See also: last week's Kernel page.

Kernel development

For those of you following Alan's patches to the prepatch system, his latest patch is 2.4.0test13pre4ac2. He's currently working on merging the 2.2.18 fixes into the 2.4.0 tree.

The current stable kernel release is still 2.2.18. One new prepatch for 2.2.19 has been released this week, 2.2.19pre3. A run_task_queue fix, resolving a lockup problem some people have been seeing, should be welcome.

Other, minor discussions. This was a relatively light week for the linux-kernel mailing list, with about half of the normal volume of posts. Here are a couple of discussion items that came up this week:

• In spite of Linus' comment, "if you think you can skip christmas with a 4-year-old in the house, think AGAIN", a bug report from Marco d'Itri regarding an innd mmap bug, plus work by Marco, Linus and others on Christmas Eve and Christmas day resulting in the finding of a slight faux-pas. It seems the only routines in the current 2.4 tree that actually write out dirty buffers are ones called under heavy memory pressure. Under certain conditions, a machine under no such pressure could be cleanly shut down, yet fail to update modified files. Expect to see a fix for this coming out soon, now that Christmas is officially over.

• Linus encourages the use of gcc 2.96 and gcc 2.97 for testing purposes. It seems that Linus' complaints about the inclusion of a relatively untested version of gcc with Red Hat 7 have caused many people to assume that the use of such compilers was globally "verboten". Linus pointed out that problems with the compiler, or with the kernel as demonstrated by the new compiler, can't be found if no one uses it. "Now, now, I'd love to see reports of expecially the new updated compiler. ... I just want people to mention the fact, so that I can correlate any bug-reports with a compiler version. Just in case. It can be important (and not just because of compiler bugs, but due to real kernel bugs that just were hidden by pure luck with other compilers)".

  So go dig up the new compilers, folks. If they can help turn up kernel bugs, then best use them before the first stable version of 2.4 is announced.

• Wait queues and wake-ups in 2.2.19pre2 were the topic of another discussion thread. Hinging upon the discussion was whether or not the 2.4 wait queue API would be back-ported, or whether the 2.2 series would continue to be unique. Underlying this was a more primary issue, that the code introducing the 2.4 behavior into the 2.2 tree was "subtly broken". Further investigation turned up some race conditions that affected both the 2.2 and 2.4 wait queue code. Fixes for those are in the works; here is Andrew Morton's draft patch for the 2.4 series.

• Daniel Phillips posted a Request for Comment on a patch that uses semaphores for daemon wakeup instead of relying on the wait queues. The primary result is to produce much cleaner code; neither performance improvements nor penalties were expected.

• Other patches and updates released this week include:

  • Jan Kara has ported his quota patch to -test12 and provided some notes on the patch, which includes both fixes and some additional functionality.

  • Rasmus Anderson posted an updated version of his rcpci45.c patch, a general cleanup that includes an update to the new PCI API.

  • Ingo Oeser has released an updated version of his OOM-Killer-API patch. This patch allows people to add the OOM handler of their choice to the 2.4 tree (The choice of an OOM handler for the 2.4 tree has not yet been determined, just discussed).

  Section Editor: Jonathan Corbet

For other kernel news, see: Kernel traffic Kernel Newsflash Kernel Trap 2.5 Status

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

Next: Distributions