Failure injection using mavlink and systemcmd

[julianoes]

Work in progress to implement failure injection using MAVLink (mavlink/mavlink#1331) as well as a systemcmd.

Unlike the SIM_BLOCK_... param this can be generic to SITL, HITL, reality.

[julianoes]

The tests fail here because mavlink/MAVSDK#1124 is not merged and released yet.

[swham86]

Are you guys still working on the failure injection? I think this feature is really cool.
I would like to point out some issues when I used this draft version by using SITL (jmavsim).

failure gyro off <-- it causes kernel panic
failure accel off <-- it causes kernel panic
failure battery off <-- not working
failure motor off <-- not working
failure rc_signal off <-- not working
failure mavlink_signal off <-- not working

I hope that the failure injection will be implemented soon! Thanks.

[julianoes]

@swham86 thanks for the interest and feedback. Yes I'm working on this, I was just on holiday last week.

What do you mean with kernel panic? Does your Linux system actually panic? Stopping gyro or accel won't work properly yet because the lockstep simulation depends on it and without it, everything stops. And the others you listed are not implemented yet, but I'm going to add more functionality soon.

[swham86]

@julianoes The kernel panic means just stopping simulation. I hope that the failure injection will be implemented soon. Thanks for your feedback!

[julianoes]

If anyone feels like reviewing, now would be the time. I plan to get this in as soon as possible. FYI @jkflying

[julianoes]

@dagar I think we can attempt to merge this.

[MaEtUgR]

I just saw this invalidated some of the documentation (JFYI @hamishwillee):
https://dev.px4.io/master/en/simulation/index.html#simulating-failsafes-and-sensorhardware-failure
https://dev.px4.io/master/en/simulation/failsafes.html#gps-loss

@julianoes Would it be possible to quickly change the param pointers to the new failsafe injection way?

[hamishwillee]

Thanks @MaEtUgR - any info on docs that need to be added here?

[MaEtUgR]

I knew how the old way worked not the new way with this pr. That's why I was even looking for the docs. I could take the effort and figure it out but if the people write it who have implemented it and already know, that would probably be much quicker and more accurate.

[hamishwillee]

Thanks @MaEtUgR

@julianoes On you > Would it be possible to quickly change the param pointers to the new failsafe injection way?

[julianoes]

@hamishwillee ignoring this right now, sorry.