Failed to load GLX after new hard drive

[equimax]

The following installation was working fine for me:

Freebsd 4.8
VGA JATON/GF2 MX400 32MB PCI RTL
Nvidia Driver patched and installed.
KDE 3.1

Then I had a hard drive crash. I run two hard drives: Drive two backs up drive one every 24 hours using an rsync routine. Normally, when drive one fails, I can just make drive two the master drive, tweek a few things and I am off and running. This time when I did this, KDE won't start (startx from command line.) It reports

"Failed to Load GLX"

Followed by some reports of missing libraries.

I'm not sure what to do to figure this out. My rsync routine does not copy /tmp directories, so if the NVIDIA driver is looking for a /tmp directory that is not there, that could be the problem.

Anyway, your ideas would be greatly appreciated.

I have not uploaded any log files. I would be happy to do that. However, if you would like to see a log, please give me some step by step instructions to generate it, or tell me where to find it. I am new at this stuff.

Thnx!

[SirDice]

Why are you still running 4.8?

[equimax]

Why are you still running 4.8?

Long Story. Why, are you just curios?

[SirDice]

Long Story. Why, are you just curios?
4.8 isn't supported by the freebsd team anymore. 4.11 (the last 4.x release) will be EOL at the end of this month. So I'm kind of curious why you're still running such an old version :D

[equimax]

QUOTE=SirDice]4.8 isn't supported by the freebsd team anymore. 4.11 (the last 4.x release) will be EOL at the end of this month. So I'm kind of curious why you're still running such an old version :D[/QUOTE]

I am a small business operator who maintains a web site and a series of office computers while serving customers, paying taxes, and doing all the other stuff that a business operator does. My FreeBSD machine is a development machine for my current web site. Much of my development time is now taken up with creating a completely new web site which will do away with the need for the FreeBSD development machine. Meanwhile, I would like to get the FreeBSD machine working again for use in the short term. I have better things to do with my time than upgrade to the latest bells and whistles if the old ones are working OK. In this case the old machine was working fine until the hard drive failure described. (nana2)

So, having said that, do you have any helpful suggestions?

[SirDice]

You could just try to reinstall the nvidia-driver, I'm assuming you already tried that? Anything in /var/log/Xorg.0.log? Or is that oldy still running XFree? Not sure when they've moved from XFree to Xorg.. Have a look at it's logfile. Missing module dir mentioned in there?

/tmp does contain some needed directories. .ICE-unix and .X11-unix both dirs should at least exist and have their sticky bit set.

[equimax]

Thanks so much for taking the time to give me your suggestions!

You could just try to reinstall the nvidia-driver, I'm assuming you already tried that?

Yep. Good suggestion though.

Anything in /var/log/Xorg.0.log? Or is that oldy still running XFree? Not sure when they've moved from XFree to Xorg.. Have a look at it's logfile. Missing
module dir mentioned in there?

Tis XFree in my case. I looked at the log and couldn't see anything that I know to be out of whack. I was going to attach a copy of the file for you to look at, but I discovered that my Samba server is not loading correctly either. (I access this forum from my Windows box.) I am beginning to think that I have more problems than first met the eye. In fact, from looking at the log, it looks like the Nvidia driver is loading. GLX is failing, but that may have been true all along.

/tmp does contain some needed directories. .ICE-unix and .X11-unix both dirs should at least exist and have their sticky bit set.

.X11-unix was there. I added .ICE-unix, but that did not change things.

Looks like I may have to punt. There seems to be more wrong than I thought.

S

[SirDice]

Can you ssh to the box from your windows machine? If so, you could copy the files using winscp. I use it all the time in combination with putty.

[equimax]

Can you ssh to the box from your windows machine? If so, you could copy the files using winscp. I use it all the time in combination with putty.

Cool! Thanks for putting me on to winscp. :) I have attached the Xfree log file.
Thanks for sticking with me on this. Looking forward to your thoughts.

[SirDice]

A few things I noticed..

(II) LoadModule: "nvidia"
(II) Loading /usr/X11R6/lib/modules/drivers/nvidia_drv.o
(II) Module nvidia: vendor="NVIDIA Corporation"
compiled for 4.0.2, module version = 1.0.4365
Module class: XFree86 Video Driver

(II) NVIDIA XFree86 Driver 1.0-4365 Wed May 28 09:23:41 PDT 2003
(II) NVIDIA Unified Driver for all NVIDIA GPUs

It does load the nvidia-driver, an old one but if it worked before it should still be ok.

Just before it fails to load GLX:
(WW) NVIDIA(0): Failed to verify AGP usage

This makes me think it might be AGP related. Did you compile the nvidia-driver with or without freebsd_agp? Is AGP in your kernel? Is it loaded as a module? On 5.x and higher this is done in /boot/loader.conf. A line agp_load="YES" will load the freebsd AGP module. IIRC 4.x should work the same but I'm not sure if agp is build into the generic kernel of 4.x though.

[equimax]

It does load the nvidia-driver, an old one but if it worked before it should still be ok.

Agreed.

Just before it fails to load GLX:
(WW) NVIDIA(0): Failed to verify AGP usage

This makes me think it might be AGP related. Did you compile the nvidia-driver with or without freebsd_agp? Is AGP in your kernel? Is it loaded as a module? On 5.x and higher this is done in /boot/loader.conf. A line agp_load="YES" will load the freebsd AGP module. IIRC 4.x should work the same but I'm not sure if agp is build into the generic kernel of 4.x though.

Hmmm.... interesting thought. I checked the current kernel and the line

Device agp

is in it and not commented out. No other reference to agp though. I added the line agp_load="YES" to my /boot/loader.conf file and rebooted to see what would happen. Unfortunately, this did not change the problem or the XFree86.0.log file errors. I have extensive notes from when I installed the NVIDIA driver originally and there is no mention there of having to do anything about AGP. My video card is not an AGP card.

So, I remain stumped. My gut is telling me that the problem must be related to something that did not copy to the backup hard drive such as a file or directory. I have done this kind of hard drive switch in the past without this kind of problem, although I can't remember if I have done it since I installed this video card and driver.

I really appreciate your time and effort on my behalf. If anything else occurs to you, please let me know.

S

[adamk75]

First, can you confirm that the nvidia kernel module is getting loaded?

Second, do you see any libglx* files in /usr/X11R6/lib/modules/extensions?

[equimax]

Thanks so much for jumping in on this question.

First, can you confirm that the nvidia kernel module is getting loaded?

Hmmm.... sorry to be a little vague on what this means. As stated earlier in this thread, the basic NVIDIA driver seems to be loading, then failing or at least Xwindow/KDE is failing. By the "nvidia kernel" do you mean something that is loading in the FreeBSD kernel or separate from that? How would I investigate to find an answer for you?

Second, do you see any libglx* files in /usr/X11R6/lib/modules/extensions?

Yes,

libglx.so@
libglx.so.1

thanks again for helping with this.

S:)

[adamk75]

You can run "kldstat" to get a list of loaded kernel modules. You should see one called 'nvidia'.

What is the file size of the libglx.so.1?

I've downloaded the 4365 driver from nvidia's website, and it looks like it should be 681008 bytes.

Have you tried reinstalling the 4365 driver? You can download it from:

http://www.nvidia.com/object/freebsd_1.0-4365.html

[equimax]

Thanks so much for the additional ideas.

You can run "kldstat" to get a list of loaded kernel modules. You should see one called 'nvidia'.

Thanks for this instruction. I ran kldstat and yes, nvidia.ko is in there.

What is the file size of the libglx.so.1?
I've downloaded the 4365 driver from nvidia's website, and it looks like it should be 681008 bytes.

Yes, the size agrees, 681008.

Have you tried reinstalling the 4365 driver? You can download it from: url]http://www.nvidia.com/object/freebsd_1.0-4365.html[/url]

I already have this TAR file. Before starting this thread, I did a make clean followed by a make setup. Then I checked to be sure the appropriate Device Section showing Driver "nvidia" exists in XF86config. Then I checked to be sure the kernel contains the needed USER_LDT support. I did not reinstall the driver patch, but it was installed originally.

The more I get in to this, the more I am suspicious that the problem is not the NVIDIA driver and the failure to install GLX has been there all along. Maybe I am simply not using GLX. I'm not even sure what it is.

The KDE failure message says "Could not start KDEinit, check installation." Previous experience suggests that this message can mean lots of different things. Maybe this is a KDE problem, not a video driver problem.

What do you think?

S:confused:

[adamk75]

I have to agree. It seems unlikely that the failure to load the glx module would cause KDE not to work properly. You can always try gnome or some other simple window manager to see if X will start with those.

I'm actually beginning to think that this might be related to not copying /tmp over to the new drive. Maybe there's something different about the permissions on /tmp on this particular machine and KDE can't create the necessary sockets.

Adam

[equimax]

I have to agree. It seems unlikely that the failure to load the glx module would cause KDE not to work properly. You can always try gnome or some other simple window manager to see if X will start with those.
Adam

Well, your input caused me to persist until I found the problem. I went back and checked my notes concerning new hard drive startup using my backup drive. I have a list of directories that are not backed up and need to be re-created. I found that I had failed to create the /var/run directory. I created it, rebooted, and KDE fired right up.

Talk about feeling stupid. I did learn some things with this exercise, but I'm sorry to have wasted the time of some great people with this thread. Thanks to all of you. Your offers of help have been greatly appreciated. Without your input, I might have given up and never found the problem.

Seth

[SirDice]

My video card is not an AGP card.
That would explain the failure to verify AGP ;)

If your system doesn't even have an AGP slot or if you don't use it you can safely remove the device agp from the kernel config.

Good to hear everything is working :)

One thing I've learned over the years, it's no use making backups if the backups don't work. Test it before you really need 'm ;)

[equimax]

That would explain the failure to verify AGP ;)

If your system doesn't even have an AGP slot or if you don't use it you can safely remove the device agp from the kernel config.

Good to hear everything is working :)

One thing I've learned over the years, it's no use making backups if the backups don't work. Test it before you really need 'm ;)


Thanks again for the help and advice. You are so right about backups. In my experience even backups that are tested and work may subsequently fail for a variety of reasons. The solution is redundancy. If one backup fails, at least you have a second or third chance.

S