Handle interrupt acknowledge register in Cortex-A53 SRE port

[sviaunxp]

Description

Let the FreeRTOS IRQ handler properly store and restore the ICCIAR register value around the vApplicationIRQHandler() call.

This pull request is essentially created so that the "SRE" port matches the original ("MMIO") port [1] so that vApplicationIRQHandler() does not depend on the port being used.

[1] https://github.com/FreeRTOS/FreeRTOS-Kernel/blob/main/portable/GCC/ARM_CA53_64_BIT/portASM.S#L310

Test Steps

Tested with an application - with the same implementation of vApplicationIRQHandler() - for both A53 ports.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[aggarg]

Whether to use ICC_IAR1_EL1 and ICC_IAR0_EL1 depends on whether the interrupt being handled is configured as Group 1 or Group 0 interrupt. This information is best known to the application and therefore, the application interrupt handler should acknowledge and mark completion of the interrupt. What do you think?

[sviaunxp]

Group 0 interrupts are always signaled as FIQs [2].
Since we are in FreeRTOS_IRQ_Handler(), we are only handling IRQs (as opposed to FIQs).
Hence, there is no reason to access ICC_IAR0_EL1 here, IMO.

[2] https://developer.arm.com/documentation/198123/0300/Arm-CoreLink-GIC-fundamentals

[codecov]

Codecov Report

Merging #392 (0949268) into main (a030d0a) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #392   +/-   ##
=======================================
  Coverage   92.13%   92.13%           
=======================================
  Files           4        4           
  Lines        1272     1272           
  Branches      342      342           
=======================================
  Hits         1172     1172           
  Misses         53       53           
  Partials       47       47           

Flag	Coverage Δ
unittests	92.13% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a030d0a...0949268. Read the comment docs.

[sviaunxp]

Hi @aggarg
Any feedback on this?
We should at least align both implementations of the Cortex-A53 ports in order to run the same application seamlessly on one or the other...

[aggarg]

Hi @sviaunxp , Apologies for missing your earlier reply. You are right that Group 0 interrupts are always signaled as FIQs. Though, in its current form, the same function can be used as both IRQ and FIQ handler? What is your motivation to do these operations in the port layer?

My motivation was to keep the interrupt grouping information out of the port layer. Let me check with others in my team and get back to you.

Thanks.

[sviaunxp]

Hi @aggarg ,

Thank you for your response.

My motivation is the following:

1. keep the application's IRQ handler as simple as possible
   (assuming FreeRTOS_IRQ_Handler() is only used to handle IRQs)
2. keep the SRE port behavior similar to the original one ("MMIO" based) in order to keep consistency across both ports.
   (hence have the same application handler to work seamlessly on either)

Feel free to touch base with your team and get back to me when you can.

[sviaunxp]

Hi @aggarg,
Have you got a chance to discuss this internally?

Thank you for your feedback.

[aggarg]

Hi @sviaunxp,

Yes and we decided to take this change. I will test it and merge it soon. Thank you for your contribution.

-Gaurav

[sviaunxp]

Hi Gaurav,

Thank you for accepting and testing this patch.
Don't hesitate to get back to me if ever you see regressions.

BR,
Stéphane.