Skip to contentSkip to conversation

NVIDIA Tegra devfreq improvements and Tegra20/30 support

PATCH Superseded Dmitry Osipenko 2 participants 54 messages Apr 15, 2019 – Apr 17, 2019

Subsystems (2)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v2 cover NVIDIA Tegra devfreq improvements and Tegra20/30 support

Hello,

I tried to utilize the Tegra devfreq driver on Tegra30 and found out that it doesn't work properly due to improper Hz<->kHz conversions made by the driver. After fixing that problem and doing some more testing I noticed that there are things that could be improved and in result here is this patchset that fixes the problems, makes some improvements and adds support for NVIDIA Tegra30 SoC's. This series was tested on Tegra30 and Tegra124 machines.

This series also adds support for the Tegra20 SoC's that do not have the ACTMON hardware unit, but can use Memory Controller counters to make decision about the required memory frequency. Please note that the patch that instantiates a platform device for the driver isn't a part of this series, I'll submit it later on because of interdependency with some other ongoing patchset.

Please review, thanks.

Changelog:

v2: The patchset was quite heavily reworked since v1, few patches we
dropped or squashed into the new ones and more patches we added. In a result more bugs and potential problems are fixed now, driver's code got more clean up.

The Tegra20 driver-addition patch is now a part of this series, it has no changes since v1.

Dmitry Osipenko (19):
PM / devfreq: tegra: Fix kHz to Hz conversion
PM / devfreq: tegra: Replace readl-writel with relaxed versions PM / devfreq: tegra: Don't ignore clk errors
PM / devfreq: tegra: Don't set EMC clock rate to maximum on probe PM / devfreq: tegra: Replace write memory barrier with the read
barrier
PM / devfreq: tegra: Fix missed error checking on devfreq
initialization failure
PM / devfreq: tegra: Register clk notifier in the end of driver's
probe
PM / devfreq: tegra: Remove OPP entries on driver removal PM / devfreq: tegra: Change interrupt request order
PM / devfreq: tegra: Drop primary interrupt handler
PM / devfreq: tegra: De-initialize properly on driver's probe error PM / devfreq: tegra: Avoid inconsistency of current frequency value PM / devfreq: tegra: Mark ACTMON's governor as immutable PM / devfreq: tegra: Move governor registration to driver's probe PM / devfreq: tegra: Synchronize IRQ after masking it in hardware PM / devfreq: tegra: Reconfigure hardware on governor's restart PM / devfreq: tegra: Support Tegra30
PM / devfreq: tegra: Enable COMPILE_TEST for the driver PM / devfreq: Introduce driver for NVIDIA Tegra20

 MAINTAINERS                       |   8 +
 drivers/devfreq/Kconfig           |  15 +-
 drivers/devfreq/Makefile          |   1 +
 drivers/devfreq/tegra-devfreq.c   | 296 +++++++++++++-----------------
 drivers/devfreq/tegra20-devfreq.c | 177 ++++++++++++++++++
 5 files changed, 324 insertions(+), 173 deletions(-)
 create mode 100644 drivers/devfreq/tegra20-devfreq.c
Signature · 1 line
2.21.0
#1 lore raw

PATCH v2 1/19 PM / devfreq: tegra: Fix kHz to Hz conversion

The kHz to Hz is incorrectly converted in a few places in the code, this results in a wrong frequency being calculated because devfreq core uses OPP frequencies that are given in Hz to clamp the rate, while tegra-devfreq gives to the core value in kHz and then it also expects to receive value in kHz from the core. In a result memory freq is always set to a value which is close to ULONG_MAX because of the bug. Hence the EMC frequency is always capped to the maximum and the driver doesn't do anything useful. This patch was tested on Tegra30 and Tegra124 SoC's, EMC frequency scaling works properly now.

---
 drivers/devfreq/tegra-devfreq.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
drivers/devfreq/tegra-devfreq.c +4 -6
index c89ba7b834ff..02905978abe1 100644
@@ -486,9 +486,9 @@ static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 {
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct dev_pm_opp *opp;
-	unsigned long rate = *freq * KHZ;
+	unsigned long rate;
 
-	opp = devfreq_recommended_opp(dev, &rate, flags);
+	opp = devfreq_recommended_opp(dev, freq, flags);
 	if (IS_ERR(opp)) {
 		dev_err(dev, "Failed to find opp for %lu KHz\n", *freq);
 		return PTR_ERR(opp);
@@ -499,8 +499,6 @@ static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 	clk_set_min_rate(tegra->emc_clock, rate);
 	clk_set_rate(tegra->emc_clock, 0);
 
-	*freq = rate;
-
 	return 0;
 }
 
@@ -510,7 +508,7 @@ static int tegra_devfreq_get_dev_status(struct device *dev,
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct tegra_devfreq_device *actmon_dev;
 
-	stat->current_frequency = tegra->cur_freq;
+	stat->current_frequency = tegra->cur_freq * KHZ;
 
 	/* To be used by the tegra governor */
 	stat->private_data = tegra;
@@ -565,7 +563,7 @@ static int tegra_governor_get_target(struct devfreq *devfreq,
 		target_freq = max(target_freq, dev->target_freq);
 	}
 
-	*freq = target_freq;
+	*freq = target_freq * KHZ;
 
 	return 0;
 }
Signature · 1 line
2.21.0
#4 lore raw

PATCH v2 2/19 PM / devfreq: tegra: Replace readl-writel with relaxed versions

In reply to Dmitry Osipenko

There is no need to insert memory barrier on each readl/writel invocation, hence use the relaxed versions.

---
 drivers/devfreq/tegra-devfreq.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/devfreq/tegra-devfreq.c +4 -4
index 02905978abe1..0c0909fba545 100644
@@ -191,23 +191,23 @@ static struct tegra_actmon_emc_ratio actmon_emc_ratios[] = {
 
 static u32 actmon_readl(struct tegra_devfreq *tegra, u32 offset)
 {
-	return readl(tegra->regs + offset);
+	return readl_relaxed(tegra->regs + offset);
 }
 
 static void actmon_writel(struct tegra_devfreq *tegra, u32 val, u32 offset)
 {
-	writel(val, tegra->regs + offset);
+	writel_relaxed(val, tegra->regs + offset);
 }
 
 static u32 device_readl(struct tegra_devfreq_device *dev, u32 offset)
 {
-	return readl(dev->regs + offset);
+	return readl_relaxed(dev->regs + offset);
 }
 
 static void device_writel(struct tegra_devfreq_device *dev, u32 val,
 			  u32 offset)
 {
-	writel(val, dev->regs + offset);
+	writel_relaxed(val, dev->regs + offset);
 }
 
 static unsigned long do_percent(unsigned long val, unsigned int pct)
Signature · 1 line
2.21.0
#6 lore raw

PATCH v2 3/19 PM / devfreq: tegra: Don't ignore clk errors

In reply to Dmitry Osipenko

The clk_set_min_rate() could fail and in this case clk_set_rate() sets rate to 0, which may drop EMC rate to minimum and make machine very difficult to use.

---
 drivers/devfreq/tegra-devfreq.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +8 -2
index 0c0909fba545..69581c9082d4 100644
@@ -487,6 +487,7 @@ static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct dev_pm_opp *opp;
 	unsigned long rate;
+	int err;
 
 	opp = devfreq_recommended_opp(dev, freq, flags);
 	if (IS_ERR(opp)) {
@@ -496,8 +497,13 @@ static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 	rate = dev_pm_opp_get_freq(opp);
 	dev_pm_opp_put(opp);
 
-	clk_set_min_rate(tegra->emc_clock, rate);
-	clk_set_rate(tegra->emc_clock, 0);
+	err = clk_set_min_rate(tegra->emc_clock, rate);
+	if (err)
+		return err;
+
+	err = clk_set_rate(tegra->emc_clock, 0);
+	if (err)
+		return err;
 
 	return 0;
 }
Signature · 1 line
2.21.0
#9 lore raw

PATCH v2 4/19 PM / devfreq: tegra: Don't set EMC clock rate to maximum on probe

In reply to Dmitry Osipenko

There is no real benefit from doing so, hence let's drop that rate setting for consistency.

---
 drivers/devfreq/tegra-devfreq.c | 2 --
 1 file changed, 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +0 -2
index 69581c9082d4..d62fb1b0d9bb 100644
@@ -648,8 +648,6 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		return PTR_ERR(tegra->emc_clock);
 	}
 
-	clk_set_rate(tegra->emc_clock, ULONG_MAX);
-
 	tegra->rate_change_nb.notifier_call = tegra_actmon_rate_notify_cb;
 	err = clk_notifier_register(tegra->emc_clock, &tegra->rate_change_nb);
 	if (err) {
Signature · 1 line
2.21.0
#12 lore raw

PATCH v2 5/19 PM / devfreq: tegra: Replace write memory barrier with the read barrier

In reply to Dmitry Osipenko

The write memory barrier isn't needed because the BUS buffer is flushed by read after write that happens after the removed wmb(), we will also use readl() instead of the relaxed version to ensure that read is indeed completed.

---
 drivers/devfreq/tegra-devfreq.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +1 -2
index d62fb1b0d9bb..f0f0d78f6cbf 100644
@@ -243,8 +243,7 @@ static void tegra_devfreq_update_wmark(struct tegra_devfreq *tegra,
 static void actmon_write_barrier(struct tegra_devfreq *tegra)
 {
 	/* ensure the update has reached the ACTMON */
-	wmb();
-	actmon_readl(tegra, ACTMON_GLB_STATUS);
+	readl(tegra->regs + ACTMON_GLB_STATUS);
 }
 
 static void actmon_isr_device(struct tegra_devfreq *tegra,
Signature · 1 line
2.21.0
#16 lore raw

PATCH v2 6/19 PM / devfreq: tegra: Fix missed error checking on devfreq initialization failure

In reply to Dmitry Osipenko

The code doesn't check for devfreq initialization failure, fix it.

---
 drivers/devfreq/tegra-devfreq.c | 4 ++++
 1 file changed, 4 insertions(+)
drivers/devfreq/tegra-devfreq.c +4 -0
index f0f0d78f6cbf..aa0478464b35 100644
@@ -707,6 +707,10 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 						 &tegra_devfreq_profile,
 						 "tegra_actmon",
 						 NULL);
+	if (IS_ERR(tegra->devfreq)) {
+		err = PTR_ERR(tegra->devfreq);
+		return err;
+	}
 
 	return 0;
 }
Signature · 1 line
2.21.0
#20 lore raw

PATCH v2 7/19 PM / devfreq: tegra: Register clk notifier in the end of driver's probe

In reply to Dmitry Osipenko

The notifier isn't unregistered if something fails after the registration, move it to the end of probe to fix the potential use-after-free problem.

---
 drivers/devfreq/tegra-devfreq.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
drivers/devfreq/tegra-devfreq.c +8 -8
index aa0478464b35..c248c18431d9 100644
@@ -647,14 +647,6 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		return PTR_ERR(tegra->emc_clock);
 	}
 
-	tegra->rate_change_nb.notifier_call = tegra_actmon_rate_notify_cb;
-	err = clk_notifier_register(tegra->emc_clock, &tegra->rate_change_nb);
-	if (err) {
-		dev_err(&pdev->dev,
-			"Failed to register rate change notifier\n");
-		return err;
-	}
-
 	reset_control_assert(tegra->reset);
 
 	err = clk_prepare_enable(tegra->clock);
@@ -712,6 +704,14 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		return err;
 	}
 
+	tegra->rate_change_nb.notifier_call = tegra_actmon_rate_notify_cb;
+	err = clk_notifier_register(tegra->emc_clock, &tegra->rate_change_nb);
+	if (err) {
+		dev_err(&pdev->dev,
+			"Failed to register rate change notifier\n");
+		return err;
+	}
+
 	return 0;
 }
 
Signature · 1 line
2.21.0
#21 lore raw

PATCH v2 8/19 PM / devfreq: tegra: Remove OPP entries on driver removal

In reply to Dmitry Osipenko

This fixes "_opp_is_duplicate" warning messages on driver's module reload.

---
 drivers/devfreq/tegra-devfreq.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +10 -2
index c248c18431d9..98ba27b1e2c4 100644
@@ -701,7 +701,7 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 						 NULL);
 	if (IS_ERR(tegra->devfreq)) {
 		err = PTR_ERR(tegra->devfreq);
-		return err;
+		goto remove_opps;
 	}
 
 	tegra->rate_change_nb.notifier_call = tegra_actmon_rate_notify_cb;
@@ -709,10 +709,15 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 	if (err) {
 		dev_err(&pdev->dev,
 			"Failed to register rate change notifier\n");
-		return err;
+		goto remove_opps;
 	}
 
 	return 0;
+
+remove_opps:
+	dev_pm_opp_remove_all_dynamic(&pdev->dev);
+
+	return err;
 }
 
 static int tegra_devfreq_remove(struct platform_device *pdev)
@@ -722,6 +727,9 @@ static int tegra_devfreq_remove(struct platform_device *pdev)
 	u32 val;
 	unsigned int i;
 
+	devm_devfreq_remove_device(&pdev->dev, tegra->devfreq);
+	dev_pm_opp_remove_all_dynamic(&pdev->dev);
+
 	for (i = 0; i < ARRAY_SIZE(actmon_device_configs); i++) {
 		val = device_readl(&tegra->devices[i], ACTMON_DEV_CTRL);
 		val &= ~ACTMON_DEV_CTRL_ENB;
Signature · 1 line
2.21.0
#22 lore raw

PATCH v2 9/19 PM / devfreq: tegra: Change interrupt request order

In reply to Dmitry Osipenko

The interrupt should be requested after devfreq initialization because interrupt handler uses the devfreq, otherwise interrupt handler will crash if interrupt fires before the initialization. This shouldn't happen in practice, but it's more correct to have interrupt available after all of interrupt prerequisites are ready.

---
 drivers/devfreq/tegra-devfreq.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
drivers/devfreq/tegra-devfreq.c +8 -8
index 98ba27b1e2c4..2a1464098200 100644
@@ -686,14 +686,6 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 
 	platform_set_drvdata(pdev, tegra);
 
-	err = devm_request_threaded_irq(&pdev->dev, irq, actmon_isr,
-					actmon_thread_isr, IRQF_SHARED,
-					"tegra-devfreq", tegra);
-	if (err) {
-		dev_err(&pdev->dev, "Interrupt request failed\n");
-		return err;
-	}
-
 	tegra_devfreq_profile.initial_freq = clk_get_rate(tegra->emc_clock);
 	tegra->devfreq = devm_devfreq_add_device(&pdev->dev,
 						 &tegra_devfreq_profile,
@@ -701,6 +693,14 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 						 NULL);
 	if (IS_ERR(tegra->devfreq)) {
 		err = PTR_ERR(tegra->devfreq);
+		return err;
+	}
+
+	err = devm_request_threaded_irq(&pdev->dev, irq, actmon_isr,
+					actmon_thread_isr, IRQF_SHARED,
+					"tegra-devfreq", tegra);
+	if (err) {
+		dev_err(&pdev->dev, "Interrupt request failed\n");
 		goto remove_opps;
 	}
 
Signature · 1 line
2.21.0
#23 lore raw

PATCH v2 10/19 PM / devfreq: tegra: Drop primary interrupt handler

In reply to Dmitry Osipenko

There is no real need in the primary interrupt handler, hence move everything to the secondary (threaded) handler. In a result locking is consistent now and there are no potential races with the interrupt handler because it is protected with the devfreq's mutex.

---
 drivers/devfreq/tegra-devfreq.c | 61 ++++++++++++---------------------
 1 file changed, 22 insertions(+), 39 deletions(-)
drivers/devfreq/tegra-devfreq.c +22 -39
index 2a1464098200..69b557df5084 100644
@@ -144,7 +144,6 @@ static struct tegra_devfreq_device_config actmon_device_configs[] = {
 struct tegra_devfreq_device {
 	const struct tegra_devfreq_device_config *config;
 	void __iomem *regs;
-	spinlock_t lock;
 
 	/* Average event count sampled in the last interrupt */
 	u32 avg_count;
@@ -249,11 +248,8 @@ static void actmon_write_barrier(struct tegra_devfreq *tegra)
 static void actmon_isr_device(struct tegra_devfreq *tegra,
 			      struct tegra_devfreq_device *dev)
 {
-	unsigned long flags;
 	u32 intr_status, dev_ctrl;
 
-	spin_lock_irqsave(&dev->lock, flags);
-
 	dev->avg_count = device_readl(dev, ACTMON_DEV_AVG_COUNT);
 	tegra_devfreq_update_avg_wmark(tegra, dev);
 
@@ -302,26 +298,6 @@ static void actmon_isr_device(struct tegra_devfreq *tegra,
 	device_writel(dev, ACTMON_INTR_STATUS_CLEAR, ACTMON_DEV_INTR_STATUS);
 
 	actmon_write_barrier(tegra);
-
-	spin_unlock_irqrestore(&dev->lock, flags);
-}
-
-static irqreturn_t actmon_isr(int irq, void *data)
-{
-	struct tegra_devfreq *tegra = data;
-	bool handled = false;
-	unsigned int i;
-	u32 val;
-
-	val = actmon_readl(tegra, ACTMON_GLB_STATUS);
-	for (i = 0; i < ARRAY_SIZE(tegra->devices); i++) {
-		if (val & tegra->devices[i].config->irq_mask) {
-			actmon_isr_device(tegra, tegra->devices + i);
-			handled = true;
-		}
-	}
-
-	return handled ? IRQ_WAKE_THREAD : IRQ_NONE;
 }
 
 static unsigned long actmon_cpu_to_emc_rate(struct tegra_devfreq *tegra,
@@ -348,35 +324,46 @@ static void actmon_update_target(struct tegra_devfreq *tegra,
 	unsigned long cpu_freq = 0;
 	unsigned long static_cpu_emc_freq = 0;
 	unsigned int avg_sustain_coef;
-	unsigned long flags;
+	u32 avg_count;
 
 	if (dev->config->avg_dependency_threshold) {
 		cpu_freq = cpufreq_get(0);
 		static_cpu_emc_freq = actmon_cpu_to_emc_rate(tegra, cpu_freq);
 	}
 
-	spin_lock_irqsave(&dev->lock, flags);
-
-	dev->target_freq = dev->avg_count / ACTMON_SAMPLING_PERIOD;
+	avg_count = dev->avg_count;
+	dev->target_freq = avg_count / ACTMON_SAMPLING_PERIOD;
 	avg_sustain_coef = 100 * 100 / dev->config->boost_up_threshold;
 	dev->target_freq = do_percent(dev->target_freq, avg_sustain_coef);
 	dev->target_freq += dev->boost_freq;
 
-	if (dev->avg_count >= dev->config->avg_dependency_threshold)
+	if (avg_count >= dev->config->avg_dependency_threshold)
 		dev->target_freq = max(dev->target_freq, static_cpu_emc_freq);
-
-	spin_unlock_irqrestore(&dev->lock, flags);
 }
 
 static irqreturn_t actmon_thread_isr(int irq, void *data)
 {
 	struct tegra_devfreq *tegra = data;
+	bool handled = false;
+	unsigned int i;
+	u32 val;
 
 	mutex_lock(&tegra->devfreq->lock);
-	update_devfreq(tegra->devfreq);
+
+	val = actmon_readl(tegra, ACTMON_GLB_STATUS);
+	for (i = 0; i < ARRAY_SIZE(tegra->devices); i++) {
+		if (val & tegra->devices[i].config->irq_mask) {
+			actmon_isr_device(tegra, tegra->devices + i);
+			handled = true;
+		}
+	}
+
+	if (handled)
+		update_devfreq(tegra->devfreq);
+
 	mutex_unlock(&tegra->devfreq->lock);
 
-	return IRQ_HANDLED;
+	return handled ? IRQ_HANDLED : IRQ_NONE;
 }
 
 static int tegra_actmon_rate_notify_cb(struct notifier_block *nb,
@@ -386,7 +373,6 @@ static int tegra_actmon_rate_notify_cb(struct notifier_block *nb,
 	struct tegra_devfreq *tegra;
 	struct tegra_devfreq_device *dev;
 	unsigned int i;
-	unsigned long flags;
 
 	if (action != POST_RATE_CHANGE)
 		return NOTIFY_OK;
@@ -398,9 +384,7 @@ static int tegra_actmon_rate_notify_cb(struct notifier_block *nb,
 	for (i = 0; i < ARRAY_SIZE(tegra->devices); i++) {
 		dev = &tegra->devices[i];
 
-		spin_lock_irqsave(&dev->lock, flags);
 		tegra_devfreq_update_wmark(tegra, dev);
-		spin_unlock_irqrestore(&dev->lock, flags);
 	}
 
 	actmon_write_barrier(tegra);
@@ -668,7 +652,6 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		dev = tegra->devices + i;
 		dev->config = actmon_device_configs + i;
 		dev->regs = tegra->regs + dev->config->offset;
-		spin_lock_init(&dev->lock);
 
 		tegra_actmon_configure_device(tegra, dev);
 	}
@@ -696,8 +679,8 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		return err;
 	}
 
-	err = devm_request_threaded_irq(&pdev->dev, irq, actmon_isr,
-					actmon_thread_isr, IRQF_SHARED,
+	err = devm_request_threaded_irq(&pdev->dev, irq, NULL,
+					actmon_thread_isr, IRQF_ONESHOT,
 					"tegra-devfreq", tegra);
 	if (err) {
 		dev_err(&pdev->dev, "Interrupt request failed\n");
Signature · 1 line
2.21.0
#27 lore raw

PATCH v2 11/19 PM / devfreq: tegra: De-initialize properly on driver's probe error

In reply to Dmitry Osipenko

Reset hardware, disable ACTMON clock, release OPP's and free IRQ before removing devfreq device since there is no guarantee that interrupt handling won't run after masking interrupt in hardware.

---
 drivers/devfreq/tegra-devfreq.c | 53 ++++++++++++++++-----------------
 1 file changed, 26 insertions(+), 27 deletions(-)
drivers/devfreq/tegra-devfreq.c +26 -27
index 69b557df5084..a668e4fbc874 100644
@@ -663,28 +663,28 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 
 	irq = platform_get_irq(pdev, 0);
 	if (irq < 0) {
-		dev_err(&pdev->dev, "Failed to get IRQ: %d\n", irq);
-		return irq;
+		err = irq;
+		dev_err(&pdev->dev, "Failed to get IRQ: %d\n", err);
+		goto remove_opps;
 	}
 
 	platform_set_drvdata(pdev, tegra);
 
 	tegra_devfreq_profile.initial_freq = clk_get_rate(tegra->emc_clock);
-	tegra->devfreq = devm_devfreq_add_device(&pdev->dev,
-						 &tegra_devfreq_profile,
-						 "tegra_actmon",
-						 NULL);
+	tegra->devfreq = devfreq_add_device(&pdev->dev,
+					    &tegra_devfreq_profile,
+					    "tegra_actmon",
+					    NULL);
 	if (IS_ERR(tegra->devfreq)) {
 		err = PTR_ERR(tegra->devfreq);
-		return err;
+		goto remove_opps;
 	}
 
-	err = devm_request_threaded_irq(&pdev->dev, irq, NULL,
-					actmon_thread_isr, IRQF_ONESHOT,
-					"tegra-devfreq", tegra);
+	err = request_threaded_irq(irq, NULL, actmon_thread_isr, IRQF_ONESHOT,
+				   "tegra-devfreq", tegra);
 	if (err) {
 		dev_err(&pdev->dev, "Interrupt request failed\n");
-		goto remove_opps;
+		goto remove_devfreq;
 	}
 
 	tegra->rate_change_nb.notifier_call = tegra_actmon_rate_notify_cb;
@@ -692,14 +692,23 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 	if (err) {
 		dev_err(&pdev->dev,
 			"Failed to register rate change notifier\n");
-		goto remove_opps;
+		goto disable_interrupt;
 	}
 
 	return 0;
 
+disable_interrupt:
+	free_irq(irq, tegra);
+
+remove_devfreq:
+	devfreq_remove_device(tegra->devfreq);
+
 remove_opps:
 	dev_pm_opp_remove_all_dynamic(&pdev->dev);
 
+	reset_control_reset(tegra->reset);
+	clk_disable_unprepare(tegra->clock);
+
 	return err;
 }
 
@@ -707,24 +716,14 @@ static int tegra_devfreq_remove(struct platform_device *pdev)
 {
 	struct tegra_devfreq *tegra = platform_get_drvdata(pdev);
 	int irq = platform_get_irq(pdev, 0);
-	u32 val;
-	unsigned int i;
-
-	devm_devfreq_remove_device(&pdev->dev, tegra->devfreq);
-	dev_pm_opp_remove_all_dynamic(&pdev->dev);
-
-	for (i = 0; i < ARRAY_SIZE(actmon_device_configs); i++) {
-		val = device_readl(&tegra->devices[i], ACTMON_DEV_CTRL);
-		val &= ~ACTMON_DEV_CTRL_ENB;
-		device_writel(&tegra->devices[i], val, ACTMON_DEV_CTRL);
-	}
-
-	actmon_write_barrier(tegra);
-
-	devm_free_irq(&pdev->dev, irq, tegra);
 
 	clk_notifier_unregister(tegra->emc_clock, &tegra->rate_change_nb);
+	free_irq(irq, tegra);
+
+	devfreq_remove_device(tegra->devfreq);
+	dev_pm_opp_remove_all_dynamic(&pdev->dev);
 
+	reset_control_reset(tegra->reset);
 	clk_disable_unprepare(tegra->clock);
 
 	return 0;
Signature · 1 line
2.21.0
#29 lore raw

PATCH v2 12/19 PM / devfreq: tegra: Avoid inconsistency of current frequency value

In reply to Dmitry Osipenko

The frequency value potentially could change in-between. It doesn't cause any real problem at all right now, but that could change in the future. Hence let's avoid the inconsistency.

---
 drivers/devfreq/tegra-devfreq.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +4 -2
index a668e4fbc874..f1a6f951813a 100644
@@ -496,13 +496,15 @@ static int tegra_devfreq_get_dev_status(struct device *dev,
 {
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct tegra_devfreq_device *actmon_dev;
+	unsigned long cur_freq;
 
-	stat->current_frequency = tegra->cur_freq * KHZ;
+	cur_freq = READ_ONCE(tegra->cur_freq);
 
 	/* To be used by the tegra governor */
 	stat->private_data = tegra;
 
 	/* The below are to be used by the other governors */
+	stat->current_frequency = cur_freq * KHZ;
 
 	actmon_dev = &tegra->devices[MCALL];
 
@@ -513,7 +515,7 @@ static int tegra_devfreq_get_dev_status(struct device *dev,
 	stat->busy_time *= 100 / BUS_SATURATION_RATIO;
 
 	/* Number of cycles in a sampling period */
-	stat->total_time = ACTMON_SAMPLING_PERIOD * tegra->cur_freq;
+	stat->total_time = ACTMON_SAMPLING_PERIOD * cur_freq;
 
 	stat->busy_time = min(stat->busy_time, stat->total_time);
 
Signature · 1 line
2.21.0
#33 lore raw

PATCH v2 13/19 PM / devfreq: tegra: Mark ACTMON's governor as immutable

In reply to Dmitry Osipenko

The ACTMON's governor supports only the Tegra's devfreq device and there is no need to use any other governor, hence let's mark Tegra governor as immutable to permanently stick it with Tegra's devfreq device.

---
 drivers/devfreq/Kconfig         | 1 -
 drivers/devfreq/tegra-devfreq.c | 1 +
 2 files changed, 1 insertion(+), 1 deletion(-)
drivers/devfreq/Kconfig +0 -1
index 6a172d338f6d..a78dffe603c1 100644
@@ -94,7 +94,6 @@ config ARM_EXYNOS_BUS_DEVFREQ
 config ARM_TEGRA_DEVFREQ
 	tristate "Tegra DEVFREQ Driver"
 	depends on ARCH_TEGRA_124_SOC
-	select DEVFREQ_GOV_SIMPLE_ONDEMAND
 	select PM_OPP
 	help
 	  This adds the DEVFREQ driver for the Tegra family of SoCs.
drivers/devfreq/tegra-devfreq.c +1 -0
index f1a6f951813a..832e4f5aa11b 100644
@@ -593,6 +593,7 @@ static struct devfreq_governor tegra_devfreq_governor = {
 	.name = "tegra_actmon",
 	.get_target_freq = tegra_governor_get_target,
 	.event_handler = tegra_governor_event_handler,
+	.immutable = true,
 };
 
 static int tegra_devfreq_probe(struct platform_device *pdev)
Signature · 1 line
2.21.0
#35 lore raw

PATCH v2 14/19 PM / devfreq: tegra: Move governor registration to driver's probe

In reply to Dmitry Osipenko

There is no need to register the ACTMON's governor separately from the driver, hence let's move the registration into the driver's probe function for consistency and to make code cleaner a tad.

---
 drivers/devfreq/tegra-devfreq.c | 44 +++++++++------------------------
 1 file changed, 12 insertions(+), 32 deletions(-)
drivers/devfreq/tegra-devfreq.c +12 -32
index 832e4f5aa11b..46c61af8ca33 100644
@@ -673,6 +673,12 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 
 	platform_set_drvdata(pdev, tegra);
 
+	err = devfreq_add_governor(&tegra_devfreq_governor);
+	if (err) {
+		dev_err(&pdev->dev, "Failed to add governor: %d\n", err);
+		goto remove_opps;
+	}
+
 	tegra_devfreq_profile.initial_freq = clk_get_rate(tegra->emc_clock);
 	tegra->devfreq = devfreq_add_device(&pdev->dev,
 					    &tegra_devfreq_profile,
@@ -680,7 +686,7 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 					    NULL);
 	if (IS_ERR(tegra->devfreq)) {
 		err = PTR_ERR(tegra->devfreq);
-		goto remove_opps;
+		goto remove_governor;
 	}
 
 	err = request_threaded_irq(irq, NULL, actmon_thread_isr, IRQF_ONESHOT,
@@ -706,6 +712,9 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 remove_devfreq:
 	devfreq_remove_device(tegra->devfreq);
 
+remove_governor:
+	devfreq_remove_governor(&tegra_devfreq_governor);
+
 remove_opps:
 	dev_pm_opp_remove_all_dynamic(&pdev->dev);
 
@@ -729,7 +738,7 @@ static int tegra_devfreq_remove(struct platform_device *pdev)
 	reset_control_reset(tegra->reset);
 	clk_disable_unprepare(tegra->clock);
 
-	return 0;
+	return devfreq_remove_governor(&tegra_devfreq_governor);
 }
 
 static const struct of_device_id tegra_devfreq_of_match[] = {
@@ -747,36 +756,7 @@ static struct platform_driver tegra_devfreq_driver = {
 		.of_match_table = tegra_devfreq_of_match,
 	},
 };
-
-static int __init tegra_devfreq_init(void)
-{
-	int ret = 0;
-
-	ret = devfreq_add_governor(&tegra_devfreq_governor);
-	if (ret) {
-		pr_err("%s: failed to add governor: %d\n", __func__, ret);
-		return ret;
-	}
-
-	ret = platform_driver_register(&tegra_devfreq_driver);
-	if (ret)
-		devfreq_remove_governor(&tegra_devfreq_governor);
-
-	return ret;
-}
-module_init(tegra_devfreq_init)
-
-static void __exit tegra_devfreq_exit(void)
-{
-	int ret = 0;
-
-	platform_driver_unregister(&tegra_devfreq_driver);
-
-	ret = devfreq_remove_governor(&tegra_devfreq_governor);
-	if (ret)
-		pr_err("%s: failed to remove governor: %d\n", __func__, ret);
-}
-module_exit(tegra_devfreq_exit)
+module_platform_driver(tegra_devfreq_driver);
 
 MODULE_LICENSE("GPL v2");
 MODULE_DESCRIPTION("Tegra devfreq driver");
Signature · 1 line
2.21.0
#37 lore raw

PATCH v2 15/19 PM / devfreq: tegra: Synchronize IRQ after masking it in hardware

In reply to Dmitry Osipenko

There is no guarantee that interrupt handling isn't running in parallel, hence it is necessary to synchronize IRQ in order to ensure that interrupt is indeed disabled.

---
 drivers/devfreq/tegra-devfreq.c | 33 ++++++++++++++++-----------------
 1 file changed, 16 insertions(+), 17 deletions(-)
drivers/devfreq/tegra-devfreq.c +16 -17
index 46c61af8ca33..cfbfbe28e7bf 100644
@@ -171,6 +171,8 @@ struct tegra_devfreq {
 	struct notifier_block	rate_change_nb;
 
 	struct tegra_devfreq_device devices[ARRAY_SIZE(actmon_device_configs)];
+
+	int irq;
 };
 
 struct tegra_actmon_emc_ratio {
@@ -432,6 +434,8 @@ static void tegra_actmon_disable_interrupts(struct tegra_devfreq *tegra)
 	}
 
 	actmon_write_barrier(tegra);
+
+	synchronize_irq(tegra->irq);
 }
 
 static void tegra_actmon_configure_device(struct tegra_devfreq *tegra,
@@ -603,7 +607,6 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 	struct resource *res;
 	unsigned int i;
 	unsigned long rate;
-	int irq;
 	int err;
 
 	tegra = devm_kzalloc(&pdev->dev, sizeof(*tegra), GFP_KERNEL);
@@ -659,18 +662,18 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		tegra_actmon_configure_device(tegra, dev);
 	}
 
+	tegra->irq = platform_get_irq(pdev, 0);
+	if (tegra->irq < 0) {
+		err = tegra->irq;
+		dev_err(&pdev->dev, "Failed to get IRQ: %d\n", err);
+		return err;
+	}
+
 	for (rate = 0; rate <= tegra->max_freq * KHZ; rate++) {
 		rate = clk_round_rate(tegra->emc_clock, rate);
 		dev_pm_opp_add(&pdev->dev, rate, 0);
 	}
 
-	irq = platform_get_irq(pdev, 0);
-	if (irq < 0) {
-		err = irq;
-		dev_err(&pdev->dev, "Failed to get IRQ: %d\n", err);
-		goto remove_opps;
-	}
-
 	platform_set_drvdata(pdev, tegra);
 
 	err = devfreq_add_governor(&tegra_devfreq_governor);
@@ -689,10 +692,11 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		goto remove_governor;
 	}
 
-	err = request_threaded_irq(irq, NULL, actmon_thread_isr, IRQF_ONESHOT,
-				   "tegra-devfreq", tegra);
+	err = devm_request_threaded_irq(&pdev->dev, tegra->irq, NULL,
+					actmon_thread_isr, IRQF_ONESHOT,
+					"tegra-devfreq", tegra);
 	if (err) {
-		dev_err(&pdev->dev, "Interrupt request failed\n");
+		dev_err(&pdev->dev, "Interrupt request failed: %d\n", err);
 		goto remove_devfreq;
 	}
 
@@ -701,14 +705,11 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 	if (err) {
 		dev_err(&pdev->dev,
 			"Failed to register rate change notifier\n");
-		goto disable_interrupt;
+		goto remove_devfreq;
 	}
 
 	return 0;
 
-disable_interrupt:
-	free_irq(irq, tegra);
-
 remove_devfreq:
 	devfreq_remove_device(tegra->devfreq);
 
@@ -727,10 +728,8 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 static int tegra_devfreq_remove(struct platform_device *pdev)
 {
 	struct tegra_devfreq *tegra = platform_get_drvdata(pdev);
-	int irq = platform_get_irq(pdev, 0);
 
 	clk_notifier_unregister(tegra->emc_clock, &tegra->rate_change_nb);
-	free_irq(irq, tegra);
 
 	devfreq_remove_device(tegra->devfreq);
 	dev_pm_opp_remove_all_dynamic(&pdev->dev);
Signature · 1 line
2.21.0
#40 lore raw

PATCH v2 16/19 PM / devfreq: tegra: Reconfigure hardware on governor's restart

In reply to Dmitry Osipenko

Move hardware configuration to governor's start/resume methods. This allows to re-initialize hardware counters and reconfigure cleanly if governor was stopped/paused. That is needed because we are not aware of all hardware changes that happened while governor was stopped and the paused state may get out of sync with reality, hence it's better to start with a clean slate after the pause. In a result there is no memory bandwidth starvation after resume from suspend-to-ram that results in display controller underflowing that happens on resume because of improper decision made by devfreq about the required memory frequency. This change also cleans up code a tad by moving hardware-configuration code into a single location.

---
 drivers/devfreq/tegra-devfreq.c | 84 ++++++++++++---------------------
 1 file changed, 31 insertions(+), 53 deletions(-)
drivers/devfreq/tegra-devfreq.c +31 -53
index cfbfbe28e7bf..f0711d5ad27d 100644
@@ -394,50 +394,6 @@ static int tegra_actmon_rate_notify_cb(struct notifier_block *nb,
 	return NOTIFY_OK;
 }
 
-static void tegra_actmon_enable_interrupts(struct tegra_devfreq *tegra)
-{
-	struct tegra_devfreq_device *dev;
-	u32 val;
-	unsigned int i;
-
-	for (i = 0; i < ARRAY_SIZE(tegra->devices); i++) {
-		dev = &tegra->devices[i];
-
-		val = device_readl(dev, ACTMON_DEV_CTRL);
-		val |= ACTMON_DEV_CTRL_AVG_ABOVE_WMARK_EN;
-		val |= ACTMON_DEV_CTRL_AVG_BELOW_WMARK_EN;
-		val |= ACTMON_DEV_CTRL_CONSECUTIVE_BELOW_WMARK_EN;
-		val |= ACTMON_DEV_CTRL_CONSECUTIVE_ABOVE_WMARK_EN;
-
-		device_writel(dev, val, ACTMON_DEV_CTRL);
-	}
-
-	actmon_write_barrier(tegra);
-}
-
-static void tegra_actmon_disable_interrupts(struct tegra_devfreq *tegra)
-{
-	struct tegra_devfreq_device *dev;
-	u32 val;
-	unsigned int i;
-
-	for (i = 0; i < ARRAY_SIZE(tegra->devices); i++) {
-		dev = &tegra->devices[i];
-
-		val = device_readl(dev, ACTMON_DEV_CTRL);
-		val &= ~ACTMON_DEV_CTRL_AVG_ABOVE_WMARK_EN;
-		val &= ~ACTMON_DEV_CTRL_AVG_BELOW_WMARK_EN;
-		val &= ~ACTMON_DEV_CTRL_CONSECUTIVE_BELOW_WMARK_EN;
-		val &= ~ACTMON_DEV_CTRL_CONSECUTIVE_ABOVE_WMARK_EN;
-
-		device_writel(dev, val, ACTMON_DEV_CTRL);
-	}
-
-	actmon_write_barrier(tegra);
-
-	synchronize_irq(tegra->irq);
-}
-
 static void tegra_actmon_configure_device(struct tegra_devfreq *tegra,
 					  struct tegra_devfreq_device *dev)
 {
@@ -461,13 +417,40 @@ static void tegra_actmon_configure_device(struct tegra_devfreq *tegra,
 		<< ACTMON_DEV_CTRL_CONSECUTIVE_BELOW_WMARK_NUM_SHIFT;
 	val |= (ACTMON_ABOVE_WMARK_WINDOW - 1)
 		<< ACTMON_DEV_CTRL_CONSECUTIVE_ABOVE_WMARK_NUM_SHIFT;
+	val |= ACTMON_DEV_CTRL_AVG_ABOVE_WMARK_EN;
+	val |= ACTMON_DEV_CTRL_AVG_BELOW_WMARK_EN;
+	val |= ACTMON_DEV_CTRL_CONSECUTIVE_BELOW_WMARK_EN;
+	val |= ACTMON_DEV_CTRL_CONSECUTIVE_ABOVE_WMARK_EN;
 	val |= ACTMON_DEV_CTRL_ENB;
 
 	device_writel(dev, val, ACTMON_DEV_CTRL);
+}
+
+static void tegra_actmon_start(struct tegra_devfreq *tegra)
+{
+	unsigned int i;
+
+	actmon_writel(tegra, ACTMON_SAMPLING_PERIOD - 1,
+		      ACTMON_GLB_PERIOD_CTRL);
+
+	for (i = 0; i < ARRAY_SIZE(tegra->devices); i++)
+		tegra_actmon_configure_device(tegra, &tegra->devices[i]);
 
 	actmon_write_barrier(tegra);
 }
 
+static void tegra_actmon_stop(struct tegra_devfreq *tegra)
+{
+	unsigned int i;
+
+	for (i = 0; i < ARRAY_SIZE(tegra->devices); i++)
+		device_writel(&tegra->devices[i], 0x00000000, ACTMON_DEV_CTRL);
+
+	actmon_write_barrier(tegra);
+
+	synchronize_irq(tegra->irq);
+}
+
 static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 				u32 flags)
 {
@@ -571,22 +554,22 @@ static int tegra_governor_event_handler(struct devfreq *devfreq,
 	switch (event) {
 	case DEVFREQ_GOV_START:
 		devfreq_monitor_start(devfreq);
-		tegra_actmon_enable_interrupts(tegra);
+		tegra_actmon_start(tegra);
 		break;
 
 	case DEVFREQ_GOV_STOP:
-		tegra_actmon_disable_interrupts(tegra);
+		tegra_actmon_stop(tegra);
 		devfreq_monitor_stop(devfreq);
 		break;
 
 	case DEVFREQ_GOV_SUSPEND:
-		tegra_actmon_disable_interrupts(tegra);
+		tegra_actmon_stop(tegra);
 		devfreq_monitor_suspend(devfreq);
 		break;
 
 	case DEVFREQ_GOV_RESUME:
 		devfreq_monitor_resume(devfreq);
-		tegra_actmon_enable_interrupts(tegra);
+		tegra_actmon_start(tegra);
 		break;
 	}
 
@@ -651,15 +634,10 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 	tegra->max_freq = clk_round_rate(tegra->emc_clock, ULONG_MAX) / KHZ;
 	tegra->cur_freq = clk_get_rate(tegra->emc_clock) / KHZ;
 
-	actmon_writel(tegra, ACTMON_SAMPLING_PERIOD - 1,
-		      ACTMON_GLB_PERIOD_CTRL);
-
 	for (i = 0; i < ARRAY_SIZE(actmon_device_configs); i++) {
 		dev = tegra->devices + i;
 		dev->config = actmon_device_configs + i;
 		dev->regs = tegra->regs + dev->config->offset;
-
-		tegra_actmon_configure_device(tegra, dev);
 	}
 
 	tegra->irq = platform_get_irq(pdev, 0);
Signature · 1 line
2.21.0
#41 lore raw

PATCH v2 17/19 PM / devfreq: tegra: Support Tegra30

In reply to Dmitry Osipenko

The devfreq driver can be used on Tegra30 without any code change and it works perfectly fine, the default Tegra124 parameters are good enough for Tegra30.

---
 drivers/devfreq/Kconfig         | 4 ++--
 drivers/devfreq/tegra-devfreq.c | 1 +
 2 files changed, 3 insertions(+), 2 deletions(-)
drivers/devfreq/Kconfig +2 -2
index a78dffe603c1..78c33ddd4eea 100644
@@ -92,8 +92,8 @@ config ARM_EXYNOS_BUS_DEVFREQ
 	  This does not yet operate with optimal voltages.
 
 config ARM_TEGRA_DEVFREQ
-	tristate "Tegra DEVFREQ Driver"
-	depends on ARCH_TEGRA_124_SOC
+	tristate "NVIDIA Tegra30+ DEVFREQ Driver"
+	depends on ARCH_TEGRA
 	select PM_OPP
 	help
 	  This adds the DEVFREQ driver for the Tegra family of SoCs.
drivers/devfreq/tegra-devfreq.c +1 -0
index f0711d5ad27d..0a2465a58cf5 100644
@@ -719,6 +719,7 @@ static int tegra_devfreq_remove(struct platform_device *pdev)
 }
 
 static const struct of_device_id tegra_devfreq_of_match[] = {
+	{ .compatible = "nvidia,tegra30-actmon" },
 	{ .compatible = "nvidia,tegra124-actmon" },
 	{ },
 };
Signature · 1 line
2.21.0
#45 lore raw

PATCH v2 18/19 PM / devfreq: tegra: Enable COMPILE_TEST for the driver

In reply to Dmitry Osipenko

The driver's compilation doesn't have any specific dependencies, hence the COMPILE_TEST option can be supported in Kconfig.

---
 drivers/devfreq/Kconfig | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/devfreq/Kconfig +1 -1
index 78c33ddd4eea..bd6652863e7d 100644
@@ -93,7 +93,7 @@ config ARM_EXYNOS_BUS_DEVFREQ
 
 config ARM_TEGRA_DEVFREQ
 	tristate "NVIDIA Tegra30+ DEVFREQ Driver"
-	depends on ARCH_TEGRA
+	depends on ARCH_TEGRA || COMPILE_TEST
 	select PM_OPP
 	help
 	  This adds the DEVFREQ driver for the Tegra family of SoCs.
Signature · 1 line
2.21.0
#49 lore raw

PATCH v2 19/19 PM / devfreq: Introduce driver for NVIDIA Tegra20

In reply to Dmitry Osipenko

Add devfreq driver for NVIDIA Tegra20 SoC's. The driver periodically reads out Memory Controller counters and adjusts memory frequency based on the memory clients activity.

---
 MAINTAINERS                       |   8 ++
 drivers/devfreq/Kconfig           |  10 ++
 drivers/devfreq/Makefile          |   1 +
 drivers/devfreq/tegra20-devfreq.c | 177 ++++++++++++++++++++++++++++++
 4 files changed, 196 insertions(+)
 create mode 100644 drivers/devfreq/tegra20-devfreq.c
MAINTAINERS +8 -0
index 80b59db1b6e4..91f475ec4545 100644
@@ -10056,6 +10056,14 @@ F:	include/linux/memblock.h
 F:	mm/memblock.c
 F:	Documentation/core-api/boot-time-mm.rst
 
+MEMORY FREQUENCY SCALING DRIVER FOR NVIDIA TEGRA20
+M:	Dmitry Osipenko <[email protected]>
+L:	[email protected]
+L:	[email protected]
+T:	git git://git.kernel.org/pub/scm/linux/kernel/git/mzx/devfreq.git
+S:	Maintained
+F:	drivers/devfreq/tegra20-devfreq.c
+
 MEMORY MANAGEMENT
 L:	[email protected]
 W:	http://www.linux-mm.org
drivers/devfreq/Kconfig +10 -0
index bd6652863e7d..af4c86c4e0f6 100644
@@ -100,6 +100,16 @@ config ARM_TEGRA_DEVFREQ
 	  It reads ACTMON counters of memory controllers and adjusts the
 	  operating frequencies and voltages with OPP support.
 
+config ARM_TEGRA20_DEVFREQ
+	tristate "NVIDIA Tegra20 DEVFREQ Driver"
+	depends on (TEGRA_MC && TEGRA20_EMC) || COMPILE_TEST
+	select DEVFREQ_GOV_SIMPLE_ONDEMAND
+	select PM_OPP
+	help
+	  This adds the DEVFREQ driver for the Tegra20 family of SoCs.
+	  It reads Memory Controller counters and adjusts the operating
+	  frequencies and voltages with OPP support.
+
 config ARM_RK3399_DMC_DEVFREQ
 	tristate "ARM RK3399 DMC DEVFREQ Driver"
 	depends on ARCH_ROCKCHIP
drivers/devfreq/Makefile +1 -0
index 32b8d4d3f12c..6fcc5596b8b7 100644
@@ -11,6 +11,7 @@ obj-$(CONFIG_DEVFREQ_GOV_PASSIVE)	+= governor_passive.o
 obj-$(CONFIG_ARM_EXYNOS_BUS_DEVFREQ)	+= exynos-bus.o
 obj-$(CONFIG_ARM_RK3399_DMC_DEVFREQ)	+= rk3399_dmc.o
 obj-$(CONFIG_ARM_TEGRA_DEVFREQ)		+= tegra-devfreq.o
+obj-$(CONFIG_ARM_TEGRA20_DEVFREQ)	+= tegra20-devfreq.o
 
 # DEVFREQ Event Drivers
 obj-$(CONFIG_PM_DEVFREQ_EVENT)		+= event/
drivers/devfreq/tegra20-devfreq.c new file +177 -0

177 lines not on this page · raw message

Signature · 1 line
2.21.0
maintainer Reviewed-by on v2 1/19
#2 lore raw

Re: [PATCH v2 01/19] PM / devfreq: tegra: Fix kHz to Hz conversion

In reply to Dmitry Osipenko

Hi,

I add one minor comment (KHZ -> hz).

Quoted 22 lines · Dmitry Osipenko

The kHz to Hz is incorrectly converted in a few places in the code, this results in a wrong frequency being calculated because devfreq core uses OPP frequencies that are given in Hz to clamp the rate, while tegra-devfreq gives to the core value in kHz and then it also expects to receive value in kHz from the core. In a result memory freq is always set to a value which is close to ULONG_MAX because of the bug. Hence the EMC frequency is always capped to the maximum and the driver doesn't do anything useful. This patch was tested on Tegra30 and Tegra124 SoC's, EMC frequency scaling works properly now.

---
 drivers/devfreq/tegra-devfreq.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
drivers/devfreq/tegra-devfreq.c +0 -0
index c89ba7b834ff..02905978abe1 100644

drivers/devfreq/tegra-devfreq.c static int tegra_devfreq_target(struct device *dev, unsigned long *freq, · line 493

Quoted 10 lines
drivers/devfreq/tegra-devfreq.c +2 -2
@@ -486,9 +486,9 @@ static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 {
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct dev_pm_opp *opp;
-	unsigned long rate = *freq * KHZ;
+	unsigned long rate;
 
-	opp = devfreq_recommended_opp(dev, &rate, flags);
+	opp = devfreq_recommended_opp(dev, freq, flags);
 	if (IS_ERR(opp)) {
 		dev_err(dev, "Failed to find opp for %lu KHz\n", *freq);

Need to change it as following:
- KHz -> Hz

Quoted 21 lines
drivers/devfreq/tegra-devfreq.c +1 -3
 		return PTR_ERR(opp);
@@ -499,8 +499,6 @@ static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 	clk_set_min_rate(tegra->emc_clock, rate);
 	clk_set_rate(tegra->emc_clock, 0);
 
-	*freq = rate;
-
 	return 0;
 }
 
@@ -510,7 +508,7 @@ static int tegra_devfreq_get_dev_status(struct device *dev,
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct tegra_devfreq_device *actmon_dev;
 
-	stat->current_frequency = tegra->cur_freq;
+	stat->current_frequency = tegra->cur_freq * KHZ;
 
 	/* To be used by the tegra governor */
 	stat->private_data = tegra;

drivers/devfreq/tegra-devfreq.c static int tegra_governor_get_target(struct devfreq *devfreq, · line 569

Quoted 8 lines
drivers/devfreq/tegra-devfreq.c +1 -1
@@ -565,7 +563,7 @@ static int tegra_governor_get_target(struct devfreq *devfreq,
 		target_freq = max(target_freq, dev->target_freq);
 	}
 
-	*freq = target_freq;
+	*freq = target_freq * KHZ;
 
 	return 0;
 }

It looks good to me if modify it in accordance with my comment.

Regards,
Chanwoo Choi

Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer
#7 lore raw

Re: [PATCH v2 03/19] PM / devfreq: tegra: Don't ignore clk errors

In reply to Dmitry Osipenko

Hi,

Quoted 23 lines · Dmitry Osipenko

The clk_set_min_rate() could fail and in this case clk_set_rate() sets rate to 0, which may drop EMC rate to minimum and make machine very difficult to use.

---
 drivers/devfreq/tegra-devfreq.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +1 -0
index 0c0909fba545..69581c9082d4 100644
@@ -487,6 +487,7 @@ static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct dev_pm_opp *opp;
 	unsigned long rate;
+	int err;
 
 	opp = devfreq_recommended_opp(dev, freq, flags);
 	if (IS_ERR(opp)) {
@@ -496,8 +497,13 @@ static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 	rate = dev_pm_opp_get_freq(opp);

drivers/devfreq/tegra-devfreq.c static int tegra_devfreq_target(struct device *dev, unsigned long *freq, · line 505

Quoted 10 lines
drivers/devfreq/tegra-devfreq.c +6 -2
 	dev_pm_opp_put(opp);
 
-	clk_set_min_rate(tegra->emc_clock, rate);
-	clk_set_rate(tegra->emc_clock, 0);
+	err = clk_set_min_rate(tegra->emc_clock, rate);
+	if (err)
+		return err;
+
+	err = clk_set_rate(tegra->emc_clock, 0);
+	if (err)

Why don't you restore the previous minimum clock for tegra->emc_clock when failed to execute 'clk_set_rate(tegra->emc_clock, 0);'?

I think that if you control the restoring the previous minimum clock, it is more nice to handle the error handling.

Quoted 5 lines
drivers/devfreq/tegra-devfreq.c +1 -0
+		return err;
 
 	return 0;
 }
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer Reviewed-by on v2 4/19
#10 lore raw

Re: [PATCH v2 04/19] PM / devfreq: tegra: Don't set EMC clock rate to maximum on probe

In reply to Dmitry Osipenko

Hi,

Quoted 13 lines · Dmitry Osipenko

There is no real benefit from doing so, hence let's drop that rate setting for consistency.

---
 drivers/devfreq/tegra-devfreq.c | 2 --
 1 file changed, 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +0 -0
index 69581c9082d4..d62fb1b0d9bb 100644

drivers/devfreq/tegra-devfreq.c static int tegra_devfreq_probe(struct platform_device *pdev) · line 652

Quoted 5 lines
drivers/devfreq/tegra-devfreq.c +0 -2
@@ -648,8 +648,6 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		return PTR_ERR(tegra->emc_clock);
 	}
 
-	clk_set_rate(tegra->emc_clock, ULONG_MAX);
-

It seems like that initialize the emc_clock as the supported maximum clock. But, if the rate for emc_clock initialized by either bootloader or clock driver and it's well working until this code, actually, it is not necessary.

Looks good to me.

Quoted 4 lines
drivers/devfreq/tegra-devfreq.c +0 -0
 	tegra->rate_change_nb.notifier_call = tegra_actmon_rate_notify_cb;
 	err = clk_notifier_register(tegra->emc_clock, &tegra->rate_change_nb);
 	if (err) {
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer
#17 lore raw

Re: [PATCH v2 06/19] PM / devfreq: tegra: Fix missed error checking on devfreq initialization failure

In reply to Dmitry Osipenko

Hi,

patch6/7/8/9 are for handling of exception handling in probe() function. Actually, I'm not sure that there are special reason to split out the patches. I think that you can squash patch6/7/8/9 to only one patch.

Also, even if patch6/7/8/9 handle the exception handling in probe(), the tegra_devfreq_probe() doesn't support the restoring sequence when fail happen. I think that if you want to fix the fail case of probe(), please add the restoring sequence about following function.
- clk_disable_unprepare()
- clk_notifier_unregister()
- dev_pm_opp_remove()

Quoted 23 lines · Dmitry Osipenko

The code doesn't check for devfreq initialization failure, fix it.

---
 drivers/devfreq/tegra-devfreq.c | 4 ++++
 1 file changed, 4 insertions(+)
drivers/devfreq/tegra-devfreq.c +4 -0
index f0f0d78f6cbf..aa0478464b35 100644
@@ -707,6 +707,10 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 						 &tegra_devfreq_profile,
 						 "tegra_actmon",
 						 NULL);
+	if (IS_ERR(tegra->devfreq)) {
+		err = PTR_ERR(tegra->devfreq);
+		return err;
+	}
 
 	return 0;
 }
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer
#24 lore raw

Re: [PATCH v2 10/19] PM / devfreq: tegra: Drop primary interrupt handler

In reply to Dmitry Osipenko

Hi,

It looks good to me to drop the primary interrupt handler but I have some comments. Please check it.

Quoted 68 lines · Dmitry Osipenko

There is no real need in the primary interrupt handler, hence move everything to the secondary (threaded) handler. In a result locking is consistent now and there are no potential races with the interrupt handler because it is protected with the devfreq's mutex.

---
 drivers/devfreq/tegra-devfreq.c | 61 ++++++++++++---------------------
 1 file changed, 22 insertions(+), 39 deletions(-)
drivers/devfreq/tegra-devfreq.c +1 -25
index 2a1464098200..69b557df5084 100644
@@ -144,7 +144,6 @@ static struct tegra_devfreq_device_config actmon_device_configs[] = {
 struct tegra_devfreq_device {
 	const struct tegra_devfreq_device_config *config;
 	void __iomem *regs;
-	spinlock_t lock;
 
 	/* Average event count sampled in the last interrupt */
 	u32 avg_count;
@@ -249,11 +248,8 @@ static void actmon_write_barrier(struct tegra_devfreq *tegra)
 static void actmon_isr_device(struct tegra_devfreq *tegra,
 			      struct tegra_devfreq_device *dev)
 {
-	unsigned long flags;
 	u32 intr_status, dev_ctrl;
 
-	spin_lock_irqsave(&dev->lock, flags);
-
 	dev->avg_count = device_readl(dev, ACTMON_DEV_AVG_COUNT);
 	tegra_devfreq_update_avg_wmark(tegra, dev);
 
@@ -302,26 +298,6 @@ static void actmon_isr_device(struct tegra_devfreq *tegra,
 	device_writel(dev, ACTMON_INTR_STATUS_CLEAR, ACTMON_DEV_INTR_STATUS);
 
 	actmon_write_barrier(tegra);
-
-	spin_unlock_irqrestore(&dev->lock, flags);
-}
-
-static irqreturn_t actmon_isr(int irq, void *data)
-{
-	struct tegra_devfreq *tegra = data;
-	bool handled = false;
-	unsigned int i;
-	u32 val;
-
-	val = actmon_readl(tegra, ACTMON_GLB_STATUS);
-	for (i = 0; i < ARRAY_SIZE(tegra->devices); i++) {
-		if (val & tegra->devices[i].config->irq_mask) {
-			actmon_isr_device(tegra, tegra->devices + i);
-			handled = true;
-		}
-	}
-
-	return handled ? IRQ_WAKE_THREAD : IRQ_NONE;
 }
 
 static unsigned long actmon_cpu_to_emc_rate(struct tegra_devfreq *tegra,
@@ -348,35 +324,46 @@ static void actmon_update_target(struct tegra_devfreq *tegra,
 	unsigned long cpu_freq = 0;
 	unsigned long static_cpu_emc_freq = 0;
 	unsigned int avg_sustain_coef;
-	unsigned long flags;
+	u32 avg_count;
 

drivers/devfreq/tegra-devfreq.c static void actmon_update_target(struct tegra_devfreq *tegra, · line 335

Quoted 10 lines
drivers/devfreq/tegra-devfreq.c +2 -3
 	if (dev->config->avg_dependency_threshold) {
 		cpu_freq = cpufreq_get(0);
 		static_cpu_emc_freq = actmon_cpu_to_emc_rate(tegra, cpu_freq);
 	}
 
-	spin_lock_irqsave(&dev->lock, flags);
-
-	dev->target_freq = dev->avg_count / ACTMON_SAMPLING_PERIOD;
+	avg_count = dev->avg_count;
+	dev->target_freq = avg_count / ACTMON_SAMPLING_PERIOD;

Actually, this change is not related to this patch. Please keep the original code.

drivers/devfreq/tegra-devfreq.c static void actmon_update_target(struct tegra_devfreq *tegra, · line 340

Quoted 6 lines
drivers/devfreq/tegra-devfreq.c +1 -1
 	avg_sustain_coef = 100 * 100 / dev->config->boost_up_threshold;
 	dev->target_freq = do_percent(dev->target_freq, avg_sustain_coef);
 	dev->target_freq += dev->boost_freq;
 
-	if (dev->avg_count >= dev->config->avg_dependency_threshold)
+	if (avg_count >= dev->config->avg_dependency_threshold)

ditto.

drivers/devfreq/tegra-devfreq.c static int tegra_actmon_rate_notify_cb(struct notifier_block *nb,

Quoted 42 lines
drivers/devfreq/tegra-devfreq.c +16 -5
 		dev->target_freq = max(dev->target_freq, static_cpu_emc_freq);
-
-	spin_unlock_irqrestore(&dev->lock, flags);
 }
 
 static irqreturn_t actmon_thread_isr(int irq, void *data)
 {
 	struct tegra_devfreq *tegra = data;
+	bool handled = false;
+	unsigned int i;
+	u32 val;
 
 	mutex_lock(&tegra->devfreq->lock);
-	update_devfreq(tegra->devfreq);
+
+	val = actmon_readl(tegra, ACTMON_GLB_STATUS);
+	for (i = 0; i < ARRAY_SIZE(tegra->devices); i++) {
+		if (val & tegra->devices[i].config->irq_mask) {
+			actmon_isr_device(tegra, tegra->devices + i);
+			handled = true;
+		}
+	}
+
+	if (handled)
+		update_devfreq(tegra->devfreq);
+
 	mutex_unlock(&tegra->devfreq->lock);
 
-	return IRQ_HANDLED;
+	return handled ? IRQ_HANDLED : IRQ_NONE;
 }
 
 static int tegra_actmon_rate_notify_cb(struct notifier_block *nb,
@@ -386,7 +373,6 @@ static int tegra_actmon_rate_notify_cb(struct notifier_block *nb,
 	struct tegra_devfreq *tegra;
 	struct tegra_devfreq_device *dev;
 	unsigned int i;
-	unsigned long flags;
 
 	if (action != POST_RATE_CHANGE)
 		return NOTIFY_OK;
@@ -398,9 +384,7 @@ static int tegra_actmon_rate_notify_cb(struct notifier_block *nb,

When I review this patch, I have a question about why tegra_actmon_rate_notify_cb is needed.

tegra_actmon_rate_notify_cb() do something
when the clock rate of emc_clock is changed.
I think that 'emc_clock' is changed by this driver. It means that the this driver can catch the change timing of emc_clock rate without notifier.

IMO, it is possible to call tegra_devfreq_update_wmark() directly without clock notifier before calling 'clk_set_min_rate()/clk_set_rate()' in the tegra_devfreq_target().

With clock notifier, it cannot restore something for tegra_devfreq_update_wmark(tegra, dev) when failed to set the rate of emc_clk by 'clk_set_min_rate()/clk_set_rate()'.

Quoted 29 lines
drivers/devfreq/tegra-devfreq.c +2 -5
 	for (i = 0; i < ARRAY_SIZE(tegra->devices); i++) {
 		dev = &tegra->devices[i];
 
-		spin_lock_irqsave(&dev->lock, flags);
 		tegra_devfreq_update_wmark(tegra, dev);
-		spin_unlock_irqrestore(&dev->lock, flags);
 	}
 
 	actmon_write_barrier(tegra);
@@ -668,7 +652,6 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		dev = tegra->devices + i;
 		dev->config = actmon_device_configs + i;
 		dev->regs = tegra->regs + dev->config->offset;
-		spin_lock_init(&dev->lock);
 
 		tegra_actmon_configure_device(tegra, dev);
 	}
@@ -696,8 +679,8 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		return err;
 	}
 
-	err = devm_request_threaded_irq(&pdev->dev, irq, actmon_isr,
-					actmon_thread_isr, IRQF_SHARED,
+	err = devm_request_threaded_irq(&pdev->dev, irq, NULL,
+					actmon_thread_isr, IRQF_ONESHOT,
 					"tegra-devfreq", tegra);
 	if (err) {
 		dev_err(&pdev->dev, "Interrupt request failed\n");
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer
#28 lore raw

Re: [PATCH v2 11/19] PM / devfreq: tegra: De-initialize properly on driver's probe error

In reply to Dmitry Osipenko

Hi Dmitry,

I already replied against patch6 about the exception handling of tegra_devfreq_probe(). This patchset split out the patch related to error handling for probe(). I think that you can squash the patches regarding of exception handling for probe() to one patch instead of split out the multiple patches.

Quoted 109 lines · Dmitry Osipenko

Reset hardware, disable ACTMON clock, release OPP's and free IRQ before removing devfreq device since there is no guarantee that interrupt handling won't run after masking interrupt in hardware.

---
 drivers/devfreq/tegra-devfreq.c | 53 ++++++++++++++++-----------------
 1 file changed, 26 insertions(+), 27 deletions(-)
drivers/devfreq/tegra-devfreq.c +26 -27
index 69b557df5084..a668e4fbc874 100644
@@ -663,28 +663,28 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 
 	irq = platform_get_irq(pdev, 0);
 	if (irq < 0) {
-		dev_err(&pdev->dev, "Failed to get IRQ: %d\n", irq);
-		return irq;
+		err = irq;
+		dev_err(&pdev->dev, "Failed to get IRQ: %d\n", err);
+		goto remove_opps;
 	}
 
 	platform_set_drvdata(pdev, tegra);
 
 	tegra_devfreq_profile.initial_freq = clk_get_rate(tegra->emc_clock);
-	tegra->devfreq = devm_devfreq_add_device(&pdev->dev,
-						 &tegra_devfreq_profile,
-						 "tegra_actmon",
-						 NULL);
+	tegra->devfreq = devfreq_add_device(&pdev->dev,
+					    &tegra_devfreq_profile,
+					    "tegra_actmon",
+					    NULL);
 	if (IS_ERR(tegra->devfreq)) {
 		err = PTR_ERR(tegra->devfreq);
-		return err;
+		goto remove_opps;
 	}
 
-	err = devm_request_threaded_irq(&pdev->dev, irq, NULL,
-					actmon_thread_isr, IRQF_ONESHOT,
-					"tegra-devfreq", tegra);
+	err = request_threaded_irq(irq, NULL, actmon_thread_isr, IRQF_ONESHOT,
+				   "tegra-devfreq", tegra);
 	if (err) {
 		dev_err(&pdev->dev, "Interrupt request failed\n");
-		goto remove_opps;
+		goto remove_devfreq;
 	}
 
 	tegra->rate_change_nb.notifier_call = tegra_actmon_rate_notify_cb;
@@ -692,14 +692,23 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 	if (err) {
 		dev_err(&pdev->dev,
 			"Failed to register rate change notifier\n");
-		goto remove_opps;
+		goto disable_interrupt;
 	}
 
 	return 0;
 
+disable_interrupt:
+	free_irq(irq, tegra);
+
+remove_devfreq:
+	devfreq_remove_device(tegra->devfreq);
+
 remove_opps:
 	dev_pm_opp_remove_all_dynamic(&pdev->dev);
 
+	reset_control_reset(tegra->reset);
+	clk_disable_unprepare(tegra->clock);
+
 	return err;
 }
 
@@ -707,24 +716,14 @@ static int tegra_devfreq_remove(struct platform_device *pdev)
 {
 	struct tegra_devfreq *tegra = platform_get_drvdata(pdev);
 	int irq = platform_get_irq(pdev, 0);
-	u32 val;
-	unsigned int i;
-
-	devm_devfreq_remove_device(&pdev->dev, tegra->devfreq);
-	dev_pm_opp_remove_all_dynamic(&pdev->dev);
-
-	for (i = 0; i < ARRAY_SIZE(actmon_device_configs); i++) {
-		val = device_readl(&tegra->devices[i], ACTMON_DEV_CTRL);
-		val &= ~ACTMON_DEV_CTRL_ENB;
-		device_writel(&tegra->devices[i], val, ACTMON_DEV_CTRL);
-	}
-
-	actmon_write_barrier(tegra);
-
-	devm_free_irq(&pdev->dev, irq, tegra);
 
 	clk_notifier_unregister(tegra->emc_clock, &tegra->rate_change_nb);
+	free_irq(irq, tegra);
+
+	devfreq_remove_device(tegra->devfreq);
+	dev_pm_opp_remove_all_dynamic(&pdev->dev);
 
+	reset_control_reset(tegra->reset);
 	clk_disable_unprepare(tegra->clock);
 
 	return 0;
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer
#30 lore raw

Re: [PATCH v2 12/19] PM / devfreq: tegra: Avoid inconsistency of current frequency value

In reply to Dmitry Osipenko

Hi,

Quoted 32 lines · Dmitry Osipenko

The frequency value potentially could change in-between. It doesn't cause any real problem at all right now, but that could change in the future. Hence let's avoid the inconsistency.

---
 drivers/devfreq/tegra-devfreq.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +3 -1
index a668e4fbc874..f1a6f951813a 100644
@@ -496,13 +496,15 @@ static int tegra_devfreq_get_dev_status(struct device *dev,
 {
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct tegra_devfreq_device *actmon_dev;
+	unsigned long cur_freq;
 
-	stat->current_frequency = tegra->cur_freq * KHZ;
+	cur_freq = READ_ONCE(tegra->cur_freq);
 
 	/* To be used by the tegra governor */
 	stat->private_data = tegra;
 
 	/* The below are to be used by the other governors */
+	stat->current_frequency = cur_freq * KHZ;
 
 	actmon_dev = &tegra->devices[MCALL];
 

drivers/devfreq/tegra-devfreq.c static int tegra_devfreq_get_dev_status(struct device *dev, · line 521

Quoted 8 lines
drivers/devfreq/tegra-devfreq.c +1 -1
@@ -513,7 +515,7 @@ static int tegra_devfreq_get_dev_status(struct device *dev,
 	stat->busy_time *= 100 / BUS_SATURATION_RATIO;
 
 	/* Number of cycles in a sampling period */
-	stat->total_time = ACTMON_SAMPLING_PERIOD * tegra->cur_freq;
+	stat->total_time = ACTMON_SAMPLING_PERIOD * cur_freq;
 
 	stat->busy_time = min(stat->busy_time, stat->total_time);
 

The read/write access of tegra->cur_freq is in the single routine of update_devfreq() as following. I think that there are no any potential problem about the inconsistency of tegra->cur_freq.

IMHO, if there are no any problem now, I'm not sure that we need to apply this patch.

update_devfreq()

{
	devfreq->governor->get_target_freq()
		devfreq_update_stats(devfreq)
			tegra_devfreq_get_dev_status()
				stat->current_frequency = tegra->cur_freq * KHZ;

	devfreq_set_target()
		tegra_devfreq_target()
			clk_set_min_rate(emc_rate, )
				tegra_actmon_rate_notify_cb()
					tegra->cur_freq = data->new_rate / KHZ;
		
			clk_set_rate(emc_rate, )
				tegra_actmon_rate_notify_cb()
					tegra->cur_freq = data->new_rate / KHZ;
}
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer Reviewed-by on v2 13/19
#34 lore raw

Re: [PATCH v2 13/19] PM / devfreq: tegra: Mark ACTMON's governor as immutable

In reply to Dmitry Osipenko

Hi,

Quoted 28 lines · Dmitry Osipenko

The ACTMON's governor supports only the Tegra's devfreq device and there is no need to use any other governor, hence let's mark Tegra governor as immutable to permanently stick it with Tegra's devfreq device.

---
 drivers/devfreq/Kconfig         | 1 -
 drivers/devfreq/tegra-devfreq.c | 1 +
 2 files changed, 1 insertion(+), 1 deletion(-)
drivers/devfreq/Kconfig +0 -1
index 6a172d338f6d..a78dffe603c1 100644
@@ -94,7 +94,6 @@ config ARM_EXYNOS_BUS_DEVFREQ
 config ARM_TEGRA_DEVFREQ
 	tristate "Tegra DEVFREQ Driver"
 	depends on ARCH_TEGRA_124_SOC
-	select DEVFREQ_GOV_SIMPLE_ONDEMAND
 	select PM_OPP
 	help
 	  This adds the DEVFREQ driver for the Tegra family of SoCs.
drivers/devfreq/tegra-devfreq.c +0 -0
index f1a6f951813a..832e4f5aa11b 100644

drivers/devfreq/tegra-devfreq.c static struct devfreq_governor tegra_devfreq_governor = { · line 599

Quoted 7 lines
drivers/devfreq/tegra-devfreq.c +1 -0
@@ -593,6 +593,7 @@ static struct devfreq_governor tegra_devfreq_governor = {
 	.name = "tegra_actmon",
 	.get_target_freq = tegra_governor_get_target,
 	.event_handler = tegra_governor_event_handler,
+	.immutable = true,
 };
 
 static int tegra_devfreq_probe(struct platform_device *pdev)

Looks good to me.

Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer Reviewed-by on v2 14/19
#36 lore raw

Re: [PATCH v2 14/19] PM / devfreq: tegra: Move governor registration to driver's probe

In reply to Dmitry Osipenko
Quoted 83 lines · Dmitry Osipenko

There is no need to register the ACTMON's governor separately from the driver, hence let's move the registration into the driver's probe function for consistency and to make code cleaner a tad.

---
 drivers/devfreq/tegra-devfreq.c | 44 +++++++++------------------------
 1 file changed, 12 insertions(+), 32 deletions(-)
drivers/devfreq/tegra-devfreq.c +11 -26
index 832e4f5aa11b..46c61af8ca33 100644
@@ -673,6 +673,12 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 
 	platform_set_drvdata(pdev, tegra);
 
+	err = devfreq_add_governor(&tegra_devfreq_governor);
+	if (err) {
+		dev_err(&pdev->dev, "Failed to add governor: %d\n", err);
+		goto remove_opps;
+	}
+
 	tegra_devfreq_profile.initial_freq = clk_get_rate(tegra->emc_clock);
 	tegra->devfreq = devfreq_add_device(&pdev->dev,
 					    &tegra_devfreq_profile,
@@ -680,7 +686,7 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 					    NULL);
 	if (IS_ERR(tegra->devfreq)) {
 		err = PTR_ERR(tegra->devfreq);
-		goto remove_opps;
+		goto remove_governor;
 	}
 
 	err = request_threaded_irq(irq, NULL, actmon_thread_isr, IRQF_ONESHOT,
@@ -706,6 +712,9 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 remove_devfreq:
 	devfreq_remove_device(tegra->devfreq);
 
+remove_governor:
+	devfreq_remove_governor(&tegra_devfreq_governor);
+
 remove_opps:
 	dev_pm_opp_remove_all_dynamic(&pdev->dev);
 
@@ -729,7 +738,7 @@ static int tegra_devfreq_remove(struct platform_device *pdev)
 	reset_control_reset(tegra->reset);
 	clk_disable_unprepare(tegra->clock);
 
-	return 0;
+	return devfreq_remove_governor(&tegra_devfreq_governor);
 }
 
 static const struct of_device_id tegra_devfreq_of_match[] = {
@@ -747,36 +756,7 @@ static struct platform_driver tegra_devfreq_driver = {
 		.of_match_table = tegra_devfreq_of_match,
 	},
 };
-
-static int __init tegra_devfreq_init(void)
-{
-	int ret = 0;
-
-	ret = devfreq_add_governor(&tegra_devfreq_governor);
-	if (ret) {
-		pr_err("%s: failed to add governor: %d\n", __func__, ret);
-		return ret;
-	}
-
-	ret = platform_driver_register(&tegra_devfreq_driver);
-	if (ret)
-		devfreq_remove_governor(&tegra_devfreq_governor);
-
-	return ret;
-}
-module_init(tegra_devfreq_init)
-
-static void __exit tegra_devfreq_exit(void)
-{
-	int ret = 0;
-
-	platform_driver_unregister(&tegra_devfreq_driver);

drivers/devfreq/tegra-devfreq.c static struct platform_driver tegra_devfreq_driver = { · line 762

Quoted 10 lines
drivers/devfreq/tegra-devfreq.c +1 -6
-
-	ret = devfreq_remove_governor(&tegra_devfreq_governor);
-	if (ret)
-		pr_err("%s: failed to remove governor: %d\n", __func__, ret);
-}
-module_exit(tegra_devfreq_exit)
+module_platform_driver(tegra_devfreq_driver);
 
 MODULE_LICENSE("GPL v2");
 MODULE_DESCRIPTION("Tegra devfreq driver");

Looks good to me.

Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer
#38 lore raw

Re: [PATCH v2 15/19] PM / devfreq: tegra: Synchronize IRQ after masking it in hardware

In reply to Dmitry Osipenko

Hi,

In this patchset,
the patch11 adds new 'disable_interrupt' goto statement and then patch15 removes 'disable_interrupt' goto statement again. Actually, it is inefficient.

If you change the order of patches,
you could remove this stuff.

Quoted 108 lines · Dmitry Osipenko

There is no guarantee that interrupt handling isn't running in parallel, hence it is necessary to synchronize IRQ in order to ensure that interrupt is indeed disabled.

---
 drivers/devfreq/tegra-devfreq.c | 33 ++++++++++++++++-----------------
 1 file changed, 16 insertions(+), 17 deletions(-)
drivers/devfreq/tegra-devfreq.c +16 -17
index 46c61af8ca33..cfbfbe28e7bf 100644
@@ -171,6 +171,8 @@ struct tegra_devfreq {
 	struct notifier_block	rate_change_nb;
 
 	struct tegra_devfreq_device devices[ARRAY_SIZE(actmon_device_configs)];
+
+	int irq;
 };
 
 struct tegra_actmon_emc_ratio {
@@ -432,6 +434,8 @@ static void tegra_actmon_disable_interrupts(struct tegra_devfreq *tegra)
 	}
 
 	actmon_write_barrier(tegra);
+
+	synchronize_irq(tegra->irq);
 }
 
 static void tegra_actmon_configure_device(struct tegra_devfreq *tegra,
@@ -603,7 +607,6 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 	struct resource *res;
 	unsigned int i;
 	unsigned long rate;
-	int irq;
 	int err;
 
 	tegra = devm_kzalloc(&pdev->dev, sizeof(*tegra), GFP_KERNEL);
@@ -659,18 +662,18 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		tegra_actmon_configure_device(tegra, dev);
 	}
 
+	tegra->irq = platform_get_irq(pdev, 0);
+	if (tegra->irq < 0) {
+		err = tegra->irq;
+		dev_err(&pdev->dev, "Failed to get IRQ: %d\n", err);
+		return err;
+	}
+
 	for (rate = 0; rate <= tegra->max_freq * KHZ; rate++) {
 		rate = clk_round_rate(tegra->emc_clock, rate);
 		dev_pm_opp_add(&pdev->dev, rate, 0);
 	}
 
-	irq = platform_get_irq(pdev, 0);
-	if (irq < 0) {
-		err = irq;
-		dev_err(&pdev->dev, "Failed to get IRQ: %d\n", err);
-		goto remove_opps;
-	}
-
 	platform_set_drvdata(pdev, tegra);
 
 	err = devfreq_add_governor(&tegra_devfreq_governor);
@@ -689,10 +692,11 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		goto remove_governor;
 	}
 
-	err = request_threaded_irq(irq, NULL, actmon_thread_isr, IRQF_ONESHOT,
-				   "tegra-devfreq", tegra);
+	err = devm_request_threaded_irq(&pdev->dev, tegra->irq, NULL,
+					actmon_thread_isr, IRQF_ONESHOT,
+					"tegra-devfreq", tegra);
 	if (err) {
-		dev_err(&pdev->dev, "Interrupt request failed\n");
+		dev_err(&pdev->dev, "Interrupt request failed: %d\n", err);
 		goto remove_devfreq;
 	}
 
@@ -701,14 +705,11 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 	if (err) {
 		dev_err(&pdev->dev,
 			"Failed to register rate change notifier\n");
-		goto disable_interrupt;
+		goto remove_devfreq;
 	}
 
 	return 0;
 
-disable_interrupt:
-	free_irq(irq, tegra);
-
 remove_devfreq:
 	devfreq_remove_device(tegra->devfreq);
 
@@ -727,10 +728,8 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 static int tegra_devfreq_remove(struct platform_device *pdev)
 {
 	struct tegra_devfreq *tegra = platform_get_drvdata(pdev);
-	int irq = platform_get_irq(pdev, 0);
 
 	clk_notifier_unregister(tegra->emc_clock, &tegra->rate_change_nb);
-	free_irq(irq, tegra);
 
 	devfreq_remove_device(tegra->devfreq);
 	dev_pm_opp_remove_all_dynamic(&pdev->dev);
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer
#46 lore raw

Re: [PATCH v2 18/19] PM / devfreq: tegra: Enable COMPILE_TEST for the driver

In reply to Dmitry Osipenko
Quoted 14 lines · Dmitry Osipenko

The driver's compilation doesn't have any specific dependencies, hence the COMPILE_TEST option can be supported in Kconfig.

---
 drivers/devfreq/Kconfig | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/devfreq/Kconfig +0 -0
index 78c33ddd4eea..bd6652863e7d 100644

drivers/devfreq/Kconfig config ARM_EXYNOS_BUS_DEVFREQ · line 99

Quoted 8 lines
drivers/devfreq/Kconfig +1 -1
@@ -93,7 +93,7 @@ config ARM_EXYNOS_BUS_DEVFREQ
 
 config ARM_TEGRA_DEVFREQ
 	tristate "NVIDIA Tegra30+ DEVFREQ Driver"
-	depends on ARCH_TEGRA
+	depends on ARCH_TEGRA || COMPILE_TEST
 	select PM_OPP
 	help
 	  This adds the DEVFREQ driver for the Tegra family of SoCs.

I think that it doesn't need to make it a separate patch. You can merge this patch to patch17 and then drop this patch.

Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer Reviewed-by on v2 17/19
#42 lore raw

Re: [PATCH v2 17/19] PM / devfreq: tegra: Support Tegra30

In reply to Dmitry Osipenko

Hi,

Quoted 15 lines · Dmitry Osipenko

The devfreq driver can be used on Tegra30 without any code change and it works perfectly fine, the default Tegra124 parameters are good enough for Tegra30.

---
 drivers/devfreq/Kconfig         | 4 ++--
 drivers/devfreq/tegra-devfreq.c | 1 +
 2 files changed, 3 insertions(+), 2 deletions(-)
drivers/devfreq/Kconfig +0 -0
index a78dffe603c1..78c33ddd4eea 100644

drivers/devfreq/Kconfig config ARM_EXYNOS_BUS_DEVFREQ · line 95

Quoted 6 lines
drivers/devfreq/Kconfig +1 -2
@@ -92,8 +92,8 @@ config ARM_EXYNOS_BUS_DEVFREQ
 	  This does not yet operate with optimal voltages.
 
 config ARM_TEGRA_DEVFREQ
-	tristate "Tegra DEVFREQ Driver"
-	depends on ARCH_TEGRA_124_SOC
+	tristate "NVIDIA Tegra30+ DEVFREQ Driver"

Looks good to me. But, I have a question.

'Tegra30+' expression is enough in order to indicate the support for both tegra30 and tegra124?

In my case, it is difficult to catch the meaning
of which tegra30+ support the kind of tegra124.

Quoted 10 lines
drivers/devfreq/Kconfig +1 -0
+	depends on ARCH_TEGRA
 	select PM_OPP
 	help
 	  This adds the DEVFREQ driver for the Tegra family of SoCs.
drivers/devfreq/tegra-devfreq.c +0 -0
index f0711d5ad27d..0a2465a58cf5 100644

drivers/devfreq/tegra-devfreq.c static int tegra_devfreq_remove(struct platform_device *pdev) · line 725

Quoted 7 lines
drivers/devfreq/tegra-devfreq.c +1 -0
@@ -719,6 +719,7 @@ static int tegra_devfreq_remove(struct platform_device *pdev)
 }
 
 static const struct of_device_id tegra_devfreq_of_match[] = {
+	{ .compatible = "nvidia,tegra30-actmon" },
 	{ .compatible = "nvidia,tegra124-actmon" },
 	{ },
 };
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer
#13 lore raw

Re: [PATCH v2 05/19] PM / devfreq: tegra: Replace write memory barrier with the read barrier

In reply to Dmitry Osipenko

Hi,

Quoted 15 lines · Dmitry Osipenko

The write memory barrier isn't needed because the BUS buffer is flushed by read after write that happens after the removed wmb(), we will also use readl() instead of the relaxed version to ensure that read is indeed completed.

---
 drivers/devfreq/tegra-devfreq.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +0 -0
index d62fb1b0d9bb..f0f0d78f6cbf 100644

drivers/devfreq/tegra-devfreq.c static void tegra_devfreq_update_wmark(struct tegra_devfreq *tegra, · line 246

Quoted 6 lines
drivers/devfreq/tegra-devfreq.c +1 -2
@@ -243,8 +243,7 @@ static void tegra_devfreq_update_wmark(struct tegra_devfreq *tegra,
 static void actmon_write_barrier(struct tegra_devfreq *tegra)
 {
 	/* ensure the update has reached the ACTMON */
-	wmb();
-	actmon_readl(tegra, ACTMON_GLB_STATUS);
+	readl(tegra->regs + ACTMON_GLB_STATUS);

I think that this meaning of actmon_write_barrier() keeps the order of 'store' assembly command without the execution change from compiler optimization by using the wmb().

But, this patch edits it as following:
The result of the following two cases are same?

[original code]

	wmb()
	read_relaxed()

[new code by this patch]

	readl_relaxed()
	rmb()
Quoted 4 lines
drivers/devfreq/tegra-devfreq.c +0 -0
 }
 
 static void actmon_isr_device(struct tegra_devfreq *tegra,
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer
#50 lore raw

Re: [PATCH v2 19/19] PM / devfreq: Introduce driver for NVIDIA Tegra20

In reply to Dmitry Osipenko

Hi,

Quoted 71 lines · Dmitry Osipenko

Add devfreq driver for NVIDIA Tegra20 SoC's. The driver periodically reads out Memory Controller counters and adjusts memory frequency based on the memory clients activity.

---
 MAINTAINERS                       |   8 ++
 drivers/devfreq/Kconfig           |  10 ++
 drivers/devfreq/Makefile          |   1 +
 drivers/devfreq/tegra20-devfreq.c | 177 ++++++++++++++++++++++++++++++
 4 files changed, 196 insertions(+)
 create mode 100644 drivers/devfreq/tegra20-devfreq.c
MAINTAINERS +8 -0
index 80b59db1b6e4..91f475ec4545 100644
@@ -10056,6 +10056,14 @@ F:	include/linux/memblock.h
 F:	mm/memblock.c
 F:	Documentation/core-api/boot-time-mm.rst
 
+MEMORY FREQUENCY SCALING DRIVER FOR NVIDIA TEGRA20
+M:	Dmitry Osipenko <[email protected]>
+L:	[email protected]
+L:	[email protected]
+T:	git git://git.kernel.org/pub/scm/linux/kernel/git/mzx/devfreq.git
+S:	Maintained
+F:	drivers/devfreq/tegra20-devfreq.c
+
 MEMORY MANAGEMENT
 L:	[email protected]
 W:	http://www.linux-mm.org
drivers/devfreq/Kconfig +10 -0
index bd6652863e7d..af4c86c4e0f6 100644
@@ -100,6 +100,16 @@ config ARM_TEGRA_DEVFREQ
 	  It reads ACTMON counters of memory controllers and adjusts the
 	  operating frequencies and voltages with OPP support.
 
+config ARM_TEGRA20_DEVFREQ
+	tristate "NVIDIA Tegra20 DEVFREQ Driver"
+	depends on (TEGRA_MC && TEGRA20_EMC) || COMPILE_TEST
+	select DEVFREQ_GOV_SIMPLE_ONDEMAND
+	select PM_OPP
+	help
+	  This adds the DEVFREQ driver for the Tegra20 family of SoCs.
+	  It reads Memory Controller counters and adjusts the operating
+	  frequencies and voltages with OPP support.
+
 config ARM_RK3399_DMC_DEVFREQ
 	tristate "ARM RK3399 DMC DEVFREQ Driver"
 	depends on ARCH_ROCKCHIP
drivers/devfreq/Makefile +1 -0
index 32b8d4d3f12c..6fcc5596b8b7 100644
@@ -11,6 +11,7 @@ obj-$(CONFIG_DEVFREQ_GOV_PASSIVE)	+= governor_passive.o
 obj-$(CONFIG_ARM_EXYNOS_BUS_DEVFREQ)	+= exynos-bus.o
 obj-$(CONFIG_ARM_RK3399_DMC_DEVFREQ)	+= rk3399_dmc.o
 obj-$(CONFIG_ARM_TEGRA_DEVFREQ)		+= tegra-devfreq.o
+obj-$(CONFIG_ARM_TEGRA20_DEVFREQ)	+= tegra20-devfreq.o
 
 # DEVFREQ Event Drivers
 obj-$(CONFIG_PM_DEVFREQ_EVENT)		+= event/
drivers/devfreq/tegra20-devfreq.c new file +0 -0
new file mode 100644
index 000000000000..18c9aad7a9d7

drivers/devfreq/tegra20-devfreq.c line 6

Quoted 6 lines
drivers/devfreq/tegra20-devfreq.c new file +6 -0
@@ -0,0 +1,177 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * NVIDIA Tegra20 devfreq driver
+ *
+ * Author: Dmitry Osipenko <[email protected]>
+ */

It doesn't any "Copyright (c) 2019 ..." sentence.

drivers/devfreq/tegra20-devfreq.c line 18

Quoted 12 lines
drivers/devfreq/tegra20-devfreq.c new file +12 -0
+
+#include <linux/clk.h>
+#include <linux/devfreq.h>
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of_device.h>
+#include <linux/platform_device.h>
+#include <linux/pm_opp.h>
+#include <linux/slab.h>
+
+#include <soc/tegra/mc.h>

I can find the '<soc/tegra/mc.h>' header file
on mainline branch. But mc.h is included in linux-next.git.

If you don't share the patch related to mc.h,
the kernel build will be failed when apply it the devfreq.git on final step. Actually, it should make the immutable branch between two related maintainers in order to remove the build fail.

Quoted 29 lines
drivers/devfreq/tegra20-devfreq.c new file +29 -0
+
+#include "governor.h"
+
+#define MC_STAT_CONTROL				0x90
+#define MC_STAT_EMC_CLOCK_LIMIT			0xa0
+#define MC_STAT_EMC_CLOCKS			0xa4
+#define MC_STAT_EMC_CONTROL			0xa8
+#define MC_STAT_EMC_COUNT			0xb8
+
+#define EMC_GATHER_CLEAR			(1 << 8)
+#define EMC_GATHER_ENABLE			(3 << 8)
+
+struct tegra_devfreq {
+	struct devfreq *devfreq;
+	struct clk *clk;
+	void __iomem *regs;
+};
+
+static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
+				u32 flags)
+{
+	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
+	struct dev_pm_opp *opp;
+	unsigned long rate;
+	int err;
+
+	opp = devfreq_recommended_opp(dev, freq, flags);
+	if (IS_ERR(opp))
+		return PTR_ERR(opp);

drivers/devfreq/tegra20-devfreq.c line 57

Quoted 10 lines
drivers/devfreq/tegra20-devfreq.c new file +10 -0
+
+	rate = dev_pm_opp_get_freq(opp);
+	dev_pm_opp_put(opp);
+
+	err = clk_set_min_rate(tegra->clk, rate);
+	if (err)
+		return err;
+
+	err = clk_set_rate(tegra->clk, 0);
+	if (err)

When fail happen, I think that you have to control
the restoring sequence for previous operation like clk_set_min_rate().

Quoted 50 lines
drivers/devfreq/tegra20-devfreq.c new file +50 -0
+		return err;
+
+	return 0;
+}
+
+static int tegra_devfreq_get_dev_status(struct device *dev,
+					struct devfreq_dev_status *stat)
+{
+	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
+
+	stat->busy_time = readl_relaxed(tegra->regs + MC_STAT_EMC_COUNT);
+	stat->total_time = readl_relaxed(tegra->regs + MC_STAT_EMC_CLOCKS) / 8;
+	stat->current_frequency = clk_get_rate(tegra->clk);
+
+	writel_relaxed(EMC_GATHER_CLEAR, tegra->regs + MC_STAT_CONTROL);
+	writel_relaxed(EMC_GATHER_ENABLE, tegra->regs + MC_STAT_CONTROL);
+
+	return 0;
+}
+
+static struct devfreq_dev_profile tegra_devfreq_profile = {
+	.polling_ms	= 500,
+	.target		= tegra_devfreq_target,
+	.get_dev_status	= tegra_devfreq_get_dev_status,
+};
+
+static struct tegra_mc *tegra_get_memory_controller(void)
+{
+	struct platform_device *pdev;
+	struct device_node *np;
+	struct tegra_mc *mc;
+
+	np = of_find_compatible_node(NULL, NULL, "nvidia,tegra20-mc-gart");
+	if (!np)
+		return ERR_PTR(-ENOENT);
+
+	pdev = of_find_device_by_node(np);
+	of_node_put(np);
+	if (!pdev)
+		return ERR_PTR(-ENODEV);
+
+	mc = platform_get_drvdata(pdev);
+	if (!mc)
+		return ERR_PTR(-EPROBE_DEFER);
+
+	return mc;
+}
+
+static int tegra_devfeq_probe(struct platform_device *pdev)
+{

drivers/devfreq/tegra20-devfreq.c line 117

Quoted 10 lines
drivers/devfreq/tegra20-devfreq.c new file +10 -0
+	struct tegra_devfreq *tegra;
+	struct tegra_mc *mc;
+	unsigned long max_rate;
+	unsigned long rate;
+	int err;
+
+	mc = tegra_get_memory_controller();
+	if (IS_ERR(mc)) {
+		err = PTR_ERR(mc);
+		dev_err(&pdev->dev, "failed to get mc: %d\n", err);

How about using 'memory controller' instead of 'mc'? Because 'mc' is not standard expression.

drivers/devfreq/tegra20-devfreq.c line 130

Quoted 13 lines
drivers/devfreq/tegra20-devfreq.c new file +13 -0
+		return err;
+	}
+
+	tegra = devm_kzalloc(&pdev->dev, sizeof(*tegra), GFP_KERNEL);
+	if (!tegra)
+		return -ENOMEM;
+
+	tegra->clk = devm_clk_get(&pdev->dev, "emc");
+	if (IS_ERR(tegra->clk)) {
+		err = PTR_ERR(tegra->clk);
+		dev_err(&pdev->dev, "failed to get emc clock: %d\n", err);
+		return err;
+	}

Don't you need to enable the 'emc' clock'?
Because this patch doesn't enable this clock.

drivers/devfreq/tegra20-devfreq.c line 143

Quoted 13 lines
drivers/devfreq/tegra20-devfreq.c new file +13 -0
+
+	tegra->regs = mc->regs;
+
+	max_rate = clk_round_rate(tegra->clk, ULONG_MAX);
+
+	for (rate = 0; rate <= max_rate; rate++) {
+		rate = clk_round_rate(tegra->clk, rate);
+		dev_pm_opp_add(&pdev->dev, rate, 0);
+	}
+
+	writel_relaxed(0x00000000, tegra->regs + MC_STAT_CONTROL);
+	writel_relaxed(0x00000000, tegra->regs + MC_STAT_EMC_CONTROL);
+	writel_relaxed(0xffffffff, tegra->regs + MC_STAT_EMC_CLOCK_LIMIT);

You better to add the comments of what are the meaning of 0x00000000/0xffffffff. Without the detailed comments, it is difficult to understand of meaning.

Quoted 16 lines
drivers/devfreq/tegra20-devfreq.c new file +16 -0
+
+	platform_set_drvdata(pdev, tegra);
+
+	tegra->devfreq = devfreq_add_device(&pdev->dev, &tegra_devfreq_profile,
+					    DEVFREQ_GOV_SIMPLE_ONDEMAND, NULL);
+	if (IS_ERR(tegra->devfreq))
+		return PTR_ERR(tegra->devfreq);
+
+	return 0;
+}
+
+static int tegra_devfreq_remove(struct platform_device *pdev)
+{
+	struct tegra_devfreq *tegra = platform_get_drvdata(pdev);
+
+	devfreq_remove_device(tegra->devfreq);

drivers/devfreq/tegra20-devfreq.c line 169

Quoted 10 lines
drivers/devfreq/tegra20-devfreq.c new file +10 -0
+	dev_pm_opp_remove_all_dynamic(&pdev->dev);
+
+	return 0;
+}
+
+static struct platform_driver tegra_devfeq_driver = {
+	.probe		= tegra_devfeq_probe,
+	.remove		= tegra_devfreq_remove,
+	.driver		= {
+		.name	= "tegra20-devfreq",

How can you bind this driver without compatible name for Devicetree? And I tried to find the name ("tegra20-devfreq") in
the MFD drivers.

Quoted 9 lines
drivers/devfreq/tegra20-devfreq.c new file +8 -0
+	},
+};
+module_platform_driver(tegra_devfeq_driver);
+
+MODULE_ALIAS("platform:tegra20-devfreq");
+MODULE_AUTHOR("Dmitry Osipenko <[email protected]>");
+MODULE_DESCRIPTION("NVIDIA Tegra20 devfreq driver");
+MODULE_LICENSE("GPL v2");
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
#3 lore raw

Re: [PATCH v2 01/19] PM / devfreq: tegra: Fix kHz to Hz conversion

In reply to Chanwoo Choi

16.04.2019 4:45, Chanwoo Choi пишет:

Quoted 3 lines

Hi,

I add one minor comment (KHZ -> hz).

Hello Chanwoo,

Thank you very much for the review!

Quoted 29 lines · Dmitry Osipenko
Quoted 28 lines

The kHz to Hz is incorrectly converted in a few places in the code, this results in a wrong frequency being calculated because devfreq core uses OPP frequencies that are given in Hz to clamp the rate, while tegra-devfreq gives to the core value in kHz and then it also expects to receive value in kHz from the core. In a result memory freq is always set to a value which is close to ULONG_MAX because of the bug. Hence the EMC frequency is always capped to the maximum and the driver doesn't do anything useful. This patch was tested on Tegra30 and Tegra124 SoC's, EMC frequency scaling works properly now.

---
 drivers/devfreq/tegra-devfreq.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
drivers/devfreq/tegra-devfreq.c +1 -1
index c89ba7b834ff..02905978abe1 100644
@@ -486,9 +486,9 @@ static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 {
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct dev_pm_opp *opp;
-	unsigned long rate = *freq * KHZ;
+	unsigned long rate;
 
Quoted 7 lines

drivers/devfreq/tegra-devfreq.c static int tegra_devfreq_target(struct device *dev, unsigned long *freq, · line 493

Quoted 4 lines
drivers/devfreq/tegra-devfreq.c +1 -1
-	opp = devfreq_recommended_opp(dev, &rate, flags);
+	opp = devfreq_recommended_opp(dev, freq, flags);
 	if (IS_ERR(opp)) {
 		dev_err(dev, "Failed to find opp for %lu KHz\n", *freq);

Need to change it as following:
- KHz -> Hz

Indeed, good catch.

Quoted 27 lines
Quoted 27 lines
drivers/devfreq/tegra-devfreq.c +2 -4
 		return PTR_ERR(opp);
@@ -499,8 +499,6 @@ static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 	clk_set_min_rate(tegra->emc_clock, rate);
 	clk_set_rate(tegra->emc_clock, 0);
 
-	*freq = rate;
-
 	return 0;
 }
 
@@ -510,7 +508,7 @@ static int tegra_devfreq_get_dev_status(struct device *dev,
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct tegra_devfreq_device *actmon_dev;
 
-	stat->current_frequency = tegra->cur_freq;
+	stat->current_frequency = tegra->cur_freq * KHZ;
 
 	/* To be used by the tegra governor */
 	stat->private_data = tegra;
@@ -565,7 +563,7 @@ static int tegra_governor_get_target(struct devfreq *devfreq,
 		target_freq = max(target_freq, dev->target_freq);
 	}
 
-	*freq = target_freq;
+	*freq = target_freq * KHZ;
 
Quoted 6 lines

drivers/devfreq/tegra-devfreq.c static int tegra_governor_get_target(struct devfreq *devfreq, · line 569

Quoted 3 lines
drivers/devfreq/tegra-devfreq.c +0 -0
 	return 0;
 }

It looks good to me if modify it in accordance with my comment.

Okay, thanks.

#8 lore raw

Re: [PATCH v2 03/19] PM / devfreq: tegra: Don't ignore clk errors

In reply to Chanwoo Choi

16.04.2019 4:52, Chanwoo Choi пишет:

Quoted 40 lines

Hi,

Quoted 23 lines · Dmitry Osipenko

The clk_set_min_rate() could fail and in this case clk_set_rate() sets rate to 0, which may drop EMC rate to minimum and make machine very difficult to use.

---
 drivers/devfreq/tegra-devfreq.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +1 -0
index 0c0909fba545..69581c9082d4 100644
@@ -487,6 +487,7 @@ static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct dev_pm_opp *opp;
 	unsigned long rate;
+	int err;
 
 	opp = devfreq_recommended_opp(dev, freq, flags);
 	if (IS_ERR(opp)) {
@@ -496,8 +497,13 @@ static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
 	rate = dev_pm_opp_get_freq(opp);

drivers/devfreq/tegra-devfreq.c static int tegra_devfreq_target(struct device *dev, unsigned long *freq, · line 505

Quoted 10 lines
drivers/devfreq/tegra-devfreq.c +6 -2
 	dev_pm_opp_put(opp);
 
-	clk_set_min_rate(tegra->emc_clock, rate);
-	clk_set_rate(tegra->emc_clock, 0);
+	err = clk_set_min_rate(tegra->emc_clock, rate);
+	if (err)
+		return err;
+
+	err = clk_set_rate(tegra->emc_clock, 0);
+	if (err)

Why don't you restore the previous minimum clock for tegra->emc_clock when failed to execute 'clk_set_rate(tegra->emc_clock, 0);'?

Quoted 2 lines

I think that if you control the restoring the previous minimum clock, it is more nice to handle the error handling.

I now see that devfreq has the previous_freq field and indeed it is possible to restore the min_rate value using it. Thanks for the suggestion.

#11 lore raw

Re: [PATCH v2 04/19] PM / devfreq: tegra: Don't set EMC clock rate to maximum on probe

In reply to Chanwoo Choi

16.04.2019 4:59, Chanwoo Choi пишет:

Quoted 26 lines

Hi,

Quoted 13 lines · Dmitry Osipenko

There is no real benefit from doing so, hence let's drop that rate setting for consistency.

---
 drivers/devfreq/tegra-devfreq.c | 2 --
 1 file changed, 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +0 -0
index 69581c9082d4..d62fb1b0d9bb 100644

drivers/devfreq/tegra-devfreq.c static int tegra_devfreq_probe(struct platform_device *pdev) · line 652

Quoted 5 lines
drivers/devfreq/tegra-devfreq.c +0 -2
@@ -648,8 +648,6 @@ static int tegra_devfreq_probe(struct platform_device *pdev)
 		return PTR_ERR(tegra->emc_clock);
 	}
 
-	clk_set_rate(tegra->emc_clock, ULONG_MAX);
-

It seems like that initialize the emc_clock as the supported maximum clock. But, if the rate for emc_clock initialized by either bootloader or clock driver and it's well working until this code, actually, it is not necessary.

Quoted 2 lines

Looks good to me.

EMC clock rate is set properly at a boot time on all Tegra's, thanks.

#14 lore raw

Re: [PATCH v2 05/19] PM / devfreq: tegra: Replace write memory barrier with the read barrier

In reply to Chanwoo Choi

16.04.2019 11:00, Chanwoo Choi пишет:

Quoted 13 lines

Hi,

Quoted 10 lines · Dmitry Osipenko

The write memory barrier isn't needed because the BUS buffer is flushed by read after write that happens after the removed wmb(), we will also use readl() instead of the relaxed version to ensure that read is indeed completed.

---
 drivers/devfreq/tegra-devfreq.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
Quoted 15 lines

drivers/devfreq/tegra-devfreq.c static void tegra_devfreq_update_wmark(struct tegra_devfreq *tegra, · line 246

Quoted 11 lines
drivers/devfreq/tegra-devfreq.c +1 -2
index d62fb1b0d9bb..f0f0d78f6cbf 100644
@@ -243,8 +243,7 @@ static void tegra_devfreq_update_wmark(struct tegra_devfreq *tegra,
 static void actmon_write_barrier(struct tegra_devfreq *tegra)
 {
 	/* ensure the update has reached the ACTMON */
-	wmb();
-	actmon_readl(tegra, ACTMON_GLB_STATUS);
+	readl(tegra->regs + ACTMON_GLB_STATUS);

I think that this meaning of actmon_write_barrier() keeps the order of 'store' assembly command without the execution change from compiler optimization by using the wmb().

The IO mapped memory is strongly-ordered on ARM, hence all readl/writel accesses are guaranteed to be ordered by default. I think wmb() here is just a cut-n-pasted relic from old downstream driver.

Quoted 10 lines

But, this patch edits it as following:
The result of the following two cases are same?

[original code]

	wmb()
	read_relaxed()

[new code by this patch]

	readl_relaxed()
	rmb()

Yes, the result is the same. The wmb() is not just about IO accesses, but about all kind of memory accesses and at least on Tegra30 it is quite expensive operation because it translates into L2XO cache syncing (arm_heavy_mb) + dsb(). It should be more efficient to flush out writes with a read-back and then wait for that read operation to be completed.

#18 lore raw

Re: [PATCH v2 06/19] PM / devfreq: tegra: Fix missed error checking on devfreq initialization failure

In reply to Chanwoo Choi

16.04.2019 5:32, Chanwoo Choi пишет:

Quoted 5 lines

Hi,

patch6/7/8/9 are for handling of exception handling in probe() function. Actually, I'm not sure that there are special reason to split out the patches. I think that you can squash patch6/7/8/9 to only one patch.

Indeed, I was rebasing and reordering patches multiple times and looks like there is no reason not to squash these patches now.

Quoted 7 lines

Also, even if patch6/7/8/9 handle the exception handling in probe(), the tegra_devfreq_probe() doesn't support the restoring sequence when fail happen. I think that if you want to fix the fail case of probe(), please add the restoring sequence about following function.
- clk_disable_unprepare()
- clk_notifier_unregister()
- dev_pm_opp_remove()

When all of 6/7/8/9 patches are applied, the clk_notifier_register() becomes the last invocation of the probe function and clk_enable() is kept at the first place of the probe order. Hence the sequence you're suggesting is already incorrect because error-unwinding order usually should be opposite to the probe order. It looks to me that the current final result of these patches is already correct.

#25 lore raw

Re: [PATCH v2 10/19] PM / devfreq: tegra: Drop primary interrupt handler

In reply to Chanwoo Choi

16.04.2019 8:56, Chanwoo Choi пишет:

Quoted 88 lines

Hi,

It looks good to me to drop the primary interrupt handler but I have some comments. Please check it.

Quoted 68 lines · Dmitry Osipenko

There is no real need in the primary interrupt handler, hence move everything to the secondary (threaded) handler. In a result locking is consistent now and there are no potential races with the interrupt handler because it is protected with the devfreq's mutex.

---
 drivers/devfreq/tegra-devfreq.c | 61 ++++++++++++---------------------
 1 file changed, 22 insertions(+), 39 deletions(-)
drivers/devfreq/tegra-devfreq.c +1 -25
index 2a1464098200..69b557df5084 100644
@@ -144,7 +144,6 @@ static struct tegra_devfreq_device_config actmon_device_configs[] = {
 struct tegra_devfreq_device {
 	const struct tegra_devfreq_device_config *config;
 	void __iomem *regs;
-	spinlock_t lock;
 
 	/* Average event count sampled in the last interrupt */
 	u32 avg_count;
@@ -249,11 +248,8 @@ static void actmon_write_barrier(struct tegra_devfreq *tegra)
 static void actmon_isr_device(struct tegra_devfreq *tegra,
 			      struct tegra_devfreq_device *dev)
 {
-	unsigned long flags;
 	u32 intr_status, dev_ctrl;
 
-	spin_lock_irqsave(&dev->lock, flags);
-
 	dev->avg_count = device_readl(dev, ACTMON_DEV_AVG_COUNT);
 	tegra_devfreq_update_avg_wmark(tegra, dev);
 
@@ -302,26 +298,6 @@ static void actmon_isr_device(struct tegra_devfreq *tegra,
 	device_writel(dev, ACTMON_INTR_STATUS_CLEAR, ACTMON_DEV_INTR_STATUS);
 
 	actmon_write_barrier(tegra);
-
-	spin_unlock_irqrestore(&dev->lock, flags);
-}
-
-static irqreturn_t actmon_isr(int irq, void *data)
-{
-	struct tegra_devfreq *tegra = data;
-	bool handled = false;
-	unsigned int i;
-	u32 val;
-
-	val = actmon_readl(tegra, ACTMON_GLB_STATUS);
-	for (i = 0; i < ARRAY_SIZE(tegra->devices); i++) {
-		if (val & tegra->devices[i].config->irq_mask) {
-			actmon_isr_device(tegra, tegra->devices + i);
-			handled = true;
-		}
-	}
-
-	return handled ? IRQ_WAKE_THREAD : IRQ_NONE;
 }
 
 static unsigned long actmon_cpu_to_emc_rate(struct tegra_devfreq *tegra,
@@ -348,35 +324,46 @@ static void actmon_update_target(struct tegra_devfreq *tegra,
 	unsigned long cpu_freq = 0;
 	unsigned long static_cpu_emc_freq = 0;
 	unsigned int avg_sustain_coef;
-	unsigned long flags;
+	u32 avg_count;
 

drivers/devfreq/tegra-devfreq.c static void actmon_update_target(struct tegra_devfreq *tegra, · line 335

Quoted 10 lines
drivers/devfreq/tegra-devfreq.c +2 -3
 	if (dev->config->avg_dependency_threshold) {
 		cpu_freq = cpufreq_get(0);
 		static_cpu_emc_freq = actmon_cpu_to_emc_rate(tegra, cpu_freq);
 	}
 
-	spin_lock_irqsave(&dev->lock, flags);
-
-	dev->target_freq = dev->avg_count / ACTMON_SAMPLING_PERIOD;
+	avg_count = dev->avg_count;
+	dev->target_freq = avg_count / ACTMON_SAMPLING_PERIOD;

Actually, this change is not related to this patch. Please keep the original code.

Quoted 8 lines

drivers/devfreq/tegra-devfreq.c static void actmon_update_target(struct tegra_devfreq *tegra, · line 340

Quoted 6 lines
drivers/devfreq/tegra-devfreq.c +1 -1
 	avg_sustain_coef = 100 * 100 / dev->config->boost_up_threshold;
 	dev->target_freq = do_percent(dev->target_freq, avg_sustain_coef);
 	dev->target_freq += dev->boost_freq;
 
-	if (dev->avg_count >= dev->config->avg_dependency_threshold)
+	if (avg_count >= dev->config->avg_dependency_threshold)

ditto.

Good catch, it's a leftover from v1 that I forgot to revert. Thank you.

[snip]

Quoted 9 lines

When I review this patch, I have a question about why tegra_actmon_rate_notify_cb is needed.

tegra_actmon_rate_notify_cb() do something
when the clock rate of emc_clock is changed.
I think that 'emc_clock' is changed by this driver. It means that the this driver can catch the change timing of emc_clock rate without notifier.

The devfreq driver isn't the only driver of the EMC clock rate. The devfreq driver changes EMC freq dynamically based of on average memory usage activity, but for some hardware units (like display controller for example) there is a requirement for a minimum memory bandwidth (isochronous transactions) and hence when display is waking up from suspend it immediately requires an amount of memory bandwidth that could be higher than ACTMON hardware unit suggests and besides the ACTMON's reaction is delayed by the sampling period (12ms).

Quoted 7 lines

IMO, it is possible to call tegra_devfreq_update_wmark() directly without clock notifier before calling 'clk_set_min_rate()/clk_set_rate()' in the tegra_devfreq_target().

With clock notifier, it cannot restore something for tegra_devfreq_update_wmark(tegra, dev) when failed to set the rate of emc_clk by 'clk_set_min_rate()/clk_set_rate()'.

The watermarks should be changed in accordance to the actual EMC clock rate. Given that EMC rate could be changed by something else than the devfreq driver, we need to re-adjust the watermarks to the actual values after the EMC rate-change completion. Please note that the clock notifier uses the POST_RATE_CHANGE event that happens only after successful completion of EMC clock rate change.

#31 lore raw

Re: [PATCH v2 12/19] PM / devfreq: tegra: Avoid inconsistency of current frequency value

In reply to Chanwoo Choi

16.04.2019 10:15, Chanwoo Choi пишет:

Quoted 40 lines

Hi,

Quoted 37 lines · Dmitry Osipenko

The frequency value potentially could change in-between. It doesn't cause any real problem at all right now, but that could change in the future. Hence let's avoid the inconsistency.

---
 drivers/devfreq/tegra-devfreq.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +4 -2
index a668e4fbc874..f1a6f951813a 100644
@@ -496,13 +496,15 @@ static int tegra_devfreq_get_dev_status(struct device *dev,
 {
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct tegra_devfreq_device *actmon_dev;
+	unsigned long cur_freq;
 
-	stat->current_frequency = tegra->cur_freq * KHZ;
+	cur_freq = READ_ONCE(tegra->cur_freq);
 
 	/* To be used by the tegra governor */
 	stat->private_data = tegra;
 
 	/* The below are to be used by the other governors */
+	stat->current_frequency = cur_freq * KHZ;
 
 	actmon_dev = &tegra->devices[MCALL];
 
@@ -513,7 +515,7 @@ static int tegra_devfreq_get_dev_status(struct device *dev,
 	stat->busy_time *= 100 / BUS_SATURATION_RATIO;
 
 	/* Number of cycles in a sampling period */
-	stat->total_time = ACTMON_SAMPLING_PERIOD * tegra->cur_freq;
+	stat->total_time = ACTMON_SAMPLING_PERIOD * cur_freq;
 
Quoted 7 lines

drivers/devfreq/tegra-devfreq.c static int tegra_devfreq_get_dev_status(struct device *dev, · line 521

Quoted 3 lines
drivers/devfreq/tegra-devfreq.c +0 -0
 	stat->busy_time = min(stat->busy_time, stat->total_time);
 

The read/write access of tegra->cur_freq is in the single routine of update_devfreq() as following. I think that there are no any potential problem about the inconsistency of tegra->cur_freq.

No, that's wrong assumption. The tegra->cur_freq is changed by the clock notifier that runs asynchronously with the devfreq driver when EMC clock rate is changed by something else in the kernel.

Quoted 4 lines

IMHO, if there are no any problem now, I'm not sure that we need to apply this patch.

update_devfreq()

Quoted 18 lines
{
	devfreq->governor->get_target_freq()
		devfreq_update_stats(devfreq)
			tegra_devfreq_get_dev_status()
				stat->current_frequency = tegra->cur_freq * KHZ;

	devfreq_set_target()
		tegra_devfreq_target()
			clk_set_min_rate(emc_rate, )
				tegra_actmon_rate_notify_cb()
					tegra->cur_freq = data->new_rate / KHZ;
		
			clk_set_rate(emc_rate, )
				tegra_actmon_rate_notify_cb()
					tegra->cur_freq = data->new_rate / KHZ;
}

The cur_freq value is changed by the clock notifier that runs asynchronously with the rest of the devfreq driver. Hence potentially compiler may generate two separate fetches of the cur_freq value, then the clock rate could be changed by other CPU core simultaneously with tegra_devfreq_get_dev_status() or kernel may re-schedule preemptively, changing the clock rate in-between of the two fetches.

#39 lore raw

Re: [PATCH v2 15/19] PM / devfreq: tegra: Synchronize IRQ after masking it in hardware

In reply to Chanwoo Choi

16.04.2019 10:41, Chanwoo Choi пишет:

Quoted 9 lines

Hi,

In this patchset,
the patch11 adds new 'disable_interrupt' goto statement and then patch15 removes 'disable_interrupt' goto statement again. Actually, it is inefficient.

If you change the order of patches,
you could remove this stuff.

Okay, I'll change that in v3. Thanks.

#43 lore raw

Re: [PATCH v2 17/19] PM / devfreq: tegra: Support Tegra30

In reply to Chanwoo Choi

16.04.2019 10:48, Chanwoo Choi пишет:

Quoted 27 lines

Hi,

Quoted 15 lines · Dmitry Osipenko

The devfreq driver can be used on Tegra30 without any code change and it works perfectly fine, the default Tegra124 parameters are good enough for Tegra30.

---
 drivers/devfreq/Kconfig         | 4 ++--
 drivers/devfreq/tegra-devfreq.c | 1 +
 2 files changed, 3 insertions(+), 2 deletions(-)
drivers/devfreq/Kconfig +0 -0
index a78dffe603c1..78c33ddd4eea 100644

drivers/devfreq/Kconfig config ARM_EXYNOS_BUS_DEVFREQ · line 95

Quoted 6 lines
drivers/devfreq/Kconfig +1 -2
@@ -92,8 +92,8 @@ config ARM_EXYNOS_BUS_DEVFREQ
 	  This does not yet operate with optimal voltages.
 
 config ARM_TEGRA_DEVFREQ
-	tristate "Tegra DEVFREQ Driver"
-	depends on ARCH_TEGRA_124_SOC
+	tristate "NVIDIA Tegra30+ DEVFREQ Driver"

Looks good to me. But, I have a question.

Quoted 5 lines

'Tegra30+' expression is enough in order to indicate the support for both tegra30 and tegra124?

In my case, it is difficult to catch the meaning
of which tegra30+ support the kind of tegra124.

The ACTMON serves Tegra30, Tegra124 and Tegra210 SoC's. The later Tegra generations also have the ACTMON hardware unit, but most of actual drivers have been moved into firmware on the later gens, including the ACTMON driver. I'll change the wording to "Tegra30/124/210" for clarity. Note that Tegra210 support isn't implemented at the moment by the driver, but that may change in the future.

Quoted 11 lines
Quoted 11 lines
drivers/devfreq/Kconfig +1 -0
+	depends on ARCH_TEGRA
 	select PM_OPP
 	help
 	  This adds the DEVFREQ driver for the Tegra family of SoCs.
drivers/devfreq/tegra-devfreq.c +0 -0
index f0711d5ad27d..0a2465a58cf5 100644
@@ -719,6 +719,7 @@ static int tegra_devfreq_remove(struct platform_device *pdev)
 }
 
Quoted 9 lines

drivers/devfreq/tegra-devfreq.c static int tegra_devfreq_remove(struct platform_device *pdev) · line 725

Quoted 6 lines
drivers/devfreq/tegra-devfreq.c +1 -0
 static const struct of_device_id tegra_devfreq_of_match[] = {
+	{ .compatible = "nvidia,tegra30-actmon" },
 	{ .compatible = "nvidia,tegra124-actmon" },
 	{ },
 };

Thanks.

#47 lore raw

Re: [PATCH v2 18/19] PM / devfreq: tegra: Enable COMPILE_TEST for the driver

In reply to Chanwoo Choi

16.04.2019 10:43, Chanwoo Choi пишет:

Quoted 15 lines · Dmitry Osipenko
Quoted 14 lines

The driver's compilation doesn't have any specific dependencies, hence the COMPILE_TEST option can be supported in Kconfig.

---
 drivers/devfreq/Kconfig | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/devfreq/Kconfig +0 -0
index 78c33ddd4eea..bd6652863e7d 100644
@@ -93,7 +93,7 @@ config ARM_EXYNOS_BUS_DEVFREQ
 
Quoted 12 lines

drivers/devfreq/Kconfig config ARM_EXYNOS_BUS_DEVFREQ · line 99

Quoted 8 lines
drivers/devfreq/Kconfig +1 -1
 config ARM_TEGRA_DEVFREQ
 	tristate "NVIDIA Tegra30+ DEVFREQ Driver"
-	depends on ARCH_TEGRA
+	depends on ARCH_TEGRA || COMPILE_TEST
 	select PM_OPP
 	help
 	  This adds the DEVFREQ driver for the Tegra family of SoCs.

I think that it doesn't need to make it a separate patch. You can merge this patch to patch17 and then drop this patch.

I'd prefer to have this change as a separate patch because this is a logically distinct change from the patch 17.

#51 lore raw

Re: [PATCH v2 19/19] PM / devfreq: Introduce driver for NVIDIA Tegra20

In reply to Chanwoo Choi

16.04.2019 11:31, Chanwoo Choi пишет:

Quoted 66 lines

Hi,

Quoted 63 lines · Dmitry Osipenko

Add devfreq driver for NVIDIA Tegra20 SoC's. The driver periodically reads out Memory Controller counters and adjusts memory frequency based on the memory clients activity.

---
 MAINTAINERS                       |   8 ++
 drivers/devfreq/Kconfig           |  10 ++
 drivers/devfreq/Makefile          |   1 +
 drivers/devfreq/tegra20-devfreq.c | 177 ++++++++++++++++++++++++++++++
 4 files changed, 196 insertions(+)
 create mode 100644 drivers/devfreq/tegra20-devfreq.c
MAINTAINERS +8 -0
index 80b59db1b6e4..91f475ec4545 100644
@@ -10056,6 +10056,14 @@ F:	include/linux/memblock.h
 F:	mm/memblock.c
 F:	Documentation/core-api/boot-time-mm.rst
 
+MEMORY FREQUENCY SCALING DRIVER FOR NVIDIA TEGRA20
+M:	Dmitry Osipenko <[email protected]>
+L:	[email protected]
+L:	[email protected]
+T:	git git://git.kernel.org/pub/scm/linux/kernel/git/mzx/devfreq.git
+S:	Maintained
+F:	drivers/devfreq/tegra20-devfreq.c
+
 MEMORY MANAGEMENT
 L:	[email protected]
 W:	http://www.linux-mm.org
drivers/devfreq/Kconfig +10 -0
index bd6652863e7d..af4c86c4e0f6 100644
@@ -100,6 +100,16 @@ config ARM_TEGRA_DEVFREQ
 	  It reads ACTMON counters of memory controllers and adjusts the
 	  operating frequencies and voltages with OPP support.
 
+config ARM_TEGRA20_DEVFREQ
+	tristate "NVIDIA Tegra20 DEVFREQ Driver"
+	depends on (TEGRA_MC && TEGRA20_EMC) || COMPILE_TEST
+	select DEVFREQ_GOV_SIMPLE_ONDEMAND
+	select PM_OPP
+	help
+	  This adds the DEVFREQ driver for the Tegra20 family of SoCs.
+	  It reads Memory Controller counters and adjusts the operating
+	  frequencies and voltages with OPP support.
+
 config ARM_RK3399_DMC_DEVFREQ
 	tristate "ARM RK3399 DMC DEVFREQ Driver"
 	depends on ARCH_ROCKCHIP
drivers/devfreq/Makefile +1 -0
index 32b8d4d3f12c..6fcc5596b8b7 100644
@@ -11,6 +11,7 @@ obj-$(CONFIG_DEVFREQ_GOV_PASSIVE)	+= governor_passive.o
 obj-$(CONFIG_ARM_EXYNOS_BUS_DEVFREQ)	+= exynos-bus.o
 obj-$(CONFIG_ARM_RK3399_DMC_DEVFREQ)	+= rk3399_dmc.o
 obj-$(CONFIG_ARM_TEGRA_DEVFREQ)		+= tegra-devfreq.o
+obj-$(CONFIG_ARM_TEGRA20_DEVFREQ)	+= tegra20-devfreq.o
 
Quoted 16 lines
Quoted 8 lines
drivers/devfreq/Makefile +0 -0
 # DEVFREQ Event Drivers
 obj-$(CONFIG_PM_DEVFREQ_EVENT)		+= event/
drivers/devfreq/tegra20-devfreq.c new file +0 -0
new file mode 100644
index 000000000000..18c9aad7a9d7

drivers/devfreq/tegra20-devfreq.c line 6

Quoted 6 lines
drivers/devfreq/tegra20-devfreq.c new file +6 -0
@@ -0,0 +1,177 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * NVIDIA Tegra20 devfreq driver
+ *
+ * Author: Dmitry Osipenko <[email protected]>
+ */

It doesn't any "Copyright (c) 2019 ..." sentence.

I'll add one in v3.

Quoted 16 lines

drivers/devfreq/tegra20-devfreq.c line 18

Quoted 12 lines
drivers/devfreq/tegra20-devfreq.c new file +12 -0
+
+#include <linux/clk.h>
+#include <linux/devfreq.h>
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of_device.h>
+#include <linux/platform_device.h>
+#include <linux/pm_opp.h>
+#include <linux/slab.h>
+
+#include <soc/tegra/mc.h>

I can find the '<soc/tegra/mc.h>' header file
on mainline branch. But mc.h is included in linux-next.git.

Quoted 4 lines

If you don't share the patch related to mc.h,
the kernel build will be failed when apply it the devfreq.git on final step. Actually, it should make the immutable branch between two related maintainers in order to remove the build fail.

The '<soc/tegra/mc.h>' header file exists since v3.18 [0], seems you just missed something.

[0] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/include/soc/tegra/mc.h?h=v3.18&id=8918465163171322c77a19d5258a95f56d89d2e4

Quoted 34 lines
Quoted 34 lines
drivers/devfreq/tegra20-devfreq.c new file +34 -0
+
+#include "governor.h"
+
+#define MC_STAT_CONTROL				0x90
+#define MC_STAT_EMC_CLOCK_LIMIT			0xa0
+#define MC_STAT_EMC_CLOCKS			0xa4
+#define MC_STAT_EMC_CONTROL			0xa8
+#define MC_STAT_EMC_COUNT			0xb8
+
+#define EMC_GATHER_CLEAR			(1 << 8)
+#define EMC_GATHER_ENABLE			(3 << 8)
+
+struct tegra_devfreq {
+	struct devfreq *devfreq;
+	struct clk *clk;
+	void __iomem *regs;
+};
+
+static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
+				u32 flags)
+{
+	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
+	struct dev_pm_opp *opp;
+	unsigned long rate;
+	int err;
+
+	opp = devfreq_recommended_opp(dev, freq, flags);
+	if (IS_ERR(opp))
+		return PTR_ERR(opp);
+
+	rate = dev_pm_opp_get_freq(opp);
+	dev_pm_opp_put(opp);
+
+	err = clk_set_min_rate(tegra->clk, rate);
Quoted 8 lines

drivers/devfreq/tegra20-devfreq.c line 57

Quoted 5 lines
drivers/devfreq/tegra20-devfreq.c new file +5 -0
+	if (err)
+		return err;
+
+	err = clk_set_rate(tegra->clk, 0);
+	if (err)

When fail happen, I think that you have to control
the restoring sequence for previous operation like clk_set_min_rate().

Okay, thanks.

Quoted 55 lines
Quoted 55 lines
drivers/devfreq/tegra20-devfreq.c new file +55 -0
+		return err;
+
+	return 0;
+}
+
+static int tegra_devfreq_get_dev_status(struct device *dev,
+					struct devfreq_dev_status *stat)
+{
+	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
+
+	stat->busy_time = readl_relaxed(tegra->regs + MC_STAT_EMC_COUNT);
+	stat->total_time = readl_relaxed(tegra->regs + MC_STAT_EMC_CLOCKS) / 8;
+	stat->current_frequency = clk_get_rate(tegra->clk);
+
+	writel_relaxed(EMC_GATHER_CLEAR, tegra->regs + MC_STAT_CONTROL);
+	writel_relaxed(EMC_GATHER_ENABLE, tegra->regs + MC_STAT_CONTROL);
+
+	return 0;
+}
+
+static struct devfreq_dev_profile tegra_devfreq_profile = {
+	.polling_ms	= 500,
+	.target		= tegra_devfreq_target,
+	.get_dev_status	= tegra_devfreq_get_dev_status,
+};
+
+static struct tegra_mc *tegra_get_memory_controller(void)
+{
+	struct platform_device *pdev;
+	struct device_node *np;
+	struct tegra_mc *mc;
+
+	np = of_find_compatible_node(NULL, NULL, "nvidia,tegra20-mc-gart");
+	if (!np)
+		return ERR_PTR(-ENOENT);
+
+	pdev = of_find_device_by_node(np);
+	of_node_put(np);
+	if (!pdev)
+		return ERR_PTR(-ENODEV);
+
+	mc = platform_get_drvdata(pdev);
+	if (!mc)
+		return ERR_PTR(-EPROBE_DEFER);
+
+	return mc;
+}
+
+static int tegra_devfeq_probe(struct platform_device *pdev)
+{
+	struct tegra_devfreq *tegra;
+	struct tegra_mc *mc;
+	unsigned long max_rate;
+	unsigned long rate;
+	int err;
Quoted 8 lines

drivers/devfreq/tegra20-devfreq.c line 117

Quoted 5 lines
drivers/devfreq/tegra20-devfreq.c new file +5 -0
+
+	mc = tegra_get_memory_controller();
+	if (IS_ERR(mc)) {
+		err = PTR_ERR(mc);
+		dev_err(&pdev->dev, "failed to get mc: %d\n", err);

How about using 'memory controller' instead of 'mc'? Because 'mc' is not standard expression.

Sounds good, thanks.

Quoted 8 lines
Quoted 8 lines
drivers/devfreq/tegra20-devfreq.c new file +8 -0
+		return err;
+	}
+
+	tegra = devm_kzalloc(&pdev->dev, sizeof(*tegra), GFP_KERNEL);
+	if (!tegra)
+		return -ENOMEM;
+
+	tegra->clk = devm_clk_get(&pdev->dev, "emc");
Quoted 8 lines

drivers/devfreq/tegra20-devfreq.c line 130

Quoted 5 lines
drivers/devfreq/tegra20-devfreq.c new file +5 -0
+	if (IS_ERR(tegra->clk)) {
+		err = PTR_ERR(tegra->clk);
+		dev_err(&pdev->dev, "failed to get emc clock: %d\n", err);
+		return err;
+	}

Don't you need to enable the 'emc' clock'?
Because this patch doesn't enable this clock.

EMC is a system critical clock (marked as critical by the clock driver), it is guaranteed to be always enabled. I think it is fine to omit clock-enabling in this case.

Quoted 9 lines
Quoted 9 lines
drivers/devfreq/tegra20-devfreq.c new file +9 -0
+
+	tegra->regs = mc->regs;
+
+	max_rate = clk_round_rate(tegra->clk, ULONG_MAX);
+
+	for (rate = 0; rate <= max_rate; rate++) {
+		rate = clk_round_rate(tegra->clk, rate);
+		dev_pm_opp_add(&pdev->dev, rate, 0);
+	}
Quoted 8 lines

drivers/devfreq/tegra20-devfreq.c line 143

Quoted 4 lines
drivers/devfreq/tegra20-devfreq.c new file +4 -0
+
+	writel_relaxed(0x00000000, tegra->regs + MC_STAT_CONTROL);
+	writel_relaxed(0x00000000, tegra->regs + MC_STAT_EMC_CONTROL);
+	writel_relaxed(0xffffffff, tegra->regs + MC_STAT_EMC_CLOCK_LIMIT);

You better to add the comments of what are the meaning of 0x00000000/0xffffffff. Without the detailed comments, it is difficult to understand of meaning.

Okay. Although the complete registers description is available in public and someone really curious could just google for the full details.

Quoted 22 lines
Quoted 22 lines
drivers/devfreq/tegra20-devfreq.c new file +22 -0
+
+	platform_set_drvdata(pdev, tegra);
+
+	tegra->devfreq = devfreq_add_device(&pdev->dev, &tegra_devfreq_profile,
+					    DEVFREQ_GOV_SIMPLE_ONDEMAND, NULL);
+	if (IS_ERR(tegra->devfreq))
+		return PTR_ERR(tegra->devfreq);
+
+	return 0;
+}
+
+static int tegra_devfreq_remove(struct platform_device *pdev)
+{
+	struct tegra_devfreq *tegra = platform_get_drvdata(pdev);
+
+	devfreq_remove_device(tegra->devfreq);
+	dev_pm_opp_remove_all_dynamic(&pdev->dev);
+
+	return 0;
+}
+
+static struct platform_driver tegra_devfeq_driver = {
Quoted 8 lines

drivers/devfreq/tegra20-devfreq.c line 169

Quoted 4 lines
drivers/devfreq/tegra20-devfreq.c new file +4 -0
+	.probe		= tegra_devfeq_probe,
+	.remove		= tegra_devfreq_remove,
+	.driver		= {
+		.name	= "tegra20-devfreq",

How can you bind this driver without compatible name for Devicetree? And I tried to find the name ("tegra20-devfreq") in
the MFD drivers.

As I wrote in the cover-letter, the device for the driver will be created by the EMC driver. I'll send out the patch that creates the device later on because of EMC driver patch-series interdependence, for now I have that patch here [1] if you're curious how it looks like.

[1] https://github.com/grate-driver/linux/commit/c34440402bb4fb365647bf43166e85562c124273

maintainer
#52 lore raw

Re: [PATCH v2 19/19] PM / devfreq: Introduce driver for NVIDIA Tegra20

Hi,

Quoted 102 lines · Dmitry Osipenko

16.04.2019 11:31, Chanwoo Choi пишет:

Quoted 66 lines

Hi,

Quoted 63 lines · Dmitry Osipenko

Add devfreq driver for NVIDIA Tegra20 SoC's. The driver periodically reads out Memory Controller counters and adjusts memory frequency based on the memory clients activity.

---
 MAINTAINERS                       |   8 ++
 drivers/devfreq/Kconfig           |  10 ++
 drivers/devfreq/Makefile          |   1 +
 drivers/devfreq/tegra20-devfreq.c | 177 ++++++++++++++++++++++++++++++
 4 files changed, 196 insertions(+)
 create mode 100644 drivers/devfreq/tegra20-devfreq.c
MAINTAINERS +8 -0
index 80b59db1b6e4..91f475ec4545 100644
@@ -10056,6 +10056,14 @@ F:	include/linux/memblock.h
 F:	mm/memblock.c
 F:	Documentation/core-api/boot-time-mm.rst
 
+MEMORY FREQUENCY SCALING DRIVER FOR NVIDIA TEGRA20
+M:	Dmitry Osipenko <[email protected]>
+L:	[email protected]
+L:	[email protected]
+T:	git git://git.kernel.org/pub/scm/linux/kernel/git/mzx/devfreq.git
+S:	Maintained
+F:	drivers/devfreq/tegra20-devfreq.c
+
 MEMORY MANAGEMENT
 L:	[email protected]
 W:	http://www.linux-mm.org
drivers/devfreq/Kconfig +10 -0
index bd6652863e7d..af4c86c4e0f6 100644
@@ -100,6 +100,16 @@ config ARM_TEGRA_DEVFREQ
 	  It reads ACTMON counters of memory controllers and adjusts the
 	  operating frequencies and voltages with OPP support.
 
+config ARM_TEGRA20_DEVFREQ
+	tristate "NVIDIA Tegra20 DEVFREQ Driver"
+	depends on (TEGRA_MC && TEGRA20_EMC) || COMPILE_TEST
+	select DEVFREQ_GOV_SIMPLE_ONDEMAND
+	select PM_OPP
+	help
+	  This adds the DEVFREQ driver for the Tegra20 family of SoCs.
+	  It reads Memory Controller counters and adjusts the operating
+	  frequencies and voltages with OPP support.
+
 config ARM_RK3399_DMC_DEVFREQ
 	tristate "ARM RK3399 DMC DEVFREQ Driver"
 	depends on ARCH_ROCKCHIP
drivers/devfreq/Makefile +1 -0
index 32b8d4d3f12c..6fcc5596b8b7 100644
@@ -11,6 +11,7 @@ obj-$(CONFIG_DEVFREQ_GOV_PASSIVE)	+= governor_passive.o
 obj-$(CONFIG_ARM_EXYNOS_BUS_DEVFREQ)	+= exynos-bus.o
 obj-$(CONFIG_ARM_RK3399_DMC_DEVFREQ)	+= rk3399_dmc.o
 obj-$(CONFIG_ARM_TEGRA_DEVFREQ)		+= tegra-devfreq.o
+obj-$(CONFIG_ARM_TEGRA20_DEVFREQ)	+= tegra20-devfreq.o
 
Quoted 16 lines
Quoted 8 lines
drivers/devfreq/Makefile +0 -0
 # DEVFREQ Event Drivers
 obj-$(CONFIG_PM_DEVFREQ_EVENT)		+= event/
drivers/devfreq/tegra20-devfreq.c new file +0 -0
new file mode 100644
index 000000000000..18c9aad7a9d7

drivers/devfreq/tegra20-devfreq.c line 6

Quoted 6 lines
drivers/devfreq/tegra20-devfreq.c new file +6 -0
@@ -0,0 +1,177 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * NVIDIA Tegra20 devfreq driver
+ *
+ * Author: Dmitry Osipenko <[email protected]>
+ */

It doesn't any "Copyright (c) 2019 ..." sentence.

I'll add one in v3.

Quoted 16 lines

drivers/devfreq/tegra20-devfreq.c line 18

Quoted 12 lines
drivers/devfreq/tegra20-devfreq.c new file +12 -0
+
+#include <linux/clk.h>
+#include <linux/devfreq.h>
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of_device.h>
+#include <linux/platform_device.h>
+#include <linux/pm_opp.h>
+#include <linux/slab.h>
+
+#include <soc/tegra/mc.h>

I can find the '<soc/tegra/mc.h>' header file
on mainline branch. But mc.h is included in linux-next.git.

Quoted 8 lines
Quoted 4 lines

If you don't share the patch related to mc.h,
the kernel build will be failed when apply it the devfreq.git on final step. Actually, it should make the immutable branch between two related maintainers in order to remove the build fail.

The '<soc/tegra/mc.h>' header file exists since v3.18 [0], seems you just missed something.

[0] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/include/soc/tegra/mc.h?h=v3.18&id=8918465163171322c77a19d5258a95f56d89d2e4

Sorry. It is my missing point. When I tried to find it, it is included in the mainline kernel.

Quoted 129 lines
Quoted 35 lines
Quoted 35 lines
drivers/devfreq/tegra20-devfreq.c new file +34 -0
+
+#include "governor.h"
+
+#define MC_STAT_CONTROL				0x90
+#define MC_STAT_EMC_CLOCK_LIMIT			0xa0
+#define MC_STAT_EMC_CLOCKS			0xa4
+#define MC_STAT_EMC_CONTROL			0xa8
+#define MC_STAT_EMC_COUNT			0xb8
+
+#define EMC_GATHER_CLEAR			(1 << 8)
+#define EMC_GATHER_ENABLE			(3 << 8)
+
+struct tegra_devfreq {
+	struct devfreq *devfreq;
+	struct clk *clk;
+	void __iomem *regs;
+};
+
+static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
+				u32 flags)
+{
+	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
+	struct dev_pm_opp *opp;
+	unsigned long rate;
+	int err;
+
+	opp = devfreq_recommended_opp(dev, freq, flags);
+	if (IS_ERR(opp))
+		return PTR_ERR(opp);
+
+	rate = dev_pm_opp_get_freq(opp);
+	dev_pm_opp_put(opp);
+
+	err = clk_set_min_rate(tegra->clk, rate);
Quoted 8 lines

drivers/devfreq/tegra20-devfreq.c line 57

Quoted 5 lines
drivers/devfreq/tegra20-devfreq.c new file +5 -0
+	if (err)
+		return err;
+
+	err = clk_set_rate(tegra->clk, 0);
+	if (err)

When fail happen, I think that you have to control
the restoring sequence for previous operation like clk_set_min_rate().

Okay, thanks.

Quoted 55 lines
Quoted 55 lines
drivers/devfreq/tegra20-devfreq.c new file +55 -0
+		return err;
+
+	return 0;
+}
+
+static int tegra_devfreq_get_dev_status(struct device *dev,
+					struct devfreq_dev_status *stat)
+{
+	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
+
+	stat->busy_time = readl_relaxed(tegra->regs + MC_STAT_EMC_COUNT);
+	stat->total_time = readl_relaxed(tegra->regs + MC_STAT_EMC_CLOCKS) / 8;
+	stat->current_frequency = clk_get_rate(tegra->clk);
+
+	writel_relaxed(EMC_GATHER_CLEAR, tegra->regs + MC_STAT_CONTROL);
+	writel_relaxed(EMC_GATHER_ENABLE, tegra->regs + MC_STAT_CONTROL);
+
+	return 0;
+}
+
+static struct devfreq_dev_profile tegra_devfreq_profile = {
+	.polling_ms	= 500,
+	.target		= tegra_devfreq_target,
+	.get_dev_status	= tegra_devfreq_get_dev_status,
+};
+
+static struct tegra_mc *tegra_get_memory_controller(void)
+{
+	struct platform_device *pdev;
+	struct device_node *np;
+	struct tegra_mc *mc;
+
+	np = of_find_compatible_node(NULL, NULL, "nvidia,tegra20-mc-gart");
+	if (!np)
+		return ERR_PTR(-ENOENT);
+
+	pdev = of_find_device_by_node(np);
+	of_node_put(np);
+	if (!pdev)
+		return ERR_PTR(-ENODEV);
+
+	mc = platform_get_drvdata(pdev);
+	if (!mc)
+		return ERR_PTR(-EPROBE_DEFER);
+
+	return mc;
+}
+
+static int tegra_devfeq_probe(struct platform_device *pdev)
+{
+	struct tegra_devfreq *tegra;
+	struct tegra_mc *mc;
+	unsigned long max_rate;
+	unsigned long rate;
+	int err;
Quoted 8 lines

drivers/devfreq/tegra20-devfreq.c line 117

Quoted 5 lines
drivers/devfreq/tegra20-devfreq.c new file +5 -0
+
+	mc = tegra_get_memory_controller();
+	if (IS_ERR(mc)) {
+		err = PTR_ERR(mc);
+		dev_err(&pdev->dev, "failed to get mc: %d\n", err);

How about using 'memory controller' instead of 'mc'? Because 'mc' is not standard expression.

Sounds good, thanks.

Quoted 8 lines
Quoted 8 lines
drivers/devfreq/tegra20-devfreq.c new file +8 -0
+		return err;
+	}
+
+	tegra = devm_kzalloc(&pdev->dev, sizeof(*tegra), GFP_KERNEL);
+	if (!tegra)
+		return -ENOMEM;
+
+	tegra->clk = devm_clk_get(&pdev->dev, "emc");
Quoted 8 lines

drivers/devfreq/tegra20-devfreq.c line 130

Quoted 5 lines
drivers/devfreq/tegra20-devfreq.c new file +5 -0
+	if (IS_ERR(tegra->clk)) {
+		err = PTR_ERR(tegra->clk);
+		dev_err(&pdev->dev, "failed to get emc clock: %d\n", err);
+		return err;
+	}

Don't you need to enable the 'emc' clock'?
Because this patch doesn't enable this clock.

Quoted 1 line

EMC is a system critical clock (marked as critical by the clock driver), it is guaranteed to be always enabled. I think it is fine to omit clock-enabling in this case.

If you think don't need to enable it due to the critical clock, instead, please add the comment about that emc clock is critical. In my case, it looks like the strange use-case because this driver doesn't contain the any enable code for clock.

Quoted 52 lines
Quoted 10 lines
Quoted 10 lines
drivers/devfreq/tegra20-devfreq.c new file +9 -0
+
+	tegra->regs = mc->regs;
+
+	max_rate = clk_round_rate(tegra->clk, ULONG_MAX);
+
+	for (rate = 0; rate <= max_rate; rate++) {
+		rate = clk_round_rate(tegra->clk, rate);
+		dev_pm_opp_add(&pdev->dev, rate, 0);
+	}
Quoted 8 lines

drivers/devfreq/tegra20-devfreq.c line 143

Quoted 4 lines
drivers/devfreq/tegra20-devfreq.c new file +4 -0
+
+	writel_relaxed(0x00000000, tegra->regs + MC_STAT_CONTROL);
+	writel_relaxed(0x00000000, tegra->regs + MC_STAT_EMC_CONTROL);
+	writel_relaxed(0xffffffff, tegra->regs + MC_STAT_EMC_CLOCK_LIMIT);

You better to add the comments of what are the meaning of 0x00000000/0xffffffff. Without the detailed comments, it is difficult to understand of meaning.

Okay. Although the complete registers description is available in public and someone really curious could just google for the full details.

Quoted 22 lines
Quoted 22 lines
drivers/devfreq/tegra20-devfreq.c new file +22 -0
+
+	platform_set_drvdata(pdev, tegra);
+
+	tegra->devfreq = devfreq_add_device(&pdev->dev, &tegra_devfreq_profile,
+					    DEVFREQ_GOV_SIMPLE_ONDEMAND, NULL);
+	if (IS_ERR(tegra->devfreq))
+		return PTR_ERR(tegra->devfreq);
+
+	return 0;
+}
+
+static int tegra_devfreq_remove(struct platform_device *pdev)
+{
+	struct tegra_devfreq *tegra = platform_get_drvdata(pdev);
+
+	devfreq_remove_device(tegra->devfreq);
+	dev_pm_opp_remove_all_dynamic(&pdev->dev);
+
+	return 0;
+}
+
+static struct platform_driver tegra_devfeq_driver = {
Quoted 8 lines

drivers/devfreq/tegra20-devfreq.c line 169

Quoted 4 lines
drivers/devfreq/tegra20-devfreq.c new file +4 -0
+	.probe		= tegra_devfeq_probe,
+	.remove		= tegra_devfreq_remove,
+	.driver		= {
+		.name	= "tegra20-devfreq",

How can you bind this driver without compatible name for Devicetree? And I tried to find the name ("tegra20-devfreq") in
the MFD drivers.

Quoted 1 line

As I wrote in the cover-letter, the device for the driver will be created by the EMC driver. I'll

send out the patch that creates the device later on because of EMC driver patch-series interdependence, for now I have that patch here [1] if you're curious how it looks like.

OK. I understand. Thanks.

Quoted 4 lines

[1] https://github.com/grate-driver/linux/commit/c34440402bb4fb365647bf43166e85562c124273

Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer Reviewed-by
#48 lore raw

Re: [PATCH v2 18/19] PM / devfreq: tegra: Enable COMPILE_TEST for the driver

In reply to Dmitry Osipenko
Quoted 29 lines · Dmitry Osipenko

16.04.2019 10:43, Chanwoo Choi пишет:

Quoted 15 lines · Dmitry Osipenko
Quoted 14 lines

The driver's compilation doesn't have any specific dependencies, hence the COMPILE_TEST option can be supported in Kconfig.

---
 drivers/devfreq/Kconfig | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/devfreq/Kconfig +0 -0
index 78c33ddd4eea..bd6652863e7d 100644
@@ -93,7 +93,7 @@ config ARM_EXYNOS_BUS_DEVFREQ
 
Quoted 12 lines

drivers/devfreq/Kconfig config ARM_EXYNOS_BUS_DEVFREQ · line 99

Quoted 8 lines
drivers/devfreq/Kconfig +1 -1
 config ARM_TEGRA_DEVFREQ
 	tristate "NVIDIA Tegra30+ DEVFREQ Driver"
-	depends on ARCH_TEGRA
+	depends on ARCH_TEGRA || COMPILE_TEST
 	select PM_OPP
 	help
 	  This adds the DEVFREQ driver for the Tegra family of SoCs.

I think that it doesn't need to make it a separate patch. You can merge this patch to patch17 and then drop this patch.

Quoted 3 lines

I'd prefer to have this change as a separate patch because this is a logically distinct change from the patch 17.

OK.

Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer
#44 lore raw

Re: [PATCH v2 17/19] PM / devfreq: tegra: Support Tegra30

In reply to Dmitry Osipenko
Quoted 28 lines · Dmitry Osipenko

16.04.2019 10:48, Chanwoo Choi пишет:

Quoted 27 lines

Hi,

Quoted 15 lines · Dmitry Osipenko

The devfreq driver can be used on Tegra30 without any code change and it works perfectly fine, the default Tegra124 parameters are good enough for Tegra30.

---
 drivers/devfreq/Kconfig         | 4 ++--
 drivers/devfreq/tegra-devfreq.c | 1 +
 2 files changed, 3 insertions(+), 2 deletions(-)
drivers/devfreq/Kconfig +0 -0
index a78dffe603c1..78c33ddd4eea 100644

drivers/devfreq/Kconfig config ARM_EXYNOS_BUS_DEVFREQ · line 95

Quoted 6 lines
drivers/devfreq/Kconfig +1 -2
@@ -92,8 +92,8 @@ config ARM_EXYNOS_BUS_DEVFREQ
 	  This does not yet operate with optimal voltages.
 
 config ARM_TEGRA_DEVFREQ
-	tristate "Tegra DEVFREQ Driver"
-	depends on ARCH_TEGRA_124_SOC
+	tristate "NVIDIA Tegra30+ DEVFREQ Driver"

Looks good to me. But, I have a question.

Quoted 7 lines
Quoted 5 lines

'Tegra30+' expression is enough in order to indicate the support for both tegra30 and tegra124?

In my case, it is difficult to catch the meaning
of which tegra30+ support the kind of tegra124.

The ACTMON serves Tegra30, Tegra124 and Tegra210 SoC's. The later Tegra generations also have the ACTMON hardware unit, but most of actual drivers have been moved into firmware on the later gens, including the ACTMON driver. I'll change the wording to "Tegra30/124/210" for clarity. Note that Tegra210 support isn't implemented at the moment by the driver, but that may change in the future.

OK. Thanks.

Quoted 25 lines
Quoted 12 lines
Quoted 12 lines
drivers/devfreq/Kconfig +1 -0
+	depends on ARCH_TEGRA
 	select PM_OPP
 	help
 	  This adds the DEVFREQ driver for the Tegra family of SoCs.
drivers/devfreq/tegra-devfreq.c +0 -0
index f0711d5ad27d..0a2465a58cf5 100644
@@ -719,6 +719,7 @@ static int tegra_devfreq_remove(struct platform_device *pdev)
 }
 
Quoted 9 lines

drivers/devfreq/tegra-devfreq.c static int tegra_devfreq_remove(struct platform_device *pdev) · line 725

Quoted 6 lines
drivers/devfreq/tegra-devfreq.c +1 -0
 static const struct of_device_id tegra_devfreq_of_match[] = {
+	{ .compatible = "nvidia,tegra30-actmon" },
 	{ .compatible = "nvidia,tegra124-actmon" },
 	{ },
 };

Thanks.

Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer
#26 lore raw

Re: [PATCH v2 10/19] PM / devfreq: tegra: Drop primary interrupt handler

In reply to Dmitry Osipenko

Hi,

Quoted 106 lines · Dmitry Osipenko

16.04.2019 8:56, Chanwoo Choi пишет:

Quoted 88 lines

Hi,

It looks good to me to drop the primary interrupt handler but I have some comments. Please check it.

Quoted 68 lines · Dmitry Osipenko

There is no real need in the primary interrupt handler, hence move everything to the secondary (threaded) handler. In a result locking is consistent now and there are no potential races with the interrupt handler because it is protected with the devfreq's mutex.

---
 drivers/devfreq/tegra-devfreq.c | 61 ++++++++++++---------------------
 1 file changed, 22 insertions(+), 39 deletions(-)
drivers/devfreq/tegra-devfreq.c +1 -25
index 2a1464098200..69b557df5084 100644
@@ -144,7 +144,6 @@ static struct tegra_devfreq_device_config actmon_device_configs[] = {
 struct tegra_devfreq_device {
 	const struct tegra_devfreq_device_config *config;
 	void __iomem *regs;
-	spinlock_t lock;
 
 	/* Average event count sampled in the last interrupt */
 	u32 avg_count;
@@ -249,11 +248,8 @@ static void actmon_write_barrier(struct tegra_devfreq *tegra)
 static void actmon_isr_device(struct tegra_devfreq *tegra,
 			      struct tegra_devfreq_device *dev)
 {
-	unsigned long flags;
 	u32 intr_status, dev_ctrl;
 
-	spin_lock_irqsave(&dev->lock, flags);
-
 	dev->avg_count = device_readl(dev, ACTMON_DEV_AVG_COUNT);
 	tegra_devfreq_update_avg_wmark(tegra, dev);
 
@@ -302,26 +298,6 @@ static void actmon_isr_device(struct tegra_devfreq *tegra,
 	device_writel(dev, ACTMON_INTR_STATUS_CLEAR, ACTMON_DEV_INTR_STATUS);
 
 	actmon_write_barrier(tegra);
-
-	spin_unlock_irqrestore(&dev->lock, flags);
-}
-
-static irqreturn_t actmon_isr(int irq, void *data)
-{
-	struct tegra_devfreq *tegra = data;
-	bool handled = false;
-	unsigned int i;
-	u32 val;
-
-	val = actmon_readl(tegra, ACTMON_GLB_STATUS);
-	for (i = 0; i < ARRAY_SIZE(tegra->devices); i++) {
-		if (val & tegra->devices[i].config->irq_mask) {
-			actmon_isr_device(tegra, tegra->devices + i);
-			handled = true;
-		}
-	}
-
-	return handled ? IRQ_WAKE_THREAD : IRQ_NONE;
 }
 
 static unsigned long actmon_cpu_to_emc_rate(struct tegra_devfreq *tegra,
@@ -348,35 +324,46 @@ static void actmon_update_target(struct tegra_devfreq *tegra,
 	unsigned long cpu_freq = 0;
 	unsigned long static_cpu_emc_freq = 0;
 	unsigned int avg_sustain_coef;
-	unsigned long flags;
+	u32 avg_count;
 

drivers/devfreq/tegra-devfreq.c static void actmon_update_target(struct tegra_devfreq *tegra, · line 335

Quoted 10 lines
drivers/devfreq/tegra-devfreq.c +2 -3
 	if (dev->config->avg_dependency_threshold) {
 		cpu_freq = cpufreq_get(0);
 		static_cpu_emc_freq = actmon_cpu_to_emc_rate(tegra, cpu_freq);
 	}
 
-	spin_lock_irqsave(&dev->lock, flags);
-
-	dev->target_freq = dev->avg_count / ACTMON_SAMPLING_PERIOD;
+	avg_count = dev->avg_count;
+	dev->target_freq = avg_count / ACTMON_SAMPLING_PERIOD;

Actually, this change is not related to this patch. Please keep the original code.

Quoted 8 lines

drivers/devfreq/tegra-devfreq.c static void actmon_update_target(struct tegra_devfreq *tegra, · line 340

Quoted 6 lines
drivers/devfreq/tegra-devfreq.c +1 -1
 	avg_sustain_coef = 100 * 100 / dev->config->boost_up_threshold;
 	dev->target_freq = do_percent(dev->target_freq, avg_sustain_coef);
 	dev->target_freq += dev->boost_freq;
 
-	if (dev->avg_count >= dev->config->avg_dependency_threshold)
+	if (avg_count >= dev->config->avg_dependency_threshold)

ditto.

Good catch, it's a leftover from v1 that I forgot to revert. Thank you.

[snip]

Quoted 4 lines

When I review this patch, I have a question about why tegra_actmon_rate_notify_cb is needed.

Quoted 7 lines
Quoted 5 lines

tegra_actmon_rate_notify_cb() do something
when the clock rate of emc_clock is changed.
I think that 'emc_clock' is changed by this driver. It means that the this driver can catch the change timing of emc_clock rate without notifier.

The devfreq driver isn't the only driver of the EMC clock rate. The devfreq driver changes EMC freq dynamically based of on average memory usage activity, but for some hardware units (like display controller for example) there is a requirement for a minimum memory bandwidth (isochronous transactions) and hence when display is waking up from suspend it immediately requires an amount of memory bandwidth that could be higher than ACTMON hardware unit suggests and besides the ACTMON's reaction is delayed by the sampling period (12ms).

Thanks for explaining it. If EMC clock is used on multiple point, I understand why the clock notifier is necessary.

Quoted 10 lines
Quoted 8 lines

IMO, it is possible to call tegra_devfreq_update_wmark() directly without clock notifier before calling 'clk_set_min_rate()/clk_set_rate()' in the tegra_devfreq_target().

With clock notifier, it cannot restore something for tegra_devfreq_update_wmark(tegra, dev) when failed to set the rate of emc_clk by 'clk_set_min_rate()/clk_set_rate()'.

The watermarks should be changed in accordance to the actual EMC clock rate. Given that EMC rate could be changed by something else than the devfreq driver, we need to re-adjust the watermarks to the actual values after the EMC rate-change completion. Please note that the clock notifier uses the POST_RATE_CHANGE event that happens only after successful completion of EMC clock rate change.

I understand the needs of clock notifier. Thanks.

Quoted 2 lines
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer Reviewed-by
#32 lore raw

Re: [PATCH v2 12/19] PM / devfreq: tegra: Avoid inconsistency of current frequency value

In reply to Dmitry Osipenko

Hi,

Quoted 67 lines · Dmitry Osipenko

16.04.2019 10:15, Chanwoo Choi пишет:

Quoted 40 lines

Hi,

Quoted 37 lines · Dmitry Osipenko

The frequency value potentially could change in-between. It doesn't cause any real problem at all right now, but that could change in the future. Hence let's avoid the inconsistency.

---
 drivers/devfreq/tegra-devfreq.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/devfreq/tegra-devfreq.c +4 -2
index a668e4fbc874..f1a6f951813a 100644
@@ -496,13 +496,15 @@ static int tegra_devfreq_get_dev_status(struct device *dev,
 {
 	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
 	struct tegra_devfreq_device *actmon_dev;
+	unsigned long cur_freq;
 
-	stat->current_frequency = tegra->cur_freq * KHZ;
+	cur_freq = READ_ONCE(tegra->cur_freq);
 
 	/* To be used by the tegra governor */
 	stat->private_data = tegra;
 
 	/* The below are to be used by the other governors */
+	stat->current_frequency = cur_freq * KHZ;
 
 	actmon_dev = &tegra->devices[MCALL];
 
@@ -513,7 +515,7 @@ static int tegra_devfreq_get_dev_status(struct device *dev,
 	stat->busy_time *= 100 / BUS_SATURATION_RATIO;
 
 	/* Number of cycles in a sampling period */
-	stat->total_time = ACTMON_SAMPLING_PERIOD * tegra->cur_freq;
+	stat->total_time = ACTMON_SAMPLING_PERIOD * cur_freq;
 
Quoted 7 lines

drivers/devfreq/tegra-devfreq.c static int tegra_devfreq_get_dev_status(struct device *dev, · line 521

Quoted 3 lines
drivers/devfreq/tegra-devfreq.c +0 -0
 	stat->busy_time = min(stat->busy_time, stat->total_time);
 

The read/write access of tegra->cur_freq is in the single routine of update_devfreq() as following. I think that there are no any potential problem about the inconsistency of tegra->cur_freq.

No, that's wrong assumption. The tegra->cur_freq is changed by the clock notifier that runs asynchronously with the devfreq driver when EMC clock rate is changed by something else in the kernel.

Quoted 16 lines

IMHO, if there are no any problem now, I'm not sure that we need to apply this patch.

update_devfreq()

{
	devfreq->governor->get_target_freq()
		devfreq_update_stats(devfreq)
			tegra_devfreq_get_dev_status()
				stat->current_frequency = tegra->cur_freq * KHZ;

	devfreq_set_target()
		tegra_devfreq_target()
			clk_set_min_rate(emc_rate, )
				tegra_actmon_rate_notify_cb()
					tegra->cur_freq = data->new_rate / KHZ;
Quoted 10 lines
Quoted 6 lines
			clk_set_rate(emc_rate, )
				tegra_actmon_rate_notify_cb()
					tegra->cur_freq = data->new_rate / KHZ;
}

The cur_freq value is changed by the clock notifier that runs asynchronously with the rest of the devfreq driver. Hence potentially compiler may generate two separate fetches of the cur_freq value, then the clock rate could be changed by other CPU core simultaneously with tegra_devfreq_get_dev_status() or kernel may re-schedule preemptively, changing the clock rate in-between of the two fetches.

Thanks. I understand why have to consider the inconsistency of clock which is used on the multiple points.

Looks good to me.

Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer Reviewed-by
#15 lore raw

Re: [PATCH v2 05/19] PM / devfreq: tegra: Replace write memory barrier with the read barrier

In reply to Dmitry Osipenko
Quoted 30 lines · Dmitry Osipenko

16.04.2019 11:00, Chanwoo Choi пишет:

Quoted 13 lines

Hi,

Quoted 10 lines · Dmitry Osipenko

The write memory barrier isn't needed because the BUS buffer is flushed by read after write that happens after the removed wmb(), we will also use readl() instead of the relaxed version to ensure that read is indeed completed.

---
 drivers/devfreq/tegra-devfreq.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
Quoted 15 lines

drivers/devfreq/tegra-devfreq.c static void tegra_devfreq_update_wmark(struct tegra_devfreq *tegra, · line 246

Quoted 11 lines
drivers/devfreq/tegra-devfreq.c +1 -2
index d62fb1b0d9bb..f0f0d78f6cbf 100644
@@ -243,8 +243,7 @@ static void tegra_devfreq_update_wmark(struct tegra_devfreq *tegra,
 static void actmon_write_barrier(struct tegra_devfreq *tegra)
 {
 	/* ensure the update has reached the ACTMON */
-	wmb();
-	actmon_readl(tegra, ACTMON_GLB_STATUS);
+	readl(tegra->regs + ACTMON_GLB_STATUS);

I think that this meaning of actmon_write_barrier() keeps the order of 'store' assembly command without the execution change from compiler optimization by using the wmb().

Quoted 1 line

The IO mapped memory is strongly-ordered on ARM, hence all readl/writel accesses are guaranteed to be ordered by default. I think wmb() here is just a cut-n-pasted relic from old downstream driver.

OK.

Quoted 8 lines
Quoted 8 lines

But, this patch edits it as following:
The result of the following two cases are same?

[original code]

	wmb()
	read_relaxed()
Quoted 7 lines
Quoted 3 lines

[new code by this patch]

	readl_relaxed()
	rmb()

Yes, the result is the same. The wmb() is not just about IO accesses, but about all kind of memory accesses and at least on Tegra30 it is quite expensive operation because it translates into L2XO cache syncing (arm_heavy_mb) + dsb(). It should be more efficient to flush out writes with a read-back and then wait for that read operation to be completed.

OK. Thanks for explanation.

Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer
#19 lore raw

Re: [PATCH v2 06/19] PM / devfreq: tegra: Fix missed error checking on devfreq initialization failure

In reply to Dmitry Osipenko

Hi,

Quoted 17 lines · Dmitry Osipenko

16.04.2019 5:32, Chanwoo Choi пишет:

Quoted 5 lines

Hi,

patch6/7/8/9 are for handling of exception handling in probe() function. Actually, I'm not sure that there are special reason to split out the patches. I think that you can squash patch6/7/8/9 to only one patch.

Indeed, I was rebasing and reordering patches multiple times and looks like there is no reason not to squash these patches now.

Quoted 7 lines

Also, even if patch6/7/8/9 handle the exception handling in probe(), the tegra_devfreq_probe() doesn't support the restoring sequence when fail happen. I think that if you want to fix the fail case of probe(), please add the restoring sequence about following function.
- clk_disable_unprepare()
- clk_notifier_unregister()
- dev_pm_opp_remove()

Quoted 1 line

When all of 6/7/8/9 patches are applied, the clk_notifier_register() becomes the last invocation of the probe function and clk_enable() is kept at the first place of the probe order. Hence the sequence you're suggesting is already incorrect because error-unwinding order usually should be opposite to the probe order. It looks to me that the current final result of these patches is already correct.

You're right. When I replied it, I have not considered the order. Sorry, it made you some confusion.

Quoted 2 lines
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
maintainer Reviewed-by on v2 2/19
#5 lore raw

Re: [PATCH v2 02/19] PM / devfreq: tegra: Replace readl-writel with relaxed versions

In reply to Dmitry Osipenko

Hi,

Quoted 31 lines · Dmitry Osipenko

There is no need to insert memory barrier on each readl/writel invocation, hence use the relaxed versions.

---
 drivers/devfreq/tegra-devfreq.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/devfreq/tegra-devfreq.c +3 -3
index 02905978abe1..0c0909fba545 100644
@@ -191,23 +191,23 @@ static struct tegra_actmon_emc_ratio actmon_emc_ratios[] = {
 
 static u32 actmon_readl(struct tegra_devfreq *tegra, u32 offset)
 {
-	return readl(tegra->regs + offset);
+	return readl_relaxed(tegra->regs + offset);
 }
 
 static void actmon_writel(struct tegra_devfreq *tegra, u32 val, u32 offset)
 {
-	writel(val, tegra->regs + offset);
+	writel_relaxed(val, tegra->regs + offset);
 }
 
 static u32 device_readl(struct tegra_devfreq_device *dev, u32 offset)
 {
-	return readl(dev->regs + offset);
+	return readl_relaxed(dev->regs + offset);

drivers/devfreq/tegra-devfreq.c static struct tegra_actmon_emc_ratio actmon_emc_ratios[] = { · line 213

Quoted 10 lines
drivers/devfreq/tegra-devfreq.c +1 -1
 }
 
 static void device_writel(struct tegra_devfreq_device *dev, u32 val,
 			  u32 offset)
 {
-	writel(val, dev->regs + offset);
+	writel_relaxed(val, dev->regs + offset);
 }
 
 static unsigned long do_percent(unsigned long val, unsigned int pct)
Signature · 3 lines
Best Regards,
Chanwoo Choi
Samsung Electronics
#53 lore raw

Re: [PATCH v2 19/19] PM / devfreq: Introduce driver for NVIDIA Tegra20

In reply to Chanwoo Choi

17.04.2019 3:26, Chanwoo Choi пишет:

Quoted 246 lines

Hi,

Quoted 102 lines · Dmitry Osipenko

16.04.2019 11:31, Chanwoo Choi пишет:

Quoted 66 lines

Hi,

Quoted 63 lines · Dmitry Osipenko

Add devfreq driver for NVIDIA Tegra20 SoC's. The driver periodically reads out Memory Controller counters and adjusts memory frequency based on the memory clients activity.

---
 MAINTAINERS                       |   8 ++
 drivers/devfreq/Kconfig           |  10 ++
 drivers/devfreq/Makefile          |   1 +
 drivers/devfreq/tegra20-devfreq.c | 177 ++++++++++++++++++++++++++++++
 4 files changed, 196 insertions(+)
 create mode 100644 drivers/devfreq/tegra20-devfreq.c
MAINTAINERS +8 -0
index 80b59db1b6e4..91f475ec4545 100644
@@ -10056,6 +10056,14 @@ F:	include/linux/memblock.h
 F:	mm/memblock.c
 F:	Documentation/core-api/boot-time-mm.rst
 
+MEMORY FREQUENCY SCALING DRIVER FOR NVIDIA TEGRA20
+M:	Dmitry Osipenko <[email protected]>
+L:	[email protected]
+L:	[email protected]
+T:	git git://git.kernel.org/pub/scm/linux/kernel/git/mzx/devfreq.git
+S:	Maintained
+F:	drivers/devfreq/tegra20-devfreq.c
+
 MEMORY MANAGEMENT
 L:	[email protected]
 W:	http://www.linux-mm.org
drivers/devfreq/Kconfig +10 -0
index bd6652863e7d..af4c86c4e0f6 100644
@@ -100,6 +100,16 @@ config ARM_TEGRA_DEVFREQ
 	  It reads ACTMON counters of memory controllers and adjusts the
 	  operating frequencies and voltages with OPP support.
 
+config ARM_TEGRA20_DEVFREQ
+	tristate "NVIDIA Tegra20 DEVFREQ Driver"
+	depends on (TEGRA_MC && TEGRA20_EMC) || COMPILE_TEST
+	select DEVFREQ_GOV_SIMPLE_ONDEMAND
+	select PM_OPP
+	help
+	  This adds the DEVFREQ driver for the Tegra20 family of SoCs.
+	  It reads Memory Controller counters and adjusts the operating
+	  frequencies and voltages with OPP support.
+
 config ARM_RK3399_DMC_DEVFREQ
 	tristate "ARM RK3399 DMC DEVFREQ Driver"
 	depends on ARCH_ROCKCHIP
drivers/devfreq/Makefile +1 -0
index 32b8d4d3f12c..6fcc5596b8b7 100644
@@ -11,6 +11,7 @@ obj-$(CONFIG_DEVFREQ_GOV_PASSIVE)	+= governor_passive.o
 obj-$(CONFIG_ARM_EXYNOS_BUS_DEVFREQ)	+= exynos-bus.o
 obj-$(CONFIG_ARM_RK3399_DMC_DEVFREQ)	+= rk3399_dmc.o
 obj-$(CONFIG_ARM_TEGRA_DEVFREQ)		+= tegra-devfreq.o
+obj-$(CONFIG_ARM_TEGRA20_DEVFREQ)	+= tegra20-devfreq.o
 
Quoted 16 lines
Quoted 8 lines
drivers/devfreq/Makefile +0 -0
 # DEVFREQ Event Drivers
 obj-$(CONFIG_PM_DEVFREQ_EVENT)		+= event/
drivers/devfreq/tegra20-devfreq.c new file +0 -0
new file mode 100644
index 000000000000..18c9aad7a9d7

drivers/devfreq/tegra20-devfreq.c line 6

Quoted 6 lines
drivers/devfreq/tegra20-devfreq.c new file +6 -0
@@ -0,0 +1,177 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * NVIDIA Tegra20 devfreq driver
+ *
+ * Author: Dmitry Osipenko <[email protected]>
+ */

It doesn't any "Copyright (c) 2019 ..." sentence.

I'll add one in v3.

Quoted 16 lines

drivers/devfreq/tegra20-devfreq.c line 18

Quoted 12 lines
drivers/devfreq/tegra20-devfreq.c new file +12 -0
+
+#include <linux/clk.h>
+#include <linux/devfreq.h>
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of_device.h>
+#include <linux/platform_device.h>
+#include <linux/pm_opp.h>
+#include <linux/slab.h>
+
+#include <soc/tegra/mc.h>

I can find the '<soc/tegra/mc.h>' header file
on mainline branch. But mc.h is included in linux-next.git.

Quoted 8 lines
Quoted 4 lines

If you don't share the patch related to mc.h,
the kernel build will be failed when apply it the devfreq.git on final step. Actually, it should make the immutable branch between two related maintainers in order to remove the build fail.

The '<soc/tegra/mc.h>' header file exists since v3.18 [0], seems you just missed something.

[0] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/include/soc/tegra/mc.h?h=v3.18&id=8918465163171322c77a19d5258a95f56d89d2e4

Sorry. It is my missing point. When I tried to find it, it is included in the mainline kernel.

Quoted 129 lines
Quoted 35 lines
Quoted 35 lines
drivers/devfreq/tegra20-devfreq.c new file +34 -0
+
+#include "governor.h"
+
+#define MC_STAT_CONTROL				0x90
+#define MC_STAT_EMC_CLOCK_LIMIT			0xa0
+#define MC_STAT_EMC_CLOCKS			0xa4
+#define MC_STAT_EMC_CONTROL			0xa8
+#define MC_STAT_EMC_COUNT			0xb8
+
+#define EMC_GATHER_CLEAR			(1 << 8)
+#define EMC_GATHER_ENABLE			(3 << 8)
+
+struct tegra_devfreq {
+	struct devfreq *devfreq;
+	struct clk *clk;
+	void __iomem *regs;
+};
+
+static int tegra_devfreq_target(struct device *dev, unsigned long *freq,
+				u32 flags)
+{
+	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
+	struct dev_pm_opp *opp;
+	unsigned long rate;
+	int err;
+
+	opp = devfreq_recommended_opp(dev, freq, flags);
+	if (IS_ERR(opp))
+		return PTR_ERR(opp);
+
+	rate = dev_pm_opp_get_freq(opp);
+	dev_pm_opp_put(opp);
+
+	err = clk_set_min_rate(tegra->clk, rate);
Quoted 8 lines

drivers/devfreq/tegra20-devfreq.c line 57

Quoted 5 lines
drivers/devfreq/tegra20-devfreq.c new file +5 -0
+	if (err)
+		return err;
+
+	err = clk_set_rate(tegra->clk, 0);
+	if (err)

When fail happen, I think that you have to control
the restoring sequence for previous operation like clk_set_min_rate().

Okay, thanks.

Quoted 55 lines
Quoted 55 lines
drivers/devfreq/tegra20-devfreq.c new file +55 -0
+		return err;
+
+	return 0;
+}
+
+static int tegra_devfreq_get_dev_status(struct device *dev,
+					struct devfreq_dev_status *stat)
+{
+	struct tegra_devfreq *tegra = dev_get_drvdata(dev);
+
+	stat->busy_time = readl_relaxed(tegra->regs + MC_STAT_EMC_COUNT);
+	stat->total_time = readl_relaxed(tegra->regs + MC_STAT_EMC_CLOCKS) / 8;
+	stat->current_frequency = clk_get_rate(tegra->clk);
+
+	writel_relaxed(EMC_GATHER_CLEAR, tegra->regs + MC_STAT_CONTROL);
+	writel_relaxed(EMC_GATHER_ENABLE, tegra->regs + MC_STAT_CONTROL);
+
+	return 0;
+}
+
+static struct devfreq_dev_profile tegra_devfreq_profile = {
+	.polling_ms	= 500,
+	.target		= tegra_devfreq_target,
+	.get_dev_status	= tegra_devfreq_get_dev_status,
+};
+
+static struct tegra_mc *tegra_get_memory_controller(void)
+{
+	struct platform_device *pdev;
+	struct device_node *np;
+	struct tegra_mc *mc;
+
+	np = of_find_compatible_node(NULL, NULL, "nvidia,tegra20-mc-gart");
+	if (!np)
+		return ERR_PTR(-ENOENT);
+
+	pdev = of_find_device_by_node(np);
+	of_node_put(np);
+	if (!pdev)
+		return ERR_PTR(-ENODEV);
+
+	mc = platform_get_drvdata(pdev);
+	if (!mc)
+		return ERR_PTR(-EPROBE_DEFER);
+
+	return mc;
+}
+
+static int tegra_devfeq_probe(struct platform_device *pdev)
+{
+	struct tegra_devfreq *tegra;
+	struct tegra_mc *mc;
+	unsigned long max_rate;
+	unsigned long rate;
+	int err;
Quoted 8 lines

drivers/devfreq/tegra20-devfreq.c line 117

Quoted 5 lines
drivers/devfreq/tegra20-devfreq.c new file +5 -0
+
+	mc = tegra_get_memory_controller();
+	if (IS_ERR(mc)) {
+		err = PTR_ERR(mc);
+		dev_err(&pdev->dev, "failed to get mc: %d\n", err);

How about using 'memory controller' instead of 'mc'? Because 'mc' is not standard expression.

Sounds good, thanks.

Quoted 16 lines

drivers/devfreq/tegra20-devfreq.c line 130

Quoted 13 lines
drivers/devfreq/tegra20-devfreq.c new file +13 -0
+		return err;
+	}
+
+	tegra = devm_kzalloc(&pdev->dev, sizeof(*tegra), GFP_KERNEL);
+	if (!tegra)
+		return -ENOMEM;
+
+	tegra->clk = devm_clk_get(&pdev->dev, "emc");
+	if (IS_ERR(tegra->clk)) {
+		err = PTR_ERR(tegra->clk);
+		dev_err(&pdev->dev, "failed to get emc clock: %d\n", err);
+		return err;
+	}

Don't you need to enable the 'emc' clock'?
Because this patch doesn't enable this clock.

Quoted 6 lines
Quoted 1 line

EMC is a system critical clock (marked as critical by the clock driver), it is guaranteed to be always enabled. I think it is fine to omit clock-enabling in this case.

If you think don't need to enable it due to the critical clock, instead, please add the comment about that emc clock is critical. In my case, it looks like the strange use-case because this driver doesn't contain the any enable code for clock.

Okay, thank you for the review again.

[snip]