Hi Abhi,

On 3/10/2011 9:06 AM, Abhijeet Dharmapurikar wrote:

Mark Brown wrote:

quoted

On Mon, Mar 07, 2011 at 10:09:48PM -0800, adharmap at codeaurora.org wrote:

quoted

+    mpp_chip->gpio_chip.label = PM8XXX_MPP_DEV_NAME;
+    mpp_chip->gpio_chip.direction_input = pm8xxx_mpp_dir_input;
+    mpp_chip->gpio_chip.direction_output = pm8xxx_mpp_dir_output;
+    mpp_chip->gpio_chip.to_irq = pm8xxx_mpp_to_irq;
+    mpp_chip->gpio_chip.get = pm8xxx_mpp_get;
+    mpp_chip->gpio_chip.set = pm8xxx_mpp_set;
+    mpp_chip->gpio_chip.dbg_show = pm8xxx_mpp_dbg_show;
+    mpp_chip->gpio_chip.ngpio = pdata->core_data.nmpps;
+    mpp_chip->gpio_chip.can_sleep = 1;
+    mpp_chip->gpio_chip.dev = &pdev->dev;
+    mpp_chip->gpio_chip.base = pdata->mpp_base;

It's seems really odd that you're adding gpiolib stuff here when you've
also got a separate gpiolib driver.  Possibly this all shouldn't
actually be split up as much as it is - there's also the issue with the
gpiolib driver needing to peer into the interrupt controller.  It might
simplify the code if things were merged more.

I dont think merging code will help here. gpio lines and mpp lines are very different piece of hardware,they have a different register map and different config attributes. They do fall under the generic 'gpiolib' umbrella, but it seems clean to keep them separate.

I agree with you that gpio code needs to call on interrupt controller ( a complicated hardware design) but I think I have a clean software implementation to manage it.

It would be good to add that MPP do have more functionality than normal GPIO. So, I expect that in future
we might have few exported APIs beyond gpiolib integration. 

---Trilok Soni

-- 
Sent by a consultant of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.