/* * This file Copyright (C) 2007-2014 Mnemosyne LLC * * It may be used under the GNU GPL versions 2 or 3 * or any future license endorsed by Mnemosyne LLC. * * $Id$ */ #include #include #include #include #include #include #include #include #include "transmission.h" #include "cache.h" #include "completion.h" #include "file.h" #include "log.h" #include "peer-io.h" #include "peer-mgr.h" #include "peer-msgs.h" #include "session.h" #include "torrent.h" #include "torrent-magnet.h" #include "tr-dht.h" #include "utils.h" #include "variant.h" #include "version.h" #ifndef EBADMSG #define EBADMSG EINVAL #endif /** *** **/ enum { BT_CHOKE = 0, BT_UNCHOKE = 1, BT_INTERESTED = 2, BT_NOT_INTERESTED = 3, BT_HAVE = 4, BT_BITFIELD = 5, BT_REQUEST = 6, BT_PIECE = 7, BT_CANCEL = 8, BT_PORT = 9, BT_FEXT_SUGGEST = 13, BT_FEXT_HAVE_ALL = 14, BT_FEXT_HAVE_NONE = 15, BT_FEXT_REJECT = 16, BT_FEXT_ALLOWED_FAST = 17, BT_LTEP = 20, LTEP_HANDSHAKE = 0, UT_PEX_ID = 1, UT_METADATA_ID = 3, MAX_PEX_PEER_COUNT = 50, MIN_CHOKE_PERIOD_SEC = 10, /* idle seconds before we send a keepalive */ KEEPALIVE_INTERVAL_SECS = 100, PEX_INTERVAL_SECS = 90, /* sec between sendPex () calls */ REQQ = 512, METADATA_REQQ = 64, MAGIC_NUMBER = 21549, /* used in lowering the outMessages queue period */ IMMEDIATE_PRIORITY_INTERVAL_SECS = 0, HIGH_PRIORITY_INTERVAL_SECS = 2, LOW_PRIORITY_INTERVAL_SECS = 10, /* number of pieces we'll allow in our fast set */ MAX_FAST_SET_SIZE = 3, /* how many blocks to keep prefetched per peer */ PREFETCH_SIZE = 18, /* when we're making requests from another peer, batch them together to send enough requests to meet our bandwidth goals for the next N seconds */ REQUEST_BUF_SECS = 10, /* defined in BEP #9 */ METADATA_MSG_TYPE_REQUEST = 0, METADATA_MSG_TYPE_DATA = 1, METADATA_MSG_TYPE_REJECT = 2 }; enum { AWAITING_BT_LENGTH, AWAITING_BT_ID, AWAITING_BT_MESSAGE, AWAITING_BT_PIECE }; typedef enum { ENCRYPTION_PREFERENCE_UNKNOWN, ENCRYPTION_PREFERENCE_YES, ENCRYPTION_PREFERENCE_NO } encryption_preference_t; /** *** **/ struct peer_request { uint32_t index; uint32_t offset; uint32_t length; }; static void blockToReq (const tr_torrent * tor, tr_block_index_t block, struct peer_request * setme) { tr_torrentGetBlockLocation (tor, block, &setme->index, &setme->offset, &setme->length); } /** *** **/ /* this is raw, unchanged data from the peer regarding * the current message that it's sending us. */ struct tr_incoming { uint8_t id; uint32_t length; /* includes the +1 for id length */ struct peer_request blockReq; /* metadata for incoming blocks */ struct evbuffer * block; /* piece data for incoming blocks */ }; /** * Low-level communication state information about a connected peer. * * This structure remembers the low-level protocol states that we're * in with this peer, such as active requests, pex messages, and so on. * Its fields are all private to peer-msgs.c. * * Data not directly involved with sending & receiving messages is * stored in tr_peer, where it can be accessed by both peermsgs and * the peer manager. * * @see struct peer_atom * @see tr_peer */ struct tr_peerMsgs { struct tr_peer peer; /* parent */ uint16_t magic_number; /* Whether or not we've choked this peer. */ bool peer_is_choked; /* whether or not the peer has indicated it will download from us. */ bool peer_is_interested; /* whether or the peer is choking us. */ bool client_is_choked; /* whether or not we've indicated to the peer that we would download from them if unchoked. */ bool client_is_interested; bool peerSupportsPex; bool peerSupportsMetadataXfer; bool clientSentLtepHandshake; bool peerSentLtepHandshake; /*bool haveFastSet;*/ int desiredRequestCount; int prefetchCount; int is_active[2]; /* how long the outMessages batch should be allowed to grow before * it's flushed -- some messages (like requests >:) should be sent * very quickly; others aren't as urgent. */ int8_t outMessagesBatchPeriod; uint8_t state; uint8_t ut_pex_id; uint8_t ut_metadata_id; uint16_t pexCount; uint16_t pexCount6; tr_port dht_port; encryption_preference_t encryption_preference; size_t metadata_size_hint; #if 0 size_t fastsetSize; tr_piece_index_t fastset[MAX_FAST_SET_SIZE]; #endif tr_torrent * torrent; tr_peer_callback callback; void * callbackData; struct evbuffer * outMessages; /* all the non-piece messages */ struct peer_request peerAskedFor[REQQ]; int peerAskedForMetadata[METADATA_REQQ]; int peerAskedForMetadataCount; tr_pex * pex; tr_pex * pex6; /*time_t clientSentPexAt;*/ time_t clientSentAnythingAt; time_t chokeChangedAt; /* when we started batching the outMessages */ time_t outMessagesBatchedAt; struct tr_incoming incoming; /* if the peer supports the Extension Protocol in BEP 10 and supplied a reqq argument, it's stored here. Otherwise, the value is zero and should be ignored. */ int64_t reqq; struct event * pexTimer; struct tr_peerIo * io; }; /** *** **/ static inline tr_session* getSession (struct tr_peerMsgs * msgs) { return msgs->torrent->session; } /** *** **/ static void myDebug (const char * file, int line, const struct tr_peerMsgs * msgs, const char * fmt, ...) TR_GNUC_PRINTF(4, 5); static void myDebug (const char * file, int line, const struct tr_peerMsgs * msgs, const char * fmt, ...) { const tr_sys_file_t fp = tr_logGetFile (); if (fp != TR_BAD_SYS_FILE) { va_list args; char timestr[64]; struct evbuffer * buf = evbuffer_new (); char * base = tr_sys_path_basename (file, NULL); char * message; evbuffer_add_printf (buf, "[%s] %s - %s [%s]: ", tr_logGetTimeStr (timestr, sizeof (timestr)), tr_torrentName (msgs->torrent), tr_peerIoGetAddrStr (msgs->io), tr_quark_get_string (msgs->peer.client, NULL)); va_start (args, fmt); evbuffer_add_vprintf (buf, fmt, args); va_end (args); evbuffer_add_printf (buf, " (%s:%d)", base, line); message = evbuffer_free_to_str (buf); tr_sys_file_write_line (fp, message, NULL); tr_free (base); tr_free (message); } } #define dbgmsg(msgs, ...) \ do \ { \ if (tr_logGetDeepEnabled ()) \ myDebug (__FILE__, __LINE__, msgs, __VA_ARGS__); \ } \ while (0) /** *** **/ static void pokeBatchPeriod (tr_peerMsgs * msgs, int interval) { if (msgs->outMessagesBatchPeriod > interval) { msgs->outMessagesBatchPeriod = interval; dbgmsg (msgs, "lowering batch interval to %d seconds", interval); } } static void dbgOutMessageLen (tr_peerMsgs * msgs) { dbgmsg (msgs, "outMessage size is now %"TR_PRIuSIZE, evbuffer_get_length (msgs->outMessages)); } static void protocolSendReject (tr_peerMsgs * msgs, const struct peer_request * req) { struct evbuffer * out = msgs->outMessages; assert (tr_peerIoSupportsFEXT (msgs->io)); evbuffer_add_uint32 (out, sizeof (uint8_t) + 3 * sizeof (uint32_t)); evbuffer_add_uint8 (out, BT_FEXT_REJECT); evbuffer_add_uint32 (out, req->index); evbuffer_add_uint32 (out, req->offset); evbuffer_add_uint32 (out, req->length); dbgmsg (msgs, "rejecting %u:%u->%u...", req->index, req->offset, req->length); dbgOutMessageLen (msgs); } static void protocolSendRequest (tr_peerMsgs * msgs, const struct peer_request * req) { struct evbuffer * out = msgs->outMessages; evbuffer_add_uint32 (out, sizeof (uint8_t) + 3 * sizeof (uint32_t)); evbuffer_add_uint8 (out, BT_REQUEST); evbuffer_add_uint32 (out, req->index); evbuffer_add_uint32 (out, req->offset); evbuffer_add_uint32 (out, req->length); dbgmsg (msgs, "requesting %u:%u->%u...", req->index, req->offset, req->length); dbgOutMessageLen (msgs); pokeBatchPeriod (msgs, IMMEDIATE_PRIORITY_INTERVAL_SECS); } static void protocolSendCancel (tr_peerMsgs * msgs, const struct peer_request * req) { struct evbuffer * out = msgs->outMessages; evbuffer_add_uint32 (out, sizeof (uint8_t) + 3 * sizeof (uint32_t)); evbuffer_add_uint8 (out, BT_CANCEL); evbuffer_add_uint32 (out, req->index); evbuffer_add_uint32 (out, req->offset); evbuffer_add_uint32 (out, req->length); dbgmsg (msgs, "cancelling %u:%u->%u...", req->index, req->offset, req->length); dbgOutMessageLen (msgs); pokeBatchPeriod (msgs, IMMEDIATE_PRIORITY_INTERVAL_SECS); } static void protocolSendPort (tr_peerMsgs *msgs, uint16_t port) { struct evbuffer * out = msgs->outMessages; dbgmsg (msgs, "sending Port %u", port); evbuffer_add_uint32 (out, 3); evbuffer_add_uint8 (out, BT_PORT); evbuffer_add_uint16 (out, port); } static void protocolSendHave (tr_peerMsgs * msgs, uint32_t index) { struct evbuffer * out = msgs->outMessages; evbuffer_add_uint32 (out, sizeof (uint8_t) + sizeof (uint32_t)); evbuffer_add_uint8 (out, BT_HAVE); evbuffer_add_uint32 (out, index); dbgmsg (msgs, "sending Have %u", index); dbgOutMessageLen (msgs); pokeBatchPeriod (msgs, LOW_PRIORITY_INTERVAL_SECS); } #if 0 static void protocolSendAllowedFast (tr_peerMsgs * msgs, uint32_t pieceIndex) { tr_peerIo * io = msgs->io; struct evbuffer * out = msgs->outMessages; assert (tr_peerIoSupportsFEXT (msgs->io)); evbuffer_add_uint32 (io, out, sizeof (uint8_t) + sizeof (uint32_t)); evbuffer_add_uint8 (io, out, BT_FEXT_ALLOWED_FAST); evbuffer_add_uint32 (io, out, pieceIndex); dbgmsg (msgs, "sending Allowed Fast %u...", pieceIndex); dbgOutMessageLen (msgs); } #endif static void protocolSendChoke (tr_peerMsgs * msgs, int choke) { struct evbuffer * out = msgs->outMessages; evbuffer_add_uint32 (out, sizeof (uint8_t)); evbuffer_add_uint8 (out, choke ? BT_CHOKE : BT_UNCHOKE); dbgmsg (msgs, "sending %s...", choke ? "Choke" : "Unchoke"); dbgOutMessageLen (msgs); pokeBatchPeriod (msgs, IMMEDIATE_PRIORITY_INTERVAL_SECS); } static void protocolSendHaveAll (tr_peerMsgs * msgs) { struct evbuffer * out = msgs->outMessages; assert (tr_peerIoSupportsFEXT (msgs->io)); evbuffer_add_uint32 (out, sizeof (uint8_t)); evbuffer_add_uint8 (out, BT_FEXT_HAVE_ALL); dbgmsg (msgs, "sending HAVE_ALL..."); dbgOutMessageLen (msgs); pokeBatchPeriod (msgs, IMMEDIATE_PRIORITY_INTERVAL_SECS); } static void protocolSendHaveNone (tr_peerMsgs * msgs) { struct evbuffer * out = msgs->outMessages; assert (tr_peerIoSupportsFEXT (msgs->io)); evbuffer_add_uint32 (out, sizeof (uint8_t)); evbuffer_add_uint8 (out, BT_FEXT_HAVE_NONE); dbgmsg (msgs, "sending HAVE_NONE..."); dbgOutMessageLen (msgs); pokeBatchPeriod (msgs, IMMEDIATE_PRIORITY_INTERVAL_SECS); } /** *** EVENTS **/ static void publish (tr_peerMsgs * msgs, tr_peer_event * e) { if (msgs->callback != NULL) (*msgs->callback) (&msgs->peer, e, msgs->callbackData); } static void fireError (tr_peerMsgs * msgs, int err) { tr_peer_event e = TR_PEER_EVENT_INIT; e.eventType = TR_PEER_ERROR; e.err = err; publish (msgs, &e); } static void fireGotBlock (tr_peerMsgs * msgs, const struct peer_request * req) { tr_peer_event e = TR_PEER_EVENT_INIT; e.eventType = TR_PEER_CLIENT_GOT_BLOCK; e.pieceIndex = req->index; e.offset = req->offset; e.length = req->length; publish (msgs, &e); } static void fireGotRej (tr_peerMsgs * msgs, const struct peer_request * req) { tr_peer_event e = TR_PEER_EVENT_INIT; e.eventType = TR_PEER_CLIENT_GOT_REJ; e.pieceIndex = req->index; e.offset = req->offset; e.length = req->length; publish (msgs, &e); } static void fireGotChoke (tr_peerMsgs * msgs) { tr_peer_event e = TR_PEER_EVENT_INIT; e.eventType = TR_PEER_CLIENT_GOT_CHOKE; publish (msgs, &e); } static void fireClientGotHaveAll (tr_peerMsgs * msgs) { tr_peer_event e = TR_PEER_EVENT_INIT; e.eventType = TR_PEER_CLIENT_GOT_HAVE_ALL; publish (msgs, &e); } static void fireClientGotHaveNone (tr_peerMsgs * msgs) { tr_peer_event e = TR_PEER_EVENT_INIT; e.eventType = TR_PEER_CLIENT_GOT_HAVE_NONE; publish (msgs, &e); } static void fireClientGotPieceData (tr_peerMsgs * msgs, uint32_t length) { tr_peer_event e = TR_PEER_EVENT_INIT; e.length = length; e.eventType = TR_PEER_CLIENT_GOT_PIECE_DATA; publish (msgs, &e); } static void firePeerGotPieceData (tr_peerMsgs * msgs, uint32_t length) { tr_peer_event e = TR_PEER_EVENT_INIT; e.length = length; e.eventType = TR_PEER_PEER_GOT_PIECE_DATA; publish (msgs, &e); } static void fireClientGotSuggest (tr_peerMsgs * msgs, uint32_t pieceIndex) { tr_peer_event e = TR_PEER_EVENT_INIT; e.eventType = TR_PEER_CLIENT_GOT_SUGGEST; e.pieceIndex = pieceIndex; publish (msgs, &e); } static void fireClientGotPort (tr_peerMsgs * msgs, tr_port port) { tr_peer_event e = TR_PEER_EVENT_INIT; e.eventType = TR_PEER_CLIENT_GOT_PORT; e.port = port; publish (msgs, &e); } static void fireClientGotAllowedFast (tr_peerMsgs * msgs, uint32_t pieceIndex) { tr_peer_event e = TR_PEER_EVENT_INIT; e.eventType = TR_PEER_CLIENT_GOT_ALLOWED_FAST; e.pieceIndex = pieceIndex; publish (msgs, &e); } static void fireClientGotBitfield (tr_peerMsgs * msgs, tr_bitfield * bitfield) { tr_peer_event e = TR_PEER_EVENT_INIT; e.eventType = TR_PEER_CLIENT_GOT_BITFIELD; e.bitfield = bitfield; publish (msgs, &e); } static void fireClientGotHave (tr_peerMsgs * msgs, tr_piece_index_t index) { tr_peer_event e = TR_PEER_EVENT_INIT; e.eventType = TR_PEER_CLIENT_GOT_HAVE; e.pieceIndex = index; publish (msgs, &e); } /** *** ALLOWED FAST SET *** For explanation, see http://www.bittorrent.org/beps/bep_0006.html **/ #if 0 size_t tr_generateAllowedSet (tr_piece_index_t * setmePieces, size_t desiredSetSize, size_t pieceCount, const uint8_t * infohash, const tr_address * addr) { size_t setSize = 0; assert (setmePieces); assert (desiredSetSize <= pieceCount); assert (desiredSetSize); assert (pieceCount); assert (infohash); assert (addr); if (addr->type == TR_AF_INET) { uint8_t w[SHA_DIGEST_LENGTH + 4], *walk=w; uint8_t x[SHA_DIGEST_LENGTH]; uint32_t ui32 = ntohl (htonl (addr->addr.addr4.s_addr) & 0xffffff00); /* (1) */ memcpy (w, &ui32, sizeof (uint32_t)); walk += sizeof (uint32_t); memcpy (walk, infohash, SHA_DIGEST_LENGTH); /* (2) */ walk += SHA_DIGEST_LENGTH; tr_sha1 (x, w, walk-w, NULL); /* (3) */ assert (sizeof (w) == walk-w); while (setSizeio); const int peerIsNeedy = msgs->peer->progress < 0.10; if (fext && peerIsNeedy && !msgs->haveFastSet) { size_t i; const struct tr_address * addr = tr_peerIoGetAddress (msgs->io, NULL); const tr_info * inf = &msgs->torrent->info; const size_t numwant = MIN (MAX_FAST_SET_SIZE, inf->pieceCount); /* build the fast set */ msgs->fastsetSize = tr_generateAllowedSet (msgs->fastset, numwant, inf->pieceCount, inf->hash, addr); msgs->haveFastSet = true; /* send it to the peer */ for (i=0; ifastsetSize; ++i) protocolSendAllowedFast (msgs, msgs->fastset[i]); } } #endif /*** **** ACTIVE ***/ static bool tr_peerMsgsCalculateActive (const tr_peerMsgs * msgs, tr_direction direction) { bool is_active; assert (tr_isPeerMsgs (msgs)); assert (tr_isDirection (direction)); if (direction == TR_CLIENT_TO_PEER) { is_active = tr_peerMsgsIsPeerInterested (msgs) && !tr_peerMsgsIsPeerChoked (msgs); if (is_active) assert (!tr_peerIsSeed (&msgs->peer)); } else /* TR_PEER_TO_CLIENT */ { if (!tr_torrentHasMetadata (msgs->torrent)) { is_active = true; } else { is_active = tr_peerMsgsIsClientInterested (msgs) && !tr_peerMsgsIsClientChoked (msgs); if (is_active) assert (!tr_torrentIsSeed (msgs->torrent)); } } return is_active; } bool tr_peerMsgsIsActive (const tr_peerMsgs * msgs, tr_direction direction) { bool is_active; assert (tr_isPeerMsgs (msgs)); assert (tr_isDirection (direction)); is_active = msgs->is_active[direction]; assert (is_active == tr_peerMsgsCalculateActive (msgs, direction)); return is_active; } static void tr_peerMsgsSetActive (tr_peerMsgs * msgs, tr_direction direction, bool is_active) { dbgmsg (msgs, "direction [%d] is_active [%d]", (int)direction, (int)is_active); if (msgs->is_active[direction] != is_active) { msgs->is_active[direction] = is_active; tr_swarmIncrementActivePeers (msgs->torrent->swarm, direction, is_active); } } void tr_peerMsgsUpdateActive (tr_peerMsgs * msgs, tr_direction direction) { const bool is_active = tr_peerMsgsCalculateActive (msgs, direction); tr_peerMsgsSetActive (msgs, direction, is_active); } /** *** INTEREST **/ static void sendInterest (tr_peerMsgs * msgs, bool b) { struct evbuffer * out = msgs->outMessages; assert (msgs); assert (tr_isBool (b)); msgs->client_is_interested = b; dbgmsg (msgs, "Sending %s", b ? "Interested" : "Not Interested"); evbuffer_add_uint32 (out, sizeof (uint8_t)); evbuffer_add_uint8 (out, b ? BT_INTERESTED : BT_NOT_INTERESTED); pokeBatchPeriod (msgs, HIGH_PRIORITY_INTERVAL_SECS); dbgOutMessageLen (msgs); } static void updateInterest (tr_peerMsgs * msgs UNUSED) { /* FIXME -- might need to poke the mgr on startup */ } void tr_peerMsgsSetInterested (tr_peerMsgs * msgs, bool b) { assert (tr_isBool (b)); if (msgs->client_is_interested != b) { sendInterest (msgs, b); tr_peerMsgsUpdateActive (msgs, TR_PEER_TO_CLIENT); } } static bool popNextMetadataRequest (tr_peerMsgs * msgs, int * piece) { if (msgs->peerAskedForMetadataCount == 0) return false; *piece = msgs->peerAskedForMetadata[0]; tr_removeElementFromArray (msgs->peerAskedForMetadata, 0, sizeof (int), msgs->peerAskedForMetadataCount--); return true; } static bool popNextRequest (tr_peerMsgs * msgs, struct peer_request * setme) { if (msgs->peer.pendingReqsToClient == 0) return false; *setme = msgs->peerAskedFor[0]; tr_removeElementFromArray (msgs->peerAskedFor, 0, sizeof (struct peer_request), msgs->peer.pendingReqsToClient--); return true; } static void cancelAllRequestsToClient (tr_peerMsgs * msgs) { struct peer_request req; const int mustSendCancel = tr_peerIoSupportsFEXT (msgs->io); while (popNextRequest (msgs, &req)) if (mustSendCancel) protocolSendReject (msgs, &req); } void tr_peerMsgsSetChoke (tr_peerMsgs * msgs, bool peer_is_choked) { const time_t now = tr_time (); const time_t fibrillationTime = now - MIN_CHOKE_PERIOD_SEC; assert (msgs != NULL); assert (tr_isBool (peer_is_choked)); if (msgs->chokeChangedAt > fibrillationTime) { dbgmsg (msgs, "Not changing choke to %d to avoid fibrillation", peer_is_choked); } else if (msgs->peer_is_choked != peer_is_choked) { msgs->peer_is_choked = peer_is_choked; if (peer_is_choked) cancelAllRequestsToClient (msgs); protocolSendChoke (msgs, peer_is_choked); msgs->chokeChangedAt = now; tr_peerMsgsUpdateActive (msgs, TR_CLIENT_TO_PEER); } } /** *** **/ void tr_peerMsgsHave (tr_peerMsgs * msgs, uint32_t index) { protocolSendHave (msgs, index); /* since we have more pieces now, we might not be interested in this peer */ updateInterest (msgs); } /** *** **/ static bool reqIsValid (const tr_peerMsgs * peer, uint32_t index, uint32_t offset, uint32_t length) { return tr_torrentReqIsValid (peer->torrent, index, offset, length); } static bool requestIsValid (const tr_peerMsgs * msgs, const struct peer_request * req) { return reqIsValid (msgs, req->index, req->offset, req->length); } void tr_peerMsgsCancel (tr_peerMsgs * msgs, tr_block_index_t block) { struct peer_request req; /*fprintf (stderr, "SENDING CANCEL MESSAGE FOR BLOCK %"TR_PRIuSIZE"\n\t\tFROM PEER %p ------------------------------------\n", (size_t)block, msgs->peer);*/ blockToReq (msgs->torrent, block, &req); protocolSendCancel (msgs, &req); } /** *** **/ static void sendLtepHandshake (tr_peerMsgs * msgs) { tr_variant val; bool allow_pex; bool allow_metadata_xfer; struct evbuffer * payload; struct evbuffer * out = msgs->outMessages; const unsigned char * ipv6 = tr_globalIPv6 (); static tr_quark version_quark = 0; if (msgs->clientSentLtepHandshake) return; if (!version_quark) version_quark = tr_quark_new (TR_NAME " " USERAGENT_PREFIX, -1); dbgmsg (msgs, "sending an ltep handshake"); msgs->clientSentLtepHandshake = true; /* decide if we want to advertise metadata xfer support (BEP 9) */ if (tr_torrentIsPrivate (msgs->torrent)) allow_metadata_xfer = false; else allow_metadata_xfer = true; /* decide if we want to advertise pex support */ if (!tr_torrentAllowsPex (msgs->torrent)) allow_pex = false; else if (msgs->peerSentLtepHandshake) allow_pex = msgs->peerSupportsPex; else allow_pex = true; tr_variantInitDict (&val, 8); tr_variantDictAddInt (&val, TR_KEY_e, getSession (msgs)->encryptionMode != TR_CLEAR_PREFERRED); if (ipv6 != NULL) tr_variantDictAddRaw (&val, TR_KEY_ipv6, ipv6, 16); if (allow_metadata_xfer && tr_torrentHasMetadata (msgs->torrent) && (msgs->torrent->infoDictLength > 0)) tr_variantDictAddInt (&val, TR_KEY_metadata_size, msgs->torrent->infoDictLength); tr_variantDictAddInt (&val, TR_KEY_p, tr_sessionGetPublicPeerPort (getSession (msgs))); tr_variantDictAddInt (&val, TR_KEY_reqq, REQQ); tr_variantDictAddInt (&val, TR_KEY_upload_only, tr_torrentIsSeed (msgs->torrent)); tr_variantDictAddQuark (&val, TR_KEY_v, version_quark); if (allow_metadata_xfer || allow_pex) { tr_variant * m = tr_variantDictAddDict (&val, TR_KEY_m, 2); if (allow_metadata_xfer) tr_variantDictAddInt (m, TR_KEY_ut_metadata, UT_METADATA_ID); if (allow_pex) tr_variantDictAddInt (m, TR_KEY_ut_pex, UT_PEX_ID); } payload = tr_variantToBuf (&val, TR_VARIANT_FMT_BENC); evbuffer_add_uint32 (out, 2 * sizeof (uint8_t) + evbuffer_get_length (payload)); evbuffer_add_uint8 (out, BT_LTEP); evbuffer_add_uint8 (out, LTEP_HANDSHAKE); evbuffer_add_buffer (out, payload); pokeBatchPeriod (msgs, IMMEDIATE_PRIORITY_INTERVAL_SECS); dbgOutMessageLen (msgs); /* cleanup */ evbuffer_free (payload); tr_variantFree (&val); } static void parseLtepHandshake (tr_peerMsgs * msgs, int len, struct evbuffer * inbuf) { int64_t i; tr_variant val, * sub; uint8_t * tmp = tr_new (uint8_t, len); const uint8_t *addr; size_t addr_len; tr_pex pex; int8_t seedProbability = -1; memset (&pex, 0, sizeof (tr_pex)); tr_peerIoReadBytes (msgs->io, inbuf, tmp, len); msgs->peerSentLtepHandshake = true; if (tr_variantFromBenc (&val, tmp, len) || !tr_variantIsDict (&val)) { dbgmsg (msgs, "GET extended-handshake, couldn't get dictionary"); tr_free (tmp); return; } dbgmsg (msgs, "here is the handshake: [%*.*s]", len, len, tmp); /* does the peer prefer encrypted connections? */ if (tr_variantDictFindInt (&val, TR_KEY_e, &i)) { msgs->encryption_preference = i ? ENCRYPTION_PREFERENCE_YES : ENCRYPTION_PREFERENCE_NO; if (i) pex.flags |= ADDED_F_ENCRYPTION_FLAG; } /* check supported messages for utorrent pex */ msgs->peerSupportsPex = false; msgs->peerSupportsMetadataXfer = false; if (tr_variantDictFindDict (&val, TR_KEY_m, &sub)) { if (tr_variantDictFindInt (sub, TR_KEY_ut_pex, &i)) { msgs->peerSupportsPex = i != 0; msgs->ut_pex_id = (uint8_t) i; dbgmsg (msgs, "msgs->ut_pex is %d", (int)msgs->ut_pex_id); } if (tr_variantDictFindInt (sub, TR_KEY_ut_metadata, &i)) { msgs->peerSupportsMetadataXfer = i != 0; msgs->ut_metadata_id = (uint8_t) i; dbgmsg (msgs, "msgs->ut_metadata_id is %d", (int)msgs->ut_metadata_id); } if (tr_variantDictFindInt (sub, TR_KEY_ut_holepunch, &i)) { /* Mysterious µTorrent extension that we don't grok. However, it implies support for µTP, so use it to indicate that. */ tr_peerMgrSetUtpFailed (msgs->torrent, tr_peerIoGetAddress (msgs->io, NULL), false); } } /* look for metainfo size (BEP 9) */ if (tr_variantDictFindInt (&val, TR_KEY_metadata_size, &i)) { tr_torrentSetMetadataSizeHint (msgs->torrent, i); msgs->metadata_size_hint = (size_t) i; } /* look for upload_only (BEP 21) */ if (tr_variantDictFindInt (&val, TR_KEY_upload_only, &i)) seedProbability = i==0 ? 0 : 100; /* get peer's listening port */ if (tr_variantDictFindInt (&val, TR_KEY_p, &i)) { pex.port = htons ((uint16_t)i); fireClientGotPort (msgs, pex.port); dbgmsg (msgs, "peer's port is now %d", (int)i); } if (tr_peerIoIsIncoming (msgs->io) && tr_variantDictFindRaw (&val, TR_KEY_ipv4, &addr, &addr_len) && (addr_len == 4)) { pex.addr.type = TR_AF_INET; memcpy (&pex.addr.addr.addr4, addr, 4); tr_peerMgrAddPex (msgs->torrent, TR_PEER_FROM_LTEP, &pex, seedProbability); } if (tr_peerIoIsIncoming (msgs->io) && tr_variantDictFindRaw (&val, TR_KEY_ipv6, &addr, &addr_len) && (addr_len == 16)) { pex.addr.type = TR_AF_INET6; memcpy (&pex.addr.addr.addr6, addr, 16); tr_peerMgrAddPex (msgs->torrent, TR_PEER_FROM_LTEP, &pex, seedProbability); } /* get peer's maximum request queue size */ if (tr_variantDictFindInt (&val, TR_KEY_reqq, &i)) msgs->reqq = i; tr_variantFree (&val); tr_free (tmp); } static void parseUtMetadata (tr_peerMsgs * msgs, int msglen, struct evbuffer * inbuf) { tr_variant dict; char * msg_end; const char * benc_end; int64_t msg_type = -1; int64_t piece = -1; int64_t total_size = 0; uint8_t * tmp = tr_new (uint8_t, msglen); tr_peerIoReadBytes (msgs->io, inbuf, tmp, msglen); msg_end = (char*)tmp + msglen; if (!tr_variantFromBencFull (&dict, tmp, msglen, NULL, &benc_end)) { tr_variantDictFindInt (&dict, TR_KEY_msg_type, &msg_type); tr_variantDictFindInt (&dict, TR_KEY_piece, &piece); tr_variantDictFindInt (&dict, TR_KEY_total_size, &total_size); tr_variantFree (&dict); } dbgmsg (msgs, "got ut_metadata msg: type %d, piece %d, total_size %d", (int)msg_type, (int)piece, (int)total_size); if (msg_type == METADATA_MSG_TYPE_REJECT) { /* NOOP */ } if ((msg_type == METADATA_MSG_TYPE_DATA) && (!tr_torrentHasMetadata (msgs->torrent)) && (msg_end - benc_end <= METADATA_PIECE_SIZE) && (piece * METADATA_PIECE_SIZE + (msg_end - benc_end) <= total_size)) { const int pieceLen = msg_end - benc_end; tr_torrentSetMetadataPiece (msgs->torrent, piece, benc_end, pieceLen); } if (msg_type == METADATA_MSG_TYPE_REQUEST) { if ((piece >= 0) && tr_torrentHasMetadata (msgs->torrent) && !tr_torrentIsPrivate (msgs->torrent) && (msgs->peerAskedForMetadataCount < METADATA_REQQ)) { msgs->peerAskedForMetadata[msgs->peerAskedForMetadataCount++] = piece; } else { tr_variant tmp; struct evbuffer * payload; struct evbuffer * out = msgs->outMessages; /* build the rejection message */ tr_variantInitDict (&tmp, 2); tr_variantDictAddInt (&tmp, TR_KEY_msg_type, METADATA_MSG_TYPE_REJECT); tr_variantDictAddInt (&tmp, TR_KEY_piece, piece); payload = tr_variantToBuf (&tmp, TR_VARIANT_FMT_BENC); /* write it out as a LTEP message to our outMessages buffer */ evbuffer_add_uint32 (out, 2 * sizeof (uint8_t) + evbuffer_get_length (payload)); evbuffer_add_uint8 (out, BT_LTEP); evbuffer_add_uint8 (out, msgs->ut_metadata_id); evbuffer_add_buffer (out, payload); pokeBatchPeriod (msgs, HIGH_PRIORITY_INTERVAL_SECS); dbgOutMessageLen (msgs); /* cleanup */ evbuffer_free (payload); tr_variantFree (&tmp); } } tr_free (tmp); } static void parseUtPex (tr_peerMsgs * msgs, int msglen, struct evbuffer * inbuf) { int loaded = 0; uint8_t * tmp = tr_new (uint8_t, msglen); tr_variant val; tr_torrent * tor = msgs->torrent; const uint8_t * added; size_t added_len; tr_peerIoReadBytes (msgs->io, inbuf, tmp, msglen); if (tr_torrentAllowsPex (tor) && ((loaded = !tr_variantFromBenc (&val, tmp, msglen)))) { if (tr_variantDictFindRaw (&val, TR_KEY_added, &added, &added_len)) { tr_pex * pex; size_t i, n; size_t added_f_len = 0; const uint8_t * added_f = NULL; tr_variantDictFindRaw (&val, TR_KEY_added_f, &added_f, &added_f_len); pex = tr_peerMgrCompactToPex (added, added_len, added_f, added_f_len, &n); n = MIN (n, MAX_PEX_PEER_COUNT); for (i=0; iio, inbuf, <ep_msgid); msglen--; if (ltep_msgid == LTEP_HANDSHAKE) { dbgmsg (msgs, "got ltep handshake"); parseLtepHandshake (msgs, msglen, inbuf); if (tr_peerIoSupportsLTEP (msgs->io)) { sendLtepHandshake (msgs); sendPex (msgs); } } else if (ltep_msgid == UT_PEX_ID) { dbgmsg (msgs, "got ut pex"); msgs->peerSupportsPex = true; parseUtPex (msgs, msglen, inbuf); } else if (ltep_msgid == UT_METADATA_ID) { dbgmsg (msgs, "got ut metadata"); msgs->peerSupportsMetadataXfer = true; parseUtMetadata (msgs, msglen, inbuf); } else { dbgmsg (msgs, "skipping unknown ltep message (%d)", (int)ltep_msgid); evbuffer_drain (inbuf, msglen); } } static int readBtLength (tr_peerMsgs * msgs, struct evbuffer * inbuf, size_t inlen) { uint32_t len; if (inlen < sizeof (len)) return READ_LATER; tr_peerIoReadUint32 (msgs->io, inbuf, &len); if (len == 0) /* peer sent us a keepalive message */ dbgmsg (msgs, "got KeepAlive"); else { msgs->incoming.length = len; msgs->state = AWAITING_BT_ID; } return READ_NOW; } static int readBtMessage (tr_peerMsgs *, struct evbuffer *, size_t); static int readBtId (tr_peerMsgs * msgs, struct evbuffer * inbuf, size_t inlen) { uint8_t id; if (inlen < sizeof (uint8_t)) return READ_LATER; tr_peerIoReadUint8 (msgs->io, inbuf, &id); msgs->incoming.id = id; dbgmsg (msgs, "msgs->incoming.id is now %d; msgs->incoming.length is %"TR_PRIuSIZE, id, (size_t)msgs->incoming.length); if (id == BT_PIECE) { msgs->state = AWAITING_BT_PIECE; return READ_NOW; } else if (msgs->incoming.length != 1) { msgs->state = AWAITING_BT_MESSAGE; return READ_NOW; } else return readBtMessage (msgs, inbuf, inlen - 1); } static void updatePeerProgress (tr_peerMsgs * msgs) { tr_peerUpdateProgress (msgs->torrent, &msgs->peer); /*updateFastSet (msgs);*/ updateInterest (msgs); } static void prefetchPieces (tr_peerMsgs *msgs) { int i; if (!getSession (msgs)->isPrefetchEnabled) return; for (i=msgs->prefetchCount; ipeer.pendingReqsToClient && ipeerAskedFor + i; if (requestIsValid (msgs, req)) { tr_cachePrefetchBlock (getSession (msgs)->cache, msgs->torrent, req->index, req->offset, req->length); ++msgs->prefetchCount; } } } static void peerMadeRequest (tr_peerMsgs * msgs, const struct peer_request * req) { const bool fext = tr_peerIoSupportsFEXT (msgs->io); const int reqIsValid = requestIsValid (msgs, req); const int clientHasPiece = reqIsValid && tr_torrentPieceIsComplete (msgs->torrent, req->index); const int peerIsChoked = msgs->peer_is_choked; bool allow = false; if (!reqIsValid) dbgmsg (msgs, "rejecting an invalid request."); else if (!clientHasPiece) dbgmsg (msgs, "rejecting request for a piece we don't have."); else if (peerIsChoked) dbgmsg (msgs, "rejecting request from choked peer"); else if (msgs->peer.pendingReqsToClient + 1 >= REQQ) dbgmsg (msgs, "rejecting request ... reqq is full"); else allow = true; if (allow) { msgs->peerAskedFor[msgs->peer.pendingReqsToClient++] = *req; prefetchPieces (msgs); } else if (fext) { protocolSendReject (msgs, req); } } static bool messageLengthIsCorrect (const tr_peerMsgs * msg, uint8_t id, uint32_t len) { switch (id) { case BT_CHOKE: case BT_UNCHOKE: case BT_INTERESTED: case BT_NOT_INTERESTED: case BT_FEXT_HAVE_ALL: case BT_FEXT_HAVE_NONE: return len == 1; case BT_HAVE: case BT_FEXT_SUGGEST: case BT_FEXT_ALLOWED_FAST: return len == 5; case BT_BITFIELD: if (tr_torrentHasMetadata (msg->torrent)) return len == (msg->torrent->info.pieceCount + 7u) / 8u + 1u; /* we don't know the piece count yet, so we can only guess whether to send true or false */ if (msg->metadata_size_hint > 0) return len <= msg->metadata_size_hint; return true; case BT_REQUEST: case BT_CANCEL: case BT_FEXT_REJECT: return len == 13; case BT_PIECE: return len > 9 && len <= 16393; case BT_PORT: return len == 3; case BT_LTEP: return len >= 2; default: return false; } } static int clientGotBlock (tr_peerMsgs * msgs, struct evbuffer * block, const struct peer_request * req); static int readBtPiece (tr_peerMsgs * msgs, struct evbuffer * inbuf, size_t inlen, size_t * setme_piece_bytes_read) { struct peer_request * req = &msgs->incoming.blockReq; assert (evbuffer_get_length (inbuf) >= inlen); dbgmsg (msgs, "In readBtPiece"); if (!req->length) { if (inlen < 8) return READ_LATER; tr_peerIoReadUint32 (msgs->io, inbuf, &req->index); tr_peerIoReadUint32 (msgs->io, inbuf, &req->offset); req->length = msgs->incoming.length - 9; dbgmsg (msgs, "got incoming block header %u:%u->%u", req->index, req->offset, req->length); return READ_NOW; } else { int err; size_t n; size_t nLeft; struct evbuffer * block_buffer; if (msgs->incoming.block == NULL) msgs->incoming.block = evbuffer_new (); block_buffer = msgs->incoming.block; /* read in another chunk of data */ nLeft = req->length - evbuffer_get_length (block_buffer); n = MIN (nLeft, inlen); tr_peerIoReadBytesToBuf (msgs->io, inbuf, block_buffer, n); fireClientGotPieceData (msgs, n); *setme_piece_bytes_read += n; dbgmsg (msgs, "got %"TR_PRIuSIZE" bytes for block %u:%u->%u ... %d remain", n, req->index, req->offset, req->length, (int)(req->length - evbuffer_get_length (block_buffer))); if (evbuffer_get_length (block_buffer) < req->length) return READ_LATER; /* pass the block along... */ err = clientGotBlock (msgs, block_buffer, req); evbuffer_drain (block_buffer, evbuffer_get_length (block_buffer)); /* cleanup */ req->length = 0; msgs->state = AWAITING_BT_LENGTH; return err ? READ_ERR : READ_NOW; } } static void updateDesiredRequestCount (tr_peerMsgs * msgs); static int readBtMessage (tr_peerMsgs * msgs, struct evbuffer * inbuf, size_t inlen) { uint32_t ui32; uint32_t msglen = msgs->incoming.length; const uint8_t id = msgs->incoming.id; #ifndef NDEBUG const size_t startBufLen = evbuffer_get_length (inbuf); #endif const bool fext = tr_peerIoSupportsFEXT (msgs->io); --msglen; /* id length */ dbgmsg (msgs, "got BT id %d, len %d, buffer size is %"TR_PRIuSIZE, (int)id, (int)msglen, inlen); if (inlen < msglen) return READ_LATER; if (!messageLengthIsCorrect (msgs, id, msglen + 1)) { dbgmsg (msgs, "bad packet - BT message #%d with a length of %d", (int)id, (int)msglen); fireError (msgs, EMSGSIZE); return READ_ERR; } switch (id) { case BT_CHOKE: dbgmsg (msgs, "got Choke"); msgs->client_is_choked = true; if (!fext) fireGotChoke (msgs); tr_peerMsgsUpdateActive (msgs, TR_PEER_TO_CLIENT); break; case BT_UNCHOKE: dbgmsg (msgs, "got Unchoke"); msgs->client_is_choked = false; tr_peerMsgsUpdateActive (msgs, TR_PEER_TO_CLIENT); updateDesiredRequestCount (msgs); break; case BT_INTERESTED: dbgmsg (msgs, "got Interested"); msgs->peer_is_interested = true; tr_peerMsgsUpdateActive (msgs, TR_CLIENT_TO_PEER); break; case BT_NOT_INTERESTED: dbgmsg (msgs, "got Not Interested"); msgs->peer_is_interested = false; tr_peerMsgsUpdateActive (msgs, TR_CLIENT_TO_PEER); break; case BT_HAVE: tr_peerIoReadUint32 (msgs->io, inbuf, &ui32); dbgmsg (msgs, "got Have: %u", ui32); if (tr_torrentHasMetadata (msgs->torrent) && (ui32 >= msgs->torrent->info.pieceCount)) { fireError (msgs, ERANGE); return READ_ERR; } /* a peer can send the same HAVE message twice... */ if (!tr_bitfieldHas (&msgs->peer.have, ui32)) { tr_bitfieldAdd (&msgs->peer.have, ui32); fireClientGotHave (msgs, ui32); } updatePeerProgress (msgs); break; case BT_BITFIELD: { uint8_t * tmp = tr_new (uint8_t, msglen); dbgmsg (msgs, "got a bitfield"); tr_peerIoReadBytes (msgs->io, inbuf, tmp, msglen); tr_bitfieldSetRaw (&msgs->peer.have, tmp, msglen, tr_torrentHasMetadata (msgs->torrent)); fireClientGotBitfield (msgs, &msgs->peer.have); updatePeerProgress (msgs); tr_free (tmp); break; } case BT_REQUEST: { struct peer_request r; tr_peerIoReadUint32 (msgs->io, inbuf, &r.index); tr_peerIoReadUint32 (msgs->io, inbuf, &r.offset); tr_peerIoReadUint32 (msgs->io, inbuf, &r.length); dbgmsg (msgs, "got Request: %u:%u->%u", r.index, r.offset, r.length); peerMadeRequest (msgs, &r); break; } case BT_CANCEL: { int i; struct peer_request r; tr_peerIoReadUint32 (msgs->io, inbuf, &r.index); tr_peerIoReadUint32 (msgs->io, inbuf, &r.offset); tr_peerIoReadUint32 (msgs->io, inbuf, &r.length); tr_historyAdd (&msgs->peer.cancelsSentToClient, tr_time (), 1); dbgmsg (msgs, "got a Cancel %u:%u->%u", r.index, r.offset, r.length); for (i=0; ipeer.pendingReqsToClient; ++i) { const struct peer_request * req = msgs->peerAskedFor + i; if ((req->index == r.index) && (req->offset == r.offset) && (req->length == r.length)) break; } if (i < msgs->peer.pendingReqsToClient) tr_removeElementFromArray (msgs->peerAskedFor, i, sizeof (struct peer_request), msgs->peer.pendingReqsToClient--); break; } case BT_PIECE: assert (0); /* handled elsewhere! */ break; case BT_PORT: dbgmsg (msgs, "Got a BT_PORT"); tr_peerIoReadUint16 (msgs->io, inbuf, &msgs->dht_port); if (msgs->dht_port > 0) tr_dhtAddNode (getSession (msgs), tr_peerAddress (&msgs->peer), msgs->dht_port, 0); break; case BT_FEXT_SUGGEST: dbgmsg (msgs, "Got a BT_FEXT_SUGGEST"); tr_peerIoReadUint32 (msgs->io, inbuf, &ui32); if (fext) fireClientGotSuggest (msgs, ui32); else { fireError (msgs, EMSGSIZE); return READ_ERR; } break; case BT_FEXT_ALLOWED_FAST: dbgmsg (msgs, "Got a BT_FEXT_ALLOWED_FAST"); tr_peerIoReadUint32 (msgs->io, inbuf, &ui32); if (fext) fireClientGotAllowedFast (msgs, ui32); else { fireError (msgs, EMSGSIZE); return READ_ERR; } break; case BT_FEXT_HAVE_ALL: dbgmsg (msgs, "Got a BT_FEXT_HAVE_ALL"); if (fext) { tr_bitfieldSetHasAll (&msgs->peer.have); assert (tr_bitfieldHasAll (&msgs->peer.have)); fireClientGotHaveAll (msgs); updatePeerProgress (msgs); } else { fireError (msgs, EMSGSIZE); return READ_ERR; } break; case BT_FEXT_HAVE_NONE: dbgmsg (msgs, "Got a BT_FEXT_HAVE_NONE"); if (fext) { tr_bitfieldSetHasNone (&msgs->peer.have); fireClientGotHaveNone (msgs); updatePeerProgress (msgs); } else { fireError (msgs, EMSGSIZE); return READ_ERR; } break; case BT_FEXT_REJECT: { struct peer_request r; dbgmsg (msgs, "Got a BT_FEXT_REJECT"); tr_peerIoReadUint32 (msgs->io, inbuf, &r.index); tr_peerIoReadUint32 (msgs->io, inbuf, &r.offset); tr_peerIoReadUint32 (msgs->io, inbuf, &r.length); if (fext) fireGotRej (msgs, &r); else { fireError (msgs, EMSGSIZE); return READ_ERR; } break; } case BT_LTEP: dbgmsg (msgs, "Got a BT_LTEP"); parseLtep (msgs, msglen, inbuf); break; default: dbgmsg (msgs, "peer sent us an UNKNOWN: %d", (int)id); tr_peerIoDrain (msgs->io, inbuf, msglen); break; } assert (msglen + 1 == msgs->incoming.length); assert (evbuffer_get_length (inbuf) == startBufLen - msglen); msgs->state = AWAITING_BT_LENGTH; return READ_NOW; } /* returns 0 on success, or an errno on failure */ static int clientGotBlock (tr_peerMsgs * msgs, struct evbuffer * data, const struct peer_request * req) { int err; tr_torrent * tor = msgs->torrent; const tr_block_index_t block = _tr_block (tor, req->index, req->offset); assert (msgs); assert (req); if (!requestIsValid (msgs, req)) { dbgmsg (msgs, "dropping invalid block %u:%u->%u", req->index, req->offset, req->length); return EBADMSG; } if (req->length != tr_torBlockCountBytes (msgs->torrent, block)) { dbgmsg (msgs, "wrong block size -- expected %u, got %d", tr_torBlockCountBytes (msgs->torrent, block), req->length); return EMSGSIZE; } dbgmsg (msgs, "got block %u:%u->%u", req->index, req->offset, req->length); if (!tr_peerMgrDidPeerRequest (msgs->torrent, &msgs->peer, block)) { dbgmsg (msgs, "we didn't ask for this message..."); return 0; } if (tr_torrentPieceIsComplete (msgs->torrent, req->index)) { dbgmsg (msgs, "we did ask for this message, but the piece is already complete..."); return 0; } /** *** Save the block **/ if ((err = tr_cacheWriteBlock (getSession (msgs)->cache, tor, req->index, req->offset, req->length, data))) return err; tr_bitfieldAdd (&msgs->peer.blame, req->index); fireGotBlock (msgs, req); return 0; } static int peerPulse (void * vmsgs); static void didWrite (tr_peerIo * io UNUSED, size_t bytesWritten, bool wasPieceData, void * vmsgs) { tr_peerMsgs * msgs = vmsgs; if (wasPieceData) firePeerGotPieceData (msgs, bytesWritten); if (tr_isPeerIo (io) && io->userData) peerPulse (msgs); } static ReadState canRead (tr_peerIo * io, void * vmsgs, size_t * piece) { ReadState ret; tr_peerMsgs * msgs = vmsgs; struct evbuffer * in = tr_peerIoGetReadBuffer (io); const size_t inlen = evbuffer_get_length (in); dbgmsg (msgs, "canRead: inlen is %"TR_PRIuSIZE", msgs->state is %d", inlen, msgs->state); if (!inlen) { ret = READ_LATER; } else if (msgs->state == AWAITING_BT_PIECE) { ret = readBtPiece (msgs, in, inlen, piece); } else switch (msgs->state) { case AWAITING_BT_LENGTH: ret = readBtLength (msgs, in, inlen); break; case AWAITING_BT_ID: ret = readBtId (msgs, in, inlen); break; case AWAITING_BT_MESSAGE: ret = readBtMessage (msgs, in, inlen); break; default: ret = READ_ERR; assert (0); } dbgmsg (msgs, "canRead: ret is %d", (int)ret); return ret; } int tr_peerMsgsIsReadingBlock (const tr_peerMsgs * msgs, tr_block_index_t block) { if (msgs->state != AWAITING_BT_PIECE) return false; return block == _tr_block (msgs->torrent, msgs->incoming.blockReq.index, msgs->incoming.blockReq.offset); } /** *** **/ static void updateDesiredRequestCount (tr_peerMsgs * msgs) { tr_torrent * const torrent = msgs->torrent; /* there are lots of reasons we might not want to request any blocks... */ if (tr_torrentIsSeed (torrent) || !tr_torrentHasMetadata (torrent) || msgs->client_is_choked || !msgs->client_is_interested) { msgs->desiredRequestCount = 0; } else { int estimatedBlocksInPeriod; unsigned int rate_Bps; unsigned int irate_Bps; const int floor = 4; const int seconds = REQUEST_BUF_SECS; const uint64_t now = tr_time_msec (); /* Get the rate limit we should use. * FIXME: this needs to consider all the other peers as well... */ rate_Bps = tr_peerGetPieceSpeed_Bps (&msgs->peer, now, TR_PEER_TO_CLIENT); if (tr_torrentUsesSpeedLimit (torrent, TR_PEER_TO_CLIENT)) rate_Bps = MIN (rate_Bps, tr_torrentGetSpeedLimit_Bps (torrent, TR_PEER_TO_CLIENT)); /* honor the session limits, if enabled */ if (tr_torrentUsesSessionLimits (torrent) && tr_sessionGetActiveSpeedLimit_Bps (torrent->session, TR_PEER_TO_CLIENT, &irate_Bps)) rate_Bps = MIN (rate_Bps, irate_Bps); /* use this desired rate to figure out how * many requests we should send to this peer */ estimatedBlocksInPeriod = (rate_Bps * seconds) / torrent->blockSize; msgs->desiredRequestCount = MAX (floor, estimatedBlocksInPeriod); /* honor the peer's maximum request count, if specified */ if (msgs->reqq > 0) if (msgs->desiredRequestCount > msgs->reqq) msgs->desiredRequestCount = msgs->reqq; } } static void updateMetadataRequests (tr_peerMsgs * msgs, time_t now) { int piece; if (msgs->peerSupportsMetadataXfer && tr_torrentGetNextMetadataRequest (msgs->torrent, now, &piece)) { tr_variant tmp; struct evbuffer * payload; struct evbuffer * out = msgs->outMessages; /* build the data message */ tr_variantInitDict (&tmp, 3); tr_variantDictAddInt (&tmp, TR_KEY_msg_type, METADATA_MSG_TYPE_REQUEST); tr_variantDictAddInt (&tmp, TR_KEY_piece, piece); payload = tr_variantToBuf (&tmp, TR_VARIANT_FMT_BENC); dbgmsg (msgs, "requesting metadata piece #%d", piece); /* write it out as a LTEP message to our outMessages buffer */ evbuffer_add_uint32 (out, 2 * sizeof (uint8_t) + evbuffer_get_length (payload)); evbuffer_add_uint8 (out, BT_LTEP); evbuffer_add_uint8 (out, msgs->ut_metadata_id); evbuffer_add_buffer (out, payload); pokeBatchPeriod (msgs, HIGH_PRIORITY_INTERVAL_SECS); dbgOutMessageLen (msgs); /* cleanup */ evbuffer_free (payload); tr_variantFree (&tmp); } } static void updateBlockRequests (tr_peerMsgs * msgs) { if (tr_torrentIsPieceTransferAllowed (msgs->torrent, TR_PEER_TO_CLIENT) && (msgs->desiredRequestCount > 0) && (msgs->peer.pendingReqsToPeer <= (msgs->desiredRequestCount * 0.66))) { int i; int n; tr_block_index_t * blocks; const int numwant = msgs->desiredRequestCount - msgs->peer.pendingReqsToPeer; assert (tr_peerMsgsIsClientInterested (msgs)); assert (!tr_peerMsgsIsClientChoked (msgs)); blocks = tr_new (tr_block_index_t, numwant); tr_peerMgrGetNextRequests (msgs->torrent, &msgs->peer, numwant, blocks, &n, false); for (i=0; itorrent, blocks[i], &req); protocolSendRequest (msgs, &req); } tr_free (blocks); } } static size_t fillOutputBuffer (tr_peerMsgs * msgs, time_t now) { int piece; size_t bytesWritten = 0; struct peer_request req; const bool haveMessages = evbuffer_get_length (msgs->outMessages) != 0; const bool fext = tr_peerIoSupportsFEXT (msgs->io); /** *** Protocol messages **/ if (haveMessages && !msgs->outMessagesBatchedAt) /* fresh batch */ { dbgmsg (msgs, "started an outMessages batch (length is %"TR_PRIuSIZE")", evbuffer_get_length (msgs->outMessages)); msgs->outMessagesBatchedAt = now; } else if (haveMessages && ((now - msgs->outMessagesBatchedAt) >= msgs->outMessagesBatchPeriod)) { const size_t len = evbuffer_get_length (msgs->outMessages); /* flush the protocol messages */ dbgmsg (msgs, "flushing outMessages... to %p (length is %"TR_PRIuSIZE")", (void*)msgs->io, len); tr_peerIoWriteBuf (msgs->io, msgs->outMessages, false); msgs->clientSentAnythingAt = now; msgs->outMessagesBatchedAt = 0; msgs->outMessagesBatchPeriod = LOW_PRIORITY_INTERVAL_SECS; bytesWritten += len; } /** *** Metadata Pieces **/ if ((tr_peerIoGetWriteBufferSpace (msgs->io, now) >= METADATA_PIECE_SIZE) && popNextMetadataRequest (msgs, &piece)) { char * data; int dataLen; bool ok = false; data = tr_torrentGetMetadataPiece (msgs->torrent, piece, &dataLen); if ((dataLen > 0) && (data != NULL)) { tr_variant tmp; struct evbuffer * payload; struct evbuffer * out = msgs->outMessages; /* build the data message */ tr_variantInitDict (&tmp, 3); tr_variantDictAddInt (&tmp, TR_KEY_msg_type, METADATA_MSG_TYPE_DATA); tr_variantDictAddInt (&tmp, TR_KEY_piece, piece); tr_variantDictAddInt (&tmp, TR_KEY_total_size, msgs->torrent->infoDictLength); payload = tr_variantToBuf (&tmp, TR_VARIANT_FMT_BENC); /* write it out as a LTEP message to our outMessages buffer */ evbuffer_add_uint32 (out, 2 * sizeof (uint8_t) + evbuffer_get_length (payload) + dataLen); evbuffer_add_uint8 (out, BT_LTEP); evbuffer_add_uint8 (out, msgs->ut_metadata_id); evbuffer_add_buffer (out, payload); evbuffer_add (out, data, dataLen); pokeBatchPeriod (msgs, HIGH_PRIORITY_INTERVAL_SECS); dbgOutMessageLen (msgs); evbuffer_free (payload); tr_variantFree (&tmp); tr_free (data); ok = true; } if (!ok) /* send a rejection message */ { tr_variant tmp; struct evbuffer * payload; struct evbuffer * out = msgs->outMessages; /* build the rejection message */ tr_variantInitDict (&tmp, 2); tr_variantDictAddInt (&tmp, TR_KEY_msg_type, METADATA_MSG_TYPE_REJECT); tr_variantDictAddInt (&tmp, TR_KEY_piece, piece); payload = tr_variantToBuf (&tmp, TR_VARIANT_FMT_BENC); /* write it out as a LTEP message to our outMessages buffer */ evbuffer_add_uint32 (out, 2 * sizeof (uint8_t) + evbuffer_get_length (payload)); evbuffer_add_uint8 (out, BT_LTEP); evbuffer_add_uint8 (out, msgs->ut_metadata_id); evbuffer_add_buffer (out, payload); pokeBatchPeriod (msgs, HIGH_PRIORITY_INTERVAL_SECS); dbgOutMessageLen (msgs); evbuffer_free (payload); tr_variantFree (&tmp); } } /** *** Data Blocks **/ if ((tr_peerIoGetWriteBufferSpace (msgs->io, now) >= msgs->torrent->blockSize) && popNextRequest (msgs, &req)) { --msgs->prefetchCount; if (requestIsValid (msgs, &req) && tr_torrentPieceIsComplete (msgs->torrent, req.index)) { int err; const uint32_t msglen = 4 + 1 + 4 + 4 + req.length; struct evbuffer * out; struct evbuffer_iovec iovec[1]; out = evbuffer_new (); evbuffer_expand (out, msglen); evbuffer_add_uint32 (out, sizeof (uint8_t) + 2 * sizeof (uint32_t) + req.length); evbuffer_add_uint8 (out, BT_PIECE); evbuffer_add_uint32 (out, req.index); evbuffer_add_uint32 (out, req.offset); evbuffer_reserve_space (out, req.length, iovec, 1); err = tr_cacheReadBlock (getSession (msgs)->cache, msgs->torrent, req.index, req.offset, req.length, iovec[0].iov_base); iovec[0].iov_len = req.length; evbuffer_commit_space (out, iovec, 1); /* check the piece if it needs checking... */ if (!err && tr_torrentPieceNeedsCheck (msgs->torrent, req.index)) if ((err = !tr_torrentCheckPiece (msgs->torrent, req.index))) tr_torrentSetLocalError (msgs->torrent, _("Please Verify Local Data! Piece #%"TR_PRIuSIZE" is corrupt."), (size_t)req.index); if (err) { if (fext) protocolSendReject (msgs, &req); } else { const size_t n = evbuffer_get_length (out); dbgmsg (msgs, "sending block %u:%u->%u", req.index, req.offset, req.length); assert (n == msglen); tr_peerIoWriteBuf (msgs->io, out, true); bytesWritten += n; msgs->clientSentAnythingAt = now; tr_historyAdd (&msgs->peer.blocksSentToPeer, tr_time (), 1); } evbuffer_free (out); if (err) { bytesWritten = 0; msgs = NULL; } } else if (fext) /* peer needs a reject message */ { protocolSendReject (msgs, &req); } if (msgs != NULL) prefetchPieces (msgs); } /** *** Keepalive **/ if ((msgs != NULL) && (msgs->clientSentAnythingAt != 0) && ((now - msgs->clientSentAnythingAt) > KEEPALIVE_INTERVAL_SECS)) { dbgmsg (msgs, "sending a keepalive message"); evbuffer_add_uint32 (msgs->outMessages, 0); pokeBatchPeriod (msgs, IMMEDIATE_PRIORITY_INTERVAL_SECS); } return bytesWritten; } static int peerPulse (void * vmsgs) { tr_peerMsgs * msgs = vmsgs; const time_t now = tr_time (); if (tr_isPeerIo (msgs->io)) { updateDesiredRequestCount (msgs); updateBlockRequests (msgs); updateMetadataRequests (msgs, now); } for (;;) if (fillOutputBuffer (msgs, now) < 1) break; return true; /* loop forever */ } void tr_peerMsgsPulse (tr_peerMsgs * msgs) { if (msgs != NULL) peerPulse (msgs); } static void gotError (tr_peerIo * io UNUSED, short what, void * vmsgs) { if (what & BEV_EVENT_TIMEOUT) dbgmsg (vmsgs, "libevent got a timeout, what=%hd", what); if (what & (BEV_EVENT_EOF | BEV_EVENT_ERROR)) dbgmsg (vmsgs, "libevent got an error! what=%hd, errno=%d (%s)", what, errno, tr_strerror (errno)); fireError (vmsgs, ENOTCONN); } static void sendBitfield (tr_peerMsgs * msgs) { void * bytes; size_t byte_count = 0; struct evbuffer * out = msgs->outMessages; assert (tr_torrentHasMetadata (msgs->torrent)); bytes = tr_torrentCreatePieceBitfield (msgs->torrent, &byte_count); evbuffer_add_uint32 (out, sizeof (uint8_t) + byte_count); evbuffer_add_uint8 (out, BT_BITFIELD); evbuffer_add (out, bytes, byte_count); dbgmsg (msgs, "sending bitfield... outMessage size is now %"TR_PRIuSIZE, evbuffer_get_length (out)); pokeBatchPeriod (msgs, IMMEDIATE_PRIORITY_INTERVAL_SECS); tr_free (bytes); } static void tellPeerWhatWeHave (tr_peerMsgs * msgs) { const bool fext = tr_peerIoSupportsFEXT (msgs->io); if (fext && tr_torrentHasAll (msgs->torrent)) { protocolSendHaveAll (msgs); } else if (fext && tr_torrentHasNone (msgs->torrent)) { protocolSendHaveNone (msgs); } else if (!tr_torrentHasNone (msgs->torrent)) { sendBitfield (msgs); } } /** *** **/ /* some peers give us error messages if we send more than this many peers in a single pex message http://wiki.theory.org/BitTorrentPeerExchangeConventions */ #define MAX_PEX_ADDED 50 #define MAX_PEX_DROPPED 50 typedef struct { tr_pex * added; tr_pex * dropped; tr_pex * elements; int addedCount; int droppedCount; int elementCount; } PexDiffs; static void pexAddedCb (void * vpex, void * userData) { PexDiffs * diffs = userData; tr_pex * pex = vpex; if (diffs->addedCount < MAX_PEX_ADDED) { diffs->added[diffs->addedCount++] = *pex; diffs->elements[diffs->elementCount++] = *pex; } } static inline void pexDroppedCb (void * vpex, void * userData) { PexDiffs * diffs = userData; tr_pex * pex = vpex; if (diffs->droppedCount < MAX_PEX_DROPPED) { diffs->dropped[diffs->droppedCount++] = *pex; } } static inline void pexElementCb (void * vpex, void * userData) { PexDiffs * diffs = userData; tr_pex * pex = vpex; diffs->elements[diffs->elementCount++] = *pex; } typedef void (tr_set_func)(void * element, void * userData); /** * @brief find the differences and commonalities in two sorted sets * @param a the first set * @param aCount the number of elements in the set 'a' * @param b the second set * @param bCount the number of elements in the set 'b' * @param compare the sorting method for both sets * @param elementSize the sizeof the element in the two sorted sets * @param in_a called for items in set 'a' but not set 'b' * @param in_b called for items in set 'b' but not set 'a' * @param in_both called for items that are in both sets * @param userData user data passed along to in_a, in_b, and in_both */ static void tr_set_compare (const void * va, size_t aCount, const void * vb, size_t bCount, int compare (const void * a, const void * b), size_t elementSize, tr_set_func in_a_cb, tr_set_func in_b_cb, tr_set_func in_both_cb, void * userData) { const uint8_t * a = va; const uint8_t * b = vb; const uint8_t * aend = a + elementSize * aCount; const uint8_t * bend = b + elementSize * bCount; while (a != aend || b != bend) { if (a == aend) { (*in_b_cb)((void*)b, userData); b += elementSize; } else if (b == bend) { (*in_a_cb)((void*)a, userData); a += elementSize; } else { const int val = (*compare)(a, b); if (!val) { (*in_both_cb)((void*)a, userData); a += elementSize; b += elementSize; } else if (val < 0) { (*in_a_cb)((void*)a, userData); a += elementSize; } else if (val > 0) { (*in_b_cb)((void*)b, userData); b += elementSize; } } } } static void sendPex (tr_peerMsgs * msgs) { if (msgs->peerSupportsPex && tr_torrentAllowsPex (msgs->torrent)) { PexDiffs diffs; PexDiffs diffs6; tr_pex * newPex = NULL; tr_pex * newPex6 = NULL; const int newCount = tr_peerMgrGetPeers (msgs->torrent, &newPex, TR_AF_INET, TR_PEERS_CONNECTED, MAX_PEX_PEER_COUNT); const int newCount6 = tr_peerMgrGetPeers (msgs->torrent, &newPex6, TR_AF_INET6, TR_PEERS_CONNECTED, MAX_PEX_PEER_COUNT); /* build the diffs */ diffs.added = tr_new (tr_pex, newCount); diffs.addedCount = 0; diffs.dropped = tr_new (tr_pex, msgs->pexCount); diffs.droppedCount = 0; diffs.elements = tr_new (tr_pex, newCount + msgs->pexCount); diffs.elementCount = 0; tr_set_compare (msgs->pex, msgs->pexCount, newPex, newCount, tr_pexCompare, sizeof (tr_pex), pexDroppedCb, pexAddedCb, pexElementCb, &diffs); diffs6.added = tr_new (tr_pex, newCount6); diffs6.addedCount = 0; diffs6.dropped = tr_new (tr_pex, msgs->pexCount6); diffs6.droppedCount = 0; diffs6.elements = tr_new (tr_pex, newCount6 + msgs->pexCount6); diffs6.elementCount = 0; tr_set_compare (msgs->pex6, msgs->pexCount6, newPex6, newCount6, tr_pexCompare, sizeof (tr_pex), pexDroppedCb, pexAddedCb, pexElementCb, &diffs6); dbgmsg ( msgs, "pex: old peer count %d+%d, new peer count %d+%d, " "added %d+%d, removed %d+%d", msgs->pexCount, msgs->pexCount6, newCount, newCount6, diffs.addedCount, diffs6.addedCount, diffs.droppedCount, diffs6.droppedCount); if (!diffs.addedCount && !diffs.droppedCount && !diffs6.addedCount && !diffs6.droppedCount) { tr_free (diffs.elements); tr_free (diffs6.elements); } else { int i; tr_variant val; uint8_t * tmp, *walk; struct evbuffer * payload; struct evbuffer * out = msgs->outMessages; /* update peer */ tr_free (msgs->pex); msgs->pex = diffs.elements; msgs->pexCount = diffs.elementCount; tr_free (msgs->pex6); msgs->pex6 = diffs6.elements; msgs->pexCount6 = diffs6.elementCount; /* build the pex payload */ tr_variantInitDict (&val, 3); /* ipv6 support: left as 3: * speed vs. likelihood? */ if (diffs.addedCount > 0) { /* "added" */ tmp = walk = tr_new (uint8_t, diffs.addedCount * 6); for (i = 0; i < diffs.addedCount; ++i) { memcpy (walk, &diffs.added[i].addr.addr, 4); walk += 4; memcpy (walk, &diffs.added[i].port, 2); walk += 2; } assert ((walk - tmp) == diffs.addedCount * 6); tr_variantDictAddRaw (&val, TR_KEY_added, tmp, walk - tmp); tr_free (tmp); /* "added.f" * unset each holepunch flag because we don't support it. */ tmp = walk = tr_new (uint8_t, diffs.addedCount); for (i = 0; i < diffs.addedCount; ++i) *walk++ = diffs.added[i].flags & ~ADDED_F_HOLEPUNCH; assert ((walk - tmp) == diffs.addedCount); tr_variantDictAddRaw (&val, TR_KEY_added_f, tmp, walk - tmp); tr_free (tmp); } if (diffs.droppedCount > 0) { /* "dropped" */ tmp = walk = tr_new (uint8_t, diffs.droppedCount * 6); for (i = 0; i < diffs.droppedCount; ++i) { memcpy (walk, &diffs.dropped[i].addr.addr, 4); walk += 4; memcpy (walk, &diffs.dropped[i].port, 2); walk += 2; } assert ((walk - tmp) == diffs.droppedCount * 6); tr_variantDictAddRaw (&val, TR_KEY_dropped, tmp, walk - tmp); tr_free (tmp); } if (diffs6.addedCount > 0) { /* "added6" */ tmp = walk = tr_new (uint8_t, diffs6.addedCount * 18); for (i = 0; i < diffs6.addedCount; ++i) { memcpy (walk, &diffs6.added[i].addr.addr.addr6.s6_addr, 16); walk += 16; memcpy (walk, &diffs6.added[i].port, 2); walk += 2; } assert ((walk - tmp) == diffs6.addedCount * 18); tr_variantDictAddRaw (&val, TR_KEY_added6, tmp, walk - tmp); tr_free (tmp); /* "added6.f" * unset each holepunch flag because we don't support it. */ tmp = walk = tr_new (uint8_t, diffs6.addedCount); for (i = 0; i < diffs6.addedCount; ++i) *walk++ = diffs6.added[i].flags & ~ADDED_F_HOLEPUNCH; assert ((walk - tmp) == diffs6.addedCount); tr_variantDictAddRaw (&val, TR_KEY_added6_f, tmp, walk - tmp); tr_free (tmp); } if (diffs6.droppedCount > 0) { /* "dropped6" */ tmp = walk = tr_new (uint8_t, diffs6.droppedCount * 18); for (i = 0; i < diffs6.droppedCount; ++i) { memcpy (walk, &diffs6.dropped[i].addr.addr.addr6.s6_addr, 16); walk += 16; memcpy (walk, &diffs6.dropped[i].port, 2); walk += 2; } assert ((walk - tmp) == diffs6.droppedCount * 18); tr_variantDictAddRaw (&val, TR_KEY_dropped6, tmp, walk - tmp); tr_free (tmp); } /* write the pex message */ payload = tr_variantToBuf (&val, TR_VARIANT_FMT_BENC); evbuffer_add_uint32 (out, 2 * sizeof (uint8_t) + evbuffer_get_length (payload)); evbuffer_add_uint8 (out, BT_LTEP); evbuffer_add_uint8 (out, msgs->ut_pex_id); evbuffer_add_buffer (out, payload); pokeBatchPeriod (msgs, HIGH_PRIORITY_INTERVAL_SECS); dbgmsg (msgs, "sending a pex message; outMessage size is now %"TR_PRIuSIZE, evbuffer_get_length (out)); dbgOutMessageLen (msgs); evbuffer_free (payload); tr_variantFree (&val); } /* cleanup */ tr_free (diffs.added); tr_free (diffs.dropped); tr_free (newPex); tr_free (diffs6.added); tr_free (diffs6.dropped); tr_free (newPex6); /*msgs->clientSentPexAt = tr_time ();*/ } } static void pexPulse (evutil_socket_t foo UNUSED, short bar UNUSED, void * vmsgs) { struct tr_peerMsgs * msgs = vmsgs; sendPex (msgs); assert (msgs->pexTimer != NULL); tr_timerAdd (msgs->pexTimer, PEX_INTERVAL_SECS, 0); } /*** **** tr_peer virtual functions ***/ static bool peermsgs_is_transferring_pieces (const struct tr_peer * peer, uint64_t now, tr_direction direction, unsigned int * setme_Bps) { unsigned int Bps = 0; if (tr_isPeerMsgs (peer)) { const tr_peerMsgs * msgs = (const tr_peerMsgs *) peer; Bps = tr_peerIoGetPieceSpeed_Bps (msgs->io, now, direction); } if (setme_Bps != NULL) *setme_Bps = Bps; return Bps > 0; } static void peermsgs_destruct (tr_peer * peer) { tr_peerMsgs * msgs = PEER_MSGS (peer); assert (msgs != NULL); tr_peerMsgsSetActive (msgs, TR_UP, false); tr_peerMsgsSetActive (msgs, TR_DOWN, false); if (msgs->pexTimer != NULL) event_free (msgs->pexTimer); if (msgs->incoming.block != NULL) evbuffer_free (msgs->incoming.block); if (msgs->io) { tr_peerIoClear (msgs->io); tr_peerIoUnref (msgs->io); /* balanced by the ref in handshakeDoneCB () */ } evbuffer_free (msgs->outMessages); tr_free (msgs->pex6); tr_free (msgs->pex); tr_peerDestruct (&msgs->peer); memset (msgs, ~0, sizeof (tr_peerMsgs)); } static const struct tr_peer_virtual_funcs my_funcs = { .destruct = peermsgs_destruct, .is_transferring_pieces = peermsgs_is_transferring_pieces }; /*** **** ***/ time_t tr_peerMsgsGetConnectionAge (const tr_peerMsgs * msgs) { assert (tr_isPeerMsgs (msgs)); return tr_peerIoGetAge (msgs->io); } bool tr_peerMsgsIsPeerChoked (const tr_peerMsgs * msgs) { assert (tr_isPeerMsgs (msgs)); return msgs->peer_is_choked; } bool tr_peerMsgsIsPeerInterested (const tr_peerMsgs * msgs) { assert (tr_isPeerMsgs (msgs)); return msgs->peer_is_interested; } bool tr_peerMsgsIsClientChoked (const tr_peerMsgs * msgs) { assert (tr_isPeerMsgs (msgs)); return msgs->client_is_choked; } bool tr_peerMsgsIsClientInterested (const tr_peerMsgs * msgs) { assert (tr_isPeerMsgs (msgs)); return msgs->client_is_interested; } bool tr_peerMsgsIsUtpConnection (const tr_peerMsgs * msgs) { assert (tr_isPeerMsgs (msgs)); return msgs->io->utp_socket != NULL; } bool tr_peerMsgsIsEncrypted (const tr_peerMsgs * msgs) { assert (tr_isPeerMsgs (msgs)); return tr_peerIoIsEncrypted (msgs->io); } bool tr_peerMsgsIsIncomingConnection (const tr_peerMsgs * msgs) { assert (tr_isPeerMsgs (msgs)); return tr_peerIoIsIncoming (msgs->io); } /*** **** ***/ bool tr_isPeerMsgs (const void * msgs) { /* FIXME: this is pretty crude */ return (msgs != NULL) && (((struct tr_peerMsgs*)msgs)->magic_number == MAGIC_NUMBER); } tr_peerMsgs * tr_peerMsgsCast (void * vm) { return tr_isPeerMsgs(vm) ? vm : NULL; } tr_peerMsgs * tr_peerMsgsNew (struct tr_torrent * torrent, struct tr_peerIo * io, tr_peer_callback callback, void * callbackData) { tr_peerMsgs * m; assert (io != NULL); m = tr_new0 (tr_peerMsgs, 1); tr_peerConstruct (&m->peer, torrent); m->peer.funcs = &my_funcs; m->magic_number = MAGIC_NUMBER; m->client_is_choked = true; m->peer_is_choked = true; m->client_is_interested = false; m->peer_is_interested = false; m->is_active[TR_UP] = false; m->is_active[TR_DOWN] = false; m->callback = callback; m->callbackData = callbackData; m->io = io; m->torrent = torrent; m->state = AWAITING_BT_LENGTH; m->outMessages = evbuffer_new (); m->outMessagesBatchedAt = 0; m->outMessagesBatchPeriod = LOW_PRIORITY_INTERVAL_SECS; if (tr_torrentAllowsPex (torrent)) { m->pexTimer = evtimer_new (torrent->session->event_base, pexPulse, m); tr_timerAdd (m->pexTimer, PEX_INTERVAL_SECS, 0); } if (tr_peerIoSupportsUTP (m->io)) { const tr_address * addr = tr_peerIoGetAddress (m->io, NULL); tr_peerMgrSetUtpSupported (torrent, addr); tr_peerMgrSetUtpFailed (torrent, addr, false); } if (tr_peerIoSupportsLTEP (m->io)) sendLtepHandshake (m); tellPeerWhatWeHave (m); if (tr_dhtEnabled (torrent->session) && tr_peerIoSupportsDHT (m->io)) { /* Only send PORT over IPv6 when the IPv6 DHT is running (BEP-32). */ const struct tr_address *addr = tr_peerIoGetAddress (m->io, NULL); if (addr->type == TR_AF_INET || tr_globalIPv6 ()) protocolSendPort (m, tr_dhtPort (torrent->session)); } tr_peerIoSetIOFuncs (m->io, canRead, didWrite, gotError, m); updateDesiredRequestCount (m); return m; }