fix bidir dshot for NXP i.MXRT boards

[321thijs123]

Solved Problem

When attempting to use bidirectional dshot on a NXP MR-VMU-RT1176 Flight Controller (FMUv6X-RT) I found that the esc_status messages would never get published. This seems to be due to bdshot_recv_mask being used to check the number of ready erpms, while this mask seem to always be set to zero before this is done.

Fixes #25853

Solution

• Check bdshot_parsed_recv_mask instead of bdshot_recv_mask when counting the number of ready erpms.

Changelog Entry

For release notes:

Bugfix: esc_status never gets published for nxp xrt boards

[PetervdPerk-NXP]

Looks to be a regression from #23863 just tested on my hardware and this fixes it.
@321thijs123 could you remove the merge commit and use rebase instead?

@dakejahl I've found another regression caused by #23863 as well.
This code you added waits for all motors, I suppose this is for your round-robin code.

PX4-Autopilot/src/drivers/dshot/DShot.cpp

Lines 314 to 317 in cfe4cc8

	// We wait until all are ready.
	if (up_bdshot_num_erpm_ready() < _num_motors) {
	return 0;
	}

But this breaks Dshot based ESC esc_online_flags detection, i.e. motor unplugs esc_status doesn't get published anymore because up_bdshot_num_erpm_ready != _num_motors.

Could we rename up_bdshot_num_erpm_ready to up_bdshot_erpm_ready which is returns a boolean true/false and then you move your round-robin code to your underlying dshot driver. This keeps the esc_online_flags functionality intact.

[321thijs123]

@PetervdPerk-NXP I removed the merge commit and used rebase now.

[PetervdPerk-NXP]

LGTM, thank you for your contribution @321thijs123

[dakejahl]

But this breaks Dshot based ESC esc_online_flags detection, i.e. motor unplugs esc_status doesn't get published anymore because up_bdshot_num_erpm_ready != _num_motors.
Could we rename up_bdshot_num_erpm_ready to up_bdshot_erpm_ready which is returns a boolean true/false and then you move your round-robin code to your underlying dshot driver. This keeps the esc_online_flags functionality intact.

These issues are fixed in #25848. It's quite a huge PR that I doubt anyone will review. I'm still trying to figure out if it's possible to break it into pieces.

Going to merge this in the meantime to fix the issues now. Thanks!