From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <tglx@linutronix.de>
Received: from mail.linutronix.de (146.0.238.70:993) by
  crypto-ml.lab.linutronix.de with IMAP4-SSL for <speck@linutronix.de>; 02 Jul
  2018 17:01:26 -0000
Received: from p4fea482e.dip0.t-ipconnect.de ([79.234.72.46]
 helo=nanos.glx-home)	by Galois.linutronix.de with esmtpsa
 (TLS1.2:DHE_RSA_AES_256_CBC_SHA256:256)	(Exim 4.80)	(envelope-from
 <tglx@linutronix.de>)	id 1fa2Cb-0007m6-JE	for speck@linutronix.de; Mon, 02
 Jul 2018 19:01:25 +0200
Date: Mon, 2 Jul 2018 19:01:22 +0200 (CEST)
From: Thomas Gleixner <tglx@linutronix.de>
Subject: Re: [patch V5 05/10] KVM magic # 5
In-Reply-To: <20180702163559.GD17137@char.US.ORACLE.com>
Message-ID: <alpine.DEB.2.21.1807021859340.1590@nanos.tec.linutronix.de>
References: <20180702154426.910579106@linutronix.de>
 <20180702160528.906801335@linutronix.de>
 <20180702163559.GD17137@char.US.ORACLE.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit
To: speck@linutronix.de
List-ID: <speck.linutronix.de>

On Mon, 2 Jul 2018, speck for Konrad Rzeszutek Wilk wrote:
> > -static void __maybe_unused vmx_l1d_flush(void)
> > +static void vmx_l1d_flush(struct kvm_vcpu *vcpu)
> >  {
> >  	int size = PAGE_SIZE << L1D_CACHE_ORDER;
> > +	bool always;
> > +
> > +	/*
> > +	 * If the mitigation mode is 'flush always', keep the flush bit
> > +	 * set, otherwise clear it. It gets set again either from
> > +	 * vcpu_run() or from one of the unsafe VMEXIT handlers.
> > +	 */
> > +	always = vmentry_l1d_flush == VMENTER_L1D_FLUSH_ALWAYS;
> > +	vcpu->arch.l1tf_flush_l1d = always;
> 
> <blinks> You did the reset of arch.l1tf_flush_l1d _after_ we have done
> this vmx_l1d_flush call, nice!! So obvious in retrospect.

:)

> > @@ -4799,6 +4800,8 @@ int kvm_read_guest_virt(struct kvm_vcpu
> >  {
> >  	u32 access = (kvm_x86_ops->get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0;
> >  
> > +	/* The gva_to_pa walker can pull in tons of pages. */
> > +	vcpu->arch.l1tf_flush_l1d = true;
> 
> 
> I think also kvm_write_guest_virt_system ? That covers vmptrs and vmread.

That fell through the cracks when picking the bits from the original
patch. Fixed.

Thanks,

	tglx