On 09/20/2010 09:30 AM, Stefan Seyfried wrote:
On Mon, 20 Sep 2010 09:11:36 -0400 Jeff Mahoney
wrote: It should be dropped. We had it because there was a very, very heavy performance overhead in it when we released SLE11 (2.6.27) but it has since been fixed.
This is a mainline fix, I assume? (I'm going to talk about cgroups this week, so this might be important information for my listeners ;-)
Yep. The issue was that there was a contended spinlock in the middle of the fast path for memalloc. For low numbers of CPUs it was hardly noticeable but for higher numbers of CPUs it had a serious impact. It was fixed[1] in mainline shortly after LPC last year and included in 2.6.32. -Jeff [1] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdif... -- Jeff Mahoney SUSE Labs -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org