]> git.ipfire.org Git - thirdparty/linux.git/blame - net/ipv4/ipmr.c
projects / thirdparty / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'io_uring-5.7-2020-05-22' of git://git.kernel.dk/linux-block
[thirdparty/linux.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
7013908c
AG		 112#define ipmr_for_each_table(mrt, net) \
113 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list, \
114 lockdep_rtnl_is_held() || \
115 list_empty(&net->ipv4.mr_tables))
f0ad0860 116
7b0db857
YM		 117static struct mr_table *ipmr_mr_table_iter(struct net *net,
118 struct mr_table *mrt)
119{
120 struct mr_table *ret;
121
122 if (!mrt)
123 ret = list_entry_rcu(net->ipv4.mr_tables.next,
124 struct mr_table, list);
125 else
126 ret = list_entry_rcu(mrt->list.next,
127 struct mr_table, list);
128
129 if (&ret->list == &net->ipv4.mr_tables)
130 return NULL;
131 return ret;
132}
133
f0ad0860
PM		 134static struct mr_table *ipmr_get_table(struct net *net, u32 id)
135{
136 struct mr_table *mrt;
137
138 ipmr_for_each_table(mrt, net) {
139 if (mrt->id == id)
140 return mrt;
141 }
142 return NULL;
143}
144
da91981b 145static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
f0ad0860
PM		 146 struct mr_table **mrt)
147{
f0ad0860 148 int err;
95f4a45d
HFS		 149 struct ipmr_result res;
150 struct fib_lookup_arg arg = {
151 .result = &res,
152 .flags = FIB_LOOKUP_NOREF,
153 };
f0ad0860 154
e58e4159
DA		 155 /* update flow if oif or iif point to device enslaved to l3mdev */
156 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
157
da91981b
DM		 158 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
159 flowi4_to_flowi(flp4), 0, &arg);
f0ad0860
PM		 160 if (err < 0)
161 return err;
162 *mrt = res.mrt;
163 return 0;
164}
165
166static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
167 int flags, struct fib_lookup_arg *arg)
168{
169 struct ipmr_result *res = arg->result;
170 struct mr_table *mrt;
1da177e4 171
f0ad0860
PM		 172 switch (rule->action) {
173 case FR_ACT_TO_TBL:
174 break;
175 case FR_ACT_UNREACHABLE:
176 return -ENETUNREACH;
177 case FR_ACT_PROHIBIT:
178 return -EACCES;
179 case FR_ACT_BLACKHOLE:
180 default:
181 return -EINVAL;
182 }
183
e58e4159
DA		 184 arg->table = fib_rule_get_table(rule, arg);
185
186 mrt = ipmr_get_table(rule->fr_net, arg->table);
51456b29 187 if (!mrt)
f0ad0860
PM		 188 return -EAGAIN;
189 res->mrt = mrt;
190 return 0;
191}
192
193static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
194{
195 return 1;
196}
197
198static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
199 FRA_GENERIC_POLICY,
200};
201
202static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
b16fb418
RP		 203 struct fib_rule_hdr *frh, struct nlattr **tb,
204 struct netlink_ext_ack *extack)
f0ad0860
PM		 205{
206 return 0;
207}
208
209static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
210 struct nlattr **tb)
211{
212 return 1;
213}
214
215static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
216 struct fib_rule_hdr *frh)
217{
218 frh->dst_len = 0;
219 frh->src_len = 0;
220 frh->tos = 0;
221 return 0;
222}
223
04a6f82c 224static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
25239cee 225 .family = RTNL_FAMILY_IPMR,
f0ad0860
PM		 226 .rule_size = sizeof(struct ipmr_rule),
227 .addr_size = sizeof(u32),
228 .action = ipmr_rule_action,
229 .match = ipmr_rule_match,
230 .configure = ipmr_rule_configure,
231 .compare = ipmr_rule_compare,
f0ad0860
PM		 232 .fill = ipmr_rule_fill,
233 .nlgroup = RTNLGRP_IPV4_RULE,
234 .policy = ipmr_rule_policy,
235 .owner = THIS_MODULE,
236};
237
238static int __net_init ipmr_rules_init(struct net *net)
239{
240 struct fib_rules_ops *ops;
241 struct mr_table *mrt;
242 int err;
243
244 ops = fib_rules_register(&ipmr_rules_ops_template, net);
245 if (IS_ERR(ops))
246 return PTR_ERR(ops);
247
248 INIT_LIST_HEAD(&net->ipv4.mr_tables);
249
250 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
1113ebbc
NA		 251 if (IS_ERR(mrt)) {
252 err = PTR_ERR(mrt);
f0ad0860
PM		 253 goto err1;
254 }
255
256 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
257 if (err < 0)
258 goto err2;
259
260 net->ipv4.mr_rules_ops = ops;
261 return 0;
262
263err2:
f243e5a7 264 ipmr_free_table(mrt);
f0ad0860
PM		 265err1:
266 fib_rules_unregister(ops);
267 return err;
268}
269
270static void __net_exit ipmr_rules_exit(struct net *net)
271{
272 struct mr_table *mrt, *next;
273
ed785309 274 rtnl_lock();
035320d5
ED		 275 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
276 list_del(&mrt->list);
acbb219d 277 ipmr_free_table(mrt);
035320d5 278 }
f0ad0860 279 fib_rules_unregister(net->ipv4.mr_rules_ops);
419df12f 280 rtnl_unlock();
f0ad0860 281}
4d65b948 282
b7a59557
JP		 283static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
284 struct netlink_ext_ack *extack)
4d65b948 285{
b7a59557 286 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
4d65b948
YG		 287}
288
289static unsigned int ipmr_rules_seq_read(struct net *net)
290{
291 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
292}
478e4c2f
YG		 293
294bool ipmr_rule_default(const struct fib_rule *rule)
295{
296 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
297}
298EXPORT_SYMBOL(ipmr_rule_default);
f0ad0860
PM		 299#else
300#define ipmr_for_each_table(mrt, net) \
301 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
302
7b0db857
YM		 303static struct mr_table *ipmr_mr_table_iter(struct net *net,
304 struct mr_table *mrt)
305{
306 if (!mrt)
307 return net->ipv4.mrt;
308 return NULL;
309}
310
f0ad0860
PM		 311static struct mr_table *ipmr_get_table(struct net *net, u32 id)
312{
313 return net->ipv4.mrt;
314}
315
da91981b 316static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
f0ad0860
PM		 317 struct mr_table **mrt)
318{
319 *mrt = net->ipv4.mrt;
320 return 0;
321}
322
323static int __net_init ipmr_rules_init(struct net *net)
324{
1113ebbc
NA		 325 struct mr_table *mrt;
326
327 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
328 if (IS_ERR(mrt))
329 return PTR_ERR(mrt);
330 net->ipv4.mrt = mrt;
331 return 0;
f0ad0860
PM		 332}
333
334static void __net_exit ipmr_rules_exit(struct net *net)
335{
ed785309 336 rtnl_lock();
acbb219d 337 ipmr_free_table(net->ipv4.mrt);
ed785309
WC		 338 net->ipv4.mrt = NULL;
339 rtnl_unlock();
f0ad0860 340}
4d65b948 341
b7a59557
JP		 342static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
343 struct netlink_ext_ack *extack)
4d65b948
YG		 344{
345 return 0;
346}
347
348static unsigned int ipmr_rules_seq_read(struct net *net)
349{
350 return 0;
351}
478e4c2f
YG		 352
353bool ipmr_rule_default(const struct fib_rule *rule)
354{
355 return true;
356}
357EXPORT_SYMBOL(ipmr_rule_default);
f0ad0860
PM		 358#endif
359
8fb472c0
NA		 360static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
361 const void *ptr)
362{
363 const struct mfc_cache_cmp_arg *cmparg = arg->key;
364 struct mfc_cache *c = (struct mfc_cache *)ptr;
365
366 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
367 cmparg->mfc_origin != c->mfc_origin;
368}
369
370static const struct rhashtable_params ipmr_rht_params = {
494fff56 371 .head_offset = offsetof(struct mr_mfc, mnode),
8fb472c0
NA		 372 .key_offset = offsetof(struct mfc_cache, cmparg),
373 .key_len = sizeof(struct mfc_cache_cmp_arg),
374 .nelem_hint = 3,
8fb472c0
NA		 375 .obj_cmpfn = ipmr_hash_cmp,
376 .automatic_shrinking = true,
377};
378
0bbbf0e7
YM		 379static void ipmr_new_table_set(struct mr_table *mrt,
380 struct net *net)
381{
382#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
383 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
384#endif
385}
386
845c9a7a
YM		 387static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
388 .mfc_mcastgrp = htonl(INADDR_ANY),
389 .mfc_origin = htonl(INADDR_ANY),
390};
391
392static struct mr_table_ops ipmr_mr_table_ops = {
393 .rht_params = &ipmr_rht_params,
394 .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
395};
396
f0ad0860
PM		 397static struct mr_table *ipmr_new_table(struct net *net, u32 id)
398{
399 struct mr_table *mrt;
1da177e4 400
1113ebbc
NA		 401 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
402 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
403 return ERR_PTR(-EINVAL);
404
f0ad0860 405 mrt = ipmr_get_table(net, id);
00db4124 406 if (mrt)
f0ad0860
PM		 407 return mrt;
408
845c9a7a 409 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
0bbbf0e7 410 ipmr_expire_process, ipmr_new_table_set);
f0ad0860 411}
1da177e4 412
acbb219d
FR		 413static void ipmr_free_table(struct mr_table *mrt)
414{
415 del_timer_sync(&mrt->ipmr_expire_timer);
ca8d4794
CS		 416 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
417 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
8fb472c0 418 rhltable_destroy(&mrt->mfc_hash);
acbb219d
FR		 419 kfree(mrt);
420}
421
1da177e4
LT		 422/* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
423
d607032d
WC		 424static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
425{
4feb88e5
BT		 426 struct net *net = dev_net(dev);
427
d607032d
WC		 428 dev_close(dev);
429
4feb88e5 430 dev = __dev_get_by_name(net, "tunl0");
d607032d 431 if (dev) {
5bc3eb7e 432 const struct net_device_ops *ops = dev->netdev_ops;
d607032d 433 struct ifreq ifr;
d607032d
WC		 434 struct ip_tunnel_parm p;
435
436 memset(&p, 0, sizeof(p));
437 p.iph.daddr = v->vifc_rmt_addr.s_addr;
438 p.iph.saddr = v->vifc_lcl_addr.s_addr;
439 p.iph.version = 4;
440 p.iph.ihl = 5;
441 p.iph.protocol = IPPROTO_IPIP;
442 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
443 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
444
5bc3eb7e
SH		 445 if (ops->ndo_do_ioctl) {
446 mm_segment_t oldfs = get_fs();
447
448 set_fs(KERNEL_DS);
449 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
450 set_fs(oldfs);
451 }
d607032d
WC		 452 }
453}
454
a0b47736
NA		 455/* Initialize ipmr pimreg/tunnel in_device */
456static bool ipmr_init_vif_indev(const struct net_device *dev)
457{
458 struct in_device *in_dev;
459
460 ASSERT_RTNL();
461
462 in_dev = __in_dev_get_rtnl(dev);
463 if (!in_dev)
464 return false;
465 ipv4_devconf_setall(in_dev);
466 neigh_parms_data_state_setall(in_dev->arp_parms);
467 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
468
469 return true;
470}
471
7ef8f65d 472static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
1da177e4
LT		 473{
474 struct net_device *dev;
475
4feb88e5 476 dev = __dev_get_by_name(net, "tunl0");
1da177e4
LT		 477
478 if (dev) {
5bc3eb7e 479 const struct net_device_ops *ops = dev->netdev_ops;
1da177e4
LT		 480 int err;
481 struct ifreq ifr;
1da177e4 482 struct ip_tunnel_parm p;
1da177e4
LT		 483
484 memset(&p, 0, sizeof(p));
485 p.iph.daddr = v->vifc_rmt_addr.s_addr;
486 p.iph.saddr = v->vifc_lcl_addr.s_addr;
487 p.iph.version = 4;
488 p.iph.ihl = 5;
489 p.iph.protocol = IPPROTO_IPIP;
490 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
ba93ef74 491 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
1da177e4 492
5bc3eb7e
SH		 493 if (ops->ndo_do_ioctl) {
494 mm_segment_t oldfs = get_fs();
495
496 set_fs(KERNEL_DS);
497 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
498 set_fs(oldfs);
a8cb16dd 499 } else {
5bc3eb7e 500 err = -EOPNOTSUPP;
a8cb16dd 501 }
1da177e4
LT		 502 dev = NULL;
503
4feb88e5
BT		 504 if (err == 0 &&
505 (dev = __dev_get_by_name(net, p.name)) != NULL) {
1da177e4 506 dev->flags |= IFF_MULTICAST;
a0b47736 507 if (!ipmr_init_vif_indev(dev))
1da177e4 508 goto failure;
00f54e68 509 if (dev_open(dev, NULL))
1da177e4 510 goto failure;
7dc00c82 511 dev_hold(dev);
1da177e4
LT		 512 }
513 }
514 return dev;
515
516failure:
1da177e4
LT		 517 unregister_netdevice(dev);
518 return NULL;
519}
520
c316c629 521#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
6fef4c0c 522static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
1da177e4 523{
4feb88e5 524 struct net *net = dev_net(dev);
f0ad0860 525 struct mr_table *mrt;
da91981b
DM		 526 struct flowi4 fl4 = {
527 .flowi4_oif = dev->ifindex,
6a662719 528 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
da91981b 529 .flowi4_mark = skb->mark,
f0ad0860
PM		 530 };
531 int err;
532
da91981b 533 err = ipmr_fib_lookup(net, &fl4, &mrt);
e40dbc51
BG		 534 if (err < 0) {
535 kfree_skb(skb);
f0ad0860 536 return err;
e40dbc51 537 }
4feb88e5 538
1da177e4 539 read_lock(&mrt_lock);
cf3677ae
PE		 540 dev->stats.tx_bytes += skb->len;
541 dev->stats.tx_packets++;
0c12295a 542 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
1da177e4
LT		 543 read_unlock(&mrt_lock);
544 kfree_skb(skb);
6ed10654 545 return NETDEV_TX_OK;
1da177e4
LT		 546}
547
ee9b9596
ND		 548static int reg_vif_get_iflink(const struct net_device *dev)
549{
550 return 0;
551}
552
007c3838
SH		 553static const struct net_device_ops reg_vif_netdev_ops = {
554 .ndo_start_xmit = reg_vif_xmit,
ee9b9596 555 .ndo_get_iflink = reg_vif_get_iflink,
007c3838
SH		 556};
557
1da177e4
LT		 558static void reg_vif_setup(struct net_device *dev)
559{
560 dev->type = ARPHRD_PIMREG;
46f25dff 561 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
1da177e4 562 dev->flags = IFF_NOARP;
70cb4a45 563 dev->netdev_ops = &reg_vif_netdev_ops;
cf124db5 564 dev->needs_free_netdev = true;
403dbb97 565 dev->features |= NETIF_F_NETNS_LOCAL;
1da177e4
LT		 566}
567
f0ad0860 568static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
1da177e4
LT		 569{
570 struct net_device *dev;
f0ad0860 571 char name[IFNAMSIZ];
1da177e4 572
f0ad0860
PM		 573 if (mrt->id == RT_TABLE_DEFAULT)
574 sprintf(name, "pimreg");
575 else
576 sprintf(name, "pimreg%u", mrt->id);
1da177e4 577
c835a677 578 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
1da177e4 579
51456b29 580 if (!dev)
1da177e4
LT		 581 return NULL;
582
403dbb97
TG		 583 dev_net_set(dev, net);
584
1da177e4
LT		 585 if (register_netdevice(dev)) {
586 free_netdev(dev);
587 return NULL;
588 }
1da177e4 589
a0b47736 590 if (!ipmr_init_vif_indev(dev))
1da177e4 591 goto failure;
00f54e68 592 if (dev_open(dev, NULL))
1da177e4
LT		 593 goto failure;
594
7dc00c82
WC		 595 dev_hold(dev);
596
1da177e4
LT		 597 return dev;
598
599failure:
1da177e4
LT		 600 unregister_netdevice(dev);
601 return NULL;
602}
c316c629
NA		 603
604/* called with rcu_read_lock() */
605static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
606 unsigned int pimlen)
607{
608 struct net_device *reg_dev = NULL;
609 struct iphdr *encap;
610
611 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
7ef8f65d 612 /* Check that:
c316c629
NA		 613 * a. packet is really sent to a multicast group
614 * b. packet is not a NULL-REGISTER
615 * c. packet is not truncated
616 */
617 if (!ipv4_is_multicast(encap->daddr) ||
618 encap->tot_len == 0 ||
619 ntohs(encap->tot_len) + pimlen > skb->len)
620 return 1;
621
622 read_lock(&mrt_lock);
623 if (mrt->mroute_reg_vif_num >= 0)
624 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
625 read_unlock(&mrt_lock);
626
627 if (!reg_dev)
628 return 1;
629
630 skb->mac_header = skb->network_header;
631 skb_pull(skb, (u8 *)encap - skb->data);
632 skb_reset_network_header(skb);
633 skb->protocol = htons(ETH_P_IP);
634 skb->ip_summed = CHECKSUM_NONE;
635
636 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
637
638 netif_rx(skb);
639
640 return NET_RX_SUCCESS;
641}
642#else
643static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
644{
645 return NULL;
646}
1da177e4
LT		 647#endif
648
b362053a
YG		 649static int call_ipmr_vif_entry_notifiers(struct net *net,
650 enum fib_event_type event_type,
651 struct vif_device *vif,
652 vifi_t vif_index, u32 tb_id)
653{
bc67a0da
YM		 654 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
655 vif, vif_index, tb_id,
656 &net->ipv4.ipmr_seq);
b362053a
YG		 657}
658
b362053a
YG		 659static int call_ipmr_mfc_entry_notifiers(struct net *net,
660 enum fib_event_type event_type,
661 struct mfc_cache *mfc, u32 tb_id)
662{
54c4cad9
YM		 663 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
664 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
b362053a
YG		 665}
666
2c53040f
BH		 667/**
668 * vif_delete - Delete a VIF entry
7dc00c82 669 * @notify: Set to 1, if the caller is a notifier_call
1da177e4 670 */
0c12295a 671static int vif_delete(struct mr_table *mrt, int vifi, int notify,
d17fa6fa 672 struct list_head *head)
1da177e4 673{
b362053a 674 struct net *net = read_pnet(&mrt->net);
1da177e4
LT		 675 struct vif_device *v;
676 struct net_device *dev;
677 struct in_device *in_dev;
678
0c12295a 679 if (vifi < 0 || vifi >= mrt->maxvif)
1da177e4
LT		 680 return -EADDRNOTAVAIL;
681
0c12295a 682 v = &mrt->vif_table[vifi];
1da177e4 683
b362053a
YG		 684 if (VIF_EXISTS(mrt, vifi))
685 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
686 mrt->id);
687
1da177e4
LT		 688 write_lock_bh(&mrt_lock);
689 dev = v->dev;
690 v->dev = NULL;
691
692 if (!dev) {
693 write_unlock_bh(&mrt_lock);
694 return -EADDRNOTAVAIL;
695 }
696
0c12295a
PM		 697 if (vifi == mrt->mroute_reg_vif_num)
698 mrt->mroute_reg_vif_num = -1;
1da177e4 699
a8cb16dd 700 if (vifi + 1 == mrt->maxvif) {
1da177e4 701 int tmp;
a8cb16dd
ED		 702
703 for (tmp = vifi - 1; tmp >= 0; tmp--) {
0c12295a 704 if (VIF_EXISTS(mrt, tmp))
1da177e4
LT		 705 break;
706 }
0c12295a 707 mrt->maxvif = tmp+1;
1da177e4
LT		 708 }
709
710 write_unlock_bh(&mrt_lock);
711
712 dev_set_allmulti(dev, -1);
713
a8cb16dd
ED		 714 in_dev = __in_dev_get_rtnl(dev);
715 if (in_dev) {
42f811b8 716 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
3b022865 717 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
d67b8c61
ND		 718 NETCONFA_MC_FORWARDING,
719 dev->ifindex, &in_dev->cnf);
1da177e4
LT		 720 ip_rt_multicast_event(in_dev);
721 }
722
a8cb16dd 723 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
d17fa6fa 724 unregister_netdevice_queue(dev, head);
1da177e4
LT		 725
726 dev_put(dev);
727 return 0;
728}
729
a8c9486b 730static void ipmr_cache_free_rcu(struct rcu_head *head)
5c0a66f5 731{
494fff56 732 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
a8c9486b 733
494fff56 734 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
5c0a66f5
BT		 735}
736
8c13af2a 737static void ipmr_cache_free(struct mfc_cache *c)
a8c9486b 738{
494fff56 739 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
a8c9486b
ED		 740}
741
1da177e4 742/* Destroy an unresolved cache entry, killing queued skbs
a8cb16dd 743 * and reporting error to netlink readers.
1da177e4 744 */
0c12295a 745static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
1da177e4 746{
8de53dfb 747 struct net *net = read_pnet(&mrt->net);
1da177e4 748 struct sk_buff *skb;
9ef1d4c7 749 struct nlmsgerr *e;
1da177e4 750
0c12295a 751 atomic_dec(&mrt->cache_resolve_queue_len);
1da177e4 752
494fff56 753 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
eddc9ec5 754 if (ip_hdr(skb)->version == 0) {
af72868b
JB		 755 struct nlmsghdr *nlh = skb_pull(skb,
756 sizeof(struct iphdr));
1da177e4 757 nlh->nlmsg_type = NLMSG_ERROR;
573ce260 758 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1da177e4 759 skb_trim(skb, nlh->nlmsg_len);
573ce260 760 e = nlmsg_data(nlh);
9ef1d4c7
PM		 761 e->error = -ETIMEDOUT;
762 memset(&e->msg, 0, sizeof(e->msg));
2942e900 763
15e47304 764 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
a8cb16dd 765 } else {
1da177e4 766 kfree_skb(skb);
a8cb16dd 767 }
1da177e4
LT		 768 }
769
5c0a66f5 770 ipmr_cache_free(c);
1da177e4
LT		 771}
772
e258beb2 773/* Timer process for the unresolved queue. */
e99e88a9 774static void ipmr_expire_process(struct timer_list *t)
1da177e4 775{
e99e88a9 776 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
494fff56 777 struct mr_mfc *c, *next;
1da177e4 778 unsigned long expires;
494fff56 779 unsigned long now;
1da177e4
LT		 780
781 if (!spin_trylock(&mfc_unres_lock)) {
0c12295a 782 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
1da177e4
LT		 783 return;
784 }
785
0c12295a 786 if (list_empty(&mrt->mfc_unres_queue))
1da177e4
LT		 787 goto out;
788
789 now = jiffies;
790 expires = 10*HZ;
1da177e4 791
0c12295a 792 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1da177e4
LT		 793 if (time_after(c->mfc_un.unres.expires, now)) {
794 unsigned long interval = c->mfc_un.unres.expires - now;
795 if (interval < expires)
796 expires = interval;
1da177e4
LT		 797 continue;
798 }
799
862465f2 800 list_del(&c->list);
494fff56
YM		 801 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
802 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
1da177e4
LT		 803 }
804
0c12295a
PM		 805 if (!list_empty(&mrt->mfc_unres_queue))
806 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
1da177e4
LT		 807
808out:
809 spin_unlock(&mfc_unres_lock);
810}
811
812/* Fill oifs list. It is called under write locked mrt_lock. */
494fff56 813static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
d658f8a0 814 unsigned char *ttls)
1da177e4
LT		 815{
816 int vifi;
817
818 cache->mfc_un.res.minvif = MAXVIFS;
819 cache->mfc_un.res.maxvif = 0;
820 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
821
0c12295a
PM		 822 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
823 if (VIF_EXISTS(mrt, vifi) &&
cf958ae3 824 ttls[vifi] && ttls[vifi] < 255) {
1da177e4
LT		 825 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
826 if (cache->mfc_un.res.minvif > vifi)
827 cache->mfc_un.res.minvif = vifi;
828 if (cache->mfc_un.res.maxvif <= vifi)
829 cache->mfc_un.res.maxvif = vifi + 1;
830 }
831 }
90b5ca17 832 cache->mfc_un.res.lastuse = jiffies;
1da177e4
LT		 833}
834
0c12295a
PM		 835static int vif_add(struct net *net, struct mr_table *mrt,
836 struct vifctl *vifc, int mrtsock)
1da177e4 837{
bccb3025 838 struct netdev_phys_item_id ppid = { };
1da177e4 839 int vifi = vifc->vifc_vifi;
0c12295a 840 struct vif_device *v = &mrt->vif_table[vifi];
1da177e4
LT		 841 struct net_device *dev;
842 struct in_device *in_dev;
d607032d 843 int err;
1da177e4
LT		 844
845 /* Is vif busy ? */
0c12295a 846 if (VIF_EXISTS(mrt, vifi))
1da177e4
LT		 847 return -EADDRINUSE;
848
849 switch (vifc->vifc_flags) {
1da177e4 850 case VIFF_REGISTER:
1973a4ea 851 if (!ipmr_pimsm_enabled())
c316c629
NA		 852 return -EINVAL;
853 /* Special Purpose VIF in PIM
1da177e4
LT		 854 * All the packets will be sent to the daemon
855 */
0c12295a 856 if (mrt->mroute_reg_vif_num >= 0)
1da177e4 857 return -EADDRINUSE;
f0ad0860 858 dev = ipmr_reg_vif(net, mrt);
1da177e4
LT		 859 if (!dev)
860 return -ENOBUFS;
d607032d
WC		 861 err = dev_set_allmulti(dev, 1);
862 if (err) {
863 unregister_netdevice(dev);
7dc00c82 864 dev_put(dev);
d607032d
WC		 865 return err;
866 }
1da177e4 867 break;
e905a9ed 868 case VIFF_TUNNEL:
4feb88e5 869 dev = ipmr_new_tunnel(net, vifc);
1da177e4
LT		 870 if (!dev)
871 return -ENOBUFS;
d607032d
WC		 872 err = dev_set_allmulti(dev, 1);
873 if (err) {
874 ipmr_del_tunnel(dev, vifc);
7dc00c82 875 dev_put(dev);
d607032d
WC		 876 return err;
877 }
1da177e4 878 break;
ee5e81f0 879 case VIFF_USE_IFINDEX:
1da177e4 880 case 0:
ee5e81f0
I		 881 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
882 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
51456b29 883 if (dev && !__in_dev_get_rtnl(dev)) {
ee5e81f0
I		 884 dev_put(dev);
885 return -EADDRNOTAVAIL;
886 }
a8cb16dd 887 } else {
ee5e81f0 888 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
a8cb16dd 889 }
1da177e4
LT		 890 if (!dev)
891 return -EADDRNOTAVAIL;
d607032d 892 err = dev_set_allmulti(dev, 1);
7dc00c82
WC		 893 if (err) {
894 dev_put(dev);
d607032d 895 return err;
7dc00c82 896 }
1da177e4
LT		 897 break;
898 default:
899 return -EINVAL;
900 }
901
a8cb16dd
ED		 902 in_dev = __in_dev_get_rtnl(dev);
903 if (!in_dev) {
d0490cfd 904 dev_put(dev);
1da177e4 905 return -EADDRNOTAVAIL;
d0490cfd 906 }
42f811b8 907 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
3b022865
DA		 908 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
909 dev->ifindex, &in_dev->cnf);
1da177e4
LT		 910 ip_rt_multicast_event(in_dev);
911
a8cb16dd 912 /* Fill in the VIF structures */
6853f21f
YM		 913 vif_device_init(v, dev, vifc->vifc_rate_limit,
914 vifc->vifc_threshold,
915 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
916 (VIFF_TUNNEL | VIFF_REGISTER));
a8cb16dd 917
bccb3025
FF		 918 err = dev_get_port_parent_id(dev, &ppid, true);
919 if (err == 0) {
920 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
921 v->dev_parent_id.id_len = ppid.id_len;
5d8b3e69
YG		 922 } else {
923 v->dev_parent_id.id_len = 0;
924 }
6853f21f 925
c354e124
JK		 926 v->local = vifc->vifc_lcl_addr.s_addr;
927 v->remote = vifc->vifc_rmt_addr.s_addr;
1da177e4
LT		 928
929 /* And finish update writing critical data */
930 write_lock_bh(&mrt_lock);
c354e124 931 v->dev = dev;
a8cb16dd 932 if (v->flags & VIFF_REGISTER)
0c12295a 933 mrt->mroute_reg_vif_num = vifi;
0c12295a
PM		 934 if (vifi+1 > mrt->maxvif)
935 mrt->maxvif = vifi+1;
1da177e4 936 write_unlock_bh(&mrt_lock);
b362053a 937 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, vifi, mrt->id);
1da177e4
LT		 938 return 0;
939}
940
a8c9486b 941/* called with rcu_read_lock() */
0c12295a 942static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
4feb88e5
BT		 943 __be32 origin,
944 __be32 mcastgrp)
1da177e4 945{
8fb472c0
NA		 946 struct mfc_cache_cmp_arg arg = {
947 .mfc_mcastgrp = mcastgrp,
948 .mfc_origin = origin
949 };
660b26dc 950
845c9a7a 951 return mr_mfc_find(mrt, &arg);
660b26dc
ND		 952}
953
954/* Look for a (*,G) entry */
955static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
956 __be32 mcastgrp, int vifi)
957{
8fb472c0
NA		 958 struct mfc_cache_cmp_arg arg = {
959 .mfc_mcastgrp = mcastgrp,
960 .mfc_origin = htonl(INADDR_ANY)
961 };
660b26dc 962
360eb5da 963 if (mcastgrp == htonl(INADDR_ANY))
845c9a7a
YM		 964 return mr_mfc_find_any_parent(mrt, vifi);
965 return mr_mfc_find_any(mrt, vifi, &arg);
660b26dc
ND		 966}
967
8fb472c0
NA		 968/* Look for a (S,G,iif) entry if parent != -1 */
969static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
970 __be32 origin, __be32 mcastgrp,
971 int parent)
972{
973 struct mfc_cache_cmp_arg arg = {
974 .mfc_mcastgrp = mcastgrp,
975 .mfc_origin = origin,
976 };
8fb472c0 977
845c9a7a 978 return mr_mfc_find_parent(mrt, &arg, parent);
8fb472c0
NA		 979}
980
7ef8f65d 981/* Allocate a multicast cache entry */
d658f8a0 982static struct mfc_cache *ipmr_cache_alloc(void)
1da177e4 983{
c354e124 984 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
a8c9486b 985
70a0dec4 986 if (c) {
494fff56
YM		 987 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
988 c->_c.mfc_un.res.minvif = MAXVIFS;
8c13af2a 989 c->_c.free = ipmr_cache_free_rcu;
494fff56 990 refcount_set(&c->_c.mfc_un.res.refcount, 1);
70a0dec4 991 }
1da177e4
LT		 992 return c;
993}
994
d658f8a0 995static struct mfc_cache *ipmr_cache_alloc_unres(void)
1da177e4 996{
c354e124 997 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
a8c9486b
ED		 998
999 if (c) {
494fff56
YM		 1000 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
1001 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
a8c9486b 1002 }
1da177e4
LT		 1003 return c;
1004}
1005
7ef8f65d 1006/* A cache entry has gone into a resolved state from queued */
0c12295a
PM		 1007static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
1008 struct mfc_cache *uc, struct mfc_cache *c)
1da177e4
LT		 1009{
1010 struct sk_buff *skb;
9ef1d4c7 1011 struct nlmsgerr *e;
1da177e4 1012
a8cb16dd 1013 /* Play the pending entries through our router */
494fff56 1014 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
eddc9ec5 1015 if (ip_hdr(skb)->version == 0) {
af72868b
JB		 1016 struct nlmsghdr *nlh = skb_pull(skb,
1017 sizeof(struct iphdr));
1da177e4 1018
7b0db857
YM		 1019 if (mr_fill_mroute(mrt, skb, &c->_c,
1020 nlmsg_data(nlh)) > 0) {
a8cb16dd
ED		 1021 nlh->nlmsg_len = skb_tail_pointer(skb) -
1022 (u8 *)nlh;
1da177e4
LT		 1023 } else {
1024 nlh->nlmsg_type = NLMSG_ERROR;
573ce260 1025 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1da177e4 1026 skb_trim(skb, nlh->nlmsg_len);
573ce260 1027 e = nlmsg_data(nlh);
9ef1d4c7
PM		 1028 e->error = -EMSGSIZE;
1029 memset(&e->msg, 0, sizeof(e->msg));
1da177e4 1030 }
2942e900 1031
15e47304 1032 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
a8cb16dd 1033 } else {
4b1f0d33 1034 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
a8cb16dd 1035 }
1da177e4
LT		 1036 }
1037}
1038
5a645dd8 1039/* Bounce a cache query up to mrouted and netlink.
1da177e4 1040 *
c316c629 1041 * Called under mrt_lock.
1da177e4 1042 */
0c12295a 1043static int ipmr_cache_report(struct mr_table *mrt,
4feb88e5 1044 struct sk_buff *pkt, vifi_t vifi, int assert)
1da177e4 1045{
c9bdd4b5 1046 const int ihl = ip_hdrlen(pkt);
c316c629 1047 struct sock *mroute_sk;
1da177e4
LT		 1048 struct igmphdr *igmp;
1049 struct igmpmsg *msg;
c316c629 1050 struct sk_buff *skb;
1da177e4
LT		 1051 int ret;
1052
c921c207 1053 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1da177e4
LT		 1054 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1055 else
1da177e4
LT		 1056 skb = alloc_skb(128, GFP_ATOMIC);
1057
132adf54 1058 if (!skb)
1da177e4
LT		 1059 return -ENOBUFS;
1060
c921c207 1061 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1da177e4 1062 /* Ugly, but we have no choice with this interface.
a8cb16dd
ED		 1063 * Duplicate old header, fix ihl, length etc.
1064 * And all this only to mangle msg->im_msgtype and
1065 * to set msg->im_mbz to "mbz" :-)
1da177e4 1066 */
878c8145
ACM		 1067 skb_push(skb, sizeof(struct iphdr));
1068 skb_reset_network_header(skb);
badff6d0 1069 skb_reset_transport_header(skb);
0272ffc4 1070 msg = (struct igmpmsg *)skb_network_header(skb);
d56f90a7 1071 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
c921c207 1072 msg->im_msgtype = assert;
1da177e4 1073 msg->im_mbz = 0;
c921c207
NA		 1074 if (assert == IGMPMSG_WRVIFWHOLE)
1075 msg->im_vif = vifi;
1076 else
1077 msg->im_vif = mrt->mroute_reg_vif_num;
eddc9ec5
ACM		 1078 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1079 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1080 sizeof(struct iphdr));
c316c629
NA		 1081 } else {
1082 /* Copy the IP header */
1083 skb_set_network_header(skb, skb->len);
1084 skb_put(skb, ihl);
1085 skb_copy_to_linear_data(skb, pkt->data, ihl);
1086 /* Flag to the kernel this is a route add */
1087 ip_hdr(skb)->protocol = 0;
1088 msg = (struct igmpmsg *)skb_network_header(skb);
1089 msg->im_vif = vifi;
1090 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1091 /* Add our header */
4df864c1 1092 igmp = skb_put(skb, sizeof(struct igmphdr));
c316c629
NA		 1093 igmp->type = assert;
1094 msg->im_msgtype = assert;
1095 igmp->code = 0;
1096 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1097 skb->transport_header = skb->network_header;
e905a9ed 1098 }
1da177e4 1099
4c968709
ED		 1100 rcu_read_lock();
1101 mroute_sk = rcu_dereference(mrt->mroute_sk);
51456b29 1102 if (!mroute_sk) {
4c968709 1103 rcu_read_unlock();
1da177e4
LT		 1104 kfree_skb(skb);
1105 return -EINVAL;
1106 }
1107
5a645dd8
JG		 1108 igmpmsg_netlink_event(mrt, skb);
1109
a8cb16dd 1110 /* Deliver to mrouted */
4c968709
ED		 1111 ret = sock_queue_rcv_skb(mroute_sk, skb);
1112 rcu_read_unlock();
70a269e6 1113 if (ret < 0) {
e87cc472 1114 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1da177e4
LT		 1115 kfree_skb(skb);
1116 }
1117
1118 return ret;
1119}
1120
7ef8f65d
NA		 1121/* Queue a packet for resolution. It gets locked cache entry! */
1122static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
4b1f0d33 1123 struct sk_buff *skb, struct net_device *dev)
1da177e4 1124{
8fb472c0
NA		 1125 const struct iphdr *iph = ip_hdr(skb);
1126 struct mfc_cache *c;
862465f2 1127 bool found = false;
1da177e4 1128 int err;
1da177e4
LT		 1129
1130 spin_lock_bh(&mfc_unres_lock);
494fff56 1131 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
e258beb2 1132 if (c->mfc_mcastgrp == iph->daddr &&
862465f2
PM		 1133 c->mfc_origin == iph->saddr) {
1134 found = true;
1da177e4 1135 break;
862465f2 1136 }
1da177e4
LT		 1137 }
1138
862465f2 1139 if (!found) {
a8cb16dd 1140 /* Create a new entry if allowable */
0079ad8e
HL		 1141 c = ipmr_cache_alloc_unres();
1142 if (!c) {
1da177e4
LT		 1143 spin_unlock_bh(&mfc_unres_lock);
1144
1145 kfree_skb(skb);
1146 return -ENOBUFS;
1147 }
1148
a8cb16dd 1149 /* Fill in the new cache entry */
494fff56 1150 c->_c.mfc_parent = -1;
eddc9ec5
ACM		 1151 c->mfc_origin = iph->saddr;
1152 c->mfc_mcastgrp = iph->daddr;
1da177e4 1153
a8cb16dd 1154 /* Reflect first query at mrouted. */
0c12295a 1155 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
494fff56 1156
4feb88e5 1157 if (err < 0) {
e905a9ed 1158 /* If the report failed throw the cache entry
1da177e4
LT		 1159 out - Brad Parker
1160 */
1161 spin_unlock_bh(&mfc_unres_lock);
1162
5c0a66f5 1163 ipmr_cache_free(c);
1da177e4
LT		 1164 kfree_skb(skb);
1165 return err;
1166 }
1167
0c12295a 1168 atomic_inc(&mrt->cache_resolve_queue_len);
494fff56 1169 list_add(&c->_c.list, &mrt->mfc_unres_queue);
8cd3ac9f 1170 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1da177e4 1171
278554bd 1172 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
494fff56
YM		 1173 mod_timer(&mrt->ipmr_expire_timer,
1174 c->_c.mfc_un.unres.expires);
1da177e4
LT		 1175 }
1176
a8cb16dd 1177 /* See if we can append the packet */
494fff56 1178 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1da177e4
LT		 1179 kfree_skb(skb);
1180 err = -ENOBUFS;
1181 } else {
4b1f0d33
DS		 1182 if (dev) {
1183 skb->dev = dev;
1184 skb->skb_iif = dev->ifindex;
1185 }
494fff56 1186 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1da177e4
LT		 1187 err = 0;
1188 }
1189
1190 spin_unlock_bh(&mfc_unres_lock);
1191 return err;
1192}
1193
7ef8f65d 1194/* MFC cache manipulation by user space mroute daemon */
1da177e4 1195
660b26dc 1196static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1da177e4 1197{
b362053a 1198 struct net *net = read_pnet(&mrt->net);
8fb472c0 1199 struct mfc_cache *c;
1da177e4 1200
8fb472c0
NA		 1201 /* The entries are added/deleted only under RTNL */
1202 rcu_read_lock();
1203 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1204 mfc->mfcc_mcastgrp.s_addr, parent);
1205 rcu_read_unlock();
1206 if (!c)
1207 return -ENOENT;
494fff56
YM		 1208 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1209 list_del_rcu(&c->_c.list);
b362053a 1210 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
8fb472c0 1211 mroute_netlink_event(mrt, c, RTM_DELROUTE);
8c13af2a 1212 mr_cache_put(&c->_c);
1da177e4 1213
8fb472c0 1214 return 0;
1da177e4
LT		 1215}
1216
0c12295a 1217static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
660b26dc 1218 struct mfcctl *mfc, int mrtsock, int parent)
1da177e4 1219{
862465f2 1220 struct mfc_cache *uc, *c;
494fff56 1221 struct mr_mfc *_uc;
8fb472c0
NA		 1222 bool found;
1223 int ret;
1da177e4 1224
a50436f2
PM		 1225 if (mfc->mfcc_parent >= MAXVIFS)
1226 return -ENFILE;
1227
8fb472c0
NA		 1228 /* The entries are added/deleted only under RTNL */
1229 rcu_read_lock();
1230 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1231 mfc->mfcc_mcastgrp.s_addr, parent);
1232 rcu_read_unlock();
1233 if (c) {
1da177e4 1234 write_lock_bh(&mrt_lock);
494fff56
YM		 1235 c->_c.mfc_parent = mfc->mfcc_parent;
1236 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1da177e4 1237 if (!mrtsock)
494fff56 1238 c->_c.mfc_flags |= MFC_STATIC;
1da177e4 1239 write_unlock_bh(&mrt_lock);
b362053a
YG		 1240 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1241 mrt->id);
8cd3ac9f 1242 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1da177e4
LT		 1243 return 0;
1244 }
1245
360eb5da 1246 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
660b26dc 1247 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1da177e4
LT		 1248 return -EINVAL;
1249
d658f8a0 1250 c = ipmr_cache_alloc();
51456b29 1251 if (!c)
1da177e4
LT		 1252 return -ENOMEM;
1253
c354e124
JK		 1254 c->mfc_origin = mfc->mfcc_origin.s_addr;
1255 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
494fff56
YM		 1256 c->_c.mfc_parent = mfc->mfcc_parent;
1257 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1da177e4 1258 if (!mrtsock)
494fff56 1259 c->_c.mfc_flags |= MFC_STATIC;
1da177e4 1260
494fff56 1261 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
8fb472c0
NA		 1262 ipmr_rht_params);
1263 if (ret) {
1264 pr_err("ipmr: rhtable insert error %d\n", ret);
1265 ipmr_cache_free(c);
1266 return ret;
1267 }
494fff56 1268 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
7ef8f65d
NA		 1269 /* Check to see if we resolved a queued list. If so we
1270 * need to send on the frames and tidy up.
1da177e4 1271 */
b0ebb739 1272 found = false;
1da177e4 1273 spin_lock_bh(&mfc_unres_lock);
494fff56
YM		 1274 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1275 uc = (struct mfc_cache *)_uc;
e258beb2 1276 if (uc->mfc_origin == c->mfc_origin &&
1da177e4 1277 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
494fff56 1278 list_del(&_uc->list);
0c12295a 1279 atomic_dec(&mrt->cache_resolve_queue_len);
b0ebb739 1280 found = true;
1da177e4
LT		 1281 break;
1282 }
1283 }
0c12295a
PM		 1284 if (list_empty(&mrt->mfc_unres_queue))
1285 del_timer(&mrt->ipmr_expire_timer);
1da177e4
LT		 1286 spin_unlock_bh(&mfc_unres_lock);
1287
b0ebb739 1288 if (found) {
0c12295a 1289 ipmr_cache_resolve(net, mrt, uc, c);
5c0a66f5 1290 ipmr_cache_free(uc);
1da177e4 1291 }
b362053a 1292 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
8cd3ac9f 1293 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1da177e4
LT		 1294 return 0;
1295}
1296
7ef8f65d 1297/* Close the multicast socket, and clear the vif tables etc */
ca8d4794 1298static void mroute_clean_tables(struct mr_table *mrt, int flags)
1da177e4 1299{
b362053a 1300 struct net *net = read_pnet(&mrt->net);
494fff56
YM		 1301 struct mr_mfc *c, *tmp;
1302 struct mfc_cache *cache;
d17fa6fa 1303 LIST_HEAD(list);
8fb472c0 1304 int i;
e905a9ed 1305
a8cb16dd 1306 /* Shut down all active vif entries */
ca8d4794
CS		 1307 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1308 for (i = 0; i < mrt->maxvif; i++) {
1309 if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1310 !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1311 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1312 continue;
1313 vif_delete(mrt, i, 0, &list);
1314 }
1315 unregister_netdevice_many(&list);
1da177e4
LT		 1316 }
1317
a8cb16dd 1318 /* Wipe the cache */
ca8d4794
CS		 1319 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1320 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1321 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1322 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1323 continue;
1324 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1325 list_del_rcu(&c->list);
494fff56 1326 cache = (struct mfc_cache *)c;
ca8d4794
CS		 1327 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1328 mrt->id);
494fff56 1329 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
ca8d4794
CS		 1330 mr_cache_put(c);
1331 }
1332 }
1333
1334 if (flags & MRT_FLUSH_MFC) {
1335 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1336 spin_lock_bh(&mfc_unres_lock);
1337 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1338 list_del(&c->list);
1339 cache = (struct mfc_cache *)c;
1340 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1341 ipmr_destroy_unres(mrt, cache);
1342 }
1343 spin_unlock_bh(&mfc_unres_lock);
1da177e4 1344 }
1da177e4
LT		 1345 }
1346}
1347
4c968709
ED		 1348/* called from ip_ra_control(), before an RCU grace period,
1349 * we dont need to call synchronize_rcu() here
1350 */
1da177e4
LT		 1351static void mrtsock_destruct(struct sock *sk)
1352{
4feb88e5 1353 struct net *net = sock_net(sk);
f0ad0860 1354 struct mr_table *mrt;
4feb88e5 1355
128aaa98 1356 rtnl_lock();
f0ad0860 1357 ipmr_for_each_table(mrt, net) {
4c968709 1358 if (sk == rtnl_dereference(mrt->mroute_sk)) {
f0ad0860 1359 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
3b022865
DA		 1360 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1361 NETCONFA_MC_FORWARDING,
d67b8c61
ND		 1362 NETCONFA_IFINDEX_ALL,
1363 net->ipv4.devconf_all);
a9b3cd7f 1364 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
ca8d4794 1365 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
f0ad0860 1366 }
1da177e4 1367 }
128aaa98 1368 rtnl_unlock();
1da177e4
LT		 1369}
1370
7ef8f65d
NA		 1371/* Socket options and virtual interface manipulation. The whole
1372 * virtual interface system is a complete heap, but unfortunately
1373 * that's how BSD mrouted happens to think. Maybe one day with a proper
1374 * MOSPF/PIM router set up we can clean this up.
1da177e4 1375 */
e905a9ed 1376
29e97d21
NA		 1377int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval,
1378 unsigned int optlen)
1da177e4 1379{
4feb88e5 1380 struct net *net = sock_net(sk);
29e97d21 1381 int val, ret = 0, parent = 0;
f0ad0860 1382 struct mr_table *mrt;
29e97d21
NA		 1383 struct vifctl vif;
1384 struct mfcctl mfc;
c921c207 1385 bool do_wrvifwhole;
29e97d21 1386 u32 uval;
f0ad0860 1387
29e97d21
NA		 1388 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1389 rtnl_lock();
5e1859fb 1390 if (sk->sk_type != SOCK_RAW ||
29e97d21
NA		 1391 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1392 ret = -EOPNOTSUPP;
1393 goto out_unlock;
1394 }
5e1859fb 1395
f0ad0860 1396 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
29e97d21
NA		 1397 if (!mrt) {
1398 ret = -ENOENT;
1399 goto out_unlock;
1400 }
132adf54 1401 if (optname != MRT_INIT) {
33d480ce 1402 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
29e97d21
NA		 1403 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1404 ret = -EACCES;
1405 goto out_unlock;
1406 }
1da177e4
LT		 1407 }
1408
132adf54
SH		 1409 switch (optname) {
1410 case MRT_INIT:
42e6b89c 1411 if (optlen != sizeof(int)) {
29e97d21 1412 ret = -EINVAL;
42e6b89c
NA		 1413 break;
1414 }
1415 if (rtnl_dereference(mrt->mroute_sk)) {
29e97d21 1416 ret = -EADDRINUSE;
29e97d21 1417 break;
42e6b89c 1418 }
132adf54
SH		 1419
1420 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1421 if (ret == 0) {
cf778b00 1422 rcu_assign_pointer(mrt->mroute_sk, sk);
4feb88e5 1423 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
3b022865
DA		 1424 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1425 NETCONFA_MC_FORWARDING,
d67b8c61
ND		 1426 NETCONFA_IFINDEX_ALL,
1427 net->ipv4.devconf_all);
132adf54 1428 }
29e97d21 1429 break;
132adf54 1430 case MRT_DONE:
29e97d21
NA		 1431 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1432 ret = -EACCES;
1433 } else {
128aaa98
KT		 1434 /* We need to unlock here because mrtsock_destruct takes
1435 * care of rtnl itself and we can't change that due to
1436 * the IP_ROUTER_ALERT setsockopt which runs without it.
1437 */
1438 rtnl_unlock();
29e97d21 1439 ret = ip_ra_control(sk, 0, NULL);
128aaa98 1440 goto out;
29e97d21
NA		 1441 }
1442 break;
132adf54
SH		 1443 case MRT_ADD_VIF:
1444 case MRT_DEL_VIF:
29e97d21
NA		 1445 if (optlen != sizeof(vif)) {
1446 ret = -EINVAL;
1447 break;
1448 }
1449 if (copy_from_user(&vif, optval, sizeof(vif))) {
1450 ret = -EFAULT;
1451 break;
1452 }
1453 if (vif.vifc_vifi >= MAXVIFS) {
1454 ret = -ENFILE;
1455 break;
1456 }
c354e124 1457 if (optname == MRT_ADD_VIF) {
4c968709
ED		 1458 ret = vif_add(net, mrt, &vif,
1459 sk == rtnl_dereference(mrt->mroute_sk));
132adf54 1460 } else {
0c12295a 1461 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
132adf54 1462 }
29e97d21 1463 break;
7ef8f65d
NA		 1464 /* Manipulate the forwarding caches. These live
1465 * in a sort of kernel/user symbiosis.
1466 */
132adf54
SH		 1467 case MRT_ADD_MFC:
1468 case MRT_DEL_MFC:
660b26dc 1469 parent = -1;
a8eceea8 1470 fallthrough;
660b26dc
ND		 1471 case MRT_ADD_MFC_PROXY:
1472 case MRT_DEL_MFC_PROXY:
29e97d21
NA		 1473 if (optlen != sizeof(mfc)) {
1474 ret = -EINVAL;
1475 break;
1476 }
1477 if (copy_from_user(&mfc, optval, sizeof(mfc))) {
1478 ret = -EFAULT;
1479 break;
1480 }
660b26dc
ND		 1481 if (parent == 0)
1482 parent = mfc.mfcc_parent;
660b26dc
ND		 1483 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1484 ret = ipmr_mfc_delete(mrt, &mfc, parent);
132adf54 1485 else
4c968709 1486 ret = ipmr_mfc_add(net, mrt, &mfc,
660b26dc
ND		 1487 sk == rtnl_dereference(mrt->mroute_sk),
1488 parent);
29e97d21 1489 break;
ca8d4794
CS		 1490 case MRT_FLUSH:
1491 if (optlen != sizeof(val)) {
1492 ret = -EINVAL;
1493 break;
1494 }
1495 if (get_user(val, (int __user *)optval)) {
1496 ret = -EFAULT;
1497 break;
1498 }
1499 mroute_clean_tables(mrt, val);
1500 break;
7ef8f65d 1501 /* Control PIM assert. */
132adf54 1502 case MRT_ASSERT:
29e97d21
NA		 1503 if (optlen != sizeof(val)) {
1504 ret = -EINVAL;
1505 break;
1506 }
1507 if (get_user(val, (int __user *)optval)) {
1508 ret = -EFAULT;
1509 break;
1510 }
1511 mrt->mroute_do_assert = val;
1512 break;
132adf54 1513 case MRT_PIM:
1973a4ea 1514 if (!ipmr_pimsm_enabled()) {
29e97d21
NA		 1515 ret = -ENOPROTOOPT;
1516 break;
1517 }
1518 if (optlen != sizeof(val)) {
1519 ret = -EINVAL;
1520 break;
1521 }
1522 if (get_user(val, (int __user *)optval)) {
1523 ret = -EFAULT;
1524 break;
1525 }
ba93ef74 1526
c921c207 1527 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
29e97d21
NA		 1528 val = !!val;
1529 if (val != mrt->mroute_do_pim) {
1530 mrt->mroute_do_pim = val;
1531 mrt->mroute_do_assert = val;
c921c207 1532 mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1da177e4 1533 }
29e97d21 1534 break;
f0ad0860 1535 case MRT_TABLE:
29e97d21
NA		 1536 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1537 ret = -ENOPROTOOPT;
1538 break;
1539 }
1540 if (optlen != sizeof(uval)) {
1541 ret = -EINVAL;
1542 break;
1543 }
1544 if (get_user(uval, (u32 __user *)optval)) {
1545 ret = -EFAULT;
1546 break;
1547 }
f0ad0860 1548
4c968709
ED		 1549 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1550 ret = -EBUSY;
1551 } else {
29e97d21 1552 mrt = ipmr_new_table(net, uval);
1113ebbc
NA		 1553 if (IS_ERR(mrt))
1554 ret = PTR_ERR(mrt);
5e1859fb 1555 else
29e97d21 1556 raw_sk(sk)->ipmr_table = uval;
4c968709 1557 }
29e97d21 1558 break;
7ef8f65d 1559 /* Spurious command, or MRT_VERSION which you cannot set. */
132adf54 1560 default:
29e97d21 1561 ret = -ENOPROTOOPT;
1da177e4 1562 }
29e97d21
NA		 1563out_unlock:
1564 rtnl_unlock();
128aaa98 1565out:
29e97d21 1566 return ret;
1da177e4
LT		 1567}
1568
7ef8f65d 1569/* Getsock opt support for the multicast routing system. */
c354e124 1570int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1da177e4
LT		 1571{
1572 int olr;
1573 int val;
4feb88e5 1574 struct net *net = sock_net(sk);
f0ad0860
PM		 1575 struct mr_table *mrt;
1576
5e1859fb
ED		 1577 if (sk->sk_type != SOCK_RAW ||
1578 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1579 return -EOPNOTSUPP;
1580
f0ad0860 1581 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
51456b29 1582 if (!mrt)
f0ad0860 1583 return -ENOENT;
1da177e4 1584
fe9ef3ce
NA		 1585 switch (optname) {
1586 case MRT_VERSION:
1587 val = 0x0305;
1588 break;
1589 case MRT_PIM:
1973a4ea 1590 if (!ipmr_pimsm_enabled())
fe9ef3ce
NA		 1591 return -ENOPROTOOPT;
1592 val = mrt->mroute_do_pim;
1593 break;
1594 case MRT_ASSERT:
1595 val = mrt->mroute_do_assert;
1596 break;
1597 default:
1da177e4 1598 return -ENOPROTOOPT;
fe9ef3ce 1599 }
1da177e4
LT		 1600
1601 if (get_user(olr, optlen))
1602 return -EFAULT;
1da177e4
LT		 1603 olr = min_t(unsigned int, olr, sizeof(int));
1604 if (olr < 0)
1605 return -EINVAL;
c354e124 1606 if (put_user(olr, optlen))
1da177e4 1607 return -EFAULT;
c354e124 1608 if (copy_to_user(optval, &val, olr))
1da177e4
LT		 1609 return -EFAULT;
1610 return 0;
1611}
1612
7ef8f65d 1613/* The IP multicast ioctl support routines. */
1da177e4
LT		 1614int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1615{
1616 struct sioc_sg_req sr;
1617 struct sioc_vif_req vr;
1618 struct vif_device *vif;
1619 struct mfc_cache *c;
4feb88e5 1620 struct net *net = sock_net(sk);
f0ad0860
PM		 1621 struct mr_table *mrt;
1622
1623 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
51456b29 1624 if (!mrt)
f0ad0860 1625 return -ENOENT;
e905a9ed 1626
132adf54
SH		 1627 switch (cmd) {
1628 case SIOCGETVIFCNT:
c354e124 1629 if (copy_from_user(&vr, arg, sizeof(vr)))
132adf54 1630 return -EFAULT;
0c12295a 1631 if (vr.vifi >= mrt->maxvif)
132adf54 1632 return -EINVAL;
5648451e 1633 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
132adf54 1634 read_lock(&mrt_lock);
0c12295a
PM		 1635 vif = &mrt->vif_table[vr.vifi];
1636 if (VIF_EXISTS(mrt, vr.vifi)) {
c354e124
JK		 1637 vr.icount = vif->pkt_in;
1638 vr.ocount = vif->pkt_out;
1639 vr.ibytes = vif->bytes_in;
1640 vr.obytes = vif->bytes_out;
1da177e4 1641 read_unlock(&mrt_lock);
1da177e4 1642
c354e124 1643 if (copy_to_user(arg, &vr, sizeof(vr)))
132adf54
SH		 1644 return -EFAULT;
1645 return 0;
1646 }
1647 read_unlock(&mrt_lock);
1648 return -EADDRNOTAVAIL;
1649 case SIOCGETSGCNT:
c354e124 1650 if (copy_from_user(&sr, arg, sizeof(sr)))
132adf54
SH		 1651 return -EFAULT;
1652
a8c9486b 1653 rcu_read_lock();
0c12295a 1654 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
132adf54 1655 if (c) {
494fff56
YM		 1656 sr.pktcnt = c->_c.mfc_un.res.pkt;
1657 sr.bytecnt = c->_c.mfc_un.res.bytes;
1658 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
a8c9486b 1659 rcu_read_unlock();
132adf54 1660
c354e124 1661 if (copy_to_user(arg, &sr, sizeof(sr)))
132adf54
SH		 1662 return -EFAULT;
1663 return 0;
1664 }
a8c9486b 1665 rcu_read_unlock();
132adf54
SH		 1666 return -EADDRNOTAVAIL;
1667 default:
1668 return -ENOIOCTLCMD;
1da177e4
LT		 1669 }
1670}
1671
709b46e8
EB		 1672#ifdef CONFIG_COMPAT
1673struct compat_sioc_sg_req {
1674 struct in_addr src;
1675 struct in_addr grp;
1676 compat_ulong_t pktcnt;
1677 compat_ulong_t bytecnt;
1678 compat_ulong_t wrong_if;
1679};
1680
ca6b8bb0
DM		 1681struct compat_sioc_vif_req {
1682 vifi_t vifi; /* Which iface */
1683 compat_ulong_t icount;
1684 compat_ulong_t ocount;
1685 compat_ulong_t ibytes;
1686 compat_ulong_t obytes;
1687};
1688
709b46e8
EB		 1689int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1690{
0033d5ad 1691 struct compat_sioc_sg_req sr;
ca6b8bb0
DM		 1692 struct compat_sioc_vif_req vr;
1693 struct vif_device *vif;
709b46e8
EB		 1694 struct mfc_cache *c;
1695 struct net *net = sock_net(sk);
1696 struct mr_table *mrt;
1697
1698 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
51456b29 1699 if (!mrt)
709b46e8
EB		 1700 return -ENOENT;
1701
1702 switch (cmd) {
ca6b8bb0
DM		 1703 case SIOCGETVIFCNT:
1704 if (copy_from_user(&vr, arg, sizeof(vr)))
1705 return -EFAULT;
1706 if (vr.vifi >= mrt->maxvif)
1707 return -EINVAL;
5648451e 1708 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
ca6b8bb0
DM		 1709 read_lock(&mrt_lock);
1710 vif = &mrt->vif_table[vr.vifi];
1711 if (VIF_EXISTS(mrt, vr.vifi)) {
1712 vr.icount = vif->pkt_in;
1713 vr.ocount = vif->pkt_out;
1714 vr.ibytes = vif->bytes_in;
1715 vr.obytes = vif->bytes_out;
1716 read_unlock(&mrt_lock);
1717
1718 if (copy_to_user(arg, &vr, sizeof(vr)))
1719 return -EFAULT;
1720 return 0;
1721 }
1722 read_unlock(&mrt_lock);
1723 return -EADDRNOTAVAIL;
709b46e8
EB		 1724 case SIOCGETSGCNT:
1725 if (copy_from_user(&sr, arg, sizeof(sr)))
1726 return -EFAULT;
1727
1728 rcu_read_lock();
1729 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1730 if (c) {
494fff56
YM		 1731 sr.pktcnt = c->_c.mfc_un.res.pkt;
1732 sr.bytecnt = c->_c.mfc_un.res.bytes;
1733 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
709b46e8
EB		 1734 rcu_read_unlock();
1735
1736 if (copy_to_user(arg, &sr, sizeof(sr)))
1737 return -EFAULT;
1738 return 0;
1739 }
1740 rcu_read_unlock();
1741 return -EADDRNOTAVAIL;
1742 default:
1743 return -ENOIOCTLCMD;
1744 }
1745}
1746#endif
1747
1da177e4
LT		 1748static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1749{
351638e7 1750 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4feb88e5 1751 struct net *net = dev_net(dev);
f0ad0860 1752 struct mr_table *mrt;
1da177e4
LT		 1753 struct vif_device *v;
1754 int ct;
e9dc8653 1755
1da177e4
LT		 1756 if (event != NETDEV_UNREGISTER)
1757 return NOTIFY_DONE;
f0ad0860
PM		 1758
1759 ipmr_for_each_table(mrt, net) {
1760 v = &mrt->vif_table[0];
1761 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1762 if (v->dev == dev)
e92036a6 1763 vif_delete(mrt, ct, 1, NULL);
f0ad0860 1764 }
1da177e4
LT		 1765 }
1766 return NOTIFY_DONE;
1767}
1768
c354e124 1769static struct notifier_block ip_mr_notifier = {
1da177e4
LT		 1770 .notifier_call = ipmr_device_event,
1771};
1772
7ef8f65d
NA		 1773/* Encapsulate a packet by attaching a valid IPIP header to it.
1774 * This avoids tunnel drivers and other mess and gives us the speed so
1775 * important for multicast video.
1da177e4 1776 */
b6a7719a
HFS		 1777static void ip_encap(struct net *net, struct sk_buff *skb,
1778 __be32 saddr, __be32 daddr)
1da177e4 1779{
8856dfa3 1780 struct iphdr *iph;
b71d1d42 1781 const struct iphdr *old_iph = ip_hdr(skb);
8856dfa3
ACM		 1782
1783 skb_push(skb, sizeof(struct iphdr));
b0e380b1 1784 skb->transport_header = skb->network_header;
8856dfa3 1785 skb_reset_network_header(skb);
eddc9ec5 1786 iph = ip_hdr(skb);
1da177e4 1787
a8cb16dd 1788 iph->version = 4;
e023dd64
ACM		 1789 iph->tos = old_iph->tos;
1790 iph->ttl = old_iph->ttl;
1da177e4
LT		 1791 iph->frag_off = 0;
1792 iph->daddr = daddr;
1793 iph->saddr = saddr;
1794 iph->protocol = IPPROTO_IPIP;
1795 iph->ihl = 5;
1796 iph->tot_len = htons(skb->len);
b6a7719a 1797 ip_select_ident(net, skb, NULL);
1da177e4
LT		 1798 ip_send_check(iph);
1799
1da177e4 1800 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
895b5c9f 1801 nf_reset_ct(skb);
1da177e4
LT		 1802}
1803
0c4b51f0
EB		 1804static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1805 struct sk_buff *skb)
1da177e4 1806{
a8cb16dd 1807 struct ip_options *opt = &(IPCB(skb)->opt);
1da177e4 1808
73186df8
DM		 1809 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1810 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1da177e4
LT		 1811
1812 if (unlikely(opt->optlen))
1813 ip_forward_options(skb);
1814
13206b6b 1815 return dst_output(net, sk, skb);
1da177e4
LT		 1816}
1817
a5bc9294
YG		 1818#ifdef CONFIG_NET_SWITCHDEV
1819static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1820 int in_vifi, int out_vifi)
1821{
1822 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1823 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1824
875e8939 1825 if (!skb->offload_l3_fwd_mark)
a5bc9294
YG		 1826 return false;
1827 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1828 return false;
1829 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1830 &in_vif->dev_parent_id);
1831}
1832#else
1833static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1834 int in_vifi, int out_vifi)
1835{
1836 return false;
1837}
1838#endif
1839
7ef8f65d 1840/* Processing handlers for ipmr_forward */
1da177e4 1841
0c12295a 1842static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
6e0735d1 1843 int in_vifi, struct sk_buff *skb, int vifi)
1da177e4 1844{
eddc9ec5 1845 const struct iphdr *iph = ip_hdr(skb);
0c12295a 1846 struct vif_device *vif = &mrt->vif_table[vifi];
1da177e4
LT		 1847 struct net_device *dev;
1848 struct rtable *rt;
31e4543d 1849 struct flowi4 fl4;
1da177e4
LT		 1850 int encap = 0;
1851
51456b29 1852 if (!vif->dev)
1da177e4
LT		 1853 goto out_free;
1854
1da177e4
LT		 1855 if (vif->flags & VIFF_REGISTER) {
1856 vif->pkt_out++;
c354e124 1857 vif->bytes_out += skb->len;
cf3677ae
PE		 1858 vif->dev->stats.tx_bytes += skb->len;
1859 vif->dev->stats.tx_packets++;
0c12295a 1860 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
69ebbf58 1861 goto out_free;
1da177e4 1862 }
1da177e4 1863
a5bc9294
YG		 1864 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1865 goto out_free;
1866
a8cb16dd 1867 if (vif->flags & VIFF_TUNNEL) {
31e4543d 1868 rt = ip_route_output_ports(net, &fl4, NULL,
78fbfd8a
DM		 1869 vif->remote, vif->local,
1870 0, 0,
1871 IPPROTO_IPIP,
1872 RT_TOS(iph->tos), vif->link);
b23dd4fe 1873 if (IS_ERR(rt))
1da177e4
LT		 1874 goto out_free;
1875 encap = sizeof(struct iphdr);
1876 } else {
31e4543d 1877 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
78fbfd8a
DM		 1878 0, 0,
1879 IPPROTO_IPIP,
1880 RT_TOS(iph->tos), vif->link);
b23dd4fe 1881 if (IS_ERR(rt))
1da177e4
LT		 1882 goto out_free;
1883 }
1884
d8d1f30b 1885 dev = rt->dst.dev;
1da177e4 1886
d8d1f30b 1887 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1da177e4 1888 /* Do not fragment multicasts. Alas, IPv4 does not
a8cb16dd
ED		 1889 * allow to send ICMP, so that packets will disappear
1890 * to blackhole.
1da177e4 1891 */
73186df8 1892 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1da177e4
LT		 1893 ip_rt_put(rt);
1894 goto out_free;
1895 }
1896
d8d1f30b 1897 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1da177e4
LT		 1898
1899 if (skb_cow(skb, encap)) {
e905a9ed 1900 ip_rt_put(rt);
1da177e4
LT		 1901 goto out_free;
1902 }
1903
1904 vif->pkt_out++;
c354e124 1905 vif->bytes_out += skb->len;
1da177e4 1906
adf30907 1907 skb_dst_drop(skb);
d8d1f30b 1908 skb_dst_set(skb, &rt->dst);
eddc9ec5 1909 ip_decrease_ttl(ip_hdr(skb));
1da177e4
LT		 1910
1911 /* FIXME: forward and output firewalls used to be called here.
a8cb16dd
ED		 1912 * What do we do with netfilter? -- RR
1913 */
1da177e4 1914 if (vif->flags & VIFF_TUNNEL) {
b6a7719a 1915 ip_encap(net, skb, vif->local, vif->remote);
1da177e4 1916 /* FIXME: extra output firewall step used to be here. --RR */
2f4c02d4
PE		 1917 vif->dev->stats.tx_packets++;
1918 vif->dev->stats.tx_bytes += skb->len;
1da177e4
LT		 1919 }
1920
9ee6c5dc 1921 IPCB(skb)->flags |= IPSKB_FORWARDED;
1da177e4 1922
7ef8f65d 1923 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1da177e4
LT		 1924 * not only before forwarding, but after forwarding on all output
1925 * interfaces. It is clear, if mrouter runs a multicasting
1926 * program, it should receive packets not depending to what interface
1927 * program is joined.
1928 * If we will not make it, the program will have to join on all
1929 * interfaces. On the other hand, multihoming host (or router, but
1930 * not mrouter) cannot join to more than one interface - it will
1931 * result in receiving multiple packets.
1932 */
29a26a56
EB		 1933 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1934 net, NULL, skb, skb->dev, dev,
1da177e4
LT		 1935 ipmr_forward_finish);
1936 return;
1937
1938out_free:
1939 kfree_skb(skb);
1da177e4
LT		 1940}
1941
0c12295a 1942static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1da177e4
LT		 1943{
1944 int ct;
0c12295a
PM		 1945
1946 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1947 if (mrt->vif_table[ct].dev == dev)
1da177e4
LT		 1948 break;
1949 }
1950 return ct;
1951}
1952
1953/* "local" means that we should preserve one skb (for local delivery) */
c4854ec8 1954static void ip_mr_forward(struct net *net, struct mr_table *mrt,
4b1f0d33 1955 struct net_device *dev, struct sk_buff *skb,
494fff56 1956 struct mfc_cache *c, int local)
1da177e4 1957{
4b1f0d33 1958 int true_vifi = ipmr_find_vif(mrt, dev);
1da177e4
LT		 1959 int psend = -1;
1960 int vif, ct;
1961
494fff56
YM		 1962 vif = c->_c.mfc_parent;
1963 c->_c.mfc_un.res.pkt++;
1964 c->_c.mfc_un.res.bytes += skb->len;
1965 c->_c.mfc_un.res.lastuse = jiffies;
1da177e4 1966
494fff56 1967 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
660b26dc
ND		 1968 struct mfc_cache *cache_proxy;
1969
1970 /* For an (*,G) entry, we only check that the incomming
1971 * interface is part of the static tree.
1972 */
845c9a7a 1973 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
660b26dc 1974 if (cache_proxy &&
494fff56 1975 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
660b26dc
ND		 1976 goto forward;
1977 }
1978
7ef8f65d 1979 /* Wrong interface: drop packet and (maybe) send PIM assert. */
4b1f0d33 1980 if (mrt->vif_table[vif].dev != dev) {
c7537967 1981 if (rt_is_output_route(skb_rtable(skb))) {
1da177e4 1982 /* It is our own packet, looped back.
a8cb16dd
ED		 1983 * Very complicated situation...
1984 *
1985 * The best workaround until routing daemons will be
1986 * fixed is not to redistribute packet, if it was
1987 * send through wrong interface. It means, that
1988 * multicast applications WILL NOT work for
1989 * (S,G), which have default multicast route pointing
1990 * to wrong oif. In any case, it is not a good
1991 * idea to use multicasting applications on router.
1da177e4
LT		 1992 */
1993 goto dont_forward;
1994 }
1995
494fff56 1996 c->_c.mfc_un.res.wrong_if++;
1da177e4 1997
0c12295a 1998 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1da177e4 1999 /* pimsm uses asserts, when switching from RPT to SPT,
a8cb16dd
ED		 2000 * so that we cannot check that packet arrived on an oif.
2001 * It is bad, but otherwise we would need to move pretty
2002 * large chunk of pimd to kernel. Ough... --ANK
1da177e4 2003 */
0c12295a 2004 (mrt->mroute_do_pim ||
494fff56 2005 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
e905a9ed 2006 time_after(jiffies,
494fff56
YM		 2007 c->_c.mfc_un.res.last_assert +
2008 MFC_ASSERT_THRESH)) {
2009 c->_c.mfc_un.res.last_assert = jiffies;
0c12295a 2010 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
c921c207
NA		 2011 if (mrt->mroute_do_wrvifwhole)
2012 ipmr_cache_report(mrt, skb, true_vifi,
2013 IGMPMSG_WRVIFWHOLE);
1da177e4
LT		 2014 }
2015 goto dont_forward;
2016 }
2017
660b26dc 2018forward:
0c12295a
PM		 2019 mrt->vif_table[vif].pkt_in++;
2020 mrt->vif_table[vif].bytes_in += skb->len;
1da177e4 2021
7ef8f65d 2022 /* Forward the frame */
494fff56
YM		 2023 if (c->mfc_origin == htonl(INADDR_ANY) &&
2024 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
660b26dc 2025 if (true_vifi >= 0 &&
494fff56 2026 true_vifi != c->_c.mfc_parent &&
660b26dc 2027 ip_hdr(skb)->ttl >
494fff56 2028 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
660b26dc
ND		 2029 /* It's an (*,*) entry and the packet is not coming from
2030 * the upstream: forward the packet to the upstream
2031 * only.
2032 */
494fff56 2033 psend = c->_c.mfc_parent;
660b26dc
ND		 2034 goto last_forward;
2035 }
2036 goto dont_forward;
2037 }
494fff56
YM		 2038 for (ct = c->_c.mfc_un.res.maxvif - 1;
2039 ct >= c->_c.mfc_un.res.minvif; ct--) {
660b26dc 2040 /* For (*,G) entry, don't forward to the incoming interface */
494fff56 2041 if ((c->mfc_origin != htonl(INADDR_ANY) ||
360eb5da 2042 ct != true_vifi) &&
494fff56 2043 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
1da177e4
LT		 2044 if (psend != -1) {
2045 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
a8cb16dd 2046
1da177e4 2047 if (skb2)
a5bc9294 2048 ipmr_queue_xmit(net, mrt, true_vifi,
6e0735d1 2049 skb2, psend);
1da177e4 2050 }
c354e124 2051 psend = ct;
1da177e4
LT		 2052 }
2053 }
660b26dc 2054last_forward:
1da177e4
LT		 2055 if (psend != -1) {
2056 if (local) {
2057 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
a8cb16dd 2058
1da177e4 2059 if (skb2)
a5bc9294 2060 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
6e0735d1 2061 psend);
1da177e4 2062 } else {
6e0735d1 2063 ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
c4854ec8 2064 return;
1da177e4
LT		 2065 }
2066 }
2067
2068dont_forward:
2069 if (!local)
2070 kfree_skb(skb);
1da177e4
LT		 2071}
2072
417da66f 2073static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
ee3f1aaf 2074{
417da66f
DM		 2075 struct rtable *rt = skb_rtable(skb);
2076 struct iphdr *iph = ip_hdr(skb);
da91981b 2077 struct flowi4 fl4 = {
417da66f
DM		 2078 .daddr = iph->daddr,
2079 .saddr = iph->saddr,
b0fe4a31 2080 .flowi4_tos = RT_TOS(iph->tos),
4fd551d7
DM		 2081 .flowi4_oif = (rt_is_output_route(rt) ?
2082 skb->dev->ifindex : 0),
2083 .flowi4_iif = (rt_is_output_route(rt) ?
1fb9489b 2084 LOOPBACK_IFINDEX :
4fd551d7 2085 skb->dev->ifindex),
b4869889 2086 .flowi4_mark = skb->mark,
ee3f1aaf
DM		 2087 };
2088 struct mr_table *mrt;
2089 int err;
2090
da91981b 2091 err = ipmr_fib_lookup(net, &fl4, &mrt);
ee3f1aaf
DM		 2092 if (err)
2093 return ERR_PTR(err);
2094 return mrt;
2095}
1da177e4 2096
7ef8f65d
NA		 2097/* Multicast packets for forwarding arrive here
2098 * Called with rcu_read_lock();
1da177e4 2099 */
1da177e4
LT		 2100int ip_mr_input(struct sk_buff *skb)
2101{
2102 struct mfc_cache *cache;
4feb88e5 2103 struct net *net = dev_net(skb->dev);
511c3f92 2104 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
f0ad0860 2105 struct mr_table *mrt;
bcfc7d33
TW		 2106 struct net_device *dev;
2107
2108 /* skb->dev passed in is the loX master dev for vrfs.
2109 * As there are no vifs associated with loopback devices,
2110 * get the proper interface that does have a vif associated with it.
2111 */
2112 dev = skb->dev;
2113 if (netif_is_l3_master(skb->dev)) {
2114 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2115 if (!dev) {
2116 kfree_skb(skb);
2117 return -ENODEV;
2118 }
2119 }
1da177e4
LT		 2120
2121 /* Packet is looped back after forward, it should not be
a8cb16dd 2122 * forwarded second time, but still can be delivered locally.
1da177e4 2123 */
4c968709 2124 if (IPCB(skb)->flags & IPSKB_FORWARDED)
1da177e4
LT		 2125 goto dont_forward;
2126
417da66f 2127 mrt = ipmr_rt_fib_lookup(net, skb);
ee3f1aaf
DM		 2128 if (IS_ERR(mrt)) {
2129 kfree_skb(skb);
2130 return PTR_ERR(mrt);
e40dbc51 2131 }
1da177e4 2132 if (!local) {
4c968709
ED		 2133 if (IPCB(skb)->opt.router_alert) {
2134 if (ip_call_ra_chain(skb))
2135 return 0;
2136 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2137 /* IGMPv1 (and broken IGMPv2 implementations sort of
2138 * Cisco IOS <= 11.2(8)) do not put router alert
2139 * option to IGMP packets destined to routable
2140 * groups. It is very bad, because it means
2141 * that we can forward NO IGMP messages.
2142 */
2143 struct sock *mroute_sk;
2144
2145 mroute_sk = rcu_dereference(mrt->mroute_sk);
2146 if (mroute_sk) {
895b5c9f 2147 nf_reset_ct(skb);
4c968709
ED		 2148 raw_rcv(mroute_sk, skb);
2149 return 0;
2150 }
1da177e4
LT		 2151 }
2152 }
2153
a8c9486b 2154 /* already under rcu_read_lock() */
0c12295a 2155 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
51456b29 2156 if (!cache) {
bcfc7d33 2157 int vif = ipmr_find_vif(mrt, dev);
660b26dc
ND		 2158
2159 if (vif >= 0)
2160 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2161 vif);
2162 }
1da177e4 2163
7ef8f65d 2164 /* No usable cache entry */
51456b29 2165 if (!cache) {
1da177e4
LT		 2166 int vif;
2167
2168 if (local) {
2169 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2170 ip_local_deliver(skb);
51456b29 2171 if (!skb2)
1da177e4 2172 return -ENOBUFS;
1da177e4
LT		 2173 skb = skb2;
2174 }
2175
a8c9486b 2176 read_lock(&mrt_lock);
bcfc7d33 2177 vif = ipmr_find_vif(mrt, dev);
1da177e4 2178 if (vif >= 0) {
4b1f0d33 2179 int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
1da177e4
LT		 2180 read_unlock(&mrt_lock);
2181
0eae88f3 2182 return err2;
1da177e4
LT		 2183 }
2184 read_unlock(&mrt_lock);
2185 kfree_skb(skb);
2186 return -ENODEV;
2187 }
2188
a8c9486b 2189 read_lock(&mrt_lock);
4b1f0d33 2190 ip_mr_forward(net, mrt, dev, skb, cache, local);
1da177e4
LT		 2191 read_unlock(&mrt_lock);
2192
2193 if (local)
2194 return ip_local_deliver(skb);
2195
2196 return 0;
2197
2198dont_forward:
2199 if (local)
2200 return ip_local_deliver(skb);
2201 kfree_skb(skb);
2202 return 0;
2203}
2204
b1879204 2205#ifdef CONFIG_IP_PIMSM_V1
7ef8f65d 2206/* Handle IGMP messages of PIMv1 */
a8cb16dd 2207int pim_rcv_v1(struct sk_buff *skb)
b1879204
IJ		 2208{
2209 struct igmphdr *pim;
4feb88e5 2210 struct net *net = dev_net(skb->dev);
f0ad0860 2211 struct mr_table *mrt;
b1879204
IJ		 2212
2213 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2214 goto drop;
2215
2216 pim = igmp_hdr(skb);
2217
417da66f 2218 mrt = ipmr_rt_fib_lookup(net, skb);
ee3f1aaf
DM		 2219 if (IS_ERR(mrt))
2220 goto drop;
0c12295a 2221 if (!mrt->mroute_do_pim ||
b1879204
IJ		 2222 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2223 goto drop;
2224
f0ad0860 2225 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
b1879204
IJ		 2226drop:
2227 kfree_skb(skb);
2228 }
1da177e4
LT		 2229 return 0;
2230}
2231#endif
2232
2233#ifdef CONFIG_IP_PIMSM_V2
a8cb16dd 2234static int pim_rcv(struct sk_buff *skb)
1da177e4
LT		 2235{
2236 struct pimreghdr *pim;
f0ad0860
PM		 2237 struct net *net = dev_net(skb->dev);
2238 struct mr_table *mrt;
1da177e4 2239
b1879204 2240 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
1da177e4
LT		 2241 goto drop;
2242
9c70220b 2243 pim = (struct pimreghdr *)skb_transport_header(skb);
56245cae 2244 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
a8cb16dd 2245 (pim->flags & PIM_NULL_REGISTER) ||
e905a9ed 2246 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
d3bc23e7 2247 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
1da177e4
LT		 2248 goto drop;
2249
417da66f 2250 mrt = ipmr_rt_fib_lookup(net, skb);
ee3f1aaf
DM		 2251 if (IS_ERR(mrt))
2252 goto drop;
f0ad0860 2253 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
b1879204
IJ		 2254drop:
2255 kfree_skb(skb);
2256 }
1da177e4
LT		 2257 return 0;
2258}
2259#endif
2260
9a1b9496
DM		 2261int ipmr_get_route(struct net *net, struct sk_buff *skb,
2262 __be32 saddr, __be32 daddr,
9f09eaea 2263 struct rtmsg *rtm, u32 portid)
1da177e4 2264{
1da177e4 2265 struct mfc_cache *cache;
9a1b9496
DM		 2266 struct mr_table *mrt;
2267 int err;
1da177e4 2268
f0ad0860 2269 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
51456b29 2270 if (!mrt)
f0ad0860
PM		 2271 return -ENOENT;
2272
a8c9486b 2273 rcu_read_lock();
9a1b9496 2274 cache = ipmr_cache_find(mrt, saddr, daddr);
51456b29 2275 if (!cache && skb->dev) {
660b26dc 2276 int vif = ipmr_find_vif(mrt, skb->dev);
1da177e4 2277
660b26dc
ND		 2278 if (vif >= 0)
2279 cache = ipmr_cache_find_any(mrt, daddr, vif);
2280 }
51456b29 2281 if (!cache) {
72287490 2282 struct sk_buff *skb2;
eddc9ec5 2283 struct iphdr *iph;
1da177e4 2284 struct net_device *dev;
a8cb16dd 2285 int vif = -1;
1da177e4 2286
1da177e4 2287 dev = skb->dev;
a8c9486b 2288 read_lock(&mrt_lock);
a8cb16dd
ED		 2289 if (dev)
2290 vif = ipmr_find_vif(mrt, dev);
2291 if (vif < 0) {
1da177e4 2292 read_unlock(&mrt_lock);
a8c9486b 2293 rcu_read_unlock();
1da177e4
LT		 2294 return -ENODEV;
2295 }
7901cd97
GN		 2296
2297 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
72287490
AK		 2298 if (!skb2) {
2299 read_unlock(&mrt_lock);
a8c9486b 2300 rcu_read_unlock();
72287490
AK		 2301 return -ENOMEM;
2302 }
2303
2cf75070 2304 NETLINK_CB(skb2).portid = portid;
e2d1bca7
ACM		 2305 skb_push(skb2, sizeof(struct iphdr));
2306 skb_reset_network_header(skb2);
eddc9ec5
ACM		 2307 iph = ip_hdr(skb2);
2308 iph->ihl = sizeof(struct iphdr) >> 2;
9a1b9496
DM		 2309 iph->saddr = saddr;
2310 iph->daddr = daddr;
eddc9ec5 2311 iph->version = 0;
4b1f0d33 2312 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
1da177e4 2313 read_unlock(&mrt_lock);
a8c9486b 2314 rcu_read_unlock();
1da177e4
LT		 2315 return err;
2316 }
2317
a8c9486b 2318 read_lock(&mrt_lock);
7b0db857 2319 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
1da177e4 2320 read_unlock(&mrt_lock);
a8c9486b 2321 rcu_read_unlock();
1da177e4
LT		 2322 return err;
2323}
2324
cb6a4e46 2325static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
65886f43
ND		 2326 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2327 int flags)
cb6a4e46
PM		 2328{
2329 struct nlmsghdr *nlh;
2330 struct rtmsg *rtm;
1eb99af5 2331 int err;
cb6a4e46 2332
65886f43 2333 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
51456b29 2334 if (!nlh)
cb6a4e46
PM		 2335 return -EMSGSIZE;
2336
2337 rtm = nlmsg_data(nlh);
2338 rtm->rtm_family = RTNL_FAMILY_IPMR;
2339 rtm->rtm_dst_len = 32;
2340 rtm->rtm_src_len = 32;
2341 rtm->rtm_tos = 0;
2342 rtm->rtm_table = mrt->id;
f3756b79
DM		 2343 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2344 goto nla_put_failure;
cb6a4e46
PM		 2345 rtm->rtm_type = RTN_MULTICAST;
2346 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
494fff56 2347 if (c->_c.mfc_flags & MFC_STATIC)
9a68ac72
ND		 2348 rtm->rtm_protocol = RTPROT_STATIC;
2349 else
2350 rtm->rtm_protocol = RTPROT_MROUTED;
cb6a4e46
PM		 2351 rtm->rtm_flags = 0;
2352
930345ea
JB		 2353 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2354 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
f3756b79 2355 goto nla_put_failure;
7b0db857 2356 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
1eb99af5
ND		 2357 /* do not break the dump if cache is unresolved */
2358 if (err < 0 && err != -ENOENT)
cb6a4e46
PM		 2359 goto nla_put_failure;
2360
053c095a
JB		 2361 nlmsg_end(skb, nlh);
2362 return 0;
cb6a4e46
PM		 2363
2364nla_put_failure:
2365 nlmsg_cancel(skb, nlh);
2366 return -EMSGSIZE;
2367}
2368
7b0db857
YM		 2369static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2370 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2371 int flags)
2372{
2373 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2374 cmd, flags);
2375}
2376
8cd3ac9f
ND		 2377static size_t mroute_msgsize(bool unresolved, int maxvif)
2378{
2379 size_t len =
2380 NLMSG_ALIGN(sizeof(struct rtmsg))
2381 + nla_total_size(4) /* RTA_TABLE */
2382 + nla_total_size(4) /* RTA_SRC */
2383 + nla_total_size(4) /* RTA_DST */
2384 ;
2385
2386 if (!unresolved)
2387 len = len
2388 + nla_total_size(4) /* RTA_IIF */
2389 + nla_total_size(0) /* RTA_MULTIPATH */
2390 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2391 /* RTA_MFC_STATS */
a9a08042 2392 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
8cd3ac9f
ND		 2393 ;
2394
2395 return len;
2396}
2397
2398static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2399 int cmd)
2400{
2401 struct net *net = read_pnet(&mrt->net);
2402 struct sk_buff *skb;
2403 int err = -ENOBUFS;
2404
494fff56
YM		 2405 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2406 mrt->maxvif),
8cd3ac9f 2407 GFP_ATOMIC);
51456b29 2408 if (!skb)
8cd3ac9f
ND		 2409 goto errout;
2410
65886f43 2411 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
8cd3ac9f
ND		 2412 if (err < 0)
2413 goto errout;
2414
2415 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2416 return;
2417
2418errout:
2419 kfree_skb(skb);
2420 if (err < 0)
2421 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2422}
2423
5a645dd8
JG		 2424static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2425{
2426 size_t len =
2427 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2428 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2429 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2430 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2431 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2432 /* IPMRA_CREPORT_PKT */
2433 + nla_total_size(payloadlen)
2434 ;
2435
2436 return len;
2437}
2438
2439static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2440{
2441 struct net *net = read_pnet(&mrt->net);
2442 struct nlmsghdr *nlh;
2443 struct rtgenmsg *rtgenm;
2444 struct igmpmsg *msg;
2445 struct sk_buff *skb;
2446 struct nlattr *nla;
2447 int payloadlen;
2448
2449 payloadlen = pkt->len - sizeof(struct igmpmsg);
2450 msg = (struct igmpmsg *)skb_network_header(pkt);
2451
2452 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2453 if (!skb)
2454 goto errout;
2455
2456 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2457 sizeof(struct rtgenmsg), 0);
2458 if (!nlh)
2459 goto errout;
2460 rtgenm = nlmsg_data(nlh);
2461 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2462 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2463 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif) ||
2464 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2465 msg->im_src.s_addr) ||
2466 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2467 msg->im_dst.s_addr))
2468 goto nla_put_failure;
2469
2470 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2471 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2472 nla_data(nla), payloadlen))
2473 goto nla_put_failure;
2474
2475 nlmsg_end(skb, nlh);
2476
2477 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2478 return;
2479
2480nla_put_failure:
2481 nlmsg_cancel(skb, nlh);
2482errout:
2483 kfree_skb(skb);
2484 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2485}
2486
d0440029
JK		 2487static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2488 const struct nlmsghdr *nlh,
2489 struct nlattr **tb,
2490 struct netlink_ext_ack *extack)
2491{
2492 struct rtmsg *rtm;
2493 int i, err;
2494
2495 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2496 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2497 return -EINVAL;
2498 }
2499
2500 if (!netlink_strict_get_check(skb))
8cb08174
JB		 2501 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2502 rtm_ipv4_policy, extack);
d0440029
JK		 2503
2504 rtm = nlmsg_data(nlh);
2505 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2506 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2507 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2508 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2509 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2510 return -EINVAL;
2511 }
2512
8cb08174
JB		 2513 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2514 rtm_ipv4_policy, extack);
d0440029
JK		 2515 if (err)
2516 return err;
2517
2518 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2519 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2520 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2521 return -EINVAL;
2522 }
2523
2524 for (i = 0; i <= RTA_MAX; i++) {
2525 if (!tb[i])
2526 continue;
2527
2528 switch (i) {
2529 case RTA_SRC:
2530 case RTA_DST:
2531 case RTA_TABLE:
2532 break;
2533 default:
2534 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2535 return -EINVAL;
2536 }
2537 }
2538
2539 return 0;
2540}
2541
4f75ba69
DS		 2542static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2543 struct netlink_ext_ack *extack)
2544{
2545 struct net *net = sock_net(in_skb->sk);
2546 struct nlattr *tb[RTA_MAX + 1];
2547 struct sk_buff *skb = NULL;
2548 struct mfc_cache *cache;
2549 struct mr_table *mrt;
4f75ba69
DS		 2550 __be32 src, grp;
2551 u32 tableid;
2552 int err;
2553
d0440029 2554 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
4f75ba69
DS		 2555 if (err < 0)
2556 goto errout;
2557
4f75ba69
DS		 2558 src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2559 grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2560 tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2561
2562 mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2e3d232e
DC		 2563 if (!mrt) {
2564 err = -ENOENT;
4f75ba69
DS		 2565 goto errout_free;
2566 }
2567
2568 /* entries are added/deleted only under RTNL */
2569 rcu_read_lock();
2570 cache = ipmr_cache_find(mrt, src, grp);
2571 rcu_read_unlock();
2572 if (!cache) {
2573 err = -ENOENT;
2574 goto errout_free;
2575 }
2576
2577 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2578 if (!skb) {
2579 err = -ENOBUFS;
2580 goto errout_free;
2581 }
2582
2583 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2584 nlh->nlmsg_seq, cache,
2585 RTM_NEWROUTE, 0);
2586 if (err < 0)
2587 goto errout_free;
2588
2589 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2590
2591errout:
2592 return err;
2593
2594errout_free:
2595 kfree_skb(skb);
2596 goto errout;
2597}
2598
cb6a4e46
PM		 2599static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2600{
4724676d 2601 struct fib_dump_filter filter = {};
cb167893 2602 int err;
4724676d 2603
e8ba330a 2604 if (cb->strict_check) {
4724676d 2605 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
effe6792 2606 &filter, cb);
e8ba330a
DA		 2607 if (err < 0)
2608 return err;
2609 }
2610
cb167893
DA		 2611 if (filter.table_id) {
2612 struct mr_table *mrt;
2613
2614 mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2615 if (!mrt) {
ae677bbb
DA		 2616 if (filter.dump_all_families)
2617 return skb->len;
2618
cb167893
DA		 2619 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2620 return -ENOENT;
2621 }
2622 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2623 &mfc_unres_lock, &filter);
2624 return skb->len ? : err;
2625 }
2626
7b0db857 2627 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
cb167893 2628 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
cb6a4e46
PM		 2629}
2630
ccbb0aa6
NA		 2631static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2632 [RTA_SRC] = { .type = NLA_U32 },
2633 [RTA_DST] = { .type = NLA_U32 },
2634 [RTA_IIF] = { .type = NLA_U32 },
2635 [RTA_TABLE] = { .type = NLA_U32 },
2636 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2637};
2638
2639static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2640{
2641 switch (rtm_protocol) {
2642 case RTPROT_STATIC:
2643 case RTPROT_MROUTED:
2644 return true;
2645 }
2646 return false;
2647}
2648
2649static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2650{
2651 struct rtnexthop *rtnh = nla_data(nla);
2652 int remaining = nla_len(nla), vifi = 0;
2653
2654 while (rtnh_ok(rtnh, remaining)) {
2655 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2656 if (++vifi == MAXVIFS)
2657 break;
2658 rtnh = rtnh_next(rtnh, &remaining);
2659 }
2660
2661 return remaining > 0 ? -EINVAL : vifi;
2662}
2663
2664/* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2665static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2666 struct mfcctl *mfcc, int *mrtsock,
c21ef3e3
DA		 2667 struct mr_table **mrtret,
2668 struct netlink_ext_ack *extack)
ccbb0aa6
NA		 2669{
2670 struct net_device *dev = NULL;
2671 u32 tblid = RT_TABLE_DEFAULT;
2672 struct mr_table *mrt;
2673 struct nlattr *attr;
2674 struct rtmsg *rtm;
2675 int ret, rem;
2676
8cb08174
JB		 2677 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2678 rtm_ipmr_policy, extack);
ccbb0aa6
NA		 2679 if (ret < 0)
2680 goto out;
2681 rtm = nlmsg_data(nlh);
2682
2683 ret = -EINVAL;
2684 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2685 rtm->rtm_type != RTN_MULTICAST ||
2686 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2687 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2688 goto out;
2689
2690 memset(mfcc, 0, sizeof(*mfcc));
2691 mfcc->mfcc_parent = -1;
2692 ret = 0;
2693 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2694 switch (nla_type(attr)) {
2695 case RTA_SRC:
2696 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2697 break;
2698 case RTA_DST:
2699 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2700 break;
2701 case RTA_IIF:
2702 dev = __dev_get_by_index(net, nla_get_u32(attr));
2703 if (!dev) {
2704 ret = -ENODEV;
2705 goto out;
2706 }
2707 break;
2708 case RTA_MULTIPATH:
2709 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2710 ret = -EINVAL;
2711 goto out;
2712 }
2713 break;
2714 case RTA_PREFSRC:
2715 ret = 1;
2716 break;
2717 case RTA_TABLE:
2718 tblid = nla_get_u32(attr);
2719 break;
2720 }
2721 }
2722 mrt = ipmr_get_table(net, tblid);
2723 if (!mrt) {
2724 ret = -ENOENT;
2725 goto out;
2726 }
2727 *mrtret = mrt;
2728 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2729 if (dev)
2730 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2731
2732out:
2733 return ret;
2734}
2735
2736/* takes care of both newroute and delroute */
c21ef3e3
DA		 2737static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2738 struct netlink_ext_ack *extack)
ccbb0aa6
NA		 2739{
2740 struct net *net = sock_net(skb->sk);
2741 int ret, mrtsock, parent;
2742 struct mr_table *tbl;
2743 struct mfcctl mfcc;
2744
2745 mrtsock = 0;
2746 tbl = NULL;
c21ef3e3 2747 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
ccbb0aa6
NA		 2748 if (ret < 0)
2749 return ret;
2750
2751 parent = ret ? mfcc.mfcc_parent : -1;
2752 if (nlh->nlmsg_type == RTM_NEWROUTE)
2753 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2754 else
2755 return ipmr_mfc_delete(tbl, &mfcc, parent);
2756}
2757
772c344d
NA		 2758static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2759{
2760 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2761
2762 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2763 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2764 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2765 mrt->mroute_reg_vif_num) ||
2766 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2767 mrt->mroute_do_assert) ||
c921c207
NA		 2768 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2769 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2770 mrt->mroute_do_wrvifwhole))
772c344d
NA		 2771 return false;
2772
2773 return true;
2774}
2775
2776static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2777{
2778 struct nlattr *vif_nest;
2779 struct vif_device *vif;
2780
2781 /* if the VIF doesn't exist just continue */
2782 if (!VIF_EXISTS(mrt, vifid))
2783 return true;
2784
2785 vif = &mrt->vif_table[vifid];
ae0be8de 2786 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
772c344d
NA		 2787 if (!vif_nest)
2788 return false;
2789 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2790 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2791 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2792 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2793 IPMRA_VIFA_PAD) ||
2794 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2795 IPMRA_VIFA_PAD) ||
2796 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2797 IPMRA_VIFA_PAD) ||
2798 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2799 IPMRA_VIFA_PAD) ||
2800 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2801 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2802 nla_nest_cancel(skb, vif_nest);
2803 return false;
2804 }
2805 nla_nest_end(skb, vif_nest);
2806
2807 return true;
2808}
2809
14fc5bb2
DA		 2810static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2811 struct netlink_ext_ack *extack)
2812{
2813 struct ifinfomsg *ifm;
2814
2815 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2816 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2817 return -EINVAL;
2818 }
2819
2820 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2821 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2822 return -EINVAL;
2823 }
2824
2825 ifm = nlmsg_data(nlh);
2826 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2827 ifm->ifi_change || ifm->ifi_index) {
2828 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2829 return -EINVAL;
2830 }
2831
2832 return 0;
2833}
2834
772c344d
NA		 2835static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2836{
2837 struct net *net = sock_net(skb->sk);
2838 struct nlmsghdr *nlh = NULL;
2839 unsigned int t = 0, s_t;
2840 unsigned int e = 0, s_e;
2841 struct mr_table *mrt;
2842
14fc5bb2
DA		 2843 if (cb->strict_check) {
2844 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2845
2846 if (err < 0)
2847 return err;
2848 }
2849
772c344d
NA		 2850 s_t = cb->args[0];
2851 s_e = cb->args[1];
2852
2853 ipmr_for_each_table(mrt, net) {
2854 struct nlattr *vifs, *af;
2855 struct ifinfomsg *hdr;
2856 u32 i;
2857
2858 if (t < s_t)
2859 goto skip_table;
2860 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2861 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2862 sizeof(*hdr), NLM_F_MULTI);
2863 if (!nlh)
2864 break;
2865
2866 hdr = nlmsg_data(nlh);
2867 memset(hdr, 0, sizeof(*hdr));
2868 hdr->ifi_family = RTNL_FAMILY_IPMR;
2869
ae0be8de 2870 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
772c344d
NA		 2871 if (!af) {
2872 nlmsg_cancel(skb, nlh);
2873 goto out;
2874 }
2875
2876 if (!ipmr_fill_table(mrt, skb)) {
2877 nlmsg_cancel(skb, nlh);
2878 goto out;
2879 }
2880
ae0be8de 2881 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
772c344d
NA		 2882 if (!vifs) {
2883 nla_nest_end(skb, af);
2884 nlmsg_end(skb, nlh);
2885 goto out;
2886 }
2887 for (i = 0; i < mrt->maxvif; i++) {
2888 if (e < s_e)
2889 goto skip_entry;
2890 if (!ipmr_fill_vif(mrt, i, skb)) {
2891 nla_nest_end(skb, vifs);
2892 nla_nest_end(skb, af);
2893 nlmsg_end(skb, nlh);
2894 goto out;
2895 }
2896skip_entry:
2897 e++;
2898 }
2899 s_e = 0;
2900 e = 0;
2901 nla_nest_end(skb, vifs);
2902 nla_nest_end(skb, af);
2903 nlmsg_end(skb, nlh);
2904skip_table:
2905 t++;
2906 }
2907
2908out:
2909 cb->args[1] = e;
2910 cb->args[0] = t;
2911
2912 return skb->len;
2913}
2914
e905a9ed 2915#ifdef CONFIG_PROC_FS
7ef8f65d
NA		 2916/* The /proc interfaces to multicast routing :
2917 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
1da177e4 2918 */
1da177e4
LT		 2919
2920static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
ba93ef74 2921 __acquires(mrt_lock)
1da177e4 2922{
3feda6b4 2923 struct mr_vif_iter *iter = seq->private;
f6bb4514 2924 struct net *net = seq_file_net(seq);
f0ad0860
PM		 2925 struct mr_table *mrt;
2926
2927 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
51456b29 2928 if (!mrt)
f0ad0860
PM		 2929 return ERR_PTR(-ENOENT);
2930
2931 iter->mrt = mrt;
f6bb4514 2932
1da177e4 2933 read_lock(&mrt_lock);
3feda6b4 2934 return mr_vif_seq_start(seq, pos);
1da177e4
LT		 2935}
2936
2937static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
ba93ef74 2938 __releases(mrt_lock)
1da177e4
LT		 2939{
2940 read_unlock(&mrt_lock);
2941}
2942
2943static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2944{
3feda6b4 2945 struct mr_vif_iter *iter = seq->private;
f0ad0860 2946 struct mr_table *mrt = iter->mrt;
f6bb4514 2947
1da177e4 2948 if (v == SEQ_START_TOKEN) {
e905a9ed 2949 seq_puts(seq,
1da177e4
LT		 2950 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2951 } else {
2952 const struct vif_device *vif = v;
6853f21f
YM		 2953 const char *name = vif->dev ?
2954 vif->dev->name : "none";
1da177e4
LT		 2955
2956 seq_printf(seq,
91e6dd82 2957 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
0c12295a 2958 vif - mrt->vif_table,
e905a9ed 2959 name, vif->bytes_in, vif->pkt_in,
1da177e4
LT		 2960 vif->bytes_out, vif->pkt_out,
2961 vif->flags, vif->local, vif->remote);
2962 }
2963 return 0;
2964}
2965
f690808e 2966static const struct seq_operations ipmr_vif_seq_ops = {
1da177e4 2967 .start = ipmr_vif_seq_start,
3feda6b4 2968 .next = mr_vif_seq_next,
1da177e4
LT		 2969 .stop = ipmr_vif_seq_stop,
2970 .show = ipmr_vif_seq_show,
2971};
2972
1da177e4
LT		 2973static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2974{
f6bb4514 2975 struct net *net = seq_file_net(seq);
f0ad0860 2976 struct mr_table *mrt;
f6bb4514 2977
f0ad0860 2978 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
51456b29 2979 if (!mrt)
f0ad0860 2980 return ERR_PTR(-ENOENT);
f6bb4514 2981
c8d61968 2982 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
1da177e4
LT		 2983}
2984
2985static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2986{
2987 int n;
2988
2989 if (v == SEQ_START_TOKEN) {
e905a9ed 2990 seq_puts(seq,
1da177e4
LT		 2991 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2992 } else {
2993 const struct mfc_cache *mfc = v;
c8d61968 2994 const struct mr_mfc_iter *it = seq->private;
f0ad0860 2995 const struct mr_table *mrt = it->mrt;
e905a9ed 2996
0eae88f3
ED		 2997 seq_printf(seq, "%08X %08X %-3hd",
2998 (__force u32) mfc->mfc_mcastgrp,
2999 (__force u32) mfc->mfc_origin,
494fff56 3000 mfc->_c.mfc_parent);
1da177e4 3001
0c12295a 3002 if (it->cache != &mrt->mfc_unres_queue) {
1ea472e2 3003 seq_printf(seq, " %8lu %8lu %8lu",
494fff56
YM		 3004 mfc->_c.mfc_un.res.pkt,
3005 mfc->_c.mfc_un.res.bytes,
3006 mfc->_c.mfc_un.res.wrong_if);
3007 for (n = mfc->_c.mfc_un.res.minvif;
3008 n < mfc->_c.mfc_un.res.maxvif; n++) {
0c12295a 3009 if (VIF_EXISTS(mrt, n) &&
494fff56 3010 mfc->_c.mfc_un.res.ttls[n] < 255)
cf958ae3 3011 seq_printf(seq,
e905a9ed 3012 " %2d:%-3d",
494fff56 3013 n, mfc->_c.mfc_un.res.ttls[n]);
1da177e4 3014 }
1ea472e2
BT		 3015 } else {
3016 /* unresolved mfc_caches don't contain
3017 * pkt, bytes and wrong_if values
3018 */
3019 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
1da177e4
LT		 3020 }
3021 seq_putc(seq, '\n');
3022 }
3023 return 0;
3024}
3025
f690808e 3026static const struct seq_operations ipmr_mfc_seq_ops = {
1da177e4 3027 .start = ipmr_mfc_seq_start,
c8d61968
YM		 3028 .next = mr_mfc_seq_next,
3029 .stop = mr_mfc_seq_stop,
1da177e4
LT		 3030 .show = ipmr_mfc_seq_show,
3031};
e905a9ed 3032#endif
1da177e4
LT		 3033
3034#ifdef CONFIG_IP_PIMSM_V2
32613090 3035static const struct net_protocol pim_protocol = {
1da177e4 3036 .handler = pim_rcv,
403dbb97 3037 .netns_ok = 1,
1da177e4
LT		 3038};
3039#endif
3040
4d65b948
YG		 3041static unsigned int ipmr_seq_read(struct net *net)
3042{
3043 ASSERT_RTNL();
3044
3045 return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3046}
3047
b7a59557
JP		 3048static int ipmr_dump(struct net *net, struct notifier_block *nb,
3049 struct netlink_ext_ack *extack)
4d65b948 3050{
cdc9f944 3051 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
b7a59557 3052 ipmr_mr_table_iter, &mrt_lock, extack);
4d65b948
YG		 3053}
3054
3055static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3056 .family = RTNL_FAMILY_IPMR,
3057 .fib_seq_read = ipmr_seq_read,
3058 .fib_dump = ipmr_dump,
3059 .owner = THIS_MODULE,
3060};
3061
ef739d8a 3062static int __net_init ipmr_notifier_init(struct net *net)
4d65b948
YG		 3063{
3064 struct fib_notifier_ops *ops;
3065
3066 net->ipv4.ipmr_seq = 0;
3067
3068 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3069 if (IS_ERR(ops))
3070 return PTR_ERR(ops);
3071 net->ipv4.ipmr_notifier_ops = ops;
3072
3073 return 0;
3074}
3075
3076static void __net_exit ipmr_notifier_exit(struct net *net)
3077{
3078 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3079 net->ipv4.ipmr_notifier_ops = NULL;
3080}
3081
7ef8f65d 3082/* Setup for IP multicast routing */
cf958ae3
BT		 3083static int __net_init ipmr_net_init(struct net *net)
3084{
f0ad0860 3085 int err;
cf958ae3 3086
4d65b948
YG		 3087 err = ipmr_notifier_init(net);
3088 if (err)
3089 goto ipmr_notifier_fail;
3090
f0ad0860
PM		 3091 err = ipmr_rules_init(net);
3092 if (err < 0)
4d65b948 3093 goto ipmr_rules_fail;
f6bb4514
BT		 3094
3095#ifdef CONFIG_PROC_FS
3096 err = -ENOMEM;
c3506372
CH		 3097 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3098 sizeof(struct mr_vif_iter)))
f6bb4514 3099 goto proc_vif_fail;
c3506372
CH		 3100 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3101 sizeof(struct mr_mfc_iter)))
f6bb4514
BT		 3102 goto proc_cache_fail;
3103#endif
2bb8b26c
BT		 3104 return 0;
3105
f6bb4514
BT		 3106#ifdef CONFIG_PROC_FS
3107proc_cache_fail:
ece31ffd 3108 remove_proc_entry("ip_mr_vif", net->proc_net);
f6bb4514 3109proc_vif_fail:
f0ad0860 3110 ipmr_rules_exit(net);
f6bb4514 3111#endif
4d65b948
YG		 3112ipmr_rules_fail:
3113 ipmr_notifier_exit(net);
3114ipmr_notifier_fail:
cf958ae3
BT		 3115 return err;
3116}
3117
3118static void __net_exit ipmr_net_exit(struct net *net)
3119{
f6bb4514 3120#ifdef CONFIG_PROC_FS
ece31ffd
G		 3121 remove_proc_entry("ip_mr_cache", net->proc_net);
3122 remove_proc_entry("ip_mr_vif", net->proc_net);
f6bb4514 3123#endif
4d65b948 3124 ipmr_notifier_exit(net);
f0ad0860 3125 ipmr_rules_exit(net);
cf958ae3
BT		 3126}
3127
3128static struct pernet_operations ipmr_net_ops = {
3129 .init = ipmr_net_init,
3130 .exit = ipmr_net_exit,
3131};
e905a9ed 3132
03d2f897 3133int __init ip_mr_init(void)
1da177e4 3134{
03d2f897
WC		 3135 int err;
3136
1da177e4
LT		 3137 mrt_cachep = kmem_cache_create("ip_mrt_cache",
3138 sizeof(struct mfc_cache),
a8c9486b 3139 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
20c2df83 3140 NULL);
03d2f897 3141
cf958ae3
BT		 3142 err = register_pernet_subsys(&ipmr_net_ops);
3143 if (err)
3144 goto reg_pernet_fail;
3145
03d2f897
WC		 3146 err = register_netdevice_notifier(&ip_mr_notifier);
3147 if (err)
3148 goto reg_notif_fail;
403dbb97
TG		 3149#ifdef CONFIG_IP_PIMSM_V2
3150 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
058bd4d2 3151 pr_err("%s: can't add PIM protocol\n", __func__);
403dbb97
TG		 3152 err = -EAGAIN;
3153 goto add_proto_fail;
3154 }
3155#endif
c7ac8679 3156 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
b97bac64 3157 ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
ccbb0aa6 3158 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
b97bac64 3159 ipmr_rtm_route, NULL, 0);
ccbb0aa6 3160 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
b97bac64 3161 ipmr_rtm_route, NULL, 0);
772c344d
NA		 3162
3163 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
b97bac64 3164 NULL, ipmr_rtm_dumplink, 0);
03d2f897 3165 return 0;
f6bb4514 3166
403dbb97
TG		 3167#ifdef CONFIG_IP_PIMSM_V2
3168add_proto_fail:
3169 unregister_netdevice_notifier(&ip_mr_notifier);
3170#endif
c3e38896 3171reg_notif_fail:
cf958ae3
BT		 3172 unregister_pernet_subsys(&ipmr_net_ops);
3173reg_pernet_fail:
c3e38896 3174 kmem_cache_destroy(mrt_cachep);
03d2f897 3175 return err;
1da177e4 3176}
A mirror of Linus' kernel repository