Re: [PATCH v3 2/2] PM / Runtime: Add pm_runtime_enable_recursive

From: Tomeu Vizoso
Date: Thu Jul 02 2015 - 10:00:28 EST


On 20 May 2015 at 16:24, Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote:
> On Wed, 20 May 2015, Tomeu Vizoso wrote:
>
>> On 19 May 2015 at 19:49, Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote:
>> > On Tue, 19 May 2015, Tomeu Vizoso wrote:
>> >
>> >> This function makes less cumbersome to enable runtime PM in a device and
>> >> its descendants.
>> >>
>> >> Signed-off-by: Tomeu Vizoso <tomeu.vizoso@xxxxxxxxxxxxx>
>> >
>> > I don't see the point of this. In the scenario you have in mind, are
>> > the device and all its descendants registered by the same
>> > subsystem/driver?
>>
>> Not quite, the scenario here is a driver (uvcvideo) that deals with a
>> specific piece of hardware and knows that all the descendants of the
>> device it's bound to are virtual.
>>
>> The subtree is:
>>
>> 1-1:1.0 (bound to uvcvideo)
>> ep_87
>> input4
>> event4
>> media0
>> video0
>
> Just because these sub-devices are virtual, it doesn't mean you can
> ignore the way they interact with runtime PM.

Fair enough, but then, how are we expected to be able to use the
direct_complete facility if the core bails out if a descendant doesn't
have runtime PM enabled?

> In the case of ep_87 this doesn't matter. Endpoint devices (like all
> devices) are in the SUSPENDED state by default when they are created,
> and they never leave that state.

I don't see why it doesn't matter for endpoints or the others. They
don't have runtime PM enabled, so no ancestor will be able to do
direct_complete.

> Does the uvcvideo interface go into runtime suspend at the right times?
> If it does then you shouldn't have anything more to worry about.

It goes in and out of runtime suspend just fine, but when the system
goes to sleep, it's runtime resumed and then suspended. And a full
resume takes typically very long for USB cameras.

>> I liked how the force_direct_complete flag played out here, but I
>> agree with Rafael that it can be abused as the PM domain or the bus
>> type weren't able to prevent going directly to complete.
>>
>> This is my testing branch, btw:
>>
>> https://git.collabora.com/cgit/user/tomeu/linux.git/log/?h=fast-resume-v5
>>
>> > If they are, can't the subsystem/driver code enable
>> > runtime PM for each of them when they are registered, by adding a
>> > single call in the right spot?
>> >
>> > If they don't all belong to the same subsystem/driver, who is going to
>> > call your pm_runtime_enable_recursive routine? No single caller will
>> > have the right to enable runtime PM for all these devices.
>>
>> Yeah, I was thinking that uvcvideo might be able to decide that, but
>> I'm not really sure about it.
>
> A possible way around the problem is to use pm_suspend_ignore_children
> on the uvcvideo interface. But I'm not sure that would be the right
> thing to do.

Would that mean that if a device has ignore_children then it could
still do direct_complete even if its descendants weren't able to?

Thanks,

Tomeu

> Alan Stern
>
> --
> 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/
--
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/