[PATCH] Input: marix-keymap - automatically allocate memory for keymap

In device tree enabled setups requiring preallocated memory for storing keymap
is quite often awkward, so let's provide an option of allocating it directly
in matrix_keypad_build_keymap().

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
---
 drivers/input/matrix-keymap.c | 22 +++++++++++++++++++++-
 1 file changed, 21 insertions(+), 1 deletion(-)

diff --git a/drivers/input/matrix-keymap.c b/drivers/input/matrix-keymap.c
index 443ad64b..da04f13 100644
--- a/drivers/input/matrix-keymap.c
+++ b/drivers/input/matrix-keymap.c

@@ -122,6 +122,11 @@ static int matrix_keypad_parse_of_keymap(const char *propname,
  * it will attempt load the keymap from property specified by @keymap_name
  * argument (or "linux,keymap" if @keymap_name is %NULL).
  *
+ * If @keymap is %NULL the function will automatically allocate managed
+ * block of memory to store the keymap. This memory will be associated with
+ * the parent device and automatically freed when device unbinds from the
+ * driver.
+ *
  * Callers are expected to set up input_dev->dev.parent before calling this
  * function.
  */

@@ -132,12 +137,27 @@ int matrix_keypad_build_keymap(const struct matrix_keymap_data *keymap_data,
 			       struct input_dev *input_dev)
 {
 	unsigned int row_shift = get_count_order(cols);
+	size_t max_keys = rows << row_shift;
 	int i;
 	int error;
 
+	if (WARN_ON(!input_dev->dev.parent))
+		return -EINVAL;
+
+	if (!keymap) {
+		keymap = devm_kzalloc(input_dev->dev.parent,
+				      max_keys * sizeof(*keymap),
+				      GFP_KERNEL);
+		if (!keymap) {
+			dev_err(input_dev->dev.parent,
+				"Unable to allocate memory for keymap");
+			return -ENOMEM;
+		}
+	}
+
 	input_dev->keycode = keymap;
 	input_dev->keycodesize = sizeof(*keymap);
-	input_dev->keycodemax = rows << row_shift;
+	input_dev->keycodemax = max_keys;
 
 	__set_bit(EV_KEY, input_dev->evbit);
 

-- 
1.7.11.7


-- 
Dmitry

On Thu, 8 Nov 2012 08:45:21 -0800
Dmitry Torokhov [off-list ref] wrote:

In my tree GFP_KERNEL isn't defined at that point. Is an include
missing in this patch or is my tree (3.6+some next cherry picks)
missing a commit?

+		if (!keymap) {
+			dev_err(input_dev->dev.parent,
+				"Unable to allocate memory for keymap");
+			return -ENOMEM;
+		}
+	}
+
 	input_dev->keycode = keymap;
 	input_dev->keycodesize = sizeof(*keymap);
-	input_dev->keycodemax = rows << row_shift;
+	input_dev->keycodemax = max_keys;
 
 	__set_bit(EV_KEY, input_dev->evbit);
 

Otherwise that look good to me.

Alban

Hi Alban,

On Wed, Nov 14, 2012 at 10:40:59AM +0100, Alban Bedel wrote:

On Thu, 8 Nov 2012 08:45:21 -0800
Dmitry Torokhov [off-list ref] wrote:

quoted

+	if (!keymap) {
+		keymap = devm_kzalloc(input_dev->dev.parent,
+				      max_keys * sizeof(*keymap),
+				      GFP_KERNEL);

In my tree GFP_KERNEL isn't defined at that point. Is an include
missing in this patch or is my tree (3.6+some next cherry picks)
missing a commit?

No, I apparetnly had an older tree when I created a patch, i have since
added include/linux/gfp.h

quoted

+		if (!keymap) {
+			dev_err(input_dev->dev.parent,
+				"Unable to allocate memory for keymap");
+			return -ENOMEM;
+		}
+	}
+
 	input_dev->keycode = keymap;
 	input_dev->keycodesize = sizeof(*keymap);
-	input_dev->keycodemax = rows << row_shift;
+	input_dev->keycodemax = max_keys;
 
 	__set_bit(EV_KEY, input_dev->evbit);
 

Otherwise that look good to me.

Great, I am putting you down as reviewed-by then.

Thanks.

-- 
Dmitry