This fixes a bug where some CCS records were written with the wrong TLS
record version.
Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Flush following the CCS after an HRR. Only flush the HRR if middlebox
compat is turned off.
Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Normally we flush immediately after writing the ClientHello. However if
we are going to write a CCS immediately because we've got early_data to
come, then we should move the flush until after the CCS.
Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

The CCS may be sent at different times based on whether or not we
sent an HRR earlier. In order to make that decision this commit
also updates things to make sure we remember whether an HRR was
used or not.
Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Ensure that after an HRR we can only negotiate TLSv1.3
Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Since we no longer do version negotiation during the processing of an HRR
we do not need the TLSv1.3 specific write transition for ClientHello
Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Clients will send a "fake" session id and servers must echo it back.
Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

The new ServerHello format is essentially now the same as the old TLSv1.2
one, but it must additionally include supported_versions. The version
field is fixed at TLSv1.2, and the version negotiation happens solely via
supported_versions.
Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4701)

Reduce RSA_MAX_PRIME_NUM to 5.
Remove no longer used RSA_MIN_PRIME_SIZE.
Make rsa_multip_cap honor RSA_MAX_PRIME_NUM.
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4905)

Reviewed-by: NAndy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4921)

Also improved documentation of TRUSTED_FIRST
Reviewed-by: NMatt Caswell <matt@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>

Fix a typo for "retrieve" and some indentation.
Reviewed-by: NKurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/4919)

Reviewed-by: NBen Kaduk <kaduk@mit.edu>
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4917)

Signed-off-by: NPatrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: NAndy Polyakov <appro@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4916)

If OPENSSL_init_crypto() hasn't been called yet when ERR_get_state()
is called, it need to be called early, so the base initialization is
done.  On some platforms (those who support DSO functionality and
don't define OPENSSL_USE_NODELETE), that includes a call of
ERR_set_mark(), which calls this function again.
Furthermore, we know that ossl_init_thread_start(), which is called
later in ERR_get_state(), calls OPENSSL_init_crypto(0, NULL), except
that's too late.
Here's what happens without an early call of OPENSSL_init_crypto():

    => ERR_get_state():
         => CRYPTO_THREAD_get_local():
         <= NULL;
         # no state is found, so it gets allocated.
         => ossl_init_thread_start():
              => OPENSSL_init_crypto():
                   # Here, base_inited is set to 1
                   # before ERR_set_mark() call
                   => ERR_set_mark():
                        => ERR_get_state():
                             => CRYPTO_THREAD_get_local():
                             <= NULL;
                             # no state is found, so it gets allocated!!!!!
                             => ossl_init_thread_start():
                                  => OPENSSL_init_crypto():
                                       # base_inited is 1,
                                       # so no more init to be done
                                  <= 1
                             <=
                             => CRYPTO_thread_set_local():
                             <=
                        <=
                   <=
              <= 1
         <=
         => CRYPTO_thread_set_local()      # previous value removed!
    <=

Result: double allocation, and we have a leak.

By calling the base OPENSSL_init_crypto() early, we get this instead:

    => ERR_get_state():
         => OPENSSL_init_crypto():
              # Here, base_inited is set to 1
              # before ERR_set_mark() call
              => ERR_set_mark():
                   => ERR_get_state():
                        => OPENSSL_init_crypto():
                             # base_inited is 1,
                             # so no more init to be done
                        <= 1
                        => CRYPTO_THREAD_get_local():
                        <= NULL;
                        # no state is found, so it gets allocated
                        # let's assume we got 0xDEADBEEF
                        => ossl_init_thread_start():
                             => OPENSSL_init_crypto():
                                  # base_inited is 1,
                                  # so no more init to be done
                             <= 1
                        <= 1
                        => CRYPTO_thread_set_local():
                        <=
                   <=
              <=
         <= 1
         => CRYPTO_THREAD_get_local():
         <= 0xDEADBEEF
    <= 0xDEADBEEF

Result: no leak.
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4913)

Reviewed-by: NAndy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4907)

This got lost somehow.  The methods to do makedepend on Windows and
VMS are hard coded for cl (Windows) and CC/DECC (VMS), because that's
what we currently support natively.
Reviewed-by: NAndy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4907)

Reviewed-by: NAndy Polyakov <appro@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4840)

Reviewed-by: NAndy Polyakov <appro@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4840)

Remove some config attributes that just duplicate values that are
already there in other attributes.

Remove the special runs of mkdef.pl and mkrc.pl from build file
templates, as these are now done via GENERATE statements in
build.info.

Remove all references to ordinal files from build file templates, as
these are now treated via the GENERATE statements in build.info.

Also remove -shared flags and similar that are there in shared-info.pl
anyway.  (in the case of darwin, it's mandatory, as -bundle and
-dynamiclib don't mix)
Reviewed-by: NAndy Polyakov <appro@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4840)

Reviewed-by: NAndy Polyakov <appro@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4840)

Because this also includes handling all sorts of non-object files when
linking a program, shared library or DSO, this also includes allowing
general recognition of files such as .res files (compiled from .rc
files), or .def / .map / .opt files (for export and possibly
versioning of public symbols only).

This does mean that there's a tangible change for all build file
templates: they must now recognise and handle the `.o` extension,
which is used internally to recognise object files internally.  This
extension was removed by common.tmpl before this change, but would
mean that the platform specific templates wouldn't know if "foo.map"
was originally "foo.map.o" (i.e. an object file in its own right) or
"foo.map" (an export definition file that should be treated as such,
not as an object file).

For the sake of simplifying things, we also modify util/mkdef.pl to
produce .def (Windows) and .opt (VMS) files that don't need additional
hackery.
Reviewed-by: NAndy Polyakov <appro@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4840)

This makes it possible to add build.info statements for using resource
files as well as linker scripts (.def for Windows, .map for Unix, and
.opt for VMS) is if they were source files.  This requires changes in
the build file templates.
Reviewed-by: NAndy Polyakov <appro@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4840)

This will replace the use of Makefile.shared

This also means a small adjustment on how the attributes dso_cflags,
dso_cxxflags and dso_lflags are treated.  They were previously treated
as an extension to shared_cflag, shared_cxxflag and shared_ldflag, but
they should really be regarded as alternatives instead, for example
for darwin, where -dynamiclib is used for shared libraries and -bundle
for DSOs.

We take the opportunity to clean out things that are redundant or
otherwise superfluous (for example the check of GNU ld on platforms
where it never existed).
Reviewed-by: NAndy Polyakov <appro@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4840)

It will return the last expression from the input file.

We also use this in read_config, which slightly changes what's
expected of Configurations/*.conf.  They do not have to assign
%targets specifically.  On the other hand, the table of configs MUST
be the last expression in each of those files.
Reviewed-by: NAndy Polyakov <appro@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4840)

This commit contains suggestion that (hopefully) improve the
documentation in ssl.pod.
Reviewed-by: NAndy Polyakov <appro@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4914)

The one in rsa.c was overlooked when fixing the same comment in
pkey.c as part of eff1752b.
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4902)

Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4899)