Jason Liu wrote:

Hi, Stefano

Hi Jason,

 MX53 has different flash header structure than mx51/35/25, which

means I need change the tools/imximage.c files

I can see something in the patch you have already sent. However, I do not know the details, and the documentation on Freescale's website is not yet released.

 I want to add one imximage_v2.c and .h file and register the

imx_v2 format to system. Do you think it's accepted or not.

I do not know, it depends if this is the best solution expressed by the ML ;-). Is there a way to get a preview about the flash header required by the MX.53 ?

I try to make some assumptions based on the flash_header.S file you delivered in your patchset, just to share some ideas.

It seems that for MX.53 all (or most) values are written with big endianess. This could be managed with a tag in the configuration file, something like:

ENDIAN BIG

If it is not set, values must be written in little endian mode as for MX.51.

Except for the endianess, the DCD table seems to be the same (two integers for each entry, address and value), but with a different magic number as for MX51 (dcd_hdr). Is there any internal name in Freescale to distinguish between the flash header for the MX51 and for the MX53 (or other processors of the family) ? Some version numbers ? It could be possible to add this information to the configuration file to inform the imximage utility which kind of header must be generated. Feel free to add any kind of additional tags to the configuration file you think can be useful.

I do not know which kind of modifications you plane to do, but I think it would be preferable to have a single utility to manage the IMX header. I mean, you can add additional files, but it would be bad if the parser and some other parts will be duplicated. I think it is better to extend the actual parser to support the MX53 changes. Some values in your flash header seems to be addresses, as

boot_data: .word 0x77800000

Is it ? Then yes, I think these values should not be hard-coded, and they can be computed by the imximage as it does currently for MX51 (check for example to the app_code_jump_vector and dcd_ptr_ptr in the code).

By the way, I posted your question on the u-boot ML, if someone else would like to contribute to this topic.

Best regards, Stefano