Slashdot Mirror
Upgrading Your Current System To Kernel 2.6
An anonymous reader writes "This white paper provides an overview of the process of moving an existing desktop system to the 2.6 kernel. It will highlight other software requirements imposed by the new kernel and administrative changes that you must make when migrating an existing system to the 2.6 kernel. It supplements previous whitepapers in the same series about Customizing the 2.6 kernel [Slashdot discussion here(1)] and porting drivers to the 2.6 kernel [Slashdot discussion here(2)] to the 2.6 kernel."
I had all kinds of trouble upgrading to 2.6. Sound and networking didn't work, as well as some of my filesystems. Luckily, I'm smart enough to setup lilo to run multiple kernels, so going back to 2.4 was no trouble to tweak my config file and recompile and try again. I never got everything totally working right. I was going to just wait until the next series of distro releases solved these problems for me, but maybe I should give it another shot.
...laptop touchpad don't bother going near 2.6 if you don't know exactly what you're doing. I still haven't got the mouse working. I can't even find anyone who can tell me how to confirm if the touchpad is a synaptic.
I'm not the only one suffering this.
Works great on my slackware desktop.
One thing that was a blocker for me to move to 2.6 was my mouse would go a lot faster while in X. I finally found that it was the new input system. Under 2.4 the default XF86Config file would have two mouse input settings, one for ps/2 and one for USB. Under 2.6 both of these were picked up regardless of the mouse being ps/2 or usb so all mouse events, clicks, etc were picked up twice. Removing one of the mouse entries made everything work as normal.
Probably everyone but me knew this, but thought I'd throw it out in case anyone else is in the same boat.
Here's my experience with 2.6 kernel (been running it since 2.6.0 was released).
DVDs look awesome. I had to tweak the hdparms for DMA, but they work great.
Ever since NVidia came out with the latest drivers, things like the UT2k4 Demo fun fantastic.
I was a little hung up on modules... seeing as I rarely use them, it wasn't a show-stopper. The conversion from modutils to module-init-tools was mostly painless.
Recently, I've been playing with MTD, and trying to get a test machine to use 12 out of the 16 megs of an AGP Voodoo3 3000 card's memory as a device I can format or use as swap. I have been unsuccessful. (2.6.3). This is also on a testing machine, not my "main" machine.
On a slightly OT note, planning on building a Mini ITX system with a Via Epia board (one of the 800 mhz ones). Should have the case this week, jury's still out on the mb.
Other than that, no complaints, it's been fantastic. I'm running 2.6.3 on 3 different machines (with different responsibilies) and it feels like there's no going back now!
FLR
I run gentoo on a dual pIII-600 with an ATI radeon7000 / SBlive / intelpro100. Kernels 2.4.18 through 2.4.22 took several hours of tweaking before I could even get the thing to boot correctly. ( 2.2 never had this problem )
Last week I took about 30 minutes and grabbed 2.6.3 did a clean/config/make, which took about the majority of that time, and booted into the fastest Linux box I have ever had.
2.6 booted with OpenGL without any tweaks pushing glgears to 1600fps and ALSA kicked in without errors on the emu10k1. Device drivers posed no issues for either the USB keyboard/mouse or hardrive or nework card.
No 'migration' was necessary for either windowmaker / enlightenment / blender / JACK or any of my other 100 some odd apps.
I've tried all versions of 2.6 (2.6.1,2,3) but still get this problem. Everything seems smoother/more responsive except for a problem I have with a game (enemy territory). In 2.4 this game usually loads a map in 20-30 seconds. In 2.6 it takes about the same time but every map thereafter gets longer and longer, until they start taking several minutes. I did a test by launching my own server and just kept reloading the same map. 1st try 32 seconds, 5th try over 4 minutes. If anything it should be slightly quicker as some would be in memory/swap. The only other game I have in linux similar to enemy territory is quake3. It has the same normal load times in both 2.4 and 2.6, but it's maps are much smaller than enemy territory's so it doesn't stress the system as much. This leads me to believe that it's more a swap/memory issue than a graphics driver issue. I did hdparm tests and dma is enabled and I'm getting about the same speed in 2.4 as in 2.6. My system is a p4 1.6 with 128 megs ram gf4ti 4200. Now, I realize 128 megs is low these days, and would probably help aleviate this problem but it seems when a system is stressed in this way 2.4 performs better than 2.6. In 2.4 I can play on a server for as much as I want but with 2.6 I usually get kicked within a couple of new maps due to it timing out. Reconnecting to the server doesnt help, but quiting ET and restarting helps for that initial map, but then the cycle repeats. I don't see a way on this forum of attaching my config, I compiled the kernel myself and have gone over it several times to see if some option could be the cause of this. The first thing I tried was turning off the preemptive kernel option, but didn't help. My system is debian based (morphix distro) and as I mentioned I compile the kernel myself, not a precompiled kernel. I also made sure X doesn't have a negative nice value. You might suggest to throw more ram at the problem and even though it might help, I shouldn't have to as 2.4 seems to get by.