I'll see if I can suss anything out this week. I'll compile and test on GENERIC as well.
On Tue, Dec 5, 2017 at 10:13 AM, Stefan Sperling <[email protected]> wrote: > On Tue, Dec 05, 2017 at 09:38:07AM -0700, Base Pr1me wrote: > > Testing environment: 3 APUs with same SSID, 2 5GHz channels and 1 2.4GHz > channel. > > > > Test 1: Boot lappy, connected to 2.4GHz spot that I was closest to. > Walked > > to 5GHz spot and waited. It switched fine. Walked to second 5GHz spot and > > waited. It did not switch. Waited longer. It did not switch. I had to > start > > manually scanning APs to get it to switch. I was not checking routing. I > > should have. > > > > Test 2: Reboot lappy next to 5GHz switch. It connected to the 2.4GHz > spot, > > which had ~35% less signal strength. Commenced checking web mail. > Waited. It > > switched to the AP I was standing by. Kept checking email. Walked back by > > 2.4GHz AP and waited. It switched. Routing data stopped. Waited. > Initiated > > `route show` and it just hung there. Had to do a manual scan and routing > > started working again. > > There seems to be an unrelated bug which is rearing its head during > these tests. 'route show' doing nothing indicates there may be some > deeper issue involved. Other parts of the network stack are likely > taking part in the observed behaviour. > > Similar issues have been reported before I published this diff, where > an iwm interface got stuck doing nothing until ifconfig down/up. > > It's unfortunate that this problem shows up at the same time as I > start working on background scan because it interferes with the > primary goal of this diff (seamless operation). > > Still unsure what's going on, but I will try to track it down. > Can you show me a way to reproduce the hangs reliably? > > Does it also happen on GENERIC instead of GENERIC.MP? >
