Fix GPIO AF settings being overridden by timerInit().

[hydra]

timerInit blindly reconfigures GPIO AF for all pins in the fullTimerHardware array regardless of if the timer pin is used already by something else.

For example, if it is called AFTER the SPI initilisation (e.g. after configureSPIBusses), then any AF settings for the SPI are overridden by timer AF, making the SPI hang when it's used.

This PR extracts the AF init code into timerIOInit and calls it early in the init code.

Targets like the SPRacingH7ZERO, which use PB3/PB4 for SPI3 gyro, now boot. With out this PR the target would hang on SPI3 access because the GPIO AF settings were reset from AF6 (SPI3) to AF1 and AF2 respectively.

[github-actions]

Do you want to test this code? You can flash it directly from Betaflight Configurator:

• Simply put #12886 (this pull request number) in the Select commit field of the Configurator firmware flasher tab (you need to Enable expert mode, Show release candidates and Development).

WARNING: It may be unstable. Use only for testing!

[hydra]

There are binaries for testing in betaflight/config#66 for the SPRacingH7ZERO and SPRacingH7NANO. The former of which requires this PR to boot.

[ledvinap]

Reasonable first step. But isn't it possible to init timers only when they are used.
Also, this PR does not affect non-hal times. Is that intentional?

[hydra]

Reasonable first step. But isn't it possible to init timers only when they are used. Also, this PR does not affect non-hal times. Is that intentional?

This is the simplest work-around to allow targets to boot until a BF maintainer, i.e. not me, comes up with a long-term solution. I looked for similar code in timer_stdperiph.c but didn't find it. I didn't check any other architectures either.

FYI: From what I can tell the code was broken in #12862 by @blckmn

The default was TIM_USE_ANY which was an alias for TIM_USE_NONE, and then when the check to see if the timer was actually used before initialising the AF was deleted everything broke.

See https://github.com/betaflight/betaflight/pull/12862/files#diff-7308a69a0b62f3e86b3f82fa62403d15d3f154261f03e1211fc04e879bdc8908L47-L48

The check that was removed was this:

for (unsigned timerIndex = 0; timerIndex < TIMER_CHANNEL_COUNT; timerIndex++) {
        const timerHardware_t *timerHardwarePtr = &TIMER_HARDWARE[timerIndex];
        if (timerHardwarePtr->usageFlags == TIM_USE_NONE) {
            continue; 
        }
        // XXX IOConfigGPIOAF in timerInit should eventually go away.
        IOConfigGPIOAF(IOGetByTag(timerHardwarePtr->tag), IOCFG_AF_PP, timerHardwarePtr->alternateFunction);
    }

See https://github.com/betaflight/betaflight/pull/12862/files#diff-e83e68d46b461bd090d5a593952e3ed2517b55f439dd79028e08d0fdf5890763L1034-L1036

Enjoy!

[blckmn]

I agree with @ledvinap. We can initialise the AF when the pin is first requested to be used with a timer.

I'm happy to clean this up given I broke it.

[hydra]

FYI, this PR is the quickest thing I could think of that would allow me to continue working on restoring the H7ZERO target. I didn't spend a lot of time on this PR, nor do I intend to.

Suggest merging as-is, and then @blckmn can do a follow-up PR which should be tested on the H7ZERO before being merged. This PR is tested on 4 H7 targets, binaries are in the linked issue.

[ledvinap]

@hydra : Can you add comments about workaround to code.

@blckmn : IMO it is best to merge this and then work on correct fix. The change is small

[blckmn]

Going to merge immediately as it fixes the issue, and I'll work on a longer term fix.

[blckmn]

Servos, PWM Motor outputs all seem to have the IOConfigGPIOAF already located within their respective init. It may be simply a case of finding any loose ends, and amending to remove the for loop entirely from HAL.

@ledvinap explains why it may have only been present for the HAL driver, a hangover.

[hydra]

@hydra : Can you add comments about workaround to code.

There's already comments in the init method in my PR. 😄

Thanks everyone for being pragmatic and getting this merged.