Stop parsing on invalid or trailing data

[fabian-z]

Breaking out of the parsing loop when it is impossible to parse the read block as either data or spare entry allows yaffshiv to be used with images extracted using a detected start offset, e.g. using binwalk without a known length.

Also adjust the brute force condition, since the user requested it and FS parsing may return zero objects for incorrect settings, making brute force impossible.

[fabian-z]

I also validated this change using real device NAND dumps and with generated test images, but input regarding different images would be welcome. Ping @ValdikSS as original code author :smile:

[ValdikSS]

Would it work for a NAND dump where not all blocks have been used just yet and are still filled with FF's?

[fabian-z]

I could not identify a regression for this patch testing with real device dumps (reverse engineering), however the current version already fails to identify settings / find files in some cases for me. When trying to build a test environment (Linux 4.14 with nandsim and yaffs kernel module), I am hitting several different cases of invalid parsing as well, regardless if using upstream or my patch. So it is currently difficult for me to provide a reproducing test case for yaffshiv.

If 0xFF blocks are of concern, the patch can be easily adjusted to work around this issue with a code block like this:

            data_empty, spare_empty = True, True
            (obj_hdr_data, obj_hdr_spare) = self.read_block()
            for b in obj_hdr_data:
                if b != 255:
                    data_empty = False
                    break
            for b in obj_hdr_spare:
                if b != 255:
                    spare_empty = False
                    break
            if data_empty and spare_empty:
                continue

I would appreciate testing with different real-world images to ensure no regressions happen here :+1: Thanks for your efforts!

[ValdikSS]

Well, don't have much images, so I can't test thoroughly.