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