linux-kernel - Re: [PATCH v3] sched/fair: update scale invariance of PELT

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 18 May 2018 14:08:51 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     Patrick Bellasi <patrick.bellasi@....com>
Cc:     Vincent Guittot <vincent.guittot@...aro.org>, mingo@...nel.org,
        linux-kernel@...r.kernel.org, dietmar.eggemann@....com,
        Morten.Rasmussen@....com, yuyang.du@...el.com, pjt@...gle.com,
        bsegall@...gle.com, "Rafael J. Wysocki" <rjw@...ysocki.net>
Subject: Re: [PATCH v3] sched/fair: update scale invariance of PELT

On Fri, May 18, 2018 at 11:17:38AM +0100, Patrick Bellasi wrote:
> On 18-May 11:36, Peter Zijlstra wrote:
> > 
> > Replying to the latest version available; given the current interest I
> > figure I'd re-read some of the old threads and look at this stuff again.
> > 
> > On Fri, Apr 28, 2017 at 04:23:55PM +0200, Vincent Guittot wrote:
> > 
> > > diff --git a/include/linux/sched.h b/include/linux/sched.h
> > > index 0978fb7..f8dde36 100644
> > > --- a/include/linux/sched.h
> > > +++ b/include/linux/sched.h
> > > @@ -313,6 +313,7 @@ struct load_weight {
> > >   */
> > >  struct sched_avg {
> > >  	u64				last_update_time;
> > > +	u64				stolen_idle_time;
> > >  	u64				load_sum;
> > >  	u32				util_sum;
> > >  	u32				period_contrib;
> > 
> > Right, so sadly Patrick stole that space with the util_est bits.
> 
> Sorry :(
> 
> However, I remember we already talked about the idea to
> update load_avg and runnable_load_avg to use u32:
> 
>    https://marc.info/?l=linux-kernel&m=151334269426419&w=2
> 

Sadly not I fear, (runnable_)load for entities is obviously limited in
range, but we use the same struct sched_avg for the cfs_rq aggregates,
and there is no real limit on the total weight of a cfs_rq.

However, I think we can shrink util_avg unconditionally, which would
give us exactly the 4 byte hole we need.

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 959a8588e365..670b53b10a68 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -402,7 +402,8 @@ struct sched_avg {
 	u32				period_contrib;
 	unsigned long			load_avg;
 	unsigned long			runnable_load_avg;
-	unsigned long			util_avg;
+	u32				util_avg;
+	u32				stolen_idle_time;
 	struct util_est			util_est;
 } ____cacheline_aligned;