It seems that with the 6.4.0-rc8a/rc9f prerelease, there are issues with a GPU being bound to the host that were not there with 6.4.0-rc7a.

 

This results in various failed to write to memory messages in the log, assuming because the ROM file cannot be passed to the GPU properly, and the well-known error 43 in Nvidia's driver under Windows.

 

This can be fixed without restarting or blacklisting the GPU in the kernel with the following commands prior booting the VM:

 

Quote

echo 0 > /sys/class/vtconsole/vtcon0/bind
echo 0 > /sys/class/vtconsole/vtcon1/bind
echo efi-framebuffer.0 > /sys/bus/platform/drivers/efi-framebuffer/unbind

 

Edited September 23, 20178 yr by realies

-rc8a is EOL, please try -rc9f

On 22/09/2017 at 11:36 PM, limetech said:

-rc8a is EOL, please try -rc9f

Can confirm the same behaviour and fix for rc9f.

Can you provide your system diagnostics?  Also, what boot mode are you using (EFI or legacy)?

Sure, EFI.

 

 

Edited February 3, 20188 yr by realies

4 hours ago, realies said:

Sure, EFI.

 

helion-diagnostics-20170929-0421.zip

 

Can you reproduce the issue in legacy boot mode?  If not, there may be issues with that which we do not have plans to address in the short term.

Just tested and can confirm, legacy boot mode does not have this issue. What seems to be the issue with efi mode? Also, any particular differences other than this when loading unRAID in EFI/leagcy boot mode?

Just tested and can confirm, legacy boot mode does not have this issue. What seems to be the issue with efi mode? Also, any particular differences other than this when loading unRAID in EFI/leagcy boot mode?

We are honestly not sure. Efi booting is a new feature in 6.4 and figuring out a way to support that method with GPU pass through just isn't a priority right now. Too many other bigger fish to fry.

Other than this issue with GPU pass through, I am not aware of any other problems specific to EFI booting.

Sent from my SM-G930P using Tapatalk