linux-kernel - Re: [RFC] cpuset: remove sched domain hooks from cpusets

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 for Android: free password hash cracker in your pocket

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

Date:	Thu, 19 Oct 2006 20:24:59 +1000
From:	Nick Piggin <nickpiggin@...oo.com.au>
To:	Paul Jackson <pj@....com>
CC:	Andrew Morton <akpm@...l.org>, Martin Bligh <mbligh@...gle.com>,
	Paul Menage <menage@...gle.com>, Simon.Derr@...l.net,
	linux-kernel@...r.kernel.org,
	Dinakar Guniguntala <dino@...ibm.com>,
	Rohit Seth <rohitseth@...gle.com>, Robin Holt <holt@....com>,
	dipankar@...ibm.com, "Siddha, Suresh B" <suresh.b.siddha@...el.com>
Subject: Re: [RFC] cpuset: remove sched domain hooks from cpusets

Paul Jackson wrote:
> From: Paul Jackson <pj@....com>
> 
> Remove the cpuset hooks that defined sched domains depending on the
> setting of the 'cpu_exclusive' flag.

Before we chuck the baby out...

> The cpu_exclusive flag can only be set on a child if it is set on
> the parent.
> 
> This made that flag painfully unsuitable for use as a flag defining
> a partitioning of a system.

Sigh, it isn't. That's simply how cpusets tried to use it.

> It was entirely unobvious to a cpuset user what partitioning of sched
> domains they would be causing when they set that one cpu_exclusive bit
> on one cpuset, because it depended on what CPUs were in the remainder
> of that cpusets siblings and child cpusets, after subtracting out
> other cpu_exclusive cpusets.

As far as a user is concerned, the cpusets is the interface. domain
partitioning is an implementation detail that just happens to make
it work better in some cases.

> Furthermore, there was no way on production systems to query the
> result.

You shouldn't need to, assuming cpusets doesn't mess it up.

Here is an untested patch. Apparently takes care of CPU hotplug too.

-- 
SUSE Labs, Novell Inc.

View attachment "sched-domains-cpusets-fixes.patch" of type "text/plain" (8468 bytes)