[LWN Logo]
[LWN.net]

Sections:
 Main page
 Security
 Kernel
 Distributions
 Development
 Commerce
 Linux in the news
 Announcements
 Back page
All in one big page

See also: last week's Kernel page.

Kernel development


The current development kernel is 2.3.26. Work is already well advanced on 2.3.27 - five pre-patches have been released so far. 2.3.27 may well be out by the time you read this.

The current stable kernel release remains 2.2.13. The indications are that this kernel is indeed stable, there have been few complaints out there. Nonetheless there are several fixes out there; these are currently available as 2.2.14pre4, which will eventually become the next stable release. Alan Cox has put out a separate set of patches as 2.2.13ac3; this is a more developmental patch which contains some features (i.e. RAID 0.90, Raw I/O) which will not make it into 2.2 soon, if ever.

Kernel crash dump analyzer released. SGI has finally released its Linux kernel crash dump analyzer. It's not a tool that most Linux users should have to use, but those of us working on kernel and driver development may find it invaluable. SGI has done the developer community a service by creating and releasing this code.

ext3 0.0.2c has been released by Stephen Tweedie; details in the announcement. Ext3, of course, provides journaling to the standard ext2 filesystem. This release fixes some difficulties and features improved documentation.

Journaling for ReiserFS has been released, some details can be found in this press release. Linux now has two journaling filesystems that can be downloaded and used now (though perhaps with a bit of caution still), and another one (XFS) coming someday.

But what is this ReiserFS? This filesystem is the result of a persistent effort by Hans Reiser and his company Namesys; more details than many would ever want can be found on the Namesys web page. Mr. Reiser has a short-term goal, being a higher-performance filesystem, and a longer-term one: completely changing the way operating systems, data structures, and name spaces are handled.

The short-term performance goals are mostly being pursued through the use of balanced tree data structures. Most current filesystems maintain directories as linear lists of file entries, perhaps with some hashing to speed lookups. ReiserFS uses a tree structure instead not just for directory information, but for the files themselves as well. In some situations, especially those where thousands of small files exist, the performance improvements can be large.

(There is also a tree implementation for the standard ext2 filesystem being worked on by Ted Ts'o, but no code is currently available).

Hans Reiser's long-term vision is more ambitious. He sees a couple of sources of evil in the way current systems are designed: the compartmentalization of namespaces and the imposition of structure on data. Unix (and Linux) systems use the filesystem as a wide-ranging namespace, but it is far from the only one. Network devices have their own space; each separate relational database or structured file is also its own namespace.

Mr. Reiser's central point is that the utility of a system is not determined by the number of components it has, but by the number of possible interconnections between those components. Distinct namespaces keep components from talking to each other, and thus greatly reduce the capabilities of the system as a whole. If everything could be pulled together into a single namespace (the filesystem), the result should be a vastly more powerful operating system.

Despite the presence of a great many words on the web site, getting a firm grasp on how the ultimate ReiserFS-based system would look is not an easy thing. The system would look something like a single, large, amorphous object database with keyword addressing. Things like relational databases or dbm files would no longer exist; instead each individual item would live directly in the filesystem, which would provide powerful indexing and searching mechanisms so that they could be found.

Thus the emphasis on performance with small files. This vision of the operating system could easily result in millions of tiny entities being poured into the filesystem. It had better be fast. Numerous other features to support this mode of operation are envisioned as well: powerful searching mechanisms, inheritance of attributes (and data) between files, etc.

The result is expected to be a system with a much greater expressive power - if users can understand and learn how to make use of it.

This discussion oversimplifies things to the point that people who actually understand the long term ReiserFS vision are probably pretty upset. Those interested in the full picture should really just set aside a substantial block of time and wander through the Namesys web pages.

Meanwhile, developing even the current ReiserFS is a lot of work. Mr. Reiser intends to fund this work through commercial consulting and support fees. The ReiserFS license has drawn a bit of criticism which appears to be unjustified: all it really says is that (1) ReiserFS is under the GPL and can only be distributed with GPL kernels, and (2) if you don't like that other licensing terms are available for a fee. Currently SuSE appears to have engaged their services, and will be shipping ReiserFS with its 6.3 release.

TiVo's kernel mods available. TiVo, which sells a Linux-based "digital VCR" box, has made its modifications to the Linux PPC kernelavailable on the web.

Other patches and updates released this week include:

Section Editor: Jon Corbet


November 11, 1999

For other kernel news, see:

 

Next: Distributions

 
Eklektix, Inc. Linux powered! Copyright © 1999 Eklektix, Inc., all rights reserved
Linux ® is a registered trademark of Linus Torvalds