From mboxrd@z Thu Jan 1 00:00:00 1970 From: Hannes Reinecke Subject: Re: [PATCHv2 05/14] libata: NCQ Encapsulation for READ LOG DMA EXT Date: Thu, 14 Apr 2016 17:59:48 +0200 Message-ID: <570FBE74.2040202@suse.com> References: <1460443678-57934-1-git-send-email-hare@suse.de> <1460443678-57934-6-git-send-email-hare@suse.de> <20160413180735.GD3676@htj.duckdns.org> <570F2E33.3080709@suse.de> <20160414154338.GB12583@htj.duckdns.org> Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20160414154338.GB12583@htj.duckdns.org> Sender: linux-ide-owner@vger.kernel.org To: Tejun Heo , Hannes Reinecke Cc: linux-ide@vger.kernel.org, "Martin K. Petersen" , Christoph Hellwig , James Bottomley , Shaun Tancheff , Damien Le Moal , linux-scsi@vger.kernel.org List-Id: linux-scsi@vger.kernel.org On 04/14/2016 05:43 PM, Tejun Heo wrote: > Hello, Hannes. > > On Thu, Apr 14, 2016 at 07:44:19AM +0200, Hannes Reinecke wrote: >> Hehe. No, it isn't, if you look closely. >> (Or make that: it _shouldn't_, and I've messed it up) >> (That's what the 'fpdma' parameter is for) >> >> The benefit is not so much for normal operations, but it'll give us >> a performance improvements on SMR drives where we need to issue >> READ LOG DMA EXT rather frequently. There we really want to have >> them as NCQ commands. > > I'm still a bit confused. Isn't it part of EH? If so, the path is > never traveled with other commands in flight and thus whether a > command is NCQ or not doesn't make any difference. The only thing > it'd do is making devices which advertise the capability but don't get > it quite right fail. > For this patch, yes, you are right. However, the ZAC enablement patches later on submit READ LOG EXT commands (for REPORT ZONES), and _they_ benefit from NCQ encapsulation. Cheers, Hannes