i2c_read_reg on phytec-som-am335x returns wrong data from EEPROM

[Ian Abbott]

On 02/08/17 16:48, Martin Hollingsworth wrote:
> Hello folks,
> I am having trouble reading from the 4k EEPROM via I2C on the phytec-som-am335x board. What I want to achieve is, that the code in board.c reads a value from EEPROM, evaluates it and then adds the value as globalvar to the userland for scripting. However if I read the data using the function from below, the i2c_read_reg() returns with SUCCESS but the data read to the buffer is incorrect. I use the following snippet of code to read the data within the physom_devices_init() function within the board.c file.
> 
> ------------
> struct i2c_adapter *adapter = i2c_get_adapter(0);
> if(adapter)
> {
>      struct i2c_client client;
>      client.adapter = adapter;
>      client.addr = 0x52;
> 	char buff[32] = {0xFF};
>      if(i2c_read_reg(&client, 0xFE0, buff, sizeof(buff)) > 0)
>      {
> 		for( ii = 0; ii < 32; ii++ )
> 		{
> 			printf( "%02X", buff[ii] );
> 		}
> 	[...]
> ------------
> 
> If I compare the EEPROM access in barebox userland, I see similar problems. All access to the EEPROM via I2C commands only work correct, if I set the "-w" (use 16bit word access) flag. In all cases the i2c_read or i2c_write functions return 0 (success) but the read or write was not executed correctly. See the example, where both reads should return the same data:
> 
> ------------
> barebox at Phytec phyCORE AM335x:/ i2c_read -b 0 -a 0x52 -r 0xFE0 -c 32 -w
> 0xaa 0xff 0xaa 0xff [...] --> OK, DATA is correct
> barebox at Phytec phyCORE AM335x:/ i2c_read -b 0 -a 0x52 -r 0xFE0 -c 32
> 0xf3 0xfd 0x54 0x23 [...] --> RETURNVALUE = 0 (SUCCESS), BUT READ DATA IS WRONG
> ------------
> 
> However if I access the EEPROM using the device definition /dev/eeprom0, as provided by the DTSI file, the usual memory commands in userland - like memset, mw and md - all work just fine.
> 
> What am I overlooking here? How do I configure i2c_read_reg() in C-Code to do WORD access to the EEPROM?
> 
> Any suggestion would be appreciated.
> Regards, Martin

I believe it's just a case of OR'ing the register address (0xFE0 in your 
case) with I2C_ADDR_16_BIT.  This tells the i2c_read_reg() and 
i2c_write_reg() functions in "drivers/i2c/i2c.c" to send a two-byte 
address (MSB followed by LSB) to the I2C client instead of the 
single-byte address that most I2C client devices expect.

-- 
-=( Ian Abbott @ MEV Ltd.    E-mail: <abbotti at mev.co.uk> )=-
-=(                          Web: http://www.mev.co.uk/  )=-