wait_us without usticker

[mikisch81]

Description

Some targets (Mainly PSA secure targets running TF-M) have no usticker.
In order to support wait_us() in these targets (mbed_die() uses it for example), wait_us() is implemented using wait_ns() which does software-loop based delays.

• only last commit is relevant, other commits are from Add wait_ns API #9812.

Fixes #9853

Depends on #9812

Pull request type

[X] Fix
[ ] Refactor
[ ] Target update
[ ] Functionality change
[ ] Docs update
[ ] Test update
[ ] Breaking change

Reviewers

@kjbracey-arm

Release Notes

Some targets have no usticker. Fallback to use busy loop wait_ns() for cases like this.

[mikisch81]

@jeromecoutant

[ciarmcom]

@mikisch81, thank you for your changes.
@kjbracey-arm @Ronny-Liu @ARMmbed/mbed-os-test @ARMmbed/mbed-os-core @ARMmbed/mbed-os-maintainers please review.

[deepikabhavnani]

@mikisch81 - Please note #9812 is not verified for Cortex-M33 devices and #9853 issue is for Cortex-M33 devices

#9812 (comment)

[kjbracey]

If you are actually working on an M33, it would be awesome if you could confirm the timing.

I was about to suggest using the wait_ns test, but I now realise that it actually also assumes that you have USTICKER - it tests against us_ticker unconditionally and against lp_ticker if available.

Might have to use a stopwatch...

[mikisch81]

@kjbracey-arm do you have a cm33 device to test it?
I can try and run the tests on the Musca_a1.
@jeromecoutant can you test it on the L5 port?

[kjbracey]

do you have a cm33 device to test it?

No.

[mikisch81]

I was about to suggest using the wait_ns test, but I now realise that it actually also assumes that you have USTICKER - it tests against us_ticker unconditionally and against lp_ticker if available.

Might have to use a stopwatch...

The greentea tests are running in the non-secure part which has usticker.

[kjbracey]

Okay, tolerances are set very high (45%) on the test as checked in - to allow for flash wait states etc, so a "pass" isn't tight enough to show whether the scaling number is definitely correct. If you just reduce those numbers right down to a couple of percent, you should be able to confirm how tight it is - tell if the 5 cycles is correct.

[kjbracey]

You probably need to rebase this now that the wait_ns PR is in - the commits are still showing up here.

[0xc0170]

Rebased

[0xc0170]

Anyone from @ARMmbed/mbed-os-core for review?

[0xc0170]

Set to needs: CI to execute tests (will come most likely after reviews)

[deepikabhavnani]

Change looks good, only concern was for M33 which is fixed in LPC PR.
Not sure if this needs rebase

We might need to document somewhere that this change is only for secure side and US_TICKER is still needed for all targets

[0xc0170]

CI started

[mbed-ci]

Test run: SUCCESS

Summary: 13 of 13 test jobs passed
Build number : 1
Build artifacts