Date: Mon, 20 Jan 1997 08:53:28 -0700 (MST)
From: Josh Turpen <snarfy AT goodnet DOT com>
To: djgpp AT delorie DOT com
Subject: Interrupt handlers & GPF's in RMCBs
Message-ID: <Pine.GSO.3.95.970120080408.4793A-100000@goodguy>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII


Can anybody tell me if my interrupt code is ok?  About once out of 10 times
my program bombs with a GPF | Stack Fault | Page Fault in RMCB.  I don't 
know what a real-mode callback is, but I think it has to do with my
interrupt code. 

# file irq.S
.globl _irq0_int

_irq0_int:  pushl %eax                     /* save clobbered regs */
            pushw %ds                      /* save clobbered regs */
	    movw %cs:___djgpp_our_DS, %ds  /* get a valid DS so that */
            movl __lwp_enable, %eax        /* we can access __lwp_enable */
            cmpl $0, %eax                  /* raise exception? */
            jne done                       /* no */
            movl $0x99, %al                /* generate exception 0x99 */
            call ___djgpp_hw_exception     /* later trap exception */
done:       movl $0x20, %eax               /* outportb(0x20,0x20) to */
            outb %al, $0x20                /* acknoledge interrupt */
            popw %ds                       /* restore DS */
            popl %eax                      /* restore EAX */
            sti                            /* restore int flag */
            iret                           /* return */


I've tried it with and without the sti, as some people told me I don't
need it while the DPMI spec says I do.  Either way it still blows up
about 1 out of 10 times.  This interrupt handler is installed with
the __dpmi_set_protected_mode_interrupt_vector, and it works great
except for the occasional hang.

I'm not even sure if I should be looking at this code for the bug,
because I really don't know what a RMCB is.  

Josh
snarfy AT goodnet DOT com