MIME-Version: 1.0
In-Reply-To: <b6c5339f0907311211w4d810b23uc6ec8027924da647@mail.gmail.com>
References: <20090731181020.GE7963@makis>
	 <43e72e890907311115h56df6ac0p1ea7165e305a10ee@mail.gmail.com>
	 <40f31dec0907311125o1654bd23jee861ba8bea57611@mail.gmail.com>
	 <b6c5339f0907311211w4d810b23uc6ec8027924da647@mail.gmail.com>
Date: Fri, 31 Jul 2009 22:52:42 +0300
Message-ID: <40f31dec0907311252n76adfde7kb9f34bf8c6baf34c@mail.gmail.com>
Subject: Re: [ath5k-devel] [PATCH 4/4] ath5k: Use SWI to trigger calibration
From: Nick Kossifidis <mickflemm@gmail.com>
To: Bob Copeland <me@bobcopeland.com>
Cc: "Luis R. Rodriguez" <mcgrof@gmail.com>,
	ath5k-devel@lists.ath5k.org, linux-wireless@vger.kernel.org,
	linville@tuxdriver.com, jirislaby@gmail.com, nbd@openwrt.org
Content-Type: text/plain; charset=UTF-8
Sender: linux-wireless-owner@vger.kernel.org

2009/7/31 Bob Copeland <me@bobcopeland.com>:
> On Fri, Jul 31, 2009 at 2:25 PM, Nick Kossifidis<mickflemm@gmail.com> wrote:
>>    Also we don't need to run calibration if we are idle (no interrupts).
>
> I think this is the big win right now...
>
>> c) Having calibration on a tasklet is better since during calibration
>>    we can't transmit or receive (antennas are detached to measure
>>    noise floor), previously calibration could run in parallel with tx/rx
>>    and interfere (packet loss).
>
> This can still happen, no?  Two tasklets can run in parallel on
> different processors, as long as they are different tasklets.
>
> In practice, this won't happen much because tasklets run on the
> cpu that scheduled them, and irq affinity is such (at least on my
> hardware) that it's almost always the same CPU.  But I think to
> make the above true it needs to stop the queues etc when doing
> calibration.
>

ACK, how do we do that ?
ieee80211_stop_queues / ieee80211_wake_queues

> Also we are missing tasklet_kill()?
>

Yup, resending...


-- 
GPG ID: 0xD21DB2DB
As you read this post global entropy rises. Have Fun ;-)
Nick