From mboxrd@z Thu Jan 1 00:00:00 1970 From: Paolo Bonzini Subject: Re: [PATCH] kvm: add halt_poll_ns module parameter Date: Tue, 10 Feb 2015 08:50:47 +0100 Message-ID: <54D9B857.7080500@redhat.com> References: <1423226937-11169-1-git-send-email-pbonzini@redhat.com> <54D92005.2060308@de.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit Cc: riel@redhat.com, mtosatti@redhat.com, rkrcmar@redhat.com, jan.kiszka@siemens.com, dmatlack@google.com To: Christian Borntraeger , linux-kernel@vger.kernel.org, kvm@vger.kernel.org Return-path: In-Reply-To: <54D92005.2060308@de.ibm.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On 09/02/2015 22:00, Christian Borntraeger wrote: > I can confirm that this also helps uperf with a 1/1 byte round trip work load > between guests on s390. And I can confirm the higher CPU load. This is normally > a no-go for the typical s390 users, which utilize their systems as much as > possible. Your check for single_task_running could actually solve that > problem because on overcommitment this will never switch to polling if the > runqueues get full. > Since this is also runtime configurable and defaults to 0 it should be pretty > painless. > > The only question is: is there a sane way of doing autotuning? The answer is: we'll see. :) I have some ideas, I have to run them through the necessary experiments. Paolo