Date: Thu, 27 Jul 2017 16:46:32 -0400
From: Tejun Heo <tj@kernel.org>
To: Michael Bringmann <mwb@linux.vnet.ibm.com>
Cc: Lai Jiangshan <jiangshanlai@gmail.com>, linux-kernel@vger.kernel.org,
        nfont@linux.vnet.ibm.com
Subject: Re: [PATCH v6] workqueue: Fix edge cases for calc of pool's cpumask
Message-ID: <20170727204632.GL742618@devbig577.frc2.facebook.com>
References: <5c421712-fb38-3020-d3fc-b9bdea792cd3@linux.vnet.ibm.com>
 <20170727183148.GH742618@devbig577.frc2.facebook.com>
 <ddf94c62-4cb2-d71b-b320-647b0c2975f0@linux.vnet.ibm.com>
 <20170727192410.GI742618@devbig577.frc2.facebook.com>
 <7c21c88d-6ce3-ae1a-d257-3042efa1bddf@linux.vnet.ibm.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <7c21c88d-6ce3-ae1a-d257-3042efa1bddf@linux.vnet.ibm.com>
User-Agent: Mutt/1.5.21 (2010-09-15)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1734
Lines: 35

Hello, Michael.

On Thu, Jul 27, 2017 at 03:15:47PM -0500, Michael Bringmann wrote:
> There is an underlying assumption in many layers / modules of the Linux
> system that CPU <-> node mapping is static.  This is despite the presence
> of features like NUMA and 'hotplug' that support the dynamic addition/
> removal of fundamental system resources like CPUs and memory.  PowerPC
> systems, however, do provide extensive features for the dynamic change
> of resources available to a system.

The text can go as-is but keeping cpu <-> node mapping static is a
trade-off rather than upper layers missing out something.  Making cpu
<-> node mapping dynamic means adding complexities and overhead to a
lot hotter paths including memory allocation.  It's just a better
trade off to keep the mapping static from arch side even if that means
we have to use more complex mapping in arch and/or allocate more
possible cpus than strictly necessary.

> Currently, there is little or no synchronization protection around the
> updating of the CPU <-> node mapping, and the export/update of this
> information for other layers / modules.  In systems which can change 
> this mapping during 'hotplug', like PowerPC, the information is changing 
> underneath all layers that might reference it.
> 
> This patch attempts to ensure that a valid, usable cpumask attribute is
> used by the workqueue infrastructure when setting up new resource pools.
> It prevents a crash that has been observed when an 'empty' cpumask is
> passed along to the worker/task scheduling code.  It is intended as an
> intermediate fix until a more fundamental review and correction of the
> issue can be done.

Thanks a lot for your patience!  Much appreciated.

-- 
tejun