net/ipv6/anycast.c

   1 /*
   2  *      Anycast support for IPv6
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      David L Stevens (dlstevens@us.ibm.com)
   7  *
   8  *      based heavily on net/ipv6/mcast.c
   9  *
  10  *      This program is free software; you can redistribute it and/or
  11  *      modify it under the terms of the GNU General Public License
  12  *      as published by the Free Software Foundation; either version
  13  *      2 of the License, or (at your option) any later version.
  14  */
  15
  16 #include <linux/capability.h>
  17 #include <linux/module.h>
  18 #include <linux/errno.h>
  19 #include <linux/types.h>
  20 #include <linux/random.h>
  21 #include <linux/string.h>
  22 #include <linux/socket.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/in6.h>
  26 #include <linux/netdevice.h>
  27 #include <linux/if_arp.h>
  28 #include <linux/route.h>
  29 #include <linux/init.h>
  30 #include <linux/proc_fs.h>
  31 #include <linux/seq_file.h>
  32 #include <linux/slab.h>
  33
  34 #include <net/net_namespace.h>
  35 #include <net/sock.h>
  36 #include <net/snmp.h>
  37
  38 #include <net/ipv6.h>
  39 #include <net/protocol.h>
  40 #include <net/if_inet6.h>
  41 #include <net/ndisc.h>
  42 #include <net/addrconf.h>
  43 #include <net/ip6_route.h>
  44
  45 #include <net/checksum.h>
  46
  47 static int ipv6_dev_ac_dec(struct net_device *dev, const struct in6_addr *addr);
  48
  49 /*
  50  *      socket join an anycast group
  51  */
  52
  53 int ipv6_sock_ac_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
  54 {
  55         struct ipv6_pinfo *np = inet6_sk(sk);
  56         struct net_device *dev = NULL;
  57         struct inet6_dev *idev;
  58         struct ipv6_ac_socklist *pac;
  59         struct net *net = sock_net(sk);
  60         int     ishost = !net->ipv6.devconf_all->forwarding;
  61         int     err = 0;
  62
  63         ASSERT_RTNL();
  64
  65         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
  66                 return -EPERM;
  67         if (ipv6_addr_is_multicast(addr))
  68                 return -EINVAL;
  69         if (ipv6_chk_addr(net, addr, NULL, 0))
  70                 return -EINVAL;
  71
  72         pac = sock_kmalloc(sk, sizeof(struct ipv6_ac_socklist), GFP_KERNEL);
  73         if (!pac)
  74                 return -ENOMEM;
  75         pac->acl_next = NULL;
  76         pac->acl_addr = *addr;
  77
  78         if (ifindex == 0) {
  79                 struct rt6_info *rt;
  80
  81                 rt = rt6_lookup(net, addr, NULL, 0, 0);
  82                 if (rt) {
  83                         dev = rt->dst.dev;
  84                         ip6_rt_put(rt);
  85                 } else if (ishost) {
  86                         err = -EADDRNOTAVAIL;
  87                         goto error;
  88                 } else {
  89                         /* router, no matching interface: just pick one */
  90                         dev = __dev_get_by_flags(net, IFF_UP,
  91                                                  IFF_UP | IFF_LOOPBACK);
  92                 }
  93         } else
  94                 dev = __dev_get_by_index(net, ifindex);
  95
  96         if (!dev) {
  97                 err = -ENODEV;
  98                 goto error;
  99         }
 100
 101         idev = __in6_dev_get(dev);
 102         if (!idev) {
 103                 if (ifindex)
 104                         err = -ENODEV;
 105                 else
 106                         err = -EADDRNOTAVAIL;
 107                 goto error;
 108         }
 109         /* reset ishost, now that we have a specific device */
 110         ishost = !idev->cnf.forwarding;
 111
 112         pac->acl_ifindex = dev->ifindex;
 113
 114         /* XXX
 115          * For hosts, allow link-local or matching prefix anycasts.
 116          * This obviates the need for propagating anycast routes while
 117          * still allowing some non-router anycast participation.
 118          */
 119         if (!ipv6_chk_prefix(addr, dev)) {
 120                 if (ishost)
 121                         err = -EADDRNOTAVAIL;
 122                 if (err)
 123                         goto error;
 124         }
 125
 126         err = __ipv6_dev_ac_inc(idev, addr);
 127         if (!err) {
 128                 pac->acl_next = np->ipv6_ac_list;
 129                 np->ipv6_ac_list = pac;
 130                 pac = NULL;
 131         }
 132
 133 error:
 134         if (pac)
 135                 sock_kfree_s(sk, pac, sizeof(*pac));
 136         return err;
 137 }
 138
 139 /*
 140  *      socket leave an anycast group
 141  */
 142 int ipv6_sock_ac_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
 143 {
 144         struct ipv6_pinfo *np = inet6_sk(sk);
 145         struct net_device *dev;
 146         struct ipv6_ac_socklist *pac, *prev_pac;
 147         struct net *net = sock_net(sk);
 148
 149         ASSERT_RTNL();
 150
 151         prev_pac = NULL;
 152         for (pac = np->ipv6_ac_list; pac; pac = pac->acl_next) {
 153                 if ((ifindex == 0 || pac->acl_ifindex == ifindex) &&
 154                      ipv6_addr_equal(&pac->acl_addr, addr))
 155                         break;
 156                 prev_pac = pac;
 157         }
 158         if (!pac)
 159                 return -ENOENT;
 160         if (prev_pac)
 161                 prev_pac->acl_next = pac->acl_next;
 162         else
 163                 np->ipv6_ac_list = pac->acl_next;
 164
 165         dev = __dev_get_by_index(net, pac->acl_ifindex);
 166         if (dev)
 167                 ipv6_dev_ac_dec(dev, &pac->acl_addr);
 168
 169         sock_kfree_s(sk, pac, sizeof(*pac));
 170         return 0;
 171 }
 172
 173 void ipv6_sock_ac_close(struct sock *sk)
 174 {
 175         struct ipv6_pinfo *np = inet6_sk(sk);
 176         struct net_device *dev = NULL;
 177         struct ipv6_ac_socklist *pac;
 178         struct net *net = sock_net(sk);
 179         int     prev_index;
 180
 181         if (!np->ipv6_ac_list)
 182                 return;
 183
 184         rtnl_lock();
 185         pac = np->ipv6_ac_list;
 186         np->ipv6_ac_list = NULL;
 187
 188         prev_index = 0;
 189         while (pac) {
 190                 struct ipv6_ac_socklist *next = pac->acl_next;
 191
 192                 if (pac->acl_ifindex != prev_index) {
 193                         dev = __dev_get_by_index(net, pac->acl_ifindex);
 194                         prev_index = pac->acl_ifindex;
 195                 }
 196                 if (dev)
 197                         ipv6_dev_ac_dec(dev, &pac->acl_addr);
 198                 sock_kfree_s(sk, pac, sizeof(*pac));
 199                 pac = next;
 200         }
 201         rtnl_unlock();
 202 }
 203
 204 static void aca_get(struct ifacaddr6 *aca)
 205 {
 206         refcount_inc(&aca->aca_refcnt);
 207 }
 208
 209 static void aca_put(struct ifacaddr6 *ac)
 210 {
 211         if (refcount_dec_and_test(&ac->aca_refcnt)) {
 212                 in6_dev_put(ac->aca_idev);
 213                 dst_release(&ac->aca_rt->dst);
 214                 kfree(ac);
 215         }
 216 }
 217
 218 static struct ifacaddr6 *aca_alloc(struct rt6_info *rt,
 219                                    const struct in6_addr *addr)
 220 {
 221         struct inet6_dev *idev = rt->rt6i_idev;
 222         struct ifacaddr6 *aca;
 223
 224         aca = kzalloc(sizeof(*aca), GFP_ATOMIC);
 225         if (!aca)
 226                 return NULL;
 227
 228         aca->aca_addr = *addr;
 229         in6_dev_hold(idev);
 230         aca->aca_idev = idev;
 231         aca->aca_rt = rt;
 232         aca->aca_users = 1;
 233         /* aca_tstamp should be updated upon changes */
 234         aca->aca_cstamp = aca->aca_tstamp = jiffies;
 235         refcount_set(&aca->aca_refcnt, 1);
 236
 237         return aca;
 238 }
 239
 240 /*
 241  *      device anycast group inc (add if not found)
 242  */
 243 int __ipv6_dev_ac_inc(struct inet6_dev *idev, const struct in6_addr *addr)
 244 {
 245         struct ifacaddr6 *aca;
 246         struct rt6_info *rt;
 247         int err;
 248
 249         ASSERT_RTNL();
 250
 251         write_lock_bh(&idev->lock);
 252         if (idev->dead) {
 253                 err = -ENODEV;
 254                 goto out;
 255         }
 256
 257         for (aca = idev->ac_list; aca; aca = aca->aca_next) {
 258                 if (ipv6_addr_equal(&aca->aca_addr, addr)) {
 259                         aca->aca_users++;
 260                         err = 0;
 261                         goto out;
 262                 }
 263         }
 264
 265         rt = addrconf_dst_alloc(idev, addr, true);
 266         if (IS_ERR(rt)) {
 267                 err = PTR_ERR(rt);
 268                 goto out;
 269         }
 270         aca = aca_alloc(rt, addr);
 271         if (!aca) {
 272                 ip6_rt_put(rt);
 273                 err = -ENOMEM;
 274                 goto out;
 275         }
 276
 277         aca->aca_next = idev->ac_list;
 278         idev->ac_list = aca;
 279
 280         /* Hold this for addrconf_join_solict() below before we unlock,
 281          * it is already exposed via idev->ac_list.
 282          */
 283         aca_get(aca);
 284         write_unlock_bh(&idev->lock);
 285
 286         ip6_ins_rt(rt);
 287
 288         addrconf_join_solict(idev->dev, &aca->aca_addr);
 289
 290         aca_put(aca);
 291         return 0;
 292 out:
 293         write_unlock_bh(&idev->lock);
 294         return err;
 295 }
 296
 297 /*
 298  *      device anycast group decrement
 299  */
 300 int __ipv6_dev_ac_dec(struct inet6_dev *idev, const struct in6_addr *addr)
 301 {
 302         struct ifacaddr6 *aca, *prev_aca;
 303
 304         ASSERT_RTNL();
 305
 306         write_lock_bh(&idev->lock);
 307         prev_aca = NULL;
 308         for (aca = idev->ac_list; aca; aca = aca->aca_next) {
 309                 if (ipv6_addr_equal(&aca->aca_addr, addr))
 310                         break;
 311                 prev_aca = aca;
 312         }
 313         if (!aca) {
 314                 write_unlock_bh(&idev->lock);
 315                 return -ENOENT;
 316         }
 317         if (--aca->aca_users > 0) {
 318                 write_unlock_bh(&idev->lock);
 319                 return 0;
 320         }
 321         if (prev_aca)
 322                 prev_aca->aca_next = aca->aca_next;
 323         else
 324                 idev->ac_list = aca->aca_next;
 325         write_unlock_bh(&idev->lock);
 326         addrconf_leave_solict(idev, &aca->aca_addr);
 327
 328         dst_hold(&aca->aca_rt->dst);
 329         ip6_del_rt(aca->aca_rt);
 330
 331         aca_put(aca);
 332         return 0;
 333 }
 334
 335 /* called with rtnl_lock() */
 336 static int ipv6_dev_ac_dec(struct net_device *dev, const struct in6_addr *addr)
 337 {
 338         struct inet6_dev *idev = __in6_dev_get(dev);
 339
 340         if (!idev)
 341                 return -ENODEV;
 342         return __ipv6_dev_ac_dec(idev, addr);
 343 }
 344
 345 void ipv6_ac_destroy_dev(struct inet6_dev *idev)
 346 {
 347         struct ifacaddr6 *aca;
 348
 349         write_lock_bh(&idev->lock);
 350         while ((aca = idev->ac_list) != NULL) {
 351                 idev->ac_list = aca->aca_next;
 352                 write_unlock_bh(&idev->lock);
 353
 354                 addrconf_leave_solict(idev, &aca->aca_addr);
 355
 356                 dst_hold(&aca->aca_rt->dst);
 357                 ip6_del_rt(aca->aca_rt);
 358
 359                 aca_put(aca);
 360
 361                 write_lock_bh(&idev->lock);
 362         }
 363         write_unlock_bh(&idev->lock);
 364 }
 365
 366 /*
 367  *      check if the interface has this anycast address
 368  *      called with rcu_read_lock()
 369  */
 370 static bool ipv6_chk_acast_dev(struct net_device *dev, const struct in6_addr *addr)
 371 {
 372         struct inet6_dev *idev;
 373         struct ifacaddr6 *aca;
 374
 375         idev = __in6_dev_get(dev);
 376         if (idev) {
 377                 read_lock_bh(&idev->lock);
 378                 for (aca = idev->ac_list; aca; aca = aca->aca_next)
 379                         if (ipv6_addr_equal(&aca->aca_addr, addr))
 380                                 break;
 381                 read_unlock_bh(&idev->lock);
 382                 return aca != NULL;
 383         }
 384         return false;
 385 }
 386
 387 /*
 388  *      check if given interface (or any, if dev==0) has this anycast address
 389  */
 390 bool ipv6_chk_acast_addr(struct net *net, struct net_device *dev,
 391                          const struct in6_addr *addr)
 392 {
 393         bool found = false;
 394
 395         rcu_read_lock();
 396         if (dev)
 397                 found = ipv6_chk_acast_dev(dev, addr);
 398         else
 399                 for_each_netdev_rcu(net, dev)
 400                         if (ipv6_chk_acast_dev(dev, addr)) {
 401                                 found = true;
 402                                 break;
 403                         }
 404         rcu_read_unlock();
 405         return found;
 406 }
 407
 408 /*      check if this anycast address is link-local on given interface or
 409  *      is global
 410  */
 411 bool ipv6_chk_acast_addr_src(struct net *net, struct net_device *dev,
 412                              const struct in6_addr *addr)
 413 {
 414         return ipv6_chk_acast_addr(net,
 415                                    (ipv6_addr_type(addr) & IPV6_ADDR_LINKLOCAL ?
 416                                     dev : NULL),
 417                                    addr);
 418 }
 419
 420 #ifdef CONFIG_PROC_FS
 421 struct ac6_iter_state {
 422         struct seq_net_private p;
 423         struct net_device *dev;
 424         struct inet6_dev *idev;
 425 };
 426
 427 #define ac6_seq_private(seq)    ((struct ac6_iter_state *)(seq)->private)
 428
 429 static inline struct ifacaddr6 *ac6_get_first(struct seq_file *seq)
 430 {
 431         struct ifacaddr6 *im = NULL;
 432         struct ac6_iter_state *state = ac6_seq_private(seq);
 433         struct net *net = seq_file_net(seq);
 434
 435         state->idev = NULL;
 436         for_each_netdev_rcu(net, state->dev) {
 437                 struct inet6_dev *idev;
 438                 idev = __in6_dev_get(state->dev);
 439                 if (!idev)
 440                         continue;
 441                 read_lock_bh(&idev->lock);
 442                 im = idev->ac_list;
 443                 if (im) {
 444                         state->idev = idev;
 445                         break;
 446                 }
 447                 read_unlock_bh(&idev->lock);
 448         }
 449         return im;
 450 }
 451
 452 static struct ifacaddr6 *ac6_get_next(struct seq_file *seq, struct ifacaddr6 *im)
 453 {
 454         struct ac6_iter_state *state = ac6_seq_private(seq);
 455
 456         im = im->aca_next;
 457         while (!im) {
 458                 if (likely(state->idev != NULL))
 459                         read_unlock_bh(&state->idev->lock);
 460
 461                 state->dev = next_net_device_rcu(state->dev);
 462                 if (!state->dev) {
 463                         state->idev = NULL;
 464                         break;
 465                 }
 466                 state->idev = __in6_dev_get(state->dev);
 467                 if (!state->idev)
 468                         continue;
 469                 read_lock_bh(&state->idev->lock);
 470                 im = state->idev->ac_list;
 471         }
 472         return im;
 473 }
 474
 475 static struct ifacaddr6 *ac6_get_idx(struct seq_file *seq, loff_t pos)
 476 {
 477         struct ifacaddr6 *im = ac6_get_first(seq);
 478         if (im)
 479                 while (pos && (im = ac6_get_next(seq, im)) != NULL)
 480                         --pos;
 481         return pos ? NULL : im;
 482 }
 483
 484 static void *ac6_seq_start(struct seq_file *seq, loff_t *pos)
 485         __acquires(RCU)
 486 {
 487         rcu_read_lock();
 488         return ac6_get_idx(seq, *pos);
 489 }
 490
 491 static void *ac6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 492 {
 493         struct ifacaddr6 *im = ac6_get_next(seq, v);
 494
 495         ++*pos;
 496         return im;
 497 }
 498
 499 static void ac6_seq_stop(struct seq_file *seq, void *v)
 500         __releases(RCU)
 501 {
 502         struct ac6_iter_state *state = ac6_seq_private(seq);
 503
 504         if (likely(state->idev != NULL)) {
 505                 read_unlock_bh(&state->idev->lock);
 506                 state->idev = NULL;
 507         }
 508         rcu_read_unlock();
 509 }
 510
 511 static int ac6_seq_show(struct seq_file *seq, void *v)
 512 {
 513         struct ifacaddr6 *im = (struct ifacaddr6 *)v;
 514         struct ac6_iter_state *state = ac6_seq_private(seq);
 515
 516         seq_printf(seq, "%-4d %-15s %pi6 %5d\n",
 517                    state->dev->ifindex, state->dev->name,
 518                    &im->aca_addr, im->aca_users);
 519         return 0;
 520 }
 521
 522 static const struct seq_operations ac6_seq_ops = {
 523         .start  =       ac6_seq_start,
 524         .next   =       ac6_seq_next,
 525         .stop   =       ac6_seq_stop,
 526         .show   =       ac6_seq_show,
 527 };
 528
 529 static int ac6_seq_open(struct inode *inode, struct file *file)
 530 {
 531         return seq_open_net(inode, file, &ac6_seq_ops,
 532                             sizeof(struct ac6_iter_state));
 533 }
 534
 535 static const struct file_operations ac6_seq_fops = {
 536         .owner          =       THIS_MODULE,
 537         .open           =       ac6_seq_open,
 538         .read           =       seq_read,
 539         .llseek         =       seq_lseek,
 540         .release        =       seq_release_net,
 541 };
 542
 543 int __net_init ac6_proc_init(struct net *net)
 544 {
 545         if (!proc_create("anycast6", S_IRUGO, net->proc_net, &ac6_seq_fops))
 546                 return -ENOMEM;
 547
 548         return 0;
 549 }
 550
 551 void ac6_proc_exit(struct net *net)
 552 {
 553         remove_proc_entry("anycast6", net->proc_net);
 554 }
 555 #endif
 556