Slashdot Mirror
Configuring the 2.6 Linux Kernel
An anonymous reader writes "This article is the first in a series by William von Hagen on using the new Linux 2.6 kernel, with a special emphasis on the primary issues in migrating existing drivers, applications, and embedded Linux deployments to a Linux distribution based on the 2.6 kernel. Bill is the author of Linux Filesystems, Hacking the TiVo, SGML for Dummies, Installing Red Hat Linux 7, and is the coauthor of The Definitive Guide to GCC (with Kurt Wall) and The Mac OS X Power Users Guide (with Brian Profitt)." This looks to be a good series for anyone planning to migrate to Linux 2.6, and having done just that myself, I'll attest to wanting more documentation along the way.
Mandrake 10 will be the first major distro use Kernel 2.6. Download the beta here.
Easy to install, just download the ISOs, burn to disk, reboot and the installer will appear.
Make sure to REPORT ALL BUGS, unless you want to see the LG incident again.
Firstly, s/summerize/summarize/, and secondly, the 'funky X configuration interfaces' you talk about are nothing more than GUI applications which have nothing to do with the kernel.
This recent trend in GUIfication of Linux is troubling, and your post illustrates exactly why. It's the same 'logic' that allowed MS to call Windows an 'Operating System' before it really was. It's the same logic that lusers use when they say that they 'can't get into the Microsoft' when they really mean there's an application problem.
Computers are NOT monolithic, they are NOT black boxes. They are boxes of legos, where you can build what you want, when you want it, and leave out the crap. You have control.
Unless you cede responsibility and control to someone else.
I want to delete my account but Slashdot doesn't allow it.
> make menuconfig
;)
I have never compiled my kernel before, and today I jsut compiled the 2.4 version, installed and later compiled 2.6.1 and installed... all without any problems.
Enough rant
I just like the config file method. Does this mean that I can keep my config file (for compiling the kernel) when I upgrade to the next stable? (keeping in mind that I have it customized)
I found this sticky at linuxquestions.org's forums to be most helpful in doing an easy and straightforward 2.6 compile on a slackware system. LinuxQuestions.org
Whats with the default config for the kernel, it's not blank, it's a specific setup that looks like it's for a P4 with a specific network card and chipset, shouldnt the kernel config be bare? this is referring to source downloaded directly from kernel.org,
Is the 2.6 kernel "Usable" yet? By this I mean getting obscure hardware to work such as my USB Midi Interface, and what about proprietary drivers such as Nvidia's, will existing code compiled for 2.4 kernel work? or will you have to recompile stuff (IPTables for example). Im running a gentoo box. I've been toying with updating the kernel to 2.6 and I've been hearing that there are a lot of radical changes to the kernel, the performance enhancements are very exciting to say the least. But what kinds of headaches am I going to have with a real world (used as a desktop as well as a server) system?
Electronic Music Made Using Linux http://soundcloud.com/polyp
I've considered setting up a wiki for the Kernel to give people a place to file what they've learned about it and share with the general public. Anyone think this would help anyone?
I've only recently started using Linux on a day-by-day basis, and after installing Debian unstable I switched directly to 2.6, without ever compiling a 2.4 kernel. That worked without a hitch, so now I'm wondering if the difference is so big. I still have an old Pentium I around the house, and I'm thinking of making this one a firewall/IDS... and so far I'm not sure if 2.6 was a little overkill for that one...
-- Power corrupts, but PowerPoint corrupts absolutely.
As far as I can tell, there is no way to use a KVM with kernel 2.6 and still have mouse wheel support. It works until you switch away and back to Linux, at which point the mouse goes completely nuts and cannot be revived. Back in 2.4, there were two hacks to revive the mouse in this situation (switch VCs, or set the mouse protocol in X to "AUTO") but neither of these work in 2.6. Windows, needless to say, has no problems. You can supposedly pass a psmouse.noext parameter to the kernel at boot time to fix the craziness, but a) this would remove mouse wheel support, and b) I never got it work anyway. If you know a workaround, please post!
Protect your browser with the Force Safe Search add-on
If more people put up articles like this to make it easier to use the popular open source software products it'd speed up rate of adoption a lot. If there was one on writing kernel modules ... ;-)
The problem with building your own kernel is that you will always make it as slim as possible. Then each time you buy a new piece of hardware (USB-mouse, SATA hd, etc) you'll need to build it again (and you've thrown away the original .config of course :). Is there a configure everything (besides what's been specified as built-in) as modules?
This is just a very loosly disguised advert for TimeSys Linux
Nothing any monkey cant work out in about five minutes (and if they cant they should not be cross compiling for embedded devices)
Since most people dont RTFA this isnt a problem, if you are one of the many... dont bother - its S**T
I think this will turn out to be a great series of white papers helping people get to grips with the process of configuring and compiling their own kernels, but I have to say that I think there would be faster progress on new kernels if the was some central repository of precompiled binary packages for the major distros throughout the development cycle.
The truth of the matter is that now linux is gaining wider acceptance, the community is filling up with more and more noobs and we should be doing more to help them understand the "new" (to them) technology. We also need to remember that not everyone who wants to use the software needs to be some sort of guru.This article is a great start to moving more people to the new code quicker, but regular up to date debs/rpms for all the current distributions will push that long even faster. I know someone will probably post saying "but there are packed versions for xxxx at somewhere.org", but they are often difficult to find for the noobs who just don't know where to look.
The 2.6 kernel is noticeably faster on my dual Athlon 2100+mp, at the user interface; X is faster than I've ever seen it before; the realtime scheduling is awesome.
In short, as soon as you can reasonably do so, I recommend you migrate to the 2.6.x kernel.
Thinking outside my Head
As opposed to, say, Microsoft or Oracle providing the software? And you don't think that the Chinese (or Indians, or Russians, or whatever) have the ability to write software themselves, that programming is somehow a uniquely American talent? If you get your head out of whatever other anatomical place it's parked in, you'd see how silly that is.
One of the advantages of Windows never mentioned in the article is the ability of Microsoft to remotely deactivate Windows XP in the case of a national emergency.
And you wonder why the rest of the world is moving away from it. Whether it's true or not, just making statements like this hurts the U.S. software industry more than anything that Stallman can say.
A known proponent of socialism, the Chinese government and RMS are natural allies.
Well golly, Bubba, it seems that the Bush administration has been getting pretty cozy with the Chinese government, too. If anyone's looking to sell out Taiwan, there's where you want to direct your vitriol. Are you aware that the President's brother, Neil Bush, is getting rich lobbying for China? No, I didn't think so...
Oops... I think I got baited!
If you are upgrading an NForce-based machine to 2.6.x, save yourself some headaches and add "noapic nolapic" to the Kernel append string. I experienced repeatable hard lockups when doing disk intensive I/O until adding those parameters.
Also, NVIDIA's nforce package is no longer necessary. The experimental forcedeth driver in 2.6.2 works quite well in my experience, and apparently an Intel sound driver works for the NForce onboard sound.
See my latest journal entry for my account of migrating MDK 9.1 to a vanilla 2.6.1 kernel.