bugreporter:

Also confirmed with different country codes and TX power settings.

bugreporter:

Also confirmed with openwrt-15.05.1-lantiq-xrx200-VGV7510KW22NOR-squashfs.image.

bugreporter:

These tests were carried out in quick succession while the device was less than 10 cm away from the client's wi-fi antennas and the LuCI web interface showed a signal strength of -12 dBm:

[localhost]$ iperf3 -c 192.168.1.1
Connecting to host 192.168.1.1, port 5201
[ 4] local 192.168.1.171 port 56724 connected to 192.168.1.1 port 5201
[ ID] Interval Transfer Bandwidth Retr Cwnd
[ 4] 0.00-1.00 sec 2.63 MBytes 22.1 Mbits/sec 0 25.5 KBytes
[ 4] 1.00-2.00 sec 2.31 MBytes 19.4 Mbits/sec 0 26.9 KBytes
[ 4] 2.00-3.00 sec 2.54 MBytes 21.3 Mbits/sec 0 26.9 KBytes
[ 4] 3.00-4.00 sec 2.61 MBytes 21.9 Mbits/sec 0 26.9 KBytes
[ 4] 4.00-5.00 sec 2.73 MBytes 22.9 Mbits/sec 0 41.0 KBytes
[ 4] 5.00-6.00 sec 3.15 MBytes 26.4 Mbits/sec 0 41.0 KBytes
[ 4] 6.00-7.00 sec 2.85 MBytes 23.9 Mbits/sec 0 63.6 KBytes
[ 4] 7.00-8.00 sec 3.34 MBytes 28.1 Mbits/sec 0 63.6 KBytes
[ 4] 8.00-9.00 sec 3.40 MBytes 28.5 Mbits/sec 0 63.6 KBytes
[ 4] 9.00-10.00 sec 3.73 MBytes 31.3 Mbits/sec 0 63.6 KBytes

iperf Done.

bjonglez:

Can you try running iperf on a computer plugged to a LAN port instead of running it on the router directly?

The CPU on this router does not seem to be really low-end, but you never know, iperf uses quite a lot of CPU.

bugreporter:

Yes, switching client and server does not make a difference:

root@router:~# iperf3 -c 192.168.1.161
Connecting to host 192.168.1.161, port 5201
[ 4] local 192.168.1.1 port 41652 connected to 192.168.1.161 port 5201
[ ID] Interval Transfer Bandwidth Retr Cwnd
[ 4] 0.00-1.01 sec 9.45 MBytes 78.8 Mbits/sec 0 90.5 KBytes
[ 4] 1.01-2.01 sec 9.85 MBytes 82.6 Mbits/sec 0 90.5 KBytes
[ 4] 2.01-3.00 sec 9.94 MBytes 83.5 Mbits/sec 0 90.5 KBytes
[ 4] 3.00-4.00 sec 9.95 MBytes 83.4 Mbits/sec 0 90.5 KBytes
[ 4] 4.00-5.01 sec 9.96 MBytes 83.3 Mbits/sec 0 90.5 KBytes
[ 4] 5.01-6.01 sec 9.91 MBytes 82.6 Mbits/sec 0 90.5 KBytes
[ 4] 6.01-7.00 sec 9.83 MBytes 83.3 Mbits/sec 0 90.5 KBytes
[ 4] 7.00-8.00 sec 9.88 MBytes 82.9 Mbits/sec 0 90.5 KBytes
[ 4] 8.00-9.00 sec 9.92 MBytes 83.0 Mbits/sec 0 90.5 KBytes
[ 4] 9.00-10.00 sec 9.85 MBytes 82.6 Mbits/sec 0 90.5 KBytes

iperf Done.


bjonglez:

Can you please try to run iperf through the router, instead of on the router?

Basically, take two computers, one on wifi, one on ethernet. Then run iperf between the two computers.

bugreporter:

Yes, does not make a difference.
After all, I originally noticed this problem because Wi-Fi would occasionally become even too slow for my already humble Internet connection.

I have further played around with settings and older firmware blobs, and checked the mtd partitioning, all to no avail.
The only other peculiarity I have noticed is that the kernel apparently recognizes the RT3062F chip as a RT3572:
[ 10.902947] ieee80211 phy0: rt2x00_set_rt: Info - RT chipset 3572, rev 0223 detected
[ 10.910726] ieee80211 phy0: rt2x00_set_rf: Info - RF chipset 0008 detected

bugreporter:

I have also checked the integrity of the 'board_config' partition to the best of my ability and manually replaced its embedded Ralink eeprom with the (endian-converted) 'RT3062.eeprom' file from the ZyXEL P-2812HNU-F1, without apparent improvement.

bjonglez:

Well, if performance was already poor in OpenWRT CC, I'm not sure what can be done (it's not a regression in LEDE).

You can try asking Daniel Golle, look at https://www.kickstarter.com/projects/1327597961/better-support-for-mt7620a-n-in-openwrt-lede/ (it's the same driver, rt2x00)

Sven:

I have the same problem too, with all of my o2 Box 6431 devices. Is there anything I can do to help fixing this?

The problem occurs in LEDE 17.01.[0-4] and in current OpenWrt snapshot reboot-5933.

The o2 Box 6431 is a really great OpenWrt device because all of its hardware is supported, but this problem makes the WiFi kind of unusable.

Thanks,
Sven

mkresin:

I've done a wireless benchmark recently and I'm not able to reproduce the wireless performance issue:

https://www.spinics.net/lists/linux-wireless/msg171568.html

https://www.spinics.net/lists/linux-wireless/msg171734.html

Whatever it is that you're facing it doesn't look like that something in general is broken.

Might be that the 2.4GHz band is just overcrouded at your place.

The limited ethernet bandwidth is a known issue.

Any traffic originating or targeting the o2 Box itself is limited to somewhat around 100Mbit/s. It is the maximum bandwidth the CPU is able to handle in software. Since routing and NAT is done in software (using netfilter framework of the linux kernel), both are hit by the 100Mbit/s cap.

To achieve more bandwidth, support for the hardware accelerated routing/NAT is required. To my knowledge nobody is working on this topic at the moment. One of the reasons is that the datasheets are only available under NDA and I'm not aware that anyone not having a contract with/being a customer of lantiq was ever able to get hands on the datasheets.

I've done some benchmarking using the kernel 4.14 flow offloading and was able to see an improvement (lantiq board with gigabit ethernet used):

HEAD

net to client 95.2 Mbits/sec routing
client to net 109 Mbits/sec routing

net to client 85.6 Mbits/sec NAT
client to net 96.2 Mbits/sec NAT

HEAD + FLOW_OFFLOADING

net to client 129 Mbits/sec routing
client to net 144 Mbits/sec routing

net to client 124 Mbits/sec NAT
client to net 144 Mbits/sec NAT