There was a check against the remaining size of the buffer, which used
to be correct, but was broken two commits ago when we started not just
copying the input but also adding to it.
Replace it with a check that the input length is not greater that what's
expected for this step. This guarantees we won't overflow the internal
buffer.
While at it, add an explicit cast to uint8_t when writing the length to
the buffer, so silence an MSVC warning. This cast is safe because we
checked that the length is no larger than 65 or 32 (depending on the
step), so in any case is fits in one byte.
This was found because some lengths had not been adjusted in the test
suite, and instead of failing cleanly, library code performed buffer
overflows. I'll fix the tests in the next commit.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
Introduce a unit test to test mbedtls_x509_get_name() and add a testcase
with a corrupt DER-encoded name that causes mbedtls_x509_get_name() to
have to cleanup things it is allocated. If it fails to do this, a memory
leak is detected under Asan builds.
Signed-off-by: David Horstmann <david.horstmann@arm.com>
This macro is specific to the Mbed TLS implementation and not part of
the public API, so it shouldn't used the PSA_ namespace.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
The format used by the mbedtls_ecjpake_xxx() APIs and that defined by
the PSA Crypto PAKE extension are quite different; the former is
tailored to the needs of TLS while the later is quite generic and plain.
Previously we only addressed some part of this impedance mismatch: the
different number of I/O rounds, but failed to address the part where the
legacy API adds some extras (length bytes, ECParameters) that shouldn't
be present in the PSA Crypto version. See comments in the code.
Add some length testing as well; would have caught the issue.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
mbedtls_x509_get_name() allocates a linked list of mbedtls_x509_name
structs but does not free these when there is an error, leaving the
caller to free them itself. Change this to cleanup these objects within
the function in case of an error.
Signed-off-by: David Horstmann <david.horstmann@arm.com>
When writing the new .data file, first write the new content, then replace
the target. This way, there isn't a temporary state in which the file is
partially written. This temporary state can be misleading if the build is
interrupted. It's annoying if you're watching changes to the output and the
changes appear as emptying the file following by the new version appearing.
Now interrupted builds don't leave a file that appears to be up to date but
isn't, and when watching the output, there's a single transition to the new
version.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
This commit codifies some conventions that result from the original design
goals and others that have emerged after starting the implementation.
* Value ranges
* Bignum parameter naming and ordering
* Sizes
* Aliasing and overlap
* Error handling
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
There was already a short introduction to _who_ should use each module, but
not to _what_ each module does.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Use mbedtls_test_read_mpi_core() to read the test data. Among other
benefits, X and Y are now allocated to their exact size, so analyzers (Asan,
Valgrind, Coverity, ...) have a chance of complaining if the tested function
overflows the buffer.
Remove TEST_CF_PUBLIC calls which are no longer necessary.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
When testing a function that is supposed to be constant-flow, we declare the
inputs as constant-flow secrets with TEST_CF_SECRET. The result of such a
function is itself a constant-flow secret, so it can't be tested with
comparison operators.
In TEST_EQUAL, TEST_LE_U and TEST_LE_S, declare the values to be compared as
public. This way, test code doesn't need to explicitly declare results as
public if they're only used by one of these macros.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
This comment (which used to be attached to the implementation, and should
not have been moved to the header file) is incorrect: the library function
mbedtls_mpi_read_string preserves leading zeros as desired, but does not
create a zero-limb object for an empty string.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
The length in bytes is used with the 'len' prefix and
the length in limbs is used with the 'limbs' prefix.
Signed-off-by: Gabor Mezei <gabor.mezei@arm.com>
