OFFICIAL Zen Kernel Discussion - zen-kernel.org

Kollin · Posted: Sat Oct 31, 2009 3:37 pm Post subject:

today i updated to 31-zen5 ant this happend:

cheater1034 · Veteran Joined: 09 Sep 2004 Posts: 1558

Ok guys, here's a new proposition for the graphically inclined.

Here is my logo (the conrad dog) - I'd like someone to turn it into a buddha/meditating/zen style logo (anyway possible)

If someone could do it, it would be great!
here's the .svg file:
http://omploader.org/vMm55Mw
_________________
IRC!: #zen-sources on irc.rizon.net
zen-kernel.org
--
Lost in android development land.

RealNC · Tux's lil' helper Joined: 13 Apr 2008 Posts: 148

Wouldn't a penguin be more logical than a dog?

cheater1034 · Veteran Joined: 09 Sep 2004 Posts: 1558

cheater1034 · Veteran Joined: 09 Sep 2004 Posts: 1558

here's an idea i've started to tender up

http://omploader.org/vMm84YQ
_________________
IRC!: #zen-sources on irc.rizon.net
zen-kernel.org
--
Lost in android development land.

rahulthewall · Veteran Joined: 01 Nov 2007 Posts: 1264 Location: Zürich

cheater1034 · Veteran Joined: 09 Sep 2004 Posts: 1558

tranquilcool · Veteran Joined: 25 Mar 2005 Posts: 1179

lockups in 32-zen6-r1
don't know why.

EDIT: solved. played too long and too hard with vm.
_________________
this is a strange strange world.

wired_ · Posted: Wed Nov 04, 2009 2:52 pm Post subject:

Hey guys

Hwoarang and I decided to move the rc* and 9999 ebuilds in tree and get rid of the overlay.

They'll stay masked, however you won't have to add an overlay for just a couple of ebuilds anymore :wink:

I just committed the ebuilds in cvs, they should be available in rsync mirrors in a few hours.

Hwoarang will remove the overlay from layman/github in ~10 days.

_________________
Alex Alexander
wired @ gentoo
stupid _ in forum nick because someone else has wired

dodo1122 · Posted: Wed Nov 04, 2009 4:59 pm Post subject:

Great

Thanks.
tranquilcool, so the lockups are not kernels fault?
_________________
#zen-sources on irc.rizon.net

kernelOfTruth · Posted: Wed Nov 04, 2009 6:24 pm Post subject: so

most probably those are not really "hardlocks" but a total interruption of all pci-e (pci express) related traffic due to heavy flushing to disc (even more if barrier-support is enabled)

soundcard output, ethernet card throughput come to an complete halt during these phases for me

the kernel-developers better improve this starvation

anyone has encountered this too ?

unfortunately this isn't fixed by BFS and BFQ - hence it's not a i/o and/or cpu-scheduler problem :idea:

_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004

tranquilcool · Veteran Joined: 25 Mar 2005 Posts: 1179

kernelOfTruth · Posted: Wed Nov 04, 2009 7:38 pm Post subject:

so you're seeing the same problem like me, tranquilcool ?

I'm glad I'm not the only one :roll:

the main problem seems to be that with pci-e those pci-lanes (those separate lines are called lanes, right ?) can't operate autonomously so if one has heavy
work/traffic to do it completely hinders the other ones from their work

IMO they (the devs) (unintentionally) are creating more and more bottlenecks and it's getting more and more difficult to track those down in the future so why not halt development in some parts, e.g. pci-e or libata / VFS subsystem during one cycle and fully concentrate in reproducing, finding, analyzing, understanding and fixing these kind of bugs ?

I'm not sure but I believe that I didn't see this at all with 32bit or it at least was much less manifest than now besides that from my remembrance there was no interruption during flushing to disc with 2.6.24 so it partly still might be related to cpu-scheduling and/or the changes that went into CFS, VFS since then
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004

dodo1122 · Posted: Wed Nov 04, 2009 8:38 pm Post subject:

Hmm, i think i'm getting this too tbh. I was doing a backup this morning and i couldn't even dmesg when it was copying files... Hopefully it'll get tracked down and fixed before .32 gets released.
_________________
#zen-sources on irc.rizon.net

tranquilcool · Veteran Joined: 25 Mar 2005 Posts: 1179

dodo1122 · Posted: Thu Nov 05, 2009 11:03 am Post subject:

It's not that btrfs is broken, it was never fixed... You just need a bigger partition, so it never gets anywhere close to 100% space used, else problems are going to show. They're working on it tho.
_________________
#zen-sources on irc.rizon.net

kernelOfTruth · Posted: Thu Nov 05, 2009 1:38 pm Post subject:

could someone please post those symptoms over at lkml if it hasn't already ? (including the halt of sound, ethernet throughput, and jerkiness of X)

that way the devs and other testers can reproduce and tackle those problems by the horns and eliminate it ASAP :mrgreen:

I would collect some data & post it there like in the past - unfortunately I currently am 200% busy so I have no time and it wouldn't likely make it there before release of 2.6.32

(it's about several important exams so I'm not making this up :wink:

)

thanks

_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004

Mr. Tao · Tux's lil' helper Joined: 20 Jul 2007 Posts: 147

Recent problems with Reiser4 corruption on 2.6.31 I reported might have been related to hardware (aka stupid sata cable). I moved to 2.6.31-zen6 and run it under reasonable load for a few days without problem.
There still is a problem with SLOB. For validation I compiled 2.6.31-zen6 with SLOB – result was complete lack of responsiveness. After switching back to SLQB everything went back to normal. config
Is ramzswap aka compcache working for anyone in recent zens? Since 2.6.29 I'm receiving:
swapon /dev/ramzswap0 wrote:

swapon: /dev/ramzswap0: read swap header failed: Invalid argument

kernelOfTruth · Posted: Thu Nov 05, 2009 7:27 pm Post subject:

Waninkoko · Guru Joined: 13 May 2005 Posts: 549

I have ported Classic RCU and Preempt RCU to .32

Use it except if you have >=4096 CPUs.

Waninkoko · Guru Joined: 13 May 2005 Posts: 549

Erm, forget what I said for now... Thanks to Preempt RCU dodo's computer is borked

EDIT: Well, I just tried both RCU subsystems on PPC (non SMP) and works perfect.

kernelOfTruth · Posted: Fri Nov 06, 2009 1:02 am Post subject:

@tranquilcool:

please try sqlb (slab allocator) instead of slab or slub

I just switched back to it after a long time and it seems to completely fix the starvation / mutual blockage

I now can transfer lots of small files via rsync from one partition to another and run a ethernet speed test during that with no visible impairment

this hasn't been possible before ...

so it's a BUG in slab, slub, slob (?) in comparison to sqlb (you can't call that a regression can you ? :roll:

)

to be more precise these symptoms are caused by a collection of bugs in CFS, CFQ (and more mildly the other i/o schedulers too) and the currently available mainline slab-allocators

so could anyone with the problems mentioned in the following amd64-thread:
https://forums.gentoo.org/viewtopic-t-793263.html?sid=1b47ecb01eea573b75b980d28bc5bc0b confirm that their problems are fixed by using BFS, BFQ and SQLB ?

thanks
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004

tranquilcool · Veteran Joined: 25 Mar 2005 Posts: 1179

cheater1034 · Veteran Joined: 09 Sep 2004 Posts: 1558

tranquilcool · Veteran Joined: 25 Mar 2005 Posts: 1179

kinda baffles why i have dhcpcd lockup during boot
with 2.6.32-rc6-zen1. it just hangs.

any help? thanks.
_________________
this is a strange strange world.