Subversion Repositories openrisc_2011-10-31

                SPI flash programming application

This software will be compiled and have another application embedded in it, 
which will then be programmed into a flash memory via the SPI protocol.

At present, only the simple_spi core communicating with the M25P10 serial flash
memory via SPI is supported (the config on ORSoC FPGA board)

First, compile the program to be put into flash, and generate a binary image of
it.

The bootloader program that loads programs from the SPI flash, at reset, reads
out the first 4 bytes (32-bit word) from the flash and expects this to contain 
the length of the image in bytes. It then enters a loop for this many bytes
and copies them to the SDRAM, before jumping to the reset vector in SDRAM.

This means the image programed into the SPI flash must contain its size at the
first word, and the remainder of the image must be correctly aligned so that
the reset vector is at address 0x100 after being loaded.

Included with this software package is a tool which will take a binary 
application and embed its size in the first word. The makefile then takes this
altered binary image, and convert it into an object file with the appropriate
section name, and is linked into the final application by a custom linker 
script also included in this path.

First the program to be loaded into the flash memory should be compiled. It 
should be, of course, tested on the board first, before being loaded into the 
flash. Once the program is tested on the board, the ELF can be taken and 
objcopy'd to a binary image. It should be noted that the image should be 
addressed from 0, and the vectors should be correctly located.

Convert to the ELF to raw binary with:

$ or32-elf-objcopy -O binary myapp.elf myapp.bin

Now, the spiflash-program.elf program should be compiled, and the environment
variable PROGRAMMINGFILE should be set to point to myapp.bin, the binary 
version of the program we are to load into the flash memory.

To ensure the correct settings for UART and other peripherals are picked up 
ensure the BOARD_PATH variable is set correctly, too.

       make spiflash-program.elf PROGRAMMINGFILE=/path/to/myapp.bin \
            BOARD_PATH=`pwd`/../../../boards/actel/ordb1a3pe1500

It's advisable to do a "make distclean" first, to clear out any previously 
compiled software for other boards.

The steps to embed the size of the program in the binary and then link it 
into the final executable are all done automatically by the Makefile.

Finally, spiflash-program.elf should be loaded onto the board and executed. A
small prompt should come up, pressing 'h' will explain the options, 'p' will
do the programming (and verification of the image) and 'v' will verify the 
image again.

Warning:
Note that in the flash on the ORSoC dev board, there appears to be 3 pages, 
either pages 130-132 or pages 180-182 which do not provide reliable data when
read from. The cause of this is unknown. If this is a big problem, be sure that
no vital information is stored on those pages.

And of course, be sure that the program being loaded fits into the SPI flash.

Once the flash memory is programmed, and if the design is configured with the
SPI flash bootloader, resetting the target should cause the new application to
be executed.

Author: Julius Baxter, julius.baxter@orsoc.se