Skip to contentSkip to conversation

OMAP: GPIO: Fix OMAP4 GPIO reg access issues

PATCH In review Charulatha V 3 participants 6 messages Apr 8, 2010 – May 12, 2010

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

PATCH v2 1/1 OMAP: GPIO: Fix OMAP4 GPIO reg access issues

Access to some of the OMAP4 GPIO registers are not properly handled. This patch fixes it.

This patch is tested on 3430SDP and 4430SDP boards

---
 arch/arm/plat-omap/gpio.c |   61 +++++++++++++++++++++++++++++++++++---------
 1 files changed, 55 insertions(+), 12 deletions(-)
arch/arm/plat-omap/gpio.c +47 -12
index 76a347b..3a9891c 100644
@@ -137,7 +137,11 @@
 #define OMAP4_GPIO_IRQSTATUSCLR1	0x0040
 #define OMAP4_GPIO_IRQWAKEN0		0x0044
 #define OMAP4_GPIO_IRQWAKEN1		0x0048
-#define OMAP4_GPIO_SYSSTATUS		0x0104
+#define OMAP4_GPIO_SYSSTATUS		0x0114
+#define OMAP4_GPIO_IRQENABLE1		0x011c
+#define OMAP4_GPIO_WAKE_EN		0x0120
+#define OMAP4_GPIO_IRQSTATUS2		0x0128
+#define OMAP4_GPIO_IRQENABLE2		0x012c
 #define OMAP4_GPIO_CTRL			0x0130
 #define OMAP4_GPIO_OE			0x0134
 #define OMAP4_GPIO_DATAIN		0x0138
@@ -148,6 +152,10 @@
 #define OMAP4_GPIO_FALLINGDETECT	0x014c
 #define OMAP4_GPIO_DEBOUNCENABLE	0x0150
 #define OMAP4_GPIO_DEBOUNCINGTIME	0x0154
+#define OMAP4_GPIO_CLEARIRQENABLE1	0x0160
+#define OMAP4_GPIO_SETIRQENABLE1	0x0164
+#define OMAP4_GPIO_CLEARWKUENA		0x0180
+#define OMAP4_GPIO_SETWKUENA		0x0184
 #define OMAP4_GPIO_CLEARDATAOUT		0x0190
 #define OMAP4_GPIO_SETDATAOUT		0x0194
 /*
@@ -591,12 +599,16 @@ static int _get_gpio_dataout(struct gpio_bank *bank, int gpio)
 		reg += OMAP7XX_GPIO_DATA_OUTPUT;
 		break;
 #endif
-#ifdef CONFIG_ARCH_OMAP2PLUS
+#if defined(CONFIG_ARCH_OMAP2) || defined(CONFIG_ARCH_OMAP3)
 	case METHOD_GPIO_24XX:
-	case METHOD_GPIO_44XX:
 		reg += OMAP24XX_GPIO_DATAOUT;
 		break;
 #endif
+#ifdef CONFIG_ARCH_OMAP4
+	case METHOD_GPIO_44XX:
+		reg += OMAP4_GPIO_DATAOUT;
+		break;
+#endif
 	default:
 		return -EINVAL;
 	}
@@ -1200,11 +1215,17 @@ static int omap_gpio_request(struct gpio_chip *chip, unsigned offset)
 #endif
 	if (!cpu_class_is_omap1()) {
 		if (!bank->mod_usage) {
+			void __iomem *reg = bank->base;
 			u32 ctrl;
-			ctrl = __raw_readl(bank->base + OMAP24XX_GPIO_CTRL);
-			ctrl &= 0xFFFFFFFE;
+
+			if (cpu_is_omap24xx() || cpu_is_omap34xx())
+				reg += OMAP24XX_GPIO_CTRL;
+			else if (cpu_is_omap44xx())
+				reg += OMAP4_GPIO_CTRL;
+			ctrl = __raw_readl(reg);
 			/* Module is enabled, clocks are not gated */
-			__raw_writel(ctrl, bank->base + OMAP24XX_GPIO_CTRL);
+			ctrl &= 0xFFFFFFFE;
+			__raw_writel(ctrl, reg);
 		}
 		bank->mod_usage |= 1 << offset;
 	}
@@ -1226,22 +1247,34 @@ static void omap_gpio_free(struct gpio_chip *chip, unsigned offset)
 		__raw_writel(1 << offset, reg);
 	}
 #endif
-#ifdef CONFIG_ARCH_OMAP2PLUS
-	if ((bank->method == METHOD_GPIO_24XX) ||
-			(bank->method == METHOD_GPIO_44XX)) {
+#if defined(CONFIG_ARCH_OMAP2) || defined(CONFIG_ARCH_OMAP3)
+	if (bank->method == METHOD_GPIO_24XX) {
 		/* Disable wake-up during idle for dynamic tick */
 		void __iomem *reg = bank->base + OMAP24XX_GPIO_CLEARWKUENA;
 		__raw_writel(1 << offset, reg);
 	}
 #endif
+#ifdef CONFIG_ARCH_OMAP4
+	if (bank->method == METHOD_GPIO_44XX) {
+		/* Disable wake-up during idle for dynamic tick */
+		void __iomem *reg = bank->base + OMAP4_GPIO_IRQWAKEN0;
+		__raw_writel(1 << offset, reg);
+	}
+#endif
 	if (!cpu_class_is_omap1()) {
 		bank->mod_usage &= ~(1 << offset);
 		if (!bank->mod_usage) {
+			void __iomem *reg = bank->base;
 			u32 ctrl;
-			ctrl = __raw_readl(bank->base + OMAP24XX_GPIO_CTRL);
+
+			if (cpu_is_omap24xx() || cpu_is_omap34xx())
+				reg += OMAP24XX_GPIO_CTRL;
+			else if (cpu_is_omap44xx())
+				reg += OMAP4_GPIO_CTRL;
+			ctrl = __raw_readl(reg);
 			/* Module is disabled, clocks are gated */
 			ctrl |= 1;
-			__raw_writel(ctrl, bank->base + OMAP24XX_GPIO_CTRL);
+			__raw_writel(ctrl, reg);
 		}
 	}
 	_reset_gpio(bank, bank->chip.base + offset);
@@ -1570,9 +1603,14 @@ static int gpio_is_input(struct gpio_bank *bank, int mask)
 		reg += OMAP7XX_GPIO_DIR_CONTROL;
 		break;
 	case METHOD_GPIO_24XX:
-	case METHOD_GPIO_44XX:
 		reg += OMAP24XX_GPIO_OE;
 		break;
+	case METHOD_GPIO_44XX:
+		reg += OMAP4_GPIO_OE;
+		break;
+	default:
+		WARN_ONCE(1, "gpio_is_input: incorrect OMAP GPIO method");
+		return -EINVAL;
 	}
 	return __raw_readl(reg) & mask;
 }
Signature · 1 line
1.6.3.3
#1 lore raw

Tony,

If there are no review comments, can you please push this patch?

-V Charulatha

Quoted 147 lines

-----Original Message-----
From: Varadarajan, Charulatha
Sent: Thursday, April 08, 2010 4:28 PM
To: [email protected]
Cc: Varadarajan, Charulatha; Kevin Hilman
Subject: [PATCH V2] OMAP: GPIO: Fix OMAP4 GPIO reg access issues

Access to some of the OMAP4 GPIO registers are not properly handled. This patch fixes it.

This patch is tested on 3430SDP and 4430SDP boards

---
 arch/arm/plat-omap/gpio.c |   61 +++++++++++++++++++++++++++++++++++---------
 1 files changed, 55 insertions(+), 12 deletions(-)
arch/arm/plat-omap/gpio.c +15 -3
index 76a347b..3a9891c 100644
@@ -137,7 +137,11 @@
 #define OMAP4_GPIO_IRQSTATUSCLR1	0x0040
 #define OMAP4_GPIO_IRQWAKEN0		0x0044
 #define OMAP4_GPIO_IRQWAKEN1		0x0048
-#define OMAP4_GPIO_SYSSTATUS		0x0104
+#define OMAP4_GPIO_SYSSTATUS		0x0114
+#define OMAP4_GPIO_IRQENABLE1		0x011c
+#define OMAP4_GPIO_WAKE_EN		0x0120
+#define OMAP4_GPIO_IRQSTATUS2		0x0128
+#define OMAP4_GPIO_IRQENABLE2		0x012c
 #define OMAP4_GPIO_CTRL			0x0130
 #define OMAP4_GPIO_OE			0x0134
 #define OMAP4_GPIO_DATAIN		0x0138
@@ -148,6 +152,10 @@
 #define OMAP4_GPIO_FALLINGDETECT	0x014c
 #define OMAP4_GPIO_DEBOUNCENABLE	0x0150
 #define OMAP4_GPIO_DEBOUNCINGTIME	0x0154
+#define OMAP4_GPIO_CLEARIRQENABLE1	0x0160
+#define OMAP4_GPIO_SETIRQENABLE1	0x0164
+#define OMAP4_GPIO_CLEARWKUENA		0x0180
+#define OMAP4_GPIO_SETWKUENA		0x0184
 #define OMAP4_GPIO_CLEARDATAOUT		0x0190
 #define OMAP4_GPIO_SETDATAOUT		0x0194
 /*
@@ -591,12 +599,16 @@ static int _get_gpio_dataout(struct gpio_bank *bank, int
 gpio)
 		reg += OMAP7XX_GPIO_DATA_OUTPUT;
 		break;
 #endif
-#ifdef CONFIG_ARCH_OMAP2PLUS
+#if defined(CONFIG_ARCH_OMAP2) || defined(CONFIG_ARCH_OMAP3)
 	case METHOD_GPIO_24XX:
-	case METHOD_GPIO_44XX:
 		reg += OMAP24XX_GPIO_DATAOUT;
 		break;
 #endif
+#ifdef CONFIG_ARCH_OMAP4
+	case METHOD_GPIO_44XX:
+		reg += OMAP4_GPIO_DATAOUT;
+		break;
+#endif
 	default:
 		return -EINVAL;
arch/arm/plat-omap/gpio.c +9 -3
 	}
@@ -1200,11 +1215,17 @@ static int omap_gpio_request(struct gpio_chip *chip,
 unsigned offset)
 #endif
 	if (!cpu_class_is_omap1()) {
 		if (!bank->mod_usage) {
+			void __iomem *reg = bank->base;
 			u32 ctrl;
-			ctrl = __raw_readl(bank->base + OMAP24XX_GPIO_CTRL);
-			ctrl &= 0xFFFFFFFE;
+
+			if (cpu_is_omap24xx() || cpu_is_omap34xx())
+				reg += OMAP24XX_GPIO_CTRL;
+			else if (cpu_is_omap44xx())
+				reg += OMAP4_GPIO_CTRL;
+			ctrl = __raw_readl(reg);
 			/* Module is enabled, clocks are not gated */
-			__raw_writel(ctrl, bank->base + OMAP24XX_GPIO_CTRL);
+			ctrl &= 0xFFFFFFFE;
+			__raw_writel(ctrl, reg);
 		}
 		bank->mod_usage |= 1 << offset;
arch/arm/plat-omap/gpio.c +17 -5
 	}
@@ -1226,22 +1247,34 @@ static void omap_gpio_free(struct gpio_chip *chip,
 unsigned offset)
 		__raw_writel(1 << offset, reg);
 	}
 #endif
-#ifdef CONFIG_ARCH_OMAP2PLUS
-	if ((bank->method == METHOD_GPIO_24XX) ||
-			(bank->method == METHOD_GPIO_44XX)) {
+#if defined(CONFIG_ARCH_OMAP2) || defined(CONFIG_ARCH_OMAP3)
+	if (bank->method == METHOD_GPIO_24XX) {
 		/* Disable wake-up during idle for dynamic tick */
 		void __iomem *reg = bank->base + OMAP24XX_GPIO_CLEARWKUENA;
 		__raw_writel(1 << offset, reg);
 	}
 #endif
+#ifdef CONFIG_ARCH_OMAP4
+	if (bank->method == METHOD_GPIO_44XX) {
+		/* Disable wake-up during idle for dynamic tick */
+		void __iomem *reg = bank->base + OMAP4_GPIO_IRQWAKEN0;
+		__raw_writel(1 << offset, reg);
+	}
+#endif
 	if (!cpu_class_is_omap1()) {
 		bank->mod_usage &= ~(1 << offset);
 		if (!bank->mod_usage) {
+			void __iomem *reg = bank->base;
 			u32 ctrl;
-			ctrl = __raw_readl(bank->base + OMAP24XX_GPIO_CTRL);
+
+			if (cpu_is_omap24xx() || cpu_is_omap34xx())
+				reg += OMAP24XX_GPIO_CTRL;
+			else if (cpu_is_omap44xx())
+				reg += OMAP4_GPIO_CTRL;
+			ctrl = __raw_readl(reg);
 			/* Module is disabled, clocks are gated */
 			ctrl |= 1;
-			__raw_writel(ctrl, bank->base + OMAP24XX_GPIO_CTRL);
+			__raw_writel(ctrl, reg);
 		}
 	}
arch/arm/plat-omap/gpio.c +6 -2
 	_reset_gpio(bank, bank->chip.base + offset);
@@ -1570,9 +1603,14 @@ static int gpio_is_input(struct gpio_bank *bank, int mask)
 		reg += OMAP7XX_GPIO_DIR_CONTROL;
 		break;
 	case METHOD_GPIO_24XX:
-	case METHOD_GPIO_44XX:
 		reg += OMAP24XX_GPIO_OE;
 		break;
+	case METHOD_GPIO_44XX:
+		reg += OMAP4_GPIO_OE;
+		break;
+	default:
+		WARN_ONCE(1, "gpio_is_input: incorrect OMAP GPIO method");
+		return -EINVAL;
 	}
 	return __raw_readl(reg) & mask;
 }
--

1.6.3.3

Acked-by on v2 1/1
#2 lore raw
In reply to Charulatha V
Quoted 7 lines · Charulatha V

Access to some of the OMAP4 GPIO registers are not properly handled. This patch fixes it.

This patch is tested on 3430SDP and 4430SDP boards

All this #ifdeffery will be gone with the in-progress hwmod conversion, but this fix should go in in the meantime.

Kevin

#3 lore raw

OMAP Runtime APIs

Kevin,

While adapting some drivers to HWMOD way, I tried using pm_runtime APIs. My understanding is that instead of using clock FW APIs or omap_device APIs, we can make use of pm_runtime APIs.

But, I observed that pm_runtime_enable() does not enable the clocks (fck/ick).

To test this, I used origin/pm-wip/mmc branch.

In omap_hsmmc_probe() function, I tried printing the values of CM_FCLKEN1_CORE and CM_ICLKEN1_CORE registers before and after pm_runtime_enable(). The values indicated that both before and after pm_runtime_enable(), the MMC clocks (fclk & iclk) are disabled.

Strangely, I do not find any crash while accessing MMC registers with mmc iclk disabled.

If I use clock FW APIs to enable these clocks (after pm_runtime_enable()) and print these values, I could see those clocks enabled properly and values printed accordingly.

I observed similar behavior with Watchdog and found that access of some watchdog registers end up in crash if I use pm_runtime APIs (as the wdt iclk is not enabled while using pm_runtime_enable()), but works fine if I use clk FW APIS to enable clks.

I am using origin/pm-wip/runtime branch with patches to adapt watchdog to HWMOD FW on top of.

I use "omap3_defconfig" and test on 3430SDP.

On mmc branch I am working on top of the following commit:
commit fe64f0fab492651d7f9296b9d830f2c11844b546

Author: Kevin Hilman <[email protected]>
Date:   Wed Jan 20 09:48:52 2010 -0800

OMAP2/3 MMC: initial conversion to runtime PM

Please clarify.

Thanks,
V Charulatha

#4 lore raw

Re: OMAP Runtime APIs

Hi Charu,

Quoted 6 lines · Charulatha

While adapting some drivers to HWMOD way, I tried using pm_runtime APIs. My understanding is that instead of using clock FW APIs or omap_device APIs, we can make use of pm_runtime APIs.

But, I observed that pm_runtime_enable() does not enable the clocks (fck/ick).

pm_runtime_enable() is not the equivalent of enabling the clocks. This function simply enables runtime PM _functionality_ for that device, it doesn't actually do any HW manipulation.

The way that clocks are enabled/disabled is by using the pm_runtime_get*() and pm_runtime_put*() calls. These do usage couting in the runtime PM core and when there's a zero use-count, then pm_runtime_idle() is called which will use omap_device API to disable the clocks. Similar for the enable path.

[...]

Quoted 12 lines

I am using origin/pm-wip/runtime branch with patches to adapt watchdog to HWMOD FW on top of.

I use "omap3_defconfig" and test on 3430SDP.

On mmc branch I am working on top of the following commit:
commit fe64f0fab492651d7f9296b9d830f2c11844b546

Author: Kevin Hilman <[email protected]>
Date:   Wed Jan 20 09:48:52 2010 -0800

OMAP2/3 MMC: initial conversion to runtime PM

I recommend you look at the details of this patch you will see that the clock enable/disable calls were replaced by runtime PM _get() and _put() calls.

The _enable() call is just done at _probe() time.

Hope that helps,

Kevin

#5 lore raw

PATCH v2 1/1 OMAP: GPIO: Fix OMAP4 GPIO reg access issues

In reply to Charulatha V

This patch has been applied to the linux-omap
by youw fwiendly patch wobot.

Branch in linux-omap: for-next

Initial commit ID (Likely to change): 27d714284a31693362bafeb124d5c400db20e2ad

PatchWorks
http://patchwork.kernel.org/patch/91247/

Git (Likely to change, and takes a while to get mirrored)
http://git.kernel.org/?p=linux/kernel/git/tmlind/linux-omap-2.6.git;a=commit;h=27d714284a31693362bafeb124d5c400db20e2ad