]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - contrib/unbound/edns-subnet/subnetmod.c
MFV r337190: 9486 reduce memory used by device removal on fragmented pools
[FreeBSD/FreeBSD.git] / contrib / unbound / edns-subnet / subnetmod.c
1 /*
2  * edns-subnet/subnetmod.c - edns subnet module. Must be called before validator
3  * and iterator.
4  *
5  * Copyright (c) 2013, NLnet Labs. All rights reserved.
6  *
7  * This software is open source.
8  * 
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 
13  * Redistributions of source code must retain the above copyright notice,
14  * this list of conditions and the following disclaimer.
15  * 
16  * Redistributions in binary form must reproduce the above copyright notice,
17  * this list of conditions and the following disclaimer in the documentation
18  * and/or other materials provided with the distribution.
19  * 
20  * Neither the name of the NLNET LABS nor the names of its contributors may
21  * be used to endorse or promote products derived from this software without
22  * specific prior written permission.
23  * 
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
25  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
26  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
27  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
28  * HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
29  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
30  * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
31  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
32  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
33  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
34  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35  */
36  /**
37  * \file
38  * subnet module for unbound.
39  */
40
41 #include "config.h"
42
43 #ifdef CLIENT_SUBNET /* keeps splint happy */
44
45 #include "edns-subnet/subnetmod.h"
46 #include "edns-subnet/edns-subnet.h"
47 #include "edns-subnet/addrtree.h"
48 #include "edns-subnet/subnet-whitelist.h"
49
50 #include "services/mesh.h"
51 #include "services/cache/dns.h"
52 #include "util/module.h"
53 #include "util/regional.h"
54 #include "util/storage/slabhash.h"
55 #include "util/config_file.h"
56 #include "util/data/msgreply.h"
57 #include "sldns/sbuffer.h"
58
59 #define ECS_MAX_TREESIZE 100
60
61 /** externally called */
62 void 
63 subnet_data_delete(void *d, void *ATTR_UNUSED(arg))
64 {
65         struct subnet_msg_cache_data *r;
66         r = (struct subnet_msg_cache_data*)d;
67         addrtree_delete(r->tree4);
68         addrtree_delete(r->tree6);
69         free(r);
70 }
71
72 /** externally called */
73 size_t 
74 msg_cache_sizefunc(void *k, void *d)
75 {
76         struct msgreply_entry *q = (struct msgreply_entry*)k;
77         struct subnet_msg_cache_data *r = (struct subnet_msg_cache_data*)d;
78         size_t s = sizeof(struct msgreply_entry) 
79                 + sizeof(struct subnet_msg_cache_data)
80                 + q->key.qname_len + lock_get_mem(&q->entry.lock);
81         s += addrtree_size(r->tree4);
82         s += addrtree_size(r->tree6);
83         return s;
84 }
85
86 /** new query for ecs module */
87 static int
88 subnet_new_qstate(struct module_qstate *qstate, int id)
89 {
90         struct subnet_qstate *sq = (struct subnet_qstate*)regional_alloc(
91                 qstate->region, sizeof(struct subnet_qstate));
92         if(!sq) 
93                 return 0;
94         qstate->minfo[id] = sq;
95         memset(sq, 0, sizeof(*sq));
96         return 1;
97 }
98
99 /** Add ecs struct to edns list, after parsing it to wire format. */
100 static void
101 ecs_opt_list_append(struct ecs_data* ecs, struct edns_option** list,
102         struct module_qstate *qstate)
103 {
104         size_t sn_octs, sn_octs_remainder;
105         sldns_buffer* buf = qstate->env->scratch_buffer;
106
107         if(ecs->subnet_validdata) {
108                 log_assert(ecs->subnet_addr_fam == EDNSSUBNET_ADDRFAM_IP4 || 
109                         ecs->subnet_addr_fam == EDNSSUBNET_ADDRFAM_IP6);
110                 log_assert(ecs->subnet_addr_fam != EDNSSUBNET_ADDRFAM_IP4 || 
111                         ecs->subnet_source_mask <=  INET_SIZE*8);
112                 log_assert(ecs->subnet_addr_fam != EDNSSUBNET_ADDRFAM_IP6 || 
113                         ecs->subnet_source_mask <= INET6_SIZE*8);
114
115                 sn_octs = ecs->subnet_source_mask / 8;
116                 sn_octs_remainder =
117                         (size_t)((ecs->subnet_source_mask % 8)>0?1:0);
118                 
119                 log_assert(sn_octs + sn_octs_remainder <= INET6_SIZE);
120                 
121                 sldns_buffer_clear(buf);
122                 sldns_buffer_write_u16(buf, ecs->subnet_addr_fam);
123                 sldns_buffer_write_u8(buf, ecs->subnet_source_mask);
124                 sldns_buffer_write_u8(buf, ecs->subnet_scope_mask);
125                 sldns_buffer_write(buf, ecs->subnet_addr, sn_octs);
126                 if(sn_octs_remainder)
127                         sldns_buffer_write_u8(buf, ecs->subnet_addr[sn_octs] & 
128                                 ~(0xFF >> (ecs->subnet_source_mask % 8)));
129                 sldns_buffer_flip(buf);
130
131                 edns_opt_list_append(list,
132                                 qstate->env->cfg->client_subnet_opcode,
133                                 sn_octs + sn_octs_remainder + 4,
134                                 sldns_buffer_begin(buf), qstate->region);
135         }
136 }
137
138 int ecs_whitelist_check(struct query_info* qinfo,
139         uint16_t ATTR_UNUSED(flags), struct module_qstate* qstate,
140         struct sockaddr_storage* addr, socklen_t addrlen,
141         uint8_t* ATTR_UNUSED(zone), size_t ATTR_UNUSED(zonelen),
142         struct regional* ATTR_UNUSED(region), int id, void* ATTR_UNUSED(cbargs))
143 {
144         struct subnet_qstate *sq;
145         struct subnet_env *sn_env;
146         
147         if(!(sq=(struct subnet_qstate*)qstate->minfo[id]))
148                 return 1;
149         sn_env = (struct subnet_env*)qstate->env->modinfo[id];
150
151         /* Cache by default, might be disabled after parsing EDNS option
152          * received from nameserver. */
153         qstate->no_cache_store = 0;
154
155         if(sq->ecs_server_out.subnet_validdata && ((sq->subnet_downstream &&
156                 qstate->env->cfg->client_subnet_always_forward) ||
157                 ecs_is_whitelisted(sn_env->whitelist, 
158                 addr, addrlen, qinfo->qname, qinfo->qname_len,
159                 qinfo->qclass))) {
160                 /* Address on whitelist or client query contains ECS option, we
161                  * want to sent out ECS. Only add option if it is not already
162                  * set. */
163                 if(!(sq->subnet_sent)) {
164                         ecs_opt_list_append(&sq->ecs_server_out,
165                                 &qstate->edns_opts_back_out, qstate);
166                         sq->subnet_sent = 1;
167                 }
168         }
169         else if(sq->subnet_sent) {
170                 /* Outgoing ECS option is set, but we don't want to sent it to
171                  * this address, remove option. */
172                 edns_opt_list_remove(&qstate->edns_opts_back_out,
173                         qstate->env->cfg->client_subnet_opcode);
174                 sq->subnet_sent = 0;
175         }
176         return 1;
177 }
178
179
180 int
181 subnetmod_init(struct module_env *env, int id)
182 {
183         struct subnet_env *sn_env = (struct subnet_env*)calloc(1,
184                 sizeof(struct subnet_env));
185         if(!sn_env) {
186                 log_err("malloc failure");
187                 return 0;
188         }
189         alloc_init(&sn_env->alloc, NULL, 0);
190         env->modinfo[id] = (void*)sn_env;
191         /* Copy msg_cache settings */
192         sn_env->subnet_msg_cache = slabhash_create(env->cfg->msg_cache_slabs,
193                 HASH_DEFAULT_STARTARRAY, env->cfg->msg_cache_size,
194                 msg_cache_sizefunc, query_info_compare, query_entry_delete,
195                 subnet_data_delete, NULL);
196         if(!sn_env->subnet_msg_cache) {
197                 log_err("subnet: could not create cache");
198                 free(sn_env);
199                 env->modinfo[id] = NULL;
200                 return 0;
201         }
202         /* whitelist for edns subnet capable servers */
203         sn_env->whitelist = ecs_whitelist_create();
204         if(!sn_env->whitelist ||
205                 !ecs_whitelist_apply_cfg(sn_env->whitelist, env->cfg)) {
206                 log_err("subnet: could not create ECS whitelist");
207                 slabhash_delete(sn_env->subnet_msg_cache);
208                 free(sn_env);
209                 env->modinfo[id] = NULL;
210                 return 0;
211         }
212
213         verbose(VERB_QUERY, "subnet: option registered (%d)",
214                 env->cfg->client_subnet_opcode);
215         /* Create new mesh state for all queries. */
216         env->unique_mesh = 1;
217         if(!edns_register_option(env->cfg->client_subnet_opcode,
218                 env->cfg->client_subnet_always_forward /* bypass cache */,
219                 0 /* no aggregation */, env)) {
220                 log_err("subnet: could not register opcode");
221                 ecs_whitelist_delete(sn_env->whitelist);
222                 slabhash_delete(sn_env->subnet_msg_cache);
223                 free(sn_env);
224                 env->modinfo[id] = NULL;
225                 return 0;
226         }
227         inplace_cb_register((void*)ecs_whitelist_check, inplace_cb_query, NULL,
228                 env, id);
229         inplace_cb_register((void*)ecs_edns_back_parsed,
230                 inplace_cb_edns_back_parsed, NULL, env, id);
231         inplace_cb_register((void*)ecs_query_response,
232                 inplace_cb_query_response, NULL, env, id);
233         lock_rw_init(&sn_env->biglock);
234         return 1;
235 }
236
237 void
238 subnetmod_deinit(struct module_env *env, int id)
239 {
240         struct subnet_env *sn_env;
241         if(!env || !env->modinfo[id])
242                 return;
243         sn_env = (struct subnet_env*)env->modinfo[id];
244         lock_rw_destroy(&sn_env->biglock);
245         inplace_cb_delete(env, inplace_cb_edns_back_parsed, id);
246         inplace_cb_delete(env, inplace_cb_query, id);
247         inplace_cb_delete(env, inplace_cb_query_response, id);
248         ecs_whitelist_delete(sn_env->whitelist);
249         slabhash_delete(sn_env->subnet_msg_cache);
250         alloc_clear(&sn_env->alloc);
251         free(sn_env);
252         env->modinfo[id] = NULL;
253 }
254
255 /** Tells client that upstream has no/improper support */
256 static void
257 cp_edns_bad_response(struct ecs_data *target, struct ecs_data *source)
258 {
259         target->subnet_scope_mask  = 0;
260         target->subnet_source_mask = source->subnet_source_mask;
261         target->subnet_addr_fam    = source->subnet_addr_fam;
262         memcpy(target->subnet_addr, source->subnet_addr, INET6_SIZE);
263         target->subnet_validdata = 1;
264 }
265
266 static void
267 delfunc(void *envptr, void *elemptr) {
268         struct reply_info *elem = (struct reply_info *)elemptr;
269         struct subnet_env *env = (struct subnet_env *)envptr;
270         reply_info_parsedelete(elem, &env->alloc);
271 }
272
273 static size_t
274 sizefunc(void *elemptr) {
275         struct reply_info *elem  = (struct reply_info *)elemptr;
276         return sizeof (struct reply_info) - sizeof (struct rrset_ref)
277                 + elem->rrset_count * sizeof (struct rrset_ref)
278                 + elem->rrset_count * sizeof (struct ub_packed_rrset_key *);
279 }
280
281 /**
282  * Select tree from cache entry based on edns data.
283  * If for address family not present it will create a new one.
284  * NULL on failure to create. */
285 static struct addrtree* 
286 get_tree(struct subnet_msg_cache_data *data, struct ecs_data *edns, 
287         struct subnet_env *env, struct config_file* cfg)
288 {
289         struct addrtree *tree;
290         if (edns->subnet_addr_fam == EDNSSUBNET_ADDRFAM_IP4) {
291                 if (!data->tree4)
292                         data->tree4 = addrtree_create(
293                                 cfg->max_client_subnet_ipv4, &delfunc,
294                                 &sizefunc, env, ECS_MAX_TREESIZE);
295                 tree = data->tree4;
296         } else {
297                 if (!data->tree6)
298                         data->tree6 = addrtree_create(
299                                 cfg->max_client_subnet_ipv6, &delfunc,
300                                 &sizefunc, env, ECS_MAX_TREESIZE);
301                 tree = data->tree6;
302         }
303         return tree;
304 }
305
306 static void
307 update_cache(struct module_qstate *qstate, int id)
308 {
309         struct msgreply_entry *mrep_entry;
310         struct addrtree *tree;
311         struct reply_info *rep;
312         struct query_info qinf;
313         struct subnet_env *sne = qstate->env->modinfo[id];
314         struct subnet_qstate *sq = (struct subnet_qstate*)qstate->minfo[id];
315         struct slabhash *subnet_msg_cache = sne->subnet_msg_cache;
316         struct ecs_data *edns = &sq->ecs_client_in;
317         size_t i;
318
319         /* We already calculated hash upon lookup */
320         hashvalue_type h = qstate->minfo[id] ? 
321                 ((struct subnet_qstate*)qstate->minfo[id])->qinfo_hash : 
322                 query_info_hash(&qstate->qinfo, qstate->query_flags);
323         /* Step 1, general qinfo lookup */
324         struct lruhash_entry *lru_entry = slabhash_lookup(subnet_msg_cache, h,
325                 &qstate->qinfo, 1);
326         int acquired_lock = (lru_entry != NULL);
327         if (!lru_entry) {
328                 qinf = qstate->qinfo;
329                 qinf.qname = memdup(qstate->qinfo.qname,
330                         qstate->qinfo.qname_len);
331                 if(!qinf.qname) {
332                         log_err("memdup failed");
333                         return;
334                 }
335                 mrep_entry = query_info_entrysetup(&qinf, NULL, h);
336                 free(qinf.qname); /* if qname 'consumed', it is set to NULL */
337                 if (!mrep_entry) {
338                         log_err("query_info_entrysetup failed");
339                         return;
340                 }
341                 lru_entry = &mrep_entry->entry;
342                 lock_rw_wrlock(&lru_entry->lock);
343                 lru_entry->data = calloc(1,
344                         sizeof(struct subnet_msg_cache_data));
345                 if (!lru_entry->data) {
346                         log_err("malloc failed");
347                         return;
348                 }
349         }
350         /* Step 2, find the correct tree */
351         if (!(tree = get_tree(lru_entry->data, edns, sne, qstate->env->cfg))) {
352                 if (acquired_lock) lock_rw_unlock(&lru_entry->lock);
353                 log_err("Subnet cache insertion failed");
354                 return;
355         }
356         lock_quick_lock(&sne->alloc.lock);
357         rep = reply_info_copy(qstate->return_msg->rep, &sne->alloc, NULL);
358         lock_quick_unlock(&sne->alloc.lock);
359         if (!rep) {
360                 if (acquired_lock) lock_rw_unlock(&lru_entry->lock);
361                 log_err("Subnet cache insertion failed");
362                 return;
363         }
364         
365         /* store RRsets */
366         for(i=0; i<rep->rrset_count; i++) {
367                 rep->ref[i].key = rep->rrsets[i];
368                 rep->ref[i].id = rep->rrsets[i]->id;
369         }
370         reply_info_set_ttls(rep, *qstate->env->now);
371         rep->flags |= (BIT_RA | BIT_QR); /* fix flags to be sensible for */
372         rep->flags &= ~(BIT_AA | BIT_CD);/* a reply based on the cache   */
373         addrtree_insert(tree, (addrkey_t*)edns->subnet_addr, 
374                 edns->subnet_source_mask, 
375                 sq->ecs_server_in.subnet_scope_mask, rep,
376                 rep->ttl, *qstate->env->now);
377         if (acquired_lock) {
378                 lock_rw_unlock(&lru_entry->lock);
379         } else {
380                 lock_rw_unlock(&lru_entry->lock);
381                 slabhash_insert(subnet_msg_cache, h, lru_entry, lru_entry->data,
382                         NULL);
383         }
384 }
385
386 /** Lookup in cache and reply true iff reply is sent. */
387 static int
388 lookup_and_reply(struct module_qstate *qstate, int id, struct subnet_qstate *sq)
389 {
390         struct lruhash_entry *e;
391         struct module_env *env = qstate->env;
392         struct subnet_env *sne = (struct subnet_env*)env->modinfo[id];
393         hashvalue_type h = query_info_hash(&qstate->qinfo, qstate->query_flags);
394         struct subnet_msg_cache_data *data;
395         struct ecs_data *ecs = &sq->ecs_client_in;
396         struct addrtree *tree;
397         struct addrnode *node;
398         uint8_t scope;
399
400         memset(&sq->ecs_client_out, 0, sizeof(sq->ecs_client_out));
401
402         if (sq) sq->qinfo_hash = h; /* Might be useful on cache miss */
403         e = slabhash_lookup(sne->subnet_msg_cache, h, &qstate->qinfo, 1);
404         if (!e) return 0; /* qinfo not in cache */
405         data = e->data;
406         tree = (ecs->subnet_addr_fam == EDNSSUBNET_ADDRFAM_IP4)?
407                 data->tree4 : data->tree6;
408         if (!tree) { /* qinfo in cache but not for this family */
409                 lock_rw_unlock(&e->lock);
410                 return 0;
411         }
412         node = addrtree_find(tree, (addrkey_t*)ecs->subnet_addr, 
413                 ecs->subnet_source_mask, *env->now);
414         if (!node) { /* plain old cache miss */
415                 lock_rw_unlock(&e->lock);
416                 return 0;
417         }
418
419         qstate->return_msg = tomsg(NULL, &qstate->qinfo,
420                 (struct reply_info *)node->elem, qstate->region, *env->now,
421                 env->scratch);
422         scope = (uint8_t)node->scope;
423         lock_rw_unlock(&e->lock);
424         
425         if (!qstate->return_msg) { /* Failed allocation or expired TTL */
426                 return 0;
427         }
428         
429         if (sq->subnet_downstream) { /* relay to interested client */
430                 sq->ecs_client_out.subnet_scope_mask = scope;
431                 sq->ecs_client_out.subnet_addr_fam = ecs->subnet_addr_fam;
432                 sq->ecs_client_out.subnet_source_mask = ecs->subnet_source_mask;
433                 memcpy(&sq->ecs_client_out.subnet_addr, &ecs->subnet_addr,
434                         INET6_SIZE);
435                 sq->ecs_client_out.subnet_validdata = 1;
436         }
437         return 1;
438 }
439
440 /**
441  * Test first bits of addresses for equality. Caller is responsible
442  * for making sure that both a and b are at least net/8 octets long.
443  * @param a: first address.
444  * @param a: seconds address.
445  * @param net: Number of bits to test.
446  * @return: 1 if equal, 0 otherwise.
447  */
448 static int 
449 common_prefix(uint8_t *a, uint8_t *b, uint8_t net)
450 {
451         size_t n = (size_t)net / 8;
452         return !memcmp(a, b, n) && ((net % 8) == 0 || a[n] == b[n]);
453 }
454
455 static enum module_ext_state
456 eval_response(struct module_qstate *qstate, int id, struct subnet_qstate *sq)
457 {
458         struct subnet_env *sne = qstate->env->modinfo[id];
459
460         struct ecs_data *c_in  = &sq->ecs_client_in; /* rcvd from client */
461         struct ecs_data *c_out = &sq->ecs_client_out;/* will send to client */
462         struct ecs_data *s_in  = &sq->ecs_server_in; /* rcvd from auth */
463         struct ecs_data *s_out = &sq->ecs_server_out;/* sent to auth */
464
465         memset(c_out, 0, sizeof(*c_out));
466
467         if (!qstate->return_msg) return module_error;
468         
469         /* We have not asked for subnet data */
470         if (!sq->subnet_sent) {
471                 if (s_in->subnet_validdata)
472                         verbose(VERB_QUERY, "subnet: received spurious data");
473                 if (sq->subnet_downstream) /* Copy back to client */
474                         cp_edns_bad_response(c_out, c_in);
475                 return module_finished;
476         }
477         
478         /* subnet sent but nothing came back */
479         if (!s_in->subnet_validdata) {
480                 /* The authority indicated no support for edns subnet. As a
481                  * consequence the answer ended up in the regular cache. It
482                  * is still usefull to put it in the edns subnet cache for
483                  * when a client explicitly asks for subnet specific answer. */
484                 verbose(VERB_QUERY, "subnet: Authority indicates no support");
485                 lock_rw_wrlock(&sne->biglock);
486                 update_cache(qstate, id);
487                 lock_rw_unlock(&sne->biglock);
488                 if (sq->subnet_downstream)
489                         cp_edns_bad_response(c_out, c_in);
490                 return module_finished;
491         }
492         
493         /* Being here means we have asked for and got a subnet specific 
494          * answer. Also, the answer from the authority is not yet cached 
495          * anywhere. */
496         
497         /* can we accept response? */
498         if(s_out->subnet_addr_fam != s_in->subnet_addr_fam ||
499                 s_out->subnet_source_mask != s_in->subnet_source_mask ||
500                 !common_prefix(s_out->subnet_addr, s_in->subnet_addr, 
501                         s_out->subnet_source_mask))
502         {
503                 /* we can not accept, restart query without option */
504                 verbose(VERB_QUERY, "subnet: forged data");
505                 s_out->subnet_validdata = 0;
506                 (void)edns_opt_list_remove(&qstate->edns_opts_back_out,
507                         qstate->env->cfg->client_subnet_opcode);
508                 sq->subnet_sent = 0;
509                 return module_restart_next;
510         }
511
512         lock_rw_wrlock(&sne->biglock);
513         update_cache(qstate, id);
514         lock_rw_unlock(&sne->biglock);
515         
516         if (sq->subnet_downstream) {
517                 /* Client wants to see the answer, echo option back
518                  * and adjust the scope. */
519                 c_out->subnet_addr_fam = c_in->subnet_addr_fam;
520                 c_out->subnet_source_mask = c_in->subnet_source_mask;
521                 memcpy(&c_out->subnet_addr, &c_in->subnet_addr, INET6_SIZE);
522                 c_out->subnet_scope_mask = s_in->subnet_scope_mask;
523                 c_out->subnet_validdata = 1;
524         }
525         return module_finished;
526 }
527
528 /** Parse EDNS opt data containing ECS */
529 static int
530 parse_subnet_option(struct edns_option* ecs_option, struct ecs_data* ecs)
531 {
532         memset(ecs, 0, sizeof(*ecs));
533         if (ecs_option->opt_len < 4)
534                 return 0;
535
536         ecs->subnet_addr_fam = sldns_read_uint16(ecs_option->opt_data);
537         ecs->subnet_source_mask = ecs_option->opt_data[2];
538         ecs->subnet_scope_mask = ecs_option->opt_data[3];
539         /* remaining bytes indicate address */
540         
541         /* validate input*/
542         /* option length matches calculated length? */
543         if (ecs_option->opt_len != (size_t)((ecs->subnet_source_mask+7)/8 + 4))
544                 return 0;
545         if (ecs_option->opt_len - 4 > INET6_SIZE || ecs_option->opt_len == 0)
546                 return 0;
547         if (ecs->subnet_addr_fam == EDNSSUBNET_ADDRFAM_IP4) {
548                 if (ecs->subnet_source_mask > 32 || ecs->subnet_scope_mask > 32)
549                         return 0;
550         } else if (ecs->subnet_addr_fam == EDNSSUBNET_ADDRFAM_IP6) {
551                 if (ecs->subnet_source_mask > 128 ||
552                         ecs->subnet_scope_mask > 128)
553                         return 0;
554         } else
555                 return 0;
556         
557         /* valid ECS data, write to ecs_data */
558         if (copy_clear(ecs->subnet_addr, INET6_SIZE, ecs_option->opt_data + 4, 
559                 ecs_option->opt_len - 4, ecs->subnet_source_mask))
560                 return 0;
561         ecs->subnet_validdata = 1;
562         return 1;
563 }
564
565 static void
566 subnet_option_from_ss(struct sockaddr_storage *ss, struct ecs_data* ecs,
567         struct config_file* cfg)
568 {
569         void* sinaddr;
570
571         /* Construct subnet option from original query */
572         if(((struct sockaddr_in*)ss)->sin_family == AF_INET) {
573                 ecs->subnet_source_mask = cfg->max_client_subnet_ipv4;
574                 ecs->subnet_addr_fam = EDNSSUBNET_ADDRFAM_IP4;
575                 sinaddr = &((struct sockaddr_in*)ss)->sin_addr;
576                 if (!copy_clear( ecs->subnet_addr, INET6_SIZE,
577                         (uint8_t *)sinaddr, INET_SIZE,
578                         ecs->subnet_source_mask)) {
579                         ecs->subnet_validdata = 1;
580                 }
581         }
582 #ifdef INET6
583         else {
584                 ecs->subnet_source_mask = cfg->max_client_subnet_ipv6;
585                 ecs->subnet_addr_fam = EDNSSUBNET_ADDRFAM_IP6;
586                 sinaddr = &((struct sockaddr_in6*)ss)->sin6_addr;
587                 if (!copy_clear( ecs->subnet_addr, INET6_SIZE,
588                         (uint8_t *)sinaddr, INET6_SIZE,
589                         ecs->subnet_source_mask)) {
590                         ecs->subnet_validdata = 1;
591                 }
592         }
593 #else
594                         /* We don't know how to handle ip6, just pass */
595 #endif /* INET6 */
596 }
597
598 int
599 ecs_query_response(struct module_qstate* qstate, struct dns_msg* response,
600         int id, void* ATTR_UNUSED(cbargs))
601 {
602         struct subnet_qstate *sq;
603         
604         if(!response || !(sq=(struct subnet_qstate*)qstate->minfo[id]))
605                 return 1;
606
607         if(sq->subnet_sent &&
608                 FLAGS_GET_RCODE(response->rep->flags) == LDNS_RCODE_REFUSED) {
609                 /* REFUSED response to ECS query, remove ECS option. */
610                 edns_opt_list_remove(&qstate->edns_opts_back_out,
611                         qstate->env->cfg->client_subnet_opcode);
612                 sq->subnet_sent = 0;
613                 memset(&sq->ecs_server_out, 0, sizeof(sq->ecs_server_out));
614         }
615         return 1;
616 }
617
618 int
619 ecs_edns_back_parsed(struct module_qstate* qstate, int id,
620         void* ATTR_UNUSED(cbargs))
621 {
622         struct subnet_qstate *sq;
623         struct edns_option* ecs_opt;
624         
625         if(!(sq=(struct subnet_qstate*)qstate->minfo[id]))
626                 return 1;
627         if((ecs_opt = edns_opt_list_find(
628                 qstate->edns_opts_back_in,
629                 qstate->env->cfg->client_subnet_opcode))) {
630                 if(parse_subnet_option(ecs_opt, &sq->ecs_server_in) &&
631                         sq->subnet_sent &&
632                         sq->ecs_server_in.subnet_validdata)
633                         /* Only skip global cache store if we sent an ECS option
634                          * and received one back. Answers from non-whitelisted
635                          * servers will end up in global cache. Answers for
636                          * queries with 0 source will not (unless nameserver
637                          * does not support ECS). */
638                         qstate->no_cache_store = 1;
639         }
640
641         return 1;
642 }
643
644 void
645 subnetmod_operate(struct module_qstate *qstate, enum module_ev event, 
646         int id, struct outbound_entry* outbound)
647 {
648         struct subnet_env *sne = qstate->env->modinfo[id];
649         struct subnet_qstate *sq = (struct subnet_qstate*)qstate->minfo[id];
650         
651         verbose(VERB_QUERY, "subnet[module %d] operate: extstate:%s "
652                 "event:%s", id, strextstate(qstate->ext_state[id]), 
653                 strmodulevent(event));
654         log_query_info(VERB_QUERY, "subnet operate: query", &qstate->qinfo);
655
656         if((event == module_event_new || event == module_event_pass) &&
657                 sq == NULL) {
658                 struct edns_option* ecs_opt;
659                 if(!subnet_new_qstate(qstate, id)) {
660                         qstate->return_msg = NULL;
661                         qstate->ext_state[id] = module_finished;
662                         return;
663                 }
664
665                 sq = (struct subnet_qstate*)qstate->minfo[id];
666
667                 if((ecs_opt = edns_opt_list_find(
668                         qstate->edns_opts_front_in,
669                         qstate->env->cfg->client_subnet_opcode))) {
670                         if(!parse_subnet_option(ecs_opt, &sq->ecs_client_in)) {
671                                 /* Wrongly formatted ECS option. RFC mandates to
672                                  * return FORMERROR. */
673                                 qstate->return_rcode = LDNS_RCODE_FORMERR;
674                                 qstate->ext_state[id] = module_finished;
675                                 return;
676                         }
677                         sq->subnet_downstream = 1;
678                 }
679                 else if(qstate->mesh_info->reply_list) {
680                         subnet_option_from_ss(
681                                 &qstate->mesh_info->reply_list->query_reply.addr,
682                                 &sq->ecs_client_in, qstate->env->cfg);
683                 }
684                 
685                 if(sq->ecs_client_in.subnet_validdata == 0) {
686                         /* No clients are interested in result or we could not
687                          * parse it, we don't do client subnet */
688                         sq->ecs_server_out.subnet_validdata = 0;
689                         verbose(VERB_ALGO, "subnet: pass to next module");
690                         qstate->ext_state[id] = module_wait_module;
691                         return;
692                 }
693
694                 lock_rw_wrlock(&sne->biglock);
695                 if (lookup_and_reply(qstate, id, sq)) {
696                         lock_rw_unlock(&sne->biglock);
697                         verbose(VERB_QUERY, "subnet: answered from cache");
698                         qstate->ext_state[id] = module_finished;
699
700                         ecs_opt_list_append(&sq->ecs_client_out,
701                                 &qstate->edns_opts_front_out, qstate);
702                         return;
703                 }
704                 lock_rw_unlock(&sne->biglock);
705                 
706                 sq->ecs_server_out.subnet_addr_fam =
707                         sq->ecs_client_in.subnet_addr_fam;
708                 sq->ecs_server_out.subnet_source_mask =
709                         sq->ecs_client_in.subnet_source_mask;
710                 /* Limit source prefix to configured maximum */
711                 if(sq->ecs_server_out.subnet_addr_fam == EDNSSUBNET_ADDRFAM_IP4 
712                         && sq->ecs_server_out.subnet_source_mask >
713                         qstate->env->cfg->max_client_subnet_ipv4)
714                         sq->ecs_server_out.subnet_source_mask =
715                                 qstate->env->cfg->max_client_subnet_ipv4;
716                 else if(sq->ecs_server_out.subnet_addr_fam == EDNSSUBNET_ADDRFAM_IP6 
717                         && sq->ecs_server_out.subnet_source_mask >
718                         qstate->env->cfg->max_client_subnet_ipv6)
719                         sq->ecs_server_out.subnet_source_mask =
720                                 qstate->env->cfg->max_client_subnet_ipv6;
721                 /* Safe to copy completely, even if the source is limited by the
722                  * configuration. ecs_opt_list_append() will limit the address.
723                  * */
724                 memcpy(&sq->ecs_server_out.subnet_addr,
725                         sq->ecs_client_in.subnet_addr, INET6_SIZE);
726                 sq->ecs_server_out.subnet_scope_mask = 0;
727                 sq->ecs_server_out.subnet_validdata = 1;
728                 if(sq->ecs_server_out.subnet_source_mask != 0 &&
729                         qstate->env->cfg->client_subnet_always_forward &&
730                         sq->subnet_downstream)
731                         /* ECS specific data required, do not look at the global
732                          * cache in other modules. */
733                         qstate->no_cache_lookup = 1;
734                 
735                 /* pass request to next module */
736                 verbose(VERB_ALGO,
737                         "subnet: not found in cache. pass to next module");
738                 qstate->ext_state[id] = module_wait_module;
739                 return;
740         }
741         /* Query handed back by next module, we have a 'final' answer */
742         if(sq && event == module_event_moddone) {
743                 qstate->ext_state[id] = eval_response(qstate, id, sq);
744                 if(qstate->ext_state[id] == module_finished) {
745                         ecs_opt_list_append(&sq->ecs_client_out,
746                                 &qstate->edns_opts_front_out, qstate);
747                 }
748                 return;
749         }
750         if(sq && outbound) {
751                 return;
752         }
753         /* We are being revisited */
754         if(event == module_event_pass || event == module_event_new) {
755                 /* Just pass it on, we already did the work */
756                 verbose(VERB_ALGO, "subnet: pass to next module");
757                 qstate->ext_state[id] = module_wait_module;
758                 return;
759         }
760         if(!sq && (event == module_event_moddone)) {
761                 /* during priming, module done but we never started */
762                 qstate->ext_state[id] = module_finished;
763                 return;
764         }
765         log_err("subnet: bad event %s", strmodulevent(event));
766         qstate->ext_state[id] = module_error;
767         return;
768 }
769
770 void
771 subnetmod_clear(struct module_qstate *ATTR_UNUSED(qstate),
772         int ATTR_UNUSED(id))
773 {
774         /* qstate has no data outside region */
775 }
776
777 void
778 subnetmod_inform_super(struct module_qstate *ATTR_UNUSED(qstate),
779         int ATTR_UNUSED(id), struct module_qstate *ATTR_UNUSED(super))
780 {
781         /* Not used */
782 }
783
784 size_t
785 subnetmod_get_mem(struct module_env *env, int id)
786 {
787         struct subnet_env *sn_env = env->modinfo[id];
788         if (!sn_env) return 0;
789         return sizeof(*sn_env) + 
790                 slabhash_get_mem(sn_env->subnet_msg_cache) +
791                 ecs_whitelist_get_mem(sn_env->whitelist);
792 }
793
794 /**
795  * The module function block 
796  */
797 static struct module_func_block subnetmod_block = {
798         "subnet", &subnetmod_init, &subnetmod_deinit, &subnetmod_operate,
799         &subnetmod_inform_super, &subnetmod_clear, &subnetmod_get_mem
800 };
801
802 struct module_func_block*
803 subnetmod_get_funcblock(void)
804 {
805         return &subnetmod_block;
806 }
807
808 /** Wrappers for static functions to unit test */
809 size_t
810 unittest_wrapper_subnetmod_sizefunc(void *elemptr)
811 {
812         return sizefunc(elemptr);
813 }
814
815 #endif  /* CLIENT_SUBNET */