See also: last week's Kernel page.

Kernel development

Long-time linux-kernel watchers know, of course, what came next: a flurry of mail regarding things that still need to go in. Items named vary from truly crucial updates to wishlist items; some of the more serious ones include:

• David Miller says he is still "a few weeks away" from getting the Sparc platform working properly.
• A number of filesystems (AFFS, HFS, NTFS, QNX, ...) are broken. Coda is apparently due for a big update that will need serious testing.
• Ramdisks are still broken.
• RAID has problems as well.
• There is a pile of race conditions with module loading and unloading that can lead to user-exploitable holes.
• A lot of device drivers still need updates. One good task for would-be kernel hackers with the right hardware would be to fix the ISA memory-mapped I/O stuff. It changed in 2.3, and most of the drivers still need to be updated.
• There are problems with unix-domain sockets, and IP problems as well.
• Various optimizations for newer processors need to be merged.
• It's still not clear if the softnet code (nice scalability stuff) will go in or not.
• The USB code has been through a lot of big changes which need to be merged.
• The NFSv3 client code is still unincorporated, and would be a real shame to have left out.
• An impassioned plea was posted to incorporate the 32-bit UID patch, in order to help make Linux usable in very large organizations.

This list will only grow, and Linus, at some point, is going to have to start saying "no" to things. Experience says he will take on that role without hesitation, meaning that not everything everybody wants is going in. This, of course, is about the most important role that Linus plays - drawing the line so that a stable, high-quality kernel can come out in the near future.

The current stable kernel release is still 2.2.13. Work continues on the 2.2.14 pre-patch; the latest is 2.2.14pre14. Assuming that more problems do not turn up, this one might just turn into the real 2.2.14. Really, honestly, we mean it this time...

Note that 2.2.14 will likely contain two versions of the Tulip ethernet driver. Evidently there is no one version that works for everybody; people with Tulip-based cards may have to experiment to find which one works for them. This is, of course, an unfortunate situation; hopefully the problems can be ironed out before too long.

An update to the 2.0 kernel is in the works after a denial-of-service problem was found in the networking code (see this week's security page for details). Alan Cox, however, is not up to the task; for some strange reason he thinks he is sufficiently busy keeping 2.2 going and trying to pull together things for 2.4. So it looks like David Weinehall will be stepping in to put together a new 2.0 release. Assuming Linus is pleased with the results, 2.0 will continue to be maintained for a while yet.

The license for BitKeeper has been released. BitKeeper is the source management system being built by Larry McVoy and others, and which will, in theory, be used by the kernel developers to better keep track of changes. The BitKeeper License has seen its share of controversy, since it contains features designed to convince proprietary software vendors to pay for commercial licenses. Details on the disagreements can be found in our BitKeeper feature, written last spring.

Note that Larry is considering a special exemption from the logging requirements of BitKeeper when used for "administrative" purposes. Such a use would be to put all of a system's /etc files into BitKeeper. A number of advantages would come from such use, including the tracking of system configuration changes and the ability to easily clone system setups.

Meanwhile, we have a license, but where is the code? Larry told LWN that a code release should follow the final version of the license by not more than a week or two, so we're getting close...

The official USB point of contact is Randy Dunlap. Linus had asked for somebody to step into that role last week, after realizing that he was getting too many uncoordinated USB (Universal Serial Bus) patches from too many people. Randy's job, as outlined in this announcement, is to take overall responsibility for the USB subsystem, make sure that all the patches work together, and filter things on their way to Linus. He has taken on quite a job - USB development is a very active area, with a lot of code being produced.

For a good overview of the current status of USB support, have a look at The Linux-USB howto v0.9 which was just posted by Brad Hards.

Announcing the Kernel Journal. Zack Brown, the guy behind Kernel Traffic, has now announced the Kernel Journal. This page consists of the latest announcements of kernel releases, patches, and other development projects.

There are currently two projects developing access control lists (ACLs) for Linux. It would appear that the two are now talking to each other, and that they have established a common mailing list. Interested parties should have a look at this note for subscription information, and pointers to both projects as well. (There is also a web archive for the mailing list which does not appear in the announcement).

Other patches and updates released this week include:

• An overview of the GNU/Linux IEEE-1394 ("firewire") subsystem has been posted by Søren Thing Andersen.

• Devfs is up to version 148; there has also been a new devfsd release as well.

• Neil Brown send out this note about the status of his work with the kernel NFS server. He has been fixing a lot of things, as well as incorporating H.J. Lu's work and that of others. According to Neil, as of 2.3.32 the kernel NFS daemon is in good shape. He has now released a patch which backports the changes to 2.2.13.

• Keith Owens released modutils 2.3.8.

Section Editor: Jon Corbet

For other kernel news, see: Kernel traffic Kernel Newsflash Kernel Trap Linux 2.5.x Porting help

Other resources: Kernel Source Reference L-K mailing list FAQ Linux-MM Linux Scalability Effort Kernel Newbies Linux Device Drivers

Next: Distributions