I tried it on three routers in my testbed (all running OpenWrt, though I don't think that's relevant here):
1) Buffalo WZR600DHP (switch0: Atheros AR8316 rev. 1 switch registered on mdio-bus.0) 2) Ubiquiti AirRouter (switch0: Atheros AR724X/AR933X built-in rev. 2 switch registered on mdio-bus.0) 3) DLink DIR860L-B1 (mtk_soc_eth 1e100000.ethernet: loaded mt7530 driver[1]) All three freaked out on a LAN port to LAN port connection and some broadcast traffic (e.g. in my case the ARP packet resulting from an attempt to ping a non-existent ip address on the network). [1] I'm pretty sure the mt7530 is the switch that is embedded in the mt7621 SoC. On Wed, Dec 25, 2019 at 6:12 PM Russell Senior <[email protected]> wrote: > This behavior is going to depend on the switch chip embedded in the SoC > on Tomas's device. The Linux kernel or its bridging behavior won't be > involved until the traffic leaves the switch. In openwrt, there is a > standard interface for configuring the switch, called swconfig, although > the future belongs to something called DSA, or distributed switch > architecture. A few devices have DSA drivers, but most are still on > swconfig. Because of the uncertainty over what the specific switch does, > empiricism wins here. It does what it does. Short of a part number and a > detailed datasheet, testing is almost certainly the most expedient approach. > > On Wed, Dec 25, 2019, 17:59 Mike C. <[email protected]> wrote: > >> Date: Wed, 25 Dec 2019 12:25:35 -0800 >> Subject: Re: [PLUG] Lan loops - follow on PoE injector >> On Wed, Dec 25, 2019 at 9:33 AM Russell Senior <[email protected] >> > >> wrote: >> >> > Only one way to be sure! ;-) >> > >> > >> Said the person on top of the building when contemplating if the fall >> would >> kill him, while another person used factual data and applied higher level >> abstract concepts such as math & physics to determine the most likely >> outcome. J/K Ha! >> >> The OP is running OpenWrt on the router so this makes it far more >> interesting, especially for Network Nerds like me, as OpenWrt is highly >> configurable providing ample opportunities for experimentation. >> >> According to this table from OpenWrt documentation, if the device type is >> set to "bridge" then Spanning Tree Protocol can either be enabled or >> disabled. (See screenshot "STP") >> >> Even more interesting is that you might be able to even get a diagram of >> the switch/router backplane that will show you how the ports are >> physically >> & logically connected. Such as this diagram I found in the OpenWrt >> documentation. (See attached screenshot "Diagram") >> >> OpenWrt documentation refers to Asus WL-500G as a switch, however based on >> the diagram the 4 LAN ports are physically bridged off Eth0 and the only >> actual switching that happens appears to be between Eth0 (LAN) and Eth2 >> (WiFi). >> >> I suspect this is a pretty common architecture for SOHO network gear. >> Physically wired up, the LAN ports are a multi-port bridge. It's designed >> to bridge 4 physical LANs into one logical LAN (VLAN 0) the default >> management vlan. >> >> OpenWrt would most certainly set the interface type to bridge and enable >> STP thereby thwarting any layer 2 looping mischief. >> >> But just disable STP in OpenWrt and have some fun! >> _______________________________________________ >> PLUG mailing list >> [email protected] >> http://lists.pdxlinux.org/mailman/listinfo/plug >> > _______________________________________________ PLUG mailing list [email protected] http://lists.pdxlinux.org/mailman/listinfo/plug
