Helium Sources (kernel patchset)

Message

Phlogiston · Post by **Phlogiston** » Mon Jun 11, 2007 9:36 am

hirakendu wrote:Nope, frankly I have no clue about undervolting and fan control. But tell me something - so without this thinkpad fan control, does your notebook cpu fan keep whining at full speed all the time? Because on my hp dv5000, I dont have any fancy things in the kernel (except the usual acpi fan support), and things seem fine - its quiet usually and whines only while doing intensive tasks.

The main point is that I can turn the fan off. And when I'm on battery and cpu is scaled down and undervolting active that the normal case

But you included the linux phc patch?

hirakendu · Post by **hirakendu** » Mon Jun 11, 2007 10:32 am

First things first - Ok, I now get it. Helium1 doesnt have phc, only thinkpad extras. Helium2 does have it. I have a feeling you are using helium(1).

Now the fun part - what on earth - you can turn off the fan

. Lol, then I must give it a try too. Although I am too afraid of burning out my cpu on my notebook

. (Especially given the fact that not so long ago, I had burnt up my good old athlon-xp 2000+. It used to run at 73 degrees celsius while playing doom3

.)

Phlogiston · Post by **Phlogiston** » Mon Jun 11, 2007 10:39 am

hirakendu wrote:First things first - Ok, I now get it. Helium1 doesnt have phc, only thinkpad extras. Helium2 does have it. I have a feeling you are using helium(1).

Now the fun part - what on earth - you can turn off the fan . Lol, then I must give it a try too. Although I am too afraid of burning out my cpu on my notebook . (Especially given the fact that not so long ago, I had burnt up my good old athlon-xp 2000+. It used to run at 73 degrees celsius while playing doom3 .)

Hehe maybe you are right about helium1/2.

The thing is that together with undervolting, the temperature is quite low on my laptop. Usually about 40-50 degrees. And the script I use just turns off the fan if the temp. is low.

Whats the difference between kamikaze and helium in general?

hirakendu · Post by **hirakendu** » Mon Jun 11, 2007 10:56 am

There is absolutely no difference

. Except for the 4k-stacks thing. In fact if you don't use ndiswrapper or reiser4 or such (which apparently need 8k stacks, infact ndiswrapper wants 16k

), you can (should/must/may/will) use kamikaze-sources itself.

(A couple of minor things. I mount iso images frequently and apparently 2.6.22 doesnt create /dev/loop* entries by default and udev is supposed to do it. So there was this patch in mm-tree to re-enable that. And the other thing is a patch (from myriad forum searches long time back :p : http://www.linuxquestions.org/questions ... p?t=489805) for the conexant audio codec on my notebook on which alsa still needs to improvise on.)

Anyway, the kernel has been working fine for me. So, what better way to share the pleasure

. (Although seems that it hasnt really been a 'pleasurable' experience for you

. )

Phlogiston · Post by **Phlogiston** » Tue Jun 12, 2007 7:02 am

Hmm now I have that problem again:

Code: Select all

request_module: runaway loop modprobe binfmt-0000

It did no occur on helium1, so maybe we can finally find out which patch causes the error. I had the same problem with kamikaze sources and ragnarok1.

hirakendu · Post by **hirakendu** » Tue Jun 12, 2007 7:31 am

Btw, what does that module do ? I modprobe it and it just quietly loads. Can't even see it in lsmod or dmesg. Incidentally, I have been running this kernel for over 3-4 days without rebooting on both my workstation and notebook. Seems to be rock stable

.

Phlogiston · Post by **Phlogiston** » Tue Jun 12, 2007 7:44 am

hirakendu wrote:Btw, what does that module do ? I modprobe it and it just quietly loads. Can't even see it in lsmod or dmesg. Incidentally, I have been running this kernel for over 3-4 days without rebooting on both my workstation and notebook. Seems to be rock stable .

I don't know about if it is really a module. But what are the differences between helium1 and 2?

hirakendu · Post by **hirakendu** » Tue Jun 12, 2007 8:15 am

Frankly, I myself forgot

. Helium1 was pretty simple - just the usual genpatches, suspend2, unionfs, reiser4, powertop, thinkpad, udf, and ipw3945. There was no phc, ck-patchset, cfs-sched. Those were the only main things I can remember missing.

Phlogiston · Post by **Phlogiston** » Tue Jun 12, 2007 8:30 am

hirakendu wrote:Frankly, I myself forgot . Helium1 was pretty simple - just the usual genpatches, suspend2, unionfs, reiser4, powertop, thinkpad, udf, and ipw3945. There was no phc, ck-patchset, cfs-sched. Those were the only main things I can remember missing.

I think its ck or cfs then...

Phlogiston · Post by **Phlogiston** » Sat Jun 16, 2007 8:20 am

Phlogiston wrote:
hirakendu wrote:Frankly, I myself forgot . Helium1 was pretty simple - just the usual genpatches, suspend2, unionfs, reiser4, powertop, thinkpad, udf, and ipw3945. There was no phc, ck-patchset, cfs-sched. Those were the only main things I can remember missing.
I think its ck or cfs then...

As far as I found out, its an issue in the vanilla kernel. I don't know what to do. I will try to boot without initrd.

martin.k · Post by **martin.k** » Sun Jun 17, 2007 8:10 pm

-ck patchset is dead !!!
http://members.optusnet.com.au/ckolivas/kernel/

hirakendu · Post by **hirakendu** » Tue Jun 26, 2007 6:22 pm

First, my condolences for ck-patchset

. Was an essential ingredient of any performance patch till now

. Then again, the recent kernels (2.6.18+) have been reasonably fast and interactive. Although stability has taken a hit - so that seems to be the major concern (for me :p) though. The good thing is, for now, 2.6.22 (rc's) seem to be really really stable

.

A small change made to the helium patches (no version revision :p). This was regarding vmware-modules unable to compile. Made the changes to include/asm-i386/page.h (syntax stuff) to get vmware compiled. (From the thread : http://www.vmware.com/community/thread. ... eID=677688 )
Also, one has to edit the vmware-modules ebuild to use vmware-any-any-update110 (instead of 109). (Details : Edit /usr/portage/eclass/vmware.eclass and look for the variable ANY_ANY="vmware-any-any-update109" and change it to 110. And do ebuild digest for appropriate ebuilds in app-emulation/vmware-modules and vmware-workstation etc.)

And a small note on ipw3945. Although the ipw3945 v1.2.1 included in the patch is pretty fine, had some troubles connecting with knetworkmanager. Also had issues with iwl3945 and knetworkmanager - seems to reset and take time to reconnect frequently (thats a networkmanager policy to look for better connections frequently ?). (Besides, iwl3945, rather the d80211 system doesn't have the LED functionality as of now?) For now, ipw3945 1.2.0 in the portage seems to work perfectly. If using this, then disable the ipw3945 in kernel. Also might require to edit the ebuild to continue compiling with the in kernel 80211 system without cribbing

.

hirakendu · Post by **hirakendu** » Sun Jul 15, 2007 11:01 am

2.6.22-helium3 = 2.6.22-kamikaze1 - 4k_stacks - hrtimers - acpi_dsdt_initrd :
http://hirakendu.mooo.com/helium-source ... .patch.bz2

hirakendu · Post by **hirakendu** » Tue Jan 22, 2008 7:45 am

Here is another new one : 2.6.24-helium0. It is based on 2.6.24-rc8-git4 and tries to be as close to vanilla as possible with the following features :

reiser4 (from mm)
unionfs
aufs
tuxonice (suspend2)
squashfs+lzma (from zen)
fbcondecor (splash)
linux-phc
latencytop
madwifi
alsa-git (from mm)

Try this at home

Patch : http://hirakendu.mooo.com/helium-source ... .patch.bz2,
Broken-out : http://hirakendu.mooo.com/helium-source ... roken-out/.

To find out if this kernel is right for you, may try out precompiled sources : http://hirakendu.mooo.com/helium-source ... ll.tar.bz2 or http://hirakendu.mooo.com/helium-source ... ed.tar.bz2.

For those looking for just reiser4 on vanilla kernel, may peep into this : http://hirakendu.mooo.com/reiser4-stuff/.

kernelOfTruth · Post by **kernelOfTruth** » Tue Jan 22, 2008 10:34 am

thanks, hirakendu !

I'll give them a test-ride

hirakendu · Post by **hirakendu** » Tue Jan 22, 2008 11:05 am

thanks. hopefully you have a vanilla experience

.

makism · Post by **makism** » Mon Jan 28, 2008 12:41 pm

just booted the .24-hellium1, i`ll stick with it for some time to check it :p

kernelOfTruth · Post by **kernelOfTruth** » Tue Jan 29, 2008 2:00 pm

just booted the .24-hellium1,

WOW! there's already 2.6.24-helium1 ?

just downloading it now

makism wrote: i`ll stick with it for some time to check it :p

++

I need a stable & fast kernel for that laptop ...

luqas · Post by **luqas** » Tue Jan 29, 2008 2:49 pm

Going to give these a spin. Love the minimalist approach of this kernel, plus I want to try out tuxonice.

Just curious as to what the git-alsa patch adds?

hirakendu · Post by **hirakendu** » Tue Jan 29, 2008 3:35 pm

Thanks for using it and appreciating the minimalist approach. (Just banking on the idea that vanilla should be enterprise class, so modify it as less as I can

.)

The git-alsa patch is not really really necessary. Just for explicit support of newer C-Media Oxygen based audio cards (which might be already there in vanilla). (There is always some new audio hardware that needs git-alsa.) Moreover, it looks it doesn't break anything (I hate regressions

). Plus I am not sure if alsa in vanilla already has the strict power saving features.

About the fastness part. Here is my config. On both my notebook (dell vostro 1400) and workstation (athlon64x2 3800+ & nv geforce 7300gs), I run a 100Hz dyn ticks kernel with no pre-empt whatsoever and no fiddling with any features. (Also using 4k stacks and slub, which are bog standard by now

.) Essentially what might be called a very 'server' type kernel. And I am reasonably happy with that. It is very stable. Merrily runs doom3 even after days of uptime. And interactive enough for my purposes. Also good for battery I suppose. There may be something 'faster', but I am very happy with this. I think after years of kernel compiling, every one settles on his/her own magic config

. Now then again, I haven't ever tried it on lesser hardware, both my machines are dual core

.

As a side remark, I have removed latencytop. It requires frame-pointers among other debug stuff to be enabled. I am ok with that, though my CLAGS have -fomit-frame-pointer. But the surprising thing is that even after doing that, when I ran the latencytop application, the list was blank

. If anyone has tried latencytop before, does this need CONFIG_EMBEDDED=y ? (general setup, configure standard kernel features for small systems).

@ KernelOfTruth : The patches in .24-h1 are same as in h0, adding the reiser4-fix-null-pointer-dereference-in-sys_truncate64.patch as you suggested in zen-sources thread. Thanks for that.

praka123 · Post by **praka123** » Fri Mar 28, 2008 1:31 am

@hirakendo:
any updates :0
what about 2.6.24?

hirakendu · Post by **hirakendu** » Fri Mar 28, 2008 4:48 am

Its up-to-date always

. 2.6.24-helium1 is the latest. (Although updated pretty frequently, name is not

.) As mentioned before, includes latest resier4 (from mm, with extras), aufs and unionfs, squashfs 3.3 with lzma, tuxonice (suspend2), fbcondecor (fbsplash), and also atl2 driver (from sabayon) for attansic l2 ethernet in eeepc. It includes 2.6.24.4. So apply the patch to vanilla 2.6.24. Also, as a sidenote, I use only even numbered kernels, no next will be 2.6.26.

hirakendu · Post by **hirakendu** » Tue Jun 03, 2008 7:20 am

Here is a newer version of helium sources patch for 2.6.25 (2.6.25-helium1) :

http://hirakendu.mooo.com/helium-source ... 5-helium1/

Minimalistic as usual (see broken-out folder for individual patches) : reiser4, squashfs+lzma, aufs, unionfs, tuxonice(suspend2), fbcondecor and attansic_l2_ethernet.

A few notes - its actually more-or-less redundant, given zen-sources. There is a 'section-mismatch' problem which happens even in vanilla 2.6.25. To my satisfaction, all extra kernel modules in portage (nvidia-drivers, ati-drivers, vmware-modules, linux-uvc, gspcav1, ndiswrapper, hsfmodem etc) work, except 'slmodem' because of gpl-symbol issue.

Although I usually only go for even versions of kernels, here is 2.6.25 for a change

. Looking forward to 2.6.26

.

hirakendu · Post by **hirakendu** » Mon Jul 14, 2008 8:37 am

There we are : 2.6.26

. And here's another helium-sources 2.6.26-helium0 (patch and precompiled sources) :

http://hirakendu.mooo.com/helium-source ... 6-helium0/

hirakendu · Post by **hirakendu** » Sun Aug 10, 2008 10:08 pm

2.6.26-helium1 :

http://hirakendu.mooo.com/helium-source ... 6-helium1/

changelog : updated to 2.6.26.2. unionfs updated. see broken-out/ folder for details. also may use precompiled sources (w/ or w/o sources) to save time or just to try out.

other notes : by far, 2.6.26.2 is the best kernel in recent times. a note on a new rig at work - has a new gigabyte ep43-ds3l motherboard. (along with e8400 wolfdale cpu and nv gf 8500gt gpu, its a blazing combo. stock 3ghz can be easily overclocked to 3.6ghz.) 2.6.26 had problems with audio codec alc888 and wouldn't load. 2.6.26.2 solves it. meanwhile, as a strange 'workaround', disabling ahci in the bios actually speeds up the hard disks

. (w/ ahci, its 55MB/s and w/o, its a raging 110MB/s.) strangely, earlier kernels like 2.6.24 wouldnt boot with ahci disabled. any case, 2.6.26.2 rocks.

if looking for something close to gentoo-sources rather than vanilla, do use hitchhiker sources (there should be very little difference). and not to mention, for lots of features and bleeding edge, zen or zenmm

.