ubiquiti(udw): the Dream Wall's seventeen LAN ports are one bridge (refs #130) #2329

Merged
jaylouisw merged 1 commit from catalogue/1120-2252 into main 2026-11-20 23:08:50 +00:00
Owner

Refs #130 (reference only - this does not answer the UDW-PRO request, which stays open and blocked).

The Dream Wall's seventeen 1G LAN sockets were seventeen routed interfaces. Ubiquiti says otherwise, on its own tech-spec table:

row value source
MAC Address Table Size 8,000 https://techspecs.ui.com/unifi/cloud-gateways/udw (200, 704,702 B, read 2026-11-20)
VLAN 802.1Q same page, Software section
1 GbE RJ45 17 (4 PoE++; 4 PoE+; 4 PoE / 1G/100M/10M) same page, Overview; and https://store.ui.com/us/en/products/udw (200, 568,234 B) __NEXT_DATA__ -> props.pageProps.collection.products[0].technicalSpecification, which also carries a "LACP Port Aggregation" row

A MAC table is a bridge table; 802.1Q tagging and a LAG are switch features. A gateway whose LAN sockets were routed interfaces would hold none of them, and Ubiquiti counts the bank as one GbE row. devices/ubiquiti/udw.json declared no l2-forward, so the two commands a UniFi user runs first were refused:

% UniFi Dream Wall-d1 cannot switch frames: it has no l2-forward capability - this port routes: use `ip address <address> <mask>`
% UniFi Dream Wall-d1 cannot bundle ports into a LAG: it has no l2-forward capability

and two hosts on eth0 and eth1 in 192.168.1.0/24 could not resolve each other's ARP - the box dropped their frames as not addressed to this device. An address on eth0 was answered on eth0 alone, so a host on eth1 could not reach its own gateway.

This adds l2-forward to the UDW row and nothing else: same 17+1+2 copper/cage census, same 12 PoE ports, same radios, same WAN role on wan1. What is now true that was not:

  • two hosts on adjacent LAN ports reach each other (they are one broadcast domain, as the vendor's Port Layout row says);
  • an address on a LAN port is answered from the whole bank, not just that port;
  • switchport access vlan <id> and channel-group <n> mode ... are accepted;
  • a LAN port is still routable with no switchport, so existing routed-LAN labs keep working, and wan1 still takes an address and keeps role: "wan".

test/udw-lan-bank-bridges.test.js gates all of that, including the census assertions, so a future edit cannot drop the bridge again without a red test.

Gates: node scripts/syntax-check.js (501 files OK), node scripts/audit-port-types.js (0 contradicting, 0 fibre-as-copper), node scripts/audit-devices.js (0 hard defects), node scripts/check-control-bytes.js (clean), node scripts/run-tests.js (421 suites, 13,959 assertions, 0 failed), test/hardware-invariants.test.js (26 passed). docs/DEVICE_REGISTRY.md regenerated - no change, the role prose already covers an l2-forward+l3-forward box.

Not done here, deliberately: the UniFi gateway siblings (udm, udm-pro-max, ucg-max, uxg-max) have the same missing capability, and each needs its own vendor rows read before it is edited - this PR does not touch them.

Refs #130 (reference only - this does not answer the UDW-PRO request, which stays open and blocked). The Dream Wall's seventeen 1G LAN sockets were seventeen routed interfaces. Ubiquiti says otherwise, on its own tech-spec table: | row | value | source | |--|--|--| | MAC Address Table Size | 8,000 | https://techspecs.ui.com/unifi/cloud-gateways/udw (200, 704,702 B, read 2026-11-20) | | VLAN | 802.1Q | same page, Software section | | 1 GbE RJ45 | 17 (4 PoE++; 4 PoE+; 4 PoE / 1G/100M/10M) | same page, Overview; and `https://store.ui.com/us/en/products/udw` (200, 568,234 B) `__NEXT_DATA__` -> `props.pageProps.collection.products[0].technicalSpecification`, which also carries a "LACP Port Aggregation" row | A MAC table is a bridge table; 802.1Q tagging and a LAG are switch features. A gateway whose LAN sockets were routed interfaces would hold none of them, and Ubiquiti counts the bank as one GbE row. `devices/ubiquiti/udw.json` declared no `l2-forward`, so the two commands a UniFi user runs first were refused: ``` % UniFi Dream Wall-d1 cannot switch frames: it has no l2-forward capability - this port routes: use `ip address <address> <mask>` % UniFi Dream Wall-d1 cannot bundle ports into a LAG: it has no l2-forward capability ``` and two hosts on eth0 and eth1 in 192.168.1.0/24 could not resolve each other's ARP - the box dropped their frames as `not addressed to this device`. An address on eth0 was answered on eth0 alone, so a host on eth1 could not reach its own gateway. This adds `l2-forward` to the UDW row and nothing else: same 17+1+2 copper/cage census, same 12 PoE ports, same radios, same WAN role on `wan1`. What is now true that was not: - two hosts on adjacent LAN ports reach each other (they are one broadcast domain, as the vendor's Port Layout row says); - an address on a LAN port is answered from the whole bank, not just that port; - `switchport access vlan <id>` and `channel-group <n> mode ...` are accepted; - a LAN port is still routable with `no switchport`, so existing routed-LAN labs keep working, and `wan1` still takes an address and keeps `role: "wan"`. `test/udw-lan-bank-bridges.test.js` gates all of that, including the census assertions, so a future edit cannot drop the bridge again without a red test. Gates: `node scripts/syntax-check.js` (501 files OK), `node scripts/audit-port-types.js` (0 contradicting, 0 fibre-as-copper), `node scripts/audit-devices.js` (0 hard defects), `node scripts/check-control-bytes.js` (clean), `node scripts/run-tests.js` (421 suites, 13,959 assertions, 0 failed), `test/hardware-invariants.test.js` (26 passed). `docs/DEVICE_REGISTRY.md` regenerated - no change, the role prose already covers an `l2-forward`+`l3-forward` box. Not done here, deliberately: the UniFi gateway siblings (`udm`, `udm-pro-max`, `ucg-max`, `uxg-max`) have the same missing capability, and each needs its own vendor rows read before it is edited - this PR does not touch them.
catalogue(ubiquiti): the Dream Wall's LAN bank is one bridge, so VLANs and LAGs are configurable (refs #130)
All checks were successful
CI / Classify the change (pull_request) Successful in 15s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 15s
CI / Test (Node 22) (pull_request) Successful in 59s
CI / Server smoke and Docker build (pull_request) Successful in 25s
CI / Test (Node 20) (pull_request) Successful in 1m0s
CI / Test (Node 18) (pull_request) Successful in 1m5s
CI / Arm auto-merge (pull_request) Successful in 5s
4379c710e8
forgejo-actions scheduled this pull request to auto merge when all checks succeed 2026-11-20 23:03:46 +00:00
jaylouisw deleted branch catalogue/1120-2252 2026-11-20 23:08:52 +00:00
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
jaylouisw/packetforge!2329
No description provided.