[2/2] cpufreq: ap806: add cpufreq driver for Armada 8K
diff mbox series

Message ID 20180924151249.32328-3-gregory.clement@bootlin.com
State Changes Requested, archived
Headers show
Series
  • Add cpufreq support for Armada 8K
Related show

Commit Message

Gregory CLEMENT Sept. 24, 2018, 3:12 p.m. UTC
Add cpufreq driver for Marvell AP-806 found on Aramda 8K.
The AP-806 has DFS (Dynamic Frequency Scaling) with coupled
clock domain for two clusters, so this driver will directly
use generic cpufreq-dt driver as backend.

Based on the work of Omri Itach <omrii@marvell.com>.

Signed-off-by: Gregory CLEMENT <gregory.clement@bootlin.com>
---
 drivers/cpufreq/Kconfig.arm         |  11 +++
 drivers/cpufreq/Makefile            |   1 +
 drivers/cpufreq/armada-8k-cpufreq.c | 148 ++++++++++++++++++++++++++++
 3 files changed, 160 insertions(+)
 create mode 100644 drivers/cpufreq/armada-8k-cpufreq.c

Comments

Viresh Kumar Oct. 4, 2018, 5:06 a.m. UTC | #1
On 24-09-18, 17:12, Gregory CLEMENT wrote:
> Add cpufreq driver for Marvell AP-806 found on Aramda 8K.
> The AP-806 has DFS (Dynamic Frequency Scaling) with coupled
> clock domain for two clusters, so this driver will directly
> use generic cpufreq-dt driver as backend.
> 
> Based on the work of Omri Itach <omrii@marvell.com>.
> 
> Signed-off-by: Gregory CLEMENT <gregory.clement@bootlin.com>
> ---
>  drivers/cpufreq/Kconfig.arm         |  11 +++
>  drivers/cpufreq/Makefile            |   1 +
>  drivers/cpufreq/armada-8k-cpufreq.c | 148 ++++++++++++++++++++++++++++
>  3 files changed, 160 insertions(+)
>  create mode 100644 drivers/cpufreq/armada-8k-cpufreq.c
> 
> diff --git a/drivers/cpufreq/Kconfig.arm b/drivers/cpufreq/Kconfig.arm
> index 0cd8eb76ad59..fd34095743a7 100644
> --- a/drivers/cpufreq/Kconfig.arm
> +++ b/drivers/cpufreq/Kconfig.arm
> @@ -25,6 +25,17 @@ config ARM_ARMADA_37XX_CPUFREQ
>  	  This adds the CPUFreq driver support for Marvell Armada 37xx SoCs.
>  	  The Armada 37xx PMU supports 4 frequency and VDD levels.
>  
> +config ARM_ARMADA_8K_CPUFREQ
> +	tristate "Armada 8K CPUFreq driver"

Since you wanted this to be tristate, have you tried

insmod armada-8k-cpufreq.ko
rmmod armada-8k-cpufreq.ko
insmod armada-8k-cpufreq.ko

?

That will show some failures I believe.

> +	depends on ARCH_MVEBU && CPUFREQ_DT
> +	help
> +	  This enables the CPUFreq driver support for Marvell
> +	  Armada8k SOCs.
> +	  Armada8K device has the AP806 which supports scaling
> +	  to any full integer divider.
> +
> +	  If in doubt, say N.
> +
>  # big LITTLE core layer and glue drivers
>  config ARM_BIG_LITTLE_CPUFREQ
>  	tristate "Generic ARM big LITTLE CPUfreq driver"
> diff --git a/drivers/cpufreq/Makefile b/drivers/cpufreq/Makefile
> index c1ffeabe4ecf..21f4f56229db 100644
> --- a/drivers/cpufreq/Makefile
> +++ b/drivers/cpufreq/Makefile
> @@ -53,6 +53,7 @@ obj-$(CONFIG_ARM_BIG_LITTLE_CPUFREQ)	+= arm_big_little.o
>  obj-$(CONFIG_ARM_DT_BL_CPUFREQ)		+= arm_big_little_dt.o
>  
>  obj-$(CONFIG_ARM_ARMADA_37XX_CPUFREQ)	+= armada-37xx-cpufreq.o
> +obj-$(CONFIG_ARM_ARMADA_8K_CPUFREQ)	+= armada-8k-cpufreq.o

Maybe add another tab before + to align with other lines ?

>  obj-$(CONFIG_ARM_BRCMSTB_AVS_CPUFREQ)	+= brcmstb-avs-cpufreq.o
>  obj-$(CONFIG_ACPI_CPPC_CPUFREQ)		+= cppc_cpufreq.o
>  obj-$(CONFIG_ARCH_DAVINCI)		+= davinci-cpufreq.o
> diff --git a/drivers/cpufreq/armada-8k-cpufreq.c b/drivers/cpufreq/armada-8k-cpufreq.c
> new file mode 100644
> index 000000000000..1c604f1d27d2
> --- /dev/null
> +++ b/drivers/cpufreq/armada-8k-cpufreq.c
> @@ -0,0 +1,148 @@
> +// SPDX-License-Identifier: GPL-2.0+
> +/*
> + * CPUFreq support for Armada 7K/8K

7K ?

> + *
> + * Copyright (C) 2018 Marvell
> + *
> + * Omri Itach <omrii@marvell.com>
> + * Gregory Clement <gregory.clement@bootlin.com>
> + */
> +
> +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
> +
> +#include <linux/clk.h>
> +#include <linux/cpu.h>
> +#include <linux/err.h>
> +#include <linux/init.h>
> +#include <linux/kernel.h>
> +#include <linux/module.h>
> +#include <linux/of.h>
> +#include <linux/platform_device.h>
> +#include <linux/pm_opp.h>
> +#include <linux/slab.h>
> +
> +#define MIN_FREQ 100000000
> +
> +/*
> + * Setup the opps list with the divider for the max frequency, that
> + * will be filled at runtime, 0 meaning 100Mhz

Add a full-stop at the end please.

I am not sure why will you want to support 100 MHz as the CPUs may end
up running on that a lot :)

> + */
> +static const int opps_div[] __initconst = {1, 2, 3, 0};
> +
> +struct opps_array {

Maybe name this freq_table.

> +	struct device *cpu_dev;
> +	unsigned int freq[ARRAY_SIZE(opps_div)];
> +};
> +
> +/* If the CPUs share the same clock, then they are in the same cluster */
> +static void __init armada_8k_get_sharing_cpus(struct clk *cur_clk,
> +				       struct cpumask *cpumask)

Maybe align better to calm down checkpatch --strict :)

> +{
> +	int cpu;
> +
> +	for_each_possible_cpu(cpu) {
> +		struct device *cpu_dev = get_cpu_device(cpu);
> +		struct clk *clk = clk_get(cpu_dev, 0);

Don't you need to do a corresponding clk_put() ?

> +
> +		if (IS_ERR(clk))
> +			dev_warn(cpu_dev, "Cannot get clock for CPU %d\n", cpu);

Why continue after this point and check the next if ? Maybe write this
as if/else.

> +
> +		if (clk_is_match(clk, cur_clk))
> +			cpumask_set_cpu(cpu, cpumask);
> +	}
> +

Remove blank line ?

> +}
> +
> +static int __init armada_8k_cpufreq_init(void)
> +{
> +	struct opps_array *opps_arrays;
> +	struct platform_device *pdev;
> +	int ret, cpu, opps_index = 0;
> +	unsigned int cur_frequency;
> +	struct device_node *node;
> +
> +	node = of_find_compatible_node(NULL, NULL, "marvell,ap806-cpu-clock");
> +	if (!node || !of_device_is_available(node))
> +		return -ENODEV;
> +
> +	opps_arrays = kcalloc(num_possible_cpus(), sizeof(*opps_arrays),
> +			      GFP_KERNEL);
> +	/*
> +	 * For each CPU, this loop registers the operating points
> +	 * supported (which are the nominal CPU frequency and full integer
> +	 * divisions of it).
> +	 */
> +	for_each_possible_cpu(cpu) {
> +		struct device *cpu_dev;
> +		struct cpumask cpus;
> +		unsigned int freq;
> +		struct clk *clk;
> +		int i;
> +
> +		cpu_dev = get_cpu_device(cpu);
> +		if (!cpu_dev) {
> +			dev_err(cpu_dev, "Cannot get CPU %d\n", cpu);
> +			continue;
> +		}
> +
> +		clk = clk_get(cpu_dev, 0);
> +		if (IS_ERR(clk)) {
> +			dev_err(cpu_dev, "Cannot get clock for CPU %d\n", cpu);
> +			ret = PTR_ERR(clk);
> +			goto free_opp_array;

What if this is the second iteration of this loop. Who will remove the
OPPs added during earlier iteration ?

> +		}
> +
> +		/* Get nominal (current) CPU frequency */
> +		cur_frequency = clk_get_rate(clk);
> +		if (!cur_frequency) {
> +			dev_err(cpu_dev,
> +				"Failed to get clock rate for CPU %d\n", cpu);
> +			ret = -EINVAL;

clk_put() ?

> +			goto free_opp_array;
> +		}
> +
> +		opps_arrays[opps_index].cpu_dev = cpu_dev;

I will add a blank line here.

> +		for (i = 0; i < ARRAY_SIZE(opps_div); i++) {
> +			if (opps_div[i])
> +				freq = cur_frequency / opps_div[i];
> +			else
> +				freq = MIN_FREQ;
> +
> +			ret = dev_pm_opp_add(cpu_dev, freq, 0);
> +			if (ret)
> +				goto remove_opp;
> +
> +			opps_arrays[opps_index].freq[i] = freq;
> +		}
> +
> +		cpumask_clear(&cpus);
> +		armada_8k_get_sharing_cpus(clk, &cpus);
> +		dev_pm_opp_set_sharing_cpus(cpu_dev, &cpus);

Don't you get any error or print messages like duplicate-OPPs ? Maybe
they are dev_dbg() :(

You have already done above for a group of CPUs and you shouldn't be
adding the OPPs again for the same group for an iteration that belongs
to the next CPU.

So if you have CPU 0-3 sharing the clock line, then you will add 4
OPPs in the first iteration and then try the same for 3 more times for
the CPU 1,2,3. You should really avoid it.

And generally the error handling should improve here.

> +	}
> +
> +	pdev = platform_device_register_simple("cpufreq-dt", -1, NULL, 0);
> +	ret = PTR_ERR_OR_ZERO(pdev);
> +	if (ret)
> +		goto remove_opp;
> +	kfree(opps_arrays);
> +	return 0;
> +
> +remove_opp:
> +	for (; opps_index >= 0; opps_index--) {
> +		int i = 0;
> +
> +		while (opps_arrays[opps_index].freq[i]) {
> +			dev_pm_opp_remove(opps_arrays[opps_index].cpu_dev,
> +					  opps_arrays[opps_index].freq[i]);
> +			i++;
> +		}
> +	}
> +free_opp_array:
> +	kfree(opps_arrays);
> +	return ret;
> +}
> +module_init(armada_8k_cpufreq_init);
> +
> +MODULE_AUTHOR("Gregory Clement <gregory.clement@bootlin.com>");
> +MODULE_DESCRIPTION("Armada 8K cpufreq driver");
> +MODULE_LICENSE("GPL");
> -- 
> 2.19.0
Gregory CLEMENT Nov. 15, 2018, 10:36 a.m. UTC | #2
Hi Viresh,


Sorry for the delay.

 On jeu., oct. 04 2018, Viresh Kumar <viresh.kumar@linaro.org> wrote:

> On 24-09-18, 17:12, Gregory CLEMENT wrote:
>> Add cpufreq driver for Marvell AP-806 found on Aramda 8K.
>> The AP-806 has DFS (Dynamic Frequency Scaling) with coupled
>> clock domain for two clusters, so this driver will directly
>> use generic cpufreq-dt driver as backend.
>> 
>> Based on the work of Omri Itach <omrii@marvell.com>.
>> 
>> Signed-off-by: Gregory CLEMENT <gregory.clement@bootlin.com>
>> ---
>>  drivers/cpufreq/Kconfig.arm         |  11 +++
>>  drivers/cpufreq/Makefile            |   1 +
>>  drivers/cpufreq/armada-8k-cpufreq.c | 148 ++++++++++++++++++++++++++++
>>  3 files changed, 160 insertions(+)
>>  create mode 100644 drivers/cpufreq/armada-8k-cpufreq.c
>> 
>> diff --git a/drivers/cpufreq/Kconfig.arm b/drivers/cpufreq/Kconfig.arm
>> index 0cd8eb76ad59..fd34095743a7 100644
>> --- a/drivers/cpufreq/Kconfig.arm
>> +++ b/drivers/cpufreq/Kconfig.arm
>> @@ -25,6 +25,17 @@ config ARM_ARMADA_37XX_CPUFREQ
>>  	  This adds the CPUFreq driver support for Marvell Armada 37xx SoCs.
>>  	  The Armada 37xx PMU supports 4 frequency and VDD levels.
>>  
>> +config ARM_ARMADA_8K_CPUFREQ
>> +	tristate "Armada 8K CPUFreq driver"
>
> Since you wanted this to be tristate, have you tried
>
> insmod armada-8k-cpufreq.ko
> rmmod armada-8k-cpufreq.ko
> insmod armada-8k-cpufreq.ko
> ?

Actually the purpose of it is to be able to have the driver available as
a module. Thanks to this the size kernel image is not increased. Once
the system is running then we can load the module and have the feature
but it is true that is no more possible to unload it. However, except
for developing or debugging, I don't see the real interest to unload
the driver.

>
> That will show some failures I believe.
>
>> +	depends on ARCH_MVEBU && CPUFREQ_DT
>> +	help
>> +	  This enables the CPUFreq driver support for Marvell
>> +	  Armada8k SOCs.
>> +	  Armada8K device has the AP806 which supports scaling
>> +	  to any full integer divider.
>> +
>> +	  If in doubt, say N.
>> +
>>  # big LITTLE core layer and glue drivers
>>  config ARM_BIG_LITTLE_CPUFREQ
>>  	tristate "Generic ARM big LITTLE CPUfreq driver"
>> diff --git a/drivers/cpufreq/Makefile b/drivers/cpufreq/Makefile
>> index c1ffeabe4ecf..21f4f56229db 100644
>> --- a/drivers/cpufreq/Makefile
>> +++ b/drivers/cpufreq/Makefile
>> @@ -53,6 +53,7 @@ obj-$(CONFIG_ARM_BIG_LITTLE_CPUFREQ)	+= arm_big_little.o
>>  obj-$(CONFIG_ARM_DT_BL_CPUFREQ)		+= arm_big_little_dt.o
>>  
>>  obj-$(CONFIG_ARM_ARMADA_37XX_CPUFREQ)	+= armada-37xx-cpufreq.o
>> +obj-$(CONFIG_ARM_ARMADA_8K_CPUFREQ)	+= armada-8k-cpufreq.o
>
> Maybe add another tab before + to align with other lines ?

I think the issue is introduced by the way the diff patch show it,
because in my file they are aligned and there is onl on tab before +. It
make sense because there is less than 8 character between ')' and '+'
on the first line.

>
>>  obj-$(CONFIG_ARM_BRCMSTB_AVS_CPUFREQ)	+= brcmstb-avs-cpufreq.o
>>  obj-$(CONFIG_ACPI_CPPC_CPUFREQ)		+= cppc_cpufreq.o
>>  obj-$(CONFIG_ARCH_DAVINCI)		+= davinci-cpufreq.o
>> diff --git a/drivers/cpufreq/armada-8k-cpufreq.c b/drivers/cpufreq/armada-8k-cpufreq.c
>> new file mode 100644
>> index 000000000000..1c604f1d27d2
>> --- /dev/null
>> +++ b/drivers/cpufreq/armada-8k-cpufreq.c
>> @@ -0,0 +1,148 @@
>> +// SPDX-License-Identifier: GPL-2.0+
>> +/*
>> + * CPUFreq support for Armada 7K/8K
>
> 7K ?

Right, for now only 8K is supported so I will fix it.

>
>> + *
>> + * Copyright (C) 2018 Marvell
>> + *
>> + * Omri Itach <omrii@marvell.com>
>> + * Gregory Clement <gregory.clement@bootlin.com>
>> + */
>> +
>> +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
>> +
>> +#include <linux/clk.h>
>> +#include <linux/cpu.h>
>> +#include <linux/err.h>
>> +#include <linux/init.h>
>> +#include <linux/kernel.h>
>> +#include <linux/module.h>
>> +#include <linux/of.h>
>> +#include <linux/platform_device.h>
>> +#include <linux/pm_opp.h>
>> +#include <linux/slab.h>
>> +
>> +#define MIN_FREQ 100000000
>> +
>> +/*
>> + * Setup the opps list with the divider for the max frequency, that
>> + * will be filled at runtime, 0 meaning 100Mhz
>
> Add a full-stop at the end please.

I am not sure to understand what you means by a "full-stop": a 0Hz
frequency?

But this case should be handle by the CPU idle driver.

> I am not sure why will you want to support 100 MHz as the CPUs may end
> up running on that a lot :)

It should depend of the policy, or am I missing something.


>
>> + */
>> +static const int opps_div[] __initconst = {1, 2, 3, 0};
>> +
>> +struct opps_array {
>
> Maybe name this freq_table.

OK

>
>> +	struct device *cpu_dev;
>> +	unsigned int freq[ARRAY_SIZE(opps_div)];
>> +};
>> +
>> +/* If the CPUs share the same clock, then they are in the same cluster */
>> +static void __init armada_8k_get_sharing_cpus(struct clk *cur_clk,
>> +				       struct cpumask *cpumask)
>
> Maybe align better to calm down checkpatch --strict :)

I use checkpatch --strict so I wonder how I missed it, but I will fix
it.

>
>> +{
>> +	int cpu;
>> +
>> +	for_each_possible_cpu(cpu) {
>> +		struct device *cpu_dev = get_cpu_device(cpu);
>> +		struct clk *clk = clk_get(cpu_dev, 0);
>
> Don't you need to do a corresponding clk_put() ?

Indeed

>
>> +
>> +		if (IS_ERR(clk))
>> +			dev_warn(cpu_dev, "Cannot get clock for CPU %d\n", cpu);
>
> Why continue after this point and check the next if ? Maybe write this
> as if/else.

Yes I can do this


>
>> +
>> +		if (clk_is_match(clk, cur_clk))
>> +			cpumask_set_cpu(cpu, cpumask);
>> +	}
>> +
>
> Remove blank line ?
>
OK (and it was also reported by checkpacth!)


>> +}
>> +
>> +static int __init armada_8k_cpufreq_init(void)
>> +{
>> +	struct opps_array *opps_arrays;
>> +	struct platform_device *pdev;
>> +	int ret, cpu, opps_index = 0;
>> +	unsigned int cur_frequency;
>> +	struct device_node *node;
>> +
>> +	node = of_find_compatible_node(NULL, NULL, "marvell,ap806-cpu-clock");
>> +	if (!node || !of_device_is_available(node))
>> +		return -ENODEV;
>> +
>> +	opps_arrays = kcalloc(num_possible_cpus(), sizeof(*opps_arrays),
>> +			      GFP_KERNEL);
>> +	/*
>> +	 * For each CPU, this loop registers the operating points
>> +	 * supported (which are the nominal CPU frequency and full integer
>> +	 * divisions of it).
>> +	 */
>> +	for_each_possible_cpu(cpu) {
>> +		struct device *cpu_dev;
>> +		struct cpumask cpus;
>> +		unsigned int freq;
>> +		struct clk *clk;
>> +		int i;
>> +
>> +		cpu_dev = get_cpu_device(cpu);
>> +		if (!cpu_dev) {
>> +			dev_err(cpu_dev, "Cannot get CPU %d\n", cpu);
>> +			continue;
>> +		}
>> +
>> +		clk = clk_get(cpu_dev, 0);
>> +		if (IS_ERR(clk)) {
>> +			dev_err(cpu_dev, "Cannot get clock for CPU %d\n", cpu);
>> +			ret = PTR_ERR(clk);
>> +			goto free_opp_array;
>
> What if this is the second iteration of this loop. Who will remove the
> OPPs added during earlier iteration ?
>
>> +		}
>> +
>> +		/* Get nominal (current) CPU frequency */
>> +		cur_frequency = clk_get_rate(clk);
>> +		if (!cur_frequency) {
>> +			dev_err(cpu_dev,
>> +				"Failed to get clock rate for CPU %d\n", cpu);
>> +			ret = -EINVAL;
>
> clk_put() ?
>
>> +			goto free_opp_array;
>> +		}
>> +
>> +		opps_arrays[opps_index].cpu_dev = cpu_dev;
>
> I will add a blank line here.
>
>> +		for (i = 0; i < ARRAY_SIZE(opps_div); i++) {
>> +			if (opps_div[i])
>> +				freq = cur_frequency / opps_div[i];
>> +			else
>> +				freq = MIN_FREQ;
>> +
>> +			ret = dev_pm_opp_add(cpu_dev, freq, 0);
>> +			if (ret)
>> +				goto remove_opp;
>> +
>> +			opps_arrays[opps_index].freq[i] = freq;
>> +		}
>> +
>> +		cpumask_clear(&cpus);
>> +		armada_8k_get_sharing_cpus(clk, &cpus);
>> +		dev_pm_opp_set_sharing_cpus(cpu_dev, &cpus);
>
> Don't you get any error or print messages like duplicate-OPPs ? Maybe
> they are dev_dbg() :(
>
> You have already done above for a group of CPUs and you shouldn't be
> adding the OPPs again for the same group for an iteration that belongs
> to the next CPU.
>
> So if you have CPU 0-3 sharing the clock line, then you will add 4
> OPPs in the first iteration and then try the same for 3 more times for
> the CPU 1,2,3. You should really avoid it.
>
> And generally the error handling should improve here.

I am going to see how to improve this function.

Thanks,

Gregory

>
>> +	}
>> +
>> +	pdev = platform_device_register_simple("cpufreq-dt", -1, NULL, 0);
>> +	ret = PTR_ERR_OR_ZERO(pdev);
>> +	if (ret)
>> +		goto remove_opp;
>> +	kfree(opps_arrays);
>> +	return 0;
>> +
>> +remove_opp:
>> +	for (; opps_index >= 0; opps_index--) {
>> +		int i = 0;
>> +
>> +		while (opps_arrays[opps_index].freq[i]) {
>> +			dev_pm_opp_remove(opps_arrays[opps_index].cpu_dev,
>> +					  opps_arrays[opps_index].freq[i]);
>> +			i++;
>> +		}
>> +	}
>> +free_opp_array:
>> +	kfree(opps_arrays);
>> +	return ret;
>> +}
>> +module_init(armada_8k_cpufreq_init);
>> +
>> +MODULE_AUTHOR("Gregory Clement <gregory.clement@bootlin.com>");
>> +MODULE_DESCRIPTION("Armada 8K cpufreq driver");
>> +MODULE_LICENSE("GPL");
>> -- 
>> 2.19.0
>
> -- 
> viresh
Viresh Kumar Nov. 16, 2018, 4:30 a.m. UTC | #3
On 15-11-18, 11:36, Gregory CLEMENT wrote:
>  On jeu., oct. 04 2018, Viresh Kumar <viresh.kumar@linaro.org> wrote:
> 
> > On 24-09-18, 17:12, Gregory CLEMENT wrote:
> >> Add cpufreq driver for Marvell AP-806 found on Aramda 8K.
> >> The AP-806 has DFS (Dynamic Frequency Scaling) with coupled
> >> clock domain for two clusters, so this driver will directly
> >> use generic cpufreq-dt driver as backend.
> >> 
> >> Based on the work of Omri Itach <omrii@marvell.com>.
> >> 
> >> Signed-off-by: Gregory CLEMENT <gregory.clement@bootlin.com>
> >> ---
> >>  drivers/cpufreq/Kconfig.arm         |  11 +++
> >>  drivers/cpufreq/Makefile            |   1 +
> >>  drivers/cpufreq/armada-8k-cpufreq.c | 148 ++++++++++++++++++++++++++++
> >>  3 files changed, 160 insertions(+)
> >>  create mode 100644 drivers/cpufreq/armada-8k-cpufreq.c
> >> 
> >> diff --git a/drivers/cpufreq/Kconfig.arm b/drivers/cpufreq/Kconfig.arm
> >> index 0cd8eb76ad59..fd34095743a7 100644
> >> --- a/drivers/cpufreq/Kconfig.arm
> >> +++ b/drivers/cpufreq/Kconfig.arm
> >> @@ -25,6 +25,17 @@ config ARM_ARMADA_37XX_CPUFREQ
> >>  	  This adds the CPUFreq driver support for Marvell Armada 37xx SoCs.
> >>  	  The Armada 37xx PMU supports 4 frequency and VDD levels.
> >>  
> >> +config ARM_ARMADA_8K_CPUFREQ
> >> +	tristate "Armada 8K CPUFreq driver"
> >
> > Since you wanted this to be tristate, have you tried
> >
> > insmod armada-8k-cpufreq.ko
> > rmmod armada-8k-cpufreq.ko
> > insmod armada-8k-cpufreq.ko
> > ?
> 
> Actually the purpose of it is to be able to have the driver available as
> a module. Thanks to this the size kernel image is not increased. Once
> the system is running then we can load the module and have the feature
> but it is true that is no more possible to unload it. However, except
> for developing or debugging, I don't see the real interest to unload
> the driver.

Sure, probably it will never get used that way but I see that as
broken. If you want to have the driver as module, better support
unloading of it as well.

> >> +++ b/drivers/cpufreq/armada-8k-cpufreq.c

> >> +/*
> >> + * Setup the opps list with the divider for the max frequency, that
> >> + * will be filled at runtime, 0 meaning 100Mhz
> >
> > Add a full-stop at the end please.
> 
> I am not sure to understand what you means by a "full-stop": a 0Hz
> frequency?
> 
> But this case should be handle by the CPU idle driver.

I meant '.' by full-stop. I wanted you to add . at the end of the
line, like this:

Setup the opps list with the divider for the max frequency, that
will be filled at runtime, 0 meaning 100Mhz.

> > I am not sure why will you want to support 100 MHz as the CPUs may end
> > up running on that a lot :)
> 
> It should depend of the policy, or am I missing something.

Sure, but with any of the general governors like
ondemand/conservative/schedutil, you will end up running at 100 MHz
when the system doesn't have high load. And 100MHz may be way too
slow.

Patch
diff mbox series

diff --git a/drivers/cpufreq/Kconfig.arm b/drivers/cpufreq/Kconfig.arm
index 0cd8eb76ad59..fd34095743a7 100644
--- a/drivers/cpufreq/Kconfig.arm
+++ b/drivers/cpufreq/Kconfig.arm
@@ -25,6 +25,17 @@  config ARM_ARMADA_37XX_CPUFREQ
 	  This adds the CPUFreq driver support for Marvell Armada 37xx SoCs.
 	  The Armada 37xx PMU supports 4 frequency and VDD levels.
 
+config ARM_ARMADA_8K_CPUFREQ
+	tristate "Armada 8K CPUFreq driver"
+	depends on ARCH_MVEBU && CPUFREQ_DT
+	help
+	  This enables the CPUFreq driver support for Marvell
+	  Armada8k SOCs.
+	  Armada8K device has the AP806 which supports scaling
+	  to any full integer divider.
+
+	  If in doubt, say N.
+
 # big LITTLE core layer and glue drivers
 config ARM_BIG_LITTLE_CPUFREQ
 	tristate "Generic ARM big LITTLE CPUfreq driver"
diff --git a/drivers/cpufreq/Makefile b/drivers/cpufreq/Makefile
index c1ffeabe4ecf..21f4f56229db 100644
--- a/drivers/cpufreq/Makefile
+++ b/drivers/cpufreq/Makefile
@@ -53,6 +53,7 @@  obj-$(CONFIG_ARM_BIG_LITTLE_CPUFREQ)	+= arm_big_little.o
 obj-$(CONFIG_ARM_DT_BL_CPUFREQ)		+= arm_big_little_dt.o
 
 obj-$(CONFIG_ARM_ARMADA_37XX_CPUFREQ)	+= armada-37xx-cpufreq.o
+obj-$(CONFIG_ARM_ARMADA_8K_CPUFREQ)	+= armada-8k-cpufreq.o
 obj-$(CONFIG_ARM_BRCMSTB_AVS_CPUFREQ)	+= brcmstb-avs-cpufreq.o
 obj-$(CONFIG_ACPI_CPPC_CPUFREQ)		+= cppc_cpufreq.o
 obj-$(CONFIG_ARCH_DAVINCI)		+= davinci-cpufreq.o
diff --git a/drivers/cpufreq/armada-8k-cpufreq.c b/drivers/cpufreq/armada-8k-cpufreq.c
new file mode 100644
index 000000000000..1c604f1d27d2
--- /dev/null
+++ b/drivers/cpufreq/armada-8k-cpufreq.c
@@ -0,0 +1,148 @@ 
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * CPUFreq support for Armada 7K/8K
+ *
+ * Copyright (C) 2018 Marvell
+ *
+ * Omri Itach <omrii@marvell.com>
+ * Gregory Clement <gregory.clement@bootlin.com>
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/clk.h>
+#include <linux/cpu.h>
+#include <linux/err.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/pm_opp.h>
+#include <linux/slab.h>
+
+#define MIN_FREQ 100000000
+
+/*
+ * Setup the opps list with the divider for the max frequency, that
+ * will be filled at runtime, 0 meaning 100Mhz
+ */
+static const int opps_div[] __initconst = {1, 2, 3, 0};
+
+struct opps_array {
+	struct device *cpu_dev;
+	unsigned int freq[ARRAY_SIZE(opps_div)];
+};
+
+/* If the CPUs share the same clock, then they are in the same cluster */
+static void __init armada_8k_get_sharing_cpus(struct clk *cur_clk,
+				       struct cpumask *cpumask)
+{
+	int cpu;
+
+	for_each_possible_cpu(cpu) {
+		struct device *cpu_dev = get_cpu_device(cpu);
+		struct clk *clk = clk_get(cpu_dev, 0);
+
+		if (IS_ERR(clk))
+			dev_warn(cpu_dev, "Cannot get clock for CPU %d\n", cpu);
+
+		if (clk_is_match(clk, cur_clk))
+			cpumask_set_cpu(cpu, cpumask);
+	}
+
+}
+
+static int __init armada_8k_cpufreq_init(void)
+{
+	struct opps_array *opps_arrays;
+	struct platform_device *pdev;
+	int ret, cpu, opps_index = 0;
+	unsigned int cur_frequency;
+	struct device_node *node;
+
+	node = of_find_compatible_node(NULL, NULL, "marvell,ap806-cpu-clock");
+	if (!node || !of_device_is_available(node))
+		return -ENODEV;
+
+	opps_arrays = kcalloc(num_possible_cpus(), sizeof(*opps_arrays),
+			      GFP_KERNEL);
+	/*
+	 * For each CPU, this loop registers the operating points
+	 * supported (which are the nominal CPU frequency and full integer
+	 * divisions of it).
+	 */
+	for_each_possible_cpu(cpu) {
+		struct device *cpu_dev;
+		struct cpumask cpus;
+		unsigned int freq;
+		struct clk *clk;
+		int i;
+
+		cpu_dev = get_cpu_device(cpu);
+		if (!cpu_dev) {
+			dev_err(cpu_dev, "Cannot get CPU %d\n", cpu);
+			continue;
+		}
+
+		clk = clk_get(cpu_dev, 0);
+		if (IS_ERR(clk)) {
+			dev_err(cpu_dev, "Cannot get clock for CPU %d\n", cpu);
+			ret = PTR_ERR(clk);
+			goto free_opp_array;
+		}
+
+		/* Get nominal (current) CPU frequency */
+		cur_frequency = clk_get_rate(clk);
+		if (!cur_frequency) {
+			dev_err(cpu_dev,
+				"Failed to get clock rate for CPU %d\n", cpu);
+			ret = -EINVAL;
+			goto free_opp_array;
+		}
+
+		opps_arrays[opps_index].cpu_dev = cpu_dev;
+		for (i = 0; i < ARRAY_SIZE(opps_div); i++) {
+			if (opps_div[i])
+				freq = cur_frequency / opps_div[i];
+			else
+				freq = MIN_FREQ;
+
+			ret = dev_pm_opp_add(cpu_dev, freq, 0);
+			if (ret)
+				goto remove_opp;
+
+			opps_arrays[opps_index].freq[i] = freq;
+		}
+
+		cpumask_clear(&cpus);
+		armada_8k_get_sharing_cpus(clk, &cpus);
+		dev_pm_opp_set_sharing_cpus(cpu_dev, &cpus);
+	}
+
+	pdev = platform_device_register_simple("cpufreq-dt", -1, NULL, 0);
+	ret = PTR_ERR_OR_ZERO(pdev);
+	if (ret)
+		goto remove_opp;
+	kfree(opps_arrays);
+	return 0;
+
+remove_opp:
+	for (; opps_index >= 0; opps_index--) {
+		int i = 0;
+
+		while (opps_arrays[opps_index].freq[i]) {
+			dev_pm_opp_remove(opps_arrays[opps_index].cpu_dev,
+					  opps_arrays[opps_index].freq[i]);
+			i++;
+		}
+	}
+free_opp_array:
+	kfree(opps_arrays);
+	return ret;
+}
+module_init(armada_8k_cpufreq_init);
+
+MODULE_AUTHOR("Gregory Clement <gregory.clement@bootlin.com>");
+MODULE_DESCRIPTION("Armada 8K cpufreq driver");
+MODULE_LICENSE("GPL");