65528 bits is more than any reasonable key until we start supporting
post-quantum cryptography.
This limit is chosen to allow bit-sizes to be stored in 16 bits, with
65535 left to indicate an invalid value. It's a whole number of bytes,
which facilitates some calculations, in particular allowing a key of
exactly PSA_CRYPTO_MAX_STORAGE_SIZE to be created but not one bit
more.
As a resource usage limit, this is arguably too large, but that's out
of scope of the current commit.
Test that key import, generation and derivation reject overly large
sizes.
Move the "core attributes" to a substructure of psa_key_attribute_t.
The motivation is to be able to use the new structure
psa_core_key_attributes_t internally.
Add a parameter to the key import method of a secure element driver to
make it report the key size in bits. This is necessary (otherwise the
core has no idea what the bit-size is), and making import report it is
easier than adding a separate method (for other key creation methods,
this information is an input, not an output).
Most driver methods are not allowed to modify the persistent data, so
the driver context structure contains a const pointer to it. Pass a
non-const pointer to the persstent data to the driver methods that
need it: init, allocate, destroy.
Pass the driver context to all driver methods except the ones that
operate on an already-setup operation context.
Rename `p_context` arguments to `op_context` to avoid confusion
between contexts.
This slightly increases storage requirements, but works in more use
cases. In particular, it allows drivers to treat choose slot numbers
with a monotonic counter that is incremented each time a key is
created, without worrying about overflow in practice.
Instead of having one giant table containing all possible methods,
represent a driver's method table as a structure containing pointers
to substructures. This way a driver that doesn't implement a certain
class of operations can use NULL for this class as a whole instead of
storing NULL for each method.
Technically we could have reused the old one for the new API, but then
we had to set an extra field during setup. The new version works when
all the fields that haven't been set explicitely are zero-initialised.
This change affects the psa_key_derivation_s structure. With the buffer
removed from the union, it is empty if MBEDTLS_MD_C is not defined.
We can avoid undefined behaviour by adding a new dummy field that is
always present or make the whole union conditional on MBEDTLS_MD_C.
In this latter case the initialiser macro has to depend on MBEDTLS_MD_C
as well. Furthermore the first structure would be either
psa_hkdf_key_derivation_t or psa_tls12_prf_key_derivation_t both of
which are very deep and would make the initialisation macro difficult
to maintain, therefore we go with the first option.
We want to make the PRF context structure depend on this flag, but
crypto_extra.h is included after crypto_struct.h and having the
option at its original place would not affect crypto_struct.h.
Add the compile time option PSA_PRE_1_0_KEY_DERIVATION. If this is not
turned on, then the function `psa_key_derivation()` is removed.
Most of the tests regarding key derivation haven't been adapted to the
new API yet and some of them have only been adapted partially. When this
new option is turned off, the tests using the old API and test cases
using the old API of partially adapted tests are skipped.
The sole purpose of this option is to make the transition to the new API
smoother. Once the transition is complete it can and should be removed
along with the old API and its implementation.
From the implementation point of view does not make much difference to
constrain the input order.
We constrain it because, this way the code is easier to review, the data
flow easier to understand and the implementations in general are easier
to validate.
Relative include paths should be avoided. The build system will
determine where to pull in includes from. Specifically, `#include
"../mbedtls/config.h"` shouldn't be used. Use `#include
"mbedtls/config.h` instead, so that the submodule-building makefiles can
change which directory to use to get mbedtls include files from.
Fixes#141
Remove the key creation functions from before the attribute-based API,
i.e. the key creation functions that worked by allocating a slot, then
setting metadata through the handle and finally creating key material.
Don't use "safe buffer size", because this it's somewhat misleading to
make it about safety: a buffer size that's too small will lead to a
runtime error, not to undefined behavior.
Convert the description of PSA_ALG_TLS12_PRF and
PSA_ALG_TLS12_PSK_TO_MS to the key derivation API that takes one input
at a time rather than the old {secret,salt,label} interface.
Define a new input category "seed".
PSA_KEY_ATTRIBUTES_INIT and psa_key_attributes_init weren't declared
in the API document, only defined in our implementation, but they are
referenced in the API document.
generate_key is a more classical name. The longer name was only
introduced to avoid confusion with getting a key from a generator,
which is key derivation, but we no longer use the generator
terminology so this reason no longer applies.
perl -i -pe 's/psa_generate_random_key/psa_generate_key/g' $(git ls-files)
“Tampering detected” was misleading because in the real world it can
also arise due to a software bug. “Corruption detected” is neutral and
more precisely reflects what can trigger the error.
perl -i -pe 's/PSA_ERROR_TAMPERING_DETECTED/PSA_ERROR_CORRUPTION_DETECTED/gi' $(git ls-files)
Move DSA-related key types and algorithms to the
implementation-specific header file. Not that we actually implement
DSA, but with domain parameters, we should be able to.
Parametrize finite-field Diffie-Hellman key types with a DH group
identifier, in the same way elliptic curve keys are parametrized with
an EC curve identifier.
Define the DH groups from the TLS registry (these are the groups from
RFC 7919).
Replicate the macro definitions and the metadata tests from elliptic
curve identifiers to DH group identifiers.
Define PSA_DH_GROUP_CUSTOM as an implementation-specific extension for
which domain parameters are used to specify the group.
Move psa_get_key_domain_parameters() and
psa_set_key_domain_parameters() out of the official API and declare
them to be implementation-specific extensions.
Expand the documentation of psa_set_key_domain_parameters() a bit to
explain how domain parameters are used.
Remove all mentions of domain parameters from the documentation of API
functions. This leaves DH and DSA effectively unusable.
Generators are mostly about key derivation (currently: only about key
derivation). "Generator" is not a commonly used term in cryptography.
So favor "derivation" as terminology.
This commit updates the function descriptions.
Generators are mostly about key derivation (currently: only about key
derivation). "Generator" is not a commonly used term in cryptography.
So favor "derivation" as terminology. Call a generator a key
derivation operation structure, since it behaves like other multipart
operation structures. Furthermore, the function names are not fully
consistent.
In this commit, I rename the functions to consistently have the prefix
"psa_key_derivation_". I used the following command:
perl -i -pe '%t = (
psa_crypto_generator_t => "psa_key_derivation_operation_t",
psa_crypto_generator_init => "psa_key_derivation_init",
psa_key_derivation_setup => "psa_key_derivation_setup",
psa_key_derivation_input_key => "psa_key_derivation_input_key",
psa_key_derivation_input_bytes => "psa_key_derivation_input_bytes",
psa_key_agreement => "psa_key_derivation_key_agreement",
psa_set_generator_capacity => "psa_key_derivation_set_capacity",
psa_get_generator_capacity => "psa_key_derivation_get_capacity",
psa_generator_read => "psa_key_derivation_output_bytes",
psa_generate_derived_key => "psa_key_derivation_output_key",
psa_generator_abort => "psa_key_derivation_abort",
PSA_CRYPTO_GENERATOR_INIT => "PSA_KEY_DERIVATION_OPERATION_INIT",
PSA_GENERATOR_UNBRIDLED_CAPACITY => "PSA_KEY_DERIVATION_UNLIMITED_CAPACITY",
); s/\b(@{[join("|", keys %t)]})\b/$t{$1}/ge' $(git ls-files)
In psa_import_key, change the order of parameters to pass
the pointer where the newly created handle will be stored last.
This is consistent with most other library functions that put inputs
before outputs.
In psa_generate_derived_key, change the order of parameters to pass
the pointer where the newly created handle will be stored last.
This is consistent with most other library functions that put inputs
before outputs.
Use individual setters for the id and lifetime fields of an attribute
structure, like the other attributes.
This commit updates the specification and adds an implementation of
the new setters.
Define a range of key identifiers for use by the application
(0..2^30-1), a range for use by implementations (2^30..2^31), and a
range that is reserved for future use (2^31..2^32-1).
Change the scope of key identifiers to be global, rather than
per lifetime. As a result, you now need to specify the lifetime of a
key only when creating it.
Declare algorithms for ChaCha20 and ChaCha20-Poly1305, and a
corresponding (common) key type.
Don't declare Poly1305 as a separate algorithm because it's a one-time
authenticator, not a MAC, so the API isn't suitable for it (no way to
use a nonce).
New macros PSA_AEAD_UPDATE_OUTPUT_SIZE, PSA_AEAD_FINISH_OUTPUT_SIZE
and PSA_AEAD_VERIFY_OUTPUT_SIZE to determine the output buffer sizes
for psa_aead_update(), psa_aead_finish() and psa_aead_verify().
Like psa_aead_finish(), psa_aead_verify() needs to produce output from
the last partial block of input if psa_aead_update() cannot produce
output byte by byte.
In psa_import_key and psa_copy_key, some information comes from the
key data (input buffer or source key) rather than from the attributes:
key size for import, key size and type and domain parameters for copy.
If an unused attribute is nonzero in the attribute structure, check
that it matches the correct value. This protects against application
errors.
There was a guarantee that psa_get_key_attributes() does not require a
subsequent psa_reset_key_attributes() to free resources as long as the
key was created with attributes having this property. This requirement
was hard to pin down because if a key is created with default
parameters, there are cases where it is difficult to ensure that the
domain parameters will be reported without allocating memory. So
remove this guarantee. Now the only case psa_reset_key_attributes() is
not required is if the attribute structure has only been modified with
certain specific setters.
Read extra data from the domain parameters in the attribute structure
instead of taking an argument on the function call.
Implement this for RSA key generation, where the public exponent can
be set as a domain parameter.
Add tests that generate RSA keys with various public exponents.
Change psa_get_domain_parameters() and psa_set_domain_parameters() to
access a psa_key_attributes_t structure rather than a key handle.
In psa_get_key_attributes(), treat the RSA public exponent as a domain
parameter and read it out. This is in preparation for removing the
`extra` parameter of psa_generate_key() and setting the RSA public
exponent for key generation via domain parameters.
In this commit, the default public exponent 65537 is not treated
specially, which allows us to verify that test code that should be
calling psa_reset_key_attributes() after retrieving the attributes of
an RSA key is doing so properly (if it wasn't, there would be a memory
leak), even if the test data happens to use an RSA key with the
default public exponent.
Instead of passing a separate parameter for the key size to
psa_generate_key and psa_generator_import_key, set it through the
attributes, like the key type and other metadata.
Types and functions that are not used in the attribute-based key
creation API are now implementation-specific extensions, kept around
until we finish transitioning to the new API.
Implement attribute querying.
Test attribute getters and setters. Use psa_get_key_attributes instead
of the deprecated functions psa_get_key_policy or
psa_get_key_information in most tests.
Implement the new, attribute-based psa_import_key and some basic
functions to access psa_key_attributes_t. Replace
psa_import_key_to_handle by psa_import_key in a few test functions.
This commit does not handle persistence attributes yet.
This commit starts a migration to a new interface for key creation.
Today, the application allocates a handle, then fills its metadata,
and finally injects key material. The new interface fills metadata
into a temporary structure, and a handle is allocated at the same time
it gets filled with both metadata and key material.
This commit was obtained by moving the declaration of the old-style
functions to crypto_extra.h and renaming them with the to_handle
suffix, adding declarations for the new-style functions in crypto.h
under their new name, and running
perl -i -pe 's/\bpsa_(import|copy|generator_import|generate)_key\b/$&_to_handle/g' library/*.c tests/suites/*.function programs/psa/*.c
perl -i -pe 's/\bpsa_get_key_lifetime\b/$&_from_handle/g' library/*.c tests/suites/*.function programs/psa/*.c
Many functions that are specific to the old interface, and which will
not remain under the same name with the new interface, are still in
crypto.h for now.
All functional tests should still pass. The documentation may have
some broken links.
This gives a little more room to encode key agreement algorithms,
while keeping enough space for key derivation algorithms.
This doesn't affect any of the already-defined algorithms.
Merge the Mbed Crypto development branch a little after
mbedcrypto-1.0.0 into the PSA Crypto API 1.0 beta branch a little
after beta 2.
Summary of merge conflicts:
* Some features (psa_copy_key, public key format without
SubjectPublicKeyInfo wrapping) went into both sides, but with a few
improvements on the implementation side. For those, take the
implementation side.
* The key derivation API changed considerably on the API side. This
merge commit generally goes with the updated API except in the tests
where it keeps some aspects of the implementation.
Due to the divergence between the two branches on key derivation and
key agreement, test_suite_psa_crypto does not compile. This will be
resolved in subsequent commits.
MBEDTLS_PSA_HAS_ITS_IO is not really useful since it doesn't actually
enable anything except the entropy seed file support, which only
requires the ITS interface and not a native implemetation. Remove it.
For DH, ECC (Weierstrass curves) and DSA, specify that the re-drawing
method is the one defined by NIST as
"key-pair generation by testing candidates", and describe it
unambiguously.
Also specify DES explicitly.
psa_generator_import_key() was only specified for "symmetric keys",
and there were some mistakes in the specification. Rewrite the
specification and extend it to other key types.
* For most private key types, specify that the function draws a byte
string repeatedly until the byte string is suitable.
* For DES, despite being a symmetric key type, re-drawing is
necessary.
* For Montgomery curves, despite being asymmetric, no re-drawing is
necessary.
* Specify the behavior for every standard key type other than RSA.
An implementation doesn't have to support all key types, but if it
does, it's better to have a standard.
crypto_extra.h has a dependency on platform_util.h for
MBEDTLS_DEPRECATED_NUMERIC_CONSTANT. Make the dependency explicit by
including platform_util.h. Although in most use cases the header should
already be included by something else, it doesn't hurt to include it
again and helps to clarify dependencies.
When building for the PSA crypto service (defined(PSA_CRYPTO_SECURE)),
define psa_key_owner_id_t as int32_t, which is how a PSA platform
encodes partition identity. Note that this only takes effect when the
build option MBEDTLS_PSA_CRYPTO_KEY_FILE_ID_ENCODES_OWNER is active.
Support this configuration in the ITS backend.
Declare the owner as psa_key_owner_id_t, of which an implementation
must be provided separately.
Make this a configuration option
MBEDTLS_PSA_CRYPTO_KEY_FILE_ID_ENCODES_OWNER, to make the conditional
compilation flow easier to follow. Declare it in config.h to
pacify check_names.sh.
Support for a specific implementation of psa_key_owner_id_t in storage
backends will come in a subsequent commit.
Differentiate between _key identifiers_, which are always `uint32_t`,
and _key file identifiers_, which are platform-dependent. Normally,
the two are the same.
In `psa/crypto_platform.h`, define `psa_app_key_id_t` (which is always
32 bits, the standard key identifier type) and
`psa_key_file_id_t` (which will be different in some service builds).
A subsequent commit will introduce a platform where the two are different.
It would make sense for the function declarations in `psa/crypto.h` to
use `psa_key_file_id_t`. However this file is currently part of the
PSA Crypto API specification, so it must stick to the standard type
`psa_key_id_t`. Hence, as long as the specification and Mbed Crypto
are not separate, use the implementation-specific file
`psa/crypto_platform.h` to define `psa_key_id_t` as `psa_key_file_id_t`.
In the library, systematically use `psa_key_file_id_t`.
perl -i -pe 's/psa_key_id_t/psa_key_file_id_t/g' library/*.[hc]
PSA spec now defines more generic PSA storage types instead of the ITS
specific ones. This is necessary in order to integrate with
the newer implementation of PSA ITS landing in Mbed OS soon.
Changes include the following:
- psa_status_t replaces psa_its_status_t
- psa_storage_info_t replaces psa_its_info_t
- psa_storage_uid_t replaces psa_its_uid_t
