Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 30 May 2006 23:36:35 -0400
From:      Olivier Gautherot <ogautherot@vtr.net>
To:        freebsd-small@freebsd.org, small@freebsd.org, current@freebsd.org
Subject:   Re: FreeBSD's embedded agenda - summary
Message-ID:  <200605302336.37320.ogautherot@vtr.net>
In-Reply-To: <3500.1148571492@critter.freebsd.dk>
References:  <3500.1148571492@critter.freebsd.dk>

next in thread | previous in thread | raw e-mail | index | archive | help
Greetings to all!

I think this thread moved a lot of ideas and points of view. I will skip 
the "ADN" of VxWorks and others as this does not apply to what I am trying to 
do ;-) I mean the Flash support.

In essence, we seem to have a versatile community with highly heterogeneous 
needs.

1) The "I-need-it-quick" audience who does not have a very restricted budget.
CompactFlash did not seem to be a show-stopper so I suppose a UFS file system 
on top of it (or FAT if UFS is not supported) should be OK. Wear-leveling is 
done inside using automatic sector renaming so it is a not an issue. I will 
call these the "industrial PCs" as the architecture they use is close to a 
real PC architecture (the simplest CompactFlash interface is IDE). To me, you 
are sorted out because you won't be holding your breath until I'm done ;-)

2) The "I-need-it-quick" audience with a restricted budget... who probably 
already went with another OS. They're sorted out and can wait now until the 
next project. I won't hold my breath until they come back ;-)

3) The "hard embedded" audience with RO filesystem could use UFS on a 
customized hardware layer (I call these "Target 1"). It should not be too 
difficult to implement but will require a BSP for the interface.

4) The "hard embedded" audience with RW filesystem definitely needs a more 
carefully planned FS (tentative name Plain-Old-Filesystem o POFS ;-) ), which 
I call Target 2. It is difficult to guess what Flash devices are used in the 
embedded market (my initial search was fairly miserable) and the interface 
seems to be often GPIO (what makes the project even more complex to plan). My 
view on this one is to provide a FS engine up to the hardware access and 
provide a form of BSP for various Flash types, densities and interfaces.

There was a suggestion to use Geom and I think it is a good thing to support 
so it is in the plan.

Once I have time to synthesize all this (I'm busy preparing classes for a 
university course on top of my day job...), I'll pass it on to George for the 
embeddedfrebsd.org website.

Thanks for this hot and very instructive thread!
Cheers 
-- 
Olivier Gautherot
olivier@gautherot.net



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200605302336.37320.ogautherot>