[fleeno]

I have a Challenge S that I previously had running IRIX and OpenBSD at one point with no issues, and recently hauled it out to tinker with.

Something has to have gone wrong with the hardware, or I'm just out of practice. I've been trying to install IRIX, old OpenBSD, or NetBSD with no luck. I've tried everything I can think of:

* Real HDD, BlueSCSI, SCSI2SD
* Installing via Reanimator over the network
* Installing via CD
* New SCSI cables, new terminators
* Swapping RAM

IRIX 6.5.22 tries to start, but gives me a "vector=UTLB Miss." I thought RAM, but unless every one I have has gone bad, I'm not sure. I have plenty of SIMMs and have tried a bunch of configurations with the same result. I returned it to the same config that I knew worked, real HDD (though not the same actual drive as when it worked (I used it for another project), though I have tried several drives), same terminator configuration.

NetBSD from CD starts, but ends up repeating a SCSI timeout over and over. OpenBSD from CD hangs after complaining about the clock being wrong.

Any ideas what I might check? I know these can have power supply issues that manifest as weird issues. Or maybe the RTC battery?

Here's an hinv with the CDROM connected:

>> hinv
                  System: IP22
                Processor: 100 Mhz R4600, with FPU
    Primary I-cache size: 16 Kbytes
    Primary D-cache size: 16 Kbytes
              Memory size: 128 Mbytes
                SCSI Disk: scsi(0)disk(1)
              SCSI CDROM: scsi(0)cdrom(6)

Here's the UTLB Miss error when trying to install via Reanimator:

Enter the name of the remote host: rbpi
Enter the remote directory: IRIX/6.5.22/ovl1/dist


1) Remote Tape  2)[Remote Directory]  3) Local CD-ROM  X) Local Tape
      *a) Remote directory IRIX/6.5.22/ovl1/dist from server rbpi.

Enter 1-4 to select source type, a to select the source, <esc> to quit,
or <enter> to start:


Setting $netaddr to 192.168.9.1 (from server rbpi)

Exception: <vector=UTLB Miss>
Status register: 0x30004803<CU1,CU0,IM7,IM4,IPL=???,MODE=KERNEL,EXL,IE>
Cause register: 0x8008<CE=0,IP8,EXC=RMISS>
Exception PC: 0x8ff9c39c, Exception RA: 0x8ff9a234
exception, bad address: 0x1c
  Saved user regs in hex (&gpda 0xa8740e48, &_regs 0xa8741048):
  arg: a8740000 43 8ffdf558 a
  tmp: a8740000 8ffa9c78 8ffb0a28 a8747fa4 887fe7d4 9fc4b06c 9fc4b074 31
  sve: a8740000 3 400000 8000000 16 3f80 0 8000000
  t8 a8740000 t9 ffffffff at ffffffff v0 ffffffff v1 ffffffff k1 0
  gp a8740000 fp ffffffff sp ffffff7f ra ffffffff

PANIC: Unexpected exception

[Press reset or ENTER to restart.]



Thanks!

[Raion]

What PROM version?

[weblacky]

I was actually going to say to check your RTC battery because this thing is basically an Indy from my understanding... If you truly have a bad battery then you also have no ethernet MAC address. That can make things really screwy.

Fix your MAC address (temporary of course) and then do a warm boot, using the reset button, not using shut down and see if that fixes the issue.

There in fact seems to be a correlation between the "Exception: <vector=UTLB Miss>" and the Mac address loss discovered via MAME PROM emulation efforts, found here: https://forums.irixnet.org/thread-1012-p...l#pid28501

[vishnu]

What it means is that what was supposed to be in the lookaside buffer wasn't there, it "missed." Obviously, that's not supposed to happen, thus the exception is thrown. When an exception is thrown, the processor unwinds the stack with the assumption that there will be code somewhere up the chain that will be able to handle the exception. In this case, that didn't happen. Unfortunately, exception handling is a criminally neglected component of software engineering.

[fleeno]

I'm thinking my RTC battery is good, as I do have a a MAC address. Here is my env and PROM version:

>> printenv
AutoLoad=Yes
TimeZone=PST8PDT
console=g
diskless=0
dbaud=9600
volume=80
sgilogo=y
autopower=y
netaddr=192.168.9.1
eaddr=08:00:69:11:11:11
cpufreq=100
SystemPartition=scsi(0)disk(1)rdisk(0)partition(8)
OSLoadPartition=scsi(0)disk(1)rdisk(0)partition(0)
OSLoadFilename=/unix
OSLoader=sash
NoAutoLoad=CONSOLE OPEN FAILED.
ConsoleOut=serial(0)
ConsoleIn=serial(0)
>> version
PROM Monitor SGI Version 5.3 Rev B10 R4X00/R5000 IP24 Feb 12, 1996 (BE)
>>

[fleeno]

I just tested the unloaded power supply voltages, and they all look good. The yellow wire reads ~8V, but according to this post it should read 8V when not under load.

https://forums.sgi.sh/index.php?threads/...mation.14/

[vishnu]

I just tested the unloaded power supply voltages, and they all look good. The yellow wire reads ~8V, but according to this post it should read 8V when not under load.

https://forums.sgi.sh/index.php?threads/...mation.14/

It's actually not too much of a challenge to build a resistor configuration to test low-voltage DC power supplies under load, you can buy everything you need from (for example) digikey or mouser. Don't forget the magic incantation, twinkle twinkle little star, I is equal to E over R. 

[robespierre]

TLB Miss exceptions are expected and occur frequently, because the TLB is far too small to hold all the pages that can be accessed. MIPS uses a software table-walk scheme, so each time a virtual memory address from a mapped segment misses in the TLB, software takes an exception to load the proper mapping into the TLB and restart the instruction.

In this panic report, the code that was executing is in kseg0 (Exception PC: 0x8ff9c39c, Exception RA: 0x8ff9a234).
The BadVAddr is 0x1c, which is a mapped virtual address in useg—and is a suspicious address, being so close to zero (null pointer). This could be caused by an uninitialized parameter used to reference a structure.

The code that was running was unprepared to handle a TLB Refill, so the handler that was put at the UTLB vector (0x80000000 or 0x00000000 physical) just dumps some information and halts. This would be the case early in Unix startup before the process machinery is set up.

[weblacky]

OK, super dumb question from rereading your initial post:

You obviously knew that you use an external terminator on the internal SCSI chain, you're not enabling internal termination on the storage devices right? That timeout error during CD boot would lead me to believe you have multiple terminations on one or both side sides of your chain turned on?

SGI machines of this vintage should have parity enabled on everything, do you have parity on everything? You're not supposed to mix this setting, either all devices have it or none of the devices have it enabled.

The terminal has a date command, so go into your PROM terminal and actually set a real date so it doesn't complain: https://techpubs.jurassic.nl/library/man.../ch09.html

What do you have externally attached? If you're using a external optical drive did you also match termination? Are you doing active or passive termination on the external?

[fleeno]

I am using a terminator on the CDROM drive, it has a terminator on one port, and the other connected to the machine. On the internal side I've tried a plain SCSI cable with termination enabled on the BlueSCSI. I've also tried the original SCSI cable that has a built-in terminator, with the BlueSCSI's termination disabled. I've also tried a Seagate drive just like the one that came with the machine, using a terminator on the internal cable. When not using the CDROM, I had a terminator on the external port.

I have in the past installed both OpenBSD and IRIX using a real drive and this same CDROM drive.

It was my understanding that you need a terminator on both the external port and the internal port, I think that's what you're saying too. These are active terminators (or at least they say they are!)

As for parity, the real drive does have its parity jumper installed. The BlueSCSI defaults to parity on. The CDROM drive I guess I'm not sure on, it's one of those Sun external drives that matches a SPARCstation. But even so I know I have parity on the drive when it's the only thing on the bus.

I did just try setting the date, which works, but OpenBSD still complains. I'm not sure if OpenBSD just halts when it doesn't like the date, or if it's something later that's hanging things up.

Thanks!

---

I just tried something I hadn't tried before, using the BlueSCSI as both the hard drive and CDROM drive. I put the NetBSD image on there as a CD, and it actually booted. However, the installer screen is all garbled and just redraws over and over. I've attached an image, it's very odd!