From mboxrd@z Thu Jan 1 00:00:00 1970 From: Marcelo Tosatti Subject: Re: [PATCH 03/10] KVM: SVM: Move nested INTR #vmexit into preemtible code Date: Wed, 7 Oct 2009 17:58:52 -0300 Message-ID: <20091007205852.GA7692@amt.cnet> References: <1254925888-13743-1-git-send-email-joerg.roedel@amd.com> <1254925888-13743-4-git-send-email-joerg.roedel@amd.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Avi Kivity , kvm@vger.kernel.org, linux-kernel@vger.kernel.org To: Joerg Roedel Return-path: Received: from mx1.redhat.com ([209.132.183.28]:10660 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S934352AbZJGU7e (ORCPT ); Wed, 7 Oct 2009 16:59:34 -0400 Content-Disposition: inline In-Reply-To: <1254925888-13743-4-git-send-email-joerg.roedel@amd.com> Sender: kvm-owner@vger.kernel.org List-ID: On Wed, Oct 07, 2009 at 04:31:21PM +0200, Joerg Roedel wrote: > This patch makes use of the KVM_REQ_VMEXIT to move the > emulation of #vmexit(INTR) out of non-preemptible code. > > Signed-off-by: Joerg Roedel > --- > arch/x86/kvm/svm.c | 18 ++++++++++++++++-- > 1 files changed, 16 insertions(+), 2 deletions(-) > > diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c > index b6ce1a9..7015680 100644 > --- a/arch/x86/kvm/svm.c > +++ b/arch/x86/kvm/svm.c > @@ -1379,8 +1379,14 @@ static inline int nested_svm_intr(struct vcpu_svm *svm) > > svm->vmcb->control.exit_code = SVM_EXIT_INTR; > > - if (nested_svm_exit_handled(svm)) { > - nsvm_printk("VMexit -> INTR\n"); > + if (svm->nested.intercept & 1ULL) { > + /* > + * The #vmexit can't be emulated here directly because this > + * code path runs with irqs and preemtion disabled and a > + * #vmexit emulation might sleep. Only set the request bit for > + * the #vmexit here. > + */ > + set_bit(KVM_REQ_VMEXIT, &svm->vcpu.requests); > return 1; > } What if you keep this internal to SVM? Proceed to svm_vcpu_run and return, do the emulation on the exit handler. Then there's no need for the request bit (VMX does that, see vmx_vcpu_run).