Close on the heels of the release of Linux 2.6.11 is the discussion of what should be included in the next revision. Andrew Morton who maintains the MM tree considered as the soft development tree from which things are later pushed to Linus lists the status for various new features here.
Linux: New Features For 2.6.12
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
How about: autoconfigure.
Your system detects what it has on board as features and dynamically updates and optimizes its performance on that system, installing the ideal configuration.
Most distros do that already, and its really not somthing that belongs in the kernel anyway. If you can’t configure your own kernel just let the distro handle it. Even Gentoo has a kernel auto-configuration option!
I Think its about time for Linus to start 2.6.7 and leave the stability and security to Allen Cox. Even though Linus is a good programmer and stuff Cox can do a damn good job on fixing bugs and stabilizing Linux.
i am not an anti Linus guy but its time for him to work on something new.
Stop all further development on wizzbang features (Infiniband) and take some time to implement decent ACPI support.
OK, it looks like I’m trolling and maybe I am. But I don’t see the point in adding features that very few people will use when a feature that everybody uses daily on their laptop is so utterly broken. I mean, laptops are just not usable without a sleep (suspend to RAM) function.
Should be priority #1.
well, from what i recall linus described acpi as a big mess of a metalanguage or something. so its not just a question of adding it ones and being done, every bios have its own quirks it seems. and it does not help that from the looks of it, ms have a acpi implementation that is either buggy or alterd compared to the standard.
so to get fully working support they more or less have to clone the windows way of dealing with acpi rather then code the standard and then go bug hunting. the bugs can be bad code in the bios, support for windows quirks or a bug in the linux kernel code, and given these facors it can be a pain to get it to “work”, let alone flawlessly.
Typo? You mean 2.7?
I’d also like to see a bigger seperation between stable and dev vanilla kernels; but they intend for distros to handle that now which is ok as long as you aren’t running say … gentoo
or run your own vanilla kernels.
I mean, laptops are just not usable without a sleep (suspend to RAM) function.
x86 laptops are barely usable anyway.
I would like to see Reiser4 in the vanilla kernel, maybe under a (experimental) or smthng like (unstalbe).
C’mon, I use reiser4 since 2.6.5 (+-). IT IS STABLE [IMO]
I’m sick of my machines breaking when I move to a new ‘STABLE’ kernel. You’d think these guys would have more sense.
ACPI should be the top priority for the next kernel release!
Yeah, but you have to be careful. My Intersil 11g card was getting flagged as Linksys, which was a Bad Thing.
It may be difficult to do, and asking for it may sound like whining, but I’m with the crowd that’s asking for better ACPI support. Every day at around 5 PM, I shut down X and press the power button on my laptop to turn it off. Every morning, I cold boot my laptop. If I was on Windows or OS X, I could just close the lid and call it a day, come back in the morning, press a button and be ready to go. Maybe then I’d bother taking my laptop on more travels. As it is, it’s just a brick that takes a minute and a half to get to a usable desktop.
Infiniband is great, and implementing it is in line with the claim that 3.x will be vastly more upward-scalable, but back here on Earth, we’re still missing some basic features. If I was an evangelist, how would I tell a Windows user that a Linux laptop is slick if it can’t suspend without crashing?
Oddly enough my laptop(thinkpad) freezes much more with Win2K than Linux (2.6.10) ….
It used to be the contrary but even if “sleep to disk” function is not correctly supported under linux, I can’t use it with windows because 2/3 time it fails freezes and loozes eveything….bad, bad ….unaccounted the times were simply shutting down Windows leads to freeze (and this is an out of the box install, no crazy device, no special driver).
I guess ACPI is a real mess for every system, hopefully standard will shift to something better because I still have to see an implementation that would not suck.
you must have many VERY unlucky machines…
I agree with the call for better ACPI support, but it’s hardly as easy as saying it – many (if not most) laptops have broken BIOS’s that deviate wildly from the standard – this is partly Microsoft’s fault, for their OS tacitly accepting it and their compiler which is producing said broken code.
I think blame should be placed on the computer manufacturers too though; releasing a machine that doesn’t conform to the ACPI spec simply shouldn’t be acceptable if they use the term ACPI to advertise it in any way. But no, they just get away with it, because most consumer protection groups don’t understand the issues involved.
Reiser4 in the kernel would be nice; apparently there are some far-reaching issues with how it should be merged.
I think it’s fairly crucial though; to put up some good opposition to WinFS (if it ever appears) or Apple’s Spotlight. The filesystem is one area where Linux has been ahead for a long time; keeping it that way would be nice 🙂
Maybe some support for sleep in the nVidia based Powerbooks will be nice. They’ve got them working on the ATI based Powerbooks so it shouldn’t be too hard to get it working on the others.
How well does power management work on Apple portables under Linux (or OS X, for that matter)? What is their version of ACPI?
Me thinks that Linus & friends have lost common sense when it comes to producing quality 2.6 kernel releases. Looks like they’re going crazy with all the new features, but losing track of what’s important: quality.
If you a add a new feature and that breaks something else, you shouldn’t have added it. If you know of something that’s broken and have something new to add, you should fix what’s broken first, then add new stuff. It’s nice if you can get something working that didn’t work before, but it’s more important that you don’t break stuff that is working now.
The Linux kernel isn’t an app that you can toy with, add some buttons to push on, or throw some nice-looking goodies on just for fun. It isn’t an app: it’s the heart of every Linux system that’s out there.
If it’s just another driver module using proven kernel infrastructure: go ahead, add it. But if it’s more essential/bigger than that: go start a new development tree, damnit.
Users shouldn’t have to fear upgrades between ‘stable’ releases. But if kernel 2.6 development continues the way it’s been done lately, I think I’ll be looking elsewhere, and give Free-/NetBSD a try. Not to say that Linux is bad; the 2.6 kernel is an impressive piece of work. But developers would do good to get their priorities straight.
I agree whole-heartedly. What if I want the latest, most security-fixed kernel, but the last 3 or so versions all have weird bugs that make my hardware not work? What, do I backport bug fixes myself? Do I wait for someone with similar hardware to backport, or make a patch that makes one of the newer versions more stable?
While some of the newer features they’re adding are exciting for some, the quirkiness is problematic and annoying for far more.
What if I want the latest, most security-fixed kernel, but the last 3 or so versions all have weird bugs that make my hardware not work? What, do I backport bug fixes myself?
I would suggest using the 2.4 line of kernels! They’re still updated, and they work very well. What do you need 2.6 for?
I dunno what ya’ll do to your kernel but 2.6 is quite stable.I have installed all 2.6 variations since 2.6.3 and compared to when 2.4 came out…2.6 is way ahead stabilitywise….imo.
2.6 works fine here, since 2.6.0
No problems what-so-ever.
I would also like to see reiser4 in mainline along with FUSE.
What’s the difference between the MM tree and the official one ?
“What’s the difference between the MM tree and the official one ?”
isnt that explained within the news itself. its a development tree but not a major unstable branch like 2.5
Is there a site that clearly states in laymens terms the differences between the 2.4 and 2.6 kernels and why you should use one over the other? What’s missing in one but available in the other? etc.
Thanks
http://www.kniggit.net/wwol26.html
http://www.osdllinuxsummit.org/presentations/t3s3_Morton_using-2.6_…
Under OS X, sleep and resume are incredible. Close the lid and the laptop goes to sleep instantly. Open the lid and it’s ready in under a second. Much faster than any Windows XP laptops I’ve come across.
If only this would be the case under Linux. At the moment, I’m shutting down my laptop whenever I move places and that’s quite inconvenient. Now instead of being instant on, I need to wait for a minute while Linux boots.
The iBooks and Powerbooks don’t use ACPI or APM. They have their own Power Management Unit (PMU). There is a kernel module that emulates APM so that most APM tools work with Macs.
Better sleep support for the Powerbooks in the next kernel version will be fantastic. You’ll get more Mac users using Linux! Sexy hardware, free (libre and gratis) OS 🙂
Thanks for the links.
The real question behind my previous post is why is Slackware 10 faster than say Fedora Core 2 by such a huge amount? Slack uses 2.4.x by default, while I think Core 2 uses 2.6.x. The difference is *really* noticable on my PC. Is it the kernel, or the extra modules and daemons that Core 2 loads? Please don’t take this as a flame war.
Thanks again
allso, isnt the “sleep” on those apple laptops what windows calles suspend? ie, park the hardive and shut down the cpu but keep power to the ram so that memory state is preserved? when a windows machine goes into sleep i think it writes the ram memory to the hardrive and then fully powers down the laptop. it may take a bit longer but its allso safer if something should happen to the battery while not in use. maybe apple have made their laptops monitor battery while sleeping so that if it gets into danger of running out it will dump ram to drive to, i dont know…
I think Linus should not add ANY features on 2.6.x branches. That is what 2.7 is for.
I wish they’d finally whip him into keeping even releases STABLE and only do bug fixes (and keep features frozen).
The whole linux development process is a mess.
oh and just to add – yes the stupid CD writing related changes ATAPI-SCSI stuff of around 2.6.7 has caused me no end of headaches.
It is unprofessional!
The kernel developers such as Morton and Torvalds have made their opinions on 2.6 clear – they’re merging more features into it rather than immediately starting a 2.7, to keep developer attention on 2.6, lower merging hassles, etc.
To those who say “but I can’t upgrade blindly to the newest 2.6 kernel!” – throughout much of the early 2.4 branch, there could be serious trouble upgrading to the latest 2.4 kernels, sometimes as severe as filesystem corruption. Upgrading to the newest kernel is for kernel developers, people who want to experiment, etc. It is not particularily tested, much less quality assured. Use a distribution which provides modified kernels with security fixes backported if you’re looking for stability.
If you want to use kernel.org kernels, go for it; but complaining that it’s not gone through rigorous QA and sometimes breaks things is like complaining about running a CVS version of a program; by all means, report bugs, but it’s not a polished release targetted at end users.
In many current systems, the general idea is to get a system every two or three years, keep up with a few security patches, and a few years later, consider upgrading. The Linux kernel is under constant development. There’s absolutely no reason for most people to track the changes and upgrade to the newest kernel just because it’s been released and perhaps slashdot announces it.
To complain that the current system is unprofessional misses the aims entirely. Debating the aims is fine; ignoring the stated ones, which appear to be thoroughly complied with, and complaining that the development process doesn’t meet your particular standards and ideas is inane.
> I think Linus should not add ANY features on 2.6.x branches.
Morton’s the maintainer of 2.6, not Linus. A case can certainly be made for not adding features.
> That is what 2.7 is for.
Morton and Linus have decided otherwise; 2.7 is for when a huge change comes up. That may be what you would like 2.7 to be for, but it’s not what 2.7 actually is for.
> I wish they’d finally whip him into keeping even releases STABLE and only do bug fixes (and keep features frozen).
Who’s they?
Even releases have never necessarily been stable; merely more stable than odd releases, for the most part.
> The whole linux development process is a mess.
Yep.
He’s not talking about automatically building a kernel (i.e. running ‘make’, installing, etc., like Gentoo can doe), he’s talking about automatically adjusting the kernel config to the actual hardware on your system.
Personally, I think it’s a great idea, but only to a point. There’s a lot more in the kernel other than hardware, and who knows if you’ll want it enabled or not.
Dunno about all this complaining about laptops not going to sleep. I’ve been running IBM Thinkpads with 2.6 since the beginning of that version and they work fine with disabling ACPI in the kernel boot line and using APM instead. Instant on, instant off, and CPUFREQD works well too. I can adjust CPU speed manually and automatically (when on battery) etc.
Do newer notebooks not support APM at all?
As for stability, I have only had one problem and that was a build of a 2.6 kernel not supporting my laptop sound but the previous and next build of the same version worked so probably a Debian bug.
Do newer notebooks not support APM at all?
Yes. Unfortunately.
ACPI support in LInux is poor because ACPI implementations are just generally broken. Linux’s problem is that its ACPI implementation actually follows the standard, while everyone else doesn’t (often to be compatible with Microsoft’s broken implementation). For example, on my laptop, using the hlt instruction to invoke throttled cpu states triggers a 1KHz tone in the power supply that is annoying enough to make me disable ACPI. It’s the fault of the hardware, not the software.
I remember the tty code was supposed to be a huge mess still in 2.6, and it was going to be fixed. maybe now is a good time?