-
Notifications
You must be signed in to change notification settings - Fork 122
Description
Hello everyone,
I have found a bug with sockperf running a pp test (udp mode).
This is the setup on my linux machine where sockperf is running with this command:
sockperf sr --daemonize
eth0: <BROADCAST,MULTICAST,UP,LOWER_UP>
inet 10.68.64.254/22 brd 10.68.67.255 scope global eth0
eth3: <BROADCAST,MULTICAST,UP,LOWER_UP>
inet 33.56.27.126/28 brd 33.56.27.127 scope global eth3
inet 33.56.27.125/32 scope global eth3
default via 10.68.64.1 dev eth0 metric 1
33.56.27.112/28 dev eth3 proto kernel scope link src 33.56.27.126
If I try to run a pp test from a client with this command:
sockperf pp -m 64 -i 10.68.64.254 --time 10 --full-rtt --mps=max
everything is fine: I can see the correct source and destination ip addresses in the udp datagrams.
Instead if I run this command from the client:
sockperf pp -m 64 -i 33.56.27.126 --time 10 --full-rtt --mps=max
I can see that the sockperf server is sending out the udp reply datagrams with the wrong source ip address: 10.68.64.254 instead of 33.56.27.126
The tcp test mode instead is fine. Indeed running the sockeperf server with this command is a workaround:
sockperf sr -i 33.56.27.126
Would be fine to have this bug fixed.