5 years agoConfigure: allow to enable afalgeng if target does not start with Linux
Sebastian Andrzej Siewior [Thu, 25 Jan 2018 21:50:55 +0000 (22:50 +0100)]
Configure: allow to enable afalgeng if target does not start with Linux

The Debian build system uses a `debian' target which sets CFLAGS and
then we have for instance debian-amd64 which inherits from
linux-x86_64 and debian [0]. So far so good.

Unless there are different suggestions how to do this, I would keep it.
However since the target name does not start with `linux', the build
system does not enable the afalg engine. So in order to get enabled, I
added a
`enable           => [ "afalgeng" ],'
to the generic linux config which sets it explicit (as suggested by
Richard Levitte). Having this set, we can check for it instead matching
the target name.

Signed-off-by: Sebastian Andrzej Siewior <>
Reviewed-by: Matt Caswell <>
Reviewed-by: Richard Levitte <>
(Merged from

5 years agoOnly update the server session cache when the session is ready
Matt Caswell [Wed, 14 Mar 2018 17:51:18 +0000 (17:51 +0000)]
Only update the server session cache when the session is ready

In TLSv1.3 the session is not ready until after the end of the handshake
when we are constructing the NewSessionTicket.

Reviewed-by: Rich Salz <>
(Merged from

5 years agoMissings OIDs for XTS added.
Pauli [Thu, 15 Mar 2018 01:01:48 +0000 (11:01 +1000)]
Missings OIDs for XTS added.

Added two missing OIDs for AES-{128,256}-XTS.

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoFix no-ec
Matt Caswell [Wed, 14 Mar 2018 14:04:18 +0000 (14:04 +0000)]
Fix no-ec

Reviewed-by: Rich Salz <>
(Merged from

5 years agoFix test_out_option
Patrick Steuer [Mon, 12 Mar 2018 12:53:21 +0000 (12:53 +0000)]
Fix test_out_option

Random path generation code in test/recipes/15-test_out_option.t
does not work: The code sets rand_path to "/test.pem". I.e. the
test will fail as expected for unprivileged user but will pass
for root user.

Signed-off-by: Patrick Steuer <>
Reviewed-by: Bernd Edlinger <>
Reviewed-by: Richard Levitte <>
(Merged from

5 years agoutil/ sort the dependency files
Richard Levitte [Wed, 14 Mar 2018 11:39:45 +0000 (12:39 +0100)]
util/ sort the dependency files

Reviewed-by: Rich Salz <>
(Merged from

5 years agoFix an error number clash
Matt Caswell [Wed, 14 Mar 2018 11:27:45 +0000 (11:27 +0000)]
Fix an error number clash

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoPut the default set of TLSv1.3 ciphersuites in a header file
Matt Caswell [Wed, 14 Mar 2018 10:43:53 +0000 (10:43 +0000)]
Put the default set of TLSv1.3 ciphersuites in a header file

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoFix clienthellotest for new TLSv1.3 ciphersuite configuration
Matt Caswell [Mon, 12 Mar 2018 10:48:32 +0000 (10:48 +0000)]
Fix clienthellotest for new TLSv1.3 ciphersuite configuration

A place in clienthellotest was missed in converting to the new mechanism
for configuration of TLSv1.3 ciphersuites.

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoUpdate CHANGES with details of TLSv1.3 ciphersuite configuration
Matt Caswell [Wed, 21 Feb 2018 17:47:12 +0000 (17:47 +0000)]
Update CHANGES with details of TLSv1.3 ciphersuite configuration

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoAdd documentation for TLSv1.3 ciphersuite configuration
Matt Caswell [Wed, 21 Feb 2018 17:23:11 +0000 (17:23 +0000)]
Add documentation for TLSv1.3 ciphersuite configuration

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoUpdate s_time to be allow configuration of TLSv1.3 ciphersuites
Matt Caswell [Mon, 19 Feb 2018 12:07:18 +0000 (12:07 +0000)]
Update s_time to be allow configuration of TLSv1.3 ciphersuites

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoSplit configuration of TLSv1.3 ciphers from older ciphers
Matt Caswell [Fri, 16 Feb 2018 11:26:02 +0000 (11:26 +0000)]
Split configuration of TLSv1.3 ciphers from older ciphers

With the current mechanism, old cipher strings that used to work in 1.1.0,
may inadvertently disable all TLSv1.3 ciphersuites causing connections to
fail. This is confusing for users.

In reality TLSv1.3 are quite different to older ciphers. They are much
simpler and there are only a small number of them so, arguably, they don't
need the same level of control that the older ciphers have.

This change splits the configuration of TLSv1.3 ciphers from older ones.
By default the TLSv1.3 ciphers are on, so you cannot inadvertently disable
them through your existing config.

Fixes #5359

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoUpdate version numbers for TLSv1.3 draft-26
Matt Caswell [Tue, 13 Mar 2018 13:13:33 +0000 (13:13 +0000)]
Update version numbers for TLSv1.3 draft-26

Reviewed-by: Ben Kaduk <>
(Merged from

5 years agoOnly allow supported_versions in a TLSv1.3 ServerHello
Matt Caswell [Tue, 13 Mar 2018 10:36:03 +0000 (10:36 +0000)]
Only allow supported_versions in a TLSv1.3 ServerHello

As per the latest text in TLSv1.3 draft-26

Reviewed-by: Ben Kaduk <>
(Merged from

5 years agoUse the TLSv1.3 record header as AAD
Matt Caswell [Mon, 26 Feb 2018 12:26:14 +0000 (12:26 +0000)]
Use the TLSv1.3 record header as AAD

As of TLSv1.3 draft-25 the record header data must be used as AAD

Reviewed-by: Ben Kaduk <>
(Merged from

5 years agoRemove '-Wextra' as default user flags for the Linux clang targets
Richard Levitte [Wed, 14 Mar 2018 08:41:35 +0000 (09:41 +0100)]
Remove '-Wextra' as default user flags for the Linux clang targets

We have '--strict-warnings' for this kind of stuff...  also, user
flags are added last, so this overrides any warning supression
--strict-warnings may put in place (for good reasons).

Fixes #5609

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoWindows build file: make sure to quote
Richard Levitte [Tue, 13 Mar 2018 21:07:02 +0000 (22:07 +0100)]
Windows build file: make sure to quote

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoRemove the temporary file in case it wasn't renamed
Richard Levitte [Tue, 13 Mar 2018 20:55:52 +0000 (21:55 +0100)]
Remove the temporary file in case it wasn't renamed

...  also, name it with "-$$" added instead of ".$$"

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoRemove debugging prints from util/
Richard Levitte [Tue, 13 Mar 2018 18:38:04 +0000 (19:38 +0100)]
Remove debugging prints from util/

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoConfigurations/*.conf: overhaul Android targets.
Andy Polyakov [Sun, 11 Mar 2018 18:08:56 +0000 (19:08 +0100)]
Configurations/*.conf: overhaul Android targets.

Move Android targets to separate file, automate sysroot setup and
add support for NDK 16.

Reviewed-by: Rich Salz <>
(Merged from

5 years agoec/curve25519.c: resolve regression with Android NDK's arm64 gcc.
Andy Polyakov [Mon, 12 Mar 2018 09:46:23 +0000 (10:46 +0100)]
ec/curve25519.c: resolve regression with Android NDK's arm64 gcc.

Unlike "upstream", Android NDK's arm64 gcc [but not clang] performs
64x64=128-bit multiplications with library calls, which appears to
have devastating impact on performance. [The condition is reduced to
__ANDROID__ [&& !__clang__], because x86_64 has corresponding
assembly module.]

Reviewed-by: Rich Salz <>
(Merged from

5 years agostore/loader_file.c: rename variables causing conflicts with Android NDK.
Andy Polyakov [Sun, 11 Mar 2018 17:47:44 +0000 (18:47 +0100)]
store/loader_file.c: rename variables causing conflicts with Android NDK.

Reviewed-by: Rich Salz <>
(Merged from

5 years agoRefactor the 'depend' target
Richard Levitte [Tue, 13 Mar 2018 16:56:20 +0000 (17:56 +0100)]
Refactor the 'depend' target

With the help of the perl script util/, which takes all
its information directly from, the dependency adding
procedure can be streamlined for all support platforms.

Reviewed-by: Rich Salz <>
(Merged from

5 years agoupdate SRP copyright notice
Tim Hudson [Tue, 13 Mar 2018 08:33:44 +0000 (18:33 +1000)]
update SRP copyright notice

As per discussion with Peter Sylvester

Reviewed-by: Matt Caswell <>
(Merged from

5 years agoopenssl rehash: no more need to massage the files on VMS
Richard Levitte [Mon, 12 Mar 2018 22:02:18 +0000 (23:02 +0100)]
openssl rehash: no more need to massage the files on VMS

OPENSSL_DIR_read() now returns unique file names on VMS, no generation
number.  We therefore do not need to handle that case in apps/rehash.c
any more.

Reviewed-by: Rich Salz <>
(Merged from

5 years agoAdjust LPdir_unix.c on VMS for OpenSSL expectations
Richard Levitte [Sun, 11 Mar 2018 22:48:04 +0000 (23:48 +0100)]
Adjust LPdir_unix.c on VMS for OpenSSL expectations

When OPENSSL_DIR_read implemented by LPdir_unix.c gets a Unixy path,
it will return file names like you'd expect them on Unix.

However, if given a path with VMS syntax, such as "[.foo]", it returns
file names with generation numbers, such as "bar.txt;1", which makes
sense for VMS expectations, but can be surprising for OpenSSL.

Our solution is to simply shave off the generation number if
OPENSSL_DIR_read() expects there should be one, and make sure not to
return the same file name twice.  Note that VMS filesystems are case
insensitive, so the check for duplicate file names are done without
regard to character case.

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoCONF inclusion test: Add VMS specific tests
Richard Levitte [Sun, 11 Mar 2018 22:47:40 +0000 (23:47 +0100)]
CONF inclusion test: Add VMS specific tests

We want to see that VMS syntax paths are treated correctly.

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoCONF: On VMS, treat VMS syntax inclusion paths correctly
Richard Levitte [Sun, 11 Mar 2018 09:14:11 +0000 (10:14 +0100)]
CONF: On VMS, treat VMS syntax inclusion paths correctly

non-VMS syntax inclusion paths get the same treatment as on Unix.

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoFree the correct type in OBJ_add_object()
Matt Caswell [Mon, 12 Mar 2018 15:24:29 +0000 (15:24 +0000)]
Free the correct type in OBJ_add_object()

We should be using ASN1_OBJECT_free() not OPENSSL_free().

Fixes #5568

Reviewed-by: Rich Salz <>
(Merged from

5 years agoFix enable-ssl-trace
Matt Caswell [Mon, 12 Mar 2018 16:08:31 +0000 (16:08 +0000)]
Fix enable-ssl-trace

Reviewed-by: Rich Salz <>
(Merged from

5 years agoIntroduce SSL_CTX_set_stateless_cookie_{generate,verify}_cb
Benjamin Saunders [Mon, 26 Feb 2018 02:39:11 +0000 (18:39 -0800)]
Introduce SSL_CTX_set_stateless_cookie_{generate,verify}_cb

These functions are similar to SSL_CTX_set_cookie_{generate,verify}_cb,
but used for the application-controlled portion of TLS1.3 stateless
handshake cookies rather than entire DTLSv1 cookies.

Reviewed-by: Ben Kaduk <>
Reviewed-by: Matt Caswell <>
(Merged from

5 years agoImprove error handling in pk7_doit
Matt Caswell [Mon, 12 Mar 2018 13:56:34 +0000 (13:56 +0000)]
Improve error handling in pk7_doit

If a mem allocation failed we would ignore it. This commit fixes it to
always check.

Reviewed-by: Rich Salz <>
(Merged from

5 years agoAdd OIDs for DSTU-4145
gmile [Mon, 12 Mar 2018 00:52:13 +0000 (20:52 -0400)]
Add OIDs for DSTU-4145

Original source:

Full list of OIDs is available on related enactment page

CLA: trivial

Reviewed-by: Tim Hudson <>
Reviewed-by: Rich Salz <>
(Merged from

5 years agoDon't negotiate TLSv1.3 with the ossl_shim
Matt Caswell [Mon, 12 Mar 2018 11:42:00 +0000 (11:42 +0000)]
Don't negotiate TLSv1.3 with the ossl_shim

The ossl_shim doesn't know about TLSv1.3 so we should disable that
protocol version for all tests for now.

This fixes the current Travis failures.

[extended tests]

Reviewed-by: Rich Salz <>
(Merged from

5 years agoSession Ticket app data
Todd Short [Wed, 15 Mar 2017 17:25:55 +0000 (13:25 -0400)]
Session Ticket app data

Adds application data into the encrypted session ticket

Reviewed-by: Paul Dale <>
Reviewed-by: Matt Caswell <>
(Merged from

5 years agomem_sec.c: portability fixup.
Andy Polyakov [Sat, 10 Mar 2018 18:38:28 +0000 (19:38 +0100)]
mem_sec.c: portability fixup.

Reviewed-by: Rich Salz <>
5 years agoFix propotype to include the const qualifier
Kurt Roeckx [Sat, 10 Mar 2018 15:32:55 +0000 (16:32 +0100)]
Fix propotype to include the const qualifier

Reviewed-by: Andy Polyakov <>
GH: #5582

5 years agoFixed a handful of typos
Alex Gaynor [Sat, 10 Mar 2018 18:13:23 +0000 (13:13 -0500)]
Fixed a handful of typos

Reviewed-by: Andy Polyakov <>
Reviewed-by: Rich Salz <>
(Merged from

5 years agoConfigurations/10-main.conf: VC-<target> cleanups.
Andy Polyakov [Fri, 9 Mar 2018 16:28:06 +0000 (17:28 +0100)]
Configurations/10-main.conf: VC-<target> cleanups.

Reviewed-by: Richard Levitte <>
(Merged from

5 years agoConfigurations/unix-Makefile.tmpl: overhaul assembler make rules.
Andy Polyakov [Fri, 9 Mar 2018 12:57:38 +0000 (13:57 +0100)]
Configurations/unix-Makefile.tmpl: overhaul assembler make rules.

So far assembly modules were built as .pl->.S->.s followed by .s->.o.
This posed a problem in build_all_generated rule if it was executed
on another computer. So we change rule sequence to .pl->.S and then

Reviewed-by: Richard Levitte <>
(Merged from

5 years agoConfigurations/descrip.mms.tmpl: Fix small errors
Richard Levitte [Fri, 9 Mar 2018 21:58:32 +0000 (22:58 +0100)]
Configurations/descrip.mms.tmpl: Fix small errors

A missing parenthesis here, a missing comma there...

Reviewed-by: Rich Salz <>
(Merged from

5 years agoConfigure: catch the build tree
Richard Levitte [Thu, 8 Mar 2018 11:01:28 +0000 (12:01 +0100)]
Configure: catch the build tree

There are things depending on  However, it's perfectly
possible that there is one in the source directory from a previous
build, and that might disrupt an out of source build.  To avoid this
conflict, make sure never to use the source tree in that
case, i.e. make the hard assumption that it's a generated file in the
build tree, which it is.

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoConfigure: don't mangle the directory again when checking DEPEND inclusion
Richard Levitte [Wed, 7 Mar 2018 14:58:04 +0000 (15:58 +0100)]
Configure: don't mangle the directory again when checking DEPEND inclusion

When generating the correct inclusion directory for DEPEND, we mangled
it to be relative to the build or the source directory.  However, the
value we handle already come with a correct directory, so we only need
to use it as is.

Fixes #5543

Reviewed-by: Tim Hudson <>
(Merged from

5 years agoRAND_DRBG: add a function for setting the reseeding defaults
Dr. Matthias St. Pierre [Fri, 9 Mar 2018 17:24:23 +0000 (18:24 +0100)]
RAND_DRBG: add a function for setting the reseeding defaults

The introduction of thread local public and private DRBG instances (#5547)
makes it very cumbersome to change the reseeding (time) intervals for
those instances. This commit provides a function to set the default
values for all subsequently created DRBG instances.

 int RAND_DRBG_set_reseed_defaults(
                                   unsigned int master_reseed_interval,
                                   unsigned int slave_reseed_interval,
                                   time_t master_reseed_time_interval,
                                   time_t slave_reseed_time_interval

The function is intended only to be used during application initialization,
before any threads are created and before any random bytes are generated.

Reviewed-by: Rich Salz <>
(Merged from

5 years agoAttempt to fix boringssl tests
Ben Kaduk [Fri, 9 Mar 2018 18:58:39 +0000 (12:58 -0600)]
Attempt to fix boringssl tests

Commit abe256e7951e6d57f8f6b4364ea696eb4ead3852 changed the config target
element from 'cxx' to 'CXX'; catch up accordingly.
Also use a space to offset the template boundary, per convention.

[extended tests]

Reviewed-by: Richard Levitte <>
(Merged from

5 years agoDocument more X509_STORE functions
Benjamin Kaduk [Mon, 17 Jul 2017 17:44:03 +0000 (12:44 -0500)]
Document more X509_STORE functions

X509_STORE_set_verify_cb_func.pod has documentation for various callbacks
and function pointers that can be set and retrieved, but neither it nor
X509_STORE_new has much documentation for the actual purpose and usage
of X509_STORE objects.  Remedy this disparity with new documentation
for adding certificates and CRLs, expected usage, and for modifying
the default verifification behavior.

Reviewed-by: Richard Levitte <>
Reviewed-by: Tim Hudson <>
Reviewed-by: Viktor Dukhovni <>
(Merged from

5 years agoFix doc-nits
Ben Kaduk [Fri, 9 Mar 2018 16:33:19 +0000 (10:33 -0600)]
Fix doc-nits

Remove a space from a whitespace-only line.

Reviewed-by: Rich Salz <>
Reviewed-by: Richard Levitte <>
(Merged from

5 years agoReuse extension_is_relevant() in should_add_extension()
Benjamin Kaduk [Tue, 14 Mar 2017 19:41:08 +0000 (14:41 -0500)]
Reuse extension_is_relevant() in should_add_extension()

At the core of things is the concept that each extension is only
defined in certain context(s) -- the ClientHello, EncryptedExtensions,
etc., and sometimes only for a specific protocol or protocol range;
we want to enforce that we only parse or generate extensions in the
context(s) for which they are defined.  There is some subtlety here,
in that the protocol version in use is not known when generating the
ClientHello (but it is known when the ClientHello extensions are
being parsed!), so the SSL_IS_TLS13() macro must be used with caution.
Nonetheless, by making assertions about whether we are acting in a
server role and whether the current context is (not) a ClientHello,
we can consolidate almost all of the logic for determining whether
an extension is permitted in a given protocol message, whether we
are generating or parsing that message.

The only logic that remains separate relates to generating the ClientHello,
as it depends on an external factor (the maximum permitted TLS version) that
is not defined in the parsing context.

Reviewed-by: Matt Caswell <>
(Merged from

5 years agoFix type error in PEM processing
Benjamin Kaduk [Tue, 30 Jan 2018 18:55:44 +0000 (12:55 -0600)]
Fix type error in PEM processing

The get_name() helper was using a variable of type size_t to hold the
result of BIO_gets(), but BIO_gets() returns int and makes use of negative
values to indicate error conditions.

Change the type of the local variable to match, and propagate that
through to other places in the file to avoid -Wsign-compare issues.

Reviewed-by: Rich Salz <>
(Merged from

5 years agoAdd SSL/SSL_CTX_use_cert_and_key()
Todd Short [Mon, 28 Dec 2015 14:13:20 +0000 (09:13 -0500)]
Add SSL/SSL_CTX_use_cert_and_key()

Add functions that will do the work of assigning certificate, privatekey
and chain certs to an SSL or SSL_CTX. If no privatekey is given, use the
publickey. This will permit the keys to pass validation for both ECDSA
and RSA. If a private key has already been set for the certificate, it
is discarded. A real private key can be set later.

This is an all-or-nothing setting of these parameters. Unlike the
SSL/SSL_CTX_use_certificate() and SSL/SSL_CTX_use_PrivateKey() functions,
the existing cert or privatekey is not modified (i.e. parameters copied).
This permits the existing cert/privatekey to be replaced.

It replaces the sequence of:
* SSL_use_certificate()
* SSL_use_privatekey()
* SSL_set1_chain()
And may actually be faster, as multiple checks are consolidated.

The private key can be NULL, if so an ENGINE module needs to contain the
actual private key that is to be used.

Note that ECDH (using the certificate's ECDSA key) ciphers do not work
without the private key being present, based on how the private key is
used in ECDH. ECDH does not offer PFS; ECDHE ciphers should be used instead.

Reviewed-by: Tim Hudson <>
Reviewed-by: Ben Kaduk <>
(Merged from

5 years agoConfigurations/50-win-onecore.conf: add Windows 10 OneCore targets.
Andy Polyakov [Sat, 3 Mar 2018 22:09:52 +0000 (23:09 +0100)]
Configurations/50-win-onecore.conf: add Windows 10 OneCore targets.

This includes even ARM targets.

Reviewed-by: Richard Levitte <>
(Merged from

5 years agoConfigurations/windows-makefile.tmpl: simplify install-path "flavour"-ing.
Andy Polyakov [Sat, 3 Mar 2018 21:26:45 +0000 (22:26 +0100)]
Configurations/windows-makefile.tmpl: simplify install-path "flavour"-ing.

$target{build_scheme} consists of fixed number of elements with 3rd
element denoting VC install-path "flavour", i.e. where to install things.
Instead of looking at 3rd, let's look at last. This allows to override
flavour from template in a simple way.

Configurations/10-main.conf: define generic "flavour" in VC-common
template. Since VC-W32 was the only recognized "flavour", remove
"flavour" definitions from all targets/templates, but VC-WIN32. And
rename VC-W32 to VC-WOW.

Reviewed-by: Richard Levitte <>
(Merged from

5 years agoConfigurations/10-main.conf: amend out-dated comments.
Andy Polyakov [Thu, 8 Mar 2018 20:11:12 +0000 (21:11 +0100)]
Configurations/10-main.conf: amend out-dated comments.

Reviewed-by: Richard Levitte <>
(Merged from

5 years agoRestore the display of options with 'openssl version -a'
Richard Levitte [Fri, 9 Mar 2018 13:28:51 +0000 (14:28 +0100)]
Restore the display of options with 'openssl version -a'

Reviewed-by: Rich Salz <>
(Merged from

5 years agoDisplay the library building flags
Richard Levitte [Fri, 9 Mar 2018 11:39:01 +0000 (12:39 +0100)]
Display the library building flags

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoConfigure et al: Move the definition of library only flags
Richard Levitte [Tue, 6 Mar 2018 20:09:05 +0000 (21:09 +0100)]
Configure et al: Move the definition of library only flags

We're currently using the attributes 'defines', 'cppflags', 'cflags'
etc quite liberally, with no regard for where that ends up.  Quite a
few of those flags are actually only relevant for the libraries
(mostly libcrypto), so it's safe to say that those could be applied to
the libraries only.

So, we move some of those flags to 'lib_defines', 'lib_cppflags',
'lib_cflags', etc.

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agospeed: add ecdhx448 to ecdh choices
Steven Noonan [Fri, 9 Mar 2018 12:15:20 +0000 (07:15 -0500)]
speed: add ecdhx448 to ecdh choices

CLA: trivial
Signed-off-by: Steven Noonan <>
Reviewed-by: Matt Caswell <>
Reviewed-by: Rich Salz <>
(Merged from

5 years agoFixed a spelling mistake in ASN1_TIME_set.pod
Alex Gaynor [Fri, 9 Mar 2018 12:11:13 +0000 (07:11 -0500)]
Fixed a spelling mistake in ASN1_TIME_set.pod

Reviewed-by: Tim Hudson <>
Reviewed-by: Rich Salz <>
(Merged from

5 years agoGive more information in the SSL_stateless return code
Matt Caswell [Thu, 8 Mar 2018 17:44:12 +0000 (17:44 +0000)]
Give more information in the SSL_stateless return code

Allow users to distinguish between an error occurring and an HRR being

Fixes #5549

Reviewed-by: Rich Salz <>
(Merged from

5 years agoTest the new PSK behaviour
Matt Caswell [Thu, 8 Mar 2018 13:45:22 +0000 (13:45 +0000)]
Test the new PSK behaviour

Reviewed-by: Rich Salz <>
(Merged from

5 years agoUpdate documentation for the new PSK behaviour
Matt Caswell [Thu, 8 Mar 2018 08:20:23 +0000 (08:20 +0000)]
Update documentation for the new PSK behaviour

Reviewed-by: Rich Salz <>
(Merged from

5 years agoFallback on old style PSK callbacks if the new style ones aren't present
Matt Caswell [Tue, 6 Mar 2018 16:41:51 +0000 (16:41 +0000)]
Fallback on old style PSK callbacks if the new style ones aren't present

Reviewed-by: Rich Salz <>
(Merged from

5 years agoTolerate TLSv1.3 PSKs that are a different size to the hash size
Matt Caswell [Tue, 6 Mar 2018 14:12:10 +0000 (14:12 +0000)]
Tolerate TLSv1.3 PSKs that are a different size to the hash size

We also default to SHA256 as per the spec if we do not have an explicit
digest defined.

Reviewed-by: Rich Salz <>
(Merged from

5 years agoMake OCSP "multi" compatible with "no-sock" builds.
Viktor Dukhovni [Thu, 8 Mar 2018 15:51:46 +0000 (10:51 -0500)]
Make OCSP "multi" compatible with "no-sock" builds.

Reviewed-by: Matt Caswell <>
5 years agoRemove useless -D_ENDIAN from MPE/iX-gcc config
Richard Levitte [Thu, 8 Mar 2018 21:03:01 +0000 (22:03 +0100)]
Remove useless -D_ENDIAN from MPE/iX-gcc config

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoReduce travis-ci log output
Michael Richardson [Tue, 6 Mar 2018 19:18:43 +0000 (14:18 -0500)]
Reduce travis-ci log output

Travis-ci log output is huge and overflows internal travis-ci view,
which makes it hard to find errors.
Redirect some output to a file and dump it only if it fails.
Remove "v" option from tar that builds and extracts the srcdist.
While running the tests manually, some non-POSIX (bashisms) with ==
vs = came to light.

Reviewed-by: Andy Polyakov <>
Reviewed-by: Richard Levitte <>
(Merged from

5 years agoConfigure: correct the check of env vars vs command line flags
Richard Levitte [Thu, 8 Mar 2018 17:49:37 +0000 (18:49 +0100)]
Configure: correct the check of env vars vs command line flags

The check to detect if env / make variables conflicted with compiler /
linker flags on the configure command line went a little too far, and
would stop the configuration process if any such command line flag was
given, with no regard for the env / make variables at all.

This change refines the check so the stop only gets triggered if any
such flags were given AND any of the corresponding variables are set.

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoMake a few more asm modules conform: last argument is output file
Richard Levitte [Sat, 10 Feb 2018 07:47:51 +0000 (08:47 +0100)]
Make a few more asm modules conform: last argument is output file

Fixes #5310

Reviewed-by: Rich Salz <>
(Merged from

5 years agoConfigure: disallow the mixture of compiling flags and env / make variables
Richard Levitte [Thu, 8 Mar 2018 13:16:18 +0000 (14:16 +0100)]
Configure: disallow the mixture of compiling flags and env / make variables

Note that this might give surprising results if someone forgets an environment
variable that has been set previously.

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoConfigurations/README: update documentation on flags
Richard Levitte [Wed, 7 Mar 2018 23:17:29 +0000 (00:17 +0100)]
Configurations/README: update documentation on flags

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoConfigure et al: get rid of last traces of plib_lflags / PLIB_LDFLAGS
Richard Levitte [Wed, 7 Mar 2018 23:16:47 +0000 (00:16 +0100)]
Configure et al: get rid of last traces of plib_lflags / PLIB_LDFLAGS

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoconfig: Pass diverse flags to Configure via the environment
Richard Levitte [Wed, 7 Mar 2018 21:44:59 +0000 (22:44 +0100)]
config: Pass diverse flags to Configure via the environment

Passing flags "discovered" by 'config' on the command line to
'Configure' mixes them up with flags given by the user.  That is
contrary to their intention, so they need to be passed in a different

Enter the environment variables __CNF_CPPDEFINES, __CNF_CPPINCLUDES,
__CNF_LDLIBS, initialised by 'config', and then used by Configure to
initialise these %config values: defines, includes, cppflags, cflags,
cxxflags, lflags, and ex_libs.

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoConfigurations/10-main.conf: Duplicate cflags to cxxflags
Richard Levitte [Wed, 7 Mar 2018 12:29:41 +0000 (13:29 +0100)]
Configurations/10-main.conf: Duplicate cflags to cxxflags

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoAdd space to asoutflag value where appropriate
Richard Levitte [Wed, 7 Mar 2018 12:29:04 +0000 (13:29 +0100)]
Add space to asoutflag value where appropriate

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoMake "make variables" config attributes for overridable flags
Richard Levitte [Tue, 6 Mar 2018 19:35:30 +0000 (20:35 +0100)]
Make "make variables" config attributes for overridable flags

With the support of "make variables" comes the possibility for the
user to override them.  However, we need to make a difference between
defaults that we use (and that should be overridable by the user) and
flags that are crucial for building OpenSSL (should not be

Typically, overridable flags are those setting optimization levels,
warnings levels, that kind of thing, while non-overridable flags are,
for example, macros that indicate aspects of how the config target
should be treated, such as L_ENDIAN and B_ENDIAN.

We do that differentiation by allowing upper case attributes in the
config targets, named exactly like the "make variables" we support,
and reserving the lower case attributes for non-overridable project

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoConfigure et al: rename all dso_* to module_* in
Richard Levitte [Wed, 7 Mar 2018 19:31:15 +0000 (20:31 +0100)]
Configure et al: rename all dso_* to module_* in

Because there are already attributes with the dso_ prefix that are
used instead of the corresponding lib_ attributes rather than in
addition to them, it gets confusing to have similar or exactly the
same attributes working with different semantics on Unix.

So we rename those by changing the prefix dso_ to module_, and having
those work just like the shared_ attributes, but for DSOs.

Reviewed-by: Andy Polyakov <>
(Merged from

5 years agoDuplicate entries ssl_handshake_tbl trace entries...
FdaSilvaYY [Thu, 8 Mar 2018 15:32:34 +0000 (10:32 -0500)]
Duplicate entries ssl_handshake_tbl trace entries...

... and add some missing known values.
Sort ssl/tls extension array list

Reviewed-by: Matt Caswell <>
Reviewed-by: Rich Salz <>
(Merged from

5 years agoFix issues in ia32 RDRAND asm leading to reduced entropy
Bryan Donlan [Wed, 7 Mar 2018 21:01:06 +0000 (16:01 -0500)]
Fix issues in ia32 RDRAND asm leading to reduced entropy

This patch fixes two issues in the ia32 RDRAND assembly code that result in a
(possibly significant) loss of entropy.

The first, less significant, issue is that, by returning success as 0 from
OPENSSL_ia32_rdrand() and OPENSSL_ia32_rdseed(), a subtle bias was introduced.
Specifically, because the assembly routine copied the remaining number of
retries over the result when RDRAND/RDSEED returned 'successful but zero', a
bias towards values 1-8 (primarily 8) was introduced.

The second, more worrying issue was that, due to a mixup in registers, when a
buffer that was not size 0 or 1 mod 8 was passed to OPENSSL_ia32_rdrand_bytes
or OPENSSL_ia32_rdseed_bytes, the last (n mod 8) bytes were all the same value.
This issue impacts only the 64-bit variant of the assembly.

This change fixes both issues by first eliminating the only use of
OPENSSL_ia32_rdrand, replacing it with OPENSSL_ia32_rdrand_bytes, and fixes the
register mixup in OPENSSL_ia32_rdrand_bytes. It also adds a sanity test for
OPENSSL_ia32_rdrand_bytes and OPENSSL_ia32_rdseed_bytes to help catch problems
of this nature in the future.

Reviewed-by: Andy Polyakov <>
Reviewed-by: Rich Salz <>
(Merged from

5 years agoFixed several readability issues in DH_generate_parameters.pod
Alex Gaynor [Thu, 8 Mar 2018 12:59:50 +0000 (07:59 -0500)]
Fixed several readability issues in DH_generate_parameters.pod

Reviewed-by: Tim Hudson <>
Reviewed-by: Matthias St. Pierre <>
(Merged from

5 years agoCorrected two typos in a man page
Alex Gaynor [Wed, 7 Mar 2018 19:01:05 +0000 (14:01 -0500)]
Corrected two typos in a man page

Reviewed-by: Tim Hudson <>
Reviewed-by: Rich Salz <>
(Merged from

5 years agoAdd GOST OIDs for Edwards parameter sets
Sergey Zhuravlev [Wed, 7 Mar 2018 17:58:38 +0000 (12:58 -0500)]
Add GOST OIDs for Edwards parameter sets

Add OIDs for parameter sets of Edwards elliptic curves.

CLA: trivial

Reviewed-by: Bernd Edlinger <>
Reviewed-by: Rich Salz <>
(Merged from

5 years agoRemove OPENSSL_USE_NODELETE guards in shlibloadtest
Matt Caswell [Tue, 6 Mar 2018 10:41:47 +0000 (10:41 +0000)]
Remove OPENSSL_USE_NODELETE guards in shlibloadtest

PR #3399 converted shlibloadtest to the new test framework. It also
seemed to add some `OPENSSL_USE_NODELETE` guards to the library
unloading part of the test. This part was added in a commit with this

    Review feedback; use single main, #ifdef ADD_TEST

    Suppose OPENSSL_USE_NODELETE (via Nick Reilly)

Strangely though there doesn't seem to be any relevant review feedback
in that PR that could justify the addition of those guards. The guards
do not appear in 1.1.0.

Having the guards changes the nature of the test, so that we only test
library unloading on platforms where OPENSSL_USE_NODELETE is set (Linux
and Windows). I can't think of any good reason for this and as it doesn't
seem to be necessary in 1.1.0 so I think we should remove them.

Reviewed-by: Richard Levitte <>
(Merged from

5 years agoImplement multi-process OCSP responder.
Viktor Dukhovni [Mon, 5 Mar 2018 20:18:04 +0000 (15:18 -0500)]
Implement multi-process OCSP responder.

With "-multi" the OCSP responder forks multiple child processes,
and respawns them as needed.  This can be used as a long-running
service, not just a demo program.  Therefore the index file is
automatically re-read when changed.  The responder also now optionally
times out client requests.

Reviewed-by: Matt Caswell <>
5 years agoPrepare to detect index changes in OCSP responder.
Viktor Dukhovni [Mon, 5 Mar 2018 19:40:02 +0000 (14:40 -0500)]
Prepare to detect index changes in OCSP responder.

Retain open file handle and previous stat data for the CA index
file, enabling detection and index reload (upcoming commit).

Check requirements before entering accept loop.

Reviewed-by: Matt Caswell <>
5 years agoConfigurations/unix-Makefile.tmpl: remove assignment of AS and ASFLAGS
Richard Levitte [Wed, 7 Mar 2018 13:52:47 +0000 (14:52 +0100)]
Configurations/unix-Makefile.tmpl: remove assignment of AS and ASFLAGS

We have never used these variables with the Unix Makefile, and there's
no reason for us to change this, so to avoid confusion, we remove them.

Reviewed-by: Rich Salz <>
(Merged from

5 years agoBIO_s_mem.pod: fix indirection for out parameter **pp
Dr. Matthias St. Pierre [Wed, 7 Mar 2018 13:37:23 +0000 (14:37 +0100)]
BIO_s_mem.pod: fix indirection for out parameter **pp

BIO_get_mem_data() and BIO_get_mem_ptr() assign to *pp, not pp

Reviewed-by: Rich Salz <>
(Merged from

5 years agocrypto/armcap.c: mask SHA512 hardware detection on iOS.
Andy Polyakov [Mon, 5 Mar 2018 16:10:03 +0000 (17:10 +0100)]
crypto/armcap.c: mask SHA512 hardware detection on iOS.

When running iOS application from command line it's impossible to
get past the failing capability detection. This is because it's
executed under debugger and iOS debugger is impossible to deal with.
[If Apple implements SHA512 in silicon, it would have to be detected
with sysctlbyname.]

Reviewed-by: Rich Salz <>
5 years agoConfigurations/10-main.conf: add -fno-common back to darwin-ppc-cc.
Andy Polyakov [Mon, 5 Mar 2018 15:55:48 +0000 (16:55 +0100)]
Configurations/10-main.conf: add -fno-common back to darwin-ppc-cc.

-fno-common was removed for all Darwin targets in
0c8734198d4282f6997965a03cd2e0ceaf207549 with rationale "it's either
'ranlib -c' or '-fno-common'." However, it's still absolutely required
in 32-bit darwin-ppc-cc. And when trying things out I didn't quite
see why it was formulated as one-or-another choice, as 'ranlib -c'
shouldn't [and doesn't] have problems with object modules without
commons. [Well, to be frank, I didn't manage to reproduce the problem
the modification was meaning to resolve either...]

Reviewed-by: Rich Salz <>
5 years agoutil/ only quote stuff that actually needs quoting
Richard Levitte [Tue, 6 Mar 2018 20:05:16 +0000 (21:05 +0100)]
util/ only quote stuff that actually needs quoting

Reviewed-by: Rich Salz <>
(Merged from

5 years Use all the "openssl_api_defines"
Richard Levitte [Tue, 6 Mar 2018 20:04:11 +0000 (21:04 +0100)] Use all the "openssl_api_defines"

Reviewed-by: Rich Salz <>
(Merged from

5 years agoConfigure: fix small spelling error
Richard Levitte [Tue, 6 Mar 2018 20:03:21 +0000 (21:03 +0100)]
Configure: fix small spelling error

Reviewed-by: Rich Salz <>
(Merged from

5 years agoAvoid unconditional store in CRYPTO_malloc.
knekritz [Tue, 6 Mar 2018 18:21:49 +0000 (13:21 -0500)]
Avoid unconditional store in CRYPTO_malloc.

Reviewed-by: Richard Levitte <>
Reviewed-by: Rich Salz <>
(Merged from

5 years agoFix a typo in the s_client man page
Alex Gaynor [Tue, 6 Mar 2018 18:07:33 +0000 (13:07 -0500)]
Fix a typo in the s_client man page

Reviewed-by: Matthias St. Pierre <>
Reviewed-by: Rich Salz <>
(Merged from

5 years agoFix a bunch of gcc warnings in packettest.c
Bernd Edlinger [Mon, 5 Mar 2018 18:27:47 +0000 (19:27 +0100)]
Fix a bunch of gcc warnings in packettest.c

Had been observed with recent gcc-8 snapshot and
$ ./config --strict-warnings enable-asan

Reviewed-by: Rich Salz <>
(Merged from

5 years agoCheck the parent DRBG's strength
Kurt Roeckx [Sun, 4 Mar 2018 12:23:05 +0000 (13:23 +0100)]
Check the parent DRBG's strength

We currently don't support the algorithm from NIST SP 800-90C
10.1.2 to use a weaker DRBG as source

Reviewed-by: Dr. Matthias St. Pierre <>
GH: #5506

5 years agobnrand_range: Always call bnrand() with the correct flag
Kurt Roeckx [Sun, 4 Mar 2018 23:29:21 +0000 (00:29 +0100)]
bnrand_range: Always call bnrand() with the correct flag

It was calling the BN_rand() when it should have call BN_priv_rand()

Reviewed-by: Tim Hudson <>
GH: #5514

5 years agoAdd support for .include directive in config files
Tomas Mraz [Tue, 27 Feb 2018 12:55:35 +0000 (13:55 +0100)]
Add support for .include directive in config files

Either files or directories of *.cnf or *.conf files
can be included.
Recursive inclusion of directories is not supported.

Reviewed-by: Andy Polyakov <>
Reviewed-by: Matt Caswell <>
(Merged from

5 years agoIf not sending key_share (no TLSv1.3), return appropriately.
Todd Short [Fri, 2 Mar 2018 14:36:28 +0000 (09:36 -0500)]
If not sending key_share (no TLSv1.3), return appropriately.

This fixes an issue raised in PR #4964 by kaduk.

Reviewed-by: Ben Kaduk <>
Reviewed-by: Matt Caswell <>
(Merged from

5 years agoFix status_request and SCT extensions
Matt Caswell [Fri, 2 Mar 2018 15:02:11 +0000 (15:02 +0000)]
Fix status_request and SCT extensions

They are valid for use in a CertificateRequest message, but we did not
allow it. If a server sent such a message using either of those two
extensions then the handshake would abort.

This corrects that error, but does not add support for actually processing
the extensions. They are simply ignored, and a TODO is inserted to add
support at a later time.

This was found during interoperability testing with btls:

Prompted by these errors I reviewed the complete list of extensions and
compared them with the latest table in draft-24 to confirm there were no
other errors of a similar type. I did not find any.

Reviewed-by: Rich Salz <>
Reviewed-by: Ben Kaduk <>
(Merged from