Linux Kernel 2.6.3 Released
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
I just dont like the idea of having a monolithic kernel with embedded drivers – most of the last 2 kernel updates are only driver fixes. Have a constant driver API, and allow drivers to be downloaded seperately from the core kernel. This way the kernel remains unchanged, and you only have to download driver updates. Same applies to the ALSA sound system. You know, microkernels aren’t such a bad idea now, are they?
Majority of experienced Linux users compile their drivers as modules. In fact, starting with kernel-2.6, you don’t need to recompile your whole kernel to update certain parts of it like the drivers.
Besides, I don’t know of any major kernel that doesn’t require some sort of inconvenience to update or ugrade. Micro, Macro, Mono kernel or whatever.
Do distros like Fedora compile drivers as modules by default?
From what I understand, distro’s like Red Hat compile with almost all options turned on since it doesn’t know what the end user’s system is … meaning you’re going to have a lot of unneeded bloat – but I could be wrong.
You are wrong
Most distributions compile everything as modules and then modprobe them as necessary(/possible).
Pretty much every distro compiles drivers as modules by default. Otherwise, the default kernels would be enormous.
@SmallStepForMan:
The drivers aren’t really part of the kernel. The source tree is just organized that way. At runtime, the drivers are loaded as dynamic modules, just like in BeOS or Windows. Now, the reason the drivers are organized this way is two-fold:
1) The kernel developers don’t like being slaves to driver compatibility. Keeping a stable binary ABI for drivers would be very complicated, and slow down development of the kernel. As the recent Windows source code story pointed out, a binary driver ABI burdens the code with lots of hacks for retaining binary compatibility. In any case, delicately maintaining binary-compatibility works a lot better for proprietory developers, who make major releases every several years, than for OSS developers, who must make improvements incrementally and in an evolutionary manner.
2) A lot of the same people who work on the kernel work on the drivers. For example, if someone changes the power-management API, he might want to go through and update some of the major drivers to reflect that change. The whole distributed development model just works a lot better when things are all integrated and in one place.
Yes, most of them.
There are, of course, a small set that are compiled into the kernel, as ide for example, by obvious reasons (how can you mount your root if your kernel doesn’t know how to read from your hd? This also apply to other devices you must access in kernel loading time).
Any bets on what the final version number is going to be? 2.6.12? I don’t usually like to play with beta OS’s so i am going to wait till the release of fedora or a version of Suse with the 2.6 kernel. I am excited about the new kernel though i hope it kicks butt.
Drivers are compiled as modules, but they are just inserted into one huge binary, if one thing fails, everything can potentially go down with it.
(This has been brought to you by Modern Operating System, Tanenbaum.)
@acobar
>how can you mount your root if your kernel doesn’t know how to read from your hd?
Not even HD-controller-modules for ide or scsi need to be compiled directly into the kernel.
There can be an initial ramdisk (initrd) containing these modules.
The boot-loader loads the initrd into RAM. The kernel loads the modules from RAM.
http://linux.maruhn.com/sec/mkinitrd.html
Anybody got the comments linus always posts to lkm?
“Ok, it’s out.
There were some minimal changes relative to the last -rc4, mostly some configuration and build fixes, but a few important one-liners too.
Happy testing,
Linus”
from linuxtoday.com
There are, of course, a small set that are compiled into the kernel, as ide for example, by obvious reasons (how can you mount your root if your kernel doesn’t know how to read from your hd? This also apply to other devices you must access in kernel loading time).
In the case of FreeBSD, the bootloader uses BIOS calls to load the kernel image from the hard drive and is also capable of linking modules into the kernel before starting it. This requres a complex three stage bootloader in which two stages are filesystem-aware, but the end result is quite nice compared to less featureful bootloaders like LILO. Thus everything in the FreeBSD kernel can be modular, and virtually everything is enabled per default driver-wise.
Is GRUB capable of linking modules with the Linux kernel before booting it?
I would think that 2.6.3 would be the 4th version in the 2.6 series since you count 2.6.0. You must not index into arrays often enough.
I’ve had no luck with the 2.6 kernel so far. Hopefully this will be the one …
Yes, Tanenbaum would say that.
He wrote Minix (a microkernel), afterall. If you recall, Minix is in Linux’s history. I won’t go into detail; you should know this history, I’m sure. The point is that Dr. Tanenbaum prefers microkernels–he has another besides Minix–and his bias should at least be noted without too much presumption.
Anyhow, it seems to me that the kernel development is doing very well at this point. The Alsa updates look fairly promising.
I can imagine what poeple would say if MS came out with 3 Service Packs in a matter of, what 6 months, and some one posted that they still couldn’t get Windows XP to run. I can imagine what people here would be saying if Microsoft came out with three Service Packs in 6 months.
What do service packs have to do with a Kernel Point release?
Thats some of the funniest stuff I’ve read in a while. Made me smile, great post. Surely being as outragious as that post is its not taken seriously by its poster, nor will it be by me. “Linux collects weapons of mass destruction” indeed! Hah! Great stuff! Maybe thats why SCO is so scared?
Majority of experienced Linux users compile their drivers as modules. In fact, starting with kernel-2.6, you don’t need to recompile your whole kernel to update certain parts of it like the drivers.
Actually, you could update a lot of drivers without recompiling a whole kernel since 2.2
it is upsetting that trackpads on soem notebooks do not work properly in 2.6.
That’s a huge changelog
The latest kernel breaks the ATI drivers as of 2.6.3_3 and up to the release. Not Linus’ fault probably, but a pain in the ass all the same.
The latest kernel breaks the ATI drivers as of 2.6.3_3 and up to the release. Not Linus’ fault probably, but a pain in the ass all the same.
That has been my problem with the 2.6 kernel all along. I can’t use ATI’s drivers which means I can’t use my dual-head setup (which I can’t live without any more), and now it would mean I couldn’t play ut2k4. Too much sacrifice. So I’m sticking with 2.4.x for now.
So that I can wait to update my kernel on my box with the Radeon. As far as I know, the Nvidia 5336 drivers still work (my other box has a GeForce); anybody given them a try with the new release yet?
I run kernel 2.6.1 and the ATI drivers and have no problems whatsoever. (ATI AIW 8500 using FireGL drivers).
I was going to use 2.6.3 when it came out, but I will hold back for now.
I don’t know about 2.6.3, but other 2.6.x releases can certainly be used with ATI drivers. It requires a little bit of work patching, but you should find everything you need on http://www.rage3d.com/board/forumdisplay.php?s=9d76602a1413bffa31ba…
http://isec.pl/vulnerabilities/isec-0014-mremap-unmap.txt
— Begin Quote —
Synopsis: Linux kernel do_mremap VMA limit local privilege escalation
vulnerability
Product: Linux kernel
Version: 2.2 up to 2.2.25, 2.4 up to 2.4.24, 2.6 up to 2.6.2
Vendor: http://www.kernel.org/
URL: http://isec.pl/vulnerabilities/isec-0014-mremap-unmap.txt
CVE: CAN-2004-0077
Author: Paul Starzetz <[email protected]>
Date: February 18, 2004
Issue:
======
A critical security vulnerability has been found in the Linux kernel
memory management code inside the mremap(2) system call due to missing
function return value check. This bug is completely unrelated to the
mremap bug disclosed on 05-01-2004 except concerning the same internal
kernel function code.
— End Quote —
Stupid ALSA.
The mindless drones will ignore the fact that there’s a security fix, but will instead latch onto the latest user-ran, executable e-mail attachment worm that is somehow a “Microsoft hole” even though it’s the users who are running it…
This is why Linux is not mainstream. Hypocrisy and anti-social behavior.
http://marc.theaimsgroup.com/?l=linux-kernel&m=107707898920941&w=2
Don’t be too quick to not upgrade because of one person saying they don’t work (especially on this site!), I know people who have them working fine.
The OSS drivers work fine in 2.4X but ALSA in 2.6X, really screws up my sound. Im just going to wait until distros start coming out with it by default.
Man, they release new kernels faster than I can say “new release”.
I’m not complaining though, progress is good
I just found out a few days ago 2.6.2 was released. I’m still using 2.6.1. Think I’ll wait a bit longer till the dust settles.
Download the new version of alsa (alsa-driver-1.0.2c, the rest are still 1.0.2) and compile them. Don’t forget to point out where are the linux sources (usually /usr/src/linux).
If you don’t put your sources there, just make a symlynk or use an option with ./configure.
After make and install, just run alsaconf and enjoy.
Hi
“The mindless drones will ignore the fact that there’s a security fix, but will instead latch onto the latest user-ran, executable e-mail attachment worm that is somehow a “Microsoft hole” even though it’s the users who are running it…
This is why Linux is not mainstream. Hypocrisy and anti-social behavior.”
1) Linux is already mainstream in the server side
2) Saying that an operating system will not go maintstream because a bunch of users are “anit social” is unreasonable
3)Outlook and Outlook express enabled people to automatically execute scriptable components and executables which resulted in the security hole. None of the email viruses are targetting any other client. Obviously outlook favored ease of use over security. This is only recently changing for the better.
4) Nobody is ignoring the kernel security fixes.
5) You assumed that and are artifically inserting windows here in a discussion about kernel 2.6 point releases.
6) None of the distributions are 2.6 based and hence the security threat if any is restricted to early adoptors and testers who are fully aware of this
7) It is well known that the stable series gets a good amount of enhacements over multiple point releases to be considered shippable
Jess
The specifications of grub are quite impressive but I never played with it. I use to take the safe way and lilo is just more appealing to me.
Someone went fishing. Don’t feed it
Thanks to the guy that pointed out the ATI howto. Took me about an hour but now I’m up and running with the 2.6.x kernel for the first time since it’s release. Dual-head (Big Desktop) is working and so is the 3D.
I’m using 2.6.3 by the way … so those of you wondering if it works or not, it seems to from my end. YMMV
You are wrong, since almost all of these features are modules. Than means no overhead/bloat if you don’t use them/not required for your hardware. Takes some diskspace, but not that many Mb.
I haven’t ever taken the time to get Alsa to work.
I will really try hard on one of these version
realeases as I realize it is the future. But
for now if you still want to try 2.6 series fear not,
the OSS stuff is all still in the kernel, it is
what I’m using, exactly the same as in the 2.4,
just in a different place in the “make menuconfig”.
Then again most everything is moved around, but
just thought I’d let you know that OSS sound is still
working in 2.6. Also it discourges use of OSS by labeling
it as “OBSELETE” or something, but it still works.
“I just dont like the idea of having a monolithic kernel with embedded drivers – most of the last 2 kernel updates are only driver fixes. Have a constant driver API, and allow drivers to be downloaded seperately from the core kernel. This way the kernel remains unchanged, and you only have to download driver updates.”
I know how you feel. It’s one of the reasons I like Darwin. Even though XNU isn’t a true microkernel (due to the inclusion of BSD), it has all of the plumbing of a microkernel, and the drivers are (mostly) seperate from the kernel itself. One of the things that always bugged me about Linux (and yes, BSD as well
is that there is no stable driver interface, and for Linux at least, there never will be (Linus is stubbornly against it).
Darwin is a complete open source OS (which appeals to me) but I’m not entirely sure how “free” it is; as in is it free in the sense of the other BSDs? I couldn’t say. The license was too long for my liking, and the current lack of drivers for my hardware kind of made it a moot point anyway. It’s still a snazy system, and I like it alot.
I am wondering if DragonFly will be moving drivers out of the kernel proper. I have read nothing that suggest that they are planning on it, but as they are moving towards a threaded message passing system, I can’t can’t see why they couldn’t try for the stable driver APIs that Linux and BSD lack, but Darwin and Windows have.
Exciting times at any rate.
“…user-ran, executable e-mail attachment worm that is somehow a “Microsoft hole” even though it’s the users who are running it.”
Could you be a little more clear as to what you mean, please? Are you trying to say that Linux users are anti-social because you somehow relate a malicious email with a totally unrelated Linux kernel fix? I’m sorry to sound so dubious, but we could relate far more technical bugs than an apparently deliberate ambiguous issue like an email problem. The do_mremap problem was, after all, a very technical and advanced issue requiring a much more profound knowledge of the underlying principles than a simple attached malicious executable in the majority of cases.
If we should be so bold as to allege anti-social behavior
in any case within our community, we should also have very
strong evidence of such, without relying on obviously
flippant alleges.
@Bascule: Grub includes a generic mechanism for loading files into memory at boot-time so the kernel can get them without having a disk-driver loaded first.
@Bill Sykes: A kernel point-release is nothing like a Windows Service Pack. The Linux kernel is improve incrementally, especially in the early stages of a new version. Windows is developed in stages, with lots of fixes and changes coming with each service pack. A kernel point release is more akin to the individual patches that come out quite often on Windows Update. Quit trying to apply your Windows knowledge to the Linux world — it doesn’t work!
Woot woot my sound finally works. 🙂
Thanx!
So all is well in ATI land? what card are u running?
And are u able to play Unreal, etc?
I run dual monitors as well so that would be great to get that going.
1) Linux is already mainstream in the server side
Mainstream–as in, mainstream. Not the server niche. Mainstream computer usage. Linux has at least another ten years in that department. I am not trolling–I seriously believe that. I could write page after page about the latest Fedora or Red Hat or Mandrake, the so-called easy distributions, and what exactly is wrong with all of them. A hint–most of the problems are fundamental stumbling blocks that are a part of Linux architecture itself. It will take years and years for these problems to evolved into more accessible components that allow for an easier and more elegant computing experience.
2) Saying that an operating system will not go maintstream because a bunch of users are “anit social” is unreasonable
No, it is not, because the attitude is reflected throughout the entire community and development efforts. Hell, I’ve had people cut me down when I simply suggested interface improvements. I was completely polite and grateful, but suddenly because I have a job and therefore don’t have the time to contribute endless patches to fix someone else’s app, I don’t “deserve” to criticize “volunteer work.” The community as a whole is immature and unprofessional. If it’s not bizarre project names like “Xouvert” and “EMACS” and “Kougar,” its anti-social bullying of end-users.
3)Outlook and Outlook express enabled people to automatically execute scriptable components and executables which resulted in the security hole. None of the email viruses are targetting any other client. Obviously outlook favored ease of use over security. This is only recently changing for the better.
Both Outlook and Outlook Express do not automatically run user attachments. An outright lie. Outlook in particular won’t even allow them to appear–instead displaying a dialog box. The e-mail viruses you speak of obviously do not target other clients, because they are targetted for Windows, the most widespread desktop OS. Give Linux that marketshare and see what happens. Ever visited linuxsecurity.com any time and witnessed the dozens of vulnerabilities announced weekly for Linux distributions? Next.
4) Nobody is ignoring the kernel security fixes.
Compare to when an executable worm goes around. It’s something a user has to run for it to infect anything. Suddenly, everyone is falling over themselves criticizing Microsoft fast enough, when it has nothing to do with them. But when major kernel vulnerabilities are announced (this isn’t the first one), there is ZERO outcry. Slashdot didn’t even stick it on their front page like they do with every single minor worm that makes an appearance on unpatched Windows systems now and then.
5) You assumed that and are artifically inserting windows here in a discussion about kernel 2.6 point releases.
I’m merely making a point. Points are always “artificially inserted.” Disregard if you don’t agree, but don’t criticize me for holding this opinion. Besides, Windows and “M$” are ALWAYS on the minds of the majority of Linux users. Read a little closer; there is already Windows discussion going on regarding service packs. Sometimes discussions go off into related tangets, because discussions are dynamic processes.
6) None of the distributions are 2.6 based and hence the security threat if any is restricted to early adoptors and testers who are fully aware of this
Slackware and others were 2.6-ready, and 2.6 has been out for quite some time. In any event, that point is irrelevant–you clearly did not read that the flaw affects all kernels up to 2.6.2 and 2.2.45, which is why a new 2.2.x kernel was also released. Next.
7) It is well known that the stable series gets a good amount of enhacements over multiple point releases to be considered shippable
The point of the test releases of 2.6 was to get it shippable. Nonetheless, “enhacements” are indeed a good thing–as are security fixes. And it is also a good thing to point out double standards; it helps to give perspective to those who are willing to accept differing viewpoints as possibly having a rational base.
I love that my previous comment got “reviewed.”
I couldn’t get ALSA to work when I first started using 2.6. I had ALSA and my soundcard driver compiled as modules and it would not work so I tried compliling both into the kernel and that did not work either. The solution was to compile ALSA into the kernel but compile the actual driver as a module. On boot it says that no soundcard is found but only because the module is not loaded yet.
I could write page after page about the latest Fedora or Red Hat or Mandrake, the so-called easy distributions, and what exactly is wrong with all of them. A hint–most of the problems are fundamental stumbling blocks that are a part of Linux architecture itself
Why don’t you go into more detail? What fundamental problems? You can’t honestly expect anyone to believe your opinion based on a total lack of facts.
Hell, I’ve had people cut me down when I simply suggested interface improvements. I was completely polite and grateful, but suddenly because I have a job and therefore don’t have the time to contribute endless patches to fix someone else’s app, I don’t “deserve” to criticize “volunteer work.”
Open source does not work the same way. You have no right to request features. You don’t have to pay for the software. If it doesn’t suit your needs then don’t use it or fix it yourself. This is the whole point of open source! Go ahead and criticize poorly written apps that you paid for but why do you have to criticize a piece of work that was provided for free? No one is forcing you to use it or pay for it. Now if you are a loyal, paying customer then you have every right to ask for changes because if they do not comply you can take your business elsewhere. That’s the great thing about choice. People like you are just pissed that “taking your business elsewhere” has no affect on the open source community so you have to bitch about them in forums.
The community as a whole is immature and unprofessional. If it’s not bizarre project names like “Xouvert” and “EMACS” and “Kougar,” its anti-social bullying of end-users.
That is a huge leap to claim the entire community is immature and unprofessional just because of the names of some programs. Windows has its share of stupid program names too (CuteFTP anyone?) but does it really matter? What’s in a name anyway? I would like to know what you consider “bullying end-users”. How is it possible for an entirely free system to bully anyone?
Ever visited linuxsecurity.com any time and witnessed the dozens of vulnerabilities announced weekly for Linux distributions? Next.
Local exploits are entirely different from remote exploits. Linux tends to suffer from the former, and Windows tends to suffer from that latter. A local exploit won’t do anything from across the web, where most attacks occur.
Compare to when an executable worm goes around. It’s something a user has to run for it to infect anything. Suddenly, everyone is falling over themselves criticizing Microsoft fast enough, when it has nothing to do with them. But when major kernel vulnerabilities are announced (this isn’t the first one), there is ZERO outcry.
What was the end result of the latest Linux vulnerability? An announcement and a fix as I recall. What was the end result of the latest worm for Windows? Billions of dollars in lost revenue and tens of thousands of zombie computers spamming the rest of the world. Does that answer your question?
I’m merely making a point. Points are always “artificially inserted.” Disregard if you don’t agree, but don’t criticize me for holding this opinion. Besides, Windows and “M$” are ALWAYS on the minds of the majority of Linux users. Read a little closer; there is already Windows discussion going on regarding service packs. Sometimes discussions go off into related tangets, because discussions are dynamic processes.
That entire paragraph was utter nonsense.
Slackware and others were 2.6-ready, and 2.6 has been out for quite some time. In any event, that point is irrelevant–you clearly did not read that the flaw affects all kernels up to 2.6.2 and 2.2.45, which is why a new 2.2.x kernel was also released. Next.
What exaclty came of that flaw? A fix. That’s it. You are making a big deal about a flaw that was found and fixed. That makes no sense.
The point of the test releases of 2.6 was to get it shippable. Nonetheless, “enhacements” are indeed a good thing–as are security fixes. And it is also a good thing to point out double standards; it helps to give perspective to those who are willing to accept differing viewpoints as possibly having a rational base.
You seem to have no rational base for your posts.
What exaclty came of that flaw? A fix. That’s it. You are making a big deal about a flaw that was found and fixed. That makes no sense.
No, I’m not, the paragraph you’re referring to is a response to the previous poster who said the flaw only affected 2.6 kernels.
Next time, try reading and comprehending.