Hello,
I am struggling to get Qbv working as expected on bsp 0.3.
I have connected eno0 ("MAC0") of two LS1028A ARDB's and a linux PC together in a LAN (with a dumb switch), and established a VLAN 1 on all such that:
LS1028ARDB #1: eno0.1 has ip 195.168.1.1/24
LS1028ARDB #2: eno0.1 has ip 195.168.1.2/24
PC: eth0.1 has ip 195.168.1.3/24
vconfig set egress_map and vconfig_set_ingress_map has been run all all such that all traffic on the VLAN ends up with a VLAN priority of 4.
Pinging the PC from either LS1028ARDB confirms in wireshark on the PC that the ICMP request and reply packets do indeed have the correct VLAN ID and priority. Similarly tcpdump run on both LS1028ARDB's show that ICMP requests arrive with the correct VLAN ID and priority.
In addition all the above equipment can ping each other on a 10.0.1.0 net (no VLAN).
A qbv.txt file has been created with the following contents:
t0 00000000b 100000
t1 10111110b 1000000000
This has been entered into tsntool as such:
tsntool>qbvset --device eno0 --entryfile /qbv.txt
tsntool>qbvget --device eno0 confirms that setup has been set
(btw: doesn't work on switch ports).
With the above setup no traffic flows whatsoever. I would expect traffic to flow on my VLAN 1 with priority 4.
If it is changed to:
t0 00000001b 100000
t1 10111110b 1000000000
then traffic on all traffic classes flow - all with the same LOW bandwidth. I would expect VLAN 1 with priority 4 to have a much better bandwidth than traffic class 0.
If it is changed to:
t0 00000001b 1000000000
t1 10111110b 100000
then traffic on all traffic classes flow - all with the same HIGH bandwidth. I would expect VLAN 1 with priority 4 to have a much lower bandwidth than on traffic class 0.
All experimentation show far show that the switch does not distinguish between traffic classes at all. It behaves as if gate bit 0 and period in the qbv.txt file are the only values that matter. Gates bit 1-7 are don't-care.
What am I missing ?
Extremely grateful for any help ... please ...