From mboxrd@z Thu Jan 1 00:00:00 1970 From: Gleb Natapov Subject: Re: [PATCH] kvm: make vcpu life cycle separated from kvm instance Date: Mon, 5 Dec 2011 12:18:42 +0200 Message-ID: <20111205101842.GH12507@redhat.com> References: <4ED212A1.3060300@redhat.com> <1322807192-14147-1-git-send-email-kernelfans@gmail.com> <4EDB4A0D.1000609@redhat.com> <4EDC8F4B.7040308@redhat.com> <20111205094240.GE12507@redhat.com> <4EDC95E0.20503@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Liu ping fan , kvm@vger.kernel.org, linux-kernel@vger.kernel.org, aliguori@us.ibm.com, jan.kiszka@web.de To: Avi Kivity Return-path: Content-Disposition: inline In-Reply-To: <4EDC95E0.20503@redhat.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On Mon, Dec 05, 2011 at 11:58:56AM +0200, Avi Kivity wrote: > On 12/05/2011 11:42 AM, Gleb Natapov wrote: > > On Mon, Dec 05, 2011 at 11:30:51AM +0200, Avi Kivity wrote: > > > On 12/05/2011 07:29 AM, Liu ping fan wrote: > > > > like this, > > > > #define kvm_for_each_vcpu(idx, cnt, vcpup, kvm) \ > > > > for (idx = 0, cnt = 0, vcpup = kvm_get_vcpu(kvm, idx); \ > > > > cnt < atomic_read(&kvm->online_vcpus) && \ > > > > idx < KVM_MAX_VCPUS; \ > > > > idx++, (vcpup == NULL)?:cnt++, vcpup = kvm_get_vcpu(kvm, idx)) \ > > > > if (vcpup == NULL) \ > > > > continue; \ > > > > else > > > > > > > > > > > > A little ugly, but have not thought a better way out :-) > > > > > > > > > > #define kvm_for_each_vcpu(vcpu, it) for (vcpu = kvm_fev_init(&it); vcpu; > > > vcpu = kvm_fev_next(&it, vcpu)) > > > > > > Though that doesn't give a good place for rcu_read_unlock(). > > > > > > > > Why not use rculist to store vcpus and use list_for_each_entry_rcu()? > > We can, but that's a bigger change. > Is it? I do not see a lot of accesses to vcpu array except those loops. -- Gleb.