From mboxrd@z Thu Jan  1 00:00:00 1970
From: corbet@lwn.net (Jonathan Corbet)
Date: Wed, 25 Sep 2013 01:15:20 -0600
Subject: [PATCH v2] [media] marvell-ccic: simplify and fix clk handling
 (a bit)
In-Reply-To: <1380049187-16036-1-git-send-email-u.kleine-koenig@pengutronix.de>
References: <1380048167-13729-1-git-send-email-u.kleine-koenig@pengutronix.de>
 <1380049187-16036-1-git-send-email-u.kleine-koenig@pengutronix.de>
Message-ID: <20130925011520.1fa24013@lwn.net>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Tue, 24 Sep 2013 20:59:47 +0200
Uwe Kleine-K?nig <u.kleine-koenig@pengutronix.de> wrote:

> The marvell-ccic does several things wrong or ineffectively in the clock
> handling and it's usage of the devm_* stuff
> 
>  - it assumes clk_get doesn't return NULL
>  - it explicitly calls devm_clk_put instead just keeping the reference
>    during it's lifetime and let the driver core call it
>  - it calls kfree, gpio_free and free_irq for resources it requested
>    using devm_kzalloc, devm_gpio_request and devm_request_irq
>    respectively.
>  - it mixes devm_ and unmanaged resources which probably results in a
>    race condition during remove

OK, all of that stuff was added this time around by Libin; my
understanding of that particular hardware is ... minimal.  The basic
idea of the patch seems sound.  I do note, though, that you've changed
the behavior of the driver somewhat.  The MIPI clock is current
obtained at power-up time and released on power-down; you've moved it
to probe time instead, and it's held for the lifetime of the driver.
Perhaps that's even better, I don't know...Libin, what do you say on
that?

The free_irq() call is also removed by a patch previously submitted by
Wei Yongjun.

> This patch fixes all but the last issue in this list. This patch doesn't
> introduce new reasons for not building, but there are already several
> bulid problems.

Care to report those?

Thanks,

jon