Raise an error on syscall failure in tls_retry_write_records

Record the errno when we get a syscall failure in
tls_retry_write_records

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23723)

Make a failure in ktls_sendfile a syscall error

a failure in ktls_sendfile results in an error in ERR_LIB_SSL, but its
really a syscall error, since ktls_sendfile just maps to a call to the
sendfile syscall.  Encode it as such

Fixes #23722

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23723)

Fix coverity 1596617

Somehow a double free slipped into conf_mod.c, remove it

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24263)

Fix coverity-1596616

Need to add a null check prior to derefencing pointer for free

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24263)

Add an Apple privacy info file for OpenSSL

Added PrivacyInfo.xcprivacy to os-dep/Apple/ dir.

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24260)

ssl/statem: Replace size_t with int and add the checks

Replace the type of variables with int to avoid implicit conversion when it is assigned by EVP_MD_get_size().
Moreover, add the checks to avoid integer overflow.

Fixes: 6594189 ("Merge early_data_info extension into early_data")
Fixes: 9368f86 ("Add TLSv1.3 client side external PSK support")
Fixes: 1053a6e ("Implement Server side of PSK extension parsing")
Signed-off-by: Jiasheng Jiang <jiasheng@purdue.edu>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23937)

Avoid duplicate default CApath lookups

Fixes #21067

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24140)

hkdf: when HMAC key is all zeros, still set a valid key length

By itself, this is no change in any computation. However, this will
unlock enforcing minimum key lengths for NIST and FIPS 140-3
requirements.

Also reading RFC8448 and RFC5869, this seems to be strictly correct
too.

Signed-off-by: Dimitri John Ledkov <dimitri.ledkov@surgut.co.uk>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24204)

Fix grammar in srp_verifier.txt

CLA: trivial

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24237)

sha512.c: Grammar Fixes and Spell Checks in Implementation Notes

Added commas for sentence openers in Implementation Notes. Fixed
spelling of "reasons" section of the notes.

CLA: trivial

Co-authored-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24241)

Fix grammar in certificates.txt

CLA: trivial

Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24238)

Update perl-actions/install-with-cpanm version in CI

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/23613)

Add Attribute Certificate suport comments to CHANGES and NEWS

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15857)

x509_acert: Add more parsing and printing tests

These have been extracted from the boucycastle test code.
Make sure that these certificates can be safely and correctly parsed
and printed.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15857)

x509_acert: Load attributes from config file section

Several of the attribute values defined for use by attribute certificates
use multi-valued data in an ASN.1 SEQUENCE. Allow reading of these values
from a configuration file, similar to how generic X.509 extensions are
handled.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15857)

fuzz: Add attribute certificate fuzz test

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15857)

x509_acert: Add simple API tests

Add a some simple API tests for reading, printing, signing
and verifying attribute certificates.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15857)

Add IETFAttrSyntax type support

The IETFAtrrSyntax type is used for the values of several attributes
defined in RFC 5755 for use with attribute certificates.
Specifically this type is used with the "Charging Identity" and
"Group" attributes.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15857)

x509_acert: Add and retrieve certificate extensions

Add API to manage attribute certificate extensions

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15857)

x509_acert: Add API to sign and verify attribute certificates

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15857)

x509_acert: Add, remove and get attribute certificate attributes

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15857)

Attribute certificate printing functions

Add functions to print an attribute certificate.  Several
attribute value types defined by the RFC 5755 specification
are multi-field values (i.e ASN1_SEQUENCE rather than an ASN1_STRING
or similar format).  Currently those values are printed using
`ASN1_item_print`.  A more user-friendly output mechanism (maybe
similar to the i2r_ functions used for X509 extensions) could be
added in future.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15857)

Attribute certificate getter and setter API

Only fields that are allowed by RFC 5755 are
accessible through this API.  Fields that are only supported
in version 1 attribute certificates (e.g. the AttCertIssuer
v1Form fields) are not implemented.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15857)

Add RFC 5755 attribute certificate support

Add support for attribute certificates (v2) as described
in RFC 5755 profile.

Attribute certificates provide a mechanism to manage authorization
information separately from the identity information provided by
public key certificates.

This initial patch adds the ASN.1 definitions
and I/O API.  Accessor functions for the certificate fields
will be added in subsequent patches.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15857)

Remove all references to FLOSS for NonStop Builds.

FLOSS is no longer a dependency for NonStop as of the deprecation of the SPT
thread model builds.

Fixes: #24214
Signed-off-by: Randall S. Becker <randall.becker@nexbridge.ca>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24217)

updating fuzz-corpora submodule

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23671)

adding a multithreaded hashtable test

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23671)

Adding hashtable fuzzer

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23671)

Introduce new internal hashtable implementation

Create a new hashtable that is more efficient than the existing LHASH_OF
implementation.  the new ossl_ht api offers several new features that
improve performance opportunistically

* A more generalized hash function.  Currently using fnv1a, provides a
  more general hash function, but can still be overridden where needed

* Improved locking and reference counting.  This hash table is
  internally locked with an RCU lock, and optionally reference counts
  elements, allowing for users to not have to create and manage their
  own read/write locks

* Lockless operation.  The hash table can be configured to operate
  locklessly on the read side, improving performance, at the sacrifice
  of the ability to grow the hash table or delete elements from it

* A filter function allowing for the retrieval of several elements at a
  time matching a given criteria without having to hold a lock
  permanently

* a doall_until iterator variant, that allows callers which need to
  iterate over the entire hash table until a given condition is met (as
  defined by the return value of the iterator callback).  This allows
  for callers attempting to do expensive cache searches for a small
  number of elements to terminate the iteration early, saving cpu cycles

* Dynamic type safety.  The hash table provides operations to set and
  get data of a specific type without having to define a type at the
  instatiation point

* Multiple data type storage.  The hash table can store multiple data
  types allowing for more flexible usage

* Ubsan safety.  Because the API deals with concrete single types
  (HT_KEY and HT_VALUE), leaving specific type casting to the call
  recipient with dynamic type validation, this implementation is safe
  from the ubsan undefined behavior warnings that require additional
  thunking on callbacks.

Testing of this new hashtable with an equivalent hash function, I can
observe approximately a 6% performance improvement in the lhash_test

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23671)

Add CRYPTO_atomic_store api

Generally we can get away with just using CRYPTO_atomic_load to do
stores by reversing the source and target variables, but doing so
creates a problem for the thread sanitizer as CRYPTO_atomic_load hard
codes an __ATOMIC_ACQUIRE constraint, which confuses tsan into thinking
that loads and stores aren't properly ordered, leading to RAW/WAR
hazzards getting reported.  Instead create a CRYPTO_atomic_store api
that is identical to the load variant, save for the fact that the value
is a unit64_t rather than a pointer that gets stored using an
__ATOMIC_RELEASE constraint, satisfying tsan.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23671)

Fix list appending in win ossl_rcu_call

The ossl_rcu_call function for windows creates a linked list loop.  fix
it to work like the pthread version properly

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23671)

Make thread sanitizer cope with rcu locks

This is unfortunate, but seems necessecary

tsan in gcc/clang tracks data races by recording memory references made
while various locks are held.  If it finds that a given address is
read/written while under lock (or under no locks without the use of
atomics), it issues a warning

this creates a specific problem for rcu, because on the write side of a
critical section, we write data under the protection of a lock, but by
definition the read side has no lock, and so rcu warns us about it,
which is really a false positive, because we know that, even if a
pointer changes its value, the data it points to will be valid.

The best way to fix it, short of implementing tsan hooks for rcu locks
in any thread sanitizer in the field, is to 'fake it'.  If thread
sanitization is activated, then in ossl_rcu_write_[lock|unlock] we add
annotations to make the sanitizer think that, after the write lock is
taken, that we immediately unlock it, and lock it right before we unlock
it again.  In this way tsan thinks there are no locks held while
referencing protected data on the read or write side.

we still need to use atomics to ensure that tsan recognizes that we are
doing atomic accesses safely, but thats ok, and we still get warnings if
we don't do that properly

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23671)

Coverity found the following issues:

1591471
1591474
1591476

which pertain to memory leaks in the conf_mod code

If an error is encountered after the module STACK_OF is duplicated or
created in the new_modules variable, we need to remember to free it in
the error path

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23462)

Fix potential divide by zero error

Coverity caught the following issues:
1591477
1591475
1591473
1591470

all of which are simmilar, in that they catch potential divide by zero
in double values.  It can't actually happen since the the threads which
increment these counters don't exit until they reach non-zero values,
but its easy to add the checks, so lets do that to ensure that we don't
change something in the future that causes it.

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23462)

Invoke tear_down when exiting test_encode_tls_sct() prematurely

Fixes #24121

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24222)

crypto/threads_pthread.c: Fix typos found by codespell

CLA: trivial

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24206)

Be more explicit about RSAES-PKCS#1v1.5 error handling

And add a note how to perform side-channel free error stack handling.

Signed-off-by: Hubert Kario <hkario@redhat.com>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24159)

Use empty renegotiate extension instead of SCSV for TLS > 1.0

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24161)

Fix missing NULL check in prov_config_test

coverity-1596500 caught a missing null check.  We should never hit it as
the test harness always sets the environment variable, but lets add the
check for safety

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24208)

fix sending error when no root CA cert update available

Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24169)

Fix migration guide mappings for i2o/o2i_ECPublicKey

Fixes #23854

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24041)

Make rcu_thread_key context-aware

Currently, rcu has a global bit of data, the CRYPTO_THREAD_LOCAL object
to store per thread data.  This works in some cases, but fails in FIPS,
becuase it contains its own copy of the global key.

So
1) Make the rcu_thr_key a per-context variable, and force
   ossl_rcu_lock_new to be context aware

2) Store a pointer to the context in the lock object

3) Use the context to get the global thread key on read/write lock

4) Use ossl_thread_start_init to properly register a cleanup on thread
   exit

5) Fix up missed calls to OSSL_thread_stop() in our tests

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24162)

OSSL_STORE: Add reference docs for the built-in Windows store implementation

Fixes openssl/project#422

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24170)

openssl fipsinstall: fix cosmetic wart

This change makes the message on failure consistent with the message on
success by trimming a single space in the error message.

CLA: trivial
Signed-off-by: Enji Cooper <yaneurabeya@gmail.com>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/24180)

Adjust tests that were depending on X25519 and X448 in fips

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/24099)

Exclude X25519 and X448 from capabilities advertised by FIPS provider

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24099)

Make X25519 and X448 FIPS unapproved

Partially fixes: #22105

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/24099)

Fix up path generation to use OPENSSL_MODULES

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24025)

Update modulepath test for provider config to skip if not present

If the p_test.so library isn't present, don't run the test

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24025)

Add test for OSSL_PROVIDER_load with module path set

Ensure that, with the modulepath setting set in a config field, that we
are able to load a provider from the path relative to OPENSSL_MODULES

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24025)

set module path from template

Modules that aren't activated at conf load time don't seem to set the
module path from the template leading to load failures.  Make sure to
set that

Fixes #24020

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24025)

QUIC TXP: Fix reserve calculations for PING frames

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24122)

Fix fragile explicit cert date tests.

The tests used localtime to format "today's" date, but then extracted a
GMT date from the cert.  The comparison breaks when run late in the
evening west of UTC, or early in the AM hours east of UTC.

Also took care of case when test runs at stroke of midnight, by
accepting either the "today" before the cert creation, or the
"today" after, should they be different.

Fixes fragile tests in #21716

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24139)

Update provider-compatibility.yml

Documentation Change: Line 34

Changed 'utl' to 'url' to correctly reflect the variables used in the releases in this file.

CLA: trivial

Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24164)

Updated list formatting, added hyperlinks, modernized syntax

Updated list formatting to allow for easier readability, Added/adjusted hyperlinks, modernized command substitution syntax

CLA: trivial

Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/24165)

.ctags.d is previous, include it in our tarballs

This is a simple change of .gitattributes, so our tarballs continue to
be a reproducible output of a util/mktar.sh (i.e. git archive with no
other funny business).

Fixes #24090

Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24156)

Use scalar ALU and vector ALU together for chacha20 stream cipher

Fixes #24070

Use scalar ALU for 1 chacha block with rvv ALU simultaneously.
The tail elements(non-multiple of block length) will be handled by
the scalar logic.

Use rvv path if the input length > chacha_block_size.

And we have about 1.2x improvement comparing with the original code.

Reviewed-by: Hongren Zheng <i@zenithal.me>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24097)

poly1305.c: fix typo on POLY1305_BLOCK_SIZE

no code change

Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24136)

Unable to run asm code on OpenBSD (amd64)

In order to get asm code running on OpenBSD we must place
all constants into .rodata sections.

davidben@ also pointed out we need to adjust `x86_64-xlate.pl` perlasm
script to adjust read-olny sections for various flavors (OSes). Those
changes were cherry-picked from boringssl.

closes #23312

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23997)

extend x86_64-xlate.pl perlasm so it can handle .rodata sections properly

For nasm/masm assembler flavors the xlate script must make sure the code
won't land in .rodata section along the data.

For masm we also need to introduce an .align option which can be passed
along section header. It's hint for masm to align rodata/rdata section
properly.

Also macos-x flavor requires small tweak to emit proper section header
for its assembler style.

Changes for masm flavor are based on SEGMENT description [1] in
MASM reference manual.

Changes for nasm flavor are based on nasm 2.14 manual chapter 7 [2].

Details behind macos-x changes can be found in 'Overview of the Mach-O
Executable Format' [3]

[1] https://learn.microsoft.com/en-us/cpp/assembler/masm/segment?view=msvc-170

[2] https://nasm.us/xdoc/2.14rc0/html/nasmdoc7.html

[3] https://developer.apple.com/library/archive/documentation/Performance/Conceptual/CodeFootprint/Articles/MachOOverview.html

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23997)

fuzz/decoder.c: Limit the EVP_PKEY_param_check on DHX keys as well

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24126)

Define KU_ constants via corresponding X509v3_KU_

Also wrap X509v3_KU_UNDEF in `#ifndef OPENSSL_NO_DEPRECATED_3_4`.

Fixes #22955

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24138)

Handle empty param in EVP_PKEY_CTX_add1_hkdf_info

Fixes #24130
The regression was introduced in PR #23456.

Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24141)

OpenSSL 3.2.0, QUIC, macOS, error 56 on connected UDP socket

current `translate_msg()` function attempts to set `->msg_name`
(and `->msg_namelen`) with `BIO`'s peer name (connection destination)
regardless if underlying socket is connected or not. Such implementation
uncovers differences in socket implementation between various OSes.

As we have learned hard way `sendmsg()` and `sendmmsg()` on `OpenBSD`
and (`MacOS` too) fail to send messages with `->msg_name` being
set on connected socket. In such case the caller receives
`EISCON` errro.

I think `translate_msg()` caller should provide a hint to indicate
whether we deal with connected (or un-connected) socket. For
connected sockets the peer's name should not be set/filled
by `translate_msg()`. On the other hand if socket is un-connected,
then `translate_msg()` must populate `->msg_name` and `->msg_namelen`
members.

The caller can use `getpeername(2)` to see if socket is
connected. If `getpeername()` succeeds then we must be dealing
with connected socket and `translate_msg()` must not set
`->msg_name` and `->msg_namelen` members. If `getpeername(2)`
fails, then `translate_msg()` must provide peer's name (destination
address) in `->msg_name` and set `->msg_namelen` accordingly.

The propposed fix introduces `is_connected()` function,
which applies `getpeername()` to socket bound to `BIO` instance.
The `dgram_sendmmsg()` uses `is_connected()` as a hint
for `translate_msg()` function, so msghdr gets initialized
with respect to socket state.

The change also modifies existing `test/quic_client_test.c`
so it also covers the case of connected socket. To keep
things simple we can introduce optional argument `connect_first`
to `./quic_client_test` function. Without `connect_first`
the test run as usual. With `connect_first` the test creates
and connects socket first. Then it passes such socket to
`BIO` sub-system to perform `QUIC` connect test as usual.

Fixes #23251

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23396)

doc/fingerprints.txt: Add the future OpenSSL release key

This will be used for future releases

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24063)

crypto/threads_pthread.c: refactor all atomics fallbacks for type safety

The atomics fallbacks were using 'void *' as a generic transport for all
possible scalar and pointer types, with the hypothesis that a pointer is
as large as the largest possible scalar type that we would use.

Then enters the use of uint64_t, which is larger than a pointer on any
32-bit system (or any system that has 32-bit pointer configurations).

We could of course choose a larger type as a generic transport.  However,
that only pushes the problem forward in time...  and it's still a hack.
It's therefore safer to reimplement the fallbacks per type that atomics
are used for, and deal with missing per type fallbacks when the need
arrises in the future.

For test build purposes, the macro USE_ATOMIC_FALLBACKS is introduced.
If OpenSSL is configured with '-DUSE_ATOMIC_FALLBACKS', the fallbacks
will be used, unconditionally.

Fixes #24096

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24123)

crypto/threads_pthread.c: Cleanup misaligned preprocessor directives

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24123)

Remove repetitive words

Signed-off-by: fanqiaojun <fanqiaojun@yeah.net>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24128)

Adding missing NULL pointer check

CLA: trivial
In the provider store API, it is not necessary to provide both open and
attach method at the same time and providing at least one of them is
enough. Adding some null pointer checks to prevent exceptions in case
of not providing both methods at the same time.

Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23703)

list_provider_info(): Fix leak on error

Fixes #24110

Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/24117)

Augment README.md in top level to indicate Makefile presence

Note that they are available but only meant as a guide to self building,
and are not used expressly to build as part of the overall openssl build

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
(Merged from https://github.com/openssl/openssl/pull/24047)

disable http3 demo on windows

The external nghttp3 library seems to have a linking issue on windows
(several missing symbols).  Disable that build in windows for now until
its fixed

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
(Merged from https://github.com/openssl/openssl/pull/24047)

make addr_len the right sign in sslecho

cygwin caught a signedness difference in this pointer.

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
(Merged from https://github.com/openssl/openssl/pull/24047)

Replace getline with fgets in sslecho demo

Windows doesn't support getline, so we need to use fgets here

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
(Merged from https://github.com/openssl/openssl/pull/24047)

dont include unistd.h on windows for sslecho

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
(Merged from https://github.com/openssl/openssl/pull/24047)

Don't include unistd.h in sconnect for windows

The platform doesn't support it

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
(Merged from https://github.com/openssl/openssl/pull/24047)

Fix signal handling in saccept for windows

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
(Merged from https://github.com/openssl/openssl/pull/24047)

Fix warnings found by clang in CI

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
(Merged from https://github.com/openssl/openssl/pull/24047)

Enable demos in select builds

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
(Merged from https://github.com/openssl/openssl/pull/24047)

fix all the warnings in our demos and make them enableable

Fix up the warnings in the demos and make them configurable with
enable-demos

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
(Merged from https://github.com/openssl/openssl/pull/24047)

Convert demos to primary build system

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
(Merged from https://github.com/openssl/openssl/pull/24047)

ossl_provider_new(): Fix memory leak on error

Fixes #24095

Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24100)

make_addressPrefix(): Fix a memory leak in error case

Fixes #24098

Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24102)

Fix duplicate mutex allocation in threads_win.c

Creating an rcu lock does a double allocation of the underlying mutex.
Not sure how asan didn't catch this, but we clearly have a duplicate
line here

Fixes #24085

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24086)

VMS: Move defining _XOPEN_SOURCE and  _XOPEN_SOURCE_EXTENDED to config target

For all other platforms that need these macros defined, that's how it's
done, so we have VMS follow suit.  That avoids a crash between in source
definitions and command line definitions on some other platforms.

Fixes #24075

Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24083)

(cherry picked from commit 7f04bb065d9d948d049ef1ef1bd4062cb7831392)

crypto/provider_core.c: Allocate activatecnt_lock

CRYPTO_atomic_add has a lock as a parameter, which is often ignored, but in
some cases (for example, when BROKEN_CLANG_ATOMICS is defined) it is required.

There is no easy way to determine if the lock is needed or not. The current
logic looks like this:

    if defined(OPENSSL_THREADS) && !defined(CRYPTO_TDEBUG) && !defined(OPENSSL_SYS_WINDOWS)
      if defined(__GNUC__) && defined(__ATOMIC_ACQ_REL) && !defined(BROKEN_CLANG_ATOMICS)
        - It works without the lock, but in general the need for the
          lock depends on __atomic_is_lock_free results
      elif defined(__sun) && (defined(__SunOS_5_10) || defined(__SunOS_5_11))
        - The lock is not needed (unless ret is NULL, which should never
          happen?)
      else
        - The lock is required
      endif
    else
      - The lock is not needed
    endif

Adding such conditions outside of crypto.h is error-prone, so it is better to
always allocate the lock, otherwise CRYPTO_atomic_add may silently fail.

Fixes #23376.

CLA: trivial
Fixes: fc570b2605 ("Avoid taking a write lock in ossl_provider_doall_activated()")
Signed-off-by: Oleg Bulatov <oleg@bulatov.me>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24081)

Change approach to SSL_pending API

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24040)

(cherry picked from commit 5a13d35f243be66f6ad914aefe99fb708812dff1)

QUIC APL: Revise SSL_pending and SSL_has_pending handling for s_client compat

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24040)

(cherry picked from commit 7c33eb1e7fd3248ad29c172b5b4c0658a7be3adc)

QUIC QSM: Add function to determine if data is waiting

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24040)

(cherry picked from commit 3c2bc702eb9287b84e8584ad427e72da0ab21ec1)

test: fix 20-test_dgst.t to use hexkey

Currently 20-test_dgst.t calls a quite bogus command:

    $ openssl dgst -sha256 -hmac -macopt hexkey:FFFF test/data.bin test/data.bin
    hexkey:FFFF: No such file or directory
    HMAC-SHA2-256(test/data.bin)= b6727b7bb251dfa65846e0a8223bdd57d244aa6d7e312cb906d8e21f2dee3a57
    HMAC-SHA2-256(test/data.bin)= b6727b7bb251dfa65846e0a8223bdd57d244aa6d7e312cb906d8e21f2dee3a57
    805B632D4A730000:error:80000002:system library:file_ctrl:No such file or directory:crypto/bio/bss_file.c:297:calling fopen(hexkey:FFF, r)
    805B632D4A730000:error:10080002:BIO routines:file_ctrl:system lib:crypto/bio/bss_file.c:300:

Does not check status code, discards stderr, and verifies the
checksums as per above. Note that the checksum is for the HMAC key
"-macopt", and `hexkey:FFFF` is attempted to be opened as a file.

See HMAC values for key `-macopt` and `hexkey:FFFF` using `openssl-mac`:

    $ openssl mac -digest SHA256 -macopt hexkey:$(printf '%s' '-macopt' | xxd -p -u) -in ./test/data.bin HMAC
    B6727B7BB251DFA65846E0A8223BDD57D244AA6D7E312CB906D8E21F2DEE3A57

    $ openssl mac -digest SHA256 -macopt hexkey:FFFF -in ./test/data.bin HMAC
    7C02D4A17D2560A5BB6763EDBF33F3A34F415398F8F2E07F04B83FFD7C087DAE

Fix this test case to actually use HMAC with hexkey:FFFF as intended.

Signed-off-by: Dimitri John Ledkov <dimitri.ledkov@surgut.co.uk>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/24068)

Document that private and pairwise checks are not bounded by key size

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/24049)

fuzz/decoder.c: Limit the key sizes on which checks are run

In particular the DH safe prime check will be limited to 8192 bits
and the private and pairwise checks are limited to 16384 bits on
any key types.

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/24049)

Fix socket descriptor checks on Windows

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24035)

Fix typos found by codespell

Reviewed-by: Todd Short <todd.short@me.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24013)

man EVP_PKEY_CTX_set_params: document params is a list

Signed-off-by: Hubert Kario <hkario@redhat.com>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23986)

aarch64: fix BTI in bsaes assembly code

Change-Id: I63f0fb2af5eb9cea515dec96485325f8efd50511

Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/23982)

Ensure proper memory barriers around ossl_rcu_deref/ossl_rcu_assign_ptr

Since the addition of macos14 M1 runners in our CI jobs we've been
seeing periodic random failures in the test_threads CI job.
Specifically we've seen instances in which the shared pointer in the
test (which points to a monotonically incrementing uint64_t went
backwards.

From taking a look at the disassembled code in the failing case, we see
that __atomic_load_n when emitted in clang 15 looks like this
0000000100120488 <_ossl_rcu_uptr_deref>:
100120488: f8bfc000     ldapr   x0, [x0]
10012048c: d65f03c0     ret

Notably, when compiling with gcc on the same system we get this output
instead:
0000000100120488 <_ossl_rcu_uptr_deref>:
100120488: f8bfc000     ldar   x0, [x0]
10012048c: d65f03c0     ret

Checking the arm docs for the difference between ldar and ldapr:
https://developer.arm.com/documentation/ddi0602/2023-09/Base-Instructions/LDAPR--Load-Acquire-RCpc-Register-
https://developer.arm.com/documentation/dui0802/b/A64-Data-Transfer-Instructions/LDAR

It seems that the ldar instruction provides a global cpu fence, not
completing until all writes in a given cpus writeback queue have
completed

Conversely, the ldapr instruction attmpts to achieve performance
improvements by honoring the Local Ordering register available in the
system coprocessor, only flushing writes in the same address region as
other cpus on the system.

I believe that on M1 virtualized cpus the ldapr is not properly ordering
writes, leading to an out of order read, despite the needed fencing.
I've opened an issue with apple on this here:
https://developer.apple.com/forums/thread/749530

I believe that it is not safe to issue an ldapr instruction unless the
programmer knows that the Local order registers are properly configured
for use on the system.

So to fix it I'm proposing with this patch that we, in the event that:
1) __APPLE__ is defined
AND
2) __clang__ is defined
AND
3) __aarch64__ is defined

during the build, that we override the ATOMIC_LOAD_N macro in the rcu
code such that it uses a custom function with inline assembly to emit
the ldar instruction rather than the ldapr instruction.  The above
conditions should get us to where this is only used on more recent MAC
cpus, and only in the case where the affected clang compiler emits the
offending instruction.

I've run this patch 10 times in our CI and failed to reproduce the
issue, whereas previously I could trigger it within 5 runs routinely.

Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23974)

Downgrade also the download-artifact action

It has to have the same version as upload-artifact.

Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/24065)

ffc/ffc_params_generate.c: Add the check for the EVP_MD_get_size()

Add the check for the EVP_MD_get_size() to avoid invalid negative numbers.

Fixes: 4f2271d58a ("Add ACVP fips module tests")
Signed-off-by: Jiasheng Jiang <jiasheng@purdue.edu>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23970)

ts/ts_rsp_sign.c: Add the check for the EVP_MD_CTX_get_size()

Add the check for the return value of EVP_MD_CTX_get_size() to avoid invalid negative numbers.

Fixes: c7235be6e3 ("RFC 3161 compliant time stamp request creation, response generation and response verification.")
Signed-off-by: Jiasheng Jiang <jiasheng@purdue.edu>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23960)