Re: [PATCH v3 2/2] clocksource: Add Low Power STM32 timers driver

Hi Benjamin,

On Tue, Jan 28, 2020 at 01:07:02PM +0100, Benjamin Gaignard wrote:

From: Benjamin Gaignard <redacted>

Implement clock event driver using low power STM32 timers.
Low power timers counter s running even in when CPU is in stop mode.

Can you fix the sentence above?

It could be used as clock event broadcaster to wake up CPUs but not like
a clocksource because each it rise an interrupt the counter restart from 0.

Please give a description of the timer (eg. e37e45934afed).
 

(silent option)

+	depends on OF && ARM && (MACH_STM32MP157 || COMPILE_TEST)

Do you really want to make this option? This timer belongs to a mfd, so if the
mfd is enabled it should enable the timer also, no ?

Please check the headers above, all are not needed IMO.

+#define LPTIM_ICR		0x04
+#define ICR_ARRMCF		BIT(1)
+
+#define LPTIM_IER		0x08
+#define IER_ARRMIE		BIT(1)
+
+#define LPTIM_CFGR		0x0C
+#define CFGR_PSC_OFFSET		9
+
+#define LPTIM_CR		0x10
+#define CR_ENABLE		BIT(0)
+#define CR_SNGSTRT		BIT(1)
+#define CR_CNTSTRT		BIT(2)
+
+#define LPTIM_CMP		0x14
+#define LPTIM_ARR		0x18
+
+#define STM32_LP_BITS		16
+#define STM32_LP_RATING		400
+#define STM32_CLKRATE		(32000 * HZ)
+#define TIMER_MAX_VAL		(BIT(STM32_LP_BITS) - 1)

STM32 prefix missing

+#define STM32_LP_MAX_PSC	7
+
+struct stm32_lp_private {
+	struct regmap *reg;
+	struct clk *clk;
+	struct clock_event_device clkevt;
+	unsigned long rate;

This field can be saved if the parameter is passed to the set_prescalar
function.

+	unsigned long period;
+};
+
+static struct stm32_lp_private*
+to_priv(struct clock_event_device *clkevt)
+{
+	return container_of(clkevt, struct stm32_lp_private, clkevt);
+}
+
+static int stm32_clkevent_lp_shutdown(struct clock_event_device *clkevt)
+{
+	struct stm32_lp_private *priv = to_priv(clkevt);
+
+	regmap_write(priv->reg, LPTIM_CR, 0);
+	regmap_write(priv->reg, LPTIM_IER, 0);
+	/* clear pending flags */
+	regmap_write(priv->reg, LPTIM_ICR, ICR_ARRMCF);
+
+	return 0;
+}
+
+static int stm32_clkevent_lp_set_timer(unsigned long evt,
+				       struct clock_event_device *clkevt,
+				       int is_periodic)
+{
+	struct stm32_lp_private *priv = to_priv(clkevt);
+
+	/* disable TIMER */
+	regmap_write(priv->reg, LPTIM_CR, 0);
+	/* enable ARR interrupt */
+	regmap_write(priv->reg, LPTIM_IER, IER_ARRMIE);
+	/* enable LPTIMER*/
+	regmap_write(priv->reg, LPTIM_CR, CR_ENABLE);
+	/* set nex event counter */
+	regmap_write(priv->reg, LPTIM_ARR, evt);
+	/* reset counter */
+	regmap_write(priv->reg, LPTIM_CMP, 0);
+
+	/* start counter */
+	if (is_periodic)
+		regmap_write(priv->reg, LPTIM_CR, CR_CNTSTRT | CR_ENABLE);
+	else
+		regmap_write(priv->reg, LPTIM_CR, CR_SNGSTRT | CR_ENABLE);

Please merge all these calls to regmap_write into a single one to reduce
contention.

+	return 0;
+}
+
+static int stm32_clkevent_lp_set_next_event(unsigned long evt,
+					    struct clock_event_device *clkevt)
+{
+	return stm32_clkevent_lp_set_timer(evt, clkevt,
+					   clockevent_state_periodic(clkevt));
+}
+
+static int stm32_clkevent_lp_set_periodic(struct clock_event_device *clkevt)
+{
+	struct stm32_lp_private *priv = to_priv(clkevt);
+
+	return stm32_clkevent_lp_set_timer(priv->period, clkevt, true);
+}
+
+static int stm32_clkevent_lp_set_oneshot(struct clock_event_device *clkevt)
+{
+	struct stm32_lp_private *priv = to_priv(clkevt);
+
+	return stm32_clkevent_lp_set_timer(priv->period, clkevt, false);
+}
+
+static irqreturn_t stm32_clkevent_lp_irq_handler(int irq, void *dev_id)
+{
+	struct clock_event_device *clkevt = (struct clock_event_device *)dev_id;
+	struct stm32_lp_private *priv = to_priv(clkevt);
+
+	regmap_write(priv->reg, LPTIM_ICR, ICR_ARRMCF);
+
+	clkevt->event_handler(clkevt);
+
+	return IRQ_HANDLED;
+}
+
+/**

This is a kerneldoc format

+ * stm32_clkevent_lp_set_prescaler - Compute and set the prescaler register
+ * @priv: a pointer to a timer-of structure
+ *
+ * Compute the prescaler to always target a 32Khz timer rate
+ */

Given these are static callbacks, you can drop the function description by
moving it to a comment in the function. 

+static void stm32_clkevent_lp_set_prescaler(struct stm32_lp_private *priv)
+{
+	int i;
+
+	for (i = 0; i <= STM32_LP_MAX_PSC; i++) {
+		if (DIV_ROUND_CLOSEST(priv->rate, 1 << i) < STM32_CLKRATE)
+			break;
+	}
+
+	regmap_write(priv->reg, LPTIM_CFGR, i << CFGR_PSC_OFFSET);
+
+	/* Adjust rate and period given the prescaler value */
+	priv->rate = DIV_ROUND_CLOSEST(priv->rate, (1 << i));
+	priv->period = DIV_ROUND_UP(priv->rate, HZ);
+}
+
+static int stm32_clkevent_lp_init(struct stm32_lp_private *priv,
+				  struct device_node *np)
+{
+	int ret;
+
+	priv->clkevt.name = np->full_name;
+	priv->clkevt.cpumask = cpu_possible_mask;
+	priv->clkevt.features = CLOCK_EVT_FEAT_PERIODIC |
+				CLOCK_EVT_FEAT_ONESHOT;
+	priv->clkevt.set_state_shutdown = stm32_clkevent_lp_shutdown;
+	priv->clkevt.set_state_periodic = stm32_clkevent_lp_set_periodic;
+	priv->clkevt.set_state_oneshot = stm32_clkevent_lp_set_oneshot;
+	priv->clkevt.set_next_event = stm32_clkevent_lp_set_next_event;
+	priv->clkevt.rating = STM32_LP_RATING;
+
+	priv->clkevt.irq = irq_of_parse_and_map(np, 0);
+	if (!priv->clkevt.irq)
+		return -EINVAL;
+
+	ret = request_irq(priv->clkevt.irq, stm32_clkevent_lp_irq_handler,
+			  IRQF_TIMER, np->full_name, &priv->clkevt);
+	if (ret)
+		return ret;

May be move the node parsing to the 'probe' function and use demv_request_irq
(or use timer-of)?

Pass the timer name as parameter to this function?

+
+	clockevents_config_and_register(&priv->clkevt, priv->rate, 0x1,
+					TIMER_MAX_VAL);
+
+	return 0;
+}
+
+static int stm32_clkevent_lp_probe(struct platform_device *pdev)
+{
+	struct stm32_lptimer *ddata = dev_get_drvdata(pdev->dev.parent);
+	struct stm32_lp_private *priv;
+	int ret;
+
+	priv = devm_kzalloc(&pdev->dev, sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+
+	priv->reg = ddata->regmap;
+	priv->clk = ddata->clk;
+	ret = clk_prepare_enable(priv->clk);
+	if (ret)
+		return -EINVAL;
+
+	priv->rate = clk_get_rate(priv->clk);
+	if (!priv->rate)
+		goto out_clk_disable;
+
+	stm32_clkevent_lp_set_prescaler(priv);
+
+	ret = stm32_clkevent_lp_init(priv, pdev->dev.parent->of_node);
+	if (ret)
+		goto out_clk_disable;
+
+	platform_set_drvdata(pdev, priv);

See if this is needed given the comment below about the removal.

+	return 0;
+
+out_clk_disable:
+	clk_disable_unprepare(priv->clk);
+
+	return -EINVAL;
+}
+
+static int stm32_clkevent_lp_remove(struct platform_device *pdev)
+{
+	struct stm32_lp_private *priv = platform_get_drvdata(pdev);
+
+	clk_disable_unprepare(priv->clk);
+	free_irq(priv->clkevt.irq, &priv->clkevt);
+
+	return 0;
+}

This is broken IMO, the time framework will continue considering the previously
registered clockevent as available. You should consider converting the mfd to a
bool instead of a tristate (and remove the dead code).

+static const struct of_device_id stm32_clkevent_lp_of_match[] = {
+	{ .compatible = "st,stm32-clkevent-lp", },
+	{},
+};
+MODULE_DEVICE_TABLE(of, stm32_clkevent_lp_of_match);
+
+static struct platform_driver stm32_clkevent_lp_driver = {
+	.probe	= stm32_clkevent_lp_probe,
+	.remove = stm32_clkevent_lp_remove,
+	.driver	= {
+		.name = "stm32-clkevent-lp",
+		.of_match_table = of_match_ptr(stm32_clkevent_lp_of_match),
+	},
+};
+module_platform_driver(stm32_clkevent_lp_driver);
+
+MODULE_ALIAS("platform:stm32-clkevent-lp");
+MODULE_DESCRIPTION("STMicroelectronics STM32 clockevent LP driver");
+MODULE_LICENSE("GPL v2");
-- 
2.15.0

-- 

 <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