Re: [PATCH 29/38] tick-sched: remove wrapper around __tick_nohz_task_switch()

[Frederic Weisbecker <fweisbec@gmail.com>]

On Tue, Apr 15, 2014 at 03:23:37PM +0530, Viresh Kumar wrote:
> On 15 April 2014 14:43, Frederic Weisbecker <fweisbec@gmail.com> wrote:
> > Yeah. But not just that.
> >
> > Using an inline saves a function call and reduce the offline case to a simple
> > condition check. But there is also the jump label that reduce the condition check
> > to an unconditional jump in the off case.
> >
> > To summarize, here's how calling tick_nohz_task_switch() maps to final C code:
> >
> > finish_task_switch()
> > {
> >        //do things before calling tick_nohz_task_switch()...
> >        // call tick_nohz_task_switch
> >        goto offcase;
> >        if (tick_nohz_full_enabled())
> >            __tick_nohz_task_switch(tsk);
> > offcase:
> >       //end of call to tick_nohz_task_switch
> >       //do things before calling tick_nohz_task_switch()...
> > }
> >
> > In the offcase, the code is like above. We don't even do the check, thanks to
> > the jump label code we unconditionally jump to what's next in finish_task_switch()
> > (there is actually nothing afterward but that's for the picture).
> >
> > Now if there is at least a CPU that is full dynticks on boot, it is enabled
> > with context_tracking_cpu_set(). Then the jump label code patches the code in
> > finish_task_switch() to turn the goto offcase into a nop. Then the condition is
> > actually verified on every call to finish_task_switch().
> >
> > So it goes beyond than just saving a function call.
> 
> Sorry, but my poor mind still couldn't understand what you are trying to
> tell me :(

Welcome to the club of the daily confused people.
I'm happy to hear I'm not alone :)

> 
> So lets clarify things one by one :)
> 
> - What do you mean by offcase? CONFIG_NO_HZ_FULL not configured
> into the kernel or it is configured but none of the CPUs is running in that
> mode?

So by offcase I mean CONFIG_NO_HZ_FULL=y but the nohz_full boot parameter
is empty, or simply not passed at all. And of course CONFIG_NO_HZ_FULL_ALL=n

This config is now likely on some distros because we want to make full
dynticks available for users who want it. But if it's not used (which is 99.999%
of the usecases), we want to minimize as much as possible its overhead.

Lets call that dynamic off-case.

> 
> - Also what does it correspond to in code: goto offcase; ? There is no labels
> or goto statements in code that I can see.. This is how the code looks to me.
> 
> > finish_task_switch()
> > {
> >        //do things before calling tick_nohz_task_switch()...
> >        // call tick_nohz_task_switch
> >        if (tick_nohz_full_enabled())
> >            __tick_nohz_task_switch(tsk);
> > }

Sure but check out the static_key_false() in the implementation of tick_nohz_full_enabled().
That's where the magic hides.

> 
> __tick_nohz_task_switch() may or maynot be available at all depending
> on CONFIG_NO_HZ_FULL is enabled into the kernel or not. But that
> was the case with tick_nohz_task_switch() as well in my patch. So
> shouldn't make a difference..
> 
> Again, sorry for not understanding what you are trying to explain here.
> I want to understand this once and for all and probably add a comment
> here as well :)

No problem, the jump label/static key code is quite tricky. And its use
can be easily missed, as in here.

Also its unfamous API naming (static_key_true/static_key_true) that is
anything but intuitive.

> 
> --
> viresh