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;
MIME-Version: 1.0
References: <000000000000886dbd05b7ffa8db@google.com> <20210104124153.0992b1f7fd1a145e193a333f@linux-foundation.org>
 <CAHk-=wi6hd8ATJ1W90goTxjgyvuoFsf0xZdAJmZ2c0dx5wcJSg@mail.gmail.com>
 <alpine.LSU.2.11.2101041839440.3466@eggly.anvils> <CAHk-=wi36CBggdRfdggACvf2hG+djM9kKnorrwsByN6uDvPExA@mail.gmail.com>
 <CAHk-=wh=5kDGukMs2sVZ8uHZJX4VL13oD5+xMAR4HvuY6QckLg@mail.gmail.com>
 <CAHk-=wgD9GK5CeHopYmRHoYS9cNuCmDMsc=+MbM_KgJ0KB+=ng@mail.gmail.com> <20210112104425.GA8760@quack2.suse.cz>
In-Reply-To: <20210112104425.GA8760@quack2.suse.cz>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Tue, 12 Jan 2021 09:35:34 -0800
Message-ID: <CAHk-=whNKdNaBn2SuWa5=gjLNC3iAcG4+Si_TPBF7R8AbkC8tQ@mail.gmail.com>
Subject: Re: kernel BUG at mm/page-writeback.c:LINE!
To:     Jan Kara <jack@suse.cz>
Cc:     Hugh Dickins <hughd@google.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Matthew Wilcox <willy@infradead.org>,
        syzbot <syzbot+2fc0712f8f8b8b8fa0ef@syzkaller.appspotmail.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Linux-MM <linux-mm@kvack.org>,
        syzkaller-bugs <syzkaller-bugs@googlegroups.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Jan 12, 2021 at 2:44 AM Jan Kara <jack@suse.cz> wrote:
>
> On Fri 08-01-21 18:04:21, Linus Torvalds wrote:
> >
> > Oh, and Michael Larabel (of phoronix) reports that that one-liner does
> > something bad to a few PostgreSQL tests, on the order of 5-10%
> > regression on some machines (but apparently not others).
>
> Do you have more details? From my experience (we do regular pgbench runs
> for various kernels in various configs in SUSE) PostgreSQL numbers tend to
> be somewhat noisy and more dependent on CPU scheduling and NUMA locality
> than anything else. But it very much depends on the exact config passed to
> pgbench so that's why I'm asking...

No, I don't really have many more details. I don't have things like
raw numbers or exact configurations, but Michael has been very
responsive if you ask, so if you are interested, just email him at
Michael Larabel <Michael@michaellarabel.com>.

It wasn't NUMA - apparently the machines he saw this on were just
plain consumer setups, and his larger machines didn't actually show
the effect. But yes, I suspect it was some scheduling artifact, and
probably just fairly random noise from just changing the scheduling
pattern a bit.

            Linus