[Bug] "differential_drive stop" produces Segmentation fault

[slgrobotics]

Describe the bug

When running a "_make px4_sitl gz_r1rover" (or _gzlawnmower) differential_drive module runs at startup by the rc.rover_differential_apps

If user wants to stop the module, by "_differentialdrive stop" - a Segmentation fault happens and whole PX4 program quits.

@PerFrivik @MaEtUgR @dagar - issue created per your request

To Reproduce

pxh> differential_drive stop

... Segmentation fault

Expected behavior

The module should stop and could be started again

Screenshot / Media

No response

Flight Log

no log needed

Software Version

current main

Flight controller

N/A - sitl

Vehicle type

Rover

How are the different components wired up (including port information)

No response

Additional context

No response

[MaEtUgR]

Thanks for the issue! I started testing during the dev call and so far I found:

• The module stops fine just like expected when running and stopping it e.g. in an gz_x500 simulation where it's not in the loop. But stoping the mc_rate_control module which is AFAIK in the path for lockstep also works fine there.
• When not calling the exit and cleanup here: https://github.com/PX4/PX4-Autopilot/blob/28db3e1c8ca97fd1bb6067be15c9caf19db8e735/src/modules/differential_drive/DifferentialDrive.cpp#L68 it stops without segfault but blocks the shell for several seconds doing so. It can then be started and stopped again but I'm not sure if that's without residues.

[PerFrivik]

Thanks, I'll take a look!

[matteodelseppia]

I think I fixed the segmentation fault. You just have to add a return statement after the exit_and_cleanup() (line 68 of DifferentialDrive.cpp), otherwise it will try to update the current parameters even if the memory for the module has already been freed.

    if (should_exit()) {
        ScheduleClear();
        exit_and_cleanup();
                return; // < add this
    }

Can I try to pull this?