[vfio-users] RmInitAdapter failed for nvidia in ubuntu guest

[Cepheus]

Am 30.06.2016 um 18:39 schrieb Alex Williamson:
> I don't see that you're doing anything about hiding KVM.
> 
> <features>
>   ...
>   <kvm>
>     <hidden state='on'/>
>   </kvm>
>   ...
> </features>


Thank you. Unfortunately, the driver still fails to initialize. The vm
"test9" did not start again after I changed kernel params "vmalloc=512M"
and later reverted the change.
A new vm "test10" with the above kvm "hidden" state behaves nearly
identical. There is not "not supported by ..." message in Xorg.0.log.
But the following:

[    11.106] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA GPU at
PCI:0:5:0.  Please
[    11.106] (EE) NVIDIA(GPU-0):     check your system's kernel log for
additional error
[    11.106] (EE) NVIDIA(GPU-0):     messages and refer to Chapter 8:
Common Problems in the
[    11.106] (EE) NVIDIA(GPU-0):     README for additional information.
[    11.106] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA
graphics device!
[    11.106] (EE) NVIDIA(0): Failing initialization of X screen 0


The dmesg messages look similar. The numbers are different:

[    4.376541] NVRM: RmInitAdapter failed! (0x2c:0x59:1159)
[    4.376913] NVRM: rm_init_adapter failed for device bearing minor
number 0

Additional questions: With "test9", I did have a login prompt in the
machine's virt-manager window on the host desktop. This was due to the
console device. With "test10", I also did leave the console device in
there, but I only get some text junk instead of a login prompt. This is
no showstopper - I enabled ssh in the guest - but I would like to know
what the reason for the difference is. Guest is Ubuntu 16.04 in both cases.

If the guest does not come up - no ping reply, no ssh - what is the best
way to know what's happening? I do not want the guest to have SPICE
graphics to not confuse the guest which display to use - output shall be
the passed through nvidia card.

Thanks
Kai