net/core/neighbour.c

   1 /*
   2  *      Generic address resolution entity
   3  *
   4  *      Authors:
   5  *      Pedro Roque             <roque@di.fc.ul.pt>
   6  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  *
  13  *      Fixes:
  14  *      Vitaly E. Lavrov        releasing NULL neighbor in neigh_add.
  15  *      Harald Welte            Add neighbour cache statistics like rtstat
  16  */
  17
  18 #include <linux/types.h>
  19 #include <linux/kernel.h>
  20 #include <linux/module.h>
  21 #include <linux/socket.h>
  22 #include <linux/sched.h>
  23 #include <linux/netdevice.h>
  24 #include <linux/proc_fs.h>
  25 #ifdef CONFIG_SYSCTL
  26 #include <linux/sysctl.h>
  27 #endif
  28 #include <linux/times.h>
  29 #include <net/neighbour.h>
  30 #include <net/dst.h>
  31 #include <net/sock.h>
  32 #include <net/netevent.h>
  33 #include <net/netlink.h>
  34 #include <linux/rtnetlink.h>
  35 #include <linux/random.h>
  36 #include <linux/string.h>
  37
  38 #define NEIGH_DEBUG 1
  39
  40 #define NEIGH_PRINTK(x...) printk(x)
  41 #define NEIGH_NOPRINTK(x...) do { ; } while(0)
  42 #define NEIGH_PRINTK0 NEIGH_PRINTK
  43 #define NEIGH_PRINTK1 NEIGH_NOPRINTK
  44 #define NEIGH_PRINTK2 NEIGH_NOPRINTK
  45
  46 #if NEIGH_DEBUG >= 1
  47 #undef NEIGH_PRINTK1
  48 #define NEIGH_PRINTK1 NEIGH_PRINTK
  49 #endif
  50 #if NEIGH_DEBUG >= 2
  51 #undef NEIGH_PRINTK2
  52 #define NEIGH_PRINTK2 NEIGH_PRINTK
  53 #endif
  54
  55 #define PNEIGH_HASHMASK         0xF
  56
  57 static void neigh_timer_handler(unsigned long arg);
  58 #ifdef CONFIG_ARPD
  59 static void neigh_app_notify(struct neighbour *n);
  60 #endif
  61 static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev);
  62 void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev);
  63
  64 static struct neigh_table *neigh_tables;
  65 #ifdef CONFIG_PROC_FS
  66 static struct file_operations neigh_stat_seq_fops;
  67 #endif
  68
  69 /*
  70    Neighbour hash table buckets are protected with rwlock tbl->lock.
  71
  72    - All the scans/updates to hash buckets MUST be made under this lock.
  73    - NOTHING clever should be made under this lock: no callbacks
  74      to protocol backends, no attempts to send something to network.
  75      It will result in deadlocks, if backend/driver wants to use neighbour
  76      cache.
  77    - If the entry requires some non-trivial actions, increase
  78      its reference count and release table lock.
  79
  80    Neighbour entries are protected:
  81    - with reference count.
  82    - with rwlock neigh->lock
  83
  84    Reference count prevents destruction.
  85
  86    neigh->lock mainly serializes ll address data and its validity state.
  87    However, the same lock is used to protect another entry fields:
  88     - timer
  89     - resolution queue
  90
  91    Again, nothing clever shall be made under neigh->lock,
  92    the most complicated procedure, which we allow is dev->hard_header.
  93    It is supposed, that dev->hard_header is simplistic and does
  94    not make callbacks to neighbour tables.
  95
  96    The last lock is neigh_tbl_lock. It is pure SMP lock, protecting
  97    list of neighbour tables. This list is used only in process context,
  98  */
  99
 100 static DEFINE_RWLOCK(neigh_tbl_lock);
 101
 102 static int neigh_blackhole(struct sk_buff *skb)
 103 {
 104         kfree_skb(skb);
 105         return -ENETDOWN;
 106 }
 107
 108 /*
 109  * It is random distribution in the interval (1/2)*base...(3/2)*base.
 110  * It corresponds to default IPv6 settings and is not overridable,
 111  * because it is really reasonable choice.
 112  */
 113
 114 unsigned long neigh_rand_reach_time(unsigned long base)
 115 {
 116         return (base ? (net_random() % base) + (base >> 1) : 0);
 117 }
 118
 119
 120 static int neigh_forced_gc(struct neigh_table *tbl)
 121 {
 122         int shrunk = 0;
 123         int i;
 124
 125         NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs);
 126
 127         write_lock_bh(&tbl->lock);
 128         for (i = 0; i <= tbl->hash_mask; i++) {
 129                 struct neighbour *n, **np;
 130
 131                 np = &tbl->hash_buckets[i];
 132                 while ((n = *np) != NULL) {
 133                         /* Neighbour record may be discarded if:
 134                          * - nobody refers to it.
 135                          * - it is not permanent
 136                          */
 137                         write_lock(&n->lock);
 138                         if (atomic_read(&n->refcnt) == 1 &&
 139                             !(n->nud_state & NUD_PERMANENT)) {
 140                                 *np     = n->next;
 141                                 n->dead = 1;
 142                                 shrunk  = 1;
 143                                 write_unlock(&n->lock);
 144                                 neigh_release(n);
 145                                 continue;
 146                         }
 147                         write_unlock(&n->lock);
 148                         np = &n->next;
 149                 }
 150         }
 151
 152         tbl->last_flush = jiffies;
 153
 154         write_unlock_bh(&tbl->lock);
 155
 156         return shrunk;
 157 }
 158
 159 static int neigh_del_timer(struct neighbour *n)
 160 {
 161         if ((n->nud_state & NUD_IN_TIMER) &&
 162             del_timer(&n->timer)) {
 163                 neigh_release(n);
 164                 return 1;
 165         }
 166         return 0;
 167 }
 168
 169 static void pneigh_queue_purge(struct sk_buff_head *list)
 170 {
 171         struct sk_buff *skb;
 172
 173         while ((skb = skb_dequeue(list)) != NULL) {
 174                 dev_put(skb->dev);
 175                 kfree_skb(skb);
 176         }
 177 }
 178
 179 static void neigh_flush_dev(struct neigh_table *tbl, struct net_device *dev)
 180 {
 181         int i;
 182
 183         for (i = 0; i <= tbl->hash_mask; i++) {
 184                 struct neighbour *n, **np = &tbl->hash_buckets[i];
 185
 186                 while ((n = *np) != NULL) {
 187                         if (dev && n->dev != dev) {
 188                                 np = &n->next;
 189                                 continue;
 190                         }
 191                         *np = n->next;
 192                         write_lock(&n->lock);
 193                         neigh_del_timer(n);
 194                         n->dead = 1;
 195
 196                         if (atomic_read(&n->refcnt) != 1) {
 197                                 /* The most unpleasant situation.
 198                                    We must destroy neighbour entry,
 199                                    but someone still uses it.
 200
 201                                    The destroy will be delayed until
 202                                    the last user releases us, but
 203                                    we must kill timers etc. and move
 204                                    it to safe state.
 205                                  */
 206                                 skb_queue_purge(&n->arp_queue);
 207                                 n->output = neigh_blackhole;
 208                                 if (n->nud_state & NUD_VALID)
 209                                         n->nud_state = NUD_NOARP;
 210                                 else
 211                                         n->nud_state = NUD_NONE;
 212                                 NEIGH_PRINTK2("neigh %p is stray.\n", n);
 213                         }
 214                         write_unlock(&n->lock);
 215                         neigh_release(n);
 216                 }
 217         }
 218 }
 219
 220 void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev)
 221 {
 222         write_lock_bh(&tbl->lock);
 223         neigh_flush_dev(tbl, dev);
 224         write_unlock_bh(&tbl->lock);
 225 }
 226
 227 int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
 228 {
 229         write_lock_bh(&tbl->lock);
 230         neigh_flush_dev(tbl, dev);
 231         pneigh_ifdown(tbl, dev);
 232         write_unlock_bh(&tbl->lock);
 233
 234         del_timer_sync(&tbl->proxy_timer);
 235         pneigh_queue_purge(&tbl->proxy_queue);
 236         return 0;
 237 }
 238
 239 static struct neighbour *neigh_alloc(struct neigh_table *tbl)
 240 {
 241         struct neighbour *n = NULL;
 242         unsigned long now = jiffies;
 243         int entries;
 244
 245         entries = atomic_inc_return(&tbl->entries) - 1;
 246         if (entries >= tbl->gc_thresh3 ||
 247             (entries >= tbl->gc_thresh2 &&
 248              time_after(now, tbl->last_flush + 5 * HZ))) {
 249                 if (!neigh_forced_gc(tbl) &&
 250                     entries >= tbl->gc_thresh3)
 251                         goto out_entries;
 252         }
 253
 254         n = kmem_cache_zalloc(tbl->kmem_cachep, GFP_ATOMIC);
 255         if (!n)
 256                 goto out_entries;
 257
 258         skb_queue_head_init(&n->arp_queue);
 259         rwlock_init(&n->lock);
 260         n->updated        = n->used = now;
 261         n->nud_state      = NUD_NONE;
 262         n->output         = neigh_blackhole;
 263         n->parms          = neigh_parms_clone(&tbl->parms);
 264         init_timer(&n->timer);
 265         n->timer.function = neigh_timer_handler;
 266         n->timer.data     = (unsigned long)n;
 267
 268         NEIGH_CACHE_STAT_INC(tbl, allocs);
 269         n->tbl            = tbl;
 270         atomic_set(&n->refcnt, 1);
 271         n->dead           = 1;
 272 out:
 273         return n;
 274
 275 out_entries:
 276         atomic_dec(&tbl->entries);
 277         goto out;
 278 }
 279
 280 static struct neighbour **neigh_hash_alloc(unsigned int entries)
 281 {
 282         unsigned long size = entries * sizeof(struct neighbour *);
 283         struct neighbour **ret;
 284
 285         if (size <= PAGE_SIZE) {
 286                 ret = kzalloc(size, GFP_ATOMIC);
 287         } else {
 288                 ret = (struct neighbour **)
 289                       __get_free_pages(GFP_ATOMIC|__GFP_ZERO, get_order(size));
 290         }
 291         return ret;
 292 }
 293
 294 static void neigh_hash_free(struct neighbour **hash, unsigned int entries)
 295 {
 296         unsigned long size = entries * sizeof(struct neighbour *);
 297
 298         if (size <= PAGE_SIZE)
 299                 kfree(hash);
 300         else
 301                 free_pages((unsigned long)hash, get_order(size));
 302 }
 303
 304 static void neigh_hash_grow(struct neigh_table *tbl, unsigned long new_entries)
 305 {
 306         struct neighbour **new_hash, **old_hash;
 307         unsigned int i, new_hash_mask, old_entries;
 308
 309         NEIGH_CACHE_STAT_INC(tbl, hash_grows);
 310
 311         BUG_ON(new_entries & (new_entries - 1));
 312         new_hash = neigh_hash_alloc(new_entries);
 313         if (!new_hash)
 314                 return;
 315
 316         old_entries = tbl->hash_mask + 1;
 317         new_hash_mask = new_entries - 1;
 318         old_hash = tbl->hash_buckets;
 319
 320         get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd));
 321         for (i = 0; i < old_entries; i++) {
 322                 struct neighbour *n, *next;
 323
 324                 for (n = old_hash[i]; n; n = next) {
 325                         unsigned int hash_val = tbl->hash(n->primary_key, n->dev);
 326
 327                         hash_val &= new_hash_mask;
 328                         next = n->next;
 329
 330                         n->next = new_hash[hash_val];
 331                         new_hash[hash_val] = n;
 332                 }
 333         }
 334         tbl->hash_buckets = new_hash;
 335         tbl->hash_mask = new_hash_mask;
 336
 337         neigh_hash_free(old_hash, old_entries);
 338 }
 339
 340 struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
 341                                struct net_device *dev)
 342 {
 343         struct neighbour *n;
 344         int key_len = tbl->key_len;
 345         u32 hash_val = tbl->hash(pkey, dev);
 346
 347         NEIGH_CACHE_STAT_INC(tbl, lookups);
 348
 349         read_lock_bh(&tbl->lock);
 350         for (n = tbl->hash_buckets[hash_val & tbl->hash_mask]; n; n = n->next) {
 351                 if (dev == n->dev && !memcmp(n->primary_key, pkey, key_len)) {
 352                         neigh_hold(n);
 353                         NEIGH_CACHE_STAT_INC(tbl, hits);
 354                         break;
 355                 }
 356         }
 357         read_unlock_bh(&tbl->lock);
 358         return n;
 359 }
 360
 361 struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, const void *pkey)
 362 {
 363         struct neighbour *n;
 364         int key_len = tbl->key_len;
 365         u32 hash_val = tbl->hash(pkey, NULL);
 366
 367         NEIGH_CACHE_STAT_INC(tbl, lookups);
 368
 369         read_lock_bh(&tbl->lock);
 370         for (n = tbl->hash_buckets[hash_val & tbl->hash_mask]; n; n = n->next) {
 371                 if (!memcmp(n->primary_key, pkey, key_len)) {
 372                         neigh_hold(n);
 373                         NEIGH_CACHE_STAT_INC(tbl, hits);
 374                         break;
 375                 }
 376         }
 377         read_unlock_bh(&tbl->lock);
 378         return n;
 379 }
 380
 381 struct neighbour *neigh_create(struct neigh_table *tbl, const void *pkey,
 382                                struct net_device *dev)
 383 {
 384         u32 hash_val;
 385         int key_len = tbl->key_len;
 386         int error;
 387         struct neighbour *n1, *rc, *n = neigh_alloc(tbl);
 388
 389         if (!n) {
 390                 rc = ERR_PTR(-ENOBUFS);
 391                 goto out;
 392         }
 393
 394         memcpy(n->primary_key, pkey, key_len);
 395         n->dev = dev;
 396         dev_hold(dev);
 397
 398         /* Protocol specific setup. */
 399         if (tbl->constructor && (error = tbl->constructor(n)) < 0) {
 400                 rc = ERR_PTR(error);
 401                 goto out_neigh_release;
 402         }
 403
 404         /* Device specific setup. */
 405         if (n->parms->neigh_setup &&
 406             (error = n->parms->neigh_setup(n)) < 0) {
 407                 rc = ERR_PTR(error);
 408                 goto out_neigh_release;
 409         }
 410
 411         n->confirmed = jiffies - (n->parms->base_reachable_time << 1);
 412
 413         write_lock_bh(&tbl->lock);
 414
 415         if (atomic_read(&tbl->entries) > (tbl->hash_mask + 1))
 416                 neigh_hash_grow(tbl, (tbl->hash_mask + 1) << 1);
 417
 418         hash_val = tbl->hash(pkey, dev) & tbl->hash_mask;
 419
 420         if (n->parms->dead) {
 421                 rc = ERR_PTR(-EINVAL);
 422                 goto out_tbl_unlock;
 423         }
 424
 425         for (n1 = tbl->hash_buckets[hash_val]; n1; n1 = n1->next) {
 426                 if (dev == n1->dev && !memcmp(n1->primary_key, pkey, key_len)) {
 427                         neigh_hold(n1);
 428                         rc = n1;
 429                         goto out_tbl_unlock;
 430                 }
 431         }
 432
 433         n->next = tbl->hash_buckets[hash_val];
 434         tbl->hash_buckets[hash_val] = n;
 435         n->dead = 0;
 436         neigh_hold(n);
 437         write_unlock_bh(&tbl->lock);
 438         NEIGH_PRINTK2("neigh %p is created.\n", n);
 439         rc = n;
 440 out:
 441         return rc;
 442 out_tbl_unlock:
 443         write_unlock_bh(&tbl->lock);
 444 out_neigh_release:
 445         neigh_release(n);
 446         goto out;
 447 }
 448
 449 struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl, const void *pkey,
 450                                     struct net_device *dev, int creat)
 451 {
 452         struct pneigh_entry *n;
 453         int key_len = tbl->key_len;
 454         u32 hash_val = *(u32 *)(pkey + key_len - 4);
 455
 456         hash_val ^= (hash_val >> 16);
 457         hash_val ^= hash_val >> 8;
 458         hash_val ^= hash_val >> 4;
 459         hash_val &= PNEIGH_HASHMASK;
 460
 461         read_lock_bh(&tbl->lock);
 462
 463         for (n = tbl->phash_buckets[hash_val]; n; n = n->next) {
 464                 if (!memcmp(n->key, pkey, key_len) &&
 465                     (n->dev == dev || !n->dev)) {
 466                         read_unlock_bh(&tbl->lock);
 467                         goto out;
 468                 }
 469         }
 470         read_unlock_bh(&tbl->lock);
 471         n = NULL;
 472         if (!creat)
 473                 goto out;
 474
 475         n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL);
 476         if (!n)
 477                 goto out;
 478
 479         memcpy(n->key, pkey, key_len);
 480         n->dev = dev;
 481         if (dev)
 482                 dev_hold(dev);
 483
 484         if (tbl->pconstructor && tbl->pconstructor(n)) {
 485                 if (dev)
 486                         dev_put(dev);
 487                 kfree(n);
 488                 n = NULL;
 489                 goto out;
 490         }
 491
 492         write_lock_bh(&tbl->lock);
 493         n->next = tbl->phash_buckets[hash_val];
 494         tbl->phash_buckets[hash_val] = n;
 495         write_unlock_bh(&tbl->lock);
 496 out:
 497         return n;
 498 }
 499
 500
 501 int pneigh_delete(struct neigh_table *tbl, const void *pkey,
 502                   struct net_device *dev)
 503 {
 504         struct pneigh_entry *n, **np;
 505         int key_len = tbl->key_len;
 506         u32 hash_val = *(u32 *)(pkey + key_len - 4);
 507
 508         hash_val ^= (hash_val >> 16);
 509         hash_val ^= hash_val >> 8;
 510         hash_val ^= hash_val >> 4;
 511         hash_val &= PNEIGH_HASHMASK;
 512
 513         write_lock_bh(&tbl->lock);
 514         for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL;
 515              np = &n->next) {
 516                 if (!memcmp(n->key, pkey, key_len) && n->dev == dev) {
 517                         *np = n->next;
 518                         write_unlock_bh(&tbl->lock);
 519                         if (tbl->pdestructor)
 520                                 tbl->pdestructor(n);
 521                         if (n->dev)
 522                                 dev_put(n->dev);
 523                         kfree(n);
 524                         return 0;
 525                 }
 526         }
 527         write_unlock_bh(&tbl->lock);
 528         return -ENOENT;
 529 }
 530
 531 static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
 532 {
 533         struct pneigh_entry *n, **np;
 534         u32 h;
 535
 536         for (h = 0; h <= PNEIGH_HASHMASK; h++) {
 537                 np = &tbl->phash_buckets[h];
 538                 while ((n = *np) != NULL) {
 539                         if (!dev || n->dev == dev) {
 540                                 *np = n->next;
 541                                 if (tbl->pdestructor)
 542                                         tbl->pdestructor(n);
 543                                 if (n->dev)
 544                                         dev_put(n->dev);
 545                                 kfree(n);
 546                                 continue;
 547                         }
 548                         np = &n->next;
 549                 }
 550         }
 551         return -ENOENT;
 552 }
 553
 554
 555 /*
 556  *      neighbour must already be out of the table;
 557  *
 558  */
 559 void neigh_destroy(struct neighbour *neigh)
 560 {
 561         struct hh_cache *hh;
 562
 563         NEIGH_CACHE_STAT_INC(neigh->tbl, destroys);
 564
 565         if (!neigh->dead) {
 566                 printk(KERN_WARNING
 567                        "Destroying alive neighbour %p\n", neigh);
 568                 dump_stack();
 569                 return;
 570         }
 571
 572         if (neigh_del_timer(neigh))
 573                 printk(KERN_WARNING "Impossible event.\n");
 574
 575         while ((hh = neigh->hh) != NULL) {
 576                 neigh->hh = hh->hh_next;
 577                 hh->hh_next = NULL;
 578
 579                 write_seqlock_bh(&hh->hh_lock);
 580                 hh->hh_output = neigh_blackhole;
 581                 write_sequnlock_bh(&hh->hh_lock);
 582                 if (atomic_dec_and_test(&hh->hh_refcnt))
 583                         kfree(hh);
 584         }
 585
 586         if (neigh->parms->neigh_destructor)
 587                 (neigh->parms->neigh_destructor)(neigh);
 588
 589         skb_queue_purge(&neigh->arp_queue);
 590
 591         dev_put(neigh->dev);
 592         neigh_parms_put(neigh->parms);
 593
 594         NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh);
 595
 596         atomic_dec(&neigh->tbl->entries);
 597         kmem_cache_free(neigh->tbl->kmem_cachep, neigh);
 598 }
 599
 600 /* Neighbour state is suspicious;
 601    disable fast path.
 602
 603    Called with write_locked neigh.
 604  */
 605 static void neigh_suspect(struct neighbour *neigh)
 606 {
 607         struct hh_cache *hh;
 608
 609         NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
 610
 611         neigh->output = neigh->ops->output;
 612
 613         for (hh = neigh->hh; hh; hh = hh->hh_next)
 614                 hh->hh_output = neigh->ops->output;
 615 }
 616
 617 /* Neighbour state is OK;
 618    enable fast path.
 619
 620    Called with write_locked neigh.
 621  */
 622 static void neigh_connect(struct neighbour *neigh)
 623 {
 624         struct hh_cache *hh;
 625
 626         NEIGH_PRINTK2("neigh %p is connected.\n", neigh);
 627
 628         neigh->output = neigh->ops->connected_output;
 629
 630         for (hh = neigh->hh; hh; hh = hh->hh_next)
 631                 hh->hh_output = neigh->ops->hh_output;
 632 }
 633
 634 static void neigh_periodic_timer(unsigned long arg)
 635 {
 636         struct neigh_table *tbl = (struct neigh_table *)arg;
 637         struct neighbour *n, **np;
 638         unsigned long expire, now = jiffies;
 639
 640         NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs);
 641
 642         write_lock(&tbl->lock);
 643
 644         /*
 645          *      periodically recompute ReachableTime from random function
 646          */
 647
 648         if (time_after(now, tbl->last_rand + 300 * HZ)) {
 649                 struct neigh_parms *p;
 650                 tbl->last_rand = now;
 651                 for (p = &tbl->parms; p; p = p->next)
 652                         p->reachable_time =
 653                                 neigh_rand_reach_time(p->base_reachable_time);
 654         }
 655
 656         np = &tbl->hash_buckets[tbl->hash_chain_gc];
 657         tbl->hash_chain_gc = ((tbl->hash_chain_gc + 1) & tbl->hash_mask);
 658
 659         while ((n = *np) != NULL) {
 660                 unsigned int state;
 661
 662                 write_lock(&n->lock);
 663
 664                 state = n->nud_state;
 665                 if (state & (NUD_PERMANENT | NUD_IN_TIMER)) {
 666                         write_unlock(&n->lock);
 667                         goto next_elt;
 668                 }
 669
 670                 if (time_before(n->used, n->confirmed))
 671                         n->used = n->confirmed;
 672
 673                 if (atomic_read(&n->refcnt) == 1 &&
 674                     (state == NUD_FAILED ||
 675                      time_after(now, n->used + n->parms->gc_staletime))) {
 676                         *np = n->next;
 677                         n->dead = 1;
 678                         write_unlock(&n->lock);
 679                         neigh_release(n);
 680                         continue;
 681                 }
 682                 write_unlock(&n->lock);
 683
 684 next_elt:
 685                 np = &n->next;
 686         }
 687
 688         /* Cycle through all hash buckets every base_reachable_time/2 ticks.
 689          * ARP entry timeouts range from 1/2 base_reachable_time to 3/2
 690          * base_reachable_time.
 691          */
 692         expire = tbl->parms.base_reachable_time >> 1;
 693         expire /= (tbl->hash_mask + 1);
 694         if (!expire)
 695                 expire = 1;
 696
 697         if (expire>HZ)
 698                 mod_timer(&tbl->gc_timer, round_jiffies(now + expire));
 699         else
 700                 mod_timer(&tbl->gc_timer, now + expire);
 701
 702         write_unlock(&tbl->lock);
 703 }
 704
 705 static __inline__ int neigh_max_probes(struct neighbour *n)
 706 {
 707         struct neigh_parms *p = n->parms;
 708         return (n->nud_state & NUD_PROBE ?
 709                 p->ucast_probes :
 710                 p->ucast_probes + p->app_probes + p->mcast_probes);
 711 }
 712
 713 static inline void neigh_add_timer(struct neighbour *n, unsigned long when)
 714 {
 715         if (unlikely(mod_timer(&n->timer, when))) {
 716                 printk("NEIGH: BUG, double timer add, state is %x\n",
 717                        n->nud_state);
 718                 dump_stack();
 719         }
 720 }
 721
 722 /* Called when a timer expires for a neighbour entry. */
 723
 724 static void neigh_timer_handler(unsigned long arg)
 725 {
 726         unsigned long now, next;
 727         struct neighbour *neigh = (struct neighbour *)arg;
 728         unsigned state;
 729         int notify = 0;
 730
 731         write_lock(&neigh->lock);
 732
 733         state = neigh->nud_state;
 734         now = jiffies;
 735         next = now + HZ;
 736
 737         if (!(state & NUD_IN_TIMER)) {
 738 #ifndef CONFIG_SMP
 739                 printk(KERN_WARNING "neigh: timer & !nud_in_timer\n");
 740 #endif
 741                 goto out;
 742         }
 743
 744         if (state & NUD_REACHABLE) {
 745                 if (time_before_eq(now,
 746                                    neigh->confirmed + neigh->parms->reachable_time)) {
 747                         NEIGH_PRINTK2("neigh %p is still alive.\n", neigh);
 748                         next = neigh->confirmed + neigh->parms->reachable_time;
 749                 } else if (time_before_eq(now,
 750                                           neigh->used + neigh->parms->delay_probe_time)) {
 751                         NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
 752                         neigh->nud_state = NUD_DELAY;
 753                         neigh->updated = jiffies;
 754                         neigh_suspect(neigh);
 755                         next = now + neigh->parms->delay_probe_time;
 756                 } else {
 757                         NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
 758                         neigh->nud_state = NUD_STALE;
 759                         neigh->updated = jiffies;
 760                         neigh_suspect(neigh);
 761                         notify = 1;
 762                 }
 763         } else if (state & NUD_DELAY) {
 764                 if (time_before_eq(now,
 765                                    neigh->confirmed + neigh->parms->delay_probe_time)) {
 766                         NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh);
 767                         neigh->nud_state = NUD_REACHABLE;
 768                         neigh->updated = jiffies;
 769                         neigh_connect(neigh);
 770                         notify = 1;
 771                         next = neigh->confirmed + neigh->parms->reachable_time;
 772                 } else {
 773                         NEIGH_PRINTK2("neigh %p is probed.\n", neigh);
 774                         neigh->nud_state = NUD_PROBE;
 775                         neigh->updated = jiffies;
 776                         atomic_set(&neigh->probes, 0);
 777                         next = now + neigh->parms->retrans_time;
 778                 }
 779         } else {
 780                 /* NUD_PROBE|NUD_INCOMPLETE */
 781                 next = now + neigh->parms->retrans_time;
 782         }
 783
 784         if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) &&
 785             atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) {
 786                 struct sk_buff *skb;
 787
 788                 neigh->nud_state = NUD_FAILED;
 789                 neigh->updated = jiffies;
 790                 notify = 1;
 791                 NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed);
 792                 NEIGH_PRINTK2("neigh %p is failed.\n", neigh);
 793
 794                 /* It is very thin place. report_unreachable is very complicated
 795                    routine. Particularly, it can hit the same neighbour entry!
 796
 797                    So that, we try to be accurate and avoid dead loop. --ANK
 798                  */
 799                 while (neigh->nud_state == NUD_FAILED &&
 800                        (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
 801                         write_unlock(&neigh->lock);
 802                         neigh->ops->error_report(neigh, skb);
 803                         write_lock(&neigh->lock);
 804                 }
 805                 skb_queue_purge(&neigh->arp_queue);
 806         }
 807
 808         if (neigh->nud_state & NUD_IN_TIMER) {
 809                 if (time_before(next, jiffies + HZ/2))
 810                         next = jiffies + HZ/2;
 811                 if (!mod_timer(&neigh->timer, next))
 812                         neigh_hold(neigh);
 813         }
 814         if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) {
 815                 struct sk_buff *skb = skb_peek(&neigh->arp_queue);
 816                 /* keep skb alive even if arp_queue overflows */
 817                 if (skb)
 818                         skb_get(skb);
 819                 write_unlock(&neigh->lock);
 820                 neigh->ops->solicit(neigh, skb);
 821                 atomic_inc(&neigh->probes);
 822                 if (skb)
 823                         kfree_skb(skb);
 824         } else {
 825 out:
 826                 write_unlock(&neigh->lock);
 827         }
 828         if (notify)
 829                 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
 830
 831 #ifdef CONFIG_ARPD
 832         if (notify && neigh->parms->app_probes)
 833                 neigh_app_notify(neigh);
 834 #endif
 835         neigh_release(neigh);
 836 }
 837
 838 int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
 839 {
 840         int rc;
 841         unsigned long now;
 842
 843         write_lock_bh(&neigh->lock);
 844
 845         rc = 0;
 846         if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))
 847                 goto out_unlock_bh;
 848
 849         now = jiffies;
 850
 851         if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) {
 852                 if (neigh->parms->mcast_probes + neigh->parms->app_probes) {
 853                         atomic_set(&neigh->probes, neigh->parms->ucast_probes);
 854                         neigh->nud_state     = NUD_INCOMPLETE;
 855                         neigh->updated = jiffies;
 856                         neigh_hold(neigh);
 857                         neigh_add_timer(neigh, now + 1);
 858                 } else {
 859                         neigh->nud_state = NUD_FAILED;
 860                         neigh->updated = jiffies;
 861                         write_unlock_bh(&neigh->lock);
 862
 863                         if (skb)
 864                                 kfree_skb(skb);
 865                         return 1;
 866                 }
 867         } else if (neigh->nud_state & NUD_STALE) {
 868                 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
 869                 neigh_hold(neigh);
 870                 neigh->nud_state = NUD_DELAY;
 871                 neigh->updated = jiffies;
 872                 neigh_add_timer(neigh,
 873                                 jiffies + neigh->parms->delay_probe_time);
 874         }
 875
 876         if (neigh->nud_state == NUD_INCOMPLETE) {
 877                 if (skb) {
 878                         if (skb_queue_len(&neigh->arp_queue) >=
 879                             neigh->parms->queue_len) {
 880                                 struct sk_buff *buff;
 881                                 buff = neigh->arp_queue.next;
 882                                 __skb_unlink(buff, &neigh->arp_queue);
 883                                 kfree_skb(buff);
 884                         }
 885                         __skb_queue_tail(&neigh->arp_queue, skb);
 886                 }
 887                 rc = 1;
 888         }
 889 out_unlock_bh:
 890         write_unlock_bh(&neigh->lock);
 891         return rc;
 892 }
 893
 894 static void neigh_update_hhs(struct neighbour *neigh)
 895 {
 896         struct hh_cache *hh;
 897         void (*update)(struct hh_cache*, struct net_device*, unsigned char *) =
 898                 neigh->dev->header_cache_update;
 899
 900         if (update) {
 901                 for (hh = neigh->hh; hh; hh = hh->hh_next) {
 902                         write_seqlock_bh(&hh->hh_lock);
 903                         update(hh, neigh->dev, neigh->ha);
 904                         write_sequnlock_bh(&hh->hh_lock);
 905                 }
 906         }
 907 }
 908
 909
 910
 911 /* Generic update routine.
 912    -- lladdr is new lladdr or NULL, if it is not supplied.
 913    -- new    is new state.
 914    -- flags
 915         NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
 916                                 if it is different.
 917         NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
 918                                 lladdr instead of overriding it
 919                                 if it is different.
 920                                 It also allows to retain current state
 921                                 if lladdr is unchanged.
 922         NEIGH_UPDATE_F_ADMIN    means that the change is administrative.
 923
 924         NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
 925                                 NTF_ROUTER flag.
 926         NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
 927                                 a router.
 928
 929    Caller MUST hold reference count on the entry.
 930  */
 931
 932 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new,
 933                  u32 flags)
 934 {
 935         u8 old;
 936         int err;
 937         int notify = 0;
 938         struct net_device *dev;
 939         int update_isrouter = 0;
 940
 941         write_lock_bh(&neigh->lock);
 942
 943         dev    = neigh->dev;
 944         old    = neigh->nud_state;
 945         err    = -EPERM;
 946
 947         if (!(flags & NEIGH_UPDATE_F_ADMIN) &&
 948             (old & (NUD_NOARP | NUD_PERMANENT)))
 949                 goto out;
 950
 951         if (!(new & NUD_VALID)) {
 952                 neigh_del_timer(neigh);
 953                 if (old & NUD_CONNECTED)
 954                         neigh_suspect(neigh);
 955                 neigh->nud_state = new;
 956                 err = 0;
 957                 notify = old & NUD_VALID;
 958                 goto out;
 959         }
 960
 961         /* Compare new lladdr with cached one */
 962         if (!dev->addr_len) {
 963                 /* First case: device needs no address. */
 964                 lladdr = neigh->ha;
 965         } else if (lladdr) {
 966                 /* The second case: if something is already cached
 967                    and a new address is proposed:
 968                    - compare new & old
 969                    - if they are different, check override flag
 970                  */
 971                 if ((old & NUD_VALID) &&
 972                     !memcmp(lladdr, neigh->ha, dev->addr_len))
 973                         lladdr = neigh->ha;
 974         } else {
 975                 /* No address is supplied; if we know something,
 976                    use it, otherwise discard the request.
 977                  */
 978                 err = -EINVAL;
 979                 if (!(old & NUD_VALID))
 980                         goto out;
 981                 lladdr = neigh->ha;
 982         }
 983
 984         if (new & NUD_CONNECTED)
 985                 neigh->confirmed = jiffies;
 986         neigh->updated = jiffies;
 987
 988         /* If entry was valid and address is not changed,
 989            do not change entry state, if new one is STALE.
 990          */
 991         err = 0;
 992         update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
 993         if (old & NUD_VALID) {
 994                 if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) {
 995                         update_isrouter = 0;
 996                         if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) &&
 997                             (old & NUD_CONNECTED)) {
 998                                 lladdr = neigh->ha;
 999                                 new = NUD_STALE;
1000                         } else
1001                                 goto out;
1002                 } else {
1003                         if (lladdr == neigh->ha && new == NUD_STALE &&
1004                             ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) ||
1005                              (old & NUD_CONNECTED))
1006                             )
1007                                 new = old;
1008                 }
1009         }
1010
1011         if (new != old) {
1012                 neigh_del_timer(neigh);
1013                 if (new & NUD_IN_TIMER) {
1014                         neigh_hold(neigh);
1015                         neigh_add_timer(neigh, (jiffies +
1016                                                 ((new & NUD_REACHABLE) ?
1017                                                  neigh->parms->reachable_time :
1018                                                  0)));
1019                 }
1020                 neigh->nud_state = new;
1021         }
1022
1023         if (lladdr != neigh->ha) {
1024                 memcpy(&neigh->ha, lladdr, dev->addr_len);
1025                 neigh_update_hhs(neigh);
1026                 if (!(new & NUD_CONNECTED))
1027                         neigh->confirmed = jiffies -
1028                                       (neigh->parms->base_reachable_time << 1);
1029                 notify = 1;
1030         }
1031         if (new == old)
1032                 goto out;
1033         if (new & NUD_CONNECTED)
1034                 neigh_connect(neigh);
1035         else
1036                 neigh_suspect(neigh);
1037         if (!(old & NUD_VALID)) {
1038                 struct sk_buff *skb;
1039
1040                 /* Again: avoid dead loop if something went wrong */
1041
1042                 while (neigh->nud_state & NUD_VALID &&
1043                        (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
1044                         struct neighbour *n1 = neigh;
1045                         write_unlock_bh(&neigh->lock);
1046                         /* On shaper/eql skb->dst->neighbour != neigh :( */
1047                         if (skb->dst && skb->dst->neighbour)
1048                                 n1 = skb->dst->neighbour;
1049                         n1->output(skb);
1050                         write_lock_bh(&neigh->lock);
1051                 }
1052                 skb_queue_purge(&neigh->arp_queue);
1053         }
1054 out:
1055         if (update_isrouter) {
1056                 neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ?
1057                         (neigh->flags | NTF_ROUTER) :
1058                         (neigh->flags & ~NTF_ROUTER);
1059         }
1060         write_unlock_bh(&neigh->lock);
1061
1062         if (notify)
1063                 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
1064 #ifdef CONFIG_ARPD
1065         if (notify && neigh->parms->app_probes)
1066                 neigh_app_notify(neigh);
1067 #endif
1068         return err;
1069 }
1070
1071 struct neighbour *neigh_event_ns(struct neigh_table *tbl,
1072                                  u8 *lladdr, void *saddr,
1073                                  struct net_device *dev)
1074 {
1075         struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev,
1076                                                  lladdr || !dev->addr_len);
1077         if (neigh)
1078                 neigh_update(neigh, lladdr, NUD_STALE,
1079                              NEIGH_UPDATE_F_OVERRIDE);
1080         return neigh;
1081 }
1082
1083 static void neigh_hh_init(struct neighbour *n, struct dst_entry *dst,
1084                           __be16 protocol)
1085 {
1086         struct hh_cache *hh;
1087         struct net_device *dev = dst->dev;
1088
1089         for (hh = n->hh; hh; hh = hh->hh_next)
1090                 if (hh->hh_type == protocol)
1091                         break;
1092
1093         if (!hh && (hh = kzalloc(sizeof(*hh), GFP_ATOMIC)) != NULL) {
1094                 seqlock_init(&hh->hh_lock);
1095                 hh->hh_type = protocol;
1096                 atomic_set(&hh->hh_refcnt, 0);
1097                 hh->hh_next = NULL;
1098                 if (dev->hard_header_cache(n, hh)) {
1099                         kfree(hh);
1100                         hh = NULL;
1101                 } else {
1102                         atomic_inc(&hh->hh_refcnt);
1103                         hh->hh_next = n->hh;
1104                         n->hh       = hh;
1105                         if (n->nud_state & NUD_CONNECTED)
1106                                 hh->hh_output = n->ops->hh_output;
1107                         else
1108                                 hh->hh_output = n->ops->output;
1109                 }
1110         }
1111         if (hh) {
1112                 atomic_inc(&hh->hh_refcnt);
1113                 dst->hh = hh;
1114         }
1115 }
1116
1117 /* This function can be used in contexts, where only old dev_queue_xmit
1118    worked, f.e. if you want to override normal output path (eql, shaper),
1119    but resolution is not made yet.
1120  */
1121
1122 int neigh_compat_output(struct sk_buff *skb)
1123 {
1124         struct net_device *dev = skb->dev;
1125
1126         __skb_pull(skb, skb->nh.raw - skb->data);
1127
1128         if (dev->hard_header &&
1129             dev->hard_header(skb, dev, ntohs(skb->protocol), NULL, NULL,
1130                              skb->len) < 0 &&
1131             dev->rebuild_header(skb))
1132                 return 0;
1133
1134         return dev_queue_xmit(skb);
1135 }
1136
1137 /* Slow and careful. */
1138
1139 int neigh_resolve_output(struct sk_buff *skb)
1140 {
1141         struct dst_entry *dst = skb->dst;
1142         struct neighbour *neigh;
1143         int rc = 0;
1144
1145         if (!dst || !(neigh = dst->neighbour))
1146                 goto discard;
1147
1148         __skb_pull(skb, skb->nh.raw - skb->data);
1149
1150         if (!neigh_event_send(neigh, skb)) {
1151                 int err;
1152                 struct net_device *dev = neigh->dev;
1153                 if (dev->hard_header_cache && !dst->hh) {
1154                         write_lock_bh(&neigh->lock);
1155                         if (!dst->hh)
1156                                 neigh_hh_init(neigh, dst, dst->ops->protocol);
1157                         err = dev->hard_header(skb, dev, ntohs(skb->protocol),
1158                                                neigh->ha, NULL, skb->len);
1159                         write_unlock_bh(&neigh->lock);
1160                 } else {
1161                         read_lock_bh(&neigh->lock);
1162                         err = dev->hard_header(skb, dev, ntohs(skb->protocol),
1163                                                neigh->ha, NULL, skb->len);
1164                         read_unlock_bh(&neigh->lock);
1165                 }
1166                 if (err >= 0)
1167                         rc = neigh->ops->queue_xmit(skb);
1168                 else
1169                         goto out_kfree_skb;
1170         }
1171 out:
1172         return rc;
1173 discard:
1174         NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n",
1175                       dst, dst ? dst->neighbour : NULL);
1176 out_kfree_skb:
1177         rc = -EINVAL;
1178         kfree_skb(skb);
1179         goto out;
1180 }
1181
1182 /* As fast as possible without hh cache */
1183
1184 int neigh_connected_output(struct sk_buff *skb)
1185 {
1186         int err;
1187         struct dst_entry *dst = skb->dst;
1188         struct neighbour *neigh = dst->neighbour;
1189         struct net_device *dev = neigh->dev;
1190
1191         __skb_pull(skb, skb->nh.raw - skb->data);
1192
1193         read_lock_bh(&neigh->lock);
1194         err = dev->hard_header(skb, dev, ntohs(skb->protocol),
1195                                neigh->ha, NULL, skb->len);
1196         read_unlock_bh(&neigh->lock);
1197         if (err >= 0)
1198                 err = neigh->ops->queue_xmit(skb);
1199         else {
1200                 err = -EINVAL;
1201                 kfree_skb(skb);
1202         }
1203         return err;
1204 }
1205
1206 static void neigh_proxy_process(unsigned long arg)
1207 {
1208         struct neigh_table *tbl = (struct neigh_table *)arg;
1209         long sched_next = 0;
1210         unsigned long now = jiffies;
1211         struct sk_buff *skb;
1212
1213         spin_lock(&tbl->proxy_queue.lock);
1214
1215         skb = tbl->proxy_queue.next;
1216
1217         while (skb != (struct sk_buff *)&tbl->proxy_queue) {
1218                 struct sk_buff *back = skb;
1219                 long tdif = NEIGH_CB(back)->sched_next - now;
1220
1221                 skb = skb->next;
1222                 if (tdif <= 0) {
1223                         struct net_device *dev = back->dev;
1224                         __skb_unlink(back, &tbl->proxy_queue);
1225                         if (tbl->proxy_redo && netif_running(dev))
1226                                 tbl->proxy_redo(back);
1227                         else
1228                                 kfree_skb(back);
1229
1230                         dev_put(dev);
1231                 } else if (!sched_next || tdif < sched_next)
1232                         sched_next = tdif;
1233         }
1234         del_timer(&tbl->proxy_timer);
1235         if (sched_next)
1236                 mod_timer(&tbl->proxy_timer, jiffies + sched_next);
1237         spin_unlock(&tbl->proxy_queue.lock);
1238 }
1239
1240 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
1241                     struct sk_buff *skb)
1242 {
1243         unsigned long now = jiffies;
1244         unsigned long sched_next = now + (net_random() % p->proxy_delay);
1245
1246         if (tbl->proxy_queue.qlen > p->proxy_qlen) {
1247                 kfree_skb(skb);
1248                 return;
1249         }
1250
1251         NEIGH_CB(skb)->sched_next = sched_next;
1252         NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED;
1253
1254         spin_lock(&tbl->proxy_queue.lock);
1255         if (del_timer(&tbl->proxy_timer)) {
1256                 if (time_before(tbl->proxy_timer.expires, sched_next))
1257                         sched_next = tbl->proxy_timer.expires;
1258         }
1259         dst_release(skb->dst);
1260         skb->dst = NULL;
1261         dev_hold(skb->dev);
1262         __skb_queue_tail(&tbl->proxy_queue, skb);
1263         mod_timer(&tbl->proxy_timer, sched_next);
1264         spin_unlock(&tbl->proxy_queue.lock);
1265 }
1266
1267
1268 struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
1269                                       struct neigh_table *tbl)
1270 {
1271         struct neigh_parms *p = kmemdup(&tbl->parms, sizeof(*p), GFP_KERNEL);
1272
1273         if (p) {
1274                 p->tbl            = tbl;
1275                 atomic_set(&p->refcnt, 1);
1276                 INIT_RCU_HEAD(&p->rcu_head);
1277                 p->reachable_time =
1278                                 neigh_rand_reach_time(p->base_reachable_time);
1279                 if (dev) {
1280                         if (dev->neigh_setup && dev->neigh_setup(dev, p)) {
1281                                 kfree(p);
1282                                 return NULL;
1283                         }
1284
1285                         dev_hold(dev);
1286                         p->dev = dev;
1287                 }
1288                 p->sysctl_table = NULL;
1289                 write_lock_bh(&tbl->lock);
1290                 p->next         = tbl->parms.next;
1291                 tbl->parms.next = p;
1292                 write_unlock_bh(&tbl->lock);
1293         }
1294         return p;
1295 }
1296
1297 static void neigh_rcu_free_parms(struct rcu_head *head)
1298 {
1299         struct neigh_parms *parms =
1300                 container_of(head, struct neigh_parms, rcu_head);
1301
1302         neigh_parms_put(parms);
1303 }
1304
1305 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
1306 {
1307         struct neigh_parms **p;
1308
1309         if (!parms || parms == &tbl->parms)
1310                 return;
1311         write_lock_bh(&tbl->lock);
1312         for (p = &tbl->parms.next; *p; p = &(*p)->next) {
1313                 if (*p == parms) {
1314                         *p = parms->next;
1315                         parms->dead = 1;
1316                         write_unlock_bh(&tbl->lock);
1317                         if (parms->dev)
1318                                 dev_put(parms->dev);
1319                         call_rcu(&parms->rcu_head, neigh_rcu_free_parms);
1320                         return;
1321                 }
1322         }
1323         write_unlock_bh(&tbl->lock);
1324         NEIGH_PRINTK1("neigh_parms_release: not found\n");
1325 }
1326
1327 void neigh_parms_destroy(struct neigh_parms *parms)
1328 {
1329         kfree(parms);
1330 }
1331
1332 void neigh_table_init_no_netlink(struct neigh_table *tbl)
1333 {
1334         unsigned long now = jiffies;
1335         unsigned long phsize;
1336
1337         atomic_set(&tbl->parms.refcnt, 1);
1338         INIT_RCU_HEAD(&tbl->parms.rcu_head);
1339         tbl->parms.reachable_time =
1340                           neigh_rand_reach_time(tbl->parms.base_reachable_time);
1341
1342         if (!tbl->kmem_cachep)
1343                 tbl->kmem_cachep =
1344                         kmem_cache_create(tbl->id, tbl->entry_size, 0,
1345                                           SLAB_HWCACHE_ALIGN|SLAB_PANIC,
1346                                           NULL, NULL);
1347         tbl->stats = alloc_percpu(struct neigh_statistics);
1348         if (!tbl->stats)
1349                 panic("cannot create neighbour cache statistics");
1350
1351 #ifdef CONFIG_PROC_FS
1352         tbl->pde = create_proc_entry(tbl->id, 0, proc_net_stat);
1353         if (!tbl->pde)
1354                 panic("cannot create neighbour proc dir entry");
1355         tbl->pde->proc_fops = &neigh_stat_seq_fops;
1356         tbl->pde->data = tbl;
1357 #endif
1358
1359         tbl->hash_mask = 1;
1360         tbl->hash_buckets = neigh_hash_alloc(tbl->hash_mask + 1);
1361
1362         phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *);
1363         tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL);
1364
1365         if (!tbl->hash_buckets || !tbl->phash_buckets)
1366                 panic("cannot allocate neighbour cache hashes");
1367
1368         get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd));
1369
1370         rwlock_init(&tbl->lock);
1371         init_timer(&tbl->gc_timer);
1372         tbl->gc_timer.data     = (unsigned long)tbl;
1373         tbl->gc_timer.function = neigh_periodic_timer;
1374         tbl->gc_timer.expires  = now + 1;
1375         add_timer(&tbl->gc_timer);
1376
1377         init_timer(&tbl->proxy_timer);
1378         tbl->proxy_timer.data     = (unsigned long)tbl;
1379         tbl->proxy_timer.function = neigh_proxy_process;
1380         skb_queue_head_init(&tbl->proxy_queue);
1381
1382         tbl->last_flush = now;
1383         tbl->last_rand  = now + tbl->parms.reachable_time * 20;
1384 }
1385
1386 void neigh_table_init(struct neigh_table *tbl)
1387 {
1388         struct neigh_table *tmp;
1389
1390         neigh_table_init_no_netlink(tbl);
1391         write_lock(&neigh_tbl_lock);
1392         for (tmp = neigh_tables; tmp; tmp = tmp->next) {
1393                 if (tmp->family == tbl->family)
1394                         break;
1395         }
1396         tbl->next       = neigh_tables;
1397         neigh_tables    = tbl;
1398         write_unlock(&neigh_tbl_lock);
1399
1400         if (unlikely(tmp)) {
1401                 printk(KERN_ERR "NEIGH: Registering multiple tables for "
1402                        "family %d\n", tbl->family);
1403                 dump_stack();
1404         }
1405 }
1406
1407 int neigh_table_clear(struct neigh_table *tbl)
1408 {
1409         struct neigh_table **tp;
1410
1411         /* It is not clean... Fix it to unload IPv6 module safely */
1412         del_timer_sync(&tbl->gc_timer);
1413         del_timer_sync(&tbl->proxy_timer);
1414         pneigh_queue_purge(&tbl->proxy_queue);
1415         neigh_ifdown(tbl, NULL);
1416         if (atomic_read(&tbl->entries))
1417                 printk(KERN_CRIT "neighbour leakage\n");
1418         write_lock(&neigh_tbl_lock);
1419         for (tp = &neigh_tables; *tp; tp = &(*tp)->next) {
1420                 if (*tp == tbl) {
1421                         *tp = tbl->next;
1422                         break;
1423                 }
1424         }
1425         write_unlock(&neigh_tbl_lock);
1426
1427         neigh_hash_free(tbl->hash_buckets, tbl->hash_mask + 1);
1428         tbl->hash_buckets = NULL;
1429
1430         kfree(tbl->phash_buckets);
1431         tbl->phash_buckets = NULL;
1432
1433         free_percpu(tbl->stats);
1434         tbl->stats = NULL;
1435
1436         return 0;
1437 }
1438
1439 int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1440 {
1441         struct ndmsg *ndm;
1442         struct nlattr *dst_attr;
1443         struct neigh_table *tbl;
1444         struct net_device *dev = NULL;
1445         int err = -EINVAL;
1446
1447         if (nlmsg_len(nlh) < sizeof(*ndm))
1448                 goto out;
1449
1450         dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
1451         if (dst_attr == NULL)
1452                 goto out;
1453
1454         ndm = nlmsg_data(nlh);
1455         if (ndm->ndm_ifindex) {
1456                 dev = dev_get_by_index(ndm->ndm_ifindex);
1457                 if (dev == NULL) {
1458                         err = -ENODEV;
1459                         goto out;
1460                 }
1461         }
1462
1463         read_lock(&neigh_tbl_lock);
1464         for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1465                 struct neighbour *neigh;
1466
1467                 if (tbl->family != ndm->ndm_family)
1468                         continue;
1469                 read_unlock(&neigh_tbl_lock);
1470
1471                 if (nla_len(dst_attr) < tbl->key_len)
1472                         goto out_dev_put;
1473
1474                 if (ndm->ndm_flags & NTF_PROXY) {
1475                         err = pneigh_delete(tbl, nla_data(dst_attr), dev);
1476                         goto out_dev_put;
1477                 }
1478
1479                 if (dev == NULL)
1480                         goto out_dev_put;
1481
1482                 neigh = neigh_lookup(tbl, nla_data(dst_attr), dev);
1483                 if (neigh == NULL) {
1484                         err = -ENOENT;
1485                         goto out_dev_put;
1486                 }
1487
1488                 err = neigh_update(neigh, NULL, NUD_FAILED,
1489                                    NEIGH_UPDATE_F_OVERRIDE |
1490                                    NEIGH_UPDATE_F_ADMIN);
1491                 neigh_release(neigh);
1492                 goto out_dev_put;
1493         }
1494         read_unlock(&neigh_tbl_lock);
1495         err = -EAFNOSUPPORT;
1496
1497 out_dev_put:
1498         if (dev)
1499                 dev_put(dev);
1500 out:
1501         return err;
1502 }
1503
1504 int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1505 {
1506         struct ndmsg *ndm;
1507         struct nlattr *tb[NDA_MAX+1];
1508         struct neigh_table *tbl;
1509         struct net_device *dev = NULL;
1510         int err;
1511
1512         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
1513         if (err < 0)
1514                 goto out;
1515
1516         err = -EINVAL;
1517         if (tb[NDA_DST] == NULL)
1518                 goto out;
1519
1520         ndm = nlmsg_data(nlh);
1521         if (ndm->ndm_ifindex) {
1522                 dev = dev_get_by_index(ndm->ndm_ifindex);
1523                 if (dev == NULL) {
1524                         err = -ENODEV;
1525                         goto out;
1526                 }
1527
1528                 if (tb[NDA_LLADDR] && nla_len(tb[NDA_LLADDR]) < dev->addr_len)
1529                         goto out_dev_put;
1530         }
1531
1532         read_lock(&neigh_tbl_lock);
1533         for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1534                 int flags = NEIGH_UPDATE_F_ADMIN | NEIGH_UPDATE_F_OVERRIDE;
1535                 struct neighbour *neigh;
1536                 void *dst, *lladdr;
1537
1538                 if (tbl->family != ndm->ndm_family)
1539                         continue;
1540                 read_unlock(&neigh_tbl_lock);
1541
1542                 if (nla_len(tb[NDA_DST]) < tbl->key_len)
1543                         goto out_dev_put;
1544                 dst = nla_data(tb[NDA_DST]);
1545                 lladdr = tb[NDA_LLADDR] ? nla_data(tb[NDA_LLADDR]) : NULL;
1546
1547                 if (ndm->ndm_flags & NTF_PROXY) {
1548                         struct pneigh_entry *pn;
1549
1550                         err = -ENOBUFS;
1551                         pn = pneigh_lookup(tbl, dst, dev, 1);
1552                         if (pn) {
1553                                 pn->flags = ndm->ndm_flags;
1554                                 err = 0;
1555                         }
1556                         goto out_dev_put;
1557                 }
1558
1559                 if (dev == NULL)
1560                         goto out_dev_put;
1561
1562                 neigh = neigh_lookup(tbl, dst, dev);
1563                 if (neigh == NULL) {
1564                         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1565                                 err = -ENOENT;
1566                                 goto out_dev_put;
1567                         }
1568
1569                         neigh = __neigh_lookup_errno(tbl, dst, dev);
1570                         if (IS_ERR(neigh)) {
1571                                 err = PTR_ERR(neigh);
1572                                 goto out_dev_put;
1573                         }
1574                 } else {
1575                         if (nlh->nlmsg_flags & NLM_F_EXCL) {
1576                                 err = -EEXIST;
1577                                 neigh_release(neigh);
1578                                 goto out_dev_put;
1579                         }
1580
1581                         if (!(nlh->nlmsg_flags & NLM_F_REPLACE))
1582                                 flags &= ~NEIGH_UPDATE_F_OVERRIDE;
1583                 }
1584
1585                 err = neigh_update(neigh, lladdr, ndm->ndm_state, flags);
1586                 neigh_release(neigh);
1587                 goto out_dev_put;
1588         }
1589
1590         read_unlock(&neigh_tbl_lock);
1591         err = -EAFNOSUPPORT;
1592
1593 out_dev_put:
1594         if (dev)
1595                 dev_put(dev);
1596 out:
1597         return err;
1598 }
1599
1600 static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
1601 {
1602         struct nlattr *nest;
1603
1604         nest = nla_nest_start(skb, NDTA_PARMS);
1605         if (nest == NULL)
1606                 return -ENOBUFS;
1607
1608         if (parms->dev)
1609                 NLA_PUT_U32(skb, NDTPA_IFINDEX, parms->dev->ifindex);
1610
1611         NLA_PUT_U32(skb, NDTPA_REFCNT, atomic_read(&parms->refcnt));
1612         NLA_PUT_U32(skb, NDTPA_QUEUE_LEN, parms->queue_len);
1613         NLA_PUT_U32(skb, NDTPA_PROXY_QLEN, parms->proxy_qlen);
1614         NLA_PUT_U32(skb, NDTPA_APP_PROBES, parms->app_probes);
1615         NLA_PUT_U32(skb, NDTPA_UCAST_PROBES, parms->ucast_probes);
1616         NLA_PUT_U32(skb, NDTPA_MCAST_PROBES, parms->mcast_probes);
1617         NLA_PUT_MSECS(skb, NDTPA_REACHABLE_TIME, parms->reachable_time);
1618         NLA_PUT_MSECS(skb, NDTPA_BASE_REACHABLE_TIME,
1619                       parms->base_reachable_time);
1620         NLA_PUT_MSECS(skb, NDTPA_GC_STALETIME, parms->gc_staletime);
1621         NLA_PUT_MSECS(skb, NDTPA_DELAY_PROBE_TIME, parms->delay_probe_time);
1622         NLA_PUT_MSECS(skb, NDTPA_RETRANS_TIME, parms->retrans_time);
1623         NLA_PUT_MSECS(skb, NDTPA_ANYCAST_DELAY, parms->anycast_delay);
1624         NLA_PUT_MSECS(skb, NDTPA_PROXY_DELAY, parms->proxy_delay);
1625         NLA_PUT_MSECS(skb, NDTPA_LOCKTIME, parms->locktime);
1626
1627         return nla_nest_end(skb, nest);
1628
1629 nla_put_failure:
1630         return nla_nest_cancel(skb, nest);
1631 }
1632
1633 static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
1634                               u32 pid, u32 seq, int type, int flags)
1635 {
1636         struct nlmsghdr *nlh;
1637         struct ndtmsg *ndtmsg;
1638
1639         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1640         if (nlh == NULL)
1641                 return -EMSGSIZE;
1642
1643         ndtmsg = nlmsg_data(nlh);
1644
1645         read_lock_bh(&tbl->lock);
1646         ndtmsg->ndtm_family = tbl->family;
1647         ndtmsg->ndtm_pad1   = 0;
1648         ndtmsg->ndtm_pad2   = 0;
1649
1650         NLA_PUT_STRING(skb, NDTA_NAME, tbl->id);
1651         NLA_PUT_MSECS(skb, NDTA_GC_INTERVAL, tbl->gc_interval);
1652         NLA_PUT_U32(skb, NDTA_THRESH1, tbl->gc_thresh1);
1653         NLA_PUT_U32(skb, NDTA_THRESH2, tbl->gc_thresh2);
1654         NLA_PUT_U32(skb, NDTA_THRESH3, tbl->gc_thresh3);
1655
1656         {
1657                 unsigned long now = jiffies;
1658                 unsigned int flush_delta = now - tbl->last_flush;
1659                 unsigned int rand_delta = now - tbl->last_rand;
1660
1661                 struct ndt_config ndc = {
1662                         .ndtc_key_len           = tbl->key_len,
1663                         .ndtc_entry_size        = tbl->entry_size,
1664                         .ndtc_entries           = atomic_read(&tbl->entries),
1665                         .ndtc_last_flush        = jiffies_to_msecs(flush_delta),
1666                         .ndtc_last_rand         = jiffies_to_msecs(rand_delta),
1667                         .ndtc_hash_rnd          = tbl->hash_rnd,
1668                         .ndtc_hash_mask         = tbl->hash_mask,
1669                         .ndtc_hash_chain_gc     = tbl->hash_chain_gc,
1670                         .ndtc_proxy_qlen        = tbl->proxy_queue.qlen,
1671                 };
1672
1673                 NLA_PUT(skb, NDTA_CONFIG, sizeof(ndc), &ndc);
1674         }
1675
1676         {
1677                 int cpu;
1678                 struct ndt_stats ndst;
1679
1680                 memset(&ndst, 0, sizeof(ndst));
1681
1682                 for_each_possible_cpu(cpu) {
1683                         struct neigh_statistics *st;
1684
1685                         st = per_cpu_ptr(tbl->stats, cpu);
1686                         ndst.ndts_allocs                += st->allocs;
1687                         ndst.ndts_destroys              += st->destroys;
1688                         ndst.ndts_hash_grows            += st->hash_grows;
1689                         ndst.ndts_res_failed            += st->res_failed;
1690                         ndst.ndts_lookups               += st->lookups;
1691                         ndst.ndts_hits                  += st->hits;
1692                         ndst.ndts_rcv_probes_mcast      += st->rcv_probes_mcast;
1693                         ndst.ndts_rcv_probes_ucast      += st->rcv_probes_ucast;
1694                         ndst.ndts_periodic_gc_runs      += st->periodic_gc_runs;
1695                         ndst.ndts_forced_gc_runs        += st->forced_gc_runs;
1696                 }
1697
1698                 NLA_PUT(skb, NDTA_STATS, sizeof(ndst), &ndst);
1699         }
1700
1701         BUG_ON(tbl->parms.dev);
1702         if (neightbl_fill_parms(skb, &tbl->parms) < 0)
1703                 goto nla_put_failure;
1704
1705         read_unlock_bh(&tbl->lock);
1706         return nlmsg_end(skb, nlh);
1707
1708 nla_put_failure:
1709         read_unlock_bh(&tbl->lock);
1710         nlmsg_cancel(skb, nlh);
1711         return -EMSGSIZE;
1712 }
1713
1714 static int neightbl_fill_param_info(struct sk_buff *skb,
1715                                     struct neigh_table *tbl,
1716                                     struct neigh_parms *parms,
1717                                     u32 pid, u32 seq, int type,
1718                                     unsigned int flags)
1719 {
1720         struct ndtmsg *ndtmsg;
1721         struct nlmsghdr *nlh;
1722
1723         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1724         if (nlh == NULL)
1725                 return -EMSGSIZE;
1726
1727         ndtmsg = nlmsg_data(nlh);
1728
1729         read_lock_bh(&tbl->lock);
1730         ndtmsg->ndtm_family = tbl->family;
1731         ndtmsg->ndtm_pad1   = 0;
1732         ndtmsg->ndtm_pad2   = 0;
1733
1734         if (nla_put_string(skb, NDTA_NAME, tbl->id) < 0 ||
1735             neightbl_fill_parms(skb, parms) < 0)
1736                 goto errout;
1737
1738         read_unlock_bh(&tbl->lock);
1739         return nlmsg_end(skb, nlh);
1740 errout:
1741         read_unlock_bh(&tbl->lock);
1742         nlmsg_cancel(skb, nlh);
1743         return -EMSGSIZE;
1744 }
1745
1746 static inline struct neigh_parms *lookup_neigh_params(struct neigh_table *tbl,
1747                                                       int ifindex)
1748 {
1749         struct neigh_parms *p;
1750
1751         for (p = &tbl->parms; p; p = p->next)
1752                 if ((p->dev && p->dev->ifindex == ifindex) ||
1753                     (!p->dev && !ifindex))
1754                         return p;
1755
1756         return NULL;
1757 }
1758
1759 static struct nla_policy nl_neightbl_policy[NDTA_MAX+1] __read_mostly = {
1760         [NDTA_NAME]             = { .type = NLA_STRING },
1761         [NDTA_THRESH1]          = { .type = NLA_U32 },
1762         [NDTA_THRESH2]          = { .type = NLA_U32 },
1763         [NDTA_THRESH3]          = { .type = NLA_U32 },
1764         [NDTA_GC_INTERVAL]      = { .type = NLA_U64 },
1765         [NDTA_PARMS]            = { .type = NLA_NESTED },
1766 };
1767
1768 static struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] __read_mostly = {
1769         [NDTPA_IFINDEX]                 = { .type = NLA_U32 },
1770         [NDTPA_QUEUE_LEN]               = { .type = NLA_U32 },
1771         [NDTPA_PROXY_QLEN]              = { .type = NLA_U32 },
1772         [NDTPA_APP_PROBES]              = { .type = NLA_U32 },
1773         [NDTPA_UCAST_PROBES]            = { .type = NLA_U32 },
1774         [NDTPA_MCAST_PROBES]            = { .type = NLA_U32 },
1775         [NDTPA_BASE_REACHABLE_TIME]     = { .type = NLA_U64 },
1776         [NDTPA_GC_STALETIME]            = { .type = NLA_U64 },
1777         [NDTPA_DELAY_PROBE_TIME]        = { .type = NLA_U64 },
1778         [NDTPA_RETRANS_TIME]            = { .type = NLA_U64 },
1779         [NDTPA_ANYCAST_DELAY]           = { .type = NLA_U64 },
1780         [NDTPA_PROXY_DELAY]             = { .type = NLA_U64 },
1781         [NDTPA_LOCKTIME]                = { .type = NLA_U64 },
1782 };
1783
1784 int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1785 {
1786         struct neigh_table *tbl;
1787         struct ndtmsg *ndtmsg;
1788         struct nlattr *tb[NDTA_MAX+1];
1789         int err;
1790
1791         err = nlmsg_parse(nlh, sizeof(*ndtmsg), tb, NDTA_MAX,
1792                           nl_neightbl_policy);
1793         if (err < 0)
1794                 goto errout;
1795
1796         if (tb[NDTA_NAME] == NULL) {
1797                 err = -EINVAL;
1798                 goto errout;
1799         }
1800
1801         ndtmsg = nlmsg_data(nlh);
1802         read_lock(&neigh_tbl_lock);
1803         for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1804                 if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
1805                         continue;
1806
1807                 if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0)
1808                         break;
1809         }
1810
1811         if (tbl == NULL) {
1812                 err = -ENOENT;
1813                 goto errout_locked;
1814         }
1815
1816         /*
1817          * We acquire tbl->lock to be nice to the periodic timers and
1818          * make sure they always see a consistent set of values.
1819          */
1820         write_lock_bh(&tbl->lock);
1821
1822         if (tb[NDTA_PARMS]) {
1823                 struct nlattr *tbp[NDTPA_MAX+1];
1824                 struct neigh_parms *p;
1825                 int i, ifindex = 0;
1826
1827                 err = nla_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS],
1828                                        nl_ntbl_parm_policy);
1829                 if (err < 0)
1830                         goto errout_tbl_lock;
1831
1832                 if (tbp[NDTPA_IFINDEX])
1833                         ifindex = nla_get_u32(tbp[NDTPA_IFINDEX]);
1834
1835                 p = lookup_neigh_params(tbl, ifindex);
1836                 if (p == NULL) {
1837                         err = -ENOENT;
1838                         goto errout_tbl_lock;
1839                 }
1840
1841                 for (i = 1; i <= NDTPA_MAX; i++) {
1842                         if (tbp[i] == NULL)
1843                                 continue;
1844
1845                         switch (i) {
1846                         case NDTPA_QUEUE_LEN:
1847                                 p->queue_len = nla_get_u32(tbp[i]);
1848                                 break;
1849                         case NDTPA_PROXY_QLEN:
1850                                 p->proxy_qlen = nla_get_u32(tbp[i]);
1851                                 break;
1852                         case NDTPA_APP_PROBES:
1853                                 p->app_probes = nla_get_u32(tbp[i]);
1854                                 break;
1855                         case NDTPA_UCAST_PROBES:
1856                                 p->ucast_probes = nla_get_u32(tbp[i]);
1857                                 break;
1858                         case NDTPA_MCAST_PROBES:
1859                                 p->mcast_probes = nla_get_u32(tbp[i]);
1860                                 break;
1861                         case NDTPA_BASE_REACHABLE_TIME:
1862                                 p->base_reachable_time = nla_get_msecs(tbp[i]);
1863                                 break;
1864                         case NDTPA_GC_STALETIME:
1865                                 p->gc_staletime = nla_get_msecs(tbp[i]);
1866                                 break;
1867                         case NDTPA_DELAY_PROBE_TIME:
1868                                 p->delay_probe_time = nla_get_msecs(tbp[i]);
1869                                 break;
1870                         case NDTPA_RETRANS_TIME:
1871                                 p->retrans_time = nla_get_msecs(tbp[i]);
1872                                 break;
1873                         case NDTPA_ANYCAST_DELAY:
1874                                 p->anycast_delay = nla_get_msecs(tbp[i]);
1875                                 break;
1876                         case NDTPA_PROXY_DELAY:
1877                                 p->proxy_delay = nla_get_msecs(tbp[i]);
1878                                 break;
1879                         case NDTPA_LOCKTIME:
1880                                 p->locktime = nla_get_msecs(tbp[i]);
1881                                 break;
1882                         }
1883                 }
1884         }
1885
1886         if (tb[NDTA_THRESH1])
1887                 tbl->gc_thresh1 = nla_get_u32(tb[NDTA_THRESH1]);
1888
1889         if (tb[NDTA_THRESH2])
1890                 tbl->gc_thresh2 = nla_get_u32(tb[NDTA_THRESH2]);
1891
1892         if (tb[NDTA_THRESH3])
1893                 tbl->gc_thresh3 = nla_get_u32(tb[NDTA_THRESH3]);
1894
1895         if (tb[NDTA_GC_INTERVAL])
1896                 tbl->gc_interval = nla_get_msecs(tb[NDTA_GC_INTERVAL]);
1897
1898         err = 0;
1899
1900 errout_tbl_lock:
1901         write_unlock_bh(&tbl->lock);
1902 errout_locked:
1903         read_unlock(&neigh_tbl_lock);
1904 errout:
1905         return err;
1906 }
1907
1908 int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
1909 {
1910         int family, tidx, nidx = 0;
1911         int tbl_skip = cb->args[0];
1912         int neigh_skip = cb->args[1];
1913         struct neigh_table *tbl;
1914
1915         family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
1916
1917         read_lock(&neigh_tbl_lock);
1918         for (tbl = neigh_tables, tidx = 0; tbl; tbl = tbl->next, tidx++) {
1919                 struct neigh_parms *p;
1920
1921                 if (tidx < tbl_skip || (family && tbl->family != family))
1922                         continue;
1923
1924                 if (neightbl_fill_info(skb, tbl, NETLINK_CB(cb->skb).pid,
1925                                        cb->nlh->nlmsg_seq, RTM_NEWNEIGHTBL,
1926                                        NLM_F_MULTI) <= 0)
1927                         break;
1928
1929                 for (nidx = 0, p = tbl->parms.next; p; p = p->next, nidx++) {
1930                         if (nidx < neigh_skip)
1931                                 continue;
1932
1933                         if (neightbl_fill_param_info(skb, tbl, p,
1934                                                      NETLINK_CB(cb->skb).pid,
1935                                                      cb->nlh->nlmsg_seq,
1936                                                      RTM_NEWNEIGHTBL,
1937                                                      NLM_F_MULTI) <= 0)
1938                                 goto out;
1939                 }
1940
1941                 neigh_skip = 0;
1942         }
1943 out:
1944         read_unlock(&neigh_tbl_lock);
1945         cb->args[0] = tidx;
1946         cb->args[1] = nidx;
1947
1948         return skb->len;
1949 }
1950
1951 static int neigh_fill_info(struct sk_buff *skb, struct neighbour *neigh,
1952                            u32 pid, u32 seq, int type, unsigned int flags)
1953 {
1954         unsigned long now = jiffies;
1955         struct nda_cacheinfo ci;
1956         struct nlmsghdr *nlh;
1957         struct ndmsg *ndm;
1958
1959         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
1960         if (nlh == NULL)
1961                 return -EMSGSIZE;
1962
1963         ndm = nlmsg_data(nlh);
1964         ndm->ndm_family  = neigh->ops->family;
1965         ndm->ndm_pad1    = 0;
1966         ndm->ndm_pad2    = 0;
1967         ndm->ndm_flags   = neigh->flags;
1968         ndm->ndm_type    = neigh->type;
1969         ndm->ndm_ifindex = neigh->dev->ifindex;
1970
1971         NLA_PUT(skb, NDA_DST, neigh->tbl->key_len, neigh->primary_key);
1972
1973         read_lock_bh(&neigh->lock);
1974         ndm->ndm_state   = neigh->nud_state;
1975         if ((neigh->nud_state & NUD_VALID) &&
1976             nla_put(skb, NDA_LLADDR, neigh->dev->addr_len, neigh->ha) < 0) {
1977                 read_unlock_bh(&neigh->lock);
1978                 goto nla_put_failure;
1979         }
1980
1981         ci.ndm_used      = now - neigh->used;
1982         ci.ndm_confirmed = now - neigh->confirmed;
1983         ci.ndm_updated   = now - neigh->updated;
1984         ci.ndm_refcnt    = atomic_read(&neigh->refcnt) - 1;
1985         read_unlock_bh(&neigh->lock);
1986
1987         NLA_PUT_U32(skb, NDA_PROBES, atomic_read(&neigh->probes));
1988         NLA_PUT(skb, NDA_CACHEINFO, sizeof(ci), &ci);
1989
1990         return nlmsg_end(skb, nlh);
1991
1992 nla_put_failure:
1993         nlmsg_cancel(skb, nlh);
1994         return -EMSGSIZE;
1995 }
1996
1997
1998 static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
1999                             struct netlink_callback *cb)
2000 {
2001         struct neighbour *n;
2002         int rc, h, s_h = cb->args[1];
2003         int idx, s_idx = idx = cb->args[2];
2004
2005         read_lock_bh(&tbl->lock);
2006         for (h = 0; h <= tbl->hash_mask; h++) {
2007                 if (h < s_h)
2008                         continue;
2009                 if (h > s_h)
2010                         s_idx = 0;
2011                 for (n = tbl->hash_buckets[h], idx = 0; n; n = n->next, idx++) {
2012                         if (idx < s_idx)
2013                                 continue;
2014                         if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid,
2015                                             cb->nlh->nlmsg_seq,
2016                                             RTM_NEWNEIGH,
2017                                             NLM_F_MULTI) <= 0) {
2018                                 read_unlock_bh(&tbl->lock);
2019                                 rc = -1;
2020                                 goto out;
2021                         }
2022                 }
2023         }
2024         read_unlock_bh(&tbl->lock);
2025         rc = skb->len;
2026 out:
2027         cb->args[1] = h;
2028         cb->args[2] = idx;
2029         return rc;
2030 }
2031
2032 int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2033 {
2034         struct neigh_table *tbl;
2035         int t, family, s_t;
2036
2037         read_lock(&neigh_tbl_lock);
2038         family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2039         s_t = cb->args[0];
2040
2041         for (tbl = neigh_tables, t = 0; tbl; tbl = tbl->next, t++) {
2042                 if (t < s_t || (family && tbl->family != family))
2043                         continue;
2044                 if (t > s_t)
2045                         memset(&cb->args[1], 0, sizeof(cb->args) -
2046                                                 sizeof(cb->args[0]));
2047                 if (neigh_dump_table(tbl, skb, cb) < 0)
2048                         break;
2049         }
2050         read_unlock(&neigh_tbl_lock);
2051
2052         cb->args[0] = t;
2053         return skb->len;
2054 }
2055
2056 void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie)
2057 {
2058         int chain;
2059
2060         read_lock_bh(&tbl->lock);
2061         for (chain = 0; chain <= tbl->hash_mask; chain++) {
2062                 struct neighbour *n;
2063
2064                 for (n = tbl->hash_buckets[chain]; n; n = n->next)
2065                         cb(n, cookie);
2066         }
2067         read_unlock_bh(&tbl->lock);
2068 }
2069 EXPORT_SYMBOL(neigh_for_each);
2070
2071 /* The tbl->lock must be held as a writer and BH disabled. */
2072 void __neigh_for_each_release(struct neigh_table *tbl,
2073                               int (*cb)(struct neighbour *))
2074 {
2075         int chain;
2076
2077         for (chain = 0; chain <= tbl->hash_mask; chain++) {
2078                 struct neighbour *n, **np;
2079
2080                 np = &tbl->hash_buckets[chain];
2081                 while ((n = *np) != NULL) {
2082                         int release;
2083
2084                         write_lock(&n->lock);
2085                         release = cb(n);
2086                         if (release) {
2087                                 *np = n->next;
2088                                 n->dead = 1;
2089                         } else
2090                                 np = &n->next;
2091                         write_unlock(&n->lock);
2092                         if (release)
2093                                 neigh_release(n);
2094                 }
2095         }
2096 }
2097 EXPORT_SYMBOL(__neigh_for_each_release);
2098
2099 #ifdef CONFIG_PROC_FS
2100
2101 static struct neighbour *neigh_get_first(struct seq_file *seq)
2102 {
2103         struct neigh_seq_state *state = seq->private;
2104         struct neigh_table *tbl = state->tbl;
2105         struct neighbour *n = NULL;
2106         int bucket = state->bucket;
2107
2108         state->flags &= ~NEIGH_SEQ_IS_PNEIGH;
2109         for (bucket = 0; bucket <= tbl->hash_mask; bucket++) {
2110                 n = tbl->hash_buckets[bucket];
2111
2112                 while (n) {
2113                         if (state->neigh_sub_iter) {
2114                                 loff_t fakep = 0;
2115                                 void *v;
2116
2117                                 v = state->neigh_sub_iter(state, n, &fakep);
2118                                 if (!v)
2119                                         goto next;
2120                         }
2121                         if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2122                                 break;
2123                         if (n->nud_state & ~NUD_NOARP)
2124                                 break;
2125                 next:
2126                         n = n->next;
2127                 }
2128
2129                 if (n)
2130                         break;
2131         }
2132         state->bucket = bucket;
2133
2134         return n;
2135 }
2136
2137 static struct neighbour *neigh_get_next(struct seq_file *seq,
2138                                         struct neighbour *n,
2139                                         loff_t *pos)
2140 {
2141         struct neigh_seq_state *state = seq->private;
2142         struct neigh_table *tbl = state->tbl;
2143
2144         if (state->neigh_sub_iter) {
2145                 void *v = state->neigh_sub_iter(state, n, pos);
2146                 if (v)
2147                         return n;
2148         }
2149         n = n->next;
2150
2151         while (1) {
2152                 while (n) {
2153                         if (state->neigh_sub_iter) {
2154                                 void *v = state->neigh_sub_iter(state, n, pos);
2155                                 if (v)
2156                                         return n;
2157                                 goto next;
2158                         }
2159                         if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2160                                 break;
2161
2162                         if (n->nud_state & ~NUD_NOARP)
2163                                 break;
2164                 next:
2165                         n = n->next;
2166                 }
2167
2168                 if (n)
2169                         break;
2170
2171                 if (++state->bucket > tbl->hash_mask)
2172                         break;
2173
2174                 n = tbl->hash_buckets[state->bucket];
2175         }
2176
2177         if (n && pos)
2178                 --(*pos);
2179         return n;
2180 }
2181
2182 static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos)
2183 {
2184         struct neighbour *n = neigh_get_first(seq);
2185
2186         if (n) {
2187                 while (*pos) {
2188                         n = neigh_get_next(seq, n, pos);
2189                         if (!n)
2190                                 break;
2191                 }
2192         }
2193         return *pos ? NULL : n;
2194 }
2195
2196 static struct pneigh_entry *pneigh_get_first(struct seq_file *seq)
2197 {
2198         struct neigh_seq_state *state = seq->private;
2199         struct neigh_table *tbl = state->tbl;
2200         struct pneigh_entry *pn = NULL;
2201         int bucket = state->bucket;
2202
2203         state->flags |= NEIGH_SEQ_IS_PNEIGH;
2204         for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) {
2205                 pn = tbl->phash_buckets[bucket];
2206                 if (pn)
2207                         break;
2208         }
2209         state->bucket = bucket;
2210
2211         return pn;
2212 }
2213
2214 static struct pneigh_entry *pneigh_get_next(struct seq_file *seq,
2215                                             struct pneigh_entry *pn,
2216                                             loff_t *pos)
2217 {
2218         struct neigh_seq_state *state = seq->private;
2219         struct neigh_table *tbl = state->tbl;
2220
2221         pn = pn->next;
2222         while (!pn) {
2223                 if (++state->bucket > PNEIGH_HASHMASK)
2224                         break;
2225                 pn = tbl->phash_buckets[state->bucket];
2226                 if (pn)
2227                         break;
2228         }
2229
2230         if (pn && pos)
2231                 --(*pos);
2232
2233         return pn;
2234 }
2235
2236 static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos)
2237 {
2238         struct pneigh_entry *pn = pneigh_get_first(seq);
2239
2240         if (pn) {
2241                 while (*pos) {
2242                         pn = pneigh_get_next(seq, pn, pos);
2243                         if (!pn)
2244                                 break;
2245                 }
2246         }
2247         return *pos ? NULL : pn;
2248 }
2249
2250 static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos)
2251 {
2252         struct neigh_seq_state *state = seq->private;
2253         void *rc;
2254
2255         rc = neigh_get_idx(seq, pos);
2256         if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2257                 rc = pneigh_get_idx(seq, pos);
2258
2259         return rc;
2260 }
2261
2262 void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags)
2263 {
2264         struct neigh_seq_state *state = seq->private;
2265         loff_t pos_minus_one;
2266
2267         state->tbl = tbl;
2268         state->bucket = 0;
2269         state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH);
2270
2271         read_lock_bh(&tbl->lock);
2272
2273         pos_minus_one = *pos - 1;
2274         return *pos ? neigh_get_idx_any(seq, &pos_minus_one) : SEQ_START_TOKEN;
2275 }
2276 EXPORT_SYMBOL(neigh_seq_start);
2277
2278 void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2279 {
2280         struct neigh_seq_state *state;
2281         void *rc;
2282
2283         if (v == SEQ_START_TOKEN) {
2284                 rc = neigh_get_idx(seq, pos);
2285                 goto out;
2286         }
2287
2288         state = seq->private;
2289         if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) {
2290                 rc = neigh_get_next(seq, v, NULL);
2291                 if (rc)
2292                         goto out;
2293                 if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2294                         rc = pneigh_get_first(seq);
2295         } else {
2296                 BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY);
2297                 rc = pneigh_get_next(seq, v, NULL);
2298         }
2299 out:
2300         ++(*pos);
2301         return rc;
2302 }
2303 EXPORT_SYMBOL(neigh_seq_next);
2304
2305 void neigh_seq_stop(struct seq_file *seq, void *v)
2306 {
2307         struct neigh_seq_state *state = seq->private;
2308         struct neigh_table *tbl = state->tbl;
2309
2310         read_unlock_bh(&tbl->lock);
2311 }
2312 EXPORT_SYMBOL(neigh_seq_stop);
2313
2314 /* statistics via seq_file */
2315
2316 static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos)
2317 {
2318         struct proc_dir_entry *pde = seq->private;
2319         struct neigh_table *tbl = pde->data;
2320         int cpu;
2321
2322         if (*pos == 0)
2323                 return SEQ_START_TOKEN;
2324
2325         for (cpu = *pos-1; cpu < NR_CPUS; ++cpu) {
2326                 if (!cpu_possible(cpu))
2327                         continue;
2328                 *pos = cpu+1;
2329                 return per_cpu_ptr(tbl->stats, cpu);
2330         }
2331         return NULL;
2332 }
2333
2334 static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2335 {
2336         struct proc_dir_entry *pde = seq->private;
2337         struct neigh_table *tbl = pde->data;
2338         int cpu;
2339
2340         for (cpu = *pos; cpu < NR_CPUS; ++cpu) {
2341                 if (!cpu_possible(cpu))
2342                         continue;
2343                 *pos = cpu+1;
2344                 return per_cpu_ptr(tbl->stats, cpu);
2345         }
2346         return NULL;
2347 }
2348
2349 static void neigh_stat_seq_stop(struct seq_file *seq, void *v)
2350 {
2351
2352 }
2353
2354 static int neigh_stat_seq_show(struct seq_file *seq, void *v)
2355 {
2356         struct proc_dir_entry *pde = seq->private;
2357         struct neigh_table *tbl = pde->data;
2358         struct neigh_statistics *st = v;
2359
2360         if (v == SEQ_START_TOKEN) {
2361                 seq_printf(seq, "entries  allocs destroys hash_grows  lookups hits  res_failed  rcv_probes_mcast rcv_probes_ucast  periodic_gc_runs forced_gc_runs\n");
2362                 return 0;
2363         }
2364
2365         seq_printf(seq, "%08x  %08lx %08lx %08lx  %08lx %08lx  %08lx  "
2366                         "%08lx %08lx  %08lx %08lx\n",
2367                    atomic_read(&tbl->entries),
2368
2369                    st->allocs,
2370                    st->destroys,
2371                    st->hash_grows,
2372
2373                    st->lookups,
2374                    st->hits,
2375
2376                    st->res_failed,
2377
2378                    st->rcv_probes_mcast,
2379                    st->rcv_probes_ucast,
2380
2381                    st->periodic_gc_runs,
2382                    st->forced_gc_runs
2383                    );
2384
2385         return 0;
2386 }
2387
2388 static struct seq_operations neigh_stat_seq_ops = {
2389         .start  = neigh_stat_seq_start,
2390         .next   = neigh_stat_seq_next,
2391         .stop   = neigh_stat_seq_stop,
2392         .show   = neigh_stat_seq_show,
2393 };
2394
2395 static int neigh_stat_seq_open(struct inode *inode, struct file *file)
2396 {
2397         int ret = seq_open(file, &neigh_stat_seq_ops);
2398
2399         if (!ret) {
2400                 struct seq_file *sf = file->private_data;
2401                 sf->private = PDE(inode);
2402         }
2403         return ret;
2404 };
2405
2406 static struct file_operations neigh_stat_seq_fops = {
2407         .owner   = THIS_MODULE,
2408         .open    = neigh_stat_seq_open,
2409         .read    = seq_read,
2410         .llseek  = seq_lseek,
2411         .release = seq_release,
2412 };
2413
2414 #endif /* CONFIG_PROC_FS */
2415
2416 #ifdef CONFIG_ARPD
2417 static inline size_t neigh_nlmsg_size(void)
2418 {
2419         return NLMSG_ALIGN(sizeof(struct ndmsg))
2420                + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2421                + nla_total_size(MAX_ADDR_LEN) /* NDA_LLADDR */
2422                + nla_total_size(sizeof(struct nda_cacheinfo))
2423                + nla_total_size(4); /* NDA_PROBES */
2424 }
2425
2426 static void __neigh_notify(struct neighbour *n, int type, int flags)
2427 {
2428         struct sk_buff *skb;
2429         int err = -ENOBUFS;
2430
2431         skb = nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC);
2432         if (skb == NULL)
2433                 goto errout;
2434
2435         err = neigh_fill_info(skb, n, 0, 0, type, flags);
2436         if (err < 0) {
2437                 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2438                 WARN_ON(err == -EMSGSIZE);
2439                 kfree_skb(skb);
2440                 goto errout;
2441         }
2442         err = rtnl_notify(skb, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2443 errout:
2444         if (err < 0)
2445                 rtnl_set_sk_err(RTNLGRP_NEIGH, err);
2446 }
2447
2448 void neigh_app_ns(struct neighbour *n)
2449 {
2450         __neigh_notify(n, RTM_GETNEIGH, NLM_F_REQUEST);
2451 }
2452
2453 static void neigh_app_notify(struct neighbour *n)
2454 {
2455         __neigh_notify(n, RTM_NEWNEIGH, 0);
2456 }
2457
2458 #endif /* CONFIG_ARPD */
2459
2460 #ifdef CONFIG_SYSCTL
2461
2462 static struct neigh_sysctl_table {
2463         struct ctl_table_header *sysctl_header;
2464         ctl_table               neigh_vars[__NET_NEIGH_MAX];
2465         ctl_table               neigh_dev[2];
2466         ctl_table               neigh_neigh_dir[2];
2467         ctl_table               neigh_proto_dir[2];
2468         ctl_table               neigh_root_dir[2];
2469 } neigh_sysctl_template __read_mostly = {
2470         .neigh_vars = {
2471                 {
2472                         .ctl_name       = NET_NEIGH_MCAST_SOLICIT,
2473                         .procname       = "mcast_solicit",
2474                         .maxlen         = sizeof(int),
2475                         .mode           = 0644,
2476                         .proc_handler   = &proc_dointvec,
2477                 },
2478                 {
2479                         .ctl_name       = NET_NEIGH_UCAST_SOLICIT,
2480                         .procname       = "ucast_solicit",
2481                         .maxlen         = sizeof(int),
2482                         .mode           = 0644,
2483                         .proc_handler   = &proc_dointvec,
2484                 },
2485                 {
2486                         .ctl_name       = NET_NEIGH_APP_SOLICIT,
2487                         .procname       = "app_solicit",
2488                         .maxlen         = sizeof(int),
2489                         .mode           = 0644,
2490                         .proc_handler   = &proc_dointvec,
2491                 },
2492                 {
2493                         .ctl_name       = NET_NEIGH_RETRANS_TIME,
2494                         .procname       = "retrans_time",
2495                         .maxlen         = sizeof(int),
2496                         .mode           = 0644,
2497                         .proc_handler   = &proc_dointvec_userhz_jiffies,
2498                 },
2499                 {
2500                         .ctl_name       = NET_NEIGH_REACHABLE_TIME,
2501                         .procname       = "base_reachable_time",
2502                         .maxlen         = sizeof(int),
2503                         .mode           = 0644,
2504                         .proc_handler   = &proc_dointvec_jiffies,
2505                         .strategy       = &sysctl_jiffies,
2506                 },
2507                 {
2508                         .ctl_name       = NET_NEIGH_DELAY_PROBE_TIME,
2509                         .procname       = "delay_first_probe_time",
2510                         .maxlen         = sizeof(int),
2511                         .mode           = 0644,
2512                         .proc_handler   = &proc_dointvec_jiffies,
2513                         .strategy       = &sysctl_jiffies,
2514                 },
2515                 {
2516                         .ctl_name       = NET_NEIGH_GC_STALE_TIME,
2517                         .procname       = "gc_stale_time",
2518                         .maxlen         = sizeof(int),
2519                         .mode           = 0644,
2520                         .proc_handler   = &proc_dointvec_jiffies,
2521                         .strategy       = &sysctl_jiffies,
2522                 },
2523                 {
2524                         .ctl_name       = NET_NEIGH_UNRES_QLEN,
2525                         .procname       = "unres_qlen",
2526                         .maxlen         = sizeof(int),
2527                         .mode           = 0644,
2528                         .proc_handler   = &proc_dointvec,
2529                 },
2530                 {
2531                         .ctl_name       = NET_NEIGH_PROXY_QLEN,
2532                         .procname       = "proxy_qlen",
2533                         .maxlen         = sizeof(int),
2534                         .mode           = 0644,
2535                         .proc_handler   = &proc_dointvec,
2536                 },
2537                 {
2538                         .ctl_name       = NET_NEIGH_ANYCAST_DELAY,
2539                         .procname       = "anycast_delay",
2540                         .maxlen         = sizeof(int),
2541                         .mode           = 0644,
2542                         .proc_handler   = &proc_dointvec_userhz_jiffies,
2543                 },
2544                 {
2545                         .ctl_name       = NET_NEIGH_PROXY_DELAY,
2546                         .procname       = "proxy_delay",
2547                         .maxlen         = sizeof(int),
2548                         .mode           = 0644,
2549                         .proc_handler   = &proc_dointvec_userhz_jiffies,
2550                 },
2551                 {
2552                         .ctl_name       = NET_NEIGH_LOCKTIME,
2553                         .procname       = "locktime",
2554                         .maxlen         = sizeof(int),
2555                         .mode           = 0644,
2556                         .proc_handler   = &proc_dointvec_userhz_jiffies,
2557                 },
2558                 {
2559                         .ctl_name       = NET_NEIGH_GC_INTERVAL,
2560                         .procname       = "gc_interval",
2561                         .maxlen         = sizeof(int),
2562                         .mode           = 0644,
2563                         .proc_handler   = &proc_dointvec_jiffies,
2564                         .strategy       = &sysctl_jiffies,
2565                 },
2566                 {
2567                         .ctl_name       = NET_NEIGH_GC_THRESH1,
2568                         .procname       = "gc_thresh1",
2569                         .maxlen         = sizeof(int),
2570                         .mode           = 0644,
2571                         .proc_handler   = &proc_dointvec,
2572                 },
2573                 {
2574                         .ctl_name       = NET_NEIGH_GC_THRESH2,
2575                         .procname       = "gc_thresh2",
2576                         .maxlen         = sizeof(int),
2577                         .mode           = 0644,
2578                         .proc_handler   = &proc_dointvec,
2579                 },
2580                 {
2581                         .ctl_name       = NET_NEIGH_GC_THRESH3,
2582                         .procname       = "gc_thresh3",
2583                         .maxlen         = sizeof(int),
2584                         .mode           = 0644,
2585                         .proc_handler   = &proc_dointvec,
2586                 },
2587                 {
2588                         .ctl_name       = NET_NEIGH_RETRANS_TIME_MS,
2589                         .procname       = "retrans_time_ms",
2590                         .maxlen         = sizeof(int),
2591                         .mode           = 0644,
2592                         .proc_handler   = &proc_dointvec_ms_jiffies,
2593                         .strategy       = &sysctl_ms_jiffies,
2594                 },
2595                 {
2596                         .ctl_name       = NET_NEIGH_REACHABLE_TIME_MS,
2597                         .procname       = "base_reachable_time_ms",
2598                         .maxlen         = sizeof(int),
2599                         .mode           = 0644,
2600                         .proc_handler   = &proc_dointvec_ms_jiffies,
2601                         .strategy       = &sysctl_ms_jiffies,
2602                 },
2603         },
2604         .neigh_dev = {
2605                 {
2606                         .ctl_name       = NET_PROTO_CONF_DEFAULT,
2607                         .procname       = "default",
2608                         .mode           = 0555,
2609                 },
2610         },
2611         .neigh_neigh_dir = {
2612                 {
2613                         .procname       = "neigh",
2614                         .mode           = 0555,
2615                 },
2616         },
2617         .neigh_proto_dir = {
2618                 {
2619                         .mode           = 0555,
2620                 },
2621         },
2622         .neigh_root_dir = {
2623                 {
2624                         .ctl_name       = CTL_NET,
2625                         .procname       = "net",
2626                         .mode           = 0555,
2627                 },
2628         },
2629 };
2630
2631 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
2632                           int p_id, int pdev_id, char *p_name,
2633                           proc_handler *handler, ctl_handler *strategy)
2634 {
2635         struct neigh_sysctl_table *t = kmemdup(&neigh_sysctl_template,
2636                                                sizeof(*t), GFP_KERNEL);
2637         const char *dev_name_source = NULL;
2638         char *dev_name = NULL;
2639         int err = 0;
2640
2641         if (!t)
2642                 return -ENOBUFS;
2643         t->neigh_vars[0].data  = &p->mcast_probes;
2644         t->neigh_vars[1].data  = &p->ucast_probes;
2645         t->neigh_vars[2].data  = &p->app_probes;
2646         t->neigh_vars[3].data  = &p->retrans_time;
2647         t->neigh_vars[4].data  = &p->base_reachable_time;
2648         t->neigh_vars[5].data  = &p->delay_probe_time;
2649         t->neigh_vars[6].data  = &p->gc_staletime;
2650         t->neigh_vars[7].data  = &p->queue_len;
2651         t->neigh_vars[8].data  = &p->proxy_qlen;
2652         t->neigh_vars[9].data  = &p->anycast_delay;
2653         t->neigh_vars[10].data = &p->proxy_delay;
2654         t->neigh_vars[11].data = &p->locktime;
2655
2656         if (dev) {
2657                 dev_name_source = dev->name;
2658                 t->neigh_dev[0].ctl_name = dev->ifindex;
2659                 t->neigh_vars[12].procname = NULL;
2660                 t->neigh_vars[13].procname = NULL;
2661                 t->neigh_vars[14].procname = NULL;
2662                 t->neigh_vars[15].procname = NULL;
2663         } else {
2664                 dev_name_source = t->neigh_dev[0].procname;
2665                 t->neigh_vars[12].data = (int *)(p + 1);
2666                 t->neigh_vars[13].data = (int *)(p + 1) + 1;
2667                 t->neigh_vars[14].data = (int *)(p + 1) + 2;
2668                 t->neigh_vars[15].data = (int *)(p + 1) + 3;
2669         }
2670
2671         t->neigh_vars[16].data  = &p->retrans_time;
2672         t->neigh_vars[17].data  = &p->base_reachable_time;
2673
2674         if (handler || strategy) {
2675                 /* RetransTime */
2676                 t->neigh_vars[3].proc_handler = handler;
2677                 t->neigh_vars[3].strategy = strategy;
2678                 t->neigh_vars[3].extra1 = dev;
2679                 /* ReachableTime */
2680                 t->neigh_vars[4].proc_handler = handler;
2681                 t->neigh_vars[4].strategy = strategy;
2682                 t->neigh_vars[4].extra1 = dev;
2683                 /* RetransTime (in milliseconds)*/
2684                 t->neigh_vars[16].proc_handler = handler;
2685                 t->neigh_vars[16].strategy = strategy;
2686                 t->neigh_vars[16].extra1 = dev;
2687                 /* ReachableTime (in milliseconds) */
2688                 t->neigh_vars[17].proc_handler = handler;
2689                 t->neigh_vars[17].strategy = strategy;
2690                 t->neigh_vars[17].extra1 = dev;
2691         }
2692
2693         dev_name = kstrdup(dev_name_source, GFP_KERNEL);
2694         if (!dev_name) {
2695                 err = -ENOBUFS;
2696                 goto free;
2697         }
2698
2699         t->neigh_dev[0].procname = dev_name;
2700
2701         t->neigh_neigh_dir[0].ctl_name = pdev_id;
2702
2703         t->neigh_proto_dir[0].procname = p_name;
2704         t->neigh_proto_dir[0].ctl_name = p_id;
2705
2706         t->neigh_dev[0].child          = t->neigh_vars;
2707         t->neigh_neigh_dir[0].child    = t->neigh_dev;
2708         t->neigh_proto_dir[0].child    = t->neigh_neigh_dir;
2709         t->neigh_root_dir[0].child     = t->neigh_proto_dir;
2710
2711         t->sysctl_header = register_sysctl_table(t->neigh_root_dir, 0);
2712         if (!t->sysctl_header) {
2713                 err = -ENOBUFS;
2714                 goto free_procname;
2715         }
2716         p->sysctl_table = t;
2717         return 0;
2718
2719         /* error path */
2720  free_procname:
2721         kfree(dev_name);
2722  free:
2723         kfree(t);
2724
2725         return err;
2726 }
2727
2728 void neigh_sysctl_unregister(struct neigh_parms *p)
2729 {
2730         if (p->sysctl_table) {
2731                 struct neigh_sysctl_table *t = p->sysctl_table;
2732                 p->sysctl_table = NULL;
2733                 unregister_sysctl_table(t->sysctl_header);
2734                 kfree(t->neigh_dev[0].procname);
2735                 kfree(t);
2736         }
2737 }
2738
2739 #endif  /* CONFIG_SYSCTL */
2740
2741 EXPORT_SYMBOL(__neigh_event_send);
2742 EXPORT_SYMBOL(neigh_changeaddr);
2743 EXPORT_SYMBOL(neigh_compat_output);
2744 EXPORT_SYMBOL(neigh_connected_output);
2745 EXPORT_SYMBOL(neigh_create);
2746 EXPORT_SYMBOL(neigh_delete);
2747 EXPORT_SYMBOL(neigh_destroy);
2748 EXPORT_SYMBOL(neigh_dump_info);
2749 EXPORT_SYMBOL(neigh_event_ns);
2750 EXPORT_SYMBOL(neigh_ifdown);
2751 EXPORT_SYMBOL(neigh_lookup);
2752 EXPORT_SYMBOL(neigh_lookup_nodev);
2753 EXPORT_SYMBOL(neigh_parms_alloc);
2754 EXPORT_SYMBOL(neigh_parms_release);
2755 EXPORT_SYMBOL(neigh_rand_reach_time);
2756 EXPORT_SYMBOL(neigh_resolve_output);
2757 EXPORT_SYMBOL(neigh_table_clear);
2758 EXPORT_SYMBOL(neigh_table_init);
2759 EXPORT_SYMBOL(neigh_table_init_no_netlink);
2760 EXPORT_SYMBOL(neigh_update);
2761 EXPORT_SYMBOL(pneigh_enqueue);
2762 EXPORT_SYMBOL(pneigh_lookup);
2763
2764 #ifdef CONFIG_ARPD
2765 EXPORT_SYMBOL(neigh_app_ns);
2766 #endif
2767 #ifdef CONFIG_SYSCTL
2768 EXPORT_SYMBOL(neigh_sysctl_register);
2769 EXPORT_SYMBOL(neigh_sysctl_unregister);
2770 #endif