net/bridge/br_mdb.c

   1 #include <linux/err.h>
   2 #include <linux/igmp.h>
   3 #include <linux/kernel.h>
   4 #include <linux/netdevice.h>
   5 #include <linux/rculist.h>
   6 #include <linux/skbuff.h>
   7 #include <linux/if_ether.h>
   8 #include <net/ip.h>
   9 #include <net/netlink.h>
  10 #if IS_ENABLED(CONFIG_IPV6)
  11 #include <net/ipv6.h>
  12 #endif
  13
  14 #include "br_private.h"
  15
  16 static int br_rports_fill_info(struct sk_buff *skb, struct netlink_callback *cb,
  17                                struct net_device *dev)
  18 {
  19         struct net_bridge *br = netdev_priv(dev);
  20         struct net_bridge_port *p;
  21         struct hlist_node *n;
  22         struct nlattr *nest;
  23
  24         if (!br->multicast_router || hlist_empty(&br->router_list))
  25                 return 0;
  26
  27         nest = nla_nest_start(skb, MDBA_ROUTER);
  28         if (nest == NULL)
  29                 return -EMSGSIZE;
  30
  31         hlist_for_each_entry_rcu(p, n, &br->router_list, rlist) {
  32                 if (p && nla_put_u32(skb, MDBA_ROUTER_PORT, p->dev->ifindex))
  33                         goto fail;
  34         }
  35
  36         nla_nest_end(skb, nest);
  37         return 0;
  38 fail:
  39         nla_nest_cancel(skb, nest);
  40         return -EMSGSIZE;
  41 }
  42
  43 static int br_mdb_fill_info(struct sk_buff *skb, struct netlink_callback *cb,
  44                             struct net_device *dev)
  45 {
  46         struct net_bridge *br = netdev_priv(dev);
  47         struct net_bridge_mdb_htable *mdb;
  48         struct nlattr *nest, *nest2;
  49         int i, err = 0;
  50         int idx = 0, s_idx = cb->args[1];
  51
  52         if (br->multicast_disabled)
  53                 return 0;
  54
  55         mdb = rcu_dereference(br->mdb);
  56         if (!mdb)
  57                 return 0;
  58
  59         nest = nla_nest_start(skb, MDBA_MDB);
  60         if (nest == NULL)
  61                 return -EMSGSIZE;
  62
  63         for (i = 0; i < mdb->max; i++) {
  64                 struct hlist_node *h;
  65                 struct net_bridge_mdb_entry *mp;
  66                 struct net_bridge_port_group *p, **pp;
  67                 struct net_bridge_port *port;
  68
  69                 hlist_for_each_entry_rcu(mp, h, &mdb->mhash[i], hlist[mdb->ver]) {
  70                         if (idx < s_idx)
  71                                 goto skip;
  72
  73                         nest2 = nla_nest_start(skb, MDBA_MDB_ENTRY);
  74                         if (nest2 == NULL) {
  75                                 err = -EMSGSIZE;
  76                                 goto out;
  77                         }
  78
  79                         for (pp = &mp->ports;
  80                              (p = rcu_dereference(*pp)) != NULL;
  81                               pp = &p->next) {
  82                                 port = p->port;
  83                                 if (port) {
  84                                         struct br_mdb_entry e;
  85                                         e.ifindex = port->dev->ifindex;
  86                                         e.state = p->state;
  87                                         e.addr.u.ip4 = p->addr.u.ip4;
  88 #if IS_ENABLED(CONFIG_IPV6)
  89                                         e.addr.u.ip6 = p->addr.u.ip6;
  90 #endif
  91                                         e.addr.proto = p->addr.proto;
  92                                         if (nla_put(skb, MDBA_MDB_ENTRY_INFO, sizeof(e), &e)) {
  93                                                 nla_nest_cancel(skb, nest2);
  94                                                 err = -EMSGSIZE;
  95                                                 goto out;
  96                                         }
  97                                 }
  98                         }
  99                         nla_nest_end(skb, nest2);
 100                 skip:
 101                         idx++;
 102                 }
 103         }
 104
 105 out:
 106         cb->args[1] = idx;
 107         nla_nest_end(skb, nest);
 108         return err;
 109 }
 110
 111 static int br_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
 112 {
 113         struct net_device *dev;
 114         struct net *net = sock_net(skb->sk);
 115         struct nlmsghdr *nlh = NULL;
 116         int idx = 0, s_idx;
 117
 118         s_idx = cb->args[0];
 119
 120         rcu_read_lock();
 121
 122         /* In theory this could be wrapped to 0... */
 123         cb->seq = net->dev_base_seq + br_mdb_rehash_seq;
 124
 125         for_each_netdev_rcu(net, dev) {
 126                 if (dev->priv_flags & IFF_EBRIDGE) {
 127                         struct br_port_msg *bpm;
 128
 129                         if (idx < s_idx)
 130                                 goto skip;
 131
 132                         nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
 133                                         cb->nlh->nlmsg_seq, RTM_GETMDB,
 134                                         sizeof(*bpm), NLM_F_MULTI);
 135                         if (nlh == NULL)
 136                                 break;
 137
 138                         bpm = nlmsg_data(nlh);
 139                         bpm->ifindex = dev->ifindex;
 140                         if (br_mdb_fill_info(skb, cb, dev) < 0)
 141                                 goto out;
 142                         if (br_rports_fill_info(skb, cb, dev) < 0)
 143                                 goto out;
 144
 145                         cb->args[1] = 0;
 146                         nlmsg_end(skb, nlh);
 147                 skip:
 148                         idx++;
 149                 }
 150         }
 151
 152 out:
 153         if (nlh)
 154                 nlmsg_end(skb, nlh);
 155         rcu_read_unlock();
 156         cb->args[0] = idx;
 157         return skb->len;
 158 }
 159
 160 static int nlmsg_populate_mdb_fill(struct sk_buff *skb,
 161                                    struct net_device *dev,
 162                                    struct br_mdb_entry *entry, u32 pid,
 163                                    u32 seq, int type, unsigned int flags)
 164 {
 165         struct nlmsghdr *nlh;
 166         struct br_port_msg *bpm;
 167         struct nlattr *nest, *nest2;
 168
 169         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*bpm), NLM_F_MULTI);
 170         if (!nlh)
 171                 return -EMSGSIZE;
 172
 173         bpm = nlmsg_data(nlh);
 174         bpm->family  = AF_BRIDGE;
 175         bpm->ifindex = dev->ifindex;
 176         nest = nla_nest_start(skb, MDBA_MDB);
 177         if (nest == NULL)
 178                 goto cancel;
 179         nest2 = nla_nest_start(skb, MDBA_MDB_ENTRY);
 180         if (nest2 == NULL)
 181                 goto end;
 182
 183         if (nla_put(skb, MDBA_MDB_ENTRY_INFO, sizeof(*entry), entry))
 184                 goto end;
 185
 186         nla_nest_end(skb, nest2);
 187         nla_nest_end(skb, nest);
 188         return nlmsg_end(skb, nlh);
 189
 190 end:
 191         nla_nest_end(skb, nest);
 192 cancel:
 193         nlmsg_cancel(skb, nlh);
 194         return -EMSGSIZE;
 195 }
 196
 197 static inline size_t rtnl_mdb_nlmsg_size(void)
 198 {
 199         return NLMSG_ALIGN(sizeof(struct br_port_msg))
 200                 + nla_total_size(sizeof(struct br_mdb_entry));
 201 }
 202
 203 static void __br_mdb_notify(struct net_device *dev, struct br_mdb_entry *entry,
 204                             int type)
 205 {
 206         struct net *net = dev_net(dev);
 207         struct sk_buff *skb;
 208         int err = -ENOBUFS;
 209
 210         skb = nlmsg_new(rtnl_mdb_nlmsg_size(), GFP_ATOMIC);
 211         if (!skb)
 212                 goto errout;
 213
 214         err = nlmsg_populate_mdb_fill(skb, dev, entry, 0, 0, type, NTF_SELF);
 215         if (err < 0) {
 216                 kfree_skb(skb);
 217                 goto errout;
 218         }
 219
 220         rtnl_notify(skb, net, 0, RTNLGRP_MDB, NULL, GFP_ATOMIC);
 221         return;
 222 errout:
 223         rtnl_set_sk_err(net, RTNLGRP_MDB, err);
 224 }
 225
 226 void br_mdb_notify(struct net_device *dev, struct net_bridge_port *port,
 227                    struct br_ip *group, int type)
 228 {
 229         struct br_mdb_entry entry;
 230
 231         entry.ifindex = port->dev->ifindex;
 232         entry.addr.proto = group->proto;
 233         entry.addr.u.ip4 = group->u.ip4;
 234 #if IS_ENABLED(CONFIG_IPV6)
 235         entry.addr.u.ip6 = group->u.ip6;
 236 #endif
 237         __br_mdb_notify(dev, &entry, type);
 238 }
 239
 240 static bool is_valid_mdb_entry(struct br_mdb_entry *entry)
 241 {
 242         if (entry->ifindex == 0)
 243                 return false;
 244
 245         if (entry->addr.proto == htons(ETH_P_IP)) {
 246                 if (!ipv4_is_multicast(entry->addr.u.ip4))
 247                         return false;
 248                 if (ipv4_is_local_multicast(entry->addr.u.ip4))
 249                         return false;
 250 #if IS_ENABLED(CONFIG_IPV6)
 251         } else if (entry->addr.proto == htons(ETH_P_IPV6)) {
 252                 if (!ipv6_is_transient_multicast(&entry->addr.u.ip6))
 253                         return false;
 254 #endif
 255         } else
 256                 return false;
 257         if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY)
 258                 return false;
 259
 260         return true;
 261 }
 262
 263 static int br_mdb_parse(struct sk_buff *skb, struct nlmsghdr *nlh,
 264                         struct net_device **pdev, struct br_mdb_entry **pentry)
 265 {
 266         struct net *net = sock_net(skb->sk);
 267         struct br_mdb_entry *entry;
 268         struct br_port_msg *bpm;
 269         struct nlattr *tb[MDBA_SET_ENTRY_MAX+1];
 270         struct net_device *dev;
 271         int err;
 272
 273         if (!capable(CAP_NET_ADMIN))
 274                 return -EPERM;
 275
 276         err = nlmsg_parse(nlh, sizeof(*bpm), tb, MDBA_SET_ENTRY, NULL);
 277         if (err < 0)
 278                 return err;
 279
 280         bpm = nlmsg_data(nlh);
 281         if (bpm->ifindex == 0) {
 282                 pr_info("PF_BRIDGE: br_mdb_parse() with invalid ifindex\n");
 283                 return -EINVAL;
 284         }
 285
 286         dev = __dev_get_by_index(net, bpm->ifindex);
 287         if (dev == NULL) {
 288                 pr_info("PF_BRIDGE: br_mdb_parse() with unknown ifindex\n");
 289                 return -ENODEV;
 290         }
 291
 292         if (!(dev->priv_flags & IFF_EBRIDGE)) {
 293                 pr_info("PF_BRIDGE: br_mdb_parse() with non-bridge\n");
 294                 return -EOPNOTSUPP;
 295         }
 296
 297         *pdev = dev;
 298
 299         if (!tb[MDBA_SET_ENTRY] ||
 300             nla_len(tb[MDBA_SET_ENTRY]) != sizeof(struct br_mdb_entry)) {
 301                 pr_info("PF_BRIDGE: br_mdb_parse() with invalid attr\n");
 302                 return -EINVAL;
 303         }
 304
 305         entry = nla_data(tb[MDBA_SET_ENTRY]);
 306         if (!is_valid_mdb_entry(entry)) {
 307                 pr_info("PF_BRIDGE: br_mdb_parse() with invalid entry\n");
 308                 return -EINVAL;
 309         }
 310
 311         *pentry = entry;
 312         return 0;
 313 }
 314
 315 static int br_mdb_add_group(struct net_bridge *br, struct net_bridge_port *port,
 316                             struct br_ip *group, unsigned char state)
 317 {
 318         struct net_bridge_mdb_entry *mp;
 319         struct net_bridge_port_group *p;
 320         struct net_bridge_port_group __rcu **pp;
 321         struct net_bridge_mdb_htable *mdb;
 322         int err;
 323
 324         mdb = mlock_dereference(br->mdb, br);
 325         mp = br_mdb_ip_get(mdb, group);
 326         if (!mp) {
 327                 mp = br_multicast_new_group(br, port, group);
 328                 err = PTR_ERR(mp);
 329                 if (IS_ERR(mp))
 330                         return err;
 331         }
 332
 333         for (pp = &mp->ports;
 334              (p = mlock_dereference(*pp, br)) != NULL;
 335              pp = &p->next) {
 336                 if (p->port == port)
 337                         return -EEXIST;
 338                 if ((unsigned long)p->port < (unsigned long)port)
 339                         break;
 340         }
 341
 342         p = br_multicast_new_port_group(port, group, *pp, state);
 343         if (unlikely(!p))
 344                 return -ENOMEM;
 345         rcu_assign_pointer(*pp, p);
 346
 347         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 348         return 0;
 349 }
 350
 351 static int __br_mdb_add(struct net *net, struct net_bridge *br,
 352                         struct br_mdb_entry *entry)
 353 {
 354         struct br_ip ip;
 355         struct net_device *dev;
 356         struct net_bridge_port *p;
 357         int ret;
 358
 359         if (!netif_running(br->dev) || br->multicast_disabled)
 360                 return -EINVAL;
 361
 362         dev = __dev_get_by_index(net, entry->ifindex);
 363         if (!dev)
 364                 return -ENODEV;
 365
 366         p = br_port_get_rtnl(dev);
 367         if (!p || p->br != br || p->state == BR_STATE_DISABLED)
 368                 return -EINVAL;
 369
 370         ip.proto = entry->addr.proto;
 371         if (ip.proto == htons(ETH_P_IP))
 372                 ip.u.ip4 = entry->addr.u.ip4;
 373 #if IS_ENABLED(CONFIG_IPV6)
 374         else
 375                 ip.u.ip6 = entry->addr.u.ip6;
 376 #endif
 377
 378         spin_lock_bh(&br->multicast_lock);
 379         ret = br_mdb_add_group(br, p, &ip, entry->state);
 380         spin_unlock_bh(&br->multicast_lock);
 381         return ret;
 382 }
 383
 384 static int br_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 385 {
 386         struct net *net = sock_net(skb->sk);
 387         struct br_mdb_entry *entry;
 388         struct net_device *dev;
 389         struct net_bridge *br;
 390         int err;
 391
 392         err = br_mdb_parse(skb, nlh, &dev, &entry);
 393         if (err < 0)
 394                 return err;
 395
 396         br = netdev_priv(dev);
 397
 398         err = __br_mdb_add(net, br, entry);
 399         if (!err)
 400                 __br_mdb_notify(dev, entry, RTM_NEWMDB);
 401         return err;
 402 }
 403
 404 static int __br_mdb_del(struct net_bridge *br, struct br_mdb_entry *entry)
 405 {
 406         struct net_bridge_mdb_htable *mdb;
 407         struct net_bridge_mdb_entry *mp;
 408         struct net_bridge_port_group *p;
 409         struct net_bridge_port_group __rcu **pp;
 410         struct br_ip ip;
 411         int err = -EINVAL;
 412
 413         if (!netif_running(br->dev) || br->multicast_disabled)
 414                 return -EINVAL;
 415
 416         if (timer_pending(&br->multicast_querier_timer))
 417                 return -EBUSY;
 418
 419         ip.proto = entry->addr.proto;
 420         if (ip.proto == htons(ETH_P_IP))
 421                 ip.u.ip4 = entry->addr.u.ip4;
 422 #if IS_ENABLED(CONFIG_IPV6)
 423         else
 424                 ip.u.ip6 = entry->addr.u.ip6;
 425 #endif
 426
 427         spin_lock_bh(&br->multicast_lock);
 428         mdb = mlock_dereference(br->mdb, br);
 429
 430         mp = br_mdb_ip_get(mdb, &ip);
 431         if (!mp)
 432                 goto unlock;
 433
 434         for (pp = &mp->ports;
 435              (p = mlock_dereference(*pp, br)) != NULL;
 436              pp = &p->next) {
 437                 if (!p->port || p->port->dev->ifindex != entry->ifindex)
 438                         continue;
 439
 440                 if (p->port->state == BR_STATE_DISABLED)
 441                         goto unlock;
 442
 443                 rcu_assign_pointer(*pp, p->next);
 444                 hlist_del_init(&p->mglist);
 445                 del_timer(&p->timer);
 446                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 447                 err = 0;
 448
 449                 if (!mp->ports && !mp->mglist &&
 450                     netif_running(br->dev))
 451                         mod_timer(&mp->timer, jiffies);
 452                 break;
 453         }
 454
 455 unlock:
 456         spin_unlock_bh(&br->multicast_lock);
 457         return err;
 458 }
 459
 460 static int br_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 461 {
 462         struct net_device *dev;
 463         struct br_mdb_entry *entry;
 464         struct net_bridge *br;
 465         int err;
 466
 467         err = br_mdb_parse(skb, nlh, &dev, &entry);
 468         if (err < 0)
 469                 return err;
 470
 471         br = netdev_priv(dev);
 472
 473         err = __br_mdb_del(br, entry);
 474         if (!err)
 475                 __br_mdb_notify(dev, entry, RTM_DELMDB);
 476         return err;
 477 }
 478
 479 void br_mdb_init(void)
 480 {
 481         rtnl_register(PF_BRIDGE, RTM_GETMDB, NULL, br_mdb_dump, NULL);
 482         rtnl_register(PF_BRIDGE, RTM_NEWMDB, br_mdb_add, NULL, NULL);
 483         rtnl_register(PF_BRIDGE, RTM_DELMDB, br_mdb_del, NULL, NULL);
 484 }
 485
 486 void br_mdb_uninit(void)
 487 {
 488         rtnl_unregister(PF_BRIDGE, RTM_GETMDB);
 489         rtnl_unregister(PF_BRIDGE, RTM_NEWMDB);
 490         rtnl_unregister(PF_BRIDGE, RTM_DELMDB);
 491 }