Hi,

 

since the 6.4 i still have several boot issues with a win10 VM with nvidia gpu passthrough.

 

mostly coming when there was a win update or i did a manual reboot.

 

VM then starts up and stays on "Memory tested ok" page ... and often restarts the memory testing and always ends up there ...

mostly a full reboot from the server resolving this, today morning now not, but i ll take a look further later if i can get it back without

reverting back to 6.3.5.

 

the 2nd win10 VM with intel gpu passthrough i never experienced this.

 

screen from system log doesnt really seem to help (at least to me )

 

and from the VM log

 

2017-09-14 04:36:41.001+0000: starting up libvirt version: 3.5.0, qemu version: 2.10.0, hostname: AlsServer
2017-09-14 04:36:41.001+0000: Domain id=2 is tainted: high-privileges
2017-09-14 04:36:41.001+0000: Domain id=2 is tainted: host-cpu
2017-09-14T04:36:41.053917Z qemu-system-x86_64: -chardev pty,id=charserial0: char device redirected to /dev/pts/1 (label charserial0)
2017-09-14T04:37:33.257990Z qemu-system-x86_64: terminating on signal 15 from pid 7621 (/usr/sbin/libvirtd)
2017-09-14 04:37:34.659+0000: shutting down, reason=destroyed
2017-09-14 04:37:37.048+0000: starting up libvirt version: 3.5.0, qemu version: 2.10.0, hostname: AlsServer
2017-09-14 04:37:37.048+0000: Domain id=3 is tainted: high-privileges
2017-09-14 04:37:37.048+0000: Domain id=3 is tainted: host-cpu
2017-09-14T04:37:37.085100Z qemu-system-x86_64: -chardev pty,id=charserial0: char device redirected to /dev/pts/1 (label charserial0)
2017-09-14T04:39:00.465732Z qemu-system-x86_64: terminating on signal 15 from pid 7621 (/usr/sbin/libvirtd)
2017-09-14 04:39:01.866+0000: shutting down, reason=destroyed

Edited September 14, 20178 yr by alturismo

Not sure this is terribly helpful, but I pass through an Nvidia 1050TI and have not observed this problem.

 

Are you using SeaBios or omvf?

 

Once this happens, is it repeatable until a server reboot? Might be some kind of locking issue on the disk image file.

i assume its more a locked PCI slot or so, cause i also cant boot then any other VM using the Nvidia card (libreelec for example), or i also cannot start a fresh install with nvidia ...

 

and i have a GTX1070, using OMVF Bios, i remember i used seabios before when i started with unraid, but last install was OMVF and worked flawlessly until 6.3.5,

issues started with 6.4x.

 

and yes, a clean reboot always solves the issue (shutdown, cold start or usually system reboot)

 

thanks anyway  next install i will try seabios again.

I also use OMVF. But I have one and only one VM with GPU passthrough.

 

Is it possible you tried to start two VMs attempting to pass through the sane GPU. Or stopped one and started the next very quickly?

nope, i use 2 mainly 2 VM´s.

 

1. win 10 intel GPU passthrough (autostarted) and always ok

2. win 10 NV GPU passthrough, manually started when 1st one is up and running, mostly running

 

my description above was only to show and test after the win10 NV GPU is not booting anymore ... then i can try whatever i want,

nothing will work with the Nvidia passthrough anymore until a reboot is done.

so to your question, no, this only occurs after a period of uptime usually ..

My guess is, somehow it stays locked under some circomstances ... like windows update reboot and ... no idea and i cant really reproduce.

most reboots here while installing or updatingare working fine ...

As its always on, some days in the morning the NV VM is just dead, and i cant reach it anymore and also, no reboot, no start, no ...

Edited September 15, 20178 yr by alturismo

Which GPU is primary? Are you loading the ROM file? My setup has only one GPU (the Nvidia) which is primary and uses the ROM file. My prior MB had onboard video and the Nvidia was secondary with no ROM file. Neither had issues like you are describing.

 

my primary GPU is the onbaord Intel (Kaby Lake).

 

and wich Rom File if i may ask ? ... i dont load any rom files seperately.

only remember rom files when setting up a openelec mashine ages ago, never since then ...

 

and this issue does not happen too often, just sometimes (especially after auto windows updates) the VM is unresponsive ...

and then theres no way to get it started again besides a complete server reboot ... when i manually update it never happened here ...

its a not a too big deal, just may someone has an idea ... why its stalling in memory test page forever with 100 % CPU usage ... and always

telling success and stays ...

The use of ROM files is explained in the Spaceinvader One (a.k.a., @gridrunner) video in passing through Nvidia primary GPU.

 

Not sure if required with onboard video, but maybe worth a try.