From: Arnd Bergmann <[email protected]>
Casting between incompatible function types causes a warning with clang-16
because it breaks control flow integrity (KCFI) rules:
sound/soc/meson/aiu.c:243:12: error: cast from 'void (*)(struct clk *)' to 'void (*)(void *)' converts to incompatible function type [-Werror,-Wcast-function-type-strict]
243 | (void(*)(void *))clk_disable_unprepare,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
include/linux/device.h:406:34: note: expanded from macro 'devm_add_action_or_reset'
406 | __devm_add_action_or_reset(dev, action, data, #action)
| ^~~~~~
sound/soc/meson/t9015.c:274:4: error: cast from 'void (*)(struct clk *)' to 'void (*)(void *)' converts to incompatible function type [-Werror,-Wcast-function-type-strict]
274 | (void(*)(void *))clk_disable_unprepare,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
include/linux/device.h:406:34: note: expanded from macro 'devm_add_action_or_reset'
406 | __devm_add_action_or_reset(dev, action, data, #action)
| ^~~~~~
These two meson drivers cast clk_disable_unprepare() into a different type
in order to have it automatically called from the driver relase. Add
trivial helpers to do the same using correct types.
Fixes: 33901f5b9b16 ("ASoC: meson: add t9015 internal DAC driver")
Fixes: 6ae9ca9ce986 ("ASoC: meson: aiu: add i2s and spdif support")
Signed-off-by: Arnd Bergmann <[email protected]>
---
sound/soc/meson/aiu.c | 9 ++++++---
sound/soc/meson/t9015.c | 9 ++++++---
2 files changed, 12 insertions(+), 6 deletions(-)
diff --git a/sound/soc/meson/aiu.c b/sound/soc/meson/aiu.c
index 7109b81cc3d0..211f360df284 100644
--- a/sound/soc/meson/aiu.c
+++ b/sound/soc/meson/aiu.c
@@ -209,6 +209,11 @@ static const char * const aiu_spdif_ids[] = {
[MCLK] = "spdif_mclk_sel"
};
+static void aiu_clk_disable(void *clk)
+{
+ clk_disable_unprepare(clk);
+}
+
static int aiu_clk_get(struct device *dev)
{
struct aiu *aiu = dev_get_drvdata(dev);
@@ -239,9 +244,7 @@ static int aiu_clk_get(struct device *dev)
return ret;
}
- ret = devm_add_action_or_reset(dev,
- (void(*)(void *))clk_disable_unprepare,
- aiu->pclk);
+ ret = devm_add_action_or_reset(dev, aiu_clk_disable, aiu->pclk);
if (ret)
dev_err(dev, "failed to add reset action on pclk");
diff --git a/sound/soc/meson/t9015.c b/sound/soc/meson/t9015.c
index 9c6b4dac6893..e0f9a603a5e9 100644
--- a/sound/soc/meson/t9015.c
+++ b/sound/soc/meson/t9015.c
@@ -243,6 +243,11 @@ static const struct regmap_config t9015_regmap_config = {
.max_register = POWER_CFG,
};
+static void t9015_clk_disable(void *clk)
+{
+ clk_disable_unprepare(clk);
+}
+
static int t9015_probe(struct platform_device *pdev)
{
struct device *dev = &pdev->dev;
@@ -270,9 +275,7 @@ static int t9015_probe(struct platform_device *pdev)
return ret;
}
- ret = devm_add_action_or_reset(dev,
- (void(*)(void *))clk_disable_unprepare,
- priv->pclk);
+ ret = devm_add_action_or_reset(dev, t9015_clk_disable, priv->pclk);
if (ret)
return ret;
--
2.39.2
On Tue 13 Feb 2024 at 11:11, Arnd Bergmann <[email protected]> wrote:
> From: Arnd Bergmann <[email protected]>
>
> Casting between incompatible function types causes a warning with clang-16
> because it breaks control flow integrity (KCFI) rules:
>
> sound/soc/meson/aiu.c:243:12: error: cast from 'void (*)(struct clk *)' to
> 'void (*)(void *)' converts to incompatible function type
> [-Werror,-Wcast-function-type-strict]
> 243 | (void(*)(void *))clk_disable_unprepare,
> | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> include/linux/device.h:406:34: note: expanded from macro 'devm_add_action_or_reset'
> 406 | __devm_add_action_or_reset(dev, action, data, #action)
> | ^~~~~~
> sound/soc/meson/t9015.c:274:4: error: cast from 'void (*)(struct clk *)' to
> 'void (*)(void *)' converts to incompatible function type
> [-Werror,-Wcast-function-type-strict]
> 274 | (void(*)(void *))clk_disable_unprepare,
> | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> include/linux/device.h:406:34: note: expanded from macro 'devm_add_action_or_reset'
> 406 | __devm_add_action_or_reset(dev, action, data, #action)
> | ^~~~~~
>
> These two meson drivers cast clk_disable_unprepare() into a different type
> in order to have it automatically called from the driver relase. Add
> trivial helpers to do the same using correct types.
>
> Fixes: 33901f5b9b16 ("ASoC: meson: add t9015 internal DAC driver")
> Fixes: 6ae9ca9ce986 ("ASoC: meson: aiu: add i2s and spdif support")
> Signed-off-by: Arnd Bergmann <[email protected]>
> ---
> sound/soc/meson/aiu.c | 9 ++++++---
> sound/soc/meson/t9015.c | 9 ++++++---
> 2 files changed, 12 insertions(+), 6 deletions(-)
>
> diff --git a/sound/soc/meson/aiu.c b/sound/soc/meson/aiu.c
> index 7109b81cc3d0..211f360df284 100644
> --- a/sound/soc/meson/aiu.c
> +++ b/sound/soc/meson/aiu.c
> @@ -209,6 +209,11 @@ static const char * const aiu_spdif_ids[] = {
> [MCLK] = "spdif_mclk_sel"
> };
>
> +static void aiu_clk_disable(void *clk)
> +{
> + clk_disable_unprepare(clk);
> +}
> +
> static int aiu_clk_get(struct device *dev)
> {
> struct aiu *aiu = dev_get_drvdata(dev);
> @@ -239,9 +244,7 @@ static int aiu_clk_get(struct device *dev)
> return ret;
> }
>
> - ret = devm_add_action_or_reset(dev,
> - (void(*)(void *))clk_disable_unprepare,
> - aiu->pclk);
> + ret = devm_add_action_or_reset(dev, aiu_clk_disable, aiu->pclk);
Hi Arnd,
This probably pre-dates the introduction of devm_clk_get_enabled()
It would probably be better to use that instead of duplicating helper
functions which do the same thing.
If for any reason it is not possible, a common helper in clk.h would
preferable I think.
> if (ret)
> dev_err(dev, "failed to add reset action on pclk");
>
> diff --git a/sound/soc/meson/t9015.c b/sound/soc/meson/t9015.c
> index 9c6b4dac6893..e0f9a603a5e9 100644
> --- a/sound/soc/meson/t9015.c
> +++ b/sound/soc/meson/t9015.c
> @@ -243,6 +243,11 @@ static const struct regmap_config t9015_regmap_config = {
> .max_register = POWER_CFG,
> };
>
> +static void t9015_clk_disable(void *clk)
> +{
> + clk_disable_unprepare(clk);
> +}
> +
> static int t9015_probe(struct platform_device *pdev)
> {
> struct device *dev = &pdev->dev;
> @@ -270,9 +275,7 @@ static int t9015_probe(struct platform_device *pdev)
> return ret;
> }
>
> - ret = devm_add_action_or_reset(dev,
> - (void(*)(void *))clk_disable_unprepare,
> - priv->pclk);
> + ret = devm_add_action_or_reset(dev, t9015_clk_disable, priv->pclk);
> if (ret)
> return ret;
--
Jerome
On Tue, Feb 13, 2024, at 11:36, Jerome Brunet wrote:
> On Tue 13 Feb 2024 at 11:11, Arnd Bergmann <[email protected]> wrote:
> This probably pre-dates the introduction of devm_clk_get_enabled()
> It would probably be better to use that instead of duplicating helper
> functions which do the same thing.
Ah, I had not thought of that interface either, so you are probably
right that this is the best way to do it.
Can you send a replacement patch then and add my Reported-by?
I also sent the same patch for drivers/nvmem/meson-efuse.c, which
I guess will also need the same treatment. I also checked and saw
that all three files already had this code in linux-6.0 when
devm_clk_get_enabled() got added.
> If for any reason it is not possible, a common helper in clk.h would
> preferable I think.
I can't think of anything that prevents us from using
devm_clk_get_enabled() here.
Arnd
On Tue 13 Feb 2024 at 11:52, "Arnd Bergmann" <[email protected]> wrote:
> On Tue, Feb 13, 2024, at 11:36, Jerome Brunet wrote:
>> On Tue 13 Feb 2024 at 11:11, Arnd Bergmann <[email protected]> wrote:
>
>> This probably pre-dates the introduction of devm_clk_get_enabled()
>> It would probably be better to use that instead of duplicating helper
>> functions which do the same thing.
>
> Ah, I had not thought of that interface either, so you are probably
> right that this is the best way to do it.
> Can you send a replacement patch then and add my Reported-by?
Sure.
How may I reproduce the problem ?
Just tried with 'Debian clang version 16.0.6 (19)', no warning.
I suppose I need to add something ?
>
> I also sent the same patch for drivers/nvmem/meson-efuse.c, which
> I guess will also need the same treatment. I also checked and saw
> that all three files already had this code in linux-6.0 when
> devm_clk_get_enabled() got added.
>
>> If for any reason it is not possible, a common helper in clk.h would
>> preferable I think.
>
> I can't think of anything that prevents us from using
> devm_clk_get_enabled() here.
>
> Arnd
--
Jerome
On Tue, Feb 13, 2024, at 12:09, Jerome Brunet wrote:
> On Tue 13 Feb 2024 at 11:52, "Arnd Bergmann" <[email protected]> wrote:
>
>> On Tue, Feb 13, 2024, at 11:36, Jerome Brunet wrote:
>>> On Tue 13 Feb 2024 at 11:11, Arnd Bergmann <[email protected]> wrote:
>>
>>> This probably pre-dates the introduction of devm_clk_get_enabled()
>>> It would probably be better to use that instead of duplicating helper
>>> functions which do the same thing.
>>
>> Ah, I had not thought of that interface either, so you are probably
>> right that this is the best way to do it.
>> Can you send a replacement patch then and add my Reported-by?
>
> Sure.
>
> How may I reproduce the problem ?
> Just tried with 'Debian clang version 16.0.6 (19)', no warning.
>
> I suppose I need to add something ?
The warning is currently only enabled at W=1 level until
we have fixed all the existing diagnostic output. I have
sent the other ones today, but for now, please test with
'make W=1'.
Arnd