diff options
author | Matt A. Tobin <mattatobin@localhost.localdomain> | 2018-02-02 04:16:08 -0500 |
---|---|---|
committer | Matt A. Tobin <mattatobin@localhost.localdomain> | 2018-02-02 04:16:08 -0500 |
commit | 5f8de423f190bbb79a62f804151bc24824fa32d8 (patch) | |
tree | 10027f336435511475e392454359edea8e25895d /netwerk/socket/nsSOCKSIOLayer.cpp | |
parent | 49ee0794b5d912db1f95dce6eb52d781dc210db5 (diff) | |
download | UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.tar UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.tar.gz UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.tar.lz UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.tar.xz UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.zip |
Add m-esr52 at 52.6.0
Diffstat (limited to 'netwerk/socket/nsSOCKSIOLayer.cpp')
-rw-r--r-- | netwerk/socket/nsSOCKSIOLayer.cpp | 1612 |
1 files changed, 1612 insertions, 0 deletions
diff --git a/netwerk/socket/nsSOCKSIOLayer.cpp b/netwerk/socket/nsSOCKSIOLayer.cpp new file mode 100644 index 000000000..22f5751fb --- /dev/null +++ b/netwerk/socket/nsSOCKSIOLayer.cpp @@ -0,0 +1,1612 @@ +/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */ +/* vim:set expandtab ts=4 sw=4 sts=4 cin: */ +/* This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ + +#include "nspr.h" +#include "private/pprio.h" +#include "nsString.h" +#include "nsCRT.h" + +#include "nsIServiceManager.h" +#include "nsIDNSService.h" +#include "nsIDNSRecord.h" +#include "nsISOCKSSocketInfo.h" +#include "nsISocketProvider.h" +#include "nsNamedPipeIOLayer.h" +#include "nsSOCKSIOLayer.h" +#include "nsNetCID.h" +#include "nsIDNSListener.h" +#include "nsICancelable.h" +#include "nsThreadUtils.h" +#include "nsIFile.h" +#include "nsIFileProtocolHandler.h" +#include "mozilla/Logging.h" +#include "mozilla/net/DNS.h" +#include "mozilla/Unused.h" + +using mozilla::LogLevel; +using namespace mozilla::net; + +static PRDescIdentity nsSOCKSIOLayerIdentity; +static PRIOMethods nsSOCKSIOLayerMethods; +static bool firstTime = true; +static bool ipv6Supported = true; + + +static mozilla::LazyLogModule gSOCKSLog("SOCKS"); +#define LOGDEBUG(args) MOZ_LOG(gSOCKSLog, mozilla::LogLevel::Debug, args) +#define LOGERROR(args) MOZ_LOG(gSOCKSLog, mozilla::LogLevel::Error , args) + +class nsSOCKSSocketInfo : public nsISOCKSSocketInfo + , public nsIDNSListener +{ + enum State { + SOCKS_INITIAL, + SOCKS_DNS_IN_PROGRESS, + SOCKS_DNS_COMPLETE, + SOCKS_CONNECTING_TO_PROXY, + SOCKS4_WRITE_CONNECT_REQUEST, + SOCKS4_READ_CONNECT_RESPONSE, + SOCKS5_WRITE_AUTH_REQUEST, + SOCKS5_READ_AUTH_RESPONSE, + SOCKS5_WRITE_USERNAME_REQUEST, + SOCKS5_READ_USERNAME_RESPONSE, + SOCKS5_WRITE_CONNECT_REQUEST, + SOCKS5_READ_CONNECT_RESPONSE_TOP, + SOCKS5_READ_CONNECT_RESPONSE_BOTTOM, + SOCKS_CONNECTED, + SOCKS_FAILED + }; + + // A buffer of 520 bytes should be enough for any request and response + // in case of SOCKS4 as well as SOCKS5 + static const uint32_t BUFFER_SIZE = 520; + static const uint32_t MAX_HOSTNAME_LEN = 255; + static const uint32_t MAX_USERNAME_LEN = 255; + static const uint32_t MAX_PASSWORD_LEN = 255; + +public: + nsSOCKSSocketInfo(); + + NS_DECL_THREADSAFE_ISUPPORTS + NS_DECL_NSISOCKSSOCKETINFO + NS_DECL_NSIDNSLISTENER + + void Init(int32_t version, + int32_t family, + nsIProxyInfo *proxy, + const char *destinationHost, + uint32_t flags); + + void SetConnectTimeout(PRIntervalTime to); + PRStatus DoHandshake(PRFileDesc *fd, int16_t oflags = -1); + int16_t GetPollFlags() const; + bool IsConnected() const { return mState == SOCKS_CONNECTED; } + void ForgetFD() { mFD = nullptr; } + void SetNamedPipeFD(PRFileDesc *fd) { mFD = fd; } + +private: + virtual ~nsSOCKSSocketInfo() + { + ForgetFD(); + HandshakeFinished(); + } + + void HandshakeFinished(PRErrorCode err = 0); + PRStatus StartDNS(PRFileDesc *fd); + PRStatus ConnectToProxy(PRFileDesc *fd); + void FixupAddressFamily(PRFileDesc *fd, NetAddr *proxy); + PRStatus ContinueConnectingToProxy(PRFileDesc *fd, int16_t oflags); + PRStatus WriteV4ConnectRequest(); + PRStatus ReadV4ConnectResponse(); + PRStatus WriteV5AuthRequest(); + PRStatus ReadV5AuthResponse(); + PRStatus WriteV5UsernameRequest(); + PRStatus ReadV5UsernameResponse(); + PRStatus WriteV5ConnectRequest(); + PRStatus ReadV5AddrTypeAndLength(uint8_t *type, uint32_t *len); + PRStatus ReadV5ConnectResponseTop(); + PRStatus ReadV5ConnectResponseBottom(); + + uint8_t ReadUint8(); + uint16_t ReadUint16(); + uint32_t ReadUint32(); + void ReadNetAddr(NetAddr *addr, uint16_t fam); + void ReadNetPort(NetAddr *addr); + + void WantRead(uint32_t sz); + PRStatus ReadFromSocket(PRFileDesc *fd); + PRStatus WriteToSocket(PRFileDesc *fd); + + bool IsLocalProxy() + { + nsAutoCString proxyHost; + mProxy->GetHost(proxyHost); + return IsHostLocalTarget(proxyHost); + } + + nsresult SetLocalProxyPath(const nsACString& aLocalProxyPath, + NetAddr* aProxyAddr) + { +#ifdef XP_UNIX + nsresult rv; + MOZ_ASSERT(aProxyAddr); + + nsCOMPtr<nsIProtocolHandler> protocolHandler( + do_GetService(NS_NETWORK_PROTOCOL_CONTRACTID_PREFIX "file", &rv)); + if (NS_WARN_IF(NS_FAILED(rv))) { + return rv; + } + + nsCOMPtr<nsIFileProtocolHandler> fileHandler( + do_QueryInterface(protocolHandler, &rv)); + if (NS_WARN_IF(NS_FAILED(rv))) { + return rv; + } + + nsCOMPtr<nsIFile> socketFile; + rv = fileHandler->GetFileFromURLSpec(aLocalProxyPath, + getter_AddRefs(socketFile)); + if (NS_WARN_IF(NS_FAILED(rv))) { + return rv; + } + + nsAutoCString path; + if (NS_WARN_IF(NS_FAILED(rv = socketFile->GetNativePath(path)))) { + return rv; + } + + if (sizeof(aProxyAddr->local.path) <= path.Length()) { + NS_WARNING("domain socket path too long."); + return NS_ERROR_FAILURE; + } + + aProxyAddr->raw.family = AF_UNIX; + strcpy(aProxyAddr->local.path, path.get()); + + return NS_OK; +#elif defined(XP_WIN) + MOZ_ASSERT(aProxyAddr); + + if (sizeof(aProxyAddr->local.path) <= aLocalProxyPath.Length()) { + NS_WARNING("pipe path too long."); + return NS_ERROR_FAILURE; + } + + aProxyAddr->raw.family = AF_LOCAL; + strcpy(aProxyAddr->local.path, PromiseFlatCString(aLocalProxyPath).get()); + return NS_OK; +#else + mozilla::Unused << aLocalProxyPath; + mozilla::Unused << aProxyAddr; + return NS_ERROR_NOT_IMPLEMENTED; +#endif + } + + bool + SetupNamedPipeLayer(PRFileDesc *fd) + { +#if defined(XP_WIN) + if (IsLocalProxy()) { + // nsSOCKSIOLayer handshaking only works under blocking mode + // unfortunately. Remember named pipe's FD to switch between modes. + SetNamedPipeFD(fd->lower); + return true; + } +#endif + return false; + } + +private: + State mState; + uint8_t * mData; + uint8_t * mDataIoPtr; + uint32_t mDataLength; + uint32_t mReadOffset; + uint32_t mAmountToRead; + nsCOMPtr<nsIDNSRecord> mDnsRec; + nsCOMPtr<nsICancelable> mLookup; + nsresult mLookupStatus; + PRFileDesc *mFD; + + nsCString mDestinationHost; + nsCOMPtr<nsIProxyInfo> mProxy; + int32_t mVersion; // SOCKS version 4 or 5 + int32_t mDestinationFamily; + uint32_t mFlags; + NetAddr mInternalProxyAddr; + NetAddr mExternalProxyAddr; + NetAddr mDestinationAddr; + PRIntervalTime mTimeout; + nsCString mProxyUsername; // Cache, from mProxy +}; + +nsSOCKSSocketInfo::nsSOCKSSocketInfo() + : mState(SOCKS_INITIAL) + , mDataIoPtr(nullptr) + , mDataLength(0) + , mReadOffset(0) + , mAmountToRead(0) + , mVersion(-1) + , mDestinationFamily(AF_INET) + , mFlags(0) + , mTimeout(PR_INTERVAL_NO_TIMEOUT) +{ + mData = new uint8_t[BUFFER_SIZE]; + + mInternalProxyAddr.raw.family = AF_INET; + mInternalProxyAddr.inet.ip = htonl(INADDR_ANY); + mInternalProxyAddr.inet.port = htons(0); + + mExternalProxyAddr.raw.family = AF_INET; + mExternalProxyAddr.inet.ip = htonl(INADDR_ANY); + mExternalProxyAddr.inet.port = htons(0); + + mDestinationAddr.raw.family = AF_INET; + mDestinationAddr.inet.ip = htonl(INADDR_ANY); + mDestinationAddr.inet.port = htons(0); +} + +/* Helper template class to statically check that writes to a fixed-size + * buffer are not going to overflow. + * + * Example usage: + * uint8_t real_buf[TOTAL_SIZE]; + * Buffer<TOTAL_SIZE> buf(&real_buf); + * auto buf2 = buf.WriteUint16(1); + * auto buf3 = buf2.WriteUint8(2); + * + * It is possible to chain them, to limit the number of (error-prone) + * intermediate variables: + * auto buf = Buffer<TOTAL_SIZE>(&real_buf) + * .WriteUint16(1) + * .WriteUint8(2); + * + * Debug builds assert when intermediate variables are reused: + * Buffer<TOTAL_SIZE> buf(&real_buf); + * auto buf2 = buf.WriteUint16(1); + * auto buf3 = buf.WriteUint8(2); // Asserts + * + * Strings can be written, given an explicit maximum length. + * buf.WriteString<MAX_STRING_LENGTH>(str); + * + * The Written() method returns how many bytes have been written so far: + * Buffer<TOTAL_SIZE> buf(&real_buf); + * auto buf2 = buf.WriteUint16(1); + * auto buf3 = buf2.WriteUint8(2); + * buf3.Written(); // returns 3. + */ +template <size_t Size> +class Buffer +{ +public: + Buffer() : mBuf(nullptr), mLength(0) {} + + explicit Buffer(uint8_t* aBuf, size_t aLength=0) + : mBuf(aBuf), mLength(aLength) {} + + template <size_t Size2> + MOZ_IMPLICIT Buffer(const Buffer<Size2>& aBuf) : mBuf(aBuf.mBuf), mLength(aBuf.mLength) { + static_assert(Size2 > Size, "Cannot cast buffer"); + } + + Buffer<Size - sizeof(uint8_t)> WriteUint8(uint8_t aValue) { + return Write(aValue); + } + + Buffer<Size - sizeof(uint16_t)> WriteUint16(uint16_t aValue) { + return Write(aValue); + } + + Buffer<Size - sizeof(uint32_t)> WriteUint32(uint32_t aValue) { + return Write(aValue); + } + + Buffer<Size - sizeof(uint16_t)> WriteNetPort(const NetAddr* aAddr) { + return WriteUint16(aAddr->inet.port); + } + + Buffer<Size - sizeof(IPv6Addr)> WriteNetAddr(const NetAddr* aAddr) { + if (aAddr->raw.family == AF_INET) { + return Write(aAddr->inet.ip); + } else if (aAddr->raw.family == AF_INET6) { + return Write(aAddr->inet6.ip.u8); + } + NS_NOTREACHED("Unknown address family"); + return *this; + } + + template <size_t MaxLength> + Buffer<Size - MaxLength> WriteString(const nsACString& aStr) { + if (aStr.Length() > MaxLength) { + return Buffer<Size - MaxLength>(nullptr); + } + return WritePtr<char, MaxLength>(aStr.Data(), aStr.Length()); + } + + size_t Written() { + MOZ_ASSERT(mBuf); + return mLength; + } + + explicit operator bool() { return !!mBuf; } +private: + template <size_t Size2> + friend class Buffer; + + template <typename T> + Buffer<Size - sizeof(T)> Write(T& aValue) { + return WritePtr<T, sizeof(T)>(&aValue, sizeof(T)); + } + + template <typename T, size_t Length> + Buffer<Size - Length> WritePtr(const T* aValue, size_t aCopyLength) { + static_assert(Size >= Length, "Cannot write that much"); + MOZ_ASSERT(aCopyLength <= Length); + MOZ_ASSERT(mBuf); + memcpy(mBuf, aValue, aCopyLength); + Buffer<Size - Length> result(mBuf + aCopyLength, mLength + aCopyLength); + mBuf = nullptr; + mLength = 0; + return result; + } + + uint8_t* mBuf; + size_t mLength; +}; + + +void +nsSOCKSSocketInfo::Init(int32_t version, int32_t family, nsIProxyInfo *proxy, const char *host, uint32_t flags) +{ + mVersion = version; + mDestinationFamily = family; + mProxy = proxy; + mDestinationHost = host; + mFlags = flags; + mProxy->GetUsername(mProxyUsername); // cache +} + +NS_IMPL_ISUPPORTS(nsSOCKSSocketInfo, nsISOCKSSocketInfo, nsIDNSListener) + +NS_IMETHODIMP +nsSOCKSSocketInfo::GetExternalProxyAddr(NetAddr * *aExternalProxyAddr) +{ + memcpy(*aExternalProxyAddr, &mExternalProxyAddr, sizeof(NetAddr)); + return NS_OK; +} + +NS_IMETHODIMP +nsSOCKSSocketInfo::SetExternalProxyAddr(NetAddr *aExternalProxyAddr) +{ + memcpy(&mExternalProxyAddr, aExternalProxyAddr, sizeof(NetAddr)); + return NS_OK; +} + +NS_IMETHODIMP +nsSOCKSSocketInfo::GetDestinationAddr(NetAddr * *aDestinationAddr) +{ + memcpy(*aDestinationAddr, &mDestinationAddr, sizeof(NetAddr)); + return NS_OK; +} + +NS_IMETHODIMP +nsSOCKSSocketInfo::SetDestinationAddr(NetAddr *aDestinationAddr) +{ + memcpy(&mDestinationAddr, aDestinationAddr, sizeof(NetAddr)); + return NS_OK; +} + +NS_IMETHODIMP +nsSOCKSSocketInfo::GetInternalProxyAddr(NetAddr * *aInternalProxyAddr) +{ + memcpy(*aInternalProxyAddr, &mInternalProxyAddr, sizeof(NetAddr)); + return NS_OK; +} + +NS_IMETHODIMP +nsSOCKSSocketInfo::SetInternalProxyAddr(NetAddr *aInternalProxyAddr) +{ + memcpy(&mInternalProxyAddr, aInternalProxyAddr, sizeof(NetAddr)); + return NS_OK; +} + +// There needs to be a means of distinguishing between connection errors +// that the SOCKS server reports when it rejects a connection request, and +// connection errors that happen while attempting to connect to the SOCKS +// server. Otherwise, Firefox will report incorrectly that the proxy server +// is refusing connections when a SOCKS request is rejected by the proxy. +// When a SOCKS handshake failure occurs, the PR error is set to +// PR_UNKNOWN_ERROR, and the real error code is returned via the OS error. +void +nsSOCKSSocketInfo::HandshakeFinished(PRErrorCode err) +{ + if (err == 0) { + mState = SOCKS_CONNECTED; +#if defined(XP_WIN) + // Switch back to nonblocking mode after finishing handshaking. + if (IsLocalProxy() && mFD) { + PRSocketOptionData opt_nonblock; + opt_nonblock.option = PR_SockOpt_Nonblocking; + opt_nonblock.value.non_blocking = PR_TRUE; + PR_SetSocketOption(mFD, &opt_nonblock); + mFD = nullptr; + } +#endif + } else { + mState = SOCKS_FAILED; + PR_SetError(PR_UNKNOWN_ERROR, err); + } + + // We don't need the buffer any longer, so free it. + delete [] mData; + mData = nullptr; + mDataIoPtr = nullptr; + mDataLength = 0; + mReadOffset = 0; + mAmountToRead = 0; + if (mLookup) { + mLookup->Cancel(NS_ERROR_FAILURE); + mLookup = nullptr; + } +} + +PRStatus +nsSOCKSSocketInfo::StartDNS(PRFileDesc *fd) +{ + MOZ_ASSERT(!mDnsRec && mState == SOCKS_INITIAL, + "Must be in initial state to make DNS Lookup"); + + nsCOMPtr<nsIDNSService> dns = do_GetService(NS_DNSSERVICE_CONTRACTID); + if (!dns) + return PR_FAILURE; + + nsCString proxyHost; + mProxy->GetHost(proxyHost); + + mFD = fd; + nsresult rv = dns->AsyncResolve(proxyHost, 0, this, + NS_GetCurrentThread(), + getter_AddRefs(mLookup)); + + if (NS_FAILED(rv)) { + LOGERROR(("socks: DNS lookup for SOCKS proxy %s failed", + proxyHost.get())); + return PR_FAILURE; + } + mState = SOCKS_DNS_IN_PROGRESS; + PR_SetError(PR_IN_PROGRESS_ERROR, 0); + return PR_FAILURE; +} + +NS_IMETHODIMP +nsSOCKSSocketInfo::OnLookupComplete(nsICancelable *aRequest, + nsIDNSRecord *aRecord, + nsresult aStatus) +{ + MOZ_ASSERT(aRequest == mLookup, "wrong DNS query"); + mLookup = nullptr; + mLookupStatus = aStatus; + mDnsRec = aRecord; + mState = SOCKS_DNS_COMPLETE; + if (mFD) { + ConnectToProxy(mFD); + ForgetFD(); + } + return NS_OK; +} + +PRStatus +nsSOCKSSocketInfo::ConnectToProxy(PRFileDesc *fd) +{ + PRStatus status; + nsresult rv; + + MOZ_ASSERT(mState == SOCKS_DNS_COMPLETE, + "Must have DNS to make connection!"); + + if (NS_FAILED(mLookupStatus)) { + PR_SetError(PR_BAD_ADDRESS_ERROR, 0); + return PR_FAILURE; + } + + // Try socks5 if the destination addrress is IPv6 + if (mVersion == 4 && + mDestinationAddr.raw.family == AF_INET6) { + mVersion = 5; + } + + nsAutoCString proxyHost; + mProxy->GetHost(proxyHost); + + int32_t proxyPort; + mProxy->GetPort(&proxyPort); + + int32_t addresses = 0; + do { + if (IsLocalProxy()) { + rv = SetLocalProxyPath(proxyHost, &mInternalProxyAddr); + if (NS_FAILED(rv)) { + LOGERROR(("socks: unable to connect to SOCKS proxy, %s", + proxyHost.get())); + return PR_FAILURE; + } + } else { + if (addresses++) { + mDnsRec->ReportUnusable(proxyPort); + } + + rv = mDnsRec->GetNextAddr(proxyPort, &mInternalProxyAddr); + // No more addresses to try? If so, we'll need to bail + if (NS_FAILED(rv)) { + LOGERROR(("socks: unable to connect to SOCKS proxy, %s", + proxyHost.get())); + return PR_FAILURE; + } + + if (MOZ_LOG_TEST(gSOCKSLog, LogLevel::Debug)) { + char buf[kIPv6CStrBufSize]; + NetAddrToString(&mInternalProxyAddr, buf, sizeof(buf)); + LOGDEBUG(("socks: trying proxy server, %s:%hu", + buf, ntohs(mInternalProxyAddr.inet.port))); + } + } + + NetAddr proxy = mInternalProxyAddr; + FixupAddressFamily(fd, &proxy); + PRNetAddr prProxy; + NetAddrToPRNetAddr(&proxy, &prProxy); + status = fd->lower->methods->connect(fd->lower, &prProxy, mTimeout); + if (status != PR_SUCCESS) { + PRErrorCode c = PR_GetError(); + + // If EINPROGRESS, return now and check back later after polling + if (c == PR_WOULD_BLOCK_ERROR || c == PR_IN_PROGRESS_ERROR) { + mState = SOCKS_CONNECTING_TO_PROXY; + return status; + } else if (IsLocalProxy()) { + LOGERROR(("socks: connect to domain socket failed (%d)", c)); + PR_SetError(PR_CONNECT_REFUSED_ERROR, 0); + mState = SOCKS_FAILED; + return status; + } + } + } while (status != PR_SUCCESS); + +#if defined(XP_WIN) + // Switch to blocking mode during handshaking + if (IsLocalProxy() && mFD) { + PRSocketOptionData opt_nonblock; + opt_nonblock.option = PR_SockOpt_Nonblocking; + opt_nonblock.value.non_blocking = PR_FALSE; + PR_SetSocketOption(mFD, &opt_nonblock); + } +#endif + + // Connected now, start SOCKS + if (mVersion == 4) + return WriteV4ConnectRequest(); + return WriteV5AuthRequest(); +} + +void +nsSOCKSSocketInfo::FixupAddressFamily(PRFileDesc *fd, NetAddr *proxy) +{ + int32_t proxyFamily = mInternalProxyAddr.raw.family; + // Do nothing if the address family is already matched + if (proxyFamily == mDestinationFamily) { + return; + } + // If the system does not support IPv6 and the proxy address is IPv6, + // We can do nothing here. + if (proxyFamily == AF_INET6 && !ipv6Supported) { + return; + } + // If the system does not support IPv6 and the destination address is + // IPv6, convert IPv4 address to IPv4-mapped IPv6 address to satisfy + // the emulation layer + if (mDestinationFamily == AF_INET6 && !ipv6Supported) { + proxy->inet6.family = AF_INET6; + proxy->inet6.port = mInternalProxyAddr.inet.port; + uint8_t *proxyp = proxy->inet6.ip.u8; + memset(proxyp, 0, 10); + memset(proxyp + 10, 0xff, 2); + memcpy(proxyp + 12,(char *) &mInternalProxyAddr.inet.ip, 4); + // mDestinationFamily should not be updated + return; + } + // There's no PR_NSPR_IO_LAYER required when using named pipe, + // we simply ignore the TCP family here. + if (SetupNamedPipeLayer(fd)) { + return; + } + + // Get an OS native handle from a specified FileDesc + PROsfd osfd = PR_FileDesc2NativeHandle(fd); + if (osfd == -1) { + return; + } + + // Create a new FileDesc with a specified family + PRFileDesc *tmpfd = PR_OpenTCPSocket(proxyFamily); + if (!tmpfd) { + return; + } + PROsfd newsd = PR_FileDesc2NativeHandle(tmpfd); + if (newsd == -1) { + PR_Close(tmpfd); + return; + } + // Must succeed because PR_FileDesc2NativeHandle succeeded + fd = PR_GetIdentitiesLayer(fd, PR_NSPR_IO_LAYER); + MOZ_ASSERT(fd); + // Swap OS native handles + PR_ChangeFileDescNativeHandle(fd, newsd); + PR_ChangeFileDescNativeHandle(tmpfd, osfd); + // Close temporary FileDesc which is now associated with + // old OS native handle + PR_Close(tmpfd); + mDestinationFamily = proxyFamily; +} + +PRStatus +nsSOCKSSocketInfo::ContinueConnectingToProxy(PRFileDesc *fd, int16_t oflags) +{ + PRStatus status; + + MOZ_ASSERT(mState == SOCKS_CONNECTING_TO_PROXY, + "Continuing connection in wrong state!"); + + LOGDEBUG(("socks: continuing connection to proxy")); + + status = fd->lower->methods->connectcontinue(fd->lower, oflags); + if (status != PR_SUCCESS) { + PRErrorCode c = PR_GetError(); + if (c != PR_WOULD_BLOCK_ERROR && c != PR_IN_PROGRESS_ERROR) { + // A connection failure occured, try another address + mState = SOCKS_DNS_COMPLETE; + return ConnectToProxy(fd); + } + + // We're still connecting + return PR_FAILURE; + } + + // Connected now, start SOCKS + if (mVersion == 4) + return WriteV4ConnectRequest(); + return WriteV5AuthRequest(); +} + +PRStatus +nsSOCKSSocketInfo::WriteV4ConnectRequest() +{ + if (mProxyUsername.Length() > MAX_USERNAME_LEN) { + LOGERROR(("socks username is too long")); + HandshakeFinished(PR_UNKNOWN_ERROR); + return PR_FAILURE; + } + + NetAddr *addr = &mDestinationAddr; + int32_t proxy_resolve; + + MOZ_ASSERT(mState == SOCKS_CONNECTING_TO_PROXY, + "Invalid state!"); + + proxy_resolve = mFlags & nsISocketProvider::PROXY_RESOLVES_HOST; + + mDataLength = 0; + mState = SOCKS4_WRITE_CONNECT_REQUEST; + + LOGDEBUG(("socks4: sending connection request (socks4a resolve? %s)", + proxy_resolve? "yes" : "no")); + + // Send a SOCKS 4 connect request. + auto buf = Buffer<BUFFER_SIZE>(mData) + .WriteUint8(0x04) // version -- 4 + .WriteUint8(0x01) // command -- connect + .WriteNetPort(addr); + + // We don't have anything more to write after the if, so we can + // use a buffer with no further writes allowed. + Buffer<0> buf3; + if (proxy_resolve) { + // Add the full name, null-terminated, to the request + // according to SOCKS 4a. A fake IP address, with the first + // four bytes set to 0 and the last byte set to something other + // than 0, is used to notify the proxy that this is a SOCKS 4a + // request. This request type works for Tor and perhaps others. + // Passwords not supported by V4. + auto buf2 = buf.WriteUint32(htonl(0x00000001)) // Fake IP + .WriteString<MAX_USERNAME_LEN>(mProxyUsername) + .WriteUint8(0x00) // Null-terminate username + .WriteString<MAX_HOSTNAME_LEN>(mDestinationHost); // Hostname + if (!buf2) { + LOGERROR(("socks4: destination host name is too long!")); + HandshakeFinished(PR_BAD_ADDRESS_ERROR); + return PR_FAILURE; + } + buf3 = buf2.WriteUint8(0x00); + } else if (addr->raw.family == AF_INET) { + // Passwords not supported by V4. + buf3 = buf.WriteNetAddr(addr) // Add the IPv4 address + .WriteString<MAX_USERNAME_LEN>(mProxyUsername) + .WriteUint8(0x00); // Null-terminate username + } else { + LOGERROR(("socks: SOCKS 4 can only handle IPv4 addresses!")); + HandshakeFinished(PR_BAD_ADDRESS_ERROR); + return PR_FAILURE; + } + + mDataLength = buf3.Written(); + return PR_SUCCESS; +} + +PRStatus +nsSOCKSSocketInfo::ReadV4ConnectResponse() +{ + MOZ_ASSERT(mState == SOCKS4_READ_CONNECT_RESPONSE, + "Handling SOCKS 4 connection reply in wrong state!"); + MOZ_ASSERT(mDataLength == 8, + "SOCKS 4 connection reply must be 8 bytes!"); + + LOGDEBUG(("socks4: checking connection reply")); + + if (ReadUint8() != 0x00) { + LOGERROR(("socks4: wrong connection reply")); + HandshakeFinished(PR_CONNECT_REFUSED_ERROR); + return PR_FAILURE; + } + + // See if our connection request was granted + if (ReadUint8() == 90) { + LOGDEBUG(("socks4: connection successful!")); + HandshakeFinished(); + return PR_SUCCESS; + } + + LOGERROR(("socks4: unable to connect")); + HandshakeFinished(PR_CONNECT_REFUSED_ERROR); + return PR_FAILURE; +} + +PRStatus +nsSOCKSSocketInfo::WriteV5AuthRequest() +{ + MOZ_ASSERT(mVersion == 5, "SOCKS version must be 5!"); + + mDataLength = 0; + mState = SOCKS5_WRITE_AUTH_REQUEST; + + // Send an initial SOCKS 5 greeting + LOGDEBUG(("socks5: sending auth methods")); + mDataLength = Buffer<BUFFER_SIZE>(mData) + .WriteUint8(0x05) // version -- 5 + .WriteUint8(0x01) // # of auth methods -- 1 + // Use authenticate iff we have a proxy username. + .WriteUint8(mProxyUsername.IsEmpty() ? 0x00 : 0x02) + .Written(); + + return PR_SUCCESS; +} + +PRStatus +nsSOCKSSocketInfo::ReadV5AuthResponse() +{ + MOZ_ASSERT(mState == SOCKS5_READ_AUTH_RESPONSE, + "Handling SOCKS 5 auth method reply in wrong state!"); + MOZ_ASSERT(mDataLength == 2, + "SOCKS 5 auth method reply must be 2 bytes!"); + + LOGDEBUG(("socks5: checking auth method reply")); + + // Check version number + if (ReadUint8() != 0x05) { + LOGERROR(("socks5: unexpected version in the reply")); + HandshakeFinished(PR_CONNECT_REFUSED_ERROR); + return PR_FAILURE; + } + + // Make sure our authentication choice was accepted, + // and continue accordingly + uint8_t authMethod = ReadUint8(); + if (mProxyUsername.IsEmpty() && authMethod == 0x00) { // no auth + LOGDEBUG(("socks5: server allows connection without authentication")); + return WriteV5ConnectRequest(); + } else if (!mProxyUsername.IsEmpty() && authMethod == 0x02) { // username/pw + LOGDEBUG(("socks5: auth method accepted by server")); + return WriteV5UsernameRequest(); + } else { // 0xFF signals error + LOGERROR(("socks5: server did not accept our authentication method")); + HandshakeFinished(PR_CONNECT_REFUSED_ERROR); + return PR_FAILURE; + } +} + +PRStatus +nsSOCKSSocketInfo::WriteV5UsernameRequest() +{ + MOZ_ASSERT(mVersion == 5, "SOCKS version must be 5!"); + + if (mProxyUsername.Length() > MAX_USERNAME_LEN) { + LOGERROR(("socks username is too long")); + HandshakeFinished(PR_UNKNOWN_ERROR); + return PR_FAILURE; + } + + nsCString password; + mProxy->GetPassword(password); + if (password.Length() > MAX_PASSWORD_LEN) { + LOGERROR(("socks password is too long")); + HandshakeFinished(PR_UNKNOWN_ERROR); + return PR_FAILURE; + } + + mDataLength = 0; + mState = SOCKS5_WRITE_USERNAME_REQUEST; + + // RFC 1929 Username/password auth for SOCKS 5 + LOGDEBUG(("socks5: sending username and password")); + mDataLength = Buffer<BUFFER_SIZE>(mData) + .WriteUint8(0x01) // version 1 (not 5) + .WriteUint8(mProxyUsername.Length()) // username length + .WriteString<MAX_USERNAME_LEN>(mProxyUsername) // username + .WriteUint8(password.Length()) // password length + .WriteString<MAX_PASSWORD_LEN>(password) // password. WARNING: Sent unencrypted! + .Written(); + + return PR_SUCCESS; +} + +PRStatus +nsSOCKSSocketInfo::ReadV5UsernameResponse() +{ + MOZ_ASSERT(mState == SOCKS5_READ_USERNAME_RESPONSE, + "Handling SOCKS 5 username/password reply in wrong state!"); + + MOZ_ASSERT(mDataLength == 2, + "SOCKS 5 username reply must be 2 bytes"); + + // Check version number, must be 1 (not 5) + if (ReadUint8() != 0x01) { + LOGERROR(("socks5: unexpected version in the reply")); + HandshakeFinished(PR_CONNECT_REFUSED_ERROR); + return PR_FAILURE; + } + + // Check whether username/password were accepted + if (ReadUint8() != 0x00) { // 0 = success + LOGERROR(("socks5: username/password not accepted")); + HandshakeFinished(PR_CONNECT_REFUSED_ERROR); + return PR_FAILURE; + } + + LOGDEBUG(("socks5: username/password accepted by server")); + + return WriteV5ConnectRequest(); +} + +PRStatus +nsSOCKSSocketInfo::WriteV5ConnectRequest() +{ + // Send SOCKS 5 connect request + NetAddr *addr = &mDestinationAddr; + int32_t proxy_resolve; + proxy_resolve = mFlags & nsISocketProvider::PROXY_RESOLVES_HOST; + + LOGDEBUG(("socks5: sending connection request (socks5 resolve? %s)", + proxy_resolve? "yes" : "no")); + + mDataLength = 0; + mState = SOCKS5_WRITE_CONNECT_REQUEST; + + auto buf = Buffer<BUFFER_SIZE>(mData) + .WriteUint8(0x05) // version -- 5 + .WriteUint8(0x01) // command -- connect + .WriteUint8(0x00); // reserved + + // We're writing a net port after the if, so we need a buffer allowing + // to write that much. + Buffer<sizeof(uint16_t)> buf2; + // Add the address to the SOCKS 5 request. SOCKS 5 supports several + // address types, so we pick the one that works best for us. + if (proxy_resolve) { + // Add the host name. Only a single byte is used to store the length, + // so we must prevent long names from being used. + buf2 = buf.WriteUint8(0x03) // addr type -- domainname + .WriteUint8(mDestinationHost.Length()) // name length + .WriteString<MAX_HOSTNAME_LEN>(mDestinationHost); // Hostname + if (!buf2) { + LOGERROR(("socks5: destination host name is too long!")); + HandshakeFinished(PR_BAD_ADDRESS_ERROR); + return PR_FAILURE; + } + } else if (addr->raw.family == AF_INET) { + buf2 = buf.WriteUint8(0x01) // addr type -- IPv4 + .WriteNetAddr(addr); + } else if (addr->raw.family == AF_INET6) { + buf2 = buf.WriteUint8(0x04) // addr type -- IPv6 + .WriteNetAddr(addr); + } else { + LOGERROR(("socks5: destination address of unknown type!")); + HandshakeFinished(PR_BAD_ADDRESS_ERROR); + return PR_FAILURE; + } + + auto buf3 = buf2.WriteNetPort(addr); // port + mDataLength = buf3.Written(); + + return PR_SUCCESS; +} + +PRStatus +nsSOCKSSocketInfo::ReadV5AddrTypeAndLength(uint8_t *type, uint32_t *len) +{ + MOZ_ASSERT(mState == SOCKS5_READ_CONNECT_RESPONSE_TOP || + mState == SOCKS5_READ_CONNECT_RESPONSE_BOTTOM, + "Invalid state!"); + MOZ_ASSERT(mDataLength >= 5, + "SOCKS 5 connection reply must be at least 5 bytes!"); + + // Seek to the address location + mReadOffset = 3; + + *type = ReadUint8(); + + switch (*type) { + case 0x01: // ipv4 + *len = 4 - 1; + break; + case 0x04: // ipv6 + *len = 16 - 1; + break; + case 0x03: // fqdn + *len = ReadUint8(); + break; + default: // wrong address type + LOGERROR(("socks5: wrong address type in connection reply!")); + return PR_FAILURE; + } + + return PR_SUCCESS; +} + +PRStatus +nsSOCKSSocketInfo::ReadV5ConnectResponseTop() +{ + uint8_t res; + uint32_t len; + + MOZ_ASSERT(mState == SOCKS5_READ_CONNECT_RESPONSE_TOP, + "Invalid state!"); + MOZ_ASSERT(mDataLength == 5, + "SOCKS 5 connection reply must be exactly 5 bytes!"); + + LOGDEBUG(("socks5: checking connection reply")); + + // Check version number + if (ReadUint8() != 0x05) { + LOGERROR(("socks5: unexpected version in the reply")); + HandshakeFinished(PR_CONNECT_REFUSED_ERROR); + return PR_FAILURE; + } + + // Check response + res = ReadUint8(); + if (res != 0x00) { + PRErrorCode c = PR_CONNECT_REFUSED_ERROR; + + switch (res) { + case 0x01: + LOGERROR(("socks5: connect failed: " + "01, General SOCKS server failure.")); + break; + case 0x02: + LOGERROR(("socks5: connect failed: " + "02, Connection not allowed by ruleset.")); + break; + case 0x03: + LOGERROR(("socks5: connect failed: 03, Network unreachable.")); + c = PR_NETWORK_UNREACHABLE_ERROR; + break; + case 0x04: + LOGERROR(("socks5: connect failed: 04, Host unreachable.")); + break; + case 0x05: + LOGERROR(("socks5: connect failed: 05, Connection refused.")); + break; + case 0x06: + LOGERROR(("socks5: connect failed: 06, TTL expired.")); + c = PR_CONNECT_TIMEOUT_ERROR; + break; + case 0x07: + LOGERROR(("socks5: connect failed: " + "07, Command not supported.")); + break; + case 0x08: + LOGERROR(("socks5: connect failed: " + "08, Address type not supported.")); + c = PR_BAD_ADDRESS_ERROR; + break; + default: + LOGERROR(("socks5: connect failed.")); + break; + } + + HandshakeFinished(c); + return PR_FAILURE; + } + + if (ReadV5AddrTypeAndLength(&res, &len) != PR_SUCCESS) { + HandshakeFinished(PR_BAD_ADDRESS_ERROR); + return PR_FAILURE; + } + + mState = SOCKS5_READ_CONNECT_RESPONSE_BOTTOM; + WantRead(len + 2); + + return PR_SUCCESS; +} + +PRStatus +nsSOCKSSocketInfo::ReadV5ConnectResponseBottom() +{ + uint8_t type; + uint32_t len; + + MOZ_ASSERT(mState == SOCKS5_READ_CONNECT_RESPONSE_BOTTOM, + "Invalid state!"); + + if (ReadV5AddrTypeAndLength(&type, &len) != PR_SUCCESS) { + HandshakeFinished(PR_BAD_ADDRESS_ERROR); + return PR_FAILURE; + } + + MOZ_ASSERT(mDataLength == 7+len, + "SOCKS 5 unexpected length of connection reply!"); + + LOGDEBUG(("socks5: loading source addr and port")); + // Read what the proxy says is our source address + switch (type) { + case 0x01: // ipv4 + ReadNetAddr(&mExternalProxyAddr, AF_INET); + break; + case 0x04: // ipv6 + ReadNetAddr(&mExternalProxyAddr, AF_INET6); + break; + case 0x03: // fqdn (skip) + mReadOffset += len; + mExternalProxyAddr.raw.family = AF_INET; + break; + } + + ReadNetPort(&mExternalProxyAddr); + + LOGDEBUG(("socks5: connected!")); + HandshakeFinished(); + + return PR_SUCCESS; +} + +void +nsSOCKSSocketInfo::SetConnectTimeout(PRIntervalTime to) +{ + mTimeout = to; +} + +PRStatus +nsSOCKSSocketInfo::DoHandshake(PRFileDesc *fd, int16_t oflags) +{ + LOGDEBUG(("socks: DoHandshake(), state = %d", mState)); + + switch (mState) { + case SOCKS_INITIAL: + if (IsLocalProxy()) { + mState = SOCKS_DNS_COMPLETE; + mLookupStatus = NS_OK; + return ConnectToProxy(fd); + } + + return StartDNS(fd); + case SOCKS_DNS_IN_PROGRESS: + PR_SetError(PR_IN_PROGRESS_ERROR, 0); + return PR_FAILURE; + case SOCKS_DNS_COMPLETE: + return ConnectToProxy(fd); + case SOCKS_CONNECTING_TO_PROXY: + return ContinueConnectingToProxy(fd, oflags); + case SOCKS4_WRITE_CONNECT_REQUEST: + if (WriteToSocket(fd) != PR_SUCCESS) + return PR_FAILURE; + WantRead(8); + mState = SOCKS4_READ_CONNECT_RESPONSE; + return PR_SUCCESS; + case SOCKS4_READ_CONNECT_RESPONSE: + if (ReadFromSocket(fd) != PR_SUCCESS) + return PR_FAILURE; + return ReadV4ConnectResponse(); + + case SOCKS5_WRITE_AUTH_REQUEST: + if (WriteToSocket(fd) != PR_SUCCESS) + return PR_FAILURE; + WantRead(2); + mState = SOCKS5_READ_AUTH_RESPONSE; + return PR_SUCCESS; + case SOCKS5_READ_AUTH_RESPONSE: + if (ReadFromSocket(fd) != PR_SUCCESS) + return PR_FAILURE; + return ReadV5AuthResponse(); + case SOCKS5_WRITE_USERNAME_REQUEST: + if (WriteToSocket(fd) != PR_SUCCESS) + return PR_FAILURE; + WantRead(2); + mState = SOCKS5_READ_USERNAME_RESPONSE; + return PR_SUCCESS; + case SOCKS5_READ_USERNAME_RESPONSE: + if (ReadFromSocket(fd) != PR_SUCCESS) + return PR_FAILURE; + return ReadV5UsernameResponse(); + case SOCKS5_WRITE_CONNECT_REQUEST: + if (WriteToSocket(fd) != PR_SUCCESS) + return PR_FAILURE; + + // The SOCKS 5 response to the connection request is variable + // length. First, we'll read enough to tell how long the response + // is, and will read the rest later. + WantRead(5); + mState = SOCKS5_READ_CONNECT_RESPONSE_TOP; + return PR_SUCCESS; + case SOCKS5_READ_CONNECT_RESPONSE_TOP: + if (ReadFromSocket(fd) != PR_SUCCESS) + return PR_FAILURE; + return ReadV5ConnectResponseTop(); + case SOCKS5_READ_CONNECT_RESPONSE_BOTTOM: + if (ReadFromSocket(fd) != PR_SUCCESS) + return PR_FAILURE; + return ReadV5ConnectResponseBottom(); + + case SOCKS_CONNECTED: + LOGERROR(("socks: already connected")); + HandshakeFinished(PR_IS_CONNECTED_ERROR); + return PR_FAILURE; + case SOCKS_FAILED: + LOGERROR(("socks: already failed")); + return PR_FAILURE; + } + + LOGERROR(("socks: executing handshake in invalid state, %d", mState)); + HandshakeFinished(PR_INVALID_STATE_ERROR); + + return PR_FAILURE; +} + +int16_t +nsSOCKSSocketInfo::GetPollFlags() const +{ + switch (mState) { + case SOCKS_DNS_IN_PROGRESS: + case SOCKS_DNS_COMPLETE: + case SOCKS_CONNECTING_TO_PROXY: + return PR_POLL_EXCEPT | PR_POLL_WRITE; + case SOCKS4_WRITE_CONNECT_REQUEST: + case SOCKS5_WRITE_AUTH_REQUEST: + case SOCKS5_WRITE_USERNAME_REQUEST: + case SOCKS5_WRITE_CONNECT_REQUEST: + return PR_POLL_WRITE; + case SOCKS4_READ_CONNECT_RESPONSE: + case SOCKS5_READ_AUTH_RESPONSE: + case SOCKS5_READ_USERNAME_RESPONSE: + case SOCKS5_READ_CONNECT_RESPONSE_TOP: + case SOCKS5_READ_CONNECT_RESPONSE_BOTTOM: + return PR_POLL_READ; + default: + break; + } + + return 0; +} + +inline uint8_t +nsSOCKSSocketInfo::ReadUint8() +{ + uint8_t rv; + MOZ_ASSERT(mReadOffset + sizeof(rv) <= mDataLength, + "Not enough space to pop a uint8_t!"); + rv = mData[mReadOffset]; + mReadOffset += sizeof(rv); + return rv; +} + +inline uint16_t +nsSOCKSSocketInfo::ReadUint16() +{ + uint16_t rv; + MOZ_ASSERT(mReadOffset + sizeof(rv) <= mDataLength, + "Not enough space to pop a uint16_t!"); + memcpy(&rv, mData + mReadOffset, sizeof(rv)); + mReadOffset += sizeof(rv); + return rv; +} + +inline uint32_t +nsSOCKSSocketInfo::ReadUint32() +{ + uint32_t rv; + MOZ_ASSERT(mReadOffset + sizeof(rv) <= mDataLength, + "Not enough space to pop a uint32_t!"); + memcpy(&rv, mData + mReadOffset, sizeof(rv)); + mReadOffset += sizeof(rv); + return rv; +} + +void +nsSOCKSSocketInfo::ReadNetAddr(NetAddr *addr, uint16_t fam) +{ + uint32_t amt = 0; + const uint8_t *ip = mData + mReadOffset; + + addr->raw.family = fam; + if (fam == AF_INET) { + amt = sizeof(addr->inet.ip); + MOZ_ASSERT(mReadOffset + amt <= mDataLength, + "Not enough space to pop an ipv4 addr!"); + memcpy(&addr->inet.ip, ip, amt); + } else if (fam == AF_INET6) { + amt = sizeof(addr->inet6.ip.u8); + MOZ_ASSERT(mReadOffset + amt <= mDataLength, + "Not enough space to pop an ipv6 addr!"); + memcpy(addr->inet6.ip.u8, ip, amt); + } + + mReadOffset += amt; +} + +void +nsSOCKSSocketInfo::ReadNetPort(NetAddr *addr) +{ + addr->inet.port = ReadUint16(); +} + +void +nsSOCKSSocketInfo::WantRead(uint32_t sz) +{ + MOZ_ASSERT(mDataIoPtr == nullptr, + "WantRead() called while I/O already in progress!"); + MOZ_ASSERT(mDataLength + sz <= BUFFER_SIZE, + "Can't read that much data!"); + mAmountToRead = sz; +} + +PRStatus +nsSOCKSSocketInfo::ReadFromSocket(PRFileDesc *fd) +{ + int32_t rc; + const uint8_t *end; + + if (!mAmountToRead) { + LOGDEBUG(("socks: ReadFromSocket(), nothing to do")); + return PR_SUCCESS; + } + + if (!mDataIoPtr) { + mDataIoPtr = mData + mDataLength; + mDataLength += mAmountToRead; + } + + end = mData + mDataLength; + + while (mDataIoPtr < end) { + rc = PR_Read(fd, mDataIoPtr, end - mDataIoPtr); + if (rc <= 0) { + if (rc == 0) { + LOGERROR(("socks: proxy server closed connection")); + HandshakeFinished(PR_CONNECT_REFUSED_ERROR); + return PR_FAILURE; + } else if (PR_GetError() == PR_WOULD_BLOCK_ERROR) { + LOGDEBUG(("socks: ReadFromSocket(), want read")); + } + break; + } + + mDataIoPtr += rc; + } + + LOGDEBUG(("socks: ReadFromSocket(), have %u bytes total", + unsigned(mDataIoPtr - mData))); + if (mDataIoPtr == end) { + mDataIoPtr = nullptr; + mAmountToRead = 0; + mReadOffset = 0; + return PR_SUCCESS; + } + + return PR_FAILURE; +} + +PRStatus +nsSOCKSSocketInfo::WriteToSocket(PRFileDesc *fd) +{ + int32_t rc; + const uint8_t *end; + + if (!mDataLength) { + LOGDEBUG(("socks: WriteToSocket(), nothing to do")); + return PR_SUCCESS; + } + + if (!mDataIoPtr) + mDataIoPtr = mData; + + end = mData + mDataLength; + + while (mDataIoPtr < end) { + rc = PR_Write(fd, mDataIoPtr, end - mDataIoPtr); + if (rc < 0) { + if (PR_GetError() == PR_WOULD_BLOCK_ERROR) { + LOGDEBUG(("socks: WriteToSocket(), want write")); + } + break; + } + + mDataIoPtr += rc; + } + + if (mDataIoPtr == end) { + mDataIoPtr = nullptr; + mDataLength = 0; + mReadOffset = 0; + return PR_SUCCESS; + } + + return PR_FAILURE; +} + +static PRStatus +nsSOCKSIOLayerConnect(PRFileDesc *fd, const PRNetAddr *addr, PRIntervalTime to) +{ + PRStatus status; + NetAddr dst; + + nsSOCKSSocketInfo * info = (nsSOCKSSocketInfo*) fd->secret; + if (info == nullptr) return PR_FAILURE; + + if (addr->raw.family == PR_AF_INET6 && + PR_IsNetAddrType(addr, PR_IpAddrV4Mapped)) { + const uint8_t *srcp; + + LOGDEBUG(("socks: converting ipv4-mapped ipv6 address to ipv4")); + + // copied from _PR_ConvertToIpv4NetAddr() + dst.raw.family = AF_INET; + dst.inet.ip = htonl(INADDR_ANY); + dst.inet.port = htons(0); + srcp = addr->ipv6.ip.pr_s6_addr; + memcpy(&dst.inet.ip, srcp + 12, 4); + dst.inet.family = AF_INET; + dst.inet.port = addr->ipv6.port; + } else { + memcpy(&dst, addr, sizeof(dst)); + } + + info->SetDestinationAddr(&dst); + info->SetConnectTimeout(to); + + do { + status = info->DoHandshake(fd, -1); + } while (status == PR_SUCCESS && !info->IsConnected()); + + return status; +} + +static PRStatus +nsSOCKSIOLayerConnectContinue(PRFileDesc *fd, int16_t oflags) +{ + PRStatus status; + + nsSOCKSSocketInfo * info = (nsSOCKSSocketInfo*) fd->secret; + if (info == nullptr) return PR_FAILURE; + + do { + status = info->DoHandshake(fd, oflags); + } while (status == PR_SUCCESS && !info->IsConnected()); + + return status; +} + +static int16_t +nsSOCKSIOLayerPoll(PRFileDesc *fd, int16_t in_flags, int16_t *out_flags) +{ + nsSOCKSSocketInfo * info = (nsSOCKSSocketInfo*) fd->secret; + if (info == nullptr) return PR_FAILURE; + + if (!info->IsConnected()) { + *out_flags = 0; + return info->GetPollFlags(); + } + + return fd->lower->methods->poll(fd->lower, in_flags, out_flags); +} + +static PRStatus +nsSOCKSIOLayerClose(PRFileDesc *fd) +{ + nsSOCKSSocketInfo * info = (nsSOCKSSocketInfo*) fd->secret; + PRDescIdentity id = PR_GetLayersIdentity(fd); + + if (info && id == nsSOCKSIOLayerIdentity) + { + info->ForgetFD(); + 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 int32_t +nsSOCKSIOLayerAcceptRead(PRFileDesc *sd, PRFileDesc **nd, PRNetAddr **raddr, void *buf, int32_t 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 != nullptr && addr != nullptr) { + NetAddr temp; + NetAddr *tempPtr = &temp; + if (info->GetExternalProxyAddr(&tempPtr) == NS_OK) { + NetAddrToPRNetAddr(tempPtr, addr); + return PR_SUCCESS; + } + } + + return PR_FAILURE; +} + +static PRStatus +nsSOCKSIOLayerGetPeerName(PRFileDesc *fd, PRNetAddr *addr) +{ + nsSOCKSSocketInfo * info = (nsSOCKSSocketInfo*) fd->secret; + + if (info != nullptr && addr != nullptr) { + NetAddr temp; + NetAddr *tempPtr = &temp; + if (info->GetDestinationAddr(&tempPtr) == NS_OK) { + NetAddrToPRNetAddr(tempPtr, addr); + return PR_SUCCESS; + } + } + + return PR_FAILURE; +} + +static PRStatus +nsSOCKSIOLayerListen(PRFileDesc *fd, int 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(int32_t family, + const char *host, + int32_t port, + nsIProxyInfo *proxy, + int32_t socksVersion, + uint32_t flags, + PRFileDesc *fd, + nsISupports** info) +{ + NS_ENSURE_TRUE((socksVersion == 4) || (socksVersion == 5), NS_ERROR_NOT_INITIALIZED); + + + if (firstTime) + { + //XXX hack until NSPR provides an official way to detect system IPv6 + // support (bug 388519) + PRFileDesc *tmpfd = PR_OpenTCPSocket(PR_AF_INET6); + if (!tmpfd) { + ipv6Supported = false; + } else { + // If the system does not support IPv6, NSPR will push + // IPv6-to-IPv4 emulation layer onto the native layer + ipv6Supported = PR_GetIdentitiesLayer(tmpfd, PR_NSPR_IO_LAYER) == tmpfd; + PR_Close(tmpfd); + } + + nsSOCKSIOLayerIdentity = PR_GetUniqueIdentity("SOCKS layer"); + nsSOCKSIOLayerMethods = *PR_GetDefaultIOMethods(); + + nsSOCKSIOLayerMethods.connect = nsSOCKSIOLayerConnect; + nsSOCKSIOLayerMethods.connectcontinue = nsSOCKSIOLayerConnectContinue; + nsSOCKSIOLayerMethods.poll = nsSOCKSIOLayerPoll; + nsSOCKSIOLayerMethods.bind = nsSOCKSIOLayerBind; + nsSOCKSIOLayerMethods.acceptread = nsSOCKSIOLayerAcceptRead; + nsSOCKSIOLayerMethods.getsockname = nsSOCKSIOLayerGetName; + nsSOCKSIOLayerMethods.getpeername = nsSOCKSIOLayerGetPeerName; + nsSOCKSIOLayerMethods.accept = nsSOCKSIOLayerAccept; + nsSOCKSIOLayerMethods.listen = nsSOCKSIOLayerListen; + nsSOCKSIOLayerMethods.close = nsSOCKSIOLayerClose; + + firstTime = false; + } + + 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, family, proxy, host, flags); + layer->secret = (PRFilePrivate*) infoObject; + + PRDescIdentity fdIdentity = PR_GetLayersIdentity(fd); +#if defined(XP_WIN) + if (fdIdentity == mozilla::net::nsNamedPipeLayerIdentity) { + // remember named pipe fd on the info object so that we can switch + // blocking and non-blocking mode on the pipe later. + infoObject->SetNamedPipeFD(fd); + } +#endif + rv = PR_PushIOLayer(fd, fdIdentity, layer); + + if (rv == PR_FAILURE) { + LOGERROR(("PR_PushIOLayer() failed. rv = %x.", rv)); + NS_RELEASE(infoObject); + PR_DELETE(layer); + return NS_ERROR_FAILURE; + } + + *info = static_cast<nsISOCKSSocketInfo*>(infoObject); + NS_ADDREF(*info); + return NS_OK; +} + +bool +IsHostLocalTarget(const nsACString& aHost) +{ +#if defined(XP_UNIX) + return StringBeginsWith(aHost, NS_LITERAL_CSTRING("file:")); +#elif defined(XP_WIN) + return IsNamedPipePath(aHost); +#else + return false; +#endif // XP_UNIX +} |