AX211 Fails to Start on a range of kernels/firmwares (5.18.9, latest)

scsich · August 31, 2022, 1:56pm

@makuoly thanks a ton for pointing this patch out.

Thank you @crogers1 for making this available. As of this AM, I can confirm that the patch referenced here works verbatim:

github.com

crogers1/xenclient-oe/blob/8c5ba0dfd3e62e6d856f80233669c5bdb8f0f563/recipes-kernel/linux/5.15/patches/iwlwifi-enable-interrupts-for-pnvm.patch

################################################################################
SHORT DESCRIPTION: 
################################################################################
Re-enable interrupts before loading PNVM.

################################################################################
LONG DESCRIPTION: 
################################################################################
For cards in the AX210+ family that advertise a SKU and try to load a PNVM,
in the non-MSIX (MSI) case the normal firmware load flow disables all interrupts
from the card except ALIVE and RX. This is primarily to enforce the driver's
assumption that we will not receive RF_KILL before the firmware has finished
loading, or else an unintelligible error is dumped.  

This unfortunately also prevents us from responding to an interrupt from the 
card after 'kicking the doorbell' (UREG_DOORBELL_TO_ISR6_PNVM). We therefore
wait until the specified timeout and report "Timeout waiting for PNVM load!"
and fail out of driver initialization.

This patch re-enables interrupts prior to loading the PNVM so we can respond

This file has been truncated. show original

For me, it works against a minor update to 5.18.19 (see version file) and insertion of the patch into series.conf and 'kernel.spec.in` from:

I’m sure the Github issue referenced could use some weigh in (2x link limit reached for me).

Happy to post RPMs for this kernel somewhere if that’s of interest to others that are frustrated with Intel on this thread.