[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20210614183801.GE68749@worktop.programming.kicks-ass.net>
Date: Mon, 14 Jun 2021 20:38:01 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Oleg Nesterov <oleg@...hat.com>
Cc: rjw@...ysocki.net, mingo@...nel.org, vincent.guittot@...aro.org,
dietmar.eggemann@....com, rostedt@...dmis.org, mgorman@...e.de,
Will Deacon <will@...nel.org>, Tejun Heo <tj@...nel.org>,
linux-kernel@...r.kernel.org, linux-pm@...r.kernel.org
Subject: Re: [PATCH] freezer,sched: Rewrite core freezer logic
On Mon, Jun 14, 2021 at 06:54:23PM +0200, Oleg Nesterov wrote:
> On 06/14, Peter Zijlstra wrote:
> > > I guess you do this to avoid freezable_schedule() in ptrace/signal_stop,
> > > and we can't use TASK_STOPPED|TASK_FREEZABLE, it should not run after
> > > thaw()... But see above, we can't rely on __frozen(parent).
> >
> > I do this because freezing puts a task in TASK_FROZEN, and that cannot
> > preserve TAKS_STOPPED or TASK_TRACED without being subject to wakups
>
> Yes, yes, this is what I tried to say.
OK, thanks for all that. Clearly I need to stare at this code longer and
harder.
One more thing; if I add additional state bits to preserve
__TASK_{TRACED,STOPPED}, then I need to figure out at thaw time if we've
missed a wakeup or not.
Do we have sufficient state for that? If so, don't we then also not have
sufficient state to tell if a task should've been TRACED/STOPPED in the
first place?
If not, I probably should add this... I'll go dig.
Powered by blists - more mailing lists