• Immutable Page
  • Info
  • Attachments

KernelPodcast20090528

Direct from Central Square in Cambridge, Massachusetts, for Thursday, May 28th, I'm Jon Masters with a summary of today's LKML traffic.

In today's issue: ARM devicetree support, ftrace, per-BDI flusher threads, and trusted boot technology.

Flattened device tree. Since the original ARM fdt patches were posted on Wednesday, a heated argument has taken place concerning the relative merits of using OpenFirmware-esque device trees, especially on embedded platforms. The idea behind fdt is that a single kernel (image) can support many different devices without all kinds of device-specific hacks, since the required information can be passed using a generic device tree structure supplied by the firmware. This, in turn is generated by the system vendor as a BLOB that the firmware keeps care of (take, for example, an example of how Xilinx have now modified their EDK tools to generate these BLOBs automatically). In many ways, it superceeds bootinfo (bdinfo) and similar structures that previously existed in the kernel. Your author is obviously one of those strongly in favor of it.

Ftrace. Steven Rostedt posted an updated ftrace tree. This version expands existing support for kernel command line function filtering using the "ftrace=function" format string on the command line to trace functions on bootup with the addition of "ftrace_notrace" and "ftrace_filter" to add filtering of those functions, akin to the regular filtering available using debugfs interfaces on booted systems. That should help reduce noise. Also on the subject of ftrace, Wu Zhangjin posted mips-specific ftrace support.

Per-BDI writeback flusher threads. Jens Axboe posted version 9 of his patch implementing per-bdi writeback flusher threads. Since the previous release, he has fixed an on-stack allocation caused hang (remember, allocating anything more than the most insignificant memory on the fixed size kernel stack is bad) that Ted T'so had reported (complete with very length compressed dmesg log), gotten rid of explicit wait queues, and added a sync_supers thread that makes sure dirty superblocks get written.

Trusted Boot Technology. James Morris posted a link to some slides from the InvisibleThings folks on the new Intel TXT support. He says that this should help people understand some of the benefits and pitfalls of this technology at the high level.

In today's announcements: The Montreal Linux Power Management mini-summit. A reminder that a mini-summit will take place concurrent with the Montreal Linux Symposium (that's OLS on the road this year). A limit of 20 participents has been set and so interested parties should email lenb@kernel.org asap.

The latest kernel release remains 2.6.30-rc7, which was released by Linus over the US Memorial Day Weekend (on Saturday evening). An rc8 release is anticipated at any moment now.

Stephen Rothwell posted a linux-next tree for May 28th. Since Wednesday, the pxa tree gained a build failure so the version from Monday was used, the timers tree gained a conflict against the sched tree, and the omap tree gained a conflict against the arm tree. The linux-next tree still fails to build in an allyesconfig configuration for powerpc. The total subtree count remains 141.

Finally today, Alan Cox likes to live dangerously. In an 8250 serial patch, he added an "Impact" line, and in refering to Linus' recent slamming of the idea, wrote "ok I just put this here to wind up Linus".

That's a summary of today's LKML traffic. For further information visit kernel.org. I'm Jon Masters.

Tell others about this page:

last edited 2009-05-29 15:50:56 by JonMasters