RE: [PATCH rtw-next v3] wifi: rtw88: usb: route bmc frames via the high queue only for DTIM delivery

[Ping-Ke Shih <pkshih@realtek.com>]

Mehmet Fide <mehmet.fide@gmail.com> wrote:
> Hi Ping-Ke,
> 
> > FYI. I checked vendor driver. Normally REG_ATIMWND sets 0xa or 0xc in AP mode
> > for chips.
> 
> Thanks, that matches what I measured: 0x04 already drained the storm, so
> 0xa leaves comfortable headroom. If you agree I will add a small third patch
> to the series that programs REG_ATIMWND to 0xa when the interface goes into
> AP mode, as a complement to the bound and the filter (each covers a
> different failure mode: the window adds drain capacity, the bound protects
> the pool if the window is not enough, the filter keeps the ordinary chatter
> off the beacon-paced path altogether).

Agree. Please do it. 

As many chips don't initialize REG_ATIMWND (0x55a) at all, you might set the
value back to 0x2 when stopping AP.

> 
> > As you saw it is for TDLS channel switching (timeslot sharing), I
> > don't think it can work in AP mode...
> > I will ask USB experts internally to see if there is an interrupt
> > to notify driver about TBTT (or beacon early).
> 
> Understood, and thank you for asking. If such a notification exists I would
> be glad to prototype IEEE80211_HW_HOST_BROADCAST_PS_BUFFERING on top of it
> as the long-term solution; the bound and filter series does not depend on it.

I'm still checking this, and will give back to you if it exists.

By the way, comments you left in patches 1/2 and 2/2 look good to me.

Ping-Ke