5 years agoFix for SRTP Memory Leak
Matt Caswell [Wed, 15 Oct 2014 00:20:38 +0000 (01:20 +0100)]
Fix for SRTP Memory Leak


This issue was reported to OpenSSL on 26th September 2014, based on an origi
issue and patch developed by the LibreSSL project. Further analysis of the i
was performed by the OpenSSL team.

The fix was developed by the OpenSSL team.

Reviewed-by: Tim Hudson <>
5 years agoFix SSL_R naming inconsistency.
Bodo Moeller [Wed, 15 Oct 2014 12:47:39 +0000 (14:47 +0200)]
Fix SSL_R naming inconsistency.

Reviewed-by: Tim Hudson <>
5 years make ECB subroutine Windows ABI compliant.
Andy Polyakov [Wed, 15 Oct 2014 09:10:08 +0000 (11:10 +0200)] make ECB subroutine Windows ABI compliant.

RT: 3553
Reviewed-by: Emilia Kasper <>
(cherry picked from commit 69d5747f90136aa026a96204f26ab39549dfc69b)

5 years agoAdd constant_time_locl.h to HEADERS,
Tim Hudson [Thu, 25 Sep 2014 06:04:35 +0000 (08:04 +0200)]
Add constant_time_locl.h to HEADERS,
so the Win32 compile picks it up correctly.

Reviewed-by: Richard Levitte <>
5 years agoAdd the constant time test to the VMS build and tests
Richard Levitte [Thu, 25 Sep 2014 04:47:23 +0000 (06:47 +0200)]
Add the constant time test to the VMS build and tests

Reviewed-by: Tim Hudson <>

5 years agoInclude "constant_time_locl.h" rather than "../constant_time_locl.h".
Richard Levitte [Wed, 24 Sep 2014 20:59:37 +0000 (22:59 +0200)]
Include "constant_time_locl.h" rather than "../constant_time_locl.h".
The different -I compiler parameters will take care of the rest...

Reviewed-by: Tim Hudson <>
5 years agoSpaces were added in some strings for better readability. However, those spaces do...
Richard Levitte [Mon, 16 Jun 2014 11:24:04 +0000 (13:24 +0200)]
Spaces were added in some strings for better readability. However, those spaces do not belong in file names, so when picking out the individual parts, remove the spaces

Reviewed-by: Tim Hudson <>
5 years agoAdjust VMS build to Unix build. Most of all, make it so the disabled
Richard Levitte [Wed, 6 Aug 2014 02:37:27 +0000 (04:37 +0200)]
Adjust VMS build to Unix build.  Most of all, make it so the disabled
algorithms MD2 and RC5 don't get built.
Also, disable building the test apps in crypto/des and crypto/pkcs7, as
they have no support at all.

Reviewed-by: Tim Hudson <>
5 years agoMake sure test/ exit gracefully, even when openssl.exe wasn't properly built.
Richard Levitte [Wed, 18 Jun 2014 11:41:54 +0000 (13:41 +0200)]
Make sure test/ exit gracefully, even when openssl.exe wasn't properly built.

Reviewed-by: Tim Hudson <>
5 years agoUpdate the VMS build according to the latest unixly build. Partly provided by Zoltan...
Richard Levitte [Wed, 6 Aug 2014 01:54:57 +0000 (03:54 +0200)]
Update the VMS build according to the latest unixly build. Partly provided by Zoltan Arpadffy <>

Reviewed-by: Tim Hudson <>
5 years agoMake sure that disabling the MAYLOSEDATA3 warning is only done when the compiler...
Richard Levitte [Sat, 14 Jun 2014 13:13:44 +0000 (15:13 +0200)]
Make sure that disabling the MAYLOSEDATA3 warning is only done when the compiler supports it. Otherwise, there are warnings about it lacking everywhere, which is quite tedious to read through while trying to check for other warnings.

Reviewed-by: Tim Hudson <>
5 years agoSupport TLS_FALLBACK_SCSV.
Bodo Moeller [Wed, 15 Oct 2014 02:05:42 +0000 (04:05 +0200)]

Reviewed-by: Rich Salz <>
5 years agoPreserve digests for SNI.
Dr. Stephen Henson [Fri, 10 Oct 2014 12:18:09 +0000 (13:18 +0100)]
Preserve digests for SNI.

SSL_set_SSL_CTX is normally called for SNI after ClientHello has
received and the digest to use for each certificate has been decided.
The original ssl->cert contains the negotiated digests and is now
copied to the new ssl->cert.

PR: 3560
Reviewed-by: Tim Hudson <>
5 years agoRemoved duplicate definition of PKCS7_type_is_encrypted
Matt Caswell [Fri, 3 Oct 2014 22:48:49 +0000 (23:48 +0100)]
Removed duplicate definition of PKCS7_type_is_encrypted

Patch supplied by Matthieu Patou <>, and modified to also
remove duplicate definition of PKCS7_type_is_digest.


Reviewed-by: Rich Salz <>
(cherry picked from commit e0fdea3e49e7454aa76bd5ecf3a3747641354c68)

5 years agoAdd additional DigestInfo checks.
Dr. Stephen Henson [Thu, 25 Sep 2014 22:28:48 +0000 (23:28 +0100)]
Add additional DigestInfo checks.

Reencode DigestInto in DER and check against the original: this
will reject any improperly encoded DigestInfo structures.

Note: this is a precautionary measure, there is no known attack
which can exploit this.

Thanks to Brian Smith for reporting this issue.
Reviewed-by: Tim Hudson <>
5 years agoAdd missing tests
Emilia Kasper [Thu, 25 Sep 2014 11:39:21 +0000 (13:39 +0200)]
Add missing tests

Accidentally omitted from commit 455b65dfab0de51c9f67b3c909311770f2b3f801

Reviewed-by: Kurt Roeckx <>
(cherry picked from commit fdc35a9d3e8cf4cfd9330d5df9883f42cf5648ad)

5 years agoUse correct function name: CMS_add1_signer()
Dr. Stephen Henson [Sat, 20 Sep 2014 00:00:55 +0000 (01:00 +0100)]
Use correct function name: CMS_add1_signer()

Reviewed-by: Matt Caswell <>
(cherry picked from commit 5886354dcca4f8445ed35b6995a035b75409590c)

5 years agocrypto/bn/bn_nist.c: work around MSC ARM compiler bug.
Andy Polyakov [Wed, 24 Sep 2014 22:42:26 +0000 (00:42 +0200)]
crypto/bn/bn_nist.c: work around MSC ARM compiler bug.

RT: 3541
Reviewed-by: Emilia Kasper <>
(cherry picked from commit 8b07c005fe006044d0e4a795421447deca3c9f2c)

5 years agoRT3425: constant-time evp_enc
Emilia Kasper [Fri, 5 Sep 2014 12:47:33 +0000 (14:47 +0200)]
RT3425: constant-time evp_enc

Do the final padding check in EVP_DecryptFinal_ex in constant time to
avoid a timing leak from padding failure.

Reviewed-by: Rich Salz <>
(cherry picked from commit 4aac102f75b517bdb56b1bcfd0a856052d559f6e)


(cherry picked from commit 738911cde68b2b3706e502cf8daf5b14738f2f42)

5 years agoRT3067: simplify patch
Emilia Kasper [Thu, 4 Sep 2014 11:04:42 +0000 (13:04 +0200)]
RT3067: simplify patch

(Original commit adb46dbc6dd7347750df2468c93e8c34bcb93a4b)

Use the new constant-time methods consistently in s3_srvr.c

Reviewed-by: Kurt Roeckx <>
(cherry picked from commit 455b65dfab0de51c9f67b3c909311770f2b3f801)

5 years agoThis change alters the processing of invalid, RSA pre-master secrets so
Adam Langley [Wed, 24 Apr 2013 18:45:44 +0000 (14:45 -0400)]
This change alters the processing of invalid, RSA pre-master secrets so
that bad encryptions are treated like random session keys in constant

(cherry picked from commit adb46dbc6dd7347750df2468c93e8c34bcb93a4b)

Reviewed-by: Rich Salz <>
5 years agoRT3066: rewrite RSA padding checks to be slightly more constant time.
Emilia Kasper [Thu, 28 Aug 2014 17:43:49 +0000 (19:43 +0200)]
RT3066: rewrite RSA padding checks to be slightly more constant time.

Also tweak s3_cbc.c to use new constant-time methods.
Also fix memory leaks from internal errors in RSA_padding_check_PKCS1_OAEP_mgf1

This patch is based on the original RT submission by Adam Langley <>,
as well as code from BoringSSL and OpenSSL.

Reviewed-by: Kurt Roeckx <>

5 years agoFixed error introduced in commit f2be92b94dad3c6cbdf79d99a324804094cf1617
Tim Hudson [Sun, 21 Sep 2014 11:54:31 +0000 (21:54 +1000)]
Fixed error introduced in commit f2be92b94dad3c6cbdf79d99a324804094cf1617
that fixed PR#3450 where an existing cast masked an issue when i was changed
from int to long in that commit

Picked up on z/linux (s390) where sizeof(int)!=sizeof(long)

Reviewed-by: Rich Salz <>
Reviewed-by: Matt Caswell <>
(cherry picked from commit b5ff559ff90124c6fd53bbb49dae5edb4e821e0a)

5 years agoRT2560: missing NULL check in ocsp_req_find_signer
Rich Salz [Wed, 10 Sep 2014 15:43:45 +0000 (11:43 -0400)]
RT2560: missing NULL check in ocsp_req_find_signer

If we don't find a signer in the internal list, then fall
through and look at the internal list; don't just return NULL.

Reviewed-by: Dr. Stephen Henson <>
(cherry picked from commit b2aa38a980e9fbf158aafe487fb729c492b241fb)

5 years agoRT3301: Discard too-long heartbeat requests
Erik Auerswald [Wed, 27 Aug 2014 02:50:34 +0000 (22:50 -0400)]
RT3301: Discard too-long heartbeat requests

Reviewed-by: Tim Hudson <>
(cherry picked from commit af4c6e348e4bad6303e7d214cdcf2536487aabe4)

5 years agopsk_client_callback, 128-byte id bug.
Adam Langley [Fri, 20 Jun 2014 19:00:00 +0000 (12:00 -0700)]
psk_client_callback, 128-byte id bug.

Fix a bug in handling of 128 byte long PSK identity in

OpenSSL supports PSK identities of up to (and including) 128 bytes in
length. PSK identity is obtained via the psk_client_callback,
implementors of which are expected to provide a NULL-terminated
identity. However, the callback is invoked with only 128 bytes of
storage thus making it impossible to return a 128 byte long identity and
the required additional NULL byte.

This CL fixes the issue by passing in a 129 byte long buffer into the
psk_client_callback. As a safety precaution, this CL also zeroes out the
buffer before passing it into the callback, uses strnlen for obtaining
the length of the identity returned by the callback, and aborts the
handshake if the identity (without the NULL terminator) is longer than
128 bytes.

(Original patch amended to achieve strnlen in a different way.)

Reviewed-by: Rich Salz <>
(cherry picked from commit be0d851732bad7370640702bc9c4a33189ede287)

5 years agoEnsure that x**0 mod 1 = 0.
Adam Langley [Tue, 23 Apr 2013 16:13:51 +0000 (12:13 -0400)]
Ensure that x**0 mod 1 = 0.

(cherry picked from commit 2b0180c37fa6ffc48ee40caa831ca398b828e680)

Reviewed-by: Ben Laurie <>
5 years agoFollowup on RT3334 fix: make sure that a directory that's the empty
Richard Levitte [Thu, 14 Aug 2014 23:24:34 +0000 (01:24 +0200)]
Followup on RT3334 fix: make sure that a directory that's the empty
string returns 0 with errno = ENOENT.

Reviewed-by: Andy Polyakov <>
(cherry picked from commit 360928b7d0f16dde70e26841bbf9e1af727e8b8f)

5 years agoRT3334: Fix crypto/LPdir_win.c
Phil Mesnier [Thu, 14 Aug 2014 17:35:07 +0000 (19:35 +0200)]
RT3334: Fix crypto/LPdir_win.c

Reviewed-by: Richard Levitte <>
Reviewed-by: Andy Polyakov <>
(cherry picked from commit 6a14fe7576e7a14a46ba14df8be8fe478536b4fb)

5 years agoMake the inline const-time functions static.
Emilia Kasper [Thu, 28 Aug 2014 17:45:55 +0000 (19:45 +0200)]
Make the inline const-time functions static.

"inline" without static is not correct as the compiler may choose to ignore it
and will then either emit an external definition, or expect one.

Reviewed-by: Geoff Thorpe <>
(cherry picked from commit 86f50b36e63275a916b147f9d8764e3c0c060fdb)

5 years work around warning.
Andy Polyakov [Sat, 30 Aug 2014 17:17:09 +0000 (19:17 +0200)] work around warning.

Reviewed-by: Rich Salz <>
(cherry picked from commit 4d86e8df6be69ed13abb73fd564f1f894eea0a98)

5 years agoAdd tags/TAGS
Rich Salz [Mon, 14 Jul 2014 15:27:16 +0000 (11:27 -0400)]
Add tags/TAGS

Reviewed-by: Tim Hudson <>
(cherry picked from commit 9d6253cfd3395dfe7147dae134579c6b16544c53)

5 years agoRT2119,3407: Updated to dgst.pod
Rich Salz [Thu, 28 Aug 2014 23:11:42 +0000 (19:11 -0400)]
RT2119,3407: Updated to dgst.pod

Re-order algorithm list.
Be consistent in command synopsis.
Add content about signing.
Add EXAMPLE section
Add some missing options: -r, -fips-fingerprint -non-fips-allow
Various other fixes.

Reviewed-by: Andy Polyakov <>
(cherry picked from commit 6aa9dbab0f9b90060b7ee609b8c3c726ce4faf21)

5 years agoRT1941: c_rehash.pod is missing
James Westby [Thu, 14 Aug 2014 14:14:35 +0000 (10:14 -0400)]
RT1941: c_rehash.pod is missing

Add the file written by James Westby, graciously contributed
under the terms of the OpenSSL license.

Reviewed-by: Andy Polyakov <>
(cherry picked from commit cf2239b3b397174a8a6b1cc84ff68aba34ed5941)

5 years agoRT2379: Bug in BIO_set_accept_port.pod
Rich Salz [Mon, 18 Aug 2014 17:00:51 +0000 (13:00 -0400)]
RT2379: Bug in BIO_set_accept_port.pod

The doc says that port can be "*" to mean any port.
That's wrong.

Reviewed-by: Dr. Stephen Henson <>
(cherry picked from commit 07e3b31fae98b985d3d2aad7066144b11833f688)

5 years agoConstant-time utilities
Emilia Kasper [Thu, 28 Aug 2014 13:33:34 +0000 (15:33 +0200)]
Constant-time utilities

Pull constant-time methods out to a separate header, add tests.

Reviewed-by: Bodo Moeller <>
(cherry picked from commit 9a9b0c0401cae443f115ff19921d347b20aa396b)


5 years agoRT2400: ASN1_STRING_to_UTF8 missing initializer
Raphael Spreitzer [Thu, 28 Aug 2014 02:53:10 +0000 (22:53 -0400)]
RT2400: ASN1_STRING_to_UTF8 missing initializer

Reviewed-by: Tim Hudson <>
(cherry picked from commit f9fb43e176ad2a914108cd2b403425dc1ebc7262)

5 years agoRT2308: Add extern "C" { ... } wrapper
Rich Salz [Thu, 28 Aug 2014 01:47:12 +0000 (21:47 -0400)]
RT2308: Add extern "C" { ... } wrapper

Add the wrapper to all public header files (Configure
generates one).  Don't bother for those that are just
lists of #define's that do renaming.

Reviewed-by: Tim Hudson <>
(cherry picked from commit 089f10e69ece75ce31540501fe0898b15e898552)

5 years agoExplicitly check for empty ASN.1 strings in d2i_ECPrivateKey
Emilia Kasper [Mon, 25 Aug 2014 10:38:16 +0000 (12:38 +0200)]
Explicitly check for empty ASN.1 strings in d2i_ECPrivateKey

The old code implicitly relies on the ASN.1 code returning a \0-prefixed buffer
when the buffer length is 0. Change this to verify explicitly that the ASN.1 string
has positive length.

Reviewed-by: Dr Stephen Henson <>
(cherry picked from commit 82dc08de54ce443c2a9ac478faffe79e76157795)

5 years agoRT3065: automatically generate a missing EC public key
Matt Caswell [Fri, 22 Aug 2014 16:04:19 +0000 (18:04 +0200)]
RT3065: automatically generate a missing EC public key

When d2i_ECPrivateKey reads a private key with a missing (optional) public key,
generate one automatically from the group and private key.

Reviewed-by: Dr Stephen Henson <>
(cherry picked from commit ed383f847156940e93f256fed78599873a4a9b28)


5 years agoRT3065: ec_private_key_dont_crash
Adam Langley [Tue, 23 Apr 2013 19:12:36 +0000 (15:12 -0400)]
RT3065: ec_private_key_dont_crash

This change saves several EC routines from crashing when an EC_KEY is
missing a public key. The public key is optional in the EC private key
format and, without this patch, running the following through `openssl
ec` causes a crash:


Reviewed-by: Dr Stephen Henson <>
(cherry picked from commit b391570bdeb386d4fd325917c248d593d3c43930)

5 years agoRT2210: Add missing EVP_cleanup to example
Mihai Militaru [Tue, 26 Aug 2014 16:35:54 +0000 (12:35 -0400)]
RT2210: Add missing EVP_cleanup to example

I also removed some trailing whitespace and cleaned
up the "see also" list.

Reviewed-by: Emilia Kasper <>
(cherry picked from commit 7b3e11c54466f1da8b707c932e308d345fd61101)

5 years agoRT1744: SSL_CTX_set_dump_dh() doc feedback
David Gatwood [Tue, 26 Aug 2014 17:02:03 +0000 (13:02 -0400)]
RT1744: SSL_CTX_set_dump_dh() doc feedback

The description of when the server creates a DH key is
confusing.  This cleans it up.
(rsalz: also removed trailing whitespace.)

Reviewed-by: Viktor Dukhovni <>
5 years agoRT1804: fix EXAMPLE in EVP_EncryptInit.pod
Jan Schaumann [Fri, 15 Aug 2014 03:00:44 +0000 (23:00 -0400)]
RT1804: fix EXAMPLE in EVP_EncryptInit.pod

The EXAMPLE that used FILE and RC2 doesn't compile due to a
few minor errors.  Tweak to use IDEA and AES-128. Remove
examples about RC2 and RC5.

Reviewed-by: Emilia Kasper <>
5 years agoTypo fixes to evp documentation.
Matt Caswell [Thu, 24 Jul 2014 05:00:11 +0000 (01:00 -0400)]
Typo fixes to evp documentation.

This patch was submitted by user "Kox" via the wiki

Reviewed-by: Tim Hudson <>
(cherry picked from commit 2dd8cb3b9593f528d9537aa6a003d5c93df1e3c5)

5 years agoRT3060: Limit the number of empty records.
Adam Langley [Tue, 19 Aug 2014 15:57:53 +0000 (17:57 +0200)]
RT3060: Limit the number of empty records.

Limit the number of empty records that will be processed consecutively
in order to prevent ssl3_get_record from never returning.

Reported by "oftc_must_be_destroyed" and George Kadianakis.

Reviewed-by: Bodo Moeller <>
(cherry picked from commit 3aac17a82fbaf2bc23ee62f24611e5883d3e7b97)

5 years agoRT3061: Don't SEGFAULT when trying to export a public DSA key as a private key.
Adam Langley [Thu, 21 Feb 2013 22:11:55 +0000 (17:11 -0500)]
RT3061: Don't SEGFAULT when trying to export a public DSA key as a private key.

Reviewed-by: Viktor Dukhovni <>
Reviewed-by: Rich Salz <>
(cherry picked from commit e19c93811f0db499c98d2888f1c0c0ab65e6238a)

5 years agoImprove EVP_PKEY_sign documentation
Emilia Kasper [Fri, 22 Aug 2014 11:16:55 +0000 (13:16 +0200)]
Improve EVP_PKEY_sign documentation

Clarify the intended use of EVP_PKEY_sign. Make the code example compile.

Reviewed-by: Dr Stephen Henson <>
(cherry picked from commit d64c533a207f7b6d86c3bc8ffb053e5f4d0c1ca0)

5 years agodefine inline for Visual Studio
Emilia Kasper [Tue, 19 Aug 2014 11:18:07 +0000 (13:18 +0200)]
define inline for Visual Studio

In Visual Studio, inline is available in C++ only, however __inline is available for C, see

Reviewed-by: Richard Levitte <>
Reviewed-by: Dr Stephen Henson <>
(cherry picked from commit f511b25a7370c775dc9fd6198dbacd1706cf242b)

5 years agoFix build when BSAES_ASM is defined but VPAES_ASM is not
Emilia Kasper [Tue, 19 Aug 2014 14:28:07 +0000 (16:28 +0200)]
Fix build when BSAES_ASM is defined but VPAES_ASM is not

Reviewed-by: Andy Polyakov <>
(cherry picked from commit da92be4d68bec81030838e3228ef0238c565af85)

5 years agoAdd volatile qualifications to two blocks of inline asm to stop GCC from
Adam Langley [Mon, 3 Jun 2013 19:45:11 +0000 (15:45 -0400)]
Add volatile qualifications to two blocks of inline asm to stop GCC from
eliminating them as dead code.

Both volatile and "memory" are used because of some concern that the compiler
may still cache values across the asm block without it, and because this was
such a painful debugging session that I wanted to ensure that it's never

(cherry picked from commit 7753a3a68431aa81b82beea4c3f5374b41454679)


Reviewed-by: Rich Salz <>
5 years agoFixed out-of-bounds read errors in ssl3_get_key_exchange.
Matt Caswell [Sat, 26 Jul 2014 22:47:40 +0000 (23:47 +0100)]
Fixed out-of-bounds read errors in ssl3_get_key_exchange.



Reviewed-by: Emilia Käsper <>
5 years agoFurther improve/fix ec_GFp_simple_points_make_affine (ecp_smpl.c) and
Bodo Moeller [Wed, 13 Aug 2014 15:37:19 +0000 (17:37 +0200)]
Further improve/fix ec_GFp_simple_points_make_affine (ecp_smpl.c) and
group_order_tests (ectest.c).  Also fix the EC_POINTs_mul documentation (ec.h).

5 years agoFix SRP ciphersuites.
Dr. Stephen Henson [Tue, 12 Aug 2014 17:33:00 +0000 (18:33 +0100)]
Fix SRP ciphersuites.

Add patch missed from backport of SRP ciphersuite fix.
Reviewed-by: Kurt Roeckx <>
5 years agoFix SRP authentication ciphersuites.
Dr. Stephen Henson [Fri, 8 Aug 2014 10:24:25 +0000 (11:24 +0100)]
Fix SRP authentication ciphersuites.

The addition of SRP authentication needs to be checked in various places
to work properly. Specifically:

A certificate is not sent.
A certificate request must not be sent.
Server key exchange message must not contain a signature.
If appropriate SRP authentication ciphersuites should be chosen.
Reviewed-by: Matt Caswell <>
(cherry picked from commit 8f5a8805b82d1ae81168b11b7f1506db9e047dec)



5 years agoTest SRP authentication ciphersuites.
Dr. Stephen Henson [Fri, 8 Aug 2014 10:19:39 +0000 (11:19 +0100)]
Test SRP authentication ciphersuites.

Reviewed-by: Matt Caswell <>
(cherry picked from commit 193c1c07165b0042abd217274a084b49459d4443)

5 years agoOnly use FIPS EC methods in FIPS mode.
Dr. Stephen Henson [Thu, 7 Aug 2014 00:08:14 +0000 (01:08 +0100)]
Only use FIPS EC methods in FIPS mode.

Reviewed-by: Tim Hudson <>
5 years agoPrepare for 1.0.1j-dev
Matt Caswell [Wed, 6 Aug 2014 21:19:10 +0000 (22:19 +0100)]
Prepare for 1.0.1j-dev

Reviewed-by: Stephen Henson <>
5 years agoPrepare for 1.0.1i release OpenSSL_1_0_1i
Matt Caswell [Wed, 6 Aug 2014 21:18:45 +0000 (22:18 +0100)]
Prepare for 1.0.1i release

Reviewed-by: Stephen Henson <>
5 years agomake update
Matt Caswell [Wed, 6 Aug 2014 21:18:45 +0000 (22:18 +0100)]
make update

Reviewed-by: Stephen Henson <>
5 years agoupdate NEWS
Dr. Stephen Henson [Wed, 6 Aug 2014 13:47:50 +0000 (14:47 +0100)]
update NEWS

Reviewed-by: Kurt Roeckx <>
5 years agoupdate CHANGES
Dr. Stephen Henson [Wed, 6 Aug 2014 13:09:51 +0000 (14:09 +0100)]
update CHANGES

Reviewed-by: Kurt Roeckx <>
5 years agoCheck SRP parameters early.
Dr. Stephen Henson [Fri, 1 Aug 2014 13:56:56 +0000 (14:56 +0100)]
Check SRP parameters early.

Check SRP parameters when they are received so we can send back an
appropriate alert.
Reviewed-by: Kurt Roeckx <>
5 years agoFix SRP buffer overrun vulnerability.
Dr. Stephen Henson [Thu, 31 Jul 2014 19:56:22 +0000 (20:56 +0100)]
Fix SRP buffer overrun vulnerability.

Invalid parameters passed to the SRP code can be overrun an internal
buffer. Add sanity check that g, A, B < N to SRP code.

Thanks to Sean Devlin and Watson Ladd of Cryptography Services, NCC
Group for reporting this issue.
Reviewed-by: Kurt Roeckx <>
5 years agoFix SRP ciphersuite DoS vulnerability.
Dr. Stephen Henson [Tue, 29 Jul 2014 20:23:30 +0000 (21:23 +0100)]
Fix SRP ciphersuite DoS vulnerability.

If a client attempted to use an SRP ciphersuite and it had not been
set up correctly it would crash with a null pointer read. A malicious
server could exploit this in a DoS attack.

Thanks to Joonas Kuorilehto and Riku Hietamäki from Codenomicon
for reporting this issue.

Reviewed-by: Tim Hudson <>
5 years agoFix race condition in ssl_parse_serverhello_tlsext
Gabor Tyukasz [Wed, 23 Jul 2014 21:42:06 +0000 (23:42 +0200)]
Fix race condition in ssl_parse_serverhello_tlsext

Reviewed-by: Tim Hudson <>
Reviewed-by: Dr. Stephen Henson <>
5 years agoFix OID handling:
Emilia Kasper [Wed, 2 Jul 2014 17:02:33 +0000 (19:02 +0200)]
Fix OID handling:

- Upon parsing, reject OIDs with invalid base-128 encoding.
- Always NUL-terminate the destination buffer in OBJ_obj2txt printing function.


Reviewed-by: Dr. Stephen Henson <>
Reviewed-by: Kurt Roeckx <>
Reviewed-by: Tim Hudson <>
5 years agoFix DTLS anonymous EC(DH) denial of service
Emilia Käsper [Thu, 24 Jul 2014 20:15:29 +0000 (22:15 +0200)]
Fix DTLS anonymous EC(DH) denial of service


Reviewed-by: Dr. Stephen Henson <>
5 years agoFix protocol downgrade bug in case of fragmented packets
David Benjamin [Wed, 23 Jul 2014 20:32:21 +0000 (22:32 +0200)]
Fix protocol downgrade bug in case of fragmented packets


Reviewed-by: Emilia Käsper <>
Reviewed-by: Bodo Möller <>
5 years agoRemove some duplicate DTLS code.
Adam Langley [Fri, 6 Jun 2014 21:47:07 +0000 (14:47 -0700)]
Remove some duplicate DTLS code.

In a couple of functions, a sequence number would be calculated twice.

Additionally, in |dtls1_process_out_of_seq_message|, we know that
|frag_len| <= |msg_hdr->msg_len| so the later tests for |frag_len <
msg_hdr->msg_len| can be more clearly written as |frag_len !=
msg_hdr->msg_len|, since that's the only remaining case.

Reviewed-by: Matt Caswell <>
Reviewed-by: Emilia Käsper <>
5 years agoApplying same fix as in dtls1_process_out_of_seq_message. A truncated DTLS fragment...
Matt Caswell [Thu, 24 Jul 2014 22:54:28 +0000 (23:54 +0100)]
Applying same fix as in dtls1_process_out_of_seq_message. A truncated DTLS fragment would cause *ok to be clear, but the return value would still be the number of bytes read.

Problem identified by Emilia Käsper, based on previous issue/patch by Adam

Reviewed-by: Emilia Käsper <>
5 years agoFix return code for truncated DTLS fragment.
Adam Langley [Fri, 6 Jun 2014 21:44:20 +0000 (14:44 -0700)]
Fix return code for truncated DTLS fragment.

Previously, a truncated DTLS fragment in
|dtls1_process_out_of_seq_message| would cause *ok to be cleared, but
the return value would still be the number of bytes read. This would
cause |dtls1_get_message| not to consider it an error and it would
continue processing as normal until the calling function noticed that
*ok was zero.

I can't see an exploit here because |dtls1_get_message| uses
|s->init_num| as the length, which will always be zero from what I can

Reviewed-by: Matt Caswell <>
Reviewed-by: Emilia Käsper <>
5 years agoFix memory leak from zero-length DTLS fragments.
Adam Langley [Fri, 6 Jun 2014 21:30:33 +0000 (14:30 -0700)]
Fix memory leak from zero-length DTLS fragments.

The |pqueue_insert| function can fail if one attempts to insert a
duplicate sequence number. When handling a fragment of an out of
sequence message, |dtls1_process_out_of_seq_message| would not call
|dtls1_reassemble_fragment| if the fragment's length was zero. It would
then allocate a fresh fragment and attempt to insert it, but ignore the
return value, leaking the fragment.

This allows an attacker to exhaust the memory of a DTLS peer.

Fixes CVE-2014-3507

Reviewed-by: Matt Caswell <>
Reviewed-by: Emilia Käsper <>
5 years agoFix DTLS handshake message size checks.
Matt Caswell [Fri, 6 Jun 2014 21:25:52 +0000 (14:25 -0700)]
Fix DTLS handshake message size checks.

In |dtls1_reassemble_fragment|, the value of
|msg_hdr->frag_off+frag_len| was being checked against the maximum
handshake message size, but then |msg_len| bytes were allocated for the
fragment buffer. This means that so long as the fragment was within the
allowed size, the pending handshake message could consume 16MB + 2MB
(for the reassembly bitmap). Approx 10 outstanding handshake messages
are allowed, meaning that an attacker could consume ~180MB per DTLS

In the non-fragmented path (in |dtls1_process_out_of_seq_message|), no
check was applied.

Fixes CVE-2014-3506

Wholly based on patch by Adam Langley with one minor amendment.

Reviewed-by: Emilia Käsper <>
5 years agoAdded comment for the frag->reassembly == NULL case as per feedback from Emilia
Matt Caswell [Thu, 24 Jul 2014 22:33:34 +0000 (23:33 +0100)]
Added comment for the frag->reassembly == NULL case as per feedback from Emilia

Reviewed-by: Emilia Käsper <>
5 years agoAvoid double free when processing DTLS packets.
Adam Langley [Fri, 6 Jun 2014 21:19:21 +0000 (14:19 -0700)]
Avoid double free when processing DTLS packets.

The |item| variable, in both of these cases, may contain a pointer to a
|pitem| structure within |s->d1->buffered_messages|. It was being freed
in the error case while still being in |buffered_messages|. When the
error later caused the |SSL*| to be destroyed, the item would be double

Thanks to Wah-Teh Chang for spotting that the fix in 1632ef74 was
inconsistent with the other error paths (but correct).

Fixes CVE-2014-3505

Reviewed-by: Matt Caswell <>
Reviewed-by: Emilia Käsper <>
5 years agomake update
Dr. Stephen Henson [Fri, 1 Aug 2014 19:44:37 +0000 (20:44 +0100)]
make update

Reviewed-by: Tim Hudson <>
5 years agoFix error discrepancy.
Dr. Stephen Henson [Thu, 31 Jul 2014 20:22:23 +0000 (21:22 +0100)]
Fix error discrepancy.

We can't rename ssleay_rand_bytes to md_rand_bytes_lock as this will cause
an error code discrepancy. Instead keep ssleay_rand_bytes and add an
extra parameter: since ssleay_rand_bytes is not part of the public API
this wont cause any binary compatibility issues.
Reviewed-by: Kurt Roeckx < >
(cherry picked from commit 8068a675a7d1a657c54546f24e673e59e6707f03)

5 years agoUpdate $default_depflags to match current defaults.
Bodo Moeller [Fri, 1 Aug 2014 17:07:57 +0000 (19:07 +0200)]
Update $default_depflags to match current defaults.

5 years agoSimplify and fix ec_GFp_simple_points_make_affine
Bodo Moeller [Fri, 1 Aug 2014 15:18:14 +0000 (17:18 +0200)]
Simplify and fix ec_GFp_simple_points_make_affine
(which didn't always handle value 0 correctly).


5 years agoAvoid multiple lock using FIPS DRBG.
Dr. Stephen Henson [Wed, 30 Jul 2014 14:13:08 +0000 (15:13 +0100)]
Avoid multiple lock using FIPS DRBG.

Don't use multiple locks when SP800-90 DRBG is used outside FIPS mode.

Reviewed-by: Rich Salz <>
(cherry picked from commit a3efe1b6e9d2aa2ce5661e4d4b97262eae743fa7)

5 years agoAdd conditional unit testing interface.
Dr. Stephen Henson [Wed, 23 Jul 2014 12:18:06 +0000 (13:18 +0100)]
Add conditional unit testing interface.

Don't call internal functions directly call them through
SSL_test_functions(). This also makes unit testing work on
Windows and platforms that don't export internal functions
from shared libraries.

By default unit testing is not enabled: it requires the compile
time option "enable-unit-test".
Reviewed-by: Geoff Thorpe <>
(cherry picked from commit e0fc7961c4fbd27577fb519d9aea2dc788742715)



5 years ago"EC_POINT_invert" was checking "dbl" function pointer instead of "invert".
Billy Brumley [Mon, 21 Jul 2014 21:08:23 +0000 (22:08 +0100)]
"EC_POINT_invert" was checking "dbl" function pointer instead of "invert".


Reviewed-by: Rich Salz <>
(cherry picked from commit cba11f57ce161fd301a72194827327128191de7e)

5 years agoRemove old unused and unmaintained demonstration code.
Tim Hudson [Mon, 21 Jul 2014 19:26:17 +0000 (05:26 +1000)]
Remove old unused and unmaintained demonstration code.

Reviewed-by: Dr. Stephen Henson <>
(cherry picked from commit 62352b8138018775a4c085a105fccd9cdcb6323f)

5 years agoMinor documentation update removing "really" and a
Tim Hudson [Mon, 21 Jul 2014 10:03:50 +0000 (20:03 +1000)]
Minor documentation update removing "really" and a
statement of opinion rather than a fact.

Reviewed-by: Dr. Stephen Henson <>
Reviewed-by: Rich Salz <>
(cherry picked from commit c8d133e4b6f1ed1b7ad3c1a6d2c62f460e26c050)

5 years agoFix documentation for RSA_set_method(3)
Dr. Stephen Henson [Sat, 19 Jul 2014 13:20:05 +0000 (14:20 +0100)]
Fix documentation for RSA_set_method(3)

Reviewed-by: Matt Caswell <>
(cherry picked from commit 197400c3f0d617d71ad8167b52fb73046d334320)

5 years agoFix typo, add reference.
Jeffrey Walton [Thu, 17 Jul 2014 10:25:02 +0000 (11:25 +0100)]
Fix typo, add reference.

Reviewed-by: Stephen Henson <>
Reviewed-by: Matt Caswell <>
(cherry picked from commit d48e78f0cf22aaddb563f4bcfccf25b1a45ac8a4)

5 years agoDisabled XTS mode in enc utility as it is not supported
Matt Caswell [Sun, 13 Jul 2014 22:28:13 +0000 (23:28 +0100)]
Disabled XTS mode in enc utility as it is not supported


Reviewed-by: Tim Hudson <>
Reviewed-by: Rich Salz <>
(cherry picked from commit 2097a17c576f2395a10b05f14490688bc5f45a07)

5 years agoAdd Matt Caswell's fingerprint, and general update on the fingerprints file to bring...
Matt Caswell [Tue, 15 Jul 2014 21:47:29 +0000 (22:47 +0100)]
Add Matt Caswell's fingerprint, and general update on the fingerprints file to bring it up to date

Reviewed-by: Tim Hudson <>
(cherry picked from commit 3bd548192a03142c80cf8bc68659d79dea20a738)

5 years agoClarify -Verify and PSK.
Dr. Stephen Henson [Tue, 15 Jul 2014 19:22:39 +0000 (20:22 +0100)]
Clarify -Verify and PSK.

(cherry picked from commit ca2015a617842fed3d36ed4dcbbf8d5e27bc5216)

5 years agoFix DTLS certificate requesting code.
Dr. Stephen Henson [Tue, 15 Jul 2014 17:21:59 +0000 (18:21 +0100)]
Fix DTLS certificate requesting code.

Use same logic when determining when to expect a client
certificate for both TLS and DTLS.

(cherry picked from commit c8d710dc5f83d69d802f941a4cc5895eb5fe3d65)

5 years agoDon't allow -www etc options with DTLS.
Dr. Stephen Henson [Tue, 15 Jul 2014 11:22:49 +0000 (12:22 +0100)]
Don't allow -www etc options with DTLS.

The options which emulate a web server don't make sense when doing DTLS.
Exit with an error if an attempt is made to use them.

(cherry picked from commit 58a2aaeade8bdecd0f9f0df41927f7cff3012547)

5 years agoAdd ECC extensions with DTLS.
Dr. Stephen Henson [Tue, 15 Jul 2014 11:20:30 +0000 (12:20 +0100)]
Add ECC extensions with DTLS.


5 years agoUse case insensitive compare for servername.
Dr. Stephen Henson [Mon, 14 Jul 2014 22:59:13 +0000 (23:59 +0100)]
Use case insensitive compare for servername.

(cherry picked from commit 1c3e9a7c67ccdc5e770829fe951e5832e600d377)

5 years agodocument -nextprotoneg option in man pages
Hubert Kario [Fri, 6 Jun 2014 14:48:43 +0000 (16:48 +0200)]
document -nextprotoneg option in man pages

Add description of the option to advertise support of
Next Protocol Negotiation extension (-nextprotoneg) to
man pages of s_client and s_server.

(cherry picked from commit 7efd0e777e65eaa6c60d85b1cc5c889f872f8fc4)



5 years agoUse more common name for GOST key exchange.
Dr. Stephen Henson [Mon, 14 Jul 2014 14:05:50 +0000 (15:05 +0100)]
Use more common name for GOST key exchange.
(cherry picked from commit 7aabd9c92fe6f0ea2a82869e5171dcc4518cee85)

5 years agoFixed valgrind complaint due to BN_consttime_swap reading uninitialised data.
Matt Caswell [Thu, 10 Jul 2014 22:47:31 +0000 (23:47 +0100)]
Fixed valgrind complaint due to BN_consttime_swap reading uninitialised data.
This is actually ok for this function, but initialised to zero anyway if
PURIFY defined.

This does have the impact of masking any *real* unitialised data reads in bn though.

Patch based on approach suggested by Rich Salz.


(cherry picked from commit 77747e2d9a5573b1dbc15e247ce18c03374c760c)

5 years agoAdd names of GOST algorithms.
Peter Mosmans [Sun, 13 Jul 2014 17:30:07 +0000 (18:30 +0100)]
Add names of GOST algorithms.

(cherry picked from commit 924e5eda2c82d737cc5a1b9c37918aa6e34825da)

5 years ago* crypto/ui/ui_lib.c: misplaced brace in switch statement.
Richard Levitte [Sun, 13 Jul 2014 17:11:29 +0000 (19:11 +0200)]
* crypto/ui/ui_lib.c: misplaced brace in switch statement.
  Detected by

(cherry picked from commit 8b5dd340919e511137696792279f595a70ae2762)

5 years agoDon't clean up uninitialised EVP_CIPHER_CTX on error (CID 483259).
Ben Laurie [Thu, 10 Jul 2014 16:47:48 +0000 (17:47 +0100)]
Don't clean up uninitialised EVP_CIPHER_CTX on error (CID 483259).

(cherry picked from commit c1d1b0114e9d370c30649e46182393dbfc00e20c)