LinuxPPC with BDM

LinuxPPC with BDM

[Brendan Simon]

Dan mentioned (in another email) that mpcbug can cause problems when
running linux.  I take it this is a generic BDM problem.  Is this true
?  In theory it shouldn't matter should it, as long as the debugger
leaves the core in the same condition just before it enters debug mode.

I have been using the BDM to download kernels and running them from RAM
(stepping through the code, etc).  Could this be the cause of the
strange behaviour I've seen on the ADS boards ?

Should I stear clear of the BDM ?  I would prefer to use it as it allows
alot of low level debugging and control.

Brendan Simon.

BTW. I have been using a beta version of GDB with support for the
Macraigor Wiggler.  I got it from Scott Howard.  Unfortunately it will
only run on a Windows machine due to the wigglers.dll which is
required.  If anyones interested in this then there is more information
on http://www.objsw.com/


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: LinuxPPC with BDM

[Richard Hendricks]

The BDM runs into problems when the underlying operating system uses the
MMUs to implement a page table.  Since the BDM doesn't understand page
tables, when it stops the core it assumes the PC is pointing into
physical memory.  This makes debugging the kernel difficult using BDM. 
Using the BDM inherently doesn't change any registers, but MPC8bug may
initialize some registers you don't want it to.

Brendan Simon wrote:
> 
> Dan mentioned (in another email) that mpcbug can cause problems when
> running linux.  I take it this is a generic BDM problem.  Is this true
> ?  In theory it shouldn't matter should it, as long as the debugger
> leaves the core in the same condition just before it enters debug mode.
> 
> I have been using the BDM to download kernels and running them from RAM
> (stepping through the code, etc).  Could this be the cause of the
> strange behaviour I've seen on the ADS boards ?
> 
> Should I stear clear of the BDM ?  I would prefer to use it as it allows
> alot of low level debugging and control.
> 
> Brendan Simon.
> 
> BTW. I have been using a beta version of GDB with support for the
> Macraigor Wiggler.  I got it from Scott Howard.  Unfortunately it will
> only run on a Windows machine due to the wigglers.dll which is
> required.  If anyones interested in this then there is more information
> on http://www.objsw.com/

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: LinuxPPC with BDM

[Dan Malek]

Richard Hendricks wrote:
> 
> The BDM runs into problems when the underlying operating system uses the
> MMUs to implement a page table.


That's the biggest one.  Also, all of the BDMs I have used make
assumptions about the target environment.  Things like some
monitor support and debug register settings also get in the way.

Motorola's mpc8bug is the only BDM I have found that would let
me download and run....no debug mind you, just get out of the way
and run the code.  You have to make sure all of the debug is
disbled before you let it go.


> > I have been using the BDM to download kernels and running them from RAM
> > (stepping through the code, etc).  Could this be the cause of the
> > strange behaviour I've seen on the ADS boards ?


Yep.  You can only disable all debugging and let it go.


> ...  Unfortunately it will
> > only run on a Windows machine due to the wigglers.dll which is
> > required.



I am working with a company right now developing some embedded
tools hosted on a Linux/PPC (possibly Linux/x86) that includes
a BDM interface.  More interesting though is the Cygnus InSight and
GDB server I have running for application development.  People
forget that it is the application that makes the difference.
Porting and compiling a kernel these days is not a challenge like
getting a real embedded product to market.



	-- Dan

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/