Creating device files

[hans]

Hi, Im busy with the nvidia driver quite some time now and

I've found in the nv_check.sh script that a few /dev/nvidiactl
/dev/nvidia0
/dev/nvidia1
etc.
devices need te be present but I only have the first two.

Does anyone know why these are missing, are the other device files (>0) obsolite?

If i do need to create the device files, can someone give me an example. I do want to do this right the first time. I'm afraid to get lost if i start trying these things whithout knowing what i'm doing.

Something else that puzzles me is the driver name. I have nvidia.o, but the above testing script searches for NVdriver.o. Is this the same driver?

I also tried to compile the driver but make complains about the compiler revision. I'm just using what is supplied whit the distro (mdk9.0) may i assume this is right or do I need to compile the kernel also to get the same version?

:confused:

[LordMorgul]

nv_check.sh is old now, and may not work correctly, use the newer installer for version 43.49 from the nvidia driver archive. You should not need to create the device files, the OS will create them if the card is present and activated.

The Installer should take care of you, just run it from a virtual terminal (ctrl-alt-F1) with X stutdown ('telinit 3' to shutdown x, 'telinit5' to turn it back on), if you still have issues with the devices, post some additional information here like which OS version, card, system specs etc.

[bwkaz]

The /dev/nvidia* devices (other than ctl) are one for each nVidia card you have installed. If your distro uses devfs (you said Mandrake 9, right? that one does use devfs), then the device files for each card will be created when you load the nvidia.o kernel module.

And yes, nv_check is way out of date. It doesn't know about devfs, and it still uses the old module name (which is NVdriver, not NVdriver.o -- not that it matters much, but anyway).

[hans]

Thanx now i don't have to search for the drivers.

I think I will have to compile the driver since My system is AMD an there is only a source version for this processor.

First i will try to ignore the message about the compiler version since I don't like to change the C version etc. If this doesn't work then I can always compile my kernel to get the same version. (just one step at a time.)

Greetings, Hans

[hans]

I did the compiling with the ignore,
then i noticed there are two gcc version on my distro compiled it again without the ignore. This seems to work as far as the compiler version.

Anyone who can read compiler output? Maybe you could take a peek in the tar file. It isn't very usefull to me.


I have an AMD 1,7 with 1Gb mem
Asus V9180 Video
Mandrake 9.0

(jvbuiten#wanadoo.nl)

[hans]

I did the compiling with the ignore,
then i noticed there are two gcc version on my distro compiled it again without the ignore. This seems to work as far as the compiler version.

Anyone who can read compiler output? Maybe you could take a peek in the tar file. It isn't very usefull to me.

Yes it is make.tar.gz, rename it and both tar -xzf as winzip will open it (This is because of the "valid" extensions)

I have an AMD 1,7 with 1Gb mem
Asus V9180 Video
Mandrake 9.0

(jvbuiten#wanadoo.nl)

[bwkaz]

Well first off, 4180 is the wrong version to be using. 4180 is the AMD64 version -- you do not have an Opteron chip. They aren't even released yet. Your chip is 32-bit.

As for the other log, install your distro's kernel source package.

Why not use 4349? (you will still probably need kernel-source, but perhaps not)

What do cat /proc/version and gcc -dumpversion both say?

[hans]

>4349?

I did first use a normal rpm that mached my kernel version exactly
I attached the two logfiles, there you can see I also tried the 4149, which is intel.

I didn't know there was already a new version since i downloaded mine a few weeks ago, better check again.

Thanks for the amd tip, I seemed to have missed something there.

Still, since the 4139 version also supports my video processor i am still worried that a will get the same error again, let's see...

[bwkaz]

You tried to use 4180 and 4191. The newest version is 4349.

4139 and 4149 do not exist.

[hans]

> 4139 and 4149 do not exist
:( wrong button i guess...

module version = 1.0.4349,

Still the same error with 4349

(EE) LoadModule: Module nvidia does not have a nvidiaModuleData data object.
(II) UnloadModule: "nvidia"
(II) Unloading /lib/modules/2.4.19-16mdkenterprise/kernel/drivers/video/nvidia.o
(EE) Failed to load module "nvidia" (invalid module, 0)

I have attached the X11 config,
the 003c version is a working vesa config
the 004a version is changed with the nvidia driver. The result (logfile) of the 004a config file is included.

Greetings

(The zip file is a .tar.gz file)

[mtrr]

ModulePath set to/lib/modules/2.4.19-16mdkenterprise/kernel/drivers/video,/usr/X11R6/......


"/lib/modules/2.4.19....."


what the heck is that doing in your config file? that's the path to
the kernel drivers, nothing to do with X11. delete that line from
your config and try again.


cheers

[hans]

[hans@roger hans]$ locate nvidia.o
/lib/modules/2.4.19-16mdkenterprise/kernel/drivers/video/nvidia.o
[hans@roger hans]$ locate -i nvdriver.o
[hans@roger hans]$

That's where i found the driver...

Message not finished yet.

[mtrr]

the module (kernel driver) name has changed from NVdriver to nvidia.o

nothing to do with the X11 driver.

just look in nvidia's readme and delete the /lib/modules/2... line
from your XF86Config($whatever)


groetjes


ps: xfree appends an _drv.o to the name you specify as driver
name in the config file
maybe you can find it now somewhere below /usr/X11R6

hint hint: /usr/X11R6/lib/modules/drivers

[hans]

You were right, now the driver loads... at least that is what the logfile is telling me.

I just removed the line with /lib/modules/2.4.19-16mdkenterprise/kernel/drivers/video,/usr/X11R6/...... and now the driver seems to load. The result are some strange characters in the bottom-right of the screen. Then it stops responding to the keyboard. I managed to ssh to it but a shutdown also crashes.
(I think I saved myself a lot of trouble when I remembered to first edit inittab, runlevel 3)

I attached a new copy of the logfile
(.tar.gz file)

[mtrr]

look in a different thread for how to set AGP to a lower value than
8x, seems to be a known issue.

look in the errata file on the nvidia ftp server, there is further info.



cheers

[hans]

Oefff, I think i did it,

Added
Option "NvAGP" "0"

and now everything seems to work.
The driver loads, checked the logfile
and I am using 1152 dot + 16 bit color

Does this mean the AGP port cannot be used properly?

Greetings, Hans

[hans]

>look in a different thread for how to set AGP to a lower value than
>8x, seems to be a known issue.
>
>look in the errata file on the nvidia ftp server, there is further info.

I will look into that later, Thanks :)