Re: [PATCH 0/8] Suspend block api (version 7)

From: Arve Hjønnevåg
Date: Tue May 18 2010 - 16:47:18 EST


2010/5/18 Rafael J. Wysocki <rjw@xxxxxxx>:
> On Tuesday 18 May 2010, Arve Hjønnevåg wrote:
>> 2010/5/17 Rafael J. Wysocki <rjw@xxxxxxx>:
>> > On Monday 17 May 2010, Arve Hjønnevåg wrote:
>> >> 2010/5/14 Rafael J. Wysocki <rjw@xxxxxxx>:
>> >> > On Friday 14 May 2010, Arve Hjønnevåg wrote:
>> >> >> This patch series adds a suspend-block api that provides the same
>> >> >> functionality as the android wakelock api. This version has some
>> >> >> changes from, or requested by, Rafael. The most notable changes are:
>> >> >> - DEFINE_SUSPEND_BLOCKER and suspend_blocker_register have been added
>> >> >>   for statically allocated suspend blockers.
>> >> >> - suspend_blocker_destroy is now called suspend_blocker_unregister
>> >> >> - The user space mandatory _INIT ioctl has been replaced with an
>> >> >>   optional _SET_NAME ioctl.
>> >> >>
>> >> >> I kept the ack and reviewed by tags on two of the patches even though
>> >> >> there were a few cosmetic changes.
>> >> >
>> >> > Thanks for the patches, I think they are in a pretty good shape now.
>> >> >
>> >> > That said, I'd like the changelogs to be a bit more descriptive, at least for
>> >> > patch [1/8].  I think it should explain (in a few words) what the purpose of
>> >> > the feature is and what problems it solves that generally a combination of
>> >> > runtime PM and cpuidle is not suitable for in your opinion.  IOW, why you
>> >> > think we need that feature.
>> >> >
>> >>
>> >> How about:
>> >>
>> >> PM: Add opportunistic suspend support.
>> >
>> > "PM: Opportunistic suspend support" would be sufficient IMO.
>> >
>> > Now, I'd start with the motivation.  Like "Power management features present
>> > in the current mainline kernel are insufficient to get maximum possible energy
>> > savings on some platforms, such as Android, because ..." (here go explanations
>> > why this is the case in your opinion).
>> >
>> > Next, "To allow Android and similar platforms to save more energy than they
>> > currently can save using the mainline kernel, introduce a mechanism by which
>> > the system is automatically suspended (i.e. put into a system-wide sleep state)
>> > whenever it's not doing useful work, called opportunistic suspend".
>> >
>> > "For this purpose introduce the suspend blockers framework allowing the
>> > kernel's power management subsystem to decide when it is desirable to suspend
>> > the system (i.e. when useful work is not being done).  Add an API that ..."
>> >
>>
>> PM: Opportunistic suspend support.
>>
>> Power management features present in the current mainline kernel are
>> insufficient to get maximum possible energy savings on some platforms,
>> such as Android, because low power states can only safely be entered
>> from idle.
>
> Do you mean CPU low-power states or system low-power states here?
>
I think either.

> I'd add more details here, because it may not be completely clear to the
> (interested) reader why entering these states only from idle affects the
> possibility to achieve maximum energy savings.  I _guess_ you mean that
> using idle is not sufficient, because there are so many wakeups on the
> systems in question that more savings are still possible.  Is that correct?
>
Yes, is this not what the next paragraph explains?

>> Suspend, in its current form, cannot be used, since wakeup
>> events that occur right after initiating suspend will not be processed
>> until another possibly unrelated event wake up the system again.
>
> I think the word "cannot" is too strong here.  I'd say "not suitable" instead

I don't think it is too strong, but I can change it.

> and I'd say what kind of wakeup events I meant more precisely (there are
> events that wake up a CPU from idle and events that wake up the system from
> suspend).

When I say wakeup events I mean events that do both. Is there another
term for this, or should I just add: "since wakeup events (events that
wake the CPU from idle and the system from suspend)..."

>
>> On some systems idle combined with runtime PM can enter the same power
>> state as suspend, but periodic wakeups increase the average power
>> consumption. Suspending the system also reduces the harm caused by
>> apps that never go idle. On other systems suspend can enter a much
>> lower power state than idle.
>>
>> To allow Android and similar platforms to save more energy than they
>> currently can save using the mainline kernel, we introduce a mechanism
>
> Just "introduce" without "we" works too.
>
>> by which the system is automatically suspended (i.e. put into a
>> system-wide sleep state) whenever it's not doing useful work, called
>> opportunistic suspend.
>
> I think we can address the Kevin's comment about "useful" and say "whenever
> it's only doing work that can be done later without noticeable effect on
> functionality".
>
>> For this purpose introduce the suspend blockers framework allowing the
>> kernel's power management subsystem to decide when it is desirable to
>> suspend the system (i.e. when useful work is not being done). Add an
>
> Perhaps remove the part in parens entirely.
>
>> API that that drivers can use to block opportunistic suspend. This is
>> needed to avoid losing wakeup events that occur right after suspend is
>> initiated.
>>
>> Adds /sys/power/policy that selects the behavior of /sys/power/state.
>> After setting the policy to opportunistic, writes to /sys/power/state
>> become non-blocking requests that specify which suspend state to enter
>> when no suspend blockers are active. A special state, "on", stops the
>> process by activating the "main" suspend blocker.
>
> ...
>
>> PM: suspend_block: Add driver to access suspend blockers from user-space
>>
>> Add a misc device, "suspend_blocker", that allows user-space processes
>> to block automatic suspend.
>>
>> Opening this device creates a suspend blocker that can be used by the
>> opener to prevent automatic suspend from occurring.  There are ioctls
>> provided for blocking and unblocking suspend and for giving the
>> suspend blocker a meaningful name.  Closing the device special file
>> causes the suspend blocker to be destroyed.
>>
>> For example, when select or poll indicates that input event are
>> available, this interface can be used by user space to block suspend
>> before it reads those events. This allows the input driver to release
>> its suspend blocker as soon as the event queue is empty. If user space
>> could not use a suspend blocker here the input driver would need to
>> delay the release of its suspend blocker until it knows (or assumes)
>> that user space has finished processing the events.
>>
>> By careful use of suspend blockers in drivers and user space system
>> code, one can arrange for the system to stay awake for extremely short
>> periods of time in reaction to events, rapidly returning to a fully
>> suspended state.
>
> That's fine by me.
>
> Thanks,
> Rafael
>



--
Arve Hjønnevåg
--
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/