Forums

Memory corruption...

Started by goelhss March 29, 2004
Hi,
I am facing some memory corruption problem in my system.
Here some DSPs(C5420) are working in conjunction with a control
processor (CP) with which they exchange data via HPI memory.

Entire memory and data is in first page of DSPs only.
I am using CCS 2.2/JTAG for debugging it.

CP sends some control and data messages by writing into buffers in
DSP,
which are read by DSP on polling basis and writes them on McBSP.
The similar data flow is in reverse direction.

Problem is that at some random instance, DSP code crashes... with
code and data memory gone bad/corrupt.

After much debugging, I am doubting CP for this.

Please help me with these.

1. Please tell me what approaches could I use to see that it's not
DSP code
that is problemetic.( like malloc/free, push and no pop.. leading
to stack overflow), stack overflow,
bad C pointers etc).

2. Just in case it's CP which is doing some problem, like writing
into code area,
how can I detect that ?

I'd really appreciate any help that you can offer.
Thanks in Advance

Regards
Vishal.



Vishal-

Is this a Cisco or Nortel SIMM module? How many C5420 DSPs? What is the C5420
Vcc
core value?

What if you have host processor write to unused (non-defined) memory area
repeatedly
-- in that case, DSPs receive no messages, no McBSP data is written, and DSPs
should
run forever without a crash. Does that occur?

-Jeff goelhss wrote:
>
> Hi,
> I am facing some memory corruption problem in my system.
> Here some DSPs(C5420) are working in conjunction with a control
> processor (CP) with which they exchange data via HPI memory.
>
> Entire memory and data is in first page of DSPs only.
> I am using CCS 2.2/JTAG for debugging it.
>
> CP sends some control and data messages by writing into buffers in
> DSP,
> which are read by DSP on polling basis and writes them on McBSP.
> The similar data flow is in reverse direction.
>
> Problem is that at some random instance, DSP code crashes... with
> code and data memory gone bad/corrupt.
>
> After much debugging, I am doubting CP for this.
>
> Please help me with these.
>
> 1. Please tell me what approaches could I use to see that it's not
> DSP code
> that is problemetic.( like malloc/free, push and no pop.. leading
> to stack overflow), stack overflow,
> bad C pointers etc).
>
> 2. Just in case it's CP which is doing some problem, like writing
> into code area,
> how can I detect that ?
>
> I'd really appreciate any help that you can offer.
> Thanks in Advance
>
> Regards
> Vishal.