]> bbs.cooldavid.org Git - net-next-2.6.git/blame_incremental - net/ipv6/ip6mr.c
git://bbs.cooldavid.org / net-next-2.6.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
bridge : Sanitize skb before it enters the IP stack
[net-next-2.6.git] / net / ipv6 / ip6mr.c
... / ...
CommitLineData
1/*
2 * Linux IPv6 multicast routing support for BSD pim6sd
3 * Based on net/ipv4/ipmr.c.
4 *
5 * (c) 2004 Mickael Hoerdt, <hoerdt@clarinet.u-strasbg.fr>
6 * LSIIT Laboratory, Strasbourg, France
7 * (c) 2004 Jean-Philippe Andriot, <jean-philippe.andriot@6WIND.com>
8 * 6WIND, Paris, France
9 * Copyright (C)2007,2008 USAGI/WIDE Project
10 * YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
11 *
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
16 *
17 */
18
19#include <asm/system.h>
20#include <asm/uaccess.h>
21#include <linux/types.h>
22#include <linux/sched.h>
23#include <linux/errno.h>
24#include <linux/timer.h>
25#include <linux/mm.h>
26#include <linux/kernel.h>
27#include <linux/fcntl.h>
28#include <linux/stat.h>
29#include <linux/socket.h>
30#include <linux/inet.h>
31#include <linux/netdevice.h>
32#include <linux/inetdevice.h>
33#include <linux/proc_fs.h>
34#include <linux/seq_file.h>
35#include <linux/init.h>
36#include <linux/slab.h>
37#include <net/protocol.h>
38#include <linux/skbuff.h>
39#include <net/sock.h>
40#include <net/raw.h>
41#include <linux/notifier.h>
42#include <linux/if_arp.h>
43#include <net/checksum.h>
44#include <net/netlink.h>
45#include <net/fib_rules.h>
46
47#include <net/ipv6.h>
48#include <net/ip6_route.h>
49#include <linux/mroute6.h>
50#include <linux/pim.h>
51#include <net/addrconf.h>
52#include <linux/netfilter_ipv6.h>
53#include <net/ip6_checksum.h>
54
55struct mr6_table {
56 struct list_head list;
57#ifdef CONFIG_NET_NS
58 struct net *net;
59#endif
60 u32 id;
61 struct sock *mroute6_sk;
62 struct timer_list ipmr_expire_timer;
63 struct list_head mfc6_unres_queue;
64 struct list_head mfc6_cache_array[MFC6_LINES];
65 struct mif_device vif6_table[MAXMIFS];
66 int maxvif;
67 atomic_t cache_resolve_queue_len;
68 int mroute_do_assert;
69 int mroute_do_pim;
70#ifdef CONFIG_IPV6_PIMSM_V2
71 int mroute_reg_vif_num;
72#endif
73};
74
75struct ip6mr_rule {
76 struct fib_rule common;
77};
78
79struct ip6mr_result {
80 struct mr6_table *mrt;
81};
82
83/* Big lock, protecting vif table, mrt cache and mroute socket state.
84 Note that the changes are semaphored via rtnl_lock.
85 */
86
87static DEFINE_RWLOCK(mrt_lock);
88
89/*
90 * Multicast router control variables
91 */
92
93#define MIF_EXISTS(_mrt, _idx) ((_mrt)->vif6_table[_idx].dev != NULL)
94
95/* Special spinlock for queue of unresolved entries */
96static DEFINE_SPINLOCK(mfc_unres_lock);
97
98/* We return to original Alan's scheme. Hash table of resolved
99 entries is changed only in process context and protected
100 with weak lock mrt_lock. Queue of unresolved entries is protected
101 with strong spinlock mfc_unres_lock.
102
103 In this case data path is free of exclusive locks at all.
104 */
105
106static struct kmem_cache *mrt_cachep __read_mostly;
107
108static struct mr6_table *ip6mr_new_table(struct net *net, u32 id);
109static void ip6mr_free_table(struct mr6_table *mrt);
110
111static int ip6_mr_forward(struct net *net, struct mr6_table *mrt,
112 struct sk_buff *skb, struct mfc6_cache *cache);
113static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
114 mifi_t mifi, int assert);
115static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
116 struct mfc6_cache *c, struct rtmsg *rtm);
117static int ip6mr_rtm_dumproute(struct sk_buff *skb,
118 struct netlink_callback *cb);
119static void mroute_clean_tables(struct mr6_table *mrt);
120static void ipmr_expire_process(unsigned long arg);
121
122#ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
123#define ip6mr_for_each_table(mrt, net) \
124 list_for_each_entry_rcu(mrt, &net->ipv6.mr6_tables, list)
125
126static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
127{
128 struct mr6_table *mrt;
129
130 ip6mr_for_each_table(mrt, net) {
131 if (mrt->id == id)
132 return mrt;
133 }
134 return NULL;
135}
136
137static int ip6mr_fib_lookup(struct net *net, struct flowi *flp,
138 struct mr6_table **mrt)
139{
140 struct ip6mr_result res;
141 struct fib_lookup_arg arg = { .result = &res, };
142 int err;
143
144 err = fib_rules_lookup(net->ipv6.mr6_rules_ops, flp, 0, &arg);
145 if (err < 0)
146 return err;
147 *mrt = res.mrt;
148 return 0;
149}
150
151static int ip6mr_rule_action(struct fib_rule *rule, struct flowi *flp,
152 int flags, struct fib_lookup_arg *arg)
153{
154 struct ip6mr_result *res = arg->result;
155 struct mr6_table *mrt;
156
157 switch (rule->action) {
158 case FR_ACT_TO_TBL:
159 break;
160 case FR_ACT_UNREACHABLE:
161 return -ENETUNREACH;
162 case FR_ACT_PROHIBIT:
163 return -EACCES;
164 case FR_ACT_BLACKHOLE:
165 default:
166 return -EINVAL;
167 }
168
169 mrt = ip6mr_get_table(rule->fr_net, rule->table);
170 if (mrt == NULL)
171 return -EAGAIN;
172 res->mrt = mrt;
173 return 0;
174}
175
176static int ip6mr_rule_match(struct fib_rule *rule, struct flowi *flp, int flags)
177{
178 return 1;
179}
180
181static const struct nla_policy ip6mr_rule_policy[FRA_MAX + 1] = {
182 FRA_GENERIC_POLICY,
183};
184
185static int ip6mr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
186 struct fib_rule_hdr *frh, struct nlattr **tb)
187{
188 return 0;
189}
190
191static int ip6mr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
192 struct nlattr **tb)
193{
194 return 1;
195}
196
197static int ip6mr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
198 struct fib_rule_hdr *frh)
199{
200 frh->dst_len = 0;
201 frh->src_len = 0;
202 frh->tos = 0;
203 return 0;
204}
205
206static const struct fib_rules_ops __net_initdata ip6mr_rules_ops_template = {
207 .family = RTNL_FAMILY_IP6MR,
208 .rule_size = sizeof(struct ip6mr_rule),
209 .addr_size = sizeof(struct in6_addr),
210 .action = ip6mr_rule_action,
211 .match = ip6mr_rule_match,
212 .configure = ip6mr_rule_configure,
213 .compare = ip6mr_rule_compare,
214 .default_pref = fib_default_rule_pref,
215 .fill = ip6mr_rule_fill,
216 .nlgroup = RTNLGRP_IPV6_RULE,
217 .policy = ip6mr_rule_policy,
218 .owner = THIS_MODULE,
219};
220
221static int __net_init ip6mr_rules_init(struct net *net)
222{
223 struct fib_rules_ops *ops;
224 struct mr6_table *mrt;
225 int err;
226
227 ops = fib_rules_register(&ip6mr_rules_ops_template, net);
228 if (IS_ERR(ops))
229 return PTR_ERR(ops);
230
231 INIT_LIST_HEAD(&net->ipv6.mr6_tables);
232
233 mrt = ip6mr_new_table(net, RT6_TABLE_DFLT);
234 if (mrt == NULL) {
235 err = -ENOMEM;
236 goto err1;
237 }
238
239 err = fib_default_rule_add(ops, 0x7fff, RT6_TABLE_DFLT, 0);
240 if (err < 0)
241 goto err2;
242
243 net->ipv6.mr6_rules_ops = ops;
244 return 0;
245
246err2:
247 kfree(mrt);
248err1:
249 fib_rules_unregister(ops);
250 return err;
251}
252
253static void __net_exit ip6mr_rules_exit(struct net *net)
254{
255 struct mr6_table *mrt, *next;
256
257 list_for_each_entry_safe(mrt, next, &net->ipv6.mr6_tables, list) {
258 list_del(&mrt->list);
259 ip6mr_free_table(mrt);
260 }
261 fib_rules_unregister(net->ipv6.mr6_rules_ops);
262}
263#else
264#define ip6mr_for_each_table(mrt, net) \
265 for (mrt = net->ipv6.mrt6; mrt; mrt = NULL)
266
267static struct mr6_table *ip6mr_get_table(struct net *net, u32 id)
268{
269 return net->ipv6.mrt6;
270}
271
272static int ip6mr_fib_lookup(struct net *net, struct flowi *flp,
273 struct mr6_table **mrt)
274{
275 *mrt = net->ipv6.mrt6;
276 return 0;
277}
278
279static int __net_init ip6mr_rules_init(struct net *net)
280{
281 net->ipv6.mrt6 = ip6mr_new_table(net, RT6_TABLE_DFLT);
282 return net->ipv6.mrt6 ? 0 : -ENOMEM;
283}
284
285static void __net_exit ip6mr_rules_exit(struct net *net)
286{
287 ip6mr_free_table(net->ipv6.mrt6);
288}
289#endif
290
291static struct mr6_table *ip6mr_new_table(struct net *net, u32 id)
292{
293 struct mr6_table *mrt;
294 unsigned int i;
295
296 mrt = ip6mr_get_table(net, id);
297 if (mrt != NULL)
298 return mrt;
299
300 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
301 if (mrt == NULL)
302 return NULL;
303 mrt->id = id;
304 write_pnet(&mrt->net, net);
305
306 /* Forwarding cache */
307 for (i = 0; i < MFC6_LINES; i++)
308 INIT_LIST_HEAD(&mrt->mfc6_cache_array[i]);
309
310 INIT_LIST_HEAD(&mrt->mfc6_unres_queue);
311
312 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
313 (unsigned long)mrt);
314
315#ifdef CONFIG_IPV6_PIMSM_V2
316 mrt->mroute_reg_vif_num = -1;
317#endif
318#ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
319 list_add_tail_rcu(&mrt->list, &net->ipv6.mr6_tables);
320#endif
321 return mrt;
322}
323
324static void ip6mr_free_table(struct mr6_table *mrt)
325{
326 del_timer(&mrt->ipmr_expire_timer);
327 mroute_clean_tables(mrt);
328 kfree(mrt);
329}
330
331#ifdef CONFIG_PROC_FS
332
333struct ipmr_mfc_iter {
334 struct seq_net_private p;
335 struct mr6_table *mrt;
336 struct list_head *cache;
337 int ct;
338};
339
340
341static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net,
342 struct ipmr_mfc_iter *it, loff_t pos)
343{
344 struct mr6_table *mrt = it->mrt;
345 struct mfc6_cache *mfc;
346
347 read_lock(&mrt_lock);
348 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++) {
349 it->cache = &mrt->mfc6_cache_array[it->ct];
350 list_for_each_entry(mfc, it->cache, list)
351 if (pos-- == 0)
352 return mfc;
353 }
354 read_unlock(&mrt_lock);
355
356 spin_lock_bh(&mfc_unres_lock);
357 it->cache = &mrt->mfc6_unres_queue;
358 list_for_each_entry(mfc, it->cache, list)
359 if (pos-- == 0)
360 return mfc;
361 spin_unlock_bh(&mfc_unres_lock);
362
363 it->cache = NULL;
364 return NULL;
365}
366
367/*
368 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
369 */
370
371struct ipmr_vif_iter {
372 struct seq_net_private p;
373 struct mr6_table *mrt;
374 int ct;
375};
376
377static struct mif_device *ip6mr_vif_seq_idx(struct net *net,
378 struct ipmr_vif_iter *iter,
379 loff_t pos)
380{
381 struct mr6_table *mrt = iter->mrt;
382
383 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
384 if (!MIF_EXISTS(mrt, iter->ct))
385 continue;
386 if (pos-- == 0)
387 return &mrt->vif6_table[iter->ct];
388 }
389 return NULL;
390}
391
392static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
393 __acquires(mrt_lock)
394{
395 struct ipmr_vif_iter *iter = seq->private;
396 struct net *net = seq_file_net(seq);
397 struct mr6_table *mrt;
398
399 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
400 if (mrt == NULL)
401 return ERR_PTR(-ENOENT);
402
403 iter->mrt = mrt;
404
405 read_lock(&mrt_lock);
406 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1)
407 : SEQ_START_TOKEN;
408}
409
410static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
411{
412 struct ipmr_vif_iter *iter = seq->private;
413 struct net *net = seq_file_net(seq);
414 struct mr6_table *mrt = iter->mrt;
415
416 ++*pos;
417 if (v == SEQ_START_TOKEN)
418 return ip6mr_vif_seq_idx(net, iter, 0);
419
420 while (++iter->ct < mrt->maxvif) {
421 if (!MIF_EXISTS(mrt, iter->ct))
422 continue;
423 return &mrt->vif6_table[iter->ct];
424 }
425 return NULL;
426}
427
428static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
429 __releases(mrt_lock)
430{
431 read_unlock(&mrt_lock);
432}
433
434static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
435{
436 struct ipmr_vif_iter *iter = seq->private;
437 struct mr6_table *mrt = iter->mrt;
438
439 if (v == SEQ_START_TOKEN) {
440 seq_puts(seq,
441 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
442 } else {
443 const struct mif_device *vif = v;
444 const char *name = vif->dev ? vif->dev->name : "none";
445
446 seq_printf(seq,
447 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
448 vif - mrt->vif6_table,
449 name, vif->bytes_in, vif->pkt_in,
450 vif->bytes_out, vif->pkt_out,
451 vif->flags);
452 }
453 return 0;
454}
455
456static const struct seq_operations ip6mr_vif_seq_ops = {
457 .start = ip6mr_vif_seq_start,
458 .next = ip6mr_vif_seq_next,
459 .stop = ip6mr_vif_seq_stop,
460 .show = ip6mr_vif_seq_show,
461};
462
463static int ip6mr_vif_open(struct inode *inode, struct file *file)
464{
465 return seq_open_net(inode, file, &ip6mr_vif_seq_ops,
466 sizeof(struct ipmr_vif_iter));
467}
468
469static const struct file_operations ip6mr_vif_fops = {
470 .owner = THIS_MODULE,
471 .open = ip6mr_vif_open,
472 .read = seq_read,
473 .llseek = seq_lseek,
474 .release = seq_release_net,
475};
476
477static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
478{
479 struct ipmr_mfc_iter *it = seq->private;
480 struct net *net = seq_file_net(seq);
481 struct mr6_table *mrt;
482
483 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
484 if (mrt == NULL)
485 return ERR_PTR(-ENOENT);
486
487 it->mrt = mrt;
488 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
489 : SEQ_START_TOKEN;
490}
491
492static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
493{
494 struct mfc6_cache *mfc = v;
495 struct ipmr_mfc_iter *it = seq->private;
496 struct net *net = seq_file_net(seq);
497 struct mr6_table *mrt = it->mrt;
498
499 ++*pos;
500
501 if (v == SEQ_START_TOKEN)
502 return ipmr_mfc_seq_idx(net, seq->private, 0);
503
504 if (mfc->list.next != it->cache)
505 return list_entry(mfc->list.next, struct mfc6_cache, list);
506
507 if (it->cache == &mrt->mfc6_unres_queue)
508 goto end_of_list;
509
510 BUG_ON(it->cache != &mrt->mfc6_cache_array[it->ct]);
511
512 while (++it->ct < MFC6_LINES) {
513 it->cache = &mrt->mfc6_cache_array[it->ct];
514 if (list_empty(it->cache))
515 continue;
516 return list_first_entry(it->cache, struct mfc6_cache, list);
517 }
518
519 /* exhausted cache_array, show unresolved */
520 read_unlock(&mrt_lock);
521 it->cache = &mrt->mfc6_unres_queue;
522 it->ct = 0;
523
524 spin_lock_bh(&mfc_unres_lock);
525 if (!list_empty(it->cache))
526 return list_first_entry(it->cache, struct mfc6_cache, list);
527
528 end_of_list:
529 spin_unlock_bh(&mfc_unres_lock);
530 it->cache = NULL;
531
532 return NULL;
533}
534
535static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
536{
537 struct ipmr_mfc_iter *it = seq->private;
538 struct mr6_table *mrt = it->mrt;
539
540 if (it->cache == &mrt->mfc6_unres_queue)
541 spin_unlock_bh(&mfc_unres_lock);
542 else if (it->cache == mrt->mfc6_cache_array)
543 read_unlock(&mrt_lock);
544}
545
546static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
547{
548 int n;
549
550 if (v == SEQ_START_TOKEN) {
551 seq_puts(seq,
552 "Group "
553 "Origin "
554 "Iif Pkts Bytes Wrong Oifs\n");
555 } else {
556 const struct mfc6_cache *mfc = v;
557 const struct ipmr_mfc_iter *it = seq->private;
558 struct mr6_table *mrt = it->mrt;
559
560 seq_printf(seq, "%pI6 %pI6 %-3hd",
561 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
562 mfc->mf6c_parent);
563
564 if (it->cache != &mrt->mfc6_unres_queue) {
565 seq_printf(seq, " %8lu %8lu %8lu",
566 mfc->mfc_un.res.pkt,
567 mfc->mfc_un.res.bytes,
568 mfc->mfc_un.res.wrong_if);
569 for (n = mfc->mfc_un.res.minvif;
570 n < mfc->mfc_un.res.maxvif; n++) {
571 if (MIF_EXISTS(mrt, n) &&
572 mfc->mfc_un.res.ttls[n] < 255)
573 seq_printf(seq,
574 " %2d:%-3d",
575 n, mfc->mfc_un.res.ttls[n]);
576 }
577 } else {
578 /* unresolved mfc_caches don't contain
579 * pkt, bytes and wrong_if values
580 */
581 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
582 }
583 seq_putc(seq, '\n');
584 }
585 return 0;
586}
587
588static const struct seq_operations ipmr_mfc_seq_ops = {
589 .start = ipmr_mfc_seq_start,
590 .next = ipmr_mfc_seq_next,
591 .stop = ipmr_mfc_seq_stop,
592 .show = ipmr_mfc_seq_show,
593};
594
595static int ipmr_mfc_open(struct inode *inode, struct file *file)
596{
597 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
598 sizeof(struct ipmr_mfc_iter));
599}
600
601static const struct file_operations ip6mr_mfc_fops = {
602 .owner = THIS_MODULE,
603 .open = ipmr_mfc_open,
604 .read = seq_read,
605 .llseek = seq_lseek,
606 .release = seq_release_net,
607};
608#endif
609
610#ifdef CONFIG_IPV6_PIMSM_V2
611
612static int pim6_rcv(struct sk_buff *skb)
613{
614 struct pimreghdr *pim;
615 struct ipv6hdr *encap;
616 struct net_device *reg_dev = NULL;
617 struct net *net = dev_net(skb->dev);
618 struct mr6_table *mrt;
619 struct flowi fl = {
620 .iif = skb->dev->ifindex,
621 .mark = skb->mark,
622 };
623 int reg_vif_num;
624
625 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
626 goto drop;
627
628 pim = (struct pimreghdr *)skb_transport_header(skb);
629 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
630 (pim->flags & PIM_NULL_REGISTER) ||
631 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
632 sizeof(*pim), IPPROTO_PIM,
633 csum_partial((void *)pim, sizeof(*pim), 0)) &&
634 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
635 goto drop;
636
637 /* check if the inner packet is destined to mcast group */
638 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
639 sizeof(*pim));
640
641 if (!ipv6_addr_is_multicast(&encap->daddr) ||
642 encap->payload_len == 0 ||
643 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
644 goto drop;
645
646 if (ip6mr_fib_lookup(net, &fl, &mrt) < 0)
647 goto drop;
648 reg_vif_num = mrt->mroute_reg_vif_num;
649
650 read_lock(&mrt_lock);
651 if (reg_vif_num >= 0)
652 reg_dev = mrt->vif6_table[reg_vif_num].dev;
653 if (reg_dev)
654 dev_hold(reg_dev);
655 read_unlock(&mrt_lock);
656
657 if (reg_dev == NULL)
658 goto drop;
659
660 skb->mac_header = skb->network_header;
661 skb_pull(skb, (u8 *)encap - skb->data);
662 skb_reset_network_header(skb);
663 skb->protocol = htons(ETH_P_IPV6);
664 skb->ip_summed = 0;
665 skb->pkt_type = PACKET_HOST;
666
667 skb_tunnel_rx(skb, reg_dev);
668
669 netif_rx(skb);
670 dev_put(reg_dev);
671 return 0;
672 drop:
673 kfree_skb(skb);
674 return 0;
675}
676
677static const struct inet6_protocol pim6_protocol = {
678 .handler = pim6_rcv,
679};
680
681/* Service routines creating virtual interfaces: PIMREG */
682
683static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
684 struct net_device *dev)
685{
686 struct net *net = dev_net(dev);
687 struct mr6_table *mrt;
688 struct flowi fl = {
689 .oif = dev->ifindex,
690 .iif = skb->skb_iif,
691 .mark = skb->mark,
692 };
693 int err;
694
695 err = ip6mr_fib_lookup(net, &fl, &mrt);
696 if (err < 0)
697 return err;
698
699 read_lock(&mrt_lock);
700 dev->stats.tx_bytes += skb->len;
701 dev->stats.tx_packets++;
702 ip6mr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, MRT6MSG_WHOLEPKT);
703 read_unlock(&mrt_lock);
704 kfree_skb(skb);
705 return NETDEV_TX_OK;
706}
707
708static const struct net_device_ops reg_vif_netdev_ops = {
709 .ndo_start_xmit = reg_vif_xmit,
710};
711
712static void reg_vif_setup(struct net_device *dev)
713{
714 dev->type = ARPHRD_PIMREG;
715 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
716 dev->flags = IFF_NOARP;
717 dev->netdev_ops = &reg_vif_netdev_ops;
718 dev->destructor = free_netdev;
719 dev->features |= NETIF_F_NETNS_LOCAL;
720}
721
722static struct net_device *ip6mr_reg_vif(struct net *net, struct mr6_table *mrt)
723{
724 struct net_device *dev;
725 char name[IFNAMSIZ];
726
727 if (mrt->id == RT6_TABLE_DFLT)
728 sprintf(name, "pim6reg");
729 else
730 sprintf(name, "pim6reg%u", mrt->id);
731
732 dev = alloc_netdev(0, name, reg_vif_setup);
733 if (dev == NULL)
734 return NULL;
735
736 dev_net_set(dev, net);
737
738 if (register_netdevice(dev)) {
739 free_netdev(dev);
740 return NULL;
741 }
742 dev->iflink = 0;
743
744 if (dev_open(dev))
745 goto failure;
746
747 dev_hold(dev);
748 return dev;
749
750failure:
751 /* allow the register to be completed before unregistering. */
752 rtnl_unlock();
753 rtnl_lock();
754
755 unregister_netdevice(dev);
756 return NULL;
757}
758#endif
759
760/*
761 * Delete a VIF entry
762 */
763
764static int mif6_delete(struct mr6_table *mrt, int vifi, struct list_head *head)
765{
766 struct mif_device *v;
767 struct net_device *dev;
768 struct inet6_dev *in6_dev;
769
770 if (vifi < 0 || vifi >= mrt->maxvif)
771 return -EADDRNOTAVAIL;
772
773 v = &mrt->vif6_table[vifi];
774
775 write_lock_bh(&mrt_lock);
776 dev = v->dev;
777 v->dev = NULL;
778
779 if (!dev) {
780 write_unlock_bh(&mrt_lock);
781 return -EADDRNOTAVAIL;
782 }
783
784#ifdef CONFIG_IPV6_PIMSM_V2
785 if (vifi == mrt->mroute_reg_vif_num)
786 mrt->mroute_reg_vif_num = -1;
787#endif
788
789 if (vifi + 1 == mrt->maxvif) {
790 int tmp;
791 for (tmp = vifi - 1; tmp >= 0; tmp--) {
792 if (MIF_EXISTS(mrt, tmp))
793 break;
794 }
795 mrt->maxvif = tmp + 1;
796 }
797
798 write_unlock_bh(&mrt_lock);
799
800 dev_set_allmulti(dev, -1);
801
802 in6_dev = __in6_dev_get(dev);
803 if (in6_dev)
804 in6_dev->cnf.mc_forwarding--;
805
806 if (v->flags & MIFF_REGISTER)
807 unregister_netdevice_queue(dev, head);
808
809 dev_put(dev);
810 return 0;
811}
812
813static inline void ip6mr_cache_free(struct mfc6_cache *c)
814{
815 kmem_cache_free(mrt_cachep, c);
816}
817
818/* Destroy an unresolved cache entry, killing queued skbs
819 and reporting error to netlink readers.
820 */
821
822static void ip6mr_destroy_unres(struct mr6_table *mrt, struct mfc6_cache *c)
823{
824 struct net *net = read_pnet(&mrt->net);
825 struct sk_buff *skb;
826
827 atomic_dec(&mrt->cache_resolve_queue_len);
828
829 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
830 if (ipv6_hdr(skb)->version == 0) {
831 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
832 nlh->nlmsg_type = NLMSG_ERROR;
833 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
834 skb_trim(skb, nlh->nlmsg_len);
835 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT;
836 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
837 } else
838 kfree_skb(skb);
839 }
840
841 ip6mr_cache_free(c);
842}
843
844
845/* Timer process for all the unresolved queue. */
846
847static void ipmr_do_expire_process(struct mr6_table *mrt)
848{
849 unsigned long now = jiffies;
850 unsigned long expires = 10 * HZ;
851 struct mfc6_cache *c, *next;
852
853 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
854 if (time_after(c->mfc_un.unres.expires, now)) {
855 /* not yet... */
856 unsigned long interval = c->mfc_un.unres.expires - now;
857 if (interval < expires)
858 expires = interval;
859 continue;
860 }
861
862 list_del(&c->list);
863 ip6mr_destroy_unres(mrt, c);
864 }
865
866 if (!list_empty(&mrt->mfc6_unres_queue))
867 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
868}
869
870static void ipmr_expire_process(unsigned long arg)
871{
872 struct mr6_table *mrt = (struct mr6_table *)arg;
873
874 if (!spin_trylock(&mfc_unres_lock)) {
875 mod_timer(&mrt->ipmr_expire_timer, jiffies + 1);
876 return;
877 }
878
879 if (!list_empty(&mrt->mfc6_unres_queue))
880 ipmr_do_expire_process(mrt);
881
882 spin_unlock(&mfc_unres_lock);
883}
884
885/* Fill oifs list. It is called under write locked mrt_lock. */
886
887static void ip6mr_update_thresholds(struct mr6_table *mrt, struct mfc6_cache *cache,
888 unsigned char *ttls)
889{
890 int vifi;
891
892 cache->mfc_un.res.minvif = MAXMIFS;
893 cache->mfc_un.res.maxvif = 0;
894 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
895
896 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
897 if (MIF_EXISTS(mrt, vifi) &&
898 ttls[vifi] && ttls[vifi] < 255) {
899 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
900 if (cache->mfc_un.res.minvif > vifi)
901 cache->mfc_un.res.minvif = vifi;
902 if (cache->mfc_un.res.maxvif <= vifi)
903 cache->mfc_un.res.maxvif = vifi + 1;
904 }
905 }
906}
907
908static int mif6_add(struct net *net, struct mr6_table *mrt,
909 struct mif6ctl *vifc, int mrtsock)
910{
911 int vifi = vifc->mif6c_mifi;
912 struct mif_device *v = &mrt->vif6_table[vifi];
913 struct net_device *dev;
914 struct inet6_dev *in6_dev;
915 int err;
916
917 /* Is vif busy ? */
918 if (MIF_EXISTS(mrt, vifi))
919 return -EADDRINUSE;
920
921 switch (vifc->mif6c_flags) {
922#ifdef CONFIG_IPV6_PIMSM_V2
923 case MIFF_REGISTER:
924 /*
925 * Special Purpose VIF in PIM
926 * All the packets will be sent to the daemon
927 */
928 if (mrt->mroute_reg_vif_num >= 0)
929 return -EADDRINUSE;
930 dev = ip6mr_reg_vif(net, mrt);
931 if (!dev)
932 return -ENOBUFS;
933 err = dev_set_allmulti(dev, 1);
934 if (err) {
935 unregister_netdevice(dev);
936 dev_put(dev);
937 return err;
938 }
939 break;
940#endif
941 case 0:
942 dev = dev_get_by_index(net, vifc->mif6c_pifi);
943 if (!dev)
944 return -EADDRNOTAVAIL;
945 err = dev_set_allmulti(dev, 1);
946 if (err) {
947 dev_put(dev);
948 return err;
949 }
950 break;
951 default:
952 return -EINVAL;
953 }
954
955 in6_dev = __in6_dev_get(dev);
956 if (in6_dev)
957 in6_dev->cnf.mc_forwarding++;
958
959 /*
960 * Fill in the VIF structures
961 */
962 v->rate_limit = vifc->vifc_rate_limit;
963 v->flags = vifc->mif6c_flags;
964 if (!mrtsock)
965 v->flags |= VIFF_STATIC;
966 v->threshold = vifc->vifc_threshold;
967 v->bytes_in = 0;
968 v->bytes_out = 0;
969 v->pkt_in = 0;
970 v->pkt_out = 0;
971 v->link = dev->ifindex;
972 if (v->flags & MIFF_REGISTER)
973 v->link = dev->iflink;
974
975 /* And finish update writing critical data */
976 write_lock_bh(&mrt_lock);
977 v->dev = dev;
978#ifdef CONFIG_IPV6_PIMSM_V2
979 if (v->flags & MIFF_REGISTER)
980 mrt->mroute_reg_vif_num = vifi;
981#endif
982 if (vifi + 1 > mrt->maxvif)
983 mrt->maxvif = vifi + 1;
984 write_unlock_bh(&mrt_lock);
985 return 0;
986}
987
988static struct mfc6_cache *ip6mr_cache_find(struct mr6_table *mrt,
989 struct in6_addr *origin,
990 struct in6_addr *mcastgrp)
991{
992 int line = MFC6_HASH(mcastgrp, origin);
993 struct mfc6_cache *c;
994
995 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
996 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
997 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
998 return c;
999 }
1000 return NULL;
1001}
1002
1003/*
1004 * Allocate a multicast cache entry
1005 */
1006static struct mfc6_cache *ip6mr_cache_alloc(void)
1007{
1008 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1009 if (c == NULL)
1010 return NULL;
1011 c->mfc_un.res.minvif = MAXMIFS;
1012 return c;
1013}
1014
1015static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
1016{
1017 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1018 if (c == NULL)
1019 return NULL;
1020 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1021 c->mfc_un.unres.expires = jiffies + 10 * HZ;
1022 return c;
1023}
1024
1025/*
1026 * A cache entry has gone into a resolved state from queued
1027 */
1028
1029static void ip6mr_cache_resolve(struct net *net, struct mr6_table *mrt,
1030 struct mfc6_cache *uc, struct mfc6_cache *c)
1031{
1032 struct sk_buff *skb;
1033
1034 /*
1035 * Play the pending entries through our router
1036 */
1037
1038 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1039 if (ipv6_hdr(skb)->version == 0) {
1040 int err;
1041 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
1042
1043 if (__ip6mr_fill_mroute(mrt, skb, c, NLMSG_DATA(nlh)) > 0) {
1044 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
1045 } else {
1046 nlh->nlmsg_type = NLMSG_ERROR;
1047 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
1048 skb_trim(skb, nlh->nlmsg_len);
1049 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE;
1050 }
1051 err = rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
1052 } else
1053 ip6_mr_forward(net, mrt, skb, c);
1054 }
1055}
1056
1057/*
1058 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
1059 * expects the following bizarre scheme.
1060 *
1061 * Called under mrt_lock.
1062 */
1063
1064static int ip6mr_cache_report(struct mr6_table *mrt, struct sk_buff *pkt,
1065 mifi_t mifi, int assert)
1066{
1067 struct sk_buff *skb;
1068 struct mrt6msg *msg;
1069 int ret;
1070
1071#ifdef CONFIG_IPV6_PIMSM_V2
1072 if (assert == MRT6MSG_WHOLEPKT)
1073 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
1074 +sizeof(*msg));
1075 else
1076#endif
1077 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
1078
1079 if (!skb)
1080 return -ENOBUFS;
1081
1082 /* I suppose that internal messages
1083 * do not require checksums */
1084
1085 skb->ip_summed = CHECKSUM_UNNECESSARY;
1086
1087#ifdef CONFIG_IPV6_PIMSM_V2
1088 if (assert == MRT6MSG_WHOLEPKT) {
1089 /* Ugly, but we have no choice with this interface.
1090 Duplicate old header, fix length etc.
1091 And all this only to mangle msg->im6_msgtype and
1092 to set msg->im6_mbz to "mbz" :-)
1093 */
1094 skb_push(skb, -skb_network_offset(pkt));
1095
1096 skb_push(skb, sizeof(*msg));
1097 skb_reset_transport_header(skb);
1098 msg = (struct mrt6msg *)skb_transport_header(skb);
1099 msg->im6_mbz = 0;
1100 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
1101 msg->im6_mif = mrt->mroute_reg_vif_num;
1102 msg->im6_pad = 0;
1103 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
1104 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
1105
1106 skb->ip_summed = CHECKSUM_UNNECESSARY;
1107 } else
1108#endif
1109 {
1110 /*
1111 * Copy the IP header
1112 */
1113
1114 skb_put(skb, sizeof(struct ipv6hdr));
1115 skb_reset_network_header(skb);
1116 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
1117
1118 /*
1119 * Add our header
1120 */
1121 skb_put(skb, sizeof(*msg));
1122 skb_reset_transport_header(skb);
1123 msg = (struct mrt6msg *)skb_transport_header(skb);
1124
1125 msg->im6_mbz = 0;
1126 msg->im6_msgtype = assert;
1127 msg->im6_mif = mifi;
1128 msg->im6_pad = 0;
1129 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
1130 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
1131
1132 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1133 skb->ip_summed = CHECKSUM_UNNECESSARY;
1134 }
1135
1136 if (mrt->mroute6_sk == NULL) {
1137 kfree_skb(skb);
1138 return -EINVAL;
1139 }
1140
1141 /*
1142 * Deliver to user space multicast routing algorithms
1143 */
1144 ret = sock_queue_rcv_skb(mrt->mroute6_sk, skb);
1145 if (ret < 0) {
1146 if (net_ratelimit())
1147 printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n");
1148 kfree_skb(skb);
1149 }
1150
1151 return ret;
1152}
1153
1154/*
1155 * Queue a packet for resolution. It gets locked cache entry!
1156 */
1157
1158static int
1159ip6mr_cache_unresolved(struct mr6_table *mrt, mifi_t mifi, struct sk_buff *skb)
1160{
1161 bool found = false;
1162 int err;
1163 struct mfc6_cache *c;
1164
1165 spin_lock_bh(&mfc_unres_lock);
1166 list_for_each_entry(c, &mrt->mfc6_unres_queue, list) {
1167 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
1168 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr)) {
1169 found = true;
1170 break;
1171 }
1172 }
1173
1174 if (!found) {
1175 /*
1176 * Create a new entry if allowable
1177 */
1178
1179 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1180 (c = ip6mr_cache_alloc_unres()) == NULL) {
1181 spin_unlock_bh(&mfc_unres_lock);
1182
1183 kfree_skb(skb);
1184 return -ENOBUFS;
1185 }
1186
1187 /*
1188 * Fill in the new cache entry
1189 */
1190 c->mf6c_parent = -1;
1191 c->mf6c_origin = ipv6_hdr(skb)->saddr;
1192 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
1193
1194 /*
1195 * Reflect first query at pim6sd
1196 */
1197 err = ip6mr_cache_report(mrt, skb, mifi, MRT6MSG_NOCACHE);
1198 if (err < 0) {
1199 /* If the report failed throw the cache entry
1200 out - Brad Parker
1201 */
1202 spin_unlock_bh(&mfc_unres_lock);
1203
1204 ip6mr_cache_free(c);
1205 kfree_skb(skb);
1206 return err;
1207 }
1208
1209 atomic_inc(&mrt->cache_resolve_queue_len);
1210 list_add(&c->list, &mrt->mfc6_unres_queue);
1211
1212 ipmr_do_expire_process(mrt);
1213 }
1214
1215 /*
1216 * See if we can append the packet
1217 */
1218 if (c->mfc_un.unres.unresolved.qlen > 3) {
1219 kfree_skb(skb);
1220 err = -ENOBUFS;
1221 } else {
1222 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1223 err = 0;
1224 }
1225
1226 spin_unlock_bh(&mfc_unres_lock);
1227 return err;
1228}
1229
1230/*
1231 * MFC6 cache manipulation by user space
1232 */
1233
1234static int ip6mr_mfc_delete(struct mr6_table *mrt, struct mf6cctl *mfc)
1235{
1236 int line;
1237 struct mfc6_cache *c, *next;
1238
1239 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1240
1241 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[line], list) {
1242 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1243 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
1244 write_lock_bh(&mrt_lock);
1245 list_del(&c->list);
1246 write_unlock_bh(&mrt_lock);
1247
1248 ip6mr_cache_free(c);
1249 return 0;
1250 }
1251 }
1252 return -ENOENT;
1253}
1254
1255static int ip6mr_device_event(struct notifier_block *this,
1256 unsigned long event, void *ptr)
1257{
1258 struct net_device *dev = ptr;
1259 struct net *net = dev_net(dev);
1260 struct mr6_table *mrt;
1261 struct mif_device *v;
1262 int ct;
1263 LIST_HEAD(list);
1264
1265 if (event != NETDEV_UNREGISTER)
1266 return NOTIFY_DONE;
1267
1268 ip6mr_for_each_table(mrt, net) {
1269 v = &mrt->vif6_table[0];
1270 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1271 if (v->dev == dev)
1272 mif6_delete(mrt, ct, &list);
1273 }
1274 }
1275 unregister_netdevice_many(&list);
1276
1277 return NOTIFY_DONE;
1278}
1279
1280static struct notifier_block ip6_mr_notifier = {
1281 .notifier_call = ip6mr_device_event
1282};
1283
1284/*
1285 * Setup for IP multicast routing
1286 */
1287
1288static int __net_init ip6mr_net_init(struct net *net)
1289{
1290 int err;
1291
1292 err = ip6mr_rules_init(net);
1293 if (err < 0)
1294 goto fail;
1295
1296#ifdef CONFIG_PROC_FS
1297 err = -ENOMEM;
1298 if (!proc_net_fops_create(net, "ip6_mr_vif", 0, &ip6mr_vif_fops))
1299 goto proc_vif_fail;
1300 if (!proc_net_fops_create(net, "ip6_mr_cache", 0, &ip6mr_mfc_fops))
1301 goto proc_cache_fail;
1302#endif
1303
1304 return 0;
1305
1306#ifdef CONFIG_PROC_FS
1307proc_cache_fail:
1308 proc_net_remove(net, "ip6_mr_vif");
1309proc_vif_fail:
1310 ip6mr_rules_exit(net);
1311#endif
1312fail:
1313 return err;
1314}
1315
1316static void __net_exit ip6mr_net_exit(struct net *net)
1317{
1318#ifdef CONFIG_PROC_FS
1319 proc_net_remove(net, "ip6_mr_cache");
1320 proc_net_remove(net, "ip6_mr_vif");
1321#endif
1322 ip6mr_rules_exit(net);
1323}
1324
1325static struct pernet_operations ip6mr_net_ops = {
1326 .init = ip6mr_net_init,
1327 .exit = ip6mr_net_exit,
1328};
1329
1330int __init ip6_mr_init(void)
1331{
1332 int err;
1333
1334 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1335 sizeof(struct mfc6_cache),
1336 0, SLAB_HWCACHE_ALIGN,
1337 NULL);
1338 if (!mrt_cachep)
1339 return -ENOMEM;
1340
1341 err = register_pernet_subsys(&ip6mr_net_ops);
1342 if (err)
1343 goto reg_pernet_fail;
1344
1345 err = register_netdevice_notifier(&ip6_mr_notifier);
1346 if (err)
1347 goto reg_notif_fail;
1348#ifdef CONFIG_IPV6_PIMSM_V2
1349 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1350 printk(KERN_ERR "ip6_mr_init: can't add PIM protocol\n");
1351 err = -EAGAIN;
1352 goto add_proto_fail;
1353 }
1354#endif
1355 rtnl_register(RTNL_FAMILY_IP6MR, RTM_GETROUTE, NULL, ip6mr_rtm_dumproute);
1356 return 0;
1357#ifdef CONFIG_IPV6_PIMSM_V2
1358add_proto_fail:
1359 unregister_netdevice_notifier(&ip6_mr_notifier);
1360#endif
1361reg_notif_fail:
1362 unregister_pernet_subsys(&ip6mr_net_ops);
1363reg_pernet_fail:
1364 kmem_cache_destroy(mrt_cachep);
1365 return err;
1366}
1367
1368void ip6_mr_cleanup(void)
1369{
1370 unregister_netdevice_notifier(&ip6_mr_notifier);
1371 unregister_pernet_subsys(&ip6mr_net_ops);
1372 kmem_cache_destroy(mrt_cachep);
1373}
1374
1375static int ip6mr_mfc_add(struct net *net, struct mr6_table *mrt,
1376 struct mf6cctl *mfc, int mrtsock)
1377{
1378 bool found = false;
1379 int line;
1380 struct mfc6_cache *uc, *c;
1381 unsigned char ttls[MAXMIFS];
1382 int i;
1383
1384 if (mfc->mf6cc_parent >= MAXMIFS)
1385 return -ENFILE;
1386
1387 memset(ttls, 255, MAXMIFS);
1388 for (i = 0; i < MAXMIFS; i++) {
1389 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1390 ttls[i] = 1;
1391
1392 }
1393
1394 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1395
1396 list_for_each_entry(c, &mrt->mfc6_cache_array[line], list) {
1397 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1398 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
1399 found = true;
1400 break;
1401 }
1402 }
1403
1404 if (found) {
1405 write_lock_bh(&mrt_lock);
1406 c->mf6c_parent = mfc->mf6cc_parent;
1407 ip6mr_update_thresholds(mrt, c, ttls);
1408 if (!mrtsock)
1409 c->mfc_flags |= MFC_STATIC;
1410 write_unlock_bh(&mrt_lock);
1411 return 0;
1412 }
1413
1414 if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1415 return -EINVAL;
1416
1417 c = ip6mr_cache_alloc();
1418 if (c == NULL)
1419 return -ENOMEM;
1420
1421 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1422 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1423 c->mf6c_parent = mfc->mf6cc_parent;
1424 ip6mr_update_thresholds(mrt, c, ttls);
1425 if (!mrtsock)
1426 c->mfc_flags |= MFC_STATIC;
1427
1428 write_lock_bh(&mrt_lock);
1429 list_add(&c->list, &mrt->mfc6_cache_array[line]);
1430 write_unlock_bh(&mrt_lock);
1431
1432 /*
1433 * Check to see if we resolved a queued list. If so we
1434 * need to send on the frames and tidy up.
1435 */
1436 found = false;
1437 spin_lock_bh(&mfc_unres_lock);
1438 list_for_each_entry(uc, &mrt->mfc6_unres_queue, list) {
1439 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1440 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1441 list_del(&uc->list);
1442 atomic_dec(&mrt->cache_resolve_queue_len);
1443 found = true;
1444 break;
1445 }
1446 }
1447 if (list_empty(&mrt->mfc6_unres_queue))
1448 del_timer(&mrt->ipmr_expire_timer);
1449 spin_unlock_bh(&mfc_unres_lock);
1450
1451 if (found) {
1452 ip6mr_cache_resolve(net, mrt, uc, c);
1453 ip6mr_cache_free(uc);
1454 }
1455 return 0;
1456}
1457
1458/*
1459 * Close the multicast socket, and clear the vif tables etc
1460 */
1461
1462static void mroute_clean_tables(struct mr6_table *mrt)
1463{
1464 int i;
1465 LIST_HEAD(list);
1466 struct mfc6_cache *c, *next;
1467
1468 /*
1469 * Shut down all active vif entries
1470 */
1471 for (i = 0; i < mrt->maxvif; i++) {
1472 if (!(mrt->vif6_table[i].flags & VIFF_STATIC))
1473 mif6_delete(mrt, i, &list);
1474 }
1475 unregister_netdevice_many(&list);
1476
1477 /*
1478 * Wipe the cache
1479 */
1480 for (i = 0; i < MFC6_LINES; i++) {
1481 list_for_each_entry_safe(c, next, &mrt->mfc6_cache_array[i], list) {
1482 if (c->mfc_flags & MFC_STATIC)
1483 continue;
1484 write_lock_bh(&mrt_lock);
1485 list_del(&c->list);
1486 write_unlock_bh(&mrt_lock);
1487
1488 ip6mr_cache_free(c);
1489 }
1490 }
1491
1492 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1493 spin_lock_bh(&mfc_unres_lock);
1494 list_for_each_entry_safe(c, next, &mrt->mfc6_unres_queue, list) {
1495 list_del(&c->list);
1496 ip6mr_destroy_unres(mrt, c);
1497 }
1498 spin_unlock_bh(&mfc_unres_lock);
1499 }
1500}
1501
1502static int ip6mr_sk_init(struct mr6_table *mrt, struct sock *sk)
1503{
1504 int err = 0;
1505 struct net *net = sock_net(sk);
1506
1507 rtnl_lock();
1508 write_lock_bh(&mrt_lock);
1509 if (likely(mrt->mroute6_sk == NULL)) {
1510 mrt->mroute6_sk = sk;
1511 net->ipv6.devconf_all->mc_forwarding++;
1512 }
1513 else
1514 err = -EADDRINUSE;
1515 write_unlock_bh(&mrt_lock);
1516
1517 rtnl_unlock();
1518
1519 return err;
1520}
1521
1522int ip6mr_sk_done(struct sock *sk)
1523{
1524 int err = -EACCES;
1525 struct net *net = sock_net(sk);
1526 struct mr6_table *mrt;
1527
1528 rtnl_lock();
1529 ip6mr_for_each_table(mrt, net) {
1530 if (sk == mrt->mroute6_sk) {
1531 write_lock_bh(&mrt_lock);
1532 mrt->mroute6_sk = NULL;
1533 net->ipv6.devconf_all->mc_forwarding--;
1534 write_unlock_bh(&mrt_lock);
1535
1536 mroute_clean_tables(mrt);
1537 err = 0;
1538 break;
1539 }
1540 }
1541 rtnl_unlock();
1542
1543 return err;
1544}
1545
1546struct sock *mroute6_socket(struct net *net, struct sk_buff *skb)
1547{
1548 struct mr6_table *mrt;
1549 struct flowi fl = {
1550 .iif = skb->skb_iif,
1551 .oif = skb->dev->ifindex,
1552 .mark = skb->mark,
1553 };
1554
1555 if (ip6mr_fib_lookup(net, &fl, &mrt) < 0)
1556 return NULL;
1557
1558 return mrt->mroute6_sk;
1559}
1560
1561/*
1562 * Socket options and virtual interface manipulation. The whole
1563 * virtual interface system is a complete heap, but unfortunately
1564 * that's how BSD mrouted happens to think. Maybe one day with a proper
1565 * MOSPF/PIM router set up we can clean this up.
1566 */
1567
1568int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1569{
1570 int ret;
1571 struct mif6ctl vif;
1572 struct mf6cctl mfc;
1573 mifi_t mifi;
1574 struct net *net = sock_net(sk);
1575 struct mr6_table *mrt;
1576
1577 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1578 if (mrt == NULL)
1579 return -ENOENT;
1580
1581 if (optname != MRT6_INIT) {
1582 if (sk != mrt->mroute6_sk && !capable(CAP_NET_ADMIN))
1583 return -EACCES;
1584 }
1585
1586 switch (optname) {
1587 case MRT6_INIT:
1588 if (sk->sk_type != SOCK_RAW ||
1589 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1590 return -EOPNOTSUPP;
1591 if (optlen < sizeof(int))
1592 return -EINVAL;
1593
1594 return ip6mr_sk_init(mrt, sk);
1595
1596 case MRT6_DONE:
1597 return ip6mr_sk_done(sk);
1598
1599 case MRT6_ADD_MIF:
1600 if (optlen < sizeof(vif))
1601 return -EINVAL;
1602 if (copy_from_user(&vif, optval, sizeof(vif)))
1603 return -EFAULT;
1604 if (vif.mif6c_mifi >= MAXMIFS)
1605 return -ENFILE;
1606 rtnl_lock();
1607 ret = mif6_add(net, mrt, &vif, sk == mrt->mroute6_sk);
1608 rtnl_unlock();
1609 return ret;
1610
1611 case MRT6_DEL_MIF:
1612 if (optlen < sizeof(mifi_t))
1613 return -EINVAL;
1614 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1615 return -EFAULT;
1616 rtnl_lock();
1617 ret = mif6_delete(mrt, mifi, NULL);
1618 rtnl_unlock();
1619 return ret;
1620
1621 /*
1622 * Manipulate the forwarding caches. These live
1623 * in a sort of kernel/user symbiosis.
1624 */
1625 case MRT6_ADD_MFC:
1626 case MRT6_DEL_MFC:
1627 if (optlen < sizeof(mfc))
1628 return -EINVAL;
1629 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1630 return -EFAULT;
1631 rtnl_lock();
1632 if (optname == MRT6_DEL_MFC)
1633 ret = ip6mr_mfc_delete(mrt, &mfc);
1634 else
1635 ret = ip6mr_mfc_add(net, mrt, &mfc, sk == mrt->mroute6_sk);
1636 rtnl_unlock();
1637 return ret;
1638
1639 /*
1640 * Control PIM assert (to activate pim will activate assert)
1641 */
1642 case MRT6_ASSERT:
1643 {
1644 int v;
1645 if (get_user(v, (int __user *)optval))
1646 return -EFAULT;
1647 mrt->mroute_do_assert = !!v;
1648 return 0;
1649 }
1650
1651#ifdef CONFIG_IPV6_PIMSM_V2
1652 case MRT6_PIM:
1653 {
1654 int v;
1655 if (get_user(v, (int __user *)optval))
1656 return -EFAULT;
1657 v = !!v;
1658 rtnl_lock();
1659 ret = 0;
1660 if (v != mrt->mroute_do_pim) {
1661 mrt->mroute_do_pim = v;
1662 mrt->mroute_do_assert = v;
1663 }
1664 rtnl_unlock();
1665 return ret;
1666 }
1667
1668#endif
1669#ifdef CONFIG_IPV6_MROUTE_MULTIPLE_TABLES
1670 case MRT6_TABLE:
1671 {
1672 u32 v;
1673
1674 if (optlen != sizeof(u32))
1675 return -EINVAL;
1676 if (get_user(v, (u32 __user *)optval))
1677 return -EFAULT;
1678 if (sk == mrt->mroute6_sk)
1679 return -EBUSY;
1680
1681 rtnl_lock();
1682 ret = 0;
1683 if (!ip6mr_new_table(net, v))
1684 ret = -ENOMEM;
1685 raw6_sk(sk)->ip6mr_table = v;
1686 rtnl_unlock();
1687 return ret;
1688 }
1689#endif
1690 /*
1691 * Spurious command, or MRT6_VERSION which you cannot
1692 * set.
1693 */
1694 default:
1695 return -ENOPROTOOPT;
1696 }
1697}
1698
1699/*
1700 * Getsock opt support for the multicast routing system.
1701 */
1702
1703int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1704 int __user *optlen)
1705{
1706 int olr;
1707 int val;
1708 struct net *net = sock_net(sk);
1709 struct mr6_table *mrt;
1710
1711 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1712 if (mrt == NULL)
1713 return -ENOENT;
1714
1715 switch (optname) {
1716 case MRT6_VERSION:
1717 val = 0x0305;
1718 break;
1719#ifdef CONFIG_IPV6_PIMSM_V2
1720 case MRT6_PIM:
1721 val = mrt->mroute_do_pim;
1722 break;
1723#endif
1724 case MRT6_ASSERT:
1725 val = mrt->mroute_do_assert;
1726 break;
1727 default:
1728 return -ENOPROTOOPT;
1729 }
1730
1731 if (get_user(olr, optlen))
1732 return -EFAULT;
1733
1734 olr = min_t(int, olr, sizeof(int));
1735 if (olr < 0)
1736 return -EINVAL;
1737
1738 if (put_user(olr, optlen))
1739 return -EFAULT;
1740 if (copy_to_user(optval, &val, olr))
1741 return -EFAULT;
1742 return 0;
1743}
1744
1745/*
1746 * The IP multicast ioctl support routines.
1747 */
1748
1749int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1750{
1751 struct sioc_sg_req6 sr;
1752 struct sioc_mif_req6 vr;
1753 struct mif_device *vif;
1754 struct mfc6_cache *c;
1755 struct net *net = sock_net(sk);
1756 struct mr6_table *mrt;
1757
1758 mrt = ip6mr_get_table(net, raw6_sk(sk)->ip6mr_table ? : RT6_TABLE_DFLT);
1759 if (mrt == NULL)
1760 return -ENOENT;
1761
1762 switch (cmd) {
1763 case SIOCGETMIFCNT_IN6:
1764 if (copy_from_user(&vr, arg, sizeof(vr)))
1765 return -EFAULT;
1766 if (vr.mifi >= mrt->maxvif)
1767 return -EINVAL;
1768 read_lock(&mrt_lock);
1769 vif = &mrt->vif6_table[vr.mifi];
1770 if (MIF_EXISTS(mrt, vr.mifi)) {
1771 vr.icount = vif->pkt_in;
1772 vr.ocount = vif->pkt_out;
1773 vr.ibytes = vif->bytes_in;
1774 vr.obytes = vif->bytes_out;
1775 read_unlock(&mrt_lock);
1776
1777 if (copy_to_user(arg, &vr, sizeof(vr)))
1778 return -EFAULT;
1779 return 0;
1780 }
1781 read_unlock(&mrt_lock);
1782 return -EADDRNOTAVAIL;
1783 case SIOCGETSGCNT_IN6:
1784 if (copy_from_user(&sr, arg, sizeof(sr)))
1785 return -EFAULT;
1786
1787 read_lock(&mrt_lock);
1788 c = ip6mr_cache_find(mrt, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1789 if (c) {
1790 sr.pktcnt = c->mfc_un.res.pkt;
1791 sr.bytecnt = c->mfc_un.res.bytes;
1792 sr.wrong_if = c->mfc_un.res.wrong_if;
1793 read_unlock(&mrt_lock);
1794
1795 if (copy_to_user(arg, &sr, sizeof(sr)))
1796 return -EFAULT;
1797 return 0;
1798 }
1799 read_unlock(&mrt_lock);
1800 return -EADDRNOTAVAIL;
1801 default:
1802 return -ENOIOCTLCMD;
1803 }
1804}
1805
1806
1807static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1808{
1809 IP6_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1810 IPSTATS_MIB_OUTFORWDATAGRAMS);
1811 return dst_output(skb);
1812}
1813
1814/*
1815 * Processing handlers for ip6mr_forward
1816 */
1817
1818static int ip6mr_forward2(struct net *net, struct mr6_table *mrt,
1819 struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1820{
1821 struct ipv6hdr *ipv6h;
1822 struct mif_device *vif = &mrt->vif6_table[vifi];
1823 struct net_device *dev;
1824 struct dst_entry *dst;
1825 struct flowi fl;
1826
1827 if (vif->dev == NULL)
1828 goto out_free;
1829
1830#ifdef CONFIG_IPV6_PIMSM_V2
1831 if (vif->flags & MIFF_REGISTER) {
1832 vif->pkt_out++;
1833 vif->bytes_out += skb->len;
1834 vif->dev->stats.tx_bytes += skb->len;
1835 vif->dev->stats.tx_packets++;
1836 ip6mr_cache_report(mrt, skb, vifi, MRT6MSG_WHOLEPKT);
1837 goto out_free;
1838 }
1839#endif
1840
1841 ipv6h = ipv6_hdr(skb);
1842
1843 fl = (struct flowi) {
1844 .oif = vif->link,
1845 .nl_u = { .ip6_u =
1846 { .daddr = ipv6h->daddr, }
1847 }
1848 };
1849
1850 dst = ip6_route_output(net, NULL, &fl);
1851 if (!dst)
1852 goto out_free;
1853
1854 skb_dst_drop(skb);
1855 skb_dst_set(skb, dst);
1856
1857 /*
1858 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1859 * not only before forwarding, but after forwarding on all output
1860 * interfaces. It is clear, if mrouter runs a multicasting
1861 * program, it should receive packets not depending to what interface
1862 * program is joined.
1863 * If we will not make it, the program will have to join on all
1864 * interfaces. On the other hand, multihoming host (or router, but
1865 * not mrouter) cannot join to more than one interface - it will
1866 * result in receiving multiple packets.
1867 */
1868 dev = vif->dev;
1869 skb->dev = dev;
1870 vif->pkt_out++;
1871 vif->bytes_out += skb->len;
1872
1873 /* We are about to write */
1874 /* XXX: extension headers? */
1875 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
1876 goto out_free;
1877
1878 ipv6h = ipv6_hdr(skb);
1879 ipv6h->hop_limit--;
1880
1881 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
1882
1883 return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dev,
1884 ip6mr_forward2_finish);
1885
1886out_free:
1887 kfree_skb(skb);
1888 return 0;
1889}
1890
1891static int ip6mr_find_vif(struct mr6_table *mrt, struct net_device *dev)
1892{
1893 int ct;
1894
1895 for (ct = mrt->maxvif - 1; ct >= 0; ct--) {
1896 if (mrt->vif6_table[ct].dev == dev)
1897 break;
1898 }
1899 return ct;
1900}
1901
1902static int ip6_mr_forward(struct net *net, struct mr6_table *mrt,
1903 struct sk_buff *skb, struct mfc6_cache *cache)
1904{
1905 int psend = -1;
1906 int vif, ct;
1907
1908 vif = cache->mf6c_parent;
1909 cache->mfc_un.res.pkt++;
1910 cache->mfc_un.res.bytes += skb->len;
1911
1912 /*
1913 * Wrong interface: drop packet and (maybe) send PIM assert.
1914 */
1915 if (mrt->vif6_table[vif].dev != skb->dev) {
1916 int true_vifi;
1917
1918 cache->mfc_un.res.wrong_if++;
1919 true_vifi = ip6mr_find_vif(mrt, skb->dev);
1920
1921 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1922 /* pimsm uses asserts, when switching from RPT to SPT,
1923 so that we cannot check that packet arrived on an oif.
1924 It is bad, but otherwise we would need to move pretty
1925 large chunk of pimd to kernel. Ough... --ANK
1926 */
1927 (mrt->mroute_do_pim ||
1928 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1929 time_after(jiffies,
1930 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1931 cache->mfc_un.res.last_assert = jiffies;
1932 ip6mr_cache_report(mrt, skb, true_vifi, MRT6MSG_WRONGMIF);
1933 }
1934 goto dont_forward;
1935 }
1936
1937 mrt->vif6_table[vif].pkt_in++;
1938 mrt->vif6_table[vif].bytes_in += skb->len;
1939
1940 /*
1941 * Forward the frame
1942 */
1943 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
1944 if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
1945 if (psend != -1) {
1946 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1947 if (skb2)
1948 ip6mr_forward2(net, mrt, skb2, cache, psend);
1949 }
1950 psend = ct;
1951 }
1952 }
1953 if (psend != -1) {
1954 ip6mr_forward2(net, mrt, skb, cache, psend);
1955 return 0;
1956 }
1957
1958dont_forward:
1959 kfree_skb(skb);
1960 return 0;
1961}
1962
1963
1964/*
1965 * Multicast packets for forwarding arrive here
1966 */
1967
1968int ip6_mr_input(struct sk_buff *skb)
1969{
1970 struct mfc6_cache *cache;
1971 struct net *net = dev_net(skb->dev);
1972 struct mr6_table *mrt;
1973 struct flowi fl = {
1974 .iif = skb->dev->ifindex,
1975 .mark = skb->mark,
1976 };
1977 int err;
1978
1979 err = ip6mr_fib_lookup(net, &fl, &mrt);
1980 if (err < 0)
1981 return err;
1982
1983 read_lock(&mrt_lock);
1984 cache = ip6mr_cache_find(mrt,
1985 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
1986
1987 /*
1988 * No usable cache entry
1989 */
1990 if (cache == NULL) {
1991 int vif;
1992
1993 vif = ip6mr_find_vif(mrt, skb->dev);
1994 if (vif >= 0) {
1995 int err = ip6mr_cache_unresolved(mrt, vif, skb);
1996 read_unlock(&mrt_lock);
1997
1998 return err;
1999 }
2000 read_unlock(&mrt_lock);
2001 kfree_skb(skb);
2002 return -ENODEV;
2003 }
2004
2005 ip6_mr_forward(net, mrt, skb, cache);
2006
2007 read_unlock(&mrt_lock);
2008
2009 return 0;
2010}
2011
2012
2013static int __ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2014 struct mfc6_cache *c, struct rtmsg *rtm)
2015{
2016 int ct;
2017 struct rtnexthop *nhp;
2018 u8 *b = skb_tail_pointer(skb);
2019 struct rtattr *mp_head;
2020
2021 /* If cache is unresolved, don't try to parse IIF and OIF */
2022 if (c->mf6c_parent >= MAXMIFS)
2023 return -ENOENT;
2024
2025 if (MIF_EXISTS(mrt, c->mf6c_parent))
2026 RTA_PUT(skb, RTA_IIF, 4, &mrt->vif6_table[c->mf6c_parent].dev->ifindex);
2027
2028 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
2029
2030 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2031 if (MIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2032 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
2033 goto rtattr_failure;
2034 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
2035 nhp->rtnh_flags = 0;
2036 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2037 nhp->rtnh_ifindex = mrt->vif6_table[ct].dev->ifindex;
2038 nhp->rtnh_len = sizeof(*nhp);
2039 }
2040 }
2041 mp_head->rta_type = RTA_MULTIPATH;
2042 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
2043 rtm->rtm_type = RTN_MULTICAST;
2044 return 1;
2045
2046rtattr_failure:
2047 nlmsg_trim(skb, b);
2048 return -EMSGSIZE;
2049}
2050
2051int ip6mr_get_route(struct net *net,
2052 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
2053{
2054 int err;
2055 struct mr6_table *mrt;
2056 struct mfc6_cache *cache;
2057 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
2058
2059 mrt = ip6mr_get_table(net, RT6_TABLE_DFLT);
2060 if (mrt == NULL)
2061 return -ENOENT;
2062
2063 read_lock(&mrt_lock);
2064 cache = ip6mr_cache_find(mrt, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
2065
2066 if (!cache) {
2067 struct sk_buff *skb2;
2068 struct ipv6hdr *iph;
2069 struct net_device *dev;
2070 int vif;
2071
2072 if (nowait) {
2073 read_unlock(&mrt_lock);
2074 return -EAGAIN;
2075 }
2076
2077 dev = skb->dev;
2078 if (dev == NULL || (vif = ip6mr_find_vif(mrt, dev)) < 0) {
2079 read_unlock(&mrt_lock);
2080 return -ENODEV;
2081 }
2082
2083 /* really correct? */
2084 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
2085 if (!skb2) {
2086 read_unlock(&mrt_lock);
2087 return -ENOMEM;
2088 }
2089
2090 skb_reset_transport_header(skb2);
2091
2092 skb_put(skb2, sizeof(struct ipv6hdr));
2093 skb_reset_network_header(skb2);
2094
2095 iph = ipv6_hdr(skb2);
2096 iph->version = 0;
2097 iph->priority = 0;
2098 iph->flow_lbl[0] = 0;
2099 iph->flow_lbl[1] = 0;
2100 iph->flow_lbl[2] = 0;
2101 iph->payload_len = 0;
2102 iph->nexthdr = IPPROTO_NONE;
2103 iph->hop_limit = 0;
2104 ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr);
2105 ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr);
2106
2107 err = ip6mr_cache_unresolved(mrt, vif, skb2);
2108 read_unlock(&mrt_lock);
2109
2110 return err;
2111 }
2112
2113 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
2114 cache->mfc_flags |= MFC_NOTIFY;
2115
2116 err = __ip6mr_fill_mroute(mrt, skb, cache, rtm);
2117 read_unlock(&mrt_lock);
2118 return err;
2119}
2120
2121static int ip6mr_fill_mroute(struct mr6_table *mrt, struct sk_buff *skb,
2122 u32 pid, u32 seq, struct mfc6_cache *c)
2123{
2124 struct nlmsghdr *nlh;
2125 struct rtmsg *rtm;
2126
2127 nlh = nlmsg_put(skb, pid, seq, RTM_NEWROUTE, sizeof(*rtm), NLM_F_MULTI);
2128 if (nlh == NULL)
2129 return -EMSGSIZE;
2130
2131 rtm = nlmsg_data(nlh);
2132 rtm->rtm_family = RTNL_FAMILY_IPMR;
2133 rtm->rtm_dst_len = 128;
2134 rtm->rtm_src_len = 128;
2135 rtm->rtm_tos = 0;
2136 rtm->rtm_table = mrt->id;
2137 NLA_PUT_U32(skb, RTA_TABLE, mrt->id);
2138 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2139 rtm->rtm_protocol = RTPROT_UNSPEC;
2140 rtm->rtm_flags = 0;
2141
2142 NLA_PUT(skb, RTA_SRC, 16, &c->mf6c_origin);
2143 NLA_PUT(skb, RTA_DST, 16, &c->mf6c_mcastgrp);
2144
2145 if (__ip6mr_fill_mroute(mrt, skb, c, rtm) < 0)
2146 goto nla_put_failure;
2147
2148 return nlmsg_end(skb, nlh);
2149
2150nla_put_failure:
2151 nlmsg_cancel(skb, nlh);
2152 return -EMSGSIZE;
2153}
2154
2155static int ip6mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2156{
2157 struct net *net = sock_net(skb->sk);
2158 struct mr6_table *mrt;
2159 struct mfc6_cache *mfc;
2160 unsigned int t = 0, s_t;
2161 unsigned int h = 0, s_h;
2162 unsigned int e = 0, s_e;
2163
2164 s_t = cb->args[0];
2165 s_h = cb->args[1];
2166 s_e = cb->args[2];
2167
2168 read_lock(&mrt_lock);
2169 ip6mr_for_each_table(mrt, net) {
2170 if (t < s_t)
2171 goto next_table;
2172 if (t > s_t)
2173 s_h = 0;
2174 for (h = s_h; h < MFC6_LINES; h++) {
2175 list_for_each_entry(mfc, &mrt->mfc6_cache_array[h], list) {
2176 if (e < s_e)
2177 goto next_entry;
2178 if (ip6mr_fill_mroute(mrt, skb,
2179 NETLINK_CB(cb->skb).pid,
2180 cb->nlh->nlmsg_seq,
2181 mfc) < 0)
2182 goto done;
2183next_entry:
2184 e++;
2185 }
2186 e = s_e = 0;
2187 }
2188 s_h = 0;
2189next_table:
2190 t++;
2191 }
2192done:
2193 read_unlock(&mrt_lock);
2194
2195 cb->args[2] = e;
2196 cb->args[1] = h;
2197 cb->args[0] = t;
2198
2199 return skb->len;
2200}
Clone of davem's net-next-2.6 tree