linux-kernel - Re: [Patch v3 3/6] sched/fair: Implement prefer sibling imbalance calculation between asymmetric groups

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, 14 Jul 2023 16:35:12 -0700
From:   Tim Chen <tim.c.chen@...ux.intel.com>
To:     Tobias Huschle <huschle@...ux.ibm.com>,
        Shrikanth Hegde <sshegde@...ux.vnet.ibm.com>
Cc:     Peter Zijlstra <peterz@...radead.org>,
        Juri Lelli <juri.lelli@...hat.com>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Ricardo Neri <ricardo.neri@...el.com>,
        "Ravi V . Shankar" <ravi.v.shankar@...el.com>,
        Ben Segall <bsegall@...gle.com>,
        Daniel Bristot de Oliveira <bristot@...hat.com>,
        Dietmar Eggemann <dietmar.eggemann@....com>,
        Len Brown <len.brown@...el.com>, Mel Gorman <mgorman@...e.de>,
        "Rafael J . Wysocki" <rafael.j.wysocki@...el.com>,
        Srinivas Pandruvada <srinivas.pandruvada@...ux.intel.com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Valentin Schneider <vschneid@...hat.com>,
        Ionela Voinescu <ionela.voinescu@....com>, x86@...nel.org,
        linux-kernel@...r.kernel.org,
        Srikar Dronamraju <srikar@...ux.vnet.ibm.com>,
        naveen.n.rao@...ux.vnet.ibm.com,
        Yicong Yang <yangyicong@...ilicon.com>,
        Barry Song <v-songbaohua@...o.com>,
        Chen Yu <yu.c.chen@...el.com>, Hillf Danton <hdanton@...a.com>
Subject: Re: [Patch v3 3/6] sched/fair: Implement prefer sibling imbalance
 calculation between asymmetric groups

On Fri, 2023-07-14 at 16:22 +0200, Tobias Huschle wrote:
> 
> > 
> > Could this work for case where number of CPU/cores would differ
> > between two sched groups in a sched domain? Such as problem pointed
> > by tobias on S390. It would be nice if this patch can work for that 
> > case
> > as well. Ran numbers for a few cases. It looks to work.
> > https://lore.kernel.org/lkml/20230704134024.GV4253@hirez.programming.kicks-ass.net/T/#rb0a7dcd28532cafc24101e1d0aed79e6342e3901
> > 
> 
> 
> Just stumbled upon this patch series as well. In this version it looks
> similar to the prototypes I played around with, but more complete.
> So I'm happy that my understanding of the load balancer was kinda 
> correct :)
> 
>  From a functional perspective this appears to address the issues we saw 
> on s390.

Glad that this patch addresses this common issue across architectures.
I did aim to address the asymmetric groups balancing in general.
Peter pointed out this problem that's inherent when he reviewed the first
version of my patchset. 

Tim

> 
> > 
> > 
> > > +	/* Take advantage of resource in an empty sched group */
> > > +	if (imbalance == 0 && local->sum_nr_running == 0 &&
> > > +	    busiest->sum_nr_running > 1)
> > > +		imbalance = 2;
> > > +
> > 
> > I don't see how this case would be true. When there are unequal number
> > of cores and local->sum_nr_ruuning
> > is 0, and busiest->sum_nr_running is atleast 2, imbalance will be 
> > atleast 1.
> > 
> > 
> > Reviewed-by: Shrikanth Hegde <sshegde@...ux.vnet.ibm.com>
> > 
> > > +	return imbalance;
> > > +}
> > > +
> > >  static inline bool
> > >  sched_reduced_capacity(struct rq *rq, struct sched_domain *sd)
> > >  {
> > > @@ -10230,14 +10265,12 @@ static inline void 
> > > calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s
> > >  		}
> > > 
> > >  		if (busiest->group_weight == 1 || sds->prefer_sibling) {
> > > -			unsigned int nr_diff = busiest->sum_nr_running;
> > >  			/*
> > >  			 * When prefer sibling, evenly spread running tasks on
> > >  			 * groups.
> > >  			 */
> > >  			env->migration_type = migrate_task;
> > > -			lsub_positive(&nr_diff, local->sum_nr_running);
> > > -			env->imbalance = nr_diff;
> > > +			env->imbalance = sibling_imbalance(env, sds, busiest, local);
> > >  		} else {
> > > 
> > >  			/*
> > > @@ -10424,7 +10457,7 @@ static struct sched_group 
> > > *find_busiest_group(struct lb_env *env)
> > >  	 * group's child domain.
> > >  	 */
> > >  	if (sds.prefer_sibling && local->group_type == group_has_spare &&
> > > -	    busiest->sum_nr_running > local->sum_nr_running + 1)
> > > +	    sibling_imbalance(env, &sds, busiest, local) > 1)
> > >  		goto force_balance;
> > > 
> > >  	if (busiest->group_type != group_overloaded) {