JTAG debugging

I agree with Ron.  JTAG debuggers are more robust.  But I am interested
in stepping through LinuxBIOS on a working target to quickly understand
the code, so I'm able to attempt a port to a new target.  I view it as a
learning tool and a productivity tool.  If you could show people how to
use  a product like the American Arium, you might get more help.

Steve

-----Original Message-----
From: Ronald G. Minnich [mailto:[e-mail]] 
Sent: Friday, October 22, 2004 3:06 PM
To: Eric W. Biederman
Cc: Kimball, Stephen; [e-mail]
Subject: Re: Makefile changes for symbols

On Fri, 22 Oct 2004, Eric W. Biederman wrote:
> Alternatively it might be work adding gdb stubs into LinuxBIOS.  And

doing
> using a remote debugger that way.  What a jtag interface provides
> is really not much more than what gdb stubs provides.  At least
> not after memory is initialized.

that's not quite correct. There are a host of cpu failures you can
detect 
with a jtag interface that gdb stubs are useless for. GDB stubs requires

that the processor be basically sane, jtag interfaces are not near as 
picky.

ron

[Ronald G. Minnich]

On Fri, 22 Oct 2004 [e-mail] wrote:
> I agree with Ron.  JTAG debuggers are more robust.  But I am interested
> in stepping through LinuxBIOS on a working target to quickly understand
> the code, so I'm able to attempt a port to a new target.  I view it as a
> learning tool and a productivity tool.  If you could show people how to
> use  a product like the American Arium, you might get more help.

well, we'll try to get to that end of next week once we're out of 
infiniband hell.

I'm building arima hdama images with the new source base and config tool.

What a nice set of changes :-)

ron

[Richard Smith]

> > is really not much more than what gdb stubs provides.  At least
> > not after memory is initialized.

A gdb stub would be pretty cool.  I could have used it lots already.

What happend to this?

http://www.mail-archive.com/[e-mail]/msg02078.html

Richard Smith  writes:
> > > is really not much more than what gdb stubs provides.  At least
> > > not after memory is initialized.
>
> A gdb stub would be pretty cool.  I could have used it lots already.
>
> What happend to this?

Good question.

I was not terribly enthusiastic as most of my time is spent setting
up the memory controller where gdb stubs simply cannot go.  At least
not until we can figure out cache as ram.  Which unfortunately tends
to be about as bad a setting up a memory controller.

So I figured romcc is generally and still do that romcc is a lot more
useful.

In any case gdb stubs are fairly simple and I have some.  So I might
just integrate them into LinuxBIOS.

The painful part is adding exception handlers, and it would be nice
to have exception handlers in any event so that mysteriously triggered
exceptions can be caught.

Eric