Hi there,
I'm running imx-kernel 5.15.32 on a i.MX 8M Plus based custom board.
I have 2 nics:
# ethtool -i eth0
driver: fec
version: 5.15.60-tq+g03954e24f9ca-rt49-t
firmware-version:
expansion-rom-version:
bus-info: eth0
supports-statistics: yes
supports-test: yes
supports-eeprom-access: no
supports-register-dump: yes
supports-priv-flags: no
# ethtool -i eth1
driver: st_gmac
version: Jan_2016
firmware-version:
expansion-rom-version:
bus-info:
supports-statistics: yes
supports-test: no
supports-eeprom-access: no
supports-register-dump: yes
supports-priv-flags: no
I'm experiencing poor results (average, and mdev) when sending ping to eth0, and better results when pinging eth1:
ping eth0 results:
...
...
64 bytes from 90.0.0.2: icmp_seq=34463 ttl=63 time=1.02 ms
64 bytes from 90.0.0.2: icmp_seq=34464 ttl=63 time=0.949 ms
--- 90.0.0.2 ping statistics ---
100000 packets transmitted, 99999 received, 0.001% packet loss, time 107149ms
rtt min/avg/max/mdev = 0.131/0.852/12.708/0.388 ms, pipe 2
ping eth1 results
...
...
64 bytes from 90.0.0.2: icmp_seq=34463 ttl=63 time=0.447 ms
64 bytes from 90.0.0.2: icmp_seq=34464 ttl=63 time=0.401 ms
--- 90.0.0.2 ping statistics ---
100000 packets transmitted, 100000 received, 0% packet loss, time 100018ms
rtt min/avg/max/mdev = 0.205/0.411/8.626/0.045 ms
Why is eth0 ping time double than eth1?
What can I do in order to achieve better performance with eth0?
Thanks a lot,
Nir.
Hi,
I did some ping tests on the NXP i.mx8m-plus EVK with realtime edge 2.2.0. Numbers are average over 500 pings
ETH0 - 1.162ms
ETH1 - 0.312ms
Any comment on this issue would be great.
Thanks,
Nir.
Hello @NG_R3
The ENTE_QoS is assigned to eth1 instance. Only eth1 support TSN feature, you can refer this knowledge base:
https://community.nxp.com/t5/i-MX-Processors-Knowledge-Base/AVB-TSN-demo-on-i-MX8MP/ta-p/1123791
The ENET_QoS IP about eth1 support hardware multi-queue and traffic control .
Hi Qmiller,
Thanks for your reply. I understand that ENET_QoS is attached to eth1, but my question regards eth0.
I did some ping tests on the NXP i.mx8m-plus EVK with realtime edge 2.2.0.
ETH0 - 1.162ms
ETH1 - 0.312ms
These numbers are similar on any i.MX8M Plus device that I could test. Custom boards and NXP's own EVK, running NXP's realtime edge.
Why is eth0 ping time double than eth1?
What can I do in order to achieve better performance with eth0?
Thanks a lot,
Nir.