[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20191030164743.GU4114@hirez.programming.kicks-ass.net>
Date: Wed, 30 Oct 2019 17:47:43 +0100
From: Peter Zijlstra <peterz@...radead.org>
To: Oleg Nesterov <oleg@...hat.com>
Cc: Will Deacon <will@...nel.org>, Ingo Molnar <mingo@...nel.org>,
Thomas Gleixner <tglx@...utronix.de>,
linux-kernel@...r.kernel.org, bigeasy@...utronix.de,
juri.lelli@...hat.com, williams@...hat.com, bristot@...hat.com,
longman@...hat.com, dave@...olabs.net, jack@...e.com
Subject: Re: [PATCH] locking/percpu_rwsem: Rewrite to not use rwsem
On Wed, Oct 30, 2019 at 04:57:20PM +0100, Oleg Nesterov wrote:
> On 10/29, Peter Zijlstra wrote:
> >
> > That said, I think cgroups use a variant of percpu-rwsem that wreck rss
> > on purpose and always take the slowpaths.
>
> I forgot (never understodd) why does Android need this.
>
> I am wondering if it makes any sense to add a config/boot or even runtime
> knob for cgroup_threadgroup_rwsem.
It isn't just Android, but Android in specific likes to move tasks
between cgroups a lot.
And moving tasks between cgroups is what requires the write-side of the
cgroup percpu-rwsem to be taken. Adding an RCU-GP to every task movement
wasn't making them happy (which I can understand).
Powered by blists - more mailing lists