From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jens Axboe Subject: Re: [RFC PATCH 1/2] block: strict rq_affinity Date: Sat, 23 Jul 2011 20:38:20 +0200 Message-ID: <4E2B151C.6080900@fusionio.com> References: <20110722205736.17420.41366.stgit@localhost6.localdomain6> <20110722205938.17420.68621.stgit@localhost6.localdomain6> <20110723014633.GA32507@infradead.org> Mime-Version: 1.0 Content-Type: text/plain; charset="ISO-8859-1" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20110723014633.GA32507@infradead.org> Sender: linux-kernel-owner@vger.kernel.org To: Christoph Hellwig Cc: Dan Williams , Roland Dreier , Dave Jiang , "linux-kernel@vger.kernel.org" , "linux-scsi@vger.kernel.org" List-Id: linux-scsi@vger.kernel.org On 2011-07-23 03:46, Christoph Hellwig wrote: > On Fri, Jul 22, 2011 at 01:59:39PM -0700, Dan Williams wrote: >> Some storage controllers benefit from completions always being steered >> to the strict requester cpu rather than the looser "per-socket" steering >> that blk_cpu_to_group() attempts by default. > > Isn't this actually dependent on the cpu, and not the storage > controller? It is, it's completely indendent of the controller used. Perhaps some drivers could have a very heavy end io completion handling causing the problem to become larger, but in general it's an artifact of the CPU and not the controller. -- Jens Axboe