🚀 go-pugleaf

In article <sbe4gINN2a9@agate.berkeley.edu>, andywang@crown.berkeley.edu (Andrew Wang) writes:
> Hello,
>     I am having a problem with keyboard interrupts crashing my
> program.  It is a control program with a real-time hardware interrupt
> driven at 39 hz. Everything works fine until a) the keyboard buffer
> fills up and the computer tries to beep or b) sometimes a single keypress 
> will cause it to crash.
> 
> This is what I have in my (QuickC) interrupt handler:
> 
> void _far _interrupt() 
> {
>    reset_timer();		/* Reset hardware timer for next interrupt */
>    clr_int();			/* Clear interrupt request from timer */
>    outp(8259control,0x20);	/* Re-enable 8259 for next interrupt? */
>    _enable();
> 
> 
>    (body of interrupt handler)
> 
> 
> }
> 
> The interrupt handler takes maybe 1 - 4 milliseconds to complete.
> The program works on a 386-33, but neither of the AMI bios 486's
> that we have (25,33 mhz).
> 
> Here are the things I've tried:
> 
>    Removing _enable() will cause the program to crash after a certain amount
>    of time. This is possibly due to other interrupts not getting serviced? 
>    Refresh?
> 
>    Ignoring the keyboard entirely (not fetching any keys) will cause the
>    computer to crash when the buffer fills up and beeps.  Quick C says never
>    to use INT10 to beep when the video BIOS routines are running, otherwise a 
>    crash may occur. My main routine is very video-intensive, but I never TRY
>    to make the computer beep.  The interrupt handler does not touch video.
> 
>    Hmm, I tried some other things but now I can't remember!
> 
> Does anyone know how to disable the keyboard interrupt? I would like to try
> to disable it during my interrupt handler. I'll be sure to post a follow-up
> if this works!
> 
> Thanks in advance for any help you might be able to provide me!!!!
> 
> Andrew
> (arrgh!)

	I share your frustration!  I had the same problem with a
	routine to do reaction times.  In it, the speaker was being
	driven under program control and as soon as the respondent
	pressed the key (the object of the whole routine), the 486
	locked up.  Solution was to replace the keyboard interrupt
	handler address with an address of my own choosing.  Then I
	could watch for a keystroke, time it, and not let the system
	(the buffer) see it.

	Hope this helps!
	R*


-- 
Robert H. Geeslin, Ed.D.                    | Voice (918) |  Department
Geeslin@VMS.OCOM.OKSTATE.EDU                | 582-2681 day|      of
doing EDucational Programming for 20 years  | 224-0065 eve|  P. and B.S.

geeslin@vms.ocom.okstate.edu writes:

>In article <sbe4gINN2a9@agate.berkeley.edu>, andywang@crown.berkeley.edu (Andrew Wang) writes:
>> Hello,
>>     I am having a problem with keyboard interrupts crashing my
>> program.  It is a control program with a real-time hardware interrupt
>> driven at 39 hz. Everything works fine until a) the keyboard buffer
>> fills up and the computer tries to beep or b) sometimes a single keypress 
>> will cause it to crash.
>> 

>	I share your frustration!  I had the same problem with a
>	routine to do reaction times.  In it, the speaker was being
>	driven under program control and as soon as the respondent
>	pressed the key (the object of the whole routine), the 486
>	locked up.  Solution was to replace the keyboard interrupt
>	handler address with an address of my own choosing.  Then I
>	could watch for a keystroke, time it, and not let the system
>	(the buffer) see it.

>	Hope this helps!
>	R*


>-- 
>Robert H. Geeslin, Ed.D.                    | Voice (918) |  Department
>Geeslin@VMS.OCOM.OKSTATE.EDU                | 582-2681 day|      of
>doing EDucational Programming for 20 years  | 224-0065 eve|  P. and B.S.

I'd like to thank you and the several other people who had good
suggestions to help me solve this problem!!! It sure was
frustrating, but all the help allowed me to fix it.

The twist is: The computer did not crash when the keyboard handler
interrupted my interrupt handler, but when my interrupt handler interrupted
the keyboard handler! I stripped out almost everything from my interrupt
handler (except the interrupt reset stuff) and it didn't help. I moved the
outp(control8259,EOI) instruction to the and it didn't help either.

The two scenarios which caused the crash (|as explained above) are
consistent with the diagnosios above. When the keyboard buffer filled
and the keyboard interrupt beeped, my interrupt would always 
interrupt the beep and crash. Also when the keyboard interrupt was
running sometimes my interrupt would occur and cause a crash.

The solution I used (to minimize jitter in my interrupt) was:

my_timer_driven_interrupt()
{
    reset_timer();
    enable_timer_interrupt();    /* On Data Acquisition card */

    ... control routine ....


    _disable();			/* 80x86 CLI */
    outp(mask8259,0xFD & inp(mask8259));   /* enable keyboard intterrupt*/
    _enable();                  /* Now read key from keyboard */
    _disable();
    outp(mask8259,0x02 | inp(mask8259));  /* disable keybd int */
 

    outp(control8259, EOI);
}


The control interrupt cannot interrupt the keyboard interrupt.

Just now, I realize if the keyboard buffer fills up and beeps, 
my interrupt may try to re-enter itself! (It runs at 39 Hz). 
Fortunately I have a loop counter that will spit out an error message
if that occurs.

Again, thanks for everyone's input! It was most helpful!

Andrew