Re: powerpc d-i daily builds reactivated, use 2.6.12 kernels, including 64bit kernels, miboot floppies dropped for now.

["vinai" <vroopcha@mcw.edu>]

On Aug 19, 2005, Sven Luther wrote

> On Sat, Aug 20, 2005 at 03:27:36AM -0400, Rick Thomas wrote:
>>
>> .....
>>
>> Hmmm...
>>
>> Would it be possible to build a boot floppy using a fixed-config
>> stripped-down version of (say) the 2.4 kernel, with just enough
>
> No, there will be no more 2.4 powerpc kernel in etch, people needing
> it absolutely will be able to use the sarge installer and upgrade for
> now.
>
> The plan is that in the year or so upto the etch release, we will
> trim down the 2.6 kernel to miboot-size, free miboot or find a
> better solution, and fix the miboot+2.6 worked once in oldenbourg
> 2004 and never since problem.
>
> Removing the legacy stuff will only encourage us to work on the real
> fix :)
>
>> features configured to get the "real" (e.g. 2.6.xxx) kernel off  one
>> of: CD, network, hard-disk, whatever media.  It asks the user where
>> the "real" kernel is, then retrieves it, loads it, and passes
>> control to it.  As proof of concept, it could use the non-free
>> miboot until a free version becomes available.
>
> Well, the sarge 2.4 miboot floppies already installs a 2.6 kernel
> (or should), and if what you are hinting at is on-the-fly takeover
> of kernels, this is not possible (yet :).
>
>> NewWorld Macs can boot off of CDs, so these boot floppies are only
>> necessary for OldWorld PowerMacs, which, being out of production,
>> have a fixed unchanging set of devices.  So the "initial boot"
>> kernel can be built with a fixed minimal unchanging configuration.
>>  Once a working config is constructed, there will never be any need
>> to update it.
>
> Well, we already build most stuff as modules, so we would just need
> the oldworld floppy driver builtin to load the initrd, and that is
> it.  A good candidate for modularisation is the apple ide driver,
> which cannot be modularized right now, and a couple of other stuff,
> feel free to experiment with it.
>
> Now, there is also talk of a quikc-that-supports-floppies, so it
> would solve the miboot-is-non-free problem, and maybe even be able
> to load the initrd itself, so we would be able to modularize the
> floppy driver even. Or maybe even a quik-that-supports-cdrom would
> be even cooler :)
>
>> What am I missing?
>
> I guess we just need someone to do the work, as usual, i have not
> the time right now, nor in the forseable future, neither has h0lger,
> so we need fresh blood with interest to work on the etch oldworld
> boot-floppies.

Hi Folks,

One thing to keep in mind is that the swim3 driver in the Old World
Pmacs hasn't been properly updated for the 2.6 series of kernels. Last
I checked, I think the driver now compiles, but I can't remember if it
even works.  I'll take a look soon.

A while back, I'd started looking at this driver with the purpose of
getting it working, and learning about kernel driver programming.  I
had to drop it because of time constraints, etc ...  But I think at
the end of August, I should be able to spend some time on this again
(i.e. getting the floppy driver working properly for 2.6)...  If I can
pull it off, I can lend a hand with boot floppies.  I'm very attached
to my 8500, and would like to keep it (fully) in service for as long
as possible :)

I had the driver somewhat cleaned up and working by replacing the old
interrupt calling routines with the updated versions. but my changes
were not SMP safe, and my attempts to put in SMP-safe interrupt calling
primitives locked up the machine when I tested the driver.  I need to
go back and figure out how the driver works, and try to do the locking
cleanly and properly.

cheers
vinai