From: Arve Hjønnevåg on
2010/4/28 Rafael J. Wysocki <rjw(a)sisk.pl>:
> On Wednesday 28 April 2010, Oleg Nesterov wrote:
>> On 04/27, Arve Hj�nnev�g wrote:
>> >
>> > Allow work to be queued that will block suspend while it is pending
>> > or executing. To get the same functionality in the calling code often
>> > requires a separate suspend_blocker for pending and executing work, or
>> > additional state and locking. This implementation does add additional
>> > state and locking, but this can be removed later if we add support for
>> > suspend blocking work to the core workqueue code.
>>
>> I think this patch is fine.
>>
>> Just one silly question,
>>
>> > +int queue_suspend_blocking_work(struct workqueue_struct *wq,
>> > + � � � � � � � � � � � � � struct suspend_blocking_work *work)
>> > +{
>> > + � int ret;
>> > + � unsigned long flags;
>> > +
>> > + � spin_lock_irqsave(&work->lock, flags);
>> > + � suspend_block(&work->suspend_blocker);
>> > + � ret = queue_work(wq, &work->work);
>> > + � if (ret)
>> > + � � � � � work->active++;
>>
>> why not
>>
>> � � � ret = queue_work(wq, &work->work);
>> � � � if (ret) {
>> � � � � � � � suspend_block(&work->suspend_blocker);
>> � � � � � � � work->active++;
>> � � � }
>>
>> ?
>>
>> Afaics, we can't race with work->func() doing unblock, we hold work-lock.
>> And this way the code looks more clear.
>
> Agreed. �Arve, any objections to doing that?
>

I need to fix the race, but I can easily fix it in
cancel_suspend_blocking_work_sync instead. If the suspend blocker is
active for a long time, and DEBUG_SUSPEND_BLOCKER is enabled, we can
tell if the work is constantly re-queued or if the workqueue is stuck.

>> Sorry, I had no chance to read the previous patches. After the quick look
>> at 1/8 I think it is OK to call suspend_block() twice, but still...
>
> It is.
>
>> Or I missed something? Just curious.
>>
>>
>> Hmm... actually, queue_work() can also fail if we race with cancel_ which
>> temporary sets WORK_STRUCT_PENDING. In that case suspend_block() won't
>> be paired by unblock ?
>>
>>
>> > +int schedule_suspend_blocking_work(struct suspend_blocking_work *work)
>> > +{
>> > ...
>> > + � ret = schedule_work(&work->work);
>>
>> Off-topic. We should probably export keventd_wq to avoid the duplications
>> like this.
>
> Please see my reply to Tejun. :-)
>
> Rafael
>



--
Arve Hj�nnev�g
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Rafael J. Wysocki on
On Thursday 29 April 2010, Arve Hj�nnev�g wrote:
> 2010/4/28 Rafael J. Wysocki <rjw(a)sisk.pl>:
> > On Wednesday 28 April 2010, Oleg Nesterov wrote:
> >> On 04/27, Arve Hj�nnev�g wrote:
> >> >
> >> > Allow work to be queued that will block suspend while it is pending
> >> > or executing. To get the same functionality in the calling code often
> >> > requires a separate suspend_blocker for pending and executing work, or
> >> > additional state and locking. This implementation does add additional
> >> > state and locking, but this can be removed later if we add support for
> >> > suspend blocking work to the core workqueue code.
> >>
> >> I think this patch is fine.
> >>
> >> Just one silly question,
> >>
> >> > +int queue_suspend_blocking_work(struct workqueue_struct *wq,
> >> > + struct suspend_blocking_work *work)
> >> > +{
> >> > + int ret;
> >> > + unsigned long flags;
> >> > +
> >> > + spin_lock_irqsave(&work->lock, flags);
> >> > + suspend_block(&work->suspend_blocker);
> >> > + ret = queue_work(wq, &work->work);
> >> > + if (ret)
> >> > + work->active++;
> >>
> >> why not
> >>
> >> ret = queue_work(wq, &work->work);
> >> if (ret) {
> >> suspend_block(&work->suspend_blocker);
> >> work->active++;
> >> }
> >>
> >> ?
> >>
> >> Afaics, we can't race with work->func() doing unblock, we hold work-lock.
> >> And this way the code looks more clear.
> >
> > Agreed. Arve, any objections to doing that?
> >
>
> I need to fix the race, but I can easily fix it in
> cancel_suspend_blocking_work_sync instead. If the suspend blocker is
> active for a long time, and DEBUG_SUSPEND_BLOCKER is enabled, we can
> tell if the work is constantly re-queued or if the workqueue is stuck.

Well, perhaps that's worth adding a comment to the code. The debug part is not
immediately visible from the code itself.

Rafael
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Arve Hjønnevåg on
2010/4/28 Rafael J. Wysocki <rjw(a)sisk.pl>:
> On Thursday 29 April 2010, Arve Hj�nnev�g wrote:
>> 2010/4/28 Rafael J. Wysocki <rjw(a)sisk.pl>:
>> > On Wednesday 28 April 2010, Oleg Nesterov wrote:
>> >> On 04/27, Arve Hj�nnev�g wrote:
>> >> >
>> >> > Allow work to be queued that will block suspend while it is pending
>> >> > or executing. To get the same functionality in the calling code often
>> >> > requires a separate suspend_blocker for pending and executing work, or
>> >> > additional state and locking. This implementation does add additional
>> >> > state and locking, but this can be removed later if we add support for
>> >> > suspend blocking work to the core workqueue code.
>> >>
>> >> I think this patch is fine.
>> >>
>> >> Just one silly question,
>> >>
>> >> > +int queue_suspend_blocking_work(struct workqueue_struct *wq,
>> >> > + � � � � � � � � � � � � � struct suspend_blocking_work *work)
>> >> > +{
>> >> > + � int ret;
>> >> > + � unsigned long flags;
>> >> > +
>> >> > + � spin_lock_irqsave(&work->lock, flags);
>> >> > + � suspend_block(&work->suspend_blocker);
>> >> > + � ret = queue_work(wq, &work->work);
>> >> > + � if (ret)
>> >> > + � � � � � work->active++;
>> >>
>> >> why not
>> >>
>> >> � � � ret = queue_work(wq, &work->work);
>> >> � � � if (ret) {
>> >> � � � � � � � suspend_block(&work->suspend_blocker);
>> >> � � � � � � � work->active++;
>> >> � � � }
>> >>
>> >> ?
>> >>
>> >> Afaics, we can't race with work->func() doing unblock, we hold work-lock.
>> >> And this way the code looks more clear.
>> >
>> > Agreed. �Arve, any objections to doing that?
>> >
>>
>> I need to fix the race, but I can easily fix it in
>> cancel_suspend_blocking_work_sync instead. If the suspend blocker is
>> active for a long time, and DEBUG_SUSPEND_BLOCKER is enabled, we can
>> tell if the work is constantly re-queued or if the workqueue is stuck.
>
> Well, perhaps that's worth adding a comment to the code. �The debug part is not
> immediately visible from the code itself.

On second thought, this only makes a difference if both conditions are
true. If we are constantly re-queuing the work but it is not stuck,
either method will show the debug message, so I used Oleg's
suggestion.

--
Arve Hj�nnev�g
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Rafael J. Wysocki on
On Thursday 29 April 2010, Arve Hj�nnev�g wrote:
> 2010/4/28 Rafael J. Wysocki <rjw(a)sisk.pl>:
> > On Thursday 29 April 2010, Arve Hj�nnev�g wrote:
> >> 2010/4/28 Rafael J. Wysocki <rjw(a)sisk.pl>:
> >> > On Wednesday 28 April 2010, Oleg Nesterov wrote:
> >> >> On 04/27, Arve Hj�nnev�g wrote:
> >> >> >
> >> >> > Allow work to be queued that will block suspend while it is pending
> >> >> > or executing. To get the same functionality in the calling code often
> >> >> > requires a separate suspend_blocker for pending and executing work, or
> >> >> > additional state and locking. This implementation does add additional
> >> >> > state and locking, but this can be removed later if we add support for
> >> >> > suspend blocking work to the core workqueue code.
> >> >>
> >> >> I think this patch is fine.
> >> >>
> >> >> Just one silly question,
> >> >>
> >> >> > +int queue_suspend_blocking_work(struct workqueue_struct *wq,
> >> >> > + struct suspend_blocking_work *work)
> >> >> > +{
> >> >> > + int ret;
> >> >> > + unsigned long flags;
> >> >> > +
> >> >> > + spin_lock_irqsave(&work->lock, flags);
> >> >> > + suspend_block(&work->suspend_blocker);
> >> >> > + ret = queue_work(wq, &work->work);
> >> >> > + if (ret)
> >> >> > + work->active++;
> >> >>
> >> >> why not
> >> >>
> >> >> ret = queue_work(wq, &work->work);
> >> >> if (ret) {
> >> >> suspend_block(&work->suspend_blocker);
> >> >> work->active++;
> >> >> }
> >> >>
> >> >> ?
> >> >>
> >> >> Afaics, we can't race with work->func() doing unblock, we hold work-lock.
> >> >> And this way the code looks more clear.
> >> >
> >> > Agreed. Arve, any objections to doing that?
> >> >
> >>
> >> I need to fix the race, but I can easily fix it in
> >> cancel_suspend_blocking_work_sync instead. If the suspend blocker is
> >> active for a long time, and DEBUG_SUSPEND_BLOCKER is enabled, we can
> >> tell if the work is constantly re-queued or if the workqueue is stuck.
> >
> > Well, perhaps that's worth adding a comment to the code. The debug part is not
> > immediately visible from the code itself.
>
> On second thought, this only makes a difference if both conditions are
> true. If we are constantly re-queuing the work but it is not stuck,
> either method will show the debug message, so I used Oleg's
> suggestion.

OK, great.

Rafael
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Pavel Machek on
Hi!

> Allow work to be queued that will block suspend while it is pending
> or executing. To get the same functionality in the calling code often
> requires a separate suspend_blocker for pending and executing work, or
> additional state and locking. This implementation does add additional
> state and locking, but this can be removed later if we add support for
> suspend blocking work to the core workqueue code.
>
> Signed-off-by: Arve Hj??nnev??g <arve(a)android.com>

Seems sane. ACK.

> +struct suspend_blocking_work {
> + struct work_struct work;
> +#ifdef CONFIG_OPPORTUNISTIC_SUSPEND

Cound we name it C_AUTO_SUSPEND... to reduce length and typo
potential?

> + struct suspend_blocker suspend_blocker;
> + work_func_t func;
> + spinlock_t lock;
> + int active;

Is the lock internal-use, or is API user allowed to use it?
Pavel

--
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/