Skip to contentSkip to conversation

mmc: mmci: Support ap_sleep in cpuidle for ux500

PATCH New Ulf Hansson 3 participants 10 messages Sep 3, 2013 – Sep 4, 2013

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

v3 cover mmc: mmci: Support ap_sleep in cpuidle for ux500

To be able to enter ap_sleep state in cpuidle were the APE power domain regulator will be cut, a register save and restore mechanism needs to be implemented. This patchset adapts to these restrictions.

Changes in v3:
- Fold in a new patch, mmc: mmci: Adapt to new pinctrl handling
- Updated patch "mmc: mmci: Use optional sleep pinctrl state" to use the new pinctrl API.
- Added Acks from Rickard Andersson.

Ulf Hansson (4):
mmc: mmci: Adapt to new pinctrl handling
mmc: mmci: Use optional sleep pinctrl state
mmc: mmci: Adapt to register write restrictions
mmc: mmci: Save and restore register context

 drivers/mmc/host/mmci.c |   78 ++++++++++++++++++++++++++++++++++++-----------
 drivers/mmc/host/mmci.h |    4 ---
 2 files changed, 61 insertions(+), 21 deletions(-)
Signature · 1 line
1.7.9.5
#1 lore raw

PATCH v3 1/4 mmc: mmci: Adapt to new pinctrl handling

There is no need for every driver to fetch a pinctrl handle and to select the default state. Instead this is handled by the device driver core, thus we can remove this piece of code from mmci.

---
 drivers/mmc/host/mmci.c |   17 -----------------
 drivers/mmc/host/mmci.h |    4 ----
 2 files changed, 21 deletions(-)
drivers/mmc/host/mmci.c +0 -17
index c3785ed..5dffbbd 100644
@@ -1510,23 +1510,6 @@ static int mmci_probe(struct amba_device *dev,
 		mmc->f_max = min(host->mclk, fmax);
 	dev_dbg(mmc_dev(mmc), "clocking block at %u Hz\n", mmc->f_max);
 
-	host->pinctrl = devm_pinctrl_get(&dev->dev);
-	if (IS_ERR(host->pinctrl)) {
-		ret = PTR_ERR(host->pinctrl);
-		goto clk_disable;
-	}
-
-	host->pins_default = pinctrl_lookup_state(host->pinctrl,
-			PINCTRL_STATE_DEFAULT);
-
-	/* enable pins to be muxed in and configured */
-	if (!IS_ERR(host->pins_default)) {
-		ret = pinctrl_select_state(host->pinctrl, host->pins_default);
-		if (ret)
-			dev_warn(&dev->dev, "could not set default pins\n");
-	} else
-		dev_warn(&dev->dev, "could not get default pinstate\n");
-
 	/* Get regulators and the supported OCR mask */
 	mmc_regulator_get_supply(mmc);
 	if (!mmc->ocr_avail)
drivers/mmc/host/mmci.h +0 -4
index 69080fa..168bc72 100644
@@ -200,10 +200,6 @@ struct mmci_host {
 	struct sg_mapping_iter	sg_miter;
 	unsigned int		size;
 
-	/* pinctrl handles */
-	struct pinctrl		*pinctrl;
-	struct pinctrl_state	*pins_default;
-
 #ifdef CONFIG_DMA_ENGINE
 	/* DMA stuff */
 	struct dma_chan		*dma_current;
Signature · 1 line
1.7.9.5
#3 lore raw

PATCH v3 2/4 mmc: mmci: Use optional sleep pinctrl state

In reply to Ulf Hansson

By optionally putting the pins into sleep state in the .runtime_suspend callback we can accomplish two things. One is to minimize current leakage from pins and thus save power, second we can prevent the IP from driving pins output in an uncontrolled manner, which may happen if the power domain drops the domain regulator.

When returning from idle, entering .runtime_resume callback, the pins are restored to default state.

---
 drivers/mmc/host/mmci.c |    2 ++
 1 file changed, 2 insertions(+)
drivers/mmc/host/mmci.c +2 -0
index 5dffbbd..c550b3e 100644
@@ -1750,6 +1750,7 @@ static int mmci_runtime_suspend(struct device *dev)
 
 	if (mmc) {
 		struct mmci_host *host = mmc_priv(mmc);
+		pinctrl_pm_select_sleep_state(dev);
 		clk_disable_unprepare(host->clk);
 	}
 
@@ -1764,6 +1765,7 @@ static int mmci_runtime_resume(struct device *dev)
 	if (mmc) {
 		struct mmci_host *host = mmc_priv(mmc);
 		clk_prepare_enable(host->clk);
+		pinctrl_pm_select_default_state(dev);
 	}
 
 	return 0;
Signature · 1 line
1.7.9.5
#4 lore raw

PATCH v3 2/4 mmc: mmci: Use optional sleep pinctrl state

Quoted 6 lines · Ulf Hansson

By optionally putting the pins into sleep state in the .runtime_suspend callback we can accomplish two things. One is to minimize current leakage from pins and thus save power, second we can prevent the IP from driving pins output in an uncontrolled manner, which may happen if the power domain drops the domain regulator.

Quoted 6 lines

When returning from idle, entering .runtime_resume callback, the pins are restored to default state.

Yours,
Linus Walleij

#5 lore raw

PATCH v3 3/4 mmc: mmci: Adapt to register write restrictions

In reply to Ulf Hansson

After a write to the MMCICLOCK register data cannot be written to this register for three feedback clock cycles. Writes to the MMCIPOWER register must be separated by three MCLK cycles. Previously no issues has been observered, but using higher ARM clock frequencies on STE- platforms has triggered this problem.

The MMCICLOCK register is written to in .set_ios and for some data transmissions for SDIO. We do not need a delay at the data transmission path, because sending and receiving data will require more than three clock cycles. Then we use a simple logic to only delay in .set_ios and thus we don't affect throughput performance.

---
 drivers/mmc/host/mmci.c |   16 ++++++++++++++++
 1 file changed, 16 insertions(+)
drivers/mmc/host/mmci.c +16 -0
index c550b3e..82afcd3 100644
@@ -189,6 +189,21 @@ static int mmci_validate_data(struct mmci_host *host,
 	return 0;
 }
 
+static void mmci_reg_delay(struct mmci_host *host)
+{
+	/*
+	 * According to the spec, at least three feedback clock cycles
+	 * of max 52 MHz must pass between two writes to the MMCICLOCK reg.
+	 * Three MCLK clock cycles must pass between two MMCIPOWER reg writes.
+	 * Worst delay time during card init is at 100 kHz => 30 us.
+	 * Worst delay time when up and running is at 25 MHz => 120 ns.
+	 */
+	if (host->cclk < 20000000)
+		udelay(30);
+	else
+		ndelay(120);
+}
+
 /*
  * This must be called with host->lock held
  */
@@ -1264,6 +1279,7 @@ static void mmci_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
 
 	mmci_set_clkreg(host, ios->clock);
 	mmci_write_pwrreg(host, pwr);
+	mmci_reg_delay(host);
 
 	spin_unlock_irqrestore(&host->lock, flags);
 
Signature · 1 line
1.7.9.5
#6 lore raw

PATCH v3 3/4 mmc: mmci: Adapt to register write restrictions

Quoted 12 lines · Ulf Hansson

After a write to the MMCICLOCK register data cannot be written to this register for three feedback clock cycles. Writes to the MMCIPOWER register must be separated by three MCLK cycles. Previously no issues has been observered, but using higher ARM clock frequencies on STE- platforms has triggered this problem.

The MMCICLOCK register is written to in .set_ios and for some data transmissions for SDIO. We do not need a delay at the data transmission path, because sending and receiving data will require more than three clock cycles. Then we use a simple logic to only delay in .set_ios and thus we don't affect throughput performance.

Quoted 4 lines

---

+ two questions below.

Quoted 11 lines
 drivers/mmc/host/mmci.c |   16 ++++++++++++++++
 1 file changed, 16 insertions(+)
drivers/mmc/host/mmci.c +0 -0
index c550b3e..82afcd3 100644
@@ -189,6 +189,21 @@ static int mmci_validate_data(struct mmci_host *host,
 	return 0;
 }
 

drivers/mmc/host/mmci.c static int mmci_validate_data(struct mmci_host *host, · line 201

Quoted 10 lines
drivers/mmc/host/mmci.c +10 -0
+static void mmci_reg_delay(struct mmci_host *host)
+{
+	/*
+	 * According to the spec, at least three feedback clock cycles
+	 * of max 52 MHz must pass between two writes to the MMCICLOCK reg.
+	 * Three MCLK clock cycles must pass between two MMCIPOWER reg writes.
+	 * Worst delay time during card init is at 100 kHz => 30 us.
+	 * Worst delay time when up and running is at 25 MHz => 120 ns.
+	 */
+	if (host->cclk < 20000000)

Shouldn't it be 25000000 ?

What about using macros ?

#define MMC_CLOCK 25000000
#define MMCINIT_DELAY 30
#define MMCUP_DELAY 120

Quoted 17 lines
drivers/mmc/host/mmci.c +6 -0
+		udelay(30);
+	else
+		ndelay(120);
+}
+
 /*
  * This must be called with host->lock held
  */
@@ -1264,6 +1279,7 @@ static void mmci_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
 
 	mmci_set_clkreg(host, ios->clock);
 	mmci_write_pwrreg(host, pwr);
+	mmci_reg_delay(host);
 
 	spin_unlock_irqrestore(&host->lock, flags);
 
Signature · 5 lines
 <http://www.linaro.org/> Linaro.org ? Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog
#7 lore raw

PATCH v3 3/4 mmc: mmci: Adapt to register write restrictions

Quoted 33 lines · Daniel Lezcano
Quoted 12 lines · Ulf Hansson

After a write to the MMCICLOCK register data cannot be written to this register for three feedback clock cycles. Writes to the MMCIPOWER register must be separated by three MCLK cycles. Previously no issues has been observered, but using higher ARM clock frequencies on STE- platforms has triggered this problem.

The MMCICLOCK register is written to in .set_ios and for some data transmissions for SDIO. We do not need a delay at the data transmission path, because sending and receiving data will require more than three clock cycles. Then we use a simple logic to only delay in .set_ios and thus we don't affect throughput performance.

Quoted 4 lines

---

+ two questions below.

Quoted 11 lines
 drivers/mmc/host/mmci.c |   16 ++++++++++++++++
 1 file changed, 16 insertions(+)
drivers/mmc/host/mmci.c +0 -0
index c550b3e..82afcd3 100644
@@ -189,6 +189,21 @@ static int mmci_validate_data(struct mmci_host *host,
      return 0;
 }
 
Quoted 12 lines

drivers/mmc/host/mmci.c static int mmci_validate_data(struct mmci_host *host, · line 201

Quoted 10 lines
drivers/mmc/host/mmci.c +10 -0
+static void mmci_reg_delay(struct mmci_host *host)
+{
+     /*
+      * According to the spec, at least three feedback clock cycles
+      * of max 52 MHz must pass between two writes to the MMCICLOCK reg.
+      * Three MCLK clock cycles must pass between two MMCIPOWER reg writes.
+      * Worst delay time during card init is at 100 kHz => 30 us.
+      * Worst delay time when up and running is at 25 MHz => 120 ns.
+      */
+     if (host->cclk < 20000000)

Shouldn't it be 25000000 ?

You are right! Thanks for spotting this! I resend a new version.

Quoted 6 lines

What about using macros ?

#define MMC_CLOCK 25000000
#define MMCINIT_DELAY 30
#define MMCUP_DELAY 120

These values wont be used anywhere else but here. Additionally, mmci is already having other hard-coded values around. Unless you feel it is important too you, I would prefer keeping this as is.

Thanks for review Daniel!

Kind regards
Ulf Hansson

Quoted 27 lines
Quoted 11 lines
drivers/mmc/host/mmci.c +5 -0
+             udelay(30);
+     else
+             ndelay(120);
+}
+
 /*
  * This must be called with host->lock held
  */

drivers/mmc/host/mmci.c static void mmci_set_ios(struct mmc_host *mmc, struct mmc_ios *ios) · line 1285

Quoted 7 lines
drivers/mmc/host/mmci.c +1 -0
@@ -1264,6 +1279,7 @@ static void mmci_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
 
      mmci_set_clkreg(host, ios->clock);
      mmci_write_pwrreg(host, pwr);
+     mmci_reg_delay(host);
 
      spin_unlock_irqrestore(&host->lock, flags);
 

--
<http://www.linaro.org/> Linaro.org ? Open source software for ARM SoCs

Follow Linaro: <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog

#8 lore raw

PATCH v3 4/4 mmc: mmci: Save and restore register context

In reply to Ulf Hansson

If a corresponding power domain exists for the device and it manages to cut the domain regulator while the device is runtime suspended, the IP loses it's registers context. We restore the context in the .runtime_resume callback from the existing register caches to adapt to this siutuation.

We also want to make sure the registers are in a known state while restoring context in the case when the power domain did not drop the power, since there are restrictions for the order of writing to these registers. To handle this, we clear the registers in the .runtime_suspend callback.

---
 drivers/mmc/host/mmci.c |   43 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 43 insertions(+)
drivers/mmc/host/mmci.c +43 -0
index 82afcd3..97e541b 100644
@@ -62,6 +62,7 @@ static unsigned int fmax = 515633;
  * @signal_direction: input/out direction of bus signals can be indicated
  * @pwrreg_clkgate: MMCIPOWER register must be used to gate the clock
  * @busy_detect: true if busy detection on dat0 is supported
+ * @pwrreg_nopower: bits in MMCIPOWER don't controls ext. power supply
  */
 struct variant_data {
 	unsigned int		clkreg;
@@ -76,6 +77,7 @@ struct variant_data {
 	bool			signal_direction;
 	bool			pwrreg_clkgate;
 	bool			busy_detect;
+	bool			pwrreg_nopower;
 };
 
 static struct variant_data variant_arm = {
@@ -109,6 +111,7 @@ static struct variant_data variant_u300 = {
 	.pwrreg_powerup		= MCI_PWR_ON,
 	.signal_direction	= true,
 	.pwrreg_clkgate		= true,
+	.pwrreg_nopower		= true,
 };
 
 static struct variant_data variant_nomadik = {
@@ -121,6 +124,7 @@ static struct variant_data variant_nomadik = {
 	.pwrreg_powerup		= MCI_PWR_ON,
 	.signal_direction	= true,
 	.pwrreg_clkgate		= true,
+	.pwrreg_nopower		= true,
 };
 
 static struct variant_data variant_ux500 = {
@@ -135,6 +139,7 @@ static struct variant_data variant_ux500 = {
 	.signal_direction	= true,
 	.pwrreg_clkgate		= true,
 	.busy_detect		= true,
+	.pwrreg_nopower		= true,
 };
 
 static struct variant_data variant_ux500v2 = {
@@ -150,6 +155,7 @@ static struct variant_data variant_ux500v2 = {
 	.signal_direction	= true,
 	.pwrreg_clkgate		= true,
 	.busy_detect		= true,
+	.pwrreg_nopower		= true,
 };
 
 static int mmci_card_busy(struct mmc_host *mmc)
@@ -1759,6 +1765,41 @@ static int mmci_resume(struct device *dev)
 #endif
 
 #ifdef CONFIG_PM_RUNTIME
+static void mmci_save(struct mmci_host *host)
+{
+	unsigned long flags;
+
+	if (host->variant->pwrreg_nopower) {
+		spin_lock_irqsave(&host->lock, flags);
+
+		writel(0, host->base + MMCIMASK0);
+		writel(0, host->base + MMCIDATACTRL);
+		writel(0, host->base + MMCIPOWER);
+		writel(0, host->base + MMCICLOCK);
+		mmci_reg_delay(host);
+
+		spin_unlock_irqrestore(&host->lock, flags);
+	}
+
+}
+
+static void mmci_restore(struct mmci_host *host)
+{
+	unsigned long flags;
+
+	if (host->variant->pwrreg_nopower) {
+		spin_lock_irqsave(&host->lock, flags);
+
+		writel(host->clk_reg, host->base + MMCICLOCK);
+		writel(host->datactrl_reg, host->base + MMCIDATACTRL);
+		writel(host->pwr_reg, host->base + MMCIPOWER);
+		writel(MCI_IRQENABLE, host->base + MMCIMASK0);
+		mmci_reg_delay(host);
+
+		spin_unlock_irqrestore(&host->lock, flags);
+	}
+}
+
 static int mmci_runtime_suspend(struct device *dev)
 {
 	struct amba_device *adev = to_amba_device(dev);
@@ -1767,6 +1808,7 @@ static int mmci_runtime_suspend(struct device *dev)
 	if (mmc) {
 		struct mmci_host *host = mmc_priv(mmc);
 		pinctrl_pm_select_sleep_state(dev);
+		mmci_save(host);
 		clk_disable_unprepare(host->clk);
 	}
 
@@ -1781,6 +1823,7 @@ static int mmci_runtime_resume(struct device *dev)
 	if (mmc) {
 		struct mmci_host *host = mmc_priv(mmc);
 		clk_prepare_enable(host->clk);
+		mmci_restore(host);
 		pinctrl_pm_select_default_state(dev);
 	}
 
Signature · 1 line
1.7.9.5
#9 lore raw

PATCH v3 4/4 mmc: mmci: Save and restore register context

Quoted 12 lines · Ulf Hansson

If a corresponding power domain exists for the device and it manages to cut the domain regulator while the device is runtime suspended, the IP loses it's registers context. We restore the context in the .runtime_resume callback from the existing register caches to adapt to this siutuation.

We also want to make sure the registers are in a known state while restoring context in the case when the power domain did not drop the power, since there are restrictions for the order of writing to these registers. To handle this, we clear the registers in the .runtime_suspend callback.

Quoted 4 lines

---

Quoted 114 lines
 drivers/mmc/host/mmci.c |   43 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 43 insertions(+)
drivers/mmc/host/mmci.c +43 -0
index 82afcd3..97e541b 100644
@@ -62,6 +62,7 @@ static unsigned int fmax = 515633;
  * @signal_direction: input/out direction of bus signals can be indicated
  * @pwrreg_clkgate: MMCIPOWER register must be used to gate the clock
  * @busy_detect: true if busy detection on dat0 is supported
+ * @pwrreg_nopower: bits in MMCIPOWER don't controls ext. power supply
  */
 struct variant_data {
 	unsigned int		clkreg;
@@ -76,6 +77,7 @@ struct variant_data {
 	bool			signal_direction;
 	bool			pwrreg_clkgate;
 	bool			busy_detect;
+	bool			pwrreg_nopower;
 };
 
 static struct variant_data variant_arm = {
@@ -109,6 +111,7 @@ static struct variant_data variant_u300 = {
 	.pwrreg_powerup		= MCI_PWR_ON,
 	.signal_direction	= true,
 	.pwrreg_clkgate		= true,
+	.pwrreg_nopower		= true,
 };
 
 static struct variant_data variant_nomadik = {
@@ -121,6 +124,7 @@ static struct variant_data variant_nomadik = {
 	.pwrreg_powerup		= MCI_PWR_ON,
 	.signal_direction	= true,
 	.pwrreg_clkgate		= true,
+	.pwrreg_nopower		= true,
 };
 
 static struct variant_data variant_ux500 = {
@@ -135,6 +139,7 @@ static struct variant_data variant_ux500 = {
 	.signal_direction	= true,
 	.pwrreg_clkgate		= true,
 	.busy_detect		= true,
+	.pwrreg_nopower		= true,
 };
 
 static struct variant_data variant_ux500v2 = {
@@ -150,6 +155,7 @@ static struct variant_data variant_ux500v2 = {
 	.signal_direction	= true,
 	.pwrreg_clkgate		= true,
 	.busy_detect		= true,
+	.pwrreg_nopower		= true,
 };
 
 static int mmci_card_busy(struct mmc_host *mmc)
@@ -1759,6 +1765,41 @@ static int mmci_resume(struct device *dev)
 #endif
 
 #ifdef CONFIG_PM_RUNTIME
+static void mmci_save(struct mmci_host *host)
+{
+	unsigned long flags;
+
+	if (host->variant->pwrreg_nopower) {
+		spin_lock_irqsave(&host->lock, flags);
+
+		writel(0, host->base + MMCIMASK0);
+		writel(0, host->base + MMCIDATACTRL);
+		writel(0, host->base + MMCIPOWER);
+		writel(0, host->base + MMCICLOCK);
+		mmci_reg_delay(host);
+
+		spin_unlock_irqrestore(&host->lock, flags);
+	}
+
+}
+
+static void mmci_restore(struct mmci_host *host)
+{
+	unsigned long flags;
+
+	if (host->variant->pwrreg_nopower) {
+		spin_lock_irqsave(&host->lock, flags);
+
+		writel(host->clk_reg, host->base + MMCICLOCK);
+		writel(host->datactrl_reg, host->base + MMCIDATACTRL);
+		writel(host->pwr_reg, host->base + MMCIPOWER);
+		writel(MCI_IRQENABLE, host->base + MMCIMASK0);
+		mmci_reg_delay(host);
+
+		spin_unlock_irqrestore(&host->lock, flags);
+	}
+}
+
 static int mmci_runtime_suspend(struct device *dev)
 {
 	struct amba_device *adev = to_amba_device(dev);
@@ -1767,6 +1808,7 @@ static int mmci_runtime_suspend(struct device *dev)
 	if (mmc) {
 		struct mmci_host *host = mmc_priv(mmc);
 		pinctrl_pm_select_sleep_state(dev);
+		mmci_save(host);
 		clk_disable_unprepare(host->clk);
 	}
 
@@ -1781,6 +1823,7 @@ static int mmci_runtime_resume(struct device *dev)
 	if (mmc) {
 		struct mmci_host *host = mmc_priv(mmc);
 		clk_prepare_enable(host->clk);
+		mmci_restore(host);
 		pinctrl_pm_select_default_state(dev);
 	}
 
Signature · 5 lines
 <http://www.linaro.org/> Linaro.org ? Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog