LKML Summary Podcast

Audio: http://media.libsyn.com/media/jcm/linux_kernel_podcast_20090916.mp3

For Wednesday, September 16th, 2009, I’m Jon Masters with a summary of today’s LKML traffic.

In today’s issue: Blackfin, Fanotify, KVM, and tracing.

Blackfin. Daniel Walker ranted about Mike Frysinger’s posting of Blackfin patches to the LKML. Mike responded that he hadn’t seen any complaints from others about large patch series being posted, and that those with “sane” mail clients wouldn’t have to deal with much more than skipping an entire thread anyway. Andrew Morton then steered the conversion towards why Blackfin wasn’t being pulled into linux-next, to which Mike replied with a request for further information about the process for getting his patches into Stephen’s tree.

Fanotify. In response to comments from Linus on the previous afternoon (in which Linus questioned “what’s so wonderful about fanotify that we would actually want yet-another-filesystem-notification-interface”), Eric Paris posted a rather detailed overview of the kinds of features offered by his interface. Chiefly, Eric cites the fact that fanotify passes an open fd with each event rather than “some arbitrary ‘watch descriptor’, along with an extensible data format, and a commitment from several anti-malware companies to use the new interface once it is available in the mainline kernel. Jamie Lokier sent a rather terse (however quite lengthy) response that criticised various features not currently available with fanotify (such as subtree notification) and suggested that his main concern was avoiding mistakes made with the previous inotify and dnotify mechanisms, while encouraging Eric to consider other use cases not covered by anti-malware consumers.

KVM. Today being a KVM Wednesday, there were a number of patches posted. First off were some rather cool looking patches from Avi Kivity implementing “just in time” MSR switching. These aim to reduce the need to perform expensive MSR update operations on guest pre-emption. KVM already optimizes to avoid such writes on every guest entry/exit, but it will now defer to the very last possible moment before performing such an update. Not to be outdone, Joerg Roedel posted some nested SVM “fixes and cleanups” aswell.

Tracing. Steven Rostedt inquired as to collective viewpoints on a dedicated tracing list, to be held on vger.kernel.org. The idea would be to have a separate place for discussions related to users of tracers and the “perf” performance counters utility. This “will not be a place for kernel development”, according to Steven, who wishes to address the issue of LKML being intimidating for those having questions, without causing development fragmentation by taking important discussions elsewhere.

In today’s pull requests: an XFS update from Alex Elder (including a large number of the fixes previously discussed from Christoph Hellwig), some networking and SPARC fixes from David Miller (containing quite a few important looking fixes in both cases), an official request to merge the latest version of Andi Kleen’s hwposion patches into 2.6.32, part two of some AMD64 EDAC updates from Borislav Petkov, some ext3 fixes from Jan Kara, an email from Roland Dreier (in which he also reminds us that he’s been showering and brushing his teeth, so shouldn’t be entirely ignored) asking Linus where his ummunotify request stands, some DLM updates for 2.6.32 from David Teigland, some tracing updates from Steven Rostedt, and the latest round of wireless patches from John Linville (who describes them as being “nothing too controversial”).

In today’s miscellaneous items: some further kmem and hwpoison bits from Fengguang Wu, a rant from Daniel Walker about Mike Frysinger’s posting of Blackfin arcihtectural patches to the LKML (which nobody else endorsed), a patch marking the SLQB allocator as “broken” on PowerPC and s390 from Pekka Enberg, a fix to use the correct export symbol to walk a system ram range in infiniband from Kamezawa Hiroyuki, a question about explicitly putting a PCI device into an ACPI D0 state from Michal Witkowski, a patch to disable preemption within stop_machine from Xiao Guangrong, a suggestion from Balbir Singh that the memcg patches in Andrew’s mm tree be merged for 2.6.32, version 3 of a patch series from Jiri Olsa implementing multiple pids within the tracing set_pid_ftrace file, a patch changing the name of the kernel thread managing an md raid device according to the type of RAID level in use (rather than the default of merely always using “5″) from Zen Chen Jin, an RFC patch series from Sheng Yang entitled “Xen Hybrid extension support” intending to allow guests to run in other than ring0 context (thus avoiding the TLB flush overhead in context switching between guest and hypervisor), some comments from Vivek Goyal concerning the fairness (or lack thereof) of the ioband patches when running on rotational media (as compared with his own IO scheduler based IO controller patches), a regression in kallsyms reported by Paul Mundt (who says that both Sam Ravnborg and Lai Jiangshan have yet to respond), version 2 of a patch adding support for walltime to ftrace from Zhao Lei, an RFC patch to handle a negative f_pos when manipulating /dev/kmem from Kamezawa Hiroyuki, some updates to the page-types utility from Fengguang Wu, version 4 of the post merge per-bdi writeback patches from Jens Axboe, the latest version of the “Memory Protection Units” (MPU – a simpler MMU alternative implementation) from Mike Frysinger and originally authored by Bernd Schmidt, a patch finally updating the MAINTAINERS file with the new location of the ARM linux mailing lists on infradead from Joe Perches, some linker script cleanups from Tim Abbott (enabling ksplice pre-requisites), a patch warning when selecting symbols with unmet direct dependencies in kbuild from Catalin Marinas, a fix for degraded performance when all inodes are under writeback from Jan Kara, a reiserfs patch also from Jan Kara, a defense of the vnet bus patches from Gregory Haskins (in which he lays out the case for kernel-to-kernel virtualizable communication), a fix for an oprofile related ring buffer regression from Christian Borntraeger, and some futex comments from Darren Hart.

Finally today, various responses came in related to the “Tricks to speed up kernel builds”. Thomas Fjellstrom says he prefers using icecream with make -jX, while David Lang provides further build analysis.

In today’s announcements: Greg Kroah-Hartman posted review patches for the stable 2.6.27.35 and 2.6.30.8 kernels, and Jakub Narebski announced that the Git User’s Survey 2009 had ended (for which results will be available soon). On that note, Junio C Hamano announced the release of git version 1.6.4.4 (which includes an important fix for users of github occasionally experiencing an HTTP 500 error response).

The latest kernel release was 2.6.31.

Andreas Mohr reported a regression in which USB autosuspend no longer functions after performing a suspend-to-RAM and resume cycle.

Stephen Rothwell posted a linux-next tree for September 16th. Since Tuesday, conflicts continue to bounce between various trees as Linus continues to perform various merges. Linus’ tree had a build failure (for which a patch was applied), the rr tree also had a build failure, while the drbd and staging trees lost their conflicts. The total sub-tree count remains steady at 140 trees in the latest compose. Stephen reminds everyone not to post patches destined for 2.6.33 until at least 2.6.32-rc1 has been released.

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