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

Re: G5 fan control and official debian kernels.



On Tue, Jun 27, 2006 at 11:01:52AM +0200, Jonas Smedegaard wrote:
> On Tue, 27 Jun 2006 10:03:04 +0200 maximilian attems wrote:
> 
> > On Tue, Jun 27, 2006 at 09:47:32AM +0200, Sven Luther wrote:
> > > On Tue, Jun 27, 2006 at 09:17:49AM +0200, maximilian attems wrote:
> 
> > > > did you check it they get loaded by the initramfs?
> > > > please tell which fan and i2c modules need to be added for ppc?
> > > 
> > > No, it is a bug in the kernel, and even loading them by hand
> > > doesn't help.
> > > 
> > > I do have the module in /etc/modules for the above attempt.
> > 
> > that does not put them in the initramfs,
> > use /etc/initramfs-tools/modules for that.
> 
> Sven: Did you check with lsmod that they were in fact loaded (and no
> other related modules was)?

I did indeed check, and indeed adding i2c-powermac with 2.6.17 (2.6.16 was
buggy in this regard apparently) helped solve this. I was not previously aware
that i2c-powermac was needed.

> Perhaps use yaird for debugging this, to make sure the least possible
> is loaded at all.

having the module hand-loaded or in /etc/modules works just fine for this.

> > > Oh, and BTW, initramfs-tools generated ramdisks attempt to load the
> > > atkbd/i8042 modules on hardware not having any ps2 support. Can
> > > anything be done about it.
> > 
> > ppc builds those modular, which makes debuggig an initramfs a pain.
> > we already had an lengthy discussion about that 1/2 year ago.
> > as you like to keep them modular i need to load them or
> > the user might not be able to do anithing in the rescue shell.

What about doing some checks for the subarch that needs them in the first
place, and only load them if needed ? Those needing it are quickly looked at,
and i believe include only :

  - CHRP/Pegasos
  - PReP
  - CHRP/32bit

Not sure about the Power 3 RS6k chrp boxes from IBM, nor more recent
power4/power5 boxes, will investigate.

> > how about fixing those drivers?!

I guess the right thing to do is to add hotplug support to those modules.
But to quote benh :

  10:06 < benh> svenl: the "right" hotplug/udev thingy is a pain in the ass
  that I haven't fully figured out yet

And i guess i fall into this category myself too, so help is welcome in this.

BTW, but the same applies to the i2c-powermac and other such modules, there is
really no reason to load them on anything but the powermac subarch.

> AFAIR this one was not about broken drivers, but working fine when
> loaded as modules, and seldom of any use on that architecture, so left
> out to save space in the kernel. Building drivers modular is all (or
> mostly) about saving space at runtime, isn't it?

Well, with udev taking ages to try loading modules afterward, the time issue
is dubious. But i couldn't agree more about this, and i still believe the
yaird approach is the best one, which should be emulated in the MODULES=dep
way of initramfs-tools.

Friendly,

Sven Luther



Reply to: