KERNEL NEWS ZACK’S KERNEL NEWS

Cleaning House Chehab put it, folks have said at various the project active would be a better main- Andrew Morton has posted a patch to times that they would fix the driver, but tainer. Josh’s initial release had about 20 schedule FUTEX_FD for removal by June then nothing happens. contributions from Alexander Viro, Adam 2007, saying that it suffers from “unfix- Adrian has also posted patches to re- DiCarlo, and Pavel Roskin, as well as able races”. This condemnation was also move the MAC89x0, ATARI_BIONET, Josh. A later 0.2 release had about the the standard used against DevFS in the ATARI_PAMSNET, and SKMC drivers. same number of patches (this time, old days, though a clear explanation of These have also been marked as broken mainly bugfixes), from Josh and Pavel, as why the races were unfixable was hard for three years (since 2.6.0), and have well as Christopher Li, Damien Lespiau, to come by. It is doubtful FUTEX_FD will not been fixed in that time. In response Nicolas Kaiser, and Oleg Nesterov. be revived. It has very few (or no) users, to these patches, Geert Uytterhoeven and according to its author, Rusty Rus- said that he had a patch lying around by Status of CREDITS sell, it was always something of a hack Matthias Urlichs, which might fix the Jarek Poplawski recently pointed out and a special case. Once it’s removed, MAC89x0 driver. However, as Matthias that the CREDITS file was very much out Rusty says, it should be possible to yet said, this would need to be checked by of date, missing a lot of people who had further simplify the futex code. In that someone before it should be included. made massive contributions over the case, it’s possible that Andrew will want Adrian has also posted patches to re- years. explained his cur- to accelerate the removal schedule. move the FB_CYBER FB_VIRGE, FB_ rent philosophy of the CREDITS file – it Meanwhile, Adrian Bunk continues RETINAZ3, FB_ATARI, FB_SUN3, and was useful five years ago, when there the largely thankless task of removing FB_PM3 drivers, also marked as broken was no version control in use for kernel old, unmaintained, and broken code for three years. This time, James Sim- development. At that time, the only way from the kernel. He posted a patch re- mons stepped up to try to maintain these to keep track of who had made contribu- cently to remove the VIDEO_ZR36120 drivers if he could get hold of the rele- tions was to list them all in a file. With driver, which has been marked as bro- vant hardware. Michael Schmitz pointed revision control, contributors are tracked ken for three years. As Mauro Carvalho out that Geert Uytterhoeven had revived in the revision system itself, in much the FB_ATARI driver with a preliminary finer detail than the CREDITS file ever The kernel patch, which Michael then updated and could. Linus regards the CREDITS file as mailing list com- submitted. a legacy – still useful for identifying prises the core of Adrian has also posted a patch to -re early contributors, but not by any means Linux development move the SKMC driver, also broken for a canonical source of information, espe- three years. No one stepped up to claim it. cially for recent work. activities. Traffic vol- umes are immense, Sparse Maintainership Status of sysctl often reaching ten Josh Triplett has taken over maintainer- I recently said in this column that sysctl thousand messages ship of Sparse from Linus Torvalds and had been deprecated and would be re- in a given week, and put out the first official release, version moved. At the time, I said that unless keeping up to date 0.1. Linus never actually put out any some big user-space application was with the entire scope of development numbered releases himself. The Sparse found to depend on sysctl, it would is a virtually impossible task for one tool parses code (ANSI C and a num- probably have a short life. person. One of the few brave souls to ber of GCC extensions), analyzing things Well, Eric W. Biederman, who origi- take on this task is Zack Brown. like which locks a bit of code acquires or nally scheduled sysctl for removal, has Our regular monthly column keeps releases, or what address space a pointer learned of some legitimate sysctl users, you abreast of the latest discussions points to. These bits of information with some interesting results. First of all, and decisions, selected and summa- don’t necessarily tell whether code is sysctl will no longer be marked as depre- rized by Zack. Zack has been publish- broken or not, but they help a human cated, and will be compiled into all ker- ing a weekly online digest, the Kernel reader analyze the code by eye. nels by default. Thus, as Eric says, “we Traffic newsletter for over five years Linus started working on Sparse in get all of the advantages of a fast inter- now. Even reading Kernel Traffic alone 2003, providing a set of features sufficient face for anything that needs it. Currently can be a time consuming task. to his needs, after which his maintainer- sys_sysctl is about 5x faster than /proc Linux Magazine now provides you ship became much less active. When it / sys, for the same string data.” with the quintessence of Linux Kernel became clear that other users wanted to However, sysctl is still on the chopping activities, straight from the horse’s add more features, he suggested that block. With all these existing users, it mouth. someone with more interest in keeping can’t just be deprecated and removed,

14 ISSUE 76 MARCH 2007 KERNEL NEWS

but the original reasons for removing it Relocatable bzImage the relocation features themselves, valu- (Andrew Morton has called the sysctl Having to load the kernel into a one- able cleanup is going into these typically code “goop”) are all still valid. So, Eric megabyte location has placed a size bur- dense and messy portions of the kernel. plans to work with distributions to iden- den on Linux development over the tify existing users and help them migrate years. Compression has helped to miti- Support for the Siemens away from sysctl, until it once again gate that burden, but always there has SX1 Phone seems feasible to deprecate and remove been the desire to break through that Vladimir Ananiev submitted a patch to the code. His current time estimate is a barrier, to have larger kernels that sup- add support for the Siemens SX1 phone. year or two, until it would make sense to port more features. However, the prob- Tony Lindgren and Pavel Machek were consider removing the code again. lem is fairly thorny. In particular, code to enthusiastic about the patch, and wanted suspend and resume a running system it to be included in the main kernel right Support for Apple Motion tends to be very demanding of any ker- away. A number of framebuffer folks also Sensor nel relocation code. expressed interest in reviewing the frame- Stelian Pop and Michael Hanselmann Eric Biederman had implemented buffer integration of this driver as well. have implemented a driver for the Apple some patches to support a relocatable Motion Sensor (AMS), found on the kernel a while ago, and Vivek Goyal at Support for the Atmel 2005 revision of the Apple PowerBooks IBM has been testing them for some MACB On-Chip Ethernet and iBooks. Various other folks also con- time. Vivek now feels the time is right to Module tributed numerous fixes. The patches get Eric’s patches into Andrew Morton’s Haavard Skinnemoen from Atmel posted had once upon a time lived in Andrew -mm tree for broader testing. Because a patch to support the Atmel MACB on- Morton’s -mm tree, but he’d removed Eric is now involved in other projects, chip ethernet module, and listed himself them after receiving conflicting patches Vivek has forward-ported the patches to as the official maintainer in the MAIN- from the developers. As Stelian put it, the latest kernel, and added his own TAINERS file. Jeff Garzik praised the things have settled down since then, and fixes and improvements. quality of the code, saying it was a “nice, is ready to return to the -mm tree for A lot of kernel hackers have jumped clean driver that was so painless to broad testing. onto this new effort, and in addition to apply to the latest kernel”.

adver tisement