From mboxrd@z Thu Jan 1 00:00:00 1970 From: Paolo Bonzini Subject: Re: [PATCH] kvm: irqchip: Break up high order allocations of kvm_irq_routing_table Date: Mon, 11 May 2015 15:27:26 +0200 Message-ID: <5550AE3E.6060705@redhat.com> References: <1431088304-11365-1-git-send-email-joro@8bytes.org> <554CE3A5.7000101@redhat.com> <20150511112522.GJ5438@suse.de> <5550964B.6020001@redhat.com> <5550A5A3.5010403@de.ibm.com> <5550A661.9000909@de.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit Cc: Joerg Roedel , Gleb Natapov , kvm@vger.kernel.org, linux-kernel@vger.kernel.org To: Christian Borntraeger , Joerg Roedel Return-path: In-Reply-To: <5550A661.9000909@de.ibm.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On 11/05/2015 14:53, Christian Borntraeger wrote: > Am 11.05.2015 um 14:50 schrieb Christian Borntraeger: > >> s390 has a route per device, but with 100 virtio-blk devices the difference seem >> pretty much on the "dont care" side. qemu aio-poll/drain code seems to cause >> much more delay since we elimited the kernel delays by using >> synchronize_srcu_expedited. > > This is ambiguous: > My point is: qemu did not get slower, it was already slow and the kernel got fast > enough that it is no longer the slowest part. Great, I'll apply the patch. Paolo