Re: 2.4.25-1 debian kernel available based on -ben1 tree, G5 tester wanted.
On Sat, Mar 06, 2004 at 11:51:16AM +0100, Carlos Perelló Marín wrote:
> El sáb, 06-03-2004 a las 09:54, Sven Luther escribió:
> > On Sat, Mar 06, 2004 at 12:35:33AM +0100, Carlos Perelló Marín wrote:
> > > El mié, 03-03-2004 a las 09:20, Sven Luther escribió:
> > > > Hello,
> > > >
> > >
> > > Hi
> > >
> > > > I have just uploaded the 2.5.25-1 kernel, and as it is NEW, made it
> > > > available at :
> > > >
> > > >
> > > > http://people.debian.org/~luther/powerpc/2.4.25-1
> > >
> > > As a comment, I prefer a heavy modular kernel (like the x86 .deb ones)
> > > but I'm not able to compile such kernel so I know it's not as easy to do
> > > like x86 kernels.
> >
> > Please have a look at the powerpc-small kernel. It is primarily geared
> > at the oldworld pmacs and its floppy based miboot booting, but there is
> > no reason it could not be used more generally.
> >
> > In fact, it was ever the intentions to move to this kernel once it has
> > been more widely tested, so please go ahead, and please provide feedback
> > on it.
> >
> > The main problem is that i don't yet have discovered the trick to build
> > the initrd which contain the modules, since if you are going to move the
> > ide & scsi & network modules to an initrd, you also need it for booting.
>
> Yes, my problem is with initrd and thus I cannot test that kernel if the
> I cannot load the initrd image :-( I will look at this in my spare time,
> we should get it working...
Just boot it using the powerpc-small debian-installer initrd, the same
one Jeremie Koenig used on his oldworld pmac.
>
> >
> > > I'm just using your packages now instead of my own compiled kernel with
> > > an iMac TFT 15" and it seems to be working with only one problem, you
> > > have compiled the lpt port and the imac does not have such port, the
> > > problem is that sometims it hangs the machine when discover or cups
> > > tries to load the parport_pc module :-(
> >
> > Mmm, this would be a bug in the discover or cups packages, don't you
> > think ? Or maybe we could add a check in the partport modules to fail to
> > load on pmacs, which are known not to have such hardware ?
>
> Perhaps it's a bug with discover or cups because they try to load a
> driver for a hardware I don't have but I don't think such driver should
> hang the computer, that's also a kernel bug. I don't understand either
> why the parport_pc is needed with powerpc kernels if they are not pc
My pegasos powerpc motherboard as a standard via south bridge, including
serial and parallel port support. I guess some prep or chrp hardware
also has it, not sure though.
> > The -powerpc kernels support a wide range of hardware, including some
> > (as mine), that does have parport support.
>
> Yes, I know, that's why I prefer a heavy modular kernel, but Do you need
> the pc specific driver? The CONFIG_PARPORT_PC help says that it's needed
> for "All IBM PC compatible computers and some Alphas have PC-style
> parallel ports." Is it the same for your computer?
Yeah, see above. But the modular approach doesn't help here, as
apparently it is discover/cups/whatever trying to load the modules while
it should not which cause the problem.
> > Please fill a bug report, so this doesn't get lost.
>
> Ok
>
> > > I had the same problem with my tibook G4 550MHz with other kernels
> > > compiled by hand.
> >
> > Yeah, i guess the bug is in discover then, not sure though.
>
> As I said, I see that discover have a bug trying to load a module I
> don't know but also I'm asking about the need of that module in powerpc
> kernels. If you need it to get working your parport then I will fill a
> bug report against kernel 2.4 (and perhaps 2.6) because hangs a machine
> loading that module, if you don't need it I will do the same but asking
> to disable that option from the menu when compiling a powerpc kernel.
Many modules make unsane assumption about the hardware, and cause havoc.
This is unavoidable i think, since the probing of the needed bit on pc
compatible hardware cause the pmac to die horribly. Also, as i
understand how discover works, it makes a match on pci ids and such, to
detect the kernels that should run.
It should be possible to add a check in the driver though, just to be
sure. Not high priority though
Friendly,
Sven Luther
Reply to: