[PATCH v2] cpufreq: scmi: Account for malformed DT in scmi_dev_used_by_cpus()

[PATCH v2] cpufreq: scmi: Account for malformed DT in scmi_dev_used_by_cpus()

[Florian Fainelli]

Broadcom STB platforms were early adopters of the SCMI framework and as
a result, not all deployed systems have a Device Tree entry where SCMI
protocol 0x13 (PERFORMANCE) is declared as a clock provider, nor are the
CPU Device Tree node(s) referencing protocol 0x13 as their clock
provider.

For those platforms, we allow the checks done by scmi_dev_used_by_cpus()
to continue, and in the event of not having done an early return, we key
off the documented compatible string and give them a pass to continue to
use scmi-cpufreq.

Fixes: 6c9bb8692272 ("cpufreq: scmi: Skip SCMI devices that aren't used by the CPUs")
Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
---
Changes in v2:

- do not leverage the SCMI quirks framework which in spirit is for
  dealing with SCMI firmware implementations, this right here,
  specifically pertains to an older Device Tree. This also makes it
  easier to backport that change into stable trees where the commit in
  the Fixes tag is already present. There is no dependency upon the
  presence of the SCMI quirks framework

 drivers/cpufreq/scmi-cpufreq.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/cpufreq/scmi-cpufreq.c b/drivers/cpufreq/scmi-cpufreq.c
index ef078426bfd5..9df4d968fdac 100644
--- a/drivers/cpufreq/scmi-cpufreq.c
+++ b/drivers/cpufreq/scmi-cpufreq.c
@@ -15,6 +15,7 @@
 #include <linux/energy_model.h>
 #include <linux/export.h>
 #include <linux/module.h>
+#include <linux/of.h>
 #include <linux/pm_opp.h>
 #include <linux/pm_qos.h>
 #include <linux/slab.h>
@@ -424,6 +425,14 @@ static bool scmi_dev_used_by_cpus(struct device *scmi_dev)
 			return true;
 	}
 
+	/* Older Broadcom STB chips had a "clocks" property for CPU node(s)
+	 * that did not match the SCMI performance protocol node, if we got
+	 * there, it means we had such an older Device Tree, therefore return
+	 * true to preserve backwards compatibility.
+	 */
+	if (of_machine_is_compatible("brcm,brcmstb"))
+		return true;
+
 	return false;
 }
 
-- 
2.34.1

Re: [PATCH v2] cpufreq: scmi: Account for malformed DT in scmi_dev_used_by_cpus()

[Viresh Kumar]

On 15-08-25, 13:57, Florian Fainelli wrote:
> Broadcom STB platforms were early adopters of the SCMI framework and as
> a result, not all deployed systems have a Device Tree entry where SCMI
> protocol 0x13 (PERFORMANCE) is declared as a clock provider, nor are the
> CPU Device Tree node(s) referencing protocol 0x13 as their clock
> provider.
> 
> For those platforms, we allow the checks done by scmi_dev_used_by_cpus()
> to continue, and in the event of not having done an early return, we key
> off the documented compatible string and give them a pass to continue to
> use scmi-cpufreq.
> 
> Fixes: 6c9bb8692272 ("cpufreq: scmi: Skip SCMI devices that aren't used by the CPUs")
> Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
> ---
> Changes in v2:
> 
> - do not leverage the SCMI quirks framework which in spirit is for
>   dealing with SCMI firmware implementations, this right here,
>   specifically pertains to an older Device Tree. This also makes it
>   easier to backport that change into stable trees where the commit in
>   the Fixes tag is already present. There is no dependency upon the
>   presence of the SCMI quirks framework
> 
>  drivers/cpufreq/scmi-cpufreq.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/drivers/cpufreq/scmi-cpufreq.c b/drivers/cpufreq/scmi-cpufreq.c
> index ef078426bfd5..9df4d968fdac 100644
> --- a/drivers/cpufreq/scmi-cpufreq.c
> +++ b/drivers/cpufreq/scmi-cpufreq.c
> @@ -15,6 +15,7 @@
>  #include <linux/energy_model.h>
>  #include <linux/export.h>
>  #include <linux/module.h>
> +#include <linux/of.h>
>  #include <linux/pm_opp.h>
>  #include <linux/pm_qos.h>
>  #include <linux/slab.h>
> @@ -424,6 +425,14 @@ static bool scmi_dev_used_by_cpus(struct device *scmi_dev)
>  			return true;
>  	}
>  
> +	/* Older Broadcom STB chips had a "clocks" property for CPU node(s)

Incorrect multi-line comment format.

> +	 * that did not match the SCMI performance protocol node, if we got
> +	 * there, it means we had such an older Device Tree, therefore return
> +	 * true to preserve backwards compatibility.
> +	 */

-- 
viresh

Re: [PATCH v2] cpufreq: scmi: Account for malformed DT in scmi_dev_used_by_cpus()

[Sudeep Holla]

On Fri, Aug 15, 2025 at 01:57:14PM -0700, Florian Fainelli wrote:
> Broadcom STB platforms were early adopters of the SCMI framework and as
> a result, not all deployed systems have a Device Tree entry where SCMI
> protocol 0x13 (PERFORMANCE) is declared as a clock provider, nor are the
> CPU Device Tree node(s) referencing protocol 0x13 as their clock
> provider.
> 
> For those platforms, we allow the checks done by scmi_dev_used_by_cpus()
> to continue, and in the event of not having done an early return, we key
> off the documented compatible string and give them a pass to continue to
> use scmi-cpufreq.
> 

With the multi-line comment fixed as suggested by Viresh.

Reviewed-by: Sudeep Holla <sudeep.holla@arm.com>

-- 
Regards,
Sudeep