Optimising the Omnia for gigabit wired throughput

Software SW tweaks
41

Seems to work only with netdev but not iface …

If you want you could install the sysstat package, once installed service sysstat start and then sar -n EDEV 1 30 --iface=wan,wan6,eth2 during the test (1 = intervall in sec | 30 = duration in sec).

42 
Average:        IFACE   rxerr/s   txerr/s    coll/s  rxdrop/s  txdrop/s  txcarr/s  rxfram/s  rxfifo/s  txfifo/s
Average:         eth2      0.00      0.00      0.00      0.00      0.00      0.00      0.00      0.00      0.00
43

did you exclude wan | wan6? It should be working this way

44

no i did not.
I used this command
sar -n EDEV 1 30 --iface=wan,wan6,eth2

45

Hmm, well only netdev then again

Anyway, that there are no drops or errors showing could mean:

  • kernel does not account error | drop frames due to some unset kconfig flag and those statistics are not available - but I am not sure whether there is a particular flag (similar to Where are iftop, iotop, nethogs and friends?)
  • frames not being dropped by the TO kernel
  • frames are somehow (erroneously) dropped by (potentially buggy) port driver (mvneta) and not reported to the kernel - but then it should likely happen also with SFP setup
  • frames being dropped by the modem for some reason

Not sure whether worth the effort, a packet dump audit might yield some insight. If you are familiar with wireshark you could could do a remote ssh capture from the W10 desktop on TO’s eth2 port and see what happens to packets during the WAN iperf3 test.
Same for the modem (port) if it provides a ssh server.

46

Hello,
I just want to say that whatever change was made in the last 2 weeks to the hbk branch did fix it for me. I get now full throughput

1 Like