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

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

Version: ~ [ linux-5.12-rc7 ] ~ [ linux-5.11.13 ] ~ [ linux-5.10.29 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.111 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.186 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.230 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.266 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.266 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ 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 #include <net/lwtunnel.h>
 36 
 37 #include <net/ip6_fib.h>
 38 #include <net/ip6_route.h>
 39 
 40 #define RT6_DEBUG 2
 41 
 42 #if RT6_DEBUG >= 3
 43 #define RT6_TRACE(x...) pr_debug(x)
 44 #else
 45 #define RT6_TRACE(x...) do { ; } while (0)
 46 #endif
 47 
 48 static struct kmem_cache *fib6_node_kmem __read_mostly;
 49 
 50 struct fib6_cleaner {
 51         struct fib6_walker w;
 52         struct net *net;
 53         int (*func)(struct rt6_info *, void *arg);
 54         int sernum;
 55         void *arg;
 56 };
 57 
 58 #ifdef CONFIG_IPV6_SUBTREES
 59 #define FWS_INIT FWS_S
 60 #else
 61 #define FWS_INIT FWS_L
 62 #endif
 63 
 64 static void fib6_prune_clones(struct net *net, struct fib6_node *fn);
 65 static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn);
 66 static struct fib6_node *fib6_repair_tree(struct net *net, struct fib6_node *fn);
 67 static int fib6_walk(struct net *net, struct fib6_walker *w);
 68 static int fib6_walk_continue(struct fib6_walker *w);
 69 
 70 /*
 71  *      A routing update causes an increase of the serial number on the
 72  *      affected subtree. This allows for cached routes to be asynchronously
 73  *      tested when modifications are made to the destination cache as a
 74  *      result of redirects, path MTU changes, etc.
 75  */
 76 
 77 static void fib6_gc_timer_cb(unsigned long arg);
 78 
 79 #define FOR_WALKERS(net, w) \
 80         list_for_each_entry(w, &(net)->ipv6.fib6_walkers, lh)
 81 
 82 static void fib6_walker_link(struct net *net, struct fib6_walker *w)
 83 {
 84         write_lock_bh(&net->ipv6.fib6_walker_lock);
 85         list_add(&w->lh, &net->ipv6.fib6_walkers);
 86         write_unlock_bh(&net->ipv6.fib6_walker_lock);
 87 }
 88 
 89 static void fib6_walker_unlink(struct net *net, struct fib6_walker *w)
 90 {
 91         write_lock_bh(&net->ipv6.fib6_walker_lock);
 92         list_del(&w->lh);
 93         write_unlock_bh(&net->ipv6.fib6_walker_lock);
 94 }
 95 
 96 static int fib6_new_sernum(struct net *net)
 97 {
 98         int new, old;
 99 
100         do {
101                 old = atomic_read(&net->ipv6.fib6_sernum);
102                 new = old < INT_MAX ? old + 1 : 1;
103         } while (atomic_cmpxchg(&net->ipv6.fib6_sernum,
104                                 old, new) != old);
105         return new;
106 }
107 
108 enum {
109         FIB6_NO_SERNUM_CHANGE = 0,
110 };
111 
112 /*
113  *      Auxiliary address test functions for the radix tree.
114  *
115  *      These assume a 32bit processor (although it will work on
116  *      64bit processors)
117  */
118 
119 /*
120  *      test bit
121  */
122 #if defined(__LITTLE_ENDIAN)
123 # define BITOP_BE32_SWIZZLE     (0x1F & ~7)
124 #else
125 # define BITOP_BE32_SWIZZLE     0
126 #endif
127 
128 static __be32 addr_bit_set(const void *token, int fn_bit)
129 {
130         const __be32 *addr = token;
131         /*
132          * Here,
133          *      1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)
134          * is optimized version of
135          *      htonl(1 << ((~fn_bit)&0x1F))
136          * See include/asm-generic/bitops/le.h.
137          */
138         return (__force __be32)(1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)) &
139                addr[fn_bit >> 5];
140 }
141 
142 static struct fib6_node *node_alloc(void)
143 {
144         struct fib6_node *fn;
145 
146         fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC);
147 
148         return fn;
149 }
150 
151 static void node_free(struct fib6_node *fn)
152 {
153         kmem_cache_free(fib6_node_kmem, fn);
154 }
155 
156 static void rt6_rcu_free(struct rt6_info *rt)
157 {
158         call_rcu(&rt->dst.rcu_head, dst_rcu_free);
159 }
160 
161 static void rt6_free_pcpu(struct rt6_info *non_pcpu_rt)
162 {
163         int cpu;
164 
165         if (!non_pcpu_rt->rt6i_pcpu)
166                 return;
167 
168         for_each_possible_cpu(cpu) {
169                 struct rt6_info **ppcpu_rt;
170                 struct rt6_info *pcpu_rt;
171 
172                 ppcpu_rt = per_cpu_ptr(non_pcpu_rt->rt6i_pcpu, cpu);
173                 pcpu_rt = *ppcpu_rt;
174                 if (pcpu_rt) {
175                         rt6_rcu_free(pcpu_rt);
176                         *ppcpu_rt = NULL;
177                 }
178         }
179 
180         free_percpu(non_pcpu_rt->rt6i_pcpu);
181         non_pcpu_rt->rt6i_pcpu = NULL;
182 }
183 
184 static void rt6_release(struct rt6_info *rt)
185 {
186         if (atomic_dec_and_test(&rt->rt6i_ref)) {
187                 rt6_free_pcpu(rt);
188                 rt6_rcu_free(rt);
189         }
190 }
191 
192 static void fib6_link_table(struct net *net, struct fib6_table *tb)
193 {
194         unsigned int h;
195 
196         /*
197          * Initialize table lock at a single place to give lockdep a key,
198          * tables aren't visible prior to being linked to the list.
199          */
200         rwlock_init(&tb->tb6_lock);
201 
202         h = tb->tb6_id & (FIB6_TABLE_HASHSZ - 1);
203 
204         /*
205          * No protection necessary, this is the only list mutatation
206          * operation, tables never disappear once they exist.
207          */
208         hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]);
209 }
210 
211 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
212 
213 static struct fib6_table *fib6_alloc_table(struct net *net, u32 id)
214 {
215         struct fib6_table *table;
216 
217         table = kzalloc(sizeof(*table), GFP_ATOMIC);
218         if (table) {
219                 table->tb6_id = id;
220                 table->tb6_root.leaf = net->ipv6.ip6_null_entry;
221                 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
222                 inet_peer_base_init(&table->tb6_peers);
223         }
224 
225         return table;
226 }
227 
228 struct fib6_table *fib6_new_table(struct net *net, u32 id)
229 {
230         struct fib6_table *tb;
231 
232         if (id == 0)
233                 id = RT6_TABLE_MAIN;
234         tb = fib6_get_table(net, id);
235         if (tb)
236                 return tb;
237 
238         tb = fib6_alloc_table(net, id);
239         if (tb)
240                 fib6_link_table(net, tb);
241 
242         return tb;
243 }
244 EXPORT_SYMBOL_GPL(fib6_new_table);
245 
246 struct fib6_table *fib6_get_table(struct net *net, u32 id)
247 {
248         struct fib6_table *tb;
249         struct hlist_head *head;
250         unsigned int h;
251 
252         if (id == 0)
253                 id = RT6_TABLE_MAIN;
254         h = id & (FIB6_TABLE_HASHSZ - 1);
255         rcu_read_lock();
256         head = &net->ipv6.fib_table_hash[h];
257         hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
258                 if (tb->tb6_id == id) {
259                         rcu_read_unlock();
260                         return tb;
261                 }
262         }
263         rcu_read_unlock();
264 
265         return NULL;
266 }
267 EXPORT_SYMBOL_GPL(fib6_get_table);
268 
269 static void __net_init fib6_tables_init(struct net *net)
270 {
271         fib6_link_table(net, net->ipv6.fib6_main_tbl);
272         fib6_link_table(net, net->ipv6.fib6_local_tbl);
273 }
274 #else
275 
276 struct fib6_table *fib6_new_table(struct net *net, u32 id)
277 {
278         return fib6_get_table(net, id);
279 }
280 
281 struct fib6_table *fib6_get_table(struct net *net, u32 id)
282 {
283           return net->ipv6.fib6_main_tbl;
284 }
285 
286 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
287                                    int flags, pol_lookup_t lookup)
288 {
289         struct rt6_info *rt;
290 
291         rt = lookup(net, net->ipv6.fib6_main_tbl, fl6, flags);
292         if (rt->dst.error == -EAGAIN) {
293                 ip6_rt_put(rt);
294                 rt = net->ipv6.ip6_null_entry;
295                 dst_hold(&rt->dst);
296         }
297 
298         return &rt->dst;
299 }
300 
301 static void __net_init fib6_tables_init(struct net *net)
302 {
303         fib6_link_table(net, net->ipv6.fib6_main_tbl);
304 }
305 
306 #endif
307 
308 static int fib6_dump_node(struct fib6_walker *w)
309 {
310         int res;
311         struct rt6_info *rt;
312 
313         for (rt = w->leaf; rt; rt = rt->dst.rt6_next) {
314                 res = rt6_dump_route(rt, w->args);
315                 if (res < 0) {
316                         /* Frame is full, suspend walking */
317                         w->leaf = rt;
318                         return 1;
319                 }
320 
321                 /* Multipath routes are dumped in one route with the
322                  * RTA_MULTIPATH attribute. Jump 'rt' to point to the
323                  * last sibling of this route (no need to dump the
324                  * sibling routes again)
325                  */
326                 if (rt->rt6i_nsiblings)
327                         rt = list_last_entry(&rt->rt6i_siblings,
328                                              struct rt6_info,
329                                              rt6i_siblings);
330         }
331         w->leaf = NULL;
332         return 0;
333 }
334 
335 static void fib6_dump_end(struct netlink_callback *cb)
336 {
337         struct net *net = sock_net(cb->skb->sk);
338         struct fib6_walker *w = (void *)cb->args[2];
339 
340         if (w) {
341                 if (cb->args[4]) {
342                         cb->args[4] = 0;
343                         fib6_walker_unlink(net, w);
344                 }
345                 cb->args[2] = 0;
346                 kfree(w);
347         }
348         cb->done = (void *)cb->args[3];
349         cb->args[1] = 3;
350 }
351 
352 static int fib6_dump_done(struct netlink_callback *cb)
353 {
354         fib6_dump_end(cb);
355         return cb->done ? cb->done(cb) : 0;
356 }
357 
358 static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
359                            struct netlink_callback *cb)
360 {
361         struct net *net = sock_net(skb->sk);
362         struct fib6_walker *w;
363         int res;
364 
365         w = (void *)cb->args[2];
366         w->root = &table->tb6_root;
367 
368         if (cb->args[4] == 0) {
369                 w->count = 0;
370                 w->skip = 0;
371 
372                 read_lock_bh(&table->tb6_lock);
373                 res = fib6_walk(net, w);
374                 read_unlock_bh(&table->tb6_lock);
375                 if (res > 0) {
376                         cb->args[4] = 1;
377                         cb->args[5] = w->root->fn_sernum;
378                 }
379         } else {
380                 if (cb->args[5] != w->root->fn_sernum) {
381                         /* Begin at the root if the tree changed */
382                         cb->args[5] = w->root->fn_sernum;
383                         w->state = FWS_INIT;
384                         w->node = w->root;
385                         w->skip = w->count;
386                 } else
387                         w->skip = 0;
388 
389                 read_lock_bh(&table->tb6_lock);
390                 res = fib6_walk_continue(w);
391                 read_unlock_bh(&table->tb6_lock);
392                 if (res <= 0) {
393                         fib6_walker_unlink(net, w);
394                         cb->args[4] = 0;
395                 }
396         }
397 
398         return res;
399 }
400 
401 static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
402 {
403         struct net *net = sock_net(skb->sk);
404         unsigned int h, s_h;
405         unsigned int e = 0, s_e;
406         struct rt6_rtnl_dump_arg arg;
407         struct fib6_walker *w;
408         struct fib6_table *tb;
409         struct hlist_head *head;
410         int res = 0;
411 
412         s_h = cb->args[0];
413         s_e = cb->args[1];
414 
415         w = (void *)cb->args[2];
416         if (!w) {
417                 /* New dump:
418                  *
419                  * 1. hook callback destructor.
420                  */
421                 cb->args[3] = (long)cb->done;
422                 cb->done = fib6_dump_done;
423 
424                 /*
425                  * 2. allocate and initialize walker.
426                  */
427                 w = kzalloc(sizeof(*w), GFP_ATOMIC);
428                 if (!w)
429                         return -ENOMEM;
430                 w->func = fib6_dump_node;
431                 cb->args[2] = (long)w;
432         }
433 
434         arg.skb = skb;
435         arg.cb = cb;
436         arg.net = net;
437         w->args = &arg;
438 
439         rcu_read_lock();
440         for (h = s_h; h < FIB6_TABLE_HASHSZ; h++, s_e = 0) {
441                 e = 0;
442                 head = &net->ipv6.fib_table_hash[h];
443                 hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
444                         if (e < s_e)
445                                 goto next;
446                         res = fib6_dump_table(tb, skb, cb);
447                         if (res != 0)
448                                 goto out;
449 next:
450                         e++;
451                 }
452         }
453 out:
454         rcu_read_unlock();
455         cb->args[1] = e;
456         cb->args[0] = h;
457 
458         res = res < 0 ? res : skb->len;
459         if (res <= 0)
460                 fib6_dump_end(cb);
461         return res;
462 }
463 
464 /*
465  *      Routing Table
466  *
467  *      return the appropriate node for a routing tree "add" operation
468  *      by either creating and inserting or by returning an existing
469  *      node.
470  */
471 
472 static struct fib6_node *fib6_add_1(struct fib6_node *root,
473                                      struct in6_addr *addr, int plen,
474                                      int offset, int allow_create,
475                                      int replace_required, int sernum)
476 {
477         struct fib6_node *fn, *in, *ln;
478         struct fib6_node *pn = NULL;
479         struct rt6key *key;
480         int     bit;
481         __be32  dir = 0;
482 
483         RT6_TRACE("fib6_add_1\n");
484 
485         /* insert node in tree */
486 
487         fn = root;
488 
489         do {
490                 key = (struct rt6key *)((u8 *)fn->leaf + offset);
491 
492                 /*
493                  *      Prefix match
494                  */
495                 if (plen < fn->fn_bit ||
496                     !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) {
497                         if (!allow_create) {
498                                 if (replace_required) {
499                                         pr_warn("Can't replace route, no match found\n");
500                                         return ERR_PTR(-ENOENT);
501                                 }
502                                 pr_warn("NLM_F_CREATE should be set when creating new route\n");
503                         }
504                         goto insert_above;
505                 }
506 
507                 /*
508                  *      Exact match ?
509                  */
510 
511                 if (plen == fn->fn_bit) {
512                         /* clean up an intermediate node */
513                         if (!(fn->fn_flags & RTN_RTINFO)) {
514                                 rt6_release(fn->leaf);
515                                 fn->leaf = NULL;
516                         }
517 
518                         fn->fn_sernum = sernum;
519 
520                         return fn;
521                 }
522 
523                 /*
524                  *      We have more bits to go
525                  */
526 
527                 /* Try to walk down on tree. */
528                 fn->fn_sernum = sernum;
529                 dir = addr_bit_set(addr, fn->fn_bit);
530                 pn = fn;
531                 fn = dir ? fn->right : fn->left;
532         } while (fn);
533 
534         if (!allow_create) {
535                 /* We should not create new node because
536                  * NLM_F_REPLACE was specified without NLM_F_CREATE
537                  * I assume it is safe to require NLM_F_CREATE when
538                  * REPLACE flag is used! Later we may want to remove the
539                  * check for replace_required, because according
540                  * to netlink specification, NLM_F_CREATE
541                  * MUST be specified if new route is created.
542                  * That would keep IPv6 consistent with IPv4
543                  */
544                 if (replace_required) {
545                         pr_warn("Can't replace route, no match found\n");
546                         return ERR_PTR(-ENOENT);
547                 }
548                 pr_warn("NLM_F_CREATE should be set when creating new route\n");
549         }
550         /*
551          *      We walked to the bottom of tree.
552          *      Create new leaf node without children.
553          */
554 
555         ln = node_alloc();
556 
557         if (!ln)
558                 return ERR_PTR(-ENOMEM);
559         ln->fn_bit = plen;
560 
561         ln->parent = pn;
562         ln->fn_sernum = sernum;
563 
564         if (dir)
565                 pn->right = ln;
566         else
567                 pn->left  = ln;
568 
569         return ln;
570 
571 
572 insert_above:
573         /*
574          * split since we don't have a common prefix anymore or
575          * we have a less significant route.
576          * we've to insert an intermediate node on the list
577          * this new node will point to the one we need to create
578          * and the current
579          */
580 
581         pn = fn->parent;
582 
583         /* find 1st bit in difference between the 2 addrs.
584 
585            See comment in __ipv6_addr_diff: bit may be an invalid value,
586            but if it is >= plen, the value is ignored in any case.
587          */
588 
589         bit = __ipv6_addr_diff(addr, &key->addr, sizeof(*addr));
590 
591         /*
592          *              (intermediate)[in]
593          *                /        \
594          *      (new leaf node)[ln] (old node)[fn]
595          */
596         if (plen > bit) {
597                 in = node_alloc();
598                 ln = node_alloc();
599 
600                 if (!in || !ln) {
601                         if (in)
602                                 node_free(in);
603                         if (ln)
604                                 node_free(ln);
605                         return ERR_PTR(-ENOMEM);
606                 }
607 
608                 /*
609                  * new intermediate node.
610                  * RTN_RTINFO will
611                  * be off since that an address that chooses one of
612                  * the branches would not match less specific routes
613                  * in the other branch
614                  */
615 
616                 in->fn_bit = bit;
617 
618                 in->parent = pn;
619                 in->leaf = fn->leaf;
620                 atomic_inc(&in->leaf->rt6i_ref);
621 
622                 in->fn_sernum = sernum;
623 
624                 /* update parent pointer */
625                 if (dir)
626                         pn->right = in;
627                 else
628                         pn->left  = in;
629 
630                 ln->fn_bit = plen;
631 
632                 ln->parent = in;
633                 fn->parent = in;
634 
635                 ln->fn_sernum = sernum;
636 
637                 if (addr_bit_set(addr, bit)) {
638                         in->right = ln;
639                         in->left  = fn;
640                 } else {
641                         in->left  = ln;
642                         in->right = fn;
643                 }
644         } else { /* plen <= bit */
645 
646                 /*
647                  *              (new leaf node)[ln]
648                  *                /        \
649                  *           (old node)[fn] NULL
650                  */
651 
652                 ln = node_alloc();
653 
654                 if (!ln)
655                         return ERR_PTR(-ENOMEM);
656 
657                 ln->fn_bit = plen;
658 
659                 ln->parent = pn;
660 
661                 ln->fn_sernum = sernum;
662 
663                 if (dir)
664                         pn->right = ln;
665                 else
666                         pn->left  = ln;
667 
668                 if (addr_bit_set(&key->addr, plen))
669                         ln->right = fn;
670                 else
671                         ln->left  = fn;
672 
673                 fn->parent = ln;
674         }
675         return ln;
676 }
677 
678 static bool rt6_qualify_for_ecmp(struct rt6_info *rt)
679 {
680         return (rt->rt6i_flags & (RTF_GATEWAY|RTF_ADDRCONF|RTF_DYNAMIC)) ==
681                RTF_GATEWAY;
682 }
683 
684 static void fib6_copy_metrics(u32 *mp, const struct mx6_config *mxc)
685 {
686         int i;
687 
688         for (i = 0; i < RTAX_MAX; i++) {
689                 if (test_bit(i, mxc->mx_valid))
690                         mp[i] = mxc->mx[i];
691         }
692 }
693 
694 static int fib6_commit_metrics(struct dst_entry *dst, struct mx6_config *mxc)
695 {
696         if (!mxc->mx)
697                 return 0;
698 
699         if (dst->flags & DST_HOST) {
700                 u32 *mp = dst_metrics_write_ptr(dst);
701 
702                 if (unlikely(!mp))
703                         return -ENOMEM;
704 
705                 fib6_copy_metrics(mp, mxc);
706         } else {
707                 dst_init_metrics(dst, mxc->mx, false);
708 
709                 /* We've stolen mx now. */
710                 mxc->mx = NULL;
711         }
712 
713         return 0;
714 }
715 
716 static void fib6_purge_rt(struct rt6_info *rt, struct fib6_node *fn,
717                           struct net *net)
718 {
719         if (atomic_read(&rt->rt6i_ref) != 1) {
720                 /* This route is used as dummy address holder in some split
721                  * nodes. It is not leaked, but it still holds other resources,
722                  * which must be released in time. So, scan ascendant nodes
723                  * and replace dummy references to this route with references
724                  * to still alive ones.
725                  */
726                 while (fn) {
727                         if (!(fn->fn_flags & RTN_RTINFO) && fn->leaf == rt) {
728                                 fn->leaf = fib6_find_prefix(net, fn);
729                                 atomic_inc(&fn->leaf->rt6i_ref);
730                                 rt6_release(rt);
731                         }
732                         fn = fn->parent;
733                 }
734                 /* No more references are possible at this point. */
735                 BUG_ON(atomic_read(&rt->rt6i_ref) != 1);
736         }
737 }
738 
739 /*
740  *      Insert routing information in a node.
741  */
742 
743 static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt,
744                             struct nl_info *info, struct mx6_config *mxc)
745 {
746         struct rt6_info *iter = NULL;
747         struct rt6_info **ins;
748         struct rt6_info **fallback_ins = NULL;
749         int replace = (info->nlh &&
750                        (info->nlh->nlmsg_flags & NLM_F_REPLACE));
751         int add = (!info->nlh ||
752                    (info->nlh->nlmsg_flags & NLM_F_CREATE));
753         int found = 0;
754         bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
755         u16 nlflags = NLM_F_EXCL;
756         int err;
757 
758         if (info->nlh && (info->nlh->nlmsg_flags & NLM_F_APPEND))
759                 nlflags |= NLM_F_APPEND;
760 
761         ins = &fn->leaf;
762 
763         for (iter = fn->leaf; iter; iter = iter->dst.rt6_next) {
764                 /*
765                  *      Search for duplicates
766                  */
767 
768                 if (iter->rt6i_metric == rt->rt6i_metric) {
769                         /*
770                          *      Same priority level
771                          */
772                         if (info->nlh &&
773                             (info->nlh->nlmsg_flags & NLM_F_EXCL))
774                                 return -EEXIST;
775 
776                         nlflags &= ~NLM_F_EXCL;
777                         if (replace) {
778                                 if (rt_can_ecmp == rt6_qualify_for_ecmp(iter)) {
779                                         found++;
780                                         break;
781                                 }
782                                 if (rt_can_ecmp)
783                                         fallback_ins = fallback_ins ?: ins;
784                                 goto next_iter;
785                         }
786 
787                         if (rt6_duplicate_nexthop(iter, rt)) {
788                                 if (rt->rt6i_nsiblings)
789                                         rt->rt6i_nsiblings = 0;
790                                 if (!(iter->rt6i_flags & RTF_EXPIRES))
791                                         return -EEXIST;
792                                 if (!(rt->rt6i_flags & RTF_EXPIRES))
793                                         rt6_clean_expires(iter);
794                                 else
795                                         rt6_set_expires(iter, rt->dst.expires);
796                                 iter->rt6i_pmtu = rt->rt6i_pmtu;
797                                 return -EEXIST;
798                         }
799                         /* If we have the same destination and the same metric,
800                          * but not the same gateway, then the route we try to
801                          * add is sibling to this route, increment our counter
802                          * of siblings, and later we will add our route to the
803                          * list.
804                          * Only static routes (which don't have flag
805                          * RTF_EXPIRES) are used for ECMPv6.
806                          *
807                          * To avoid long list, we only had siblings if the
808                          * route have a gateway.
809                          */
810                         if (rt_can_ecmp &&
811                             rt6_qualify_for_ecmp(iter))
812                                 rt->rt6i_nsiblings++;
813                 }
814 
815                 if (iter->rt6i_metric > rt->rt6i_metric)
816                         break;
817 
818 next_iter:
819                 ins = &iter->dst.rt6_next;
820         }
821 
822         if (fallback_ins && !found) {
823                 /* No ECMP-able route found, replace first non-ECMP one */
824                 ins = fallback_ins;
825                 iter = *ins;
826                 found++;
827         }
828 
829         /* Reset round-robin state, if necessary */
830         if (ins == &fn->leaf)
831                 fn->rr_ptr = NULL;
832 
833         /* Link this route to others same route. */
834         if (rt->rt6i_nsiblings) {
835                 unsigned int rt6i_nsiblings;
836                 struct rt6_info *sibling, *temp_sibling;
837 
838                 /* Find the first route that have the same metric */
839                 sibling = fn->leaf;
840                 while (sibling) {
841                         if (sibling->rt6i_metric == rt->rt6i_metric &&
842                             rt6_qualify_for_ecmp(sibling)) {
843                                 list_add_tail(&rt->rt6i_siblings,
844                                               &sibling->rt6i_siblings);
845                                 break;
846                         }
847                         sibling = sibling->dst.rt6_next;
848                 }
849                 /* For each sibling in the list, increment the counter of
850                  * siblings. BUG() if counters does not match, list of siblings
851                  * is broken!
852                  */
853                 rt6i_nsiblings = 0;
854                 list_for_each_entry_safe(sibling, temp_sibling,
855                                          &rt->rt6i_siblings, rt6i_siblings) {
856                         sibling->rt6i_nsiblings++;
857                         BUG_ON(sibling->rt6i_nsiblings != rt->rt6i_nsiblings);
858                         rt6i_nsiblings++;
859                 }
860                 BUG_ON(rt6i_nsiblings != rt->rt6i_nsiblings);
861         }
862 
863         /*
864          *      insert node
865          */
866         if (!replace) {
867                 if (!add)
868                         pr_warn("NLM_F_CREATE should be set when creating new route\n");
869 
870 add:
871                 nlflags |= NLM_F_CREATE;
872                 err = fib6_commit_metrics(&rt->dst, mxc);
873                 if (err)
874                         return err;
875 
876                 rt->dst.rt6_next = iter;
877                 *ins = rt;
878                 rt->rt6i_node = fn;
879                 atomic_inc(&rt->rt6i_ref);
880                 if (!info->skip_notify)
881                         inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
882                 info->nl_net->ipv6.rt6_stats->fib_rt_entries++;
883 
884                 if (!(fn->fn_flags & RTN_RTINFO)) {
885                         info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
886                         fn->fn_flags |= RTN_RTINFO;
887                 }
888 
889         } else {
890                 int nsiblings;
891 
892                 if (!found) {
893                         if (add)
894                                 goto add;
895                         pr_warn("NLM_F_REPLACE set, but no existing node found!\n");
896                         return -ENOENT;
897                 }
898 
899                 err = fib6_commit_metrics(&rt->dst, mxc);
900                 if (err)
901                         return err;
902 
903                 *ins = rt;
904                 rt->rt6i_node = fn;
905                 rt->dst.rt6_next = iter->dst.rt6_next;
906                 atomic_inc(&rt->rt6i_ref);
907                 if (!info->skip_notify)
908                         inet6_rt_notify(RTM_NEWROUTE, rt, info, NLM_F_REPLACE);
909                 if (!(fn->fn_flags & RTN_RTINFO)) {
910                         info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
911                         fn->fn_flags |= RTN_RTINFO;
912                 }
913                 nsiblings = iter->rt6i_nsiblings;
914                 fib6_purge_rt(iter, fn, info->nl_net);
915                 rt6_release(iter);
916 
917                 if (nsiblings) {
918                         /* Replacing an ECMP route, remove all siblings */
919                         ins = &rt->dst.rt6_next;
920                         iter = *ins;
921                         while (iter) {
922                                 if (iter->rt6i_metric > rt->rt6i_metric)
923                                         break;
924                                 if (rt6_qualify_for_ecmp(iter)) {
925                                         *ins = iter->dst.rt6_next;
926                                         fib6_purge_rt(iter, fn, info->nl_net);
927                                         rt6_release(iter);
928                                         nsiblings--;
929                                 } else {
930                                         ins = &iter->dst.rt6_next;
931                                 }
932                                 iter = *ins;
933                         }
934                         WARN_ON(nsiblings != 0);
935                 }
936         }
937 
938         return 0;
939 }
940 
941 static void fib6_start_gc(struct net *net, struct rt6_info *rt)
942 {
943         if (!timer_pending(&net->ipv6.ip6_fib_timer) &&
944             (rt->rt6i_flags & (RTF_EXPIRES | RTF_CACHE)))
945                 mod_timer(&net->ipv6.ip6_fib_timer,
946                           jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
947 }
948 
949 void fib6_force_start_gc(struct net *net)
950 {
951         if (!timer_pending(&net->ipv6.ip6_fib_timer))
952                 mod_timer(&net->ipv6.ip6_fib_timer,
953                           jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
954 }
955 
956 /*
957  *      Add routing information to the routing tree.
958  *      <destination addr>/<source addr>
959  *      with source addr info in sub-trees
960  */
961 
962 int fib6_add(struct fib6_node *root, struct rt6_info *rt,
963              struct nl_info *info, struct mx6_config *mxc)
964 {
965         struct fib6_node *fn, *pn = NULL;
966         int err = -ENOMEM;
967         int allow_create = 1;
968         int replace_required = 0;
969         int sernum = fib6_new_sernum(info->nl_net);
970 
971         if (WARN_ON_ONCE((rt->dst.flags & DST_NOCACHE) &&
972                          !atomic_read(&rt->dst.__refcnt)))
973                 return -EINVAL;
974 
975         if (info->nlh) {
976                 if (!(info->nlh->nlmsg_flags & NLM_F_CREATE))
977                         allow_create = 0;
978                 if (info->nlh->nlmsg_flags & NLM_F_REPLACE)
979                         replace_required = 1;
980         }
981         if (!allow_create && !replace_required)
982                 pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n");
983 
984         fn = fib6_add_1(root, &rt->rt6i_dst.addr, rt->rt6i_dst.plen,
985                         offsetof(struct rt6_info, rt6i_dst), allow_create,
986                         replace_required, sernum);
987         if (IS_ERR(fn)) {
988                 err = PTR_ERR(fn);
989                 fn = NULL;
990                 goto out;
991         }
992 
993         pn = fn;
994 
995 #ifdef CONFIG_IPV6_SUBTREES
996         if (rt->rt6i_src.plen) {
997                 struct fib6_node *sn;
998 
999                 if (!fn->subtree) {
1000                         struct fib6_node *sfn;
1001 
1002                         /*
1003                          * Create subtree.
1004                          *
1005                          *              fn[main tree]
1006                          *              |
1007                          *              sfn[subtree root]
1008                          *                 \
1009                          *                  sn[new leaf node]
1010                          */
1011 
1012                         /* Create subtree root node */
1013                         sfn = node_alloc();
1014                         if (!sfn)
1015                                 goto st_failure;
1016 
1017                         sfn->leaf = info->nl_net->ipv6.ip6_null_entry;
1018                         atomic_inc(&info->nl_net->ipv6.ip6_null_entry->rt6i_ref);
1019                         sfn->fn_flags = RTN_ROOT;
1020                         sfn->fn_sernum = sernum;
1021 
1022                         /* Now add the first leaf node to new subtree */
1023 
1024                         sn = fib6_add_1(sfn, &rt->rt6i_src.addr,
1025                                         rt->rt6i_src.plen,
1026                                         offsetof(struct rt6_info, rt6i_src),
1027                                         allow_create, replace_required, sernum);
1028 
1029                         if (IS_ERR(sn)) {
1030                                 /* If it is failed, discard just allocated
1031                                    root, and then (in st_failure) stale node
1032                                    in main tree.
1033                                  */
1034                                 node_free(sfn);
1035                                 err = PTR_ERR(sn);
1036                                 goto st_failure;
1037                         }
1038 
1039                         /* Now link new subtree to main tree */
1040                         sfn->parent = fn;
1041                         fn->subtree = sfn;
1042                 } else {
1043                         sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr,
1044                                         rt->rt6i_src.plen,
1045                                         offsetof(struct rt6_info, rt6i_src),
1046                                         allow_create, replace_required, sernum);
1047 
1048                         if (IS_ERR(sn)) {
1049                                 err = PTR_ERR(sn);
1050                                 goto st_failure;
1051                         }
1052                 }
1053 
1054                 if (!fn->leaf) {
1055                         fn->leaf = rt;
1056                         atomic_inc(&rt->rt6i_ref);
1057                 }
1058                 fn = sn;
1059         }
1060 #endif
1061 
1062         err = fib6_add_rt2node(fn, rt, info, mxc);
1063         if (!err) {
1064                 fib6_start_gc(info->nl_net, rt);
1065                 if (!(rt->rt6i_flags & RTF_CACHE))
1066                         fib6_prune_clones(info->nl_net, pn);
1067                 rt->dst.flags &= ~DST_NOCACHE;
1068         }
1069 
1070 out:
1071         if (err) {
1072 #ifdef CONFIG_IPV6_SUBTREES
1073                 /*
1074                  * If fib6_add_1 has cleared the old leaf pointer in the
1075                  * super-tree leaf node we have to find a new one for it.
1076                  */
1077                 if (pn != fn && pn->leaf == rt) {
1078                         pn->leaf = NULL;
1079                         atomic_dec(&rt->rt6i_ref);
1080                 }
1081                 if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) {
1082                         pn->leaf = fib6_find_prefix(info->nl_net, pn);
1083 #if RT6_DEBUG >= 2
1084                         if (!pn->leaf) {
1085                                 WARN_ON(pn->leaf == NULL);
1086                                 pn->leaf = info->nl_net->ipv6.ip6_null_entry;
1087                         }
1088 #endif
1089                         atomic_inc(&pn->leaf->rt6i_ref);
1090                 }
1091 #endif
1092                 if (!(rt->dst.flags & DST_NOCACHE))
1093                         dst_free(&rt->dst);
1094         }
1095         return err;
1096 
1097 #ifdef CONFIG_IPV6_SUBTREES
1098         /* Subtree creation failed, probably main tree node
1099            is orphan. If it is, shoot it.
1100          */
1101 st_failure:
1102         if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)))
1103                 fib6_repair_tree(info->nl_net, fn);
1104         if (!(rt->dst.flags & DST_NOCACHE))
1105                 dst_free(&rt->dst);
1106         return err;
1107 #endif
1108 }
1109 
1110 /*
1111  *      Routing tree lookup
1112  *
1113  */
1114 
1115 struct lookup_args {
1116         int                     offset;         /* key offset on rt6_info       */
1117         const struct in6_addr   *addr;          /* search key                   */
1118 };
1119 
1120 static struct fib6_node *fib6_lookup_1(struct fib6_node *root,
1121                                        struct lookup_args *args)
1122 {
1123         struct fib6_node *fn;
1124         __be32 dir;
1125 
1126         if (unlikely(args->offset == 0))
1127                 return NULL;
1128 
1129         /*
1130          *      Descend on a tree
1131          */
1132 
1133         fn = root;
1134 
1135         for (;;) {
1136                 struct fib6_node *next;
1137 
1138                 dir = addr_bit_set(args->addr, fn->fn_bit);
1139 
1140                 next = dir ? fn->right : fn->left;
1141 
1142                 if (next) {
1143                         fn = next;
1144                         continue;
1145                 }
1146                 break;
1147         }
1148 
1149         while (fn) {
1150                 if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) {
1151                         struct rt6key *key;
1152 
1153                         key = (struct rt6key *) ((u8 *) fn->leaf +
1154                                                  args->offset);
1155 
1156                         if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
1157 #ifdef CONFIG_IPV6_SUBTREES
1158                                 if (fn->subtree) {
1159                                         struct fib6_node *sfn;
1160                                         sfn = fib6_lookup_1(fn->subtree,
1161                                                             args + 1);
1162                                         if (!sfn)
1163                                                 goto backtrack;
1164                                         fn = sfn;
1165                                 }
1166 #endif
1167                                 if (fn->fn_flags & RTN_RTINFO)
1168                                         return fn;
1169                         }
1170                 }
1171 #ifdef CONFIG_IPV6_SUBTREES
1172 backtrack:
1173 #endif
1174                 if (fn->fn_flags & RTN_ROOT)
1175                         break;
1176 
1177                 fn = fn->parent;
1178         }
1179 
1180         return NULL;
1181 }
1182 
1183 struct fib6_node *fib6_lookup(struct fib6_node *root, const struct in6_addr *daddr,
1184                               const struct in6_addr *saddr)
1185 {
1186         struct fib6_node *fn;
1187         struct lookup_args args[] = {
1188                 {
1189                         .offset = offsetof(struct rt6_info, rt6i_dst),
1190                         .addr = daddr,
1191                 },
1192 #ifdef CONFIG_IPV6_SUBTREES
1193                 {
1194                         .offset = offsetof(struct rt6_info, rt6i_src),
1195                         .addr = saddr,
1196                 },
1197 #endif
1198                 {
1199                         .offset = 0,    /* sentinel */
1200                 }
1201         };
1202 
1203         fn = fib6_lookup_1(root, daddr ? args : args + 1);
1204         if (!fn || fn->fn_flags & RTN_TL_ROOT)
1205                 fn = root;
1206 
1207         return fn;
1208 }
1209 
1210 /*
1211  *      Get node with specified destination prefix (and source prefix,
1212  *      if subtrees are used)
1213  */
1214 
1215 
1216 static struct fib6_node *fib6_locate_1(struct fib6_node *root,
1217                                        const struct in6_addr *addr,
1218                                        int plen, int offset)
1219 {
1220         struct fib6_node *fn;
1221 
1222         for (fn = root; fn ; ) {
1223                 struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset);
1224 
1225                 /*
1226                  *      Prefix match
1227                  */
1228                 if (plen < fn->fn_bit ||
1229                     !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
1230                         return NULL;
1231 
1232                 if (plen == fn->fn_bit)
1233                         return fn;
1234 
1235                 /*
1236                  *      We have more bits to go
1237                  */
1238                 if (addr_bit_set(addr, fn->fn_bit))
1239                         fn = fn->right;
1240                 else
1241                         fn = fn->left;
1242         }
1243         return NULL;
1244 }
1245 
1246 struct fib6_node *fib6_locate(struct fib6_node *root,
1247                               const struct in6_addr *daddr, int dst_len,
1248                               const struct in6_addr *saddr, int src_len)
1249 {
1250         struct fib6_node *fn;
1251 
1252         fn = fib6_locate_1(root, daddr, dst_len,
1253                            offsetof(struct rt6_info, rt6i_dst));
1254 
1255 #ifdef CONFIG_IPV6_SUBTREES
1256         if (src_len) {
1257                 WARN_ON(saddr == NULL);
1258                 if (fn && fn->subtree)
1259                         fn = fib6_locate_1(fn->subtree, saddr, src_len,
1260                                            offsetof(struct rt6_info, rt6i_src));
1261         }
1262 #endif
1263 
1264         if (fn && fn->fn_flags & RTN_RTINFO)
1265                 return fn;
1266 
1267         return NULL;
1268 }
1269 
1270 
1271 /*
1272  *      Deletion
1273  *
1274  */
1275 
1276 static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn)
1277 {
1278         if (fn->fn_flags & RTN_ROOT)
1279                 return net->ipv6.ip6_null_entry;
1280 
1281         while (fn) {
1282                 if (fn->left)
1283                         return fn->left->leaf;
1284                 if (fn->right)
1285                         return fn->right->leaf;
1286 
1287                 fn = FIB6_SUBTREE(fn);
1288         }
1289         return NULL;
1290 }
1291 
1292 /*
1293  *      Called to trim the tree of intermediate nodes when possible. "fn"
1294  *      is the node we want to try and remove.
1295  */
1296 
1297 static struct fib6_node *fib6_repair_tree(struct net *net,
1298                                            struct fib6_node *fn)
1299 {
1300         int children;
1301         int nstate;
1302         struct fib6_node *child, *pn;
1303         struct fib6_walker *w;
1304         int iter = 0;
1305 
1306         for (;;) {
1307                 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter);
1308                 iter++;
1309 
1310                 WARN_ON(fn->fn_flags & RTN_RTINFO);
1311                 WARN_ON(fn->fn_flags & RTN_TL_ROOT);
1312                 WARN_ON(fn->leaf);
1313 
1314                 children = 0;
1315                 child = NULL;
1316                 if (fn->right)
1317                         child = fn->right, children |= 1;
1318                 if (fn->left)
1319                         child = fn->left, children |= 2;
1320 
1321                 if (children == 3 || FIB6_SUBTREE(fn)
1322 #ifdef CONFIG_IPV6_SUBTREES
1323                     /* Subtree root (i.e. fn) may have one child */
1324                     || (children && fn->fn_flags & RTN_ROOT)
1325 #endif
1326                     ) {
1327                         fn->leaf = fib6_find_prefix(net, fn);
1328 #if RT6_DEBUG >= 2
1329                         if (!fn->leaf) {
1330                                 WARN_ON(!fn->leaf);
1331                                 fn->leaf = net->ipv6.ip6_null_entry;
1332                         }
1333 #endif
1334                         atomic_inc(&fn->leaf->rt6i_ref);
1335                         return fn->parent;
1336                 }
1337 
1338                 pn = fn->parent;
1339 #ifdef CONFIG_IPV6_SUBTREES
1340                 if (FIB6_SUBTREE(pn) == fn) {
1341                         WARN_ON(!(fn->fn_flags & RTN_ROOT));
1342                         FIB6_SUBTREE(pn) = NULL;
1343                         nstate = FWS_L;
1344                 } else {
1345                         WARN_ON(fn->fn_flags & RTN_ROOT);
1346 #endif
1347                         if (pn->right == fn)
1348                                 pn->right = child;
1349                         else if (pn->left == fn)
1350                                 pn->left = child;
1351 #if RT6_DEBUG >= 2
1352                         else
1353                                 WARN_ON(1);
1354 #endif
1355                         if (child)
1356                                 child->parent = pn;
1357                         nstate = FWS_R;
1358 #ifdef CONFIG_IPV6_SUBTREES
1359                 }
1360 #endif
1361 
1362                 read_lock(&net->ipv6.fib6_walker_lock);
1363                 FOR_WALKERS(net, w) {
1364                         if (!child) {
1365                                 if (w->root == fn) {
1366                                         w->root = w->node = NULL;
1367                                         RT6_TRACE("W %p adjusted by delroot 1\n", w);
1368                                 } else if (w->node == fn) {
1369                                         RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate);
1370                                         w->node = pn;
1371                                         w->state = nstate;
1372                                 }
1373                         } else {
1374                                 if (w->root == fn) {
1375                                         w->root = child;
1376                                         RT6_TRACE("W %p adjusted by delroot 2\n", w);
1377                                 }
1378                                 if (w->node == fn) {
1379                                         w->node = child;
1380                                         if (children&2) {
1381                                                 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1382                                                 w->state = w->state >= FWS_R ? FWS_U : FWS_INIT;
1383                                         } else {
1384                                                 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1385                                                 w->state = w->state >= FWS_C ? FWS_U : FWS_INIT;
1386                                         }
1387                                 }
1388                         }
1389                 }
1390                 read_unlock(&net->ipv6.fib6_walker_lock);
1391 
1392                 node_free(fn);
1393                 if (pn->fn_flags & RTN_RTINFO || FIB6_SUBTREE(pn))
1394                         return pn;
1395 
1396                 rt6_release(pn->leaf);
1397                 pn->leaf = NULL;
1398                 fn = pn;
1399         }
1400 }
1401 
1402 static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp,
1403                            struct nl_info *info)
1404 {
1405         struct fib6_walker *w;
1406         struct rt6_info *rt = *rtp;
1407         struct net *net = info->nl_net;
1408 
1409         RT6_TRACE("fib6_del_route\n");
1410 
1411         /* Unlink it */
1412         *rtp = rt->dst.rt6_next;
1413         rt->rt6i_node = NULL;
1414         net->ipv6.rt6_stats->fib_rt_entries--;
1415         net->ipv6.rt6_stats->fib_discarded_routes++;
1416 
1417         /* Reset round-robin state, if necessary */
1418         if (fn->rr_ptr == rt)
1419                 fn->rr_ptr = NULL;
1420 
1421         /* Remove this entry from other siblings */
1422         if (rt->rt6i_nsiblings) {
1423                 struct rt6_info *sibling, *next_sibling;
1424 
1425                 list_for_each_entry_safe(sibling, next_sibling,
1426                                          &rt->rt6i_siblings, rt6i_siblings)
1427                         sibling->rt6i_nsiblings--;
1428                 rt->rt6i_nsiblings = 0;
1429                 list_del_init(&rt->rt6i_siblings);
1430         }
1431 
1432         /* Adjust walkers */
1433         read_lock(&net->ipv6.fib6_walker_lock);
1434         FOR_WALKERS(net, w) {
1435                 if (w->state == FWS_C && w->leaf == rt) {
1436                         RT6_TRACE("walker %p adjusted by delroute\n", w);
1437                         w->leaf = rt->dst.rt6_next;
1438                         if (!w->leaf)
1439                                 w->state = FWS_U;
1440                 }
1441         }
1442         read_unlock(&net->ipv6.fib6_walker_lock);
1443 
1444         rt->dst.rt6_next = NULL;
1445 
1446         /* If it was last route, expunge its radix tree node */
1447         if (!fn->leaf) {
1448                 fn->fn_flags &= ~RTN_RTINFO;
1449                 net->ipv6.rt6_stats->fib_route_nodes--;
1450                 fn = fib6_repair_tree(net, fn);
1451         }
1452 
1453         fib6_purge_rt(rt, fn, net);
1454 
1455         if (!info->skip_notify)
1456                 inet6_rt_notify(RTM_DELROUTE, rt, info, 0);
1457         rt6_release(rt);
1458 }
1459 
1460 int fib6_del(struct rt6_info *rt, struct nl_info *info)
1461 {
1462         struct net *net = info->nl_net;
1463         struct fib6_node *fn = rt->rt6i_node;
1464         struct rt6_info **rtp;
1465 
1466 #if RT6_DEBUG >= 2
1467         if (rt->dst.obsolete > 0) {
1468                 WARN_ON(fn);
1469                 return -ENOENT;
1470         }
1471 #endif
1472         if (!fn || rt == net->ipv6.ip6_null_entry)
1473                 return -ENOENT;
1474 
1475         WARN_ON(!(fn->fn_flags & RTN_RTINFO));
1476 
1477         if (!(rt->rt6i_flags & RTF_CACHE)) {
1478                 struct fib6_node *pn = fn;
1479 #ifdef CONFIG_IPV6_SUBTREES
1480                 /* clones of this route might be in another subtree */
1481                 if (rt->rt6i_src.plen) {
1482                         while (!(pn->fn_flags & RTN_ROOT))
1483                                 pn = pn->parent;
1484                         pn = pn->parent;
1485                 }
1486 #endif
1487                 fib6_prune_clones(info->nl_net, pn);
1488         }
1489 
1490         /*
1491          *      Walk the leaf entries looking for ourself
1492          */
1493 
1494         for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->dst.rt6_next) {
1495                 if (*rtp == rt) {
1496                         fib6_del_route(fn, rtp, info);
1497                         return 0;
1498                 }
1499         }
1500         return -ENOENT;
1501 }
1502 
1503 /*
1504  *      Tree traversal function.
1505  *
1506  *      Certainly, it is not interrupt safe.
1507  *      However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1508  *      It means, that we can modify tree during walking
1509  *      and use this function for garbage collection, clone pruning,
1510  *      cleaning tree when a device goes down etc. etc.
1511  *
1512  *      It guarantees that every node will be traversed,
1513  *      and that it will be traversed only once.
1514  *
1515  *      Callback function w->func may return:
1516  *      0 -> continue walking.
1517  *      positive value -> walking is suspended (used by tree dumps,
1518  *      and probably by gc, if it will be split to several slices)
1519  *      negative value -> terminate walking.
1520  *
1521  *      The function itself returns:
1522  *      0   -> walk is complete.
1523  *      >0  -> walk is incomplete (i.e. suspended)
1524  *      <0  -> walk is terminated by an error.
1525  */
1526 
1527 static int fib6_walk_continue(struct fib6_walker *w)
1528 {
1529         struct fib6_node *fn, *pn;
1530 
1531         for (;;) {
1532                 fn = w->node;
1533                 if (!fn)
1534                         return 0;
1535 
1536                 if (w->prune && fn != w->root &&
1537                     fn->fn_flags & RTN_RTINFO && w->state < FWS_C) {
1538                         w->state = FWS_C;
1539                         w->leaf = fn->leaf;
1540                 }
1541                 switch (w->state) {
1542 #ifdef CONFIG_IPV6_SUBTREES
1543                 case FWS_S:
1544                         if (FIB6_SUBTREE(fn)) {
1545                                 w->node = FIB6_SUBTREE(fn);
1546                                 continue;
1547                         }
1548                         w->state = FWS_L;
1549 #endif
1550                 case FWS_L:
1551                         if (fn->left) {
1552                                 w->node = fn->left;
1553                                 w->state = FWS_INIT;
1554                                 continue;
1555                         }
1556                         w->state = FWS_R;
1557                 case FWS_R:
1558                         if (fn->right) {
1559                                 w->node = fn->right;
1560                                 w->state = FWS_INIT;
1561                                 continue;
1562                         }
1563                         w->state = FWS_C;
1564                         w->leaf = fn->leaf;
1565                 case FWS_C:
1566                         if (w->leaf && fn->fn_flags & RTN_RTINFO) {
1567                                 int err;
1568 
1569                                 if (w->skip) {
1570                                         w->skip--;
1571                                         goto skip;
1572                                 }
1573 
1574                                 err = w->func(w);
1575                                 if (err)
1576                                         return err;
1577 
1578                                 w->count++;
1579                                 continue;
1580                         }
1581 skip:
1582                         w->state = FWS_U;
1583                 case FWS_U:
1584                         if (fn == w->root)
1585                                 return 0;
1586                         pn = fn->parent;
1587                         w->node = pn;
1588 #ifdef CONFIG_IPV6_SUBTREES
1589                         if (FIB6_SUBTREE(pn) == fn) {
1590                                 WARN_ON(!(fn->fn_flags & RTN_ROOT));
1591                                 w->state = FWS_L;
1592                                 continue;
1593                         }
1594 #endif
1595                         if (pn->left == fn) {
1596                                 w->state = FWS_R;
1597                                 continue;
1598                         }
1599                         if (pn->right == fn) {
1600                                 w->state = FWS_C;
1601                                 w->leaf = w->node->leaf;
1602                                 continue;
1603                         }
1604 #if RT6_DEBUG >= 2
1605                         WARN_ON(1);
1606 #endif
1607                 }
1608         }
1609 }
1610 
1611 static int fib6_walk(struct net *net, struct fib6_walker *w)
1612 {
1613         int res;
1614 
1615         w->state = FWS_INIT;
1616         w->node = w->root;
1617 
1618         fib6_walker_link(net, w);
1619         res = fib6_walk_continue(w);
1620         if (res <= 0)
1621                 fib6_walker_unlink(net, w);
1622         return res;
1623 }
1624 
1625 static int fib6_clean_node(struct fib6_walker *w)
1626 {
1627         int res;
1628         struct rt6_info *rt;
1629         struct fib6_cleaner *c = container_of(w, struct fib6_cleaner, w);
1630         struct nl_info info = {
1631                 .nl_net = c->net,
1632         };
1633 
1634         if (c->sernum != FIB6_NO_SERNUM_CHANGE &&
1635             w->node->fn_sernum != c->sernum)
1636                 w->node->fn_sernum = c->sernum;
1637 
1638         if (!c->func) {
1639                 WARN_ON_ONCE(c->sernum == FIB6_NO_SERNUM_CHANGE);
1640                 w->leaf = NULL;
1641                 return 0;
1642         }
1643 
1644         for (rt = w->leaf; rt; rt = rt->dst.rt6_next) {
1645                 res = c->func(rt, c->arg);
1646                 if (res < 0) {
1647                         w->leaf = rt;
1648                         res = fib6_del(rt, &info);
1649                         if (res) {
1650 #if RT6_DEBUG >= 2
1651                                 pr_debug("%s: del failed: rt=%p@%p err=%d\n",
1652                                          __func__, rt, rt->rt6i_node, res);
1653 #endif
1654                                 continue;
1655                         }
1656                         return 0;
1657                 }
1658                 WARN_ON(res != 0);
1659         }
1660         w->leaf = rt;
1661         return 0;
1662 }
1663 
1664 /*
1665  *      Convenient frontend to tree walker.
1666  *
1667  *      func is called on each route.
1668  *              It may return -1 -> delete this route.
1669  *                            0  -> continue walking
1670  *
1671  *      prune==1 -> only immediate children of node (certainly,
1672  *      ignoring pure split nodes) will be scanned.
1673  */
1674 
1675 static void fib6_clean_tree(struct net *net, struct fib6_node *root,
1676                             int (*func)(struct rt6_info *, void *arg),
1677                             bool prune, int sernum, void *arg)
1678 {
1679         struct fib6_cleaner c;
1680 
1681         c.w.root = root;
1682         c.w.func = fib6_clean_node;
1683         c.w.prune = prune;
1684         c.w.count = 0;
1685         c.w.skip = 0;
1686         c.func = func;
1687         c.sernum = sernum;
1688         c.arg = arg;
1689         c.net = net;
1690 
1691         fib6_walk(net, &c.w);
1692 }
1693 
1694 static void __fib6_clean_all(struct net *net,
1695                              int (*func)(struct rt6_info *, void *),
1696                              int sernum, void *arg)
1697 {
1698         struct fib6_table *table;
1699         struct hlist_head *head;
1700         unsigned int h;
1701 
1702         rcu_read_lock();
1703         for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
1704                 head = &net->ipv6.fib_table_hash[h];
1705                 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
1706                         write_lock_bh(&table->tb6_lock);
1707                         fib6_clean_tree(net, &table->tb6_root,
1708                                         func, false, sernum, arg);
1709                         write_unlock_bh(&table->tb6_lock);
1710                 }
1711         }
1712         rcu_read_unlock();
1713 }
1714 
1715 void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *),
1716                     void *arg)
1717 {
1718         __fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg);
1719 }
1720 
1721 static int fib6_prune_clone(struct rt6_info *rt, void *arg)
1722 {
1723         if (rt->rt6i_flags & RTF_CACHE) {
1724                 RT6_TRACE("pruning clone %p\n", rt);
1725                 return -1;
1726         }
1727 
1728         return 0;
1729 }
1730 
1731 static void fib6_prune_clones(struct net *net, struct fib6_node *fn)
1732 {
1733         fib6_clean_tree(net, fn, fib6_prune_clone, true,
1734                         FIB6_NO_SERNUM_CHANGE, NULL);
1735 }
1736 
1737 static void fib6_flush_trees(struct net *net)
1738 {
1739         int new_sernum = fib6_new_sernum(net);
1740 
1741         __fib6_clean_all(net, NULL, new_sernum, NULL);
1742 }
1743 
1744 /*
1745  *      Garbage collection
1746  */
1747 
1748 struct fib6_gc_args
1749 {
1750         int                     timeout;
1751         int                     more;
1752 };
1753 
1754 static int fib6_age(struct rt6_info *rt, void *arg)
1755 {
1756         struct fib6_gc_args *gc_args = arg;
1757         unsigned long now = jiffies;
1758 
1759         /*
1760          *      check addrconf expiration here.
1761          *      Routes are expired even if they are in use.
1762          *
1763          *      Also age clones. Note, that clones are aged out
1764          *      only if they are not in use now.
1765          */
1766 
1767         if (rt->rt6i_flags & RTF_EXPIRES && rt->dst.expires) {
1768                 if (time_after(now, rt->dst.expires)) {
1769                         RT6_TRACE("expiring %p\n", rt);
1770                         return -1;
1771                 }
1772                 gc_args->more++;
1773         } else if (rt->rt6i_flags & RTF_CACHE) {
1774                 if (atomic_read(&rt->dst.__refcnt) == 0 &&
1775                     time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
1776                         RT6_TRACE("aging clone %p\n", rt);
1777                         return -1;
1778                 } else if (rt->rt6i_flags & RTF_GATEWAY) {
1779                         struct neighbour *neigh;
1780                         __u8 neigh_flags = 0;
1781 
1782                         neigh = dst_neigh_lookup(&rt->dst, &rt->rt6i_gateway);
1783                         if (neigh) {
1784                                 neigh_flags = neigh->flags;
1785                                 neigh_release(neigh);
1786                         }
1787                         if (!(neigh_flags & NTF_ROUTER)) {
1788                                 RT6_TRACE("purging route %p via non-router but gateway\n",
1789                                           rt);
1790                                 return -1;
1791                         }
1792                 }
1793                 gc_args->more++;
1794         }
1795 
1796         return 0;
1797 }
1798 
1799 void fib6_run_gc(unsigned long expires, struct net *net, bool force)
1800 {
1801         struct fib6_gc_args gc_args;
1802         unsigned long now;
1803 
1804         if (force) {
1805                 spin_lock_bh(&net->ipv6.fib6_gc_lock);
1806         } else if (!spin_trylock_bh(&net->ipv6.fib6_gc_lock)) {
1807                 mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ);
1808                 return;
1809         }
1810         gc_args.timeout = expires ? (int)expires :
1811                           net->ipv6.sysctl.ip6_rt_gc_interval;
1812 
1813         gc_args.more = icmp6_dst_gc();
1814 
1815         fib6_clean_all(net, fib6_age, &gc_args);
1816         now = jiffies;
1817         net->ipv6.ip6_rt_last_gc = now;
1818 
1819         if (gc_args.more)
1820                 mod_timer(&net->ipv6.ip6_fib_timer,
1821                           round_jiffies(now
1822                                         + net->ipv6.sysctl.ip6_rt_gc_interval));
1823         else
1824                 del_timer(&net->ipv6.ip6_fib_timer);
1825         spin_unlock_bh(&net->ipv6.fib6_gc_lock);
1826 }
1827 
1828 static void fib6_gc_timer_cb(unsigned long arg)
1829 {
1830         fib6_run_gc(0, (struct net *)arg, true);
1831 }
1832 
1833 static int __net_init fib6_net_init(struct net *net)
1834 {
1835         size_t size = sizeof(struct hlist_head) * FIB6_TABLE_HASHSZ;
1836 
1837         spin_lock_init(&net->ipv6.fib6_gc_lock);
1838         rwlock_init(&net->ipv6.fib6_walker_lock);
1839         INIT_LIST_HEAD(&net->ipv6.fib6_walkers);
1840         setup_timer(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, (unsigned long)net);
1841 
1842         net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL);
1843         if (!net->ipv6.rt6_stats)
1844                 goto out_timer;
1845 
1846         /* Avoid false sharing : Use at least a full cache line */
1847         size = max_t(size_t, size, L1_CACHE_BYTES);
1848 
1849         net->ipv6.fib_table_hash = kzalloc(size, GFP_KERNEL);
1850         if (!net->ipv6.fib_table_hash)
1851                 goto out_rt6_stats;
1852 
1853         net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl),
1854                                           GFP_KERNEL);
1855         if (!net->ipv6.fib6_main_tbl)
1856                 goto out_fib_table_hash;
1857 
1858         net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN;
1859         net->ipv6.fib6_main_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry;
1860         net->ipv6.fib6_main_tbl->tb6_root.fn_flags =
1861                 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1862         inet_peer_base_init(&net->ipv6.fib6_main_tbl->tb6_peers);
1863 
1864 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1865         net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl),
1866                                            GFP_KERNEL);
1867         if (!net->ipv6.fib6_local_tbl)
1868                 goto out_fib6_main_tbl;
1869         net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL;
1870         net->ipv6.fib6_local_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry;
1871         net->ipv6.fib6_local_tbl->tb6_root.fn_flags =
1872                 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1873         inet_peer_base_init(&net->ipv6.fib6_local_tbl->tb6_peers);
1874 #endif
1875         fib6_tables_init(net);
1876 
1877         return 0;
1878 
1879 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1880 out_fib6_main_tbl:
1881         kfree(net->ipv6.fib6_main_tbl);
1882 #endif
1883 out_fib_table_hash:
1884         kfree(net->ipv6.fib_table_hash);
1885 out_rt6_stats:
1886         kfree(net->ipv6.rt6_stats);
1887 out_timer:
1888         return -ENOMEM;
1889 }
1890 
1891 static void fib6_net_exit(struct net *net)
1892 {
1893         rt6_ifdown(net, NULL);
1894         del_timer_sync(&net->ipv6.ip6_fib_timer);
1895 
1896 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1897         inetpeer_invalidate_tree(&net->ipv6.fib6_local_tbl->tb6_peers);
1898         kfree(net->ipv6.fib6_local_tbl);
1899 #endif
1900         inetpeer_invalidate_tree(&net->ipv6.fib6_main_tbl->tb6_peers);
1901         kfree(net->ipv6.fib6_main_tbl);
1902         kfree(net->ipv6.fib_table_hash);
1903         kfree(net->ipv6.rt6_stats);
1904 }
1905 
1906 static struct pernet_operations fib6_net_ops = {
1907         .init = fib6_net_init,
1908         .exit = fib6_net_exit,
1909 };
1910 
1911 int __init fib6_init(void)
1912 {
1913         int ret = -ENOMEM;
1914 
1915         fib6_node_kmem = kmem_cache_create("fib6_nodes",
1916                                            sizeof(struct fib6_node),
1917                                            0, SLAB_HWCACHE_ALIGN,
1918                                            NULL);
1919         if (!fib6_node_kmem)
1920                 goto out;
1921 
1922         ret = register_pernet_subsys(&fib6_net_ops);
1923         if (ret)
1924                 goto out_kmem_cache_create;
1925 
1926         ret = __rtnl_register(PF_INET6, RTM_GETROUTE, NULL, inet6_dump_fib,
1927                               NULL);
1928         if (ret)
1929                 goto out_unregister_subsys;
1930 
1931         __fib6_flush_trees = fib6_flush_trees;
1932 out:
1933         return ret;
1934 
1935 out_unregister_subsys:
1936         unregister_pernet_subsys(&fib6_net_ops);
1937 out_kmem_cache_create:
1938         kmem_cache_destroy(fib6_node_kmem);
1939         goto out;
1940 }
1941 
1942 void fib6_gc_cleanup(void)
1943 {
1944         unregister_pernet_subsys(&fib6_net_ops);
1945         kmem_cache_destroy(fib6_node_kmem);
1946 }
1947 
1948 #ifdef CONFIG_PROC_FS
1949 
1950 struct ipv6_route_iter {
1951         struct seq_net_private p;
1952         struct fib6_walker w;
1953         loff_t skip;
1954         struct fib6_table *tbl;
1955         int sernum;
1956 };
1957 
1958 static int ipv6_route_seq_show(struct seq_file *seq, void *v)
1959 {
1960         struct rt6_info *rt = v;
1961         struct ipv6_route_iter *iter = seq->private;
1962 
1963         seq_printf(seq, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
1964 
1965 #ifdef CONFIG_IPV6_SUBTREES
1966         seq_printf(seq, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
1967 #else
1968         seq_puts(seq, "00000000000000000000000000000000 00 ");
1969 #endif
1970         if (rt->rt6i_flags & RTF_GATEWAY)
1971                 seq_printf(seq, "%pi6", &rt->rt6i_gateway);
1972         else
1973                 seq_puts(seq, "00000000000000000000000000000000");
1974 
1975         seq_printf(seq, " %08x %08x %08x %08x %8s\n",
1976                    rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
1977                    rt->dst.__use, rt->rt6i_flags,
1978                    rt->dst.dev ? rt->dst.dev->name : "");
1979         iter->w.leaf = NULL;
1980         return 0;
1981 }
1982 
1983 static int ipv6_route_yield(struct fib6_walker *w)
1984 {
1985         struct ipv6_route_iter *iter = w->args;
1986 
1987         if (!iter->skip)
1988                 return 1;
1989 
1990         do {
1991                 iter->w.leaf = iter->w.leaf->dst.rt6_next;
1992                 iter->skip--;
1993                 if (!iter->skip && iter->w.leaf)
1994                         return 1;
1995         } while (iter->w.leaf);
1996 
1997         return 0;
1998 }
1999 
2000 static void ipv6_route_seq_setup_walk(struct ipv6_route_iter *iter,
2001                                       struct net *net)
2002 {
2003         memset(&iter->w, 0, sizeof(iter->w));
2004         iter->w.func = ipv6_route_yield;
2005         iter->w.root = &iter->tbl->tb6_root;
2006         iter->w.state = FWS_INIT;
2007         iter->w.node = iter->w.root;
2008         iter->w.args = iter;
2009         iter->sernum = iter->w.root->fn_sernum;
2010         INIT_LIST_HEAD(&iter->w.lh);
2011         fib6_walker_link(net, &iter->w);
2012 }
2013 
2014 static struct fib6_table *ipv6_route_seq_next_table(struct fib6_table *tbl,
2015                                                     struct net *net)
2016 {
2017         unsigned int h;
2018         struct hlist_node *node;
2019 
2020         if (tbl) {
2021                 h = (tbl->tb6_id & (FIB6_TABLE_HASHSZ - 1)) + 1;
2022                 node = rcu_dereference_bh(hlist_next_rcu(&tbl->tb6_hlist));
2023         } else {
2024                 h = 0;
2025                 node = NULL;
2026         }
2027 
2028         while (!node && h < FIB6_TABLE_HASHSZ) {
2029                 node = rcu_dereference_bh(
2030                         hlist_first_rcu(&net->ipv6.fib_table_hash[h++]));
2031         }
2032         return hlist_entry_safe(node, struct fib6_table, tb6_hlist);
2033 }
2034 
2035 static void ipv6_route_check_sernum(struct ipv6_route_iter *iter)
2036 {
2037         if (iter->sernum != iter->w.root->fn_sernum) {
2038                 iter->sernum = iter->w.root->fn_sernum;
2039                 iter->w.state = FWS_INIT;
2040                 iter->w.node = iter->w.root;
2041                 WARN_ON(iter->w.skip);
2042                 iter->w.skip = iter->w.count;
2043         }
2044 }
2045 
2046 static void *ipv6_route_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2047 {
2048         int r;
2049         struct rt6_info *n;
2050         struct net *net = seq_file_net(seq);
2051         struct ipv6_route_iter *iter = seq->private;
2052 
2053         if (!v)
2054                 goto iter_table;
2055 
2056         n = ((struct rt6_info *)v)->dst.rt6_next;
2057         if (n) {
2058                 ++*pos;
2059                 return n;
2060         }
2061 
2062 iter_table:
2063         ipv6_route_check_sernum(iter);
2064         read_lock(&iter->tbl->tb6_lock);
2065         r = fib6_walk_continue(&iter->w);
2066         read_unlock(&iter->tbl->tb6_lock);
2067         if (r > 0) {
2068                 if (v)
2069                         ++*pos;
2070                 return iter->w.leaf;
2071         } else if (r < 0) {
2072                 fib6_walker_unlink(net, &iter->w);
2073                 return NULL;
2074         }
2075         fib6_walker_unlink(net, &iter->w);
2076 
2077         iter->tbl = ipv6_route_seq_next_table(iter->tbl, net);
2078         if (!iter->tbl)
2079                 return NULL;
2080 
2081         ipv6_route_seq_setup_walk(iter, net);
2082         goto iter_table;
2083 }
2084 
2085 static void *ipv6_route_seq_start(struct seq_file *seq, loff_t *pos)
2086         __acquires(RCU_BH)
2087 {
2088         struct net *net = seq_file_net(seq);
2089         struct ipv6_route_iter *iter = seq->private;
2090 
2091         rcu_read_lock_bh();
2092         iter->tbl = ipv6_route_seq_next_table(NULL, net);
2093         iter->skip = *pos;
2094 
2095         if (iter->tbl) {
2096                 ipv6_route_seq_setup_walk(iter, net);
2097                 return ipv6_route_seq_next(seq, NULL, pos);
2098         } else {
2099                 return NULL;
2100         }
2101 }
2102 
2103 static bool ipv6_route_iter_active(struct ipv6_route_iter *iter)
2104 {
2105         struct fib6_walker *w = &iter->w;
2106         return w->node && !(w->state == FWS_U && w->node == w->root);
2107 }
2108 
2109 static void ipv6_route_seq_stop(struct seq_file *seq, void *v)
2110         __releases(RCU_BH)
2111 {
2112         struct net *net = seq_file_net(seq);
2113         struct ipv6_route_iter *iter = seq->private;
2114 
2115         if (ipv6_route_iter_active(iter))
2116                 fib6_walker_unlink(net, &iter->w);
2117 
2118         rcu_read_unlock_bh();
2119 }
2120 
2121 static const struct seq_operations ipv6_route_seq_ops = {
2122         .start  = ipv6_route_seq_start,
2123         .next   = ipv6_route_seq_next,
2124         .stop   = ipv6_route_seq_stop,
2125         .show   = ipv6_route_seq_show
2126 };
2127 
2128 int ipv6_route_open(struct inode *inode, struct file *file)
2129 {
2130         return seq_open_net(inode, file, &ipv6_route_seq_ops,
2131                             sizeof(struct ipv6_route_iter));
2132 }
2133 
2134 #endif /* CONFIG_PROC_FS */
2135 

~ [ 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