linux-kernel - Re: [PATCH v4] sched/deadline: Fix sched_dl_global

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] [day] [month] [year] [list]

Message-ID: <20201005170607.GA24115@iZj6chx1xj0e0buvshuecpZ>
Date:   Tue, 6 Oct 2020 01:06:07 +0800
From:   Peng Liu <iwtbavbm@...il.com>
To:     Juri Lelli <juri.lelli@...hat.com>
Cc:     mingo@...hat.com, peterz@...radead.org, vincent.guittot@...aro.org,
        dietmar.eggemann@....com, rostedt@...dmis.org, bsegall@...gle.com,
        mgorman@...e.de, bristot@...hat.com, valentin.schneider@....com,
        raistlin@...ux.it, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v4] sched/deadline: Fix sched_dl_global_validate()

On Mon, Oct 05, 2020 at 11:11:00AM +0200, Juri Lelli wrote:

...

> >  	/*
> >  	 * Here we want to check the bandwidth not being set to some
> >  	 * value smaller than the currently allocated bandwidth in
> >  	 * any of the root_domains.
> > -	 *
> > -	 * FIXME: Cycling on all the CPUs is overdoing, but simpler than
> > -	 * cycling on root_domains... Discussion on different/better
> > -	 * solutions is welcome!
> 
> So, this patch changes 2 things: it actually fixes the problem and it
> optimizes scanning of root domains. Even though the changes are limited,
> I'd be more comfortable if we split them in two (fix + optimization).
> 
> Would you be up for doing it?

OK, I will split them.

> 
> >  struct dl_bandwidth {
> >  	raw_spinlock_t		dl_runtime_lock;
> > @@ -801,6 +795,8 @@ struct root_domain {
> >  	struct dl_bw		dl_bw;
> >  	struct cpudl		cpudl;
> >  
> > +	u64			visit_gen;
> > +
> 
> I think this deserves a comment explaining what it is and how it's used.
> 
> Also, do we really need an u64? Maybe an smaller type would be OK as
> well and could fit in the 4 bytes hole that seems available after
> dlo_count, I'm thinking.
> 
> Thanks,
> Juri
> 

u32 is also fine, after all consecutively 2^32 - 1 times validation
failure followed by a success in sched_dl_bandwidth_validate() is still
a very very unlikely case.

On a 64-bit machine, dlo_count should be aligned on a 8 bytes boundary,
there would be a 4 bytes hole between dlo_mask and dl_bw, put the new
member there seems reasonable. I'm OK with this.