HiJohannes, On Tue, 2020-05-26 at 14:38 +0200, Johannes Berg wrote: > Hi James, > > > > I guess you could "fix" that by running with wmediumd? > > > > We do run with wmediumd, although I hadn't looking into this much. > > Well, I was sort of assuming that even just running wmediumd would > introduce the latency of another userspace roundtrip, and, perhaps > more > importantly, disconnect the code paths. > > I didn't really think you need any additional delay, and yes, I don't > think wmediumd supports adding additional delay, but it does add its > own > fake medium acquisition delay. > > So maybe that means my theory was wrong? I was actually incorrect thinking we used wmediumd, we don't :) We have our own hwsim thing that is effectively taking the place of wmediumd. And with this I added a very small delay to resending out the frames and it totally fixed the issue. Thanks, James > > johannes >