Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933461AbdGSXHs (ORCPT ); Wed, 19 Jul 2017 19:07:48 -0400 Received: from mail-pg0-f68.google.com ([74.125.83.68]:32952 "EHLO mail-pg0-f68.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753598AbdGSXGh (ORCPT ); Wed, 19 Jul 2017 19:06:37 -0400 Content-Type: text/plain; charset=utf-8 Mime-Version: 1.0 (Mac OS X Mail 10.3 \(3273\)) Subject: Re: [PATCH v2] KVM: VMX: Fix invalid guest state detection after task-switch emulation From: Nadav Amit In-Reply-To: Date: Wed, 19 Jul 2017 16:06:34 -0700 Cc: =?utf-8?B?UmFkaW0gS3LEjW3DocWZ?= , "linux-kernel@vger.kernel.org" , kvm , Paolo Bonzini , Wanpeng Li Message-Id: References: <1499757180-20439-1-git-send-email-wanpeng.li@hotmail.com> <20170719112926.GA13502@potion> <20170719161952.GC17303@potion> <539DB8ED-1315-41DA-8F61-1036D44E5E48@gmail.com> To: Wanpeng Li X-Mailer: Apple Mail (2.3273) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by nfs id v6JN9VLN003376 Content-Length: 4501 Lines: 106 Wanpeng Li wrote: > 2017-07-20 6:53 GMT+08:00 Nadav Amit : >> Wanpeng Li wrote: >> >>> 2017-07-20 0:25 GMT+08:00 Nadav Amit : >>>> Radim Krčmář wrote: >>>> >>>>> 2017-07-19 08:14-0700, Nadav Amit: >>>>>> Radim Krčmář wrote: >>>>>>> @@ -2363,6 +2368,8 @@ static unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu) >>>>>>> >>>>>>> static void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags) >>>>>>> { >>>>>>> + unsigned long old_rflags = to_vmx(vcpu)->rflags; >>>>>> >>>>>> It assumes rflags was decached from the VMCS before. Probably it is true, but… >>>>> >>>>> Right, it's better to use accessors everywhere, thanks. >>>>> The line should read: >>>>> >>>>> + unsigned long old_rflags = vmx_get_rflags(vcpu); >>>>> >>>>> ---8<--- >>>>> This can be reproduced by EPT=1, unrestricted_guest=N, emulate_invalid_state=Y >>>>> or EPT=0, the trace of kvm-unit-tests/taskswitch2.flat is like below, it >>>>> tries to emulate invalid guest state task-switch: >>>>> >>>>> kvm_exit: reason TASK_SWITCH rip 0x0 info 40000058 0 >>>>> kvm_emulate_insn: 42000:0:0f 0b (0x2) >>>>> kvm_emulate_insn: 42000:0:0f 0b (0x2) failed >>>>> kvm_inj_exception: #UD (0x0) >>>>> kvm_entry: vcpu 0 >>>>> kvm_exit: reason TASK_SWITCH rip 0x0 info 40000058 0 >>>>> kvm_emulate_insn: 42000:0:0f 0b (0x2) >>>>> kvm_emulate_insn: 42000:0:0f 0b (0x2) failed >>>>> kvm_inj_exception: #UD (0x0) >>>>> >>>>> It appears that the task-switch emulation updates rflags (and vm86 flag) >>>>> only after the segments are loaded, causing vmx->emulation_required to >>>>> be set, when in fact invalid guest state emulation is not needed. >>>>> >>>>> This patch fixes it by updating vmx->emulation_required after the rflags >>>>> (and vm86 flag) is updated. >>>>> >>>>> Suggested-by: Nadav Amit >>>>> Signed-off-by: Wanpeng Li >>>>> [Wanpeng wrote the commit message with initial patch and Radim moved the >>>>> update to vmx_set_rflags and added Paolo's suggestion for the check.] >>>>> Signed-off-by: Radim Krčmář >>>>> --- >>>>> arch/x86/kvm/vmx.c | 15 ++++++++++----- >>>>> 1 file changed, 10 insertions(+), 5 deletions(-) >>>>> >>>>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c >>>>> index 84e62acf2dd8..a776aea0043a 100644 >>>>> --- a/arch/x86/kvm/vmx.c >>>>> +++ b/arch/x86/kvm/vmx.c >>>>> @@ -2326,6 +2326,11 @@ static void vmx_vcpu_put(struct kvm_vcpu *vcpu) >>>>> __vmx_load_host_state(to_vmx(vcpu)); >>>>> } >>>>> >>>>> +static bool emulation_required(struct kvm_vcpu *vcpu) >>>>> +{ >>>>> + return emulate_invalid_guest_state && !guest_state_valid(vcpu); >>>>> +} >>>>> + >>>>> static void vmx_decache_cr0_guest_bits(struct kvm_vcpu *vcpu); >>>>> >>>>> /* >>>>> @@ -2363,6 +2368,8 @@ static unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu) >>>>> >>>>> static void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags) >>>>> { >>>>> + unsigned long old_rflags = vmx_get_rflags(vcpu); >>>>> + >>>>> __set_bit(VCPU_EXREG_RFLAGS, (ulong *)&vcpu->arch.regs_avail); >>>>> to_vmx(vcpu)->rflags = rflags; >>>>> if (to_vmx(vcpu)->rmode.vm86_active) { >>>>> @@ -2370,6 +2377,9 @@ static void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags) >>>>> rflags |= X86_EFLAGS_IOPL | X86_EFLAGS_VM; >>>>> } >>>>> vmcs_writel(GUEST_RFLAGS, rflags); >>>>> + >>>>> + if ((old_rflags ^ rflags) & X86_EFLAGS_VM) >>>>> + to_vmx(vcpu)->emulation_required = emulation_required(vcpu); >>>> >>>> Sorry for not pointing it before, but here you compare the old_rflags with >>>> the new rflags but after you already “massaged” it. So the value you compare >>>> with is not what the guest “sees”. >>> >>> So you mean we should use unsigned long old_rflags = >>> vmcs_readl(GUEST_RFLAGS); right? >> >> No. The problem is not with old_rflags now, but with rflags. If vm86_active, >> then rflags is changed and you don’t compare the guest-visible rflags >> anymore. > > Ah, I see. So we should compare the old_flags with the > rmode->save_rflags(guest-visible rflags) instead of the rflags (shadow > rflags), right? Not exactly, since rmode->save_rflags are invalid if !vm86_active. Instead, I think you should have a save_rflags variable on the stack that would hold the rflags before “massaging” and use it instead.