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
Content-Transfer-Encoding: quoted-printable
From:   Andy Lutomirski <luto@amacapital.net>
Mime-Version: 1.0 (1.0)
Subject: Re: Candidate Linux ABI for Intel AMX and hypothetical new related features
Date:   Mon, 29 Mar 2021 11:14:46 -0700
Message-Id: <0E03D727-3A22-4E29-AB6E-D6D5065615C7@amacapital.net>
References: <CAJvTdKkKhCFpaWm1hb8r3GHx10KBRQvpJNTtYPSAc6m28A7sQA@mail.gmail.com>
Cc:     Greg KH <gregkh@linuxfoundation.org>,
        Andy Lutomirski <luto@kernel.org>,
        "Bae, Chang Seok" <chang.seok.bae@intel.com>,
        Dave Hansen <dave.hansen@intel.com>, X86 ML <x86@kernel.org>,
        LKML <linux-kernel@vger.kernel.org>,
        libc-alpha <libc-alpha@sourceware.org>,
        Florian Weimer <fweimer@redhat.com>,
        Rich Felker <dalias@libc.org>, Kyle Huey <me@kylehuey.com>,
        Keno Fischer <keno@juliacomputing.com>,
        Linux API <linux-api@vger.kernel.org>
In-Reply-To: <CAJvTdKkKhCFpaWm1hb8r3GHx10KBRQvpJNTtYPSAc6m28A7sQA@mail.gmail.com>
To:     Len Brown <lenb@kernel.org>
Precedence: bulk


> On Mar 29, 2021, at 9:39 AM, Len Brown <lenb@kernel.org> wrote:
>=20
> =EF=BB=BF
>>=20
>> In particular, the library may use instructions that main() doesn't know e=
xist.
>=20
> And so I'll ask my question another way.
>=20
> How is it okay to change the value of XCR0 during the run time of a progra=
m?
>=20
> I submit that it is not, and that is a deal-killer for a request/release A=
PI.
>=20
> eg.  main() doesn't know that the math library wants to use AMX,
> and neither does the threading library.  So main() doesn't know to
> call the API before either library is invoked.  The threading library star=
ts up
> and creates user-space threads based on the initial value from XCR0.
> Then the math library calls the API, which adds bits to XCRO,
> and then the user-space context switch in the threading library corrupts d=
ata
> because the new XCR0 size doesn't match the initial size.
>=20

In the most extreme case, userspace could require that every loaded DSO be t=
agged with a new ELF note indicating support for dynamic XCR0 before changin=
g XCR0.

I would like to remind everyone that kernel enablement of AVX512 *already* b=
roke old userspace. AMX will further break something. At least with dynamic X=
CR0 we can make the breakage opt-in.

The ISA could have helped here by allowing the non-compacted XSTATE format t=
o be frozen even in the face of changing XCR0.  But it didn=E2=80=99t.  At t=
he end of the day, we are faced with the fact that XSTATE is a poor design, a=
nd we have to make the best of it.=