See also: last week's Kernel page.

Kernel development

The current stable version remains at 2.2.10. Alan Cox's patches for this kernel are up to ac9. There were a lot of active reports of oopsen with the 2.2.10 kernel in the last week, but not a lot of clarity as to whether the specific causes were from hardware issues or from the kernel itself. Presumably the fact that 2.2.10 has not yet been released is an indication that no one is confident the file corruption problems have been found and resolved.

Alan enable "slab poisoning" in his patches for the 2.2.10 kernel series with his ac7 patch. To quote Alan, "The slab poisoner puts 5a5a5a5a through memory to try and trip up code that doesn't clear memory properly or uses memory that has been freed." So far, this has caught bugs in several drivers, including the ide-tape, AMD scsi and ALSA drivers. None of them were serious, but well worth squashing, nonetheless.

Discussion of albods raged on unabated this week, but no new milestones seem to have been reached.

A second topic, hotly discussed but without particular issue, was how to improve oops reporting for Linux. We mentioned last week the availability of the kmsgdump patch, to allow kernel messages to be dumped to a floppy disk. Discussion started there and moved to the question of why such dumps could not go directly to a hard disk, or more creatively, why they could not be broadcast via a network card. All of these methods seemed preferable to many to the default method of writing an oops out by hand or hooking up a serial cable. However, the concensus seemed to be that writing to a hard disk was too dangerous given the PC architecture. The ethernet option was intriguing to some people and Henning P. Schmiedeh did mention he might take a crack at the code needed to do this.

Hans Reiser asked for some documentation on the new wait.h that came with 2.3.8. Ingo Molnar responded with a list of four major changes/goals introduced into the waitqueues.

Linux access control has been a hot topic lately. One site with relevant information is http://research-cistw.saic.com/cace/, which contains descriptions of and comparisons of the different means of access control. It also has links to access control projects, such as the Domain and Type Enforcement Project. [Thanks to "Don"]

Improving large memory support was the reason for this DMA patch from H.J. Lu. Douglas Gilbert then pointed out that Jaroslav Kysela has already posted a patch for the same problem, and suggested that the two should talk.

Various patches and updates released this week:

• H.J. Lu's knfsd patch is up to version 1.4.4 and he has also released an updated spec file for it, per request.

• devfs 113 from Richard Gooch is out, along with a backport of the patch to the 2.2.X kernel series, provided by Zoltan Balaton.

• A patch to add hw_addr to /proc/net, was posted by Martin Neumann.

• DVD read-write support for the Panasonic DVD-RAM was enabled by this patch from Nathan Laredo.

• A patch to put old-style F_EXLCK locking back into the 2.2.X kernel series was released by David Parsons, for those still working with a.out programs.

• Matthias Welwarsky announced he had rewritten AX.25 support for the 2.3.9 kernel.

Section Editor: Jon Corbet

For other kernel news, see: Kernelnotes Kernel traffic Kernel Newsflash

Next: Distributions