module Netmcore_mutex:sig
..end
Netmcore_heap
), for example in the header field of
Netmcore_array
or somewhere else in heap-allocated
data structures.
In order to ensure that the mutex is in the heap, the
special function create
must be used to initialize it
there. As create
requires a mutator as argument, this is
only possible by calling create
from the callback of
Netmcore_heap.modify
.
Mutexes are special values, and cannot be copied or moved.
Mutexes are implemented on top of semaphores. For mutex types with deadlock handling, the process is considered as the owner, and process ID is the identifier.
Only mutexes of type `Normal
are thread-safe, i.e. can be used
when the worker processes also use threads internally.
typemutex_type =
[ `Errorcheck | `Normal | `Recursive ]
`Normal
mutexes have no checks for deadlocks.`Errorcheck
mutexes check whether the process locks the mutex
again, and fail in this case. Also, only the owning process can
unlock a mutex.`Recursive
mutexes allow that the owner locks several times.
The same number of unlock requests need to be issued to give
up the ownershiptype
mutex
val dummy : unit -> mutex
mutex
-typed variablesval create : Netmcore_heap.mutator -> mutex_type -> mutex
create m t
: Creates a mutex of type t
, and
pushes it to the heap, using the mutator m
.
After being pushed to the heap, the mutex can be used. It is
nonsense to copy it outside the heap.
val lock : mutex -> unit
val unlock : mutex -> unit
val destroy : mutex -> unit