Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Content-Type: text/plain;
        charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 13.4 \(3608.80.23.2.2\))
Subject: Re: [RFC PATCH 1/3] kvm: x86: Rename KVM_DEBUGREG_RELOAD to
 KVM_DEBUGREG_NEED_RELOAD
From:   Nadav Amit <nadav.amit@gmail.com>
In-Reply-To: <85cb5946-2109-28a0-578d-bed31d1b8298@redhat.com>
Date:   Sat, 25 Apr 2020 09:54:07 -0700
Cc:     Xiaoyao Li <xiaoyao.li@intel.com>,
        Sean Christopherson <sean.j.christopherson@intel.com>,
        kvm <kvm@vger.kernel.org>, Nadav Amit <namit@cs.technion.ac.il>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        linux-kernel@vger.kernel.org
Content-Transfer-Encoding: quoted-printable
Message-Id: <08C6D1FB-A4F7-49FA-AC46-5323C104840A@gmail.com>
References: <20200416101509.73526-1-xiaoyao.li@intel.com>
 <20200416101509.73526-2-xiaoyao.li@intel.com>
 <85cb5946-2109-28a0-578d-bed31d1b8298@redhat.com>
To:     Paolo Bonzini <pbonzini@redhat.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

> On Apr 25, 2020, at 1:07 AM, Paolo Bonzini <pbonzini@redhat.com> =
wrote:
>=20
> On 16/04/20 12:15, Xiaoyao Li wrote:
>> To make it more clear that the flag means DRn (except DR7) need to be
>> reloaded before vm entry.
>>=20
>> Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
>=20
> I wonder if KVM_DEBUGREG_RELOAD is needed at all.  It should be easy =
to
> write selftests for it, using the testcase in commit message
> 172b2386ed16 and the information in commit ae561edeb421.

I must be missing something, since I did not follow this thread and =
other
KVM changes very closely.

Yet, for the record, I added KVM_DEBUGREG_RELOAD due to real experienced
issues that I had while running Intel=E2=80=99s fuzzing tests on KVM: =
IIRC, the DRs
were not reloaded after an INIT event that clears them.

Personally, I would prefer that a test for that, if added, would be =
added
to KVM-unit-tests, based on Liran=E2=80=99s INIT test. This would allow =
to confirm
bare-metal behaves as the VM.