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