|
The Linux kernel mailing list comprises the core of Linux development activities. Traffic volumes are immense, often reaching ten thousand messages in a given week, and keeping up to date with the entire scope of development is a virtually impossible task for one person. One of the few brave souls to take on this task is Zack Brown. Our regular monthly column keeps you abreast of the latest discussions and decisions, selected and summarized by Zack. Zack has been publishing a weekly online digest, the Kernel Traffic newsletter for over five years now. Even reading Kernel Traffic alone can be a time consuming task. Linux Magazine now provides you with the quintessence of Linux Kernel activities, straight from the horse's mouth. |
Sage Weil announced the Ceph distributed network filesystem, licensed under the LGPL. This filesystem automatically balances data between any number of nodes and has other cool features like full POSIX semantics. The filesystem provides a configurable amount of data replication so nodes can go up and down without disrupting the system, as long as there are enough nodes to hold all the data.
Sage originally began the work as the subject of his PhD thesis, so Ceph has been under development for some time; however, the client portion has, until recently, been implemented under FUSE (Filesystem in Userspace ). This has made it easier to develop the application but makes for slower performance and some problems that could only be addressed in-kernel. Sage has chosen to make his initial announcement now because he's finally started the in-kernel replacement for the FUSE-based client. This will improve performance and generally be a more correct approach.
Jacopo Antonello has submitted code to support a backlight dimmer in the style of that on the MacBook, but there was quite a bit of resistance to the patch. Pavel Machek felt a good bit of it could be done in user space, and Arjan van de Ven felt Jacopo's patch introduced an unnecessary timing infrastructure and other bits of infrastructure that would all need justification if they were to make it into the kernel.
Jacopo disagreed with these assessments and the discussion did not have any real conclusion, but it is possible some backlight dimmer functionality will be added to the kernel at some point.
Adrian Bunk has proposed doing away with the entire concept of an "experimental" kernel feature. Adrian's planning to submit a patch to remove that configuration option, along with all dependencies on it by the various drivers; he had several reasons for this. First, enabling the "experimental" option in their default kernels has become standard for distributions, just because so many drivers depend on it; it doesn't give the user any protection that way.
Also, the drivers themselves often continue to have the dependency long after they are usable and no longer truly experimental. These things defeat the whole purpose of having the "experimental" option, which was to protect users from seeing choices for kernel features that were not really ready. Because the option must always be enabled nowadays, that protection no longer exists.
Stefan Richter and Pierre Ossman also agreed with Adrian's plan, and no dissenting voice has been heard so far. Still, this would be a relatively major change to the way kernel drivers are organized and presented; you can expect to see some lively debate before this change makes it into the kernel.
Bryan Wu submitted a patch to update Mike Frysinger's email address in the Blackfin driver entries of the MAINTAINERS file. While he was at it, he cavalierly removed Aubrey Li as maintainer of the Blackfin Serial Driver and listed Sonic Zhang instead. There must have been some offline discussion about that, because there were no objections or comments on linux-kernel when he did this.
Adrian Bunk sent in a patch to remove the entire MTRR entry from the MAINTAINERS file. Richard Gooch had been listed as the maintainer, but Adrian said he hadn't seen Richard doing any work on it for a while. Adrian added that the "X86 ARCHITECTURE" entry covered the same code anyway. No one spoke out against the patch, so the MTRR entry will probably be removed.
Grant Likely posted a patch listing himself as the official maintainer of the SystemACE driver.
Jiri Slaby is making some moves to incorporate his webcam driver into the kernel sources, although at the moment, he only wants them in Andrew Morton's -mm tree while the v4l library gets ready to support what he needs. Jiri's STK11XX driver will support a variety of webcams that come built-in on some laptops, including stk1125, stk1135, and stkdcnew webcams. Andrew responded with some technical objections, and Jiri started addressing them. No significant opposition was made concerning the new driver.
Arjan Opmeer took a shot at producing an Elantech touchpad driver; he reverse-engineered the Windows driver under QEmu, observing its data stream well enough to get something that seemed to work for him. Arjan needed testers to confirm that his guesses and assumptions were actually correct; he posted his patch, not as a submission to the kernel, but just to get a few more eyeballs on the driver behavior.
Adrian Bunk submitted a patch to remove the documentation for the Gracilis PackeTwin device driver. The driver itself has been out of the kernel for years. Alan Cox agreed the thing was dead and should be expunged, and David S. Miller applied the patch in readiness to send up to Linus Torvalds. Adrian also removed some routing architecture he said was so outdated there was no point in keeping it. Alexey Kuznetsov agreed, and David applied this patch as well.
Adrian also found and removed an ancient alert about a bug that once existed in NCSA Telnet, and David applied this. Adrian also removed the docs for the COMX driver, for which the code has been gone for more than three years. Alan acknowledged the removal, and David applied it along with the others.
Adrian also found a doc dating back to 1997 explaining ways to tune the kernel for better performance. The file was way out of date, so Adrian removed it and David applied the patch. Adrian also found and removed a file that gave advice on configuring network drivers. The file was old, and its information was covered better elsewhere, so he yanked it out. This time, Jeff Garzik applied the patch.
Recently, Dave Jones noticed that one of the large architecture code merges had resulted in a lot of file name changes in the source tree and that a lot of these files had actually listed their own file names in their code comments. Rather than try to keep these file names and path names up to date, Dave figured it would be better to remove them altogether, so he submitted a patch to do this for many X86 architecture files.
Tim Bird announced that the Linux-tiny project was back in business and will be maintained by Michael Opdenacker. The project aims to reduce the kernel's RAM and disk footprint so that smaller and smaller systems can run Linux, or at least the same small systems can continue to run the latest versions.
The Linux-tiny project will gather together the patches and feed them upstream to Linus Torvalds. Andi Kleen said it would make more sense to submit patches to the kernel maintainers. Andrew Morton wants to see such patches submitted directly to him (and CCing linux-kernel and celinux-dev). Andrew said creating a new patch collection should be a last resort, saved for patches that will have a difficult time getting into the mainline kernel.
Michael agreed, but Tim pointed out that a lot of patches were gathered together already, and someone needed to get them ready for submission. Tim promised to send the patches to Andrew. The discussion then veered off into considering actual patches that might make the kernel smaller.
| INFO |
|
[1] Kernelnewbies Japan: http://lists.kernelnewbies.org/mailman/listinfo/jp-kernelnewbies
|