Re: [PATCH] wifi: mt76: mt7921u: Don't reset the USB device on probe
From: Devin Wittmayer
Date: Thu Nov 12 2026 - 19:32:23 EST
On Thu, Nov 12, 2026 at 05:39:33PM +0400, Alexey Charkov wrote:
> Bluetooth: hci0: urb 00000000c4b7b871 failed to resubmit (2)
> Bluetooth: hci0: Opcode 0x0c03 failed: -110
I get the same timeout on an Alfa AWUS036AXML (MT7921AU, x86) every
time Bluetooth and Wi-Fi probe back to back, and never with the patch:
v7.4-rc1 6.18.52
stock patched stock patched
Bluetooth reset times out
port disabled and enabled 5 of 5 0 of 5 5 of 5 0 of 5
reset and reauthorize 5 of 5 0 of 5 5 of 5 0 of 5
Wi-Fi interface appears in 8.8 s 1.9 s 8.8 s 1.9 s
reauthorize at USB 3.0 - - 5 of 5 0 of 5
(high-speed USB unless noted)
Both ways put the two probes back to back with the Wi-Fi driver
loaded. Wi-Fi behaves as on stock otherwise, unloading under traffic
included.
> Drop the reset. The USB core resets the port in hub_port_init() before
> addressing any newly connected device, using the same hub_port_reset()
> that usb_reset_device() ends up in, so the probe-time call repeats a
> reset that has just happened. A rebind without re-enumeration resets the
> Wi-Fi subsystem through mt792xu_wfsys_reset() on both disconnect and
> probe.
In the port case the probe finds the Wi-Fi firmware still running (I
logged the ready field). The patched driver comes up from there every
time, and from a cold plug-in at USB 3.0. On stock the field is still
set after the reset.
The second hunk does not apply to 6.18.52, where the probe still takes
a device reference. I dropped only the reset line.
Tested-by: Devin Wittmayer <lucid_duck@xxxxxxxxxxxxx> # Alfa AWUS036AXML