CODEWITHSUNDEEP
cembeddedfirmware
CodeWarrior
CodeWarrior

Reputation: 1

Need to migrate my firmware image to ROM mask

I need to create a ROM mask that provides some functions. However, it should be possible to overwrite the functions for providing firmware patches. Therefore, a patch table should be located in a Flash memory that may be overwritten by later firmware upgrades, whereas the main part of the firmware is located in mask ROM and cannot be modified later.

Does anyone have any idea how this is done? what is the best practice to create patch tables?

Upvotes: 0

Views: 447

Answers (1)

Michael Dorgan
Michael Dorgan

Reputation: 12515

Patch table - Basically your ROM has one more built in jumps to a RAM (Flash) adddresses built into it. The flash always has some sort of jump back to the loction just after your ROM jump call to return to your original code. You now have the ability to change your program's behavior from RAM. This assumes you are allowed to run code from RAM of course. If not, then only data tables can be changed on the fly.

Now, just having a single jump on startup allows you to modify starting state code like version numbers or other global/constant data, but that may not be enough. You can add another jump to flash that is run every "so often" to your code as well so that it can update program state after it is running - something like once a vblank or once a application loop.

The above should give you enough lee-way to change data that has changed over time since release or perhaps to fix a slight logic error, but it won't allow you to whole-sale change functions. To do that, you need more code. And that more code depends on just how much RAM you can use.

For instance, if you had enough RAM, and running from Flash RAM didn't hurt perforamnce too much (and is allowed), you could make things very flexible by copying all your key ROM functions into RAM on startup, then calling your RAM patch jump described above, which would allow new code stored elsewhere in your RAM patch to overwrite any previously copied code. If you took this approach, you would also want to make sure that you left some extra space around your original functions to allow new padded functions a bit of room to grow. The original code to be copied to RAM could be stored compressed as well to save ROM space. This allows you to change anything after the fact. It also introduces a very easy way to exploit your code if something else is allowed to write to RAM so beware.

Hope this helps.

Upvotes: 1

Related Questions