In a SysV-like environment, one has the two types mcontext_t and
ucontext_t defined in <ucontext.h> and the four functions
getcontext(), setcontext(), makecontext() and
swapcontext() that allow user-level context switching between multiple
threads of control within a process.
The mcontext_t type is machine-dependent and opaque. The
ucontext_t type is a structure that has at least the following fields:
with sigset_t and stack_t defined in <signal.h>. Here
uc_link points to the context that will be resumed when the current
context terminates (in case the current context was created using
makecontext()), uc_sigmask is the set of signals blocked in this
context (see sigprocmask(2)), uc_stack is the stack used by this
context (see sigaltstack(2)), and uc_mcontext is the
machine-specific representation of the saved context, that includes the
calling thread's machine registers.
The function getcontext() initializes the structure pointed at by
ucp to the currently active context.
The function setcontext() restores the user context pointed at by
ucp. A successful call does not return. The context should have been
obtained by a call of getcontext(), or makecontext(), or passed
as third argument to a signal handler.
If the context was obtained by a call of getcontext(), program execution
continues as if this call just returned.
If the context was obtained by a call of makecontext(), program execution
continues by a call to the function func specified as the second
argument of that call to makecontext(). When the function func
returns, we continue with the uc_link member of the structure
ucp specified as the first argument of that call to
makecontext(). When this member is NULL, the thread exits.
If the context was obtained by a call to a signal handler, then old standard
text says that "program execution continues with the program instruction
following the instruction interrupted by the signal". However, this
sentence was removed in SUSv2, and the present verdict is "the result is
The earliest incarnation of this mechanism was the
setjmp()/longjmp() mechanism. Since that does not define the
handling of the signal context, the next stage was the
sigsetjmp()/siglongjmp() pair. The present mechanism gives much
more control. On the other hand, there is no easy way to detect whether a
return from getcontext() is from the first call, or via a
setcontext() call. The user has to invent her own bookkeeping device,
and a register variable won't do since registers are restored.
When a signal occurs, the current user context is saved and a new context is
created by the kernel for the signal handler. Do not leave the handler using
longjmp() - it is undefined what would happen with contexts. Use
siglongjmp() or setcontext() instead.