The 2.6.26 kernel changed the DMA behavior of the snd_hda_intel driver. On many chipsets, this resulted in extremely high events/0 CPU usage (I observed >30% CPU usage on my quad-core machine in some circumstances). The discussion of this bug begins here: http://linux.derkeiler.com/Mailing-Lists/Kernel/2008-06/msg02731.html and concludes http://linux.derkeiler.com/Mailing-Lists/Kernel/2008-06/msg03976.html and http://git.kernel.org/?p=linux/kernel/git/tiwai/sound-2.6.git;a=commitdiff;h=0c6341ace5efb14a0da08482e299b3c158e0d5fb Essentially, the problem was that ATI and NVidia's versions of hda turned out to behave somewhat differently from Intel's (the snd_hda_intel driver is used for many vendors' sound cards, not just Intel's). The fixes have been integrated upstream (in Takashi Iwai's git tree). I suggest that this bug should also be fixed in the current 2.6.26 gentoo-sources.
Created attachment 165878 [details, diff] hda-intel-bdl_pos_adj.patch Patch for 2.6.26 to fix ALSA CPU usage. Derived from the relevant parts of Takashi Iwai's git tree.
Thanks for the report and the patch. Assigning to maintainers.
There is a handful of dependent patches that this patch requires. This will take some more effort and testing so as not to introduce any new problems with adding multiple patches. http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=history;f=sound/pci/hda/hda_intel.c;h=1c53e337ecb2dce93b833dcce526f64288cc9136;hb=HEAD
Can someone please test in gentoo-sources-2.6.27 and port the result here
(In reply to comment #4) > Can someone please test in gentoo-sources-2.6.27 and port the result here > /port/post
(In reply to comment #4) > Can someone please test in gentoo-sources-2.6.27 and port the result here The patch has already been integrated into 2.6.27. I tested gentoo-sources-2.6.27 on the same machine that was suffering from this bug with 2.6.26, and the events/0 CPU usage is normal (just like with patched 2.6.26).
Closing as gentoo-sources-2.6.27 has been release
