From mboxrd@z Thu Jan 1 00:00:00 1970 From: James Bottomley Subject: Re: [PATCH 7/7] sg_ring: convert core ATA code to sg_ring. Date: Wed, 26 Dec 2007 11:12:59 -0600 Message-ID: <1198689179.3317.16.camel@localhost.localdomain> References: <200712191731.26512.rusty@rustcorp.com.au> <200712191836.06769.rusty@rustcorp.com.au> <200712191837.04697.rusty@rustcorp.com.au> <200712191838.27363.rusty@rustcorp.com.au> <47721294.7030401@gmail.com> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <47721294.7030401@gmail.com> Sender: linux-ide-owner@vger.kernel.org To: Tejun Heo Cc: Rusty Russell , Jens Axboe , linux-kernel@vger.kernel.org, linux-scsi@vger.kernel.org, linux-ide@vger.kernel.org List-Id: linux-scsi@vger.kernel.org On Wed, 2007-12-26 at 17:36 +0900, Tejun Heo wrote: > (PS, I haven't followed the sg chaining discussion. Why is sg chaining > an optional feature? Performance overhead on low end machines?) The idea of SG chaining is to allow drivers that wish to take advantage of it to increase their transfer lengths beyond MAX_HW_SEGMENTS*PAGE_SIZE by using chaining. However, drivers that stay below MAX_HW_SEGMENTS for the scatterlist length don't need to be altered. The ultimate goal (well, perhaps more wish) is to have all drivers converted, so SCSI can use something small for the default scatterlist sizing and dump all the sglist mempool stuff (although this may never be reached). James