Linux Kernel 2.6.32 Released

diegocg writes "Linus Torvalds has officially released the version 2.6.32 of the Linux kernel. New features include virtualization memory de-duplication, a rewrite of the writeback code faster and more scalable, many important Btrfs improvements and speedups, ATI R600/R700 3D and KMS support and other graphic improvements, a CFQ low latency mode, tracing improvements including a 'perf timechart' tool that tries to be a better bootchart, soft limits in the memory controller, support for the S+Core architecture, support for Intel Moorestown and its new firmware interface, run-time power management support, and many other improvements and new drivers. See the full changelog for more details."

Comments

Llacking in terminology.

[c0l0]

I'm not perfectly happy with the term "virtualization memory de-duplication". Linux 2.6.32 introduces what is called "KSM", an acronym that is not to be confused with "KMS (Kernel Mode Setting)" and expands to "Kernel Samepage Merging" (though other possibilities with similar meaning have already emerged). It does not target virtualization or hypervisors in general (and QEMU/KVM in particular) alone. KSM can help save memory for all workloads where many processes share a great lot of data in memory, as with KSM, you can just mark a region of memory as (potentially) shared between processes, and have redundant parts of that region collapse into a single one. KSM automagically branches out a distinct, exclusively modified copy if one of the processes sharing those pages decides to modify a certain part of the data on its own. From what I've seen until now, all that's needed to have an app benefit from KSM is a call to madvise(2) with some special magic, and you're good to go.

I really like how Linux is evolving in the 2.6 line. Now if LVM snapshot merging really makes it into 2.6.33, I'll be an even more happy gnu-penguin a few months down the road!

Re:Llacking in terminology.

[Luyseyal]

But that wasn't enough, so I had my balls cut off.

Laugh all you want, but I know an AIX kernel hacker who did just that.

-l

Re:Llacking in terminology.

[1s44c]

I'm not perfectly happy with the term "virtualization memory de-duplication".

The term is a little nonspecific. However KSM is truly wonderful and I look forward to saving a ton of physical memory over my KVM machines when the kvm/qemu userland tools catch up.

This is already in redhat's virtualization stuff.

Re:Llacking in terminology.

[Anonymous Coward]

You sir, have made my day.

Re:Llacking in terminology.

[Bert64]

I have a system running a 2.6.32-rc6 kernel with KSM and the latest kvm (which includes support for this, but its turned off by default)... Because i run a number of virtual images that boot the same kernel and system libs (different apps ofcourse), it saved me over 1gb of memory on the host.

Re:Llacking in terminology.

[Anonymous Coward]

SMIT happens?

Re:Llacking in terminology.

[Anonymous Coward]

^^ change "fucking penguins" to "being fucked by Penguins"

There, fixed that for you...

Re:Llacking in terminology.

[ichigo 2.0]

How much memory did the images use in total before the change (i.e. how big was the savings in %)?

Of course, every byte is precious as long as it doesn't affect performance, but it would be interesting to know how much more images one can expect to run on one computer. :)

Re:Llacking in terminology.

[icannotthinkofaname]

But that wasn't enough, so I had my balls cut off.

But did you chop your balls off LIKE A BOSS?

Re:Llacking in terminology.

[Hatta]

kvm (which includes support for this,

Does each application need to support KSM? I can't just run two instances of the same arbitrary application and let the kernel figure out what to do?

Re:Llacking in terminology.

[Anonymous Coward]

You think that's bad? Take a look at the Linux 2012 issue.

Re:Llacking in terminology.

[haruchai]

  At least one M$ lover has gone the distance - http://en.wikipedia.org/wiki/Ina_Fried

Re:Llacking in terminology.

[Avtuunaaja]

No, the application needs to call madvise(2) for the memory regions it thinks it can share.

To make this work automatically for all similar pages, the kernel would have to compare every page in memory against every other page in memory, not something you want to do.

Note that something similar already works for normal applications -- shared libraries and program images are shared between applications until their memory regions are written to. If a program forks, every page it has is shared until it's written to.

Re:Llacking in terminology.

[buchner.johannes]

That sounds interesting. Wouldn't this also resolve the negative effect of static libraries to a large part (having multiple copies of the same code in memory)?

Re:Llacking in terminology.

[Anonymous Coward]

This is the interwebz. Next time please hyperlink something like madvise(2) in an informative type post.

Re:Llacking in terminology.

[Bert64]

total memory usage was about 14gb across 10 images

ATI chipsets

[zoward]

2.6.32's KMS and R600/700 improvements are expected to give a huge 3D performance boost to the open source ATI drivers - can't wait to test this!

Re:ATI chipsets

[Ritz_Just_Ritz]

I'm excited about the ATI improvements making it into the kernel too. Wonder if Ubuntu Karmic will pick up the new kernel after some testing?

Re:ATI chipsets

[Cyberax]

No, you'll have to wait for Lucid Lynx (Ubuntu 10.04) for these changes.

Re:ATI chipsets

[Jojoba86]

Superb! I've had my laptop for 2 years and now may finally be able to get 3D graphics working without a major headache.

Re:ATI chipsets

[Anonymous Coward]

No, as Ubuntu Releases are version-stable, and backport security fixes only (Firefox being the exception of that rule). You may install the kernel from the mainline kernel PPA though: http://kernel.ubuntu.com/~kernel-ppa/mainline/
Just fetch the .deb that fits your architecture, and install it via `sudo dpkg -i /path/to/your/downloaded/archive.deb`.

Re:ATI chipsets

[Doug Neal]

2.6.32's KMS and R600/700 improvements are expected to give a huge 3D performance boost to the open source ATI drivers - can't wait to test this!

This is indeed excellent although it needs to be backed up by support from the X driver. Currently I am running Ubuntu Karmic on a Radeon HD 3600 series card (RV635, which counts as an R600 series - quite confusing) and 3D support sucks. Both the "radeon" and "radeonhd" drivers only have basic support for these chips - desktop effects don't really work.

I was using the fglrx driver on Jaunty, which worked OK, but it seems to be getting worse with every release. In Karmic it was so broken I just gave up on it. It seems to play a lot better with Compiz/GNOME than with KDE for some reason.

Re:ATI chipsets

[L4t3r4lu5]

You're modded funny, but this is why I use Windows. Since I moved into my own house and can put cabling where I want (negating the horrific experiences I've had with wireless networking), 3D graphics issues are the only thing stopping me migrating to linux.

Re:ATI chipsets

[erko]

If you're ok with not-so-open drivers, nvidia 3D cards have worked for years. I am waiting for quality open source 3D linux drivers, but until then, at least 3D can and has worked reliably on linux. (the nvidia-settings tool is reasonable enough that you generally don't need to edit config files)

Re:ATI chipsets

[RubberDuckie]

The Fedora team has backported the KMS and R600/700 improvements to FC12, which I've been running for a few weeks now. While it's better than nothing, 3d performance still has a way to go. The performance of my old Heretic II game is still unacceptably slow.

The ATI drivers usually took the sacrifice of a goat to get them to work, but their performance was far superior. Too bad ATI won't support recent releases of Fedora.

Re:ATI chipsets

[bcmm]

I've been using the RCs of this kernel, and the Radeon r600 support is already much faster and more stable than fglrx.

Re:ATI chipsets

[Jojoba86]

I don't know why I was modded funny or you were modded troll, I was being serious. I'm not a complete linux n00b but I've never got my ATi card (R600) working with 3D acceleration, I'm lucky to get smooth 2D graphics. The current version of openSUSE completely locks up after installation due to the graphics driver. For hardware 2 years old it shouldn't be this hard, but good job to those working on this and ATi for opening up the specs.

Re:ATI chipsets

[Jojoba86]

...at least 3D can and has worked reliably on linux...

No, you mean your 3D card has worked reliably on linux. A laptop with an R600 chipset is not easy to get working.

Re:ATI chipsets

[erko]

...A laptop with an R600 chipset is not easy to get working.

You have a good point -- graphics card choices in laptops are limited. But, instead of using windows, or linux without 3D, I just no longer consider laptops with ATI cards (which leaves nvidia choices for now).
I feel your pain though -- I had a T43 with X300 card for a couple years, and that experience is unfortunately burned in my memory.

Re:ATI chipsets

[Anonymous Coward]

I'm using the RC of this kernel and the edger ppa of x.org, and the r700 support is much slower and slightly less stable than fglrx for me.

GUI responsiveness with compositing enabled is MUCH faster though. No more full second lags when maximizing/minimizing.

Re:ATI chipsets

[fritsd]

it's very nice but rough around the edges, i.e. not fully OpenGL 2 yet. Openarena plays well but Nexuiz is ah.. extra challenging at the moment :-)

Re:ATI chipsets

[Anonymous Coward]

everything seems to play a lot better with Compiz/GNOME than KDE. I've certainly had that experience with nVidia as well. I still use KDE 4, though; just because I've been able to customize it much to my liking. But there's still weird visual bugs than when I run Compiz/GNOME on this same computer...

Re:ATI chipsets

[Anonymous Coward]

>This is indeed excellent although it needs to be backed up by support from the X driver. Currently I am running Ubuntu Karmic on a Radeon HD 3600 series card (RV635, which counts as an R600 series - quite confusing) and 3D support sucks. Both the "radeon" and "radeonhd" drivers only have basic support for these chips - desktop effects don't really work.

Ubuntu Karmic doesn't have kernel 2.6.32, and hence doesn't have support for R600/R700 3D in any driver.

As far as I can tell, not even a rolling release like Arch has yet updated the kernel and X packages to the new versions which support R600/R700 3D graphics acceleration. It shouldn't be long now, though, for rolloing release distributions to incorporate this code.

Patience, young padawan.

Re:ATI chipsets

[disi]

I tested one of the 2.6.32-rc and it actual works, but don't expect much performance :) Make sure you use >=mesa-7.6 and a decent radeon xorg driver... works also great with xorg-server-1.7 while ati-drivers don't...

Re:ATI chipsets

[bcmm]

To be honest, compositing is the bit of responsiveness and stability that I've tested most.

Re:ATI chipsets

[L4t3r4lu5]

Troll (n):

"In Internet slang, a troll is someone who posts controversial, inflammatory, extraneous, or off-topic messages in an online community, such as an online discussion forum, chat room or blog, with the primary intent of provoking other users into an emotional response[1] or of otherwise disrupting normal on-topic discussion."

Karma butchery aside, everyone who modded this post troll is an idiot. Since when was expression of opinion and description of personal experience in any way inflammatory or extraneous? This is my experience with Linux. I am not trolling.

My 8800 GTX has not had working 3D graphics in:
- Ubuntu since 6.10 (not tried latest, downloaded yesterday)
- Mandriva (since is was Mandrake. Graphics card may not have been the same at that time, but I try each version)
- Fedora Core (Since RC4).

I may have more luck with latest Ubuntu as I will have Internet access on the same box (wireless setup never worked). I would post regarding success / failure, but I'd probably be modded "redundant." It's obvious that it already works; You all say so, so it must be true.

Re:ATI chipsets

[Rysc]

Your definition of troll is specific where it should be non-specific. A troll is someone who says something for the purpose of getting a rise out of the reader, hopefully inspiring a reply, hopefully an angry reply. "3D graphics under Linux suck, this is why I use Windows" is a classic troll in any pro-Linux place (like slashdot!)

You may not have intended it as a troll and may have merely been voicing an opinion, but it certainly reads as inflammatory and certainly looks like a troll. Why didn't you say "I never could get 3D working satisfactorily on Linux, which is why I still use Windows"? This doesn't imply that Linux is deficient and Windows is superior, so it's less likely to read like trolling.

Re:ATI chipsets

[DuckDodgers]

While many Linux users don't (and shouldn't need to) care about installing and compiling a kernel from source code, it's not that difficult to do. If you want to try out the new kernel, that's a relatively easy way. It will still take you a few hours though - mostly for reading instructions on how to do it.

Re:ATI chipsets

[Nutria]

My 8800 GTX has not had working 3D graphics

That's an Nvidia card?

If so, then it's your fault. Nvidia has had a very good binary driver for 5+ years, and 10s (hundreds?) of thousands of people have used it ever since. I'm even using it in the unsupported configuration of 64-bit kernel and 32-bit userland. Google Earth rocks with the binary driver...

If, OTOH, this is some weird ATI card, never mind...

Re:ATI chipsets

[Hurricane78]

We had the same problem on Gentoo. ATi is to blame. Because the reason of your problems is, that the driver is not compatible with the newest versions of Xorg (<=1.7).

We here had to mask xorg-1.7.x and all those packages, and reinstall the latest non-masked ones, to get it to work again:

>=x11-apps/xinput-1.5.0
>=x11-base/xorg-drivers-1.7
>=x11-base/xorg-server-1.7.1
>=x11-libs/libX11-1.3.2
>=x11-libs/libXScrnSaver-1.2.0
>=x11-libs/libXext-1.1.1
>=x11-libs/libXi-1.3
>=x11-libs/libXinerama-1.1
>=x11-libs/libXtst-1.1.0
>=x11-libs/libXxf86dga-1.1.1
>=x11-libs/libXxf86vm-1.1.0
>=x11-proto/bigreqsproto-1.1.0
>=x11-proto/fixesproto-4.1.1
>=x11-proto/inputproto-2.0
>=x11-proto/recordproto-1.14
>=x11-proto/scrnsaverproto-1.2.0
>=x11-proto/xcmiscproto-1.2.0
>=x11-proto/xextproto-7.1.1
>=x11-proto/xf86bigfontproto-1.2.0
>=x11-proto/xf86dgaproto-2.1
>=x11-proto/xf86driproto-2.1.0
>=x11-proto/xf86vidmodeproto-2.3
>=x11-proto/xineramaproto-1.2

Yes, it’s a huge nasty messy pile of shit. And I’m very happy to dump fglrx forever, as soon as the radeon driver gains full 3D functionality.

Re:ATI chipsets

[Hurricane78]

Ah, damnit. I meant >=1.7! That’s what you get for posting without previewing, once a year. ;)

Does it Fix XKCD 619?

[sheepweevil]

All of these features are cool and all, but does it solve the well-known XKCD 619 bug?

Re:Does it Fix XKCD 619?

[Anonymous Coward]

http://imgur.com/73EAu - RESOLVED, FIXED quite a while now!

Re:Does it Fix XKCD 619?

[Cyberax]

Yep, it's fixed.

See here: http://imgur.com/73EAu

Re:Does it Fix XKCD 619?

[Anonymous Coward]

4 months is "quite a while now"?

Re:Does it Fix XKCD 619?

[Ambassador Kosh]

There is nothing the kernel developers can do about this. On my machine which is a dual 2218 opteron with Dual nvidia 8800gts video cards running 4 monitors just playing flash a single screen will bring one cpu core to its knees. Normal sized flash videos will bring one cpu core to its knees and will sometimes drop frames. The same video if saved to a local file and played with xine/mplayer etc will use up 1-2% cpu power at the lowest cpu freq (1 ghz).

Linux is perfectly capable of smooth video playback at low cpu usage. It is flash that is not capable of that and nothing we can do about it. Flash is proprietary and only adobe can really fix it. If they don't choose to fix it then nothing we do is going to make those videos play any better. I am running the latest version of the flash player and it still does not play any better. Flash is a massive cpu hog.

Kernel developers should work on solving problems they can actually solve. The Xorg and desktop developers should solve problems they can solve. All they can do is try and provide tools that flash can use to play videos better but since those tools already exist I don't see much that can really be done. Even on windows flash is a dog, netflix is a good demonstration of that. I am not like silverlight but it is massively faster at playing and at higher quality then flash is. An older system that could not play flash fullscreened at 800x600 from netflix can stream hd streams at 1600x1200 with no slowdowns at all with the silverlight player and low cpu usage.

Adobe needs to fix it. Nobody else is to blame for it and nobody else can fix it.

Re:Does it Fix XKCD 619?

[Shark]

It makes one wonder, with Microsoft encroaching on Adobe's turf like this, shouldn't they at least try to cover their ass and give great support on non-MS platforms? I really don't get what's going on as far as their strategy goes... Or maybe there just isn't any strategic planning.

Btrfs: kill off ext# please!

[Anonymous Coward]

I'm glad to see Btrfs improving so rapidly. I hope popular distros start including support for it, but more importantly, start using it as the default filesystem.

It's time for the ext-based filesystems to die. They are a technology that was obsolete a decade ago.

ReiserFS was set to kill them off, but unfortunately found another victim first... JFS and XFS only work well in certain high-end niches. But Btrfs is much better as an all-around filesystem, which is why it has a chance to finally put an end to ext-based filesystems.

Re:Btrfs: kill off ext# please!

[Jack Malmostoso]

ReiserFS was set to kill them off, but unfortunately found another victim first

Too soon!

Re:Btrfs: kill off ext# please!

[gzipped_tar]

Fedora 12 already supports btrfs as an experimental feature, but there's still a long way to go in the near future.

And what about Reiser4?

Re:Btrfs: kill off ext# please!

[moosesocks]

How does Btrfs compare to ZFS? I've been using ZFS-on-FUSE, and absolutely love the incredible data integrity and volume management features that it provides. The new support for deduplication will also be wonderful once implemented.

Of course, the performance and the idea of trusting my data to FUSE leave much to be desired.

(On the downside, I'm peeved that Btrfs is GPL licensed, which will prevent it from becoming "the one true filesystem" from here on out. Windows users will be stuck with NTFS, Linux users will get Btrfs, Mac users will get whatever apple is secretly working on, and the BSD/Solaris camp will get to keep ZFS. None of them will be compatible, and FAT32 somehow remains the only viable option for removable media.)

Re:Btrfs: kill off ext# please!

[psm321]

I'll stick with my tried-and-true ext3. Tried reiserfs several times with lots of problems (unrelated to the creator's other problems). Tried jfs and had massive corruption and freezes that would lock up most of the kernel. Never had a problem with ext2 or ext3. Given the fact that everybody else also seems to think that reiser, jfs, etc. are also great alternatives, it makes me wary of btrfs.

Re:Btrfs: kill off ext# please!

[Anonymous Coward]

http://lwn.net/Articles/342892/

Hope that give you the answers you were looking for.

Re:Btrfs: kill off ext# please!

[Abreu]

What prevents other, non-GPL operating systems from using Btrfs?

Writing drivers for a filesystem is not a "derivative work" is it?

Re:Btrfs: kill off ext# please!

[von_rick]

Since EXT2/3 and recently EXT4 have some level of popularity among users, there are applications in Windows to mount them at startup if you are using a dual boot system. There wasn't anything wrong with ReiserFX or JFS or any other filesystems that I tried - but EXT# was the only one which could be easily used under Windows.

Re:Btrfs: kill off ext# please!

[Fished]

It wouldn't be a derivative work to write a driver if you did so from scratch. But to do so from scratch is... shall we say a "non-trivial problem." It would be better to have a BSD licensed filesystem that could be relicensed as appropriate--GPL for linux, proprietary for Windows and Mac, BSD for .. ahem ... BSD, etc.

Re:Btrfs: kill off ext# please!

[mcgrew]

And what about Reiser4?

She's dead, Jim.

Re:Btrfs: kill off ext# please!

[gzipped_tar]

Well joking aside, I heard rumors that Reiser4 is being considered by Linux devs and likely to be merged in the stable kernel soon. Any news on that?

Re:Btrfs: kill off ext# please!

[sim82]

None of them will be compatible, and FAT32 somehow remains the only viable option for removable media.)

So in one of your possible ideal worlds, I would have to use ZFS on a 2GB SD card? Sounds interesting.

Re:Btrfs: kill off ext# please!

[Urkki]

What prevents other, non-GPL operating systems from using Btrfs?

Writing drivers for a filesystem is not a "derivative work" is it?

How good, accurate and up-to-date is the specification of btrfs?

If only real, accurate "specification" is the source code, then it's damn hard to create a compatible and reliable new implementation from scratch. File systems are complex, concurrent (meaning many files being accessed simultaneously) and performance-critical as well as reliability-critical. Getting it right is hard, while getting it wrong is bad, so there needs to be really good reasons to even try to do it, instead of using something that already works.

This is especially true while development is still continuing (as is case with btrfs).

Re:Btrfs: kill off ext# please!

[LOLLinux]

Why would it need to be under a proprietary license for Windows? BSD code can live perfectly fine along side of the proprietary code in windows.

Re:Btrfs: kill off ext# please!

[Abreu]

Nevermind that, I think that the whole objection to a BSD license in this case would be that such a license could not prevent MS (or Oracle, or Apple) from "embracing and extending" the whole filesystem so that the "standard that everybody uses" is no longer free

Re:Btrfs: kill off ext# please!

[LOLLinux]

And such an attitude goes contrary to the spirit of the BSD license.

Re:Btrfs: kill off ext# please!

[Bert64]

MS will never support a filesystem they don't control unless forced to, and certainly won't make it the default...

BSD, Solaris and OSX all support UFS, as does Linux.... Linux also supports the hfs+ filesystem currently used by OSX, not sure if bsd/solaris do but there are bsd licensed drivers for it so no reason not to.

Re:Btrfs: kill off ext# please!

[Bert64]

But even assuming ms would use an open filesystem, they would want to alter it to make it incompatible with everyone else's implementation... And they can't do that very well if people are able to download the source.

Re:Btrfs: kill off ext# please!

[Bert64]

I would prefer to use EXT2 on small sdcards, so as to support filesystem permissions...
Or how about something like jffs2 - a filesystem actually designed for flash media.

FAT32 is a pretty garbage filesystem, and it's patent encumbered, an open filesystem without the weaknesses of fat32 and which is supported everywhere would be extremely useful. It won't happen tho so long as MS have sufficient market share to bury any open filesystem, they want people locked in to their proprietary patented filesystems and use their weight to prevent any other fs from gaining traction.

Re:Btrfs: kill off ext# please!

[larry bagina]

DragonFly BSD has HAMMER.

Re:Btrfs: kill off ext# please!

[Anonymous Coward]

Fedora 12 is not a suitable system for use. I tried it recently, and the installer crashed before even getting to the disk partitioning screen. This was on a 3-year-old system that has been running Slackware and then Ubuntu its entire life, so I'm pretty sure the problem is with Fedora.

Re:Btrfs: kill off ext# please!

[LOLLinux]

But even assuming ms would use an open filesystem, they would want to alter it to make it incompatible with everyone else's implementation...

So what? People take GPLed programs and make incompatible forks with it often. Why is it wrong if Microsoft wants to do the same with BSD code it would want to use?

Re:Btrfs: kill off ext# please!

[phantomcircuit]

I've been using ZFS-on-FUSE

Are you insane? You probably just cut the performance of your drives by 90%.

Of course, the performance and the idea of trusting my data to FUSE leave much to be desired.

Oh sorry you're informed AND insane. :P

Re:Btrfs: kill off ext# please!

[ProfMobius]

Fedora12 is perfectly stable. Using it right now on 3 computers. Two of them are Intel based, one is AMD based. And they have all three of ATI, NVidia and Intel graphic chips. Didn't have any problem at all and the installer is quite stable.

Re:Btrfs: kill off ext# please!

[reub2000]

I highly doubt we'll be seeing Microsoft make any effort in terms of impliment anything that would make it play nice with Linux or any other operating system.

Re:Btrfs: kill off ext# please!

[diegocg]

Not that long, in my opinion. I've been using Btrfs as my root filesystem (except for /boot and /home, but anyway) since the day after Fedora 12 was released (two weeks). I haven't had any problem at all with it - no hangs, not even a simple printk low-periority warning. It seems to be quite stable - i'm not surprised that they are considering (according to LWN) declaring it "ready for early adopters" in the next kernel release.

Re:Btrfs: kill off ext# please!

[Anonymous Coward]

I don't think full UNIX file ownership and permissions make a lot of sense on removable media, because there's no direct equivalence of user ids between different systems. Unless you go to great lengths to ensure there is. It'd be nice if Linux had a filesystem specifically designed for removable media, that encoded stuff like read-only and executable flags, but didn't include ownership.

Re:Btrfs: kill off ext# please!

[Anonymous Coward]

MS doesn't give a shit about that "spirit".

Re:Btrfs: kill off ext# please!

[Anonymous Coward]

Unless there is a really big reason to jump filesystems like there was with ZFS and the added functionality it gave on the LVM layer, I wouldn't just move to a completely new filesystem for production use until it has been around a few years. A glitch on that level may mean data loss in ways unforseen, especially if backup programs don't detect the changes and assume that a corrupted file is fine.

ZFS is an exception. It gives a lot of functionality like snapshots, file migration, RAID-Z, and a slew of other things. If it wasn't for this fact, I'd wait a year or so for general use, and perhaps 3-5 years for production critical use.

Re:Btrfs: kill off ext# please!

[Abreu]

I think I have a (theoretical) solution to that.

Lets say that the copyright owners of btrfs create the windows/osx/solaris/aix drivers?

Or more realistically, if the copyright owners of btrfs grant an interested third party a special license to create a lgpl'd btrfs driver?

Re:Btrfs: kill off ext# please!

[Urkki]

I think I have a (theoretical) solution to that.

Lets say that the copyright owners of btrfs create the windows/osx/solaris/aix drivers?

Or more realistically, if the copyright owners of btrfs grant an interested third party a special license to create a lgpl'd btrfs driver?

There can't be a "special license" to do LGPL version. Once such version is out, well, it's out. So they could as well make the whole thing LGPL (which IMHO would be a good idea).

But if there are a lot of developers, getting everybody to agree (or even to reach everybody) is a lot of work. Replacing=rewriting the code of those who don't agree might be on option, depending on how much of it there is.

But I'm pretty sure they actually thought about it and chose GPL over LGPL because they wanted to, so they're not likely to change (Google might find the answer).

Re:Btrfs: kill off ext# please!

[TeknoHog]

It's time for the ext-based filesystems to die. They are a technology that was obsolete a decade ago.

ReiserFS was set to kill them off, but unfortunately found another victim first... JFS and XFS only work well in certain high-end niches.

In my experience, JFS offers most of the benefits of ReiserFS, while being lighter on CPU. So it is definitely not just for the high end. It has also turned out more stable than Reiser, though in the recent years this has evened out.

On some of my machines there have been consistent problems with using JFS on the root partition, but this may be due to the init scripts. No data has been lost, though, and on non-root partitions JFS has consistently been rock solid for me. This includes a number of x86, PowerPC and ARM machines.

Re:Btrfs: kill off ext# please!

[serviscope_minor]

If you look at the filesystem benchmarks, JFS is often not the fastest, but scores best in term of CPU usage. I've found that on a netbook which has a very fast disk (ie flash) and not much CPU, JFS is actually the best option. YMMV of course, and I came to this conclusion before ext4 was released, and I haven't trried the pre-release ones like btrfs.

Re:Btrfs: kill off ext# please!

[compro01]

Unless I am out of date, there's only really ext2 in windows via ext2fsd, which will mount ext3, but sans journaling, and will only mount ext4 if you disable extents, which is one of the major features.

Re:Btrfs: kill off ext# please!

[Mad Merlin]

(On the downside, I'm peeved that Btrfs is GPL licensed, which will prevent it from becoming "the one true filesystem" from here on out. Windows users will be stuck with NTFS, Linux users will get Btrfs, Mac users will get whatever apple is secretly working on, and the BSD/Solaris camp will get to keep ZFS. None of them will be compatible, and FAT32 somehow remains the only viable option for removable media.)

You may as well stop holding your breath now. Microsoft will never support a general purpose filesystem that they didn't personally develop, and Microsoft will never develop a filesystem that doesn't have draconian licensing. There will never be "one true filesystem" until Microsoft is destroyed. Though it'll probably also be necessary for Apple to disappear as well.

Re:Btrfs: kill off ext# please!

[Abreu]

There can't be a "special license" to do LGPL version. Once such version is out, well, it's out.

Why not? If the copyright owner* grants a license to do a LGPL driver for a non-Linux OS, it doesn't necessarily mean the entire project has to be cross-licensed to LGPL, does it?

And yes, for the record, cross-licensing the entire thing as GPL/LGPL might be a good idea too.

* This is assuming "the copyright owner" is just one person or a small group of people who agree on doing such a thing.

Re:Btrfs: kill off ext# please!

[bill_mcgonigle]

BSD, Solaris and OSX all support UFS, as does Linux.... Linux also supports the hfs+ filesystem currently used by OSX, not sure if bsd/solaris do but there are bsd licensed drivers for it so no reason not to.

Linux only sort-of, depends on flavor. I can't reliably mount a CF card r/w from pfSense (FreeBSD) under linux.

Re:Btrfs: kill off ext# please!

[moosesocks]

I think he was referring to the parent. The whole point of the BSD license is to not give a rat's a** about who does what with the code.

One could easily add an extra clause to the standard BSD license that states that any derivative work must be fully compatible with the reference Btrfs implementation in order to bear the Btrfs name. Many projects include such naming clauses.

This sidesteps the "extend and embrace" problem completely.

Re:Btrfs: kill off ext# please!

[sandGorgons]

Do check out ZFS-Fuse . Development has accelerated over the past couple of months and it is very usable with decent performance.

Atleast one commercial offering is using ZFS-FUSE in its products - however, no idea whether it is using a custom non-community build.

I am using ZFS-Fuse on Ubuntu Hardy with 2 hard disks in a mirrored configuration, serving its files over Samba. We do incremental backups everyday (which the filesystem supports) and are quite happy with it.

Re:Btrfs: kill off ext# please!

[tehcyder]

Oh sorry you're informed AND insane.

Welcome to slashdot...

Big speedups for media workstations..

[delire]

This 'Per-backing-device writeback' is pretty significant. I'm sure the feature film and database industries will love it especially:

The new system has much better performance in several workloads: A benchmark with two processes doing streaming writes to a 32 GB file to 5 SATA drives pushed into a LVM stripe set, XFS was 40% faster, and Btrfs 26% faster. A sample ffsb workload that does random writes to files was found to be about 8% faster on a simple SATA drive during the benchmark phase. File layout is much smoother on the vmstat stats. A SSD based writeback test on XFS performs over 20% better as well, with the throughput being very stable around 1GB/sec, where pdflush only manages 750MB/sec and fluctuates wildly while doing so. Random buffered writes to many files behave a lot better as well, as does random mmap'ed writes. A streaming vs random writer benchmark went from a few MB/s to ~120 MB/s. In short, performance improves in many important workloads.

Is this a hoax, or what?

[Sockatume]

rewrite of the writeback code

So you didn't de-lace the interace or uncabulate the turboencabulator? I'm now about 85% convinced that the open source movement is just making shit up.

Re:Is this a hoax, or what?

[Anonymous Coward]

sure it can do all that stuff... but does it run linux?

Re:Is this a hoax, or what?

[riffzifnab]

and I'm now about 93.8523% convinced you're making up statistics.

Re:Is this a hoax, or what?

[Abreu]

Linus reversed the polarity of the electron flow

Re:Is this a hoax, or what?

[butalearner]

So you didn't de-lace the interace or uncabulate the turboencabulator?

Dude, don't be ridiculous, of course they uncabulated the turboencabulator, how else could they contraplectify the apoplectifier?

Re:Is this a hoax, or what?

[Anonymous Coward]

Well, you would be 100% wrong. He just estimated his confidence in his belief being true, which isn't a statistic, just an opinion.

does KSM mean the death of Xen?

[trybywrench]

If KSM puts the KVM module on par with Xen in terms of performance then I think the writing is on the wall for Xen's demise.

Re:does KSM mean the death of Xen?

[1s44c]

If KSM puts the KVM module on par with Xen in terms of performance then I think the writing is on the wall for Xen's demise.

No. Not at all. KSM saves memory but hurts performance. It shares memory across virtual machines to save memory.

Xen can't share memory across virtual machines, it's just not put together like that.

Performance is about identical for KVM and XEN.

Re:does KSM mean the death of Xen?

[FlyingGuy]

Being somewhat ignorant of the inner workings of XEN, VMWare, KVM and the like the very idea that VM's would share memory at all seems rather risky in terms of them being sandboxed from each other. Beside a hypervisor being able to allow many VM's to run basically any OS, it would also seem that there is a security element involved eg: running Windows in a VM and Linux in another and NetWare in yet another the three would not have the ability to know the other were there and therefor be safe from being hacked into from Windows malware or affected by an ABEND ( basically a kernel panic ) from NetWare.

I can see the Hypervisor being able to move each VM's memory footprint around, but why would you desire a VM to mingle memory with another VM, or am I missing some salient fact(s) and or point?

Re:does KSM mean the death of Xen?

[diegocg]

Note that KSM also works for Xen.

KVM may not mean the death of XEN, but it has been a long time since it replaced it as the de-facto Linux virtualization system.

Re:does KSM mean the death of Xen?

[Bert64]

Instead of storing multiple copies of the same data in memory, it stores a single read-only copy and points the others to it. If you try to write to it, it traps, creates a new read/write instance which is exclusive to you and then points you at it...

Shared libraries work in much the same way. Shared libraries been implemented pretty securely for many years now.

Re:does KSM mean the death of Xen?

[FlyingGuy]

I grock the lib concept, each program gets it's own data segment and the code is run in a single image ( if you will ) and that conserves memory. Each data area is private to the program unless they explicitly share it through some IPC mechanism.

This is interesting as it seems like a way to write malware. If I wanted to deliberately run the machine into the ground I could just look for those data area's and keep attempting to write to them and force the OS to keep duplicating them over and over again. Now you could do the same thing by just having a program continue to allocate memory for itself, but this seems to a be a way to do so without having anything being able to detect that you are allocating memory on your own, just using the OS as your proxy to eventually suck up all the machine resources.

Re:does KSM mean the death of Xen?

[Hatta]

KSM saves memory but hurts performance.

If I have plenty of memory, can I easily disable KSM?

Re:does KSM mean the death of Xen?

[ichigo 2.0]

This isn't possible, as the guest OS cannot get more memory allocated to itself than has been assigned to it. Let's say you have five guest machines that "share" all their memory through this new feature. If you gain access to one of the machines and completely rewrite all the memory it has inside it, you still have only doubled the memory usage as the host os allocates the new memory for the guest os. Rewriting in the same memory multiple times will not increase memory usage any more than that as it will still be writing in the same memory area the host os allocated. Even if the attacker was able to gain access to all five machines and rewrote their memories, the host os would still only have five times as much memory usage as in the beginning (which is pretty much the way it is now without KSM).

So worry not, this should not be exploitable. *knocks on wood*

Re:does KSM mean the death of Xen?

[Anonymous Coward]

If I have plenty of memory, can I easily disable KSM?

It's easier than that: You can simply not enable it in the first place.

But yeah, assuming that it will (at some time) become the default, you will most likely (I have no reference to a credible guarantee) be able to disable it. I can't see any reason for not allowing you to do so.

Re:does KSM mean the death of Xen?

[Anonymous Coward]

If KSM puts the KVM module on par with Xen in terms of performance then I think the writing is on the wall for Xen's demise.

No. Not at all. KSM saves memory but hurts performance. It shares memory across virtual machines to save memory.

Xen can't share memory across virtual machines, it's just not put together like that.

Actually, this exact feature is scheduled for the next release of Xen, plus a few other neat tricks such as sharing memory at sub-page granularity and transparently compressing pages which aren't used very often. See, e.g., here for more details.

(And, yes, it is kind of vapour ware at the moment, but it's slightly less vapourish than normal in the sense that actual honest-to-God working code does exist somewhere, even if it's not production-quality yet.)

Re:does KSM mean the death of Xen?

[Anonymous Coward]

Actually, there's a research paper based upon sharing memory across virtual machines in Xen... Provides quite a benefit at a small but appreciable cost. Check it out:

http://www.usenix.org/events/osdi08/tech/full_papers/gupta/gupta.pdf

Re:does KSM mean the death of Xen?

[Anonymous Coward]

This was gone over above, KSM means the death of you decadent western nerds and you socialist 'FOSS' operating system :O

Re:does KSM mean the death of Xen?

[1s44c]

KSM saves memory but hurts performance.

If I have plenty of memory, can I easily disable KSM?

Disabling or enabling this should be no harder than one command line option to qemu.

Disabling it is the right thing to do if you don't need the memory saving benefit. It will waste CPU and throw your stuff out of data caches. I think most people will be more concerned with saving memory though, I know I am.

Re:does KSM mean the death of Xen?

[pasikarkkainen]

If KSM puts the KVM module on par with Xen in terms of performance then I think the writing is on the wall for Xen's demise.

No. Not at all. KSM saves memory but hurts performance. It shares memory across virtual machines to save memory.

Xen can't share memory across virtual machines, it's just not put together like that.

Performance is about identical for KVM and XEN.

Actually Xen has some form of memory overcommit nowadays.. check tmem patches from Oracle. I think they'll be part of upcoming Xen 4.0 release.

People work on the "easy" problems

[Fished]

Like the strip, and it raises a valid point. The bottom line is that kernel development advances more quickly than user interface and applications for the same reason that physics advanced more quickly than say ... psychology. That is, because developing a faster kernel is a much easier problem than developing a fun, usable desktop environment. It's easier to write, easier to test, and easier to debug. People tend to gravitate towards problems that they think they can solve--and ignore the problems they don't understand or don't want to deal with.

Personally, I think that the best way forward for Linux on the desktop would be to take GNUstep to the next level. There's a LOT of code there already written, and with a bit more work you might be able to have source-level compatibility with Mac OS X--which would give you access to a bunch of commercial apps. And, most importantly, the ability of the OpenStep API to produce a world class desktop--best in the world in fact--is proven. After 10 years, I don't think that either KDE or GNOME have really done all that much for Linux on the desktop... it's time to try a different approach.

Of course, I'm just kibbitzing, not bringing code. So what right do I have to say anything?

Re:People work on the "easy" problems

[tulcod]

Looks like you didn't get your psychology right. The problem is that creating a desktop environment is, in fact, much /easier/ to create than it is to enhance the kernel, and that makes it extremely boring. Desktop environments are trivial, but dull, to make. They are a perfect example of a job you should be getting paid for.

Re:People work on the "easy" problems

[suggsjc]

People tend to gravitate towards problems that they think they can solve--and ignore the problems they don't understand or don't want to deal with.

I think that should have read

Engineers tend to gravitate towards problems that they think they can solve--and ignore the problems they don't understand or don't want to deal with.

Re:People work on the "easy" problems

[Brian Gordon]

Yesterday I started xmoto while a video was playing and my system slowed to a crawl. I could move the mouse at a frame every few seconds, but nothing else responded, even trying to change to a virtual console and zapping X.

I don't know who's to blame but I do know that this wouldn't happen to a Haiku or SkyOS user. I'm tired of the Linux kernel; it's really not that great. Everyone seems obsessed with C, going as far as to spawn these kind of monstrosities just to force modern features into a traditional platform. Give me a stable microkernel with user-mode graphics so the above bug can never happen. Don't break it every few months and fix it later like Linux.

Maybe we just need a systems technology reboot. So much of GNU/Linux is just horribly broken and needs a sanity check, particularly desktop environment stuff. We can probably do a much better job if we start over.

Re:People work on the "easy" problems

[DragonWriter]

Personally, I think that the best way forward for Linux on the desktop would be to take GNUstep to the next level.
[...]
After 10 years, I don't think that either KDE or GNOME have really done all that much for Linux on the desktop...

Purely technical solutions to marketing and promotional problems rarely work, so its unsurprising that GNOME and KDE have done much for Linux on the desktop, since their marketing and promotional efforts are pretty minor. Of course, switch technical approaches to focus on GNUstep has the same problem.

And, most importantly, the ability of the OpenStep API to produce a world class desktop--best in the world in fact--is proven.

That the Mac OS X desktop is "best in the world" is a subjective statement on a matter of taste, not a "fact".

In terms of facts, on the marketing and promotional end where Linux has been unsuccessful, Mac OS X has been more successful, though far from as successful as Microsoft Windows.

Re:People work on the "easy" problems

[Anonymous Coward]

You are confusing tough (challenging) with tough (laborious).

Re:People work on the "easy" problems

[shutdown -p now]

That is, because developing a faster kernel is a much easier problem than developing a fun, usable desktop environment.

I disagree, it's not an easier problem. It is, however, a much more interesting problem to solve, especially to skilled hackers.

One other aspect here is that the target audience is bigger for the kernel. Desktop uptake is still very low, but kernel is used by any device that runs Linux, whether it's a router, a smartphone, a server, or a netbook. This has a side effect of kernel hacking being better financed than desktop development, as there are more commercial players interested specifically in the kernel, who couldn't care less about KDE or Gnome.

Re:People work on the "easy" problems

[theCoder]

...developing a faster kernel is a much easier problem than developing a fun, usable desktop environment.

While I agree with tulcod's response -- kernel development is usually much harder than desktop development. However, there is one important difference. A faster kernel is a measurable goal. While you might be able to make a "fun, usable desktop environment" for a single person, and maybe even for a good percentage of the population, you will never, ever satisfy everybody. Half the people want more options and more control and half want a simpler, less cluttered interface. Some people want freedom, others want to lock it down. Some people want blue, others want brown. Some people want menus, others want icons. If you're doing your job well, you can compromise enough to make everyone satisfied, if not completely happy, but that's really hard to do. A faster kernel seems easy by comparison.

Personally, I think Gnome is going too far down the simplification road. Every release it seems they take away more features and options. This last they struck at GDM (the login screen). Before that, they nerfed session management. A few releases earlier they took away screen saver control. And I'm still annoyed that I can't have different backgrounds on different desktops, and I that's been broken since Gnome 2.0!

Of course, I'm just kibbitzing, not bringing code.

Heh, agreed, same here :) Maybe one of these days Gnome will push me over the edge and make me write my own WM.

Re:People work on the "easy" problems

[javilon]

Well, GNOME and KDE ( I prefer one of them but it is not relevant to this post ) have done lots for Linux on the desktop. I have been running it for a number of years because I find it more pleasant to use than Windows. And I am not alone.

And the millions of people using it are doing so against active attacks from a number of organizations. Mainly closed software companies, and also (mainly in the past) political organizations and governments.

Re:People work on the "easy" problems

[abigor]

Very true. Desktop Linux is a bit player when compared to the overall use of Linux, and it's truly a hobbyist's domain.

Re:People work on the "easy" problems

[Anonymous Coward]

what do you propose? to rewrite the kernel in python? sorry, but something needs to run on the hw itself eventually, nevermind that the language used has little to do with your complaints. User mode graphics are ok for basic desktop use, but forget it if you want decent performance for 3d.

do you know any skyOS or haiku users? I don't.

Re:People work on the "easy" problems

[Anonymous Coward]

"The bottom line is that kernel development advances more quickly than user interface and applications for the same reason that physics advanced more quickly than say ... psychology."

Wow!! I guess, kernel development has nothing to do with UI in same way as physics have nothing to do with psychology. So I guess that analogy, being a double negative, is correct. Just wow.

The only thing in common between psychology and physics is you can call both science if you stretch the definition of science. In same way, kernel and UI are as similar since they both generally run on a given computer system.

"source-level compatibility with Mac OS X - which would give you access to a bunch of commercial apps"

No it would not. You are looking for ABI compatible, not source-level compatible (or even API compatible). And who the hell would want to duplicate the nightmare in OS X programming where Apple couldn't even decide if they were going to go with Carbon or Cacoa for 6t4-bit? Then of course, they axed one in-spite of what they were saying previously.

    http://labs.trolltech.com/blogs/2008/03/03/qtmac-cocoa-port-alpha-released/

"After 10 years, I don't think that either KDE or GNOME have really done all that much for Linux on the desktop... it's time to try a different approach."

Please, entertain us. Go on, what new UI paradigm have you managed to come up with and why haven't I heard of it? Or is it Microsoft's fault?

"Of course, I'm just kibbitzing, not bringing code. So what right do I have to say anything?"

Exactly. It's just more whine from people that do nothing. Have you ever thought that people run windows is because that's all they know and because 3rd party apps come for Windows? And 3rd party apps don't come for Linux because there isn't enough market share. And there isn't enough market share because there is no 3rd party apps for Linux.

Re:People work on the "easy" problems

[bcmm]

You're missing the bit where Flash is closed-source and the people that want it to work properly can't make it happen, whereas the people who can make it work don't want it to happen.

Re:People work on the "easy" problems

[mcrbids]

Desktop uptake is still very low, but kernel is used by any device that runs Linux, whether it's a router, a smartphone, a server, or a netbook. This has a side effect of kernel hacking being better financed than desktop development, as there are more commercial players interested specifically in the kernel, who couldn't care less about KDE or Gnome.

If I hadn't already replied in this article, I probably would have modded you up. This point is hard for many to understand, but it's quite possible that the total number of Linux kernel installs may already rival the total number of Windows installs, since the types of devices that use Linux are so variable.

Along with its home turf of Internet servers that are otherwise invisible to the end user, it's commonly used on routers, printers, phones, sensors, calculators, digital cameras and camcorders, DVRs and "media centers", so-called "network-accessible" hard drives, digital picture frames, and too many other fixed-function "embedded" devices to name. The combination of stability, driver availability, consistent programming interface, footprint, and cost (free!) is a 1-2-3-4-5 punch to nearly any industry that relies on information processing.

Linux is just about *everywhere* but the desktop nowadays.

Re:People work on the "easy" problems

[Hatta]

Making a good desktop environment is extremely challenging, as shown by the fact that no one has made one yet.

Re:People work on the "easy" problems

[serviscope_minor]

I agree with you about the monstrosoties. The kernel is full of classes and virtual functions. C++ provides exactly those features without the syntactic overhead and assosciated bugs. Because they are a built in feature, everyone does them the same way. Not only that but it uses less memory (kinder on the cache) because it stores only one copy of each vtable. It also provides *exactly* the same features as C for when they are really needed.

But people keep complaining about the "complexity" of C++ while happily implementing 3/4 of the C++ compiler by hand in C every time they want to do something simple like instantiate a class. Not only that but they come up with a vaeiety of strange and perverse arguments as to why C is better. See the infamous LKML C++ threads for some fine examples of straw men, ad homenm and general wrong headedness.

As for user mode graphics, for many years, X11 provided pretty much pure user mode graphics. The faster 3D accelerated ones tend to be more tied to the kernel.

Re:People work on the "easy" problems

[V!NCENT]

"I'm tired of the Linux kernel; it's really not that great."
Linux has always been more or less the entire FLOSS pool. Nothing in it is meant for one goal. You get all these different goals. Yes it's far from elegant in that respect.

But... It's an ecosystem where people throw in stuff. It's like evolution. You start with crap. Make countless modifications. The distro's then choose what's important. Everything that sucks dies. Everything is better than the version before sticks.

Linux was a bunch of crap realy, but it is now starting to realy mature.

GNU/Linux in itself isn't great but the entire process is.

It takes a while to go from just one fish in the ocean to a human form (go ahead, make jokes), and then even beyond.

Linux can run on so much different HW, just because it has no clear goal. No goal... it's a bunch of junk you say? Well it runs great...

Re:People work on the "easy" problems

[21mhz]

I'm tired of the Linux kernel; it's really not that great. Everyone seems obsessed with C, going as far as to spawn these kind of monstrosities just to force modern features into a traditional platform.

What, has GObject made it into the kernel?..

Ahh, IHBT.

Re:People work on the "easy" problems

[aztracker1]

I wouldn't mind seeing GnuStep underpinnings with a Mono binding that adds in XAML support. Seems to me like a decent base platform for higher level abstractions.

Re:People work on the "easy" problems

[Anonymous Coward]

This has a side effect of kernel hacking being better financed than desktop development, as there are more commercial players interested specifically in the kernel, who couldn't care less about KDE or Gnome.

Interesting, innit? It's not that you get what you pay for, but even that there's nothing to be gotten when no one pays at all.

Only paid devs are pros. Fact. Not that hackers in basements suck that much at coding, more that as soon as they get any marketable skills, they get a JOB and stop coding for free. Then one in a billion ends up getting paid to insert useless features in Linux Corporate Editions (AppArmor? Database snapshots?) instead of Flash, mp3/divx codecs, 3D support, and Linuxes for Netbookses that DO wake up the WiFi card when unsuspended.

Bah, it's not like anyone will ever get Linux pre-installed on everyone's computer. Give me OSX any day...
Face it :You'd have to be very high on very dangerous drugs to even begin to think that any Linux desktop is anywhere near as usable as OSX. That's a fact too. And Apple took a BSD, an even more obscure and arcane Unix than Linux, and turned it into the world's best OS, the one everyone wet-dreams about when they're banging their chairs on their keyboards because of Windows or Linux. Why doesn't Canonical or some such get their heads out their asses and build a dev team that will simply Make It Work Well Together? No, not ever. They couldn't even sell it. "Other means of" getting nothing for your work. Yeah. IBM pays for Apache, Novell for Samba, blah blah blah - no one pays for porting Aqua and Finder or anything as usable. Strange that.

Using Linux on a desktop is fighting a losing battle from the start. Linux is a great educational project. It certainly can be used in embedded devices, as long as you don't use code that some BigCorp pays for, because those (only those) have Teams Of Lawyers On Salary that will bankrupt you in court before they oblige you to publish the code (so you'll disclose the changes their moron code-pissers couldn't drool by themselves). Great for servers, because if you blow enough cash you can get the nice little labels that say everything is compatible together AND then call the Certified Technician when it doesn't work.

But for clients? "Oh noes, we'd have to maintain tens of drivers for tens of chip models! Woes!" Tens? Mobo chipsets are not that varied. Intel distributes about four versions of INF update for each Windows from XP to 7, supporting ICH4 to 10 (and if you have a 440BX, it's about time you trash your PII), nvidia has stopped making them, VIA and SIS each have one arch and a couple variations per gen, all driver-compatible in such a majority of specs that only coding the common ones would lose about zero functionality. Anyone else?
Oh, yeah, graph cards. Old mantra about protecting patents that might be lurking in the code... not their own IP, blah blah blah... shut up, gimme that NDA *signs* now gimme the specs, now let me torrent them, okay now they're public knowledge and a billion geeks duplicate them. Come protect not-quite-your IP now.

Lost the point several times. Bah. Linux sucks, buy a mac.

Re:People work on the "easy" problems

[DragonWriter]

I disagree, it's not an easier problem. It is, however, a much more interesting problem to solve, especially to skilled hackers.

Whether or not its an easier problem to solve, overall, its an easier problem for the kind of people who actually write code to define concretely, and validate solutions to, since the skill set needed to do that with that problem is closely related to the skill set of programmers. This is important, because to successfully solve a problem (or, in the case of problems that progres can be made by degrees, to produce a series of improvements in a problem area) requires the ability to concretely define what it means to address the problem, and the ability to validate that something actually meets that definition.

One of the things about the open source community is that, by its nature, its heavily weighted in favor of coders. Concretely defining requirements for a "fun, usable desktop environment" takes human factors engineering skills that are almost completely orthogonal to programming (and you still need programmers to write the code.) Consequently, I think that the "faster kernel" problem is a lot easier problem for the open source community to solve, even if its not an easier problem in some kind of general sense.

Re:People work on the "easy" problems

[Anonymous Coward]

Personally, I think that the best way forward for Linux on the desktop would be to take GNUstep to the next level.

Sounds like Etoile. Not source compatible with OS X, but some of the Apple stuff makes its way over (Clang, LLVM).

Re:People work on the "easy" problems

[ToasterMonkey]

No it would not. You are looking for ABI compatible, not source-level compatible (or even API compatible). And who the hell would want to duplicate the nightmare in OS X programming where Apple couldn't even decide if they were going to go with Carbon or Cacoa for 6t4-bit? Then of course, they axed one in-spite of what they were saying previously.

        http://labs.trolltech.com/blogs/2008/03/03/qtmac-cocoa-port-alpha-released/

It was not a this or that decision, Carbon is a legacy API that got left at 32-bit, and Cocoa was always the one to use going forward. In the end it forced developers like Trolltech to port to Cocoa which is A Good Thing TM.

Re:People work on the "easy" problems

[Viol8]

"X11 provided pretty much pure user mode graphics."

User mode graphics running as root - which meant if X went down it could still take your box with it. Or more likely just lock up the video display which for most people is the same thing.

Re:People work on the "easy" problems

[Viol8]

"Maybe one of these days Gnome will push me over the edge and make me write my own WM."

I did just that because I got sick of bloatware desktops. Problem is that even with the Xlib programming manuals at hand and example code from another WM I still couldn't create something that appeared to work properly. It was 95% there but had enough issues to make it a pain to use so in the end I went back to KDE. There are so many tricks and gotchas and undocumented things you have to know when doing low level Xlib that its really more hassle than its worth. And thats if you just stick to core Xlib - if you start using extensions then you can kiss your sanity goodby.

Re:People work on the "easy" problems

[buchner.johannes]

Personally, I think that the best way forward for Linux on the desktop would be to take GNUstep to the next level. [...] After 10 years, I don't think that either KDE or GNOME have really done all that much for Linux on the desktop... it's time to try a different approach.

First step: Make GNUstep (1) look nice and (2) take care of usability. That is what you missed about KDE and GNOME: They plan and think about usability, a whole concept of a desktop, and nice look & feel.
Otherwise you won't have users and thus no developers.

Re:People work on the "easy" problems

[Tim C]

That is, because developing a faster kernel is a much easier problem than developing a fun, usable desktop environment. It's easier to write, easier to test, and easier to debug.

It's also pure coding - in order to develop an aesthetically pleasing, easy to use desktop environment you'll also need graphic design, usability and information architecture skills. Most programmers don't have much in the way of them, and there are fewer professionals in those fields who are willing to give up their time to work on FOSS projects.

Re:People work on the "easy" problems

[Frantactical Fruke]

I wish someone would finally notice that the Flash player is a closed, commercial product, so the fact that its Linux version sucks hardly reflects on the quality of free software or its development model. Strangely enough, the only piece of software on my computer that truly sucks dump trucks through a straw is that Flash plug-in. It also happens to be the only bit of non-free software here...

Re:People work on the "easy" problems

[Trepidity]

I don't think it's solely limited to engineers--- there's a tendency in most fields to over-emphasize the importance of your area of expertise and assume others are of marginal importance or can be somehow safely bracketed. I'm in an area of academia with a lot of CS/humanities collaboration (and where there should be more), and it's quite evident on both sides--- the CS people generally don't understand or take the humanities parts seriously enough, and the humanities folk tend not to understand or take the tech side seriously enough. That's why a huge proportion of "new media art" is either people with solid tech and half-assed art, or people with good art ideas and half-assed tech.

Re:People work on the "easy" problems

[marcosdumay]

Oh, yes. Most people don't wat to solve anything at all.

Re:People work on the "easy" problems

[tehcyder]

Making a good desktop environment is extremely challenging, as shown by the fact that no one has made one yet.

Apart from Microsoft Bob, of course.

Re:People work on the "easy" problems

[Anonymous Coward]

That is, because developing a faster kernel is a much easier problem than developing a fun, usable desktop environment.

One other aspect here is that the target audience is bigger for the kernel. Desktop uptake is still very low...

BECAUSE THE DESKTOP IS SO BLOODY UNUSABLE!!!

Jeez.

Read what you just wrote...talk about not seeing the wood for the bloody trees.

'Fix' the desktop problem and you will get more adopters..you don't need a fancy-shmancy kernel to browse email, surf the web or even play your flash videos.

I'm still with the XKCD had it and still has it bang on as far as I am concerned.

Re:People work on the "easy" problems

[shutdown -p now]

BECAUSE THE DESKTOP IS SO BLOODY UNUSABLE!!!

Jeez.

Read what you just wrote...talk about not seeing the wood for the bloody trees.

'Fix' the desktop problem and you will get more adopters

It's precisely the vicious circle I was speaking of. Desktop Linux is used by very few people because it's worse than what competitors (MS and Apple) have to offer. To make it better, more talented people need to work on it. But talented people tend to go for either: 1) fame, so they work on things that are used by more people (i.e. kernel & server-side, not desktop; 2) money, so they get hired by some company which makes them work on things for which there is existing strong demand (i.e. kernel & server-side); 3) "just for fun", in which case they tackle the problems which are harder from engineering POV (i.e. kernel & server-side - massive scalability or somesuch).

Ultimately, you end up with few good developers working on the desktop. Perhaps what's worse is that you end up with no commercial players working on the desktop, because those guys would at least have some coherent usability goals and single-mindedly pursue them, not add more eye candy and clever hacks.

KMS

[shentino]

Kernel Mode Switching is great except for the fact that all 3 major video card vendors decided to nix VGA console support.

Re:KMS

[diegocg]

And how is that related to KMS? KMS is about moving mode switching to kernel, which is neccesary even for the most modern 3D cards. And the memory manager that was implemented to make it possible is also neccesary to implement correctly a modern graphic stack.

Re:KMS

[shentino]

Yes.

However, ATI, NVidia, and Intel have all three decided to deprecate the VGA text console if used with KMS enabled.

So if you use KMS with one of those cards, kiss text mode goodbye becasue the powers that be refuse to support it.

Re:KMS

[ettlz]

So don't load the DRI modules. Or are they doing away with the video BIOS all together?

Re:KMS

[True Grit]

kiss text mode goodbye becasue the powers that be refuse to support it.

The ATI/Radeon driver, at least, provides its own hi-res console mode (all in the kernel, not dependent on Xorg, works with fbcon to act as a framebuffer driver), so why would you want the old VGA (low-res, low-refresh) mode anyway?

For special/embedded systems, they can always leave out that specific driver (ATI/NV/Intel) that does KMS, and keep the old VGA console code, its still in the kernel & available as an option.

You haven't lost anything, you just can't use both at the same time now, e.g. you can't (AFAICT) have the old VGA and the new KMS support builtin to the kernel at the same time.

Re:KMS

[diegocg]

Duh. You can use instead a fb-based text console, so where is the problem?

Re:KMS

[shentino]

Loading fbcon has a couple of ugly side effects

1. Since we're pushing pixels instead of cells, things slow down an order of magnitude or two

2. Once the console drivers have bound onto fbcon and off of vgacon, there's no way to get vgacon back without rebooting.

Re:KMS

[shentino]

fbcon sucks

Re:KMS

[Anonymous Coward]

You use a framebuffer, the way every other modern computing platform other than the PC has always done it. Even if VGA itself goes away, you can easily emulate it on top of OpenGL. Take a look at cell phones: lots of them are running on 3D hardware right now and nothing else.

Re:KMS

[PeterBrett]

[Citation needed]

Re:KMS

[True Grit]

1. Since we're pushing pixels instead of cells, things slow down an order of magnitude or two

Except the framebuffer driver you're now using is hardware-accelerated, which more than makes up for switching from chars to pixels. This was why so much effort went into all those chip specific fb drivers that are available. They're all much faster & more convenient than plain & dumb vgacon.

Sounds like you've never seen one of those hardware-accelerated console modes, or you'd realize just how slow/inefficient plain vgacon actually is. Its a night vs. day kind of difference.

2. Once the console drivers have bound onto fbcon and off of vgacon, there's no way to get vgacon back without rebooting.

That's true, but again that brings us back to my original question: why would you want to keep vgacon around when you have a hardware-accelerated, hi-res, hi-refresh fb driver replacement already there?

This, a unified, lowlevel graphics driver that is used by both fbcon & xorg, which gives the same advantages to both the Xorg and console modes, and now allows more cooperation between the two, is what many people have been wanting for a *long* time.

One file system to rule them all

[DrYak]

On the downside, I'm peeved that Btrfs is GPL licensed, which will prevent it from becoming "the one true filesystem" from here on out.

Well, ZFS itself has a GPL-non-compatible license, but that doesn't prevent it from being usable in Linux as an independent user-space process through FUSE.
The same approach could be imagined under non-GPL-compatible OS: have the GPL implementation as a standalone userspace daemon.
(Which is not a bad idea - give more freedom to upgrade)

Windows users will be stuck with NTFS

No matter what. Even if some kernel guru released a tri licensed LGPL/BSD/Proprietary perfect file system, Microsoft will still be using NTFS and promising WinFS soon for whatever the next version of Windows is.
They have a strong case of NIH-Syndrome.

None of them will be compatible, and FAT32 somehow remains the only viable option for removable media.)

For removable media, UDF could be a good candidate too. It's getting widespread availability, specially since Microsoft added support for writing on Vista and Win7.

Re:One file system to rule them all

[Dr.Dubious DDQ]

"For removable media, UDF could be a good candidate too. It's getting widespread availability, specially since Microsoft added support for writing on Vista and Win7."

Getting slightly off-topic, but after the FAT patent-trolling recently this interests me.

I went and dug up the sadly-neglected udftools package and installed it. Sure enough, the following command (found with a bit of Googling) seems to produce a filesystem on my SD card that can be read from and written to just fine by Linux, Mac OSX (Leopard), and Vista at least:
mkudffs --media-type=hd --blocksize=512 /dev/sdc
(Obviously one should substitute the appropriate device there...) No partition table needed. I'm not sure if the "blocksize=512" is necessary, but it appeared that it might be.

The only real problems I see are the fact that udftools appears to have been abandoned half a decade ago, so there is no fsck tool[1], and that all those elderly Windows XP machines apparently can't write to this format without 3rd-party software.

Looks promising to me, if development on udftools is picked up by anyone (and in the meantime is definitely usable), so thanks for the tip...

[1] Well, udftools does include a "udffsck", but it quite literally does nothing - the source code is a stub that consists entirely of a single "void main()" function that returns 0 immediately. Mac OSX has a newfs_udf, but also appears to lack a way of repairing/validating the filesystem.

Re:One file system to rule them all

[aztracker1]

I really think that MS has gotten a lot better about their NIH approaches. Though they tend to buy out, and hire in anyone developing cool open-source that happens to work in windows.

Re:One file system to rule them all

[Anonymous Coward]

On the downside, I'm peeved that Btrfs is GPL licensed, which will prevent it from becoming "the one true filesystem" from here on out.

Well, ZFS itself has a GPL-non-compatible license, but that doesn't prevent it from being usable in Linux as an independent user-space process through FUSE.
The same approach could be imagined under non-GPL-compatible OS: have the GPL implementation as a standalone userspace daemon.
(Which is not a bad idea - give more freedom to upgrade)

err... What? And you put up with THAT?
Do you know how horrible it is that your CPU is doing nothing but manage I/O instead of letting specialized controllers do that? It's like you're back before DMA. When $200 chips were doing IO all day because no one could actually buy SCSI and no corp in the PC industry would dream of building an efficient way to access storage media, since they would sell no more SCSI.
And that's not even beginning to talk about the - yes, the performance. You LIKE a filesystem that you need an "helper application" to access... if I was your boss in IT I'd fire you with extreme prejudice. Performance is non-negotiable.

Seriously, what if I just sent a patch for Linux "fixes bug #666-666-666 : using FS drivers in Userspace is teh evil, because it CRAWLS LIKE DEAD SLUGS" that implements ZFS as a kernel driver? They'd reject it? Over license shit? What if it was NTFS-3G? (btw, for what horrible reason is that not in kernel yet? NIH, or "we love to make our users suffer"?)

Windows users will be stuck with NTFS

No matter what. Even if some kernel guru released a tri licensed LGPL/BSD/Proprietary perfect file system, Microsoft will still be using NTFS and promising WinFS soon for whatever the next version of Windows is.
They have a strong case of NIH-Syndrome.

None of them will be compatible, and FAT32 somehow remains the only viable option for removable media.)

For removable media, UDF could be a good candidate too. It's getting widespread availability, specially since Microsoft added support for writing on Vista and Win7.

Yeah. And Linux has no "I won't let you access your files" bug this time? (NTFS-3G + mount point permissions) This is gonna be SO MUCH FUN in the next five years when I'll still have to repair computers (read : reinstall XP Pirate Ed. and deactivate Windows Update), when I'll have UDF usbkeys and NTFS HDs and incompatible OSes. Glad to leave that world for embedded development.

kvm and ksm

[xming]

kvm (with a little patch) supports it, running it right now with 5 guests and have 53K pages which are shared. # cat /sys/kernel/mm/ksm/pages_sharing 53714 That's ~200MB for about 1,5GB memory used on the host. Now I can't figured out how many times those pages are shared, so I can't calculated the actual memory saved (it's between 200MB and 4x200MB).

ATI support

[PenquinCoder]

Finally?? I mean how long has ATI been open source, and we're just now getting support for the newer R600/R700 devices now?? I hope this kernel release also addresses the issue of the HD audio in the combined A/V R700 chips.

Re:ATI support

[Anonymous Coward]

The 2D specs were released in September 2007. The 3D specs were released in January 2009. Drivers do not write themselves immediately just because the specs are out, it still takes some time. But it's getting there, and they won't go away like the closed drivers will, the moment the manufacturer feels it's no longer profitable to maintain them.

Re:ATI support

[Dan Ost]

Mod parent +1 informative.

How come I never have mod points for a truly deserving post?

Re:ATI support

[Anonymous Coward]

The developers are also rolling out modernized code for all the radeon chips. So there was more than a little catching up to do. I bet R800 is supported even more quickly. This is on top of releasing a closed source driver.

Yeah, that is fine and all but the big question is

[Anonymous Coward]

Does it run Linux?

stable?

[burris]

Lots of new features, I thought the 2.X.Y versions where X is an even number are supposed to be "stable." There isn't even a 2.7 branch.

Re:stable?

[JSBiff]

I don't think they've been following that convention for like, oh, 6 years or something. You're right that once upon a time that was the way it worked, but afaik, not any more.

Re:stable?

[Changa_MC]

Lots of new features, I thought the 2.X.Y versions where X is an even number are supposed to be "stable." There isn't even a 2.7 branch.

Future versions of Linux will start with 2.6 -- only a major restructure and rewrite of the entire kernel would change that. We now denote changes using 2.6.Y.Z
This is analogous to the way all versions of MacOS are version 10.x (X.x??) There is no need to say the two point six part -- this is Linux kernel 32.

There is also no longer such a thing as stable/unstable branches, 2.6.32 is a new feature release, stability testing falls to the distributions.

Re:stable?

[ConceptJunkie]

They stopped doing that a long time ago. Now it seems that the stable and development versions are the same.

KSM

[svtdragon]

Linux 2.6.32 introduces what is called "KSM"

WHAT!? I know Linux users are pretty militant (myself among them), but to implement terrorism in the kernel?

Please tell me it's at least built as a *module* by default!

Re:KSM

[Trepidity]

It's actually that Linux users are sadomasochistic; KSM is a nested acronym standing for "kernel S&M".

HA!! Windows is at version 7 ALREADY !! Version 2?

[Anonymous Coward]

Windows is version 7. That's WAY MORE than linux. Linux suxors !! use old obsolete shit OS, dumasses to the max !!

Re:HA!! Windows is at version 7 ALREADY !! Version

[Anonymous Coward]

But is it?
A 7th version of a crap OS is still crap.

What i've been most curious about...

[pjr.cc]

Theres not alot of info around about it, but i'm dying to see the new dm-replicator module. Theres not huge amounts of information available about it and bits and pieces have been floating around for a while... it basically looks like something that could replace drbd as a more sensible mechanism of doing replication...

Personally I hope they let it do local replication as well cause the one thing i've always wanted is to replicate my laptops hd onto an occasionally-pluged-in usb disk with the ability to snapshot just the usb disk now and then... That would be fantasic.

Re:What i've been most curious about...

[pjr.cc]

actually, i just decided to have a look at patchwerk and as luck would have it, there was a patch posted that included doco.. Very worth reading... Cant wait till this makes it into mainstream.

http://patchwork.kernel.org/patch/63991/

multiplement implementations

[OrangeTide]

If Btrfs's design proves to be good, there is not a reason why there can't be both GPL and non-GPL implementations written for it. I think one of the things for universal filesystem to be successful is to have something that has more than one implementation.

FAT32 will have to die in the market when people get sick of files over 2gb getting truncated. The end is near for FAT.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Comment removed

[account_deleted]

Comment removed based on user account deletion

time saving makefile

[inode_buddha]

I'm very interested in the new make target. Specifically, "make localmodconfig". It seems that this new target will check your current .config, and also check whatever modules are currently loaded. It then creates a new config file which only builds the modules you are currently using. This could be a great time and space saving, as opposed to building everything and the kitchen sink as distros tend to do. It gives you a fairly easy and sane way to truly tweak your kernel to fit your box, or script it to fit a whole bunch of non-similar boxes.

Re:time saving makefile

[bcmm]

That's sounds potentially very useful, but beware that if it works the way you're describing it, it could remove, for example, support for USB MSC if your USB stick wasn't plugged in when you did it.

Re:time saving makefile

[gringer]

There's also a "make localyesconfig" that will be even more useful for me, particularly for removing the need for initrd. I can now do a "make localyesconfig", and not have to try to guess what particular combination of compiled-in options is required for the computer to start up, then add in the additional things as modules.

Re:time saving makefile

[ynef]

Fair enough, that obviously would be very bad. What if there could be a program that monitors your system for a period of time (including e.g. DBus events), and uses this to get a pretty good idea what your system actually requires from the kernel and provides a "smart" kernel config for you?

However, I realize that the main reason companies such as Canonical don't want this stuff to be in Ubuntu is because it would create a support nightmare. They want your kernel to be the same standard one that they use, otherwise dealing with "my program XYZ crashes all the time, I'm paying you, fix it now, dammit!" would be horrible, even more so than today.

Re:time saving makefile

[bcmm]

How would dbus events help to identify required kernel modules? The reliable way would be to make the kernel able to notify some userspace thing when a module is loaded; the easy way would be to have a daemon run lsmod every few minutes.

And why the Ubuntu reference? Ubuntu people usually aren't people that build their own kernels anyway.

Of course, even if you can monitor usage perfectly, that doesn't cover keeping a module around that you might use very rarely (having it not use up RAM being a major advantage of the module system). For example, you might keep support for HFS+ (MacOS X's filesystem) in a module in case you ever want to grab files from a friend's Mac-formatted iPod, but not use that modules for months at a time.

Re:HA!! Windows is at version 7 ALREADY !! Version

[mitoyarzun]

Windows is version 7. That's WAY MORE than linux. Linux suxors !! use old obsolete shit OS, dumasses to the max !!

I wish I had mod points, I find this funny!

Fine discussion

[Anonymous Coward]

A hilarious statement coming from a Loonix turd.

Dear sir,
What a wonderful argument !
I am interested in your ideas and wish to subscribe to your newsletter.

Wrong about JFFS2

[andrewd18]

JFFS2 is designed for unmanaged NAND flash, not flash cards with built-in controllers that emulate IDE drives. Therefore you can't use it on SD cards, CF cards, or anything that has a built-in memory controller.

See this Maemo development thread: http://www.gossamer-threads.com/lists/maemo/developers/36921

Let me be the first to say...

[Anonymous Coward]

This violates my patents. Let me go see which one I can claim against it!

Money train!

Re:HA!! Windows is at version 7 ALREADY !! Version

[Tubal-Cain]

7? Mac reached version 10 a long time ago.

Made in China

[FreakyGreenLeaky]

With comments like ...but it takes too many time to compile everything., I can see they're letting it slip that, secretly, the kernel has actually been developed in China all along. Linus is just a male stripper they hired to pose for pictures (with beer in hand).

Nice new features, but...

[Anonymous Coward]

Does it have support for smooth full-screen flash video yet?

Re:HA!! Windows is at version 7 ALREADY !! Version

[haruchai]

  It's worse than you think - the Windows official version is 6.1.7600; Linux is only at 2.6.32.

Re:HA!! Windows is at version 7 ALREADY !! Version

[Hucko]

Remember that Linus is European? The period is thousand place mark. The Linux kernel number system is actually shorthand for 2,600,320.

Reiser4 info here as of Nov 10 2009

[westyvw]

http://www.phoronix.com/scan.php?page=news_item&px=NzY4OQ