This collection of posts was originally posted by Dave Olson. It has been added to by the SGI FAQ group. From news.yale.edu!yale!yale.edu!newsserver.jvnc.net!howland.reston.ans.net!europa.eng.gtefsd.com!uunet!olivea!sgigate.sgi.com!fido.asd.sgi.com!zuni.esd.sgi.com!anchor!olson Thu Dec 16 08:30:50 1993 Newsgroups: comp.sys.sgi.hardware Path: news.yale.edu!yale!yale.edu!newsserver.jvnc.net!howland.reston.ans.net!europa.eng.gtefsd.com!uunet!olivea!sgigate.sgi.com!fido.asd.sgi.com!zuni.esd.sgi.com!anchor!olson From: olson@anchor.esd.sgi.com (Dave Olson) Subject: Re: adding 3rd party DAT drive Message-ID: Keywords: DAT Sender: news@zuni.esd.sgi.com (Net News) Organization: Silicon Graphics, Inc. Mountain View, CA References: Date: Tue, 14 Dec 93 06:45:17 GMT Lines: 182 In mg26101@i9.msi.umn.edu (Ronald Sakaguchi) writes: | I've recently purchased an HP 2.0GB DAT drive and am having problems | getting the system to recognize it. Reading the recent posts, I've | modified the master.d/tpsc to add a listing for HP 35470A. The system | stills thinks that it is 'unknown.' What other changes need to be | made? Would someone mind listing again how to change the tpsc file? | BTW, I've rebooted the system and recreated the tape files under /dev. | Still no luck. You probably didn't get the change to tpsc correct. Did you change the string lengths as well as the strings? Here's the mail/postings I've saved on the HP DAT drives on SGI systems. I make no claims as to correctness, but it seems to have worked for other folks. ============= Newsgroups: comp.periphs.scsi Subject: Re: HP35480A DAT drive - need to disable non-aligned disconnects Date: 4 May 1993 00:00:33 GMT |> | The errors I got were |> | unix: dks0d1s0 (/): Aborted command: Synchronous transfer error,. |> |> I only know of one device that actually returns that error, and that |> is the IBM 1.2 GB (OEM 663, in our case) drive. We have seen this on |> a few systems, and have so far *always* tracked it down to a bad connector |> for that drive, either on the sled (almost always), or on the backplane. I tried re-enabling synchronous mode with the "Host connectivity options" (DIP switches 7 and 8) set to 0 (=SUN). This apparently makes a difference as things are now working correctly. To summarize, I've got the DIP switches set as follows: switch 1 1 = compression enabled 2 1 = compression host controllable 3 0 = data phase disconnect disabled 4 1 = parity enabled 5 1 = default to variable block mode 6 1 = enable non-immediate file marks 7&8 00= Sun mode (11=hp, 10=dec) >From comp.sys.sgi.hardware Fri Aug 20 13:07:07 1993 Newsgroups: comp.sys.sgi.hardware Path: zuni.esd.sgi.com!fido.asd.sgi.com!sgi!decwrl!spool.mu.edu!torn!nott!nrcnet0!rss30.di.nrc.ca!ng Subject: Re: New DAT drive problem (Summary) Date: Fri, 20 Aug 93 18:02:56 GMT : I'm trying to install a new HP 35480A SCSI DAT Drive on an R4000. : However, the system doesn't recognize it as an DAT. What steps : should be taken to make it work? I couldn't find any pointers : from the manual nor the FAQ. Eventually I've managed to have the HP35480A DAT drive working. The following is the summary: 1. Set the dip switches on the bottom of the drive: 1 1 = compression enabled 2 1 = compression host controllable 3 0 = data phase disconnect disabled 4 1 = parity enabled 5 1 = default to variable block mode 6 1 = enable non-immediate file marks 7&8 00 = Sun mode (11=HP, 10=Dec) 2. Update /usr/sysgen/master.d/tpsc such that the system will recognize the drive as a "DAT" instead of an "unknown" scsi device by inserting an entry after the SGI DAT drive ("ARCHIVE", "Python"): { DATTAPE, TPDAT, 2, 8, "HP", "HP35480A", 0, 0, {0, 0, 0, 0}, MTCAN_BSF|MTCAN_BSR|MTCAN_APPEND|MTCAN_SETMK|MTCAN_PART|MTCAN_PREV| MTCAN_SYNC|MTCAN_SPEOD|MTCAN_CHKRDY|MTCAN_VAR|MTCAN_SETSZ| MTCAN_SILI|MTCAN_SEEK|MTCAN_CHTYPEANY, /* minimum delay on i/o is 4 minutes, because when a retry is * performed, the drive retries a number of times, and then * rewinds to BOT, repositions, and tries again. */ 40, 4*60, 4*60, 5*60, 512, 128*512 }, N.B. According to , the maximum block transfer size is only 128*512 because of SGI's DMA boundary restriction. [Alexis Cousein notes that this is not true: *Sun* can't handle DMA of more than 64K (128*512), but SGI can handle 512*512, which is why it's the default.] 3. Issue "autoconfig" and reboot the system to have the drive properly recognized. 4. Issue "/dev/MAKEDEV tps". 5. If you like, "ln /dev/mt/tpsSdDnsv /dev/dat" etc. where S is the SCSI controller id and D is the SCSI address of the drive. 6. To complete the picture, the SCSI address selection jumpers of the HP 35480A DAT drive is used to differentiate dual density. From the info sheet from my vendor: Desired SCSI addr Desired density Enter setting ----------------- --------------- ------------- 4 hi 0 4 lo 1 5 hi 2 5 lo 3 6 hi 4 6 lo 5 7 hi 6 7 lo 7 Also the consensus is this should go to the FAQ. Thanks again. >From comp.sys.sgi.hardware Tue Aug 24 10:43:25 1993 Newsgroups: comp.sys.sgi.hardware Subject: Re: New DAT drive problem (Summary) Date: Tue, 24 Aug 93 14:25:37 GMT : : 6. To complete the picture, the SCSI address selection jumpers of the : HP 35480A DAT drive is used to differentiate dual density. : : From the info sheet from my vendor: : : Desired SCSI addr Desired density Enter setting : ----------------- --------------- ------------- : 4 hi 0 : 4 lo 1 : 5 hi 2 : 5 lo 3 : 6 hi 4 : 6 lo 5 : 7 hi 6 : 7 lo 7 : Due to numerous email messages questioning about the jumper settings above, I have to explain more details. The complete text of the fax obtained from the vendor (not HP): Due to a lack of software on Unix based systems to support dual density DAT tape drives, it is necessary to modify the SCSI address wiring. Consequently, you will have to identify the desired SCSI address and density, (high or low), using the table provided below to compensate. [Table above inserted here.] Specifically, the modification consists of re-mapping the three address select lines such that one of the three can be used to designate density, (high or low). Please refer to the diagram below. The remaining two lines while still used for address selection have been reassigned to accomodate the change. In particular, of the three lines bit0 is now assigned as density select. Bit1 and bit2 have been re-mapped as bit0 and 1 respectively, with the actual bit2 wired "high". Therefore, with the address select lines thus modified it is necessary to use the table above to obtain the correct address. bit 2 1 0 +----+----+----+ +----+ | 1 | | | | | density +----+----+----+ +----+ ^ ^ ^ | | | +---+ +--+ +---------- | | | +----+----+----+ | 1 | | | +----+----+----+ bit 2 1 0 Further notes: At first I thought all the 35480A DAT drives sold for SGIs (or other Unix boxes) were modified as above. However, it might not be the case owing to the number of inquiries I received. We have an external unit with a SCSI address selector on the enclosure. I've experimented with various settings and it did conform to the data sheet above. Hope the picture is clearer now. -- The most beautiful things in the world are | Dave Olson those from which all excess weight has been | Silicon Graphics removed. -Henry Ford | olson@sgi.com From: wbau0931@mailszrz.zrz.TU-Berlin.DE (Wolfgang Baumann) Newsgroups: comp.sys.sgi.hardware Subject: Re: HP35480A & Irix 5.2 : dma_map problem Date: 9 Nov 1994 21:11:20 GMT Organization: Technical University Berlin, Germany Kris Boulez (kris@bionmr1.bionmr1.rug.ac.be) wrote: : I installed a HP35480A DAT on our 4D410 (irix 5.2) following the : guidelines given in the faq. I am able to backup and restore using : the command in the Toolchest (which seems to be 'bru') without any : problem. When I want to use tar however I can create the archive (tar : cvf from the command line or using the program that pops up when : double clicking the DAT icon on the desktop), but when I want to list : or extract the archive I get the following errors in the SYSLOG file. : : Nov 6 17:10:16 4A:bionmr1 unix: WARNING: dma_map: address not word aligned : Nov 6 17:10:16 2A:bionmr1 unix: wd93 SCSI Bus=1 ID=5 LUN=0: SCSI cmd=0x8 disconnected on non-word boundary (addr=c009a402, 0x7bfe left), can't DMA. Resetting SCSI bus : Nov 6 17:10:16 1A:bionmr1 unix: ALERT: Integral SCSI bus 1 reset [HELP=SCSI_RESET] Other reported that problem,too. From a paper provided by our DAT jukebox vendor: Set the configuration switches to on / on / off / off / on / off / off / ON The crucial on is the eighth which disables disconnects on non-word boundaries. Also you want to limit your default block sizes to 64 kB blocks by changing the last line of the entry in /usr/var/sysgen/master.d/scsi to 40, 4*60, 4*60, 5*60, 512, 64*512 }, I think this should be added to the FAQ. -- Wolfgang W. Baumann a.k.a. Karl Hansen internet: Wolfgang.Baumann@pi.tu-berlin.de "keep on running" / beer & cake +------------------------------------------------------------------------------+ | "Two beer or not two beer - that is no question! " (Bill Shakesbeer: Hamlet) | +------------------------------------------------------------------------------+ Technische Universitaet Berlin office: +49 (30) 314-24733 Hermann-Foettinger-Institut fuer sekr.: -23359 Thermo- und Fluiddynamik fax: -21101 Mueller-Breslau-Str. 8 D-10623 Berlin, Germany