From mboxrd@z Thu Jan  1 00:00:00 1970
From: Michal Kazior <michal.kazior@tieto.com>
Date: Mon, 29 Apr 2013 08:07:43 +0200
Subject: [ath9k-devel] [PATCH 20/20] ath10k: Remove unneeded locks
 during HTT RX attach
In-Reply-To: <1367163098-22787-20-git-send-email-sujith@msujith.org>
References: <1367163098-22787-1-git-send-email-sujith@msujith.org>
	<1367163098-22787-20-git-send-email-sujith@msujith.org>
Message-ID: <517E0E2F.7000408@tieto.com>
List-Id: <ath9k-devel.lists.ath9k.org>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: ath9k-devel@lists.ath9k.org

On 28/04/13 17:31, Sujith Manoharan wrote:
> From: Sujith Manoharan <c_manoha@qca.qualcomm.com>
>
> We can dispense with the lockdep assert since the refill
> timer just calls the replenish routine which is held under
> the ring lock.

Doesn't sound good to me.

The lockdep is there for a reason. This explicitly states the function 
must be under the htt rx lock and makes sure this is verified. This 
should not be removed.

If (which can be arguably near impossible) the initial 
ath10k_htt_rx_ring_fill_n() in htt rx attach fires of a timer and we get 
a crazy pre-emption between mod_timer() and goto we will
corrupt the htt rx ring state without the lock here.


-- Pozdrawiam / Best regards, Michal Kazior.