SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/ipv4/fib_hash.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              IPv4 FIB: lookup engine and maintenance routines.
   7  *
   8  * Version:     $Id: fib_hash.c,v 1.13 2001/10/31 21:55:54 davem Exp $
   9  *
  10  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  11  *
  12  *              This program is free software; you can redistribute it and/or
  13  *              modify it under the terms of the GNU General Public License
  14  *              as published by the Free Software Foundation; either version
  15  *              2 of the License, or (at your option) any later version.
  16  */
  17
  18 #include <asm/uaccess.h>
  19 #include <asm/system.h>
  20 #include <linux/bitops.h>
  21 #include <linux/types.h>
  22 #include <linux/kernel.h>
  23 #include <linux/mm.h>
  24 #include <linux/string.h>
  25 #include <linux/socket.h>
  26 #include <linux/sockios.h>
  27 #include <linux/errno.h>
  28 #include <linux/in.h>
  29 #include <linux/inet.h>
  30 #include <linux/inetdevice.h>
  31 #include <linux/netdevice.h>
  32 #include <linux/if_arp.h>
  33 #include <linux/proc_fs.h>
  34 #include <linux/skbuff.h>
  35 #include <linux/netlink.h>
  36 #include <linux/init.h>
  37
  38 #include <net/net_namespace.h>
  39 #include <net/ip.h>
  40 #include <net/protocol.h>
  41 #include <net/route.h>
  42 #include <net/tcp.h>
  43 #include <net/sock.h>
  44 #include <net/ip_fib.h>
  45
  46 #include "fib_lookup.h"
  47
  48 static struct kmem_cache *fn_hash_kmem __read_mostly;
  49 static struct kmem_cache *fn_alias_kmem __read_mostly;
  50
  51 struct fib_node {
  52         struct hlist_node       fn_hash;
  53         struct list_head        fn_alias;
  54         __be32                  fn_key;
  55 };
  56
  57 struct fn_zone {
  58         struct fn_zone          *fz_next;       /* Next not empty zone  */
  59         struct hlist_head       *fz_hash;       /* Hash table pointer   */
  60         int                     fz_nent;        /* Number of entries    */
  61
  62         int                     fz_divisor;     /* Hash divisor         */
  63         u32                     fz_hashmask;    /* (fz_divisor - 1)     */
  64 #define FZ_HASHMASK(fz)         ((fz)->fz_hashmask)
  65
  66         int                     fz_order;       /* Zone order           */
  67         __be32                  fz_mask;
  68 #define FZ_MASK(fz)             ((fz)->fz_mask)
  69 };
  70
  71 /* NOTE. On fast computers evaluation of fz_hashmask and fz_mask
  72  * can be cheaper than memory lookup, so that FZ_* macros are used.
  73  */
  74
  75 struct fn_hash {
  76         struct fn_zone  *fn_zones[33];
  77         struct fn_zone  *fn_zone_list;
  78 };
  79
  80 static inline u32 fn_hash(__be32 key, struct fn_zone *fz)
  81 {
  82         u32 h = ntohl(key)>>(32 - fz->fz_order);
  83         h ^= (h>>20);
  84         h ^= (h>>10);
  85         h ^= (h>>5);
  86         h &= FZ_HASHMASK(fz);
  87         return h;
  88 }
  89
  90 static inline __be32 fz_key(__be32 dst, struct fn_zone *fz)
  91 {
  92         return dst & FZ_MASK(fz);
  93 }
  94
  95 static DEFINE_RWLOCK(fib_hash_lock);
  96 static unsigned int fib_hash_genid;
  97
  98 #define FZ_MAX_DIVISOR ((PAGE_SIZE<<MAX_ORDER) / sizeof(struct hlist_head))
  99
 100 static struct hlist_head *fz_hash_alloc(int divisor)
 101 {
 102         unsigned long size = divisor * sizeof(struct hlist_head);
 103
 104         if (size <= PAGE_SIZE) {
 105                 return kmalloc(size, GFP_KERNEL);
 106         } else {
 107                 return (struct hlist_head *)
 108                         __get_free_pages(GFP_KERNEL, get_order(size));
 109         }
 110 }
 111
 112 /* The fib hash lock must be held when this is called. */
 113 static inline void fn_rebuild_zone(struct fn_zone *fz,
 114                                    struct hlist_head *old_ht,
 115                                    int old_divisor)
 116 {
 117         int i;
 118
 119         for (i = 0; i < old_divisor; i++) {
 120                 struct hlist_node *node, *n;
 121                 struct fib_node *f;
 122
 123                 hlist_for_each_entry_safe(f, node, n, &old_ht[i], fn_hash) {
 124                         struct hlist_head *new_head;
 125
 126                         hlist_del(&f->fn_hash);
 127
 128                         new_head = &fz->fz_hash[fn_hash(f->fn_key, fz)];
 129                         hlist_add_head(&f->fn_hash, new_head);
 130                 }
 131         }
 132 }
 133
 134 static void fz_hash_free(struct hlist_head *hash, int divisor)
 135 {
 136         unsigned long size = divisor * sizeof(struct hlist_head);
 137
 138         if (size <= PAGE_SIZE)
 139                 kfree(hash);
 140         else
 141                 free_pages((unsigned long)hash, get_order(size));
 142 }
 143
 144 static void fn_rehash_zone(struct fn_zone *fz)
 145 {
 146         struct hlist_head *ht, *old_ht;
 147         int old_divisor, new_divisor;
 148         u32 new_hashmask;
 149
 150         old_divisor = fz->fz_divisor;
 151
 152         switch (old_divisor) {
 153         case 16:
 154                 new_divisor = 256;
 155                 break;
 156         case 256:
 157                 new_divisor = 1024;
 158                 break;
 159         default:
 160                 if ((old_divisor << 1) > FZ_MAX_DIVISOR) {
 161                         printk(KERN_CRIT "route.c: bad divisor %d!\n", old_divisor);
 162                         return;
 163                 }
 164                 new_divisor = (old_divisor << 1);
 165                 break;
 166         }
 167
 168         new_hashmask = (new_divisor - 1);
 169
 170 #if RT_CACHE_DEBUG >= 2
 171         printk("fn_rehash_zone: hash for zone %d grows from %d\n", fz->fz_order, old_divisor);
 172 #endif
 173
 174         ht = fz_hash_alloc(new_divisor);
 175
 176         if (ht) {
 177                 memset(ht, 0, new_divisor * sizeof(struct hlist_head));
 178
 179                 write_lock_bh(&fib_hash_lock);
 180                 old_ht = fz->fz_hash;
 181                 fz->fz_hash = ht;
 182                 fz->fz_hashmask = new_hashmask;
 183                 fz->fz_divisor = new_divisor;
 184                 fn_rebuild_zone(fz, old_ht, old_divisor);
 185                 fib_hash_genid++;
 186                 write_unlock_bh(&fib_hash_lock);
 187
 188                 fz_hash_free(old_ht, old_divisor);
 189         }
 190 }
 191
 192 static inline void fn_free_node(struct fib_node * f)
 193 {
 194         kmem_cache_free(fn_hash_kmem, f);
 195 }
 196
 197 static inline void fn_free_alias(struct fib_alias *fa)
 198 {
 199         fib_release_info(fa->fa_info);
 200         kmem_cache_free(fn_alias_kmem, fa);
 201 }
 202
 203 static struct fn_zone *
 204 fn_new_zone(struct fn_hash *table, int z)
 205 {
 206         int i;
 207         struct fn_zone *fz = kzalloc(sizeof(struct fn_zone), GFP_KERNEL);
 208         if (!fz)
 209                 return NULL;
 210
 211         if (z) {
 212                 fz->fz_divisor = 16;
 213         } else {
 214                 fz->fz_divisor = 1;
 215         }
 216         fz->fz_hashmask = (fz->fz_divisor - 1);
 217         fz->fz_hash = fz_hash_alloc(fz->fz_divisor);
 218         if (!fz->fz_hash) {
 219                 kfree(fz);
 220                 return NULL;
 221         }
 222         memset(fz->fz_hash, 0, fz->fz_divisor * sizeof(struct hlist_head *));
 223         fz->fz_order = z;
 224         fz->fz_mask = inet_make_mask(z);
 225
 226         /* Find the first not empty zone with more specific mask */
 227         for (i=z+1; i<=32; i++)
 228                 if (table->fn_zones[i])
 229                         break;
 230         write_lock_bh(&fib_hash_lock);
 231         if (i>32) {
 232                 /* No more specific masks, we are the first. */
 233                 fz->fz_next = table->fn_zone_list;
 234                 table->fn_zone_list = fz;
 235         } else {
 236                 fz->fz_next = table->fn_zones[i]->fz_next;
 237                 table->fn_zones[i]->fz_next = fz;
 238         }
 239         table->fn_zones[z] = fz;
 240         fib_hash_genid++;
 241         write_unlock_bh(&fib_hash_lock);
 242         return fz;
 243 }
 244
 245 static int
 246 fn_hash_lookup(struct fib_table *tb, const struct flowi *flp, struct fib_result *res)
 247 {
 248         int err;
 249         struct fn_zone *fz;
 250         struct fn_hash *t = (struct fn_hash*)tb->tb_data;
 251
 252         read_lock(&fib_hash_lock);
 253         for (fz = t->fn_zone_list; fz; fz = fz->fz_next) {
 254                 struct hlist_head *head;
 255                 struct hlist_node *node;
 256                 struct fib_node *f;
 257                 __be32 k = fz_key(flp->fl4_dst, fz);
 258
 259                 head = &fz->fz_hash[fn_hash(k, fz)];
 260                 hlist_for_each_entry(f, node, head, fn_hash) {
 261                         if (f->fn_key != k)
 262                                 continue;
 263
 264                         err = fib_semantic_match(&f->fn_alias,
 265                                                  flp, res,
 266                                                  f->fn_key, fz->fz_mask,
 267                                                  fz->fz_order);
 268                         if (err <= 0)
 269                                 goto out;
 270                 }
 271         }
 272         err = 1;
 273 out:
 274         read_unlock(&fib_hash_lock);
 275         return err;
 276 }
 277
 278 static int fn_hash_last_dflt=-1;
 279
 280 static void
 281 fn_hash_select_default(struct fib_table *tb, const struct flowi *flp, struct fib_result *res)
 282 {
 283         int order, last_idx;
 284         struct hlist_node *node;
 285         struct fib_node *f;
 286         struct fib_info *fi = NULL;
 287         struct fib_info *last_resort;
 288         struct fn_hash *t = (struct fn_hash*)tb->tb_data;
 289         struct fn_zone *fz = t->fn_zones[0];
 290
 291         if (fz == NULL)
 292                 return;
 293
 294         last_idx = -1;
 295         last_resort = NULL;
 296         order = -1;
 297
 298         read_lock(&fib_hash_lock);
 299         hlist_for_each_entry(f, node, &fz->fz_hash[0], fn_hash) {
 300                 struct fib_alias *fa;
 301
 302                 list_for_each_entry(fa, &f->fn_alias, fa_list) {
 303                         struct fib_info *next_fi = fa->fa_info;
 304
 305                         if (fa->fa_scope != res->scope ||
 306                             fa->fa_type != RTN_UNICAST)
 307                                 continue;
 308
 309                         if (next_fi->fib_priority > res->fi->fib_priority)
 310                                 break;
 311                         if (!next_fi->fib_nh[0].nh_gw ||
 312                             next_fi->fib_nh[0].nh_scope != RT_SCOPE_LINK)
 313                                 continue;
 314                         fa->fa_state |= FA_S_ACCESSED;
 315
 316                         if (fi == NULL) {
 317                                 if (next_fi != res->fi)
 318                                         break;
 319                         } else if (!fib_detect_death(fi, order, &last_resort,
 320                                                      &last_idx, &fn_hash_last_dflt)) {
 321                                 if (res->fi)
 322                                         fib_info_put(res->fi);
 323                                 res->fi = fi;
 324                                 atomic_inc(&fi->fib_clntref);
 325                                 fn_hash_last_dflt = order;
 326                                 goto out;
 327                         }
 328                         fi = next_fi;
 329                         order++;
 330                 }
 331         }
 332
 333         if (order <= 0 || fi == NULL) {
 334                 fn_hash_last_dflt = -1;
 335                 goto out;
 336         }
 337
 338         if (!fib_detect_death(fi, order, &last_resort, &last_idx, &fn_hash_last_dflt)) {
 339                 if (res->fi)
 340                         fib_info_put(res->fi);
 341                 res->fi = fi;
 342                 atomic_inc(&fi->fib_clntref);
 343                 fn_hash_last_dflt = order;
 344                 goto out;
 345         }
 346
 347         if (last_idx >= 0) {
 348                 if (res->fi)
 349                         fib_info_put(res->fi);
 350                 res->fi = last_resort;
 351                 if (last_resort)
 352                         atomic_inc(&last_resort->fib_clntref);
 353         }
 354         fn_hash_last_dflt = last_idx;
 355 out:
 356         read_unlock(&fib_hash_lock);
 357 }
 358
 359 /* Insert node F to FZ. */
 360 static inline void fib_insert_node(struct fn_zone *fz, struct fib_node *f)
 361 {
 362         struct hlist_head *head = &fz->fz_hash[fn_hash(f->fn_key, fz)];
 363
 364         hlist_add_head(&f->fn_hash, head);
 365 }
 366
 367 /* Return the node in FZ matching KEY. */
 368 static struct fib_node *fib_find_node(struct fn_zone *fz, __be32 key)
 369 {
 370         struct hlist_head *head = &fz->fz_hash[fn_hash(key, fz)];
 371         struct hlist_node *node;
 372         struct fib_node *f;
 373
 374         hlist_for_each_entry(f, node, head, fn_hash) {
 375                 if (f->fn_key == key)
 376                         return f;
 377         }
 378
 379         return NULL;
 380 }
 381
 382 static int fn_hash_insert(struct fib_table *tb, struct fib_config *cfg)
 383 {
 384         struct fn_hash *table = (struct fn_hash *) tb->tb_data;
 385         struct fib_node *new_f, *f;
 386         struct fib_alias *fa, *new_fa;
 387         struct fn_zone *fz;
 388         struct fib_info *fi;
 389         u8 tos = cfg->fc_tos;
 390         __be32 key;
 391         int err;
 392
 393         if (cfg->fc_dst_len > 32)
 394                 return -EINVAL;
 395
 396         fz = table->fn_zones[cfg->fc_dst_len];
 397         if (!fz && !(fz = fn_new_zone(table, cfg->fc_dst_len)))
 398                 return -ENOBUFS;
 399
 400         key = 0;
 401         if (cfg->fc_dst) {
 402                 if (cfg->fc_dst & ~FZ_MASK(fz))
 403                         return -EINVAL;
 404                 key = fz_key(cfg->fc_dst, fz);
 405         }
 406
 407         fi = fib_create_info(cfg);
 408         if (IS_ERR(fi))
 409                 return PTR_ERR(fi);
 410
 411         if (fz->fz_nent > (fz->fz_divisor<<1) &&
 412             fz->fz_divisor < FZ_MAX_DIVISOR &&
 413             (cfg->fc_dst_len == 32 ||
 414              (1 << cfg->fc_dst_len) > fz->fz_divisor))
 415                 fn_rehash_zone(fz);
 416
 417         f = fib_find_node(fz, key);
 418
 419         if (!f)
 420                 fa = NULL;
 421         else
 422                 fa = fib_find_alias(&f->fn_alias, tos, fi->fib_priority);
 423
 424         /* Now fa, if non-NULL, points to the first fib alias
 425          * with the same keys [prefix,tos,priority], if such key already
 426          * exists or to the node before which we will insert new one.
 427          *
 428          * If fa is NULL, we will need to allocate a new one and
 429          * insert to the head of f.
 430          *
 431          * If f is NULL, no fib node matched the destination key
 432          * and we need to allocate a new one of those as well.
 433          */
 434
 435         if (fa && fa->fa_tos == tos &&
 436             fa->fa_info->fib_priority == fi->fib_priority) {
 437                 struct fib_alias *fa_orig;
 438
 439                 err = -EEXIST;
 440                 if (cfg->fc_nlflags & NLM_F_EXCL)
 441                         goto out;
 442
 443                 if (cfg->fc_nlflags & NLM_F_REPLACE) {
 444                         struct fib_info *fi_drop;
 445                         u8 state;
 446
 447                         if (fi->fib_treeref > 1)
 448                                 goto out;
 449
 450                         write_lock_bh(&fib_hash_lock);
 451                         fi_drop = fa->fa_info;
 452                         fa->fa_info = fi;
 453                         fa->fa_type = cfg->fc_type;
 454                         fa->fa_scope = cfg->fc_scope;
 455                         state = fa->fa_state;
 456                         fa->fa_state &= ~FA_S_ACCESSED;
 457                         fib_hash_genid++;
 458                         write_unlock_bh(&fib_hash_lock);
 459
 460                         fib_release_info(fi_drop);
 461                         if (state & FA_S_ACCESSED)
 462                                 rt_cache_flush(-1);
 463                         rtmsg_fib(RTM_NEWROUTE, key, fa, cfg->fc_dst_len, tb->tb_id,
 464                                   &cfg->fc_nlinfo, NLM_F_REPLACE);
 465                         return 0;
 466                 }
 467
 468                 /* Error if we find a perfect match which
 469                  * uses the same scope, type, and nexthop
 470                  * information.
 471                  */
 472                 fa_orig = fa;
 473                 fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
 474                 list_for_each_entry_continue(fa, &f->fn_alias, fa_list) {
 475                         if (fa->fa_tos != tos)
 476                                 break;
 477                         if (fa->fa_info->fib_priority != fi->fib_priority)
 478                                 break;
 479                         if (fa->fa_type == cfg->fc_type &&
 480                             fa->fa_scope == cfg->fc_scope &&
 481                             fa->fa_info == fi)
 482                                 goto out;
 483                 }
 484                 if (!(cfg->fc_nlflags & NLM_F_APPEND))
 485                         fa = fa_orig;
 486         }
 487
 488         err = -ENOENT;
 489         if (!(cfg->fc_nlflags & NLM_F_CREATE))
 490                 goto out;
 491
 492         err = -ENOBUFS;
 493         new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
 494         if (new_fa == NULL)
 495                 goto out;
 496
 497         new_f = NULL;
 498         if (!f) {
 499                 new_f = kmem_cache_alloc(fn_hash_kmem, GFP_KERNEL);
 500                 if (new_f == NULL)
 501                         goto out_free_new_fa;
 502
 503                 INIT_HLIST_NODE(&new_f->fn_hash);
 504                 INIT_LIST_HEAD(&new_f->fn_alias);
 505                 new_f->fn_key = key;
 506                 f = new_f;
 507         }
 508
 509         new_fa->fa_info = fi;
 510         new_fa->fa_tos = tos;
 511         new_fa->fa_type = cfg->fc_type;
 512         new_fa->fa_scope = cfg->fc_scope;
 513         new_fa->fa_state = 0;
 514
 515         /*
 516          * Insert new entry to the list.
 517          */
 518
 519         write_lock_bh(&fib_hash_lock);
 520         if (new_f)
 521                 fib_insert_node(fz, new_f);
 522         list_add_tail(&new_fa->fa_list,
 523                  (fa ? &fa->fa_list : &f->fn_alias));
 524         fib_hash_genid++;
 525         write_unlock_bh(&fib_hash_lock);
 526
 527         if (new_f)
 528                 fz->fz_nent++;
 529         rt_cache_flush(-1);
 530
 531         rtmsg_fib(RTM_NEWROUTE, key, new_fa, cfg->fc_dst_len, tb->tb_id,
 532                   &cfg->fc_nlinfo, 0);
 533         return 0;
 534
 535 out_free_new_fa:
 536         kmem_cache_free(fn_alias_kmem, new_fa);
 537 out:
 538         fib_release_info(fi);
 539         return err;
 540 }
 541
 542
 543 static int fn_hash_delete(struct fib_table *tb, struct fib_config *cfg)
 544 {
 545         struct fn_hash *table = (struct fn_hash*)tb->tb_data;
 546         struct fib_node *f;
 547         struct fib_alias *fa, *fa_to_delete;
 548         struct fn_zone *fz;
 549         __be32 key;
 550
 551         if (cfg->fc_dst_len > 32)
 552                 return -EINVAL;
 553
 554         if ((fz  = table->fn_zones[cfg->fc_dst_len]) == NULL)
 555                 return -ESRCH;
 556
 557         key = 0;
 558         if (cfg->fc_dst) {
 559                 if (cfg->fc_dst & ~FZ_MASK(fz))
 560                         return -EINVAL;
 561                 key = fz_key(cfg->fc_dst, fz);
 562         }
 563
 564         f = fib_find_node(fz, key);
 565
 566         if (!f)
 567                 fa = NULL;
 568         else
 569                 fa = fib_find_alias(&f->fn_alias, cfg->fc_tos, 0);
 570         if (!fa)
 571                 return -ESRCH;
 572
 573         fa_to_delete = NULL;
 574         fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
 575         list_for_each_entry_continue(fa, &f->fn_alias, fa_list) {
 576                 struct fib_info *fi = fa->fa_info;
 577
 578                 if (fa->fa_tos != cfg->fc_tos)
 579                         break;
 580
 581                 if ((!cfg->fc_type ||
 582                      fa->fa_type == cfg->fc_type) &&
 583                     (cfg->fc_scope == RT_SCOPE_NOWHERE ||
 584                      fa->fa_scope == cfg->fc_scope) &&
 585                     (!cfg->fc_protocol ||
 586                      fi->fib_protocol == cfg->fc_protocol) &&
 587                     fib_nh_match(cfg, fi) == 0) {
 588                         fa_to_delete = fa;
 589                         break;
 590                 }
 591         }
 592
 593         if (fa_to_delete) {
 594                 int kill_fn;
 595
 596                 fa = fa_to_delete;
 597                 rtmsg_fib(RTM_DELROUTE, key, fa, cfg->fc_dst_len,
 598                           tb->tb_id, &cfg->fc_nlinfo, 0);
 599
 600                 kill_fn = 0;
 601                 write_lock_bh(&fib_hash_lock);
 602                 list_del(&fa->fa_list);
 603                 if (list_empty(&f->fn_alias)) {
 604                         hlist_del(&f->fn_hash);
 605                         kill_fn = 1;
 606                 }
 607                 fib_hash_genid++;
 608                 write_unlock_bh(&fib_hash_lock);
 609
 610                 if (fa->fa_state & FA_S_ACCESSED)
 611                         rt_cache_flush(-1);
 612                 fn_free_alias(fa);
 613                 if (kill_fn) {
 614                         fn_free_node(f);
 615                         fz->fz_nent--;
 616                 }
 617
 618                 return 0;
 619         }
 620         return -ESRCH;
 621 }
 622
 623 static int fn_flush_list(struct fn_zone *fz, int idx)
 624 {
 625         struct hlist_head *head = &fz->fz_hash[idx];
 626         struct hlist_node *node, *n;
 627         struct fib_node *f;
 628         int found = 0;
 629
 630         hlist_for_each_entry_safe(f, node, n, head, fn_hash) {
 631                 struct fib_alias *fa, *fa_node;
 632                 int kill_f;
 633
 634                 kill_f = 0;
 635                 list_for_each_entry_safe(fa, fa_node, &f->fn_alias, fa_list) {
 636                         struct fib_info *fi = fa->fa_info;
 637
 638                         if (fi && (fi->fib_flags&RTNH_F_DEAD)) {
 639                                 write_lock_bh(&fib_hash_lock);
 640                                 list_del(&fa->fa_list);
 641                                 if (list_empty(&f->fn_alias)) {
 642                                         hlist_del(&f->fn_hash);
 643                                         kill_f = 1;
 644                                 }
 645                                 fib_hash_genid++;
 646                                 write_unlock_bh(&fib_hash_lock);
 647
 648                                 fn_free_alias(fa);
 649                                 found++;
 650                         }
 651                 }
 652                 if (kill_f) {
 653                         fn_free_node(f);
 654                         fz->fz_nent--;
 655                 }
 656         }
 657         return found;
 658 }
 659
 660 static int fn_hash_flush(struct fib_table *tb)
 661 {
 662         struct fn_hash *table = (struct fn_hash *) tb->tb_data;
 663         struct fn_zone *fz;
 664         int found = 0;
 665
 666         for (fz = table->fn_zone_list; fz; fz = fz->fz_next) {
 667                 int i;
 668
 669                 for (i = fz->fz_divisor - 1; i >= 0; i--)
 670                         found += fn_flush_list(fz, i);
 671         }
 672         return found;
 673 }
 674
 675
 676 static inline int
 677 fn_hash_dump_bucket(struct sk_buff *skb, struct netlink_callback *cb,
 678                      struct fib_table *tb,
 679                      struct fn_zone *fz,
 680                      struct hlist_head *head)
 681 {
 682         struct hlist_node *node;
 683         struct fib_node *f;
 684         int i, s_i;
 685
 686         s_i = cb->args[4];
 687         i = 0;
 688         hlist_for_each_entry(f, node, head, fn_hash) {
 689                 struct fib_alias *fa;
 690
 691                 list_for_each_entry(fa, &f->fn_alias, fa_list) {
 692                         if (i < s_i)
 693                                 goto next;
 694
 695                         if (fib_dump_info(skb, NETLINK_CB(cb->skb).pid,
 696                                           cb->nlh->nlmsg_seq,
 697                                           RTM_NEWROUTE,
 698                                           tb->tb_id,
 699                                           fa->fa_type,
 700                                           fa->fa_scope,
 701                                           f->fn_key,
 702                                           fz->fz_order,
 703                                           fa->fa_tos,
 704                                           fa->fa_info,
 705                                           NLM_F_MULTI) < 0) {
 706                                 cb->args[4] = i;
 707                                 return -1;
 708                         }
 709                 next:
 710                         i++;
 711                 }
 712         }
 713         cb->args[4] = i;
 714         return skb->len;
 715 }
 716
 717 static inline int
 718 fn_hash_dump_zone(struct sk_buff *skb, struct netlink_callback *cb,
 719                    struct fib_table *tb,
 720                    struct fn_zone *fz)
 721 {
 722         int h, s_h;
 723
 724         s_h = cb->args[3];
 725         for (h=0; h < fz->fz_divisor; h++) {
 726                 if (h < s_h) continue;
 727                 if (h > s_h)
 728                         memset(&cb->args[4], 0,
 729                                sizeof(cb->args) - 4*sizeof(cb->args[0]));
 730                 if (fz->fz_hash == NULL ||
 731                     hlist_empty(&fz->fz_hash[h]))
 732                         continue;
 733                 if (fn_hash_dump_bucket(skb, cb, tb, fz, &fz->fz_hash[h])<0) {
 734                         cb->args[3] = h;
 735                         return -1;
 736                 }
 737         }
 738         cb->args[3] = h;
 739         return skb->len;
 740 }
 741
 742 static int fn_hash_dump(struct fib_table *tb, struct sk_buff *skb, struct netlink_callback *cb)
 743 {
 744         int m, s_m;
 745         struct fn_zone *fz;
 746         struct fn_hash *table = (struct fn_hash*)tb->tb_data;
 747
 748         s_m = cb->args[2];
 749         read_lock(&fib_hash_lock);
 750         for (fz = table->fn_zone_list, m=0; fz; fz = fz->fz_next, m++) {
 751                 if (m < s_m) continue;
 752                 if (m > s_m)
 753                         memset(&cb->args[3], 0,
 754                                sizeof(cb->args) - 3*sizeof(cb->args[0]));
 755                 if (fn_hash_dump_zone(skb, cb, tb, fz) < 0) {
 756                         cb->args[2] = m;
 757                         read_unlock(&fib_hash_lock);
 758                         return -1;
 759                 }
 760         }
 761         read_unlock(&fib_hash_lock);
 762         cb->args[2] = m;
 763         return skb->len;
 764 }
 765
 766 #ifdef CONFIG_IP_MULTIPLE_TABLES
 767 struct fib_table * fib_hash_init(u32 id)
 768 #else
 769 struct fib_table * __init fib_hash_init(u32 id)
 770 #endif
 771 {
 772         struct fib_table *tb;
 773
 774         if (fn_hash_kmem == NULL)
 775                 fn_hash_kmem = kmem_cache_create("ip_fib_hash",
 776                                                  sizeof(struct fib_node),
 777                                                  0, SLAB_HWCACHE_ALIGN,
 778                                                  NULL);
 779
 780         if (fn_alias_kmem == NULL)
 781                 fn_alias_kmem = kmem_cache_create("ip_fib_alias",
 782                                                   sizeof(struct fib_alias),
 783                                                   0, SLAB_HWCACHE_ALIGN,
 784                                                   NULL);
 785
 786         tb = kmalloc(sizeof(struct fib_table) + sizeof(struct fn_hash),
 787                      GFP_KERNEL);
 788         if (tb == NULL)
 789                 return NULL;
 790
 791         tb->tb_id = id;
 792         tb->tb_lookup = fn_hash_lookup;
 793         tb->tb_insert = fn_hash_insert;
 794         tb->tb_delete = fn_hash_delete;
 795         tb->tb_flush = fn_hash_flush;
 796         tb->tb_select_default = fn_hash_select_default;
 797         tb->tb_dump = fn_hash_dump;
 798         memset(tb->tb_data, 0, sizeof(struct fn_hash));
 799         return tb;
 800 }
 801
 802 /* ------------------------------------------------------------------------ */
 803 #ifdef CONFIG_PROC_FS
 804
 805 struct fib_iter_state {
 806         struct fn_zone  *zone;
 807         int             bucket;
 808         struct hlist_head *hash_head;
 809         struct fib_node *fn;
 810         struct fib_alias *fa;
 811         loff_t pos;
 812         unsigned int genid;
 813         int valid;
 814 };
 815
 816 static struct fib_alias *fib_get_first(struct seq_file *seq)
 817 {
 818         struct fib_iter_state *iter = seq->private;
 819         struct fn_hash *table = (struct fn_hash *) ip_fib_main_table->tb_data;
 820
 821         iter->bucket    = 0;
 822         iter->hash_head = NULL;
 823         iter->fn        = NULL;
 824         iter->fa        = NULL;
 825         iter->pos       = 0;
 826         iter->genid     = fib_hash_genid;
 827         iter->valid     = 1;
 828
 829         for (iter->zone = table->fn_zone_list; iter->zone;
 830              iter->zone = iter->zone->fz_next) {
 831                 int maxslot;
 832
 833                 if (!iter->zone->fz_nent)
 834                         continue;
 835
 836                 iter->hash_head = iter->zone->fz_hash;
 837                 maxslot = iter->zone->fz_divisor;
 838
 839                 for (iter->bucket = 0; iter->bucket < maxslot;
 840                      ++iter->bucket, ++iter->hash_head) {
 841                         struct hlist_node *node;
 842                         struct fib_node *fn;
 843
 844                         hlist_for_each_entry(fn,node,iter->hash_head,fn_hash) {
 845                                 struct fib_alias *fa;
 846
 847                                 list_for_each_entry(fa,&fn->fn_alias,fa_list) {
 848                                         iter->fn = fn;
 849                                         iter->fa = fa;
 850                                         goto out;
 851                                 }
 852                         }
 853                 }
 854         }
 855 out:
 856         return iter->fa;
 857 }
 858
 859 static struct fib_alias *fib_get_next(struct seq_file *seq)
 860 {
 861         struct fib_iter_state *iter = seq->private;
 862         struct fib_node *fn;
 863         struct fib_alias *fa;
 864
 865         /* Advance FA, if any. */
 866         fn = iter->fn;
 867         fa = iter->fa;
 868         if (fa) {
 869                 BUG_ON(!fn);
 870                 list_for_each_entry_continue(fa, &fn->fn_alias, fa_list) {
 871                         iter->fa = fa;
 872                         goto out;
 873                 }
 874         }
 875
 876         fa = iter->fa = NULL;
 877
 878         /* Advance FN. */
 879         if (fn) {
 880                 struct hlist_node *node = &fn->fn_hash;
 881                 hlist_for_each_entry_continue(fn, node, fn_hash) {
 882                         iter->fn = fn;
 883
 884                         list_for_each_entry(fa, &fn->fn_alias, fa_list) {
 885                                 iter->fa = fa;
 886                                 goto out;
 887                         }
 888                 }
 889         }
 890
 891         fn = iter->fn = NULL;
 892
 893         /* Advance hash chain. */
 894         if (!iter->zone)
 895                 goto out;
 896
 897         for (;;) {
 898                 struct hlist_node *node;
 899                 int maxslot;
 900
 901                 maxslot = iter->zone->fz_divisor;
 902
 903                 while (++iter->bucket < maxslot) {
 904                         iter->hash_head++;
 905
 906                         hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) {
 907                                 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
 908                                         iter->fn = fn;
 909                                         iter->fa = fa;
 910                                         goto out;
 911                                 }
 912                         }
 913                 }
 914
 915                 iter->zone = iter->zone->fz_next;
 916
 917                 if (!iter->zone)
 918                         goto out;
 919
 920                 iter->bucket = 0;
 921                 iter->hash_head = iter->zone->fz_hash;
 922
 923                 hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) {
 924                         list_for_each_entry(fa, &fn->fn_alias, fa_list) {
 925                                 iter->fn = fn;
 926                                 iter->fa = fa;
 927                                 goto out;
 928                         }
 929                 }
 930         }
 931 out:
 932         iter->pos++;
 933         return fa;
 934 }
 935
 936 static struct fib_alias *fib_get_idx(struct seq_file *seq, loff_t pos)
 937 {
 938         struct fib_iter_state *iter = seq->private;
 939         struct fib_alias *fa;
 940
 941         if (iter->valid && pos >= iter->pos && iter->genid == fib_hash_genid) {
 942                 fa   = iter->fa;
 943                 pos -= iter->pos;
 944         } else
 945                 fa = fib_get_first(seq);
 946
 947         if (fa)
 948                 while (pos && (fa = fib_get_next(seq)))
 949                         --pos;
 950         return pos ? NULL : fa;
 951 }
 952
 953 static void *fib_seq_start(struct seq_file *seq, loff_t *pos)
 954 {
 955         void *v = NULL;
 956
 957         read_lock(&fib_hash_lock);
 958         if (ip_fib_main_table)
 959                 v = *pos ? fib_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 960         return v;
 961 }
 962
 963 static void *fib_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 964 {
 965         ++*pos;
 966         return v == SEQ_START_TOKEN ? fib_get_first(seq) : fib_get_next(seq);
 967 }
 968
 969 static void fib_seq_stop(struct seq_file *seq, void *v)
 970 {
 971         read_unlock(&fib_hash_lock);
 972 }
 973
 974 static unsigned fib_flag_trans(int type, __be32 mask, struct fib_info *fi)
 975 {
 976         static const unsigned type2flags[RTN_MAX + 1] = {
 977                 [7] = RTF_REJECT, [8] = RTF_REJECT,
 978         };
 979         unsigned flags = type2flags[type];
 980
 981         if (fi && fi->fib_nh->nh_gw)
 982                 flags |= RTF_GATEWAY;
 983         if (mask == htonl(0xFFFFFFFF))
 984                 flags |= RTF_HOST;
 985         flags |= RTF_UP;
 986         return flags;
 987 }
 988
 989 /*
 990  *      This outputs /proc/net/route.
 991  *
 992  *      It always works in backward compatibility mode.
 993  *      The format of the file is not supposed to be changed.
 994  */
 995 static int fib_seq_show(struct seq_file *seq, void *v)
 996 {
 997         struct fib_iter_state *iter;
 998         char bf[128];
 999         __be32 prefix, mask;
1000         unsigned flags;
1001         struct fib_node *f;
1002         struct fib_alias *fa;
1003         struct fib_info *fi;
1004
1005         if (v == SEQ_START_TOKEN) {
1006                 seq_printf(seq, "%-127s\n", "Iface\tDestination\tGateway "
1007                            "\tFlags\tRefCnt\tUse\tMetric\tMask\t\tMTU"
1008                            "\tWindow\tIRTT");
1009                 goto out;
1010         }
1011
1012         iter    = seq->private;
1013         f       = iter->fn;
1014         fa      = iter->fa;
1015         fi      = fa->fa_info;
1016         prefix  = f->fn_key;
1017         mask    = FZ_MASK(iter->zone);
1018         flags   = fib_flag_trans(fa->fa_type, mask, fi);
1019         if (fi)
1020                 snprintf(bf, sizeof(bf),
1021                          "%s\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u",
1022                          fi->fib_dev ? fi->fib_dev->name : "*", prefix,
1023                          fi->fib_nh->nh_gw, flags, 0, 0, fi->fib_priority,
1024                          mask, (fi->fib_advmss ? fi->fib_advmss + 40 : 0),
1025                          fi->fib_window,
1026                          fi->fib_rtt >> 3);
1027         else
1028                 snprintf(bf, sizeof(bf),
1029                          "*\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u",
1030                          prefix, 0, flags, 0, 0, 0, mask, 0, 0, 0);
1031         seq_printf(seq, "%-127s\n", bf);
1032 out:
1033         return 0;
1034 }
1035
1036 static const struct seq_operations fib_seq_ops = {
1037         .start  = fib_seq_start,
1038         .next   = fib_seq_next,
1039         .stop   = fib_seq_stop,
1040         .show   = fib_seq_show,
1041 };
1042
1043 static int fib_seq_open(struct inode *inode, struct file *file)
1044 {
1045         return seq_open_private(file, &fib_seq_ops,
1046                         sizeof(struct fib_iter_state));
1047 }
1048
1049 static const struct file_operations fib_seq_fops = {
1050         .owner          = THIS_MODULE,
1051         .open           = fib_seq_open,
1052         .read           = seq_read,
1053         .llseek         = seq_lseek,
1054         .release        = seq_release_private,
1055 };
1056
1057 int __init fib_proc_init(void)
1058 {
1059         if (!proc_net_fops_create(&init_net, "route", S_IRUGO, &fib_seq_fops))
1060                 return -ENOMEM;
1061         return 0;
1062 }
1063
1064 void __init fib_proc_exit(void)
1065 {
1066         proc_net_remove(&init_net, "route");
1067 }
1068 #endif /* CONFIG_PROC_FS */