transfers kinda work now. request algo broken tho

This commit is contained in:
Green Sky 2023-01-16 03:50:47 +01:00
parent 8b554bd80f
commit e89f1be660
No known key found for this signature in database
3 changed files with 177 additions and 53 deletions

View File

@ -43,7 +43,7 @@ namespace std { // inject
struct FTInfoSHA1 { struct FTInfoSHA1 {
std::string file_name; std::string file_name;
uint64_t file_size {0}; uint64_t file_size {0};
static constexpr size_t chunk_size {4*1024}; // 4KiB for now static constexpr size_t chunk_size {64*1024}; // 64KiB for now
std::vector<SHA1Digest> chunks; std::vector<SHA1Digest> chunks;
std::vector<uint8_t> toBuffer(void) const; std::vector<uint8_t> toBuffer(void) const;

View File

@ -2,6 +2,8 @@
#include "../tox_client.hpp" #include "../tox_client.hpp"
#include "../hash_utils.hpp"
#include <iostream> #include <iostream>
#include <tuple> #include <tuple>
@ -23,16 +25,21 @@ SHA1::SHA1(
_sha1_info_hash(std::move(sha1_info_hash)), _sha1_info_hash(std::move(sha1_info_hash)),
_have_chunk(std::move(have_chunk)) _have_chunk(std::move(have_chunk))
{ {
assert(_have_chunk.size() == _sha1_info.chunks.size());
_have_all = true; _have_all = true;
_have_count = 0; _have_count = 0;
for (const bool it : _have_chunk) { for (size_t i = 0; i < _have_chunk.size(); i++) {
if (!it) { if (_have_chunk[i]) {
_have_all = false;
} else {
_have_count++; _have_count++;
} else {
_have_all = false;
_chunk_want_queue.push_back(i);
} }
} }
// if not sequential, shuffle _chunk_want_queue
// build lookup table // build lookup table
for (size_t i = 0; i < _sha1_info.chunks.size(); i++) { for (size_t i = 0; i < _sha1_info.chunks.size(); i++) {
_chunk_hash_to_index[_sha1_info.chunks[i]] = i; _chunk_hash_to_index[_sha1_info.chunks[i]] = i;
@ -40,7 +47,7 @@ SHA1::SHA1(
} }
bool SHA1::iterate(float delta) { bool SHA1::iterate(float delta) {
// do ongoing transfers, send data?, timeout { // timer and timeouts
// info // info
for (auto it = _transfers_requested_info.begin(); it != _transfers_requested_info.end();) { for (auto it = _transfers_requested_info.begin(); it != _transfers_requested_info.end();) {
float& time_since_remove_activity = std::get<float>(*it); float& time_since_remove_activity = std::get<float>(*it);
@ -48,7 +55,7 @@ bool SHA1::iterate(float delta) {
// if we have not heard for 10sec, timeout // if we have not heard for 10sec, timeout
if (time_since_remove_activity >= 10.f) { if (time_since_remove_activity >= 10.f) {
std::cerr << "SHA1 info tansfer timed out " << std::get<0>(*it) << ":" << std::get<1>(*it) << "." << std::get<2>(*it) << "\n"; std::cerr << "SHA1 info tansfer timed out " << std::get<0>(*it) << ":" << std::get<1>(*it) << "." << int(std::get<2>(*it)) << "\n";
it = _transfers_requested_info.erase(it); it = _transfers_requested_info.erase(it);
} else { } else {
it++; it++;
@ -61,7 +68,7 @@ bool SHA1::iterate(float delta) {
// if we have not heard for 10sec, timeout // if we have not heard for 10sec, timeout
if (time_since_remove_activity >= 10.f) { if (time_since_remove_activity >= 10.f) {
std::cerr << "SHA1 sending chunk tansfer timed out " << std::get<0>(*it) << ":" << std::get<1>(*it) << "." << std::get<2>(*it) << "\n"; std::cerr << "SHA1 sending chunk tansfer timed out " << std::get<0>(*it) << ":" << std::get<1>(*it) << "." << int(std::get<2>(*it)) << "\n";
it = _transfers_sending_chunk.erase(it); it = _transfers_sending_chunk.erase(it);
} else { } else {
it++; it++;
@ -74,15 +81,16 @@ bool SHA1::iterate(float delta) {
// if we have not heard for 10sec, timeout // if we have not heard for 10sec, timeout
if (time_since_remove_activity >= 10.f) { if (time_since_remove_activity >= 10.f) {
std::cerr << "SHA1 receiving chunk tansfer timed out " << std::get<0>(*it) << ":" << std::get<1>(*it) << "." << std::get<2>(*it) << "\n"; std::cerr << "SHA1 receiving chunk tansfer timed out " << std::get<0>(*it) << ":" << std::get<1>(*it) << "." << int(std::get<2>(*it)) << "\n";
it = _transfers_receiving_chunk.erase(it); it = _transfers_receiving_chunk.erase(it);
} else { } else {
it++; it++;
} }
} }
}
// if we have not reached the total cap for transfers // if we have not reached the total cap for transfers
if (true) { if (_transfers_requested_info.size() + _transfers_sending_chunk.size() < _max_concurrent_out) {
// for each peer? transfer cap per peer? // for each peer? transfer cap per peer?
// first check requests for info // first check requests for info
@ -111,11 +119,7 @@ bool SHA1::iterate(float delta) {
const auto [group_number, peer_number, chunk_hash] = _queue_requested_chunk.front(); const auto [group_number, peer_number, chunk_hash] = _queue_requested_chunk.front();
size_t chunk_index = chunkIndex(chunk_hash).value(); size_t chunk_index = chunkIndex(chunk_hash).value();
size_t chunk_file_size = _sha1_info.chunk_size; size_t chunk_file_size = chunkSize(chunk_index);
if (chunk_index+1 == _sha1_info.chunks.size()) {
// last chunk
chunk_file_size = _sha1_info.file_size - chunk_index * _sha1_info.chunk_size;
}
uint8_t transfer_id {0}; uint8_t transfer_id {0};
@ -138,6 +142,33 @@ bool SHA1::iterate(float delta) {
} }
} }
if (!_have_all && !_chunk_want_queue.empty() && _transfers_receiving_chunk.size() < _max_concurrent_in) {
// send out request, no burst tho
std::vector<std::pair<uint32_t, uint32_t>> target_peers;
_tcl.forEachGroup([&target_peers, this](uint32_t group_number) {
_tcl.forEachGroupPeer(group_number, [&target_peers, group_number](uint32_t peer_number) {
target_peers.push_back({group_number, peer_number});
});
});
if (!target_peers.empty()) {
//if (_distrib.max() != target_peers.size()) {
//std::uniform_int_distribution<size_t> new_dist{0, target_peers.size()-1};
//_distrib.param(new_dist.param());
//}
//size_t target_index = _distrib(_rng);
size_t target_index = _rng()%target_peers.size();
auto [group_number, peer_number] = target_peers.at(target_index);
size_t chunk_index = _chunk_want_queue.front();
_chunks_requested.emplace(chunk_index);
_chunk_want_queue.pop_front();
_tcl.sendFT1RequestPrivate(group_number, peer_number, NGC_FT1_file_kind::HASH_SHA1_CHUNK, _sha1_info.chunks[chunk_index].data.data(), 20);
}
}
// TODO: unmap and remap the file every couple of minutes to keep ram usage down? // TODO: unmap and remap the file every couple of minutes to keep ram usage down?
// TODO: when to stop? // TODO: when to stop?
return false; return false;
@ -227,10 +258,89 @@ void SHA1::onFT1ReceiveRequestSHA1Chunk(uint32_t group_number, uint32_t peer_num
} }
bool SHA1::onFT1ReceiveInitSHA1Chunk(uint32_t group_number, uint32_t peer_number, const uint8_t* file_id, size_t file_id_size, const uint8_t transfer_id, const size_t file_size) { bool SHA1::onFT1ReceiveInitSHA1Chunk(uint32_t group_number, uint32_t peer_number, const uint8_t* file_id, size_t file_id_size, const uint8_t transfer_id, const size_t file_size) {
if (_transfers_receiving_chunk.size() >= _max_concurrent_in) {
// reject, max tf in
return false; return false;
} }
if (file_id_size != 20) {
std::cerr << "SHA1 got request for sha1_chunk of wrong size!!\n";
return false;
}
SHA1Digest incomming_hash(file_id, file_id_size);
if (haveChunk(incomming_hash)) {
std::cout << "SHA1 ignoring init for chunk we allready have " << incomming_hash << "\n";
return false;
}
auto chunk_i_opt = chunkIndex(incomming_hash);
if (!chunk_i_opt.has_value()) {
std::cout << "SHA1 ignoring init for unrelated chunk " << incomming_hash << "\n";
return false;
}
size_t chunk_index = chunk_i_opt.value();
// check transfers
for (const auto& it : _transfers_receiving_chunk) {
if (std::get<4>(it) == chunk_index) {
// allready in transition
return false;
}
}
_transfers_receiving_chunk.push_back(
std::make_tuple(
group_number, peer_number,
transfer_id,
0.f,
chunk_index
)
);
return true;
}
void SHA1::onFT1ReceiveDataSHA1Chunk(uint32_t group_number, uint32_t peer_number, uint8_t transfer_id, size_t data_offset, const uint8_t* data, size_t data_size) { void SHA1::onFT1ReceiveDataSHA1Chunk(uint32_t group_number, uint32_t peer_number, uint8_t transfer_id, size_t data_offset, const uint8_t* data, size_t data_size) {
// check transfers
for (auto it = _transfers_receiving_chunk.begin(); it != _transfers_receiving_chunk.end(); it++) {
if (std::get<0>(*it) == group_number && std::get<1>(*it) == peer_number && std::get<2>(*it) == transfer_id) {
std::get<float>(*it) = 0.f; // time
const size_t chunk_index = std::get<4>(*it);
size_t file_offset = chunk_index * _sha1_info.chunk_size;
// TODO: optimize
for (size_t i = 0; i < data_size; i++) {
_file_map[file_offset+data_offset+i] = data[i];
}
size_t chunk_file_size = chunkSize(chunk_index);
// if last data block
if (data_offset + data_size == chunk_file_size) {
// hash and verify
SHA1Digest test_hash = hash_sha1(_file_map.data()+file_offset, chunk_file_size);
if (test_hash != _sha1_info.chunks[chunk_index]) {
std::cerr << "SHA1 received chunks's hash does not match!, discarding\n";
_transfers_receiving_chunk.erase(it);
break;
}
_have_chunk[chunk_index] = true;
_have_count++;
_have_all = _have_count == _sha1_info.chunks.size();
std::cout << "SHA1 chunk received " << std::get<0>(*it) << ":" << std::get<1>(*it) << "." << int(std::get<2>(*it)) << " " << chunk_index << " (" << 100.f * float(_have_count) / _sha1_info.chunks.size() << "%)\n";
_transfers_receiving_chunk.erase(it);
}
break;
}
}
} }
void SHA1::onFT1SendDataSHA1Chunk(uint32_t group_number, uint32_t peer_number, uint8_t transfer_id, size_t data_offset, uint8_t* data, size_t data_size) { void SHA1::onFT1SendDataSHA1Chunk(uint32_t group_number, uint32_t peer_number, uint8_t transfer_id, size_t data_offset, uint8_t* data, size_t data_size) {
@ -248,15 +358,9 @@ void SHA1::onFT1SendDataSHA1Chunk(uint32_t group_number, uint32_t peer_number, u
data[i] = _file_map[file_offset+data_offset+i]; data[i] = _file_map[file_offset+data_offset+i];
} }
size_t chunk_file_size = _sha1_info.chunk_size;
if (chunk_index+1 == _sha1_info.chunks.size()) {
// last chunk
chunk_file_size = _sha1_info.file_size - chunk_index * _sha1_info.chunk_size;
}
// if last data block // if last data block
if (data_offset + data_size == chunk_file_size) { if (data_offset + data_size == chunkSize(chunk_index)) {
std::cout << "SHA1 chunk sent " << std::get<0>(*it) << ":" << std::get<1>(*it) << "." << std::get<2>(*it) << " " << chunk_index << "\n"; std::cout << "SHA1 chunk sent " << std::get<0>(*it) << ":" << std::get<1>(*it) << "." << int(std::get<2>(*it)) << " " << chunk_index << "\n";
_transfers_sending_chunk.erase(it); _transfers_sending_chunk.erase(it);
} }
@ -308,6 +412,15 @@ std::optional<size_t> SHA1::chunkIndex(const SHA1Digest& hash) const {
} }
} }
size_t SHA1::chunkSize(size_t chunk_index) const {
if (chunk_index+1 == _sha1_info.chunks.size()) {
// last chunk
return _sha1_info.file_size - chunk_index * _sha1_info.chunk_size;
} else {
return _sha1_info.chunk_size;
}
}
bool SHA1::haveChunk(const SHA1Digest& hash) const { bool SHA1::haveChunk(const SHA1Digest& hash) const {
if (_have_all) { if (_have_all) {
return true; return true;

View File

@ -7,8 +7,10 @@
#include <mio/mio.hpp> #include <mio/mio.hpp>
#include <unordered_map> #include <unordered_map>
#include <set>
#include <vector> #include <vector>
#include <deque> #include <deque>
#include <random>
namespace States { namespace States {
@ -50,6 +52,7 @@ struct SHA1 final : public StateI {
void queueUpRequestChunk(uint32_t group_number, uint32_t peer_number, const SHA1Digest& hash); void queueUpRequestChunk(uint32_t group_number, uint32_t peer_number, const SHA1Digest& hash);
std::optional<size_t> chunkIndex(const SHA1Digest& hash) const; std::optional<size_t> chunkIndex(const SHA1Digest& hash) const;
size_t chunkSize(size_t chunk_index) const;
bool haveChunk(const SHA1Digest& hash) const; bool haveChunk(const SHA1Digest& hash) const;
private: private:
@ -62,6 +65,14 @@ struct SHA1 final : public StateI {
std::vector<bool> _have_chunk; std::vector<bool> _have_chunk;
bool _have_all {false}; bool _have_all {false};
size_t _have_count {0}; size_t _have_count {0};
std::deque<size_t> _chunk_want_queue;
std::set<size_t> _chunks_requested;
const size_t _max_concurrent_out {4};
const size_t _max_concurrent_in {4};
std::minstd_rand _rng {1337};
std::uniform_int_distribution<size_t> _distrib;
std::unordered_map<SHA1Digest, size_t> _chunk_hash_to_index; std::unordered_map<SHA1Digest, size_t> _chunk_hash_to_index;