Linux 3.0

An anonymous reader writes "In a post to the kernel mailing list, Rob Landley, sitting in for the floating Linus, cracks the whip over what will be in Linux 3.0. His orders are on Linux and main."

Comments

The floating Linus?

[mccalli]

...sitting in for the floating Linus...

He's achieved a transcendental state now? What are the kernel people going to do when he finally ascends to Nirvana?

Cheers,
Ian

Re:The floating Linus?

[SkulkCU]

For those wondering, he's on a boat.

Either way, he doesn't exactly have his feet planted firmly on the ground...

Damn.

[packeteer]

And 2.4.19 is STILL compiling on my 50 mhz box...

My most anticipated feature

[Sheetrock]

LKCD: Linux Kernel Crash Dumps. Really, I wish this had been there for the first half of 2.4 (testing-pre?). Supposedly it'll be able to save an image of kernel memory when the kernel panics to a special partition so that it can be recovered after reboot allowing easy analysis of the image. This alone should cut down greatly on the amount of work required to submit bug reports.

Re:My most anticipated feature

[NecrosisLabs]

What would be great would be if it automatically piped it to the screen, with some form of high contrast text, like white text against a blue background...

Re:My most anticipated feature

[pcidevel]

I develop drivers for Solaris, HP-UX, and Linux (for a PCI card that my company makes). I would love to see a Kernel dump for a particular problem I am trying to debug at the moment! :)

Of course, stupid mistakes in Solaris or HP-UX kills their kernel and results in waiting for the machine to reboot. Stupid mistakes in Linux results in a kernel panic with the output sent to the syslog (9 times out of 10 bad code doesn't kill the entire Kernel, so no waiting on the machine to restart), so I definately think that Linux has the upper hand as far as handling poor kernel space code.

Re:My most anticipated feature

[dracken]

Bah! all you geeks are struggling to make the first gen BSOD. Look at Cowboyneal's article last friday

Posted by Cowboyneal on Sunday October 20, @11:16AM
from the yippee-new-windows-features! dept.

In a surprise announcement today, Microsoft President Steve Ballmer revealed that the Redmond based company will allow computer resellers and end-users to customize the appearance of the Blue Screen of Death (abbreviated BSOD), the screen that displays when the Windows operating system crashes.

The move comes as the result of numerous focus groups and customer surveys done by Microsoft. Thousands of Microsoft customers were asked, "What do you spend the most time doing on your computer?" A surprising number of respondents said, "Staring at a Blue Screen of Death". At 54 percent, it was the top answer, beating the second place answer "Downloading Pornography" by an easy 12 points.

"We immediately recognized this as a great opportunity for ourselves, our channel partners, and especially our customers." explained the excited Ballmer to a room full of reporters.

Immense video displays were used to show images of the new customizable BSOD screen side-by-side with the older static version. Users can select from a collection of "BSOD Themes", allowing them to instead have a Mauve Screen of Death or even a Paisley Screen of Death. Graphics and multimedia content can now be incorporated into the screen, making the BSOD the perfect conduit for delivering product information and entertainment to Windows users.

The Blue Screen of Death is by far the most recognized feature of the Windows (tm) operating system, and as a result, Microsoft has historically insisted on total control over its look-and-feel. This recent departure from that policy reflects Microsoft's recognition of the Windows desktop itself as the "ultimate information portal." By default, the new BSOD will be configured to show a random selection of Microsoft product information whenever the system crashes. Microsoft channel partners can negotiate with Microsoft for the right to customize the BSOD on systems they ship.

Major computer resellers such as Compaq, Gateway, and Dell are already lining up for premier placement on the new and improved BSOD.

Ballmer concluded by getting a dig in against the Open Source community. "This just goes to show that Microsoft continues to innovate at a much faster pace than open source. I have yet to see any evidence that Linux even has a BSOD, let alone a customizable one."

Why do i care?

[nurb432]

Not to sound like a troll, or flaming developers, but seriously, from a users standpoint, why do i care?

What i have now works great, give me concrete reasons i should worry about a new release.

Now as a developer i DO care.. I'm just looking at this from the stand point of a normal user ( my customers ) who hear the same stuff from M$ or apple.. 'new and improved, you must upgrade now'... And we used that as a selling point for Linux..

Re:Why do i care?

[dattaway]

why do i care?

The kernel is the framework that connects to the hardware. Its like the chassis on your car, designed to give all components a secure mounting point to connect to each other. Without the kernel, all your component software would still be functional, but not to useful as they would be laying on the ground in a proof of concept state. The kernel supplies all the hardware to hook things up and make them into a fully functional machine.

The init process and scripts, libraries, and applications are the engine, powertrain components, interior, and all the other details to make a complete operating system. The kernel is simply the framework and body to make it all possible. Compile options allow you to have lightweight race car or a dumptruck.

Re:Already got a beta version....

[cnkeller]

Here is a good place to find out about the state of various features...

Let me be the first to come out with the bad joke

[GreyWolf3000]

Again, some of the things on this list won't make it into 3.0. It's just candidates. But everything that is NOT on this list in about 7 days is probably going to become 3.1 material by default.

Speak now, or till 3.1 hold your peace...

Ok, lets all acknowledge the obvious cracks at 3.11 (like what happened with Windows). Let's sort of communally agree that we're not going to find 'em funny, before a really dumb thread enters the picture, okay?

Not Version Bloat.

[muixA]

Before anyone gets going on it.
There have been MAJOR features added to this Kernel.
Including
- UML
- New VM
- New Scheduler
- Finer SMP Locking
- At least 2 new Journaled FS (Reiserfs4 and XFS?)
- A new POSIX thread library/API.

Does anyone know if ALSA will be included?

We will finally be able to forget about the 1980's style /dev/dsp :)

--
Matt

Re:Not Version Bloat.

[rweir]

ALSA is defintely in, as is things like USB2, Access Control Lists, new NTFS support (it doesn't completely trash partitions now!), hot pluggable CPUs, software suspend (hit a key combo and save the whole system state to the harddrive), support for drives >2TB, and a whole lot more.

BTW, I have to love a community where this sort of thing is discussed on a site called KernelNewbies:)

Reiser4

[KagatoLNX]

I *sooooo* hope the Hans gets off his butt and gets ReiserFS 4 in this one. If you follow the LKML closely (or just read the Kernel Traffic, http://kt.zork.net/kernel-traffic/latest.html) then you may have heard he's sweating a bit on getting it in.

Reiser4 may just revolutionize the way the some people do stuff. I mean, next system I want to be able to do:

# cat /etc/passwd
root:x:0:0:root:/root:/bin/bash
jim: x:100:100:jim:/home/jim:/bin/bash
# cat /etc/passwd/jim/uid
100
# echo /bin/ksh > /etc/passwd/jim/shell
# cat /etc/passwd
root:x:0:0:root:/root:/bin/bash
jim: x:100:100:jim:/home/jim:/bin/ksh

All that *and* have transactional data commits with a very small performance hit!

(ReiserFS Trolls: Go ahead, bring it on!)

Re:Reiser4

[IamTheRealMike]

However, I'm not clear what feature of Reiser4 it is that you're demonstrating. Could you provide an explanation to accompany your example?

The feature he's referring to is the ability to treat files also as directories, and it promises to be a majorly cool enhancement to Linux that rewrites one of the most basic assumptions of OS design. The idea is that by improving the power of the filing system layer, and boosting the performance of the FS for small files, the need to have databases layered on top (and even /etc/passwd is a database) is eliminated. This in turn leads to a more powerful OS as that power is made generic, so being available to everything. It's better explained by Reiser himself.

One problem - 2.6/3.0 won't have that ability. What it will have is a special system call reiser4() that Hans can play with. You won't be able to "cat /etc/passwd/mike/group" anytime soon, unfortunately, this kind of major change takes a long time to work its way though the system. The reiser4 call will allow Hans to experiment with the new semantics before we even start to think about merging with the actual kernel.

Why is small file performance so important (this is the area where RFS kicks the ass of, well, pretty much everything else)? Because there are quite a lot of files out there which would actually be better stored as lots of small files. /etc/passwd is one good example, there are others. The reason they aren't currently stored as files is because traditionally filing systems have sucked when you have lots and lots of very small files, and we're talking like perhaps 5 byte files here. Reiser4 has some extremely clever algorithms in it, which mean it's good at small files but also large files too.

Of course, this is just the start of a much bigger picture, that'll see the filing system become something akin to a searchable knowledge store. Unfortunately, it's not going to happen quickly. For starters, if you were to suggest to the maintainer of app foo that they should store their data as lots of small files, they's say "no way, some of my users are on ext3, or xfs, or jfs" etc. Reiser has great vision, but he's not the only player in this field, and I have a nasty suspicion that the goal of exploding out large files into filing system structures could prove to be difficult while other filing systems are prevalent. Let's hope not, eh?

New console layer

[jaymzter]

I for one am totally psyched about re-writing the console sublayer. It's so aesthetically annoying to be running a multi-headed system, yet be reserved to only one head when on a tty. I think this has a high geek factor

Re:Linux 3.0

[cperciva]

Linux ** (just take care of all the letter names at once)

If you only want to take care of two letter tags, shouldn't that be:

Linux ??

Re:3.0?

[4of12]

Of course there's long been a convention in Linux land that less stable development kernels have odd numbers like 2.1, 2.3 and 2.5, while even numbers denote the stable series meant for pedestrian users. [Although many could argue that the VM switch during 2.4 did not exactly belong to a stable series.]

Anyway, if we're going to have an odd number major version, then all I can say is

"Get ready for a ride"...

Re:Is media automount in the kernel yet?

[Ektanoor]

Are you from Mars? There are two versions of automount in Linux and there is something called "supermount". But if you use Linux for servers, then you are deeply wrong if you put these things into action. The automount feature is not embedded to Linux just for that reason. A good secured server should in no way give a chance to automount third party media. Only the admin should do it and he shall have a chance to do it flexibly and correctly. Believe me, that this is the true way of administration. Maybe where you work people may think it is too bad that Linux doesn't automount every piece of crap that may either trash the system or give a chance for information leaks. But, on my years of sysadmin I consider that this is one of the best features not only of Linux but of the whole *NIX family.

DOS didn't have automount.

[Inoshiro]

DOS had stateless device access. Until you tried to look at a device, DOS would not touch the device (floppy drive, hd, or CDROM drive). But when you did change to the device, it would try and read in its base directory and bootsector.

Windows emulates its behaviour towards floppy disk drives, as you will find out very painfully if you click on the A: on a computer without a floppy drive (which, for me, is all of them), or without a disk in the drive.

Automount only works on hardware that gives feedback on when media is inserted (such as a CDROM drive). To prevent Badness (TM) in the blocklayer, the automount has traditionally been eschewed in favour of explicit mount. Why? Try removing a CD that's being read from in Win9x, and watch the blue-screen "Please insert CD labelled ..." as the kernel catches a block layer exception. This can be worked around by adding drive locks every time the drive is accessed, but it's generally considered to be a hairy problem best solved by having a smarter user.

Of course, many distributions include the (separate) automount patch anyways, and people who want this behaviour won't be rolling their own kernels any time soon.

What should be embedded into Linux kernel 3.0

[Ektanoor]

Well, this is what some people really wish:

Internet Explorer.
GUI.
The Eternal Flat Desktop for dummies.
Word, Excel, PowerPoint, Solitaire.
Palladin .NET
WYSINAWYG
WYGINAWYW
Easter Eggs
Make desktop "user ready". Forget the flame.
Forget the bugs, claim the features.
Add 100Kb EULA into the kernel itself.
Sell it and yourself to Bill Gates.
Rename it to Windows.
Sell it for $400 and threaten everyone who will not follow you.
Write a small text, anonymously authored - "Why I switched from Linux to Windows" and claim how your customers are deeply satisfied.

More complete list

[awptic]

This page contains a complete list of every new feature that has gone into 2.5, and other features waiting to be integrated and their status:

http://kernelnewbies.org/status/latest.html

Re:Linux 3.0

[den_erpel]

No, this time Linux is correct since the topic is about the kernel.

If you are talking about the Operating System, you should address it as GNU/Linux (same as you have GNU/Mach).

e.g. Debian, SuSE, Redhat, ... distribute GNU/Linux (you would not do much with a Linux distribution).

You must have amnesia RMS, since you learnt us to cite:
GNU is the operating system and Linux is one of its kernels

Prepare to reboot your box next year...

[DocSnyder]

How could my server ever reach 1000 days of uptime with Linus throwing out new major kernel releases every two years? ;-)