Linux 2.6.36 Released

diegocg writes "Version 2.6.36 of the Linux kernel has been released. This version includes support for the Tilera architecture, a new filesystem notification interface called fanotify, CIFS local caching, support for Intel Intelligent Power Sharing in i3/5 systems, integration of the kernel debugger and KMS, inclusion of the AppArmor security system, a redesign of workqueues optimized for concurrency, and several new drivers and small improvements. See the full changelog here for more details."

Comments

TFA

[pinkushun]

The one post where 90% of /. users will actually read TFA

what i like about kernel releases...

[underqualified]

linus trolling on everyone that disagrees with him...

Fanotify disabled in this version

[Anonymous Coward]

Because of desagreement in the ABI the fanotify is disabled in this kernel.

news for gnurds?

[digitalsushi]

This is why I come here.

Actually, I'll come back in 4 hours and read the top comments not modded funny. That's why I come here.

They should make a slashdot that's just about linux projects, nasa/physics stuff, and DIY routers. Like slashdot vintage. It'd be classy. Elastic band jeans and plaid tie dress code. God I miss the good old days. *pours mad dog 20/20 on anti-static carpeting*

Did you even read Linus' release notice?

[Anonymous Coward]

fanotify syscalls are disabled because people still can't agree on the API.

Whether a file has changed = complex?

[CarpetShark]

What the hell is it with file notification? It never seems to be reliable or stable. There was inotify, dnotify, fsnotify, fam, gamin, incrond... and since fam/gamin always ended up using 100% CPU or causing other problems, I've just avoided the whole idea, even though I regularly think of situations that I could use incrond in.

I would have thought that setting a flag/triggering an event when a file was altered would be a matter of adding a small queue/bit system for events and about one line of code to vfs functions that modify files, but obviously not.

So... does anyone use incrond and get good, reliable results? Will fanotify help at all?

Re:Whether a file has changed = complex?

[abigor]

fanotify seems interesting because it allows you to watch for arbitrary events on a global basis without specifying which file descriptors you want to watch. It seems to pass the actual file descriptor back to the original object. You read events via getsockopt().

So, for example, it's very easy to say "notify me of all file open events", for example.

You can also do the opposite of a global watch, of course, and have it watch for specific file events, much like inotify.

As I understand it, fanotify came about because of vendor demand for an efficient, non-hackish way to watch for arbitrary filesystem events without ever specifying precise files, paths, etc. An example would be malware vendors.

Re:Whether a file has changed = complex?

[Late Adopter]

There was inotify, dnotify, fsnotify, fam, gamin, incrond... and since fam/gamin always ended up using 100% CPU or causing other problems

Of those, only inotify and dnotify were userspace-facing solutions in the stock kernel. Fsnotify was a backend, intentionally preparing the way for fanotify, and it was never intended to be used directly. Fam and gamin are third-party, and unless you know you specifically need them, you should avoid. Incrond is a great userspace program to use inotify... but not an alternative to anything in that list.

Dnotify was something of an embarrassment, but inotify's been with us a while and it's worked well. Fanotify is an evolution of that, to fix architectural problems that have led to race conditions and scalability concerns. Inotify (and dnotify) is being reimplemented on top of it, so if the inotify interface doesn't cause any problems for you, plan to continue using it (and incrond if you like)!

Re:Whether a file has changed = complex?

[GooberToo]

Is this a case of linux kernel developers copying features that are in the Mac OS X darwin kernel?

Doubtful. This is really an area of natural evolution. Meaning, first to market, if that is in fact really the case, hardly means everyone else is copying what is really an obvious and extremely simple idea.

For example, planes needed to go faster. Solution, make them more aerodynamic. When everyone started making planes more aerodynamic, does that really mean everyone copied the first to do so? Hardly. It means, they all understood the problem and someone was simply first to market.

Now if you have something which indicates the APIs on OSX are particularly clever in exposing this feature and that everyone is copying those APIs, you might have a point, but I don't see you arguing that position.

Believe it or not, humanity frequently, independently, suffers from a natural progression of ideas. The fact that this occurs more or less validates no one is copying.

Case in point: I want to know when a file changes. In what ways can a file change? Those are natural progressions in seeking a solution to an extremely common problem. If you answer my question, does that you too are copying?

Re:And yet?

[CarpetShark]

But can it play fullscreen flash video smoothly yet?

The problem that prevents flash from playing fullscreen is that it's closed source crap, not that Linux is in any way incomplete.

But ...

[Osgeld]

I thought linux was up to version 10.10? (Maverick something)

Re:But ...

[Randwulf]

Ubuntu, a distribution that uses (some version) of the Linux kernel, is up to 10.10 (Maverick Meerkat). The latest version of the Linux kernel to be released is 2.6.36.

Re:But ...

[CannonballHead]

most likely: woosh :)

Re:Great...

[Anonymous Coward]

Hey. It's still me, Anonymous Coward. I take back everything I just said. Linux is awesome. And my penis is small.

Re:And yet?

[LingNoi]

http://freshmeat.net/projects/loonix/

Loonix is a custom Linux distribution meant for server applications.

I'm guessing it doesn't play flash since it's a server distribution. Silly question really.

Re:And yet?

[underqualified]

linux doesn't have poor support for flash. flash has poor support for linux.

Re:And yet?

[chrb]

No, the problem isn't incompleteness it's the fact that one has to traverse a jungle of incompatible audio and video APIs to make sure it even works at all across the various distributions.

Then how come mplayer works on every common Linux distribution, and has been able to do smooth fullscreen video for as long as I can remember?

Apple Version

[neoshroom]

How come Slashdot keeps posting stuff about Linux? Where are our Apple-related news? Lion, iLife '11, FaceTime for Mac and new MacBook Air notebooks were announced yesterday! We never speak about Apple it's always Linux, Linux, Linux! *

* for the slow-minded, this is a parody of the "Apple news again? We never get any Linux news!" posts. As long as it's not freakin' Microsoft, I'm fine with it.

Here, I'll make an Apple user feel right at home:

The newest version of Linux, Snow Penguin, has been released and this changes everything! This version includes support for the Tilera architecture, a beautiful new filesystem notification interface called iNotify, Spacewarp local caching, support for Intel Intelligent Power Sharing so your computer will otomaticaly [spelled correctly] turn off unused appliances in your house to save you thousands of dollars in power bills every year, developer improvements and a revolutionary AppArmor security system. It's speedy. It works–better. See the full keynote for more details.