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
Content-Transfer-Encoding: quoted-printable
From:   Andy Lutomirski <luto@amacapital.net>
Mime-Version: 1.0 (1.0)
Subject: Re: [PATCH 2/2] KVM: VMX: Extend VMX's #AC handding
Date:   Sat, 1 Feb 2020 09:56:11 -0800
Message-Id: <A2622E15-756D-434D-AF64-4F67781C0A74@amacapital.net>
References: <b2e2310d-2228-45c2-8174-048e18a46bb6@intel.com>
Cc:     Sean Christopherson <sean.j.christopherson@intel.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        "H. Peter Anvin" <hpa@zytor.com>,
        Paolo Bonzini <pbonzini@redhat.com>, x86@kernel.org,
        linux-kernel@vger.kernel.org, kvm@vger.kernel.org
In-Reply-To: <b2e2310d-2228-45c2-8174-048e18a46bb6@intel.com>
To:     Xiaoyao Li <xiaoyao.li@intel.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


> On Feb 1, 2020, at 8:58 AM, Xiaoyao Li <xiaoyao.li@intel.com> wrote:
>=20
> =EF=BB=BFOn 2/1/2020 5:33 AM, Andy Lutomirski wrote:
>>>> On Jan 31, 2020, at 1:04 PM, Sean Christopherson <sean.j.christopherson=
@intel.com> wrote:
>>>=20
>>> =EF=BB=BFOn Fri, Jan 31, 2020 at 12:57:51PM -0800, Andy Lutomirski wrote=
:
>>>>=20
>>>>>> On Jan 31, 2020, at 12:18 PM, Sean Christopherson <sean.j.christopher=
son@intel.com> wrote:
>>>>>=20
>>>>> This is essentially what I proposed a while back.  KVM would allow ena=
bling
>>>>> split-lock #AC in the guest if and only if SMT is disabled or the enab=
le bit
>>>>> is per-thread, *or* the host is in "warn" mode (can live with split-lo=
ck #AC
>>>>> being randomly disabled/enabled) and userspace has communicated to KVM=
 that
>>>>> it is pinning vCPUs.
>>>>=20
>>>> How about covering the actual sensible case: host is set to fatal?  In t=
his
>>>> mode, the guest gets split lock detection whether it wants it or not. H=
ow do
>>>> we communicate this to the guest?
>>>=20
>>> KVM doesn't advertise split-lock #AC to the guest and returns -EFAULT to=
 the
>>> userspace VMM if the guest triggers a split-lock #AC.
>>>=20
>>> Effectively the same behavior as any other userspace process, just that K=
VM
>>> explicitly returns -EFAULT instead of the process getting a SIGBUS.
>> Which helps how if the guest is actually SLD-aware?
>> I suppose we could make the argument that, if an SLD-aware guest gets #AC=
 at CPL0, it=E2=80=99s a bug, but it still seems rather nicer to forward the=
 #AC to the guest instead of summarily killing it.
>=20
> If KVM does advertise split-lock detection to the guest, then kvm/host can=
 know whether a guest is SLD-aware by checking guest's MSR_TEST_CTRL.SPLIT_L=
OCK_DETECT bit.
>=20
> - If guest's MSR_TEST_CTRL.SPLIT_LOCK_DETECT is set, it indicates guest is=
 SLD-aware so KVM forwards #AC to guest.
>=20

I disagree. If you advertise split-lock detection with the current core capa=
bility bit, it should *work*.  And it won=E2=80=99t.  The choices you=E2=80=99=
re actually giving the guest are:

a) Guest understands SLD and wants it on.  The guest gets the same behavior a=
s in bare metal.

b) Guest does not understand. Guest gets killed if it screws up as described=
 below.

> - If not set. It may be a old guest or a malicious guest or a guest withou=
t SLD support, and we cannot figure it out. So we have to kill the guest whe=
n host is SLD-fatal, and let guest survive when SLD-WARN for old sane buggy g=
uest.

All true, but the result of running a Linux guest in SLD-warn mode will be b=
roken.

>=20
> In a word, all the above is on the condition that KVM advertise split-lock=
 detection to guest. But this patch doesn't do this. Maybe I should add that=
 part in v2.

I think you should think the details all the way through, and I think you=E2=
=80=99re likely to determine that the Intel architecture team needs to do *s=
omething* to clean up this mess.

There are two independent problems here.  First, SLD *can=E2=80=99t* be virt=
ualized sanely because it=E2=80=99s per-core not per-thread. Second, most us=
ers *won=E2=80=99t want* to virtualize it correctly even if they could: if a=
 guest is allowed to do split locks, it can DoS the system.

So I think there should be an architectural way to tell a guest that SLD is o=
n whether it likes it or not. And the guest, if booted with sld=3Dwarn, can p=
rint a message saying =E2=80=9Chaha, actually SLD is fatal=E2=80=9D and carr=
y on.

>=20
>> ISTM, on an SLD-fatal host with an SLD-aware guest, the host should tell t=
he guest =E2=80=9Chey, you may not do split locks =E2=80=94 SLD is forced on=
=E2=80=9D and the guest should somehow acknowledge it so that it sees the ar=
chitectural behavior instead of something we made up.  Hence my suggestion.
>=20