[PATCH 1/2] remoteproc: qcom_q6v5_wcss: fix parsing of qcom,halt-regs

[PATCH 1/2] remoteproc: qcom_q6v5_wcss: fix parsing of qcom,halt-regs

[Alexandru Gagniuc]

The "qcom,halt-regs" consists of a phandle reference followed by the
three offsets within syscon for halt registers. Thus, we need to
request 4 integers from of_property_read_variable_u32_array(), with
the halt_reg ofsets at indexes 1, 2, and 3. Offset 0 is the phandle.

With MAX_HALT_REG at 3, of_property_read_variable_u32_array() returns
-EOVERFLOW, causing .probe() to fail.

Increase MAX_HALT_REG to 4, and update the indexes accordingly.

Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
---
 drivers/remoteproc/qcom_q6v5_wcss.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/remoteproc/qcom_q6v5_wcss.c b/drivers/remoteproc/qcom_q6v5_wcss.c
index 07c88623f5978..23ec87827d4f8 100644
--- a/drivers/remoteproc/qcom_q6v5_wcss.c
+++ b/drivers/remoteproc/qcom_q6v5_wcss.c
@@ -85,7 +85,7 @@
 #define TCSR_WCSS_CLK_MASK	0x1F
 #define TCSR_WCSS_CLK_ENABLE	0x14
 
-#define MAX_HALT_REG		3
+#define MAX_HALT_REG		4
 enum {
 	WCSS_IPQ8074,
 	WCSS_QCS404,
@@ -864,9 +864,9 @@ static int q6v5_wcss_init_mmio(struct q6v5_wcss *wcss,
 		return -EINVAL;
 	}
 
-	wcss->halt_q6 = halt_reg[0];
-	wcss->halt_wcss = halt_reg[1];
-	wcss->halt_nc = halt_reg[2];
+	wcss->halt_q6 = halt_reg[1];
+	wcss->halt_wcss = halt_reg[2];
+	wcss->halt_nc = halt_reg[3];
 
 	return 0;
 }
-- 
2.45.1

[PATCH 2/2] remoteproc: qcom_q6v5_wcss: use optional reset for wcss_q6_bcr_reset

[Alexandru Gagniuc]

The "wcss_q6_bcr_reset" is not used on IPQ8074, and IPQ6018. Use
devm_reset_control_get_optional_exclusive() for this reset so that
probe() does not fail on platforms where it is not used.

Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
---
 drivers/remoteproc/qcom_q6v5_wcss.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/remoteproc/qcom_q6v5_wcss.c b/drivers/remoteproc/qcom_q6v5_wcss.c
index 23ec87827d4f8..6e28744ce3f75 100644
--- a/drivers/remoteproc/qcom_q6v5_wcss.c
+++ b/drivers/remoteproc/qcom_q6v5_wcss.c
@@ -811,7 +811,8 @@ static int q6v5_wcss_init_reset(struct q6v5_wcss *wcss,
 		}
 	}
 
-	wcss->wcss_q6_bcr_reset = devm_reset_control_get_exclusive(dev, "wcss_q6_bcr_reset");
+	wcss->wcss_q6_bcr_reset = devm_reset_control_get_optional_exclusive(dev,
+							"wcss_q6_bcr_reset");
 	if (IS_ERR(wcss->wcss_q6_bcr_reset)) {
 		dev_err(wcss->dev, "unable to acquire wcss_q6_bcr_reset\n");
 		return PTR_ERR(wcss->wcss_q6_bcr_reset);
-- 
2.45.1

Re: [PATCH 1/2] remoteproc: qcom_q6v5_wcss: fix parsing of qcom,halt-regs

[Bjorn Andersson]

On Fri, Nov 28, 2025 at 07:32:05PM -0600, Alexandru Gagniuc wrote:
> The "qcom,halt-regs" consists of a phandle reference followed by the
> three offsets within syscon for halt registers. Thus, we need to
> request 4 integers from of_property_read_variable_u32_array(), with
> the halt_reg ofsets at indexes 1, 2, and 3. Offset 0 is the phandle.
> 
> With MAX_HALT_REG at 3, of_property_read_variable_u32_array() returns
> -EOVERFLOW, causing .probe() to fail.
> 
> Increase MAX_HALT_REG to 4, and update the indexes accordingly.
> 

Good catch, thanks

Fixes: 0af65b9b915e ("remoteproc: qcom: wcss: Add non pas wcss Q6 support for QCS404")

Regards,
Bjorn

> Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
> ---
>  drivers/remoteproc/qcom_q6v5_wcss.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/remoteproc/qcom_q6v5_wcss.c b/drivers/remoteproc/qcom_q6v5_wcss.c
> index 07c88623f5978..23ec87827d4f8 100644
> --- a/drivers/remoteproc/qcom_q6v5_wcss.c
> +++ b/drivers/remoteproc/qcom_q6v5_wcss.c
> @@ -85,7 +85,7 @@
>  #define TCSR_WCSS_CLK_MASK	0x1F
>  #define TCSR_WCSS_CLK_ENABLE	0x14
>  
> -#define MAX_HALT_REG		3
> +#define MAX_HALT_REG		4
>  enum {
>  	WCSS_IPQ8074,
>  	WCSS_QCS404,
> @@ -864,9 +864,9 @@ static int q6v5_wcss_init_mmio(struct q6v5_wcss *wcss,
>  		return -EINVAL;
>  	}
>  
> -	wcss->halt_q6 = halt_reg[0];
> -	wcss->halt_wcss = halt_reg[1];
> -	wcss->halt_nc = halt_reg[2];
> +	wcss->halt_q6 = halt_reg[1];
> +	wcss->halt_wcss = halt_reg[2];
> +	wcss->halt_nc = halt_reg[3];
>  
>  	return 0;
>  }
> -- 
> 2.45.1
> 

Re: [PATCH 2/2] remoteproc: qcom_q6v5_wcss: use optional reset for wcss_q6_bcr_reset

[Konrad Dybcio]

On 11/29/25 2:32 AM, Alexandru Gagniuc wrote:
> The "wcss_q6_bcr_reset" is not used on IPQ8074, and IPQ6018. Use
> devm_reset_control_get_optional_exclusive() for this reset so that
> probe() does not fail on platforms where it is not used.
> 
> Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
> ---

This reset is not even used (or documented anywhere).. the closest
I can find is "wcss_q6_reset" in qcom,q6v5.txt, initially documented
in:

Fixes: 3a3d4163e0bf ("remoteproc: qcom: Introduce Hexagon V5 based WCSS driver")

which claims it was made for.. IPQ8074

Peeking at the GCC driver, this reset is most likely present as
GCC_WCSS_Q6_BCR

Konrad

Re: [PATCH 2/2] remoteproc: qcom_q6v5_wcss: use optional reset for wcss_q6_bcr_reset

[mr.nuke.me]

On 12/1/25 5:21 AM, Konrad Dybcio wrote:
> On 11/29/25 2:32 AM, Alexandru Gagniuc wrote:
>> The "wcss_q6_bcr_reset" is not used on IPQ8074, and IPQ6018. Use
>> devm_reset_control_get_optional_exclusive() for this reset so that
>> probe() does not fail on platforms where it is not used.
>>
>> Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
>> ---
> 
> This reset is not even used (or documented anywhere).. the closest
> I can find is "wcss_q6_reset" in qcom,q6v5.txt, initially documented
> in:
> 
> Fixes: 3a3d4163e0bf ("remoteproc: qcom: Introduce Hexagon V5 based WCSS driver")
> 
> which claims it was made for.. IPQ8074
> 
> Peeking at the GCC driver, this reset is most likely present as
> GCC_WCSS_Q6_BCR

The downstream kernel uses GCC_WCSS_Q6_BCR for ipq8074 [1] and ipq6018 [2].
They only use of ->wcss_q6_bcr_reset in the QCN404 path, which does not use
->wcss_q6_reset. So maybe we can get away with calling it "wcss_q6_reset",
store the pointer in ->wcss_q6_reset, and drop '.wcss_q6_reset_required'

Which path do you think is the most appropriate?

Alex

[1] https://git.codelinaro.org/clo/qsdk/oss/kernel/linux-ipq-5.4/-/blob/NHSS.QSDK.12.5.r5/arch/arm64/boot/dts/qcom/ipq8074.dtsi?ref_type=heads#L1060
[2] https://git.codelinaro.org/clo/qsdk/oss/kernel/linux-ipq-5.4/-/blob/NHSS.QSDK.12.5.r5/arch/arm64/boot/dts/qcom/ipq6018.dtsi?ref_type=heads#L1440
[3] https://git.codelinaro.org/clo/qsdk/oss/kernel/linux-ipq-5.4/-/blob/NHSS.QSDK.12.5.r5/drivers/remoteproc/qcom_q6v5_wcss.c#L967

Re: [PATCH 2/2] remoteproc: qcom_q6v5_wcss: use optional reset for wcss_q6_bcr_reset

[Konrad Dybcio]

On 12/1/25 10:04 PM, mr.nuke.me@gmail.com wrote:
> On 12/1/25 5:21 AM, Konrad Dybcio wrote:
>> On 11/29/25 2:32 AM, Alexandru Gagniuc wrote:
>>> The "wcss_q6_bcr_reset" is not used on IPQ8074, and IPQ6018. Use
>>> devm_reset_control_get_optional_exclusive() for this reset so that
>>> probe() does not fail on platforms where it is not used.
>>>
>>> Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
>>> ---
>>
>> This reset is not even used (or documented anywhere).. the closest
>> I can find is "wcss_q6_reset" in qcom,q6v5.txt, initially documented
>> in:
>>
>> Fixes: 3a3d4163e0bf ("remoteproc: qcom: Introduce Hexagon V5 based WCSS driver")
>>
>> which claims it was made for.. IPQ8074
>>
>> Peeking at the GCC driver, this reset is most likely present as
>> GCC_WCSS_Q6_BCR
> 
> The downstream kernel uses GCC_WCSS_Q6_BCR for ipq8074 [1] and ipq6018 [2].
> They only use of ->wcss_q6_bcr_reset in the QCN404 path, which does not use
> ->wcss_q6_reset. So maybe we can get away with calling it "wcss_q6_reset",
> store the pointer in ->wcss_q6_reset, and drop '.wcss_q6_reset_required'

"BCR reset" is like saying "PIN number", both of the ones you mentioned are
referring to the same thing

Konrad

Re: [PATCH 1/2] remoteproc: qcom_q6v5_wcss: fix parsing of qcom,halt-regs

[Alex G.]

On 11/29/25 3:18 PM, Bjorn Andersson wrote:
> On Fri, Nov 28, 2025 at 07:32:05PM -0600, Alexandru Gagniuc wrote:
>> The "qcom,halt-regs" consists of a phandle reference followed by th
>> three offsets within syscon for halt registers. Thus, we need to
>> request 4 integers from of_property_read_variable_u32_array(), with
>> the halt_reg ofsets at indexes 1, 2, and 3. Offset 0 is the phandle.
>>
>> With MAX_HALT_REG at 3, of_property_read_variable_u32_array() returns
>> -EOVERFLOW, causing .probe() to fail.
>>
>> Increase MAX_HALT_REG to 4, and update the indexes accordingly.
>>
> 
> Good catch, thanks
> 
> Fixes: 0af65b9b915e ("remoteproc: qcom: wcss: Add non pas wcss Q6 support for QCS404")

Hi Bjorn,

I noticed that v1 of this series is included in the pull for v6.19 [1], 
even though there is a v2 [2] with some of your and Konrad's feedback 
included. I wanted to check if this is your intention. I am okay to go 
with v1, and am happy to submit any further improvements after the merge 
window.

Alex

[1] 
https://lore.kernel.org/linux-remoteproc/20251205200342.119676-1-andersson@kernel.org/T/#u
[2] 
lore.kernel.org/linux-remoteproc/20251202162626.1135615-1-mr.nuke.me@gmail.com/#r> 
Regards,
> Bjorn
> 
>> Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
>> ---
>>   drivers/remoteproc/qcom_q6v5_wcss.c | 8 ++++----
>>   1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/remoteproc/qcom_q6v5_wcss.c b/drivers/remoteproc/qcom_q6v5_wcss.c
>> index 07c88623f5978..23ec87827d4f8 100644
>> --- a/drivers/remoteproc/qcom_q6v5_wcss.c
>> +++ b/drivers/remoteproc/qcom_q6v5_wcss.c
>> @@ -85,7 +85,7 @@
>>   #define TCSR_WCSS_CLK_MASK	0x1F
>>   #define TCSR_WCSS_CLK_ENABLE	0x14
>>   
>> -#define MAX_HALT_REG		3
>> +#define MAX_HALT_REG		4
>>   enum {
>>   	WCSS_IPQ8074,
>>   	WCSS_QCS404,
>> @@ -864,9 +864,9 @@ static int q6v5_wcss_init_mmio(struct q6v5_wcss *wcss,
>>   		return -EINVAL;
>>   	}
>>   
>> -	wcss->halt_q6 = halt_reg[0];
>> -	wcss->halt_wcss = halt_reg[1];
>> -	wcss->halt_nc = halt_reg[2];
>> +	wcss->halt_q6 = halt_reg[1];
>> +	wcss->halt_wcss = halt_reg[2];
>> +	wcss->halt_nc = halt_reg[3];
>>   
>>   	return 0;
>>   }
>> -- 
>> 2.45.1
>>

Re: [PATCH 1/2] remoteproc: qcom_q6v5_wcss: fix parsing of qcom,halt-regs

[Bjorn Andersson]

On Fri, Dec 05, 2025 at 08:19:57PM -0600, Alex G. wrote:
> On 11/29/25 3:18 PM, Bjorn Andersson wrote:
> > On Fri, Nov 28, 2025 at 07:32:05PM -0600, Alexandru Gagniuc wrote:
> > > The "qcom,halt-regs" consists of a phandle reference followed by th
> > > three offsets within syscon for halt registers. Thus, we need to
> > > request 4 integers from of_property_read_variable_u32_array(), with
> > > the halt_reg ofsets at indexes 1, 2, and 3. Offset 0 is the phandle.
> > > 
> > > With MAX_HALT_REG at 3, of_property_read_variable_u32_array() returns
> > > -EOVERFLOW, causing .probe() to fail.
> > > 
> > > Increase MAX_HALT_REG to 4, and update the indexes accordingly.
> > > 
> > 
> > Good catch, thanks
> > 
> > Fixes: 0af65b9b915e ("remoteproc: qcom: wcss: Add non pas wcss Q6 support for QCS404")
> 
> Hi Bjorn,
> 
> I noticed that v1 of this series is included in the pull for v6.19 [1], even
> though there is a v2 [2] with some of your and Konrad's feedback included. I
> wanted to check if this is your intention. I am okay to go with v1, and am
> happy to submit any further improvements after the merge window.
> 

That's strange, I should have sent you a "thank you" email when I
applied it. I added my Fixes, and must have missed Konrad's input (or it
hadn't shown up yet?)

Without looking at Konrad's feedback, please send a new patch on top of
what I did merge if the changes would be beneficial.

Thanks,
Bjorn

> Alex
> 
> [1] https://lore.kernel.org/linux-remoteproc/20251205200342.119676-1-andersson@kernel.org/T/#u
> [2] lore.kernel.org/linux-remoteproc/20251202162626.1135615-1-mr.nuke.me@gmail.com/#r>
> Regards,
> > Bjorn
> > 
> > > Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
> > > ---
> > >   drivers/remoteproc/qcom_q6v5_wcss.c | 8 ++++----
> > >   1 file changed, 4 insertions(+), 4 deletions(-)
> > > 
> > > diff --git a/drivers/remoteproc/qcom_q6v5_wcss.c b/drivers/remoteproc/qcom_q6v5_wcss.c
> > > index 07c88623f5978..23ec87827d4f8 100644
> > > --- a/drivers/remoteproc/qcom_q6v5_wcss.c
> > > +++ b/drivers/remoteproc/qcom_q6v5_wcss.c
> > > @@ -85,7 +85,7 @@
> > >   #define TCSR_WCSS_CLK_MASK	0x1F
> > >   #define TCSR_WCSS_CLK_ENABLE	0x14
> > > -#define MAX_HALT_REG		3
> > > +#define MAX_HALT_REG		4
> > >   enum {
> > >   	WCSS_IPQ8074,
> > >   	WCSS_QCS404,
> > > @@ -864,9 +864,9 @@ static int q6v5_wcss_init_mmio(struct q6v5_wcss *wcss,
> > >   		return -EINVAL;
> > >   	}
> > > -	wcss->halt_q6 = halt_reg[0];
> > > -	wcss->halt_wcss = halt_reg[1];
> > > -	wcss->halt_nc = halt_reg[2];
> > > +	wcss->halt_q6 = halt_reg[1];
> > > +	wcss->halt_wcss = halt_reg[2];
> > > +	wcss->halt_nc = halt_reg[3];
> > >   	return 0;
> > >   }
> > > -- 
> > > 2.45.1
> > > 
>