Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:3 as permitted sender) client-ip=2620:137:e000::3:3;
MIME-Version: 1.0
References: <202310172247.b9959bd4-oliver.sang@intel.com> <2023101759-front-ember-6354@gregkh>
In-Reply-To: <2023101759-front-ember-6354@gregkh>
From:   Max Kellermann <max.kellermann@ionos.com>
Date:   Wed, 18 Oct 2023 08:31:51 +0200
Message-ID: <CAKPOu+_tQj361LAGNyw+wLnrw=Y709zcA5kjX_95c_4HOv6wxg@mail.gmail.com>
Subject: Re: [linux-next:master] [drivers/char/mem] 1b057bd800:
 stress-ng.splice.ops_per_sec -99.8% regression
To:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc:     kernel test robot <oliver.sang@intel.com>, oe-lkp@lists.linux.dev,
        lkp@intel.com, Linux Memory Management List <linux-mm@kvack.org>,
        linux-kernel@vger.kernel.org, ying.huang@intel.com,
        feng.tang@intel.com, fengwei.yin@intel.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Tue, Oct 17, 2023 at 6:57=E2=80=AFPM Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
> So everything now goes faster, right?  -99.8% regression means 99.8%
> faster?

That's what I thought, too, and sounds reasonable considering this
test is described as "stress copying of /dev/zero to /dev/null",
but... it's not what that test actually does. Contrary to the
description, it doesn't use /dev/zero at all, neither does it use
/dev/full. So it shouldn't be affected by my patch at all.

strace of that test's setup:

 pipe2([4, 5], 0)                =3D 0
 openat(AT_FDCWD, "/dev/null", O_WRONLY) =3D 6

Then it loops:

 vmsplice(5, [{iov_base=3D"\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0=
\0\0\0\0\0\0\0\0"...,
iov_len=3D65536}], 1, 0) =3D 65536
 splice(4, NULL, 6, NULL, 65536, SPLICE_F_MOVE) =3D 65536
 write(5, "\334\360U\300~\361\20jV\367\263,\221\3724\332>7\31H2|\20\254\314=
\212y\275\334I\304\207"...,
4096) =3D 4096
 vmsplice(4, [{iov_base=3D"\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0=
\0\0\0\0\0\0\0\0"...,
iov_len=3D4096}], 1, 0) =3D 4096

I don't get it.