Comment 17 for bug 855199

Things are a bit more complicated than I thought, and you should disregard what I said in comments 12, 13, and 14. Apologies.

The regression appears dependent on the kernel configuration. As I was building test images, I was shutting off many drivers and options that I hoped were irrelevant , to cut down the long build times. Some of those options *are* relevant; I have not narrowed it down yet.

So! The regression does first appear on March 16, in this commit:

  Merge branch 'irq-core-for-linus'

  http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=5f6fb45466b2273ffb91c9cf209f164f666c33b1

I have built that kernel with a fuller config that shows the bug, and a stripped down one that does not. When I next get time, I'll work to pin down the significant options.

I built the previous snapshot, with the same config as the buggy build, but it had not problems. So, for my machines at least, the "5f6fb45" commit appears to contain the regression.

mfg, I have before-and-after images, that you can test if you'd like. I"m not sure if 80M of debs are proper to attach to the bug report?