LKML Summary Podcast

Audio: http://media.libsyn.com/media/jcm/linux_kernel_podcast_20091130.mp3

For Monday, November 30th, 2009, I’m Jon Masters with a summary of today’s LKML traffic.

In today’s issue: Block IO controller, gdb, infra-red, interrupts, and networking.

Block IO Controller. Vivek Goyal posted version 4 of his latest Block IO Controller patches (which, you may remember were rewritten post-Kernel Summit to factor in the goals of a variety of other parties). The patches enable tasks to be grouped and assigned various IO bandwidth allowances. Vivek is clearly interested in seeing them hit mainline soon. The latest version incorporates a fair amount of testing and cleanups.

Gdb. Srikar Dronamraju posted an RFC patch implementing an in-kernel gdbstub using the utrace infrastructure. The current patch implements a /proc/pid/gdb entry per process that can be accessed remotely via the GDB serial protocol.

Infra-red. Jarod Wilson, freshly back from the US Thanksgiving weekend chimed in on the ongoing “should we create a raw input interface for IR” thread, noting that he would contact a few IR vendors he has connections with and see what level of support there would be for participating upstream. Alan Cox wondered whether it was really a good idea to shove 50K of IR protocol code into the kernel, but Mauro Carvalho Chehab noted that the same could be said of mice, keyboards, and serial consoles. None of this needs to be in the kernel, but it may be required for user interation, or maybe during boot.

Interrupts. Thomas Gleixner pondered aloud why will still need IRQF_DISABLED interrupt handlers (interrupt handlers that specify whether they must be run without interrupts enabled on a global level at the time) and in particular why it is even a good idea to be running any interrupt handlers while interrupts are enabled. He traces back some of the history and boils the reasoning for having interrupts enabled down to two things: 1). handlers that rely upon jiffies being updated (of which there are few at best). 2). handlers that run for a long time. The latter typically need to be fixed. Thomas requests input from others about killing off IRQF_DISABLED. Ben Herrenschmidt and others made a number of interesting points.

Network. David Miller posted some network updates, mostly to drivers. The changes seemed innocuous enough, until William Allen Simpson popped up and noted that they had made a huge number of changes to coding style and this causes various patches he is working on to break. Specifically, the format of conditional statements was “sweeping[ly]” changed to use the trailing form in which multiple conditions appear on the same line of an “if” statement. As William points out, this causes issues with patches (which now have to affect multiple lines, perhaps increasing conflicts), and CodingStyle is mute on the topic. He would like to see some kind of guidance as to the best practice. In fairness, William didn’t copy David Miller on his grumbling.

Finally today, Dan Carpenter popped up again to let us know that he’s still in Africa, and still on some fairly expensive bandwidth ($0.05USD per MB), with the constraint that he can only buy it in 200MB increments. This isn’t enough to clone a complete kernel git tree, so he would like it if someone could post an rsyncable tree somewhere that he could use to seed his repository. Once he has a tree, the daily changesets shouldn’t be a huge issue. As Avi Kivity noted, the git trees on kernel.org are already available over HTTP.

In today’s announcements: util-linux-ng version 2.16.2. Karel Zak announced the latest version of the util-linux-ng utilities has been released. This is a stable maintainance release, as opposed to the recently announced .17 rc.

The latest kernel release is 2.6.32-rc8.

Stephen Rothwell posted a linux-next tree for November 30th. Since Friday, the pxa, pcmcia, trivial, and sysctl trees had conflicts, while the sparc, v4l-dvb, net, rr, and osd trees lost theirs. The sub-tree count remained steady at 154 trees.

That’s a summary of today’s Linux Kernel Mailing List traffic, for further information visit www.kernel.org. I’m Jon Masters.