From mboxrd@z Thu Jan 1 00:00:00 1970 From: Avi Kivity Subject: Re: [PATCH 0/4] KVM: srcu-less dirty logging Date: Tue, 28 Feb 2012 12:03:27 +0200 Message-ID: <4F4CA66F.8090109@redhat.com> References: <20120223203300.241510a6.yoshikawa.takuya@oss.ntt.co.jp> <1330003524.11248.11.camel@twins> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: Takuya Yoshikawa , mtosatti@redhat.com, kvm@vger.kernel.org, linux-kernel@vger.kernel.org, paulmck@linux.vnet.ibm.com To: Peter Zijlstra Return-path: In-Reply-To: <1330003524.11248.11.camel@twins> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On 02/23/2012 03:25 PM, Peter Zijlstra wrote: > On Thu, 2012-02-23 at 20:33 +0900, Takuya Yoshikawa wrote: > > - Stop allocating extra dirty bitmap buffer area > > > > According to Peter, mmu_notifier has become preemptible. If we can > > change mmu_lock from spin_lock to mutex_lock, as Avi said before, this > > would be staightforward because we can use __put_user() right after > > xchg() with the mmu_lock held > > So the 'only' thing to consider is running the end result with lockdep > enabled since the mmu locks are rather deep in the nesting tree its very > easy to accidentally cause inversions. There will be an inversion for sure, if __put_user() faults and triggers an mmu notifier (perhaps directly, perhaps through an allocation that triggers a swap). -- error compiling committee.c: too many arguments to function