[Make-wifi-fast] hacking and slashing at the ath9k's retries

[Make-wifi-fast] hacking and slashing at the ath9k's retries

[Dave Taht]

I don't see a reason to have ATH_MAX_SW_RETRIES of 30. ?

ATH_TXMAXTRY appears to be unused.

I am also sufficiently bugged at the VI queues behavior to go and fix
it finally (which also ties into a few other things)

--- a/drivers/net/wireless/ath/ath9k/ath9k.h
+++ b/drivers/net/wireless/ath/ath9k/ath9k.h
@@ -91,8 +91,8 @@ int ath_descdma_setup(struct ath_softc *sc, struct
ath_descdma *dd,
 #define ATH_RXBUF               512
 #define ATH_TXBUF               512
 #define ATH_TXBUF_RESERVE       5
-#define ATH_TXMAXTRY            13
-#define ATH_MAX_SW_RETRIES      30
+#define ATH_TXMAXTRY            2
+#define ATH_MAX_SW_RETRIES      4

 #define TID_TO_WME_AC(_tid)                            \
        ((((_tid) == 0) || ((_tid) == 3)) ? IEEE80211_AC_BE :   \
d@dancer:~/git/linux/airtime-9$



-- 
Dave Täht
Let's go make home routers and wifi faster! With better software!
http://blog.cerowrt.org

Re: [Make-wifi-fast] hacking and slashing at the ath9k's retries

[Toke Høiland-Jørgensen]

Dave Taht <dave.taht@gmail.com> writes:

> I don't see a reason to have ATH_MAX_SW_RETRIES of 30. ?

Not really no, but in practice it doesn't make much of a difference:
Minstrel will build fairly long retry chains, so before ever getting to
the point where the max of 30 is tested against, a packet can be retried
a dozen times or more...

-Toke

Re: [Make-wifi-fast] hacking and slashing at the ath9k's retries

[Dave Taht]

On Sat, Nov 19, 2016 at 11:29 AM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
> Dave Taht <dave.taht@gmail.com> writes:
>
>> I don't see a reason to have ATH_MAX_SW_RETRIES of 30. ?
>
> Not really no, but in practice it doesn't make much of a difference:
> Minstrel will build fairly long retry chains, so before ever getting to
> the point where the max of 30 is tested against, a packet can be retried
> a dozen times or more...

That's next. :) I wanted to fix the !@#@! VI queue first.

Am looking over minstrel blues-also.

But while I'm hacking and slashing here, I can't think of any reason
why we'd ever want a non-aggregate queue depth of > 2. And I think the
complete_poll_interval is too long for short packets, but I'm not sure
what happens in terms of the jiffies conversion is.

-#define ATH_NON_AGGR_MIN_QDEPTH    8
-#define ATH_TX_COMPLETE_POLL_INT   1000
+#define ATH_NON_AGGR_MIN_QDEPTH    3
+#define ATH_TX_COMPLETE_POLL_INT   500


> -Toke



-- 
Dave Täht
Let's go make home routers and wifi faster! With better software!
http://blog.cerowrt.org