Skip to contentSkip to conversation

OMAP3+: PM: SR fixes

PATCH New Nishanth Menon 3 participants 7 messages Jul 22, 2011

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

cover OMAP3+: PM: SR fixes

Hi Kevin,
Please find attached a couple of patches collected as part of SR stabilization around idle and a slight improvement for suspend path. The second patch could probably help at a later date as part of moving the SmartReflex driver off as a seperate independent driver of it's own.

These are based on Linus master v3.0-rc7

Colin Cross (1):
OMAP3+: PM: SR: use put_sync_suspend for disabling

Nishanth Menon (1):
OMAP2+: PM: SR: add suspend/resume handlers

 arch/arm/mach-omap2/smartreflex.c |   89 ++++++++++++++++++++++++++++++++++++-
 1 files changed, 88 insertions(+), 1 deletions(-)
Signature · 2 lines
Regards,
Nishanth Menon
#1 lore raw

PATCH 1/2 OMAP3+: PM: SR: use put_sync_suspend for disabling

From: Colin Cross <[email protected]>

omap_sr_disable_reset_volt is called with irqs off in omapx_enter_sleep, as part of idle sequence, this eventually calls sr_disable and pm_runtime_put_sync. pm_runtime_put_sync calls rpm_idle, which will enable interrupts in order to call the callback. In this short interval when interrupts are enabled, scenarios such as the following can occur:
while interrupts are enabled, the timer interrupt that is supposed to wake the device out of idle occurs and is acked, so when the CPU finally goes to off, the timer is already gone, missing a wakeup event.

Further, as the documentation for runtime states:"
However, subsystems can use the pm_runtime_irq_safe() helper function to tell the PM core that a device's ->runtime_suspend() and ->runtime_resume() callbacks should be invoked in atomic context with interrupts disabled (->runtime_idle() is still invoked the default way)."

Hence, replace pm_runtime_put_sync with pm_runtime_put_sync_suspend to invoke the suspend handler and shut off the fclk for SmartReflex module instead of using the idle handler in interrupt disabled context.

---
 arch/arm/mach-omap2/smartreflex.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
arch/arm/mach-omap2/smartreflex.c +1 -1
index fb7dc52..33a027f 100644
@@ -622,7 +622,7 @@ void sr_disable(struct voltagedomain *voltdm)
 			sr_v2_disable(sr);
 	}
 
-	pm_runtime_put_sync(&sr->pdev->dev);
+	pm_runtime_put_sync_suspend(&sr->pdev->dev);
 }
 
 /**
Signature · 1 line
1.7.4.1
#2 lore raw

PATCH 1/2 OMAP3+: PM: SR: use put_sync_suspend for disabling

Quoted 17 lines · Nishanth Menon

From: Colin Cross <[email protected]>

omap_sr_disable_reset_volt is called with irqs off in omapx_enter_sleep, as part of idle sequence, this eventually calls sr_disable and pm_runtime_put_sync. pm_runtime_put_sync calls rpm_idle, which will enable interrupts in order to call the callback. In this short interval when interrupts are enabled, scenarios such as the following can occur:
while interrupts are enabled, the timer interrupt that is supposed to wake the device out of idle occurs and is acked, so when the CPU finally goes to off, the timer is already gone, missing a wakeup event.

Further, as the documentation for runtime states:"
However, subsystems can use the pm_runtime_irq_safe() helper function to tell the PM core that a device's ->runtime_suspend() and ->runtime_resume() callbacks should be invoked in atomic context with interrupts disabled (->runtime_idle() is still invoked the default way)."

Quoted 6 lines

Hence, replace pm_runtime_put_sync with pm_runtime_put_sync_suspend to invoke the suspend handler and shut off the fclk for SmartReflex module instead of using the idle handler in interrupt disabled context.

Great catch!

Looking through Documentation/power/runtime_pm.txt, I see (now) that it is well documented that _put_sync_suspend() is safe to use from interrupts-disabled context, but _put_sync() is not on that list.

Queuing this as a fix for v3.1.

Kevin

#3 lore raw

PATCH 2/2 OMAP2+: PM: SR: add suspend/resume handlers

In reply to Nishanth Menon

Suspend and Resume paths are safe enough to do it in the standard LDM suspend/resume handlers where one can sleep. Add suspend/resume handlers for SmartReflex.

---
 arch/arm/mach-omap2/smartreflex.c |   87 +++++++++++++++++++++++++++++++++++++
 1 files changed, 87 insertions(+), 0 deletions(-)
arch/arm/mach-omap2/smartreflex.c +87 -0
index 33a027f..fb90bd2 100644
@@ -39,6 +39,7 @@ struct omap_sr {
 	int				ip_type;
 	int				nvalue_count;
 	bool				autocomp_active;
+	bool				is_suspended;
 	u32				clk_length;
 	u32				err_weight;
 	u32				err_minlimit;
@@ -684,6 +685,12 @@ void omap_sr_enable(struct voltagedomain *voltdm)
 	if (!sr->autocomp_active)
 		return;
 
+	if (sr->is_suspended) {
+		dev_dbg(&sr->pdev->dev, "%s: in suspended state\n", __func__);
+		return;
+	}
+
+
 	if (!sr_class || !(sr_class->enable) || !(sr_class->configure)) {
 		dev_warn(&sr->pdev->dev, "%s: smartreflex class driver not"
 			"registered\n", __func__);
@@ -717,6 +724,11 @@ void omap_sr_disable(struct voltagedomain *voltdm)
 	if (!sr->autocomp_active)
 		return;
 
+	if (sr->is_suspended) {
+		dev_dbg(&sr->pdev->dev, "%s: in suspended state\n", __func__);
+		return;
+	}
+
 	if (!sr_class || !(sr_class->disable)) {
 		dev_warn(&sr->pdev->dev, "%s: smartreflex class driver not"
 			"registered\n", __func__);
@@ -750,6 +762,11 @@ void omap_sr_disable_reset_volt(struct voltagedomain *voltdm)
 	if (!sr->autocomp_active)
 		return;
 
+	if (sr->is_suspended) {
+		dev_dbg(&sr->pdev->dev, "%s: in suspended state\n", __func__);
+		return;
+	}
+
 	if (!sr_class || !(sr_class->disable)) {
 		dev_warn(&sr->pdev->dev, "%s: smartreflex class driver not"
 			"registered\n", __func__);
@@ -808,6 +825,11 @@ static int omap_sr_autocomp_store(void *data, u64 val)
 		return -EINVAL;
 	}
 
+	if (sr_info->is_suspended) {
+		pr_warning("%s: in suspended state\n", __func__);
+		return -EBUSY;
+	}
+
 	if (!val)
 		sr_stop_vddautocomp(sr_info);
 	else
@@ -998,8 +1020,73 @@ static int __devexit omap_sr_remove(struct platform_device *pdev)
 	return 0;
 }
 
+static int omap_sr_suspend(struct platform_device *pdev, pm_message_t state)
+{
+	struct omap_sr_data *pdata = pdev->dev.platform_data;
+	struct omap_sr *sr_info;
+
+	if (!pdata) {
+		dev_err(&pdev->dev, "%s: platform data missing\n", __func__);
+		return -EINVAL;
+	}
+
+	sr_info = _sr_lookup(pdata->voltdm);
+	if (IS_ERR(sr_info)) {
+		dev_warn(&pdev->dev, "%s: omap_sr struct not found\n",
+			__func__);
+		return -EINVAL;
+	}
+
+	if (!sr_info->autocomp_active)
+		return 0;
+
+	if (sr_info->is_suspended)
+		return 0;
+
+	omap_sr_disable_reset_volt(pdata->voltdm);
+	sr_info->is_suspended = true;
+	/* Flag the same info to the other CPUs */
+	smp_wmb();
+
+	return 0;
+}
+
+static int omap_sr_resume(struct platform_device *pdev)
+{
+	struct omap_sr_data *pdata = pdev->dev.platform_data;
+	struct omap_sr *sr_info;
+
+	if (!pdata) {
+		dev_err(&pdev->dev, "%s: platform data missing\n", __func__);
+		return -EINVAL;
+	}
+
+	sr_info = _sr_lookup(pdata->voltdm);
+	if (IS_ERR(sr_info)) {
+		dev_warn(&pdev->dev, "%s: omap_sr struct not found\n",
+			__func__);
+		return -EINVAL;
+	}
+
+	if (!sr_info->autocomp_active)
+		return 0;
+
+	if (!sr_info->is_suspended)
+		return 0;
+
+	sr_info->is_suspended = false;
+	/* Flag the same info to the other CPUs */
+	smp_wmb();
+	omap_sr_enable(pdata->voltdm);
+
+	return 0;
+}
+
+
 static struct platform_driver smartreflex_driver = {
 	.remove         = omap_sr_remove,
+	.suspend	= omap_sr_suspend,
+	.resume		= omap_sr_resume,
 	.driver		= {
 		.name	= "smartreflex",
 	},
Signature · 1 line
1.7.4.1
#4 lore raw

PATCH 2/2 OMAP2+: PM: SR: add suspend/resume handlers

Hi,

Quoted 22 lines · Nishanth Menon

Suspend and Resume paths are safe enough to do it in the standard LDM suspend/resume handlers where one can sleep. Add suspend/resume handlers for SmartReflex.

---
 arch/arm/mach-omap2/smartreflex.c |   87 +++++++++++++++++++++++++++++++++++++
 1 files changed, 87 insertions(+), 0 deletions(-)
arch/arm/mach-omap2/smartreflex.c +1 -0
index 33a027f..fb90bd2 100644
@@ -39,6 +39,7 @@ struct omap_sr {
 	int				ip_type;
 	int				nvalue_count;
 	bool				autocomp_active;
+	bool				is_suspended;
 	u32				clk_length;
 	u32				err_weight;
 	u32				err_minlimit;

arch/arm/mach-omap2/smartreflex.c void omap_sr_enable(struct voltagedomain *voltdm) · line 691

Quoted 7 lines
arch/arm/mach-omap2/smartreflex.c +4 -0
@@ -684,6 +685,12 @@ void omap_sr_enable(struct voltagedomain *voltdm)
 	if (!sr->autocomp_active)
 		return;
 
+	if (sr->is_suspended) {
+		dev_dbg(&sr->pdev->dev, "%s: in suspended state\n", __func__);
+		return;
+	}

I wonder why you get these called if you're in suspend state. If this is called by some other driver, then shouldn't you pm_runtime_get_sync(); do_whatever_you_need_to_do(); and pm_runtime_put(); rather then just returning ?

Signature · 8 lines
balbi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 490 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20110722/1df9ca98/attachment.sig>
#5 lore raw

PATCH 2/2 OMAP2+: PM: SR: add suspend/resume handlers

Quoted 28 lines · Felipe Balbi

Hi,

Quoted 25 lines · Nishanth Menon

Suspend and Resume paths are safe enough to do it in the standard LDM suspend/resume handlers where one can sleep. Add suspend/resume handlers for SmartReflex.

---
?arch/arm/mach-omap2/smartreflex.c | ? 87 +++++++++++++++++++++++++++++++++++++ ?1 files changed, 87 insertions(+), 0 deletions(-)

arch/arm/mach-omap2/smartreflex.c +1 -0
index 33a027f..fb90bd2 100644
@@ -39,6 +39,7 @@ struct omap_sr {
 ? ? ? int ? ? ? ? ? ? ? ? ? ? ? ? ? ? ip_type;
 ? ? ? int ? ? ? ? ? ? ? ? ? ? ? ? ? ? nvalue_count;
 ? ? ? bool ? ? ? ? ? ? ? ? ? ? ? ? ? ?autocomp_active;
+ ? ? bool ? ? ? ? ? ? ? ? ? ? ? ? ? ?is_suspended;
 ? ? ? u32 ? ? ? ? ? ? ? ? ? ? ? ? ? ? clk_length;
 ? ? ? u32 ? ? ? ? ? ? ? ? ? ? ? ? ? ? err_weight;
 ? ? ? u32 ? ? ? ? ? ? ? ? ? ? ? ? ? ? err_minlimit;
@@ -684,6 +685,12 @@ void omap_sr_enable(struct voltagedomain *voltdm)
 ? ? ? if (!sr->autocomp_active)
 ? ? ? ? ? ? ? return;
 
Quoted 9 lines

arch/arm/mach-omap2/smartreflex.c void omap_sr_enable(struct voltagedomain *voltdm) · line 691

Quoted 4 lines
arch/arm/mach-omap2/smartreflex.c +4 -0
+ ? ? if (sr->is_suspended) {
+ ? ? ? ? ? ? dev_dbg(&sr->pdev->dev, "%s: in suspended state\n", __func__);
+ ? ? ? ? ? ? return;
+ ? ? }

I wonder why you get these called if you're in suspend state. If this is called by some other driver, then shouldn't you pm_runtime_get_sync(); do_whatever_you_need_to_do(); and pm_runtime_put(); rather then just returning ?

because we have two state machines running in parallel - cpu idle and dvfs(cpufreq, and other dependent domain voltage scaling) and SR driver is just one - it does it's own get_sync and put_sync_suspend. we cannot guarentee that the SR enable/disable calls can be properly sequenced unless we use suspend lockouts. SmartReflex driver is an independent entity of it's own - yeah we can do the same as we have done historically in the omap[34]_idle paths(which is disable SR after we have disabled interrupts), but in case of suspend(unlike in idle), we do *know* that SR will have to be disabled - hence doing it earlier as part of standard LDM suspend sequences is more opportunistic.

Regards,
Nishanth Menon

#6 lore raw

PATCH 2/2 OMAP2+: PM: SR: add suspend/resume handlers

In reply to Nishanth Menon
Quoted 1 line · Nishanth Menon

Suspend and Resume paths are safe enough to do it in

What is 'it' ?

Quoted 2 lines

the standard LDM suspend/resume handlers where one can sleep. Add suspend/resume handlers for SmartReflex.

Minor comments on the code below, but this changelog doesn't read well, or at least I can't make any sense of it.

[...]

arch/arm/mach-omap2/smartreflex.c void omap_sr_enable(struct voltagedomain *voltdm) · line 693

Quoted 10 lines
arch/arm/mach-omap2/smartreflex.c +6 -0
@@ -684,6 +685,12 @@ void omap_sr_enable(struct voltagedomain *voltdm)
 	if (!sr->autocomp_active)
 		return;
 
+	if (sr->is_suspended) {
+		dev_dbg(&sr->pdev->dev, "%s: in suspended state\n", __func__);
+		return;
+	}
+
+

extra blank line

Quoted 3 lines
arch/arm/mach-omap2/smartreflex.c +0 -0
 	if (!sr_class || !(sr_class->enable) || !(sr_class->configure)) {
 		dev_warn(&sr->pdev->dev, "%s: smartreflex class driver not"
 			"registered\n", __func__);

[...]

arch/arm/mach-omap2/smartreflex.c static int __devexit omap_sr_remove(struct platform_device *pdev) · line 1089

Quoted 4 lines

static struct platform_driver smartreflex_driver = {

arch/arm/mach-omap2/smartreflex.c +2 -0
 	.remove         = omap_sr_remove,
+	.suspend	= omap_sr_suspend,
+	.resume		= omap_sr_resume,

You're using the legacy methods here, please use dev_pm_ops.

That means you'll need to create a struct dev_pm_ops and fill in these mehods there (and note the dev_pm_ops methods don't have a 'state' argument.

Also, when implementing suspend/resume, you need to make sure the hibernate callbacks are populated also. They should be populated with the same callbacks, so the best way to do this is to use SIMPLE_DEV_PM_OPS() (see <linux/pm.h>). That macro also takes care of the !CONFIG_PM case as well.

IOW, the result would look someting like this (not even compile tested):

static SIMPLE_DEV_PM_OPS(omap_sr_pm_ops, omap_sr_suspend, omap_sr_resume)

static struct platform_driver smartreflex_driver = {
	.remove         = omap_sr_remove,
	.driver		= {
		.name	= "smartreflex",
                .pm     = &omap_sr_pm_ops,
	},
};

Kevin