RE: gcc compile

[Yves Rutschle <y.rutschle@indigovision.com>]

Hallo,

> I have tested the code on the eb01. After swi the proc is in SV.

Therefore it's not because of the simulator :)

> 0x2018d28 <swi_handler__Fi>:	mov	r12, sp
> 0x2018d2c <swi_handler__Fi+4>:	stmdb	sp!, {r11, r12, lr, pc}
> 0x2018d30 <swi_handler__Fi+8>:	sub	r11, r12, #4	; 0x4
> 0x2018d34 <swi_handler__Fi+12>:	sub	sp, sp, #12	; 0xc
[snipped swi routine]
> 0x2018d98 <swi_handler__Fi+112>:	ldmdb	r11, {r11, sp, pc}

This code won't set CPSR back to SPSR; Try to add a hat
at the end of the return instruction:
	ldmdb r11, {r11, sp, pc}^
(the hat activates the copy of SPSR to CPSR same as movs pc,lr)

>Don?t know what I am doing wrong. This does also not work:
>void swi_handler(int reason) {
>  switch(reason) {
>  case REASON_CHANGE_TO_SV :
>    asm volatile ("mrs r0,cpsr": : :"r0");
>    asm volatile ("orr r0, r0, #192": : :"r0");
>    asm volatile ("msr cpsr, r0");
>  default :
>    break;
>  }
>  asm volatile("movs    pc, lr");
>}

Huh. What do you think happens to the stack when you leave
like that? :-) (hint: you pushed 4 registers when entering
the routine...)

I don't think you can create a whole swi handler in C (unless
gcc has extensions I don't know about. the arm C doesn't
anyway).

I'd advise you to use an assembly top-level handler such
as:

	stmdf sp!, {r0-r3, lr}
	ldr	r0, [lr,#-4]
	bic	r0, r0, #0xff000000
	mov	r1, sp			; r1 = SWI number
	mrs	r2, spsr
	stmfd	sp!, {r2}			; keep spsr in case of nested swi's
	bl	C_Handler
	ldmfd	sp!, {r2}
	msr	spsr, r2			; restore spsr
	ldmfd	sp!, {r0-r3, pc}^		; return, restore cpsr from spsr

(this is straight from ADS' developer guide, by the way)

Then your C handler looks like:

void C_Handler( int swi_number, unsigned int * stack )
{
    printf( "got swi %d\n", swi_number );
}

Hope that helps!

Yves.






------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sourceware.cygnus.com