How to detect infinite loop in program flow

[mailmail9116]

Posted on October 14, 2013 at 19:48

Hi ALL ,

I have a question that is bugging me lately and i could not find any good solution for the problem.

Suppose I have a function , that may or may not , depend on the input and the amount of fixed bugs ;) , to got stuck somewhere inside in an infinite loop .

Is there any way for the cortex to recover from this infinite loop , without doing a hardware reset ?

Thanks

Michael

[Tesla DeLorean]

Posted on October 14, 2013 at 20:01

Watchdog Interrupt?

[dthedens23]

Posted on October 14, 2013 at 21:32

Any function that may get stuck in an infinite loop should be coded with some kind of timeout or maximum number of re-try.

Better yet, do not baby sit the hardware with CPU expensive loops and polling hardware status bits.

Use the hardware to it's full advantage, by using interrupts and DMA.

Use I/O pins to driver external devices power (via a FET or some such) so that if a part stops responding, power cycle it.  (it may latch up from ESD events)

Use state machine drivers, if a device gets stuck in some state, then re-init it or power cycle it.

Non of the firmware that any vendor supplies is production worthy!  They provide examples of perfectly working code, for perfectly working hardware in a perfect world, which you an I know, it does not exist.

Read up on ''correct'' vs ''robust'' code.  I also learned a lot from Safety Critical Design Patterns

 

[emalund]

Posted on October 15, 2013 at 14:56

re watchdog interrupt.

the WD is NOT a fix, it is NOT a method, it is NOT a remedy.

If a device of my design creates a watchdo interrupt, it is recorded with all details and the unit report ''maintenance required''

do you want to have to use your fire extinguisher?

[Tesla DeLorean]

Posted on October 15, 2013 at 15:33

It does however represent the closest thing to a ''stupid code running'' detector.

The processor runs code blindly, it makes no cognitive decisions about whether a code sequence is sensible or not, that's the job of the coder.

[frankmeyer9]

Posted on October 15, 2013 at 16:11

the WD is NOT a fix, it is NOT a method, it is NOT a remedy.

 

It should of course not be used as such, but it is a proved method to deal with transient EMI problems. Sometimes, system integrity may get lost without SW designers's fault ...