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
Mime-Version: 1.0 (1.0)
Subject: Re: [PATCH] x86: Call fixup_exception() before notify_die() in math_error()
From:   Andy Lutomirski <luto@amacapital.net>
In-Reply-To: <alpine.DEB.2.21.1806190820210.1692@nanos.tec.linutronix.de>
Date:   Tue, 19 Jun 2018 08:23:03 -0700
Cc:     Andy Lutomirski <luto@kernel.org>, Siarhei.Liakh@concurrent-rt.com,
        LKML <linux-kernel@vger.kernel.org>, X86 ML <x86@kernel.org>,
        Ingo Molnar <mingo@redhat.com>,
        "H. Peter Anvin" <hpa@zytor.com>, Borislav Petkov <bpetkov@suse.de>
Content-Transfer-Encoding: quoted-printable
Message-Id: <048143ED-D567-4B6B-ABF4-EFE5F1B87607@amacapital.net>
References: <DM5PR11MB201156F1CAB2592B07C79A03B17D0@DM5PR11MB2011.namprd11.prod.outlook.com> <CALCETrXUdLZTnf0EF2LVM4ru2G6nkFX--gQ=6JpBRNZ7Gm+MsQ@mail.gmail.com> <alpine.DEB.2.21.1806190820210.1692@nanos.tec.linutronix.de>
To:     Thomas Gleixner <tglx@linutronix.de>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


> On Jun 18, 2018, at 11:23 PM, Thomas Gleixner <tglx@linutronix.de> wrote:
>=20
>> On Mon, 18 Jun 2018, Andy Lutomirski wrote:
>>=20
>> On Thu, Jun 14, 2018 at 10:10 PM Siarhei Liakh
>> <Siarhei.Liakh@concurrent-rt.com> wrote:
>>>=20
>>> fpu__drop() has an explicit fwait which under some conditions can trigge=
r
>>> a fixable FPU exception while in kernel. Thus, we should attempt to fixu=
p
>>> the exception first, and only call notify_die() if the fixup failed just=

>>> like in do_general_protection(). The original call sequence incorrectly
>>> triggers KDB entry on debug kernels under particular FPU-intensive
>>> workloads. This issue had been privately observed, fixed, and tested
>>> on 4.9.98, while this patch brings the fix to the upstream.
>>=20
>> Reviewed-by: Andy Lutomirski <luto@kernel.org>
>>=20
>> With the caveat that you are perpetuating what is arguably a bug in
>> some of the other entries: math_error() can now be called with IRQs
>> off and return with IRQs on.  If we actually start asserting good
>> behavior in the entry code, we'll need to fix this.
>=20
> Confused. math_error() is still invoked with interrupts off. What's
> different now is that notify_die() is called with interrupts conditionally=

> enabled while upstream it's always called with interrupts disabled.

True, but I don=E2=80=99t think that matters. What I=E2=80=99m grumbling abo=
ut is that we can do cond_local_irq_enable() and then return without local_i=
rq_disable().

Anyway, I think the patch is fine as is. We can unsuck the entry IRQ handlin=
g another day.

>=20
> Thanks,
>=20
>    tglx
>=20
>=20
>=20