reentrancy

Reentrancy is the ability of code to be entered again before its previous execution completes. In programming, a reentrant function can be interrupted (by an interrupt, a signal, a callback, or another thread) and called again, re-entering while it is already running. This requires that the function not rely on or corrupt shared state across invocations.

Reentrant code is characterized by its use of local state and avoidance of mutable global data. It typically does not hold locks across calls or depend on non-thread-safe resources. It should not assume a single-threaded environment and must manage memory and external interactions in a way that does not break if another invocation occurs mid-execution.

Non-reentrant code relies on shared or persistent state, global variables, or resources that can be modified by concurrent invocations. Such code is prone to data races, corruption, or inconsistent results if re-entry occurs.

In software design, reentrancy is desirable for robust interrupt handling and asynchronous contexts, but it requires careful structuring. Non-reentrant patterns can lead to subtle bugs when callbacks or signals interrupt ongoing work.

In blockchain and smart contracts, reentrancy is a well-known vulnerability. A contract that calls an external contract before updating its own state can be re-entered by the callee to manipulate funds or state. Mitigations include following the checks-effects-interactions pattern, updating state before external calls, using reentrancy guards, and preferring pull payments or limited gas transfers.

See also: reentrancy guard, checks-effects-interactions pattern.