]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - net/ipv4/ipmr.c
projects / mirror_ubuntu-jammy-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'mips_5.3' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux
[mirror_ubuntu-jammy-kernel.git] / net / ipv4 / ipmr.c
CommitLineData
2874c5fd 1// SPDX-License-Identifier: GPL-2.0-or-later
1da177e4
LT		 2/*
3 * IP multicast routing support for mrouted 3.6/3.8
4 *
113aa838 5 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT		 6 * Linux Consultancy and Custom Driver Development
7 *
1da177e4
LT		 8 * Fixes:
9 * Michael Chastain : Incorrect size of copying.
10 * Alan Cox : Added the cache manager code
11 * Alan Cox : Fixed the clone/copy bug and device race.
12 * Mike McLagan : Routing by source
13 * Malcolm Beattie : Buffer handling fixes.
14 * Alexey Kuznetsov : Double buffer free and other fixes.
15 * SVR Anand : Fixed several multicast bugs and problems.
16 * Alexey Kuznetsov : Status, optimisations and more.
17 * Brad Parker : Better behaviour on mrouted upcall
18 * overflow.
19 * Carlos Picoto : PIMv1 Support
20 * Pavlin Ivanov Radoslavov: PIMv2 Registers must checksum only PIM header
f77f13e2 21 * Relax this requirement to work with older peers.
1da177e4
LT		 22 */
23
7c0f6ba6 24#include <linux/uaccess.h>
1da177e4 25#include <linux/types.h>
08009a76 26#include <linux/cache.h>
4fc268d2 27#include <linux/capability.h>
1da177e4 28#include <linux/errno.h>
1da177e4
LT		 29#include <linux/mm.h>
30#include <linux/kernel.h>
31#include <linux/fcntl.h>
32#include <linux/stat.h>
33#include <linux/socket.h>
34#include <linux/in.h>
35#include <linux/inet.h>
36#include <linux/netdevice.h>
37#include <linux/inetdevice.h>
38#include <linux/igmp.h>
39#include <linux/proc_fs.h>
40#include <linux/seq_file.h>
41#include <linux/mroute.h>
42#include <linux/init.h>
46f25dff 43#include <linux/if_ether.h>
5a0e3ad6 44#include <linux/slab.h>
457c4cbc 45#include <net/net_namespace.h>
1da177e4
LT		 46#include <net/ip.h>
47#include <net/protocol.h>
48#include <linux/skbuff.h>
14c85021 49#include <net/route.h>
1da177e4
LT		 50#include <net/icmp.h>
51#include <net/udp.h>
52#include <net/raw.h>
53#include <linux/notifier.h>
54#include <linux/if_arp.h>
55#include <linux/netfilter_ipv4.h>
709b46e8 56#include <linux/compat.h>
bc3b2d7f 57#include <linux/export.h>
0eb71a9d 58#include <linux/rhashtable.h>
c5441932 59#include <net/ip_tunnels.h>
1da177e4 60#include <net/checksum.h>
dc5fc579 61#include <net/netlink.h>
f0ad0860 62#include <net/fib_rules.h>
d67b8c61 63#include <linux/netconf.h>
3c618c1d 64#include <net/rtnh.h>
1da177e4 65
5648451e
GS		 66#include <linux/nospec.h>
67
f0ad0860
PM		 68struct ipmr_rule {
69 struct fib_rule common;
70};
71
72struct ipmr_result {
73 struct mr_table *mrt;
74};
75
1da177e4 76/* Big lock, protecting vif table, mrt cache and mroute socket state.
a8cb16dd 77 * Note that the changes are semaphored via rtnl_lock.
1da177e4
LT		 78 */
79
80static DEFINE_RWLOCK(mrt_lock);
81
7ef8f65d 82/* Multicast router control variables */
1da177e4 83
1da177e4
LT		 84/* Special spinlock for queue of unresolved entries */
85static DEFINE_SPINLOCK(mfc_unres_lock);
86
87/* We return to original Alan's scheme. Hash table of resolved
a8cb16dd
ED		 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.
1da177e4
LT		 93 */
94
08009a76 95static struct kmem_cache *mrt_cachep __ro_after_init;
1da177e4 96
f0ad0860 97static struct mr_table *ipmr_new_table(struct net *net, u32 id);
acbb219d
FR		 98static void ipmr_free_table(struct mr_table *mrt);
99
c4854ec8 100static void ip_mr_forward(struct net *net, struct mr_table *mrt,
4b1f0d33
DS		 101 struct net_device *dev, struct sk_buff *skb,
102 struct mfc_cache *cache, int local);
0c12295a 103static int ipmr_cache_report(struct mr_table *mrt,
4feb88e5 104 struct sk_buff *pkt, vifi_t vifi, int assert);
8cd3ac9f
ND		 105static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
106 int cmd);
5a645dd8 107static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt);
ca8d4794 108static void mroute_clean_tables(struct mr_table *mrt, int flags);
e99e88a9 109static void ipmr_expire_process(struct timer_list *t);
f0ad0860
PM		 110
111#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
112#define ipmr_for_each_table(mrt, net) \
113 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
114
7b0db857
YM		 115static struct mr_table *ipmr_mr_table_iter(struct net *net,
116 struct mr_table *mrt)
117{
118 struct mr_table *ret;
119
120 if (!mrt)
121 ret = list_entry_rcu(net->ipv4.mr_tables.next,
122 struct mr_table, list);
123 else
124 ret = list_entry_rcu(mrt->list.next,
125 struct mr_table, list);
126
127 if (&ret->list == &net->ipv4.mr_tables)
128 return NULL;
129 return ret;
130}
131
f0ad0860
PM		 132static struct mr_table *ipmr_get_table(struct net *net, u32 id)
133{
134 struct mr_table *mrt;
135
136 ipmr_for_each_table(mrt, net) {
137 if (mrt->id == id)
138 return mrt;
139 }
140 return NULL;
141}
142
da91981b 143static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
f0ad0860
PM		 144 struct mr_table **mrt)
145{
f0ad0860 146 int err;
95f4a45d
HFS		 147 struct ipmr_result res;
148 struct fib_lookup_arg arg = {
149 .result = &res,
150 .flags = FIB_LOOKUP_NOREF,
151 };
f0ad0860 152
e58e4159
DA		 153 /* update flow if oif or iif point to device enslaved to l3mdev */
154 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
155
da91981b
DM		 156 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
157 flowi4_to_flowi(flp4), 0, &arg);
f0ad0860
PM		 158 if (err < 0)
159 return err;
160 *mrt = res.mrt;
161 return 0;
162}
163
164static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
165 int flags, struct fib_lookup_arg *arg)
166{
167 struct ipmr_result *res = arg->result;
168 struct mr_table *mrt;
1da177e4 169
f0ad0860
PM		 170 switch (rule->action) {
171 case FR_ACT_TO_TBL:
172 break;
173 case FR_ACT_UNREACHABLE:
174 return -ENETUNREACH;
175 case FR_ACT_PROHIBIT:
176 return -EACCES;
177 case FR_ACT_BLACKHOLE:
178 default:
179 return -EINVAL;
180 }
181
e58e4159
DA		 182 arg->table = fib_rule_get_table(rule, arg);
183
184 mrt = ipmr_get_table(rule->fr_net, arg->table);
51456b29 185 if (!mrt)
f0ad0860
PM		 186 return -EAGAIN;
187 res->mrt = mrt;
188 return 0;
189}
190
191static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
192{
193 return 1;
194}
195
196static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
197 FRA_GENERIC_POLICY,
198};
199
200static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
b16fb418
RP		 201 struct fib_rule_hdr *frh, struct nlattr **tb,
202 struct netlink_ext_ack *extack)
f0ad0860
PM		 203{
204 return 0;
205}
206
207static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
208 struct nlattr **tb)
209{
210 return 1;
211}
212
213static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
214 struct fib_rule_hdr *frh)
215{
216 frh->dst_len = 0;
217 frh->src_len = 0;
218 frh->tos = 0;
219 return 0;
220}
221
04a6f82c 222static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
25239cee 223 .family = RTNL_FAMILY_IPMR,
f0ad0860
PM		 224 .rule_size = sizeof(struct ipmr_rule),
225 .addr_size = sizeof(u32),
226 .action = ipmr_rule_action,
227 .match = ipmr_rule_match,
228 .configure = ipmr_rule_configure,
229 .compare = ipmr_rule_compare,
f0ad0860
PM		 230 .fill = ipmr_rule_fill,
231 .nlgroup = RTNLGRP_IPV4_RULE,
232 .policy = ipmr_rule_policy,
233 .owner = THIS_MODULE,
234};
235
236static int __net_init ipmr_rules_init(struct net *net)
237{
238 struct fib_rules_ops *ops;
239 struct mr_table *mrt;
240 int err;
241
242 ops = fib_rules_register(&ipmr_rules_ops_template, net);
243 if (IS_ERR(ops))
244 return PTR_ERR(ops);
245
246 INIT_LIST_HEAD(&net->ipv4.mr_tables);
247
248 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
1113ebbc
NA		 249 if (IS_ERR(mrt)) {
250 err = PTR_ERR(mrt);
f0ad0860
PM		 251 goto err1;
252 }
253
254 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
255 if (err < 0)
256 goto err2;
257
258 net->ipv4.mr_rules_ops = ops;
259 return 0;
260
261err2:
f243e5a7 262 ipmr_free_table(mrt);
f0ad0860
PM		 263err1:
264 fib_rules_unregister(ops);
265 return err;
266}
267
268static void __net_exit ipmr_rules_exit(struct net *net)
269{
270 struct mr_table *mrt, *next;
271
ed785309 272 rtnl_lock();
035320d5
ED		 273 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
274 list_del(&mrt->list);
acbb219d 275 ipmr_free_table(mrt);
035320d5 276 }
f0ad0860 277 fib_rules_unregister(net->ipv4.mr_rules_ops);
419df12f 278 rtnl_unlock();
f0ad0860 279}
4d65b948
YG		 280
281static int ipmr_rules_dump(struct net *net, struct notifier_block *nb)
282{
283 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR);
284}
285
286static unsigned int ipmr_rules_seq_read(struct net *net)
287{
288 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
289}
478e4c2f
YG		 290
291bool ipmr_rule_default(const struct fib_rule *rule)
292{
293 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
294}
295EXPORT_SYMBOL(ipmr_rule_default);
f0ad0860
PM		 296#else
297#define ipmr_for_each_table(mrt, net) \
298 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
299
7b0db857
YM		 300static struct mr_table *ipmr_mr_table_iter(struct net *net,
301 struct mr_table *mrt)
302{
303 if (!mrt)
304 return net->ipv4.mrt;
305 return NULL;
306}
307
f0ad0860
PM		 308static struct mr_table *ipmr_get_table(struct net *net, u32 id)
309{
310 return net->ipv4.mrt;
311}
312
da91981b 313static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
f0ad0860
PM		 314 struct mr_table **mrt)
315{
316 *mrt = net->ipv4.mrt;
317 return 0;
318}
319
320static int __net_init ipmr_rules_init(struct net *net)
321{
1113ebbc
NA		 322 struct mr_table *mrt;
323
324 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
325 if (IS_ERR(mrt))
326 return PTR_ERR(mrt);
327 net->ipv4.mrt = mrt;
328 return 0;
f0ad0860
PM		 329}
330
331static void __net_exit ipmr_rules_exit(struct net *net)
332{
ed785309 333 rtnl_lock();
acbb219d 334 ipmr_free_table(net->ipv4.mrt);
ed785309
WC		 335 net->ipv4.mrt = NULL;
336 rtnl_unlock();
f0ad0860 337}
4d65b948
YG		 338
339static int ipmr_rules_dump(struct net *net, struct notifier_block *nb)
340{
341 return 0;
342}
343
344static unsigned int ipmr_rules_seq_read(struct net *net)
345{
346 return 0;
347}
478e4c2f
YG		 348
349bool ipmr_rule_default(const struct fib_rule *rule)
350{
351 return true;
352}
353EXPORT_SYMBOL(ipmr_rule_default);
f0ad0860
PM		 354#endif
355
8fb472c0
NA		 356static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
357 const void *ptr)
358{
359 const struct mfc_cache_cmp_arg *cmparg = arg->key;
360 struct mfc_cache *c = (struct mfc_cache *)ptr;
361
362 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
363 cmparg->mfc_origin != c->mfc_origin;
364}
365
366static const struct rhashtable_params ipmr_rht_params = {
494fff56 367 .head_offset = offsetof(struct mr_mfc, mnode),
8fb472c0
NA		 368 .key_offset = offsetof(struct mfc_cache, cmparg),
369 .key_len = sizeof(struct mfc_cache_cmp_arg),
370 .nelem_hint = 3,
8fb472c0
NA		 371 .obj_cmpfn = ipmr_hash_cmp,
372 .automatic_shrinking = true,
373};
374
0bbbf0e7
YM		 375static void ipmr_new_table_set(struct mr_table *mrt,
376 struct net *net)
377{
378#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
379 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
380#endif
381}
382
845c9a7a
YM		 383static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
384 .mfc_mcastgrp = htonl(INADDR_ANY),
385 .mfc_origin = htonl(INADDR_ANY),
386};
387
388static struct mr_table_ops ipmr_mr_table_ops = {
389 .rht_params = &ipmr_rht_params,
390 .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
391};
392
f0ad0860
PM		 393static struct mr_table *ipmr_new_table(struct net *net, u32 id)
394{
395 struct mr_table *mrt;
1da177e4 396
1113ebbc
NA		 397 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
398 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
399 return ERR_PTR(-EINVAL);
400
f0ad0860 401 mrt = ipmr_get_table(net, id);
00db4124 402 if (mrt)
f0ad0860
PM		 403 return mrt;
404
845c9a7a 405 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
0bbbf0e7 406 ipmr_expire_process, ipmr_new_table_set);
f0ad0860 407}
1da177e4 408
acbb219d
FR		 409static void ipmr_free_table(struct mr_table *mrt)
410{
411 del_timer_sync(&mrt->ipmr_expire_timer);
ca8d4794
CS		 412 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
413 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
8fb472c0 414 rhltable_destroy(&mrt->mfc_hash);
acbb219d
FR		 415 kfree(mrt);
416}
417
1da177e4
LT		 418/* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
419
d607032d
WC		 420static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
421{
4feb88e5
BT		 422 struct net *net = dev_net(dev);
423
d607032d
WC		 424 dev_close(dev);
425
4feb88e5 426 dev = __dev_get_by_name(net, "tunl0");
d607032d 427 if (dev) {
5bc3eb7e 428 const struct net_device_ops *ops = dev->netdev_ops;
d607032d 429 struct ifreq ifr;
d607032d
WC		 430 struct ip_tunnel_parm p;
431
432 memset(&p, 0, sizeof(p));
433 p.iph.daddr = v->vifc_rmt_addr.s_addr;
434 p.iph.saddr = v->vifc_lcl_addr.s_addr;
435 p.iph.version = 4;
436 p.iph.ihl = 5;
437 p.iph.protocol = IPPROTO_IPIP;
438 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
439 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
440
5bc3eb7e
SH		 441 if (ops->ndo_do_ioctl) {
442 mm_segment_t oldfs = get_fs();
443
444 set_fs(KERNEL_DS);
445 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
446 set_fs(oldfs);
447 }
d607032d
WC		 448 }
449}
450
a0b47736
NA		 451/* Initialize ipmr pimreg/tunnel in_device */
452static bool ipmr_init_vif_indev(const struct net_device *dev)
453{
454 struct in_device *in_dev;
455
456 ASSERT_RTNL();
457
458 in_dev = __in_dev_get_rtnl(dev);
459 if (!in_dev)
460 return false;
461 ipv4_devconf_setall(in_dev);
462 neigh_parms_data_state_setall(in_dev->arp_parms);
463 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
464
465 return true;
466}
467
7ef8f65d 468static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
1da177e4
LT		 469{
470 struct net_device *dev;
471
4feb88e5 472 dev = __dev_get_by_name(net, "tunl0");
1da177e4
LT		 473
474 if (dev) {
5bc3eb7e 475 const struct net_device_ops *ops = dev->netdev_ops;
1da177e4
LT		 476 int err;
477 struct ifreq ifr;
1da177e4 478 struct ip_tunnel_parm p;
1da177e4
LT		 479
480 memset(&p, 0, sizeof(p));
481 p.iph.daddr = v->vifc_rmt_addr.s_addr;
482 p.iph.saddr = v->vifc_lcl_addr.s_addr;
483 p.iph.version = 4;
484 p.iph.ihl = 5;
485 p.iph.protocol = IPPROTO_IPIP;
486 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
ba93ef74 487 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
1da177e4 488
5bc3eb7e
SH		 489 if (ops->ndo_do_ioctl) {
490 mm_segment_t oldfs = get_fs();
491
492 set_fs(KERNEL_DS);
493 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
494 set_fs(oldfs);
a8cb16dd 495 } else {
5bc3eb7e 496 err = -EOPNOTSUPP;
a8cb16dd 497 }
1da177e4
LT		 498 dev = NULL;
499
4feb88e5
BT		 500 if (err == 0 &&
501 (dev = __dev_get_by_name(net, p.name)) != NULL) {
1da177e4 502 dev->flags |= IFF_MULTICAST;
a0b47736 503 if (!ipmr_init_vif_indev(dev))
1da177e4 504 goto failure;
00f54e68 505 if (dev_open(dev, NULL))
1da177e4 506 goto failure;
7dc00c82 507 dev_hold(dev);
1da177e4
LT		 508 }
509 }
510 return dev;
511
512failure:
1da177e4
LT		 513 unregister_netdevice(dev);
514 return NULL;
515}
516
c316c629 517#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
6fef4c0c 518static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
1da177e4 519{
4feb88e5 520 struct net *net = dev_net(dev);
f0ad0860 521 struct mr_table *mrt;
da91981b
DM		 522 struct flowi4 fl4 = {
523 .flowi4_oif = dev->ifindex,
6a662719 524 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
da91981b 525 .flowi4_mark = skb->mark,
f0ad0860
PM		 526 };
527 int err;
528
da91981b 529 err = ipmr_fib_lookup(net, &fl4, &mrt);
e40dbc51
BG		 530 if (err < 0) {
531 kfree_skb(skb);
f0ad0860 532 return err;
e40dbc51 533 }
4feb88e5 534
1da177e4 535 read_lock(&mrt_lock);
cf3677ae
PE		 536 dev->stats.tx_bytes += skb->len;
537 dev->stats.tx_packets++;
0c12295a 538 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
1da177e4
LT		 539 read_unlock(&mrt_lock);
540 kfree_skb(skb);
6ed10654 541 return NETDEV_TX_OK;
1da177e4
LT		 542}
543
ee9b9596
ND		 544static int reg_vif_get_iflink(const struct net_device *dev)
545{
546 return 0;
547}
548
007c3838
SH		 549static const struct net_device_ops reg_vif_netdev_ops = {
550 .ndo_start_xmit = reg_vif_xmit,
ee9b9596 551 .ndo_get_iflink = reg_vif_get_iflink,
007c3838
SH		 552};
553
1da177e4
LT		 554static void reg_vif_setup(struct net_device *dev)
555{
556 dev->type = ARPHRD_PIMREG;
46f25dff 557 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
1da177e4 558 dev->flags = IFF_NOARP;
70cb4a45 559 dev->netdev_ops = &reg_vif_netdev_ops;
cf124db5 560 dev->needs_free_netdev = true;
403dbb97 561 dev->features |= NETIF_F_NETNS_LOCAL;
1da177e4
LT		 562}
563
f0ad0860 564static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
1da177e4
LT		 565{
566 struct net_device *dev;
f0ad0860 567 char name[IFNAMSIZ];
1da177e4 568
f0ad0860
PM		 569 if (mrt->id == RT_TABLE_DEFAULT)
570 sprintf(name, "pimreg");
571 else
572 sprintf(name, "pimreg%u", mrt->id);
1da177e4 573
c835a677 574 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
1da177e4 575
51456b29 576 if (!dev)
1da177e4
LT		 577 return NULL;
578
403dbb97
TG		 579 dev_net_set(dev, net);
580
1da177e4
LT		 581 if (register_netdevice(dev)) {
582 free_netdev(dev);
583 return NULL;
584 }
1da177e4 585
a0b47736 586 if (!ipmr_init_vif_indev(dev))
1da177e4 587 goto failure;
00f54e68 588 if (dev_open(dev, NULL))
1da177e4
LT		 589 goto failure;
590
7dc00c82
WC		 591 dev_hold(dev);
592
1da177e4
LT		 593 return dev;
594
595failure:
1da177e4
LT		 596 unregister_netdevice(dev);
597 return NULL;
598}
c316c629
NA		 599
600/* called with rcu_read_lock() */
601static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
602 unsigned int pimlen)
603{
604 struct net_device *reg_dev = NULL;
605 struct iphdr *encap;
606
607 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
7ef8f65d 608 /* Check that:
c316c629
NA		 609 * a. packet is really sent to a multicast group
610 * b. packet is not a NULL-REGISTER
611 * c. packet is not truncated
612 */
613 if (!ipv4_is_multicast(encap->daddr) ||
614 encap->tot_len == 0 ||
615 ntohs(encap->tot_len) + pimlen > skb->len)
616 return 1;
617
618 read_lock(&mrt_lock);
619 if (mrt->mroute_reg_vif_num >= 0)
620 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
621 read_unlock(&mrt_lock);
622
623 if (!reg_dev)
624 return 1;
625
626 skb->mac_header = skb->network_header;
627 skb_pull(skb, (u8 *)encap - skb->data);
628 skb_reset_network_header(skb);
629 skb->protocol = htons(ETH_P_IP);
630 skb->ip_summed = CHECKSUM_NONE;
631
632 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
633
634 netif_rx(skb);
635
636 return NET_RX_SUCCESS;
637}
638#else
639static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
640{
641 return NULL;
642}
1da177e4
LT		 643#endif
644
b362053a
YG		 645static int call_ipmr_vif_entry_notifiers(struct net *net,
646 enum fib_event_type event_type,
647 struct vif_device *vif,
648 vifi_t vif_index, u32 tb_id)
649{
bc67a0da
YM		 650 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
651 vif, vif_index, tb_id,
652 &net->ipv4.ipmr_seq);
b362053a
YG		 653}
654
b362053a
YG		 655static int call_ipmr_mfc_entry_notifiers(struct net *net,
656 enum fib_event_type event_type,
657 struct mfc_cache *mfc, u32 tb_id)
658{
54c4cad9
YM		 659 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
660 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
b362053a
YG		 661}
662
2c53040f
BH		 663/**
664 * vif_delete - Delete a VIF entry
7dc00c82 665 * @notify: Set to 1, if the caller is a notifier_call
1da177e4 666 */
0c12295a 667static int vif_delete(struct mr_table *mrt, int vifi, int notify,
d17fa6fa 668 struct list_head *head)
1da177e4 669{
b362053a 670 struct net *net = read_pnet(&mrt->net);
1da177e4
LT		 671 struct vif_device *v;
672 struct net_device *dev;
673 struct in_device *in_dev;
674
0c12295a 675 if (vifi < 0 || vifi >= mrt->maxvif)
1da177e4
LT		 676 return -EADDRNOTAVAIL;
677
0c12295a 678 v = &mrt->vif_table[vifi];
1da177e4 679
b362053a
YG		 680 if (VIF_EXISTS(mrt, vifi))
681 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
682 mrt->id);
683
1da177e4
LT		 684 write_lock_bh(&mrt_lock);
685 dev = v->dev;
686 v->dev = NULL;
687
688 if (!dev) {
689 write_unlock_bh(&mrt_lock);
690 return -EADDRNOTAVAIL;
691 }
692
0c12295a
PM		 693 if (vifi == mrt->mroute_reg_vif_num)
694 mrt->mroute_reg_vif_num = -1;
1da177e4 695
a8cb16dd 696 if (vifi + 1 == mrt->maxvif) {
1da177e4 697 int tmp;
a8cb16dd
ED		 698
699 for (tmp = vifi - 1; tmp >= 0; tmp--) {
0c12295a 700 if (VIF_EXISTS(mrt, tmp))
1da177e4
LT		 701 break;
702 }
0c12295a 703 mrt->maxvif = tmp+1;
1da177e4
LT		 704 }
705
706 write_unlock_bh(&mrt_lock);
707
708 dev_set_allmulti(dev, -1);
709
a8cb16dd
ED		 710 in_dev = __in_dev_get_rtnl(dev);
711 if (in_dev) {
42f811b8 712 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
3b022865 713 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
d67b8c61
ND		 714 NETCONFA_MC_FORWARDING,
715 dev->ifindex, &in_dev->cnf);
1da177e4
LT		 716 ip_rt_multicast_event(in_dev);
717 }
718
a8cb16dd 719 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
d17fa6fa 720 unregister_netdevice_queue(dev, head);
1da177e4
LT		 721
722 dev_put(dev);
723 return 0;
724}
725
a8c9486b 726static void ipmr_cache_free_rcu(struct rcu_head *head)
5c0a66f5 727{
494fff56 728 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
a8c9486b 729
494fff56 730 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
5c0a66f5
BT		 731}
732
8c13af2a 733static void ipmr_cache_free(struct mfc_cache *c)
a8c9486b 734{
494fff56 735 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
a8c9486b
ED		 736}
737
1da177e4 738/* Destroy an unresolved cache entry, killing queued skbs
a8cb16dd 739 * and reporting error to netlink readers.
1da177e4 740 */
0c12295a 741static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
1da177e4 742{
8de53dfb 743 struct net *net = read_pnet(&mrt->net);
1da177e4 744 struct sk_buff *skb;
9ef1d4c7 745 struct nlmsgerr *e;
1da177e4 746
0c12295a 747 atomic_dec(&mrt->cache_resolve_queue_len);
1da177e4 748
494fff56 749 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
eddc9ec5 750 if (ip_hdr(skb)->version == 0) {
af72868b
JB		 751 struct nlmsghdr *nlh = skb_pull(skb,
752 sizeof(struct iphdr));
1da177e4 753 nlh->nlmsg_type = NLMSG_ERROR;
573ce260 754 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1da177e4 755 skb_trim(skb, nlh->nlmsg_len);
573ce260 756 e = nlmsg_data(nlh);
9ef1d4c7
PM		 757 e->error = -ETIMEDOUT;
758 memset(&e->msg, 0, sizeof(e->msg));
2942e900 759
15e47304 760 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
a8cb16dd 761 } else {
1da177e4 762 kfree_skb(skb);
a8cb16dd 763 }
1da177e4
LT		 764 }
765
5c0a66f5 766 ipmr_cache_free(c);
1da177e4
LT		 767}
768
e258beb2 769/* Timer process for the unresolved queue. */
e99e88a9 770static void ipmr_expire_process(struct timer_list *t)
1da177e4 771{
e99e88a9 772 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
494fff56 773 struct mr_mfc *c, *next;
1da177e4 774 unsigned long expires;
494fff56 775 unsigned long now;
1da177e4
LT		 776
777 if (!spin_trylock(&mfc_unres_lock)) {
0c12295a 778 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
1da177e4
LT		 779 return;
780 }
781
0c12295a 782 if (list_empty(&mrt->mfc_unres_queue))
1da177e4
LT		 783 goto out;
784
785 now = jiffies;
786 expires = 10*HZ;
1da177e4 787
0c12295a 788 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1da177e4
LT		 789 if (time_after(c->mfc_un.unres.expires, now)) {
790 unsigned long interval = c->mfc_un.unres.expires - now;
791 if (interval < expires)
792 expires = interval;
1da177e4
LT		 793 continue;
794 }
795
862465f2 796 list_del(&c->list);
494fff56
YM		 797 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
798 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
1da177e4
LT		 799 }
800
0c12295a
PM		 801 if (!list_empty(&mrt->mfc_unres_queue))
802 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
1da177e4
LT		 803
804out:
805 spin_unlock(&mfc_unres_lock);
806}
807
808/* Fill oifs list. It is called under write locked mrt_lock. */
494fff56 809static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
d658f8a0 810 unsigned char *ttls)
1da177e4
LT		 811{
812 int vifi;
813
814 cache->mfc_un.res.minvif = MAXVIFS;
815 cache->mfc_un.res.maxvif = 0;
816 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
817
0c12295a
PM		 818 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
819 if (VIF_EXISTS(mrt, vifi) &&
cf958ae3 820 ttls[vifi] && ttls[vifi] < 255) {
1da177e4
LT		 821 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
822 if (cache->mfc_un.res.minvif > vifi)
823 cache->mfc_un.res.minvif = vifi;
824 if (cache->mfc_un.res.maxvif <= vifi)
825 cache->mfc_un.res.maxvif = vifi + 1;
826 }
827 }
90b5ca17 828 cache->mfc_un.res.lastuse = jiffies;
1da177e4
LT		 829}
830
0c12295a
PM		 831static int vif_add(struct net *net, struct mr_table *mrt,
832 struct vifctl *vifc, int mrtsock)
1da177e4 833{
bccb3025 834 struct netdev_phys_item_id ppid = { };
1da177e4 835 int vifi = vifc->vifc_vifi;
0c12295a 836 struct vif_device *v = &mrt->vif_table[vifi];
1da177e4
LT		 837 struct net_device *dev;
838 struct in_device *in_dev;
d607032d 839 int err;
1da177e4
LT		 840
841 /* Is vif busy ? */
0c12295a 842 if (VIF_EXISTS(mrt, vifi))
1da177e4
LT		 843 return -EADDRINUSE;
844
845 switch (vifc->vifc_flags) {
1da177e4 846 case VIFF_REGISTER:
1973a4ea 847 if (!ipmr_pimsm_enabled())
c316c629
NA		 848 return -EINVAL;
849 /* Special Purpose VIF in PIM
1da177e4
LT		 850 * All the packets will be sent to the daemon
851 */
0c12295a 852 if (mrt->mroute_reg_vif_num >= 0)
1da177e4 853 return -EADDRINUSE;
f0ad0860 854 dev = ipmr_reg_vif(net, mrt);
1da177e4
LT		 855 if (!dev)
856 return -ENOBUFS;
d607032d
WC		 857 err = dev_set_allmulti(dev, 1);
858 if (err) {
859 unregister_netdevice(dev);
7dc00c82 860 dev_put(dev);
d607032d
WC		 861 return err;
862 }
1da177e4 863 break;
e905a9ed 864 case VIFF_TUNNEL:
4feb88e5 865 dev = ipmr_new_tunnel(net, vifc);
1da177e4
LT		 866 if (!dev)
867 return -ENOBUFS;
d607032d
WC		 868 err = dev_set_allmulti(dev, 1);
869 if (err) {
870 ipmr_del_tunnel(dev, vifc);
7dc00c82 871 dev_put(dev);
d607032d
WC		 872 return err;
873 }
1da177e4 874 break;
ee5e81f0 875 case VIFF_USE_IFINDEX:
1da177e4 876 case 0:
ee5e81f0
I		 877 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
878 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
51456b29 879 if (dev && !__in_dev_get_rtnl(dev)) {
ee5e81f0
I		 880 dev_put(dev);
881 return -EADDRNOTAVAIL;
882 }
a8cb16dd 883 } else {
ee5e81f0 884 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
a8cb16dd 885 }
1da177e4
LT		 886 if (!dev)
887 return -EADDRNOTAVAIL;
d607032d 888 err = dev_set_allmulti(dev, 1);
7dc00c82
WC		 889 if (err) {
890 dev_put(dev);
d607032d 891 return err;
7dc00c82 892 }
1da177e4
LT		 893 break;
894 default:
895 return -EINVAL;
896 }
897
a8cb16dd
ED		 898 in_dev = __in_dev_get_rtnl(dev);
899 if (!in_dev) {
d0490cfd 900 dev_put(dev);
1da177e4 901 return -EADDRNOTAVAIL;
d0490cfd 902 }
42f811b8 903 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
3b022865
DA		 904 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
905 dev->ifindex, &in_dev->cnf);
1da177e4
LT		 906 ip_rt_multicast_event(in_dev);
907
a8cb16dd 908 /* Fill in the VIF structures */
6853f21f
YM		 909 vif_device_init(v, dev, vifc->vifc_rate_limit,
910 vifc->vifc_threshold,
911 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
912 (VIFF_TUNNEL | VIFF_REGISTER));
a8cb16dd 913
bccb3025
FF		 914 err = dev_get_port_parent_id(dev, &ppid, true);
915 if (err == 0) {
916 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
917 v->dev_parent_id.id_len = ppid.id_len;
5d8b3e69
YG		 918 } else {
919 v->dev_parent_id.id_len = 0;
920 }
6853f21f 921
c354e124
JK		 922 v->local = vifc->vifc_lcl_addr.s_addr;
923 v->remote = vifc->vifc_rmt_addr.s_addr;
1da177e4
LT		 924
925 /* And finish update writing critical data */
926 write_lock_bh(&mrt_lock);
c354e124 927 v->dev = dev;
a8cb16dd 928 if (v->flags & VIFF_REGISTER)
0c12295a 929 mrt->mroute_reg_vif_num = vifi;
0c12295a
PM		 930 if (vifi+1 > mrt->maxvif)
931 mrt->maxvif = vifi+1;
1da177e4 932 write_unlock_bh(&mrt_lock);
b362053a 933 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, vifi, mrt->id);
1da177e4
LT		 934 return 0;
935}
936
a8c9486b 937/* called with rcu_read_lock() */
0c12295a 938static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
4feb88e5
BT		 939 __be32 origin,
940 __be32 mcastgrp)
1da177e4 941{
8fb472c0
NA		 942 struct mfc_cache_cmp_arg arg = {
943 .mfc_mcastgrp = mcastgrp,
944 .mfc_origin = origin
945 };
660b26dc 946
845c9a7a 947 return mr_mfc_find(mrt, &arg);
660b26dc
ND		 948}
949
950/* Look for a (*,G) entry */
951static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
952 __be32 mcastgrp, int vifi)
953{
8fb472c0
NA		 954 struct mfc_cache_cmp_arg arg = {
955 .mfc_mcastgrp = mcastgrp,
956 .mfc_origin = htonl(INADDR_ANY)
957 };
660b26dc 958
360eb5da 959 if (mcastgrp == htonl(INADDR_ANY))
845c9a7a
YM		 960 return mr_mfc_find_any_parent(mrt, vifi);
961 return mr_mfc_find_any(mrt, vifi, &arg);
660b26dc
ND		 962}
963
8fb472c0
NA		 964/* Look for a (S,G,iif) entry if parent != -1 */
965static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
966 __be32 origin, __be32 mcastgrp,
967 int parent)
968{
969 struct mfc_cache_cmp_arg arg = {
970 .mfc_mcastgrp = mcastgrp,
971 .mfc_origin = origin,
972 };
8fb472c0 973
845c9a7a 974 return mr_mfc_find_parent(mrt, &arg, parent);
8fb472c0
NA		 975}
976
7ef8f65d 977/* Allocate a multicast cache entry */
d658f8a0 978static struct mfc_cache *ipmr_cache_alloc(void)
1da177e4 979{
c354e124 980 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
a8c9486b 981
70a0dec4 982 if (c) {
494fff56
YM		 983 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
984 c->_c.mfc_un.res.minvif = MAXVIFS;
8c13af2a 985 c->_c.free = ipmr_cache_free_rcu;
494fff56 986 refcount_set(&c->_c.mfc_un.res.refcount, 1);
70a0dec4 987 }
1da177e4
LT		 988 return c;
989}
990
d658f8a0 991static struct mfc_cache *ipmr_cache_alloc_unres(void)
1da177e4 992{
c354e124 993 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
a8c9486b
ED		 994
995 if (c) {
494fff56
YM		 996 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
997 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
a8c9486b 998 }
1da177e4
LT		 999 return c;
1000}
1001
7ef8f65d 1002/* A cache entry has gone into a resolved state from queued */
0c12295a
PM		 1003static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
1004 struct mfc_cache *uc, struct mfc_cache *c)
1da177e4
LT		 1005{
1006 struct sk_buff *skb;
9ef1d4c7 1007 struct nlmsgerr *e;
1da177e4 1008
a8cb16dd 1009 /* Play the pending entries through our router */
494fff56 1010 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
eddc9ec5 1011 if (ip_hdr(skb)->version == 0) {
af72868b
JB		 1012 struct nlmsghdr *nlh = skb_pull(skb,
1013 sizeof(struct iphdr));
1da177e4 1014
7b0db857
YM		 1015 if (mr_fill_mroute(mrt, skb, &c->_c,
1016 nlmsg_data(nlh)) > 0) {
a8cb16dd
ED		 1017 nlh->nlmsg_len = skb_tail_pointer(skb) -
1018 (u8 *)nlh;
1da177e4
LT		 1019 } else {
1020 nlh->nlmsg_type = NLMSG_ERROR;
573ce260 1021 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1da177e4 1022 skb_trim(skb, nlh->nlmsg_len);
573ce260 1023 e = nlmsg_data(nlh);
9ef1d4c7
PM		 1024 e->error = -EMSGSIZE;
1025 memset(&e->msg, 0, sizeof(e->msg));
1da177e4 1026 }
2942e900 1027
15e47304 1028 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
a8cb16dd 1029 } else {
4b1f0d33 1030 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
a8cb16dd 1031 }
1da177e4
LT		 1032 }
1033}
1034
5a645dd8 1035/* Bounce a cache query up to mrouted and netlink.
1da177e4 1036 *
c316c629 1037 * Called under mrt_lock.
1da177e4 1038 */
0c12295a 1039static int ipmr_cache_report(struct mr_table *mrt,
4feb88e5 1040 struct sk_buff *pkt, vifi_t vifi, int assert)
1da177e4 1041{
c9bdd4b5 1042 const int ihl = ip_hdrlen(pkt);
c316c629 1043 struct sock *mroute_sk;
1da177e4
LT		 1044 struct igmphdr *igmp;
1045 struct igmpmsg *msg;
c316c629 1046 struct sk_buff *skb;
1da177e4
LT		 1047 int ret;
1048
c921c207 1049 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1da177e4
LT		 1050 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1051 else
1da177e4
LT		 1052 skb = alloc_skb(128, GFP_ATOMIC);
1053
132adf54 1054 if (!skb)
1da177e4
LT		 1055 return -ENOBUFS;
1056
c921c207 1057 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1da177e4 1058 /* Ugly, but we have no choice with this interface.
a8cb16dd
ED		 1059 * Duplicate old header, fix ihl, length etc.
1060 * And all this only to mangle msg->im_msgtype and
1061 * to set msg->im_mbz to "mbz" :-)
1da177e4 1062 */
878c8145
ACM		 1063 skb_push(skb, sizeof(struct iphdr));
1064 skb_reset_network_header(skb);
badff6d0 1065 skb_reset_transport_header(skb);
0272ffc4 1066 msg = (struct igmpmsg *)skb_network_header(skb);
d56f90a7 1067 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
c921c207 1068 msg->im_msgtype = assert;
1da177e4 1069 msg->im_mbz = 0;
c921c207
NA		 1070 if (assert == IGMPMSG_WRVIFWHOLE)
1071 msg->im_vif = vifi;
1072 else
1073 msg->im_vif = mrt->mroute_reg_vif_num;
eddc9ec5
ACM		 1074 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1075 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1076 sizeof(struct iphdr));
c316c629
NA		 1077 } else {
1078 /* Copy the IP header */
1079 skb_set_network_header(skb, skb->len);
1080 skb_put(skb, ihl);
1081 skb_copy_to_linear_data(skb, pkt->data, ihl);
1082 /* Flag to the kernel this is a route add */
1083 ip_hdr(skb)->protocol = 0;
1084 msg = (struct igmpmsg *)skb_network_header(skb);
1085 msg->im_vif = vifi;
1086 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1087 /* Add our header */
4df864c1 1088 igmp = skb_put(skb, sizeof(struct igmphdr));
c316c629
NA		 1089 igmp->type = assert;
1090 msg->im_msgtype = assert;
1091 igmp->code = 0;
1092 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1093 skb->transport_header = skb->network_header;
e905a9ed 1094 }
1da177e4 1095
4c968709
ED		 1096 rcu_read_lock();
1097 mroute_sk = rcu_dereference(mrt->mroute_sk);
51456b29 1098 if (!mroute_sk) {
4c968709 1099 rcu_read_unlock();
1da177e4
LT		 1100 kfree_skb(skb);
1101 return -EINVAL;
1102 }
1103
5a645dd8
JG		 1104 igmpmsg_netlink_event(mrt, skb);
1105
a8cb16dd 1106 /* Deliver to mrouted */
4c968709
ED		 1107 ret = sock_queue_rcv_skb(mroute_sk, skb);
1108 rcu_read_unlock();
70a269e6 1109 if (ret < 0) {
e87cc472 1110 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1da177e4
LT		 1111 kfree_skb(skb);
1112 }
1113
1114 return ret;
1115}
1116
7ef8f65d
NA		 1117/* Queue a packet for resolution. It gets locked cache entry! */
1118static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
4b1f0d33 1119 struct sk_buff *skb, struct net_device *dev)
1da177e4 1120{
8fb472c0
NA		 1121 const struct iphdr *iph = ip_hdr(skb);
1122 struct mfc_cache *c;
862465f2 1123 bool found = false;
1da177e4 1124 int err;
1da177e4
LT		 1125
1126 spin_lock_bh(&mfc_unres_lock);
494fff56 1127 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
e258beb2 1128 if (c->mfc_mcastgrp == iph->daddr &&
862465f2
PM		 1129 c->mfc_origin == iph->saddr) {
1130 found = true;
1da177e4 1131 break;
862465f2 1132 }
1da177e4
LT		 1133 }
1134
862465f2 1135 if (!found) {
a8cb16dd 1136 /* Create a new entry if allowable */
0c12295a 1137 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
d658f8a0 1138 (c = ipmr_cache_alloc_unres()) == NULL) {
1da177e4
LT		 1139 spin_unlock_bh(&mfc_unres_lock);
1140
1141 kfree_skb(skb);
1142 return -ENOBUFS;
1143 }
1144
a8cb16dd 1145 /* Fill in the new cache entry */
494fff56 1146 c->_c.mfc_parent = -1;
eddc9ec5
ACM		 1147 c->mfc_origin = iph->saddr;
1148 c->mfc_mcastgrp = iph->daddr;
1da177e4 1149
a8cb16dd 1150 /* Reflect first query at mrouted. */
0c12295a 1151 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
494fff56 1152
4feb88e5 1153 if (err < 0) {
e905a9ed 1154 /* If the report failed throw the cache entry
1da177e4
LT		 1155 out - Brad Parker
1156 */
1157 spin_unlock_bh(&mfc_unres_lock);
1158
5c0a66f5 1159 ipmr_cache_free(c);
1da177e4
LT		 1160 kfree_skb(skb);
1161 return err;
1162 }
1163
0c12295a 1164 atomic_inc(&mrt->cache_resolve_queue_len);
494fff56 1165 list_add(&c->_c.list, &mrt->mfc_unres_queue);
8cd3ac9f 1166 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1da177e4 1167
278554bd 1168 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
494fff56
YM		 1169 mod_timer(&mrt->ipmr_expire_timer,
1170 c->_c.mfc_un.unres.expires);
1da177e4
LT		 1171 }
1172
a8cb16dd 1173 /* See if we can append the packet */
494fff56 1174 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1da177e4
LT		 1175 kfree_skb(skb);
1176 err = -ENOBUFS;
1177 } else {
4b1f0d33
DS		 1178 if (dev) {
1179 skb->dev = dev;
1180 skb->skb_iif = dev->ifindex;
1181 }
494fff56 1182 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1da177e4
LT		 1183 err = 0;
1184 }
1185
1186 spin_unlock_bh(&mfc_unres_lock);
1187 return err;
1188}
1189
7ef8f65d 1190/* MFC cache manipulation by user space mroute daemon */
1da177e4 1191
660b26dc 1192static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1da177e4 1193{
b362053a 1194 struct net *net = read_pnet(&mrt->net);
8fb472c0 1195 struct mfc_cache *c;
1da177e4 1196
8fb472c0
NA		 1197 /* The entries are added/deleted only under RTNL */
1198 rcu_read_lock();
1199 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1200 mfc->mfcc_mcastgrp.s_addr, parent);
1201 rcu_read_unlock();
1202 if (!c)
1203 return -ENOENT;
494fff56
YM		 1204 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1205 list_del_rcu(&c->_c.list);
b362053a 1206 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
8fb472c0 1207 mroute_netlink_event(mrt, c, RTM_DELROUTE);
8c13af2a 1208 mr_cache_put(&c->_c);
1da177e4 1209
8fb472c0 1210 return 0;
1da177e4
LT		 1211}
1212
0c12295a 1213static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
660b26dc 1214 struct mfcctl *mfc, int mrtsock, int parent)
1da177e4 1215{
862465f2 1216 struct mfc_cache *uc, *c;
494fff56 1217 struct mr_mfc *_uc;
8fb472c0
NA		 1218 bool found;
1219 int ret;
1da177e4 1220
a50436f2
PM		 1221 if (mfc->mfcc_parent >= MAXVIFS)
1222 return -ENFILE;
1223
8fb472c0
NA		 1224 /* The entries are added/deleted only under RTNL */
1225 rcu_read_lock();
1226 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1227 mfc->mfcc_mcastgrp.s_addr, parent);
1228 rcu_read_unlock();
1229 if (c) {
1da177e4 1230 write_lock_bh(&mrt_lock);
494fff56
YM		 1231 c->_c.mfc_parent = mfc->mfcc_parent;
1232 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1da177e4 1233 if (!mrtsock)
494fff56 1234 c->_c.mfc_flags |= MFC_STATIC;
1da177e4 1235 write_unlock_bh(&mrt_lock);
b362053a
YG		 1236 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1237 mrt->id);
8cd3ac9f 1238 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1da177e4
LT		 1239 return 0;
1240 }
1241
360eb5da 1242 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
660b26dc 1243 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1da177e4
LT		 1244 return -EINVAL;
1245
d658f8a0 1246 c = ipmr_cache_alloc();
51456b29 1247 if (!c)
1da177e4
LT		 1248 return -ENOMEM;
1249
c354e124
JK		 1250 c->mfc_origin = mfc->mfcc_origin.s_addr;
1251 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
494fff56
YM		 1252 c->_c.mfc_parent = mfc->mfcc_parent;
1253 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1da177e4 1254 if (!mrtsock)
494fff56 1255 c->_c.mfc_flags |= MFC_STATIC;
1da177e4 1256
494fff56 1257 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
8fb472c0
NA		 1258 ipmr_rht_params);
1259 if (ret) {
1260 pr_err("ipmr: rhtable insert error %d\n", ret);
1261 ipmr_cache_free(c);
1262 return ret;
1263 }
494fff56 1264 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
7ef8f65d
NA		 1265 /* Check to see if we resolved a queued list. If so we
1266 * need to send on the frames and tidy up.
1da177e4 1267 */
b0ebb739 1268 found = false;
1da177e4 1269 spin_lock_bh(&mfc_unres_lock);
494fff56
YM		 1270 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1271 uc = (struct mfc_cache *)_uc;
e258beb2 1272 if (uc->mfc_origin == c->mfc_origin &&
1da177e4 1273 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
494fff56 1274 list_del(&_uc->list);
0c12295a 1275 atomic_dec(&mrt->cache_resolve_queue_len);
b0ebb739 1276 found = true;
1da177e4
LT		 1277 break;
1278 }
1279 }
0c12295a
PM		 1280 if (list_empty(&mrt->mfc_unres_queue))
1281 del_timer(&mrt->ipmr_expire_timer);
1da177e4
LT		 1282 spin_unlock_bh(&mfc_unres_lock);
1283
b0ebb739 1284 if (found) {
0c12295a 1285 ipmr_cache_resolve(net, mrt, uc, c);
5c0a66f5 1286 ipmr_cache_free(uc);
1da177e4 1287 }
b362053a 1288 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
8cd3ac9f 1289 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1da177e4
LT		 1290 return 0;
1291}
1292
7ef8f65d 1293/* Close the multicast socket, and clear the vif tables etc */
ca8d4794 1294static void mroute_clean_tables(struct mr_table *mrt, int flags)
1da177e4 1295{
b362053a 1296 struct net *net = read_pnet(&mrt->net);
494fff56
YM		 1297 struct mr_mfc *c, *tmp;
1298 struct mfc_cache *cache;
d17fa6fa 1299 LIST_HEAD(list);
8fb472c0 1300 int i;
e905a9ed 1301
a8cb16dd 1302 /* Shut down all active vif entries */
ca8d4794
CS		 1303 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1304 for (i = 0; i < mrt->maxvif; i++) {
1305 if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1306 !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1307 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1308 continue;
1309 vif_delete(mrt, i, 0, &list);
1310 }
1311 unregister_netdevice_many(&list);
1da177e4
LT		 1312 }
1313
a8cb16dd 1314 /* Wipe the cache */
ca8d4794
CS		 1315 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1316 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1317 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1318 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1319 continue;
1320 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1321 list_del_rcu(&c->list);
494fff56 1322 cache = (struct mfc_cache *)c;
ca8d4794
CS		 1323 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1324 mrt->id);
494fff56 1325 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
ca8d4794
CS		 1326 mr_cache_put(c);
1327 }
1328 }
1329
1330 if (flags & MRT_FLUSH_MFC) {
1331 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1332 spin_lock_bh(&mfc_unres_lock);
1333 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1334 list_del(&c->list);
1335 cache = (struct mfc_cache *)c;
1336 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1337 ipmr_destroy_unres(mrt, cache);
1338 }
1339 spin_unlock_bh(&mfc_unres_lock);
1da177e4 1340 }
1da177e4
LT		 1341 }
1342}
1343
4c968709
ED		 1344/* called from ip_ra_control(), before an RCU grace period,
1345 * we dont need to call synchronize_rcu() here
1346 */
1da177e4
LT		 1347static void mrtsock_destruct(struct sock *sk)
1348{
4feb88e5 1349 struct net *net = sock_net(sk);
f0ad0860 1350 struct mr_table *mrt;
4feb88e5 1351
128aaa98 1352 rtnl_lock();
f0ad0860 1353 ipmr_for_each_table(mrt, net) {
4c968709 1354 if (sk == rtnl_dereference(mrt->mroute_sk)) {
f0ad0860 1355 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
3b022865
DA		 1356 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1357 NETCONFA_MC_FORWARDING,
d67b8c61
ND		 1358 NETCONFA_IFINDEX_ALL,
1359 net->ipv4.devconf_all);
a9b3cd7f 1360 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
ca8d4794 1361 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
f0ad0860 1362 }
1da177e4 1363 }
128aaa98 1364 rtnl_unlock();
1da177e4
LT		 1365}
1366
7ef8f65d
NA		 1367/* Socket options and virtual interface manipulation. The whole
1368 * virtual interface system is a complete heap, but unfortunately
1369 * that's how BSD mrouted happens to think. Maybe one day with a proper
1370 * MOSPF/PIM router set up we can clean this up.
1da177e4 1371 */
e905a9ed 1372
29e97d21
NA		 1373int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval,
1374 unsigned int optlen)
1da177e4 1375{
4feb88e5 1376 struct net *net = sock_net(sk);
29e97d21 1377 int val, ret = 0, parent = 0;
f0ad0860 1378 struct mr_table *mrt;
29e97d21
NA		 1379 struct vifctl vif;
1380 struct mfcctl mfc;
c921c207 1381 bool do_wrvifwhole;
29e97d21 1382 u32 uval;
f0ad0860 1383
29e97d21
NA		 1384 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1385 rtnl_lock();
5e1859fb 1386 if (sk->sk_type != SOCK_RAW ||
29e97d21
NA		 1387 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1388 ret = -EOPNOTSUPP;
1389 goto out_unlock;
1390 }
5e1859fb 1391
f0ad0860 1392 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
29e97d21
NA		 1393 if (!mrt) {
1394 ret = -ENOENT;
1395 goto out_unlock;
1396 }
132adf54 1397 if (optname != MRT_INIT) {
33d480ce 1398 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
29e97d21
NA		 1399 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1400 ret = -EACCES;
1401 goto out_unlock;
1402 }
1da177e4
LT		 1403 }
1404
132adf54
SH		 1405 switch (optname) {
1406 case MRT_INIT:
42e6b89c 1407 if (optlen != sizeof(int)) {
29e97d21 1408 ret = -EINVAL;
42e6b89c
NA		 1409 break;
1410 }
1411 if (rtnl_dereference(mrt->mroute_sk)) {
29e97d21 1412 ret = -EADDRINUSE;
29e97d21 1413 break;
42e6b89c 1414 }
132adf54
SH		 1415
1416 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1417 if (ret == 0) {
cf778b00 1418 rcu_assign_pointer(mrt->mroute_sk, sk);
4feb88e5 1419 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
3b022865
DA		 1420 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1421 NETCONFA_MC_FORWARDING,
d67b8c61
ND		 1422 NETCONFA_IFINDEX_ALL,
1423 net->ipv4.devconf_all);
132adf54 1424 }
29e97d21 1425 break;
132adf54 1426 case MRT_DONE:
29e97d21
NA		 1427 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1428 ret = -EACCES;
1429 } else {
128aaa98
KT		 1430 /* We need to unlock here because mrtsock_destruct takes
1431 * care of rtnl itself and we can't change that due to
1432 * the IP_ROUTER_ALERT setsockopt which runs without it.
1433 */
1434 rtnl_unlock();
29e97d21 1435 ret = ip_ra_control(sk, 0, NULL);
128aaa98 1436 goto out;
29e97d21
NA		 1437 }
1438 break;
132adf54
SH		 1439 case MRT_ADD_VIF:
1440 case MRT_DEL_VIF:
29e97d21
NA		 1441 if (optlen != sizeof(vif)) {
1442 ret = -EINVAL;
1443 break;
1444 }
1445 if (copy_from_user(&vif, optval, sizeof(vif))) {
1446 ret = -EFAULT;
1447 break;
1448 }
1449 if (vif.vifc_vifi >= MAXVIFS) {
1450 ret = -ENFILE;
1451 break;
1452 }
c354e124 1453 if (optname == MRT_ADD_VIF) {
4c968709
ED		 1454 ret = vif_add(net, mrt, &vif,
1455 sk == rtnl_dereference(mrt->mroute_sk));
132adf54 1456 } else {
0c12295a 1457 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
132adf54 1458 }
29e97d21 1459 break;
7ef8f65d
NA		 1460 /* Manipulate the forwarding caches. These live
1461 * in a sort of kernel/user symbiosis.
1462 */
132adf54
SH		 1463 case MRT_ADD_MFC:
1464 case MRT_DEL_MFC:
660b26dc 1465 parent = -1;
fcfd6dfa 1466 /* fall through */
660b26dc
ND		 1467 case MRT_ADD_MFC_PROXY:
1468 case MRT_DEL_MFC_PROXY:
29e97d21
NA		 1469 if (optlen != sizeof(mfc)) {
1470 ret = -EINVAL;
1471 break;
1472 }
1473 if (copy_from_user(&mfc, optval, sizeof(mfc))) {
1474 ret = -EFAULT;
1475 break;
1476 }
660b26dc
ND		 1477 if (parent == 0)
1478 parent = mfc.mfcc_parent;
660b26dc
ND		 1479 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1480 ret = ipmr_mfc_delete(mrt, &mfc, parent);
132adf54 1481 else
4c968709 1482 ret = ipmr_mfc_add(net, mrt, &mfc,
660b26dc
ND		 1483 sk == rtnl_dereference(mrt->mroute_sk),
1484 parent);
29e97d21 1485 break;
ca8d4794
CS		 1486 case MRT_FLUSH:
1487 if (optlen != sizeof(val)) {
1488 ret = -EINVAL;
1489 break;
1490 }
1491 if (get_user(val, (int __user *)optval)) {
1492 ret = -EFAULT;
1493 break;
1494 }
1495 mroute_clean_tables(mrt, val);
1496 break;
7ef8f65d 1497 /* Control PIM assert. */
132adf54 1498 case MRT_ASSERT:
29e97d21
NA		 1499 if (optlen != sizeof(val)) {
1500 ret = -EINVAL;
1501 break;
1502 }
1503 if (get_user(val, (int __user *)optval)) {
1504 ret = -EFAULT;
1505 break;
1506 }
1507 mrt->mroute_do_assert = val;
1508 break;
132adf54 1509 case MRT_PIM:
1973a4ea 1510 if (!ipmr_pimsm_enabled()) {
29e97d21
NA		 1511 ret = -ENOPROTOOPT;
1512 break;
1513 }
1514 if (optlen != sizeof(val)) {
1515 ret = -EINVAL;
1516 break;
1517 }
1518 if (get_user(val, (int __user *)optval)) {
1519 ret = -EFAULT;
1520 break;
1521 }
ba93ef74 1522
c921c207 1523 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
29e97d21
NA		 1524 val = !!val;
1525 if (val != mrt->mroute_do_pim) {
1526 mrt->mroute_do_pim = val;
1527 mrt->mroute_do_assert = val;
c921c207 1528 mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1da177e4 1529 }
29e97d21 1530 break;
f0ad0860 1531 case MRT_TABLE:
29e97d21
NA		 1532 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1533 ret = -ENOPROTOOPT;
1534 break;
1535 }
1536 if (optlen != sizeof(uval)) {
1537 ret = -EINVAL;
1538 break;
1539 }
1540 if (get_user(uval, (u32 __user *)optval)) {
1541 ret = -EFAULT;
1542 break;
1543 }
f0ad0860 1544
4c968709
ED		 1545 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1546 ret = -EBUSY;
1547 } else {
29e97d21 1548 mrt = ipmr_new_table(net, uval);
1113ebbc
NA		 1549 if (IS_ERR(mrt))
1550 ret = PTR_ERR(mrt);
5e1859fb 1551 else
29e97d21 1552 raw_sk(sk)->ipmr_table = uval;
4c968709 1553 }
29e97d21 1554 break;
7ef8f65d 1555 /* Spurious command, or MRT_VERSION which you cannot set. */
132adf54 1556 default:
29e97d21 1557 ret = -ENOPROTOOPT;
1da177e4 1558 }
29e97d21
NA		 1559out_unlock:
1560 rtnl_unlock();
128aaa98 1561out:
29e97d21 1562 return ret;
1da177e4
LT		 1563}
1564
7ef8f65d 1565/* Getsock opt support for the multicast routing system. */
c354e124 1566int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1da177e4
LT		 1567{
1568 int olr;
1569 int val;
4feb88e5 1570 struct net *net = sock_net(sk);
f0ad0860
PM		 1571 struct mr_table *mrt;
1572
5e1859fb
ED		 1573 if (sk->sk_type != SOCK_RAW ||
1574 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1575 return -EOPNOTSUPP;
1576
f0ad0860 1577 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
51456b29 1578 if (!mrt)
f0ad0860 1579 return -ENOENT;
1da177e4 1580
fe9ef3ce
NA		 1581 switch (optname) {
1582 case MRT_VERSION:
1583 val = 0x0305;
1584 break;
1585 case MRT_PIM:
1973a4ea 1586 if (!ipmr_pimsm_enabled())
fe9ef3ce
NA		 1587 return -ENOPROTOOPT;
1588 val = mrt->mroute_do_pim;
1589 break;
1590 case MRT_ASSERT:
1591 val = mrt->mroute_do_assert;
1592 break;
1593 default:
1da177e4 1594 return -ENOPROTOOPT;
fe9ef3ce 1595 }
1da177e4
LT		 1596
1597 if (get_user(olr, optlen))
1598 return -EFAULT;
1da177e4
LT		 1599 olr = min_t(unsigned int, olr, sizeof(int));
1600 if (olr < 0)
1601 return -EINVAL;
c354e124 1602 if (put_user(olr, optlen))
1da177e4 1603 return -EFAULT;
c354e124 1604 if (copy_to_user(optval, &val, olr))
1da177e4
LT		 1605 return -EFAULT;
1606 return 0;
1607}
1608
7ef8f65d 1609/* The IP multicast ioctl support routines. */
1da177e4
LT		 1610int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1611{
1612 struct sioc_sg_req sr;
1613 struct sioc_vif_req vr;
1614 struct vif_device *vif;
1615 struct mfc_cache *c;
4feb88e5 1616 struct net *net = sock_net(sk);
f0ad0860
PM		 1617 struct mr_table *mrt;
1618
1619 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
51456b29 1620 if (!mrt)
f0ad0860 1621 return -ENOENT;
e905a9ed 1622
132adf54
SH		 1623 switch (cmd) {
1624 case SIOCGETVIFCNT:
c354e124 1625 if (copy_from_user(&vr, arg, sizeof(vr)))
132adf54 1626 return -EFAULT;
0c12295a 1627 if (vr.vifi >= mrt->maxvif)
132adf54 1628 return -EINVAL;
5648451e 1629 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
132adf54 1630 read_lock(&mrt_lock);
0c12295a
PM		 1631 vif = &mrt->vif_table[vr.vifi];
1632 if (VIF_EXISTS(mrt, vr.vifi)) {
c354e124
JK		 1633 vr.icount = vif->pkt_in;
1634 vr.ocount = vif->pkt_out;
1635 vr.ibytes = vif->bytes_in;
1636 vr.obytes = vif->bytes_out;
1da177e4 1637 read_unlock(&mrt_lock);
1da177e4 1638
c354e124 1639 if (copy_to_user(arg, &vr, sizeof(vr)))
132adf54
SH		 1640 return -EFAULT;
1641 return 0;
1642 }
1643 read_unlock(&mrt_lock);
1644 return -EADDRNOTAVAIL;
1645 case SIOCGETSGCNT:
c354e124 1646 if (copy_from_user(&sr, arg, sizeof(sr)))
132adf54
SH		 1647 return -EFAULT;
1648
a8c9486b 1649 rcu_read_lock();
0c12295a 1650 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
132adf54 1651 if (c) {
494fff56
YM		 1652 sr.pktcnt = c->_c.mfc_un.res.pkt;
1653 sr.bytecnt = c->_c.mfc_un.res.bytes;
1654 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
a8c9486b 1655 rcu_read_unlock();
132adf54 1656
c354e124 1657 if (copy_to_user(arg, &sr, sizeof(sr)))
132adf54
SH		 1658 return -EFAULT;
1659 return 0;
1660 }
a8c9486b 1661 rcu_read_unlock();
132adf54
SH		 1662 return -EADDRNOTAVAIL;
1663 default:
1664 return -ENOIOCTLCMD;
1da177e4
LT		 1665 }
1666}
1667
709b46e8
EB		 1668#ifdef CONFIG_COMPAT
1669struct compat_sioc_sg_req {
1670 struct in_addr src;
1671 struct in_addr grp;
1672 compat_ulong_t pktcnt;
1673 compat_ulong_t bytecnt;
1674 compat_ulong_t wrong_if;
1675};
1676
ca6b8bb0
DM		 1677struct compat_sioc_vif_req {
1678 vifi_t vifi; /* Which iface */
1679 compat_ulong_t icount;
1680 compat_ulong_t ocount;
1681 compat_ulong_t ibytes;
1682 compat_ulong_t obytes;
1683};
1684
709b46e8
EB		 1685int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1686{
0033d5ad 1687 struct compat_sioc_sg_req sr;
ca6b8bb0
DM		 1688 struct compat_sioc_vif_req vr;
1689 struct vif_device *vif;
709b46e8
EB		 1690 struct mfc_cache *c;
1691 struct net *net = sock_net(sk);
1692 struct mr_table *mrt;
1693
1694 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
51456b29 1695 if (!mrt)
709b46e8
EB		 1696 return -ENOENT;
1697
1698 switch (cmd) {
ca6b8bb0
DM		 1699 case SIOCGETVIFCNT:
1700 if (copy_from_user(&vr, arg, sizeof(vr)))
1701 return -EFAULT;
1702 if (vr.vifi >= mrt->maxvif)
1703 return -EINVAL;
5648451e 1704 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
ca6b8bb0
DM		 1705 read_lock(&mrt_lock);
1706 vif = &mrt->vif_table[vr.vifi];
1707 if (VIF_EXISTS(mrt, vr.vifi)) {
1708 vr.icount = vif->pkt_in;
1709 vr.ocount = vif->pkt_out;
1710 vr.ibytes = vif->bytes_in;
1711 vr.obytes = vif->bytes_out;
1712 read_unlock(&mrt_lock);
1713
1714 if (copy_to_user(arg, &vr, sizeof(vr)))
1715 return -EFAULT;
1716 return 0;
1717 }
1718 read_unlock(&mrt_lock);
1719 return -EADDRNOTAVAIL;
709b46e8
EB		 1720 case SIOCGETSGCNT:
1721 if (copy_from_user(&sr, arg, sizeof(sr)))
1722 return -EFAULT;
1723
1724 rcu_read_lock();
1725 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1726 if (c) {
494fff56
YM		 1727 sr.pktcnt = c->_c.mfc_un.res.pkt;
1728 sr.bytecnt = c->_c.mfc_un.res.bytes;
1729 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
709b46e8
EB		 1730 rcu_read_unlock();
1731
1732 if (copy_to_user(arg, &sr, sizeof(sr)))
1733 return -EFAULT;
1734 return 0;
1735 }
1736 rcu_read_unlock();
1737 return -EADDRNOTAVAIL;
1738 default:
1739 return -ENOIOCTLCMD;
1740 }
1741}
1742#endif
1743
1da177e4
LT		 1744static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1745{
351638e7 1746 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4feb88e5 1747 struct net *net = dev_net(dev);
f0ad0860 1748 struct mr_table *mrt;
1da177e4
LT		 1749 struct vif_device *v;
1750 int ct;
e9dc8653 1751
1da177e4
LT		 1752 if (event != NETDEV_UNREGISTER)
1753 return NOTIFY_DONE;
f0ad0860
PM		 1754
1755 ipmr_for_each_table(mrt, net) {
1756 v = &mrt->vif_table[0];
1757 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1758 if (v->dev == dev)
e92036a6 1759 vif_delete(mrt, ct, 1, NULL);
f0ad0860 1760 }
1da177e4
LT		 1761 }
1762 return NOTIFY_DONE;
1763}
1764
c354e124 1765static struct notifier_block ip_mr_notifier = {
1da177e4
LT		 1766 .notifier_call = ipmr_device_event,
1767};
1768
7ef8f65d
NA		 1769/* Encapsulate a packet by attaching a valid IPIP header to it.
1770 * This avoids tunnel drivers and other mess and gives us the speed so
1771 * important for multicast video.
1da177e4 1772 */
b6a7719a
HFS		 1773static void ip_encap(struct net *net, struct sk_buff *skb,
1774 __be32 saddr, __be32 daddr)
1da177e4 1775{
8856dfa3 1776 struct iphdr *iph;
b71d1d42 1777 const struct iphdr *old_iph = ip_hdr(skb);
8856dfa3
ACM		 1778
1779 skb_push(skb, sizeof(struct iphdr));
b0e380b1 1780 skb->transport_header = skb->network_header;
8856dfa3 1781 skb_reset_network_header(skb);
eddc9ec5 1782 iph = ip_hdr(skb);
1da177e4 1783
a8cb16dd 1784 iph->version = 4;
e023dd64
ACM		 1785 iph->tos = old_iph->tos;
1786 iph->ttl = old_iph->ttl;
1da177e4
LT		 1787 iph->frag_off = 0;
1788 iph->daddr = daddr;
1789 iph->saddr = saddr;
1790 iph->protocol = IPPROTO_IPIP;
1791 iph->ihl = 5;
1792 iph->tot_len = htons(skb->len);
b6a7719a 1793 ip_select_ident(net, skb, NULL);
1da177e4
LT		 1794 ip_send_check(iph);
1795
1da177e4
LT		 1796 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1797 nf_reset(skb);
1798}
1799
0c4b51f0
EB		 1800static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1801 struct sk_buff *skb)
1da177e4 1802{
a8cb16dd 1803 struct ip_options *opt = &(IPCB(skb)->opt);
1da177e4 1804
73186df8
DM		 1805 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1806 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1da177e4
LT		 1807
1808 if (unlikely(opt->optlen))
1809 ip_forward_options(skb);
1810
13206b6b 1811 return dst_output(net, sk, skb);
1da177e4
LT		 1812}
1813
a5bc9294
YG		 1814#ifdef CONFIG_NET_SWITCHDEV
1815static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1816 int in_vifi, int out_vifi)
1817{
1818 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1819 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1820
875e8939 1821 if (!skb->offload_l3_fwd_mark)
a5bc9294
YG		 1822 return false;
1823 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1824 return false;
1825 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1826 &in_vif->dev_parent_id);
1827}
1828#else
1829static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1830 int in_vifi, int out_vifi)
1831{
1832 return false;
1833}
1834#endif
1835
7ef8f65d 1836/* Processing handlers for ipmr_forward */
1da177e4 1837
0c12295a 1838static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
6e0735d1 1839 int in_vifi, struct sk_buff *skb, int vifi)
1da177e4 1840{
eddc9ec5 1841 const struct iphdr *iph = ip_hdr(skb);
0c12295a 1842 struct vif_device *vif = &mrt->vif_table[vifi];
1da177e4
LT		 1843 struct net_device *dev;
1844 struct rtable *rt;
31e4543d 1845 struct flowi4 fl4;
1da177e4
LT		 1846 int encap = 0;
1847
51456b29 1848 if (!vif->dev)
1da177e4
LT		 1849 goto out_free;
1850
1da177e4
LT		 1851 if (vif->flags & VIFF_REGISTER) {
1852 vif->pkt_out++;
c354e124 1853 vif->bytes_out += skb->len;
cf3677ae
PE		 1854 vif->dev->stats.tx_bytes += skb->len;
1855 vif->dev->stats.tx_packets++;
0c12295a 1856 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
69ebbf58 1857 goto out_free;
1da177e4 1858 }
1da177e4 1859
a5bc9294
YG		 1860 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1861 goto out_free;
1862
a8cb16dd 1863 if (vif->flags & VIFF_TUNNEL) {
31e4543d 1864 rt = ip_route_output_ports(net, &fl4, NULL,
78fbfd8a
DM		 1865 vif->remote, vif->local,
1866 0, 0,
1867 IPPROTO_IPIP,
1868 RT_TOS(iph->tos), vif->link);
b23dd4fe 1869 if (IS_ERR(rt))
1da177e4
LT		 1870 goto out_free;
1871 encap = sizeof(struct iphdr);
1872 } else {
31e4543d 1873 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
78fbfd8a
DM		 1874 0, 0,
1875 IPPROTO_IPIP,
1876 RT_TOS(iph->tos), vif->link);
b23dd4fe 1877 if (IS_ERR(rt))
1da177e4
LT		 1878 goto out_free;
1879 }
1880
d8d1f30b 1881 dev = rt->dst.dev;
1da177e4 1882
d8d1f30b 1883 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1da177e4 1884 /* Do not fragment multicasts. Alas, IPv4 does not
a8cb16dd
ED		 1885 * allow to send ICMP, so that packets will disappear
1886 * to blackhole.
1da177e4 1887 */
73186df8 1888 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1da177e4
LT		 1889 ip_rt_put(rt);
1890 goto out_free;
1891 }
1892
d8d1f30b 1893 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1da177e4
LT		 1894
1895 if (skb_cow(skb, encap)) {
e905a9ed 1896 ip_rt_put(rt);
1da177e4
LT		 1897 goto out_free;
1898 }
1899
1900 vif->pkt_out++;
c354e124 1901 vif->bytes_out += skb->len;
1da177e4 1902
adf30907 1903 skb_dst_drop(skb);
d8d1f30b 1904 skb_dst_set(skb, &rt->dst);
eddc9ec5 1905 ip_decrease_ttl(ip_hdr(skb));
1da177e4
LT		 1906
1907 /* FIXME: forward and output firewalls used to be called here.
a8cb16dd
ED		 1908 * What do we do with netfilter? -- RR
1909 */
1da177e4 1910 if (vif->flags & VIFF_TUNNEL) {
b6a7719a 1911 ip_encap(net, skb, vif->local, vif->remote);
1da177e4 1912 /* FIXME: extra output firewall step used to be here. --RR */
2f4c02d4
PE		 1913 vif->dev->stats.tx_packets++;
1914 vif->dev->stats.tx_bytes += skb->len;
1da177e4
LT		 1915 }
1916
9ee6c5dc 1917 IPCB(skb)->flags |= IPSKB_FORWARDED;
1da177e4 1918
7ef8f65d 1919 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1da177e4
LT		 1920 * not only before forwarding, but after forwarding on all output
1921 * interfaces. It is clear, if mrouter runs a multicasting
1922 * program, it should receive packets not depending to what interface
1923 * program is joined.
1924 * If we will not make it, the program will have to join on all
1925 * interfaces. On the other hand, multihoming host (or router, but
1926 * not mrouter) cannot join to more than one interface - it will
1927 * result in receiving multiple packets.
1928 */
29a26a56
EB		 1929 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1930 net, NULL, skb, skb->dev, dev,
1da177e4
LT		 1931 ipmr_forward_finish);
1932 return;
1933
1934out_free:
1935 kfree_skb(skb);
1da177e4
LT		 1936}
1937
0c12295a 1938static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1da177e4
LT		 1939{
1940 int ct;
0c12295a
PM		 1941
1942 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1943 if (mrt->vif_table[ct].dev == dev)
1da177e4
LT		 1944 break;
1945 }
1946 return ct;
1947}
1948
1949/* "local" means that we should preserve one skb (for local delivery) */
c4854ec8 1950static void ip_mr_forward(struct net *net, struct mr_table *mrt,
4b1f0d33 1951 struct net_device *dev, struct sk_buff *skb,
494fff56 1952 struct mfc_cache *c, int local)
1da177e4 1953{
4b1f0d33 1954 int true_vifi = ipmr_find_vif(mrt, dev);
1da177e4
LT		 1955 int psend = -1;
1956 int vif, ct;
1957
494fff56
YM		 1958 vif = c->_c.mfc_parent;
1959 c->_c.mfc_un.res.pkt++;
1960 c->_c.mfc_un.res.bytes += skb->len;
1961 c->_c.mfc_un.res.lastuse = jiffies;
1da177e4 1962
494fff56 1963 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
660b26dc
ND		 1964 struct mfc_cache *cache_proxy;
1965
1966 /* For an (*,G) entry, we only check that the incomming
1967 * interface is part of the static tree.
1968 */
845c9a7a 1969 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
660b26dc 1970 if (cache_proxy &&
494fff56 1971 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
660b26dc
ND		 1972 goto forward;
1973 }
1974
7ef8f65d 1975 /* Wrong interface: drop packet and (maybe) send PIM assert. */
4b1f0d33 1976 if (mrt->vif_table[vif].dev != dev) {
c7537967 1977 if (rt_is_output_route(skb_rtable(skb))) {
1da177e4 1978 /* It is our own packet, looped back.
a8cb16dd
ED		 1979 * Very complicated situation...
1980 *
1981 * The best workaround until routing daemons will be
1982 * fixed is not to redistribute packet, if it was
1983 * send through wrong interface. It means, that
1984 * multicast applications WILL NOT work for
1985 * (S,G), which have default multicast route pointing
1986 * to wrong oif. In any case, it is not a good
1987 * idea to use multicasting applications on router.
1da177e4
LT		 1988 */
1989 goto dont_forward;
1990 }
1991
494fff56 1992 c->_c.mfc_un.res.wrong_if++;
1da177e4 1993
0c12295a 1994 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1da177e4 1995 /* pimsm uses asserts, when switching from RPT to SPT,
a8cb16dd
ED		 1996 * so that we cannot check that packet arrived on an oif.
1997 * It is bad, but otherwise we would need to move pretty
1998 * large chunk of pimd to kernel. Ough... --ANK
1da177e4 1999 */
0c12295a 2000 (mrt->mroute_do_pim ||
494fff56 2001 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
e905a9ed 2002 time_after(jiffies,
494fff56
YM		 2003 c->_c.mfc_un.res.last_assert +
2004 MFC_ASSERT_THRESH)) {
2005 c->_c.mfc_un.res.last_assert = jiffies;
0c12295a 2006 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
c921c207
NA		 2007 if (mrt->mroute_do_wrvifwhole)
2008 ipmr_cache_report(mrt, skb, true_vifi,
2009 IGMPMSG_WRVIFWHOLE);
1da177e4
LT		 2010 }
2011 goto dont_forward;
2012 }
2013
660b26dc 2014forward:
0c12295a
PM		 2015 mrt->vif_table[vif].pkt_in++;
2016 mrt->vif_table[vif].bytes_in += skb->len;
1da177e4 2017
7ef8f65d 2018 /* Forward the frame */
494fff56
YM		 2019 if (c->mfc_origin == htonl(INADDR_ANY) &&
2020 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
660b26dc 2021 if (true_vifi >= 0 &&
494fff56 2022 true_vifi != c->_c.mfc_parent &&
660b26dc 2023 ip_hdr(skb)->ttl >
494fff56 2024 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
660b26dc
ND		 2025 /* It's an (*,*) entry and the packet is not coming from
2026 * the upstream: forward the packet to the upstream
2027 * only.
2028 */
494fff56 2029 psend = c->_c.mfc_parent;
660b26dc
ND		 2030 goto last_forward;
2031 }
2032 goto dont_forward;
2033 }
494fff56
YM		 2034 for (ct = c->_c.mfc_un.res.maxvif - 1;
2035 ct >= c->_c.mfc_un.res.minvif; ct--) {
660b26dc 2036 /* For (*,G) entry, don't forward to the incoming interface */
494fff56 2037 if ((c->mfc_origin != htonl(INADDR_ANY) ||
360eb5da 2038 ct != true_vifi) &&
494fff56 2039 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
1da177e4
LT		 2040 if (psend != -1) {
2041 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
a8cb16dd 2042
1da177e4 2043 if (skb2)
a5bc9294 2044 ipmr_queue_xmit(net, mrt, true_vifi,
6e0735d1 2045 skb2, psend);
1da177e4 2046 }
c354e124 2047 psend = ct;
1da177e4
LT		 2048 }
2049 }
660b26dc 2050last_forward:
1da177e4
LT		 2051 if (psend != -1) {
2052 if (local) {
2053 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
a8cb16dd 2054
1da177e4 2055 if (skb2)
a5bc9294 2056 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
6e0735d1 2057 psend);
1da177e4 2058 } else {
6e0735d1 2059 ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
c4854ec8 2060 return;
1da177e4
LT		 2061 }
2062 }
2063
2064dont_forward:
2065 if (!local)
2066 kfree_skb(skb);
1da177e4
LT		 2067}
2068
417da66f 2069static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
ee3f1aaf 2070{
417da66f
DM		 2071 struct rtable *rt = skb_rtable(skb);
2072 struct iphdr *iph = ip_hdr(skb);
da91981b 2073 struct flowi4 fl4 = {
417da66f
DM		 2074 .daddr = iph->daddr,
2075 .saddr = iph->saddr,
b0fe4a31 2076 .flowi4_tos = RT_TOS(iph->tos),
4fd551d7
DM		 2077 .flowi4_oif = (rt_is_output_route(rt) ?
2078 skb->dev->ifindex : 0),
2079 .flowi4_iif = (rt_is_output_route(rt) ?
1fb9489b 2080 LOOPBACK_IFINDEX :
4fd551d7 2081 skb->dev->ifindex),
b4869889 2082 .flowi4_mark = skb->mark,
ee3f1aaf
DM		 2083 };
2084 struct mr_table *mrt;
2085 int err;
2086
da91981b 2087 err = ipmr_fib_lookup(net, &fl4, &mrt);
ee3f1aaf
DM		 2088 if (err)
2089 return ERR_PTR(err);
2090 return mrt;
2091}
1da177e4 2092
7ef8f65d
NA		 2093/* Multicast packets for forwarding arrive here
2094 * Called with rcu_read_lock();
1da177e4 2095 */
1da177e4
LT		 2096int ip_mr_input(struct sk_buff *skb)
2097{
2098 struct mfc_cache *cache;
4feb88e5 2099 struct net *net = dev_net(skb->dev);
511c3f92 2100 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
f0ad0860 2101 struct mr_table *mrt;
bcfc7d33
TW		 2102 struct net_device *dev;
2103
2104 /* skb->dev passed in is the loX master dev for vrfs.
2105 * As there are no vifs associated with loopback devices,
2106 * get the proper interface that does have a vif associated with it.
2107 */
2108 dev = skb->dev;
2109 if (netif_is_l3_master(skb->dev)) {
2110 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2111 if (!dev) {
2112 kfree_skb(skb);
2113 return -ENODEV;
2114 }
2115 }
1da177e4
LT		 2116
2117 /* Packet is looped back after forward, it should not be
a8cb16dd 2118 * forwarded second time, but still can be delivered locally.
1da177e4 2119 */
4c968709 2120 if (IPCB(skb)->flags & IPSKB_FORWARDED)
1da177e4
LT		 2121 goto dont_forward;
2122
417da66f 2123 mrt = ipmr_rt_fib_lookup(net, skb);
ee3f1aaf
DM		 2124 if (IS_ERR(mrt)) {
2125 kfree_skb(skb);
2126 return PTR_ERR(mrt);
e40dbc51 2127 }
1da177e4 2128 if (!local) {
4c968709
ED		 2129 if (IPCB(skb)->opt.router_alert) {
2130 if (ip_call_ra_chain(skb))
2131 return 0;
2132 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2133 /* IGMPv1 (and broken IGMPv2 implementations sort of
2134 * Cisco IOS <= 11.2(8)) do not put router alert
2135 * option to IGMP packets destined to routable
2136 * groups. It is very bad, because it means
2137 * that we can forward NO IGMP messages.
2138 */
2139 struct sock *mroute_sk;
2140
2141 mroute_sk = rcu_dereference(mrt->mroute_sk);
2142 if (mroute_sk) {
2143 nf_reset(skb);
2144 raw_rcv(mroute_sk, skb);
2145 return 0;
2146 }
1da177e4
LT		 2147 }
2148 }
2149
a8c9486b 2150 /* already under rcu_read_lock() */
0c12295a 2151 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
51456b29 2152 if (!cache) {
bcfc7d33 2153 int vif = ipmr_find_vif(mrt, dev);
660b26dc
ND		 2154
2155 if (vif >= 0)
2156 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2157 vif);
2158 }
1da177e4 2159
7ef8f65d 2160 /* No usable cache entry */
51456b29 2161 if (!cache) {
1da177e4
LT		 2162 int vif;
2163
2164 if (local) {
2165 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2166 ip_local_deliver(skb);
51456b29 2167 if (!skb2)
1da177e4 2168 return -ENOBUFS;
1da177e4
LT		 2169 skb = skb2;
2170 }
2171
a8c9486b 2172 read_lock(&mrt_lock);
bcfc7d33 2173 vif = ipmr_find_vif(mrt, dev);
1da177e4 2174 if (vif >= 0) {
4b1f0d33 2175 int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
1da177e4
LT		 2176 read_unlock(&mrt_lock);
2177
0eae88f3 2178 return err2;
1da177e4
LT		 2179 }
2180 read_unlock(&mrt_lock);
2181 kfree_skb(skb);
2182 return -ENODEV;
2183 }
2184
a8c9486b 2185 read_lock(&mrt_lock);
4b1f0d33 2186 ip_mr_forward(net, mrt, dev, skb, cache, local);
1da177e4
LT		 2187 read_unlock(&mrt_lock);
2188
2189 if (local)
2190 return ip_local_deliver(skb);
2191
2192 return 0;
2193
2194dont_forward:
2195 if (local)
2196 return ip_local_deliver(skb);
2197 kfree_skb(skb);
2198 return 0;
2199}
2200
b1879204 2201#ifdef CONFIG_IP_PIMSM_V1
7ef8f65d 2202/* Handle IGMP messages of PIMv1 */
a8cb16dd 2203int pim_rcv_v1(struct sk_buff *skb)
b1879204
IJ		 2204{
2205 struct igmphdr *pim;
4feb88e5 2206 struct net *net = dev_net(skb->dev);
f0ad0860 2207 struct mr_table *mrt;
b1879204
IJ		 2208
2209 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2210 goto drop;
2211
2212 pim = igmp_hdr(skb);
2213
417da66f 2214 mrt = ipmr_rt_fib_lookup(net, skb);
ee3f1aaf
DM		 2215 if (IS_ERR(mrt))
2216 goto drop;
0c12295a 2217 if (!mrt->mroute_do_pim ||
b1879204
IJ		 2218 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2219 goto drop;
2220
f0ad0860 2221 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
b1879204
IJ		 2222drop:
2223 kfree_skb(skb);
2224 }
1da177e4
LT		 2225 return 0;
2226}
2227#endif
2228
2229#ifdef CONFIG_IP_PIMSM_V2
a8cb16dd 2230static int pim_rcv(struct sk_buff *skb)
1da177e4
LT		 2231{
2232 struct pimreghdr *pim;
f0ad0860
PM		 2233 struct net *net = dev_net(skb->dev);
2234 struct mr_table *mrt;
1da177e4 2235
b1879204 2236 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
1da177e4
LT		 2237 goto drop;
2238
9c70220b 2239 pim = (struct pimreghdr *)skb_transport_header(skb);
56245cae 2240 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
a8cb16dd 2241 (pim->flags & PIM_NULL_REGISTER) ||
e905a9ed 2242 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
d3bc23e7 2243 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
1da177e4
LT		 2244 goto drop;
2245
417da66f 2246 mrt = ipmr_rt_fib_lookup(net, skb);
ee3f1aaf
DM		 2247 if (IS_ERR(mrt))
2248 goto drop;
f0ad0860 2249 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
b1879204
IJ		 2250drop:
2251 kfree_skb(skb);
2252 }
1da177e4
LT		 2253 return 0;
2254}
2255#endif
2256
9a1b9496
DM		 2257int ipmr_get_route(struct net *net, struct sk_buff *skb,
2258 __be32 saddr, __be32 daddr,
9f09eaea 2259 struct rtmsg *rtm, u32 portid)
1da177e4 2260{
1da177e4 2261 struct mfc_cache *cache;
9a1b9496
DM		 2262 struct mr_table *mrt;
2263 int err;
1da177e4 2264
f0ad0860 2265 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
51456b29 2266 if (!mrt)
f0ad0860
PM		 2267 return -ENOENT;
2268
a8c9486b 2269 rcu_read_lock();
9a1b9496 2270 cache = ipmr_cache_find(mrt, saddr, daddr);
51456b29 2271 if (!cache && skb->dev) {
660b26dc 2272 int vif = ipmr_find_vif(mrt, skb->dev);
1da177e4 2273
660b26dc
ND		 2274 if (vif >= 0)
2275 cache = ipmr_cache_find_any(mrt, daddr, vif);
2276 }
51456b29 2277 if (!cache) {
72287490 2278 struct sk_buff *skb2;
eddc9ec5 2279 struct iphdr *iph;
1da177e4 2280 struct net_device *dev;
a8cb16dd 2281 int vif = -1;
1da177e4 2282
1da177e4 2283 dev = skb->dev;
a8c9486b 2284 read_lock(&mrt_lock);
a8cb16dd
ED		 2285 if (dev)
2286 vif = ipmr_find_vif(mrt, dev);
2287 if (vif < 0) {
1da177e4 2288 read_unlock(&mrt_lock);
a8c9486b 2289 rcu_read_unlock();
1da177e4
LT		 2290 return -ENODEV;
2291 }
72287490
AK		 2292 skb2 = skb_clone(skb, GFP_ATOMIC);
2293 if (!skb2) {
2294 read_unlock(&mrt_lock);
a8c9486b 2295 rcu_read_unlock();
72287490
AK		 2296 return -ENOMEM;
2297 }
2298
2cf75070 2299 NETLINK_CB(skb2).portid = portid;
e2d1bca7
ACM		 2300 skb_push(skb2, sizeof(struct iphdr));
2301 skb_reset_network_header(skb2);
eddc9ec5
ACM		 2302 iph = ip_hdr(skb2);
2303 iph->ihl = sizeof(struct iphdr) >> 2;
9a1b9496
DM		 2304 iph->saddr = saddr;
2305 iph->daddr = daddr;
eddc9ec5 2306 iph->version = 0;
4b1f0d33 2307 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
1da177e4 2308 read_unlock(&mrt_lock);
a8c9486b 2309 rcu_read_unlock();
1da177e4
LT		 2310 return err;
2311 }
2312
a8c9486b 2313 read_lock(&mrt_lock);
7b0db857 2314 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
1da177e4 2315 read_unlock(&mrt_lock);
a8c9486b 2316 rcu_read_unlock();
1da177e4
LT		 2317 return err;
2318}
2319
cb6a4e46 2320static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
65886f43
ND		 2321 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2322 int flags)
cb6a4e46
PM		 2323{
2324 struct nlmsghdr *nlh;
2325 struct rtmsg *rtm;
1eb99af5 2326 int err;
cb6a4e46 2327
65886f43 2328 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
51456b29 2329 if (!nlh)
cb6a4e46
PM		 2330 return -EMSGSIZE;
2331
2332 rtm = nlmsg_data(nlh);
2333 rtm->rtm_family = RTNL_FAMILY_IPMR;
2334 rtm->rtm_dst_len = 32;
2335 rtm->rtm_src_len = 32;
2336 rtm->rtm_tos = 0;
2337 rtm->rtm_table = mrt->id;
f3756b79
DM		 2338 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2339 goto nla_put_failure;
cb6a4e46
PM		 2340 rtm->rtm_type = RTN_MULTICAST;
2341 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
494fff56 2342 if (c->_c.mfc_flags & MFC_STATIC)
9a68ac72
ND		 2343 rtm->rtm_protocol = RTPROT_STATIC;
2344 else
2345 rtm->rtm_protocol = RTPROT_MROUTED;
cb6a4e46
PM		 2346 rtm->rtm_flags = 0;
2347
930345ea
JB		 2348 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2349 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
f3756b79 2350 goto nla_put_failure;
7b0db857 2351 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
1eb99af5
ND		 2352 /* do not break the dump if cache is unresolved */
2353 if (err < 0 && err != -ENOENT)
cb6a4e46
PM		 2354 goto nla_put_failure;
2355
053c095a
JB		 2356 nlmsg_end(skb, nlh);
2357 return 0;
cb6a4e46
PM		 2358
2359nla_put_failure:
2360 nlmsg_cancel(skb, nlh);
2361 return -EMSGSIZE;
2362}
2363
7b0db857
YM		 2364static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2365 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2366 int flags)
2367{
2368 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2369 cmd, flags);
2370}
2371
8cd3ac9f
ND		 2372static size_t mroute_msgsize(bool unresolved, int maxvif)
2373{
2374 size_t len =
2375 NLMSG_ALIGN(sizeof(struct rtmsg))
2376 + nla_total_size(4) /* RTA_TABLE */
2377 + nla_total_size(4) /* RTA_SRC */
2378 + nla_total_size(4) /* RTA_DST */
2379 ;
2380
2381 if (!unresolved)
2382 len = len
2383 + nla_total_size(4) /* RTA_IIF */
2384 + nla_total_size(0) /* RTA_MULTIPATH */
2385 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2386 /* RTA_MFC_STATS */
a9a08042 2387 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
8cd3ac9f
ND		 2388 ;
2389
2390 return len;
2391}
2392
2393static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2394 int cmd)
2395{
2396 struct net *net = read_pnet(&mrt->net);
2397 struct sk_buff *skb;
2398 int err = -ENOBUFS;
2399
494fff56
YM		 2400 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2401 mrt->maxvif),
8cd3ac9f 2402 GFP_ATOMIC);
51456b29 2403 if (!skb)
8cd3ac9f
ND		 2404 goto errout;
2405
65886f43 2406 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
8cd3ac9f
ND		 2407 if (err < 0)
2408 goto errout;
2409
2410 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2411 return;
2412
2413errout:
2414 kfree_skb(skb);
2415 if (err < 0)
2416 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2417}
2418
5a645dd8
JG		 2419static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2420{
2421 size_t len =
2422 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2423 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2424 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2425 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2426 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2427 /* IPMRA_CREPORT_PKT */
2428 + nla_total_size(payloadlen)
2429 ;
2430
2431 return len;
2432}
2433
2434static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2435{
2436 struct net *net = read_pnet(&mrt->net);
2437 struct nlmsghdr *nlh;
2438 struct rtgenmsg *rtgenm;
2439 struct igmpmsg *msg;
2440 struct sk_buff *skb;
2441 struct nlattr *nla;
2442 int payloadlen;
2443
2444 payloadlen = pkt->len - sizeof(struct igmpmsg);
2445 msg = (struct igmpmsg *)skb_network_header(pkt);
2446
2447 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2448 if (!skb)
2449 goto errout;
2450
2451 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2452 sizeof(struct rtgenmsg), 0);
2453 if (!nlh)
2454 goto errout;
2455 rtgenm = nlmsg_data(nlh);
2456 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2457 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2458 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif) ||
2459 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2460 msg->im_src.s_addr) ||
2461 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2462 msg->im_dst.s_addr))
2463 goto nla_put_failure;
2464
2465 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2466 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2467 nla_data(nla), payloadlen))
2468 goto nla_put_failure;
2469
2470 nlmsg_end(skb, nlh);
2471
2472 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2473 return;
2474
2475nla_put_failure:
2476 nlmsg_cancel(skb, nlh);
2477errout:
2478 kfree_skb(skb);
2479 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2480}
2481
d0440029
JK		 2482static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2483 const struct nlmsghdr *nlh,
2484 struct nlattr **tb,
2485 struct netlink_ext_ack *extack)
2486{
2487 struct rtmsg *rtm;
2488 int i, err;
2489
2490 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2491 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2492 return -EINVAL;
2493 }
2494
2495 if (!netlink_strict_get_check(skb))
8cb08174
JB		 2496 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2497 rtm_ipv4_policy, extack);
d0440029
JK		 2498
2499 rtm = nlmsg_data(nlh);
2500 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2501 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2502 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2503 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2504 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2505 return -EINVAL;
2506 }
2507
8cb08174
JB		 2508 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2509 rtm_ipv4_policy, extack);
d0440029
JK		 2510 if (err)
2511 return err;
2512
2513 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2514 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2515 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2516 return -EINVAL;
2517 }
2518
2519 for (i = 0; i <= RTA_MAX; i++) {
2520 if (!tb[i])
2521 continue;
2522
2523 switch (i) {
2524 case RTA_SRC:
2525 case RTA_DST:
2526 case RTA_TABLE:
2527 break;
2528 default:
2529 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2530 return -EINVAL;
2531 }
2532 }
2533
2534 return 0;
2535}
2536
4f75ba69
DS		 2537static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2538 struct netlink_ext_ack *extack)
2539{
2540 struct net *net = sock_net(in_skb->sk);
2541 struct nlattr *tb[RTA_MAX + 1];
2542 struct sk_buff *skb = NULL;
2543 struct mfc_cache *cache;
2544 struct mr_table *mrt;
4f75ba69
DS		 2545 __be32 src, grp;
2546 u32 tableid;
2547 int err;
2548
d0440029 2549 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
4f75ba69
DS		 2550 if (err < 0)
2551 goto errout;
2552
4f75ba69
DS		 2553 src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2554 grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2555 tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2556
2557 mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2e3d232e
DC		 2558 if (!mrt) {
2559 err = -ENOENT;
4f75ba69
DS		 2560 goto errout_free;
2561 }
2562
2563 /* entries are added/deleted only under RTNL */
2564 rcu_read_lock();
2565 cache = ipmr_cache_find(mrt, src, grp);
2566 rcu_read_unlock();
2567 if (!cache) {
2568 err = -ENOENT;
2569 goto errout_free;
2570 }
2571
2572 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2573 if (!skb) {
2574 err = -ENOBUFS;
2575 goto errout_free;
2576 }
2577
2578 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2579 nlh->nlmsg_seq, cache,
2580 RTM_NEWROUTE, 0);
2581 if (err < 0)
2582 goto errout_free;
2583
2584 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2585
2586errout:
2587 return err;
2588
2589errout_free:
2590 kfree_skb(skb);
2591 goto errout;
2592}
2593
cb6a4e46
PM		 2594static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2595{
4724676d 2596 struct fib_dump_filter filter = {};
cb167893 2597 int err;
4724676d 2598
e8ba330a 2599 if (cb->strict_check) {
4724676d 2600 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
effe6792 2601 &filter, cb);
e8ba330a
DA		 2602 if (err < 0)
2603 return err;
2604 }
2605
cb167893
DA		 2606 if (filter.table_id) {
2607 struct mr_table *mrt;
2608
2609 mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2610 if (!mrt) {
ae677bbb
DA		 2611 if (filter.dump_all_families)
2612 return skb->len;
2613
cb167893
DA		 2614 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2615 return -ENOENT;
2616 }
2617 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2618 &mfc_unres_lock, &filter);
2619 return skb->len ? : err;
2620 }
2621
7b0db857 2622 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
cb167893 2623 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
cb6a4e46
PM		 2624}
2625
ccbb0aa6
NA		 2626static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2627 [RTA_SRC] = { .type = NLA_U32 },
2628 [RTA_DST] = { .type = NLA_U32 },
2629 [RTA_IIF] = { .type = NLA_U32 },
2630 [RTA_TABLE] = { .type = NLA_U32 },
2631 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2632};
2633
2634static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2635{
2636 switch (rtm_protocol) {
2637 case RTPROT_STATIC:
2638 case RTPROT_MROUTED:
2639 return true;
2640 }
2641 return false;
2642}
2643
2644static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2645{
2646 struct rtnexthop *rtnh = nla_data(nla);
2647 int remaining = nla_len(nla), vifi = 0;
2648
2649 while (rtnh_ok(rtnh, remaining)) {
2650 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2651 if (++vifi == MAXVIFS)
2652 break;
2653 rtnh = rtnh_next(rtnh, &remaining);
2654 }
2655
2656 return remaining > 0 ? -EINVAL : vifi;
2657}
2658
2659/* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2660static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2661 struct mfcctl *mfcc, int *mrtsock,
c21ef3e3
DA		 2662 struct mr_table **mrtret,
2663 struct netlink_ext_ack *extack)
ccbb0aa6
NA		 2664{
2665 struct net_device *dev = NULL;
2666 u32 tblid = RT_TABLE_DEFAULT;
2667 struct mr_table *mrt;
2668 struct nlattr *attr;
2669 struct rtmsg *rtm;
2670 int ret, rem;
2671
8cb08174
JB		 2672 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2673 rtm_ipmr_policy, extack);
ccbb0aa6
NA		 2674 if (ret < 0)
2675 goto out;
2676 rtm = nlmsg_data(nlh);
2677
2678 ret = -EINVAL;
2679 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2680 rtm->rtm_type != RTN_MULTICAST ||
2681 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2682 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2683 goto out;
2684
2685 memset(mfcc, 0, sizeof(*mfcc));
2686 mfcc->mfcc_parent = -1;
2687 ret = 0;
2688 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2689 switch (nla_type(attr)) {
2690 case RTA_SRC:
2691 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2692 break;
2693 case RTA_DST:
2694 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2695 break;
2696 case RTA_IIF:
2697 dev = __dev_get_by_index(net, nla_get_u32(attr));
2698 if (!dev) {
2699 ret = -ENODEV;
2700 goto out;
2701 }
2702 break;
2703 case RTA_MULTIPATH:
2704 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2705 ret = -EINVAL;
2706 goto out;
2707 }
2708 break;
2709 case RTA_PREFSRC:
2710 ret = 1;
2711 break;
2712 case RTA_TABLE:
2713 tblid = nla_get_u32(attr);
2714 break;
2715 }
2716 }
2717 mrt = ipmr_get_table(net, tblid);
2718 if (!mrt) {
2719 ret = -ENOENT;
2720 goto out;
2721 }
2722 *mrtret = mrt;
2723 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2724 if (dev)
2725 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2726
2727out:
2728 return ret;
2729}
2730
2731/* takes care of both newroute and delroute */
c21ef3e3
DA		 2732static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2733 struct netlink_ext_ack *extack)
ccbb0aa6
NA		 2734{
2735 struct net *net = sock_net(skb->sk);
2736 int ret, mrtsock, parent;
2737 struct mr_table *tbl;
2738 struct mfcctl mfcc;
2739
2740 mrtsock = 0;
2741 tbl = NULL;
c21ef3e3 2742 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
ccbb0aa6
NA		 2743 if (ret < 0)
2744 return ret;
2745
2746 parent = ret ? mfcc.mfcc_parent : -1;
2747 if (nlh->nlmsg_type == RTM_NEWROUTE)
2748 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2749 else
2750 return ipmr_mfc_delete(tbl, &mfcc, parent);
2751}
2752
772c344d
NA		 2753static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2754{
2755 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2756
2757 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2758 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2759 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2760 mrt->mroute_reg_vif_num) ||
2761 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2762 mrt->mroute_do_assert) ||
c921c207
NA		 2763 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2764 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2765 mrt->mroute_do_wrvifwhole))
772c344d
NA		 2766 return false;
2767
2768 return true;
2769}
2770
2771static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2772{
2773 struct nlattr *vif_nest;
2774 struct vif_device *vif;
2775
2776 /* if the VIF doesn't exist just continue */
2777 if (!VIF_EXISTS(mrt, vifid))
2778 return true;
2779
2780 vif = &mrt->vif_table[vifid];
ae0be8de 2781 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
772c344d
NA		 2782 if (!vif_nest)
2783 return false;
2784 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2785 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2786 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2787 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2788 IPMRA_VIFA_PAD) ||
2789 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2790 IPMRA_VIFA_PAD) ||
2791 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2792 IPMRA_VIFA_PAD) ||
2793 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2794 IPMRA_VIFA_PAD) ||
2795 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2796 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2797 nla_nest_cancel(skb, vif_nest);
2798 return false;
2799 }
2800 nla_nest_end(skb, vif_nest);
2801
2802 return true;
2803}
2804
14fc5bb2
DA		 2805static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2806 struct netlink_ext_ack *extack)
2807{
2808 struct ifinfomsg *ifm;
2809
2810 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2811 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2812 return -EINVAL;
2813 }
2814
2815 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2816 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2817 return -EINVAL;
2818 }
2819
2820 ifm = nlmsg_data(nlh);
2821 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2822 ifm->ifi_change || ifm->ifi_index) {
2823 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2824 return -EINVAL;
2825 }
2826
2827 return 0;
2828}
2829
772c344d
NA		 2830static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2831{
2832 struct net *net = sock_net(skb->sk);
2833 struct nlmsghdr *nlh = NULL;
2834 unsigned int t = 0, s_t;
2835 unsigned int e = 0, s_e;
2836 struct mr_table *mrt;
2837
14fc5bb2
DA		 2838 if (cb->strict_check) {
2839 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2840
2841 if (err < 0)
2842 return err;
2843 }
2844
772c344d
NA		 2845 s_t = cb->args[0];
2846 s_e = cb->args[1];
2847
2848 ipmr_for_each_table(mrt, net) {
2849 struct nlattr *vifs, *af;
2850 struct ifinfomsg *hdr;
2851 u32 i;
2852
2853 if (t < s_t)
2854 goto skip_table;
2855 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2856 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2857 sizeof(*hdr), NLM_F_MULTI);
2858 if (!nlh)
2859 break;
2860
2861 hdr = nlmsg_data(nlh);
2862 memset(hdr, 0, sizeof(*hdr));
2863 hdr->ifi_family = RTNL_FAMILY_IPMR;
2864
ae0be8de 2865 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
772c344d
NA		 2866 if (!af) {
2867 nlmsg_cancel(skb, nlh);
2868 goto out;
2869 }
2870
2871 if (!ipmr_fill_table(mrt, skb)) {
2872 nlmsg_cancel(skb, nlh);
2873 goto out;
2874 }
2875
ae0be8de 2876 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
772c344d
NA		 2877 if (!vifs) {
2878 nla_nest_end(skb, af);
2879 nlmsg_end(skb, nlh);
2880 goto out;
2881 }
2882 for (i = 0; i < mrt->maxvif; i++) {
2883 if (e < s_e)
2884 goto skip_entry;
2885 if (!ipmr_fill_vif(mrt, i, skb)) {
2886 nla_nest_end(skb, vifs);
2887 nla_nest_end(skb, af);
2888 nlmsg_end(skb, nlh);
2889 goto out;
2890 }
2891skip_entry:
2892 e++;
2893 }
2894 s_e = 0;
2895 e = 0;
2896 nla_nest_end(skb, vifs);
2897 nla_nest_end(skb, af);
2898 nlmsg_end(skb, nlh);
2899skip_table:
2900 t++;
2901 }
2902
2903out:
2904 cb->args[1] = e;
2905 cb->args[0] = t;
2906
2907 return skb->len;
2908}
2909
e905a9ed 2910#ifdef CONFIG_PROC_FS
7ef8f65d
NA		 2911/* The /proc interfaces to multicast routing :
2912 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
1da177e4 2913 */
1da177e4
LT		 2914
2915static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
ba93ef74 2916 __acquires(mrt_lock)
1da177e4 2917{
3feda6b4 2918 struct mr_vif_iter *iter = seq->private;
f6bb4514 2919 struct net *net = seq_file_net(seq);
f0ad0860
PM		 2920 struct mr_table *mrt;
2921
2922 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
51456b29 2923 if (!mrt)
f0ad0860
PM		 2924 return ERR_PTR(-ENOENT);
2925
2926 iter->mrt = mrt;
f6bb4514 2927
1da177e4 2928 read_lock(&mrt_lock);
3feda6b4 2929 return mr_vif_seq_start(seq, pos);
1da177e4
LT		 2930}
2931
2932static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
ba93ef74 2933 __releases(mrt_lock)
1da177e4
LT		 2934{
2935 read_unlock(&mrt_lock);
2936}
2937
2938static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2939{
3feda6b4 2940 struct mr_vif_iter *iter = seq->private;
f0ad0860 2941 struct mr_table *mrt = iter->mrt;
f6bb4514 2942
1da177e4 2943 if (v == SEQ_START_TOKEN) {
e905a9ed 2944 seq_puts(seq,
1da177e4
LT		 2945 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2946 } else {
2947 const struct vif_device *vif = v;
6853f21f
YM		 2948 const char *name = vif->dev ?
2949 vif->dev->name : "none";
1da177e4
LT		 2950
2951 seq_printf(seq,
91e6dd82 2952 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
0c12295a 2953 vif - mrt->vif_table,
e905a9ed 2954 name, vif->bytes_in, vif->pkt_in,
1da177e4
LT		 2955 vif->bytes_out, vif->pkt_out,
2956 vif->flags, vif->local, vif->remote);
2957 }
2958 return 0;
2959}
2960
f690808e 2961static const struct seq_operations ipmr_vif_seq_ops = {
1da177e4 2962 .start = ipmr_vif_seq_start,
3feda6b4 2963 .next = mr_vif_seq_next,
1da177e4
LT		 2964 .stop = ipmr_vif_seq_stop,
2965 .show = ipmr_vif_seq_show,
2966};
2967
1da177e4
LT		 2968static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2969{
f6bb4514 2970 struct net *net = seq_file_net(seq);
f0ad0860 2971 struct mr_table *mrt;
f6bb4514 2972
f0ad0860 2973 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
51456b29 2974 if (!mrt)
f0ad0860 2975 return ERR_PTR(-ENOENT);
f6bb4514 2976
c8d61968 2977 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
1da177e4
LT		 2978}
2979
2980static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2981{
2982 int n;
2983
2984 if (v == SEQ_START_TOKEN) {
e905a9ed 2985 seq_puts(seq,
1da177e4
LT		 2986 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2987 } else {
2988 const struct mfc_cache *mfc = v;
c8d61968 2989 const struct mr_mfc_iter *it = seq->private;
f0ad0860 2990 const struct mr_table *mrt = it->mrt;
e905a9ed 2991
0eae88f3
ED		 2992 seq_printf(seq, "%08X %08X %-3hd",
2993 (__force u32) mfc->mfc_mcastgrp,
2994 (__force u32) mfc->mfc_origin,
494fff56 2995 mfc->_c.mfc_parent);
1da177e4 2996
0c12295a 2997 if (it->cache != &mrt->mfc_unres_queue) {
1ea472e2 2998 seq_printf(seq, " %8lu %8lu %8lu",
494fff56
YM		 2999 mfc->_c.mfc_un.res.pkt,
3000 mfc->_c.mfc_un.res.bytes,
3001 mfc->_c.mfc_un.res.wrong_if);
3002 for (n = mfc->_c.mfc_un.res.minvif;
3003 n < mfc->_c.mfc_un.res.maxvif; n++) {
0c12295a 3004 if (VIF_EXISTS(mrt, n) &&
494fff56 3005 mfc->_c.mfc_un.res.ttls[n] < 255)
cf958ae3 3006 seq_printf(seq,
e905a9ed 3007 " %2d:%-3d",
494fff56 3008 n, mfc->_c.mfc_un.res.ttls[n]);
1da177e4 3009 }
1ea472e2
BT		 3010 } else {
3011 /* unresolved mfc_caches don't contain
3012 * pkt, bytes and wrong_if values
3013 */
3014 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
1da177e4
LT		 3015 }
3016 seq_putc(seq, '\n');
3017 }
3018 return 0;
3019}
3020
f690808e 3021static const struct seq_operations ipmr_mfc_seq_ops = {
1da177e4 3022 .start = ipmr_mfc_seq_start,
c8d61968
YM		 3023 .next = mr_mfc_seq_next,
3024 .stop = mr_mfc_seq_stop,
1da177e4
LT		 3025 .show = ipmr_mfc_seq_show,
3026};
e905a9ed 3027#endif
1da177e4
LT		 3028
3029#ifdef CONFIG_IP_PIMSM_V2
32613090 3030static const struct net_protocol pim_protocol = {
1da177e4 3031 .handler = pim_rcv,
403dbb97 3032 .netns_ok = 1,
1da177e4
LT		 3033};
3034#endif
3035
4d65b948
YG		 3036static unsigned int ipmr_seq_read(struct net *net)
3037{
3038 ASSERT_RTNL();
3039
3040 return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3041}
3042
3043static int ipmr_dump(struct net *net, struct notifier_block *nb)
3044{
cdc9f944
YM		 3045 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3046 ipmr_mr_table_iter, &mrt_lock);
4d65b948
YG		 3047}
3048
3049static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3050 .family = RTNL_FAMILY_IPMR,
3051 .fib_seq_read = ipmr_seq_read,
3052 .fib_dump = ipmr_dump,
3053 .owner = THIS_MODULE,
3054};
3055
ef739d8a 3056static int __net_init ipmr_notifier_init(struct net *net)
4d65b948
YG		 3057{
3058 struct fib_notifier_ops *ops;
3059
3060 net->ipv4.ipmr_seq = 0;
3061
3062 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3063 if (IS_ERR(ops))
3064 return PTR_ERR(ops);
3065 net->ipv4.ipmr_notifier_ops = ops;
3066
3067 return 0;
3068}
3069
3070static void __net_exit ipmr_notifier_exit(struct net *net)
3071{
3072 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3073 net->ipv4.ipmr_notifier_ops = NULL;
3074}
3075
7ef8f65d 3076/* Setup for IP multicast routing */
cf958ae3
BT		 3077static int __net_init ipmr_net_init(struct net *net)
3078{
f0ad0860 3079 int err;
cf958ae3 3080
4d65b948
YG		 3081 err = ipmr_notifier_init(net);
3082 if (err)
3083 goto ipmr_notifier_fail;
3084
f0ad0860
PM		 3085 err = ipmr_rules_init(net);
3086 if (err < 0)
4d65b948 3087 goto ipmr_rules_fail;
f6bb4514
BT		 3088
3089#ifdef CONFIG_PROC_FS
3090 err = -ENOMEM;
c3506372
CH		 3091 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3092 sizeof(struct mr_vif_iter)))
f6bb4514 3093 goto proc_vif_fail;
c3506372
CH		 3094 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3095 sizeof(struct mr_mfc_iter)))
f6bb4514
BT		 3096 goto proc_cache_fail;
3097#endif
2bb8b26c
BT		 3098 return 0;
3099
f6bb4514
BT		 3100#ifdef CONFIG_PROC_FS
3101proc_cache_fail:
ece31ffd 3102 remove_proc_entry("ip_mr_vif", net->proc_net);
f6bb4514 3103proc_vif_fail:
f0ad0860 3104 ipmr_rules_exit(net);
f6bb4514 3105#endif
4d65b948
YG		 3106ipmr_rules_fail:
3107 ipmr_notifier_exit(net);
3108ipmr_notifier_fail:
cf958ae3
BT		 3109 return err;
3110}
3111
3112static void __net_exit ipmr_net_exit(struct net *net)
3113{
f6bb4514 3114#ifdef CONFIG_PROC_FS
ece31ffd
G		 3115 remove_proc_entry("ip_mr_cache", net->proc_net);
3116 remove_proc_entry("ip_mr_vif", net->proc_net);
f6bb4514 3117#endif
4d65b948 3118 ipmr_notifier_exit(net);
f0ad0860 3119 ipmr_rules_exit(net);
cf958ae3
BT		 3120}
3121
3122static struct pernet_operations ipmr_net_ops = {
3123 .init = ipmr_net_init,
3124 .exit = ipmr_net_exit,
3125};
e905a9ed 3126
03d2f897 3127int __init ip_mr_init(void)
1da177e4 3128{
03d2f897
WC		 3129 int err;
3130
1da177e4
LT		 3131 mrt_cachep = kmem_cache_create("ip_mrt_cache",
3132 sizeof(struct mfc_cache),
a8c9486b 3133 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
20c2df83 3134 NULL);
03d2f897 3135
cf958ae3
BT		 3136 err = register_pernet_subsys(&ipmr_net_ops);
3137 if (err)
3138 goto reg_pernet_fail;
3139
03d2f897
WC		 3140 err = register_netdevice_notifier(&ip_mr_notifier);
3141 if (err)
3142 goto reg_notif_fail;
403dbb97
TG		 3143#ifdef CONFIG_IP_PIMSM_V2
3144 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
058bd4d2 3145 pr_err("%s: can't add PIM protocol\n", __func__);
403dbb97
TG		 3146 err = -EAGAIN;
3147 goto add_proto_fail;
3148 }
3149#endif
c7ac8679 3150 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
b97bac64 3151 ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
ccbb0aa6 3152 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
b97bac64 3153 ipmr_rtm_route, NULL, 0);
ccbb0aa6 3154 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
b97bac64 3155 ipmr_rtm_route, NULL, 0);
772c344d
NA		 3156
3157 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
b97bac64 3158 NULL, ipmr_rtm_dumplink, 0);
03d2f897 3159 return 0;
f6bb4514 3160
403dbb97
TG		 3161#ifdef CONFIG_IP_PIMSM_V2
3162add_proto_fail:
3163 unregister_netdevice_notifier(&ip_mr_notifier);
3164#endif
c3e38896 3165reg_notif_fail:
cf958ae3
BT		 3166 unregister_pernet_subsys(&ipmr_net_ops);
3167reg_pernet_fail:
c3e38896 3168 kmem_cache_destroy(mrt_cachep);
03d2f897 3169 return err;
1da177e4 3170}
Ubuntu Jammy Linux kernel mirror