[lkml]   [2010]   [Oct]   [26]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectFwd: 2.6.36 io bring the system to its knees
"And yes, we'd very much like to fix such slowdowns via heuristics as
well (detecting large sequential IO and not letting it poison the
existing cache), so good bugreports and reproducing testcases sent to and people willing to try out
experimental kernel patches would definitely be welcome.




I'll be rather quick & to the point here.

I get & run stable kernels the same day they appear on in
hope to get away from these annoying, ignored, neglected slowdowns.

.config attached - I have Lenovo ThinkPad T400, Core2Duo T9400, 4Gb
DDR2, w/integrated GM45, iwlagn for the intel 5300 wifi, CFS, ext2 for
swap partition, ext3 for boot, ext4 for everything else.
All the hardware I have runs linux natively.
No kernel helped me from the days of 2.6.28.x upto 2.6.36. The dubbed
slowdown fixes never worked for me.
The kernel config choices are rather typical : NO_HZ, I don't go for
1000Hz and use 100 or 250Hz and voluntary preemption.
Regarding the userland:
Love choices, hence nothing but Gentoo + KDE4. Multilib. Some relevant
info here:

emerge --info
Portage (default/linux/amd64/10.0/desktop, gcc-4.5.1,
glibc-2.11.2-r0, 2.6.36 x86_64)
System uname: Linux-2.6.36-x86_64-Intel-R-_Core-TM-2_Duo_CPU_T9400_@_2.53GHz-with-gentoo-1.12.13
Timestamp of tree: Tue, 26 Oct 2010 10:30:01 +0000
app-shells/bash:     4.1_p7
dev-java/java-config: 2.1.11
dev-lang/python:     2.5.4-r4, 2.6.5-r3, 3.1.2-r4
dev-util/cmake:      2.8.1-r2
sys-apps/baselayout: 1.12.13
sys-apps/sandbox:    2.3-r1
sys-devel/autoconf:  2.13, 2.65-r1
sys-devel/automake:  1.7.9-r1, 1.8.5-r4, 1.9.6-r3, 1.10.3, 1.11.1
sys-devel/binutils:  2.20.1-r1
sys-devel/gcc:       4.5.1
sys-devel/gcc-config: 1.4.1
sys-devel/libtool:   2.2.10
sys-devel/make:      3.81-r2
CFLAGS="-O2 -pipe -march=native"
CONFIG_PROTECT="/etc /usr/share/X11/xkb /usr/share/config /var/lib/hsqldb"
CONFIG_PROTECT_MASK="/etc/ca-certificates.conf /etc/env.d
/etc/env.d/java/ /etc/fonts/fonts.conf /etc/gconf
/etc/php/apache2-php5/ext-active/ /etc/php/cgi-php5/ext-active/
/etc/php/cli-php5/ext-active/ /etc/revdep-rebuild /etc/sandbox.d
CXXFLAGS="-O2 -pipe -march=native"

Now, I know, Ingo said he wants : "good bugreports and reproducing
testcases" and my testcase is very real life and rather replicates my
typical use of computer these days:

- VirtualBox running XP only to look at some 2007 ppts ( the Ooo3
doens't cut it )
- JuK ( or VLC ) KDE's music player - some music in the background
- Chromium browser, with bunch of tabs with J2EE/J2SE javadocs, eats
out some significant swap space
- bash terminals
- ktorrent
- PDFs opened in okular, Adobe reader
- sync'ing portage tree & emerging new ebuilds
- Netbeans, Eclipse, apache, vsftd, sshd, tomcat and the whole 9 yards.

How do I notice slowdowns ? The JuK lags so badly that it can't play
any music, VBox running XP usually trashes the disk, the mouse pointer
freezes, kwin effects freeze for few seconds.
How can I make it much worse ? I can try & run disk clean up under XP
that is running in VBox, with folder compression. On top of it if I
start copying big files in linux ( 700MB avis, etc ), GUI effects
freeze,  mouse pointer freezes for few seconds sometimes.

And this is on 2.6.36 that is supposed to cure these "features". From
this perspective, 2.6.36 is no better than any previous stable kernel
I've tried.
Two threads copying the same big file ( 700Mb avi ) to different
folders, PLUS couple of "dd if=/dev/zero of=test.10g bs=1M
count=10000" lead to the same freezing sound/pointer/WM effects.

best, Aidar
PS: and yes, I do follow the problem here :
This is a monumental failure for kernel development project and FLOSS
in general.
Poor management, no leadership/championship, no responsibility,
neglect. It all shows why you can't rely on community driven project
or projects. Fact. If you deny this, you are not truthful with
yourself. This might be in one of the Harvard Business Case studies.
Just think about what would happen if this gets embedded into some
kind of pacesetter...

PS: I am a huge adept of FLOSS. I really think that RMS, GNU, Linus
and many others got it right. But it is a shame to have this "feature"
for so long with so many people affected and yet, neglected.
[unhandled content-type:image/png][unhandled content-type:application/octet-stream][unhandled content-type:application/x-wine-extension-ini]
 \ /
  Last update: 2010-10-26 15:03    [W:0.109 / U:4.420 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site