Jump to content

SGI IRIS 3000 Troubleshooting Guide

From RetroTechCollection
SGI IRIS 2400 Turbo cabinet back panel (68020 course workbook)
Back panel of the IRIS 2400 Turbo, which uses the same processor, memory and graphics boards as the 3000: status display, reset and halt buttons, configuration switch, ports 1 to 4 and Ethernet (SGI 68020 course workbook)

Fault-finding on the SGI IRIS 3000, from SGI's IRIS Series 3000 Owner's Guide, its field-service courses and the SGI Field Engineering technical aids reproduced in them, and Jonathan Levine's 1993–94 IRIS FAQ.[1][2][3] The IRIS 2000 Turbo models use the same IP2, IM1 and graphics boards and behave the same way; problems particular to the 2000 series are on SGI IRIS 2000 Troubleshooting Guide.[4]

Start-up

[edit | edit source]

At power-on the IP2 first tries to talk to the keyboard. If that works, the LEDs above the keyboard's number row run from left to right, the keyboard beeps and the PROM monitor prompt iris> appears on the screen.[3][5] If the IP2 cannot reach the keyboard, it assumes either a faulty keyboard or a server without graphics and sends the monitor prompt to port 2 instead; Levine advises keeping a terminal connected to port 2 for this reason.[3] A workstation that comes up only on port 2 with the keyboard connected may have /dev/syscon linked to /dev/ttyd1; SGI's technical aid IO-012 shows how to check the inode and device numbers with ls -li in /dev.[6]

If the keyboard does not start, check its power before replacing it. Its LEDs run from +12 V, but it also needs −12 V. On the "Control Panel / Port 1" connector, +12 V is on pins 9, 21 and 22 and −12 V on pin 18, with pins 1, 4, 5, 6 and 7 at ground. If a supply is missing, check the two fuses along the top edge of the I/O panel board, under the power compartment cover, with a meter; Levine found one that looked intact but was open inside its base.[3]

Status display

[edit | edit source]

The single hexadecimal digit on the back panel has no published key. Levine recorded on a running 3130: 5 when idle at the PROM monitor, 2 and 3 alternating while booting or running UNIX, and F then 6 then 5 while shutting down.[3]

Boot failures

[edit | edit source]
  • Wait for the disk to reach speed before typing b at the PROM monitor: the owner's guide says 20 seconds, Levine 30. Booting too soon gives the same failure as a missing boot file, without harm.[5][3]
  • md.0:defaultboot No such file or directory usually means someone deleted the file defaultboot from the root of the boot disk, and the system has to be rebuilt from a bootable mkboot tape. A defaultboot that loads and then crashes may be a damaged copy. The PROM monitor's ls lists the root file system, so you can check that bin, etc, lib, tmp, usr and defaultboot are there.[3]
  • A system tape from a machine with a different disk controller (ST-506, ESDI or SMD) carries the wrong disk driver.[3]
  • An RS-232 device can hang the system during boot because IRIS raises RTS (pin 4) before the port's getty has started. SGI's workarounds are to connect and switch the device on before booting, or (preferred) to use a three-wire null modem cable.[7]

Error messages

[edit | edit source]

When the kernel stops with a panic: message, the device abbreviation in it points to the failing hardware:[1]

Device abbreviations in panic messages (SGI Table B-1)
Abbreviation Device
si Hard disk on the 3030, 3120B and 3130 (ESDI, Storager)
md Hard disk on the 3010, 3020, 3110, 3115 and 3120 (ST-506, DSD)
mt Tape drive (any type)
qic Quarter-inch cartridge tape
mf Floppy disk drive
dsd Tape, small disk and floppy controller
nx Ethernet

[1] SGI's advice for any panic is to run sync, reboot and run fsck, and to call for service if the message comes back. A panic naming none of these devices points to another board (processor, memory or backplane) or to the kernel.[1] Other messages listed in Appendix B of the owner's guide:[8]

  • parity error: a memory parity error, with the physical address in a preceding message.
  • iinit: the root file system could not be read; the disk or the root file system is damaged.
  • dsd: ccb timeout, dsd: cib timeout or md[0-1][a-h]: couldn't start! on ST-506 machines: a faulty disk controller.
  • dsd0: hard error: a bad connection between disk and controller, a bad disk, or hard disk errors.
  • sicmd: timeout ... on ESDI machines: a faulty controller, bad disk cables or wrong controller jumpers.
  • soft error ... on block NN: a recoverable error, possibly a bad track; call for service if it repeats.
  • (***No label***): the disk has no label.
  • qic0: can't get status during init: the tape drive has no power, or its control cable is disconnected or reversed.

Disks

[edit | edit source]

Levine reports that problems with large ESDI drives on the Interphase Storager come from the drive's data rate, not its size: the Storager II handles 12.5 MHz, so a 20 Mbit/s drive may fail where a 10 Mbit/s one of similar size works. He says this is anecdotal and not from his own experience.[3] He could not get Seagate ST-4096 drives to work with the DSD controller.[3]

Stand-alone diagnostics

[edit | edit source]

SGI's field diagnostics for the IP2, IM1 and FP1 boards are on the IRIS 3000 diagnostic tape. Boot them from the PROM monitor with b ct0:ip2diag; their output appears on an ASCII terminal on port 2. memtest tests memory from the bottom limit set by lowtest to the top limit set by hightest, and fptest, fpwalktest and fpdiag test the FP1, which can also be affected by the IP2 and IM1 boards.[2] The tests are pass or fail. If they pass but memory errors still appear under UNIX, SGI's course says to raise and lower the +5 V and −5 V outputs by about 5% and run them again.[2]

Video

[edit | edit source]
  • No sync, oversize characters or every second character missing after changing switch settings: Levine's fix was the bottom two jumpers (BPRN and BPRO) on IP2 header H1, between U3 and U4 by the P1 connector; they may need fitting or removing.[3]
  • The RGB and sync lines must be terminated in 75 ohms at the last monitor in the chain only. The 3000-series monitors use terminators on a beaded chain, the older 2400 Turbo monitors switches on the back panel.[9]
  • A line across the bottom of the screen after power-on: SGI traced this to the convergence circuit when the monitor is plugged into the cabinet's switched outlet. Switch the monitor off and on again, or plug it into the wall.[10]
  • In European video format, DC4 boards 5000-090-05 and -06 without deviation authorisation DA#104 lose the last 12 pixels of each line; the fix is new PROMs at U17D and U18D.[11]
  • For an unstable genlock or NTSC output, SGI's technical aid IO-015 lists the DC4 clock jumper (INT/EXT at H22D, which must be on EXT), jumper E23 on the CG1 (which must be removed) and an etching fault on early DC4 boards at U14A.[12]

Ethernet

[edit | edit source]

The Excelan EXOS/201 board has a diagnostic LED on its front edge. Levine reports that it blinks steadily when all is well and blinks a coded B9 when the checksum of the Ethernet address PROM is wrong.[3]

[edit | edit source]

References

[edit | edit source]
  1. ↑ 1.0 1.1 1.2 1.3 SGI, 007-5220-030, Appendix B, Table B-1 and introduction, pp. B-1 and B-2.
  2. ↑ 2.0 2.1 2.2 Silicon Graphics, IRIS 3000 Series Workstation Differences (field-service study guide), lesson DIAG10, pp. 8 to 14 (File:SGI IRIS Series 3000 Workstation Differences Course.pdf).
  3. ↑ 3.00 3.01 3.02 3.03 3.04 3.05 3.06 3.07 3.08 3.09 3.10 3.11 Jonathan Levine, "Silicon Graphics IRIS 2000/3000 FAQ", revisions of May 1993 to June 1994, sections 4.1.2 to 4.1.4, 4.2.1, 4.2.2, 4.3.1 and 4.4.1.3, bitsavers.org/pdf/sgi/iris/IRIS_FAQ.txt.
  4. ↑ Silicon Graphics, IRIS Series 3000 Owner's Guide, version 3.0, 007-5220-030, 1987, section 1.2, p. 1-2 (File:SGI IRIS Series 3000 Owner's Guide (007-5220-030).pdf).
  5. ↑ 5.0 5.1 SGI, 007-5220-030, section 3.1, pp. 3-1 to 3-3. PROM monitor prompt; wait 20 seconds for the disk.
  6. ↑ SGI Field Engineering, Technical Aid IO-012 R1, "ttyd1 gets assigned as syscon", 29 July 1985, in IRIS Workstation Course, 68020 Based, Student Workbook part 2 (File:SGI IRIS Workstation Course 68020 Based Student Workbook Part 2.pdf).
  7. ↑ SGI Field Engineering, Technical Aid IO-019, "RS-232 devices are hanging the system during boot", 14 May 1986, in Student Workbook part 2.
  8. ↑ SGI, 007-5220-030, Appendix B, sections B.1.1 to B.1.3, pp. B-2 to B-5.
  9. ↑ SGI, IRIS 3000 Series Workstation Differences, lesson COMP1, step 5, p. 5.
  10. ↑ SGI Field Engineering, Technical Aid GEN-033, "Lines at bottom of Monitor", 16 December 1985, in Student Workbook part 2.
  11. ↑ SGI Field Engineering, Technical Aid IO-017, "Euro Video Display Problem", models 2x00 and 30x0, in Student Workbook part 2.
  12. ↑ SGI Field Engineering, Technical Aid IO-015, "NTSC problems", 16 December 1985, in Student Workbook part 2.


Retrieved from "https://retrotechwiki.com/index.php?title=SGI_IRIS_3000_Troubleshooting_Guide&oldid=29601"