Linux Kernel 2.6.4 Released

justinarthur writes "The Linux kernel version 2.6.4 has been released at 03:16 UTC. Included in the changes from version 2.6.3 are fixes to XFS support, Wide Area Networking, USB connectivity, and IEEE1394 connectivity. To download a copy, it is recommended that one utilizes a Linux Kernel Archives mirror. Linus Torvalds' announcement to the Linux Kernel Mailing list concerning this release is available here." Reader k-zed points out that Linux 1.0 was released in March 1994, ten years ago.

Comments

Slackers.

[grub]

Hmm.. I don't see it on ftp.sco.com yet. What lousy service for $699.

Re:Slackers.

[pe1rxq]

You better not tell them you were once a customer...
they seem to have developped a habit of suing them.

Jeroen

10 years?

[110010001000]

10 years and that guy is only on version 2?

Re:10 years?

[BabyDave]

For comparison, in 10 years Microsoft went from Windows 3.0 to Windows 2000! Clear proof that Microsoft's development process is far superior.

Re:10 years?

[happyfrogcow]

Finally, someone got a versioning scheme that held true and made sense. If only all "Version 2" software worked this well, we'd be blessed and probably have robots serving us Coronas and margaritas while we graze on a 7 layer dip and chips.

Re:10 years?

[jd]

Beats the GPL. That's been around 20-30 years, and is also on version 2.

2 is clearly the tar-baby of versions.

Re:10 years?

[TheTomcat]

According to this , GPL v1.0 was released in 1989. RMS released the GNU Manifesto in 1985.

So, it HAS been around a long time, but not 20-30 years.

(and yes, I know it was a joke).

S

Yes but...

[Bilange]

Does it run Linux?

(Ok, sorry. I know its not funny anymore.)

My 486 laptop...

[Anonymous Coward]

... hasn't even gotten finished compiling the last 2.6 kernel release *grumbling* *adding yet another patch to my to-do list*

Just when...

[Cytlid]

... I was thinking "I don't need this kernel upgrade, 2.6.3 has been working great for me..." I find in the changelog:

[IRDA]: Add stir4200 driver.

doh... finally added support for one of my usb-irda dongles.

Damn.

Re:Just when...

[damballah]

Speaking of changelogs, this is the funniest one I've found so far for 2.6.4:

[PATCH] kthread primitive

From: Rusty Russell

These two patches provide the framework for stopping kernel threads to
allow hotplug CPU. This one just adds kthread.c and kthread.h, next
one uses it.

Most importantly, adds a Monty Python quote to the kernel.

Re:Just when...

[FueledByRamen]

Haha, I have a runner up here:

Maurice van der Stee noted that he got an oops on a HPFS filesystem when
saving an edited file..

<stares at the code>
<blinks>
<wonders whereTF do we assign hpfs1_i and hpfs2_i if both inodes are non-NULL>
<finds the patch in question>
<stares at jgarzik>

This fixes it. That, BTW, means that *nobody* had ever tried to use
hpfs r/w since 2.5.3-pre3.

Poor HPFS. Nobody ever loved you...

Re:Just when...

[Winter]

Could not find it in 2.6.4, but a patch to kernel/kthread.c to 2.6.1-rc1-bk6 shows these lines:

+/* "to look upon me as her own dad -- in a very real, and legally
+ binding sense." - Michael Palin */

Which is from the Holy Grail

Did you know that..

[Anonymous Coward]

The name of this release amongst the core developers was "Heathen Chemistry.". Alan Cox came up with it - it's was inside joke about british pop/rock phenomenon.

Ack!

[MalaclypseTheYounger]

Oh no! Someone leaked their source! Call Microsoft, maybe they can help track down who leaked this to the internet...

oh, wait... nevermind.

Why yes it does!

[Medievalist]

usermode linux runs linux on linux.

So, the answer is yes.

new kernel

[tacocat]

Odd. I'm still stuck on my 2.4.xx version. I tried to upgrade a few distros to 2.6 and things didn't go very well (kernel panic)

It seems to me that the number of users who have picked up 2.6 x compared to the number that picked up 2.4 from 2.2 has greatly diminished on many of the distro mailing lists. From this it seems that either the migration is uglier than anticipated, or that more people are just willing to sit back and wait for their distro to provide them with all their needs.

Who will be the first to ship kernel 2.6 by default?

Re:new kernel

[Anonymous Coward]

Mandrake 10 is (look for the story from yesterday) Also you can get 2.6 from gentoo and debian, though not default

Re:new kernel

[peterf]

Some people are waiting for their distros to ship 2.6, but I think 2.4 is works for most users, there is no urgent need to upgrade.

This doesn't mean 2.6 won't be a success, I am sure it will be, but I think it means we have left the 'revolution' and joined the 'evolution'.

Re:new kernel

[TheAcousticMotrbiker]

The Knoppix that came with the C'T from a month ago already had a 2.6 kernel.

Suse 9.0 came with a 2.6 kernel (though neither had them as a default)

Re:new kernel

[DarkBlackFox]

2.6 does offer a number of improvements to 2.4 (as it should, being a later release), including native ATAPI (non-SCSI emulated) CDRW support (which a lot of new users get borked on and complain about), non-dangerous NTFS support (including writing, which is huge when migrating from Windows), much easier build process (simply 'make'), easier module selection layout, faster boot times, etc. etc.

I was happy as a clam switching to 2.6, and haven't looked back to 2.4 much since.

Re:new kernel

[damballah]

For MandrakeLinux 10, it is the default kernel. They also provide the latest from the 2.4 series, but you'll have to install that by hand. By "integration", they probably mean dropping devfsd in favor of udev... " I assumed the 2.6 was some addon I have to waste precious brain cells and time trying to figrure out how to install." Mandrake has had RPMs for the 2.6 series for some time now.

Re:new kernel

[MeanSolutions]

The problem with Linux lately is the annoying lack of stability and binary backward compatability.

I have noticed very little, if anything, breaking when I moved from 2.4.21 to 2.6.x on SuSE 9.0, a distro compiled for and shipped with a 2.4 kernel.


For example, the new threading libraries break all sorts of applications until you recompile the apps against the new library. This is particularly painful with commercial applications or for companies that need to provide support.

Sun, IBM, HP, etc have all been able to enhance the functionality of their proprietary Unix systems without breaking binary compatability. It is a shame that the linux kernel people do not care to do so.


That has absolutely nothing to do with the kernel "Linux". That has to do with updates to glibc and pthreads libraries. There is nothing preventing you from running a glibc2.1 with a kernel 2.0.5, 2.2.10, 2.4.18 or 2.6.4 running on top of it.

Re:new kernel

[TheTomcat]

- A responsive scheduler and preemptible kernel (sure, most users won't have a clue what it DOES, but they'll notice the difference).
- ALSA (in-kernel) -- for the first time I can remember on Linux, the mixer (non OpenSoundSystem) labels my inputs/outputs properly.

S

wow!

[matticus]

this is the first time i've installed the kernel and had it running before the slashdot announcement!
i just checked the new one after lunch. blasted centrino ultralights need all this new stuff in them.
I think i'll go celebrate "I beat slashdot's unnecessary kernel release announcement day!"

Reiser4

[Anonymous Coward]

Wonder if we'll see reiser4 in 2.6.
You can test it now, but it is very experimental.

Maybee they'll merge it with 2.7

Augh!

[asit ler]

I haven't gotten 2.6.3 compiled yet, and here comes 2.6.4. Hell, I'm still running 2.6.0-gentoo. What's with this heightened release schedule? I mean, gcc is only so fast on my machine.

Re:remove SCO code

[the_mad_poster]

i guess linus wouldn't make a statement about it now, since there's the lawsuit going on

I think the fact that SCO hasn't actually shown it bears more weight on it having not been removed. It's hard to remove something that isn't there, and it effectively isn't there unless SCO can actually show it to somebody.

Fixes my ICH5 SATA Problems

[xheliox]

2.6.4-rc1 ChangeLog:

[libata] catch, and ack, spurious DMA interrupts

Hardware issue on Intel ICH5 requires an additional ack sequence over and above the normal IDE DMA interrupt ack requirements. Issue described in post to freebsd list: http://www.mail-archive.com/freebsd-stable@freebsd .org/msg58421.html

Since the bug workaround only requires a single additional PIO or MMIO read in the interrupt handler, it is applied to all chipsets using the standard libata interrupt handler.

Credit for research the issue, creating the patch, and testing the patch all go to Jon Burgess.
---------

Woo, this is very exciting. If you had problems with SATA & ICH5... this probably fixes those problems.

just add love...

[encrypted]

all it needs now is some love and it'll be ready for my machine.

did you report it?

[Anonymous Coward]

Copy down the numbers from the kernel panic.
I know it's a pain, but we really need this.
If you're terribly lazy, just get EIP, ESP,
and any names you see.

Mail that to linux-kernel@vger.kernel.org and
expect a few questions about your hardware.

That's not so difficult, is it? This gets the
bug fixed so that the next release will run on
your system.

Re:did you report it?

[pe1rxq]

Without the symbol map the EIP and ESP values are pretty useless since they vary on compile time options (and even the compiler).

What is usefull is the EIP (and call trace) with together with the symbol table, or easier the 2.6 kernels have a build in back trace that spits out the function names itself. (If you want some fun do it yourself for once with the call trace and a symbol table, bonus points for doing hex math on a piece of paper instead of a calculator)
Just write down this list of functionnames and send it, that way the developpers immediatly known wath code path your pc took before going boom.

Jeroen

Re:In 94, I was using Windows 3.1

[jcupitt65]

Linux is just the kernel. You'll find the NT kernel hasn't changed that dramatically recently either.

Maybe you mean the desktop experience? That's provided by KDE/GNOME/fluxbox whatever, and it's very clear what innovation is going on there if you look at KDE 3.2 vs KDE 1 (for example).

Think of it as version 26

[Chemisor]

In the OSS world, major releases are counted in the minor numbers, so 2.6 is what a commercial company would have called 26.

Re:Think of it as version 26

[Frit Mock]

26th release is the wrong answer! ;)

First of all, for Linux uneven numbers are development-versions.
Second the 1.X versions stopped at 1.3
Third, version 1.0 surely was not the 10th major release ... not even real swedish programmers begin counting at 0, like their programs do ;)

However, this truly is the forth Service Pack for the Linux 2.6 ... within a few month they fix more, than M$ ever fixed in w2k ;)

Re:Think of it as version 26

[Apiakun]

Wait...

There's a Forth Service pack for 2.6?

Sweet!

Re:In 94, I was using Windows 3.1

[Astatine]

Aren't you confusing kernel and distribution? Microsoft Windows is like a distribution (it's a complete running system). How different are Slackware 1 and Slackware 9 for instance?

If you looked at what's happened to the NT kernel during those 10 years, I reckon it would also look like "10 years of incremental patches". Apart from the graphics renderer turning up in it, that is ;-)

Re:In 94, I was using Windows 3.1

[FictionPimp]

Its kinda funny, I run windows on one of my machines, and I'm constantly installing patches. Not to mention my 3 BIG patches the ones that cost me a hundred or so dallors. My 98 - 2000 patch and my 2000 to xp patch. The process is still the same, the only difference is the linux changes come bit by bit instead of in a bulk jump on a cd. Personally, i would rather get my updates in small increments. That why I can pick and choose what updates I want. If 2.6.4 doesnt add anything I need, maybe I'll hold off to the next one. And besides, it beats spending cash.

Re:In 94, I was using Windows 3.1

[JimDabell]

Now, I'm using XP, and it's easy to see how 10 years of development have consistently innovated-not in a direction I'd like, and not in a safe or secure manner, but it's innovation.

Incremental improvements inspired by others is not innovation. What great new features have Microsoft ever introduced to the world? It has to be more than a couple to claim the status of "consistently innovating".

What has Linux got but 10 years of incremental patches?

Things like ReiserFS? Major sections of the kernel have been completely rewritten since 1.0. The scheduler, the module loading system, the /dev handling (static /dev to devfs to udev), the network subsystem. Anybody even remotely familiar with the kernel wouldn't make the claim you are making.

I feel the "if it's open source, it's perfect" ideology

That's a straw man argument. People don't claim that open source automatically makes something perfect.

is the reason Windows is still the desktop OS of choice.

Desktop OS of choice for some people. It's certainly not the desktop OS choice for me.

Linux fanatics need to be able to step back and look objectively at Linux's many flaws

Zealots of any kind aren't objective. But people who use and develop Linux at home or work are not automatically zealots as you seem to be implying.

Re:In 94, I was using Windows 3.1

[Angstroem]

May I ask in what way those 10 years of development "have consistently innovated"?

What else than "incremental patches" does Microsoft deliver, especially in the days where there is no week with new IE and OE exploits being announced and eventually (after months) being fixed?

The reason why "Windows is still the desktop OS of choice" is just because it comes preinstalled with any vanilla PC you can buy out there. Because it will run the games people copy from their friends. Because it runs Microsoft Office.

If Microsoft released Office for Linux I'm more than sure that numerous offices will switch to Linux and if it's just to escape the virus race which cost them plenty of time, money, and nerves.

Partitioning md devices

[rpozz]

Finally they've included mdpart. This means anyone with a SATA RAID motherboard can use its full potential. Excellent :-)

I tried it last night..

[schon]

I don't even know of anyone running 2.4 anymore actually. I mean, why not upgrade?

I finally got around to compiling 2.6.3 last night; ran into some issues.

its much more responsive than 2.4 for desktop use

The desktop is definitely more responsive, but (at least for me) at the expense of everything else. MPlayer, xmms, and anything that's remotely timing-intensive is unusable (xmms actually skips while playing MP3s, and Mplayer prints the message "Your computer is TOO SLOW to play this file" when playing anything I've got. Note that everything works fine under 2.4.)

I went through the various mailing lists looking for suggestions, with no luck; every suggestion is OK (checked drive DMA, kernel settings, X nice level, etc.) - interestingly enough, one post I read said to try glxgears.. I did, and it runs better under 2.6 - constant frame rate, regardless of what else I'm doing, whereas in 2.4, even moving the mouse drops the frame rate.)

So it's back to 2.4 for me. I'll probably try 2.6.4, to see if the situation has improved, but for the mean time, I'll stick with 2.4.

How many have upgraded and then gone back?

[steelerguy]

So I went to the 2.6 series when they first came out. I was very happy with it at work, there was a noticable bump in speed...mainly in starting applications.

At home it was another story. Sure the speed increases I noticed at work were still there but there were some fairly large problems.

First, neither my DVD reader or CD burner were assigned /dev entries. So no reading CD's, DVD, or writing CD's. I honestly don't do it that much anyway, so I didn't spend a lot of time trouble shooting it. Plus after a day at work trouble shooting problems I don't feel like doing it at home.

Second, I have not been able to mount my USB flash drive. It is an MP3 player which I changed CD's on weekly so I am not listening to the same stuff at the gym every day. Well after a few weeks of Outkast it was time for a change so I sat down to fix the problem. Two hours later, I just went back to the 2.4 kernel.

I have gotten as far as getting the kernel to assign sda to my usb device but it never creates an entry in /dev and the /sys stuff is not mountable. Enabling USB debugging just shows me that things are messed up but does not really help much.

I also started to get annoyed with all the SCSI emulation needed to mount a USB storage device. I don't understand how Linus can hate SCSI emulation so much when it comes to burning CD's yet it is perfectly acceptable to use it to mount a USB disk. Seems a bit hypocritical, but then again...he did sort of invent Linux so I guess I can cut him some slack.

So all in all, I have been disappointed in the 2.6.x series of kernels and if they are the one's that are supposed to take the desktop market by storm then I think Linux on the desktop is in trouble. It is no wonder Redhat and SuSE are staying away from it for the most part right now. It is going to take both of them a lot of work to get everything working properly I would imagine.

Am I the only one who went back?

Re:How many have upgraded and then gone back?

[Anonymous Coward]

Supposedly, USB Mass Storage is basically SCSI protocols piped over USB, so it requires SCSI emulation.

Re:How many have upgraded and then gone back?

[Abcd1234]

Are you using devfs? Did you try to migrate to udev (since devfs is deprecated)? Did you try to manually create those device entries and just use them, so you can at least make sure the device drivers work?

As for /sys, you compiled sysfs into the kernel, right? If so, it should be automatically mounted (well, at least, it is on my box). You also need the following fstab entry, in order to mount usbfs:

none /sys/bus/usb usbfs defaults 0 0

Help me understand...

[GooberToo]

...why kernels are still not offered up via torrents yet?

Seems like it would help a lot.

Right now, I can't even connect to a use mirror. Grrr.

Re:Ready for the desktop?

[MrNybbles]

drwtsn32 wrote, apt-get update && apt-get upgrade

That is not normally enough to get a 2.6.x kernel working correctly with the rest of the system.

Debian Distros Only!

First edit your apt sources file so you are upgrading to unstable. (Insert Windows Joke Here!) (I don't know offhand if any other Debian branch has the right stuff for Linux Kernle 2.6.x)

apt-get update && apt-get install module-init-tools && apt-get upgrade

apt-get upgrade may not upgrade module-init-tools for some reason. You might also want to run "apt-get install udev" if you have the hotplug stuff built into your kernel. Other things may need to be done for your system. This was enough for mine.

The debian command dselect may do a better job of Upgrading your debian system as far as conflict resolution is concerned, but I really don't like the user interface to it. If you want to know more about debian packages check out http://www.debian.org/distrib/packages.

A complete Debian 2.4.x to 2.6.x upgrade guide would be nice. Anyone know of one?

Nforce2 APIC issues still not resolved :-(

[motown]

Even Kernel 2.4.6 still locks up frequently on my Asus A7N8X Deluxe motherboard unless I specify the options "noapic nolapic" at boot time. Then the system runs flawlessly (even with ACPI-support).

I read somewhere that the problem currently lies in the BIOS, rather than in the kernel, and that some vendors have already released proper BIOS updates that add a "C1 disconnect" option, which supposedly does the trick.

Unfortunately, Asus has released no such update as of yet.

Does anyone here (perhaps one of the kernel developers involved) have any more details on this?

Can this problem eventually be solved in the kernel, even without any BIOS updates?

After all, as far as I understood it, the BIOS pretty much takes a back seat as soon as the kernel is running, right?

Re:Anyone know what Rusty is talking about here?

[Svennig]

And the quote is:

/* "to look upon me as her own dad -- in a very real, and legally binding sense." - Michael Palin */