ARM 946E-S Abort failure (derived from csb337 bsp)

[gds]

I have been trying to just run the simple_main and/or hello_world 
examples. My console port (polled in/out) is coded, connected and seems 
to work so I can see some output. Also, am able to debug with 
arm-rtems4.8-insight debugger using a jtag debugger (Macraigor mpDemon 
with built-in gdb server).

Occasionally I am able to get all the way to main but more often I see 
an abort failure somewhere along the line. Seem rather flaky and 
inconsistent and not repeatable.

Here is what I often see printed usually before main or after exiting 
main():

INSN_LDM1
data_abort at address 0x847C, instruction: 0xE8916FF4,   spsr = 0xD1
active thread thread 0x0A010001
Previous sp=0x000179EC lr=0x00008488 and actual cpsr=600000D7
  0x00015BFC 0x00020F58 0x00000013 0x00015D14 0x600000D1 0x00000000
  0x00000000 0x00000000 0x00000000 0x00000000 0x00000000 0x00000000
  0x00000000 0x00000000 0x00000000 0x00000000 0x00000000 0x00000000
  0x00000000 0x00000000 0x00000000 0x00000000 0x00000000 0x00000000
  0x00000000 0x00000000 0x00000000 0x00000000 0x00000000 0x00000000
  0x00000000 0x00000000 0x00000000 0x00000000 0x00000000 0x00000000
  0x00000000 0x00000000 0x00000000 0x00000000 0x00000000 0x00000000
  0x00000000 0x00000000 0x00000000 0x00000000 0x00000000 0x00000000

The code at address 0x847c is in _Context_switch and looks like this 
(from arm-rtems4.8-objdump):

00008474 <_CPU_Context_switch>:
     8474:       e10f2000        mrs     r2, CPSR
     8478:       e8806ff4        stm     r0, {r2, r4, r5, r6, r7, r8, 
r9, sl, fp, sp, lr}

0000847c <_restore>:
     847c:       e8916ff4        ldm     r1, {r2, r4, r5, r6, r7, r8, 
r9, sl, fp, sp, lr}
     8480:       e129f002        msr     CPSR_fc, r2
     8484:       e1a0f00e        mov     pc, lr

00008488 <_CPU_Context_restore>:
     8488:       e1a01000        mov     r1, r0
     848c:       eafffffa        b       847c <_restore>

However, if I step through this code it does not fail. Also, if I set a 
breakpoint at 0x847c it never hits it before printing the exception 
message indicting "data abort".  I think "data abort" means I am 
accessing data outside of my ram space or possibly accessing an 
instruction that is not on a 32-bit word boundary.

I also see other failures.  Sometimes I see invalid SWI (software 
interrupt) detected. However, I think this code doubles for detecting 
undefined instructions, which may be what triggers it. (I don't see any 
software interrupts used in rtems, at least with ARM.)

Any suggestions on what I should be looking for here to debug this? I 
think I should be able to run to main and exit main with no errors printed.

Thanks,
-gene