* Felipe Balbi [off-list ref] [120910 23:02]:

quoted

 
+#ifdef CONFIG_ARCH_OMAP1
+#define omap_kp_24xx()		0
+#else
+#define omap_kp_24xx()		1
+#endif

I would rather use revision detection or different driver names (if
revision register is broken).

Hmm actually looks like we can actually remove all the omap2+
support as we no longer have any users for this one. I think I
already converted the last one to matrix-keypad a while back.
 

quoted

+static struct omap_kp *omap_kp;

please don't. This will prevent multiple instances of this driver. Even
though I don't think we will ever have an omap with multiple keypad
instances, it's still not a good practice IMHO.

Also, this ends up being "hidden" (if you have a better work let me
know) in most functions since they either pass omap_kp as argument or
define a local omap_kp variable.

Yeah good point, I'll update that and remove the omap2+ support
for this driver.
 

Sourav, is the revision register on this IP working fine across multiple
OMAPs ?

Sounds like no need for that, as we're no longer using this for
omap2+..
 

quoted

@@ -253,9 +256,9 @@ static ssize_t omap_kp_enable_store(struct device *dev, struct device_attribute
 	mutex_lock(&kp_enable_mutex);
 	if (state != kp_enable) {
 		if (state)
-			enable_irq(INT_KEYBOARD);
+			enable_irq(omap_kp->irq);
 		else
-			disable_irq(INT_KEYBOARD);
+			disable_irq(omap_kp->irq);

GREAT!! :-)

Heh yeah that nice way to do it :)
 

quoted

 static int __devinit omap_kp_probe(struct platform_device *pdev)
 {
-	struct omap_kp *omap_kp;

???? I don't see the point for that global omap_kp, actually ...

Yes you're right. Will send an updated one tomorrow.

Regards,

Tony