Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Content-Type: text/plain; charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 10.3 \(3273\))
Subject: Re: [PATCH 2/2] KVM: VMX: Add Force Emulation Prefix for "emulate the
 next instruction"
From:   Nikita Leshenko <nikita.leshchenko@oracle.com>
In-Reply-To: <d34c1d3c-324e-4645-9d48-09c94212aa5e@default>
Date:   Tue, 27 Mar 2018 12:05:25 +0300
Cc:     pbonzini@redhat.com, kernellwp@gmail.com, rkrcmar@redhat.com,
        andrew.cooper3@citrix.com, linux-kernel@vger.kernel.org,
        kvm@vger.kernel.org
Content-Transfer-Encoding: quoted-printable
Message-Id: <95B06D42-EFDF-4750-AD03-98D6C78378D5@oracle.com>
References: <d34c1d3c-324e-4645-9d48-09c94212aa5e@default>
To:     Liran Alon <liran.alon@oracle.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

What you are essentially trying to do is create a PV interface to access
the x86 emulator.
Why not use a simple hypercall (VMCALL) to accomplish this instead of
inventing yet another PV method?

Something like =E2=80=9CKVM_HC_EMULATE_NEXT_INSTRUCTION=E2=80=9D in =
kvm_emulate_hypercall
should do the trick (however it needs to be placed before the check for
CPL>0 so that user mode code can test the emulator too).

Nikita

> On 27 Mar 2018, at 11:26, Liran Alon <liran.alon@oracle.com> wrote:
>=20
>=20
> ----- pbonzini@redhat.com wrote:
>=20
>> On 27/03/2018 09:52, Liran Alon wrote:
>>> In addition, I think this module parameter should be in kvm module
>>> (not kvm_intel) and you should add similar logic to kvm_amd module
>> (SVM)
>>=20
>> If you can move handle_ud to x86.c, then it makes sense to have the
>> module parameter in the kvm module.  I haven't checked.
>=20
> I don't see a reason why you couldn't do that.
>=20
>>=20
>> Otherwise you would have to EXPORT_SYMBOL_GPL the variable; in the
>=20
> This is what I did for enable_vmware_backdoor module parameter.
> I think this is what should be done in this case as-well.
>=20
>> end
>> it's just a debugging tool, so it'd be simpler to just add it
>> separately
>> to kvm_intel and kvm_amd.
>=20
> I agree it's just a debugging tool. But no reason for it to be used =
differently
> when running tests on Intel CPU vs. AMD CPU.
> I think the effort to fix this is low.
>=20
>>=20
>> Paolo
>=20
>=20