Re: [patch 07/15] clocksource: Provide unbind interface in sysfs

[John Stultz <john.stultz@linaro.org>]

On 05/15/2013 02:47 AM, Thomas Gleixner wrote:
> On Mon, 29 Apr 2013, John Stultz wrote:
>> On 04/25/2013 01:31 PM, Thomas Gleixner wrote:
>>> With the module refcount held for the current clocksource there is no
>>> way to unload the module.
>>>
>>> Provide a sysfs interface which allows to unbind the clocksource. One
>>> could argue that the clocksource override could be (ab)used to do so,
>>> but the clocksource override cannot be used from the kernel itself,
>>> while an unbind function can be used to programmatically check whether
>>> a clocksource can be shutdown or not.
>>>
>>> The unbind functionality uses the new skip current feature of
>>> clocksource_select and verifies that a fallback clocksource has been
>>> installed. If the clocksource which should be unbound is the current
>>> clocksource and no fallback can be found, unbind returns -EBUSY.
>>>
>>> This does not support the unbinding of a clocksource which is used as
>>> the watchdog clocksource. No point in fostering crappy hardware.
>> So.. if the clocksource you want to unbind is the highest rated continuous
>> clocksource that doesn't need a watchdog (basically what's likely to be in-use
>> and required to be unbinded), its likely to be selected as the watchdog
>> already.
>>
>> ie: on a system that has only HPET/ACPI_PM, you can't unbind HPET, since its a
>> watchdog.
> No. The thing is that I only prevent unbinding if it is used as the
> watchdog. In the above HPET/PM scenario both are potential watchdogs,
> but w/o a user it's valid to unbind one of them.
>
> What I need to prevent is:
>
> TSC is current clocksource and we only have ACPI_PM as watchdog and
> its used. So now you try to unbind ACPI_PM then the TSC would be left
> w/o a watchdog instance. That's what I'm preventing. Will reword the
> changelog accordingly.

You might double check the logic, because I feel like I actually hit 
this issue where on my vm machine w/ only hpet/acpi_pm I couldn't unbind 
the hpet.

Though its been a few weeks, so maybe I'm confusing things?
Let me know if you want me to try to reproduce it.

thanks
-john