From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ming Lei Subject: Re: [PATCH V7 4/6] blk-mq: introduce .get_budget and .put_budget in blk_mq_ops Date: Sat, 14 Oct 2017 01:29:55 +0800 Message-ID: <20171013172950.GA32319@ming.t460p> References: <20171012183704.22326-1-ming.lei@redhat.com> <20171012183704.22326-5-ming.lei@redhat.com> <9a741c03-90a3-e583-ddde-0ed71c8570a2@kernel.dk> <20171013001919.GA24715@ming.t460p> <6efdb459-8746-562d-06dc-5b3e172076e1@kernel.dk> <20171013160731.GA30899@ming.t460p> <1507912260.13345.0.camel@wdc.com> <20171013164519.GF30899@ming.t460p> <1507914527.13345.17.camel@wdc.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <1507914527.13345.17.camel@wdc.com> Sender: linux-kernel-owner@vger.kernel.org To: Bart Van Assche Cc: "linux-kernel@vger.kernel.org" , "linux-block@vger.kernel.org" , "hch@infradead.org" , "tom81094@gmail.com" , "himanshu.madhani@cavium.com" , "paolo.valente@linaro.org" , "axboe@kernel.dk" , "linux-scsi@vger.kernel.org" , "oleksandr@natalenko.name" , "john.garry@huawei.com" , "osandov@fb.com" , "loberman@redhat.com" , "james.smart@broadcom.com" List-Id: linux-scsi@vger.kernel.org On Fri, Oct 13, 2017 at 05:08:52PM +0000, Bart Van Assche wrote: > On Sat, 2017-10-14 at 00:45 +0800, Ming Lei wrote: > > On Fri, Oct 13, 2017 at 04:31:04PM +0000, Bart Van Assche wrote: > > > On Sat, 2017-10-14 at 00:07 +0800, Ming Lei wrote: > > > > Actually it is in hot path, for example, lpfc and qla2xx's queue depth is 3, > > > > > > Sorry but I doubt whether that is correct. More in general, I don't know any modern > > > storage HBA for which the default queue depth is so low. > > > > You can grep: > > > > [ming@ming linux]$ git grep -n cmd_per_lun ./drivers/scsi/ | grep -E "qla2xxx|lpfc" > > Such a low queue depth will result in suboptimal performance for adapters > that communicate over a storage network. I think that's a bug and that both > adapters support much higher cmd_per_lun values. > > (+James Smart) > > James, can you explain us why commit 445cf4f4d2aa decreased LPFC_CMD_PER_LUN > from 30 to 3? Was that perhaps a workaround for a bug in a specific target > implementation? > > (+Himanshu Madhani) > > Himanshu, do you perhaps know whether it is safe to increase cmd_per_lun for > the qla2xxx initiator driver to the scsi_host->can_queue value? ->can_queue is size of the whole tag space shared by all LUNs, looks it isn't reasonable to increase cmd_per_lun to .can_queue. > > > Even SRP/IB isn't big too, just 32. > > The default value for ib_srp for cmd_per_lun is 62 but that value can be > overridden by selecting another value in /etc/default/srp_daemon.conf. Note: > a lower value is selected if after SRP login it becomes clear that the target > queue depth is lower than the cmd_per_lun value requested by the user. This > is a performance optimization and avoids that the SRP target system has to > send back BUSY responses to the initiator. OK, thanks for sharing, I just read it as 32 in Laurence's machine. -- Ming