[PATCH RESEND v1 0/3] Fixes and cleanup for CS35L41 HDA

[PATCH RESEND v1 0/3] Fixes and cleanup for CS35L41 HDA

[Stefan Binding]

Several minor issues were found during additional testing and
static analysis. These patches fix these minor issues.

Stefan Binding (3):
  ALSA: hda: cs35l41: Clean up Firmware Load Controls
  ALSA: hda: cs35l41: Fix endian conversions
  ALSA: hda/realtek: Delete cs35l41 component master during free

 sound/pci/hda/cs35l41_hda.c   | 40 ++++++++++++++++-------------------
 sound/pci/hda/patch_realtek.c |  2 ++
 2 files changed, 20 insertions(+), 22 deletions(-)

-- 
2.34.1

[PATCH RESEND v1 1/3] ALSA: hda: cs35l41: Clean up Firmware Load Controls

[Stefan Binding]

Ensure Firmware Load control and Firmware Type control
returns 1 when the value changes.

Remove fw_mutex from firmware load control put, since it is
unnecessary, and prevents any possibility of mutex inversion.

Signed-off-by: Stefan Binding <sbinding@opensource.cirrus.com>
---
 sound/pci/hda/cs35l41_hda.c | 24 ++++++++++--------------
 1 file changed, 10 insertions(+), 14 deletions(-)

diff --git a/sound/pci/hda/cs35l41_hda.c b/sound/pci/hda/cs35l41_hda.c
index b5210abb5141f..d100189e15b83 100644
--- a/sound/pci/hda/cs35l41_hda.c
+++ b/sound/pci/hda/cs35l41_hda.c
@@ -835,34 +835,26 @@ static int cs35l41_fw_load_ctl_put(struct snd_kcontrol *kcontrol,
 				   struct snd_ctl_elem_value *ucontrol)
 {
 	struct cs35l41_hda *cs35l41 = snd_kcontrol_chip(kcontrol);
-	unsigned int ret = 0;
-
-	mutex_lock(&cs35l41->fw_mutex);
 
 	if (cs35l41->request_fw_load == ucontrol->value.integer.value[0])
-		goto err;
+		return 0;
 
 	if (cs35l41->fw_request_ongoing) {
 		dev_dbg(cs35l41->dev, "Existing request not complete\n");
-		ret = -EBUSY;
-		goto err;
+		return -EBUSY;
 	}
 
 	/* Check if playback is ongoing when initial request is made */
 	if (cs35l41->playback_started) {
 		dev_err(cs35l41->dev, "Cannot Load/Unload firmware during Playback\n");
-		ret = -EBUSY;
-		goto err;
+		return -EBUSY;
 	}
 
 	cs35l41->fw_request_ongoing = true;
 	cs35l41->request_fw_load = ucontrol->value.integer.value[0];
 	schedule_work(&cs35l41->fw_load_work);
 
-err:
-	mutex_unlock(&cs35l41->fw_mutex);
-
-	return ret;
+	return 1;
 }
 
 static int cs35l41_fw_type_ctl_get(struct snd_kcontrol *kcontrol,
@@ -881,8 +873,12 @@ static int cs35l41_fw_type_ctl_put(struct snd_kcontrol *kcontrol,
 	struct cs35l41_hda *cs35l41 = snd_kcontrol_chip(kcontrol);
 
 	if (ucontrol->value.enumerated.item[0] < HDA_CS_DSP_NUM_FW) {
-		cs35l41->firmware_type = ucontrol->value.enumerated.item[0];
-		return 0;
+		if (cs35l41->firmware_type != ucontrol->value.enumerated.item[0]) {
+			cs35l41->firmware_type = ucontrol->value.enumerated.item[0];
+			return 1;
+		} else {
+			return 0;
+		}
 	}
 
 	return -EINVAL;
-- 
2.34.1

[PATCH RESEND v1 2/3] ALSA: hda: cs35l41: Fix endian conversions

[Stefan Binding]

Found during static analysis, ensure variables are correct
types before endian conversion.

Signed-off-by: Stefan Binding <sbinding@opensource.cirrus.com>
---
 sound/pci/hda/cs35l41_hda.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/sound/pci/hda/cs35l41_hda.c b/sound/pci/hda/cs35l41_hda.c
index d100189e15b83..b02462ae21f04 100644
--- a/sound/pci/hda/cs35l41_hda.c
+++ b/sound/pci/hda/cs35l41_hda.c
@@ -308,8 +308,8 @@ static int cs35l41_request_firmware_files(struct cs35l41_hda *cs35l41,
 }
 
 #if IS_ENABLED(CONFIG_EFI)
-static int cs35l41_apply_calibration(struct cs35l41_hda *cs35l41, unsigned int ambient,
-				     unsigned int r0, unsigned int status, unsigned int checksum)
+static int cs35l41_apply_calibration(struct cs35l41_hda *cs35l41, __be32 ambient, __be32 r0,
+				     __be32 status, __be32 checksum)
 {
 	int ret;
 
@@ -379,10 +379,10 @@ static int cs35l41_save_calibration(struct cs35l41_hda *cs35l41)
 
 				/* Calibration can only be applied whilst the DSP is not running */
 				ret = cs35l41_apply_calibration(cs35l41,
-								cpu_to_be32(cl->calAmbient),
-								cpu_to_be32(cl->calR),
-								cpu_to_be32(cl->calStatus),
-								cpu_to_be32(cl->calR + 1));
+								(__be32)cpu_to_be32(cl->calAmbient),
+								(__be32)cpu_to_be32(cl->calR),
+								(__be32)cpu_to_be32(cl->calStatus),
+								(__be32)cpu_to_be32(cl->calR + 1));
 			}
 		}
 		vfree(data);
@@ -745,7 +745,7 @@ static int cs35l41_runtime_resume(struct device *dev)
 
 static int cs35l41_smart_amp(struct cs35l41_hda *cs35l41)
 {
-	int halo_sts;
+	__be32 halo_sts;
 	int ret;
 
 	ret = cs35l41_init_dsp(cs35l41);
@@ -773,7 +773,7 @@ static int cs35l41_smart_amp(struct cs35l41_hda *cs35l41)
 				&halo_sts, sizeof(halo_sts));
 
 	if (ret) {
-		dev_err(cs35l41->dev, "Timeout waiting for HALO Core to start. State: %d\n",
+		dev_err(cs35l41->dev, "Timeout waiting for HALO Core to start. State: %u\n",
 			 halo_sts);
 		goto clean_dsp;
 	}
-- 
2.34.1

Re: [PATCH RESEND v1 2/3] ALSA: hda: cs35l41: Fix endian conversions

[Takashi Iwai]

On Thu, 25 May 2023 15:59:54 +0200,
Stefan Binding wrote:
> 
> @@ -379,10 +379,10 @@ static int cs35l41_save_calibration(struct cs35l41_hda *cs35l41)
>  
>  				/* Calibration can only be applied whilst the DSP is not running */
>  				ret = cs35l41_apply_calibration(cs35l41,
> -								cpu_to_be32(cl->calAmbient),
> -								cpu_to_be32(cl->calR),
> -								cpu_to_be32(cl->calStatus),
> -								cpu_to_be32(cl->calR + 1));
> +								(__be32)cpu_to_be32(cl->calAmbient),
> +								(__be32)cpu_to_be32(cl->calR),
> +								(__be32)cpu_to_be32(cl->calStatus),
> +								(__be32)cpu_to_be32(cl->calR + 1));

Do we really need those cast?  Even if yes, it must be with __force
prefix for the endian cast in general.


thanks,

Takashi

Re: [PATCH RESEND v1 2/3] ALSA: hda: cs35l41: Fix endian conversions

[Stefan Binding]

Hi Takashi,

On 05/06/2023 08:21, Takashi Iwai wrote:
> On Thu, 25 May 2023 15:59:54 +0200,
> Stefan Binding wrote:
>> @@ -379,10 +379,10 @@ static int cs35l41_save_calibration(struct cs35l41_hda *cs35l41)
>>   
>>   				/* Calibration can only be applied whilst the DSP is not running */
>>   				ret = cs35l41_apply_calibration(cs35l41,
>> -								cpu_to_be32(cl->calAmbient),
>> -								cpu_to_be32(cl->calR),
>> -								cpu_to_be32(cl->calStatus),
>> -								cpu_to_be32(cl->calR + 1));
>> +								(__be32)cpu_to_be32(cl->calAmbient),
>> +								(__be32)cpu_to_be32(cl->calR),
>> +								(__be32)cpu_to_be32(cl->calStatus),
>> +								(__be32)cpu_to_be32(cl->calR + 1));
> Do we really need those cast?  Even if yes, it must be with __force
> prefix for the endian cast in general.

These casts were added because we found some warnings when we ran the 
static analyzer sparse locally.
I think these warnings are very minor, and we can drop this patch if you 
prefer?

Thanks,

Stefan

>
> thanks,
>
> Takashi

Re: [PATCH RESEND v1 2/3] ALSA: hda: cs35l41: Fix endian conversions

[Takashi Iwai]

On Mon, 05 Jun 2023 14:50:54 +0200,
Stefan Binding wrote:
> 
> Hi Takashi,
> 
> On 05/06/2023 08:21, Takashi Iwai wrote:
> > On Thu, 25 May 2023 15:59:54 +0200,
> > Stefan Binding wrote:
> >> @@ -379,10 +379,10 @@ static int cs35l41_save_calibration(struct cs35l41_hda *cs35l41)
> >>     				/* Calibration can only be applied
> >> whilst the DSP is not running */
> >>   				ret = cs35l41_apply_calibration(cs35l41,
> >> -								cpu_to_be32(cl->calAmbient),
> >> -								cpu_to_be32(cl->calR),
> >> -								cpu_to_be32(cl->calStatus),
> >> -								cpu_to_be32(cl->calR + 1));
> >> +								(__be32)cpu_to_be32(cl->calAmbient),
> >> +								(__be32)cpu_to_be32(cl->calR),
> >> +								(__be32)cpu_to_be32(cl->calStatus),
> >> +								(__be32)cpu_to_be32(cl->calR + 1));
> > Do we really need those cast?  Even if yes, it must be with __force
> > prefix for the endian cast in general.
> 
> These casts were added because we found some warnings when we ran the
> static analyzer sparse locally.
> I think these warnings are very minor, and we can drop this patch if
> you prefer?

The warnings must be bogus, or maybe pointing to other things?
The cpu_to_be32() macro itself must return a __be32 value, hence it
makes no sense to add an extra cast .

If the static analysis still shows such a warning, it should be fixed
differently -- either fix the analyzer or fix the cpu_to_be32() macro
itself.

The changes of the argument types to __be32 are fine.  I'm arguing
only about those unnecessary cast.


thanks,

Takashi

Re: [PATCH RESEND v1 2/3] ALSA: hda: cs35l41: Fix endian conversions

[Stefan Binding]

Hi Takashi,

On 05/06/2023 14:17, Takashi Iwai wrote:
> On Mon, 05 Jun 2023 14:50:54 +0200,
> Stefan Binding wrote:
>> Hi Takashi,
>>
>> On 05/06/2023 08:21, Takashi Iwai wrote:
>>> On Thu, 25 May 2023 15:59:54 +0200,
>>> Stefan Binding wrote:
>>>> @@ -379,10 +379,10 @@ static int cs35l41_save_calibration(struct cs35l41_hda *cs35l41)
>>>>      				/* Calibration can only be applied
>>>> whilst the DSP is not running */
>>>>    				ret = cs35l41_apply_calibration(cs35l41,
>>>> -								cpu_to_be32(cl->calAmbient),
>>>> -								cpu_to_be32(cl->calR),
>>>> -								cpu_to_be32(cl->calStatus),
>>>> -								cpu_to_be32(cl->calR + 1));
>>>> +								(__be32)cpu_to_be32(cl->calAmbient),
>>>> +								(__be32)cpu_to_be32(cl->calR),
>>>> +								(__be32)cpu_to_be32(cl->calStatus),
>>>> +								(__be32)cpu_to_be32(cl->calR + 1));
>>> Do we really need those cast?  Even if yes, it must be with __force
>>> prefix for the endian cast in general.
>> These casts were added because we found some warnings when we ran the
>> static analyzer sparse locally.
>> I think these warnings are very minor, and we can drop this patch if
>> you prefer?
> The warnings must be bogus, or maybe pointing to other things?
> The cpu_to_be32() macro itself must return a __be32 value, hence it
> makes no sense to add an extra cast .
>
> If the static analysis still shows such a warning, it should be fixed
> differently -- either fix the analyzer or fix the cpu_to_be32() macro
> itself.
>
> The changes of the argument types to __be32 are fine.  I'm arguing
> only about those unnecessary cast.

You are correct, I double checked and the cast is not needed. I'll push 
up a v2.

Thanks,

Stefan

>
>
> thanks,
>
> Takashi

[PATCH RESEND v1 3/3] ALSA: hda/realtek: Delete cs35l41 component master during free

[Stefan Binding]

This ensures that the driver is properly cleaned up when freed.

Signed-off-by: Stefan Binding <sbinding@opensource.cirrus.com>
---
 sound/pci/hda/patch_realtek.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/sound/pci/hda/patch_realtek.c b/sound/pci/hda/patch_realtek.c
index 7b5f194513c7b..e3774903918fe 100644
--- a/sound/pci/hda/patch_realtek.c
+++ b/sound/pci/hda/patch_realtek.c
@@ -6757,6 +6757,8 @@ static void cs35l41_generic_fixup(struct hda_codec *cdc, int action, const char
 		else
 			spec->gen.pcm_playback_hook = comp_generic_playback_hook;
 		break;
+	case HDA_FIXUP_ACT_FREE:
+		component_master_del(dev, &comp_master_ops);
 	}
 }
 
-- 
2.34.1