Remake the way dynamic zlib is loaded

Instead of absolute hard coding of the libz library name, have it use
the macro LIBZ, which is set to defaults we know in case it's
undefined.

This allows our configuration to set something that's sane on current
or older platforms, and allows the user to override it by defining
LIBZ themselves.

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

Configuration: Simplify the way vms information is retrieved

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

Fix zlib configuration options.

The macros ZLIB and ZLIB_SHARED weren't appropriately defined,
deviating wrongly from how they worked in earlier OpenSSL versions.
So, restore it so that ZLIB is defined if configured "enable-zlib" and
so that ZLIB and ZLIB_SHARED are defined if configured
"enable-zlib-dynamic".

Additionally, correct the interpretation of the --with-zlib-lib value
on Windows and VMS, where it's used to indicate the actual zlib
zlib library file / logical name, as that can differ depending on zlib
version and packaging on those platforms.

Finally for Windows and VMS, we also define the macro LIBZ with that
file name / logical name when configured "zlib-dynamic", so the
compression unit can pick it up and use it.

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

Fix no-tls1_1 and no-tls1_2

The above config options were failing in test_ssl_old.

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

Rename int_*() functions to *_int()

There is a preference for suffixes to indicate that a function is internal
rather than prefixes. Note: the suffix is only required to disambiguate
internal functions and public symbols with the same name (but different
case)

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

Tweak to documentation

Tweak to documentation following feedback

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

Misc fix ups to deprecate explicit de-init documentation

Documentation fix ups as a result of feedback received.

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

Fix explicit de-init macros

The no-op de-init macros may fail because of extraneous ";", so we use
a slightly different construct instead.

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

Fix symbols missing from shared build

libssl needs to have access to some internal libcrypto symbols.

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

Rename lots of *_intern or *_internal function to int_*

There was a lot of naming inconsistency, so we try and standardise on
one form.

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

Add a CHANGES entry for the deprecated cleanup functions

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

make update

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

Deprecate CONF_modules_free() and make it a no-op

CONF_modules_free() should not be called expicitly - we should leave
auto-deinit to clean this up instead.

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

Deprecate ENGINE_cleanup() and make it a no-op

ENGINE_cleanup() should not be called expicitly - we should leave
auto-deinit to clean this up instead.

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

Deprecate OBJ_cleanup() and make it a no-op

OBJ_cleanup() should not be called expicitly - we should leave
auto-deinit to clean this up instead.

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

Deprecate EVP_cleanup() and make it a no-op

EVP_cleanup() should not be called expicitly - we should leave
auto-deinit to clean this up instead.

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

Deprecate BIO_sock_cleanup() and make it a no-op

BIO_sock_cleanup() should not be called expicitly - we should leave
auto-deinit to clean this up instead.

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

Deprecate CRYPTO_cleanup_all_ex_data() and make it a no-op

CRYPTO_cleanup_all_ex_data() should not be called expicitly - we should
leave auto-deinit to clean this up instead.

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

Deprecate RAND_cleanup() and make it a no-op

RAND_cleanup() should not be called expicitly - we should leave
auto-deinit to clean this up instead.

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

Fix conditional compile logic in speed.c

The conditional compile logic wasn't quite right in speed.c for when
both OPENSSL_NO_DSA and OPENSSL_NO_EC are defined.

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

Deprecate SSL_COMP_free_compression_methods() and make it a no-op

SSL_COMP_free_compression_methods() should not be called expicitly - we
should leave auto-deinit to clean this up instead.

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

Deprecate ERR_free_strings() and make it a no-op

ERR_free_strings() should not be called expicitly - we should leave
auto-deinit to clean this up instead.

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

Deprecate COMP_zlib_cleanup() and make it a no-op

COMP_zlib_cleanup() should not be called expicitly - we should leave
auto-deinit to clean this up instead.

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

X509_PUBKEY docs

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

Fix setting of debug callback in apps/enc.c

Setting the debug callback on the in and out BIOs were done too early.
No debug callback was set on the zlib filter BIO.

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

Add SSL_CTX_get_ciphers()

Add an accessor for SSL_CTX.

Since libssl was made opaque, there is no way for users to access the
cipher_list, while users can set the cipher_list by
SSL_CTX_set_cipher_list().

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

Fix the no-posix-io option

Fix a compile failure with no-posix-io

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

Remove an unused function

The function pqueue_print is not exported and is never called. Therefore
we should delete it.

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

Fix the no-tls option

The no-tls option was failing in the tests. This fixes it.

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

Fix the no-nextprotoneg option

Misc fixes to get no-nextprotoneg config option working again.

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

BIO: respect opening in text mode

When a file is opened with BIO_new_file(), make sure that the internal
mode TEXT vs BINARY setting reflects what's given in the mode string.

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

Disable some sanitizer checks without PEDANTIC

Code without PEDANTIC has intentional "undefined" behaviour. To get best
coverage for both PEDANTIC and non-PEDANTIC codepaths, run the sanitizer
builds in two different configurations:
1) Without PEDANTIC but with alignment checks disabled.
2) With PEDANTIC.

To not overload Travis too much, run one build with clang and the other
with gcc (chosen at random).

Also remove a micro-optimization in CAST code to be able to
-fsanitize=shift. Whether shift sanitization is meaningful for crypto or
an obstacle is debatable but since this appears to be the only offender,
we might as well keep the check for now.

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

Fix warnings exposed by clang-3.8

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

Adjust --strict-warnings builds in Travis

In Travis, do --strict-warnings on BUILDONLY configurations. This
ensures that the tests run even if --strict-warnings fail, and avoids
hiding unrelated test failures.

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

Don't check the generated ssl-tests configs on VMS

The simple reason is that the pre-generated files are mainly for Unix.
The VMS variants look slightly different, so comparing will always fail.

Reviewed-by: Emilia Käsper <emilia@openssl.org>

VMS doesn't have directory separators in the same sense as Unix

Don't add / in file specs on VMS.  The directory "separator" is part
of the directory spec.

Reviewed-by: Emilia Käsper <emilia@openssl.org>

RSA: the docs still talk about RSA_PKCS1_SSLeay

Rename the function to RSA_PKCS1_OpenSSL.

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

Add CHANGES entry for DH and DH_METHOD opacity

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

Add documentation for following DH and DH_METHOD opacity

A number of new functions have been added following the DH and DH_METHOD
opacity commits. This commit provides documentation for those functions.

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

Fix double free bug in error path

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

make update

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

Make DH_METHOD opaque

Move the dh_method structure into an internal header file and provide
relevant accessors for the internal fields.

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

make update

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

Make DH opaque

Move the dh_st structure into an internal header file and provide
relevant accessors for the internal fields.

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

Add SSL_DANE typedef for consistency.

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

Add OCSP to mkdef

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

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>

Clean-up *_DEBUG options.

Since NDEBUG is defined unconditionally on command line for release
builds, we can omit *_DEBUG options in favour of effective "all-on"
in debug builds exercised though CI.

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

Configurations/10-main.conf: omit now redundant -D_WINDLL.

... and refine /MT vs. /MD flag handling.

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

windows-makefile.tmpl: clean up after DLL link failure.

Without proper cleanup after DLL link failure second attempt to
run nmake would actually proceed and failure will be "shifted" to
run time. This is because libcrypto.lib import library is generated
even if DLL link fails.

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

test/[dane|evp_]test.c: BIO-fy file I/O.

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

bio/bss_file.c: since VS2015 one can't tell apart own and "alien" FILE
pointers, except for minimal std[in|out|err].

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

make update

Signed-off-by: Rob Percival <robpercival@google.com>
Reviewed-by: Emilia Käsper <emilia@openssl.org>

Suppress CT callback as appropriate

Suppress CT callbacks with aNULL or PSK ciphersuites that involve
no certificates.  Ditto when the certificate chain is validated via
DANE-TA(2) or DANE-EE(3) TLSA records.  Also skip SCT processing
when the chain is fails verification.

Move and consolidate CT callbacks from libcrypto to libssl.  We
also simplify the interface to SSL_{,CTX_}_enable_ct() which can
specify either a permissive mode that just collects information or
a strict mode that requires at least one valid SCT or else asks to
abort the connection.

Simplified SCT processing and options in s_client(1) which now has
just a simple pair of "-noct" vs. "-ct" options, the latter enables
the permissive callback so that we can complete the handshake and
report all relevant information.  When printing SCTs, print the
validation status if set and not valid.

Signed-off-by: Rob Percival <robpercival@google.com>
Reviewed-by: Emilia Käsper <emilia@openssl.org>

Fix client verify mode to check SSL_VERIFY_PEER

The original check for != SSL_VERIFY_NONE can give surprising results
when flags SSL_VERIFY_PEER is not set, but other flags are.  Note
that SSL_VERIFY_NONE (0) is not a flag bit, it is rather the absense
of all other flag bits.

Signed-off-by: Rob Percival <robpercival@google.com>
Reviewed-by: Emilia Käsper <emilia@openssl.org>

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>

OpenSSL::Test: when moving directory, affect env as well

The environment variables TOP, SRCTOP, BLDTOP, ...  are used to affect
the testing framework.  However, subprocesses may want to use them as
well, and therefore need their values corrected when we move to a
different directory.

Reviewed-by: Emilia Käsper <emilia@openssl.org>

Allow generate_ssl_tests.pl to find testlib

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

Remove redundant symlink

We used to symlink generate_ssl_tests.pl to the build directory.
Now that the build scripts look for sources in both directories, this
is no longer necessary (see commit
fbd361eaf84446e8d6860ab2b7ecf9d04585f2ef).

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

Better use BIO_snprintf() than snprintf(), in case the later isn't available

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

Fix forgotten adaptation to opaque RSA and RSA_METHOD

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

always use field names

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

Perl: foreach (@list) { code } is better

Reviewed-by: Emilia Käsper <emilia@openssl.org>

Add apps/tsget and test/ssltest_old to .gitignore

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

make update

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

Update CHANGES with the new about RSA and RSA_METHOD

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

Document RSA_METHOD creators/destructor/accessors/writers

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

Make the RSA_METHOD structure opaque

Move rsa_meth_st away from public headers.
Add RSA_METHOD creator/destructor functions.
Add RSA_METHOD accessor/writer functions.
Adapt all other source to use the creator, destructor, accessors and writers.

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

Document RSA accessors/writers

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

Make the RSA structure opaque

Move rsa_st away from public headers.
Add accessor/writer functions for the public RSA data.
Adapt all other source to use the accessors and writers.

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

Fix no-ocsp

Misc fixes for no-ocsp

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

Remove some OPENSSL_NO_ASYNC guards in init

When config'd with "no-async" the ASYNC_NULL implementation is used, so
async symbols still exist. We should still init the NULL implementation so
that when we get the async ctx it is NULL rather than undefined.

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

Perl cleanup: don't create lists unnecessarily

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

VMS: Fix special case for [.test]ssltest_old.c

[.test]ssltest.c was renamed to [.test]ssltest_old.c, reflect that in
descrip.mms.tmpl.

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

Configure: add BLAKE_DEBUG to --strict-warnings set.

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

Add missing mem leak test activation and checks

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

Add a check of the FD_SETSIZE before the call to select()

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

Rename ssltest -> ssltest_old

ssltest_old.c is deprecated. New tests should use ssl_test.c, and the
recipes in 80-test_ssl_new.t

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

testutil: return 1 on success

Require that test methods return 1 on success (not 0). This is more
customary for OpenSSL.

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

fix memory leak in ca

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

New SSL test framework

Currently, SSL tests are configured via command-line switches to
ssltest.c. This results in a lot of duplication between ssltest.c and
apps, and a complex setup. ssltest.c is also simply old and needs
maintenance.

Instead, we already have a way to configure SSL servers and clients, so
we leverage that. SSL tests can now be configured from a configuration
file. Test servers and clients are configured using the standard
ssl_conf module. Additional test settings are configured via a test
configuration.

Moreover, since the CONF language involves unnecessary boilerplate, the
test conf itself is generated from a shorter Perl syntax.

The generated testcase files are checked in to the repo to make
it easier to verify that the intended test cases are in fact run; and to
simplify debugging failures.

To demonstrate the approach, min/max protocol tests are converted to the
new format. This change also fixes MinProtocol and MaxProtocol
handling. It was previously requested that an SSL_CTX have both the
server and client flags set for these commands; this clearly can never work.

Guide to this PR:
 - test/ssl_test.c - test framework
 - test/ssl_test_ctx.* - test configuration structure
 - test/handshake_helper.* - new SSL test handshaking code
 - test/ssl-tests/ - test configurations
 - test/generate_ssl_tests.pl - script for generating CONF-style test
   configurations from perl inputs

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

Fix a shadow symbol warning

... comes from c5137473bdc7.
Fix Travis builds.

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

set exec attribute for .pl files

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

fix perl shebang

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

use whitespace more consistently

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

Revert "various spelling fixes"

This reverts commit 620d540bd47a96fb6905fbbdd8ea5167a8841a3e.
It wasn't reviewed.

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

Revert "Fix an error code spelling."

This reverts commit 2b0bcfaf834e2fb7cd52888d7330b247e3878115.
It wasn't reviewed.

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

Fix an error code spelling.

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

various spelling fixes

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

Fix a possible leak on NETSCAPE_SPKI_verify failure.

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

Use X509_REQ_get0_pubkey

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

Add X509_REQ_get0_pubkey method

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

Fix two leaks in X509_REQ_to_X509

Issue #182

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

Make sure the rand_byte buffer in padlock engine is cleansed.

Submitted by Michael McConville <mmcco@mykolab.com>

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

crypto/poly1305: don't break carry chains.

RT#4483

[poly1305-armv4.pl: remove redundant #ifdef __thumb2__]
[poly1305-ppc*.pl: presumably more accurate benchmark results]

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

Fix memory leaks in ASN.1

These leaks affect 1.1.0 dev branch only; introduced around commit
f93ad22f6adb00e722c130e792799467f3927b56

Found with LibFuzzer

Reviewed-by: Ben Laurie <ben@openssl.org>

Removed no-ops for the old locking API

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

Fix mixed declarations and code

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

Don't shadow known symbols write, read, puts, gets

It was harmless in this case, but best avoid the annoying warnings.

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

Makefile et al template: only modify static library with new object files

Previously, we updated the static libraries (libcrypto.a on Unix,
libcrypto.lib on Windows) with all the object files, regardless of if
they were rebuilt or not.  With this change, we only update them with
the object files were rebuilt.

NOTE: this does not apply on VMS, as the expansion of $? may be too
large for a command line.

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