Re: How to test MediaTek mt7921!

(edited with ChatGPT)

Hi Bjoern,

384753

I tested an MT7922 on a Lenovo ThinkCentre Tiny 950q.

Hardware:
  PCI ID:      14c3:0616
  Subsystem:   17aa:e0c6
  ASIC:        79220010
  MAC:         c8:94:02:a3:8a:d9

System:
  FreeBSD 16.0-CURRENT
  main commit: b5c1ab6c8f2
  Kernel: main-n288503-b5c1ab6c8f28-dirty
  Applied: D58320, D58498, D58499
  Firmware: wifi-firmware-mt76-kmod-mt7921-20260410

Firmware initialization now succeeds on CURRENT:

  mt79210: successfully loaded firmware image
      'mediatek/WIFI_RAM_CODE_MT7922_1.bin'
  mt79210: ASIC revision: 79220010
  mt79210: successfully loaded firmware image
      'mediatek/WIFI_MT7922_patch_mcu_1_1_hdr.bin'
  mt79210: HW/SW Version: 0x8a108a10,
      Build Time: 20260224103145a
  mt79210: WM Firmware Version: ____000000,
      Build Time: 20260224103233

A wlan0 VAP can also be created successfully:

  wlan0: parent interface: mt79210
  country US
  supported channels include 1–11, 36–48 and 149–165

The remaining failure is scanning. Both ifconfig and
wpa_supplicant scans return immediately with no BSS entries:

  wlan0: Event SCAN_RESULTS received
  Received 0 bytes of scan results (0 BSSes)

The scan completes in less than one millisecond, suggesting that no
actual hardware scan occurs or its results are not delivered.

I also attempted to bypass scanning using ap_scan=2, a fixed channel,
SSID and BSSID:

  SSID: Confusion
  BSSID: e8:9f:80:c7:55:f7
  Channel: 149 / 5745 MHz

Association still fails because net80211 has no BSS node in its scan
cache:

  ioctl[SIOCS80211, op=21, val=0, arg_len=42]:
      No such file or directory
  wlan0: Association request to the driver failed
  Authentication with 00:00:00:00:00:00 timed out

After repeated scan/association attempts, the firmware becomes
unresponsive:

  mt79210: driver own failed
  mt79210: Message 00000010 timeout
  mt79210: Failed to get patch semaphore
  mt79210: chip reset failed

A complete cold power cycle is then required.

During the first VAP creation attempt I also saw WITNESS warnings about
sleeping allocations while lkpi-priv-dma was held. The stack included:

  linux_dma_map_phys_common
  mt76_dma_tx_queue_skb_raw
  mt76_mcu_skb_send_and_get_msg
  mt76_connac_mcu_set_mac_enable
  mt7921_start
  lkpi_ic_vap_create

Please let me know which debugfs data, boot logs or additional debug
options would be most useful. I can rebuild and test patches on this
machine.

Regards,
Raul