Releasing rc3 on a Saturday, not to catch any people trying to game the rc release timing, but simply because I'll be traveling for the kernel summit tomorrow.
It turns out that the bug that we thought was due to the new virtually mapped stacks during the rc2 release wasn't due to that at all, but a block request queuing race condition. So people who turned off the new feature weren't actually avoiding it at all, but probably the only people who hit it were people like DaveJ who are doing stress-testing. But it's all fixed now, and we should be all set. There's a lot of other fixes all over too. The diffstat looks a bit unusual, with xfs showing up almost as much as all the random driver fixes, but it's mainly fixes for the new reflink code merged during this release that you wouldn't have hit anyway unless you are using the fancy new stuff. Other than that? All the usual random driver fixes, arch updates, etc. The appended shortlog isn't exactly tiny (as expected, rc3 was bigger than rc2), but it's fairly easy to just skim it and get some feeling for the kinds of things that have been going on. So go out and test. Linus