settings_pack load_pack_from_dict(bdecode_node const& settings)
{
settings_pack pack;
for (int i = 0; i < settings.dict_size(); ++i)
{
string_view key;
bdecode_node val;
std::tie(key, val) = settings.dict_at(i);
switch (val.type())
{
case bdecode_node::dict_t:
case bdecode_node::list_t:
continue;
case bdecode_node::int_t:
{
bool found = false;
for (int k = 0; k < sizeof(int_settings) / sizeof(int_settings[0]); ++k)
{
if (key != int_settings[k].name) continue;
pack.set_int(settings_pack::int_type_base + k, val.int_value());
found = true;
break;
}
if (found) continue;
for (int k = 0; k < sizeof(bool_settings) / sizeof(bool_settings[0]); ++k)
{
if (key != bool_settings[k].name) continue;
pack.set_bool(settings_pack::bool_type_base + k, val.int_value() != 0);
break;
}
}
break;
case bdecode_node::string_t:
for (int k = 0; k < sizeof(str_settings) / sizeof(str_settings[0]); ++k)
{
if (key != str_settings[k].name) continue;
pack.set_str(settings_pack::string_type_base + k, val.string_value().to_string());
break;
}
break;
case bdecode_node::none_t:
break;
}
}
return pack;
}
// convert a bdecode_node into an old skool entry
void entry::operator=(bdecode_node const& e)
{
switch (e.type())
{
case bdecode_node::string_t:
this->string() = e.string_value();
break;
case bdecode_node::int_t:
this->integer() = e.int_value();
break;
case bdecode_node::dict_t:
{
dictionary_type& d = this->dict();
for (int i = 0; i < e.dict_size(); ++i)
{
std::pair<std::string, bdecode_node> elem = e.dict_at(i);
d[elem.first] = elem.second;
}
break;
}
case bdecode_node::list_t:
{
list_type& l = this->list();
for (int i = 0; i < e.list_size(); ++i)
{
l.push_back(entry());
l.back() = e.list_at(i);
}
break;
}
case bdecode_node::none_t:
destruct();
break;
}
}
bool item::assign(bdecode_node const& v, span<char const> salt
, sequence_number const seq, public_key const& pk, signature const& sig)
{
TORRENT_ASSERT(v.data_section().size() <= 1000);
if (!verify_mutable_item(v.data_section(), salt, seq, pk, sig))
return false;
m_pk = pk;
m_sig = sig;
if (salt.size() > 0)
m_salt.assign(salt.data(), salt.size());
else
m_salt.clear();
m_seq = seq;
m_mutable = true;
m_value = v;
return true;
}
void render_variant(std::string& out, bdecode_node const& e)
{
switch (e.type())
{
case bdecode_node::dict_t:
print_dict(out);
for (int i = 0; i < e.dict_size(); ++i)
{
std::pair<lt::string_view, bdecode_node> item = e.dict_at(i);
const bool duplicate = g_seed == 1;
const bool skipped = g_seed == 2;
g_seed -= 2;
if (duplicate)
{
print_string(out, item.first.to_string());
render_variant(out, item.second);
}
if (!skipped)
{
print_string(out, item.first.to_string());
render_variant(out, item.second);
}
render_arbitrary_item(out);
}
print_terminate(out);
break;
case bdecode_node::list_t:
print_list(out);
for (int i = 0; i < e.list_size(); ++i)
{
const bool duplicate = g_seed == 1;
const bool skipped = g_seed == 2;
g_seed -= 2;
if (duplicate) render_variant(out, e.list_at(i));
render_arbitrary_item(out);
if (!skipped) render_variant(out, e.list_at(i));
}
print_terminate(out);
break;
case bdecode_node::int_t:
print_int(out, e.int_value());
break;
case bdecode_node::string_t:
print_string(out, e.string_value().to_string());
break;
default:
abort();
}
}
// root_dir is the name of the torrent, unless this is a single file
// torrent, in which case it's empty.
bool extract_files(bdecode_node const& list, file_storage& target
, std::string const& root_dir, ptrdiff_t info_ptr_diff, error_code& ec)
{
if (list.type() != bdecode_node::list_t)
{
ec = errors::torrent_file_parse_failed;
return false;
}
target.reserve(list.list_size());
// this is the counter used to name pad files
int pad_file_cnt = 0;
for (int i = 0, end(list.list_size()); i < end; ++i)
{
if (!extract_single_file(list.list_at(i), target, root_dir
, info_ptr_diff, false, pad_file_cnt, ec))
return false;
}
return true;
}
void get_item::got_data(bdecode_node const& v,
public_key const& pk,
sequence_number const seq,
signature const& sig)
{
// we received data!
// if no data_callback, we needn't care about the data we get.
// only put_immutable_item no data_callback
if (!m_data_callback) return;
// for get_immutable_item
if (m_immutable)
{
// If m_data isn't empty, we should have post alert.
if (!m_data.empty()) return;
sha1_hash incoming_target = item_target_id(v.data_section());
if (incoming_target != target()) return;
m_data.assign(v);
// There can only be one true immutable item with a given id
// Now that we've got it and the user doesn't want to do a put
// there's no point in continuing to query other nodes
m_data_callback(m_data, true);
done();
return;
}
// immutable data should have been handled before this line, only mutable
// data can reach here, which means pk, sig and seq must be valid.
std::string const salt_copy(m_data.salt());
sha1_hash const incoming_target = item_target_id(salt_copy, pk);
if (incoming_target != target()) return;
// this is mutable data. If it passes the signature
// check, remember it. Just keep the version with
// the highest sequence number.
if (m_data.empty() || m_data.seq() < seq)
{
if (!m_data.assign(v, salt_copy, seq, pk, sig))
return;
// for get_item, we should call callback when we get data,
// even if the date is not authoritative, we can update later.
// so caller can get response ASAP without waiting transaction
// time-out (15 seconds).
// for put_item, the callback function will do nothing
// if the data is non-authoritative.
m_data_callback(m_data, false);
}
}
// TODO: 2 returning a bool here is redundant. Instead this function should
// return the peer_entry
bool extract_peer_info(bdecode_node const& info, peer_entry& ret, error_code& ec)
{
// extract peer id (if any)
if (info.type() != bdecode_node::dict_t)
{
ec.assign(errors::invalid_peer_dict, get_libtorrent_category());
return false;
}
bdecode_node i = info.dict_find_string("peer id");
if (i && i.string_length() == 20)
{
std::copy(i.string_ptr(), i.string_ptr()+20, ret.pid.begin());
}
else
{
// if there's no peer_id, just initialize it to a bunch of zeroes
std::fill_n(ret.pid.begin(), 20, 0);
}
// extract ip
i = info.dict_find_string("ip");
if (i == 0)
{
ec.assign(errors::invalid_tracker_response, get_libtorrent_category());
return false;
}
ret.hostname = i.string_value();
// extract port
i = info.dict_find_int("port");
if (i == 0)
{
ec.assign(errors::invalid_tracker_response, get_libtorrent_category());
return false;
}
ret.port = boost::uint16_t(i.int_value());
return true;
}
void on_read(lt::error_code const& ec, std::size_t bytes_transferred)
{
if (ec) return;
using libtorrent::entry;
using libtorrent::bdecode;
int pos;
error_code err;
// since the simulation is single threaded, we can get away with just
// allocating a single of these
static bdecode_node msg;
int ret = bdecode(m_buffer, m_buffer + bytes_transferred, msg, err, &pos, 10, 500);
if (ret != 0) return;
if (msg.type() != bdecode_node::dict_t) return;
libtorrent::dht::msg m(msg, m_ep);
dht().incoming(m);
sock().async_receive_from(asio::mutable_buffers_1(m_buffer, sizeof(m_buffer))
, m_ep, boost::bind(&dht_node::on_read, this, _1, _2));
}
bool item::assign(bdecode_node const& v
, std::pair<char const*, int> salt
, boost::uint64_t seq, char const* pk, char const* sig)
{
TORRENT_ASSERT(v.data_section().second <= 1000);
if (pk && sig)
{
if (!verify_mutable_item(v.data_section(), salt, seq, pk, sig))
return false;
memcpy(m_pk.c_array(), pk, item_pk_len);
memcpy(m_sig.c_array(), sig, item_sig_len);
if (salt.second > 0)
m_salt.assign(salt.first, salt.second);
else
m_salt.clear();
m_seq = seq;
m_mutable = true;
}
else
m_mutable = false;
m_value = v;
return true;
}
bool verify_message(bdecode_node const& message, key_desc_t const desc[]
, bdecode_node ret[], int size, char* error, int error_size)
{
// get a non-root bdecode_node that still
// points to the root. message should not be copied
bdecode_node msg = message.non_owning();
// clear the return buffer
for (int i = 0; i < size; ++i)
ret[i].clear();
// when parsing child nodes, this is the stack
// of bdecode_nodes to return to
bdecode_node stack[5];
int stack_ptr = -1;
if (msg.type() != bdecode_node::dict_t)
{
snprintf(error, error_size, "not a dictionary");
return false;
}
++stack_ptr;
stack[stack_ptr] = msg;
for (int i = 0; i < size; ++i)
{
key_desc_t const& k = desc[i];
// fprintf(stderr, "looking for %s in %s\n", k.name, print_entry(*msg).c_str());
ret[i] = msg.dict_find(k.name);
// none_t means any type
if (ret[i] && ret[i].type() != k.type && k.type != bdecode_node::none_t)
ret[i].clear();
if (ret[i] == 0 && (k.flags & key_desc_t::optional) == 0)
{
// the key was not found, and it's not an optional key
snprintf(error, error_size, "missing '%s' key", k.name);
return false;
}
if (k.size > 0
&& ret[i]
&& k.type == bdecode_node::string_t)
{
bool invalid = false;
if (k.flags & key_desc_t::size_divisible)
invalid = (ret[i].string_length() % k.size) != 0;
else
invalid = ret[i].string_length() != k.size;
if (invalid)
{
// the string was not of the required size
ret[i].clear();
if ((k.flags & key_desc_t::optional) == 0)
{
snprintf(error, error_size, "invalid value for '%s'", k.name);
return false;
}
}
}
if (k.flags & key_desc_t::parse_children)
{
TORRENT_ASSERT(k.type == bdecode_node::dict_t);
if (ret[i])
{
++stack_ptr;
TORRENT_ASSERT(stack_ptr < int(sizeof(stack) / sizeof(stack[0])));
msg = ret[i];
stack[stack_ptr] = msg;
}
else
{
// skip all children
while (i < size && (desc[i].flags & key_desc_t::last_child) == 0) ++i;
// if this assert is hit, desc is incorrect
TORRENT_ASSERT(i < size);
}
}
else if (k.flags & key_desc_t::last_child)
{
TORRENT_ASSERT(stack_ptr > 0);
// this can happen if the specification passed
// in is unbalanced. i.e. contain more last_child
// nodes than parse_children
if (stack_ptr == 0) return false;
--stack_ptr;
msg = stack[stack_ptr];
}
}
return true;
}
int bdecode(char const* start, char const* end, bdecode_node& ret
, error_code& ec, int* error_pos, int depth_limit, int token_limit)
{
ec.clear();
ret.clear();
if (end - start > bdecode_token::max_offset)
{
if (error_pos) *error_pos = 0;
ec = make_error_code(bdecode_errors::limit_exceeded);
return -1;
}
// this is the stack of bdecode_token indices, into m_tokens.
// sp is the stack pointer, as index into the array, stack
int sp = 0;
stack_frame* stack = TORRENT_ALLOCA(stack_frame, depth_limit);
char const* const orig_start = start;
if (start == end) return 0;
while (start <= end)
{
if (start >= end) TORRENT_FAIL_BDECODE(bdecode_errors::unexpected_eof);
if (sp >= depth_limit)
TORRENT_FAIL_BDECODE(bdecode_errors::depth_exceeded);
--token_limit;
if (token_limit < 0)
TORRENT_FAIL_BDECODE(bdecode_errors::limit_exceeded);
// look for a new token
const char t = *start;
const int current_frame = sp;
// if we're currently parsing a dictionary, assert that
// every other node is a string.
if (current_frame > 0
&& ret.m_tokens[stack[current_frame-1].token].type == bdecode_token::dict)
{
if (stack[current_frame-1].state == 0)
{
// the current parent is a dict and we are parsing a key.
// only allow a digit (for a string) or 'e' to terminate
if (!numeric(t) && t != 'e')
TORRENT_FAIL_BDECODE(bdecode_errors::expected_digit);
}
}
switch (t)
{
case 'd':
stack[sp++] = ret.m_tokens.size();
// we push it into the stack so that we know where to fill
// in the next_node field once we pop this node off the stack.
// i.e. get to the node following the dictionary in the buffer
ret.m_tokens.push_back(bdecode_token(start - orig_start
, bdecode_token::dict));
++start;
break;
case 'l':
stack[sp++] = ret.m_tokens.size();
// we push it into the stack so that we know where to fill
// in the next_node field once we pop this node off the stack.
// i.e. get to the node following the list in the buffer
ret.m_tokens.push_back(bdecode_token(start - orig_start
, bdecode_token::list));
++start;
break;
case 'i':
{
char const* int_start = start;
bdecode_errors::error_code_enum e = bdecode_errors::no_error;
// +1 here to point to the first digit, rather than 'i'
start = check_integer(start + 1, end, e);
if (e)
{
// in order to gracefully terminate the tree,
// make sure the end of the previous token is set correctly
if (error_pos) *error_pos = start - orig_start;
error_pos = NULL;
start = int_start;
TORRENT_FAIL_BDECODE(e);
}
ret.m_tokens.push_back(bdecode_token(int_start - orig_start
, 1, bdecode_token::integer, 1));
TORRENT_ASSERT(*start == 'e');
// skip 'e'
++start;
break;
}
case 'e':
{
// this is the end of a list or dict
if (sp == 0)
TORRENT_FAIL_BDECODE(bdecode_errors::unexpected_eof);
if (sp > 0
&& ret.m_tokens[stack[sp-1].token].type == bdecode_token::dict
&& stack[sp-1].state == 1)
{
// this means we're parsing a dictionary and about to parse a
// value associated with a key. Instad, we got a termination
TORRENT_FAIL_BDECODE(bdecode_errors::expected_value);
}
// insert the end-of-sequence token
ret.m_tokens.push_back(bdecode_token(start - orig_start, 1
, bdecode_token::end));
// and back-patch the start of this sequence with the offset
// to the next token we'll insert
int top = stack[sp-1].token;
// subtract the token's own index, since this is a relative
// offset
if (ret.m_tokens.size() - top > bdecode_token::max_next_item)
TORRENT_FAIL_BDECODE(bdecode_errors::limit_exceeded);
ret.m_tokens[top].next_item = ret.m_tokens.size() - top;
// and pop it from the stack.
--sp;
++start;
break;
}
default:
{
// this is the case for strings. The start character is any
// numeric digit
if (!numeric(t))
TORRENT_FAIL_BDECODE(bdecode_errors::expected_value);
boost::int64_t len = t - '0';
char const* str_start = start;
++start;
bdecode_errors::error_code_enum e = bdecode_errors::no_error;
start = parse_int(start, end, ':', len, e);
if (e)
TORRENT_FAIL_BDECODE(e);
// remaining buffer size excluding ':'
const ptrdiff_t buff_size = end - start - 1;
if (len > buff_size)
TORRENT_FAIL_BDECODE(bdecode_errors::unexpected_eof);
if (len < 0)
TORRENT_FAIL_BDECODE(bdecode_errors::overflow);
// skip ':'
++start;
if (start >= end) TORRENT_FAIL_BDECODE(bdecode_errors::unexpected_eof);
// the bdecode_token only has 8 bits to keep the header size
// in. If it overflows, fail!
if (start - str_start - 2 > detail::bdecode_token::max_header)
TORRENT_FAIL_BDECODE(bdecode_errors::limit_exceeded);
ret.m_tokens.push_back(bdecode_token(str_start - orig_start
, 1, bdecode_token::string, start - str_start));
start += len;
break;
}
}
if (current_frame > 0
&& ret.m_tokens[stack[current_frame-1].token].type == bdecode_token::dict)
{
// the next item we parse is the opposite
stack[current_frame-1].state = ~stack[current_frame-1].state;
}
// this terminates the top level node, we're done!
if (sp == 0) break;
}
done:
// if parse failed, sp will be greater than 1
// unwind the stack by inserting terminator to make whatever we have
// so far valid
while (sp > 0) {
TORRENT_ASSERT(ec);
--sp;
// we may need to insert a dummy token to properly terminate the tree,
// in case we just parsed a key to a dict and failed in the value
if (ret.m_tokens[stack[sp].token].type == bdecode_token::dict
&& stack[sp].state == 1)
{
// insert an empty dictionary as the value
ret.m_tokens.push_back(bdecode_token(start - orig_start
, 2, bdecode_token::dict));
ret.m_tokens.push_back(bdecode_token(start - orig_start
, bdecode_token::end));
}
int top = stack[sp].token;
TORRENT_ASSERT(ret.m_tokens.size() - top <= bdecode_token::max_next_item);
ret.m_tokens[top].next_item = ret.m_tokens.size() - top;
ret.m_tokens.push_back(bdecode_token(start - orig_start, 1, bdecode_token::end));
}
ret.m_tokens.push_back(bdecode_token(start - orig_start, 0
, bdecode_token::end));
ret.m_token_idx = 0;
ret.m_buffer = orig_start;
ret.m_buffer_size = start - orig_start;
ret.m_root_tokens = &ret.m_tokens[0];
return ec ? -1 : 0;
}
std::string print_entry(bdecode_node const& e
, bool single_line, int indent)
{
char indent_str[200];
using std::memset;
memset(indent_str, ' ', 200);
indent_str[0] = ',';
indent_str[1] = '\n';
indent_str[199] = 0;
if (indent < 197 && indent >= 0) indent_str[indent+2] = 0;
std::string ret;
switch (e.type())
{
case bdecode_node::none_t: return "none";
case bdecode_node::int_t:
{
char str[100];
snprintf(str, sizeof(str), "%" PRId64, e.int_value());
return str;
}
case bdecode_node::string_t:
{
print_string(ret, e.string_ptr(), e.string_length(), single_line);
return ret;
}
case bdecode_node::list_t:
{
ret += '[';
bool one_liner = line_longer_than(e, 200) != -1 || single_line;
if (!one_liner) ret += indent_str + 1;
for (int i = 0; i < e.list_size(); ++i)
{
if (i == 0 && one_liner) ret += " ";
ret += print_entry(e.list_at(i), single_line, indent + 2);
if (i < e.list_size() - 1) ret += (one_liner?", ":indent_str);
else ret += (one_liner?" ":indent_str+1);
}
ret += "]";
return ret;
}
case bdecode_node::dict_t:
{
ret += "{";
bool one_liner = line_longer_than(e, 200) != -1 || single_line;
if (!one_liner) ret += indent_str+1;
for (int i = 0; i < e.dict_size(); ++i)
{
if (i == 0 && one_liner) ret += " ";
std::pair<std::string, bdecode_node> ent = e.dict_at(i);
print_string(ret, ent.first.c_str(), ent.first.size(), true);
ret += ": ";
ret += print_entry(ent.second, single_line, indent + 2);
if (i < e.dict_size() - 1) ret += (one_liner?", ":indent_str);
else ret += (one_liner?" ":indent_str+1);
}
ret += "}";
return ret;
}
}
return ret;
}
// 'top_level' is extracting the file for a single-file torrent. The
// distinction is that the filename is found in "name" rather than
// "path"
// root_dir is the name of the torrent, unless this is a single file
// torrent, in which case it's empty.
bool extract_single_file(bdecode_node const& dict, file_storage& files
, std::string const& root_dir, ptrdiff_t info_ptr_diff, bool top_level
, int& pad_file_cnt, error_code& ec)
{
if (dict.type() != bdecode_node::dict_t) return false;
boost::uint32_t file_flags = get_file_attributes(dict);
// symlinks have an implied "size" of zero. i.e. they use up 0 bytes of
// the torrent payload space
boost::int64_t const file_size = (file_flags & file_storage::flag_symlink)
? 0
: dict.dict_find_int_value("length", -1);
if (file_size < 0 )
{
ec = errors::torrent_invalid_length;
return false;
}
boost::int64_t const mtime = dict.dict_find_int_value("mtime", 0);
std::string path = root_dir;
std::string path_element;
char const* filename = NULL;
int filename_len = 0;
if (top_level)
{
// prefer the name.utf-8 because if it exists, it is more likely to be
// correctly encoded
bdecode_node p = dict.dict_find_string("name.utf-8");
if (!p) p = dict.dict_find_string("name");
if (!p || p.string_length() == 0)
{
ec = errors::torrent_missing_name;
return false;
}
filename = p.string_ptr() + info_ptr_diff;
filename_len = p.string_length();
while (filename_len > 0 && filename[0] == TORRENT_SEPARATOR)
{
filename += 1;
filename_len -= 1;
}
sanitize_append_path_element(path, p.string_ptr(), p.string_length());
}
else
{
bdecode_node p = dict.dict_find_list("path.utf-8");
if (!p) p = dict.dict_find_list("path");
if (p && p.list_size() > 0)
{
std::size_t const orig_path_len = path.size();
int const preallocate = path.size() + path_length(p, ec);
if (ec) return false;
path.reserve(preallocate);
for (int i = 0, end(p.list_size()); i < end; ++i)
{
bdecode_node e = p.list_at(i);
if (i == end - 1)
{
filename = e.string_ptr() + info_ptr_diff;
filename_len = e.string_length();
}
while (filename_len > 0 && filename[0] == TORRENT_SEPARATOR)
{
filename += 1;
filename_len -= 1;
}
sanitize_append_path_element(path, e.string_ptr(), e.string_length());
}
// if all path elements were sanitized away, we need to use another
// name instead
if (path.size() == orig_path_len)
{
path += TORRENT_SEPARATOR;
path += "_";
}
}
else if (file_flags & file_storage::flag_pad_file)
{
// pad files don't need a path element, we'll just store them
// under the .pad directory
char cnt[10];
snprintf(cnt, sizeof(cnt), "%d", pad_file_cnt);
path = combine_path(".pad", cnt);
++pad_file_cnt;
}
else
{
ec = errors::torrent_missing_name;
return false;
}
}
// bitcomet pad file
if (path.find("_____padding_file_") != std::string::npos)
file_flags = file_storage::flag_pad_file;
bdecode_node fh = dict.dict_find_string("sha1");
char const* filehash = NULL;
if (fh && fh.string_length() == 20)
filehash = fh.string_ptr() + info_ptr_diff;
std::string symlink_path;
if (file_flags & file_storage::flag_symlink)
{
if (bdecode_node s_p = dict.dict_find_list("symlink path"))
{
int const preallocate = path_length(s_p, ec);
if (ec) return false;
symlink_path.reserve(preallocate);
for (int i = 0, end(s_p.list_size()); i < end; ++i)
{
bdecode_node const& n = s_p.list_at(i);
sanitize_append_path_element(symlink_path, n.string_ptr()
, n.string_length());
}
}
}
else
{
file_flags &= ~file_storage::flag_symlink;
}
if (filename_len > path.length()
|| path.compare(path.size() - filename_len, filename_len, filename
, filename_len) != 0)
{
// if the filename was sanitized and differ, clear it to just use path
filename = NULL;
filename_len = 0;
}
files.add_file_borrow(filename, filename_len, path, file_size, file_flags, filehash
, mtime, symlink_path);
return true;
}
void get_item_observer::reply(msg const& m)
{
public_key pk;
signature sig;
sequence_number seq{0};
bdecode_node const r = m.message.dict_find_dict("r");
if (!r)
{
#ifndef TORRENT_DISABLE_LOGGING
get_observer()->log(dht_logger::traversal, "[%p] missing response dict"
, static_cast<void*>(algorithm()));
#endif
timeout();
return;
}
bdecode_node const k = r.dict_find_string("k");
if (k && k.string_length() == public_key::len)
std::memcpy(pk.bytes.data(), k.string_ptr(), public_key::len);
bdecode_node const s = r.dict_find_string("sig");
if (s && s.string_length() == signature::len)
std::memcpy(sig.bytes.data(), s.string_ptr(), signature::len);
bdecode_node const q = r.dict_find_int("seq");
if (q)
{
seq = sequence_number(q.int_value());
}
else if (k && s)
{
timeout();
return;
}
bdecode_node v = r.dict_find("v");
if (v)
{
static_cast<get_item*>(algorithm())->got_data(v, pk, seq, sig);
}
find_data_observer::reply(m);
}
