TRNG on L4 pulls rug out from USB

Description

Target: STM32L4x
Scope: USB and TRNG co-existence

In system_clock.c, the USB/SDMMC/TRNG source clock is set to either MSI or PLLSAI1, depending on the clock source (HSE/HSI use PLLSAI1 for 48MHz source, and MSI uses MSI for 48MHz source).

trng_api.c does not take this into account, and incorrectly sets the 48MHz clock source, in contradiction to what it was set to in system_clock.c

Bug Reproduction

1. In json config, choose USE_PLL_MSI for clock source
2. Use trng functionality (such as mbedtls_hardware_poll function call)
3. USB operation ceases as soon as trng_init is called, as the 48MHz clock source is incorrectly changed to RCC_RNGCLKSOURCE_PLL, invalidating how it was set up in system_clock.c

Recommend a change in trng_api.c to choose the correct 48MHz source as it was set in system_clock.c, or remove this code as the 48MHz clock source was already set in system_clock.c, correct?

#if ((CLOCK_SOURCE) & USE_PLL_MSI)
    PeriphClkInitStruct.RngClockSelection = RCC_RNGCLKSOURCE_MSI;
#else
    PeriphClkInitStruct.RngClockSelection = RCC_RNGCLKSOURCE_PLLSAI1;
#endif

Issue request type

[ ] Question
[ ] Enhancement
[X] Bug

[ciarmcom]

Internal Jira reference: https://jira.arm.com/browse/MBOCUSTRIA-194

Specific targets system_clock.c:

https://github.com/ARMmbed/mbed-os/blob/master/targets/TARGET_STM/TARGET_STM32L4/TARGET_STM32L475xG/TARGET_DISCO_L475VG_IOT01A/system_clock.c

[bcostm]

Hi
Thanks for this analysis and reporting

I also agree to remove the TRNG clock setting in the trng_api.c file and add it in the system_clock.c file instead (like we do for the USB). Just maybe add a "#ifdef DEVICE_TRNG". I will send a PR so you can test it and comment.