From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jan Kiszka <jan.kiszka@siemens.com>
Subject: Re: [PATCH 9/9] [SVM] inject NMI after IRET from a previous NMI,
 not before.
Date: Tue, 05 May 2009 11:25:13 +0200
Message-ID: <4A0005F9.4050306@siemens.com>
References: <1241511275-2261-1-git-send-email-gleb@redhat.com> <1241511275-2261-9-git-send-email-gleb@redhat.com> <49FFFCA0.3070507@siemens.com> <20090505090314.GN9795@redhat.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Cc: avi@redhat.com, kvm@vger.kernel.org
To: Gleb Natapov <gleb@redhat.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from lizzard.sbs.de ([194.138.37.39]:20798 "EHLO lizzard.sbs.de"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1751965AbZEEJZe (ORCPT <rfc822;kvm@vger.kernel.org>);
	Tue, 5 May 2009 05:25:34 -0400
In-Reply-To: <20090505090314.GN9795@redhat.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

Gleb Natapov wrote:
> On Tue, May 05, 2009 at 10:45:20AM +0200, Jan Kiszka wrote:
>>> @@ -2331,8 +2358,16 @@ static void enable_nmi_window(struct kvm_vcpu *vcpu)
>>>  {
>>>  	struct vcpu_svm *svm = to_svm(vcpu);
>>>  
>>> -	if (svm->vmcb->control.int_state & SVM_INTERRUPT_SHADOW_MASK)
>>> -		enable_irq_window(vcpu);
>>> +	if ((svm->vcpu.arch.hflags & (HF_NMI_MASK | HF_IRET_MASK))
>>> +	    == HF_NMI_MASK)
>>> +		return; /* IRET will cause a vm exit */
>>> +
>>> +	/* Something prevents NMI from been injected. Single step over
>>> +	   possible problem (IRET or exception injection or interrupt
>>> +	   shadow) */
>>> +	vcpu->arch.singlestep = true;
>>> +	svm->vmcb->save.rflags |= (X86_EFLAGS_TF | X86_EFLAGS_RF);
>> Can you single-step like this out of an IRQ handler? I mean, IRET will
>> restore the flags from the stack, and those settings should be
>> overwritten. Or am I missing something?
>>
> It seems to be working :) Shouldn't CPU checks single step before
> executing IRET and thus using old flags value? It is interesting to
> check what rflag value is immediately after IRET.

Hmm, guess I have to re-read some manuals. But regarding
rflags-after-iret, I think it should be cleared due to that restoring
from the stack.

Jan

-- 
Siemens AG, Corporate Technology, CT SE 2
Corporate Competence Center Embedded Linux