From owner-freebsd-questions Tue May 9 10:18:43 1995 Return-Path: questions-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id KAA19973 for questions-outgoing; Tue, 9 May 1995 10:18:43 -0700 Received: from estienne.cs.berkeley.edu (estienne.CS.Berkeley.EDU [128.32.42.147]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id KAA19967 for ; Tue, 9 May 1995 10:18:42 -0700 Received: from localhost (localhost [127.0.0.1]) by estienne.cs.berkeley.edu (8.6.11/8.6.9) with SMTP id KAA05598; Tue, 9 May 1995 10:18:07 -0700 Message-Id: <199505091718.KAA05598@estienne.cs.berkeley.edu> X-Authentication-Warning: estienne.cs.berkeley.edu: Host localhost didn't use HELO protocol To: Julian Elischer cc: msmith@atrad.adelaide.edu.au (Michael Smith), questions@FreeBSD.org Subject: Re: Tagged Command queueing... In-reply-to: Your message of "Mon, 08 May 1995 22:33:09 PDT." <199505090533.WAA18061@ref.tfs.com> Date: Tue, 09 May 1995 10:18:07 -0700 From: "Justin T. Gibbs" Sender: questions-owner@FreeBSD.org Precedence: bulk >> >> Greetings people! >> >> I currently run a couple of Ultrastor 34F controllers, which have >> proven extremely reliable and (going by the numbers posted here off and on) >> very quick. The literature for these controllers indicates that they >> support TCQ, and the disks I'm using support it as well. >The disk driver will queue several commands to the board at once >for each drive. (if the driver tells it it can) >I assume that the board will do TQ if it get's the oportunity.. >> ... >As for teh ultrastore controller, >look for: > /* > * fill in the prototype scsi_link. > */ > uha->sc_link.adapter_unit = unit; > uha->sc_link.adapter_targ = uha->our_id; > uha->sc_link.adapter = &uha_switch; > uha->sc_link.device = &uha_dev; > uha->sc_link.flags = SDEV_BOUNCE; > >and add to it: > uha->sc_link.opennings = 2; > >this will tell teh scsi code it can send upto two commands at >a time to each disk... >try iozone before and after and let me know if it changes? > >(I hadn't realised that that driver was single-threading it's commands..) > >take a backup first :) it might confuse the driver/scsi-system >if they can't handle 2 requests.. > >(later we can try 3,4,5 :) Before you try going above 4, you might want to fix the ccb allocation code so that running out of ccbs isn't terminal. Take a look at the buslogic and aic7xxx drivers to see what I mean. > >julian > -- Justin T. Gibbs ============================================== TCS Instructional Group - Programmer/Analyst 1 Cory | Po | Danube | Volga | Parker | Torus ==============================================