Executed ./scripts/bump_version.sh --version 2.25.0 --so-crypto 6
Increasing the SO version of the crypto library, because the openless
API improvement came with API/ABI incompatibilities. For example
- the size of psa_key_handle_t changed
- the type of a parameter in 18 public functions has changed from
psa_key_handle_t to mbedtls_svc_key_id_t
Signed-off-by: Janos Follath <janos.follath@arm.com>
T1 is set to a 2-limb value. The first operation that takes it as
input is mbedtls_mpi_mul_int, which makes it grow to 3 limbs. Later it
is shifted left, which causes it to grow again. Set its size to the
final size from the start. This saves two calls to calloc(), at the
expense of a slowdown in some operations involving T1 as input since
it now has more leading zeros.
Setting T1 to 3 limbs initially instead of 2 saves about 6% of the
calloc() calls in test_suite_ecp and does not incur a performance
penalty. Setting T1 to A->n + 2 limbs instead of 2 saves about 20% of
the calloc calls and does not cause a measurable performance
difference on my Linux PC.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Note a possible microoptimization in mbedtls_mpi_mul_hlp that I tried
in the hope of reducing the number of allocations, but turned out to
be counterproductive.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Rewrite mbedtls_mpi_mul_int to call mpi_mul_hlp directly rather than
create a temporary mpi object. This has the benefit of not performing
an allocation when the multiplication is in place (mpi operand aliased
with the result) and the result mpi is large enough.
This saves about 40% of the calloc() calls in test_suite_ecp. There is
no measurable performance difference on my Linux PC.
The cost is a few bytes in bignum.o.
When there is no aliasing, or when there is aliasing but the mpi
object needs to be enlarged, the performance difference is negligible.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
mbedtls_mpi_sub_abs systematically allocated a new mpi when the result
was aliased with the right operand (i.e. X = A - X). This aliasing
very commonly happens during ECP operations. Rewrite the function to
allocate only if the result might not fit otherwise.
This costs a few bytes of code size in bignum.o, and might make
mbedtls_mpi_sub_abs very very slightly slower when no reallocation is
done. However, there is a substantial performance gain in ECP
operations with Montgomery curves (10-20% on my PC).
test_suite_ecp drops from 1422794 to 1271506 calls to calloc().
This commit also fixes a bug whereby mbedtls_mpi_sub_abs would leak
memory when X == B (so TB was in use) and the result was negative.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Reduce the code size, stack consumption and heap consumption in
fix_negative by encoding the special-case subtraction manually.
* Code size: ecp_curves.o goes down from 7837B down to 7769 in a
sample Cortex-M0 build with all curves enabled. The savings come
from not having to set up C in INIT (which is used many times) and
from not having to catch errors in fix_negative.
* Stack consumption: get rid of C on the stack.
* Heap: mbedtls_mpi_sub_abs with destination == second operand would
make a heap allocation. The new code doesn't do any heap allocation.
* Performance: no measurable difference.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
If you pass a curve name to the benchmark program, the ECDH and ECDSA
benchmarks will only run for that particular curve. By default, all
curves are benchmarked.
To simplify the implementation, if you pass multiple curves, only the
last one will be benchmarked.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
This algorithm replaces the pre-existing stream cipher algorithms.
The underlying stream cipher is determined by the key type.
Signed-off-by: Bence Szépkúti <bence.szepkuti@arm.com>
If a random driver has a built-in entropy source and doesn't need an
external entropy source, make the driver author declare this
explicitly, rather than it being a less secure default.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
The get_random entry point was allowed to return partial data on both
PSA_SUCCESS and PSA_ERROR_INSUFFICIENT_ENTROPY, but there was no
meaningful difference between the two. Keep it simple: PSA_SUCCESS is
success but can be partial, and PSA_ERROR_INSUFFICIENT_ENTROPY is an
error with no output.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Explicitly recommend that the driver accounts for environmental
conditions that can affect the amount of entropy.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
A random generation driver does not need to support entropy injection.
This will limit it to platforms where the RNG peripheral is the sole
entropy source and without an RNG seed saved into persistent storage.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
If an RNG peripheral includes an entropy source, it would presumably
declare "initial_entropy_size" and "reseed_entropy_size" to be 0. In
this case, don't require the core to call "add_entropy".
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
The `get_entropy` entry point can be provided by multiple transparent
drivers, and the core will call all of them. But apart from that,
`get_entropy` doesn't involve an opaque key or a location, so it can
be in a transparent driver.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Transparent drivers may provide a DRBG interface through "add_entropy"
and "get_random" entry points. This interface may also be used with a
non-deterministic generator, for chips that include a TRNG.
Opaque driver may provide a "get_entropy" entry point.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Allow the core to call the "get_builtin_key" entry point to retrieve
the attributes of a built-in key. This is useful to implement
psa_get_key_attributes(), and also when the key data buffer's size
depends on the key type and size.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
In order to remove large buffers from the stack, the der data is written
into the same buffer that the pem is eventually written into, however
although the pem data is zero terminated, there is now data left in the
buffer after the zero termination, which can cause
mbedtls_x509_crt_parse to fail to parse the same buffer if passed back
in. Patches also applied to mbedtls_pk_write_pubkey_pem, and
mbedtls_pk_write_key_pem, which use similar methods of writing der data
to the same buffer, and tests modified to hopefully catch any future
regression on this.
Signed-off-by: Paul Elliott <paul.elliott@arm.com>
Having a time stamp identifying each revision of the document is
useful, but it's also a pain because it creates a conflict whenever
there are multiple pending changes at the same time. The gain isn't
worth the pain, so I'm removing the time stamp.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
