[U-Boot-Users] [AT91] Problems configuring bit-banged I2C
Hi, all!
I use the following patch on my AT91SAM9260-based custom board to configure bit-banged I2C:
diff --git a/board/atmel/at91sam9260ek/at91sam9260ek.c b/board/atmel/at91sam9260ek/at91sam9260ek.c index b30aad8..99fc2bd 100644 --- a/board/atmel/at91sam9260ek/at91sam9260ek.c +++ b/board/atmel/at91sam9260ek/at91sam9260ek.c @ -234,3 +234,46 @@ void reset_phy(void) #endif } #endif + +#define GPIO_I2C_SCL AT91_PIN_PA24 +#define GPIO_I2C_SDA AT91_PIN_PA23 + +void at91sam9260_i2c_init(void) +{ + __raw_writel(pin_to_mask(GPIO_I2C_SCL), pin_to_controller(GPIO_I2C_SCL) + PIO_IDR); + __raw_writel(pin_to_mask(GPIO_I2C_SCL), pin_to_controller(GPIO_I2C_SCL) + PIO_PUER); + __raw_writel(pin_to_mask(GPIO_I2C_SCL), pin_to_controller(GPIO_I2C_SCL) + PIO_PER); + __raw_writel(pin_to_mask(GPIO_I2C_SCL), pin_to_controller(GPIO_I2C_SCL) + PIO_MDER); + __raw_writel(pin_to_mask(GPIO_I2C_SCL), pin_to_controller(GPIO_I2C_SCL) + PIO_SODR); + __raw_writel(pin_to_mask(GPIO_I2C_SCL), pin_to_controller(GPIO_I2C_SCL) + PIO_OER); + + __raw_writel(pin_to_mask(GPIO_I2C_SDA), pin_to_controller(GPIO_I2C_SDA) + PIO_IDR); + __raw_writel(pin_to_mask(GPIO_I2C_SDA), pin_to_controller(GPIO_I2C_SDA) + PIO_PUER); + __raw_writel(pin_to_mask(GPIO_I2C_SDA), pin_to_controller(GPIO_I2C_SDA) + PIO_PER); + __raw_writel(pin_to_mask(GPIO_I2C_SDA), pin_to_controller(GPIO_I2C_SDA) + PIO_MDER); + __raw_writel(pin_to_mask(GPIO_I2C_SDA), pin_to_controller(GPIO_I2C_SDA) + PIO_SODR); + __raw_writel(pin_to_mask(GPIO_I2C_SDA), pin_to_controller(GPIO_I2C_SDA) + PIO_OER); +} +void at91sam9260_i2c_scl(unsigned long bit) +{ + if(bit) + __raw_writel(pin_to_mask(GPIO_I2C_SCL), pin_to_controller(GPIO_I2C_SCL) + PIO_SODR); + else + __raw_writel(pin_to_mask(GPIO_I2C_SCL), pin_to_controller(GPIO_I2C_SCL) + PIO_CODR); +} + +void at91sam9260_i2c_sda(unsigned long bit) +{ + printf("Setting %u\n", bit); + if(bit) + __raw_writel(pin_to_mask(GPIO_I2C_SDA), pin_to_controller(GPIO_I2C_SDA) + PIO_SODR); + else + __raw_writel(pin_to_mask(GPIO_I2C_SDA), pin_to_controller(GPIO_I2C_SDA) + PIO_CODR); +} + +int at91sam9260_i2c_read(void) +{ + int c = (__raw_readl(pin_to_controller(GPIO_I2C_SDA) + PIO_PDSR) & pin_to_mask(GPIO_I2C_SDA)) ? 1 : 0; + printf("Reading %d\n", c); + return c; +} diff --git a/include/configs/at91sam9260ek.h b/include/configs/at91sam9260ek.h index fd9932f..5e66c9e 100644 --- a/include/configs/at91sam9260ek.h +++ b/include/configs/at91sam9260ek.h @@ -85,6 +85,8 @@
#define CONFIG_CMD_NAND 1 #define CONFIG_CMD_USB 1 +#define CONFIG_CMD_I2C 1 +#define CONFIG_I2C_CMD_TREE 1
/* SDRAM */ #define CONFIG_NR_DRAM_BANKS 1 @ -192,4 +194,20 @@ #error CONFIG_USE_IRQ not supported #endif
+/* + * Software (bit-bang) I2C driver configuration + */ +#define CONFIG_SOFT_I2C 1 +#define CFG_I2C_SPEED 1000000 +#define I2C_INIT at91sam9260_i2c_init() + +#define I2C_ACTIVE +#define I2C_TRISTATE +#define I2C_SCL(bit) at91sam9260_i2c_scl(bit) +#define I2C_SDA(bit) at91sam9260_i2c_sda(bit) +#define I2C_DELAY udelay(2) +#define I2C_READ at91sam9260_i2c_read() +#define CFG_I2C_SLAVE 0 +#define DEBUG_I2C #endif +
But I'm unable to read proper values from lines - they always return 0s. Any ideas on fixing? Linux i2c-gpio driver works perfectly.
Thanks a lot, S.
On 03:36 Sat 24 May , Sergey Lapin wrote:
Hi, all!
I use the following patch on my AT91SAM9260-based custom board to configure bit-banged I2C:
diff --git a/board/atmel/at91sam9260ek/at91sam9260ek.c b/board/atmel/at91sam9260ek/at91sam9260ek.c index b30aad8..99fc2bd 100644 --- a/board/atmel/at91sam9260ek/at91sam9260ek.c +++ b/board/atmel/at91sam9260ek/at91sam9260ek.c @ -234,3 +234,46 @@ void reset_phy(void) #endif } #endif
Why don't you use the gpio function?
void at91sam9260_i2c_init(void) { at91_set_GPIO_periph(AT91_PIN_PA23, 0); at91_set_multi_drive(AT91_PIN_PA23, 1);
at91_set_A_periph(AT91_PIN_PA24, 0); at91_set_multi_drive(AT91_PIN_PA24, 1); }
void at91sam9260_i2c_scl(unsigned long bit) { at91_set_gpio_value(AT91_PIN_PA24, bit); }
void at91sam9260_i2c_sda(unsigned long bit) { at91_set_gpio_value(AT91_PIN_PA23, bit) } int at91sam9260_i2c_read(void) { retrun at91_get_gpio_value(AT91_PIN_PA23); } void at91sam9260_i2c_active() { gpio_direction_output(AT91_PIN_PA23, 0); }
void at91sam9260_i2c_tristate() { gpio_direction_input(AT91_PIN_PA23); }
diff --git a/include/configs/at91sam9260ek.h b/include/configs/at91sam9260ek.h index fd9932f..5e66c9e 100644 --- a/include/configs/at91sam9260ek.h +++ b/include/configs/at91sam9260ek.h @@ -85,6 +85,8 @@
#define CONFIG_CMD_NAND 1 #define CONFIG_CMD_USB 1 +#define CONFIG_CMD_I2C 1 +#define CONFIG_I2C_CMD_TREE 1
/* SDRAM */ #define CONFIG_NR_DRAM_BANKS 1 @ -192,4 +194,20 @@ #error CONFIG_USE_IRQ not supported #endif
+/*
- Software (bit-bang) I2C driver configuration
- */
+#define CONFIG_SOFT_I2C 1 +#define CFG_I2C_SPEED 1000000 +#define I2C_INIT at91sam9260_i2c_init()
#define I2C_ACTIVE at91sam9260_i2c_active() #define I2C_TRISTATE at91sam9260_i2c_tristate()
+#define I2C_SCL(bit) at91sam9260_i2c_scl(bit) +#define I2C_SDA(bit) at91sam9260_i2c_sda(bit) +#define I2C_DELAY udelay(2) +#define I2C_READ at91sam9260_i2c_read() +#define CFG_I2C_SLAVE 0 +#define DEBUG_I2C #endif
I sync it will be better to use the hardware stack which is not too complicate to implement
Best Regards, J.
Why don't you use the gpio function?
Well, I tried to use them but that lead to same result.
void at91sam9260_i2c_init(void) { at91_set_GPIO_periph(AT91_PIN_PA23, 0); at91_set_multi_drive(AT91_PIN_PA23, 1);
at91_set_A_periph(AT91_PIN_PA24, 0);
Why?
at91_set_multi_drive(AT91_PIN_PA24, 1); }
void at91sam9260_i2c_scl(unsigned long bit) { at91_set_gpio_value(AT91_PIN_PA24, bit); }
void at91sam9260_i2c_sda(unsigned long bit) { at91_set_gpio_value(AT91_PIN_PA23, bit) } int at91sam9260_i2c_read(void) { retrun at91_get_gpio_value(AT91_PIN_PA23); } void at91sam9260_i2c_active() { gpio_direction_output(AT91_PIN_PA23, 0); }
void at91sam9260_i2c_tristate() { gpio_direction_input(AT91_PIN_PA23); }
Thanks a lot for your help, but this did not work either. iprobe returns full 0x0 - 0x7f range as devices. Linux i2c-gpio works properly and returns proper set of devices (0x68, 0x50). Is it me or there's bug somewhere? I can see output values with oscilloscope. I wonder what happens.
I sync it will be better to use the hardware stack which is not too complicate to implement
As commented in Linux kernel about i2c_at91, it is very buggy, and I was afraid that it could not work for me at all. i2c-gpio works in Linux. What prevents soft i2c from working ont the same CPU? This stuff is a mystery for me :(
Sergey Lapin slapin@fan.ossfans.org wrote:
But I'm unable to read proper values from lines - they always return 0s. Any ideas on fixing? Linux i2c-gpio driver works perfectly.
Please try this patch:
http://www.nabble.com/-PATCH--soft_i2c%3A-Pull-SDA-high-before-reading-p1727...
You shouldn't need to do anything in the I2C_ACTIVE and I2C_TRISTATE hooks since the PIO is properly configured in open drain mode.
Haavard
On Sat, May 24, 2008 at 05:24:00PM +0200, Haavard Skinnemoen wrote:
Sergey Lapin slapin@fan.ossfans.org wrote:
But I'm unable to read proper values from lines - they always return 0s. Any ideas on fixing? Linux i2c-gpio driver works perfectly.
Please try this patch:
http://www.nabble.com/-PATCH--soft_i2c%3A-Pull-SDA-high-before-reading-p1727...
You shouldn't need to do anything in the I2C_ACTIVE and I2C_TRISTATE hooks since the PIO is properly configured in open drain mode.
Alas, I was unable to make bit-banged I2C to work even with this patch applied. As I mentioned above, it works in Linux, but I'd like to be able to set time to my RTC from u-boot. I have checked that pins are configured properly from status registers (pull-ups are enabled, multi-drive mode is enabled (open drain), etc.). I could read only zeroes from SDA line in u-boot, and using oscilloscope, I see data on line. I can't find differencies in setup with i2c-gpio from Linux. Any ideas?
Thanks a lot, S.
participants (4)
-
Haavard Skinnemoen -
Jean-Christophe PLAGNIOL-VILLARD -
Sergey Lapin -
Sergey Lapin