Re: [ECOS] ROMRAM redboot

[Andrew Lunn <andrew dot lunn at ascom dot ch>]

> Something else which might be just as useful would be to disable the
> 'printf()'s in the Flash code which is what I think Andrew is
> concerned with.  These could be conditional #ifdef CYGPKG_REDBOOT
 
> Note: we've thought about extending the flash functions to not do
> the prints directly, but rather through call backs.  (Maybe Andrew
> even suggested that).  It's not been done yet, but still under
> consideration.

No, im not concerned with this since we wrote our own flash API 6
months or more before RedHat wrote theirs. If its not broken, don't
fix it is my philosophy, so we are still using ours, not RedHats.

What more of a problem is things like ethernet frames received on eth0
and being passed to the RedBoot stack. If this happens during a flash
operation, we are dead. 

As for me suggesting it, i think not. I think i actually suggested
changing the flash API to use virtual vectors into RedBoot. That way
RedBoot knows when flash operations are going on, and it can replace
all its vectors to a NOP function in RAM while the operation is
happening. Unfortunatly, this only partialy solves the problem. This
does not allow the application to upgrade RedBoot, which is something
a field upgrade probably requires.

My solution to the problem is to make RedBoot ROMRAM. Its then safe in
terms of FLASH operations becasue its in RAM, and its safe to upgrade
it because its currently using a copy.

        Andrew