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