Re: [ECOS] RedBoot and interrupts

[Grant Edwards <grante at visi dot com>]

On Wed, Feb 12, 2003 at 04:41:20PM -0500, Mark Salter wrote:
> >>>>> Grant Edwards writes:
> 
> > What sort of problems are to be expected if an Ethernet driver
> > uses interrupts under RedBoot?
> 
> There will be problems with the gdb stub. The stub relies on
> interrupts being turned off by the master enable in the CPSR.
> If you needed interrupts for the stub comm channel, then that
> enable in the CPSR would need to allow interrupts. This would
> also allow interrupts used by the program being debugged. This
> could complicate the debugging because gdb assumes that the
> program being debug is completely stopped when say a breakpoint
> is hit. It would also prevent you from being able to debug
> interrupt handlers.

Thanks, I hadn't thought about that.

Debug issues with the stub I could live with.  It's a Linux
system and 99.9% of the debugging is going to be done under
Linux using the normal ARM gdb.  The primay need for RedBoot
using the built-in Ethernet would be to load a new kernel
and/or initrd and burn them into flash.

> I think it best to use the eepro board

For the eval board, it's fine, but our production HW (custom
boards) won't have the eepro board.

> or to use polling with the builtin ports.

That would be my first choice, but it depends on how much
development effort that's going to take. 

It will be possible to telnet (or ssh) into the board when it's
running Linux and update the kernel/initrd images in flash, but
that doesn't help you much once you've got a wonky kernel into
flash.  At that point, being able to use RedBoot to update
flash would be nice.  As a fallback, the serial console can
still be used to update flash, so it's not a total brick if you
get flash corrupted.

-- 
Grant Edwards
grante@visi.com

-- 
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss