Re: [avr-gcc-list] New prosessor goes into stack violation!

[Per Arnold Blåsmo]

Ben L. Titzer wrote:

> Perhaps the size of the initialized data section is larger than the 
> available RAM left over after allocating a stack? What is the value of 
> the SP register at this point?
>
> -B
>
> On Jan 7, 2005, at 9:22 AM, E. Weddington wrote:
>
> > Per Arnold Blåsmo wrote:
> >
> > > Hi you gurus :-)
> > >
> > > I need some help her to get the code for AT90SC3232CS running.
> > > As some of you might have seen, I have tried to make avr-libc work 
> > > for the secureAVR chip AT90SC3232CS.
> > >
> > > My compiler chain works, and I am abel to make a small "hello world" 
> > > app. work.
> > > But when running a larger app, the program crashes in the startup code.
> > >
> > > It starts the __do_copy_data  routine and enter the  
> > > .do_copy_data_loop.  But after  a number of iterations it crashes and
> > > says "stack violation".  I am not sure that is  the correct error, 
> > > because on the emulator it light up "illegal address".
> > >
> > > It happens when the operation 'ST X+, R0' executes and X=0x00fe.
> >
> > Could this address be in the interrupt vector table for the device?
> >
> > Perhaps your header file did not properly account for all the vectors 
> > or all the correct defintions?
> >
> > Eric
> >
> > _______________________________________________
> > avr-gcc-list mailing list
> > address@hidden
> > http://www.avr1.org/mailman/listinfo/avr-gcc-list
> ========================================================
> I was born not knowing and have had only a little time to change that 
> here and there.
> --Richard Feynman
No, this address is not, as I can understand, in the interrupt vector 
table. There are relatively few interrupt vector in this processor 
compared to the ordinary AVR processors. Since it originally is used in 
smart card applications it also has very limited I/O.

Since this is in the startup code and no calls or anything has been 
done, the SP has not changed from its initial value. The only thing it 
has done some far is to set the SP and start copying from the end of the 
program area in flash to SRAM.

As you can see from the disassembly this is ordinary avr-gcc startup code:

Disassembly of section .text:

00000000 <__vectors>:
  0:    0c 94 0e 00     jmp    0x1c
  4:    0c 94 27 00     jmp    0x4e
  8:    0c 94 27 00     jmp    0x4e
  c:    0c 94 27 00     jmp    0x4e
 10:    0c 94 27 00     jmp    0x4e
 14:    0c 94 27 00     jmp    0x4e
 18:    0c 94 27 00     jmp    0x4e

0000001c <__ctors_end>:
 1c:    11 24           eor    r1, r1
 1e:    1f be           out    0x3f, r1    ; 63
 20:    cf ef           ldi    r28, 0xFF    ; 255
 22:    cd bf           out    0x3d, r28    ; 61

00000024 <__do_copy_data>:
 24:    12 e0           ldi    r17, 0x02    ; 2
 26:    a0 e6           ldi    r26, 0x60    ; 96
 28:    b0 e0           ldi    r27, 0x00    ; 0
 2a:    ec e2           ldi    r30, 0x2C    ; 44
 2c:    fe e0           ldi    r31, 0x0E    ; 14
 2e:    02 c0           rjmp    .+4          ; 0x34

00000030 <.do_copy_data_loop>:
 30:    05 90           lpm    r0, Z+
 32:    0d 92           st    X+, r0     <------This is where it 
crashes  X=0x00fe

00000034 <.do_copy_data_start>:
 34:    a6 38           cpi    r26, 0x86    ; 134
 36:    b1 07           cpc    r27, r17
 38:    d9 f7           brne    .-10         ; 0x30

0000003a <__do_clear_bss>:
 3a:    13 e0           ldi    r17, 0x03    ; 3
 3c:    a6 e8           ldi    r26, 0x86    ; 134
 3e:    b2 e0           ldi    r27, 0x02    ; 2
 40:    01 c0           rjmp    .+2          ; 0x44

Per A.
--
Per Arnold Blåsmo
e-mail: address@hidden
Get Thunderbird <http://www.mozilla.org/products/thunderbird/>