linux-kernel - Re: [tip:sched/core] sched/isolation: Document the isolcpus= flags

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]

Message-ID: <20171030161101.wprohopz5eg7snb4@hirez.programming.kicks-ass.net>
Date:   Mon, 30 Oct 2017 17:11:01 +0100
From:   Peter Zijlstra <peterz@...radead.org>
To:     Christopher Lameter <cl@...ux.com>
Cc:     cmetcalf@...lanox.com, linux-kernel@...r.kernel.org,
        tglx@...utronix.de, torvalds@...ux-foundation.org, hpa@...or.com,
        riel@...hat.com, mingo@...nel.org, efault@....de,
        frederic@...nel.org, kernellwp@...il.com,
        paulmck@...ux.vnet.ibm.com, lcapitulino@...hat.com,
        linux-tip-commits@...r.kernel.org
Subject: Re: [tip:sched/core] sched/isolation: Document the isolcpus= flags

On Mon, Oct 30, 2017 at 10:48:04AM -0500, Christopher Lameter wrote:
> On Fri, 27 Oct 2017, Peter Zijlstra wrote:
> 
> > I _strongly_ object to this statement, isolcpus is _not_ the preferred
> > way, cpusets are.
> >
> > And yes, while cpusets suffers some problems, we _should_ really fix
> > those and not promote this piece of shit isolcpus crap.
> 
> Well low level control at the processor level is important and this allows
> controlling activities on a processor that is supposed to be dedicated to
> certain activities without OS interaction.
> 
> isolcpus is the *right* approach here because you are micromanaging the OS
> and are putting dedicated pieces of software on each core.

That is what you want, and cpusets should allow for that just fine.

> A cgroup suggests that threads would be scheduled over multiple cores
> which is *not* what you want.

No, that suggestion is false. cpusets should allow you to isolate
individual CPUs just fine.

> cgroup has to do something with containers

Sod containers. That's just modern group think. cpusets existed long
before all that wankery and it should very well retain the original use
cases.

That said, I know there's problems with cpusets, and those should be
fixed. But that doesn't mean isolcpus is anything other than a vile
hack.