How to install PGX32 on Sparc 10 Ultra?

I'm trying to install a PGX32 video card on my Sparc 10 Ultra running Solaris 10.

I've got the Oracle installation guide for it and I got as far as:

"4. Insert the CD-ROM labeled �GFX OpenWindows for Solaris 2� into the drive."

I don't have a CD-ROM by that name and a web search for that turned up nothing. :confused:

Further down in the guide, they offer the helpful advice: "To complete an installation from FTP files, you must first obtain the packages from Sun. Contact SunService for the correct packages and the site where the packages are located." So that's not going to work either.

I did an fbconfig -list but all that returned was /dev/fbs/m640 which is the onboard PGX graphics. The monitor connected to the PGX32 is displaying nothing.

Can someone help me get this thing going? Pretty please? Thanks!

Here's a link to all the current Oracle manuals for that old system.

Ultra 10 Workstation Product Library Documentation

But reviewing this manual:

Indicates what versions of the OpenBoot prom and which device drivers are required. It then gives the entire procedure to get that card working. If you need any files, getting them from Oracle will be pretty hard unless you have access to their support network thanks to your employer. You'll probably have better results with Google.

Document intsrdb/20238

---------- Post updated at 08:59 PM ---------- Previous update was at 08:43 PM ----------

Download SUN Driver Updates

Yup - I've been through that document. Since I don't have an employer, that route is out. I've been all over the Oracle website and while they have extensive documentation, they don't seem to offer anything for download except Solaris iso's. I searched Google high and low but found nothing. That's why I was hoping someone here might be able to help.

# pkginfo | grep TSI

returns nothing, telling me the files I need aren't installed (that would have been nice).

Anyway, I do appreciate the replies.

Yes, but, the PGX32 is not really a Sun product.

So chase the drivers down elsewhere (under it's real model name).

pgx32

---------- Post updated at 10:05 PM ---------- Previous update was at 09:05 PM ----------

Another link that might help...........

PGX32 card problems

1 Like

Hah - I'll admit I hadn't thought of that idea. The card is really a Techsource Raptor GFX-FP. So I went to Techsource dot com and found packages there for a GFX-8P (not clear if that's the same thing) for Solaris 8 and 9 but not 10. No inidcation of whether they're for Sparc or x86 either. But I downloaded them anyway on the off chance something there might work for me. I also sent them an email requesting help.

That link to netbsd.org opens a nice description of the pm2fb driver for PGX32 but I don't see anywhere to download it.

I did find pb2fb.c and pm2fb.h for Sparc64 over at JK's Raid (the poster still won't let me insert URL's), so I guess that's something. But now I have to compile this. I'm more of a Lisp programmer than a C programmer. I guess that will be my last resort if the techsource stuff doesn't pan out.

Thanks for the pointers.

I've been looking around various forums for this problem.

Some say that the drivers are on the Solaris 10 distribution CD/DVD in the guise of packages TSIgfx and TSIgfxw. Also mentioned are TSIpgx and TSIpgxw.

Can you try:

 
pkginfo <package name>

for all 4 of these to see if any exist on your system.

Could also try:

pkgadd <package name>

to try to install them.

Also, do note that you will need to reconfigure the kernel after installing the driver to get Solaris to look for it.

 
boot -r

It also seems to be the case that if the driver for PGX32 is installed you'll have /dev/fbs present.

Thanks very much for the reply. All excellent suggestions. Unfortunately, the drivers are nowhere to be found on my system. However, in a stroke of amazing luck, I just got a very nice reply to the email I sent from a gentleman at techsource who said they removed the drivers from their web site due to lack of interest. He put them back up temporarily just for me, so now I've finally got the official files I need to do this install. :slight_smile:

I have some other stuff to do today but I'll report back with the reaults once I get going on it.

Good news Michele. It's like they always say "Where there's a shell there's a way"

Well looks like I spoke too soon. This is going from bad to worse. I installed the drivers from Techsource and now I can't even start the gnome windowing system. If I let the machine boot, I end up at a blank X-windows screen with an "X" cursor. If I login with root, stop the gui login with ^C and do a gfcconfig -i, it says "No valid TSI devices found! Exiting".

I may have brought this on myself because during the driver install it complained about some other mystery package I don't have called "SUNWowrqd " being required.

So now I have a real mess on my hands. How do I get back to just being able to login to the Sun Java desktop?

The easiest way to go back is to boot from cd and restore the filesystem from backup but, before you do that, let's try and move forward.

Firstly, after putting the driver on, did you tell the kernel to reconfigure?

 
boot -r

Previous discussions on the web indicate that the new graphics adapter won't be seen until the kernel reconfigures.

( you could also add the 'verbose' switch to the boot command to see if it tells you anything you didn't already know as it boots.

 
boot -rv

)

I assume that you now have two graphics adapters in the system? Are you saying that, if you connect monitors to both adapters, that you don't get the GUI on either?

One of the links I provided previously explained how to tell the service processor about the new adapter and switch off the old one. (I assume the old one is on-board so can't be taken out?)

Remember to try kernel reconfigure after each of these (no harm in repeatedly telling Solaris to reconfigure).

Try installing the missing package from the distribution CD/DVD. It might just be on the media. If not Google for it. Post if you can't find it.

Good suggestions. Here's the details - this install was apparently doomed form the start because while executing the GFX install_all script I got this message:

Installing Raptor GFX-C System Software/Device Driver v1.4 as <TSIgfc>

## Installing part 1 of 1.
/a/platform/sun4u/kernel/drv/gfxc
[ verifying class <none> ]
## Executing postinstall script.
No such file or directory
Cnnot access file (/a/etc/minor_perm).

Failed add_drv!

pkginstall: ERROR: postinstall script did not complete successfully

Installation of <TSIgfc> on zone <bristol-jes> failed.
## Restoring state of global zone <bristol-jes>

I continued on anyway in the hope that everything would somehow magically sort itself otu but of course it didn't. I did a boot -r but that didn't matter since the new device didn't get installed. A check of fbconfig -prconf confirmed that my only driver is the built-in 8 bit one.

So next I tried to address the missing SUNWowrqd package. I did find one on an old Solaris 2.3 CD I had lying around but when I went to do the pkginstall, it said all the files it was wanting to install were already in the system. So I aborted that. But what's odd is if I do a pkginfo | grep SUNWow, it does not find SUNWowrqd.

Anyway I then uninstalled the GFX drivers, rebooted again and at least now I can log back in to my Sun Java desktop.

So now I think the problem becomes why am I getting this "No such file or directory" error. I looked that up in Google and didn't find any answers. I'm getting ready to chuck this whole machine out the window.

"One of the links I provided previously explained how to tell the service processor about the new adapter and switch off the old one. (I assume the old one is on-board so can't be taken out?"

That's right - the PGX device is right on the mobo and the PGX32 occupies a PCI slot. The hardware knows it's there because

# prtdiag -v

Shows Bus# PCI-2 , slot 1, Name=TSI,gfxp, Model=GFXP

I'm sorry, I must have missed that link, but I think we need to get the new driver at least installed before going any further.

Please post what your system comprises in the way of zones.

Your post shows there are zones. Global + non-global?

Which zone(s) are you trying to install this driver on???

If you pkgadd from the global zone without the -G switch, it will try and install onto all zones. I think this might be your problem.

1 Like

You may be on to something here. I had never even heard of "zones" until I tried doing this installation. The system I have was configured by the prvious owner, so I've got whatever he set up.

"Which zone(s) are you trying to install this driver on???"

That's a very good question. I have no idea. I just typed "./install_all". How do I know what zone I'm in? Right now I think I'm in the Twilight Zone.

"If you pkgadd from the global zone without the -G switch, it will try and install onto all zones. I think this might be your problem. "

Well yes, there's no -G switch in the script. I'm doing it from the console logged in as root, if that makes any difference. You think I should try it again but with -G?

What does zoneadm list -cv show?

Zones make up a sort of Solaris Virtualisation environment similar to VMware. Are you familiar with that?

The global zone is the base OS and a non-global zone is equivalent to a guest OS on VMware.

A zone lives in a container, so zones and containers can usually be considered as the same thing (unless your configuration is unusual).

If you are accessing the console then, by default, you are logging into the global zone. If so, please post the output of:

 
zoneadm list -cv
 

That will tell us what you're looking at.

However, this type of virtualisation usually hosts multiple systems so trying to change the graphics adapter (on all zones) is unusual.

Please post why you are trying to do this. What is the app?

Depending on your response this could be getting tricky but there are some heavy-weight members on this forum who will contribute if they read this thread.

Certainly, the fact that your driver install error includes mention of zone is significant. The error about /a/etc/minor_perm access is common where zones are involved.

Please post as much of the story as possible so this community can help you.

Also, let me invite all Solaris expert members to "pile in". (Thanks Bartus11 - you beat me to this post).

1 Like
                                                 *********   S U C C E S S !!!   ************

Eureka! Finally at long last I've got my system running off the PGX32 board! In retrospect, it was quite simple - just a matter of knowing which magic spells to cast. This is what did it:

  1. I used the drivers in the file pgxOWv30_s9.tar on the techsource.com web site.
  2. Instead of using their install_all script, I typed in the pkgadd commnd myself, adding the -G switch, ie.(from a root console login)
    # pkgadd -G -d ./ TSIpgxx.u TSIpgx.u TSIpgxw
  3. Reboot with a boot -- -r, then STOP-a to get an OK prompt.
  4. From the OK prompt, do
    OK setenv pcib-probe-list 1,3
    OK reset

And hey presto, the system rebooted with the console coming up on my big monitor connected to the external PGX32 card!

Note that these drivers are clearly listed as being for "Solaris 8 and 9 only", but they work for 10 also (and just to be clear, this is Solaris for Sparc). They also talk about OBP 3.17, but I've got 3.25 and again, they work fine anyway. One thing that was not necessary was messing around with /etc/dt/config/Xservers. I created one of these according to some instructions I found and it really screwed up the system. There is no need to even have that file present to make this all work.

I think the first key to this puzzle was going to techsource for the drivers and the final one was using the -G switch in pkgadd. That caused the drivers to install cleanly without any errors. So many many many thanks to everyone who took the time to follow this thread and offer suggestions, and especially to hicksd8, who is now my hero. :):):slight_smile:

---------- Post updated at 01:46 AM ---------- Previous update was at 12:51 AM ----------

For completeness,

bash-3.00$  zoneadm list -cv 
 ID NAME             STATUS         PATH 
  0 global           running        / 
  - bristol-jes      installed      /data/zones/bristol-jes 

Brilliant! What a result. Well done!

You're obviously an IT professional, and ten out of ten for tenacity, but you may be distracted right now with all the effort so I'll say it anyway (not wishing to upset you), backup the whole thing asap.

You don't want to go through that again!!!

All the best.

"backup the whole thing asap"

Ha - you read my mind. That was the very next thing I was going to tackle. What I'd like to do is backup my entire boot drive such that I can restore it to a new disk in the event of a catastophic failure. I'm sure the drive in there now (a 9 GB 5 1/4" Seagate) is the original one that came with the machine. I probably shouldn't even say anything, but it's not clear how much more life is left in it. It's nearly full anyway so it would be nice to replace it with a larger one.

I have three possible avenues for backups: a mounted nfs file server on my lan, a working USB card on the Sun, and enough free space on my 40 GB secondary drive to do a backup there.

I don't know if this should be a new thread but if you have any suggestions, I'd very much appreciate hearing them.

(And I actually am an IT professional, or at least I was until I retired. I started using Sun workstations way back in the pre-Sparc Sun 4 days working for a large company in R&D. But we always had sysadmins to handle stuff like backups and installing new cards, so that part is all new to me).

Well I could write loads about this but I'm not going to. With your experience (you used to reading man pages) I only need to give you the clues. (That's what I've been doing up 'til now and it's all you needed).

Backup is a matter of choice but, given the options you list, I would go for the NFS remote mount. I assume, given the age of the machine, that your filesystems are ufs.

I would use ufsdump with switches 0f (meaning 0 for full backup and f to determine filename/pathname). This will create a file for the filesystem dump on the remote system. You can then run this file to tape (or copy it to a further system with a tape drive) and take it offsite.

To recover, in brief,

  1. Fit new drive in system.
  2. Boot from cdrom
  3. Use 'format' to set up filesystem slices and swap
  4. Use 'newfs' to write filesystems on the new disk slices.
  5. Mount the new (empty) filesystem (under /a)
  6. Use 'ifconfig' to manually plumb, set ip address, and mark as 'up' the network interface.
  7. Mount the remote nfs system containing the ufsdump (under /mnt).
  8. ufsrestore the filesystem(s)

If there is more than one filesystem on your system you can back them all up this way. You can still restore individual files if needed.

Tip: Get the vtocs (virtual table of contents; partitioning sizes in DOS terms) by running 'prtvtoc' on your system and recording the output. This is invaluable in a recovery situation.

Any other questions just ask.

Hope that helps.

PS: To be absolutely purist, you should use 'fssnap' to create a snapshot of your filesystem (this will output a device name for the snapshot), then mount that device read-only, and then ufsdump that read-only filesystem. If the system is not quiesent, then you're still covered. What you will be backing up would be a stationary snapshot and not a moving filesystem. See man pages and Google that.