X-Git-Url: http://git.ipfire.org/?p=people%2Fms%2Flibloc.git;a=blobdiff_plain;f=src%2Fdatabase.c;h=73d23945203b0d2d2b4168f7ecf7903b410d6a0b;hp=9db875b87012e7e6327757b0ffb39006f42850ca;hb=f10ebc2df581d27bea60091001a9aba30301c2ea;hpb=5df10c90415a660d463764b1c7a757f3275f4043 diff --git a/src/database.c b/src/database.c index 9db875b..73d2394 100644 --- a/src/database.c +++ b/src/database.c @@ -15,42 +15,229 @@ */ #include +#include #include +#include #include #include #include #include #include +#include #include #include #include #include +#include +#include #include - -#include "libloc-private.h" -#include "as.h" -#include "database.h" -#include "stringpool.h" +#include +#include +#include struct loc_database { struct loc_ctx* ctx; int refcount; - FILE* file; unsigned int version; time_t created_at; off_t vendor; off_t description; // ASes in the database - struct loc_as** as; + struct loc_database_as_v0* as_v0; size_t as_count; + // Network tree + struct loc_database_network_node_v0* network_nodes_v0; + size_t network_nodes_count; + + // Networks + struct loc_database_network_v0* networks_v0; + size_t networks_count; + struct loc_stringpool* pool; }; -LOC_EXPORT int loc_database_new(struct loc_ctx* ctx, struct loc_database** database, size_t pool_size) { +static int loc_database_read_magic(struct loc_database* db, FILE* f) { + struct loc_database_magic magic; + + // Read from file + size_t bytes_read = fread(&magic, 1, sizeof(magic), f); + + // Check if we have been able to read enough data + if (bytes_read < sizeof(magic)) { + ERROR(db->ctx, "Could not read enough data to validate magic bytes\n"); + DEBUG(db->ctx, "Read %zu bytes, but needed %zu\n", bytes_read, sizeof(magic)); + return -ENOMSG; + } + + // Compare magic bytes + if (memcmp(LOC_DATABASE_MAGIC, magic.magic, strlen(LOC_DATABASE_MAGIC)) == 0) { + DEBUG(db->ctx, "Magic value matches\n"); + + // Parse version + db->version = be16toh(magic.version); + DEBUG(db->ctx, "Database version is %u\n", db->version); + + return 0; + } + + ERROR(db->ctx, "Database format is not compatible\n"); + + // Return an error + return 1; +} + +static int loc_database_read_as_section_v0(struct loc_database* db, + FILE* f, const struct loc_database_header_v0* header) { + off_t as_offset = be32toh(header->as_offset); + size_t as_length = be32toh(header->as_length); + + DEBUG(db->ctx, "Reading AS section from %jd (%zu bytes)\n", as_offset, as_length); + + if (as_length > 0) { + db->as_v0 = mmap(NULL, as_length, PROT_READ, + MAP_SHARED, fileno(f), as_offset); + + if (db->as_v0 == MAP_FAILED) + return -errno; + } + + db->as_count = as_length / sizeof(*db->as_v0); + + INFO(db->ctx, "Read %zu ASes from the database\n", db->as_count); + + return 0; +} + +static int loc_database_read_network_nodes_section_v0(struct loc_database* db, + FILE* f, const struct loc_database_header_v0* header) { + off_t network_nodes_offset = be32toh(header->network_tree_offset); + size_t network_nodes_length = be32toh(header->network_tree_length); + + DEBUG(db->ctx, "Reading network nodes section from %jd (%zu bytes)\n", + network_nodes_offset, network_nodes_length); + + if (network_nodes_length > 0) { + db->network_nodes_v0 = mmap(NULL, network_nodes_length, PROT_READ, + MAP_SHARED, fileno(f), network_nodes_offset); + + if (db->network_nodes_v0 == MAP_FAILED) + return -errno; + } + + db->network_nodes_count = network_nodes_length / sizeof(*db->network_nodes_v0); + + INFO(db->ctx, "Read %zu network nodes from the database\n", db->network_nodes_count); + + return 0; +} + +static int loc_database_read_networks_section_v0(struct loc_database* db, + FILE* f, const struct loc_database_header_v0* header) { + off_t networks_offset = be32toh(header->network_data_offset); + size_t networks_length = be32toh(header->network_data_length); + + DEBUG(db->ctx, "Reading networks section from %jd (%zu bytes)\n", + networks_offset, networks_length); + + if (networks_length > 0) { + db->networks_v0 = mmap(NULL, networks_length, PROT_READ, + MAP_SHARED, fileno(f), networks_offset); + + if (db->networks_v0 == MAP_FAILED) + return -errno; + } + + db->networks_count = networks_length / sizeof(*db->networks_v0); + + INFO(db->ctx, "Read %zu networks from the database\n", db->networks_count); + + return 0; +} + +static int loc_database_read_header_v0(struct loc_database* db, FILE* f) { + struct loc_database_header_v0 header; + + // Read from file + size_t size = fread(&header, 1, sizeof(header), f); + + if (size < sizeof(header)) { + ERROR(db->ctx, "Could not read enough data for header\n"); + return -ENOMSG; + } + + // Copy over data + db->created_at = be64toh(header.created_at); + db->vendor = be32toh(header.vendor); + db->description = be32toh(header.description); + + // Open pool + off_t pool_offset = be32toh(header.pool_offset); + size_t pool_length = be32toh(header.pool_length); + + int r = loc_stringpool_open(db->ctx, &db->pool, + f, pool_length, pool_offset); + if (r) + return r; + + // AS section + r = loc_database_read_as_section_v0(db, f, &header); + if (r) + return r; + + // Network Nodes + r = loc_database_read_network_nodes_section_v0(db, f, &header); + if (r) + return r; + + // Networks + r = loc_database_read_networks_section_v0(db, f, &header); + if (r) + return r; + + return 0; +} + +static int loc_database_read_header(struct loc_database* db, FILE* f) { + switch (db->version) { + case 0: + return loc_database_read_header_v0(db, f); + + default: + ERROR(db->ctx, "Incompatible database version: %u\n", db->version); + return 1; + } +} + +static int loc_database_read(struct loc_database* db, FILE* f) { + clock_t start = clock(); + + // Read magic bytes + int r = loc_database_read_magic(db, f); + if (r) + return r; + + // Read the header + r = loc_database_read_header(db, f); + if (r) + return r; + + clock_t end = clock(); + + INFO(db->ctx, "Opened database in %.8fs\n", + (double)(end - start) / CLOCKS_PER_SEC); + + return 0; +} + +LOC_EXPORT int loc_database_new(struct loc_ctx* ctx, struct loc_database** database, FILE* f) { + // Fail on invalid file handle + if (!f) + return -EINVAL; + struct loc_database* db = calloc(1, sizeof(*db)); if (!db) return -ENOMEM; @@ -59,13 +246,9 @@ LOC_EXPORT int loc_database_new(struct loc_ctx* ctx, struct loc_database** datab db->ctx = loc_ref(ctx); db->refcount = 1; - // Save creation time - db->created_at = time(NULL); + DEBUG(db->ctx, "Database object allocated at %p\n", db); - DEBUG(db->ctx, "Database allocated at %p\n", db); - - // Create string pool - int r = loc_stringpool_new(db->ctx, &db->pool, pool_size); + int r = loc_database_read(db, f); if (r) { loc_database_unref(db); return r; @@ -76,14 +259,6 @@ LOC_EXPORT int loc_database_new(struct loc_ctx* ctx, struct loc_database** datab return 0; } -LOC_EXPORT int loc_database_open(struct loc_ctx* ctx, struct loc_database** database, FILE* f) { - int r = loc_database_new(ctx, database, 0); - if (r) - return r; - - return loc_database_read(*database, f); -} - LOC_EXPORT struct loc_database* loc_database_ref(struct loc_database* db) { db->refcount++; @@ -91,21 +266,32 @@ LOC_EXPORT struct loc_database* loc_database_ref(struct loc_database* db) { } static void loc_database_free(struct loc_database* db) { + int r; + DEBUG(db->ctx, "Releasing database %p\n", db); - // Remove references to all ASes - if (db->as) { - for (unsigned int i = 0; i < db->as_count; i++) { - loc_as_unref(db->as[i]); - } - free(db->as); + // Removing all ASes + if (db->as_v0) { + r = munmap(db->as_v0, db->as_count * sizeof(*db->as_v0)); + if (r) + ERROR(db->ctx, "Could not unmap AS section: %s\n", strerror(errno)); } - loc_stringpool_unref(db->pool); + // Remove mapped network sections + if (db->networks_v0) { + r = munmap(db->networks_v0, db->networks_count * sizeof(*db->networks_v0)); + if (r) + ERROR(db->ctx, "Could not unmap networks section: %s\n", strerror(errno)); + } + + // Remove mapped network nodes section + if (db->network_nodes_v0) { + r = munmap(db->network_nodes_v0, db->network_nodes_count * sizeof(*db->network_nodes_v0)); + if (r) + ERROR(db->ctx, "Could not unmap network nodes section: %s\n", strerror(errno)); + } - // Close file - if (db->file) - fclose(db->file); + loc_stringpool_unref(db->pool); loc_unref(db->ctx); free(db); @@ -127,328 +313,263 @@ LOC_EXPORT const char* loc_database_get_vendor(struct loc_database* db) { return loc_stringpool_get(db->pool, db->vendor); } -LOC_EXPORT int loc_database_set_vendor(struct loc_database* db, const char* vendor) { - // Add the string to the string pool - off_t offset = loc_stringpool_add(db->pool, vendor); - if (offset < 0) - return offset; - - db->vendor = offset; - return 0; -} - LOC_EXPORT const char* loc_database_get_description(struct loc_database* db) { return loc_stringpool_get(db->pool, db->description); } -LOC_EXPORT int loc_database_set_description(struct loc_database* db, const char* description) { - // Add the string to the string pool - off_t offset = loc_stringpool_add(db->pool, description); - if (offset < 0) - return offset; - - db->description = offset; - return 0; -} - LOC_EXPORT size_t loc_database_count_as(struct loc_database* db) { return db->as_count; } -static int loc_database_has_as(struct loc_database* db, struct loc_as* as) { - for (unsigned int i = 0; i < db->as_count; i++) { - if (loc_as_cmp(as, db->as[i]) == 0) - return i; - } - - return -1; -} - -static int __loc_as_cmp(const void* as1, const void* as2) { - return loc_as_cmp(*(struct loc_as**)as1, *(struct loc_as**)as2); -} +// Returns the AS at position pos +static int loc_database_fetch_as(struct loc_database* db, struct loc_as** as, off_t pos) { + if ((size_t)pos >= db->as_count) + return -EINVAL; -static void loc_database_sort_ases(struct loc_database* db) { - qsort(db->as, db->as_count, sizeof(*db->as), __loc_as_cmp); -} + DEBUG(db->ctx, "Fetching AS at position %jd\n", pos); -static struct loc_as* __loc_database_add_as(struct loc_database* db, struct loc_as* as) { - // Check if AS exists already - int i = loc_database_has_as(db, as); - if (i >= 0) { - loc_as_unref(as); - - // Select already existing AS - as = db->as[i]; + int r; + switch (db->version) { + case 0: + r = loc_as_new_from_database_v0(db->ctx, db->pool, as, db->as_v0 + pos); + break; - return loc_as_ref(as); + default: + return -1; } - db->as_count++; - - // Make space for the new entry - db->as = realloc(db->as, sizeof(*db->as) * db->as_count); - - // Add the new entry at the end - db->as[db->as_count - 1] = loc_as_ref(as); - - // Sort everything - loc_database_sort_ases(db); + if (r == 0) { + DEBUG(db->ctx, "Got AS%u\n", loc_as_get_number(*as)); + } - return as; + return r; } -LOC_EXPORT struct loc_as* loc_database_add_as(struct loc_database* db, uint32_t number) { - struct loc_as* as; - int r = loc_as_new(db->ctx, db->pool, &as, number); - if (r) - return NULL; +// Performs a binary search to find the AS in the list +LOC_EXPORT int loc_database_get_as(struct loc_database* db, struct loc_as** as, uint32_t number) { + off_t lo = 0; + off_t hi = db->as_count - 1; - return __loc_database_add_as(db, as); -} + // Save start time + clock_t start = clock(); -static int loc_database_read_magic(struct loc_database* db) { - struct loc_database_magic magic; + while (lo <= hi) { + off_t i = (lo + hi) / 2; - // Read from file - size_t bytes_read = fread(&magic, 1, sizeof(magic), db->file); + // Fetch AS in the middle between lo and hi + int r = loc_database_fetch_as(db, as, i); + if (r) + return r; - // Check if we have been able to read enough data - if (bytes_read < sizeof(magic)) { - ERROR(db->ctx, "Could not read enough data to validate magic bytes\n"); - DEBUG(db->ctx, "Read %zu bytes, but needed %zu\n", bytes_read, sizeof(magic)); - return -ENOMSG; - } + // Check if this is a match + uint32_t as_number = loc_as_get_number(*as); + if (as_number == number) { + clock_t end = clock(); - // Compare magic bytes - if (memcmp(LOC_DATABASE_MAGIC, magic.magic, strlen(LOC_DATABASE_MAGIC)) == 0) { - DEBUG(db->ctx, "Magic value matches\n"); + // Log how fast this has been + DEBUG(db->ctx, "Found AS%u in %.8fs\n", as_number, + (double)(end - start) / CLOCKS_PER_SEC); - // Parse version - db->version = ntohs(magic.version); - DEBUG(db->ctx, "Database version is %u\n", db->version); + return 0; + } - return 0; + // If it wasn't, we release the AS and + // adjust our search pointers + loc_as_unref(*as); + + if (as_number < number) { + lo = i + 1; + } else + hi = i - 1; } - ERROR(db->ctx, "Database format is not compatible\n"); + // Nothing found + *as = NULL; - // Return an error return 1; } -static int loc_database_read_as_section_v0(struct loc_database* db, - off_t as_offset, size_t as_length) { - struct loc_database_as_v0 dbobj; - - // Read from the start of the section - int r = fseek(db->file, as_offset, SEEK_SET); - if (r) - return r; +// Returns the network at position pos +static int loc_database_fetch_network(struct loc_database* db, struct loc_network** network, struct in6_addr* address, off_t pos) { + if ((size_t)pos >= db->networks_count) + return -EINVAL; - // Read all ASes - size_t as_count = as_length / sizeof(dbobj); - for (unsigned int i = 0; i < as_count; i++) { - size_t bytes_read = fread(&dbobj, 1, sizeof(dbobj), db->file); - if (bytes_read < sizeof(dbobj)) { - ERROR(db->ctx, "Could not read an AS object\n"); - return -ENOMSG; - } + DEBUG(db->ctx, "Fetching network at position %jd\n", pos); - // Allocate a new AS - struct loc_as* as; - r = loc_as_new_from_database_v0(db->ctx, db->pool, &as, &dbobj); - if (r) - return r; + int r; + switch (db->version) { + case 0: + r = loc_network_new_from_database_v0(db->ctx, network, address, db->networks_v0 + pos); + break; - // Attach it to the database - as = __loc_database_add_as(db, as); - loc_as_unref(as); + default: + return -1; } - INFO(db->ctx, "Read %zu ASes from the database\n", db->as_count); + if (r == 0) { + char* string = loc_network_str(*network); + DEBUG(db->ctx, "Got network %s\n", string); + free(string); + } - return 0; + return r; } -static int loc_database_read_header_v0(struct loc_database* db) { - struct loc_database_header_v0 header; - - // Read from file - size_t size = fread(&header, 1, sizeof(header), db->file); +static int __loc_database_lookup_leaf_node(struct loc_database* db, const struct in6_addr* address, + struct loc_network** network, struct in6_addr* network_address, + const struct loc_database_network_node_v0* node) { + // Check if this node is a leaf node + if (node->zero != htobe32(0xffffffff)) + return 1; - if (size < sizeof(header)) { - ERROR(db->ctx, "Could not read enough data for header\n"); - return -ENOMSG; - } + DEBUG(db->ctx, "Node is a leaf: %jd\n", node - db->network_nodes_v0); - // Copy over data - db->created_at = be64toh(header.created_at); - db->vendor = ntohl(header.vendor); - db->description = ntohl(header.description); - - // Open pool - off_t pool_offset = ntohl(header.pool_offset); - size_t pool_length = ntohl(header.pool_length); - - int r = loc_stringpool_read(db->pool, db->file, pool_offset, pool_length); + // Fetch the network + int r = loc_database_fetch_network(db, network, + network_address, be32toh(node->one)); if (r) return r; - // AS section - off_t as_offset = ntohl(header.as_offset); - size_t as_length = ntohl(header.as_length); + // Check if the given IP address is inside the network + r = loc_network_match_address(*network, address); + if (r) { + DEBUG(db->ctx, "Searched address is not part of the network\n"); - r = loc_database_read_as_section_v0(db, as_offset, as_length); - if (r) - return r; + loc_network_unref(*network); + *network = NULL; + return 1; + } + // A network was found and the IP address matches return 0; } -static int loc_database_read_header(struct loc_database* db) { - switch (db->version) { - case 0: - return loc_database_read_header_v0(db); +// Returns the highest result available +static int __loc_database_lookup_max(struct loc_database* db, const struct in6_addr* address, + struct loc_network** network, struct in6_addr* network_address, + const struct loc_database_network_node_v0* node, int level) { - default: - ERROR(db->ctx, "Incompatible database version: %u\n", db->version); - return 1; - } -} + // If the node is a leaf node, we end here + int r = __loc_database_lookup_leaf_node(db, address, network, network_address, node); + if (r <= 0) + return r; -LOC_EXPORT int loc_database_read(struct loc_database* db, FILE* f) { - // Copy the file pointer and work on that so we don't care if - // the calling function closes the file - int fd = fileno(f); + off_t node_index; - // Make a copy - fd = dup(fd); + // Try to go down the ones path first + if (node->one) { + node_index = be32toh(node->one); + in6_addr_set_bit(network_address, level, 1); - // Retrieve a file pointer - db->file = fdopen(fd, "r"); - if (!db->file) - return -errno; + // Check boundaries + if (node_index > 0 && (size_t)node_index <= db->network_nodes_count) { + r = __loc_database_lookup_max(db, address, network, network_address, + db->network_nodes_v0 + node_index, level + 1); - int r = fseek(db->file, 0, SEEK_SET); - if (r) - return r; + // Abort when match was found or error + if (r <= 0) + return r; + } + } - // Read magic bytes - r = loc_database_read_magic(db); - if (r) - return r; + // ... and if that fails, we try to go down one step on a zero + // branch and then try the ones again... + if (node->zero) { + node_index = be32toh(node->zero); + in6_addr_set_bit(network_address, level, 0); - // Read the header - r = loc_database_read_header(db); - if (r) - return r; + // Check boundaries + if (node_index > 0 && (size_t)node_index <= db->network_nodes_count) { + r = __loc_database_lookup_max(db, address, network, network_address, + db->network_nodes_v0 + node_index, level + 1); - return 0; -} - -static void loc_database_make_magic(struct loc_database* db, struct loc_database_magic* magic) { - // Copy magic bytes - for (unsigned int i = 0; i < strlen(LOC_DATABASE_MAGIC); i++) - magic->magic[i] = LOC_DATABASE_MAGIC[i]; + // Abort when match was found or error + if (r <= 0) + return r; + } + } - // Set version - magic->version = htons(LOC_DATABASE_VERSION); + // End of path + return 1; } -static void loc_database_align_page_boundary(off_t* offset, FILE* f) { - // Move to next page boundary - while (*offset % LOC_DATABASE_PAGE_SIZE > 0) - *offset += fwrite("", 1, 1, f); -} +// Searches for an exact match along the path +static int __loc_database_lookup(struct loc_database* db, const struct in6_addr* address, + struct loc_network** network, struct in6_addr* network_address, + const struct loc_database_network_node_v0* node, int level) { + // If the node is a leaf node, we end here + int r = __loc_database_lookup_leaf_node(db, address, network, network_address, node); + if (r <= 0) + return r; -static int loc_database_write_pool(struct loc_database* db, struct loc_database_header_v0* header, off_t* offset, FILE* f) { - // Save the offset of the pool section - DEBUG(db->ctx, "Pool starts at %jd bytes\n", *offset); - header->pool_offset = htonl(*offset); + off_t node_index; - // Write the pool - size_t pool_length = loc_stringpool_write(db->pool, f); - *offset += pool_length; + // Follow the path + int bit = in6_addr_get_bit(address, level); + in6_addr_set_bit(network_address, level, bit); - DEBUG(db->ctx, "Pool has a length of %zu bytes\n", pool_length); - header->pool_length = htonl(pool_length); + if (bit == 0) + node_index = be32toh(node->zero); + else + node_index = be32toh(node->one); - return 0; -} + // If we point back to root, the path ends here + if (node_index == 0) { + DEBUG(db->ctx, "Tree ends here\n"); + return 1; + } -static int loc_database_write_as_section(struct loc_database* db, - struct loc_database_header_v0* header, off_t* offset, FILE* f) { - DEBUG(db->ctx, "AS section starts at %jd bytes\n", *offset); - header->as_offset = htonl(*offset); + // Check boundaries + if ((size_t)node_index >= db->network_nodes_count) + return -EINVAL; - size_t as_length = 0; + // Move on to the next node + r = __loc_database_lookup(db, address, network, network_address, + db->network_nodes_v0 + node_index, level + 1); - struct loc_database_as_v0 dbas; - for (unsigned int i = 0; i < db->as_count; i++) { - // Convert AS into database format - loc_as_to_database_v0(db->as[i], &dbas); + // End here if a result was found + if (r == 0) + return r; - // Write to disk - offset += fwrite(&dbas, 1, sizeof(dbas), f); - as_length += sizeof(dbas); - } + // Raise any errors + else if (r < 0) + return r; - DEBUG(db->ctx, "AS section has a length of %zu bytes\n", as_length); - header->as_length = htonl(as_length); + DEBUG(db->ctx, "Could not find an exact match at %u\n", level); - return 0; + // If nothing was found, we have to search for an inexact match + return __loc_database_lookup_max(db, address, network, network_address, node, level); } -LOC_EXPORT int loc_database_write(struct loc_database* db, FILE* f) { - struct loc_database_magic magic; - loc_database_make_magic(db, &magic); - - // Make the header - struct loc_database_header_v0 header; - header.created_at = htobe64(db->created_at); - header.vendor = htonl(db->vendor); - header.description = htonl(db->description); - - int r; - off_t offset = 0; +LOC_EXPORT int loc_database_lookup(struct loc_database* db, + struct in6_addr* address, struct loc_network** network) { + struct in6_addr network_address; + memset(&network_address, 0, sizeof(network_address)); - // Start writing at the beginning of the file - r = fseek(f, 0, SEEK_SET); - if (r) - return r; + *network = NULL; - // Write the magic - offset += fwrite(&magic, 1, sizeof(magic), f); + // Save start time + clock_t start = clock(); - // Skip the space we need to write the header later - r = fseek(f, sizeof(header), SEEK_CUR); - if (r) { - DEBUG(db->ctx, "Could not seek to position after header\n"); - return r; - } - offset += sizeof(header); + int r = __loc_database_lookup(db, address, network, &network_address, + db->network_nodes_v0, 0); - loc_database_align_page_boundary(&offset, f); + clock_t end = clock(); - // Write all ASes - r = loc_database_write_as_section(db, &header, &offset, f); - if (r) - return r; + // Log how fast this has been + DEBUG(db->ctx, "Executed network search in %.8fs\n", + (double)(end - start) / CLOCKS_PER_SEC); - loc_database_align_page_boundary(&offset, f); + return r; +} - // Write pool - r = loc_database_write_pool(db, &header, &offset, f); - if (r) - return r; +LOC_EXPORT int loc_database_lookup_from_string(struct loc_database* db, + const char* string, struct loc_network** network) { + struct in6_addr address; - // Write the header - r = fseek(f, sizeof(magic), SEEK_SET); + int r = loc_parse_address(db->ctx, string, &address); if (r) return r; - offset += fwrite(&header, 1, sizeof(header), f); - - return 0; + return loc_database_lookup(db, &address, network); }