?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
mm-only debug patch...
[mmotm.git] / net / ipv4 / ipmr.c
blobc757f0b4b74cb8df24cda2c61c3d1a4a7e495ec9
1 /*
2 * IP multicast routing support for mrouted 3.6/3.8
3 *
4 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
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 *
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
29 #include <asm/system.h>
30 #include <asm/uaccess.h>
31 #include <linux/types.h>
32 #include <linux/capability.h>
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>
49 #include <linux/if_ether.h>
50 #include <net/net_namespace.h>
51 #include <net/ip.h>
52 #include <net/protocol.h>
53 #include <linux/skbuff.h>
54 #include <net/route.h>
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>
64 #include <net/netlink.h>
65
66 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
67 #define CONFIG_IP_PIMSM 1
68 #endif
69
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
80 #define VIF_EXISTS(_net, _idx) ((_net)->ipv4.vif_table[_idx].dev != NULL)
81
82 static struct mfc_cache *mfc_unres_queue; /* Queue of unresolved entries */
83
84 /* Special spinlock for queue of unresolved entries */
85 static DEFINE_SPINLOCK(mfc_unres_lock);
86
87 /* We return to original Alan's scheme. Hash table of resolved
88 entries is changed only in process context and protected
89 with weak lock mrt_lock. Queue of unresolved entries is protected
90 with strong spinlock mfc_unres_lock.
91
92 In this case data path is free of exclusive locks at all.
93 */
94
95 static struct kmem_cache *mrt_cachep __read_mostly;
96
97 static int ip_mr_forward(struct sk_buff *skb, struct mfc_cache *cache, int local);
98 static int ipmr_cache_report(struct net *net,
99 struct sk_buff *pkt, vifi_t vifi, int assert);
100 static int ipmr_fill_mroute(struct sk_buff *skb, struct mfc_cache *c, struct rtmsg *rtm);
101
102 static struct timer_list ipmr_expire_timer;
103
104 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
105
106 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
107 {
108 struct net *net = dev_net(dev);
109
110 dev_close(dev);
111
112 dev = __dev_get_by_name(net, "tunl0");
113 if (dev) {
114 const struct net_device_ops *ops = dev->netdev_ops;
115 struct ifreq ifr;
116 struct ip_tunnel_parm p;
117
118 memset(&p, 0, sizeof(p));
119 p.iph.daddr = v->vifc_rmt_addr.s_addr;
120 p.iph.saddr = v->vifc_lcl_addr.s_addr;
121 p.iph.version = 4;
122 p.iph.ihl = 5;
123 p.iph.protocol = IPPROTO_IPIP;
124 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
125 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
126
127 if (ops->ndo_do_ioctl) {
128 mm_segment_t oldfs = get_fs();
129
130 set_fs(KERNEL_DS);
131 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
132 set_fs(oldfs);
133 }
134 }
135 }
136
137 static
138 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
139 {
140 struct net_device *dev;
141
142 dev = __dev_get_by_name(net, "tunl0");
143
144 if (dev) {
145 const struct net_device_ops *ops = dev->netdev_ops;
146 int err;
147 struct ifreq ifr;
148 struct ip_tunnel_parm p;
149 struct in_device *in_dev;
150
151 memset(&p, 0, sizeof(p));
152 p.iph.daddr = v->vifc_rmt_addr.s_addr;
153 p.iph.saddr = v->vifc_lcl_addr.s_addr;
154 p.iph.version = 4;
155 p.iph.ihl = 5;
156 p.iph.protocol = IPPROTO_IPIP;
157 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
158 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
159
160 if (ops->ndo_do_ioctl) {
161 mm_segment_t oldfs = get_fs();
162
163 set_fs(KERNEL_DS);
164 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
165 set_fs(oldfs);
166 } else
167 err = -EOPNOTSUPP;
168
169 dev = NULL;
170
171 if (err == 0 &&
172 (dev = __dev_get_by_name(net, p.name)) != NULL) {
173 dev->flags |= IFF_MULTICAST;
174
175 in_dev = __in_dev_get_rtnl(dev);
176 if (in_dev == NULL)
177 goto failure;
178
179 ipv4_devconf_setall(in_dev);
180 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
181
182 if (dev_open(dev))
183 goto failure;
184 dev_hold(dev);
185 }
186 }
187 return dev;
188
189 failure:
190 /* allow the register to be completed before unregistering. */
191 rtnl_unlock();
192 rtnl_lock();
193
194 unregister_netdevice(dev);
195 return NULL;
196 }
197
198 #ifdef CONFIG_IP_PIMSM
199
200 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
201 {
202 struct net *net = dev_net(dev);
203
204 read_lock(&mrt_lock);
205 dev->stats.tx_bytes += skb->len;
206 dev->stats.tx_packets++;
207 ipmr_cache_report(net, skb, net->ipv4.mroute_reg_vif_num,
208 IGMPMSG_WHOLEPKT);
209 read_unlock(&mrt_lock);
210 kfree_skb(skb);
211 return NETDEV_TX_OK;
212 }
213
214 static const struct net_device_ops reg_vif_netdev_ops = {
215 .ndo_start_xmit = reg_vif_xmit,
216 };
217
218 static void reg_vif_setup(struct net_device *dev)
219 {
220 dev->type = ARPHRD_PIMREG;
221 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
222 dev->flags = IFF_NOARP;
223 dev->netdev_ops = &reg_vif_netdev_ops,
224 dev->destructor = free_netdev;
225 dev->features |= NETIF_F_NETNS_LOCAL;
226 }
227
228 static struct net_device *ipmr_reg_vif(struct net *net)
229 {
230 struct net_device *dev;
231 struct in_device *in_dev;
232
233 dev = alloc_netdev(0, "pimreg", reg_vif_setup);
234
235 if (dev == NULL)
236 return NULL;
237
238 dev_net_set(dev, net);
239
240 if (register_netdevice(dev)) {
241 free_netdev(dev);
242 return NULL;
243 }
244 dev->iflink = 0;
245
246 rcu_read_lock();
247 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
248 rcu_read_unlock();
249 goto failure;
250 }
251
252 ipv4_devconf_setall(in_dev);
253 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
254 rcu_read_unlock();
255
256 if (dev_open(dev))
257 goto failure;
258
259 dev_hold(dev);
260
261 return dev;
262
263 failure:
264 /* allow the register to be completed before unregistering. */
265 rtnl_unlock();
266 rtnl_lock();
267
268 unregister_netdevice(dev);
269 return NULL;
270 }
271 #endif
272
273 /*
274 * Delete a VIF entry
275 * @notify: Set to 1, if the caller is a notifier_call
276 */
277
278 static int vif_delete(struct net *net, int vifi, int notify)
279 {
280 struct vif_device *v;
281 struct net_device *dev;
282 struct in_device *in_dev;
283
284 if (vifi < 0 || vifi >= net->ipv4.maxvif)
285 return -EADDRNOTAVAIL;
286
287 v = &net->ipv4.vif_table[vifi];
288
289 write_lock_bh(&mrt_lock);
290 dev = v->dev;
291 v->dev = NULL;
292
293 if (!dev) {
294 write_unlock_bh(&mrt_lock);
295 return -EADDRNOTAVAIL;
296 }
297
298 #ifdef CONFIG_IP_PIMSM
299 if (vifi == net->ipv4.mroute_reg_vif_num)
300 net->ipv4.mroute_reg_vif_num = -1;
301 #endif
302
303 if (vifi+1 == net->ipv4.maxvif) {
304 int tmp;
305 for (tmp=vifi-1; tmp>=0; tmp--) {
306 if (VIF_EXISTS(net, tmp))
307 break;
308 }
309 net->ipv4.maxvif = tmp+1;
310 }
311
312 write_unlock_bh(&mrt_lock);
313
314 dev_set_allmulti(dev, -1);
315
316 if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) {
317 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
318 ip_rt_multicast_event(in_dev);
319 }
320
321 if (v->flags&(VIFF_TUNNEL|VIFF_REGISTER) && !notify)
322 unregister_netdevice(dev);
323
324 dev_put(dev);
325 return 0;
326 }
327
328 static inline void ipmr_cache_free(struct mfc_cache *c)
329 {
330 release_net(mfc_net(c));
331 kmem_cache_free(mrt_cachep, c);
332 }
333
334 /* Destroy an unresolved cache entry, killing queued skbs
335 and reporting error to netlink readers.
336 */
337
338 static void ipmr_destroy_unres(struct mfc_cache *c)
339 {
340 struct sk_buff *skb;
341 struct nlmsgerr *e;
342 struct net *net = mfc_net(c);
343
344 atomic_dec(&net->ipv4.cache_resolve_queue_len);
345
346 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
347 if (ip_hdr(skb)->version == 0) {
348 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
349 nlh->nlmsg_type = NLMSG_ERROR;
350 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
351 skb_trim(skb, nlh->nlmsg_len);
352 e = NLMSG_DATA(nlh);
353 e->error = -ETIMEDOUT;
354 memset(&e->msg, 0, sizeof(e->msg));
355
356 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
357 } else
358 kfree_skb(skb);
359 }
360
361 ipmr_cache_free(c);
362 }
363
364
365 /* Single timer process for all the unresolved queue. */
366
367 static void ipmr_expire_process(unsigned long dummy)
368 {
369 unsigned long now;
370 unsigned long expires;
371 struct mfc_cache *c, **cp;
372
373 if (!spin_trylock(&mfc_unres_lock)) {
374 mod_timer(&ipmr_expire_timer, jiffies+HZ/10);
375 return;
376 }
377
378 if (mfc_unres_queue == NULL)
379 goto out;
380
381 now = jiffies;
382 expires = 10*HZ;
383 cp = &mfc_unres_queue;
384
385 while ((c=*cp) != NULL) {
386 if (time_after(c->mfc_un.unres.expires, now)) {
387 unsigned long interval = c->mfc_un.unres.expires - now;
388 if (interval < expires)
389 expires = interval;
390 cp = &c->next;
391 continue;
392 }
393
394 *cp = c->next;
395
396 ipmr_destroy_unres(c);
397 }
398
399 if (mfc_unres_queue != NULL)
400 mod_timer(&ipmr_expire_timer, jiffies + expires);
401
402 out:
403 spin_unlock(&mfc_unres_lock);
404 }
405
406 /* Fill oifs list. It is called under write locked mrt_lock. */
407
408 static void ipmr_update_thresholds(struct mfc_cache *cache, unsigned char *ttls)
409 {
410 int vifi;
411 struct net *net = mfc_net(cache);
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
417 for (vifi = 0; vifi < net->ipv4.maxvif; vifi++) {
418 if (VIF_EXISTS(net, vifi) &&
419 ttls[vifi] && ttls[vifi] < 255) {
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 net *net, struct vifctl *vifc, int mrtsock)
430 {
431 int vifi = vifc->vifc_vifi;
432 struct vif_device *v = &net->ipv4.vif_table[vifi];
433 struct net_device *dev;
434 struct in_device *in_dev;
435 int err;
436
437 /* Is vif busy ? */
438 if (VIF_EXISTS(net, vifi))
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 (net->ipv4.mroute_reg_vif_num >= 0)
449 return -EADDRINUSE;
450 dev = ipmr_reg_vif(net);
451 if (!dev)
452 return -ENOBUFS;
453 err = dev_set_allmulti(dev, 1);
454 if (err) {
455 unregister_netdevice(dev);
456 dev_put(dev);
457 return err;
458 }
459 break;
460 #endif
461 case VIFF_TUNNEL:
462 dev = ipmr_new_tunnel(net, vifc);
463 if (!dev)
464 return -ENOBUFS;
465 err = dev_set_allmulti(dev, 1);
466 if (err) {
467 ipmr_del_tunnel(dev, vifc);
468 dev_put(dev);
469 return err;
470 }
471 break;
472
473 case VIFF_USE_IFINDEX:
474 case 0:
475 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
476 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
477 if (dev && dev->ip_ptr == NULL) {
478 dev_put(dev);
479 return -EADDRNOTAVAIL;
480 }
481 } else
482 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
483
484 if (!dev)
485 return -EADDRNOTAVAIL;
486 err = dev_set_allmulti(dev, 1);
487 if (err) {
488 dev_put(dev);
489 return err;
490 }
491 break;
492 default:
493 return -EINVAL;
494 }
495
496 if ((in_dev = __in_dev_get_rtnl(dev)) == NULL)
497 return -EADDRNOTAVAIL;
498 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
499 ip_rt_multicast_event(in_dev);
500
501 /*
502 * Fill in the VIF structures
503 */
504 v->rate_limit = vifc->vifc_rate_limit;
505 v->local = vifc->vifc_lcl_addr.s_addr;
506 v->remote = vifc->vifc_rmt_addr.s_addr;
507 v->flags = vifc->vifc_flags;
508 if (!mrtsock)
509 v->flags |= VIFF_STATIC;
510 v->threshold = vifc->vifc_threshold;
511 v->bytes_in = 0;
512 v->bytes_out = 0;
513 v->pkt_in = 0;
514 v->pkt_out = 0;
515 v->link = dev->ifindex;
516 if (v->flags&(VIFF_TUNNEL|VIFF_REGISTER))
517 v->link = dev->iflink;
518
519 /* And finish update writing critical data */
520 write_lock_bh(&mrt_lock);
521 v->dev = dev;
522 #ifdef CONFIG_IP_PIMSM
523 if (v->flags&VIFF_REGISTER)
524 net->ipv4.mroute_reg_vif_num = vifi;
525 #endif
526 if (vifi+1 > net->ipv4.maxvif)
527 net->ipv4.maxvif = vifi+1;
528 write_unlock_bh(&mrt_lock);
529 return 0;
530 }
531
532 static struct mfc_cache *ipmr_cache_find(struct net *net,
533 __be32 origin,
534 __be32 mcastgrp)
535 {
536 int line = MFC_HASH(mcastgrp, origin);
537 struct mfc_cache *c;
538
539 for (c = net->ipv4.mfc_cache_array[line]; c; c = c->next) {
540 if (c->mfc_origin==origin && c->mfc_mcastgrp==mcastgrp)
541 break;
542 }
543 return c;
544 }
545
546 /*
547 * Allocate a multicast cache entry
548 */
549 static struct mfc_cache *ipmr_cache_alloc(struct net *net)
550 {
551 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
552 if (c == NULL)
553 return NULL;
554 c->mfc_un.res.minvif = MAXVIFS;
555 mfc_net_set(c, net);
556 return c;
557 }
558
559 static struct mfc_cache *ipmr_cache_alloc_unres(struct net *net)
560 {
561 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
562 if (c == NULL)
563 return NULL;
564 skb_queue_head_init(&c->mfc_un.unres.unresolved);
565 c->mfc_un.unres.expires = jiffies + 10*HZ;
566 mfc_net_set(c, net);
567 return c;
568 }
569
570 /*
571 * A cache entry has gone into a resolved state from queued
572 */
573
574 static void ipmr_cache_resolve(struct mfc_cache *uc, struct mfc_cache *c)
575 {
576 struct sk_buff *skb;
577 struct nlmsgerr *e;
578
579 /*
580 * Play the pending entries through our router
581 */
582
583 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
584 if (ip_hdr(skb)->version == 0) {
585 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
586
587 if (ipmr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) {
588 nlh->nlmsg_len = (skb_tail_pointer(skb) -
589 (u8 *)nlh);
590 } else {
591 nlh->nlmsg_type = NLMSG_ERROR;
592 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
593 skb_trim(skb, nlh->nlmsg_len);
594 e = NLMSG_DATA(nlh);
595 e->error = -EMSGSIZE;
596 memset(&e->msg, 0, sizeof(e->msg));
597 }
598
599 rtnl_unicast(skb, mfc_net(c), NETLINK_CB(skb).pid);
600 } else
601 ip_mr_forward(skb, c, 0);
602 }
603 }
604
605 /*
606 * Bounce a cache query up to mrouted. We could use netlink for this but mrouted
607 * expects the following bizarre scheme.
608 *
609 * Called under mrt_lock.
610 */
611
612 static int ipmr_cache_report(struct net *net,
613 struct sk_buff *pkt, vifi_t vifi, int assert)
614 {
615 struct sk_buff *skb;
616 const int ihl = ip_hdrlen(pkt);
617 struct igmphdr *igmp;
618 struct igmpmsg *msg;
619 int ret;
620
621 #ifdef CONFIG_IP_PIMSM
622 if (assert == IGMPMSG_WHOLEPKT)
623 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
624 else
625 #endif
626 skb = alloc_skb(128, GFP_ATOMIC);
627
628 if (!skb)
629 return -ENOBUFS;
630
631 #ifdef CONFIG_IP_PIMSM
632 if (assert == IGMPMSG_WHOLEPKT) {
633 /* Ugly, but we have no choice with this interface.
634 Duplicate old header, fix ihl, length etc.
635 And all this only to mangle msg->im_msgtype and
636 to set msg->im_mbz to "mbz" :-)
637 */
638 skb_push(skb, sizeof(struct iphdr));
639 skb_reset_network_header(skb);
640 skb_reset_transport_header(skb);
641 msg = (struct igmpmsg *)skb_network_header(skb);
642 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
643 msg->im_msgtype = IGMPMSG_WHOLEPKT;
644 msg->im_mbz = 0;
645 msg->im_vif = net->ipv4.mroute_reg_vif_num;
646 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
647 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
648 sizeof(struct iphdr));
649 } else
650 #endif
651 {
652
653 /*
654 * Copy the IP header
655 */
656
657 skb->network_header = skb->tail;
658 skb_put(skb, ihl);
659 skb_copy_to_linear_data(skb, pkt->data, ihl);
660 ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */
661 msg = (struct igmpmsg *)skb_network_header(skb);
662 msg->im_vif = vifi;
663 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
664
665 /*
666 * Add our header
667 */
668
669 igmp=(struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
670 igmp->type =
671 msg->im_msgtype = assert;
672 igmp->code = 0;
673 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
674 skb->transport_header = skb->network_header;
675 }
676
677 if (net->ipv4.mroute_sk == NULL) {
678 kfree_skb(skb);
679 return -EINVAL;
680 }
681
682 /*
683 * Deliver to mrouted
684 */
685 ret = sock_queue_rcv_skb(net->ipv4.mroute_sk, skb);
686 if (ret < 0) {
687 if (net_ratelimit())
688 printk(KERN_WARNING "mroute: pending queue full, dropping entries.\n");
689 kfree_skb(skb);
690 }
691
692 return ret;
693 }
694
695 /*
696 * Queue a packet for resolution. It gets locked cache entry!
697 */
698
699 static int
700 ipmr_cache_unresolved(struct net *net, vifi_t vifi, struct sk_buff *skb)
701 {
702 int err;
703 struct mfc_cache *c;
704 const struct iphdr *iph = ip_hdr(skb);
705
706 spin_lock_bh(&mfc_unres_lock);
707 for (c=mfc_unres_queue; c; c=c->next) {
708 if (net_eq(mfc_net(c), net) &&
709 c->mfc_mcastgrp == iph->daddr &&
710 c->mfc_origin == iph->saddr)
711 break;
712 }
713
714 if (c == NULL) {
715 /*
716 * Create a new entry if allowable
717 */
718
719 if (atomic_read(&net->ipv4.cache_resolve_queue_len) >= 10 ||
720 (c = ipmr_cache_alloc_unres(net)) == NULL) {
721 spin_unlock_bh(&mfc_unres_lock);
722
723 kfree_skb(skb);
724 return -ENOBUFS;
725 }
726
727 /*
728 * Fill in the new cache entry
729 */
730 c->mfc_parent = -1;
731 c->mfc_origin = iph->saddr;
732 c->mfc_mcastgrp = iph->daddr;
733
734 /*
735 * Reflect first query at mrouted.
736 */
737 err = ipmr_cache_report(net, skb, vifi, IGMPMSG_NOCACHE);
738 if (err < 0) {
739 /* If the report failed throw the cache entry
740 out - Brad Parker
741 */
742 spin_unlock_bh(&mfc_unres_lock);
743
744 ipmr_cache_free(c);
745 kfree_skb(skb);
746 return err;
747 }
748
749 atomic_inc(&net->ipv4.cache_resolve_queue_len);
750 c->next = mfc_unres_queue;
751 mfc_unres_queue = c;
752
753 mod_timer(&ipmr_expire_timer, c->mfc_un.unres.expires);
754 }
755
756 /*
757 * See if we can append the packet
758 */
759 if (c->mfc_un.unres.unresolved.qlen>3) {
760 kfree_skb(skb);
761 err = -ENOBUFS;
762 } else {
763 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
764 err = 0;
765 }
766
767 spin_unlock_bh(&mfc_unres_lock);
768 return err;
769 }
770
771 /*
772 * MFC cache manipulation by user space mroute daemon
773 */
774
775 static int ipmr_mfc_delete(struct net *net, struct mfcctl *mfc)
776 {
777 int line;
778 struct mfc_cache *c, **cp;
779
780 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
781
782 for (cp = &net->ipv4.mfc_cache_array[line];
783 (c = *cp) != NULL; cp = &c->next) {
784 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
785 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr) {
786 write_lock_bh(&mrt_lock);
787 *cp = c->next;
788 write_unlock_bh(&mrt_lock);
789
790 ipmr_cache_free(c);
791 return 0;
792 }
793 }
794 return -ENOENT;
795 }
796
797 static int ipmr_mfc_add(struct net *net, struct mfcctl *mfc, int mrtsock)
798 {
799 int line;
800 struct mfc_cache *uc, *c, **cp;
801
802 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
803
804 for (cp = &net->ipv4.mfc_cache_array[line];
805 (c = *cp) != NULL; cp = &c->next) {
806 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
807 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr)
808 break;
809 }
810
811 if (c != NULL) {
812 write_lock_bh(&mrt_lock);
813 c->mfc_parent = mfc->mfcc_parent;
814 ipmr_update_thresholds(c, mfc->mfcc_ttls);
815 if (!mrtsock)
816 c->mfc_flags |= MFC_STATIC;
817 write_unlock_bh(&mrt_lock);
818 return 0;
819 }
820
821 if (!ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
822 return -EINVAL;
823
824 c = ipmr_cache_alloc(net);
825 if (c == NULL)
826 return -ENOMEM;
827
828 c->mfc_origin = mfc->mfcc_origin.s_addr;
829 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
830 c->mfc_parent = mfc->mfcc_parent;
831 ipmr_update_thresholds(c, mfc->mfcc_ttls);
832 if (!mrtsock)
833 c->mfc_flags |= MFC_STATIC;
834
835 write_lock_bh(&mrt_lock);
836 c->next = net->ipv4.mfc_cache_array[line];
837 net->ipv4.mfc_cache_array[line] = c;
838 write_unlock_bh(&mrt_lock);
839
840 /*
841 * Check to see if we resolved a queued list. If so we
842 * need to send on the frames and tidy up.
843 */
844 spin_lock_bh(&mfc_unres_lock);
845 for (cp = &mfc_unres_queue; (uc=*cp) != NULL;
846 cp = &uc->next) {
847 if (net_eq(mfc_net(uc), net) &&
848 uc->mfc_origin == c->mfc_origin &&
849 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
850 *cp = uc->next;
851 atomic_dec(&net->ipv4.cache_resolve_queue_len);
852 break;
853 }
854 }
855 if (mfc_unres_queue == NULL)
856 del_timer(&ipmr_expire_timer);
857 spin_unlock_bh(&mfc_unres_lock);
858
859 if (uc) {
860 ipmr_cache_resolve(uc, c);
861 ipmr_cache_free(uc);
862 }
863 return 0;
864 }
865
866 /*
867 * Close the multicast socket, and clear the vif tables etc
868 */
869
870 static void mroute_clean_tables(struct net *net)
871 {
872 int i;
873
874 /*
875 * Shut down all active vif entries
876 */
877 for (i = 0; i < net->ipv4.maxvif; i++) {
878 if (!(net->ipv4.vif_table[i].flags&VIFF_STATIC))
879 vif_delete(net, i, 0);
880 }
881
882 /*
883 * Wipe the cache
884 */
885 for (i=0; i<MFC_LINES; i++) {
886 struct mfc_cache *c, **cp;
887
888 cp = &net->ipv4.mfc_cache_array[i];
889 while ((c = *cp) != NULL) {
890 if (c->mfc_flags&MFC_STATIC) {
891 cp = &c->next;
892 continue;
893 }
894 write_lock_bh(&mrt_lock);
895 *cp = c->next;
896 write_unlock_bh(&mrt_lock);
897
898 ipmr_cache_free(c);
899 }
900 }
901
902 if (atomic_read(&net->ipv4.cache_resolve_queue_len) != 0) {
903 struct mfc_cache *c, **cp;
904
905 spin_lock_bh(&mfc_unres_lock);
906 cp = &mfc_unres_queue;
907 while ((c = *cp) != NULL) {
908 if (!net_eq(mfc_net(c), net)) {
909 cp = &c->next;
910 continue;
911 }
912 *cp = c->next;
913
914 ipmr_destroy_unres(c);
915 }
916 spin_unlock_bh(&mfc_unres_lock);
917 }
918 }
919
920 static void mrtsock_destruct(struct sock *sk)
921 {
922 struct net *net = sock_net(sk);
923
924 rtnl_lock();
925 if (sk == net->ipv4.mroute_sk) {
926 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
927
928 write_lock_bh(&mrt_lock);
929 net->ipv4.mroute_sk = NULL;
930 write_unlock_bh(&mrt_lock);
931
932 mroute_clean_tables(net);
933 }
934 rtnl_unlock();
935 }
936
937 /*
938 * Socket options and virtual interface manipulation. The whole
939 * virtual interface system is a complete heap, but unfortunately
940 * that's how BSD mrouted happens to think. Maybe one day with a proper
941 * MOSPF/PIM router set up we can clean this up.
942 */
943
944 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
945 {
946 int ret;
947 struct vifctl vif;
948 struct mfcctl mfc;
949 struct net *net = sock_net(sk);
950
951 if (optname != MRT_INIT) {
952 if (sk != net->ipv4.mroute_sk && !capable(CAP_NET_ADMIN))
953 return -EACCES;
954 }
955
956 switch (optname) {
957 case MRT_INIT:
958 if (sk->sk_type != SOCK_RAW ||
959 inet_sk(sk)->num != IPPROTO_IGMP)
960 return -EOPNOTSUPP;
961 if (optlen != sizeof(int))
962 return -ENOPROTOOPT;
963
964 rtnl_lock();
965 if (net->ipv4.mroute_sk) {
966 rtnl_unlock();
967 return -EADDRINUSE;
968 }
969
970 ret = ip_ra_control(sk, 1, mrtsock_destruct);
971 if (ret == 0) {
972 write_lock_bh(&mrt_lock);
973 net->ipv4.mroute_sk = sk;
974 write_unlock_bh(&mrt_lock);
975
976 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
977 }
978 rtnl_unlock();
979 return ret;
980 case MRT_DONE:
981 if (sk != net->ipv4.mroute_sk)
982 return -EACCES;
983 return ip_ra_control(sk, 0, NULL);
984 case MRT_ADD_VIF:
985 case MRT_DEL_VIF:
986 if (optlen != sizeof(vif))
987 return -EINVAL;
988 if (copy_from_user(&vif, optval, sizeof(vif)))
989 return -EFAULT;
990 if (vif.vifc_vifi >= MAXVIFS)
991 return -ENFILE;
992 rtnl_lock();
993 if (optname == MRT_ADD_VIF) {
994 ret = vif_add(net, &vif, sk == net->ipv4.mroute_sk);
995 } else {
996 ret = vif_delete(net, vif.vifc_vifi, 0);
997 }
998 rtnl_unlock();
999 return ret;
1000
1001 /*
1002 * Manipulate the forwarding caches. These live
1003 * in a sort of kernel/user symbiosis.
1004 */
1005 case MRT_ADD_MFC:
1006 case MRT_DEL_MFC:
1007 if (optlen != sizeof(mfc))
1008 return -EINVAL;
1009 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1010 return -EFAULT;
1011 rtnl_lock();
1012 if (optname == MRT_DEL_MFC)
1013 ret = ipmr_mfc_delete(net, &mfc);
1014 else
1015 ret = ipmr_mfc_add(net, &mfc, sk == net->ipv4.mroute_sk);
1016 rtnl_unlock();
1017 return ret;
1018 /*
1019 * Control PIM assert.
1020 */
1021 case MRT_ASSERT:
1022 {
1023 int v;
1024 if (get_user(v,(int __user *)optval))
1025 return -EFAULT;
1026 net->ipv4.mroute_do_assert = (v) ? 1 : 0;
1027 return 0;
1028 }
1029 #ifdef CONFIG_IP_PIMSM
1030 case MRT_PIM:
1031 {
1032 int v;
1033
1034 if (get_user(v,(int __user *)optval))
1035 return -EFAULT;
1036 v = (v) ? 1 : 0;
1037
1038 rtnl_lock();
1039 ret = 0;
1040 if (v != net->ipv4.mroute_do_pim) {
1041 net->ipv4.mroute_do_pim = v;
1042 net->ipv4.mroute_do_assert = v;
1043 }
1044 rtnl_unlock();
1045 return ret;
1046 }
1047 #endif
1048 /*
1049 * Spurious command, or MRT_VERSION which you cannot
1050 * set.
1051 */
1052 default:
1053 return -ENOPROTOOPT;
1054 }
1055 }
1056
1057 /*
1058 * Getsock opt support for the multicast routing system.
1059 */
1060
1061 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1062 {
1063 int olr;
1064 int val;
1065 struct net *net = sock_net(sk);
1066
1067 if (optname != MRT_VERSION &&
1068 #ifdef CONFIG_IP_PIMSM
1069 optname!=MRT_PIM &&
1070 #endif
1071 optname!=MRT_ASSERT)
1072 return -ENOPROTOOPT;
1073
1074 if (get_user(olr, optlen))
1075 return -EFAULT;
1076
1077 olr = min_t(unsigned int, olr, sizeof(int));
1078 if (olr < 0)
1079 return -EINVAL;
1080
1081 if (put_user(olr, optlen))
1082 return -EFAULT;
1083 if (optname == MRT_VERSION)
1084 val = 0x0305;
1085 #ifdef CONFIG_IP_PIMSM
1086 else if (optname == MRT_PIM)
1087 val = net->ipv4.mroute_do_pim;
1088 #endif
1089 else
1090 val = net->ipv4.mroute_do_assert;
1091 if (copy_to_user(optval, &val, olr))
1092 return -EFAULT;
1093 return 0;
1094 }
1095
1096 /*
1097 * The IP multicast ioctl support routines.
1098 */
1099
1100 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1101 {
1102 struct sioc_sg_req sr;
1103 struct sioc_vif_req vr;
1104 struct vif_device *vif;
1105 struct mfc_cache *c;
1106 struct net *net = sock_net(sk);
1107
1108 switch (cmd) {
1109 case SIOCGETVIFCNT:
1110 if (copy_from_user(&vr, arg, sizeof(vr)))
1111 return -EFAULT;
1112 if (vr.vifi >= net->ipv4.maxvif)
1113 return -EINVAL;
1114 read_lock(&mrt_lock);
1115 vif = &net->ipv4.vif_table[vr.vifi];
1116 if (VIF_EXISTS(net, vr.vifi)) {
1117 vr.icount = vif->pkt_in;
1118 vr.ocount = vif->pkt_out;
1119 vr.ibytes = vif->bytes_in;
1120 vr.obytes = vif->bytes_out;
1121 read_unlock(&mrt_lock);
1122
1123 if (copy_to_user(arg, &vr, sizeof(vr)))
1124 return -EFAULT;
1125 return 0;
1126 }
1127 read_unlock(&mrt_lock);
1128 return -EADDRNOTAVAIL;
1129 case SIOCGETSGCNT:
1130 if (copy_from_user(&sr, arg, sizeof(sr)))
1131 return -EFAULT;
1132
1133 read_lock(&mrt_lock);
1134 c = ipmr_cache_find(net, sr.src.s_addr, sr.grp.s_addr);
1135 if (c) {
1136 sr.pktcnt = c->mfc_un.res.pkt;
1137 sr.bytecnt = c->mfc_un.res.bytes;
1138 sr.wrong_if = c->mfc_un.res.wrong_if;
1139 read_unlock(&mrt_lock);
1140
1141 if (copy_to_user(arg, &sr, sizeof(sr)))
1142 return -EFAULT;
1143 return 0;
1144 }
1145 read_unlock(&mrt_lock);
1146 return -EADDRNOTAVAIL;
1147 default:
1148 return -ENOIOCTLCMD;
1149 }
1150 }
1151
1152
1153 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1154 {
1155 struct net_device *dev = ptr;
1156 struct net *net = dev_net(dev);
1157 struct vif_device *v;
1158 int ct;
1159
1160 if (!net_eq(dev_net(dev), net))
1161 return NOTIFY_DONE;
1162
1163 if (event != NETDEV_UNREGISTER)
1164 return NOTIFY_DONE;
1165 v = &net->ipv4.vif_table[0];
1166 for (ct = 0; ct < net->ipv4.maxvif; ct++, v++) {
1167 if (v->dev == dev)
1168 vif_delete(net, ct, 1);
1169 }
1170 return NOTIFY_DONE;
1171 }
1172
1173
1174 static struct notifier_block ip_mr_notifier = {
1175 .notifier_call = ipmr_device_event,
1176 };
1177
1178 /*
1179 * Encapsulate a packet by attaching a valid IPIP header to it.
1180 * This avoids tunnel drivers and other mess and gives us the speed so
1181 * important for multicast video.
1182 */
1183
1184 static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr)
1185 {
1186 struct iphdr *iph;
1187 struct iphdr *old_iph = ip_hdr(skb);
1188
1189 skb_push(skb, sizeof(struct iphdr));
1190 skb->transport_header = skb->network_header;
1191 skb_reset_network_header(skb);
1192 iph = ip_hdr(skb);
1193
1194 iph->version = 4;
1195 iph->tos = old_iph->tos;
1196 iph->ttl = old_iph->ttl;
1197 iph->frag_off = 0;
1198 iph->daddr = daddr;
1199 iph->saddr = saddr;
1200 iph->protocol = IPPROTO_IPIP;
1201 iph->ihl = 5;
1202 iph->tot_len = htons(skb->len);
1203 ip_select_ident(iph, skb_dst(skb), NULL);
1204 ip_send_check(iph);
1205
1206 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1207 nf_reset(skb);
1208 }
1209
1210 static inline int ipmr_forward_finish(struct sk_buff *skb)
1211 {
1212 struct ip_options * opt = &(IPCB(skb)->opt);
1213
1214 IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
1215
1216 if (unlikely(opt->optlen))
1217 ip_forward_options(skb);
1218
1219 return dst_output(skb);
1220 }
1221
1222 /*
1223 * Processing handlers for ipmr_forward
1224 */
1225
1226 static void ipmr_queue_xmit(struct sk_buff *skb, struct mfc_cache *c, int vifi)
1227 {
1228 struct net *net = mfc_net(c);
1229 const struct iphdr *iph = ip_hdr(skb);
1230 struct vif_device *vif = &net->ipv4.vif_table[vifi];
1231 struct net_device *dev;
1232 struct rtable *rt;
1233 int encap = 0;
1234
1235 if (vif->dev == NULL)
1236 goto out_free;
1237
1238 #ifdef CONFIG_IP_PIMSM
1239 if (vif->flags & VIFF_REGISTER) {
1240 vif->pkt_out++;
1241 vif->bytes_out += skb->len;
1242 vif->dev->stats.tx_bytes += skb->len;
1243 vif->dev->stats.tx_packets++;
1244 ipmr_cache_report(net, skb, vifi, IGMPMSG_WHOLEPKT);
1245 goto out_free;
1246 }
1247 #endif
1248
1249 if (vif->flags&VIFF_TUNNEL) {
1250 struct flowi fl = { .oif = vif->link,
1251 .nl_u = { .ip4_u =
1252 { .daddr = vif->remote,
1253 .saddr = vif->local,
1254 .tos = RT_TOS(iph->tos) } },
1255 .proto = IPPROTO_IPIP };
1256 if (ip_route_output_key(net, &rt, &fl))
1257 goto out_free;
1258 encap = sizeof(struct iphdr);
1259 } else {
1260 struct flowi fl = { .oif = vif->link,
1261 .nl_u = { .ip4_u =
1262 { .daddr = iph->daddr,
1263 .tos = RT_TOS(iph->tos) } },
1264 .proto = IPPROTO_IPIP };
1265 if (ip_route_output_key(net, &rt, &fl))
1266 goto out_free;
1267 }
1268
1269 dev = rt->u.dst.dev;
1270
1271 if (skb->len+encap > dst_mtu(&rt->u.dst) && (ntohs(iph->frag_off) & IP_DF)) {
1272 /* Do not fragment multicasts. Alas, IPv4 does not
1273 allow to send ICMP, so that packets will disappear
1274 to blackhole.
1275 */
1276
1277 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
1278 ip_rt_put(rt);
1279 goto out_free;
1280 }
1281
1282 encap += LL_RESERVED_SPACE(dev) + rt->u.dst.header_len;
1283
1284 if (skb_cow(skb, encap)) {
1285 ip_rt_put(rt);
1286 goto out_free;
1287 }
1288
1289 vif->pkt_out++;
1290 vif->bytes_out += skb->len;
1291
1292 skb_dst_drop(skb);
1293 skb_dst_set(skb, &rt->u.dst);
1294 ip_decrease_ttl(ip_hdr(skb));
1295
1296 /* FIXME: forward and output firewalls used to be called here.
1297 * What do we do with netfilter? -- RR */
1298 if (vif->flags & VIFF_TUNNEL) {
1299 ip_encap(skb, vif->local, vif->remote);
1300 /* FIXME: extra output firewall step used to be here. --RR */
1301 vif->dev->stats.tx_packets++;
1302 vif->dev->stats.tx_bytes += skb->len;
1303 }
1304
1305 IPCB(skb)->flags |= IPSKB_FORWARDED;
1306
1307 /*
1308 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1309 * not only before forwarding, but after forwarding on all output
1310 * interfaces. It is clear, if mrouter runs a multicasting
1311 * program, it should receive packets not depending to what interface
1312 * program is joined.
1313 * If we will not make it, the program will have to join on all
1314 * interfaces. On the other hand, multihoming host (or router, but
1315 * not mrouter) cannot join to more than one interface - it will
1316 * result in receiving multiple packets.
1317 */
1318 NF_HOOK(PF_INET, NF_INET_FORWARD, skb, skb->dev, dev,
1319 ipmr_forward_finish);
1320 return;
1321
1322 out_free:
1323 kfree_skb(skb);
1324 return;
1325 }
1326
1327 static int ipmr_find_vif(struct net_device *dev)
1328 {
1329 struct net *net = dev_net(dev);
1330 int ct;
1331 for (ct = net->ipv4.maxvif-1; ct >= 0; ct--) {
1332 if (net->ipv4.vif_table[ct].dev == dev)
1333 break;
1334 }
1335 return ct;
1336 }
1337
1338 /* "local" means that we should preserve one skb (for local delivery) */
1339
1340 static int ip_mr_forward(struct sk_buff *skb, struct mfc_cache *cache, int local)
1341 {
1342 int psend = -1;
1343 int vif, ct;
1344 struct net *net = mfc_net(cache);
1345
1346 vif = cache->mfc_parent;
1347 cache->mfc_un.res.pkt++;
1348 cache->mfc_un.res.bytes += skb->len;
1349
1350 /*
1351 * Wrong interface: drop packet and (maybe) send PIM assert.
1352 */
1353 if (net->ipv4.vif_table[vif].dev != skb->dev) {
1354 int true_vifi;
1355
1356 if (skb_rtable(skb)->fl.iif == 0) {
1357 /* It is our own packet, looped back.
1358 Very complicated situation...
1359
1360 The best workaround until routing daemons will be
1361 fixed is not to redistribute packet, if it was
1362 send through wrong interface. It means, that
1363 multicast applications WILL NOT work for
1364 (S,G), which have default multicast route pointing
1365 to wrong oif. In any case, it is not a good
1366 idea to use multicasting applications on router.
1367 */
1368 goto dont_forward;
1369 }
1370
1371 cache->mfc_un.res.wrong_if++;
1372 true_vifi = ipmr_find_vif(skb->dev);
1373
1374 if (true_vifi >= 0 && net->ipv4.mroute_do_assert &&
1375 /* pimsm uses asserts, when switching from RPT to SPT,
1376 so that we cannot check that packet arrived on an oif.
1377 It is bad, but otherwise we would need to move pretty
1378 large chunk of pimd to kernel. Ough... --ANK
1379 */
1380 (net->ipv4.mroute_do_pim ||
1381 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1382 time_after(jiffies,
1383 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1384 cache->mfc_un.res.last_assert = jiffies;
1385 ipmr_cache_report(net, skb, true_vifi, IGMPMSG_WRONGVIF);
1386 }
1387 goto dont_forward;
1388 }
1389
1390 net->ipv4.vif_table[vif].pkt_in++;
1391 net->ipv4.vif_table[vif].bytes_in += skb->len;
1392
1393 /*
1394 * Forward the frame
1395 */
1396 for (ct = cache->mfc_un.res.maxvif-1; ct >= cache->mfc_un.res.minvif; ct--) {
1397 if (ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1398 if (psend != -1) {
1399 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1400 if (skb2)
1401 ipmr_queue_xmit(skb2, cache, psend);
1402 }
1403 psend = ct;
1404 }
1405 }
1406 if (psend != -1) {
1407 if (local) {
1408 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1409 if (skb2)
1410 ipmr_queue_xmit(skb2, cache, psend);
1411 } else {
1412 ipmr_queue_xmit(skb, cache, psend);
1413 return 0;
1414 }
1415 }
1416
1417 dont_forward:
1418 if (!local)
1419 kfree_skb(skb);
1420 return 0;
1421 }
1422
1423
1424 /*
1425 * Multicast packets for forwarding arrive here
1426 */
1427
1428 int ip_mr_input(struct sk_buff *skb)
1429 {
1430 struct mfc_cache *cache;
1431 struct net *net = dev_net(skb->dev);
1432 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1433
1434 /* Packet is looped back after forward, it should not be
1435 forwarded second time, but still can be delivered locally.
1436 */
1437 if (IPCB(skb)->flags&IPSKB_FORWARDED)
1438 goto dont_forward;
1439
1440 if (!local) {
1441 if (IPCB(skb)->opt.router_alert) {
1442 if (ip_call_ra_chain(skb))
1443 return 0;
1444 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP){
1445 /* IGMPv1 (and broken IGMPv2 implementations sort of
1446 Cisco IOS <= 11.2(8)) do not put router alert
1447 option to IGMP packets destined to routable
1448 groups. It is very bad, because it means
1449 that we can forward NO IGMP messages.
1450 */
1451 read_lock(&mrt_lock);
1452 if (net->ipv4.mroute_sk) {
1453 nf_reset(skb);
1454 raw_rcv(net->ipv4.mroute_sk, skb);
1455 read_unlock(&mrt_lock);
1456 return 0;
1457 }
1458 read_unlock(&mrt_lock);
1459 }
1460 }
1461
1462 read_lock(&mrt_lock);
1463 cache = ipmr_cache_find(net, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1464
1465 /*
1466 * No usable cache entry
1467 */
1468 if (cache == NULL) {
1469 int vif;
1470
1471 if (local) {
1472 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1473 ip_local_deliver(skb);
1474 if (skb2 == NULL) {
1475 read_unlock(&mrt_lock);
1476 return -ENOBUFS;
1477 }
1478 skb = skb2;
1479 }
1480
1481 vif = ipmr_find_vif(skb->dev);
1482 if (vif >= 0) {
1483 int err = ipmr_cache_unresolved(net, vif, skb);
1484 read_unlock(&mrt_lock);
1485
1486 return err;
1487 }
1488 read_unlock(&mrt_lock);
1489 kfree_skb(skb);
1490 return -ENODEV;
1491 }
1492
1493 ip_mr_forward(skb, cache, local);
1494
1495 read_unlock(&mrt_lock);
1496
1497 if (local)
1498 return ip_local_deliver(skb);
1499
1500 return 0;
1501
1502 dont_forward:
1503 if (local)
1504 return ip_local_deliver(skb);
1505 kfree_skb(skb);
1506 return 0;
1507 }
1508
1509 #ifdef CONFIG_IP_PIMSM
1510 static int __pim_rcv(struct sk_buff *skb, unsigned int pimlen)
1511 {
1512 struct net_device *reg_dev = NULL;
1513 struct iphdr *encap;
1514 struct net *net = dev_net(skb->dev);
1515
1516 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
1517 /*
1518 Check that:
1519 a. packet is really destinted to a multicast group
1520 b. packet is not a NULL-REGISTER
1521 c. packet is not truncated
1522 */
1523 if (!ipv4_is_multicast(encap->daddr) ||
1524 encap->tot_len == 0 ||
1525 ntohs(encap->tot_len) + pimlen > skb->len)
1526 return 1;
1527
1528 read_lock(&mrt_lock);
1529 if (net->ipv4.mroute_reg_vif_num >= 0)
1530 reg_dev = net->ipv4.vif_table[net->ipv4.mroute_reg_vif_num].dev;
1531 if (reg_dev)
1532 dev_hold(reg_dev);
1533 read_unlock(&mrt_lock);
1534
1535 if (reg_dev == NULL)
1536 return 1;
1537
1538 skb->mac_header = skb->network_header;
1539 skb_pull(skb, (u8*)encap - skb->data);
1540 skb_reset_network_header(skb);
1541 skb->dev = reg_dev;
1542 skb->protocol = htons(ETH_P_IP);
1543 skb->ip_summed = 0;
1544 skb->pkt_type = PACKET_HOST;
1545 skb_dst_drop(skb);
1546 reg_dev->stats.rx_bytes += skb->len;
1547 reg_dev->stats.rx_packets++;
1548 nf_reset(skb);
1549 netif_rx(skb);
1550 dev_put(reg_dev);
1551
1552 return 0;
1553 }
1554 #endif
1555
1556 #ifdef CONFIG_IP_PIMSM_V1
1557 /*
1558 * Handle IGMP messages of PIMv1
1559 */
1560
1561 int pim_rcv_v1(struct sk_buff * skb)
1562 {
1563 struct igmphdr *pim;
1564 struct net *net = dev_net(skb->dev);
1565
1566 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
1567 goto drop;
1568
1569 pim = igmp_hdr(skb);
1570
1571 if (!net->ipv4.mroute_do_pim ||
1572 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
1573 goto drop;
1574
1575 if (__pim_rcv(skb, sizeof(*pim))) {
1576 drop:
1577 kfree_skb(skb);
1578 }
1579 return 0;
1580 }
1581 #endif
1582
1583 #ifdef CONFIG_IP_PIMSM_V2
1584 static int pim_rcv(struct sk_buff * skb)
1585 {
1586 struct pimreghdr *pim;
1587
1588 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
1589 goto drop;
1590
1591 pim = (struct pimreghdr *)skb_transport_header(skb);
1592 if (pim->type != ((PIM_VERSION<<4)|(PIM_REGISTER)) ||
1593 (pim->flags&PIM_NULL_REGISTER) ||
1594 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
1595 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
1596 goto drop;
1597
1598 if (__pim_rcv(skb, sizeof(*pim))) {
1599 drop:
1600 kfree_skb(skb);
1601 }
1602 return 0;
1603 }
1604 #endif
1605
1606 static int
1607 ipmr_fill_mroute(struct sk_buff *skb, struct mfc_cache *c, struct rtmsg *rtm)
1608 {
1609 int ct;
1610 struct rtnexthop *nhp;
1611 struct net *net = mfc_net(c);
1612 struct net_device *dev = net->ipv4.vif_table[c->mfc_parent].dev;
1613 u8 *b = skb_tail_pointer(skb);
1614 struct rtattr *mp_head;
1615
1616 if (dev)
1617 RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex);
1618
1619 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
1620
1621 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
1622 if (c->mfc_un.res.ttls[ct] < 255) {
1623 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
1624 goto rtattr_failure;
1625 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
1626 nhp->rtnh_flags = 0;
1627 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
1628 nhp->rtnh_ifindex = net->ipv4.vif_table[ct].dev->ifindex;
1629 nhp->rtnh_len = sizeof(*nhp);
1630 }
1631 }
1632 mp_head->rta_type = RTA_MULTIPATH;
1633 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
1634 rtm->rtm_type = RTN_MULTICAST;
1635 return 1;
1636
1637 rtattr_failure:
1638 nlmsg_trim(skb, b);
1639 return -EMSGSIZE;
1640 }
1641
1642 int ipmr_get_route(struct net *net,
1643 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
1644 {
1645 int err;
1646 struct mfc_cache *cache;
1647 struct rtable *rt = skb_rtable(skb);
1648
1649 read_lock(&mrt_lock);
1650 cache = ipmr_cache_find(net, rt->rt_src, rt->rt_dst);
1651
1652 if (cache == NULL) {
1653 struct sk_buff *skb2;
1654 struct iphdr *iph;
1655 struct net_device *dev;
1656 int vif;
1657
1658 if (nowait) {
1659 read_unlock(&mrt_lock);
1660 return -EAGAIN;
1661 }
1662
1663 dev = skb->dev;
1664 if (dev == NULL || (vif = ipmr_find_vif(dev)) < 0) {
1665 read_unlock(&mrt_lock);
1666 return -ENODEV;
1667 }
1668 skb2 = skb_clone(skb, GFP_ATOMIC);
1669 if (!skb2) {
1670 read_unlock(&mrt_lock);
1671 return -ENOMEM;
1672 }
1673
1674 skb_push(skb2, sizeof(struct iphdr));
1675 skb_reset_network_header(skb2);
1676 iph = ip_hdr(skb2);
1677 iph->ihl = sizeof(struct iphdr) >> 2;
1678 iph->saddr = rt->rt_src;
1679 iph->daddr = rt->rt_dst;
1680 iph->version = 0;
1681 err = ipmr_cache_unresolved(net, vif, skb2);
1682 read_unlock(&mrt_lock);
1683 return err;
1684 }
1685
1686 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
1687 cache->mfc_flags |= MFC_NOTIFY;
1688 err = ipmr_fill_mroute(skb, cache, rtm);
1689 read_unlock(&mrt_lock);
1690 return err;
1691 }
1692
1693 #ifdef CONFIG_PROC_FS
1694 /*
1695 * The /proc interfaces to multicast routing /proc/ip_mr_cache /proc/ip_mr_vif
1696 */
1697 struct ipmr_vif_iter {
1698 struct seq_net_private p;
1699 int ct;
1700 };
1701
1702 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
1703 struct ipmr_vif_iter *iter,
1704 loff_t pos)
1705 {
1706 for (iter->ct = 0; iter->ct < net->ipv4.maxvif; ++iter->ct) {
1707 if (!VIF_EXISTS(net, iter->ct))
1708 continue;
1709 if (pos-- == 0)
1710 return &net->ipv4.vif_table[iter->ct];
1711 }
1712 return NULL;
1713 }
1714
1715 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
1716 __acquires(mrt_lock)
1717 {
1718 struct net *net = seq_file_net(seq);
1719
1720 read_lock(&mrt_lock);
1721 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
1722 : SEQ_START_TOKEN;
1723 }
1724
1725 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1726 {
1727 struct ipmr_vif_iter *iter = seq->private;
1728 struct net *net = seq_file_net(seq);
1729
1730 ++*pos;
1731 if (v == SEQ_START_TOKEN)
1732 return ipmr_vif_seq_idx(net, iter, 0);
1733
1734 while (++iter->ct < net->ipv4.maxvif) {
1735 if (!VIF_EXISTS(net, iter->ct))
1736 continue;
1737 return &net->ipv4.vif_table[iter->ct];
1738 }
1739 return NULL;
1740 }
1741
1742 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
1743 __releases(mrt_lock)
1744 {
1745 read_unlock(&mrt_lock);
1746 }
1747
1748 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
1749 {
1750 struct net *net = seq_file_net(seq);
1751
1752 if (v == SEQ_START_TOKEN) {
1753 seq_puts(seq,
1754 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
1755 } else {
1756 const struct vif_device *vif = v;
1757 const char *name = vif->dev ? vif->dev->name : "none";
1758
1759 seq_printf(seq,
1760 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
1761 vif - net->ipv4.vif_table,
1762 name, vif->bytes_in, vif->pkt_in,
1763 vif->bytes_out, vif->pkt_out,
1764 vif->flags, vif->local, vif->remote);
1765 }
1766 return 0;
1767 }
1768
1769 static const struct seq_operations ipmr_vif_seq_ops = {
1770 .start = ipmr_vif_seq_start,
1771 .next = ipmr_vif_seq_next,
1772 .stop = ipmr_vif_seq_stop,
1773 .show = ipmr_vif_seq_show,
1774 };
1775
1776 static int ipmr_vif_open(struct inode *inode, struct file *file)
1777 {
1778 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
1779 sizeof(struct ipmr_vif_iter));
1780 }
1781
1782 static const struct file_operations ipmr_vif_fops = {
1783 .owner = THIS_MODULE,
1784 .open = ipmr_vif_open,
1785 .read = seq_read,
1786 .llseek = seq_lseek,
1787 .release = seq_release_net,
1788 };
1789
1790 struct ipmr_mfc_iter {
1791 struct seq_net_private p;
1792 struct mfc_cache **cache;
1793 int ct;
1794 };
1795
1796
1797 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
1798 struct ipmr_mfc_iter *it, loff_t pos)
1799 {
1800 struct mfc_cache *mfc;
1801
1802 it->cache = net->ipv4.mfc_cache_array;
1803 read_lock(&mrt_lock);
1804 for (it->ct = 0; it->ct < MFC_LINES; it->ct++)
1805 for (mfc = net->ipv4.mfc_cache_array[it->ct];
1806 mfc; mfc = mfc->next)
1807 if (pos-- == 0)
1808 return mfc;
1809 read_unlock(&mrt_lock);
1810
1811 it->cache = &mfc_unres_queue;
1812 spin_lock_bh(&mfc_unres_lock);
1813 for (mfc = mfc_unres_queue; mfc; mfc = mfc->next)
1814 if (net_eq(mfc_net(mfc), net) &&
1815 pos-- == 0)
1816 return mfc;
1817 spin_unlock_bh(&mfc_unres_lock);
1818
1819 it->cache = NULL;
1820 return NULL;
1821 }
1822
1823
1824 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
1825 {
1826 struct ipmr_mfc_iter *it = seq->private;
1827 struct net *net = seq_file_net(seq);
1828
1829 it->cache = NULL;
1830 it->ct = 0;
1831 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
1832 : SEQ_START_TOKEN;
1833 }
1834
1835 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1836 {
1837 struct mfc_cache *mfc = v;
1838 struct ipmr_mfc_iter *it = seq->private;
1839 struct net *net = seq_file_net(seq);
1840
1841 ++*pos;
1842
1843 if (v == SEQ_START_TOKEN)
1844 return ipmr_mfc_seq_idx(net, seq->private, 0);
1845
1846 if (mfc->next)
1847 return mfc->next;
1848
1849 if (it->cache == &mfc_unres_queue)
1850 goto end_of_list;
1851
1852 BUG_ON(it->cache != net->ipv4.mfc_cache_array);
1853
1854 while (++it->ct < MFC_LINES) {
1855 mfc = net->ipv4.mfc_cache_array[it->ct];
1856 if (mfc)
1857 return mfc;
1858 }
1859
1860 /* exhausted cache_array, show unresolved */
1861 read_unlock(&mrt_lock);
1862 it->cache = &mfc_unres_queue;
1863 it->ct = 0;
1864
1865 spin_lock_bh(&mfc_unres_lock);
1866 mfc = mfc_unres_queue;
1867 while (mfc && !net_eq(mfc_net(mfc), net))
1868 mfc = mfc->next;
1869 if (mfc)
1870 return mfc;
1871
1872 end_of_list:
1873 spin_unlock_bh(&mfc_unres_lock);
1874 it->cache = NULL;
1875
1876 return NULL;
1877 }
1878
1879 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
1880 {
1881 struct ipmr_mfc_iter *it = seq->private;
1882 struct net *net = seq_file_net(seq);
1883
1884 if (it->cache == &mfc_unres_queue)
1885 spin_unlock_bh(&mfc_unres_lock);
1886 else if (it->cache == net->ipv4.mfc_cache_array)
1887 read_unlock(&mrt_lock);
1888 }
1889
1890 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
1891 {
1892 int n;
1893 struct net *net = seq_file_net(seq);
1894
1895 if (v == SEQ_START_TOKEN) {
1896 seq_puts(seq,
1897 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
1898 } else {
1899 const struct mfc_cache *mfc = v;
1900 const struct ipmr_mfc_iter *it = seq->private;
1901
1902 seq_printf(seq, "%08lX %08lX %-3hd",
1903 (unsigned long) mfc->mfc_mcastgrp,
1904 (unsigned long) mfc->mfc_origin,
1905 mfc->mfc_parent);
1906
1907 if (it->cache != &mfc_unres_queue) {
1908 seq_printf(seq, " %8lu %8lu %8lu",
1909 mfc->mfc_un.res.pkt,
1910 mfc->mfc_un.res.bytes,
1911 mfc->mfc_un.res.wrong_if);
1912 for (n = mfc->mfc_un.res.minvif;
1913 n < mfc->mfc_un.res.maxvif; n++ ) {
1914 if (VIF_EXISTS(net, n) &&
1915 mfc->mfc_un.res.ttls[n] < 255)
1916 seq_printf(seq,
1917 " %2d:%-3d",
1918 n, mfc->mfc_un.res.ttls[n]);
1919 }
1920 } else {
1921 /* unresolved mfc_caches don't contain
1922 * pkt, bytes and wrong_if values
1923 */
1924 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
1925 }
1926 seq_putc(seq, '\n');
1927 }
1928 return 0;
1929 }
1930
1931 static const struct seq_operations ipmr_mfc_seq_ops = {
1932 .start = ipmr_mfc_seq_start,
1933 .next = ipmr_mfc_seq_next,
1934 .stop = ipmr_mfc_seq_stop,
1935 .show = ipmr_mfc_seq_show,
1936 };
1937
1938 static int ipmr_mfc_open(struct inode *inode, struct file *file)
1939 {
1940 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
1941 sizeof(struct ipmr_mfc_iter));
1942 }
1943
1944 static const struct file_operations ipmr_mfc_fops = {
1945 .owner = THIS_MODULE,
1946 .open = ipmr_mfc_open,
1947 .read = seq_read,
1948 .llseek = seq_lseek,
1949 .release = seq_release_net,
1950 };
1951 #endif
1952
1953 #ifdef CONFIG_IP_PIMSM_V2
1954 static const struct net_protocol pim_protocol = {
1955 .handler = pim_rcv,
1956 .netns_ok = 1,
1957 };
1958 #endif
1959
1960
1961 /*
1962 * Setup for IP multicast routing
1963 */
1964 static int __net_init ipmr_net_init(struct net *net)
1965 {
1966 int err = 0;
1967
1968 net->ipv4.vif_table = kcalloc(MAXVIFS, sizeof(struct vif_device),
1969 GFP_KERNEL);
1970 if (!net->ipv4.vif_table) {
1971 err = -ENOMEM;
1972 goto fail;
1973 }
1974
1975 /* Forwarding cache */
1976 net->ipv4.mfc_cache_array = kcalloc(MFC_LINES,
1977 sizeof(struct mfc_cache *),
1978 GFP_KERNEL);
1979 if (!net->ipv4.mfc_cache_array) {
1980 err = -ENOMEM;
1981 goto fail_mfc_cache;
1982 }
1983
1984 #ifdef CONFIG_IP_PIMSM
1985 net->ipv4.mroute_reg_vif_num = -1;
1986 #endif
1987
1988 #ifdef CONFIG_PROC_FS
1989 err = -ENOMEM;
1990 if (!proc_net_fops_create(net, "ip_mr_vif", 0, &ipmr_vif_fops))
1991 goto proc_vif_fail;
1992 if (!proc_net_fops_create(net, "ip_mr_cache", 0, &ipmr_mfc_fops))
1993 goto proc_cache_fail;
1994 #endif
1995 return 0;
1996
1997 #ifdef CONFIG_PROC_FS
1998 proc_cache_fail:
1999 proc_net_remove(net, "ip_mr_vif");
2000 proc_vif_fail:
2001 kfree(net->ipv4.mfc_cache_array);
2002 #endif
2003 fail_mfc_cache:
2004 kfree(net->ipv4.vif_table);
2005 fail:
2006 return err;
2007 }
2008
2009 static void __net_exit ipmr_net_exit(struct net *net)
2010 {
2011 #ifdef CONFIG_PROC_FS
2012 proc_net_remove(net, "ip_mr_cache");
2013 proc_net_remove(net, "ip_mr_vif");
2014 #endif
2015 kfree(net->ipv4.mfc_cache_array);
2016 kfree(net->ipv4.vif_table);
2017 }
2018
2019 static struct pernet_operations ipmr_net_ops = {
2020 .init = ipmr_net_init,
2021 .exit = ipmr_net_exit,
2022 };
2023
2024 int __init ip_mr_init(void)
2025 {
2026 int err;
2027
2028 mrt_cachep = kmem_cache_create("ip_mrt_cache",
2029 sizeof(struct mfc_cache),
2030 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
2031 NULL);
2032 if (!mrt_cachep)
2033 return -ENOMEM;
2034
2035 err = register_pernet_subsys(&ipmr_net_ops);
2036 if (err)
2037 goto reg_pernet_fail;
2038
2039 setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0);
2040 err = register_netdevice_notifier(&ip_mr_notifier);
2041 if (err)
2042 goto reg_notif_fail;
2043 #ifdef CONFIG_IP_PIMSM_V2
2044 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2045 printk(KERN_ERR "ip_mr_init: can't add PIM protocol\n");
2046 err = -EAGAIN;
2047 goto add_proto_fail;
2048 }
2049 #endif
2050 return 0;
2051
2052 #ifdef CONFIG_IP_PIMSM_V2
2053 add_proto_fail:
2054 unregister_netdevice_notifier(&ip_mr_notifier);
2055 #endif
2056 reg_notif_fail:
2057 del_timer(&ipmr_expire_timer);
2058 unregister_pernet_subsys(&ipmr_net_ops);
2059 reg_pernet_fail:
2060 kmem_cache_destroy(mrt_cachep);
2061 return err;
2062 }
mm of the moment repository