curl/lib/vtls/schannel.c

2121 строка
72 KiB
C
Исходник Обычный вид История

/***************************************************************************
* _ _ ____ _
* Project ___| | | | _ \| |
* / __| | | | |_) | |
* | (__| |_| | _ <| |___
* \___|\___/|_| \_\_____|
*
2016-04-03 21:28:34 +03:00
* Copyright (C) 2012 - 2016, Marc Hoersken, <info@marc-hoersken.de>
* Copyright (C) 2012, Mark Salisbury, <mark.salisbury@hp.com>
* Copyright (C) 2012 - 2018, Daniel Stenberg, <daniel@haxx.se>, et al.
*
* This software is licensed as described in the file COPYING, which
* you should have received as part of this distribution. The terms
* are also available at https://curl.haxx.se/docs/copyright.html.
*
* You may opt to use, copy, modify, merge, publish, distribute and/or sell
* copies of the Software, and permit persons to whom the Software is
* furnished to do so, under the terms of the COPYING file.
*
* This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
* KIND, either express or implied.
*
***************************************************************************/
/*
* Source file for all SChannel-specific code for the TLS/SSL layer. No code
* but vtls.c should ever call or use these functions.
*
*/
/*
* Based upon the PolarSSL implementation in polarssl.c and polarssl.h:
* Copyright (C) 2010, 2011, Hoi-Ho Chan, <hoiho.chan@gmail.com>
*
* Based upon the CyaSSL implementation in cyassl.c and cyassl.h:
* Copyright (C) 1998 - 2012, Daniel Stenberg, <daniel@haxx.se>, et al.
*
* Thanks for code and inspiration!
*/
build: fix circular header inclusion with other packages This commit renames lib/setup.h to lib/curl_setup.h and renames lib/setup_once.h to lib/curl_setup_once.h. Removes the need and usage of a header inclusion guard foreign to libcurl. [1] Removes the need and presence of an alarming notice we carried in old setup_once.h [2] ---------------------------------------- 1 - lib/setup_once.h used __SETUP_ONCE_H macro as header inclusion guard up to commit ec691ca3 which changed this to HEADER_CURL_SETUP_ONCE_H, this single inclusion guard is enough to ensure that inclusion of lib/setup_once.h done from lib/setup.h is only done once. Additionally lib/setup.h has always used __SETUP_ONCE_H macro to protect inclusion of setup_once.h even after commit ec691ca3, this was to avoid a circular header inclusion triggered when building a c-ares enabled version with c-ares sources available which also has a setup_once.h header. Commit ec691ca3 exposes the real nature of __SETUP_ONCE_H usage in lib/setup.h, it is a header inclusion guard foreign to libcurl belonging to c-ares's setup_once.h The renaming this commit does, fixes the circular header inclusion, and as such removes the need and usage of a header inclusion guard foreign to libcurl. Macro __SETUP_ONCE_H no longer used in libcurl. 2 - Due to the circular interdependency of old lib/setup_once.h and the c-ares setup_once.h header, old file lib/setup_once.h has carried back from 2006 up to now days an alarming and prominent notice about the need of keeping libcurl's and c-ares's setup_once.h in sync. Given that this commit fixes the circular interdependency, the need and presence of mentioned notice is removed. All mentioned interdependencies come back from now old days when the c-ares project lived inside a curl subdirectory. This commit removes last traces of such fact.
2013-01-06 22:06:49 +04:00
#include "curl_setup.h"
#ifdef USE_SCHANNEL
#ifndef USE_WINDOWS_SSPI
# error "Can't compile SCHANNEL support without SSPI."
#endif
vtls: encapsulate SSL backend-specific data So far, all of the SSL backends' private data has been declared as part of the ssl_connect_data struct, in one big #if .. #elif .. #endif block. This can only work as long as the SSL backend is a compile-time option, something we want to change in the next commits. Therefore, let's encapsulate the exact data needed by each SSL backend into a private struct, and let's avoid bleeding any SSL backend-specific information into urldata.h. This is also necessary to allow multiple SSL backends to be compiled in at the same time, as e.g. OpenSSL's and CyaSSL's headers cannot be included in the same .c file. To avoid too many malloc() calls, we simply append the private structs to the connectdata struct in allocate_conn(). This requires us to take extra care of alignment issues: struct fields often need to be aligned on certain boundaries e.g. 32-bit values need to be stored at addresses that divide evenly by 4 (= 32 bit / 8 bit-per-byte). We do that by assuming that no SSL backend's private data contains any fields that need to be aligned on boundaries larger than `long long` (typically 64-bit) would need. Under this assumption, we simply add a dummy field of type `long long` to the `struct connectdata` struct. This field will never be accessed but acts as a placeholder for the four instances of ssl_backend_data instead. the size of each ssl_backend_data struct is stored in the SSL backend-specific metadata, to allow allocate_conn() to know how much extra space to allocate, and how to initialize the ssl[sockindex]->backend and proxy_ssl[sockindex]->backend pointers. This would appear to be a little complicated at first, but is really necessary to encapsulate the private data of each SSL backend correctly. And we need to encapsulate thusly if we ever want to allow selecting CyaSSL and OpenSSL at runtime, as their headers cannot be included within the same .c file (there are just too many conflicting definitions and declarations for that). Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
2017-07-28 23:09:35 +03:00
#include <schnlsp.h>
#include <schannel.h>
#include "curl_sspi.h"
#include "schannel.h"
2013-12-25 14:20:39 +04:00
#include "vtls.h"
#include "sendf.h"
#include "connect.h" /* for the connect timeout */
#include "strerror.h"
#include "select.h" /* for the socket readyness */
#include "inet_pton.h" /* for IP addr SNI check */
#include "curl_multibyte.h"
#include "warnless.h"
#include "x509asn1.h"
#include "curl_printf.h"
#include "system_win32.h"
#include "hostcheck.h"
/* The last #include file should be: */
#include "curl_memory.h"
#include "memdebug.h"
/* ALPN requires version 8.1 of the Windows SDK, which was
shipped with Visual Studio 2013, aka _MSC_VER 1800:
https://technet.microsoft.com/en-us/library/hh831771%28v=ws.11%29.aspx
*/
#if defined(_MSC_VER) && (_MSC_VER >= 1800) && !defined(_USING_V110_SDK71_)
# define HAS_ALPN 1
#endif
#ifndef UNISP_NAME_A
#define UNISP_NAME_A "Microsoft Unified Security Protocol Provider"
#endif
#ifndef UNISP_NAME_W
#define UNISP_NAME_W L"Microsoft Unified Security Protocol Provider"
#endif
#ifndef UNISP_NAME
#ifdef UNICODE
#define UNISP_NAME UNISP_NAME_W
#else
#define UNISP_NAME UNISP_NAME_A
#endif
#endif
#ifdef UNICODE
#define CURL_CERT_STORE_PROV_SYSTEM CERT_STORE_PROV_SYSTEM_W
#else
#define CURL_CERT_STORE_PROV_SYSTEM CERT_STORE_PROV_SYSTEM_A
#endif
#ifndef SP_PROT_SSL2_CLIENT
#define SP_PROT_SSL2_CLIENT 0x00000008
#endif
#ifndef SP_PROT_SSL3_CLIENT
#define SP_PROT_SSL3_CLIENT 0x00000008
#endif
#ifndef SP_PROT_TLS1_CLIENT
#define SP_PROT_TLS1_CLIENT 0x00000080
#endif
#ifndef SP_PROT_TLS1_0_CLIENT
#define SP_PROT_TLS1_0_CLIENT SP_PROT_TLS1_CLIENT
#endif
#ifndef SP_PROT_TLS1_1_CLIENT
#define SP_PROT_TLS1_1_CLIENT 0x00000200
#endif
#ifndef SP_PROT_TLS1_2_CLIENT
#define SP_PROT_TLS1_2_CLIENT 0x00000800
#endif
#ifndef SECBUFFER_ALERT
#define SECBUFFER_ALERT 17
#endif
/* Both schannel buffer sizes must be > 0 */
#define CURL_SCHANNEL_BUFFER_INIT_SIZE 4096
#define CURL_SCHANNEL_BUFFER_FREE_SIZE 1024
#define CERT_THUMBPRINT_STR_LEN 40
#define CERT_THUMBPRINT_DATA_LEN 20
/* Uncomment to force verbose output
* #define infof(x, y, ...) printf(y, __VA_ARGS__)
* #define failf(x, y, ...) printf(y, __VA_ARGS__)
*/
#ifndef CALG_SHA_256
# define CALG_SHA_256 0x0000800c
#endif
vtls: encapsulate SSL backend-specific data So far, all of the SSL backends' private data has been declared as part of the ssl_connect_data struct, in one big #if .. #elif .. #endif block. This can only work as long as the SSL backend is a compile-time option, something we want to change in the next commits. Therefore, let's encapsulate the exact data needed by each SSL backend into a private struct, and let's avoid bleeding any SSL backend-specific information into urldata.h. This is also necessary to allow multiple SSL backends to be compiled in at the same time, as e.g. OpenSSL's and CyaSSL's headers cannot be included in the same .c file. To avoid too many malloc() calls, we simply append the private structs to the connectdata struct in allocate_conn(). This requires us to take extra care of alignment issues: struct fields often need to be aligned on certain boundaries e.g. 32-bit values need to be stored at addresses that divide evenly by 4 (= 32 bit / 8 bit-per-byte). We do that by assuming that no SSL backend's private data contains any fields that need to be aligned on boundaries larger than `long long` (typically 64-bit) would need. Under this assumption, we simply add a dummy field of type `long long` to the `struct connectdata` struct. This field will never be accessed but acts as a placeholder for the four instances of ssl_backend_data instead. the size of each ssl_backend_data struct is stored in the SSL backend-specific metadata, to allow allocate_conn() to know how much extra space to allocate, and how to initialize the ssl[sockindex]->backend and proxy_ssl[sockindex]->backend pointers. This would appear to be a little complicated at first, but is really necessary to encapsulate the private data of each SSL backend correctly. And we need to encapsulate thusly if we ever want to allow selecting CyaSSL and OpenSSL at runtime, as their headers cannot be included within the same .c file (there are just too many conflicting definitions and declarations for that). Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
2017-07-28 23:09:35 +03:00
/* Structs to store Schannel handles */
struct curl_schannel_cred {
CredHandle cred_handle;
TimeStamp time_stamp;
int refcount;
};
struct curl_schannel_ctxt {
CtxtHandle ctxt_handle;
TimeStamp time_stamp;
};
struct ssl_backend_data {
struct curl_schannel_cred *cred;
struct curl_schannel_ctxt *ctxt;
SecPkgContext_StreamSizes stream_sizes;
size_t encdata_length, decdata_length;
size_t encdata_offset, decdata_offset;
unsigned char *encdata_buffer, *decdata_buffer;
/* encdata_is_incomplete: if encdata contains only a partial record that
can't be decrypted without another Curl_read_plain (that is, status is
SEC_E_INCOMPLETE_MESSAGE) then set this true. after Curl_read_plain writes
more bytes into encdata then set this back to false. */
bool encdata_is_incomplete;
unsigned long req_flags, ret_flags;
CURLcode recv_unrecoverable_err; /* schannel_recv had an unrecoverable err */
bool recv_sspi_close_notify; /* true if connection closed by close_notify */
bool recv_connection_closed; /* true if connection closed, regardless how */
bool use_alpn; /* true if ALPN is used for this connection */
};
#define BACKEND connssl->backend
static Curl_recv schannel_recv;
static Curl_send schannel_send;
static CURLcode pkp_pin_peer_pubkey(struct connectdata *conn, int sockindex,
const char *pinnedpubkey);
#ifdef _WIN32_WCE
static CURLcode verify_certificate(struct connectdata *conn, int sockindex);
#endif
static void InitSecBuffer(SecBuffer *buffer, unsigned long BufType,
void *BufDataPtr, unsigned long BufByteSize)
{
buffer->cbBuffer = BufByteSize;
buffer->BufferType = BufType;
buffer->pvBuffer = BufDataPtr;
}
static void InitSecBufferDesc(SecBufferDesc *desc, SecBuffer *BufArr,
unsigned long NumArrElem)
{
desc->ulVersion = SECBUFFER_VERSION;
desc->pBuffers = BufArr;
desc->cBuffers = NumArrElem;
}
static CURLcode
set_ssl_version_min_max(SCHANNEL_CRED *schannel_cred, struct connectdata *conn)
{
struct Curl_easy *data = conn->data;
long ssl_version = SSL_CONN_CONFIG(version);
long ssl_version_max = SSL_CONN_CONFIG(version_max);
long i = ssl_version;
switch(ssl_version_max) {
case CURL_SSLVERSION_MAX_NONE:
ssl_version_max = ssl_version << 16;
break;
case CURL_SSLVERSION_MAX_DEFAULT:
ssl_version_max = CURL_SSLVERSION_MAX_TLSv1_2;
break;
}
for(; i <= (ssl_version_max >> 16); ++i) {
switch(i) {
case CURL_SSLVERSION_TLSv1_0:
schannel_cred->grbitEnabledProtocols |= SP_PROT_TLS1_0_CLIENT;
break;
case CURL_SSLVERSION_TLSv1_1:
schannel_cred->grbitEnabledProtocols |= SP_PROT_TLS1_1_CLIENT;
break;
case CURL_SSLVERSION_TLSv1_2:
schannel_cred->grbitEnabledProtocols |= SP_PROT_TLS1_2_CLIENT;
break;
case CURL_SSLVERSION_TLSv1_3:
failf(data, "Schannel: TLS 1.3 is not yet supported");
return CURLE_SSL_CONNECT_ERROR;
}
}
return CURLE_OK;
}
static CURLcode
get_cert_location(TCHAR *path, DWORD *store_name, TCHAR **store_path,
TCHAR **thumbprint)
{
TCHAR *sep;
size_t store_name_len;
sep = _tcschr(path, TEXT('\\'));
if(sep == NULL)
return CURLE_SSL_CONNECT_ERROR;
store_name_len = sep - path;
if(_tcsnccmp(path, TEXT("CurrentUser"), store_name_len) == 0)
*store_name = CERT_SYSTEM_STORE_CURRENT_USER;
else if(_tcsnccmp(path, TEXT("LocalMachine"), store_name_len) == 0)
*store_name = CERT_SYSTEM_STORE_LOCAL_MACHINE;
else if(_tcsnccmp(path, TEXT("CurrentService"), store_name_len) == 0)
*store_name = CERT_SYSTEM_STORE_CURRENT_SERVICE;
else if(_tcsnccmp(path, TEXT("Services"), store_name_len) == 0)
*store_name = CERT_SYSTEM_STORE_SERVICES;
else if(_tcsnccmp(path, TEXT("Users"), store_name_len) == 0)
*store_name = CERT_SYSTEM_STORE_USERS;
else if(_tcsnccmp(path, TEXT("CurrentUserGroupPolicy"),
store_name_len) == 0)
*store_name = CERT_SYSTEM_STORE_CURRENT_USER_GROUP_POLICY;
else if(_tcsnccmp(path, TEXT("LocalMachineGroupPolicy"),
store_name_len) == 0)
*store_name = CERT_SYSTEM_STORE_LOCAL_MACHINE_GROUP_POLICY;
else if(_tcsnccmp(path, TEXT("LocalMachineEnterprise"),
store_name_len) == 0)
*store_name = CERT_SYSTEM_STORE_LOCAL_MACHINE_ENTERPRISE;
else
return CURLE_SSL_CONNECT_ERROR;
*store_path = sep + 1;
sep = _tcschr(*store_path, TEXT('\\'));
if(sep == NULL)
return CURLE_SSL_CONNECT_ERROR;
*sep = 0;
*thumbprint = sep + 1;
if(_tcslen(*thumbprint) != CERT_THUMBPRINT_STR_LEN)
return CURLE_SSL_CONNECT_ERROR;
return CURLE_OK;
}
static CURLcode
schannel_connect_step1(struct connectdata *conn, int sockindex)
{
ssize_t written = -1;
struct Curl_easy *data = conn->data;
struct ssl_connect_data *connssl = &conn->ssl[sockindex];
SecBuffer outbuf;
SecBufferDesc outbuf_desc;
SecBuffer inbuf;
SecBufferDesc inbuf_desc;
#ifdef HAS_ALPN
unsigned char alpn_buffer[128];
#endif
SCHANNEL_CRED schannel_cred;
PCCERT_CONTEXT client_certs[1] = { NULL };
SECURITY_STATUS sspi_status = SEC_E_OK;
struct curl_schannel_cred *old_cred = NULL;
struct in_addr addr;
#ifdef ENABLE_IPV6
struct in6_addr addr6;
#endif
TCHAR *host_name;
CURLcode result;
char * const hostname = SSL_IS_PROXY() ? conn->http_proxy.host.name :
conn->host.name;
infof(data, "schannel: SSL/TLS connection with %s port %hu (step 1/3)\n",
hostname, conn->remote_port);
if(Curl_verify_windows_version(5, 1, PLATFORM_WINNT,
VERSION_LESS_THAN_EQUAL)) {
/* SChannel in Windows XP (OS version 5.1) uses legacy handshakes and
algorithms that may not be supported by all servers. */
infof(data, "schannel: WinSSL version is old and may not be able to "
"connect to some servers due to lack of SNI, algorithms, etc.\n");
}
#ifdef HAS_ALPN
/* ALPN is only supported on Windows 8.1 / Server 2012 R2 and above.
Also it doesn't seem to be supported for Wine, see curl bug #983. */
BACKEND->use_alpn = conn->bits.tls_enable_alpn &&
!GetProcAddress(GetModuleHandleA("ntdll"),
"wine_get_version") &&
Curl_verify_windows_version(6, 3, PLATFORM_WINNT,
VERSION_GREATER_THAN_EQUAL);
#else
BACKEND->use_alpn = false;
#endif
BACKEND->cred = NULL;
/* check for an existing re-usable credential handle */
if(SSL_SET_OPTION(primary.sessionid)) {
Curl_ssl_sessionid_lock(conn);
proxy: Support HTTPS proxy and SOCKS+HTTP(s) * HTTPS proxies: An HTTPS proxy receives all transactions over an SSL/TLS connection. Once a secure connection with the proxy is established, the user agent uses the proxy as usual, including sending CONNECT requests to instruct the proxy to establish a [usually secure] TCP tunnel with an origin server. HTTPS proxies protect nearly all aspects of user-proxy communications as opposed to HTTP proxies that receive all requests (including CONNECT requests) in vulnerable clear text. With HTTPS proxies, it is possible to have two concurrent _nested_ SSL/TLS sessions: the "outer" one between the user agent and the proxy and the "inner" one between the user agent and the origin server (through the proxy). This change adds supports for such nested sessions as well. A secure connection with a proxy requires its own set of the usual SSL options (their actual descriptions differ and need polishing, see TODO): --proxy-cacert FILE CA certificate to verify peer against --proxy-capath DIR CA directory to verify peer against --proxy-cert CERT[:PASSWD] Client certificate file and password --proxy-cert-type TYPE Certificate file type (DER/PEM/ENG) --proxy-ciphers LIST SSL ciphers to use --proxy-crlfile FILE Get a CRL list in PEM format from the file --proxy-insecure Allow connections to proxies with bad certs --proxy-key KEY Private key file name --proxy-key-type TYPE Private key file type (DER/PEM/ENG) --proxy-pass PASS Pass phrase for the private key --proxy-ssl-allow-beast Allow security flaw to improve interop --proxy-sslv2 Use SSLv2 --proxy-sslv3 Use SSLv3 --proxy-tlsv1 Use TLSv1 --proxy-tlsuser USER TLS username --proxy-tlspassword STRING TLS password --proxy-tlsauthtype STRING TLS authentication type (default SRP) All --proxy-foo options are independent from their --foo counterparts, except --proxy-crlfile which defaults to --crlfile and --proxy-capath which defaults to --capath. Curl now also supports %{proxy_ssl_verify_result} --write-out variable, similar to the existing %{ssl_verify_result} variable. Supported backends: OpenSSL, GnuTLS, and NSS. * A SOCKS proxy + HTTP/HTTPS proxy combination: If both --socks* and --proxy options are given, Curl first connects to the SOCKS proxy and then connects (through SOCKS) to the HTTP or HTTPS proxy. TODO: Update documentation for the new APIs and --proxy-* options. Look for "Added in 7.XXX" marks.
2016-11-16 20:49:15 +03:00
if(!Curl_ssl_getsessionid(conn, (void **)&old_cred, NULL, sockindex)) {
BACKEND->cred = old_cred;
infof(data, "schannel: re-using existing credential handle\n");
/* increment the reference counter of the credential/session handle */
BACKEND->cred->refcount++;
infof(data, "schannel: incremented credential handle refcount = %d\n",
BACKEND->cred->refcount);
}
Curl_ssl_sessionid_unlock(conn);
}
if(!BACKEND->cred) {
/* setup Schannel API options */
memset(&schannel_cred, 0, sizeof(schannel_cred));
schannel_cred.dwVersion = SCHANNEL_CRED_VERSION;
proxy: Support HTTPS proxy and SOCKS+HTTP(s) * HTTPS proxies: An HTTPS proxy receives all transactions over an SSL/TLS connection. Once a secure connection with the proxy is established, the user agent uses the proxy as usual, including sending CONNECT requests to instruct the proxy to establish a [usually secure] TCP tunnel with an origin server. HTTPS proxies protect nearly all aspects of user-proxy communications as opposed to HTTP proxies that receive all requests (including CONNECT requests) in vulnerable clear text. With HTTPS proxies, it is possible to have two concurrent _nested_ SSL/TLS sessions: the "outer" one between the user agent and the proxy and the "inner" one between the user agent and the origin server (through the proxy). This change adds supports for such nested sessions as well. A secure connection with a proxy requires its own set of the usual SSL options (their actual descriptions differ and need polishing, see TODO): --proxy-cacert FILE CA certificate to verify peer against --proxy-capath DIR CA directory to verify peer against --proxy-cert CERT[:PASSWD] Client certificate file and password --proxy-cert-type TYPE Certificate file type (DER/PEM/ENG) --proxy-ciphers LIST SSL ciphers to use --proxy-crlfile FILE Get a CRL list in PEM format from the file --proxy-insecure Allow connections to proxies with bad certs --proxy-key KEY Private key file name --proxy-key-type TYPE Private key file type (DER/PEM/ENG) --proxy-pass PASS Pass phrase for the private key --proxy-ssl-allow-beast Allow security flaw to improve interop --proxy-sslv2 Use SSLv2 --proxy-sslv3 Use SSLv3 --proxy-tlsv1 Use TLSv1 --proxy-tlsuser USER TLS username --proxy-tlspassword STRING TLS password --proxy-tlsauthtype STRING TLS authentication type (default SRP) All --proxy-foo options are independent from their --foo counterparts, except --proxy-crlfile which defaults to --crlfile and --proxy-capath which defaults to --capath. Curl now also supports %{proxy_ssl_verify_result} --write-out variable, similar to the existing %{ssl_verify_result} variable. Supported backends: OpenSSL, GnuTLS, and NSS. * A SOCKS proxy + HTTP/HTTPS proxy combination: If both --socks* and --proxy options are given, Curl first connects to the SOCKS proxy and then connects (through SOCKS) to the HTTP or HTTPS proxy. TODO: Update documentation for the new APIs and --proxy-* options. Look for "Added in 7.XXX" marks.
2016-11-16 20:49:15 +03:00
if(conn->ssl_config.verifypeer) {
#ifdef _WIN32_WCE
/* certificate validation on CE doesn't seem to work right; we'll
do it following a more manual process. */
schannel_cred.dwFlags = SCH_CRED_MANUAL_CRED_VALIDATION |
SCH_CRED_IGNORE_NO_REVOCATION_CHECK |
SCH_CRED_IGNORE_REVOCATION_OFFLINE;
#else
schannel_cred.dwFlags = SCH_CRED_AUTO_CRED_VALIDATION;
proxy: Support HTTPS proxy and SOCKS+HTTP(s) * HTTPS proxies: An HTTPS proxy receives all transactions over an SSL/TLS connection. Once a secure connection with the proxy is established, the user agent uses the proxy as usual, including sending CONNECT requests to instruct the proxy to establish a [usually secure] TCP tunnel with an origin server. HTTPS proxies protect nearly all aspects of user-proxy communications as opposed to HTTP proxies that receive all requests (including CONNECT requests) in vulnerable clear text. With HTTPS proxies, it is possible to have two concurrent _nested_ SSL/TLS sessions: the "outer" one between the user agent and the proxy and the "inner" one between the user agent and the origin server (through the proxy). This change adds supports for such nested sessions as well. A secure connection with a proxy requires its own set of the usual SSL options (their actual descriptions differ and need polishing, see TODO): --proxy-cacert FILE CA certificate to verify peer against --proxy-capath DIR CA directory to verify peer against --proxy-cert CERT[:PASSWD] Client certificate file and password --proxy-cert-type TYPE Certificate file type (DER/PEM/ENG) --proxy-ciphers LIST SSL ciphers to use --proxy-crlfile FILE Get a CRL list in PEM format from the file --proxy-insecure Allow connections to proxies with bad certs --proxy-key KEY Private key file name --proxy-key-type TYPE Private key file type (DER/PEM/ENG) --proxy-pass PASS Pass phrase for the private key --proxy-ssl-allow-beast Allow security flaw to improve interop --proxy-sslv2 Use SSLv2 --proxy-sslv3 Use SSLv3 --proxy-tlsv1 Use TLSv1 --proxy-tlsuser USER TLS username --proxy-tlspassword STRING TLS password --proxy-tlsauthtype STRING TLS authentication type (default SRP) All --proxy-foo options are independent from their --foo counterparts, except --proxy-crlfile which defaults to --crlfile and --proxy-capath which defaults to --capath. Curl now also supports %{proxy_ssl_verify_result} --write-out variable, similar to the existing %{ssl_verify_result} variable. Supported backends: OpenSSL, GnuTLS, and NSS. * A SOCKS proxy + HTTP/HTTPS proxy combination: If both --socks* and --proxy options are given, Curl first connects to the SOCKS proxy and then connects (through SOCKS) to the HTTP or HTTPS proxy. TODO: Update documentation for the new APIs and --proxy-* options. Look for "Added in 7.XXX" marks.
2016-11-16 20:49:15 +03:00
/* TODO s/data->set.ssl.no_revoke/SSL_SET_OPTION(no_revoke)/g */
if(data->set.ssl.no_revoke)
schannel_cred.dwFlags |= SCH_CRED_IGNORE_NO_REVOCATION_CHECK |
SCH_CRED_IGNORE_REVOCATION_OFFLINE;
else
schannel_cred.dwFlags |= SCH_CRED_REVOCATION_CHECK_CHAIN;
#endif
proxy: Support HTTPS proxy and SOCKS+HTTP(s) * HTTPS proxies: An HTTPS proxy receives all transactions over an SSL/TLS connection. Once a secure connection with the proxy is established, the user agent uses the proxy as usual, including sending CONNECT requests to instruct the proxy to establish a [usually secure] TCP tunnel with an origin server. HTTPS proxies protect nearly all aspects of user-proxy communications as opposed to HTTP proxies that receive all requests (including CONNECT requests) in vulnerable clear text. With HTTPS proxies, it is possible to have two concurrent _nested_ SSL/TLS sessions: the "outer" one between the user agent and the proxy and the "inner" one between the user agent and the origin server (through the proxy). This change adds supports for such nested sessions as well. A secure connection with a proxy requires its own set of the usual SSL options (their actual descriptions differ and need polishing, see TODO): --proxy-cacert FILE CA certificate to verify peer against --proxy-capath DIR CA directory to verify peer against --proxy-cert CERT[:PASSWD] Client certificate file and password --proxy-cert-type TYPE Certificate file type (DER/PEM/ENG) --proxy-ciphers LIST SSL ciphers to use --proxy-crlfile FILE Get a CRL list in PEM format from the file --proxy-insecure Allow connections to proxies with bad certs --proxy-key KEY Private key file name --proxy-key-type TYPE Private key file type (DER/PEM/ENG) --proxy-pass PASS Pass phrase for the private key --proxy-ssl-allow-beast Allow security flaw to improve interop --proxy-sslv2 Use SSLv2 --proxy-sslv3 Use SSLv3 --proxy-tlsv1 Use TLSv1 --proxy-tlsuser USER TLS username --proxy-tlspassword STRING TLS password --proxy-tlsauthtype STRING TLS authentication type (default SRP) All --proxy-foo options are independent from their --foo counterparts, except --proxy-crlfile which defaults to --crlfile and --proxy-capath which defaults to --capath. Curl now also supports %{proxy_ssl_verify_result} --write-out variable, similar to the existing %{ssl_verify_result} variable. Supported backends: OpenSSL, GnuTLS, and NSS. * A SOCKS proxy + HTTP/HTTPS proxy combination: If both --socks* and --proxy options are given, Curl first connects to the SOCKS proxy and then connects (through SOCKS) to the HTTP or HTTPS proxy. TODO: Update documentation for the new APIs and --proxy-* options. Look for "Added in 7.XXX" marks.
2016-11-16 20:49:15 +03:00
if(data->set.ssl.no_revoke)
infof(data, "schannel: disabled server certificate revocation "
"checks\n");
else
infof(data, "schannel: checking server certificate revocation\n");
}
else {
schannel_cred.dwFlags = SCH_CRED_MANUAL_CRED_VALIDATION |
SCH_CRED_IGNORE_NO_REVOCATION_CHECK |
SCH_CRED_IGNORE_REVOCATION_OFFLINE;
infof(data, "schannel: disabled server certificate revocation checks\n");
}
proxy: Support HTTPS proxy and SOCKS+HTTP(s) * HTTPS proxies: An HTTPS proxy receives all transactions over an SSL/TLS connection. Once a secure connection with the proxy is established, the user agent uses the proxy as usual, including sending CONNECT requests to instruct the proxy to establish a [usually secure] TCP tunnel with an origin server. HTTPS proxies protect nearly all aspects of user-proxy communications as opposed to HTTP proxies that receive all requests (including CONNECT requests) in vulnerable clear text. With HTTPS proxies, it is possible to have two concurrent _nested_ SSL/TLS sessions: the "outer" one between the user agent and the proxy and the "inner" one between the user agent and the origin server (through the proxy). This change adds supports for such nested sessions as well. A secure connection with a proxy requires its own set of the usual SSL options (their actual descriptions differ and need polishing, see TODO): --proxy-cacert FILE CA certificate to verify peer against --proxy-capath DIR CA directory to verify peer against --proxy-cert CERT[:PASSWD] Client certificate file and password --proxy-cert-type TYPE Certificate file type (DER/PEM/ENG) --proxy-ciphers LIST SSL ciphers to use --proxy-crlfile FILE Get a CRL list in PEM format from the file --proxy-insecure Allow connections to proxies with bad certs --proxy-key KEY Private key file name --proxy-key-type TYPE Private key file type (DER/PEM/ENG) --proxy-pass PASS Pass phrase for the private key --proxy-ssl-allow-beast Allow security flaw to improve interop --proxy-sslv2 Use SSLv2 --proxy-sslv3 Use SSLv3 --proxy-tlsv1 Use TLSv1 --proxy-tlsuser USER TLS username --proxy-tlspassword STRING TLS password --proxy-tlsauthtype STRING TLS authentication type (default SRP) All --proxy-foo options are independent from their --foo counterparts, except --proxy-crlfile which defaults to --crlfile and --proxy-capath which defaults to --capath. Curl now also supports %{proxy_ssl_verify_result} --write-out variable, similar to the existing %{ssl_verify_result} variable. Supported backends: OpenSSL, GnuTLS, and NSS. * A SOCKS proxy + HTTP/HTTPS proxy combination: If both --socks* and --proxy options are given, Curl first connects to the SOCKS proxy and then connects (through SOCKS) to the HTTP or HTTPS proxy. TODO: Update documentation for the new APIs and --proxy-* options. Look for "Added in 7.XXX" marks.
2016-11-16 20:49:15 +03:00
if(!conn->ssl_config.verifyhost) {
SSL: Several SSL-backend related fixes axTLS: This will make the axTLS backend perform the RFC2818 checks, honoring the VERIFYHOST setting similar to the OpenSSL backend. Generic for OpenSSL and axTLS: Move the hostcheck and cert_hostcheck functions from the lib/ssluse.c files to make them genericly available for both the OpenSSL, axTLS and other SSL backends. They are now in the new lib/hostcheck.c file. CyaSSL: CyaSSL now also has the RFC2818 checks enabled by default. There is a limitation that the verifyhost can not be enabled exclusively on the Subject CN field comparison. This SSL backend will thus behave like the NSS and the GnuTLS (meaning: RFC2818 ok, or bust). In other words: setting verifyhost to 0 or 1 will disable the Subject Alt Names checks too. Schannel: Updated the schannel information messages: Split the IP address usage message from the verifyhost setting and changed the message about disabling SNI (Server Name Indication, used in HTTP virtual hosting) into a message stating that the Subject Alternative Names checks are being disabled when verifyhost is set to 0 or 1. As a side effect of switching off the RFC2818 related servername checks with SCH_CRED_NO_SERVERNAME_CHECK (http://msdn.microsoft.com/en-us/library/aa923430.aspx) the SNI feature is being disabled. This effect is not documented in MSDN, but Wireshark output clearly shows the effect (details on the libcurl maillist). PolarSSL: Fix the prototype change in PolarSSL of ssl_set_session() and the move of the peer_cert from the ssl_context to the ssl_session. Found this change in the PolarSSL SVN between r1316 and r1317 where the POLARSSL_VERSION_NUMBER was at 0x01010100. But to accommodate the Ubuntu PolarSSL version 1.1.4 the check is to discriminate between lower then PolarSSL version 1.2.0 and 1.2.0 and higher. Note: The PolarSSL SVN trunk jumped from version 1.1.1 to 1.2.0. Generic: All the SSL backends are fixed and checked to work with the ssl.verifyhost as a boolean, which is an internal API change.
2012-11-03 05:06:51 +04:00
schannel_cred.dwFlags |= SCH_CRED_NO_SERVERNAME_CHECK;
infof(data, "schannel: verifyhost setting prevents Schannel from "
"comparing the supplied target name with the subject "
"names in server certificates.\n");
}
proxy: Support HTTPS proxy and SOCKS+HTTP(s) * HTTPS proxies: An HTTPS proxy receives all transactions over an SSL/TLS connection. Once a secure connection with the proxy is established, the user agent uses the proxy as usual, including sending CONNECT requests to instruct the proxy to establish a [usually secure] TCP tunnel with an origin server. HTTPS proxies protect nearly all aspects of user-proxy communications as opposed to HTTP proxies that receive all requests (including CONNECT requests) in vulnerable clear text. With HTTPS proxies, it is possible to have two concurrent _nested_ SSL/TLS sessions: the "outer" one between the user agent and the proxy and the "inner" one between the user agent and the origin server (through the proxy). This change adds supports for such nested sessions as well. A secure connection with a proxy requires its own set of the usual SSL options (their actual descriptions differ and need polishing, see TODO): --proxy-cacert FILE CA certificate to verify peer against --proxy-capath DIR CA directory to verify peer against --proxy-cert CERT[:PASSWD] Client certificate file and password --proxy-cert-type TYPE Certificate file type (DER/PEM/ENG) --proxy-ciphers LIST SSL ciphers to use --proxy-crlfile FILE Get a CRL list in PEM format from the file --proxy-insecure Allow connections to proxies with bad certs --proxy-key KEY Private key file name --proxy-key-type TYPE Private key file type (DER/PEM/ENG) --proxy-pass PASS Pass phrase for the private key --proxy-ssl-allow-beast Allow security flaw to improve interop --proxy-sslv2 Use SSLv2 --proxy-sslv3 Use SSLv3 --proxy-tlsv1 Use TLSv1 --proxy-tlsuser USER TLS username --proxy-tlspassword STRING TLS password --proxy-tlsauthtype STRING TLS authentication type (default SRP) All --proxy-foo options are independent from their --foo counterparts, except --proxy-crlfile which defaults to --crlfile and --proxy-capath which defaults to --capath. Curl now also supports %{proxy_ssl_verify_result} --write-out variable, similar to the existing %{ssl_verify_result} variable. Supported backends: OpenSSL, GnuTLS, and NSS. * A SOCKS proxy + HTTP/HTTPS proxy combination: If both --socks* and --proxy options are given, Curl first connects to the SOCKS proxy and then connects (through SOCKS) to the HTTP or HTTPS proxy. TODO: Update documentation for the new APIs and --proxy-* options. Look for "Added in 7.XXX" marks.
2016-11-16 20:49:15 +03:00
switch(conn->ssl_config.version) {
case CURL_SSLVERSION_DEFAULT:
case CURL_SSLVERSION_TLSv1:
schannel_cred.grbitEnabledProtocols = SP_PROT_TLS1_0_CLIENT |
SP_PROT_TLS1_1_CLIENT |
SP_PROT_TLS1_2_CLIENT;
break;
case CURL_SSLVERSION_TLSv1_0:
case CURL_SSLVERSION_TLSv1_1:
case CURL_SSLVERSION_TLSv1_2:
case CURL_SSLVERSION_TLSv1_3:
{
result = set_ssl_version_min_max(&schannel_cred, conn);
if(result != CURLE_OK)
return result;
break;
}
case CURL_SSLVERSION_SSLv3:
schannel_cred.grbitEnabledProtocols = SP_PROT_SSL3_CLIENT;
break;
case CURL_SSLVERSION_SSLv2:
schannel_cred.grbitEnabledProtocols = SP_PROT_SSL2_CLIENT;
break;
default:
failf(data, "Unrecognized parameter passed via CURLOPT_SSLVERSION");
return CURLE_SSL_CONNECT_ERROR;
}
/* client certificate */
if(data->set.ssl.cert) {
DWORD cert_store_name;
TCHAR *cert_store_path;
TCHAR *cert_thumbprint_str;
CRYPT_HASH_BLOB cert_thumbprint;
BYTE cert_thumbprint_data[CERT_THUMBPRINT_DATA_LEN];
HCERTSTORE cert_store;
TCHAR *cert_path = Curl_convert_UTF8_to_tchar(data->set.ssl.cert);
if(!cert_path)
return CURLE_OUT_OF_MEMORY;
result = get_cert_location(cert_path, &cert_store_name,
&cert_store_path, &cert_thumbprint_str);
if(result != CURLE_OK) {
Curl_unicodefree(cert_path);
return result;
}
cert_store = CertOpenStore(CURL_CERT_STORE_PROV_SYSTEM, 0,
(HCRYPTPROV)NULL,
cert_store_name, cert_store_path);
if(!cert_store) {
Curl_unicodefree(cert_path);
return CURLE_SSL_CONNECT_ERROR;
}
cert_thumbprint.pbData = cert_thumbprint_data;
cert_thumbprint.cbData = CERT_THUMBPRINT_DATA_LEN;
if(!CryptStringToBinary(cert_thumbprint_str, CERT_THUMBPRINT_STR_LEN,
CRYPT_STRING_HEXRAW,
cert_thumbprint_data, &cert_thumbprint.cbData,
NULL, NULL)) {
Curl_unicodefree(cert_path);
return CURLE_SSL_CONNECT_ERROR;
}
client_certs[0] = CertFindCertificateInStore(
cert_store, X509_ASN_ENCODING | PKCS_7_ASN_ENCODING, 0,
CERT_FIND_HASH, &cert_thumbprint, NULL);
Curl_unicodefree(cert_path);
if(client_certs[0]) {
schannel_cred.cCreds = 1;
schannel_cred.paCred = client_certs;
}
CertCloseStore(cert_store, 0);
}
/* allocate memory for the re-usable credential handle */
BACKEND->cred = (struct curl_schannel_cred *)
calloc(1, sizeof(struct curl_schannel_cred));
if(!BACKEND->cred) {
failf(data, "schannel: unable to allocate memory");
if(client_certs[0])
CertFreeCertificateContext(client_certs[0]);
return CURLE_OUT_OF_MEMORY;
}
BACKEND->cred->refcount = 1;
2016-02-03 07:09:25 +03:00
/* https://msdn.microsoft.com/en-us/library/windows/desktop/aa374716.aspx
*/
sspi_status =
s_pSecFn->AcquireCredentialsHandle(NULL, (TCHAR *)UNISP_NAME,
SECPKG_CRED_OUTBOUND, NULL,
&schannel_cred, NULL, NULL,
&BACKEND->cred->cred_handle,
&BACKEND->cred->time_stamp);
if(client_certs[0])
CertFreeCertificateContext(client_certs[0]);
if(sspi_status != SEC_E_OK) {
if(sspi_status == SEC_E_WRONG_PRINCIPAL)
failf(data, "schannel: SNI or certificate check failed: %s",
Curl_sspi_strerror(conn, sspi_status));
else
failf(data, "schannel: AcquireCredentialsHandle failed: %s",
Curl_sspi_strerror(conn, sspi_status));
Curl_safefree(BACKEND->cred);
return CURLE_SSL_CONNECT_ERROR;
}
}
/* Warn if SNI is disabled due to use of an IP address */
if(Curl_inet_pton(AF_INET, hostname, &addr)
#ifdef ENABLE_IPV6
|| Curl_inet_pton(AF_INET6, hostname, &addr6)
#endif
) {
infof(data, "schannel: using IP address, SNI is not supported by OS.\n");
}
#ifdef HAS_ALPN
if(BACKEND->use_alpn) {
int cur = 0;
int list_start_index = 0;
unsigned int *extension_len = NULL;
unsigned short* list_len = NULL;
/* The first four bytes will be an unsigned int indicating number
of bytes of data in the rest of the the buffer. */
extension_len = (unsigned int *)(&alpn_buffer[cur]);
cur += sizeof(unsigned int);
/* The next four bytes are an indicator that this buffer will contain
ALPN data, as opposed to NPN, for example. */
*(unsigned int *)&alpn_buffer[cur] =
SecApplicationProtocolNegotiationExt_ALPN;
cur += sizeof(unsigned int);
/* The next two bytes will be an unsigned short indicating the number
of bytes used to list the preferred protocols. */
list_len = (unsigned short*)(&alpn_buffer[cur]);
cur += sizeof(unsigned short);
list_start_index = cur;
#ifdef USE_NGHTTP2
if(data->set.httpversion >= CURL_HTTP_VERSION_2) {
memcpy(&alpn_buffer[cur], NGHTTP2_PROTO_ALPN, NGHTTP2_PROTO_ALPN_LEN);
cur += NGHTTP2_PROTO_ALPN_LEN;
infof(data, "schannel: ALPN, offering %s\n", NGHTTP2_PROTO_VERSION_ID);
}
#endif
alpn_buffer[cur++] = ALPN_HTTP_1_1_LENGTH;
memcpy(&alpn_buffer[cur], ALPN_HTTP_1_1, ALPN_HTTP_1_1_LENGTH);
cur += ALPN_HTTP_1_1_LENGTH;
infof(data, "schannel: ALPN, offering %s\n", ALPN_HTTP_1_1);
*list_len = curlx_uitous(cur - list_start_index);
*extension_len = *list_len + sizeof(unsigned int) + sizeof(unsigned short);
InitSecBuffer(&inbuf, SECBUFFER_APPLICATION_PROTOCOLS, alpn_buffer, cur);
InitSecBufferDesc(&inbuf_desc, &inbuf, 1);
}
else
{
InitSecBuffer(&inbuf, SECBUFFER_EMPTY, NULL, 0);
InitSecBufferDesc(&inbuf_desc, &inbuf, 1);
}
#else /* HAS_ALPN */
InitSecBuffer(&inbuf, SECBUFFER_EMPTY, NULL, 0);
InitSecBufferDesc(&inbuf_desc, &inbuf, 1);
#endif
/* setup output buffer */
InitSecBuffer(&outbuf, SECBUFFER_EMPTY, NULL, 0);
InitSecBufferDesc(&outbuf_desc, &outbuf, 1);
/* setup request flags */
BACKEND->req_flags = ISC_REQ_SEQUENCE_DETECT | ISC_REQ_REPLAY_DETECT |
ISC_REQ_CONFIDENTIALITY | ISC_REQ_ALLOCATE_MEMORY |
ISC_REQ_STREAM;
/* allocate memory for the security context handle */
BACKEND->ctxt = (struct curl_schannel_ctxt *)
calloc(1, sizeof(struct curl_schannel_ctxt));
if(!BACKEND->ctxt) {
failf(data, "schannel: unable to allocate memory");
return CURLE_OUT_OF_MEMORY;
}
host_name = Curl_convert_UTF8_to_tchar(hostname);
if(!host_name)
return CURLE_OUT_OF_MEMORY;
/* Schannel InitializeSecurityContext:
https://msdn.microsoft.com/en-us/library/windows/desktop/aa375924.aspx
At the moment we don't pass inbuf unless we're using ALPN since we only
use it for that, and Wine (for which we currently disable ALPN) is giving
us problems with inbuf regardless. https://github.com/curl/curl/issues/983
*/
sspi_status = s_pSecFn->InitializeSecurityContext(
&BACKEND->cred->cred_handle, NULL, host_name, BACKEND->req_flags, 0, 0,
(BACKEND->use_alpn ? &inbuf_desc : NULL),
0, &BACKEND->ctxt->ctxt_handle,
&outbuf_desc, &BACKEND->ret_flags, &BACKEND->ctxt->time_stamp);
Curl_unicodefree(host_name);
if(sspi_status != SEC_I_CONTINUE_NEEDED) {
if(sspi_status == SEC_E_WRONG_PRINCIPAL)
failf(data, "schannel: SNI or certificate check failed: %s",
Curl_sspi_strerror(conn, sspi_status));
else
failf(data, "schannel: initial InitializeSecurityContext failed: %s",
Curl_sspi_strerror(conn, sspi_status));
Curl_safefree(BACKEND->ctxt);
return CURLE_SSL_CONNECT_ERROR;
}
infof(data, "schannel: sending initial handshake data: "
"sending %lu bytes...\n", outbuf.cbBuffer);
/* send initial handshake data which is now stored in output buffer */
result = Curl_write_plain(conn, conn->sock[sockindex], outbuf.pvBuffer,
outbuf.cbBuffer, &written);
s_pSecFn->FreeContextBuffer(outbuf.pvBuffer);
if((result != CURLE_OK) || (outbuf.cbBuffer != (size_t) written)) {
failf(data, "schannel: failed to send initial handshake data: "
"sent %zd of %lu bytes", written, outbuf.cbBuffer);
return CURLE_SSL_CONNECT_ERROR;
}
infof(data, "schannel: sent initial handshake data: "
"sent %zd bytes\n", written);
BACKEND->recv_unrecoverable_err = CURLE_OK;
BACKEND->recv_sspi_close_notify = false;
BACKEND->recv_connection_closed = false;
BACKEND->encdata_is_incomplete = false;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
/* continue to second handshake step */
connssl->connecting_state = ssl_connect_2;
return CURLE_OK;
}
static CURLcode
schannel_connect_step2(struct connectdata *conn, int sockindex)
{
int i;
ssize_t nread = -1, written = -1;
struct Curl_easy *data = conn->data;
struct ssl_connect_data *connssl = &conn->ssl[sockindex];
unsigned char *reallocated_buffer;
size_t reallocated_length;
SecBuffer outbuf[3];
SecBufferDesc outbuf_desc;
SecBuffer inbuf[2];
SecBufferDesc inbuf_desc;
SECURITY_STATUS sspi_status = SEC_E_OK;
TCHAR *host_name;
CURLcode result;
2012-06-21 00:26:51 +04:00
bool doread;
char * const hostname = SSL_IS_PROXY() ? conn->http_proxy.host.name :
conn->host.name;
const char *pubkey_ptr;
2012-06-21 00:26:51 +04:00
doread = (connssl->connecting_state != ssl_connect_2_writing) ? TRUE : FALSE;
infof(data, "schannel: SSL/TLS connection with %s port %hu (step 2/3)\n",
hostname, conn->remote_port);
if(!BACKEND->cred || !BACKEND->ctxt)
return CURLE_SSL_CONNECT_ERROR;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
/* buffer to store previously received and decrypted data */
if(BACKEND->decdata_buffer == NULL) {
BACKEND->decdata_offset = 0;
BACKEND->decdata_length = CURL_SCHANNEL_BUFFER_INIT_SIZE;
BACKEND->decdata_buffer = malloc(BACKEND->decdata_length);
if(BACKEND->decdata_buffer == NULL) {
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
failf(data, "schannel: unable to allocate memory");
return CURLE_OUT_OF_MEMORY;
}
}
/* buffer to store previously received and encrypted data */
if(BACKEND->encdata_buffer == NULL) {
BACKEND->encdata_is_incomplete = false;
BACKEND->encdata_offset = 0;
BACKEND->encdata_length = CURL_SCHANNEL_BUFFER_INIT_SIZE;
BACKEND->encdata_buffer = malloc(BACKEND->encdata_length);
if(BACKEND->encdata_buffer == NULL) {
failf(data, "schannel: unable to allocate memory");
return CURLE_OUT_OF_MEMORY;
}
}
/* if we need a bigger buffer to read a full message, increase buffer now */
if(BACKEND->encdata_length - BACKEND->encdata_offset <
CURL_SCHANNEL_BUFFER_FREE_SIZE) {
/* increase internal encrypted data buffer */
reallocated_length = BACKEND->encdata_offset +
CURL_SCHANNEL_BUFFER_FREE_SIZE;
reallocated_buffer = realloc(BACKEND->encdata_buffer,
reallocated_length);
if(reallocated_buffer == NULL) {
failf(data, "schannel: unable to re-allocate memory");
return CURLE_OUT_OF_MEMORY;
}
else {
BACKEND->encdata_buffer = reallocated_buffer;
BACKEND->encdata_length = reallocated_length;
}
}
for(;;) {
if(doread) {
/* read encrypted handshake data from socket */
result = Curl_read_plain(conn->sock[sockindex],
(char *) (BACKEND->encdata_buffer +
BACKEND->encdata_offset),
BACKEND->encdata_length -
BACKEND->encdata_offset,
&nread);
if(result == CURLE_AGAIN) {
if(connssl->connecting_state != ssl_connect_2_writing)
connssl->connecting_state = ssl_connect_2_reading;
infof(data, "schannel: failed to receive handshake, "
"need more data\n");
return CURLE_OK;
}
else if((result != CURLE_OK) || (nread == 0)) {
failf(data, "schannel: failed to receive handshake, "
"SSL/TLS connection failed");
return CURLE_SSL_CONNECT_ERROR;
}
/* increase encrypted data buffer offset */
BACKEND->encdata_offset += nread;
BACKEND->encdata_is_incomplete = false;
infof(data, "schannel: encrypted data got %zd\n", nread);
}
infof(data, "schannel: encrypted data buffer: offset %zu length %zu\n",
BACKEND->encdata_offset, BACKEND->encdata_length);
/* setup input buffers */
InitSecBuffer(&inbuf[0], SECBUFFER_TOKEN, malloc(BACKEND->encdata_offset),
curlx_uztoul(BACKEND->encdata_offset));
InitSecBuffer(&inbuf[1], SECBUFFER_EMPTY, NULL, 0);
InitSecBufferDesc(&inbuf_desc, inbuf, 2);
/* setup output buffers */
InitSecBuffer(&outbuf[0], SECBUFFER_TOKEN, NULL, 0);
InitSecBuffer(&outbuf[1], SECBUFFER_ALERT, NULL, 0);
InitSecBuffer(&outbuf[2], SECBUFFER_EMPTY, NULL, 0);
InitSecBufferDesc(&outbuf_desc, outbuf, 3);
if(inbuf[0].pvBuffer == NULL) {
failf(data, "schannel: unable to allocate memory");
return CURLE_OUT_OF_MEMORY;
}
/* copy received handshake data into input buffer */
memcpy(inbuf[0].pvBuffer, BACKEND->encdata_buffer,
BACKEND->encdata_offset);
host_name = Curl_convert_UTF8_to_tchar(hostname);
if(!host_name)
return CURLE_OUT_OF_MEMORY;
2016-02-03 07:09:25 +03:00
/* https://msdn.microsoft.com/en-us/library/windows/desktop/aa375924.aspx
*/
sspi_status = s_pSecFn->InitializeSecurityContext(
&BACKEND->cred->cred_handle, &BACKEND->ctxt->ctxt_handle,
host_name, BACKEND->req_flags, 0, 0, &inbuf_desc, 0, NULL,
&outbuf_desc, &BACKEND->ret_flags, &BACKEND->ctxt->time_stamp);
Curl_unicodefree(host_name);
/* free buffer for received handshake data */
Curl_safefree(inbuf[0].pvBuffer);
/* check if the handshake was incomplete */
if(sspi_status == SEC_E_INCOMPLETE_MESSAGE) {
BACKEND->encdata_is_incomplete = true;
connssl->connecting_state = ssl_connect_2_reading;
infof(data, "schannel: received incomplete message, need more data\n");
return CURLE_OK;
}
/* If the server has requested a client certificate, attempt to continue
the handshake without one. This will allow connections to servers which
request a client certificate but do not require it. */
if(sspi_status == SEC_I_INCOMPLETE_CREDENTIALS &&
!(BACKEND->req_flags & ISC_REQ_USE_SUPPLIED_CREDS)) {
BACKEND->req_flags |= ISC_REQ_USE_SUPPLIED_CREDS;
connssl->connecting_state = ssl_connect_2_writing;
infof(data, "schannel: a client certificate has been requested\n");
return CURLE_OK;
}
/* check if the handshake needs to be continued */
if(sspi_status == SEC_I_CONTINUE_NEEDED || sspi_status == SEC_E_OK) {
for(i = 0; i < 3; i++) {
/* search for handshake tokens that need to be send */
if(outbuf[i].BufferType == SECBUFFER_TOKEN && outbuf[i].cbBuffer > 0) {
infof(data, "schannel: sending next handshake data: "
"sending %lu bytes...\n", outbuf[i].cbBuffer);
/* send handshake token to server */
result = Curl_write_plain(conn, conn->sock[sockindex],
outbuf[i].pvBuffer, outbuf[i].cbBuffer,
&written);
if((result != CURLE_OK) ||
(outbuf[i].cbBuffer != (size_t) written)) {
failf(data, "schannel: failed to send next handshake data: "
"sent %zd of %lu bytes", written, outbuf[i].cbBuffer);
return CURLE_SSL_CONNECT_ERROR;
}
}
/* free obsolete buffer */
if(outbuf[i].pvBuffer != NULL) {
s_pSecFn->FreeContextBuffer(outbuf[i].pvBuffer);
}
}
}
else {
if(sspi_status == SEC_E_WRONG_PRINCIPAL)
failf(data, "schannel: SNI or certificate check failed: %s",
Curl_sspi_strerror(conn, sspi_status));
else
failf(data, "schannel: next InitializeSecurityContext failed: %s",
Curl_sspi_strerror(conn, sspi_status));
return sspi_status == SEC_E_UNTRUSTED_ROOT ?
CURLE_SSL_CACERT : CURLE_SSL_CONNECT_ERROR;
}
/* check if there was additional remaining encrypted data */
if(inbuf[1].BufferType == SECBUFFER_EXTRA && inbuf[1].cbBuffer > 0) {
infof(data, "schannel: encrypted data length: %lu\n", inbuf[1].cbBuffer);
/*
There are two cases where we could be getting extra data here:
1) If we're renegotiating a connection and the handshake is already
complete (from the server perspective), it can encrypted app data
(not handshake data) in an extra buffer at this point.
2) (sspi_status == SEC_I_CONTINUE_NEEDED) We are negotiating a
connection and this extra data is part of the handshake.
We should process the data immediately; waiting for the socket to
be ready may fail since the server is done sending handshake data.
*/
/* check if the remaining data is less than the total amount
and therefore begins after the already processed data */
if(BACKEND->encdata_offset > inbuf[1].cbBuffer) {
memmove(BACKEND->encdata_buffer,
(BACKEND->encdata_buffer + BACKEND->encdata_offset) -
inbuf[1].cbBuffer, inbuf[1].cbBuffer);
BACKEND->encdata_offset = inbuf[1].cbBuffer;
if(sspi_status == SEC_I_CONTINUE_NEEDED) {
doread = FALSE;
continue;
}
}
}
else {
BACKEND->encdata_offset = 0;
}
break;
}
/* check if the handshake needs to be continued */
if(sspi_status == SEC_I_CONTINUE_NEEDED) {
connssl->connecting_state = ssl_connect_2_reading;
return CURLE_OK;
}
/* check if the handshake is complete */
if(sspi_status == SEC_E_OK) {
connssl->connecting_state = ssl_connect_3;
infof(data, "schannel: SSL/TLS handshake complete\n");
}
pubkey_ptr = SSL_IS_PROXY() ?
data->set.str[STRING_SSL_PINNEDPUBLICKEY_PROXY] :
data->set.str[STRING_SSL_PINNEDPUBLICKEY_ORIG];
if(pubkey_ptr) {
result = pkp_pin_peer_pubkey(conn, sockindex, pubkey_ptr);
if(result) {
failf(data, "SSL: public key does not match pinned public key!");
return result;
}
}
#ifdef _WIN32_WCE
/* Windows CE doesn't do any server certificate validation.
We have to do it manually. */
proxy: Support HTTPS proxy and SOCKS+HTTP(s) * HTTPS proxies: An HTTPS proxy receives all transactions over an SSL/TLS connection. Once a secure connection with the proxy is established, the user agent uses the proxy as usual, including sending CONNECT requests to instruct the proxy to establish a [usually secure] TCP tunnel with an origin server. HTTPS proxies protect nearly all aspects of user-proxy communications as opposed to HTTP proxies that receive all requests (including CONNECT requests) in vulnerable clear text. With HTTPS proxies, it is possible to have two concurrent _nested_ SSL/TLS sessions: the "outer" one between the user agent and the proxy and the "inner" one between the user agent and the origin server (through the proxy). This change adds supports for such nested sessions as well. A secure connection with a proxy requires its own set of the usual SSL options (their actual descriptions differ and need polishing, see TODO): --proxy-cacert FILE CA certificate to verify peer against --proxy-capath DIR CA directory to verify peer against --proxy-cert CERT[:PASSWD] Client certificate file and password --proxy-cert-type TYPE Certificate file type (DER/PEM/ENG) --proxy-ciphers LIST SSL ciphers to use --proxy-crlfile FILE Get a CRL list in PEM format from the file --proxy-insecure Allow connections to proxies with bad certs --proxy-key KEY Private key file name --proxy-key-type TYPE Private key file type (DER/PEM/ENG) --proxy-pass PASS Pass phrase for the private key --proxy-ssl-allow-beast Allow security flaw to improve interop --proxy-sslv2 Use SSLv2 --proxy-sslv3 Use SSLv3 --proxy-tlsv1 Use TLSv1 --proxy-tlsuser USER TLS username --proxy-tlspassword STRING TLS password --proxy-tlsauthtype STRING TLS authentication type (default SRP) All --proxy-foo options are independent from their --foo counterparts, except --proxy-crlfile which defaults to --crlfile and --proxy-capath which defaults to --capath. Curl now also supports %{proxy_ssl_verify_result} --write-out variable, similar to the existing %{ssl_verify_result} variable. Supported backends: OpenSSL, GnuTLS, and NSS. * A SOCKS proxy + HTTP/HTTPS proxy combination: If both --socks* and --proxy options are given, Curl first connects to the SOCKS proxy and then connects (through SOCKS) to the HTTP or HTTPS proxy. TODO: Update documentation for the new APIs and --proxy-* options. Look for "Added in 7.XXX" marks.
2016-11-16 20:49:15 +03:00
if(conn->ssl_config.verifypeer)
return verify_certificate(conn, sockindex);
#endif
return CURLE_OK;
}
static CURLcode
schannel_connect_step3(struct connectdata *conn, int sockindex)
{
CURLcode result = CURLE_OK;
struct Curl_easy *data = conn->data;
struct ssl_connect_data *connssl = &conn->ssl[sockindex];
SECURITY_STATUS sspi_status = SEC_E_OK;
CERT_CONTEXT *ccert_context = NULL;
#ifndef CURL_DISABLE_VERBOSE_STRINGS
const char * const hostname = SSL_IS_PROXY() ? conn->http_proxy.host.name :
conn->host.name;
#endif
#ifdef HAS_ALPN
SecPkgContext_ApplicationProtocol alpn_result;
#endif
DEBUGASSERT(ssl_connect_3 == connssl->connecting_state);
infof(data, "schannel: SSL/TLS connection with %s port %hu (step 3/3)\n",
hostname, conn->remote_port);
if(!BACKEND->cred)
return CURLE_SSL_CONNECT_ERROR;
/* check if the required context attributes are met */
if(BACKEND->ret_flags != BACKEND->req_flags) {
if(!(BACKEND->ret_flags & ISC_RET_SEQUENCE_DETECT))
failf(data, "schannel: failed to setup sequence detection");
if(!(BACKEND->ret_flags & ISC_RET_REPLAY_DETECT))
failf(data, "schannel: failed to setup replay detection");
if(!(BACKEND->ret_flags & ISC_RET_CONFIDENTIALITY))
failf(data, "schannel: failed to setup confidentiality");
if(!(BACKEND->ret_flags & ISC_RET_ALLOCATED_MEMORY))
failf(data, "schannel: failed to setup memory allocation");
if(!(BACKEND->ret_flags & ISC_RET_STREAM))
failf(data, "schannel: failed to setup stream orientation");
return CURLE_SSL_CONNECT_ERROR;
}
#ifdef HAS_ALPN
if(BACKEND->use_alpn) {
sspi_status = s_pSecFn->QueryContextAttributes(&BACKEND->ctxt->ctxt_handle,
SECPKG_ATTR_APPLICATION_PROTOCOL, &alpn_result);
if(sspi_status != SEC_E_OK) {
failf(data, "schannel: failed to retrieve ALPN result");
return CURLE_SSL_CONNECT_ERROR;
}
if(alpn_result.ProtoNegoStatus ==
SecApplicationProtocolNegotiationStatus_Success) {
infof(data, "schannel: ALPN, server accepted to use %.*s\n",
alpn_result.ProtocolIdSize, alpn_result.ProtocolId);
#ifdef USE_NGHTTP2
if(alpn_result.ProtocolIdSize == NGHTTP2_PROTO_VERSION_ID_LEN &&
!memcmp(NGHTTP2_PROTO_VERSION_ID, alpn_result.ProtocolId,
NGHTTP2_PROTO_VERSION_ID_LEN)) {
conn->negnpn = CURL_HTTP_VERSION_2;
}
else
#endif
if(alpn_result.ProtocolIdSize == ALPN_HTTP_1_1_LENGTH &&
!memcmp(ALPN_HTTP_1_1, alpn_result.ProtocolId,
ALPN_HTTP_1_1_LENGTH)) {
conn->negnpn = CURL_HTTP_VERSION_1_1;
}
}
else
infof(data, "ALPN, server did not agree to a protocol\n");
}
#endif
/* save the current session data for possible re-use */
if(SSL_SET_OPTION(primary.sessionid)) {
bool incache;
struct curl_schannel_cred *old_cred = NULL;
Curl_ssl_sessionid_lock(conn);
proxy: Support HTTPS proxy and SOCKS+HTTP(s) * HTTPS proxies: An HTTPS proxy receives all transactions over an SSL/TLS connection. Once a secure connection with the proxy is established, the user agent uses the proxy as usual, including sending CONNECT requests to instruct the proxy to establish a [usually secure] TCP tunnel with an origin server. HTTPS proxies protect nearly all aspects of user-proxy communications as opposed to HTTP proxies that receive all requests (including CONNECT requests) in vulnerable clear text. With HTTPS proxies, it is possible to have two concurrent _nested_ SSL/TLS sessions: the "outer" one between the user agent and the proxy and the "inner" one between the user agent and the origin server (through the proxy). This change adds supports for such nested sessions as well. A secure connection with a proxy requires its own set of the usual SSL options (their actual descriptions differ and need polishing, see TODO): --proxy-cacert FILE CA certificate to verify peer against --proxy-capath DIR CA directory to verify peer against --proxy-cert CERT[:PASSWD] Client certificate file and password --proxy-cert-type TYPE Certificate file type (DER/PEM/ENG) --proxy-ciphers LIST SSL ciphers to use --proxy-crlfile FILE Get a CRL list in PEM format from the file --proxy-insecure Allow connections to proxies with bad certs --proxy-key KEY Private key file name --proxy-key-type TYPE Private key file type (DER/PEM/ENG) --proxy-pass PASS Pass phrase for the private key --proxy-ssl-allow-beast Allow security flaw to improve interop --proxy-sslv2 Use SSLv2 --proxy-sslv3 Use SSLv3 --proxy-tlsv1 Use TLSv1 --proxy-tlsuser USER TLS username --proxy-tlspassword STRING TLS password --proxy-tlsauthtype STRING TLS authentication type (default SRP) All --proxy-foo options are independent from their --foo counterparts, except --proxy-crlfile which defaults to --crlfile and --proxy-capath which defaults to --capath. Curl now also supports %{proxy_ssl_verify_result} --write-out variable, similar to the existing %{ssl_verify_result} variable. Supported backends: OpenSSL, GnuTLS, and NSS. * A SOCKS proxy + HTTP/HTTPS proxy combination: If both --socks* and --proxy options are given, Curl first connects to the SOCKS proxy and then connects (through SOCKS) to the HTTP or HTTPS proxy. TODO: Update documentation for the new APIs and --proxy-* options. Look for "Added in 7.XXX" marks.
2016-11-16 20:49:15 +03:00
incache = !(Curl_ssl_getsessionid(conn, (void **)&old_cred, NULL,
sockindex));
if(incache) {
if(old_cred != BACKEND->cred) {
infof(data, "schannel: old credential handle is stale, removing\n");
/* we're not taking old_cred ownership here, no refcount++ is needed */
Curl_ssl_delsessionid(conn, (void *)old_cred);
incache = FALSE;
}
}
if(!incache) {
result = Curl_ssl_addsessionid(conn, (void *)BACKEND->cred,
proxy: Support HTTPS proxy and SOCKS+HTTP(s) * HTTPS proxies: An HTTPS proxy receives all transactions over an SSL/TLS connection. Once a secure connection with the proxy is established, the user agent uses the proxy as usual, including sending CONNECT requests to instruct the proxy to establish a [usually secure] TCP tunnel with an origin server. HTTPS proxies protect nearly all aspects of user-proxy communications as opposed to HTTP proxies that receive all requests (including CONNECT requests) in vulnerable clear text. With HTTPS proxies, it is possible to have two concurrent _nested_ SSL/TLS sessions: the "outer" one between the user agent and the proxy and the "inner" one between the user agent and the origin server (through the proxy). This change adds supports for such nested sessions as well. A secure connection with a proxy requires its own set of the usual SSL options (their actual descriptions differ and need polishing, see TODO): --proxy-cacert FILE CA certificate to verify peer against --proxy-capath DIR CA directory to verify peer against --proxy-cert CERT[:PASSWD] Client certificate file and password --proxy-cert-type TYPE Certificate file type (DER/PEM/ENG) --proxy-ciphers LIST SSL ciphers to use --proxy-crlfile FILE Get a CRL list in PEM format from the file --proxy-insecure Allow connections to proxies with bad certs --proxy-key KEY Private key file name --proxy-key-type TYPE Private key file type (DER/PEM/ENG) --proxy-pass PASS Pass phrase for the private key --proxy-ssl-allow-beast Allow security flaw to improve interop --proxy-sslv2 Use SSLv2 --proxy-sslv3 Use SSLv3 --proxy-tlsv1 Use TLSv1 --proxy-tlsuser USER TLS username --proxy-tlspassword STRING TLS password --proxy-tlsauthtype STRING TLS authentication type (default SRP) All --proxy-foo options are independent from their --foo counterparts, except --proxy-crlfile which defaults to --crlfile and --proxy-capath which defaults to --capath. Curl now also supports %{proxy_ssl_verify_result} --write-out variable, similar to the existing %{ssl_verify_result} variable. Supported backends: OpenSSL, GnuTLS, and NSS. * A SOCKS proxy + HTTP/HTTPS proxy combination: If both --socks* and --proxy options are given, Curl first connects to the SOCKS proxy and then connects (through SOCKS) to the HTTP or HTTPS proxy. TODO: Update documentation for the new APIs and --proxy-* options. Look for "Added in 7.XXX" marks.
2016-11-16 20:49:15 +03:00
sizeof(struct curl_schannel_cred),
sockindex);
if(result) {
Curl_ssl_sessionid_unlock(conn);
failf(data, "schannel: failed to store credential handle");
return result;
}
else {
/* this cred session is now also referenced by sessionid cache */
BACKEND->cred->refcount++;
infof(data, "schannel: stored credential handle in session cache\n");
}
}
Curl_ssl_sessionid_unlock(conn);
}
if(data->set.ssl.certinfo) {
sspi_status = s_pSecFn->QueryContextAttributes(&BACKEND->ctxt->ctxt_handle,
SECPKG_ATTR_REMOTE_CERT_CONTEXT, &ccert_context);
if((sspi_status != SEC_E_OK) || (ccert_context == NULL)) {
failf(data, "schannel: failed to retrieve remote cert context");
return CURLE_SSL_CONNECT_ERROR;
}
result = Curl_ssl_init_certinfo(data, 1);
if(!result) {
if(((ccert_context->dwCertEncodingType & X509_ASN_ENCODING) != 0) &&
(ccert_context->cbCertEncoded > 0)) {
const char *beg = (const char *) ccert_context->pbCertEncoded;
const char *end = beg + ccert_context->cbCertEncoded;
result = Curl_extract_certinfo(conn, 0, beg, end);
}
}
CertFreeCertificateContext(ccert_context);
if(result)
return result;
}
connssl->connecting_state = ssl_connect_done;
return CURLE_OK;
}
static CURLcode
schannel_connect_common(struct connectdata *conn, int sockindex,
bool nonblocking, bool *done)
{
CURLcode result;
struct Curl_easy *data = conn->data;
struct ssl_connect_data *connssl = &conn->ssl[sockindex];
curl_socket_t sockfd = conn->sock[sockindex];
time_t timeout_ms;
int what;
/* check if the connection has already been established */
if(ssl_connection_complete == connssl->state) {
*done = TRUE;
return CURLE_OK;
}
if(ssl_connect_1 == connssl->connecting_state) {
/* check out how much more time we're allowed */
timeout_ms = Curl_timeleft(data, NULL, TRUE);
if(timeout_ms < 0) {
/* no need to continue if time already is up */
failf(data, "SSL/TLS connection timeout");
return CURLE_OPERATION_TIMEDOUT;
}
result = schannel_connect_step1(conn, sockindex);
if(result)
return result;
}
while(ssl_connect_2 == connssl->connecting_state ||
ssl_connect_2_reading == connssl->connecting_state ||
ssl_connect_2_writing == connssl->connecting_state) {
/* check out how much more time we're allowed */
timeout_ms = Curl_timeleft(data, NULL, TRUE);
if(timeout_ms < 0) {
/* no need to continue if time already is up */
failf(data, "SSL/TLS connection timeout");
return CURLE_OPERATION_TIMEDOUT;
}
/* if ssl is expecting something, check if it's available. */
if(connssl->connecting_state == ssl_connect_2_reading
|| connssl->connecting_state == ssl_connect_2_writing) {
curl_socket_t writefd = ssl_connect_2_writing ==
connssl->connecting_state ? sockfd : CURL_SOCKET_BAD;
curl_socket_t readfd = ssl_connect_2_reading ==
connssl->connecting_state ? sockfd : CURL_SOCKET_BAD;
what = Curl_socket_check(readfd, CURL_SOCKET_BAD, writefd,
nonblocking ? 0 : timeout_ms);
if(what < 0) {
/* fatal error */
failf(data, "select/poll on SSL/TLS socket, errno: %d", SOCKERRNO);
return CURLE_SSL_CONNECT_ERROR;
}
else if(0 == what) {
if(nonblocking) {
*done = FALSE;
return CURLE_OK;
}
else {
/* timeout */
failf(data, "SSL/TLS connection timeout");
return CURLE_OPERATION_TIMEDOUT;
}
}
/* socket is readable or writable */
}
/* Run transaction, and return to the caller if it failed or if
* this connection is part of a multi handle and this loop would
* execute again. This permits the owner of a multi handle to
* abort a connection attempt before step2 has completed while
* ensuring that a client using select() or epoll() will always
* have a valid fdset to wait on.
*/
result = schannel_connect_step2(conn, sockindex);
if(result || (nonblocking &&
(ssl_connect_2 == connssl->connecting_state ||
ssl_connect_2_reading == connssl->connecting_state ||
ssl_connect_2_writing == connssl->connecting_state)))
return result;
} /* repeat step2 until all transactions are done. */
if(ssl_connect_3 == connssl->connecting_state) {
result = schannel_connect_step3(conn, sockindex);
if(result)
return result;
}
if(ssl_connect_done == connssl->connecting_state) {
connssl->state = ssl_connection_complete;
conn->recv[sockindex] = schannel_recv;
conn->send[sockindex] = schannel_send;
*done = TRUE;
}
else
*done = FALSE;
/* reset our connection state machine */
connssl->connecting_state = ssl_connect_1;
return CURLE_OK;
}
static ssize_t
schannel_send(struct connectdata *conn, int sockindex,
const void *buf, size_t len, CURLcode *err)
{
ssize_t written = -1;
size_t data_len = 0;
unsigned char *data = NULL;
struct ssl_connect_data *connssl = &conn->ssl[sockindex];
SecBuffer outbuf[4];
SecBufferDesc outbuf_desc;
SECURITY_STATUS sspi_status = SEC_E_OK;
CURLcode result;
/* check if the maximum stream sizes were queried */
if(BACKEND->stream_sizes.cbMaximumMessage == 0) {
sspi_status = s_pSecFn->QueryContextAttributes(
&BACKEND->ctxt->ctxt_handle,
SECPKG_ATTR_STREAM_SIZES,
&BACKEND->stream_sizes);
if(sspi_status != SEC_E_OK) {
*err = CURLE_SEND_ERROR;
return -1;
}
}
/* check if the buffer is longer than the maximum message length */
if(len > BACKEND->stream_sizes.cbMaximumMessage) {
len = BACKEND->stream_sizes.cbMaximumMessage;
}
/* calculate the complete message length and allocate a buffer for it */
data_len = BACKEND->stream_sizes.cbHeader + len +
BACKEND->stream_sizes.cbTrailer;
data = (unsigned char *) malloc(data_len);
if(data == NULL) {
*err = CURLE_OUT_OF_MEMORY;
return -1;
}
/* setup output buffers (header, data, trailer, empty) */
InitSecBuffer(&outbuf[0], SECBUFFER_STREAM_HEADER,
data, BACKEND->stream_sizes.cbHeader);
InitSecBuffer(&outbuf[1], SECBUFFER_DATA,
data + BACKEND->stream_sizes.cbHeader, curlx_uztoul(len));
InitSecBuffer(&outbuf[2], SECBUFFER_STREAM_TRAILER,
data + BACKEND->stream_sizes.cbHeader + len,
BACKEND->stream_sizes.cbTrailer);
InitSecBuffer(&outbuf[3], SECBUFFER_EMPTY, NULL, 0);
InitSecBufferDesc(&outbuf_desc, outbuf, 4);
/* copy data into output buffer */
memcpy(outbuf[1].pvBuffer, buf, len);
2016-02-03 07:09:25 +03:00
/* https://msdn.microsoft.com/en-us/library/windows/desktop/aa375390.aspx */
sspi_status = s_pSecFn->EncryptMessage(&BACKEND->ctxt->ctxt_handle, 0,
&outbuf_desc, 0);
/* check if the message was encrypted */
if(sspi_status == SEC_E_OK) {
written = 0;
/* send the encrypted message including header, data and trailer */
len = outbuf[0].cbBuffer + outbuf[1].cbBuffer + outbuf[2].cbBuffer;
/*
It's important to send the full message which includes the header,
encrypted payload, and trailer. Until the client receives all the
data a coherent message has not been delivered and the client
can't read any of it.
If we wanted to buffer the unwritten encrypted bytes, we would
tell the client that all data it has requested to be sent has been
sent. The unwritten encrypted bytes would be the first bytes to
send on the next invocation.
Here's the catch with this - if we tell the client that all the
bytes have been sent, will the client call this method again to
send the buffered data? Looking at who calls this function, it
seems the answer is NO.
*/
/* send entire message or fail */
while(len > (size_t)written) {
ssize_t this_write;
time_t timeleft;
int what;
this_write = 0;
timeleft = Curl_timeleft(conn->data, NULL, FALSE);
if(timeleft < 0) {
/* we already got the timeout */
failf(conn->data, "schannel: timed out sending data "
"(bytes sent: %zd)", written);
*err = CURLE_OPERATION_TIMEDOUT;
written = -1;
break;
}
what = SOCKET_WRITABLE(conn->sock[sockindex], timeleft);
if(what < 0) {
/* fatal error */
failf(conn->data, "select/poll on SSL socket, errno: %d", SOCKERRNO);
*err = CURLE_SEND_ERROR;
written = -1;
break;
}
else if(0 == what) {
failf(conn->data, "schannel: timed out sending data "
"(bytes sent: %zd)", written);
*err = CURLE_OPERATION_TIMEDOUT;
written = -1;
break;
}
/* socket is writable */
result = Curl_write_plain(conn, conn->sock[sockindex], data + written,
len - written, &this_write);
if(result == CURLE_AGAIN)
continue;
else if(result != CURLE_OK) {
*err = result;
written = -1;
break;
}
written += this_write;
}
}
else if(sspi_status == SEC_E_INSUFFICIENT_MEMORY) {
*err = CURLE_OUT_OF_MEMORY;
}
else{
*err = CURLE_SEND_ERROR;
}
Curl_safefree(data);
if(len == (size_t)written)
/* Encrypted message including header, data and trailer entirely sent.
The return value is the number of unencrypted bytes that were sent. */
written = outbuf[1].cbBuffer;
return written;
}
static ssize_t
schannel_recv(struct connectdata *conn, int sockindex,
char *buf, size_t len, CURLcode *err)
{
size_t size = 0;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
ssize_t nread = -1;
struct Curl_easy *data = conn->data;
struct ssl_connect_data *connssl = &conn->ssl[sockindex];
unsigned char *reallocated_buffer;
size_t reallocated_length;
bool done = FALSE;
SecBuffer inbuf[4];
SecBufferDesc inbuf_desc;
SECURITY_STATUS sspi_status = SEC_E_OK;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
/* we want the length of the encrypted buffer to be at least large enough
that it can hold all the bytes requested and some TLS record overhead. */
size_t min_encdata_length = len + CURL_SCHANNEL_BUFFER_FREE_SIZE;
/****************************************************************************
* Don't return or set BACKEND->recv_unrecoverable_err unless in the cleanup.
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
* The pattern for return error is set *err, optional infof, goto cleanup.
*
* Our priority is to always return as much decrypted data to the caller as
* possible, even if an error occurs. The state of the decrypted buffer must
* always be valid. Transfer of decrypted data to the caller's buffer is
* handled in the cleanup.
*/
infof(data, "schannel: client wants to read %zu bytes\n", len);
*err = CURLE_OK;
if(len && len <= BACKEND->decdata_offset) {
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
infof(data, "schannel: enough decrypted data is already available\n");
goto cleanup;
}
else if(BACKEND->recv_unrecoverable_err) {
*err = BACKEND->recv_unrecoverable_err;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
infof(data, "schannel: an unrecoverable error occurred in a prior call\n");
goto cleanup;
}
else if(BACKEND->recv_sspi_close_notify) {
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
/* once a server has indicated shutdown there is no more encrypted data */
infof(data, "schannel: server indicated shutdown in a prior call\n");
goto cleanup;
}
else if(!len) {
/* It's debatable what to return when !len. Regardless we can't return
immediately because there may be data to decrypt (in the case we want to
decrypt all encrypted cached data) so handle !len later in cleanup.
*/
; /* do nothing */
}
else if(!BACKEND->recv_connection_closed) {
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
/* increase enc buffer in order to fit the requested amount of data */
size = BACKEND->encdata_length - BACKEND->encdata_offset;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
if(size < CURL_SCHANNEL_BUFFER_FREE_SIZE ||
BACKEND->encdata_length < min_encdata_length) {
reallocated_length = BACKEND->encdata_offset +
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
CURL_SCHANNEL_BUFFER_FREE_SIZE;
if(reallocated_length < min_encdata_length) {
reallocated_length = min_encdata_length;
}
reallocated_buffer = realloc(BACKEND->encdata_buffer,
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
reallocated_length);
if(reallocated_buffer == NULL) {
*err = CURLE_OUT_OF_MEMORY;
failf(data, "schannel: unable to re-allocate memory");
goto cleanup;
}
BACKEND->encdata_buffer = reallocated_buffer;
BACKEND->encdata_length = reallocated_length;
size = BACKEND->encdata_length - BACKEND->encdata_offset;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
infof(data, "schannel: encdata_buffer resized %zu\n",
BACKEND->encdata_length);
}
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
infof(data, "schannel: encrypted data buffer: offset %zu length %zu\n",
BACKEND->encdata_offset, BACKEND->encdata_length);
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
/* read encrypted data from socket */
*err = Curl_read_plain(conn->sock[sockindex],
(char *)(BACKEND->encdata_buffer +
BACKEND->encdata_offset),
size, &nread);
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
if(*err) {
nread = -1;
if(*err == CURLE_AGAIN)
infof(data, "schannel: Curl_read_plain returned CURLE_AGAIN\n");
else if(*err == CURLE_RECV_ERROR)
infof(data, "schannel: Curl_read_plain returned CURLE_RECV_ERROR\n");
else
infof(data, "schannel: Curl_read_plain returned error %d\n", *err);
}
else if(nread == 0) {
BACKEND->recv_connection_closed = true;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
infof(data, "schannel: server closed the connection\n");
2015-05-02 23:21:25 +03:00
}
else if(nread > 0) {
BACKEND->encdata_offset += (size_t)nread;
BACKEND->encdata_is_incomplete = false;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
infof(data, "schannel: encrypted data got %zd\n", nread);
}
}
infof(data, "schannel: encrypted data buffer: offset %zu length %zu\n",
BACKEND->encdata_offset, BACKEND->encdata_length);
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
/* decrypt loop */
while(BACKEND->encdata_offset > 0 && sspi_status == SEC_E_OK &&
(!len || BACKEND->decdata_offset < len ||
BACKEND->recv_connection_closed)) {
/* prepare data buffer for DecryptMessage call */
InitSecBuffer(&inbuf[0], SECBUFFER_DATA, BACKEND->encdata_buffer,
curlx_uztoul(BACKEND->encdata_offset));
/* we need 3 more empty input buffers for possible output */
InitSecBuffer(&inbuf[1], SECBUFFER_EMPTY, NULL, 0);
InitSecBuffer(&inbuf[2], SECBUFFER_EMPTY, NULL, 0);
InitSecBuffer(&inbuf[3], SECBUFFER_EMPTY, NULL, 0);
InitSecBufferDesc(&inbuf_desc, inbuf, 4);
2016-02-03 07:09:25 +03:00
/* https://msdn.microsoft.com/en-us/library/windows/desktop/aa375348.aspx
*/
sspi_status = s_pSecFn->DecryptMessage(&BACKEND->ctxt->ctxt_handle,
&inbuf_desc, 0, NULL);
/* check if everything went fine (server may want to renegotiate
or shutdown the connection context) */
if(sspi_status == SEC_E_OK || sspi_status == SEC_I_RENEGOTIATE ||
sspi_status == SEC_I_CONTEXT_EXPIRED) {
/* check for successfully decrypted data, even before actual
renegotiation or shutdown of the connection context */
if(inbuf[1].BufferType == SECBUFFER_DATA) {
infof(data, "schannel: decrypted data length: %lu\n",
inbuf[1].cbBuffer);
/* increase buffer in order to fit the received amount of data */
size = inbuf[1].cbBuffer > CURL_SCHANNEL_BUFFER_FREE_SIZE ?
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
inbuf[1].cbBuffer : CURL_SCHANNEL_BUFFER_FREE_SIZE;
if(BACKEND->decdata_length - BACKEND->decdata_offset < size ||
BACKEND->decdata_length < len) {
/* increase internal decrypted data buffer */
reallocated_length = BACKEND->decdata_offset + size;
/* make sure that the requested amount of data fits */
if(reallocated_length < len) {
reallocated_length = len;
}
reallocated_buffer = realloc(BACKEND->decdata_buffer,
reallocated_length);
if(reallocated_buffer == NULL) {
*err = CURLE_OUT_OF_MEMORY;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
failf(data, "schannel: unable to re-allocate memory");
goto cleanup;
}
BACKEND->decdata_buffer = reallocated_buffer;
BACKEND->decdata_length = reallocated_length;
}
/* copy decrypted data to internal buffer */
size = inbuf[1].cbBuffer;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
if(size) {
memcpy(BACKEND->decdata_buffer + BACKEND->decdata_offset,
inbuf[1].pvBuffer, size);
BACKEND->decdata_offset += size;
}
infof(data, "schannel: decrypted data added: %zu\n", size);
infof(data, "schannel: decrypted data cached: offset %zu length %zu\n",
BACKEND->decdata_offset, BACKEND->decdata_length);
}
/* check for remaining encrypted data */
if(inbuf[3].BufferType == SECBUFFER_EXTRA && inbuf[3].cbBuffer > 0) {
infof(data, "schannel: encrypted data length: %lu\n",
inbuf[3].cbBuffer);
/* check if the remaining data is less than the total amount
* and therefore begins after the already processed data
*/
if(BACKEND->encdata_offset > inbuf[3].cbBuffer) {
/* move remaining encrypted data forward to the beginning of
buffer */
memmove(BACKEND->encdata_buffer,
(BACKEND->encdata_buffer + BACKEND->encdata_offset) -
inbuf[3].cbBuffer, inbuf[3].cbBuffer);
BACKEND->encdata_offset = inbuf[3].cbBuffer;
}
infof(data, "schannel: encrypted data cached: offset %zu length %zu\n",
BACKEND->encdata_offset, BACKEND->encdata_length);
}
else {
/* reset encrypted buffer offset, because there is no data remaining */
BACKEND->encdata_offset = 0;
}
/* check if server wants to renegotiate the connection context */
if(sspi_status == SEC_I_RENEGOTIATE) {
infof(data, "schannel: remote party requests renegotiation\n");
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
if(*err && *err != CURLE_AGAIN) {
infof(data, "schannel: can't renogotiate, an error is pending\n");
goto cleanup;
}
if(BACKEND->encdata_offset) {
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
*err = CURLE_RECV_ERROR;
infof(data, "schannel: can't renogotiate, "
"encrypted data available\n");
goto cleanup;
}
/* begin renegotiation */
infof(data, "schannel: renegotiating SSL/TLS connection\n");
connssl->state = ssl_connection_negotiating;
connssl->connecting_state = ssl_connect_2_writing;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
*err = schannel_connect_common(conn, sockindex, FALSE, &done);
if(*err) {
infof(data, "schannel: renegotiation failed\n");
goto cleanup;
}
/* now retry receiving data */
sspi_status = SEC_E_OK;
infof(data, "schannel: SSL/TLS connection renegotiated\n");
continue;
}
/* check if the server closed the connection */
else if(sspi_status == SEC_I_CONTEXT_EXPIRED) {
/* In Windows 2000 SEC_I_CONTEXT_EXPIRED (close_notify) is not
returned so we have to work around that in cleanup. */
BACKEND->recv_sspi_close_notify = true;
if(!BACKEND->recv_connection_closed) {
BACKEND->recv_connection_closed = true;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
infof(data, "schannel: server closed the connection\n");
}
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
goto cleanup;
}
}
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
else if(sspi_status == SEC_E_INCOMPLETE_MESSAGE) {
BACKEND->encdata_is_incomplete = true;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
if(!*err)
*err = CURLE_AGAIN;
infof(data, "schannel: failed to decrypt data, need more data\n");
goto cleanup;
}
else {
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
*err = CURLE_RECV_ERROR;
infof(data, "schannel: failed to read data from server: %s\n",
Curl_sspi_strerror(conn, sspi_status));
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
goto cleanup;
}
}
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
infof(data, "schannel: encrypted data buffer: offset %zu length %zu\n",
BACKEND->encdata_offset, BACKEND->encdata_length);
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
infof(data, "schannel: decrypted data buffer: offset %zu length %zu\n",
BACKEND->decdata_offset, BACKEND->decdata_length);
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
cleanup:
/* Warning- there is no guarantee the encdata state is valid at this point */
infof(data, "schannel: schannel_recv cleanup\n");
/* Error if the connection has closed without a close_notify.
Behavior here is a matter of debate. We don't want to be vulnerable to a
truncation attack however there's some browser precedent for ignoring the
close_notify for compatibility reasons.
Additionally, Windows 2000 (v5.0) is a special case since it seems it doesn't
return close_notify. In that case if the connection was closed we assume it
was graceful (close_notify) since there doesn't seem to be a way to tell.
*/
if(len && !BACKEND->decdata_offset && BACKEND->recv_connection_closed &&
!BACKEND->recv_sspi_close_notify) {
bool isWin2k = Curl_verify_windows_version(5, 0, PLATFORM_WINNT,
VERSION_EQUAL);
if(isWin2k && sspi_status == SEC_E_OK)
BACKEND->recv_sspi_close_notify = true;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
else {
*err = CURLE_RECV_ERROR;
infof(data, "schannel: server closed abruptly (missing close_notify)\n");
}
}
/* Any error other than CURLE_AGAIN is an unrecoverable error. */
if(*err && *err != CURLE_AGAIN)
BACKEND->recv_unrecoverable_err = *err;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
size = len < BACKEND->decdata_offset ? len : BACKEND->decdata_offset;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
if(size) {
memcpy(buf, BACKEND->decdata_buffer, size);
memmove(BACKEND->decdata_buffer, BACKEND->decdata_buffer + size,
BACKEND->decdata_offset - size);
BACKEND->decdata_offset -= size;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
infof(data, "schannel: decrypted data returned %zu\n", size);
infof(data, "schannel: decrypted data buffer: offset %zu length %zu\n",
BACKEND->decdata_offset, BACKEND->decdata_length);
*err = CURLE_OK;
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
return (ssize_t)size;
}
if(!*err && !BACKEND->recv_connection_closed)
schannel: schannel_recv overhaul This commit is several drafts squashed together. The changes from each draft are noted below. If any changes are similar and possibly contradictory the change in the latest draft takes precedence. Bug: https://github.com/bagder/curl/issues/244 Reported-by: Chris Araman %% %% Draft 1 %% - return 0 if len == 0. that will have to be documented. - continue on and process the caches regardless of raw recv - if decrypted data will be returned then set the error code to CURLE_OK and return its count - if decrypted data will not be returned and the connection has closed (eg nread == 0) then return 0 and CURLE_OK - if decrypted data will not be returned and the connection *hasn't* closed then set the error code to CURLE_AGAIN --only if an error code isn't already set-- and return -1 - narrow the Win2k workaround to only Win2k %% %% Draft 2 %% - Trying out a change in flow to handle corner cases. %% %% Draft 3 %% - Back out the lazier decryption change made in draft2. %% %% Draft 4 %% - Some formatting and branching changes - Decrypt all encrypted cached data when len == 0 - Save connection closed state - Change special Win2k check to use connection closed state %% %% Draft 5 %% - Default to CURLE_AGAIN in cleanup if an error code wasn't set and the connection isn't closed. %% %% Draft 6 %% - Save the last error only if it is an unrecoverable error. Prior to this I saved the last error state in all cases; unfortunately the logic to cover that in all cases would lead to some muddle and I'm concerned that could then lead to a bug in the future so I've replaced it by only recording an unrecoverable error and that state will persist. - Do not recurse on renegotiation. Instead we'll continue on to process any trailing encrypted data received during the renegotiation only. - Move the err checks in cleanup after the check for decrypted data. In either case decrypted data is always returned but I think it's easier to understand when those err checks come after the decrypted data check. %% %% Draft 7 %% - Regardless of len value go directly to cleanup if there is an unrecoverable error or a close_notify was already received. Prior to this change we only acknowledged those two states if len != 0. - Fix a bug in connection closed behavior: Set the error state in the cleanup, because we don't know for sure it's an error until that time. - (Related to above) In the case the connection is closed go "greedy" with the decryption to make sure all remaining encrypted data has been decrypted even if it is not needed at that time by the caller. This is necessary because we can only tell if the connection closed gracefully (close_notify) once all encrypted data has been decrypted. - Do not renegotiate when an unrecoverable error is pending. %% %% Draft 8 %% - Don't show 'server closed the connection' info message twice. - Show an info message if server closed abruptly (missing close_notify).
2015-06-17 07:17:03 +03:00
*err = CURLE_AGAIN;
/* It's debatable what to return when !len. We could return whatever error we
got from decryption but instead we override here so the return is consistent.
*/
if(!len)
*err = CURLE_OK;
return *err ? -1 : 0;
}
static CURLcode Curl_schannel_connect_nonblocking(struct connectdata *conn,
int sockindex, bool *done)
{
return schannel_connect_common(conn, sockindex, TRUE, done);
}
static CURLcode Curl_schannel_connect(struct connectdata *conn, int sockindex)
{
CURLcode result;
bool done = FALSE;
result = schannel_connect_common(conn, sockindex, FALSE, &done);
if(result)
return result;
DEBUGASSERT(done);
return CURLE_OK;
}
static bool Curl_schannel_data_pending(const struct connectdata *conn,
int sockindex)
{
const struct ssl_connect_data *connssl = &conn->ssl[sockindex];
if(connssl->use) /* SSL/TLS is in use */
return (BACKEND->decdata_offset > 0 ||
(BACKEND->encdata_offset > 0 && !BACKEND->encdata_is_incomplete));
else
return FALSE;
}
static void Curl_schannel_close(struct connectdata *conn, int sockindex)
{
if(conn->ssl[sockindex].use)
/* if the SSL/TLS channel hasn't been shut down yet, do that now. */
Curl_ssl_shutdown(conn, sockindex);
}
static void Curl_schannel_session_free(void *ptr)
{
/* this is expected to be called under sessionid lock */
struct curl_schannel_cred *cred = ptr;
cred->refcount--;
if(cred->refcount == 0) {
s_pSecFn->FreeCredentialsHandle(&cred->cred_handle);
Curl_safefree(cred);
}
}
static int Curl_schannel_shutdown(struct connectdata *conn, int sockindex)
{
2016-02-03 07:09:25 +03:00
/* See https://msdn.microsoft.com/en-us/library/windows/desktop/aa380138.aspx
* Shutting Down an Schannel Connection
*/
struct Curl_easy *data = conn->data;
struct ssl_connect_data *connssl = &conn->ssl[sockindex];
char * const hostname = SSL_IS_PROXY() ? conn->http_proxy.host.name :
conn->host.name;
infof(data, "schannel: shutting down SSL/TLS connection with %s port %hu\n",
hostname, conn->remote_port);
if(BACKEND->cred && BACKEND->ctxt) {
SecBufferDesc BuffDesc;
SecBuffer Buffer;
SECURITY_STATUS sspi_status;
SecBuffer outbuf;
SecBufferDesc outbuf_desc;
CURLcode result;
TCHAR *host_name;
DWORD dwshut = SCHANNEL_SHUTDOWN;
InitSecBuffer(&Buffer, SECBUFFER_TOKEN, &dwshut, sizeof(dwshut));
InitSecBufferDesc(&BuffDesc, &Buffer, 1);
sspi_status = s_pSecFn->ApplyControlToken(&BACKEND->ctxt->ctxt_handle,
&BuffDesc);
if(sspi_status != SEC_E_OK)
failf(data, "schannel: ApplyControlToken failure: %s",
Curl_sspi_strerror(conn, sspi_status));
host_name = Curl_convert_UTF8_to_tchar(hostname);
if(!host_name)
return CURLE_OUT_OF_MEMORY;
/* setup output buffer */
InitSecBuffer(&outbuf, SECBUFFER_EMPTY, NULL, 0);
InitSecBufferDesc(&outbuf_desc, &outbuf, 1);
sspi_status = s_pSecFn->InitializeSecurityContext(
&BACKEND->cred->cred_handle,
&BACKEND->ctxt->ctxt_handle,
host_name,
BACKEND->req_flags,
0,
0,
NULL,
0,
&BACKEND->ctxt->ctxt_handle,
&outbuf_desc,
&BACKEND->ret_flags,
&BACKEND->ctxt->time_stamp);
Curl_unicodefree(host_name);
if((sspi_status == SEC_E_OK) || (sspi_status == SEC_I_CONTEXT_EXPIRED)) {
/* send close message which is in output buffer */
ssize_t written;
result = Curl_write_plain(conn, conn->sock[sockindex], outbuf.pvBuffer,
outbuf.cbBuffer, &written);
s_pSecFn->FreeContextBuffer(outbuf.pvBuffer);
if((result != CURLE_OK) || (outbuf.cbBuffer != (size_t) written)) {
infof(data, "schannel: failed to send close msg: %s"
" (bytes written: %zd)\n", curl_easy_strerror(result), written);
}
}
}
/* free SSPI Schannel API security context handle */
if(BACKEND->ctxt) {
infof(data, "schannel: clear security context handle\n");
s_pSecFn->DeleteSecurityContext(&BACKEND->ctxt->ctxt_handle);
Curl_safefree(BACKEND->ctxt);
}
/* free SSPI Schannel API credential handle */
if(BACKEND->cred) {
Curl_ssl_sessionid_lock(conn);
Curl_schannel_session_free(BACKEND->cred);
Curl_ssl_sessionid_unlock(conn);
BACKEND->cred = NULL;
}
/* free internal buffer for received encrypted data */
if(BACKEND->encdata_buffer != NULL) {
Curl_safefree(BACKEND->encdata_buffer);
BACKEND->encdata_length = 0;
BACKEND->encdata_offset = 0;
BACKEND->encdata_is_incomplete = false;
}
/* free internal buffer for received decrypted data */
if(BACKEND->decdata_buffer != NULL) {
Curl_safefree(BACKEND->decdata_buffer);
BACKEND->decdata_length = 0;
BACKEND->decdata_offset = 0;
}
return CURLE_OK;
}
static int Curl_schannel_init(void)
{
return (Curl_sspi_global_init() == CURLE_OK ? 1 : 0);
}
static void Curl_schannel_cleanup(void)
{
Curl_sspi_global_cleanup();
}
static size_t Curl_schannel_version(char *buffer, size_t size)
{
size = snprintf(buffer, size, "WinSSL");
return size;
}
static CURLcode Curl_schannel_random(struct Curl_easy *data UNUSED_PARAM,
unsigned char *entropy, size_t length)
{
HCRYPTPROV hCryptProv = 0;
2017-09-06 22:11:55 +03:00
(void)data;
if(!CryptAcquireContext(&hCryptProv, NULL, NULL, PROV_RSA_FULL,
CRYPT_VERIFYCONTEXT | CRYPT_SILENT))
return CURLE_FAILED_INIT;
if(!CryptGenRandom(hCryptProv, (DWORD)length, entropy)) {
CryptReleaseContext(hCryptProv, 0UL);
return CURLE_FAILED_INIT;
}
CryptReleaseContext(hCryptProv, 0UL);
return CURLE_OK;
}
static CURLcode pkp_pin_peer_pubkey(struct connectdata *conn, int sockindex,
const char *pinnedpubkey)
{
SECURITY_STATUS status;
struct Curl_easy *data = conn->data;
struct ssl_connect_data *connssl = &conn->ssl[sockindex];
CERT_CONTEXT *pCertContextServer = NULL;
const char *x509_der;
DWORD x509_der_len;
curl_X509certificate x509_parsed;
curl_asn1Element *pubkey;
/* Result is returned to caller */
CURLcode result = CURLE_SSL_PINNEDPUBKEYNOTMATCH;
/* if a path wasn't specified, don't pin */
if(!pinnedpubkey)
return CURLE_OK;
do {
status = s_pSecFn->QueryContextAttributes(&BACKEND->ctxt->ctxt_handle,
SECPKG_ATTR_REMOTE_CERT_CONTEXT,
&pCertContextServer);
if((status != SEC_E_OK) || (pCertContextServer == NULL)) {
failf(data, "schannel: Failed to read remote certificate context: %s",
Curl_sspi_strerror(conn, status));
break; /* failed */
}
if(!(((pCertContextServer->dwCertEncodingType & X509_ASN_ENCODING) != 0) &&
(pCertContextServer->cbCertEncoded > 0)))
break;
x509_der = (const char *)pCertContextServer->pbCertEncoded;
x509_der_len = pCertContextServer->cbCertEncoded;
memset(&x509_parsed, 0, sizeof x509_parsed);
if(Curl_parseX509(&x509_parsed, x509_der, x509_der + x509_der_len))
break;
pubkey = &x509_parsed.subjectPublicKeyInfo;
if(!pubkey->header || pubkey->end <= pubkey->header) {
failf(data, "SSL: failed retrieving public key from server certificate");
break;
}
result = Curl_pin_peer_pubkey(data,
pinnedpubkey,
(const unsigned char *)pubkey->header,
(size_t)(pubkey->end - pubkey->header));
if(result) {
failf(data, "SSL: public key does not match pinned public key!");
}
} while(0);
if(pCertContextServer)
CertFreeCertificateContext(pCertContextServer);
return result;
}
#ifdef _WIN32_WCE
static CURLcode verify_certificate(struct connectdata *conn, int sockindex)
{
SECURITY_STATUS status;
struct Curl_easy *data = conn->data;
struct ssl_connect_data *connssl = &conn->ssl[sockindex];
CURLcode result = CURLE_OK;
CERT_CONTEXT *pCertContextServer = NULL;
const CERT_CHAIN_CONTEXT *pChainContext = NULL;
const char * const conn_hostname = SSL_IS_PROXY() ?
conn->http_proxy.host.name :
conn->host.name;
status = s_pSecFn->QueryContextAttributes(&BACKEND->ctxt->ctxt_handle,
SECPKG_ATTR_REMOTE_CERT_CONTEXT,
&pCertContextServer);
if((status != SEC_E_OK) || (pCertContextServer == NULL)) {
failf(data, "schannel: Failed to read remote certificate context: %s",
Curl_sspi_strerror(conn, status));
result = CURLE_PEER_FAILED_VERIFICATION;
}
if(result == CURLE_OK) {
CERT_CHAIN_PARA ChainPara;
memset(&ChainPara, 0, sizeof(ChainPara));
ChainPara.cbSize = sizeof(ChainPara);
if(!CertGetCertificateChain(NULL,
pCertContextServer,
NULL,
pCertContextServer->hCertStore,
&ChainPara,
proxy: Support HTTPS proxy and SOCKS+HTTP(s) * HTTPS proxies: An HTTPS proxy receives all transactions over an SSL/TLS connection. Once a secure connection with the proxy is established, the user agent uses the proxy as usual, including sending CONNECT requests to instruct the proxy to establish a [usually secure] TCP tunnel with an origin server. HTTPS proxies protect nearly all aspects of user-proxy communications as opposed to HTTP proxies that receive all requests (including CONNECT requests) in vulnerable clear text. With HTTPS proxies, it is possible to have two concurrent _nested_ SSL/TLS sessions: the "outer" one between the user agent and the proxy and the "inner" one between the user agent and the origin server (through the proxy). This change adds supports for such nested sessions as well. A secure connection with a proxy requires its own set of the usual SSL options (their actual descriptions differ and need polishing, see TODO): --proxy-cacert FILE CA certificate to verify peer against --proxy-capath DIR CA directory to verify peer against --proxy-cert CERT[:PASSWD] Client certificate file and password --proxy-cert-type TYPE Certificate file type (DER/PEM/ENG) --proxy-ciphers LIST SSL ciphers to use --proxy-crlfile FILE Get a CRL list in PEM format from the file --proxy-insecure Allow connections to proxies with bad certs --proxy-key KEY Private key file name --proxy-key-type TYPE Private key file type (DER/PEM/ENG) --proxy-pass PASS Pass phrase for the private key --proxy-ssl-allow-beast Allow security flaw to improve interop --proxy-sslv2 Use SSLv2 --proxy-sslv3 Use SSLv3 --proxy-tlsv1 Use TLSv1 --proxy-tlsuser USER TLS username --proxy-tlspassword STRING TLS password --proxy-tlsauthtype STRING TLS authentication type (default SRP) All --proxy-foo options are independent from their --foo counterparts, except --proxy-crlfile which defaults to --crlfile and --proxy-capath which defaults to --capath. Curl now also supports %{proxy_ssl_verify_result} --write-out variable, similar to the existing %{ssl_verify_result} variable. Supported backends: OpenSSL, GnuTLS, and NSS. * A SOCKS proxy + HTTP/HTTPS proxy combination: If both --socks* and --proxy options are given, Curl first connects to the SOCKS proxy and then connects (through SOCKS) to the HTTP or HTTPS proxy. TODO: Update documentation for the new APIs and --proxy-* options. Look for "Added in 7.XXX" marks.
2016-11-16 20:49:15 +03:00
(data->set.ssl.no_revoke ? 0 :
CERT_CHAIN_REVOCATION_CHECK_CHAIN),
NULL,
&pChainContext)) {
failf(data, "schannel: CertGetCertificateChain failed: %s",
Curl_sspi_strerror(conn, GetLastError()));
pChainContext = NULL;
result = CURLE_PEER_FAILED_VERIFICATION;
}
if(result == CURLE_OK) {
CERT_SIMPLE_CHAIN *pSimpleChain = pChainContext->rgpChain[0];
DWORD dwTrustErrorMask = ~(DWORD)(CERT_TRUST_IS_NOT_TIME_NESTED);
dwTrustErrorMask &= pSimpleChain->TrustStatus.dwErrorStatus;
if(dwTrustErrorMask) {
if(dwTrustErrorMask & CERT_TRUST_IS_REVOKED)
failf(data, "schannel: CertGetCertificateChain trust error"
" CERT_TRUST_IS_REVOKED");
else if(dwTrustErrorMask & CERT_TRUST_IS_PARTIAL_CHAIN)
failf(data, "schannel: CertGetCertificateChain trust error"
" CERT_TRUST_IS_PARTIAL_CHAIN");
else if(dwTrustErrorMask & CERT_TRUST_IS_UNTRUSTED_ROOT)
failf(data, "schannel: CertGetCertificateChain trust error"
" CERT_TRUST_IS_UNTRUSTED_ROOT");
else if(dwTrustErrorMask & CERT_TRUST_IS_NOT_TIME_VALID)
failf(data, "schannel: CertGetCertificateChain trust error"
" CERT_TRUST_IS_NOT_TIME_VALID");
else
failf(data, "schannel: CertGetCertificateChain error mask: 0x%08x",
dwTrustErrorMask);
result = CURLE_PEER_FAILED_VERIFICATION;
}
}
}
if(result == CURLE_OK) {
proxy: Support HTTPS proxy and SOCKS+HTTP(s) * HTTPS proxies: An HTTPS proxy receives all transactions over an SSL/TLS connection. Once a secure connection with the proxy is established, the user agent uses the proxy as usual, including sending CONNECT requests to instruct the proxy to establish a [usually secure] TCP tunnel with an origin server. HTTPS proxies protect nearly all aspects of user-proxy communications as opposed to HTTP proxies that receive all requests (including CONNECT requests) in vulnerable clear text. With HTTPS proxies, it is possible to have two concurrent _nested_ SSL/TLS sessions: the "outer" one between the user agent and the proxy and the "inner" one between the user agent and the origin server (through the proxy). This change adds supports for such nested sessions as well. A secure connection with a proxy requires its own set of the usual SSL options (their actual descriptions differ and need polishing, see TODO): --proxy-cacert FILE CA certificate to verify peer against --proxy-capath DIR CA directory to verify peer against --proxy-cert CERT[:PASSWD] Client certificate file and password --proxy-cert-type TYPE Certificate file type (DER/PEM/ENG) --proxy-ciphers LIST SSL ciphers to use --proxy-crlfile FILE Get a CRL list in PEM format from the file --proxy-insecure Allow connections to proxies with bad certs --proxy-key KEY Private key file name --proxy-key-type TYPE Private key file type (DER/PEM/ENG) --proxy-pass PASS Pass phrase for the private key --proxy-ssl-allow-beast Allow security flaw to improve interop --proxy-sslv2 Use SSLv2 --proxy-sslv3 Use SSLv3 --proxy-tlsv1 Use TLSv1 --proxy-tlsuser USER TLS username --proxy-tlspassword STRING TLS password --proxy-tlsauthtype STRING TLS authentication type (default SRP) All --proxy-foo options are independent from their --foo counterparts, except --proxy-crlfile which defaults to --crlfile and --proxy-capath which defaults to --capath. Curl now also supports %{proxy_ssl_verify_result} --write-out variable, similar to the existing %{ssl_verify_result} variable. Supported backends: OpenSSL, GnuTLS, and NSS. * A SOCKS proxy + HTTP/HTTPS proxy combination: If both --socks* and --proxy options are given, Curl first connects to the SOCKS proxy and then connects (through SOCKS) to the HTTP or HTTPS proxy. TODO: Update documentation for the new APIs and --proxy-* options. Look for "Added in 7.XXX" marks.
2016-11-16 20:49:15 +03:00
if(conn->ssl_config.verifyhost) {
TCHAR cert_hostname_buff[256];
DWORD len;
/* TODO: Fix this for certificates with multiple alternative names.
Right now we're only asking for the first preferred alternative name.
Instead we'd need to do all via CERT_NAME_SEARCH_ALL_NAMES_FLAG
(if WinCE supports that?) and run this section in a loop for each.
https://msdn.microsoft.com/en-us/library/windows/desktop/aa376086.aspx
curl: (51) schannel: CertGetNameString() certificate hostname
(.google.com) did not match connection (google.com)
*/
len = CertGetNameString(pCertContextServer,
CERT_NAME_DNS_TYPE,
CERT_NAME_DISABLE_IE4_UTF8_FLAG,
NULL,
cert_hostname_buff,
256);
if(len > 0) {
const char *cert_hostname;
/* Comparing the cert name and the connection hostname encoded as UTF-8
* is acceptable since both values are assumed to use ASCII
* (or some equivalent) encoding
*/
cert_hostname = Curl_convert_tchar_to_UTF8(cert_hostname_buff);
if(!cert_hostname) {
result = CURLE_OUT_OF_MEMORY;
}
else{
int match_result;
match_result = Curl_cert_hostcheck(cert_hostname, conn->host.name);
if(match_result == CURL_HOST_MATCH) {
infof(data,
"schannel: connection hostname (%s) validated "
"against certificate name (%s)\n",
conn->host.name,
cert_hostname);
result = CURLE_OK;
}
else{
failf(data,
"schannel: connection hostname (%s) "
"does not match certificate name (%s)",
conn->host.name,
cert_hostname);
result = CURLE_PEER_FAILED_VERIFICATION;
}
Curl_unicodefree(cert_hostname);
}
}
else {
failf(data,
"schannel: CertGetNameString did not provide any "
"certificate name information");
result = CURLE_PEER_FAILED_VERIFICATION;
}
}
}
if(pChainContext)
CertFreeCertificateChain(pChainContext);
if(pCertContextServer)
CertFreeCertificateContext(pCertContextServer);
return result;
}
#endif /* _WIN32_WCE */
static void Curl_schannel_checksum(const unsigned char *input,
size_t inputlen,
unsigned char *checksum,
size_t checksumlen,
DWORD provType,
const unsigned int algId)
{
HCRYPTPROV hProv = 0;
HCRYPTHASH hHash = 0;
DWORD cbHashSize = 0;
DWORD dwHashSizeLen = (DWORD)sizeof(cbHashSize);
DWORD dwChecksumLen = (DWORD)checksumlen;
/* since this can fail in multiple ways, zero memory first so we never
* return old data
*/
memset(checksum, 0, checksumlen);
if(!CryptAcquireContext(&hProv, NULL, NULL, provType,
CRYPT_VERIFYCONTEXT))
return; /* failed */
do {
if(!CryptCreateHash(hProv, algId, 0, 0, &hHash))
break; /* failed */
if(!CryptHashData(hHash, (const BYTE*)input, (DWORD)inputlen, 0))
break; /* failed */
/* get hash size */
if(!CryptGetHashParam(hHash, HP_HASHSIZE, (BYTE *)&cbHashSize,
&dwHashSizeLen, 0))
break; /* failed */
/* check hash size */
if(checksumlen < cbHashSize)
break; /* failed */
if(CryptGetHashParam(hHash, HP_HASHVAL, checksum, &dwChecksumLen, 0))
break; /* failed */
} while(0);
if(hHash)
CryptDestroyHash(hHash);
if(hProv)
CryptReleaseContext(hProv, 0);
}
static CURLcode Curl_schannel_md5sum(unsigned char *input,
size_t inputlen,
unsigned char *md5sum,
size_t md5len)
{
Curl_schannel_checksum(input, inputlen, md5sum, md5len,
PROV_RSA_FULL, CALG_MD5);
return CURLE_OK;
}
static CURLcode Curl_schannel_sha256sum(const unsigned char *input,
size_t inputlen,
unsigned char *sha256sum,
size_t sha256len)
{
Curl_schannel_checksum(input, inputlen, sha256sum, sha256len,
PROV_RSA_AES, CALG_SHA_256);
return CURLE_OK;
}
static void *Curl_schannel_get_internals(struct ssl_connect_data *connssl,
CURLINFO info UNUSED_PARAM)
{
(void)info;
return &BACKEND->ctxt->ctxt_handle;
}
const struct Curl_ssl Curl_ssl_schannel = {
{ CURLSSLBACKEND_SCHANNEL, "schannel" }, /* info */
0, /* have_ca_path */
1, /* have_certinfo */
1, /* have_pinnedpubkey */
0, /* have_ssl_ctx */
0, /* support_https_proxy */
vtls: encapsulate SSL backend-specific data So far, all of the SSL backends' private data has been declared as part of the ssl_connect_data struct, in one big #if .. #elif .. #endif block. This can only work as long as the SSL backend is a compile-time option, something we want to change in the next commits. Therefore, let's encapsulate the exact data needed by each SSL backend into a private struct, and let's avoid bleeding any SSL backend-specific information into urldata.h. This is also necessary to allow multiple SSL backends to be compiled in at the same time, as e.g. OpenSSL's and CyaSSL's headers cannot be included in the same .c file. To avoid too many malloc() calls, we simply append the private structs to the connectdata struct in allocate_conn(). This requires us to take extra care of alignment issues: struct fields often need to be aligned on certain boundaries e.g. 32-bit values need to be stored at addresses that divide evenly by 4 (= 32 bit / 8 bit-per-byte). We do that by assuming that no SSL backend's private data contains any fields that need to be aligned on boundaries larger than `long long` (typically 64-bit) would need. Under this assumption, we simply add a dummy field of type `long long` to the `struct connectdata` struct. This field will never be accessed but acts as a placeholder for the four instances of ssl_backend_data instead. the size of each ssl_backend_data struct is stored in the SSL backend-specific metadata, to allow allocate_conn() to know how much extra space to allocate, and how to initialize the ssl[sockindex]->backend and proxy_ssl[sockindex]->backend pointers. This would appear to be a little complicated at first, but is really necessary to encapsulate the private data of each SSL backend correctly. And we need to encapsulate thusly if we ever want to allow selecting CyaSSL and OpenSSL at runtime, as their headers cannot be included within the same .c file (there are just too many conflicting definitions and declarations for that). Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
2017-07-28 23:09:35 +03:00
sizeof(struct ssl_backend_data),
Curl_schannel_init, /* init */
Curl_schannel_cleanup, /* cleanup */
Curl_schannel_version, /* version */
Curl_none_check_cxn, /* check_cxn */
Curl_schannel_shutdown, /* shutdown */
Curl_schannel_data_pending, /* data_pending */
Curl_schannel_random, /* random */
Curl_none_cert_status_request, /* cert_status_request */
Curl_schannel_connect, /* connect */
Curl_schannel_connect_nonblocking, /* connect_nonblocking */
Curl_schannel_get_internals, /* get_internals */
Curl_schannel_close, /* close_one */
Curl_none_close_all, /* close_all */
Curl_schannel_session_free, /* session_free */
Curl_none_set_engine, /* set_engine */
Curl_none_set_engine_default, /* set_engine_default */
Curl_none_engines_list, /* engines_list */
Curl_none_false_start, /* false_start */
Curl_schannel_md5sum, /* md5sum */
Curl_schannel_sha256sum /* sha256sum */
};
#endif /* USE_SCHANNEL */