oweals/openssl.git
8 years agoRemove some documentation for functions not in 1.0.x
Matt Caswell [Tue, 3 May 2016 11:45:45 +0000 (12:45 +0100)]
Remove some documentation for functions not in 1.0.x

A few functions in the recently added EVP_EncodeInit docs don't apply to
the 1.0.x branches.

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoAdd documentation for EVP_EncodeInit() and similar functions
Matt Caswell [Mon, 25 Apr 2016 10:54:30 +0000 (11:54 +0100)]
Add documentation for EVP_EncodeInit() and similar functions

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoEnsure EVP_EncodeUpdate handles an output length that is too long
Matt Caswell [Mon, 25 Apr 2016 08:06:29 +0000 (09:06 +0100)]
Ensure EVP_EncodeUpdate handles an output length that is too long

With the EVP_EncodeUpdate function it is the caller's responsibility to
determine how big the output buffer should be. The function writes the
amount actually used to |*outl|. However this could go negative with a
sufficiently large value for |inl|. We add a check for this error
condition.

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoAvoid overflow in EVP_EncodeUpdate
Matt Caswell [Fri, 4 Mar 2016 10:17:17 +0000 (10:17 +0000)]
Avoid overflow in EVP_EncodeUpdate

An overflow can occur in the EVP_EncodeUpdate function which is used for
Base64 encoding of binary data. If an attacker is able to supply very large
amounts of input data then a length check can overflow resulting in a heap
corruption. Due to the very large amounts of data involved this will most
likely result in a crash.

Internally to OpenSSL the EVP_EncodeUpdate function is primarly used by the
PEM_write_bio* family of functions. These are mainly used within the
OpenSSL command line applications, so any application which processes
data from an untrusted source and outputs it as a PEM file should be
considered vulnerable to this issue.

User applications that call these APIs directly with large amounts of
untrusted data may also be vulnerable.

Issue reported by Guido Vranken.

CVE-2016-2105

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoPrevent EBCDIC overread for very long strings
Matt Caswell [Thu, 28 Apr 2016 09:46:55 +0000 (10:46 +0100)]
Prevent EBCDIC overread for very long strings

ASN1 Strings that are over 1024 bytes can cause an overread in
applications using the X509_NAME_oneline() function on EBCDIC systems.
This could result in arbitrary stack data being returned in the buffer.

Issue reported by Guido Vranken.

CVE-2016-2176

Reviewed-by: Andy Polyakov <appro@openssl.org>
8 years agoFix encrypt overflow
Matt Caswell [Thu, 3 Mar 2016 23:36:23 +0000 (23:36 +0000)]
Fix encrypt overflow

An overflow can occur in the EVP_EncryptUpdate function. If an attacker is
able to supply very large amounts of input data after a previous call to
EVP_EncryptUpdate with a partial block then a length check can overflow
resulting in a heap corruption.

Following an analysis of all OpenSSL internal usage of the
EVP_EncryptUpdate function all usage is one of two forms.

The first form is like this:
EVP_EncryptInit()
EVP_EncryptUpdate()

i.e. where the EVP_EncryptUpdate() call is known to be the first called
function after an EVP_EncryptInit(), and therefore that specific call
must be safe.

The second form is where the length passed to EVP_EncryptUpdate() can be
seen from the code to be some small value and therefore there is no
possibility of an overflow.

Since all instances are one of these two forms, I believe that there can
be no overflows in internal code due to this problem.

It should be noted that EVP_DecryptUpdate() can call EVP_EncryptUpdate()
in certain code paths. Also EVP_CipherUpdate() is a synonym for
EVP_EncryptUpdate(). Therefore I have checked all instances of these
calls too, and came to the same conclusion, i.e. there are no instances
in internal usage where an overflow could occur.

This could still represent a security issue for end user code that calls
this function directly.

CVE-2016-2106

Issue reported by Guido Vranken.

Reviewed-by: Tim Hudson <tjh@openssl.org>
(cherry picked from commit 3f3582139fbb259a1c3cbb0a25236500a409bf26)

8 years agoFix i2d_X509_AUX: pp can be NULL.
Dr. Stephen Henson [Mon, 2 May 2016 16:33:50 +0000 (17:33 +0100)]
Fix i2d_X509_AUX: pp can be NULL.

Reported by David Benjamin

Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit 05aef4bbdbc18e7b9490512cdee41e8a608bcc0e)

8 years agoRemove confusing comment.
TJ Saunders [Fri, 29 Apr 2016 14:40:28 +0000 (07:40 -0700)]
Remove confusing comment.

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
8 years agoIssue #719:
TJ Saunders [Wed, 23 Mar 2016 18:55:53 +0000 (11:55 -0700)]
Issue #719:

If no serverinfo extension is found in some cases, do not abort the handshake,
but simply omit/skip that extension.

Check for already-registered serverinfo callbacks during serverinfo
registration.

Update SSL_CTX_use_serverinfo() documentation to mention the need to reload the
same serverinfo per certificate, for servers with multiple server certificates.

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
8 years agoDon't free ret->data if malloc fails.
Dr. Stephen Henson [Wed, 27 Apr 2016 19:27:41 +0000 (20:27 +0100)]
Don't free ret->data if malloc fails.

Issue reported by Guido Vranken.

Reviewed-by: Matt Caswell <matt@openssl.org>
8 years agoAdd checks to X509_NAME_oneline()
Dr. Stephen Henson [Thu, 28 Apr 2016 18:45:44 +0000 (19:45 +0100)]
Add checks to X509_NAME_oneline()

Sanity check field lengths and sums to avoid potential overflows and reject
excessively large X509_NAME structures.

Issue reported by Guido Vranken.

Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit 77076dc944f76e821e4eae3a6563b853ce00c0ed)

Conflicts:
crypto/x509/x509_err.c
crypto/x509/x509_obj.c

8 years agoSanity check buffer length.
Dr. Stephen Henson [Thu, 28 Apr 2016 12:09:27 +0000 (13:09 +0100)]
Sanity check buffer length.

Reject zero length buffers passed to X509_NAME_onelne().

Issue reported by Guido Vranken.

Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit b33d1141b6dcce947708b984c5e9e91dad3d675d)

8 years agoAdd size limit to X509_NAME structure.
Dr. Stephen Henson [Thu, 28 Apr 2016 11:55:29 +0000 (12:55 +0100)]
Add size limit to X509_NAME structure.

This adds an explicit limit to the size of an X509_NAME structure. Some
part of OpenSSL (e.g. TLS) already effectively limit the size due to
restrictions on certificate size.

Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit 295f3a24919157e2f9021d0b1709353710ad63db)

8 years agoReject inappropriate private key encryption ciphers.
Dr. Stephen Henson [Sat, 23 Apr 2016 12:33:05 +0000 (13:33 +0100)]
Reject inappropriate private key encryption ciphers.

The traditional private key encryption algorithm doesn't function
properly if the IV length of the cipher is zero. These ciphers
(e.g. ECB mode) are not suitable for private key encryption
anyway.

Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit d78df5dfd650e6de159a19a033513481064644f5)

8 years agoEnsure we check i2d_X509 return val
Matt Caswell [Mon, 25 Apr 2016 15:05:55 +0000 (16:05 +0100)]
Ensure we check i2d_X509 return val

The i2d_X509() function can return a negative value on error. Therefore
we should make sure we check it.

Issue reported by Yuan Jochen Kang.

Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit 446ba8de9af9aa4fa3debc7c76a38f4efed47a62)

8 years agoFix a signed/unsigned warning
Matt Caswell [Mon, 25 Apr 2016 16:45:11 +0000 (17:45 +0100)]
Fix a signed/unsigned warning

This causes a compilation failure when using --strict-warnings in 1.0.2
and 1.0.1

Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
8 years agoFix NULL deref in apps/pkcs7
Rich Salz [Mon, 25 Apr 2016 12:56:54 +0000 (08:56 -0400)]
Fix NULL deref in apps/pkcs7

Thanks to Brian Carpenter for finding and reporting this.

Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit 79356a83b78a2d936dcd022847465d9ebf6c67b1)

8 years agos390x assembly pack: cache capability query results.
Andy Polyakov [Fri, 15 Apr 2016 14:39:22 +0000 (16:39 +0200)]
s390x assembly pack: cache capability query results.

IBM argues that in certain scenarios capability query is really
expensive. At the same time it's asserted that query results can
be safely cached, because disabling CPACF is incompatible with
reboot-free operation.

Reviewed-by: Tim Hudson <tjh@openssl.org>
(cherry picked from commit 670ad0fbf6ebcf113e278d8174081a7e2d2fa44c)

8 years agoFix buffer overrun in ASN1_parse().
Viktor Dukhovni [Wed, 20 Apr 2016 02:23:24 +0000 (22:23 -0400)]
Fix buffer overrun in ASN1_parse().

Backport of commits:

        79c7f74d6cefd5d32fa20e69195ad3de834ce065
bdcd660e33710079b495cf5cc6a1aaa5d2dcd317

from master.

Reviewed-by: Matt Caswell <matt@openssl.org>
8 years agoHarden ASN.1 BIO handling of large amounts of data.
Dr. Stephen Henson [Mon, 11 Apr 2016 12:57:20 +0000 (13:57 +0100)]
Harden ASN.1 BIO handling of large amounts of data.

If the ASN.1 BIO is presented with a large length field read it in
chunks of increasing size checking for EOF on each read. This prevents
small files allocating excessive amounts of data.

CVE-2016-2109

Thanks to Brian Carpenter for reporting this issue.

Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
(cherry picked from commit c62981390d6cf9e3d612c489b8b77c2913b25807)

8 years agoPPC assembly pack: remove branch hints.
Andy Polyakov [Thu, 31 Mar 2016 16:47:17 +0000 (18:47 +0200)]
PPC assembly pack: remove branch hints.

As it turns out branch hints grew as kind of a misconception. In
addition their interpretation by GNU assembler is affected by
assembler flags and can end up with opposite meaning on different
processors. As we have to loose quite a lot on misinterprerations,
especially on newer processors, we just omit them altogether.

Reviewed-by: Tim Hudson <tjh@openssl.org>
(cherry picked from commit 20b88bb160865b0c2998056fe841b1cbacd6d4c7)

8 years agoFix memory leak on invalid CertificateRequest.
David Benjamin [Mon, 14 Mar 2016 19:03:07 +0000 (15:03 -0400)]
Fix memory leak on invalid CertificateRequest.

Free up parsed X509_NAME structure if the CertificateRequest message
contains excess data.

The security impact is considered insignificant. This is a client side
only leak and a large number of connections to malicious servers would
be needed to have a significant impact.

This was found by libFuzzer.

Reviewed-by: Emilia Käsper <emilia@openssl.org>
Reviewed-by: Stephen Henson <steve@openssl.org>
8 years agoFix ALPN
Todd Short [Sat, 5 Mar 2016 13:47:55 +0000 (08:47 -0500)]
Fix ALPN

* Perform ALPN after the SNI callback; the SSL_CTX may change due to
  that processing
* Add flags to indicate that we actually sent ALPN, to properly error
  out if unexpectedly received.
* document ALPN functions
* unit tests

Backport of commit 817cd0d52f0462039d1fe60462150be7f59d2002

Reviewed-by: Emilia Käsper <emilia@openssl.org>
Reviewed-by: Dr. Stephen Henson <steve@openssl.org>
8 years agoFix FIPS SSLv2 test
Dr. Stephen Henson [Sat, 26 Mar 2016 15:00:53 +0000 (15:00 +0000)]
Fix FIPS SSLv2 test

Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
8 years agoFix the no-comp option for Windows
Matt Caswell [Thu, 17 Mar 2016 12:55:02 +0000 (12:55 +0000)]
Fix the no-comp option for Windows

no-comp on Windows was not actually suppressing compilation of the code,
although it was suppressing its use.

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoAdd a check for a failed malloc
Matt Caswell [Tue, 15 Mar 2016 11:51:48 +0000 (11:51 +0000)]
Add a check for a failed malloc

Ensure we check for a NULL return from OPENSSL_malloc

Issue reported by Guido Vranken.

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoEnsure that memory allocated for the ticket is freed
Matt Caswell [Tue, 15 Mar 2016 11:38:56 +0000 (11:38 +0000)]
Ensure that memory allocated for the ticket is freed

If a call to EVP_DecryptUpdate fails then a memory leak could occur.
Ensure that the memory is freed appropriately.

Issue reported by Guido Vranken.

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoFix a potential double free in EVP_DigestInit_ex
Matt Caswell [Mon, 14 Mar 2016 17:06:19 +0000 (17:06 +0000)]
Fix a potential double free in EVP_DigestInit_ex

There is a potential double free in EVP_DigestInit_ex. This is believed
to be reached only as a result of programmer error - but we should fix it
anyway.

Issue reported by Guido Vranken.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(cherry picked from commit ffe9150b1508a0ffc9e724f975691f24eb045c05)

8 years agoAdd no-ssl2-method
Kurt Roeckx [Wed, 9 Mar 2016 17:10:52 +0000 (18:10 +0100)]
Add no-ssl2-method

Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
MR: #2341

8 years agoTouch the correct variables for the system; shlib_wrap.sh on Solaris
Richard Levitte [Wed, 9 Mar 2016 10:36:32 +0000 (11:36 +0100)]
Touch the correct variables for the system; shlib_wrap.sh on Solaris

If there is cause to think LD_LIBRARY_PATH_32 and LD_PRELOAD_32 are
appropriate variables to touch, do so.  Otherwise, touch the usual
LD_LIBRARY_PATH and LD_PRELOAD.  This covers for older installations
that don't have a mix of 32-bit and 64-bit libs.

Reviewed-by: Andy Polyakov <appro@openssl.org>
(cherry picked from commit a772e9d01a81dae132cb03107292b3ecc725e5af)

8 years agoRecognise 32-bit Solaris in util/shlib_wrap.sh
Richard Levitte [Sun, 6 Mar 2016 21:36:57 +0000 (22:36 +0100)]
Recognise 32-bit Solaris in util/shlib_wrap.sh

Submitted by Erik Forsberg <erik@efca.com>

Reviewed-by: Andy Polyakov <appro@openssl.org>
(cherry picked from commit 86e75432835ec3ef7700a7103f9847e3c3d4350d)

8 years agoexpose SSLv2 method prototypes
Viktor Dukhovni [Tue, 8 Mar 2016 20:33:02 +0000 (15:33 -0500)]
expose SSLv2 method prototypes

Reviewed-by: Kurt Roeckx <kurt@openssl.org>
8 years agoRetain SSLv2 methods as functions that return NULL
Viktor Dukhovni [Mon, 7 Mar 2016 21:10:38 +0000 (21:10 +0000)]
Retain SSLv2 methods as functions that return NULL

This improves ABI compatibility when symbol resolution is not lazy.

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoRemove LOW from the default
Kurt Roeckx [Sun, 10 Jan 2016 12:23:43 +0000 (13:23 +0100)]
Remove LOW from the default

Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
8 years agoDon't shift serial number into sign bit
Dr. Stephen Henson [Fri, 4 Mar 2016 18:04:46 +0000 (18:04 +0000)]
Don't shift serial number into sign bit

Reviewed-by: Rich Salz <rsalz@openssl.org>
8 years agobn/asm/x86[_64]-mont*.pl: complement alloca with page-walking.
Andy Polyakov [Fri, 4 Mar 2016 10:39:11 +0000 (11:39 +0100)]
bn/asm/x86[_64]-mont*.pl: complement alloca with page-walking.

Some OSes, *cough*-dows, insist on stack being "wired" to
physical memory in strictly sequential manner, i.e. if stack
allocation spans two pages, then reference to farmost one can
be punishable by SEGV. But page walking can do good even on
other OSes, because it guarantees that villain thread hits
the guard page before it can make damage to innocent one...

Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit adc4f1fc25b2cac90076f1e1695b05b7aeeae501)

8 years agoperlasm/x86_64-xlate.pl: handle binary constants early.
Andy Polyakov [Fri, 4 Mar 2016 10:32:26 +0000 (11:32 +0100)]
perlasm/x86_64-xlate.pl: handle binary constants early.

Not all assemblers of "gas" flavour handle binary constants, e.g.
seasoned MacOS Xcode doesn't, so give them a hand.

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
(cherry picked from commit 6e42e3ff9cde43830555549fdafa2a8b37b9485b)

8 years agoSanity check PVK file fields.
Dr. Stephen Henson [Thu, 3 Mar 2016 23:37:36 +0000 (23:37 +0000)]
Sanity check PVK file fields.

PVK files with abnormally large length or salt fields can cause an
integer overflow which can result in an OOB read and heap corruption.
However this is an rarely used format and private key files do not
normally come from untrusted sources the security implications not
significant.

Fix by limiting PVK length field to 100K and salt to 10K: these should be
more than enough to cover any files encountered in practice.

Issue reported by Guido Vranken.

Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit 5f57abe2b150139b8b057313d52b1fe8f126c952)

8 years agoRT4358: Extra ] in doc/ocsp.pod
Eric S. Raymond [Mon, 29 Feb 2016 19:54:47 +0000 (14:54 -0500)]
RT4358: Extra ] in doc/ocsp.pod

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoPrepare for 1.0.2h-dev
Matt Caswell [Tue, 1 Mar 2016 13:37:56 +0000 (13:37 +0000)]
Prepare for 1.0.2h-dev

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoPrepare for 1.0.2g release OpenSSL_1_0_2g
Matt Caswell [Tue, 1 Mar 2016 13:36:54 +0000 (13:36 +0000)]
Prepare for 1.0.2g release

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agomake update
Matt Caswell [Tue, 1 Mar 2016 13:36:54 +0000 (13:36 +0000)]
make update

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoEnsure mk1mf.pl is aware of no-weak-ssl-ciphers option
Matt Caswell [Tue, 1 Mar 2016 12:08:33 +0000 (12:08 +0000)]
Ensure mk1mf.pl is aware of no-weak-ssl-ciphers option

Update mk1mf.pl to properly handle no-weak-ssl-ciphers

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoUpdate CHANGES and NEWS for new release
Matt Caswell [Tue, 1 Mar 2016 11:00:48 +0000 (11:00 +0000)]
Update CHANGES and NEWS for new release

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agobn/asm/x86_64-mont5.pl: unify gather procedure in hardly used path
Andy Polyakov [Tue, 26 Jan 2016 15:50:10 +0000 (16:50 +0100)]
bn/asm/x86_64-mont5.pl: unify gather procedure in hardly used path
and reorganize/harmonize post-conditions.

Additional hardening following on from CVE-2016-0702

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from master)

8 years agocrypto/bn/x86_64-mont5.pl: constant-time gather procedure.
Andy Polyakov [Mon, 25 Jan 2016 22:41:01 +0000 (23:41 +0100)]
crypto/bn/x86_64-mont5.pl: constant-time gather procedure.

At the same time remove miniscule bias in final subtraction.
Performance penalty varies from platform to platform, and even with
key length. For rsa2048 sign it was observed to be 4% for Sandy
Bridge and 7% on Broadwell.

CVE-2016-0702

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from master)

8 years agobn/asm/rsaz-avx2.pl: constant-time gather procedure.
Andy Polyakov [Mon, 25 Jan 2016 22:25:40 +0000 (23:25 +0100)]
bn/asm/rsaz-avx2.pl: constant-time gather procedure.

Performance penalty is 2%.

CVE-2016-0702

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from master)

8 years agobn/asm/rsax-x86_64.pl: constant-time gather procedure.
Andy Polyakov [Mon, 25 Jan 2016 22:06:45 +0000 (23:06 +0100)]
bn/asm/rsax-x86_64.pl: constant-time gather procedure.

Performance penalty is 2% on Linux and 5% on Windows.

CVE-2016-0702

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from master)

8 years agobn/bn_exp.c: constant-time MOD_EXP_CTIME_COPY_FROM_PREBUF.
Andy Polyakov [Mon, 25 Jan 2016 19:38:38 +0000 (20:38 +0100)]
bn/bn_exp.c: constant-time MOD_EXP_CTIME_COPY_FROM_PREBUF.

Performance penalty varies from platform to platform, and even
key length. For rsa2048 sign it was observed to reach almost 10%.

CVE-2016-0702

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from master)

Resolved conflicts:
crypto/bn/bn_exp.c

8 years agoDisable EXPORT and LOW SSLv3+ ciphers by default
Viktor Dukhovni [Fri, 19 Feb 2016 18:05:11 +0000 (13:05 -0500)]
Disable EXPORT and LOW SSLv3+ ciphers by default

Reviewed-by: Emilia Käsper <emilia@openssl.org>
8 years agoAdd a test for SSLv2 configuration
Matt Caswell [Fri, 19 Feb 2016 11:38:25 +0000 (11:38 +0000)]
Add a test for SSLv2 configuration

SSLv2 should be off by default. You can only turn it on if you have called
SSL_CTX_clear_options(SSL_OP_NO_SSLv2) or
SSL_clear_options(SSL_OP_NO_SSLv2). You should not be able to inadvertantly
turn it on again via SSL_CONF without having done that first.

Reviewed-by: Emilia Käsper <emilia@openssl.org>
8 years agoBring SSL method documentation up to date
Viktor Dukhovni [Thu, 18 Feb 2016 04:38:55 +0000 (23:38 -0500)]
Bring SSL method documentation up to date

Reviewed-by: Emilia Käsper <emilia@openssl.org>
8 years agoDisable SSLv2 default build, default negotiation and weak ciphers.
Viktor Dukhovni [Thu, 18 Feb 2016 02:07:48 +0000 (21:07 -0500)]
Disable SSLv2 default build, default negotiation and weak ciphers.

SSLv2 is by default disabled at build-time.  Builds that are not
configured with "enable-ssl2" will not support SSLv2.  Even if
"enable-ssl2" is used, users who want to negotiate SSLv2 via the
version-flexible SSLv23_method() will need to explicitly call either
of:

    SSL_CTX_clear_options(ctx, SSL_OP_NO_SSLv2);
or
    SSL_clear_options(ssl, SSL_OP_NO_SSLv2);

as appropriate.  Even if either of those is used, or the application
explicitly uses the version-specific SSLv2_method() or its client
or server variants, SSLv2 ciphers vulnerable to exhaustive search
key recovery have been removed.  Specifically, the SSLv2 40-bit
EXPORT ciphers, and SSLv2 56-bit DES are no longer available.

Mitigation for CVE-2016-0800

Reviewed-by: Emilia Käsper <emilia@openssl.org>
8 years agoFix BN_hex2bn/BN_dec2bn NULL ptr/heap corruption
Matt Caswell [Mon, 22 Feb 2016 10:27:18 +0000 (10:27 +0000)]
Fix BN_hex2bn/BN_dec2bn NULL ptr/heap corruption

In the BN_hex2bn function the number of hex digits is calculated using
an int value |i|. Later |bn_expand| is called with a value of |i * 4|.
For large values of |i| this can result in |bn_expand| not allocating any
memory because |i * 4| is negative. This leaves ret->d as NULL leading
to a subsequent NULL ptr deref. For very large values of |i|, the
calculation |i * 4| could be a positive value smaller than |i|. In this
case memory is allocated to ret->d, but it is insufficiently sized
leading to heap corruption. A similar issue exists in BN_dec2bn.

This could have security consequences if BN_hex2bn/BN_dec2bn is ever
called by user applications with very large untrusted hex/dec data. This is
anticipated to be a rare occurrence.

All OpenSSL internal usage of this function uses data that is not expected
to be untrusted, e.g. config file data or application command line
arguments. If user developed applications generate config file data based
on untrusted data then it is possible that this could also lead to security
consequences. This is also anticipated to be a rare.

Issue reported by Guido Vranken.

CVE-2016-0797

Reviewed-by: Andy Polyakov <appro@openssl.org>
8 years agoFix invalid free
Kurt Roeckx [Sun, 28 Feb 2016 10:40:12 +0000 (11:40 +0100)]
Fix invalid free

Also removes an extra initializer, the master branch has a extra field.

Reviewed-by: Rich Salz <rsalz@openssl.org>
8 years agoFix two possible leaks
FdaSilvaYY [Fri, 29 Jan 2016 18:53:28 +0000 (19:53 +0100)]
Fix two possible leaks

Backport of 98637bd

Reviewed-by: Kurt Roeckx <kurt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
8 years agoFix possible memory leak on BUF_MEM_grow_clean failure
FdaSilvaYY [Fri, 29 Jan 2016 18:49:38 +0000 (19:49 +0100)]
Fix possible memory leak on BUF_MEM_grow_clean failure

backport of 3eb70c5ebae6f2b5fd6034ed5af14910c8479688
shorter changes

Reviewed-by: Kurt Roeckx <kurt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
8 years agomodes/asm/ghash-x86_64.pl: refine GNU assembler version detection.
Andy Polyakov [Wed, 24 Feb 2016 19:07:06 +0000 (20:07 +0100)]
modes/asm/ghash-x86_64.pl: refine GNU assembler version detection.

Even though AVX support was added in GAS 2.19 vpclmulqdq was apparently
added in 2.20.

Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit d3cdab17362b9ab0f1a0525b736c862937a9cd2f)

8 years agoRevert "Don't check RSA_FLAG_SIGN_VER."
Kurt Roeckx [Sat, 20 Feb 2016 14:29:45 +0000 (15:29 +0100)]
Revert "Don't check RSA_FLAG_SIGN_VER."

This reverts commit 6656ba7152dfe4bba865e327dd362ea08544aa80.

This broke existing engines that didn't properly implement the sign and verify
functions.

Reviewed-by: Richard Levitte <levitte@openssl.org>
MR: #2077

8 years agoFix memory issues in BIO_*printf functions
Matt Caswell [Thu, 25 Feb 2016 13:09:46 +0000 (13:09 +0000)]
Fix memory issues in BIO_*printf functions

The internal |fmtstr| function used in processing a "%s" format string
in the BIO_*printf functions could overflow while calculating the length
of a string and cause an OOB read when printing very long strings.

Additionally the internal |doapr_outch| function can attempt to write to
an OOB memory location (at an offset from the NULL pointer) in the event of
a memory allocation failure. In 1.0.2 and below this could be caused where
the size of a buffer to be allocated is greater than INT_MAX. E.g. this
could be in processing a very long "%s" format string. Memory leaks can also
occur.

These issues will only occur on certain platforms where sizeof(size_t) >
sizeof(int). E.g. many 64 bit systems. The first issue may mask the second
issue dependent on compiler behaviour.

These problems could enable attacks where large amounts of untrusted data
is passed to the BIO_*printf functions. If applications use these functions
in this way then they could be vulnerable. OpenSSL itself uses these
functions when printing out human-readable dumps of ASN.1 data. Therefore
applications that print this data could be vulnerable if the data is from
untrusted sources. OpenSSL command line applications could also be
vulnerable where they print out ASN.1 data, or if untrusted data is passed
as command line arguments.

Libssl is not considered directly vulnerable. Additionally certificates etc
received via remote connections via libssl are also unlikely to be able to
trigger these issues because of message size limits enforced within libssl.

CVE-2016-0799

Issue reported by Guido Vranken.

Reviewed-by: Andy Polyakov <appro@openssl.org>
8 years agoCVE-2016-0798: avoid memory leak in SRP
Emilia Kasper [Wed, 24 Feb 2016 11:59:59 +0000 (12:59 +0100)]
CVE-2016-0798: avoid memory leak in SRP

The SRP user database lookup method SRP_VBASE_get_by_user had confusing
memory management semantics; the returned pointer was sometimes newly
allocated, and sometimes owned by the callee. The calling code has no
way of distinguishing these two cases.

Specifically, SRP servers that configure a secret seed to hide valid
login information are vulnerable to a memory leak: an attacker
connecting with an invalid username can cause a memory leak of around
300 bytes per connection.

Servers that do not configure SRP, or configure SRP but do not configure
a seed are not vulnerable.

In Apache, the seed directive is known as SSLSRPUnknownUserSeed.

To mitigate the memory leak, the seed handling in SRP_VBASE_get_by_user
is now disabled even if the user has configured a seed.

Applications are advised to migrate to SRP_VBASE_get1_by_user. However,
note that OpenSSL makes no strong guarantees about the
indistinguishability of valid and invalid logins. In particular,
computations are currently not carried out in constant time.

Reviewed-by: Rich Salz <rsalz@openssl.org>
8 years agoec/asm/ecp_nistz256-x86_64.pl: get corner case logic right.
Andy Polyakov [Wed, 3 Feb 2016 10:26:11 +0000 (11:26 +0100)]
ec/asm/ecp_nistz256-x86_64.pl: get corner case logic right.

RT#4284

Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit d93753412b455907b4dc09427ccd2382209d9af2)

8 years agoGH714: missing field initialisation
FdaSilvaYY [Fri, 19 Feb 2016 22:28:52 +0000 (23:28 +0100)]
GH714: missing field initialisation

Signed-off-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Kurt Roeckx <kurt@openssl.org>
8 years agoRT4175: Fix PKCS7_verify() regression with Authenticode signatures
David Woodhouse [Tue, 23 Feb 2016 14:00:03 +0000 (14:00 +0000)]
RT4175: Fix PKCS7_verify() regression with Authenticode signatures

This is a partial revert of commit c8491de39 ("GH354: Memory leak fixes"),
which was cherry-picked from commit 55500ea7c in OpenSSL 1.1.

That commit introduced a change in behaviour which is a regression for
software implementing Microsoft Authenticode — which requires a PKCS#7
signature to be validated against explicit external data, even though
it's a non-detached signature with its own embedded data.

The is fixed differently in OpenSSL 1.1 by commit 6b2ebe433 ("Add
PKCS7_NO_DUAL_CONTENT flag"), but that approach isn't viable in the
1.0.2 stable branch, so just comment the offending check back out again.

Signed-off-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
8 years agoDon't strip object files on Cygwin
Corinna Vinschen [Sat, 16 Jan 2016 18:36:15 +0000 (19:36 +0100)]
Don't strip object files on Cygwin

  Building for the Cygwin distro requires to be able to build debuginfo
  files.  This in turn requires to build object files without stripping.
  The stripping is performed by the next step after building which creates
  the debuginfo files.

Signed-off-by: Corinna Vinschen <vinschen@redhat.com>
(cherry picked from commit 42b8f1427a8cdf320e51374a5fbfefc74bec0519)

Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoRecognise Cygwin-x86_64 in config
Richard Levitte [Mon, 22 Feb 2016 10:02:03 +0000 (11:02 +0100)]
Recognise Cygwin-x86_64 in config

In response to RT#4326

Reviewed-by: Rich Salz <rsalz@openssl.org>
8 years agoFix double free in DSA private key parsing.
Dr. Stephen Henson [Thu, 18 Feb 2016 12:47:23 +0000 (12:47 +0000)]
Fix double free in DSA private key parsing.

Fix double free bug when parsing malformed DSA private keys.

Thanks to Adam Langley (Google/BoringSSL) for discovering this bug using
libFuzzer.

CVE-2016-0705

Reviewed-by: Emilia Käsper <emilia@openssl.org>
8 years agoMakefile.shared: limit .dll image base pinning to FIPS builds.
Andy Polyakov [Fri, 19 Feb 2016 12:40:58 +0000 (13:40 +0100)]
Makefile.shared: limit .dll image base pinning to FIPS builds.

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoevp/e_des3.c: address compiler warning.
Andy Polyakov [Thu, 18 Feb 2016 13:08:55 +0000 (14:08 +0100)]
evp/e_des3.c: address compiler warning.

In backporting from master one modification was mistreated.

RT#4210

Reviewed-by: Matt Caswell <matt@openssl.org>
8 years agotypo
Dr. Stephen Henson [Thu, 18 Feb 2016 02:49:14 +0000 (02:49 +0000)]
typo

Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit f6fb7f1856d443185c23f1a5968c08b4269dd37d)

8 years agoSwitch to FIPS implementation for CMAC.
Dr. Stephen Henson [Tue, 16 Feb 2016 22:17:43 +0000 (22:17 +0000)]
Switch to FIPS implementation for CMAC.

Reviewed-by: Tim Hudson <tjh@openssl.org>
8 years agoevp/e_des[3].c: address compiler warnings, fix formatting.
Andy Polyakov [Fri, 12 Feb 2016 13:46:02 +0000 (14:46 +0100)]
evp/e_des[3].c: address compiler warnings, fix formatting.

RT#4210

(1.0.2-specific adaptation of 7687f5255011a5a3ca75e8c5427683d58ae411c0)

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agomodes/ctr128.c: pay attention to ecount_buf alignment in CRYPTO_ctr128_encrypt.
Andy Polyakov [Fri, 12 Feb 2016 13:07:27 +0000 (14:07 +0100)]
modes/ctr128.c: pay attention to ecount_buf alignment in CRYPTO_ctr128_encrypt.

It's never problem if CRYPTO_ctr128_encrypt is called from EVP, because
buffer in question is always aligned within EVP_CIPHER_CTX structure.

RT#4218

Reviewed-by: Richard Levitte <levitte@openssl.org>
(cherry picked from commit 5e4bbeb49fb6522d858703201b5adee9611e7b7b)

8 years agoRT 3854: Update apps/req
Emilia Kasper [Tue, 2 Feb 2016 16:12:45 +0000 (17:12 +0100)]
RT 3854: Update apps/req

Change the default keysize to 2048 bits, and the minimum to 512 bits.

Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit a76265574398944d686d2d0de9bacca162f555ca)

8 years agoperlasm/x86_64-xlate.pl: pass pure constants verbatim.
Andy Polyakov [Wed, 10 Feb 2016 14:11:40 +0000 (15:11 +0100)]
perlasm/x86_64-xlate.pl: pass pure constants verbatim.

RT#3885

Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit fd7dc201d3b9d43972de6a0e659f7ef6421c99cc)

8 years agoutil/mk1mf.pl: use LINK_CMD instead of LINK variable.
Andy Polyakov [Wed, 3 Feb 2016 17:21:00 +0000 (18:21 +0100)]
util/mk1mf.pl: use LINK_CMD instead of LINK variable.

Trouble is that LINK variable assignment in make-file interferes with
LINK environment variable, which can be used to modify Microsoft's
LINK.EXE behaviour.

RT#4289

Reviewed-by: Richard Levitte <levitte@openssl.org>
(cherry picked from commit d44bb1c31ca00f4359090daa15659c0dd1a08f0d)

Resolved conflicts:
util/pl/VC-32.pl

8 years agoms/uplink-x86.pl: make it work.
Andy Polyakov [Tue, 9 Feb 2016 10:53:11 +0000 (11:53 +0100)]
ms/uplink-x86.pl: make it work.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(cherry picked from commit 740b2b9a6cf31b02916a4d18f868e8a95934c083)

8 years agoHandle SSL_shutdown while in init more appropriately #2
Matt Caswell [Tue, 2 Feb 2016 10:05:43 +0000 (10:05 +0000)]
Handle SSL_shutdown while in init more appropriately #2

Previous commit f73c737c7 attempted to "fix" a problem with the way
SSL_shutdown() behaved whilst in mid-handshake. The original behaviour had
SSL_shutdown() return immediately having taken no action if called mid-
handshake with a return value of 1 (meaning everything was shutdown
successfully). In fact the shutdown has not been successful.

Commit f73c737c7 changed that to send a close_notify anyway and then
return. This seems to be causing some problems for some applications so
perhaps a better (much simpler) approach is revert to the previous
behaviour (no attempt at a shutdown), but return -1 (meaning the shutdown
was not successful).

This also fixes a bug where SSL_shutdown always returns 0 when shutdown
*very* early in the handshake (i.e. we are still using SSLv23_method).

Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
8 years agoif no comparison function set make sk_sort no op
Dr. Stephen Henson [Sat, 6 Feb 2016 17:07:58 +0000 (17:07 +0000)]
if no comparison function set make sk_sort no op

Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit 0ca2e82ab1575f9c4eed259c4d66ca9372a45bf5)

8 years agoFix missing ok=0 with locally blacklisted CAs
Viktor Dukhovni [Tue, 2 Feb 2016 09:35:27 +0000 (04:35 -0500)]
Fix missing ok=0 with locally blacklisted CAs

Also in X509_verify_cert() avoid using "i" not only as a loop
counter, but also as a trust outcome and as an error ordinal.

Finally, make sure that all "goto end" jumps return an error, with
"end" renamed to "err" accordingly.

[ The 1.1.0 version of X509_verify_cert() is major rewrite,
  which addresses these issues in a more systemic way. ]

Reviewed-by: Rich Salz <rsalz@openssl.org>
8 years agoRestore xmm7 from the correct address on win64
Kurt Roeckx [Wed, 3 Feb 2016 17:54:02 +0000 (18:54 +0100)]
Restore xmm7 from the correct address on win64

Reviewed-by: Richard Levitte <levitte@openssl.org>
RT: #4288, MR: #1831
(cherry picked from commit df057ea6c8a20e4babc047689507dfafde59ffd6)

8 years agoFix pkeyutl/rsautl empty encrypt-input/decrypt-output handling
Viktor Dukhovni [Tue, 2 Feb 2016 04:37:42 +0000 (23:37 -0500)]
Fix pkeyutl/rsautl empty encrypt-input/decrypt-output handling

Also fix option processing in pkeyutl to allow use of (formerly)
"out-of-order" switches that were needless implementation limitations.

RT2018

Reviewed-by: Rich Salz <rsalz@openssl.org>
8 years agoGH611: s_client help message bug
Rich Salz [Tue, 2 Feb 2016 17:16:10 +0000 (12:16 -0500)]
GH611: s_client help message bug

Reviewed-by: Emilia Käsper <emilia@openssl.org>
8 years agoGH554: Improve pkeyutl doc
Hubert Kario [Mon, 1 Feb 2016 16:14:34 +0000 (11:14 -0500)]
GH554: Improve pkeyutl doc

Signed-off-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Dr. Stephen Henson <steve@openssl.org>
(cherry picked from commit 53619f9f40740ae8b256998574758aa191635db8)

8 years agoRT4129: BUF_new_mem_buf should take const void *
Daniel Kahn Gillmor [Sun, 31 Jan 2016 01:38:08 +0000 (20:38 -0500)]
RT4129: BUF_new_mem_buf should take const void *

Signed-off-by: Rich Salz <rsalz@akamai.com>
Reviewed-by: Dr. Stephen Henson <steve@openssl.org>
(cherry picked from commit 8ab31975bacb9c907261088937d3aa4102e3af84)

8 years agoGH102: Add volatile to CRYPTO_memcmp
Rich Salz [Fri, 29 Jan 2016 17:09:05 +0000 (12:09 -0500)]
GH102: Add volatile to CRYPTO_memcmp

Can't hurt and seems to prevent problems from some over-aggressive
(LTO?) compilers.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(cherry picked from commit 98ab57644f44d2d83595c2d0f69138a284d6096b)

8 years agoBackport SHA2 support for capi engine
Dr. Stephen Henson [Thu, 21 Jan 2016 14:29:16 +0000 (14:29 +0000)]
Backport SHA2 support for capi engine

Reviewed-by: Andy Polyakov <appro@openssl.org>
8 years agoAdd have_precompute_mult tests
Matt Caswell [Wed, 4 Nov 2015 22:54:29 +0000 (22:54 +0000)]
Add have_precompute_mult tests

Add tests for have_precompute_mult for the optimised curves (nistp224,
nistp256 and nistp521) if present

Reviewed-by: Richard Levitte <levitte@openssl.org>
(cherry picked from commit 8ce4e7e605577cb5818de068e2c6da60901cddba)

8 years agoFix bug in nistp224/256/521 where have_precompute_mult always returns 0
Matt Caswell [Wed, 4 Nov 2015 17:30:22 +0000 (17:30 +0000)]
Fix bug in nistp224/256/521 where have_precompute_mult always returns 0

During precomputation if the group given is well known then we memcpy a
well known precomputation. However we go the wrong label in the code and
don't store the data properly. Consequently if we call have_precompute_mult
the data isn't there and we return 0.

RT#3600

Reviewed-by: Richard Levitte <levitte@openssl.org>
(cherry picked from commit 615614c8862fb89dcf1551a4e113be0789dddf5f)

8 years agoAdd missing return value checks
Matt Caswell [Fri, 29 Jan 2016 09:40:03 +0000 (09:40 +0000)]
Add missing return value checks

The function DH_check_pub_key() was missing some return value checks in
some calls to BN functions.

RT#4278

Reviewed-by: Andy Polyakov <appro@openssl.org>
(cherry picked from commit f5a12207eccfd814bde68b880a96910dfa25f164)

8 years agoCorrect value of DH_CHECK_PUBKEY_INVALID
Matt Caswell [Fri, 29 Jan 2016 09:38:06 +0000 (09:38 +0000)]
Correct value of DH_CHECK_PUBKEY_INVALID

A new return value for DH_check_pub_key was recently added:
DH_CHECK_PUBKEY_INVALID. As this is a flag which can be ORed with other
return values it should have been set to the value 4 not 3.

RT#4278

Reviewed-by: Andy Polyakov <appro@openssl.org>
(cherry picked from commit cb389fe80462e20daba30835a9e86354451bd14f)

8 years agoFix CHANGES entry about DSA_generate_parameters_ex
Kurt Roeckx [Wed, 27 Jan 2016 19:31:57 +0000 (20:31 +0100)]
Fix CHANGES entry about DSA_generate_parameters_ex

Reviewed-by: Viktor Dukhovni <openssl-users@dukhovni.org>
8 years agoCorrect number of arguments in BIO_get_conn_int_port macro
Richard Levitte [Thu, 28 Jan 2016 16:55:11 +0000 (17:55 +0100)]
Correct number of arguments in BIO_get_conn_int_port macro

Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit 41a28cb2944a4e1c9d13889757a3bd9f72abeca1)

8 years agoPrepare for 1.0.2g-dev
Matt Caswell [Thu, 28 Jan 2016 13:58:24 +0000 (13:58 +0000)]
Prepare for 1.0.2g-dev

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoPrepare for 1.0.2f release OpenSSL_1_0_2f
Matt Caswell [Thu, 28 Jan 2016 13:57:22 +0000 (13:57 +0000)]
Prepare for 1.0.2f release

Reviewed-by: Richard Levitte <levitte@openssl.org>
8 years agoUpdate CHANGES and NEWS for release
Matt Caswell [Wed, 27 Jan 2016 13:52:29 +0000 (13:52 +0000)]
Update CHANGES and NEWS for release

Add details about the latest issues into CHANGES and NEWS ready for the
next release.

Reviewed-by: Rich Salz <rsalz@openssl.org>
8 years agoAdd a test for small subgroup attacks on DH/DHE
Matt Caswell [Wed, 20 Jan 2016 11:56:28 +0000 (11:56 +0000)]
Add a test for small subgroup attacks on DH/DHE

Following on from the previous commit, add a test to ensure that
DH_compute_key correctly fails if passed a bad y such that:

y^q (mod p) != 1

Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
8 years agoAlways generate DH keys for ephemeral DH cipher suites
Matt Caswell [Thu, 17 Dec 2015 02:57:20 +0000 (02:57 +0000)]
Always generate DH keys for ephemeral DH cipher suites

Modified version of the commit ffaef3f15 in the master branch by Stephen
Henson. This makes the SSL_OP_SINGLE_DH_USE option a no-op and always
generates a new DH key for every handshake regardless.

CVE-2016-0701 (fix part 2 or 2)

Issue reported by Antonio Sanso

Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
8 years agoPrevent small subgroup attacks on DH/DHE
Matt Caswell [Mon, 18 Jan 2016 11:31:58 +0000 (11:31 +0000)]
Prevent small subgroup attacks on DH/DHE

Historically OpenSSL only ever generated DH parameters based on "safe"
primes. More recently (in version 1.0.2) support was provided for
generating X9.42 style parameter files such as those required for RFC
5114 support. The primes used in such files may not be "safe". Where an
application is using DH configured with parameters based on primes that
are not "safe" then an attacker could use this fact to find a peer's
private DH exponent. This attack requires that the attacker complete
multiple handshakes in which the peer uses the same DH exponent.

A simple mitigation is to ensure that y^q (mod p) == 1

CVE-2016-0701 (fix part 1 of 2)

Issue reported by Antonio Sanso.

Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
8 years agoBetter SSLv2 cipher-suite enforcement
Viktor Dukhovni [Thu, 31 Dec 2015 03:44:51 +0000 (22:44 -0500)]
Better SSLv2 cipher-suite enforcement

Based on patch by: Nimrod Aviram <nimrod.aviram@gmail.com>

CVE-2015-3197

Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>