Hi Stefan, You are completely right. I responded shortly only to confirm this finding and it was not meant as a kind of bug report.
I will dive deeper into the issue when I have the time for that. Thanks for Your work and response. On 07 Nov 22:12, Stefan Sperling wrote: > On Tue, Nov 07, 2017 at 08:57:32PM +0000, Uwe Werler wrote: > > I can confirm the massively decreased throughput. I upgraded my Alix now > > from > > 6.1 to 6.2. In 6.1 in both modes 11n and 11g I got at least ~16 Mbit > > measured > > via tcpbench in 6.1 in both directions. Now with 6.2 I get ~4,5 MBit only. > > Issue noted, stamped, and shelved. > > I am sorry to say this but both of your reports are rather useless. > There is nothing anyone could do based on such statements. > All that could be done is guessing, which is a waste of time. > > Run some serious experiments, obtain some understanding of how this stuff > works (anyone can download the spec for free from IEEE or read an introductory > book) and give us concrete data which shows the before/after delta. > Make sure to account for environmental conditions. Document your findings > with captures of frames (data, management, control, and action frames) from > the air, such that we can use tcpdump or wireshark to view your captures > and learn what's actually going on in your wifi before/after. > > I am not saying that I do not believe you. There are bugs in this driver, > no doubt. But the known remaining problems in athn(4) have reasons which > are so far entirely opaque to me. It may take some serious and time consuming > debugging at the lower levels to figure them out. We may need to compare > every line of code with Linux/FreeBSD and find the crucial differences. > This hardware uses no firmware which means it exposes all the complicated > details. Which makes debugging rather hard. > > I don't plan to volunteer my time doing such tedious work in the near future. > My athn APs work fine for me. > > If the situation is really unbearable for you, well, you have access to > all the hardware and code and project history you need to help us fix it. > > The fact that neither of you have taken time to try to even just find a > specific commit up to 6.2 which made it worse for you already indicates > you're not going to invest much time into this. > And if you're not going to, then why should anyone else? > --
