Prev: [PATCH] ACPI: skip checking BM_STS if the BIOS doesn't ask for it
Next: generic changes for memblock
From: Tejun Heo on 27 Jul 2010 04:20
Hello,

On 07/26/2010 09:57 PM, Michael S. Tsirkin wrote:
>> For freeze, it probably is okay but for stop, I think it's better to
>> keep the semantics straight forward.
>
> What are the semantics then? What do we want stop followed
> by queue and flush to do?

One scenario I can think of is the following.

kthread_worker allows kthreads to be attached and stopped anytime, so
if the caller stops the current worker while flushing is pending and
attaches a new worker, the flushing which was pending will never
happen.

But, in general, it's nasty to allow execution and its completion to
be separated. Things like that are likely to bite us back in obscure
ways. I think it would be silly to have such oddity in generic code
when it can be avoided without too much trouble.

Thanks.

--
tejun
--
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: Tejun Heo on 27 Jul 2010 04:30
Hello,

On 07/26/2010 10:19 PM, Michael S. Tsirkin wrote:
> Let's try to define what do we want to achieve then. Do you want
> code that flushes workers not to block when workers are frozen? How
> will we handle work submitted when worker is frozen?

As I wrote earlier, it's not necessarily about correctness but rather
avoiding unnecessary surprises and of course flushing can and should
stall if the queue is frozen but let's not separate execution of a
work and its completion with something which can take undeterminate
amount of time.

Thanks.

--
tejun
--
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: Tejun Heo on 28 Jul 2010 03:50
On 07/27/2010 09:19 PM, Michael S. Tsirkin wrote:
>> Thinking a bit more about it, it kind of sucks that queueing to
>> another worker from worker->func() breaks flush. Maybe the right
>> thing to do there is using atomic_t for done_seq?
>
> I don't believe it will help: we might have:
>
> worker1 runs work
> work requeues itself queued index = 1
> worker1 reads queued index = 1
> worker2 runs work
> work requeues itself queued index = 2
> worker2 runs work
> worker2 reads queued index = 2
> worker2 writes done index = 2
> worker1 writes done index = 1
>
> As you see, done index got moved back.

Yeah, I think the flushing logic should be moved to the worker. Are
you interested in doing it w/ your change?

Thanks.

--
tejun
--
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: Tejun Heo on 28 Jul 2010 08:10
Hello,

On 07/28/2010 12:48 PM, Michael S. Tsirkin wrote:
> I'm unsure how flush_work operates under these conditions. E.g. in
> workqueue.c, this seems to work by keeping a pointer to current
> workqueue in the work. But what prevents us from destroying the
> workqueue when work might not be running?

In cmwq, work points to the gcwq it was on, which keeps track of all
the works in progress, so flushing work which is on a destroyed
workqueue should be fine, but in the original implementation, it would
end up accessing freed memory.

> Is this currently broken if you use multiple workqueues
> for the same work? If yes, I propose we do as I did,
> making flush_work get worker pointer, and only flushing
> on that worker.

The original semantics of workqueue is that flush_work() guarantees
that the work has finished executing on the workqueue it was last
queued on. Adding @worker to flush_work() is okay, I think.

Thanks.

--
tejun
--
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/
First  |  Prev  | 
Pages: 1 2 3
Prev: [PATCH] ACPI: skip checking BM_STS if the BIOS doesn't ask for it
Next: generic changes for memblock