PiZeroW: brcmfmac driver reload conflicts with g_multi mass-storage

[macmpi]

Describe the bug

With g_multi modprobed with file=/dev/mmcblk0p1 and available usb0 interface, lets unload/reload brcmfmac driver (rmmod brcmfmac_wcc brcmfmac && modprobe brcmfmac)
After a little while (about 30sec), g_multi goes down with following in dmesg:

[   66.461661] dwc2 20980000.usb: dwc2_hsotg_ep_stop_xfr: timeout GINTSTS.GOUTNAKEFF
[   66.461877] dwc2 20980000.usb: dwc2_hsotg_ep_stop_xfr: timeout DOEPCTL.EPDisable
[   66.462087] dwc2 20980000.usb: dwc2_hsotg_ep_stop_xfr: timeout GINTSTS.GOUTNAKEFF
[   66.462268] dwc2 20980000.usb: dwc2_hsotg_ep_stop_xfr: timeout DOEPCTL.EPDisable
[   66.462477] dwc2 20980000.usb: dwc2_hsotg_ep_stop_xfr: timeout GINTSTS.GOUTNAKEFF
[   66.462659] dwc2 20980000.usb: dwc2_hsotg_ep_stop_xfr: timeout DOEPCTL.EPDisable
[   66.462842] dwc2 20980000.usb: dwc2_hsotg_ep_stop_xfr: timeout GINTSTS.GOUTNAKEFF
[   66.463019] dwc2 20980000.usb: dwc2_hsotg_ep_stop_xfr: timeout DOEPCTL.EPDisable
[   66.512705] dwc2 20980000.usb: new device is high-speed
[   71.631564] dwc2 20980000.usb: new device is high-speed

Reversing steps (first unload/reload brcmfmac, and then modprobing g_multi does NOT cause the issue.
Without any mass-storage file associated the issue does not happen either.
(same overall issue pattern if g_multi is replaced by equivalent configfs setup).

This issue happens with PiZeroW but does NOT with PiZero2W.

Steps to reproduce the behaviour

see description

Device (s)

Raspberry Pi Zero W / WH

System

• kernel 6.12.62
• $ vcgencmd version
  Dec 8 2025 19:42:54
  Copyright (c) 2012 Broadcom
  version 2226a853bb9f5fd80392e3a4a89e457aeca88008 (clean) (release) (start_cd)

Logs

$ dmesg | grep brcmfmac
[   15.067323] brcmfmac: F1 signature read @0x18000000=0x1541a9a6
[   15.073757] brcmfmac: brcmf_fw_alloc_request: using brcm/brcmfmac43430-sdio for chip BCM43430/1
[   15.078738] usbcore: registered new interface driver brcmfmac
[   15.711846] brcmfmac: brcmf_c_process_txcap_blob: no txcap_blob available (err=-2)
[   15.722132] brcmfmac: brcmf_c_preinit_dcmds: Firmware: BCM43430/1 wl0: Jul 19 2021 03:24:18 version 7.45.98 (TOB) (56df937 CY) FWID 01-8e14b897
[   21.520429] brcmfmac: brcmf_cfg80211_set_power_mgmt: power save enabled

Additional context

No response

[pelwell]

I've labelled this a Curiosity, as it is unlikely to affect people and there is a clear workaround.

[macmpi]

In fact the driver reload situation is an easy way to trigger the (fairly common) root issue: agreed the presentation may sound quite niche at first, but the failure is more generic, with no obvious workarounds in most cases...

It's probably more a timing conflict that pops-up in (many) various instances that impacts users: we indeed see those dwc2_hsotg_ep_stop_xfr: timeout GINTSTS.GOUTNAKEFF problems in many different reports.
For instance I sometimes see same when starting a boot service that setup gadgets (with no brcmfmac driver reload per say ... and therefore no workaround).
It does not impact all devices (e.g. PiZero2W not impacted), so it may be some low-level device configuration issue.

Unsure it may relate to the stall=0/1 limitations: interested on your feedback on this too.