[kdb] Invalidate registers if values are on the stack?

[Keith Owens]

On Fri, 6 Feb 2009 12:31:33 -0600, 
Martin Hicks <mort at sgi.com> wrote:
>
>why does bb_start_block0_special in arch/x86/kdb/kdba_bt.c automatically
>invalidate the register if the NS specifies that the contents of the
>register is also on the stack?
>
>        for (i = 0; i < r->mem_size; ++i) {
>                value = r->mem[i].value;
>                if (test_bit(value, r->skip_mem.bits)) {
>                        /* this memory entry is not defined for this label */
>                        continue;
>                }
>                bb_memory_set_reg_value(BBRG_RSP, r->mem[i].offset_address,
>                                        value, 0);
>                bb_reg_set_undef(value);
>        }

Not also, instead of.  If a saved register is passed on stack then we
do not want the register that originally contained that value to be
considered as part of the data flow.