diff options
Diffstat (limited to '')
-rw-r--r-- | rtrlib/pfx/trie/trie-pfx.c | 645 |
1 files changed, 645 insertions, 0 deletions
diff --git a/rtrlib/pfx/trie/trie-pfx.c b/rtrlib/pfx/trie/trie-pfx.c new file mode 100644 index 0000000..ffb2416 --- /dev/null +++ b/rtrlib/pfx/trie/trie-pfx.c @@ -0,0 +1,645 @@ +/* + * This file is part of RTRlib. + * + * This file is subject to the terms and conditions of the MIT license. + * See the file LICENSE in the top level directory for more details. + * + * Website: http://rtrlib.realmv6.org/ + */ + +#include "trie-pfx.h" + +#include "rtrlib/lib/alloc_utils_private.h" +#include "rtrlib/lib/ip_private.h" +#include "rtrlib/pfx/pfx_private.h" +#include "rtrlib/pfx/trie/trie_private.h" +#include "rtrlib/rtrlib_export_private.h" + +#include <assert.h> +#include <pthread.h> +#include <stdbool.h> +#include <stdlib.h> + +struct data_elem { + uint32_t asn; + uint8_t max_len; + const struct rtr_socket *socket; +}; + +struct node_data { + unsigned int len; + struct data_elem *ary; +}; + +struct copy_cb_args { + struct pfx_table *pfx_table; + const struct rtr_socket *socket; + bool error; +}; + +struct notify_diff_cb_args { + struct pfx_table *old_table; + struct pfx_table *new_table; + const struct rtr_socket *socket; + pfx_update_fp pfx_update_fp; + bool added; +}; + +static struct trie_node *pfx_table_get_root(const struct pfx_table *pfx_table, const enum lrtr_ip_version ver); +static int pfx_table_del_elem(struct node_data *data, const unsigned int index); +static int pfx_table_create_node(struct trie_node **node, const struct pfx_record *record); +static int pfx_table_append_elem(struct node_data *data, const struct pfx_record *record); +static struct data_elem *pfx_table_find_elem(const struct node_data *data, const struct pfx_record *record, + unsigned int *index); +static bool pfx_table_elem_matches(struct node_data *data, const uint32_t asn, const uint8_t prefix_len); +static void pfx_table_notify_clients(struct pfx_table *pfx_table, const struct pfx_record *record, const bool added); +static int pfx_table_remove_id(struct pfx_table *pfx_table, struct trie_node **root, struct trie_node *node, + const struct rtr_socket *socket, const unsigned int level); +static int pfx_table_node2pfx_record(struct trie_node *node, struct pfx_record records[], const unsigned int ary_len); +static void pfx_table_free_reason(struct pfx_record **reason, unsigned int *reason_len); + +void pfx_table_notify_clients(struct pfx_table *pfx_table, const struct pfx_record *record, const bool added) +{ + if (pfx_table->update_fp) + pfx_table->update_fp(pfx_table, *record, added); +} + +RTRLIB_EXPORT void pfx_table_init(struct pfx_table *pfx_table, pfx_update_fp update_fp) +{ + pfx_table->ipv4 = NULL; + pfx_table->ipv6 = NULL; + pfx_table->update_fp = update_fp; + pthread_rwlock_init(&(pfx_table->lock), NULL); +} + +void pfx_table_free_without_notify(struct pfx_table *pfx_table) +{ + pfx_table->update_fp = NULL; + pfx_table_free(pfx_table); +} + +RTRLIB_EXPORT void pfx_table_free(struct pfx_table *pfx_table) +{ + for (int i = 0; i < 2; i++) { + struct trie_node *root = (i == 0 ? pfx_table->ipv4 : pfx_table->ipv6); + + if (root) { + struct trie_node *rm_node; + + pthread_rwlock_wrlock(&(pfx_table->lock)); + do { + struct node_data *data = (struct node_data *)(root->data); + + for (unsigned int j = 0; j < data->len; j++) { + struct pfx_record record = {data->ary[j].asn, (root->prefix), root->len, + data->ary[j].max_len, data->ary[j].socket}; + pfx_table_notify_clients(pfx_table, &record, false); + } + rm_node = (trie_remove(root, &(root->prefix), root->len, 0)); + assert(rm_node); + lrtr_free(((struct node_data *)rm_node->data)->ary); + lrtr_free(rm_node->data); + lrtr_free(rm_node); + } while (rm_node != root); + if (i == 0) + pfx_table->ipv4 = NULL; + else + pfx_table->ipv6 = NULL; + + pthread_rwlock_unlock(&(pfx_table->lock)); + } + } + pthread_rwlock_destroy(&(pfx_table->lock)); +} + +int pfx_table_append_elem(struct node_data *data, const struct pfx_record *record) +{ + struct data_elem *tmp = lrtr_realloc(data->ary, sizeof(struct data_elem) * ((data->len) + 1)); + + if (!tmp) + return PFX_ERROR; + data->len++; + data->ary = tmp; + data->ary[data->len - 1].asn = record->asn; + data->ary[data->len - 1].max_len = record->max_len; + data->ary[data->len - 1].socket = record->socket; + return PFX_SUCCESS; +} + +int pfx_table_create_node(struct trie_node **node, const struct pfx_record *record) +{ + int err; + + *node = lrtr_malloc(sizeof(struct trie_node)); + if (!*node) + return PFX_ERROR; + + (*node)->prefix = record->prefix; + (*node)->len = record->min_len; + (*node)->lchild = NULL; + (*node)->rchild = NULL; + (*node)->parent = NULL; + + (*node)->data = lrtr_malloc(sizeof(struct node_data)); + if (!(*node)->data) { + err = PFX_ERROR; + goto free_node; + } + + ((struct node_data *)(*node)->data)->len = 0; + ((struct node_data *)(*node)->data)->ary = NULL; + + err = pfx_table_append_elem(((struct node_data *)(*node)->data), record); + if (err) + goto free_node_data; + + return PFX_SUCCESS; + +free_node_data: + lrtr_free((*node)->data); +free_node: + lrtr_free(*node); + + return err; +} + +struct data_elem *pfx_table_find_elem(const struct node_data *data, const struct pfx_record *record, + unsigned int *index) +{ + for (unsigned int i = 0; i < data->len; i++) { + if (data->ary[i].asn == record->asn && data->ary[i].max_len == record->max_len && + data->ary[i].socket == record->socket) { + if (index) + *index = i; + return &(data->ary[i]); + } + } + return NULL; +} + +// returns pfx_table->ipv4 if record version is LRTR_IPV4 else pfx_table->ipv6 +inline struct trie_node *pfx_table_get_root(const struct pfx_table *pfx_table, const enum lrtr_ip_version ver) +{ + return ver == LRTR_IPV4 ? pfx_table->ipv4 : pfx_table->ipv6; +} + +int pfx_table_del_elem(struct node_data *data, const unsigned int index) +{ + struct data_elem *tmp; + struct data_elem deleted_elem = data->ary[index]; + + // if index is not the last elem in the list, move all other elems backwards in the array + if (index != data->len - 1) { + for (unsigned int i = index; i < data->len - 1; i++) + data->ary[i] = data->ary[i + 1]; + } + + data->len--; + if (!data->len) { + lrtr_free(data->ary); + data->ary = NULL; + return PFX_SUCCESS; + } + + tmp = lrtr_realloc(data->ary, sizeof(struct data_elem) * data->len); + if (!tmp) { + data->ary[data->len] = deleted_elem; + data->len++; + return PFX_ERROR; + } + + data->ary = tmp; + + return PFX_SUCCESS; +} + +RTRLIB_EXPORT int pfx_table_add(struct pfx_table *pfx_table, const struct pfx_record *record) +{ + pthread_rwlock_wrlock(&(pfx_table->lock)); + + struct trie_node *root = pfx_table_get_root(pfx_table, record->prefix.ver); + unsigned int lvl = 0; + + if (root) { + bool found; + struct trie_node *node = trie_lookup_exact(root, &(record->prefix), record->min_len, &lvl, &found); + + if (found) { // node with prefix exists + if (pfx_table_find_elem(node->data, record, NULL)) { + pthread_rwlock_unlock(&pfx_table->lock); + return PFX_DUPLICATE_RECORD; + } + // append record to note_data array + int rtval = pfx_table_append_elem(node->data, record); + + pthread_rwlock_unlock(&pfx_table->lock); + if (rtval == PFX_SUCCESS) + pfx_table_notify_clients(pfx_table, record, true); + return rtval; + } + + // no node with same prefix and prefix_len found + struct trie_node *new_node = NULL; + + if (pfx_table_create_node(&new_node, record) == PFX_ERROR) { + pthread_rwlock_unlock(&pfx_table->lock); + return PFX_ERROR; + } + trie_insert(node, new_node, lvl); + pthread_rwlock_unlock(&pfx_table->lock); + pfx_table_notify_clients(pfx_table, record, true); + return PFX_SUCCESS; + } + + // tree is empty, record will be the root_node + struct trie_node *new_node = NULL; + + if (pfx_table_create_node(&new_node, record) == PFX_ERROR) { + pthread_rwlock_unlock(&pfx_table->lock); + return PFX_ERROR; + } + if (record->prefix.ver == LRTR_IPV4) + pfx_table->ipv4 = new_node; + else + pfx_table->ipv6 = new_node; + + pthread_rwlock_unlock(&pfx_table->lock); + pfx_table_notify_clients(pfx_table, record, true); + return PFX_SUCCESS; +} + +RTRLIB_EXPORT int pfx_table_remove(struct pfx_table *pfx_table, const struct pfx_record *record) +{ + pthread_rwlock_wrlock(&(pfx_table->lock)); + struct trie_node *root = pfx_table_get_root(pfx_table, record->prefix.ver); + + unsigned int lvl = 0; // tree depth were node was found + bool found; + struct trie_node *node = trie_lookup_exact(root, &(record->prefix), record->min_len, &lvl, &found); + + if (!found) { + pthread_rwlock_unlock(&pfx_table->lock); + return PFX_RECORD_NOT_FOUND; + } + + unsigned int index; + struct data_elem *elem = pfx_table_find_elem(node->data, record, &index); + + if (!elem) { + pthread_rwlock_unlock(&pfx_table->lock); + return PFX_RECORD_NOT_FOUND; + } + + struct node_data *ndata = (struct node_data *)node->data; + + if (pfx_table_del_elem(ndata, index) == PFX_ERROR) { + pthread_rwlock_unlock(&pfx_table->lock); + return PFX_ERROR; + } + + if (ndata->len == 0) { + node = trie_remove(node, &(record->prefix), record->min_len, lvl); + assert(node); + + if (node == root) { + if (record->prefix.ver == LRTR_IPV4) + pfx_table->ipv4 = NULL; + else + pfx_table->ipv6 = NULL; + } + assert(((struct node_data *)node->data)->len == 0); + lrtr_free(node->data); + lrtr_free(node); + } + pthread_rwlock_unlock(&pfx_table->lock); + + pfx_table_notify_clients(pfx_table, record, false); + + return PFX_SUCCESS; +} + +bool pfx_table_elem_matches(struct node_data *data, const uint32_t asn, const uint8_t prefix_len) +{ + for (unsigned int i = 0; i < data->len; i++) { + if (data->ary[i].asn != 0 && data->ary[i].asn == asn && prefix_len <= data->ary[i].max_len) + return true; + } + return false; +} + +int pfx_table_node2pfx_record(struct trie_node *node, struct pfx_record *records, const unsigned int ary_len) +{ + struct node_data *data = node->data; + + if (ary_len < data->len) + return PFX_ERROR; + + for (unsigned int i = 0; i < data->len; i++) { + records[i].asn = data->ary[i].asn; + records[i].prefix = node->prefix; + records[i].min_len = node->len; + records[i].max_len = data->ary[i].max_len; + records[i].socket = data->ary[i].socket; + } + return data->len; +} + +inline void pfx_table_free_reason(struct pfx_record **reason, unsigned int *reason_len) +{ + if (reason) { + lrtr_free(*reason); + *reason = NULL; + } + if (reason_len) + *reason_len = 0; +} + +RTRLIB_EXPORT int pfx_table_validate_r(struct pfx_table *pfx_table, struct pfx_record **reason, + unsigned int *reason_len, const uint32_t asn, const struct lrtr_ip_addr *prefix, + const uint8_t prefix_len, enum pfxv_state *result) +{ + // assert(reason_len == NULL || *reason_len == 0); + // assert(reason == NULL || *reason == NULL); + + pthread_rwlock_rdlock(&(pfx_table->lock)); + struct trie_node *root = pfx_table_get_root(pfx_table, prefix->ver); + + if (!root) { + pthread_rwlock_unlock(&pfx_table->lock); + *result = BGP_PFXV_STATE_NOT_FOUND; + pfx_table_free_reason(reason, reason_len); + return PFX_SUCCESS; + } + + unsigned int lvl = 0; + struct trie_node *node = trie_lookup(root, prefix, prefix_len, &lvl); + + if (!node) { + pthread_rwlock_unlock(&pfx_table->lock); + *result = BGP_PFXV_STATE_NOT_FOUND; + pfx_table_free_reason(reason, reason_len); + return PFX_SUCCESS; + } + + if (reason_len && reason) { + *reason_len = ((struct node_data *)node->data)->len; + *reason = lrtr_realloc(*reason, *reason_len * sizeof(struct pfx_record)); + if (!*reason) { + pthread_rwlock_unlock(&pfx_table->lock); + pfx_table_free_reason(reason, reason_len); + return PFX_ERROR; + } + if (pfx_table_node2pfx_record(node, *reason, *reason_len) == PFX_ERROR) { + pthread_rwlock_unlock(&pfx_table->lock); + pfx_table_free_reason(reason, reason_len); + return PFX_ERROR; + } + } + + while (!pfx_table_elem_matches(node->data, asn, prefix_len)) { + if (lrtr_ip_addr_is_zero(lrtr_ip_addr_get_bits( + prefix, lvl++, + 1))) //post-incr lvl, trie_lookup is performed on child_nodes => parent lvl + 1 + node = trie_lookup(node->lchild, prefix, prefix_len, &lvl); + else + node = trie_lookup(node->rchild, prefix, prefix_len, &lvl); + + if (!node) { + pthread_rwlock_unlock(&pfx_table->lock); + *result = BGP_PFXV_STATE_INVALID; + return PFX_SUCCESS; + } + + if (reason_len && reason) { + unsigned int r_len_old = *reason_len; + *reason_len += ((struct node_data *)node->data)->len; + *reason = lrtr_realloc(*reason, *reason_len * sizeof(struct pfx_record)); + struct pfx_record *start = *reason + r_len_old; + + if (!*reason) { + pthread_rwlock_unlock(&pfx_table->lock); + pfx_table_free_reason(reason, reason_len); + return PFX_ERROR; + } + if (pfx_table_node2pfx_record(node, start, ((struct node_data *)node->data)->len) == + PFX_ERROR) { + pthread_rwlock_unlock(&pfx_table->lock); + pfx_table_free_reason(reason, reason_len); + return PFX_ERROR; + } + } + } + + pthread_rwlock_unlock(&pfx_table->lock); + *result = BGP_PFXV_STATE_VALID; + return PFX_SUCCESS; +} + +RTRLIB_EXPORT int pfx_table_validate(struct pfx_table *pfx_table, const uint32_t asn, const struct lrtr_ip_addr *prefix, + const uint8_t prefix_len, enum pfxv_state *result) +{ + return pfx_table_validate_r(pfx_table, NULL, NULL, asn, prefix, prefix_len, result); +} + +RTRLIB_EXPORT int pfx_table_src_remove(struct pfx_table *pfx_table, const struct rtr_socket *socket) +{ + for (unsigned int i = 0; i < 2; i++) { + struct trie_node **root = (i == 0 ? &(pfx_table->ipv4) : &(pfx_table->ipv6)); + + pthread_rwlock_wrlock(&(pfx_table->lock)); + if (*root) { + int rtval = pfx_table_remove_id(pfx_table, root, *root, socket, 0); + + if (rtval == PFX_ERROR) { + pthread_rwlock_unlock(&pfx_table->lock); + return PFX_ERROR; + } + } + pthread_rwlock_unlock(&pfx_table->lock); + } + return PFX_SUCCESS; +} + +int pfx_table_remove_id(struct pfx_table *pfx_table, struct trie_node **root, struct trie_node *node, + const struct rtr_socket *socket, const unsigned int level) +{ + assert(node); + assert(root); + assert(*root); + bool check_node = true; + + while (check_node) { + // data from removed node are replaced from data from child nodes (if children exists), + // same node must be checked again if it was replaced with previous child node data + struct node_data *data = node->data; + + for (unsigned int i = 0; i < data->len; i++) { + while (data->len > i && data->ary[i].socket == socket) { + struct pfx_record record = {data->ary[i].asn, node->prefix, node->len, + data->ary[i].max_len, data->ary[i].socket}; + if (pfx_table_del_elem(data, i) == PFX_ERROR) + return PFX_ERROR; + pfx_table_notify_clients(pfx_table, &record, false); + } + } + if (data->len == 0) { + struct trie_node *rm_node = trie_remove(node, &(node->prefix), node->len, level); + + assert(rm_node); + assert(((struct node_data *)rm_node->data)->len == 0); + lrtr_free(((struct node_data *)rm_node->data)); + lrtr_free(rm_node); + + if (rm_node == *root) { + *root = NULL; + return PFX_SUCCESS; + } else if (rm_node == node) { + return PFX_SUCCESS; + } + } else { + check_node = false; + } + } + + if (node->lchild) { + if (pfx_table_remove_id(pfx_table, root, node->lchild, socket, level + 1) == PFX_ERROR) + return PFX_ERROR; + } + if (node->rchild) + return pfx_table_remove_id(pfx_table, root, node->rchild, socket, level + 1); + return PFX_SUCCESS; +} + +static void pfx_table_for_each_rec(struct trie_node *n, pfx_for_each_fp fp, void *data) +{ + struct pfx_record pfxr; + struct node_data *nd; + + assert(n); + assert(fp); + + nd = (struct node_data *)n->data; + assert(nd); + + if (n->lchild) + pfx_table_for_each_rec(n->lchild, fp, data); + + for (unsigned int i = 0; i < nd->len; i++) { + pfxr.asn = nd->ary[i].asn; + pfxr.prefix = n->prefix; + pfxr.min_len = n->len; + pfxr.max_len = nd->ary[i].max_len; + pfxr.socket = nd->ary[i].socket; + fp(&pfxr, data); + } + + if (n->rchild) + pfx_table_for_each_rec(n->rchild, fp, data); +} + +RTRLIB_EXPORT void pfx_table_for_each_ipv4_record(struct pfx_table *pfx_table, pfx_for_each_fp fp, void *data) +{ + assert(pfx_table); + + if (!pfx_table->ipv4) + return; + + pthread_rwlock_rdlock(&(pfx_table->lock)); + pfx_table_for_each_rec(pfx_table->ipv4, fp, data); + pthread_rwlock_unlock(&pfx_table->lock); +} + +RTRLIB_EXPORT void pfx_table_for_each_ipv6_record(struct pfx_table *pfx_table, pfx_for_each_fp fp, void *data) +{ + assert(pfx_table); + + if (!pfx_table->ipv6) + return; + + pthread_rwlock_rdlock(&(pfx_table->lock)); + pfx_table_for_each_rec(pfx_table->ipv6, fp, data); + pthread_rwlock_unlock(&pfx_table->lock); +} + +static void pfx_table_copy_cb(const struct pfx_record *record, void *data) +{ + struct copy_cb_args *args = data; + + if (record->socket != args->socket) { + if (pfx_table_add(args->pfx_table, record) != PFX_SUCCESS) + args->error = true; + } +} + +int pfx_table_copy_except_socket(struct pfx_table *src_table, struct pfx_table *dst_table, + const struct rtr_socket *socket) +{ + struct copy_cb_args args = {dst_table, socket, false}; + + pfx_table_for_each_ipv4_record(src_table, pfx_table_copy_cb, &args); + if (args.error) + return PFX_ERROR; + + pfx_table_for_each_ipv6_record(src_table, pfx_table_copy_cb, &args); + if (args.error) + return PFX_ERROR; + + return PFX_SUCCESS; +} + +void pfx_table_swap(struct pfx_table *a, struct pfx_table *b) +{ + struct trie_node *ipv4_tmp; + struct trie_node *ipv6_tmp; + + pthread_rwlock_wrlock(&(a->lock)); + pthread_rwlock_wrlock(&(b->lock)); + + ipv4_tmp = a->ipv4; + ipv6_tmp = a->ipv6; + + a->ipv4 = b->ipv4; + a->ipv6 = b->ipv6; + + b->ipv4 = ipv4_tmp; + b->ipv6 = ipv6_tmp; + + pthread_rwlock_unlock(&(b->lock)); + pthread_rwlock_unlock(&(a->lock)); +} + +static void pfx_table_notify_diff_cb(const struct pfx_record *record, void *data) +{ + struct notify_diff_cb_args *args = data; + + if (args->socket == record->socket && args->added) { + if (pfx_table_remove(args->old_table, record) != PFX_SUCCESS) + pfx_table_notify_clients(args->new_table, record, args->added); + } else if (args->socket == record->socket && !args->added) { + pfx_table_notify_clients(args->new_table, record, args->added); + } +} + +void pfx_table_notify_diff(struct pfx_table *new_table, struct pfx_table *old_table, const struct rtr_socket *socket) +{ + pfx_update_fp old_table_fp; + struct notify_diff_cb_args args = {old_table, new_table, socket, new_table->update_fp, true}; + + // Disable update callback for old_table table + old_table_fp = old_table->update_fp; + old_table->update_fp = NULL; + + // Iterate new_table and try to delete every prefix from the given socket in old_table + // If the prefix could not be removed it was added in new_table and the update cb must be called + pfx_table_for_each_ipv4_record(new_table, pfx_table_notify_diff_cb, &args); + pfx_table_for_each_ipv6_record(new_table, pfx_table_notify_diff_cb, &args); + + // Iterate over old_table table and search and remove remaining prefixes from the socket + // issue a remove notification for every one of them, because they are not present in new_table. + args.added = false; + pfx_table_for_each_ipv4_record(old_table, pfx_table_notify_diff_cb, &args); + pfx_table_for_each_ipv6_record(old_table, pfx_table_notify_diff_cb, &args); + + // Restore original state of old_tables update_fp + old_table->update_fp = old_table_fp; +} |