Control: severity -1 normal
Control: tags -1 moreinfo

On Sun, 2016-07-10 at 20:19 +0200, Peter Keel wrote:
> Package: nvidia-kernel-dkms
> Version: 367.27-1
> Severity: serious
> Justification: 6
> 
> Dear Maintainer,
> 
> Installing a new kernel 4.6.3 (self-compiled), and installing the
> experimental nvidia-drivers a few days later, led to the drivers
> compiled without module versions. 
> 
> 
> Commandline used: 
> dkms install nvidia-current -v 367.27 -k 4.6.3 --force --verbose 
> 
> Note that this was working on the earlier 4.5.7 kernel for which the
> symbols were generated. So I expect either compiler or compile-chain
> to be the culprit. 
> 
> 
> What happened is this: 
> 
> /lib/modules/4.6.3/updates/dkms# modprobe --dump-modversions 
> nvidia-current.ko 
> /lib/modules/4.6.3/updates/dkms#
> 
> There are no versions. Normal modules (ALSO dkms ones, for the same kernel)
> look like this: 
> 
> /lib/modules/4.6.3/updates/dkms# modprobe --dump-modversions 
> compat_xtables.ko 
> 0x83ca59d8    module_layout
> 0x449ad0a7    memcmp
> /lib/modules/4.6.3/updates/dkms#
> 
>    * What was the outcome of this action?
> 
> [   68.125562] nvidia: no symbol version for module_layout
> 
>    * What outcome did you expect instead?
> 
> [  978.452077] NVRM: loading NVIDIA UNIX x86_64 Kernel Module  367.27  Thu 
> Jun  9 18:53:27 PDT 2016
> 
> (This I actually did by allowing the kernel to force-load modules; and to 
> load the same unversioned module via "modprobe --force" -- not a 
> sustainable solution. 

I am running the same driver with 4.6 from Jessie backports, and I'm not
seeing that problem. Are you sure it's not due to the custom kernel?

Could you please try with the kernel from the repositories so that we
can narrow it down a bit more?

Kind regards,
Luca Boccassi

Attachment: signature.asc
Description: This is a digitally signed message part

Reply via email to