Date: Tue, 10 Nov 1998 21:23:35 -0700 (MST) From: "Kenneth D. Merry" <ken@plutotech.com> To: M.S.Powell@ais.salford.ac.uk (Mark Powell) Cc: freebsd-scsi@FreeBSD.ORG Subject: Re: 3.0 scsi parity and tagged openings problems Message-ID: <199811110423.VAA14192@panzer.plutotech.com> In-Reply-To: <Pine.BSF.3.96.981110164026.19530A-100000@plato.salford.ac.uk> from Mark Powell at "Nov 10, 98 05:09:06 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
Mark Powell wrote... > > Just did a fresh install of 3.0-RELEASE on a machine that previously ran > 2.2.7-STABLE. It's a Viglen XX server with two 7880 chips on the > motherboard. It has three Fujitsu M2934 wide drives on each scsi bus, > connected thus: > > 7880 -> (da0) 2934 : (da1) 2934 : (da2) 2934 : CDROM : TAPE > 7880 -> (da3) 2934 : (da4) 2934 : (da5) 2934 > > I have the system on da0 and want to make a ccd of drives da1 and da3. > However, when I try to newfs the newly created ccd... > > # newfs /dev/rccd0a > Warning: 1280 sector(s) in last cylinder unallocated > /dev/ccd0a: 17013504 sectors in 4154 cylinders of 1 tracks, 4096 > sectors > 8307.4MB in 260 cyl groups (16 c/g, 32.00MB/g, 7936 i/g) > super-block backups (for fsck -b #) at: > n1, n2, n3, n4 etc. > Nov 10 15:48:55 titan /kernel: (da3:ahc1:0:4:0): tagged openings now 125 > Nov 10 15:48:55 titan /kernel: (da3:ahc1:0:4:0): tagged openings now 125 > Nov 10 15:48:55 titan /kernel: (da3:ahc1:0:4:0): tagged openings now 124 > Nov 10 15:48:55 titan /kernel: (da3:ahc1:0:4:0): tagged openings now 124 > Nov 10 15:48:55 titan /kernel: (da3:ahc1:0:4:0): tagged openings now 122 > Nov 10 15:48:55 titan /kernel: (da3:ahc1:0:4:0): tagged openings now 122 > Nov 10 15:48:55 titan /kernel: (da3:ahc1:0:4:0): tagged openings now 121 > etc. etc. until > Nov 10 15:48:55 titan /kernel: (da3:ahc1:0:4:0): tagged openings now 64 > Nov 10 15:48:55 titan /kernel: (da3:ahc1:0:4:0): tagged openings now 63 > Nov 10 15:48:58 titan /kernel: (da2:ahc0:0:1:0): tagged openings now 2 > Nov 10 15:48:58 titan /kernel: (da2:ahc0:0:1:0): tagged openings now 2 It looks like one of your disks at least doesn't like tagged queueing. The interesting thing is that it's one of the drives that is *not* in your CCD array. (you said that only da1 and da3 are in the array) Were you doing I/O to da2 at the time? What's on that disk? I would suggest that you: - disable tagged queueing for that disk or - get new, non-broken firmware for that disk Since you've got Fujitsu drives, it's questionable whether or not you'll be able to get non-broken firmware for them. You can disable tagged queueing one of two ways: - put a quirk entry in cam_xpt.c. There are examples of how to do this in there. (Just look for the Western Digital entries..) - use camcontrol to edit mode page 10 (0x0a) for the drive(s) in question, and set the DQue bit to 1. > Then the block numbers stop. First time it happened I thought the machine > had hung, so I sshed in and rebooted it. Upon the console appeared: > > syncing disks x y x ... > giving up > ahc0: data parity error detected during address or write data phase That's a PCI parity error. (I don't know what caused it, or why it appeared, etc..) > Next time I tried it, I noticed that it hadn't actually hung, but was > still printing a block number every couple of minutes or so. I killed it > and after a few minutes got my prompt back. Then the same newfs command > worked perfectly? > Any ideas what caused that? Bad firmware, most likely. Try disabling tagged queueing for the disk or disks in question and see if it helps any. The reason you probably didn't see this on 2.2.7-stable is because the old SCSI layer only allowed 4 tags at a time to go to a drive. It may be that your drive has trouble whenever its transaction queue gets full and it has to return queue full status. The old SCSI code probably wouldn't have filled the transaction queue, but CAM will send as many transactions as it can to the disk. Ken -- Kenneth Merry ken@plutotech.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199811110423.VAA14192>