Slashdot is powered by your submissions, so send in your scoop

 


Forgot your password?
Close
typodupeerror
×
Linux Software

Vanilla Kernel 2.6 Stability vs 2.4? 129

Posted by Cliff from the to-upgrade-or-not dept.
chromis asks: "I am a 'Linux-from-scratch' like Linux user. I maintain my system for almost 4 years that way. I'm still using kernel 2.4, and I'm a little bit afraid for updating to 2.6 because of the problems like stability issues, driver subsystem problems, etc. I once tried 2.6.0 a long time ago, but I experienced random freezes which I could not diagnose. We all know about the development model issues, and I often read complaints about current kernel development practices. Now that kernel 2.6.13 is out, I really want to ask Slashdot: if you are a vanilla 2.6 kernel user, how are your experiences with these plain kernel.org 2.6 kernels? Is it really as bad as some people claim, or is 2.6 only usable when using a distro from a large vendor like Red Hat, SuSE, etc? I really would like to upgrade to the new vanilla 2.6 kernel eventually, but I'm a little hesitant. Any advice?"
This discussion has been archived. No new comments can be posted.

Vanilla Kernel 2.6 Stability vs 2.4? More

Vanilla Kernel 2.6 Stability vs 2.4?

Comments Filter:

  • This I know (Score:5, Funny)

    by xactuary ( 746078 ) on Tuesday August 30, 2005 @09:28PM (#13441614)
    I am running the older kernel and it kept me from getting first post.

  • Works for me (Score:4, Informative)

    by meowsqueak ( 599208 ) on Tuesday August 30, 2005 @09:30PM (#13441629)
    I use the vanilla kernels with moderately modern hardware (up to about 4 years old) and I have no stability issues whatsoever. I tend to stay within one or two versions of the bleeding edge release.

    The advantages of the 2.6 kernels (udev, nptl, device driver model) outweigh the disadvantages (i.e. risk) for my situation, in my opinion.

    That said, I still use linux-2.4 on my headless server, mainly because I haven't been bothered to upgrade it recently. It works fine, so I see little point in changing it.

    • Re:Works for me (Score:4, Interesting)

      by apoc.famine ( 621563 ) <apoc.famine@NOSPAM.gmail.com> on Tuesday August 30, 2005 @10:31PM (#13442102) Journal
      To really answer your question, what they (not me - I don't know crap about linux kernels, other than which one I've just upgraded to) need to know is what you'e using your computer(s) for. Are the advantages of the 2.6 kernel as listed above advantages for you?

      Personally, I moved from Win2k to linux (gentoo) due to instability in 3rd party software which I could get decent replacements for in linux. I generally run the lastest gentoo kernel, and haven't noticed any instability other than what my n00bishness has artificially created. Not that that helps you at all, I'm guesing. ;)

      • Comment removed (Score:5, Informative)

        by account_deleted ( 4530225 ) on Tuesday August 30, 2005 @11:09PM (#13442359)
        Comment removed based on user account deletion
        • i run 2.4 on my web"server" (it's a p1 200), and 2.6 on my dad's ibook (it's older, dual boots gentoo and os 9) and my workstation, which is about a year old. i like 2.4 on the server because i don't have to update it much (same goes for apache 1.3.x and php 4.x), and the last bzImage i made was 493k :)

          i just rebooted the server for a ram upgrade and general cleaning, but before that it was up 54 days. i don't want to have to move the server duties to my workstation or my dad's laptop because they released
    • IMHO, there is very little reason to stay with the 2.4 kernel, and many reasons to migrate to the 2.6 kernel. A caveat here, though. It really depends more upon your "comfort level" with tweeking the kernel and rebuilding it, as well as the current distribution you run.

      An explaination is in order. Red Hat, as well as some other linux distributions, have a tendency to back-port wanted new features into older kernels. In my experience, the mish-mash of shared libs required to achieve back-porting makes th

  • Seems ok. (Score:5, Informative)

    by ColaMan ( 37550 ) on Tuesday August 30, 2005 @09:41PM (#13441729) Journal
    I've used various incarnations of 2.6 on my mythtv box. It's under fairly high load, with memory, video and disk intensive processes, has high PCI utilisation (2 capture cards, sometimes running at once). It runs 24/7, sometimes hot enough to get the CPU temp alarm beeping.

    Number of times it's had a kernel panic over the last year? Zero. Good enough for me.

    And as other posters have said, the advantages with hardware, latency patches, acpi support help too.
    • Just a question : what stops you from adding this new kernel to your grub or LILO and stress test it in any condition you think will cause a freeze ?

      I mean there are few things as easy as installing a new kernel and then removing it later if it doesn't satisfy you...

    • Re:Seems ok. (Score:4, Informative)

      by spagetti_code ( 773137 ) on Tuesday August 30, 2005 @10:44PM (#13442182)
      Its not all rosy. The kernel version (2.6.11.7) used in Knoppmyth has an issue [google.com] with USB hard drives - hence my extra HDDs plugged into the myth box were attached using firewire (not that its a problem).

      My point is that the 2.6 kernel is not without its flaws - it depends which particular kernel version you get and which warts it has.

      My advice is if it ain't broke...
      Do you really *need* the hassle of the upgrade?

    • Re:Seems ok. (Score:3, Insightful)

      by extrasolar ( 28341 )
      But what's your uptime?
      • Uptime of that box is in the order of a few weeks to a month, before power issues or my tinkering ends in a reboot (hmmm, I wonder if I can get bios wakeup going in myth? (much rebooting later).... not yet).

        Personally, if it's stable for a few weeks , then you've only really got a couple of rarely-encountered corner cases that could cause it to fall over. And I rarely encounter them :-)

  • Gentoo 2.6.13 (Score:3, Informative)

    by Professional Slacker ( 761130 ) on Tuesday August 30, 2005 @09:45PM (#13441756) Homepage
    I can't speak for the vanilla kernel, but the Gentoo 2.6.13 kernel borked my system something fierce. The init process grinds to an virtual stop just after loading the kernel, it took a minute and a half to set the host name, I still haven't had the patience let it finish booting. But that's the risk of using a fresh kernel. 2.6.12 didn't give me any trouble.

    • Re: (Score:2, Interesting)

      by account_deleted ( 4530225 )
      Comment removed based on user account deletion

      • Re:Gentoo 2.6.13 (Score:5, Informative)

        by rincebrain ( 776480 ) on Tuesday August 30, 2005 @09:55PM (#13441854) Homepage
        Seconding. Gentoo user of vanilla sources straight from kernel.org, and the gentoo sources are a lot more unstable than the vanilla sources.

        I use vanilla sources regularly, and haven't experienced any problems as a result. I recommend them to all of my Linux friends, too.
        • What the hell are the gentoo people doing to the kernel? The Debian kernel maintainers patch the kernel to fix bugs but it has never disrupted the stability of my system. I also use vanilla while the latest kernel isnt packaged yet so I've had my share of uptime with vanilla kernels as well. Does adding some bugs somehow better the 1337ness of gentoo?

          • Re:Gentoo 2.6.13 (Score:5, Interesting)

            by rincebrain ( 776480 ) on Tuesday August 30, 2005 @11:32PM (#13442515) Homepage
            Gentoo seems to add unnecessary patches to the kernel that break things. I've had at least three cases in which using a vanilla kernel instead of gentoo-sources fixed the problem that was being experienced by the Gentoo user.

            The problem is that Gentoo doesn't add patches to fix known kernel bugs, they add patches to resolve user problems.

            Guess what that does to stability. :)

            Debian packaged kernels, like msot of Debian's stable branch, are very...stable, in contrast.
          • Comment removed based on user account deletion
        • Well, you say that, but:

          # uname -r
          2.6.8-gentoo-r3
          # uptime
            09:55:26 up 299 days, 19:41, 2 users, load average: 0.01, 0.08, 0.10

          # uname -r
          2.6.7-gentoo-r11
          # uptime
            10:00:01 up 252 days, 19:58, 1 user, load average: 0.57, 0.29, 0.18

          So clearly Gentoo kernels aren't necessarily the kiss of death for stability. Maybe it's a recent development in Gentoo-land.
      • Yeah, I don't get that about Gentoo at all. Back when I was taking it for a spin I saw some of the wierdest behaviour with thier kernels, yet they wouldn't put the latest plain ol' kernel.org release in the portage tree.
      • I use the vanilla kernels under Gentoo ever since I spent a week trying to get multiple targets on a firewire to ATA adaptor to work (on PPC), and finally found that it works perfectly under the vanilla kernel(s) of the same versions. I don't know how many times I rebooted, rebuilt kernels + modules, and patched various sources that week, but I'm still pretty pissed at the Gentoo kernel patching idiots for stealing that time from me.

    • Re:Gentoo 2.6.13 (Score:5, Interesting)

      by thegrassyknowl ( 762218 ) on Tuesday August 30, 2005 @10:17PM (#13441999)
      I've heard of several issues with Gentoo kernels on Multi-processor systems.

      I run a 2.6.12 on my desktop, and we are playing with 2.6.12 in a high-load embedded system. They both seem to work well enough here. I've only ever seen kernel panics when my network switch fails (damn dicky power connector; been meaning to replace) and the NFS-mounted root on the embedded box goes away.

      I recently upgraded my laptop to 2.6.13 and it brought all manner of problems (wireless didn't work anymore. Sound problems that were fixed in 2.6.12 reappeared, etc). I think most of my problems are with the IPW2200 driver modules I have loaded, so I just rolled back to 2.6.12 where it all works well.

      Stick with 2.6.12 for now if you're scared of problems. I can safely say that it is pretty damned reliable.

      • Re:Gentoo 2.6.13 (Score:4, Informative)

        by thegrassyknowl ( 762218 ) on Tuesday August 30, 2005 @10:20PM (#13442022)

        I've heard of several issues with Gentoo kernels on Multi-processor systems.

        I just remembered the exact problems I'd heard of, and it wasn't Gentoo-specific (but it only appeared on Gentoo for some strange reason)... It was a CPU freq scaling thing with AMD64 CPUS. Apparantly the latest driver is broken and when the frequency scales down in one CPU the kernel detects a loss of sync and panics, instead of realising that the CPU frequency is scaling and compensating for it.

        It's fixed by disabling CPU frequency scaling. Apparantly AMD are working on a PowerNow patch for it, but that is just hear-say AFIK.

        • I wonder if there's the same bug in x86? Because I do use SMP (for a HT P4), and cpu scaling. I'll see what happen if I build a 2.6.13 without cpu scaling. Thanks for the information.
          • Yes in fact the SMP/Scaling conflict does carry over in to x86. I removed SMP from the kernel, booted over to the new kernel and life was better than it was the last time I tried to use 2.6.13. However without SMP the modem that I don't use shat it self. An other module is getting the axe it looks like.
    • I'm using Gentoo sources (2.6.11 something and 2.6.9 something) on my laptop and server, no problems. With regards to 2.4 vs 2.6, the biggest improvement I've noticed is the addition of preemption, it really makes a huge (perceived) difference for a desktop machine. ACPI support is nice too, I'm not sure how it is in 2.4, I always used APM with 2.4.

    • Re:Gentoo 2.6.13 (Score:2, Interesting)

      by M1000 ( 21853 )
      As I am typing this, I'm about to do make modules_install of gentoo's 2.6.13 kernel ;-)

      You do know that this release dropped support of devfs ? From now on, you'll need a udev system.

      Took me half an hour to convert to udev on 2.6.12, and everything went right.

      But again, I as reboot on 2.6.13 after typing this, maybe I'll regret it ;-)
      • Lucky you, I made the move to udev, and all was fine and dandy except devfs refused to bow out, I ended resigning that despite removing it from the .config, running make clean && make && make modules_install, and then unmerging devfsd, I'd just have to keep using the nodevfs, devfs=nomount, and udev boot flags. Udev works fine for me under 2.6.13 (and 2.6.12 boot flags aside), but I can't get the ndiswrapper working yet, so it's back to 2.6.12 for the time being. This is the first kernel upg
        • (2.6.13 works ok, btw)

          1. make sure you drop devfs and devfs mount at boot from kernel
          2. make a `make mrproper` (save your .config elsewhere before)
          3. I see that I'm still having devfsd installed; don't seems to be a problem in my case.
          4. I edited /etc/rc and told the file to use udev, and not auto. (not required, but I wanted to be sure).

          --

          Decided to compile 2.6.13 with the default timer frequency at 250 hz (was at 1000hz with 2.6 before)

          I also selected the premption level to 2 (desktop).

          --

          Only bug so far:

          i

      • this release dropped support of devfs ? From now on, you'll need a udev system.

        My god. My laziness has finally to wake up. As obsolete as devfs is, it still worked ok for me on my Gentoo...

    • Re:Gentoo 2.6.13 (Score:2, Informative)

      by keltor ( 99721 ) *
      Before people start down the decrying Gentoo path too much, realize that Gentoo offers a bunch of different kernel choices, one of which is called gentoo-sources and that is what this poster is referring to, not Gentoo using 2.6.13 or anything else. Just this one source tree. I use vanilla sources on a unstable (as in latest greatest version of software not stability) and I have not had a problem. My uptime is typically the time between when I notice one kernel version and when I notice the next kernel v
  • I've been using 2.6 for quite a while now (maybe 2.6.3) and for the most part I've liked it. I've had maybe on oops ever and its stable and fast.
  • (For reference: Distro: Gentoo x86_64)

    No real awful speedbumps I couldn't get around except one: Around the 2.6.10 series, ran into some strange data corruption problems with the data in extremely large files (~1GB range per file) when transferring them. Never managed to pin it down strictly to the kernel, so I can't blame it for certain.

    Apart from the fact that I need one out-of-tree driver, no issues with vanilla 2.6 ever since 2.6.11.

  • very stable (Score:3, Interesting)

    by Paul Jakma ( 2677 ) on Tuesday August 30, 2005 @09:49PM (#13441800) Homepage Journal
    Core 2.6 seems very stable to me. There's always variation in drivers though, but even those are better in 2.6, afaict, eg:
    # uptime
      02:44:06 up 173 days, 8:46, 7 users, load average: 0.59, 0.30, 0.28
    # uname -r
    2.6.10-1.770_FC3
    Only occasional power outages and required kernel upgrades have taken it down. 2.4 was reliable too on that hardware though.

    • Re:very stable (Score:5, Informative)

      by ivan256 ( 17499 ) * on Tuesday August 30, 2005 @09:58PM (#13441874)
      Doesn't the FC3 at the end of the version string mean "Fedora Core 3"?

      That's not a vanilla kernel, it's a patched up kernel from RedHat.
      • That's not a vanilla kernel, it's a patched up kernel from RedHat.

        Bah, yes. I completely missed the "vanilla versus patched-up" aspect of the question. I should go to sleep.

        Note however that FC tries to /not/ patch-up the kernel, like RH used to do for RHL. Only obvious fixes that are already on their way into vanilla are /supposed/ to be allowed into FC (AFAIK anyway).

        So, let me correct myself: No idea about "vanilla", but FC kernels have been rock solid for me, and FC kernels should be pretty close to "va
  • i have used the 2.6 kernel series since it came out. I run Gentoo and i have a few debian systems. I try to keep up with the latest version. the gentoo system has 2.6.13 installed and i have no problems.

  • >uname -a && uptime
    Linux maverick 2.6.6 #15 SMP Fri Jun 4 19:58:51 EDT 2004 i686 GNU/Linux
      21:47:15 up 92 days, 25 min, 2 users, load average: 1.40, 1.38, 1.37

    It would be longer, but that's exactly the amount of time it's been since I moved into my house.
    • Your uptime should never be greater then the time since the last git snapshot release :)
    • It would be longer, but that's exactly the amount of time it's been since I moved into my house.
      Pfft, that's no excuse.
      • Heh. You jest...

        I briefly considered moving the machine connected to the UPS the entire way... The UPS can power it for about 7 hours (APS SmartUPS4000, AMD Athlon XP-M 2400+ with PowerNow enabled), but the UPS alone was heavy enough to carry on the stairs, so it ended up getting switched off... It's not like it would have been connected to the network for those few hours anyway, and unless it's online and serving I don't consider it "up" anyway.

        • Re:Does this help? (Score:4, Funny)

          by GoRK ( 10018 ) on Wednesday August 31, 2005 @12:53AM (#13443025) Homepage Journal
          I moved an old 486/33 server once back in the day while it was in the middle of the OS install... We had started in one building in the room we thought we were getting, but we ended up getting a room on the entire opposite side of campus. Since installing a linux distro in this day was not exactly a speedy process, we decided to go ahead and move it while it worked. Three of us carried it.. One fellow on the CPU, one on the UPS, and one on the monitor -- we decided that if we were going to move a running computer, UPS alarms blazing it would be pointless for people to see us doing it without the screen powered up and scrolling mounds of text. We decided to hand carry it also as we though rolling it on a dolly might harm the running hard drive due to vibration.
  • 2.4 will give you a more stable kernel due to being tested better. I am still sticking with 2.4 on my servers, but just because I'm lazy. :)

    I've been running 2.6 on all desktop & laptop systems without problems since 2.6.9 (about a year). I certainly would not want to give up the better interactivity, better MM performance, wide hardware support, and features like udev.

    2.6.x will have have hickups now and again because that is where the development occurs. That's why a few kernel hackers (Chris Wrigh
  • You will need the appropriate hotplug utilities, get them here. [kernel.org]

    The 2.6.8 kernel had an issue with CD writing (only root could do this). This has been corrected in later kernels. You may have to delv into the udev rules to get things setup the way you like. Read the fine HOWTO [reactivated.net] on writing rules for udev.

    My cd-rom did not get recognized after boot unless the ide-cd module was called before udev started. There was a mixup with tty and pty in the default udev rules around the time of switch between 2.6.7 and 2.6.8 and it obliterated the 'less' and 'man' commands. How convenient is it that I can't run 'man udev.rules'?
     
    I believe this has also been remedied since then. If in doubt I suggest taking the following steps.
    1. wget ftp://ftp.kernel.org/pub/linux/kernel/v2.4/linux-2 .6.x.tar.bz2 [kernel.org]
    2. tar jxvf linux-2.6.x.tar.bz2 && cd linux-2.6.x
    3. less README
    4. cd Documentation && less post-halloween-2.6.txt

     
    Title of the second document is "The post-halloween document. v0.48 (aka, 2.6 - what to expect)". That should tell you everything you need to know about upgrading from 2.4.

  • what /boot is for.. (Score:2, Informative)

    by Daxster ( 854610 )
    Why not put both into /boot and add both to your bootloader? There are lots of tutorials for doing that, for 'testing' in case a newly compiled kernel doesn't work.
    I have this on my Slackware -current computer. A vanilla 2.4, vanilla 2.6.10, and compiled 2.6.10 which I use and have had no issues with. I plan on compiling 2.6.13 soon, to keep up to date with bug fixes and improvements.
  • What's a good walkthrough on installing linux from strach?

  • I've been using the latest 2.6 kernel, patched with Gentoo [gentoo.org] and Suspend2 [suspend2.net] patches. I started with 2.6.9, and it had some ACPI problems, but once I upgraded my BIOS to the latest version and upgraded the kernel to 2.6.10, everything worked well. Other than those specific ACPI issues, I've had no general stability problems. Everything works well.

    I used to run Slackware, and I have to say that when I upgraded it from a 2.4 kernel to 2.6, the system did perform better. I think that if people just upgrade cau

  • Something doesn't add up here... (Score:5, Insightful)

    by bergeron76 ( 176351 ) * on Tuesday August 30, 2005 @10:26PM (#13442062) Homepage
    You're an LFS Linux user ("Linux-from-scratch like" Linux user as you put it), and you're concerned about upgrading your kernel?

    Since you cross-compiled and built your libraries, compiler, toolchain, etc from scratch, why are you worried about upgrading your kernel? Surely, you know that it's trivial to modify your boot loader so you can boot multiple kernels and try them out without consequence to your system.

    Second, why are you interested in using a Vanilla 2.6 kernel if you built your entire Linux system from scratch?

    Please pardon me if I'm mistaken, but you certainly don't sound like a 'tweaker'. Your question is analogous to: "I'm a die hard car tuner, I've modded my hotrod and tweaked my cam's, changed my gear ratios and added 2 inches to my manifold: Should I use premium gasoline in my new Hot Rod?"

    • What's the problem? I run LFS on a server, specifically because it's rock solid. The actual compilation is no big deal, nor is modifying the bootloader, but why should I, if I can get information first?

      The car analogy is actually very good -- should I use premium gas, given the number of horror stories I've heard about it? Many tweakers also like to do their research first... after all, why make your own mistakes if you can learn from others'.
      • You should just try it. Maybe it will run better. Maybe it won't. You won't know until you try. Different people will have vastly different experiences with the same things in life. I mean sheesh, just look at sex for chrissakes! People generally build a test server first and work it under various loads to see if it will perform as well. If you think you can squeeze a good 10-20% decrease in server load by going to a newer kernel you should likely be looking at buying new hardware anyways if your little box
        • > I mean sheesh, just look at sex for chrissakes!

          Whoa whoa, I was following you until you brought sex into the mix. Now I'm completely lost :)

          Do you mean that we should watch what the pornstars do when they want to upgrade their Linux kernel?

          (Unrelated side note: If you've ever bought anything from IKEA, you know that's its hard to get the damn stickers with the barcodes off the bottom of the item you bought. A few days after visiting IKEA, I noticed the funniest thing in a pr0n movie -- they were usi

    • Re:Something doesn't add up here... (Score:4, Interesting)

      by chromis ( 738106 ) on Wednesday August 31, 2005 @06:15AM (#13444179) Homepage
      I understand why you ask this question :)

      Well, i'm not a real tweaker in the sense that I compile and tune everything for maximum performance. I rather tune the system to my specific software needs and stability in the sense of "if i don't ask for x, i don't have x". My system is very basic and i have a good overview. It contains only things that I need. I really like to put some effort in installing software so that I am aware of all it's features, dependencies and caveats. I like to do this by hand and by reading documentation from the software authors themselves. Yes, perhaps it is a tedious approach but it works very nice for me and i have a system which i can really trust. For me, this is the power of Open source actually.

      Before I upgrade to a major version (be it a major GCC version - I worked with gcc 2.95.2+some patch for a long, long time before i upgraded to 3+, or in this case the kernel), I always spend some time researching if the upgrade is worthwhile and good.

      So, yes: I cross-compiled and built libraries myself ofcourse, but i always try to choose stable versions. Also with kernels: i never tried an odd (2.1, 2.3, 2.5) kernel release.

      In case of the kernel, I am little bit confused because of the development model (no 2.7), fast development cycles, in relation to the comments and complaints I sometimes read on the internet and here on Slashdot. Regarding kernel stability, it is my understanding that 'stability should be guaranteed by vendors' ie. 'use a vendor kernel'. I am my own vendor, so to speak. Hence my question.

      In my years of experience, i know that critical parts of the system (toolchain, kernel) can produce very strange problems not directly noticable in a week of testing.

      Yes, such risks are always present when using free software, but software from a stable chain always worked perfectly for me. Especially software where no-one complains about :)

      I found it very difficult to find information regarding this, hence i tried Ask Slashdot.

      • I ran the vanilla kernel on Gentoo for a while and didn't have a single crash in months of uptime. That was back around 2.6.9 or so, on a server. Does that help?

        (I'm now running a mixture of Gentoo and Debian kernels. No crashes from them either.)
    • "I'm an old slashdot user. I'm afraid people on slashdot wouldnt understand my perfectly understandable and clear point in a computer related discussion, so I make an analogy with cars."
  • This is just my personal opinion, but I'm not really comfortable with 2.6 until they make a 2.7 to toss all the gee-whiz development stuff in to, which they seem to still be using 2.6 for. Until then, mentally I classify 2.6 as a development/unstable kernel, and don't use it for much. I have a couple friends that have had good luck, and some that have horrid luck. One of our servers at work has some very, very odd issues with 2.6, and there are others that won't run 2.4 without patched drivers and the li

  • not perfect (Score:3, Informative)

    by Bodhidharma ( 22913 ) <jimliedeka@NOSpAm.gmail.com> on Tuesday August 30, 2005 @10:44PM (#13442180)
    I like the 2.6 kernel over the 2.4 kernel because I can play MP3s and Oggs without skips every time I refocus the window.

    On the down side, I'm running Ubuntu 5.04 on a Sony S270 laptop. I use the 2.6.11 when I want sound to work at all and 2.6.10 when I want my touch pad to work right. I've tried a couple of custom compiles of 2.6.10 and 2.6.11 but haven't gotten either to work right yet.

              Jim

    • I like the 2.6 kernel over the 2.4 kernel because I can play MP3s and Oggs without skips every time I refocus the window.

      I have very good reasons to believe that MP3/Ogg skips are not a kernel problem, but a set up problem.

      On the down side, I'm running Ubuntu 5.04 on a Sony S270 laptop. I use the 2.6.11 when I want sound to work at all and 2.6.10 when I want my touch pad to work right. I've tried a couple of custom compiles of 2.6.10 and 2.6.11 but haven't gotten either to work right yet.

      I have a

  • Faster (more responsive) and udev works for me (Score:3, Informative)

    by mc_barron ( 546164 ) on Tuesday August 30, 2005 @11:05PM (#13442317) Homepage
    Have grown to really love 2.6.x kernels. Started back in the low single digits, just upgraded to 13. NO stability issues for me, ever (at least not due to the kernel). The important things to me in 2.6: udev and better response time (switching between windows, etc). Just recently got udev working just the way I like it - can't imagine going back to the old devfs.
  • So far, my experience with 2.6 has shown that older hardware and 2.6 are a potentially dangerous combination. 2.6 appears to work fine on newer hardware. [Note: I am specifically referring to generic x86 systems.] On older hardware, it can be hit or miss.

    There are some great features in 2.6, however for a production environment where it just has to work, the safer bet is 2.4.

    For me, it's really pretty annoying. Used to, you could generally count on "released" kernels to be pretty stable (1.2, 2.0), but that
  • I once tried 2.6.0 a long time ago, but I experienced random freezes which I could not diagnose.

    I tried 1.1 a "long time ago". (11 years or so.) If I'd known that 2.6.0 was available back then, I'd have tried it. (I still keep the Slackware 2.1 CD handy for quicky 486 installs.)

  • My experiences with vanilla 2.6 (Score:3, Interesting)

    by schon ( 31600 ) on Tuesday August 30, 2005 @11:30PM (#13442508)
    I run Slackware, and just started using 2.6 in production a couple of months ago (about when 2.6.12 came out.)

    So far, I'd recommend staying away from udev - it's just way too flaky for words - it seems OK if your hardware doesn't change, but when you start hotplugging and the device nodes don't show up unless you "sudo /etc/rc.d/rc.udev restart", it gets old very quickly.

    Stability-wise it's OK, I'm using it on two desktops, three servers and my laptop, and haven't had a crash or oops. (Although I've only been running it for a couple of months.)

    General desktop performance (KDE) is OK - I saw no noticeable difference from 2.4.

    NWN is noticeably slower however - there seems to be a lot more disk thrashing while playing, even though swap is unused and there is a ton of free RAM (I think I might need to tweak something in /sys/block/hda/queue.) For the time being, I've switched back to 2.4 for NWN.
    • So far, I'd recommend staying away from udev

      And you'd suggest what as an alternative? The pile of broken crap that is devfs? The whole point of udev is that new devices do show up; it sounds like you have some misconfiguration issue.

      • you'd suggest what as an alternative?

        How about static device files in /dev? You know, like how the Linux world worked for years before people wanted to mimic FreeBSD?

        I can live with /dev/sda being there even when my thumbdrive isn't connected. It doesn't bother me one whit.

        The whole point of udev is that new devices do show up

        Yes, and it's currently broken (at least in my experience.) That makes it rather pointless, no?

        it sounds like you have some misconfiguration issue

        That's entirely possible, however t
  • I've been using Gentoo for a long time now, and I've compiled and installed every release (including every minor release) since, probably, 2.6.9 . To be honest, I'm not using the vanilla sources, but the "ck" patchset which includes the staircase scheduler (compiled specifically for amd64). Even with these changes, I've found the 2.6 kernels to be the most stable software you can run on a computer. I can't remember one time that my system has crashed (unless I had broken something through my own stupidity).

  • 2.6 is nice, kinda, maybe... (Score:3, Interesting)

    by SilverspurG ( 844751 ) * on Tuesday August 30, 2005 @11:58PM (#13442682) Homepage Journal
    I was on the same track. I was stuck in 2.4 land for a long time just because I had gotten my systems to the point where every piece of hardware worked and I knew how to get it all working again if I upgraded my kernel. Like you, I had trouble with the 2.6 kernel upgrade. I tried it once (circa 2.6.4) and it was a catastrophe for my wireless cards (madwifi and centrino). Finally I let Debian sid put in 2.6.12, and it seems all the 3rd party drivers have upgraded to the 2.6 bandwagon.

    Configuration: I could run through the 2.4 configure tree in 20 mins or less. It takes me at least twice that in 2.6. Too much IP and an effed up broken patent/copyright system creating too many incompatible devices at levels that aren't easily segmented into kernel layers.

    Compilation: Yeah. It takes a lot longer.

    Performance: I noticed that mouse response in X is a lot faster. That's probably an artificial representative, though. I haven't really noticed load or response times to be much different from 2.4 to 2.6. Running on 400 MHz machines, I still notice this when it actually improves.

    Modules: On a Debian 2.4 kernel I had maybe 12 modules loaded. On a LFS 2.4 kernel I had maybe 4. On Debian 2.6 kernel I have 91 modules loaded and many of them are for hardware which I don't have (see the section on configuration: there are too many devices which look the same to the kernel but are different due to IP pissing matches).

    Udev: I hate it. I don't hotplug. I don't want to hotplug. Hotplugging is evil. My system shouldn't be doing anything with a device until I say I'm good and ready for it. Except for hotplugging, there's no real need for udev.

    Mostly I'm upgrading to 2.6 because I can't afford to be left in the dust.

    PS. No real LFS'er would call it Linux-from-scratch. Lose the hyphens.
    • On a LFS 2.4 kernel I had maybe 4. On Debian 2.6 kernel I have 91 modules loaded and many of them are for hardware which I don't have

      Could you not have started your comment with that line? Then I could have avoided wasting my time reading the rest of it.

      Are you saying that you can successfully build an LFS system, but you can't work out how to stop kernel modules from loading?

      What exactly did you learn from LFS?

    • I don't hotplug. I don't want to hotplug. Hotplugging is evil. My system shouldn't be doing anything with a device until I say I'm good and ready for it. Except for hotplugging, there's no real need for udev.

      I'm with you on this. That's why I'm pissed off at the lack of support for devfs from 2.6.13

      My god, the Linux kernel still supports dinosaur-era things like Minix file systems or m68k cpus (and it's good it supports them IMHO) but suddenly stops to support the device filesystem management it had un

      • You're telling me that devfs was entirely removed from 2.6.13? Great... Just fucking great. Gentoo previously defaulted to devfs, and I've got a bunch of Gentoo boxes in production.

        That's just fucking fantastic.

        I had read previously that devfs was going to remain for the life of 2.6. Of course, with the constant turmoil happening in 2.6 development, I really shouldn't be surprised.

        Yay, more fun stuff to deal with down the line.

        • I read it on /. comments, but now I checked the 2.6.13 changelog. It seems it's true. Sigh. See for example below.

          [PATCH] devfs: remove devfs from Kconfig preventing it from being built

          Here's a much smaller patch to simply disable devfs from the build. If this goes well, and there are no complaints for a few weeks, I'll resend my big "devfs-die-die-die" series of patches that rip the whole thing out of the kernel tree.

      • Because nobody stepped forward to maintain devfs! A number of people offered but nobody actually followed through. That should tell you something right there.

        The version of devfs that you use right now is horrifyingly buggy, especially on SMP systems. Switching to udev will take less time in the long run than trying to the keep bloaty and rusted devfs code working. Switching to udev is generally very easy.

        • I don't say devfs is good, just that it works on my machine right now and I don't feel committed to change. They could leave devfs in the kernel, perhaps saying something like *deprecated-use at your own risk* and dropping it in 2.8, not in 2.6.13

  • but I've run vanilla 2.6 kernels since 2.6.5 on four different slackware installs at work and at home: a year old dell dimension P-IV box with an nvidia graphics card, an ancient dell 300 MHz P-II machine, an even more ancient 75 MHz P-I toshiba laptop, and a newish 32 bit AMD / low-end Asus motherboard / ATI graphics card pc built from parts.

    Never had any hardware problems with any of them. (Although I haven't upgraded the two older machines since around 6.5 or so - and it's possible I just happened to lu
  • Built a Hylafax http://hylafax.org/ [hylafax.org] system on top of the latest v6.1, LFS http://linuxfromscratch.org./ [linuxfromscratch.org.]
    Details:
    3GHz Intel Pentium 4 Processor, 1Gb RAM
    11,878.40 BogoMIPS Total, 250Gb Hard Drive
        GCC 3.4.3
        Samba 3.0.14a
        HylaFAX 4.2.1

    Gotta say it's way ahead of expectations.
    I won't touch another distro now for my mission critical.
    Although, Knoppix, http://www.knoppix.org/ [knoppix.org] and Ubuntoo, http://www.ubuntulinux.org/ [ubuntulinux.org] are great "insert CD and run" distros, for workstations.
    Working with SlackWare seems effortless also, http://www.slackware.org/ [slackware.org].

    Was fortunate enough to meet the fine gent who started the LFS project: Gerard Beekmans
    Highly recommended support for the project, even if it's just $5 for a beer via donations :->, http://www.linuxfromscratch.org/hints/contribute.h tml [linuxfromscratch.org] or a much needed "hints" writeup, http://www.linuxfromscratch.org/hints/ [linuxfromscratch.org].

  • I've been using FC4 and was using FC3 before that. I've had no problems with the 2.6.12 kernel that I'm currently using. No stability issues. I was having a few issues before this specific revision but sometimes its also hard to figure out if its KDE, GNOME, a specific application, or the kernel. I honestly think that at this point in time its just as stable as 2.4.x but keep in mind that any operating system can freeze at any time for something that nobody has seen before so switching to 2.6.x at this poin

  • No problem here - I have bog standard hardware and everything works perfectly.

    ---
    jon_edwards@spanners4us.com
  • changing to 2.6.13 later this week

    Uptime------------System----Boot up

    62 days, 14:02:05 Linux 2.6.9 Wed Jun 29 19:06:18 2005
    90 days, 18:28:05 Linux 2.6.9 Tue Mar 22 20:54:33 2005
    28 days, 08:41:03 Linux 2.6.9 Tue Feb 22 12:02:50 2005
    31 days, 14:05:41 Linux 2.6.9 Fri Jan 21 19:42:01 2005
    49 days, 07:58:12 Linux 2.6.9 Fri Dec 3 11:40:11 2004
    31 days, 06:14:18 Linux 2.6.8-rc2 Mon Oct 4 19:32:10 2004
    39 days, 16:12:23 Linux 2.6.8-rc2 Thu Aug 26 03:14:37 2004
    33 days, 16:05:05 Lin
  • If you want a balance between bleeding edge and stability use the last bugfix release of the previous 2.6 kernel.

    i.e. use kernel 2.6.12.6 because it should be more stable than the new 2.6.13.

  • ...for last 2 months or so, like for a salvation. I want to make that auth system using iButtons [ibutton.com] and .13 is the first to include full, useful system. (1-wire protocol was present in the kernels before, and in userspace even earlier, but only with .13 it's mature enough to be usable.)

  • It's stable (Score:3, Informative)

    by anpe ( 217106 ) on Wednesday August 31, 2005 @03:48AM (#13443744)
    Just don't upgrade _right now_ to bleeding edge a bleeding edge kernel (2.6.13 in that case). Wait for the dust to settle (two or three weeks) and upgrade. I've done that since early 2.6.0 releases and it works like charm.
    Note: You can install triple dotted releases (2.6.x.y asap as they only contain minor upgrades or security fixes)
  • I had freezes in 2.6 a long time ago. It was caused by a combination of nforce and IO-APIC. All I had to do what disable APIC in the kernel until the next kernel version was released with a fix. Other than that I've had no problems with 2.6.

    Stop being paranoid. If you configure the kernel correctly it wont freeze. And for all the great things 2.6 has to offer I don't know why anyone would choose to use an older kernel if they have a choice.

  • Rule Number One: for any software, hardware, computer, vehicle, anything -- never buy version .0 unless you are willing to suffer instabilities.

    Rule Number Two: See Rule Number One

    You said you suffered instabilities in the 2.6.0 release. No .. duh!

    I thought you would be more concerned with their change in practice to do away with the odd/even stable/development model that they used up to kernel 2.4. As I understand it, now all the development problems are rolled into the kernel intended for public use

  • 2.6.13 is very stable (Score:3, Informative)

    by objorkum ( 695401 ) on Wednesday August 31, 2005 @06:49AM (#13444271) Homepage
    I am using a plain vanilla kernel (2.6.13) on my Slackware 10.1 system, and it's very very stable. The earliest 2.6 kernels were a bit unstable for me, but since 2.6.8 or so, they have been very very stable. I notice that my self-configured 2.6.13 kernel is faster than the Slackware vanilla kernel (2.4.29). GNOME responses faster to my actions, for example.

  • oh my god (Score:3, Funny)

    by XO ( 250276 ) <blade.eric@NospAM.gmail.com> on Wednesday August 31, 2005 @07:18AM (#13444353) Homepage Journal
    self-sufficient linux user for 4 years doing Linux from Scratch, and you don't know how to flip between kernels?

    I ran continuously from 2.5.56 (after my SCSI driver started working again) until 2.6.13 ? or so that was out about 4 weeks ago.

    Had virtually no kernel related problems.

    Then I switched to Windows.

    Life is much easier now.

  • I switched from 2.4.x to 2.6.7 when it came out, and haven't had a single problem. Some of the newer ones seem iffy, but 2.6.7 is rock-solid as far as I can tell.

    I still prefer to use 2.4.x series kernels in servers, just to have that extra insurance (the servers I run have very little need for the features in 2.6.7), but desktop machines get 2.6.7 all the way.

  • So far, so good, though I recommend waiting for the stable patch series to iron out any brown-paper-bag bugs:

    root@prodserver:~$ uptime
      08:22:38 up 89 days, 21:59, 2 users, load average: 0.08, 0.06, 0.01
    root@prodserver:~$ uname -a
    Linux nli-aus-srv01 2.6.11.11 #1 SMP Thu Jun 2 09:36:16 CDT 2005 i686 GNU/Linux
  • I've just updated three machines to 2.6.13, and on one of them, the kernel ate all ReiserFS partitions for breakfast (i.e. at the time of booting). It was a small corruption at the top of the filesystem trees, and easy to recover (reiserfsck --rebuild-tree and a little manual work), but still not nice. I'm pretty sure it had to do with a bug in the EPIA MII BIOS, because my other machines were unaffected. VIA even provided a beta BIOS update [viaarena.com] and now all is well.

    The machine has had related problems with e

  • I have several machines with older 2.6 kernels and more than a year of uptime so for me it's been stable. I also have a few that I just upgraded to 2.6.12.4 recently that seem to be working fine. You really do need to make sure your modutils and maybe other user land tools (does anyone have a list?) is up to date and be cautious about enabling some of the new features but I havn't run into any unresolvable problems since switching. One annoyance is that you cannot just slap a kernel on a floppy and boot
  • I was using a 2.4 kernel and some nVidia drivers. I upgraded to Fedora Core 3, which had the 2.6.9 kernel I believe. Configuring the nVidia drivers was a nightmare, I ended up using an unofficial version patched by someone, version 6111 I think.

    The first thing I noticed with 2.6 is that something changed for the worse in memory management - closing programs which eat up a lot of RAM make the system completely unresponsive for several seconds. Then I started getting crashes (kernel oops) when doing OpenGL
  • I am a Gentoo user and my linux box is my baby. I am very careful about what I load and am constantly making sure that I use what I have installed and only really install things that I feel confident will not aversely affect the performance and stability of my system. I stuck with 2.4 until recently, diligently upgrading when bugfixes were released. I had my kernel configuration down to a science. Less than the 20 minutes you described. All this new stuff--ALSA, udev, ACPI, new module loading tools, et
  • The biggest problems I have with using a very-recent kerenl is that sometimes some of the not-included modules which I need to compile from source just don't seem to like my newer kernel. At the moment I'm having a bugger of a time getting 'unionfs' to compile with a 2.6.13 kernel, and I do remember at one time that my NVidia drivers didn't like certain 2.6 kernels either.

    Once something compiles though, I generally don't have kernel problems with common hardware.

Slashdot Top Deals

If you think the system is working, ask someone who's waiting for a prompt.

Close