Re: [PATCH] tty: Only wakeup the line discipline idle queue when queue is active

[Preeti U Murthy <preeti@linux.vnet.ibm.com>]

Hi Ivo,
Can you explain how this problem could create a scheduler overhead?
I am a little confused, because as far as i know,scheduler does not come
in the picture of the wake up path right? select_task_rq() in
try_to_wake_up() is where the scheduler comes in,and this is after the
task wakes up.

On 01/03/2013 03:19 PM, Ivo Sieben wrote:
> Oleg, Peter, Ingo, Andi & Preeti,
> 
> 2013/1/2 Jiri Slaby <jslaby@suse.cz>:
>> On 01/02/2013 04:21 PM, Ivo Sieben wrote:
>>> I don't understand your responses: do you suggest to implement this
>>> "if active" behavior in:
>>> * A new wake_up function called wake_up_if_active() that is part of
>>> the waitqueue layer?
>>
>> Sounds good.
>>
>> --
>> js
>> suse labs
> 
> I want to ask you 'scheduler' people for your opinion:
> 
> Maybe you remember my previous patch where I suggested an extra
> 'waitqueue empty' check before entering the critical section of the
> wakeup() function (If you do not remember see
> https://lkml.org/lkml/2012/10/25/159)
> 
> Finally Oleg responded that a lot of callers do
> 
> 	if (waitqueue_active(q))
> 		wake_up(...);
> 
> what made my patch pointless and adds a memory barrier. I then decided
> to also implement the 'waitqueue_active' approach for my problem.
> 
> But now I get a review comment by Jiri that he would like to hide this
> 'if active behavior' in a wake_up_if_active() kind of function. I
> think he is right that implementing this check in the wakeup function
> would clean things up, right?
> 
> I would like to have your opinion on the following two suggestions:
> - We still can do the original patch on the wake_up() that I
> suggested. I then can do an additional code cleanup patch that removes
> the double 'waitqueue_active' call (a quick grep found about 150 of
> these waitqueue active calls) on several places in the code.

I think this is a good move.

> - Or - as an alternative - I could add extra _if_active() versions of
> all wake_up() functions, that implement this extra test.

Why add 'extra' if_active versions? Why not optimize this within the
existing wake_up() functions?
> 
> Regards,
> Ivo

Regards
Preeti U Murthy
>