From mboxrd@z Thu Jan 1 00:00:00 1970 From: Liran Alon Subject: Re: [PATCH] KVM: nVMX: Fix races when sending nested PI while dest enters/leaves L2 Date: Sat, 11 Nov 2017 00:59:51 +0200 Message-ID: <5A062F67.7050709@ORACLE.COM> References: <1510252040-5609-1-git-send-email-liran.alon@oracle.com> <20171110180616.GE15561@flask> <0b202d55-6c87-e9b9-49aa-74aa7e83ff6b@redhat.com> <20171110223018.GM2189@flask> <5A062C8A.8000307@ORACLE.COM> <1774540798.19563775.1510354270639.JavaMail.zimbra@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8bit Cc: =?UTF-8?B?UmFkaW0gS3LEjW3DocWZ?= , kvm@vger.kernel.org, jmattson@google.com, wanpeng li , idan brown , Krish Sadhukhan To: Paolo Bonzini Return-path: Received: from aserp1040.oracle.com ([141.146.126.69]:31231 "EHLO aserp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754310AbdKJXAB (ORCPT ); Fri, 10 Nov 2017 18:00:01 -0500 In-Reply-To: <1774540798.19563775.1510354270639.JavaMail.zimbra@redhat.com> Sender: kvm-owner@vger.kernel.org List-ID: On 11/11/17 00:51, Paolo Bonzini wrote: > > Liran Alon wrote: >> On 10/11/17 23:37, Paolo Bonzini wrote: >>> On 10/11/2017 19:06, Radim Krčmář wrote: >>>>> /* the PIR and ON have been set by L1. */ >>>>> if (!kvm_vcpu_trigger_posted_interrupt(vcpu, true)) { >>>> This would still fail on the exiting case. >>>> >>>> If one VCPU was just after a VM exit, then the sender would see it >>>> IN_GUEST_MODE, send the posted notification and return true, but the >>>> notification would do nothing >>> >>> It would cause *something*---a vmexit because the vector doesn't match >>> the L1 posted interrupt. Then smp_kvm_posted_intr_nested_ipi would be >>> invoked from vmx_handle_external_intr. >>> >>> Could we detect the vector in vmx_handle_external_intr and set >>> pi_pending+KVM_REQ_EVENT? Or invoke a function in KVM from >>> smp_kvm_posted_intr_nested_ipi? Or would both be insane?... >> >> Consider the case sender sees vcpu->mode==IN_GUEST_MODE and before it >> sends the physical IPI, dest CPU exits from guest and continues in L0 >> all the way until vcpu_enter_guest() and pass the part it checks for >> KVM_REQ_EVENT but before it disables interrupts. Then sender sends the >> physical IPI which is received in host-context and therefore runs >> smp_kvm_posted_intr_nested_ipi() which sets KVM_REQ_EVENT & >> pi_pending=true. But without Radim's suggestion of checking pi_pending >> after interrupts disabled, this is too late as dest CPU will not check >> these again until next exit from L2 guest. > > The destination CPU checks vcpu->requests before entering the guest, so this > case would be handled fine (see Documentation/virtual/kvm/vcpu-requests.rst > for an explanation of IN_GUEST_MODE, vcpu->requests, and PIR.ON). > Oh I see. You are right. My bad. So I think your suggestion also works and may be simpler but I would further think about it before sending a v2. Thanks for clarifying this. -Liran > Thanks, > > Paolo > >> I hope I didn't misunderstand something here :)