?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
md: factor out updating of 'recovery_offset'.
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv4 / ipmr.c
blob630a56df7b47ee80c9c77dcbb62048dc73ed9fc5
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 case 0:
473 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
474 if (!dev)
475 return -EADDRNOTAVAIL;
476 err = dev_set_allmulti(dev, 1);
477 if (err) {
478 dev_put(dev);
479 return err;
480 }
481 break;
482 default:
483 return -EINVAL;
484 }
485
486 if ((in_dev = __in_dev_get_rtnl(dev)) == NULL)
487 return -EADDRNOTAVAIL;
488 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
489 ip_rt_multicast_event(in_dev);
490
491 /*
492 * Fill in the VIF structures
493 */
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;
498 if (!mrtsock)
499 v->flags |= VIFF_STATIC;
500 v->threshold = vifc->vifc_threshold;
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);
511 v->dev = dev;
512 #ifdef CONFIG_IP_PIMSM
513 if (v->flags&VIFF_REGISTER)
514 net->ipv4.mroute_reg_vif_num = vifi;
515 #endif
516 if (vifi+1 > net->ipv4.maxvif)
517 net->ipv4.maxvif = vifi+1;
518 write_unlock_bh(&mrt_lock);
519 return 0;
520 }
521
522 static struct mfc_cache *ipmr_cache_find(struct net *net,
523 __be32 origin,
524 __be32 mcastgrp)
525 {
526 int line = MFC_HASH(mcastgrp, origin);
527 struct mfc_cache *c;
528
529 for (c = net->ipv4.mfc_cache_array[line]; c; c = c->next) {
530 if (c->mfc_origin==origin && c->mfc_mcastgrp==mcastgrp)
531 break;
532 }
533 return c;
534 }
535
536 /*
537 * Allocate a multicast cache entry
538 */
539 static struct mfc_cache *ipmr_cache_alloc(struct net *net)
540 {
541 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
542 if (c == NULL)
543 return NULL;
544 c->mfc_un.res.minvif = MAXVIFS;
545 mfc_net_set(c, net);
546 return c;
547 }
548
549 static struct mfc_cache *ipmr_cache_alloc_unres(struct net *net)
550 {
551 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
552 if (c == NULL)
553 return NULL;
554 skb_queue_head_init(&c->mfc_un.unres.unresolved);
555 c->mfc_un.unres.expires = jiffies + 10*HZ;
556 mfc_net_set(c, net);
557 return c;
558 }
559
560 /*
561 * A cache entry has gone into a resolved state from queued
562 */
563
564 static void ipmr_cache_resolve(struct mfc_cache *uc, struct mfc_cache *c)
565 {
566 struct sk_buff *skb;
567 struct nlmsgerr *e;
568
569 /*
570 * Play the pending entries through our router
571 */
572
573 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
574 if (ip_hdr(skb)->version == 0) {
575 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
576
577 if (ipmr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) {
578 nlh->nlmsg_len = (skb_tail_pointer(skb) -
579 (u8 *)nlh);
580 } else {
581 nlh->nlmsg_type = NLMSG_ERROR;
582 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
583 skb_trim(skb, nlh->nlmsg_len);
584 e = NLMSG_DATA(nlh);
585 e->error = -EMSGSIZE;
586 memset(&e->msg, 0, sizeof(e->msg));
587 }
588
589 rtnl_unicast(skb, mfc_net(c), NETLINK_CB(skb).pid);
590 } else
591 ip_mr_forward(skb, c, 0);
592 }
593 }
594
595 /*
596 * Bounce a cache query up to mrouted. We could use netlink for this but mrouted
597 * expects the following bizarre scheme.
598 *
599 * Called under mrt_lock.
600 */
601
602 static int ipmr_cache_report(struct net *net,
603 struct sk_buff *pkt, vifi_t vifi, int assert)
604 {
605 struct sk_buff *skb;
606 const int ihl = ip_hdrlen(pkt);
607 struct igmphdr *igmp;
608 struct igmpmsg *msg;
609 int ret;
610
611 #ifdef CONFIG_IP_PIMSM
612 if (assert == IGMPMSG_WHOLEPKT)
613 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
614 else
615 #endif
616 skb = alloc_skb(128, GFP_ATOMIC);
617
618 if (!skb)
619 return -ENOBUFS;
620
621 #ifdef CONFIG_IP_PIMSM
622 if (assert == IGMPMSG_WHOLEPKT) {
623 /* Ugly, but we have no choice with this interface.
624 Duplicate old header, fix ihl, length etc.
625 And all this only to mangle msg->im_msgtype and
626 to set msg->im_mbz to "mbz" :-)
627 */
628 skb_push(skb, sizeof(struct iphdr));
629 skb_reset_network_header(skb);
630 skb_reset_transport_header(skb);
631 msg = (struct igmpmsg *)skb_network_header(skb);
632 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
633 msg->im_msgtype = IGMPMSG_WHOLEPKT;
634 msg->im_mbz = 0;
635 msg->im_vif = net->ipv4.mroute_reg_vif_num;
636 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
637 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
638 sizeof(struct iphdr));
639 } else
640 #endif
641 {
642
643 /*
644 * Copy the IP header
645 */
646
647 skb->network_header = skb->tail;
648 skb_put(skb, ihl);
649 skb_copy_to_linear_data(skb, pkt->data, ihl);
650 ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */
651 msg = (struct igmpmsg *)skb_network_header(skb);
652 msg->im_vif = vifi;
653 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
654
655 /*
656 * Add our header
657 */
658
659 igmp=(struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
660 igmp->type =
661 msg->im_msgtype = assert;
662 igmp->code = 0;
663 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
664 skb->transport_header = skb->network_header;
665 }
666
667 if (net->ipv4.mroute_sk == NULL) {
668 kfree_skb(skb);
669 return -EINVAL;
670 }
671
672 /*
673 * Deliver to mrouted
674 */
675 ret = sock_queue_rcv_skb(net->ipv4.mroute_sk, skb);
676 if (ret < 0) {
677 if (net_ratelimit())
678 printk(KERN_WARNING "mroute: pending queue full, dropping entries.\n");
679 kfree_skb(skb);
680 }
681
682 return ret;
683 }
684
685 /*
686 * Queue a packet for resolution. It gets locked cache entry!
687 */
688
689 static int
690 ipmr_cache_unresolved(struct net *net, vifi_t vifi, struct sk_buff *skb)
691 {
692 int err;
693 struct mfc_cache *c;
694 const struct iphdr *iph = ip_hdr(skb);
695
696 spin_lock_bh(&mfc_unres_lock);
697 for (c=mfc_unres_queue; c; c=c->next) {
698 if (net_eq(mfc_net(c), net) &&
699 c->mfc_mcastgrp == iph->daddr &&
700 c->mfc_origin == iph->saddr)
701 break;
702 }
703
704 if (c == NULL) {
705 /*
706 * Create a new entry if allowable
707 */
708
709 if (atomic_read(&net->ipv4.cache_resolve_queue_len) >= 10 ||
710 (c = ipmr_cache_alloc_unres(net)) == NULL) {
711 spin_unlock_bh(&mfc_unres_lock);
712
713 kfree_skb(skb);
714 return -ENOBUFS;
715 }
716
717 /*
718 * Fill in the new cache entry
719 */
720 c->mfc_parent = -1;
721 c->mfc_origin = iph->saddr;
722 c->mfc_mcastgrp = iph->daddr;
723
724 /*
725 * Reflect first query at mrouted.
726 */
727 err = ipmr_cache_report(net, skb, vifi, IGMPMSG_NOCACHE);
728 if (err < 0) {
729 /* If the report failed throw the cache entry
730 out - Brad Parker
731 */
732 spin_unlock_bh(&mfc_unres_lock);
733
734 ipmr_cache_free(c);
735 kfree_skb(skb);
736 return err;
737 }
738
739 atomic_inc(&net->ipv4.cache_resolve_queue_len);
740 c->next = mfc_unres_queue;
741 mfc_unres_queue = c;
742
743 mod_timer(&ipmr_expire_timer, c->mfc_un.unres.expires);
744 }
745
746 /*
747 * See if we can append the packet
748 */
749 if (c->mfc_un.unres.unresolved.qlen>3) {
750 kfree_skb(skb);
751 err = -ENOBUFS;
752 } else {
753 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
754 err = 0;
755 }
756
757 spin_unlock_bh(&mfc_unres_lock);
758 return err;
759 }
760
761 /*
762 * MFC cache manipulation by user space mroute daemon
763 */
764
765 static int ipmr_mfc_delete(struct net *net, struct mfcctl *mfc)
766 {
767 int line;
768 struct mfc_cache *c, **cp;
769
770 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
771
772 for (cp = &net->ipv4.mfc_cache_array[line];
773 (c = *cp) != NULL; cp = &c->next) {
774 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
775 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr) {
776 write_lock_bh(&mrt_lock);
777 *cp = c->next;
778 write_unlock_bh(&mrt_lock);
779
780 ipmr_cache_free(c);
781 return 0;
782 }
783 }
784 return -ENOENT;
785 }
786
787 static int ipmr_mfc_add(struct net *net, struct mfcctl *mfc, int mrtsock)
788 {
789 int line;
790 struct mfc_cache *uc, *c, **cp;
791
792 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
793
794 for (cp = &net->ipv4.mfc_cache_array[line];
795 (c = *cp) != NULL; cp = &c->next) {
796 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
797 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr)
798 break;
799 }
800
801 if (c != NULL) {
802 write_lock_bh(&mrt_lock);
803 c->mfc_parent = mfc->mfcc_parent;
804 ipmr_update_thresholds(c, mfc->mfcc_ttls);
805 if (!mrtsock)
806 c->mfc_flags |= MFC_STATIC;
807 write_unlock_bh(&mrt_lock);
808 return 0;
809 }
810
811 if (!ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
812 return -EINVAL;
813
814 c = ipmr_cache_alloc(net);
815 if (c == NULL)
816 return -ENOMEM;
817
818 c->mfc_origin = mfc->mfcc_origin.s_addr;
819 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
820 c->mfc_parent = mfc->mfcc_parent;
821 ipmr_update_thresholds(c, mfc->mfcc_ttls);
822 if (!mrtsock)
823 c->mfc_flags |= MFC_STATIC;
824
825 write_lock_bh(&mrt_lock);
826 c->next = net->ipv4.mfc_cache_array[line];
827 net->ipv4.mfc_cache_array[line] = c;
828 write_unlock_bh(&mrt_lock);
829
830 /*
831 * Check to see if we resolved a queued list. If so we
832 * need to send on the frames and tidy up.
833 */
834 spin_lock_bh(&mfc_unres_lock);
835 for (cp = &mfc_unres_queue; (uc=*cp) != NULL;
836 cp = &uc->next) {
837 if (net_eq(mfc_net(uc), net) &&
838 uc->mfc_origin == c->mfc_origin &&
839 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
840 *cp = uc->next;
841 atomic_dec(&net->ipv4.cache_resolve_queue_len);
842 break;
843 }
844 }
845 if (mfc_unres_queue == NULL)
846 del_timer(&ipmr_expire_timer);
847 spin_unlock_bh(&mfc_unres_lock);
848
849 if (uc) {
850 ipmr_cache_resolve(uc, c);
851 ipmr_cache_free(uc);
852 }
853 return 0;
854 }
855
856 /*
857 * Close the multicast socket, and clear the vif tables etc
858 */
859
860 static void mroute_clean_tables(struct net *net)
861 {
862 int i;
863
864 /*
865 * Shut down all active vif entries
866 */
867 for (i = 0; i < net->ipv4.maxvif; i++) {
868 if (!(net->ipv4.vif_table[i].flags&VIFF_STATIC))
869 vif_delete(net, i, 0);
870 }
871
872 /*
873 * Wipe the cache
874 */
875 for (i=0; i<MFC_LINES; i++) {
876 struct mfc_cache *c, **cp;
877
878 cp = &net->ipv4.mfc_cache_array[i];
879 while ((c = *cp) != NULL) {
880 if (c->mfc_flags&MFC_STATIC) {
881 cp = &c->next;
882 continue;
883 }
884 write_lock_bh(&mrt_lock);
885 *cp = c->next;
886 write_unlock_bh(&mrt_lock);
887
888 ipmr_cache_free(c);
889 }
890 }
891
892 if (atomic_read(&net->ipv4.cache_resolve_queue_len) != 0) {
893 struct mfc_cache *c, **cp;
894
895 spin_lock_bh(&mfc_unres_lock);
896 cp = &mfc_unres_queue;
897 while ((c = *cp) != NULL) {
898 if (!net_eq(mfc_net(c), net)) {
899 cp = &c->next;
900 continue;
901 }
902 *cp = c->next;
903
904 ipmr_destroy_unres(c);
905 }
906 spin_unlock_bh(&mfc_unres_lock);
907 }
908 }
909
910 static void mrtsock_destruct(struct sock *sk)
911 {
912 struct net *net = sock_net(sk);
913
914 rtnl_lock();
915 if (sk == net->ipv4.mroute_sk) {
916 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
917
918 write_lock_bh(&mrt_lock);
919 net->ipv4.mroute_sk = NULL;
920 write_unlock_bh(&mrt_lock);
921
922 mroute_clean_tables(net);
923 }
924 rtnl_unlock();
925 }
926
927 /*
928 * Socket options and virtual interface manipulation. The whole
929 * virtual interface system is a complete heap, but unfortunately
930 * that's how BSD mrouted happens to think. Maybe one day with a proper
931 * MOSPF/PIM router set up we can clean this up.
932 */
933
934 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
935 {
936 int ret;
937 struct vifctl vif;
938 struct mfcctl mfc;
939 struct net *net = sock_net(sk);
940
941 if (optname != MRT_INIT) {
942 if (sk != net->ipv4.mroute_sk && !capable(CAP_NET_ADMIN))
943 return -EACCES;
944 }
945
946 switch (optname) {
947 case MRT_INIT:
948 if (sk->sk_type != SOCK_RAW ||
949 inet_sk(sk)->num != IPPROTO_IGMP)
950 return -EOPNOTSUPP;
951 if (optlen != sizeof(int))
952 return -ENOPROTOOPT;
953
954 rtnl_lock();
955 if (net->ipv4.mroute_sk) {
956 rtnl_unlock();
957 return -EADDRINUSE;
958 }
959
960 ret = ip_ra_control(sk, 1, mrtsock_destruct);
961 if (ret == 0) {
962 write_lock_bh(&mrt_lock);
963 net->ipv4.mroute_sk = sk;
964 write_unlock_bh(&mrt_lock);
965
966 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
967 }
968 rtnl_unlock();
969 return ret;
970 case MRT_DONE:
971 if (sk != net->ipv4.mroute_sk)
972 return -EACCES;
973 return ip_ra_control(sk, 0, NULL);
974 case MRT_ADD_VIF:
975 case MRT_DEL_VIF:
976 if (optlen != sizeof(vif))
977 return -EINVAL;
978 if (copy_from_user(&vif, optval, sizeof(vif)))
979 return -EFAULT;
980 if (vif.vifc_vifi >= MAXVIFS)
981 return -ENFILE;
982 rtnl_lock();
983 if (optname == MRT_ADD_VIF) {
984 ret = vif_add(net, &vif, sk == net->ipv4.mroute_sk);
985 } else {
986 ret = vif_delete(net, vif.vifc_vifi, 0);
987 }
988 rtnl_unlock();
989 return ret;
990
991 /*
992 * Manipulate the forwarding caches. These live
993 * in a sort of kernel/user symbiosis.
994 */
995 case MRT_ADD_MFC:
996 case MRT_DEL_MFC:
997 if (optlen != sizeof(mfc))
998 return -EINVAL;
999 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1000 return -EFAULT;
1001 rtnl_lock();
1002 if (optname == MRT_DEL_MFC)
1003 ret = ipmr_mfc_delete(net, &mfc);
1004 else
1005 ret = ipmr_mfc_add(net, &mfc, sk == net->ipv4.mroute_sk);
1006 rtnl_unlock();
1007 return ret;
1008 /*
1009 * Control PIM assert.
1010 */
1011 case MRT_ASSERT:
1012 {
1013 int v;
1014 if (get_user(v,(int __user *)optval))
1015 return -EFAULT;
1016 net->ipv4.mroute_do_assert = (v) ? 1 : 0;
1017 return 0;
1018 }
1019 #ifdef CONFIG_IP_PIMSM
1020 case MRT_PIM:
1021 {
1022 int v;
1023
1024 if (get_user(v,(int __user *)optval))
1025 return -EFAULT;
1026 v = (v) ? 1 : 0;
1027
1028 rtnl_lock();
1029 ret = 0;
1030 if (v != net->ipv4.mroute_do_pim) {
1031 net->ipv4.mroute_do_pim = v;
1032 net->ipv4.mroute_do_assert = v;
1033 }
1034 rtnl_unlock();
1035 return ret;
1036 }
1037 #endif
1038 /*
1039 * Spurious command, or MRT_VERSION which you cannot
1040 * set.
1041 */
1042 default:
1043 return -ENOPROTOOPT;
1044 }
1045 }
1046
1047 /*
1048 * Getsock opt support for the multicast routing system.
1049 */
1050
1051 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1052 {
1053 int olr;
1054 int val;
1055 struct net *net = sock_net(sk);
1056
1057 if (optname != MRT_VERSION &&
1058 #ifdef CONFIG_IP_PIMSM
1059 optname!=MRT_PIM &&
1060 #endif
1061 optname!=MRT_ASSERT)
1062 return -ENOPROTOOPT;
1063
1064 if (get_user(olr, optlen))
1065 return -EFAULT;
1066
1067 olr = min_t(unsigned int, olr, sizeof(int));
1068 if (olr < 0)
1069 return -EINVAL;
1070
1071 if (put_user(olr, optlen))
1072 return -EFAULT;
1073 if (optname == MRT_VERSION)
1074 val = 0x0305;
1075 #ifdef CONFIG_IP_PIMSM
1076 else if (optname == MRT_PIM)
1077 val = net->ipv4.mroute_do_pim;
1078 #endif
1079 else
1080 val = net->ipv4.mroute_do_assert;
1081 if (copy_to_user(optval, &val, olr))
1082 return -EFAULT;
1083 return 0;
1084 }
1085
1086 /*
1087 * The IP multicast ioctl support routines.
1088 */
1089
1090 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1091 {
1092 struct sioc_sg_req sr;
1093 struct sioc_vif_req vr;
1094 struct vif_device *vif;
1095 struct mfc_cache *c;
1096 struct net *net = sock_net(sk);
1097
1098 switch (cmd) {
1099 case SIOCGETVIFCNT:
1100 if (copy_from_user(&vr, arg, sizeof(vr)))
1101 return -EFAULT;
1102 if (vr.vifi >= net->ipv4.maxvif)
1103 return -EINVAL;
1104 read_lock(&mrt_lock);
1105 vif = &net->ipv4.vif_table[vr.vifi];
1106 if (VIF_EXISTS(net, vr.vifi)) {
1107 vr.icount = vif->pkt_in;
1108 vr.ocount = vif->pkt_out;
1109 vr.ibytes = vif->bytes_in;
1110 vr.obytes = vif->bytes_out;
1111 read_unlock(&mrt_lock);
1112
1113 if (copy_to_user(arg, &vr, sizeof(vr)))
1114 return -EFAULT;
1115 return 0;
1116 }
1117 read_unlock(&mrt_lock);
1118 return -EADDRNOTAVAIL;
1119 case SIOCGETSGCNT:
1120 if (copy_from_user(&sr, arg, sizeof(sr)))
1121 return -EFAULT;
1122
1123 read_lock(&mrt_lock);
1124 c = ipmr_cache_find(net, sr.src.s_addr, sr.grp.s_addr);
1125 if (c) {
1126 sr.pktcnt = c->mfc_un.res.pkt;
1127 sr.bytecnt = c->mfc_un.res.bytes;
1128 sr.wrong_if = c->mfc_un.res.wrong_if;
1129 read_unlock(&mrt_lock);
1130
1131 if (copy_to_user(arg, &sr, sizeof(sr)))
1132 return -EFAULT;
1133 return 0;
1134 }
1135 read_unlock(&mrt_lock);
1136 return -EADDRNOTAVAIL;
1137 default:
1138 return -ENOIOCTLCMD;
1139 }
1140 }
1141
1142
1143 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1144 {
1145 struct net_device *dev = ptr;
1146 struct net *net = dev_net(dev);
1147 struct vif_device *v;
1148 int ct;
1149
1150 if (!net_eq(dev_net(dev), net))
1151 return NOTIFY_DONE;
1152
1153 if (event != NETDEV_UNREGISTER)
1154 return NOTIFY_DONE;
1155 v = &net->ipv4.vif_table[0];
1156 for (ct = 0; ct < net->ipv4.maxvif; ct++, v++) {
1157 if (v->dev == dev)
1158 vif_delete(net, ct, 1);
1159 }
1160 return NOTIFY_DONE;
1161 }
1162
1163
1164 static struct notifier_block ip_mr_notifier = {
1165 .notifier_call = ipmr_device_event,
1166 };
1167
1168 /*
1169 * Encapsulate a packet by attaching a valid IPIP header to it.
1170 * This avoids tunnel drivers and other mess and gives us the speed so
1171 * important for multicast video.
1172 */
1173
1174 static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr)
1175 {
1176 struct iphdr *iph;
1177 struct iphdr *old_iph = ip_hdr(skb);
1178
1179 skb_push(skb, sizeof(struct iphdr));
1180 skb->transport_header = skb->network_header;
1181 skb_reset_network_header(skb);
1182 iph = ip_hdr(skb);
1183
1184 iph->version = 4;
1185 iph->tos = old_iph->tos;
1186 iph->ttl = old_iph->ttl;
1187 iph->frag_off = 0;
1188 iph->daddr = daddr;
1189 iph->saddr = saddr;
1190 iph->protocol = IPPROTO_IPIP;
1191 iph->ihl = 5;
1192 iph->tot_len = htons(skb->len);
1193 ip_select_ident(iph, skb_dst(skb), NULL);
1194 ip_send_check(iph);
1195
1196 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1197 nf_reset(skb);
1198 }
1199
1200 static inline int ipmr_forward_finish(struct sk_buff *skb)
1201 {
1202 struct ip_options * opt = &(IPCB(skb)->opt);
1203
1204 IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
1205
1206 if (unlikely(opt->optlen))
1207 ip_forward_options(skb);
1208
1209 return dst_output(skb);
1210 }
1211
1212 /*
1213 * Processing handlers for ipmr_forward
1214 */
1215
1216 static void ipmr_queue_xmit(struct sk_buff *skb, struct mfc_cache *c, int vifi)
1217 {
1218 struct net *net = mfc_net(c);
1219 const struct iphdr *iph = ip_hdr(skb);
1220 struct vif_device *vif = &net->ipv4.vif_table[vifi];
1221 struct net_device *dev;
1222 struct rtable *rt;
1223 int encap = 0;
1224
1225 if (vif->dev == NULL)
1226 goto out_free;
1227
1228 #ifdef CONFIG_IP_PIMSM
1229 if (vif->flags & VIFF_REGISTER) {
1230 vif->pkt_out++;
1231 vif->bytes_out += skb->len;
1232 vif->dev->stats.tx_bytes += skb->len;
1233 vif->dev->stats.tx_packets++;
1234 ipmr_cache_report(net, skb, vifi, IGMPMSG_WHOLEPKT);
1235 goto out_free;
1236 }
1237 #endif
1238
1239 if (vif->flags&VIFF_TUNNEL) {
1240 struct flowi fl = { .oif = vif->link,
1241 .nl_u = { .ip4_u =
1242 { .daddr = vif->remote,
1243 .saddr = vif->local,
1244 .tos = RT_TOS(iph->tos) } },
1245 .proto = IPPROTO_IPIP };
1246 if (ip_route_output_key(net, &rt, &fl))
1247 goto out_free;
1248 encap = sizeof(struct iphdr);
1249 } else {
1250 struct flowi fl = { .oif = vif->link,
1251 .nl_u = { .ip4_u =
1252 { .daddr = iph->daddr,
1253 .tos = RT_TOS(iph->tos) } },
1254 .proto = IPPROTO_IPIP };
1255 if (ip_route_output_key(net, &rt, &fl))
1256 goto out_free;
1257 }
1258
1259 dev = rt->u.dst.dev;
1260
1261 if (skb->len+encap > dst_mtu(&rt->u.dst) && (ntohs(iph->frag_off) & IP_DF)) {
1262 /* Do not fragment multicasts. Alas, IPv4 does not
1263 allow to send ICMP, so that packets will disappear
1264 to blackhole.
1265 */
1266
1267 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
1268 ip_rt_put(rt);
1269 goto out_free;
1270 }
1271
1272 encap += LL_RESERVED_SPACE(dev) + rt->u.dst.header_len;
1273
1274 if (skb_cow(skb, encap)) {
1275 ip_rt_put(rt);
1276 goto out_free;
1277 }
1278
1279 vif->pkt_out++;
1280 vif->bytes_out += skb->len;
1281
1282 skb_dst_drop(skb);
1283 skb_dst_set(skb, &rt->u.dst);
1284 ip_decrease_ttl(ip_hdr(skb));
1285
1286 /* FIXME: forward and output firewalls used to be called here.
1287 * What do we do with netfilter? -- RR */
1288 if (vif->flags & VIFF_TUNNEL) {
1289 ip_encap(skb, vif->local, vif->remote);
1290 /* FIXME: extra output firewall step used to be here. --RR */
1291 vif->dev->stats.tx_packets++;
1292 vif->dev->stats.tx_bytes += skb->len;
1293 }
1294
1295 IPCB(skb)->flags |= IPSKB_FORWARDED;
1296
1297 /*
1298 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1299 * not only before forwarding, but after forwarding on all output
1300 * interfaces. It is clear, if mrouter runs a multicasting
1301 * program, it should receive packets not depending to what interface
1302 * program is joined.
1303 * If we will not make it, the program will have to join on all
1304 * interfaces. On the other hand, multihoming host (or router, but
1305 * not mrouter) cannot join to more than one interface - it will
1306 * result in receiving multiple packets.
1307 */
1308 NF_HOOK(PF_INET, NF_INET_FORWARD, skb, skb->dev, dev,
1309 ipmr_forward_finish);
1310 return;
1311
1312 out_free:
1313 kfree_skb(skb);
1314 return;
1315 }
1316
1317 static int ipmr_find_vif(struct net_device *dev)
1318 {
1319 struct net *net = dev_net(dev);
1320 int ct;
1321 for (ct = net->ipv4.maxvif-1; ct >= 0; ct--) {
1322 if (net->ipv4.vif_table[ct].dev == dev)
1323 break;
1324 }
1325 return ct;
1326 }
1327
1328 /* "local" means that we should preserve one skb (for local delivery) */
1329
1330 static int ip_mr_forward(struct sk_buff *skb, struct mfc_cache *cache, int local)
1331 {
1332 int psend = -1;
1333 int vif, ct;
1334 struct net *net = mfc_net(cache);
1335
1336 vif = cache->mfc_parent;
1337 cache->mfc_un.res.pkt++;
1338 cache->mfc_un.res.bytes += skb->len;
1339
1340 /*
1341 * Wrong interface: drop packet and (maybe) send PIM assert.
1342 */
1343 if (net->ipv4.vif_table[vif].dev != skb->dev) {
1344 int true_vifi;
1345
1346 if (skb_rtable(skb)->fl.iif == 0) {
1347 /* It is our own packet, looped back.
1348 Very complicated situation...
1349
1350 The best workaround until routing daemons will be
1351 fixed is not to redistribute packet, if it was
1352 send through wrong interface. It means, that
1353 multicast applications WILL NOT work for
1354 (S,G), which have default multicast route pointing
1355 to wrong oif. In any case, it is not a good
1356 idea to use multicasting applications on router.
1357 */
1358 goto dont_forward;
1359 }
1360
1361 cache->mfc_un.res.wrong_if++;
1362 true_vifi = ipmr_find_vif(skb->dev);
1363
1364 if (true_vifi >= 0 && net->ipv4.mroute_do_assert &&
1365 /* pimsm uses asserts, when switching from RPT to SPT,
1366 so that we cannot check that packet arrived on an oif.
1367 It is bad, but otherwise we would need to move pretty
1368 large chunk of pimd to kernel. Ough... --ANK
1369 */
1370 (net->ipv4.mroute_do_pim ||
1371 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1372 time_after(jiffies,
1373 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1374 cache->mfc_un.res.last_assert = jiffies;
1375 ipmr_cache_report(net, skb, true_vifi, IGMPMSG_WRONGVIF);
1376 }
1377 goto dont_forward;
1378 }
1379
1380 net->ipv4.vif_table[vif].pkt_in++;
1381 net->ipv4.vif_table[vif].bytes_in += skb->len;
1382
1383 /*
1384 * Forward the frame
1385 */
1386 for (ct = cache->mfc_un.res.maxvif-1; ct >= cache->mfc_un.res.minvif; ct--) {
1387 if (ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1388 if (psend != -1) {
1389 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1390 if (skb2)
1391 ipmr_queue_xmit(skb2, cache, psend);
1392 }
1393 psend = ct;
1394 }
1395 }
1396 if (psend != -1) {
1397 if (local) {
1398 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1399 if (skb2)
1400 ipmr_queue_xmit(skb2, cache, psend);
1401 } else {
1402 ipmr_queue_xmit(skb, cache, psend);
1403 return 0;
1404 }
1405 }
1406
1407 dont_forward:
1408 if (!local)
1409 kfree_skb(skb);
1410 return 0;
1411 }
1412
1413
1414 /*
1415 * Multicast packets for forwarding arrive here
1416 */
1417
1418 int ip_mr_input(struct sk_buff *skb)
1419 {
1420 struct mfc_cache *cache;
1421 struct net *net = dev_net(skb->dev);
1422 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1423
1424 /* Packet is looped back after forward, it should not be
1425 forwarded second time, but still can be delivered locally.
1426 */
1427 if (IPCB(skb)->flags&IPSKB_FORWARDED)
1428 goto dont_forward;
1429
1430 if (!local) {
1431 if (IPCB(skb)->opt.router_alert) {
1432 if (ip_call_ra_chain(skb))
1433 return 0;
1434 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP){
1435 /* IGMPv1 (and broken IGMPv2 implementations sort of
1436 Cisco IOS <= 11.2(8)) do not put router alert
1437 option to IGMP packets destined to routable
1438 groups. It is very bad, because it means
1439 that we can forward NO IGMP messages.
1440 */
1441 read_lock(&mrt_lock);
1442 if (net->ipv4.mroute_sk) {
1443 nf_reset(skb);
1444 raw_rcv(net->ipv4.mroute_sk, skb);
1445 read_unlock(&mrt_lock);
1446 return 0;
1447 }
1448 read_unlock(&mrt_lock);
1449 }
1450 }
1451
1452 read_lock(&mrt_lock);
1453 cache = ipmr_cache_find(net, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1454
1455 /*
1456 * No usable cache entry
1457 */
1458 if (cache == NULL) {
1459 int vif;
1460
1461 if (local) {
1462 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1463 ip_local_deliver(skb);
1464 if (skb2 == NULL) {
1465 read_unlock(&mrt_lock);
1466 return -ENOBUFS;
1467 }
1468 skb = skb2;
1469 }
1470
1471 vif = ipmr_find_vif(skb->dev);
1472 if (vif >= 0) {
1473 int err = ipmr_cache_unresolved(net, vif, skb);
1474 read_unlock(&mrt_lock);
1475
1476 return err;
1477 }
1478 read_unlock(&mrt_lock);
1479 kfree_skb(skb);
1480 return -ENODEV;
1481 }
1482
1483 ip_mr_forward(skb, cache, local);
1484
1485 read_unlock(&mrt_lock);
1486
1487 if (local)
1488 return ip_local_deliver(skb);
1489
1490 return 0;
1491
1492 dont_forward:
1493 if (local)
1494 return ip_local_deliver(skb);
1495 kfree_skb(skb);
1496 return 0;
1497 }
1498
1499 #ifdef CONFIG_IP_PIMSM
1500 static int __pim_rcv(struct sk_buff *skb, unsigned int pimlen)
1501 {
1502 struct net_device *reg_dev = NULL;
1503 struct iphdr *encap;
1504 struct net *net = dev_net(skb->dev);
1505
1506 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
1507 /*
1508 Check that:
1509 a. packet is really destinted to a multicast group
1510 b. packet is not a NULL-REGISTER
1511 c. packet is not truncated
1512 */
1513 if (!ipv4_is_multicast(encap->daddr) ||
1514 encap->tot_len == 0 ||
1515 ntohs(encap->tot_len) + pimlen > skb->len)
1516 return 1;
1517
1518 read_lock(&mrt_lock);
1519 if (net->ipv4.mroute_reg_vif_num >= 0)
1520 reg_dev = net->ipv4.vif_table[net->ipv4.mroute_reg_vif_num].dev;
1521 if (reg_dev)
1522 dev_hold(reg_dev);
1523 read_unlock(&mrt_lock);
1524
1525 if (reg_dev == NULL)
1526 return 1;
1527
1528 skb->mac_header = skb->network_header;
1529 skb_pull(skb, (u8*)encap - skb->data);
1530 skb_reset_network_header(skb);
1531 skb->dev = reg_dev;
1532 skb->protocol = htons(ETH_P_IP);
1533 skb->ip_summed = 0;
1534 skb->pkt_type = PACKET_HOST;
1535 skb_dst_drop(skb);
1536 reg_dev->stats.rx_bytes += skb->len;
1537 reg_dev->stats.rx_packets++;
1538 nf_reset(skb);
1539 netif_rx(skb);
1540 dev_put(reg_dev);
1541
1542 return 0;
1543 }
1544 #endif
1545
1546 #ifdef CONFIG_IP_PIMSM_V1
1547 /*
1548 * Handle IGMP messages of PIMv1
1549 */
1550
1551 int pim_rcv_v1(struct sk_buff * skb)
1552 {
1553 struct igmphdr *pim;
1554 struct net *net = dev_net(skb->dev);
1555
1556 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
1557 goto drop;
1558
1559 pim = igmp_hdr(skb);
1560
1561 if (!net->ipv4.mroute_do_pim ||
1562 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
1563 goto drop;
1564
1565 if (__pim_rcv(skb, sizeof(*pim))) {
1566 drop:
1567 kfree_skb(skb);
1568 }
1569 return 0;
1570 }
1571 #endif
1572
1573 #ifdef CONFIG_IP_PIMSM_V2
1574 static int pim_rcv(struct sk_buff * skb)
1575 {
1576 struct pimreghdr *pim;
1577
1578 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
1579 goto drop;
1580
1581 pim = (struct pimreghdr *)skb_transport_header(skb);
1582 if (pim->type != ((PIM_VERSION<<4)|(PIM_REGISTER)) ||
1583 (pim->flags&PIM_NULL_REGISTER) ||
1584 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
1585 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
1586 goto drop;
1587
1588 if (__pim_rcv(skb, sizeof(*pim))) {
1589 drop:
1590 kfree_skb(skb);
1591 }
1592 return 0;
1593 }
1594 #endif
1595
1596 static int
1597 ipmr_fill_mroute(struct sk_buff *skb, struct mfc_cache *c, struct rtmsg *rtm)
1598 {
1599 int ct;
1600 struct rtnexthop *nhp;
1601 struct net *net = mfc_net(c);
1602 struct net_device *dev = net->ipv4.vif_table[c->mfc_parent].dev;
1603 u8 *b = skb_tail_pointer(skb);
1604 struct rtattr *mp_head;
1605
1606 if (dev)
1607 RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex);
1608
1609 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
1610
1611 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
1612 if (c->mfc_un.res.ttls[ct] < 255) {
1613 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
1614 goto rtattr_failure;
1615 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
1616 nhp->rtnh_flags = 0;
1617 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
1618 nhp->rtnh_ifindex = net->ipv4.vif_table[ct].dev->ifindex;
1619 nhp->rtnh_len = sizeof(*nhp);
1620 }
1621 }
1622 mp_head->rta_type = RTA_MULTIPATH;
1623 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
1624 rtm->rtm_type = RTN_MULTICAST;
1625 return 1;
1626
1627 rtattr_failure:
1628 nlmsg_trim(skb, b);
1629 return -EMSGSIZE;
1630 }
1631
1632 int ipmr_get_route(struct net *net,
1633 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
1634 {
1635 int err;
1636 struct mfc_cache *cache;
1637 struct rtable *rt = skb_rtable(skb);
1638
1639 read_lock(&mrt_lock);
1640 cache = ipmr_cache_find(net, rt->rt_src, rt->rt_dst);
1641
1642 if (cache == NULL) {
1643 struct sk_buff *skb2;
1644 struct iphdr *iph;
1645 struct net_device *dev;
1646 int vif;
1647
1648 if (nowait) {
1649 read_unlock(&mrt_lock);
1650 return -EAGAIN;
1651 }
1652
1653 dev = skb->dev;
1654 if (dev == NULL || (vif = ipmr_find_vif(dev)) < 0) {
1655 read_unlock(&mrt_lock);
1656 return -ENODEV;
1657 }
1658 skb2 = skb_clone(skb, GFP_ATOMIC);
1659 if (!skb2) {
1660 read_unlock(&mrt_lock);
1661 return -ENOMEM;
1662 }
1663
1664 skb_push(skb2, sizeof(struct iphdr));
1665 skb_reset_network_header(skb2);
1666 iph = ip_hdr(skb2);
1667 iph->ihl = sizeof(struct iphdr) >> 2;
1668 iph->saddr = rt->rt_src;
1669 iph->daddr = rt->rt_dst;
1670 iph->version = 0;
1671 err = ipmr_cache_unresolved(net, vif, skb2);
1672 read_unlock(&mrt_lock);
1673 return err;
1674 }
1675
1676 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
1677 cache->mfc_flags |= MFC_NOTIFY;
1678 err = ipmr_fill_mroute(skb, cache, rtm);
1679 read_unlock(&mrt_lock);
1680 return err;
1681 }
1682
1683 #ifdef CONFIG_PROC_FS
1684 /*
1685 * The /proc interfaces to multicast routing /proc/ip_mr_cache /proc/ip_mr_vif
1686 */
1687 struct ipmr_vif_iter {
1688 struct seq_net_private p;
1689 int ct;
1690 };
1691
1692 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
1693 struct ipmr_vif_iter *iter,
1694 loff_t pos)
1695 {
1696 for (iter->ct = 0; iter->ct < net->ipv4.maxvif; ++iter->ct) {
1697 if (!VIF_EXISTS(net, iter->ct))
1698 continue;
1699 if (pos-- == 0)
1700 return &net->ipv4.vif_table[iter->ct];
1701 }
1702 return NULL;
1703 }
1704
1705 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
1706 __acquires(mrt_lock)
1707 {
1708 struct net *net = seq_file_net(seq);
1709
1710 read_lock(&mrt_lock);
1711 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
1712 : SEQ_START_TOKEN;
1713 }
1714
1715 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1716 {
1717 struct ipmr_vif_iter *iter = seq->private;
1718 struct net *net = seq_file_net(seq);
1719
1720 ++*pos;
1721 if (v == SEQ_START_TOKEN)
1722 return ipmr_vif_seq_idx(net, iter, 0);
1723
1724 while (++iter->ct < net->ipv4.maxvif) {
1725 if (!VIF_EXISTS(net, iter->ct))
1726 continue;
1727 return &net->ipv4.vif_table[iter->ct];
1728 }
1729 return NULL;
1730 }
1731
1732 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
1733 __releases(mrt_lock)
1734 {
1735 read_unlock(&mrt_lock);
1736 }
1737
1738 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
1739 {
1740 struct net *net = seq_file_net(seq);
1741
1742 if (v == SEQ_START_TOKEN) {
1743 seq_puts(seq,
1744 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
1745 } else {
1746 const struct vif_device *vif = v;
1747 const char *name = vif->dev ? vif->dev->name : "none";
1748
1749 seq_printf(seq,
1750 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
1751 vif - net->ipv4.vif_table,
1752 name, vif->bytes_in, vif->pkt_in,
1753 vif->bytes_out, vif->pkt_out,
1754 vif->flags, vif->local, vif->remote);
1755 }
1756 return 0;
1757 }
1758
1759 static const struct seq_operations ipmr_vif_seq_ops = {
1760 .start = ipmr_vif_seq_start,
1761 .next = ipmr_vif_seq_next,
1762 .stop = ipmr_vif_seq_stop,
1763 .show = ipmr_vif_seq_show,
1764 };
1765
1766 static int ipmr_vif_open(struct inode *inode, struct file *file)
1767 {
1768 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
1769 sizeof(struct ipmr_vif_iter));
1770 }
1771
1772 static const struct file_operations ipmr_vif_fops = {
1773 .owner = THIS_MODULE,
1774 .open = ipmr_vif_open,
1775 .read = seq_read,
1776 .llseek = seq_lseek,
1777 .release = seq_release_net,
1778 };
1779
1780 struct ipmr_mfc_iter {
1781 struct seq_net_private p;
1782 struct mfc_cache **cache;
1783 int ct;
1784 };
1785
1786
1787 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
1788 struct ipmr_mfc_iter *it, loff_t pos)
1789 {
1790 struct mfc_cache *mfc;
1791
1792 it->cache = net->ipv4.mfc_cache_array;
1793 read_lock(&mrt_lock);
1794 for (it->ct = 0; it->ct < MFC_LINES; it->ct++)
1795 for (mfc = net->ipv4.mfc_cache_array[it->ct];
1796 mfc; mfc = mfc->next)
1797 if (pos-- == 0)
1798 return mfc;
1799 read_unlock(&mrt_lock);
1800
1801 it->cache = &mfc_unres_queue;
1802 spin_lock_bh(&mfc_unres_lock);
1803 for (mfc = mfc_unres_queue; mfc; mfc = mfc->next)
1804 if (net_eq(mfc_net(mfc), net) &&
1805 pos-- == 0)
1806 return mfc;
1807 spin_unlock_bh(&mfc_unres_lock);
1808
1809 it->cache = NULL;
1810 return NULL;
1811 }
1812
1813
1814 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
1815 {
1816 struct ipmr_mfc_iter *it = seq->private;
1817 struct net *net = seq_file_net(seq);
1818
1819 it->cache = NULL;
1820 it->ct = 0;
1821 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
1822 : SEQ_START_TOKEN;
1823 }
1824
1825 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1826 {
1827 struct mfc_cache *mfc = v;
1828 struct ipmr_mfc_iter *it = seq->private;
1829 struct net *net = seq_file_net(seq);
1830
1831 ++*pos;
1832
1833 if (v == SEQ_START_TOKEN)
1834 return ipmr_mfc_seq_idx(net, seq->private, 0);
1835
1836 if (mfc->next)
1837 return mfc->next;
1838
1839 if (it->cache == &mfc_unres_queue)
1840 goto end_of_list;
1841
1842 BUG_ON(it->cache != net->ipv4.mfc_cache_array);
1843
1844 while (++it->ct < MFC_LINES) {
1845 mfc = net->ipv4.mfc_cache_array[it->ct];
1846 if (mfc)
1847 return mfc;
1848 }
1849
1850 /* exhausted cache_array, show unresolved */
1851 read_unlock(&mrt_lock);
1852 it->cache = &mfc_unres_queue;
1853 it->ct = 0;
1854
1855 spin_lock_bh(&mfc_unres_lock);
1856 mfc = mfc_unres_queue;
1857 while (mfc && !net_eq(mfc_net(mfc), net))
1858 mfc = mfc->next;
1859 if (mfc)
1860 return mfc;
1861
1862 end_of_list:
1863 spin_unlock_bh(&mfc_unres_lock);
1864 it->cache = NULL;
1865
1866 return NULL;
1867 }
1868
1869 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
1870 {
1871 struct ipmr_mfc_iter *it = seq->private;
1872 struct net *net = seq_file_net(seq);
1873
1874 if (it->cache == &mfc_unres_queue)
1875 spin_unlock_bh(&mfc_unres_lock);
1876 else if (it->cache == net->ipv4.mfc_cache_array)
1877 read_unlock(&mrt_lock);
1878 }
1879
1880 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
1881 {
1882 int n;
1883 struct net *net = seq_file_net(seq);
1884
1885 if (v == SEQ_START_TOKEN) {
1886 seq_puts(seq,
1887 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
1888 } else {
1889 const struct mfc_cache *mfc = v;
1890 const struct ipmr_mfc_iter *it = seq->private;
1891
1892 seq_printf(seq, "%08lX %08lX %-3hd",
1893 (unsigned long) mfc->mfc_mcastgrp,
1894 (unsigned long) mfc->mfc_origin,
1895 mfc->mfc_parent);
1896
1897 if (it->cache != &mfc_unres_queue) {
1898 seq_printf(seq, " %8lu %8lu %8lu",
1899 mfc->mfc_un.res.pkt,
1900 mfc->mfc_un.res.bytes,
1901 mfc->mfc_un.res.wrong_if);
1902 for (n = mfc->mfc_un.res.minvif;
1903 n < mfc->mfc_un.res.maxvif; n++ ) {
1904 if (VIF_EXISTS(net, n) &&
1905 mfc->mfc_un.res.ttls[n] < 255)
1906 seq_printf(seq,
1907 " %2d:%-3d",
1908 n, mfc->mfc_un.res.ttls[n]);
1909 }
1910 } else {
1911 /* unresolved mfc_caches don't contain
1912 * pkt, bytes and wrong_if values
1913 */
1914 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
1915 }
1916 seq_putc(seq, '\n');
1917 }
1918 return 0;
1919 }
1920
1921 static const struct seq_operations ipmr_mfc_seq_ops = {
1922 .start = ipmr_mfc_seq_start,
1923 .next = ipmr_mfc_seq_next,
1924 .stop = ipmr_mfc_seq_stop,
1925 .show = ipmr_mfc_seq_show,
1926 };
1927
1928 static int ipmr_mfc_open(struct inode *inode, struct file *file)
1929 {
1930 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
1931 sizeof(struct ipmr_mfc_iter));
1932 }
1933
1934 static const struct file_operations ipmr_mfc_fops = {
1935 .owner = THIS_MODULE,
1936 .open = ipmr_mfc_open,
1937 .read = seq_read,
1938 .llseek = seq_lseek,
1939 .release = seq_release_net,
1940 };
1941 #endif
1942
1943 #ifdef CONFIG_IP_PIMSM_V2
1944 static const struct net_protocol pim_protocol = {
1945 .handler = pim_rcv,
1946 .netns_ok = 1,
1947 };
1948 #endif
1949
1950
1951 /*
1952 * Setup for IP multicast routing
1953 */
1954 static int __net_init ipmr_net_init(struct net *net)
1955 {
1956 int err = 0;
1957
1958 net->ipv4.vif_table = kcalloc(MAXVIFS, sizeof(struct vif_device),
1959 GFP_KERNEL);
1960 if (!net->ipv4.vif_table) {
1961 err = -ENOMEM;
1962 goto fail;
1963 }
1964
1965 /* Forwarding cache */
1966 net->ipv4.mfc_cache_array = kcalloc(MFC_LINES,
1967 sizeof(struct mfc_cache *),
1968 GFP_KERNEL);
1969 if (!net->ipv4.mfc_cache_array) {
1970 err = -ENOMEM;
1971 goto fail_mfc_cache;
1972 }
1973
1974 #ifdef CONFIG_IP_PIMSM
1975 net->ipv4.mroute_reg_vif_num = -1;
1976 #endif
1977
1978 #ifdef CONFIG_PROC_FS
1979 err = -ENOMEM;
1980 if (!proc_net_fops_create(net, "ip_mr_vif", 0, &ipmr_vif_fops))
1981 goto proc_vif_fail;
1982 if (!proc_net_fops_create(net, "ip_mr_cache", 0, &ipmr_mfc_fops))
1983 goto proc_cache_fail;
1984 #endif
1985 return 0;
1986
1987 #ifdef CONFIG_PROC_FS
1988 proc_cache_fail:
1989 proc_net_remove(net, "ip_mr_vif");
1990 proc_vif_fail:
1991 kfree(net->ipv4.mfc_cache_array);
1992 #endif
1993 fail_mfc_cache:
1994 kfree(net->ipv4.vif_table);
1995 fail:
1996 return err;
1997 }
1998
1999 static void __net_exit ipmr_net_exit(struct net *net)
2000 {
2001 #ifdef CONFIG_PROC_FS
2002 proc_net_remove(net, "ip_mr_cache");
2003 proc_net_remove(net, "ip_mr_vif");
2004 #endif
2005 kfree(net->ipv4.mfc_cache_array);
2006 kfree(net->ipv4.vif_table);
2007 }
2008
2009 static struct pernet_operations ipmr_net_ops = {
2010 .init = ipmr_net_init,
2011 .exit = ipmr_net_exit,
2012 };
2013
2014 int __init ip_mr_init(void)
2015 {
2016 int err;
2017
2018 mrt_cachep = kmem_cache_create("ip_mrt_cache",
2019 sizeof(struct mfc_cache),
2020 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
2021 NULL);
2022 if (!mrt_cachep)
2023 return -ENOMEM;
2024
2025 err = register_pernet_subsys(&ipmr_net_ops);
2026 if (err)
2027 goto reg_pernet_fail;
2028
2029 setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0);
2030 err = register_netdevice_notifier(&ip_mr_notifier);
2031 if (err)
2032 goto reg_notif_fail;
2033 #ifdef CONFIG_IP_PIMSM_V2
2034 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2035 printk(KERN_ERR "ip_mr_init: can't add PIM protocol\n");
2036 err = -EAGAIN;
2037 goto add_proto_fail;
2038 }
2039 #endif
2040 return 0;
2041
2042 #ifdef CONFIG_IP_PIMSM_V2
2043 add_proto_fail:
2044 unregister_netdevice_notifier(&ip_mr_notifier);
2045 #endif
2046 reg_notif_fail:
2047 del_timer(&ipmr_expire_timer);
2048 unregister_pernet_subsys(&ipmr_net_ops);
2049 reg_pernet_fail:
2050 kmem_cache_destroy(mrt_cachep);
2051 return err;
2052 }
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree