return errors for unsupported operations

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

Remove reference to ECDSA_OpenSSL.

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

Move and adapt ECDSA sign and verify functions.

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

modify ecdsatest to use accessor

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

Add ECDSA_SIG accessor.

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

move ECDSA_SIG prototypes

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

make errors

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

extend EC_KEY_METHOD for signing support

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

adapt ossl_ecdsa.c to crypto/ec

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

move ECDSA_SIG definition

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

Move ECDSA implementation to crypto/ec

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

Move ECDSA_SIG ASN.1 to crypto/ec

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

Add set methods.

Add set_group, set_public and set_private methods. An EC_KEY_METHOD can use
these to perform any appropriate operation when the key components are set,
such as caching data in some more convenient ENGINE specific format or
returning an error if the parameters are invalid or the operation is
not supported.

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

EC_KEY_METHOD copy support

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

EC_KEY_METHOD init and finish support

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

ENGINE fixes

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

remove ECDH from mkdef.pl

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

remove ECDH_METHOD typedef

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

remove ecdh.h header

Remove redundant ecdh.h header and any references to it.

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

remove ECDH error loading

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

Remove crypto/ecdh update Makefile.org

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

remove ECDH_METHOD from ENGINE

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

make errors

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

Add compute key support to EC_KEY_METHOD

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

Adapt ecdh_compute_key

Rename ecdh_compute_key into ossl_ecdh_compute_key and modify it
to use EC error codes. Remove superfluous old ECDH functions.

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

move ECDH implementation to crypto/ec

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

Move ECDH_KDF_X9_62 to crypto/ec

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

EC_KEY_METHOD keygen support.

Add keygen to EC_KEY_METHOD. Redirect EC_KEY_generate_key through
method and set the current EC key generation function as the default.

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

make errors

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

Support for EC_KEY_METHOD.

Add EC_KEY_METHOD. This is part of the EC revision and will make EC behave
more like other algorithms. Specifically:

EC_KEY_METHOD is part of EC_KEY.
It is part of ENGINE.
Default or key specific implementations can be provided to redirect some
or all operations.

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

Link library for backtrace() on BSD...

Reviewed-by: Rich Salz <rsalz@openssl.org>

NEWS: Add a bit of precision regarding removal of cipher suites

Suggested by Benjamin Kaduk <bkaduk@akamai.com>

Reviewed-by: Matt Caswell <matt@openssl.org>

Additional NEWS

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

Refer to website for acknowledgements.

Reviewed-by: Steve Marquess <marquess@openssl.com>

Not all 'find's know -xtype, use -type instead

Reviewed-by: Rich Salz <rsalz@openssl.org>

Adapt the OS X build to use the OS X tar

As part of this, move release creation to a script to be called from
.travis.yml.  That makes it much easier to test outside of travis.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Make it possible to affect the way dists are made

Introducing DISTTARVARS to propagate changed variables down to the
tar-making target.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Update NEWS

Reviewed-by: Matt Caswell <matt@openssl.org>

Extended master secret test script.

Reviewed-by: Matt Caswell <matt@openssl.org>

Add extms extension

Reviewed-by: Matt Caswell <matt@openssl.org>

TLSProxy update

Add function to delete extensions and fix ClientHello repacking.

Reviewed-by: Matt Caswell <matt@openssl.org>

update errors

Reviewed-by: Matt Caswell <matt@openssl.org>

Extended master secret fixes and checks.

Add new flag TLS1_FLAGS_RECEIVED_EXTMS which is set when the peer sends
the extended master secret extension.

Server now sends extms if and only if the client sent extms.

Check consistency of extms extension when resuming sessions following (where
practical) RFC7627.

Reviewed-by: Matt Caswell <matt@openssl.org>

Fix merge error

Commit 6140f0365 added some new ctrl constants. However due to a
merge error one of these values was duplicated with an existing
value.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Add some new cipher ctrl constants

These are needed for GOST

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>

Cleanup the EVP_MD_CTX before exit rather than after

Reviewed-by: Matt Caswell <matt@openssl.org>

Remove double semi (;)

When in the middle of declarations, some C compilers will complain.

Reviewed-by: Matt Caswell <matt@openssl.org>

Remove typedef of HMAC_CTX from crypto/hmac/hmac_lcl.h

This is already defined in include/openssl/ossl_typ.h.

Reviewed-by: Matt Caswell <matt@openssl.org>

Fix clang complaints about uninitialised variables.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Change tar owner and group to just 0

It seems like some tar versions don't like the name:id form for
--owner and --group.  The closest known anonymous user being 0 (root),
that seems to be the most appropriate user/group to assign ownership
to.  It matters very little when unpacking either way.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Add an entry in CHANGES

Reviewed-by: Rich Salz <rsalz@openssl.org>

Document EVP_MD constructors, destructors and manipulators

Reviewed-by: Rich Salz <rsalz@openssl.org>

Document the HMAC changes

Reviewed-by: Rich Salz <rsalz@openssl.org>

Document the EVP_MD_CTX changes

Reviewed-by: Rich Salz <rsalz@openssl.org>

make update

Reviewed-by: Rich Salz <rsalz@openssl.org>

Cleanup: fix all sources that used HMAC_CTX_init

Reviewed-by: Rich Salz <rsalz@openssl.org>

Cleanup: rename HMAC_CTX_init to HMAC_CTX_reset

Reviewed-by: Rich Salz <rsalz@openssl.org>

Cleanup: support EVP_MD_CTX_(create|init|destroy) for deprecated use

Reviewed-by: Rich Salz <rsalz@openssl.org>

Cleanup: fix all sources that used EVP_MD_CTX_(create|init|destroy)

Reviewed-by: Rich Salz <rsalz@openssl.org>

Cleanup: rename EVP_MD_CTX_(create|init|destroy) to EVP_MD_CTX_(new|reset|free)

Looking over names, it seems like we usually use names ending with
_new and _free as object constructors and destructors.  Also, since
EVP_MD_CTX_init is now used to reset a EVP_MD_CTX, it might as well be
named accordingly.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Cleanup: Remove M_EVP_MD_* macros

These macros were only meant for crypto/evp, and are now entirely
unused.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Adapt the rest of the source to the removal of (EVP_MD_CTX|HMAC_CTX)_cleanup

Reviewed-by: Rich Salz <rsalz@openssl.org>

Remove HMAC_CTX_cleanup and combine its functionality into EVP_MD_CTX_init

This follows the same idea as the combination of EVP_MD_CTX_cleanup
and EVP_MD_CTX_init into one function.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Remove EVP_MD_CTX_cleanup and put its functionality into EVP_MD_CTX_init

The idea is that with EVP_MD_CTX_create() and EVP_MD_CTX_destroy(),
EVP_MD_CTX_cleanup and EVP_MD_CTX_init is not used the same as before.
Instead, we need a single function that can be used to reinitialise an
existing EVP_MD_CTX that's been created with EVP_MD_CTX_create()
previously.  Combining EVP_MD_CTX_cleanup and EVP_MD_CTX_init into
that one function is the answer.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Adapt the rest of the source to the opaque HMAC_CTX

Reviewed-by: Rich Salz <rsalz@openssl.org>

Make the definition of HMAC_CTX opaque

This moves the definition to crypto/hmac/hmac_lcl.h.  Constructor and
destructor added, and the typedef moved to include/openssl/ossl_typ.h.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Add inclusion of internal/evp_int.h to all crypto/ files that need it

These are the files that add new EVP_MDs.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Adapt all engines that add new EVP_MDs

Reviewed-by: Rich Salz <rsalz@openssl.org>

Have the few apps that accessed EVP_MD directly use accessors instead

Reviewed-by: Rich Salz <rsalz@openssl.org>

Make the definition of EVP_MD opaque

This moves the definition to crypto/include/internal/evp_int.h and
defines all the necessary method creators, destructors, writers and
accessors.  The name standard for the latter is inspired from the
corresponding functions to manipulate UI methods.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Adjust all accesses to EVP_MD_CTX to use accessor functions.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Document the changed HMAC API.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Adapt HMAC to the EVP_MD_CTX changes

This change required some special treatment, as HMAC is intertwined
with EVP_MD.  For now, all local HMAC_CTX variables MUST be
initialised with HMAC_CTX_EMPTY, or whatever happens to be on the
stack will be mistaken for actual pointers to EVP_MD_CTX.  This will
change as soon as HMAC_CTX becomes opaque.

Also, since HMAC_CTX_init() can fail now, its return type changes from
void to int, and it will return 0 on failure, 1 on success.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Have other crypto/evp files include evp_locl.h

Note: this does not include the files in crypto/evp that are just
instanciations of EVP_MD.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Make the definition of EVP_MD_CTX opaque

This moves the definitionto crypto/evp/evp_locl.h, along with a few
associated accessor macros.  A few accessor/writer functions added.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Do not add symlinks in the source release

Reviewed-by: Rich Salz <rsalz@openssl.org>

In travis, build from a "source release" rather than from the build tree

Reviewed-by: Rich Salz <rsalz@openssl.org>

Small changes to creating dists

Make TARFILE include ../ instead of having that hard coded all over the place.
When transforming file names in TAR_COMMAND, use $(NAME) instead of openssl-$(VERSION)

Reviewed-by: Rich Salz <rsalz@openssl.org>

Fix and update versions in CHANGES and NEWS

Reviewed-by: Matt Caswell <matt@openssl.org>

ARMv4 assembly pack: allow Thumb2 even in iOS build,
and engage it in most modules.

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

Fix typo and improve a bit of text

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

Really disable 56-bit (single-DES) ciphers

Reviewed-by: Kurt Roeckx <kurt@openssl.org>

Remove support for all 40 and 56 bit ciphers.

Reviewed-by: Rich Salz <rsalz@openssl.org>
MR: #364

Remove SSL_{CTX_}set_ecdh_auto() and always enable ECDH

Reviewed-by: Dr. Stephen Henson <steve@openssl.org>

Make SSL_{CTX}_set_tmp_ecdh() call SSL_{CTX_}set1_curves()

SSL_{CTX}_set_tmp_ecdh() allows to set 1 EC curve and then tries to use it.  On
the other hand SSL_{CTX_}set1_curves() allows you to set a list of curves, but
only when SSL_{CTX_}set_ecdh_auto() was called to turn it on.

Reviewed-by: Dr. Stephen Henson <steve@openssl.org>

Remove support for SSL_{CTX_}set_tmp_ecdh_callback().

This only gets used to set a specific curve without actually checking that the
peer supports it or not and can therefor result in handshake failures that can
be avoided by selecting a different cipher.

Reviewed-by: Dr. Stephen Henson <steve@openssl.org>

Fix EAP FAST in the new state machine

The new state machine code missed an allowed transition when resuming a
session via EAP FAST. This commits adds the missing check for the
transition.

Reviewed-by: Andy Polyakov <appro@openssl.org>

Revert unnecessary SSL_CIPHER_get_bits API change

Reviewed-by: Rich Salz <rsalz@openssl.org>

Run test/run_tests.pl directly in the test_ordinals target

Running 'make TEST=test_ordinals test' starts the whole build process,
which wasn't desired for this target.  Instead, we take a shortcut.

Reviewed-by: Rich Salz <rsalz@openssl.org>

bn/asm/x86_64-mont5.pl: fix carry propagating bug (CVE-2015-3193).

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

perlasm/ppc-xlate.pl: comply with ABIs that specify vrsave as reserved.

RT#4162

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

modes/ocb128.c: fix sanitizer warning.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Fix ./Configure reconf

'./Configure reconf' hasn't been working for a while, because a perl
lable needs to be immediately followed by a block.

Reviewed-by: Andy Polyakov <appro@openssl.org>

Remove RSA_FLAG_SIGN_VER flag.

Remove RSA_FLAG_SIGN_VER: this was origininally used to retain binary
compatibility after RSA_METHOD was extended to include rsa_sign and
rsa_verify fields. It is no longer needed.

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

Move the backtrace memleak options to a separate variable

The contents of this variable ($memleak_devteam_backtrace) is added to
$cflags unless we build for a platform we know doesn't support gcc's
-rdynamic och backtrace() and friends.

Reviewed-by: Andy Polyakov <appro@openssl.org>

make update

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

Remove legacy sign/verify from EVP_MD.

Remove sign/verify and required_pkey_type fields of EVP_MD: these are a
legacy from when digests were linked to public key types. All signing is
now handled by the corresponding EVP_PKEY_METHOD.

Only allow supported digest types in RSA EVP_PKEY_METHOD: other algorithms
already block unsupported types.

Remove now obsolete EVP_dss1() and EVP_ecdsa().

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

Run test_ordinals after update

Catch a common 'make update' failure: conflicting ordinals.

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

_BSD_SOURCE is deprecated, use _DEFAULT_SOURCE instead

The feature_test_macros(7) manual tells us that _BSD_SOURCE is
deprecated since glibc 2.20 and that the compiler will warn about it
being used, unless _DEFAULT_SOURCE is defined as well.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Add backtrace to memory leak output

This is an option for builds with gcc and --strict-warnings.

Reviewed-by: Rich Salz <rsalz@openssl.org>