[commit] Respect "red zone" for SPU inferior calls

[commit] Respect "red zone" for SPU inferior calls

[Ulrich Weigand]

Hello,

GDB neglected to respect the ABI-defined "red zone" of 2000 bytes on the SPU,
which could cause stack corruption when calling an inferior function while
within prologue code of another function.  Fixed by the patch below.  

Tested on spu-elf with no regressions, applied to mainline.

Bye,
Ulrich


ChangeLog:

	* spu-tdep.c (spu_gdbarch_init): Call set_gdbarch_frame_red_zone_size.

Index: gdb/spu-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/spu-tdep.c,v
retrieving revision 1.31
diff -c -p -r1.31 spu-tdep.c
*** gdb/spu-tdep.c	12 Jul 2008 22:42:52 -0000	1.31
--- gdb/spu-tdep.c	6 Aug 2008 18:25:35 -0000
*************** spu_gdbarch_init (struct gdbarch_info in
*** 2043,2048 ****
--- 2043,2049 ----
    /* Inferior function calls.  */
    set_gdbarch_call_dummy_location (gdbarch, ON_STACK);
    set_gdbarch_frame_align (gdbarch, spu_frame_align);
+   set_gdbarch_frame_red_zone_size (gdbarch, 2000);
    set_gdbarch_push_dummy_call (gdbarch, spu_push_dummy_call);
    set_gdbarch_dummy_id (gdbarch, spu_dummy_id);
    set_gdbarch_return_value (gdbarch, spu_return_value);
-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com

Re: [commit] Respect "red zone" for SPU inferior calls

[Mark Kettenis]

> Date: Wed, 6 Aug 2008 20:29:28 +0200 (CEST)
> From: "Ulrich Weigand" <uweigand@de.ibm.com>
> 
> Hello,
> 
> GDB neglected to respect the ABI-defined "red zone" of 2000 bytes on the SPU,
> which could cause stack corruption when calling an inferior function while
> within prologue code of another function.  Fixed by the patch below.

Do I read that right, is that reallu *decimal* 2000?  Seems a bit ood
to me.

Re: [commit] Respect "red zone" for SPU inferior calls

[Ulrich Weigand]

Mark Kettenis wrote:

> > From: "Ulrich Weigand" <uweigand@de.ibm.com>
> > GDB neglected to respect the ABI-defined "red zone" of 2000 bytes on the SPU,
> > which could cause stack corruption when calling an inferior function while
> > within prologue code of another function.  Fixed by the patch below.
> 
> Do I read that right, is that reallu *decimal* 2000?  Seems a bit ood
> to me.

It may be odd, but this is indeed what the ABI specifies:
http://www-01.ibm.com/chips/techlib/techlib.nsf/techdocs/02E544E65760B0BF87257060006F8F20/$file/SPU_ABI-Specification_1.8.pdf

  "Storing to memory using the stack pointer plus an offset must never be
   done with an offset less than -2000 (-125*16). This allows interrupt
   handlers to use the application stack by first adding -2000 to the
   stack pointer."   (p. 9)

This also agrees with what GCC implements.

Bye,
Ulrich

-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com