From owner-freebsd-scsi Tue Mar 5 9:31:23 2002 Delivered-To: freebsd-scsi@freebsd.org Received: from panzer.kdm.org (panzer.kdm.org [216.160.178.169]) by hub.freebsd.org (Postfix) with ESMTP id 942EB37B416 for ; Tue, 5 Mar 2002 09:31:06 -0800 (PST) Received: (from ken@localhost) by panzer.kdm.org (8.11.6/8.9.1) id g25HV3h17997; Tue, 5 Mar 2002 10:31:03 -0700 (MST) (envelope-from ken) Date: Tue, 5 Mar 2002 10:31:03 -0700 From: "Kenneth D. Merry" To: Carl Reisinger Cc: freebsd-scsi@FreeBSD.ORG Subject: Re: Writing > 64KB records to SCSI Tape, FreeBSD 4.1 (?) Message-ID: <20020305103103.A17761@panzer.kdm.org> References: <200203050641.AOD06786@mail.mirapoint.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5.1i In-Reply-To: <200203050641.AOD06786@mail.mirapoint.com>; from cer@mirapoint.com on Mon, Mar 04, 2002 at 10:42:34PM -0800 Sender: owner-freebsd-scsi@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.org On Mon, Mar 04, 2002 at 22:42:34 -0800, Carl Reisinger wrote: > >I believe that's pretty close to what an Adaptec 1542 can > handle, so that's > >why we've left it at that size. (From a CAM standpoint > that is.) > > I haven't laid hands or eyes on a 1542 in nearly 8 years. > But from watching this list it seems a fair number of people > have cards that I haven't seen in years. Yes, I think Matthew Dodd's Adaptec 4944 may be one of the rarest pieces of hardware I've heard about in a while. IIRC, *very* few of them were made. > > > >I would suggest just bumping si_iosize_max up to 128K in > your copy of the > >sa(4) driver and see if that does the trick. > > > > That's what I would like to do, but my knowledge of the sa > driver, and FreeBSD drivers in general, is slim. > > Since sending my original message I have spent more time > looking over the code. If I understand things correctly the > dev parameter to saopen is actually a pointer to a specinfo > structure and I can just change the value of si_iosize_max > in saopen. True? > > If I want to get real fancy I can set si_iosize_max in > samount after the max record size supported by the drive is > found. (use the min of MAXPHYS and the drive max) I would suggest changing si_iosize_max in saregister() after the various make_dev(). Make sure you change it for each dev_t created. That way, you only end up changing the size once for each device, instead of every time you open/close. > (Now if there was a way to query the HBA for its max > transfer size). We've been planning on doing that, although it may end up being a little more complicated than just asking for a max transfer size. You may have a maximum size, a maximum recommended size, a maximum number of S/G elements, and so forth. Basically the idea would to to figure out what the size restrictions along the I/O path are. > The good news, for me, is that the controllers I need to > support is very limited. aic7899 and a few Symbios > controllers. The aic7899 seems to handle 16MB transfers. > Don't know about the Symbios chips. You should be fine with most any modern controller. The 7899 and Symbios controllers should be fine. > >> NOTE: While most of the kernel is based on 4.1-STABLE, I > >> believe the CAM subsystem and scsi_sa has been lifted > from > >> 4.3. (The behaviour is the same under 4.5) > > > >Sounds "interesting". :) > > > > We try to limit "changes", but every so often we need a fix > from a later release. And sometimes that involves back > porting a fair amount of code. I forget what the 4.1 CAM > problem was that caused us to pull in the 4.3 code. I know how it is. :) (Been there, done that, rather.) Ken -- Kenneth Merry ken@kdm.org To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message