kern/175446: [ath] high volumes of PHY errors lead to BB/MAC hangs and resets

[Adrian Chadd]

>Number:         175446
>Category:       kern
>Synopsis:       [ath] high volumes of PHY errors lead to BB/MAC hangs and resets
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    freebsd-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Sun Jan 20 06:30:00 UTC 2013
>Closed-Date:
>Last-Modified:
>Originator:     Adrian Chadd
>Release:        
>Organization:
>Environment:
>Description:
There seem to be some situations where high volumes of PHY errors lead to BB and MAC hangs.

These have been observed on the AR9220/AR9280 and later chips. It doesn't seem to affect the AR5416/AR9160 for some reason (and not the earlier chips as well.)

This PR will track the changes made to try and trace down and fix the specific issue(s) surrounding this particular issue.
>How-To-Repeat:
Enable radar/spectral scan and do some traffic.
>Fix:
The specific fixes I've been told about:

* AR9280, AR9285, AR9287 - don't do PHY error accounting, use the hardware counters;
* The bug here has to do with zero-length DMA frames and some issues with them in the RX FIFO;
* Radar and spectral scan frames don't cause issues - they're not zero-length payload frames.

>Release-Note:
>Audit-Trail:
>Unformatted: