]> git.ipfire.org Git - thirdparty/freeswitch.git/commitdiff
FS-9775: Search for both v4 and v6 should now be implemented, untested currently
authorShane Bryldt <astaelan@gmail.com>
Sat, 17 Dec 2016 23:10:58 +0000 (23:10 +0000)
committerMike Jerris <mike@jerris.com>
Wed, 25 Jan 2017 20:59:37 +0000 (14:59 -0600)
libs/libks/src/dht/ks_dht-int.h
libs/libks/src/dht/ks_dht.c
libs/libks/src/dht/ks_dht.h
libs/libks/src/dht/ks_dht_datagram.c
libs/libks/src/dht/ks_dht_endpoint.c
libs/libks/src/dht/ks_dht_message.c
libs/libks/src/dht/ks_dht_search.c
libs/libks/src/dht/ks_dht_storageitem.c
libs/libks/src/dht/ks_dht_transaction.c
libs/libks/test/testdht2.c

index 22b80530df9c723a86cb798619c79bf7d688f71d..24bec0c531fc933bcb135aee3528ed340f18decd 100644 (file)
@@ -267,6 +267,7 @@ KS_DECLARE(ks_status_t) ks_dht_search_create(ks_dht_search_t **search, ks_pool_t
 KS_DECLARE(void) ks_dht_search_destroy(ks_dht_search_t **search);
 
 KS_DECLARE(ks_status_t) ks_dht_search_callback_add(ks_dht_search_t *search, ks_dht_search_callback_t callback);
+KS_DECLARE(void) ks_dht_search_expire(ks_dht_search_t *search, ks_hash_t *pending, int32_t *active);
 
 KS_DECLARE(ks_status_t) ks_dht_search_pending_create(ks_dht_search_pending_t **pending, ks_pool_t *pool, const ks_dht_nodeid_t *nodeid);
 KS_DECLARE(void) ks_dht_search_pending_destroy(ks_dht_search_pending_t **pending);
index a1b4c65fc353eea86cf2e3ba7adcfcb2f6553b4b..aaaef29fc5fad306a549a3462c2eeff7611059bd 100644 (file)
@@ -83,13 +83,6 @@ KS_DECLARE(ks_status_t) ks_dht_create(ks_dht_t **dht, ks_pool_t *pool, ks_thread
        ks_assert(d->registry_error);
        // @todo register 301 error for internal get/put CAS hash mismatch retry handler
 
-       /**
-        * Default these to FALSE, binding will set them TRUE when a respective address is bound.
-        * @todo these may not be useful anymore they are from legacy code
-        */
-    d->bind_ipv4 = KS_FALSE;
-       d->bind_ipv6 = KS_FALSE;
-
        /**
         * Initialize the data used to track endpoints to NULL, binding will handle latent allocations.
         * The endpoints and endpoints_poll arrays are maintained in parallel to optimize polling.
@@ -154,13 +147,16 @@ KS_DECLARE(ks_status_t) ks_dht_create(ks_dht_t **dht, ks_pool_t *pool, ks_thread
        /**
         * Create the hash to store searches.
         */
-       ks_hash_create(&d->search_hash, KS_HASH_MODE_ARBITRARY, KS_HASH_FLAG_NOLOCK | KS_HASH_FLAG_DUP_CHECK, d->pool);
-       ks_assert(d->search_hash);
+       ks_hash_create(&d->searches4_hash, KS_HASH_MODE_ARBITRARY, KS_HASH_FLAG_NOLOCK | KS_HASH_FLAG_DUP_CHECK, d->pool);
+       ks_assert(d->searches4_hash);
+       ks_hash_create(&d->searches6_hash, KS_HASH_MODE_ARBITRARY, KS_HASH_FLAG_NOLOCK | KS_HASH_FLAG_DUP_CHECK, d->pool);
+       ks_assert(d->searches6_hash);
 
        /**
-        * The search hash uses arbitrary key size, which requires the key size be provided.
+        * The searches hash uses arbitrary key size, which requires the key size be provided.
         */
-       ks_hash_set_keysize(d->search_hash, KS_DHT_NODEID_SIZE);
+       ks_hash_set_keysize(d->searches4_hash, KS_DHT_NODEID_SIZE);
+       ks_hash_set_keysize(d->searches6_hash, KS_DHT_NODEID_SIZE);
        
        /**
         * The opaque write tokens require some entropy for generating which needs to change periodically but accept tokens using the last two secrets.
@@ -225,14 +221,23 @@ KS_DECLARE(void) ks_dht_destroy(ks_dht_t **dht)
        /**
         * Cleanup the search hash and it's contents if it is allocated.
         */
-       if (d->search_hash) {
-               for (it = ks_hash_first(d->search_hash, KS_UNLOCKED); it; it = ks_hash_next(&it)) {
+       if (d->searches6_hash) {
+               for (it = ks_hash_first(d->searches6_hash, KS_UNLOCKED); it; it = ks_hash_next(&it)) {
                        ks_dht_search_t *val;
 
                        ks_hash_this_val(it, (void **)&val);
                        ks_dht_search_destroy(&val);
                }
-               ks_hash_destroy(&d->search_hash);
+               ks_hash_destroy(&d->searches6_hash);
+       }
+       if (d->searches4_hash) {
+               for (it = ks_hash_first(d->searches4_hash, KS_UNLOCKED); it; it = ks_hash_next(&it)) {
+                       ks_dht_search_t *val;
+
+                       ks_hash_this_val(it, (void **)&val);
+                       ks_dht_search_destroy(&val);
+               }
+               ks_hash_destroy(&d->searches4_hash);
        }
 
        /**
@@ -285,30 +290,18 @@ KS_DECLARE(void) ks_dht_destroy(ks_dht_t **dht)
        /**
         * Cleanup the array of endpoint pointers if it is allocated.
         */
-       if (d->endpoints) {
-               ks_pool_free(d->pool, &d->endpoints);
-               d->endpoints = NULL;
-       }
+       if (d->endpoints) ks_pool_free(d->pool, &d->endpoints);
 
        /**
         * Cleanup the array of endpoint polling data if it is allocated.
         */
-       if (d->endpoints_poll) {
-               ks_pool_free(d->pool, &d->endpoints_poll);
-               d->endpoints_poll = NULL;
-       }
+       if (d->endpoints_poll) ks_pool_free(d->pool, &d->endpoints_poll);
 
        /**
         * Cleanup the endpoints hash if it is allocated.
         */
        if (d->endpoints_hash) ks_hash_destroy(&d->endpoints_hash);
 
-       /**
-        * Probably don't need this
-        */
-       d->bind_ipv4 = KS_FALSE;
-       d->bind_ipv6 = KS_FALSE;
-
        /**
         * Cleanup the type, query, and error registries if they have been allocated.
         */
@@ -474,12 +467,6 @@ KS_DECLARE(ks_status_t) ks_dht_bind(ks_dht_t *dht, const ks_dht_nodeid_t *nodeid
                return KS_STATUS_FAIL;
        }
 
-       /**
-        * Legacy code, this can probably go away
-        */
-       dht->bind_ipv4 |= addr->family == AF_INET;
-       dht->bind_ipv6 |= addr->family == AF_INET6;
-
        /**
         * Attempt to open a UDP datagram socket for the given address family.
         */
@@ -494,8 +481,7 @@ KS_DECLARE(ks_status_t) ks_dht_bind(ks_dht_t *dht, const ks_dht_nodeid_t *nodeid
        /**
         * Attempt to bind the socket to the desired local address.
         */
-       // @todo shouldn't ks_addr_bind take a const addr *?
-       if ((ret = ks_addr_bind(sock, (ks_sockaddr_t *)addr)) != KS_STATUS_SUCCESS) goto done;
+       if ((ret = ks_addr_bind(sock, addr)) != KS_STATUS_SUCCESS) goto done;
 
        /**
         * Allocate the endpoint to track the local socket.
@@ -616,6 +602,54 @@ KS_DECLARE(void) ks_dht_pulse(ks_dht_t *dht, int32_t timeout)
        if (dht->rt_ipv6) ks_dhtrt_process_table(dht->rt_ipv6);
 }
 
+KS_DECLARE(void) ks_dht_pulse_expirations_searches(ks_dht_t *dht, ks_hash_t *searches)
+{
+       ks_hash_iterator_t *it = NULL;
+       ks_time_t now = ks_time_now();
+
+       ks_assert(dht);
+       ks_assert(searches);
+
+       ks_hash_write_lock(searches);
+       for (it = ks_hash_first(searches, KS_UNLOCKED); it; it = ks_hash_next(&it)) {
+               const void *key = NULL;
+               ks_dht_search_t *value = NULL;
+               int32_t active = 0;
+
+               ks_hash_this(it, &key, NULL, (void **)&value);
+
+               ks_mutex_lock(value->mutex);
+               for (ks_hash_iterator_t *i = ks_hash_first(value->pending, KS_UNLOCKED); i; i = ks_hash_next(&i)) {
+                       const void *k = NULL;
+                       ks_dht_search_pending_t *v = NULL;
+
+                       ks_hash_this(i, &k, NULL, (void **)&v);
+
+                       if (v->finished) continue;
+
+                       if (v->expiration <= now) {
+                               char id_buf[KS_DHT_NODEID_SIZE * 2 + 1];
+                               char id2_buf[KS_DHT_NODEID_SIZE * 2 + 1];
+                               ks_log(KS_LOG_DEBUG,
+                                          "Search for %s pending find_node to %s has expired without response\n",
+                                          ks_dht_hexid(&value->target, id_buf),
+                                          ks_dht_hexid(&v->nodeid, id2_buf));
+                               v->finished = KS_TRUE;
+                               continue;
+                       }
+                       active++;
+               }
+               ks_mutex_unlock(value->mutex);
+
+               if (active == 0) {
+                       for (int32_t index = 0; index < value->callbacks_size; ++index) value->callbacks[index](dht, value);
+                       ks_hash_remove(searches, (void *)key);
+                       ks_dht_search_destroy(&value);
+               }
+       }
+       ks_hash_write_unlock(searches);
+}
+
 KS_DECLARE(void) ks_dht_pulse_expirations(ks_dht_t *dht)
 {
        ks_hash_iterator_t *it = NULL;
@@ -645,44 +679,8 @@ KS_DECLARE(void) ks_dht_pulse_expirations(ks_dht_t *dht)
        }
        ks_hash_write_unlock(dht->transactions_hash);
 
-       ks_hash_write_lock(dht->search_hash);
-       for (it = ks_hash_first(dht->search_hash, KS_UNLOCKED); it; it = ks_hash_next(&it)) {
-               const void *search_key = NULL;
-               ks_dht_search_t *search_value = NULL;
-
-               ks_hash_this(it, &search_key, NULL, (void **)&search_value);
-
-               ks_hash_write_lock(search_value->pending);
-               for (ks_hash_iterator_t *i = ks_hash_first(search_value->pending, KS_UNLOCKED); i; i = ks_hash_next(&i)) {
-                       const void *pending_key = NULL;
-                       ks_dht_search_pending_t *pending_value = NULL;
-                       ks_bool_t pending_remove = KS_FALSE;
-
-                       ks_hash_this(i, &pending_key, NULL, (void **)&pending_value);
-
-                       if (pending_value->finished) pending_remove = KS_TRUE;
-                       else if (pending_value->expiration <= now) {
-                               char id_buf[KS_DHT_NODEID_SIZE * 2 + 1];
-                               char id2_buf[KS_DHT_NODEID_SIZE * 2 + 1];
-                               ks_log(KS_LOG_DEBUG,
-                                          "Search for %s pending find_node to %s has expired without response\n",
-                                          ks_dht_hexid(&search_value->target, id_buf),
-                                          ks_dht_hexid(&pending_value->nodeid, id2_buf));
-                               pending_remove = KS_TRUE;
-                       }
-                       if (pending_remove) {
-                               ks_hash_remove(search_value->pending, (void *)pending_key);
-                               ks_dht_search_pending_destroy(&pending_value);
-                       }
-               }
-               ks_hash_write_unlock(search_value->pending);
-               if (ks_hash_count(search_value->pending) == 0) {
-                       for (int32_t index = 0; index < search_value->callbacks_size; ++index) search_value->callbacks[index](dht, search_value);
-                       ks_hash_remove(dht->search_hash, (void *)search_key);
-                       ks_dht_search_destroy(&search_value);
-               }
-       }
-       ks_hash_write_unlock(dht->search_hash);
+       if (dht->rt_ipv4) ks_dht_pulse_expirations_searches(dht, dht->searches4_hash);
+       if (dht->rt_ipv6) ks_dht_pulse_expirations_searches(dht, dht->searches6_hash);
 
        if (dht->token_secret_expiration && dht->token_secret_expiration <= now) {
                dht->token_secret_expiration = ks_time_now() + (KS_DHT_TOKENSECRET_EXPIRATION * 1000);
@@ -771,7 +769,7 @@ KS_DECLARE(ks_status_t) ks_dht_utility_compact_addressinfo(const ks_sockaddr_t *
        memcpy(buffer + (*buffer_length), paddr, sizeof(uint32_t));
        *buffer_length += addr_len;
 
-       memcpy(buffer + (*buffer_length), (const void *)&port, sizeof(uint16_t));
+       memcpy(buffer + (*buffer_length), &port, sizeof(uint16_t));
        *buffer_length += sizeof(uint16_t);
 
        return KS_STATUS_SUCCESS;
@@ -801,8 +799,7 @@ KS_DECLARE(ks_status_t) ks_dht_utility_expand_addressinfo(const uint8_t *buffer,
        port = *((uint16_t *)(buffer + *buffer_length));
        *buffer_length += sizeof(uint16_t);
 
-       // @todo ks_addr_set_raw second parameter should be const?
-       return ks_addr_set_raw(address, (void *)paddr, port, address->family);
+       return ks_addr_set_raw(address, paddr, port, address->family);
 }
 
 KS_DECLARE(ks_status_t) ks_dht_utility_compact_nodeinfo(const ks_dht_nodeid_t *nodeid,
@@ -824,7 +821,7 @@ KS_DECLARE(ks_status_t) ks_dht_utility_compact_nodeinfo(const ks_dht_nodeid_t *n
                return KS_STATUS_NO_MEM;
        }
 
-       memcpy(buffer + (*buffer_length), (void *)nodeid, KS_DHT_NODEID_SIZE);
+       memcpy(buffer + (*buffer_length), nodeid->id, KS_DHT_NODEID_SIZE);
        *buffer_length += KS_DHT_NODEID_SIZE;
 
        return ks_dht_utility_compact_addressinfo(address, buffer, buffer_length, buffer_size);
@@ -953,7 +950,7 @@ KS_DECLARE(ks_bool_t) ks_dht_token_verify(ks_dht_t *dht, ks_sockaddr_t *raddr, k
 KS_DECLARE(ks_status_t) ks_dht_send(ks_dht_t *dht, ks_dht_message_t *message)
 {
        // @todo calculate max IPV6 payload size?
-       char buf[1000];
+       char buf[1001];
        ks_size_t buf_len;
 
        ks_assert(dht);
@@ -963,8 +960,11 @@ KS_DECLARE(ks_status_t) ks_dht_send(ks_dht_t *dht, ks_dht_message_t *message)
 
        // @todo blacklist check
 
-       // @todo use different encode function to check if all data was encoded, do not send large incomplete messages
        buf_len = ben_encode2(buf, sizeof(buf), message->data);
+       if (buf_len >= sizeof(buf)) {
+               ks_log(KS_LOG_DEBUG, "Dropping message that is too large\n");
+               return KS_STATUS_FAIL;
+       }
 
        ks_log(KS_LOG_DEBUG, "Sending message to %s %d\n", message->raddr.host, message->raddr.port);
        ks_log(KS_LOG_DEBUG, "%s\n", ben_print(message->data));
@@ -1192,13 +1192,15 @@ KS_DECLARE(ks_status_t) ks_dht_process_response(ks_dht_t *dht, ks_dht_message_t
 
 
 KS_DECLARE(ks_status_t) ks_dht_search(ks_dht_t *dht,
-                                                                         int family,
+                                                                         int32_t family,
                                                                          ks_dht_nodeid_t *target,
                                                                          ks_dht_search_callback_t callback,
                                                                          ks_dht_search_t **search)
 {
+       ks_bool_t locked_searches = KS_FALSE;
        ks_bool_t locked_search = KS_FALSE;
-       ks_bool_t locked_pending = KS_FALSE;
+       ks_hash_t *searches = NULL;
+       ks_dhtrt_routetable_t *rt = NULL;
        ks_dht_search_t *s = NULL;
        ks_bool_t inserted = KS_FALSE;
        ks_bool_t allocated = KS_FALSE;
@@ -1211,12 +1213,27 @@ KS_DECLARE(ks_status_t) ks_dht_search(ks_dht_t *dht,
 
        if (search) *search = NULL;
 
-       // @todo start write lock on search_hash and hold until after inserting
+       if (family == AF_INET) {
+               if (!dht->rt_ipv4) {
+                       ret = KS_STATUS_FAIL;
+                       goto done;
+               }
+               searches = dht->searches4_hash;
+               rt = dht->rt_ipv4;
+       } else {
+               if (!dht->rt_ipv6) {
+                       ret = KS_STATUS_FAIL;
+                       goto done;
+               }
+               searches = dht->searches6_hash;
+               rt = dht->rt_ipv6;
+       }
+
        // check hash for target to see if search already exists
-       ks_hash_write_lock(dht->search_hash);
-       locked_search = KS_TRUE;
+       ks_hash_write_lock(searches);
+       locked_searches = KS_TRUE;
 
-       s = ks_hash_search(dht->search_hash, target->id, KS_UNLOCKED);
+       s = ks_hash_search(searches, target->id, KS_UNLOCKED);
 
        // if search does not exist, create new search and store in hash by target
        if (!s) {
@@ -1230,49 +1247,56 @@ KS_DECLARE(ks_status_t) ks_dht_search(ks_dht_t *dht,
        // if the search is old then bail out and return successfully
        if (!allocated) goto done;
 
-       if ((ret = ks_hash_insert(dht->search_hash, s->target.id, s)) == KS_STATUS_SUCCESS) goto done;
+       // everything past this point until final cleanup is only for when a search of the target does not already exist
+
+       if ((ret = ks_hash_insert(searches, s->target.id, s)) == KS_STATUS_SUCCESS) goto done;
        inserted = KS_TRUE;
 
-       // lock pending before unlocking the search hash to prevent this search from being used before we finish setting it up
-       ks_hash_write_lock(s->pending);
-       locked_pending = KS_TRUE;
+       // lock search before unlocking the searches_hash to prevent this search from being used before we finish setting it up
+       ks_mutex_lock(s->mutex);
+       locked_search = KS_TRUE;
 
-       // release search hash lock now, but pending is still locked
-       ks_hash_write_unlock(dht->search_hash);
-       locked_search = KS_FALSE;
+       // release searches_hash lock now, but search is still locked
+       ks_hash_write_unlock(searches);
+       locked_searches = KS_FALSE;
 
        // find closest good nodes to target locally and store as the closest results
-    query.nodeid = *target;
+       query.nodeid = *target;
        query.type = KS_DHT_REMOTE;
        query.max = KS_DHT_SEARCH_RESULTS_MAX_SIZE;
        query.family = family;
        query.count = 0;
-       ks_dhtrt_findclosest_nodes(family == AF_INET ? dht->rt_ipv4 : dht->rt_ipv6, &query);
+       ks_dhtrt_findclosest_nodes(rt, &query);
        for (int32_t i = 0; i < query.count; ++i) {
                ks_dht_node_t *n = query.nodes[i];
                ks_dht_search_pending_t *pending = NULL;
-               
-               s->results[i] = n->nodeid;
-               ks_dht_utility_nodeid_xor(&s->distances[i], &n->nodeid, &s->target);
-               // add to pending with expiration
+
+               // always take the initial local closest good nodes as results, they are already good nodes that are closest with no results yet
+               s->results[s->results_length] = n->nodeid;
+               ks_dht_utility_nodeid_xor(&s->distances[s->results_length], &n->nodeid, &s->target);
+               s->results_length++;
+
+               pending = ks_hash_search(s->pending, n->nodeid.id, KS_UNLOCKED);
+               if (pending) continue; // skip duplicates, this really shouldn't happen on a new search but we sanity check
+
+               // add to pending with expiration, if any of this fails it's almost catastrophic so just bail out and fail the entire search attempt
+               // there are no probable causes for a failure but check them anyway
                if ((ret = ks_dht_search_pending_create(&pending, s->pool, &n->nodeid)) != KS_STATUS_SUCCESS) goto done;
                if ((ret = ks_hash_insert(s->pending, n->nodeid.id, pending)) != KS_STATUS_SUCCESS) {
                        ks_dht_search_pending_destroy(&pending);
                        goto done;
                }
                if ((ret = ks_dht_send_findnode(dht, NULL, &n->addr, target)) != KS_STATUS_SUCCESS) goto done;
-               // increment here in case we end up bailing out; execute with what it has or destroy the search?
-               s->results_length++;
        }
-       // @todo release query nodes
-       ks_hash_write_unlock(s->pending);
-       locked_pending = KS_FALSE;
+       // @todo release closest local query node locks
+       ks_mutex_unlock(s->mutex);
+       locked_search = KS_FALSE;
 
        if (search) *search = s;
 
  done:
-       if (locked_search) ks_hash_write_unlock(dht->search_hash);
-       if (locked_pending) ks_hash_write_unlock(s->pending);
+       if (locked_searches) ks_hash_write_unlock(searches);
+       if (locked_search) ks_mutex_unlock(s->mutex);
        if (ret != KS_STATUS_SUCCESS) {
                if (!inserted && s) ks_dht_search_destroy(&s);
                *search = NULL;
@@ -1513,7 +1537,13 @@ KS_DECLARE(ks_status_t) ks_dht_send_findnode(ks_dht_t *dht, ks_dht_endpoint_t *e
 
        ben_dict_set(a, ben_blob("id", 2), ben_blob(message->endpoint->nodeid.id, KS_DHT_NODEID_SIZE));
        ben_dict_set(a, ben_blob("target", 6), ben_blob(targetid->id, KS_DHT_NODEID_SIZE));
-       // @todo produce "want" value if both families are bound
+       // Only request both v4 and v6 if we have both interfaces bound and are looking for our own node id, aka bootstrapping
+       if (dht->rt_ipv4 && dht->rt_ipv6 && !memcmp(message->endpoint->nodeid.id, targetid->id, KS_DHT_NODEID_SIZE)) {
+               struct bencode *want = ben_list();
+               ben_list_append_str(want, "n4");
+               ben_list_append_str(want, "n6");
+               ben_dict_set(a, ben_blob("want", 4), want);
+       }
 
        ks_log(KS_LOG_DEBUG, "Sending message query find_node\n");
        ks_q_push(dht->send_q, (void *)message);
@@ -1579,7 +1609,7 @@ KS_DECLARE(ks_status_t) ks_dht_process_query_findnode(ks_dht_t *dht, ks_dht_mess
        query.max = 8; // should be like KS_DHTRT_BUCKET_SIZE
        if (want4) {
                query.family = AF_INET;
-               ks_dhtrt_findclosest_nodes(routetable, &query);
+               ks_dhtrt_findclosest_nodes(dht->rt_ipv4, &query);
 
                for (int32_t i = 0; i < query.count; ++i) {
                        ks_dht_node_t *qn = query.nodes[i];
@@ -1592,10 +1622,11 @@ KS_DECLARE(ks_status_t) ks_dht_process_query_findnode(ks_dht_t *dht, ks_dht_mess
 
                        ks_log(KS_LOG_DEBUG, "Compacted ipv4 nodeinfo for %s (%s %d)\n", ks_dht_hexid(&qn->nodeid, id_buf), qn->addr.host, qn->addr.port);
                }
+               // @todo release query nodes
        }
        if (want6) {
                query.family = AF_INET6;
-               ks_dhtrt_findclosest_nodes(routetable, &query);
+               ks_dhtrt_findclosest_nodes(dht->rt_ipv6, &query);
 
                for (int32_t i = 0; i < query.count; ++i) {
                        ks_dht_node_t *qn = query.nodes[i];
@@ -1633,6 +1664,8 @@ KS_DECLARE(ks_status_t) ks_dht_process_response_findnode(ks_dht_t *dht, ks_dht_m
 {
        ks_dht_nodeid_t *id;
        struct bencode *n;
+       //ks_bool_t n4 = KS_FALSE;
+       //ks_bool_t n6 = KS_FALSE;
        const uint8_t *nodes = NULL;
        const uint8_t *nodes6 = NULL;
        size_t nodes_size = 0;
@@ -1642,6 +1675,7 @@ KS_DECLARE(ks_status_t) ks_dht_process_response_findnode(ks_dht_t *dht, ks_dht_m
        ks_dhtrt_routetable_t *routetable = NULL;
        ks_dht_node_t *node = NULL;
        char id_buf[KS_DHT_NODEID_SIZE * 2 + 1];
+       ks_hash_t *searches = NULL;
        ks_dht_search_t *search = NULL;
        ks_status_t ret = KS_STATUS_SUCCESS;
 
@@ -1653,11 +1687,13 @@ KS_DECLARE(ks_status_t) ks_dht_process_response_findnode(ks_dht_t *dht, ks_dht_m
 
        n = ben_dict_get_by_str(message->args, "nodes");
        if (n) {
+               //n4 = KS_TRUE;
                nodes = (const uint8_t *)ben_str_val(n);
                nodes_size = ben_str_len(n);
        }
        n = ben_dict_get_by_str(message->args, "nodes6");
        if (n) {
+               //n6 = KS_TRUE;
                nodes6 = (const uint8_t *)ben_str_val(n);
                nodes6_size = ben_str_len(n);
        }
@@ -1671,14 +1707,18 @@ KS_DECLARE(ks_status_t) ks_dht_process_response_findnode(ks_dht_t *dht, ks_dht_m
        ks_log(KS_LOG_DEBUG, "Touching node %s\n", ks_dht_hexid(id, id_buf));
        if ((ret = ks_dhtrt_touch_node(routetable, *id)) != KS_STATUS_SUCCESS) goto done;
 
-       ks_hash_read_lock(dht->search_hash);
-       search = ks_hash_search(dht->search_hash, message->transaction->target.id, KS_UNLOCKED);
-       ks_hash_read_unlock(dht->search_hash);
+       searches = message->raddr.family == AF_INET ? dht->searches4_hash : dht->searches6_hash;
+
+       ks_hash_read_lock(searches);
+       search = ks_hash_search(searches, message->transaction->target.id, KS_UNLOCKED);
        if (search) {
-               ks_dht_search_pending_t *pending = ks_hash_search(search->pending, id->id, KS_READLOCKED);
-               ks_hash_read_unlock(search->pending);
+               ks_dht_search_pending_t *pending = NULL;
+
+               ks_mutex_lock(search->mutex);
+               pending = ks_hash_search(search->pending, id->id, KS_UNLOCKED);
                if (pending) pending->finished = KS_TRUE;
        }
+       ks_hash_read_unlock(searches);
 
        while (nodes_len < nodes_size) {
                ks_dht_nodeid_t nid;
@@ -1697,7 +1737,7 @@ KS_DECLARE(ks_status_t) ks_dht_process_response_findnode(ks_dht_t *dht, ks_dht_m
                ks_dhtrt_create_node(dht->rt_ipv4, nid, KS_DHT_REMOTE, addr.host, addr.port, &node);
                ks_dhtrt_release_node(node);
 
-               if (search) {
+               if (search && message->raddr.family == AF_INET && !ks_hash_search(search->pending, nid.id, KS_UNLOCKED)) {
                        ks_dht_nodeid_t distance;
                        int32_t results_index = -1;
                        
@@ -1707,10 +1747,10 @@ KS_DECLARE(ks_status_t) ks_dht_process_response_findnode(ks_dht_t *dht, ks_dht_m
                                search->results_length++;
                        } else {
                                for (int32_t index = 0; index < search->results_length; ++index) {
-                                       // Check if new node is closer than this existing result
+                                       // Check if new node is closer than this previous result
                                        if (memcmp(distance.id, search->distances[index].id, KS_DHT_NODEID_SIZE) < 0) {
                                                // If this is the first node that is further then keep it
-                                               // Else if two or more nodes are further, and this existing result is further than the previous one then keep it
+                                               // Else if two or more nodes are further, and this previous result is further than the current one then keep the current result
                                                if (results_index < 0) results_index = index;
                                                else if (memcmp(search->distances[index].id, search->distances[results_index].id, KS_DHT_NODEID_SIZE) > 0) results_index = index;
                                        }
@@ -1757,12 +1797,55 @@ KS_DECLARE(ks_status_t) ks_dht_process_response_findnode(ks_dht_t *dht, ks_dht_m
                ks_log(KS_LOG_DEBUG, "Creating node %s\n", ks_dht_hexid(&nid, id_buf));
                ks_dhtrt_create_node(dht->rt_ipv6, nid, KS_DHT_REMOTE, addr.host, addr.port, &node);
                ks_dhtrt_release_node(node);
+
+               if (search && message->raddr.family == AF_INET6 && !ks_hash_search(search->pending, nid.id, KS_UNLOCKED)) {
+                       ks_dht_nodeid_t distance;
+                       int32_t results_index = -1;
+                       
+                       ks_dht_utility_nodeid_xor(&distance, &nid, &search->target);
+                       if (search->results_length < KS_DHT_SEARCH_RESULTS_MAX_SIZE) {
+                               results_index = search->results_length;
+                               search->results_length++;
+                       } else {
+                               for (int32_t index = 0; index < search->results_length; ++index) {
+                                       // Check if new node is closer than this previous result
+                                       if (memcmp(distance.id, search->distances[index].id, KS_DHT_NODEID_SIZE) < 0) {
+                                               // If this is the first node that is further then keep it
+                                               // Else if two or more nodes are further, and this previous result is further than the current one then keep the current result
+                                               if (results_index < 0) results_index = index;
+                                               else if (memcmp(search->distances[index].id, search->distances[results_index].id, KS_DHT_NODEID_SIZE) > 0) results_index = index;
+                                       }
+                               }
+                       }
+
+                       if (results_index >= 0) {
+                               char id2_buf[KS_DHT_NODEID_SIZE * 2 + 1];
+                               char id3_buf[KS_DHT_NODEID_SIZE * 2 + 1];
+                               ks_dht_search_pending_t *pending = NULL;
+
+                               ks_log(KS_LOG_DEBUG,
+                                          "Set closer node id %s (%s) in search of target id %s at results index %d\n",
+                                          ks_dht_hexid(&nid, id_buf),
+                                          ks_dht_hexid(&distance, id2_buf),
+                                          ks_dht_hexid(&search->target, id3_buf),
+                                          results_index);
+                               search->results[results_index] = nid;
+                               search->distances[results_index] = distance;
+
+                               if ((ret = ks_dht_search_pending_create(&pending, search->pool, &nid)) != KS_STATUS_SUCCESS) goto done;
+                               if ((ret = ks_hash_insert(search->pending, nid.id, pending)) != KS_STATUS_SUCCESS) {
+                                       ks_dht_search_pending_destroy(&pending);
+                                       goto done;
+                               }
+                               if ((ret = ks_dht_send_findnode(dht, NULL, &addr, &search->target)) != KS_STATUS_SUCCESS) goto done;
+                       }
+               }
        }
-       // @todo repeat above for ipv6 table
 
        ks_log(KS_LOG_DEBUG, "Message response find_node is reached\n");
 
  done:
+       if(search) ks_mutex_unlock(search->mutex);
        return ret;
 }
 
index c9f60ef1819b2340c93248a41a3fa3658ee8a8bd..51b4f84d9f55f3b89bcda3a3935b0659fbc7ef3f 100644 (file)
@@ -149,17 +149,6 @@ struct ks_dht_transaction_s {
        ks_bool_t finished;
 };
 
-// Check if search already exists for the target id, if so add another callback, must be a popular target id
-// Otherwise create new search, set target id, add callback, and insert the search into the dht search_hash with target id key
-// Get closest local nodes to target id, check against results, send_findnode for closer nodes and add to pending hash with queried node id
-// Upon receiving find_node response, check target id against dht search_hash, check responding node id against pending hash, set finished for purging
-// Update results if responding node id is closer than any current result, or the results are not full
-// Check response nodes against results, send_findnode for closer nodes and add to pending hash with an expiration
-// Pulse expirations purges expired and finished from pending hash, once hash is empty callbacks are called providing results array
-// Note:
-// During the lifetime of a search, the ks_dht_node_t's must be kept alive
-// Do a query touch on nodes prior to being added to pending, this should reset timeout and keep the nodes alive long enough even if they are dubious
-// Nodes which land in results are known good with recent response to find_nodes and should be around for a while before route table worries about cleanup
 struct ks_dht_search_s {
        ks_pool_t *pool;
        ks_mutex_t *mutex;
@@ -208,9 +197,6 @@ struct ks_dht_s {
        ks_hash_t *registry_query;
        ks_hash_t *registry_error;
 
-       ks_bool_t bind_ipv4;
-       ks_bool_t bind_ipv6;
-
        ks_dht_endpoint_t **endpoints;
        int32_t endpoints_size;
        ks_hash_t *endpoints_hash;
@@ -230,7 +216,8 @@ struct ks_dht_s {
        ks_dhtrt_routetable_t *rt_ipv4;
        ks_dhtrt_routetable_t *rt_ipv6;
 
-       ks_hash_t *search_hash;
+       ks_hash_t *searches4_hash;
+       ks_hash_t *searches6_hash;
 
        volatile uint32_t token_secret_current;
        volatile uint32_t token_secret_previous;
@@ -300,7 +287,7 @@ KS_DECLARE(ks_status_t) ks_dht_register_error(ks_dht_t *dht, const char *value,
  * @param dht pointer to the dht instance
  * @param nodeid pointer to a nodeid for this endpoint, may be NULL to generate one randomly
  * @param addr pointer to the local address information
- * @param dereferenced out pointer to the allocated endpoint, may be NULL to ignore endpoint output
+ * @param endpoint dereferenced out pointer to the allocated endpoint, may be NULL to ignore endpoint output
  * @return The ks_status_t result: KS_STATUS_SUCCESS, KS_STATUS_FAIL, ...
  * @see ks_socket_option
  * @see ks_addr_bind
@@ -320,6 +307,25 @@ KS_DECLARE(ks_status_t) ks_dht_bind(ks_dht_t *dht, const ks_dht_nodeid_t *nodeid
  */
 KS_DECLARE(void) ks_dht_pulse(ks_dht_t *dht, int32_t timeout);
 
+/**
+ * Create a network search of the closest nodes to a target.
+ * @param dht pointer to the dht instance
+ * @param family either AF_INET or AF_INET6 for the appropriate network to search
+ * @param target pointer to the nodeid for the target to be searched
+ * @param callback an optional callback to add to the search when it is finished
+ * @param search dereferenced out pointer to the allocated search, may be NULL to ignore search output
+ * @return The ks_status_t result: KS_STATUS_SUCCESS, KS_STATUS_FAIL
+ * @see ks_dht_search_create
+ * @see ks_dht_search_callback_add
+ * @see ks_hash_insert
+ * @see ks_dht_search_pending_create
+ * @see ks_dht_send_findnode
+ */
+KS_DECLARE(ks_status_t) ks_dht_search(ks_dht_t *dht,
+                                                                         int32_t family,
+                                                                         ks_dht_nodeid_t *target,
+                                                                         ks_dht_search_callback_t callback,
+                                                                         ks_dht_search_t **search);
 
 /**
  *
index a13c090d184720a1ae05afbd6ea5b7806ef6febe..fe98cffc3c5c8575bfba7083f8137db903ac424e 100644 (file)
@@ -31,8 +31,7 @@ KS_DECLARE(ks_status_t) ks_dht_datagram_create(ks_dht_datagram_t **datagram,
 
        // done:
        if (ret != KS_STATUS_SUCCESS) {
-               if (dg) ks_dht_datagram_destroy(&dg);
-               *datagram = NULL;
+               ks_dht_datagram_destroy(datagram);
        }
        return ret;
 }
@@ -46,9 +45,7 @@ KS_DECLARE(void) ks_dht_datagram_destroy(ks_dht_datagram_t **datagram)
 
        dg = *datagram;
 
-       ks_pool_free(dg->pool, &dg);
-
-       *datagram = NULL;
+       ks_pool_free(dg->pool, datagram);
 }
 
 /* For Emacs:
index 4907d15f342c37bba66a76cf6c04a8513003befc..d8c2965fc420e226b70a6416ed89a4eaaaafdefa 100644 (file)
@@ -30,8 +30,7 @@ KS_DECLARE(ks_status_t) ks_dht_endpoint_create(ks_dht_endpoint_t **endpoint,
 
        // done:
        if (ret != KS_STATUS_SUCCESS) {
-               if (ep) ks_dht_endpoint_destroy(&ep);
-               *endpoint = NULL;
+               if (ep) ks_dht_endpoint_destroy(endpoint);
        }
        return ret;
 }
@@ -49,9 +48,8 @@ KS_DECLARE(void) ks_dht_endpoint_destroy(ks_dht_endpoint_t **endpoint)
        ep = *endpoint;
 
        if (ep->sock != KS_SOCK_INVALID) ks_socket_close(&ep->sock);
-       ks_pool_free(ep->pool, &ep);
 
-       *endpoint = NULL;
+       ks_pool_free(ep->pool, endpoint);
 }
 
 /* For Emacs:
index 3430c2853d43bfcece0c9e25d95a6658ca5aae5b..75780f46c53bb42e90533c4c9026e7413aecdf12 100644 (file)
@@ -26,8 +26,7 @@ KS_DECLARE(ks_status_t) ks_dht_message_create(ks_dht_message_t **message,
 
        // done:
        if (ret != KS_STATUS_SUCCESS) {
-               if (m) ks_dht_message_destroy(&m);
-               *message = NULL;
+               ks_dht_message_destroy(message);
        }
        return ret;
 }
@@ -45,9 +44,8 @@ KS_DECLARE(void) ks_dht_message_destroy(ks_dht_message_t **message)
                ben_free(m->data);
                m->data = NULL;
        }
-       ks_pool_free(m->pool, &(*message));
 
-       *message = NULL;
+       ks_pool_free(m->pool, message);
 }
 
 
index 1945fcd2949af8fe1d8888a1dd841e591b083a7d..e15d726f3ba296769c8fd66cbca1f76e49333f74 100644 (file)
@@ -27,8 +27,7 @@ KS_DECLARE(ks_status_t) ks_dht_search_create(ks_dht_search_t **search, ks_pool_t
 
        // done:
        if (ret != KS_STATUS_SUCCESS) {
-               if (s) ks_dht_search_destroy(&s);
-               *search = NULL;
+               if (s) ks_dht_search_destroy(search);
        }
        return ret;
 }
@@ -58,9 +57,7 @@ KS_DECLARE(void) ks_dht_search_destroy(ks_dht_search_t **search)
        }
        if (s->mutex) ks_mutex_destroy(&s->mutex);
 
-       ks_pool_free(s->pool, &s);
-
-       *search = NULL;
+       ks_pool_free(s->pool, search);
 }
 
 KS_DECLARE(ks_status_t) ks_dht_search_callback_add(ks_dht_search_t *search, ks_dht_search_callback_t callback)
@@ -100,10 +97,9 @@ KS_DECLARE(ks_status_t) ks_dht_search_pending_create(ks_dht_search_pending_t **p
 
        // done:
        if (ret != KS_STATUS_SUCCESS) {
-               if (p) ks_dht_search_pending_destroy(&p);
-               *pending = NULL;
+               if (p) ks_dht_search_pending_destroy(pending);
        }
-       return KS_STATUS_SUCCESS;
+       return ret;
 }
 
 KS_DECLARE(void) ks_dht_search_pending_destroy(ks_dht_search_pending_t **pending)
@@ -115,9 +111,7 @@ KS_DECLARE(void) ks_dht_search_pending_destroy(ks_dht_search_pending_t **pending
 
        p = *pending;
 
-       ks_pool_free(p->pool, &p);
-
-       *pending = NULL;
+       ks_pool_free(p->pool, pending);
 }
 
 /* For Emacs:
index c20cdd1e9c79c35e7fa29d1f7d51968ae077f804..ce34bde9f9220587f86b503151dbf72da5f7dd6e 100644 (file)
@@ -32,8 +32,7 @@ KS_DECLARE(ks_status_t) ks_dht_storageitem_create_immutable(ks_dht_storageitem_t
 
        // done:
        if (ret != KS_STATUS_SUCCESS) {
-               if (si) ks_dht_storageitem_destroy(&si);
-               *item = NULL;
+               if (si) ks_dht_storageitem_destroy(item);
        }
        return ret;
 }
@@ -83,8 +82,7 @@ KS_DECLARE(ks_status_t) ks_dht_storageitem_create_mutable(ks_dht_storageitem_t *
 
        // done:
        if (ret != KS_STATUS_SUCCESS) {
-               if (si) ks_dht_storageitem_destroy(&si);
-               *item = NULL;
+               if (si) ks_dht_storageitem_destroy(item);
        }
        return ret;
 }
@@ -105,9 +103,8 @@ KS_DECLARE(void) ks_dht_storageitem_destroy(ks_dht_storageitem_t **item)
                ben_free(si->v);
                si->v = NULL;
        }
-       ks_pool_free(si->pool, &si);
 
-       *item = NULL;
+       ks_pool_free(si->pool, item);
 }
 
 /* For Emacs:
index 91db11447d8f5a267803b017cc8f4b79f9923dc1..6d6b7baf3585661982bdbdd29494a0bf6ee73739 100644 (file)
@@ -25,8 +25,7 @@ KS_DECLARE(ks_status_t) ks_dht_transaction_create(ks_dht_transaction_t **transac
 
        // done:
        if (ret != KS_STATUS_SUCCESS) {
-               if (t) ks_dht_transaction_destroy(&t);
-               *transaction = NULL;
+               if (t) ks_dht_transaction_destroy(transaction);
        }
        return ret;
 }
@@ -40,9 +39,7 @@ KS_DECLARE(void) ks_dht_transaction_destroy(ks_dht_transaction_t **transaction)
 
        t = *transaction;
 
-       ks_pool_free(t->pool, &t);
-
-       *transaction = NULL;
+       ks_pool_free(t->pool, transaction);
 }
 
 /* For Emacs:
index 76ce6abe227dc3d1fadea83b4b9a2660c2a30883..36dc09e43106341947c20cae6e0131a89f1a4b76 100644 (file)
@@ -149,9 +149,9 @@ int main() {
 
   diag("Pulsing for route table pings\n"); // Wait for route table pinging to catch up
   for (int i = 0; i < 10; ++i) {
-         diag("DHT 1\n");
+         //diag("DHT 1\n");
          ks_dht_pulse(dht1, 100);
-         diag("DHT 2\n");
+         //diag("DHT 2\n");
          ks_dht_pulse(dht2, 100);
   }
   ok(ks_dhtrt_find_node(dht1->rt_ipv4, ep2->nodeid) != NULL); // The node should be good by now, and thus be returned as good
@@ -174,9 +174,9 @@ int main() {
   
   diag("Pulsing for route table pings\n"); // Wait for route table pinging to catch up
   for (int i = 0; i < 10; ++i) {
-         diag("DHT 1\n");
+         //diag("DHT 1\n");
          ks_dht_pulse(dht1, 100);
-         diag("DHT 2\n");
+         //diag("DHT 2\n");
          ks_dht_pulse(dht2, 100);
   }
   ok(ks_dhtrt_find_node(dht3->rt_ipv4, ep2->nodeid) != NULL); // The node should be good by now, and thus be returned as good