On Fri, Apr 21, 2006 at 06:53:46PM +0200, Thomas Renninger wrote:
> On Friday 21 April 2006 17:18, Bjorn Helgaas wrote:
> > On Friday 21 April 2006 04:43, Thomas Renninger wrote:
> > > If booting with acpi=off and loading the processor module, it
> > > may happen that the kernel oopses when trying to unload the module again.
> > > Better exit all ACPI modules immediately if booted with acpi=off.
> > > ...
> > > @@ -289,31 +289,28 @@ static int __init acpi_ac_init(void)
> > > {
> > > int result = 0;
> > >
> > > - ACPI_FUNCTION_TRACE("acpi_ac_init");
> > > + if (acpi_disabled)
> > > + return -ENODEV;
> >
> > This doesn't feel like the right solution to me.
> > acpi_bus_register_driver() already returns -ENODEV
> > if acpi_disabled, and that ought to be enough so the
> > drivers can handle this correctly.
> >
> > I don't think we should clutter every ACPI driver with
> > the same check for acpi_disabled.
> Yes, you are right.
> >
> > Did you investigate the oops when unloading the processor module?
> It's in unregister_driver.
> > I wonder if that driver has a bug there. It has a mysterious
> > comment that "we keep the driver loaded even when ACPI is not
> > running" for powernow-k8. I don't understand that at all, but
> > it sounds like it requires extra care to make sure everything
> > works right.
> AFAIK processor module has no functions if acpi=off.
> Some cpufreq drivers still work. The ones which have the steps
> for each CPU model compiled in and that should still work.
>
> There should be no functions missing when booting acpi=off and processor
> module cannot be loaded.
> But I might be wrong here...
>
> Hmm, as even proc is deleted, the module probably should not be loaded...
>
> I think this one should go in then:
>
> Subject: Exit ACPI processor module if acpi is disabled
No. For example powernow-k8 and speedstep-centrino do have a dependency on a
function which is exported by processor_perflib.c. THis means you cannot
load these modules (which can run without ACPI support) if loading the
processor module fails. Therefore its init function must not fail.
Dominik
-
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
