Advice on using kernel module "GPIO Buttons" ( "GPIO Keys") with DM8168 ?

Ed Sutton

My goal is to use the "GPIO Keys" or gpio-keys driver rather than /sys/class/gpio. I understand the gpio-keys solution will allow a button press to generate an event as if it came from a keyboard.

Here is what I have done so far:

I connected a push button to GPO_IO6 (pin G5).
I verified that pushing button pulls voltage down to ground.
I configured and built a new uImage kernel using menuconfig to add support for InputDeviceSupport-Keyboards->"GPIO Buttons" in addition to /sys/class/gpio/ and Basic memory-mapped GPI controllers support.
After boot I see only a mouse under /dev/input/

--- GPIO Support

[ ] Debug GPIO calls

[*] /sys/class/gpio/... (sysfs interface)

*** Memory mapped GPIO expanders: ***

<*> Basic memory-mapped GPIO controllers support

< > IT8761E GPIO support

When using the sys interface, I see no change in value when pressing or releasing the button:

/sys/devices/virtual/gpio/gpio6# cat value

When I read PINCTRL297 associated with GPO6 it is set to 0x02:

root@dm816x-evm:/dev/input# devmem2 0x48140CA0

/dev/mem opened.

Memory mapped at address 0x40258000.

Read at address 0x48140CA0 (0x40258ca0): 0x00000002

I tried setting PINCTRL297 to 000 MUXMODE[2:0]. I assumed it should be in this mode with no change in the sys cat value reported when I push a button.

Using devmen2 to read from GPIO0 @ 0x48032000:

GPIO_DATAIN offset 138h, the state remains the same regardless of button status.

root@dm816x-evm:/sys/devices/virtual/gpio/gpio6# devmem2 0x48032138

/dev/mem opened.

Memory mapped at address 0x400e4000.

Read at address 0x48032138 (0x400e4138): 0x00000008

The value read at GPIO_OE offset 134h indicates GP06 is configured as an input:

root@dm816x-evm:/sys/devices/virtual/gpio/gpio6# devmem2 0x48032134

/dev/mem opened.

Memory mapped at address 0x40250000.

Read at address 0x48032134 (0x40250134): 0xFFEFFFF0

Are there additional kernel configuration required? Or is there a possible conflict between loading multiple GPIO associated drivers such as /sys/class/gpio and "GPIO Buttons"? I see no GPIO messages in the boot up either good or bad. I am running out of ideas. Any tips or suggestions are much appreciated.

Thanks in advance for any direction,

-Ed

over 12 years ago

0 Pavel Botev over 12 years ago

TI__Guru**** 170625 points

Hi Ed,

The first thing we should verify is that the PINCTRL297[2:0] MUXMODE is 0x0 (GP0[6]), not 0x2 (GPMC_A[23]). From your post, I can not be sure that you achieve to set 0x0 in PINCTRL297[2:0] MUXMODE bit field. Could you please confirm that you achieved? If not, I will check how this can be set to 0x0.

Regards,

Pavel

0 Ed Sutton over 12 years ago in reply to Pavel Botev

Intellectual 885 points

Hi Pavel,

Once again thank you for your valued assistance.

It appears that writing 0x0 to PINCTRL297 is not accepted. It remains at 0x02.

When I use devmen2 to read PINCTRL297 it is 0x2.
When I write a 0x0 to PINCTRL297 the readback displays 0 and appears it succeeded.
When I read read PINCTRL297 a second time it is back to 0x2?

Output captured from script running in console:

PINCTRL297 @ 0x48140CA0

Read PINCTRL

devmem2 0x48140CA0 w

/dev/mem opened.

Memory mapped at address 0x4030c000.

Read at address 0x48140CA0 (0x4030cca0): 0x00000002

Configure PINCTRL to MUXMODE 0x0 (GPO[6])

devmem2 0x48140CA0 w 0x0

/dev/mem opened.

Memory mapped at address 0x4024f000.

Read at address 0x48140CA0 (0x4024fca0): 0x00000002

Write at address 0x48140CA0 (0x4024fca0): 0x00000000, readback 0x00000000

Read PINCTRL after set to 0x0

devmem2 0x48140CA0 w

/dev/mem opened.

Memory mapped at address 0x400ae000.

Read at address 0x48140CA0 (0x400aeca0): 0x00000002

*** Error PINCTRL returned 0x02, expected to be 0x0

0 Pavel Botev over 12 years ago in reply to Ed Sutton

TI__Guru**** 170625 points

Ed,

I also can not change the value of PINCTRL297 to 0x0 from user space. Thus we need to change it from the linux kernel. You should modify the below file as:

{EZSDK}/board-support/linux-2.6.37-psp04.04.00.01/arch/arm/mach-omap2/board-ti8168evm.c

static struct omap_board_mux board_mux[] __initdata = {

   /* PIN mux for non-muxed NOR */
   TI816X_MUX(TIM7_OUT, OMAP_MUX_MODE1),   /* gpmc_a12 */
   TI816X_MUX(UART1_CTSN, OMAP_MUX_MODE1),   /* gpmc_a13 */
   TI816X_MUX(UART1_RTSN, OMAP_MUX_MODE1),   /* gpmc_a14 */
   TI816X_MUX(UART2_RTSN, OMAP_MUX_MODE1),   /* gpmc_a15 */
   /* REVISIT: why 2 lines configured as gpmc_a15 */
   TI816X_MUX(SC1_RST, OMAP_MUX_MODE1),   /* gpmc_a15 */
   TI816X_MUX(UART2_CTSN, OMAP_MUX_MODE1),   /* gpmc_a16 */
   TI816X_MUX(UART0_RIN, OMAP_MUX_MODE1),   /* gpmc_a17 */
   TI816X_MUX(UART0_DCDN, OMAP_MUX_MODE1),   /* gpmc_a18 */
   TI816X_MUX(UART0_DSRN, OMAP_MUX_MODE1),   /* gpmc_a19 */
   TI816X_MUX(UART0_DTRN, OMAP_MUX_MODE1),   /* gpmc_a20 */
   TI816X_MUX(SPI_SCS3, OMAP_MUX_MODE1),   /* gpmc_a21 */
   TI816X_MUX(SPI_SCS2, OMAP_MUX_MODE1),   /* gpmc_a22 */
   //TI816X_MUX(GP0_IO6, OMAP_MUX_MODE2),   /* gpmc_a23 */
   TI816X_MUX(TIM6_OUT, OMAP_MUX_MODE1),   /* gpmc-a24 */
   TI816X_MUX(SC0_DATA, OMAP_MUX_MODE1),   /* gpmc_a25 */
   /* for controlling high address */
   TI816X_MUX(GPMC_A27, OMAP_MUX_MODE1),   /* gpio-20 */
   { .reg_offset = OMAP_MUX_TERMINATOR },
};

Now you should build new uImage and use it when booting the board. The value should be 0x0.

Regards,

Pavel

0 Ed Sutton over 12 years ago in reply to Pavel Botev

Intellectual 885 points

Thank you Pavel! PICCTRL297 is now MUXMODE 0x0 and I can access GP0_I06 status using devmem2 and see the button status change.

When rebuilding the uImage is it good practice to do a clean first? Or a "make clean" and a "make distclean" command? And should the CROSS_COMPILE and ARCH arguments be present? All I did was:

cd board-support/linux-2.6.37-psp04.04.00.01/

make CROSS_COMPILE=arm-none-linux-gnueabi- ARCH=arm uImage

Hopefully I can make sense on how to use gpio-keys now and connect GP0_I04 (brighness+) and GP0_I05 (brightness-) to the PWM control you had helped me get working.

According to this article, I expected to see /dev/input/event* folder to appear but it did not:

http://docs.blackfin.uclinux.org/doku.php?id=linux-kernel:drivers:gpio-keys

Could there be a gpio-keys board support source file that needs editing first?  There are many source files under /arch/arm that include gpio_keys.h but I cannot find one associated with the ti8168 EVM.

/board-support/linux-2.6.37-psp04.04.00.01/arch/arm

There are 8 that source files that include gpio_keys.h under mach-omap2:

find mach-omap2 -iname '*' | xargs grep 'gpio_keys.h' -sl

mach-omap2/board-rx51-peripherals.c

mach-omap2/board-omap3beagle.c

mach-omap2/board-ldp.c

mach-omap2/board-omap3stalker.c

mach-omap2/board-omap3touchbook.c

mach-omap2/board-devkit8000.c

mach-omap2/board-omap3pandora.c

mach-omap2/board-4430sdp.c

-Ed

0 Ed Sutton over 12 years ago in reply to Ed Sutton

Intellectual 885 points

I am getting the feeling the answer may be "there is no existing gpio-keys support built-in for the TI DM8168 EVM". Is this correct?

And if I want support for gpio-keys I will need to add it myself to:

${EZSDK}/board-support/linux-2.6.37-psp04.04.00.01/arch/arm/mach-omap2/board-ti8168evm.c

At first this sounds like a daunting task to me. However after looking at this ADI article on GPIO-keys, is it possible adding support requires only a few paragraphs of code?

${EZSDK}/board-support/linux-2.6.37-psp04.04.00.01/arch/blackfin/mach-bf533/boards/stamp.c

#if defined(CONFIG_KEYBOARD_GPIO) || defined(CONFIG_KEYBOARD_GPIO_MODULE)

#include <linux/input.h>

#include <linux/gpio_keys.h>

static struct gpio_keys_button bfin_gpio_keys_table[] = {

{BTN_0, GPIO_PF5, 0, "gpio-keys: BTN0"},

{BTN_1, GPIO_PF6, 0, "gpio-keys: BTN1"},

{BTN_2, GPIO_PF8, 0, "gpio-keys: BTN2"},

};

static struct gpio_keys_platform_data bfin_gpio_keys_data = {

.buttons = bfin_gpio_keys_table,

.nbuttons = ARRAY_SIZE(bfin_gpio_keys_table),

};

static struct platform_device bfin_device_gpiokeys = {

.name = "gpio-keys",

.dev = {

.platform_data = &bfin_gpio_keys_data,

};

#endif

I am guessing it is not that straight forward on a non BlackFin processor?

Or maybe it is straight forward? The board-omap3beagle.c implementation only appears to be a few paragraphs of data structures:

static struct gpio_keys_button gpio_buttons[] = {

{

.code = BTN_EXTRA,

.gpio = 7,

.desc = "user",

.wakeup = 1,

};

static struct gpio_keys_platform_data gpio_key_info = {

.buttons = gpio_buttons,

.nbuttons = ARRAY_SIZE(gpio_buttons),

};

static struct platform_device keys_gpio = {

.name = "gpio-keys",

.id = -1,

.dev = {

.platform_data = &gpio_key_info,

};

static struct platform_device *omap3_beagle_devices[] __initdata = {

&leds_gpio,

&keys_gpio,

&beagle_dss_device,

};

Could I be on the right track?

-Ed

0 Ed Sutton over 12 years ago in reply to Ed Sutton

Intellectual 885 points

In an attempt to add gpio-keys support to the board-ti8168evm.c I copied and modified source from board-ti8168evm.c. It sort of works. Or at least I can see a new /dev/input/event0. Also it has taken over the /sys/class/gpio4 - gpio6 access which now reports that it can not be opened.

However, when I cat /dev/input/event0 and press my button (normally high) nothing happens. I tried macking active_low=1 with same result.

cat /sys/devices/platform/gpio-keys/input/input0/event0/dev

13:64

cat /proc/bus/input/devices

I: Bus=0019 Vendor=0001 Product=0001 Version=0100

N: Name="gpio-keys"

P: Phys=gpio-keys/input0

S: Sysfs=/devices/platform/gpio-keys/input/input0

U: Uniq=

H: Handlers=kbd event0

B: EV=3

B: KEY=3 0 0 0 40 0 0 0

// EES - Add support for GPIO (gpio-keys)

// Copied from board-omap3beagle.c

static struct gpio_keys_button gpio_buttons[] = {

{

.code = KEY_BRIGHTNESSUP,

.gpio = 4,

.active_low = 1,

.desc = "brightness+",

.wakeup = 1,

{

.code = KEY_BRIGHTNESSDOWN,

.gpio = 5,

.active_low = 1,

.desc = "brightness-",

.wakeup = 1,

{

.code = KEY_HOME,

.gpio = 6,

.active_low = 1,

.desc = "home",

.wakeup = 1,

};

static void __init ti8168_evm_init(void)

{

// EES

platform_add_devices(ti8168_devices, ARRAY_SIZE(ti8168_devices));

I will keep investigating.

-Ed

0 Ed Sutton over 12 years ago in reply to Ed Sutton

Intellectual 885 points

I checked the button was pulling voltage to ground again. I guess that was not really necessary since devmem2 could see the button status change. Or at least devmem2 could see the change before I added GPIO keys support. My input handlers display as:

cat /proc/bus/input/handlers

N: Number=0 Name=kbd

N: Number=1 Name=sysrq (filter)

N: Number=2 Name=mousedev Minor=32

N: Number=3 Name=evdev Minor=64

0 Pavel Botev over 12 years ago in reply to Ed Sutton

TI__Guru**** 170625 points

Hi Ed,

Regarding gpio_keys, I searched our DM81xx forums and wiki pages, but did not find anything useful. The same with the Linux forum.

I found something that might be in help in the Android forum:

http://e2e.ti.com/support/embedded/android/f/509/p/134217/489984.aspx

http://e2e.ti.com/support/embedded/android/f/509/p/207734/747366.aspx

Please check if these two threads can help. I will continue to investigate and let you know when I have something.

Regards,

Pavel

0 Ed Sutton over 12 years ago in reply to Pavel Botev

Intellectual 885 points

Hi Pavel,

It is working now. Thank you for your help! I had seen those links but re-reading them made me think it sounded like a mux mode problem even though you had showed me how get it working using devmem2. Adding the config for "GPIO Buttons" ( a.k.a gpio-keys ) apparently changes the mux mode or something that had been working previously with devmem2.

Now I can cat device input events and the event is actually working when I press GP0_IO6 defined as KEY_HOME.

The following first 2 lines are from GP0_IO6 button down and the last 2 lines are the button up.

root@dm816x-evm:~# cat /dev/input/event0 |hexdump -C                            
00000000  5e 57 76 50 bb 5f 09 00  01 00 66 00 01 00 00 00  |^WvP._....f.....|  
00000010  5e 57 76 50 c1 5f 09 00  00 00 00 00 00 00 00 00  |^WvP._..........|  
00000020  5f 57 76 50 5c 50 08 00  01 00 66 00 00 00 00 00  |_WvP\P....f.....|  
00000030  5f 57 76 50 61 50 08 00  00 00 00 00 00 00 00 00  |_WvPaP..........|

It was a puzzling problem. You showed me how to get the devmem2 interface working by commenting out the omap_board_mux board_mux[] data for GPO_IO6 in the board-ti8168evm.c file. This worked and devmem2 could definitely display button state changes. I am not sure if commenting out made it default to mode0 or whatever was defined in the mux81xx.c file. Incidentally I am not sure if mux81xx.c just defines the default state or simply declares the mux capable I/O. I think it is the later.

Anyway after adding an explicit OMAP_MUX_MODE0 statements to omap_board_mux board_mux[] for the desired GPIO all is good!

File: board-ti8168evm.c

#ifdef CONFIG_OMAP_MUX
static struct omap_board_mux board_mux[] __initdata = {

	/* PIN mux for non-muxed NOR */
	TI816X_MUX(TIM7_OUT, OMAP_MUX_MODE1),	/* gpmc_a12 */
	TI816X_MUX(UART1_CTSN, OMAP_MUX_MODE1),	/* gpmc_a13 */
	TI816X_MUX(UART1_RTSN, OMAP_MUX_MODE1),	/* gpmc_a14 */
	TI816X_MUX(UART2_RTSN, OMAP_MUX_MODE1),	/* gpmc_a15 */
	/* REVISIT: why 2 lines configured as gpmc_a15 */
	TI816X_MUX(SC1_RST, OMAP_MUX_MODE1),	/* gpmc_a15 */
	TI816X_MUX(UART2_CTSN, OMAP_MUX_MODE1),	/* gpmc_a16 */
	TI816X_MUX(UART0_RIN, OMAP_MUX_MODE1),	/* gpmc_a17 */
	TI816X_MUX(UART0_DCDN, OMAP_MUX_MODE1),	/* gpmc_a18 */
	TI816X_MUX(UART0_DSRN, OMAP_MUX_MODE1),	/* gpmc_a19 */
	TI816X_MUX(UART0_DTRN, OMAP_MUX_MODE1),	/* gpmc_a20 */
	TI816X_MUX(SPI_SCS3, OMAP_MUX_MODE1),	/* gpmc_a21 */
	TI816X_MUX(SPI_SCS2, OMAP_MUX_MODE1),	/* gpmc_a22 */

        // EES - GP0_IO4, GP0_IO5, GP0_IO6 are connected to the bezel buttons on the prototype
        // Configure GPIO used by gpio-keys ( CONFIG_KEYBOARD_GPIO=y ) to OMAP_MUX_MODE0
        TI816X_MUX(GP0_IO4, OMAP_MUX_MODE0),
        TI816X_MUX(GP0_IO5, OMAP_MUX_MODE0),	/* gpmc_a24 */
        TI816X_MUX(GP0_IO6, OMAP_MUX_MODE0),	/* gpmc_a23 */

	TI816X_MUX(TIM6_OUT, OMAP_MUX_MODE1),	/* gpmc-a24 */
	TI816X_MUX(SC0_DATA, OMAP_MUX_MODE1),	/* gpmc_a25 */
	/* for controlling high address */
	TI816X_MUX(GPMC_A27, OMAP_MUX_MODE1),	/* gpio-20 */
	{ .reg_offset = OMAP_MUX_TERMINATOR },
};
#else

So with your help, all is good once again! This embedded Linux stuff is starting to become more enjoyable finally.

-Ed

Processors

Processors forum

Advice on using kernel module "GPIO Buttons" ( "GPIO Keys") with DM8168 ?

Output captured from script running in console:

File: board-ti8168evm.c