On Fri, Jul 13, 2012 at 11:17:31AM +0200, Wolfram Sang wrote:

On Fri, Jul 13, 2012 at 12:09:01PM +0530, Jayachandran C wrote:

quoted

From: Ganesan Ramalingam <ganesanr-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>

Deprecate 'regstep' property and use the standard 'reg-shift' property
for register offset shifts. 'regstep' will still be supported as an
optional property, but will give a warning when used.

Signed-off-by: Ganesan Ramalingam <ganesanr-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
Signed-off-by: Jayachandran C <jayachandranc-oSioyQM9ZPnuBjGU1YDckgC/G2K4zDHf@public.gmane.org>
---

[...]

quoted

+		}
+	} else
+		i2c->reg_shift = val;

Now you are assigning an u32 to an int. Regstep will never be really a
huge number, but still...

Also, braces around else-block according to coding style.

quoted

 
-	val = of_get_property(pdev->dev.of_node, "clock-frequency", NULL);
-	if (!val) {
+	if (of_property_read_u32(np, "clock-frequency", &val)) {
 		dev_err(&pdev->dev,
 			"Missing required parameter 'clock-frequency'\n");
 		return -ENODEV;
 	}
-	i2c->clock_khz = be32_to_cpup(val) / 1000;
+	i2c->clock_khz = val / 1000;

Okay, "/ 1000" will guarantee that the divided value will fit into an
int. Yet, what do we gain by not using u32?

I will send out a new version of this patch using u32. In this case I
felt that there is no need to use a fixed width type when an int would
do.

Regards,
JC.