linux-kernel - Re: [PATCH] sched_rt: Migrate equal priority tasks to available CPUs

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: <1315947117.26295.22.camel@gandalf.stny.rr.com>
Date:	Tue, 13 Sep 2011 16:51:57 -0400
From:	Steven Rostedt <rostedt@...dmis.org>
To:	Shawn Bohrer <sbohrer@...advisors.com>
Cc:	Ingo Molnar <mingo@...e.hu>, Peter Zijlstra <peterz@...radead.org>,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH] sched_rt: Migrate equal priority tasks to available
 CPUs

On Tue, 2011-09-13 at 15:06 -0500, Shawn Bohrer wrote:

> Alright, I'm still having trouble making a case that actually resulted
> in some CPUs left completely idle while others are loaded, but the
> rough test case below does still show the benefit of the patch.

Actually, I have a reliable reproducer. I took my rt-migrate-test.c that
I created for testing the original push/pull schedule algorithm and
added a -e switch. It requires >2 CPUs and does the following. Keeps the
highest priority task high, and then #cpus-1 tasks at the same priority,
and a #cpus + 1 task at the lowest priority. If the lowest priority task
runs before any of the other tasks, then it fails the test.

./rt-migrate-test -e -c

It sometimes takes a few iterations to trigger the failure, but it
doesn't take too long.

I ran a version with your patch and I could not trigger the bug. I
consider this a real regression.

Acked-by: Steven Rostedt <rostedt@...dmis.org>
Tested-by: Steven Rostedt <rostedt@...dmis.org>

-- Steve

View attachment "rt-migrate-test.c" of type "text/x-csrc" (13658 bytes)