From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christoph Hellwig Subject: Re: [PATCH 06/17] scsi: add support for per-host cmd pools Date: Fri, 7 Feb 2014 04:46:40 -0800 Message-ID: <20140207124640.GC30510@infradead.org> References: <20140205123930.150608699@bombadil.infradead.org> <20140205124019.946278511@bombadil.infradead.org> <52F4A8C4.8080602@cs.wisc.edu> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from bombadil.infradead.org ([198.137.202.9]:33605 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932426AbaBGMqp (ORCPT ); Fri, 7 Feb 2014 07:46:45 -0500 Content-Disposition: inline In-Reply-To: <52F4A8C4.8080602@cs.wisc.edu> Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: Mike Christie Cc: Christoph Hellwig , Jens Axboe , James Bottomley , Nicholas Bellinger , linux-scsi@vger.kernel.org On Fri, Feb 07, 2014 at 03:35:00AM -0600, Mike Christie wrote: > It seems there is a issue with using the cmd_size to indicate the driver > has its own cmd pool and also using that for scsi mq enabled drivers to > indicate that we want the LLD's struct allocated by blk/scsi mq. > > If a driver sets cmd_size for only the scsi/blk mq purpose, this patch > wants the driver to also setup a cmd pool which I do not think is used > when doing scsi/blk mq. I don't quite understand what you mean. cmd_size means that each scsi_cmnd passed to the driver will have additional per-driver data. For the old path it's implemented by creating a slab cache and storing it in the host template to easily find it, for blk-mq it is used to increase the allocation in the block core as scsi doesn't do it's own allocations in this case. Very different implementation underneath, but same effect for the driver.