In the implementation of NVMalloc, some different free list_heads may point to memory regions with overlapping areas. For example, the figure above is extracted from your paper, in this figure, the second 1-bl list_head points to the chunk with one block length, the first 2-bl list_head points to the chunk with two blocks length, and the 3-bl list_head points to the chunk with three blocks length, these chunks have some overlapping areas, so these list_heads are related. I think when a chunk is allocated, the list_head that points to this chunk need to be released, at the same time, the related list_heads need to be released too. However, in your code, I only found that you free the former list_head, and the related list_heads are not freed. This will cause the later allocations may allocate to the memory blocks that is in use.

In order to check this problem, I added a judgement statement in function makeOne(), this function is called in getFreeLocation(). When a free memory is allocated, makeOne() will modify the corresponding state value in bitmap. If the allocated memory is in use, bitmap[index] & mask will be 1, I will print the error and exit the program. In test I did encounter this error. So I am wondering if you have omitted this point. If so, please let me know.