Make the handling of output and input formats consistent

Most of all, we needed to sort out which ones are binary and which
ones are text, and make sure they are treated accordingly and
consistently so

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

Make the verify_extra test location agnostic

Depending on platform, verify_extra_test may fail because it relies on
test/ being the current working directory.  Make it get all the required
files on the command line instead to solve that issue.

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

make update

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

Avoid direct X509 structure access

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

Replace X509 macros with functions

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

Cleanup testtsa script

Add RUN function; remove single-use functions and use their body inline.

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

Make TS structures opaque.

Most of the accessors existed and were already used so it was easy.
TS_VERIFY_CTX didn't have accessors/settors so I added the simple and
obvious ones, and changed the app to use them.  Also, within crypto/ts,
replaced the functions with direct access to the structure members
since we generally aren't opaque within a directory.

Also fix RT3901.

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

RT3955: Reduce some stack usage

Use malloc/free instead of big onstack buffers.

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

In X509_STORE_CTX_init, cleanup on failure

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

RT3951: Add X509_V_FLAG_NO_CHECK_TIME to suppress time check

In some environments, such as firmware, the current system time is entirely
meaningless. Provide a clean mechanism to suppress the checks against it.

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

Use the shlib wrapper when running nptest

Since there seems to be no way to avoid linking to libssl and libcrypto,
just wrap the test. This unbreaks "shared" builds when using clang and/or
OS X.

Signed-off-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Ben Laurie <ben@openssl.org>

make X509_REVOKED opaque

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

Supply a build rule for the recently added nptest

Commit d4ab70f27cb7e518e6a9d6323c996cc3feb7496b added a test program
to check that the NULL pointer is represented as all zero bits, but
did not specify a build rule for that new executable.  On many platforms,
the implicit rule sufficed, since nptest is a very simple program, but
for at least darwin-i386-cc, an explicit rule is needed.  On darwin-i386-cc,
the implicit rule targetted a 64-bit executable, but the object file
containing the definition of main was a 32-bit object, which the linker
excluded from consideration, resulting in a link failure due to no
definition for _main.

Add the missing build rule to fix the build on such platforms.

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

Match SUITEB strings at start of cipher list.

PR#4009.

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

Test for NULL ptr == 0

Add a test to ensure that "char *p = NULL" is equivalent to
all-bytes-zero.

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

RT3998: Allow scrypt to be disabled

This does 64-bit division and multiplication, and on 32-bit platforms
pulls in libgcc symbols (and MSVC does similar) which may not be
available.  Mostly done by David Woodhouse.

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

More zalloc nits

Found on GitHub by dimman

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

Revert "OPENSSL_NO_xxx cleanup: RFC3779"

This reverts the non-cleanup parts of commit c73ad69017. We do actually
have a reasonable use case for OPENSSL_NO_RFC3779 in the EDK2 UEFI
build, since we don't have a strspn() function in our runtime environment
and we don't want the RFC3779 functionality anyway.

In addition, it changes the default behaviour of the Configure script so
that RFC3779 support isn't disabled by default. It was always disabled
from when it was first added in 2006, right up until the point where
OPENSSL_NO_RFC3779 was turned into a no-op, and the code in the
Configure script was left *trying* to disable it, but not actually
working.

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

remove 0 assignments.

After openssl_zalloc, cleanup more "set to 0/NULL" assignments.
Many are from github feedback.

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

Add UEFI flag for rand build

Add OPENSSL_SYS_UEFI flag for RAND handling;

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

Win32 build fix: include internal/numbers.h to get UIN32_MAX

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

RT3674: Make no-cms build work.

Also has changes from from David Woodhouse <David.Woodhouse@intel.com>
and some tweaks from me.

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

make update

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

PBE lookup test

Add test to check PBE lookups: these can fail if the PBE table is not
correctly orders. Add to "make test".

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

Add initial Travis CI configuration

Closes #63

Signed-off-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Ben Laurie <ben@openssl.org>

Two changes at ones lead to a confused libeay.num.  Fix

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

Check OPENSSL_gmtime_diff

It's test code that only runs on 64bit time_t machines.
Move it to a standalone test/gmdifftest

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

Add and use OPENSSL_zalloc

There are many places (nearly 50) where we malloc and then memset.
Add an OPENSSL_zalloc routine to encapsulate that.
(Missed one conversion; thanks Richard)
Also fixes GH328

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

make update

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

make X509_CRL opaque

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

Extend ciphersuite test coverage.

Add support for testing ECDSA and DSA ciphersuites.

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

Fix CCM support in DTLS

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

Update ssltest certificate handling.

Use SSL_CONF for certificate handling is ssltest.c, this changes the
behaviour slightly: the -cert and -key options are no longer recognised
and a default certificate file is not used.

This change means that -s_cert and -c_cert can be used mode than once
to support use of multiple certificates.

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

Create DSA and ECDSA certificates.

If supported create DSA and ECDSA certificates and test them.

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

delete unused structure

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

Move OPENSSL_ITEM to store.h

The only place that uses it.

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

Cleaner handling of "cnid" in do_x509_check

Avoid using cnid = 0, use NID_undef instead, and return early instead
of trying to find an instance of that in the subject DN.

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

Better handling of verify param id peername field

Initialize pointers in param id by the book (explicit NULL assignment,
rather than just memset 0).

In x509_verify_param_zero() set peername to NULL after freeing it.

In x509_vfy.c's internal check_hosts(), avoid potential leak of
possibly already non-NULL peername.  This is only set when a check
succeeds, so don't need to do this repeatedly in the loop.

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

Fix DTLS1.2 buffers

Fix the setup of DTLS1.2 buffers to take account of the Header

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

Fix the comment for POINT_CONVERSION_UNCOMPRESSED

The |z| value should be 0x04 not 0x02

RT#3838

Signed-off-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Emilia Käsper <emilia@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>

make update

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

Document extension functions

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

functions to retrieve certificate flags

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

use uint32_t for certificate flags

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

RT3984: Fix clang compiler warning on Mac OS X where %ld is used for uint64_t.

clang suggests %llu instead, but it isn't clear that is portable on
all platforms.

C99 and above define a handy macro for us, so we try to use that
definition and fall back to current definition if needed (though we
switch to 'u' for unsigned).

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

apps/speed.c: fix memory leak

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

RT4002: check for NULL cipher in p12_crpt.c

The NULL cipher case can't actually happen because we have no
EVP_PBE_CTL combinations where cipher_nid is -1 and keygen is
PKCS12_PBE_keyivgen. But make the code more obviously correct.

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

RT 3493: fix RSA test

- Pass in the right ciphertext length to ensure we're indeed testing
  ciphertext corruption (and not truncation).
- Only test one mutation per byte to not make the test too slow.
- Add a separate test for truncated ciphertexts.

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

RT3767: openssl_button.gif should be PNG

No, we should just delete it.  And updated the README

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

Ignore .dir-locals.el

Because we recently encourage people to have a .dir-locals.el, it's a good
idea to ignore it on a git level.

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

make update

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

Add X509_up_ref function.

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

Add X509_CRL_up_ref function

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

check bn_new return value

Slightly modified from the original PR.
Signed-off-by: Rich Salz <rsalz@akamai.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>

make X509_REQ opaque

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

make X509_CERT_AUX opaque

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

Remove asn1-kludge option.

Remove asn1-kludge option from the req utility. It was a decade old
workaround for CAs and software which required an invalid encoding
of PKCS#10 certificate requests: omitting the attributes field even
though it is not OPTIONAL.

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

GH367 follow-up, for more clarity

Signed-off-by: Rich Salz <rsalz@akamai.com>
Reviewed-by: Emilia Käsper <emilia@openssl.org>

BN_mod_exp_mont_consttime: check for zero modulus.

Don't dereference |d| when |top| is zero. Also test that various BIGNUM methods behave correctly on zero/even inputs.

Follow-up to b11980d79a52ec08844f08bea0e66c04b691840b

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

Remove auto-fill-mode

Apparently, emacs sees changes to auto-fill-mode as insecure

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

Add an example .dir-locals.el

This file, when copied to .dir-locals.el in the OpenSSL source top,
will make sure that the CC mode style "OpenSSL-II" will be used for
all C files.

Additionally, I makes sure that tabs are never used as indentation
character, regardless of the emacs mode, and that the fill column is
78.

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

Add emacs CC mode style for OpenSSL

This hopefully conforms closely enough to the current code style.

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

Fix spurious bntest failures.

BN_bntest_rand generates a single-word zero BIGNUM with quite a large probability.

A zero BIGNUM in turn will end up having a NULL |d|-buffer, which we shouldn't dereference without checking.

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

Remove the "times" directory.

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

Fix 4c42ebd; forgot to inutil util/libeay.num

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

GH354: Memory leak fixes

Fix more potential leaks in X509_verify_cert()
Fix memory leak in ClientHello test
Fix memory leak in gost2814789 test
Fix potential memory leak in PKCS7_verify()
Fix potential memory leaks in X509_add1_reject_object()
Refactor to use "goto err" in cleanup.

Signed-off-by: Rich Salz <rsalz@akamai.com>
Reviewed-by: Emilia Käsper <emilia@openssl.org>

GH367: Fix dsa keygen for too-short seed

If the seed value for dsa key generation is too short (< qsize),
return an error. Also update the documentation.

Signed-off-by: Rich Salz <rsalz@akamai.com>
Reviewed-by: Emilia Käsper <emilia@openssl.org>

Fix memory over-read

Fix from David Baggett via tweet.

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

Various doc fixes.

Make all mention of digest algorithm use "any supported algorithm"

RT2071, some new manpages from Victor B. Wagner <vitus@cryptocom.ru>:
    X509_LOOKUP_hash_dir.pod
    X509_check_ca.pod
    X509_check_issued.pod

RT 1600:
    Remove references to non-existant objects(3)
    Add RETURN VALUES to BIO_do_accept page.

RT1818:
    RSA_sign Can return values other than 0 on failure.

RT3634:
    Fix AES CBC aliases (Steffen Nurpmeso <sdaoden@yandex.com>)

RT3678:
    Some clarifications to BIO_new_pair
    (Devchandra L Meetei <dlmeetei@gmail.com>)

RT3787:
    Fix some EVP_ function return values
    (Laetitia Baudoin <lbaudoin@google.com>)

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

GH350: -help text few s_client and s_server flags

add -help description of sigalgs, client_sigalgs, curves
and named_curve

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

Fixed problem with multiple load-unload of comp zlib

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

GH correct organizationalUnitName

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

Clean up reset of read/write sequences

Use sizeof instead of an explicit size, and use the functions for the
purpose.

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

PACKET: add methods for reading length-prefixed TLS vectors.

Rewrite ssl3_get_client_hello to use the new methods.

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

Fix SSLv2-compatible ClientHello processing.

If the client challenge is less than 32 bytes, it is padded with leading - not trailing - zero bytes.

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

PACKET: constify where possible

The PACKET should hold a 'const unsigned char*' underneath as well
but the legacy code passes the record buffer around as 'unsigned char*'
(to callbacks, too) so that's a bigger refactor.

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

GH371: Print debug info for ALPN extension

Also known as RT 4106
Signed-off-by: Rich Salz <rsalz@akamai.com>
Reviewed-by: Matt Caswell <matt@openssl.org>

Remove _locked memory functions.

Undocumented, unused, unnecessary (replaced by secure arena).

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

BN_bin2bn handle leading zero's

If a binary sequence is all zero's, call BN_zero.

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

Add NewSessionTicket test suite

Add a set of tests for checking that NewSessionTicket messages are
behaving as expected.

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

Fix TLSProxy end of test detection

Previously TLSProxy would detect a successful handshake once it saw the
server Finished message. This causes problems with abbreviated handshakes,
or if the client fails to process a message from the last server flight.

This change additionally sends some application data and finishes when the
client sends a CloseNotify.

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

Fix DTLS session ticket renewal

A DTLS client will abort a handshake if the server attempts to renew the
session ticket. This is caused by a state machine discrepancy between DTLS
and TLS discovered during the state machine rewrite work.

The bug can be demonstrated as follows:

Start a DTLS s_server instance:
openssl s_server -dtls

Start a client and obtain a session but no ticket:
openssl s_client -dtls -sess_out session.pem -no_ticket

Now start a client reusing the session, but allow a ticket:
openssl s_client -dtls -sess_in session.pem

The client will abort the handshake.

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

Ignore generated *.S ARM assembly files

Signed-off-by: Rich Salz <rsalz@akamai.com>
Reviewed-by: Emilia Käsper <emilia@openssl.org>

RT4019: Duplicate -hmac flag in dgst.pod

Signed-off-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Emilia Käsper <emilia@openssl.org>

GH372: Remove duplicate flags

Signed-off-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Emilia Käsper <emilia@openssl.org>

Small cleanup of crypto.pod

Came up on the mailing list, from Ken Goldman.

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

GH337: Need backslash before leading #

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

RT4015: Add missing date to CHANGES

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

More test cases.

Add DSA tests.

Add tests to verify signatures against public keys. This will also check
that a public key is read in correctly.

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

Add DSA digest length checks.

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

Fix L<> content in manpages

L<foo|foo> is sub-optimal  If the xref is the same as the title,
which is what we do, then you only need L<foo>.  This fixes all
1457 occurrences in 349 files.  Approximately.  (And pod used to
need both.)

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

Add new types to indent.pro

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

Add new GOST OIDs

Add new OIDs for latest GOST updates

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

Restore previous behaviour of only running one algorithm when -evp alg is used.

Submitted by: Eric Young <eay@pobox.com>
Reviewed-by: Ben Laurie <ben@openssl.org>

restore usage of -elapsed that was disabled in the ifdef reorg

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

GH345: Remove stderr output

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

Move FAQ to the web.

Best hope of keeping current.

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

PACKETise CertificateRequest

Process CertificateRequest messages using the PACKET API

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

PACKETise ClientKeyExchange processing

Use the new PACKET code to process the CKE message

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

PACKETise NewSessionTicket

Process NewSessionTicket messages using the new PACKET API

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