From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christoph Hellwig Subject: Re: [PATCH 00/10] mpt3sas: full mq support Date: Tue, 7 Feb 2017 16:34:46 +0100 Message-ID: <20170207153446.GA22228@lst.de> References: <1485854760-122683-1-git-send-email-hare@suse.de> <20170207131909.GH19247@lst.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from verein.lst.de ([213.95.11.211]:41239 "EHLO newverein.lst.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754868AbdBGPes (ORCPT ); Tue, 7 Feb 2017 10:34:48 -0500 Content-Disposition: inline In-Reply-To: Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: Hannes Reinecke Cc: Christoph Hellwig , "Martin K. Petersen" , James Bottomley , linux-scsi@vger.kernel.org, Sathya Prakash , Kashyap Desai , mpt-fusionlinux.pdl@broadcom.com On Tue, Feb 07, 2017 at 03:38:51PM +0100, Hannes Reinecke wrote: > The SCSI passthrough commands pass in pre-formatted SGLs, so the driver > just has to map them. > If we were converting that we first have to re-format the > (driver-specific) SGLs into linux sg lists, only to have them converted > back into driver-specific ones once queuecommand is called. > You sure it's worth the effort? > > The driver already reserves some tags for precisely this use-case, so it > won't conflict with normal I/O operation. > So where's the problem with that? If it was an entirely separate path that would be easy, but it's not - see all the poking into the tag maps that your patch 8 includes. If it was just a few tags on the side not interacting with the scsi or blk-mq it wouldn't be such a problem.