[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20220804095150.GG3493@suse.de>
Date: Thu, 4 Aug 2022 10:51:50 +0100
From: Mel Gorman <mgorman@...e.de>
To: Ingo Molnar <mingo@...nel.org>
Cc: Libo Chen <libo.chen@...cle.com>,
K Prateek Nayak <kprateek.nayak@....com>,
peterz@...radead.org, vincent.guittot@...aro.org,
tim.c.chen@...ux.intel.com, 21cnbao@...il.com,
dietmar.eggemann@....com, linux-kernel@...r.kernel.org,
tglx@...utronix.de
Subject: Re: [PATCH] sched/fair: no sync wakeup from interrupt context
On Thu, Aug 04, 2022 at 10:55:12AM +0200, Ingo Molnar wrote:
>
> * Libo Chen <libo.chen@...cle.com> wrote:
>
> > Oh yes, I have no issue with holding the patch back until the regression
> > is fully understood. I was just a little confused on your reference to
> > Mel's comments. [...]
>
> Yeah, that was just me getting confused about which change Mel was
> referring to, as I was looking for external confirmation saying what I was
> thinking about the patch: in_task()/in_interrupt() heuristics rarely do
> well. ;-)
>
Even though I was referring to something else, the reported regression
is still a regression. Prateek's report and what I brought up in
https://lore.kernel.org/all/20220715100738.GD3493@suse.de/ are both simply
examples where changes to affine wakeup can have surprising results.
--
Mel Gorman
SUSE Labs
Powered by blists - more mailing lists