Pixhawk 6C external I2C bus hack

[julianoes]

This contains two changes:

1. This swaps the internal baro and mag back to the bus which is both internal an external but configured as external for this case, in the same way as Update fmu-v6c rc.board_sensors #20151.

2. Also, it adds a hack for fmu-v6c regarding mag calibration: In order for the mag calibration to treat the mag as an internal one, as well as sensors_status_mag and sensors_status_baro to be correct, we need to insert this override to change these two sensors specifically to internal.

FYI @vincentpoont2

Tests:

Mag calibration:

...
MAG 396321 EN: 1, offset: [-0.039 -0.009 -0.025], scale: [0.991 1.069 0.981], Internal

sensor status topics:

nsh> listener sensors_status_mag

TOPIC: sensors_status_mag
 sensors_status_mag
    timestamp: 233004294 (0.008210 seconds ago)
    device_id_primary: 396321 (Type: 0x06, I2C:4 (0x0C))
    device_ids: [396321, 0, 0, 0]
    inconsistency: [0.0000, nan, nan, nan]
    healthy: [True, False, False, False]
    priority: [50, 0, 0, 0]
    enabled: [True, False, False, False]
    external: [False, False, False, False]

nsh> listener sensors_status_baro

TOPIC: sensors_status_baro
 sensors_status_baro
    timestamp: 241376486 (0.003908 seconds ago)
    device_id_primary: 4028193 (Type: 0x3D, I2C:4 (0x77))
    device_ids: [4028193, 0, 0, 0]
    inconsistency: [0.0000, nan, nan, nan]
    healthy: [True, False, False, False]
    priority: [50, 0, 0, 0]
    enabled: [True, False, False, False]
    external: [False, False, False, False]

[dagar]

In terms of something we can live with, rather than a board define slipping into common code maybe we could flip it around so that DeviceExternal() respects something like an override list (or call) that the board support could optionally provide?

[vincentpoont2]

If this goes in, then we should we remove this #20198 ?

[julianoes]

@dagar I've updated it, have another look. I believe this is cleaner and doesn't have the board specific ifdef.

[julianoes]

@vincentpoont2 yes, this replaces #20198, and #20151.

[vincentpoont2]

This #20198 just got merged and reverted fmuv6c I2C4 back to "internal" so that the board would start Baro & Mag automatically using Main FW.

PX4-Autopilot/boards/px4/fmu-v6c/src/i2c.cpp

Line 39 in 1080855

initI2CBusInternal(4),

@julianoes Does this need to move back to "external" with this hack?

[davids5]

LGTM

[davids5]

In terms of something we can live with, rather than a board define slipping into common code maybe we could flip it around so that DeviceExternal() respects something like an override list (or call) that the board support could optionally provide?

It may need to be expanded later to be a call out that the board could override but that is not worth the code space now.

[julianoes]

@vincentpoont2 yes, it needs to go back to external.

[vincentpoont2]

Everything look good on our test with this PR. @julianoes

Board sensors: /etc/init.d/rc.board_sensors
bmi055_accel #0 on SPI bus 1 rotation 4
bmi055_gyro #0 on SPI bus 1 rotation 4
icm42688p #0 on SPI bus 1 rotation 6
ms5611 #0 on I2C bus 4 (external) address 0x77
ist8310 #0 on I2C bus 4 (external) address 0xC
ist8310 #1 on I2C bus 1 (external) address 0xE rotation 10
ms4525do #0 on I2C bus 4 (external) address 0x28

[davids5]

Looks good to me

[julianoes]

@dagar I've addressed your two comments, fixed up the original commit, rebased on master and force pushed.

[dagar]

That looks good, I was just hoping to keep it a little more consistent with px4_i2c_bus_external.

How's this?
338ca2c

I wanted to make sure we capture the same notion of external consistently (I2C::external(), etc).

[julianoes]

@dagar your "hack" is much nicer. I will revert my changes, cherry-pick it here, and test it.

[julianoes]

Tested, the calibration is correct, and the sensors_status_mag/sensors_status_baro topics as well.

[davids5]

Much cleaner. What is the CI failure?

[julianoes]

@davids5 looks unrelated. I'll create a separate PR to fix it.

/Users/runner/work/PX4-Autopilot/PX4-Autopilot/platforms/posix/src/px4/common/lockstep_scheduler/src/lockstep_scheduler.cpp:53:55: fatal error: format specifies type 'unsigned long' but the argument has type 'uint64_t' (aka 'unsigned long long') [-Wformat]
                PX4_INFO("setting initial absolute time to %lu us", time_us);

[julianoes]

@davids5 nevermind, it's already fixed: c707997.