On a machine acting as a bridge, I have only one dummynet pipe configured with PLR. I add two rules, hoping that I will generate symmetric packet loss on UL & DL for a single host behind the bridge. However, I end up with double the packet loss in UL, while in DL the packet loss is exacty what I configured. And I can't figure out at all where my mistake lies... my config: Code:ipfw pipe 7 config plr 0.004 #--> 0.4%, right? ipfw add 300 pipe 7 out dst-ip 10.x.y.z/30 #--> I also tried only IP@ 10.x.x.2, for the # specific host. it's anyway alone in that subnet ipfw add 400 pipe 7 out src-ip 10.x.y.z/30 # obviously x,y,z identical to the above ones I test with iperf, one linux and one windows machine (first command line for the server, second line for client): Code:iperf -s -u -l 1400 iperf -c 10.x.y.z -f m -u -i 5 -b 3m -l 1400 -t 100 (tried with other values for bandwidth, no difference. need -t 100 for a good average since my PLR is very low). -when running this in one direction, my PLR is the desired one: ~0.4% (0.39% to 0.41%). -when running this in the other direction (I switch my iperf commands and adjust the IP@ for the server, the machines remain on the same side of the bridge, just the iperf server/client are switched) I get double the packet loss: ~0.8% (0.76%-0.85%). same result if I run iperf with the -r command (also for reversed direction testing). If I put prob 0.5 to the 400 rule (which is the problematic one), I get the correct result of 0.4% --> which is strange... I checked that only one rule is matched when running the iperf test (via ipfw show, I see what rule increments packet count) and also that the pipe is configured with the correct value (ipfw pipe show). if I change rule 400 to: Code:ipfw add 400 pipe 7 in src-ip 10.x.y.z/30 it is not matched at all... which is again weird. I get then PLR 0% and no packets hit the rule. I use a dual port Intel card (em0, em1) configured as a bridge, no IP@ for any of the ports. sysctl.conf: Code:net.link.ether.ipfw=0 net.link.bridge.ipfw=1 It must be smtg obvious, but I cannot see it.