From owner-freebsd-current Fri Sep 8 03:35:39 1995 Return-Path: current-owner Received: (from majordom@localhost) by freefall.freebsd.org (8.6.12/8.6.6) id DAA29521 for current-outgoing; Fri, 8 Sep 1995 03:35:39 -0700 Received: from Sysiphos (Sysiphos.MI.Uni-Koeln.DE [134.95.212.10]) by freefall.freebsd.org (8.6.12/8.6.6) with SMTP id DAA29504 for ; Fri, 8 Sep 1995 03:35:13 -0700 Received: by Sysiphos id AA28390 (5.67b/IDA-1.5 for current@freebsd.org); Fri, 8 Sep 1995 10:58:15 +0200 Message-Id: <199509080858.AA28390@Sysiphos> From: se@zpr.uni-koeln.de (Stefan Esser) Date: Fri, 8 Sep 1995 10:58:14 +0200 In-Reply-To: Jean-Marc Zucconi "Re: Another NCR problem" (Sep 8, 0:57) X-Mailer: Mail User's Shell (7.2.6 alpha(2) 7/9/95) To: Jean-Marc Zucconi Subject: Re: Another NCR problem Cc: current@freebsd.org Sender: current-owner@freebsd.org Precedence: bulk On Sep 8, 0:57, Jean-Marc Zucconi wrote: } Subject: Re: Another NCR problem } > Please always include the probe messages ! } } OK here they are: } Sep 6 19:30:35 qix /kernel: Probing for devices on the pci0 bus: } Sep 6 19:30:36 qix /kernel: configuration mode 1 allows 32 devices. } Sep 6 19:30:36 qix /kernel: chip0 rev 1 on pci0:0 } Sep 6 19:30:36 qix /kernel: chip1 rev 2 on pci0:7 } Sep 6 19:30:36 qix /kernel: vga0 rev 0 int a irq 12 on pci0:9 } Sep 6 19:30:36 qix /kernel: ncr0 rev 2 int a irq 11 on pci0:10 } Sep 6 19:30:37 qix /kernel: (ncr0:0:0): "MICROP 4110-09TBCU0322J HT01" type 0 fixed SCSI 2 } Sep 6 19:30:37 qix /kernel: (ncr0:1:0): "MICROP 2210-09MQ1001901 HQ30" type 0 fixed SCSI 2 } Sep 6 19:30:38 qix /kernel: (ncr0:2:0): "WANGTEK 5150ES SCSI ES41 B170" type 1 removable SCSI 1 } Sep 6 19:30:38 qix /kernel: (ncr0:3:0): "SONY CD-ROM CDU-8003A 1.9a" type 5 removable SCSI 2 } Sep 6 19:30:38 qix /kernel: (ncr0:4:0): "QUANTUM LIGHTNING 540S 241E" type 0 fixed SCSI 2 } Sep 6 19:30:39 qix /kernel: (ncr0:5:0): "SONY CD-ROM CDU-55S 1.0f" type 5 removable SCSI 2 } Sep 6 19:30:39 qix /kernel: pci0: uses 33554688 bytes of memory from f7fff000 upto f9ffffff. } Sep 6 19:30:39 qix /kernel: pci0: uses 256 bytes of I/O space from e400 upto e4ff. Looks fine ... } The machine is a P90 (Asus P55TP4XE Mainboard, Asus SC-200 scsi } controler, Diamond 968 Vram video) Is the Asus SC-200 at one end of the SCSI cable, and if Yes: Does it include an active terminator ? (The NCR8150S board does, but many other 53c810 based boards don't ...) } This happens some time after I start the system, and when there is } some activity on the scsi bus: I was doing a backup of the } disk(lun#4), but I already wrote 4 tapes w/o any problem. I see. } > Can you send a boot message log of the } > last version that did work ? } } No: this is a new machine (only up since sunday) } > } $ ncrcontrol -i } > } T:L Vendor Device Rev Speed Max Wide Tags } > } 0:0 MICROP 4110-09TBCU0322J HT01 10.0 10.0 8 3 } > } 1:0 MICROP 2210-09MQ1001901 HQ30 10.0 10.0 8 4 } > } 2:0 WANGTEK 5150ES SCSI ES41 B170 ? 5.0 ? - } > } 3:0 SONY CD-ROM CDU-8003A 1.9a 4.0 10.0 8 - } > } 4:0 QUANTUM LIGHTNING 540S 241E 10.0 10.0 8 - } > } 5:0 SONY CD-ROM CDU-55S 1.0f 3.6 10.0 8 - } } > Well, quite a number of devices ... } > How long is your SCSI bus cable, BTW ? } } targets 0,1,2 and 4 are internal (cable length ~ 60 cm - can't be } shortened!); the cdroms are external (length ~ 1.7m + 90cm). There } were no reads to the cdroms when the problem occured. Well, lets see, what the last public SCSI-2 draft says on this topic: # 5.2 Cable requirements # # The characteristic impedance of the cable should be no less than 90 ohms # and no greater than 140 ohms. The characteristic impedance of the cable # used when implementing the fast synchronous data transfer option is # defined in 5.2.3. [ ... ] # NOTES # 4 To minimize discontinuities and signal reflections, cables of different # impedances should not be used in the same bus. Implementations may require # trade-offs in shielding effectiveness, cable length, the number of loads, # transfer rates, and cost to achieve satisfactory system operation. [ ... ] # 5.2.1 Single-ended cable # # A 50-conductor flat cable or 25-signal twisted-pair cable shall be used # for the A cable. A 68-conductor flat cable or 34-signal twisted-pair # cable shall be used for the B cable if the wide SCSI option is # implemented. The maximum cumulative cable length shall be 6,0 m. If # twisted-pair cables are used, then twisted pairs in the cable shall be # wired to physically opposing contacts in the connector. # # A stub length of no more than 0,1 m is allowed off the mainline # interconnection within any connected equipment or from any connected # point. # # NOTE 7 Stub clustering should be avoided. Stubs should be spaced at least # 0,3 m apart. The interesting points are NOTE 4 and NOTE 7. If you mix internal and external cables, you most probably have different impedances. And with 4 devices on a 60 cm cable, you obviously can't have them at least 0.3 m apart ... The maximum length of 6 m does in general not work with transfer rates beyond 5MHz. I've had lots of problems with a workstation, that got an external disk and later an additional tape drive in distinct boxes. Adding the second box made the system crash within a few minutes of SCSI bus activity, independently of which device or set of devices was active. (I've seen FAST SCSI disk OEM manual list max. length vs. speed and in one case (a 1GB drive) I remember, the limit was 6 m at 2.2MHz, and 2 m at 5 Mhz ...) We switched to differential SCSI on all machines with any kind of external FAST devices for that reason, two years ago. } > Did you add devices lately ? } } No, but I can easily remove the cdroms. Yes, please do so! } > Have you tried to boot a kernel built with: } } > options "SCSI_NCR_MAX_SYNC=0" } } No. I will do it now. Maybe a limit of 5MHz (SCSI_NCR_MAX_SYNC=5000) will suffice ... Regards, STefan -- Stefan Esser, Zentrum fuer Paralleles Rechnen Tel: +49 221 4706021 Universitaet zu Koeln, Weyertal 80, 50931 Koeln FAX: +49 221 4705160 ============================================================================== http://www.zpr.uni-koeln.de/staff/esser/esser.html