Merge a development version of Mbed TLS 2.16.0 that doesn't have
parameter validation into development.
The following conflicts were resolved:
- Update ChangeLog to include release notes merged from development so
far, with a version of "2.14.0+01b34fb316a5" and release date of
"xxxx-xx-xx" to show this is not a released version, but instead a
snapshot of the development branch equivalent to version of the 2.14.0
with additional commits from the mbedtls/development branch up through
01b34fb316 included. Entries added for unreleased versions of Mbed
Crypto remain at the top of the file for Mbed TLS 2.xx.x.
- Replace the Mbed Crypto version of
mbedtls_rsa_rsaes_pkcs1_v15_decrypt() with the version from Mbed TLS
which fixes timing variations and memory access variations that could
lead to a Bleichenbacher-style padding oracle attack. This will
prevent using psa_asymmetric_decrypt() with zero-length output buffers
until a follow up commit is made to restore this capability.
- In ssl_srv.c, include changes for both the new ECDH interface and
opaque PSK as already added to development previously.
Context: There are two public key writing functions in Mbed TLS. First,
mbedtls_pk_write_pubkey(), which exports a public key in the form of a
SubjectPublicKey structure containing the raw keying material
(for example, EC point coordinates for an EC public key, without
reference to the underlying curve). Secondly, mbedtls_pk_write_pubkey_der(),
which exports a public key in the form of a SubjectPublicKeyInfo structure,
wrapping the SubjectPublicKey structure by additional information
identifying the type of public key (and for ECC, e.g., it'd also contain
the ECC group identifier). The implementation of mbedtls_pk_write_pubkey_der()
calls mbedtls_pk_write_pubkey() first and then adds the corresponding
algorithm identifier wrapper.
Both of these functions need to be provided for PSA-based opaque PK contexts,
based on PSA's public key export function.
Previously, PSA used the SubjectPublicKeyInfo structure as its export format,
so mbedtls_pk_write_pubkey_der() could be easily implemented, while
mbedtls_pk_write_pubkey() would need to trim the output of the PSA export.
The previous implementation of mbedtls_pk_write_pubkey() is not quite right
because it calls PSA export doesn't do any trimming, hence exporting the large
SubjectPublicKeyInfo structure instead of the small SubjectPublicKey.
mbedtls_pk_write_pubkey_der(), in turn, immediately returns after calling
mbedtls_pk_write_pubkey(), hence also returning the SubjectPublicKeyInfo
structure, which is correct.
By now, the PSA public key export format has changed to the smaller
SubjectPublicKey structure. This means that, now, mbedtls_pk_write_pubkey()
can be implemented by just calling the PSA export, and that
mbedtls_pk_write_pubkey_der() needs to add the algorithm information around
it, just as in the other types of PK contexts. While not correct for the
old format, the existing code for mbedtls_pk_write_pubkey() is therefore
correct for the new PSA public key format, and needs no change apart from
the missing pointer shift in the last commit.
The implementation of mbedtls_pk_write_pubkey_der() needs a special code
path for PSA-based opaque PK contexts, as the PK context only contains
the PSA key handle, and the PSA API needs to be used to extract the
underlying EC curve to be able to write the AlgorithmParameter structure
that's part of the SubjectPublicKeyInfo structure.
That's what this commit does, (hopefully) making both
mbedtls_pk_write_pubkey() and mbedtls_pk_write_pubkey_der() export
the correctly formatted public key based on the new PSA public key format.
The SSL module accesses ECDH context members directly to print debug
information. This can't work with the new context, where we can't make
assumptions about the implementation of the context. This commit adds
new debug functions to complete the encapsulation of the ECDH context
and work around the problem.
We want to support alternative software implementations and we extend
the ECDH context to enable this. The actual functional change that makes
use of the new context is out of scope for this commit.
Changing the context breaks the API and therefore it has to be
excluded from the default configuration by a compile time flag.
We add the compile time flag to the module header instead of
`config.h`, because this is not a standalone feature, it only
enables adding new implementations in the future.
The new context features a union of the individual implementations
and a selector that chooses the implementation in use. An alternative
is to use an opaque context and function pointers, like for example the
PK module does it, but it is more dangerous, error prone and tedious to
implement.
We leave the group ID and the point format at the top level of the
structure, because they are very simple and adding an abstraction
layer around them away does not come with any obvious benefit.
Other alternatives considered:
- Using the module level replacement mechanism in the ECP module. This
would have made the use of the replacement feature more difficult and
the benefit limited.
- Replacing our Montgomery implementations with a new one directly. This
would have prevented using Montgomery curves across implementations.
(For example use implementation A for Curve448 and implementation B for
Curve22519.) Also it would have been inflexible and limited to
Montgomery curves.
- Encoding the implementation selector and the alternative context in
`mbedtls_ecp_point` somehow and rewriting `mbedtls_ecp_mul()` to
dispatch between implementations. This would have been a dangerous and
ugly hack, and very likely to break legacy applications.
- Same as above just with hardcoding the selector and using a compile
time option to make the selection. Rejected for the same reasons as
above.
- Using the PK module to provide to provide an entry point for
alternative implementations. Like most of the above options this
wouldn't have come with a new compile time option, but conceptually
would have been very out of place and would have meant much more work to
complete the abstraction around the context.
In retrospect:
- We could have used the group ID as the selector, but this would have
made the code less flexible and only marginally simpler. On the other
hand it would have allowed to get rid of the compile time option if a
tight integration of the alternative is possible. (It does not seem
possible at this point.)
- We could have used the same approach we do in this commit to the
`mbedtls_ecp_point` structure. Completing the abstraction around this
structure would have been a much bigger and much riskier code change
with increase in memory footprint, potential decrease in performance
and no immediate benefit.
In the future we want to support alternative ECDH implementations. We
can't make assumptions about the structure of the context they might
use, and therefore shouldn't access the members of
`mbedtls_ecdh_context`.
Currently the lifecycle of the context can't be done without direct
manipulation. This commit adds `mbedtls_ecdh_setup()` to complete
covering the context lifecycle with functions.
`mbedtls_ecp_tls_read_group()` both parses the group ID and loads the
group into the structure provided. We want to support alternative
implementations of ECDH in the future and for that we need to parse the
group ID without populating an `mbedtls_ecp_group` structure (because
alternative implementations might not use that).
This commit moves the part that parses the group ID to a new function.
There is no need to test the new function directly, because the tests
for `mbedtls_ecp_tls_read_group()` are already implicitly testing it.
There is no intended change in behaviour in this commit.
Revived from a previous PR by Gilles, see:
https://github.com/ARMmbed/mbedtls/pull/1293/files#diff-568ef321d275f2035b8b26a70ee9af0bR71
This will be useful in eliminating temporary stack buffers for transcoding the
signature: in order to do that in place we need to be able to make assumptions
about the size of the output buffer, which this macro will provide. (See next
commit.)
It's better for names in the API to describe the "what" (opaque keys) rather
than the "how" (using PSA), at least since we don't intend to have multiple
function doing the same "what" in different ways in the foreseeable future.
Unfortunately the can_do wrapper does not receive the key context as an
argument, so it cannot check psa_get_key_information(). Later we might want to
change our internal structures to fix this, but for now we'll just restrict
opaque PSA keys to be ECDSA keypairs, as this is the only thing we need for
now. It also simplifies testing a bit (no need to test each key type).
While at it, clarify who's responsible for destroying the underlying key. That
can't be us because some keys cannot be destroyed and we wouldn't know. So
let's leave that up to the caller.
The sanity checking script tests/scripts/check-names.sh uses a
simple state machine paired with a sequence of `sed` commands to
extract enumeration constants from the code. This code, however,
doesn't work properly when using multiline comments in enumerations
such as recently done in the constants MBEDTLS_CIPHER_PSA_KEY_XXX.
This commit doesn't attempt to make check-names.sh more robust
but instead uses /* ... */ comment indicators in each comment line,
while silences check-names.sh.
Increasing the robustness of check-names.sh is instead tracked
in #2210.
For AEAD ciphers, the information contained in mbedtls_cipher_info
is not enough to deduce a PSA algorithm value of type psa_algorithm_t.
This is because mbedtls_cipher_info doesn't contain the AEAD tag
length, while values of type psa_algorithm_t do.
This commit adds the AEAD tag length as a separate parameter
to mbedtls_cipher_setup_psa(). For Non-AEAD ciphers, the value
must be 0.
This approach is preferred over passing psa_algorithm_t directly
in order to keep the changes in existing code using the cipher layer
small.
This commit implements the internal key slot management performed
by PSA-based cipher contexts. Specifically, `mbedtls_cipher_setkey()`
wraps the provided raw key material into a key slot, and
`mbedtls_cipher_free()` destroys that key slot.
This field determines whether a cipher context should
use an external implementation of the PSA Crypto API for
cryptographic operations, or Mbed TLS' own crypto library.
The commit also adds dummy implementations for the cipher API.
This commit adds a field `psk_opaque` to the handshake parameter
struct `mbedtls_ssl_handshake_params` which indicates if the user
has configured the use of an opaque PSK.
This commit adds two public API functions
mbedtls_ssl_conf_psk_opaque()
mbedtls_ssl_set_hs_psk_opaque()
which allow to configure the use of opaque, PSA-maintained PSKs
at configuration time or run time.
