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>
In-Reply-To: <CAHk-=wig6VizZHtRznz7uAWa-hHWjrCNANZ9B+1G=aTWPiVH4g@mail.gmail.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Fri, 26 May 2023 10:00:22 -0700
Message-ID: <CAHk-=whkci5ck5Him8Lx5ECKHEtj=bipYmOCGe8DWrrp8uDq5g@mail.gmail.com>
Subject: Re: x86 copy performance regression
To:     Eric Dumazet <edumazet@google.com>
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 9:56=E2=80=AFAM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> Ahh, I tested it in an allmodconfig build, but only building that one
> file, and not trying to link it. And gas was apparent;y perfectly
> happy just leaving that undefined feature as a relocation.

Oh, never mind. Even with that fixed, objtool is very unhappy about my
hack, because it knows about short jumps, and despite me encoding it
as a sequence of bytes, objtool will just decode it anyway and say
"that's not right".

Grr, I tried so hard to get the exact asm I wanted, but our
sanity-checking catches my cleverness and stops me in my tracks.

Let me go look at it some more. I *really* didn't want to make the
code worse for ERSM.

             Linus