boringssl/ssl/ssl_lib.cc

2916 lines
84 KiB
C++
Raw Normal View History

/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
* All rights reserved.
*
* This package is an SSL implementation written
* by Eric Young (eay@cryptsoft.com).
* The implementation was written so as to conform with Netscapes SSL.
*
* This library is free for commercial and non-commercial use as long as
* the following conditions are aheared to. The following conditions
* apply to all code found in this distribution, be it the RC4, RSA,
* lhash, DES, etc., code; not just the SSL code. The SSL documentation
* included with this distribution is covered by the same copyright terms
* except that the holder is Tim Hudson (tjh@cryptsoft.com).
*
* Copyright remains Eric Young's, and as such any Copyright notices in
* the code are not to be removed.
* If this package is used in a product, Eric Young should be given attribution
* as the author of the parts of the library used.
* This can be in the form of a textual message at program startup or
* in documentation (online or textual) provided with the package.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* "This product includes cryptographic software written by
* Eric Young (eay@cryptsoft.com)"
* The word 'cryptographic' can be left out if the rouines from the library
* being used are not cryptographic related :-).
* 4. If you include any Windows specific code (or a derivative thereof) from
* the apps directory (application code) you must include an acknowledgement:
* "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
*
* THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* The licence and distribution terms for any publically available version or
* derivative of this code cannot be changed. i.e. this code cannot simply be
* copied and put under another distribution licence
* [including the GNU Public Licence.]
*/
/* ====================================================================
* Copyright (c) 1998-2007 The OpenSSL Project. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
*
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the
* distribution.
*
* 3. All advertising materials mentioning features or use of this
* software must display the following acknowledgment:
* "This product includes software developed by the OpenSSL Project
* for use in the OpenSSL Toolkit. (http://www.openssl.org/)"
*
* 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
* endorse or promote products derived from this software without
* prior written permission. For written permission, please contact
* openssl-core@openssl.org.
*
* 5. Products derived from this software may not be called "OpenSSL"
* nor may "OpenSSL" appear in their names without prior written
* permission of the OpenSSL Project.
*
* 6. Redistributions of any form whatsoever must retain the following
* acknowledgment:
* "This product includes software developed by the OpenSSL Project
* for use in the OpenSSL Toolkit (http://www.openssl.org/)"
*
* THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
* EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE OpenSSL PROJECT OR
* ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
* STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
* OF THE POSSIBILITY OF SUCH DAMAGE.
* ====================================================================
*
* This product includes cryptographic software written by Eric Young
* (eay@cryptsoft.com). This product includes software written by Tim
* Hudson (tjh@cryptsoft.com).
*
*/
/* ====================================================================
* Copyright 2002 Sun Microsystems, Inc. ALL RIGHTS RESERVED.
* ECC cipher suite support in OpenSSL originally developed by
* SUN MICROSYSTEMS, INC., and contributed to the OpenSSL project.
*/
/* ====================================================================
* Copyright 2005 Nokia. All rights reserved.
*
* The portions of the attached software ("Contribution") is developed by
* Nokia Corporation and is licensed pursuant to the OpenSSL open source
* license.
*
* The Contribution, originally written by Mika Kousa and Pasi Eronen of
* Nokia Corporation, consists of the "PSK" (Pre-Shared Key) ciphersuites
* support (see RFC 4279) to OpenSSL.
*
* No patent licenses or other rights except those expressly stated in
* the OpenSSL open source license shall be deemed granted or received
* expressly, by implication, estoppel, or otherwise.
*
* No assurances are provided by Nokia that the Contribution does not
* infringe the patent or other intellectual property rights of any third
* party or that the license provides you with all the necessary rights
* to make use of the Contribution.
*
* THE SOFTWARE IS PROVIDED "AS IS" WITHOUT WARRANTY OF ANY KIND. IN
* ADDITION TO THE DISCLAIMERS INCLUDED IN THE LICENSE, NOKIA
* SPECIFICALLY DISCLAIMS ANY LIABILITY FOR CLAIMS BROUGHT BY YOU OR ANY
* OTHER ENTITY BASED ON INFRINGEMENT OF INTELLECTUAL PROPERTY RIGHTS OR
* OTHERWISE. */
#include <openssl/ssl.h>
#include <assert.h>
#include <stdlib.h>
#include <string.h>
#include <openssl/bytestring.h>
#include <openssl/crypto.h>
#include <openssl/err.h>
#include <openssl/lhash.h>
#include <openssl/mem.h>
#include <openssl/rand.h>
#include "internal.h"
#include "../crypto/internal.h"
#if defined(OPENSSL_WINDOWS)
#include <sys/timeb.h>
#else
#include <sys/socket.h>
#include <sys/time.h>
#endif
BSSL_NAMESPACE_BEGIN
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
// |SSL_R_UNKNOWN_PROTOCOL| is no longer emitted, but continue to define it
// to avoid downstream churn.
OPENSSL_DECLARE_ERROR_REASON(SSL, UNKNOWN_PROTOCOL)
// The following errors are no longer emitted, but are used in nginx without
// #ifdefs.
OPENSSL_DECLARE_ERROR_REASON(SSL, BLOCK_CIPHER_PAD_IS_WRONG)
OPENSSL_DECLARE_ERROR_REASON(SSL, NO_CIPHERS_SPECIFIED)
// Some error codes are special. Ensure the make_errors.go script never
// regresses this.
static_assert(SSL_R_TLSV1_ALERT_NO_RENEGOTIATION ==
SSL_AD_NO_RENEGOTIATION + SSL_AD_REASON_OFFSET,
"alert reason code mismatch");
// kMaxHandshakeSize is the maximum size, in bytes, of a handshake message.
static const size_t kMaxHandshakeSize = (1u << 24) - 1;
static CRYPTO_EX_DATA_CLASS g_ex_data_class_ssl =
CRYPTO_EX_DATA_CLASS_INIT_WITH_APP_DATA;
static CRYPTO_EX_DATA_CLASS g_ex_data_class_ssl_ctx =
CRYPTO_EX_DATA_CLASS_INIT_WITH_APP_DATA;
bool CBBFinishArray(CBB *cbb, Array<uint8_t> *out) {
uint8_t *ptr;
size_t len;
if (!CBB_finish(cbb, &ptr, &len)) {
OPENSSL_PUT_ERROR(SSL, ERR_R_INTERNAL_ERROR);
return false;
}
out->Reset(ptr, len);
return true;
}
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
void ssl_reset_error_state(SSL *ssl) {
// Functions which use |SSL_get_error| must reset I/O and error state on
// entry.
ssl->s3->rwstate = SSL_NOTHING;
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
ERR_clear_error();
ERR_clear_system_error();
}
void ssl_set_read_error(SSL* ssl) {
ssl->s3->read_shutdown = ssl_shutdown_error;
ssl->s3->read_error.reset(ERR_save_state());
}
static bool check_read_error(const SSL *ssl) {
if (ssl->s3->read_shutdown == ssl_shutdown_error) {
ERR_restore_state(ssl->s3->read_error.get());
return false;
}
return true;
}
bool ssl_can_write(const SSL *ssl) {
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
return !SSL_in_init(ssl) || ssl->s3->hs->can_early_write;
}
bool ssl_can_read(const SSL *ssl) {
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
return !SSL_in_init(ssl) || ssl->s3->hs->can_early_read;
}
ssl_open_record_t ssl_open_handshake(SSL *ssl, size_t *out_consumed,
uint8_t *out_alert, Span<uint8_t> in) {
*out_consumed = 0;
if (!check_read_error(ssl)) {
*out_alert = 0;
return ssl_open_record_error;
}
auto ret = ssl->method->open_handshake(ssl, out_consumed, out_alert, in);
if (ret == ssl_open_record_error) {
ssl_set_read_error(ssl);
}
return ret;
}
ssl_open_record_t ssl_open_change_cipher_spec(SSL *ssl, size_t *out_consumed,
uint8_t *out_alert,
Span<uint8_t> in) {
*out_consumed = 0;
if (!check_read_error(ssl)) {
*out_alert = 0;
return ssl_open_record_error;
}
auto ret =
ssl->method->open_change_cipher_spec(ssl, out_consumed, out_alert, in);
if (ret == ssl_open_record_error) {
ssl_set_read_error(ssl);
}
return ret;
}
ssl_open_record_t ssl_open_app_data(SSL *ssl, Span<uint8_t> *out,
size_t *out_consumed, uint8_t *out_alert,
Span<uint8_t> in) {
*out_consumed = 0;
if (!check_read_error(ssl)) {
*out_alert = 0;
return ssl_open_record_error;
}
auto ret = ssl->method->open_app_data(ssl, out, out_consumed, out_alert, in);
if (ret == ssl_open_record_error) {
ssl_set_read_error(ssl);
}
return ret;
}
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
void ssl_update_cache(SSL_HANDSHAKE *hs, int mode) {
SSL *const ssl = hs->ssl;
SSL_CTX *ctx = ssl->session_ctx.get();
// Never cache sessions with empty session IDs.
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
if (ssl->s3->established_session->session_id_length == 0 ||
ssl->s3->established_session->not_resumable ||
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
(ctx->session_cache_mode & mode) != mode) {
return;
}
// Clients never use the internal session cache.
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
int use_internal_cache = ssl->server && !(ctx->session_cache_mode &
SSL_SESS_CACHE_NO_INTERNAL_STORE);
// A client may see new sessions on abbreviated handshakes if the server
// decides to renew the ticket. Once the handshake is completed, it should be
// inserted into the cache.
if (ssl->s3->established_session.get() != ssl->session.get() ||
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
(!ssl->server && hs->ticket_expected)) {
if (use_internal_cache) {
SSL_CTX_add_session(ctx, ssl->s3->established_session.get());
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
}
if (ctx->new_session_cb != NULL) {
UniquePtr<SSL_SESSION> ref = UpRef(ssl->s3->established_session);
if (ctx->new_session_cb(ssl, ref.get())) {
// |new_session_cb|'s return value signals whether it took ownership.
ref.release();
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
}
}
}
if (use_internal_cache &&
!(ctx->session_cache_mode & SSL_SESS_CACHE_NO_AUTO_CLEAR)) {
// Automatically flush the internal session cache every 255 connections.
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
int flush_cache = 0;
CRYPTO_MUTEX_lock_write(&ctx->lock);
ctx->handshakes_since_cache_flush++;
if (ctx->handshakes_since_cache_flush >= 255) {
flush_cache = 1;
ctx->handshakes_since_cache_flush = 0;
}
CRYPTO_MUTEX_unlock_write(&ctx->lock);
if (flush_cache) {
struct OPENSSL_timeval now;
ssl_get_current_time(ssl, &now);
SSL_CTX_flush_sessions(ctx, now.tv_sec);
}
}
}
static int cbb_add_hex(CBB *cbb, const uint8_t *in, size_t in_len) {
static const char hextable[] = "0123456789abcdef";
uint8_t *out;
if (!CBB_add_space(cbb, &out, in_len * 2)) {
return 0;
}
for (size_t i = 0; i < in_len; i++) {
*(out++) = (uint8_t)hextable[in[i] >> 4];
*(out++) = (uint8_t)hextable[in[i] & 0xf];
}
return 1;
}
int ssl_log_secret(const SSL *ssl, const char *label, const uint8_t *secret,
size_t secret_len) {
if (ssl->ctx->keylog_callback == NULL) {
return 1;
}
ScopedCBB cbb;
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
uint8_t *out;
size_t out_len;
if (!CBB_init(cbb.get(), strlen(label) + 1 + SSL3_RANDOM_SIZE * 2 + 1 +
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
secret_len * 2 + 1) ||
!CBB_add_bytes(cbb.get(), (const uint8_t *)label, strlen(label)) ||
!CBB_add_bytes(cbb.get(), (const uint8_t *)" ", 1) ||
!cbb_add_hex(cbb.get(), ssl->s3->client_random, SSL3_RANDOM_SIZE) ||
!CBB_add_bytes(cbb.get(), (const uint8_t *)" ", 1) ||
!cbb_add_hex(cbb.get(), secret, secret_len) ||
!CBB_add_u8(cbb.get(), 0 /* NUL */) ||
!CBB_finish(cbb.get(), &out, &out_len)) {
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
return 0;
}
ssl->ctx->keylog_callback(ssl, (const char *)out);
OPENSSL_free(out);
return 1;
}
void ssl_do_info_callback(const SSL *ssl, int type, int value) {
void (*cb)(const SSL *ssl, int type, int value) = NULL;
if (ssl->info_callback != NULL) {
cb = ssl->info_callback;
} else if (ssl->ctx->info_callback != NULL) {
cb = ssl->ctx->info_callback;
}
if (cb != NULL) {
cb(ssl, type, value);
}
}
void ssl_do_msg_callback(SSL *ssl, int is_write, int content_type,
Span<const uint8_t> in) {
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
if (ssl->msg_callback == NULL) {
return;
}
// |version| is zero when calling for |SSL3_RT_HEADER| and |SSL2_VERSION| for
// a V2ClientHello.
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
int version;
switch (content_type) {
case 0:
// V2ClientHello
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
version = SSL2_VERSION;
break;
case SSL3_RT_HEADER:
version = 0;
break;
default:
version = SSL_version(ssl);
}
ssl->msg_callback(is_write, version, content_type, in.data(), in.size(), ssl,
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
ssl->msg_callback_arg);
}
void ssl_get_current_time(const SSL *ssl, struct OPENSSL_timeval *out_clock) {
// TODO(martinkr): Change callers to |ssl_ctx_get_current_time| and drop the
// |ssl| arg from |current_time_cb| if possible.
ssl_ctx_get_current_time(ssl->ctx.get(), out_clock);
}
void ssl_ctx_get_current_time(const SSL_CTX *ctx,
struct OPENSSL_timeval *out_clock) {
if (ctx->current_time_cb != NULL) {
// TODO(davidben): Update current_time_cb to use OPENSSL_timeval. See
// https://crbug.com/boringssl/155.
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
struct timeval clock;
ctx->current_time_cb(nullptr /* ssl */, &clock);
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
if (clock.tv_sec < 0) {
assert(0);
out_clock->tv_sec = 0;
out_clock->tv_usec = 0;
} else {
out_clock->tv_sec = (uint64_t)clock.tv_sec;
out_clock->tv_usec = (uint32_t)clock.tv_usec;
}
return;
}
#if defined(BORINGSSL_UNSAFE_DETERMINISTIC_MODE)
out_clock->tv_sec = 1234;
out_clock->tv_usec = 1234;
#elif defined(OPENSSL_WINDOWS)
struct _timeb time;
_ftime(&time);
if (time.time < 0) {
assert(0);
out_clock->tv_sec = 0;
out_clock->tv_usec = 0;
} else {
out_clock->tv_sec = time.time;
out_clock->tv_usec = time.millitm * 1000;
}
#else
struct timeval clock;
gettimeofday(&clock, NULL);
if (clock.tv_sec < 0) {
assert(0);
out_clock->tv_sec = 0;
out_clock->tv_usec = 0;
} else {
out_clock->tv_sec = (uint64_t)clock.tv_sec;
out_clock->tv_usec = (uint32_t)clock.tv_usec;
}
#endif
}
void SSL_CTX_set_handoff_mode(SSL_CTX *ctx, bool on) {
ctx->handoff = on;
}
static bool ssl_can_renegotiate(const SSL *ssl) {
if (ssl->server || SSL_is_dtls(ssl)) {
return false;
}
if (ssl_protocol_version(ssl) >= TLS1_3_VERSION) {
return false;
}
// The config has already been shed.
if (!ssl->config) {
return false;
}
switch (ssl->renegotiate_mode) {
case ssl_renegotiate_ignore:
case ssl_renegotiate_never:
return false;
case ssl_renegotiate_freely:
return true;
case ssl_renegotiate_once:
return ssl->s3->total_renegotiations == 0;
}
assert(0);
return false;
}
static void ssl_maybe_shed_handshake_config(SSL *ssl) {
if (ssl->s3->hs != nullptr ||
ssl->config == nullptr ||
!ssl->config->shed_handshake_config ||
ssl_can_renegotiate(ssl)) {
return;
}
ssl->config.reset();
}
void SSL_set_handoff_mode(SSL *ssl, bool on) {
if (!ssl->config) {
return;
}
ssl->config->handoff = on;
}
BSSL_NAMESPACE_END
Move libssl's internals into the bssl namespace. This is horrible, but everything else I tried was worse. The goal with this CL is to take the extern "C" out of ssl/internal.h and move most symbols to namespace bssl, so we can start using C++ helpers and destructors without worry. Complications: - Public API functions must be extern "C" and match their declaration in ssl.h, which is unnamespaced. C++ really does not want you to interleave namespaced and unnamespaced things. One can actually write a namespaced extern "C" function, but this means, from C++'s perspective, the function is namespaced. Trying to namespace the public header would worked but ended up too deep a rabbithole. - Our STACK_OF macros do not work right in namespaces. - The typedefs for our exposed but opaque types are visible in the header files and copied into consuming projects as forward declarations. We ultimately want to give SSL a destructor, but clobbering an unnamespaced ssl_st::~ssl_st seems bad manners. - MSVC complains about ambiguous names if one typedefs SSL to bssl::SSL. This CL opts for: - ssl/*.cc must begin with #define BORINGSSL_INTERNAL_CXX_TYPES. This informs the public headers to create forward declarations which are compatible with our namespaces. - For now, C++-defined type FOO ends up at bssl::FOO with a typedef outside. Later I imagine we'll rename many of them. - Internal functions get namespace bssl, so we stop worrying about stomping the tls1_prf symbol. Exported C functions are stuck as they are. Rather than try anything weird, bite the bullet and reorder files which have a mix of public and private functions. I expect that over time, the public functions will become fairly small as we move logic to more idiomatic C++. Files without any public C functions can just be written normally. - To avoid MSVC troubles, some bssl types are renamed to CPlusPlusStyle in advance of them being made idiomatic C++. Bug: 132 Change-Id: Ic931895e117c38b14ff8d6e5a273e868796c7581 Reviewed-on: https://boringssl-review.googlesource.com/18124 Reviewed-by: David Benjamin <davidben@google.com>
2017-07-18 21:34:25 +01:00
using namespace bssl;
int SSL_library_init(void) {
CRYPTO_library_init();
return 1;
}
Switch OPENSSL_VERSION_NUMBER to 1.1.0. Although we are derived from 1.0.2, we mimic 1.1.0 in some ways around our FOO_up_ref functions and opaque libssl types. This causes some difficulties when porting third-party code as any OPENSSL_VERSION_NUMBER checks for 1.1.0 APIs we have will be wrong. Moreover, adding accessors without changing OPENSSL_VERSION_NUMBER can break external projects. It is common to implement a compatibility version of an accessor under #ifdef as a static function. This then conflicts with our headers if we, unlike OpenSSL 1.0.2, have this function. This change switches OPENSSL_VERSION_NUMBER to 1.1.0 and atomically adds enough accessors for software with 1.1.0 support already. The hope is this will unblock hiding SSL_CTX and SSL_SESSION, which will be especially useful with C++-ficiation. The cost is we will hit some growing pains as more 1.1.0 consumers enter the ecosystem and we converge on the right set of APIs to import from upstream. It does not remove any 1.0.2 APIs, so we will not require that all projects support 1.1.0. The exception is APIs which changed in 1.1.0 but did not change the function signature. Those are breaking changes. Specifically: - SSL_CTX_sess_set_get_cb is now const-correct. - X509_get0_signature is now const-correct. For C++ consumers only, this change temporarily includes an overload hack for SSL_CTX_sess_set_get_cb that keeps the old callback working. This is a workaround for Node not yet supporting OpenSSL 1.1.0. The version number is set at (the as yet unreleased) 1.1.0g to denote that this change includes https://github.com/openssl/openssl/pull/4384. Bug: 91 Change-Id: I5eeb27448a6db4c25c244afac37f9604d9608a76 Reviewed-on: https://boringssl-review.googlesource.com/10340 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: Adam Langley <agl@google.com>
2016-08-12 19:48:19 +01:00
int OPENSSL_init_ssl(uint64_t opts, const OPENSSL_INIT_SETTINGS *settings) {
CRYPTO_library_init();
return 1;
}
static uint32_t ssl_session_hash(const SSL_SESSION *sess) {
return ssl_hash_session_id(
MakeConstSpan(sess->session_id, sess->session_id_length));
}
static int ssl_session_cmp(const SSL_SESSION *a, const SSL_SESSION *b) {
if (a->session_id_length != b->session_id_length) {
return 1;
}
return OPENSSL_memcmp(a->session_id, b->session_id, a->session_id_length);
}
ssl_ctx_st::ssl_ctx_st(const SSL_METHOD *ssl_method)
: method(ssl_method->method),
x509_method(ssl_method->x509_method),
retain_only_sha256_of_client_certs(false),
quiet_shutdown(false),
ocsp_stapling_enabled(false),
signed_cert_timestamps_enabled(false),
channel_id_enabled(false),
grease_enabled(false),
allow_unknown_alpn_protos(false),
ed25519_enabled(false),
rsa_pss_rsae_certs_enabled(true),
false_start_allowed_without_alpn(false),
ignore_tls13_downgrade(false),
handoff(false),
enable_early_data(false) {
CRYPTO_MUTEX_init(&lock);
CRYPTO_new_ex_data(&ex_data);
}
ssl_ctx_st::~ssl_ctx_st() {
// Free the internal session cache. Note that this calls the caller-supplied
// remove callback, so we must do it before clearing ex_data. (See ticket
// [openssl.org #212].)
SSL_CTX_flush_sessions(this, 0);
CRYPTO_free_ex_data(&g_ex_data_class_ssl_ctx, this, &ex_data);
CRYPTO_MUTEX_cleanup(&lock);
lh_SSL_SESSION_free(sessions);
x509_method->ssl_ctx_free(this);
}
SSL_CTX *SSL_CTX_new(const SSL_METHOD *method) {
if (method == NULL) {
OPENSSL_PUT_ERROR(SSL, SSL_R_NULL_SSL_METHOD_PASSED);
return nullptr;
}
UniquePtr<SSL_CTX> ret = MakeUnique<SSL_CTX>(method);
if (!ret) {
return nullptr;
}
ret->cert = MakeUnique<CERT>(method->x509_method);
ret->sessions = lh_SSL_SESSION_new(ssl_session_hash, ssl_session_cmp);
ret->client_CA.reset(sk_CRYPTO_BUFFER_new_null());
if (ret->cert == nullptr ||
ret->sessions == nullptr ||
ret->client_CA == nullptr ||
!ret->x509_method->ssl_ctx_new(ret.get())) {
return nullptr;
}
if (!SSL_CTX_set_strict_cipher_list(ret.get(), SSL_DEFAULT_CIPHER_LIST) ||
// Lock the SSL_CTX to the specified version, for compatibility with
// legacy uses of SSL_METHOD.
!SSL_CTX_set_max_proto_version(ret.get(), method->version) ||
!SSL_CTX_set_min_proto_version(ret.get(), method->version)) {
OPENSSL_PUT_ERROR(SSL, ERR_R_INTERNAL_ERROR);
return nullptr;
}
return ret.release();
}
int SSL_CTX_up_ref(SSL_CTX *ctx) {
CRYPTO_refcount_inc(&ctx->references);
return 1;
}
void SSL_CTX_free(SSL_CTX *ctx) {
if (ctx == NULL ||
!CRYPTO_refcount_dec_and_test_zero(&ctx->references)) {
return;
}
ctx->~ssl_ctx_st();
OPENSSL_free(ctx);
}
ssl_st::ssl_st(SSL_CTX *ctx_arg)
: method(ctx_arg->method),
max_send_fragment(ctx_arg->max_send_fragment),
msg_callback(ctx_arg->msg_callback),
msg_callback_arg(ctx_arg->msg_callback_arg),
tls13_variant(ctx_arg->tls13_variant),
ctx(UpRef(ctx_arg)),
session_ctx(UpRef(ctx_arg)),
options(ctx->options),
mode(ctx->mode),
max_cert_list(ctx->max_cert_list),
server(false),
quiet_shutdown(ctx->quiet_shutdown),
enable_early_data(ctx->enable_early_data) {
CRYPTO_new_ex_data(&ex_data);
}
ssl_st::~ssl_st() {
CRYPTO_free_ex_data(&g_ex_data_class_ssl, this, &ex_data);
// |config| refers to |this|, so we must release it earlier.
config.reset();
if (method != NULL) {
method->ssl_free(this);
}
}
SSL *SSL_new(SSL_CTX *ctx) {
if (ctx == nullptr) {
OPENSSL_PUT_ERROR(SSL, SSL_R_NULL_SSL_CTX);
return nullptr;
}
UniquePtr<SSL> ssl = MakeUnique<SSL>(ctx);
if (ssl == nullptr) {
return nullptr;
}
ssl->config = MakeUnique<SSL_CONFIG>(ssl.get());
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (ssl->config == nullptr) {
return nullptr;
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
}
ssl->config->conf_min_version = ctx->conf_min_version;
ssl->config->conf_max_version = ctx->conf_max_version;
ssl->config->cert = ssl_cert_dup(ctx->cert.get());
if (ssl->config->cert == nullptr) {
return nullptr;
}
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
ssl->config->verify_mode = ctx->verify_mode;
ssl->config->verify_callback = ctx->default_verify_callback;
ssl->config->custom_verify_callback = ctx->custom_verify_callback;
ssl->config->retain_only_sha256_of_client_certs =
ctx->retain_only_sha256_of_client_certs;
if (!ssl->config->supported_group_list.CopyFrom(ctx->supported_group_list) ||
!ssl->config->alpn_client_proto_list.CopyFrom(
ctx->alpn_client_proto_list) ||
!ssl->config->verify_sigalgs.CopyFrom(ctx->verify_sigalgs)) {
return nullptr;
}
if (ctx->psk_identity_hint) {
ssl->config->psk_identity_hint.reset(
BUF_strdup(ctx->psk_identity_hint.get()));
if (ssl->config->psk_identity_hint == nullptr) {
return nullptr;
}
}
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
ssl->config->psk_client_callback = ctx->psk_client_callback;
ssl->config->psk_server_callback = ctx->psk_server_callback;
ssl->config->channel_id_enabled = ctx->channel_id_enabled;
ssl->config->channel_id_private = UpRef(ctx->channel_id_private);
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
ssl->config->signed_cert_timestamps_enabled =
ctx->signed_cert_timestamps_enabled;
ssl->config->ocsp_stapling_enabled = ctx->ocsp_stapling_enabled;
ssl->config->handoff = ctx->handoff;
ssl->config->ignore_tls13_downgrade = ctx->ignore_tls13_downgrade;
if (!ssl->method->ssl_new(ssl.get()) ||
!ssl->ctx->x509_method->ssl_new(ssl->s3->hs.get())) {
return nullptr;
}
return ssl.release();
}
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
SSL_CONFIG::SSL_CONFIG(SSL *ssl_arg)
: ssl(ssl_arg),
signed_cert_timestamps_enabled(false),
ocsp_stapling_enabled(false),
channel_id_enabled(false),
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
retain_only_sha256_of_client_certs(false),
handoff(false),
shed_handshake_config(false),
ignore_tls13_downgrade(false),
jdk11_workaround(false) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
assert(ssl);
}
SSL_CONFIG::~SSL_CONFIG() {
if (ssl->ctx != nullptr) {
ssl->ctx->x509_method->ssl_config_free(this);
}
}
void SSL_free(SSL *ssl) {
Delete(ssl);
}
void SSL_set_connect_state(SSL *ssl) {
ssl->server = false;
ssl->do_handshake = ssl_client_handshake;
}
void SSL_set_accept_state(SSL *ssl) {
ssl->server = true;
ssl->do_handshake = ssl_server_handshake;
}
void SSL_set0_rbio(SSL *ssl, BIO *rbio) {
ssl->rbio.reset(rbio);
}
void SSL_set0_wbio(SSL *ssl, BIO *wbio) {
ssl->wbio.reset(wbio);
}
void SSL_set_bio(SSL *ssl, BIO *rbio, BIO *wbio) {
// For historical reasons, this function has many different cases in ownership
// handling.
// If nothing has changed, do nothing
if (rbio == SSL_get_rbio(ssl) && wbio == SSL_get_wbio(ssl)) {
return;
}
// If the two arguments are equal, one fewer reference is granted than
// taken.
if (rbio != NULL && rbio == wbio) {
BIO_up_ref(rbio);
}
// If only the wbio is changed, adopt only one reference.
if (rbio == SSL_get_rbio(ssl)) {
SSL_set0_wbio(ssl, wbio);
return;
}
// There is an asymmetry here for historical reasons. If only the rbio is
// changed AND the rbio and wbio were originally different, then we only adopt
// one reference.
if (wbio == SSL_get_wbio(ssl) && SSL_get_rbio(ssl) != SSL_get_wbio(ssl)) {
SSL_set0_rbio(ssl, rbio);
return;
}
// Otherwise, adopt both references.
SSL_set0_rbio(ssl, rbio);
SSL_set0_wbio(ssl, wbio);
}
BIO *SSL_get_rbio(const SSL *ssl) { return ssl->rbio.get(); }
BIO *SSL_get_wbio(const SSL *ssl) { return ssl->wbio.get(); }
size_t SSL_quic_max_handshake_flight_len(const SSL *ssl,
enum ssl_encryption_level_t level) {
// Limits flights to 16K by default when there are no large
// (certificate-carrying) messages.
static const size_t kDefaultLimit = 16384;
switch (level) {
case ssl_encryption_initial:
return kDefaultLimit;
case ssl_encryption_early_data:
// QUIC does not send EndOfEarlyData.
return 0;
case ssl_encryption_handshake:
if (ssl->server) {
// Servers may receive Certificate message if configured to request
// client certificates.
if (!!(ssl->config->verify_mode & SSL_VERIFY_PEER) &&
ssl->max_cert_list > kDefaultLimit) {
return ssl->max_cert_list;
}
} else {
// Clients may receive both Certificate message and a CertificateRequest
// message.
if (2*ssl->max_cert_list > kDefaultLimit) {
return 2*ssl->max_cert_list;
}
}
return kDefaultLimit;
case ssl_encryption_application:
// Note there is not actually a bound on the number of NewSessionTickets
// one may send in a row. This level may need more involved flow
// control. See https://github.com/quicwg/base-drafts/issues/1834.
return kDefaultLimit;
}
return 0;
}
enum ssl_encryption_level_t SSL_quic_read_level(const SSL *ssl) {
return ssl->s3->read_level;
}
enum ssl_encryption_level_t SSL_quic_write_level(const SSL *ssl) {
return ssl->s3->write_level;
}
int SSL_provide_quic_data(SSL *ssl, enum ssl_encryption_level_t level,
const uint8_t *data, size_t len) {
if (ssl->ctx->quic_method == nullptr) {
OPENSSL_PUT_ERROR(SSL, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
return 0;
}
if (level != ssl->s3->read_level) {
OPENSSL_PUT_ERROR(SSL, SSL_R_WRONG_ENCRYPTION_LEVEL_RECEIVED);
return 0;
}
size_t new_len = (ssl->s3->hs_buf ? ssl->s3->hs_buf->length : 0) + len;
if (new_len < len ||
new_len > SSL_quic_max_handshake_flight_len(ssl, level)) {
OPENSSL_PUT_ERROR(SSL, SSL_R_EXCESSIVE_MESSAGE_SIZE);
return 0;
}
return tls_append_handshake_data(ssl, MakeConstSpan(data, len));
}
int SSL_do_handshake(SSL *ssl) {
ssl_reset_error_state(ssl);
Clear the error queue on entry to core SSL operations. OpenSSL historically made some poor API decisions. Rather than returning a status enum in SSL_read, etc., these functions must be paired with SSL_get_error which determines the cause of the last error's failure. This requires SSL_read communicate with SSL_get_error with some stateful flag, rwstate. Further, probably as workarounds for bugs elsewhere, SSL_get_error does not trust rwstate. Among other quirks, if the error queue is non-empty, SSL_get_error overrides rwstate and returns a value based on that. This requires that SSL_read, etc., be called with an empty error queue. (Or we hit one of the spurious ERR_clear_error calls in the handshake state machine, likely added as further self-workarounds.) Since requiring callers consistently clear the error queue everywhere is unreasonable (crbug.com/567501), clear ERR_clear_error *once* at the entry point. Until/unless[*] we make SSL_get_error sane, this is the most reasonable way to get to the point that clearing the error queue on error is optional. With those in place, the calls in the handshake state machine are no longer needed. (I suspect all the ERR_clear_system_error calls can also go, but I'll investigate and think about that separately.) [*] I'm not even sure it's possible anymore, thanks to the possibility of BIO_write pushing to the error queue. BUG=567501,593963 Change-Id: I564ace199e5a4a74b2554ad3335e99cd17120741 Reviewed-on: https://boringssl-review.googlesource.com/7455 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-03-14 18:25:46 +00:00
if (ssl->do_handshake == NULL) {
OPENSSL_PUT_ERROR(SSL, SSL_R_CONNECTION_TYPE_NOT_SET);
return -1;
}
if (!SSL_in_init(ssl)) {
return 1;
}
// Run the handshake.
SSL_HANDSHAKE *hs = ssl->s3->hs.get();
bool early_return = false;
int ret = ssl_run_handshake(hs, &early_return);
ssl_do_info_callback(
ssl, ssl->server ? SSL_CB_ACCEPT_EXIT : SSL_CB_CONNECT_EXIT, ret);
if (ret <= 0) {
return ret;
}
// Destroy the handshake object if the handshake has completely finished.
if (!early_return) {
ssl->s3->hs.reset();
ssl_maybe_shed_handshake_config(ssl);
}
return 1;
}
int SSL_connect(SSL *ssl) {
if (ssl->do_handshake == NULL) {
// Not properly initialized yet
SSL_set_connect_state(ssl);
}
Clear the error queue on entry to core SSL operations. OpenSSL historically made some poor API decisions. Rather than returning a status enum in SSL_read, etc., these functions must be paired with SSL_get_error which determines the cause of the last error's failure. This requires SSL_read communicate with SSL_get_error with some stateful flag, rwstate. Further, probably as workarounds for bugs elsewhere, SSL_get_error does not trust rwstate. Among other quirks, if the error queue is non-empty, SSL_get_error overrides rwstate and returns a value based on that. This requires that SSL_read, etc., be called with an empty error queue. (Or we hit one of the spurious ERR_clear_error calls in the handshake state machine, likely added as further self-workarounds.) Since requiring callers consistently clear the error queue everywhere is unreasonable (crbug.com/567501), clear ERR_clear_error *once* at the entry point. Until/unless[*] we make SSL_get_error sane, this is the most reasonable way to get to the point that clearing the error queue on error is optional. With those in place, the calls in the handshake state machine are no longer needed. (I suspect all the ERR_clear_system_error calls can also go, but I'll investigate and think about that separately.) [*] I'm not even sure it's possible anymore, thanks to the possibility of BIO_write pushing to the error queue. BUG=567501,593963 Change-Id: I564ace199e5a4a74b2554ad3335e99cd17120741 Reviewed-on: https://boringssl-review.googlesource.com/7455 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-03-14 18:25:46 +00:00
return SSL_do_handshake(ssl);
}
int SSL_accept(SSL *ssl) {
if (ssl->do_handshake == NULL) {
// Not properly initialized yet
SSL_set_accept_state(ssl);
}
Clear the error queue on entry to core SSL operations. OpenSSL historically made some poor API decisions. Rather than returning a status enum in SSL_read, etc., these functions must be paired with SSL_get_error which determines the cause of the last error's failure. This requires SSL_read communicate with SSL_get_error with some stateful flag, rwstate. Further, probably as workarounds for bugs elsewhere, SSL_get_error does not trust rwstate. Among other quirks, if the error queue is non-empty, SSL_get_error overrides rwstate and returns a value based on that. This requires that SSL_read, etc., be called with an empty error queue. (Or we hit one of the spurious ERR_clear_error calls in the handshake state machine, likely added as further self-workarounds.) Since requiring callers consistently clear the error queue everywhere is unreasonable (crbug.com/567501), clear ERR_clear_error *once* at the entry point. Until/unless[*] we make SSL_get_error sane, this is the most reasonable way to get to the point that clearing the error queue on error is optional. With those in place, the calls in the handshake state machine are no longer needed. (I suspect all the ERR_clear_system_error calls can also go, but I'll investigate and think about that separately.) [*] I'm not even sure it's possible anymore, thanks to the possibility of BIO_write pushing to the error queue. BUG=567501,593963 Change-Id: I564ace199e5a4a74b2554ad3335e99cd17120741 Reviewed-on: https://boringssl-review.googlesource.com/7455 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-03-14 18:25:46 +00:00
return SSL_do_handshake(ssl);
}
static int ssl_do_post_handshake(SSL *ssl, const SSLMessage &msg) {
if (ssl_protocol_version(ssl) >= TLS1_3_VERSION) {
return tls13_post_handshake(ssl, msg);
}
// Check for renegotiation on the server before parsing to use the correct
// error. Renegotiation is triggered by a different message for servers.
if (ssl->server) {
OPENSSL_PUT_ERROR(SSL, SSL_R_NO_RENEGOTIATION);
ssl_send_alert(ssl, SSL3_AL_FATAL, SSL_AD_NO_RENEGOTIATION);
return 0;
}
if (msg.type != SSL3_MT_HELLO_REQUEST || CBS_len(&msg.body) != 0) {
ssl_send_alert(ssl, SSL3_AL_FATAL, SSL_AD_DECODE_ERROR);
OPENSSL_PUT_ERROR(SSL, SSL_R_BAD_HELLO_REQUEST);
return 0;
}
if (ssl->renegotiate_mode == ssl_renegotiate_ignore) {
return 1; // Ignore the HelloRequest.
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
}
if (!ssl_can_renegotiate(ssl) ||
// Renegotiation is only supported at quiescent points in the application
// protocol, namely in HTTPS, just before reading the HTTP response.
// Require the record-layer be idle and avoid complexities of sending a
// handshake record while an application_data record is being written.
!ssl->s3->write_buffer.empty() ||
ssl->s3->write_shutdown != ssl_shutdown_none) {
OPENSSL_PUT_ERROR(SSL, SSL_R_NO_RENEGOTIATION);
ssl_send_alert(ssl, SSL3_AL_FATAL, SSL_AD_NO_RENEGOTIATION);
return 0;
}
// Begin a new handshake.
if (ssl->s3->hs != nullptr) {
OPENSSL_PUT_ERROR(SSL, ERR_R_INTERNAL_ERROR);
return 0;
}
ssl->s3->hs = ssl_handshake_new(ssl);
if (ssl->s3->hs == nullptr) {
return 0;
}
ssl->s3->total_renegotiations++;
return 1;
}
int SSL_process_quic_post_handshake(SSL *ssl) {
ssl_reset_error_state(ssl);
if (SSL_in_init(ssl)) {
OPENSSL_PUT_ERROR(SSL, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
return 0;
}
// Replay post-handshake message errors.
if (!check_read_error(ssl)) {
return 0;
}
// Process any buffered post-handshake messages.
SSLMessage msg;
while (ssl->method->get_message(ssl, &msg)) {
// Handle the post-handshake message and try again.
if (!ssl_do_post_handshake(ssl, msg)) {
ssl_set_read_error(ssl);
return 0;
}
ssl->method->next_message(ssl);
}
return 1;
}
static int ssl_read_impl(SSL *ssl) {
ssl_reset_error_state(ssl);
Clear the error queue on entry to core SSL operations. OpenSSL historically made some poor API decisions. Rather than returning a status enum in SSL_read, etc., these functions must be paired with SSL_get_error which determines the cause of the last error's failure. This requires SSL_read communicate with SSL_get_error with some stateful flag, rwstate. Further, probably as workarounds for bugs elsewhere, SSL_get_error does not trust rwstate. Among other quirks, if the error queue is non-empty, SSL_get_error overrides rwstate and returns a value based on that. This requires that SSL_read, etc., be called with an empty error queue. (Or we hit one of the spurious ERR_clear_error calls in the handshake state machine, likely added as further self-workarounds.) Since requiring callers consistently clear the error queue everywhere is unreasonable (crbug.com/567501), clear ERR_clear_error *once* at the entry point. Until/unless[*] we make SSL_get_error sane, this is the most reasonable way to get to the point that clearing the error queue on error is optional. With those in place, the calls in the handshake state machine are no longer needed. (I suspect all the ERR_clear_system_error calls can also go, but I'll investigate and think about that separately.) [*] I'm not even sure it's possible anymore, thanks to the possibility of BIO_write pushing to the error queue. BUG=567501,593963 Change-Id: I564ace199e5a4a74b2554ad3335e99cd17120741 Reviewed-on: https://boringssl-review.googlesource.com/7455 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-03-14 18:25:46 +00:00
if (ssl->do_handshake == NULL) {
OPENSSL_PUT_ERROR(SSL, SSL_R_UNINITIALIZED);
return -1;
}
// Replay post-handshake message errors.
if (!check_read_error(ssl)) {
return -1;
}
while (ssl->s3->pending_app_data.empty()) {
// Complete the current handshake, if any. False Start will cause
// |SSL_do_handshake| to return mid-handshake, so this may require multiple
// iterations.
while (!ssl_can_read(ssl)) {
int ret = SSL_do_handshake(ssl);
if (ret < 0) {
return ret;
}
if (ret == 0) {
OPENSSL_PUT_ERROR(SSL, SSL_R_SSL_HANDSHAKE_FAILURE);
return -1;
}
}
// Process any buffered post-handshake messages.
SSLMessage msg;
if (ssl->method->get_message(ssl, &msg)) {
// If we received an interrupt in early read (EndOfEarlyData), loop again
// for the handshake to process it.
if (SSL_in_init(ssl)) {
ssl->s3->hs->can_early_read = false;
continue;
}
// Handle the post-handshake message and try again.
if (!ssl_do_post_handshake(ssl, msg)) {
ssl_set_read_error(ssl);
return -1;
}
ssl->method->next_message(ssl);
continue; // Loop again. We may have begun a new handshake.
}
uint8_t alert = SSL_AD_DECODE_ERROR;
size_t consumed = 0;
auto ret = ssl_open_app_data(ssl, &ssl->s3->pending_app_data, &consumed,
&alert, ssl->s3->read_buffer.span());
bool retry;
int bio_ret = ssl_handle_open_record(ssl, &retry, ret, consumed, alert);
if (bio_ret <= 0) {
return bio_ret;
}
if (!retry) {
assert(!ssl->s3->pending_app_data.empty());
ssl->s3->key_update_count = 0;
}
}
return 1;
}
Clear the error queue on entry to core SSL operations. OpenSSL historically made some poor API decisions. Rather than returning a status enum in SSL_read, etc., these functions must be paired with SSL_get_error which determines the cause of the last error's failure. This requires SSL_read communicate with SSL_get_error with some stateful flag, rwstate. Further, probably as workarounds for bugs elsewhere, SSL_get_error does not trust rwstate. Among other quirks, if the error queue is non-empty, SSL_get_error overrides rwstate and returns a value based on that. This requires that SSL_read, etc., be called with an empty error queue. (Or we hit one of the spurious ERR_clear_error calls in the handshake state machine, likely added as further self-workarounds.) Since requiring callers consistently clear the error queue everywhere is unreasonable (crbug.com/567501), clear ERR_clear_error *once* at the entry point. Until/unless[*] we make SSL_get_error sane, this is the most reasonable way to get to the point that clearing the error queue on error is optional. With those in place, the calls in the handshake state machine are no longer needed. (I suspect all the ERR_clear_system_error calls can also go, but I'll investigate and think about that separately.) [*] I'm not even sure it's possible anymore, thanks to the possibility of BIO_write pushing to the error queue. BUG=567501,593963 Change-Id: I564ace199e5a4a74b2554ad3335e99cd17120741 Reviewed-on: https://boringssl-review.googlesource.com/7455 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-03-14 18:25:46 +00:00
int SSL_read(SSL *ssl, void *buf, int num) {
int ret = SSL_peek(ssl, buf, num);
if (ret <= 0) {
return ret;
}
// TODO(davidben): In DTLS, should the rest of the record be discarded? DTLS
// is not a stream. See https://crbug.com/boringssl/65.
ssl->s3->pending_app_data =
ssl->s3->pending_app_data.subspan(static_cast<size_t>(ret));
if (ssl->s3->pending_app_data.empty()) {
ssl->s3->read_buffer.DiscardConsumed();
}
return ret;
Clear the error queue on entry to core SSL operations. OpenSSL historically made some poor API decisions. Rather than returning a status enum in SSL_read, etc., these functions must be paired with SSL_get_error which determines the cause of the last error's failure. This requires SSL_read communicate with SSL_get_error with some stateful flag, rwstate. Further, probably as workarounds for bugs elsewhere, SSL_get_error does not trust rwstate. Among other quirks, if the error queue is non-empty, SSL_get_error overrides rwstate and returns a value based on that. This requires that SSL_read, etc., be called with an empty error queue. (Or we hit one of the spurious ERR_clear_error calls in the handshake state machine, likely added as further self-workarounds.) Since requiring callers consistently clear the error queue everywhere is unreasonable (crbug.com/567501), clear ERR_clear_error *once* at the entry point. Until/unless[*] we make SSL_get_error sane, this is the most reasonable way to get to the point that clearing the error queue on error is optional. With those in place, the calls in the handshake state machine are no longer needed. (I suspect all the ERR_clear_system_error calls can also go, but I'll investigate and think about that separately.) [*] I'm not even sure it's possible anymore, thanks to the possibility of BIO_write pushing to the error queue. BUG=567501,593963 Change-Id: I564ace199e5a4a74b2554ad3335e99cd17120741 Reviewed-on: https://boringssl-review.googlesource.com/7455 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-03-14 18:25:46 +00:00
}
Clear the error queue on entry to core SSL operations. OpenSSL historically made some poor API decisions. Rather than returning a status enum in SSL_read, etc., these functions must be paired with SSL_get_error which determines the cause of the last error's failure. This requires SSL_read communicate with SSL_get_error with some stateful flag, rwstate. Further, probably as workarounds for bugs elsewhere, SSL_get_error does not trust rwstate. Among other quirks, if the error queue is non-empty, SSL_get_error overrides rwstate and returns a value based on that. This requires that SSL_read, etc., be called with an empty error queue. (Or we hit one of the spurious ERR_clear_error calls in the handshake state machine, likely added as further self-workarounds.) Since requiring callers consistently clear the error queue everywhere is unreasonable (crbug.com/567501), clear ERR_clear_error *once* at the entry point. Until/unless[*] we make SSL_get_error sane, this is the most reasonable way to get to the point that clearing the error queue on error is optional. With those in place, the calls in the handshake state machine are no longer needed. (I suspect all the ERR_clear_system_error calls can also go, but I'll investigate and think about that separately.) [*] I'm not even sure it's possible anymore, thanks to the possibility of BIO_write pushing to the error queue. BUG=567501,593963 Change-Id: I564ace199e5a4a74b2554ad3335e99cd17120741 Reviewed-on: https://boringssl-review.googlesource.com/7455 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-03-14 18:25:46 +00:00
int SSL_peek(SSL *ssl, void *buf, int num) {
if (ssl->ctx->quic_method != nullptr) {
OPENSSL_PUT_ERROR(SSL, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
return 0;
}
int ret = ssl_read_impl(ssl);
if (ret <= 0) {
return ret;
}
if (num <= 0) {
return num;
}
size_t todo =
std::min(ssl->s3->pending_app_data.size(), static_cast<size_t>(num));
OPENSSL_memcpy(buf, ssl->s3->pending_app_data.data(), todo);
return static_cast<int>(todo);
}
int SSL_write(SSL *ssl, const void *buf, int num) {
ssl_reset_error_state(ssl);
Clear the error queue on entry to core SSL operations. OpenSSL historically made some poor API decisions. Rather than returning a status enum in SSL_read, etc., these functions must be paired with SSL_get_error which determines the cause of the last error's failure. This requires SSL_read communicate with SSL_get_error with some stateful flag, rwstate. Further, probably as workarounds for bugs elsewhere, SSL_get_error does not trust rwstate. Among other quirks, if the error queue is non-empty, SSL_get_error overrides rwstate and returns a value based on that. This requires that SSL_read, etc., be called with an empty error queue. (Or we hit one of the spurious ERR_clear_error calls in the handshake state machine, likely added as further self-workarounds.) Since requiring callers consistently clear the error queue everywhere is unreasonable (crbug.com/567501), clear ERR_clear_error *once* at the entry point. Until/unless[*] we make SSL_get_error sane, this is the most reasonable way to get to the point that clearing the error queue on error is optional. With those in place, the calls in the handshake state machine are no longer needed. (I suspect all the ERR_clear_system_error calls can also go, but I'll investigate and think about that separately.) [*] I'm not even sure it's possible anymore, thanks to the possibility of BIO_write pushing to the error queue. BUG=567501,593963 Change-Id: I564ace199e5a4a74b2554ad3335e99cd17120741 Reviewed-on: https://boringssl-review.googlesource.com/7455 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-03-14 18:25:46 +00:00
if (ssl->ctx->quic_method != nullptr) {
OPENSSL_PUT_ERROR(SSL, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
return 0;
}
if (ssl->do_handshake == NULL) {
OPENSSL_PUT_ERROR(SSL, SSL_R_UNINITIALIZED);
return -1;
}
if (ssl->s3->write_shutdown != ssl_shutdown_none) {
OPENSSL_PUT_ERROR(SSL, SSL_R_PROTOCOL_IS_SHUTDOWN);
return -1;
}
int ret = 0;
bool needs_handshake = false;
do {
// If necessary, complete the handshake implicitly.
if (!ssl_can_write(ssl)) {
ret = SSL_do_handshake(ssl);
if (ret < 0) {
return ret;
}
if (ret == 0) {
OPENSSL_PUT_ERROR(SSL, SSL_R_SSL_HANDSHAKE_FAILURE);
return -1;
}
}
ret = ssl->method->write_app_data(ssl, &needs_handshake,
(const uint8_t *)buf, num);
} while (needs_handshake);
return ret;
}
int SSL_shutdown(SSL *ssl) {
ssl_reset_error_state(ssl);
Clear the error queue on entry to core SSL operations. OpenSSL historically made some poor API decisions. Rather than returning a status enum in SSL_read, etc., these functions must be paired with SSL_get_error which determines the cause of the last error's failure. This requires SSL_read communicate with SSL_get_error with some stateful flag, rwstate. Further, probably as workarounds for bugs elsewhere, SSL_get_error does not trust rwstate. Among other quirks, if the error queue is non-empty, SSL_get_error overrides rwstate and returns a value based on that. This requires that SSL_read, etc., be called with an empty error queue. (Or we hit one of the spurious ERR_clear_error calls in the handshake state machine, likely added as further self-workarounds.) Since requiring callers consistently clear the error queue everywhere is unreasonable (crbug.com/567501), clear ERR_clear_error *once* at the entry point. Until/unless[*] we make SSL_get_error sane, this is the most reasonable way to get to the point that clearing the error queue on error is optional. With those in place, the calls in the handshake state machine are no longer needed. (I suspect all the ERR_clear_system_error calls can also go, but I'll investigate and think about that separately.) [*] I'm not even sure it's possible anymore, thanks to the possibility of BIO_write pushing to the error queue. BUG=567501,593963 Change-Id: I564ace199e5a4a74b2554ad3335e99cd17120741 Reviewed-on: https://boringssl-review.googlesource.com/7455 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-03-14 18:25:46 +00:00
if (ssl->do_handshake == NULL) {
OPENSSL_PUT_ERROR(SSL, SSL_R_UNINITIALIZED);
return -1;
}
// If we are in the middle of a handshake, silently succeed. Consumers often
// call this function before |SSL_free|, whether the handshake succeeded or
// not. We assume the caller has already handled failed handshakes.
if (SSL_in_init(ssl)) {
return 1;
}
if (ssl->quiet_shutdown) {
// Do nothing if configured not to send a close_notify.
ssl->s3->write_shutdown = ssl_shutdown_close_notify;
ssl->s3->read_shutdown = ssl_shutdown_close_notify;
return 1;
}
// This function completes in two stages. It sends a close_notify and then it
// waits for a close_notify to come in. Perform exactly one action and return
// whether or not it succeeds.
if (ssl->s3->write_shutdown != ssl_shutdown_close_notify) {
// Send a close_notify.
if (ssl_send_alert(ssl, SSL3_AL_WARNING, SSL_AD_CLOSE_NOTIFY) <= 0) {
return -1;
}
} else if (ssl->s3->alert_dispatch) {
// Finish sending the close_notify.
if (ssl->method->dispatch_alert(ssl) <= 0) {
return -1;
}
} else if (ssl->s3->read_shutdown != ssl_shutdown_close_notify) {
if (SSL_is_dtls(ssl)) {
// Bidirectional shutdown doesn't make sense for an unordered
// transport. DTLS alerts also aren't delivered reliably, so we may even
// time out because the peer never received our close_notify. Report to
// the caller that the channel has fully shut down.
if (ssl->s3->read_shutdown == ssl_shutdown_error) {
ERR_restore_state(ssl->s3->read_error.get());
return -1;
}
ssl->s3->read_shutdown = ssl_shutdown_close_notify;
} else {
// Process records until an error, close_notify, or application data.
if (ssl_read_impl(ssl) > 0) {
// We received some unexpected application data.
OPENSSL_PUT_ERROR(SSL, SSL_R_APPLICATION_DATA_ON_SHUTDOWN);
return -1;
}
if (ssl->s3->read_shutdown != ssl_shutdown_close_notify) {
return -1;
}
}
}
// Return 0 for unidirectional shutdown and 1 for bidirectional shutdown.
return ssl->s3->read_shutdown == ssl_shutdown_close_notify;
}
int SSL_send_fatal_alert(SSL *ssl, uint8_t alert) {
if (ssl->s3->alert_dispatch) {
if (ssl->s3->send_alert[0] != SSL3_AL_FATAL ||
ssl->s3->send_alert[1] != alert) {
// We are already attempting to write a different alert.
OPENSSL_PUT_ERROR(SSL, SSL_R_PROTOCOL_IS_SHUTDOWN);
return -1;
}
return ssl->method->dispatch_alert(ssl);
}
return ssl_send_alert(ssl, SSL3_AL_FATAL, alert);
}
int SSL_set_quic_transport_params(SSL *ssl, const uint8_t *params,
size_t params_len) {
return ssl->config && ssl->config->quic_transport_params.CopyFrom(
MakeConstSpan(params, params_len));
}
void SSL_get_peer_quic_transport_params(const SSL *ssl,
const uint8_t **out_params,
size_t *out_params_len) {
*out_params = ssl->s3->peer_quic_transport_params.data();
*out_params_len = ssl->s3->peer_quic_transport_params.size();
}
void SSL_CTX_set_early_data_enabled(SSL_CTX *ctx, int enabled) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
ctx->enable_early_data = !!enabled;
}
void SSL_CTX_set_tls13_variant(SSL_CTX *ctx, enum tls13_variant_t variant) {
ctx->tls13_variant = variant;
}
void SSL_set_tls13_variant(SSL *ssl, enum tls13_variant_t variant) {
ssl->tls13_variant = variant;
}
void SSL_set_early_data_enabled(SSL *ssl, int enabled) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
ssl->enable_early_data = !!enabled;
}
int SSL_in_early_data(const SSL *ssl) {
if (ssl->s3->hs == NULL) {
return 0;
}
return ssl->s3->hs->in_early_data;
}
int SSL_early_data_accepted(const SSL *ssl) {
return ssl->s3->early_data_accepted;
}
void SSL_reset_early_data_reject(SSL *ssl) {
SSL_HANDSHAKE *hs = ssl->s3->hs.get();
if (hs == NULL ||
hs->wait != ssl_hs_early_data_rejected) {
abort();
}
hs->wait = ssl_hs_ok;
hs->in_early_data = false;
hs->early_session.reset();
// Discard any unfinished writes from the perspective of |SSL_write|'s
// retry. The handshake will transparently flush out the pending record
// (discarded by the server) to keep the framing correct.
ssl->s3->wpend_pending = false;
}
static int bio_retry_reason_to_error(int reason) {
switch (reason) {
case BIO_RR_CONNECT:
return SSL_ERROR_WANT_CONNECT;
case BIO_RR_ACCEPT:
return SSL_ERROR_WANT_ACCEPT;
default:
return SSL_ERROR_SYSCALL;
}
}
int SSL_get_error(const SSL *ssl, int ret_code) {
if (ret_code > 0) {
return SSL_ERROR_NONE;
}
// Make things return SSL_ERROR_SYSCALL when doing SSL_do_handshake etc,
// where we do encode the error
uint32_t err = ERR_peek_error();
if (err != 0) {
if (ERR_GET_LIB(err) == ERR_LIB_SYS) {
return SSL_ERROR_SYSCALL;
}
return SSL_ERROR_SSL;
}
if (ret_code == 0) {
if (ssl->s3->read_shutdown == ssl_shutdown_close_notify) {
return SSL_ERROR_ZERO_RETURN;
}
// An EOF was observed which violates the protocol, and the underlying
// transport does not participate in the error queue. Bubble up to the
// caller.
return SSL_ERROR_SYSCALL;
}
switch (ssl->s3->rwstate) {
case SSL_PENDING_SESSION:
return SSL_ERROR_PENDING_SESSION;
case SSL_CERTIFICATE_SELECTION_PENDING:
return SSL_ERROR_PENDING_CERTIFICATE;
case SSL_HANDOFF:
return SSL_ERROR_HANDOFF;
case SSL_HANDBACK:
return SSL_ERROR_HANDBACK;
case SSL_READING: {
if (ssl->ctx->quic_method) {
return SSL_ERROR_WANT_READ;
}
BIO *bio = SSL_get_rbio(ssl);
if (BIO_should_read(bio)) {
return SSL_ERROR_WANT_READ;
}
if (BIO_should_write(bio)) {
// TODO(davidben): OpenSSL historically checked for writes on the read
// BIO. Can this be removed?
return SSL_ERROR_WANT_WRITE;
}
if (BIO_should_io_special(bio)) {
return bio_retry_reason_to_error(BIO_get_retry_reason(bio));
}
break;
}
case SSL_WRITING: {
BIO *bio = SSL_get_wbio(ssl);
if (BIO_should_write(bio)) {
return SSL_ERROR_WANT_WRITE;
}
if (BIO_should_read(bio)) {
// TODO(davidben): OpenSSL historically checked for reads on the write
// BIO. Can this be removed?
return SSL_ERROR_WANT_READ;
}
if (BIO_should_io_special(bio)) {
return bio_retry_reason_to_error(BIO_get_retry_reason(bio));
}
break;
}
case SSL_X509_LOOKUP:
return SSL_ERROR_WANT_X509_LOOKUP;
case SSL_CHANNEL_ID_LOOKUP:
return SSL_ERROR_WANT_CHANNEL_ID_LOOKUP;
case SSL_PRIVATE_KEY_OPERATION:
return SSL_ERROR_WANT_PRIVATE_KEY_OPERATION;
case SSL_PENDING_TICKET:
return SSL_ERROR_PENDING_TICKET;
case SSL_EARLY_DATA_REJECTED:
return SSL_ERROR_EARLY_DATA_REJECTED;
case SSL_CERTIFICATE_VERIFY:
return SSL_ERROR_WANT_CERTIFICATE_VERIFY;
}
return SSL_ERROR_SYSCALL;
}
uint32_t SSL_CTX_set_options(SSL_CTX *ctx, uint32_t options) {
ctx->options |= options;
return ctx->options;
}
uint32_t SSL_CTX_clear_options(SSL_CTX *ctx, uint32_t options) {
ctx->options &= ~options;
return ctx->options;
}
uint32_t SSL_CTX_get_options(const SSL_CTX *ctx) { return ctx->options; }
uint32_t SSL_set_options(SSL *ssl, uint32_t options) {
ssl->options |= options;
return ssl->options;
}
uint32_t SSL_clear_options(SSL *ssl, uint32_t options) {
ssl->options &= ~options;
return ssl->options;
}
uint32_t SSL_get_options(const SSL *ssl) { return ssl->options; }
uint32_t SSL_CTX_set_mode(SSL_CTX *ctx, uint32_t mode) {
ctx->mode |= mode;
return ctx->mode;
}
uint32_t SSL_CTX_clear_mode(SSL_CTX *ctx, uint32_t mode) {
ctx->mode &= ~mode;
return ctx->mode;
}
uint32_t SSL_CTX_get_mode(const SSL_CTX *ctx) { return ctx->mode; }
uint32_t SSL_set_mode(SSL *ssl, uint32_t mode) {
ssl->mode |= mode;
return ssl->mode;
}
uint32_t SSL_clear_mode(SSL *ssl, uint32_t mode) {
ssl->mode &= ~mode;
return ssl->mode;
}
uint32_t SSL_get_mode(const SSL *ssl) { return ssl->mode; }
void SSL_CTX_set0_buffer_pool(SSL_CTX *ctx, CRYPTO_BUFFER_POOL *pool) {
ctx->pool = pool;
}
int SSL_get_tls_unique(const SSL *ssl, uint8_t *out, size_t *out_len,
size_t max_out) {
*out_len = 0;
OPENSSL_memset(out, 0, max_out);
// tls-unique is not defined for TLS 1.3.
if (!ssl->s3->initial_handshake_complete ||
ssl_protocol_version(ssl) >= TLS1_3_VERSION) {
return 0;
}
// The tls-unique value is the first Finished message in the handshake, which
// is the client's in a full handshake and the server's for a resumption. See
// https://tools.ietf.org/html/rfc5929#section-3.1.
const uint8_t *finished = ssl->s3->previous_client_finished;
size_t finished_len = ssl->s3->previous_client_finished_len;
if (ssl->session != NULL) {
// tls-unique is broken for resumed sessions unless EMS is used.
if (!ssl->session->extended_master_secret) {
return 0;
}
finished = ssl->s3->previous_server_finished;
finished_len = ssl->s3->previous_server_finished_len;
}
*out_len = finished_len;
if (finished_len > max_out) {
*out_len = max_out;
}
OPENSSL_memcpy(out, finished, *out_len);
return 1;
}
static int set_session_id_context(CERT *cert, const uint8_t *sid_ctx,
size_t sid_ctx_len) {
if (sid_ctx_len > sizeof(cert->sid_ctx)) {
OPENSSL_PUT_ERROR(SSL, SSL_R_SSL_SESSION_ID_CONTEXT_TOO_LONG);
return 0;
}
static_assert(sizeof(cert->sid_ctx) < 256, "sid_ctx too large");
cert->sid_ctx_length = (uint8_t)sid_ctx_len;
OPENSSL_memcpy(cert->sid_ctx, sid_ctx, sid_ctx_len);
return 1;
}
int SSL_CTX_set_session_id_context(SSL_CTX *ctx, const uint8_t *sid_ctx,
size_t sid_ctx_len) {
return set_session_id_context(ctx->cert.get(), sid_ctx, sid_ctx_len);
}
int SSL_set_session_id_context(SSL *ssl, const uint8_t *sid_ctx,
size_t sid_ctx_len) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return 0;
}
return set_session_id_context(ssl->config->cert.get(), sid_ctx, sid_ctx_len);
}
const uint8_t *SSL_get0_session_id_context(const SSL *ssl, size_t *out_len) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
assert(ssl->config);
*out_len = 0;
return NULL;
}
*out_len = ssl->config->cert->sid_ctx_length;
return ssl->config->cert->sid_ctx;
}
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
void SSL_certs_clear(SSL *ssl) {
if (!ssl->config) {
return;
}
ssl_cert_clear_certs(ssl->config->cert.get());
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
}
int SSL_get_fd(const SSL *ssl) { return SSL_get_rfd(ssl); }
int SSL_get_rfd(const SSL *ssl) {
int ret = -1;
BIO *b = BIO_find_type(SSL_get_rbio(ssl), BIO_TYPE_DESCRIPTOR);
if (b != NULL) {
BIO_get_fd(b, &ret);
}
return ret;
}
int SSL_get_wfd(const SSL *ssl) {
int ret = -1;
BIO *b = BIO_find_type(SSL_get_wbio(ssl), BIO_TYPE_DESCRIPTOR);
if (b != NULL) {
BIO_get_fd(b, &ret);
}
return ret;
}
int SSL_set_fd(SSL *ssl, int fd) {
BIO *bio = BIO_new(BIO_s_socket());
if (bio == NULL) {
OPENSSL_PUT_ERROR(SSL, ERR_R_BUF_LIB);
return 0;
}
BIO_set_fd(bio, fd, BIO_NOCLOSE);
SSL_set_bio(ssl, bio, bio);
return 1;
}
int SSL_set_wfd(SSL *ssl, int fd) {
BIO *rbio = SSL_get_rbio(ssl);
if (rbio == NULL || BIO_method_type(rbio) != BIO_TYPE_SOCKET ||
BIO_get_fd(rbio, NULL) != fd) {
BIO *bio = BIO_new(BIO_s_socket());
if (bio == NULL) {
OPENSSL_PUT_ERROR(SSL, ERR_R_BUF_LIB);
return 0;
}
BIO_set_fd(bio, fd, BIO_NOCLOSE);
SSL_set0_wbio(ssl, bio);
} else {
// Copy the rbio over to the wbio.
BIO_up_ref(rbio);
SSL_set0_wbio(ssl, rbio);
}
return 1;
}
int SSL_set_rfd(SSL *ssl, int fd) {
BIO *wbio = SSL_get_wbio(ssl);
if (wbio == NULL || BIO_method_type(wbio) != BIO_TYPE_SOCKET ||
BIO_get_fd(wbio, NULL) != fd) {
BIO *bio = BIO_new(BIO_s_socket());
if (bio == NULL) {
OPENSSL_PUT_ERROR(SSL, ERR_R_BUF_LIB);
return 0;
}
BIO_set_fd(bio, fd, BIO_NOCLOSE);
SSL_set0_rbio(ssl, bio);
} else {
// Copy the wbio over to the rbio.
BIO_up_ref(wbio);
SSL_set0_rbio(ssl, wbio);
}
return 1;
}
static size_t copy_finished(void *out, size_t out_len, const uint8_t *in,
size_t in_len) {
if (out_len > in_len) {
out_len = in_len;
}
OPENSSL_memcpy(out, in, out_len);
return in_len;
}
size_t SSL_get_finished(const SSL *ssl, void *buf, size_t count) {
if (!ssl->s3->initial_handshake_complete ||
ssl_protocol_version(ssl) >= TLS1_3_VERSION) {
return 0;
}
if (ssl->server) {
return copy_finished(buf, count, ssl->s3->previous_server_finished,
ssl->s3->previous_server_finished_len);
}
return copy_finished(buf, count, ssl->s3->previous_client_finished,
ssl->s3->previous_client_finished_len);
}
size_t SSL_get_peer_finished(const SSL *ssl, void *buf, size_t count) {
if (!ssl->s3->initial_handshake_complete ||
ssl_protocol_version(ssl) >= TLS1_3_VERSION) {
return 0;
}
if (ssl->server) {
return copy_finished(buf, count, ssl->s3->previous_client_finished,
ssl->s3->previous_client_finished_len);
}
return copy_finished(buf, count, ssl->s3->previous_server_finished,
ssl->s3->previous_server_finished_len);
}
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
int SSL_get_verify_mode(const SSL *ssl) {
if (!ssl->config) {
assert(ssl->config);
return -1;
}
return ssl->config->verify_mode;
}
int SSL_get_extms_support(const SSL *ssl) {
// TLS 1.3 does not require extended master secret and always reports as
// supporting it.
if (!ssl->s3->have_version) {
return 0;
}
if (ssl_protocol_version(ssl) >= TLS1_3_VERSION) {
return 1;
}
// If the initial handshake completed, query the established session.
if (ssl->s3->established_session != NULL) {
return ssl->s3->established_session->extended_master_secret;
}
// Otherwise, query the in-progress handshake.
if (ssl->s3->hs != NULL) {
return ssl->s3->hs->extended_master_secret;
}
assert(0);
return 0;
}
int SSL_CTX_get_read_ahead(const SSL_CTX *ctx) { return 0; }
int SSL_get_read_ahead(const SSL *ssl) { return 0; }
int SSL_CTX_set_read_ahead(SSL_CTX *ctx, int yes) { return 1; }
int SSL_set_read_ahead(SSL *ssl, int yes) { return 1; }
int SSL_pending(const SSL *ssl) {
return static_cast<int>(ssl->s3->pending_app_data.size());
}
int SSL_CTX_check_private_key(const SSL_CTX *ctx) {
return ssl_cert_check_private_key(ctx->cert.get(),
ctx->cert->privatekey.get());
}
int SSL_check_private_key(const SSL *ssl) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return 0;
}
return ssl_cert_check_private_key(ssl->config->cert.get(),
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
ssl->config->cert->privatekey.get());
}
long SSL_get_default_timeout(const SSL *ssl) {
return SSL_DEFAULT_SESSION_TIMEOUT;
}
int SSL_renegotiate(SSL *ssl) {
// Caller-initiated renegotiation is not supported.
OPENSSL_PUT_ERROR(SSL, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
return 0;
}
int SSL_renegotiate_pending(SSL *ssl) {
return SSL_in_init(ssl) && ssl->s3->initial_handshake_complete;
}
int SSL_total_renegotiations(const SSL *ssl) {
return ssl->s3->total_renegotiations;
}
size_t SSL_CTX_get_max_cert_list(const SSL_CTX *ctx) {
return ctx->max_cert_list;
}
void SSL_CTX_set_max_cert_list(SSL_CTX *ctx, size_t max_cert_list) {
if (max_cert_list > kMaxHandshakeSize) {
max_cert_list = kMaxHandshakeSize;
}
ctx->max_cert_list = (uint32_t)max_cert_list;
}
size_t SSL_get_max_cert_list(const SSL *ssl) {
return ssl->max_cert_list;
}
void SSL_set_max_cert_list(SSL *ssl, size_t max_cert_list) {
if (max_cert_list > kMaxHandshakeSize) {
max_cert_list = kMaxHandshakeSize;
}
ssl->max_cert_list = (uint32_t)max_cert_list;
}
int SSL_CTX_set_max_send_fragment(SSL_CTX *ctx, size_t max_send_fragment) {
if (max_send_fragment < 512) {
max_send_fragment = 512;
}
if (max_send_fragment > SSL3_RT_MAX_PLAIN_LENGTH) {
max_send_fragment = SSL3_RT_MAX_PLAIN_LENGTH;
}
ctx->max_send_fragment = (uint16_t)max_send_fragment;
return 1;
}
int SSL_set_max_send_fragment(SSL *ssl, size_t max_send_fragment) {
if (max_send_fragment < 512) {
max_send_fragment = 512;
}
if (max_send_fragment > SSL3_RT_MAX_PLAIN_LENGTH) {
max_send_fragment = SSL3_RT_MAX_PLAIN_LENGTH;
}
ssl->max_send_fragment = (uint16_t)max_send_fragment;
return 1;
}
int SSL_set_mtu(SSL *ssl, unsigned mtu) {
if (!SSL_is_dtls(ssl) || mtu < dtls1_min_mtu()) {
return 0;
}
ssl->d1->mtu = mtu;
return 1;
}
int SSL_get_secure_renegotiation_support(const SSL *ssl) {
if (!ssl->s3->have_version) {
return 0;
}
return ssl_protocol_version(ssl) >= TLS1_3_VERSION ||
ssl->s3->send_connection_binding;
}
size_t SSL_CTX_sess_number(const SSL_CTX *ctx) {
MutexReadLock lock(const_cast<CRYPTO_MUTEX *>(&ctx->lock));
return lh_SSL_SESSION_num_items(ctx->sessions);
}
unsigned long SSL_CTX_sess_set_cache_size(SSL_CTX *ctx, unsigned long size) {
unsigned long ret = ctx->session_cache_size;
ctx->session_cache_size = size;
return ret;
}
unsigned long SSL_CTX_sess_get_cache_size(const SSL_CTX *ctx) {
return ctx->session_cache_size;
}
int SSL_CTX_set_session_cache_mode(SSL_CTX *ctx, int mode) {
int ret = ctx->session_cache_mode;
ctx->session_cache_mode = mode;
return ret;
}
int SSL_CTX_get_session_cache_mode(const SSL_CTX *ctx) {
return ctx->session_cache_mode;
}
int SSL_CTX_get_tlsext_ticket_keys(SSL_CTX *ctx, void *out, size_t len) {
if (out == NULL) {
return 48;
}
if (len != 48) {
OPENSSL_PUT_ERROR(SSL, SSL_R_INVALID_TICKET_KEYS_LENGTH);
return 0;
}
// The default ticket keys are initialized lazily. Trigger a key
// rotation to initialize them.
if (!ssl_ctx_rotate_ticket_encryption_key(ctx)) {
return 0;
}
uint8_t *out_bytes = reinterpret_cast<uint8_t *>(out);
MutexReadLock lock(&ctx->lock);
OPENSSL_memcpy(out_bytes, ctx->ticket_key_current->name, 16);
OPENSSL_memcpy(out_bytes + 16, ctx->ticket_key_current->hmac_key, 16);
OPENSSL_memcpy(out_bytes + 32, ctx->ticket_key_current->aes_key, 16);
return 1;
}
int SSL_CTX_set_tlsext_ticket_keys(SSL_CTX *ctx, const void *in, size_t len) {
if (in == NULL) {
return 48;
}
if (len != 48) {
OPENSSL_PUT_ERROR(SSL, SSL_R_INVALID_TICKET_KEYS_LENGTH);
return 0;
}
auto key = MakeUnique<TicketKey>();
if (!key) {
return 0;
}
const uint8_t *in_bytes = reinterpret_cast<const uint8_t *>(in);
OPENSSL_memcpy(key->name, in_bytes, 16);
OPENSSL_memcpy(key->hmac_key, in_bytes + 16, 16);
OPENSSL_memcpy(key->aes_key, in_bytes + 32, 16);
// Disable automatic key rotation for manually-configured keys. This is now
// the caller's responsibility.
key->next_rotation_tv_sec = 0;
ctx->ticket_key_current = std::move(key);
ctx->ticket_key_prev.reset();
return 1;
}
int SSL_CTX_set_tlsext_ticket_key_cb(
SSL_CTX *ctx, int (*callback)(SSL *ssl, uint8_t *key_name, uint8_t *iv,
EVP_CIPHER_CTX *ctx, HMAC_CTX *hmac_ctx,
int encrypt)) {
ctx->ticket_key_cb = callback;
return 1;
}
int SSL_CTX_set1_curves(SSL_CTX *ctx, const int *curves, size_t curves_len) {
return tls1_set_curves(&ctx->supported_group_list,
MakeConstSpan(curves, curves_len));
}
int SSL_set1_curves(SSL *ssl, const int *curves, size_t curves_len) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return 0;
}
return tls1_set_curves(&ssl->config->supported_group_list,
MakeConstSpan(curves, curves_len));
}
int SSL_CTX_set1_curves_list(SSL_CTX *ctx, const char *curves) {
return tls1_set_curves_list(&ctx->supported_group_list, curves);
}
int SSL_set1_curves_list(SSL *ssl, const char *curves) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return 0;
}
return tls1_set_curves_list(&ssl->config->supported_group_list, curves);
}
uint16_t SSL_get_curve_id(const SSL *ssl) {
// TODO(davidben): This checks the wrong session if there is a renegotiation
// in progress.
SSL_SESSION *session = SSL_get_session(ssl);
if (session == NULL) {
return 0;
}
return session->group_id;
}
int SSL_CTX_set_tmp_dh(SSL_CTX *ctx, const DH *dh) {
return 1;
}
int SSL_set_tmp_dh(SSL *ssl, const DH *dh) {
return 1;
}
STACK_OF(SSL_CIPHER) *SSL_CTX_get_ciphers(const SSL_CTX *ctx) {
return ctx->cipher_list->ciphers.get();
}
int SSL_CTX_cipher_in_group(const SSL_CTX *ctx, size_t i) {
if (i >= sk_SSL_CIPHER_num(ctx->cipher_list->ciphers.get())) {
return 0;
}
return ctx->cipher_list->in_group_flags[i];
}
STACK_OF(SSL_CIPHER) *SSL_get_ciphers(const SSL *ssl) {
if (ssl == NULL) {
return NULL;
}
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (ssl->config == NULL) {
assert(ssl->config);
return NULL;
}
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
return ssl->config->cipher_list ? ssl->config->cipher_list->ciphers.get()
: ssl->ctx->cipher_list->ciphers.get();
}
const char *SSL_get_cipher_list(const SSL *ssl, int n) {
if (ssl == NULL) {
return NULL;
}
STACK_OF(SSL_CIPHER) *sk = SSL_get_ciphers(ssl);
if (sk == NULL || n < 0 || (size_t)n >= sk_SSL_CIPHER_num(sk)) {
return NULL;
}
const SSL_CIPHER *c = sk_SSL_CIPHER_value(sk, n);
if (c == NULL) {
return NULL;
}
return c->name;
}
int SSL_CTX_set_cipher_list(SSL_CTX *ctx, const char *str) {
return ssl_create_cipher_list(&ctx->cipher_list, str, false /* not strict */);
}
int SSL_CTX_set_strict_cipher_list(SSL_CTX *ctx, const char *str) {
return ssl_create_cipher_list(&ctx->cipher_list, str, true /* strict */);
}
int SSL_set_cipher_list(SSL *ssl, const char *str) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return 0;
}
return ssl_create_cipher_list(&ssl->config->cipher_list, str,
false /* not strict */);
}
int SSL_set_strict_cipher_list(SSL *ssl, const char *str) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return 0;
}
return ssl_create_cipher_list(&ssl->config->cipher_list, str,
true /* strict */);
}
const char *SSL_get_servername(const SSL *ssl, const int type) {
if (type != TLSEXT_NAMETYPE_host_name) {
return NULL;
}
// Historically, |SSL_get_servername| was also the configuration getter
// corresponding to |SSL_set_tlsext_host_name|.
if (ssl->hostname != nullptr) {
return ssl->hostname.get();
}
return ssl->s3->hostname.get();
}
int SSL_get_servername_type(const SSL *ssl) {
if (SSL_get_servername(ssl, TLSEXT_NAMETYPE_host_name) == NULL) {
return -1;
}
return TLSEXT_NAMETYPE_host_name;
}
void SSL_CTX_set_custom_verify(
SSL_CTX *ctx, int mode,
enum ssl_verify_result_t (*callback)(SSL *ssl, uint8_t *out_alert)) {
ctx->verify_mode = mode;
ctx->custom_verify_callback = callback;
}
void SSL_set_custom_verify(
SSL *ssl, int mode,
enum ssl_verify_result_t (*callback)(SSL *ssl, uint8_t *out_alert)) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return;
}
ssl->config->verify_mode = mode;
ssl->config->custom_verify_callback = callback;
}
void SSL_CTX_enable_signed_cert_timestamps(SSL_CTX *ctx) {
ctx->signed_cert_timestamps_enabled = true;
}
void SSL_enable_signed_cert_timestamps(SSL *ssl) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return;
}
ssl->config->signed_cert_timestamps_enabled = true;
}
void SSL_CTX_enable_ocsp_stapling(SSL_CTX *ctx) {
ctx->ocsp_stapling_enabled = true;
}
void SSL_enable_ocsp_stapling(SSL *ssl) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return;
}
ssl->config->ocsp_stapling_enabled = true;
}
void SSL_get0_signed_cert_timestamp_list(const SSL *ssl, const uint8_t **out,
size_t *out_len) {
SSL_SESSION *session = SSL_get_session(ssl);
if (ssl->server || !session || !session->signed_cert_timestamp_list) {
*out_len = 0;
*out = NULL;
return;
}
*out = CRYPTO_BUFFER_data(session->signed_cert_timestamp_list.get());
*out_len = CRYPTO_BUFFER_len(session->signed_cert_timestamp_list.get());
}
void SSL_get0_ocsp_response(const SSL *ssl, const uint8_t **out,
size_t *out_len) {
SSL_SESSION *session = SSL_get_session(ssl);
if (ssl->server || !session || !session->ocsp_response) {
*out_len = 0;
*out = NULL;
return;
}
*out = CRYPTO_BUFFER_data(session->ocsp_response.get());
*out_len = CRYPTO_BUFFER_len(session->ocsp_response.get());
}
int SSL_set_tlsext_host_name(SSL *ssl, const char *name) {
ssl->hostname.reset();
if (name == nullptr) {
return 1;
}
size_t len = strlen(name);
if (len == 0 || len > TLSEXT_MAXLEN_host_name) {
OPENSSL_PUT_ERROR(SSL, SSL_R_SSL3_EXT_INVALID_SERVERNAME);
return 0;
}
ssl->hostname.reset(BUF_strdup(name));
if (ssl->hostname == nullptr) {
OPENSSL_PUT_ERROR(SSL, ERR_R_MALLOC_FAILURE);
return 0;
}
return 1;
}
int SSL_CTX_set_tlsext_servername_callback(
SSL_CTX *ctx, int (*callback)(SSL *ssl, int *out_alert, void *arg)) {
ctx->servername_callback = callback;
return 1;
}
int SSL_CTX_set_tlsext_servername_arg(SSL_CTX *ctx, void *arg) {
ctx->servername_arg = arg;
return 1;
}
int SSL_select_next_proto(uint8_t **out, uint8_t *out_len, const uint8_t *peer,
unsigned peer_len, const uint8_t *supported,
unsigned supported_len) {
const uint8_t *result;
int status;
// For each protocol in peer preference order, see if we support it.
for (unsigned i = 0; i < peer_len;) {
for (unsigned j = 0; j < supported_len;) {
if (peer[i] == supported[j] &&
OPENSSL_memcmp(&peer[i + 1], &supported[j + 1], peer[i]) == 0) {
// We found a match
result = &peer[i];
status = OPENSSL_NPN_NEGOTIATED;
goto found;
}
j += supported[j];
j++;
}
i += peer[i];
i++;
}
// There's no overlap between our protocols and the peer's list.
result = supported;
status = OPENSSL_NPN_NO_OVERLAP;
found:
*out = (uint8_t *)result + 1;
*out_len = result[0];
return status;
}
void SSL_get0_next_proto_negotiated(const SSL *ssl, const uint8_t **out_data,
unsigned *out_len) {
*out_data = ssl->s3->next_proto_negotiated.data();
*out_len = ssl->s3->next_proto_negotiated.size();
}
void SSL_CTX_set_next_protos_advertised_cb(
SSL_CTX *ctx,
int (*cb)(SSL *ssl, const uint8_t **out, unsigned *out_len, void *arg),
void *arg) {
ctx->next_protos_advertised_cb = cb;
ctx->next_protos_advertised_cb_arg = arg;
}
void SSL_CTX_set_next_proto_select_cb(
SSL_CTX *ctx, int (*cb)(SSL *ssl, uint8_t **out, uint8_t *out_len,
const uint8_t *in, unsigned in_len, void *arg),
void *arg) {
ctx->next_proto_select_cb = cb;
ctx->next_proto_select_cb_arg = arg;
}
int SSL_CTX_set_alpn_protos(SSL_CTX *ctx, const uint8_t *protos,
unsigned protos_len) {
// Note this function's calling convention is backwards.
return ctx->alpn_client_proto_list.CopyFrom(MakeConstSpan(protos, protos_len))
? 0
: 1;
}
int SSL_set_alpn_protos(SSL *ssl, const uint8_t *protos, unsigned protos_len) {
// Note this function's calling convention is backwards.
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return 1;
}
return ssl->config->alpn_client_proto_list.CopyFrom(
MakeConstSpan(protos, protos_len))
? 0
: 1;
}
void SSL_CTX_set_alpn_select_cb(SSL_CTX *ctx,
int (*cb)(SSL *ssl, const uint8_t **out,
uint8_t *out_len, const uint8_t *in,
unsigned in_len, void *arg),
void *arg) {
ctx->alpn_select_cb = cb;
ctx->alpn_select_cb_arg = arg;
}
void SSL_get0_alpn_selected(const SSL *ssl, const uint8_t **out_data,
unsigned *out_len) {
if (SSL_in_early_data(ssl) && !ssl->server) {
*out_data = ssl->s3->hs->early_session->early_alpn.data();
*out_len = ssl->s3->hs->early_session->early_alpn.size();
} else {
*out_data = ssl->s3->alpn_selected.data();
*out_len = ssl->s3->alpn_selected.size();
}
}
void SSL_CTX_set_allow_unknown_alpn_protos(SSL_CTX *ctx, int enabled) {
ctx->allow_unknown_alpn_protos = !!enabled;
}
int SSL_CTX_add_cert_compression_alg(SSL_CTX *ctx, uint16_t alg_id,
ssl_cert_compression_func_t compress,
ssl_cert_decompression_func_t decompress) {
assert(compress != nullptr || decompress != nullptr);
for (const auto *alg : ctx->cert_compression_algs.get()) {
if (alg->alg_id == alg_id) {
return 0;
}
}
UniquePtr<CertCompressionAlg> alg = MakeUnique<CertCompressionAlg>();
if (alg == nullptr) {
return 0;
}
alg->alg_id = alg_id;
alg->compress = compress;
alg->decompress = decompress;
if (ctx->cert_compression_algs == nullptr) {
ctx->cert_compression_algs.reset(sk_CertCompressionAlg_new_null());
if (ctx->cert_compression_algs == nullptr) {
return 0;
}
}
if (!PushToStack(ctx->cert_compression_algs.get(), std::move(alg))) {
if (sk_CertCompressionAlg_num(ctx->cert_compression_algs.get()) == 0) {
ctx->cert_compression_algs.reset();
}
return 0;
}
return 1;
}
void SSL_CTX_set_tls_channel_id_enabled(SSL_CTX *ctx, int enabled) {
ctx->channel_id_enabled = !!enabled;
}
int SSL_CTX_enable_tls_channel_id(SSL_CTX *ctx) {
SSL_CTX_set_tls_channel_id_enabled(ctx, 1);
return 1;
}
void SSL_set_tls_channel_id_enabled(SSL *ssl, int enabled) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return;
}
ssl->config->channel_id_enabled = !!enabled;
}
int SSL_enable_tls_channel_id(SSL *ssl) {
SSL_set_tls_channel_id_enabled(ssl, 1);
return 1;
}
static int is_p256_key(EVP_PKEY *private_key) {
const EC_KEY *ec_key = EVP_PKEY_get0_EC_KEY(private_key);
return ec_key != NULL &&
EC_GROUP_get_curve_name(EC_KEY_get0_group(ec_key)) ==
NID_X9_62_prime256v1;
}
int SSL_CTX_set1_tls_channel_id(SSL_CTX *ctx, EVP_PKEY *private_key) {
if (!is_p256_key(private_key)) {
OPENSSL_PUT_ERROR(SSL, SSL_R_CHANNEL_ID_NOT_P256);
return 0;
}
ctx->channel_id_private = UpRef(private_key);
ctx->channel_id_enabled = true;
return 1;
}
int SSL_set1_tls_channel_id(SSL *ssl, EVP_PKEY *private_key) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return 0;
}
if (!is_p256_key(private_key)) {
OPENSSL_PUT_ERROR(SSL, SSL_R_CHANNEL_ID_NOT_P256);
return 0;
}
ssl->config->channel_id_private = UpRef(private_key);
ssl->config->channel_id_enabled = true;
return 1;
}
size_t SSL_get_tls_channel_id(SSL *ssl, uint8_t *out, size_t max_out) {
if (!ssl->s3->channel_id_valid) {
return 0;
}
OPENSSL_memcpy(out, ssl->s3->channel_id, (max_out < 64) ? max_out : 64);
return 64;
}
int SSL_set_token_binding_params(SSL *ssl, const uint8_t *params, size_t len) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return 0;
}
if (len > 256) {
OPENSSL_PUT_ERROR(SSL, ERR_R_OVERFLOW);
return 0;
}
return ssl->config->token_binding_params.CopyFrom(MakeConstSpan(params, len));
}
int SSL_is_token_binding_negotiated(const SSL *ssl) {
return ssl->s3->token_binding_negotiated;
}
uint8_t SSL_get_negotiated_token_binding_param(const SSL *ssl) {
return ssl->s3->negotiated_token_binding_param;
}
size_t SSL_get0_certificate_types(const SSL *ssl, const uint8_t **out_types) {
Span<const uint8_t> types;
if (!ssl->server && ssl->s3->hs != nullptr) {
types = ssl->s3->hs->certificate_types;
}
*out_types = types.data();
return types.size();
}
size_t SSL_get0_peer_verify_algorithms(const SSL *ssl,
const uint16_t **out_sigalgs) {
Span<const uint16_t> sigalgs;
if (ssl->s3->hs != nullptr) {
sigalgs = ssl->s3->hs->peer_sigalgs;
}
*out_sigalgs = sigalgs.data();
return sigalgs.size();
}
EVP_PKEY *SSL_get_privatekey(const SSL *ssl) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
assert(ssl->config);
return NULL;
}
if (ssl->config->cert != NULL) {
return ssl->config->cert->privatekey.get();
}
return NULL;
}
EVP_PKEY *SSL_CTX_get0_privatekey(const SSL_CTX *ctx) {
if (ctx->cert != NULL) {
return ctx->cert->privatekey.get();
}
return NULL;
}
const SSL_CIPHER *SSL_get_current_cipher(const SSL *ssl) {
const SSL_SESSION *session = SSL_get_session(ssl);
return session == nullptr ? nullptr : session->cipher;
}
int SSL_session_reused(const SSL *ssl) {
return ssl->s3->session_reused || SSL_in_early_data(ssl);
}
const COMP_METHOD *SSL_get_current_compression(SSL *ssl) { return NULL; }
const COMP_METHOD *SSL_get_current_expansion(SSL *ssl) { return NULL; }
int SSL_get_server_tmp_key(SSL *ssl, EVP_PKEY **out_key) { return 0; }
void SSL_CTX_set_quiet_shutdown(SSL_CTX *ctx, int mode) {
ctx->quiet_shutdown = (mode != 0);
}
int SSL_CTX_get_quiet_shutdown(const SSL_CTX *ctx) {
return ctx->quiet_shutdown;
}
void SSL_set_quiet_shutdown(SSL *ssl, int mode) {
ssl->quiet_shutdown = (mode != 0);
}
int SSL_get_quiet_shutdown(const SSL *ssl) { return ssl->quiet_shutdown; }
void SSL_set_shutdown(SSL *ssl, int mode) {
// It is an error to clear any bits that have already been set. (We can't try
// to get a second close_notify or send two.)
assert((SSL_get_shutdown(ssl) & mode) == SSL_get_shutdown(ssl));
if (mode & SSL_RECEIVED_SHUTDOWN &&
ssl->s3->read_shutdown == ssl_shutdown_none) {
ssl->s3->read_shutdown = ssl_shutdown_close_notify;
}
if (mode & SSL_SENT_SHUTDOWN &&
ssl->s3->write_shutdown == ssl_shutdown_none) {
ssl->s3->write_shutdown = ssl_shutdown_close_notify;
}
}
int SSL_get_shutdown(const SSL *ssl) {
int ret = 0;
if (ssl->s3->read_shutdown != ssl_shutdown_none) {
// Historically, OpenSSL set |SSL_RECEIVED_SHUTDOWN| on both close_notify
// and fatal alert.
ret |= SSL_RECEIVED_SHUTDOWN;
}
if (ssl->s3->write_shutdown == ssl_shutdown_close_notify) {
// Historically, OpenSSL set |SSL_SENT_SHUTDOWN| on only close_notify.
ret |= SSL_SENT_SHUTDOWN;
}
return ret;
}
SSL_CTX *SSL_get_SSL_CTX(const SSL *ssl) { return ssl->ctx.get(); }
SSL_CTX *SSL_set_SSL_CTX(SSL *ssl, SSL_CTX *ctx) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return NULL;
}
if (ssl->ctx.get() == ctx) {
return ssl->ctx.get();
}
// One cannot change the X.509 callbacks during a connection.
if (ssl->ctx->x509_method != ctx->x509_method) {
assert(0);
return NULL;
}
UniquePtr<CERT> new_cert = ssl_cert_dup(ctx->cert.get());
if (!new_cert) {
return nullptr;
}
ssl->config->cert = std::move(new_cert);
ssl->ctx = UpRef(ctx);
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
ssl->enable_early_data = ssl->ctx->enable_early_data;
return ssl->ctx.get();
}
void SSL_set_info_callback(SSL *ssl,
void (*cb)(const SSL *ssl, int type, int value)) {
ssl->info_callback = cb;
}
void (*SSL_get_info_callback(const SSL *ssl))(const SSL *ssl, int type,
int value) {
return ssl->info_callback;
}
int SSL_state(const SSL *ssl) {
return SSL_in_init(ssl) ? SSL_ST_INIT : SSL_ST_OK;
}
void SSL_set_state(SSL *ssl, int state) { }
char *SSL_get_shared_ciphers(const SSL *ssl, char *buf, int len) {
if (len <= 0) {
return NULL;
}
buf[0] = '\0';
return buf;
}
int SSL_CTX_set_quic_method(SSL_CTX *ctx, const SSL_QUIC_METHOD *quic_method) {
if (ctx->method->is_dtls) {
return 0;
}
ctx->quic_method = quic_method;
return 1;
}
int SSL_get_ex_new_index(long argl, void *argp, CRYPTO_EX_unused *unused,
CRYPTO_EX_dup *dup_unused, CRYPTO_EX_free *free_func) {
int index;
if (!CRYPTO_get_ex_new_index(&g_ex_data_class_ssl, &index, argl, argp,
free_func)) {
return -1;
}
return index;
}
int SSL_set_ex_data(SSL *ssl, int idx, void *data) {
return CRYPTO_set_ex_data(&ssl->ex_data, idx, data);
}
void *SSL_get_ex_data(const SSL *ssl, int idx) {
return CRYPTO_get_ex_data(&ssl->ex_data, idx);
}
int SSL_CTX_get_ex_new_index(long argl, void *argp, CRYPTO_EX_unused *unused,
CRYPTO_EX_dup *dup_unused,
CRYPTO_EX_free *free_func) {
int index;
if (!CRYPTO_get_ex_new_index(&g_ex_data_class_ssl_ctx, &index, argl, argp,
free_func)) {
return -1;
}
return index;
}
int SSL_CTX_set_ex_data(SSL_CTX *ctx, int idx, void *data) {
return CRYPTO_set_ex_data(&ctx->ex_data, idx, data);
}
void *SSL_CTX_get_ex_data(const SSL_CTX *ctx, int idx) {
return CRYPTO_get_ex_data(&ctx->ex_data, idx);
}
int SSL_want(const SSL *ssl) { return ssl->s3->rwstate; }
void SSL_CTX_set_tmp_rsa_callback(SSL_CTX *ctx,
RSA *(*cb)(SSL *ssl, int is_export,
int keylength)) {}
void SSL_set_tmp_rsa_callback(SSL *ssl, RSA *(*cb)(SSL *ssl, int is_export,
int keylength)) {}
void SSL_CTX_set_tmp_dh_callback(SSL_CTX *ctx,
DH *(*cb)(SSL *ssl, int is_export,
int keylength)) {}
void SSL_set_tmp_dh_callback(SSL *ssl, DH *(*cb)(SSL *ssl, int is_export,
int keylength)) {}
static int use_psk_identity_hint(UniquePtr<char> *out,
const char *identity_hint) {
if (identity_hint != NULL && strlen(identity_hint) > PSK_MAX_IDENTITY_LEN) {
OPENSSL_PUT_ERROR(SSL, SSL_R_DATA_LENGTH_TOO_LONG);
return 0;
}
// Clear currently configured hint, if any.
out->reset();
// Treat the empty hint as not supplying one. Plain PSK makes it possible to
// send either no hint (omit ServerKeyExchange) or an empty hint, while
// ECDHE_PSK can only spell empty hint. Having different capabilities is odd,
// so we interpret empty and missing as identical.
if (identity_hint != NULL && identity_hint[0] != '\0') {
out->reset(BUF_strdup(identity_hint));
if (*out == nullptr) {
return 0;
}
}
return 1;
}
int SSL_CTX_use_psk_identity_hint(SSL_CTX *ctx, const char *identity_hint) {
return use_psk_identity_hint(&ctx->psk_identity_hint, identity_hint);
}
int SSL_use_psk_identity_hint(SSL *ssl, const char *identity_hint) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return 0;
}
return use_psk_identity_hint(&ssl->config->psk_identity_hint, identity_hint);
}
const char *SSL_get_psk_identity_hint(const SSL *ssl) {
if (ssl == NULL) {
return NULL;
}
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (ssl->config == NULL) {
assert(ssl->config);
return NULL;
}
return ssl->config->psk_identity_hint.get();
}
const char *SSL_get_psk_identity(const SSL *ssl) {
if (ssl == NULL) {
return NULL;
}
SSL_SESSION *session = SSL_get_session(ssl);
if (session == NULL) {
return NULL;
}
return session->psk_identity.get();
}
void SSL_set_psk_client_callback(
SSL *ssl, unsigned (*cb)(SSL *ssl, const char *hint, char *identity,
unsigned max_identity_len, uint8_t *psk,
unsigned max_psk_len)) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return;
}
ssl->config->psk_client_callback = cb;
}
void SSL_CTX_set_psk_client_callback(
SSL_CTX *ctx, unsigned (*cb)(SSL *ssl, const char *hint, char *identity,
unsigned max_identity_len, uint8_t *psk,
unsigned max_psk_len)) {
ctx->psk_client_callback = cb;
}
void SSL_set_psk_server_callback(
SSL *ssl, unsigned (*cb)(SSL *ssl, const char *identity, uint8_t *psk,
unsigned max_psk_len)) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return;
}
ssl->config->psk_server_callback = cb;
}
void SSL_CTX_set_psk_server_callback(
SSL_CTX *ctx, unsigned (*cb)(SSL *ssl, const char *identity,
uint8_t *psk, unsigned max_psk_len)) {
ctx->psk_server_callback = cb;
}
void SSL_CTX_set_msg_callback(SSL_CTX *ctx,
void (*cb)(int write_p, int version,
int content_type, const void *buf,
size_t len, SSL *ssl, void *arg)) {
ctx->msg_callback = cb;
}
void SSL_CTX_set_msg_callback_arg(SSL_CTX *ctx, void *arg) {
ctx->msg_callback_arg = arg;
}
void SSL_set_msg_callback(SSL *ssl,
void (*cb)(int write_p, int version, int content_type,
const void *buf, size_t len, SSL *ssl,
void *arg)) {
ssl->msg_callback = cb;
}
void SSL_set_msg_callback_arg(SSL *ssl, void *arg) {
ssl->msg_callback_arg = arg;
}
void SSL_CTX_set_keylog_callback(SSL_CTX *ctx,
void (*cb)(const SSL *ssl, const char *line)) {
ctx->keylog_callback = cb;
}
void (*SSL_CTX_get_keylog_callback(const SSL_CTX *ctx))(const SSL *ssl,
const char *line) {
return ctx->keylog_callback;
}
void SSL_CTX_set_current_time_cb(SSL_CTX *ctx,
void (*cb)(const SSL *ssl,
struct timeval *out_clock)) {
ctx->current_time_cb = cb;
}
int SSL_is_init_finished(const SSL *ssl) {
return !SSL_in_init(ssl);
}
int SSL_in_init(const SSL *ssl) {
// This returns false once all the handshake state has been finalized, to
// allow callbacks and getters based on SSL_in_init to return the correct
// values.
SSL_HANDSHAKE *hs = ssl->s3->hs.get();
return hs != nullptr && !hs->handshake_finalized;
}
int SSL_in_false_start(const SSL *ssl) {
if (ssl->s3->hs == NULL) {
return 0;
}
return ssl->s3->hs->in_false_start;
}
int SSL_cutthrough_complete(const SSL *ssl) {
return SSL_in_false_start(ssl);
}
void SSL_get_structure_sizes(size_t *ssl_size, size_t *ssl_ctx_size,
size_t *ssl_session_size) {
*ssl_size = sizeof(SSL);
*ssl_ctx_size = sizeof(SSL_CTX);
*ssl_session_size = sizeof(SSL_SESSION);
}
int SSL_is_server(const SSL *ssl) { return ssl->server; }
int SSL_is_dtls(const SSL *ssl) { return ssl->method->is_dtls; }
void SSL_CTX_set_select_certificate_cb(
SSL_CTX *ctx,
enum ssl_select_cert_result_t (*cb)(const SSL_CLIENT_HELLO *)) {
ctx->select_certificate_cb = cb;
}
void SSL_CTX_set_dos_protection_cb(SSL_CTX *ctx,
int (*cb)(const SSL_CLIENT_HELLO *)) {
ctx->dos_protection_cb = cb;
}
void SSL_CTX_set_reverify_on_resume(SSL_CTX *ctx, int enabled) {
ctx->reverify_on_resume = !!enabled;
}
void SSL_set_renegotiate_mode(SSL *ssl, enum ssl_renegotiate_mode_t mode) {
ssl->renegotiate_mode = mode;
// Check if |ssl_can_renegotiate| has changed and the configuration may now be
// shed. HTTP clients may initially allow renegotiation for HTTP/1.1, and then
// disable after the handshake once the ALPN protocol is known to be HTTP/2.
ssl_maybe_shed_handshake_config(ssl);
}
int SSL_get_ivs(const SSL *ssl, const uint8_t **out_read_iv,
const uint8_t **out_write_iv, size_t *out_iv_len) {
size_t write_iv_len;
if (!ssl->s3->aead_read_ctx->GetIV(out_read_iv, out_iv_len) ||
!ssl->s3->aead_write_ctx->GetIV(out_write_iv, &write_iv_len) ||
*out_iv_len != write_iv_len) {
return 0;
}
return 1;
}
static uint64_t be_to_u64(const uint8_t in[8]) {
return (((uint64_t)in[0]) << 56) | (((uint64_t)in[1]) << 48) |
(((uint64_t)in[2]) << 40) | (((uint64_t)in[3]) << 32) |
(((uint64_t)in[4]) << 24) | (((uint64_t)in[5]) << 16) |
(((uint64_t)in[6]) << 8) | ((uint64_t)in[7]);
}
uint64_t SSL_get_read_sequence(const SSL *ssl) {
// TODO(davidben): Internally represent sequence numbers as uint64_t.
if (SSL_is_dtls(ssl)) {
// max_seq_num already includes the epoch.
assert(ssl->d1->r_epoch == (ssl->d1->bitmap.max_seq_num >> 48));
return ssl->d1->bitmap.max_seq_num;
}
return be_to_u64(ssl->s3->read_sequence);
}
uint64_t SSL_get_write_sequence(const SSL *ssl) {
uint64_t ret = be_to_u64(ssl->s3->write_sequence);
if (SSL_is_dtls(ssl)) {
assert((ret >> 48) == 0);
ret |= ((uint64_t)ssl->d1->w_epoch) << 48;
}
return ret;
}
uint16_t SSL_get_peer_signature_algorithm(const SSL *ssl) {
// TODO(davidben): This checks the wrong session if there is a renegotiation
// in progress.
SSL_SESSION *session = SSL_get_session(ssl);
if (session == NULL) {
return 0;
}
return session->peer_signature_algorithm;
}
size_t SSL_get_client_random(const SSL *ssl, uint8_t *out, size_t max_out) {
if (max_out == 0) {
return sizeof(ssl->s3->client_random);
}
if (max_out > sizeof(ssl->s3->client_random)) {
max_out = sizeof(ssl->s3->client_random);
}
OPENSSL_memcpy(out, ssl->s3->client_random, max_out);
return max_out;
}
size_t SSL_get_server_random(const SSL *ssl, uint8_t *out, size_t max_out) {
if (max_out == 0) {
return sizeof(ssl->s3->server_random);
}
if (max_out > sizeof(ssl->s3->server_random)) {
max_out = sizeof(ssl->s3->server_random);
}
OPENSSL_memcpy(out, ssl->s3->server_random, max_out);
return max_out;
}
const SSL_CIPHER *SSL_get_pending_cipher(const SSL *ssl) {
SSL_HANDSHAKE *hs = ssl->s3->hs.get();
if (hs == NULL) {
return NULL;
}
return hs->new_cipher;
}
void SSL_set_retain_only_sha256_of_client_certs(SSL *ssl, int enabled) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return;
}
ssl->config->retain_only_sha256_of_client_certs = !!enabled;
}
void SSL_CTX_set_retain_only_sha256_of_client_certs(SSL_CTX *ctx, int enabled) {
ctx->retain_only_sha256_of_client_certs = !!enabled;
}
void SSL_CTX_set_grease_enabled(SSL_CTX *ctx, int enabled) {
ctx->grease_enabled = !!enabled;
}
int32_t SSL_get_ticket_age_skew(const SSL *ssl) {
return ssl->s3->ticket_age_skew;
}
void SSL_CTX_set_false_start_allowed_without_alpn(SSL_CTX *ctx, int allowed) {
ctx->false_start_allowed_without_alpn = !!allowed;
}
int SSL_is_tls13_downgrade(const SSL *ssl) { return ssl->s3->tls13_downgrade; }
void SSL_CTX_set_ignore_tls13_downgrade(SSL_CTX *ctx, int ignore) {
ctx->ignore_tls13_downgrade = !!ignore;
}
void SSL_set_ignore_tls13_downgrade(SSL *ssl, int ignore) {
if (!ssl->config) {
return;
}
ssl->config->ignore_tls13_downgrade = !!ignore;
}
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
void SSL_set_shed_handshake_config(SSL *ssl, int enable) {
if (!ssl->config) {
return;
}
ssl->config->shed_handshake_config = !!enable;
}
void SSL_set_jdk11_workaround(SSL *ssl, int enable) {
if (!ssl->config) {
return;
}
ssl->config->jdk11_workaround = !!enable;
}
int SSL_clear(SSL *ssl) {
SSL_CONFIG: new struct for sheddable handshake configuration. |SSL_CONFIG| is a container for bits of configuration that are unneeded after the handshake completes. By default it is retained for the life of the |SSL|, but it may be shed at the caller's option by calling SSL_set_shed_handshake_config(). This is incompatible with renegotiation, and with SSL_clear(). |SSL_CONFIG| is reachable by |ssl->config| and by |hs->config|. The latter is always non-NULL. To avoid null checks, I've changed the signature of a number of functions from |SSL*| arguments to |SSL_HANDSHAKE*| arguments. When configuration has been shed, setters that touch |SSL_CONFIG| return an error value if that is possible. Setters that return |void| do nothing. Getters that request |SSL_CONFIG| values will fail with an |assert| if the configuration has been shed. When asserts are compiled out, they will return an error value. The aim of this commit is to simplify analysis of split-handshakes by making it obvious that some bits of state have no effects beyond the handshake. It also cuts down on memory usage. Of note: |SSL_CTX| is still reachable after the configuration has been shed, and a couple things need to be retained only for the sake of post-handshake hooks. Perhaps these can be fixed in time. Change-Id: Idf09642e0518945b81a1e9fcd7331cc9cf7cc2d6 Bug: 123 Reviewed-on: https://boringssl-review.googlesource.com/27644 Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2018-04-13 23:51:30 +01:00
if (!ssl->config) {
return 0; // SSL_clear may not be used after shedding config.
}
// In OpenSSL, reusing a client |SSL| with |SSL_clear| causes the previously
// established session to be offered the next time around. wpa_supplicant
// depends on this behavior, so emulate it.
UniquePtr<SSL_SESSION> session;
if (!ssl->server && ssl->s3->established_session != NULL) {
session = UpRef(ssl->s3->established_session);
}
// The ssl->d1->mtu is simultaneously configuration (preserved across
// clear) and connection-specific state (gets reset).
//
// TODO(davidben): Avoid this.
unsigned mtu = 0;
if (ssl->d1 != NULL) {
mtu = ssl->d1->mtu;
}
ssl->method->ssl_free(ssl);
if (!ssl->method->ssl_new(ssl)) {
return 0;
}
if (SSL_is_dtls(ssl) && (SSL_get_options(ssl) & SSL_OP_NO_QUERY_MTU)) {
ssl->d1->mtu = mtu;
}
if (session != nullptr) {
SSL_set_session(ssl, session.get());
}
return 1;
}
int SSL_CTX_sess_connect(const SSL_CTX *ctx) { return 0; }
int SSL_CTX_sess_connect_good(const SSL_CTX *ctx) { return 0; }
int SSL_CTX_sess_connect_renegotiate(const SSL_CTX *ctx) { return 0; }
int SSL_CTX_sess_accept(const SSL_CTX *ctx) { return 0; }
int SSL_CTX_sess_accept_renegotiate(const SSL_CTX *ctx) { return 0; }
int SSL_CTX_sess_accept_good(const SSL_CTX *ctx) { return 0; }
int SSL_CTX_sess_hits(const SSL_CTX *ctx) { return 0; }
int SSL_CTX_sess_cb_hits(const SSL_CTX *ctx) { return 0; }
int SSL_CTX_sess_misses(const SSL_CTX *ctx) { return 0; }
int SSL_CTX_sess_timeouts(const SSL_CTX *ctx) { return 0; }
int SSL_CTX_sess_cache_full(const SSL_CTX *ctx) { return 0; }
int SSL_num_renegotiations(const SSL *ssl) {
return SSL_total_renegotiations(ssl);
}
int SSL_CTX_need_tmp_RSA(const SSL_CTX *ctx) { return 0; }
int SSL_need_tmp_RSA(const SSL *ssl) { return 0; }
int SSL_CTX_set_tmp_rsa(SSL_CTX *ctx, const RSA *rsa) { return 1; }
int SSL_set_tmp_rsa(SSL *ssl, const RSA *rsa) { return 1; }
void ERR_load_SSL_strings(void) {}
void SSL_load_error_strings(void) {}
int SSL_cache_hit(SSL *ssl) { return SSL_session_reused(ssl); }
int SSL_CTX_set_tmp_ecdh(SSL_CTX *ctx, const EC_KEY *ec_key) {
if (ec_key == NULL || EC_KEY_get0_group(ec_key) == NULL) {
OPENSSL_PUT_ERROR(SSL, ERR_R_PASSED_NULL_PARAMETER);
return 0;
}
int nid = EC_GROUP_get_curve_name(EC_KEY_get0_group(ec_key));
return SSL_CTX_set1_curves(ctx, &nid, 1);
}
int SSL_set_tmp_ecdh(SSL *ssl, const EC_KEY *ec_key) {
if (ec_key == NULL || EC_KEY_get0_group(ec_key) == NULL) {
OPENSSL_PUT_ERROR(SSL, ERR_R_PASSED_NULL_PARAMETER);
return 0;
}
int nid = EC_GROUP_get_curve_name(EC_KEY_get0_group(ec_key));
return SSL_set1_curves(ssl, &nid, 1);
}
void SSL_CTX_set_ticket_aead_method(SSL_CTX *ctx,
const SSL_TICKET_AEAD_METHOD *aead_method) {
ctx->ticket_aead_method = aead_method;
}
int SSL_set_tlsext_status_type(SSL *ssl, int type) {
if (!ssl->config) {
return 0;
}
ssl->config->ocsp_stapling_enabled = type == TLSEXT_STATUSTYPE_ocsp;
return 1;
}
int SSL_get_tlsext_status_type(const SSL *ssl) {
if (ssl->server) {
SSL_HANDSHAKE *hs = ssl->s3->hs.get();
return hs != nullptr && hs->ocsp_stapling_requested
? TLSEXT_STATUSTYPE_ocsp
: TLSEXT_STATUSTYPE_nothing;
}
return ssl->config != nullptr && ssl->config->ocsp_stapling_enabled
? TLSEXT_STATUSTYPE_ocsp
: TLSEXT_STATUSTYPE_nothing;
}
int SSL_set_tlsext_status_ocsp_resp(SSL *ssl, uint8_t *resp, size_t resp_len) {
if (SSL_set_ocsp_response(ssl, resp, resp_len)) {
OPENSSL_free(resp);
return 1;
}
return 0;
}
size_t SSL_get_tlsext_status_ocsp_resp(const SSL *ssl, const uint8_t **out) {
size_t ret;
SSL_get0_ocsp_response(ssl, out, &ret);
return ret;
}
int SSL_CTX_set_tlsext_status_cb(SSL_CTX *ctx,
int (*callback)(SSL *ssl, void *arg)) {
ctx->legacy_ocsp_callback = callback;
return 1;
}
int SSL_CTX_set_tlsext_status_arg(SSL_CTX *ctx, void *arg) {
ctx->legacy_ocsp_callback_arg = arg;
return 1;
}