[Raion]

I have uploaded a copy of the L3 software for Linux here: http://usftp.irixnet.org/sgi-tools/l3-em...nux.tar.gz

It apparently requires a very old version of Linux. If anyone could document what versions of Linux can run it, that would be interesting to see. I'm curious as to how far the flexibility of it goes.

[indy99]

I had it running for several years before a kernel update killed it. Pretty sure I installed around kernel version 2.6.12, and seems like 2.6.17 broke the USB wrapper SGI used to talk to the L1. The distribution was packaged for Redhat, which is what the re-badged SGI PCs it was intended to run on came pre-loaded with.

Pretty slick piece of software; even gives you a graphical interface that looks just like the L2 LCD display panel.

[jan-jaap]

I  anyone could document what versions of Linux can run it, that would be interesting to see. I'm curious as to how far the flexibility of it goes.

I think it needs RedHat 6. Not RedHat Enteprise Linux 6 (RHEL).

RedHat 6

If you want to be period-correct you get an SGI 230 PC to run it. They used those with the big multi-rack O3Ks where an L3 makes sense. These days it's probably better to run it in a dedicated VM.

[Raion]

I was sure it was of that vintage.

[mgtremaine]

Redhat 6.2 the super kernel.. You can run on alpha and sparc also. I have a box set somewhere.

-Mike

[jwhat]

Hi SGI'ers,

I have previously got the L3 and L2 emulator running on old IBM PC, but really wanted to get this going in VM.
My attempt to do this with VMWare failed as I could not get USB to work.

I have now managed to get this running on QEMU / KVM with Wacom USB Tablet for mouse (so you can use VNC to access it).

I have documented notes on how to get this running on QEMU / KVM here: https://tips.graphica.com.au/creaky-old-...ore-linux/

You can see I used Fedora Core R1 (yellow) which has a Linux 2.4 kernel which is required to work with SGI USB drivers.

A bit of archaeological digging required to get this running, only remaining step to to plug O350 L1 USB into server and see if I can capture it with the emulator.

Cheers from Oz,

John.

[Raion]

Hi John,

Thank you for your hard work.

[weblacky]

On the CD scripts, claims RedHat 6.2. That's on Archive.org...I know I have that CD set, last set I bought was Shrike Redhat 7 back in the early 2000s...

https://archive.org/details/redhat-6.2_release

I realize all consumer emulation really goes only back to Redhat 7, I cannot readily find an emulator/virtualization that claims full Red Hat 6.2 support. VirtualBox will do 7 support...that's the oldest I found. I'd probably just use an old desktop PC or something if I needed to keep this around. By yeah, a VM image would be great.

[jwhat]

Hi Weblacky,

Fedora Core R1 (yarrow) runs fine on physical machine with CD-IST-3.24 which was version I got via Nekochan link back in the old days.

My physical machine is IBM ThinkCentre with Pentium 4 with an Adaptec SCSI card, so it is around same era as the SGI software.

I just moved from spinning disk to SCSI2SD V6 and that works very well.

With emulation setup I have tried to get all the VM settings which make the machine look likes it of same era to ensure it works (hence using IDE, coreduo etc).
The critical configuration is USB, which must be set to USB 2 or machine fails during hardware detection on install boot.

Cheers from Oz,


John

[jwhat]

Hi SGI & Irixers,

I have played some more with getting L3/L2 working via QEMU / KVM.

I reached the point where I have O350 L1 USB plugged into my QEMU / KVM host, but the L1 is cycling up and down on its connection, which means it is not stable for long enough for me to re-direct the USB Device to the Fedora Core L3/L2 VM.

You can see cycle via the dmesg :

>> $ dmesg | grep usb
>> [ 1777.251564] usb 3-3: USB disconnect, device number 95
>> [ 1779.561037] usb 3-3: new full-speed USB device number 96 using xhci_hcd
>> [ 1779.712779] usb 3-3: New USB device found, idVendor=065e, idProduct=1234, bcdDevice= 1.00
>> [ 1779.712783] usb 3-3: New USB device strings: Mfr=1, Product=2, SerialNumber=3
>> [ 1779.712785] usb 3-3: Product: SN1 L1 System Controller
>> [ 1779.712788] usb 3-3: Manufacturer: Silicon Graphics, Inc.
>> [ 1779.712789] usb 3-3: SerialNumber: 00000000
>> [ 1781.291231] usb 3-3: USB disconnect, device number 96
>> [ 1783.601071] usb 3-3: new full-speed USB device number 97 using xhci_hcd
>> [ 1783.756886] usb 3-3: New USB device found, idVendor=065e, idProduct=1234, bcdDevice= 1.00
>> [ 1783.756890] usb 3-3: New USB device strings: Mfr=1, Product=2, SerialNumber=3
>> [ 1783.756892] usb 3-3: Product: SN1 L1 System Controller
>> [ 1783.756895] usb 3-3: Manufacturer: Silicon Graphics, Inc.
>> [ 1783.756897] usb 3-3: SerialNumber: 00000000
>> [ 1785.331243] usb 3-3: USB disconnect, device number 97
>> [ 1787.645129] usb 3-3: new full-speed USB device number 98 using xhci_hcd
>> [ 1787.796866] usb 3-3: New USB device found, idVendor=065e, idProduct=1234, bcdDevice= 1.00
>> ...
>> ...
>> ...

So thought maybe if I build the sgil1 USB driver (it is included on the IST CD with GPL) and installed it on the QEMU / KVM host then this could take ownership of L1 and then I could re-direct this to the VM.

So to cut a long story short I have hacked the sgil1 code and scripts (see github here: https://github.com/zebity/sgil1 ) and now have sgil1 driver loading on Ubuntu 20.04 (linux 5.4) host:

>> $ dmesg | grep usb
>> [ 0.440280] usbcore: registered new interface driver usbfs
>> [ 0.440280] usbcore: registered new interface driver hub
>> [ 0.440280] usbcore: registered new device driver usb
>> ...
>> ...
>> ...
>> [47347.049644] usbcore: registered new interface driver sgil1
>> [48846.445478] usbcore: deregistering interface driver sgil1
>> [53621.919319] usbcore: registered new interface driver sgil1

The kernel complains when you do modprobe to load the KLM, due to it being unsigned but otherwise it loads up ok.

I did my initial build and test with Ubuntu 20.04 development VM, but then copied compiled KLM over onto test server and powered up O350 and hence L1.
I can see that that driver loads and connects to the L1 and reports a lot of configuration information, but then L3/L2 VM appears to freeze and then entire machine becomes partially responsive.

I could not get the KLM to unload and could not stop the L3/L3 VM and when I did reboot of machine the reboot never completed, so I had to power it off via the BMC controller (which is Intel server equivalent of L1 controller).

While doing this I had a few other ideas:

1. Could you avoid loading sgil1 altogether by just getting right udev rules for the L1 devices ?
2. Does introduction of virtualisation mean the sgil1 code needs to have some extra "things" put into code to allow it to work /
(I ask as both VM and the host behaviour was affected
3. Maybe just throw away the SGI code and write a very simple dummy plug driver ?
4. Write udev script so that on L1 connect it automatically redirects usb to VM ?

Anyway I am posting this as it currently is, as it might be some SGI'ers who also have much more in-depth knowledge of Linux kernel and USB Driver layer than me so maybe we could collaborate via GitHub code base.

Future thoughts ... if we can get the sgil1 driver going on modern Ubuntu then it opens up an new world of possible ways to more easy manage the SGI's which have L1 (ie O3000, O300, O350, Tezro & Fuel).

Cheers from Oz,

John.