Linux Kernel 2.6.18 Released
Submitted by 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
What is a typical changelog size for the linux kernel?
Never mind, I took a look at the others. This seems to be par for the course.
Edited to reflect the fact that I investigated this.
Edited 2006-09-20 07:14
I’d argue that your initial reaction is still valid.
The diff against 2.6.17 is 950,000-some lines long. And that’s typical for the 2.6. kernel. I don’t know how they can call 2.6 “stable” when nearly every minor revision has nearly a million-line diff. That’s a lot of changes for “stable”.
Here’s a nice changelog:
http://kernelnewbies.org/LinuxChanges
– New wireless driver included, zd1211: a lot of WiFi adapters are based on this chipset
– A lot of fixes for Mactels
– Support for iSight on iMac G5
– Hotplug and many other fixes for the SATA driver
– Audio support for many machines (Sony laptops and Mac Mini)
Enjoy!
I never bothered looking at size of the kernel before but now I’m shocked. With all those drivers, there is around 210+ MB of source code!
It must be a real job maintaining all that code!
and yet its done
heh, much of the code dont need much changes done to it over most smaller time periods (device drivers mostly) unless something closer to the core of the kernel changes that is important to how the other code works.
and much of those drivers are maintained by the same people, organizations and corporations that use them, they just happen to use linux as the basis for all the rest to ease the workload.
it just shows the power of cooperation i guess.
and then you trow in the source for x.org or xfree, gnome or kde (with their software packages) and things realy start to make you wonder how its all financed
I just installed 2.6.18 on FC5,a much cleaner compile.
Have we heard any rumors/rumblings yet about the next version of the kernel and what that might bring?
Just curious……………..
AFAIK a 2.8 kernel has not been taken in consideration yet !
Edited 2006-09-20 11:13
What the hell kind of article is this? No summary of changes, no description of any kind. Just “Linux update” with a tar.bz2 and changelog link. If I wanted that I’d subscribe to just kernel.org and forget sites like this.
Really, is that the best you guys can do? Is this how low OSNews is sinking just to announce things ahead of the rest? Why don’t you just stop summarizing articles altogether!
This was very, very lazy Eugenia.
Here is a much more informative article on this topic:
http://kerneltrap.org/node/7144
Edited 2006-09-20 11:07
While I agree that this Eugenia’s post is a little sparse, to say the least, the one you linked to doesn’t actually say much else, other than “Linus talked like a pirate.”
I imagine an actual OSNews article containing only Linus’s email summarising changes (and maybe mentioning pirates) would have attracted even more heated critisism for being a waste of time.
Edited 2006-09-20 12:08
Yarr! A new kernel off the port bow! Give her a broadside and prepare to be boarded.
This isn’t a direct reply to you Schmeggma. I just had to borrow your reply link, as the OSN editors have yet again decided to stoop to mod trolling themselves.
WTF is up with modding every post the editors disagree with to -5 and removing the reply link. Do you guys really think this kind kind of bullshit behavior is going ot keep users coming to your site? f–k it, I’ve had enough of this site, the troll editors, and all the other bullshit. I’ll be moving on now to another news site, and if I don’t find one to my liking I’ll build one myself that kicks OSN ass.
What do you want, the 3mb changelog posted on the front page?
Other comments showed how to summarize the changes nicely or at least give helpfull links.
People… some would bitch if you gave them a new kernel.
beyond the kernel and its minor changes that is not visible to users and rather concentrate on other aspects of the OS like eg the interface, the stability of applications, the availability of commercial drivers, the integration and other things as well.
I would give the kernel the crown, but maybe 2 years ago, now it is the base we all feel comfortable with.
How many times I wished open office would be faster, and much stable, how many times I wished the browsers be as fast as in windows, or how many times we wish installing application would be more simple and straight forward rather than compiling in many distros.
I wish all other parts of the OS would achieve what the kernel has achieved in short time!
Give also some credit to HP, IBM, Sun and others who donate hardware devices in thousands of dollar to kernel development!
To achieve better integration we must avoid integration. The best kinds of integration are applications that work together without being tied to each other.
Unix is about lots of small programs all working together to create what ever you need.
Integration should always be done at the user level because each user has completely different needs. By integrating at the development level you just get applications that don’t fit a user’s needs.
I <3 tar,gzip,ssh,grep,mail,rsync.
Linux has DBus for that. DBus is used by HAL (which help managing your hardware), which uses Udev, which itself is based on lots of improvements from the kernel.
DBus basically allows “applications that work together without being tied to each other”.
It couldn’t have been achieved without sysfs whih is in the kernel, and would have been less efficient if Linux hadn’t the unstable API policy.
>Unix is about lots of small programs all working together to create what ever you need.
And integration is all about their developers talking to each other, not doing duplicate work and cooperating on common issues. It makes sense for some small programs to merge and for some it doesn’t. It depends on many factors. Integration is done to get more features but can also bring new set of issues.
I don’t believe in that “do one thing and do it well” paradigm as it doesn’t work in most real world cases (nor in operatig system design). It’s an oversimplification. For example it is more useful to have a grocery with both potatoes and tomatoes instead of having two selling just one thing on adjacent places. But it may be better to have a phone and a PDA instead of a bulky and ugly combo device which is neither a good phone nor a good PDA.
…but if they were selling the Potmatoes would you be happy if you just wanted the potato?
But it may be better to have a phone and a PDA instead of a bulky and ugly combo device which is neither a good phone nor a good PDA.
My Treo resents that comment
You’d better move beyond your total lack of clue and giving advice at the same time.
If you got even a little clue, you’d know that the “minor” changes you talk about allow you to have a faster system and disks even under load, to suspend without problem, to autodetect and autoconfigure devices like USB without the need to read the red tape on your device (true PnP), to have live updating of your FS (downloads), to have more Wifi adapter supported, … a lot of thing very visible to any users. Then, there are other things visible to specific users.
What is this trend of denying title of “user” to people that are not Joe using his desktop ?
The fact that some parts of ivtv drivers are now integrated in the kernel tree will be visible to lots of MythTV users. They are not lesser users than Joe !
I would give the kernel the crown, but maybe 2 years ago, now it is the base we all feel comfortable with
A lot of people are still not comfortable with the kernel, as a lot of “unstable internal API” trollfests or “recompiling the kernel to update a driver” trollfests show us.
How many times I wished open office would be faster, and much stable, how many times I wished the browsers be as fast as in windows
Please, don’t downgrade my browsers to the speed they have on Windows ! I’m always amazed at how slow they are on Windows, IE being the worst. Sometimes, I wonder if the Firefox works on Windows, being so slow.
or how many times we wish installing application would be more simple and straight forward rather than compiling in many distros
Who are “we” ? Every one of my users install apps on Linux, which they couldn’t do on Windows !
“compiling in many distros” ?!! You just showed you’re a troll. You couldn’t help it, could you ? Or are you really that clueless ?
I wish all other parts of the OS would achieve what the kernel has achieved in short time!
Which is ?
Give also some credit to HP, IBM, Sun and others who donate hardware devices in thousands of dollar to kernel development!
You’ve got to be kidding. You think they don’t gain a lot from that ?
This is work like everywhere else. I cringe to have to remind you that very few people on earth actually buy the hardware devices that you talk about, which are for enterprises. And so, if these companies want to sell Linux on their expensive and powerful hardware, they’d better give the Linux kernel devs the hardware necessary for that. No credit due there, as these changes for big iron make people angry more than anything.
No credit due there,
I give IBM credit for their fine developerworks articles.
“If you got even a little clue, you’d know that the “minor” changes you talk about allow you to have a faster system and disks even under load”
False.
Faster disks, then let me tell you the story of very High I/O Wait you can experience on all linux distros since kernel 2.4 till kernel 2.6.18; the story goes like this, if you just copy a huge folder like 1 GB internally on your disk your CPU no matter what AMD, Intel, X2, Dual core, HT on will reach 50% through a samba job on the system, just one, then your CPU raeches 70%, do a third job and your system is unusable at all due to 100% CPU saturation. I also confirmed this on Intel hot E6400 CPU and still no change. Now even though I dislike windows and I use it as my just file server and gaming, I admit that you could access your disk like a maniac without even getting CPU >20% unless you have compressed the volume then you CPU under the same previous tests would reach to 30%.
I guess, my friend that you lack experience, or you never even tried to benchmark linux at all. I benchmarked linux like crazy, with SPECviewperf and other small tools that is compatible with windows as well and I admit sadly that windows beats linux in performance, and gladly not stability.
Who’s problem is this, is it the drivers for disks, noop, because the kernel guys confirmed its related to file system (ext3 or Reiserfs), and this is true as I have repeated again the previous tests on JFS and XFS and found that JFS beats both linux file systems while XFS beats all of the three to a point reaching MacOSX native journaled file system, but again nothing to reach windows performance (of course assuming its a fresh install and not infected, which I always do to my testing environments).
“to have more Wifi adapter supported”
Why just Wifi Adapter why not also web camera which the kernel basically supports almost none of the most common two major webcam manufacturers (logitech and Creative)
Sound support on almost all cards with synthesizers never were able to play .mid files due to some error messages of “null sound device” again never was fixed since 2.4 kernel till now 2.6.18. It would be then wise to not expect much from the kernel but rather from OEMs to give kernel modules!
I wish you stop claiming things and misleading readers, and start to backup you claims with real life benchmarks and results like I did before, and you are welcomed to repeat the previous tests and write down your experiences for us, and thanks for all.
What a load of tripe. Just for kicks, I moved 20GB off one of my disks to another in the root filesystem. My oldish CPU only hit 30% with ReiserFS, which is a cpu-intensive filesystem.
While not very informative for end users, Linus’ announcement of 2.6.18 is pretty amusing:
http://kerneltrap.org/node/7144
Ben Dooks:
[ARM] 3784/1: S3C2413: keelhaul config for MACH_S3C2413/MACH_SMDK2413
[ARM] 3785/1: S3C2412: Fix idle code as default uses wrong clocks
[ARM] 3793/1: S3C2412: fix barnacles in wrong serial info struct
[ARM] 3794/1: S3C24XX: do not defined set_irq_wake when no CONFIG_PM
…
It’s good to know someone’s fixing the barnacles in the Linux source code.
Davy Jones:
Didn’t do anything, the scurvy lad. Ahoy!
Eh, I thought the Monkees were lip-synching anyway.
On a (slightly) more serious note it’s somewhat refreshing to see that the Linux kernel developers have a sense of humor about what they do.
I know Microsoft definitely does, given all those videos (like the “Microsoft Redesigns the iPod” video they made as a cautionary tale for their packaging department)
http://video.google.com/videoplay?docid=4313772690011721857
One question I probably ought to be able to find elsewhere: Has anyone added Minix 3’s MinixFS to the Linux kernel drivers yet (or will they ever)? I know Minix 3 uses a new system Linux can’t read.
Hey, they’ve fixed the IPW2200 problem with WPA 🙂
I’ll finally being able to use wifi again, that’s a great news.
Hope my Realtek 888 is fixed.
Hope my Realtek 888 is fixed.
here you go!
Patch from Realtek:
– Add support of ALC888 codec
– ….
I don’t know if it supports Realtek ALC850?? Anyone?
Edited 2006-09-20 22:45
2.6.18 is supposed to have a lot of fixes for users with nVidia AM2 motherboards, and quite a few for Intels 965 chipset.
distros’s that have just released with a 2.6.17 kernel may have dropped a bollock as the premier chipsets for the respective AMD and Intel platforms will remain essentially unusable to the common man.
still SUSE 10.2 intends to use the .18 kernel, and maybe fedora 6 will do so too?
I couldn’t wait for FC6 and compiled the damn thing 🙂
It reminds me that this kernel should have a fix for a broken PATA driver for a chipset which is used on the best Core 2 Duo motherboards (ASUS, MSI), and makes, for example, Ubuntu live CD unbootable on these, unless you have a SATA CD drive of course.
for i am not, and must await SUSE 10.2
I’ve never benchmarked CFQ with our loads versus AS. A bit late now, but I’m wondering if this will cause our shop some performance issues (initially, I’m aware I can change it back to AS).
Edited 2006-09-20 15:28
I’ve never benchmarked CFQ with our loads versus AS. A bit late now, but I’m wondering if this will cause our shop some performance issues (initially, I’m aware I can change it back to AS).
Well, CFQ looks better than AS here -> http://www.redhat.com/magazine/008jun05/features/schedulers/ so presumably you will have some performance gain
Well, CFQ looks better than AS here
Excellent article! Much appreciated.
I did a summarization of some of the longer lists out there here:
http://www.breakitdownblog.com/2006/09/20/linux-kernel-2618-release…
I don’t if that helps some of you that don’t want to read through the huge dump changelog or the longer summaries out there. I didn’t cover everything, just the bigger items I thought people care more about.
from http://www.heise.de/open/artikel/77566:
/quote The most important innovations of Linux 2.6.18
With Linux 2.6.18 released the developers the next larger revision of the Linux Kernels. It brings numerous innovations with improvements on the process scheduler and the SATA drivers as well as numerous reconditioned or new subsystems and drivers back.
After three months development time has Linus Torvalds with the Kernel 2.6.18 those again a larger revision of Linux released. Thus it lasted those up to the completion somewhat longer than the normally anvisierten eight weeks – dreiwöchige time-out the Linux father probably is to it with debt.
Announcement
The numerous innovations of the recent Kernel version might not feel most users directly, since changes do like those at the scheduler and the Locking mechanisms rather unnoticed under the hood their service. As with each Kernel version there were however numerous detail improvements important in the sum, a heap new and revised hardware drivers and some partial completely reconditioned and improved subsystems.
So supported the Netfilter Framework for putting on Firewalls now Connection Tracking with minutes SIP (session Initiation Protocol), used for IP-Telefonie. The Netfilter code for minutes H.323 controlled now besides call Forwarding. Within the NFS range there were numerous improvements as for instance support for Direct I/O (O_DIRECT). Erstmal offers the Kernel 2.6.18 of Intel in new server chip corroding existing I/O Acceleration Technology (I/O RK), which serves about for the optimization of Gigabit Ethernet network transfers. New gene Eric Segmentation off load (GSO), the TCP Segmentation off load is to improve the throughput on network hardware (TSO) do not control.
Normally under /usr/include/linux/ being the header files of the Kernels for system LIBRARies and hardware near user space programs were revised and released from unnecessary ballast. They leave themselves now with “make headers_install“install, whereby the Installationsscript removes directly all exclusively for the Kernel of relevant sections.
The documentation describes now, which ABIs (Application Binary interface) are reliable to the user space, which are still in a test and a development phase and which are to be removed sooner or later. As standard IO scheduler now CFQ is used. The file system JFFS2 now supports the Extended of attribute (EAs), approximately needed for SELinux. CIFS can besides with SMB servers communicate, which work still with unencrypted passwords – for instance OS/2 or Windows 95. That SMBFS for a long time any longer not maintained applies now also officially as ungewartet and might be erased – now there CIFS to a large extent everything can, which also the SMBFS controls – sooner or later from the Kernel. Thus it also the hardly still maintained DEVICE file system was issued devfs, that after various discussions in the past months after first preparations with 2.6.17 with 2.6.18-RC1 completely over board thrown became.
Numerous improvements with the process scheduler were added; under them also the extension SMPnice, which is to provide with processes with different priority for a fairer distribution on multiprocessor systems. One other innovation makes possible for the scheduler with dual core/multiprocessor systems to distribute the processes during small system utilization in such a way the fact that a processor or a one of the cores makes the work while other in the current savings mode continues to remain…..
Another way to view changes is to just copy the .config from the previous kernel into the new one and run “make oldconfig”. As a rule I always do this with a newer kernel. One thing that often doesn’t get mentioned with updated summaries are (usually many) small additions to netfilter, such as adding new types of filters and tracking methods.
The zd1211 driver that was added is for USB wifi devices only. I still think Linux desperately needs a better wifi system in general and much better support for 802.11g PCI and PCMCIA cards one typically finds in stores. All these different projects providing drivers for different chipsets which behave differently is a little crazy.
Core2 boards with Intel 965 chipsets (975 too?) don’t have native IDE support, on a lot of boards this is added with an extra chip. Often, it is a Jmicron 363. Until now this was only usable in Linux with certain kernels, and a boot option all-generic-ide (which also made my SATA drives appear as IDE.
2.6.18 allows IDE drives attached to the Jmicron chip to function without needing any boot parameters, and doesn’t screw with the SATA attached to the ich8. Tested and working great here, which makes me rather happy
The danger of being a relatively early adopter.. fortunately that danger has passed now