kris/boringssl
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
44bedc348d
boringssl/ssl/tls_record.c

472 lines
17 KiB
C
Raw Normal View History

Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
/* 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) 1998-2002 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). */
#include <openssl/ssl.h>
#include <assert.h>
Tidy up keyblock and CCS logic slightly. Move the actual SSL_AEAD_CTX swap into the record layer. Also revise the intermediate state we store between setup_key_block and change_cipher_state. With SSL_AEAD_CTX_new abstracted out, keeping the EVP_AEAD around doesn't make much sense. Just store enough to partition the key block. Change-Id: I773fb46a2cb78fa570f00c0a89339c15bbb1d719 Reviewed-on: https://boringssl-review.googlesource.com/6832 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:34:23 +00:00
#include <string.h>
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
#include <openssl/bytestring.h>
#include <openssl/err.h>
Push alert handling down into the record functions. Alert handling is more-or-less identical across all contexts. Push it down from read_bytes into the low-level record functions. This also deduplicates the code shared between TLS and DTLS. Now the only type mismatch managed by read_bytes is if we get handshake data in read_app_data. Change-Id: Ia8331897b304566e66d901899cfbf31d2870194e Reviewed-on: https://boringssl-review.googlesource.com/8124 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-02 20:42:01 +01:00
#include <openssl/mem.h>
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
#include "internal.h"
Remove in-place TLS record assembly for now. Decrypting is very easy to do in-place, but encrypting in-place is a hassle. The rules actually were wrong due to record-splitting. The aliasing prefix and the alignment prefix actually differ by 1. Take it out for now in preparation for tightening the aliasing rules. If we decide to do in-place encrypt later, probably it'd be more useful to return header + in-place ciphertext + trailer. (That, in turn, needs a scatter/gather thing on the AEAD thanks to TLS 1.3's padding and record type construction.) We may also wish to rethink how record-splitting works here. Change-Id: I0187d39c541e76ef933b7c2c193323164fd8a156 Reviewed-on: https://boringssl-review.googlesource.com/8230 Reviewed-by: Adam Langley <agl@google.com>
2016-06-08 23:31:24 +01:00
#include "../crypto/internal.h"
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
Reject empty records of unexpected type. The old empty record logic discarded the records at a very low-level. Let the error bubble up to ssl3_read_bytes so the type mismatch logic may kick in before the empty record is skipped. Add tests for when the record in question is application data, before before the handshake and post ChangeCipherSpec. BUG=521840 Change-Id: I47dff389cda65d6672b9be39d7d89490331063fa Reviewed-on: https://boringssl-review.googlesource.com/5754 Reviewed-by: Adam Langley <agl@google.com>
2015-08-22 06:35:43 +01:00
/* kMaxEmptyRecords is the number of consecutive, empty records that will be
* processed. Without this limit an attacker could send empty records at a
* faster rate than we can process and cause record processing to loop
* forever. */
static const uint8_t kMaxEmptyRecords = 32;
Push alert handling down into the record functions. Alert handling is more-or-less identical across all contexts. Push it down from read_bytes into the low-level record functions. This also deduplicates the code shared between TLS and DTLS. Now the only type mismatch managed by read_bytes is if we get handshake data in read_app_data. Change-Id: Ia8331897b304566e66d901899cfbf31d2870194e Reviewed-on: https://boringssl-review.googlesource.com/8124 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-02 20:42:01 +01:00
/* kMaxWarningAlerts is the number of consecutive warning alerts that will be
* processed. */
static const uint8_t kMaxWarningAlerts = 4;
Remove |need_record_splitting| from |SSL3_STATE|. It is redundant given the other state in the connection. Change-Id: I5dc71627132659ab4316a5ea360c9ca480fb7c6c Reviewed-on: https://boringssl-review.googlesource.com/6646 Reviewed-by: Adam Langley <agl@google.com>
2015-12-06 21:07:47 +00:00
/* ssl_needs_record_splitting returns one if |ssl|'s current outgoing cipher
* state needs record-splitting and zero otherwise. */
static int ssl_needs_record_splitting(const SSL *ssl) {
Replace enc_flags with normalized version checks. This removes the various non-PRF checks from SSL3_ENC_METHOD so that can have a clearer purpose. It also makes TLS 1.0 through 1.2's SSL3_ENC_METHOD tables identical and gives us an assert to ensure nothing accesses the version bits before version negotiation. Accordingly, ssl_needs_record_splitting was reordered slightly so we don't rely on enc_method being initialized to TLS 1.2 pre-version-negotiation. This leaves alert_value as the only part of SSL3_ENC_METHOD which may be accessed before version negotiation. Change-Id: If9e299e2ef5511b5fa442b2af654eed054c3e675 Reviewed-on: https://boringssl-review.googlesource.com/6842 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-30 05:08:49 +00:00
return ssl->s3->aead_write_ctx != NULL &&
ssl3_protocol_version(ssl) < TLS1_1_VERSION &&
Remove |need_record_splitting| from |SSL3_STATE|. It is redundant given the other state in the connection. Change-Id: I5dc71627132659ab4316a5ea360c9ca480fb7c6c Reviewed-on: https://boringssl-review.googlesource.com/6646 Reviewed-by: Adam Langley <agl@google.com>
2015-12-06 21:07:47 +00:00
(ssl->mode & SSL_MODE_CBC_RECORD_SPLITTING) != 0 &&
Move aead_{read,write}_ctx and next_proto_negotiated into ssl->s3. Both are connection state rather than configuration state. Notably this cuts down more of SSL_clear that can't just use ssl_free + ssl_new. Change-Id: I3c05b3ae86d4db8bd75f1cd21656f57fc5b55ca9 Reviewed-on: https://boringssl-review.googlesource.com/6835 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:56:49 +00:00
SSL_CIPHER_is_block_cipher(ssl->s3->aead_write_ctx->cipher);
Remove |need_record_splitting| from |SSL3_STATE|. It is redundant given the other state in the connection. Change-Id: I5dc71627132659ab4316a5ea360c9ca480fb7c6c Reviewed-on: https://boringssl-review.googlesource.com/6646 Reviewed-by: Adam Langley <agl@google.com>
2015-12-06 21:07:47 +00:00
}
Move ssl3_record_sequence_update with the other record-layer bits. Change-Id: I045a4d3e304872b8c97231dcde5bca7753a878fb Reviewed-on: https://boringssl-review.googlesource.com/6831 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:11:39 +00:00
int ssl_record_sequence_update(uint8_t *seq, size_t seq_len) {
size_t i;
for (i = seq_len - 1; i < seq_len; i--) {
++seq[i];
if (seq[i] != 0) {
return 1;
}
}
OPENSSL_PUT_ERROR(SSL, ERR_R_OVERFLOW);
return 0;
}
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
size_t ssl_record_prefix_len(const SSL *ssl) {
if (SSL_IS_DTLS(ssl)) {
return DTLS1_RT_HEADER_LENGTH +
Move aead_{read,write}_ctx and next_proto_negotiated into ssl->s3. Both are connection state rather than configuration state. Notably this cuts down more of SSL_clear that can't just use ssl_free + ssl_new. Change-Id: I3c05b3ae86d4db8bd75f1cd21656f57fc5b55ca9 Reviewed-on: https://boringssl-review.googlesource.com/6835 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:56:49 +00:00
SSL_AEAD_CTX_explicit_nonce_len(ssl->s3->aead_read_ctx);
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
} else {
return SSL3_RT_HEADER_LENGTH +
Move aead_{read,write}_ctx and next_proto_negotiated into ssl->s3. Both are connection state rather than configuration state. Notably this cuts down more of SSL_clear that can't just use ssl_free + ssl_new. Change-Id: I3c05b3ae86d4db8bd75f1cd21656f57fc5b55ca9 Reviewed-on: https://boringssl-review.googlesource.com/6835 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:56:49 +00:00
SSL_AEAD_CTX_explicit_nonce_len(ssl->s3->aead_read_ctx);
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
}
}
Remove in-place TLS record assembly for now. Decrypting is very easy to do in-place, but encrypting in-place is a hassle. The rules actually were wrong due to record-splitting. The aliasing prefix and the alignment prefix actually differ by 1. Take it out for now in preparation for tightening the aliasing rules. If we decide to do in-place encrypt later, probably it'd be more useful to return header + in-place ciphertext + trailer. (That, in turn, needs a scatter/gather thing on the AEAD thanks to TLS 1.3's padding and record type construction.) We may also wish to rethink how record-splitting works here. Change-Id: I0187d39c541e76ef933b7c2c193323164fd8a156 Reviewed-on: https://boringssl-review.googlesource.com/8230 Reviewed-by: Adam Langley <agl@google.com>
2016-06-08 23:31:24 +01:00
size_t ssl_seal_align_prefix_len(const SSL *ssl) {
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
if (SSL_IS_DTLS(ssl)) {
return DTLS1_RT_HEADER_LENGTH +
Move aead_{read,write}_ctx and next_proto_negotiated into ssl->s3. Both are connection state rather than configuration state. Notably this cuts down more of SSL_clear that can't just use ssl_free + ssl_new. Change-Id: I3c05b3ae86d4db8bd75f1cd21656f57fc5b55ca9 Reviewed-on: https://boringssl-review.googlesource.com/6835 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:56:49 +00:00
SSL_AEAD_CTX_explicit_nonce_len(ssl->s3->aead_write_ctx);
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
} else {
size_t ret = SSL3_RT_HEADER_LENGTH +
Move aead_{read,write}_ctx and next_proto_negotiated into ssl->s3. Both are connection state rather than configuration state. Notably this cuts down more of SSL_clear that can't just use ssl_free + ssl_new. Change-Id: I3c05b3ae86d4db8bd75f1cd21656f57fc5b55ca9 Reviewed-on: https://boringssl-review.googlesource.com/6835 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:56:49 +00:00
SSL_AEAD_CTX_explicit_nonce_len(ssl->s3->aead_write_ctx);
Remove |need_record_splitting| from |SSL3_STATE|. It is redundant given the other state in the connection. Change-Id: I5dc71627132659ab4316a5ea360c9ca480fb7c6c Reviewed-on: https://boringssl-review.googlesource.com/6646 Reviewed-by: Adam Langley <agl@google.com>
2015-12-06 21:07:47 +00:00
if (ssl_needs_record_splitting(ssl)) {
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
ret += SSL3_RT_HEADER_LENGTH;
Move aead_{read,write}_ctx and next_proto_negotiated into ssl->s3. Both are connection state rather than configuration state. Notably this cuts down more of SSL_clear that can't just use ssl_free + ssl_new. Change-Id: I3c05b3ae86d4db8bd75f1cd21656f57fc5b55ca9 Reviewed-on: https://boringssl-review.googlesource.com/6835 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:56:49 +00:00
ret += ssl_cipher_get_record_split_len(ssl->s3->aead_write_ctx->cipher);
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
}
return ret;
}
}
size_t ssl_max_seal_overhead(const SSL *ssl) {
Adding TLS 1.3 Record Layer. In TLS 1.3, the actual record type is hidden within the encrypted data and the record layer defaults to using a TLS 1.0 {3, 1} record version for compatibility. Additionally the record layer no longer checks the minor version of the record layer to maintain compatibility with the TLS 1.3 spec. Change-Id: If2c08e48baab170c1658e0715c33929d36c9be3a Reviewed-on: https://boringssl-review.googlesource.com/8091 Reviewed-by: David Benjamin <davidben@google.com>
2016-06-01 19:07:09 +01:00
size_t ret = SSL_AEAD_CTX_max_overhead(ssl->s3->aead_write_ctx);
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
if (SSL_IS_DTLS(ssl)) {
Adding TLS 1.3 Record Layer. In TLS 1.3, the actual record type is hidden within the encrypted data and the record layer defaults to using a TLS 1.0 {3, 1} record version for compatibility. Additionally the record layer no longer checks the minor version of the record layer to maintain compatibility with the TLS 1.3 spec. Change-Id: If2c08e48baab170c1658e0715c33929d36c9be3a Reviewed-on: https://boringssl-review.googlesource.com/8091 Reviewed-by: David Benjamin <davidben@google.com>
2016-06-01 19:07:09 +01:00
ret += DTLS1_RT_HEADER_LENGTH;
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
} else {
Adding TLS 1.3 Record Layer. In TLS 1.3, the actual record type is hidden within the encrypted data and the record layer defaults to using a TLS 1.0 {3, 1} record version for compatibility. Additionally the record layer no longer checks the minor version of the record layer to maintain compatibility with the TLS 1.3 spec. Change-Id: If2c08e48baab170c1658e0715c33929d36c9be3a Reviewed-on: https://boringssl-review.googlesource.com/8091 Reviewed-by: David Benjamin <davidben@google.com>
2016-06-01 19:07:09 +01:00
ret += SSL3_RT_HEADER_LENGTH;
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
}
Adding TLS 1.3 Record Layer. In TLS 1.3, the actual record type is hidden within the encrypted data and the record layer defaults to using a TLS 1.0 {3, 1} record version for compatibility. Additionally the record layer no longer checks the minor version of the record layer to maintain compatibility with the TLS 1.3 spec. Change-Id: If2c08e48baab170c1658e0715c33929d36c9be3a Reviewed-on: https://boringssl-review.googlesource.com/8091 Reviewed-by: David Benjamin <davidben@google.com>
2016-06-01 19:07:09 +01:00
/* TLS 1.3 needs an extra byte for the encrypted record type. */
if (ssl->s3->have_version &&
ssl3_protocol_version(ssl) >= TLS1_3_VERSION) {
ret += 1;
}
if (!SSL_IS_DTLS(ssl) && ssl_needs_record_splitting(ssl)) {
ret *= 2;
}
return ret;
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
}
Make tls_open_record always in-place. The business with ssl_record_prefix_len is rather a hassle. Instead, have tls_open_record always decrypt in-place and give back a CBS to where the body is. This way the caller doesn't need to do an extra check all to avoid creating an invalid pointer and underflow in subtraction. Change-Id: I4e12b25a760870d8f8a503673ab00a2d774fc9ee Reviewed-on: https://boringssl-review.googlesource.com/8173 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-06 23:54:51 +01:00
enum ssl_open_record_t tls_open_record(SSL *ssl, uint8_t *out_type, CBS *out,
size_t *out_consumed, uint8_t *out_alert,
uint8_t *in, size_t in_len) {
Push alert handling down into the record functions. Alert handling is more-or-less identical across all contexts. Push it down from read_bytes into the low-level record functions. This also deduplicates the code shared between TLS and DTLS. Now the only type mismatch managed by read_bytes is if we get handshake data in read_app_data. Change-Id: Ia8331897b304566e66d901899cfbf31d2870194e Reviewed-on: https://boringssl-review.googlesource.com/8124 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-02 20:42:01 +01:00
*out_consumed = 0;
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
CBS cbs;
CBS_init(&cbs, in, in_len);
/* Decode the record header. */
uint8_t type;
uint16_t version, ciphertext_len;
if (!CBS_get_u8(&cbs, &type) ||
!CBS_get_u16(&cbs, &version) ||
!CBS_get_u16(&cbs, &ciphertext_len)) {
*out_consumed = SSL3_RT_HEADER_LENGTH;
return ssl_open_record_partial;
}
Adding TLS 1.3 Record Layer. In TLS 1.3, the actual record type is hidden within the encrypted data and the record layer defaults to using a TLS 1.0 {3, 1} record version for compatibility. Additionally the record layer no longer checks the minor version of the record layer to maintain compatibility with the TLS 1.3 spec. Change-Id: If2c08e48baab170c1658e0715c33929d36c9be3a Reviewed-on: https://boringssl-review.googlesource.com/8091 Reviewed-by: David Benjamin <davidben@google.com>
2016-06-01 19:07:09 +01:00
/* Check that the major version in the record matches. As of TLS 1.3, the
* minor version is no longer checked. */
if ((version >> 8) != SSL3_VERSION_MAJOR) {
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
OPENSSL_PUT_ERROR(SSL, SSL_R_WRONG_VERSION_NUMBER);
*out_alert = SSL_AD_PROTOCOL_VERSION;
return ssl_open_record_error;
}
/* Check the ciphertext length. */
Remove SSL_OP_MICROSOFT_BIG_SSLV3_BUFFER. This dates to SSLeay 0.8.0 (or earlier). The use counter sees virtually no hits. Change-Id: Iff4c8899d5cb0ba4afca113c66d15f1d980ffe41 Reviewed-on: https://boringssl-review.googlesource.com/6558 Reviewed-by: Adam Langley <agl@google.com>
2015-11-19 01:41:11 +00:00
if (ciphertext_len > SSL3_RT_MAX_ENCRYPTED_LENGTH) {
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
OPENSSL_PUT_ERROR(SSL, SSL_R_ENCRYPTED_LENGTH_TOO_LONG);
*out_alert = SSL_AD_RECORD_OVERFLOW;
return ssl_open_record_error;
}
/* Extract the body. */
CBS body;
if (!CBS_get_bytes(&cbs, &body, ciphertext_len)) {
*out_consumed = SSL3_RT_HEADER_LENGTH + (size_t)ciphertext_len;
return ssl_open_record_partial;
}
Add helper functions for info_callback and msg_callback. This is getting a little repetitive. Change-Id: Ib0fa8ab10149557c2d728b88648381b9368221d9 Reviewed-on: https://boringssl-review.googlesource.com/8126 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: Adam Langley <agl@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-02 01:16:03 +01:00
ssl_do_msg_callback(ssl, 0 /* read */, 0, SSL3_RT_HEADER, in,
SSL3_RT_HEADER_LENGTH);
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
Make tls_open_record always in-place. The business with ssl_record_prefix_len is rather a hassle. Instead, have tls_open_record always decrypt in-place and give back a CBS to where the body is. This way the caller doesn't need to do an extra check all to avoid creating an invalid pointer and underflow in subtraction. Change-Id: I4e12b25a760870d8f8a503673ab00a2d774fc9ee Reviewed-on: https://boringssl-review.googlesource.com/8173 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-06 23:54:51 +01:00
/* Decrypt the body in-place. */
if (!SSL_AEAD_CTX_open(ssl->s3->aead_read_ctx, out, type, version,
ssl->s3->read_sequence, (uint8_t *)CBS_data(&body),
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
CBS_len(&body))) {
OPENSSL_PUT_ERROR(SSL, SSL_R_DECRYPTION_FAILED_OR_BAD_RECORD_MAC);
*out_alert = SSL_AD_BAD_RECORD_MAC;
return ssl_open_record_error;
}
Push alert handling down into the record functions. Alert handling is more-or-less identical across all contexts. Push it down from read_bytes into the low-level record functions. This also deduplicates the code shared between TLS and DTLS. Now the only type mismatch managed by read_bytes is if we get handshake data in read_app_data. Change-Id: Ia8331897b304566e66d901899cfbf31d2870194e Reviewed-on: https://boringssl-review.googlesource.com/8124 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-02 20:42:01 +01:00
*out_consumed = in_len - CBS_len(&cbs);
Move ssl3_record_sequence_update with the other record-layer bits. Change-Id: I045a4d3e304872b8c97231dcde5bca7753a878fb Reviewed-on: https://boringssl-review.googlesource.com/6831 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:11:39 +00:00
if (!ssl_record_sequence_update(ssl->s3->read_sequence, 8)) {
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
*out_alert = SSL_AD_INTERNAL_ERROR;
return ssl_open_record_error;
}
Adding TLS 1.3 Record Layer. In TLS 1.3, the actual record type is hidden within the encrypted data and the record layer defaults to using a TLS 1.0 {3, 1} record version for compatibility. Additionally the record layer no longer checks the minor version of the record layer to maintain compatibility with the TLS 1.3 spec. Change-Id: If2c08e48baab170c1658e0715c33929d36c9be3a Reviewed-on: https://boringssl-review.googlesource.com/8091 Reviewed-by: David Benjamin <davidben@google.com>
2016-06-01 19:07:09 +01:00
/* TLS 1.3 hides the record type inside the encrypted data. */
if (ssl->s3->have_version &&
ssl3_protocol_version(ssl) >= TLS1_3_VERSION &&
ssl->s3->aead_read_ctx != NULL) {
Make tls_open_record always in-place. The business with ssl_record_prefix_len is rather a hassle. Instead, have tls_open_record always decrypt in-place and give back a CBS to where the body is. This way the caller doesn't need to do an extra check all to avoid creating an invalid pointer and underflow in subtraction. Change-Id: I4e12b25a760870d8f8a503673ab00a2d774fc9ee Reviewed-on: https://boringssl-review.googlesource.com/8173 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-06 23:54:51 +01:00
do {
if (!CBS_get_last_u8(out, &type)) {
OPENSSL_PUT_ERROR(SSL, SSL_R_DECRYPTION_FAILED_OR_BAD_RECORD_MAC);
*out_alert = SSL_AD_DECRYPT_ERROR;
return ssl_open_record_error;
}
} while (type == 0);
Adding TLS 1.3 Record Layer. In TLS 1.3, the actual record type is hidden within the encrypted data and the record layer defaults to using a TLS 1.0 {3, 1} record version for compatibility. Additionally the record layer no longer checks the minor version of the record layer to maintain compatibility with the TLS 1.3 spec. Change-Id: If2c08e48baab170c1658e0715c33929d36c9be3a Reviewed-on: https://boringssl-review.googlesource.com/8091 Reviewed-by: David Benjamin <davidben@google.com>
2016-06-01 19:07:09 +01:00
}
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
/* Check the plaintext length. */
Make tls_open_record always in-place. The business with ssl_record_prefix_len is rather a hassle. Instead, have tls_open_record always decrypt in-place and give back a CBS to where the body is. This way the caller doesn't need to do an extra check all to avoid creating an invalid pointer and underflow in subtraction. Change-Id: I4e12b25a760870d8f8a503673ab00a2d774fc9ee Reviewed-on: https://boringssl-review.googlesource.com/8173 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-06 23:54:51 +01:00
if (CBS_len(out) > SSL3_RT_MAX_PLAIN_LENGTH) {
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
OPENSSL_PUT_ERROR(SSL, SSL_R_DATA_LENGTH_TOO_LONG);
*out_alert = SSL_AD_RECORD_OVERFLOW;
return ssl_open_record_error;
}
Reject empty records of unexpected type. The old empty record logic discarded the records at a very low-level. Let the error bubble up to ssl3_read_bytes so the type mismatch logic may kick in before the empty record is skipped. Add tests for when the record in question is application data, before before the handshake and post ChangeCipherSpec. BUG=521840 Change-Id: I47dff389cda65d6672b9be39d7d89490331063fa Reviewed-on: https://boringssl-review.googlesource.com/5754 Reviewed-by: Adam Langley <agl@google.com>
2015-08-22 06:35:43 +01:00
/* Limit the number of consecutive empty records. */
Make tls_open_record always in-place. The business with ssl_record_prefix_len is rather a hassle. Instead, have tls_open_record always decrypt in-place and give back a CBS to where the body is. This way the caller doesn't need to do an extra check all to avoid creating an invalid pointer and underflow in subtraction. Change-Id: I4e12b25a760870d8f8a503673ab00a2d774fc9ee Reviewed-on: https://boringssl-review.googlesource.com/8173 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-06 23:54:51 +01:00
if (CBS_len(out) == 0) {
Reject empty records of unexpected type. The old empty record logic discarded the records at a very low-level. Let the error bubble up to ssl3_read_bytes so the type mismatch logic may kick in before the empty record is skipped. Add tests for when the record in question is application data, before before the handshake and post ChangeCipherSpec. BUG=521840 Change-Id: I47dff389cda65d6672b9be39d7d89490331063fa Reviewed-on: https://boringssl-review.googlesource.com/5754 Reviewed-by: Adam Langley <agl@google.com>
2015-08-22 06:35:43 +01:00
ssl->s3->empty_record_count++;
if (ssl->s3->empty_record_count > kMaxEmptyRecords) {
OPENSSL_PUT_ERROR(SSL, SSL_R_TOO_MANY_EMPTY_FRAGMENTS);
*out_alert = SSL_AD_UNEXPECTED_MESSAGE;
return ssl_open_record_error;
}
/* Apart from the limit, empty records are returned up to the caller. This
* allows the caller to reject records of the wrong type. */
} else {
ssl->s3->empty_record_count = 0;
}
Push alert handling down into the record functions. Alert handling is more-or-less identical across all contexts. Push it down from read_bytes into the low-level record functions. This also deduplicates the code shared between TLS and DTLS. Now the only type mismatch managed by read_bytes is if we get handshake data in read_app_data. Change-Id: Ia8331897b304566e66d901899cfbf31d2870194e Reviewed-on: https://boringssl-review.googlesource.com/8124 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-02 20:42:01 +01:00
if (type == SSL3_RT_ALERT) {
Make tls_open_record always in-place. The business with ssl_record_prefix_len is rather a hassle. Instead, have tls_open_record always decrypt in-place and give back a CBS to where the body is. This way the caller doesn't need to do an extra check all to avoid creating an invalid pointer and underflow in subtraction. Change-Id: I4e12b25a760870d8f8a503673ab00a2d774fc9ee Reviewed-on: https://boringssl-review.googlesource.com/8173 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-06 23:54:51 +01:00
return ssl_process_alert(ssl, out_alert, CBS_data(out), CBS_len(out));
Push alert handling down into the record functions. Alert handling is more-or-less identical across all contexts. Push it down from read_bytes into the low-level record functions. This also deduplicates the code shared between TLS and DTLS. Now the only type mismatch managed by read_bytes is if we get handshake data in read_app_data. Change-Id: Ia8331897b304566e66d901899cfbf31d2870194e Reviewed-on: https://boringssl-review.googlesource.com/8124 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-02 20:42:01 +01:00
}
ssl->s3->warning_alert_count = 0;
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
*out_type = type;
return ssl_open_record_success;
}
static int do_seal_record(SSL *ssl, uint8_t *out, size_t *out_len,
size_t max_out, uint8_t type, const uint8_t *in,
size_t in_len) {
if (max_out < SSL3_RT_HEADER_LENGTH) {
OPENSSL_PUT_ERROR(SSL, SSL_R_BUFFER_TOO_SMALL);
return 0;
}
Remove in-place TLS record assembly for now. Decrypting is very easy to do in-place, but encrypting in-place is a hassle. The rules actually were wrong due to record-splitting. The aliasing prefix and the alignment prefix actually differ by 1. Take it out for now in preparation for tightening the aliasing rules. If we decide to do in-place encrypt later, probably it'd be more useful to return header + in-place ciphertext + trailer. (That, in turn, needs a scatter/gather thing on the AEAD thanks to TLS 1.3's padding and record type construction.) We may also wish to rethink how record-splitting works here. Change-Id: I0187d39c541e76ef933b7c2c193323164fd8a156 Reviewed-on: https://boringssl-review.googlesource.com/8230 Reviewed-by: Adam Langley <agl@google.com>
2016-06-08 23:31:24 +01:00
/* Either |in| and |out| don't alias or |in| aligns with the
* ciphertext. |tls_seal_record| forbids aliasing, but TLS 1.3 aliases them
* internally. */
assert(!buffers_alias(in, in_len, out, max_out) ||
in ==
out + SSL3_RT_HEADER_LENGTH +
SSL_AEAD_CTX_explicit_nonce_len(ssl->s3->aead_write_ctx));
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
out[0] = type;
Adding TLS 1.3 Record Layer. In TLS 1.3, the actual record type is hidden within the encrypted data and the record layer defaults to using a TLS 1.0 {3, 1} record version for compatibility. Additionally the record layer no longer checks the minor version of the record layer to maintain compatibility with the TLS 1.3 spec. Change-Id: If2c08e48baab170c1658e0715c33929d36c9be3a Reviewed-on: https://boringssl-review.googlesource.com/8091 Reviewed-by: David Benjamin <davidben@google.com>
2016-06-01 19:07:09 +01:00
/* The TLS record-layer version number is meaningless and, starting in
* TLS 1.3, is frozen at TLS 1.0. But for historical reasons, SSL 3.0
* ClientHellos should use SSL 3.0 and pre-TLS-1.3 expects the version
* to change after version negotiation. */
uint16_t wire_version = TLS1_VERSION;
if (ssl->version == SSL3_VERSION ||
(ssl->s3->have_version && ssl3_protocol_version(ssl) < TLS1_3_VERSION)) {
wire_version = ssl->version;
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
}
out[1] = wire_version >> 8;
out[2] = wire_version & 0xff;
size_t ciphertext_len;
Move aead_{read,write}_ctx and next_proto_negotiated into ssl->s3. Both are connection state rather than configuration state. Notably this cuts down more of SSL_clear that can't just use ssl_free + ssl_new. Change-Id: I3c05b3ae86d4db8bd75f1cd21656f57fc5b55ca9 Reviewed-on: https://boringssl-review.googlesource.com/6835 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:56:49 +00:00
if (!SSL_AEAD_CTX_seal(ssl->s3->aead_write_ctx, out + SSL3_RT_HEADER_LENGTH,
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
&ciphertext_len, max_out - SSL3_RT_HEADER_LENGTH,
type, wire_version, ssl->s3->write_sequence, in,
in_len) ||
Move ssl3_record_sequence_update with the other record-layer bits. Change-Id: I045a4d3e304872b8c97231dcde5bca7753a878fb Reviewed-on: https://boringssl-review.googlesource.com/6831 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:11:39 +00:00
!ssl_record_sequence_update(ssl->s3->write_sequence, 8)) {
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
return 0;
}
if (ciphertext_len >= 1 << 16) {
OPENSSL_PUT_ERROR(SSL, ERR_R_OVERFLOW);
return 0;
}
out[3] = ciphertext_len >> 8;
out[4] = ciphertext_len & 0xff;
*out_len = SSL3_RT_HEADER_LENGTH + ciphertext_len;
Add helper functions for info_callback and msg_callback. This is getting a little repetitive. Change-Id: Ib0fa8ab10149557c2d728b88648381b9368221d9 Reviewed-on: https://boringssl-review.googlesource.com/8126 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: Adam Langley <agl@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-02 01:16:03 +01:00
ssl_do_msg_callback(ssl, 1 /* write */, 0, SSL3_RT_HEADER, out,
SSL3_RT_HEADER_LENGTH);
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
return 1;
}
int tls_seal_record(SSL *ssl, uint8_t *out, size_t *out_len, size_t max_out,
uint8_t type, const uint8_t *in, size_t in_len) {
Remove in-place TLS record assembly for now. Decrypting is very easy to do in-place, but encrypting in-place is a hassle. The rules actually were wrong due to record-splitting. The aliasing prefix and the alignment prefix actually differ by 1. Take it out for now in preparation for tightening the aliasing rules. If we decide to do in-place encrypt later, probably it'd be more useful to return header + in-place ciphertext + trailer. (That, in turn, needs a scatter/gather thing on the AEAD thanks to TLS 1.3's padding and record type construction.) We may also wish to rethink how record-splitting works here. Change-Id: I0187d39c541e76ef933b7c2c193323164fd8a156 Reviewed-on: https://boringssl-review.googlesource.com/8230 Reviewed-by: Adam Langley <agl@google.com>
2016-06-08 23:31:24 +01:00
if (buffers_alias(in, in_len, out, max_out)) {
OPENSSL_PUT_ERROR(SSL, SSL_R_OUTPUT_ALIASES_INPUT);
return 0;
}
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
size_t frag_len = 0;
Adding TLS 1.3 Record Layer. In TLS 1.3, the actual record type is hidden within the encrypted data and the record layer defaults to using a TLS 1.0 {3, 1} record version for compatibility. Additionally the record layer no longer checks the minor version of the record layer to maintain compatibility with the TLS 1.3 spec. Change-Id: If2c08e48baab170c1658e0715c33929d36c9be3a Reviewed-on: https://boringssl-review.googlesource.com/8091 Reviewed-by: David Benjamin <davidben@google.com>
2016-06-01 19:07:09 +01:00
/* TLS 1.3 hides the actual record type inside the encrypted data. */
if (ssl->s3->have_version &&
ssl3_protocol_version(ssl) >= TLS1_3_VERSION &&
Fix TLS 1.3 seal logic. Check against the write encryption state, not the read state. Change-Id: Ib3d8e02800e37bd089ef02c67a0b7e5dc009b1a5 Reviewed-on: https://boringssl-review.googlesource.com/8330 Reviewed-by: Nick Harper <nharper@chromium.org> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-16 22:03:03 +01:00
ssl->s3->aead_write_ctx != NULL) {
Adding TLS 1.3 Record Layer. In TLS 1.3, the actual record type is hidden within the encrypted data and the record layer defaults to using a TLS 1.0 {3, 1} record version for compatibility. Additionally the record layer no longer checks the minor version of the record layer to maintain compatibility with the TLS 1.3 spec. Change-Id: If2c08e48baab170c1658e0715c33929d36c9be3a Reviewed-on: https://boringssl-review.googlesource.com/8091 Reviewed-by: David Benjamin <davidben@google.com>
2016-06-01 19:07:09 +01:00
size_t padding = SSL3_RT_HEADER_LENGTH + 1;
if (in_len > in_len + padding || max_out < in_len + padding) {
OPENSSL_PUT_ERROR(SSL, SSL_R_BUFFER_TOO_SMALL);
return 0;
}
memmove(out + SSL3_RT_HEADER_LENGTH, in, in_len);
out[SSL3_RT_HEADER_LENGTH + in_len] = type;
in = out + SSL3_RT_HEADER_LENGTH;
type = SSL3_RT_APPLICATION_DATA;
in_len++;
}
Remove |need_record_splitting| from |SSL3_STATE|. It is redundant given the other state in the connection. Change-Id: I5dc71627132659ab4316a5ea360c9ca480fb7c6c Reviewed-on: https://boringssl-review.googlesource.com/6646 Reviewed-by: Adam Langley <agl@google.com>
2015-12-06 21:07:47 +00:00
if (type == SSL3_RT_APPLICATION_DATA && in_len > 1 &&
ssl_needs_record_splitting(ssl)) {
Remove in-place TLS record assembly for now. Decrypting is very easy to do in-place, but encrypting in-place is a hassle. The rules actually were wrong due to record-splitting. The aliasing prefix and the alignment prefix actually differ by 1. Take it out for now in preparation for tightening the aliasing rules. If we decide to do in-place encrypt later, probably it'd be more useful to return header + in-place ciphertext + trailer. (That, in turn, needs a scatter/gather thing on the AEAD thanks to TLS 1.3's padding and record type construction.) We may also wish to rethink how record-splitting works here. Change-Id: I0187d39c541e76ef933b7c2c193323164fd8a156 Reviewed-on: https://boringssl-review.googlesource.com/8230 Reviewed-by: Adam Langley <agl@google.com>
2016-06-08 23:31:24 +01:00
if (!do_seal_record(ssl, out, &frag_len, max_out, type, in, 1)) {
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
return 0;
}
in++;
in_len--;
out += frag_len;
max_out -= frag_len;
Disable all TLS crypto in fuzzer mode. Both sides' signature and Finished checks still occur, but the results are ignored. Also, all ciphers behave like the NULL cipher. Conveniently, this isn't that much code since all ciphers and their size computations funnel into SSL_AEAD_CTX. This does carry some risk that we'll mess up this code. Up until now, we've tried to avoid test-only changes to the SSL stack. There is little risk that anyone will ship a BORINGSSL_UNSAFE_FUZZER_MODE build for anything since it doesn't interop anyway. There is some risk that we'll end up messing up the disableable checks. However, both skipped checks have negative tests in runner (see tests that set InvalidSKXSignature and BadFinished). For good measure, I've added a server variant of the existing BadFinished test to this CL, although they hit the same code. Change-Id: I37f6b4d62b43bc08fab7411965589b423d86f4b8 Reviewed-on: https://boringssl-review.googlesource.com/7287 Reviewed-by: Adam Langley <agl@google.com>
2016-03-02 03:57:40 +00:00
#if !defined(BORINGSSL_UNSAFE_FUZZER_MODE)
Move aead_{read,write}_ctx and next_proto_negotiated into ssl->s3. Both are connection state rather than configuration state. Notably this cuts down more of SSL_clear that can't just use ssl_free + ssl_new. Change-Id: I3c05b3ae86d4db8bd75f1cd21656f57fc5b55ca9 Reviewed-on: https://boringssl-review.googlesource.com/6835 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:56:49 +00:00
assert(SSL3_RT_HEADER_LENGTH + ssl_cipher_get_record_split_len(
ssl->s3->aead_write_ctx->cipher) ==
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
frag_len);
Disable all TLS crypto in fuzzer mode. Both sides' signature and Finished checks still occur, but the results are ignored. Also, all ciphers behave like the NULL cipher. Conveniently, this isn't that much code since all ciphers and their size computations funnel into SSL_AEAD_CTX. This does carry some risk that we'll mess up this code. Up until now, we've tried to avoid test-only changes to the SSL stack. There is little risk that anyone will ship a BORINGSSL_UNSAFE_FUZZER_MODE build for anything since it doesn't interop anyway. There is some risk that we'll end up messing up the disableable checks. However, both skipped checks have negative tests in runner (see tests that set InvalidSKXSignature and BadFinished). For good measure, I've added a server variant of the existing BadFinished test to this CL, although they hit the same code. Change-Id: I37f6b4d62b43bc08fab7411965589b423d86f4b8 Reviewed-on: https://boringssl-review.googlesource.com/7287 Reviewed-by: Adam Langley <agl@google.com>
2016-03-02 03:57:40 +00:00
#endif
Factor out the buffering and low-level record code. This begins decoupling the transport from the SSL state machine. The buffering logic is hidden behind an opaque API. Fields like ssl->packet and ssl->packet_length are gone. ssl3_get_record and dtls1_get_record now call low-level tls_open_record and dtls_open_record functions that unpack a single record independent of who owns the buffer. Both may be called in-place. This removes ssl->rstate which was redundant with the buffer length. Future work will push the buffer up the stack until it is above the handshake. Then we can expose SSL_open and SSL_seal APIs which act like *_open_record but return a slightly larger enum due to other events being possible. Likewise the handshake state machine will be detached from its buffer. The existing SSL_read, SSL_write, etc., APIs will be implemented on top of SSL_open, etc., combined with ssl_read_buffer_* and ssl_write_buffer_*. (Which is why ssl_read_buffer_extend still tries to abstract between TLS's and DTLS's fairly different needs.) The new buffering logic does not support read-ahead (removed previously) since it lacks a memmove on ssl_read_buffer_discard for TLS, but this could be added if desired. The old buffering logic wasn't quite right anyway; it tried to avoid the memmove in some cases and could get stuck too far into the buffer and not accept records. (The only time the memmove is optional is in DTLS or if enough of the record header is available to know that the entire next record would fit in the buffer.) The new logic also now actually decrypts the ciphertext in-place again, rather than almost in-place when there's an explicit nonce/IV. (That accidentally switched in https://boringssl-review.googlesource.com/#/c/4792/; see 3d59e04bce96474099ba76786a2337e99ae14505.) BUG=468889 Change-Id: I403c1626253c46897f47c7ae93aeab1064b767b2 Reviewed-on: https://boringssl-review.googlesource.com/5715 Reviewed-by: Adam Langley <agl@google.com>
2015-07-29 02:34:45 +01:00
}
if (!do_seal_record(ssl, out, out_len, max_out, type, in, in_len)) {
return 0;
}
*out_len += frag_len;
return 1;
}
Tidy up keyblock and CCS logic slightly. Move the actual SSL_AEAD_CTX swap into the record layer. Also revise the intermediate state we store between setup_key_block and change_cipher_state. With SSL_AEAD_CTX_new abstracted out, keeping the EVP_AEAD around doesn't make much sense. Just store enough to partition the key block. Change-Id: I773fb46a2cb78fa570f00c0a89339c15bbb1d719 Reviewed-on: https://boringssl-review.googlesource.com/6832 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:34:23 +00:00
void ssl_set_read_state(SSL *ssl, SSL_AEAD_CTX *aead_ctx) {
if (SSL_IS_DTLS(ssl)) {
ssl->d1->r_epoch++;
memset(&ssl->d1->bitmap, 0, sizeof(ssl->d1->bitmap));
}
memset(ssl->s3->read_sequence, 0, sizeof(ssl->s3->read_sequence));
Move aead_{read,write}_ctx and next_proto_negotiated into ssl->s3. Both are connection state rather than configuration state. Notably this cuts down more of SSL_clear that can't just use ssl_free + ssl_new. Change-Id: I3c05b3ae86d4db8bd75f1cd21656f57fc5b55ca9 Reviewed-on: https://boringssl-review.googlesource.com/6835 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:56:49 +00:00
SSL_AEAD_CTX_free(ssl->s3->aead_read_ctx);
ssl->s3->aead_read_ctx = aead_ctx;
Tidy up keyblock and CCS logic slightly. Move the actual SSL_AEAD_CTX swap into the record layer. Also revise the intermediate state we store between setup_key_block and change_cipher_state. With SSL_AEAD_CTX_new abstracted out, keeping the EVP_AEAD around doesn't make much sense. Just store enough to partition the key block. Change-Id: I773fb46a2cb78fa570f00c0a89339c15bbb1d719 Reviewed-on: https://boringssl-review.googlesource.com/6832 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:34:23 +00:00
}
void ssl_set_write_state(SSL *ssl, SSL_AEAD_CTX *aead_ctx) {
if (SSL_IS_DTLS(ssl)) {
ssl->d1->w_epoch++;
memcpy(ssl->d1->last_write_sequence, ssl->s3->write_sequence,
sizeof(ssl->s3->write_sequence));
}
memset(ssl->s3->write_sequence, 0, sizeof(ssl->s3->write_sequence));
Move aead_{read,write}_ctx and next_proto_negotiated into ssl->s3. Both are connection state rather than configuration state. Notably this cuts down more of SSL_clear that can't just use ssl_free + ssl_new. Change-Id: I3c05b3ae86d4db8bd75f1cd21656f57fc5b55ca9 Reviewed-on: https://boringssl-review.googlesource.com/6835 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:56:49 +00:00
SSL_AEAD_CTX_free(ssl->s3->aead_write_ctx);
ssl->s3->aead_write_ctx = aead_ctx;
Tidy up keyblock and CCS logic slightly. Move the actual SSL_AEAD_CTX swap into the record layer. Also revise the intermediate state we store between setup_key_block and change_cipher_state. With SSL_AEAD_CTX_new abstracted out, keeping the EVP_AEAD around doesn't make much sense. Just store enough to partition the key block. Change-Id: I773fb46a2cb78fa570f00c0a89339c15bbb1d719 Reviewed-on: https://boringssl-review.googlesource.com/6832 Reviewed-by: Adam Langley <alangley@gmail.com>
2015-12-25 20:34:23 +00:00
}
Push alert handling down into the record functions. Alert handling is more-or-less identical across all contexts. Push it down from read_bytes into the low-level record functions. This also deduplicates the code shared between TLS and DTLS. Now the only type mismatch managed by read_bytes is if we get handshake data in read_app_data. Change-Id: Ia8331897b304566e66d901899cfbf31d2870194e Reviewed-on: https://boringssl-review.googlesource.com/8124 Reviewed-by: Steven Valdez <svaldez@google.com> Reviewed-by: David Benjamin <davidben@google.com>
2016-06-02 20:42:01 +01:00
enum ssl_open_record_t ssl_process_alert(SSL *ssl, uint8_t *out_alert,
const uint8_t *in, size_t in_len) {
/* Alerts records may not contain fragmented or multiple alerts. */
if (in_len != 2) {
*out_alert = SSL_AD_DECODE_ERROR;
OPENSSL_PUT_ERROR(SSL, SSL_R_BAD_ALERT);
return ssl_open_record_error;
}
ssl_do_msg_callback(ssl, 0 /* read */, ssl->version, SSL3_RT_ALERT, in, in_len);
const uint8_t alert_level = in[0];
const uint8_t alert_descr = in[1];
uint16_t alert = (alert_level << 8) | alert_descr;
ssl_do_info_callback(ssl, SSL_CB_READ_ALERT, alert);
if (alert_level == SSL3_AL_WARNING) {
if (alert_descr == SSL_AD_CLOSE_NOTIFY) {
ssl->s3->recv_shutdown = ssl_shutdown_close_notify;
return ssl_open_record_close_notify;
}
ssl->s3->warning_alert_count++;
if (ssl->s3->warning_alert_count > kMaxWarningAlerts) {
*out_alert = SSL_AD_UNEXPECTED_MESSAGE;
OPENSSL_PUT_ERROR(SSL, SSL_R_TOO_MANY_WARNING_ALERTS);
return ssl_open_record_error;
}
return ssl_open_record_discard;
}
if (alert_level == SSL3_AL_FATAL) {
ssl->s3->recv_shutdown = ssl_shutdown_fatal_alert;
SSL_CTX_remove_session(ssl->ctx, ssl->session);
char tmp[16];
OPENSSL_PUT_ERROR(SSL, SSL_AD_REASON_OFFSET + alert_descr);
BIO_snprintf(tmp, sizeof(tmp), "%d", alert_descr);
ERR_add_error_data(2, "SSL alert number ", tmp);
return ssl_open_record_fatal_alert;
}
*out_alert = SSL_AD_ILLEGAL_PARAMETER;
OPENSSL_PUT_ERROR(SSL, SSL_R_UNKNOWN_ALERT_TYPE);
return ssl_open_record_error;
}
Reference in New Issue Copy Permalink