On Fri, May 12, 2017 at 11:14:35AM +0100, Caolan McMahon wrote: > >> Please test -current. > >> > >> Or you could try the following diffs on a 6.1 source tree but I have > >> not tested that and don't want to support it, so you are on your own. > >> https://marc.info/?l=openbsd-tech&m=149399149502787&q=raw > >> https://marc.info/?l=openbsd-tech&m=149399299003411&q=raw > > Thanks Stefan, just knowing there are relevant commits in -current is good. > > > > Come to think of it, an easy workaround might be to disable 11n mode: > > > > ifconfig iwm0 mode 11g > > > > That should make it work in 6.1. > > 11g has better range than 11n in 6.1 (the first patch above addresses this). > > You are probably far away from the AP, right? > > I'm reasonably far away from the AP now, but I've also tested sitting > right next to the AP and I hit the same speed issue. > I'm also running in 11g mode at the moment in the hope that it would > help - I have subjectively noticed fewer stalls but networking > continues to be slow.
For now, I won't be able to help you beyond what I have already suggested. I am quite sure that the driver itself is generally fine, though it may have problems in specific cases. Please try other environments and access points with the same laptop and I hope you'll find that it can perform well. It can be slow for any number of reasons and it's hard to debug remotely. Any bugs left in the driver at this stage are quite subtle and will require details to figure out (that means I would need lot more than "it is slow", such as captured frame exchanges which demonstrate a particular problem). Perhaps it will become better in future releases if more parts of the 802.11n spec get implemented since our current implementation is still incomplete. When you test the Linux driver you're testing a much more complete 802.11n implementation so one can't really draw a direct comparison.