On 11/18/2024 15:11, Jacobo Pantoja wrote:
Thanks for answering
On Mon, 18 Nov 2024 at 02:20, Mario Limonciello
<mario.limonciello@xxxxxxx> wrote:
Hi Jacobo,
On 11/17/2024 11:42, Jacobo Pantoja wrote:
Hi Mario / crypto mailing list:
I'm trying to pass-through my AMD 5600G's integrated CPU; I can do it
easily with Linux guest, but I'm being unable to do so with a Windows
11 guest (which is my end goal)
What do you mean "pass through the CPU"? What exactly is "working" with
Linux guests and what exactly is "failing" with Windows ones?
Is this related to passing through the graphics PCI device from the APU
and having problems with that perhaps?
Yes, sorry, it's "GPU" pass-through. I typed it right on the subject
but not in the body
I've noted in my dmesg the following line:
"ccp: unable to access the device: you might be running a broken BIOS"
Are you trying to pass through the PCI device for the PSP to a guest?
What is your goal with it?
Just want to pass the GPU, nothing else
Tracing it a bit on the internet, I found a couple of fwupd commits
done by you stating that in some platforms this is expected (e.g.
0x1649) [1]
Comparing, in my motherboard I see that the Platform Security
Processor is 1022:15DF, being that last number in the same code you
applied the patch... but I cannot understand whether the ccp message
is expected on this platform (chipset is B450, if it adds info) or
not; and if this could be related to the pass-through problems.
Any hints would be more than welcome
Those messages are referring to some cryptographic acceleration IP
offered by the PSP on some SoCs.
Not all BIOSes all access to it and it really is a case by case basis if
it's expected behavior or not. When it's not accessible that "just"
means that you can't use that acceleration feature. There are other
features the PCI PSP driver exports such as TEE, security attributes,
dynamic boost control, SEV etc. Not all platforms support all features.
If you're just shooting in the dark for your issue based on the warning
about the BIOS not offering CCP this is probably the wrong tree to bark
up. If it's actually related it would be good at least for me to
understand what that message has to do with a Windows guest.
I'm just trying to discard potential issues in trying to achieve the GPU
pass-through. I'm pretty sure it is related to the way the GPU driver
starts the hardware in Windows, because when the guest is Linux,
just passing through the VFCT ACPI table is enough to get the GPU
working. But in Windows there's something different and I was wondering
if this CCP "issue" could be somehow related.
But I get from your message that this is most likely unrelated to my
GPU pass-through problem.
Best regards
Got it. Yes; this CCP issue should be unrelated to issues with GPU pass
through and Windows.
