Kernel Log: The second day of Kernel Summit 2008; criticism of Ubuntu at the opening of the Linux Plumbers Conference
Following the release on Tuesday of details about debates and decisions at this year's Kernel Summit, LWN.net has now reported on the second and final day of the conference. Linus Torvalds, Andrew Morton, and some 80 other important and nearly all male kernel developers discussed their approaches to the further development of Linux and exchanged experiences. Like the [ticker:uk_11545 LWN.net articles covering the first day], Jonathan Corbet's reports on the second day of the kernel Summit are available exclusively to LWN.net subscribers until the 25th of September. As has been the case in previous years, other sources of information on the summit are scarce – it often takes weeks or months for additional information on decisions reached at the conference to trickle out on blogs or posts on the Linux Kernel Mailing List (LKML).
Tracing tools were at the top of the agenda on the second day of the conference. According to a highly detailed report in LWN.net, there was a good deal of discussion about the SystemTap, Dtrace and LTTng analysis tools. Some of the kernel developers tried their hands at SystemTap, but even some of them lacked the endurance to set up and maintain the diagnostic tool. Torvalds considered SystemTap to be much too complicated. Instead, he made an appeal to improve the current, much simpler kernel tools for tracing before adding new ones. However, kernel hackers were unable to agree on a concrete plan on how to proceed. Some of the kernel hackers wanted to continue to discuss the topic at the Linux Plumbers Conference, which started immediately after the Kernel Summit at the same location.
Changes in the handling of the kernel thread were also debated. That gave rise to a discussion about thread interrupts – these are used in the realtime tree (RT) to allow hardware interrupts to be run in their own threads in order to avoid disrupting other work in the system. Until now, a number of developers have viewed this concept, which has a profound influence on how interrupts are handled, with a critical eye, but following a few debates about details there are no longer any significant objections to the idea. This paves the way for the code for threaded interrupts to be incorporated into the main development branch, but first the code has to be adapted so that threaded interrupts are not used by default, but only for hardware devices where it makes sense and is unlikely to cause any problems.
LWN.net reporter Jonathan Corbet led a session on the topic of kernel documentation himself. During that session, kernel hackers decided to remove completely outdated documents from kernel sources, since they can point new developers in the wrong direction. In addition, Torvalds called for better release notes for the new kernel versions, but it is not clear who is going to be responsible for this.
Once again, there were also mini summits on special topics. In one, Matt Mackall presented his Bloatwatch project, which closely analyses the growth of text and file structures in Linux. In another mini summit, Network administrator David S. Miller described in broad strokes the Linux TX Multiqueue Implementation through which the network stack in version 2.6.27 offers basic functions for network hardware with more transmit queues (LWN.net article).
James Bottomley led a session on the topic of "Fixing the Kernel Janitor's Project" before kernel developers had the chance to discuss it after the Kernel Summit. This project is designed to allow programmers to orient themselves in the world of the kernel and learn how to contribute to Linux development. New kernel hackers often work from the documentation or advice of other project members and established kernel hackers do not always appreciate the results. Torvalds thought that newbies ought not to begin by correcting code that triggers compiler warnings, as the project advises, since such corrections often contain new bugs.
According to the description in LWN.net, the discussion proceeded to make the point that testing new kernels, and fixing any bugs found, would be an especially good task for newcomers. But at the end of the day, there was no concrete decision on how to proceed – while a mentoring project that would actively help newbies was suggested, nobody expressed an interest in taking charge of the idea. Someone also made the comment that distributions should make current versions of the kernel available for testing to make it easier for testers to try them out.
The Kernel Summit's closing event in the evening doubled as the opening party for the Linux Plumbers Conference. Participants at this community event voted in six of the ten new members of the Linux Foundation's Technical Advisory Board (TAB). James Bottomley, Kristen Carlson Accardi, Chris Mason, Dave Jones, Chris Wright and Christoph Hellwig were voted in. Hellwig, however, will have to stand for re-election again in a year, since he occupies the seat that Olaf Kirch prematurely resigned.
Well-known kernel developer Greg Kroah-Hartman made a keynote address on Wednesday at the opening of the Linux Plumbers Conference. In the address, he voiced sharp criticism for Ubuntu sponsor Canonical, whose employees, he said, contributed too little to the further development of the kernel and other open-source projects that constitute the basic software for Linux distributions. At the same time, Kroah-Hartman had to admit that Canonical had contributed more than five or six patches over the past few years, as he had claimed in a talk some months ago. In fact, according to his calculations the company had contributed a hundred patches between Linux 2.6.15 and 2.6.27-rc6; during the same period, Red Hat developers had contributed 11,846 patches and Novell developers 7222.
Canonical developers had also contributed very little to the development of X.org, GCC and binutils -- Red Hat developers were also the most active on these scores. Details and additional figures can be found in the presentation slides that Greg Kroah-Hartman released on his blog. Canonical CTO Matt Zimmerman took a stand in his blog on Greg Kroah-Hartman's keynote speech. He sharply criticised Greg Kroah-Hartman and called into question a number of the well-known kernel developer's conclusions. Other blogs are also taking a closer look at the keynote speech by Kroah-Hartman.
Kroah-Hartman's criticism is nothing new – other kernel developers have repeatedly criticised Canonical and Ubuntu developers in the past for their failure to take part in the further development of the main development branch and have accused them of not contributing their own improvements to the Ubuntu kernel. An article on LWN.net that deals with the keynote, available only to subscribers to LWN.net till September 25th, reports on many of these criticisms. In a previous article in response to the original talk, and readable by non-subscribers, Corbet has also defended Ubuntu. With his keynote address at the Linux Plumbers Conference, though, Kroah-Hartman has pushed these criticisms into the limelight.
- AMD has released version 8.9 of its graphics driver, known as Catalyst" or "Fglrx", for x86-32 and x86-64 Linux. These include a number of corrections and increase the number of supported Linux distributions.
- Harald Welte, who started work for VIA a few weeks ago as an open source consultant, explained VIA's open source plans in an interview on Slashdot.
Further background and information about developments in the Linux kernel and its environment can also be found in previous issues of the kernel log at heise open:
- Kernel-Log: Hard core Linux developers discuss the future of Linux at the Kernel Summit
- Kernel-Log: New stable and developer kernel, Mesa 7.1 and X-Server 1.5 released
- Kernel Log: New stable and pre-release kernels, Ubuntu 8.10 with 2.6.27?
- Kernel Log: New video drivers for AMD, Intel, Nvidia and VIA hardware
- Kernel Log: Kernel development explained, new Synaptics driver, Linux 2.6.27-rc3 published
- Kernel Log: Ath9k driver for Atheros Wifi in 2.6.27; reading material and videos for Linux experts