Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <CANn89iKUbyrJ=r2+_kK+sb2ZSSHifFZ7QkPLDpAtkJ8v4WUumA@mail.gmail.com>
 <CAHk-=whqNMUPbjCyMjyxfH_5-Xass=DrMkPT5ZTJbFrtU=qDEQ@mail.gmail.com>
 <CANn89i+bExb_P6A9ROmwqNgGdO5o8wawVZ5r3MHnz0qfhxvTtA@mail.gmail.com>
 <CAHk-=wig6VizZHtRznz7uAWa-hHWjrCNANZ9B+1G=aTWPiVH4g@mail.gmail.com>
 <CAHk-=whkci5ck5Him8Lx5ECKHEtj=bipYmOCGe8DWrrp8uDq5g@mail.gmail.com>
 <CAHk-=whtDupvWtj_ow11wU4_u=KvifTqno=5mW1VofyehjdVRA@mail.gmail.com>
 <CANn89i+u8jvfSQAQ=_JY0be56deJNhKgDWbqpDAvfm-i34qX9A@mail.gmail.com> <CAHk-=wh16fVwO2yZ4Fx0kyRHsNDhGddzNxfQQz2+x08=CPvk_Q@mail.gmail.com>
In-Reply-To: <CAHk-=wh16fVwO2yZ4Fx0kyRHsNDhGddzNxfQQz2+x08=CPvk_Q@mail.gmail.com>
From:   Eric Dumazet <edumazet@google.com>
Date:   Fri, 26 May 2023 19:51:22 +0200
Message-ID: <CANn89iJ3=OiZEABRQQLL6z+J-Wy8AvTJz6NPLQDOtzREiiYb4Q@mail.gmail.com>
Subject: Re: x86 copy performance regression
To:     Linus Torvalds <torvalds@linux-foundation.org>
Cc:     LKML <linux-kernel@vger.kernel.org>,
        netdev <netdev@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Fri, May 26, 2023 at 7:40=E2=80=AFPM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> On Fri, May 26, 2023 at 10:25=E2=80=AFAM Eric Dumazet <edumazet@google.co=
m> wrote:
> >
> > arch/x86/lib/copy_user_64.S:34:2: error: invalid instruction mnemonic
> > 'alternative'
>
> Ok, that's just odd. For me, assembler mnemonics - very much including
> macros - are case-insensitive.
>
> It's actually documented that way, with the example given is for a
> macro that is declared as "sum" and then used as "SUM":
>
>    https://sourceware.org/binutils/docs/as/Macro.html
>
> And if you want to use macros as pseudo-instructions, that's what you
> want, since typically assembler instructions are not case sensitive.
>
> But yeah, your build environment is clearly different, and yes, we
> declare the macro with all caps, and other places use it that way too.
>
> Clang?

Yes, we use clang here ...

Hmmm

[   25.532236] RIP: 0010:0xffffffffa5a85134
[   25.536173] Code: Unable to access opcode bytes at 0xffffffffa5a8510a.
[   25.542720] RSP: 0000:ffff92f08159bcd8 EFLAGS: 00050206
[   25.547960] RAX: 00007ffc3b16c318 RBX: 0000000000000000 RCX: 00000000000=
00170
[   25.555118] RDX: 0000000000000170 RSI: ffff92f0944d4c28 RDI: 00007ffc3b1=
6c1a8
[   25.562275] RBP: ffff92f08159bce0 R08: fefefefefefefeff R09: 00000000000=
0002c
[   25.569432] R10: 000000000000002c R11: ffff92f0944d5bb0 R12: 00007ffc3b1=
6cff2
[   25.576588] R13: 00007ffc3b16c1a8 R14: 0000000000000001 R15: ffff92f0944=
d4ac0
[   25.583746] FS:  0000000000000000(0000) GS:ffff934e404c0000(0000)
knlGS:0000000000000000
[   25.591862] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   25.597624] CR2: ffffffffa5a8510a CR3: 000000010e33c003 CR4: 00000000003=
706e0
[   25.604780] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 00000000000=
00000
[   25.611936] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 00000000000=
00400
[   25.619092] Call Trace:
[   25.621545]  <TASK>
[   25.623648]  ? _copy_to_user+0x20/0x30
[   25.627409]  create_elf_tables+0x528/0x5a0
[   25.631520]  load_elf_binary+0x9e7/0xce0
[   25.635453]  bprm_execve+0x2bf/0x5c0
[   25.639040]  kernel_execve+0x2ad/0x2d0
[   25.642799]  run_init_process+0xa9/0xb0
[   25.646648]  ? rest_init+0xc0/0xc0
[   25.650059]  kernel_init+0x82/0x1a0
[   25.653558]  ret_from
_fork+0x1f/0x30
[   25.657145]  </TASK>