Submitted by Linux 3.0 has been released. Technically, the release of version 3.0 of the Linux kernel shouldn’t be too eventful, since the jump in version number doesn’t actually signify any huge change or whatever; the only reason behind the bump to 3.0 is to come to saner version numbering. Still, man, it’s like, totally version 3 of the Linux kernel.
So yeah, the year of Linux on the desktop never came, but in its place came a lot of other, possibly far more impressive stuff. The year of Linux on the server – dominating. The year of Linux on smartphones – Linux is godlike here. The year of Linux on high-performance computing – Linux aced this market. And a whole load of other things, like embedded uses for Torvalds’ baby.
Now we’re at version 3.0, but it’s a relatively minor release. “As already mentioned several times, there are no special landmark features or incompatibilities related to the version number change, it’s simply a way to drop an inconvenient numbering system in honor of twenty years of Linux,” Torvalds writes, “In fact, the 3.0 merge window was calmer than most, and apart from some excitement from RCU I’d have called it really smooth. Which is not to say that there may not be bugs, but if anything, there are hopefully fewer than usual, rather than the normal ‘.0’ problems.”
The changes include Btrfs data scrubbing and automatic defragmentation, XEN Dom0 support, unprivileged ICMP_ECHO, wake on WLAN, Berkeley Packet Filter JIT filtering, a memcached-like system for the page cache, a sendmmsg() syscall that batches sendmsg() calls, the setns() a syscall that allows better handling of light virtualization systems such as containers, new hardware support such as Microsoft Kinect and AMD Llano Fusion APUs, and many other drivers and small changes. Full changelog here.
Sure there have been kernel releases with more enhancements, but if you care for Xen dom0 then this is totally worth the new major release.
Way faster BTRFS and netfilter rules is also very important for some people.
I always hated the 2.6.32.57-wtf version numbering so this is a great release.
And you can follow Linus’ rants on G+ now, which is strange but fun.
Edit: And Happy Birthday, Linux. MS would also like to congratulate. http://www.h-online.com/open/news/item/Linux-receives-20th-birthday…
Edited 2011-07-22 09:13 UTC
That’s a weird video and doesn’t feel like “Happy Birthday” message at all.
Though I don’t know why I’m surprised by this.
I thought it was kinda neat and fun and a lot friendlier than you’d expect, all things considered.
I think so too. Any peace overture is nice at this point, though this “What went wrong” nonsense could have used some more PR work ie. “Microsoft and Linux got off to a rocky start, but the past is the past, and the future is now, and we predict great things from this powerful alliance…” So on and so forth. At least they’re not comparing Linux to the Matrix anymore.
G+, or the fact that you can follow his rants?
It’s actually pretty neat, I added him to my Circles there, makes it easy to follow his antics
Bah I can’t join Google+ yet because OSAlert uses Google Apps, so I don’t have a Google Profile >:(.
You can create a new gmail account and use that for google +. I’d send an invite.
I know, I already have an old regular Gmail account.
The problem is that I’d only get stuck, and would need to move to my real email address once Profiles IS added to Apps users. Kind of annoying.
I’m in a somewhat similar situation. It appears as if you can switch the google account that google + is associated. Not sure if that really works. I’m hesitant to try, but maybe I’ll get bored this weekend and see if it works.
This reminds me a conversation on a forum, someone interesting in trying Linux and asking where to download it. A geek responded http://www.kernel.org
True story.
What’s with the strange links?
It seems Tom has been hooked on LoL recently
Which is strange since I could swear he said gaming is for consoles and not PC (lol is pc only game)
leagueoflegends.com
I’m so yesterday.
I know, right? Like, wow. Man. 3.0. Dude, that’s like, wow, man. Cool.
Okay, serious question: If version numbers still meant what they used to [which they don’t (I’m looking at you, Chrome, and now, Firefox)], which of the 2.6.xx kernels would/should have been the first 3.0? Any suggestions? Personally, with the rapid release schedules much software has today, I think version numbers are meaningless, but if you had to choose, which would you choose, and why?
For me, being most interested in the graphics stack, I’d say the wide-spread inclusion of KMS, as far as my recent memory goes, but I know I’m forgetting other milestones. What are other milestones that might have made a good 3.0 release?
The way things work you should compare the Linux version numbering to that of webkit, which is at version 534.20. webkit is then tested and tuned for a specific browser release, which gets a more “real” version number in the classical sense, with a more normal release procedure. The Linux kernel in itself is not a product and doesn’t have very thorough release management, so it doesn’t get normal version numbers. The distributions that commit to a certain version of the kernel and patch/test that one do however have normal release procedures.
Removal of the BKL (Big Kernel Lock)?
I certainly don’t mean to denigrate Linux, but I think in an ideal world you wouldn’t know what version your kernel was at, anymore than people know without looking what version of WebKit or Gecko their browser is on.
Now, thats a great point!
Software that would “just work” really would be nice. I think thats sort of what Java meant to do when they started out, but look how far that got. Seems like .net does an pretty good job of handling different versions of stuff, but no offense to the Mono folks, I don’t think .net will get any traction outside of MS land.
And to a large degree, this is that ideal world. I have both Fedora and Ubuntu machines at home, and I have no real idea what kernel version either of them is running. Obviously it’s going to be a recent version of 2.6.x, but beyond that I haven’t a clue…
I like the FreeBSD numbering system, Release Version.Revision. Ideally the version number should indicate the baseline functions, and the revision indicates how those functions have changed in regards to bug fixes.
Decade.release.build makes more sense for the Linux kernel, and I think that’s would be a reasonable system.
I’m embracing the apostasy. They didn’t just change the version meaning willy nilly. They moved to shorter release cycles so each version has less changes, making it theoretically easier to upgrade with less worry.
If anything these version number radicals are forcing us to actually investigate releases rather than assuming the company/org has done it for us by blessing it with a version number to indicate their confidence in it.
I have not so fond memories of being screwed over by point releases because I assumed it was a minor change. And it was, a minor change that went undetected until it was an obscenely obvious problem with devastating consequences. Its not a shame on them for choosing a version number, its a shame on me for not testing it sufficiently.
The H coverage, the best around, as usual
http://www.h-online.com/open/features/What-s-new-in-Linux-3-0-12795…
Read the in detail too.
Anyway, I think I will skip. BTRFS had a lot of “huh, umm, do we need this code? Let’s drop it and see” kind of changes. I use it since I gave up on ZFS, so this make me a little worried. Distributions start to ship BTRFS by default, I hope it will help.
Edited 2011-07-22 12:33 UTC
The big question for me is: Does Btrfs finally have a fsck that can fix errors? If there isn’t such a tool/feature I don’t use the filesystem.
It doesn’t. Its still understood to not be ready for prime time. The online fsck is probably big enough news that you’ll hear about it. No idea when it will happen.There was an effort that looked promising at the beginning of the year, but they scrapped it for whatever design/performance reason. I’m guessing its not trivial.
But, I completely agree. I’m not playing with it until it has online fsck error fixing.
The last news I’ve seen yesterday* is that they have one, but they are doing extensive testing.
But btrfs will probably have 3 things:
– the one that was added in the Linux 3.0: scrubbing: it will check the checksums and data on the disk and try to fix any problems it can find. This is something which can run will the filesystem is in use I guess.
– seems someone posted a patch on the btrfs mailinglist recently to do online filesystem checks in the kernel. To check if the meta-data is still consistent.
– and the offline fsck already mentioned above.
* I can’t seem to find it anymore though
BTRFS is a COW File System. If it get in invalid state, it should be able to reverse itself and by itself to any previous valid state for the affected files.
If it work, I can’t tell, it worked fine until now (I run integrity check against saved md5 checksum and file counter for unused directories). In theory, this kind of file system is much more resistant than classic file system like Ext or Apple HFS+.
Offline check can be nice, but they should be useless no mather how they work. It’s the job for the online integrity checked.
It doesn’t work.
My entire btrfs filesystem got corrupted the other day because of battery running out. Without fsck, there was no way to make it work again.
Oooo. Ouch!
Not quite. It supports COW snapshots, so you should (in theory) be able to fall back to a previous snapshot. But that doesn’t help much if you’re not using snapshots, and don’t have anything to fall back to.
ZFS does that, but ZFS does not rollback to the latest snapshot. Instead, ZFS rolls back to the latest change.
And BTRFS needing fsck? Why is that? As I have understood it, fsck on ordinary Linux filesystems need the filesystem to be offline while doing fsck. I heard about one big raid which took one week to fsck, when it was a data problem. It would be faster to recreate the raid from backup than waiting one week.
scrub in ZFS allows you to use the raid while controlling the data. There is no fsck in ZFS, only scrub. Why having scrub and fsck in BTRFS? Beats me.
There were not a lot of changes in the 2.6.. -> 3.0 jump, IRC, I think there were some big changes in the 2.0 -> 2.2 version,
So, historically, what do you suppose were the most significant changes / version jumps in Linux?
2.4 -> 2.6
Linus’s baby is a cyborg?
Technically he’s an android.
p.s., Mac goes up to 11.0!!
Don’t hold up traffic, nothing to see here. Same old same old.
I downloaded and compiled 3.0 and found it to be noticeably slower than the 2.6.32 kernel I normally run. And some things (graphics and boot time) are noticeably slower in 3.0 than 2.6.38.
Personally I’m not liking this trend, about every six months I grab a new kernel release and it seems each one has slightly worse performance than the one before it.
Anyone else noticed a performance change (increase or decrease)?
Can’t say I have, that said I haven’t done any proper benchmarks. Boot time for me ‘seems’ to be faster these days but again I have no hard data to back that up with.
I did notice that with the upgrade from .37->.39 I did, my system launched using 13 megabytes less of ram than it did prior to the upgrade. Not much to shout about these days I know, but I was just shocked at the notion of a newer version of something using less memory, particularly given that my system is quite lean as it is.
jessesmith,
“Personally I’m not liking this trend, about every six months I grab a new kernel release and it seems each one has slightly worse performance than the one before it.
Anyone else noticed a performance change (increase or decrease)?”
I’m curious myself, can anyone else who’s planning to install the new kernel do some before/after benchmarks?
Network/disk/fs/ram.
I would do it myself, but I was not planning on installing the new kernel until it’s included in the distros.
I have noticed that my laptop in particular feels slower after every new ubuntu install, but I use a desktop pc for all the heavy work so I haven’t had much reason to investigate.
It’s probably at least also (if not largely) because you expect and actively search for it; a placebo, basically (not anything “bad” …on the contrary, a very normal, very human thing; our neural network, over the half a billion years of its evolution, didn’t really encounter any selection pressures depending on accurate representation, over disjointed moments, of such minute subtleties of perception – but OTOH being focused on the stimuli of the moment to the point of over-sensitiveness, was certainly adaptive)
Hm, I guess setting up an ABX test (say, some script essentially randomizing kernel version used) should be relatively trivial.
Edited 2011-07-29 04:58 UTC
So, the real major feature of 3.0 is in fact a build system that can cope with this new numbering scheme.
Is there really anything that can be done in kernel worth a major 3.0.0 version number as per old numbering rules? Looks exactly like Solaris dropping 2.7 for 7.
Still I kinda hoped Linus would use that opportunity to introduce something revolutionary he might have been picking at in his basement, some kind of a skunk works project.
Now Linux 3 is going to make every headline but it is sad there is nothing to showcase. Another missed opportunity for Linux on the desktop publicity wise?
Edited 2011-07-23 09:56 UTC
If you go by the old rules, nothing short of a complete rewrite would be worthy of moving to 3.0. Maybe changing to a microkernel design, or something like that.
Anything short of that would just be grounds for a 2.8.x series.
That’s exactly what it is. Technically, it’s just another release in the 2.6 series, but Linux and co figured that with the patch version getting to 40, it was time to do something different.