[<prev] [next>] [day] [month] [year] [list]
Message-ID: <484F1CE1.3010701@qualcomm.com>
Date: Tue, 10 Jun 2008 17:31:29 -0700
From: Max Krasnyansky <maxk@...lcomm.com>
To: LKML <linux-kernel@...r.kernel.org>
CC: linux-rt-users@...r.kernel.org
Subject: linux v2.6.25.6.cpuisol1
This is the latest version of my CPU isolation tree.
Git tree is available at:
git://git.kernel.org/pub/scm/linux/kernel/git/maxk/cpuisol-2.6.25.y.git
http://www.kernel.org/pub/scm/linux/kernel/git/maxk/cpuisol-2.6.25.y.git
Full patch is at:
http://kernel.org/pub/linux/kernel/people/maxk/cpuisol/patches/v2.6.25.6.cpuisol1.patch
In order to simplify overall system setup I wrote a simple tool for system
partitioning - "syspart". You can get the latest version from here
http://kernel.org/pub/linux/kernel/people/maxk/cpuisol/syspart/
git://git.kernel.org/pub/scm/linux/kernel/git/maxk/syspart.git
This tool takes care of setting up IRQ affinities and cpusets, moving kernel
and user space tasks, bring CPUs off/online, etc. Please note that you want to
disable daemons like "irqbalance" to avoid resource partitioning conflicts.
For more info see syspart/README.
--
I also have some sample kernel configs that have been tested and are know to
work. You can pick one of them as base for your kernel:
http://kernel.org/pub/linux/kernel/people/maxk/cpuisol/configs/
Required kernel features:
CONFIG_CGROUP
CONFIG_CPUSETS
CONFIG_CPU_HOTPLUG
Following features are required if you intend on chewing ~100% in the user
space SCHED_FIFO threads running on the isolated CPUs (note that OPROFILE is
known to not work for this use case):
CONFIG_SLUB or CONFIG_SLOB
You may also want to disable the following features in order to achieve the
lowest possible latency:
!CONFIG_HIGH_RES_TIMERS
!CONFIG_NO_HZ
!CONFIG_CPU_FREQ
And of course you want to refer to
http://rt.wiki.kernel.org/index.php/HOWTO:_Build_an_RT-application
for general guidelines on how to build RT applications.
---
This version is based on a completely different approach (compared to previous
cpuisol tree releases). Instead of the centralized cpu_isolation_map it's now
based on the existing kernel features like cpusets. The number of changes is
therefor very small.
Shortlog:
Max Krasnyansky (6):
sched: Fix memory leak in the cpu hotplug handing logic
sched: CPU hotplug events must not destroy scheduler domains created by
the cpusets
genirq: Expose default irq affinity mask (take 3)
sysctl: Add generic cpumask parser for sysctls
stopmachine: Introduce stopmachine_cpus sysctl
Linux 2.6.25.6.cpuisol1
sched: changes are simple bug fixes and are on their way into the mainline.
genirq: stuff has been ACKed in general is pending final ACK from genirq
maintainer.
stopmachine change is a temporary hack to allow for module insertion and removal.
Diffstat:
Documentation/IRQ-affinity.txt | 37 ++++++++++++++----
Documentation/filesystems/proc.txt | 29 +++++++++-----
Makefile | 2 -
arch/alpha/kernel/irq.c | 5 +-
include/linux/irq.h | 14 ++-----
include/linux/sysctl.h | 2 +
kernel/cpuset.c | 6 +++
kernel/irq/manage.c | 28 +++++++++++++-
kernel/irq/proc.c | 59 +++++++++++++++++++++++++++--
kernel/sched.c | 23 +++++++++++
kernel/stop_machine.c | 5 ++
kernel/sysctl.c | 73 +++++++++++++++++++++++++++++++++++
12 files changed, 244 insertions(+), 39 deletions(-)
Comments, questions, suggestions and bug reports are welcome.
Max
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists