Skip to contentSkip to conversation

ASoC: SOF: amd: fix building without CONFIG_ACPI

PATCH In review Arnd Bergmann 4 participants 4 messages Sep 25

Subsystems (3)
Sound ASoC Sound - sound open firmware (SOF) drivers
lore
Share & tools
lore

Arnd Bergmann

author
#0 lore raw

PATCH 1/1 ASoC: SOF: amd: fix building without CONFIG_ACPI

From: Arnd Bergmann <[email protected]>

When ACPI is disabled, this driver runs into a build failure:

sound/soc/sof/amd/acp.c:1374:13: error: call to undeclared function 'acpi_find_child_device'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
1374 | pdm_dev = acpi_find_child_device(adev, ACP7X_DMIC_ADDR, 0);

The code block is already guarded by an 'if (adev)' check that can only be true if ACPI is enabled, so add another #ifdef to match this.

Fixes 203ea76c4b0a ("ASoC: SOF: amd: add ACP7.B/7.F PDM controller scan and pdata propagation")
Signed-off-by Arnd Bergmann <[email protected]>

---
I also tried adding ACPI Kconfig dependencies, but that quickly got out of hand since the acp module is selected by a number of drivers that then all need the same dependency.

It may be possible to work around this in the ACPI code by adding a stub macro for acpi_find_child_device() that does nothing.

Signed-off-by Arnd Bergmann <[email protected]>
---
 sound/soc/sof/amd/acp.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
sound/soc/sof/amd/acp.c +7 -2
index b059039c9e0d..4b2d7cb244c9 100644
@@ -1059,6 +1059,7 @@ int amd_sof_acp_resume(struct snd_sof_dev *sdev)
 }
 EXPORT_SYMBOL_NS(amd_sof_acp_resume, "SND_SOC_SOF_AMD_COMMON");
 
+#ifdef CONFIG_ACPI
 static void acp_sof_scan_pdm_devices(struct snd_sof_dev *sdev,
 				     struct acpi_device *pdm_dev)
 {

@@ -1083,6 +1084,7 @@ static void acp_sof_scan_pdm_devices(struct snd_sof_dev *sdev,
 	}
 	fwnode_handle_put(child);
 }
+#endif
 
 #if IS_ENABLED(CONFIG_SND_SOC_SOF_AMD_SOUNDWIRE)
 static int acp_sof_scan_sdw_devices(struct snd_sof_dev *sdev, u64 addr)

@@ -1299,9 +1301,7 @@ int amd_sof_acp7x_probe(struct snd_sof_dev *sdev)
 	struct pci_dev *pci = to_pci_dev(sdev->dev);
 	struct acp_dev_data *adata;
 	const struct sof_amd_acp_desc *chip;
-	const union acpi_object *obj;
 	struct acpi_device *adev;
-	struct acpi_device *pdm_dev;
 	unsigned int addr;
 	unsigned int irqflags;
 	int ret;

@@ -1369,7 +1369,11 @@ int amd_sof_acp7x_probe(struct snd_sof_dev *sdev)
 		adata->is_sdw_dev = true;
 
 skip_soundwire:
+#ifdef CONFIG_ACPI
 	if (adev) {
+		struct acpi_device *pdm_dev;
+		const union acpi_object *obj;
+
 		/* DMIC ACPI child address is 2 on ACP7x platforms */
 		pdm_dev = acpi_find_child_device(adev, ACP7X_DMIC_ADDR, 0);
 		if (pdm_dev)

@@ -1379,6 +1383,7 @@ int amd_sof_acp7x_probe(struct snd_sof_dev *sdev)
 					   ACPI_TYPE_INTEGER, &obj))
 			adata->acp_sof_signed_firmware_image = obj->integer.value;
 	}
+#endif
 
 	sdev->dsp_box.offset = 0;
 	sdev->dsp_box.size = BOX_SIZE_512;
Signature · 1 line
2.53.0
#1 lore raw
Quoted 13 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

When ACPI is disabled, this driver runs into a build failure:

sound/soc/sof/amd/acp.c:1374:13: error: call to undeclared function 'acpi_find_child_device'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
1374 | pdm_dev = acpi_find_child_device(adev, ACP7X_DMIC_ADDR, 0);

The code block is already guarded by an 'if (adev)' check that can only be true if ACPI is enabled, so add another #ifdef to match this.

Fixes 203ea76c4b0a ("ASoC: SOF: amd: add ACP7.B/7.F PDM controller scan and pdata propagation")
Signed-off-by Arnd Bergmann <[email protected]>
Quoted 7 lines

---
I also tried adding ACPI Kconfig dependencies, but that quickly got out of hand since the acp module is selected by a number of drivers that then all need the same dependency.

It may be possible to work around this in the ACPI code by adding a stub macro for acpi_find_child_device() that does nothing.

I reproduced the issue as you reported. The stub approach seems to work too. I've posted that option here for Mark and Rafael to weigh out, thanks for the idea.

https://lore.kernel.org/linux-acpi/[email protected]/

Quoted 56 lines
Signed-off-by Arnd Bergmann <[email protected]>
---
  sound/soc/sof/amd/acp.c | 9 +++++++--
  1 file changed, 7 insertions(+), 2 deletions(-)
sound/soc/sof/amd/acp.c +7 -2
index b059039c9e0d..4b2d7cb244c9 100644
@@ -1059,6 +1059,7 @@ int amd_sof_acp_resume(struct snd_sof_dev *sdev)
  }
  EXPORT_SYMBOL_NS(amd_sof_acp_resume, "SND_SOC_SOF_AMD_COMMON");
 
+#ifdef CONFIG_ACPI
  static void acp_sof_scan_pdm_devices(struct snd_sof_dev *sdev,
  				     struct acpi_device *pdm_dev)
  {

@@ -1083,6 +1084,7 @@ static void acp_sof_scan_pdm_devices(struct snd_sof_dev *sdev,
  	}
  	fwnode_handle_put(child);
  }
+#endif
 
  #if IS_ENABLED(CONFIG_SND_SOC_SOF_AMD_SOUNDWIRE)
  static int acp_sof_scan_sdw_devices(struct snd_sof_dev *sdev, u64 addr)

@@ -1299,9 +1301,7 @@ int amd_sof_acp7x_probe(struct snd_sof_dev *sdev)
  	struct pci_dev *pci = to_pci_dev(sdev->dev);
  	struct acp_dev_data *adata;
  	const struct sof_amd_acp_desc *chip;
-	const union acpi_object *obj;
  	struct acpi_device *adev;
-	struct acpi_device *pdm_dev;
  	unsigned int addr;
  	unsigned int irqflags;
  	int ret;

@@ -1369,7 +1369,11 @@ int amd_sof_acp7x_probe(struct snd_sof_dev *sdev)
  		adata->is_sdw_dev = true;
 
  skip_soundwire:
+#ifdef CONFIG_ACPI
  	if (adev) {
+		struct acpi_device *pdm_dev;
+		const union acpi_object *obj;
+
  		/* DMIC ACPI child address is 2 on ACP7x platforms */
  		pdm_dev = acpi_find_child_device(adev, ACP7X_DMIC_ADDR, 0);
  		if (pdm_dev)

@@ -1379,6 +1383,7 @@ int amd_sof_acp7x_probe(struct snd_sof_dev *sdev)
  					   ACPI_TYPE_INTEGER, &obj))
  			adata->acp_sof_signed_firmware_image = obj->integer.value;
  	}
+#endif
 
  	sdev->dsp_box.offset = 0;
  	sdev->dsp_box.size = BOX_SIZE_512;
reviewer
#2 lore raw
Quoted 10 lines · Mario Limonciello
Quoted 7 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

When ACPI is disabled, this driver runs into a build failure:

sound/soc/sof/amd/acp.c:1374:13: error: call to undeclared function 'acpi_find_child_device'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration] 1374 |                 pdm_dev = acpi_find_child_device(adev, ACP7X_DMIC_ADDR, 0);

Hi Arnd,

Thanks for catching and fixing this issue.
As the author of the original patch, I didn't receive the original bug report. Was it sent to a limited recipient set, or could there have been a mailing list delivery issue?

-
Vijendar

Quoted 73 lines
Quoted 6 lines

The code block is already guarded by an 'if (adev)' check that can only be true if ACPI is enabled, so add another #ifdef to match this.

Fixes 203ea76c4b0a ("ASoC: SOF: amd: add ACP7.B/7.F PDM controller scan and pdata propagation")
Signed-off-by Arnd Bergmann <[email protected]>
Quoted 7 lines

---
I also tried adding ACPI Kconfig dependencies, but that quickly got out of hand since the acp module is selected by a number of drivers that then all need the same dependency.

It may be possible to work around this in the ACPI code by adding a stub macro for acpi_find_child_device() that does nothing.

I reproduced the issue as you reported.  The stub approach seems to work too.  I've posted that option here for Mark and Rafael to weigh out, thanks for the idea.

https://lore.kernel.org/linux-acpi/[email protected]/

Quoted 53 lines
Signed-off-by Arnd Bergmann <[email protected]>

---
  sound/soc/sof/amd/acp.c | 9 +++++++--

  1 file changed, 7 insertions(+), 2 deletions(-)
sound/soc/sof/amd/acp.c +0 -0
index b059039c9e0d..4b2d7cb244c9 100644
@@ -1059,6 +1059,7 @@ int amd_sof_acp_resume(struct snd_sof_dev *sdev)
   }
   EXPORT_SYMBOL_NS(amd_sof_acp_resume, "SND_SOC_SOF_AMD_COMMON");

  +#ifdef CONFIG_ACPI
  static void acp_sof_scan_pdm_devices(struct snd_sof_dev *sdev, struct acpi_device *pdm_dev)
  {

sound/soc/sof/amd/acp.c +1 -2
@@ -1083,6 +1084,7 @@ static void acp_sof_scan_pdm_devices(struct 
 snd_sof_dev *sdev,
       }
       fwnode_handle_put(child);
   }
+#endif
     #if IS_ENABLED(CONFIG_SND_SOC_SOF_AMD_SOUNDWIRE)
   static int acp_sof_scan_sdw_devices(struct snd_sof_dev *sdev, u64 addr)

@@ -1299,9 +1301,7 @@ int amd_sof_acp7x_probe(struct snd_sof_dev *sdev)
       struct pci_dev *pci = to_pci_dev(sdev->dev);
       struct acp_dev_data *adata;
       const struct sof_amd_acp_desc *chip;
-    const union acpi_object *obj;
       struct acpi_device *adev;
-    struct acpi_device *pdm_dev;
       unsigned int addr;
       unsigned int irqflags;
       int ret;

@@ -1369,7 +1369,11 @@ int amd_sof_acp7x_probe(struct snd_sof_dev *sdev)
           adata->is_sdw_dev = true;

    skip_soundwire:
+#ifdef CONFIG_ACPI
      if (adev) {
+        struct acpi_device *pdm_dev;
+        const union acpi_object *obj;
+
          /* DMIC ACPI child address is 2 on ACP7x platforms */ pdm_dev = acpi_find_child_device(adev, ACP7X_DMIC_ADDR, 0); if (pdm_dev)

sound/soc/sof/amd/acp.c +1 -0
@@ -1379,6 +1383,7 @@ int amd_sof_acp7x_probe(struct snd_sof_dev *sdev)
                          ACPI_TYPE_INTEGER, &obj))
               adata->acp_sof_signed_firmware_image = obj->integer.value;
       }
+#endif
         sdev->dsp_box.offset = 0;
       sdev->dsp_box.size = BOX_SIZE_512;
#3 lore raw
Quoted 12 lines · Mukunda,Vijendar
Quoted 11 lines · Mario Limonciello
Quoted 10 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

When ACPI is disabled, this driver runs into a build failure:

sound/soc/sof/amd/acp.c:1374:13: error: call to undeclared function 'acpi_find_child_device'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration] 1374 |                 pdm_dev = acpi_find_child_device(adev, ACP7X_DMIC_ADDR, 0);

Quoted 6 lines

Hi Arnd,

Thanks for catching and fixing this issue.
As the author of the original patch, I didn't receive the original bug report. Was it sent to a limited recipient set, or could there have been a mailing list delivery issue?

No idea, my report is in the archive, and I assume that is how Mario saw it:

https://lore.kernel.org/linux-sound/[email protected]/

I only ran into the problem on today's linux-next during randconfig testing.

Arnd