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;
Date:   Fri, 23 Oct 2020 16:39:46 -0700
User-Agent: K-9 Mail for Android
In-Reply-To: <bca28d6e33a3475193478e762214c6ea@AcuMS.aculab.com>
References: <20201023203154.27335-1-linux@rasmusvillemoes.dk> <CAHk-=wj1m3cvS-3dOYzNavYWLFu=9fwo0-6HTHJhG-X5B73gZg@mail.gmail.com> <8820745F-E761-42E6-8A70-7B04EE70692C@zytor.com> <CAHk-=wg9L3EZk=cBjt5R3LkE8Y6swwOZ8sxhpQYcJO3Fj1wLbQ@mail.gmail.com> <bca28d6e33a3475193478e762214c6ea@AcuMS.aculab.com>
MIME-Version: 1.0
Content-Type: text/plain;
 charset=utf-8
Content-Transfer-Encoding: quoted-printable
Subject: RE: [PATCH] x86/uaccess: fix code generation in put_user()
To:     David Laight <David.Laight@ACULAB.COM>,
        "'Linus Torvalds'" <torvalds@linux-foundation.org>
CC:     Rasmus Villemoes <linux@rasmusvillemoes.dk>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        the arch/x86 maintainers <x86@kernel.org>,
        Sean Christopherson <sean.j.christopherson@intel.com>,
        Naresh Kamboju <naresh.kamboju@linaro.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
From:   hpa@zytor.com
Message-ID: <3F1D3946-E7B8-4F77-9189-D49E1ECE0B0D@zytor.com>
Precedence: bulk

On October 23, 2020 2:52:16 PM PDT, David Laight <David=2ELaight@ACULAB=2EC=
OM> wrote:
>From: Linus Torvalds
>> Sent: 23 October 2020 22:11
>>=20
>> On Fri, Oct 23, 2020 at 2:00 PM <hpa@zytor=2Ecom> wrote:
>> >
>> > There is no same reason to mess around with hacks when we are
>talking about dx:ax, though=2E
>>=20
>> Sure there is=2E
>>=20
>> "A" doesn't actually mean %edx:%eax like you seem to think=2E
>>=20
>> It actually means %eax OR %edx, and then if given a 64-bit value, it
>> will use the combination (with %edx being the high bits)=2E
>>=20
>> So using "A" unconditionally doesn't work - it gives random behavior
>> for 32-bit (or smaller) types=2E
>>=20
>> Or you'd have to cast the value to always be 64-bit, and have the
>> extra code generation=2E
>>=20
>> IOW, an unconditional "A" is wrong=2E
>>=20
>> And the alternative is to just duplicate things, and go back to the
>> explicit size testing, but honestly, I really think that's much worse
>> than relying on a documented feature of "register asm()" that gcc
>> _documents_ is for this kind of inline asm use=2E
>
>Could do_put_user() do an initial check for 64 bit
>then expand a different #define that contains the actual
>code passing either "a" or "A" for the constriant=2E
>
>Apart from another level of indirection nothing is duplicated=2E
>
>	David
>
>-
>Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes,
>MK1 1PT, UK
>Registration No: 1397386 (Wales)

Maybe #define ASM_AX64 or some such=2E
--=20
Sent from my Android device with K-9 Mail=2E Please excuse my brevity=2E