol pim_protocol = { | |
1918 | .handler = pim_rcv, | |
1919 | }; | |
1920 | #endif | |
1921 | ||
1922 | ||
1923 | /* | |
1924 | * Setup for IP multicast routing | |
1925 | */ | |
cf958ae3 BT |
1926 | static int __net_init ipmr_net_init(struct net *net) |
1927 | { | |
1928 | int err = 0; | |
1929 | ||
1930 | net->ipv4.vif_table = kcalloc(MAXVIFS, sizeof(struct vif_device), | |
1931 | GFP_KERNEL); | |
1932 | if (!net->ipv4.vif_table) { | |
1933 | err = -ENOMEM; | |
1934 | goto fail; | |
1935 | } | |
2bb8b26c BT |
1936 | |
1937 | /* Forwarding cache */ | |
1938 | net->ipv4.mfc_cache_array = kcalloc(MFC_LINES, | |
1939 | sizeof(struct mfc_cache *), | |
1940 | GFP_KERNEL); | |
1941 | if (!net->ipv4.mfc_cache_array) { | |
1942 | err = -ENOMEM; | |
1943 | goto fail_mfc_cache; | |
1944 | } | |
1945 | return 0; | |
1946 | ||
1947 | fail_mfc_cache: | |
1948 | kfree(net->ipv4.vif_table); | |
cf958ae3 BT |
1949 | fail: |
1950 | return err; | |
1951 | } | |
1952 | ||
1953 | static void __net_exit ipmr_net_exit(struct net *net) | |
1954 | { | |
2bb8b26c | 1955 | kfree(net->ipv4.mfc_cache_array); |
cf958ae3 BT |
1956 | kfree(net->ipv4.vif_table); |
1957 | } | |
1958 | ||
1959 | static struct pernet_operations ipmr_net_ops = { | |
1960 | .init = ipmr_net_init, | |
1961 | .exit = ipmr_net_exit, | |
1962 | }; | |
e905a9ed | 1963 | |
03d2f897 | 1964 | int __init ip_mr_init(void) |
1da177e4 | 1965 | { |
03d2f897 WC |
1966 | int err; |
1967 | ||
1da177e4 LT |
1968 | mrt_cachep = kmem_cache_create("ip_mrt_cache", |
1969 | sizeof(struct mfc_cache), | |
e5d679f3 | 1970 | 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, |
20c2df83 | 1971 | NULL); |
03d2f897 WC |
1972 | if (!mrt_cachep) |
1973 | return -ENOMEM; | |
1974 | ||
cf958ae3 BT |
1975 | err = register_pernet_subsys(&ipmr_net_ops); |
1976 | if (err) | |
1977 | goto reg_pernet_fail; | |
1978 | ||
b24b8a24 | 1979 | setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0); |
03d2f897 WC |
1980 | err = register_netdevice_notifier(&ip_mr_notifier); |
1981 | if (err) | |
1982 | goto reg_notif_fail; | |
e905a9ed | 1983 | #ifdef CONFIG_PROC_FS |
03d2f897 WC |
1984 | err = -ENOMEM; |
1985 | if (!proc_net_fops_create(&init_net, "ip_mr_vif", 0, &ipmr_vif_fops)) | |
1986 | goto proc_vif_fail; | |
1987 | if (!proc_net_fops_create(&init_net, "ip_mr_cache", 0, &ipmr_mfc_fops)) | |
1988 | goto proc_cache_fail; | |
e905a9ed | 1989 | #endif |
03d2f897 | 1990 | return 0; |
03d2f897 | 1991 | #ifdef CONFIG_PROC_FS |
03d2f897 WC |
1992 | proc_cache_fail: |
1993 | proc_net_remove(&init_net, "ip_mr_vif"); | |
c3e38896 BT |
1994 | proc_vif_fail: |
1995 | unregister_netdevice_notifier(&ip_mr_notifier); | |
03d2f897 | 1996 | #endif |
c3e38896 BT |
1997 | reg_notif_fail: |
1998 | del_timer(&ipmr_expire_timer); | |
cf958ae3 BT |
1999 | unregister_pernet_subsys(&ipmr_net_ops); |
2000 | reg_pernet_fail: | |
c3e38896 | 2001 | kmem_cache_destroy(mrt_cachep); |
03d2f897 | 2002 | return err; |
1da177e4 | 2003 | } |