/* * This file is part of the Distributed Network Block Device 3 * * Copyright(c) 2011-2012 Johann Latocha * * This file may be licensed under the terms of of the * GNU General Public License Version 2 (the ``GPL''). * * Software distributed under the License is distributed * on an ``AS IS'' basis, WITHOUT WARRANTY OF ANY KIND, either * express or implied. See the GPL for the specific language * governing rights and limitations. * * You should have received a copy of the GPL along with this * program. If not, go to http://www.gnu.org/licenses/gpl.html * or write to the Free Software Foundation, Inc., * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. * */ #include "helper.h" #include "image.h" #include "uplink.h" #include "locks.h" #include "rpc.h" #include "altservers.h" #include "../shared/sockhelper.h" #include "../shared/timing.h" #include "../shared/protocol.h" #include "../serialize.h" #include #ifdef __linux__ #include #endif #ifdef __FreeBSD__ #include #include #include #endif #include #include #include static dnbd3_client_t *_clients[SERVER_MAX_CLIENTS]; static int _num_clients = 0; static pthread_spinlock_t _clients_lock; static char nullbytes[500]; static atomic_uint_fast64_t totalBytesSent = 0; // Adding and removing clients -- list management static bool addToList(dnbd3_client_t *client); static void removeFromList(dnbd3_client_t *client); static dnbd3_client_t* freeClientStruct(dnbd3_client_t *client); static inline bool recv_request_header(int sock, dnbd3_request_t *request) { ssize_t ret, fails = 0; #ifdef AFL_MODE sock = 0; #endif // Read request header from socket while ( ( ret = recv( sock, request, sizeof(*request), MSG_WAITALL ) ) != sizeof(*request) ) { if ( errno == EINTR && ++fails < 10 ) continue; if ( ret >= 0 || ++fails > SOCKET_TIMEOUT_CLIENT_RETRIES ) return false; if ( errno == EAGAIN ) continue; logadd( LOG_DEBUG2, "Error receiving request: Could not read message header (%d/%d, e=%d)\n", (int)ret, (int)sizeof(*request), errno ); return false; } // Make sure all bytes are in the right order (endianness) fixup_request( *request ); if ( request->magic != dnbd3_packet_magic ) { logadd( LOG_DEBUG2, "Magic in client request incorrect (cmd: %d, len: %d)\n", (int)request->cmd, (int)request->size ); return false; } // Payload sanity check if ( request->cmd != CMD_GET_BLOCK && request->size > MAX_PAYLOAD ) { logadd( LOG_WARNING, "Client tries to send a packet of type %d with %d bytes payload. Dropping client.", (int)request->cmd, (int)request->size ); return false; } return true; } static inline bool recv_request_payload(int sock, uint32_t size, serialized_buffer_t *payload) { #ifdef AFL_MODE sock = 0; #endif if ( size == 0 ) { logadd( LOG_ERROR, "Called recv_request_payload() to receive 0 bytes" ); return false; } if ( size > MAX_PAYLOAD ) { logadd( LOG_ERROR, "Called recv_request_payload() for more bytes than the passed buffer could hold!" ); return false; } if ( sock_recv( sock, payload->buffer, size ) != (ssize_t)size ) { logadd( LOG_DEBUG1, "Could not receive request payload of length %d\n", (int)size ); return false; } // Prepare payload buffer for reading serializer_reset_read( payload, size ); return true; } /** * Send reply with optional payload. payload can be null. The caller has to * acquire the sendMutex first. */ static inline bool send_reply(int sock, dnbd3_reply_t *reply, void *payload) { const uint32_t size = reply->size; fixup_reply( *reply ); if ( sock_sendAll( sock, reply, sizeof(dnbd3_reply_t), 1 ) != sizeof(dnbd3_reply_t) ) { logadd( LOG_DEBUG1, "Sending reply header to client failed" ); return false; } if ( size != 0 && payload != NULL ) { if ( sock_sendAll( sock, payload, size, 1 ) != (ssize_t)size ) { logadd( LOG_DEBUG1, "Sending payload of %"PRIu32" bytes to client failed", size ); return false; } } return true; } /** * Send given amount of null bytes. The caller has to acquire the sendMutex first. */ static inline bool sendPadding( const int fd, uint32_t bytes ) { ssize_t ret; while ( bytes >= sizeof(nullbytes) ) { ret = sock_sendAll( fd, nullbytes, sizeof(nullbytes), 2 ); if ( ret <= 0 ) return false; bytes -= (uint32_t)ret; } return sock_sendAll( fd, nullbytes, bytes, 2 ) == (ssize_t)bytes; } void net_init() { spin_init( &_clients_lock, PTHREAD_PROCESS_PRIVATE ); } void* net_handleNewConnection(void *clientPtr) { dnbd3_client_t * const client = (dnbd3_client_t *)clientPtr; dnbd3_request_t request; // Await data from client. Since this is a fresh connection, we expect data right away sock_setTimeout( client->sock, _clientTimeout ); do { #ifdef AFL_MODE const int ret = (int)recv( 0, &request, sizeof(request), MSG_WAITALL ); #else const int ret = (int)recv( client->sock, &request, sizeof(request), MSG_WAITALL ); #endif // It's expected to be a real dnbd3 client // Check request for validity. This implicitly dictates that all HTTP requests are more than 24 bytes... if ( ret != (int)sizeof(request) ) { logadd( LOG_DEBUG2, "Error receiving request: Could not read message header (%d/%d, e=%d)", (int)ret, (int)sizeof(request), errno ); goto fail_preadd; } if ( request.magic != dnbd3_packet_magic ) { // Let's see if this looks like an HTTP request if ( ((char*)&request)[0] == 'G' || ((char*)&request)[0] == 'P' ) { // Close enough... rpc_sendStatsJson( client->sock, &client->host, &request, ret ); } else { logadd( LOG_DEBUG1, "Magic in client handshake incorrect" ); } goto fail_preadd; } // Magic OK, untangle byte order if required fixup_request( request ); if ( request.cmd != CMD_SELECT_IMAGE ) { logadd( LOG_WARNING, "Client sent != CMD_SELECT_IMAGE in handshake (got cmd=%d, size=%d), dropping client.", (int)request.cmd, (int)request.size ); goto fail_preadd; } } while (0); // Fully init client struct spin_init( &client->lock, PTHREAD_PROCESS_PRIVATE ); pthread_mutex_init( &client->sendMutex, NULL ); spin_lock( &client->lock ); host_to_string( &client->host, client->hostName, HOSTNAMELEN ); client->hostName[HOSTNAMELEN-1] = '\0'; spin_unlock( &client->lock ); client->bytesSent = 0; if ( !addToList( client ) ) { freeClientStruct( client ); logadd( LOG_WARNING, "Could not add new client to list when connecting" ); return NULL; } dnbd3_reply_t reply; dnbd3_image_t *image = NULL; int image_file = -1; int num; bool bOk = false; bool hasName = false; serialized_buffer_t payload; uint16_t rid, client_version; uint64_t start, end; dnbd3_server_entry_t server_list[NUMBER_SERVERS]; // Set to zero to make valgrind happy memset( &reply, 0, sizeof(reply) ); memset( &payload, 0, sizeof(payload) ); reply.magic = dnbd3_packet_magic; // Receive first packet's payload if ( recv_request_payload( client->sock, request.size, &payload ) ) { char *image_name; client_version = serializer_get_uint16( &payload ); image_name = serializer_get_string( &payload ); rid = serializer_get_uint16( &payload ); const uint8_t flags = serializer_get_uint8( &payload ); client->isServer = ( flags & FLAGS8_SERVER ); if ( request.size < 3 || !image_name || client_version < MIN_SUPPORTED_CLIENT ) { if ( client_version < MIN_SUPPORTED_CLIENT ) { logadd( LOG_DEBUG1, "Client %s too old", client->hostName ); } else { logadd( LOG_DEBUG1, "Incomplete handshake received from %s", client->hostName ); } } else { if ( !client->isServer || !_isProxy ) { // Is a normal client, or we're not proxy image = image_getOrLoad( image_name, rid ); } else if ( _backgroundReplication != BGR_FULL && ( flags & FLAGS8_BG_REP ) ) { // We're a proxy, client is another proxy, we don't do BGR, but connecting proxy does... // Reject, as this would basically force this proxy to do BGR too. image = image_get( image_name, rid, true ); if ( image != NULL && image->cache_map != NULL ) { // Only exception is if the image is complete locally image = image_release( image ); } } else if ( _lookupMissingForProxy ) { // No BGR mismatch and we're told to lookup missing images on a known uplink server // if the requesting client is a proxy image = image_getOrLoad( image_name, rid ); } else { // No BGR mismatch, but don't lookup if image is unknown locally image = image_get( image_name, rid, true ); } spin_lock( &client->lock ); client->image = image; spin_unlock( &client->lock ); if ( image == NULL ) { //logadd( LOG_DEBUG1, "Client requested non-existent image '%s' (rid:%d), rejected\n", image_name, (int)rid ); } else if ( !image->working ) { logadd( LOG_DEBUG1, "Client %s requested non-working image '%s' (rid:%d), rejected\n", client->hostName, image_name, (int)rid ); } else { bool penalty; // Image is fine so far, but occasionally drop a client if the uplink for the image is clogged or unavailable bOk = true; if ( image->cache_map != NULL ) { spin_lock( &image->lock ); if ( image->uplink == NULL || image->uplink->cacheFd == -1 || image->uplink->queueLen > SERVER_UPLINK_QUEUELEN_THRES ) { bOk = ( rand() % 4 ) == 1; } penalty = bOk && image->uplink != NULL && image->uplink->cacheFd == -1; spin_unlock( &image->lock ); if ( penalty ) { // Wait 100ms if local caching is not working so this usleep( 100000 ); // server gets a penalty and is less likely to be selected } } if ( bOk ) { spin_lock( &image->lock ); image_file = image->readFd; if ( !client->isServer ) { // Only update immediately if this is a client. Servers are handled on disconnect. timing_get( &image->atime ); } spin_unlock( &image->lock ); serializer_reset_write( &payload ); serializer_put_uint16( &payload, client_version < 3 ? client_version : PROTOCOL_VERSION ); // XXX: Since messed up fuse client was messed up before :( serializer_put_string( &payload, image->name ); serializer_put_uint16( &payload, (uint16_t)image->rid ); serializer_put_uint64( &payload, image->virtualFilesize ); reply.cmd = CMD_SELECT_IMAGE; reply.size = serializer_get_written_length( &payload ); if ( !send_reply( client->sock, &reply, &payload ) ) { bOk = false; } } } } } if ( bOk ) { // add artificial delay if applicable if ( client->isServer && _serverPenalty != 0 ) { usleep( _serverPenalty ); } else if ( !client->isServer && _clientPenalty != 0 ) { usleep( _clientPenalty ); } // client handling mainloop while ( recv_request_header( client->sock, &request ) ) { if ( _shutdown ) break; switch ( request.cmd ) { case CMD_GET_BLOCK:; const uint64_t offset = request.offset_small; // Copy to full uint64 to prevent repeated masking if ( offset >= image->virtualFilesize ) { // Sanity check logadd( LOG_WARNING, "Client %s requested non-existent block", client->hostName ); reply.size = 0; reply.cmd = CMD_ERROR; send_reply( client->sock, &reply, NULL ); break; } if ( offset + request.size > image->virtualFilesize ) { // Sanity check logadd( LOG_WARNING, "Client %s requested data block that extends beyond image size", client->hostName ); reply.size = 0; reply.cmd = CMD_ERROR; send_reply( client->sock, &reply, NULL ); break; } if ( request.size != 0 && image->cache_map != NULL ) { // This is a proxyed image, check if we need to relay the request... start = offset & ~(uint64_t)(DNBD3_BLOCK_SIZE - 1); end = (offset + request.size + DNBD3_BLOCK_SIZE - 1) & ~(uint64_t)(DNBD3_BLOCK_SIZE - 1); bool isCached = true; spin_lock( &image->lock ); // Check again as we only aquired the lock just now if ( image->cache_map != NULL ) { const uint64_t firstByteInMap = start >> 15; const uint64_t lastByteInMap = (end - 1) >> 15; uint64_t pos; // Middle - quick checking if ( isCached ) { pos = firstByteInMap + 1; while ( pos < lastByteInMap ) { if ( image->cache_map[pos] != 0xff ) { isCached = false; break; } ++pos; } } // First byte if ( isCached ) { pos = start; do { const int map_x = (pos >> 12) & 7; // mod 8 const uint8_t bit_mask = (uint8_t)( 1 << map_x ); if ( (image->cache_map[firstByteInMap] & bit_mask) == 0 ) { isCached = false; break; } pos += DNBD3_BLOCK_SIZE; } while ( firstByteInMap == (pos >> 15) && pos < end ); } // Last byte - only check if request spans multiple bytes in cache map if ( isCached && firstByteInMap != lastByteInMap ) { pos = lastByteInMap << 15; while ( pos < end ) { assert( lastByteInMap == (pos >> 15) ); const int map_x = (pos >> 12) & 7; // mod 8 const uint8_t bit_mask = (uint8_t)( 1 << map_x ); if ( (image->cache_map[lastByteInMap] & bit_mask) == 0 ) { isCached = false; break; } pos += DNBD3_BLOCK_SIZE; } } } spin_unlock( &image->lock ); if ( !isCached ) { if ( !uplink_request( client, request.handle, offset, request.size, request.hops ) ) { logadd( LOG_DEBUG1, "Could not relay uncached request from %s to upstream proxy, disabling image %s:%d", client->hostName, image->name, image->rid ); image->working = false; goto exit_client_cleanup; } break; // DONE, exit request.cmd switch } } reply.cmd = CMD_GET_BLOCK; reply.size = request.size; reply.handle = request.handle; fixup_reply( reply ); const bool lock = image->uplink != NULL; if ( lock ) pthread_mutex_lock( &client->sendMutex ); // Send reply header if ( send( client->sock, &reply, sizeof(dnbd3_reply_t), (request.size == 0 ? 0 : MSG_MORE) ) != sizeof(dnbd3_reply_t) ) { if ( lock ) pthread_mutex_unlock( &client->sendMutex ); logadd( LOG_DEBUG1, "Sending CMD_GET_BLOCK reply header to %s failed", client->hostName ); goto exit_client_cleanup; } if ( request.size != 0 ) { // Send payload if request length > 0 size_t done = 0; off_t foffset = (off_t)offset; size_t realBytes; if ( offset + request.size <= image->realFilesize ) { realBytes = request.size; } else { realBytes = (size_t)(image->realFilesize - offset); } while ( done < realBytes ) { // TODO: Should we consider EOPNOTSUPP on BSD for sendfile and fallback to read/write? // Linux would set EINVAL or ENOSYS instead, which it unfortunately also does for a couple of other failures :/ // read/write would kill performance anyways so a fallback would probably be of little use either way. #ifdef AFL_MODE char buf[1000]; size_t cnt = realBytes - done; if ( cnt > 1000 ) { cnt = 1000; } const ssize_t sent = pread( image_file, buf, cnt, foffset ); if ( sent > 0 ) { //write( client->sock, buf, sent ); // This is not verified in any way, so why even do it... } else { const int err = errno; #elif defined(__linux__) const ssize_t sent = sendfile( client->sock, image_file, &foffset, realBytes - done ); if ( sent <= 0 ) { const int err = errno; #elif defined(__FreeBSD__) off_t sent; const int ret = sendfile( image_file, client->sock, foffset, realBytes - done, NULL, &sent, 0 ); if ( ret == -1 || sent == 0 ) { const int err = errno; if ( ret == -1 ) { if ( err == EAGAIN || err == EINTR ) { // EBUSY? manpage doesn't explicitly mention *sent here.. But then again we dont set the according flag anyways done += sent; continue; } sent = -1; } #endif if ( lock ) pthread_mutex_unlock( &client->sendMutex ); if ( sent == -1 ) { if ( err != EPIPE && err != ECONNRESET && err != ESHUTDOWN && err != EAGAIN && err != EWOULDBLOCK ) { logadd( LOG_DEBUG1, "sendfile to %s failed (image to net. sent %d/%d, errno=%d)", client->hostName, (int)done, (int)realBytes, err ); } if ( err == EBADF || err == EFAULT || err == EINVAL || err == EIO ) { logadd( LOG_INFO, "Disabling %s:%d", image->name, image->rid ); image->working = false; } } goto exit_client_cleanup; } done += sent; } if ( request.size > (uint32_t)realBytes ) { if ( !sendPadding( client->sock, request.size - (uint32_t)realBytes ) ) { if ( lock ) pthread_mutex_unlock( &client->sendMutex ); goto exit_client_cleanup; } } } if ( lock ) pthread_mutex_unlock( &client->sendMutex ); // Global per-client counter client->bytesSent += request.size; // Increase counter for statistics. break; case CMD_GET_SERVERS: // Build list of known working alt servers num = altservers_getListForClient( &client->host, server_list, NUMBER_SERVERS ); reply.cmd = CMD_GET_SERVERS; reply.size = (uint32_t)( num * sizeof(dnbd3_server_entry_t) ); pthread_mutex_lock( &client->sendMutex ); send_reply( client->sock, &reply, server_list ); pthread_mutex_unlock( &client->sendMutex ); goto set_name; break; case CMD_KEEPALIVE: reply.cmd = CMD_KEEPALIVE; reply.size = 0; pthread_mutex_lock( &client->sendMutex ); send_reply( client->sock, &reply, NULL ); pthread_mutex_unlock( &client->sendMutex ); set_name: ; if ( !hasName ) { hasName = true; setThreadName( client->hostName ); } break; case CMD_SET_CLIENT_MODE: client->isServer = false; break; case CMD_GET_CRC32: reply.cmd = CMD_GET_CRC32; pthread_mutex_lock( &client->sendMutex ); if ( image->crc32 == NULL ) { reply.size = 0; send_reply( client->sock, &reply, NULL ); } else { const uint32_t size = reply.size = (uint32_t)( (IMGSIZE_TO_HASHBLOCKS(image->realFilesize) + 1) * sizeof(uint32_t) ); send_reply( client->sock, &reply, NULL ); send( client->sock, &image->masterCrc32, sizeof(uint32_t), MSG_MORE ); send( client->sock, image->crc32, size - sizeof(uint32_t), 0 ); } pthread_mutex_unlock( &client->sendMutex ); break; default: logadd( LOG_ERROR, "Unknown command from client %s: %d", client->hostName, (int)request.cmd ); break; } } } exit_client_cleanup: ; // First remove from list, then add to counter to prevent race condition removeFromList( client ); totalBytesSent += client->bytesSent; // Access time, but only if client didn't just probe if ( image != NULL ) { spin_lock( &image->lock ); if ( client->bytesSent > DNBD3_BLOCK_SIZE * 10 ) { timing_get( &image->atime ); } spin_unlock( &image->lock ); } freeClientStruct( client ); // This will also call image_release on client->image return NULL ; fail_preadd: ; close( client->sock ); free( client ); return NULL; } /** * Get list of all clients. */ struct json_t* net_getListAsJson() { json_t *jsonClients = json_array(); json_t *clientStats; int imgId, isServer; uint64_t bytesSent; char host[HOSTNAMELEN]; host[HOSTNAMELEN-1] = '\0'; spin_lock( &_clients_lock ); for ( int i = 0; i < _num_clients; ++i ) { dnbd3_client_t * const client = _clients[i]; if ( client == NULL || client->image == NULL ) continue; spin_lock( &client->lock ); // Unlock so we give other threads a chance to access the client list. // We might not get an atomic snapshot of the currently connected clients, // but that doesn't really make a difference anyways. spin_unlock( &_clients_lock ); strncpy( host, client->hostName, HOSTNAMELEN - 1 ); imgId = client->image->id; isServer = (int)client->isServer; bytesSent = client->bytesSent; spin_unlock( &client->lock ); clientStats = json_pack( "{sssisisI}", "address", host, "imageId", imgId, "isServer", isServer, "bytesSent", (json_int_t)bytesSent ); json_array_append_new( jsonClients, clientStats ); spin_lock( &_clients_lock ); } spin_unlock( &_clients_lock ); return jsonClients; } /** * Get number of clients connected, total bytes sent, or both. * we don't unlock the list while iterating or we might get an * incorrect result if a client is disconnecting while iterating. */ void net_getStats(int *clientCount, int *serverCount, uint64_t *bytesSent) { int cc = 0, sc = 0; uint64_t bs = 0; spin_lock( &_clients_lock ); for ( int i = 0; i < _num_clients; ++i ) { const dnbd3_client_t * const client = _clients[i]; if ( client == NULL || client->image == NULL ) continue; if ( client->isServer ) { sc += 1; } else { cc += 1; } bs += client->bytesSent; } spin_unlock( &_clients_lock ); if ( clientCount != NULL ) { *clientCount = cc; } if ( serverCount != NULL ) { *serverCount = sc; } if ( bytesSent != NULL ) { *bytesSent = totalBytesSent + bs; } } void net_disconnectAll() { int i; spin_lock( &_clients_lock ); for (i = 0; i < _num_clients; ++i) { if ( _clients[i] == NULL ) continue; dnbd3_client_t * const client = _clients[i]; spin_lock( &client->lock ); if ( client->sock >= 0 ) shutdown( client->sock, SHUT_RDWR ); spin_unlock( &client->lock ); } spin_unlock( &_clients_lock ); } void net_waitForAllDisconnected() { int retries = 10, count, i; do { count = 0; spin_lock( &_clients_lock ); for (i = 0; i < _num_clients; ++i) { if ( _clients[i] == NULL ) continue; count++; } spin_unlock( &_clients_lock ); if ( count != 0 ) { logadd( LOG_INFO, "%d clients still active...\n", count ); sleep( 1 ); } } while ( count != 0 && --retries > 0 ); _num_clients = 0; } /* +++ * Client list. * * Adding and removing clients. */ /** * Remove a client from the clients array * Locks on: _clients_lock */ static void removeFromList(dnbd3_client_t *client) { int i; spin_lock( &_clients_lock ); for ( i = _num_clients - 1; i >= 0; --i ) { if ( _clients[i] == client ) { _clients[i] = NULL; } if ( _clients[i] == NULL && i + 1 == _num_clients ) --_num_clients; } spin_unlock( &_clients_lock ); } /** * Free the client struct recursively. * !! Make sure to call this function after removing the client from _dnbd3_clients !! * Locks on: _clients[].lock, _images[].lock * might call functions that lock on _images, _image[], uplink.queueLock, client.sendMutex */ static dnbd3_client_t* freeClientStruct(dnbd3_client_t *client) { spin_lock( &client->lock ); pthread_mutex_lock( &client->sendMutex ); if ( client->sock != -1 ) close( client->sock ); client->sock = -1; pthread_mutex_unlock( &client->sendMutex ); if ( client->image != NULL ) { spin_lock( &client->image->lock ); if ( client->image->uplink != NULL ) uplink_removeClient( client->image->uplink, client ); spin_unlock( &client->image->lock ); client->image = image_release( client->image ); } spin_unlock( &client->lock ); spin_destroy( &client->lock ); pthread_mutex_destroy( &client->sendMutex ); free( client ); return NULL ; } //###// /** * Add client to the clients array. * Locks on: _clients_lock */ static bool addToList(dnbd3_client_t *client) { int i; spin_lock( &_clients_lock ); for (i = 0; i < _num_clients; ++i) { if ( _clients[i] != NULL ) continue; _clients[i] = client; spin_unlock( &_clients_lock ); return true; } if ( _num_clients >= _maxClients ) { spin_unlock( &_clients_lock ); logadd( LOG_ERROR, "Maximum number of clients reached!" ); return false; } _clients[_num_clients++] = client; spin_unlock( &_clients_lock ); return true; }