Re: [PATCH v2 3/3] thermal: Make THERMAL_HWMON implementation fully internal

[Jean Delvare]

Hi Guenter,

On Tue, 26 Apr 2011 14:00:31 -0700, Guenter Roeck wrote:
> On Tue, 2011-04-26 at 15:39 -0400, Jean Delvare wrote:
> > Sorry I realize I have been inaccurate. thermal_sys indeed depends on
> > the processor module, and that's what prevents me from unloading it.
> > It's the processor module which has a reference count of 1, and no
> > dependency, so I have no idea how I could unload it.
> > 
> You mean the ACPI processor driver ? This comment might explain it:
> 
> /*
>  * We keep the driver loaded even when ACPI is not running.
>  * This is needed for the powernow-k8 driver, that works even without
>  * ACPI, but needs symbols from this driver
>  */

I doubt it. I don't have powernow-k8 loaded on the affected systems.
And if I did, that would be a regular inter-module dependency that
would be listed by lsmod.

> > (...)
> > I'm not sure. I don't see any relevant call to try_module_get under
> > drivers/acpi, and I'm not aware of any other way to increase the
> > reference count.
> > 
> What happens if the calling code (such as, in my case here, the acpi
> video code) gets built into the kernel ? Would that force the module to
> be and remain loaded ?

For regular inter-module dependencies, that's not even possible. You
can't build into the kernel a driver which depends on a symbol exported
by a driver which is built as a module. It will break at link time.

For reference counts increased by try_module_get(), it is possible, but
that doesn't necessarily make a difference from the all-modules case:
the reference count can be increased at any time, not only during
driver initialization, so the corresponding module_put doesn't
necessarily happen during module unloading (which indeed never happens
for a built-in driver.)

I tried booting in initlevel 1, but it didn't help, the reference count
of module processor is stuck to 1. So it's not user-space causing it...
it's something in the kernel itself. But I really don't have the time
to investigate this further, especially when I have no idea where to
look next.

-- 
Jean Delvare
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/