Re: [PATCH] wifi: brcmfmac: avoid sleeping tx locks in netpoll context
From: Arend van Spriel
Date: Fri Oct 09 2026 - 08:04:54 EST
On 09/10/2026 09:30, Karl Mehltretter wrote:
On Tue, Oct 06, 2026 at 10:27:27AM +0100, Arend van Spriel wrote:To be honest I ported the whole shenanigan to brcmfmac, but I do not recall any projects used the functionality in production. It is mainly intended for SDIO devices.
Lockdep instrumentation in fcmode=1 and 2 confirmed that cleanup enters
without txq_lock.
This is an observation unrelated to the patch, right? You mentioned in the
patch it only covers fcmode=0, right?
Yes, the cleanup observation is unrelated to the fcmode=0 fix. I
mentioned it to explain why I kept brcmu_pktq_penq() for normal calls.
I'm looking at a separate brcmfmac FWS-path RT/netpoll fix for modes
1 and 2. The prototype defers IRQ-disabled transmit to the FWS worker.
Are modes 1 and 2 used or recommended on SDIO devices? That would
help me prioritise the follow-up.
Regards,
Arend