Document the test strategy

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
This commit is contained in:
Gilles Peskine 2022-12-14 19:00:56 +01:00
parent d878d1c638
commit 071f47343b

View File

@ -3,6 +3,38 @@
* functions. Due to the complexity of how these functions are tested,
* we test all the layers in a single test suite, unlike the way other
* functions are tested with each layer in its own test suite.
*
* Test strategy
* =============
*
* There are three main goals for testing random() functions:
* - Parameter validation.
* - Correctness of outputs (well-formed, in range).
* - Distribution of outputs.
*
* We test parameter validation in a standard way, with unit tests with
* positive and negative cases:
* - mbedtls_mpi_core_random(): negative cases for mpi_core_random_basic.
* - mbedtls_mpi_mod_raw_random(), mbedtls_mpi_mod_random(): negative
* cases for mpi_mod_random_validation.
* - mbedtls_mpi_random(): mpi_random_fail.
*
* We test the correctness of outputs in positive tests:
* - mbedtls_mpi_core_random(): positive cases for mpi_core_random_basic,
* and mpi_random_many.
* - mbedtls_mpi_mod_raw_random(), mbedtls_mpi_mod_random(): tested indirectly
* via mpi_mod_random_values.
* - mbedtls_mpi_random(): mpi_random_sizes, plus indirectly via
* mpi_random_values.
*
* We test the distribution of outputs only for mbedtls_mpi_core_random(),
* in mpi_random_many, which runs the function multiple times. This also
* helps in validating the output range, through test cases with a small
* range where any output out of range would be very likely to lead to a
* test failure. For the other functions, we validate the distribution
* indirectly by testing that these functions consume the random generator
* in the same way as mbedtls_mpi_core_random(). This is done in
* mpi_mod_random_values and mpi_legacy_random_values.
*/
#include "mbedtls/bignum.h"