In a multithreaded computing environment, hazard pointers are one approach to solving the problems posed by dynamic memory management of the nodes in a lock-free data structure. These problems generally arise only in environments that don't have automatic garbage collection.[1]
Any lock-free data structure that uses the compare-and-swap primitive must deal with the ABA problem. For example, in a lock-free stack represented as an intrusively linked list, one thread may be attempting to pop an item from the front of the stack (A → B → C). It remembers the second-from-top value "B", and then performs
Furthermore, any lock-free algorithm containing code of the formthis->head and deallocate it, meaning that the memory access through currentNode on the second line reads deallocated memory (which may in fact already be in use by some other thread for a completely different purpose).
Hazard pointers can be used to address both of these problems. In a hazard-pointer system, each thread keeps a list of hazard pointers indicating which nodes the thread is currently accessing. (In many systems this "list" may be probably limited to only one[1] [2] or two elements.) Nodes on the hazard pointer list must not be modified or deallocated by any other thread.
When a thread wishes to remove a node, it places it on a list of nodes "to be freed later", but does not actually deallocate the node's memory until no other thread's hazard list contains the pointer. This manual garbage collection can be done by a dedicated garbage-collection thread (if the list "to be freed later" is shared by all the threads); alternatively, cleaning up the "to be freed" list can be done by each worker thread as part of an operation such as "pop" (in which case each worker thread can be responsible for its own "to be freed" list).
In 2002, Maged Michael of IBM filed an application for a U.S. patent on the hazard pointer technique, but the application was abandoned in 2010.
Alternatives to hazard pointers include reference counting.[1]