... whoa, no I didn't write the FreeBSD driver by scratch. The FreeBSD driver and HAL more closely resembles the internal codebase and I've made a point of trying to not have the HAL drift very much. I mostly wrote the 11n TX aggregation from scratch but I definitely can't take credit for all the work from Sam Leffler and the team of engineers at Atheros who worked on that HAL and driver. :-)
Now, as for the reason: It's quite possible there's a totally logical reason for the AR5416 NF cal pauses that go on in ath9k. The driver loops over the AGC register to see if the NF cal completed or not. It may or may not take a long time. If it's consistently taking a long time it may stop other tasks from running. I haven't checked to see whether the NF cal code is running inside a spinlock_bh() context which would block other tasklets from running on that CPU. That could be causing you some grief. Adrian _______________________________________________ ath9k-devel mailing list ath9k-devel@lists.ath9k.org https://lists.ath9k.org/mailman/listinfo/ath9k-devel