This is the mail archive of the
ecos-discuss@sources.redhat.com
mailing list for the eCos project.
RE: loading images with redboot
- To: "Ecos-Discuss (E-mail)" <ecos-discuss at sources dot redhat dot com>
- Subject: RE: [ECOS] loading images with redboot
- From: "Dan Conti" <danc at iobjects dot com>
- Date: Tue, 9 Oct 2001 22:47:26 -0700
> -----Original Message-----
> From: Gary Thomas [mailto:gthomas@redhat.com]
> Sent: Tuesday, October 09, 2001 10:16 PM
> Subject: RE: [ECOS] loading images with redboot
>
> > I'm still at a loss as to why i can't do 'go 0x20040' from inside
> > redboot. My eventual goal is to not have a debugger attached to this
> > image. The image itself is apparently being transferred
> properly, since
> > i can attach gdb and run it.
>
> I'm a little perplexed as well - obviously we do this all the
> time too!
>
> Can you get any hint what the program is doing when you run
> it that way?
> There really should be no difference (modulo virtual vector
> I/O, but it
> doesn't seem that this is your problem) when running it one way vs the
> other.
I wrote some assembly in warm_reset to initialize the serial port and
write out some data to it. This seems to have executed ok, so it's
loading and jumping in with no problems. i can shuffle the code around
and see where things start to fail. I'm wondering about the virtual
vector part of it, i just noticed my ram kernel doesn't claim the comms
virtual vector, but when redboot switches to gdb mode it adjusts this
vector. Could this be the cause? I have all the expertise of a rock when
it comes to virtual vectors.
In this situation, where i use redboot to load then execute the image,
does the image become standalone as soon as it starts executing?