Message-ID: <1475526171.4622.9.camel@redhat.com>
Subject: Re: [PATCH RFC 4/5] x86,fpu: lazily skip FPU restore when still
 loaded
From: Rik van Riel <riel@redhat.com>
To: Dave Hansen <dave.hansen@linux.intel.com>,
        linux-kernel@vger.kernel.org
Cc: x86@kernel.org, tglx@linutronix.de, pbonzini@redhat.com,
        mingo@redhat.com, luto@kernel.org, hpa@zytor.com, bp@suse.de
Date: Mon, 03 Oct 2016 16:22:51 -0400
In-Reply-To: <57F2B9E9.7030806@linux.intel.com>
References: <1475353895-22175-1-git-send-email-riel@redhat.com>
         <1475353895-22175-5-git-send-email-riel@redhat.com>
         <57F2B9E9.7030806@linux.intel.com>
Content-Type: multipart/signed; micalg="pgp-sha256";
        protocol="application/pgp-signature"; boundary="=-seNJk+RhPJMPoBKBFRFZ"
Mime-Version: 1.0
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2059
Lines: 62


--=-seNJk+RhPJMPoBKBFRFZ
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Mon, 2016-10-03 at 13:04 -0700, Dave Hansen wrote:
> On 10/01/2016 01:31 PM, riel@redhat.com wrote:
> >=20
> > =C2=A0/*
> > + * Check whether an FPU's register set is still loaded in the CPU.
> > + */
> > +static inline bool fpu_lazy_skip_restore(struct fpu *fpu)
> > +{
> > +	bool still_loaded =3D (fpu->fpstate_active &&
> > +			=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0fpu->last_cpu =3D=3D
> > raw_smp_processor_id() &&
> > +			=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0__this_cpu_read(fpu_fpregs_owner_ctx)
> > =3D=3D fpu);
> > +
> > +	fpu->fpregs_active =3D still_loaded;
> > +	return still_loaded;
> > +}
> I wonder if we should call this something more along the lines of
> fpregs_activate_fast(), which returns if it managed to do the
> activation
> fast or not.=C2=A0=C2=A0I _think_ that's more along the lines of what it =
is
> actually doing.=C2=A0=C2=A0The fact that it can be lazy is really an
> implementation detail.
>=20
> What are the preempt rules with this thing?=C2=A0=C2=A0This needs to be c=
alled
> in
> preempt-disabled contexts, right?

Indeed, all the FPU context switching code needs=C2=A0
to be called in preempt-disabled contexts.

You do not want to get preempted halfway through
saving or restoring floating point registers.

--=20
All rights reversed

--=-seNJk+RhPJMPoBKBFRFZ
Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: This is a digitally signed message part
Content-Transfer-Encoding: 7bit

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2

iQEcBAABCAAGBQJX8r4bAAoJEM553pKExN6DooUH/0/pL6asT/8qFt+RcH0PZtt5
GvMoLoMd8NkVbNJBqqcmW0Rg3HhlzdnAAvO9BdDffiFyANOdWjNJIxr/Jd2XTKdv
cTRRK0brLJ6yEacWWr+avBvrAZ0QxGapZr47Bgve6NV/avyUgphvfNIrR5aWizE6
8NYvCS4vH9muejejZE7OPJa1ScXPYjwr/cKRXk/FJYEL6P8445YsgsU+xgF9Jl3T
E9uSfq3LfrXMz1n/eomjfbjtqwAoiHGRtpH6u7Nk95U8CsQWIRG1QQxxHXvjdcDE
sib/l3aNt51YYAJCicXuzxb88PlHgD8hxVxbfQm66xITX4HhJla9R5PTeLQ8/xQ=
=ONSB
-----END PGP SIGNATURE-----

--=-seNJk+RhPJMPoBKBFRFZ--