Updated kernel26ck

Izuil · 2007-05-13 22:56:54

Well, solved the problem so I'm happy now

danym · 2007-05-14 13:15:40

I've recently tried your last kernel26ck almost with success. By almost success, i mean that everything seemed to work except networkmanager was unable to connect to wireless. With kernel26, networkmanager is working well.

lucke · 2007-05-14 13:39:53

There has been a report of networkmanager misbehaviour with sd.

Other than that, ck2 is out with swap prefetch fixes.

Last edited by lucke (2007-05-14 13:44:45)

swiergot · 2007-05-14 19:14:15

B wrote:

I don't think the maintainer is gonna provide all add-on drivers you can think of .

Not all, only those requested

lucke wrote:

There has been a report of networkmanager misbehaviour with sd.

Yep, no fix available so far.

lucke wrote:

Other than that, ck2 is out with swap prefetch fixes.

Updated.

broch · 2007-05-18 18:59:35

SD - replaces staircase
CFS - replaces Ingo's scheduler
enabling in the kernel does really nothing. You have to set nice for X (SD limit is -1, CFS limit is -19, best/stable at -10)

to take advantage of SD you also need to modify /etc/sysctl.conf and add
kernel.rr_interval = 400

CFS does not require editing sysctl.conf

both schedulers are designed for desktop only

personally I found CFS better, SD caused sound stuttering and jumpy mouse under heavy load. However both schedulers are at early stage of development so best is to check which works better for the specific configuration

Last edited by broch (2007-05-19 12:47:40)

henryouly · 2007-05-19 07:18:42

hi swiergot, thanks for your effort, do you have time to build ndiswrapper module in against your ck kernel?

swiergot · 2007-05-19 08:26:18

Sure, done.

lucke · 2007-05-19 09:57:22

broch, you've got (almost) all your facts wrong.

ekerazha · 2007-05-19 11:41:29

broch wrote:

best/stable at -10

Why should -10 be better than -19?

both schedulers are designed for desktop only

Absolutely not:

From: Ingo Molnar [email blocked]
To: linux-kernel
Subject: [Announce] [patch] Modular Scheduler Core and Completely Fair Scheduler [CFS]
Date: Fri, 13 Apr 2007 22:21:00 +0200
[announce] [patch] Modular Scheduler Core and Completely Fair Scheduler [CFS]
i'm pleased to announce the first release of the "Modular Scheduler Core
and Completely Fair Scheduler [CFS]" patchset:
http://redhat.com/~mingo/cfs-scheduler/ … +cfs.patch
This project is a complete rewrite of the Linux task scheduler. My goal
is to address various feature requests and to fix deficiencies in the
vanilla scheduler that were suggested/found in the past few years, both
for desktop scheduling and for server scheduling workloads.

Last edited by ekerazha (2007-05-19 11:42:07)

mykey · 2007-05-19 18:17:16

Is there any chance to get the reiser4 patch into this kernel since beyond is gone?

swiergot · 2007-05-19 20:08:37

I'm affraid no. Basically kernel26ck is the official kernel26 plus the ck patchset. So as long as reiser4 doesn't make it into kernel26, it will not be included in kernel26ck.

I don't know if inclusion of reiser4 to Arch's kernel was already discussed. I'm not familiar with kernel internals but it seems to me that the very few changes that the reiser4 patch does to the kernel are not harmful. Surely the Arch kernel has many patches that modify the vanilla kernel much more deeply. I'll search the forum and the mailing list and see what can be done.

broch · 2007-05-21 13:32:47

both for desktop scheduling and for server scheduling workloads.

in theory, yes
have you seen any benchmarks proving this in real world tests? No? Then cfs is safe for desktop, unless it is proven otherwise. Including the fact that this is still early testing phase so schedulers (both SD and CFS) thend to crash.

Why should -10 be better than -19?

see benchmarks on lkml. Maybe in theory -19 looks better, but this is not the case in the benchmarks + lower stability (also lkml)

broch, you've got (almost) all your facts wrong.

which ones?

Last edited by broch (2007-05-21 13:33:37)

mykey · 2007-05-22 07:41:08

@ swiergot

Thank you for looking into this - kernel26mm (which seems to be the only one with the reiser4 patch) just doesn't run nicely over here and a professionally tweaked kernel with the patch would just be really appreciated

swiergot · 2007-05-22 10:35:20

Unfortunatelly reiser4 won't be included in the official kernel package (see here). I might however create a separate kernel package (or packages) with reiser4.

mykey · 2007-05-23 07:35:15

The opinions in the mailing list reflect the sad discussion about including reiser4 in vanilla - yes hans reiser seems to be one of a kind and yes people simply seem to not understand what he is coding - IMHO.

I really like reiser4 - patched my own kernels also before I started using arch and was suprised actually that the patch was included in beyond (which I honestly consider one of the best kernels at this point in time thanks to iphitus). I had my issues with the patched version of grub too which I was not able to solve - but then I don't understand why people stick to a piece of bootloader that is not even developed any longer - so I am using lilo which works flawlessly as long as I use it in an ext2 boot partition.

Since I am not a coder I simply would appreciate a kernel package that includes the patch somewhere - it does not have to be 'official' as long as it exsists.

Thanks again for your approach to the problem.

robbel · 2007-05-24 18:01:29

Hi there,

would it be possible for you to create a package of the fglrx drivers that work with your latest ck patches?

Thanks!
Philipp

tomk · 2007-05-24 20:32:20

Hi ck people

Apologies for my unavoidable absence, and thanks to swiergot for keeping the flag flying. I'm building 2.6.21-ck2 now, so expect it in [extra] tomorrow, along with the usual assortment of external modules.

Re ck's future, I'll keep producing packages as long as Con keeps maintaining the patchset.

3nd3r · 2007-05-25 05:44:31

I am using this kernel and it runs great Thanks for this

swiergot · 2007-05-25 10:16:45

Welcome back, Tom!

PS In the meantime, fglrx has been added to the repo.

skottish · 2007-05-29 03:09:41

Bumping to let the community know that swiergot's package (built from the ABS tree) works well on 64 bit. I've had it up and running for days without any noticeable problems.

aRcHaTe · 2007-05-29 13:29:03

can it you build it with fbsplash?

cpu · 2007-05-30 10:47:18

Would be nice to have 2.6.21-ck2 in x86_64 repo...

aRcHaTe · 2007-05-30 23:34:02

did anybody could compile ck kernel with fbsplash patch? mine gives error a PKGBUILD would be apreciated

ekerazha · 2007-06-01 19:04:10

broch wrote:

in theory, yes
have you seen any benchmarks proving this in real world tests? No? Then cfs is safe for desktop, unless it is proven otherwise. Including the fact that this is still early testing phase so schedulers (both SD and CFS) thend to crash.

Have you seen any benchmark contrary to this? No? Have you seen any benchmark proving that CFS is "safe for desktop"? Is there a "benchmark" for "safety"? Does "good-for-desktops" necessarily exclude "good-for-servers"?

Ingo Molnar, the CFS developer, said "both for desktop scheduling and for server scheduling workloads": do you know anything about CFS that Ingo doesn't know?

see benchmarks on lkml. Maybe in theory -19 looks better, but this is not the case in the benchmarks + lower stability (also lkml)

What? Where? When?

mic64 · 2007-06-02 11:43:57

Hi

I have a problem with the ck kernel, but also with the standard one.
I can remember the days with older kernels, where I could do a lot of stuff at the same time.
example: encoding, rar some stuff and still do the normal stuff like surf the web or browse files.

Now it is almost impossible to do this.
As soon I start rar or mencoder the system behaves like crap.
Opening a konqueror takes minutes, watching a clip stutters like hell.

My System C2D E6300 @ 2,8 GhZ
2 Gb RAM

What is this?

Last edited by mic64 (2007-06-02 11:47:18)

Arch Linux

#26 2007-05-13 22:56:54

Re: Updated kernel26ck

#27 2007-05-14 13:15:40

Re: Updated kernel26ck

#28 2007-05-14 13:39:53

Re: Updated kernel26ck

#29 2007-05-14 19:14:15

Re: Updated kernel26ck

#30 2007-05-18 18:59:35

Re: Updated kernel26ck

#31 2007-05-19 07:18:42

Re: Updated kernel26ck

#32 2007-05-19 08:26:18

Re: Updated kernel26ck

#33 2007-05-19 09:57:22

Re: Updated kernel26ck

#34 2007-05-19 11:41:29

Re: Updated kernel26ck

#35 2007-05-19 18:17:16

Re: Updated kernel26ck

#36 2007-05-19 20:08:37

Re: Updated kernel26ck

#37 2007-05-21 13:32:47

Re: Updated kernel26ck

#38 2007-05-22 07:41:08

Re: Updated kernel26ck

#39 2007-05-22 10:35:20

Re: Updated kernel26ck

#40 2007-05-23 07:35:15

Re: Updated kernel26ck

#41 2007-05-24 18:01:29

Re: Updated kernel26ck

#42 2007-05-24 20:32:20

Re: Updated kernel26ck

#43 2007-05-25 05:44:31

Re: Updated kernel26ck

#44 2007-05-25 10:16:45

Re: Updated kernel26ck

#45 2007-05-29 03:09:41

Re: Updated kernel26ck

#46 2007-05-29 13:29:03

Re: Updated kernel26ck

#47 2007-05-30 10:47:18

Re: Updated kernel26ck

#48 2007-05-30 23:34:02

Re: Updated kernel26ck

#49 2007-06-01 19:04:10

Re: Updated kernel26ck

#50 2007-06-02 11:43:57

Re: Updated kernel26ck

Board footer