STM32: TRNG: remove call to deprecated HAL_RNG_GetRandomNumber #5179
Conversation
LMESTM
force-pushed
the
trng_deprecated_call
branch
from
6deaa96 to
73d9da1
Compare
|
LGTM |
LMESTM
force-pushed
the
trng_deprecated_call
branch
from
73d9da1 to
b0854a3
Compare
targets/TARGET_STM/trng_api.c
Outdated
There was a problem hiding this comment.
line 36 - uses atomic access, not here when you are setting it?
There was a problem hiding this comment.
In the free case, I don't have to test the value, I just have to set it to unused, so I preferred to keep it simple: as there is only 1 user, once freed, it will always be 0.
An alternative is to throw an error in case of unbalanced, (like in sleep manager):
core_util_critical_section_enter();
if (users == 0) {
core_util_critical_section_exit();
error("Unbalanced number of trng free (< 0)");
}
core_util_atomic_decr_u16(&users, 1);
core_util_critical_section_exit();
but that seemed overkill to me ...
I can do the change anyway if people think it's of added value.
There was a problem hiding this comment.
OK, can be as it is.
Why init can be called only once - one owner limitation?
There was a problem hiding this comment.
yes - one HW only and only one client in MBED in features/mbedtls/platform/src/mbed_trng.c, so I'll stick to it for now.
|
ARM Internal Ref: IOTSSL-1790 |
|
bump for review |
|
Hello @0xc0170 Kind regards |
targets/TARGET_STM/trng_api.c
Outdated
There was a problem hiding this comment.
what about corner cases where output length will not be a multiple of 4? In this case you will overflow the buffer
There was a problem hiding this comment.
I thought I'd covered it, but clearly missed here - will fix - thx Ron !
LMESTM
force-pushed
the
trng_deprecated_call
branch
from
b0854a3 to
91c6b42
Compare
|
@RonEld thanks for review - I pushed an update to the code which should address the issue |
|
thanks again @RonEld |
|
@0xc0170 Could you please wait for my input? I'll have a look at it quickly. |
targets/TARGET_STM/trng_api.c
Outdated
There was a problem hiding this comment.
What happens if we call HAL_RNG_GenerateRandomNumber() without calling trng_init() first?
There was a problem hiding this comment.
HAL would return an error in such case.
targets/TARGET_STM/trng_api.c
Outdated
There was a problem hiding this comment.
I know that I am being paranoid, but could we please wipe the random[] buffer before returning.
There was a problem hiding this comment.
sure - will do so and post an update
There was a problem hiding this comment.
ok I made an update and now random is erased immediately after use.
Build : SUCCESSBuild number : 119 Triggering tests/test mbed-os |
Test : SUCCESSBuild number : 50 |
LMESTM
force-pushed
the
trng_deprecated_call
branch
from
91c6b42 to
8251ff4
Compare
Build : SUCCESSBuild number : 157 Triggering tests/test mbed-os |
targets/TARGET_STM/trng_api.c
Outdated
There was a problem hiding this comment.
I suspect this way most of the compilers would optimise the out the instructions zeroising the array, declaring random[] to be volatile might be necessary to prevent it.
There was a problem hiding this comment.
yet another good point - thanks !
done
HAL_RNG_GetRandomNumber is a deprecated API and replaced here with a call to HAL_RNG_GenerateRandomNumber. Doing so, we also rework the driver to use the 4 bytes returned by a call to HAL_RNG_GenerateRandomNumber instead of 1 byte out of 4. HAL_RNG_GenerateRandomNumber was not returning any error code, so now we can also check the return code.
There is only 1 RNG HW IP and we do not support more than one driver user at a time, so let's ensure this is the case and raise an error if needed.
LMESTM
force-pushed
the
trng_deprecated_call
branch
from
8251ff4 to
849749f
Compare
|
/morph build |
|
/morph build |
Build : SUCCESSBuild number : 289 Triggering tests/morph test |
Build : SUCCESSBuild number : 292 Triggering tests/morph test |
Test : SUCCESSBuild number : 131 |
Description
HAL_RNG_GetRandomNumber is a deprecated API and replaced here with a call to HAL_RNG_GenerateRandomNumber. Doing so, we also rework the driver to use the 4 bytes returned by a call to HAL_RNG_GenerateRandomNumber instead of 1 byte out of 4.
HAL_RNG_GenerateRandomNumber was not returning any error code, so now we can also check the return code.
A second commit ensures that there is a single user at a time.
Status
READY
Tests