t4t5/owly: AirDrop with an iPhone from a non-Apple card, and the MT7925 is on its
tested list. MIT, 3.2k stars, new to our list. The hardware table says the MT7925 (mt7925e, Framework 13,
14c3:0717) works in both directions in shared mode, meaning the card stays associated while AWDL runs on the AP's
channel (6, 44 or 149). The repo carries iOS 26 captures from that card (tests/captures/*-mt7925e-ch44-1.pcap), a
protocol spec (docs/SPEC.md) and the driver notes in crates/owly-radio/drivers.toml. What they report about the
card:
- injected frames leave at most one every 2.5 ms, so sending is slower than receiving (about 490 kB/s against 1.3 MB/s);
- after a monitor interface is removed the station stops receiving 3 to 165 s later (an mt76 firmware bug), so owly reconnects at the end of every session;
- the MT7927 (
14c3:7927) never hears the phone in shared mode (owly#44); - the MT7921 and MT7922 only work in an exclusive, opt-in mode, one slot in sixteen at 1 Mbit/s (owly#14).
Why it matters: our #2 (no AWDL
interface) has had no proof that the MT7925 radio and firmware can send and receive AWDL frames at all. This is that
proof on Linux, with captures to compare our frames against. It does not make AirDrop work on macOS: owly runs the
whole protocol itself, while macOS runs it in IO80211AWDLPeerManager and wants the driver for radio, window events
and frame delivery. The injection rate limit and the post-monitor receive stall are things MtkWlan would hit too.
Confidence: High that it works on Linux (their table and captures), Unverified on whether the same firmware behaviour
holds under our MtkWlan.
DrogaBox/RTL8814Native: AWDL got further. About 90 commits in
20 hours, nearly all documentation. Against 25G241 their layer now creates awdl0, and
IO80211AWDLPeerManager::configure finishes, enableAWDL runs the whole ioctl chain (155, 129, 117) and wifip2pd
registers the AirPlay and RAOP searches
(AWDL_STAGE3_CONFIGURE_RESULTS.md).
The ioctl table says which setters are fatal: 125, 122, 132, 135, 134, 116 (inside interfaceStateChange) and 117;
most others are only logged. They stop there: the manager runs a solo election with no peers, and nothing can be
checked without a peer on the air
(AWDL_ESTADO_Y_BLOQUEO.md).
They plan to use an ESP32-S3 running ESP32Drop (0BSD) as a test peer and
list owly, winawdl, omdrop and seemoo's owl as protocol references. The project is GPL-2.0 (Realtek header), so
ideas only. Everything is for 25G241 and our 25G332 needs each offset rechecked. Confidence: Medium (their own
hardware results, one machine, not reproduced by us).
It also does WPA3-SAE inside its driver
(SAE_HANDOFF.md): group 19, AKM 8, PMF,
SA Query, tested against one SAE-only hotspot and one transition router. Our README says no WPA3-SAE, and this is
an independent record of what that takes (its PHASE5_IO80211.md has an "SAE gates" section we have not read yet).
Confidence: Medium.
Discovery: five EFI repos and a Realtek RTW89 port (FranciumOS-org/AirPort_RTW89, OCLP legacy stack, no WPA3 on its tested chips, no AWDL). kgp-macPro/AirportItlwm-Tahoe pushed Intel BE200 results (legacy stack, no bearing on the native ABI). Also new on our list: raph559/winawdl (MIT, AWDL on a Realtek card under Windows) and ESP32Drop, as above.