You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
bitcoin/src/netbase.cpp

875 lines
30 KiB

// Copyright (c) 2009-2010 Satoshi Nakamoto
// Copyright (c) 2009-2022 The Bitcoin Core developers
// Distributed under the MIT software license, see the accompanying
// file COPYING or http://www.opensource.org/licenses/mit-license.php.
#if defined(HAVE_CONFIG_H)
#include <config/bitcoin-config.h>
#endif
#include <netbase.h>
#include <compat/compat.h>
#include <logging.h>
#include <sync.h>
#include <tinyformat.h>
#include <util/sock.h>
#include <util/strencodings.h>
#include <util/string.h>
#include <util/time.h>
#include <atomic>
#include <chrono>
#include <cstdint>
#include <functional>
#include <limits>
#include <memory>
#if HAVE_SOCKADDR_UN
#include <sys/un.h>
#endif
// Settings
static GlobalMutex g_proxyinfo_mutex;
static Proxy proxyInfo[NET_MAX] GUARDED_BY(g_proxyinfo_mutex);
static Proxy nameProxy GUARDED_BY(g_proxyinfo_mutex);
int nConnectTimeout = DEFAULT_CONNECT_TIMEOUT;
bool fNameLookup = DEFAULT_NAME_LOOKUP;
// Need ample time for negotiation for very slow proxies such as Tor
std::chrono::milliseconds g_socks5_recv_timeout = 20s;
CThreadInterrupt g_socks5_interrupt;
ReachableNets g_reachable_nets;
std::vector<CNetAddr> WrappedGetAddrInfo(const std::string& name, bool allow_lookup)
{
addrinfo ai_hint{};
// We want a TCP port, which is a streaming socket type
ai_hint.ai_socktype = SOCK_STREAM;
ai_hint.ai_protocol = IPPROTO_TCP;
// We don't care which address family (IPv4 or IPv6) is returned
ai_hint.ai_family = AF_UNSPEC;
// If we allow lookups of hostnames, use the AI_ADDRCONFIG flag to only
// return addresses whose family we have an address configured for.
//
// If we don't allow lookups, then use the AI_NUMERICHOST flag for
// getaddrinfo to only decode numerical network addresses and suppress
// hostname lookups.
ai_hint.ai_flags = allow_lookup ? AI_ADDRCONFIG : AI_NUMERICHOST;
addrinfo* ai_res{nullptr};
const int n_err{getaddrinfo(name.c_str(), nullptr, &ai_hint, &ai_res)};
if (n_err != 0) {
return {};
}
// Traverse the linked list starting with ai_trav.
addrinfo* ai_trav{ai_res};
std::vector<CNetAddr> resolved_addresses;
while (ai_trav != nullptr) {
if (ai_trav->ai_family == AF_INET) {
assert(ai_trav->ai_addrlen >= sizeof(sockaddr_in));
resolved_addresses.emplace_back(reinterpret_cast<sockaddr_in*>(ai_trav->ai_addr)->sin_addr);
}
if (ai_trav->ai_family == AF_INET6) {
assert(ai_trav->ai_addrlen >= sizeof(sockaddr_in6));
const sockaddr_in6* s6{reinterpret_cast<sockaddr_in6*>(ai_trav->ai_addr)};
resolved_addresses.emplace_back(s6->sin6_addr, s6->sin6_scope_id);
}
ai_trav = ai_trav->ai_next;
}
freeaddrinfo(ai_res);
return resolved_addresses;
}
DNSLookupFn g_dns_lookup{WrappedGetAddrInfo};
enum Network ParseNetwork(const std::string& net_in) {
std::string net = ToLower(net_in);
if (net == "ipv4") return NET_IPV4;
if (net == "ipv6") return NET_IPV6;
if (net == "onion") return NET_ONION;
if (net == "tor") {
LogPrintf("Warning: net name 'tor' is deprecated and will be removed in the future. You should use 'onion' instead.\n");
return NET_ONION;
}
if (net == "i2p") {
return NET_I2P;
}
if (net == "cjdns") {
return NET_CJDNS;
}
return NET_UNROUTABLE;
}
std::string GetNetworkName(enum Network net)
{
switch (net) {
case NET_UNROUTABLE: return "not_publicly_routable";
case NET_IPV4: return "ipv4";
case NET_IPV6: return "ipv6";
case NET_ONION: return "onion";
case NET_I2P: return "i2p";
case NET_CJDNS: return "cjdns";
case NET_INTERNAL: return "internal";
case NET_MAX: assert(false);
} // no default case, so the compiler can warn about missing cases
assert(false);
}
std::vector<std::string> GetNetworkNames(bool append_unroutable)
{
std::vector<std::string> names;
for (int n = 0; n < NET_MAX; ++n) {
const enum Network network{static_cast<Network>(n)};
if (network == NET_UNROUTABLE || network == NET_INTERNAL) continue;
names.emplace_back(GetNetworkName(network));
}
if (append_unroutable) {
names.emplace_back(GetNetworkName(NET_UNROUTABLE));
}
return names;
}
static std::vector<CNetAddr> LookupIntern(const std::string& name, unsigned int nMaxSolutions, bool fAllowLookup, DNSLookupFn dns_lookup_function)
{
if (!ContainsNoNUL(name)) return {};
{
CNetAddr addr;
// From our perspective, onion addresses are not hostnames but rather
// direct encodings of CNetAddr much like IPv4 dotted-decimal notation
// or IPv6 colon-separated hextet notation. Since we can't use
// getaddrinfo to decode them and it wouldn't make sense to resolve
// them, we return a network address representing it instead. See
// CNetAddr::SetSpecial(const std::string&) for more details.
if (addr.SetSpecial(name)) return {addr};
}
std::vector<CNetAddr> addresses;
for (const CNetAddr& resolved : dns_lookup_function(name, fAllowLookup)) {
if (nMaxSolutions > 0 && addresses.size() >= nMaxSolutions) {
break;
}
/* Never allow resolving to an internal address. Consider any such result invalid */
if (!resolved.IsInternal()) {
addresses.push_back(resolved);
}
}
return addresses;
}
std::vector<CNetAddr> LookupHost(const std::string& name, unsigned int nMaxSolutions, bool fAllowLookup, DNSLookupFn dns_lookup_function)
{
if (!ContainsNoNUL(name)) return {};
std::string strHost = name;
if (strHost.empty()) return {};
if (strHost.front() == '[' && strHost.back() == ']') {
strHost = strHost.substr(1, strHost.size() - 2);
}
return LookupIntern(strHost, nMaxSolutions, fAllowLookup, dns_lookup_function);
}
std::optional<CNetAddr> LookupHost(const std::string& name, bool fAllowLookup, DNSLookupFn dns_lookup_function)
{
const std::vector<CNetAddr> addresses{LookupHost(name, 1, fAllowLookup, dns_lookup_function)};
return addresses.empty() ? std::nullopt : std::make_optional(addresses.front());
}
std::vector<CService> Lookup(const std::string& name, uint16_t portDefault, bool fAllowLookup, unsigned int nMaxSolutions, DNSLookupFn dns_lookup_function)
{
if (name.empty() || !ContainsNoNUL(name)) {
return {};
}
uint16_t port{portDefault};
std::string hostname;
SplitHostPort(name, port, hostname);
const std::vector<CNetAddr> addresses{LookupIntern(hostname, nMaxSolutions, fAllowLookup, dns_lookup_function)};
if (addresses.empty()) return {};
std::vector<CService> services;
services.reserve(addresses.size());
for (const auto& addr : addresses)
services.emplace_back(addr, port);
return services;
}
std::optional<CService> Lookup(const std::string& name, uint16_t portDefault, bool fAllowLookup, DNSLookupFn dns_lookup_function)
{
const std::vector<CService> services{Lookup(name, portDefault, fAllowLookup, 1, dns_lookup_function)};
return services.empty() ? std::nullopt : std::make_optional(services.front());
}
CService LookupNumeric(const std::string& name, uint16_t portDefault, DNSLookupFn dns_lookup_function)
{
if (!ContainsNoNUL(name)) {
return {};
}
// "1.2:345" will fail to resolve the ip, but will still set the port.
// If the ip fails to resolve, re-init the result.
return Lookup(name, portDefault, /*fAllowLookup=*/false, dns_lookup_function).value_or(CService{});
}
bool IsUnixSocketPath(const std::string& name)
{
#if HAVE_SOCKADDR_UN
if (name.find(ADDR_PREFIX_UNIX) != 0) return false;
// Split off "unix:" prefix
std::string str{name.substr(ADDR_PREFIX_UNIX.length())};
// Path size limit is platform-dependent
// see https://manpages.ubuntu.com/manpages/xenial/en/man7/unix.7.html
if (str.size() + 1 > sizeof(((sockaddr_un*)nullptr)->sun_path)) return false;
return true;
#else
return false;
#endif
}
/** SOCKS version */
enum SOCKSVersion: uint8_t {
SOCKS4 = 0x04,
SOCKS5 = 0x05
};
/** Values defined for METHOD in RFC1928 */
enum SOCKS5Method: uint8_t {
NOAUTH = 0x00, //!< No authentication required
GSSAPI = 0x01, //!< GSSAPI
USER_PASS = 0x02, //!< Username/password
NO_ACCEPTABLE = 0xff, //!< No acceptable methods
};
/** Values defined for CMD in RFC1928 */
enum SOCKS5Command: uint8_t {
CONNECT = 0x01,
BIND = 0x02,
UDP_ASSOCIATE = 0x03
};
/** Values defined for REP in RFC1928 */
enum SOCKS5Reply: uint8_t {
SUCCEEDED = 0x00, //!< Succeeded
GENFAILURE = 0x01, //!< General failure
NOTALLOWED = 0x02, //!< Connection not allowed by ruleset
NETUNREACHABLE = 0x03, //!< Network unreachable
HOSTUNREACHABLE = 0x04, //!< Network unreachable
CONNREFUSED = 0x05, //!< Connection refused
TTLEXPIRED = 0x06, //!< TTL expired
CMDUNSUPPORTED = 0x07, //!< Command not supported
ATYPEUNSUPPORTED = 0x08, //!< Address type not supported
};
/** Values defined for ATYPE in RFC1928 */
enum SOCKS5Atyp: uint8_t {
IPV4 = 0x01,
DOMAINNAME = 0x03,
IPV6 = 0x04,
};
/** Status codes that can be returned by InterruptibleRecv */
enum class IntrRecvError {
OK,
Timeout,
Disconnected,
NetworkError,
Interrupted
};
/**
* Try to read a specified number of bytes from a socket. Please read the "see
* also" section for more detail.
*
* @param data The buffer where the read bytes should be stored.
* @param len The number of bytes to read into the specified buffer.
* @param timeout The total timeout for this read.
* @param sock The socket (has to be in non-blocking mode) from which to read bytes.
*
* @returns An IntrRecvError indicating the resulting status of this read.
* IntrRecvError::OK only if all of the specified number of bytes were
* read.
*
* @see This function can be interrupted by calling g_socks5_interrupt().
* Sockets can be made non-blocking with Sock::SetNonBlocking().
*/
static IntrRecvError InterruptibleRecv(uint8_t* data, size_t len, std::chrono::milliseconds timeout, const Sock& sock)
{
auto curTime{Now<SteadyMilliseconds>()};
const auto endTime{curTime + timeout};
while (len > 0 && curTime < endTime) {
ssize_t ret = sock.Recv(data, len, 0); // Optimistically try the recv first
if (ret > 0) {
len -= ret;
data += ret;
} else if (ret == 0) { // Unexpected disconnection
return IntrRecvError::Disconnected;
} else { // Other error or blocking
int nErr = WSAGetLastError();
if (nErr == WSAEINPROGRESS || nErr == WSAEWOULDBLOCK || nErr == WSAEINVAL) {
// Only wait at most MAX_WAIT_FOR_IO at a time, unless
// we're approaching the end of the specified total timeout
const auto remaining = std::chrono::milliseconds{endTime - curTime};
const auto timeout = std::min(remaining, std::chrono::milliseconds{MAX_WAIT_FOR_IO});
if (!sock.Wait(timeout, Sock::RECV)) {
return IntrRecvError::NetworkError;
}
} else {
return IntrRecvError::NetworkError;
}
}
if (g_socks5_interrupt) {
return IntrRecvError::Interrupted;
}
curTime = Now<SteadyMilliseconds>();
}
return len == 0 ? IntrRecvError::OK : IntrRecvError::Timeout;
}
/** Convert SOCKS5 reply to an error message */
static std::string Socks5ErrorString(uint8_t err)
{
switch(err) {
case SOCKS5Reply::GENFAILURE:
return "general failure";
case SOCKS5Reply::NOTALLOWED:
return "connection not allowed";
case SOCKS5Reply::NETUNREACHABLE:
return "network unreachable";
case SOCKS5Reply::HOSTUNREACHABLE:
return "host unreachable";
case SOCKS5Reply::CONNREFUSED:
return "connection refused";
case SOCKS5Reply::TTLEXPIRED:
return "TTL expired";
case SOCKS5Reply::CMDUNSUPPORTED:
return "protocol error";
case SOCKS5Reply::ATYPEUNSUPPORTED:
return "address type not supported";
default:
return "unknown";
}
}
bool Socks5(const std::string& strDest, uint16_t port, const ProxyCredentials* auth, const Sock& sock)
{
try {
IntrRecvError recvr;
LogPrint(BCLog::NET, "SOCKS5 connecting %s\n", strDest);
if (strDest.size() > 255) {
LogError("Hostname too long\n");
return false;
}
// Construct the version identifier/method selection message
std::vector<uint8_t> vSocks5Init;
vSocks5Init.push_back(SOCKSVersion::SOCKS5); // We want the SOCK5 protocol
if (auth) {
vSocks5Init.push_back(0x02); // 2 method identifiers follow...
vSocks5Init.push_back(SOCKS5Method::NOAUTH);
vSocks5Init.push_back(SOCKS5Method::USER_PASS);
} else {
vSocks5Init.push_back(0x01); // 1 method identifier follows...
vSocks5Init.push_back(SOCKS5Method::NOAUTH);
}
sock.SendComplete(vSocks5Init, g_socks5_recv_timeout, g_socks5_interrupt);
uint8_t pchRet1[2];
if (InterruptibleRecv(pchRet1, 2, g_socks5_recv_timeout, sock) != IntrRecvError::OK) {
LogPrintf("Socks5() connect to %s:%d failed: InterruptibleRecv() timeout or other failure\n", strDest, port);
return false;
}
if (pchRet1[0] != SOCKSVersion::SOCKS5) {
LogError("Proxy failed to initialize\n");
return false;
}
if (pchRet1[1] == SOCKS5Method::USER_PASS && auth) {
// Perform username/password authentication (as described in RFC1929)
std::vector<uint8_t> vAuth;
vAuth.push_back(0x01); // Current (and only) version of user/pass subnegotiation
if (auth->username.size() > 255 || auth->password.size() > 255) {
LogError("Proxy username or password too long\n");
return false;
}
vAuth.push_back(auth->username.size());
vAuth.insert(vAuth.end(), auth->username.begin(), auth->username.end());
vAuth.push_back(auth->password.size());
vAuth.insert(vAuth.end(), auth->password.begin(), auth->password.end());
sock.SendComplete(vAuth, g_socks5_recv_timeout, g_socks5_interrupt);
LogPrint(BCLog::PROXY, "SOCKS5 sending proxy authentication %s:%s\n", auth->username, auth->password);
uint8_t pchRetA[2];
if (InterruptibleRecv(pchRetA, 2, g_socks5_recv_timeout, sock) != IntrRecvError::OK) {
LogError("Error reading proxy authentication response\n");
return false;
}
if (pchRetA[0] != 0x01 || pchRetA[1] != 0x00) {
LogError("Proxy authentication unsuccessful\n");
return false;
}
} else if (pchRet1[1] == SOCKS5Method::NOAUTH) {
// Perform no authentication
} else {
LogError("Proxy requested wrong authentication method %02x\n", pchRet1[1]);
return false;
}
std::vector<uint8_t> vSocks5;
vSocks5.push_back(SOCKSVersion::SOCKS5); // VER protocol version
vSocks5.push_back(SOCKS5Command::CONNECT); // CMD CONNECT
vSocks5.push_back(0x00); // RSV Reserved must be 0
vSocks5.push_back(SOCKS5Atyp::DOMAINNAME); // ATYP DOMAINNAME
vSocks5.push_back(strDest.size()); // Length<=255 is checked at beginning of function
vSocks5.insert(vSocks5.end(), strDest.begin(), strDest.end());
vSocks5.push_back((port >> 8) & 0xFF);
vSocks5.push_back((port >> 0) & 0xFF);
sock.SendComplete(vSocks5, g_socks5_recv_timeout, g_socks5_interrupt);
uint8_t pchRet2[4];
if ((recvr = InterruptibleRecv(pchRet2, 4, g_socks5_recv_timeout, sock)) != IntrRecvError::OK) {
if (recvr == IntrRecvError::Timeout) {
/* If a timeout happens here, this effectively means we timed out while connecting
* to the remote node. This is very common for Tor, so do not print an
* error message. */
return false;
} else {
LogError("Error while reading proxy response\n");
return false;
}
}
if (pchRet2[0] != SOCKSVersion::SOCKS5) {
LogError("Proxy failed to accept request\n");
return false;
}
if (pchRet2[1] != SOCKS5Reply::SUCCEEDED) {
// Failures to connect to a peer that are not proxy errors
LogPrintf("Socks5() connect to %s:%d failed: %s\n", strDest, port, Socks5ErrorString(pchRet2[1]));
return false;
}
if (pchRet2[2] != 0x00) { // Reserved field must be 0
LogError("Error: malformed proxy response\n");
return false;
}
uint8_t pchRet3[256];
switch (pchRet2[3]) {
case SOCKS5Atyp::IPV4: recvr = InterruptibleRecv(pchRet3, 4, g_socks5_recv_timeout, sock); break;
case SOCKS5Atyp::IPV6: recvr = InterruptibleRecv(pchRet3, 16, g_socks5_recv_timeout, sock); break;
case SOCKS5Atyp::DOMAINNAME: {
recvr = InterruptibleRecv(pchRet3, 1, g_socks5_recv_timeout, sock);
if (recvr != IntrRecvError::OK) {
LogError("Error reading from proxy\n");
return false;
}
int nRecv = pchRet3[0];
recvr = InterruptibleRecv(pchRet3, nRecv, g_socks5_recv_timeout, sock);
break;
}
default: {
LogError("Error: malformed proxy response\n");
return false;
}
}
if (recvr != IntrRecvError::OK) {
LogError("Error reading from proxy\n");
return false;
}
if (InterruptibleRecv(pchRet3, 2, g_socks5_recv_timeout, sock) != IntrRecvError::OK) {
LogError("Error reading from proxy\n");
return false;
}
LogPrint(BCLog::NET, "SOCKS5 connected %s\n", strDest);
return true;
} catch (const std::runtime_error& e) {
LogError("Error during SOCKS5 proxy handshake: %s\n", e.what());
return false;
}
}
std::unique_ptr<Sock> CreateSockOS(sa_family_t address_family)
{
// Not IPv4, IPv6 or UNIX
if (address_family == AF_UNSPEC) return nullptr;
int protocol{IPPROTO_TCP};
#if HAVE_SOCKADDR_UN
if (address_family == AF_UNIX) protocol = 0;
#endif
// Create a socket in the specified address family.
SOCKET hSocket = socket(address_family, SOCK_STREAM, protocol);
if (hSocket == INVALID_SOCKET) {
return nullptr;
}
auto sock = std::make_unique<Sock>(hSocket);
// Ensure that waiting for I/O on this socket won't result in undefined
// behavior.
if (!sock->IsSelectable()) {
LogPrintf("Cannot create connection: non-selectable socket created (fd >= FD_SETSIZE ?)\n");
return nullptr;
}
#ifdef SO_NOSIGPIPE
int set = 1;
// Set the no-sigpipe option on the socket for BSD systems, other UNIXes
// should use the MSG_NOSIGNAL flag for every send.
if (sock->SetSockOpt(SOL_SOCKET, SO_NOSIGPIPE, (void*)&set, sizeof(int)) == SOCKET_ERROR) {
LogPrintf("Error setting SO_NOSIGPIPE on socket: %s, continuing anyway\n",
NetworkErrorString(WSAGetLastError()));
}
#endif
// Set the non-blocking option on the socket.
if (!sock->SetNonBlocking()) {
LogPrintf("Error setting socket to non-blocking: %s\n", NetworkErrorString(WSAGetLastError()));
return nullptr;
}
#if HAVE_SOCKADDR_UN
if (address_family == AF_UNIX) return sock;
#endif
// Set the no-delay option (disable Nagle's algorithm) on the TCP socket.
const int on{1};
if (sock->SetSockOpt(IPPROTO_TCP, TCP_NODELAY, &on, sizeof(on)) == SOCKET_ERROR) {
LogPrint(BCLog::NET, "Unable to set TCP_NODELAY on a newly created socket, continuing anyway\n");
}
return sock;
}
std::function<std::unique_ptr<Sock>(const sa_family_t&)> CreateSock = CreateSockOS;
template<typename... Args>
static void LogConnectFailure(bool manual_connection, const char* fmt, const Args&... args) {
std::string error_message = tfm::format(fmt, args...);
if (manual_connection) {
LogPrintf("%s\n", error_message);
} else {
LogPrint(BCLog::NET, "%s\n", error_message);
}
}
static bool ConnectToSocket(const Sock& sock, struct sockaddr* sockaddr, socklen_t len, const std::string& dest_str, bool manual_connection)
{
// Connect to `sockaddr` using `sock`.
if (sock.Connect(sockaddr, len) == SOCKET_ERROR) {
int nErr = WSAGetLastError();
// WSAEINVAL is here because some legacy version of winsock uses it
if (nErr == WSAEINPROGRESS || nErr == WSAEWOULDBLOCK || nErr == WSAEINVAL)
{
// Connection didn't actually fail, but is being established
// asynchronously. Thus, use async I/O api (select/poll)
// synchronously to check for successful connection with a timeout.
const Sock::Event requested = Sock::RECV | Sock::SEND;
Sock::Event occurred;
if (!sock.Wait(std::chrono::milliseconds{nConnectTimeout}, requested, &occurred)) {
LogPrintf("wait for connect to %s failed: %s\n",
dest_str,
NetworkErrorString(WSAGetLastError()));
return false;
} else if (occurred == 0) {
LogPrint(BCLog::NET, "connection attempt to %s timed out\n", dest_str);
return false;
}
// Even if the wait was successful, the connect might not
// have been successful. The reason for this failure is hidden away
// in the SO_ERROR for the socket in modern systems. We read it into
// sockerr here.
int sockerr;
socklen_t sockerr_len = sizeof(sockerr);
if (sock.GetSockOpt(SOL_SOCKET, SO_ERROR, (sockopt_arg_type)&sockerr, &sockerr_len) ==
SOCKET_ERROR) {
LogPrintf("getsockopt() for %s failed: %s\n", dest_str, NetworkErrorString(WSAGetLastError()));
return false;
}
if (sockerr != 0) {
LogConnectFailure(manual_connection,
"connect() to %s failed after wait: %s",
dest_str,
NetworkErrorString(sockerr));
return false;
}
}
#ifdef WIN32
else if (WSAGetLastError() != WSAEISCONN)
#else
else
#endif
{
LogConnectFailure(manual_connection, "connect() to %s failed: %s", dest_str, NetworkErrorString(WSAGetLastError()));
return false;
}
}
return true;
}
std::unique_ptr<Sock> ConnectDirectly(const CService& dest, bool manual_connection)
{
auto sock = CreateSock(dest.GetSAFamily());
if (!sock) {
LogPrintLevel(BCLog::NET, BCLog::Level::Error, "Cannot create a socket for connecting to %s\n", dest.ToStringAddrPort());
return {};
}
// Create a sockaddr from the specified service.
struct sockaddr_storage sockaddr;
socklen_t len = sizeof(sockaddr);
if (!dest.GetSockAddr((struct sockaddr*)&sockaddr, &len)) {
LogPrintf("Cannot get sockaddr for %s: unsupported network\n", dest.ToStringAddrPort());
return {};
}
if (!ConnectToSocket(*sock, (struct sockaddr*)&sockaddr, len, dest.ToStringAddrPort(), manual_connection)) {
return {};
}
return sock;
}
std::unique_ptr<Sock> Proxy::Connect() const
{
if (!IsValid()) return {};
if (!m_is_unix_socket) return ConnectDirectly(proxy, /*manual_connection=*/true);
#if HAVE_SOCKADDR_UN
auto sock = CreateSock(AF_UNIX);
if (!sock) {
LogPrintLevel(BCLog::NET, BCLog::Level::Error, "Cannot create a socket for connecting to %s\n", m_unix_socket_path);
return {};
}
const std::string path{m_unix_socket_path.substr(ADDR_PREFIX_UNIX.length())};
struct sockaddr_un addrun;
memset(&addrun, 0, sizeof(addrun));
addrun.sun_family = AF_UNIX;
// leave the last char in addrun.sun_path[] to be always '\0'
memcpy(addrun.sun_path, path.c_str(), std::min(sizeof(addrun.sun_path) - 1, path.length()));
socklen_t len = sizeof(addrun);
if(!ConnectToSocket(*sock, (struct sockaddr*)&addrun, len, path, /*manual_connection=*/true)) {
return {};
}
return sock;
#else
return {};
#endif
}
bool SetProxy(enum Network net, const Proxy &addrProxy) {
assert(net >= 0 && net < NET_MAX);
if (!addrProxy.IsValid())
return false;
LOCK(g_proxyinfo_mutex);
proxyInfo[net] = addrProxy;
return true;
}
bool GetProxy(enum Network net, Proxy &proxyInfoOut) {
assert(net >= 0 && net < NET_MAX);
LOCK(g_proxyinfo_mutex);
if (!proxyInfo[net].IsValid())
return false;
proxyInfoOut = proxyInfo[net];
return true;
}
bool SetNameProxy(const Proxy &addrProxy) {
if (!addrProxy.IsValid())
return false;
LOCK(g_proxyinfo_mutex);
nameProxy = addrProxy;
return true;
}
bool GetNameProxy(Proxy &nameProxyOut) {
LOCK(g_proxyinfo_mutex);
if(!nameProxy.IsValid())
return false;
nameProxyOut = nameProxy;
return true;
}
bool HaveNameProxy() {
LOCK(g_proxyinfo_mutex);
return nameProxy.IsValid();
}
bool IsProxy(const CNetAddr &addr) {
LOCK(g_proxyinfo_mutex);
for (int i = 0; i < NET_MAX; i++) {
if (addr == static_cast<CNetAddr>(proxyInfo[i].proxy))
return true;
}
return false;
}
std::unique_ptr<Sock> ConnectThroughProxy(const Proxy& proxy,
const std::string& dest,
uint16_t port,
bool& proxy_connection_failed)
{
// first connect to proxy server
auto sock = proxy.Connect();
if (!sock) {
proxy_connection_failed = true;
return {};
}
// do socks negotiation
if (proxy.m_randomize_credentials) {
ProxyCredentials random_auth;
static std::atomic_int counter(0);
random_auth.username = random_auth.password = strprintf("%i", counter++);
if (!Socks5(dest, port, &random_auth, *sock)) {
return {};
}
} else {
if (!Socks5(dest, port, nullptr, *sock)) {
return {};
}
}
return sock;
}
CSubNet LookupSubNet(const std::string& subnet_str)
{
CSubNet subnet;
assert(!subnet.IsValid());
if (!ContainsNoNUL(subnet_str)) {
return subnet;
}
const size_t slash_pos{subnet_str.find_last_of('/')};
const std::string str_addr{subnet_str.substr(0, slash_pos)};
std::optional<CNetAddr> addr{LookupHost(str_addr, /*fAllowLookup=*/false)};
if (addr.has_value()) {
addr = static_cast<CNetAddr>(MaybeFlipIPv6toCJDNS(CService{addr.value(), /*port=*/0}));
if (slash_pos != subnet_str.npos) {
const std::string netmask_str{subnet_str.substr(slash_pos + 1)};
uint8_t netmask;
if (ParseUInt8(netmask_str, &netmask)) {
// Valid number; assume CIDR variable-length subnet masking.
subnet = CSubNet{addr.value(), netmask};
} else {
// Invalid number; try full netmask syntax. Never allow lookup for netmask.
const std::optional<CNetAddr> full_netmask{LookupHost(netmask_str, /*fAllowLookup=*/false)};
if (full_netmask.has_value()) {
subnet = CSubNet{addr.value(), full_netmask.value()};
}
}
} else {
// Single IP subnet (<ipv4>/32 or <ipv6>/128).
subnet = CSubNet{addr.value()};
}
}
return subnet;
}
bool IsBadPort(uint16_t port)
{
/* Don't forget to update doc/p2p-bad-ports.md if you change this list. */
switch (port) {
case 1: // tcpmux
case 7: // echo
case 9: // discard
case 11: // systat
case 13: // daytime
case 15: // netstat
case 17: // qotd
case 19: // chargen
case 20: // ftp data
case 21: // ftp access
case 22: // ssh
case 23: // telnet
case 25: // smtp
case 37: // time
case 42: // name
case 43: // nicname
case 53: // domain
case 69: // tftp
case 77: // priv-rjs
case 79: // finger
case 87: // ttylink
case 95: // supdup
case 101: // hostname
case 102: // iso-tsap
case 103: // gppitnp
case 104: // acr-nema
case 109: // pop2
case 110: // pop3
case 111: // sunrpc
case 113: // auth
case 115: // sftp
case 117: // uucp-path
case 119: // nntp
case 123: // NTP
case 135: // loc-srv /epmap
case 137: // netbios
case 139: // netbios
case 143: // imap2
case 161: // snmp
case 179: // BGP
case 389: // ldap
case 427: // SLP (Also used by Apple Filing Protocol)
case 465: // smtp+ssl
case 512: // print / exec
case 513: // login
case 514: // shell
case 515: // printer
case 526: // tempo
case 530: // courier
case 531: // chat
case 532: // netnews
case 540: // uucp
case 548: // AFP (Apple Filing Protocol)
case 554: // rtsp
case 556: // remotefs
case 563: // nntp+ssl
case 587: // smtp (rfc6409)
case 601: // syslog-conn (rfc3195)
case 636: // ldap+ssl
case 989: // ftps-data
case 990: // ftps
case 993: // ldap+ssl
case 995: // pop3+ssl
case 1719: // h323gatestat
case 1720: // h323hostcall
case 1723: // pptp
case 2049: // nfs
case 3659: // apple-sasl / PasswordServer
case 4045: // lockd
case 5060: // sip
case 5061: // sips
case 6000: // X11
case 6566: // sane-port
case 6665: // Alternate IRC
case 6666: // Alternate IRC
case 6667: // Standard IRC
case 6668: // Alternate IRC
case 6669: // Alternate IRC
case 6697: // IRC + TLS
case 10080: // Amanda
return true;
}
return false;
}
CService MaybeFlipIPv6toCJDNS(const CService& service)
{
CService ret{service};
if (ret.IsIPv6() && ret.HasCJDNSPrefix() && g_reachable_nets.Contains(NET_CJDNS)) {
ret.m_net = NET_CJDNS;
}
return ret;
}