[alsa-devel] [PATCH v3] ASoC: wm8960: Let wm8960 codec driver manage its own MCLK
When we want to use wm8960 codec, we should enable its MCLK in machine driver. It's reasonable for wm8960 codec driver to manage its own MCLK to save power.
If platform data is not supplied, we will get MCLK from device tree and check the return code to handle probe deferral.
Enable runtime power management, and auto enable/disable MCLK in pm_runtime resume and suspend. When wm8960 codec is being used, it will triger resume() to enable MCLK. When codec is not being used, it will triger suspend() to disable MCLK.
Signed-off-by: Zidan Wang b50113@freescale.com --- include/sound/wm8960.h | 1 + sound/soc/codecs/wm8960.c | 56 ++++++++++++++++++++++++++++++++++++++++++++--- 2 files changed, 54 insertions(+), 3 deletions(-)
diff --git a/include/sound/wm8960.h b/include/sound/wm8960.h index e8ce8ee..d7e84ed 100644 --- a/include/sound/wm8960.h +++ b/include/sound/wm8960.h @@ -16,6 +16,7 @@ #define WM8960_DRES_MAX 3
struct wm8960_data { + struct clk *mclk; bool capless; /* Headphone outputs configured in capless mode */
bool shared_lrclk; /* DAC and ADC LRCLKs are wired together */ diff --git a/sound/soc/codecs/wm8960.c b/sound/soc/codecs/wm8960.c index 031a1ae..ea5f0b7 100644 --- a/sound/soc/codecs/wm8960.c +++ b/sound/soc/codecs/wm8960.c @@ -15,6 +15,8 @@ #include <linux/init.h> #include <linux/delay.h> #include <linux/pm.h> +#include <linux/clk.h> +#include <linux/pm_runtime.h> #include <linux/i2c.h> #include <linux/slab.h> #include <sound/core.h> @@ -978,7 +980,7 @@ static const struct regmap_config wm8960_regmap = { .volatile_reg = wm8960_volatile, };
-static void wm8960_set_pdata_from_of(struct i2c_client *i2c, +static int wm8960_set_pdata_from_of(struct i2c_client *i2c, struct wm8960_data *pdata) { const struct device_node *np = i2c->dev.of_node; @@ -988,6 +990,15 @@ static void wm8960_set_pdata_from_of(struct i2c_client *i2c,
if (of_property_read_bool(np, "wlf,shared-lrclk")) pdata->shared_lrclk = true; + + pdata->mclk = devm_clk_get(&i2c->dev, NULL); + + if (IS_ERR(pdata->mclk)) { + if (PTR_ERR(pdata->mclk) == -EPROBE_DEFER) + return -EPROBE_DEFER; + } + + return 0; }
static int wm8960_i2c_probe(struct i2c_client *i2c, @@ -1008,8 +1019,11 @@ static int wm8960_i2c_probe(struct i2c_client *i2c,
if (pdata) memcpy(&wm8960->pdata, pdata, sizeof(struct wm8960_data)); - else if (i2c->dev.of_node) - wm8960_set_pdata_from_of(i2c, &wm8960->pdata); + else if (i2c->dev.of_node) { + ret = wm8960_set_pdata_from_of(i2c, &wm8960->pdata); + if (ret != 0) + return ret; + }
ret = wm8960_reset(wm8960->regmap); if (ret != 0) { @@ -1041,6 +1055,9 @@ static int wm8960_i2c_probe(struct i2c_client *i2c,
i2c_set_clientdata(i2c, wm8960);
+ pm_runtime_enable(&i2c->dev); + pm_request_idle(&i2c->dev); + ret = snd_soc_register_codec(&i2c->dev, &soc_codec_dev_wm8960, &wm8960_dai, 1);
@@ -1053,6 +1070,38 @@ static int wm8960_i2c_remove(struct i2c_client *client) return 0; }
+#ifdef CONFIG_PM_RUNTIME +static int wm8960_runtime_resume(struct device *dev) +{ + struct wm8960_priv *wm8960 = dev_get_drvdata(dev); + int ret; + + if (!IS_ERR(wm8960->pdata.mclk)) { + ret = clk_prepare_enable(wm8960->pdata.mclk); + if (ret) { + dev_err(dev, "Failed to enable MCLK: %d\n", ret); + return ret; + } + } + + return 0; +} + +static int wm8960_runtime_suspend(struct device *dev) +{ + struct wm8960_priv *wm8960 = dev_get_drvdata(dev); + + if (!IS_ERR(wm8960->pdata.mclk)) + clk_disable_unprepare(wm8960->pdata.mclk); + + return 0; +} +#endif + +static const struct dev_pm_ops wm8960_pm = { + SET_RUNTIME_PM_OPS(wm8960_runtime_suspend, wm8960_runtime_resume, NULL) +}; + static const struct i2c_device_id wm8960_i2c_id[] = { { "wm8960", 0 }, { } @@ -1070,6 +1119,7 @@ static struct i2c_driver wm8960_i2c_driver = { .name = "wm8960", .owner = THIS_MODULE, .of_match_table = wm8960_of_match, + .pm = &wm8960_pm, }, .probe = wm8960_i2c_probe, .remove = wm8960_i2c_remove,
On Thu, Dec 04, 2014 at 08:41:19PM +0800, Zidan Wang wrote:
When we want to use wm8960 codec, we should enable its MCLK in machine driver. It's reasonable for wm8960 codec driver to manage its own MCLK to save power.
If platform data is not supplied, we will get MCLK from device tree and check the return code to handle probe deferral.
Enable runtime power management, and auto enable/disable MCLK in pm_runtime resume and suspend. When wm8960 codec is being used, it will triger resume() to enable MCLK. When codec is not being used, it will triger suspend() to disable MCLK.
Signed-off-by: Zidan Wang b50113@freescale.com
include/sound/wm8960.h | 1 + sound/soc/codecs/wm8960.c | 56 ++++++++++++++++++++++++++++++++++++++++++++--- 2 files changed, 54 insertions(+), 3 deletions(-)
diff --git a/include/sound/wm8960.h b/include/sound/wm8960.h index e8ce8ee..d7e84ed 100644 --- a/include/sound/wm8960.h +++ b/include/sound/wm8960.h @@ -16,6 +16,7 @@ #define WM8960_DRES_MAX 3
struct wm8960_data {
- struct clk *mclk;
Is this really pdata? Would the pdata entry to locate the clock not be a string holding the clock name that you call clk_get on, rather than a clk pointer itself? Probably this should go in wm8960_priv instead.
Thanks, Charles
On Fri, Dec 05, 2014 at 05:01:56PM +0000, Charles Keepax wrote:
On Thu, Dec 04, 2014 at 08:41:19PM +0800, Zidan Wang wrote:
struct wm8960_data {
- struct clk *mclk;
Is this really pdata? Would the pdata entry to locate the clock not be a string holding the clock name that you call clk_get on, rather than a clk pointer itself? Probably this should go in wm8960_priv instead.
There should be no platform data of any kind for the clock, the name should be fixed and defined in terms of the name of the clock on the device requesting it. The code is actually doing the right thing here, it's just that it's put the variable into the platform data because currently the driver just uses the platform data as the driver data.
So I should move mclk to wm8960_priv. And doing wm8960_priv->mclk = devm_clk_get(&i2c->dev, "codec_mclk") in wm8960_i2c_probe.
Best Regards, Zidan
-----Original Message----- From: Mark Brown [mailto:broonie@kernel.org] Sent: Saturday, December 06, 2014 1:40 AM To: Charles Keepax Cc: Wang Zidan-B50113; lgirdwood@gmail.com; perex@perex.cz; tiwai@suse.de; lars@metafoo.de; Xiubo Li-B47053; patches@opensource.wolfsonmicro.com; alsa-devel@alsa-project.org; linux-kernel@vger.kernel.org Subject: Re: [alsa-devel][PATCH v3] ASoC: wm8960: Let wm8960 codec driver manage its own MCLK
On Fri, Dec 05, 2014 at 05:01:56PM +0000, Charles Keepax wrote:
On Thu, Dec 04, 2014 at 08:41:19PM +0800, Zidan Wang wrote:
struct wm8960_data {
- struct clk *mclk;
Is this really pdata? Would the pdata entry to locate the clock not be a string holding the clock name that you call clk_get on, rather than a clk pointer itself? Probably this should go in wm8960_priv instead.
There should be no platform data of any kind for the clock, the name should be fixed and defined in terms of the name of the clock on the device requesting it. The code is actually doing the right thing here, it's just that it's put the variable into the platform data because currently the driver just uses the platform data as the driver data.
On Mon, Dec 08, 2014 at 09:49:51AM +0000, Zidan Wang wrote:
So I should move mclk to wm8960_priv. And doing wm8960_priv->mclk = devm_clk_get(&i2c->dev, "codec_mclk") in wm8960_i2c_probe.
Best Regards, Zidan
-----Original Message-----
From the sounds of Mark's comment I believe he is happy with it
where it is, and if he is happy so am I.
Thanks, Charles
On Mon, Dec 08, 2014 at 09:49:51AM +0000, Zidan Wang wrote:
So I should move mclk to wm8960_priv. And doing wm8960_priv->mclk = devm_clk_get(&i2c->dev, "codec_mclk") in wm8960_i2c_probe.
Yes, please.
On Mon, Dec 08, 2014 at 11:14:45AM +0000, Mark Brown wrote:
On Mon, Dec 08, 2014 at 09:49:51AM +0000, Zidan Wang wrote:
So I should move mclk to wm8960_priv. And doing wm8960_priv->mclk = devm_clk_get(&i2c->dev, "codec_mclk") in wm8960_i2c_probe.
Yes, please.
Apologies I totally misread you last email.
Thanks, Charles
On Thu, Dec 04, 2014 at 08:41:19PM +0800, Zidan Wang wrote:
When we want to use wm8960 codec, we should enable its MCLK in machine driver. It's reasonable for wm8960 codec driver to manage its own MCLK to save power.
If platform data is not supplied, we will get MCLK from device tree and check the return code to handle probe deferral.
Enable runtime power management, and auto enable/disable MCLK in pm_runtime resume and suspend. When wm8960 codec is being used, it will triger resume() to enable MCLK. When codec is not being used, it will triger suspend() to disable MCLK.
Signed-off-by: Zidan Wang b50113@freescale.com
Acked-by: Charles Keepax ckeepax@opensource.wolfsonmicro.com
Thanks, Charles
participants (3)
-
Charles Keepax
-
Mark Brown
-
Zidan Wang