extra NPU driver

[timota]

Hi,

I’ve installed the latest 6.18 kernel and so far it appears to be working well — I haven’t noticed any issues.

However, I’m missing a driver/module for the following device:

Processing accelerators [1200]: Intel Corporation Arrow Lake NPU [8086:ad1d] (rev 01)

Would it be possible to include support for this device in the kernel build?

Thank you.

[thor2002ro]

i will add it next one

[timota]

Excellent, thank you — really looking forward to giving this a try!

[timota]

Hi @thor2002ro.

I have tried a new rc6 kernel. And see log messages that I did not see before on 6.18.x kernel

pcieport 0000:80:1b.4: AER: Correctable error message received
pcieport 0000:80:1b.4: AER: Uncorrectable (Non-Fatal) error message received
pcieport 0000:80:1b.4: PCIe Bus Error:
    severity=Uncorrectable (Non-Fatal)
    type=Transaction Layer
    (Receiver ID)
device [8086:7f44] error status/mask=...
[21] ACSViolation
TLP Header: 0x34000000 ...
thunderbolt 0000:84:00.0: AER: can't recover (no error_detected callback)
xhci_hcd 0000:97:00.0: AER: can't recover (no error_detected callback)
pcieport 0000:80:1b.4: AER: device recovery failed

device behind it is: Intel PCIe Root Port / PCH bridge on 700/800-series platforms (Raptor Lake / Meteor Lake class).

[thor2002ro]

Well that looks serios... pcie devices shouldn't trigger aer errors like that.... 2 possibilities there's an phisical problem or there's a regression in the kernel.... In some small cases could be the power saving... You could try "pcie_aspmoff" at boot
I have an amd system... I can't test stuff like this.... Just wait for the final 6.19 baybe it will be fine....

[timota]

It is definitely not hw issue. I don't see this messages on stock kernel and on your previous on 6.18.x. Only in 6.19. It makes me think that some changes been done in 6.19.x

[timota]

Also, when aspm=off no logs, So this clearly indicates that something in a kernel changed

[thor2002ro]

try 6.18.7 latest stable

[timota]

Hi @thor2002ro,

Apologies for the delayed response — I wasn’t able to test the kernel during the week. So far, everything looks good on 6.18.7 with no issues observed.

PS: I also tested the latest 6.19-rc7 and the issue is still present.

Thank you for your work and support.

[thor2002ro]

yah 6.19 has some aggressive pcie power saving on somewhere if I enable aspm full with super power saving pcie just dies on my ryzen :)))) even the gpu driver panics