If errno is ENXIO in BSS_new_file(), set BIO_R_NO_SUCH_FILE

VMS sets that errno when the device part of a file spec is malformed
or a logical name that doesn't exist.

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

Travis: Do asan and msan with shared libraries

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

Fix mem leaks during auto-deinit

Certain functions are automatically called during auto-deinit in order
to deallocate resources. However, if we have never entered a function which
marks lib crypto as inited then they never get called. This can happen if
the user only ever makes use of a small sub-set of functions that don't hit
the auto-init code.

This commit ensures all such resources deallocated by these functions also
init libcrypto when they are initially allocated.

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

Ensure trace recognises X25519

Using the -trace option to s_server or s_client was incorrectly printing
UNKNOWN for the X25519 curve.

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

engines/afalg: make it compile with backward compatibility headers.

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

Configurations/10-main.cf: AIX "facelift".

Improve interchangeability of aix*-gcc targets by linking shared
libraries with -static-libgcc, and address linking problems with
vendor compiler.

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

chacha/asm/chacha-ppc.pl: add missing .text directive.

RT#4667

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

Add missing debug strings.

Found by turning -Wswitch-enum on.

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

Allow asan, msan and ubsan to be configured with shared libraries

The background story is that util/shlib_wrap.sh was setting LD_PRELOAD
or similar platform dependent variables, just in case the shared
libraries were built with -rpath.  Unfortunately, this doesn't work
too well with asan, msan or ubsan.

So, the solution is to forbid the combination of shared libraries,
-rpath and any of the sanity analyzers we can configure.

This changes util/shlib_wrap.sh so it only contains the code that sets
LD_PRELOAD when -rpath has been used when configuring.

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

Avoid EVP_PKEY_cmp() crash on EC keys without public component

Some hardware devices don't provide the public EC_POINT data. The only
way for X509_check_private_key() to validate that the key matches a
given certificate is to actually perform a sign operation and then
verify it using the public key in the certificate.

Maybe that can come later, as discussed in issue 1532. But for now let's
at least make it fail gracefully and not crash.

GH: 1532

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1547)
(cherry picked from commit 92ed7fa575a80955f3bb6efefca9bf576a953586)

Configure: Reorganise the checking of disabled options

The way we figured out what options are crypto algorithms and what are
something other was somewhat sketchy.  This change bases the
distinction on available sdirs instead.

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

GH1537: Avoid double-free in the EVP_PKEY API

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

Misc BN fixes

Never output -0; make "negative zero" an impossibility.
Do better checking on BN_rand top/bottom requirements and #bits.
Update doc.
Ignoring trailing garbage in BN_asc2bn.

Port this commit from boringSSL: https://boringssl.googlesource.com/boringssl/+/899b9b19a4cd3fe526aaf5047ab9234cdca19f7d%5E!/
        Ensure |BN_div| never gives negative zero in the no_branch code.

        Have |bn_correct_top| fix |bn->neg| if the input is zero so that we
        don't have negative zeros lying around.

        Thanks to Brian Smith for noticing.

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

Correct detection of group end in map file when testing symbol presence

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

Unix build: have the makedepend and cc actions in one recipe

In the case of using an independent makedepend, we had split that into
two separate recipes, one depending on the other.  However, there are
cases where the makedepend recipe was always trying, but doesn't
update the time stamp of the .d file because there are no actual
changes, and thereby causing constant updates of the object files.

This change makes one recipe that takes care of both makedepend och
cc, thereby avoiding these extra updates.

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

modes/asm/ghash-armv4.pl: improve interoperability with Android NDK.

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

Configure: clarify and refine -static.

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

Configurations/10-main.conf: add android64-mips64 target.

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

MIPS assembly pack: adapt it for MIPS[32|64]R6.

MIPS[32|64]R6 is binary and source incompatible with previous MIPS ISA
specifications. Fortunately it's still possible to resolve differences
in source code with standard pre-processor and switching to trap-free
version of addition and subtraction instructions.

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

Configurations/10-main.conf: add android64-x86_64 target.

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

Configurations/10-main.conf: fix omittions in commentary.

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

Configurations/10-main.conf: remove solaris-x86-cc target.

Since vendor assembler can't assemble our modules with -KPIC flag,
it, assembly support, was not available as an option. But this
means lack of side-channel resistant code, which is incompatible
with security by todays standards.

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

Move 05-test_fuzz.t to 90-test_fuzz.t

This adheres much better to the documentation in test/README

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

Document the enhanced tests specification

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

Revert "INSTALL: add information on option no-fuzz-test"

This reverts commit 7f9ae88817ddf0aac5c6bd95d9a5af1c54ed5bbf.

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

Revert "Make it possible to disable fuzz testing"

This reverts commit eb40eaed727500bf4a15f848c99e37edd18e142e.

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

Add a more versatile test chooser

So far, the test runner (test/run_tests.pl) could get a list of tests
to run, and if non were given, it assumes all available tests should
be performed.

However, that makes skipping just one or two tests a bit of a pain.
This change makes the possibilities more versatile, run_checker.pl
takes these arguments and will process them in the given order,
starting with an empty set of tests to perform:

    alltests            The current set becomes the whole set of
                        available tests.
    test_xxx            Adds 'test_xxx' to the current set.
    -test_xxx           Removes 'test_xxx' from the current set.  If
                        nothing has been added to the set before this
                        argument, the current set is first initialised
                        to the whole set of available tests, then
                        'test_xxx' is removed from the current set.
    list                Display all available tests, then stop.

If no arguments are given, 'alltests' is assumed.

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

INSTALL: add information on option no-fuzz-test

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

INSTALL: add information on the environment variable BUILDFILE

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

INSTALL: clarify OPENSSL_LOCAL_CONFIG_DIR

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

Have Configure's HASH or TABLE produce complete lists

Because some targets execute perl code that might die, we risk
incomplete lists.  Make it so dying doesn't happen when we're listing
targets.

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

Configure's print_table_entry printed incorrect information

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

Make it possible to disable fuzz testing

These tests take a very long time on some platforms, and arent't
always strictly necessary.  This makes it possible to turn them
off.  The necessary binaries are still built, though, in case
someone still wants to do a manual run.

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

crypto/bn/*: x86[_64] division instruction doesn't handle constants, change constraint from 'g' to 'r'.

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

The Perl interpreter might be in a path with spaces, so maybe quote it

Note: some shells do not like the command verb to be quoted, so we avoid
it unless it's actually necessary.

RT#4665

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

Code cleanup UI

Remove NULL check on parameter, and use NULL not ! on buffer.

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

Add some CertStatus tests

The previous commit revealed a long standing problem where CertStatus
processing was broken in DTLS. This would have been revealed by better
testing - so add some!

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

Ensure the CertStatus message adds a DTLS message header where needed

The function tls_construct_cert_status() is called by both TLS and DTLS
code. However it only ever constructed a TLS message header for the message
which obviously failed in DTLS.

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

Configure: save away the value of OPENSSL_LOCAL_CONFIG_DIR for reconf

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

Configure: Redo the logic for finding build file templates

Build file templates would be looked up like this if the user gave us
an additional directory to look for configuration files and build file
templates:

    $OPENSSL_LOCAL_CONFIG_DIR/$OSTYPE-Makefile.tmpl
    $SOURCEDIR/Configurations/$OSTYPE-Makefile.tmpl
    $OPENSSL_LOCAL_CONFIG_DIR/Makefile.tmpl
    $SOURCEDIR/Configurations/Makefile.tmpl

So for example, if the user created his own Makefile.tmpl and tried to
use it with a unixly config, it would never be user because we have a
unix-Makefile.tmpl in our Configurations directory.  This is clearly
wrong, and this change makes it look in this order instead:

    $OPENSSL_LOCAL_CONFIG_DIR/$OSTYPE-Makefile.tmpl
    $OPENSSL_LOCAL_CONFIG_DIR/Makefile.tmpl
    $SOURCEDIR/Configurations/$OSTYPE-Makefile.tmpl
    $SOURCEDIR/Configurations/Makefile.tmpl

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

Configure: clean away temporary section of code

We've done away with Makefile as source of information and now use
configdata.pm exclusively.

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

Make it possible for the user to specify a different default build file

Make sure the information is kept for reconfiguration too.

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

Add ecp_nistz256-ppc64 module.

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

perlasm/ppc-xlate.pl: recognize .type directive.

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

bn/asm/ppc.pl: harmonize .size directive in bn_mul_words.

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

Remove comment tags from structs (coding style)

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

Use uppercase name for PCT_ enum

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

VMS: honor --openssldir setting

Because of a perl operator priority mixup, the --openssldir argument
wasn't honored.

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

Remove trailing zeros

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

Improve the definition of STITCHED_CALL in e_rc4_hmac_md5.c

The definition of STITCHED_CALL relies on OPENSSL_NO_ASM.  However,
when a configuration simply lacks the assembler implementation for RC4
(which is where we have implemented the stitched call), OPENSSL_NO_ASM
isn't implemented.  Better, then, to rely on specific macros that
indicated that RC4 (and MD5) are implemented in assembler.

For this to work properly, we must also make sure Configure adds the
definition of RC4_ASM among the C flags.

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

Remove note from CHANGES about EC DRBG

EC DRBG support was added in 7fdcb457 in 2011 and then later removed.
However the CHANGES entry for its original addition was left behind.
This just removes the spurious CHANGES entry.

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

Update CHANGES, NEWS, README and opensslv.h on master

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

Fix uninit read in sslapitest

msan detected an uninit read.

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

CHANGES: mention Windows UTF-8 opt-in option.

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

Windows: UTF-8 opt-in for command-line arguments and console input.

User can make Windows openssl.exe to treat command-line arguments
and console input as UTF-8 By setting OPENSSL_WIN32_UTF8 environment
variable (to any value). This is likely to be required for data
interchangeability with other OSes and PKCS#12 containers generated
with Windows CryptoAPI.

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

Support broken PKCS#12 key generation.

OpenSSL versions before 1.1.0 didn't convert non-ASCII
UTF8 PKCS#12 passwords to Unicode correctly.

To correctly decrypt older files, if MAC verification fails
with the supplied password attempt to use the broken format
which is compatible with earlier versions of OpenSSL.

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

Don't switch password formats using global state.

To avoid possible race conditions don't switch password format using
global state in crypto/pkcs12

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

Fix an uninitialised read on an error path

Found by Coverity.

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

NEWS: add a number of the types that were made opaque

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

Un-delete still documented X509_STORE_CTX_set_verify

It should not have been removed.

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

Configurations/10-main.conf: fix solaris64-*-cc link problems.

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

ec/asm/ecp_nistz256-x86_64.pl: /cmovb/cmovc/ as nasm doesn't recognize cmovb.

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

Clarify the error messages in 08f6ae5b28

Ensure it is clear to the user why there has been an error.

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

Fix no-ec2m

The new curves test did not take into account no-ec2m

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

CRYPTO_atomic_add(): check that the object is lock free

If not, fall back to our own code, using the given mutex

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

CRYPTO_atomic_add(): use acquire release memory order rather than relaxed

For increments, the relaxed model is fine.  For decrements, it's
recommended to use the acquire release model.  We therefore go for the
latter.

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

Check for __GNUC__ to use GNU C atomic buildins

Note: we trust any other compiler that fully implements GNU extension
to define __GNUC__

RT#4642

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

Trust RSA_check_key() to return correct values

In apps/rsa.c, we were second guessing RSA_check_key() to leave error
codes lying around without returning -1 properly.  However, this also
catches other errors that are lying around and that we should not care
about.

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

Avoid overflow in MDC2_Update()

Thanks to Shi Lei for reporting this issue.

CVE-2016-6303

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

Put DES into "not default" category.

Add CVE to CHANGES

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

To avoid SWEET32 attack, move 3DES to weak

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

Typo fixes

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

Updates the CT_POLICY_EVAL_CTX POD

Ownership semantics and function names have changed.

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

Correct documentation about SCT setters resetting validation status

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

Removes the SCT_verify* POD

SCT_verify_v1 has been removed and SCT_verify is no longer part of the
public API.

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

Documents the SCT validation functions

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

Removes {o2i,i2o}_SCT_signature from PODs

These functions have been removed from the public API.

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

Documents the CTLOG functions

CTLOG_new_null() has been removed from the code, so it has also been
removed from this POD.

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

Document the i2o and o2i SCT functions

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

Removes d2i_SCT_LIST.pod

This is covered by d2i_X509.pod.

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

Document that SCT_set_source returns 0 on failure.

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

Clarifies the format of a log's public key in the CONF file

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

Refer to OPENSSLDIR rather than "the OpenSSL install directory"

The prior wording was less accurate.
See https://github.com/openssl/openssl/pull/1372#discussion_r73127000.

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

Adds history section to CT PODs

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

Fixes final issue in CT PODs highlighted by util/find-doc-nits.pl

Fixes complaint "ct missing from SYNOPSIS".

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

Renames CT_POLICY_EVAL_CTX.pod to CT_POLICY_EVAL_CTX_new.pod

util/fix-doc-nits.pl complains that
"CT_POLICY_EVAL_CTX (filename) missing from NAME section".

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

Adds copyright section to ct.pod

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

Adds newline after =cut in PODs

util/find-doc-nits.pl complains that the file "doesn't end with =cut".

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

Adds missing function names to NAME section of PODs

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

Add enum definitions to CT pods

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

Remove unnecessary bold tags in CT pods

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

Add comment about calling CT_POLICY_EVAL_CTX_free

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

Fix comment about what SCT_LIST_validate does.

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

Add SSL_get0_peer_scts to ssl.pod

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

Fix comment about return value of ct_extract_tls_extension_scts

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

First draft of CT documentation

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

Remove some dead code from rec_layer_s3.c

It is never valid to call ssl3_read_bytes with
type == SSL3_RT_CHANGE_CIPHER_SPEC, and in fact we check for valid values
for type near the beginning of the function. Therefore this check will never
be true and can be removed.

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

Remove useless assignment

The variable assignment c1 is never read before it is overwritten.

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

Fix some resource leaks in the apps

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

ec/ecp_nistz256: harmonize is_infinity with ec_GFp_simple_is_at_infinity.

RT#4625

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