- Modified blk_mq_start_request to be asynchronously
called from lcd, we don't wait for its response in the LCD
to issue a matching end_request.
- Refactored the mmap implementation with vm_insert_page,
as remap_pfn_range marks the pte as special. This is not good
for get_user_pages because it expects a actual struct page
backing the memory but remap_pfn_range assumes that memory
is device memory and marks the pte as special.
- Add a bunch of benchmarking code for profiling

Signed-off-by: Abhiram Balasubramanian <abhiram@cs.utah.edu>