split_send_fragment should always be less than or equal to max_send_fragment

A bug meant that SSL_CTRL_SET_MAX_SEND_FRAGMENT was not adjusting
split_send_fragment properly.

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

Fix BIO_CTRL_DUP for an SSL BIO

The variables in the BIO weren't being duplicated properly.

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

Fix passwd seg fault

Passing the -stdin arg to the passwd command line app *and* supply a
password on the command line causes a seg fault.

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

SPARCv9 assembly pack: fine-tune run-time switch.

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

camellia/build.info: fix typo

RT#4520

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

Remove dead-code infinite loop

Commit d32f5d8733df9938727710d4194e92813c421ef1 added a 'goto end;' statement
at the end of the code block for the 'end' label.  Fortunately, it was after a
return statement, so no infinite loop occurred, but it is still dead code.

Remove the extra goto statement as cleanup.

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

Add ex_data functions for X509_STORE

Add X509_STORE_{set,get}_ex_data() function and
X509_STORE_get_ex_new_index() macro.

X509_STORE has ex_data and the documentation also mentions them but they
are not actually implemented.

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

Ensure we check i2d_X509 return val

The i2d_X509() function can return a negative value on error. Therefore
we should make sure we check it.

Issue reported by Yuan Jochen Kang.

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

*.der files are binary.

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

Remove the never-functional no-sct

It was added as part of 2df84dd3299ff25fa078ca7ffbdeaac65b361feb
but has never actually been used for anything; presumably it was
a typo for one of SCTP or CT.

This removes the last '??' entry from INSTALL.

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

Remove some OPENSSL_NO_SHA that snuck in

Commit d064e6ab52ac8e7b80b2a5d11b31bca583b769c7 removed all the
OPENSSL_NO_SHA guards, but commit
a50ad1daaa68c109ea1a14225a7aba8660526101 regenerated some due to the
sha entries in the %md_disabler table in apps/progs.pl.

Update %md_disabler to reflect that sha is not disableable, and
remove OPENSSL_NO_SHA for good.

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

crypto/poly1305/asm: chase overflow bit on x86 and ARM platforms.

Even though no test could be found to trigger this, paper-n-pencil
estimate suggests that x86 and ARM inner loop lazy reductions can
loose a bit in H4>>*5+H0 step.

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

make update

And recycle some disused slots.

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

Added missing X509_STORE_CTX_set_error_depth() accessor

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

Document the enhancements for DEPEND and INCLUDE and use a better example

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

Build system: add include directories and dependencies for generators

In the case of generating a file like this:

    GENERATE[foo.S]=mkfoo.pl arg1 arg2

the 'mkfoo.pl' generator itself might need to include other files,
such as perl modules within our source tree.  We can reuse already
existing syntax for it, like this:

    INCLUDE[mkfoo.pl]=module/path

or:

    DEPEND[mkfoo.pl]=modules/mymodule.pm

This change implements the support for such constructs, and for the
DEPEND statement, for any value that indicates a perl module (.pm
file), it will automatically infer an INCLUDE statement for its
directory, just like it does for C header files, so you won't have do
write this:

    DEPEND[mkfoo.pl]=modules/mymodule.pm
    INCLUDE[mkfoo.pl]=modules

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

Fix NULL deref in apps/pkcs7

Thanks to Brian Carpenter for finding and reporting this.

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

API compat macros for renamed X509_STORE_CTX functions

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

s390x assembly pack: cache capability query results.

IBM argues that in certain scenarios capability query is really
expensive. At the same time it's asserted that query results can
be safely cached, because disabling CPACF is incompatible with
reboot-free operation.

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

crypto/sparc_arch.h: reserve more SPARCv9 capability bits.

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

chacha/asm/chacha-ppc.pl: get misalignment corner case right on big-endian.

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

Fix EC_KEY_set_private_key() to call key->group->meth->set_private()

Fix a bug introduced by 6903e2e7e9a4 (Extended EC_METHOD customisation
support., 2016-02-01). key->meth->set_private() is wrongly called where
it should call key->group->meth->set_private().

PR#4517

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

add test for CVE-2016-2109

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

In d2i_test return error for malloc failure.

Bad ASN.1 data should never be able to trigger a malloc failure so return
an error in d2i_test if a malloc failure occurs.

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

Harden ASN.1 BIO handling of large amounts of data.

If the ASN.1 BIO is presented with a large length field read it in
chunks of increasing size checking for EOF on each read. This prevents
small files allocating excessive amounts of data.

CVE-2016-2109

Thanks to Brian Carpenter for reporting this issue.

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

Warn when doing an out-of-source build and finding in-source build artifacts

The reason to warn is that configuration *may* pick up on
configuration header files that are in the source tree, that might be
for a wildly different configuration than what is expected in the
current out-of-source configuration.

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

make update

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

Enabled DANE only when at least one TLSA RR was added

It is up to the caller of SSL_dane_tlsa_add() to take appropriate
action when no records are added successfully or adding some records
triggers an internal error (negative return value).

With this change the caller can continue with PKIX if desired when
none of the TLSA records are usable, or take some appropriate action
if DANE is required.

Also fixed the internal ssl_dane_dup() function to properly initialize
the TLSA RR stack in the target SSL handle.  Errors in ssl_dane_dup()
are no longer ignored.

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

Don't set peer_tmp until we have finished constructing it

If we fail halfway through constructing the peer_tmp EVP_PKEY but we have
already stored it in s->s3->peer_tmp then if anything tries to use it then
it will likely fail. This was causing s_client to core dump in the
sslskewith0p test. s_client was trying to print out the connection
parameters that it had negotiated so far. Arguably s_client should not do
that if the connection has failed...but given it is existing functionality
it's easier to fix libssl.

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

Extensions to d2i_test.

Using ASN1_ITEM tables in d2i_test: this then uses consistent names and
makes it easier to extend.

Add bio, reencode and compare tests.

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

Fix capi engine for no-dsa

The capi engine was failing to compile on Windows if the no-dsa option
was selected.

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

Fix no-dsa on Windows/VMS

The no-dsa option was failing on Windows because some symbols were not
correctly flagged in libcrypto.num. Problem found due to the new symbol
consistency test.

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

Fix the indentation of OPENSSL_NO_STDIO in pem.h

Some pre-processor macros were incorrectly indented

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

Fix no-cmac on Windows/VMS

no-cmac was failing on Windows/VMS due to libcrypto.num not marking the
CMAC functions properly. Found due to the new symbol consistency test.

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

Unified copyright for test recipes

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

Add a best effort test to check shared library consistency

Our main development platforms are of the Unix family, which doesn't
have the same strictness regarding a shared library being consistent
with the contents of the ld script (.map file, on Linux and Solaris)
as Windows is with the contents of the .def file or VMS is with the
linker symb_vector option.

To eliminate surprises, we therefore need to make sure to check that
the contents of the .map file is matched with the shared library, at
least to check that the shared library isn't missing any symbols that
should be present.

This test isn't absolutely perfect, as it will only check the symbols
that would be present on Linux / Solaris and will therefore miss those
that would only appear on Windows or VMS.  On the other hand, those
platform specific are few and far apart in time, so it's not likely
they will pose a problem.

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

Fix CRYPTO_clear_realloc() bug.

If allocation in CRYPTO_clear_realloc() fails don't free up the original
buffer: this is consistent with the behaviour of realloc(3) and is expected
in other places in OpenSSL.

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

Update the Configurations READMEs

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

Fix no-ocsp on Windows (and probably VMS)

The ocsp.h file did not have appropriate guards causing link failures on
Windows.

GH Issue 900

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

Remove some unneccessary assignments to argc

openssl.c and ts.c assign the value of opt_num_rest() to argc, but then
only use the value once.

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

Remove some unused argc assignments

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

Fix a missing return value check in v3_addr

All other instances of extract_min_max are checked for an error return,
except this one.

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

Add missing return value check in pkcs8 app

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

Add missing return value checks

Also correct the return value from the the "prime" application

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

Fix missing break in option parsing

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

Remove some dead code

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

Fix some code maintenance issues

Various instances of variables being written to, but then never read.

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

BIO socket connect failure was not handled correctly.
The state was always set to BIO_CONN_S_OK.

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

Fix missing IDEA renames (windows build)

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

Copyright consolidation script

With Richard Levitte.

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

Travis: _srcdist, not _srcdir

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

Fix no-ui on Windows

Ensure public functions have appropriate guards in header files.

GH Issue 899

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

Copyright consolidation; .pm and Configure

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

Update copyright; generated files.

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

Remove --classic build entirely

The Unix build was the last to retain the classic build scheme.  The
new unified scheme has matured enough, even though some details may
need polishing.

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

Fix no-sock on Windows

Link errors were occurring on Windows because the header files were not
correctly guarding some functions with OPENSSL_NO_SOCK

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

Include winsock2.h even if compiling no-sock

We need the struct timeval definition from winsock2.h even if we're not
going to call any socket functions.

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

Don't use select on Windows

Windows "select" only works for sockets so don't use it to wait for async.

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

Cascade no-dgram from no-sock in Configure not e_os.h

e_os.h was defining OPENSSL_NO_DGRAM if OPENSSL_NO_SOCK was defined.
This causes link problems on Windows because the generated .def files
still contain the DGRAM symbols even though they have not been compiled.

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

Fix no-dgram on Windows

Link errors were occurring on Windows because the header files were not
correctly guarding some functions with OPENSSL_NO_DGRAM

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

Copyright consolidation: perl files

Add copyright to most .pl files
This does NOT cover any .pl file that has other copyright in it.
Most of those are Andy's but some are public domain.
Fix typo's in some existing files.

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

Fix warnings installing pod files

Fixes some links in the pod files

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

Configurations: fix typo in 50-masm.conf.

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

evp/aes_aes.c: engage Fujitsu SPARC64 X AES support.

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

sparcv9cap.c: add Fujitsu SPARC64 X AES capability detection.

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

Add AES assembly module for Fujitsu SPARC64 X/X+.

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

Configuration: add unsupported masm target.

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

perlasm/x86_64-xlate.pl: make latest ml64 work.

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

poly1305/asm/poly1305-x86_64.pl: not all assemblers manage << in constants.

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

Typo, asoutflag -> perlasm_scheme

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

Only allow Microsoft assembler with no-asm on Windows

This also restores the possibility to have ml used with VC-WIN32 with
no-asm, which was lost during the mk1mf -> unified transition.

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

Build: Make 'test' depend on 'tests'

Because we have a directory 'test', the target 'test' may be confusing
to make.  However, if we make it depend on 'tests', which doesn't
exist, make should never fail to run the actions.

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

Prepare for 1.1.0-pre6-dev

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

Prepare for 1.1.0-pre5 release

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

Unsigned chars can't be negative

Fix a problem where an unsigned char was being checked to see if it was
negative.

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

Fix PKCS7 memory leak

Commit f0e0fd51f was a bit over-zealous in removing a call to
X509_STORE_CTX_cleanup(). The call in question was in a loop and was
required to cleanup resources used on each iteration of the loop. Removing
this resulted in a memory leak.

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

Make string_to_hex/hex_to_string public

Give the API new names, document it.

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

Rename some lowercase API's

Make OBJ_name_cmp internal
Rename idea_xxx to IDEA_xxx
Rename get_rfc_xxx to BN_get_rfc_xxx
Rename v3_addr and v3_asid functions to X509v3_...

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

NOTES files will contain more requirements, make that explicit

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

Haiku x86-64 doesn't need RC4_CHAR.

* add a comment about Haiku being built with no-asm.

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

Free methods on destroy.

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

Free engine on error.

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

Opacity.

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

Signed/unsigned compare.

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

Restore OCSP_basic_verify() error return semantics

Recently, OCSP_basic_verify() was changed to always return 0 on error,
when it would previously return 0 on error and < 0 on fatal error.
This restores the previous semantics back.

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

Remove declaration of function that isn't implemented

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

Add X509_STORE_CTX_set0_untrusted function.

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

Small OCSP fixup

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

Make many X509_xxx types opaque.

Make X509_OBJECT, X509_STORE_CTX, X509_STORE, X509_LOOKUP,
and X509_LOOKUP_METHOD opaque.
Remove unused X509_CERT_FILE_CTX

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

Don't use a default for --with-zlib-lib on Windows with option 'zlib'

To begin with, the default should have been the import library, not
the DLL itself.  However, we don't know what directory it's installed
in either way, so we may as well demand the full path from the user

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

ex_libs settings have to be added to, not overriden, on Windows too

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

Update Travis and Appveyor for change to shared by default

Update the Travis and Appveyor builds to explicitly state no-shared where
applicable.

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

Make "shared" builds the default

In most cases we expect that people will be using shared libraries not
static ones, therefore we make that the default.

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

Fix ex_data locks issue

Travis identified a problem with freeing the ex_data locks which wasn't
quite right in ff2344052. Trying to fix it identified a further problem:
the ex_data locks are cleaned up by OPENSSL_cleanup(), which is called
explicitly by CRYPTO_mem_leaks(), but then later the BIO passed to
CRYPTO_mem_leaks() is freed. An attempt is then made to use the ex_data
lock already freed.

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

Add Haiku support.

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

Fix installation on VMS for no-stdio and no-autoalginit

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

Fix installation for no-stdio and no-autoalginit

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

Fix no-stdio and no-autoalginit

no-stdio does not work with the apps. Since the tests also need the apps
it doesn't support that either. Therefore we disable building of both.

no-autoalginit is not compatible with the apps because it requires explicit
loading of the algorithms, and the apps don't do that. Therefore we disable
building the apps for this option. Similarly the tests depend on the apps
so we also disable the tests. Finally the whole point about no-autoalginit
is to avoid excessive executable sizes when doing static linking. Therefore
we disable "shared" if this option is selected.

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

The err_cleanup() funtion is internal so shouldn't be exported

Running a "make update" wanted to add err_cleanup to libcrypto.num which
is wrong.

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

Write POD page.

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