• Immutable Page
  • Info
  • Attachments

KernelPodcast20090519

For Tuesday, May 19th 2009, I'm Jon Masters with a summary of today's LKML traffic.

In today's issue: Xen, tracepoints, RAID6, and MCE.

Xen. Jeremy Fitzhardinge's hopes for a smooth merge of dom0 patches were dashed again. This time, the comments concerned its intrusive changes to DMA mapping and swiotlb vs. Xen IOMMU. Fujita Tomonori stated that, if Xen dom0 support had to be merged, he really prefered some code duplication rather than increased complexity from, as he put it, "ugly unified code".

Tracepoints. Jason Baron noticed a performance quirk in the tracepoint code. It turns out that arguments that are passed as macros or that perform dereferences and needlessly evaluated prior to the initial tracepoint on/off check comparison. This obviously impacts the likely tracepoint off case with unnecessary overhead. Jason proposes to fix this by introducing a new macro and a slight change to each callsite.

RAID6. Dan Williams posted version two of a 3 topic series aimed at adding asychronous RAID6 acceleration by extending the async_tx API to include support for offloading RAID6 operations to hardware. The latest version of the patch series addresses some concerns from Peter Anvin about the recovery code, Neil Brown's concerns regarding kernel stack usage, and a number of other items. The code is also updated to order to pass a new series of RAID6 recovery self tests that have been included in-kernel.

MCE. Andi Kleen posted a series of 20 patches against the mce2 branch of the tip x86 tree. This ongoing cleanup work against the x86 tree aims to cleanup and unify the 32 and 64-bit MCE handling code, implement MCE inject (that is to say artificially synthesized Machine Check Exceptions), and ensure all the old bug work arounds are synchronized between the different codebases. It is all intended to be included into the 2.6.31 merge window.

In today's announcements: No major announcements today. Most developers are quietly preparing for the impending 2.6.31 merge window onslaught.

Stephen Rothwell posted a linux-next tree for May 19th. Since Monday, a new edac-amd tree was introduced and immediately dropped due to a build problem, the net tree gained a conflict against Linus' tree, the vfs tree gained a conflict against the ext4 tree, and the block tree lost a conflict. Stephen posted a number of emails citing individual tree failures, including net and vfs (with potential fixes including several changes to an FCoE driver). He also raised some miscellaneous items in a separate mail, suggesting that he will continue to cross-post about build failures to both the linux-next and the LKML lists. He suggests that most developers are likely preparing their trees for the next merge window and thus that he should be prepared for the inevitable forthcoming onslaught.

The latest kernel release is 2.6.30-rc6. An rc7 is expected to be released before week's end. Likely on Friday afternoon.

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 2009-05-20 07:28:09 by JonMasters