From owner-freebsd-stable Fri Feb 21 6:32:51 2003 Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 54EEB37B418 for ; Fri, 21 Feb 2003 06:32:47 -0800 (PST) Received: from isp1.viapraga.cz (isp1.viapraga.cz [212.67.64.2]) by mx1.FreeBSD.org (Postfix) with ESMTP id 80CBB43FBD for ; Fri, 21 Feb 2003 06:32:46 -0800 (PST) (envelope-from Frantisek.Rysanek@pragonet.cz) Received: from (nat.pragonet.cz [212.67.66.62]) by isp1.viapraga.cz (Sendmail) with ESMTP id h1LEWjgN002586; Fri, 21 Feb 2003 15:32:46 +0100 Received: from localhost (frr@localhost) by rysanekf2.pragonet.cz (8.9.3/8.9.3) with ESMTP id PAA02168; Fri, 21 Feb 2003 15:34:58 +0100 Date: Fri, 21 Feb 2003 15:34:58 +0100 (CET) From: Frantisek Rysanek To: stable@freebsd.org Cc: scott_long@btc.adaptec.com Subject: bug/misbehavior report - Adaptec ASR2120 (aac/aacp) under 4.8-pre-release1 Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Dear gentlemen, I'd like to report my experience with the Adaptec ASR2120 RAID controller. Summary ======= OS Version: 4.8-PRE1 (from current.freebsd.org, FEB.20th) vs. 4.7 Hardware device: Adaptec ASR2120 - U320 SCSI RAID controller Driver: aac (device aac and aacp) Problem: until 4.7, the aac driver didn't work with the ASR2120 (due to the ASR2120 being slightly different than previous aac chips - something with DMA addressing) in 4.8-PRE1, the installer kernel works fine, but the GENERIC kernel installed on the RAID volume doesn't boot due to a quirk between aac/aacp devices. Suggested correction: disable device aacp in the installer and GENERIC kernels. Either for the whole aac driver in general, or as a QUIRK option specific to ASR2120 - or disable the initial bus reset done by aacp, that might just make it. Our current workaround: we had a 5.0 installed on a stand-alone disk - we booted 5.0, mounted the RAID volume with 4.8-PRE1 installed, copied the installer kernel from the installer floppy, booted that, built a custom 4.8-PRE1 kernel with aacp commented out. It's been working ever since. More information: the aacp (pass through device) seems to send a bus reset to the "RAID-private" bus, apart from messing with the individual devices. The RAID controller detects the reset in the first place and the RAID operation gets disrupted for a while, probably due to the disk drives recovering from the reset. This temporary disruption is enough to prevent the kernel from booting off the aac device - it says stuff like "SCSI command timed out". The same happens with 5.0 too, only it usually manages to boot - unless the RAID's performance is further hampered e.g. by a background RAID volume rebuild during boot. Took me a while to find this out. In case this is a design decision: does anyone need RAID pass-through in the installer? Is anyone still using SCSI CD-ROM/floppy drive, having only the RAID controller available, therefore forced to attempt the pass-through? Most machines I know nowadays are equipped with a legacy floppy controller and an IDE CD-ROM drive. I can imagine that, OTOH, quite a number of people would like to install FreeBSD on a RAID drive and boot off that drive - right away, without cumbersome workarounds to get the damn thing to work. As far as I understand, the ASR2120 supports a SCSI CD-ROM as a direct-access device out of the box. I'm eagerly awaiting 4.8-RELEASE with a hope of getting the ASR2120 RAID to work under FreeBSD 4.x. The behavior of 4.8-PRE1 urges me to notify you of this partial misbehavior while there's some time to tweak it. Thanks for the great job you're doing. Frank Rysanek To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message