[nater1217]

Hello!
I recently acquired 2 Octanes, one with SE graphics and one with MXE. I noticed that the one with MXE graphics had no hard disks installed, and I'm guessing it didn't work when it was put into storage.

When the MXE graphics board is installed, the system Kernel Panics. Originally, it would get to the point when X would load before panic-ing, but now, it panics immediately after booting. I can't even start the installer now. I did not touch the system to cause this change. It happened randomly just after I had finished installing Irix 6.5.26.

I know its the graphics board because the system boots without it installed or with the SE board installed. Moving the MXE board to the other octane cause it to do the same thing. The problem follows the graphics board.
I did try removing the TRAMs with no change in symptoms. It passes all IDE checks.

I attached two pictures showing the kernel panics. The 'Data Bus Error ' is the one it does now. It looks like the PC and Cause is the same, but the address changes every time.

Any Ideas? I tried (Carefully) blowing air on the compression connector to clean it. There was also some fuzz (not the gold fuzz stuff) on it that I carefully removed with a needle. I did not touch the gold fuzz.

[TruHobbyist]

Could you post a cat /var/adm/crash/analysis.N where N is the last (highest) number?

[nater1217]

The file is attached. This is from a few months ago when I was messing with it on a different install of Irix. It looks like I didn't wait for it to make the file the most recent time I was messing with it.
Also just as a reminder, I don't even get to this point anymore. It panics as soon as it tries to boot

[TruHobbyist]

This:

9 mgrasReleasePoolUserMemory[../graphics/MGRAS/mgras_dma.c: 584, 0xa80000002036d778]

may indicate non-working DMA, which is bad (no data transfer to/from the board).

And this:

<0>PANIC: KERNEL FAULT
    PC: 0xa80000002036d778 ep: 0xffffffffffffbc48
    EXC code:128, `Software detected SEGV '
    Bad addr: 0x3e8, cause: 0x8<CE=0
    <6>

may indicate an offset (0x3e8) into some struct field with a base address of 0 (NULL pointer <-). My educated guess: base addresses in a driver that are [not properly initialized] *AND* that are [not expected to fail] (these ones would get caught early on if the programmer really expected them) would be another symptom for a hardware fault.

Hopefully you can get a cheap replacement board in the forum and bring your Octane back to life.

[vishnu]

Trace stops because bad PC 0xfffffffffffffff9 found

Something is really confusing your system, fffffffffffffff9 in base 10 is 18,446,744,073,709,551,609 !

[nater1217]

Well I tried a Hail Mary by taking off the compression connector from the board and reattaching it. After doing that, I'm back to the original kernel panic and I can at least boot.

The analysis file looks pretty similar, same cause, same BADVADDR. I attached it, but I don't think there's any new info in there. I already know the answer is get a new graphics board. 

Can I at least install one of the TRAMs on my SE board? They are part number 030-1263

[vishnu]

The debugger is pointing out exactly where it's crashing, line 640 of this file, which presumably permits the MGRAS system direct memory access, which is to say, access to RAM that bypasses the CPU, but it's only meaningful if there's someone at SGI who can take a look at it and try to figure out what's going wrong. Sadly, those days are long gone. 

../graphics/MGRAS/mgras_dma.c: 640

[Raion]

The best thing to do would be to have someone decompile that object file with matching C, but that just points to perceived damage on the card and not a fault with the operating system.

You can see that they kept the debug symbols in.

As for the OP, yes, if your cards are mxe and sse then the parts will interchange. If one is an "i" and not an "e" then no, there's no luck there.

Hold on to the board because you never know when somebody can come along and diagnose it better.

If you can't hold the trash part then I would be willing to take it if you're in the United States for a small payment towards you, though I would ask you to do it through the marketplace so that we both can get feedback and that we can get more data about prices

[vishnu]

The best thing to do would be to have someone decompile that object file with matching C, but that just points to perceived damage on the card and not a fault with the operating system.

You can see that they kept the debug symbols in.

As for the OP, yes, if your cards are mxe and sse then the parts will interchange. If one is an "i" and not an "e" then no, there's no luck there.

Hold on to the board because you never know when somebody can come along and diagnose it better.

If you can't hold the trash part then I would be willing to take it if you're in the United States for a small payment towards you, though I would ask you to do it through the marketplace so that we both can get feedback and that we can get more data about prices

I don't have any of my SGIs up at the moment so this only amounts to conjecture, but running strings and piping it to grep on any of the error messages show by the debugger, on the Xsgi binary could possibly be illuminating.

[nater1217]

Its been a year, but I just wanted to update I recently got another Octane that had an SSE board in it and its showing the exact same crash... MgrasDownload: error in downloading GE ucode. It's blowing my mind that I have 2 boards with the same issue.

Now before attempting an Irix install on this new board, I put the TRAM modules from my board onto it to make it an MXE, but once I did that I just got a red light and no boot at all. I should've attempted to install it as an SSE, because now I'm wondering if a bad TRAM has killed it.

I put one of the TRAMs on an SE board, and also got a red light. Thankfully after removing it, everything appears to be working again.

These darn Octanes are so fragile.