From mboxrd@z Thu Jan  1 00:00:00 1970
From: Paolo Bonzini <pbonzini@redhat.com>
Subject: Re: [PATCH] kvm: x86: disable KVM_FAST_MMIO_BUS
Date: Wed, 16 Aug 2017 10:17:10 -0400 (EDT)
Message-ID: <1530586086.975916.1502893030374.JavaMail.zimbra@redhat.com>
References: <20170816112249.28939-1-pbonzini@redhat.com> <20170816120702.GC6408@flask> <49976600-3d63-5ac4-cc13-1465b0ecf9f5@redhat.com> <20170816170329-mutt-send-email-mst@kernel.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
Cc: Radim =?utf-8?B?S3LEjW3DocWZ?= <rkrcmar@redhat.com>,
        linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        stable@vger.kernel.org
To: "Michael S. Tsirkin" <mst@redhat.com>
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <20170816170329-mutt-send-email-mst@kernel.org>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: kvm.vger.kernel.org


> We actually know what to expect (a write) so we could maybe
> optimize this some more with a dedicated function just for this.

We don't know the addressing mode, the size or the source (immediate
vs. register), so no.

KVM is already doing a single translation and read no matter how long
the instruction (unless it cross page boundaries).

Paolo

> > 
> > However, the eventfd is written before decoding, while full emulation
> > would write it after. So while VCPU thread latency is worse compared to
> > skip_emulated_instruction, latency to the iothread remains small.
> > 
> > Paolo
>