TheUnboundDeveloper/AM-Reaper v3.3.9-beta-RT-BE96U

Reaper v3.3.9-beta — RT-BE96U

on GitHub

v3.3.9 — hardware QoS never shapes a shared LAN trunk, the ZenWiFi client list fills, one networkmap build on every model

• GT-BE98 / GT-BE98 Pro main routers were unusable on v3.3.8 (field, three GT-BE98 Pro units). v3.3.7 made
  hardware QoS resolve an 802.1Q WAN device to its parent port so a PPPoE line is shaped on its physical port. On
  the boards with the external 2.5G switch (GT-BE98, GT-BE98 Pro, GT-BE96, GT-BE19000) the "2.5G WAN/LAN-1" jack is
  vlan4094 on eth1, and eth1 is also LAN-2..4 and a wired AiMesh backhaul - so the whole shaper (queues
  rebuilt, PI2, the port shaper at the upload rate) landed on the LAN trunk: load 17, ksoftirqd and bcmsw_rx
  burning, 45 of 2400 flows accelerated, eight daemons stuck on rtnl_lock, a wired node unable to rejoin. v3.3.6
  had pointed tmctl at vlan4094 itself, which it refuses, so hardware QoS on that jack was silently inert. A
  resolved parent that is a bridge member or in lan_ifnames is now never shaped: both engines refuse to arm with
  one syslog line, and the QoS page says so next to the engine choice. A PPPoE session over an ISP VLAN on a
  dedicated WAN port still shapes its parent. Confirmed by the reporter on test images. Host tests
  test_hwqos_lan_trunk.py, test_hwqos_pppoe.py.
• ZenWiFi BQ16 / BQ16 Pro: the client list was empty from the first build. Every page fed by
  get_clientlist() - the Dashboard counts, the stock client list, the AiMesh topology's per-node counts - read
  networkmap's shared-memory client table through a struct with one member gated on RTCONFIG_CAPTIVE_PORTAL.
  The prebuilt networkmap (the same ASUS build on every BCM4916 model) writes that member; the BQ16 and BQ16 Pro
  are the only models built without Captive Portal, so their httpd read the entry counters 256 bytes early: a
  zero count. The member is now unconditional (an ABI pin, like the bwdpi block next to it). Host test
  test_networkmap_abi_pin.py compiles the table under both flag sets and requires one size.
• GT-BE19000: the clean room ships the same networkmap build as every other model. Its platform archive
  carried the GPL 39274 networkmap, whose client table predates mlo_links/is_re - the defect that emptied the
  GT-BE98 list at v1.5.9. ci/container_build.sh now copies canon's hash-pinned reference binary over the model's
  prebuild dir after the archive is unpacked and proves it by hash (test_networkmap_prebuilt_parity.py). Not yet
  seen on hardware; the model has no router-mode tester.

Images & checksums (RT-BE96U)

Two flashable images: + AI Advisor (default) and Standard (noMCP, all AI components compiled out entirely). Flash the *_nand_squashfs.pkgtb via Administration > Firmware Upgrade.

This is a beta release. Its filename carries _BETA and the router reports the same string on the dashboard and the About page, so you can always tell which channel a flashed image came from. Stable releases carry no marker.

Verify a download against the attached SHA256SUMS-RT-BE96U-Reaper_v3.3.9.txt.

Corresponding source & reproducibility

The RT-BE96U image for v3.3.9-beta is built from this repository at tag v3.3.9-beta-RT-BE96U: the pinned Asuswrt-Merlin base (3006.102.8-beta2, a7ebfa133a) plus the complete patch series. The tag freezes the exact source that produced it.

• Patches: patches/
• Provenance record: provenance/manifest.json
• How to verify: docs/REPRODUCIBILITY.md and docs/SOURCE-AVAILABILITY.md

The auto-attached Source code (zip/tar.gz) asset below is this repository at tag v3.3.9-beta-RT-BE96U (patches + docs).