Re: OLinuXino (was: Re: Raspberry PI)

[Jochen Kunz <jkunz%unixag-kl.fh-kl.de@localhost>]

On Fri, 28 Sep 2012 11:26:37 +0300
Petri Laakso <petri.laakso%asd.fi@localhost> wrote:

> I'd like see some day that bootimx23 is able to load kernel from the NetBSD 
> partition
> on the SD card. Would make kernel upgrades very easy as user doesn't need
> to learn about elftosb. It could be possible after we learn how actually to 
> use MMC/SD
> interface in the kernel (to mount root for example)
> Armed with that experience, it would be easier to work on same stuff in 
> bootloader.
Sure. A bootloader, that reads an ELF kernel image from a FFS
partition, is the way to go - in the future. It also needs a way to
pass boot parameters like -s to the kernel. But I would like to keep
the possibility to combine bootloader and kernel into a single
sb-image. This is required for booting via USB. At the current state of
the affair booting a sb-image with an embedded kernel from a dedicated
sd-card partition is perfectely OK.

> Once in a while bootloader gets stuck. Possibly because some brownouts on SoC 
> power
> lines or whatever reasons. Bootloader most likely needs an another look, but 
> for now
> it does its job.
Yes. I am sure things will shake out when more people get involved. 

> After nasty stuff was done and I'd look into -current I saw this 
> arm32/arm32_kvminit.c
> which seems to provide an interface to do that all nasty stuff which all 
> these evbarm/*/*_machdep.c
> usually do. I like the idea.
Well. Just use it. ;-)

> > I sugest this as the next steps:
> > - You keep hacking on the kernel stuff. :-)
> Yup, next step is to integrate *spl() stuff with interrupt controller to
> make clocks tick.
Yes, as interrupts and timing is a precondition for all the other
device drivers. Also: Once interrupts work multiple people will be
able to hack on several device drivers in parallel. I.e. you can hack
on the SD/MMC driver while I hack on the USB host driver.

I agree: A driver for the SD/MMC card controler would be the most
important, as this carries the root file system. Once that works,
support for the SD/MMC card controler in the bootloader.

> > - I'll commit it to the NetBSD CVS repository, if the portmaster has no
> >   objections. It is very early stuff and unfinished. But I like to get
> >   it into the official CVS repository early to ensure that nothing gets
> >   lost and to give others an easy way to contribute.
> That's the start. "Scaffoldings" are in place so it is relatively
> easy to dive in and make improvements.
Yes. And my experience from similar projects taught me:
"Commit early, or it will fail and get lost."
"Commit early, or someone else will replicate your work due to not
knowing about yor work."

> iMX233-OLinuXino-MICRO looks very exciting part
> because its price and loaded with NetBSD (Lua + GPIO) it could control
> almost anything ;-)
That is the reason why I am interrsted in the Micro. It would nicly fit
into applications where a 8 bit AVR is to small. E.g. when a full blown
TCP-IP is needed or "lots of" RAM. Why should I use a low end ARM
microcontroler like an AT91SAM7X, if I can get a full blown Unix machine
for effectively the same money in an only slightly larger footprint?
(OK, that is oversimplified.)
-- 


\end{Jochen}

\ref{http://www.unixag-kl.fh-kl.de/~jkunz/}