Received-SPF: pass (google.com: domain of linux-kernel+bounces-172555-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Date: Tue, 7 May 2024 17:21:02 -0700
In-Reply-To: <b4892d4cb7fea466fd82bcaf72ad3b29d28ce778.camel@intel.com>
Precedence: bulk
Mime-Version: 1.0
References: <20240404121327.3107131-1-pbonzini@redhat.com> <20240404121327.3107131-8-pbonzini@redhat.com>
 <43d1ade0461868016165e964e2bc97f280aee9d4.camel@intel.com>
 <ZhSYEVCHqSOpVKMh@google.com> <b4892d4cb7fea466fd82bcaf72ad3b29d28ce778.camel@intel.com>
Message-ID: <ZjrFblWaPNwXe0my@google.com>
Subject: Re: [PATCH v5 07/17] KVM: x86: add fields to struct kvm_arch for CoCo features
From: Sean Christopherson <seanjc@google.com>
To: Rick P Edgecombe <rick.p.edgecombe@intel.com>
Cc: "kvm@vger.kernel.org" <kvm@vger.kernel.org>, "pbonzini@redhat.com" <pbonzini@redhat.com>, 
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>, Yan Y Zhao <yan.y.zhao@intel.com>, 
	"michael.roth@amd.com" <michael.roth@amd.com>, Isaku Yamahata <isaku.yamahata@intel.com>
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable

On Tue, May 07, 2024, Rick P Edgecombe wrote:
> On Mon, 2024-04-08 at 18:21 -0700, Sean Christopherson wrote:
> > > - Give other name for this check like zap_from_leafs (or better name?=
)
> > > =C2=A0=C2=A0 The implementation is same to kvm_gfn_shared_mask() with=
 comment.
> > > =C2=A0=C2=A0 - Or we can add a boolean variable to struct kvm
> >=20
> > If we _don't_ hardcode the behavior, a per-memslot flag or a per-VM
> > capability (and thus boolean) is likely the way to go.=C2=A0 My off-the=
-cuff
> > vote is probably for a per-memslot flag.
>=20
> Hi Sean,
>=20
> Can you elaborate on the reason for a per-memslot flag? We are discussing=
 this
> design point internally, and also the intersection with the previous atte=
mpts to
> do something similar with a per-vm flag[0].
>=20
> I'm wondering if the intention is to try to make a memslot flag, so it ca=
n be
> expanded for the normal VM usage.

Sure, I'll go with that answer.  Like I said, off-the-cuff.

There's no concrete motiviation, it's more that _if_ we're going to expose =
a knob
to userspace, then I'd prefer to make it as precise as possible to minimize=
 the
changes of KVM ending up back in ABI hell again.

> Because the discussion on the original attempts, it seems safer to keep t=
his
> behavior more limited (TDX only) for now.  And for TDX's usage a struct k=
vm
> bool fits best because all memslots need to be set to zap_leafs_only =3D =
true,
> anyway.

No they don't.  They might be set that way in practice for QEMU, but it's n=
ot
strictly required.  E.g. nothing would prevent a VMM from exposing a shared=
-only
memslot to a guest.  The memslots that burned KVM the first time around wer=
e
related to VFIO devices, and I wouldn't put it past someone to be crazy eno=
ugh
to expose an passhtrough an untrusted device to a TDX guest.

> It's simpler for userspace, and less possible situations to worry about f=
or KVM.
>=20
> [0] https://lore.kernel.org/kvm/20200703025047.13987-1-sean.j.christopher=
son@intel.com/