Pool and reuse shared memory blocks. This saves having to reallocate and reshare pages of memory, which is slow. A block is half the size of a page, since network packets usually don't require more than this.
make grant_access is a shared pool of blocks. When a new page is first allocated,
grant_access is called to share it.
val use : t -> (id:Cstruct.uint16 -> Xen_os.Xen.Gntref.t -> Cstruct.t -> ('a * unit Lwt.t) Lwt.t) -> ('a * unit Lwt.t) Lwt.t
use t fn calls
fn ~id gref block with a free shared block of memory and that block's unique ID (note: the
gref is NOT unique). The function should return a thread that indicates when the request has been added to the queue, by returning a result value and a second thread indicating when the block can be returned to the pool.
val shutdown : t -> unit
Unshare and free all the pages of the pool once all outstanding requests have completed. Calling
shutdown will return an error.