Skip to contentSkip to conversation

watchdog: qcom: Add suspend/resume support

PATCH In review Sai Prakash Ranjan 5 participants 10 messages Jan 17, 2019 – Jan 18, 2019

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 watchdog: qcom: Add suspend/resume support

This adds the support for qcom watchdog suspend and resume when entering and exiting deep sleep states. Otherwise having watchdog active after suspend would result in unwanted crashes/resets if resume happens after a long time.

---

v2:
* Use __maybe_unused instead of #ifdef as suggested by Brian and Guenter

---
 drivers/watchdog/qcom-wdt.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)
drivers/watchdog/qcom-wdt.c +23 -0
index 780971318810..5dfd604477a4 100644
@@ -245,6 +245,28 @@ static int qcom_wdt_remove(struct platform_device *pdev)
 	return 0;
 }
 
+static int __maybe_unused qcom_wdt_suspend(struct device *dev)
+{
+	struct qcom_wdt *wdt = dev_get_drvdata(dev);
+
+	if (watchdog_active(&wdt->wdd))
+		qcom_wdt_stop(&wdt->wdd);
+
+	return 0;
+}
+
+static int __maybe_unused qcom_wdt_resume(struct device *dev)
+{
+	struct qcom_wdt *wdt = dev_get_drvdata(dev);
+
+	if (watchdog_active(&wdt->wdd))
+		qcom_wdt_start(&wdt->wdd);
+
+	return 0;
+}
+
+static SIMPLE_DEV_PM_OPS(qcom_wdt_pm_ops, qcom_wdt_suspend, qcom_wdt_resume);
+
 static const struct of_device_id qcom_wdt_of_table[] = {
 	{ .compatible = "qcom,kpss-timer", .data = reg_offset_data_apcs_tmr },
 	{ .compatible = "qcom,scss-timer", .data = reg_offset_data_apcs_tmr },
@@ -259,6 +281,7 @@ static struct platform_driver qcom_watchdog_driver = {
 	.driver	= {
 		.name		= KBUILD_MODNAME,
 		.of_match_table	= qcom_wdt_of_table,
+		.pm		= &qcom_wdt_pm_ops,
 	},
 };
 module_platform_driver(qcom_watchdog_driver);
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#1 lore raw
Quoted 7 lines · Sai Prakash Ranjan

This adds the support for qcom watchdog suspend and resume when entering and exiting deep sleep states. Otherwise having watchdog active after suspend would result in unwanted crashes/resets if resume happens after a long time.

Quoted 10 lines

---

v2:
* Use __maybe_unused instead of #ifdef as suggested by Brian and Guenter

---
  drivers/watchdog/qcom-wdt.c | 23 +++++++++++++++++++++++
  1 file changed, 23 insertions(+)

Adding Brian.

Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
maintainer Reviewed-by on v2 1/1
#2 lore raw
In reply to Sai Prakash Ranjan
Quoted 6 lines · Sai Prakash Ranjan

This adds the support for qcom watchdog suspend and resume when entering and exiting deep sleep states. Otherwise having watchdog active after suspend would result in unwanted crashes/resets if resume happens after a long time.

Quoted 55 lines

---

v2:
* Use __maybe_unused instead of #ifdef as suggested by Brian and Guenter

---
 drivers/watchdog/qcom-wdt.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)
drivers/watchdog/qcom-wdt.c +23 -0
index 780971318810..5dfd604477a4 100644
@@ -245,6 +245,28 @@ static int qcom_wdt_remove(struct platform_device *pdev)
 	return 0;
 }
 
+static int __maybe_unused qcom_wdt_suspend(struct device *dev)
+{
+	struct qcom_wdt *wdt = dev_get_drvdata(dev);
+
+	if (watchdog_active(&wdt->wdd))
+		qcom_wdt_stop(&wdt->wdd);
+
+	return 0;
+}
+
+static int __maybe_unused qcom_wdt_resume(struct device *dev)
+{
+	struct qcom_wdt *wdt = dev_get_drvdata(dev);
+
+	if (watchdog_active(&wdt->wdd))
+		qcom_wdt_start(&wdt->wdd);
+
+	return 0;
+}
+
+static SIMPLE_DEV_PM_OPS(qcom_wdt_pm_ops, qcom_wdt_suspend, qcom_wdt_resume);
+
 static const struct of_device_id qcom_wdt_of_table[] = {
 	{ .compatible = "qcom,kpss-timer", .data = reg_offset_data_apcs_tmr },
 	{ .compatible = "qcom,scss-timer", .data = reg_offset_data_apcs_tmr },
@@ -259,6 +281,7 @@ static struct platform_driver qcom_watchdog_driver = {
 	.driver	= {
 		.name		= KBUILD_MODNAME,
 		.of_match_table	= qcom_wdt_of_table,
+		.pm		= &qcom_wdt_pm_ops,
 	},
 };
 module_platform_driver(qcom_watchdog_driver);
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#3 lore raw
Quoted 10 lines · Guenter Roeck
Quoted 6 lines · Sai Prakash Ranjan

This adds the support for qcom watchdog suspend and resume when entering and exiting deep sleep states. Otherwise having watchdog active after suspend would result in unwanted crashes/resets if resume happens after a long time.

Thanks Guenter. Should I send a v3 with your reviewed-by?

- Sai

Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
maintainer
#4 lore raw
Quoted 7 lines · Sai Prakash Ranjan
Quoted 6 lines · Guenter Roeck
Quoted 5 lines · Sai Prakash Ranjan

This adds the support for qcom watchdog suspend and resume when entering and exiting deep sleep states. Otherwise having watchdog active after suspend would result in unwanted crashes/resets if resume happens after a long time.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

Thanks Guenter. Should I send a v3 with your reviewed-by?

I add it to my watchdog-next branch and Wim will pick it up from there. No need to resend.

Guenter

#5 lore raw
Quoted 10 lines · Guenter Roeck
Quoted 9 lines · Sai Prakash Ranjan
Quoted 8 lines · Guenter Roeck
Quoted 6 lines · Sai Prakash Ranjan

This adds the support for qcom watchdog suspend and resume when entering and exiting deep sleep states. Otherwise having watchdog active after suspend would result in unwanted crashes/resets if resume happens after a long time.

Quoted 9 lines
Quoted 5 lines
Quoted 2 lines

Thanks Guenter. Should I send a v3 with your reviewed-by?

I add it to my watchdog-next branch and Wim will pick it up from there. No need to resend.

Ok thanks.

- Sai

Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#6 lore raw
In reply to Sai Prakash Ranjan

Quoting Sai Prakash Ranjan (2019-01-17 07:19:42)

Quoted 17 lines
drivers/watchdog/qcom-wdt.c +9 -0
index 780971318810..5dfd604477a4 100644
@@ -245,6 +245,28 @@ static int qcom_wdt_remove(struct platform_device *pdev)
        return 0;
 }
 
+static int __maybe_unused qcom_wdt_suspend(struct device *dev)
+{
+       struct qcom_wdt *wdt = dev_get_drvdata(dev);
+
+       if (watchdog_active(&wdt->wdd))
+               qcom_wdt_stop(&wdt->wdd);
+
+       return 0;
+}

drivers/watchdog/qcom-wdt.c static int qcom_wdt_remove(struct platform_device *pdev) · line 266

Quoted 10 lines
drivers/watchdog/qcom-wdt.c +10 -0
+
+static int __maybe_unused qcom_wdt_resume(struct device *dev)
+{
+       struct qcom_wdt *wdt = dev_get_drvdata(dev);
+
+       if (watchdog_active(&wdt->wdd))
+               qcom_wdt_start(&wdt->wdd);
+
+       return 0;
+}

This looks fairly generic. For example, the Mediatek driver also stops and starts (but also pings after starting). Grepping for 'active' in drivers/watchdog/ finds more examples. Could there be some functions in watchdog core that do the common things like watchdog_stop() and watchdog_start() and watchdog_start_and_ping()? Or maybe a bit can be set during registration so that the 'struct class watchdog_class' can get PM ops to stop and start on suspend/resume of the watchdog character device?

Nothing is wrong with the patch, I'm just bemoaning the amount of code duplication here.

#7 lore raw
Quoted 39 lines · Stephen Boyd

Quoting Sai Prakash Ranjan (2019-01-17 07:19:42)

Quoted 17 lines
drivers/watchdog/qcom-wdt.c +9 -0
index 780971318810..5dfd604477a4 100644
@@ -245,6 +245,28 @@ static int qcom_wdt_remove(struct platform_device *pdev)
        return 0;
 }
 
+static int __maybe_unused qcom_wdt_suspend(struct device *dev)
+{
+       struct qcom_wdt *wdt = dev_get_drvdata(dev);
+
+       if (watchdog_active(&wdt->wdd))
+               qcom_wdt_stop(&wdt->wdd);
+
+       return 0;
+}

drivers/watchdog/qcom-wdt.c static int qcom_wdt_remove(struct platform_device *pdev) · line 266

Quoted 10 lines
drivers/watchdog/qcom-wdt.c +10 -0
+
+static int __maybe_unused qcom_wdt_resume(struct device *dev)
+{
+       struct qcom_wdt *wdt = dev_get_drvdata(dev);
+
+       if (watchdog_active(&wdt->wdd))
+               qcom_wdt_start(&wdt->wdd);
+
+       return 0;
+}

This looks fairly generic. For example, the Mediatek driver also stops and starts (but also pings after starting). Grepping for 'active' in drivers/watchdog/ finds more examples. Could there be some functions in watchdog core that do the common things like watchdog_stop() and watchdog_start() and watchdog_start_and_ping()? Or maybe a bit can be set during registration so that the 'struct class watchdog_class' can get PM ops to stop and start on suspend/resume of the watchdog character device?

Quoted 3 lines

Nothing is wrong with the patch, I'm just bemoaning the amount of code duplication here.

Patch(es) to add the functionality to the watchdog core are welcome; it should be possible to move the functionality into the core (maybe to be enabled with a new watchdog API call). Doing it using the class device sounds like an excellent idea. This should be straightforward to implement, though the question of "should we ping on resume or not" might be an endless source for bike shedding fun.

Guenter

#8 lore raw
Quoted 44 lines · Guenter Roeck
Quoted 39 lines · Stephen Boyd

Quoting Sai Prakash Ranjan (2019-01-17 07:19:42)

Quoted 17 lines
drivers/watchdog/qcom-wdt.c +9 -0
index 780971318810..5dfd604477a4 100644
@@ -245,6 +245,28 @@ static int qcom_wdt_remove(struct platform_device *pdev)
        return 0;
 }
 
+static int __maybe_unused qcom_wdt_suspend(struct device *dev)
+{
+       struct qcom_wdt *wdt = dev_get_drvdata(dev);
+
+       if (watchdog_active(&wdt->wdd))
+               qcom_wdt_stop(&wdt->wdd);
+
+       return 0;
+}

drivers/watchdog/qcom-wdt.c static int qcom_wdt_remove(struct platform_device *pdev) · line 266

Quoted 10 lines
drivers/watchdog/qcom-wdt.c +10 -0
+
+static int __maybe_unused qcom_wdt_resume(struct device *dev)
+{
+       struct qcom_wdt *wdt = dev_get_drvdata(dev);
+
+       if (watchdog_active(&wdt->wdd))
+               qcom_wdt_start(&wdt->wdd);
+
+       return 0;
+}

This looks fairly generic. For example, the Mediatek driver also stops and starts (but also pings after starting). Grepping for 'active' in drivers/watchdog/ finds more examples. Could there be some functions in watchdog core that do the common things like watchdog_stop() and watchdog_start() and watchdog_start_and_ping()? Or maybe a bit can be set during registration so that the 'struct class watchdog_class' can get PM ops to stop and start on suspend/resume of the watchdog character device?

Quoted 3 lines

Nothing is wrong with the patch, I'm just bemoaning the amount of code duplication here.

Quoted 8 lines

Patch(es) to add the functionality to the watchdog core are welcome; it should be possible to move the functionality into the core (maybe to be enabled with a new watchdog API call). Doing it using the class device sounds like an excellent idea. This should be straightforward to implement, though the question of "should we ping on resume or not" might be an endless source for bike shedding fun.

Guenter

It could be controlled by a currently unused bit in the watchdog_info.options. Then people can agree to disagree.

Signature · 3 lines
-----------------------------------------------------------------------------
Jerry Hoemann                  Software Engineer   Hewlett Packard Enterprise
-----------------------------------------------------------------------------
#9 lore raw
Quoted 53 lines · Jerry Hoemann
Quoted 51 lines · Guenter Roeck
Quoted 39 lines · Stephen Boyd

Quoting Sai Prakash Ranjan (2019-01-17 07:19:42)

Quoted 17 lines
drivers/watchdog/qcom-wdt.c +9 -0
index 780971318810..5dfd604477a4 100644
@@ -245,6 +245,28 @@ static int qcom_wdt_remove(struct platform_device *pdev)
        return 0;
 }
 
+static int __maybe_unused qcom_wdt_suspend(struct device *dev)
+{
+       struct qcom_wdt *wdt = dev_get_drvdata(dev);
+
+       if (watchdog_active(&wdt->wdd))
+               qcom_wdt_stop(&wdt->wdd);
+
+       return 0;
+}

drivers/watchdog/qcom-wdt.c static int qcom_wdt_remove(struct platform_device *pdev) · line 266

Quoted 10 lines
drivers/watchdog/qcom-wdt.c +10 -0
+
+static int __maybe_unused qcom_wdt_resume(struct device *dev)
+{
+       struct qcom_wdt *wdt = dev_get_drvdata(dev);
+
+       if (watchdog_active(&wdt->wdd))
+               qcom_wdt_start(&wdt->wdd);
+
+       return 0;
+}

This looks fairly generic. For example, the Mediatek driver also stops and starts (but also pings after starting). Grepping for 'active' in drivers/watchdog/ finds more examples. Could there be some functions in watchdog core that do the common things like watchdog_stop() and watchdog_start() and watchdog_start_and_ping()? Or maybe a bit can be set during registration so that the 'struct class watchdog_class' can get PM ops to stop and start on suspend/resume of the watchdog character device?

Quoted 3 lines

Nothing is wrong with the patch, I'm just bemoaning the amount of code duplication here.

Patch(es) to add the functionality to the watchdog core are welcome; it should be possible to move the functionality into the core (maybe to be enabled with a new watchdog API call). Doing it using the class device sounds like an excellent idea. This should be straightforward to implement, though the question of "should we ping on resume or not" might be an endless source for bike shedding fun.

Quoted 5 lines
Quoted 1 line

Guenter

It could be controlled by a currently unused bit in the watchdog_info.options. Then people can agree to disagree.

I thought about that, but that would result in inconsistent behavior across drivers. Of course we could be all fancy and make it runtime configurable via ioctl.

Guenter