Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1444372 > unrolled thread

Re: [ANNOUNCE] 4.6.4-rt8

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [ANNOUNCE] 4.6.4-rt8 Steven Rostedt <rostedt@goodmis.org> - 2016-07-15 17:30 +0200
    Re: [ANNOUNCE] 4.6.4-rt8 Sebastian Andrzej Siewior <bigeasy@linutronix.de> - 2016-07-15 17:40 +0200
      Re: [ANNOUNCE] 4.6.4-rt8 Steven Rostedt <rostedt@goodmis.org> - 2016-07-15 17:40 +0200

#1444372 — Re: [ANNOUNCE] 4.6.4-rt8

On Fri, 15 Jul 2016 13:48:04 +0200
Sebastian Andrzej Siewior <bigeasy@linutronix.de> wrote:

> Dear RT folks!
> 
> I'm pleased to announce the v4.6.4-rt8 patch set. 
> 
> Changes since v4.6.4-rt8:
> 
>   - Import Thomas' timer rework known as "timer: Refactor the timer
>     wheel" patch set which made its way into the -TIP tree. With this
>     changes we get NOHZ_FULL working. Finally.

Does this mean that there's no more 1 second tick?

> 
>   - Avoid warning of an unused symbol in the !RT case
>     (preemptible_lazy())
> 
>   - Replace the "trace event preempt count" fixup with Steven's version.
> 
> Known issues
> 	- CPU hotplug got a little better but can deadlock.

Hmm, better but can deadlock.

-- Steve

[toc] | [next] | [standalone]

#1444384

* Steven Rostedt | 2016-07-15 11:22:44 [-0400]:

>> Changes since v4.6.4-rt8:
>> 
>>   - Import Thomas' timer rework known as "timer: Refactor the timer
>>     wheel" patch set which made its way into the -TIP tree. With this
>>     changes we get NOHZ_FULL working. Finally.
>
>Does this mean that there's no more 1 second tick?

From tracing there a timer wakeup each second but before we had a wakeup
each jiffy because we could not lookup the next timer without taking the
base lock which we can do now.
However if there is room for improvement to have a wakeup every two or
five seconds I am all for it because from RT point of view we should be
able to do the same things as the non-RT version.

Sebastian

[toc] | [prev] | [next] | [standalone]

#1444387

On Fri, 15 Jul 2016 17:31:22 +0200
Sebastian Andrzej Siewior <bigeasy@linutronix.de> wrote:

> * Steven Rostedt | 2016-07-15 11:22:44 [-0400]:
> 
> >> Changes since v4.6.4-rt8:
> >> 
> >>   - Import Thomas' timer rework known as "timer: Refactor the timer
> >>     wheel" patch set which made its way into the -TIP tree. With this
> >>     changes we get NOHZ_FULL working. Finally.  
> >
> >Does this mean that there's no more 1 second tick?  
> 
> >From tracing there a timer wakeup each second but before we had a wakeup  
> each jiffy because we could not lookup the next timer without taking the
> base lock which we can do now.
> However if there is room for improvement to have a wakeup every two or
> five seconds I am all for it because from RT point of view we should be
> able to do the same things as the non-RT version.

Ah, that's right. I forgot that NOHZ_FULL was broken in -rt, which
sorta defeated the point, as NOHZ_FULL is a birth child of -rt in the
first place.

OK, thanks for clarifying.

-- Steve

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web