On Mon, 2007-06-18 at 15:51 +0200, Nils N. Gjerdevik wrote:
Joakim Tjernlund wrote:
On Mon, 2007-06-18 at 13:39 +0200, Nils Gjerdevik wrote:
Hi, I'm trying to store the u-boot environment in an EEPROM (at24c128) on a custom at91rm9200 based board. This fails when using the included HARD_I2C driver, and since there are known problems with the I2C controller on this uC, I'm trying to set up SOFT_I2C instead, without success so far... This is what I've put in my config file:
#define I2C_INIT \ do { \ *AT91C_PIOA_PER = AT91C_PA25_TWD | AT91C_PA26_TWCK; \ *AT91C_PIOA_ODR = AT91C_PA25_TWD; \ *AT91C_PIOA_OER = AT91C_PA26_TWCK; \ }while(0)
#define I2C_ACTIVE (*AT91C_PIOA_OER = AT91C_PA25_TWD)
#define I2C_TRISTATE (*AT91C_PIOA_ODR = AT91C_PA25_TWD)
#define I2C_READ ((*AT91C_PIOA_PDSR & AT91C_PA25_TWD) != 0)
#define I2C_SDA(bit) \ if(bit) { \ *AT91C_PIOA_SODR = AT91C_PA25_TWD; \ } else { \ *AT91C_PIOA_CODR = AT91C_PA25_TWD; \ }
#define I2C_SCL(bit) \ if(bit) { \ *AT91C_PIOA_SODR = AT91C_PA26_TWCK; \ } else { \ *AT91C_PIOA_CODR = AT91C_PA26_TWCK; \ }
#define I2C_DELAY udelay(5)
Can anyone see what's wrong with this code? I can access the EEPROM without problems in Linux, using the bit-banging driver, so the board design should be OK. The code above is(loosely) based on the Linux bit-banging driver.
Regards, Nils
Hi Nils Don't know anything about at91rm9200, but you could test the below patch I sent a few weeks ago. I have tested it now and it works for me.
I think the README w.r.t I2C_TRISTATE is OK as is(don't change it). I do think the soft i2c driver is broken in several places w.r.t IC2_ACTIVE/I2C_TRISTATE and I2C reset sequence. The below patch is an attempt to fix it, but I havn't tested it.
Hi Joakim, I'm afraid your patch didn't help. All reads just return 0x00, so I guess it's related to my definitions for the at91rm9200 and not the generic soft_i2c code.
Nils
Nils, try changing ACTIVATE and TRISTATE to toggle direction, TRISTATE should set the pin to input and ACTIVE to output. I think you should keep my patch while doing so. If you can use a true open drain output, then ACTIVATE and TRISTATE should be empty.
Jocke