Skip to contentSkip to conversation

ARM: vexpress: Set-up shared OPP table instead of individual for each CPU

PATCH In review Sudeep Holla 2 participants 5 messages Nov 28, 2019 – Dec 2, 2019

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 ARM: vexpress: Set-up shared OPP table instead of individual for each CPU

Currently we add individual copy of same OPP table for each CPU within the cluster. This is redundant and doesn't reflect the reality.

We can't use core cpumask to set policy->cpus in ve_spc_cpufreq_init() anymore as it gets called via cpuhp_cpufreq_online()->cpufreq_online() ->cpufreq_driver->init() and the cpumask gets updated upon CPU hotplug operations. It also may cause issues when the vexpress_spc_cpufreq driver is built as a module.

Since ve_spc_clk_init is built-in device initcall, we should be able to use the same topology_core_cpumask to set the opp sharing cpumask via dev_pm_opp_set_sharing_cpus and use the same later in the driver via dev_pm_opp_get_sharing_cpus.

---
 arch/arm/mach-vexpress/spc.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
arch/arm/mach-vexpress/spc.c +11 -1
index 354e0e7025ae..1da11bdb1dfb 100644
@@ -551,8 +551,9 @@ static struct clk *ve_spc_clk_register(struct device *cpu_dev)
 
 static int __init ve_spc_clk_init(void)
 {
-	int cpu;
+	int cpu, cluster;
 	struct clk *clk;
+	bool init_opp_table[MAX_CLUSTERS] = { false };
 
 	if (!info)
 		return 0; /* Continue only if SPC is initialised */
@@ -578,8 +579,17 @@ static int __init ve_spc_clk_init(void)
 			continue;
 		}
 
+		cluster = topology_physical_package_id(cpu_dev->id);
+		if (init_opp_table[cluster])
+			continue;
+
 		if (ve_init_opp_table(cpu_dev))
 			pr_warn("failed to initialise cpu%d opp table\n", cpu);
+		else if (dev_pm_opp_set_sharing_cpus(cpu_dev,
+			 topology_core_cpumask(cpu_dev->id)))
+			pr_warn("failed to mark OPPs shared for cpu%d\n", cpu);
+		else
+			init_opp_table[cluster] = true;
 	}
 
 	platform_device_register_simple("vexpress-spc-cpufreq", -1, NULL, 0);
Signature · 1 line
2.17.1
#1 lore raw

PATCH 2/2 cpufreq: vexpress-spc: Switch cpumask from topology core to OPP sharing

Since commit ca74b316df96 ("arm: Use common cpu_topology structure and functions.") the core cpumask has to be modified during cpu hotplug operations. So using them to set up cpufreq policy cpumask may be incorrect as it may contain only cpus that are online at that instance.

Instead, we can use the cpumask setup by OPP library that contains all the cpus sharing OPP table using dev_pm_opp_get_sharing_cpus.

---
 drivers/cpufreq/vexpress-spc-cpufreq.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/cpufreq/vexpress-spc-cpufreq.c +1 -1
index 506e3f2bf53a..83c85d3d67e3 100644
@@ -434,7 +434,7 @@ static int ve_spc_cpufreq_init(struct cpufreq_policy *policy)
 	if (cur_cluster < MAX_CLUSTERS) {
 		int cpu;
 
-		cpumask_copy(policy->cpus, topology_core_cpumask(policy->cpu));
+		dev_pm_opp_get_sharing_cpus(cpu_dev, policy->cpus);
 
 		for_each_cpu(cpu, policy->cpus)
 			per_cpu(physical_cluster, cpu) = cur_cluster;
Signature · 1 line
2.17.1
#2 lore raw

Re: [PATCH 2/2] cpufreq: vexpress-spc: Switch cpumask from topology core to OPP sharing

Hi Viresh, Rafael,

Quoted 9 lines · Sudeep Holla

Since commit ca74b316df96 ("arm: Use common cpu_topology structure and functions.") the core cpumask has to be modified during cpu hotplug operations. So using them to set up cpufreq policy cpumask may be incorrect as it may contain only cpus that are online at that instance.

Instead, we can use the cpumask setup by OPP library that contains all the cpus sharing OPP table using dev_pm_opp_get_sharing_cpus.

Cc Viresh Kumar <[email protected]>

This can go independently via PM tree and I can take 1/2 via SoC tree (as the file is being moved). The problem will be fixed only after both lands, but this alone won't break the build. Or if you guys provide Ack, I can take both together via ARM SoC team. Let me know.

--
Regards,
Sudeep

maintainer Acked-by
#3 lore raw

Re: [PATCH 2/2] cpufreq: vexpress-spc: Switch cpumask from topology core to OPP sharing

Quoted 8 lines · Sudeep Holla

Hi Viresh, Rafael,

Quoted 5 lines · Sudeep Holla

Since commit ca74b316df96 ("arm: Use common cpu_topology structure and functions.") the core cpumask has to be modified during cpu hotplug operations. So using them to set up cpufreq policy cpumask may be incorrect as it may contain only cpus that are online at that instance.

Quoted 8 lines
Quoted 4 lines

Instead, we can use the cpumask setup by OPP library that contains all the cpus sharing OPP table using dev_pm_opp_get_sharing_cpus.

Cc Viresh Kumar <[email protected]>

This can go independently via PM tree and I can take 1/2 via SoC tree (as the file is being moved). The problem will be fixed only after both lands, but this alone won't break the build.

Yes, but it will break cpufreq for sure as shared-cpus won't be set by anyone.

Quoted 2 lines

Or if you guys provide
Ack, I can take both together via ARM SoC team. Let me know.

I was planning to take them through cpufreq tree, but that fine if you can do that.

Signature · 1 line
viresh
#4 lore raw

Re: [PATCH 2/2] cpufreq: vexpress-spc: Switch cpumask from topology core to OPP sharing

Quoted 12 lines · Viresh Kumar
Quoted 11 lines · Sudeep Holla

Hi Viresh, Rafael,

Quoted 8 lines · Sudeep Holla

Since commit ca74b316df96 ("arm: Use common cpu_topology structure and functions.") the core cpumask has to be modified during cpu hotplug operations. So using them to set up cpufreq policy cpumask may be incorrect as it may contain only cpus that are online at that instance.

Instead, we can use the cpumask setup by OPP library that contains all the cpus sharing OPP table using dev_pm_opp_get_sharing_cpus.

Quoted 8 lines
Quoted 5 lines
Quoted 1 line
Cc Viresh Kumar <[email protected]>

This can go independently via PM tree and I can take 1/2 via SoC tree (as the file is being moved). The problem will be fixed only after both lands, but this alone won't break the build.

Yes, but it will break cpufreq for sure as shared-cpus won't be set by anyone.

It's already broke on hotplug :) but yes works on boot at-least.

Quoted 6 lines
Quoted 2 lines

Or if you guys provide
Ack, I can take both together via ARM SoC team. Let me know.

I was planning to take them through cpufreq tree, but that fine if you can do that.

Arnd moved and compressed few folders in arch/arm/mach-*, and vexpress was merged into versatile. Not sure if it's planned for v5.5, so to be cautious, I preferred to take it via ARM SoC.

Quoted 1 line

Thanks, I will take them together via ARM SoC

--
Regards,
Sudeep