Do not set load_crypto_strings_inited when OPENSSL_NO_ERR is defined

Only set the load_crypto_strings_inited to 1 when err_load_crypto_strings_int was called.

This solves the following issue:
- openssl is built with no-err
- load_crypto_strings_inited is set to 1 during the OPENSSL_init_crypto call
- During the cleanup: OPENSSL_cleanup, err_free_strings_int is called because load_crypto_strings_inited == 1
- err_free_strings_int calls do_err_strings_init because it has never been called
- Now do_err_strings_init calls OPENSSL_init_crypto
- But since we are in the cleanup (stopped == 1) this results in an error:
  CRYPTOerr(CRYPTO_F_OPENSSL_INIT_CRYPTO, ERR_R_INIT_FAIL);
- which then tries to initialize everything we are trying to clean up: ERR_get_state, ossl_init_thread_start, etc
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1654)

Add error checking, small nit on ouput

Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1632)

Fix config option 'no-deprecated'

crypto/asn1/asn1_item_list.c needed including dh.h and rsa.h directly.
The reason is that they are not included by x509.h when configured
'no-deprecated'

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1741)

Add missing .pod extension to EVP_PKEY_CTX_set_tls1_prf_md

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

Fix not-c code

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1742)

Fix strict-warnings build

crypto/s390xcap.c: internal/cryptlib.h needs to be included for
OPENSSL_cpuid_setup function prototype is located there to avoid
build error due to -Werror=missing-prototypes.

Signed-off-by: Patrick Steuer <psteuer@mail.de>
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
CLA: trivial

Fix strict-warnings build

crypto/evp/e_aes.c: Types of inp and out parameters of
AES_xts_en/decrypt functions need to be changed from char to
unsigned char to avoid build error due to
'-Werror=incompatible-pointer-types'.

crypto/aes/asm/aes-s390x.pl: Comments need to reflect the above
change.

Signed-off-by: Patrick Steuer <psteuer@mail.de>
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
CLA: trivial

Fix strict-warnings build

crypto/asn1/a_strex.c: Type of width variable in asn1_valid_host
function  needs to be changed from char to signed char to avoid
build error due to '-Werror=type-limits'.

Signed-off-by: Patrick Steuer <psteuer@mail.de>
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
CLA: trivial

Add SSL_OP_NO_ENCRYPT_THEN_MAC

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

Fix encrypt-then-mac implementation for DTLS

OpenSSL 1.1.0 will negotiate EtM on DTLS but will then not actually *do* it.

If we use DTLSv1.2 that will hopefully be harmless since we'll tend to use
an AEAD ciphersuite anyway. But if we're using DTLSv1, then we certainly
will end up using CBC, so EtM is relevant — and we fail to interoperate with
anything that implements EtM correctly.

Fixing it in HEAD and 1.1.0c will mean that 1.1.0[ab] are incompatible with
1.1.0c+... for the limited case of non-AEAD ciphers, where they're *already*
incompatible with other implementations due to this bug anyway. That seems
reasonable enough, so let's do it. The only alternative is just to turn it
off for ever... which *still* leaves 1.0.0[ab] failing to communicate with
non-OpenSSL implementations anyway.

Tested against itself as well as against GnuTLS both with and without EtM.

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

Add Postgres support to -starttls

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

Fix embedded string handling.

Don't rely on embedded flag to free strings correctly: it wont be
set if there is a malloc failure during initialisation.

Thanks to Guido Vranken for reporting this issue.

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1725)

fix invalid use of incomplete type X509_STORE_CTX

CLA: trivial

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

Allow older versions in the *.num files

In 1.1.0 we only allowed a strictly increasing version number in the *.num
files, i.e. you could never introduce a symbol at the end of the *.num file
with a lower version number than the one preceding it. This made sense for
1.1.0. However in master we may be introducing symbols for backport to
1.1.0. Therefore it is ok in master to have a symbol for version 1.1.0c
coming after a symbol for version 1.1.1.

This commit fixes the check in mkdef.pl to be a bit looser to allow this.

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

Fix signatures of EVP_Digest{Sign,Verify}Update

These are implemented as macros delegating to `EVP_DigestUpdate`, which
takes a `size_t` as its third argument, not an `unsigned int`.

CLA: trivial

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

Ensure we handle len == 0 in ERR_err_string_n

If len == 0 in a call to ERR_error_string_n() then we can read beyond the
end of the buffer. Really applications should not be calling this function
with len == 0, but we shouldn't be letting it through either!

Thanks to Agostino Sarubbo for reporting this issue. Agostino's blog on
this issue is available here:
https://blogs.gentoo.org/ago/2016/10/14/openssl-libcrypto-stack-based-buffer-overflow-in-err_error_string_n-err-c/

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

Use clang++ for C++ for the linux-x86_64-clang target

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

Engine afalg: properly set operation type also on big endian.

Copy the whole ALG_OP_TYPE to CMSG_DATA.

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

Fix style issue

Reviewed-by: Kurt Roeckx <kurt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1694)

Constify command options

Reviewed-by: Kurt Roeckx <kurt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1694)

Fix C++ support: set $target{cxx} correctly

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

Fix typo

I think the second "VC-WIN32" should be "VC-WIN64".
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
CLA: trivial

Configure: remove superfluous 0x

The number is taken from the OPENSSL_VERSION_NUMBER which is already
in the hex form.

CLA: trivial

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1706)

Add memory leak detection to d2i_test

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1707)

Fix copy-paste test labels

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

Remove automatic RPATH - Add a CHANGES entry

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

Remove automatic RPATH - adapt shlib_wrap.sh

Looking for something starting with '-Wl,-rpath,' isn't good enough,
as someone might give something like '-Wl,--enable-new-dtags,-rpath,/PATH'.
Looking for ',-rpath,' should be safe enough.

We could remove the preloading stuff entirely, but just in case the
user has chosen to given RPATH setting arguments at configuration,
we'd better make sure testing will still work.  Fair warning, there
are some configuration options that do not work with preloaded OpenSSL
libraries, such as the sanity checking ones.

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

Remove automatic RPATH - add user rpath support

Make Configure recognise -rpath and -R to support user added rpaths
for OSF1 and Solaris.  For convenience, add a variable LIBRPATH in the
Unix Makefile, which the users can use as follows:

    ./config [options] -Wl,-rpath,\$(LIBRPATH)

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

Remove automatic RPATH

Before OpenSSL 1.1.0, binaries were installed in a non-standard
location by default, and runpath directories were therefore added in
those binaries, to make sure the executables would be able to find the
shared libraries they were linked with.

With OpenSSL 1.1.0 and on, binaries are installed in standard
directories by default, and the addition of runpath directories is
therefore not needed any more.

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

RT is put out to pasture

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1702)

Add C++ settings in the Linux config targets

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

Add support for C++ in Configurations/unix-Makefile.tmpl

Note that it relies on a trick from Configure, where file names for
object files made from C++ source get '.cc' replaced with '_cc.o' to
recognise them.  This is needed so the correct compiler is used when
linking binaries.

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

Add support for C++ in Configure

A note: this will form object file names by changing '.cc' to
'_cc.o'.  This will permit other configuration code to recognise these
object files were built for C++ rather than C.

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

Fix X509_NAME decode for malloc failures.

The original X509_NAME decode free code was buggy: this
could result in double free or leaks if a malloc failure
occurred.

Simplify and fix the logic.

Thanks to Guido Vranken for reporting this issue.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1691)

doc/crypto/OPENSSL_ia32cap.pod: update assembler requirements.

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

.travis.yml: minor optimizations:

- make 'pip install --user cpp-coveralls' conditional;
- limit no-stdio to single build per operating environment;
- omit enable-asan duplicate;

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

.travis.yml: install wine on demand.

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

Fix up bn_prime.pl formatting.

Align at 5 characters, not 4. There are 5-digit numbers in the output.
Also avoid emitting an extra blank line and trailing whitespace.

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

Remove trailing whitespace from some files.

The prevailing style seems to not have trailing whitespace, but a few
lines do. This is mostly in the perlasm files, but a few C files got
them after the reformat. This is the result of:

  find . -name '*.pl' | xargs sed -E -i '' -e 's/( |'$'\t'')*$//'
  find . -name '*.c' | xargs sed -E -i '' -e 's/( |'$'\t'')*$//'
  find . -name '*.h' | xargs sed -E -i '' -e 's/( |'$'\t'')*$//'

Then bn_prime.h was excluded since this is a generated file.

Note mkerr.pl has some changes in a heredoc for some help output, but
other lines there lack trailing whitespace too.

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

Add some missing types to indent.pro

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

Update fuzz corpora

New minimal fuzz corpora for asn1, asn1parse, bndiv, crl and x509

Reviewed-by: Andy Polyakov <appro@openssl.org>
GH: #1678

Fix linebreaks in the tls_construct_client_certificate function

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

Add a typedef for the construction function

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

Move setting of the handshake header up one more level

We now set the handshake header, and close the packet directly in the
write_state_machine. This is now possible because it is common for all
messages.

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

Remove the special case processing for finished construction

tls_construct_finished() used to have different arguments to all of the
other construction functions. It doesn't anymore, so there is no neeed to
treat it as a special case.

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

Harmonise setting the header and closing construction

Ensure all message types work the same way including CCS so that the state
machine doesn't need to know about special cases. Put all the special logic
into ssl_set_handshake_header() and ssl_close_construct_packet().

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

Don't set the handshake header in every message

Move setting the handshake header up a level into the state machine code
in order to reduce boilerplate.

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

Move init of the WPACKET into write_state_machine()

Instead of initialising, finishing and cleaning up the WPACKET in every
message construction function, we should do it once in
write_state_machine().

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

Remove untrue comment.

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

Make dependencies if Makefile is new.

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

Rename ssl_set_handshake_header2()

ssl_set_handshake_header2() was only ever a temporary name while we had
to have ssl_set_handshake_header() for code that hadn't been converted to
WPACKET yet. No code remains that needed that so we can rename it.

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

Remove ssl_set_handshake_header()

Remove the old ssl_set_handshake_header() implementations. Later we will
rename ssl_set_handshake_header2() to ssl_set_handshake_header().

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

Remove the tls12_get_sigandhash_old() function

This is no longer needed now that all messages use WPACKET

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

fix memory leak

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

Don't use DES when disabled.

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

fix memory leak

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

Add SRP test vectors from RFC5054

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

SRP code tidy.

Tidy up srp_Calc_k and SRP_Calc_u by making them a special case of
srp_Calc_xy which performs SHA1(PAD(x) | PAD(y)).

This addresses an OCAP Audit issue.

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

Convert NewSessionTicket construction to WPACKET

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

Fix an error in packet_locl.h

A convenience macro was using the wrong underlying function.

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

Convert CertStatus message construction to WPACKET

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

Fix mis-named macro in packet_locl.h

A couple of the WPACKET_sub_memcpy* macros were mis-named.

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

Convert SeverDone construction to WPACKET

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

make update

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

Print <ABSENT> if a STACK is NULL.

If a STACK (corresponding to SEQUENCE OF or SET OF) is NULL then the
field is absent as opposed to empty (present but has zero elements).

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

add item list support to d2i_test

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

ASN1_ITEM should use type name not structure name.

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

Add -item option to asn1parse

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

Add ASN1_ITEM lookup and enumerate functions.

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

Fix missing NULL checks in NewSessionTicket construction

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

Fix an mis-matched function code so that "make update" doesn't fail

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

Add an example of usage to the WPACKET_reserve_bytes() documentation

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

Address style feedback comments

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

Fix a bug in CKE construction for PSK

In plain PSK we don't need to do anymore construction after the preamble.
We weren't detecting this case and treating it as an unknown cipher.

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

Convert ServerKeyExchange construction to WPACKET

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

Add the WPACKET_reserve_bytes() function

WPACKET_allocate_bytes() requires you to know the size of the data you
are allocating for, before you create it. Sometimes this isn't the case,
for example we know the maximum size that a signature will be before we
create it, but not the actual size. WPACKET_reserve_bytes() enables us to
reserve bytes in the WPACKET, but not count them as written yet. We then
subsequently need to acall WPACKET_allocate_bytes to actually count them as
written.

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

Remove tls12_copy_sigalgs_old()

This was a temporary function needed during the conversion to WPACKET. All
callers have now been converted to the new way of doing this so this
function is no longer required.

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

Convert CertificateRequest construction to WPACKET

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

Address style feedback comments

Merge declarations of same type together.

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

Fix a bug in the construction of the ClienHello SRTP extension

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

Fix heartbeat compilation error

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

Delete some unneeded code

Some functions were being called from both code that used WPACKETs and code
that did not. Now that more code has been converted to use WPACKETs some of
that duplication can be removed.

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

Convert ServerHello construction to WPACKET

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

Fix an Uninit read in DTLS

If we have a handshake fragment waiting then dtls1_read_bytes() was not
correctly setting the value of recvd_type, leading to an uninit read.

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

Fix no-dtls

The new large message test in sslapitest needs OPENSSL_NO_DTLS guards

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

apps/apps.c: initialize and de-initialize engine around key loading

Before loading a key from an engine, it may need to be initialized.
When done loading the key, we must de-initialize the engine.
(if the engine is already initialized somehow, only the reference
counter will be incremented then decremented)

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

Revert "Call ENGINE_init() before trying to use keys from engine"

This reverts commit 0a72002993b8619fd0642d19af3364bafbd9a06c.
This fails to call ENGINE_finish; an alternate fix is forthcoming.

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

Call ENGINE_init() before trying to use keys from engine

When I said before that s_client "used to work in 1.0.2" that was only
partly true. It worked for engines which provided a default generic
method for some key type, because it called ENGINE_set_default() and
that ended up being an implicit initialisation and functional refcount.

But an engine which doesn't provide generic methods doesn't get initialised,
and then when you try to use it you get an error:

cannot load client certificate private key file from engine
140688147056384:error:26096075:engine routines:ENGINE_load_private_key:not initialised:crypto/engine/eng_pkey.c:66:
unable to load client certificate private key file

cf. https://github.com/OpenSC/libp11/issues/107 (in which we discover
that engine_pkcs11 *used* to provide generic methods that OpenSSL would
try to use for ephemeral DH keys when negotiating ECDHE cipher suites in
TLS, and that didn't work out very well.)

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1639)

Restore '-keyform engine' support for s_client

This used to work in 1.0.2 but disappeared when the argument parsing was
revamped.

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1639)

Add DTLS renegotiation tests

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

Extend the renegotiation tests

Add the ability to test both server initiated and client initiated reneg.

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

Update README.ssltest.md

Add update for testing renegotiation. Also change info on CTLOG_FILE
environment variable - which always seems to be required.

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

Add support for testing renegotiation

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

Fix a bug in Renegotiation extension construction

The conversion to WPACKET broke the construction of the renegotiation
extension.

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

Convert HelloRequest construction to WPACKET

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

Add missing parameter.

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

Switch back to assuming TLS 1.2.

The TLSProxy::Record->new call hard-codes a version, like
70-test_sslrecords.t.

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

Address review comments.

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

Don't test quite so many of them.

Avoid making the CI blow up.

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

Test CBC mode padding.

This is a regression test for
https://github.com/openssl/openssl/pull/1431. It tests a
maximally-padded record with each possible invalid offset.

This required fixing a bug in Message.pm where the client sending a
fatal alert followed by close_notify was still treated as success.

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