ARM: pxa: palmtreo: a few treo fixes

Some miscellaneous fixes to the treo code.

Mike Dunn (3):
ARM: pxa: palmtreo: fix lcd initialization on treo680
ARM: pxa: palmtreo: fix #ifdefs for leds-gpio device
ARM: pxa: palmtreo: add docg4 device initialization

 arch/arm/mach-pxa/include/mach/palmtreo.h |    3 +
 arch/arm/mach-pxa/palmtreo.c              |  102 ++++++++++++++++++++++++++---
 2 files changed, 96 insertions(+), 9 deletions(-)

1.7.8.6

This patch gets the LCD working on my Palm Treo680. Changing the Palm Centro code is avoided (I don't have a Centro for testing at the moment), but some initialization is shuffled around a bit.

This basically adds some code that manages the three gpios that interface to the lcd on the Treo 680. The precise role of each gpio in the hardware architecture is not entirely clear to me; this patch is the result of trial-and-error and observing how the PalmOS code initializes the lcd.

The need for this patch is not evident when Linux is loaded from PalmOS, because at that point the lcd-related gpios have already been configured. But when booting the kernel by other means, this patch is required unless the bootloader has performed the necessary initialializations.

---

I suspect that at least some of these changes may apply to the Centro as well. When I get a hold of one, I will sort this all out.

 arch/arm/mach-pxa/include/mach/palmtreo.h |    3 +
 arch/arm/mach-pxa/palmtreo.c              |   59 +++++++++++++++++++++++++++--
 2 files changed, 58 insertions(+), 4 deletions(-)

index 2d3f14e..3266194 100644
@@ -44,6 +44,9 @@
 #define GPIO_NR_TREO680_VIBRATE_EN	44
 #define GPIO_NR_TREO680_KEYB_BL		24
 #define GPIO_NR_TREO680_BT_EN		43
+#define GPIO_NR_TREO680_LCD_POWER	77
+#define GPIO_NR_TREO680_LCD_EN		86
+#define GPIO_NR_TREO680_LCD_EN_N	25
 #endif /* CONFIG_MACH_TREO680 */
 
 /* Centro685 specific GPIOs */

1.7.8.6

Dear Mike Dunn,

I think you should Cc Haojian (done)

I think I have a spare device somewhere around. I might get it available to you somehow eventually -- but that'll need some further discussion. Let's postpone it for now.

Yay, good old times ;-)

You mean U-Boot ? Makes sense.
[...]

Yes thanks you Marek, and my aplogies Haojian. Actually I was negligent on the CC's of all these patches. When I repost, I'll expand the CC list.

[...]

I plan to start checking ebay soon. The Treos seemed widely available last I checked (at least the 680), and fairly cheap.

[...]

Well my u-boot port does initialize the lcd. But generally, it shouldn't be assumed. I discovered the need for this when I loaded the kernel with the emulator. It's actually the secondary program loader that does the initialization (when it displays the Palm logo).

I'm not 100% sure, but I think that the lcd controller has to be initialized while the gpios are driven high, then after that, driving one of them low turns the lcd on. This is why palm27x_lcd_init() is called from within treo680_lcd_init(). But yes, it could be cleaner if the hardware permits. I'll experiment some more.

Yes, this needs to be cleaned up. The label "lcd power" is thrown around too much, sometimes when actually referring to the backlight. And power management is another whole effort. I'm hoping to be able to take things step by step.

[...]

Thanks for having a look Marek. I'll see if the gpio initialization can be separated from the lcd initialization.

MIke

The #ifdefs around the leds-gpio device platform data are erroneous. Currently the device is not instantiated on the centro unless CONFIG_MACH_TREO680 is defined.

---

The #ifdefs within palmtreo_leds_init() are a little ugly, but I think the only alternative would be to instantiate the leds-gpio device platform data for both treo680 and centro, which would require some of the #defines in palmtreo.h to be moved outside their containing #ifdef CONFIG_MACH_{TREO680,CENTRO}.

 arch/arm/mach-pxa/palmtreo.c |   16 +++++++++-------
 1 files changed, 9 insertions(+), 7 deletions(-)

index dcc38a1..cb9cc7b 100644
@@ -357,7 +357,9 @@ static struct gpio_led_platform_data treo680_gpio_led_info = {
 	.leds		= treo680_gpio_leds,
 	.num_leds	= ARRAY_SIZE(treo680_gpio_leds),
 };
+#endif
 
+#ifdef CONFIG_MACH_CENTRO
 static struct gpio_led centro_gpio_leds[] = {
 	{
 		.name			= "centro:vibra:vibra",

@@ -381,25 +383,25 @@ static struct gpio_led_platform_data centro_gpio_led_info = {
 	.leds		= centro_gpio_leds,
 	.num_leds	= ARRAY_SIZE(centro_gpio_leds),
 };
+#endif
 
 static struct platform_device palmtreo_leds = {
 	.name   = "leds-gpio",
 	.id     = -1,
-	.dev    = {
-		.platform_data  = &treo680_gpio_led_info,
-	}
 };
 
 static void __init palmtreo_leds_init(void)
 {
+#ifdef CONFIG_MACH_CENTRO
 	if (machine_is_centro())
 		palmtreo_leds.dev.platform_data = &centro_gpio_led_info;
-
+#endif
+#ifdef CONFIG_MACH_TREO680
+	if (machine_is_treo680())
+		palmtreo_leds.dev.platform_data = &treo680_gpio_led_info;
+#endif
 	platform_device_register(&palmtreo_leds);
 }
-#else
-static inline void palmtreo_leds_init(void) {}
-#endif
 
 /******************************************************************************
  * Machine init

1.7.8.6

Dear Mike Dunn,

Shall we not just kill all those ifdefs and let GCC decide what to optimize out? I think this kind of simple stuff should just be optimized out anyway.

Best regards,
Marek Vasut

You're suggesting that all the #ifdef CONFIG_MACH_{TREO680,CENTRO} be removed from palmtreo.h? Yes, this might be better. I'll look into it.

Thanks again,
Mike

This patch adds initialization of the docg4 nand flash device to the treo680.

---
 arch/arm/mach-pxa/palmtreo.c |   31 +++++++++++++++++++++++++++++++
 1 files changed, 31 insertions(+), 0 deletions(-)

index cb9cc7b..f3e2d08 100644
@@ -404,6 +404,36 @@ static void __init palmtreo_leds_init(void)
 }
 
 /******************************************************************************
+ * diskonchip docg4 flash
+ ******************************************************************************/
+#if defined(CONFIG_MACH_TREO680)
+/* REVISIT: does the centro have this device also? */
+#if defined(CONFIG_MTD_NAND_DOCG4) || defined(CONFIG_MTD_NAND_DOCG4_MODULE)
+static struct resource docg4_resources[] = {
+	{
+		.start	= 0x00000000,
+		.end	= 0x00001FFF,
+		.flags	= IORESOURCE_MEM,
+	},
+};
+
+static struct platform_device palmtreo680_docg4_flashdevice = {
+	.name   = "docg4",
+	.id     = -1,
+	.resource = docg4_resources,
+	.num_resources = ARRAY_SIZE(docg4_resources),
+};
+
+static void __init palmtreo680_docg4_flash_init(void)
+{
+	platform_device_register(&palmtreo680_docg4_flashdevice);
+}
+#else
+static inline void palmtreo680_docg4_flash_init(void) {}
+#endif
+#endif
+
+/******************************************************************************
  * Machine init
  ******************************************************************************/
 static void __init treo_reserve(void)

@@ -482,6 +512,7 @@ static void __init treo680_init(void)
 	treo680_lcd_init();
 	palm27x_mmc_init(GPIO_NR_TREO_SD_DETECT_N, GPIO_NR_TREO680_SD_READONLY,
 			GPIO_NR_TREO680_SD_POWER, 0);
+	palmtreo680_docg4_flash_init();
 }
 #endif
 

1.7.8.6

Dear Mike Dunn,

Best regards,
Marek Vasut

Hello.

    #if IS_ENABLED(CONFIG_MTD_NAND_DOCG4)

Thanks Sergei. I see this in kconfig.h now. This will be another patch to replace all the "if defined" in this file.

Mike