~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/ipv6/ip6_fib.c

Version: ~ [ linux-5.8 ] ~ [ linux-5.7.12 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.55 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.136 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.191 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.232 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.232 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  *      Linux INET6 implementation
  3  *      Forwarding Information Database
  4  *
  5  *      Authors:
  6  *      Pedro Roque             <roque@di.fc.ul.pt>
  7  *
  8  *      This program is free software; you can redistribute it and/or
  9  *      modify it under the terms of the GNU General Public License
 10  *      as published by the Free Software Foundation; either version
 11  *      2 of the License, or (at your option) any later version.
 12  *
 13  *      Changes:
 14  *      Yuji SEKIYA @USAGI:     Support default route on router node;
 15  *                              remove ip6_null_entry from the top of
 16  *                              routing table.
 17  *      Ville Nuorvala:         Fixed routing subtrees.
 18  */
 19 
 20 #define pr_fmt(fmt) "IPv6: " fmt
 21 
 22 #include <linux/errno.h>
 23 #include <linux/types.h>
 24 #include <linux/net.h>
 25 #include <linux/route.h>
 26 #include <linux/netdevice.h>
 27 #include <linux/in6.h>
 28 #include <linux/init.h>
 29 #include <linux/list.h>
 30 #include <linux/slab.h>
 31 
 32 #include <net/ipv6.h>
 33 #include <net/ndisc.h>
 34 #include <net/addrconf.h>
 35 
 36 #include <net/ip6_fib.h>
 37 #include <net/ip6_route.h>
 38 
 39 #define RT6_DEBUG 2
 40 
 41 #if RT6_DEBUG >= 3
 42 #define RT6_TRACE(x...) pr_debug(x)
 43 #else
 44 #define RT6_TRACE(x...) do { ; } while (0)
 45 #endif
 46 
 47 static struct kmem_cache *fib6_node_kmem __read_mostly;
 48 
 49 struct fib6_cleaner {
 50         struct fib6_walker w;
 51         struct net *net;
 52         int (*func)(struct rt6_info *, void *arg);
 53         int sernum;
 54         void *arg;
 55 };
 56 
 57 static DEFINE_RWLOCK(fib6_walker_lock);
 58 
 59 #ifdef CONFIG_IPV6_SUBTREES
 60 #define FWS_INIT FWS_S
 61 #else
 62 #define FWS_INIT FWS_L
 63 #endif
 64 
 65 static void fib6_prune_clones(struct net *net, struct fib6_node *fn);
 66 static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn);
 67 static struct fib6_node *fib6_repair_tree(struct net *net, struct fib6_node *fn);
 68 static int fib6_walk(struct fib6_walker *w);
 69 static int fib6_walk_continue(struct fib6_walker *w);
 70 
 71 /*
 72  *      A routing update causes an increase of the serial number on the
 73  *      affected subtree. This allows for cached routes to be asynchronously
 74  *      tested when modifications are made to the destination cache as a
 75  *      result of redirects, path MTU changes, etc.
 76  */
 77 
 78 static void fib6_gc_timer_cb(unsigned long arg);
 79 
 80 static LIST_HEAD(fib6_walkers);
 81 #define FOR_WALKERS(w) list_for_each_entry(w, &fib6_walkers, lh)
 82 
 83 static void fib6_walker_link(struct fib6_walker *w)
 84 {
 85         write_lock_bh(&fib6_walker_lock);
 86         list_add(&w->lh, &fib6_walkers);
 87         write_unlock_bh(&fib6_walker_lock);
 88 }
 89 
 90 static void fib6_walker_unlink(struct fib6_walker *w)
 91 {
 92         write_lock_bh(&fib6_walker_lock);
 93         list_del(&w->lh);
 94         write_unlock_bh(&fib6_walker_lock);
 95 }
 96 
 97 static int fib6_new_sernum(struct net *net)
 98 {
 99         int new, old;
100 
101         do {
102                 old = atomic_read(&net->ipv6.fib6_sernum);
103                 new = old < INT_MAX ? old + 1 : 1;
104         } while (atomic_cmpxchg(&net->ipv6.fib6_sernum,
105                                 old, new) != old);
106         return new;
107 }
108 
109 enum {
110         FIB6_NO_SERNUM_CHANGE = 0,
111 };
112 
113 /*
114  *      Auxiliary address test functions for the radix tree.
115  *
116  *      These assume a 32bit processor (although it will work on
117  *      64bit processors)
118  */
119 
120 /*
121  *      test bit
122  */
123 #if defined(__LITTLE_ENDIAN)
124 # define BITOP_BE32_SWIZZLE     (0x1F & ~7)
125 #else
126 # define BITOP_BE32_SWIZZLE     0
127 #endif
128 
129 static __be32 addr_bit_set(const void *token, int fn_bit)
130 {
131         const __be32 *addr = token;
132         /*
133          * Here,
134          *      1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)
135          * is optimized version of
136          *      htonl(1 << ((~fn_bit)&0x1F))
137          * See include/asm-generic/bitops/le.h.
138          */
139         return (__force __be32)(1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)) &
140                addr[fn_bit >> 5];
141 }
142 
143 static struct fib6_node *node_alloc(void)
144 {
145         struct fib6_node *fn;
146 
147         fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC);
148 
149         return fn;
150 }
151 
152 static void node_free(struct fib6_node *fn)
153 {
154         kmem_cache_free(fib6_node_kmem, fn);
155 }
156 
157 static void rt6_release(struct rt6_info *rt)
158 {
159         if (atomic_dec_and_test(&rt->rt6i_ref))
160                 dst_free(&rt->dst);
161 }
162 
163 static void fib6_free_table(struct fib6_table *table)
164 {
165         inetpeer_invalidate_tree(&table->tb6_peers);
166         kfree(table);
167 }
168 
169 static void fib6_link_table(struct net *net, struct fib6_table *tb)
170 {
171         unsigned int h;
172 
173         /*
174          * Initialize table lock at a single place to give lockdep a key,
175          * tables aren't visible prior to being linked to the list.
176          */
177         rwlock_init(&tb->tb6_lock);
178 
179         h = tb->tb6_id & (FIB6_TABLE_HASHSZ - 1);
180 
181         /*
182          * No protection necessary, this is the only list mutatation
183          * operation, tables never disappear once they exist.
184          */
185         hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]);
186 }
187 
188 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
189 
190 static struct fib6_table *fib6_alloc_table(struct net *net, u32 id)
191 {
192         struct fib6_table *table;
193 
194         table = kzalloc(sizeof(*table), GFP_ATOMIC);
195         if (table) {
196                 table->tb6_id = id;
197                 table->tb6_root.leaf = net->ipv6.ip6_null_entry;
198                 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
199                 inet_peer_base_init(&table->tb6_peers);
200         }
201 
202         return table;
203 }
204 
205 struct fib6_table *fib6_new_table(struct net *net, u32 id)
206 {
207         struct fib6_table *tb;
208 
209         if (id == 0)
210                 id = RT6_TABLE_MAIN;
211         tb = fib6_get_table(net, id);
212         if (tb)
213                 return tb;
214 
215         tb = fib6_alloc_table(net, id);
216         if (tb)
217                 fib6_link_table(net, tb);
218 
219         return tb;
220 }
221 
222 struct fib6_table *fib6_get_table(struct net *net, u32 id)
223 {
224         struct fib6_table *tb;
225         struct hlist_head *head;
226         unsigned int h;
227 
228         if (id == 0)
229                 id = RT6_TABLE_MAIN;
230         h = id & (FIB6_TABLE_HASHSZ - 1);
231         rcu_read_lock();
232         head = &net->ipv6.fib_table_hash[h];
233         hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
234                 if (tb->tb6_id == id) {
235                         rcu_read_unlock();
236                         return tb;
237                 }
238         }
239         rcu_read_unlock();
240 
241         return NULL;
242 }
243 
244 static void __net_init fib6_tables_init(struct net *net)
245 {
246         fib6_link_table(net, net->ipv6.fib6_main_tbl);
247         fib6_link_table(net, net->ipv6.fib6_local_tbl);
248 }
249 #else
250 
251 struct fib6_table *fib6_new_table(struct net *net, u32 id)
252 {
253         return fib6_get_table(net, id);
254 }
255 
256 struct fib6_table *fib6_get_table(struct net *net, u32 id)
257 {
258           return net->ipv6.fib6_main_tbl;
259 }
260 
261 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
262                                    int flags, pol_lookup_t lookup)
263 {
264         return (struct dst_entry *) lookup(net, net->ipv6.fib6_main_tbl, fl6, flags);
265 }
266 
267 static void __net_init fib6_tables_init(struct net *net)
268 {
269         fib6_link_table(net, net->ipv6.fib6_main_tbl);
270 }
271 
272 #endif
273 
274 static int fib6_dump_node(struct fib6_walker *w)
275 {
276         int res;
277         struct rt6_info *rt;
278 
279         for (rt = w->leaf; rt; rt = rt->dst.rt6_next) {
280                 res = rt6_dump_route(rt, w->args);
281                 if (res < 0) {
282                         /* Frame is full, suspend walking */
283                         w->leaf = rt;
284                         return 1;
285                 }
286                 WARN_ON(res == 0);
287         }
288         w->leaf = NULL;
289         return 0;
290 }
291 
292 static void fib6_dump_end(struct netlink_callback *cb)
293 {
294         struct fib6_walker *w = (void *)cb->args[2];
295 
296         if (w) {
297                 if (cb->args[4]) {
298                         cb->args[4] = 0;
299                         fib6_walker_unlink(w);
300                 }
301                 cb->args[2] = 0;
302                 kfree(w);
303         }
304         cb->done = (void *)cb->args[3];
305         cb->args[1] = 3;
306 }
307 
308 static int fib6_dump_done(struct netlink_callback *cb)
309 {
310         fib6_dump_end(cb);
311         return cb->done ? cb->done(cb) : 0;
312 }
313 
314 static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
315                            struct netlink_callback *cb)
316 {
317         struct fib6_walker *w;
318         int res;
319 
320         w = (void *)cb->args[2];
321         w->root = &table->tb6_root;
322 
323         if (cb->args[4] == 0) {
324                 w->count = 0;
325                 w->skip = 0;
326 
327                 read_lock_bh(&table->tb6_lock);
328                 res = fib6_walk(w);
329                 read_unlock_bh(&table->tb6_lock);
330                 if (res > 0) {
331                         cb->args[4] = 1;
332                         cb->args[5] = w->root->fn_sernum;
333                 }
334         } else {
335                 if (cb->args[5] != w->root->fn_sernum) {
336                         /* Begin at the root if the tree changed */
337                         cb->args[5] = w->root->fn_sernum;
338                         w->state = FWS_INIT;
339                         w->node = w->root;
340                         w->skip = w->count;
341                 } else
342                         w->skip = 0;
343 
344                 read_lock_bh(&table->tb6_lock);
345                 res = fib6_walk_continue(w);
346                 read_unlock_bh(&table->tb6_lock);
347                 if (res <= 0) {
348                         fib6_walker_unlink(w);
349                         cb->args[4] = 0;
350                 }
351         }
352 
353         return res;
354 }
355 
356 static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
357 {
358         struct net *net = sock_net(skb->sk);
359         unsigned int h, s_h;
360         unsigned int e = 0, s_e;
361         struct rt6_rtnl_dump_arg arg;
362         struct fib6_walker *w;
363         struct fib6_table *tb;
364         struct hlist_head *head;
365         int res = 0;
366 
367         s_h = cb->args[0];
368         s_e = cb->args[1];
369 
370         w = (void *)cb->args[2];
371         if (!w) {
372                 /* New dump:
373                  *
374                  * 1. hook callback destructor.
375                  */
376                 cb->args[3] = (long)cb->done;
377                 cb->done = fib6_dump_done;
378 
379                 /*
380                  * 2. allocate and initialize walker.
381                  */
382                 w = kzalloc(sizeof(*w), GFP_ATOMIC);
383                 if (!w)
384                         return -ENOMEM;
385                 w->func = fib6_dump_node;
386                 cb->args[2] = (long)w;
387         }
388 
389         arg.skb = skb;
390         arg.cb = cb;
391         arg.net = net;
392         w->args = &arg;
393 
394         rcu_read_lock();
395         for (h = s_h; h < FIB6_TABLE_HASHSZ; h++, s_e = 0) {
396                 e = 0;
397                 head = &net->ipv6.fib_table_hash[h];
398                 hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
399                         if (e < s_e)
400                                 goto next;
401                         res = fib6_dump_table(tb, skb, cb);
402                         if (res != 0)
403                                 goto out;
404 next:
405                         e++;
406                 }
407         }
408 out:
409         rcu_read_unlock();
410         cb->args[1] = e;
411         cb->args[0] = h;
412 
413         res = res < 0 ? res : skb->len;
414         if (res <= 0)
415                 fib6_dump_end(cb);
416         return res;
417 }
418 
419 /*
420  *      Routing Table
421  *
422  *      return the appropriate node for a routing tree "add" operation
423  *      by either creating and inserting or by returning an existing
424  *      node.
425  */
426 
427 static struct fib6_node *fib6_add_1(struct fib6_node *root,
428                                      struct in6_addr *addr, int plen,
429                                      int offset, int allow_create,
430                                      int replace_required, int sernum)
431 {
432         struct fib6_node *fn, *in, *ln;
433         struct fib6_node *pn = NULL;
434         struct rt6key *key;
435         int     bit;
436         __be32  dir = 0;
437 
438         RT6_TRACE("fib6_add_1\n");
439 
440         /* insert node in tree */
441 
442         fn = root;
443 
444         do {
445                 key = (struct rt6key *)((u8 *)fn->leaf + offset);
446 
447                 /*
448                  *      Prefix match
449                  */
450                 if (plen < fn->fn_bit ||
451                     !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) {
452                         if (!allow_create) {
453                                 if (replace_required) {
454                                         pr_warn("Can't replace route, no match found\n");
455                                         return ERR_PTR(-ENOENT);
456                                 }
457                                 pr_warn("NLM_F_CREATE should be set when creating new route\n");
458                         }
459                         goto insert_above;
460                 }
461 
462                 /*
463                  *      Exact match ?
464                  */
465 
466                 if (plen == fn->fn_bit) {
467                         /* clean up an intermediate node */
468                         if (!(fn->fn_flags & RTN_RTINFO)) {
469                                 rt6_release(fn->leaf);
470                                 fn->leaf = NULL;
471                         }
472 
473                         fn->fn_sernum = sernum;
474 
475                         return fn;
476                 }
477 
478                 /*
479                  *      We have more bits to go
480                  */
481 
482                 /* Try to walk down on tree. */
483                 fn->fn_sernum = sernum;
484                 dir = addr_bit_set(addr, fn->fn_bit);
485                 pn = fn;
486                 fn = dir ? fn->right : fn->left;
487         } while (fn);
488 
489         if (!allow_create) {
490                 /* We should not create new node because
491                  * NLM_F_REPLACE was specified without NLM_F_CREATE
492                  * I assume it is safe to require NLM_F_CREATE when
493                  * REPLACE flag is used! Later we may want to remove the
494                  * check for replace_required, because according
495                  * to netlink specification, NLM_F_CREATE
496                  * MUST be specified if new route is created.
497                  * That would keep IPv6 consistent with IPv4
498                  */
499                 if (replace_required) {
500                         pr_warn("Can't replace route, no match found\n");
501                         return ERR_PTR(-ENOENT);
502                 }
503                 pr_warn("NLM_F_CREATE should be set when creating new route\n");
504         }
505         /*
506          *      We walked to the bottom of tree.
507          *      Create new leaf node without children.
508          */
509 
510         ln = node_alloc();
511 
512         if (!ln)
513                 return ERR_PTR(-ENOMEM);
514         ln->fn_bit = plen;
515 
516         ln->parent = pn;
517         ln->fn_sernum = sernum;
518 
519         if (dir)
520                 pn->right = ln;
521         else
522                 pn->left  = ln;
523 
524         return ln;
525 
526 
527 insert_above:
528         /*
529          * split since we don't have a common prefix anymore or
530          * we have a less significant route.
531          * we've to insert an intermediate node on the list
532          * this new node will point to the one we need to create
533          * and the current
534          */
535 
536         pn = fn->parent;
537 
538         /* find 1st bit in difference between the 2 addrs.
539 
540            See comment in __ipv6_addr_diff: bit may be an invalid value,
541            but if it is >= plen, the value is ignored in any case.
542          */
543 
544         bit = __ipv6_addr_diff(addr, &key->addr, sizeof(*addr));
545 
546         /*
547          *              (intermediate)[in]
548          *                /        \
549          *      (new leaf node)[ln] (old node)[fn]
550          */
551         if (plen > bit) {
552                 in = node_alloc();
553                 ln = node_alloc();
554 
555                 if (!in || !ln) {
556                         if (in)
557                                 node_free(in);
558                         if (ln)
559                                 node_free(ln);
560                         return ERR_PTR(-ENOMEM);
561                 }
562 
563                 /*
564                  * new intermediate node.
565                  * RTN_RTINFO will
566                  * be off since that an address that chooses one of
567                  * the branches would not match less specific routes
568                  * in the other branch
569                  */
570 
571                 in->fn_bit = bit;
572 
573                 in->parent = pn;
574                 in->leaf = fn->leaf;
575                 atomic_inc(&in->leaf->rt6i_ref);
576 
577                 in->fn_sernum = sernum;
578 
579                 /* update parent pointer */
580                 if (dir)
581                         pn->right = in;
582                 else
583                         pn->left  = in;
584 
585                 ln->fn_bit = plen;
586 
587                 ln->parent = in;
588                 fn->parent = in;
589 
590                 ln->fn_sernum = sernum;
591 
592                 if (addr_bit_set(addr, bit)) {
593                         in->right = ln;
594                         in->left  = fn;
595                 } else {
596                         in->left  = ln;
597                         in->right = fn;
598                 }
599         } else { /* plen <= bit */
600 
601                 /*
602                  *              (new leaf node)[ln]
603                  *                /        \
604                  *           (old node)[fn] NULL
605                  */
606 
607                 ln = node_alloc();
608 
609                 if (!ln)
610                         return ERR_PTR(-ENOMEM);
611 
612                 ln->fn_bit = plen;
613 
614                 ln->parent = pn;
615 
616                 ln->fn_sernum = sernum;
617 
618                 if (dir)
619                         pn->right = ln;
620                 else
621                         pn->left  = ln;
622 
623                 if (addr_bit_set(&key->addr, plen))
624                         ln->right = fn;
625                 else
626                         ln->left  = fn;
627 
628                 fn->parent = ln;
629         }
630         return ln;
631 }
632 
633 static bool rt6_qualify_for_ecmp(struct rt6_info *rt)
634 {
635         return (rt->rt6i_flags & (RTF_GATEWAY|RTF_ADDRCONF|RTF_DYNAMIC)) ==
636                RTF_GATEWAY;
637 }
638 
639 static int fib6_commit_metrics(struct dst_entry *dst,
640                                struct nlattr *mx, int mx_len)
641 {
642         struct nlattr *nla;
643         int remaining;
644         u32 *mp;
645 
646         if (dst->flags & DST_HOST) {
647                 mp = dst_metrics_write_ptr(dst);
648         } else {
649                 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_ATOMIC);
650                 if (!mp)
651                         return -ENOMEM;
652                 dst_init_metrics(dst, mp, 0);
653         }
654 
655         nla_for_each_attr(nla, mx, mx_len, remaining) {
656                 int type = nla_type(nla);
657 
658                 if (type) {
659                         if (type > RTAX_MAX)
660                                 return -EINVAL;
661 
662                         mp[type - 1] = nla_get_u32(nla);
663                 }
664         }
665         return 0;
666 }
667 
668 static void fib6_purge_rt(struct rt6_info *rt, struct fib6_node *fn,
669                           struct net *net)
670 {
671         if (atomic_read(&rt->rt6i_ref) != 1) {
672                 /* This route is used as dummy address holder in some split
673                  * nodes. It is not leaked, but it still holds other resources,
674                  * which must be released in time. So, scan ascendant nodes
675                  * and replace dummy references to this route with references
676                  * to still alive ones.
677                  */
678                 while (fn) {
679                         if (!(fn->fn_flags & RTN_RTINFO) && fn->leaf == rt) {
680                                 fn->leaf = fib6_find_prefix(net, fn);
681                                 atomic_inc(&fn->leaf->rt6i_ref);
682                                 rt6_release(rt);
683                         }
684                         fn = fn->parent;
685                 }
686                 /* No more references are possible at this point. */
687                 BUG_ON(atomic_read(&rt->rt6i_ref) != 1);
688         }
689 }
690 
691 /*
692  *      Insert routing information in a node.
693  */
694 
695 static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt,
696                             struct nl_info *info, struct nlattr *mx, int mx_len)
697 {
698         struct rt6_info *iter = NULL;
699         struct rt6_info **ins;
700         int replace = (info->nlh &&
701                        (info->nlh->nlmsg_flags & NLM_F_REPLACE));
702         int add = (!info->nlh ||
703                    (info->nlh->nlmsg_flags & NLM_F_CREATE));
704         int found = 0;
705         bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
706         int err;
707 
708         ins = &fn->leaf;
709 
710         for (iter = fn->leaf; iter; iter = iter->dst.rt6_next) {
711                 /*
712                  *      Search for duplicates
713                  */
714 
715                 if (iter->rt6i_metric == rt->rt6i_metric) {
716                         /*
717                          *      Same priority level
718                          */
719                         if (info->nlh &&
720                             (info->nlh->nlmsg_flags & NLM_F_EXCL))
721                                 return -EEXIST;
722                         if (replace) {
723                                 found++;
724                                 break;
725                         }
726 
727                         if (iter->dst.dev == rt->dst.dev &&
728                             iter->rt6i_idev == rt->rt6i_idev &&
729                             ipv6_addr_equal(&iter->rt6i_gateway,
730                                             &rt->rt6i_gateway)) {
731                                 if (rt->rt6i_nsiblings)
732                                         rt->rt6i_nsiblings = 0;
733                                 if (!(iter->rt6i_flags & RTF_EXPIRES))
734                                         return -EEXIST;
735                                 if (!(rt->rt6i_flags & RTF_EXPIRES))
736                                         rt6_clean_expires(iter);
737                                 else
738                                         rt6_set_expires(iter, rt->dst.expires);
739                                 return -EEXIST;
740                         }
741                         /* If we have the same destination and the same metric,
742                          * but not the same gateway, then the route we try to
743                          * add is sibling to this route, increment our counter
744                          * of siblings, and later we will add our route to the
745                          * list.
746                          * Only static routes (which don't have flag
747                          * RTF_EXPIRES) are used for ECMPv6.
748                          *
749                          * To avoid long list, we only had siblings if the
750                          * route have a gateway.
751                          */
752                         if (rt_can_ecmp &&
753                             rt6_qualify_for_ecmp(iter))
754                                 rt->rt6i_nsiblings++;
755                 }
756 
757                 if (iter->rt6i_metric > rt->rt6i_metric)
758                         break;
759 
760                 ins = &iter->dst.rt6_next;
761         }
762 
763         /* Reset round-robin state, if necessary */
764         if (ins == &fn->leaf)
765                 fn->rr_ptr = NULL;
766 
767         /* Link this route to others same route. */
768         if (rt->rt6i_nsiblings) {
769                 unsigned int rt6i_nsiblings;
770                 struct rt6_info *sibling, *temp_sibling;
771 
772                 /* Find the first route that have the same metric */
773                 sibling = fn->leaf;
774                 while (sibling) {
775                         if (sibling->rt6i_metric == rt->rt6i_metric &&
776                             rt6_qualify_for_ecmp(sibling)) {
777                                 list_add_tail(&rt->rt6i_siblings,
778                                               &sibling->rt6i_siblings);
779                                 break;
780                         }
781                         sibling = sibling->dst.rt6_next;
782                 }
783                 /* For each sibling in the list, increment the counter of
784                  * siblings. BUG() if counters does not match, list of siblings
785                  * is broken!
786                  */
787                 rt6i_nsiblings = 0;
788                 list_for_each_entry_safe(sibling, temp_sibling,
789                                          &rt->rt6i_siblings, rt6i_siblings) {
790                         sibling->rt6i_nsiblings++;
791                         BUG_ON(sibling->rt6i_nsiblings != rt->rt6i_nsiblings);
792                         rt6i_nsiblings++;
793                 }
794                 BUG_ON(rt6i_nsiblings != rt->rt6i_nsiblings);
795         }
796 
797         /*
798          *      insert node
799          */
800         if (!replace) {
801                 if (!add)
802                         pr_warn("NLM_F_CREATE should be set when creating new route\n");
803 
804 add:
805                 if (mx) {
806                         err = fib6_commit_metrics(&rt->dst, mx, mx_len);
807                         if (err)
808                                 return err;
809                 }
810                 rt->dst.rt6_next = iter;
811                 *ins = rt;
812                 rt->rt6i_node = fn;
813                 atomic_inc(&rt->rt6i_ref);
814                 inet6_rt_notify(RTM_NEWROUTE, rt, info);
815                 info->nl_net->ipv6.rt6_stats->fib_rt_entries++;
816 
817                 if (!(fn->fn_flags & RTN_RTINFO)) {
818                         info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
819                         fn->fn_flags |= RTN_RTINFO;
820                 }
821 
822         } else {
823                 if (!found) {
824                         if (add)
825                                 goto add;
826                         pr_warn("NLM_F_REPLACE set, but no existing node found!\n");
827                         return -ENOENT;
828                 }
829                 if (mx) {
830                         err = fib6_commit_metrics(&rt->dst, mx, mx_len);
831                         if (err)
832                                 return err;
833                 }
834                 *ins = rt;
835                 rt->rt6i_node = fn;
836                 rt->dst.rt6_next = iter->dst.rt6_next;
837                 atomic_inc(&rt->rt6i_ref);
838                 inet6_rt_notify(RTM_NEWROUTE, rt, info);
839                 if (!(fn->fn_flags & RTN_RTINFO)) {
840                         info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
841                         fn->fn_flags |= RTN_RTINFO;
842                 }
843                 fib6_purge_rt(iter, fn, info->nl_net);
844                 rt6_release(iter);
845         }
846 
847         return 0;
848 }
849 
850 static void fib6_start_gc(struct net *net, struct rt6_info *rt)
851 {
852         if (!timer_pending(&net->ipv6.ip6_fib_timer) &&
853             (rt->rt6i_flags & (RTF_EXPIRES | RTF_CACHE)))
854                 mod_timer(&net->ipv6.ip6_fib_timer,
855                           jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
856 }
857 
858 void fib6_force_start_gc(struct net *net)
859 {
860         if (!timer_pending(&net->ipv6.ip6_fib_timer))
861                 mod_timer(&net->ipv6.ip6_fib_timer,
862                           jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
863 }
864 
865 /*
866  *      Add routing information to the routing tree.
867  *      <destination addr>/<source addr>
868  *      with source addr info in sub-trees
869  */
870 
871 int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info,
872              struct nlattr *mx, int mx_len)
873 {
874         struct fib6_node *fn, *pn = NULL;
875         int err = -ENOMEM;
876         int allow_create = 1;
877         int replace_required = 0;
878         int sernum = fib6_new_sernum(info->nl_net);
879 
880         if (info->nlh) {
881                 if (!(info->nlh->nlmsg_flags & NLM_F_CREATE))
882                         allow_create = 0;
883                 if (info->nlh->nlmsg_flags & NLM_F_REPLACE)
884                         replace_required = 1;
885         }
886         if (!allow_create && !replace_required)
887                 pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n");
888 
889         fn = fib6_add_1(root, &rt->rt6i_dst.addr, rt->rt6i_dst.plen,
890                         offsetof(struct rt6_info, rt6i_dst), allow_create,
891                         replace_required, sernum);
892         if (IS_ERR(fn)) {
893                 err = PTR_ERR(fn);
894                 fn = NULL;
895                 goto out;
896         }
897 
898         pn = fn;
899 
900 #ifdef CONFIG_IPV6_SUBTREES
901         if (rt->rt6i_src.plen) {
902                 struct fib6_node *sn;
903 
904                 if (!fn->subtree) {
905                         struct fib6_node *sfn;
906 
907                         /*
908                          * Create subtree.
909                          *
910                          *              fn[main tree]
911                          *              |
912                          *              sfn[subtree root]
913                          *                 \
914                          *                  sn[new leaf node]
915                          */
916 
917                         /* Create subtree root node */
918                         sfn = node_alloc();
919                         if (!sfn)
920                                 goto st_failure;
921 
922                         sfn->leaf = info->nl_net->ipv6.ip6_null_entry;
923                         atomic_inc(&info->nl_net->ipv6.ip6_null_entry->rt6i_ref);
924                         sfn->fn_flags = RTN_ROOT;
925                         sfn->fn_sernum = sernum;
926 
927                         /* Now add the first leaf node to new subtree */
928 
929                         sn = fib6_add_1(sfn, &rt->rt6i_src.addr,
930                                         rt->rt6i_src.plen,
931                                         offsetof(struct rt6_info, rt6i_src),
932                                         allow_create, replace_required, sernum);
933 
934                         if (IS_ERR(sn)) {
935                                 /* If it is failed, discard just allocated
936                                    root, and then (in st_failure) stale node
937                                    in main tree.
938                                  */
939                                 node_free(sfn);
940                                 err = PTR_ERR(sn);
941                                 goto st_failure;
942                         }
943 
944                         /* Now link new subtree to main tree */
945                         sfn->parent = fn;
946                         fn->subtree = sfn;
947                 } else {
948                         sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr,
949                                         rt->rt6i_src.plen,
950                                         offsetof(struct rt6_info, rt6i_src),
951                                         allow_create, replace_required, sernum);
952 
953                         if (IS_ERR(sn)) {
954                                 err = PTR_ERR(sn);
955                                 goto st_failure;
956                         }
957                 }
958 
959                 if (!fn->leaf) {
960                         fn->leaf = rt;
961                         atomic_inc(&rt->rt6i_ref);
962                 }
963                 fn = sn;
964         }
965 #endif
966 
967         err = fib6_add_rt2node(fn, rt, info, mx, mx_len);
968         if (!err) {
969                 fib6_start_gc(info->nl_net, rt);
970                 if (!(rt->rt6i_flags & RTF_CACHE))
971                         fib6_prune_clones(info->nl_net, pn);
972         }
973 
974 out:
975         if (err) {
976 #ifdef CONFIG_IPV6_SUBTREES
977                 /*
978                  * If fib6_add_1 has cleared the old leaf pointer in the
979                  * super-tree leaf node we have to find a new one for it.
980                  */
981                 if (pn != fn && pn->leaf == rt) {
982                         pn->leaf = NULL;
983                         atomic_dec(&rt->rt6i_ref);
984                 }
985                 if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) {
986                         pn->leaf = fib6_find_prefix(info->nl_net, pn);
987 #if RT6_DEBUG >= 2
988                         if (!pn->leaf) {
989                                 WARN_ON(pn->leaf == NULL);
990                                 pn->leaf = info->nl_net->ipv6.ip6_null_entry;
991                         }
992 #endif
993                         atomic_inc(&pn->leaf->rt6i_ref);
994                 }
995 #endif
996                 dst_free(&rt->dst);
997         }
998         return err;
999 
1000 #ifdef CONFIG_IPV6_SUBTREES
1001         /* Subtree creation failed, probably main tree node
1002            is orphan. If it is, shoot it.
1003          */
1004 st_failure:
1005         if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)))
1006                 fib6_repair_tree(info->nl_net, fn);
1007         dst_free(&rt->dst);
1008         return err;
1009 #endif
1010 }
1011 
1012 /*
1013  *      Routing tree lookup
1014  *
1015  */
1016 
1017 struct lookup_args {
1018         int                     offset;         /* key offset on rt6_info       */
1019         const struct in6_addr   *addr;          /* search key                   */
1020 };
1021 
1022 static struct fib6_node *fib6_lookup_1(struct fib6_node *root,
1023                                        struct lookup_args *args)
1024 {
1025         struct fib6_node *fn;
1026         __be32 dir;
1027 
1028         if (unlikely(args->offset == 0))
1029                 return NULL;
1030 
1031         /*
1032          *      Descend on a tree
1033          */
1034 
1035         fn = root;
1036 
1037         for (;;) {
1038                 struct fib6_node *next;
1039 
1040                 dir = addr_bit_set(args->addr, fn->fn_bit);
1041 
1042                 next = dir ? fn->right : fn->left;
1043 
1044                 if (next) {
1045                         fn = next;
1046                         continue;
1047                 }
1048                 break;
1049         }
1050 
1051         while (fn) {
1052                 if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) {
1053                         struct rt6key *key;
1054 
1055                         key = (struct rt6key *) ((u8 *) fn->leaf +
1056                                                  args->offset);
1057 
1058                         if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
1059 #ifdef CONFIG_IPV6_SUBTREES
1060                                 if (fn->subtree) {
1061                                         struct fib6_node *sfn;
1062                                         sfn = fib6_lookup_1(fn->subtree,
1063                                                             args + 1);
1064                                         if (!sfn)
1065                                                 goto backtrack;
1066                                         fn = sfn;
1067                                 }
1068 #endif
1069                                 if (fn->fn_flags & RTN_RTINFO)
1070                                         return fn;
1071                         }
1072                 }
1073 #ifdef CONFIG_IPV6_SUBTREES
1074 backtrack:
1075 #endif
1076                 if (fn->fn_flags & RTN_ROOT)
1077                         break;
1078 
1079                 fn = fn->parent;
1080         }
1081 
1082         return NULL;
1083 }
1084 
1085 struct fib6_node *fib6_lookup(struct fib6_node *root, const struct in6_addr *daddr,
1086                               const struct in6_addr *saddr)
1087 {
1088         struct fib6_node *fn;
1089         struct lookup_args args[] = {
1090                 {
1091                         .offset = offsetof(struct rt6_info, rt6i_dst),
1092                         .addr = daddr,
1093                 },
1094 #ifdef CONFIG_IPV6_SUBTREES
1095                 {
1096                         .offset = offsetof(struct rt6_info, rt6i_src),
1097                         .addr = saddr,
1098                 },
1099 #endif
1100                 {
1101                         .offset = 0,    /* sentinel */
1102                 }
1103         };
1104 
1105         fn = fib6_lookup_1(root, daddr ? args : args + 1);
1106         if (!fn || fn->fn_flags & RTN_TL_ROOT)
1107                 fn = root;
1108 
1109         return fn;
1110 }
1111 
1112 /*
1113  *      Get node with specified destination prefix (and source prefix,
1114  *      if subtrees are used)
1115  */
1116 
1117 
1118 static struct fib6_node *fib6_locate_1(struct fib6_node *root,
1119                                        const struct in6_addr *addr,
1120                                        int plen, int offset)
1121 {
1122         struct fib6_node *fn;
1123 
1124         for (fn = root; fn ; ) {
1125                 struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset);
1126 
1127                 /*
1128                  *      Prefix match
1129                  */
1130                 if (plen < fn->fn_bit ||
1131                     !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
1132                         return NULL;
1133 
1134                 if (plen == fn->fn_bit)
1135                         return fn;
1136 
1137                 /*
1138                  *      We have more bits to go
1139                  */
1140                 if (addr_bit_set(addr, fn->fn_bit))
1141                         fn = fn->right;
1142                 else
1143                         fn = fn->left;
1144         }
1145         return NULL;
1146 }
1147 
1148 struct fib6_node *fib6_locate(struct fib6_node *root,
1149                               const struct in6_addr *daddr, int dst_len,
1150                               const struct in6_addr *saddr, int src_len)
1151 {
1152         struct fib6_node *fn;
1153 
1154         fn = fib6_locate_1(root, daddr, dst_len,
1155                            offsetof(struct rt6_info, rt6i_dst));
1156 
1157 #ifdef CONFIG_IPV6_SUBTREES
1158         if (src_len) {
1159                 WARN_ON(saddr == NULL);
1160                 if (fn && fn->subtree)
1161                         fn = fib6_locate_1(fn->subtree, saddr, src_len,
1162                                            offsetof(struct rt6_info, rt6i_src));
1163         }
1164 #endif
1165 
1166         if (fn && fn->fn_flags & RTN_RTINFO)
1167                 return fn;
1168 
1169         return NULL;
1170 }
1171 
1172 
1173 /*
1174  *      Deletion
1175  *
1176  */
1177 
1178 static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn)
1179 {
1180         if (fn->fn_flags & RTN_ROOT)
1181                 return net->ipv6.ip6_null_entry;
1182 
1183         while (fn) {
1184                 if (fn->left)
1185                         return fn->left->leaf;
1186                 if (fn->right)
1187                         return fn->right->leaf;
1188 
1189                 fn = FIB6_SUBTREE(fn);
1190         }
1191         return NULL;
1192 }
1193 
1194 /*
1195  *      Called to trim the tree of intermediate nodes when possible. "fn"
1196  *      is the node we want to try and remove.
1197  */
1198 
1199 static struct fib6_node *fib6_repair_tree(struct net *net,
1200                                            struct fib6_node *fn)
1201 {
1202         int children;
1203         int nstate;
1204         struct fib6_node *child, *pn;
1205         struct fib6_walker *w;
1206         int iter = 0;
1207 
1208         for (;;) {
1209                 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter);
1210                 iter++;
1211 
1212                 WARN_ON(fn->fn_flags & RTN_RTINFO);
1213                 WARN_ON(fn->fn_flags & RTN_TL_ROOT);
1214                 WARN_ON(fn->leaf != NULL);
1215 
1216                 children = 0;
1217                 child = NULL;
1218                 if (fn->right)
1219                         child = fn->right, children |= 1;
1220                 if (fn->left)
1221                         child = fn->left, children |= 2;
1222 
1223                 if (children == 3 || FIB6_SUBTREE(fn)
1224 #ifdef CONFIG_IPV6_SUBTREES
1225                     /* Subtree root (i.e. fn) may have one child */
1226                     || (children && fn->fn_flags & RTN_ROOT)
1227 #endif
1228                     ) {
1229                         fn->leaf = fib6_find_prefix(net, fn);
1230 #if RT6_DEBUG >= 2
1231                         if (!fn->leaf) {
1232                                 WARN_ON(!fn->leaf);
1233                                 fn->leaf = net->ipv6.ip6_null_entry;
1234                         }
1235 #endif
1236                         atomic_inc(&fn->leaf->rt6i_ref);
1237                         return fn->parent;
1238                 }
1239 
1240                 pn = fn->parent;
1241 #ifdef CONFIG_IPV6_SUBTREES
1242                 if (FIB6_SUBTREE(pn) == fn) {
1243                         WARN_ON(!(fn->fn_flags & RTN_ROOT));
1244                         FIB6_SUBTREE(pn) = NULL;
1245                         nstate = FWS_L;
1246                 } else {
1247                         WARN_ON(fn->fn_flags & RTN_ROOT);
1248 #endif
1249                         if (pn->right == fn)
1250                                 pn->right = child;
1251                         else if (pn->left == fn)
1252                                 pn->left = child;
1253 #if RT6_DEBUG >= 2
1254                         else
1255                                 WARN_ON(1);
1256 #endif
1257                         if (child)
1258                                 child->parent = pn;
1259                         nstate = FWS_R;
1260 #ifdef CONFIG_IPV6_SUBTREES
1261                 }
1262 #endif
1263 
1264                 read_lock(&fib6_walker_lock);
1265                 FOR_WALKERS(w) {
1266                         if (!child) {
1267                                 if (w->root == fn) {
1268                                         w->root = w->node = NULL;
1269                                         RT6_TRACE("W %p adjusted by delroot 1\n", w);
1270                                 } else if (w->node == fn) {
1271                                         RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate);
1272                                         w->node = pn;
1273                                         w->state = nstate;
1274                                 }
1275                         } else {
1276                                 if (w->root == fn) {
1277                                         w->root = child;
1278                                         RT6_TRACE("W %p adjusted by delroot 2\n", w);
1279                                 }
1280                                 if (w->node == fn) {
1281                                         w->node = child;
1282                                         if (children&2) {
1283                                                 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1284                                                 w->state = w->state >= FWS_R ? FWS_U : FWS_INIT;
1285                                         } else {
1286                                                 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1287                                                 w->state = w->state >= FWS_C ? FWS_U : FWS_INIT;
1288                                         }
1289                                 }
1290                         }
1291                 }
1292                 read_unlock(&fib6_walker_lock);
1293 
1294                 node_free(fn);
1295                 if (pn->fn_flags & RTN_RTINFO || FIB6_SUBTREE(pn))
1296                         return pn;
1297 
1298                 rt6_release(pn->leaf);
1299                 pn->leaf = NULL;
1300                 fn = pn;
1301         }
1302 }
1303 
1304 static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp,
1305                            struct nl_info *info)
1306 {
1307         struct fib6_walker *w;
1308         struct rt6_info *rt = *rtp;
1309         struct net *net = info->nl_net;
1310 
1311         RT6_TRACE("fib6_del_route\n");
1312 
1313         /* Unlink it */
1314         *rtp = rt->dst.rt6_next;
1315         rt->rt6i_node = NULL;
1316         net->ipv6.rt6_stats->fib_rt_entries--;
1317         net->ipv6.rt6_stats->fib_discarded_routes++;
1318 
1319         /* Reset round-robin state, if necessary */
1320         if (fn->rr_ptr == rt)
1321                 fn->rr_ptr = NULL;
1322 
1323         /* Remove this entry from other siblings */
1324         if (rt->rt6i_nsiblings) {
1325                 struct rt6_info *sibling, *next_sibling;
1326 
1327                 list_for_each_entry_safe(sibling, next_sibling,
1328                                          &rt->rt6i_siblings, rt6i_siblings)
1329                         sibling->rt6i_nsiblings--;
1330                 rt->rt6i_nsiblings = 0;
1331                 list_del_init(&rt->rt6i_siblings);
1332         }
1333 
1334         /* Adjust walkers */
1335         read_lock(&fib6_walker_lock);
1336         FOR_WALKERS(w) {
1337                 if (w->state == FWS_C && w->leaf == rt) {
1338                         RT6_TRACE("walker %p adjusted by delroute\n", w);
1339                         w->leaf = rt->dst.rt6_next;
1340                         if (!w->leaf)
1341                                 w->state = FWS_U;
1342                 }
1343         }
1344         read_unlock(&fib6_walker_lock);
1345 
1346         rt->dst.rt6_next = NULL;
1347 
1348         /* If it was last route, expunge its radix tree node */
1349         if (!fn->leaf) {
1350                 fn->fn_flags &= ~RTN_RTINFO;
1351                 net->ipv6.rt6_stats->fib_route_nodes--;
1352                 fn = fib6_repair_tree(net, fn);
1353         }
1354 
1355         fib6_purge_rt(rt, fn, net);
1356 
1357         inet6_rt_notify(RTM_DELROUTE, rt, info);
1358         rt6_release(rt);
1359 }
1360 
1361 int fib6_del(struct rt6_info *rt, struct nl_info *info)
1362 {
1363         struct net *net = info->nl_net;
1364         struct fib6_node *fn = rt->rt6i_node;
1365         struct rt6_info **rtp;
1366 
1367 #if RT6_DEBUG >= 2
1368         if (rt->dst.obsolete > 0) {
1369                 WARN_ON(fn != NULL);
1370                 return -ENOENT;
1371         }
1372 #endif
1373         if (!fn || rt == net->ipv6.ip6_null_entry)
1374                 return -ENOENT;
1375 
1376         WARN_ON(!(fn->fn_flags & RTN_RTINFO));
1377 
1378         if (!(rt->rt6i_flags & RTF_CACHE)) {
1379                 struct fib6_node *pn = fn;
1380 #ifdef CONFIG_IPV6_SUBTREES
1381                 /* clones of this route might be in another subtree */
1382                 if (rt->rt6i_src.plen) {
1383                         while (!(pn->fn_flags & RTN_ROOT))
1384                                 pn = pn->parent;
1385                         pn = pn->parent;
1386                 }
1387 #endif
1388                 fib6_prune_clones(info->nl_net, pn);
1389         }
1390 
1391         /*
1392          *      Walk the leaf entries looking for ourself
1393          */
1394 
1395         for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->dst.rt6_next) {
1396                 if (*rtp == rt) {
1397                         fib6_del_route(fn, rtp, info);
1398                         return 0;
1399                 }
1400         }
1401         return -ENOENT;
1402 }
1403 
1404 /*
1405  *      Tree traversal function.
1406  *
1407  *      Certainly, it is not interrupt safe.
1408  *      However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1409  *      It means, that we can modify tree during walking
1410  *      and use this function for garbage collection, clone pruning,
1411  *      cleaning tree when a device goes down etc. etc.
1412  *
1413  *      It guarantees that every node will be traversed,
1414  *      and that it will be traversed only once.
1415  *
1416  *      Callback function w->func may return:
1417  *      0 -> continue walking.
1418  *      positive value -> walking is suspended (used by tree dumps,
1419  *      and probably by gc, if it will be split to several slices)
1420  *      negative value -> terminate walking.
1421  *
1422  *      The function itself returns:
1423  *      0   -> walk is complete.
1424  *      >0  -> walk is incomplete (i.e. suspended)
1425  *      <0  -> walk is terminated by an error.
1426  */
1427 
1428 static int fib6_walk_continue(struct fib6_walker *w)
1429 {
1430         struct fib6_node *fn, *pn;
1431 
1432         for (;;) {
1433                 fn = w->node;
1434                 if (!fn)
1435                         return 0;
1436 
1437                 if (w->prune && fn != w->root &&
1438                     fn->fn_flags & RTN_RTINFO && w->state < FWS_C) {
1439                         w->state = FWS_C;
1440                         w->leaf = fn->leaf;
1441                 }
1442                 switch (w->state) {
1443 #ifdef CONFIG_IPV6_SUBTREES
1444                 case FWS_S:
1445                         if (FIB6_SUBTREE(fn)) {
1446                                 w->node = FIB6_SUBTREE(fn);
1447                                 continue;
1448                         }
1449                         w->state = FWS_L;
1450 #endif
1451                 case FWS_L:
1452                         if (fn->left) {
1453                                 w->node = fn->left;
1454                                 w->state = FWS_INIT;
1455                                 continue;
1456                         }
1457                         w->state = FWS_R;
1458                 case FWS_R:
1459                         if (fn->right) {
1460                                 w->node = fn->right;
1461                                 w->state = FWS_INIT;
1462                                 continue;
1463                         }
1464                         w->state = FWS_C;
1465                         w->leaf = fn->leaf;
1466                 case FWS_C:
1467                         if (w->leaf && fn->fn_flags & RTN_RTINFO) {
1468                                 int err;
1469 
1470                                 if (w->skip) {
1471                                         w->skip--;
1472                                         goto skip;
1473                                 }
1474 
1475                                 err = w->func(w);
1476                                 if (err)
1477                                         return err;
1478 
1479                                 w->count++;
1480                                 continue;
1481                         }
1482 skip:
1483                         w->state = FWS_U;
1484                 case FWS_U:
1485                         if (fn == w->root)
1486                                 return 0;
1487                         pn = fn->parent;
1488                         w->node = pn;
1489 #ifdef CONFIG_IPV6_SUBTREES
1490                         if (FIB6_SUBTREE(pn) == fn) {
1491                                 WARN_ON(!(fn->fn_flags & RTN_ROOT));
1492                                 w->state = FWS_L;
1493                                 continue;
1494                         }
1495 #endif
1496                         if (pn->left == fn) {
1497                                 w->state = FWS_R;
1498                                 continue;
1499                         }
1500                         if (pn->right == fn) {
1501                                 w->state = FWS_C;
1502                                 w->leaf = w->node->leaf;
1503                                 continue;
1504                         }
1505 #if RT6_DEBUG >= 2
1506                         WARN_ON(1);
1507 #endif
1508                 }
1509         }
1510 }
1511 
1512 static int fib6_walk(struct fib6_walker *w)
1513 {
1514         int res;
1515 
1516         w->state = FWS_INIT;
1517         w->node = w->root;
1518 
1519         fib6_walker_link(w);
1520         res = fib6_walk_continue(w);
1521         if (res <= 0)
1522                 fib6_walker_unlink(w);
1523         return res;
1524 }
1525 
1526 static int fib6_clean_node(struct fib6_walker *w)
1527 {
1528         int res;
1529         struct rt6_info *rt;
1530         struct fib6_cleaner *c = container_of(w, struct fib6_cleaner, w);
1531         struct nl_info info = {
1532                 .nl_net = c->net,
1533         };
1534 
1535         if (c->sernum != FIB6_NO_SERNUM_CHANGE &&
1536             w->node->fn_sernum != c->sernum)
1537                 w->node->fn_sernum = c->sernum;
1538 
1539         if (!c->func) {
1540                 WARN_ON_ONCE(c->sernum == FIB6_NO_SERNUM_CHANGE);
1541                 w->leaf = NULL;
1542                 return 0;
1543         }
1544 
1545         for (rt = w->leaf; rt; rt = rt->dst.rt6_next) {
1546                 res = c->func(rt, c->arg);
1547                 if (res < 0) {
1548                         w->leaf = rt;
1549                         res = fib6_del(rt, &info);
1550                         if (res) {
1551 #if RT6_DEBUG >= 2
1552                                 pr_debug("%s: del failed: rt=%p@%p err=%d\n",
1553                                          __func__, rt, rt->rt6i_node, res);
1554 #endif
1555                                 continue;
1556                         }
1557                         return 0;
1558                 }
1559                 WARN_ON(res != 0);
1560         }
1561         w->leaf = rt;
1562         return 0;
1563 }
1564 
1565 /*
1566  *      Convenient frontend to tree walker.
1567  *
1568  *      func is called on each route.
1569  *              It may return -1 -> delete this route.
1570  *                            0  -> continue walking
1571  *
1572  *      prune==1 -> only immediate children of node (certainly,
1573  *      ignoring pure split nodes) will be scanned.
1574  */
1575 
1576 static void fib6_clean_tree(struct net *net, struct fib6_node *root,
1577                             int (*func)(struct rt6_info *, void *arg),
1578                             bool prune, int sernum, void *arg)
1579 {
1580         struct fib6_cleaner c;
1581 
1582         c.w.root = root;
1583         c.w.func = fib6_clean_node;
1584         c.w.prune = prune;
1585         c.w.count = 0;
1586         c.w.skip = 0;
1587         c.func = func;
1588         c.sernum = sernum;
1589         c.arg = arg;
1590         c.net = net;
1591 
1592         fib6_walk(&c.w);
1593 }
1594 
1595 static void __fib6_clean_all(struct net *net,
1596                              int (*func)(struct rt6_info *, void *),
1597                              int sernum, void *arg)
1598 {
1599         struct fib6_table *table;
1600         struct hlist_head *head;
1601         unsigned int h;
1602 
1603         rcu_read_lock();
1604         for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
1605                 head = &net->ipv6.fib_table_hash[h];
1606                 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
1607                         write_lock_bh(&table->tb6_lock);
1608                         fib6_clean_tree(net, &table->tb6_root,
1609                                         func, false, sernum, arg);
1610                         write_unlock_bh(&table->tb6_lock);
1611                 }
1612         }
1613         rcu_read_unlock();
1614 }
1615 
1616 void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *),
1617                     void *arg)
1618 {
1619         __fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg);
1620 }
1621 
1622 static int fib6_prune_clone(struct rt6_info *rt, void *arg)
1623 {
1624         if (rt->rt6i_flags & RTF_CACHE) {
1625                 RT6_TRACE("pruning clone %p\n", rt);
1626                 return -1;
1627         }
1628 
1629         return 0;
1630 }
1631 
1632 static void fib6_prune_clones(struct net *net, struct fib6_node *fn)
1633 {
1634         fib6_clean_tree(net, fn, fib6_prune_clone, true,
1635                         FIB6_NO_SERNUM_CHANGE, NULL);
1636 }
1637 
1638 static void fib6_flush_trees(struct net *net)
1639 {
1640         int new_sernum = fib6_new_sernum(net);
1641 
1642         __fib6_clean_all(net, NULL, new_sernum, NULL);
1643 }
1644 
1645 /*
1646  *      Garbage collection
1647  */
1648 
1649 static struct fib6_gc_args
1650 {
1651         int                     timeout;
1652         int                     more;
1653 } gc_args;
1654 
1655 static int fib6_age(struct rt6_info *rt, void *arg)
1656 {
1657         unsigned long now = jiffies;
1658 
1659         /*
1660          *      check addrconf expiration here.
1661          *      Routes are expired even if they are in use.
1662          *
1663          *      Also age clones. Note, that clones are aged out
1664          *      only if they are not in use now.
1665          */
1666 
1667         if (rt->rt6i_flags & RTF_EXPIRES && rt->dst.expires) {
1668                 if (time_after(now, rt->dst.expires)) {
1669                         RT6_TRACE("expiring %p\n", rt);
1670                         return -1;
1671                 }
1672                 gc_args.more++;
1673         } else if (rt->rt6i_flags & RTF_CACHE) {
1674                 if (atomic_read(&rt->dst.__refcnt) == 0 &&
1675                     time_after_eq(now, rt->dst.lastuse + gc_args.timeout)) {
1676                         RT6_TRACE("aging clone %p\n", rt);
1677                         return -1;
1678                 } else if (rt->rt6i_flags & RTF_GATEWAY) {
1679                         struct neighbour *neigh;
1680                         __u8 neigh_flags = 0;
1681 
1682                         neigh = dst_neigh_lookup(&rt->dst, &rt->rt6i_gateway);
1683                         if (neigh) {
1684                                 neigh_flags = neigh->flags;
1685                                 neigh_release(neigh);
1686                         }
1687                         if (!(neigh_flags & NTF_ROUTER)) {
1688                                 RT6_TRACE("purging route %p via non-router but gateway\n",
1689                                           rt);
1690                                 return -1;
1691                         }
1692                 }
1693                 gc_args.more++;
1694         }
1695 
1696         return 0;
1697 }
1698 
1699 static DEFINE_SPINLOCK(fib6_gc_lock);
1700 
1701 void fib6_run_gc(unsigned long expires, struct net *net, bool force)
1702 {
1703         unsigned long now;
1704 
1705         if (force) {
1706                 spin_lock_bh(&fib6_gc_lock);
1707         } else if (!spin_trylock_bh(&fib6_gc_lock)) {
1708                 mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ);
1709                 return;
1710         }
1711         gc_args.timeout = expires ? (int)expires :
1712                           net->ipv6.sysctl.ip6_rt_gc_interval;
1713 
1714         gc_args.more = icmp6_dst_gc();
1715 
1716         fib6_clean_all(net, fib6_age, NULL);
1717         now = jiffies;
1718         net->ipv6.ip6_rt_last_gc = now;
1719 
1720         if (gc_args.more)
1721                 mod_timer(&net->ipv6.ip6_fib_timer,
1722                           round_jiffies(now
1723                                         + net->ipv6.sysctl.ip6_rt_gc_interval));
1724         else
1725                 del_timer(&net->ipv6.ip6_fib_timer);
1726         spin_unlock_bh(&fib6_gc_lock);
1727 }
1728 
1729 static void fib6_gc_timer_cb(unsigned long arg)
1730 {
1731         fib6_run_gc(0, (struct net *)arg, true);
1732 }
1733 
1734 static int __net_init fib6_net_init(struct net *net)
1735 {
1736         size_t size = sizeof(struct hlist_head) * FIB6_TABLE_HASHSZ;
1737 
1738         setup_timer(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, (unsigned long)net);
1739 
1740         net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL);
1741         if (!net->ipv6.rt6_stats)
1742                 goto out_timer;
1743 
1744         /* Avoid false sharing : Use at least a full cache line */
1745         size = max_t(size_t, size, L1_CACHE_BYTES);
1746 
1747         net->ipv6.fib_table_hash = kzalloc(size, GFP_KERNEL);
1748         if (!net->ipv6.fib_table_hash)
1749                 goto out_rt6_stats;
1750 
1751         net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl),
1752                                           GFP_KERNEL);
1753         if (!net->ipv6.fib6_main_tbl)
1754                 goto out_fib_table_hash;
1755 
1756         net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN;
1757         net->ipv6.fib6_main_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry;
1758         net->ipv6.fib6_main_tbl->tb6_root.fn_flags =
1759                 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1760         inet_peer_base_init(&net->ipv6.fib6_main_tbl->tb6_peers);
1761 
1762 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1763         net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl),
1764                                            GFP_KERNEL);
1765         if (!net->ipv6.fib6_local_tbl)
1766                 goto out_fib6_main_tbl;
1767         net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL;
1768         net->ipv6.fib6_local_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry;
1769         net->ipv6.fib6_local_tbl->tb6_root.fn_flags =
1770                 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1771         inet_peer_base_init(&net->ipv6.fib6_local_tbl->tb6_peers);
1772 #endif
1773         fib6_tables_init(net);
1774 
1775         return 0;
1776 
1777 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1778 out_fib6_main_tbl:
1779         kfree(net->ipv6.fib6_main_tbl);
1780 #endif
1781 out_fib_table_hash:
1782         kfree(net->ipv6.fib_table_hash);
1783 out_rt6_stats:
1784         kfree(net->ipv6.rt6_stats);
1785 out_timer:
1786         return -ENOMEM;
1787 }
1788 
1789 static void fib6_net_exit(struct net *net)
1790 {
1791         unsigned int i;
1792 
1793         rt6_ifdown(net, NULL);
1794         del_timer_sync(&net->ipv6.ip6_fib_timer);
1795 
1796         for (i = 0; i < FIB6_TABLE_HASHSZ; i++) {
1797                 struct hlist_head *head = &net->ipv6.fib_table_hash[i];
1798                 struct hlist_node *tmp;
1799                 struct fib6_table *tb;
1800 
1801                 hlist_for_each_entry_safe(tb, tmp, head, tb6_hlist) {
1802                         hlist_del(&tb->tb6_hlist);
1803                         fib6_free_table(tb);
1804                 }
1805         }
1806 
1807         kfree(net->ipv6.fib_table_hash);
1808         kfree(net->ipv6.rt6_stats);
1809 }
1810 
1811 static struct pernet_operations fib6_net_ops = {
1812         .init = fib6_net_init,
1813         .exit = fib6_net_exit,
1814 };
1815 
1816 int __init fib6_init(void)
1817 {
1818         int ret = -ENOMEM;
1819 
1820         fib6_node_kmem = kmem_cache_create("fib6_nodes",
1821                                            sizeof(struct fib6_node),
1822                                            0, SLAB_HWCACHE_ALIGN,
1823                                            NULL);
1824         if (!fib6_node_kmem)
1825                 goto out;
1826 
1827         ret = register_pernet_subsys(&fib6_net_ops);
1828         if (ret)
1829                 goto out_kmem_cache_create;
1830 
1831         ret = __rtnl_register(PF_INET6, RTM_GETROUTE, NULL, inet6_dump_fib,
1832                               NULL);
1833         if (ret)
1834                 goto out_unregister_subsys;
1835 
1836         __fib6_flush_trees = fib6_flush_trees;
1837 out:
1838         return ret;
1839 
1840 out_unregister_subsys:
1841         unregister_pernet_subsys(&fib6_net_ops);
1842 out_kmem_cache_create:
1843         kmem_cache_destroy(fib6_node_kmem);
1844         goto out;
1845 }
1846 
1847 void fib6_gc_cleanup(void)
1848 {
1849         unregister_pernet_subsys(&fib6_net_ops);
1850         kmem_cache_destroy(fib6_node_kmem);
1851 }
1852 
1853 #ifdef CONFIG_PROC_FS
1854 
1855 struct ipv6_route_iter {
1856         struct seq_net_private p;
1857         struct fib6_walker w;
1858         loff_t skip;
1859         struct fib6_table *tbl;
1860         int sernum;
1861 };
1862 
1863 static int ipv6_route_seq_show(struct seq_file *seq, void *v)
1864 {
1865         struct rt6_info *rt = v;
1866         struct ipv6_route_iter *iter = seq->private;
1867 
1868         seq_printf(seq, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
1869 
1870 #ifdef CONFIG_IPV6_SUBTREES
1871         seq_printf(seq, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
1872 #else
1873         seq_puts(seq, "00000000000000000000000000000000 00 ");
1874 #endif
1875         if (rt->rt6i_flags & RTF_GATEWAY)
1876                 seq_printf(seq, "%pi6", &rt->rt6i_gateway);
1877         else
1878                 seq_puts(seq, "00000000000000000000000000000000");
1879 
1880         seq_printf(seq, " %08x %08x %08x %08x %8s\n",
1881                    rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
1882                    rt->dst.__use, rt->rt6i_flags,
1883                    rt->dst.dev ? rt->dst.dev->name : "");
1884         iter->w.leaf = NULL;
1885         return 0;
1886 }
1887 
1888 static int ipv6_route_yield(struct fib6_walker *w)
1889 {
1890         struct ipv6_route_iter *iter = w->args;
1891 
1892         if (!iter->skip)
1893                 return 1;
1894 
1895         do {
1896                 iter->w.leaf = iter->w.leaf->dst.rt6_next;
1897                 iter->skip--;
1898                 if (!iter->skip && iter->w.leaf)
1899                         return 1;
1900         } while (iter->w.leaf);
1901 
1902         return 0;
1903 }
1904 
1905 static void ipv6_route_seq_setup_walk(struct ipv6_route_iter *iter)
1906 {
1907         memset(&iter->w, 0, sizeof(iter->w));
1908         iter->w.func = ipv6_route_yield;
1909         iter->w.root = &iter->tbl->tb6_root;
1910         iter->w.state = FWS_INIT;
1911         iter->w.node = iter->w.root;
1912         iter->w.args = iter;
1913         iter->sernum = iter->w.root->fn_sernum;
1914         INIT_LIST_HEAD(&iter->w.lh);
1915         fib6_walker_link(&iter->w);
1916 }
1917 
1918 static struct fib6_table *ipv6_route_seq_next_table(struct fib6_table *tbl,
1919                                                     struct net *net)
1920 {
1921         unsigned int h;
1922         struct hlist_node *node;
1923 
1924         if (tbl) {
1925                 h = (tbl->tb6_id & (FIB6_TABLE_HASHSZ - 1)) + 1;
1926                 node = rcu_dereference_bh(hlist_next_rcu(&tbl->tb6_hlist));
1927         } else {
1928                 h = 0;
1929                 node = NULL;
1930         }
1931 
1932         while (!node && h < FIB6_TABLE_HASHSZ) {
1933                 node = rcu_dereference_bh(
1934                         hlist_first_rcu(&net->ipv6.fib_table_hash[h++]));
1935         }
1936         return hlist_entry_safe(node, struct fib6_table, tb6_hlist);
1937 }
1938 
1939 static void ipv6_route_check_sernum(struct ipv6_route_iter *iter)
1940 {
1941         if (iter->sernum != iter->w.root->fn_sernum) {
1942                 iter->sernum = iter->w.root->fn_sernum;
1943                 iter->w.state = FWS_INIT;
1944                 iter->w.node = iter->w.root;
1945                 WARN_ON(iter->w.skip);
1946                 iter->w.skip = iter->w.count;
1947         }
1948 }
1949 
1950 static void *ipv6_route_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1951 {
1952         int r;
1953         struct rt6_info *n;
1954         struct net *net = seq_file_net(seq);
1955         struct ipv6_route_iter *iter = seq->private;
1956 
1957         if (!v)
1958                 goto iter_table;
1959 
1960         n = ((struct rt6_info *)v)->dst.rt6_next;
1961         if (n) {
1962                 ++*pos;
1963                 return n;
1964         }
1965 
1966 iter_table:
1967         ipv6_route_check_sernum(iter);
1968         read_lock(&iter->tbl->tb6_lock);
1969         r = fib6_walk_continue(&iter->w);
1970         read_unlock(&iter->tbl->tb6_lock);
1971         if (r > 0) {
1972                 if (v)
1973                         ++*pos;
1974                 return iter->w.leaf;
1975         } else if (r < 0) {
1976                 fib6_walker_unlink(&iter->w);
1977                 return NULL;
1978         }
1979         fib6_walker_unlink(&iter->w);
1980 
1981         iter->tbl = ipv6_route_seq_next_table(iter->tbl, net);
1982         if (!iter->tbl)
1983                 return NULL;
1984 
1985         ipv6_route_seq_setup_walk(iter);
1986         goto iter_table;
1987 }
1988 
1989 static void *ipv6_route_seq_start(struct seq_file *seq, loff_t *pos)
1990         __acquires(RCU_BH)
1991 {
1992         struct net *net = seq_file_net(seq);
1993         struct ipv6_route_iter *iter = seq->private;
1994 
1995         rcu_read_lock_bh();
1996         iter->tbl = ipv6_route_seq_next_table(NULL, net);
1997         iter->skip = *pos;
1998 
1999         if (iter->tbl) {
2000                 ipv6_route_seq_setup_walk(iter);
2001                 return ipv6_route_seq_next(seq, NULL, pos);
2002         } else {
2003                 return NULL;
2004         }
2005 }
2006 
2007 static bool ipv6_route_iter_active(struct ipv6_route_iter *iter)
2008 {
2009         struct fib6_walker *w = &iter->w;
2010         return w->node && !(w->state == FWS_U && w->node == w->root);
2011 }
2012 
2013 static void ipv6_route_seq_stop(struct seq_file *seq, void *v)
2014         __releases(RCU_BH)
2015 {
2016         struct ipv6_route_iter *iter = seq->private;
2017 
2018         if (ipv6_route_iter_active(iter))
2019                 fib6_walker_unlink(&iter->w);
2020 
2021         rcu_read_unlock_bh();
2022 }
2023 
2024 static const struct seq_operations ipv6_route_seq_ops = {
2025         .start  = ipv6_route_seq_start,
2026         .next   = ipv6_route_seq_next,
2027         .stop   = ipv6_route_seq_stop,
2028         .show   = ipv6_route_seq_show
2029 };
2030 
2031 int ipv6_route_open(struct inode *inode, struct file *file)
2032 {
2033         return seq_open_net(inode, file, &ipv6_route_seq_ops,
2034                             sizeof(struct ipv6_route_iter));
2035 }
2036 
2037 #endif /* CONFIG_PROC_FS */
2038 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp