From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jens Axboe Subject: Re: [PATCH] fix for Incorrect number of segments after building list problem Date: Wed, 20 Oct 2004 18:11:40 +0200 Sender: linux-scsi-owner@vger.kernel.org Message-ID: <20041020161139.GR10531@suse.de> References: <1097790683.1717.47.camel@mulgrave> <20041020143940.GK10531@suse.de> <20041020150708.GN10531@suse.de> <1098287481.2008.6.camel@mulgrave> <20041020155825.GP10531@suse.de> <1098288465.2008.13.camel@mulgrave> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from ns.virtualhost.dk ([195.184.98.160]:58274 "EHLO virtualhost.dk") by vger.kernel.org with ESMTP id S267866AbUJTQMT (ORCPT ); Wed, 20 Oct 2004 12:12:19 -0400 Content-Disposition: inline In-Reply-To: <1098288465.2008.13.camel@mulgrave> List-Id: linux-scsi@vger.kernel.org To: James Bottomley Cc: 'Dave Olien' , SCSI Mailing List On Wed, Oct 20 2004, James Bottomley wrote: > On Wed, 2004-10-20 at 10:58, Jens Axboe wrote: > > Ah yes, now I remember. How is the fix for that coming along, btw? > > Erm, it didn't cross my mind until I realised what this problem was > about. All SCSI drivers are setting 65536 (the default) for the max > segment length. However, this is being violated wholesale by all the > IOMMUs and it all works because their length descriptors are 32 bit > anyway ... > > However, I'll add it to the list and see if I can get something out. My > slight problem is testing. Only SATA and IDE have this short descriptor > problem, so I've nothing really to test the actual fix with. Just set some articial limit (say, 0x3fff) and add a debug check in init_io() to check the violation? -- Jens Axboe