Home > Linux > Linux Kernel 2.6.7 Released Linux Kernel 2.6.7 Released Submitted by Davide Abrigo 2004-06-16 Linux 49 Comments Kernel 2.6.7 released today. Grab it from kernel.org or from one mirror. Read the changelog here. About The Author Eugenia Loli Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker. Follow me on Twitter @EugeniaLoli 49 Comments 2004-06-16 9:59 am After the rash of security related releases (see 2.6.1 – 2.6.4 or so), it’s nice to see a release that is just for standard bug fixing. It also looks like they have a patch in that should improve battery life, so it laptops will possibly last even longer than when using other operating systems (ACPI/APM patch). Hopefully we have seen the last of the remote-root mremap() vulnerability (and hopefully there aren’t any similar aremap, bremap, etc functions with similar undiscovered remote root vulnerabilites). I feel like I can sleep better with my production servers on 2.6.x now, at least. 2004-06-16 10:15 am That list is so long….I am not into this stuff but does anyone know if the partitioning bug has been fixed? You know, the one that stuffs up your partition table. 2004-06-16 10:18 am Actually, I don’t think that was a bug. Think about it, what better way to get people to stop using Windows, than to NOT ALLOW it? Hmm? You can conveniently later say “I dropped a cheetoh on the keyboard causing a typo in the master boot record calculation recursive memoization function” to disguise your evil malice. It kinds of reminds me of the days when Linux had almost zero support for most modems (Linux users liked to call them winmodems to show their contempt). My personal theory was that this was done so that no one could go online and complain about their experience with Linux. Can’t get the internet working? Can’t complain, problem solved. Then again, maybe I have been hanging out at slashdot for too long. 2004-06-16 10:19 am in parted. 2004-06-16 10:34 am you don’t need linux at ALL for that bug to appear – i’ve been helping people on IRC with exactly the same problem, with a new hard disk, with exactly the same cure (i.e. forcing LBA in the BIOS) and “winmodem” isn’t a term used for contempt, it’s a term used for… well… http://www.softsaver.com/WebArt/3compcmciamodem.jpg 2004-06-16 10:56 am This too is a security related release! It fixes the floating-point-exception handling, which could lead to DOS-attacks. 2004-06-16 11:00 am Does this release fix the recent kernel bug mentioned at: http://linuxreviews.org/news/2004-06-11_kernel_crash/index.html 2004-06-16 11:05 am That’s a ton of USB patches — hopefully my Prism2 USB wireless card will work now … 2004-06-16 11:31 am Yes – the FPU exploit is fixed. 2004-06-16 11:39 am > It kinds of reminds me of the days when Linux had almost > zero support for most modems (Linux users liked to call > them winmodems to show their contempt). Linux has had support for modems since before version 1.0. What you’re refering to is “software modems”. Regular hardware modems follow well defined standards for modem I/O and contain their own CPU to handle all the internal low level I/O operations that need to be done to support modem transfers. Software modems, OTOH, try to be cheaper by not including a CPU. Your own PC’s CPU does all low level I/O operations through special drivers that were written for Windows. Lucent, who I believe made the first software modem, also coined the term winmodem. Linux users had no say in this. The hardware portion of a winmodem uses a non-standard undisclosed interface. From my understanding, no two software modems from different manufacturers had the same interface. This made it very difficult for people to write winmodem device drivers for Linux. Linux users avoided these types of modems since they were useless under Linux. BTW, many Windows users avoided winmodems too. The Windows drivers tended to be buggy, and because they ran on your PC’s CPU, they also affected the performance of your machine. 2004-06-16 11:40 am Er, there was a brand or model name called “Winmodem” from wince the name came from. It was software based. All hardware modems, to my knowledge, are supported under Linux and have been for a very long time (since 2.0 at least) – it was the software based modems that gave Linux developers hell. The hardware companies refused to release specs, so much had to be reversed engineered. “Winmodems” were not called that out of contempt, it was just a generic name for software based modems since “Winmodem” was, perhaps still is, a very popular brand/name. Linux’s issues with the softmodems was not Linux’s fault, the developers were rarely helped out by the makers. It would appear the situation is better now, but I’ve been on broadband for years now and hadn’t had to mess with dialup in quite sometime. evidence for what I’ve said – http://www.usr.com/support/overview-template.asp?prod=s-win US Robotics apparently had a product line called “Winmodem,” likely the one I remember. 2004-06-16 11:48 am Hopefully we have seen the last of the remote-root mremap() vulnerability The mremap vulnerability was a local exploit, not a remote one. 2004-06-16 11:56 am Hopefully we have seen the last of the remote-root mremap() vulnerability The mremap vulnerability was a local exploit, not a remote one. Don’t try, he does not or does not want to understand the difference, as seen in many discussions before. 2004-06-16 12:11 pm Not that I think it’s a bad thing 2004-06-16 12:35 pm Sheesh – that is one ‘phat’ changelog. BTW, “anonymous”, this wasn’t exactly just a bug-fix release. It actually has a lot of major new things in it, like the scheduling domains, the anonVM stuff, big ALSA update, and much more. Obviously, a lot of the major stuff went into -rc1, which was some weeks ago now. Anyway, it will be interesting to see what comes next. If 2.6 actually starts settling down now, then we can be sure that 2.7 is not far away! 2004-06-16 1:04 pm I should have known that I got trolled when I read this statement: I feel like I can sleep better with my production servers on 2.6.x now, at least. Who is using 2.6 on a production server? Commercial distros still default to 2.4. 2.6 has been plenty stable for me but I don’t quite think it should be on a production server until more bugs are ironed out, and neither do most. I highly doubt you are even running a home server, nevermind a production server. 2004-06-16 1:29 pm Since a regular old modem is nothing more than a serial interface as far as the software is concerned, Linux has supported modems ever since it supported serial interfaces. A regular old hardware modem doesn’t require any driver at all beyond simple serial support. Only Linus himself and possibly a few early contributors will remember a time when Linux didn’t even support a simple serial interface, and the instant that support was added, the VAST majority of modems in the world worked on linux. The more recent software modems (called winmodems by their MANUFACTURERS, not by linux users) on the other hand need drivers written, just like a lot of other hardware does. They offload much of the work to the CPU that regular modems did themselves, making winmodems cheaper to produce. If Linux ever had near zero modem support, it was more then a decade ago, back in the pre-1.0 days. Very very earily in the pre-1.0 days. 2004-06-16 1:34 pm I hope this is better than the 2.6.5 release. I downloaded it that for my FC2 install and my PC would not boot. It would freeze when it tried to load the USB controller driver. I had to go back to using the 2.6.4 kernel. This was on a nForce2 mobo. 2004-06-16 1:37 pm Thankfully they fixed the hard-disc spin-down on reboot issue (I’ve checked the changelogs). 2004-06-16 2:43 pm “The more recent software modems (called winmodems by their MANUFACTURERS, not by linux users) on the other hand need drivers written, just like a lot of other hardware does. They offload much of the work to the CPU that regular modems did themselves, making winmodems cheaper to produce. ” He should have also informed you that a quality modem is faster than a winmodem. Personally, I feel embarrassed for ever buying a winmodem and if I used dial-up now I’d be getting a nice US Robotics. 2004-06-16 3:20 pm Hmmm…is anyone getting a kernel panic with kernel-2.6.7.x-rc1.1 (2.1, 3.1) on Mandrake 10? 2004-06-16 3:48 pm Look thruogh your kernel configuration, you are probably missing something. include things you’re not sure about, keep in mind not to make the IDE/ATA and the filesystem (for your root volume) drivers as modules (you need them to boot, don’t you?). then recompile and try again. 2004-06-16 4:46 pm I’m somewhat at low risk with one of my servers, but I’ve been waiting for the “next” release of the kernels that have the FPU exploit fixed. It seems that this one has it done, but what about the 2.4 series? This server is running Debian/stable, and I want to stick with the 2.4 series (dont want to compile binutils and such). Anyway — Any idea which 2.4 kernel version has the FPU fix, or which is slated to have it? 2004-06-16 5:07 pm “Anyway — Any idea which 2.4 kernel version has the FPU fix, or which is slated to have it?” many of the distros have released a version that fixes this thing. the current 2.4 release doesnt have it afaik 2004-06-16 5:09 pm “Partitioning By OSNews Reader (IP: 144.139.86.—) – Posted on 2004-06-16 10:15:06 That list is so long….I am not into this stuff but does anyone know if the partitioning bug has been fixed? You know, the one that stuffs up your partition table.” its NOT a kernel bug. the kernel doesnt handle partitioning any more. its the responsibility of parted and boot loaders. so the bug is in parted here which tries to “fix” the table which windows doesnt recognise 2004-06-16 5:11 pm “Timer+FPU save/restore bug fixed? By Patrick Mackinlay (IP: —.uknet.spacesurfer.com) – Posted on 2004-06-16 11:00:07 Does this release fix the recent kernel bug mentioned at: http://linuxreviews.org/news/2004-06-11_kernel_crash/index.html ” yes. it does 2004-06-16 5:31 pm Well, I suppose it’s more a Fedora problem or an issue with your specific motherboard rather than a problem with all 2.6.5 + nForce2 boards. I’ve got two boards here running Linux with the lastest vanilla kernel (2.6.6, soon 2.6.7 but previously 2.6.5) and never encountered issues with USB. I don’t really know how Fedora handles kernel modules as I never used it but you should select the OCHI driver instead of the UHCI if you can. 2004-06-16 5:56 pm Thanks everyone on the kernel team! I have ~20 servers running 2.6 now since last November. With very minor exceptions (a buggy amd-64 board), everything has been smooth and 2.6.7 seems to be the best kernel yet. 2004-06-16 6:10 pm Linux users liked to call them winmodems to show their contempt Actually, the manufacturers called them winmodems. Thusly, that what they were. They were dumb boards which pushed most of the real work onto the CPU. As drivers were commonly only available for windows, the name aptly fit. You seem to be forgetting the markets toe into winprinters too. It was the same concept. Put a dumb, brainless printer together, make the CPU do all of the work, and only make drivers available for the windows platform. Again, the name was driven my the market and not by Linux Zealots. The name fit. I think you’ve been misinformed. 2004-06-16 7:00 pm Do the “mkinitrd” command as root. It will make a image with the sutible modlues in. Speaking of modules will kernel 2.8 or 3.0 become more modular ? 2004-06-16 7:09 pm Are you just repeating buzzwords that you know nothing about? “Speaking of modules will kernel 2.8 or 3.0 become more modular ?” What does that mean? Microkernel? No. Let us all know what you think “modular” means, please. Love to hear it. 2004-06-16 7:31 pm Modular, means to me. 1> someting to is not built in one unit. 2> Something that has its functions/drivers in separate parts that can be changed with ease. I maybe wrong in my use of words because I am only human. Please by all means give me the correct use of the word. I am not flamming the Linux kernel as its my main OS that has, for a long time given me trouble free operation of my computer. (Plus it come fire proof as standard) 2004-06-16 7:50 pm It’s nice to see more support for the VIA Nehemiah and the other chips on my board (a mini-itx board). Longhaul support means my computer will use less electricity as the CPU doesn’t have to run at full speed all the time regardless of how much work it’s actually doing (as it did before the update). Significant updates to the sound chip and USB controller drivers are nice. But what I am really waiting for with baited breath is the graphics chips (Unichrome) stuff, once that’s in there, I can run with full hardware support on a vanilla kernel, which would be very nice, as right now there are only patches for fairly old revisions of 2.6 for my graphics hardware, which then of course do not include the new longhaul and other stuff. These patches have been around a while — I wish I knew why it was taking to long to put them in. Erik 2004-06-16 8:03 pm Linux users liked to call them winmodems to show their contempt Trust me. It wasn’t just Linux users. Winmodems were and are to this very day, contemptible, evil things. 2004-06-16 8:03 pm Aaron, don’t bother replying. These people know about Linux’s defficiencies, and do nothing to fix them. They also like to pretend they don’t exist. If you mention a problem with Linux (like it’s multiple remote-root vulnerabilites), they will not try to address the issue, they will only make personal attacks at you in a vainglorious attempt to change the subject before anyone notices their precious emperor has no clothes. Truly, sad times we live in. You disagree with the group-think, and you are a troll, you are guided from the herd, and verbally executed. I guess that GPL isn’t “free as in thought.” 2004-06-16 8:14 pm “Linux has had support for modems since before version 1.0. What you’re refering to is “software modems”.” Yep. If you had an internal modem, you’d had a fair chance you had one of those. If you had an external however, you were almost always lucky. Those worked, and they worked fine. It doesn’t apply on cable “modems” or ADSL “modems” because those work totally different, and the question is wether those could even be called modems or wether router is a better name. 2004-06-16 8:19 pm “He should have also informed you that a quality modem is faster than a winmodem.” True, a software modem requires drivers, the software does the work. A hardware modem does not. It’s like RTL’s vs 3C’s. You pay for quality Why isn’t Linux kernel 2.6.x modular enough? 2004-06-16 8:34 pm “Why isn’t Linux kernel 2.6.x modular enough?” what exactly do you mean by that. 2004-06-16 8:36 pm I think he is implying: “Why isn’t Linux kernel 2.6.x modular enough?” 2004-06-16 8:37 pm Trust me. It wasn’t just Linux users. Winmodems were and are to this very day, contemptible, evil things. You can say that again. I only tried a winmodem once back in 1997. I had an hardware modem (USR 33.6k) but I wanted to “upgrade” to 56k. The stupid softmodem kept disconnecting me while I was doing heavy work on my computer because it didn’t had enough dedicated CPU time. Fortunately, I was able to get a refund and bought a real 56k modem instead. I’m now on broadband since 2000 but I won’t ever recommend a softmodem, not even for someone with a multi-GHz computer (I had a P1 200 back in the day) or my worst enemy. 2004-06-16 8:44 pm “I think he is implying: “Why isn’t Linux kernel 2.6.x modular enough?”” you said the exact same thing. 2004-06-16 8:45 pm Trolling I see. “Aaron, don’t bother replying. These people know about Linux’s defficiencies, and do nothing to fix them.” Who’s “they”? The people actually doing the work, or some “mythical” person you pulled out of nowere? “They also like to pretend they don’t exist. If you mention a problem with Linux (like it’s multiple remote-root vulnerabilites), they will not try to address the issue,” Despite proof to the contrary. “they will only make personal attacks at you in a vainglorious attempt to change the subject before anyone notices their precious emperor has no clothes.” Here we go with “they” again. “Truly, sad times we live in. You disagree with the group-think, and you are a troll, you are guided from the herd, and verbally executed. I guess that GPL isn’t “free as in thought.” Truelly sad times when people can come on a forum and dispense FUD, with the expectation that their words will be accepted without question. Sounds like your no freer than the people you accuse. Anyway the original post was talking about “modular” and the reply requested clarification (could have been more polite, but then everyone here should have thick skins by now). 2004-06-16 8:48 pm >>>Why isn’t Linux kernel 2.6.x modular enough? <-; IMHO its just as modular as it needs to be. I heard that the kernel is not modular enough. When compiling the kernels most of drivers are as modules that can be loaded and unloaded. I have been reading that modules should not be loaded on kernel space but user space. My first reaction was the switch back and forth would cause a slight slow down. Am I right in thinking that this goes back to the old micro vs mono ? 2004-06-16 9:00 pm ” I have been reading that modules should not be loaded on kernel space but user space. My first reaction was the switch back and forth would cause a slight slow down. Am I right in thinking that this goes back to the old micro vs mono ? ” kernel modules cannot be safely loaded in user space. its very racy unless you have a micro kernel but then everything including the fs would be a module so yes you are right 2004-06-17 12:11 am To see how modular linux can be, you just need to run a generic initrd kernel from your distro maker. Running Debian’s 2.6 kernels I have about 50 or so modules loaded at once, providing very nearly all of the kernel’s services. Right now though, on a home built kernel, I have 6 loaded, and only two of them are loaded at boot. If linux were to become significantly more modular than it is now, it would be a lot more than just a point release… 2004-06-17 2:53 am If you mention a problem with Linux (like it’s multiple remote-root vulnerabilites)… which don’t exist at all. mremap was a local, I repeat, local exploit. Local being the opposite of remote. Maybe you just don’t understand what the word remote means. Please get a clue before you make yourself look even more foolish. 2004-06-17 4:02 am Had problems with my Dual AMD rig with the 2.6.5 and 2.6.6 kernels giving kernel panics and the likes and hap hazzard USB support (major changes to USB subsystem I believe in 2.6.5). Fell back to 2.6.4 and have been runing smoothly since. I will try 2.6.7 to see what has changed as ASLA in 2.6.4 is far from perfect for me. 2004-06-17 3:20 pm What’s that a week from the vunarbility becomeing known and it being fixed. Microsoft take note! http://www.codefish.net.au/modules.php?op=modload&name=News&file=ar… 2004-06-17 5:27 pm ” If you mention a problem with Linux (like it’s multiple remote-root vulnerabilites)… which don’t exist at all.” Actually, there have been remote exploits in the past. As far as I know, they were patched very quickly. Local exploits are nasty, but let’s be honest…if someone has physical access to your machine, you’re boned anyway.