Commen=
t # 3
on bug 10794=
7
from Kent Ross
I have this issue, or one very like it. I have a system with b=
oth a Vega64 and
a GTX 980ti. The 980ti is controlled by the vfio-pci driver (preempting
nouveau, nvidiafb, etc. on boot) and is therefore not in use by this host
system.
If a display is plugged into the 980ti at boot time, booting will always fa=
il
with a black screen and more failures from drm. If a display is not plugged=
in,
booting will almost always work without a problem, with fewer errors from d=
rm.
Even with no display plugged into the 980ti, when the system's displays
(plugged into the vega64) resume when they have been sleeping (such as on t=
he
lock screen) there is a substantial chance all displays will be completely
black and drm failures can be seen in dmesg. If a display is plugged into t=
he
980ti when the main displays wake, the outcome is much less likely to be go=
od.
In the case where the system is returning from sleeping screens it fortunat=
ely
does not fully necessitate a complete reboot, as the displays will sleep ag=
ain
in a few seconds and it can be tried again until success. Nonetheless, as t=
his
system is intended to be used as a VM host with a passed-through monitor, t=
his
bug is a showstopper.
I have full dmesg logs from boot to an unusable black screen, and from boot=
to
a login screen.