Control flow

In software, control flow (or flow of control) describes how execution progresses from one command to the next. In many contexts, such as machine code and an imperative programming language, control progresses sequentially (to the command located immediately after the currently executing command) except when a command transfers control to another point – in which case the command is classified as a control flow command. Depending on context, other terms are used instead of command. For example, in machine code, the typical term is instruction and in an imperative language, the typical term is statement. Although an imperative language encodes control flow explicitly, languages of other programming paradigms are less focused on control flow. A declarative language specifies desired results without prescribing an order of operations. A functional language uses both language constructs and functions to control flow even though they are usually not called control flow statements. For a central processing unit instruction set, a control flow instruction often alters the program counter and is either an unconditional branch (a.k.a. jump) or a conditional branch. An alternative approach is predication which conditionally enables instructions instead of branching. An asynchronous control flow transfer such as an interrupt or a signal alters the normal flow of control to a handler before returning control to where it was interrupted. One way to attack software is to redirect the flow of execution. A variety of control-flow integrity techniques, including stack canaries, buffer overflow protection, shadow stacks, and vtable pointer verification, are used to defend against these attacks.