Bus Error (nVidia)

I'm in the process of rebuilding my multimedia player Ding. I upgraded the Kernel (2.6.18) and the OS (Debian testing), but have a strange problem with building the nvidia drivers:

ding:/usr/src# sh NVIDIA-Linux-x86-1.0-8774-pkg1.run
Verifying archive integrity... OK
Uncompressing NVIDIA Accelerated Graphics Driver for Linux-x86 1.0-8774...
NVIDIA-Linux-x86-1.0-8774-pkg1.run: line 926:  8406 Bus error               $script $scriptargs $*

I havn't the slightest idea what causes this. I also tried older driver packages with the same result. An strace brought the following output:

gettimeofday({1160596514, 741277}, NULL) = 0
getpid()                                = 6913
open("/tmp/nv-VUiCVZ", O_RDWR|O_CREAT|O_EXCL, 0600) = 4
lseek(4, 61404, SEEK_SET)               = 61404
write(4, "\0", 1)                       = 1
mmap2(NULL, 61405, PROT_READ|PROT_WRITE, MAP_SHARED, 4, 0) = 0xb7e00000
--- SIGBUS (Bus error) @ 0 (0) ---

So this mmap2 call seems to cause a SIGBUS signal, but what can I do to fix this? Googling didn't brought up anything useful :-(

If anyone has an idea please post a comment.

Tags:: nvidia,; linux,; troubleshooting

Similar posts:

Posted on Wednesday, October the 11th 2006 (5 years ago).

As an additional note it may be interesting to know, that the root file system of the PC is mounted via NFS (using the NFS kernel server). Some googling told me that this might make trouble with the mmap call, but I find this strange because it did just work fine before (Kernel 2.4.17 I think).

2006-10-14 11:53:38

Andi

Okay, just for the record: I finally solved the problem by downgrading the Kernel to 2.6.17.14.

2006-10-15 23:21:29

Andi

Andi, thanks for the information - I was running into the same situation. Exact same scenario, too - decided to upgrade my mediabox, which is set up very much like yours. This post saved me a few hours of frustration - downgrading to 2.6.17.14 worked perfectly. Thanks!

2006-11-02 18:30:00

James Baicoianu

As James said, thank you for the info. This also solved my exact same problem. Most people never post the final fix action, but thankfully you did! :)

2007-01-12 00:05:21

Paul

I had the very same problem and managed to solve it. Perhaps this way works for other people too.
It should work when your server and client have the same kernel (2.6.19.2 in my case) and the server is able to run the client's binaries (both x86 in my case).

1) Login to your NFS server
2) mount -o bind /proc /export/proc (where /export is / for your NFS client)
3) chroot /export
4) sh NVIDIA-Linux-x86-1.0-*-pkg1.run -x
5) cd NVIDIA-Linux-x86-1.0-*-pgk1
6) ./nvidia-installer
(this might warn about a different GCC version installed than used for the (NFS server's) kernel, ignore that)
This will crash when trying to load the nvidia kernel module (if your server has no nvidia graphics card).
7) cp usr/src/nv/nvidia.ko /lib/modules/2.6.*/kernel/drivers/video
8) [on the client!] modprobe nvidia
(this should work!)
9) [on the server!] ./nvidia-installer --no-kernel-module
(this should install everything apart from the already obtained kernel module)

Done!

Please comment via E-Mail.

Bye,
Carsten

2007-01-14 20:11:37

Carsten Otto

Step 8 should include a "depmod -a" before the modprobe.

2007-01-14 20:30:05

Carsten Otto

I tried the above technique and got it to work with kernel 2.6.19-1 (with fedora patches). Way to go

2007-02-05 09:01:06

Max Behensky

This seems to be fixed in the latest kernel, 2.6.20. I just ran the stock installer and it worked, whereas with 2.6.19 ir required the above song and dance.

2007-02-06 07:59:58

Max Behensky