Re: [RFC 15/16] sched/fair: Account kthread runtime debt for CFS bandwidth

From: Tejun Heo <tj@kernel.org>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>,
	Alexander Duyck <alexanderduyck@fb.com>,
	Alex Williamson <alex.williamson@redhat.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	Ben Segall <bsegall@google.com>,
	Cornelia Huck <cohuck@redhat.com>,
	Dan Williams <dan.j.williams@intel.com>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	Dietmar Eggemann <dietmar.eggemann@arm.com>,
	Herbert Xu <herbert@gondor.apana.org.au>,
	Ingo Molnar <mingo@redhat.com>, Jason Gunthorpe <jgg@nvidia.com>,
	Johannes Weiner <hannes@cmpxchg.org>,
	Josh Triplett <josh@joshtriplett.org>,
	Michal Hocko <mhocko@suse.com>, Nico Pache <npache@redhat.com>,
	Pasha Tatashin <pasha.tatashin@soleen.com>,
	Steffen Klassert <steffen.klassert@secunet.com>,
	Steve Sistare <steven.sistare@oracle.com>,
	Tim Chen <tim.c.chen@linux.intel.com>,
	Vincent Guittot <vincent.guittot@linaro.org>,
	linux-mm@kvack.org, kvm@vger.kernel.org,
	linux-kernel@vger.kernel.org, linux-crypto@vger.kernel.org
Subject: Re: [RFC 15/16] sched/fair: Account kthread runtime debt for CFS bandwidth
Date: Fri, 14 Jan 2022 06:30:34 -1000	[thread overview]
Message-ID: <YeGlKpumqRxT9L7A@slm.duckdns.org> (raw)
In-Reply-To: <YeFDC0mV3yurUFbl@hirez.programming.kicks-ass.net>

Hello,

On Fri, Jan 14, 2022 at 10:31:55AM +0100, Peter Zijlstra wrote:
> So part of the problem I have with this is that these external things
> can consume all the bandwidth and basically indefinitely starve the
> group.
> 
> This is doulby so if you're going to account things like softirq network
> processing.

So, anything which is accounted this way should slow down / stall the
originator so that backcharges can't run away. In many cases, these
connections are already there - e.g. if you keep charging socket rx buffers
to a cgroup, the processes in the cgroup will slow down and the backpressure
on the network socket will slow down the incoming packets. Sometimes, the
backpressure propagation needs to be added explicitly - e.g. filesystem
metadata writes can run away because something can keeping on issuing
metadata updates without getting throttled neither on memory or io side. So,
for situations like that, we need to add an explicit mechanism to throttle
the originator asynchronously.

Thanks.

-- 
tejun