check_chain_extensions(): Add check that Basic Constraints of CA cert are marked critical

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12478)

Extend X509 cert checks and error reporting in v3_{purp,crld}.c and x509_{set,vfy}.c

add various checks for malformedness to static check_chain_extensions() in x509_vfc.c
improve error reporting of X509v3_cache_extensions() in v3_purp.c
add error reporting to x509_init_sig_info() in x509_set.c
improve static setup_dp() and related functions in v3_purp.c and v3_crld.c
add test case for non-conforming cert from https://tools.ietf.org/html/rfc8410#section-10.2

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12478)

apps/cmp.c: Improve safeguard assertion on consistency of cmp_options[] and cmp_vars[]

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12836)

apps_ui.c: Correct password prompt for ui_method

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12493)

apps_ui.c: Correct handling of empty password from -passin

This is done in analogy to commit ca3245a61989009a99931748723d12e30d0a66b2

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12493)

apps_ui.c: Improve error handling and return value of setup_ui_method()

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12493)

Fix fipsinstall module path

If a path is specified with the -module option it will use this path to load the library when the provider is activated,
instead of also having to set the environment variable OPENSSL_MODULES.

Added a platform specific opt_path_end() function that uses existing functionality used by opt_progname().

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12761)

STORE: Fix OSSL_STORE_attach() to check |ui_method| before use

ossl_pw_set_ui_method() demands that the passed |ui_method| be
non-NULL, and OSSL_STORE_attach() didn't check it beforehand.

While we're at it, we remove the passphrase caching that's set at the
library level, and trust the implementations to deal with that on
their own as needed.

Fixes #12830

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12831)

Add/harmonize multi-valued RDN support and doc of ca, cmp, req, storeutl, and x509 apps

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12769)

X509_NAME_cmp(): Clearly document its semantics, referencing relevant RFCs

Fixes #12765

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12769)

X509_NAME_add_entry_by_txt.pod: Improve documentation w.r.t. multi-valued RDNs (containing sets of AVAs)

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12769)

X509_NAME_cmp: restrict normal return values to {-1,0,1} to avoid confusion with -2 for error

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12769)

X509_NAME_oneline(): Fix output of multi-valued RDNs, escaping '/' and '+' in values

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12769)

X509_NAME_print_ex.pod: re-format lines to fit within 80 chars limit

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12769)

app_load_config_bio(): fix crash on error

It turns out that the CONF_modules_load(conf, NULL, 0) call is just wrong.

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12817)

Fix an EVP_MD_CTX leak

If we initialise an EVP_MD_CTX with a legacy MD, and then reuse the same
EVP_MD_CTX with a provided MD then we end up leaking the md_data.

We need to ensure we free the md_data if we change to a provided MD.

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12779)

Diverse build.info: Adjust paths

Fixes #12815

Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/12816)

bugfix in apps/cmp.c and cmp_client.c: inconsistencies on retrieving extraCerts in code and doc

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12822)

bugfix in ossl_cmp_msg_protect(): set senderKID and extend extraCerts also for unprotected CMP requests

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12822)

bugfix in ossl_cmp_msg_add_extraCerts(): should include cert chain when using PBM

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12822)

test/cmp_{client,msg}_test.c: minor code cleanup

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12655)

test/recipes/81-test_cmp_cli_data/Mock/server.cnf: minor cleanup

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12655)

81-test_cmp_cli: Make test output files all different according to #11080

Also some minor improvements mostly of test cases regarding PKCS#10 CSR input

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12655)

81-test_cmp_cli.t: Stop unlinking test output files according to #11080

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12655)

apps.c: Fix mem leaks on error in load_certs() and load_crls()

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12823)

apps/cmp.c: clear leftover errors on loading libengines.so etc.

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12824)

apps.c: Fix diagnostics and return value of load_key_certs_crls() on error

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12824)

Replace all wrong usages of 'B<...>' (typically by 'I<...>') in OSSL_CMP_CTX_new.pod

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12741)

Clean up CMP chain building for CMP signer, TLS client, and newly enrolled certs

* Use strenghtened cert chain building, verifying chain using optional trust store
  while making sure that no certificate status (e.g., CRL) checks are done
* Use OSSL_CMP_certConf_cb() by default and move its doc to OSSL_CMP_CTX_new.pod
* Simplify certificate and cert store loading in apps/cmp.c

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12741)

Fix markdown nits in NOTES-Windows.txt

And add a comment that this file is in markdown, but has a .txt
extension on purpose.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12805)

Support writing RSA keys using the traditional format again

Fixes: #6855
Reviewed-by: Richard Levitte <levitte@openssl.org>
GH: #8743

ENCODER: Refactor provider implementations, and some cleanup

The encoder implementations were implemented by unnecessarily copying
code into numerous topical source files, making them hard to maintain.
This changes merges all those into two source files, one that encodes
into DER and PEM, the other to text.

Diverse small cleanups are included.

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12803)

Fix up issue on AIX caused by broken compiler handling of macro expansion

Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12812)

s_time: check return values better

Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/12808)

In a non-shared build, don't include the md5 object files in legacy provider

Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/11961)

TLS fixes for CBC mode and no-deprecated

Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/11961)

TLS: remove legacy code path supporting special CBC mode

Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/11961)

legacy: include MD5 code in legacy provider

Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/11961)

Deprecate SHA and MD5 again.

This reverts commit a978dc3bffb63e6bfc40fe6955e8798bdffb4e7e.

Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/11961)

Allow unauthenticated CMP server if missing -trusted, -srvcert, and -secret options

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12806)

Add 4 new OIDs for PKIX key purposes and 3 new CMP information types

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12807)

TEST: modify test/endecode_test.c to not use legacy keys

Now that PEM_write_bio_PrivateKey_traditional() can handle
provider-native EVP_PKEYs, we don't need to use explicitly legacy
ones.

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12738)

PEM: Make PEM_write_bio_PrivateKey_traditional() handle provider-native keys

PEM_write_bio_PrivateKey_traditional() didn't handle provider-native
keys very well.  Originally, it would simply use the corresponding
encoder, which is likely to output modern PEM (not "traditional").

PEM_write_bio_PrivateKey_traditional() is now changed to try and get a
legacy copy of the input EVP_PKEY, and use that copy for traditional
output, if it has such support.

Internally, evp_pkey_copy_downgraded() is added, to be used when
evp_pkey_downgrade() is too intrusive for what it's needed for.

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12738)

Add CMS AuthEnvelopedData with AES-GCM support

Add the AuthEnvelopedData as defined in RFC 5083 with AES-GCM
parameter as defined in RFC 5084.

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/8024)

apps/cmp.c: Allow default HTTP path (aka CMP alias) given with -server option

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12786)

apps/cmp.c: Use enhanced OSSL_HTTP_parse_url(), removing parse_addr() and atoint()

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12786)

OSSL_HTTP_parse_url(): add optional port number return parameter and strengthen documentation

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12786)

EVP: Move the functions and controls for setting and getting distid

Those functions were located in the EC files, but is really broader
than that, even thought currently only used for SM2.  They should
therefore be in a more central location, which was also indicated by
diverse TODOs.

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12789)

EVP: Expand the use of EVP_PKEY_CTX_md()

Setting a hash function was reserved for signature operations.
However, it turns out that SM2 uses a hash function for encryption and
decryption as well.
Therefore, EVP_PKEY_CTX_md() must be called with an expanded operation
type combination that includes EVP_PKEY_OP_TYPE_CRYPT when used in a
generic way.

For SM2, test/recipes/30-test_evp_data/evppkey_sm2.txt is expanded to
test decryption both with an implicit and an explicit digest.

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12789)

EVP: Add support for delayed EVP_PKEY operation parameters

They get called "delayed parameters" because they may make it to the
implementation at a later time than when they're given.

This currently only covers the distinguished ID, as that's the only
EVP_PKEY operation parameter so far that has been possible to give
before the operation has been initialized.

This includes a re-implementation of EVP_PKEY_CTX_set1_id(),
EVP_PKEY_CTX_get1_id(), and EVP_PKEY_CTX_get1_id_len().

Also, the more rigorous controls of keytype and optype are restored.

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12789)

New GOST PKCS12 standard support

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12780)

Fix test/evp_extra_test.c

Because EVP_PKEY_CTX_new_from_name() could return a non-NULL context
with no value in it, the lack of legacy implementation when OpenSSL
was configured with 'no-ec' went through undetected.  This adds the
necessary guards to skip a test of SM2 in that case.

Reviewed-by: Paul Yang <kaishen.yy@antfin.com>
(Merged from https://github.com/openssl/openssl/pull/12785)

EVP: Don't shadow EVP_PKEY_CTX_new* error records

There are places that add an ERR_R_MALLOC_FAILURE record when any of
EVP_PKEY_CTX_new*() return NULL, which is 1) inaccurate, and 2)
shadows the more accurate error record generated when trying to create
the EVP_PKEY_CTX.

Reviewed-by: Paul Yang <kaishen.yy@antfin.com>
(Merged from https://github.com/openssl/openssl/pull/12785)

EVP: Preserve the EVP_PKEY id in a few more spots

As long as there are internal legacy keys for EVP_PKEY, we need to preserve
the EVP_PKEY numeric identity when generating a key, and when creating the
EVP_PKEY_CTX.

For added consistency, the EVP_PKEY_CTX contructor tries a little
harder to find a EVP_PKEY_METHOD.  Otherwise, we may run into
situations where the EVP_PKEY_CTX ends up having no associated methods
at all.

Reviewed-by: Paul Yang <kaishen.yy@antfin.com>
(Merged from https://github.com/openssl/openssl/pull/12785)

Use return code for 'which command' checks

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12775)

Fix memory leaks in conf_def.c

Fixes #12471
CLA: trivial

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12533)

Building: Build Unix static libraries one object file at a time

We're hitting problems that the 'ar' command line becomes too long for
some 'make' versions, or the shell it uses.

We therefore change the way we create a static library by doing so one
object file at a time.  This is slower, but has better guarantees to
work properly on limited systems.

Fixes #12116

Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/12706)

DOC: Fix check of EVP_PKEY_fromdata{,_init} in examples

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12389)

DOC: Modify one example in EVP_PKEY_fromdata(3)

The example to create an EC key from user data didn't show what one
could expect and application to do, especially with regard to how it's
done with raw EC functions.  We therefore refactor it to make proper
use of a BIGNUM where expected, and also use OSSL_PARAM_BLD(3) for
easier handling of the OSSL_PARAM array.

Fixes #12388

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12389)

Cleanup deprecation of ENGINE_setup_bsd_cryptodev

CLA: trivial

Reviewed-by: Ben Kaduk <kaduk@mit.edu>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12793)

Slightly abstract ktls_start() to reduce OS-specific #ifdefs.

Instead of passing the length in from the caller, compute the length
to pass to setsockopt() inside of ktls_start().  This isolates the
OS-specific behavior to ktls.h and removes it from the socket BIO
implementations.

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/12782)

Remove unused dummy functions from ktls.h.

The KTLS functions are always used under #ifndef OPENSSL_NO_KTLS, so
the dummy functions were never used.

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/12782)

Fix the socket BIO control methods to use ktls_crypto_info_t.

This is mostly a cosmetic cleanup I missed when adding the
ktls_crypto_info_t type.  However, while fixing this I noticed that
the changes to extract the size from crypto_info from the wrapper
structure for Linux KTLS had not been propagated from bss_sock.c to
bss_conn.c, so I've fixed that to use the correct length.

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/12782)

X509_STORE_CTX_print_verify_cb(): add AKID and SKID output for (non-)trusted certs

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12787)

OSSL_CMP_CTX: rename field and its getter/setter from 'untrusted_certs' to 'untrusted

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12788)

Strengthen chain building for CMP

* Add -own_trusted option to CMP app
* Add OSSL_CMP_CTX_build_cert_chain()
* Add optional trust store arg to ossl_cmp_build_cert_chain()
* Extend the tests in cmp_protect_test.c and the documentation accordingly

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12791)

Add OSSL_CMP_CTX_get1_newChain() and related CLI option -chainout

Also simplify certificate saving in apps/cmp.c

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12790)

test/drbgtest: improve the reseed after fork test

Issue #12377 demonstrated that it is not sufficient to verify that
after a fork a reseeding is triggered in the child. This commit
enhances the test by collecting the output of the public and private
drbg for the parent and all children and checking for duplicates.
In case of duplicates, it prints an error message and displays a
sorted output.

The analysis of #12377 (see [1]) showed that due to an error in the
resetting of the AES-CTR (issue #12405, fixed by #12413), it could
happen that only the first n bytes (n=1,...15) of the children's
random output were identical.

This test is optimized to detect this issue by only comparing the first
byte of the sampled data (i.e., the first 'column' of the output).
The number of samples is chosen high enough to keep the chance of false
positives low. The test is executed sixteen times, each time advancing
the internal counter by requesting a single extra byte of random data.

Another, more general test splits the entire sampled random data into
two-byte chunks and counts their collisions. If a certain threshold
is exceeded, it reports an error.

[1] https://github.com/openssl/openssl/issues/12377#issuecomment-656207334

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12407)

Fix coverity CID #1454815 - NULL ptr dereference in initthread.c

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/12708)

Fix coverity CID #1452769 & #1452771 - Arg passed to function that cannot be negative in cms_ess.c

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/12708)

Fix coverity CID #1457935 - Check return value in ffc_params.c for BIO_indent/BIO_puts calls.

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/12708)

Fix coverity CID #1465967 & #1465968 - fix NULL dereference in dh_ameth.c

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/12708)

Fix coverity CID #1466371 - fix dereference before NULL check.

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/12708)

Fix coverity CID #1466375 - Remove dead code.

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/12708)

Fix coverity CID #1466377 - resource leak due to early return in ec_get_params().

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/12708)

Fix coverity CID #1466378 - Incorrect expression in ec_backend.c

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/12708)

ASN1: Make ASN1_item_verify_ctx() work with provider-native keys

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

EC: Remove one error record that shadows another

In EC_GROUP_new_from_params(), ERR_R_EC_LIB was reported if
group_new_from_name() returned NULL.  However, this shadows a possible
EC_R_INVALID_CURVE, making that harder to detect, which happens to be
important to do in test/evp_test.c.

This also extends key_unsupported() in test/evp_test.c to check for
this error alongside the check for EC_R_UNKNOWN_GROUP.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

TEST: Ensure that the base provider i activated when needed

The fips providers can't be activated alone if encoding, decoding or
STORE are going to be used.

To enable this, we selectively use test/fips-and-base.cnf instead of
test/fips.cnf in our test recipes.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

Revert "TEST: separate out NIST ECC tests from non-NIST"

This file split turned out to be a mistake as soon as the fetching
error reporting got properly sorted.

This reverts commit e6ed04a9dcc2ead94e35c4a7400b9c998b5ad9ac.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

EVP: Don't report malloc failure in new_raw_key_int()

On failure by EVP_PKEY_CTX_new_from_name(), this function reported
ERR_R_MALLOC_FAILURE.  However, that's not necessarily true, as it can
fail because the algorithm isn't present.

Either way, EVP_PKEY_CTX_new_from_name() records more accurate errors
on its own, and one of them - EVP_R_FETCH_FAILED - is significant for
test/evp_test.c.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

TEST: have key_unsupported() in evp_test.c look at the last error

key_unsupported() looked at the first error in the queue to see if a
key algorithm is supported or not.  However, there are situations
where the errors it looks for is preceded by others.  It's much safer
to look at the last recorded error.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

STORE: Stop the flood of errors

The old 'file:' loader was recently changed to stop the flood of
repeated nested ASN.1 errors when trying to decode a DER blob in
diverse ways.

That is now reproduced in ossl_store_handle_load_result()

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

CORE: Fix small bug in passphrase caching

Passphrase caching didn't allocate memory when it got to cache an
empty string, leading to a crash.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

STORE: Fix potential memory leak

When closing an OSSL_STORE_CTX, also clear the passphrase data.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

"Downgrade" provider-native keys to legacy where needed

Some sub-systems and openssl sub-commands do not yet deal cleanly with
purely provider-native EVP_PKEYs.  We compensate that by "downgrading"
keys in select places, or ensure that the 'file:' scheme ENGINE loader
is activated.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

EVP: Downgrade EVP_PKEYs in EVP_PKEY2PKCS8()

EVP_PKEY2PKCS8() relies on the presence of an EVP_PKEY_ASN1_METHOD,
which requires "downgrading" the EVP_PKEY to have a legacy internal
key.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

TEST: Modify test/recipes/90-test_store.t for use with different 'file:' loaders

We want to perform the same tests with a provider implementation, and
also make sure that an ENGINE implementation works as advertised.

OSSL_STORE_open() / OSSL_STORE_open_wirh_libctx() work in such a way
that they look for internal / engine implementations first, and only
failing that, they will try to fetch a provider implementation.  This
ensures that when we do specify an engine, it gets exceptional priority.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

STORE: Deprecate legacy / ENGINE functions

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

STORE: Add a built-in 'file:' storemgmt implementation (loader)

This replaces the older 'file:' loader that is now an engine.

It's still possible to use the older 'file:' loader by explicitly
using the engine, and tests will remain for it as long as ENGINEs are
still supported (even through deprecated).

To support this storemgmt implementation, a few internal OSSL_DECODER
modifications are needed:

-   An internal function that implements most of
    OSSL_DECODER_CTX_new_by_EVP_PKEY(), but operates on an already
    existing OSSL_DECODER_CTX instead of allocating a new one.
-   Allow direct creation of a OSSL_DECODER from an OSSL_ALGORITHM.
    It isn't attached to any provider, and is only used internally, to
    simply catch any DER encoded object to be passed back to the
    object callback with no further checking.  This implementation
    becomes the last resort decoder, when all "normal"
    decodation attempts (i.e. those that are supposed to result
    in an OpenSSL object of some sort) have failed.

Because file_store_attach() uses BIO_tell(), we must also support
BIO_ctrl() as a libcrypto upcall.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

STORE: Move the built-in 'file:' loader to become an engine module

From this point on, this engine must be specifically specified.

To replace the internal EMBEDDED hack with something unique for the
new module, functions to create application specific OSSL_STORE_INFO
types were added.

Furthermore, the following function had to be exported:

ossl_do_blob_header()
ossl_do_PVK_header()
asn1_d2i_read_bio()

Finally, evp_pkcs82pkey_int() has become public under a new name,
EVP_PKCS82PKEY_with_libctx()

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

OSSL_ENCODER / OSSL_DECODER post-rename cleanup

There are a few remaining spots where 'deser' wasn't changed to 'decoder'

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

ASN1: Fix d2i_KeyParams() to advance |pp| like all other d2i functions do

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12587)

Ensure EVP_MAC_update() passes the length even if it is 0

We leave it up to the EVP_MAC implemenations what to do with an update
where the data length is 0. In the TLS HMAC implemenation this is still
signficant.

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12732)

Convert ssl3_cbc_digest_record() to use EVP_MD_is_a()

Previously it used EVP_MD_type(), which doesn't work when called inside
the FIPs module.

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12732)

Update the EVP_PKEY MAC documentation

Include more information about the new HMAC parameter. Also fill in some
missing documentation about the EVP_PKEY MAC bridge.

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12732)

Enable PKEY MAC bridge signature algs to take ctx params

The underlying MAC implementations may take ctx params. Therefore we allow
the bridge to pass these through.

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12732)

Make ssl3_cbc_digest_record() use the real data_size

Previously we passed it the data plus mac size. Now we just pass it the
data size. We already know the mac size.

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12732)

Start using the provider side TLS HMAC implementation

This commit just moves the TLS1 and above implementation to use the TLS
HMAC implementation in the providers.

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12732)

Add an HMAC implementation that is TLS aware

The TLS HMAC implementation should take care to calculate the MAC in
constant time in the case of MAC-Then-Encrypt where we have a variable
amount of padding.

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12732)