Skip to contentSkip to conversation

ARM: s3c64xx: cpuidle: convert to platform driver

PATCH New Daniel Lezcano 2 participants 9 messages Oct 25, 2013 – Oct 30, 2013

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 ARM: s3c64xx: cpuidle: convert to platform driver

The driver is tied with the pm low level code making difficult to split the driver into a more arch independent code. The platform driver allows to move the standby callback into the platform data field and use a simple driver with no more dependency on the low level code.

The standby callback has a portion of code to set the standby method and the effective cpu_do_idle switching the cpu to the right mode. As this code is redundant in the cpu suspend code, it has been factored out when implementing the standby methdod.

By this way, the driver is ready to be moved out to the drivers/cpuidle.

---
 arch/arm/mach-s3c64xx/cpuidle.c |   38 ++++++++++++++++----------------------
 arch/arm/mach-s3c64xx/pm.c      |   33 ++++++++++++++++++++++++++-------
 2 files changed, 42 insertions(+), 29 deletions(-)
arch/arm/mach-s3c64xx/cpuidle.c +16 -22
index 3c8ab07..8022f5f 100644
@@ -9,33 +9,16 @@
  * published by the Free Software Foundation.
 */
 
-#include <linux/kernel.h>
-#include <linux/init.h>
 #include <linux/cpuidle.h>
-#include <linux/io.h>
-#include <linux/export.h>
-#include <linux/time.h>
+#include <linux/platform_device.h>
 
-#include <asm/proc-fns.h>
-
-#include <mach/map.h>
-
-#include "regs-sys.h"
-#include "regs-syscon-power.h"
+static void (*s3c64xx_standby)(void);
 
 static int s3c64xx_enter_idle(struct cpuidle_device *dev,
 			      struct cpuidle_driver *drv,
 			      int index)
 {
-	unsigned long tmp;
-
-	/* Setup PWRCFG to enter idle mode */
-	tmp = __raw_readl(S3C64XX_PWR_CFG);
-	tmp &= ~S3C64XX_PWRCFG_CFG_WFI_MASK;
-	tmp |= S3C64XX_PWRCFG_CFG_WFI_IDLE;
-	__raw_writel(tmp, S3C64XX_PWR_CFG);
-
-	cpu_do_idle();
+	s3c64xx_standby();
 
 	return index;
 }
@@ -56,8 +39,19 @@ static struct cpuidle_driver s3c64xx_cpuidle_driver = {
 	.state_count = 1,
 };
 
-static int __init s3c64xx_init_cpuidle(void)
+static int s3c64xx_cpuidle_probe(struct platform_device *dev)
 {
+	s3c64xx_standby =  (void *)(dev->dev.platform_data);
+
 	return cpuidle_register(&s3c64xx_cpuidle_driver, NULL);
 }
-device_initcall(s3c64xx_init_cpuidle);
+
+static struct platform_driver s3c64xx_driver_cpuidle = {
+	.driver = {
+		.name = "cpuidle-s3c64xx",
+		.owner = THIS_MODULE,
+	},
+	.probe = s3c64xx_cpuidle_probe,
+};
+
+module_platform_driver(s3c64xx_driver_cpuidle);
arch/arm/mach-s3c64xx/pm.c +26 -7
index 6a1f91f..534fb4e 100644
@@ -18,6 +18,7 @@
 #include <linux/io.h>
 #include <linux/gpio.h>
 #include <linux/pm_domain.h>
+#include <linux/platform_device.h>
 
 #include <mach/map.h>
 #include <mach/irqs.h>
@@ -259,6 +260,22 @@ void s3c_pm_save_core(void)
 	s3c_pm_do_save(core_save, ARRAY_SIZE(core_save));
 }
 
+static void s3c64xx_set_standby(void)
+{
+	unsigned long tmp;
+
+	tmp = __raw_readl(S3C64XX_PWR_CFG);
+	tmp &= ~S3C64XX_PWRCFG_CFG_WFI_MASK;
+	tmp |= S3C64XX_PWRCFG_CFG_WFI_SLEEP;
+	__raw_writel(tmp, S3C64XX_PWR_CFG);
+}
+
+static void s3c64xx_standby(void)
+{
+	s3c64xx_set_standby();
+	cpu_do_idle();
+}
+
 /* since both s3c6400 and s3c6410 share the same sleep pm calls, we
  * put the per-cpu code in here until any new cpu comes along and changes
  * this.
@@ -269,11 +286,7 @@ static int s3c64xx_cpu_suspend(unsigned long arg)
 	unsigned long tmp;
 
 	/* set our standby method to sleep */
-
-	tmp = __raw_readl(S3C64XX_PWR_CFG);
-	tmp &= ~S3C64XX_PWRCFG_CFG_WFI_MASK;
-	tmp |= S3C64XX_PWRCFG_CFG_WFI_SLEEP;
-	__raw_writel(tmp, S3C64XX_PWR_CFG);
+	s3c64xx_set_standby();
 
 	/* clear any old wakeup */
 
@@ -348,6 +361,13 @@ int __init s3c64xx_pm_init(void)
 	return 0;
 }
 
+static struct platform_device s3c64xx_cpuidle_device = {
+	.name = "cpuidle-s3c64xx",
+	.dev = {
+		.platform_data = s3c64xx_standby,
+	},
+};
+
 static __init int s3c64xx_pm_initcall(void)
 {
 	pm_cpu_prep = s3c64xx_pm_prepare;
@@ -364,8 +384,7 @@ static __init int s3c64xx_pm_initcall(void)
 	gpio_direction_output(S3C64XX_GPN(14), 0);
 	gpio_direction_output(S3C64XX_GPN(15), 0);
 #endif
-
-	return 0;
+	return platform_device_register(&s3c64xx_cpuidle_device);
 }
 arch_initcall(s3c64xx_pm_initcall);
 
Signature · 1 line
1.7.9.5
#1 lore raw

PATCH 2/2 ARM: s3c64xx: cpuidle: move driver to drivers/cpuidle

There is no more dependency with the arch dependant code, thanks to the platform driver conversion.

Move the driver to the drivers/cpuidle like the other drivers.

---
 arch/arm/mach-s3c64xx/Makefile                     |    1 -
 drivers/cpuidle/Kconfig.arm                        |    7 +++++++
 drivers/cpuidle/Makefile                           |    1 +
 .../cpuidle.c => drivers/cpuidle/cpuidle-s3c64xx.c |    0
 4 files changed, 8 insertions(+), 1 deletion(-)
 rename arch/arm/mach-s3c64xx/cpuidle.c => drivers/cpuidle/cpuidle-s3c64xx.c (100%)
arch/arm/mach-s3c64xx/Makefile +0 -1
index 31d0c91..e7a3ee7 100644
@@ -22,7 +22,6 @@ obj-$(CONFIG_CPU_S3C6410)	+= s3c6410.o
 # PM
 
 obj-$(CONFIG_PM)		+= pm.o irq-pm.o sleep.o
-obj-$(CONFIG_CPU_IDLE)		+= cpuidle.o
 
 # DMA support
 
drivers/cpuidle/Kconfig.arm +7 -0
index f23bd75..5b2f84d 100644
@@ -44,3 +44,10 @@ config ARM_AT91_CPUIDLE
 	depends on ARCH_AT91
 	help
 	  Select this to enable cpuidle for AT91 processors
+
+config ARM_S3C64XX_CPUIDLE
+	bool "Cpu Idle Driver for the s3c64xx processors"
+	default y
+	depends on ARCH_S3C64XX
+	help
+	  Select this to enable cpuidle for s3c64xx processors
drivers/cpuidle/Makefile +1 -0
index 527be28..2768020 100644
@@ -13,3 +13,4 @@ obj-$(CONFIG_ARM_KIRKWOOD_CPUIDLE)	+= cpuidle-kirkwood.o
 obj-$(CONFIG_ARM_ZYNQ_CPUIDLE)		+= cpuidle-zynq.o
 obj-$(CONFIG_ARM_U8500_CPUIDLE)         += cpuidle-ux500.o
 obj-$(CONFIG_ARM_AT91_CPUIDLE)          += cpuidle-at91.o
+obj-$(CONFIG_ARM_S3C64XX_CPUIDLE)	+= cpuidle-s3c64xx.o
drivers/cpuidle/cpuidle-s3c64xx.c renamed from arch/arm/mach-s3c64xx/cpuidle.c +0 -0
similarity index 100%
rename from arch/arm/mach-s3c64xx/cpuidle.c
rename to drivers/cpuidle/cpuidle-s3c64xx.c
Signature · 1 line
1.7.9.5
#2 lore raw

PATCH 1/2 ARM: s3c64xx: cpuidle: convert to platform driver

In reply to Daniel Lezcano

Hi Daniel,

[Sending again, without HTML part. Sorry for the noise.]

Quoted 11 lines · On Friday 25 of October 2013 09:11:13 Daniel Lezcano wrote:

The driver is tied with the pm low level code making difficult to split the driver into a more arch independent code. The platform driver allows to move the standby callback into the platform data field and use a simple driver with no more dependency on the low level code.

The standby callback has a portion of code to set the standby method and the effective cpu_do_idle switching the cpu to the right mode. As this code is redundant in the cpu suspend code, it has been factored out when implementing the standby methdod.

By this way, the driver is ready to be moved out to the drivers/cpuidle.

The idea itself is quite good, but unfortunately I have to NAK this. Please see details in comments below.

Quoted 14 lines
---
 arch/arm/mach-s3c64xx/cpuidle.c |   38

++++++++++++++++---------------------- arch/arm/mach-s3c64xx/pm.c |
33 ++++++++++++++++++++++++++------- 2 files changed, 42
insertions(+), 29 deletions(-)

(quoted diff) +0 -0

b/arch/arm/mach-s3c64xx/cpuidle.c index 3c8ab07..8022f5f 100644

arch/arm/mach-s3c64xx/cpuidle.c +0 -0
@@ -9,33 +9,16 @@
  * published by the Free Software Foundation.
 */

[snip]

arch/arm/mach-s3c64xx/cpuidle.c line 35

Quoted 10 lines

static int s3c64xx_enter_idle(struct cpuidle_device *dev,

arch/arm/mach-s3c64xx/cpuidle.c +0 -6
 			      struct cpuidle_driver *drv,
 			      int index)
 {
-	unsigned long tmp;
-
-	/* Setup PWRCFG to enter idle mode */
-	tmp = __raw_readl(S3C64XX_PWR_CFG);
-	tmp &= ~S3C64XX_PWRCFG_CFG_WFI_MASK;
-	tmp |= S3C64XX_PWRCFG_CFG_WFI_IDLE;

Note the S3C64XX_PWRCFG_CFG_WFI_IDLE flag here.

Quoted 7 lines
arch/arm/mach-s3c64xx/cpuidle.c +1 -3
-	__raw_writel(tmp, S3C64XX_PWR_CFG);
-
-	cpu_do_idle();
+	s3c64xx_standby();
 
 	return index;
 }

[snip]

arch/arm/mach-s3c64xx/cpuidle.c

Quoted 7 lines
arch/arm/mach-s3c64xx/cpuidle.c +7 -0
+static void s3c64xx_set_standby(void)
+{
+	unsigned long tmp;
+
+	tmp = __raw_readl(S3C64XX_PWR_CFG);
+	tmp &= ~S3C64XX_PWRCFG_CFG_WFI_MASK;
+	tmp |= S3C64XX_PWRCFG_CFG_WFI_SLEEP;

Note the S3C64XX_PWRCFG_CFG_WFI_SLEEP flag here, which is different from the S3C64XX_PWRCFG_CFG_WFI_IDLE flag above.

Quoted 13 lines
arch/arm/mach-s3c64xx/cpuidle.c +9 -0
+	__raw_writel(tmp, S3C64XX_PWR_CFG);
+}
+
+static void s3c64xx_standby(void)
+{
+	s3c64xx_set_standby();
+	cpu_do_idle();
+}
+
 /* since both s3c6400 and s3c6410 share the same sleep pm calls, we
  * put the per-cpu code in here until any new cpu comes along and

changes * this.

arch/arm/mach-s3c64xx/cpuidle.c static int s3c64xx_cpu_suspend(unsigned long arg) · line 275

Quoted 7 lines
arch/arm/mach-s3c64xx/cpuidle.c +0 -4
@@ -269,11 +286,7 @@ static int s3c64xx_cpu_suspend(unsigned long arg)
 	unsigned long tmp;
 
 	/* set our standby method to sleep */
-
-	tmp = __raw_readl(S3C64XX_PWR_CFG);
-	tmp &= ~S3C64XX_PWRCFG_CFG_WFI_MASK;
-	tmp |= S3C64XX_PWRCFG_CFG_WFI_SLEEP;

Finally note the S3C64XX_PWRCFG_CFG_WFI_SLEEP flag here again.

I believe it should be visible now what's wrong with this patch. To make sure it is, let me explain how the system controller of S3C64xx handles WFI requests.

When the CPU issues WFI request to the syscon, it takes an action depending on how it is configured. A bit field is there in one of syscon registers (S3C64XX_PWR_CFG) that selects what action to perform in case of WFI request.

You can program the syscon to ignore the request, enter IDLE mode, enter STOP mode or enter SLEEP mode. As the names suggest, for cpuidle, it needs to be programmed for IDLE mode and for system-wide sleep it needs to be set to SLEEP mode. STOP mode is not very useful as it has mostly the same effect that can be achieved by performing fine-grained clock and power gating of peripherals manually, so it is unused by Linux.

Now, my take on the issue you are trying to solve would be a bit different. Since the S3C64xx does not have any interesting cpuidle modes, just a normal, clock-gated WFI mode, it does not need to have a cpuidle driver at all. All that is needed is simply setting up the S3C64XX_PWRCFG_CFG_WFI field to S3C64XX_PWRCFG_CFG_WFI_IDLE early at boot-up, then set it to S3C64XX_PWRCFG_CFG_WFI_SLEEP just before entering the sleep mode and restore it back to S3C64XX_PWRCFG_CFG_WFI_IDLE after waking up.

Best regards,
Tomasz

#3 lore raw

PATCH 1/2 ARM: s3c64xx: cpuidle: convert to platform driver

Quoted 26 lines · Tomasz Figa

Hi Daniel,

[Sending again, without HTML part. Sorry for the noise.]

Quoted 11 lines · On Friday 25 of October 2013 09:11:13 Daniel Lezcano wrote:

The driver is tied with the pm low level code making difficult to split the driver into a more arch independent code. The platform driver allows to move the standby callback into the platform data field and use a simple driver with no more dependency on the low level code.

The standby callback has a portion of code to set the standby method and the effective cpu_do_idle switching the cpu to the right mode. As this code is redundant in the cpu suspend code, it has been factored out when implementing the standby methdod.

By this way, the driver is ready to be moved out to the drivers/cpuidle.

The idea itself is quite good, but unfortunately I have to NAK this. Please see details in comments below.

Quoted 6 lines
---
  arch/arm/mach-s3c64xx/cpuidle.c |   38

++++++++++++++++---------------------- arch/arm/mach-s3c64xx/pm.c |
33 ++++++++++++++++++++++++++------- 2 files changed, 42
insertions(+), 29 deletions(-)

[ ... ]

Quoted 6 lines

arch/arm/mach-s3c64xx/pm.c static int s3c64xx_cpu_suspend(unsigned long arg) · line 275

Quoted 4 lines
arch/arm/mach-s3c64xx/pm.c +0 -4
-
-	tmp = __raw_readl(S3C64XX_PWR_CFG);
-	tmp &= ~S3C64XX_PWRCFG_CFG_WFI_MASK;
-	tmp |= S3C64XX_PWRCFG_CFG_WFI_SLEEP;

Finally note the S3C64XX_PWRCFG_CFG_WFI_SLEEP flag here again.

Ouch ! I missed it. Thanks for spotting the problem.

Quoted 9 lines

I believe it should be visible now what's wrong with this patch. To make sure it is, let me explain how the system controller of S3C64xx handles WFI requests.

When the CPU issues WFI request to the syscon, it takes an action depending on how it is configured. A bit field is there in one of syscon registers (S3C64XX_PWR_CFG) that selects what action to perform in case of WFI request.

Quoted 6 lines

You can program the syscon to ignore the request, enter IDLE mode, enter STOP mode or enter SLEEP mode. As the names suggest, for cpuidle, it needs to be programmed for IDLE mode and for system-wide sleep it needs to be set to SLEEP mode. STOP mode is not very useful as it has mostly the same effect that can be achieved by performing fine-grained clock and power gating of peripherals manually, so it is unused by Linux.

Yes, this is what I assumed but I missed the CFG_WFI_SLEEP flag, my eyes read CFG_WFI_IDLE.

Quoted 7 lines

Now, my take on the issue you are trying to solve would be a bit different. Since the S3C64xx does not have any interesting cpuidle modes, just a normal, clock-gated WFI mode, it does not need to have a cpuidle driver at all. All that is needed is simply setting up the S3C64XX_PWRCFG_CFG_WFI field to S3C64XX_PWRCFG_CFG_WFI_IDLE early at boot-up, then set it to S3C64XX_PWRCFG_CFG_WFI_SLEEP just before entering the sleep mode and restore it back to S3C64XX_PWRCFG_CFG_WFI_IDLE after waking up.

So you are suggesting to remove the cpuidle driver ?

Won't it be worth to add a new WFI_SLEEP state to the cpuidle driver ?

Thanks for the review.

-- 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
#4 lore raw

PATCH 1/2 ARM: s3c64xx: cpuidle: convert to platform driver

Quoted 79 lines · On Friday 25 of October 2013 21:13:35 Daniel Lezcano wrote:
Quoted 36 lines · Tomasz Figa

Hi Daniel,

[Sending again, without HTML part. Sorry for the noise.]

Quoted 14 lines · On Friday 25 of October 2013 09:11:13 Daniel Lezcano wrote:

The driver is tied with the pm low level code making difficult to split
the driver into a more arch independent code. The platform driver allows
to move the standby callback into the platform data field and use a simple driver with no more dependency on the low level code.

The standby callback has a portion of code to set the standby method and
the effective cpu_do_idle switching the cpu to the right mode. As this code is redundant in the cpu suspend code, it has been factored out when
implementing the standby methdod.

Quoted 2 lines

By this way, the driver is ready to be moved out to the drivers/cpuidle.

The idea itself is quite good, but unfortunately I have to NAK this. Please see details in comments below.

Quoted 11 lines

---

  arch/arm/mach-s3c64xx/cpuidle.c |   38

++++++++++++++++---------------------- arch/arm/mach-s3c64xx/pm.c |

   33 ++++++++++++++++++++++++++------- 2 files changed, 42

insertions(+), 29 deletions(-)

[ ... ]

Quoted 6 lines

arch/arm/mach-s3c64xx/pm.c static int s3c64xx_cpu_suspend(unsigned long arg) · line 275

Quoted 4 lines
arch/arm/mach-s3c64xx/pm.c +0 -4
-
-	tmp = __raw_readl(S3C64XX_PWR_CFG);
-	tmp &= ~S3C64XX_PWRCFG_CFG_WFI_MASK;
-	tmp |= S3C64XX_PWRCFG_CFG_WFI_SLEEP;

Finally note the S3C64XX_PWRCFG_CFG_WFI_SLEEP flag here again.

Ouch ! I missed it. Thanks for spotting the problem.

Quoted 10 lines

I believe it should be visible now what's wrong with this patch. To make
sure it is, let me explain how the system controller of S3C64xx handles WFI requests.

When the CPU issues WFI request to the syscon, it takes an action depending on how it is configured. A bit field is there in one of syscon registers (S3C64XX_PWR_CFG) that selects what action to perform in case of WFI request.

Quoted 8 lines

You can program the syscon to ignore the request, enter IDLE mode, enter
STOP mode or enter SLEEP mode. As the names suggest, for cpuidle, it needs to be programmed for IDLE mode and for system-wide sleep it needs to be set to SLEEP mode. STOP mode is not very useful as it has mostly the same effect that can be achieved by performing fine-grained clock and power gating of peripherals manually, so it is unused by Linux.

Yes, this is what I assumed but I missed the CFG_WFI_SLEEP flag, my eyes read CFG_WFI_IDLE.

Quoted 8 lines

Now, my take on the issue you are trying to solve would be a bit different. Since the S3C64xx does not have any interesting cpuidle modes, just a normal, clock-gated WFI mode, it does not need to have a cpuidle driver at all. All that is needed is simply setting up the S3C64XX_PWRCFG_CFG_WFI field to S3C64XX_PWRCFG_CFG_WFI_IDLE early at boot-up, then set it to S3C64XX_PWRCFG_CFG_WFI_SLEEP just before entering the sleep mode and restore it back to S3C64XX_PWRCFG_CFG_WFI_IDLE after waking up.

Quoted 1 line

So you are suggesting to remove the cpuidle driver ?

Exactly.

Quoted 2 lines

Won't it be worth to add a new WFI_SLEEP state to the cpuidle driver ?

I don't think so. How a suspend-to-RAM specific thing like WFI_SLEEP could be relevant to a cpuidle driver? (Unless there are some plans to consolidate STR with cpuidle that I haven't heard about...)

Best regards,
Tomasz

#5 lore raw

PATCH 1/2 ARM: s3c64xx: cpuidle: convert to platform driver

Quoted 80 lines · Tomasz Figa
Quoted 79 lines · On Friday 25 of October 2013 21:13:35 Daniel Lezcano wrote:
Quoted 36 lines · Tomasz Figa

Hi Daniel,

[Sending again, without HTML part. Sorry for the noise.]

Quoted 14 lines · On Friday 25 of October 2013 09:11:13 Daniel Lezcano wrote:

The driver is tied with the pm low level code making difficult to split
the driver into a more arch independent code. The platform driver allows
to move the standby callback into the platform data field and use a simple driver with no more dependency on the low level code.

The standby callback has a portion of code to set the standby method and
the effective cpu_do_idle switching the cpu to the right mode. As this code is redundant in the cpu suspend code, it has been factored out when
implementing the standby methdod.

Quoted 2 lines

By this way, the driver is ready to be moved out to the drivers/cpuidle.

The idea itself is quite good, but unfortunately I have to NAK this. Please see details in comments below.

Quoted 11 lines

---

   arch/arm/mach-s3c64xx/cpuidle.c |   38

++++++++++++++++---------------------- arch/arm/mach-s3c64xx/pm.c |

    33 ++++++++++++++++++++++++++------- 2 files changed, 42

insertions(+), 29 deletions(-)

[ ... ]

Quoted 6 lines

arch/arm/mach-s3c64xx/pm.c static int s3c64xx_cpu_suspend(unsigned long arg) · line 275

Quoted 4 lines
arch/arm/mach-s3c64xx/pm.c +0 -4
-
-	tmp = __raw_readl(S3C64XX_PWR_CFG);
-	tmp &= ~S3C64XX_PWRCFG_CFG_WFI_MASK;
-	tmp |= S3C64XX_PWRCFG_CFG_WFI_SLEEP;

Finally note the S3C64XX_PWRCFG_CFG_WFI_SLEEP flag here again.

Ouch ! I missed it. Thanks for spotting the problem.

Quoted 10 lines

I believe it should be visible now what's wrong with this patch. To make
sure it is, let me explain how the system controller of S3C64xx handles WFI requests.

When the CPU issues WFI request to the syscon, it takes an action depending on how it is configured. A bit field is there in one of syscon registers (S3C64XX_PWR_CFG) that selects what action to perform in case of WFI request.

Quoted 8 lines

You can program the syscon to ignore the request, enter IDLE mode, enter
STOP mode or enter SLEEP mode. As the names suggest, for cpuidle, it needs to be programmed for IDLE mode and for system-wide sleep it needs to be set to SLEEP mode. STOP mode is not very useful as it has mostly the same effect that can be achieved by performing fine-grained clock and power gating of peripherals manually, so it is unused by Linux.

Yes, this is what I assumed but I missed the CFG_WFI_SLEEP flag, my eyes read CFG_WFI_IDLE.

Quoted 8 lines

Now, my take on the issue you are trying to solve would be a bit different. Since the S3C64xx does not have any interesting cpuidle modes, just a normal, clock-gated WFI mode, it does not need to have a cpuidle driver at all. All that is needed is simply setting up the S3C64XX_PWRCFG_CFG_WFI field to S3C64XX_PWRCFG_CFG_WFI_IDLE early at boot-up, then set it to S3C64XX_PWRCFG_CFG_WFI_SLEEP just before entering the sleep mode and restore it back to S3C64XX_PWRCFG_CFG_WFI_IDLE after waking up.

Quoted 3 lines
Quoted 1 line

So you are suggesting to remove the cpuidle driver ?

Exactly.

I see.

Quoted 5 lines
Quoted 1 line

Won't it be worth to add a new WFI_SLEEP state to the cpuidle driver ?

I don't think so. How a suspend-to-RAM specific thing like WFI_SLEEP could be relevant to a cpuidle driver? (Unless there are some plans to consolidate STR with cpuidle that I haven't heard about...)

I finally found a documentation for the s3c6410x and the description of the different modes. Indeed, the sleep mode is not adequate for a cpuidle state. What about the 'stop' and 'deep stop' state ?

What is 'STR' ?

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

PATCH 1/2 ARM: s3c64xx: cpuidle: convert to platform driver

On 10/25/2013 03:23 PM, Daniel Lezcano wrote:

[ ... ]

Quoted 10 lines
Quoted 6 lines
Quoted 1 line

Won't it be worth to add a new WFI_SLEEP state to the cpuidle driver ?

I don't think so. How a suspend-to-RAM specific thing like WFI_SLEEP could
be relevant to a cpuidle driver? (Unless there are some plans to consolidate STR with cpuidle that I haven't heard about...)

I finally found a documentation for the s3c6410x and the description of the different modes. Indeed, the sleep mode is not adequate for a cpuidle state. What about the 'stop' and 'deep stop' state ?

Hi Thomas,

just a reminder about the question above, so I can go ahead: fix what you pointed out or remove the driver directly.

You mentionned in the previous email the STOP is not usful because it can be controlled by manually outside of the cpuidle driver. But I see in the documentation, the stop states power gates the cpu and deep-stop stops the regulator.

If these states have to been added later, still it worth to remove the driver ?

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
#7 lore raw

PATCH 1/2 ARM: s3c64xx: cpuidle: convert to platform driver

Hi Daniel,

Quoted 19 lines · On Wednesday 30 of October 2013 14:43:51 Daniel Lezcano wrote:

On 10/25/2013 03:23 PM, Daniel Lezcano wrote:

[ ... ]

Quoted 8 lines
Quoted 7 lines
Quoted 2 lines

Won't it be worth to add a new WFI_SLEEP state to the cpuidle driver ?

I don't think so. How a suspend-to-RAM specific thing like WFI_SLEEP could
be relevant to a cpuidle driver? (Unless there are some plans to consolidate STR with cpuidle that I haven't heard about...)

Quoted 4 lines

I finally found a documentation for the s3c6410x and the description of
the different modes. Indeed, the sleep mode is not adequate for a cpuidle state. What about the 'stop' and 'deep stop' state ?

Hi Thomas,

Quoted 7 lines

just a reminder about the question above, so I can go ahead: fix what you pointed out or remove the driver directly.

You mentionned in the previous email the STOP is not usful because it can be controlled by manually outside of the cpuidle driver. But I see in the documentation, the stop states power gates the cpu and deep-stop stops the regulator.

STOP clock-gates the CPU and DEEP-STOP power-gates it by stopping the regulator.

There are some interesting aspects of those modes, like memory self- refresh, PLL power-off and system-wide down clocking, but I believe they are too tricky to handle (coupling with device PM, stopped timers) and with too little possible power saving to justify the effort of adding support for them.

In the end, I haven't seen support for them implemented even in strange vendor kernels used even on production devices, like Android phones.

Quoted 3 lines

If these states have to been added later, still it worth to remove the driver ?

I don't think that anybody is even going to add support for them and by anybody I should probably mean myself, since I'm currently the only person actively adding new code for this platform, as a part of my hobby.

Best regards,
Tomasz

#8 lore raw

PATCH 1/2 ARM: s3c64xx: cpuidle: convert to platform driver

Quoted 43 lines · Tomasz Figa

Hi Daniel,

Quoted 19 lines · On Wednesday 30 of October 2013 14:43:51 Daniel Lezcano wrote:

On 10/25/2013 03:23 PM, Daniel Lezcano wrote:

[ ... ]

Quoted 8 lines
Quoted 7 lines
Quoted 2 lines

Won't it be worth to add a new WFI_SLEEP state to the cpuidle driver ?

I don't think so. How a suspend-to-RAM specific thing like WFI_SLEEP could
be relevant to a cpuidle driver? (Unless there are some plans to consolidate STR with cpuidle that I haven't heard about...)

Quoted 4 lines

I finally found a documentation for the s3c6410x and the description of
the different modes. Indeed, the sleep mode is not adequate for a cpuidle state. What about the 'stop' and 'deep stop' state ?

Hi Thomas,

Quoted 7 lines

just a reminder about the question above, so I can go ahead: fix what you pointed out or remove the driver directly.

You mentionned in the previous email the STOP is not usful because it can be controlled by manually outside of the cpuidle driver. But I see in the documentation, the stop states power gates the cpu and deep-stop stops the regulator.

STOP clock-gates the CPU and DEEP-STOP power-gates it by stopping the regulator.

There are some interesting aspects of those modes, like memory self- refresh, PLL power-off and system-wide down clocking, but I believe they are too tricky to handle (coupling with device PM, stopped timers) and with too little possible power saving to justify the effort of adding support for them.

In the end, I haven't seen support for them implemented even in strange vendor kernels used even on production devices, like Android phones.

Quoted 1 line
Quoted 6 lines
Quoted 2 lines

If these states have to been added later, still it worth to remove the driver ?

I don't think that anybody is even going to add support for them and by anybody I should probably mean myself, since I'm currently the only person actively adding new code for this platform, as a part of my hobby.

Ok, I will kill the driver then.

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