#include <signal.h> int kill(pid_t pid, int sig);
For a process to have permission to send a signal to a process designated by pid, unless the sending process has appropriate privileges, the real or effective user ID of the sending process shall match the real or saved set-user-ID of the receiving process.
If pid is greater than 0, sig shall be sent to the process whose process ID is equal to pid.
If pid is 0, sig shall be sent to all processes (excluding an unspecified set of system processes) whose process group ID is equal to the process group ID of the sender, and for which the process has permission to send a signal.
If pid is -1, sig shall be sent to all processes (excluding an unspecified set of system processes) for which the process has permission to send that signal.
If pid is negative, but not -1, sig shall be sent to all processes (excluding an unspecified set of system processes) whose process group ID is equal to the absolute value of pid, and for which the process has permission to send a signal.
If the value of pid causes sig to be generated for the sending process, and if sig is not blocked for the calling thread and if no other thread has sig unblocked or is waiting in a sigwait() function for sig, either sig or at least one pending unblocked signal shall be delivered to the sending thread before kill() returns.
The user ID tests described above shall not be applied when sending SIGCONT to a process that is a member of the same session as the sending process.
An implementation that provides extended security controls may impose further implementation-defined restrictions on the sending of signals, including the null signal. In particular, the system may deny the existence of some or all of the processes specified by pid.
Some implementations provide semantic extensions to the kill() function when the absolute value of pid is greater than some maximum, or otherwise special, value. Negative values are a flag to kill(). Since most implementations return [ESRCH] in this case, this behavior is not included in this volume of POSIX.1-2008, although a conforming implementation could provide such an extension.
The unspecified processes to which a signal cannot be sent may include the scheduler or init.
There was initially strong sentiment to specify that, if pid specifies that a signal be sent to the calling process and that signal is not blocked, that signal would be delivered before kill() returns. This would permit a process to call kill() and be guaranteed that the call never return. However, historical implementations that provide only the signal() function make only the weaker guarantee in this volume of POSIX.1-2008, because they only deliver one signal each time a process enters the kernel. Modifications to such implementations to support the sigaction() function generally require entry to the kernel following return from a signal-catching function, in order to restore the signal mask. Such modifications have the effect of satisfying the stronger requirement, at least when sigaction() is used, but not necessarily when signal() is used. The standard developers considered making the stronger requirement except when signal() is used, but felt this would be unnecessarily complex. Implementors are encouraged to meet the stronger requirement whenever possible. In practice, the weaker requirement is the same, except in the rare case when two signals arrive during a very short window. This reasoning also applies to a similar requirement for sigprocmask().
In 4.2 BSD, the SIGCONT signal can be sent to any descendant process regardless of user-ID security checks. This allows a job control shell to continue a job even if processes in the job have altered their user IDs (as in the su command). In keeping with the addition of the concept of sessions, similar functionality is provided by allowing the SIGCONT signal to be sent to any process in the same session regardless of user ID security checks. This is less restrictive than BSD in the sense that ancestor processes (in the same session) can now be the recipient. It is more restrictive than BSD in the sense that descendant processes that form new sessions are now subject to the user ID checks. A similar relaxation of security is not necessary for the other job control signals since those signals are typically sent by the terminal driver in recognition of special characters being typed; the terminal driver bypasses all security checks.
In secure implementations, a process may be restricted from sending a signal to a process having a different security label. In order to prevent the existence or nonexistence of a process from being used as a covert channel, such processes should appear nonexistent to the sender; that is, [ESRCH] should be returned, rather than [EPERM], if pid refers only to such processes.
Existing implementations vary on the result of a kill() with pid indicating an inactive process (a terminated process that has not been waited for by its parent). Some indicate success on such a call (subject to permission checking), while others give an error of [ESRCH]. Since the definition of process lifetime in this volume of POSIX.1-2008 covers inactive processes, the [ESRCH] error as described is inappropriate in this case. In particular, this means that an application cannot have a parent process check for termination of a particular child with kill(). (Usually this is done with the null signal; this can be done reliably with waitpid().)
There is some belief that the name kill() is misleading, since the function is not always intended to cause process termination. However, the name is common to all historical implementations, and any change would be in conflict with the goal of minimal changes to existing application code.
Any typographical or formatting errors that appear in this page are most likely to have been introduced during the conversion of the source files to man page format. To report such errors, see https://www.kernel.org/doc/man-pages/reporting_bugs.html .