Date: Thu, 28 Apr 2016 21:08:06 +0200
From: Radim =?utf-8?B?S3LEjW3DocWZ?= <rkrcmar@redhat.com>
To: Bruce Rogers <brogers@suse.com>
Cc: kvm@vger.kernel.org, linux-kernel@vger.kernel.org, namit@cs.technion.ac.il,
        stable@vger.kernel.org
Subject: Re: [PATCH] KVM: x86: fix ordering of cr0 initialization code in
 vmx_cpu_reset
Message-ID: <20160428190805.GA9655@potion>
References: <1461351363-28094-1-git-send-email-brogers@suse.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1461351363-28094-1-git-send-email-brogers@suse.com>
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1451
Lines: 33

2016-04-22 12:56-0600, Bruce Rogers:
> Commit d28bc9dd25ce reversed the order of two lines which initialize cr0,
> allowing the current (old) cr0 value to mess up vcpu initialization.
> This was observed in the checks for cr0 X86_CR0_WP bit in the context of
> kvm_mmu_reset_context(). Besides, setting vcpu->arch.cr0 after vmx_set_cr0()
> is completely redundant. Change the order back to ensure proper vcpu
> initialization.
> 
> The combination of booting with ovmf firmware when guest vcpus > 1 and kvm's
> ept=N option being set results in a VM-entry failure. This patch fixes that.

Greg pointed out missing,
  Cc: stable@vger.kernel.org
when stable@vger.kernel.org was Cc'd. Adding
  Fixes: d28bc9dd25ce ("KVM: x86: INIT and reset sequences are different")
would be nice too (even when it is redundant).

> Signed-off-by: Bruce Rogers <brogers@suse.com>
> ---
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> @@ -5046,8 +5046,8 @@ static void vmx_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event)
>  	cr0 = X86_CR0_NW | X86_CR0_CD | X86_CR0_ET;
> -	vmx_set_cr0(vcpu, cr0); /* enter rmode */
>  	vmx->vcpu.arch.cr0 = cr0;
> +	vmx_set_cr0(vcpu, cr0); /* enter rmode */

So vmx_set_cr0() has a code that depends on vmx->vcpu.arch.cr0 being
already set the to new value.  Do you know what function is it?

I think we better set vmx->vcpu.arch.cr0 early in vmx_set_cr0().
Or do other callsites somehow depend on the old cr0 value?

Thanks.