Re: [ECOS] exception when calling reset_platform

[Iztok Zupet <iz at vsr dot si>]

On Friday 14 June 2002 14:41, Roland Caßebohm wrote:
> On Friday, 14. June 2002 14:05, Iztok Zupet wrote:
> > On Friday 14 June 2002 12:45, Roland Caßebohm wrote:
> > > On Thursday, 13. June 2002 19:16, Iztok Zupet wrote:
> > > > On Thursday 13 June 2002 18:20, Roland Caßebohm wrote:
> > > > > Hi,
> > > > >
> > > > > my ARM7TDMI (SNDS100 Board) makes an abort prefetch exception when
> > > > > the application calls reset_platform().
> > > > >
> > > > > I added two nop's because I thought the processor prefetches the
> > > > > init_done value, but this doesn't help:
> > > > >
> > > > > ----------
> > > > >         .global reset_platform
> > > > >         .type   reset_platform,function
> > > > > reset_platform:
> > > > > #ifdef CYGSEM_HAL_ROM_MONITOR
> > > > >         // initialize CPSR (machine state register)
> > > > >         mov
> > > > > r0,#(CPSR_IRQ_DISABLE|CPSR_FIQ_DISABLE|CPSR_SUPERVISOR_MODE) msr
> > > > > cpsr,r0
> > > > >         b       warm_reset
> > > > > #else
> > > > >         mov     r0,#0
> > > > >         mov     pc,r0           // Jump to reset vector
> > > > >         //roland
> > > > >         mov     r0,r0           // nop
> > > > >         mov     r0,r0           // nop
> > > > > #endif
> > > > >
> > > > > init_done:
> > > > >         .long   0xDEADB00B
> > > > > ----------
> > > > >
> > > > > Does anybody know what's wrong?
> > > > >
> > > > > Roland
> > > >
> > > > Hi:
> > > >
> > > >   Looks like that the abort prefetch hapens somewhere else, not in
> > > > the code You've sent. Take a look at the reset vector and the code
> > > > that executes there ( if it is in RAM). It may be corrupt for some
> > > > reason. Anyhow, can You figure out the address of the exeption?
> > > >
> > > > iz
> > >
> > > The exception seems to be at 0x31638:
> > >
> > > void vs_reboot(void)
> > > {
> > >
> > >     ...
> > >
> > >    31638:       ebffbb55        bl      20394 <reset_platform>
> > > }
> > >    3163c:       e91ba800        ldmdb   r11, {r11, sp, pc}
> >
> > Strange. Since the abort prefetch happens here (it happens only beacuse
> > of the ABORT signal asserted) it means that <reset_platform> is no longer
> > in valid memory that can be reached by cpu. What kind of memory mapping
> > do You have? Take a look at the linker xxx.ldi script in the platform/hal
> > directory if they if it's correct for  Your memory layout.
> >
> > iz
>
> RAM is mapped to 0x0 and the size is 0x200000. The function reset_platform
> should be and is in this array (0x20394).
>
> Roland

Now, try to replace the 
	bl	<reset_platform>
with
        mov     r0,#0
        mov     pc,r0           // Jump to reset vector

to see what happens. I dont have any other idea. Do You have any kind of 
memory management on board?

iz


-- 
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss