Skip to contentSkip to conversation

ARM: cpuidle for 4.13

GIT PULL In review Daniel Lezcano 4 participants 10 messages Jun 12, 2017 – Jun 22, 2017 CPU idle time management framework

lore
Share & tools
lore
#0 lore raw

Hi Rafael,

this is a pull request for a fix already and the asymmetric idle state support for HMP systems. It is based on V4.12-rc5.

- Make per cpuidle driver for the ARM cpuidle driver in order to support SMP, HMP and future DynamIQ cpus (Daniel Lezcano)

- Release of node on error (Christophe Jaillet)

Thanks

-- Daniel

The following changes since commit 32c1431eea4881a6b17bd7c639315010aeefa452:

Linux 4.12-rc5 (2017-06-11 16:48:20 -0700)

are available in the git repository at:

http://[email protected]/people/daniel.lezcano/linux.git cpuidle/4.13

for you to fetch changes up to 610c28be6046786ccec22c827481fd865d4db8e0:

ARM: cpuidle: Support asymmetric idle definition (2017-06-12 14:57:21 +0200)

----------------------------------------------------------------
Christophe Jaillet (1):
cpuidle: dt: Add missing 'of_node_put()'

Daniel Lezcano (1):
ARM: cpuidle: Support asymmetric idle definition

 drivers/cpuidle/Kconfig.arm      |  1 +
 drivers/cpuidle/cpuidle-arm.c    | 62 ++++++++++++++++++++++++++++++++++++++------------------------
 drivers/cpuidle/dt_idle_states.c |  4 +++-
 3 files changed, 42 insertions(+), 25 deletions(-)
Signature · 5 lines
 <http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog
#1 lore raw

PATCH 1/2 cpuidle: dt: Add missing 'of_node_put()'

From: Christophe Jaillet <[email protected]>

'of_node_put()' should be called on pointer returned by 'of_parse_phandle()' when done. In this function this is done in all path except this 'continue', so add it.

Fixes 97735da074fd ("drivers: cpuidle: Add status property to ARM idle states")
---
 drivers/cpuidle/dt_idle_states.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/cpuidle/dt_idle_states.c +3 -1
index ffca4fc..ae8eb03 100644
@@ -180,8 +180,10 @@ int dt_init_idle_driver(struct cpuidle_driver *drv,
 		if (!state_node)
 			break;
 
-		if (!of_device_is_available(state_node))
+		if (!of_device_is_available(state_node)) {
+			of_node_put(state_node);
 			continue;
+		}
 
 		if (!idle_state_valid(state_node, i, cpumask)) {
 			pr_warn("%s idle state not valid, bailing out\n",
Signature · 1 line
2.7.4
#2 lore raw

PATCH 2/2 ARM: cpuidle: Support asymmetric idle definition

Some hardware have clusters with different idle states. The current code does not support this and fails as it expects all the idle states to be identical.

Because of this, the Mediatek mtk8173 had to create the same idle state for a big.Little system and now the Hisilicon 960 is facing the same situation.

Solve this by simply assuming the multiple driver will be needed for all the platforms using the ARM generic cpuidle driver which makes sense because of the different topologies we can support with a single kernel for ARM32 or ARM64.

Every CPU has its own driver, so every single CPU can specify in the DT the idle states.

This simple approach allows to support the future dynamIQ system, current SMP and HMP.

Tested on:
- 96boards: Hikey 620
- 96boards: Hikey 960
- 96boards: dragonboard410c
- Mediatek 8173

---
 drivers/cpuidle/Kconfig.arm   |  1 +
 drivers/cpuidle/cpuidle-arm.c | 62 ++++++++++++++++++++++++++-----------------
 2 files changed, 39 insertions(+), 24 deletions(-)
drivers/cpuidle/Kconfig.arm +1 -0
index 21340e0..f521448 100644
@@ -4,6 +4,7 @@
 config ARM_CPUIDLE
         bool "Generic ARM/ARM64 CPU idle Driver"
         select DT_IDLE_STATES
+	select CPU_IDLE_MULTIPLE_DRIVERS
         help
           Select this to enable generic cpuidle driver for ARM.
           It provides a generic idle driver whose idle states are configured
drivers/cpuidle/cpuidle-arm.c +38 -24

91 lines not on this page · raw message

Signature · 1 line
2.7.4
Acked-by on v1 1/2
#8 lore raw

Re: [PATCH 1/2] cpuidle: dt: Add missing 'of_node_put()'

In reply to Daniel Lezcano
Quoted 9 lines · Daniel Lezcano

From: Christophe Jaillet <[email protected]>

'of_node_put()' should be called on pointer returned by 'of_parse_phandle()' when done. In this function this is done in all path except this 'continue', so add it.

Fixes 97735da074fd ("drivers: cpuidle: Add status property to ARM idle states")
Quoted 23 lines
---
 drivers/cpuidle/dt_idle_states.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/cpuidle/dt_idle_states.c +3 -1
index ffca4fc..ae8eb03 100644
@@ -180,8 +180,10 @@ int dt_init_idle_driver(struct cpuidle_driver *drv,
 		if (!state_node)
 			break;
 
-		if (!of_device_is_available(state_node))
+		if (!of_device_is_available(state_node)) {
+			of_node_put(state_node);
 			continue;
+		}
 
 		if (!idle_state_valid(state_node, i, cpumask)) {
 			pr_warn("%s idle state not valid, bailing out\n",
Signature · 1 line
2.7.4
#3 lore raw

Re: [PATCH 2/2] ARM: cpuidle: Support asymmetric idle definition

In reply to Daniel Lezcano
Quoted 22 lines · Daniel Lezcano

Some hardware have clusters with different idle states. The current code does not support this and fails as it expects all the idle states to be identical.

Because of this, the Mediatek mtk8173 had to create the same idle state for a big.Little system and now the Hisilicon 960 is facing the same situation.

Solve this by simply assuming the multiple driver will be needed for all the platforms using the ARM generic cpuidle driver which makes sense because of the different topologies we can support with a single kernel for ARM32 or ARM64.

Every CPU has its own driver, so every single CPU can specify in the DT the idle states.

This simple approach allows to support the future dynamIQ system, current SMP and HMP.

Tested on:
- 96boards: Hikey 620
- 96boards: Hikey 960
- 96boards: dragonboard410c
- Mediatek 8173

Quoted 4 lines

There seems to have been quite some discussion regarding this one and I'm not sure about the resolution of it.

I'd feel more comfortable with an ACK or Reviewed-by from Sudeep or Lorenzo here.

Quoted 107 lines
---
 drivers/cpuidle/Kconfig.arm   |  1 +
 drivers/cpuidle/cpuidle-arm.c | 62 ++++++++++++++++++++++++++-----------------
 2 files changed, 39 insertions(+), 24 deletions(-)
drivers/cpuidle/Kconfig.arm +1 -0
index 21340e0..f521448 100644
@@ -4,6 +4,7 @@
 config ARM_CPUIDLE
         bool "Generic ARM/ARM64 CPU idle Driver"
         select DT_IDLE_STATES
+	select CPU_IDLE_MULTIPLE_DRIVERS
         help
           Select this to enable generic cpuidle driver for ARM.
           It provides a generic idle driver whose idle states are configured
drivers/cpuidle/cpuidle-arm.c +35 -22
index f440d38..7080c38 100644
@@ -18,6 +18,7 @@
 #include <linux/module.h>
 #include <linux/of.h>
 #include <linux/slab.h>
+#include <linux/topology.h>
 
 #include <asm/cpuidle.h>
 
@@ -44,7 +45,7 @@ static int arm_enter_idle_state(struct cpuidle_device *dev,
 	return CPU_PM_CPU_IDLE_ENTER(arm_cpuidle_suspend, idx);
 }
 
-static struct cpuidle_driver arm_idle_driver = {
+static struct cpuidle_driver arm_idle_driver __initdata = {
 	.name = "arm_idle",
 	.owner = THIS_MODULE,
 	/*
@@ -80,30 +81,42 @@ static const struct of_device_id arm_idle_state_match[] __initconst = {
 static int __init arm_idle_init(void)
 {
 	int cpu, ret;
-	struct cpuidle_driver *drv = &arm_idle_driver;
+	struct cpuidle_driver *drv;
 	struct cpuidle_device *dev;
 
-	/*
-	 * Initialize idle states data, starting at index 1.
-	 * This driver is DT only, if no DT idle states are detected (ret == 0)
-	 * let the driver initialization fail accordingly since there is no
-	 * reason to initialize the idle driver if only wfi is supported.
-	 */
-	ret = dt_init_idle_driver(drv, arm_idle_state_match, 1);
-	if (ret <= 0)
-		return ret ? : -ENODEV;
-
-	ret = cpuidle_register_driver(drv);
-	if (ret) {
-		pr_err("Failed to register cpuidle driver\n");
-		return ret;
-	}
-
-	/*
-	 * Call arch CPU operations in order to initialize
-	 * idle states suspend back-end specific data
-	 */
 	for_each_possible_cpu(cpu) {
+
+		drv = kmemdup(&arm_idle_driver, sizeof(*drv), GFP_KERNEL);
+		if (!drv) {
+			ret = -ENOMEM;
+			goto out_fail;
+		}
+
+		drv->cpumask = (struct cpumask *)cpumask_of(cpu);
+
+		/*
+		 * Initialize idle states data, starting at index 1.  This
+		 * driver is DT only, if no DT idle states are detected (ret
+		 * == 0) let the driver initialization fail accordingly since
+		 * there is no reason to initialize the idle driver if only
+		 * wfi is supported.
+		 */
+		ret = dt_init_idle_driver(drv, arm_idle_state_match, 1);
+		if (ret <= 0) {
+			ret = ret ? : -ENODEV;
+			goto out_fail;
+		}
+
+		ret = cpuidle_register_driver(drv);
+		if (ret) {
+			pr_err("Failed to register cpuidle driver\n");
+			goto out_fail;
+		}
+
+		/*
+		 * Call arch CPU operations in order to initialize
+		 * idle states suspend back-end specific data
+		 */
 		ret = arm_cpuidle_init(cpu);
 
 		/*
@@ -141,10 +154,11 @@ static int __init arm_idle_init(void)
 		dev = per_cpu(cpuidle_devices, cpu);
 		cpuidle_unregister_device(dev);
 		kfree(dev);

drivers/cpuidle/cpuidle-arm.c static int __init arm_idle_init(void) · line 164

Quoted 10 lines
drivers/cpuidle/cpuidle-arm.c +3 -2
+		drv = cpuidle_get_driver();
+		cpuidle_unregister_driver(drv);
+		kfree(drv);
 	}
 
-	cpuidle_unregister_driver(drv);
-
 	return ret;
 }
 device_initcall(arm_idle_init);

Thanks,
Rafael

#9 lore raw

Re: [PATCH 1/2] cpuidle: dt: Add missing 'of_node_put()'

In reply to Daniel Lezcano
Quoted 9 lines · Daniel Lezcano

From: Christophe Jaillet <[email protected]>

'of_node_put()' should be called on pointer returned by 'of_parse_phandle()' when done. In this function this is done in all path except this 'continue', so add it.

Fixes 97735da074fd ("drivers: cpuidle: Add status property to ARM idle states")

I've already applied this one.

Thanks,
Rafael

#4 lore raw

Re: [PATCH 2/2] ARM: cpuidle: Support asymmetric idle definition

In reply to Rafael J. Wysocki
Quoted 28 lines · Rafael J. Wysocki
Quoted 22 lines · Daniel Lezcano

Some hardware have clusters with different idle states. The current code does not support this and fails as it expects all the idle states to be identical.

Because of this, the Mediatek mtk8173 had to create the same idle state for a big.Little system and now the Hisilicon 960 is facing the same situation.

Solve this by simply assuming the multiple driver will be needed for all the platforms using the ARM generic cpuidle driver which makes sense because of the different topologies we can support with a single kernel for ARM32 or ARM64.

Every CPU has its own driver, so every single CPU can specify in the DT the idle states.

This simple approach allows to support the future dynamIQ system, current SMP and HMP.

Tested on:
- 96boards: Hikey 620
- 96boards: Hikey 960
- 96boards: dragonboard410c
- Mediatek 8173

Quoted 4 lines
Quoted 4 lines

There seems to have been quite some discussion regarding this one and I'm not sure about the resolution of it.

I'd feel more comfortable with an ACK or Reviewed-by from Sudeep or Lorenzo here.

I understand.

Sudeep it is ok with the patch [1] without an explicit acked-by.

-- Daniel

[1] https://www.spinics.net/lists/kernel/msg2525980.html

Acked-by on v1 2/2
#7 lore raw

Re: [PATCH 2/2] ARM: cpuidle: Support asymmetric idle definition

In reply to Daniel Lezcano

Hi Daniel,

Quoted 16 lines · Daniel Lezcano

Some hardware have clusters with different idle states. The current code does not support this and fails as it expects all the idle states to be identical.

Because of this, the Mediatek mtk8173 had to create the same idle state for a big.Little system and now the Hisilicon 960 is facing the same situation.

Solve this by simply assuming the multiple driver will be needed for all the platforms using the ARM generic cpuidle driver which makes sense because of the different topologies we can support with a single kernel for ARM32 or ARM64.

Every CPU has its own driver, so every single CPU can specify in the DT the idle states.

This simple approach allows to support the future dynamIQ system, current SMP and HMP.

Quoted 7 lines

Tested on:
- 96boards: Hikey 620
- 96boards: Hikey 960
- 96boards: dragonboard410c
- Mediatek 8173

Cc Sudeep Holla <[email protected]>

Sorry for the delay, as I mentioned earlier I would like to add the minimum change to avoid this on platforms that don't require this. But that can be done later, I will try to come up with simple solution when I get time. Though I am not 100% happy ;), I am fine with this change for now:

Signature · 2 lines
Regards,
Sudeep
#5 lore raw

Re: [PATCH 2/2] ARM: cpuidle: Support asymmetric idle definition

In reply to Rafael J. Wysocki
Quoted 28 lines · Rafael J. Wysocki
Quoted 22 lines · Daniel Lezcano

Some hardware have clusters with different idle states. The current code does not support this and fails as it expects all the idle states to be identical.

Because of this, the Mediatek mtk8173 had to create the same idle state for a big.Little system and now the Hisilicon 960 is facing the same situation.

Solve this by simply assuming the multiple driver will be needed for all the platforms using the ARM generic cpuidle driver which makes sense because of the different topologies we can support with a single kernel for ARM32 or ARM64.

Every CPU has its own driver, so every single CPU can specify in the DT the idle states.

This simple approach allows to support the future dynamIQ system, current SMP and HMP.

Tested on:
- 96boards: Hikey 620
- 96boards: Hikey 960
- 96boards: dragonboard410c
- Mediatek 8173

Quoted 4 lines
Quoted 4 lines

There seems to have been quite some discussion regarding this one and I'm not sure about the resolution of it.

I'd feel more comfortable with an ACK or Reviewed-by from Sudeep or Lorenzo here.

Hi Rafael,

just a gentle reminder, Sudeep acked the patch.

Thanks.

-- Daniel

Signature · 5 lines
 <http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog
#6 lore raw

Re: [PATCH 2/2] ARM: cpuidle: Support asymmetric idle definition

Quoted 32 lines · Daniel Lezcano
Quoted 31 lines · Rafael J. Wysocki
Quoted 22 lines · Daniel Lezcano

Some hardware have clusters with different idle states. The current code does not support this and fails as it expects all the idle states to be identical.

Because of this, the Mediatek mtk8173 had to create the same idle state for a big.Little system and now the Hisilicon 960 is facing the same situation.

Solve this by simply assuming the multiple driver will be needed for all the platforms using the ARM generic cpuidle driver which makes sense because of the different topologies we can support with a single kernel for ARM32 or ARM64.

Every CPU has its own driver, so every single CPU can specify in the DT the idle states.

This simple approach allows to support the future dynamIQ system, current SMP and HMP.

Tested on:
- 96boards: Hikey 620
- 96boards: Hikey 960
- 96boards: dragonboard410c
- Mediatek 8173

Quoted 4 lines

There seems to have been quite some discussion regarding this one and I'm not sure about the resolution of it.

Quoted 6 lines
Quoted 1 line

I'd feel more comfortable with an ACK or Reviewed-by from Sudeep or Lorenzo here.

Hi Rafael,

just a gentle reminder, Sudeep acked the patch.

Yes, I'll get to it later today, most likely.

Thanks,
Rafael