engine: fix double free on error path.

In function try_decode_PKCS8Encrypted, p8 is freed via X509_SIG_free() at line 481.
If function new_EMBEDDED() returns a null pointer at line 483, the execution will goto nop8.
In the nop8 branch, p8 is freed again at line 491.

Bug reported by @Yunlongs

Fixes #14915

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14921)

ts: fix double free on error path.

In function int_ts_RESP_verify_token, if (flags & TS_VFY_DATA) is true, function ts_compute_imprint() will be called at line 299.
In the implementation of ts_compute_imprint, it allocates md_alg at line 406.
But after the allocation, if the execution goto err, then md_alg will be freed in the first time by X509_ALGOR_free at line 439.

After that, ts_compute_imprint returns 0 and the execution goto err branch of int_ts_RESP_verify_token.
In the err branch, md_alg will be freed in the second time at line 320.

Bug reported by @Yunlongs

Fixes #14914

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14921)

srp: fix double free,

In function SRP_create_verifier_ex, it calls SRP_create_verifier_BN_ex(..., &v, ..) at line 653.
In the implementation of SRP_create_verifier_BN_ex(), *verify (which is the paremeter of v) is allocated a pointer via BN_new() at line 738.
And *verify is freed via BN_clear_free() at line 743, and return 0.
Then the execution continues up to goto err at line 655, and the freed v is freed again at line 687.

Bug reported by @Yunlongs

Fixes #14913

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14921)

params_dup: fix off by one error that allows array overreach.

The end of loop test allows the index to go one step too far to be able to
terminate the param array but the end of list record is still added.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14922)

Improve ossl_cmp_build_cert_chain(); publish it as X509_build_chain()

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14128)

Flip ordering back

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/14219)

Fetch before get-by-name

This causes tests to break.  Pushing it to help others debug.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/14219)

Fetch and free cipher and md's

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/14219)

Fix compile errors on s390.

Commit f6c95e46c03025b2694241e1ad785d8bd3ac083b added an "origin" field to
EVP_CIPHER and EVP_MD structures but did not update the s390 specific
implementations.  Update these to fix compile errors on s390.

Signed-off-by: Juergen Christ <jchrist@linux.ibm.com>
Reviewed-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14926)

Remove extra trailing semicolon

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14903)

Update krb5 module to latest release

Fixes #14902

Also add workaround of `sudo hostname localhost` for the
intermittent test failures seen in CI.

Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/14872)

PKCS12 etc.: Add hints on using -legacy and -provider-path options

Fixes #14790

Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14891)

Add a test for OSSL_LIB_CTX_set0_default

Also includes testing for OSSL_LIB_CTX_get0_global_default().

Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14890)

Add the function OSSL_LIB_CTX_get0_global_default()

An API function for obtaining the global default lib ctx.

Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14890)

Change the semantics of OSSL_LIB_CTX_set0_default() NULL handling

Change things so that passing NULL to OSSL_LIB_CTX_set0_default() means
keep the current library context unchanged.

This has the advantage of simplifying error handling, e.g. you can call
OSSL_LIB_CTX_set0_default in an error/finalisation block safe in the
knowledge the if the "prevctx" was never set then it will be a no-op (like
calling a "free" function with NULL).

Fixes #14593

Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14890)

Remove a TODO(3.0) from keymgmt_lib.c

The TODO suggest a possible refactoring. The refactoring doesn't seem
necessary at this stage. If it is required later it can be done without
affecting external APIs - so just remove the TODO.

Fixes #14397

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14888)

Don't worry about magic in the Makefile for 3.0

We remove a TODO(3.0) from the unix Makefile template. The current
approach works. It can be improved later.

Fixes #14403

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14887)

Fix some TODO(3.0) occurrences in ssl/t1_lib.c

One was related to probing for the combination of signature and hash
algorithm together. This is currently not easily possible. The TODO(3.0)
is converted to a normal comment and I've raised the problem as issue
number #14885 as something to resolve post 3.0.

The other TODO was a hard coded limit on the number of groups that could
be registered. This has been amended so that there is no limit.

Fixes #14333

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14886)

Detect low-level engine and app method based keys

The low-level engine and app method based keys have to be treated
as foreign and must be used with old legacy pmeths.

Fixes #14632

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14859)

Remove crypt32.lib from C++Builder configuration

`import32.lib` serves the purpose for most Windows API libraries, including this one. For example, with a GNU `grep` utility:

>tdump %BDS%\lib\win32c\release\import32.lib | grep -B 3 -A 1 CertOpenStore
171E32 COMENT  Purge: Yes, List: Yes, Class: 160 (0A0h), SubClass: 1 (01h)
    Dynamic link import (IMPDEF)
      Imported by: name
      Internal Name: CertOpenStore
      Module Name: CRYPT32.dll

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Link with uplink module

The Clang-based `bcc32c.exe` expects AT&T syntax for inline assembly.
References:
 - http://docwiki.embarcadero.com/RADStudio/Sydney/en/Differences_Between_Clang-enhanced_C%2B%2B_Compilers_and_Previous-Generation_C%2B%2B_Compilers#Inline_Assembly
 - https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html
 - https://sourceware.org/binutils/docs/as/i386_002dVariations.html

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Link with .def files

MSVC's `link.exe` automatically finds `__cdecl` C functions (which are
decorated with a leading underscore by the compiler) when they are
mentioned in a `.def` file without the leading underscore.  This is an
[under-documented feature][1] of MSVC's `link.exe`.  C++Builder's
`ilink32.exe` doesn't do this, and thus needs the name-translation in
the `.def` file.  Then `implib.exe` needs to be told to re-add it.

(The Clang-based `bcc32c.exe` doesn't implement the [`-vu` or `-u-`][2]
options to skip adding the leading underscore to `__cdecl` C function
names, so this is the only way to have things work with non-underscored
export names in the DLLs.)

[1]: https://github.com/MicrosoftDocs/cpp-docs/issues/2653
[2]: http://docwiki.embarcadero.com/RADStudio/Sydney/en/Options_Not_Supported_by_Clang-enhanced_C%2B%2B_Compilers#BCC32_Options_that_Are_Not_Supported_by_Clang-enhanced_C.2B.2B_Compilers

Also silence linker warnings on duplicate symbols and ensure that error-
case cleanup in link rules work in C++Builder's `make.exe`.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Generate dependency information

The Clang-based `bcc32c.exe` doesn't implement the `-Hp` option, so we
have to use [`cpp32.exe`][1] instead.  Therefore, change the dependency-
emitting command to use `$(CPP)` instead of `$(CC)`, which which also
uncovered the [existing bug of `2>&1` before `> $dep`][2].  Also
C++Builder's `make.exe` doesn't implement `2>&1` in its command runner,
so wrap the whole line in a `cmd /C`.

[1]: http://docwiki.embarcadero.com/RADStudio/Sydney/en/CPP32.EXE,_the_C_Compiler_Preprocessor
[2]: https://ss64.com/nt/syntax-redirection.html

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Avoid more MSVC-specific C runtime library functions

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Build resource files

We need to compile with [brcc32.exe][1] and link with [ilink32.exe][2].
The latter expects the `.res` files to be given in the final comma-
separated section in the command line (after the `.def` file).

[1]: http://docwiki.embarcadero.com/RADStudio/Sydney/en/BRCC32.EXE,_the_Resource_Compiler
[2]: http://docwiki.embarcadero.com/RADStudio/Sydney/en/Using_ILINK32_and_ILINK64_on_the_Command_Line#Command-Line_Elements

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Support DLL builds + Fix C RTL variants

We need to generate a import library without the version in the
filename. MSVC's `link.exe` accommodates this with the [`/implib:`
option][1], while C++Builder needs a separate run of [`implib.exe`][2].
Also fix the variants of the [C runtime library and startup objects][3].

[1]: https://docs.microsoft.com/en-us/cpp/build/reference/implib-name-import-library?view=msvc-160
[2]: http://docwiki.embarcadero.com/RADStudio/Sydney/en/IMPLIB.EXE,_the_Import_Library_Tool_for_Win32
[3]: http://docwiki.embarcadero.com/RADStudio/Sydney/en/Static_Runtime_Libraries

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Ensure cw32mt.lib and import32.lib are linked to in no-sock mode

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Document C++Builder usage in NOTES-WINDOWS.md

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Replace "ld_wildcard_args" with "bin_lflags"

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Add explanation + bugtracker link for quoted dependency workarounds

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Use cmd.exe to export env vars before commands

Microsoft NMake runs the entire command block in a single `cmd.exe`
session, so environment variables exported with `set` are seen by the
following commands. C++Builder's `make.exe` doesn't use `cmd.exe` but
executes each command line separately. While it emulates only certain
shell features itself, `set` isn't one of them.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Resurrect and modernize C++Builder config

Similar configuration to what was deleted in 8e56a422, updated for the
unified build scheme. Use the [Clang-based `bcc32c.exe`][1] to avoid a
[bug][2] with the old one, and increase the `tlib.exe` page size to 256.
Also avoid MSVC-specific C runtime library functions.

[1]: http://docwiki.embarcadero.com/RADStudio/Sydney/en/Win32_Clang-enhanced_Compilers
[2]: https://quality.embarcadero.com/browse/RSP-31630

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Avoid redirection to quoted filename

C++Builder's `make.exe` cannot handle a command output redirection to a
quoted filename, saying: "The system cannot find the file specified."

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Generalize link rule in windows-makefile.tmpl

C++Builder's [`ilink32.exe`][1] expects its different types of input /
output files to be in a particular sequence and comma-separated -- even
in the response file as experimentation revealed, contrary to what is
documented. The lines of the response file need to end with `+`.

Also [`setargv.obj`][2] is MSVC-specific. The C++Builder equivalent is
[`wildargs.obj`][3].

[1]: http://docwiki.embarcadero.com/RADStudio/Sydney/en/Using_ILINK32_and_ILINK64_on_the_Command_Line
[2]: https://docs.microsoft.com/en-us/cpp/c-language/expanding-wildcard-arguments?view=msvc-160
[3]: http://docwiki.embarcadero.com/RADStudio/Sydney/en/Wildcard_Arguments

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Ensure at least one command if no dependencies

C++Builder's `make.exe` complains if a target has no dependencies (e.g.
after variable expansion) and no lines of commands. Ensure there is a
blank command line if the dependency list is entirely made of variables.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Avoid quoting dependency filepaths in build tree

C++Builder's `make.exe` has a bug in finding the rule of a quoted
dependency that doesn't exist in the filesystem. So for example:

    A: "src\B" "out\C"
     touch $@
    out\C:
     mkdir out
     touch $@

leads to:
    Fatal: '"out\C"' does not exist - don't know how to make it

This happens even with the `-N` option, and is different behavior from
Microsoft NMake which documents the feature of [quoted filepaths][1].

Commit cb663908 quoted all dependency filepaths, in case they are used
in a out-of-source build. The quoting is not done for target names,
however, which implies that the build directory is still expected to not
have spaces. It follows that we only need to quote non-generated source
files in dependency lists, since generated source files will be created
in the build directory.

Change the logic accordingly as a workaround, so that it works at least
for in-source builds with C++Builder's `make.exe`.

[1]: https://docs.microsoft.com/cpp/build/reference/long-filenames-in-a-makefile?view=msvc-160

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Generalize delimiter in archiver response file

While [`lib.exe` of MSVC][1] expects newline-delimited response file
lines, [`tlib.exe` of C++Builder][2] expects lines to end with `&` in
order to read the next line.

[1]: https://docs.microsoft.com/cpp/build/reference/running-lib?view=msvc-160#lib-command-files
[2]: http://docwiki.embarcadero.com/RADStudio/Sydney/en/TLIB.EXE,_the_Library_Manager#Response_Files

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Avoid space between "-I" and include directory

The space is problematic with C++Builder (`bcc32.exe`). MSVC (`cl.exe`)
doesn't care.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Move VS Tools configuration to VC-common target

The Windows toolchain is composed of utilites from both the Windows SDK
and the Visual Studio Build Tools. Move the configuration of the latter
utilities into the `VC-common` template inside `10-main.conf`, while
leaving the former utilities in `BASE_Windows`. This allows for other
Windows compilers to be configured inheriting from `BASE_Windows`.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Avoid "&&" in windows-makefile.tmpl

The `make.exe` utility shipped with Embarcadero C++Builder has a special
meaning for the [`&&` operator in commands][1]. This meaning is retained
even in Microsoft NMake compatibility mode (`make -N`). Split all
commands using this shell (`cmd.exe`) operator into two commands.

[1]: http://docwiki.embarcadero.com/RADStudio/Sydney/en/MAKE_Rules_(Explicit_and_Implicit)_and_Commands#MAKE_Command_Operators

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13540)

Add test for the IV handling of DES based ciphers

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/14882)

Do IV reset also for DES and 3DES implementations

Fixes #14704

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/14882)

Change the default MANSUFFIX

We now use the MANSUFFIX "ossl" by default.

Fixes #14318

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14847)

Add DHX FIPS 186-4 domain parameter validation example

Fixes #14369

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14896)

CORE: pre-populate the namemap with legacy OIDs too

This also pre-populates the namemap with names derived from the
internal EVP_PKEY_ASN1_METHODs.  This requires attention, as they
contain aliases that we may want (RSA == rsaEncryption), as well as
aliases that we absolutely do not want (SM2 == EC).

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14498)

TEST: Use OSSL_MAX_NAME_SIZE instead of arbitrary number of mdname

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14498)

CORE: Register all legacy "names" when generating the initial namemap

When generating the initial namemap from EVP cipher and digest names,
we din't do it quite as thoroughly as necessary, which meant that so
called "long names" weren't necessarily registered, and if anyone ever
tried to check the algorithm of an EVP_CIPHER or EVP_MD using a so
called "long name" would fail.

This doesn't deal with the fact that "long names" have a distinct role
as human readable descriptors, useful for printouts.  Further changes
are necessary to deal with this.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14498)

Modify OBJ_nid2sn(OBJ_obj2nid(...)) occurences to use OBJ_obj2txt()

The intention is to allow for OIDs for which libcrypto has no
information, but are still fetchable for OSSL_ALGORITHM
implementations that specify an OID amongst their names.

Fixes #14278

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14498)

TEST: Modify how the retrieved digest name for SM2 digestsign is checked

Because of aliases, retrieved names won't always match one specific
string.  A safer way to check is to fetch the digest from the
retrieved name and check it's the expected one with the help of
EVP_MD_is_a().

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14498)

TEST: Modify testutil's run_tests to display NOSUBTEST cases individually

When test cases were added with ADD_ALL_TESTS_NOSUBTEST(), all those
iteration verdicts were summarized as if it was one single case.  This
modification gets each iteration verdict displayed separately instead.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14498)

TEST: Modify test/evp_fetch_prov_test.c to also fetch by OID

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14498)

PROV: Add OIDs we know to all provider applicable algorithms

The OIDs were extracted with the help of libcrypto's ASN1 OID database.

While doing this, we move all the names strings to macro definitions,
to avoid duplication and conflicting names declarations.  Those macros
are all in providers/implementations/include/prov/names.h

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14498)

Add "origin" field to EVP_CIPHER, EVP_MD

Add a "where did this EVP_{CIPHER,MD} come from" flag: global, via fetch,
or via EVP_{CIPHER,MD}_meth_new.  Update EVP_{CIPHER,MD}_free to handle all
three origins. The flag is deliberately right before some function pointers,
so that compile-time failures (int/pointer) will occur, as opposed to
taking a bit in the existing "flags" field.  The "global variable" flag
is non-zero, so the default case of using OPENSSL_zalloc (for provider
ciphers), will do the right thing. Ref-counting is a no-op for
Make up_ref no-op for global MD and CIPHER objects

Deprecate EVP_MD_CTX_md().  Added EVP_MD_CTX_get0_md() (same semantics as
the deprecated function) and EVP_MD_CTX_get1_md().  Likewise, deprecate
EVP_CIPHER_CTX_cipher() in favor of EVP_CIPHER_CTX_get0_cipher(), and add
EVP_CIPHER_CTX_get1_CIPHER().

Refactor EVP_MD_free() and EVP_MD_meth_free() to call new common
evp_md_free_int() function.
Refactor EVP_CIPHER_free() and EVP_CIPHER_meth_free() to call new common
evp_cipher_free_int() function.

Also change some flags tests to explicit test == or != zero. E.g.,
        if (flags & x) --> if ((flags & x) != 0)
        if (!(flags & x)) --> if ((flags & x) == 0)
Only done for those lines where "get0_cipher" calls were made.

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14193)

Standard style for all EVP_xxx_free routines

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14193)

Fix typo in aesccm.c

CLA: trivial

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14894)

apps/cmp.c: Fix TLS hostname checking in case -server provides more than hostname

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14889)

OSSL_CMP_CTX_new(): Fix distinction of out-of-memory and other errors

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14889)

Fix naming for EVP_RAND_CTX_gettable functions.

Change:

    EVP_RAND_gettable_ctx_params -> EVP_RAND_CTX_gettable_params
    EVP_RAND_settable_ctx_params -> EVP_RAND_CTX_settable_params

Which brings them in line with the other similar functions for other algorithm
types.

Fixes #14880

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14893)

Sanity check provider up-calls

When we store references to up-calls for future reference we run a sanity
check to make sure we either previously haven't set these values or they
are the same as last time. We don't support the scenario where an
application is linked against multiple versions of libcrypto but using a
shared fips.so file. This would result in different up-calls for different
calls to OSSL_provider_init(), which we currently can't handle.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14814)

Store some FIPS global variables in the FIPS_GLOBAL structure

We had some FIPS global variables that were based on values from the
config file. In theory if two instances of the fips module are loaded
they could be based on different config files which would cause this to
fail. Instead we store them in the FIPS_GLOBAL structure.

Fixes #14364

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14814)

lifecycle: update master lifecycle transition spreadsheet fixing the ettable issue

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14867)

lifecycle: correct [sg]ettable to [sg]et

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14867)

demos: Add clean target for bio/Makefile

CLA: trivial

Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14661)

Add more negative checks for integers passed to OPENSSL_malloc().

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14830)

SipHash: Fix CTRL API for the digest size.

Adds a mapping from "digestsize" to "size" for setting a param.

Fixes #14370

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14850)

Github workflows: re-implement a no-shared build

We do this both on Ubuntu and MacOS X

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/14753)

Add EVP_PKEY_todata() and EVP_PKEY_export() functions.

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14800)

Add OID for mdc2WithRSASignature and remove related TODO 3.0

Fixes #14366

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14874)

Rename EVP_PKEY_get0_first_alg_name to EVP_PKEY_get0_type_name

We use type elsewhere and documenting the 'first' in the
name of the call is a little bit superfluous making the
name too mouthful.

Also rename EVP_PKEY_typenames_do_all to
EVP_PKEY_type_names_do_all to keep the words separated by
underscore.

Fixes #14701

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14868)

changes: note that some ctrl calls have a different error return.

Providers do not distinguish between invalid and other errors via the return
code.

Fixes #14442

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14864)

Do not allow creating empty RSA keys by duplication

Also avoid crashing in rsa_get_params on empty keys.

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/14793)

Remove keymgmt_copy function from the provider API

It is superceded by the keymgmt_dup.

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/14793)

Add selection support to the provider keymgmt_dup function

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/14793)

Implement provider-side keymgmt_dup function

To avoid mutating key data add OSSL_FUNC_KEYMGMT_DUP function
to the provider API and implement it for all asym-key key
managements.

Use it when copying everything to an empty EVP_PKEY
which is the case with EVP_PKEY_dup().

Fixes #14658

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/14793)

80-test_cmp_http.t: Extend diagnostics of mock server launch

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14839)

80-test_cmp_http.t: Silence check for availability of 'kill' and 'lsof' commands

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14839)

80-test_cmp_http.t: Fix resumption when skipping after mock server launch failed

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14839)

apps/cmp: Add generic random state options, e.g., for nonce generation

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14842)

openssl-cmp.pod.in: Fix missing provider options description

Also correct layout of engines description

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14842)

cmp_util.c: Fix OSSL_CMP_log_open() in case OPENSSL_NO_TRACE

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14842)

APPS and TEST: Make sure prog name is set for usage output

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14841)

APPS: make apps strict on app_RAND_load() and app_RAND_write() failure

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14840)

Document the invariants for the empty X509_NAME encoding

Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/14832)

X509_NAME_cmp: if canon_enclen is 0 for both names return 0

We do not care whether canon_enc is NULL in this case.

Fixes #14813

Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/14832)

bio_printf: add \0 terminators for error returns in floating point conversions.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14829)

bio: note that BIO_sprintf null terminates on insufficient space.

Fixes: #14772
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14829)

bio: add a malloc failed error to BIO_print

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14829)

Add some additional NULL checks to prevent segfaults.

Fixes #14809

PR #14752 attempted to pass the libctx, propq in a few places related to
X509 signing. There were a few places that needed additional NULL checks so that they behavethe same as they did before.

OCSP_basic_sign() was changed to call EVP_DigestSignInit_ex() which passed the parameter EVP_MD_name(dgst). Since dgst can be NULL EVP_MD_name() was segfaulting.
Adding an additional NULL check EVP_MD_name() resolves this issue.

The other NULL checks are required to produce errors rather than
segfaults if the certificate is NULL.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14826)

Add domain parameter match check for DH and ECDH key exchange.

Fixes #14808

Validation checks were moved into EVP_PKEY_derive_set_peer() which broke
an external negative test. Originally the old code was semi working by checking the peers public key was in the range of other parties p. It was not actually ever
checking that the domain parameters were consistent between the 2
parties. It now checks the parameters match as well as validating the
peers public key.

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14823)

Remove a TODO(3.0) from X509_PUBKEY_set

The comment talks about the EVP_PKEY that is contained within an
X509_PUBKEY object and whether it has to be exactly the same as the one
passed by the caller in X509_PUBKEY_set(). IMO it does, so the TODO should
be dropped.

Fixes #14378

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14845)

crypto: raise error on malloc failure
clean a few style nits.

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14806)

nits: fix a few typo in template code

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14806)

doc: Fix formatting

CLA: trivial

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14835)

Handle set_alpn_protos inputs better.

It's possible to set an invalid protocol list that will be sent in a
ClientHello. This validates the inputs to make sure this does not
happen.

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14815)

util/wrap.pl: use the apps/openssl.cnf from the source tree

The `make install_fips` target failed

    msp@debian:~/src/openssl$ make install_fips
    *** Installing FIPS module
    install providers/fips.so -> /opt/openssl-dev/lib/ossl-modules/fips.so
    *** Installing FIPS module configuration
    fipsinstall /opt/openssl-dev/ssl/fipsmodule.cnf
    FATAL: Startup failure (dev note: apps_startup()) for ./apps/openssl
    ... No such file or directory:crypto/conf/conf_def.c:771:calling stat(fipsmodule.cnf)
    ...
    make: *** [Makefile:3341: install_fips] Error 1

because the `openssl fipsinstall` command was loading a previously installed
configuration file instead of the copy shipped with the source tree.

    msp@debian:~/src/openssl$ strace -f make install_fips |& grep openssl.cnf
    [pid 128683] openat(AT_FDCWD, "/opt/openssl-dev/ssl/openssl.cnf", O_RDONLY) = 3

This issue reveiled a more general problem, which applies to the tests as well:
unless openssl is installed, the openssl app must not use any preinstalled
configuration file. This holds in particular when the preinstalled configuration
file load providers, which caused the above failure.

The most consistent way to achieve this behaviour is to set the OPENSSL_CONF
environment variable to the correct location in the util/wrap.pl perl wrapper.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14136)

apps: call ERR_print_errors when OSSL_PROVIDER_load fails

The ERR_print_errors often displays the reason why the provider
couldn't be loaded. Hence it is quite important for debugging.

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14818)

Note deprecated function/macros with no replacement.

These functions are deprecated with no replacement specified:

    DH_clear_flags, DH_get_1024_160, DH_get_2048_224, DH_get_2048_256,
    DH_set_flags, DH_test_flags, DSA_clear_flags, DSA_dup_DH,
    DSAparams_dup, DSA_set_flags, DSA_test_flags, RSA_blinding_off,
    RSA_blinding_on, RSA_clear_flags, RSA_get_version, RSAPrivateKey_dup,
    RSAPublicKey_dup, RSA_set_flags, RSA_setup_blinding and
    RSA_test_flags.

The flags that are going are:

    DH_FLAG_CACHE_MONT_P, DSA_FLAG_CACHE_MONT_P,
    RSA_FLAG_BLINDING, RSA_FLAG_CACHE_PRIVATE, RSA_FLAG_CACHE_PUBLIC,
    RSA_FLAG_EXT_PKEY, RSA_FLAG_NO_BLINDING, RSA_FLAG_THREAD_SAFE and
    RSA_METHOD_FLAG_NO_CHECK.

These two flags are "readable" via EVP_is_a().  They are not writable:

    DH_FLAG_TYPE_DHX and DH_FLAG_TYPE_DH.

Fixes #14616

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/14824)

Fix windows compiler error in kmac_prov.c

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14851)

Add FIPS Self test for AES_ECB decrypt

Fixes #14807

Compliance with IG 9.4 requires that an inverse cipher function be
tested if one is implemented. Just running AES_GCM encrypt/decrypt does not meet this
requirement (Since only ECB, CBC, XTS, KW, KWP support the inverse
function during decryption mode).

Added a mode to the cipher test so that the AES_GCM only does an encrypt
and AES_ECB only does a decrypt. TDES still does both.

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14825)

Remove the function EVP_PKEY_set_alias_type

OTC recently voted that EVP_PKEY types will be immutable in 3.0. This
means that EVP_PKEY_set_alias_type can no longer work and should be
removed entirely (applications will need to be rewritten not to use it).

It was primarily used for SM2 which no longer needs this call.
Applications should generate SM2 keys directly (without going via an EC
key first), or otherwise when loading keys they should automatically be
detected as SM2 keys.

Fixes #14379

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14803)