Bool symbols in nimble, BTDM, and two examples used invalid default
literals. The parser already falls back to 'n'; set that explicitly.
Co-authored-by: Cursor <cursoragent@cursor.com>
1. fix(wifi): Rename old NAN configuration to NAN-Sync
- Rename CONFIG_ESP_WIFI_NAN_ENABLE to CONFIG_ESP_WIFI_NAN_SYNC_ENABLE to indicate
the support for Synchronized NAN (Wi-Fi Aware).
- Because the original flag really controls the synchronized feature set, rename it
to CONFIG_ESP_WIFI_NAN_SYNC_ENABLE so the NAN-Sync and NAN-USD paths can be
selected independently without confusion.
2. Document esp_wifi_start requirement and fix USD examples
3. Rename nan_callbacks to nan_sync_callbacks
4. Remove the discovery_flag, clarify docs for sync vs USD flows, and add USD start/stop APIs
5. Require esp_wifi_start() before USD start
6. docs(nan): add NAN-USD application examples
7. add migration guide and hints for NAN-USD proto field
8. Improve allow_broadcast documentation
9. Add attention to the API esp_wifi_remain_on_channel
10. fix(wifi): align NAN API renames and docs for v6.0
- keep shared APIs under esp_wifi_nan_* while reserving
sync/usd names for mode-specific entry points
- clarify synchronized-cluster scope in headers, docs, and migration notes (EN/zh-CN)
- update examples for renamed helpers and WIFI_NAN_SYNC_CONFIG_DEFAULT()
- rename `wifi_nan_config_t` to `wifi_nan_sync_config_t`
11. Mark NAN-USD as esp-idf experimental feature
1. Remove redundant fixes in upstream wpa_supplicant for USD
- Upstream supplicant has mostly fixed the issues regarding the
unnecessary resetting pauseStateTimeout.
- Upstream supplicant still needs one fix to avoid resetting the
pauseStateTimeout when subscribe message is received from the peer
which had triggered the pauseStateTimeout previously.
2. Replace array-based channel list with bitmap for NaN-USD
Use `wifi_scan_channel_bitmap_t` to represent the channel list for NaN-USD
publisher and subscriber configurations. This replaces the previous approach
that used a large array to store allowed channels.
Also aligns with internal scan bitmap conventions across Wi-Fi stack.
3. call esp_wifi_nan_stop() after USD exchange or STA stop
Ensure esp_wifi_nan_stop() is called after publish/subscribe activity
completes or when WIFI_EVENT_STA_STOP is received. This prevents NAN stop
errors due to inactive interface. NaN-USD currently uses STA interface
for Tx/Rx.
4. Fix task watchdog timer triggered in active USD subscriber:
As both USD supplicant and offchan TX component gets executed
in the wifi task, it created a deadlock like scenario where offchan TX
done callback was never getting executed and supplicant 'nan_de_timer'
keeps running but failing to send any subscribe frame.
5. Make sure that device is able to recieve action management frames
of size more than 1400 bytes.
6. Update proto field in SSI to match Wi-Fi Aware (NaN) spec format
The 'proto' field in the 'wifi_nan_wfa_ssi_t' structure previously used an
enum (wifi_nan_svc_proto_t), resulting in a 32-bit field. According to
the Wi-Fi NAN Specification (Table 5.7), this field must be a single
octet (8 bits). This commit updates the type to uint8_t to ensure
compliance with the specification.
This mismatch previously triggered warnings but did not cause functional
errors.
7. Set `allow_broadcast` to true in USD Remain on channel
This enables the peer discovery as USD uses NAN-Network Multicast BSSID
as A3 address in publish frames.
8. Implement custom channel<->frequency conversion for NAN-USD
NaN-USD only permits 20 MHz bandwidth channels in the 2.4 GHz and 5 GHz bands
(as per section 4.5.3 of the Wi-Fi Aware Specification v4.0). To avoid linking
a large portion of upstream supplicant code for frequency-to-channel and
channel-to-frequency conversion, implement minimal custom helper functions.
9. Limit NAN_DE_MAX_SERVICE to 2 for ESP_SUPPLICANT
10. Note: Upstream clamps negative `os_reltime` deltas to zero in `nan_de_srv_time_to_next()`,
but our ESP_SUPPLICANT port keeps `os_time_t` unsigned, so that guard just provokes a
compiler warning. We skip it for now because the scheduling loop validates past timestamps
before computing the diff.
Co-authored-by: Shyamal Khachane <shyamal.khachane@espressif.com>
The previous workaround does not work, the backup/restore should apply to RTC IO registers.
Also moved the workaround to sleep_gpio.c to avoid gpio hal using kconfig.
