Signed-off-by: Ronald Cron <[email protected]>

Add ALPN information in session tickets

Add testing for concurrently loading/using/destroying the same key

Expand MSVC to Visual Studio and announce the moving of the solution
files.

Signed-off-by: Bence Szépkúti <[email protected]>

All supported versions of Visual Studio support AESNI, so drop the
version number.

Signed-off-by: Bence Szépkúti <[email protected]>

Signed-off-by: Waleed Elmelegy <[email protected]>

Signed-off-by: Waleed Elmelegy <[email protected]>

…ompat-tests

TLS 1.3: Resumption and early data compatibility tests

Test the behavior of mbedtls_pk_get_psa_attributes() and
mbedtls_pk_import_into_psa() with respect to lifetime. In particular, test
that they work with persistent keys as documented.

Test cases generated by the following script:
```
for old in [('transparent', '0:0:1'),
            ('opaque volatile [export]', '1:0:1'),
            ('opaque volatile [copy]', '1:0:0'),
            ('opaque persistent [export]', '1:1:1'),
            ('opaque persistent [copy]', '1:1:0')]:
    for to_public in [('pair', '0'),
                      ('public', '1')]:
        for to_persistent in [('volatile', '0'),
                              ('persistent', '1')]:
            depends = ('\ndepends_on:MBEDTLS_USE_PSA_CRYPTO'
                       if old[0].startswith('opaque')
                       else '')
            print(f"""\
PSA import into PSA: {old[0]} -> {to_persistent[0]} {to_public[0]}{depends}
pk_import_into_psa_lifetime:{old[1]}:{to_public[1]}:{to_persistent[1]}
""")
```

Signed-off-by: Gilles Peskine <[email protected]>

Drop Support for MSVC 2013, 2015 and Arm Compiler 5

Signed-off-by: Paul Elliott <[email protected]>

Signed-off-by: Ronald Cron <[email protected]>

…-nego-testing

TLS: Improve server version negotiation testing

Signed-off-by: Ronald Cron <[email protected]>

Signed-off-by: Ronald Cron <[email protected]>

Signed-off-by: Ronald Cron <[email protected]>

…eiving-early-data

Check ALPN when receiving early data

Catch potential invalid calls to init.

Signed-off-by: Paul Elliott <[email protected]>

Signed-off-by: Paul Elliott <[email protected]>

Signed-off-by: Paul Elliott <[email protected]>

TLS 1.3: Documentation update for 3.6 release

Adds missing transition and italicises internal functions

Signed-off-by: Ryan Everett <[email protected]>

A few typo fixes, extrapolations and extra details.

Signed-off-by: Ryan Everett <[email protected]>

Make PSA global_data thread safe

Signed-off-by: Paul Elliott <[email protected]>

…est-lifetime

pk_import_into_psa: test persistent keys

Co-authored-by: Paul Elliott <[email protected]>
Signed-off-by: Ryan Everett <[email protected]>

Add changelog entry for threading MVP

Signed-off-by: Dave Rodgman <[email protected]>

Signed-off-by: Dave Rodgman <[email protected]>

…ty-docs

Update psa-thread-safety.md to reflect version 3.6 changes

Signed-off-by: Ronald Cron <[email protected]>

Signed-off-by: Ronald Cron <[email protected]>

Signed-off-by: Ronald Cron <[email protected]>

Signed-off-by: Ronald Cron <[email protected]>

Signed-off-by: Dave Rodgman <[email protected]>

We were successful in adding transparent memory-poisoning testing, so
simplify to the real design decision we made.

Signed-off-by: David Horstmann <[email protected]>

The actual functions were called mbedtls_test_memory_poison()
and mbedtls_test_memory_unpoison(). Update the design section to
reflect this.

Signed-off-by: David Horstmann <[email protected]>

Since this is just an example, remove specific-sounding references to
mbedtls_psa_core_poison_memory() and replace with more abstract and
generic-sounding memory_poison_hook() and memory_unpoison_hook().

Signed-off-by: David Horstmann <[email protected]>

Signed-off-by: Paul Elliott <[email protected]>

Namely LOCAL_INPUT_DECLARE() and friends

Signed-off-by: David Horstmann <[email protected]>

Codestyle autogen fix

Signed-off-by: Dave Rodgman <[email protected]>

Signed-off-by: David Horstmann <[email protected]>

The configuration of memory poisoning is now performed via
compile-time detection setting MBEDTLS_MEMORY_CAN_POISON. Update
the design to take account of this.

Signed-off-by: David Horstmann <[email protected]>

Add a note that validation of validation was implemented in metatest.c
and explain briefly what that program is for.

Signed-off-by: David Horstmann <[email protected]>

Try to perform verify_ext() using the opaque context when the
key type is MBEDTLS_PK_RSASSA_PSS. This currently leads to a
crash while running the test suite and this will be fixed by
the next commit.

Signed-off-by: Valerio Setti <[email protected]>

…A PSS

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Dave Rodgman <[email protected]>

…_changelog

Add issues fixed to threading MVP changelog entry

This question has been resolved, as we know that we can test
transparently.

Signed-off-by: David Horstmann <[email protected]>

Explain this option and the way it relates to the copying macros.

Signed-off-by: David Horstmann <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

…g-design-doc

Rewrite PSA shared memory design document

Signed-off-by: Mingjie Shen <[email protected]>

…_sign_ext()

Signed-off-by: Valerio Setti <[email protected]>

Enable TLS 1.3 by default

Signed-off-by: Steven WdV <[email protected]>

[Bugfix] Fix null dereference in `mbedtls_pk_verify_ext()`

Fix potential stack buffer overread when
checking PSK binders.

Signed-off-by: Ronald Cron <[email protected]>

The behavior of the functions is kept intact. Changes concern:
- generate the initial PK context using PSA parameters only; this
  allows to remove 1 input parameter for the test function.
- add/fix comments.

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

…C are defined

This bug was not found until now because:
- !PK_[WRITE|PARSE]_C is only tested in component_full_no_pkparse_pkwrite()
- the test only case concerning RSA key had MBEDTLS_PK_WRITE_C as dependency
  so it was not executed in that component.

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

tls13: srv: Fix potential stack buffer overread

…psa_sign()

Previously only only PKCS1 v1.5 was tested.

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Minos Galanakis <[email protected]>

Use mbedtls_test_key_consistency_psa_pk() to verify that the
generated PK contexts match with the original PSA keys instead
of doing sign/verify and encrypt/decrypt.

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

… length

This is also used in pk_psa_sign() to properly size buffers holding
the public key.

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

…lopment-restricted-merge-19032024

 Merge development on restricted 10/03/2024

If the public key is exported with mbedtls_pk_write_pubkey_der()
it should be re-imported with mbedtls_pk_parse_public_key().
Alternative options (when PK_WRITE is not defined), i.e.
mbedtls_ecp_point_write_binary() and mbedtls_rsa_write_pubkey(),
export the key in a different format which cannot be parsed by
pk_parse module so mbedtls_ecp_point_read_binary() and
mbedtls_rsa_parse_pubkey() should be used respectively in this
case.

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Ronald Cron <[email protected]>

mbedtls_pk_setup_opaque always uses PKCS#1v1.5 for RSA keys

Update BRANCHES for 3.6

…ompat_sh

Fix fake cases listed of compat.sh

Signed-off-by: Minos Galanakis <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

Signed-off-by: Valerio Setti <[email protected]>

…on-tests-titles

ssl-opt.sh: Improve version selection test titles

Unify consistency tests for mbedtls_pk_import_into_psa and mbedtls_pk_copy_from_psa

The #ifdef guard in the get_builtin_key() should be
PSA_CRYPTO_ACCELERATOR_DRIVER_PRESENT to allow for
multiple drivers to be plugged into the wrapper.

Signed-off-by: Antonio de Angelis <[email protected]>

Signed-off-by: Minos Galanakis <[email protected]>

Signed-off-by: Minos Galanakis <[email protected]>

./scripts/bump_version.sh --version 3.6.0 --so-crypto 16 --so-x509 7  --so-tls 21

Signed-off-by: Minos Galanakis <[email protected]>

In case of opaque keys skip the check of the supported primary/enrollment
algorithms. Just try to perfom the signature and if the wrapped key
does not support RSA PSS the operation will fail automatically.

Signed-off-by: Valerio Setti <[email protected]>

…xt()

If the wrapped key has a PKCS1 v1.5 signature algorithm, then try
to call sign_ext() to perform PSA RSS. Of course this will fail
because it's not supported by the wrapped key.

Signed-off-by: Valerio Setti <[email protected]>

This also updates use-psa-crypto.md accordingly.

Signed-off-by: Valerio Setti <[email protected]>