gecko/netwerk/socket/nsSOCKSIOLayer.cpp
2011-03-24 21:37:53 +13:00

942 lines
30 KiB
C++

/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*-
*
* ***** BEGIN LICENSE BLOCK *****
* Version: MPL 1.1/GPL 2.0/LGPL 2.1
*
* The contents of this file are subject to the Mozilla Public License Version
* 1.1 (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
* http://www.mozilla.org/MPL/
*
* Software distributed under the License is distributed on an "AS IS" basis,
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
* for the specific language governing rights and limitations under the
* License.
*
* The Original Code is mozilla.org code.
*
* The Initial Developer of the Original Code is
* Netscape Communications Corporation.
* Portions created by the Initial Developer are Copyright (C) 1998
* the Initial Developer. All Rights Reserved.
*
* Contributor(s):
* Justin Bradford <jab@atdot.org>
* Bradley Baetz <bbaetz@acm.org>
* Darin Fisher <darin@meer.net>
* Malcolm Smith <malsmith@cs.rmit.edu.au>
*
* Alternatively, the contents of this file may be used under the terms of
* either the GNU General Public License Version 2 or later (the "GPL"), or
* the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
* in which case the provisions of the GPL or the LGPL are applicable instead
* of those above. If you wish to allow use of your version of this file only
* under the terms of either the GPL or the LGPL, and not to allow others to
* use your version of this file under the terms of the MPL, indicate your
* decision by deleting the provisions above and replace them with the notice
* and other provisions required by the GPL or the LGPL. If you do not delete
* the provisions above, a recipient may use your version of this file under
* the terms of any one of the MPL, the GPL or the LGPL.
*
* ***** END LICENSE BLOCK ***** */
#include "nspr.h"
#include "nsString.h"
#include "nsCRT.h"
#include "nsIServiceManager.h"
#include "nsIDNSService.h"
#include "nsIDNSRecord.h"
#include "nsISOCKSSocketInfo.h"
#include "nsISocketProvider.h"
#include "nsSOCKSIOLayer.h"
#include "nsNetCID.h"
static PRDescIdentity nsSOCKSIOLayerIdentity;
static PRIOMethods nsSOCKSIOLayerMethods;
static PRBool firstTime = PR_TRUE;
#if defined(PR_LOGGING)
static PRLogModuleInfo *gSOCKSLog;
#define LOGDEBUG(args) PR_LOG(gSOCKSLog, PR_LOG_DEBUG, args)
#define LOGERROR(args) PR_LOG(gSOCKSLog, PR_LOG_ERROR , args)
#else
#define LOGDEBUG(args)
#define LOGERROR(args)
#endif
class nsSOCKSSocketInfo : public nsISOCKSSocketInfo
{
public:
nsSOCKSSocketInfo();
virtual ~nsSOCKSSocketInfo() {}
NS_DECL_ISUPPORTS
NS_DECL_NSISOCKSSOCKETINFO
void Init(PRInt32 version,
const char *proxyHost,
PRInt32 proxyPort,
const char *destinationHost,
PRUint32 flags);
const nsCString &DestinationHost() { return mDestinationHost; }
const nsCString &ProxyHost() { return mProxyHost; }
PRInt32 ProxyPort() { return mProxyPort; }
PRInt32 Version() { return mVersion; }
PRUint32 Flags() { return mFlags; }
private:
nsCString mDestinationHost;
nsCString mProxyHost;
PRInt32 mProxyPort;
PRInt32 mVersion; // SOCKS version 4 or 5
PRUint32 mFlags;
PRNetAddr mInternalProxyAddr;
PRNetAddr mExternalProxyAddr;
PRNetAddr mDestinationAddr;
};
nsSOCKSSocketInfo::nsSOCKSSocketInfo()
: mProxyPort(-1)
, mVersion(-1)
, mFlags(0)
{
PR_InitializeNetAddr(PR_IpAddrAny, 0, &mInternalProxyAddr);
PR_InitializeNetAddr(PR_IpAddrAny, 0, &mExternalProxyAddr);
PR_InitializeNetAddr(PR_IpAddrAny, 0, &mDestinationAddr);
}
void
nsSOCKSSocketInfo::Init(PRInt32 version, const char *proxyHost, PRInt32 proxyPort, const char *host, PRUint32 flags)
{
mVersion = version;
mProxyHost = proxyHost;
mProxyPort = proxyPort;
mDestinationHost = host;
mFlags = flags;
}
NS_IMPL_THREADSAFE_ISUPPORTS1(nsSOCKSSocketInfo, nsISOCKSSocketInfo)
NS_IMETHODIMP
nsSOCKSSocketInfo::GetExternalProxyAddr(PRNetAddr * *aExternalProxyAddr)
{
memcpy(*aExternalProxyAddr, &mExternalProxyAddr, sizeof(PRNetAddr));
return NS_OK;
}
NS_IMETHODIMP
nsSOCKSSocketInfo::SetExternalProxyAddr(PRNetAddr *aExternalProxyAddr)
{
memcpy(&mExternalProxyAddr, aExternalProxyAddr, sizeof(PRNetAddr));
return NS_OK;
}
NS_IMETHODIMP
nsSOCKSSocketInfo::GetDestinationAddr(PRNetAddr * *aDestinationAddr)
{
memcpy(*aDestinationAddr, &mDestinationAddr, sizeof(PRNetAddr));
return NS_OK;
}
NS_IMETHODIMP
nsSOCKSSocketInfo::SetDestinationAddr(PRNetAddr *aDestinationAddr)
{
memcpy(&mDestinationAddr, aDestinationAddr, sizeof(PRNetAddr));
return NS_OK;
}
NS_IMETHODIMP
nsSOCKSSocketInfo::GetInternalProxyAddr(PRNetAddr * *aInternalProxyAddr)
{
memcpy(*aInternalProxyAddr, &mInternalProxyAddr, sizeof(PRNetAddr));
return NS_OK;
}
NS_IMETHODIMP
nsSOCKSSocketInfo::SetInternalProxyAddr(PRNetAddr *aInternalProxyAddr)
{
memcpy(&mInternalProxyAddr, aInternalProxyAddr, sizeof(PRNetAddr));
return NS_OK;
}
static PRInt32
pr_RecvAll(PRFileDesc *fd, unsigned char *buf, PRInt32 amount, PRIntn flags,
PRIntervalTime *timeout)
{
PRInt32 bytesRead = 0;
PRInt32 offset = 0;
while (offset < amount) {
PRIntervalTime start_time = PR_IntervalNow();
bytesRead = PR_Recv(fd, buf + offset, amount - offset, flags, *timeout);
PRIntervalTime elapsed = PR_IntervalNow() - start_time;
if (elapsed > *timeout) {
*timeout = 0;
} else {
*timeout -= elapsed;
}
if (bytesRead > 0) {
offset += bytesRead;
} else if (bytesRead == 0 || offset != 0) {
return offset;
} else {
return bytesRead;
}
if (*timeout == 0) {
LOGERROR(("PR_Recv() timed out. amount = %d. offset = %d.",
amount, offset));
return offset;
}
}
return offset;
}
static PRInt32
pr_Send(PRFileDesc *fd, const void *buf, PRInt32 amount, PRIntn flags,
PRIntervalTime *timeout)
{
PRIntervalTime start_time = PR_IntervalNow();
PRInt32 retval = PR_Send(fd, buf, amount, flags, *timeout);
PRIntervalTime elapsed = PR_IntervalNow() - start_time;
if (elapsed > *timeout) {
*timeout = 0;
LOGERROR(("PR_Send() timed out. amount = %d. retval = %d.",
amount, retval));
return retval;
} else {
*timeout -= elapsed;
}
if (retval <= 0) {
LOGERROR(("PR_Send() failed. amount = %d. retval = %d.",
amount, retval));
}
return retval;
}
// Negotiate a SOCKS 5 connection. Assumes the TCP connection to the socks
// server port has been established.
static nsresult
ConnectSOCKS5(PRFileDesc *fd, const PRNetAddr *addr, PRNetAddr *extAddr, PRIntervalTime timeout)
{
int request_len = 0;
int response_len = 0;
int desired_len = 0;
unsigned char request[22];
unsigned char response[262];
NS_ENSURE_TRUE(fd, NS_ERROR_NOT_INITIALIZED);
NS_ENSURE_TRUE(addr, NS_ERROR_NOT_INITIALIZED);
NS_ENSURE_TRUE(extAddr, NS_ERROR_NOT_INITIALIZED);
request[0] = 0x05; // SOCKS version 5
request[1] = 0x01; // number of auth procotols we recognize
// auth protocols
request[2] = 0x00; // no authentication required
// compliant implementations MUST implement GSSAPI
// and SHOULD implement username/password and MAY
// implement CHAP
// TODO: we don't implement these
//request[3] = 0x01; // GSSAPI
//request[4] = 0x02; // username/password
//request[5] = 0x03; // CHAP
request_len = 2 + request[1];
int write_len = pr_Send(fd, request, request_len, 0, &timeout);
if (write_len != request_len) {
return NS_ERROR_FAILURE;
}
// get the server's response.
desired_len = 2;
response_len = pr_RecvAll(fd, response, desired_len, 0, &timeout);
if (response_len < desired_len) {
LOGERROR(("pr_RecvAll() failed. response_len = %d.", response_len));
return NS_ERROR_FAILURE;
}
if (response[0] != 0x05) {
// it's a either not SOCKS or not our version
LOGERROR(("Not a SOCKS 5 reply. Expected: 5; received: %x", response[0]));
return NS_ERROR_FAILURE;
}
switch (response[1]) {
case 0x00:
// no auth
break;
case 0x01:
// GSSAPI
// TODO: implement
LOGERROR(("Server want to use GSSAPI to authenticate, but we don't support it."));
return NS_ERROR_FAILURE;
case 0x02:
// username/password
// TODO: implement
LOGERROR(("Server want to use username/password to authenticate, but we don't support it."));
return NS_ERROR_FAILURE;
case 0x03:
// CHAP
// TODO: implement?
LOGERROR(("Server want to use CHAP to authenticate, but we don't support it."));
return NS_ERROR_FAILURE;
default:
// unrecognized auth method
LOGERROR(("Uncrecognized authentication method received: %x", response[1]));
return NS_ERROR_FAILURE;
}
// we are now authenticated, so lets tell
// the server where to connect to
request_len = 0;
request[0] = 0x05; // SOCKS version 5
request[1] = 0x01; // CONNECT command
request[2] = 0x00; // obligatory reserved field (perfect for MS tampering!)
// get destination port
PRInt32 destPort = PR_ntohs(PR_NetAddrInetPort(addr));
nsSOCKSSocketInfo * info = (nsSOCKSSocketInfo*) fd->secret;
if (info->Flags() & nsISocketProvider::PROXY_RESOLVES_HOST) {
LOGDEBUG(("using server to resolve hostnames rather than resolving it first\n"));
// if the PROXY_RESOLVES_HOST flag is set, we assume
// that the transport wants us to pass the SOCKS server the
// hostname and port and let it do the name resolution.
// the real destination hostname and port was stored
// in our info object earlier when this layer was created.
const nsCString& destHost = info->DestinationHost();
LOGDEBUG(("host:port -> %s:%li", destHost.get(), destPort));
request[3] = 0x03; // encoding of destination address (3 == hostname)
int host_len = destHost.Length();
if (host_len > 255) {
// SOCKS5 transmits the length of the hostname in a single char.
// This gives us an absolute limit of 255 chars in a hostname, and
// there's nothing we can do to extend it. I don't think many
// hostnames will ever be bigger than this, so hopefully it's an
// uneventful abort condition.
LOGERROR (("Hostname too big for SOCKS5."));
return NS_ERROR_INVALID_ARG;
}
request[4] = (char) host_len;
request_len = 5;
// Send the initial header first...
write_len = pr_Send(fd, request, request_len, 0, &timeout);
if (write_len != request_len) {
// bad write
return NS_ERROR_FAILURE;
}
// Now send the hostname...
write_len = pr_Send(fd, destHost.get(), host_len, 0, &timeout);
if (write_len != host_len) {
// bad write
return NS_ERROR_FAILURE;
}
// There's no data left because we just sent it.
request_len = 0;
} else if (PR_NetAddrFamily(addr) == PR_AF_INET) {
request[3] = 0x01; // encoding of destination address (1 == IPv4)
request_len = 8; // 4 for address, 4 SOCKS headers
char * ip = (char*)(&addr->inet.ip);
request[4] = *ip++;
request[5] = *ip++;
request[6] = *ip++;
request[7] = *ip++;
} else if (PR_NetAddrFamily(addr) == PR_AF_INET6) {
request[3] = 0x04; // encoding of destination address (4 == IPv6)
request_len = 20; // 16 for address, 4 SOCKS headers
char * ip = (char*)(&addr->ipv6.ip.pr_s6_addr);
request[4] = *ip++; request[5] = *ip++;
request[6] = *ip++; request[7] = *ip++;
request[8] = *ip++; request[9] = *ip++;
request[10] = *ip++; request[11] = *ip++;
request[12] = *ip++; request[13] = *ip++;
request[14] = *ip++; request[15] = *ip++;
request[16] = *ip++; request[17] = *ip++;
request[18] = *ip++; request[19] = *ip++;
// we're going to test to see if this address can
// be mapped back into IPv4 without loss. if so,
// we'll use IPv4 instead, as reliable SOCKS server
// support for IPv6 is probably questionable.
if (PR_IsNetAddrType(addr, PR_IpAddrV4Mapped)) {
request[3] = 0x01; // ipv4 encoding
request[4] = request[16];
request[5] = request[17];
request[6] = request[18];
request[7] = request[19];
request_len -= 12;
}
} else {
// Unknown address type
LOGERROR(("Don't know what kind of IP address this is."));
return NS_ERROR_FAILURE;
}
// add the destination port to the request
request[request_len] = (unsigned char)(destPort >> 8);
request[request_len+1] = (unsigned char)destPort;
request_len += 2;
write_len = pr_Send(fd, request, request_len, 0, &timeout);
if (write_len != request_len) {
// bad write
return NS_ERROR_FAILURE;
}
desired_len = 5;
response_len = pr_RecvAll(fd, response, desired_len, 0, &timeout);
if (response_len < desired_len) { // bad read
LOGERROR(("pr_RecvAll() failed getting connect command reply. response_len = %d.", response_len));
return NS_ERROR_FAILURE;
}
if (response[0] != 0x05) {
// bad response
LOGERROR(("Not a SOCKS 5 reply. Expected: 5; received: %x", response[0]));
return NS_ERROR_FAILURE;
}
switch(response[1]) {
case 0x00: break; // success
case 0x01: LOGERROR(("SOCKS 5 server rejected connect request: 01, General SOCKS server failure."));
return NS_ERROR_FAILURE;
case 0x02: LOGERROR(("SOCKS 5 server rejected connect request: 02, Connection not allowed by ruleset."));
return NS_ERROR_FAILURE;
case 0x03: LOGERROR(("SOCKS 5 server rejected connect request: 03, Network unreachable."));
return NS_ERROR_FAILURE;
case 0x04: LOGERROR(("SOCKS 5 server rejected connect request: 04, Host unreachable."));
return NS_ERROR_FAILURE;
case 0x05: LOGERROR(("SOCKS 5 server rejected connect request: 05, Connection refused."));
return NS_ERROR_FAILURE;
case 0x06: LOGERROR(("SOCKS 5 server rejected connect request: 06, TTL expired."));
return NS_ERROR_FAILURE;
case 0x07: LOGERROR(("SOCKS 5 server rejected connect request: 07, Command not supported."));
return NS_ERROR_FAILURE;
case 0x08: LOGERROR(("SOCKS 5 server rejected connect request: 08, Address type not supported."));
return NS_ERROR_FAILURE;
default: LOGERROR(("SOCKS 5 server rejected connect request: %x.", response[1]));
return NS_ERROR_FAILURE;
}
switch (response[3]) {
case 0x01: // IPv4
desired_len = 4 + 2 - 1;
break;
case 0x03: // FQDN
desired_len = response[4] + 2;
break;
case 0x04: // IPv6
desired_len = 16 + 2 - 1;
break;
default: // unknown format
return NS_ERROR_FAILURE;
break;
}
response_len = pr_RecvAll(fd, response + 5, desired_len, 0, &timeout);
if (response_len < desired_len) { // bad read
LOGERROR(("pr_RecvAll() failed getting connect command reply. response_len = %d.", response_len));
return NS_ERROR_FAILURE;
}
response_len += 5;
// get external bound address (this is what
// the outside world sees as "us")
char *ip = nsnull;
PRUint16 extPort = 0;
switch (response[3]) {
case 0x01: // IPv4
extPort = (response[8] << 8) | response[9];
PR_SetNetAddr(PR_IpAddrAny, PR_AF_INET, extPort, extAddr);
ip = (char*)(&extAddr->inet.ip);
*ip++ = response[4];
*ip++ = response[5];
*ip++ = response[6];
*ip++ = response[7];
break;
case 0x04: // IPv6
extPort = (response[20] << 8) | response[21];
PR_SetNetAddr(PR_IpAddrAny, PR_AF_INET6, extPort, extAddr);
ip = (char*)(&extAddr->ipv6.ip.pr_s6_addr);
*ip++ = response[4]; *ip++ = response[5];
*ip++ = response[6]; *ip++ = response[7];
*ip++ = response[8]; *ip++ = response[9];
*ip++ = response[10]; *ip++ = response[11];
*ip++ = response[12]; *ip++ = response[13];
*ip++ = response[14]; *ip++ = response[15];
*ip++ = response[16]; *ip++ = response[17];
*ip++ = response[18]; *ip++ = response[19];
break;
case 0x03: // FQDN
// if we get here, we don't know our external address.
// however, as that's possibly not critical to the user,
// we let it slide.
extPort = (response[response_len - 2] << 8) |
response[response_len - 1];
PR_InitializeNetAddr(PR_IpAddrNull, extPort, extAddr);
break;
}
return NS_OK;
}
// Negotiate a SOCKS 4 connection. Assumes the TCP connection to the socks
// server port has been established.
static nsresult
ConnectSOCKS4(PRFileDesc *fd, const PRNetAddr *addr, PRIntervalTime timeout)
{
int request_len = 0;
int write_len;
int response_len = 0;
int desired_len = 0;
char *ip = nsnull;
unsigned char request[12];
unsigned char response[10];
NS_ENSURE_TRUE(fd, NS_ERROR_NOT_INITIALIZED);
NS_ENSURE_TRUE(addr, NS_ERROR_NOT_INITIALIZED);
request[0] = 0x04; // SOCKS version 4
request[1] = 0x01; // CD command code -- 1 for connect
// destination port
PRInt32 destPort = PR_ntohs(PR_NetAddrInetPort(addr));
// store the port
request[2] = (unsigned char)(destPort >> 8);
request[3] = (unsigned char)destPort;
// username
request[8] = 'M';
request[9] = 'O';
request[10] = 'Z';
request[11] = 0x00;
request_len = 12;
nsSOCKSSocketInfo * info = (nsSOCKSSocketInfo*) fd->secret;
if (info->Flags() & nsISocketProvider::PROXY_RESOLVES_HOST) {
LOGDEBUG(("using server to resolve hostnames rather than resolving it first\n"));
// if the PROXY_RESOLVES_HOST flag is set, we assume that the
// transport wants us to pass the SOCKS server the hostname
// and port and let it do the name resolution.
// an extension to SOCKS 4, called 4a, specifies a way
// to do this, so we'll try that and hope the
// server supports it.
// the real destination hostname and port was stored
// in our info object earlier when this layer was created.
const nsCString& destHost = info->DestinationHost();
LOGDEBUG(("host:port -> %s:%li\n", destHost.get(), destPort));
// the IP portion of the query is set to this special address.
request[4] = 0;
request[5] = 0;
request[6] = 0;
request[7] = 1;
write_len = pr_Send(fd, request, request_len, 0, &timeout);
if (write_len != request_len) {
return NS_ERROR_FAILURE;
}
// Remember the NULL.
int host_len = destHost.Length() + 1;
write_len = pr_Send(fd, destHost.get(), host_len, 0, &timeout);
if (write_len != host_len) {
return NS_ERROR_FAILURE;
}
// No data to send, just sent it.
request_len = 0;
} else if (PR_NetAddrFamily(addr) == PR_AF_INET) { // IPv4
// store the ip
ip = (char*)(&addr->inet.ip);
request[4] = *ip++;
request[5] = *ip++;
request[6] = *ip++;
request[7] = *ip++;
} else if (PR_NetAddrFamily(addr) == PR_AF_INET6) { // IPv6
// IPv4 address encoded in an IPv6 address
if (PR_IsNetAddrType(addr, PR_IpAddrV4Mapped)) {
// store the ip
ip = (char*)(&addr->ipv6.ip.pr_s6_addr[12]);
request[4] = *ip++;
request[5] = *ip++;
request[6] = *ip++;
request[7] = *ip++;
} else {
LOGERROR(("IPv6 is not supported in SOCKS 4."));
return NS_ERROR_FAILURE; // SOCKS 4 can't do IPv6
}
} else {
LOGERROR(("Don't know what kind of IP address this is."));
return NS_ERROR_FAILURE; // don't recognize this type
}
if (request_len > 0) {
write_len = pr_Send(fd, request, request_len, 0, &timeout);
if (write_len != request_len) {
return NS_ERROR_FAILURE;
}
}
// get the server's response
desired_len = 8; // size of the response
response_len = pr_RecvAll(fd, response, desired_len, 0, &timeout);
if (response_len < desired_len) {
LOGERROR(("pr_RecvAll() failed. response_len = %d.", response_len));
return NS_ERROR_FAILURE;
}
if ((response[0] != 0x00) && (response[0] != 0x04)) {
// Novell BorderManager sends a response of type 4, should be zero
// According to the spec. Cope with this brokenness.
// it's not a SOCKS 4 reply or version 0 of the reply code
LOGERROR(("Not a SOCKS 4 reply. Expected: 0; received: %x.", response[0]));
return NS_ERROR_FAILURE;
}
if (response[1] != 0x5A) { // = 90: request granted
// connect request not granted
LOGERROR(("Connection request refused. Expected: 90; received: %d.", response[1]));
return NS_ERROR_FAILURE;
}
return NS_OK;
}
static PRStatus
nsSOCKSIOLayerConnect(PRFileDesc *fd, const PRNetAddr *addr, PRIntervalTime /*timeout*/)
{
PRStatus status;
nsSOCKSSocketInfo * info = (nsSOCKSSocketInfo*) fd->secret;
if (info == NULL) return PR_FAILURE;
// First, we need to look up our proxy...
const nsCString &proxyHost = info->ProxyHost();
if (proxyHost.IsEmpty())
return PR_FAILURE;
PRInt32 socksVersion = info->Version();
LOGDEBUG(("nsSOCKSIOLayerConnect SOCKS %u; proxyHost: %s.", socksVersion, proxyHost.get()));
// Sync resolve the proxy hostname.
PRNetAddr proxyAddr;
nsCOMPtr<nsIDNSRecord> rec;
nsresult rv;
{
nsCOMPtr<nsIDNSService> dns = do_GetService(NS_DNSSERVICE_CONTRACTID);
if (!dns)
return PR_FAILURE;
rv = dns->Resolve(proxyHost, 0, getter_AddRefs(rec));
if (NS_FAILED(rv))
return PR_FAILURE;
}
info->SetInternalProxyAddr(&proxyAddr);
// For now, we'll do this as a blocking connect,
// but with nspr 4.1, the necessary functions to
// do a non-blocking connect will be available
// Preserve the non-blocking state of the socket
PRBool nonblocking;
PRSocketOptionData sockopt;
sockopt.option = PR_SockOpt_Nonblocking;
status = PR_GetSocketOption(fd, &sockopt);
if (PR_SUCCESS != status) {
LOGERROR(("PR_GetSocketOption() failed. status = %x.", status));
return status;
}
// Store blocking option
nonblocking = sockopt.value.non_blocking;
sockopt.option = PR_SockOpt_Nonblocking;
sockopt.value.non_blocking = PR_FALSE;
status = PR_SetSocketOption(fd, &sockopt);
if (PR_SUCCESS != status) {
LOGERROR(("PR_SetSocketOption() failed. status = %x.", status));
return status;
}
// Now setup sockopts, so we can restore the value later.
sockopt.option = PR_SockOpt_Nonblocking;
sockopt.value.non_blocking = nonblocking;
// This connectWait should be long enough to connect to local proxy
// servers, but not much longer. Since this protocol negotiation
// uses blocking network calls, the app can appear to hang for a maximum
// of this time if the user presses the STOP button during the SOCKS
// connection negotiation. Note that this value only applies to the
// connecting to the SOCKS server: once the SOCKS connection has been
// established, the value is not used anywhere else.
PRIntervalTime connectWait = PR_SecondsToInterval(10);
// Connect to the proxy server.
PRInt32 addresses = 0;
do {
rv = rec->GetNextAddr(info->ProxyPort(), &proxyAddr);
if (NS_FAILED(rv)) {
status = PR_FAILURE;
break;
}
++addresses;
status = fd->lower->methods->connect(fd->lower, &proxyAddr, connectWait);
} while (PR_SUCCESS != status);
if (PR_SUCCESS != status) {
LOGERROR(("Failed to TCP connect to the proxy server (%s): timeout = %d, status = %x, tried %d addresses.", proxyHost.get(), connectWait, status, addresses));
PR_SetSocketOption(fd, &sockopt);
return status;
}
// We are now connected to the SOCKS proxy server.
// Now we will negotiate a connection to the desired server.
// External IP address returned from ConnectSOCKS5(). Not supported in SOCKS4.
PRNetAddr extAddr;
PR_InitializeNetAddr(PR_IpAddrNull, 0, &extAddr);
NS_ASSERTION((socksVersion == 4) || (socksVersion == 5), "SOCKS Version must be selected");
// Try to connect via SOCKS 5.
if (socksVersion == 5) {
rv = ConnectSOCKS5(fd, addr, &extAddr, connectWait);
if (NS_FAILED(rv)) {
PR_SetSocketOption(fd, &sockopt);
return PR_FAILURE;
}
}
// Try to connect via SOCKS 4.
else {
rv = ConnectSOCKS4(fd, addr, connectWait);
if (NS_FAILED(rv)) {
PR_SetSocketOption(fd, &sockopt);
return PR_FAILURE;
}
}
info->SetDestinationAddr((PRNetAddr*)addr);
info->SetExternalProxyAddr(&extAddr);
// restore non-blocking option
PR_SetSocketOption(fd, &sockopt);
// we're set-up and connected.
// this socket can be used as normal now.
return PR_SUCCESS;
}
static PRStatus
nsSOCKSIOLayerClose(PRFileDesc *fd)
{
nsSOCKSSocketInfo * info = (nsSOCKSSocketInfo*) fd->secret;
PRDescIdentity id = PR_GetLayersIdentity(fd);
if (info && id == nsSOCKSIOLayerIdentity)
{
NS_RELEASE(info);
fd->identity = PR_INVALID_IO_LAYER;
}
return fd->lower->methods->close(fd->lower);
}
static PRFileDesc*
nsSOCKSIOLayerAccept(PRFileDesc *fd, PRNetAddr *addr, PRIntervalTime timeout)
{
// TODO: implement SOCKS support for accept
return fd->lower->methods->accept(fd->lower, addr, timeout);
}
static PRInt32
nsSOCKSIOLayerAcceptRead(PRFileDesc *sd, PRFileDesc **nd, PRNetAddr **raddr, void *buf, PRInt32 amount, PRIntervalTime timeout)
{
// TODO: implement SOCKS support for accept, then read from it
return sd->lower->methods->acceptread(sd->lower, nd, raddr, buf, amount, timeout);
}
static PRStatus
nsSOCKSIOLayerBind(PRFileDesc *fd, const PRNetAddr *addr)
{
// TODO: implement SOCKS support for bind (very similar to connect)
return fd->lower->methods->bind(fd->lower, addr);
}
static PRStatus
nsSOCKSIOLayerGetName(PRFileDesc *fd, PRNetAddr *addr)
{
nsSOCKSSocketInfo * info = (nsSOCKSSocketInfo*) fd->secret;
if (info != NULL && addr != NULL) {
if (info->GetExternalProxyAddr(&addr) == NS_OK)
return PR_SUCCESS;
}
return PR_FAILURE;
}
static PRStatus
nsSOCKSIOLayerGetPeerName(PRFileDesc *fd, PRNetAddr *addr)
{
nsSOCKSSocketInfo * info = (nsSOCKSSocketInfo*) fd->secret;
if (info != NULL && addr != NULL) {
if (info->GetDestinationAddr(&addr) == NS_OK)
return PR_SUCCESS;
}
return PR_FAILURE;
}
static PRStatus
nsSOCKSIOLayerListen(PRFileDesc *fd, PRIntn backlog)
{
// TODO: implement SOCKS support for listen
return fd->lower->methods->listen(fd->lower, backlog);
}
// add SOCKS IO layer to an existing socket
nsresult
nsSOCKSIOLayerAddToSocket(PRInt32 family,
const char *host,
PRInt32 port,
const char *proxyHost,
PRInt32 proxyPort,
PRInt32 socksVersion,
PRUint32 flags,
PRFileDesc *fd,
nsISupports** info)
{
NS_ENSURE_TRUE((socksVersion == 4) || (socksVersion == 5), NS_ERROR_NOT_INITIALIZED);
if (firstTime)
{
nsSOCKSIOLayerIdentity = PR_GetUniqueIdentity("SOCKS layer");
nsSOCKSIOLayerMethods = *PR_GetDefaultIOMethods();
nsSOCKSIOLayerMethods.connect = nsSOCKSIOLayerConnect;
nsSOCKSIOLayerMethods.bind = nsSOCKSIOLayerBind;
nsSOCKSIOLayerMethods.acceptread = nsSOCKSIOLayerAcceptRead;
nsSOCKSIOLayerMethods.getsockname = nsSOCKSIOLayerGetName;
nsSOCKSIOLayerMethods.getpeername = nsSOCKSIOLayerGetPeerName;
nsSOCKSIOLayerMethods.accept = nsSOCKSIOLayerAccept;
nsSOCKSIOLayerMethods.listen = nsSOCKSIOLayerListen;
nsSOCKSIOLayerMethods.close = nsSOCKSIOLayerClose;
firstTime = PR_FALSE;
#if defined(PR_LOGGING)
gSOCKSLog = PR_NewLogModule("SOCKS");
#endif
}
LOGDEBUG(("Entering nsSOCKSIOLayerAddToSocket()."));
PRFileDesc * layer;
PRStatus rv;
layer = PR_CreateIOLayerStub(nsSOCKSIOLayerIdentity, &nsSOCKSIOLayerMethods);
if (! layer)
{
LOGERROR(("PR_CreateIOLayerStub() failed."));
return NS_ERROR_FAILURE;
}
nsSOCKSSocketInfo * infoObject = new nsSOCKSSocketInfo();
if (!infoObject)
{
// clean up IOLayerStub
LOGERROR(("Failed to create nsSOCKSSocketInfo()."));
PR_DELETE(layer);
return NS_ERROR_FAILURE;
}
NS_ADDREF(infoObject);
infoObject->Init(socksVersion, proxyHost, proxyPort, host, flags);
layer->secret = (PRFilePrivate*) infoObject;
rv = PR_PushIOLayer(fd, PR_GetLayersIdentity(fd), layer);
if (NS_FAILED(rv))
{
LOGERROR(("PR_PushIOLayer() failed. rv = %x.", rv));
NS_RELEASE(infoObject);
PR_DELETE(layer);
return NS_ERROR_FAILURE;
}
*info = infoObject;
NS_ADDREF(*info);
return NS_OK;
}