[2/2] Input: nomadik-ske-keypad - start using the apb_pclk

Commit Message

Ulf Hansson Nov. 1, 2012, 2:20 p.m. UTC

From: Ulf Hansson <[e-mail]>

Previously this clock was handled internally by the clockdriver,
but now this is separate clk. So we need take care of it.

Signed-off-by: Ulf Hansson <[e-mail]>
---
 drivers/input/keyboard/nomadik-ske-keypad.c |   22 ++++++++++++++++++++--
 1 file changed, 20 insertions(+), 2 deletions(-)

Comments

Linus Walleij Nov. 4, 2012, 6:12 p.m. UTC | #1

On Thu, Nov 1, 2012 at 3:20 PM, Ulf Hansson <[e-mail]> wrote:

> From: Ulf Hansson <[e-mail]>
>
> Previously this clock was handled internally by the clockdriver,
> but now this is separate clk. So we need take care of it.
>
> Signed-off-by: Ulf Hansson <[e-mail]>

So this is a silicon block clock and falls into the category of things
we've been discussing...

If I understand correctly, the only "real" solution is to
implement the PM domains and have these enable the clocks.

An alternative may be to move this driver over to the AMBA bus,
because I think this device actually has primecell registers.
Then the bus will take care of the pclk for starters.

Yours,
Linus Walleij

Ulf Hansson Nov. 5, 2012, 12:25 p.m. UTC | #2

On 4 November 2012 19:12, Linus Walleij <[e-mail]> wrote:
> On Thu, Nov 1, 2012 at 3:20 PM, Ulf Hansson <[e-mail]> wrote:
>
>> From: Ulf Hansson <[e-mail]>
>>
>> Previously this clock was handled internally by the clockdriver,
>> but now this is separate clk. So we need take care of it.
>>
>> Signed-off-by: Ulf Hansson <[e-mail]>
>
> So this is a silicon block clock and falls into the category of things
> we've been discussing...
>
> If I understand correctly, the only "real" solution is to
> implement the PM domains and have these enable the clocks.

Agree.

Although, since the pm_domain not yet exist, this as a way forward for
now - to fix what is broken.
When the pm_domain is in place and when we decide to fold in the clock
handling in there, we can move it.

>
> An alternative may be to move this driver over to the AMBA bus,
> because I think this device actually has primecell registers.
> Then the bus will take care of the pclk for starters.

You are definitely right, this driver can be converted into using the AMBA bus.
Although, do you think that should be done _instead_ of going ahead
with this patch or do you want that to be handled as a next and a
separate step?

Kind regards
Ulf Hansson

Linus Walleij Nov. 5, 2012, 12:37 p.m. UTC | #3

On Mon, Nov 5, 2012 at 1:25 PM, Ulf Hansson <[e-mail]> wrote:
> On 4 November 2012 19:12, Linus Walleij <[e-mail]> wrote:

>> If I understand correctly, the only "real" solution is to
>> implement the PM domains and have these enable the clocks.
>
> Agree.
>
> Although, since the pm_domain not yet exist, this as a way forward for
> now - to fix what is broken.

True. Fixing regressions is more important.
Acked-by: Linus Walleij <[e-mail]>

> When the pm_domain is in place and when we decide to fold in the clock
> handling in there, we can move it.

OK for me.

>> An alternative may be to move this driver over to the AMBA bus,
>> because I think this device actually has primecell registers.
>> Then the bus will take care of the pclk for starters.
>
> You are definitely right, this driver can be converted into using the AMBA bus.
> Although, do you think that should be done _instead_ of going ahead
> with this patch or do you want that to be handled as a next and a
> separate step?

No, just a suggestion of possibilities...

Yours,
Linus Walleij

Patch

diff --git a/drivers/input/keyboard/nomadik-ske-keypad.c b/drivers/input/keyboard/nomadik-ske-keypad.c
index 95dcc9b..a1a9375 100644
--- a/drivers/input/keyboard/nomadik-ske-keypad.c
+++ b/drivers/input/keyboard/nomadik-ske-keypad.c
@@ -67,6 +67,7 @@  struct ske_keypad {
 	const struct ske_keypad_platform_data *board;
 	unsigned short keymap[SKE_KPD_NUM_ROWS * SKE_KPD_NUM_COLS];
 	struct clk *clk;
+	struct clk *pclk;
 	spinlock_t ske_keypad_lock;
 };
 
@@ -271,11 +272,18 @@  static int __init ske_keypad_probe(struct platform_device *pdev)
 		goto err_free_mem_region;
 	}
 
+	keypad->pclk = clk_get(&pdev->dev, "apb_pclk");
+	if (IS_ERR(keypad->pclk)) {
+		dev_err(&pdev->dev, "failed to get pclk\n");
+		error = PTR_ERR(keypad->pclk);
+		goto err_iounmap;
+	}
+
 	keypad->clk = clk_get(&pdev->dev, NULL);
 	if (IS_ERR(keypad->clk)) {
 		dev_err(&pdev->dev, "failed to get clk\n");
 		error = PTR_ERR(keypad->clk);
-		goto err_iounmap;
+		goto err_pclk;
 	}
 
 	input->id.bustype = BUS_HOST;
@@ -294,10 +302,16 @@  static int __init ske_keypad_probe(struct platform_device *pdev)
 	if (!plat->no_autorepeat)
 		__set_bit(EV_REP, input->evbit);
 
+	error = clk_prepare_enable(keypad->pclk);
+	if (error) {
+		dev_err(&pdev->dev, "Failed to prepare/enable pclk\n");
+		goto err_clk;
+	}
+
 	error = clk_prepare_enable(keypad->clk);
 	if (error) {
 		dev_err(&pdev->dev, "Failed to prepare/enable clk\n");
-		goto err_clk;
+		goto err_pclk_disable;
 	}
 
 
@@ -336,8 +350,12 @@  err_free_irq:
 	free_irq(keypad->irq, keypad);
 err_clk_disable:
 	clk_disable_unprepare(keypad->clk);
+err_pclk_disable:
+	clk_disable_unprepare(keypad->pclk);
 err_clk:
 	clk_put(keypad->clk);
+err_pclk:
+	clk_put(keypad->pclk);
 err_iounmap:
 	iounmap(keypad->reg_base);
 err_free_mem_region: