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  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  */
  15
  16 #include <asm/uaccess.h>
  17 #include <asm/system.h>
  18 #include <linux/bitops.h>
  19 #include <linux/types.h>
  20 #include <linux/kernel.h>
  21 #include <linux/mm.h>
  22 #include <linux/string.h>
  23 #include <linux/socket.h>
  24 #include <linux/sockios.h>
  25 #include <linux/errno.h>
  26 #include <linux/in.h>
  27 #include <linux/inet.h>
  28 #include <linux/inetdevice.h>
  29 #include <linux/netdevice.h>
  30 #include <linux/if_arp.h>
  31 #include <linux/proc_fs.h>
  32 #include <linux/skbuff.h>
  33 #include <linux/netlink.h>
  34 #include <linux/init.h>
  35 #include <linux/slab.h>
  36
  37 #include <net/net_namespace.h>
  38 #include <net/ip.h>
  39 #include <net/protocol.h>
  40 #include <net/route.h>
  41 #include <net/tcp.h>
  42 #include <net/sock.h>
  43 #include <net/ip_fib.h>
  44
  45 #include "fib_lookup.h"
  46
  47 static struct kmem_cache *fn_hash_kmem __read_mostly;
  48 static struct kmem_cache *fn_alias_kmem __read_mostly;
  49
  50 struct fib_node {
  51         struct hlist_node       fn_hash;
  52         struct list_head        fn_alias;
  53         __be32                  fn_key;
  54         struct fib_alias        fn_embedded_alias;
  55 };
  56
  57 #define EMBEDDED_HASH_SIZE (L1_CACHE_BYTES / sizeof(struct hlist_head))
  58
  59 struct fn_zone {
  60         struct fn_zone __rcu    *fz_next;       /* Next not empty zone  */
  61         struct hlist_head __rcu *fz_hash;       /* Hash table pointer   */
  62         seqlock_t               fz_lock;
  63         u32                     fz_hashmask;    /* (fz_divisor - 1)     */
  64
  65         u8                      fz_order;       /* Zone order (0..32)   */
  66         u8                      fz_revorder;    /* 32 - fz_order        */
  67         __be32                  fz_mask;        /* inet_make_mask(order) */
  68 #define FZ_MASK(fz)             ((fz)->fz_mask)
  69
  70         struct hlist_head       fz_embedded_hash[EMBEDDED_HASH_SIZE];
  71
  72         int                     fz_nent;        /* Number of entries    */
  73         int                     fz_divisor;     /* Hash size (mask+1)   */
  74 };
  75
  76 struct fn_hash {
  77         struct fn_zone          *fn_zones[33];
  78         struct fn_zone __rcu    *fn_zone_list;
  79 };
  80
  81 static inline u32 fn_hash(__be32 key, struct fn_zone *fz)
  82 {
  83         u32 h = ntohl(key) >> fz->fz_revorder;
  84         h ^= (h>>20);
  85         h ^= (h>>10);
  86         h ^= (h>>5);
  87         h &= fz->fz_hashmask;
  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 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 kzalloc(size, GFP_KERNEL);
 106
 107         return (struct hlist_head *)
 108                 __get_free_pages(GFP_KERNEL | __GFP_ZERO, get_order(size));
 109 }
 110
 111 /* The fib hash lock must be held when this is called. */
 112 static inline void fn_rebuild_zone(struct fn_zone *fz,
 113                                    struct hlist_head *old_ht,
 114                                    int old_divisor)
 115 {
 116         int i;
 117
 118         for (i = 0; i < old_divisor; i++) {
 119                 struct hlist_node *node, *n;
 120                 struct fib_node *f;
 121
 122                 hlist_for_each_entry_safe(f, node, n, &old_ht[i], fn_hash) {
 123                         struct hlist_head *new_head;
 124
 125                         hlist_del_rcu(&f->fn_hash);
 126
 127                         new_head = rcu_dereference_protected(fz->fz_hash, 1) +
 128                                    fn_hash(f->fn_key, fz);
 129                         hlist_add_head_rcu(&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         new_divisor = old_divisor = fz->fz_divisor;
 151
 152         switch (old_divisor) {
 153         case EMBEDDED_HASH_SIZE:
 154                 new_divisor *= EMBEDDED_HASH_SIZE;
 155                 break;
 156         case EMBEDDED_HASH_SIZE*EMBEDDED_HASH_SIZE:
 157                 new_divisor *= (EMBEDDED_HASH_SIZE/2);
 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(KERN_DEBUG "fn_rehash_zone: hash for zone %d grows from %d\n",
 172                fz->fz_order, old_divisor);
 173 #endif
 174
 175         ht = fz_hash_alloc(new_divisor);
 176
 177         if (ht) {
 178                 struct fn_zone nfz;
 179
 180                 memcpy(&nfz, fz, sizeof(nfz));
 181
 182                 write_seqlock_bh(&fz->fz_lock);
 183                 old_ht = rcu_dereference_protected(fz->fz_hash, 1);
 184                 RCU_INIT_POINTER(nfz.fz_hash, ht);
 185                 nfz.fz_hashmask = new_hashmask;
 186                 nfz.fz_divisor = new_divisor;
 187                 fn_rebuild_zone(&nfz, old_ht, old_divisor);
 188                 fib_hash_genid++;
 189                 rcu_assign_pointer(fz->fz_hash, ht);
 190                 fz->fz_hashmask = new_hashmask;
 191                 fz->fz_divisor = new_divisor;
 192                 write_sequnlock_bh(&fz->fz_lock);
 193
 194                 if (old_ht != fz->fz_embedded_hash) {
 195                         synchronize_rcu();
 196                         fz_hash_free(old_ht, old_divisor);
 197                 }
 198         }
 199 }
 200
 201 static void fn_free_node_rcu(struct rcu_head *head)
 202 {
 203         struct fib_node *f = container_of(head, struct fib_node, fn_embedded_alias.rcu);
 204
 205         kmem_cache_free(fn_hash_kmem, f);
 206 }
 207
 208 static inline void fn_free_node(struct fib_node *f)
 209 {
 210         call_rcu(&f->fn_embedded_alias.rcu, fn_free_node_rcu);
 211 }
 212
 213 static void fn_free_alias_rcu(struct rcu_head *head)
 214 {
 215         struct fib_alias *fa = container_of(head, struct fib_alias, rcu);
 216
 217         kmem_cache_free(fn_alias_kmem, fa);
 218 }
 219
 220 static inline void fn_free_alias(struct fib_alias *fa, struct fib_node *f)
 221 {
 222         fib_release_info(fa->fa_info);
 223         if (fa == &f->fn_embedded_alias)
 224                 fa->fa_info = NULL;
 225         else
 226                 call_rcu(&fa->rcu, fn_free_alias_rcu);
 227 }
 228
 229 static struct fn_zone *
 230 fn_new_zone(struct fn_hash *table, int z)
 231 {
 232         int i;
 233         struct fn_zone *fz = kzalloc(sizeof(struct fn_zone), GFP_KERNEL);
 234         if (!fz)
 235                 return NULL;
 236
 237         seqlock_init(&fz->fz_lock);
 238         fz->fz_divisor = z ? EMBEDDED_HASH_SIZE : 1;
 239         fz->fz_hashmask = fz->fz_divisor - 1;
 240         RCU_INIT_POINTER(fz->fz_hash, fz->fz_embedded_hash);
 241         fz->fz_order = z;
 242         fz->fz_revorder = 32 - z;
 243         fz->fz_mask = inet_make_mask(z);
 244
 245         /* Find the first not empty zone with more specific mask */
 246         for (i = z + 1; i <= 32; i++)
 247                 if (table->fn_zones[i])
 248                         break;
 249         if (i > 32) {
 250                 /* No more specific masks, we are the first. */
 251                 rcu_assign_pointer(fz->fz_next,
 252                                    rtnl_dereference(table->fn_zone_list));
 253                 rcu_assign_pointer(table->fn_zone_list, fz);
 254         } else {
 255                 rcu_assign_pointer(fz->fz_next,
 256                                    rtnl_dereference(table->fn_zones[i]->fz_next));
 257                 rcu_assign_pointer(table->fn_zones[i]->fz_next, fz);
 258         }
 259         table->fn_zones[z] = fz;
 260         fib_hash_genid++;
 261         return fz;
 262 }
 263
 264 int fib_table_lookup(struct fib_table *tb,
 265                      const struct flowi *flp, struct fib_result *res,
 266                      int fib_flags)
 267 {
 268         int err;
 269         struct fn_zone *fz;
 270         struct fn_hash *t = (struct fn_hash *)tb->tb_data;
 271
 272         rcu_read_lock();
 273         for (fz = rcu_dereference(t->fn_zone_list);
 274              fz != NULL;
 275              fz = rcu_dereference(fz->fz_next)) {
 276                 struct hlist_head *head;
 277                 struct hlist_node *node;
 278                 struct fib_node *f;
 279                 __be32 k;
 280                 unsigned int seq;
 281
 282                 do {
 283                         seq = read_seqbegin(&fz->fz_lock);
 284                         k = fz_key(flp->fl4_dst, fz);
 285
 286                         head = rcu_dereference(fz->fz_hash) + fn_hash(k, fz);
 287                         hlist_for_each_entry_rcu(f, node, head, fn_hash) {
 288                                 if (f->fn_key != k)
 289                                         continue;
 290
 291                                 err = fib_semantic_match(&f->fn_alias,
 292                                                  flp, res,
 293                                                  fz->fz_order, fib_flags);
 294                                 if (err <= 0)
 295                                         goto out;
 296                         }
 297                 } while (read_seqretry(&fz->fz_lock, seq));
 298         }
 299         err = 1;
 300 out:
 301         rcu_read_unlock();
 302         return err;
 303 }
 304
 305 void fib_table_select_default(struct fib_table *tb,
 306                               const struct flowi *flp, struct fib_result *res)
 307 {
 308         int order, last_idx;
 309         struct hlist_node *node;
 310         struct fib_node *f;
 311         struct fib_info *fi = NULL;
 312         struct fib_info *last_resort;
 313         struct fn_hash *t = (struct fn_hash *)tb->tb_data;
 314         struct fn_zone *fz = t->fn_zones[0];
 315         struct hlist_head *head;
 316
 317         if (fz == NULL)
 318                 return;
 319
 320         last_idx = -1;
 321         last_resort = NULL;
 322         order = -1;
 323
 324         rcu_read_lock();
 325         head = rcu_dereference(fz->fz_hash);
 326         hlist_for_each_entry_rcu(f, node, head, fn_hash) {
 327                 struct fib_alias *fa;
 328
 329                 list_for_each_entry_rcu(fa, &f->fn_alias, fa_list) {
 330                         struct fib_info *next_fi = fa->fa_info;
 331
 332                         if (fa->fa_scope != res->scope ||
 333                             fa->fa_type != RTN_UNICAST)
 334                                 continue;
 335
 336                         if (next_fi->fib_priority > res->fi->fib_priority)
 337                                 break;
 338                         if (!next_fi->fib_nh[0].nh_gw ||
 339                             next_fi->fib_nh[0].nh_scope != RT_SCOPE_LINK)
 340                                 continue;
 341
 342                         fib_alias_accessed(fa);
 343
 344                         if (fi == NULL) {
 345                                 if (next_fi != res->fi)
 346                                         break;
 347                         } else if (!fib_detect_death(fi, order, &last_resort,
 348                                                 &last_idx, tb->tb_default)) {
 349                                 fib_result_assign(res, fi);
 350                                 tb->tb_default = order;
 351                                 goto out;
 352                         }
 353                         fi = next_fi;
 354                         order++;
 355                 }
 356         }
 357
 358         if (order <= 0 || fi == NULL) {
 359                 tb->tb_default = -1;
 360                 goto out;
 361         }
 362
 363         if (!fib_detect_death(fi, order, &last_resort, &last_idx,
 364                                 tb->tb_default)) {
 365                 fib_result_assign(res, fi);
 366                 tb->tb_default = order;
 367                 goto out;
 368         }
 369
 370         if (last_idx >= 0)
 371                 fib_result_assign(res, last_resort);
 372         tb->tb_default = last_idx;
 373 out:
 374         rcu_read_unlock();
 375 }
 376
 377 /* Insert node F to FZ. */
 378 static inline void fib_insert_node(struct fn_zone *fz, struct fib_node *f)
 379 {
 380         struct hlist_head *head = rtnl_dereference(fz->fz_hash) + fn_hash(f->fn_key, fz);
 381
 382         hlist_add_head_rcu(&f->fn_hash, head);
 383 }
 384
 385 /* Return the node in FZ matching KEY. */
 386 static struct fib_node *fib_find_node(struct fn_zone *fz, __be32 key)
 387 {
 388         struct hlist_head *head = rtnl_dereference(fz->fz_hash) + fn_hash(key, fz);
 389         struct hlist_node *node;
 390         struct fib_node *f;
 391
 392         hlist_for_each_entry_rcu(f, node, head, fn_hash) {
 393                 if (f->fn_key == key)
 394                         return f;
 395         }
 396
 397         return NULL;
 398 }
 399
 400
 401 static struct fib_alias *fib_fast_alloc(struct fib_node *f)
 402 {
 403         struct fib_alias *fa = &f->fn_embedded_alias;
 404
 405         if (fa->fa_info != NULL)
 406                 fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
 407         return fa;
 408 }
 409
 410 /* Caller must hold RTNL. */
 411 int fib_table_insert(struct fib_table *tb, struct fib_config *cfg)
 412 {
 413         struct fn_hash *table = (struct fn_hash *) tb->tb_data;
 414         struct fib_node *new_f = NULL;
 415         struct fib_node *f;
 416         struct fib_alias *fa, *new_fa;
 417         struct fn_zone *fz;
 418         struct fib_info *fi;
 419         u8 tos = cfg->fc_tos;
 420         __be32 key;
 421         int err;
 422
 423         if (cfg->fc_dst_len > 32)
 424                 return -EINVAL;
 425
 426         fz = table->fn_zones[cfg->fc_dst_len];
 427         if (!fz && !(fz = fn_new_zone(table, cfg->fc_dst_len)))
 428                 return -ENOBUFS;
 429
 430         key = 0;
 431         if (cfg->fc_dst) {
 432                 if (cfg->fc_dst & ~FZ_MASK(fz))
 433                         return -EINVAL;
 434                 key = fz_key(cfg->fc_dst, fz);
 435         }
 436
 437         fi = fib_create_info(cfg);
 438         if (IS_ERR(fi))
 439                 return PTR_ERR(fi);
 440
 441         if (fz->fz_nent > (fz->fz_divisor<<1) &&
 442             fz->fz_divisor < FZ_MAX_DIVISOR &&
 443             (cfg->fc_dst_len == 32 ||
 444              (1 << cfg->fc_dst_len) > fz->fz_divisor))
 445                 fn_rehash_zone(fz);
 446
 447         f = fib_find_node(fz, key);
 448
 449         if (!f)
 450                 fa = NULL;
 451         else
 452                 fa = fib_find_alias(&f->fn_alias, tos, fi->fib_priority);
 453
 454         /* Now fa, if non-NULL, points to the first fib alias
 455          * with the same keys [prefix,tos,priority], if such key already
 456          * exists or to the node before which we will insert new one.
 457          *
 458          * If fa is NULL, we will need to allocate a new one and
 459          * insert to the head of f.
 460          *
 461          * If f is NULL, no fib node matched the destination key
 462          * and we need to allocate a new one of those as well.
 463          */
 464
 465         if (fa && fa->fa_tos == tos &&
 466             fa->fa_info->fib_priority == fi->fib_priority) {
 467                 struct fib_alias *fa_first, *fa_match;
 468
 469                 err = -EEXIST;
 470                 if (cfg->fc_nlflags & NLM_F_EXCL)
 471                         goto out;
 472
 473                 /* We have 2 goals:
 474                  * 1. Find exact match for type, scope, fib_info to avoid
 475                  * duplicate routes
 476                  * 2. Find next 'fa' (or head), NLM_F_APPEND inserts before it
 477                  */
 478                 fa_match = NULL;
 479                 fa_first = fa;
 480                 fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
 481                 list_for_each_entry_continue(fa, &f->fn_alias, fa_list) {
 482                         if (fa->fa_tos != tos)
 483                                 break;
 484                         if (fa->fa_info->fib_priority != fi->fib_priority)
 485                                 break;
 486                         if (fa->fa_type == cfg->fc_type &&
 487                             fa->fa_scope == cfg->fc_scope &&
 488                             fa->fa_info == fi) {
 489                                 fa_match = fa;
 490                                 break;
 491                         }
 492                 }
 493
 494                 if (cfg->fc_nlflags & NLM_F_REPLACE) {
 495                         u8 state;
 496
 497                         fa = fa_first;
 498                         if (fa_match) {
 499                                 if (fa == fa_match)
 500                                         err = 0;
 501                                 goto out;
 502                         }
 503                         err = -ENOBUFS;
 504                         new_fa = fib_fast_alloc(f);
 505                         if (new_fa == NULL)
 506                                 goto out;
 507
 508                         new_fa->fa_tos = fa->fa_tos;
 509                         new_fa->fa_info = fi;
 510                         new_fa->fa_type = cfg->fc_type;
 511                         new_fa->fa_scope = cfg->fc_scope;
 512                         state = fa->fa_state;
 513                         new_fa->fa_state = state & ~FA_S_ACCESSED;
 514                         fib_hash_genid++;
 515                         list_replace_rcu(&fa->fa_list, &new_fa->fa_list);
 516
 517                         fn_free_alias(fa, f);
 518                         if (state & FA_S_ACCESSED)
 519                                 rt_cache_flush(cfg->fc_nlinfo.nl_net, -1);
 520                         rtmsg_fib(RTM_NEWROUTE, key, new_fa, cfg->fc_dst_len,
 521                                   tb->tb_id, &cfg->fc_nlinfo, NLM_F_REPLACE);
 522                         return 0;
 523                 }
 524
 525                 /* Error if we find a perfect match which
 526                  * uses the same scope, type, and nexthop
 527                  * information.
 528                  */
 529                 if (fa_match)
 530                         goto out;
 531
 532                 if (!(cfg->fc_nlflags & NLM_F_APPEND))
 533                         fa = fa_first;
 534         }
 535
 536         err = -ENOENT;
 537         if (!(cfg->fc_nlflags & NLM_F_CREATE))
 538                 goto out;
 539
 540         err = -ENOBUFS;
 541
 542         if (!f) {
 543                 new_f = kmem_cache_zalloc(fn_hash_kmem, GFP_KERNEL);
 544                 if (new_f == NULL)
 545                         goto out;
 546
 547                 INIT_HLIST_NODE(&new_f->fn_hash);
 548                 INIT_LIST_HEAD(&new_f->fn_alias);
 549                 new_f->fn_key = key;
 550                 f = new_f;
 551         }
 552
 553         new_fa = fib_fast_alloc(f);
 554         if (new_fa == NULL)
 555                 goto out;
 556
 557         new_fa->fa_info = fi;
 558         new_fa->fa_tos = tos;
 559         new_fa->fa_type = cfg->fc_type;
 560         new_fa->fa_scope = cfg->fc_scope;
 561         new_fa->fa_state = 0;
 562
 563         /*
 564          * Insert new entry to the list.
 565          */
 566
 567         if (new_f)
 568                 fib_insert_node(fz, new_f);
 569         list_add_tail_rcu(&new_fa->fa_list,
 570                  (fa ? &fa->fa_list : &f->fn_alias));
 571         fib_hash_genid++;
 572
 573         if (new_f)
 574                 fz->fz_nent++;
 575         rt_cache_flush(cfg->fc_nlinfo.nl_net, -1);
 576
 577         rtmsg_fib(RTM_NEWROUTE, key, new_fa, cfg->fc_dst_len, tb->tb_id,
 578                   &cfg->fc_nlinfo, 0);
 579         return 0;
 580
 581 out:
 582         if (new_f)
 583                 kmem_cache_free(fn_hash_kmem, new_f);
 584         fib_release_info(fi);
 585         return err;
 586 }
 587
 588 int fib_table_delete(struct fib_table *tb, struct fib_config *cfg)
 589 {
 590         struct fn_hash *table = (struct fn_hash *)tb->tb_data;
 591         struct fib_node *f;
 592         struct fib_alias *fa, *fa_to_delete;
 593         struct fn_zone *fz;
 594         __be32 key;
 595
 596         if (cfg->fc_dst_len > 32)
 597                 return -EINVAL;
 598
 599         if ((fz  = table->fn_zones[cfg->fc_dst_len]) == NULL)
 600                 return -ESRCH;
 601
 602         key = 0;
 603         if (cfg->fc_dst) {
 604                 if (cfg->fc_dst & ~FZ_MASK(fz))
 605                         return -EINVAL;
 606                 key = fz_key(cfg->fc_dst, fz);
 607         }
 608
 609         f = fib_find_node(fz, key);
 610
 611         if (!f)
 612                 fa = NULL;
 613         else
 614                 fa = fib_find_alias(&f->fn_alias, cfg->fc_tos, 0);
 615         if (!fa)
 616                 return -ESRCH;
 617
 618         fa_to_delete = NULL;
 619         fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
 620         list_for_each_entry_continue(fa, &f->fn_alias, fa_list) {
 621                 struct fib_info *fi = fa->fa_info;
 622
 623                 if (fa->fa_tos != cfg->fc_tos)
 624                         break;
 625
 626                 if ((!cfg->fc_type ||
 627                      fa->fa_type == cfg->fc_type) &&
 628                     (cfg->fc_scope == RT_SCOPE_NOWHERE ||
 629                      fa->fa_scope == cfg->fc_scope) &&
 630                     (!cfg->fc_protocol ||
 631                      fi->fib_protocol == cfg->fc_protocol) &&
 632                     fib_nh_match(cfg, fi) == 0) {
 633                         fa_to_delete = fa;
 634                         break;
 635                 }
 636         }
 637
 638         if (fa_to_delete) {
 639                 int kill_fn;
 640
 641                 fa = fa_to_delete;
 642                 rtmsg_fib(RTM_DELROUTE, key, fa, cfg->fc_dst_len,
 643                           tb->tb_id, &cfg->fc_nlinfo, 0);
 644
 645                 kill_fn = 0;
 646                 list_del_rcu(&fa->fa_list);
 647                 if (list_empty(&f->fn_alias)) {
 648                         hlist_del_rcu(&f->fn_hash);
 649                         kill_fn = 1;
 650                 }
 651                 fib_hash_genid++;
 652
 653                 if (fa->fa_state & FA_S_ACCESSED)
 654                         rt_cache_flush(cfg->fc_nlinfo.nl_net, -1);
 655                 fn_free_alias(fa, f);
 656                 if (kill_fn) {
 657                         fn_free_node(f);
 658                         fz->fz_nent--;
 659                 }
 660
 661                 return 0;
 662         }
 663         return -ESRCH;
 664 }
 665
 666 static int fn_flush_list(struct fn_zone *fz, int idx)
 667 {
 668         struct hlist_head *head = rtnl_dereference(fz->fz_hash) + idx;
 669         struct hlist_node *node, *n;
 670         struct fib_node *f;
 671         int found = 0;
 672
 673         hlist_for_each_entry_safe(f, node, n, head, fn_hash) {
 674                 struct fib_alias *fa, *fa_node;
 675                 int kill_f;
 676
 677                 kill_f = 0;
 678                 list_for_each_entry_safe(fa, fa_node, &f->fn_alias, fa_list) {
 679                         struct fib_info *fi = fa->fa_info;
 680
 681                         if (fi && (fi->fib_flags&RTNH_F_DEAD)) {
 682                                 list_del_rcu(&fa->fa_list);
 683                                 if (list_empty(&f->fn_alias)) {
 684                                         hlist_del_rcu(&f->fn_hash);
 685                                         kill_f = 1;
 686                                 }
 687                                 fib_hash_genid++;
 688
 689                                 fn_free_alias(fa, f);
 690                                 found++;
 691                         }
 692                 }
 693                 if (kill_f) {
 694                         fn_free_node(f);
 695                         fz->fz_nent--;
 696                 }
 697         }
 698         return found;
 699 }
 700
 701 /* caller must hold RTNL. */
 702 int fib_table_flush(struct fib_table *tb)
 703 {
 704         struct fn_hash *table = (struct fn_hash *) tb->tb_data;
 705         struct fn_zone *fz;
 706         int found = 0;
 707
 708         for (fz = rtnl_dereference(table->fn_zone_list);
 709              fz != NULL;
 710              fz = rtnl_dereference(fz->fz_next)) {
 711                 int i;
 712
 713                 for (i = fz->fz_divisor - 1; i >= 0; i--)
 714                         found += fn_flush_list(fz, i);
 715         }
 716         return found;
 717 }
 718
 719 void fib_free_table(struct fib_table *tb)
 720 {
 721         struct fn_hash *table = (struct fn_hash *) tb->tb_data;
 722         struct fn_zone *fz, *next;
 723
 724         next = table->fn_zone_list;
 725         while (next != NULL) {
 726                 fz = next;
 727                 next = fz->fz_next;
 728
 729                 if (fz->fz_hash != fz->fz_embedded_hash)
 730                         fz_hash_free(fz->fz_hash, fz->fz_divisor);
 731
 732                 kfree(fz);
 733         }
 734
 735         kfree(tb);
 736 }
 737
 738 static inline int
 739 fn_hash_dump_bucket(struct sk_buff *skb, struct netlink_callback *cb,
 740                      struct fib_table *tb,
 741                      struct fn_zone *fz,
 742                      struct hlist_head *head)
 743 {
 744         struct hlist_node *node;
 745         struct fib_node *f;
 746         int i, s_i;
 747
 748         s_i = cb->args[4];
 749         i = 0;
 750         hlist_for_each_entry_rcu(f, node, head, fn_hash) {
 751                 struct fib_alias *fa;
 752
 753                 list_for_each_entry_rcu(fa, &f->fn_alias, fa_list) {
 754                         if (i < s_i)
 755                                 goto next;
 756
 757                         if (fib_dump_info(skb, NETLINK_CB(cb->skb).pid,
 758                                           cb->nlh->nlmsg_seq,
 759                                           RTM_NEWROUTE,
 760                                           tb->tb_id,
 761                                           fa->fa_type,
 762                                           fa->fa_scope,
 763                                           f->fn_key,
 764                                           fz->fz_order,
 765                                           fa->fa_tos,
 766                                           fa->fa_info,
 767                                           NLM_F_MULTI) < 0) {
 768                                 cb->args[4] = i;
 769                                 return -1;
 770                         }
 771 next:
 772                         i++;
 773                 }
 774         }
 775         cb->args[4] = i;
 776         return skb->len;
 777 }
 778
 779 static inline int
 780 fn_hash_dump_zone(struct sk_buff *skb, struct netlink_callback *cb,
 781                    struct fib_table *tb,
 782                    struct fn_zone *fz)
 783 {
 784         int h, s_h;
 785         struct hlist_head *head = rcu_dereference(fz->fz_hash);
 786
 787         if (head == NULL)
 788                 return skb->len;
 789         s_h = cb->args[3];
 790         for (h = s_h; h < fz->fz_divisor; h++) {
 791                 if (hlist_empty(head + h))
 792                         continue;
 793                 if (fn_hash_dump_bucket(skb, cb, tb, fz, head + h) < 0) {
 794                         cb->args[3] = h;
 795                         return -1;
 796                 }
 797                 memset(&cb->args[4], 0,
 798                        sizeof(cb->args) - 4*sizeof(cb->args[0]));
 799         }
 800         cb->args[3] = h;
 801         return skb->len;
 802 }
 803
 804 int fib_table_dump(struct fib_table *tb, struct sk_buff *skb,
 805                    struct netlink_callback *cb)
 806 {
 807         int m = 0, s_m;
 808         struct fn_zone *fz;
 809         struct fn_hash *table = (struct fn_hash *)tb->tb_data;
 810
 811         s_m = cb->args[2];
 812         rcu_read_lock();
 813         for (fz = rcu_dereference(table->fn_zone_list);
 814              fz != NULL;
 815              fz = rcu_dereference(fz->fz_next), m++) {
 816                 if (m < s_m)
 817                         continue;
 818                 if (fn_hash_dump_zone(skb, cb, tb, fz) < 0) {
 819                         cb->args[2] = m;
 820                         rcu_read_unlock();
 821                         return -1;
 822                 }
 823                 memset(&cb->args[3], 0,
 824                        sizeof(cb->args) - 3*sizeof(cb->args[0]));
 825         }
 826         rcu_read_unlock();
 827         cb->args[2] = m;
 828         return skb->len;
 829 }
 830
 831 void __init fib_hash_init(void)
 832 {
 833         fn_hash_kmem = kmem_cache_create("ip_fib_hash", sizeof(struct fib_node),
 834                                          0, SLAB_PANIC, NULL);
 835
 836         fn_alias_kmem = kmem_cache_create("ip_fib_alias", sizeof(struct fib_alias),
 837                                           0, SLAB_PANIC, NULL);
 838
 839 }
 840
 841 struct fib_table *fib_hash_table(u32 id)
 842 {
 843         struct fib_table *tb;
 844
 845         tb = kmalloc(sizeof(struct fib_table) + sizeof(struct fn_hash),
 846                      GFP_KERNEL);
 847         if (tb == NULL)
 848                 return NULL;
 849
 850         tb->tb_id = id;
 851         tb->tb_default = -1;
 852
 853         memset(tb->tb_data, 0, sizeof(struct fn_hash));
 854         return tb;
 855 }
 856
 857 /* ------------------------------------------------------------------------ */
 858 #ifdef CONFIG_PROC_FS
 859
 860 struct fib_iter_state {
 861         struct seq_net_private p;
 862         struct fn_zone  *zone;
 863         int             bucket;
 864         struct hlist_head *hash_head;
 865         struct fib_node *fn;
 866         struct fib_alias *fa;
 867         loff_t pos;
 868         unsigned int genid;
 869         int valid;
 870 };
 871
 872 static struct fib_alias *fib_get_first(struct seq_file *seq)
 873 {
 874         struct fib_iter_state *iter = seq->private;
 875         struct fib_table *main_table;
 876         struct fn_hash *table;
 877
 878         main_table = fib_get_table(seq_file_net(seq), RT_TABLE_MAIN);
 879         table = (struct fn_hash *)main_table->tb_data;
 880
 881         iter->bucket    = 0;
 882         iter->hash_head = NULL;
 883         iter->fn        = NULL;
 884         iter->fa        = NULL;
 885         iter->pos       = 0;
 886         iter->genid     = fib_hash_genid;
 887         iter->valid     = 1;
 888
 889         for (iter->zone = rcu_dereference(table->fn_zone_list);
 890              iter->zone != NULL;
 891              iter->zone = rcu_dereference(iter->zone->fz_next)) {
 892                 int maxslot;
 893
 894                 if (!iter->zone->fz_nent)
 895                         continue;
 896
 897                 iter->hash_head = rcu_dereference(iter->zone->fz_hash);
 898                 maxslot = iter->zone->fz_divisor;
 899
 900                 for (iter->bucket = 0; iter->bucket < maxslot;
 901                      ++iter->bucket, ++iter->hash_head) {
 902                         struct hlist_node *node;
 903                         struct fib_node *fn;
 904
 905                         hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) {
 906                                 struct fib_alias *fa;
 907
 908                                 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
 909                                         iter->fn = fn;
 910                                         iter->fa = fa;
 911                                         goto out;
 912                                 }
 913                         }
 914                 }
 915         }
 916 out:
 917         return iter->fa;
 918 }
 919
 920 static struct fib_alias *fib_get_next(struct seq_file *seq)
 921 {
 922         struct fib_iter_state *iter = seq->private;
 923         struct fib_node *fn;
 924         struct fib_alias *fa;
 925
 926         /* Advance FA, if any. */
 927         fn = iter->fn;
 928         fa = iter->fa;
 929         if (fa) {
 930                 BUG_ON(!fn);
 931                 list_for_each_entry_continue(fa, &fn->fn_alias, fa_list) {
 932                         iter->fa = fa;
 933                         goto out;
 934                 }
 935         }
 936
 937         fa = iter->fa = NULL;
 938
 939         /* Advance FN. */
 940         if (fn) {
 941                 struct hlist_node *node = &fn->fn_hash;
 942                 hlist_for_each_entry_continue(fn, node, fn_hash) {
 943                         iter->fn = fn;
 944
 945                         list_for_each_entry(fa, &fn->fn_alias, fa_list) {
 946                                 iter->fa = fa;
 947                                 goto out;
 948                         }
 949                 }
 950         }
 951
 952         fn = iter->fn = NULL;
 953
 954         /* Advance hash chain. */
 955         if (!iter->zone)
 956                 goto out;
 957
 958         for (;;) {
 959                 struct hlist_node *node;
 960                 int maxslot;
 961
 962                 maxslot = iter->zone->fz_divisor;
 963
 964                 while (++iter->bucket < maxslot) {
 965                         iter->hash_head++;
 966
 967                         hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) {
 968                                 list_for_each_entry(fa, &fn->fn_alias, fa_list) {
 969                                         iter->fn = fn;
 970                                         iter->fa = fa;
 971                                         goto out;
 972                                 }
 973                         }
 974                 }
 975
 976                 iter->zone = rcu_dereference(iter->zone->fz_next);
 977
 978                 if (!iter->zone)
 979                         goto out;
 980
 981                 iter->bucket = 0;
 982                 iter->hash_head = rcu_dereference(iter->zone->fz_hash);
 983
 984                 hlist_for_each_entry(fn, node, iter->hash_head, fn_hash) {
 985                         list_for_each_entry(fa, &fn->fn_alias, fa_list) {
 986                                 iter->fn = fn;
 987                                 iter->fa = fa;
 988                                 goto out;
 989                         }
 990                 }
 991         }
 992 out:
 993         iter->pos++;
 994         return fa;
 995 }
 996
 997 static struct fib_alias *fib_get_idx(struct seq_file *seq, loff_t pos)
 998 {
 999         struct fib_iter_state *iter = seq->private;
1000         struct fib_alias *fa;
1001
1002         if (iter->valid && pos >= iter->pos && iter->genid == fib_hash_genid) {
1003                 fa   = iter->fa;
1004                 pos -= iter->pos;
1005         } else
1006                 fa = fib_get_first(seq);
1007
1008         if (fa)
1009                 while (pos && (fa = fib_get_next(seq)))
1010                         --pos;
1011         return pos ? NULL : fa;
1012 }
1013
1014 static void *fib_seq_start(struct seq_file *seq, loff_t *pos)
1015         __acquires(RCU)
1016 {
1017         void *v = NULL;
1018
1019         rcu_read_lock();
1020         if (fib_get_table(seq_file_net(seq), RT_TABLE_MAIN))
1021                 v = *pos ? fib_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1022         return v;
1023 }
1024
1025 static void *fib_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1026 {
1027         ++*pos;
1028         return v == SEQ_START_TOKEN ? fib_get_first(seq) : fib_get_next(seq);
1029 }
1030
1031 static void fib_seq_stop(struct seq_file *seq, void *v)
1032         __releases(RCU)
1033 {
1034         rcu_read_unlock();
1035 }
1036
1037 static unsigned fib_flag_trans(int type, __be32 mask, struct fib_info *fi)
1038 {
1039         static const unsigned type2flags[RTN_MAX + 1] = {
1040                 [7] = RTF_REJECT,
1041                 [8] = RTF_REJECT,
1042         };
1043         unsigned flags = type2flags[type];
1044
1045         if (fi && fi->fib_nh->nh_gw)
1046                 flags |= RTF_GATEWAY;
1047         if (mask == htonl(0xFFFFFFFF))
1048                 flags |= RTF_HOST;
1049         flags |= RTF_UP;
1050         return flags;
1051 }
1052
1053 /*
1054  *      This outputs /proc/net/route.
1055  *
1056  *      It always works in backward compatibility mode.
1057  *      The format of the file is not supposed to be changed.
1058  */
1059 static int fib_seq_show(struct seq_file *seq, void *v)
1060 {
1061         struct fib_iter_state *iter;
1062         int len;
1063         __be32 prefix, mask;
1064         unsigned flags;
1065         struct fib_node *f;
1066         struct fib_alias *fa;
1067         struct fib_info *fi;
1068
1069         if (v == SEQ_START_TOKEN) {
1070                 seq_printf(seq, "%-127s\n", "Iface\tDestination\tGateway "
1071                            "\tFlags\tRefCnt\tUse\tMetric\tMask\t\tMTU"
1072                            "\tWindow\tIRTT");
1073                 goto out;
1074         }
1075
1076         iter    = seq->private;
1077         f       = iter->fn;
1078         fa      = iter->fa;
1079         fi      = fa->fa_info;
1080         prefix  = f->fn_key;
1081         mask    = FZ_MASK(iter->zone);
1082         flags   = fib_flag_trans(fa->fa_type, mask, fi);
1083         if (fi)
1084                 seq_printf(seq,
1085                          "%s\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u%n",
1086                          fi->fib_dev ? fi->fib_dev->name : "*", prefix,
1087                          fi->fib_nh->nh_gw, flags, 0, 0, fi->fib_priority,
1088                          mask, (fi->fib_advmss ? fi->fib_advmss + 40 : 0),
1089                          fi->fib_window,
1090                          fi->fib_rtt >> 3, &len);
1091         else
1092                 seq_printf(seq,
1093                          "*\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u%n",
1094                          prefix, 0, flags, 0, 0, 0, mask, 0, 0, 0, &len);
1095
1096         seq_printf(seq, "%*s\n", 127 - len, "");
1097 out:
1098         return 0;
1099 }
1100
1101 static const struct seq_operations fib_seq_ops = {
1102         .start  = fib_seq_start,
1103         .next   = fib_seq_next,
1104         .stop   = fib_seq_stop,
1105         .show   = fib_seq_show,
1106 };
1107
1108 static int fib_seq_open(struct inode *inode, struct file *file)
1109 {
1110         return seq_open_net(inode, file, &fib_seq_ops,
1111                             sizeof(struct fib_iter_state));
1112 }
1113
1114 static const struct file_operations fib_seq_fops = {
1115         .owner          = THIS_MODULE,
1116         .open           = fib_seq_open,
1117         .read           = seq_read,
1118         .llseek         = seq_lseek,
1119         .release        = seq_release_net,
1120 };
1121
1122 int __net_init fib_proc_init(struct net *net)
1123 {
1124         if (!proc_net_fops_create(net, "route", S_IRUGO, &fib_seq_fops))
1125                 return -ENOMEM;
1126         return 0;
1127 }
1128
1129 void __net_exit fib_proc_exit(struct net *net)
1130 {
1131         proc_net_remove(net, "route");
1132 }
1133 #endif /* CONFIG_PROC_FS */