oom_get_cached_sff with length exceeding page boundary repeats same page

[donboll]

oom_get_cached_sff will accept any length. If the length exceeds the end of the page, it will call itself recursively with the length reduced by the amount already read from the page cache. The recursion does not increment the page number, so the data will keep coming from the same page.

Note, length less than 0 should return an exception as an error.

Probably should just restrict length to be positive, and offset + length to be <= 128 (no crossing page boundaries)