RT-AX86U 2.5G LAN port errors

[panni]

Hey,

I've tested three cables so far on the 2.5G LAN port of my RT-AX86U (386.7_2), using iperf3 with 10 parallel streams.
I'm seeing these errors on the router with all three cables (and about 1.24 GBit throughput only), when sending data from the AX86U:

`iperf3 -c IP -P 10` on the router, `iperf3 -s` on the other machine, which is connected through a 2.5G switch.

Sep 23 15:15:08 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=15361
Sep 23 15:15:10 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=16385
Sep 23 15:15:13 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=17409
Sep 23 15:15:16 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=18433
Sep 23 15:15:22 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=19457
Sep 23 15:15:25 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=20481
Sep 23 15:15:29 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=21505

Has anyone else observed this? It's pretty obvious that in TX the router has issues. All three cables are at least CAT 5e certified and 10m in length.

This only happens with 10 parallel streams. With a single stream the errors don't show up, but the speed is still way below 2.5G.

RX is doing 2.3 GBit without any errors. Including overhead, that's 2.5G.

Edit: All filtering disabled, trendmicro declined.

 

[ColinTaylor]

Yes, I wrote about it here. It doesn't just affect the 2.5GbE port but also the regular 1GbE ports.

This only affects router to LAN data transfers, not WAN to LAN or LAN to LAN.

 

[panni]

Yes, I wrote about it here. It doesn't just affect the 2.5GbE port but also the regular 1GbE ports.

This only affects router to LAN data transfers, not WAN to LAN or LAN to LAN.

Hmm, not sure this is the same thing. Using the 1G ports I can't reproduce the issue, only on the 2.5G one.

I should see the same issues on the 1G ports if this was the same issue as the one you've described, right?

 

[ColinTaylor]

I should see the same issues on the 1G ports if this was the same issue as the one you've described, right?

Yes (if it's a router to LAN test). I don't use the 2.5GbE port but can create exactly the same errors as you when I use your iperf command:

Sep 23 14:36:54 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=1
Sep 23 14:36:55 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=1025
Sep 23 14:36:56 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=2049
Sep 23 14:36:57 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=3073
Sep 23 14:36:59 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=4097
Sep 23 14:37:00 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=5121
Sep 23 14:37:01 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=6145
Sep 23 14:37:02 kernel: ERR: rdpa_cpu_tx_port_enet_lan#213: rdpa_cpu_tx_port_enet_lan failed. rdd_rc=120 tx_rdd_error_count=7169

It would be easy to test if it's the same problem, just issue the command I suggested in that other post and see if your errors go away:

ethtool -K br0 tx off sg off tso off gso off

EDIT: Sorry, ethtool isn't part of the firmware, it has to be added from Entware.

 

[panni]

This only affects router to LAN data transfers, not WAN to LAN or LAN to LAN.

This is very helpful. I was beginning to think I had to ditch the 2.5G port, but if this doesn't affect LAN to LAN and WAN to LAN, then it should be fine.

I'm using a 1G WAN connection which saturates the 1G connection to other devices on my network (streaming boxes), when under full bore. That's why I wanted to use the 2.5G port.

My current setup is:
[2.5G switch (PC 1 (NAS, streaming), PC 2)] <- 1G -> [RT-AX86U <-> 1G WAN + a couple of streaming devices that need to connect to PC1]. As with the upgraded 1G connection, a download thread from PC1 can saturate the link between it and the RT-AX86U, restricting the bandwidth of the streaming devices.

Edit: I think the above is wrong. Only an upload thread of PC 1/2 would hinder the streaming devices from achieving their necessary downlink speeds from PC1.
I may not need the 2.5G port anyways.

Or am I looking at this wrong?

 

[ColinTaylor]

Try running the iperf client on another LAN device rather than the router.

 

[ColinTaylor]

Edit: I think the above is wrong. Only an upload thread of PC 1/2 would hinder the streaming devices from achieving their necessary downlink speeds from PC1.
I may not need the 2.5G port anyways.

Or am I looking at this wrong?

I think you are correct in that the large volumes of data are going in different directions. So it's not so much an issue of bandwidth but rather QoS. i.e. even though the streaming device is pulling most of it's data from the NAS it still has to constantly send ACK's back to the NAS to tell it that it received that data. If the NAS's downlink is saturated then the ACK's might be delayed.

However, most "streaming" protocols are designed to work over high latency links so the above shouldn't be a problem. The data is obtained in large chucks and then buffered. So in reality the data isn't actually "streamed" like it would be for VoIP for example.