Problems with ath at kern.hz=100

[Bengt Ahlgren]

Bengt Ahlgren <bengta at sics.se> writes:

> Sam Leffler <sam at freebsd.org> writes:
>
>> Bengt Ahlgren wrote:
>>> Sam Leffler <sam at freebsd.org> writes:
>>>
>>> [...]
>>>
>>>   
>>>> You've removed all the context of the original problem; I can't recall
>>>> what you were trying to fix.  The ps q drops might be caused by a bug
>>>> that was fixed (I think in RELENG_7).  I don't see what version of
>>>> code you're running so can't tell.
>>>>     
>>>
>>> Sorry, a summary of the problem:
>>>
>>> The system frequently comes into a state where outgoing packets are
>>> held somewhere.  The condition persists for seconds to minutes.  It
>>> goes away by itself, but some of the packets are usually lost.  This
>>> occurs on ath, FreeBSD 7.1-REL, no SMP, with kern.hz=100 (but not with
>>> hz=1000).  Original mail with description can be found here:
>>>
>>> http://lists.freebsd.org/pipermail/freebsd-mobile/2009-February/011343.html
>>>
>>> I have done some more investigation, inserting debug prints in
>>> ath_start (if_ath.c 1.177.2.2.2.1), one just after IFQ_DRV_DEQUEUE
>>> (line 1576), another just before the call to ieee80211_pwrsave (line
>>> 1614).
>>>
>>> Packets are dequeued, but ieee80211_pwrsave is called and nothing
>>> happens.  Below is the output from "athdebug xmit" with these extra
>>> two DPRINTF:s during the queue-up.  Then after a while packets are
>>> sent again.
>>>
>>> Bengt
>>>
>>>   
>> <...debug output removed...>
>>
>> Did you try RELENG_7? ap mode power save was fixed post 7.1 release.
>
> I can do that, but I am not using ap mode.  Will it still be useful?

Hmm, I see no difference in ath between 7.1-REL and RELENG_7 (besides
CVS ID strings)???  Are you meaning HEAD?  If so, can I just drop in
src/sys/dev/ath from HEAD in 7.1R?

Bengt