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