Re: [RFC PATCH 0/4] wifi: rtw89: add NL80211_IFTYPE_P2P_DEVICE support

[andres parra <andres.parrab@gmail.com>]

Hi Ping-Ke,

Thanks for the quick, detailed reply -- really appreciate it. I applied
patches 11-14 from your series (git am -p6 straight onto a clean sync
of core.c/core.h/chan.c/chan.h/debug.c/fw.c/ps.c/regd.c/util.h against
current mainline, since I don't have a full rtw-next checkout set up
yet) and tested on the same RTL8922AE hardware.

Results:

- Built clean, loaded clean, zero WARN/BUG in dmesg through the whole
  test.
- Real P2P-Device wdev appears (confirmed via `iw dev`), same as with
  my own patch.
- Real P2P discovery, GO negotiation (P2P-GO-NEG-SUCCESS), and full
  group formation against my actual Wi-Fi Display TV, tested twice:
  once with raw wpa_cli, once through a real WFD casting application
  (different encoder/bitrate settings each time). Both completed
  successfully, real DHCP lease on the resulting P2P-Client interface,
  real video actually displaying on the TV.
- No regression to normal STA association/roaming with your patches
  loaded.

On the two things I asked about:

1. Cross-band cost: confirmed reproducible with your patches too, and
   it matches your explanation exactly -- forced the STA link onto
   5GHz while the P2P-Client link stayed on 2.4GHz (same channel it
   started on), and casting became noticeably choppier, then smooth
   again once STA moved back to 2.4GHz. Consistent with TDMA timeslot
   sharing rather than a bug in either patch set.

2. Same-band channel-switch disconnect: reproduced it with your
   patches too, while writing this email. My AP did a real, ordinary
   channel switch on the STA link (CTRL-EVENT-STARTED-CHANNEL-SWITCH
   -> CTRL-EVENT-CHANNEL-SWITCH, 2462 MHz -> 2412 MHz, i.e. channel 11
   to channel 1, both 2.4 GHz), while the P2P-Client link was still
   active on channel 11. Result, same signature as with my own patch:

     p2p-wlanX-Y: CTRL-EVENT-BEACON-LOSS
     p2p-wlanX-Y: CTRL-EVENT-DISCONNECTED bssid=<GO> reason=4 locally_generated=1
     P2P-GROUP-REMOVED p2p-wlanX-Y client reason=IDLE

   No automatic recovery -- the group is gone, casting stops, no crash
   or kernel WARN/BUG (dmesg clean, system otherwise fully stable,
   STA link itself reconnects fine on the new channel). So this one is
   confirmed **not** fixed by patches 11-14 -- same behavior as
   before your series was applied. Happy to test whatever you'd
   suggest, or dig into rtw89_entity_mgnt/chan.c myself if that's
   useful, now that patch 11 gave me a clearer read on that code.

Test flow, as asked, for reference:
  wpa_cli -i p2p-dev-wlan0 p2p_find
  wpa_cli -i p2p-dev-wlan0 p2p_connect <TV MAC> pbc go_intent=0
  (watch for P2P-GO-NEG-SUCCESS / P2P-GROUP-STARTED in
  journalctl -u wpa_supplicant)
  -- then wait for or induce a real AP-side channel switch on the STA
  link (I couldn't force mine reliably; this one happened on its own
  from the AP), watch for CTRL-EVENT-CHANNEL-SWITCH on the STA link
  followed by whatever happens to the P2P-Client link.

Given your team already has a complete, working implementation with
real fixes I wouldn't have found on my own (the 6 GHz regd recalc
issue, the entity-pause/tracking interaction), I'll hold off on
pushing my own patches further -- happy to keep helping test yours on
real hardware instead, since that seems like the more useful thing I
can offer here. Thanks again for taking the time to look at this and
point me in the right direction.

Andres