---
layout: post
status: publish
published: true
title: ! 'Userspace responsiveness .bashrc alternative (Ubuntu 10.10: Working)'
wordpress_id: 437
wordpress_url: http://www.martineve.com/?p=437
date: !binary |-
MjAxMC0xMS0xOSAxOTo0OTowMyArMDEwMA==
date_gmt: !binary |-
MjAxMC0xMS0xOSAxOTo0OTowMyArMDEwMA==
categories:
- Technology
- Linux
tags:
- Technology
- Ubuntu
comments:
- id: 3982
author: Alex Filonov
author_email: muddlin@gmail.com
author_url: ''
date: !binary |-
MjAxMC0xMi0wOCAwMjozOTo1NiArMDEwMA==
date_gmt: !binary |-
MjAxMC0xMi0wOCAwMjozOTo1NiArMDEwMA==
content: ! "Maybe working on 10.10. But on 8.04 I'm getting:\r\n\r\n\r\nalex@gazelle:~$
mkdir -p -m 0700 /dev/cgroup/cpu/user/$$ \r\nmkdir: cannot create directory `/dev/cgroup/cpu/user/6806':
Invalid argument"
- id: 6172
author: foobar
author_email: totolepabo@yahoo.fr
author_url: http://yahoo.fr
date: !binary |-
MjAxMS0wMi0wOCAyMjoyNToxNSArMDEwMA==
date_gmt: !binary |-
MjAxMS0wMi0wOCAyMjoyNToxNSArMDEwMA==
content: ! "Please, can we stop this ? This \"incredible tuning\" is nothing but
a pile of shit. It will ONLY work if you start ALL your processes from a terminal,
or anything having a PS1 prompt.\r\n\r\nfoo@bar:~$ for PID in `cat /dev/cgroup/cpu/user/*/tasks`;do
cut -f1 -d' ' /proc/$PID/cmdline;done\r\n\r\nbash\r\nbash\r\nscreen-Sfoo\r\nSCREEN-Sfoo\r\n/bin/bash\r\nrtorrent\r\nbash\r\nssh\r\nssh\r\n\r\nWOOOO
i'm SO HAPPY my shells are more responsive now !!\r\n\r\nPlease learn a little
bit people."
- id: 6173
author: Martin Paul Eve
author_email: martin@martineve.com
author_url: ''
date: !binary |-
MjAxMS0wMi0wOSAwOTozODowNCArMDEwMA==
date_gmt: !binary |-
MjAxMS0wMi0wOSAwOTozODowNCArMDEwMA==
content: ! "The only reason I approved this comment was so that your rudeness and
unhelpful manner can be made visible. How on earth do you think that adopting
such a condescending, sarcastic tone is likely to help anyone?\r\n\r\nIn response
to your \"argument\" (not quite sure why I'm bothering) I never claimed otherwise.
As is quite clearly discussed in the Slashdot comment thread, this patch mostly
benefits those who run large compile jobs in the background. As I fall into that
group (mostly compiling AOSP), I thought I would share the userspace solution
to a patch that was accepted for kernel merge. Just because you have nothing better
to do on the command line than run rtorrent doesn't mean the situation is the
same for all of us."
- id: 6174
author: foobar
author_email: totolepabo@yahoo.fr
author_url: http://yahoo.fr
date: !binary |-
MjAxMS0wMi0wOSAyMTo0NDo1NSArMDEwMA==
date_gmt: !binary |-
MjAxMS0wMi0wOSAyMTo0NDo1NSArMDEwMA==
content: ! "I am really grateful you did your highness :)\r\n\r\nThe only reason
i'm being so rude on this comment is that there's a BUNCH of blogs fapping to
this while they don't have a clue of what it really does.\r\n\r\nNow, you _are_
right, this will effectively boost up shell jobs (or anything having its parent
given a PS1), but unfortunately, not everyone catch this, and you can read pretty
much everywhere \"WOW THIS IS INCREDIBLE, FIREFOX IS NOW SO MUCH RESPONSIVE ON
UBUNTU !!1!\". That's called a \"placebo effect\".\r\n\r\nNevertheless, thank
you for your reply (and no, i'm not being sarcastic here), it feels good to read
some common sense.\r\n\r\nKeep up the good work !"
- id: 6175
author: Martin Paul Eve
author_email: martin@martineve.com
author_url: ''
date: !binary |-
MjAxMS0wMi0xMCAwODoyNjo0OSArMDEwMA==
date_gmt: !binary |-
MjAxMS0wMi0xMCAwODoyNjo0OSArMDEwMA==
content: Thanks for this reply; I'm sorry if my tone was also rude in the reply.
I agree, there is a certain amount of misinformation circulating as to where this
will benefit users -- many have also pointed out, though, that this mostly benefits...
Linus and his usage model!
---
<p>Recent discussion on the lkm has lead to Linus giving the go-ahead to a large kernel patch that massively increases responsiveness when multi-tasking on Linux desktop. A <a href="http://slashdot.org/story/10/11/18/2246213/Alternative-To-the-200-Line-Linux-Kernel-Patch">slashdot article</a> today also points out that the same result can be achieved by modifications userspace and that this is, indeed, how the patch was tested.</p>
<p>The original instructions can be found at <a href="http://www.webupd8.org/2010/11/alternative-to-200-lines-kernel-patch.html">webupd8</a>, but, for Ubuntu systems, they are incomplete.</p>
<p>Here's the correct information.</p>
<p>/etc/rc.local should contain:</p>
<p><code>mkdir -p /dev/cgroup/cpu<br />
mount -t cgroup cgroup /dev/cgroup/cpu -o cpu<br />
mkdir -m 0777 /dev/cgroup/cpu/user<br />
echo "/usr/local/sbin/cgroup_clean" > /dev/cgroup/cpu/release_agent</code></p>
<p>and should be executable (sudo chmod +x /etc/rc.local).</p>
<p>Your ~/.bashrc should contain:</p>
<p><code>if [ "$PS1" ] ; then<br />
mkdir -p -m 0700 /dev/cgroup/cpu/user/$$ > /dev/null 2>&1<br />
echo $$ > /dev/cgroup/cpu/user/$$/tasks<br />
echo "1" > /dev/cgroup/cpu/user/$$/notify_on_release<br />
fi</code></p>
<p>directly at the top.</p>
<p>Finally (and this is where the original instructions were incomplete), /usr/local/sbin/cgroup_clean should contain:</p>
<p><code>#!/bin/sh<br />
if [ "$1" != "/user" ]; then<br />
rmdir /dev/cgroup/cpu/$1<br />
fi </code></p>
<p>as suggested by an <a href="http://www.webupd8.org/2010/11/alternative-to-200-lines-kernel-patch.html#comment-99182030">anonymous comment</a> on the original. Finally, ensure the cgroup cleanup script is executable: sudo chmod +x /usr/local/sbin/cgroup_clean</p>
<p>Reboot your system and you too can have an <a href="http://www.phoronix.com/scan.php?page=article&item=linux_2637_video&num=2">ultra-responsive desktop</a>!</p>