[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <dd39029d93bb4de5ed485b5d4181fc19d4c0c4f0.camel@gmx.de>
Date: Sun, 28 Nov 2021 13:35:06 +0100
From: Mike Galbraith <efault@....de>
To: Alexey Avramov <hakavlad@...ox.lv>,
Mel Gorman <mgorman@...hsingularity.net>
Cc: Andrew Morton <akpm@...ux-foundation.org>,
Michal Hocko <mhocko@...e.com>,
Vlastimil Babka <vbabka@...e.cz>,
Rik van Riel <riel@...riel.com>,
Darrick Wong <djwong@...nel.org>, regressions@...ts.linux.dev,
Linux-fsdevel <linux-fsdevel@...r.kernel.org>,
Linux-MM <linux-mm@...ck.org>,
LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 1/1] mm: vmscan: Reduce throttling due to a failure to
make progress
On Sun, 2021-11-28 at 12:39 +0100, Mike Galbraith wrote:
> On Sun, 2021-11-28 at 11:00 +0100, Mike Galbraith wrote:
> > On Sun, 2021-11-28 at 04:26 +0900, Alexey Avramov wrote:
> > > I will present the results of the new tests here.
> > >
> > > TLDR;
> > > =====
> > > No one Mel's patch doesn't prevent stalls in my tests.
> >
> > Seems there may be a problem with the THROTTLE_WRITEBACK bits..
>
> Disregard. I backed hacklet out again, and one-liner that had ground
> my box to fine powder was harmless. Hmm.
Grr, no, disregard the disregard, it's just annoyingly variable, may
perform akin to 5.15 but most likely won't. With bandaid, it _seems_
to be consistently on par with 5.15.
Bandaid likely just break the writeback bits all to pieces, but that's
ok, if breakage makes things work better, they need more breaking :)
-Mike
Powered by blists - more mailing lists