The behavior of the global variable "pbl_cmd_initaddr" seems wrong, and prevent to use some address depending on the size of the SPL. This is maybe due to some alignment constraints but I didn't find any documentation that would explain it.
The variable is initialized like this: 288 pbl_cmd_initaddr = params->addr & PBL_ADDR_24BIT_MASK; 289 pbl_cmd_initaddr |= PBL_ACS_CONT_CMD; 290 pbl_cmd_initaddr |= uboot_size;
Then, during the computation of the addresses of the 64 bytes chunk, it is used like this: 84 next_pbl_cmd = pbl_cmd_initaddr - uboot_size;
So basically, if the size of the SPL is using n bits, we can't use the first n bits of the address range. In my particular case, I wanted to copy the SPL at address 0x1000_0a00 and mkimage computed 0x1000_0800 instead.
Why is "size" embedded like this in "pbl_cmd_initaddr" when it is a global variable too ?
Best,
Vincent