Turris Omnia on a 300/100 line occasionally (once every 2-4 months) shows a connection speed of 100/100 (speedtest.net). In two cases, the problem with slower downloads disappeared after restarting the router.
How can I analyze the problem, if it occurs, in more detail ?
At the moment the connection is fine and through I found the message
Jan 9 13:56:02 Turris_JB kernel: [24085.751275] Marvell 88E1540 mv88e6xxx-1:04: Downshift occurred from negotiated speed 1Gbps to actual speed 100Mbps, check cabling!
Jan 9 13:56:02 Turris_JB kernel: [24085.764018] mv88e6085 f1072004.mdio-mii:10 lan4: Link is Up - 100Mbps/Full - flow control off
LAN 4 was the line where the technicians’ laptop connected when checking the connection - I had to allow them access to the network (IP bound to MAC). Her measurements were also fine
So as the messages tell you, something is marginal either the ethernet cable of one of the ports on either side. For 1 Gbps link rate ethernet needs all 4 wire pairs, while for 100 Mbps only 2 wire pairs are needed. If even a single wire has marginal contact (either broken in the cable itself or a dirty contact or a bent spring in the socket) the link might degrade down to 100 Mbps. I would bet that instead of restarting the router you could also:
a) unplug and replug the problematic cable in the affected LAN port
b) use ethtool to force 1 Gbps renegotiation (not sure about the invocation and currently not at home)
In my experience cables actually do get bad so consider replacing that cable and check both involved sockets on either end of that cable as well.
Also check with ethtool -S lan4 for errors they might indicate problem with cabling as well.
I have a link at home that has the same problem. So I forced 1Gbps there on one side and thats it. Left autoneg on but advertising only 1Gbps. Cable couldnt be replaced cuz its in the walls. So ot is what it is
The problem on LAN 4 is more just for illustration, the port is not commonly used and the measurements that the provider’s technicians made on it (as a free port) were fine. They measured 10% better than 300/100
I routinely take measurements on another LAN port. The problem will not, I think, be with the specific port, but in general. This should be demonstrated by the following: start monitoring the status systematically using LibreSpeed directly on the router.
At the time of the fault, I will check the interface in reForis
It doesn’t bother me that much, in normal operation I don’t know about it. But I’ll start looking into it.