[ath][CURRENT] Triangle-shaped network performance

[Adrian Chadd]

On 25 March 2012 23:03, Lytochkin Boris <lytboris at gmail.com> wrote:

>> 10 beacons a second is fine. 100 beacons a second ?
> bintval = 1000 is 1 beacon per second. Actually I was sure that Cisco
> AP have this setting at 3k.

Hm. Weird. 1 a second seems a bit infrequent. I'm not sure what/why
this is a problem. I haven't really thought about it.. :)


>> That almost feels like some weird interrupt handling issue.
> Any ideas how to proove this theory?

Not immediately, no. I'd start by enabling interrupt debugging and see
how often RX/TX completion interrupts are coming in.


Adrian