search

ARMmbed / mbed-os

Arm Mbed OS is a platform operating system designed for the internet of things
https://mbed.com
Other
4.68k stars 2.98k forks source link

TRNG Greentea Test requires manual RESET for target EP_AGORA #10765

Closed trowbridgec closed 5 years ago

trowbridgec commented 5 years ago

Description

While running the Mbed OS Greentea tests to validate adding the EP_AGORA target, I found that I had to manually press the RESET button during the tests-mbed_hal-trng test in order for it to pass (see PR #10629).

The 'Agora' (EP_AGORA) board does not have a DAPLink circuit on board; we debug and program it via an nRF52840 DK running a modified version of DAPLink that forces the interface chip to use the external debug port. I'm not positive, but that could be related?

Below I've attached some verbose logs of the test failing on GCC_ARM and IAR. It seems like the host test hangs waiting after sending the 2nd sync. If I manually RESET the target right after the sync is sent, the test continues and passes without issue.

Issue request type

[ ] Question
[ ] Enhancement
[x] Bug
0xc0170 commented 5 years ago

This looks like the interface/host test problem , not related to trng - however do not understand how this is not seen in other tests (the start sequence is different, should not be prior sync packet?).

ciarmcom commented 5 years ago

Internal Jira reference: https://jira.arm.com/browse/MBOCUSTRIA-1261

linlingao commented 5 years ago

The TRNG test initiates a hardware reset in the middle of the test to verify the randomly generated data is indeed random. By default, it requires DAPLink to intercept break to initiate the reset. If there's no DAPLink, it will fail. If it hangs after the 2nd sync, it typically means the reset did not go through. There is another alternative to initiates reset using pyOCD but that requires some hack.

0Grit commented 5 years ago

@flit any insight?

flit commented 5 years ago

I think @linlingao explained it. If the serial port is not going through DAPLink, then the serial break sent by the test host will not be intercepted to reset the target.

0Grit commented 5 years ago

Sounds like the hack method might be less desirable than getting hard reset working with the current or other daplink adapter.

linlingao commented 5 years ago

Yeah, I would recommend getting DAPLink to work since it's an integral part of Mbed.

trowbridgec commented 5 years ago

That makes sense because we are using a separate UART for communicating with the DUT (utilizing the mbedls mocking/override functionality).

linlingao commented 5 years ago

@trowbridgec Do you mind closing this?

0Grit commented 5 years ago

@linlingao @trowbridgec is PTO today.

Go ahead and close. Since we have debugger on a baseboard board, it is resolvable without source changes.