2.6.8-rc2-nitro3 "Shaken, not stirred"

Message

seppe · Post by **seppe** » Tue Jul 27, 2004 2:32 pm

Rainmaker > did you changed PORTAGE_NICENESS in /etc/make.conf? Because the Staircase scheduler of this kernel is very sensitive to nice values. The goal of this scheduler is to handle nice values like they should (nice value 0 gets 20 times less time to do something then nice value 20). Read CK's notes at http://members.optusnet.com.au/ckolivas/kernel/ and his annoucement at http://kerneltrap.org/node/view/2744#comment-7871

blaster999 · Post by **blaster999** » Tue Jul 27, 2004 4:33 pm

Everything works, except for bootsplash

How to make it work with vesafb-tng (I didn't manage to cleanly revert it

)? What kernel parameters should I pass?

scaba · Post by **scaba** » Tue Jul 27, 2004 4:54 pm

on your kernel line put something like

Code: Select all

video=vesafb:ywrap,mtrr,1024x768-16@60

can't get it to work, though

blaster999 · Post by **blaster999** » Tue Jul 27, 2004 5:11 pm

Thanks for the tip! However the results are even stranger: at first it complains that no video mode has been set, after I press the spacebar it starts in 80x25 mode, after some seconds it goes to 1024x768 framebuffer (no fb image though) and only at the end of booting the background shows up. Is there a way to revert vesafb-tng without breaking the kernel (last time I tried there was a kernel panic

)?

Rainmaker · Post by **Rainmaker** » Tue Jul 27, 2004 5:14 pm

Nope, not using PORTAGE_NICENESS

I am renicing reiser4 on bootup, as recommended by love-sources though... Might try without it.

GentooBox · Post by **GentooBox** » Tue Jul 27, 2004 7:16 pm

PickledOnion wrote:
GentooBox wrote:seppe - I dont know if you are doing it already, but it would be a good idea if you tested (if it compiles) the kernel before releasing it.

You can test it by:
Code: Select all
make allyesconfig
make
You would rarely get a heavily patched kernel to compile that way.

Take the reiser4 patch for example - if you choose all the options it will not compile, at least that was the case a few weeks ago.

Also, how does you test it for single processor support and so on - choose make allyesconfig, but then go and uncheck some options?

You can't get it to work for all possible combinations and an allyes just takes you on an endless journey trying to find the conflicts.

Out of interest - have you tried an allyes on a vanilla 2.6.7 straight from kernel.org? It doesn't work

Why release a kernel that dont work ?

You dont need to run the kernel you test (make allyesconfig).
if something fails, then find out why it fails, and fix it.
then try again. - In this nitro kernel version, the only thing that dont compile is reiser4, lirc and vesa on an AMD64.

I reversed vesa, reiserfs4 and unchecked lirc - and it compiles fine.

being a kernel maintainer is a lot more than just patching a kernel.

Pink · Post by **Pink** » Wed Jul 28, 2004 7:48 am

You misunderstood my point - doing an allyesconfig will make no difference as it will work for some people and not for others.

Actually you proved my point rather nicely (thankyou

) by showing that if you reversed some patches it compiled fine (presumably by doing an allyesconfig otherwise is was a pointless post) and yet on other computers it is running just dandy.

As to your question of 'Why release a kernel that dont work ?'. Um, who siad the vanilla 2.6.7 doesn't work? I think you missed my point (again) as I said doing an allyes doesn't work on that kernel on my machine, thus showing that it makes no difference doing an allyes as it works for some and not others, I used the vanilla kernel as an example as it is an obvously working and well constructed kernel. A fairly simple and understandable point I thought.

As to your last point - I assume that is a little dig at me - please could you point to the post (in any thread) where I even suggest I am a kernel maintainer - (I'm not, I've got no programming expereince and I have never claimed or even suggested to be one - so, again, um, what was your point?)

MrApples · Post by **MrApples** » Wed Jul 28, 2004 8:07 am

i actually think he was referring the seppe, which is not to say it was warranted

joki · Post by **joki** » Wed Jul 28, 2004 9:10 am

nice kernel!
compiled cleanly and works smoothly

big thx

seppe · Post by **seppe** » Wed Jul 28, 2004 10:51 am

@GentooBox: I tested this kernel on my machine, and it worked perfectly. Vesafb-tng worked, lirc worked, win4lin worked, the ck patches worked, etc ..

The only thing I hadn't tested was reiser4, simply because I don't have a reiser4 partition .. I'd like to wait to the final version instead of risking my precious Gentoo box now. But I did compiled reiser4 support in my kernel, and it compiled cleanly here without any problems.
Bootsplash didn't worked perfectly, but I thought that was due to etc-update who overwrote my bootsplash conf file the last time I updated my system.

BUT everything compiled cleanly here at my pentium 3 with 800Mhz, and I had no problems with running it. Too bad I don't own a AMD64, then I would have a killer system and maybe I got the same errors as you and then I wouldn't have released this kernel.

My point: how can I know that the compilation fails on other architecture like the AMD 64? I don't know, but I think I can't

If it compiles cleanly and runs without problems on the architecture(s) *I* own, then it looks stable for me.

Btw, read the warning message when you emerge nitro-sources, it says that this is an experimental patch set etc ..

GentooBox · Post by **GentooBox** » Wed Jul 28, 2004 12:15 pm

seppe wrote:@GentooBox: I tested this kernel on my machine, and it worked perfectly. Vesafb-tng worked, lirc worked, win4lin worked, the ck patches worked, etc ..

The only thing I hadn't tested was reiser4, simply because I don't have a reiser4 partition .. I'd like to wait to the final version instead of risking my precious Gentoo box now. But I did compiled reiser4 support in my kernel, and it compiled cleanly here without any problems.
Bootsplash didn't worked perfectly, but I thought that was due to etc-update who overwrote my bootsplash conf file the last time I updated my system.

BUT everything compiled cleanly here at my pentium 3 with 800Mhz, and I had no problems with running it. Too bad I don't own a AMD64, then I would have a killer system and maybe I got the same errors as you and then I wouldn't have released this kernel.

My point: how can I know that the compilation fails on other architecture like the AMD 64? I don't know, but I think I can't If it compiles cleanly and runs without problems on the architecture(s) *I* own, then it looks stable for me.

Btw, read the warning message when you emerge nitro-sources, it says that this is an experimental patch set etc ..

Oh, i am very sorry if i assaulted your kernel, that was not planned.

I was only suggesting that it would be an idea to test the kernel to see if it even compiled.
and since its such a heavy patched kernel, the best thing you can do is a "make allyesconfig"

i can see that you tested the kernel, good job.

Its better to release a kernel that compile on a x86 cpu than releasing a kernel that dont

I am running your kernel right now, and its faster than my normal kernel from gentoo-dev-sources.

I dont know why everything expermental just have to fail on AMD64

- I've got used to it.

seppe · Post by **seppe** » Wed Jul 28, 2004 12:28 pm

Oh, i am very sorry if i assaulted your kernel, that was not planned

Hey, you weren't assaulting it, and I said suggestions and comments were welcome

The AMD64 architecture is just different to a 'normal' x86 architecture, and compatibility looks still like a problem from time to time.

I try to test the kernel as much as possible, but as I said before .. I can't know how it compiles and runs on a different architecture than I have

But thanks to the 'make allyesconfig' tip, maybe I'll try that. Although that's actually not a good idea .. why would I need ALL possible drivers and options? I just need to test the patches I provide in this patchset, not the things the base kernel (2.6.8-rc2) provides .. that is not my job, but the official 2.6 kernel maintainer(s)

GentooBox · Post by **GentooBox** » Wed Jul 28, 2004 12:41 pm

seppe wrote:
Oh, i am very sorry if i assaulted your kernel, that was not planned
Hey, you weren't assaulting it, and I said suggestions and comments were welcome

The AMD64 architecture is just different to a 'normal' x86 architecture, and compatibility looks still like a problem from time to time.

I try to test the kernel as much as possible, but as I said before .. I can't know how it compiles and runs on a different architecture than I have

But thanks to the 'make allyesconfig' tip, maybe I'll try that. Although that's actually not a good idea .. why would I need ALL possible drivers and options? I just need to test the patches I provide in this patchset, not the things the base kernel (2.6.8-rc2) provides .. that is not my job, but the official 2.6 kernel maintainer(s)

Is it a bad idea to be sure ?

Well, unless you know excatly what have been patched you dont need allyesconfig but only check the options that have been patched.

scaba · Post by **scaba** » Wed Jul 28, 2004 1:01 pm

seppe wrote:The new nitro-sources patchset is here!

2.6.8-rc2-nitro3 working fine here. thanks very much for these enhanced ck-sources. looking forward to the next one

blaster999 wrote:Thanks for the tip! However the results are even stranger: at first it complains that no video mode has been set, after I press the spacebar it starts in 80x25 mode, after some seconds it goes to 1024x768 framebuffer (no fb image though) and only at the end of booting the background shows up. Is there a way to revert vesafb-tng without breaking the kernel (last time I tried there was a kernel panic )?

i don't know, sorry. i'm just leaving bootsplash out at the moment. framebuffer works fine w/out bootsplash for me.

Realmaker · Post by **Realmaker** » Wed Jul 28, 2004 7:04 pm

Since i rebooted with this patch-set, sim city 4 isn't working fluently anymore under cedega

Pain · Post by **Pain** » Wed Jul 28, 2004 10:27 pm

I tried this kernel and it toasted my system

Somehow my kernel tree became corrupt so I has to fsck and rebuild it. After that I found all my system files were no longer installed. emerge -s whatever reported no glibc no xorg no nothing and nothing would emerge, had to use the live cd. All part of the fun of new kernels. System going again but despite the trouble won't stop me trying new stuff.

. It was almost fun

Pain.

FirechilD · Post by **FirechilD** » Wed Jul 28, 2004 11:28 pm

- had a soundproblem witch ut2k4 (im using alsa) -> it was constantly chopped into pieces
- firefox and other apps crashed while their were in use

switched back to the love-sources and everthing is fine ^ ^

system specs:

amd 2800+ barton
asus a7n8x deluxe
audigy 2 zs

scoobydu · Post by **scoobydu** » Fri Jul 30, 2004 12:12 pm

Nice patchset

Tried your latest nitro4 and get the same errors as nitro3;

The fbmem error ... (seems to be amd64 specific)

uhci hangs as well, which is really annoying, as my keyboard is attached to it!?

Not sure it the uhci is a 2.6.8 thing or a nitro specific problem ..

Oh well, back to 2.6.7 for now

hotplainrice · Post by **hotplainrice** » Fri Jul 30, 2004 12:53 pm

Is it possible to patch this kernel with swsusp ?

Sorry cos i dont do programming, i dont get the concept well.

scaba · Post by **scaba** » Fri Jul 30, 2004 2:50 pm

scoobydu wrote:Nice patchset

Tried your latest nitro4 and get the same errors as nitro3; [...]

did i somehow miss the nitro4 thread?

seppe · Post by **seppe** » Fri Jul 30, 2004 2:53 pm

The nitro4 thread is coming

Elementl · Post by **Elementl** » Sun Aug 22, 2004 4:32 am

I realize this may be a newbie question -
Why dont I show the nitro sources when i do an emerge search?
Am I missing something?

d0nju4n · Post by **d0nju4n** » Sun Aug 22, 2004 6:29 am

Elementl wrote:I realize this may be a newbie question -
Why dont I show the nitro sources when i do an emerge search?
Am I missing something?

Elementl: There is a nice set of instructions about using portage overlay to install these sources here.

It was in the first post of this thread...

blk_jack · Post by **blk_jack** » Sun Aug 22, 2004 10:34 am

Maybe you guys should write a detailed HOWTO on what kind of NICE levels people should be having. There are a *lot* (read: most) of kernels that suggest you change the nice levels of certain applications (X, xmms, etc) to -10 or -15 even. For instance, I have the verynice daemon setup to change the nice levels of some of my applications to negative. This works VERY well for love-sources and any other kernels that DON'T use the staircase scheduler.

Here's my problem, though. I recently installed nitro-sources to check it out and the kernel went nuts with all my default nice values. Applications (galeon, rhythmbox, etc) were choppy and slow as hell. I could barely play mp3s without having them skip when I switched windows and all my games in winex were slower and less fluent.

I know it's not your 'job' to detail these kinds of things, but a little heads up or a general reference would be nice. I bet you lose more than a few people trying out nitro-sources for the first time just because you don't give a simple link to information on nice levels and the difference between schedulers in your kernel posts.

Food for thought.

2.6.8-rc2-nitro3 "Shaken, not stirred"

Newb question

Re: Newb question