dm_eeprom: look up chosen bootcount device in the device tree

[thom-nic]

instead of using compile-time constants for bus/address/offset

Example device tree definition:

	chosen {
		// see: u-boot/drivers/bootcount/bootcount-uclass.c
		u-boot,bootcount-device = <&bootcount_i2c_eeprom>;
	};
	bootcount_i2c_eeprom: bc_i2c_eeprom {
		// see: u-boot/drivers/bootcount/bootcount_dm_i2c.c
		compatible = "u-boot,bootcount-i2c";
		i2cbcdev = <&eeprom0>;
		offset = <0x30>;
	};

[thom-nic]

Example debut output:

DEBUG: Read from /proc/device-tree/compatible: phytec,am625-phyboard-voltserver
DEBUG: Read from /proc/device-tree/compatible: phytec,am62-phycore-som
DEBUG: Read from /proc/device-tree/compatible: ti,am625
DEBUG: Discovering DM I2C EEPROM bootcount device...
DEBUG:  Found bootcount-device phandle 2
DEBUG:  Found bootcount node /sys/firmware/devicetree/base/bc_i2c_eeprom
DEBUG:  Using offset 0x30
DEBUG:  Found i2cbcdev phandle 90
DEBUG:  Found eeprom node /sys/firmware/devicetree/base/bus@f0000/i2c@20000000/eeprom@50
DEBUG:  EEPROM I2C address 0x50
DEBUG:  Checking device 0-0050 ...
DEBUG:  Matched EEPROM device 0-0050 (bus 0 addr 0x50)
Detected DM I2C EEPROM
DEBUG: Action=detect

[thom-nic]

@passgat I can't request you as a reviewer but I wonder if you'd be interested in looking at this. I am considering whether this implementation is sufficient or should I use libfdt instead.

Edit: upon further reflection I would rather not depend on libfdt unless it can be statically linked, in order to make it easy to include in initramfs. Given that the sysfs method that I'm using now seems to be reliable without any external dependency, I am apt to keep it as-is.