Previously, PSA used SubjectPublicKeyInfo structures to serialize EC public keys.
This has recently been changed to using ECPoint structures instead, but the wrapper
making PSA ECDSA verification available through Mbed TLS' PK API hasn't yet been
adapted accordingly - which is what this commit does.
Luckily, Mbed TLS' PK API offers two functions mbedtls_pk_write_pubkey()
and mbedtls_pk_write_pubkey_der(), the latter exporting a SubjectPublicKeyInfo
structure and the former exporting an ECPoint structure in case of EC public
keys. For the adaptation of the ECDSA wrapper ecdsa_verify_wrap() it is therefore
sufficient to use mbedtls_pk_write_pubkey() instead of mbedtls_pk_write_pubkey_der().
Previously, PSA used SubjectPublicKeyInfo structures to serialize EC public keys.
This has recently been changed to using ECPoint structures instead, but the wrapper
making PSA ECDSA verification available through Mbed TLS' PK API hasn't yet been
adapted accordingly - which is what this commit does.
Luckily, Mbed TLS' PK API offers two functions mbedtls_pk_write_pubkey()
and mbedtls_pk_write_pubkey_der(), the latter exporting a SubjectPublicKeyInfo
structure and the former exporting an ECPoint structure in case of EC public
keys. For the adaptation of the ECDSA wrapper ecdsa_verify_wrap() it is therefore
sufficient to use mbedtls_pk_write_pubkey() instead of mbedtls_pk_write_pubkey_der().
The file oid.c had conditional inclusion of functions based on a config.h
define that belongs to X.509, which is backwards. For now, just include those
functions unconditionally and rely on the linker to garbage-collect them if
not used.
In the longer term X.509-specific functions are likely to be removed from
libmbedcrypto, but at this step the goal is to preserve the API (and even ABI)
of libmbedcrypto for as long as possible while separating the source trees of
Mbed Crypto and Mbed TLS.
As agreed during the workshop, temporarily move definitions to oid.h even if
they might not semantically belong here, as a short-term measure allowing to
build libmbecrypto on its own (without X.509 files present in the source tree)
but still provide all the things Mbed TLS currently expects, and more
specifically preserve the API and ABI exposed by libmbedtls.
Remove extra status handling code from psa_import_key_into_slot(). This
helps save a tiny amount of code space, but mainly serves to improve the
readability of the code.
Move pk-using code to inside psa_import_rsa_key(). This aligns the shape
of psa_import_rsa_key() to match that of psa_import_ec_private_key() and
psa_import_ec_public_key().
Remove front matter from our EC key format, to make it just the contents
of an ECPoint as defined by SEC1 section 2.3.3.
As a consequence of the simplification, remove the restriction on not
being able to use an ECDH key with ECDSA. There is no longer any OID
specified when importing a key, so we can't reject importing of an ECDH
key for the purpose of ECDSA based on the OID.
Use the PSA-native status type in psa_key_agreement_ecdh() in
preparation for us calling PSA functions (and not just Mbed TLS
functions) and still being able to return a psa_status_t (without having
to translate it to a Mbed TLS error and then back again).
Remove pkcs-1 and rsaEncryption front matter from RSA public keys. Move
code that was shared between RSA and other key types (like EC keys) to
be used only with non-RSA keys.
Remove the type and bits arguments to psa_allocate_key() and
psa_create_key(). They can be useful if the implementation wants to
know exactly how much space to allocate for the slot, but many
implementations (including ours) don't care, and it's possible to work
around their lack by deferring size-dependent actions to the time when
the key material is created. They are a burden to applications and
make the API more complex, and the benefits aren't worth it.
Change the API and adapt the implementation, the units test and the
sample code accordingly.
Add a new function mbedtls_asn1_write_named_bitstring() that removes
trailing 0s at the end of DER encoded bitstrings. The function is
implemented according to Hanno Becker's suggestions.
This commit also changes the functions x509write_crt_set_ns_cert_type
and crt_set_key_usage to call the new function as the use named
bitstrings instead of the regular bitstrings.
When MBEDTLS_PSA_CRYPTO_SPM is defined, the code is being built for SPM (Secure Partition Manager)
integration which separates the code into two parts: NSPE (Non-Secure Processing Environment) and SPE
(Secure Processing Environment). When building for the SPE, an additional header file should be included.
You can use PSA_ALG_ANY_HASH to build the algorithm value for a
hash-and-sign algorithm in a policy. Then the policy allows usage with
this hash-and-sign family with any hash.
Test that PSA_ALG_ANY_HASH-based policies allow a specific hash, but
not a different hash-and-sign family. Test that PSA_ALG_ANY_HASH is
not valid for operations, only in policies.
Previously we weren't initializing the freshly allocated ECP keypair
when importing private EC keys. This didn't seem to cause problems, at
least according to our current test coverage, but it's better to ensure
we don't have a partially initialized object by explicitly initializing
the keypair.
Return the error code if failed, instead of returning value `1`.
If not failed, return the call of the underlying function,
in `mbedtls_ecdsa_genkey()`.
Use `cmake -D CMAKE_BUILD_TYPE=Asan` rather than manually setting
`-fsanitize=address`. This lets cmake determine the necessary compiler
and linker flags.
With UNSAFE_BUILD on, force -Wno-error. This is necessary to build
with MBEDTLS_TEST_NULL_ENTROPY.
Use `cmake -D CMAKE_BUILD_TYPE=Asan` rather than manually setting
`-fsanitize=address`. This lets cmake determine the necessary compiler
and linker flags.
With UNSAFE_BUILD on, force -Wno-error. This is necessary to build
with MBEDTLS_TEST_NULL_ENTROPY.
Add new initializers for key policies and use them in our docs, example
programs, tests, and library code. Prefer using the macro initializers
due to their straightforwardness.
mbedtls_mpi_read_binary() calls memcpy() with the source pointer being
the source pointer passed to mbedtls_mpi_read_binary(), the latter may
be NULL if the buffer length is 0 (and this happens e.g. in the ECJPAKE
test suite). The behavior of memcpy(), in contrast, is undefined when
called with NULL source buffer, even if the length of the copy operation
is 0.
This commit fixes this by explicitly checking that the source pointer is
not NULL before calling memcpy(), and skipping the call otherwise.
Context: The function `mbedtls_mpi_fill_random()` uses a temporary stack
buffer to hold the random data before reading it into the target MPI.
Problem: This is inefficient both computationally and memory-wise.
Memory-wise, it may lead to a stack overflow on constrained devices with
limited stack.
Fix: This commit introduces the following changes to get rid of the
temporary stack buffer entirely:
1. It modifies the call to the PRNG to output the random data directly
into the target MPI's data buffer.
This alone, however, constitutes a change of observable behaviour:
The previous implementation guaranteed to interpret the bytes emitted by
the PRNG in a big-endian fashion, while rerouting the PRNG output into the
target MPI's limb array leads to an interpretation that depends on the
endianness of the host machine.
As a remedy, the following change is applied, too:
2. Reorder the bytes emitted from the PRNG within the target MPI's
data buffer to ensure big-endian semantics.
Luckily, the byte reordering was already implemented as part of
`mbedtls_mpi_read_binary()`, so:
3. Extract bigendian-to-host byte reordering from
`mbedtls_mpi_read_binary()` to a separate internal function
`mpi_bigendian_to_host()` to be used by `mbedtls_mpi_read_binary()`
and `mbedtls_mpi_fill_random()`.
The calls to cipher_finish didn't actually do anything:
- the cipher mode is always ECB
- in that case cipher_finish() only sets *olen to zero, and returns either 0
or an error depending on whether there was pending data
- olen is a local variable in the caller, so setting it to zero right before
returning is not essential
- the return value of cipher_finis() was not checked by the caller so that's
not useful either
- the cipher layer does not have ALT implementations so the behaviour
described above is unconditional on ALT implementations (in particular,
cipher_finish() can't be useful to hardware as (with ECB) it doesn't call any
functions from lower-level modules that could release resources for example)
Since the calls are causing issues with parameter validation, and were no
serving any functional purpose, it's simpler to just remove them.
Somehow, mbedtls_sha256_ret() is defined even if MBEDTLS_SHA256_ALT
is set, and it is using SHA256_VALIDATE_RET. The documentation should
be enhanced to indicate that MBEDTLS_SHA256_ALT does _not_ replace
the entire module, but only the core SHA-256 functions.
Somehow, mbedtls_sha512_ret() is defined even if MBEDTLS_SHA512_ALT
is set, and it is using SHA512_VALIDATE_RET. The documentation should
be enhanced to indicate that MBEDTLS_SHA512_ALT does _not_ replace
the entire module, but only the core SHA-512 functions.
Somehow, mbedtls_sha1_ret() is defined even if MBEDTLS_SHA1_ALT
is set, and it is using SHA1_VALIDATE_RET. The documentation should
be enhanced to indicate that MBEDTLS_SHA1_ALT does _not_ replace
the entire module, but only the core SHA-1 functions.
Document when a context must be initialized or not, when it must be
set up or not, and whether it needs a private key or a public key will
do.
The implementation is sometimes more liberal than the documentation,
accepting a non-set-up context as a context that can't perform the
requested information. This preserves backward compatibility.
The MPI_VALIDATE_RET() macro cannot be used for parameter
validation of mbedtls_mpi_lsb() because this function returns
a size_t.
Use the underlying MBEDTLS_INTERNAL_VALIDATE_RET() insteaed,
returning 0 on failure.
Also, add a test for this behaviour.
A 0-length buffer for the key is a legitimate edge case. Ensure that
it works, even with buf=NULL. Document the key and keylen parameters.
There are already test cases for parsing an empty buffer. A subsequent
commit will add tests for writing to an empty buffer.
Add checks for null pointers under MBEDTLS_CHECK_PARAMS.
In functions that perform operations with a context, only check if the
context pointer is non-null under MBEDTLS_CHECK_PARAMS. In the default
configuration, unconditionally dereference the context pointer.
In functions that query a context, support NULL as a
pointer-to-context argument, and return the same value as for a
context which has been initialized but not set up.
- The validity of the input and output parameters is checked by
parameter validation.
- A PRNG is required in public mode only (even though it's also
recommended in private mode), so move the check to the
corresponding branch.
The check was already done later when calling ECB, (as evidenced by the tests
passing, which have a call with data_unit set to NULL), but it's more readable
to have it here too, and more helpful when debugging.
Some of the documentation is obsolete in its reference to key slots
when it should discuss key handles. This may require a further pass,
possibly with some reorganization of error codes.
Update the documentation of functions that modify key slots (key
material creation and psa_set_key_policy()) to discuss how they affect
storage.
Move psa_load_persistent_key_into_slot,
psa_internal_make_key_persistent and psa_internal_release_key_slot to
the slot management module.
Expose psa_import_key_into_slot from the core.
After this commit, there are no longer any functions declared in
psa_crypto_slot_management.h and defined in psa_crypto.c. There are
still function calls in both directions between psa_crypto.c and
psa_crypto_slot_management.c.
Move the key slot array and its initialization and wiping to the slot
management module.
Also move the lowest-level key slot access function psa_get_key_slot
and the auxiliary function for slot allocation
psa_internal_allocate_key_slot to the slot management module.
Since Mbed TLS 2.10, there is a single copy of
mbedtls_platform_zeroize for the whole library instead of one per
module. Update the PSA crypto module accordingly.
This commit finishes the removal of support for direct access to key
slots in psa_crypto.c.
This marks the end of the necessary phase of the transition to key
handles. The code should subsequently be refactored to move key slot
management from psa_crypto.c to psa_crypto_slot_management.c.
This commit marks the beginning of the removal of support for direct
access to key slots. From this commit on, programs that use
psa_key_slot_t will no longer compile.
Subsequent commits will remove the now-unused legacy support in
psa_crypto.c.
The code only worked if psa_key_id_t (formerly psa_key_slot_t)
promoted to int and every value fit in int. Now the code only assumes
that psa_key_id_t is less wide than unsigned long, which is the case
since psa_key_id_t is a 32-bit type in our implementation.
Move the persistent storage implementation from psa_key_slot_t to
psa_key_id_t. For the most part, this just means changing the types of
function arguments.
Update the documentation of some functions to reflect the fact that
the slot identifier is purely a storage identifier and is not related
to how the slot is designated in memory.
Many places in the code called psa_remove_key_data_from_memory (which
preserves metadata for the sake of failues in psa_import_key) followed
by clearing the slot data. Use an auxiliary function for this.
Access the slot directly rather than going through psa_get_key_slot.
Unlike other places where key slots are accessed through
psa_get_key_slot, here, we know where all the slots are and there are
no policy or permission considerations.
This resolves a memory leak: allocated slots were not getting freed
because psa_get_key_slot rejected the attempt of accessing them
directly rather than via a handle.
Implement psa_allocate_key, psa_open_key, psa_create_key,
psa_close_key.
Add support for keys designated to handles to psa_get_key_slot, and
thereby to the whole API.
Allocated and non-allocated keys can coexist. This is a temporary
stage in order to transition from the use of direct slot numbers to
allocated handles only. Once all the tests and sample programs have
been migrated to use handles, the implementation will be simplified
and made more robust with support for handles only.
At the moment, the in-storage slot identifier is the in-memory slot
number. But track them separately, to prepare for API changes that
will let them be different (psa_open_key, psa_create_key).
The function called through the macro MBEDTLS_PARAM_FAILED() must be supplied by
users and makes no sense as a library function, apart from debug and test.
Function calls to alternative implementations have to follow certain
rules in order to preserve correct functionality. To avoid accidentally
breaking these rules we state them explicitly in the ECP module for
ourselves and every contributor to see.
We initialized the ECC hardware before calling
mbedtls_ecp_mul_shortcuts(). This in turn calls
mbedtls_ecp_mul_restartable(), which initializes and frees the hardware
too. This issue has been introduced by recent changes and caused some
accelerators to hang.
We move the initialization after the mbedtle_ecp_mul_shortcuts() calls
to avoid double initialization.
The SSL module accesses ECDH context members directly. This can't work
with the new context, where we can't make any assumption about the
implementation of the context.
This commit makes use of the new functions to avoid accessing ECDH
members directly. The only members that are still accessed directly are
the group ID and the point format and they are independent from the
implementation.
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.
The functionality from public API functions are moved to
`xxx_internal()` functions. The public API functions are modified to do
basic parameter validation and dispatch the call to the right
implementation.
There is no intended change in behaviour when
`MBEDTLS_ECDH_LEGACY_CONTEXT` is enabled.
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.
Add missing compilation guards that broke the build if either GCM or
CCM was not defined.
Add missing guards on test cases that require GCM or CBC.
The build and tests now pass for any subset of {MBEDTLS_CCM_C,
MBEDTLS_GCM_C}. There are still unused variables warnings if neither
is defined.
Refactor `mpi_write_hlp()` to not be recursive, to fix stack overflows.
Iterate over the `mbedtls_mpi` division of the radix requested,
until it is zero. Each iteration, put the residue in the next LSB
of the output buffer. Fixes#2190
Refactor mbedtls_ctr_drbg_update_seed_file and
mbedtls_hmac_drbg_update_seed_file to make the error logic clearer.
The new code does not use fseek, so it works with non-seekable files.
Add a function to configure entropy sources. For testing only.
Use it to test that the library initialization fails properly if there is no
entropy source.
There is a probability that r will be encoded as 31 or less bytes in DER,
so additional padding is added in such case.
Added a signature-part extraction function to tidy up the code further.
