[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <d3c7edcb-0499-182b-acda-c5f1f4d4e7b3@redhat.com>
Date: Wed, 30 Jan 2019 18:15:41 +0100
From: Paolo Bonzini <pbonzini@...hat.com>
To: Eric Northup <digitaleric@...gle.com>, jingqi.liu@...el.com
Cc: rkrcmar@...hat.com, Thomas Gleixner <tglx@...utronix.de>,
Ingo Molnar <mingo@...hat.com>, bp@...en8.de,
"H. Peter Anvin" <hpa@...or.com>,
the arch/x86 maintainers <x86@...nel.org>,
KVM <kvm@...r.kernel.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
tao3.xu@...el.com
Subject: Re: [PATCH v2 0/2] KVM: x86: expose direct stores instructions into
VM.
On 30/11/18 21:41, Eric Northup wrote:
> On Mon, Nov 5, 2018 at 10:01 PM Liu Jingqi <jingqi.liu@...el.com> wrote:
>>
>> Direct stores instructions MOVDIRI and MOVDIR64B will be available in
>> Tremont and other future x86 processors,
>> and need to be exposed to guest VM.
>
> It seems like KVM's emulator should be able to complete these
> instructions to emulated MMIO before exposing CPUID to guests in any
> default or supported configurations.
>
> It'll be much simpler for usermode to implement that property if the
> KVM-reported supported CPUID table doesn't get updated before the KVM
> emulator does.
We already do not support emulation for many CPUID bits we expose (AVX
and AVX512 above all), so that's not a huge problem.
I suppose these instructions are unlikely to be used on MMIO areas such
as legacy VGA VRAM.
Paolo
Powered by blists - more mailing lists