On Mon, Mar 26, 2012 at 09:26:01AM -0600, Stephen Warren wrote:

On 03/26/2012 03:45 AM, Viresh Kumar wrote:

quoted

matrix_keyboard_of_fill_keymap() routines allocates memory using kzalloc()
routine. It is freed on call to matrix_keyboard_of_free_keymap() routine.

This patch converts these kzalloc() calls to devm_kzalloc() and thus we don't
require matrix_keyboard_of_free_keymap() anymore.

Signed-off-by: Viresh Kumar <redacted>

The Tegra-related parts,
Acked-by: Stephen Warren <redacted>

But that said, Dmitry specifically requested that the devm_* function
not be used during initial review of this code; IIRC he preferred just
doing the explicit frees. I can't find a link right now though.

Right, there is no need to haul allocated memory past the building of
the "real" keymap anyway.

Actually I want to change the interface so that we could do something
like:

	if (pdata->keymap) {
		matrix_keypad_build_keymap(...)
	} else {
		error = matrix_keypad_build_of_keymap();
		if (error)
			goto blah;
	}

So that one could override OF data with plaform data if needed and no
freeing of intermediate platform-like keymap structure is requited.
Or maybe encapsulate OF parsing directly into matrix_keypad_build_keymap
if passed keymap data is NULL...

Thanks.

-- 
Dmitry