Floating exception error caused by changing certain parameters' value

[ktskt]

Bug report

Issue details Hello, developers.

I found multiple floating point exception issue by changing parameters From my observation, buggy codes for each parameter are located in the similar places. (Although their locations are not identical) I put each parameter and value. And, I describe its error message as follows. I used some extreme values. But, if this exception happens, it indicates that there is a potential problem in the certain condition. So, I think it should be patched.

Also, I think it is necessary to put the range check against unreasonable values IMO.

1. param set PSC_VELXY_I -10000000000000000000000000 Program received signal SIGFPE, Arithmetic exception. 0x000000000045abed in Vector2::length() const () (gdb) info stack

   0 0x000000000045abed in Vector2::length() const ()

   1 0x00000000004b9455 in AC_PID_2D::get_i_shrink() ()

   2 0x00000000004b5361 in AC_PosControl::run_xy_controller(float, float) ()

   3 0x00000000004b5f44 in AC_PosControl::update_xy_controller(float) ()

   4 0x00000000004be38e in AC_WPNav::update_wpnav() ()

   5 0x000000000041ef52 in Copter::ModeGuided::takeoff_run() ()

   6 0x00000000004072ae in Copter::fast_loop() ()

   7 0x000000000046defe in AP_Scheduler::loop() ()

   8 0x0000000000406a30 in Copter::loop() ()

   9 0x00000000004d0e30 in HAL_SITL::run(int, char const, AP_HAL::HAL::Callbacks*) const ()

   10 0x000000000040339e in main ()

2. param set ATC_RAT_RLL_D -10000000000000000000000000 Program received signal SIGFPE, Arithmetic exception. 0x00000000004af251 in AC_AttitudeControl::control_monitor_update() () (gdb) info stack

   0 0x00000000004af251 in AC_AttitudeControl::control_monitor_update() ()

   1 0x000000000040727d in Copter::fast_loop() ()

   2 0x000000000046defe in AP_Scheduler::loop() ()

   3 0x0000000000406a30 in Copter::loop() ()

   4 0x00000000004d0e30 in HAL_SITL::run(int, char const, AP_HAL::HAL::Callbacks*) const ()

   5 0x000000000040339e in main ()

3. param set ATC_RAT_YAW_D -10000000000000000000000000 Program received signal SIGFPE, Arithmetic exception. 0x00000000004af2f2 in AC_AttitudeControl::control_monitor_update() () (gdb) info stack

   0 0x00000000004af2f2 in AC_AttitudeControl::control_monitor_update() ()

   1 0x000000000040727d in Copter::fast_loop() ()

   2 0x000000000046defe in AP_Scheduler::loop() ()

   3 0x0000000000406a30 in Copter::loop() ()

   4 0x00000000004d0e30 in HAL_SITL::run(int, char const, AP_HAL::HAL::Callbacks*) const ()

   5 0x000000000040339e in main ()

4. param set ATC_RAT_PIT_D -10000000000000000000000000 Program received signal SIGFPE, Arithmetic exception. 0x00000000004af2b7 in AC_AttitudeControl::control_monitor_update() () (gdb) info stack

   0 0x00000000004af2b7 in AC_AttitudeControl::control_monitor_update() ()

   1 0x000000000040727d in Copter::fast_loop() ()

   2 0x000000000046defe in AP_Scheduler::loop() ()

   3 0x0000000000406a30 in Copter::loop() ()

   4 0x00000000004d0e30 in HAL_SITL::run(int, char const, AP_HAL::HAL::Callbacks*) const ()

   5 0x000000000040339e in main ()

5. param set PSC_VELXY_P -10000000000000000000000000 Program received signal SIGFPE, Arithmetic exception. 0x00000000004b54df in AC_PosControl::run_xy_controller(float, float) () (gdb) info stack

   0 0x00000000004b54df in AC_PosControl::run_xy_controller(float, float) ()

   1 0x00000000004b5f44 in AC_PosControl::update_xy_controller(float) ()

   2 0x00000000004be38e in AC_WPNav::update_wpnav() ()

   3 0x000000000041ef52 in Copter::ModeGuided::takeoff_run() ()

   4 0x00000000004072ae in Copter::fast_loop() ()

   5 0x000000000046defe in AP_Scheduler::loop() ()

   6 0x0000000000406a30 in Copter::loop() ()

   7 0x00000000004d0e30 in HAL_SITL::run(int, char const, AP_HAL::HAL::Callbacks*) const ()

   8 0x000000000040339e in main ()

6. param set ATC_RAT_RLL_P -10000000000000000000000000 Program received signal SIGFPE, Arithmetic exception. 0x00000000004af229 in AC_AttitudeControl::control_monitor_update() () (gdb) info stack

   0 0x00000000004af229 in AC_AttitudeControl::control_monitor_update() ()

   1 0x000000000040727d in Copter::fast_loop() ()

   2 0x000000000046defe in AP_Scheduler::loop() ()

   3 0x0000000000406a30 in Copter::loop() ()

   4 0x00000000004d0e30 in HAL_SITL::run(int, char const, AP_HAL::HAL::Callbacks*) const ()

   5 0x000000000040339e in main ()

7. param set ATC_RAT_PIT_P -10000000000000000000000000 Program received signal SIGFPE, Arithmetic exception. 0x00000000004af28e in AC_AttitudeControl::control_monitor_update() () (gdb) info stack

   0 0x00000000004af28e in AC_AttitudeControl::control_monitor_update() ()

   1 0x000000000040727d in Copter::fast_loop() ()

   2 0x000000000046defe in AP_Scheduler::loop() ()

   3 0x0000000000406a30 in Copter::loop() ()

   4 0x00000000004d0e30 in HAL_SITL::run(int, char const, AP_HAL::HAL::Callbacks*) const ()

   5 0x000000000040339e in main ()

8. param set ATC_ANG_RLL_P -10000000000000000000000000 ERROR: Floating point exception - aborting (I changed parameter during AUTO mode)

9. param set ATC_ANG_PIT_P -10000000000000000000000000 ERROR: Floating point exception - aborting (I changed parameter during AUTO mode)

10. param set ATC_ANG_YAW_P -10000000000000000000000000 ERROR: Floating point exception - aborting (I changed parameter during AUTO mode)

11. param set ATC_RAT_RLL_FF -10000000000000000000000000 ERROR: Floating point exception - aborting (I changed parameter during AUTO mode)

12. param set ATC_RAT_YAW_FF -10000000000000000000000000 ERROR: Floating point exception - aborting (I changed parameter during AUTO mode)

Version ArduCopter 3.6.9rc2 (the most recent stable 3.6 version)

Platform [ ] All [ ] AntennaTracker [O ] Copter [ ] Plane [ ] Rover [ ] Submarine

Airframe type Quadcopter

Hardware type SITL

[Jaaaky]

Yes, I can confirm this on plane too. Caught one yesterday.

[rmackay9]

We discussed this on the dev call and a few points came out of this:

• this seems like the output from a Fuzz Test. These tests can be a useful way to find issues where bad user input could crash the system
• we should be careful to distinguish between a vehicle crash and an unrecoverable flight controller lock-up.
  • many parameter changes could cause a vehicle to crash. For example changing the AHRS_ORIENT to 8 (roll 180) will cause the EKF to think the vehicle is upside down and it will certainly crash. Similarly setting an attitude control gain to a very high number will often cause the vehicle to oscillate so badly that it crashes. Because AP supports such a wide variety of vehicles it is likely impossible to protect against all these possible failures.
  • we should never have an unrecoverable flight controller lock-up. This is a situation where the flight controller stops responding to input completely. A floating point exception could cause this but we handle these exceptions different in SITL vs a real-vehicle. To simulate what happens on a real-vehicle (in SITL) set the SIM_FLOAT_EXCEPT parameter to 0.

So @ktskt could you try testing again after setting SIM_FLOAT_EXCEPT = 0?

In general we are keen to find unrecoverable flight lock-ups but less interested in param changes that cause the vehicle to crash.

[IamPete1]

I'm going to close this, please re-open if your able to re-test with SIM_FLOAT_EXCEPT = 0. We would also be interested if you can give us some information on how you ran the testing, it might be something we cold include as part of AP's autotesting.

[figuernd]

Param testing seems to be an issue with pursuing. Just ran into an ardusub bug failing with a floating point exception.

After doing some gdb tracing found that AP_Logger_File::periodic_1Hz() relies on FILE_RATEMAX and DISARM_RATEMAX, which do not have defaults. Setting those to 0 fixed.