[PATCH v2 0/3] cpufreq: qcom: handle ipq806x with no SMEM

[PATCH v2 0/3] cpufreq: qcom: handle ipq806x with no SMEM

[Christian Marangi]

This small series handle a small device family of ipq806x
devices (Google OnHub) that doesn't have SMEM init.

We improve the SMEM driver and apply a workaround in
the cpufreq driver.

Changes v2:
- Rename error macro to INIT_ERR_PTR
- Return -ENODEV from smem probe
- Restructure if condition in cpufreq driver

Christian Marangi (3):
  err.h: add INIT_ERR_PTR macro
  soc: qcom: smem: better track SMEM uninitialized state
  cpufreq: qcom-nvmem: add compatible fallback for ipq806x for no SMEM

 drivers/cpufreq/qcom-cpufreq-nvmem.c | 17 +++++++++++++++--
 drivers/soc/qcom/smem.c              | 26 +++++++++++++++-----------
 include/linux/err.h                  |  8 ++++++++
 3 files changed, 38 insertions(+), 13 deletions(-)

-- 
2.51.0

[PATCH v2 1/3] err.h: add INIT_ERR_PTR macro

[Christian Marangi]

Add INIT_ERR_PTR macro to initialize static variables with error
pointers. This might be useful for specific case where there is a static
variable initialized to an error condition and then later set to the
real handle once probe finish/completes.

This is to handle compilation problems like:

error: initializer element is not constant

where ERR_PTR can't be used.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
---
 include/linux/err.h | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/include/linux/err.h b/include/linux/err.h
index 1d60aa86db53..8c37be0620ab 100644
--- a/include/linux/err.h
+++ b/include/linux/err.h
@@ -41,6 +41,14 @@ static inline void * __must_check ERR_PTR(long error)
 	return (void *) error;
 }
 
+/**
+ * INIT_ERR_PTR - Init a const error pointer.
+ * @error: A negative error code.
+ *
+ * Like ERR_PTR(), but usable to initialize static variables.
+ */
+#define INIT_ERR_PTR(error) ((void *)(error))
+
 /* Return the pointer in the percpu address space. */
 #define ERR_PTR_PCPU(error) ((void __percpu *)(unsigned long)ERR_PTR(error))
 
-- 
2.51.0

Re: [PATCH v2 1/3] err.h: add INIT_ERR_PTR macro

[Andy Shevchenko]

On Fri, Oct 31, 2025 at 02:08:32PM +0100, Christian Marangi wrote:
> Add INIT_ERR_PTR macro to initialize static variables with error

INIT_ERR_PTR()

> pointers. This might be useful for specific case where there is a static
> variable initialized to an error condition and then later set to the
> real handle once probe finish/completes.
> 
> This is to handle compilation problems like:
> 
> error: initializer element is not constant
> 
> where ERR_PTR can't be used.

ERR_PTR()

In case you need a new version, please address the above.
Anyways, LGTM now,
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

-- 
With Best Regards,
Andy Shevchenko

[PATCH v2 2/3] soc: qcom: smem: better track SMEM uninitialized state

[Christian Marangi]

There is currently a problem where, in the specific case of SMEM not
initialized by SBL, any SMEM API wrongly returns PROBE_DEFER
communicating wrong info to any user of this API.

A better way to handle this would be to track the SMEM state and return
a different kind of error than PROBE_DEFER.

Rework the __smem handle to always init it to the error pointer
-EPROBE_DEFER following what is already done by the SMEM API.
If we detect that the SBL didn't initialized SMEM, set the __smem handle
to the error pointer -ENODEV.
Also rework the SMEM API to handle the __smem handle to be an error
pointer and return it appropriately.

This way user of the API can react and return a proper error or use
fallback way for the failing API.

While at it, change the return error when SMEM is not initialized by SBL
also to -ENODEV to make it consistent with the __smem handle and use
dev_err_probe() helper to return the message.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
---
 drivers/soc/qcom/smem.c | 26 +++++++++++++++-----------
 1 file changed, 15 insertions(+), 11 deletions(-)

diff --git a/drivers/soc/qcom/smem.c b/drivers/soc/qcom/smem.c
index 592819701809..5b7626528284 100644
--- a/drivers/soc/qcom/smem.c
+++ b/drivers/soc/qcom/smem.c
@@ -353,8 +353,12 @@ static void *cached_entry_to_item(struct smem_private_entry *e)
 	return p - le32_to_cpu(e->size);
 }
 
-/* Pointer to the one and only smem handle */
-static struct qcom_smem *__smem;
+/*
+ * Pointer to the one and only smem handle.
+ * Init to -EPROBE_DEFER to signal SMEM still has to be probed.
+ * Can be set to -ENODEV if SMEM is not initialized by SBL.
+ */
+static struct qcom_smem *__smem = INIT_ERR_PTR(-EPROBE_DEFER);
 
 /* Timeout (ms) for the trylock of remote spinlocks */
 #define HWSPINLOCK_TIMEOUT	1000
@@ -508,8 +512,8 @@ int qcom_smem_alloc(unsigned host, unsigned item, size_t size)
 	unsigned long flags;
 	int ret;
 
-	if (!__smem)
-		return -EPROBE_DEFER;
+	if (IS_ERR(__smem))
+		return PTR_ERR(__smem);
 
 	if (item < SMEM_ITEM_LAST_FIXED) {
 		dev_err(__smem->dev,
@@ -685,10 +689,10 @@ static void *qcom_smem_get_private(struct qcom_smem *smem,
 void *qcom_smem_get(unsigned host, unsigned item, size_t *size)
 {
 	struct smem_partition *part;
-	void *ptr = ERR_PTR(-EPROBE_DEFER);
+	void *ptr;
 
-	if (!__smem)
-		return ptr;
+	if (IS_ERR(__smem))
+		return __smem;
 
 	if (WARN_ON(item >= __smem->item_count))
 		return ERR_PTR(-EINVAL);
@@ -723,8 +727,8 @@ int qcom_smem_get_free_space(unsigned host)
 	struct smem_header *header;
 	unsigned ret;
 
-	if (!__smem)
-		return -EPROBE_DEFER;
+	if (IS_ERR(__smem))
+		return PTR_ERR(__smem);
 
 	if (host < SMEM_HOST_COUNT && __smem->partitions[host].virt_base) {
 		part = &__smem->partitions[host];
@@ -1181,8 +1185,8 @@ static int qcom_smem_probe(struct platform_device *pdev)
 	header = smem->regions[0].virt_base;
 	if (le32_to_cpu(header->initialized) != 1 ||
 	    le32_to_cpu(header->reserved)) {
-		dev_err(&pdev->dev, "SMEM is not initialized by SBL\n");
-		return -EINVAL;
+		__smem = ERR_PTR(-ENODEV);
+		return dev_err_probe(&pdev->dev, PTR_ERR(__smem), "SMEM is not initialized by SBL\n");
 	}
 
 	hwlock_id = of_hwspin_lock_get_id(pdev->dev.of_node, 0);
-- 
2.51.0

[PATCH v2 3/3] cpufreq: qcom-nvmem: add compatible fallback for ipq806x for no SMEM

[Christian Marangi]

On some IPQ806x SoC SMEM might be not initialized by SBL. This is the
case for some Google devices (the OnHub family) that can't make use of
SMEM to detect the SoC ID.

To handle these specific case, check if the SMEM is not initialized (by
checking if the qcom_smem_get_soc_id returns -ENODEV) and fallback to
OF machine compatible checking to identify the SoC variant.

Notice that the checking order is important as the machine compatible
are normally defined with the specific one following the generic SoC
(for example compatible = "qcom,ipq8065", "qcom,ipq8064").

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
---
 drivers/cpufreq/qcom-cpufreq-nvmem.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/drivers/cpufreq/qcom-cpufreq-nvmem.c b/drivers/cpufreq/qcom-cpufreq-nvmem.c
index 3a8ed723a23e..5a9bd780a4f3 100644
--- a/drivers/cpufreq/qcom-cpufreq-nvmem.c
+++ b/drivers/cpufreq/qcom-cpufreq-nvmem.c
@@ -257,8 +257,8 @@ static int qcom_cpufreq_ipq8064_name_version(struct device *cpu_dev,
 					     char **pvs_name,
 					     struct qcom_cpufreq_drv *drv)
 {
+	int msm_id = -1, ret = 0;
 	int speed = 0, pvs = 0;
-	int msm_id, ret = 0;
 	u8 *speedbin;
 	size_t len;
 
@@ -275,8 +275,21 @@ static int qcom_cpufreq_ipq8064_name_version(struct device *cpu_dev,
 	get_krait_bin_format_a(cpu_dev, &speed, &pvs, speedbin);
 
 	ret = qcom_smem_get_soc_id(&msm_id);
-	if (ret)
+	if (ret == -ENODEV) {
+		/* Fallback to compatible match with no SMEM initialized */
+		ret = 0;
+		if (of_machine_is_compatible("qcom,ipq8062"))
+			msm_id = QCOM_ID_IPQ8062;
+		else if (of_machine_is_compatible("qcom,ipq8065") ||
+			 of_machine_is_compatible("qcom,ipq8069"))
+			msm_id = QCOM_ID_IPQ8065;
+		else if (of_machine_is_compatible("qcom,ipq8064") ||
+			 of_machine_is_compatible("qcom,ipq8066") ||
+			 of_machine_is_compatible("qcom,ipq8068"))
+			msm_id = QCOM_ID_IPQ8064;
+	} else if (ret) {
 		goto exit;
+	}
 
 	switch (msm_id) {
 	case QCOM_ID_IPQ8062:
-- 
2.51.0

Re: [PATCH v2 3/3] cpufreq: qcom-nvmem: add compatible fallback for ipq806x for no SMEM

[Andy Shevchenko]

On Fri, Oct 31, 2025 at 02:08:34PM +0100, Christian Marangi wrote:
> On some IPQ806x SoC SMEM might be not initialized by SBL. This is the
> case for some Google devices (the OnHub family) that can't make use of
> SMEM to detect the SoC ID.
> 
> To handle these specific case, check if the SMEM is not initialized (by
> checking if the qcom_smem_get_soc_id returns -ENODEV) and fallback to
> OF machine compatible checking to identify the SoC variant.
> 
> Notice that the checking order is important as the machine compatible
> are normally defined with the specific one following the generic SoC
> (for example compatible = "qcom,ipq8065", "qcom,ipq8064").

...

> +		if (of_machine_is_compatible("qcom,ipq8062"))
> +			msm_id = QCOM_ID_IPQ8062;
> +		else if (of_machine_is_compatible("qcom,ipq8065") ||
> +			 of_machine_is_compatible("qcom,ipq8069"))
> +			msm_id = QCOM_ID_IPQ8065;
> +		else if (of_machine_is_compatible("qcom,ipq8064") ||
> +			 of_machine_is_compatible("qcom,ipq8066") ||
> +			 of_machine_is_compatible("qcom,ipq8068"))
> +			msm_id = QCOM_ID_IPQ8064;

A nit-pick (in case you need a new version of the series): I would expect
the conditionals be sorted by assigned value.

		if (of_machine_is_compatible("qcom,ipq8062"))
			msm_id = QCOM_ID_IPQ8062;
		else if (of_machine_is_compatible("qcom,ipq8064") ||
			 of_machine_is_compatible("qcom,ipq8066") ||
			 of_machine_is_compatible("qcom,ipq8068"))
			msm_id = QCOM_ID_IPQ8064;
		else if (of_machine_is_compatible("qcom,ipq8065") ||
			 of_machine_is_compatible("qcom,ipq8069"))
			msm_id = QCOM_ID_IPQ8065;

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 3/3] cpufreq: qcom-nvmem: add compatible fallback for ipq806x for no SMEM

[Christian Marangi]

On Fri, Oct 31, 2025 at 03:26:46PM +0200, Andy Shevchenko wrote:
> On Fri, Oct 31, 2025 at 02:08:34PM +0100, Christian Marangi wrote:
> > On some IPQ806x SoC SMEM might be not initialized by SBL. This is the
> > case for some Google devices (the OnHub family) that can't make use of
> > SMEM to detect the SoC ID.
> > 
> > To handle these specific case, check if the SMEM is not initialized (by
> > checking if the qcom_smem_get_soc_id returns -ENODEV) and fallback to
> > OF machine compatible checking to identify the SoC variant.
> > 
> > Notice that the checking order is important as the machine compatible
> > are normally defined with the specific one following the generic SoC
> > (for example compatible = "qcom,ipq8065", "qcom,ipq8064").
> 
> ...
> 
> > +		if (of_machine_is_compatible("qcom,ipq8062"))
> > +			msm_id = QCOM_ID_IPQ8062;
> > +		else if (of_machine_is_compatible("qcom,ipq8065") ||
> > +			 of_machine_is_compatible("qcom,ipq8069"))
> > +			msm_id = QCOM_ID_IPQ8065;
> > +		else if (of_machine_is_compatible("qcom,ipq8064") ||
> > +			 of_machine_is_compatible("qcom,ipq8066") ||
> > +			 of_machine_is_compatible("qcom,ipq8068"))
> > +			msm_id = QCOM_ID_IPQ8064;
> 
> A nit-pick (in case you need a new version of the series): I would expect
> the conditionals be sorted by assigned value.
> 
> 		if (of_machine_is_compatible("qcom,ipq8062"))
> 			msm_id = QCOM_ID_IPQ8062;
> 		else if (of_machine_is_compatible("qcom,ipq8064") ||
> 			 of_machine_is_compatible("qcom,ipq8066") ||
> 			 of_machine_is_compatible("qcom,ipq8068"))
> 			msm_id = QCOM_ID_IPQ8064;
> 		else if (of_machine_is_compatible("qcom,ipq8065") ||
> 			 of_machine_is_compatible("qcom,ipq8069"))
> 			msm_id = QCOM_ID_IPQ8065;
>

Hi as said in the commit, parsing 65/69 before 64 is needed as we might
have compatible like

"qcom,ipq8065","qcom,ipq8064" so we might incorrectly parse msm_id
ipq8064.

> -- 
> With Best Regards,
> Andy Shevchenko
> 
> 

-- 
	Ansuel

Re: [PATCH v2 3/3] cpufreq: qcom-nvmem: add compatible fallback for ipq806x for no SMEM

[Andy Shevchenko]

On Fri, Oct 31, 2025 at 03:19:12PM +0100, Christian Marangi wrote:
> On Fri, Oct 31, 2025 at 03:26:46PM +0200, Andy Shevchenko wrote:
> > On Fri, Oct 31, 2025 at 02:08:34PM +0100, Christian Marangi wrote:

...

> > > +		if (of_machine_is_compatible("qcom,ipq8062"))
> > > +			msm_id = QCOM_ID_IPQ8062;
> > > +		else if (of_machine_is_compatible("qcom,ipq8065") ||
> > > +			 of_machine_is_compatible("qcom,ipq8069"))
> > > +			msm_id = QCOM_ID_IPQ8065;
> > > +		else if (of_machine_is_compatible("qcom,ipq8064") ||
> > > +			 of_machine_is_compatible("qcom,ipq8066") ||
> > > +			 of_machine_is_compatible("qcom,ipq8068"))
> > > +			msm_id = QCOM_ID_IPQ8064;
> > 
> > A nit-pick (in case you need a new version of the series): I would expect
> > the conditionals be sorted by assigned value.
> > 
> > 		if (of_machine_is_compatible("qcom,ipq8062"))
> > 			msm_id = QCOM_ID_IPQ8062;
> > 		else if (of_machine_is_compatible("qcom,ipq8064") ||
> > 			 of_machine_is_compatible("qcom,ipq8066") ||
> > 			 of_machine_is_compatible("qcom,ipq8068"))
> > 			msm_id = QCOM_ID_IPQ8064;
> > 		else if (of_machine_is_compatible("qcom,ipq8065") ||
> > 			 of_machine_is_compatible("qcom,ipq8069"))
> > 			msm_id = QCOM_ID_IPQ8065;
> >
> 
> Hi as said in the commit, parsing 65/69 before 64 is needed as we might
> have compatible like
> 
> "qcom,ipq8065","qcom,ipq8064" so we might incorrectly parse msm_id
> ipq8064.

Oh, this is unfortunate. Wouldn't it be possible to use some API that returns
an index (or an error if not found) of the compatible? I believe we have a such
for the regular 'compatible' properties.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 3/3] cpufreq: qcom-nvmem: add compatible fallback for ipq806x for no SMEM

[Christian Marangi]

On Fri, Oct 31, 2025 at 04:34:22PM +0200, Andy Shevchenko wrote:
> On Fri, Oct 31, 2025 at 03:19:12PM +0100, Christian Marangi wrote:
> > On Fri, Oct 31, 2025 at 03:26:46PM +0200, Andy Shevchenko wrote:
> > > On Fri, Oct 31, 2025 at 02:08:34PM +0100, Christian Marangi wrote:
> 
> ...
> 
> > > > +		if (of_machine_is_compatible("qcom,ipq8062"))
> > > > +			msm_id = QCOM_ID_IPQ8062;
> > > > +		else if (of_machine_is_compatible("qcom,ipq8065") ||
> > > > +			 of_machine_is_compatible("qcom,ipq8069"))
> > > > +			msm_id = QCOM_ID_IPQ8065;
> > > > +		else if (of_machine_is_compatible("qcom,ipq8064") ||
> > > > +			 of_machine_is_compatible("qcom,ipq8066") ||
> > > > +			 of_machine_is_compatible("qcom,ipq8068"))
> > > > +			msm_id = QCOM_ID_IPQ8064;
> > > 
> > > A nit-pick (in case you need a new version of the series): I would expect
> > > the conditionals be sorted by assigned value.
> > > 
> > > 		if (of_machine_is_compatible("qcom,ipq8062"))
> > > 			msm_id = QCOM_ID_IPQ8062;
> > > 		else if (of_machine_is_compatible("qcom,ipq8064") ||
> > > 			 of_machine_is_compatible("qcom,ipq8066") ||
> > > 			 of_machine_is_compatible("qcom,ipq8068"))
> > > 			msm_id = QCOM_ID_IPQ8064;
> > > 		else if (of_machine_is_compatible("qcom,ipq8065") ||
> > > 			 of_machine_is_compatible("qcom,ipq8069"))
> > > 			msm_id = QCOM_ID_IPQ8065;
> > >
> > 
> > Hi as said in the commit, parsing 65/69 before 64 is needed as we might
> > have compatible like
> > 
> > "qcom,ipq8065","qcom,ipq8064" so we might incorrectly parse msm_id
> > ipq8064.
> 
> Oh, this is unfortunate. Wouldn't it be possible to use some API that returns
> an index (or an error if not found) of the compatible? I believe we have a such
> for the regular 'compatible' properties.
> 

Well also using something like checking for the virst compatible might
be problematic as real device have something like "netgear,r7800",
"qcom,ipq8065","qcom,ipq8064".

I will check if I can implement some alternative logic to have
consistent order.

-- 
	Ansuel

Re: [PATCH v2 3/3] cpufreq: qcom-nvmem: add compatible fallback for ipq806x for no SMEM

[Dmitry Baryshkov]

On Sat, Nov 01, 2025 at 12:50:09PM +0100, Christian Marangi wrote:
> On Fri, Oct 31, 2025 at 04:34:22PM +0200, Andy Shevchenko wrote:
> > On Fri, Oct 31, 2025 at 03:19:12PM +0100, Christian Marangi wrote:
> > > On Fri, Oct 31, 2025 at 03:26:46PM +0200, Andy Shevchenko wrote:
> > > > On Fri, Oct 31, 2025 at 02:08:34PM +0100, Christian Marangi wrote:
> > 
> > ...
> > 
> > > > > +		if (of_machine_is_compatible("qcom,ipq8062"))
> > > > > +			msm_id = QCOM_ID_IPQ8062;
> > > > > +		else if (of_machine_is_compatible("qcom,ipq8065") ||
> > > > > +			 of_machine_is_compatible("qcom,ipq8069"))
> > > > > +			msm_id = QCOM_ID_IPQ8065;
> > > > > +		else if (of_machine_is_compatible("qcom,ipq8064") ||
> > > > > +			 of_machine_is_compatible("qcom,ipq8066") ||
> > > > > +			 of_machine_is_compatible("qcom,ipq8068"))
> > > > > +			msm_id = QCOM_ID_IPQ8064;
> > > > 
> > > > A nit-pick (in case you need a new version of the series): I would expect
> > > > the conditionals be sorted by assigned value.
> > > > 
> > > > 		if (of_machine_is_compatible("qcom,ipq8062"))
> > > > 			msm_id = QCOM_ID_IPQ8062;
> > > > 		else if (of_machine_is_compatible("qcom,ipq8064") ||
> > > > 			 of_machine_is_compatible("qcom,ipq8066") ||
> > > > 			 of_machine_is_compatible("qcom,ipq8068"))
> > > > 			msm_id = QCOM_ID_IPQ8064;
> > > > 		else if (of_machine_is_compatible("qcom,ipq8065") ||
> > > > 			 of_machine_is_compatible("qcom,ipq8069"))
> > > > 			msm_id = QCOM_ID_IPQ8065;
> > > >
> > > 
> > > Hi as said in the commit, parsing 65/69 before 64 is needed as we might
> > > have compatible like
> > > 
> > > "qcom,ipq8065","qcom,ipq8064" so we might incorrectly parse msm_id
> > > ipq8064.
> > 
> > Oh, this is unfortunate. Wouldn't it be possible to use some API that returns
> > an index (or an error if not found) of the compatible? I believe we have a such
> > for the regular 'compatible' properties.
> > 
> 
> Well also using something like checking for the virst compatible might
> be problematic as real device have something like "netgear,r7800",
> "qcom,ipq8065","qcom,ipq8064".
> 
> I will check if I can implement some alternative logic to have
> consistent order.

See drivers/soc/qcom/qcom_pd_mapper.c. You can use of_match_node() on
the root device.

-- 
With best wishes
Dmitry

Re: [PATCH v2 3/3] cpufreq: qcom-nvmem: add compatible fallback for ipq806x for no SMEM

[Andy Shevchenko]

On Fri, Oct 31, 2025 at 02:08:34PM +0100, Christian Marangi wrote:
> On some IPQ806x SoC SMEM might be not initialized by SBL. This is the
> case for some Google devices (the OnHub family) that can't make use of
> SMEM to detect the SoC ID.
> 
> To handle these specific case, check if the SMEM is not initialized (by
> checking if the qcom_smem_get_soc_id returns -ENODEV) and fallback to
> OF machine compatible checking to identify the SoC variant.
> 
> Notice that the checking order is important as the machine compatible
> are normally defined with the specific one following the generic SoC
> (for example compatible = "qcom,ipq8065", "qcom,ipq8064").

FWIW,
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 3/3] cpufreq: qcom-nvmem: add compatible fallback for ipq806x for no SMEM

[Bjorn Andersson]

On Fri, Oct 31, 2025 at 02:08:34PM +0100, Christian Marangi wrote:
> On some IPQ806x SoC SMEM might be not initialized by SBL. This is the
> case for some Google devices (the OnHub family) that can't make use of
> SMEM to detect the SoC ID.
> 
> To handle these specific case, check if the SMEM is not initialized (by
> checking if the qcom_smem_get_soc_id returns -ENODEV) and fallback to
> OF machine compatible checking to identify the SoC variant.
> 
> Notice that the checking order is important as the machine compatible
> are normally defined with the specific one following the generic SoC
> (for example compatible = "qcom,ipq8065", "qcom,ipq8064").
> 
> Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>

Reviewed-by: Bjorn Andersson <andersson@kernel.org>

And as mentioned in v1, this (cpufreq) patch can be merged independently
of the first two patches. So please merge it through the cpufreq tree.

Regards,
Bjorn

> ---
>  drivers/cpufreq/qcom-cpufreq-nvmem.c | 17 +++++++++++++++--
>  1 file changed, 15 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/cpufreq/qcom-cpufreq-nvmem.c b/drivers/cpufreq/qcom-cpufreq-nvmem.c
> index 3a8ed723a23e..5a9bd780a4f3 100644
> --- a/drivers/cpufreq/qcom-cpufreq-nvmem.c
> +++ b/drivers/cpufreq/qcom-cpufreq-nvmem.c
> @@ -257,8 +257,8 @@ static int qcom_cpufreq_ipq8064_name_version(struct device *cpu_dev,
>  					     char **pvs_name,
>  					     struct qcom_cpufreq_drv *drv)
>  {
> +	int msm_id = -1, ret = 0;
>  	int speed = 0, pvs = 0;
> -	int msm_id, ret = 0;
>  	u8 *speedbin;
>  	size_t len;
>  
> @@ -275,8 +275,21 @@ static int qcom_cpufreq_ipq8064_name_version(struct device *cpu_dev,
>  	get_krait_bin_format_a(cpu_dev, &speed, &pvs, speedbin);
>  
>  	ret = qcom_smem_get_soc_id(&msm_id);
> -	if (ret)
> +	if (ret == -ENODEV) {
> +		/* Fallback to compatible match with no SMEM initialized */
> +		ret = 0;
> +		if (of_machine_is_compatible("qcom,ipq8062"))
> +			msm_id = QCOM_ID_IPQ8062;
> +		else if (of_machine_is_compatible("qcom,ipq8065") ||
> +			 of_machine_is_compatible("qcom,ipq8069"))
> +			msm_id = QCOM_ID_IPQ8065;
> +		else if (of_machine_is_compatible("qcom,ipq8064") ||
> +			 of_machine_is_compatible("qcom,ipq8066") ||
> +			 of_machine_is_compatible("qcom,ipq8068"))
> +			msm_id = QCOM_ID_IPQ8064;
> +	} else if (ret) {
>  		goto exit;
> +	}
>  
>  	switch (msm_id) {
>  	case QCOM_ID_IPQ8062:
> -- 
> 2.51.0
> 

Re: [PATCH v2 3/3] cpufreq: qcom-nvmem: add compatible fallback for ipq806x for no SMEM

[Christian Marangi]

On Sat, Nov 01, 2025 at 12:42:55PM -0500, Bjorn Andersson wrote:
> On Fri, Oct 31, 2025 at 02:08:34PM +0100, Christian Marangi wrote:
> > On some IPQ806x SoC SMEM might be not initialized by SBL. This is the
> > case for some Google devices (the OnHub family) that can't make use of
> > SMEM to detect the SoC ID.
> > 
> > To handle these specific case, check if the SMEM is not initialized (by
> > checking if the qcom_smem_get_soc_id returns -ENODEV) and fallback to
> > OF machine compatible checking to identify the SoC variant.
> > 
> > Notice that the checking order is important as the machine compatible
> > are normally defined with the specific one following the generic SoC
> > (for example compatible = "qcom,ipq8065", "qcom,ipq8064").
> > 
> > Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
> 
> Reviewed-by: Bjorn Andersson <andersson@kernel.org>
> 
> And as mentioned in v1, this (cpufreq) patch can be merged independently
> of the first two patches. So please merge it through the cpufreq tree.
> 

I will send a new revision just for this patch so I can use
of_match_node()

Should be ok since it hasn't been picked right?

> 
> > ---
> >  drivers/cpufreq/qcom-cpufreq-nvmem.c | 17 +++++++++++++++--
> >  1 file changed, 15 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/cpufreq/qcom-cpufreq-nvmem.c b/drivers/cpufreq/qcom-cpufreq-nvmem.c
> > index 3a8ed723a23e..5a9bd780a4f3 100644
> > --- a/drivers/cpufreq/qcom-cpufreq-nvmem.c
> > +++ b/drivers/cpufreq/qcom-cpufreq-nvmem.c
> > @@ -257,8 +257,8 @@ static int qcom_cpufreq_ipq8064_name_version(struct device *cpu_dev,
> >  					     char **pvs_name,
> >  					     struct qcom_cpufreq_drv *drv)
> >  {
> > +	int msm_id = -1, ret = 0;
> >  	int speed = 0, pvs = 0;
> > -	int msm_id, ret = 0;
> >  	u8 *speedbin;
> >  	size_t len;
> >  
> > @@ -275,8 +275,21 @@ static int qcom_cpufreq_ipq8064_name_version(struct device *cpu_dev,
> >  	get_krait_bin_format_a(cpu_dev, &speed, &pvs, speedbin);
> >  
> >  	ret = qcom_smem_get_soc_id(&msm_id);
> > -	if (ret)
> > +	if (ret == -ENODEV) {
> > +		/* Fallback to compatible match with no SMEM initialized */
> > +		ret = 0;
> > +		if (of_machine_is_compatible("qcom,ipq8062"))
> > +			msm_id = QCOM_ID_IPQ8062;
> > +		else if (of_machine_is_compatible("qcom,ipq8065") ||
> > +			 of_machine_is_compatible("qcom,ipq8069"))
> > +			msm_id = QCOM_ID_IPQ8065;
> > +		else if (of_machine_is_compatible("qcom,ipq8064") ||
> > +			 of_machine_is_compatible("qcom,ipq8066") ||
> > +			 of_machine_is_compatible("qcom,ipq8068"))
> > +			msm_id = QCOM_ID_IPQ8064;
> > +	} else if (ret) {
> >  		goto exit;
> > +	}
> >  
> >  	switch (msm_id) {
> >  	case QCOM_ID_IPQ8062:
> > -- 
> > 2.51.0
> > 

-- 
	Ansuel

Re: [PATCH v2 3/3] cpufreq: qcom-nvmem: add compatible fallback for ipq806x for no SMEM

[Konrad Dybcio]

On 11/4/25 12:59 PM, Christian Marangi wrote:
> On Sat, Nov 01, 2025 at 12:42:55PM -0500, Bjorn Andersson wrote:
>> On Fri, Oct 31, 2025 at 02:08:34PM +0100, Christian Marangi wrote:
>>> On some IPQ806x SoC SMEM might be not initialized by SBL. This is the
>>> case for some Google devices (the OnHub family) that can't make use of
>>> SMEM to detect the SoC ID.
>>>
>>> To handle these specific case, check if the SMEM is not initialized (by
>>> checking if the qcom_smem_get_soc_id returns -ENODEV) and fallback to
>>> OF machine compatible checking to identify the SoC variant.
>>>
>>> Notice that the checking order is important as the machine compatible
>>> are normally defined with the specific one following the generic SoC
>>> (for example compatible = "qcom,ipq8065", "qcom,ipq8064").
>>>
>>> Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
>>
>> Reviewed-by: Bjorn Andersson <andersson@kernel.org>
>>
>> And as mentioned in v1, this (cpufreq) patch can be merged independently
>> of the first two patches. So please merge it through the cpufreq tree.
>>
> 
> I will send a new revision just for this patch so I can use
> of_match_node()
> 
> Should be ok since it hasn't been picked right?

Yes, this is desired, even

Konrad

Re: (subset) [PATCH v2 0/3] cpufreq: qcom: handle ipq806x with no SMEM

[Bjorn Andersson]

On Fri, 31 Oct 2025 14:08:31 +0100, Christian Marangi wrote:
> This small series handle a small device family of ipq806x
> devices (Google OnHub) that doesn't have SMEM init.
> 
> We improve the SMEM driver and apply a workaround in
> the cpufreq driver.
> 
> Changes v2:
> - Rename error macro to INIT_ERR_PTR
> - Return -ENODEV from smem probe
> - Restructure if condition in cpufreq driver
> 
> [...]

Applied, thanks!

[1/3] err.h: add INIT_ERR_PTR macro
      commit: 652a86b24c5ac444afaf7625c9340d55aab7f105
[2/3] soc: qcom: smem: better track SMEM uninitialized state
      commit: 7a94d5f31b549e18f908cb669c59f066f45a21c7

Best regards,
-- 
Bjorn Andersson <andersson@kernel.org>