[SOLVED] NVIDIA-Linux-x86-173.14.39-pkg1.run

[coreplayer2]

I've noticed the two config files which are listed in the failed test are totally different from each other.  While the set parameters of the config appear similar, each line has an additional textural prefix.. not sure if that matters?

[Juanito]

I understood that "test -e filename1 -a -e filename2"  should return true if filename1 AND filename2 exist. I don't believe the files need to be identical.

[Juanito]

I just loaded the test nvidia extension (I don't have nvidia hardware) with Xorg-7.7-3d.

If Xorg-7.7-3d is loaded first then the start-up script from the test nvidia extension will not overwrite the libGL and libglx symlinks.

..which is a good thing as Xorg-7.7 will not start if I manually point the symlinks to libGL.so.173.14.39 and libglx.so.173.14.39.

Loading the nvidia module (which I wouldn't expect to work without nvidia hardware) gives this:

Code: [Select]

$ sudo modprobe nvidia
modprobe: can't load module nvidia (kernel.tclocal/drivers/video/nvidia.ko.gz): unknown symbol in module, or unknown parameter

$ modinfo nvidia
filename:       kernel.tclocal/drivers/video/nvidia.ko.gz
license:        NVIDIA
alias:          char-major-195-*
alias:          pci:v000010DEd*sv*sd*bc03sc02i00*
alias:          pci:v000010DEd*sv*sd*bc03sc00i00*
depends:        agpgart
vermagic:       3.16.6-tinycore SMP mod_unload 486

$ lsmod | grep agpgart
agpgart                24576  2 drm,intel_gtt

[curaga]

dmesg will tell you which symbol.

[Juanito]

Indeed:

Code: [Select]

nvidia: Unknown symbol acpi_os_wait_events_complete (err 0)
..and so one more thing that needs patching

I see also that Xorg will probably not detect nvidia hardware and something like this is required:

Code: [Select]

$ cat /etc/X11/xorg.conf.d/20-nvidia.conf

Section "Device"
        Identifier "My GPU"
        Driver "nvidia"
EndSection

[Juanito]

The acpi patch is at:

http://tinycorelinux.net/6.x/x86/tcz/src/nvidia_driver/NVIDIA-Linux-x86-173.14.39-pkg1.run_acpi.patch

@Stone.Giant - the test extension with the recompiled nvidia driver is also reposted.

I've now gone about as far as I can without nvidia hardware:

Code: [Select]

$ sudo modprobe nvidia
modprobe: can't load module nvidia (kernel.tclocal/drivers/video/nvidia.ko.gz): No such device

$ dmesg | grep -i nvidia
nvidia: module license 'NVIDIA' taints kernel.
NVRM: No NVIDIA graphics adapter found!


[coreplayer2]

I understood that "test -e filename1 -a -e filename2"  should return true if filename1 AND filename2 exist. I don't believe the files need to be identical.

It was just a thought,  I assumed that was likely the case though hadn't found it in the script yet. meanwhile have been looking for clues as to why neither file was not at the path expected


Sent from my iPad using Tapatalk HD

[Juanito]

Both files were in the expected place in the prepared kernel source, that's why I'm puzzled as to why the test fails 

[coreplayer2]

I've created symlinks to the src files at every possible expected location, yet...     I'm missing something still !

I can extract the very latest nvidia sources, and build all driver modules yet the driver continues to segfault during loading..

Aaaarg!

[curaga]

@coreplayer2

Are you also using an old, unsupported driver? The 173 legacy driver was last updated 2013...

[coreplayer2]

@ curaga   A summary of hardware
Among my PC's I have 3 with Nvidia video cards,

My main PC uses one of the latest Maxwell GM204 chips (out for almost a year) supported by driver v346.72

Other PC's use older hardware which are supported by the latest legacy 340.76 (almost all older Nvidia hardware are supported by this Legacy  Linux driver)
Geforce GTX 560
Geforce GTX 260m



The legacy v340.76 is the driver we should be creating with kernel 3.16.6 to support all but the very latest Nvidia hardware.
The driver v346.72 supports the latest hardware http://www.nvidia.com/download/driverResults.aspx/81760/en-us

it is the Nvidia driver 346.72 in 3.16.6 on my latest hardware (which is fully supported by this driver) that fails the the config test, yet if built segfaults when installed

To assist, I have attempted to run Nvidia driver 173 on my oldest Geforce GTX 200 series hardware, so far without success.  While I can not find more info on the NVIDIA-Linux-x86-173.14.39,  I found NVIDIA-Linux-x86-173.14.30 info in the archives which supports the desktop version but not the mobile version of the hardware I have.  The driver usually reports if the hardware is not supported, but not so in this case.

[curaga]

Oh, if the latest fails, you should report the bug to nvidia.

[coreplayer2]

Yes have done that already.  Their latest 352.09  beta driver compiles and installs in tx-6 / 3.16.6 without error,  however the driver segfaults when Xorg is run

I wrote a script to gather all the log data after the driver crashes before the inevitable REISUB..  which I'm collecting now

Code: [Select]

[drm] Initialized nvidia-drm 0.0.0 20150116 for 0000:01:00.0 on minor 0
NVRM: loading NVIDIA UNIX x86 Kernel Module  352.09  Tue May 12 13:54:22 PDT 2015
nvidia-settings[19438]: segfault at b8c5d984 ip b8c5d984 sp bf9c5eec error 14


[Juanito]

I've uploaded the nvidia-340.76 driver to the same place if you'd like to test - note that I did not create symlinks to the nvidia libGL and libglx.

[coreplayer2]

Ok thx


Sent from my iPhone using Tapatalk