watchthelight labs

mt7925-wifi: 2026-10-29 (window 2026-10-28 11:14 to 2026-10-29 10:00 UTC)

Two new MT7921/MT7922 driver repos appeared. Neither one drives the MT7925 or contains code from this project. The one mt76 commit in the window does not change anything in MtkWlan.

Notable

Arxusion/AirportMtkWlan (repo, created 2026-10-28, GPL-2.0, release v1.5.1): a re-upload of the SimpleMtkWlan / AirportMtkWlan line with all 61 commits of history kept: laobamac's init, then mavizao, flxxxxddd, pengubow and kitsuneislife, then three new commits by Arxusion. It covers MT7921, RZ608 and MT7922/RZ616 only and builds the Sequoia and Tahoe front ends.

  • The new work: a WPA2/WPA3 transition-mode join fix (226361a) that negotiates PMF only when the AP's node wants it, leaves the VHT capability element out of 2.4 GHz association requests, and keeps deep sleep off during association. It also adds a release that removes the CI workflow and CLAUDE.md (6740e5d), and a SwiftUI installer that writes to the EFI partition (a483221).
  • Hardware: the README says MT7921 is "verified operational", but no tester, machine or measurement is named for the new commits. Medium confidence that the changes ran on hardware at all.
  • In 6740e5d, when the user sets no country code and none can be derived, the country code falls back to "IN" (with a comment saying this unlocks all 5 GHz channels). That is a regulatory-domain problem for users outside India. It is worth knowing before anyone points an MT7921 user at this build. High confidence (it is in the diff).
  • Our code: not used. I searched the whole tree for watchthelight, <wtl>, MtkWlanNative, org.watchthelight and Legion and found nothing. Its few MT7925_* register defines date from laobamac's initial import (1647d81). Its front end is laobamac's itlwm WCL code, not our IO80211 native approach. CREDITS.md and the README credit every upstream author. There is nothing for us to ask for. High confidence.
  • For us: our README comparison ("as of 2026-10-20") says AirportMtkWlan supports "macOS 15.2 to 15.x only". That is now out of date: this line also claims Tahoe. The 2.4 GHz VHT element is an idea we could check in assocIes, since we add VHT Capabilities whenever the AP advertises VHT, whatever the band. Idea only (GPL). Low confidence that it matters.

chenhggu6475/MT7922-itlmt7922-Hackintosh (repo, created 2026-10-29): a re-upload of hbyonly1's itlmt7922 0.2.2 (an MT7922 port of thanhnndev's itlmt7925, Clear BSD), plus two commits for an experimental WPA-EAP (PEAP/MSCHAPv2) branch (c89e7c2). The README says this was tested on eduroam on an RZ616 under macOS 15.8.1 in Ethernet mode. It credits thanhnndev and hbyonly1 and contains nothing of ours. If it holds up, it is the first enterprise-auth lead in the itlmt79xx family; our README lists "no enterprise". Medium confidence.

openwrt/mt76 59b149d (commit): the RX PN replay counters of an MLD peer now live on the peer's default wcid, so a frame replayed on another link no longer passes. MtkWlan has a single link and keeps its CCMP replay counters per key and TID in MtkSec.cpp, so nothing changes for us today. This only matters if MLO is ever added. High confidence.

Web sweep

Searches for MT7925 macOS, MT7922 Tahoe, AirportMtkWlan and MediaTek Wi-Fi 7 hackintosh found nothing new in the last two days.