/* Copyright (c) 2003-2014, Arvid Norberg All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. * Neither the name of the author nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include #include #include #include #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING #include // for va_start, va_end #include // for vsnprintf #endif #include "libtorrent/peer_connection.hpp" #include "libtorrent/identify_client.hpp" #include "libtorrent/entry.hpp" #include "libtorrent/bencode.hpp" #include "libtorrent/alert_types.hpp" #include "libtorrent/invariant_check.hpp" #include "libtorrent/io.hpp" #include "libtorrent/file.hpp" #include "libtorrent/version.hpp" #include "libtorrent/extensions.hpp" #include "libtorrent/aux_/session_interface.hpp" #include "libtorrent/policy.hpp" #include "libtorrent/socket_type.hpp" #include "libtorrent/assert.hpp" #include "libtorrent/broadcast_socket.hpp" #include "libtorrent/torrent.hpp" #include "libtorrent/peer_info.hpp" #include "libtorrent/bt_peer_connection.hpp" #include "libtorrent/network_thread_pool.hpp" #include "libtorrent/error.hpp" #include "libtorrent/alloca.hpp" #include "libtorrent/disk_interface.hpp" #include "libtorrent/bandwidth_manager.hpp" #include "libtorrent/request_blocks.hpp" // for request_a_block #include "libtorrent/performance_counters.hpp" // for counters #include "libtorrent/alert_manager.hpp" // for alert_manageralert_manager #ifdef TORRENT_DEBUG #include #endif #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_LOGGING || defined TORRENT_ERROR_LOGGING #include "libtorrent/escape_string.hpp" #include "libtorrent/socket_io.hpp" #endif //#define TORRENT_CORRUPT_DATA using boost::shared_ptr; namespace libtorrent { enum { // the limits of the download queue size min_request_queue = 2, }; int round_up8(int v) { return ((v & 7) == 0) ? v : v + (8 - (v & 7)); } #if defined TORRENT_REQUEST_LOGGING void write_request_log(FILE* f, sha1_hash const& ih , peer_connection* p, peer_request const& r) { // the event format in the log is: // uint64_t timestamp (microseconds) // uint64_t info-hash prefix // uint32_t peer identifier // uint32_t piece // uint32_t start offset // uint32_t length char event[32]; char* ptr = event; detail::write_uint64(time_now_hires().time_since_epoch().count(), ptr); memcpy(ptr, &ih[0], 8); ptr += 8; detail::write_uint32(uintptr_t(p) & 0xffffffff, ptr); detail::write_uint32(r.piece, ptr); detail::write_uint32(r.start, ptr); detail::write_uint32(r.length, ptr); int ret = fwrite(event, 1, sizeof(event), f); if (ret != sizeof(event)) { fprintf(stderr, "ERROR writing to request log: (%d) %s\n" , errno, strerror(errno)); } } #endif #if TORRENT_USE_ASSERTS bool peer_connection::is_single_thread() const { boost::shared_ptr t = m_torrent.lock(); if (!t) return true; return t->is_single_thread(); } #endif // outbound connection peer_connection::peer_connection(peer_connection_args const& pack) : peer_connection_hot_members(pack.tor, *pack.ses, *pack.sett) , m_socket(pack.s) , m_peer_info(pack.peerinfo) , m_counters(*pack.stats_counters) , m_num_pieces(0) , m_recv_start(0) , m_max_out_request_queue(m_settings.get_int(settings_pack::max_out_request_queue)) , m_remote(pack.endp) , m_disk_thread(*pack.disk_thread) , m_allocator(*pack.allocator) , m_ios(*pack.ios) , m_work(m_ios) , m_last_piece(time_now()) , m_last_request(time_now()) , m_last_incoming_request(min_time()) , m_last_unchoke(time_now()) , m_last_unchoked(time_now()) , m_last_choke(min_time()) , m_last_receive(time_now()) , m_last_sent(time_now()) , m_requested(min_time()) , m_remote_dl_update(time_now()) , m_connect(time_now()) , m_became_uninterested(time_now()) , m_became_uninteresting(time_now()) , m_downloaded_at_last_round(0) , m_uploaded_at_last_round(0) , m_uploaded_at_last_unchoke(0) , m_soft_packet_size(0) , m_outstanding_bytes(0) , m_disk_recv_buffer(*pack.allocator, 0) , m_last_seen_complete(0) , m_receiving_block(piece_block::invalid) , m_timeout_extend(0) , m_extension_outstanding_bytes(0) , m_queued_time_critical(0) , m_recv_end(0) , m_disk_recv_buffer_size(0) , m_reading_bytes(0) , m_picker_options(0) , m_num_invalid_requests(0) , m_remote_pieces_dled(0) , m_remote_dl_rate(0) , m_outstanding_writing_bytes(0) , m_download_rate_peak(0) , m_upload_rate_peak(0) , m_desired_queue_size(2) , m_speed(slow) , m_prefer_whole_pieces(0) , m_disk_read_failures(0) , m_outstanding_piece_verification(0) , m_outgoing(!pack.tor.expired()) , m_received_listen_port(false) , m_fast_reconnect(false) , m_failed(false) , m_connected(pack.tor.expired()) , m_request_large_blocks(false) , m_share_mode(false) , m_upload_only(false) , m_bitfield_received(false) , m_no_download(false) , m_sent_suggests(false) , m_holepunch_mode(false) , m_peer_choked(true) , m_have_all(false) , m_peer_interested(false) , m_need_interest_update(false) , m_has_metadata(true) , m_exceeded_limit(false) #if TORRENT_USE_ASSERTS , m_in_constructor(true) , m_disconnect_started(false) , m_initialized(false) , m_in_use(1337) , m_received_in_piece(0) , m_destructed(false) , m_socket_is_writing(false) #endif { m_counters.inc_stats_counter(counters::num_tcp_peers + m_socket->type() - 1); if (m_connected) m_counters.inc_stats_counter(counters::num_peers_connected); else if (m_connecting) m_counters.inc_stats_counter(counters::num_peers_half_open); m_superseed_piece[0] = -1; m_superseed_piece[1] = -1; boost::shared_ptr t = m_torrent.lock(); // if t is NULL, we better not be connecting, since // we can't decrement the connecting counter TORRENT_ASSERT(t || !m_connecting); if (m_connecting && t) t->inc_num_connecting(); m_est_reciprocation_rate = m_settings.get_int(settings_pack::default_est_reciprocation_rate); m_channel_state[upload_channel] = peer_info::bw_idle; m_channel_state[download_channel] = peer_info::bw_idle; m_quota[0] = 0; m_quota[1] = 0; TORRENT_ASSERT(pack.peerinfo == 0 || pack.peerinfo->banned == false); #ifndef TORRENT_DISABLE_RESOLVE_COUNTRIES std::fill(m_country, m_country + 2, 0); #endif #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING error_code ec; TORRENT_ASSERT(m_socket->remote_endpoint(ec) == m_remote || ec); tcp::endpoint local_ep = m_socket->local_endpoint(ec); std::string log_name = "[" + local_ep.address().to_string(ec) + "#" + to_string(local_ep.port()).elems + "]-" "[" + m_remote.address().to_string(ec) + "#" + to_string(m_remote.port()).elems + "]"; if (t) log_name = combine_path(to_hex(t->info_hash().to_string()) , log_name); m_logger = m_ses.create_log(log_name, m_ses.listen_port()); peer_log("%s [ ep: %s type: %s seed: %d p: %p local: %s]" , m_outgoing ? ">>> OUTGOING_CONNECTION" : "<<< INCOMING CONNECTION" , print_endpoint(m_remote).c_str() , m_socket->type_name() , m_peer_info ? m_peer_info->seed : 0, m_peer_info , print_endpoint(local_ep).c_str()); #endif #ifdef TORRENT_DEBUG piece_failed = false; #endif std::fill(m_peer_id.begin(), m_peer_id.end(), 0); } int peer_connection::timeout() const { TORRENT_ASSERT(is_single_thread()); int ret = m_settings.get_int(settings_pack::peer_timeout); #if TORRENT_USE_I2P if (m_peer_info && m_peer_info->is_i2p_addr) { // quadruple the timeout for i2p peers ret *= 4; } #endif return ret; } void peer_connection::increase_est_reciprocation_rate() { TORRENT_ASSERT(is_single_thread()); m_est_reciprocation_rate += m_est_reciprocation_rate * m_settings.get_int(settings_pack::increase_est_reciprocation_rate) / 100; } void peer_connection::decrease_est_reciprocation_rate() { TORRENT_ASSERT(is_single_thread()); m_est_reciprocation_rate -= m_est_reciprocation_rate * m_settings.get_int(settings_pack::decrease_est_reciprocation_rate) / 100; } int peer_connection::get_priority(int channel) const { TORRENT_ASSERT(is_single_thread()); TORRENT_ASSERT(channel >= 0 && channel < 2); int prio = 1; for (int i = 0; i < num_classes(); ++i) { int class_prio = m_ses.peer_classes().at(class_at(i))->priority[channel]; if (prio < class_prio) prio = class_prio; } boost::shared_ptr t = associated_torrent().lock(); if (t) { for (int i = 0; i < t->num_classes(); ++i) { int class_prio = m_ses.peer_classes().at(t->class_at(i))->priority[channel]; if (prio < class_prio) prio = class_prio; } } return prio; } void peer_connection::reset_choke_counters() { TORRENT_ASSERT(is_single_thread()); m_downloaded_at_last_round= m_statistics.total_payload_download(); m_uploaded_at_last_round = m_statistics.total_payload_upload(); } void peer_connection::start() { TORRENT_ASSERT(is_single_thread()); TORRENT_ASSERT(m_peer_info == 0 || m_peer_info->connection == this); boost::shared_ptr t = m_torrent.lock(); if (!m_outgoing) { tcp::socket::non_blocking_io ioc(true); error_code ec; m_socket->io_control(ioc, ec); if (ec) { disconnect(ec, op_iocontrol); return; } m_remote = m_socket->remote_endpoint(ec); if (ec) { disconnect(ec, op_getpeername); return; } m_local = m_socket->local_endpoint(ec); if (ec) { disconnect(ec, op_getname); return; } if (m_remote.address().is_v4() && m_settings.get_int(settings_pack::peer_tos) != 0) { m_socket->set_option(type_of_service(m_settings.get_int(settings_pack::peer_tos)), ec); #if defined TORRENT_VERBOSE_LOGGING peer_log(">>> SET_TOS[ tos: %d e: %s ]", m_settings.get_int(settings_pack::peer_tos), ec.message().c_str()); #endif } #if TORRENT_USE_IPV6 && defined IPV6_TCLASS else if (m_remote.address().is_v6() && m_settings.get_int(settings_pack::peer_tos) != 0) { m_socket->set_option(traffic_class(m_settings.get_int(settings_pack::peer_tos)), ec); } #endif } #if defined TORRENT_VERBOSE_LOGGING peer_log("*** SET_PEER_CLASS [ a: %s ]", print_address(m_remote.address()).c_str()); #endif m_ses.set_peer_classes(this, m_remote.address(), m_socket->type()); #if defined TORRENT_VERBOSE_LOGGING for (int i = 0; i < num_classes(); ++i) { peer_log("*** CLASS [ %s ]", m_ses.peer_classes().at(class_at(i))->label.c_str()); } #endif if (t && t->ready_for_connections()) { init(); } // if this is an incoming connection, we're done here if (!m_connecting) return; #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log(">>> OPEN [ protocol: %s ]", (m_remote.address().is_v4()?"IPv4":"IPv6")); #endif error_code ec; m_socket->open(m_remote.protocol(), ec); if (ec) { disconnect(ec, op_sock_open); return; } tcp::endpoint bound_ip = m_ses.bind_outgoing_socket(*m_socket , m_remote.address(), ec); #if defined TORRENT_VERBOSE_LOGGING peer_log(">>> BIND [ dst: %s ec: %s ]", print_endpoint(bound_ip).c_str() , ec.message().c_str()); #endif if (ec) { disconnect(ec, op_sock_bind); return; } #if defined TORRENT_VERBOSE_LOGGING peer_log(">>> ASYNC_CONNECT [ dst: %s ]", print_endpoint(m_remote).c_str()); #endif #if defined TORRENT_ASIO_DEBUGGING add_outstanding_async("peer_connection::on_connection_complete"); #endif #if defined TORRENT_LOGGING || defined TORRENT_ERROR_LOGGING if (t) t->debug_log("START connect [%p] (%d)", this, int(t->num_peers())); #endif m_socket->async_connect(m_remote , boost::bind(&peer_connection::on_connection_complete, self(), _1)); m_connect = time_now_hires(); sent_syn(m_remote.address().is_v6()); if (t && t->alerts().should_post()) { t->alerts().post_alert(peer_connect_alert( t->get_handle(), remote(), pid(), m_socket->type())); } #if defined TORRENT_VERBOSE_LOGGING peer_log("*** LOCAL ENDPOINT[ e: %s ]", print_endpoint(m_socket->local_endpoint(ec)).c_str()); #endif } void peer_connection::update_interest() { TORRENT_ASSERT(is_single_thread()); if (!m_need_interest_update) { // we're the first to request an interest update // post a message in order to delay it enough for // any potential other messages already in the queue // to not trigger another one. This effectively defer // the update until the current message queue is // flushed m_ios.post(boost::bind(&peer_connection::do_update_interest, self())); } m_need_interest_update = true; } void peer_connection::do_update_interest() { TORRENT_ASSERT(is_single_thread()); TORRENT_ASSERT(m_need_interest_update); m_need_interest_update = false; boost::shared_ptr t = m_torrent.lock(); if (!t) return; // if m_have_piece is 0, it means the connections // have not been initialized yet. The interested // flag will be updated once they are. if (m_have_piece.size() == 0) { #if defined TORRENT_VERBOSE_LOGGING peer_log("*** UPDATE_INTEREST [ connections not initialized ]"); #endif return; } if (!t->ready_for_connections()) { #if defined TORRENT_VERBOSE_LOGGING peer_log("*** UPDATE_INTEREST [ not ready for connections ]"); #endif return; } bool interested = false; if (!t->is_upload_only()) { t->need_picker(); piece_picker const& p = t->picker(); int num_pieces = p.num_pieces(); for (int j = 0; j != num_pieces; ++j) { if (m_have_piece[j] && t->piece_priority(j) > 0 && !p.has_piece_passed(j)) { interested = true; #if defined TORRENT_VERBOSE_LOGGING peer_log("*** UPDATE_INTEREST [ interesting, piece: %d ]", j); #endif break; } } } #if defined TORRENT_VERBOSE_LOGGING if (!interested) { peer_log("*** UPDATE_INTEREST [ not interesting ]"); } #endif if (!interested) send_not_interested(); else t->peer_is_interesting(*this); TORRENT_ASSERT(in_handshake() || is_interesting() == interested); } #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING void peer_connection::peer_log(char const* fmt, ...) const { TORRENT_ASSERT(is_single_thread()); if (!m_logger) return; va_list v; va_start(v, fmt); char usr[400]; vsnprintf(usr, sizeof(usr), fmt, v); va_end(v); char buf[450]; snprintf(buf, sizeof(buf), "%s: %s\n", time_now_string(), usr); (*m_logger) << buf; } #endif #ifndef TORRENT_DISABLE_EXTENSIONS void peer_connection::add_extension(boost::shared_ptr ext) { TORRENT_ASSERT(is_single_thread()); m_extensions.push_back(ext); } peer_plugin const* peer_connection::find_plugin(char const* type) { TORRENT_ASSERT(is_single_thread()); for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if (strcmp((*i)->type(), type) == 0) return (*i).get(); } return 0; } #endif void peer_connection::send_allowed_set() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); if (t->super_seeding()) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** SKIPPING ALLOWED SET BECAUSE OF SUPER SEEDING"); #endif return; } if (upload_only()) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** SKIPPING ALLOWED SET BECAUSE PEER IS UPLOAD ONLY"); #endif return; } int num_allowed_pieces = m_settings.get_int(settings_pack::allowed_fast_set_size); if (num_allowed_pieces == 0) return; int num_pieces = t->torrent_file().num_pieces(); if (num_allowed_pieces >= num_pieces) { // this is a special case where we have more allowed // fast pieces than pieces in the torrent. Just send // an allowed fast message for every single piece for (int i = 0; i < num_pieces; ++i) { // there's no point in offering fast pieces // that the peer already has if (has_piece(i)) continue; #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> ALLOWED_FAST [ %d ]", i); #endif write_allow_fast(i); TORRENT_ASSERT(std::find(m_accept_fast.begin() , m_accept_fast.end(), i) == m_accept_fast.end()); if (m_accept_fast.empty()) { m_accept_fast.reserve(10); m_accept_fast_piece_cnt.reserve(10); } m_accept_fast.push_back(i); m_accept_fast_piece_cnt.push_back(0); } return; } std::string x; address const& addr = m_remote.address(); if (addr.is_v4()) { address_v4::bytes_type bytes = addr.to_v4().to_bytes(); x.assign((char*)&bytes[0], bytes.size()); } #if TORRENT_USE_IPV6 else { address_v6::bytes_type bytes = addr.to_v6().to_bytes(); x.assign((char*)&bytes[0], bytes.size()); } #endif x.append((char*)&t->torrent_file().info_hash()[0], 20); sha1_hash hash = hasher(x.c_str(), x.size()).final(); for (;;) { char* p = (char*)&hash[0]; for (int i = 0; i < 5; ++i) { int piece = detail::read_uint32(p) % num_pieces; if (std::find(m_accept_fast.begin(), m_accept_fast.end(), piece) == m_accept_fast.end()) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> ALLOWED_FAST [ %d ]", piece); #endif write_allow_fast(piece); if (m_accept_fast.empty()) { m_accept_fast.reserve(10); m_accept_fast_piece_cnt.reserve(10); } m_accept_fast.push_back(piece); m_accept_fast_piece_cnt.push_back(0); if (int(m_accept_fast.size()) >= num_allowed_pieces || int(m_accept_fast.size()) == num_pieces) return; } } hash = hasher((char*)&hash[0], 20).final(); } } void peer_connection::on_metadata_impl() { TORRENT_ASSERT(is_single_thread()); boost::shared_ptr t = associated_torrent().lock(); m_have_piece.resize(t->torrent_file().num_pieces(), m_have_all); m_num_pieces = m_have_piece.count(); // now that we know how many pieces there are // remove any invalid allowed_fast and suggest pieces // now that we know what the number of pieces are for (std::vector::iterator i = m_allowed_fast.begin(); i != m_allowed_fast.end();) { if (*i < m_num_pieces) { ++i; continue; } i = m_allowed_fast.erase(i); } for (std::vector::iterator i = m_suggested_pieces.begin(); i != m_suggested_pieces.end();) { if (*i < m_num_pieces) { ++i; continue; } i = m_suggested_pieces.erase(i); } on_metadata(); if (m_disconnecting) return; } void peer_connection::init() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); TORRENT_ASSERT(t->valid_metadata()); TORRENT_ASSERT(t->ready_for_connections()); m_have_piece.resize(t->torrent_file().num_pieces(), m_have_all); if (m_have_all) m_num_pieces = t->torrent_file().num_pieces(); #if TORRENT_USE_ASSERTS TORRENT_ASSERT(!m_initialized); m_initialized = true; #endif // now that we have a piece_picker, // update it with this peer's pieces TORRENT_ASSERT(m_num_pieces == m_have_piece.count()); if (m_num_pieces == int(m_have_piece.size())) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** THIS IS A SEED [ p: %p ]", m_peer_info); #endif TORRENT_ASSERT(m_have_piece.all_set()); TORRENT_ASSERT(m_have_piece.count() == m_have_piece.size()); TORRENT_ASSERT(m_have_piece.size() == t->torrent_file().num_pieces()); // if this is a web seed. we don't have a peer_info struct t->set_seed(m_peer_info, true); m_upload_only = true; t->peer_has_all(this); #if TORRENT_USE_INVARIANT_CHECKS if (t && t->has_picker()) t->picker().check_peer_invariant(m_have_piece, this); #endif if (t->is_upload_only()) send_not_interested(); else t->peer_is_interesting(*this); return; } // if we're a seed, we don't keep track of piece availability if (t->has_picker()) { t->peer_has(m_have_piece, this); bool interesting = false; for (int i = 0; i < int(m_have_piece.size()); ++i) { if (m_have_piece[i]) { // if the peer has a piece and we don't, the peer is interesting if (!t->have_piece(i) && t->picker().piece_priority(i) != 0) interesting = true; } } if (interesting) t->peer_is_interesting(*this); else send_not_interested(); } else { update_interest(); } } peer_connection::~peer_connection() { m_counters.inc_stats_counter(counters::num_tcp_peers + m_socket->type() - 1, -1); // INVARIANT_CHECK; TORRENT_ASSERT(!m_in_constructor); TORRENT_ASSERT(m_disconnecting); TORRENT_ASSERT(m_disconnect_started); TORRENT_ASSERT(!m_destructed); #if TORRENT_USE_ASSERTS m_destructed = true; #endif #if TORRENT_USE_ASSERTS m_in_use = 0; #endif // decrement the stats counter set_endgame(false); if (m_interesting) m_counters.inc_stats_counter(counters::num_peers_down_interested, -1); if (m_peer_interested) m_counters.inc_stats_counter(counters::num_peers_up_interested, -1); if (!m_choked) { m_counters.inc_stats_counter(counters::num_peers_up_unchoked_all, -1); if (!ignore_unchoke_slots()) m_counters.inc_stats_counter(counters::num_peers_up_unchoked, -1); } if (!m_peer_choked) m_counters.inc_stats_counter(counters::num_peers_down_unchoked, -1); if (m_connected) m_counters.inc_stats_counter(counters::num_peers_connected, -1); m_connected = false; if (!m_download_queue.empty()) m_counters.inc_stats_counter(counters::num_peers_down_requests, -1); // defensive boost::shared_ptr t = m_torrent.lock(); // if t is NULL, we better not be connecting, since // we can't decrement the connecting counter TORRENT_ASSERT(t || !m_connecting); // we should really have dealt with this already TORRENT_ASSERT(!m_connecting); if (m_connecting) { m_counters.inc_stats_counter(counters::num_peers_half_open, -1); if (t) t->dec_num_connecting(); m_connecting = false; } m_disk_recv_buffer_size = 0; #ifndef TORRENT_DISABLE_EXTENSIONS m_extensions.clear(); #endif #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** CONNECTION CLOSED"); #endif TORRENT_ASSERT(m_request_queue.empty()); TORRENT_ASSERT(m_download_queue.empty()); #if TORRENT_USE_ASSERTS if (m_peer_info) TORRENT_ASSERT(m_peer_info->connection == 0); #endif } bool peer_connection::on_parole() const { return peer_info_struct() && peer_info_struct()->on_parole; } int peer_connection::picker_options() const { TORRENT_ASSERT(is_single_thread()); int ret = m_picker_options; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); if (!t) return 0; if (t->num_time_critical_pieces() > 0) { ret |= piece_picker::time_critical_mode; } if (t->is_sequential_download()) { ret |= piece_picker::sequential; } else if (t->num_have() < m_settings.get_int(settings_pack::initial_picker_threshold)) { // if we have fewer pieces than a certain threshols // don't pick rare pieces, just pick random ones, // and prioritize finishing them ret |= piece_picker::prioritize_partials; } else { ret |= piece_picker::rarest_first | piece_picker::speed_affinity; } if (m_snubbed) { // snubbed peers should request // the common pieces first, just to make // it more likely for all snubbed peers to // request blocks from the same piece ret |= piece_picker::reverse; } if (m_settings.get_bool(settings_pack::prioritize_partial_pieces)) ret |= piece_picker::prioritize_partials; if (on_parole()) ret |= piece_picker::on_parole | piece_picker::prioritize_partials; // only one of rarest_first and sequential can be set. TORRENT_ASSERT((ret & piece_picker::rarest_first) ? 1 : 0 + (ret & piece_picker::sequential) ? 1 : 0 <= 1); return ret; } void peer_connection::fast_reconnect(bool r) { TORRENT_ASSERT(is_single_thread()); if (!peer_info_struct() || peer_info_struct()->fast_reconnects > 1) return; m_fast_reconnect = r; peer_info_struct()->last_connected = (boost::uint16_t)m_ses.session_time(); int rewind = m_settings.get_int(settings_pack::min_reconnect_time) * m_settings.get_int(settings_pack::max_failcount); if (peer_info_struct()->last_connected < rewind) peer_info_struct()->last_connected = 0; else peer_info_struct()->last_connected -= rewind; if (peer_info_struct()->fast_reconnects < 15) ++peer_info_struct()->fast_reconnects; } void peer_connection::received_piece(int index) { TORRENT_ASSERT(is_single_thread()); // dont announce during handshake if (in_handshake()) return; #if defined TORRENT_VERBOSE_LOGGING peer_log("<<< RECEIVED [ piece: %d ]", index); #endif // remove suggested pieces once we have them std::vector::iterator i = std::find( m_suggested_pieces.begin(), m_suggested_pieces.end(), index); if (i != m_suggested_pieces.end()) m_suggested_pieces.erase(i); // remove allowed fast pieces i = std::find(m_allowed_fast.begin(), m_allowed_fast.end(), index); if (i != m_allowed_fast.end()) m_allowed_fast.erase(i); if (has_piece(index)) { // if we got a piece that this peer has // it might have been the last interesting // piece this peer had. We might not be // interested anymore update_interest(); if (is_disconnecting()) return; } if (disconnect_if_redundant()) return; #if TORRENT_USE_ASSERTS boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); #endif } void peer_connection::announce_piece(int index) { TORRENT_ASSERT(is_single_thread()); // dont announce during handshake if (in_handshake()) return; if (has_piece(index)) { // optimization, don't send have messages // to peers that already have the piece if (!m_settings.get_bool(settings_pack::send_redundant_have)) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> HAVE [ piece: %d ] SUPRESSED", index); #endif return; } } if (disconnect_if_redundant()) return; #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> HAVE [ piece: %d ]", index); #endif write_have(index); #if TORRENT_USE_ASSERTS boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); #endif } bool peer_connection::has_piece(int i) const { TORRENT_ASSERT(is_single_thread()); boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); TORRENT_ASSERT(t->valid_metadata()); TORRENT_ASSERT(i >= 0); TORRENT_ASSERT(i < t->torrent_file().num_pieces()); return m_have_piece[i]; } std::vector const& peer_connection::request_queue() const { TORRENT_ASSERT(is_single_thread()); return m_request_queue; } std::vector const& peer_connection::download_queue() const { TORRENT_ASSERT(is_single_thread()); return m_download_queue; } std::vector const& peer_connection::upload_queue() const { TORRENT_ASSERT(is_single_thread()); return m_requests; } time_duration peer_connection::download_queue_time(int extra_bytes) const { TORRENT_ASSERT(is_single_thread()); boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); int rate = 0; // if we haven't received any data recently, the current download rate // is not representative if (time_now() - m_last_piece > seconds(30) && m_download_rate_peak > 0) { rate = m_download_rate_peak; } else if (time_now() - m_last_unchoked < seconds(5) && m_statistics.total_payload_upload() < 2 * 0x4000) { // if we're have only been unchoked for a short period of time, // we don't know what rate we can get from this peer. Instead of assuming // the lowest possible rate, assume the average. int peers_with_requests = stats_counters()[counters::num_peers_down_requests]; // avoid division by 0 if (peers_with_requests == 0) peers_with_requests = 1; // TODO: this should be the global download rate rate = t->statistics().transfer_rate(stat::download_payload) / peers_with_requests; } else { // current download rate in bytes per seconds rate = m_statistics.transfer_rate(stat::download_payload); } // avoid division by zero if (rate < 50) rate = 50; // average of current rate and peak // rate = (rate + m_download_rate_peak) / 2; return milliseconds((m_outstanding_bytes + m_queued_time_critical * t->block_size() * 1000) / rate); } void peer_connection::add_stat(size_type downloaded, size_type uploaded) { TORRENT_ASSERT(is_single_thread()); m_statistics.add_stat(downloaded, uploaded); } void peer_connection::received_bytes(int bytes_payload, int bytes_protocol) { TORRENT_ASSERT(is_single_thread()); m_statistics.received_bytes(bytes_payload, bytes_protocol); if (m_ignore_stats) return; boost::shared_ptr t = m_torrent.lock(); if (!t) return; t->received_bytes(bytes_payload, bytes_protocol); } void peer_connection::sent_bytes(int bytes_payload, int bytes_protocol) { TORRENT_ASSERT(is_single_thread()); m_statistics.sent_bytes(bytes_payload, bytes_protocol); #ifndef TORRENT_DISABLE_EXTENSIONS if (bytes_payload) { for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { (*i)->sent_payload(bytes_payload); } } #endif if (m_ignore_stats) return; boost::shared_ptr t = m_torrent.lock(); if (!t) return; t->sent_bytes(bytes_payload, bytes_protocol); } void peer_connection::trancieve_ip_packet(int bytes, bool ipv6) { TORRENT_ASSERT(is_single_thread()); m_statistics.trancieve_ip_packet(bytes, ipv6); if (m_ignore_stats) return; boost::shared_ptr t = m_torrent.lock(); if (!t) return; t->trancieve_ip_packet(bytes, ipv6); } void peer_connection::sent_syn(bool ipv6) { TORRENT_ASSERT(is_single_thread()); m_statistics.sent_syn(ipv6); if (m_ignore_stats) return; boost::shared_ptr t = m_torrent.lock(); if (!t) return; t->sent_syn(ipv6); } void peer_connection::received_synack(bool ipv6) { TORRENT_ASSERT(is_single_thread()); m_statistics.received_synack(ipv6); if (m_ignore_stats) return; boost::shared_ptr t = m_torrent.lock(); if (!t) return; t->received_synack(ipv6); } bitfield const& peer_connection::get_bitfield() const { TORRENT_ASSERT(is_single_thread()); return m_have_piece; } void peer_connection::received_valid_data(int index) { TORRENT_ASSERT(is_single_thread()); // this fails because we haven't had time to disconnect // seeds yet, and we might have just become one // INVARIANT_CHECK; #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { TORRENT_TRY { (*i)->on_piece_pass(index); } TORRENT_CATCH(std::exception&) {} } #endif } bool peer_connection::received_invalid_data(int index, bool single_peer) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { TORRENT_TRY { (*i)->on_piece_failed(index); } TORRENT_CATCH(std::exception&) {} } #endif return true; } // verifies a piece to see if it is valid (is within a valid range) // and if it can correspond to a request generated by libtorrent. bool peer_connection::verify_piece(const peer_request& p) const { TORRENT_ASSERT(is_single_thread()); boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); TORRENT_ASSERT(t->valid_metadata()); torrent_info const& ti = t->torrent_file(); return p.piece >= 0 && p.piece < ti.num_pieces() && p.start >= 0 && p.start < ti.piece_length() && t->to_req(piece_block(p.piece, p.start / t->block_size())) == p; } void peer_connection::attach_to_torrent(sha1_hash const& ih, bool allow_encrypted) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; #if defined TORRENT_LOGGING || defined TORRENT_ERROR_LOGGING m_connect_time = time_now_hires(); #endif TORRENT_ASSERT(!m_disconnecting); TORRENT_ASSERT(m_torrent.expired()); boost::weak_ptr wpt = m_ses.find_torrent(ih); boost::shared_ptr t = wpt.lock(); #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING // now that we know which torrent this peer belongs // to. Move the log file into its directory error_code ec; std::string log_name = combine_path(to_hex(ih.to_string()) , m_remote.address().to_string(ec) + "_" + to_string(m_remote.port()).elems); m_logger->move_log_file(m_ses.get_log_path(), log_name, m_ses.listen_port()); #endif if (t && t->is_aborted()) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** the torrent has been aborted"); #endif t.reset(); } if (!t) { t = m_ses.delay_load_torrent(ih, this); #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING if (t) peer_log("*** Delay loaded torrent: %s:", to_hex(ih.to_string()).c_str()); #endif } if (!t) { // we couldn't find the torrent! #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** couldn't find a torrent with the given info_hash: %s torrents:", to_hex(ih.to_string()).c_str()); m_ses.log_all_torrents(this); #endif disconnect(errors::invalid_info_hash, op_bittorrent, 1); return; } if (t->is_paused() && m_settings.get_bool(settings_pack::incoming_starts_queued_torrents) && !t->is_aborted()) { t->resume(); } if (t->is_paused() || t->is_aborted()) { // paused torrents will not accept // incoming connections unless they are auto managed // and inconing_starts_queued_torrents is true // torrents that have errors should always reject // incoming peers #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("rejected connection to paused torrent"); #endif disconnect(errors::torrent_paused, op_bittorrent, 2); return; } #if TORRENT_USE_I2P i2p_stream* i2ps = m_socket->get(); if (!i2ps && t->torrent_file().is_i2p() && !m_settings.get_bool(settings_pack::allow_i2p_mixed)) { // the torrent is an i2p torrent, the peer is a regular peer // and we don't allow mixed mode. Disconnect the peer. #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("rejected regular connection to i2p torrent"); #endif disconnect(errors::peer_banned, op_bittorrent, 2); return; } #endif // TORRENT_USE_I2P TORRENT_ASSERT(m_torrent.expired()); // check to make sure we don't have another connection with the same // info_hash and peer_id. If we do. close this connection. t->attach_peer(this); if (m_disconnecting) return; // it's important to assign the torrent after successfully attaching. // if the peer disconnects while attaching, it's not a proper member // of the torrent and peer_connection::disconnect() will fail if it // think it is m_torrent = t; if (m_exceeded_limit) { // find a peer in some torrent (presumably the one with most peers) // and disconnect the lowest ranking peer boost::weak_ptr torr = m_ses.find_disconnect_candidate_torrent(); boost::shared_ptr other_t = torr.lock(); if (other_t) { if (other_t->num_peers() <= t->num_peers()) { disconnect(errors::too_many_connections, op_bittorrent); return; } // find the lowest ranking peer and disconnect that peer_connection* p = other_t->find_lowest_ranking_peer(); p->disconnect(errors::too_many_connections, op_bittorrent); peer_disconnected_other(); } else { disconnect(errors::too_many_connections, op_bittorrent); return; } } TORRENT_ASSERT(!m_torrent.expired()); // if the torrent isn't ready to accept // connections yet, we'll have to wait with // our initialization if (t->ready_for_connections()) init(); TORRENT_ASSERT(!m_torrent.expired()); // assume the other end has no pieces // if we don't have valid metadata yet, // leave the vector unallocated TORRENT_ASSERT(m_num_pieces == 0); m_have_piece.clear_all(); TORRENT_ASSERT(!m_torrent.expired()); } boost::uint32_t peer_connection::peer_rank() const { TORRENT_ASSERT(is_single_thread()); return m_peer_info == NULL ? 0 : m_peer_info->rank(m_ses.external_address(), m_ses.listen_port()); } // message handlers // ----------------------------- // --------- KEEPALIVE --------- // ----------------------------- void peer_connection::incoming_keepalive() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== KEEPALIVE"); #endif } // ----------------------------- // ----------- CHOKE ----------- // ----------------------------- void peer_connection::set_endgame(bool b) { TORRENT_ASSERT(is_single_thread()); if (m_endgame_mode == b) return; m_endgame_mode = b; if (m_endgame_mode) m_counters.inc_stats_counter(counters::num_peers_end_game); else m_counters.inc_stats_counter(counters::num_peers_end_game, -1); } void peer_connection::incoming_choke() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_choke()) return; } #endif if (is_disconnecting()) return; #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== CHOKE"); #endif if (m_peer_choked == false) m_counters.inc_stats_counter(counters::num_peers_down_unchoked, -1); m_peer_choked = true; set_endgame(false); clear_request_queue(); } void peer_connection::clear_request_queue() { TORRENT_ASSERT(is_single_thread()); boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); if (!t->has_picker()) { m_request_queue.clear(); return; } // clear the requests that haven't been sent yet if (peer_info_struct() == 0 || !peer_info_struct()->on_parole) { // if the peer is not in parole mode, clear the queued // up block requests piece_picker& p = t->picker(); for (std::vector::const_iterator i = m_request_queue.begin() , end(m_request_queue.end()); i != end; ++i) { p.abort_download(i->block, peer_info_struct()); } m_request_queue.clear(); m_queued_time_critical = 0; } } bool match_request(peer_request const& r, piece_block const& b, int block_size) { if (int(b.piece_index) != r.piece) return false; if (int(b.block_index) != r.start / block_size) return false; if (r.start % block_size != 0) return false; return true; } // ----------------------------- // -------- REJECT PIECE ------- // ----------------------------- void peer_connection::incoming_reject_request(peer_request const& r) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== REJECT_PIECE [ piece: %d | s: %x | l: %x ]" , r.piece, r.start, r.length); #endif #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_reject(r)) return; } #endif if (is_disconnecting()) return; std::vector::iterator i = std::find_if( m_download_queue.begin(), m_download_queue.end() , boost::bind(match_request, boost::cref(r), boost::bind(&pending_block::block, _1) , t->block_size())); if (i != m_download_queue.end()) { pending_block b = *i; bool remove_from_picker = !i->timed_out && !i->not_wanted; m_download_queue.erase(i); TORRENT_ASSERT(m_outstanding_bytes >= r.length); m_outstanding_bytes -= r.length; if (m_outstanding_bytes < 0) m_outstanding_bytes = 0; if (m_download_queue.empty()) m_counters.inc_stats_counter(counters::num_peers_down_requests, -1); // if the peer is in parole mode, keep the request if (peer_info_struct() && peer_info_struct()->on_parole) { // we should only add it if the block is marked as // busy in the piece-picker if (remove_from_picker) m_request_queue.insert(m_request_queue.begin(), b); } else if (!t->is_seed() && remove_from_picker) { piece_picker& p = t->picker(); p.abort_download(b.block, peer_info_struct()); } #if TORRENT_USE_INVARIANT_CHECKS check_invariant(); #endif } #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING else { peer_log("*** PIECE NOT IN REQUEST QUEUE"); } #endif if (has_peer_choked()) { // if we're choked and we got a rejection of // a piece in the allowed fast set, remove it // from the allow fast set. std::vector::iterator i = std::find( m_allowed_fast.begin(), m_allowed_fast.end(), r.piece); if (i != m_allowed_fast.end()) m_allowed_fast.erase(i); } else { std::vector::iterator i = std::find(m_suggested_pieces.begin() , m_suggested_pieces.end(), r.piece); if (i != m_suggested_pieces.end()) m_suggested_pieces.erase(i); } if (m_request_queue.empty() && m_download_queue.size() < 2) { if (request_a_block(*t, *this)) m_counters.inc_stats_counter(counters::reject_piece_picks); send_block_requests(); } } // ----------------------------- // ------- SUGGEST PIECE ------- // ----------------------------- void peer_connection::incoming_suggest(int index) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== SUGGEST_PIECE [ piece: %d ]", index); #endif boost::shared_ptr t = m_torrent.lock(); if (!t) return; #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_suggest(index)) return; } #endif if (is_disconnecting()) return; if (index < 0) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("<== INVALID_SUGGEST_PIECE [ %d ]", index); #endif return; } if (t->valid_metadata()) { if (index >= int(m_have_piece.size())) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("<== INVALID_SUGGEST [ %d | s: %d ]" , index, int(m_have_piece.size())); #endif return; } // if we already have the piece, we can // ignore this message if (t->have_piece(index)) return; } if (int(m_suggested_pieces.size()) > m_settings.get_int(settings_pack::max_suggest_pieces)) m_suggested_pieces.erase(m_suggested_pieces.begin()); m_suggested_pieces.push_back(index); #ifdef TORRENT_VERBOSE_LOGGING peer_log("** SUGGEST_PIECE [ piece: %d added to set: %d ]", index, int(m_suggested_pieces.size())); #endif } // ----------------------------- // ---------- UNCHOKE ---------- // ----------------------------- void peer_connection::incoming_unchoke() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); #if defined TORRENT_LOGGING || defined TORRENT_ERROR_LOGGING m_unchoke_time = time_now_hires(); t->debug_log("UNCHOKE [%p] (%d ms)", this, int(total_milliseconds(m_unchoke_time - m_bitfield_time))); #endif #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_unchoke()) return; } #endif #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== UNCHOKE"); #endif if (m_peer_choked) m_counters.inc_stats_counter(counters::num_peers_down_unchoked); m_peer_choked = false; m_last_unchoked = time_now(); if (is_disconnecting()) return; if (is_interesting()) { if (request_a_block(*t, *this)) m_counters.inc_stats_counter(counters::unchoke_piece_picks); send_block_requests(); } } // ----------------------------- // -------- INTERESTED --------- // ----------------------------- void peer_connection::incoming_interested() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_interested()) return; } #endif #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== INTERESTED"); #endif if (m_peer_interested == false) m_counters.inc_stats_counter(counters::num_peers_up_interested); m_peer_interested = true; if (is_disconnecting()) return; // if the peer is ready to download stuff, it must have metadata m_has_metadata = true; disconnect_if_redundant(); if (is_disconnecting()) return; if (t->graceful_pause()) { #if defined TORRENT_VERBOSE_LOGGING peer_log("DID NOT UNCHOKE [ graceful pause mode ]"); #endif return; } if (!is_choked()) { // the reason to send an extra unchoke message here is that // because of the handshake-round-trip optimization, we may // end up sending an unchoke before the other end sends us // an interested message. This may confuse clients, not reacting // to the first unchoke, and then not check whether it's unchoked // when sending the interested message. If the other end's client // has this problem, sending another unchoke here will kick it // to react to the fact that it's unchoked. #ifdef TORRENT_VERBOSE_LOGGING peer_log("SENDING REDUNDANT UNCHOKE"); #endif write_unchoke(); return; } maybe_unchoke_this_peer(); } void peer_connection::maybe_unchoke_this_peer() { TORRENT_ASSERT(is_single_thread()); if (ignore_unchoke_slots()) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("ABOUT TO UNCHOKE [ peer ignores unchoke slots ]"); #endif // if this peer is expempted from the choker // just unchoke it immediately send_unchoke(); } else if (m_ses.preemptive_unchoke()) { // if the peer is choked and we have upload slots left, // then unchoke it. Another condition that has to be met // is that the torrent doesn't keep track of the individual // up/down ratio for each peer (ratio == 0) or (if it does // keep track) this particular connection isn't a leecher. // If the peer was choked because it was leeching, don't // unchoke it again. // The exception to this last condition is if we're a seed. // In that case we don't care if people are leeching, they // can't pay for their downloads anyway. boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); t->unchoke_peer(*this); } #if defined TORRENT_VERBOSE_LOGGING else { peer_log("DID NOT UNCHOKE [ the number of uploads (%d) " "is more than or equal to the limit (%d) ]" , m_ses.num_uploads(), m_settings.get_int(settings_pack::unchoke_slots_limit)); } #endif } // ----------------------------- // ------ NOT INTERESTED ------- // ----------------------------- void peer_connection::incoming_not_interested() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_not_interested()) return; } #endif m_became_uninterested = time_now(); #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== NOT_INTERESTED"); #endif if (m_peer_interested) m_counters.inc_stats_counter(counters::num_peers_up_interested, -1); m_peer_interested = false; if (is_disconnecting()) return; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); choke_this_peer(); } void peer_connection::choke_this_peer() { TORRENT_ASSERT(is_single_thread()); if (is_choked()) return; if (ignore_unchoke_slots()) { send_choke(); return; } boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); if (m_peer_info && m_peer_info->optimistically_unchoked) { m_peer_info->optimistically_unchoked = false; m_counters.inc_stats_counter(counters::num_peers_up_unchoked_optimistic, -1); t->trigger_optimistic_unchoke(); } t->choke_peer(*this); t->trigger_unchoke(); } // ----------------------------- // ----------- HAVE ------------ // ----------------------------- void peer_connection::incoming_have(int index) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_have(index)) return; } #endif if (is_disconnecting()) return; // if we haven't received a bitfield, it was // probably omitted, which is the same as 'have_none' if (!m_bitfield_received) incoming_have_none(); #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== HAVE [ piece: %d ]", index); #endif if (is_disconnecting()) return; if (!t->valid_metadata() && index >= int(m_have_piece.size())) { if (index < 131072) { // if we don't have metadata // and we might not have received a bitfield // extend the bitmask to fit the new // have message m_have_piece.resize(index + 1, false); } else { // unless the index > 64k, in which case // we just ignore it return; } } // if we got an invalid message, abort if (index >= int(m_have_piece.size()) || index < 0) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** ERROR: [ have-metadata have_piece.size: %d ]", index, int(m_have_piece.size())); #endif disconnect(errors::invalid_have, op_bittorrent, 2); return; } if (t->super_seeding() && !m_settings.get_bool(settings_pack::strict_super_seeding)) { // if we're superseeding and the peer just told // us that it completed the piece we're superseeding // to it, change the superseeding piece for this peer // if the peer optimizes out redundant have messages // this will be handled when the peer sends not-interested // instead. if (super_seeded_piece(index)) { superseed_piece(index, t->get_piece_to_super_seed(m_have_piece)); } } if (m_have_piece[index]) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log(" got redundant HAVE message for index: %d", index); #endif return; } m_have_piece.set_bit(index); ++m_num_pieces; // if the peer is downloading stuff, it must have metadata m_has_metadata = true; // only update the piece_picker if // we have the metadata and if // we're not a seed (in which case // we won't have a piece picker) if (!t->valid_metadata()) return; t->peer_has(index, this); // this will disregard all have messages we get within // the first two seconds. Since some clients implements // lazy bitfields, these will not be reliable to use // for an estimated peer download rate. if (!peer_info_struct() || m_ses.session_time() - peer_info_struct()->last_connected > 2) { // update bytes downloaded since last timer ++m_remote_pieces_dled; } // it's important to not disconnect before we have // updated the piece picker, otherwise we will incorrectly // decrement the piece count without first incrementing it if (is_seed()) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** THIS IS A SEED [ p: %p ]", m_peer_info); #endif TORRENT_ASSERT(m_have_piece.all_set()); TORRENT_ASSERT(m_have_piece.count() == m_have_piece.size()); TORRENT_ASSERT(m_have_piece.size() == t->torrent_file().num_pieces()); t->seen_complete(); t->set_seed(m_peer_info, true); m_upload_only = true; #if TORRENT_USE_INVARIANT_CHECKS if (t && t->has_picker()) t->picker().check_peer_invariant(m_have_piece, this); #endif disconnect_if_redundant(); if (is_disconnecting()) return; } if (!t->has_piece_passed(index) && !t->is_seed() && !is_interesting() && (!t->has_picker() || t->picker().piece_priority(index) != 0)) t->peer_is_interesting(*this); // if we're super seeding, this might mean that somebody // forwarded this piece. In which case we need to give // a new piece to that peer if (t->super_seeding() && m_settings.get_bool(settings_pack::strict_super_seeding) && (!super_seeded_piece(index) || t->num_peers() == 1)) { for (torrent::peer_iterator i = t->begin() , end(t->end()); i != end; ++i) { peer_connection* p = *i; if (!p->super_seeded_piece(index)) continue; if (!p->has_piece(index)) continue; p->superseed_piece(index, t->get_piece_to_super_seed(p->get_bitfield())); } } } // ----------------------------- // -------- DONT HAVE ---------- // ----------------------------- void peer_connection::incoming_dont_have(int index) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_dont_have(index)) return; } #endif if (is_disconnecting()) return; #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== DONT_HAVE [ piece: %d ]", index); #endif // if we got an invalid message, abort if (index >= int(m_have_piece.size()) || index < 0) { disconnect(errors::invalid_dont_have, op_bittorrent, 2); return; } if (!m_have_piece[index]) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log(" got redundant DONT_HAVE message for index: %d", index); #endif return; } bool was_seed = is_seed(); m_have_piece.clear_bit(index); TORRENT_ASSERT(m_num_pieces > 0); --m_num_pieces; // only update the piece_picker if // we have the metadata and if // we're not a seed (in which case // we won't have a piece picker) if (!t->valid_metadata()) return; t->peer_lost(index, this); if (was_seed) t->set_seed(m_peer_info, false); } // ----------------------------- // --------- BITFIELD ---------- // ----------------------------- void peer_connection::incoming_bitfield(bitfield const& bits) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_bitfield(bits)) return; } #endif if (is_disconnecting()) return; #ifdef TORRENT_VERBOSE_LOGGING std::string bitfield_str; bitfield_str.resize(bits.size()); for (int i = 0; i < int(bits.size()); ++i) bitfield_str[i] = bits[i] ? '1' : '0'; peer_log("<== BITFIELD [ %s ]", bitfield_str.c_str()); #endif // if we don't have the metedata, we cannot // verify the bitfield size if (t->valid_metadata() && (bits.size() + 7) / 8 != (m_have_piece.size() + 7) / 8) { disconnect(errors::invalid_bitfield_size, op_bittorrent, 2); return; } if (m_bitfield_received) { // if we've already received a bitfield message // we first need to count down all the pieces // we believe the peer has first t->peer_lost(m_have_piece, this); } m_bitfield_received = true; #if defined TORRENT_LOGGING || defined TORRENT_ERROR_LOGGING m_bitfield_time = time_now_hires(); t->debug_log("HANDSHAKE [%p] (%d ms)", this, int(total_milliseconds(m_bitfield_time - m_connect_time))); #endif // if we don't have metadata yet // just remember the bitmask // don't update the piecepicker // (since it doesn't exist yet) if (!t->ready_for_connections()) { #ifdef TORRENT_VERBOSE_LOGGING if (m_num_pieces == int(bits.size())) peer_log("*** THIS IS A SEED [ p: %p ]", m_peer_info); #endif m_have_piece = bits; m_num_pieces = bits.count(); t->set_seed(m_peer_info, m_num_pieces == int(bits.size())); #if TORRENT_USE_INVARIANT_CHECKS if (t && t->has_picker()) t->picker().check_peer_invariant(m_have_piece, this); #endif return; } TORRENT_ASSERT(t->valid_metadata()); int num_pieces = bits.count(); if (num_pieces == int(m_have_piece.size())) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** THIS IS A SEED [ p: %p ]", m_peer_info); #endif t->set_seed(m_peer_info, true); m_upload_only = true; m_have_piece.set_all(); m_num_pieces = num_pieces; t->peer_has_all(this); TORRENT_ASSERT(m_have_piece.all_set()); TORRENT_ASSERT(m_have_piece.count() == m_have_piece.size()); TORRENT_ASSERT(m_have_piece.size() == t->torrent_file().num_pieces()); #if TORRENT_USE_INVARIANT_CHECKS if (t && t->has_picker()) t->picker().check_peer_invariant(m_have_piece, this); #endif if (!t->is_upload_only()) t->peer_is_interesting(*this); disconnect_if_redundant(); return; } // let the torrent know which pieces the // peer has // if we're a seed, we don't keep track of piece availability bool interesting = false; t->peer_has(bits, this); if (!t->is_upload_only()) { for (int i = 0; i < (int)m_have_piece.size(); ++i) { bool have = bits[i]; if (!have || m_have_piece[i]) continue; // if we don't have a picker, the assumption is that the piece // priority is 1, or that we're a seed, but in that case have_piece // would have returned true. if (!t->have_piece(i) && (!t->has_picker() || t->picker().piece_priority(i) != 0)) interesting = true; } } m_have_piece = bits; m_num_pieces = num_pieces; if (interesting) t->peer_is_interesting(*this); else if (upload_only() && can_disconnect(error_code(errors::upload_upload_connection, get_libtorrent_category()))) disconnect(errors::upload_upload_connection, op_bittorrent); } bool peer_connection::disconnect_if_redundant() { TORRENT_ASSERT(is_single_thread()); if (m_disconnecting) return false; // we cannot disconnect in a constructor TORRENT_ASSERT(m_in_constructor == false); if (!m_settings.get_bool(settings_pack::close_redundant_connections)) return false; boost::shared_ptr t = m_torrent.lock(); if (!t) return false; // if we don't have the metadata yet, don't disconnect // also, if the peer doesn't have metadata we shouldn't // disconnect it, since it may want to request the // metadata from us if (!t->valid_metadata() || !has_metadata()) return false; // don't close connections in share mode, we don't know if we need them if (t->share_mode()) return false; if (m_upload_only && t->is_upload_only() && can_disconnect(error_code(errors::upload_upload_connection, get_libtorrent_category()))) { disconnect(errors::upload_upload_connection, op_bittorrent); return true; } if (m_upload_only && !m_interesting && m_bitfield_received && t->are_files_checked() && can_disconnect(error_code(errors::uninteresting_upload_peer, get_libtorrent_category()))) { disconnect(errors::uninteresting_upload_peer, op_bittorrent); return true; } return false; } bool peer_connection::can_disconnect(error_code const& ec) const { TORRENT_ASSERT(is_single_thread()); #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::const_iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if (!(*i)->can_disconnect(ec)) return false; } #endif return true; } // ----------------------------- // ---------- REQUEST ---------- // ----------------------------- void peer_connection::incoming_request(peer_request const& r) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); m_counters.inc_stats_counter(counters::piece_requests); #if defined TORRENT_VERBOSE_LOGGING peer_log("<== REQUEST [ piece: %d s: %x l: %x ]" , r.piece, r.start, r.length); #endif if (t->super_seeding() && !super_seeded_piece(r.piece)) { m_counters.inc_stats_counter(counters::invalid_piece_requests); ++m_num_invalid_requests; #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** INVALID_REQUEST [ piece not superseeded " "i: %d t: %d n: %d h: %d ss1: %d ss2: %d ]" , m_peer_interested , int(t->torrent_file().piece_size(r.piece)) , t->torrent_file().num_pieces() , t->has_piece_passed(r.piece) , m_superseed_piece[0] , m_superseed_piece[1]); #endif write_reject_request(r); if (t->alerts().should_post()) { t->alerts().post_alert(invalid_request_alert( t->get_handle(), m_remote, m_peer_id, r)); } return; } // if we haven't received a bitfield, it was // probably omitted, which is the same as 'have_none' if (!m_bitfield_received) incoming_have_none(); if (is_disconnecting()) return; #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_request(r)) return; } #endif if (is_disconnecting()) return; if (!t->valid_metadata()) { m_counters.inc_stats_counter(counters::invalid_piece_requests); // if we don't have valid metadata yet, // we shouldn't get a request #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** INVALID_REQUEST [ we don't have metadata yet ]"); peer_log("==> REJECT_PIECE [ piece: %d | s: %x | l: %x ] no metadata" , r.piece, r.start, r.length); #endif write_reject_request(r); return; } if (int(m_requests.size()) > m_settings.get_int(settings_pack::max_allowed_in_request_queue)) { m_counters.inc_stats_counter(counters::max_piece_requests); // don't allow clients to abuse our // memory consumption. // ignore requests if the client // is making too many of them. #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** INVALID_REQUEST [ incoming request queue full %d ]" , int(m_requests.size())); peer_log("==> REJECT_PIECE [ piece: %d | s: %x | l: %x ] too many requests" , r.piece, r.start, r.length); #endif write_reject_request(r); return; } int fast_idx = -1; std::vector::iterator fast_iter = std::find(m_accept_fast.begin() , m_accept_fast.end(), r.piece); if (fast_iter != m_accept_fast.end()) fast_idx = fast_iter - m_accept_fast.begin(); // make sure this request // is legal and that the peer // is not choked if (r.piece < 0 || r.piece >= t->torrent_file().num_pieces() || (!t->has_piece_passed(r.piece) && !t->is_predictive_piece(r.piece) && !t->seed_mode()) || r.start < 0 || r.start >= t->torrent_file().piece_size(r.piece) || r.length <= 0 || r.length + r.start > t->torrent_file().piece_size(r.piece) || !m_peer_interested || r.length > t->block_size()) { m_counters.inc_stats_counter(counters::invalid_piece_requests); #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** INVALID_REQUEST [ " "i: %d t: %d n: %d h: %d block_limit: %d ]" , m_peer_interested , int(t->torrent_file().piece_size(r.piece)) , t->torrent_file().num_pieces() , t->has_piece_passed(r.piece) , t->block_size()); peer_log("==> REJECT_PIECE [ piece: %d | s: %d | l: %d ] invalid request" , r.piece , r.start , r.length); #endif write_reject_request(r); ++m_num_invalid_requests; if (t->alerts().should_post()) { t->alerts().post_alert(invalid_request_alert( t->get_handle(), m_remote, m_peer_id, r)); } // every ten invalid request, remind the peer that it's choked if (!m_peer_interested && m_num_invalid_requests % 10 == 0 && m_choked) { if (m_num_invalid_requests > 300 && !m_peer_choked && can_disconnect(error_code(errors::too_many_requests_when_choked, get_libtorrent_category()))) { disconnect(errors::too_many_requests_when_choked, op_bittorrent, 2); return; } #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> CHOKE"); #endif write_choke(); } return; } // if we have choked the client // ignore the request const int blocks_per_piece = static_cast( (t->torrent_file().piece_length() + t->block_size() - 1) / t->block_size()); // disconnect peers that downloads more than foo times an allowed // fast piece if (m_choked && fast_idx != -1 && m_accept_fast_piece_cnt[fast_idx] >= 3 * blocks_per_piece && can_disconnect(error_code(errors::too_many_requests_when_choked, get_libtorrent_category()))) { disconnect(errors::too_many_requests_when_choked, op_bittorrent, 2); return; } if (m_choked && fast_idx == -1) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** REJECTING REQUEST [ peer choked and piece not in allowed fast set ]"); peer_log(" ==> REJECT_PIECE [ piece: %d | s: %d | l: %d ] peer choked" , r.piece, r.start, r.length); #endif m_counters.inc_stats_counter(counters::choked_piece_requests); write_reject_request(r); // allow peers to send request up to 2 seconds after getting choked, // then disconnect them if (time_now() - seconds(2) > m_last_choke && can_disconnect(error_code(errors::too_many_requests_when_choked, get_libtorrent_category()))) { disconnect(errors::too_many_requests_when_choked, op_bittorrent, 2); return; } } else { // increase the allowed fast set counter if (fast_idx != -1) ++m_accept_fast_piece_cnt[fast_idx]; if (m_requests.empty()) m_counters.inc_stats_counter(counters::num_peers_up_requests); m_requests.push_back(r); #ifdef TORRENT_REQUEST_LOGGING FILE* log = m_ses.get_request_log(); if (log) write_request_log(log, t->info_hash(), this, r); #endif m_last_incoming_request = time_now(); fill_send_buffer(); } } // reject all requests to this piece void peer_connection::reject_piece(int index) { TORRENT_ASSERT(is_single_thread()); for (std::vector::iterator i = m_requests.begin() , end(m_requests.end()); i != end; ++i) { peer_request const& r = *i; if (r.piece != index) continue; write_reject_request(r); i = m_requests.erase(i); if (m_requests.empty()) m_counters.inc_stats_counter(counters::num_peers_up_requests, -1); } } void peer_connection::incoming_piece_fragment(int bytes) { TORRENT_ASSERT(is_single_thread()); m_last_piece = time_now(); TORRENT_ASSERT(m_outstanding_bytes >= bytes); m_outstanding_bytes -= bytes; if (m_outstanding_bytes < 0) m_outstanding_bytes = 0; boost::shared_ptr t = associated_torrent().lock(); #if TORRENT_USE_ASSERTS TORRENT_ASSERT(m_received_in_piece + bytes <= t->block_size()); m_received_in_piece += bytes; #endif // progress of this torrent increased t->state_updated(); #if TORRENT_USE_INVARIANT_CHECKS check_invariant(); #endif } void peer_connection::start_receive_piece(peer_request const& r) { TORRENT_ASSERT(is_single_thread()); #if TORRENT_USE_INVARIANT_CHECKS check_invariant(); #endif #if TORRENT_USE_ASSERTS buffer::const_interval recv_buffer = receive_buffer(); int recv_pos = recv_buffer.end - recv_buffer.begin; TORRENT_ASSERT(recv_pos >= 9); #endif boost::shared_ptr t = associated_torrent().lock(); TORRENT_ASSERT(t); if (!verify_piece(r)) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** INVALID_PIECE [ piece: %d s: %d l: %d ]" , r.piece, r.start, r.length); #endif disconnect(errors::invalid_piece, op_bittorrent, 2); return; } piece_block b(r.piece, r.start / t->block_size()); m_receiving_block = b; bool in_req_queue = false; for (std::vector::iterator i = m_download_queue.begin() , end(m_download_queue.end()); i != end; ++i) { if (i->block != b) continue; in_req_queue = true; if (i->receiving == false) { i->receiving = true; // if send_buffer_offset is greater then or equal to 0, it means // the callback of the send operation when we sent this // request hasn't come back yet, and we're already // receiving the response from it. Count the rtt as 0. int rtt = (i->send_buffer_offset >= 0) ? 0 : int(total_milliseconds(time_now_hires() - i->request_time)); m_rtt.add_sample(rtt); #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** RTT: %d ms [%d +/- %d ms]", rtt, m_rtt.mean() , m_rtt.avg_deviation()); #endif } break; } // if this is not in the request queue, we have to // assume our outstanding bytes includes this piece too // if we're disconnecting, we shouldn't add pieces if (!in_req_queue && !m_disconnecting) { for (std::vector::iterator i = m_request_queue.begin() , end(m_request_queue.end()); i != end; ++i) { if (i->block != b) continue; in_req_queue = true; if (i - m_request_queue.begin() < m_queued_time_critical) --m_queued_time_critical; m_request_queue.erase(i); break; } if (m_download_queue.empty()) m_counters.inc_stats_counter(counters::num_peers_down_requests); m_download_queue.insert(m_download_queue.begin(), b); if (!in_req_queue) { if (t->alerts().should_post()) { t->alerts().post_alert(unwanted_block_alert(t->get_handle(), m_remote , m_peer_id, b.block_index, b.piece_index)); } #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** The block we just got was not in the request queue ***"); #endif TORRENT_ASSERT(m_download_queue.front().block == b); m_download_queue.front().not_wanted = true; } m_outstanding_bytes += r.length; } } #if TORRENT_USE_INVARIANT_CHECKS struct check_postcondition { check_postcondition(boost::shared_ptr const& t_ , bool init_check = true): t(t_) { if (init_check) check(); } ~check_postcondition() { check(); } void check() { if (!t->is_seed()) { const int blocks_per_piece = static_cast( (t->torrent_file().piece_length() + t->block_size() - 1) / t->block_size()); std::vector const& dl_queue = t->picker().get_download_queue(); for (std::vector::const_iterator i = dl_queue.begin(); i != dl_queue.end(); ++i) { TORRENT_ASSERT(i->finished <= blocks_per_piece); } } } shared_ptr t; }; #endif // ----------------------------- // ----------- PIECE ----------- // ----------------------------- void peer_connection::incoming_piece(peer_request const& p, char const* data) { TORRENT_ASSERT(is_single_thread()); bool exceeded = false; char* buffer = m_allocator.allocate_disk_buffer(exceeded, self(), "receive buffer"); if (buffer == 0) { disconnect(errors::no_memory, op_alloc_recvbuf); return; } if (exceeded) { if ((m_channel_state[download_channel] & peer_info::bw_disk) == 0) m_counters.inc_stats_counter(counters::num_peers_down_disk); m_channel_state[download_channel] |= peer_info::bw_disk; #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** exceeded disk buffer watermark"); #endif } disk_buffer_holder holder(m_allocator, buffer); std::memcpy(buffer, data, p.length); incoming_piece(p, holder); } void peer_connection::incoming_piece(peer_request const& p, disk_buffer_holder& data) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); TORRENT_ASSERT(!m_disk_recv_buffer); TORRENT_ASSERT(m_disk_recv_buffer_size == 0); // we're not receiving any block right now m_receiving_block = piece_block::invalid; #ifdef TORRENT_CORRUPT_DATA // corrupt all pieces from certain peers if (m_remote.address().is_v4() && (m_remote.address().to_v4().to_ulong() & 0xf) == 0) { data.get()[0] = ~data.get()[0]; } #endif // if we haven't received a bitfield, it was // probably omitted, which is the same as 'have_none' if (!m_bitfield_received) incoming_have_none(); if (is_disconnecting()) return; update_desired_queue_size(); #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_piece(p, data)) { #if TORRENT_USE_ASSERTS TORRENT_ASSERT(m_received_in_piece == p.length); m_received_in_piece = 0; #endif return; } } #endif if (is_disconnecting()) return; #if TORRENT_USE_INVARIANT_CHECKS check_postcondition post_checker_(t); #if defined TORRENT_EXPENSIVE_INVARIANT_CHECKS t->check_invariant(); #endif #endif #ifdef TORRENT_VERBOSE_LOGGING hasher h; h.update(data.get(), p.length); peer_log("<== PIECE [ piece: %d | s: %x | l: %x | ds: %d | qs: %d | q: %d | hash: %s ]" , p.piece, p.start, p.length, statistics().download_rate() , int(m_desired_queue_size), int(m_download_queue.size()) , to_hex(h.final().to_string()).c_str()); #endif if (p.length == 0) { if (t->alerts().should_post()) { t->alerts().post_alert(peer_error_alert(t->get_handle(), m_remote , m_peer_id, op_bittorrent, errors::peer_sent_empty_piece)); } // This is used as a reject-request by bitcomet incoming_reject_request(p); return; } // if we're already seeding, don't bother, // just ignore it if (t->is_seed()) { #if TORRENT_USE_ASSERTS TORRENT_ASSERT(m_received_in_piece == p.length); m_received_in_piece = 0; #endif if (!m_download_queue.empty()) { m_download_queue.erase(m_download_queue.begin()); if (m_download_queue.empty()) m_counters.inc_stats_counter(counters::num_peers_down_requests, -1); } t->add_redundant_bytes(p.length, torrent::piece_seed); return; } ptime now = time_now(); t->need_picker(); piece_picker& picker = t->picker(); piece_block block_finished(p.piece, p.start / t->block_size()); TORRENT_ASSERT(verify_piece(p)); std::vector::iterator b = std::find_if( m_download_queue.begin() , m_download_queue.end() , has_block(block_finished)); if (b == m_download_queue.end()) { if (t->alerts().should_post()) { t->alerts().post_alert(unwanted_block_alert(t->get_handle(), m_remote , m_peer_id, block_finished.block_index, block_finished.piece_index)); } #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** The block we just got was not in the request queue ***"); #endif #if TORRENT_USE_ASSERTS TORRENT_ASSERT_VAL(m_received_in_piece == p.length, m_received_in_piece); m_received_in_piece = 0; #endif t->add_redundant_bytes(p.length, torrent::piece_unknown); // the bytes of the piece we just completed have been deducted from // m_outstanding_bytes as we received it, in incoming_piece_fragment. // however, it now turns out the piece we received wasn't in the // download queue, so we still have the same number of pieces in the // download queue, which is why we need to add the bytes back. m_outstanding_bytes += p.length; #if TORRENT_USE_INVARIANT_CHECKS check_invariant(); #endif return; } #if TORRENT_USE_ASSERTS TORRENT_ASSERT_VAL(m_received_in_piece == p.length, m_received_in_piece); m_received_in_piece = 0; #endif // if the block we got is already finished, then ignore it if (picker.is_downloaded(block_finished)) { torrent::wasted_reason_t reason; if (b->timed_out) reason = torrent::piece_timed_out; else if (b->not_wanted) reason = torrent::piece_cancelled; else if (b->busy) reason = torrent::piece_end_game; else reason = torrent::piece_unknown; t->add_redundant_bytes(p.length, reason); m_download_queue.erase(b); if (m_download_queue.empty()) m_counters.inc_stats_counter(counters::num_peers_down_requests, -1); m_timeout_extend = 0; if (m_disconnecting) return; if (!m_download_queue.empty()) m_requested = now; if (request_a_block(*t, *this)) m_counters.inc_stats_counter(counters::incoming_redundant_piece_picks); send_block_requests(); return; } if (total_seconds(now - m_requested) < m_settings.get_int(settings_pack::request_timeout) && m_snubbed) { m_snubbed = false; if (t->alerts().should_post()) { t->alerts().post_alert(peer_unsnubbed_alert(t->get_handle() , m_remote, m_peer_id)); } } #if defined TORRENT_LOGGING || defined TORRENT_ERROR_LOGGING t->debug_log("PIECE [%p] (%d ms) (%d)", this , int(total_milliseconds(time_now_hires() - m_unchoke_time)), t->num_have()); #endif #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** FILE ASYNC WRITE [ piece: %d | s: %x | l: %x ]" , p.piece, p.start, p.length); #endif m_download_queue.erase(b); if (m_download_queue.empty()) m_counters.inc_stats_counter(counters::num_peers_down_requests, -1); if (t->is_deleted()) return; if (!t->need_loaded()) { t->add_redundant_bytes(p.length, torrent::piece_unknown); return; } t->inc_refcount("async_write"); m_disk_thread.async_write(&t->storage(), p, data , boost::bind(&peer_connection::on_disk_write_complete , self(), _1, p, t)); boost::uint64_t write_queue_size = m_counters.inc_stats_counter( counters::queued_write_bytes, p.length); m_outstanding_writing_bytes += p.length; boost::uint64_t max_queue_size = m_settings.get_int( settings_pack::max_queued_disk_bytes); if (write_queue_size > max_queue_size && write_queue_size - p.length < max_queue_size && m_settings.get_int(settings_pack::cache_size) > 5 && t->alerts().should_post()) { t->alerts().post_alert(performance_alert(t->get_handle() , performance_alert::too_high_disk_queue_limit)); } if (!m_download_queue.empty()) { m_timeout_extend = (std::max)(m_timeout_extend - m_settings.get_int(settings_pack::request_timeout), 0); m_requested += seconds(m_settings.get_int(settings_pack::request_timeout)); if (m_requested > now) m_requested = now; } else { m_timeout_extend = 0; } bool was_finished = picker.is_piece_finished(p.piece); // did we request this block from any other peers? bool multi = picker.num_peers(block_finished) > 1; // fprintf(stderr, "peer_connection mark_as_writing peer: %p piece: %d block: %d\n" // , peer_info_struct(), block_finished.piece_index, block_finished.block_index); picker.mark_as_writing(block_finished, peer_info_struct()); TORRENT_ASSERT(picker.num_peers(block_finished) == 0); // if we requested this block from other peers, cancel it now if (multi) t->cancel_block(block_finished); if (m_settings.get_int(settings_pack::predictive_piece_announce)) { int piece = block_finished.piece_index; piece_picker::downloading_piece st; t->picker().piece_info(piece, st); int num_blocks = t->picker().blocks_in_piece(piece); if (st.requested > 0 && st.writing + st.finished + st.requested == num_blocks) { std::vector d; t->picker().get_requestors(d, piece); if (d.size() == 1) { // only make predictions if all remaining // blocks are requested from the same peer torrent_peer* p = (torrent_peer*)d[0]; if (p->connection) { // we have a connection. now, what is the current // download rate from this peer, and how many blocks // do we have left to download? boost::int64_t rate = p->connection->statistics().download_payload_rate(); boost::int64_t bytes_left = boost::int64_t(st.requested) * t->block_size(); // the settings unit is milliseconds, so calculate the // number of milliseconds worth of bytes left in the piece if (rate > 1000 && (bytes_left * 1000) / rate < m_settings.get_int(settings_pack::predictive_piece_announce)) { // we predict we will complete this piece very soon. t->predicted_have_piece(piece, (bytes_left * 1000) / rate); } } } } } TORRENT_ASSERT(picker.num_peers(block_finished) == 0); #if TORRENT_USE_INVARIANT_CHECKS \ && defined TORRENT_EXPENSIVE_INVARIANT_CHECKS t->check_invariant(); #endif #if TORRENT_USE_ASSERTS piece_picker::downloading_piece pi; picker.piece_info(p.piece, pi); int num_blocks = picker.blocks_in_piece(p.piece); TORRENT_ASSERT(pi.writing + pi.finished + pi.requested <= num_blocks); TORRENT_ASSERT(picker.is_piece_finished(p.piece) == (pi.writing + pi.finished == num_blocks)); #endif // did we just finish the piece? // this means all blocks are either written // to disk or are in the disk write cache if (picker.is_piece_finished(p.piece) && !was_finished) { #if TORRENT_USE_INVARIANT_CHECKS check_postcondition post_checker2_(t, false); #endif t->verify_piece(p.piece); } if (is_disconnecting()) return; if (request_a_block(*t, *this)) m_counters.inc_stats_counter(counters::incoming_piece_picks); send_block_requests(); } void peer_connection::on_disk_write_complete(disk_io_job const* j , peer_request p, boost::shared_ptr t) { TORRENT_ASSERT(is_single_thread()); torrent_ref_holder h(t.get(), "async_write"); if (t) t->dec_refcount("async_write"); #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** FILE ASYNC WRITE COMPLETE [ ret: %d | piece: %d | s: %x | l: %x | e: %s ]" , j->ret, p.piece, p.start, p.length, j->error.ec.message().c_str()); #endif m_counters.inc_stats_counter(counters::queued_write_bytes, -p.length); m_outstanding_writing_bytes -= p.length; TORRENT_ASSERT(m_outstanding_writing_bytes >= 0); // flush send buffer at the end of // this burst of disk events // m_ses.cork_burst(this); INVARIANT_CHECK; if (!t) { disconnect(j->error.ec, op_file_write); return; } t->schedule_storage_tick(); // in case the outstanding bytes just dropped down // to allow to receive more data setup_receive(read_async); piece_block block_finished(p.piece, p.start / t->block_size()); if (j->ret < 0) { // handle_disk_error may disconnect us t->handle_disk_error(j, this); return; } TORRENT_ASSERT(j->ret == p.length); if (!t->has_picker()) return; piece_picker& picker = t->picker(); TORRENT_ASSERT(p.piece == j->piece); TORRENT_ASSERT(p.start == j->d.io.offset); TORRENT_ASSERT(picker.num_peers(block_finished) == 0); if (j->ret == -1 && j->error.ec == boost::system::errc::operation_canceled) { TORRENT_ASSERT(false); // how do we get here? picker.mark_as_canceled(block_finished, peer_info_struct()); return; } // fprintf(stderr, "peer_connection mark_as_finished peer: %p piece: %d block: %d\n" // , peer_info_struct(), block_finished.piece_index, block_finished.block_index); picker.mark_as_finished(block_finished, peer_info_struct()); t->maybe_done_flushing(); if (t->alerts().should_post()) { t->alerts().post_alert(block_finished_alert(t->get_handle(), remote(), pid(), block_finished.block_index, block_finished.piece_index)); } disconnect_if_redundant(); if (m_disconnecting) return; #if TORRENT_USE_ASSERTS if (t->has_picker()) { const std::vector& q = picker.get_download_queue(); for (std::vector::const_iterator i = q.begin(), end(q.end()); i != end; ++i) { if (i->index != block_finished.piece_index) continue; piece_picker::downloading_piece const& p = *i; TORRENT_ASSERT(p.info[block_finished.block_index].state == piece_picker::block_info::state_finished); } } #endif if (t->is_aborted()) return; } // ----------------------------- // ---------- CANCEL ----------- // ----------------------------- void peer_connection::incoming_cancel(peer_request const& r) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_cancel(r)) return; } #endif if (is_disconnecting()) return; #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== CANCEL [ piece: %d | s: %x | l: %x ]", r.piece, r.start, r.length); #endif std::vector::iterator i = std::find(m_requests.begin(), m_requests.end(), r); if (i != m_requests.end()) { m_counters.inc_stats_counter(counters::cancelled_piece_requests); m_requests.erase(i); if (m_requests.empty()) m_counters.inc_stats_counter(counters::num_peers_up_requests, -1); #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> REJECT_PIECE [ piece: %d s: %x l: %x ] cancelled" , r.piece , r.start , r.length); #endif write_reject_request(r); } else { // TODO: 3 since we throw away the queue entry once we issue // the disk job, this may happen. Instead, we should keep the // queue entry around, mark it as having been requested from // disk and once the disk job comes back, discard it if it has // been cancelled. Maybe even be able to cancel disk jobs? #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** GOT CANCEL NOT IN THE QUEUE"); #endif } } // ----------------------------- // --------- DHT PORT ---------- // ----------------------------- void peer_connection::incoming_dht_port(int listen_port) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== DHT_PORT [ p: %d ]", listen_port); #endif #ifndef TORRENT_DISABLE_DHT m_ses.add_dht_node(udp::endpoint( m_remote.address(), listen_port)); #endif } // ----------------------------- // --------- HAVE ALL ---------- // ----------------------------- void peer_connection::incoming_have_all() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); // we cannot disconnect in a constructor, and // this function may end up doing that TORRENT_ASSERT(m_in_constructor == false); #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== HAVE_ALL"); #endif #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_have_all()) return; } #endif if (is_disconnecting()) return; if (m_bitfield_received) t->peer_lost(m_have_piece, this); m_have_all = true; #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** THIS IS A SEED [ p: %p ]", m_peer_info); #endif t->set_seed(m_peer_info, true); m_upload_only = true; m_bitfield_received = true; #if defined TORRENT_LOGGING || defined TORRENT_ERROR_LOGGING m_bitfield_time = time_now_hires(); t->debug_log("HANDSHAKE [%p] (%d ms)", this, int(total_milliseconds(m_bitfield_time - m_connect_time))); #endif // if we don't have metadata yet // just remember the bitmask // don't update the piecepicker // (since it doesn't exist yet) if (!t->ready_for_connections()) { // assume seeds are interesting when we // don't even have the metadata t->peer_is_interesting(*this); disconnect_if_redundant(); return; } TORRENT_ASSERT(!m_have_piece.empty()); m_have_piece.set_all(); m_num_pieces = m_have_piece.size(); t->peer_has_all(this); #if TORRENT_USE_INVARIANT_CHECKS if (t && t->has_picker()) t->picker().check_peer_invariant(m_have_piece, this); #endif TORRENT_ASSERT(m_have_piece.all_set()); TORRENT_ASSERT(m_have_piece.count() == m_have_piece.size()); TORRENT_ASSERT(m_have_piece.size() == t->torrent_file().num_pieces()); // if we're finished, we're not interested if (t->is_upload_only()) send_not_interested(); else t->peer_is_interesting(*this); disconnect_if_redundant(); } // ----------------------------- // --------- HAVE NONE --------- // ----------------------------- void peer_connection::incoming_have_none() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== HAVE_NONE"); #endif boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_have_none()) return; } #endif if (is_disconnecting()) return; if (m_bitfield_received) t->peer_lost(m_have_piece, this); t->set_seed(m_peer_info, false); m_bitfield_received = true; #if defined TORRENT_LOGGING || defined TORRENT_ERROR_LOGGING m_bitfield_time = time_now_hires(); t->debug_log("HANDSHAKE [%p] (%d ms)", this, int(total_milliseconds(m_bitfield_time - m_connect_time))); #endif m_have_piece.clear_all(); m_num_pieces = 0; // if the peer is ready to download stuff, it must have metadata m_has_metadata = true; // we're never interested in a peer that doesn't have anything send_not_interested(); TORRENT_ASSERT(!m_have_piece.empty() || !t->ready_for_connections()); disconnect_if_redundant(); } // ----------------------------- // ------- ALLOWED FAST -------- // ----------------------------- void peer_connection::incoming_allowed_fast(int index) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); #ifdef TORRENT_LOGGING { ptime now = time_now_hires(); t->debug_log("ALLOW FAST [%p] (%d ms)", this, int(total_milliseconds(now - m_connect_time))); if (m_peer_choked) m_unchoke_time = now; } #endif #ifdef TORRENT_VERBOSE_LOGGING peer_log("<== ALLOWED_FAST [ %d ]", index); #endif #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((*i)->on_allowed_fast(index)) return; } #endif if (is_disconnecting()) return; if (index < 0) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("<== INVALID_ALLOWED_FAST [ %d ]", index); #endif return; } if (t->valid_metadata()) { if (index >= int(m_have_piece.size())) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("<== INVALID_ALLOWED_FAST [ %d | s: %d ]" , index, int(m_have_piece.size())); #endif return; } // if we already have the piece, we can // ignore this message if (t->have_piece(index)) return; } // if we don't have the metadata, we'll verify // this piece index later m_allowed_fast.push_back(index); // if the peer has the piece and we want // to download it, request it if (int(m_have_piece.size()) > index && m_have_piece[index] && !t->has_piece_passed(index) && t->valid_metadata() && t->has_picker() && t->picker().piece_priority(index) > 0) { t->peer_is_interesting(*this); } } std::vector const& peer_connection::allowed_fast() { TORRENT_ASSERT(is_single_thread()); boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); // TODO: sort the allowed fast set in priority order return m_allowed_fast; } bool peer_connection::can_request_time_critical() const { TORRENT_ASSERT(is_single_thread()); if (has_peer_choked() || !is_interesting()) return false; if ((int)m_download_queue.size() + (int)m_request_queue.size() > m_desired_queue_size * 2) return false; if (on_parole()) return false; if (m_disconnecting) return false; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); if (t->upload_mode()) return false; // ignore snubbed peers, since they're not likely to return pieces in a // timely manner anyway if (m_snubbed) return false; return true; } bool peer_connection::make_time_critical(piece_block const& block) { TORRENT_ASSERT(is_single_thread()); std::vector::iterator rit = std::find_if(m_request_queue.begin() , m_request_queue.end(), has_block(block)); if (rit == m_request_queue.end()) return false; #if TORRENT_USE_ASSERTS boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); TORRENT_ASSERT(t->has_picker()); TORRENT_ASSERT(t->picker().is_requested(block)); #endif // ignore it if it's already time critical if (rit - m_request_queue.begin() < m_queued_time_critical) return false; pending_block b = *rit; m_request_queue.erase(rit); m_request_queue.insert(m_request_queue.begin() + m_queued_time_critical, b); ++m_queued_time_critical; return true; } bool peer_connection::add_request(piece_block const& block, int flags) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); TORRENT_ASSERT(!m_disconnecting); TORRENT_ASSERT(t->valid_metadata()); TORRENT_ASSERT(block.block_index != piece_block::invalid.block_index); TORRENT_ASSERT(block.piece_index != piece_block::invalid.piece_index); TORRENT_ASSERT(int(block.piece_index) < t->torrent_file().num_pieces()); TORRENT_ASSERT(int(block.block_index) < t->torrent_file().piece_size(block.piece_index)); TORRENT_ASSERT(!t->picker().is_requested(block) || (t->picker().num_peers(block) > 0)); TORRENT_ASSERT(!t->have_piece(block.piece_index)); TORRENT_ASSERT(std::find_if(m_download_queue.begin(), m_download_queue.end() , has_block(block)) == m_download_queue.end()); TORRENT_ASSERT(std::find(m_request_queue.begin(), m_request_queue.end() , block) == m_request_queue.end()); if (t->upload_mode()) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** PIECE_PICKER [ not_picking: %d,%d upload_mode ]" , block.piece_index, block.block_index); #endif return false; } if (m_disconnecting) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** PIECE_PICKER [ not_picking: %d,%d disconnecting ]" , block.piece_index, block.block_index); #endif return false; } piece_picker::piece_state_t state; peer_speed_t speed = peer_speed(); char const* speedmsg = 0; if (speed == fast) { speedmsg = "fast"; state = piece_picker::fast; } else if (speed == medium) { speedmsg = "medium"; state = piece_picker::medium; } else { speedmsg = "slow"; state = piece_picker::slow; } if ((flags & req_busy) && !(flags & req_time_critical)) { // this block is busy (i.e. it has been requested // from another peer already). Only allow one busy // request in the pipeline at the time // this rule does not apply to time critical pieces, // in which case we are allowed to pick more than one // busy blocks for (std::vector::const_iterator i = m_download_queue.begin() , end(m_download_queue.end()); i != end; ++i) { if (i->busy) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** PIECE_PICKER [ not_picking: %d,%d already in download queue & busy ]" , block.piece_index, block.block_index); #endif return false; } } for (std::vector::const_iterator i = m_request_queue.begin() , end(m_request_queue.end()); i != end; ++i) { if (i->busy) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** PIECE_PICKER [ not_picking: %d,%d already in request queue & busy ]" , block.piece_index, block.block_index); #endif return false; } } } if (!t->picker().mark_as_downloading(block, peer_info_struct(), state)) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** PIECE_PICKER [ not_picking: %d,%d failed to mark_as_downloading ]" , block.piece_index, block.block_index); #endif return false; } if (t->alerts().should_post()) { t->alerts().post_alert(block_downloading_alert(t->get_handle(), remote(), pid(), speedmsg, block.block_index, block.piece_index)); } pending_block pb(block); pb.busy = (flags & req_busy) ? true : false; if (flags & req_time_critical) { m_request_queue.insert(m_request_queue.begin() + m_queued_time_critical , pb); ++m_queued_time_critical; } else { m_request_queue.push_back(pb); } return true; } void peer_connection::cancel_all_requests() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); // this peer might be disconnecting if (!t) return; TORRENT_ASSERT(t->valid_metadata()); #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** CANCEL ALL REQUESTS"); #endif while (!m_request_queue.empty()) { t->picker().abort_download(m_request_queue.back().block, peer_info_struct()); m_request_queue.pop_back(); } m_queued_time_critical = 0; // make a local temporary copy of the download queue, since it // may be modified when we call write_cancel (for peers that don't // support the FAST extensions). std::vector temp_copy = m_download_queue; for (std::vector::iterator i = temp_copy.begin() , end(temp_copy.end()); i != end; ++i) { piece_block b = i->block; int block_offset = b.block_index * t->block_size(); int block_size = (std::min)(t->torrent_file().piece_size(b.piece_index)-block_offset, t->block_size()); TORRENT_ASSERT(block_size > 0); TORRENT_ASSERT(block_size <= t->block_size()); // we can't cancel the piece if we've started receiving it if (m_receiving_block == b) continue; peer_request r; r.piece = b.piece_index; r.start = block_offset; r.length = block_size; #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> CANCEL [ piece: %d s: %d l: %d b: %d ]" , b.piece_index, block_offset, block_size, b.block_index); #endif write_cancel(r); } } void peer_connection::cancel_request(piece_block const& block, bool force) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); // this peer might be disconnecting if (!t) return; TORRENT_ASSERT(t->valid_metadata()); TORRENT_ASSERT(block.block_index != piece_block::invalid.block_index); TORRENT_ASSERT(block.piece_index != piece_block::invalid.piece_index); TORRENT_ASSERT(int(block.piece_index) < t->torrent_file().num_pieces()); TORRENT_ASSERT(int(block.block_index) < t->torrent_file().piece_size(block.piece_index)); // if all the peers that requested this block has been // cancelled, then just ignore the cancel. if (!t->picker().is_requested(block)) return; std::vector::iterator it = std::find_if(m_download_queue.begin(), m_download_queue.end(), has_block(block)); if (it == m_download_queue.end()) { std::vector::iterator rit = std::find_if(m_request_queue.begin() , m_request_queue.end(), has_block(block)); // when a multi block is received, it is cancelled // from all peers, so if this one hasn't requested // the block, just ignore to cancel it. if (rit == m_request_queue.end()) return; if (rit - m_request_queue.begin() < m_queued_time_critical) --m_queued_time_critical; t->picker().abort_download(block, peer_info_struct()); m_request_queue.erase(rit); // since we found it in the request queue, it means it hasn't been // sent yet, so we don't have to send a cancel. return; } int block_offset = block.block_index * t->block_size(); int block_size = (std::min)(t->torrent_file().piece_size(block.piece_index)-block_offset, t->block_size()); TORRENT_ASSERT(block_size > 0); TORRENT_ASSERT(block_size <= t->block_size()); it->not_wanted = true; if (force) t->picker().abort_download(block, peer_info_struct()); if (m_outstanding_bytes < block_size) return; peer_request r; r.piece = block.piece_index; r.start = block_offset; r.length = block_size; #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> CANCEL [ piece: %d s: %d l: %d b: %d ]" , block.piece_index, block_offset, block_size, block.block_index); #endif write_cancel(r); } bool peer_connection::send_choke() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; if (m_choked) { TORRENT_ASSERT(m_peer_info == NULL || m_peer_info->optimistically_unchoked == false); return false; } if (m_peer_info && m_peer_info->optimistically_unchoked) { m_peer_info->optimistically_unchoked = false; m_counters.inc_stats_counter(counters::num_peers_up_unchoked_optimistic, -1); } #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> CHOKE"); #endif write_choke(); m_counters.inc_stats_counter(counters::num_peers_up_unchoked_all, -1); if (!ignore_unchoke_slots()) m_counters.inc_stats_counter(counters::num_peers_up_unchoked, -1); m_choked = true; m_last_choke = time_now(); m_num_invalid_requests = 0; // reject the requests we have in the queue // except the allowed fast pieces for (std::vector::iterator i = m_requests.begin(); i != m_requests.end();) { if (std::find(m_accept_fast.begin(), m_accept_fast.end(), i->piece) != m_accept_fast.end()) { ++i; continue; } peer_request const& r = *i; m_counters.inc_stats_counter(counters::choked_piece_requests); #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> REJECT_PIECE [ piece: %d s: %d l: %d ] choking" , r.piece , r.start , r.length); #endif write_reject_request(r); i = m_requests.erase(i); if (m_requests.empty()) m_counters.inc_stats_counter(counters::num_peers_up_requests, -1); } return true; } bool peer_connection::send_unchoke() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; if (!m_choked) return false; boost::shared_ptr t = m_torrent.lock(); if (!t->ready_for_connections()) return false; if (!m_sent_suggests) { std::vector const& ret = t->get_suggested_pieces(); for (std::vector::const_iterator i = ret.begin() , end(ret.end()); i != end; ++i) { TORRENT_ASSERT(i->piece_index >= 0); // this can happen if a piece fail to be // flushed to disk for whatever reason if (!t->has_piece_passed(i->piece_index)) continue; send_suggest(i->piece_index); } m_sent_suggests = true; } m_last_unchoke = time_now(); write_unchoke(); m_counters.inc_stats_counter(counters::num_peers_up_unchoked_all); if (!ignore_unchoke_slots()) m_counters.inc_stats_counter(counters::num_peers_up_unchoked); m_choked = false; m_uploaded_at_last_unchoke = m_statistics.total_payload_upload(); #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> UNCHOKE"); #endif return true; } void peer_connection::send_interested() { TORRENT_ASSERT(is_single_thread()); if (m_interesting) return; boost::shared_ptr t = m_torrent.lock(); if (!t->ready_for_connections()) return; m_interesting = true; m_counters.inc_stats_counter(counters::num_peers_down_interested); write_interested(); #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> INTERESTED"); #endif } void peer_connection::send_not_interested() { TORRENT_ASSERT(is_single_thread()); // we cannot disconnect in a constructor, and // this function may end up doing that TORRENT_ASSERT(m_in_constructor == false); if (!m_interesting) { disconnect_if_redundant(); return; } boost::shared_ptr t = m_torrent.lock(); if (!t->ready_for_connections()) return; m_interesting = false; m_counters.inc_stats_counter(counters::num_peers_down_interested, -1); disconnect_if_redundant(); if (m_disconnecting) return; write_not_interested(); m_became_uninteresting = time_now(); #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> NOT_INTERESTED"); #endif } void peer_connection::send_suggest(int piece) { TORRENT_ASSERT(is_single_thread()); if (m_connecting) return; if (in_handshake()) return; // don't suggest a piece that the peer already has // don't suggest anything to a peer that isn't interested if (has_piece(piece) || !m_peer_interested) return; // we cannot suggest a piece we don't have! #if TORRENT_USE_ASSERTS boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); TORRENT_ASSERT(t->has_piece_passed(piece)); #endif TORRENT_ASSERT(piece >= 0 && piece < t->torrent_file().num_pieces()); if (m_sent_suggested_pieces.empty()) { boost::shared_ptr t = m_torrent.lock(); m_sent_suggested_pieces.resize(t->torrent_file().num_pieces(), false); } TORRENT_ASSERT(piece >= 0 && piece < m_sent_suggested_pieces.size()); if (m_sent_suggested_pieces[piece]) return; m_sent_suggested_pieces.set_bit(piece); write_suggest(piece); } void peer_connection::send_block_requests() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); if (m_disconnecting) return; if (t->graceful_pause() && m_outstanding_bytes == 0) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** GRACEFUL PAUSE [ NO MORE DOWNLOAD ]"); #endif disconnect(errors::torrent_paused, op_bittorrent); return; } // we can't download pieces in these states if (t->state() == torrent_status::checking_files || t->state() == torrent_status::checking_resume_data || t->state() == torrent_status::downloading_metadata || t->state() == torrent_status::allocating) return; if ((int)m_download_queue.size() >= m_desired_queue_size || t->upload_mode()) return; bool empty_download_queue = m_download_queue.empty(); while (!m_request_queue.empty() && ((int)m_download_queue.size() < m_desired_queue_size || m_queued_time_critical > 0)) { pending_block block = m_request_queue.front(); m_request_queue.erase(m_request_queue.begin()); if (m_queued_time_critical) --m_queued_time_critical; // if we're a seed, we don't have a piece picker // so we don't have to worry about invariants getting // out of sync with it if (!t->has_picker()) continue; // this can happen if a block times out, is re-requested and // then arrives "unexpectedly" if (t->picker().is_finished(block.block) || t->picker().is_downloaded(block.block)) { t->picker().abort_download(block.block, peer_info_struct()); continue; } int block_offset = block.block.block_index * t->block_size(); int block_size = (std::min)(t->torrent_file().piece_size( block.block.piece_index) - block_offset, t->block_size()); TORRENT_ASSERT(block_size > 0); TORRENT_ASSERT(block_size <= t->block_size()); peer_request r; r.piece = block.block.piece_index; r.start = block_offset; r.length = block_size; if (m_download_queue.empty()) m_counters.inc_stats_counter(counters::num_peers_down_requests); TORRENT_ASSERT(verify_piece(t->to_req(block.block))); block.send_buffer_offset = m_send_buffer.size(); m_download_queue.push_back(block); m_outstanding_bytes += block_size; #if TORRENT_USE_INVARIANT_CHECKS check_invariant(); #endif /* #ifdef TORRENT_VERBOSE_LOGGING (*m_logger) << time_now_string() << " *** REQUEST-QUEUE** [ " "piece: " << block.piece_index << " | " "block: " << block.block_index << " ]\n"; #endif */ // if we are requesting large blocks, merge the smaller // blocks that are in the same piece into larger requests if (m_request_large_blocks) { int blocks_per_piece = t->torrent_file().piece_length() / t->block_size(); while (!m_request_queue.empty()) { // check to see if this block is connected to the previous one // if it is, merge them, otherwise, break this merge loop pending_block const& front = m_request_queue.front(); if (front.block.piece_index * blocks_per_piece + front.block.block_index != block.block.piece_index * blocks_per_piece + block.block.block_index + 1) break; block = m_request_queue.front(); m_request_queue.erase(m_request_queue.begin()); TORRENT_ASSERT(verify_piece(t->to_req(block.block))); if (m_download_queue.empty()) m_counters.inc_stats_counter(counters::num_peers_down_requests); block.send_buffer_offset = m_send_buffer.size(); m_download_queue.push_back(block); if (m_queued_time_critical) --m_queued_time_critical; #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** MERGING REQUEST [ piece: %d block: %d ]" , block.block.piece_index, block.block.block_index); #endif block_offset = block.block.block_index * t->block_size(); block_size = (std::min)(t->torrent_file().piece_size( block.block.piece_index) - block_offset, t->block_size()); TORRENT_ASSERT(block_size > 0); TORRENT_ASSERT(block_size <= t->block_size()); r.length += block_size; m_outstanding_bytes += block_size; #if TORRENT_USE_INVARIANT_CHECKS check_invariant(); #endif } } // the verification will fail for coalesced blocks TORRENT_ASSERT(verify_piece(r) || m_request_large_blocks); #ifndef TORRENT_DISABLE_EXTENSIONS bool handled = false; for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { if ((handled = (*i)->write_request(r))) break; } if (is_disconnecting()) return; if (!handled) #endif { write_request(r); m_last_request = time_now(); } #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> REQUEST [ piece: %d | s: %x | l: %x | ds: %d B/s | " "dqs: %d rqs: %d blk: %s ]" , r.piece, r.start, r.length, statistics().download_rate() , int(m_desired_queue_size), int(m_download_queue.size()) , m_request_large_blocks?"large":"single"); #endif } m_last_piece = time_now(); if (!m_download_queue.empty() && empty_download_queue) { // This means we just added a request to this connection m_requested = time_now(); #if defined TORRENT_LOGGING || defined TORRENT_ERROR_LOGGING t->debug_log("REQUEST [%p] (%d ms)", this , int(total_milliseconds(time_now_hires() - m_unchoke_time))); #endif } } void peer_connection::connect_failed(error_code const& e) { TORRENT_ASSERT(is_single_thread()); TORRENT_ASSERT(e); #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("CONNECTION FAILED: %s", print_endpoint(m_remote).c_str()); #endif #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_LOGGING || defined TORRENT_ERROR_LOGGING m_ses.session_log(" CONNECTION FAILED: %s", print_endpoint(m_remote).c_str()); #endif m_counters.inc_stats_counter(counters::connect_timeouts); boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(!m_connecting || t); if (m_connecting) { m_counters.inc_stats_counter(counters::num_peers_half_open, -1); if (t) t->dec_num_connecting(); m_connecting = false; } // a connection attempt using uTP just failed // mark this peer as not supporting uTP // we'll never try it again (unless we're trying holepunch) if (is_utp(*m_socket) && m_peer_info && m_peer_info->supports_utp && !m_holepunch_mode) { m_peer_info->supports_utp = false; // reconnect immediately using TCP torrent_peer* pi = peer_info_struct(); boost::shared_ptr t = m_torrent.lock(); fast_reconnect(true); disconnect(e, op_connect, 0); if (t && pi) t->connect_to_peer(pi, true); return; } if (m_holepunch_mode) fast_reconnect(true); #ifndef TORRENT_DISABLE_EXTENSIONS if ((!is_utp(*m_socket) || !m_settings.get_bool(settings_pack::enable_outgoing_tcp)) && m_peer_info && m_peer_info->supports_holepunch && !m_holepunch_mode) { boost::shared_ptr t = m_torrent.lock(); // see if we can try a holepunch bt_peer_connection* p = t->find_introducer(remote()); if (p) p->write_holepunch_msg(bt_peer_connection::hp_rendezvous, remote(), 0); } #endif disconnect(e, op_connect, 1); return; } // the error argument defaults to 0, which means deliberate disconnect // 1 means unexpected disconnect/error // 2 protocol error (client sent something invalid) void peer_connection::disconnect(error_code const& ec, operation_t op, int error) { TORRENT_ASSERT(is_single_thread()); #if TORRENT_USE_ASSERTS m_disconnect_started = true; #endif if (m_disconnecting) return; // while being disconnected, it's possible that our torrent_peer // pointer gets cleared. Make sure we save it to be able to keep // proper books in the piece_picker (when debugging is enabled) torrent_peer* self_peer = peer_info_struct(); #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING switch (error) { case 0: peer_log("*** CONNECTION CLOSED [op: %d] %s", op, ec.message().c_str()); break; case 1: peer_log("*** CONNECTION FAILED [op: %d] %s", op, ec.message().c_str()); break; case 2: peer_log("*** PEER ERROR [op: %d] %s", op, ec.message().c_str()); break; } #endif if ((m_channel_state[upload_channel] & peer_info::bw_network) == 0) { // make sure we free up all send buffers that are owned // by the disk thread m_send_buffer.clear(); m_disk_recv_buffer.reset(); m_disk_recv_buffer_size = 0; } // we cannot do this in a constructor TORRENT_ASSERT(m_in_constructor == false); if (error > 0) m_failed = true; if (m_connected) m_counters.inc_stats_counter(counters::num_peers_connected, -1); m_connected = false; // for incoming connections, we get invalid argument errors // when asking for the remote endpoint and the socket already // closed, which is an edge case, but possible to happen when // a peer makes a TCP and uTP connection in parallel. // for outgoing connections however, why would we get this? // TORRENT_ASSERT(ec != error::invalid_argument || !m_outgoing); m_counters.inc_stats_counter(counters::disconnected_peers); if (error == 2) m_counters.inc_stats_counter(counters::error_peers); if (ec == error::connection_reset) m_counters.inc_stats_counter(counters::connreset_peers); else if (ec == error::eof) m_counters.inc_stats_counter(counters::eof_peers); else if (ec == error::connection_refused) m_counters.inc_stats_counter(counters::connrefused_peers); else if (ec == error::connection_aborted) m_counters.inc_stats_counter(counters::connaborted_peers); else if (ec == error::no_permission) m_counters.inc_stats_counter(counters::perm_peers); else if (ec == error::no_buffer_space) m_counters.inc_stats_counter(counters::buffer_peers); else if (ec == error::host_unreachable) m_counters.inc_stats_counter(counters::unreachable_peers); else if (ec == error::broken_pipe) m_counters.inc_stats_counter(counters::broken_pipe_peers); else if (ec == error::address_in_use) m_counters.inc_stats_counter(counters::addrinuse_peers); else if (ec == error::access_denied) m_counters.inc_stats_counter(counters::no_access_peers); else if (ec == error::invalid_argument) m_counters.inc_stats_counter(counters::invalid_arg_peers); else if (ec == error::operation_aborted) m_counters.inc_stats_counter(counters::aborted_peers); else if (ec == error_code(errors::upload_upload_connection) || ec == error_code(errors::uninteresting_upload_peer) || ec == error_code(errors::torrent_aborted) || ec == error_code(errors::self_connection) || ec == error_code(errors::torrent_paused)) m_counters.inc_stats_counter(counters::uninteresting_peers); if (ec == error_code(errors::timed_out) || ec == error::timed_out) m_counters.inc_stats_counter(counters::transport_timeout_peers); if (ec == error_code(errors::timed_out_inactivity) || ec == error_code(errors::timed_out_no_request) || ec == error_code(errors::timed_out_no_interest)) m_counters.inc_stats_counter(counters::timeout_peers); if (ec == error_code(errors::no_memory)) m_counters.inc_stats_counter(counters::no_memory_peers); if (ec == error_code(errors::too_many_connections)) m_counters.inc_stats_counter(counters::too_many_peers); if (ec == error_code(errors::timed_out_no_handshake)) m_counters.inc_stats_counter(counters::connect_timeouts); if (error > 0) { if (is_utp(*m_socket)) m_counters.inc_stats_counter(counters::error_utp_peers); else m_counters.inc_stats_counter(counters::error_tcp_peers); if (m_outgoing) m_counters.inc_stats_counter(counters::error_outgoing_peers); else m_counters.inc_stats_counter(counters::error_incoming_peers); #ifndef TORRENT_DISABLE_ENCRYPTION if (type() == bittorrent_connection) { bt_peer_connection* bt = static_cast(this); if (bt->supports_encryption()) m_counters.inc_stats_counter( counters::error_encrypted_peers); if (bt->rc4_encrypted() && bt->supports_encryption()) m_counters.inc_stats_counter(counters::error_rc4_peers); } #endif // TORRENT_DISABLE_ENCRYPTION } boost::shared_ptr me(self()); INVARIANT_CHECK; if (m_channel_state[upload_channel] & peer_info::bw_disk) { m_counters.inc_stats_counter(counters::num_peers_up_disk, -1); m_channel_state[upload_channel] &= ~peer_info::bw_disk; } if (m_channel_state[download_channel] & peer_info::bw_disk) { m_counters.inc_stats_counter(counters::num_peers_down_disk, -1); m_channel_state[download_channel] &= ~peer_info::bw_disk; } boost::shared_ptr t = m_torrent.lock(); if (m_connecting) { m_counters.inc_stats_counter(counters::num_peers_half_open, -1); if (t) t->dec_num_connecting(); m_connecting = false; } torrent_handle handle; if (t) handle = t->get_handle(); #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { (*i)->on_disconnect(ec); } #endif if (ec == error::address_in_use && m_settings.get_int(settings_pack::outgoing_port) != 0 && t) { if (t->alerts().should_post()) t->alerts().post_alert(performance_alert( handle, performance_alert::too_few_outgoing_ports)); } if (t) { if (ec) { if ((error > 1 || ec.category() == get_socks_category()) && t->alerts().should_post()) { t->alerts().post_alert( peer_error_alert(handle, remote(), pid(), op, ec)); } else if (error <= 1 && t->alerts().should_post()) { t->alerts().post_alert( peer_disconnected_alert(handle, remote(), pid(), op, ec)); } } // make sure we keep all the stats! if (!m_ignore_stats) { // report any partially received payload as redundant boost::optional pbp = downloading_piece_progress(); if (pbp && pbp->bytes_downloaded > 0 && pbp->bytes_downloaded < pbp->full_block_bytes) { t->add_redundant_bytes(pbp->bytes_downloaded, torrent::piece_closing); } } if (t->has_picker()) { piece_picker& picker = t->picker(); while (!m_download_queue.empty()) { pending_block& qe = m_download_queue.back(); if (!qe.timed_out && !qe.not_wanted) picker.abort_download(qe.block, self_peer); m_outstanding_bytes -= t->to_req(qe.block).length; if (m_outstanding_bytes < 0) m_outstanding_bytes = 0; m_download_queue.pop_back(); } while (!m_request_queue.empty()) { pending_block& qe = m_request_queue.back(); if (!qe.timed_out && !qe.not_wanted) picker.abort_download(qe.block, self_peer); m_request_queue.pop_back(); } } else { m_download_queue.clear(); m_request_queue.clear(); m_outstanding_bytes = 0; } m_queued_time_critical = 0; #if TORRENT_USE_INVARIANT_CHECKS check_invariant(); #endif t->remove_peer(this); } else { TORRENT_ASSERT(m_download_queue.empty()); TORRENT_ASSERT(m_request_queue.empty()); } #if defined TORRENT_DEBUG && defined TORRENT_EXPENSIVE_INVARIANT_CHECKS // since this connection doesn't have a torrent reference // no torrent should have a reference to this connection either TORRENT_ASSERT(!m_ses.any_torrent_has_peer(this)); #endif m_disconnecting = true; error_code e; async_shutdown(*m_socket, m_socket); m_ses.close_connection(this, ec); } bool peer_connection::ignore_unchoke_slots() const { TORRENT_ASSERT(is_single_thread()); if (num_classes() == 0) return true; if (m_ses.ignore_unchoke_slots_set(*this)) return true; boost::shared_ptr t = m_torrent.lock(); if (t && m_ses.ignore_unchoke_slots_set(*t)) return true; return false; } // defined in upnp.cpp bool is_local(address const& a); bool peer_connection::on_local_network() const { TORRENT_ASSERT(is_single_thread()); if (libtorrent::is_local(m_remote.address()) || is_loopback(m_remote.address())) return true; return false; } void peer_connection::get_peer_info(peer_info& p) const { TORRENT_ASSERT(is_single_thread()); TORRENT_ASSERT(!associated_torrent().expired()); ptime now = time_now(); p.download_rate_peak = m_download_rate_peak; p.upload_rate_peak = m_upload_rate_peak; p.rtt = m_rtt.mean(); p.down_speed = statistics().download_rate(); p.up_speed = statistics().upload_rate(); p.payload_down_speed = statistics().download_payload_rate(); p.payload_up_speed = statistics().upload_payload_rate(); p.pid = pid(); p.ip = remote(); p.pending_disk_bytes = m_outstanding_writing_bytes; p.pending_disk_read_bytes = m_reading_bytes; p.send_quota = m_quota[upload_channel]; p.receive_quota = m_quota[download_channel]; p.num_pieces = m_num_pieces; if (m_download_queue.empty()) p.request_timeout = -1; else p.request_timeout = int(total_seconds(m_requested - now) + m_settings.get_int(settings_pack::request_timeout) + m_timeout_extend); p.download_queue_time = download_queue_time(); p.queue_bytes = m_outstanding_bytes; #ifndef TORRENT_DISABLE_RESOLVE_COUNTRIES p.country[0] = m_country[0]; p.country[1] = m_country[1]; #else std::fill(p.country, p.country + 2, 0); #endif p.total_download = statistics().total_payload_download(); p.total_upload = statistics().total_payload_upload(); #ifndef TORRENT_NO_DEPRECATE p.upload_limit = -1; p.download_limit = -1; p.load_balancing = 0; #endif p.download_queue_length = int(download_queue().size() + m_request_queue.size()); p.requests_in_buffer = int(std::count_if(m_download_queue.begin() , m_download_queue.end() , boost::bind(&pending_block::send_buffer_offset, _1) >= 0)); p.target_dl_queue_length = int(desired_queue_size()); p.upload_queue_length = int(upload_queue().size()); p.timed_out_requests = 0; p.busy_requests = 0; for (std::vector::const_iterator i = m_download_queue.begin() , end(m_download_queue.end()); i != end; ++i) { if (i->timed_out) ++p.timed_out_requests; if (i->busy) ++p.busy_requests; } if (boost::optional ret = downloading_piece_progress()) { p.downloading_piece_index = ret->piece_index; p.downloading_block_index = ret->block_index; p.downloading_progress = ret->bytes_downloaded; p.downloading_total = ret->full_block_bytes; } else { p.downloading_piece_index = -1; p.downloading_block_index = -1; p.downloading_progress = 0; p.downloading_total = 0; } p.pieces = get_bitfield(); p.last_request = now - m_last_request; p.last_active = now - (std::max)(m_last_sent, m_last_receive); // this will set the flags so that we can update them later p.flags = 0; get_specific_peer_info(p); p.flags |= is_seed() ? peer_info::seed : 0; p.flags |= m_snubbed ? peer_info::snubbed : 0; p.flags |= m_upload_only ? peer_info::upload_only : 0; p.flags |= m_endgame_mode ? peer_info::endgame_mode : 0; p.flags |= m_holepunch_mode ? peer_info::holepunched : 0; if (peer_info_struct()) { torrent_peer* pi = peer_info_struct(); TORRENT_ASSERT(pi->in_use); p.source = pi->source; p.failcount = pi->failcount; p.num_hashfails = pi->hashfails; p.flags |= pi->on_parole ? peer_info::on_parole : 0; p.flags |= pi->optimistically_unchoked ? peer_info::optimistic_unchoke : 0; #ifndef TORRENT_NO_DEPRECATE p.inet_as = 0xffff; #endif } else { p.source = 0; p.failcount = 0; p.num_hashfails = 0; p.inet_as = 0xffff; } p.remote_dl_rate = m_remote_dl_rate; p.send_buffer_size = m_send_buffer.capacity(); p.used_send_buffer = m_send_buffer.size(); p.receive_buffer_size = m_recv_buffer.capacity() + m_disk_recv_buffer_size; p.used_receive_buffer = m_recv_pos; p.write_state = m_channel_state[upload_channel]; p.read_state = m_channel_state[download_channel]; // pieces may be empty if we don't have metadata yet if (p.pieces.size() == 0) { p.progress = 0.f; p.progress_ppm = 0; } else { #if TORRENT_NO_FPU p.progress = 0.f; #else p.progress = (float)p.pieces.count() / (float)p.pieces.size(); #endif p.progress_ppm = boost::uint64_t(p.pieces.count()) * 1000000 / p.pieces.size(); } p.estimated_reciprocation_rate = m_est_reciprocation_rate; error_code ec; p.local_endpoint = get_socket()->local_endpoint(ec); } // allocates a disk buffer of size 'disk_buffer_size' and replaces the // end of the current receive buffer with it. i.e. the receive pos // must be <= packet_size - disk_buffer_size // the disk buffer can be accessed through release_disk_receive_buffer() // when it is queried, the responsibility to free it is transferred // to the caller bool peer_connection::allocate_disk_receive_buffer(int disk_buffer_size) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; TORRENT_ASSERT(m_packet_size > 0); TORRENT_ASSERT(m_recv_pos <= m_packet_size - disk_buffer_size); TORRENT_ASSERT(!m_disk_recv_buffer); TORRENT_ASSERT(m_disk_recv_buffer_size == 0); TORRENT_ASSERT(disk_buffer_size <= 16 * 1024); if (disk_buffer_size == 0) return true; if (disk_buffer_size > 16 * 1024) { disconnect(errors::invalid_piece_size, op_bittorrent, 2); return false; } // first free the old buffer m_disk_recv_buffer.reset(); // then allocate a new one bool exceeded = false; m_disk_recv_buffer.reset(m_allocator.allocate_disk_buffer(exceeded, self(), "receive buffer")); if (!m_disk_recv_buffer) { disconnect(errors::no_memory, op_alloc_recvbuf); return false; } if (exceeded) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** exceeded disk buffer watermark"); #endif if ((m_channel_state[download_channel] & peer_info::bw_disk) == 0) m_counters.inc_stats_counter(counters::num_peers_down_disk); m_channel_state[download_channel] |= peer_info::bw_disk; } m_disk_recv_buffer_size = disk_buffer_size; return true; } char* peer_connection::release_disk_receive_buffer() { TORRENT_ASSERT(is_single_thread()); if (!m_disk_recv_buffer) return 0; TORRENT_ASSERT(m_disk_recv_buffer_size <= m_recv_end); TORRENT_ASSERT(m_recv_start <= m_recv_end - m_disk_recv_buffer_size); m_recv_end -= m_disk_recv_buffer_size; m_disk_recv_buffer_size = 0; return m_disk_recv_buffer.release(); } // size = the packet size to remove from the receive buffer // packet_size = the next packet size to receive in the buffer // offset = the offset into the receive buffer where to remove `size` bytes void peer_connection::cut_receive_buffer(int size, int packet_size, int offset) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; TORRENT_ASSERT(packet_size > 0); TORRENT_ASSERT(int(m_recv_buffer.size()) >= size); TORRENT_ASSERT(int(m_recv_buffer.size()) >= m_recv_pos); TORRENT_ASSERT(m_recv_pos >= size + offset); TORRENT_ASSERT(offset >= 0); TORRENT_ASSERT(m_recv_buffer.size() >= m_recv_end); TORRENT_ASSERT(m_recv_start <= m_recv_end); TORRENT_ASSERT(size >= 0); if (offset > 0) { TORRENT_ASSERT(m_recv_start - size <= m_recv_end); if (size > 0) std::memmove(&m_recv_buffer[0] + m_recv_start + offset , &m_recv_buffer[0] + m_recv_start + offset + size , m_recv_end - m_recv_start - size - offset); m_recv_pos -= size; m_recv_end -= size; #ifdef TORRENT_DEBUG std::fill(m_recv_buffer.begin() + m_recv_end, m_recv_buffer.end(), 0xcc); #endif } else { TORRENT_ASSERT(m_recv_start + size <= m_recv_end); m_recv_start += size; m_recv_pos -= size; } m_packet_size = packet_size; } // the purpose of this function is to free up and cut off all messages // in the receive buffer that have been parsed and processed. void peer_connection::normalize_receive_buffer() { TORRENT_ASSERT(is_single_thread()); TORRENT_ASSERT(m_recv_end >= m_recv_start); if (m_recv_start == 0) return; if (m_recv_end > m_recv_start) std::memmove(&m_recv_buffer[0], &m_recv_buffer[0] + m_recv_start, m_recv_end - m_recv_start); m_recv_end -= m_recv_start; m_recv_start = 0; #ifdef TORRENT_DEBUG std::fill(m_recv_buffer.begin() + m_recv_end, m_recv_buffer.end(), 0xcc); #endif } void peer_connection::superseed_piece(int replace_piece, int new_piece) { TORRENT_ASSERT(is_single_thread()); if (new_piece == -1) { if (m_superseed_piece[0] == -1) return; m_superseed_piece[0] = -1; m_superseed_piece[1] = -1; #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** ending super seed mode"); #endif boost::shared_ptr t = m_torrent.lock(); assert(t); // this will either send a full bitfield or // a have-all message, effectively terminating // super-seeding, since the peer may pick any piece write_bitfield(); return; } assert(!has_piece(new_piece)); #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> HAVE [ piece: %d ] (super seed)", new_piece); #endif write_have(new_piece); if (replace_piece >= 0) { // move the piece we're replacing to the tail if (m_superseed_piece[0] == replace_piece) std::swap(m_superseed_piece[0], m_superseed_piece[1]); } m_superseed_piece[1] = m_superseed_piece[0]; m_superseed_piece[0] = new_piece; } void peer_connection::update_desired_queue_size() { TORRENT_ASSERT(is_single_thread()); if (m_snubbed) { m_desired_queue_size = 1; return; } int download_rate = statistics().download_payload_rate(); // calculate the desired download queue size const int queue_time = m_settings.get_int(settings_pack::request_queue_time); // (if the latency is more than this, the download will stall) // so, the queue size is queue_time * down_rate / 16 kiB // (16 kB is the size of each request) // the minimum number of requests is 2 and the maximum is 48 // the block size doesn't have to be 16. So we first query the // torrent for it boost::shared_ptr t = m_torrent.lock(); const int block_size = t->block_size(); TORRENT_ASSERT(block_size > 0); m_desired_queue_size = queue_time * download_rate / block_size; if (m_desired_queue_size > m_max_out_request_queue) m_desired_queue_size = m_max_out_request_queue; if (m_desired_queue_size < min_request_queue) m_desired_queue_size = min_request_queue; } void peer_connection::second_tick(int tick_interval_ms) { TORRENT_ASSERT(is_single_thread()); ptime now = time_now(); boost::shared_ptr me(self()); // the invariant check must be run before me is destructed // in case the peer got disconnected INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); int warning = 0; // drain the IP overhead from the bandwidth limiters if (m_settings.get_bool(settings_pack::rate_limit_ip_overhead) && t) { warning |= m_ses.use_quota_overhead(*this, m_statistics.download_ip_overhead() , m_statistics.upload_ip_overhead()); warning |= m_ses.use_quota_overhead(*t, m_statistics.download_ip_overhead() , m_statistics.upload_ip_overhead()); } if (warning && t->alerts().should_post()) { for (int channel = 0; channel < 2; ++channel) { if ((warning & (1 << channel)) == 0) continue; t->alerts().post_alert(performance_alert(t->get_handle() , channel == peer_connection::download_channel ? performance_alert::download_limit_too_low : performance_alert::upload_limit_too_low)); } } if (!t || m_disconnecting) { TORRENT_ASSERT(t || !m_connecting); if (m_connecting) { m_counters.inc_stats_counter(counters::num_peers_half_open, -1); if (t) t->dec_num_connecting(); m_connecting = false; } disconnect(errors::torrent_aborted, op_bittorrent); return; } if (m_endgame_mode && m_interesting && m_download_queue.empty() && m_request_queue.empty() && now - seconds(5) >= m_last_request) { // this happens when we're in strict end-game // mode and the peer could not request any blocks // because they were all taken but there were still // unrequested blocks. Now, 5 seconds later, there // might not be any unrequested blocks anymore, so // we should try to pick another block to see // if we can pick a busy one m_last_request = now; if (request_a_block(*t, *this)) m_counters.inc_stats_counter(counters::end_game_piece_picks); if (m_disconnecting) return; send_block_requests(); } if (t->super_seeding() && !m_peer_interested && m_became_uninterested + seconds(10) < now) { // maybe we need to try another piece, to see if the peer // become interested in us then superseed_piece(-1, t->get_piece_to_super_seed(m_have_piece)); } on_tick(); if (is_disconnecting()) return; #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { (*i)->tick(); } if (is_disconnecting()) return; #endif // if the peer hasn't said a thing for a certain // time, it is considered to have timed out time_duration d; d = (std::min)(now - m_last_receive, now - m_last_sent); if (m_connecting) { int connect_timeout = m_settings.get_int(settings_pack::peer_connect_timeout); if (m_peer_info) connect_timeout += 3 * m_peer_info->failcount; // SSL and i2p handshakes are slow if (is_ssl(*m_socket)) connect_timeout += 10; #if TORRENT_USE_I2P if (is_i2p(*m_socket)) connect_timeout += 20; #endif if (d > seconds(connect_timeout) && can_disconnect(error_code(errors::timed_out, get_libtorrent_category()))) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** CONNECT FAILED [ waited %d seconds ] ***", int(total_seconds(d))); #endif connect_failed(errors::timed_out); return; } } // if we can't read, it means we're blocked on the rate-limiter // or the disk, not the peer itself. In this case, don't blame // the peer and disconnect it bool may_timeout = (m_channel_state[download_channel] & peer_info::bw_network) != 0; // TODO: 2 use a deadline_timer for timeouts. Don't rely on second_tick()! // Hook this up to connect timeout as well. This would improve performance // because of less work in second_tick(), and might let use remove ticking // entirely eventually if (may_timeout && d > seconds(timeout()) && !m_connecting && m_reading_bytes == 0 && can_disconnect(error_code(errors::timed_out_inactivity, get_libtorrent_category()))) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** LAST ACTIVITY [ %d seconds ago ] ***", int(total_seconds(d))); #endif disconnect(errors::timed_out_inactivity, op_bittorrent); return; } // do not stall waiting for a handshake if (may_timeout && !m_connecting && in_handshake() && d > seconds(m_settings.get_int(settings_pack::handshake_timeout))) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** NO HANDSHAKE [ waited %d seconds ] ***", int(total_seconds(d))); #endif disconnect(errors::timed_out_no_handshake, op_bittorrent); return; } // disconnect peers that we unchoked, but // they didn't send a request within 60 seconds. // but only if we're a seed d = now - (std::max)(m_last_unchoke, m_last_incoming_request); if (may_timeout && !m_connecting && m_requests.empty() && m_reading_bytes == 0 && !m_choked && m_peer_interested && t && t->is_upload_only() && d > seconds(60) && can_disconnect(error_code(errors::timed_out_no_request, get_libtorrent_category()))) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** NO REQUEST [ waited %d seconds ] ***", int(total_seconds(d))); #endif disconnect(errors::timed_out_no_request, op_bittorrent); return; } // if the peer hasn't become interested and we haven't // become interested in the peer for 10 minutes, it // has also timed out. time_duration d1; time_duration d2; d1 = now - m_became_uninterested; d2 = now - m_became_uninteresting; time_duration time_limit = seconds( m_settings.get_int(settings_pack::inactivity_timeout)); // don't bother disconnect peers we haven't been interested // in (and that hasn't been interested in us) for a while // unless we have used up all our connection slots if (may_timeout && !m_interesting && !m_peer_interested && d1 > time_limit && d2 > time_limit && (m_ses.num_connections() >= m_settings.get_int(settings_pack::connections_limit) || (t && t->num_peers() >= t->max_connections())) && can_disconnect(error_code(errors::timed_out_no_interest, get_libtorrent_category()))) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** MUTUAL NO INTEREST [ t1: %d t2: %d ]" , total_seconds(d1), total_seconds(d2)); #endif disconnect(errors::timed_out_no_interest, op_bittorrent); return; } // TODO: 3 instead of using settings_pack::request_timeout, use // m_rtt.mean() + m_rtt.avg_deviation() * 2 or something like that. // the configuration option could hopefully be removed if (may_timeout && !m_download_queue.empty() && m_quota[download_channel] > 0 && now > m_requested + seconds(m_settings.get_int(settings_pack::request_timeout) + m_timeout_extend)) { snub_peer(); } // if we haven't sent something in too long, send a keep-alive keep_alive(); m_statistics.second_tick(tick_interval_ms); if (m_statistics.upload_payload_rate() > m_upload_rate_peak) { m_upload_rate_peak = m_statistics.upload_payload_rate(); } if (m_statistics.download_payload_rate() > m_download_rate_peak) { m_download_rate_peak = m_statistics.download_payload_rate(); } if (is_disconnecting()) return; if (!t->ready_for_connections()) return; update_desired_queue_size(); if (m_desired_queue_size == m_max_out_request_queue && t->alerts().should_post()) { t->alerts().post_alert(performance_alert(t->get_handle() , performance_alert::outstanding_request_limit_reached)); } int piece_timeout = m_settings.get_int(settings_pack::piece_timeout); if (!m_download_queue.empty() && m_quota[download_channel] > 0 && now - m_last_piece > seconds(piece_timeout + m_timeout_extend)) { // this peer isn't sending the pieces we've // requested (this has been observed by BitComet) // in this case we'll clear our download queue and // re-request the blocks. #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** PIECE_REQUEST TIMED OUT [ %d time: %d to: %d extend: %d ]" , (int)m_download_queue.size(), total_seconds(now - m_last_piece) , piece_timeout, m_timeout_extend); #endif snub_peer(); } // update once every minute if (now - m_remote_dl_update >= seconds(60)) { boost::int64_t piece_size = t->torrent_file().piece_length(); if (m_remote_dl_rate > 0) m_remote_dl_rate = int((m_remote_dl_rate * 2 / 3) + ((boost::int64_t(m_remote_pieces_dled) * piece_size / 3) / 60)); else m_remote_dl_rate = int(boost::int64_t(m_remote_pieces_dled) * piece_size / 60); m_remote_pieces_dled = 0; m_remote_dl_update = now; } fill_send_buffer(); } void peer_connection::snub_peer() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); if (!m_snubbed) { m_snubbed = true; if (t->alerts().should_post()) { t->alerts().post_alert(peer_snubbed_alert(t->get_handle() , m_remote, m_peer_id)); } } m_desired_queue_size = 1; if (on_parole()) { m_timeout_extend += m_settings.get_int(settings_pack::request_timeout); return; } if (!t->has_picker()) return; piece_picker& picker = t->picker(); // first, if we have any unsent requests, just // wipe those out while (!m_request_queue.empty()) { t->picker().abort_download(m_request_queue.back().block, peer_info_struct()); m_request_queue.pop_back(); } m_queued_time_critical = 0; TORRENT_ASSERT(!m_download_queue.empty()); // request a new block before removing the previous // one, in order to prevent it from // picking the same block again, stalling the // same piece indefinitely. m_desired_queue_size = 2; if (request_a_block(*t, *this)) m_counters.inc_stats_counter(counters::snubbed_piece_picks); // the block we just picked (potentially) // hasn't been put in m_download_queue yet. // it's in m_request_queue and will be sent // once send_block_requests() is called. m_desired_queue_size = 1; // time out the last request eligible // block in the queue int i = m_download_queue.size() - 1; for (; i >= 0; --i) { if (!m_download_queue[i].timed_out && !m_download_queue[i].not_wanted) break; } if (i >= 0) { pending_block& qe = m_download_queue[i]; piece_block r = qe.block; // only time out a request if it blocks the piece // from being completed (i.e. no free blocks to // request from it) piece_picker::downloading_piece p; picker.piece_info(qe.block.piece_index, p); int free_blocks = picker.blocks_in_piece(qe.block.piece_index) - p.finished - p.writing - p.requested; if (free_blocks > 0) { m_timeout_extend += m_settings.get_int(settings_pack::request_timeout); send_block_requests(); return; } if (t->alerts().should_post()) { t->alerts().post_alert(block_timeout_alert(t->get_handle() , remote(), pid(), qe.block.block_index, qe.block.piece_index)); } qe.timed_out = true; picker.abort_download(r, peer_info_struct()); } send_block_requests(); } int peer_connection::preferred_caching() const { TORRENT_ASSERT(is_single_thread()); int line_size = 0; if (m_settings.get_bool(settings_pack::guided_read_cache)) { boost::shared_ptr t = m_torrent.lock(); int upload_rate = m_statistics.upload_payload_rate(); if (upload_rate == 0) upload_rate = 1; int num_uploads = m_ses.num_uploads(); if (num_uploads == 0) num_uploads = 1; // assume half of the cache is write cache if we're downloading // this torrent as well int cache_size = m_settings.get_int(settings_pack::cache_size) / num_uploads; if (!t->is_upload_only()) cache_size /= 2; // cache_size is the amount of cache we have per peer. The // cache line should not be greater than this line_size = cache_size; } return line_size; } void peer_connection::fill_send_buffer() { TORRENT_ASSERT(is_single_thread()); #ifdef TORRENT_EXPENSIVE_INVARIANT_CHECKS INVARIANT_CHECK; #endif bool sent_a_piece = false; boost::shared_ptr t = m_torrent.lock(); if (!t || t->is_aborted()) return; // only add new piece-chunks if the send buffer is small enough // otherwise there will be no end to how large it will be! boost::uint64_t upload_rate = int(m_statistics.upload_rate()); int buffer_size_watermark = int(upload_rate * m_settings.get_int(settings_pack::send_buffer_watermark_factor) / 100); if (buffer_size_watermark < m_settings.get_int(settings_pack::send_buffer_low_watermark)) { buffer_size_watermark = m_settings.get_int(settings_pack::send_buffer_low_watermark); } else if (buffer_size_watermark > m_settings.get_int(settings_pack::send_buffer_watermark)) { buffer_size_watermark = m_settings.get_int(settings_pack::send_buffer_watermark); } // don't just pop the front element here, since in seed mode one request may // be blocked because we have to verify the hash first, so keep going with the // next request. However, only let each peer have one hash verification outstanding // at any given time for (int i = 0; i < m_requests.size() && (send_buffer_size() + m_reading_bytes < buffer_size_watermark); ++i) { TORRENT_ASSERT(t->ready_for_connections()); peer_request& r = m_requests[i]; TORRENT_ASSERT(r.piece >= 0); TORRENT_ASSERT(r.piece < (int)m_have_piece.size()); // TORRENT_ASSERT(t->have_piece(r.piece)); TORRENT_ASSERT(r.start + r.length <= t->torrent_file().piece_size(r.piece)); TORRENT_ASSERT(r.length > 0 && r.start >= 0); if (t->is_deleted()) { #if defined TORRENT_VERBOSE_LOGGING peer_log("==> REJECT_PIECE [ piece: %d s: %x l: %x ] torrent deleted" , r.piece , r.start , r.length); #endif write_reject_request(r); continue; } if (t->seed_mode() && !t->verified_piece(r.piece)) { // we're still verifying the hash of this piece // so we can't return it yet. if (t->verifying_piece(r.piece)) continue; // only have three outstanding hash check per peer if (m_outstanding_piece_verification >= 3) continue; ++m_outstanding_piece_verification; #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** SEED-MODE FILE ASYNC HASH [ piece: %d ]", r.piece); #endif // this means we're in seed mode and we haven't yet // verified this piece (r.piece) if (!t->need_loaded()) return; t->inc_refcount("async_seed_hash"); m_disk_thread.async_hash(&t->storage(), r.piece, 0 , boost::bind(&peer_connection::on_seed_mode_hashed, self(), _1) , this); t->verifying(r.piece); continue; } // in seed mode, we might end up accepting a request // which it later turns out we cannot serve, if we ended // up not having that piece if (!t->has_piece_passed(r.piece)) { // we don't have this piece yet, but we anticipate to have // it very soon, so we have told our peers we have it. // hold off on sending it. If the piece fails later // we will reject this request if (t->is_predictive_piece(r.piece)) continue; #if defined TORRENT_VERBOSE_LOGGING peer_log("==> REJECT_PIECE [ piece: %d s: %x l: %x ] piece not passed hash check" , r.piece , r.start , r.length); #endif write_reject_request(r); } else { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** FILE ASYNC READ [ piece: %d | s: %x | l: %x ]" , r.piece, r.start, r.length); #endif m_reading_bytes += r.length; sent_a_piece = true; // the callback function may be called immediately, instead of being posted if (!t->need_loaded()) return; t->inc_refcount("async_read"); m_disk_thread.async_read(&t->storage(), r , boost::bind(&peer_connection::on_disk_read_complete , self(), _1, r, time_now_hires()), this); } m_requests.erase(m_requests.begin() + i); if (m_requests.empty()) m_counters.inc_stats_counter(counters::num_peers_up_requests, -1); --i; } if (t->share_mode() && sent_a_piece) t->recalc_share_mode(); } // this is called when a previously unchecked piece has been // checked, while in seed-mode void peer_connection::on_seed_mode_hashed(disk_io_job const* j) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); torrent_ref_holder h(t.get(), "async_seed_hash"); if (t) t->dec_refcount("async_seed_hash"); TORRENT_ASSERT(m_outstanding_piece_verification > 0); --m_outstanding_piece_verification; if (!t || t->is_aborted()) return; if (j->error) { t->handle_disk_error(j, this); t->leave_seed_mode(false); return; } // we're using the piece hashes here, we need the torrent to be loaded if (!t->need_loaded()) return; if (!m_settings.get_bool(settings_pack::disable_hash_checks) && sha1_hash(j->d.piece_hash) != t->torrent_file().hash_for_piece(j->piece)) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** SEED-MODE FILE HASH [ piece: %d failed ]", j->piece); #endif t->leave_seed_mode(false); } else { TORRENT_ASSERT(t->verifying_piece(j->piece)); if (t->seed_mode()) t->verified(j->piece); #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** SEED-MODE FILE HASH [ piece: %d passed ]", j->piece); #endif if (t) { if (t->seed_mode() && t->all_verified()) t->leave_seed_mode(true); } } // try to service the requests again, now that the piece // has been verified fill_send_buffer(); } void peer_connection::on_disk_read_complete(disk_io_job const* j , peer_request r, ptime issue_time) { TORRENT_ASSERT(is_single_thread()); // return value: // 0: success, piece passed hash check // -1: disk failure int disk_rtt = int(total_microseconds(time_now_hires() - issue_time)); #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** FILE ASYNC READ COMPLETE [ ret: %d | piece: %d | s: %x | l: %x" " | b: %p | c: %s | e: %s | rtt: %d us ]" , j->ret, r.piece, r.start, r.length, j->buffer , (j->flags & disk_io_job::cache_hit ? "cache hit" : "cache miss") , j->error.ec.message().c_str(), disk_rtt); #endif m_reading_bytes -= r.length; boost::shared_ptr t = m_torrent.lock(); torrent_ref_holder h(t.get(), "async_read"); if (t) t->dec_refcount("async_read"); if (j->ret < 0) { if (!t) { disconnect(j->error.ec, op_file_read); return; } TORRENT_ASSERT(j->buffer == 0); write_dont_have(r.piece); write_reject_request(r); if (t->alerts().should_post()) t->alerts().post_alert(file_error_alert(j->error.ec , t->resolve_filename(j->error.file) , j->error.operation_str(), t->get_handle())); ++m_disk_read_failures; if (m_disk_read_failures > 100) disconnect(j->error.ec, op_file_read); return; } // we're only interested in failures in a row. // if we every now and then successfully send a // block, the peer is still useful m_disk_read_failures = 0; TORRENT_ASSERT(j->ret == r.length); // even if we're disconnecting, we need to free this block // otherwise the disk thread will hang, waiting for the network // thread to be done with it disk_buffer_holder buffer(m_allocator, *j); if (m_disconnecting) return; // flush send buffer at the end of // this burst of disk events // m_ses.cork_burst(this); if (!t) { disconnect(j->error.ec, op_file_read); return; } if (j->ret != r.length) { // handle_disk_error may disconnect us t->handle_disk_error(j, this); return; } #if defined TORRENT_VERBOSE_LOGGING peer_log("==> PIECE [ piece: %d s: %x l: %x ]" , r.piece, r.start, r.length); #endif m_counters.blend_stats_counter(counters::request_latency, disk_rtt, 5); // we probably just pulled this piece into the cache. // if it's rare enough to make it into the suggested piece // push another piece out if (m_settings.get_int(settings_pack::suggest_mode) == settings_pack::suggest_read_cache && (j->flags & disk_io_job::cache_hit) == 0) { t->add_suggest_piece(r.piece); } write_piece(r, buffer); } void peer_connection::assign_bandwidth(int channel, int amount) { TORRENT_ASSERT(is_single_thread()); #ifdef TORRENT_VERBOSE_LOGGING peer_log("%s ASSIGN BANDWIDHT [ bytes: %d ]" , channel == upload_channel ? ">>>" : "<<<", amount); #endif TORRENT_ASSERT(amount > 0 || is_disconnecting()); m_quota[channel] += amount; TORRENT_ASSERT(m_channel_state[channel] & peer_info::bw_limit); m_channel_state[channel] &= ~peer_info::bw_limit; #if TORRENT_USE_INVARIANT_CHECKS check_invariant(); #endif if (is_disconnecting()) return; if (channel == upload_channel) { setup_send(); } else if (channel == download_channel) { setup_receive(); } } // the number of bytes we expect to receive, or want to send // channel either refer to upload or download. This is used // by the rate limiter to allocate quota for this peer int peer_connection::wanted_transfer(int channel) { TORRENT_ASSERT(is_single_thread()); shared_ptr t = m_torrent.lock(); if (channel == download_channel) { return (std::max)((std::max)(m_outstanding_bytes , m_packet_size - m_recv_pos) + 30 , int(boost::int64_t(m_statistics.download_rate()) * 2 / (1000 / m_settings.get_int(settings_pack::tick_interval)))); } else { return (std::max)((std::max)(m_reading_bytes , m_send_buffer.size()) , int((boost::int64_t(m_statistics.upload_rate()) * 2 * m_settings.get_int(settings_pack::tick_interval)) / 1000)); } } int peer_connection::request_bandwidth(int channel, int bytes) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; // we can only have one outstanding bandwidth request at a time if (m_channel_state[channel] & peer_info::bw_limit) return 0; shared_ptr t = m_torrent.lock(); bytes = (std::max)(wanted_transfer(channel), bytes); // we already have enough quota if (m_quota[channel] >= bytes) return 0; // deduct the bytes we already have quota for bytes -= m_quota[channel]; int priority = get_priority(channel); int max_channels = num_classes() + (t ? t->num_classes() : 0) + 2; bandwidth_channel** channels = TORRENT_ALLOCA(bandwidth_channel*, max_channels); // collect the pointers to all bandwidth channels // that apply to this torrent int c = 0; c += m_ses.copy_pertinent_channels(*this, channel , channels + c, max_channels - c); if (t) { c += m_ses.copy_pertinent_channels(*t, channel , channels + c, max_channels - c); } #ifdef TORRENT_DEBUG // make sure we don't have duplicates std::set unique_classes; for (int i = 0; i < c; ++i) { TORRENT_ASSERT(unique_classes.count(channels[i]) == 0); unique_classes.insert(channels[i]); } #endif #ifdef TORRENT_VERBOSE_LOGGING peer_log("%s REQUEST_BANDWIDTH [ bytes: %d quota: %d wanted_transfer: %d " "prio: %d num_channels: %d ]" , channel == download_channel ? "<<<" : ">>>", bytes , m_quota[channel], wanted_transfer(channel), priority, c); #endif TORRENT_ASSERT((m_channel_state[channel] & peer_info::bw_limit) == 0); bandwidth_manager* manager = m_ses.get_bandwidth_manager(channel); int ret = manager->request_bandwidth(self() , bytes, priority, channels, c); if (ret == 0) m_channel_state[channel] |= peer_info::bw_limit; else { #ifdef TORRENT_VERBOSE_LOGGING peer_log("%s ASSIGN BANDWIDTH [ bytes: %d ]" , channel == download_channel ? "<<<" : ">>>", ret); #endif m_quota[channel] += ret; } return ret; } void peer_connection::uncork_socket() { TORRENT_ASSERT(is_single_thread()); if (!m_corked) return; m_corked = false; setup_send(); } void peer_connection::setup_send() { TORRENT_ASSERT(is_single_thread()); if (m_disconnecting) return; // we may want to request more quota at this point request_bandwidth(upload_channel); if (m_channel_state[upload_channel] & peer_info::bw_network) return; if (m_quota[upload_channel] == 0 && !m_send_buffer.empty() && !m_connecting) { return; } int quota_left = m_quota[upload_channel]; if (m_send_buffer.empty() && m_reading_bytes > 0 && quota_left > 0) { if ((m_channel_state[upload_channel] & peer_info::bw_disk) == 0) m_counters.inc_stats_counter(counters::num_peers_up_disk); m_channel_state[upload_channel] |= peer_info::bw_disk; #ifdef TORRENT_VERBOSE_LOGGING peer_log(">>> waiting for disk [outstanding: %d]", m_reading_bytes); #endif if (!m_connecting && !m_requests.empty() && m_reading_bytes > m_settings.get_int(settings_pack::send_buffer_watermark) - 0x4000) { boost::shared_ptr t = m_torrent.lock(); // we're stalled on the disk. We want to write and we can write // but our send buffer is empty, waiting to be refilled from the disk // this either means the disk is slower than the network connection // or that our send buffer watermark is too small, because we can // send it all before the disk gets back to us. That's why we only // trigger this if we've also filled the allowed send buffer. The // first request would not fill it all the way up because of the // upload rate being virtually 0. If m_requests is empty, it doesn't // matter anyway, because we don't have any more requests from the // peer to hang on to the disk if (t && t->alerts().should_post()) { t->alerts().post_alert(performance_alert(t->get_handle() , performance_alert::send_buffer_watermark_too_low)); } } } else { if (m_channel_state[upload_channel] & peer_info::bw_disk) m_counters.inc_stats_counter(counters::num_peers_up_disk, -1); m_channel_state[upload_channel] &= ~peer_info::bw_disk; } if (!can_write()) { #ifdef TORRENT_VERBOSE_LOGGING if (m_send_buffer.empty()) { peer_log(">>> SEND BUFFER DEPLETED [" " quota: %d buf: %d connecting: %s disconnecting: %s pending_disk: %d ]" , m_quota[upload_channel] , int(m_send_buffer.size()), m_connecting?"yes":"no" , m_disconnecting?"yes":"no", m_reading_bytes); } else { peer_log(">>> CANNOT WRITE [" " quota: %d buf: %d connecting: %s disconnecting: %s pending_disk: %d ]" , m_quota[upload_channel] , int(m_send_buffer.size()), m_connecting?"yes":"no" , m_disconnecting?"yes":"no", m_reading_bytes); } #endif return; } // send the actual buffer int amount_to_send = m_send_buffer.size(); if (amount_to_send > quota_left) amount_to_send = quota_left; TORRENT_ASSERT(amount_to_send > 0); if (m_corked) { #ifdef TORRENT_VERBOSE_LOGGING peer_log(">>> CORKED WRITE [ bytes: %d ]", amount_to_send); #endif return; } TORRENT_ASSERT((m_channel_state[upload_channel] & peer_info::bw_network) == 0); #ifdef TORRENT_VERBOSE_LOGGING peer_log(">>> ASYNC_WRITE [ bytes: %d ]", amount_to_send); #endif std::vector const& vec = m_send_buffer.build_iovec(amount_to_send); #if defined TORRENT_ASIO_DEBUGGING add_outstanding_async("peer_connection::on_send_data"); #endif #if TORRENT_USE_ASSERTS TORRENT_ASSERT(!m_socket_is_writing); m_socket_is_writing = true; #endif // uTP sockets aren't thread safe... if (is_utp(*m_socket)) { m_socket->async_write_some(vec, make_write_handler(boost::bind( &peer_connection::on_send_data, self(), _1, _2))); } else { socket_job j; j.type = socket_job::write_job; j.vec = &vec; j.peer = self(); m_ses.post_socket_job(j); } m_channel_state[upload_channel] |= peer_info::bw_network; } void peer_connection::on_disk() { TORRENT_ASSERT(is_single_thread()); if ((m_channel_state[download_channel] & peer_info::bw_disk) == 0) return; boost::shared_ptr me(self()); #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** dropped below disk buffer watermark"); #endif m_counters.inc_stats_counter(counters::num_peers_down_disk, -1); m_channel_state[download_channel] &= ~peer_info::bw_disk; setup_receive(read_async); } void peer_connection::on_allocate_disk_buffer(char* buffer, int buffer_size) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; TORRENT_ASSERT(!m_disk_recv_buffer); TORRENT_ASSERT(m_channel_state[download_channel] & peer_info::bw_disk); m_disk_recv_buffer.reset(buffer); m_disk_recv_buffer_size = buffer_size; m_counters.inc_stats_counter(counters::num_peers_down_disk, -1); m_channel_state[download_channel] &= ~peer_info::bw_disk; setup_receive(read_async); } void peer_connection::setup_receive(sync_t sync) { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; if (m_disconnecting) return; // we may want to request more quota at this point request_bandwidth(download_channel); if (m_channel_state[download_channel] & peer_info::bw_network) return; if (m_quota[download_channel] == 0 && !m_connecting) { return; } if (!can_read()) { #ifdef TORRENT_VERBOSE_LOGGING peer_log("<<< CANNOT READ [ quota: %d " "can-write-to-disk: %s queue-limit: %d disconnecting: %s " " connecting: %s ]" , m_quota[download_channel] , ((m_channel_state[download_channel] & peer_info::bw_disk)?"no":"yes") , m_settings.get_int(settings_pack::max_queued_disk_bytes) , (m_disconnecting?"yes":"no") , (m_connecting?"yes":"no")); #endif // if we block reading, waiting for the disk, we will wake up // by the disk_io_thread posting a message every time it drops // from being at or exceeding the limit down to below the limit return; } error_code ec; try_read(read_async, ec); } size_t peer_connection::try_read(sync_t s, error_code& ec) { TORRENT_ASSERT(is_single_thread()); TORRENT_ASSERT(m_connected); if (m_quota[download_channel] == 0) { ec = asio::error::would_block; return 0; } if (!can_read()) { ec = asio::error::would_block; return 0; } int max_receive = m_packet_size - m_recv_pos; boost::array vec; int num_bufs = 0; // only apply the contiguous receive buffer when we don't have any // outstanding requests. When we're likely to receive pieces, we'll // save more time from avoiding copying data from the socket if ((m_settings.get_bool(settings_pack::contiguous_recv_buffer) || m_download_queue.empty()) && !m_disk_recv_buffer) { if (s == read_sync) { ec = asio::error::would_block; return 0; } TORRENT_ASSERT((m_channel_state[download_channel] & peer_info::bw_network) == 0); m_channel_state[download_channel] |= peer_info::bw_network; #ifdef TORRENT_VERBOSE_LOGGING peer_log("<<< ASYNC_READ [ ]"); #endif #if defined TORRENT_ASIO_DEBUGGING add_outstanding_async("peer_connection::on_receive_data_nb"); #endif m_socket->async_read_some(asio::null_buffers(), make_read_handler( boost::bind(&peer_connection::on_receive_data_nb, self(), _1, _2))); return 0; } TORRENT_ASSERT(m_packet_size > 0); TORRENT_ASSERT(max_receive >= 0); if (m_recv_pos >= m_soft_packet_size) m_soft_packet_size = 0; if (m_soft_packet_size && max_receive > m_soft_packet_size - m_recv_pos) max_receive = m_soft_packet_size - m_recv_pos; int quota_left = m_quota[download_channel]; if (max_receive > quota_left) max_receive = quota_left; if (max_receive == 0) { ec = asio::error::would_block; return 0; } TORRENT_ASSERT(m_recv_pos >= 0); TORRENT_ASSERT(m_packet_size > 0); int regular_buffer_size = m_packet_size - m_disk_recv_buffer_size; if (int(m_recv_buffer.size()) < regular_buffer_size) m_recv_buffer.resize(round_up8(regular_buffer_size)); if (!m_disk_recv_buffer || regular_buffer_size >= m_recv_pos + max_receive) { // only receive into regular buffer TORRENT_ASSERT(m_recv_pos + max_receive <= int(m_recv_buffer.size())); vec[0] = asio::buffer(&m_recv_buffer[m_recv_pos], max_receive); num_bufs = 1; } else if (m_recv_pos >= regular_buffer_size) { // only receive into disk buffer TORRENT_ASSERT(m_recv_pos - regular_buffer_size >= 0); TORRENT_ASSERT(m_recv_pos - regular_buffer_size + max_receive <= m_disk_recv_buffer_size); vec[0] = asio::buffer(m_disk_recv_buffer.get() + m_recv_pos - regular_buffer_size, max_receive); num_bufs = 1; } else { // receive into both regular and disk buffer TORRENT_ASSERT(max_receive + m_recv_pos > regular_buffer_size); TORRENT_ASSERT(m_recv_pos < regular_buffer_size); TORRENT_ASSERT(max_receive - regular_buffer_size + m_recv_pos <= m_disk_recv_buffer_size); vec[0] = asio::buffer(&m_recv_buffer[m_recv_pos] , regular_buffer_size - m_recv_pos); vec[1] = asio::buffer(m_disk_recv_buffer.get() , max_receive - regular_buffer_size + m_recv_pos); num_bufs = 2; } if (s == read_async) { TORRENT_ASSERT((m_channel_state[download_channel] & peer_info::bw_network) == 0); m_channel_state[download_channel] |= peer_info::bw_network; #ifdef TORRENT_VERBOSE_LOGGING peer_log("<<< ASYNC_READ [ max: %d bytes ]", max_receive); #endif // utp sockets aren't thread safe... #if defined TORRENT_ASIO_DEBUGGING add_outstanding_async("peer_connection::on_receive_data"); #endif if (is_utp(*m_socket)) { if (num_bufs == 1) { m_socket->async_read_some( asio::mutable_buffers_1(vec[0]), make_read_handler( boost::bind(&peer_connection::on_receive_data, self(), _1, _2))); } else { m_socket->async_read_some( vec, make_read_handler( boost::bind(&peer_connection::on_receive_data, self(), _1, _2))); } } else { socket_job j; j.type = socket_job::read_job; j.peer = self(); if (num_bufs == 1) { j.recv_buf = asio::buffer_cast(vec[0]); j.buf_size = asio::buffer_size(vec[0]); } else { j.read_vec = vec; } m_ses.post_socket_job(j); } return 0; } size_t ret = 0; if (num_bufs == 1) { ret = m_socket->read_some(asio::mutable_buffers_1(vec[0]), ec); } else { ret = m_socket->read_some(vec, ec); } // this is weird. You would imagine read_some() would do this if (ret == 0 && !ec) ec = asio::error::eof; #ifdef TORRENT_VERBOSE_LOGGING peer_log("<<< SYNC_READ [ max: %d ret: %d e: %s ]", max_receive, ret, ec ? ec.message().c_str() : ""); #endif return ret; } #ifndef TORRENT_DISABLE_ENCRYPTION // returns the last 'bytes' from the receive buffer std::pair peer_connection::wr_recv_buffers(int bytes) { TORRENT_ASSERT(is_single_thread()); TORRENT_ASSERT(bytes <= m_recv_pos); std::pair vec; int regular_buffer_size = m_packet_size - m_disk_recv_buffer_size; TORRENT_ASSERT(regular_buffer_size >= 0); if (!m_disk_recv_buffer || regular_buffer_size >= m_recv_pos) { vec.first = buffer::interval(&m_recv_buffer[0] + m_recv_start + m_recv_pos - bytes, &m_recv_buffer[0] + m_recv_start + m_recv_pos); vec.second = buffer::interval(0,0); } else if (m_recv_pos - bytes >= regular_buffer_size) { vec.first = buffer::interval(m_disk_recv_buffer.get() + m_recv_pos - regular_buffer_size - bytes, m_disk_recv_buffer.get() + m_recv_pos - regular_buffer_size); vec.second = buffer::interval(0,0); } else { TORRENT_ASSERT(m_recv_pos - bytes < regular_buffer_size); TORRENT_ASSERT(m_recv_pos > regular_buffer_size); vec.first = buffer::interval(&m_recv_buffer[0] + m_recv_start + m_recv_pos - bytes , &m_recv_buffer[0] + m_recv_start + regular_buffer_size); vec.second = buffer::interval(m_disk_recv_buffer.get() , m_disk_recv_buffer.get() + m_recv_pos - regular_buffer_size); } TORRENT_ASSERT(vec.first.left() + vec.second.left() == bytes); return vec; } #endif void peer_connection::reset_recv_buffer(int packet_size) { TORRENT_ASSERT(is_single_thread()); TORRENT_ASSERT(m_recv_buffer.size() >= m_recv_end); TORRENT_ASSERT(packet_size > 0); if (m_recv_end > m_packet_size) { cut_receive_buffer(m_packet_size, packet_size); return; } m_recv_pos = 0; m_recv_start = 0; m_recv_end = 0; m_packet_size = packet_size; } void peer_connection::append_send_buffer(char* buffer, int size , chained_buffer::free_buffer_fun destructor, void* userdata , block_cache_reference ref, bool encrypted) { TORRENT_ASSERT(is_single_thread()); // bittorrent connections should never use this function, since // they might be encrypted and this would circumvent the actual // encryption. bt_peer_connection overrides this function with // its own version. TORRENT_ASSERT(encrypted || type() != bittorrent_connection); m_send_buffer.append_buffer(buffer, size, size, destructor , userdata, ref); } void peer_connection::append_const_send_buffer(char const* buffer, int size , chained_buffer::free_buffer_fun destructor, void* userdata , block_cache_reference ref) { TORRENT_ASSERT(is_single_thread()); m_send_buffer.append_buffer((char*)buffer, size, size, destructor , userdata, ref); } void session_free_buffer(char* buffer, void* userdata, block_cache_reference) { aux::session_interface* ses = (aux::session_interface*)userdata; ses->free_buffer(buffer); } void peer_connection::send_buffer(char const* buf, int size, int flags , void (*fun)(char*, int, void*), void* userdata) { TORRENT_ASSERT(is_single_thread()); int free_space = m_send_buffer.space_in_last_buffer(); if (free_space > size) free_space = size; if (free_space > 0) { char* dst = m_send_buffer.append(buf, free_space); TORRENT_ASSERT(dst != 0); if (fun) fun(dst, free_space, userdata); size -= free_space; buf += free_space; } if (size <= 0) return; int i = 0; while (size > 0) { char* chain_buf = m_ses.allocate_buffer(); if (chain_buf == 0) { disconnect(errors::no_memory, op_alloc_sndbuf); return; } const int alloc_buf_size = m_ses.send_buffer_size(); int buf_size = (std::min)(alloc_buf_size, size); memcpy(chain_buf, buf, buf_size); if (fun) fun(chain_buf, buf_size, userdata); buf += buf_size; size -= buf_size; m_send_buffer.append_buffer(chain_buf, alloc_buf_size, buf_size , &session_free_buffer, &m_ses); ++i; } setup_send(); } template struct set_to_zero { set_to_zero(T& v, bool cond): m_val(v), m_cond(cond) {} void fire() { if (!m_cond) return; m_cond = false; m_val = 0; } ~set_to_zero() { if (m_cond) m_val = 0; } private: T& m_val; bool m_cond; }; void peer_connection::on_receive_data_nb(const error_code& error , std::size_t bytes_transferred) { TORRENT_ASSERT(is_single_thread()); #if defined TORRENT_ASIO_DEBUGGING complete_async("peer_connection::on_receive_data_nb"); #endif // leave this bit set until we're done looping, reading from the socket. // that way we don't trigger any async read calls until the end of this // function. TORRENT_ASSERT(m_channel_state[download_channel] & peer_info::bw_network); // nb is short for null_buffers. In this mode we don't actually // allocate a receive buffer up-front, but get notified when // we can read from the socket, and then determine how much there // is to read. error_code ec; std::size_t buffer_size = m_socket->available(ec); if (ec) { disconnect(ec, op_available); return; } #ifdef TORRENT_VERBOSE_LOGGING peer_log("<<< READ_AVAILABLE [ bytes: %d ]", buffer_size); #endif // at this point the ioctl told us the socket doesn't have any // pending bytes. This probably means some error happened. // in order to find out though, we need to initiate a read // operation if (buffer_size == 0) { // try to read one byte. The socket is non-blocking anyway // so worst case, we'll fail with EWOULDBLOCK buffer_size = 1; } else { if (buffer_size > m_quota[download_channel]) { request_bandwidth(download_channel, buffer_size); buffer_size = m_quota[download_channel]; } // we're already waiting to get some more // quota from the bandwidth manager if (buffer_size == 0) { // allow reading from the socket again TORRENT_ASSERT(m_channel_state[download_channel] & peer_info::bw_network); m_channel_state[download_channel] &= ~peer_info::bw_network; return; } } if (buffer_size > 2097152) buffer_size = 2097152; m_recv_buffer.resize(m_recv_pos + buffer_size); TORRENT_ASSERT(m_recv_start == 0); // utp sockets aren't thread safe... if (is_utp(*m_socket)) { bytes_transferred = m_socket->read_some(asio::buffer(&m_recv_buffer[m_recv_pos] , buffer_size), ec); if (ec) { if (ec == boost::asio::error::try_again || ec == boost::asio::error::would_block) { // allow reading from the socket again TORRENT_ASSERT(m_channel_state[download_channel] & peer_info::bw_network); m_channel_state[download_channel] &= ~peer_info::bw_network; setup_receive(read_async); return; } disconnect(ec, op_sock_read); return; } } else { #if defined TORRENT_ASIO_DEBUGGING add_outstanding_async("peer_connection::on_receive_data"); #endif socket_job j; j.type = socket_job::read_job; j.recv_buf = &m_recv_buffer[m_recv_pos]; j.buf_size = buffer_size; j.peer = self(); m_ses.post_socket_job(j); return; } receive_data_impl(error, bytes_transferred, 0); } // -------------------------- // RECEIVE DATA // -------------------------- // nb is true if this callback is due to a null_buffers() // invocation of async_read_some(). In that case, we need // to disregard bytes_transferred. // at all exit points of this function, one of the following MUST hold: // 1. the socket is disconnecting // 2. m_channel_state[download_channel] & peer_info::bw_network == 0 void peer_connection::on_receive_data(const error_code& error , std::size_t bytes_transferred) { TORRENT_ASSERT(is_single_thread()); #if defined TORRENT_ASIO_DEBUGGING complete_async("peer_connection::on_receive_data"); #endif // leave this bit set until we're done looping, reading from the socket. // that way we don't trigger any async read calls until the end of this // function. TORRENT_ASSERT(m_channel_state[download_channel] & peer_info::bw_network); receive_data_impl(error, bytes_transferred, 10); } void peer_connection::receive_data_impl(const error_code& error , std::size_t bytes_transferred, int read_loops) { TORRENT_ASSERT(is_single_thread()); #ifdef TORRENT_VERBOSE_LOGGING peer_log("<<< ON_RECEIVE_DATA [ bytes: %d error: %s ]" , bytes_transferred, error.message().c_str()); #endif // submit all disk jobs later m_ses.deferred_submit_jobs(); // keep ourselves alive in until this function exits in // case we disconnect // this needs to be created before the invariant check, // to keep the object alive through the exit check boost::shared_ptr me(self()); // flush the send buffer at the end of this function cork _c(*this); INVARIANT_CHECK; int bytes_in_loop = bytes_transferred; if (error) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("*** ERROR [ in peer_connection::on_receive_data error: %s ]" , error.message().c_str()); #endif trancieve_ip_packet(bytes_in_loop, m_remote.address().is_v6()); on_receive(error, bytes_transferred); disconnect(error, op_sock_read); return; } m_counters.inc_stats_counter(counters::on_read_counter); m_ses.received_buffer(bytes_transferred); #ifdef TORRENT_VERBOSE_LOGGING peer_log("<<< ON_RECEIVE_DATA [ bytes: %d error: %s ]" , bytes_transferred, error.message().c_str()); #endif if (m_extension_outstanding_bytes > 0) m_extension_outstanding_bytes -= (std::min)(m_extension_outstanding_bytes, int(bytes_transferred)); int num_loops = 0; do { #ifdef TORRENT_VERBOSE_LOGGING peer_log("<<< read %d bytes", int(bytes_transferred)); #endif // correct the dl quota usage, if not all of the buffer was actually read TORRENT_ASSERT(int(bytes_transferred) <= m_quota[download_channel]); m_quota[download_channel] -= bytes_transferred; if (m_disconnecting) { trancieve_ip_packet(bytes_in_loop, m_remote.address().is_v6()); return; } TORRENT_ASSERT(m_packet_size > 0); TORRENT_ASSERT(bytes_transferred > 0); m_recv_end += bytes_transferred; TORRENT_ASSERT(m_recv_pos <= int(m_recv_buffer.size() + m_disk_recv_buffer_size)); int bytes = bytes_transferred; int sub_transferred = 0; do { INVARIANT_CHECK; #if TORRENT_USE_ASSERTS size_type cur_payload_dl = m_statistics.last_payload_downloaded(); size_type cur_protocol_dl = m_statistics.last_protocol_downloaded(); #endif int packet_size = m_soft_packet_size ? m_soft_packet_size : m_packet_size; int limit = packet_size > m_recv_pos ? packet_size - m_recv_pos : packet_size; sub_transferred = (std::min)(bytes, limit); m_recv_pos += sub_transferred; on_receive(error, sub_transferred); bytes -= sub_transferred; TORRENT_ASSERT(sub_transferred > 0); #if TORRENT_USE_ASSERTS TORRENT_ASSERT(m_statistics.last_payload_downloaded() - cur_payload_dl >= 0); TORRENT_ASSERT(m_statistics.last_protocol_downloaded() - cur_protocol_dl >= 0); size_type stats_diff = m_statistics.last_payload_downloaded() - cur_payload_dl + m_statistics.last_protocol_downloaded() - cur_protocol_dl; TORRENT_ASSERT(stats_diff == int(sub_transferred)); #endif if (m_disconnecting) return; } while (bytes > 0 && sub_transferred > 0); normalize_receive_buffer(); TORRENT_ASSERT(m_recv_pos == m_recv_end); TORRENT_ASSERT(m_packet_size > 0); if (m_peer_choked && m_recv_pos == 0 && (m_recv_buffer.capacity() - m_packet_size) > 128) { // round up to an even 8 bytes since that's the RC4 blocksize buffer(round_up8(m_packet_size)).swap(m_recv_buffer); } if (m_recv_pos >= m_soft_packet_size) m_soft_packet_size = 0; if (num_loops > read_loops) break; error_code ec; bytes_transferred = try_read(read_sync, ec); TORRENT_ASSERT(bytes_transferred > 0 || ec); if (ec == asio::error::would_block || ec == asio::error::try_again) break; if (ec) { trancieve_ip_packet(bytes_in_loop, m_remote.address().is_v6()); disconnect(ec, op_sock_read); return; } bytes_in_loop += bytes_transferred; ++num_loops; } while (bytes_transferred > 0); m_last_receive = time_now(); if (is_seed()) { boost::shared_ptr t = m_torrent.lock(); if (t) t->seen_complete(); } trancieve_ip_packet(bytes_in_loop, m_remote.address().is_v6()); // allow reading from the socket again TORRENT_ASSERT(m_channel_state[download_channel] & peer_info::bw_network); m_channel_state[download_channel] &= ~peer_info::bw_network; setup_receive(read_async); } bool peer_connection::can_write() const { TORRENT_ASSERT(is_single_thread()); // if we have requests or pending data to be sent or announcements to be made // we want to send data return !m_send_buffer.empty() && m_quota[upload_channel] > 0 && !m_connecting; } bool peer_connection::can_read() { TORRENT_ASSERT(is_single_thread()); INVARIANT_CHECK; boost::shared_ptr t = m_torrent.lock(); bool bw_limit = m_quota[download_channel] > 0; if (!bw_limit) return false; if (m_outstanding_bytes > 0) { // if we're expecting to download piece data, we might not // want to read from the socket in case we're out of disk // cache space right now if (m_channel_state[download_channel] & peer_info::bw_disk) return false; /* // if we already have a disk buffer, we might as well use it // if contiguous recv buffer is true, don't apply this logic, but // actually wait until we try to allocate a buffer and exceed the limit if (m_disk_recv_buffer == NULL && !m_settings.get_bool(settings_pack::contiguous_recv_buffer)) { m_disk_recv_buffer.reset(m_ses.async_allocate_disk_buffer("receive buffer", boost::bind(&peer_connection::on_allocate_disk_buffer, self(), _1, #error buffer_size))); if (m_disk_recv_buffer == NULL) { m_counters.inc_stats_counter(counters::num_peers_down_disk); const_cast(this)->m_channel_state[download_channel] |= peer_info::bw_disk; #ifdef TORRENT_VERBOSE_LOGGING peer_log("*** exceeded disk buffer watermark"); #endif return false; } } */ } return !m_connecting && !m_disconnecting; } void peer_connection::on_connection_complete(error_code const& e) { TORRENT_ASSERT(is_single_thread()); #if defined TORRENT_ASIO_DEBUGGING complete_async("peer_connection::on_connection_complete"); #endif ptime completed = time_now_hires(); INVARIANT_CHECK; m_rtt.add_sample(int(total_milliseconds(completed - m_connect))); #if defined TORRENT_LOGGING || defined TORRENT_ERROR_LOGGING { boost::shared_ptr t = m_torrent.lock(); t->debug_log("END connect [%p] (RTT: %d ms)", this, m_rtt.mean()); m_connect_time = completed; } #endif #ifdef TORRENT_USE_OPENSSL // add this RTT to the PRNG seed, to add more unpredictability boost::uint64_t now = total_microseconds(completed - m_connect); // assume 12 bits of entropy (i.e. about 8 milliseconds) RAND_add(&now, 8, 1.5); #endif // if t is NULL, we better not be connecting, since // we can't decrement the connecting counter boost::shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t || !m_connecting); if (m_connecting) { m_counters.inc_stats_counter(counters::num_peers_half_open, -1); if (t) t->dec_num_connecting(); m_connecting = false; } if (m_disconnecting) return; if (e) { connect_failed(e); return; } TORRENT_ASSERT(!m_connected); m_connected = true; m_counters.inc_stats_counter(counters::num_peers_connected); if (m_disconnecting) return; m_last_receive = time_now(); error_code ec; m_local = m_socket->local_endpoint(ec); if (ec) { disconnect(ec, op_getname); return; } // if there are outgoing interfaces specified, verify this // peer is correctly bound to on of them if (!m_settings.get_str(settings_pack::outgoing_interfaces).empty()) { if (!m_ses.verify_bound_address(m_local.address() , is_utp(*m_socket), ec)) { if (ec) { disconnect(ec, op_get_interface); return; } disconnect(error_code( boost::system::errc::no_such_device, generic_category()) , op_connect); return; } } if (is_utp(*m_socket) && m_peer_info) { m_peer_info->confirmed_supports_utp = true; m_peer_info->supports_utp = false; } // this means the connection just succeeded received_synack(m_remote.address().is_v6()); TORRENT_ASSERT(m_socket); #if defined TORRENT_VERBOSE_LOGGING peer_log(">>> COMPLETED [ ep: %s rtt: %d ]" , print_endpoint(m_remote).c_str(), m_rtt.mean()); #endif // set the socket to non-blocking, so that we can // read the entire buffer on each read event we get tcp::socket::non_blocking_io ioc(true); #if defined TORRENT_VERBOSE_LOGGING peer_log("*** SET NON-BLOCKING"); #endif m_socket->io_control(ioc, ec); if (ec) { disconnect(ec, op_iocontrol); return; } if (m_remote == m_socket->local_endpoint(ec)) { // if the remote endpoint is the same as the local endpoint, we're connected // to ourselves if (m_peer_info && t) t->ban_peer(m_peer_info); disconnect(errors::self_connection, op_bittorrent, 1); return; } if (m_remote.address().is_v4() && m_settings.get_int(settings_pack::peer_tos) != 0) { error_code ec; m_socket->set_option(type_of_service(m_settings.get_int(settings_pack::peer_tos)), ec); #if defined TORRENT_VERBOSE_LOGGING peer_log(">>> SET_TOS[ tos: %d e: %s ]", m_settings.get_int(settings_pack::peer_tos), ec.message().c_str()); #endif } #if TORRENT_USE_IPV6 && defined IPV6_TCLASS else if (m_remote.address().is_v6() && m_settings.get_int(settings_pack::peer_tos) != 0) { m_socket->set_option(traffic_class(m_settings.get_int(settings_pack::peer_tos)), ec); } #endif #ifndef TORRENT_DISABLE_EXTENSIONS for (extension_list_t::iterator i = m_extensions.begin() , end(m_extensions.end()); i != end; ++i) { (*i)->on_connected(); } #endif on_connected(); setup_send(); setup_receive(); } // -------------------------- // SEND DATA // -------------------------- void peer_connection::on_send_data(error_code const& error , std::size_t bytes_transferred) { TORRENT_ASSERT(is_single_thread()); m_counters.inc_stats_counter(counters::on_write_counter); m_ses.sent_buffer(bytes_transferred); #if TORRENT_USE_ASSERTS TORRENT_ASSERT(m_socket_is_writing); m_socket_is_writing = false; #endif // submit all disk jobs when we've processed all messages // in the current message queue m_ses.deferred_submit_jobs(); #if defined TORRENT_VERBOSE_LOGGING peer_log("*** ON_SEND_DATA [ bytes: %d error: %s ]" , int(bytes_transferred), error.message().c_str()); #endif INVARIANT_CHECK; #if defined TORRENT_ASIO_DEBUGGING complete_async("peer_connection::on_send_data"); #endif // keep ourselves alive in until this function exits in // case we disconnect boost::shared_ptr me(self()); TORRENT_ASSERT(m_channel_state[upload_channel] & peer_info::bw_network); m_send_buffer.pop_front(bytes_transferred); ptime now = time_now_hires(); for (std::vector::iterator i = m_download_queue.begin() , end(m_download_queue.end()); i != end; ++i) { if (i->send_buffer_offset < 0) continue; i->send_buffer_offset -= bytes_transferred; if (i->send_buffer_offset >= 0) continue; i->request_time = now; i->send_buffer_offset = -1; } m_channel_state[upload_channel] &= ~peer_info::bw_network; TORRENT_ASSERT(int(bytes_transferred) <= m_quota[upload_channel]); m_quota[upload_channel] -= bytes_transferred; trancieve_ip_packet(bytes_transferred, m_remote.address().is_v6()); #ifdef TORRENT_VERBOSE_LOGGING peer_log(">>> wrote %d bytes", int(bytes_transferred)); #endif if (error) { #if defined TORRENT_VERBOSE_LOGGING || defined TORRENT_ERROR_LOGGING peer_log("**ERROR**: %s [in peer_connection::on_send_data]", error.message().c_str()); #endif disconnect(error, op_sock_write); return; } if (m_disconnecting) { // make sure we free up all send buffers that are owned // by the disk thread m_send_buffer.clear(); m_disk_recv_buffer.reset(); return; } TORRENT_ASSERT(!m_connecting); TORRENT_ASSERT(bytes_transferred > 0); m_last_sent = now; #if TORRENT_USE_ASSERTS size_type cur_payload_ul = m_statistics.last_payload_uploaded(); size_type cur_protocol_ul = m_statistics.last_protocol_uploaded(); #endif on_sent(error, bytes_transferred); #if TORRENT_USE_ASSERTS TORRENT_ASSERT(m_statistics.last_payload_uploaded() - cur_payload_ul >= 0); TORRENT_ASSERT(m_statistics.last_protocol_uploaded() - cur_protocol_ul >= 0); size_type stats_diff = m_statistics.last_payload_uploaded() - cur_payload_ul + m_statistics.last_protocol_uploaded() - cur_protocol_ul; TORRENT_ASSERT(stats_diff == int(bytes_transferred)); #endif fill_send_buffer(); setup_send(); } #if TORRENT_USE_INVARIANT_CHECKS struct peer_count_t { peer_count_t(): num_peers(0), num_peers_with_timeouts(0), num_peers_with_nowant(0), num_not_requested(0) {} int num_peers; int num_peers_with_timeouts; int num_peers_with_nowant; int num_not_requested; // std::vector peers; }; void peer_connection::check_invariant() const { TORRENT_ASSERT(is_single_thread()); TORRENT_ASSERT(m_in_use == 1337); TORRENT_ASSERT(m_queued_time_critical <= int(m_request_queue.size())); TORRENT_ASSERT(m_recv_end >= m_recv_start); TORRENT_ASSERT(m_accept_fast.size() == m_accept_fast_piece_cnt.size()); TORRENT_ASSERT(bool(m_disk_recv_buffer) == (m_disk_recv_buffer_size > 0)); for (int i = 0; i < 2; ++i) { if (m_channel_state[i] & peer_info::bw_limit) { // if we're waiting for bandwidth, we should be in the // bandwidth manager's queue TORRENT_ASSERT(m_ses.get_bandwidth_manager(i)->is_queued(this)); } } boost::shared_ptr t = m_torrent.lock(); #if TORRENT_USE_INVARIANT_CHECKS \ && !defined TORRENT_NO_EXPENSIVE_INVARIANT_CHECK if (t && t->has_picker() && !m_disconnecting) t->picker().check_peer_invariant(m_have_piece, this); #endif if (!m_disconnect_started && m_initialized) { // none of this matters if we're disconnecting anyway if (t->is_finished()) TORRENT_ASSERT(!is_interesting() || m_need_interest_update); if (is_seed()) TORRENT_ASSERT(upload_only()); } if (m_disconnecting) { TORRENT_ASSERT(m_download_queue.empty()); TORRENT_ASSERT(m_request_queue.empty()); TORRENT_ASSERT(m_disconnect_started); } else if (!m_in_constructor) { TORRENT_ASSERT(m_ses.has_peer((peer_connection*)this)); } TORRENT_ASSERT(m_outstanding_bytes >= 0); if (t && t->valid_metadata() && !m_disconnecting) { torrent_info const& ti = t->torrent_file(); // if the piece is fully downloaded, we might have popped it from the // download queue already int outstanding_bytes = 0; // bool in_download_queue = false; int block_size = t->block_size(); piece_block last_block(ti.num_pieces()-1 , (ti.piece_size(ti.num_pieces()-1) + block_size - 1) / block_size); for (std::vector::const_iterator i = m_download_queue.begin() , end(m_download_queue.end()); i != end; ++i) { TORRENT_ASSERT(i->block.piece_index <= last_block.piece_index); TORRENT_ASSERT(i->block.piece_index < last_block.piece_index || i->block.block_index <= last_block.block_index); if (m_received_in_piece && i == m_download_queue.begin()) { // in_download_queue = true; // this assert is not correct since block may have different sizes // and may not be returned in the order they were requested // TORRENT_ASSERT(t->to_req(i->block).length >= m_received_in_piece); outstanding_bytes += t->to_req(i->block).length - m_received_in_piece; } else { outstanding_bytes += t->to_req(i->block).length; } } //if (p && p->bytes_downloaded < p->full_block_bytes) TORRENT_ASSERT(in_download_queue); TORRENT_ASSERT(m_outstanding_bytes == outstanding_bytes); } std::set unique; std::transform(m_download_queue.begin(), m_download_queue.end() , std::inserter(unique, unique.begin()), boost::bind(&pending_block::block, _1)); std::transform(m_request_queue.begin(), m_request_queue.end() , std::inserter(unique, unique.begin()), boost::bind(&pending_block::block, _1)); TORRENT_ASSERT(unique.size() == m_download_queue.size() + m_request_queue.size()); if (m_peer_info) { TORRENT_ASSERT(m_peer_info->prev_amount_upload == 0); TORRENT_ASSERT(m_peer_info->prev_amount_download == 0); TORRENT_ASSERT(m_peer_info->connection == this || m_peer_info->connection == 0); if (m_peer_info->optimistically_unchoked) TORRENT_ASSERT(!is_choked()); } TORRENT_ASSERT(m_have_piece.count() == m_num_pieces); if (!t) { #ifdef TORRENT_EXPENSIVE_INVARIANT_CHECKS // since this connection doesn't have a torrent reference // no torrent should have a reference to this connection either TORRENT_ASSERT(!m_ses.any_torrent_has_peer(this)); #endif return; } if (t->ready_for_connections() && m_initialized) TORRENT_ASSERT(t->torrent_file().num_pieces() == int(m_have_piece.size())); // in share mode we don't close redundant connections if (m_settings.get_bool(settings_pack::close_redundant_connections) && !t->share_mode()) { bool ok_to_disconnect = can_disconnect(error_code(errors::upload_upload_connection, get_libtorrent_category())) || can_disconnect(error_code(errors::uninteresting_upload_peer, get_libtorrent_category())) || can_disconnect(error_code(errors::too_many_requests_when_choked, get_libtorrent_category())) || can_disconnect(error_code(errors::timed_out_no_interest, get_libtorrent_category())) || can_disconnect(error_code(errors::timed_out_no_request, get_libtorrent_category())) || can_disconnect(error_code(errors::timed_out_inactivity, get_libtorrent_category())); // make sure upload only peers are disconnected if (t->is_upload_only() && m_upload_only && t->valid_metadata() && has_metadata() && ok_to_disconnect) TORRENT_ASSERT(m_disconnect_started || t->graceful_pause() || t->has_error()); if (m_upload_only && !m_interesting && m_bitfield_received && t->are_files_checked() && t->valid_metadata() && has_metadata() && ok_to_disconnect) TORRENT_ASSERT(m_disconnect_started); } if (!m_disconnect_started && m_initialized && m_settings.get_bool(settings_pack::close_redundant_connections)) { // none of this matters if we're disconnecting anyway if (t->is_upload_only() && !m_need_interest_update) TORRENT_ASSERT(!m_interesting || t->graceful_pause() || t->has_error()); if (is_seed()) TORRENT_ASSERT(m_upload_only); } #ifdef TORRENT_EXPENSIVE_INVARIANT_CHECKS if (t->has_picker()) { std::map num_requests; for (torrent::const_peer_iterator i = t->begin(); i != t->end(); ++i) { // make sure this peer is not a dangling pointer TORRENT_ASSERT(m_ses.has_peer(*i)); peer_connection const& p = *(*i); for (std::vector::const_iterator i = p.request_queue().begin() , end(p.request_queue().end()); i != end; ++i) { ++num_requests[i->block].num_peers; ++num_requests[i->block].num_peers_with_timeouts; ++num_requests[i->block].num_peers_with_nowant; ++num_requests[i->block].num_not_requested; // num_requests[i->block].peers.push_back(&p); } for (std::vector::const_iterator i = p.download_queue().begin() , end(p.download_queue().end()); i != end; ++i) { if (!i->not_wanted && !i->timed_out) ++num_requests[i->block].num_peers; if (i->timed_out) ++num_requests[i->block].num_peers_with_timeouts; if (i->not_wanted) ++num_requests[i->block].num_peers_with_nowant; // num_requests[i->block].peers.push_back(&p); } } for (std::map::iterator i = num_requests.begin() , end(num_requests.end()); i != end; ++i) { piece_block b = i->first; peer_count_t const& pc = i->second; int count = pc.num_peers; int count_with_timeouts = pc.num_peers_with_timeouts; int count_with_nowant = pc.num_peers_with_nowant; (void)count_with_timeouts; (void)count_with_nowant; int picker_count = t->picker().num_peers(b); if (!t->picker().is_downloaded(b)) TORRENT_ASSERT(picker_count == count); } } #endif /* if (t->has_picker() && !t->is_aborted()) { for (std::vector::const_iterator i = m_download_queue.begin() , end(m_download_queue.end()); i != end; ++i) { pending_block const& pb = *i; if (pb.timed_out || pb.not_wanted) continue; TORRENT_ASSERT(t->picker().get_block_state(pb.block) != piece_picker::block_info::state_none); TORRENT_ASSERT(complete); } } */ // extremely expensive invariant check /* if (!t->is_seed()) { piece_picker& p = t->picker(); const std::vector& dlq = p.get_download_queue(); const int blocks_per_piece = static_cast( t->torrent_file().piece_length() / t->block_size()); for (std::vector::const_iterator i = dlq.begin(); i != dlq.end(); ++i) { for (int j = 0; j < blocks_per_piece; ++j) { if (std::find(m_request_queue.begin(), m_request_queue.end() , piece_block(i->index, j)) != m_request_queue.end() || std::find(m_download_queue.begin(), m_download_queue.end() , piece_block(i->index, j)) != m_download_queue.end()) { TORRENT_ASSERT(i->info[j].peer == m_remote); } else { TORRENT_ASSERT(i->info[j].peer != m_remote || i->info[j].finished); } } } } */ } #endif peer_connection::peer_speed_t peer_connection::peer_speed() { TORRENT_ASSERT(is_single_thread()); shared_ptr t = m_torrent.lock(); TORRENT_ASSERT(t); int download_rate = int(statistics().download_payload_rate()); int torrent_download_rate = int(t->statistics().download_payload_rate()); if (download_rate > 512 && download_rate > torrent_download_rate / 16) m_speed = fast; else if (download_rate > 4096 && download_rate > torrent_download_rate / 64) m_speed = medium; else if (download_rate < torrent_download_rate / 15 && m_speed == fast) m_speed = medium; else m_speed = slow; return peer_connection::peer_speed_t(m_speed); } void peer_connection::keep_alive() { TORRENT_ASSERT(is_single_thread()); #ifdef TORRENT_EXPENSIVE_INVARIANT_CHECKS INVARIANT_CHECK; #endif time_duration d; d = time_now() - m_last_sent; if (total_seconds(d) < timeout() / 2) return; if (m_connecting) return; if (in_handshake()) return; // if the last send has not completed yet, do not send a keep // alive if (m_channel_state[upload_channel] & peer_info::bw_network) return; #ifdef TORRENT_VERBOSE_LOGGING peer_log("==> KEEPALIVE"); #endif m_last_sent = time_now(); write_keepalive(); } bool peer_connection::is_seed() const { TORRENT_ASSERT(is_single_thread()); // if m_num_pieces == 0, we probably don't have the // metadata yet. boost::shared_ptr t = m_torrent.lock(); return m_num_pieces == (int)m_have_piece.size() && m_num_pieces > 0 && t && t->valid_metadata(); } void peer_connection::set_share_mode(bool u) { TORRENT_ASSERT(is_single_thread()); // if the peer is a seed, ignore share mode messages if (is_seed()) return; m_share_mode = u; } void peer_connection::set_upload_only(bool u) { TORRENT_ASSERT(is_single_thread()); // if the peer is a seed, don't allow setting // upload_only to false if (m_upload_only || is_seed()) return; m_upload_only = u; boost::shared_ptr t = associated_torrent().lock(); t->set_seed(m_peer_info, u); disconnect_if_redundant(); } }