[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 release is 2.4.0-test9. Linus released this kernel just prior to taking a trip to Germany, so it may be the last for a little while. This is also the point at which Linus had said that he would no longer accept any patches that were not fixes for "urgent" bugs. The freeze is getting tighter.

The current stable kernel release is 2.2.17. The 2.2.18 prepatch is up to 2.2.18pre15 currently. This patch is in the "bug squash" mode, and has a few small problems - for example, the PPC and Sparc architectures do not build. There's a few other things to be dealt with as well, so the official 2.2.18 release is still somewhat distant.

If you install Red Hat 7, be sure to install the "kgcc" package and use it when building kernels. The gcc package in this distribution is a little too new to be used for this task (see this week's Distributions Page for more).

Fixing the 2GHz limit. It turns out that the Linux kernel has a built in limit that will cause it to break on processors with a clock speed greater than 2GHz. Since processors that run at well over 1GHz are already available, the day when this limit will matter is not that far away.

Fortunately, the problem is easy to fix. It's just a matter of changing the way the udelay() function does its work. The fix has already gone into the 2.2.18pre series, and will likely show up before too long in the 2.4.0-test kernels as well. When the blazingly fast new processors show up, Linux will be ready.

The Kernel Wiki wants your help. Gary Lawrence Murphy is looking to get 10 minutes worth of time from everybody who knows something about the internals of the Linux kernel. His project, known as KernelWiki, is to completely document the internals of the 2.4 kernel in some sort of reasonable time frame. In typical Wiki fashion, the Kernel Wiki allows anybody to add content to the site. With luck, enough knowledgeable people will take up the challenge and something useful will result.

Recent developments with filesystems. A few different filesystem issues have come up over the last week. They include:

  • Soft Updates. Kirk McKusick developed the "soft updates" technique for BSD a few years ago. It is a (relatively) simple set of write ordering guarantees that tries to keep the filesystem as consistent as possible. Experience from users shows that soft updates helps a lot in crash recovery.

    The question that came up is whether there were any plans to port soft updates to Linux. The general consensus is that the real answer is going to be full journaling filesystems and that there is no need, at this point, for partial measures like soft updates. There are people who would like to see soft updates go in, but it's unlikely in a time when large virtual filesystem changes are already planned to support journaling and better memory management.

  • ext3: Stephen Tweedie has released ext3-0.0.4, the latest snapshot of his journaling extension to the ext2 filesystem. This patch includes metadata-only journaling, which helps get the performance closer to that of straight ext2.

    This release is not meant for any but the most hardcore of users, however. That's because a complete implementation of metadata-only journaling requires an incompatible superblock change. Soon a 0.0.5 release will come out with the new superblock; at that point, going back to earlier ext3 implementations will be more difficult (but still possible). Meanwhile, the ext3-0.0.4 release comes with even fewer guarantees than usual.

  • TUX2: The TUX2 phase tree filesystem was covered in the August 31 LWN Kernel Page; it is a different approach to the creation of a crash-proof filesystem. Daniel Phillips, TUX2's creator, has run across a set of patents held by Network Appliance that would seem to cover the TUX2 approach. If Network Appliance were to attempt to enforce these patents - which it is not currently doing - it would obviously create trouble for the inclusion of TUX2 in the kernel.

    See Daniel Phillips' posting for a rather strongly worded discussion of the patents and what he thinks of them. He claims prior art for all of the techniques covered in the patents - several years prior. So if it comes to a fight he should win. But there are only so many of these fights that the free software world can afford to fight. Software patents remain a serious problem.

In the middle of all this, the ReiserFS group has been strangely quiet...

A reminder on ECN. Recent 2.4.0-test kernels support the Explicit Congestion Notification (ECN) extension; the September 14 Kernel Page describes this change somewhat. Unfortunately, some firewalls out there on the net react poorly to systems that try to use ECN, with the result that many systems are simply unreachable to ECN-capable hosts. LinuxToday.com was recently cited as being one of the affected sites.

If you are running a recent 2.4.0-test kernel and are experiencing difficulties in connecting to certain sites, you should try turning off ECN. A simple command like:

echo 0 > /proc/sys/net/ipv4/tcp_ecn
will do the trick.

TUX 1.0 (kernel HTTP server) released. The first stable release of the TUX 1.0 kernel-based web server has been announced. TUX is the server which produced such great SPECWeb numbers last June, and which still holds the record for the fastest performance. For those who would like to learn more, LWN looked at how TUX works in the September 7 kernel page.

Other patches and updates released this week include:

Section Editor: Jonathan Corbet


October 5, 2000

For other kernel news, see:

Other resources:

 

Next: Distributions

 
Eklektix, Inc. Linux powered! Copyright © 2000 Eklektix, Inc., all rights reserved
Linux ® is a registered trademark of Linus Torvalds