Date: Tue, 9 Apr 2013 09:18:38 -0700
From: Guenter Roeck <linux@roeck-us.net>
To: Wim Van Sebroeck <wim@iguana.be>
Cc: Samuel Ortiz <sameo@linux.intel.com>, linux-watchdog@vger.kernel.org,
        linux-kernel@vger.kernel.org,
        =?iso-8859-1?Q?P=E1draig?= Brady <P@draigBrady.com>
Subject: Re: mfd: Core driver for Winbond chips
Message-ID: <20130409161838.GB27050@roeck-us.net>
References: <1362957300-18223-1-git-send-email-linux@roeck-us.net>
 <20130322205258.GE7867@spo001.leaseweb.com>
 <20130323002810.GA26245@roeck-us.net>
 <20130323174914.GA21563@roeck-us.net>
 <20130409093700.GM24058@zurbaran>
 <20130409113639.GA8731@roeck-us.net>
 <20130409114544.GD7867@spo001.leaseweb.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20130409114544.GD7867@spo001.leaseweb.com>
User-Agent: Mutt/1.5.21 (2010-09-15)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1901
Lines: 37

On Tue, Apr 09, 2013 at 01:45:44PM +0200, Wim Van Sebroeck wrote:
> Hi Guenter,
> 
> > I was waiting for feedback from Wim, who submitted a similar driver, about his
> > thoughts. Key question is how to reserve access to the shared resource - either
> > through an exported function in the mfd driver requesting a mutex, or through
> > request_muxed_region(). I am going back and forth myself on which one is better.
> > 
> > Maybe it does not really matter, but using a function has the slight advantage
> > that it auto-loads and locks the mfd module while one of its client modules
> > is loaded. If we use request_muxed_region, that is not the case and the client
> > module must use another means to request and lock the mfd module.
> > 
> > Maybe you have an opinion ?
> 
> This is indeed the main issue that has to be solved. Both options will work.
> I like the auto-load and lock, but I need to look at the request_muxed_region
> code again first before I can see what the possible drawbacks are :-).
> 
One drawback of using request_muxed_region is that it needs a return value
from superio_enter. Also, it needs some code in the client driver init function
to ensure that the mfd driver gets loaded, and possibly a call to __module_get()
in the client driver probe function to keep the mfd driver loaded.

winbond_superio_enter() would not need a return value and could use
devm_request_region. We could also consider allocating the hwmon memory space in
the mfd driver and pass it as resource to the client drivers, which would remove
a few more lines of code from those.

Overall I am slightly in favor of using an exported function.

Guenter
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/