Re: Linux drivers in gnumach
On Tue, 22 Aug 2000, Marcus Brinkmann wrote:
> I hope you'll be asked a third time, because then it becomes the truth :)
> Ugh, even oskit-mach contains i386 specific code, but less then gnumach
> without oskit. So, it might actually be easier to port oskit-mach, and you
> might be able to make use of the ancient efforts.
Perhaps. I'm looking at both, but working more on gnumach really. If
everyone keeps nudging us to work on oskit so much, why does anyone bother
working on gnumach? Just curious...
Also, today I was adding the Elf64 stuff to the various elf.h headers in
gnumach. I was wondering why the current structure was used (defining the
types of Elf??_* directly in the machine-specific elf.h file) rather than
defining them like __u16 Elf??_* in the main elf.h and just typedef'ing
__u16 (eg) in a machine-specific types header? Just curious again since
it would save a ton of duplication of effort if another port needed
something like Elf128.
> But on the other hand, oskit containes current drivers from bsd and linux,
> both are ported to many architectures.
> And OSKit is ported to StrongArm SA-110, so the seperation of architecture
> dependant and independant stuff should be already done (although I did not
> verify this). I am not sure how good the seperation is in gnumach, though.
> Maybe it is good enough already.
Gnumach's seperation is not bad. There are some things (see above) that
seem odd, but in general, it seems fairly easy to figure out what's
arch-specific. I'll have to take another look at oskit to see how good
they've got things sorted out.
> A google search on oskit alpha showed:
> * Alan Au will be working on a partical port of OSKIT to L4/Alpha.
> Is this the contact you mentioned? In the announcement of OS Kit:
> "We are also planning to do an Alpha port of the OSKit. More on that later."
I believe we mailed alanau, but got no responce as of yet. I'm a bit
disappointed that we couldn't find more info on that port since
modernising that seems like the way to go.
> Well, in other words: More people might be interested in porting OSKit to
> Alpha than in gnumach, so you could share the workload.
True, but so far, the only ones that seem interested in seeing OSKit
getting ported to Alpha are people without Alphas (I'm presuming they want
it for what I call "the coolness factor"). I haven't heard of any more
serious offers to assist with OSKit than I have wrt gnumach, so it still
looks like the workload would be the same.
> But talk is cheap, and I don't have an alpha, so don't let me distract you.
> I can try to help with the glue code if it really needs an update, but no
Ok, thanks. We just got the old code checked into our seperate CVS tree
(we wanted to do this until we get everying integrated better into a
structure that was uniform with the rest of gnumach). I'm now fixing up
header files and will move on to the hard-core stuff soon (fingers