The impending release of a new stable kernel promises greater adoption for Linux, as it becomes more reliable and scalable over a larger variety of processors. This is a new article that highlights some of the changes, both big and small, with some code samples.
Towards Linux 2.6
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
2.6-test5-mm4 runs great on my system. EVERYTHING is so much faster than 2.4. Even webpages render faster. The big thing though is the multitasking.
I can compile somthing,listen to audio, surf, and use GIMP without any slowdown or lag. VERY impressive, and my machine is hardly top of the line.
I am having a couple issues with DevFS though, hopefully it will be fixed (if its the kernels problem).
I tried running 2.6.0-test5 on an Athlon MP system and it “crashed” after about 8 hours under heavy load. The crash was odd… I don’t believe a kernel panic ever occured, but for whatever reason the interrupt handler simply stopped working and all drivers entered infinite loops waiting for interrupts.
It’s going to be awhile before 2.6 is ready for real world production use, especially on SMP systems…
I am having a couple issues with DevFS though, hopefully it will be fixed (if its the kernels problem)
Are you running gentoo by chance? I tried running a 2.5.xx kernel(not too long ago) and there were issues with DevFS.
I can compile somthing,listen to audio, surf, and use GIMP without any slowdown or lag. VERY impressive, and my machine is hardly top of the line.
I’ve heard that several times, so I’m very excited about this release (too lazy to try the test releases). There’s nothing worse than untarring when listening to music and then everything just stopping for 5 seconds 😉
This new linux kernel is starting to sound very BeOS like. Granted the mechanics and techniques sound different, but the end-user benefits sound the same. Can anyone tell me if we are going to see the BeOS responsiveness on Linux anytime soon?
No, ArchLinux.
http://www.archlinux.org
It seems to be a common problem, at least on the kernel mailing list.
The kernel cant find the RootFS, even though there is the proper entry in LILO.
Most of the stuff was way over my head, but the new audio core sounds promising
This is a known error in the mm4 patch for 2.6.0-test5, and there is a specific patch for it.
If you understand the patch format, it’s easier to just edit and save the file init/do_mounts.h in the linux-2.6.0-test5 source directory yourself. There are 2 lines that look like this:
Code:
if (!sys_stat64(name, &stat) != 0)
simply remove the ! from sys_stat64 in each occurance (only the first !sys, NOT the last != 0).
Then recompile your kernel. Worked for me.
Robert Crawford
Hey, that worked perfectly. Thanks!!
I only found one instance though.
Could you by chance point me to the exact patch location, in case I have to do this again.
Thanks again.
😀
>Can anyone tell me if we are going to see the BeOS responsiveness on Linux anytime soon?
It is a guess of course, but I think that a big part of BeOS responsiveness was because of application’s design, so I don’t think that kernel improvements will be sufficient..
A non-responsive kernel can prevent apps responsiveness, but a kernel with low-latency is not enough, you need the “cleverly designed” apps too..
“I am having a couple issues with DevFS though, hopefully it will be fixed (if its the kernels problem)”
I’ve heard that current devfs is supposedly deprecated in 2.5/2.6, in favor of a user space implementation or something.
“It is a guess of course, but I think that a big part of BeOS responsiveness was because of application’s design, so I don’t think that kernel improvements will be sufficient.. ”
I think the BeOS app_server also plays a key role in responsiveness. In some cases, applications automatically benefit from the design of the App/Interface Kit – for example, applications automatically spawn a new thread when its main window is created.
nice -n 20 make
With that I can play mp3’s while I’m surfing the web or even playing in an emulator with no loss. All while running my Cel 1.6 on it’s battery. 😉
A user-space implementation sounds pretty unlikely if you ask me.
After all that would me quite inefficient and you would still have to publish the information to the use-space agent somehow, then you could publish it outright.
That would almost be like realising /proc in user-space.
>> I’ve heard that current devfs is supposedly deprecated in 2.5/2.6, in favor of a user space implementation or something.
Yep, afaik devfs is going to be replaced with udev.
See:
http://lwn.net/Articles/41444/
http://lwn.net/Articles/28897/
yeah thats great and all, but it is a hack and should not be considered a working solution.
HAHAHAHA
That was funny. 🙂
man nice, it’s not a hack. You are just telling the scheduler what to do, otherwise your default ulimit is going to allow your process to hog the entire CPU.
quote from article…
A lot of work in improving threading performance has gone into the 2.5 kernel. The improved threading model in 2.6 was also done by Ingo Molnar. Based on a 1:1 threading model (one kernel thread for one user thread), it includes in-kernel support for the new Native Posix Threading Library (or NPTL), which was developed by Molnar in cooperation with Ulrich Drepper.
Does Ingo work for redhat or something, or did he improve the work redhat has done, because I was lead to believe that redhat developed this and back ported it back to the 2.4 kernel for redhat 9.
Am I right or wrong here?
No, it wasn’t done by Redhat at all.
They just ported it to the 2.4 kernel.
Thats the beauty of OpenSource, its open for anyone to use. 😉
I don’t like many of the decisions made by the kernel folks. As long as thet remain with a monolithic kernel, removing functionality like that provided by devfs to a userspace program seems awefully silly as the kernel is the program that deals with that sort of thing.
/sys? doesn’t /proc already do what /sys is supposed to do (via /proc/sys) in addition to many other things? Fill me in here, I am at a loss over this one.
The threading improvements are certainly nice though, as is the new security architecture and the inclusion of ALSA.
Wait a minute I see where I got things confused, it was sponsored and back ported by redhat.
http://kerneltrap.org/node/view/422
that every time when a new version of Linux kernel is released they comment like it becomes more reliable and scalable over a larger variety of processors.
How much does Linux need to become more reliable and scalable over such a larger variety of processors? With that much refinement, it should already become the super high class.. totally the top notche platform on Earth at this time. :p
I am just making a joke from the comment. No, not to flame some one or some products, to be a troll nor to start an OS war. I know that same thing going on MS products as well with words like ‘innovative’ and/or ‘new technology’
Anyway even on other platforms, I guess it is not any bad to listen to music while surfing and doing some creative works on GIMP..for example.. like myself have done HTML/CSS coding on vi while surfing with mozilla-firebird, looking my html work done so far on mozilla firebird (on another tab) and IE on win98 in vmware, looking at the original web design png image file on GIMP, cutting images from the design, listening to MP3, and chatting on XChat2 and Gaim all on my P4 1.5G FreeBSD 5.1-current machine. No slowdown at all (well, vmware was a bit slow though)
The difference is that You get the COMPLETE changelog that has every single line of code that has been changed(added).
With MS ALL you get is buzzwords.
So basically, if Linus was blowing smoke, people would realize it much quicker.
Oh well yes, that is true for Linux and other Open Source OSes.
I hope all of Open Source Developers keep up their good works. Cheers!
Well, they keep using that line because Linux keeps getting more scalable. A (very) rough breakdown:
2.0 was a nice uniprocessor kernel for a moderately loaded server.
2.2 was nice for a dual-processor SMP machine, on a slightly more heavily loaded server.
2.4 scaled up to 4+ processors, and down to embedded machines and PDAs.
2.6 scales up to machines with a dozen processors or so (like the SGI Altrix servers) and down to desktop machines.
2.8 will concentrate on scaling Linux to massive NUMA architectures with 100+ CPUs.
The patch can be found in recent posts on the google kernel group, or the Gentoo forums, but I had a problem with copy/pasting it into a text file. I lost the formatting, and it wouldn’t apply to the already patched 2.6.0-test5-mm4 kernel, so I had to do the hand edit. In this case it was easy, as the “fix patch” is really small. Probably will be fixed in mm5. I opened it in kwrite, and it was obvious what needed to be changed (- lines were in green, replace lines, +, was in red), and you couild see that all that changes is the !. On the Gentoo forum, do a search for author wrc1944, and look at the post for 2.6.0-test5, and the patch is in that thread.
Robert Crawford
Nice, simple & easy-to-understand info! Thx
I am not a kernel programmer nor a kernel hacker, therefore your comment was just easy to get.
Even I told “it should already become the super high class.. totally the top notche platform on Earth at this time”, though I really feel that Linux needs to become more mature and more powerful. *I know*
It’s going to be awhile before 2.6 is ready for real world production use, especially on SMP systems…
rofl.
Your problem is an interesting one (I assume you did file a bug report and/or posted about it), but the above statement is rubbish. You should know better than to make grand, sweeping statements off of a single anecdote.
2.6-test is already very stable, largely due to the work of groups like the OSDL who’ve been running automated regression and load tests. 2.6 will be stable and “production ready” long before the next iteration of your favorite os.