2513 lines
67 KiB
C++
2513 lines
67 KiB
C++
// This file Copyright © 2007-2023 Mnemosyne LLC.
|
|
// It may be used under GPLv2 (SPDX: GPL-2.0-only), GPLv3 (SPDX: GPL-3.0-only),
|
|
// or any future license endorsed by Mnemosyne LLC.
|
|
// License text can be found in the licenses/ folder.
|
|
|
|
#include <algorithm>
|
|
#include <array>
|
|
#include <cerrno> /* error codes ERANGE, ... */
|
|
#include <chrono>
|
|
#include <climits> /* INT_MAX */
|
|
#include <cmath>
|
|
#include <cstdint>
|
|
#include <ctime> // time_t
|
|
#include <deque>
|
|
#include <map>
|
|
#include <iterator> // std::back_inserter
|
|
#include <memory>
|
|
#include <optional>
|
|
#include <tuple> // std::tie
|
|
#include <utility>
|
|
#include <vector>
|
|
|
|
#include <fmt/format.h>
|
|
|
|
#define LIBTRANSMISSION_PEER_MODULE
|
|
#include "transmission.h"
|
|
|
|
#include "announcer.h"
|
|
#include "bandwidth.h"
|
|
#include "blocklist.h"
|
|
#include "cache.h"
|
|
#include "clients.h"
|
|
#include "completion.h"
|
|
#include "crypto-utils.h"
|
|
#include "handshake.h"
|
|
#include "log.h"
|
|
#include "net.h"
|
|
#include "peer-io.h"
|
|
#include "peer-mgr-active-requests.h"
|
|
#include "peer-mgr-wishlist.h"
|
|
#include "peer-mgr.h"
|
|
#include "peer-msgs.h"
|
|
#include "session.h"
|
|
#include "timer.h"
|
|
#include "torrent.h"
|
|
#include "torrent-magnet.h"
|
|
#include "tr-assert.h"
|
|
#include "tr-utp.h"
|
|
#include "utils.h"
|
|
#include "webseed.h"
|
|
|
|
using namespace std::literals;
|
|
|
|
// use for bitwise operations w/peer_atom.flags2
|
|
static auto constexpr MyflagBanned = int{ 1 };
|
|
|
|
// use for bitwise operations w/peer_atom.flags2
|
|
// unreachable for now... but not banned.
|
|
// if they try to connect to us it's okay
|
|
static auto constexpr MyflagUnreachable = int{ 2 };
|
|
|
|
static auto constexpr CancelHistorySec = int{ 60 };
|
|
|
|
// ---
|
|
|
|
class HandshakeMediator final : public tr_handshake::Mediator
|
|
{
|
|
private:
|
|
[[nodiscard]] static std::optional<TorrentInfo> torrent(tr_torrent* tor)
|
|
{
|
|
if (tor == nullptr)
|
|
{
|
|
return {};
|
|
}
|
|
|
|
auto info = TorrentInfo{};
|
|
info.info_hash = tor->infoHash();
|
|
info.client_peer_id = tor->peer_id();
|
|
info.id = tor->id();
|
|
info.is_done = tor->isDone();
|
|
return info;
|
|
}
|
|
|
|
public:
|
|
explicit HandshakeMediator(tr_session& session) noexcept
|
|
: session_{ session }
|
|
{
|
|
}
|
|
|
|
[[nodiscard]] std::optional<TorrentInfo> torrent(tr_sha1_digest_t const& info_hash) const override
|
|
{
|
|
return torrent(session_.torrents().get(info_hash));
|
|
}
|
|
|
|
[[nodiscard]] std::optional<TorrentInfo> torrent_from_obfuscated(tr_sha1_digest_t const& info_hash) const override
|
|
{
|
|
return torrent(tr_torrentFindFromObfuscatedHash(&session_, info_hash));
|
|
}
|
|
|
|
[[nodiscard]] bool allows_dht() const override
|
|
{
|
|
return session_.allowsDHT();
|
|
}
|
|
|
|
[[nodiscard]] bool allows_tcp() const override
|
|
{
|
|
return session_.allowsTCP();
|
|
}
|
|
|
|
void set_utp_failed(tr_sha1_digest_t const& info_hash, tr_address const& addr) override
|
|
{
|
|
if (auto* const tor = session_.torrents().get(info_hash); tor != nullptr)
|
|
{
|
|
tr_peerMgrSetUtpFailed(tor, addr, true);
|
|
}
|
|
}
|
|
|
|
[[nodiscard]] bool is_peer_known_seed(tr_torrent_id_t tor_id, tr_address const& addr) const override;
|
|
|
|
[[nodiscard]] libtransmission::TimerMaker& timer_maker() override
|
|
{
|
|
return session_.timerMaker();
|
|
}
|
|
|
|
[[nodiscard]] size_t pad(void* setme, size_t maxlen) const override
|
|
{
|
|
auto const len = tr_rand_int(maxlen);
|
|
tr_rand_buffer(setme, len);
|
|
return len;
|
|
}
|
|
|
|
private:
|
|
tr_session& session_;
|
|
};
|
|
|
|
/**
|
|
* Peer information that should be kept even before we've connected and
|
|
* after we've disconnected. These are kept in a pool of peer_atoms to decide
|
|
* which ones would make good candidates for connecting to, and to watch out
|
|
* for banned peers.
|
|
*
|
|
* @see tr_peer
|
|
* @see tr_peerMsgs
|
|
*/
|
|
struct peer_atom
|
|
{
|
|
peer_atom(tr_address addr_in, tr_port port_in, uint8_t flags_in, uint8_t from)
|
|
: addr{ addr_in }
|
|
, port{ port_in }
|
|
, fromFirst{ from }
|
|
, fromBest{ from }
|
|
, flags{ flags_in }
|
|
{
|
|
++n_atoms;
|
|
}
|
|
|
|
peer_atom(peer_atom&&) = delete;
|
|
peer_atom(peer_atom const&) = delete;
|
|
peer_atom& operator=(peer_atom&&) = delete;
|
|
peer_atom& operator=(peer_atom const&) = delete;
|
|
|
|
~peer_atom()
|
|
{
|
|
[[maybe_unused]] auto const n_prev = n_atoms--;
|
|
TR_ASSERT(n_prev > 0U);
|
|
}
|
|
|
|
[[nodiscard]] static auto atom_count() noexcept
|
|
{
|
|
return n_atoms.load();
|
|
}
|
|
|
|
[[nodiscard]] constexpr auto isSeed() const noexcept
|
|
{
|
|
return (flags & ADDED_F_SEED_FLAG) != 0;
|
|
}
|
|
|
|
[[nodiscard]] auto display_name() const
|
|
{
|
|
return addr.display_name(port);
|
|
}
|
|
|
|
[[nodiscard]] bool isBlocklisted(tr_session const* session) const
|
|
{
|
|
if (blocklisted_)
|
|
{
|
|
return *blocklisted_;
|
|
}
|
|
|
|
auto const value = session->addressIsBlocked(addr);
|
|
blocklisted_ = value;
|
|
return value;
|
|
}
|
|
|
|
[[nodiscard]] constexpr int getReconnectIntervalSecs(time_t const now) const noexcept
|
|
{
|
|
auto sec = int{};
|
|
bool const unreachable = (this->flags2 & MyflagUnreachable) != 0;
|
|
|
|
/* if we were recently connected to this peer and transferring piece
|
|
* data, try to reconnect to them sooner rather that later -- we don't
|
|
* want network troubles to get in the way of a good peer. */
|
|
if (!unreachable && now - this->piece_data_time <= MinimumReconnectIntervalSecs * 2)
|
|
{
|
|
sec = MinimumReconnectIntervalSecs;
|
|
}
|
|
/* otherwise, the interval depends on how many times we've tried
|
|
* and failed to connect to the peer */
|
|
else
|
|
{
|
|
auto step = this->num_fails;
|
|
|
|
/* penalize peers that were unreachable the last time we tried */
|
|
if (unreachable)
|
|
{
|
|
step += 2;
|
|
}
|
|
|
|
switch (step)
|
|
{
|
|
case 0:
|
|
sec = 0;
|
|
break;
|
|
|
|
case 1:
|
|
sec = 10;
|
|
break;
|
|
|
|
case 2:
|
|
sec = 60 * 2;
|
|
break;
|
|
|
|
case 3:
|
|
sec = 60 * 15;
|
|
break;
|
|
|
|
case 4:
|
|
sec = 60 * 30;
|
|
break;
|
|
|
|
case 5:
|
|
sec = 60 * 60;
|
|
break;
|
|
|
|
default:
|
|
sec = 60 * 120;
|
|
break;
|
|
}
|
|
}
|
|
|
|
return sec;
|
|
}
|
|
|
|
void setBlocklistedDirty()
|
|
{
|
|
blocklisted_.reset();
|
|
}
|
|
|
|
[[nodiscard]] constexpr std::optional<bool> isReachable() const
|
|
{
|
|
if ((flags2 & MyflagUnreachable) != 0)
|
|
{
|
|
return false;
|
|
}
|
|
|
|
if ((flags & ADDED_F_CONNECTABLE) != 0)
|
|
{
|
|
return true;
|
|
}
|
|
|
|
return std::nullopt;
|
|
}
|
|
|
|
tr_address const addr;
|
|
|
|
tr_port port = {};
|
|
|
|
uint16_t num_fails = {};
|
|
|
|
time_t time = {}; /* when the peer's connection status last changed */
|
|
time_t piece_data_time = {};
|
|
|
|
time_t lastConnectionAttemptAt = {};
|
|
time_t lastConnectionAt = {};
|
|
|
|
uint8_t const fromFirst; /* where the peer was first found */
|
|
uint8_t fromBest; /* the "best" value of where the peer has been found */
|
|
uint8_t flags = {}; /* these match the added_f flags */
|
|
uint8_t flags2 = {}; /* flags that aren't defined in added_f */
|
|
|
|
bool utp_failed = false; /* We recently failed to connect over µTP */
|
|
bool is_connected = false;
|
|
|
|
private:
|
|
mutable std::optional<bool> blocklisted_;
|
|
|
|
// the minimum we'll wait before attempting to reconnect to a peer
|
|
static auto constexpr MinimumReconnectIntervalSecs = int{ 5 };
|
|
|
|
static auto inline n_atoms = std::atomic<size_t>{};
|
|
};
|
|
|
|
using Handshakes = std::map<tr_address, tr_handshake>;
|
|
|
|
#define tr_logAddDebugSwarm(swarm, msg) tr_logAddDebugTor((swarm)->tor, msg)
|
|
#define tr_logAddTraceSwarm(swarm, msg) tr_logAddTraceTor((swarm)->tor, msg)
|
|
|
|
/** @brief Opaque, per-torrent data structure for peer connection information */
|
|
class tr_swarm
|
|
{
|
|
public:
|
|
[[nodiscard]] auto unique_lock() const
|
|
{
|
|
return tor->unique_lock();
|
|
}
|
|
|
|
tr_swarm(tr_peerMgr* manager_in, tr_torrent* tor_in) noexcept
|
|
: manager{ manager_in }
|
|
, tor{ tor_in }
|
|
{
|
|
rebuildWebseeds();
|
|
}
|
|
|
|
tr_swarm(tr_swarm&&) = delete;
|
|
tr_swarm(tr_swarm const&) = delete;
|
|
tr_swarm& operator=(tr_swarm&&) = delete;
|
|
tr_swarm& operator=(tr_swarm const&) = delete;
|
|
|
|
~tr_swarm()
|
|
{
|
|
auto const lock = unique_lock();
|
|
TR_ASSERT(!is_running);
|
|
TR_ASSERT(std::empty(outgoing_handshakes));
|
|
TR_ASSERT(std::empty(peers));
|
|
}
|
|
|
|
[[nodiscard]] bool peer_is_in_use(peer_atom const& atom) const;
|
|
|
|
void cancelOldRequests()
|
|
{
|
|
auto const now = tr_time();
|
|
auto const oldest = now - RequestTtlSecs;
|
|
|
|
for (auto const& [block, peer] : active_requests.sentBefore(oldest))
|
|
{
|
|
maybeSendCancelRequest(peer, block, nullptr);
|
|
active_requests.remove(block, peer);
|
|
}
|
|
}
|
|
|
|
void cancelAllRequestsForBlock(tr_block_index_t block, tr_peer const* no_notify)
|
|
{
|
|
for (auto* peer : active_requests.remove(block))
|
|
{
|
|
maybeSendCancelRequest(peer, block, no_notify);
|
|
}
|
|
}
|
|
|
|
[[nodiscard]] uint16_t countActiveWebseeds(uint64_t now) const noexcept
|
|
{
|
|
if (!tor->isRunning || tor->isDone())
|
|
{
|
|
return {};
|
|
}
|
|
|
|
return std::count_if(
|
|
std::begin(webseeds),
|
|
std::end(webseeds),
|
|
[&now](auto const& webseed) { return webseed->isTransferringPieces(now, TR_DOWN, nullptr); });
|
|
}
|
|
|
|
[[nodiscard]] auto peerCount() const noexcept
|
|
{
|
|
return std::size(peers);
|
|
}
|
|
|
|
void stop()
|
|
{
|
|
auto const lock = unique_lock();
|
|
|
|
is_running = false;
|
|
removeAllPeers();
|
|
outgoing_handshakes.clear();
|
|
}
|
|
|
|
void removePeer(tr_peer* peer)
|
|
{
|
|
auto const lock = unique_lock();
|
|
|
|
auto* const atom = peer->atom;
|
|
TR_ASSERT(atom != nullptr);
|
|
|
|
atom->time = tr_time();
|
|
|
|
if (auto iter = std::find(std::begin(peers), std::end(peers), peer); iter != std::end(peers))
|
|
{
|
|
peers.erase(iter);
|
|
}
|
|
|
|
--stats.peer_count;
|
|
--stats.peer_from_count[atom->fromFirst];
|
|
|
|
TR_ASSERT(stats.peer_count == peerCount());
|
|
|
|
delete peer;
|
|
}
|
|
|
|
void removeAllPeers()
|
|
{
|
|
auto tmp = peers;
|
|
|
|
for (auto* peer : tmp)
|
|
{
|
|
removePeer(peer);
|
|
}
|
|
|
|
TR_ASSERT(stats.peer_count == 0);
|
|
}
|
|
|
|
void updateEndgame()
|
|
{
|
|
/* we consider ourselves to be in endgame if the number of bytes
|
|
we've got requested is >= the number of bytes left to download */
|
|
is_endgame_ = uint64_t(std::size(active_requests)) * tr_block_info::BlockSize >= tor->leftUntilDone();
|
|
}
|
|
|
|
[[nodiscard]] constexpr auto isEndgame() const noexcept
|
|
{
|
|
return is_endgame_;
|
|
}
|
|
|
|
void addStrike(tr_peer* peer) const
|
|
{
|
|
tr_logAddTraceSwarm(
|
|
this,
|
|
fmt::format("increasing peer {} strike count to {}", peer->display_name(), peer->strikes + 1));
|
|
|
|
if (++peer->strikes >= MaxBadPiecesPerPeer)
|
|
{
|
|
peer->atom->flags2 |= MyflagBanned;
|
|
peer->do_purge = true;
|
|
tr_logAddTraceSwarm(this, fmt::format("banning peer {}", peer->display_name()));
|
|
}
|
|
}
|
|
|
|
void rebuildWebseeds()
|
|
{
|
|
auto const n = tor->webseedCount();
|
|
|
|
webseeds.clear();
|
|
webseeds.reserve(n);
|
|
for (size_t i = 0; i < n; ++i)
|
|
{
|
|
webseeds.emplace_back(tr_webseedNew(tor, tor->webseed(i), &tr_swarm::peerCallbackFunc, this));
|
|
}
|
|
webseeds.shrink_to_fit();
|
|
|
|
stats.active_webseed_count = 0;
|
|
}
|
|
|
|
[[nodiscard]] TR_CONSTEXPR20 auto isAllSeeds() const noexcept
|
|
{
|
|
if (!pool_is_all_seeds_)
|
|
{
|
|
pool_is_all_seeds_ = std::all_of(std::begin(pool), std::end(pool), [](auto const& atom) { return atom.isSeed(); });
|
|
}
|
|
|
|
return *pool_is_all_seeds_;
|
|
}
|
|
|
|
void markAllSeedsFlagDirty() noexcept
|
|
{
|
|
pool_is_all_seeds_.reset();
|
|
}
|
|
|
|
[[nodiscard]] peer_atom* get_existing_atom(tr_address const& addr) noexcept
|
|
{
|
|
auto const iter = std::find_if(
|
|
std::begin(pool),
|
|
std::end(pool),
|
|
[&addr](auto const& atom) { return atom.addr == addr; });
|
|
return iter != std::end(pool) ? &*iter : nullptr;
|
|
}
|
|
|
|
[[nodiscard]] peer_atom const* get_existing_atom(tr_address const& addr) const noexcept
|
|
{
|
|
auto const iter = std::find_if(
|
|
std::begin(pool),
|
|
std::end(pool),
|
|
[&addr](auto const& atom) { return atom.addr == addr; });
|
|
return iter != std::end(pool) ? &*iter : nullptr;
|
|
}
|
|
|
|
[[nodiscard]] bool peer_is_a_seed(tr_address const& addr) const noexcept
|
|
{
|
|
auto const* const atom = get_existing_atom(addr);
|
|
return atom != nullptr && atom->isSeed();
|
|
}
|
|
|
|
peer_atom* ensure_atom_exists(tr_address const& addr, tr_port const port, uint8_t const flags, uint8_t const from)
|
|
{
|
|
TR_ASSERT(addr.is_valid());
|
|
TR_ASSERT(from < TR_PEER_FROM__MAX);
|
|
|
|
peer_atom* atom = get_existing_atom(addr);
|
|
|
|
if (atom == nullptr)
|
|
{
|
|
atom = &pool.emplace_back(addr, port, flags, from);
|
|
}
|
|
else
|
|
{
|
|
atom->fromBest = std::min(atom->fromBest, from);
|
|
atom->flags |= flags;
|
|
}
|
|
|
|
markAllSeedsFlagDirty();
|
|
|
|
return atom;
|
|
}
|
|
|
|
void mark_atom_as_seed(peer_atom& atom)
|
|
{
|
|
tr_logAddTraceSwarm(this, fmt::format("marking peer {} as a seed", atom.display_name()));
|
|
atom.flags |= ADDED_F_SEED_FLAG;
|
|
markAllSeedsFlagDirty();
|
|
}
|
|
|
|
static void peerCallbackFunc(tr_peer* peer, tr_peer_event const& event, void* vs)
|
|
{
|
|
TR_ASSERT(peer != nullptr);
|
|
auto* s = static_cast<tr_swarm*>(vs);
|
|
auto const lock = s->unique_lock();
|
|
|
|
switch (event.type)
|
|
{
|
|
case tr_peer_event::Type::ClientSentPieceData:
|
|
{
|
|
auto const now = tr_time();
|
|
auto* const tor = s->tor;
|
|
|
|
tor->uploadedCur += event.length;
|
|
tr_announcerAddBytes(tor, TR_ANN_UP, event.length);
|
|
tor->setDateActive(now);
|
|
tor->setDirty();
|
|
tor->session->addUploaded(event.length);
|
|
|
|
if (peer->atom != nullptr)
|
|
{
|
|
peer->atom->piece_data_time = now;
|
|
}
|
|
|
|
break;
|
|
}
|
|
|
|
case tr_peer_event::Type::ClientGotPieceData:
|
|
{
|
|
auto const now = tr_time();
|
|
auto* const tor = s->tor;
|
|
|
|
tor->downloadedCur += event.length;
|
|
tor->setDateActive(now);
|
|
tor->setDirty();
|
|
tor->session->addDownloaded(event.length);
|
|
|
|
if (peer->atom != nullptr)
|
|
{
|
|
peer->atom->piece_data_time = now;
|
|
}
|
|
|
|
break;
|
|
}
|
|
|
|
case tr_peer_event::Type::ClientGotHave:
|
|
case tr_peer_event::Type::ClientGotHaveAll:
|
|
case tr_peer_event::Type::ClientGotHaveNone:
|
|
case tr_peer_event::Type::ClientGotBitfield:
|
|
/* TODO: if we don't need these, should these events be removed? */
|
|
/* noop */
|
|
break;
|
|
|
|
case tr_peer_event::Type::ClientGotRej:
|
|
s->active_requests.remove(s->tor->pieceLoc(event.pieceIndex, event.offset).block, peer);
|
|
break;
|
|
|
|
case tr_peer_event::Type::ClientGotChoke:
|
|
s->active_requests.remove(peer);
|
|
break;
|
|
|
|
case tr_peer_event::Type::ClientGotPort:
|
|
if (peer->atom != nullptr)
|
|
{
|
|
peer->atom->port = event.port;
|
|
}
|
|
|
|
break;
|
|
|
|
case tr_peer_event::Type::ClientGotSuggest:
|
|
case tr_peer_event::Type::ClientGotAllowedFast:
|
|
// not currently supported
|
|
break;
|
|
|
|
case tr_peer_event::Type::ClientGotBlock:
|
|
{
|
|
auto* const tor = s->tor;
|
|
auto const loc = tor->pieceLoc(event.pieceIndex, event.offset);
|
|
s->cancelAllRequestsForBlock(loc.block, peer);
|
|
peer->blocks_sent_to_client.add(tr_time(), 1);
|
|
tr_torrentGotBlock(tor, loc.block);
|
|
break;
|
|
}
|
|
|
|
case tr_peer_event::Type::Error:
|
|
if (event.err == ERANGE || event.err == EMSGSIZE || event.err == ENOTCONN)
|
|
{
|
|
/* some protocol error from the peer */
|
|
peer->do_purge = true;
|
|
tr_logAddDebugSwarm(
|
|
s,
|
|
fmt::format(
|
|
"setting {} do_purge flag because we got an ERANGE, EMSGSIZE, or ENOTCONN error",
|
|
peer->display_name()));
|
|
}
|
|
else
|
|
{
|
|
tr_logAddDebugSwarm(s, fmt::format("unhandled error: {}", tr_strerror(event.err)));
|
|
}
|
|
|
|
break;
|
|
}
|
|
}
|
|
|
|
Handshakes outgoing_handshakes;
|
|
|
|
mutable tr_swarm_stats stats = {};
|
|
|
|
uint8_t optimistic_unchoke_time_scaler = 0;
|
|
|
|
bool is_running = false;
|
|
|
|
tr_peerMgr* const manager;
|
|
|
|
tr_torrent* const tor;
|
|
|
|
ActiveRequests active_requests;
|
|
|
|
// depends-on: active_requests
|
|
std::vector<std::unique_ptr<tr_peer>> webseeds;
|
|
|
|
// depends-on: active_requests
|
|
std::vector<tr_peerMsgs*> peers;
|
|
|
|
// tr_peers hold pointers to the items in this container,
|
|
// so use a deque instead of vector to prevent insertion from
|
|
// invalidating those pointers
|
|
std::deque<peer_atom> pool;
|
|
|
|
tr_peerMsgs* optimistic = nullptr; /* the optimistic peer, or nullptr if none */
|
|
|
|
time_t lastCancel = 0;
|
|
|
|
private:
|
|
static void maybeSendCancelRequest(tr_peer* peer, tr_block_index_t block, tr_peer const* muted)
|
|
{
|
|
auto* msgs = dynamic_cast<tr_peerMsgs*>(peer);
|
|
if (msgs != nullptr && msgs != muted)
|
|
{
|
|
peer->cancels_sent_to_peer.add(tr_time(), 1);
|
|
msgs->cancel_block_request(block);
|
|
}
|
|
}
|
|
|
|
// number of bad pieces a peer is allowed to send before we ban them
|
|
static auto constexpr MaxBadPiecesPerPeer = int{ 5 };
|
|
|
|
// how long we'll let requests we've made linger before we cancel them
|
|
static auto constexpr RequestTtlSecs = int{ 90 };
|
|
|
|
mutable std::optional<bool> pool_is_all_seeds_;
|
|
|
|
bool is_endgame_ = false;
|
|
};
|
|
|
|
struct tr_peerMgr
|
|
{
|
|
explicit tr_peerMgr(tr_session* session_in)
|
|
: session{ session_in }
|
|
, handshake_mediator_{ *session }
|
|
, bandwidth_timer_{ session->timerMaker().create([this]() { bandwidthPulse(); }) }
|
|
, rechoke_timer_{ session->timerMaker().create([this]() { rechokePulseMarshall(); }) }
|
|
, refill_upkeep_timer_{ session->timerMaker().create([this]() { refillUpkeep(); }) }
|
|
{
|
|
bandwidth_timer_->startRepeating(BandwidthPeriod);
|
|
rechoke_timer_->startRepeating(RechokePeriod);
|
|
refill_upkeep_timer_->startRepeating(RefillUpkeepPeriod);
|
|
}
|
|
|
|
tr_peerMgr(tr_peerMgr&&) = delete;
|
|
tr_peerMgr(tr_peerMgr const&) = delete;
|
|
tr_peerMgr& operator=(tr_peerMgr&&) = delete;
|
|
tr_peerMgr& operator=(tr_peerMgr const&) = delete;
|
|
|
|
[[nodiscard]] auto unique_lock() const
|
|
{
|
|
return session->unique_lock();
|
|
}
|
|
|
|
~tr_peerMgr()
|
|
{
|
|
auto const lock = unique_lock();
|
|
incoming_handshakes.clear();
|
|
}
|
|
|
|
void rechokeSoon() noexcept
|
|
{
|
|
rechoke_timer_->setInterval(100ms);
|
|
}
|
|
|
|
void bandwidthPulse();
|
|
void rechokePulse() const;
|
|
void reconnectPulse();
|
|
void refillUpkeep() const;
|
|
void makeNewPeerConnections(size_t max);
|
|
|
|
[[nodiscard]] tr_swarm* get_existing_swarm(tr_sha1_digest_t const& hash) const
|
|
{
|
|
auto* const tor = session->torrents().get(hash);
|
|
return tor == nullptr ? nullptr : tor->swarm;
|
|
}
|
|
|
|
tr_session* const session;
|
|
Handshakes incoming_handshakes;
|
|
|
|
HandshakeMediator handshake_mediator_;
|
|
|
|
private:
|
|
void rechokePulseMarshall()
|
|
{
|
|
rechokePulse();
|
|
rechoke_timer_->setInterval(RechokePeriod);
|
|
}
|
|
|
|
std::unique_ptr<libtransmission::Timer> const bandwidth_timer_;
|
|
std::unique_ptr<libtransmission::Timer> const rechoke_timer_;
|
|
std::unique_ptr<libtransmission::Timer> const refill_upkeep_timer_;
|
|
|
|
static auto constexpr BandwidthPeriod = 500ms;
|
|
static auto constexpr RechokePeriod = 10s;
|
|
static auto constexpr RefillUpkeepPeriod = 10s;
|
|
|
|
// how frequently to decide which peers live and die
|
|
static auto constexpr ReconnectPeriodMsec = int{ 500 };
|
|
|
|
// max number of peers to ask for per second overall.
|
|
// this throttle is to avoid overloading the router
|
|
static auto constexpr MaxConnectionsPerSecond = size_t{ 12 };
|
|
};
|
|
|
|
// --- tr_peer virtual functions
|
|
|
|
tr_peer::tr_peer(tr_torrent const* tor, peer_atom* atom_in)
|
|
: session{ tor->session }
|
|
, swarm{ tor->swarm }
|
|
, atom{ atom_in }
|
|
, blame{ tor->blockCount() }
|
|
{
|
|
}
|
|
|
|
tr_peer::~tr_peer()
|
|
{
|
|
if (swarm != nullptr)
|
|
{
|
|
swarm->active_requests.remove(this);
|
|
}
|
|
|
|
if (atom != nullptr)
|
|
{
|
|
atom->is_connected = false;
|
|
}
|
|
}
|
|
|
|
// ---
|
|
|
|
tr_peerMgr* tr_peerMgrNew(tr_session* session)
|
|
{
|
|
return new tr_peerMgr{ session };
|
|
}
|
|
|
|
void tr_peerMgrFree(tr_peerMgr* manager)
|
|
{
|
|
delete manager;
|
|
}
|
|
|
|
// ---
|
|
|
|
void tr_peerMgrOnBlocklistChanged(tr_peerMgr* mgr)
|
|
{
|
|
/* we cache whether or not a peer is blocklisted...
|
|
since the blocklist has changed, erase that cached value */
|
|
for (auto* const tor : mgr->session->torrents())
|
|
{
|
|
for (auto& atom : tor->swarm->pool)
|
|
{
|
|
atom.setBlocklistedDirty();
|
|
}
|
|
}
|
|
}
|
|
|
|
// ---
|
|
|
|
void tr_peerMgrSetUtpSupported(tr_torrent* tor, tr_address const& addr)
|
|
{
|
|
if (auto* const atom = tor->swarm->get_existing_atom(addr); atom != nullptr)
|
|
{
|
|
atom->flags |= ADDED_F_UTP_FLAGS;
|
|
}
|
|
}
|
|
|
|
void tr_peerMgrSetUtpFailed(tr_torrent* tor, tr_address const& addr, bool failed)
|
|
{
|
|
if (auto* const atom = tor->swarm->get_existing_atom(addr); atom != nullptr)
|
|
{
|
|
atom->utp_failed = failed;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* REQUESTS
|
|
*
|
|
* There are two data structures associated with managing block requests:
|
|
*
|
|
* 1. tr_swarm::active_requests, an opaque class that tracks what requests
|
|
* we currently have, i.e. which blocks and from which peers.
|
|
* This is used for cancelling requests that have been waiting
|
|
* for too long and avoiding duplicate requests.
|
|
*
|
|
* 2. tr_swarm::pieces, an array of "struct weighted_piece" which lists the
|
|
* pieces that we want to request. It's used to decide which blocks to
|
|
* return next when tr_peerMgrGetBlockRequests() is called.
|
|
*/
|
|
|
|
// --- struct block_request
|
|
|
|
// TODO: if we keep this, add equivalent API to ActiveRequest
|
|
void tr_peerMgrClientSentRequests(tr_torrent* torrent, tr_peer* peer, tr_block_span_t span)
|
|
{
|
|
auto const now = tr_time();
|
|
|
|
for (tr_block_index_t block = span.begin; block < span.end; ++block)
|
|
{
|
|
torrent->swarm->active_requests.add(block, peer, now);
|
|
}
|
|
}
|
|
|
|
std::vector<tr_block_span_t> tr_peerMgrGetNextRequests(tr_torrent* torrent, tr_peer const* peer, size_t numwant)
|
|
{
|
|
class MediatorImpl final : public Wishlist::Mediator
|
|
{
|
|
public:
|
|
MediatorImpl(tr_torrent const* torrent_in, tr_peer const* peer_in)
|
|
: torrent_{ torrent_in }
|
|
, swarm_{ torrent_in->swarm }
|
|
, peer_{ peer_in }
|
|
{
|
|
}
|
|
|
|
MediatorImpl(MediatorImpl&&) = delete;
|
|
MediatorImpl(MediatorImpl const&) = delete;
|
|
MediatorImpl& operator=(MediatorImpl&&) = delete;
|
|
MediatorImpl& operator=(MediatorImpl const&) = delete;
|
|
|
|
~MediatorImpl() override = default;
|
|
|
|
[[nodiscard]] bool clientCanRequestBlock(tr_block_index_t block) const override
|
|
{
|
|
return !torrent_->hasBlock(block) && !swarm_->active_requests.has(block, peer_);
|
|
}
|
|
|
|
[[nodiscard]] bool clientCanRequestPiece(tr_piece_index_t piece) const override
|
|
{
|
|
return torrent_->pieceIsWanted(piece) && peer_->hasPiece(piece);
|
|
}
|
|
|
|
[[nodiscard]] bool isEndgame() const override
|
|
{
|
|
return swarm_->isEndgame();
|
|
}
|
|
|
|
[[nodiscard]] size_t countActiveRequests(tr_block_index_t block) const override
|
|
{
|
|
return swarm_->active_requests.count(block);
|
|
}
|
|
|
|
[[nodiscard]] size_t countMissingBlocks(tr_piece_index_t piece) const override
|
|
{
|
|
return torrent_->countMissingBlocksInPiece(piece);
|
|
}
|
|
|
|
[[nodiscard]] tr_block_span_t blockSpan(tr_piece_index_t piece) const override
|
|
{
|
|
return torrent_->blockSpanForPiece(piece);
|
|
}
|
|
|
|
[[nodiscard]] tr_piece_index_t countAllPieces() const override
|
|
{
|
|
return torrent_->pieceCount();
|
|
}
|
|
|
|
[[nodiscard]] tr_priority_t priority(tr_piece_index_t piece) const override
|
|
{
|
|
return torrent_->piecePriority(piece);
|
|
}
|
|
|
|
private:
|
|
tr_torrent const* const torrent_;
|
|
tr_swarm const* const swarm_;
|
|
tr_peer const* const peer_;
|
|
};
|
|
|
|
torrent->swarm->updateEndgame();
|
|
auto const mediator = MediatorImpl{ torrent, peer };
|
|
return Wishlist{ mediator }.next(numwant);
|
|
}
|
|
|
|
// --- Piece List Manipulation / Accessors
|
|
|
|
bool tr_peerMgrDidPeerRequest(tr_torrent const* tor, tr_peer const* peer, tr_block_index_t block)
|
|
{
|
|
return tor->swarm->active_requests.has(block, peer);
|
|
}
|
|
|
|
size_t tr_peerMgrCountActiveRequestsToPeer(tr_torrent const* tor, tr_peer const* peer)
|
|
{
|
|
return tor->swarm->active_requests.count(peer);
|
|
}
|
|
|
|
void tr_peerMgr::refillUpkeep() const
|
|
{
|
|
auto const lock = unique_lock();
|
|
|
|
for (auto* const tor : session->torrents())
|
|
{
|
|
tor->swarm->cancelOldRequests();
|
|
}
|
|
}
|
|
|
|
void tr_peerMgrPieceCompleted(tr_torrent* tor, tr_piece_index_t p)
|
|
{
|
|
bool piece_came_from_peers = false;
|
|
|
|
for (auto* const peer : tor->swarm->peers)
|
|
{
|
|
// notify the peer that we now have this piece
|
|
peer->on_piece_completed(p);
|
|
|
|
if (!piece_came_from_peers)
|
|
{
|
|
piece_came_from_peers = peer->blame.test(p);
|
|
}
|
|
}
|
|
|
|
if (piece_came_from_peers) /* webseed downloads don't belong in announce totals */
|
|
{
|
|
tr_announcerAddBytes(tor, TR_ANN_DOWN, tor->pieceSize(p));
|
|
}
|
|
|
|
// bookkeeping
|
|
tor->set_needs_completeness_check();
|
|
}
|
|
|
|
namespace
|
|
{
|
|
namespace handshake_helpers
|
|
{
|
|
void create_bit_torrent_peer(tr_torrent* tor, std::shared_ptr<tr_peerIo> io, struct peer_atom* atom, tr_quark client)
|
|
{
|
|
TR_ASSERT(atom != nullptr);
|
|
TR_ASSERT(tr_isTorrent(tor));
|
|
TR_ASSERT(tor->swarm != nullptr);
|
|
|
|
tr_swarm* swarm = tor->swarm;
|
|
|
|
auto* peer = tr_peerMsgsNew(tor, atom, std::move(io), &tr_swarm::peerCallbackFunc, swarm);
|
|
peer->client = client;
|
|
atom->is_connected = true;
|
|
|
|
swarm->peers.push_back(peer);
|
|
|
|
++swarm->stats.peer_count;
|
|
++swarm->stats.peer_from_count[atom->fromFirst];
|
|
|
|
TR_ASSERT(swarm->stats.peer_count == swarm->peerCount());
|
|
TR_ASSERT(swarm->stats.peer_from_count[atom->fromFirst] <= swarm->stats.peer_count);
|
|
|
|
// TODO is this needed?
|
|
// isn't it already initialized in tr_peerMsgsImpl's ctor?
|
|
peer->update_active(TR_UP);
|
|
peer->update_active(TR_DOWN);
|
|
}
|
|
|
|
/* FIXME: this is kind of a mess. */
|
|
[[nodiscard]] bool on_handshake_done(tr_peerMgr* manager, tr_handshake::Result const& result)
|
|
{
|
|
TR_ASSERT(result.io != nullptr);
|
|
|
|
bool const ok = result.is_connected;
|
|
bool success = false;
|
|
|
|
auto* const s = manager->get_existing_swarm(result.io->torrent_hash());
|
|
|
|
auto const [addr, port] = result.io->socket_address();
|
|
|
|
if (result.io->is_incoming())
|
|
{
|
|
manager->incoming_handshakes.erase(addr);
|
|
}
|
|
else if (s != nullptr)
|
|
{
|
|
s->outgoing_handshakes.erase(addr);
|
|
}
|
|
|
|
auto const lock = manager->unique_lock();
|
|
|
|
if (!ok || s == nullptr || !s->is_running)
|
|
{
|
|
if (s != nullptr)
|
|
{
|
|
struct peer_atom* atom = s->get_existing_atom(addr);
|
|
|
|
if (atom != nullptr)
|
|
{
|
|
++atom->num_fails;
|
|
|
|
if (!result.read_anything_from_peer)
|
|
{
|
|
tr_logAddTraceSwarm(
|
|
s,
|
|
fmt::format(
|
|
"marking peer {} as unreachable... num_fails is {}",
|
|
atom->display_name(),
|
|
atom->num_fails));
|
|
atom->flags2 |= MyflagUnreachable;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
else /* looking good */
|
|
{
|
|
struct peer_atom* atom = s->ensure_atom_exists(addr, port, 0, TR_PEER_FROM_INCOMING);
|
|
|
|
atom->time = tr_time();
|
|
atom->piece_data_time = 0;
|
|
atom->lastConnectionAt = tr_time();
|
|
|
|
if (!result.io->is_incoming())
|
|
{
|
|
atom->flags |= ADDED_F_CONNECTABLE;
|
|
atom->flags2 &= ~MyflagUnreachable;
|
|
}
|
|
|
|
/* In principle, this flag specifies whether the peer groks µTP,
|
|
not whether it's currently connected over µTP. */
|
|
if (result.io->is_utp())
|
|
{
|
|
atom->flags |= ADDED_F_UTP_FLAGS;
|
|
}
|
|
|
|
if ((atom->flags2 & MyflagBanned) != 0)
|
|
{
|
|
tr_logAddTraceSwarm(s, fmt::format("banned peer {} tried to reconnect", atom->display_name()));
|
|
}
|
|
else if (result.io->is_incoming() && s->peerCount() >= s->tor->peerLimit())
|
|
{
|
|
/* too many peers already */
|
|
}
|
|
else if (atom->is_connected)
|
|
{
|
|
// we're already connected to this peer; do nothing
|
|
}
|
|
else
|
|
{
|
|
auto client = tr_quark{ TR_KEY_NONE };
|
|
if (result.peer_id)
|
|
{
|
|
auto buf = std::array<char, 128>{};
|
|
tr_clientForId(std::data(buf), sizeof(buf), *result.peer_id);
|
|
client = tr_quark_new(std::data(buf));
|
|
}
|
|
|
|
result.io->set_bandwidth(&s->tor->bandwidth_);
|
|
create_bit_torrent_peer(s->tor, result.io, atom, client);
|
|
|
|
success = true;
|
|
}
|
|
}
|
|
|
|
return success;
|
|
}
|
|
} // namespace handshake_helpers
|
|
} // namespace
|
|
|
|
void tr_peerMgrAddIncoming(tr_peerMgr* manager, tr_peer_socket&& socket)
|
|
{
|
|
using namespace handshake_helpers;
|
|
|
|
TR_ASSERT(manager->session != nullptr);
|
|
auto const lock = manager->unique_lock();
|
|
|
|
tr_session* session = manager->session;
|
|
|
|
if (session->addressIsBlocked(socket.address()))
|
|
{
|
|
tr_logAddTrace(fmt::format("Banned IP address '{}' tried to connect to us", socket.display_name()));
|
|
socket.close();
|
|
}
|
|
else if (manager->incoming_handshakes.count(socket.address()) != 0U)
|
|
{
|
|
socket.close();
|
|
}
|
|
else /* we don't have a connection to them yet... */
|
|
{
|
|
auto address = socket.address();
|
|
manager->incoming_handshakes.try_emplace(
|
|
address,
|
|
&manager->handshake_mediator_,
|
|
tr_peerIo::new_incoming(session, &session->top_bandwidth_, std::move(socket)),
|
|
session->encryptionMode(),
|
|
[manager](tr_handshake::Result const& result) { return on_handshake_done(manager, result); });
|
|
}
|
|
}
|
|
|
|
void tr_peerMgrSetSwarmIsAllSeeds(tr_torrent* tor)
|
|
{
|
|
auto const lock = tor->unique_lock();
|
|
|
|
auto* const swarm = tor->swarm;
|
|
|
|
for (auto& atom : swarm->pool)
|
|
{
|
|
swarm->mark_atom_as_seed(atom);
|
|
}
|
|
|
|
swarm->markAllSeedsFlagDirty();
|
|
}
|
|
|
|
size_t tr_peerMgrAddPex(tr_torrent* tor, uint8_t from, tr_pex const* pex, size_t n_pex)
|
|
{
|
|
size_t n_used = 0;
|
|
tr_swarm* s = tor->swarm;
|
|
auto const lock = s->manager->unique_lock();
|
|
|
|
for (tr_pex const* const end = pex + n_pex; pex != end; ++pex)
|
|
{
|
|
if (tr_isPex(pex) && /* safeguard against corrupt data */
|
|
!s->manager->session->addressIsBlocked(pex->addr) && pex->is_valid_for_peers())
|
|
{
|
|
s->ensure_atom_exists(pex->addr, pex->port, pex->flags, from);
|
|
++n_used;
|
|
}
|
|
}
|
|
|
|
return n_used;
|
|
}
|
|
|
|
std::vector<tr_pex> tr_pex::from_compact_ipv4(
|
|
void const* compact,
|
|
size_t compact_len,
|
|
uint8_t const* added_f,
|
|
size_t added_f_len)
|
|
{
|
|
size_t const n = compact_len / 6;
|
|
auto const* walk = static_cast<std::byte const*>(compact);
|
|
auto pex = std::vector<tr_pex>(n);
|
|
|
|
for (size_t i = 0; i < n; ++i)
|
|
{
|
|
std::tie(pex[i].addr, walk) = tr_address::from_compact_ipv4(walk);
|
|
std::tie(pex[i].port, walk) = tr_port::fromCompact(walk);
|
|
|
|
if (added_f != nullptr && n == added_f_len)
|
|
{
|
|
pex[i].flags = added_f[i];
|
|
}
|
|
}
|
|
|
|
return pex;
|
|
}
|
|
|
|
std::vector<tr_pex> tr_pex::from_compact_ipv6(
|
|
void const* compact,
|
|
size_t compact_len,
|
|
uint8_t const* added_f,
|
|
size_t added_f_len)
|
|
{
|
|
size_t const n = compact_len / 18;
|
|
auto const* walk = static_cast<std::byte const*>(compact);
|
|
auto pex = std::vector<tr_pex>(n);
|
|
|
|
for (size_t i = 0; i < n; ++i)
|
|
{
|
|
std::tie(pex[i].addr, walk) = tr_address::from_compact_ipv6(walk);
|
|
std::tie(pex[i].port, walk) = tr_port::fromCompact(walk);
|
|
|
|
if (added_f != nullptr && n == added_f_len)
|
|
{
|
|
pex[i].flags = added_f[i];
|
|
}
|
|
}
|
|
|
|
return pex;
|
|
}
|
|
|
|
// ---
|
|
|
|
void tr_peerMgrGotBadPiece(tr_torrent* tor, tr_piece_index_t piece_index)
|
|
{
|
|
auto* const swarm = tor->swarm;
|
|
auto const byte_count = tor->pieceSize(piece_index);
|
|
|
|
for (auto* const peer : swarm->peers)
|
|
{
|
|
if (peer->blame.test(piece_index))
|
|
{
|
|
tr_logAddTraceSwarm(
|
|
swarm,
|
|
fmt::format(
|
|
"peer {} contributed to corrupt piece ({}); now has {} strikes",
|
|
peer->display_name(),
|
|
piece_index,
|
|
peer->strikes + 1));
|
|
swarm->addStrike(peer);
|
|
}
|
|
}
|
|
|
|
tr_announcerAddBytes(tor, TR_ANN_CORRUPT, byte_count);
|
|
}
|
|
|
|
namespace
|
|
{
|
|
namespace get_peers_helpers
|
|
{
|
|
|
|
/* better goes first */
|
|
struct CompareAtomsByUsefulness
|
|
{
|
|
[[nodiscard]] constexpr static int compare(peer_atom const& a, peer_atom const& b) noexcept // <=>
|
|
{
|
|
if (a.piece_data_time != b.piece_data_time)
|
|
{
|
|
return a.piece_data_time > b.piece_data_time ? -1 : 1;
|
|
}
|
|
|
|
if (a.fromBest != b.fromBest)
|
|
{
|
|
return a.fromBest < b.fromBest ? -1 : 1;
|
|
}
|
|
|
|
if (a.num_fails != b.num_fails)
|
|
{
|
|
return a.num_fails < b.num_fails ? -1 : 1;
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
[[nodiscard]] constexpr bool operator()(peer_atom const& a, peer_atom const& b) const noexcept
|
|
{
|
|
return compare(a, b) < 0;
|
|
}
|
|
|
|
[[nodiscard]] constexpr bool operator()(peer_atom const* a, peer_atom const* b) const noexcept
|
|
{
|
|
return compare(*a, *b) < 0;
|
|
}
|
|
};
|
|
|
|
[[nodiscard]] bool isAtomInteresting(tr_torrent const* tor, peer_atom const& atom)
|
|
{
|
|
if (tor->isDone() && atom.isSeed())
|
|
{
|
|
return false;
|
|
}
|
|
|
|
if (tor->swarm->peer_is_in_use(atom))
|
|
{
|
|
return true;
|
|
}
|
|
|
|
if (atom.isBlocklisted(tor->session))
|
|
{
|
|
return false;
|
|
}
|
|
|
|
if ((atom.flags2 & MyflagBanned) != 0)
|
|
{
|
|
return false;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
} // namespace get_peers_helpers
|
|
} // namespace
|
|
|
|
std::vector<tr_pex> tr_peerMgrGetPeers(tr_torrent const* tor, uint8_t address_type, uint8_t list_mode, size_t max_peer_count)
|
|
{
|
|
using namespace get_peers_helpers;
|
|
|
|
TR_ASSERT(tr_isTorrent(tor));
|
|
auto const lock = tor->unique_lock();
|
|
|
|
TR_ASSERT(address_type == TR_AF_INET || address_type == TR_AF_INET6);
|
|
TR_ASSERT(list_mode == TR_PEERS_CONNECTED || list_mode == TR_PEERS_INTERESTING);
|
|
|
|
tr_swarm const* s = tor->swarm;
|
|
|
|
// build a list of atoms
|
|
|
|
auto atoms = std::vector<peer_atom const*>{};
|
|
if (list_mode == TR_PEERS_CONNECTED) /* connected peers only */
|
|
{
|
|
atoms.reserve(s->peerCount());
|
|
std::transform(
|
|
std::begin(s->peers),
|
|
std::end(s->peers),
|
|
std::back_inserter(atoms),
|
|
[](auto const* peer) { return peer->atom; });
|
|
}
|
|
else /* TR_PEERS_INTERESTING */
|
|
{
|
|
for (auto const& atom : s->pool)
|
|
{
|
|
if (isAtomInteresting(tor, atom))
|
|
{
|
|
atoms.push_back(&atom);
|
|
}
|
|
}
|
|
}
|
|
|
|
std::sort(std::begin(atoms), std::end(atoms), CompareAtomsByUsefulness{});
|
|
|
|
// add the first N of them into our return list
|
|
|
|
auto const n = std::min(std::size(atoms), max_peer_count);
|
|
auto pex = std::vector<tr_pex>{};
|
|
pex.reserve(n);
|
|
|
|
for (size_t i = 0; i < std::size(atoms) && std::size(pex) < n; ++i)
|
|
{
|
|
auto const* const atom = atoms[i];
|
|
|
|
if (atom->addr.type == address_type)
|
|
{
|
|
TR_ASSERT(atom->addr.is_valid());
|
|
pex.emplace_back(atom->addr, atom->port, atom->flags);
|
|
}
|
|
}
|
|
|
|
std::sort(std::begin(pex), std::end(pex));
|
|
return pex;
|
|
}
|
|
|
|
void tr_peerMgrStartTorrent(tr_torrent* tor)
|
|
{
|
|
TR_ASSERT(tr_isTorrent(tor));
|
|
auto const lock = tor->unique_lock();
|
|
|
|
tr_swarm* const swarm = tor->swarm;
|
|
|
|
swarm->is_running = true;
|
|
|
|
swarm->manager->rechokeSoon();
|
|
}
|
|
|
|
void tr_peerMgrStopTorrent(tr_torrent* tor)
|
|
{
|
|
TR_ASSERT(tr_isTorrent(tor));
|
|
|
|
tor->swarm->stop();
|
|
}
|
|
|
|
void tr_peerMgrAddTorrent(tr_peerMgr* manager, tr_torrent* tor)
|
|
{
|
|
TR_ASSERT(tr_isTorrent(tor));
|
|
auto const lock = tor->unique_lock();
|
|
TR_ASSERT(tor->swarm == nullptr);
|
|
|
|
tor->swarm = new tr_swarm{ manager, tor };
|
|
}
|
|
|
|
void tr_peerMgrRemoveTorrent(tr_torrent* tor)
|
|
{
|
|
TR_ASSERT(tr_isTorrent(tor));
|
|
auto const lock = tor->unique_lock();
|
|
|
|
tor->swarm->stop();
|
|
delete tor->swarm;
|
|
tor->swarm = nullptr;
|
|
}
|
|
|
|
void tr_peerMgrOnTorrentGotMetainfo(tr_torrent* tor)
|
|
{
|
|
auto* const swarm = tor->swarm;
|
|
|
|
/* the webseed list may have changed... */
|
|
swarm->rebuildWebseeds();
|
|
|
|
/* some peer_msgs' progress fields may not be accurate if we
|
|
didn't have the metadata before now... so refresh them all... */
|
|
for (auto* peer : swarm->peers)
|
|
{
|
|
peer->onTorrentGotMetainfo();
|
|
|
|
if (peer->isSeed())
|
|
{
|
|
swarm->mark_atom_as_seed(*peer->atom);
|
|
}
|
|
}
|
|
|
|
/* update the bittorrent peers' willingness... */
|
|
for (auto* peer : swarm->peers)
|
|
{
|
|
peer->update_active(TR_UP);
|
|
peer->update_active(TR_DOWN);
|
|
}
|
|
}
|
|
|
|
int8_t tr_peerMgrPieceAvailability(tr_torrent const* tor, tr_piece_index_t piece)
|
|
{
|
|
if (!tor->hasMetainfo())
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
if (tor->isSeed() || tor->hasPiece(piece))
|
|
{
|
|
return -1;
|
|
}
|
|
|
|
auto const& peers = tor->swarm->peers;
|
|
return std::count_if(std::begin(peers), std::end(peers), [piece](auto const* peer) { return peer->hasPiece(piece); });
|
|
}
|
|
|
|
void tr_peerMgrTorrentAvailability(tr_torrent const* tor, int8_t* tab, unsigned int n_tabs)
|
|
{
|
|
TR_ASSERT(tr_isTorrent(tor));
|
|
TR_ASSERT(tab != nullptr);
|
|
TR_ASSERT(n_tabs > 0);
|
|
|
|
std::fill_n(tab, n_tabs, int8_t{});
|
|
|
|
auto const interval = tor->pieceCount() / static_cast<float>(n_tabs);
|
|
for (tr_piece_index_t i = 0; i < n_tabs; ++i)
|
|
{
|
|
auto const piece = static_cast<tr_piece_index_t>(i * interval);
|
|
tab[i] = tr_peerMgrPieceAvailability(tor, piece);
|
|
}
|
|
}
|
|
|
|
tr_swarm_stats tr_swarmGetStats(tr_swarm const* swarm)
|
|
{
|
|
TR_ASSERT(swarm != nullptr);
|
|
auto& stats = swarm->stats;
|
|
stats.active_webseed_count = swarm->countActiveWebseeds(tr_time_msec());
|
|
return stats;
|
|
}
|
|
|
|
void tr_swarmIncrementActivePeers(tr_swarm* swarm, tr_direction direction, bool is_active)
|
|
{
|
|
int n = swarm->stats.active_peer_count[direction];
|
|
|
|
if (is_active)
|
|
{
|
|
++n;
|
|
}
|
|
else
|
|
{
|
|
--n;
|
|
}
|
|
|
|
TR_ASSERT(n >= 0);
|
|
TR_ASSERT(n <= swarm->stats.peer_count);
|
|
|
|
swarm->stats.active_peer_count[direction] = n;
|
|
}
|
|
|
|
/* count how many bytes we want that connected peers have */
|
|
uint64_t tr_peerMgrGetDesiredAvailable(tr_torrent const* tor)
|
|
{
|
|
TR_ASSERT(tr_isTorrent(tor));
|
|
|
|
// common shortcuts...
|
|
|
|
if (!tor->isRunning || tor->isStopping || tor->isDone() || !tor->hasMetainfo())
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
tr_swarm const* const swarm = tor->swarm;
|
|
if (swarm == nullptr || swarm->peerCount() == 0U)
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
auto available = swarm->peers.front()->has();
|
|
for (auto const* const peer : swarm->peers)
|
|
{
|
|
available |= peer->has();
|
|
}
|
|
|
|
if (available.hasAll())
|
|
{
|
|
return tor->leftUntilDone();
|
|
}
|
|
|
|
auto desired_available = uint64_t{};
|
|
|
|
for (tr_piece_index_t i = 0, n = tor->pieceCount(); i < n; ++i)
|
|
{
|
|
if (tor->pieceIsWanted(i) && available.test(i))
|
|
{
|
|
desired_available += tor->countMissingBytesInPiece(i);
|
|
}
|
|
}
|
|
|
|
TR_ASSERT(desired_available <= tor->totalSize());
|
|
return desired_available;
|
|
}
|
|
|
|
tr_webseed_view tr_peerMgrWebseed(tr_torrent const* tor, size_t i)
|
|
{
|
|
TR_ASSERT(tr_isTorrent(tor));
|
|
TR_ASSERT(tor->swarm != nullptr);
|
|
size_t const n = std::size(tor->swarm->webseeds);
|
|
TR_ASSERT(i < n);
|
|
|
|
return i >= n ? tr_webseed_view{} : tr_webseedView(tor->swarm->webseeds[i].get());
|
|
}
|
|
|
|
namespace
|
|
{
|
|
namespace peer_stat_helpers
|
|
{
|
|
|
|
[[nodiscard]] auto getPeerStats(tr_peerMsgs const* peer, time_t now, uint64_t now_msec)
|
|
{
|
|
auto stats = tr_peer_stat{};
|
|
auto const* const atom = peer->atom;
|
|
|
|
auto const [addr, port] = peer->socketAddress();
|
|
|
|
addr.display_name(stats.addr, sizeof(stats.addr));
|
|
stats.client = peer->client.c_str();
|
|
stats.port = port.host();
|
|
stats.from = atom->fromFirst;
|
|
stats.progress = peer->percentDone();
|
|
stats.isUTP = peer->is_utp_connection();
|
|
stats.isEncrypted = peer->is_encrypted();
|
|
stats.rateToPeer_KBps = tr_toSpeedKBps(peer->get_piece_speed_bytes_per_second(now_msec, TR_CLIENT_TO_PEER));
|
|
stats.rateToClient_KBps = tr_toSpeedKBps(peer->get_piece_speed_bytes_per_second(now_msec, TR_PEER_TO_CLIENT));
|
|
stats.peerIsChoked = peer->is_peer_choked();
|
|
stats.peerIsInterested = peer->is_peer_interested();
|
|
stats.clientIsChoked = peer->is_client_choked();
|
|
stats.clientIsInterested = peer->is_client_interested();
|
|
stats.isIncoming = peer->is_incoming_connection();
|
|
stats.isDownloadingFrom = peer->is_active(TR_PEER_TO_CLIENT);
|
|
stats.isUploadingTo = peer->is_active(TR_CLIENT_TO_PEER);
|
|
stats.isSeed = peer->isSeed();
|
|
|
|
stats.blocksToPeer = peer->blocks_sent_to_peer.count(now, CancelHistorySec);
|
|
stats.blocksToClient = peer->blocks_sent_to_client.count(now, CancelHistorySec);
|
|
stats.cancelsToPeer = peer->cancels_sent_to_peer.count(now, CancelHistorySec);
|
|
stats.cancelsToClient = peer->cancels_sent_to_client.count(now, CancelHistorySec);
|
|
|
|
stats.activeReqsToPeer = peer->activeReqCount(TR_CLIENT_TO_PEER);
|
|
stats.activeReqsToClient = peer->activeReqCount(TR_PEER_TO_CLIENT);
|
|
|
|
char* pch = stats.flagStr;
|
|
|
|
if (stats.isUTP)
|
|
{
|
|
*pch++ = 'T';
|
|
}
|
|
|
|
if (peer->swarm->optimistic == peer)
|
|
{
|
|
*pch++ = 'O';
|
|
}
|
|
|
|
if (stats.isDownloadingFrom)
|
|
{
|
|
*pch++ = 'D';
|
|
}
|
|
else if (stats.clientIsInterested)
|
|
{
|
|
*pch++ = 'd';
|
|
}
|
|
|
|
if (stats.isUploadingTo)
|
|
{
|
|
*pch++ = 'U';
|
|
}
|
|
else if (stats.peerIsInterested)
|
|
{
|
|
*pch++ = 'u';
|
|
}
|
|
|
|
if (!stats.clientIsChoked && !stats.clientIsInterested)
|
|
{
|
|
*pch++ = 'K';
|
|
}
|
|
|
|
if (!stats.peerIsChoked && !stats.peerIsInterested)
|
|
{
|
|
*pch++ = '?';
|
|
}
|
|
|
|
if (stats.isEncrypted)
|
|
{
|
|
*pch++ = 'E';
|
|
}
|
|
|
|
if (stats.from == TR_PEER_FROM_DHT)
|
|
{
|
|
*pch++ = 'H';
|
|
}
|
|
else if (stats.from == TR_PEER_FROM_PEX)
|
|
{
|
|
*pch++ = 'X';
|
|
}
|
|
|
|
if (stats.isIncoming)
|
|
{
|
|
*pch++ = 'I';
|
|
}
|
|
|
|
*pch = '\0';
|
|
|
|
return stats;
|
|
}
|
|
|
|
} // namespace peer_stat_helpers
|
|
} // namespace
|
|
|
|
tr_peer_stat* tr_peerMgrPeerStats(tr_torrent const* tor, size_t* setme_count)
|
|
{
|
|
using namespace peer_stat_helpers;
|
|
|
|
TR_ASSERT(tr_isTorrent(tor));
|
|
TR_ASSERT(tor->swarm->manager != nullptr);
|
|
|
|
auto const n = tor->swarm->peerCount();
|
|
auto* const ret = new tr_peer_stat[n];
|
|
|
|
auto const now = tr_time();
|
|
auto const now_msec = tr_time_msec();
|
|
std::transform(
|
|
std::begin(tor->swarm->peers),
|
|
std::end(tor->swarm->peers),
|
|
ret,
|
|
[&now, &now_msec](auto const* peer) { return getPeerStats(peer, now, now_msec); });
|
|
|
|
*setme_count = n;
|
|
return ret;
|
|
}
|
|
|
|
void tr_peerMgrClearInterest(tr_torrent* tor)
|
|
{
|
|
TR_ASSERT(tr_isTorrent(tor));
|
|
auto const lock = tor->unique_lock();
|
|
|
|
auto& peers = tor->swarm->peers;
|
|
std::for_each(std::begin(peers), std::end(peers), [](auto* const peer) { peer->set_interested(false); });
|
|
}
|
|
|
|
namespace
|
|
{
|
|
namespace update_interest_helpers
|
|
{
|
|
/* does this peer have any pieces that we want? */
|
|
[[nodiscard]] bool isPeerInteresting(
|
|
tr_torrent const* const tor,
|
|
std::vector<bool> const& piece_is_interesting,
|
|
tr_peerMsgs const* const peer)
|
|
{
|
|
/* these cases should have already been handled by the calling code... */
|
|
TR_ASSERT(!tor->isDone());
|
|
TR_ASSERT(tor->clientCanDownload());
|
|
|
|
if (peer->isSeed())
|
|
{
|
|
return true;
|
|
}
|
|
|
|
for (tr_piece_index_t i = 0; i < tor->pieceCount(); ++i)
|
|
{
|
|
if (piece_is_interesting[i] && peer->hasPiece(i))
|
|
{
|
|
return true;
|
|
}
|
|
}
|
|
|
|
return false;
|
|
}
|
|
|
|
// determine which peers to show interest in
|
|
void updateInterest(tr_swarm* swarm)
|
|
{
|
|
// sometimes this function isn't necessary
|
|
auto const* const tor = swarm->tor;
|
|
if (tor->isDone() || !tor->clientCanDownload())
|
|
{
|
|
return;
|
|
}
|
|
|
|
if (auto const peer_count = swarm->peerCount(); peer_count > 0)
|
|
{
|
|
int const n = tor->pieceCount();
|
|
|
|
// build a bitfield of interesting pieces...
|
|
auto piece_is_interesting = std::vector<bool>{};
|
|
piece_is_interesting.resize(n);
|
|
for (int i = 0; i < n; ++i)
|
|
{
|
|
piece_is_interesting[i] = tor->pieceIsWanted(i) && !tor->hasPiece(i);
|
|
}
|
|
|
|
for (auto* const peer : swarm->peers)
|
|
{
|
|
peer->set_interested(isPeerInteresting(tor, piece_is_interesting, peer));
|
|
}
|
|
}
|
|
}
|
|
} // namespace update_interest_helpers
|
|
} // namespace
|
|
|
|
// ---
|
|
|
|
namespace
|
|
{
|
|
namespace rechoke_uploads_helpers
|
|
{
|
|
struct ChokeData
|
|
{
|
|
ChokeData(tr_peerMsgs* msgs_in, int rate_in, uint8_t salt_in, bool is_interested_in, bool was_choked_in, bool is_choked_in)
|
|
: msgs{ msgs_in }
|
|
, rate{ rate_in }
|
|
, salt{ salt_in }
|
|
, is_interested{ is_interested_in }
|
|
, was_choked{ was_choked_in }
|
|
, is_choked{ is_choked_in }
|
|
{
|
|
}
|
|
|
|
tr_peerMsgs* msgs;
|
|
int rate;
|
|
uint8_t salt;
|
|
bool is_interested;
|
|
bool was_choked;
|
|
bool is_choked;
|
|
|
|
[[nodiscard]] constexpr auto compare(ChokeData const& that) const noexcept // <=>
|
|
{
|
|
if (this->rate != that.rate) // prefer higher overall speeds
|
|
{
|
|
return this->rate > that.rate ? -1 : 1;
|
|
}
|
|
|
|
if (this->was_choked != that.was_choked) // prefer unchoked
|
|
{
|
|
return this->was_choked ? 1 : -1;
|
|
}
|
|
|
|
if (this->salt != that.salt) // random order
|
|
{
|
|
return this->salt < that.salt ? -1 : 1;
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
[[nodiscard]] constexpr auto operator<(ChokeData const& that) const noexcept
|
|
{
|
|
return compare(that) < 0;
|
|
}
|
|
};
|
|
|
|
/* get a rate for deciding which peers to choke and unchoke. */
|
|
[[nodiscard]] auto getRateBps(tr_torrent const* tor, tr_peer const* peer, uint64_t now)
|
|
{
|
|
if (tor->isDone())
|
|
{
|
|
return peer->get_piece_speed_bytes_per_second(now, TR_CLIENT_TO_PEER);
|
|
}
|
|
|
|
/* downloading a private torrent... take upload speed into account
|
|
* because there may only be a small window of opportunity to share */
|
|
if (tor->isPrivate())
|
|
{
|
|
return peer->get_piece_speed_bytes_per_second(now, TR_PEER_TO_CLIENT) +
|
|
peer->get_piece_speed_bytes_per_second(now, TR_CLIENT_TO_PEER);
|
|
}
|
|
|
|
/* downloading a public torrent */
|
|
return peer->get_piece_speed_bytes_per_second(now, TR_PEER_TO_CLIENT);
|
|
}
|
|
|
|
// an optimistically unchoked peer is immune from rechoking
|
|
// for this many calls to rechokeUploads().
|
|
auto constexpr OptimisticUnchokeMultiplier = uint8_t{ 4 };
|
|
|
|
void rechokeUploads(tr_swarm* s, uint64_t const now)
|
|
{
|
|
auto const lock = s->unique_lock();
|
|
|
|
auto const peer_count = s->peerCount();
|
|
auto& peers = s->peers;
|
|
auto choked = std::vector<ChokeData>{};
|
|
choked.reserve(peer_count);
|
|
auto const* const session = s->manager->session;
|
|
bool const choke_all = !s->tor->clientCanUpload();
|
|
bool const is_maxed_out = s->tor->bandwidth_.is_maxed_out(TR_UP, now);
|
|
|
|
/* an optimistic unchoke peer's "optimistic"
|
|
* state lasts for N calls to rechokeUploads(). */
|
|
if (s->optimistic_unchoke_time_scaler > 0)
|
|
{
|
|
--s->optimistic_unchoke_time_scaler;
|
|
}
|
|
else
|
|
{
|
|
s->optimistic = nullptr;
|
|
}
|
|
|
|
/* sort the peers by preference and rate */
|
|
auto salter = tr_salt_shaker{};
|
|
for (auto* const peer : peers)
|
|
{
|
|
if (peer->isSeed())
|
|
{
|
|
/* choke seeds and partial seeds */
|
|
peer->set_choke(true);
|
|
}
|
|
else if (choke_all)
|
|
{
|
|
/* choke everyone if we're not uploading */
|
|
peer->set_choke(true);
|
|
}
|
|
else if (peer != s->optimistic)
|
|
{
|
|
choked.emplace_back(
|
|
peer,
|
|
getRateBps(s->tor, peer, now),
|
|
salter(),
|
|
peer->is_peer_interested(),
|
|
peer->is_peer_choked(),
|
|
true);
|
|
}
|
|
}
|
|
|
|
std::sort(std::begin(choked), std::end(choked));
|
|
|
|
/**
|
|
* Reciprocation and number of uploads capping is managed by unchoking
|
|
* the N peers which have the best upload rate and are interested.
|
|
* This maximizes the client's download rate. These N peers are
|
|
* referred to as downloaders, because they are interested in downloading
|
|
* from the client.
|
|
*
|
|
* Peers which have a better upload rate (as compared to the downloaders)
|
|
* but aren't interested get unchoked. If they become interested, the
|
|
* downloader with the worst upload rate gets choked. If a client has
|
|
* a complete file, it uses its upload rate rather than its download
|
|
* rate to decide which peers to unchoke.
|
|
*
|
|
* If our bandwidth is maxed out, don't unchoke any more peers.
|
|
*/
|
|
auto checked_choke_count = size_t{ 0U };
|
|
auto unchoked_interested = size_t{ 0U };
|
|
|
|
for (auto& item : choked)
|
|
{
|
|
if (unchoked_interested >= session->uploadSlotsPerTorrent())
|
|
{
|
|
break;
|
|
}
|
|
|
|
item.is_choked = is_maxed_out ? item.was_choked : false;
|
|
|
|
++checked_choke_count;
|
|
|
|
if (item.is_interested)
|
|
{
|
|
++unchoked_interested;
|
|
}
|
|
}
|
|
|
|
/* optimistic unchoke */
|
|
if (s->optimistic == nullptr && !is_maxed_out && checked_choke_count < std::size(choked))
|
|
{
|
|
auto rand_pool = std::vector<ChokeData*>{};
|
|
|
|
for (auto i = checked_choke_count, n = std::size(choked); i < n; ++i)
|
|
{
|
|
if (choked[i].is_interested)
|
|
{
|
|
rand_pool.push_back(&choked[i]);
|
|
}
|
|
}
|
|
|
|
if (auto const n = std::size(rand_pool); n != 0)
|
|
{
|
|
auto* c = rand_pool[tr_rand_int(n)];
|
|
c->is_choked = false;
|
|
s->optimistic = c->msgs;
|
|
s->optimistic_unchoke_time_scaler = OptimisticUnchokeMultiplier;
|
|
}
|
|
}
|
|
|
|
for (auto& item : choked)
|
|
{
|
|
item.msgs->set_choke(item.is_choked);
|
|
}
|
|
}
|
|
} // namespace rechoke_uploads_helpers
|
|
} // namespace
|
|
|
|
void tr_peerMgr::rechokePulse() const
|
|
{
|
|
using namespace update_interest_helpers;
|
|
using namespace rechoke_uploads_helpers;
|
|
|
|
auto const lock = unique_lock();
|
|
auto const now = tr_time_msec();
|
|
|
|
for (auto* const tor : session->torrents())
|
|
{
|
|
if (tor->isRunning)
|
|
{
|
|
// possibly stop torrents that have seeded enough
|
|
tr_torrentCheckSeedLimit(tor);
|
|
}
|
|
|
|
if (tor->isRunning)
|
|
{
|
|
if (auto* const swarm = tor->swarm; swarm->stats.peer_count > 0)
|
|
{
|
|
rechokeUploads(swarm, now);
|
|
updateInterest(swarm);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
// --- Life and Death
|
|
|
|
namespace
|
|
{
|
|
namespace disconnect_helpers
|
|
{
|
|
// when many peers are available, keep idle ones this long
|
|
auto constexpr MinUploadIdleSecs = time_t{ 60 };
|
|
|
|
// when few peers are available, keep idle ones this long
|
|
auto constexpr MaxUploadIdleSecs = time_t{ 60 * 5 };
|
|
|
|
[[nodiscard]] bool shouldPeerBeClosed(tr_swarm const* s, tr_peerMsgs const* peer, size_t peer_count, time_t const now)
|
|
{
|
|
/* if it's marked for purging, close it */
|
|
if (peer->do_purge)
|
|
{
|
|
tr_logAddTraceSwarm(s, fmt::format("purging peer {} because its do_purge flag is set", peer->display_name()));
|
|
return true;
|
|
}
|
|
|
|
auto const* tor = s->tor;
|
|
auto const* const atom = peer->atom;
|
|
|
|
/* disconnect if we're both seeds and enough time has passed for PEX */
|
|
if (tor->isDone() && peer->isSeed())
|
|
{
|
|
return !tor->allowsPex() || now - atom->time >= 30;
|
|
}
|
|
|
|
/* disconnect if it's been too long since piece data has been transferred.
|
|
* this is on a sliding scale based on number of available peers... */
|
|
{
|
|
auto const relax_strictness_if_fewer_than_n = static_cast<size_t>(std::lround(tor->peerLimit() * 0.9));
|
|
/* if we have >= relaxIfFewerThan, strictness is 100%.
|
|
* if we have zero connections, strictness is 0% */
|
|
float const strictness = peer_count >= relax_strictness_if_fewer_than_n ?
|
|
1.0 :
|
|
peer_count / (float)relax_strictness_if_fewer_than_n;
|
|
auto const lo = MinUploadIdleSecs;
|
|
auto const hi = MaxUploadIdleSecs;
|
|
time_t const limit = hi - (hi - lo) * strictness;
|
|
time_t const idle_time = now - std::max(atom->time, atom->piece_data_time);
|
|
|
|
if (idle_time > limit)
|
|
{
|
|
tr_logAddTraceSwarm(
|
|
s,
|
|
fmt::format(
|
|
"purging peer {} because it's been {} secs since we shared anything",
|
|
peer->display_name(),
|
|
idle_time));
|
|
return true;
|
|
}
|
|
}
|
|
|
|
return false;
|
|
}
|
|
|
|
void closePeer(tr_peer* peer)
|
|
{
|
|
TR_ASSERT(peer != nullptr);
|
|
auto const* const s = peer->swarm;
|
|
|
|
/* if we transferred piece data, then they might be good peers,
|
|
so reset their `num_fails' weight to zero. otherwise we connected
|
|
to them fruitlessly, so mark it as another fail */
|
|
if (auto* const atom = peer->atom; atom->piece_data_time != 0)
|
|
{
|
|
tr_logAddTraceSwarm(s, fmt::format("resetting atom {} num_fails to 0", peer->display_name()));
|
|
atom->num_fails = 0;
|
|
}
|
|
else
|
|
{
|
|
++atom->num_fails;
|
|
tr_logAddTraceSwarm(s, fmt::format("incremented atom {} num_fails to {}", peer->display_name(), atom->num_fails));
|
|
}
|
|
|
|
tr_logAddTraceSwarm(s, fmt::format("removing bad peer {}", peer->display_name()));
|
|
peer->swarm->removePeer(peer);
|
|
}
|
|
|
|
struct ComparePeerByActivity
|
|
{
|
|
[[nodiscard]] constexpr static int compare(tr_peer const* a, tr_peer const* b) // <=>
|
|
{
|
|
if (a->do_purge != b->do_purge)
|
|
{
|
|
return a->do_purge ? 1 : -1;
|
|
}
|
|
|
|
/* the one to give us data more recently goes first */
|
|
if (a->atom->piece_data_time != b->atom->piece_data_time)
|
|
{
|
|
return a->atom->piece_data_time > b->atom->piece_data_time ? -1 : 1;
|
|
}
|
|
|
|
/* the one we connected to most recently goes first */
|
|
if (a->atom->time != b->atom->time)
|
|
{
|
|
return a->atom->time > b->atom->time ? -1 : 1;
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
[[nodiscard]] constexpr bool operator()(tr_peer const* a, tr_peer const* b) const // less than
|
|
{
|
|
return compare(a, b) < 0;
|
|
}
|
|
};
|
|
|
|
[[nodiscard]] auto getPeersToClose(tr_swarm const* const swarm, time_t const now_sec)
|
|
{
|
|
auto peers_to_close = std::vector<tr_peer*>{};
|
|
|
|
auto const peer_count = swarm->peerCount();
|
|
for (auto* peer : swarm->peers)
|
|
{
|
|
if (shouldPeerBeClosed(swarm, peer, peer_count, now_sec))
|
|
{
|
|
peers_to_close.push_back(peer);
|
|
}
|
|
}
|
|
|
|
return peers_to_close;
|
|
}
|
|
|
|
void closeBadPeers(tr_swarm* s, time_t const now_sec)
|
|
{
|
|
for (auto* peer : getPeersToClose(s, now_sec))
|
|
{
|
|
closePeer(peer);
|
|
}
|
|
}
|
|
|
|
void enforceSwarmPeerLimit(tr_swarm* swarm, size_t max)
|
|
{
|
|
// do we have too many peers?
|
|
if (auto const n = swarm->peerCount(); n <= max)
|
|
{
|
|
return;
|
|
}
|
|
|
|
// close all but the `max` most active
|
|
auto peers = swarm->peers;
|
|
std::partial_sort(std::begin(peers), std::begin(peers) + max, std::end(peers), ComparePeerByActivity{});
|
|
std::for_each(std::begin(peers) + max, std::end(peers), closePeer);
|
|
}
|
|
|
|
void enforceSessionPeerLimit(tr_session* session)
|
|
{
|
|
// No need to disconnect if we are under the peer limit
|
|
auto const max = session->peerLimit();
|
|
if (tr_peerMsgs::size() <= max)
|
|
{
|
|
return;
|
|
}
|
|
|
|
// Make a list of all the peers.
|
|
auto peers = std::vector<tr_peer*>{};
|
|
peers.reserve(tr_peerMsgs::size());
|
|
for (auto const* const tor : session->torrents())
|
|
{
|
|
peers.insert(std::end(peers), std::begin(tor->swarm->peers), std::end(tor->swarm->peers));
|
|
}
|
|
|
|
TR_ASSERT(tr_peerMsgs::size() == std::size(peers));
|
|
if (std::size(peers) > max)
|
|
{
|
|
std::partial_sort(std::begin(peers), std::begin(peers) + max, std::end(peers), ComparePeerByActivity{});
|
|
std::for_each(std::begin(peers) + max, std::end(peers), closePeer);
|
|
}
|
|
}
|
|
} // namespace disconnect_helpers
|
|
} // namespace
|
|
|
|
void tr_peerMgr::reconnectPulse()
|
|
{
|
|
using namespace disconnect_helpers;
|
|
|
|
auto const lock = session->unique_lock();
|
|
auto const now_sec = tr_time();
|
|
|
|
// remove crappy peers
|
|
for (auto* const tor : session->torrents())
|
|
{
|
|
auto* const swarm = tor->swarm;
|
|
|
|
if (!swarm->is_running)
|
|
{
|
|
swarm->removeAllPeers();
|
|
}
|
|
else
|
|
{
|
|
closeBadPeers(swarm, now_sec);
|
|
}
|
|
}
|
|
|
|
// if we're over the per-torrent peer limits, cull some peers
|
|
for (auto* const tor : session->torrents())
|
|
{
|
|
if (tor->isRunning)
|
|
{
|
|
enforceSwarmPeerLimit(tor->swarm, tor->peerLimit());
|
|
}
|
|
}
|
|
|
|
// if we're over the per-session peer limits, cull some peers
|
|
enforceSessionPeerLimit(session);
|
|
|
|
// try to make new peer connections
|
|
auto const max_connections_per_pulse = int(MaxConnectionsPerSecond * (ReconnectPeriodMsec / 1000.0));
|
|
makeNewPeerConnections(max_connections_per_pulse);
|
|
}
|
|
|
|
// --- Bandwidth Allocation
|
|
|
|
namespace
|
|
{
|
|
namespace bandwidth_helpers
|
|
{
|
|
|
|
void pumpAllPeers(tr_peerMgr* mgr)
|
|
{
|
|
for (auto* const tor : mgr->session->torrents())
|
|
{
|
|
for (auto* const peer : tor->swarm->peers)
|
|
{
|
|
peer->pulse();
|
|
}
|
|
}
|
|
}
|
|
|
|
void queuePulse(tr_session* session, tr_direction dir)
|
|
{
|
|
TR_ASSERT(session != nullptr);
|
|
TR_ASSERT(tr_isDirection(dir));
|
|
|
|
if (!session->queueEnabled(dir))
|
|
{
|
|
return;
|
|
}
|
|
|
|
auto const n = session->countQueueFreeSlots(dir);
|
|
for (auto* tor : session->getNextQueuedTorrents(dir, n))
|
|
{
|
|
tr_torrentStartNow(tor);
|
|
session->onQueuedTorrentStarted(tor);
|
|
}
|
|
}
|
|
|
|
} // namespace bandwidth_helpers
|
|
} // namespace
|
|
|
|
void tr_peerMgr::bandwidthPulse()
|
|
{
|
|
using namespace bandwidth_helpers;
|
|
|
|
auto const lock = unique_lock();
|
|
|
|
pumpAllPeers(this);
|
|
|
|
// allocate bandwidth to the peers
|
|
static auto constexpr Msec = std::chrono::duration_cast<std::chrono::milliseconds>(BandwidthPeriod).count();
|
|
session->top_bandwidth_.allocate(Msec);
|
|
|
|
// torrent upkeep
|
|
for (auto* const tor : session->torrents())
|
|
{
|
|
tor->do_idle_work();
|
|
tr_torrentMagnetDoIdleWork(tor);
|
|
}
|
|
|
|
/* pump the queues */
|
|
queuePulse(session, TR_UP);
|
|
queuePulse(session, TR_DOWN);
|
|
|
|
reconnectPulse();
|
|
}
|
|
|
|
// ---
|
|
|
|
bool tr_swarm::peer_is_in_use(peer_atom const& atom) const
|
|
{
|
|
return atom.is_connected || outgoing_handshakes.count(atom.addr) != 0U ||
|
|
manager->incoming_handshakes.count(atom.addr) != 0U;
|
|
}
|
|
|
|
namespace
|
|
{
|
|
namespace connect_helpers
|
|
{
|
|
/* is this atom someone that we'd want to initiate a connection to? */
|
|
[[nodiscard]] bool isPeerCandidate(tr_torrent const* tor, peer_atom const& atom, time_t const now)
|
|
{
|
|
// have we already tried and failed to connect?
|
|
if (auto const reachable = atom.isReachable(); reachable && !*reachable)
|
|
{
|
|
return false;
|
|
}
|
|
|
|
// not if we're both seeds
|
|
if (tor->isDone() && atom.isSeed())
|
|
{
|
|
return false;
|
|
}
|
|
|
|
// not if we've already got a connection to them...
|
|
if (tor->swarm->peer_is_in_use(atom))
|
|
{
|
|
return false;
|
|
}
|
|
|
|
// not if we just tried them already
|
|
if (now - atom.time < atom.getReconnectIntervalSecs(now))
|
|
{
|
|
return false;
|
|
}
|
|
|
|
// not if they're blocklisted
|
|
if (atom.isBlocklisted(tor->session))
|
|
{
|
|
return false;
|
|
}
|
|
|
|
// not if they're banned...
|
|
if ((atom.flags2 & MyflagBanned) != 0)
|
|
{
|
|
return false;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
struct peer_candidate
|
|
{
|
|
uint64_t score;
|
|
tr_torrent* tor;
|
|
peer_atom* atom;
|
|
};
|
|
|
|
[[nodiscard]] bool torrentWasRecentlyStarted(tr_torrent const* tor)
|
|
{
|
|
return difftime(tr_time(), tor->startDate) < 120;
|
|
}
|
|
|
|
[[nodiscard]] constexpr uint64_t addValToKey(uint64_t value, int width, uint64_t addme)
|
|
{
|
|
value = value << (uint64_t)width;
|
|
value |= addme;
|
|
return value;
|
|
}
|
|
|
|
/* smaller value is better */
|
|
[[nodiscard]] uint64_t getPeerCandidateScore(tr_torrent const* tor, peer_atom const& atom, uint8_t salt)
|
|
{
|
|
auto i = uint64_t{};
|
|
auto score = uint64_t{};
|
|
bool const failed = atom.lastConnectionAt < atom.lastConnectionAttemptAt;
|
|
|
|
/* prefer peers we've connected to, or never tried, over peers we failed to connect to. */
|
|
i = failed ? 1 : 0;
|
|
score = addValToKey(score, 1, i);
|
|
|
|
/* prefer the one we attempted least recently (to cycle through all peers) */
|
|
i = atom.lastConnectionAttemptAt;
|
|
score = addValToKey(score, 32, i);
|
|
|
|
/* prefer peers belonging to a torrent of a higher priority */
|
|
switch (tor->getPriority())
|
|
{
|
|
case TR_PRI_HIGH:
|
|
i = 0;
|
|
break;
|
|
|
|
case TR_PRI_NORMAL:
|
|
i = 1;
|
|
break;
|
|
|
|
case TR_PRI_LOW:
|
|
i = 2;
|
|
break;
|
|
}
|
|
|
|
score = addValToKey(score, 4, i);
|
|
|
|
/* prefer recently-started torrents */
|
|
i = torrentWasRecentlyStarted(tor) ? 0 : 1;
|
|
score = addValToKey(score, 1, i);
|
|
|
|
/* prefer torrents we're downloading with */
|
|
i = tor->isDone() ? 1 : 0;
|
|
score = addValToKey(score, 1, i);
|
|
|
|
/* prefer peers that are known to be connectible */
|
|
i = (atom.flags & ADDED_F_CONNECTABLE) != 0 ? 0 : 1;
|
|
score = addValToKey(score, 1, i);
|
|
|
|
/* prefer peers that we might be able to upload to */
|
|
i = (atom.flags & ADDED_F_SEED_FLAG) == 0 ? 0 : 1;
|
|
score = addValToKey(score, 1, i);
|
|
|
|
/* Prefer peers that we got from more trusted sources.
|
|
* lower `fromBest` values indicate more trusted sources */
|
|
score = addValToKey(score, 4, atom.fromBest);
|
|
|
|
/* salt */
|
|
score = addValToKey(score, 8, salt);
|
|
|
|
return score;
|
|
}
|
|
|
|
/** @return an array of all the atoms we might want to connect to */
|
|
[[nodiscard]] std::vector<peer_candidate> getPeerCandidates(tr_session* session, size_t max)
|
|
{
|
|
auto const now = tr_time();
|
|
auto const now_msec = tr_time_msec();
|
|
|
|
// leave 5% of connection slots for incoming connections -- ticket #2609
|
|
if (auto const max_candidates = static_cast<size_t>(session->peerLimit() * 0.95); max_candidates <= tr_peerMsgs::size())
|
|
{
|
|
return {};
|
|
}
|
|
|
|
auto candidates = std::vector<peer_candidate>{};
|
|
candidates.reserve(peer_atom::atom_count());
|
|
|
|
/* populate the candidate array */
|
|
auto salter = tr_salt_shaker{};
|
|
for (auto* const tor : session->torrents())
|
|
{
|
|
auto* const swarm = tor->swarm;
|
|
|
|
if (!swarm->is_running)
|
|
{
|
|
continue;
|
|
}
|
|
|
|
/* if everyone in the swarm is seeds and pex is disabled because
|
|
* the torrent is private, then don't initiate connections */
|
|
bool const seeding = tor->isDone();
|
|
if (seeding && swarm->isAllSeeds() && tor->isPrivate())
|
|
{
|
|
continue;
|
|
}
|
|
|
|
/* if we've already got enough peers in this torrent... */
|
|
if (tor->peerLimit() <= swarm->peerCount())
|
|
{
|
|
continue;
|
|
}
|
|
|
|
/* if we've already got enough speed in this torrent... */
|
|
if (seeding && tor->bandwidth_.is_maxed_out(TR_UP, now_msec))
|
|
{
|
|
continue;
|
|
}
|
|
|
|
for (auto& atom : swarm->pool)
|
|
{
|
|
if (isPeerCandidate(tor, atom, now))
|
|
{
|
|
candidates.push_back({ getPeerCandidateScore(tor, atom, salter()), tor, &atom });
|
|
}
|
|
}
|
|
}
|
|
|
|
// only keep the best `max` candidates
|
|
if (std::size(candidates) > max)
|
|
{
|
|
std::partial_sort(
|
|
std::begin(candidates),
|
|
std::begin(candidates) + max,
|
|
std::end(candidates),
|
|
[](auto const& a, auto const& b) { return a.score < b.score; });
|
|
candidates.resize(max);
|
|
}
|
|
|
|
return candidates;
|
|
}
|
|
|
|
void initiateConnection(tr_peerMgr* mgr, tr_swarm* s, peer_atom& atom)
|
|
{
|
|
using namespace handshake_helpers;
|
|
|
|
auto const now = tr_time();
|
|
bool utp = mgr->session->allowsUTP() && !atom.utp_failed;
|
|
|
|
if (atom.fromFirst == TR_PEER_FROM_PEX)
|
|
{
|
|
/* PEX has explicit signalling for µTP support. If an atom
|
|
originally came from PEX and doesn't have the µTP flag, skip the
|
|
µTP connection attempt. Are we being optimistic here? */
|
|
utp = utp && (atom.flags & ADDED_F_UTP_FLAGS) != 0;
|
|
}
|
|
|
|
auto* const session = mgr->session;
|
|
|
|
if (tr_peer_socket::limit_reached(session) || (!utp && !session->allowsTCP()))
|
|
{
|
|
return;
|
|
}
|
|
|
|
tr_logAddTraceSwarm(
|
|
s,
|
|
fmt::format("Starting an OUTGOING {} connection with {}", utp ? " µTP" : "TCP", atom.display_name()));
|
|
|
|
auto peer_io = tr_peerIo::new_outgoing(
|
|
session,
|
|
&session->top_bandwidth_,
|
|
atom.addr,
|
|
atom.port,
|
|
s->tor->infoHash(),
|
|
s->tor->completeness == TR_SEED,
|
|
utp);
|
|
|
|
if (!peer_io)
|
|
{
|
|
tr_logAddTraceSwarm(s, fmt::format("peerIo not created; marking peer {} as unreachable", atom.display_name()));
|
|
atom.flags2 |= MyflagUnreachable;
|
|
++atom.num_fails;
|
|
}
|
|
else
|
|
{
|
|
s->outgoing_handshakes.try_emplace(
|
|
atom.addr,
|
|
&mgr->handshake_mediator_,
|
|
peer_io,
|
|
session->encryptionMode(),
|
|
[mgr](tr_handshake::Result const& result) { return on_handshake_done(mgr, result); });
|
|
}
|
|
|
|
atom.lastConnectionAttemptAt = now;
|
|
atom.time = now;
|
|
}
|
|
} // namespace connect_helpers
|
|
} // namespace
|
|
|
|
void tr_peerMgr::makeNewPeerConnections(size_t max)
|
|
{
|
|
using namespace connect_helpers;
|
|
|
|
auto const lock = session->unique_lock();
|
|
|
|
for (auto& candidate : getPeerCandidates(session, max))
|
|
{
|
|
initiateConnection(this, candidate.tor->swarm, *candidate.atom);
|
|
}
|
|
}
|
|
|
|
// ---
|
|
|
|
bool HandshakeMediator::is_peer_known_seed(tr_torrent_id_t tor_id, tr_address const& addr) const
|
|
{
|
|
auto const* const tor = session_.torrents().get(tor_id);
|
|
return tor != nullptr && tor->swarm != nullptr && tor->swarm->peer_is_a_seed(addr);
|
|
}
|