[PATCH 0/5] ASoC: clock provider clean-up
The purpose of this patchset it remove the use the clk member of 'struct clk_hw' in ASoC. 'struct clk' is a per-user reference to an actual clock. In the future, the clk member in 'struct clk_hw' may go away.
The usage of this member by a clock provider usually falls into either of following categories: * Mis-usage of the clock consumer API by a clock provider. * Clock provider also being a user of its own clocks. In this case the provider should request a 'struct clk' through the appropriate API instead of poking in 'struct clk_hw' internals.
Jerome Brunet (5): ASoC: stm32: properly get clk from the provider ASoC: wcd934x: use the clock provider API ASoC: rt5682: clock driver must use the clock provider API ASoC: lpass: use the clock provider API ASoC: da7219: properly get clk from the provider
sound/soc/codecs/da7219.c | 5 ++++- sound/soc/codecs/lpass-va-macro.c | 2 +- sound/soc/codecs/lpass-wsa-macro.c | 9 +++------ sound/soc/codecs/rt5682.c | 6 +++--- sound/soc/codecs/wcd934x.c | 6 ++++-- sound/soc/stm/stm32_sai_sub.c | 5 ++++- 6 files changed, 19 insertions(+), 14 deletions(-)
Instead of using the clk embedded in the clk_hw (which is meant to go away), a clock provider which need to interact with its own clock should request clk reference through the clock provider API.
Signed-off-by: Jerome Brunet jbrunet@baylibre.com --- sound/soc/stm/stm32_sai_sub.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/sound/soc/stm/stm32_sai_sub.c b/sound/soc/stm/stm32_sai_sub.c index 3aa1cf262402..c1561237ee24 100644 --- a/sound/soc/stm/stm32_sai_sub.c +++ b/sound/soc/stm/stm32_sai_sub.c @@ -484,7 +484,10 @@ static int stm32_sai_add_mclk_provider(struct stm32_sai_sub_data *sai) dev_err(dev, "mclk register returned %d\n", ret); return ret; } - sai->sai_mclk = hw->clk; + + sai->sai_mclk = devm_clk_hw_get_clk(dev, hw, NULL); + if (IS_ERR(sai->sai_mclk)) + return PTR_ERR(sai->sai_mclk);
/* register mclk provider */ return devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get, hw);
Quoting Jerome Brunet (2021-04-10 04:13:52)
Instead of using the clk embedded in the clk_hw (which is meant to go away), a clock provider which need to interact with its own clock should request clk reference through the clock provider API.
Signed-off-by: Jerome Brunet jbrunet@baylibre.com
Reviewed-by: Stephen Boyd sboyd@kernel.org
Clock providers should use the clk_hw API
Signed-off-by: Jerome Brunet jbrunet@baylibre.com --- sound/soc/codecs/wcd934x.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/sound/soc/codecs/wcd934x.c b/sound/soc/codecs/wcd934x.c index 5fe403307b72..ae3ea136a9f8 100644 --- a/sound/soc/codecs/wcd934x.c +++ b/sound/soc/codecs/wcd934x.c @@ -2116,11 +2116,13 @@ static struct clk *wcd934x_register_mclk_output(struct wcd934x_codec *wcd) wcd->hw.init = &init;
hw = &wcd->hw; - ret = clk_hw_register(wcd->dev->parent, hw); + ret = devm_clk_hw_register(wcd->dev->parent, hw); if (ret) return ERR_PTR(ret);
- of_clk_add_provider(np, of_clk_src_simple_get, hw->clk); + ret = devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get, hw); + if (ret) + return ERR_PTR(ret);
return NULL; }
Quoting Jerome Brunet (2021-04-10 04:13:53)
Clock providers should use the clk_hw API
It sort of already is :)
Signed-off-by: Jerome Brunet jbrunet@baylibre.com
Reviewed-by: Stephen Boyd sboyd@kernel.org
Clock drivers ops should not the clk API but the clock provider (clk_hw) instead.
Signed-off-by: Jerome Brunet jbrunet@baylibre.com --- sound/soc/codecs/rt5682.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/sound/soc/codecs/rt5682.c b/sound/soc/codecs/rt5682.c index 0e2a10ed11da..2eee02ac8d49 100644 --- a/sound/soc/codecs/rt5682.c +++ b/sound/soc/codecs/rt5682.c @@ -2634,7 +2634,7 @@ static int rt5682_wclk_set_rate(struct clk_hw *hw, unsigned long rate, container_of(hw, struct rt5682_priv, dai_clks_hw[RT5682_DAI_WCLK_IDX]); struct snd_soc_component *component = rt5682->component; - struct clk *parent_clk; + struct clk_hw *parent_hw; const char * const clk_name = clk_hw_get_name(hw); int pre_div; unsigned int clk_pll2_out; @@ -2649,8 +2649,8 @@ static int rt5682_wclk_set_rate(struct clk_hw *hw, unsigned long rate, * * It will set the codec anyway by assuming mclk is 48MHz. */ - parent_clk = clk_get_parent(hw->clk); - if (!parent_clk) + parent_hw = clk_hw_get_parent(hw); + if (!parent_hw) dev_warn(component->dev, "Parent mclk of wclk not acquired in driver. Please ensure mclk was provided as %d Hz.\n", CLK_PLL2_FIN);
Quoting Jerome Brunet (2021-04-10 04:13:54)
Clock drivers ops should not the clk API but the clock provider (clk_hw) instead.
Signed-off-by: Jerome Brunet jbrunet@baylibre.com
sound/soc/codecs/rt5682.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/sound/soc/codecs/rt5682.c b/sound/soc/codecs/rt5682.c index 0e2a10ed11da..2eee02ac8d49 100644 --- a/sound/soc/codecs/rt5682.c +++ b/sound/soc/codecs/rt5682.c @@ -2634,7 +2634,7 @@ static int rt5682_wclk_set_rate(struct clk_hw *hw, unsigned long rate, container_of(hw, struct rt5682_priv, dai_clks_hw[RT5682_DAI_WCLK_IDX]); struct snd_soc_component *component = rt5682->component;
struct clk *parent_clk;
struct clk_hw *parent_hw; const char * const clk_name = clk_hw_get_name(hw); int pre_div; unsigned int clk_pll2_out;
@@ -2649,8 +2649,8 @@ static int rt5682_wclk_set_rate(struct clk_hw *hw, unsigned long rate, * * It will set the codec anyway by assuming mclk is 48MHz. */
parent_clk = clk_get_parent(hw->clk);
if (!parent_clk)
parent_hw = clk_hw_get_parent(hw);
if (!parent_hw) dev_warn(component->dev, "Parent mclk of wclk not acquired in driver. Please ensure mclk was provided as %d Hz.\n", CLK_PLL2_FIN);
Can this code be removed? I don't know why we care to check if the clk has a parent or not.
On Mon 12 Apr 2021 at 22:27, Stephen Boyd sboyd@kernel.org wrote:
Quoting Jerome Brunet (2021-04-10 04:13:54)
Clock drivers ops should not the clk API but the clock provider (clk_hw) instead.
Signed-off-by: Jerome Brunet jbrunet@baylibre.com
sound/soc/codecs/rt5682.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/sound/soc/codecs/rt5682.c b/sound/soc/codecs/rt5682.c index 0e2a10ed11da..2eee02ac8d49 100644 --- a/sound/soc/codecs/rt5682.c +++ b/sound/soc/codecs/rt5682.c @@ -2634,7 +2634,7 @@ static int rt5682_wclk_set_rate(struct clk_hw *hw, unsigned long rate, container_of(hw, struct rt5682_priv, dai_clks_hw[RT5682_DAI_WCLK_IDX]); struct snd_soc_component *component = rt5682->component;
struct clk *parent_clk;
struct clk_hw *parent_hw; const char * const clk_name = clk_hw_get_name(hw); int pre_div; unsigned int clk_pll2_out;
@@ -2649,8 +2649,8 @@ static int rt5682_wclk_set_rate(struct clk_hw *hw, unsigned long rate, * * It will set the codec anyway by assuming mclk is 48MHz. */
parent_clk = clk_get_parent(hw->clk);
if (!parent_clk)
parent_hw = clk_hw_get_parent(hw);
if (!parent_hw) dev_warn(component->dev, "Parent mclk of wclk not acquired in driver. Please ensure mclk was provided as %d Hz.\n", CLK_PLL2_FIN);
Can this code be removed? I don't know why we care to check if the clk has a parent or not.
I'm focusing on removing "hw->clk" where they are - w/o changing too much what the driver does. I don't have the HW nor the story behind it and there is about 50 more drivers to be fixed ... thankfully, most are in drivers/clk/ ;)
Here, at least the clock consummer API is not longer used within a clock ops, which is not great considering the locking scheme (among other things)
Clock providers should be registered using the clk_hw API.
Signed-off-by: Jerome Brunet jbrunet@baylibre.com --- sound/soc/codecs/lpass-va-macro.c | 2 +- sound/soc/codecs/lpass-wsa-macro.c | 9 +++------ 2 files changed, 4 insertions(+), 7 deletions(-)
diff --git a/sound/soc/codecs/lpass-va-macro.c b/sound/soc/codecs/lpass-va-macro.c index 5294c57b2cd4..56b887301172 100644 --- a/sound/soc/codecs/lpass-va-macro.c +++ b/sound/soc/codecs/lpass-va-macro.c @@ -1343,7 +1343,7 @@ static int va_macro_register_fsgen_output(struct va_macro *va) if (ret) return ret;
- return of_clk_add_provider(np, of_clk_src_simple_get, va->hw.clk); + return devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get, &va->hw); }
static int va_macro_validate_dmic_sample_rate(u32 dmic_sample_rate, diff --git a/sound/soc/codecs/lpass-wsa-macro.c b/sound/soc/codecs/lpass-wsa-macro.c index e79a70386b4b..acb95e83c788 100644 --- a/sound/soc/codecs/lpass-wsa-macro.c +++ b/sound/soc/codecs/lpass-wsa-macro.c @@ -2337,10 +2337,9 @@ static const struct clk_ops swclk_gate_ops = { .recalc_rate = swclk_recalc_rate, };
-static struct clk *wsa_macro_register_mclk_output(struct wsa_macro *wsa) +static int wsa_macro_register_mclk_output(struct wsa_macro *wsa) { struct device *dev = wsa->dev; - struct device_node *np = dev->of_node; const char *parent_clk_name; const char *clk_name = "mclk"; struct clk_hw *hw; @@ -2358,11 +2357,9 @@ static struct clk *wsa_macro_register_mclk_output(struct wsa_macro *wsa) hw = &wsa->hw; ret = clk_hw_register(wsa->dev, hw); if (ret) - return ERR_PTR(ret); - - of_clk_add_provider(np, of_clk_src_simple_get, hw->clk); + return ret;
- return NULL; + return devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get, hw); }
static const struct snd_soc_component_driver wsa_macro_component_drv = {
Thanks Jerome for the patch,
On 10/04/2021 12:13, Jerome Brunet wrote:
Clock providers should be registered using the clk_hw API.
Signed-off-by: Jerome Brunet jbrunet@baylibre.com
sound/soc/codecs/lpass-va-macro.c | 2 +- sound/soc/codecs/lpass-wsa-macro.c | 9 +++------ 2 files changed, 4 insertions(+), 7 deletions(-)
diff --git a/sound/soc/codecs/lpass-va-macro.c b/sound/soc/codecs/lpass-va-macro.c index 5294c57b2cd4..56b887301172 100644 --- a/sound/soc/codecs/lpass-va-macro.c +++ b/sound/soc/codecs/lpass-va-macro.c @@ -1343,7 +1343,7 @@ static int va_macro_register_fsgen_output(struct va_macro *va) if (ret) return ret;
- return of_clk_add_provider(np, of_clk_src_simple_get, va->hw.clk);
- return devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get, &va->hw);
Now that we convert this to devm, You missed error path and driver remove where we delete clk provider. This should be removed as well as part of this patch.
This applies to both wsa and va macro.
Thanks, srini
}
static int va_macro_validate_dmic_sample_rate(u32 dmic_sample_rate, diff --git a/sound/soc/codecs/lpass-wsa-macro.c b/sound/soc/codecs/lpass-wsa-macro.c index e79a70386b4b..acb95e83c788 100644 --- a/sound/soc/codecs/lpass-wsa-macro.c +++ b/sound/soc/codecs/lpass-wsa-macro.c @@ -2337,10 +2337,9 @@ static const struct clk_ops swclk_gate_ops = { .recalc_rate = swclk_recalc_rate, };
-static struct clk *wsa_macro_register_mclk_output(struct wsa_macro *wsa) +static int wsa_macro_register_mclk_output(struct wsa_macro *wsa) { struct device *dev = wsa->dev;
- struct device_node *np = dev->of_node; const char *parent_clk_name; const char *clk_name = "mclk"; struct clk_hw *hw;
@@ -2358,11 +2357,9 @@ static struct clk *wsa_macro_register_mclk_output(struct wsa_macro *wsa) hw = &wsa->hw; ret = clk_hw_register(wsa->dev, hw); if (ret)
return ERR_PTR(ret);
- of_clk_add_provider(np, of_clk_src_simple_get, hw->clk);
return ret;
- return NULL;
return devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get, hw); }
static const struct snd_soc_component_driver wsa_macro_component_drv = {
On Mon 12 Apr 2021 at 11:38, Srinivas Kandagatla srinivas.kandagatla@linaro.org wrote:
Thanks Jerome for the patch,
On 10/04/2021 12:13, Jerome Brunet wrote:
Clock providers should be registered using the clk_hw API. Signed-off-by: Jerome Brunet jbrunet@baylibre.com
sound/soc/codecs/lpass-va-macro.c | 2 +- sound/soc/codecs/lpass-wsa-macro.c | 9 +++------ 2 files changed, 4 insertions(+), 7 deletions(-) diff --git a/sound/soc/codecs/lpass-va-macro.c b/sound/soc/codecs/lpass-va-macro.c index 5294c57b2cd4..56b887301172 100644 --- a/sound/soc/codecs/lpass-va-macro.c +++ b/sound/soc/codecs/lpass-va-macro.c @@ -1343,7 +1343,7 @@ static int va_macro_register_fsgen_output(struct va_macro *va) if (ret) return ret;
- return of_clk_add_provider(np, of_clk_src_simple_get, va->hw.clk);
- return devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get, &va->hw);
Now that we convert this to devm, You missed error path and driver remove where we delete clk provider. This should be removed as well as part of this patch.
Indeed. I should not have switched to devm here - It was not really the purpose of the patch. Habits I guess.
Do you prefer I stick with devm (with the suggested fix) or revert to the no-devm way for the v2 ? It makes no difference to me TBH.
This applies to both wsa and va macro.
Thanks, srini
} static int va_macro_validate_dmic_sample_rate(u32 dmic_sample_rate, diff --git a/sound/soc/codecs/lpass-wsa-macro.c b/sound/soc/codecs/lpass-wsa-macro.c index e79a70386b4b..acb95e83c788 100644 --- a/sound/soc/codecs/lpass-wsa-macro.c +++ b/sound/soc/codecs/lpass-wsa-macro.c @@ -2337,10 +2337,9 @@ static const struct clk_ops swclk_gate_ops = { .recalc_rate = swclk_recalc_rate, }; -static struct clk *wsa_macro_register_mclk_output(struct wsa_macro *wsa) +static int wsa_macro_register_mclk_output(struct wsa_macro *wsa) { struct device *dev = wsa->dev;
- struct device_node *np = dev->of_node; const char *parent_clk_name; const char *clk_name = "mclk"; struct clk_hw *hw;
@@ -2358,11 +2357,9 @@ static struct clk *wsa_macro_register_mclk_output(struct wsa_macro *wsa) hw = &wsa->hw; ret = clk_hw_register(wsa->dev, hw); if (ret)
return ERR_PTR(ret);
- of_clk_add_provider(np, of_clk_src_simple_get, hw->clk);
return ret;
- return NULL;
- return devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get, hw); } static const struct snd_soc_component_driver wsa_macro_component_drv
= {
On 12/04/2021 13:17, Jerome Brunet wrote:
- return of_clk_add_provider(np, of_clk_src_simple_get, va->hw.clk);
- return devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get, &va->hw);
Now that we convert this to devm, You missed error path and driver remove where we delete clk provider. This should be removed as well as part of this patch.
Indeed. I should not have switched to devm here - It was not really the purpose of the patch. Habits I guess.
Do you prefer I stick with devm (with the suggested fix) or revert to the no-devm way for the v2 ? It makes no difference to me TBH.
devm should be good.
--srini
Instead of using the clk embedded in the clk_hw (which is meant to go away), a clock provider which need to interact with its own clock should request clk reference through the clock provider API.
Signed-off-by: Jerome Brunet jbrunet@baylibre.com --- sound/soc/codecs/da7219.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/sound/soc/codecs/da7219.c b/sound/soc/codecs/da7219.c index 13009d08b09a..bd3c523a8617 100644 --- a/sound/soc/codecs/da7219.c +++ b/sound/soc/codecs/da7219.c @@ -2181,7 +2181,10 @@ static int da7219_register_dai_clks(struct snd_soc_component *component) ret); goto err; } - da7219->dai_clks[i] = dai_clk_hw->clk; + + da7219->dai_clks[i] = devm_clk_hw_get_clk(dev, dai_clk_hw, NULL); + if (IS_ERR(da7219->dai_clks[i])) + return PTR_ERR(da7219->dai_clks[i]);
/* For DT setup onecell data, otherwise create lookup */ if (np) {
Quoting Jerome Brunet (2021-04-10 04:13:56)
Instead of using the clk embedded in the clk_hw (which is meant to go away), a clock provider which need to interact with its own clock should request clk reference through the clock provider API.
Signed-off-by: Jerome Brunet jbrunet@baylibre.com
Reviewed-by: Stephen Boyd sboyd@kernel.org
participants (3)
-
Jerome Brunet
-
Srinivas Kandagatla
-
Stephen Boyd