Deco X20 Firmware Instability — Multicast Table Overflow and Wi-Fi Driver Resets (Devices Disappear in App)

 

Description of Issue

Since creating a new network on 13 Oct 2025, all devices appear offline in the Deco app even though connectivity remains active.
System logs show a repeating sequence of multicast table overflows, Wi-Fi driver read failures, and cloud sync errors.
The pattern points to a firmware-level instability when high multicast/mDNS/SSDP traffic is present (Sonos, LG TVs, Home Assistant, ESP devices, etc.).

 

Observed Behavior (20:17 – 20:22 AEST)

1. Multicast/IGMP Flood Detected

Mon Oct 13 20:17:40 2025 daemon.err imsnoop: mcManagerUpdateIFEntry: if node table is full
Mon Oct 13 20:17:40 2025 daemon.err imsnoop: mcManagerMCProcess: Flood table is changed, update multicast flood table
Mon Oct 13 20:17:40 2025 daemon.err imsnoop: mcManagerMCProcess: Flood table is non-empty, update multicast flood table, entry count is 4
Mon Oct 13 20:17:40 2025 user.notice IMPROXY: INFO[imp_verify_multicast_addr@258]: Group address 239.255.255.250 belongs to IPv4 Local Scope.
Mon Oct 13 20:17:40 2025 user.notice IMPROXY: INFO[imp_verify_multicast_addr@245]: Group address 224.0.0.251 is Local Network Control Block

2. Wi-Fi Driver Failure

Mon Oct 13 20:17:17 2025 daemon.err nrd[21129]: _get_vap_status: fgets failed. cmd wl -i  bss
Mon Oct 13 20:21:56 2025 daemon.err nrd[21129]: _get_vap_status: fgets failed. cmd wl -i  bss
Mon Oct 13 20:21:25 2025 daemon.notice nrd[21129]: ar_apinfo_collect[line 1107]: AP sync timer expires!

3. Deco Cloud (AUC) and HTTP Daemon Errors

Mon Oct 13 20:21:15 2025 daemon.notice url-class[10008]: [ERROR] Communication with AUC server failed with error code: 104. HTTP response code: 400
Mon Oct 13 20:21:43 2025 daemon.notice url-class[10008]: [ERROR] Query failed for URL 'wpad.localdomain'
Mon Oct 13 20:21:49 2025 daemon.notice url-class[10008]: [ERROR] Query failed for URL 'recent.svc.cloud.microsoft'
Mon Oct 13 20:21:59 2025 daemon.err uhttpd[7501]: sh: iwconfig: not found

4. UPnP / Bridge Failures

Mon Oct 13 20:20:22 2025 daemon.warn miniupnpd[19600]: upnp_event_process_notify: connect failed: Connection timed out
Mon Oct 13 20:22:18 2025 daemon.warn miniupnpd[19600]: upnp_event_process_notify: connect failed: No route to host

5. DHCP Continues During Fault (showing system still alive)

Mon Oct 13 20:22:43 2025 daemon.info udhcpd[17149]: Sending ACK to 192.168.68.115
Mon Oct 13 20:22:43 2025 daemon.err udhcpd[17149]: my ntp server ip:27 124 125 250 be83f1bc

 

Analysis Summary

• Subsystems involved:
  Multicast manager (imsnoop / improxy), Wi-Fi driver (nrd), DHCP daemon (udhcpd), cloud connector (url-class), and local web server (uhttpd).

• Primary trigger:
  imsnoop multicast table overflow (if node table is full) → driver thread lock (_get_vap_status fgets failed) → AUC communication failures.

• Effect:
  Clients stay online but disappear from the Deco app; management daemons and cloud services repeatedly reset.
  All symptoms are consistent with multicast table memory exhaustion under high mDNS/SSDP load.

Environment

• Model: TP-Link Deco X20 (Host in Ethernet backhaul mesh)

• Devices connected:118 devices including 40 LXC containers on my Proxmox Server (Sonos, LG TVs, Home Assistant server, ESP devices, NAS, Reolink cameras).

• Network load: High multicast traffic (239.255.255.250 / 224.0.0.251).

• Issue reproducible after every 12–24 hours of uptime.

Steps to Reproduce

1. Operate the Deco X20 with > 80 active devices, including Sonos and Home Assistant.

2. Observe logs until imsnoop “if node table is full” appears.

3. Within 1–2 minutes, nrd fgets failed and url-class AUC error 104 occur.

4. Devices vanish from Deco app while internet remains functional.

Recommendations
What do i need to do?
 

Thank you.
This issue appears to stem from multicast table overflow leading to driver resets and cloud sync errors. Please forward to engineering for firmware review and potential fix confirmation.

 

These 20:57 logs confirm the same cascading fault sequence continuing 40 minutes after the initial crash, showing the Deco X20 host never recovered and stayed in a degraded state.

🔍 Additional Failure Evidence – 20:57 AEST

1. Wireless Driver Loop (unchanged fault)

Mon Oct 13 20:57:13 2025 daemon.err nrd[23448]: _get_vap_status: fgets failed. cmd wl -i  bss
Mon Oct 13 20:57:13 2025 daemon.notice nrd[23448]: ar_apinfo_collect[line 1107]: AP sync timer expires!
Mon Oct 13 20:57:44 2025 daemon.err nrd[23448]: _get_vap_status: fgets failed. cmd wl -i  bss

→ Wi-Fi driver remains hung, unable to read virtual interface state for over 40 minutes. The AP sync timer confirms it is retrying without success.

2. Persistent Multicast Flooding

Mon Oct 13 20:57:13 2025 user.notice IMPROXY: INFO[imp_verify_multicast_addr@245]: Group address 224.0.0.251 is Local Network Control Block
Mon Oct 13 20:57:13 2025 user.notice IMPROXY: INFO[imp_verify_multicast_addr@258]: Group address 239.255.255.250 belongs to IPv4 Local Scope.

→ improxy still looping through mDNS/SSDP multicast processing (224.0.0.251, 239.255.255.250).
Confirms multicast table cleanup never executed after overflow.

3. DHCP Keeps Cycling Through Entire Lease Pool

Mon Oct 13 20:57:13 2025 daemon.info udhcpd[17149]: fscanf_num:3 device_mac:BC-24-11-71-4F-91 device_ip:192.168.68.120 device_hostname:homebox
Mon Oct 13 20:57:13 2025 daemon.err udhcpd[17149]: Sending OFFER of 192.168.68.76

→ DHCP daemon still re-enumerating and re-ACKing leases for nearly every device—consistent with the host continuously resetting its LAN sessions internally.

4. HTTP and Cloud Subsystem Still Faulted

Mon Oct 13 20:57:17 2025 daemon.err url-class[25536]: after send to auc
Mon Oct 13 20:57:27 2025 daemon.err uhttpd[7501]: sh: iwconfig: not found
Mon Oct 13 20:57:49 2025 daemon.err uhttpd[7501]: uci: Entry not found

→ url-class (AUC cloud agent) remains in retry loop; uhttpd continues failing local Wi-Fi config calls (iwconfig not found).
Indicates management and cloud layers frozen but the base routing engine still operational.

5. IPv6 Router Advertisement Warnings

Mon Oct 13 20:57:13 2025 daemon.warn radvd[12491]: RDNSS address ... is not advertised by us

→ Router Advertisement Daemon confused by inconsistent IPv6 broadcast state—secondary symptom of bridge instability, not a root cause.

6. Guest VLAN Process Restart

Mon Oct 13 20:57:49 2025 user.info lua: luci_app:guest-eth guest network need set vlan state true

→ Suggests background process attempting to reinitialize guest VLAN but still within same frozen state.

🧩 Consolidated Diagnosis

The pattern across 20:17 → 20:30 → 20:57 shows a non-recovering firmware-level network stack deadlock:

• Multicast flood (improxy / imsnoop) overflows interface tables.

• Wi-Fi driver (nrd) enters loop unable to read interface.

• DHCP continues issuing ACKs, masking the deeper fault.

• HTTP & Cloud subsystems (uhttpd, url-class) hang indefinitely.

• Only full power-cycle clears the condition.

✅ Final Ticket Addendum (to append to TP-Link Support Request)

Update – 20:57 AEST (13 Oct 2025)
Issue persisted for over 40 minutes after initial failure.
Logs again show repeating improxy multicast events (224.0.0.251 / 239.255.255.250), nrd _get_vap_status fgets failed, and continuous url-class[xxxx]: after send to auc retries.
DHCP (udhcpd) continues to reprocess the entire lease pool, and the web daemon (uhttpd) fails multiple commands (iwconfig not found, uci entry not found).
No recovery or reboot occurred during this time.

This confirms that the Deco X20 host’s firmware enters an unrecoverable state under sustained multicast load.
Please escalate to engineering for firmware inspection of:

• Multicast table cleanup logic (imsnoop/improxy memory release)

• Wi-Fi driver watchdog (nrd handling of wl -i bss)

• Cloud/AUC retry throttling loop

• Optional feature to disable multicast proxy or enable IGMP snooping isolation