From mboxrd@z Thu Jan 1 00:00:00 1970 From: James Bottomley Subject: Re: [patch] qla1280 update for 2.6.1-mm4 Date: 19 Jan 2004 12:03:11 -0500 Sender: linux-scsi-owner@vger.kernel.org Message-ID: <1074531798.1894.17.camel@mulgrave> References: <16395.51364.833229.536552@gargle.gargle.HOWL> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Return-path: Received: from stat1.steeleye.com ([65.114.3.130]:7398 "EHLO hancock.sc.steeleye.com") by vger.kernel.org with ESMTP id S265377AbUASRD3 (ORCPT ); Mon, 19 Jan 2004 12:03:29 -0500 In-Reply-To: <16395.51364.833229.536552@gargle.gargle.HOWL> List-Id: linux-scsi@vger.kernel.org To: Jes Sorensen Cc: SCSI Mailing List , Andrew Morton , jeremy@sgi.com On Mon, 2004-01-19 at 07:08, Jes Sorensen wrote: > I am attaching the latest patch for qla1280, which includes Andrew's and > James' patches (modulo the 64 bit enable part) as well changes to make > it handle pci_set_dma_mask() correctly and switch to only use one of the > two SCSI command issuing versions depending on whether the driver is > compiled for 64 or 32 bit DMA. The old code effectively did this anyway, > but with this change it is no longer compiling in the part not used. OK, put it in the tree, thanks. Something to think about (I'm not asking you to change it) is whether you really want to use 64 bit descriptors on all 64 bit machines. The way the qla1280 single issue queue is organised it looks like we eat up slots (and, worse, overflow slots) quite a bit faster in 64 bit addressing mode. Therefore, it seems like the driver would operate more efficiently in 32 bit mode, so it might be worth a runtime check even on 64 bit machines to see whether the addressing mode is really necessary (i.e. no physical memory > 4GB for instance). James