On Mon, Feb 05, 2018 at 11:07:18AM +0000, Kevin Chadwick wrote: > On Mon, 5 Feb 2018 11:21:27 +0100 > > However, I can still reproduce device timeouts easily by running > > tcpbench through an AR9271 in hostap mode. I don't know yet what's > > causing the problem. > > (Not using these for production btw) > > I have unplugged my AR9271 due to these timeouts. Due to the chromecast > (below) I have a so far unneeded script now to bring it down and up > via a cronjob every minute upon a timeout in /var/log/messages. Pretty > simple but I can provide it if anyone wants it. > > I get very few timeouts on AR5008-3NG (AR5416+AR2133) but seemed to get > many suddenly with a chromecast and the athn would not recover like > usual. It seems to have stopped and since Google update chromecasts > silently I can't tell why or test further and so haven't said anything. > > The AR9271 has TxR:S of 1:1:1 compared to 3x3:2, perhaps it has a > smaller buffer or just a more constrained power supply like you > previously mentioned that struggles with many RX/TX? > > The unconfirmed possible cause was a flood of UDP MDNS packets.
Our athn(4) driver is not running periodic calibration. Which is very bad and can result in problems at the PHY layer such as eventual deafness and inability to transmit. That's the one problem to solve before looking for other causes.
