Skip to contentSkip to conversation

watchdog: add support for QCOM WDT

PATCH Superseded Josh Cartwright 4 participants 9 messages Sep 23, 2014 – Sep 24, 2014

Subsystems (3)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v2 cover watchdog: add support for QCOM WDT

This patchset provides support for the Watchdog Timer (WDT) found in the Krait Processor Sub-system (KPSS) of the MSM8960, APQ8064, and IPQ8064 chips.

This driver is implemented ontop of WATCHDOG_CORE, and therefore its primary interface is through userspace. The implemantion is currently very basic (i.e. it doesn't support PRETIMEOUT, even though it could be implemented through the WDT's BARK functionality). It should also be fairly easy to extend this driver in the future to support newer chipsets as well.

Patch 3 also extends the driver to also register a restart_notifier, making it possible for the WDT to act as a restart mechanism if more favorable mechanisms don't work. This is important for some boards which don't support PS_HOLD, like the IPQ8064-based AP148 board.

Changes since v1:
- Make use of clock API instead of using a 'clock-frequency' property
- Setup default timeout of 30 seconds when one is not specified
- Add remove() function to allow for module unloading
- Don't acquire/release watchdog lock on restart
- Don't bail completely if restart_handler registration fails

Josh Cartwright (3):
watchdog: qcom: add support for KPSS WDT
watchdog: qcom: document device tree bindings
watchdog: qcom: register a restart notifier

 .../devicetree/bindings/watchdog/qcom-wdt.txt      |  22 +++
 drivers/watchdog/Kconfig                           |  13 ++
 drivers/watchdog/Makefile                          |   1 +
 drivers/watchdog/qcom-wdt.c                        | 204 +++++++++++++++++++++
 4 files changed, 240 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/watchdog/qcom-wdt.txt
 create mode 100644 drivers/watchdog/qcom-wdt.c
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation
#1 lore raw

PATCH v2 1/3 watchdog: qcom: add support for KPSS WDT

Add a driver for the watchdog timer block found in the Krait Processor Subsystem (KPSS) on the MSM8960, APQ8064, and IPQ8064.

---
 drivers/watchdog/Kconfig    |  13 ++++
 drivers/watchdog/Makefile   |   1 +
 drivers/watchdog/qcom-wdt.c | 176 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 190 insertions(+)
 create mode 100644 drivers/watchdog/qcom-wdt.c
drivers/watchdog/Kconfig +13 -0
index 1d1330a..0479e3f 100644
@@ -443,6 +443,19 @@ config TEGRA_WATCHDOG
 	  To compile this driver as a module, choose M here: the
 	  module will be called tegra_wdt.
 
+config QCOM_WDT
+	tristate "QCOM watchdog"
+	depends on HAS_IOMEM
+	depends on ARCH_QCOM
+	select WATCHDOG_CORE
+	help
+	  Say Y here to include Watchdog timer support for the watchdog found
+	  on QCOM chipsets.  Currently supported targets are the MSM8960,
+	  APQ8064, and IPQ8064.
+
+	  To compile this driver as a module, choose M here: the
+	  module will be called qcom_wdt.
+
 # AVR32 Architecture
 
 config AT32AP700X_WDT
drivers/watchdog/Makefile +1 -0
index 468c320..d645448 100644
@@ -57,6 +57,7 @@ obj-$(CONFIG_RETU_WATCHDOG) += retu_wdt.o
 obj-$(CONFIG_BCM2835_WDT) += bcm2835_wdt.o
 obj-$(CONFIG_MOXART_WDT) += moxart_wdt.o
 obj-$(CONFIG_SIRFSOC_WATCHDOG) += sirfsoc_wdt.o
+obj-$(CONFIG_QCOM_WDT) += qcom-wdt.o
 obj-$(CONFIG_BCM_KONA_WDT) += bcm_kona_wdt.o
 obj-$(CONFIG_TEGRA_WATCHDOG) += tegra_wdt.o
 
drivers/watchdog/qcom-wdt.c new file +176 -0

176 lines not on this page · raw message

Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation
maintainer
#2 lore raw

Re: [PATCH v2 1/3] watchdog: qcom: add support for KPSS WDT

Quoted 4 lines · Josh Cartwright

Add a driver for the watchdog timer block found in the Krait Processor Subsystem (KPSS) on the MSM8960, APQ8064, and IPQ8064.

Hi Josh,

looks much better. Couple of comments inline.

Thanks,
Guenter

Quoted 162 lines
---
 drivers/watchdog/Kconfig    |  13 ++++
 drivers/watchdog/Makefile   |   1 +
 drivers/watchdog/qcom-wdt.c | 176 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 190 insertions(+)
 create mode 100644 drivers/watchdog/qcom-wdt.c
drivers/watchdog/Kconfig +13 -0
index 1d1330a..0479e3f 100644
@@ -443,6 +443,19 @@ config TEGRA_WATCHDOG
 	  To compile this driver as a module, choose M here: the
 	  module will be called tegra_wdt.
 
+config QCOM_WDT
+	tristate "QCOM watchdog"
+	depends on HAS_IOMEM
+	depends on ARCH_QCOM
+	select WATCHDOG_CORE
+	help
+	  Say Y here to include Watchdog timer support for the watchdog found
+	  on QCOM chipsets.  Currently supported targets are the MSM8960,
+	  APQ8064, and IPQ8064.
+
+	  To compile this driver as a module, choose M here: the
+	  module will be called qcom_wdt.
+
 # AVR32 Architecture
 
 config AT32AP700X_WDT
drivers/watchdog/Makefile +1 -0
index 468c320..d645448 100644
@@ -57,6 +57,7 @@ obj-$(CONFIG_RETU_WATCHDOG) += retu_wdt.o
 obj-$(CONFIG_BCM2835_WDT) += bcm2835_wdt.o
 obj-$(CONFIG_MOXART_WDT) += moxart_wdt.o
 obj-$(CONFIG_SIRFSOC_WATCHDOG) += sirfsoc_wdt.o
+obj-$(CONFIG_QCOM_WDT) += qcom-wdt.o
 obj-$(CONFIG_BCM_KONA_WDT) += bcm_kona_wdt.o
 obj-$(CONFIG_TEGRA_WATCHDOG) += tegra_wdt.o
 
drivers/watchdog/qcom-wdt.c new file +113 -0
new file mode 100644
index 0000000..d5e46e2
@@ -0,0 +1,176 @@
+/* Copyright (c) 2014, The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ */
+#include <linux/clk.h>
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/watchdog.h>
+
+#define WDT_RST		0x0
+#define WDT_EN		0x8
+#define WDT_BITE_TIME	0x24
+
+struct qcom_wdt {
+	struct watchdog_device	wdd;
+	struct clk		*clk;
+	void __iomem		*base;
+};
+
+static inline
+struct qcom_wdt *to_qcom_wdt(struct watchdog_device *wdd)
+{
+	return container_of(wdd, struct qcom_wdt, wdd);
+}
+
+static int qcom_wdt_start(struct watchdog_device *wdd)
+{
+	struct qcom_wdt *wdt = to_qcom_wdt(wdd);
+	unsigned long bite_time;
+
+	bite_time = wdd->timeout * clk_get_rate(wdt->clk);
+
+	writel(0, wdt->base + WDT_EN);
+	writel(1, wdt->base + WDT_RST);
+	writel(bite_time, wdt->base + WDT_BITE_TIME);
+	writel(1, wdt->base + WDT_EN);
+	return 0;
+}
+
+static int qcom_wdt_stop(struct watchdog_device *wdd)
+{
+	struct qcom_wdt *wdt = to_qcom_wdt(wdd);
+
+	writel(0, wdt->base + WDT_EN);
+	return 0;
+}
+
+static int qcom_wdt_ping(struct watchdog_device *wdd)
+{
+	struct qcom_wdt *wdt = to_qcom_wdt(wdd);
+
+	writel(1, wdt->base + WDT_RST);
+	return 0;
+}
+
+static int qcom_wdt_set_timeout(struct watchdog_device *wdd,
+				unsigned int timeout)
+{
+	wdd->timeout = timeout;
+	return qcom_wdt_start(wdd);
+}
+
+static const struct watchdog_ops qcom_wdt_ops = {
+	.start		= qcom_wdt_start,
+	.stop		= qcom_wdt_stop,
+	.ping		= qcom_wdt_ping,
+	.set_timeout	= qcom_wdt_set_timeout,
+	.owner		= THIS_MODULE,
+};
+
+static const struct watchdog_info qcom_wdt_info = {
+	.options	= WDIOF_KEEPALIVEPING
+			| WDIOF_MAGICCLOSE
+			| WDIOF_SETTIMEOUT,
+	.identity	= KBUILD_MODNAME,
+};
+
+static int qcom_wdt_probe(struct platform_device *pdev)
+{
+	struct qcom_wdt *wdt;
+	struct resource *res;
+	unsigned long freq;
+	int ret;
+
+	wdt = devm_kzalloc(&pdev->dev, sizeof(*wdt), GFP_KERNEL);
+	if (!wdt)
+		return -ENOMEM;
+
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	wdt->base = devm_ioremap_resource(&pdev->dev, res);
+	if (IS_ERR(wdt->base))
+		return PTR_ERR(wdt->base);
+
+	wdt->clk = devm_clk_get(&pdev->dev, NULL);
+	if (IS_ERR(wdt->clk))
+		return PTR_ERR(wdt->clk);
+
+	ret = clk_prepare_enable(wdt->clk);
+	if (ret) {
+		dev_err(&pdev->dev, "failed to setup clock\n");
+		return ret;
+	}

drivers/watchdog/qcom-wdt.c line 123

Quoted 10 lines
drivers/watchdog/qcom-wdt.c new file +10 -0
+
+	/*
+	 * We use the clock rate to calculate the max timeout, so ensure it's
+	 * not zero to avoid a divide-by-zero exception.
+	 */
+	freq = clk_get_rate(wdt->clk);
+	if (freq == 0) {
+		dev_err(&pdev->dev, "invalid clock rate\n");
+		return -EINVAL;
+	}

This will need clk_disable_unprepare().

Since you are reading the frequency here, it might make sense to store it in struct qcom_wdt so you don't have to call clk_get_rate() again in the start function.

drivers/watchdog/qcom-wdt.c line 129

Quoted 6 lines
drivers/watchdog/qcom-wdt.c new file +6 -0
+
+	wdt->wdd.dev = &pdev->dev;
+	wdt->wdd.info = &qcom_wdt_info;
+	wdt->wdd.ops = &qcom_wdt_ops;
+	wdt->wdd.min_timeout = 1;
+	wdt->wdd.max_timeout = 0x10000000U / freq;

What if the frequency turns out to be larger than 8947848 Hz ? Then your maximum timeout is below the default timeout. And if the frequency is larger than 268435456 Hz, the maximum timeout would be 0.

drivers/watchdog/qcom-wdt.c line 136

Quoted 7 lines
drivers/watchdog/qcom-wdt.c new file +7 -0
+
+	/*
+	 * If 'timeout-sec' unspecified in devicetree, assume a 30 second
+	 * default.
+	 */
+	if (watchdog_init_timeout(&wdt->wdd, 0, &pdev->dev))
+		wdt->wdd.timeout = 30;

You can just initialize timeout above with 30 seconds. Saves you the if statement here.

drivers/watchdog/qcom-wdt.c line 140

Quoted 4 lines
drivers/watchdog/qcom-wdt.c new file +4 -0
+
+	ret = watchdog_register_device(&wdt->wdd);
+	if (ret) {
+		dev_err(&pdev->dev, "failed to register watchdog\n");

This will need a clk_disable_unprepare().

Given that this is needed twice, you might want to consider using error exit code below, as suggested in CodingStyle.

Quoted 44 lines
drivers/watchdog/qcom-wdt.c new file +36 -0
+		return ret;
+	}
+
+	platform_set_drvdata(pdev, wdt);
+	return 0;
+}
+
+static int qcom_wdt_remove(struct platform_device *pdev)
+{
+	struct qcom_wdt *wdt = platform_get_drvdata(pdev);
+
+	watchdog_unregister_device(&wdt->wdd);
+	clk_disable_unprepare(wdt->clk);
+	return 0;
+}
+
+static const struct of_device_id qcom_wdt_of_table[] = {
+	{ .compatible = "qcom,kpss-wdt-msm8960", },
+	{ .compatible = "qcom,kpss-wdt-apq8064", },
+	{ .compatible = "qcom,kpss-wdt-ipq8064", },
+	{ },
+};
+MODULE_DEVICE_TABLE(of, qcom_wdt_of_table);
+
+static struct platform_driver qcom_watchdog_driver = {
+	.probe	= qcom_wdt_probe,
+	.remove	= qcom_wdt_remove,
+	.driver	= {
+		.name		= KBUILD_MODNAME,
+		.of_match_table	= qcom_wdt_of_table,
+	},
+};
+module_platform_driver(qcom_watchdog_driver);
+
+MODULE_DESCRIPTION("QCOM KPSS Watchdog Driver");
+MODULE_LICENSE("GPL v2");
Signature · 7 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation

--
To unsubscribe from this list: send the line "unsubscribe linux-watchdog" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#3 lore raw

Re: [PATCH v2 1/3] watchdog: qcom: add support for KPSS WDT

Quoted 9 lines · Guenter Roeck
Quoted 4 lines · Josh Cartwright

Add a driver for the watchdog timer block found in the Krait Processor Subsystem (KPSS) on the MSM8960, APQ8064, and IPQ8064.

Hi Josh,

looks much better. Couple of comments inline.

Thanks for another review!

[..]

Quoted 19 lines
Quoted 19 lines

+++ b/drivers/watchdog/Kconfig

drivers/watchdog/Kconfig +13 -0
@@ -443,6 +443,19 @@ config TEGRA_WATCHDOG
 	  To compile this driver as a module, choose M here: the
 	  module will be called tegra_wdt.
 
+config QCOM_WDT
+	tristate "QCOM watchdog"
+	depends on HAS_IOMEM
+	depends on ARCH_QCOM
+	select WATCHDOG_CORE
+	help
+	  Say Y here to include Watchdog timer support for the watchdog found
+	  on QCOM chipsets.  Currently supported targets are the MSM8960,
+	  APQ8064, and IPQ8064.
+
+	  To compile this driver as a module, choose M here: the
+	  module will be called qcom_wdt.
+
 # AVR32 Architecture

[..]

Quoted 29 lines
Quoted 29 lines
drivers/watchdog/qcom-wdt.c new file +29 -0
+static int qcom_wdt_probe(struct platform_device *pdev)
+{
+	struct qcom_wdt *wdt;
+	struct resource *res;
+	unsigned long freq;
+	int ret;
+
+	wdt = devm_kzalloc(&pdev->dev, sizeof(*wdt), GFP_KERNEL);
+	if (!wdt)
+		return -ENOMEM;
+
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	wdt->base = devm_ioremap_resource(&pdev->dev, res);
+	if (IS_ERR(wdt->base))
+		return PTR_ERR(wdt->base);
+
+	wdt->clk = devm_clk_get(&pdev->dev, NULL);
+	if (IS_ERR(wdt->clk))
+		return PTR_ERR(wdt->clk);
+
+	ret = clk_prepare_enable(wdt->clk);
+	if (ret) {
+		dev_err(&pdev->dev, "failed to setup clock\n");
+		return ret;
+	}
+
+	/*
+	 * We use the clock rate to calculate the max timeout, so ensure it's
+	 * not zero to avoid a divide-by-zero exception.
Quoted 8 lines

drivers/watchdog/qcom-wdt.c line 123

Quoted 6 lines
drivers/watchdog/qcom-wdt.c new file +6 -0
+	 */
+	freq = clk_get_rate(wdt->clk);
+	if (freq == 0) {
+		dev_err(&pdev->dev, "invalid clock rate\n");
+		return -EINVAL;
+	}

This will need clk_disable_unprepare().

Yep. Nice catch. Will fix.

Quoted 3 lines

Since you are reading the frequency here, it might make sense to store it in struct qcom_wdt so you don't have to call clk_get_rate() again in the start function.

Yeah, it doesn't save much, but I'll go ahead and add it.

Quoted 10 lines

drivers/watchdog/qcom-wdt.c line 129

Quoted 5 lines
drivers/watchdog/qcom-wdt.c new file +5 -0
+	wdt->wdd.dev = &pdev->dev;
+	wdt->wdd.info = &qcom_wdt_info;
+	wdt->wdd.ops = &qcom_wdt_ops;
+	wdt->wdd.min_timeout = 1;
+	wdt->wdd.max_timeout = 0x10000000U / freq;

What if the frequency turns out to be larger than 8947848 Hz ? Then your maximum timeout is below the default timeout. And if the frequency is larger than 268435456 Hz, the maximum timeout would be 0.

Yes, I should be doing more sanity checking. I'll do so.

Quoted 9 lines

drivers/watchdog/qcom-wdt.c line 136

Quoted 6 lines
drivers/watchdog/qcom-wdt.c new file +6 -0
+	/*
+	 * If 'timeout-sec' unspecified in devicetree, assume a 30 second
+	 * default.
+	 */
+	if (watchdog_init_timeout(&wdt->wdd, 0, &pdev->dev))
+		wdt->wdd.timeout = 30;

You can just initialize timeout above with 30 seconds. Saves you the if statement here.

Great. Thanks.

Quoted 8 lines

drivers/watchdog/qcom-wdt.c line 140

Quoted 3 lines
drivers/watchdog/qcom-wdt.c new file +3 -0
+	ret = watchdog_register_device(&wdt->wdd);
+	if (ret) {
+		dev_err(&pdev->dev, "failed to register watchdog\n");

This will need a clk_disable_unprepare().

Given that this is needed twice, you might want to consider using error exit code below, as suggested in CodingStyle.

Indeed. Will do.

Thanks again,
Josh

Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation
#4 lore raw

PATCH v2 2/3 watchdog: qcom: document device tree bindings

In reply to Josh Cartwright

The Qualcomm Krait Processor Sub-system (KPSS) contains one or more instances of the WDT. Provide documentation on how to describe these in the device tree.

---
 .../devicetree/bindings/watchdog/qcom-wdt.txt      | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/watchdog/qcom-wdt.txt
Documentation/devicetree/bindings/watchdog/qcom-wdt.txt new file +22 -0
new file mode 100644
index 0000000..6a1d758
@@ -0,0 +1,22 @@
+Qualcomm Krait Processor Sub-system (KPSS) Watchdog
+---------------------------------------------------
+
+Required properties :
+- compatible : shall contain only one of the following:
+
+			"qcom,kpss-wdt-msm8960"
+			"qcom,kpss-wdt-apq8064"
+			"qcom,kpss-wdt-ipq8064"
+
+- reg : shall contain base register location and length
+- clocks : shall contain the input clock phandle
+- timeout-sec : shall contain the default watchdog timeout in seconds,
+                if unset, the default timeout is 30 seconds
+
+Example:
+	watchdog at 208a038 {
+		compatible = "qcom,kpss-wdt-ipq8064";
+		reg = <0x0208a038 0x40>;
+		clocks = <&sleep_clk>;
+		timeout-sec = <10>;
+	};
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation
#5 lore raw

Re: [PATCH v2 2/3] watchdog: qcom: document device tree bindings

Quoted 1 line · On Tuesday 23 September 2014 18:04:37 Josh Cartwright wrote:

+- clocks : shall contain the input clock phandle

Just nitpicking, but this is not just a phandle, it's a clock descriptor, which is a phandle followed by a set of arguments, which may be empty, depending the on the clock controller implementation.

I would just call it 'the input clock', rather than 'input clock phandle'.

Arnd

#6 lore raw

Re: [PATCH v2 2/3] watchdog: qcom: document device tree bindings

Quoted 7 lines · Arnd Bergmann
Quoted 1 line · On Tuesday 23 September 2014 18:04:37 Josh Cartwright wrote:

+- clocks : shall contain the input clock phandle

Just nitpicking, but this is not just a phandle, it's a clock descriptor, which is a phandle followed by a set of arguments, which may be empty, depending the on the clock controller implementation.

We're _extremely_ inconsistent with terminology, so it's not confusing people don't know what to put here.

Elsewhere people say "phandle + clock specifier" (which is correct), "clock reference" (which is ok, but sounds like a phandle), etc. From what I recall some documentation says some *-specifier properties include the phandle too.

It would be really nice if we had one consistent way of refering to phandle+args style properties.

Quoted 3 lines

I would just call it 'the input clock', rather than 'input clock phandle'.

That's fine by me, given clocks is a well-understood standard property.

Mark.

#7 lore raw

PATCH v2 3/3 watchdog: qcom: register a restart notifier

In reply to Josh Cartwright

The WDT's BITE_TIME warm-reset behavior can be leveraged as a last resort mechanism for triggering chip reset. Usually, other restart methods (such as PS_HOLD) are preferrable for issuing a more complete reset of the chip. As such, keep the priority of the watchdog notifier low.

---
 drivers/watchdog/qcom-wdt.c | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)
drivers/watchdog/qcom-wdt.c +28 -0
index d5e46e2..eba92ef 100644
@@ -16,6 +16,7 @@
 #include <linux/module.h>
 #include <linux/of.h>
 #include <linux/platform_device.h>
+#include <linux/reboot.h>
 #include <linux/watchdog.h>
 
 #define WDT_RST		0x0
@@ -25,6 +26,7 @@
 struct qcom_wdt {
 	struct watchdog_device	wdd;
 	struct clk		*clk;
+	struct notifier_block	restart_nb;
 	void __iomem		*base;
 };
 
@@ -86,6 +88,23 @@ static const struct watchdog_info qcom_wdt_info = {
 	.identity	= KBUILD_MODNAME,
 };
 
+static int qcom_wdt_restart(struct notifier_block *nb, unsigned long action,
+			    void *data)
+{
+	struct qcom_wdt *wdt = container_of(nb, struct qcom_wdt, restart_nb);
+
+	/*
+	 * Trigger watchdog bite:
+	 *    Setup BITE_TIME to be very low, and enable WDT.
+	 *      0x31F3 => 390ms @ 32kHz, also value at reset
+	 */
+	writel(0, wdt->base + WDT_EN);
+	writel(1, wdt->base + WDT_RST);
+	writel(0x31F3, wdt->base + WDT_BITE_TIME);
+	writel(1, wdt->base + WDT_EN);
+	return NOTIFY_DONE;
+}
+
 static int qcom_wdt_probe(struct platform_device *pdev)
 {
 	struct qcom_wdt *wdt;
@@ -141,6 +160,14 @@ static int qcom_wdt_probe(struct platform_device *pdev)
 		return ret;
 	}
 
+	/*
+	 * WDT restart notifier has priority 0 (use as a last resort)
+	 */
+	wdt->restart_nb.notifier_call = qcom_wdt_restart;
+	ret = register_restart_handler(&wdt->restart_nb);
+	if (ret)
+		dev_err(&pdev->dev, "failed to setup restart handler\n");
+
 	platform_set_drvdata(pdev, wdt);
 	return 0;
 }
@@ -149,6 +176,7 @@ static int qcom_wdt_remove(struct platform_device *pdev)
 {
 	struct qcom_wdt *wdt = platform_get_drvdata(pdev);
 
+	unregister_restart_handler(&wdt->restart_nb);
 	watchdog_unregister_device(&wdt->wdd);
 	clk_disable_unprepare(wdt->clk);
 	return 0;
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation
maintainer
#8 lore raw

Re: [PATCH v2 3/3] watchdog: qcom: register a restart notifier

Quoted 34 lines · Josh Cartwright

The WDT's BITE_TIME warm-reset behavior can be leveraged as a last resort mechanism for triggering chip reset. Usually, other restart methods (such as PS_HOLD) are preferrable for issuing a more complete reset of the chip. As such, keep the priority of the watchdog notifier low.

---
 drivers/watchdog/qcom-wdt.c | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)
drivers/watchdog/qcom-wdt.c +2 -0
index d5e46e2..eba92ef 100644
@@ -16,6 +16,7 @@
 #include <linux/module.h>
 #include <linux/of.h>
 #include <linux/platform_device.h>
+#include <linux/reboot.h>
 #include <linux/watchdog.h>
 
 #define WDT_RST		0x0
@@ -25,6 +26,7 @@
 struct qcom_wdt {
 	struct watchdog_device	wdd;
 	struct clk		*clk;
+	struct notifier_block	restart_nb;
 	void __iomem		*base;
 };
 
@@ -86,6 +88,23 @@ static const struct watchdog_info qcom_wdt_info = {
 	.identity	= KBUILD_MODNAME,
 };

drivers/watchdog/qcom-wdt.c static const struct watchdog_info qcom_wdt_info = { · line 99

Quoted 10 lines
drivers/watchdog/qcom-wdt.c +9 -0
 
+static int qcom_wdt_restart(struct notifier_block *nb, unsigned long action,
+			    void *data)
+{
+	struct qcom_wdt *wdt = container_of(nb, struct qcom_wdt, restart_nb);
+
+	/*
+	 * Trigger watchdog bite:
+	 *    Setup BITE_TIME to be very low, and enable WDT.
+	 *      0x31F3 => 390ms @ 32kHz, also value at reset

What happens if you set it to a lower value ? If you set it to, say, 0x100, would the timeout be ~8ms ? Wouldn't that make more sense ? Or at least something like 0x800, which should make it around 64 ms, or 0x400 for 32 ms.

[ In other words, I don't really see the point for a 390 ms wait here ]

drivers/watchdog/qcom-wdt.c static const struct watchdog_info qcom_wdt_info = { · line 104

Quoted 5 lines
drivers/watchdog/qcom-wdt.c +5 -0
+	 */
+	writel(0, wdt->base + WDT_EN);
+	writel(1, wdt->base + WDT_RST);
+	writel(0x31F3, wdt->base + WDT_BITE_TIME);
+	writel(1, wdt->base + WDT_EN);

You should have some mdelay() here, larger than the expected timeout.

Thanks,
Guenter

Quoted 37 lines
drivers/watchdog/qcom-wdt.c +12 -0
+	return NOTIFY_DONE;
+}
+
 static int qcom_wdt_probe(struct platform_device *pdev)
 {
 	struct qcom_wdt *wdt;
@@ -141,6 +160,14 @@ static int qcom_wdt_probe(struct platform_device *pdev)
 		return ret;
 	}
 
+	/*
+	 * WDT restart notifier has priority 0 (use as a last resort)
+	 */
+	wdt->restart_nb.notifier_call = qcom_wdt_restart;
+	ret = register_restart_handler(&wdt->restart_nb);
+	if (ret)
+		dev_err(&pdev->dev, "failed to setup restart handler\n");
+
 	platform_set_drvdata(pdev, wdt);
 	return 0;
 }
@@ -149,6 +176,7 @@ static int qcom_wdt_remove(struct platform_device *pdev)
 {
 	struct qcom_wdt *wdt = platform_get_drvdata(pdev);
 
+	unregister_restart_handler(&wdt->restart_nb);
 	watchdog_unregister_device(&wdt->wdd);
 	clk_disable_unprepare(wdt->clk);
 	return 0;
Signature · 7 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation

--
To unsubscribe from this list: send the line "unsubscribe linux-watchdog" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html