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


Groups > linux.kernel > #1372332 > unrolled thread

Re: [PATCH] sched/deadline: No need to check NULL later_mask

Started byPeter Zijlstra <peterz@infradead.org>
First post2016-04-06 11:40 +0200
Last post2016-04-06 15:10 +0200
Articles 2 — 2 participants

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: [PATCH] sched/deadline: No need to check NULL later_mask Peter Zijlstra <peterz@infradead.org> - 2016-04-06 11:40 +0200
    Re: [PATCH] sched/deadline: No need to check NULL later_mask Xunlei Pang <xpang@redhat.com> - 2016-04-06 15:10 +0200

#1372332 — Re: [PATCH] sched/deadline: No need to check NULL later_mask

FromPeter Zijlstra <peterz@infradead.org>
Date2016-04-06 11:40 +0200
SubjectRe: [PATCH] sched/deadline: No need to check NULL later_mask
Message-ID<rkS6m-1Kx-15@gated-at.bofh.it>
On Sat, Apr 02, 2016 at 06:14:28PM +0800, Xunlei Pang wrote:
> Your proposal is very nice!
> 
> At the sched_init() stage we only have one (to be "idle") task and with irq disabled,
> no scheduling will happen, and the cpu_possible_mask was already initiated, so it's
> safe to simply move them there.
> 
> Also, how about rt&deadline sharing a percpu mask? Because only one of them can
> use the mask at a moment, operations are always under some spin_lock_irqsave().
> 
> I made a new patch below, slightly tested by running tens of rt&dl tasks for a while,
> are you fine with it?

Yep, looks fine. Please submit as a proper patch.

Thanks!

[toc] | [next] | [standalone]


#1372516

FromXunlei Pang <xpang@redhat.com>
Date2016-04-06 15:10 +0200
Message-ID<rkVnA-4fv-23@gated-at.bofh.it>
In reply to#1372332
On 2016/04/06 at 17:30, Peter Zijlstra wrote:
> On Sat, Apr 02, 2016 at 06:14:28PM +0800, Xunlei Pang wrote:
>> Your proposal is very nice!
>>
>> At the sched_init() stage we only have one (to be "idle") task and with irq disabled,
>> no scheduling will happen, and the cpu_possible_mask was already initiated, so it's
>> safe to simply move them there.
>>
>> Also, how about rt&deadline sharing a percpu mask? Because only one of them can
>> use the mask at a moment, operations are always under some spin_lock_irqsave().
>>
>> I made a new patch below, slightly tested by running tens of rt&dl tasks for a while,
>> are you fine with it?
> Yep, looks fine. Please submit as a proper patch.

Will do, thanks!

Regards,
Xunlei

>
> Thanks!

[toc] | [prev] | [standalone]


Back to top | Article view | linux.kernel


csiph-web