From owner-freebsd-scsi Tue Jul 10 5:52: 5 2001 Delivered-To: freebsd-scsi@freebsd.org Received: from ns5.pacific.net.au (ns5.pacific.net.au [203.143.252.30]) by hub.freebsd.org (Postfix) with ESMTP id 6CF4837B403 for ; Tue, 10 Jul 2001 05:52:01 -0700 (PDT) (envelope-from mckay@thehub.com.au) Received: from dungeon.home (ppp203.dyn249.pacific.net.au [203.143.249.203]) by ns5.pacific.net.au (8.9.0/8.9.1) with ESMTP id WAA21620; Tue, 10 Jul 2001 22:51:54 +1000 (EST) Received: from dungeon.home (localhost [127.0.0.1]) by dungeon.home (8.11.3/8.11.1) with ESMTP id f6ACqMF26033; Tue, 10 Jul 2001 22:52:22 +1000 (EST) (envelope-from mckay) Message-Id: <200107101252.f6ACqMF26033@dungeon.home> To: Matthew Jacob Cc: mckay@thehub.com.au, freebsd-scsi@freebsd.org, Norman Czarczinski Subject: Re: Trouble with SCSI Streamer References: <01070421461600.00594@amd.local> In-Reply-To: from Matthew Jacob at "Wed, 04 Jul 2001 19:48:05 +0000" Date: Tue, 10 Jul 2001 22:52:22 +1000 From: Stephen McKay 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 Wednesday, 4th July 2001, Matthew Jacob wrote: >On Wed, 4 Jul 2001, Norman Czarczinski wrote: > >> On Wednesday 04 July 2001 19:34, you wrote: >> > Well, that's one theory shot down. Can you explicitly try changing the >> > blocksize to 1k, test, back to 512 and test again? >> >> That was it. With 1k it works. After switching back to 512 the problem came >> back. > >Okay- sounds like a bug in sa(4) then- and wierdness in the drive. It >shouldn't allow you to set it to 512 bytes if that doesn't work. This is deja vu time! We had this discussion a couple of years ago, if I recall correctly. QIC tapes in fixed block mode at densities of QIC-320 and above are in 1k blocks, and below that they are 512 byte blocks. It is not linked to a drive type, but to the tape density. Instead of getting to the bottom of this and making the sa driver know everything there was to know about QIC drives, the resolution was to use variable length blocks on QIC-320 and above. At the time, I had some philosophical quibbles with this approach, but it seems quite practical in retrospect. Indeed, checking the pre-cam st driver reveals the default is as I just described. But checking the cam sa driver reveals quirks that set 1024 byte fixed blocks on some drives, and more code that sets 512 byte blocks for lots of densities, including ones that it shouldn't. That's not right. I think a useful exercise would be to copy more of the QIC heuristics from the old st driver into the sa driver. Stephen. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message