[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[MiNT] .xdd for HW dependent things (was: Re: Shutdown() discussion)



Hi!

This seems like a good way to go....

On Wed, 10 Dec 2003, Petr Stehlik wrote:

> On Wed, 2003-12-10 at 17:01, Odd Skancke wrote:
> >  I think perhaps the kernel should have generic shutdown code, and rely on
> > some sort of XDD to actually perform the shutdown. This is a hardware
> > specific thing, right?
>
> Right. The question is how much of the hardware specific things belong
> to the kernel and how much should be loadable modules.
>
> Let's take the Linux, for example: there is a module called "apm" that
> does the very thing we are discussing here - turns off the ATX PSU.
> Without this module loaded the OS just halts.
>
> So I think it's a smart idea.
>
> But the question of implementation is still the same: are you OK with
> having ARAnyM specific code in APM.XDD or do you want us to develop
> APM-ARANYM.XDD? If the former, what is the difference between this and
> implementing it in the kernel itself? The CVS tree is the same, after
> all. And if the latter (APM-machine.XDD) - can the kernel load the right
> XDD at boot time automagically? Or is user required to enable/disable
> the XDDs with an XBOOT style tool? That would be a step back, IMHO.

Well, in fact we don't have automagic module loading. You need to rename
the hostfs.xfs in case you don't want the kernel to try to load it. Even
though it should be safe, because it checks the __NF cookie itself and
exits if not.

If you don't want the ext2 partitions to be mounted then you have to
rename the ext2.xfs to something else (XBOOT style :().

In this area we can only make steps further IMHO. ;)

OK, let's define an .xdd to the most of the HW dependent issues from the
kernel. This way the kernel might get smaller (I personally doubt it
however).... So... at least for newly apearing features we should go this
way.

Other proposals?

regards

STan