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