Linux Signals List — Numbers, Names and Default Actions

Linux signal reference

Type a name or number to find a signal. The last column is the exit code a shell reports when a process dies from that signal (128 + number).

Signal name, number or keyword

No.	Name	Default	Meaning / cause	Exit code
1	SIGHUP	Terminate	Hangup: the controlling terminal closed. Daemons reuse it to mean “reload configuration”.	129
2	SIGINT	Terminate	Interrupt from the keyboard (Ctrl-C).	130
3	SIGQUIT	Core dump	Quit from the keyboard (Ctrl-\); dumps core.	131
4	SIGILL	Core dump	Illegal instruction.	132
5	SIGTRAP	Core dump	Trace/breakpoint trap (debuggers).	133
6	SIGABRT	Core dump	Abort signal from abort().	134
7	SIGBUS	Core dump	Bus error: bad memory access (alignment, mmap).	135
8	SIGFPE	Core dump	Floating-point exception, including integer divide by zero.	136
9	SIGKILL	Terminate	Kill signal. Cannot be caught, blocked or ignored. Last resort; the process gets no chance to clean up.	137
10	SIGUSR1	Terminate	User-defined signal 1. nginx: reopen logs; dd: print progress; Apache: graceful restart.	138
11	SIGSEGV	Core dump	Segmentation fault: invalid memory reference.	139
12	SIGUSR2	Terminate	User-defined signal 2. nginx: binary upgrade; Gunicorn: re-exec.	140
13	SIGPIPE	Terminate	Write to a pipe with no reader. Default kills the process, which is why `yes \| head` stops.	141
14	SIGALRM	Terminate	Timer signal from alarm().	142
15	SIGTERM	Terminate	Termination request. The default for kill, systemd stop and docker stop; programs should catch it and shut down cleanly.	143
16	SIGSTKFLT	Terminate	Stack fault on coprocessor (unused).	144
17	SIGCHLD	Ignore	Child stopped or terminated.	145
18	SIGCONT	Continue	Continue if stopped (fg, kill -CONT).	146
19	SIGSTOP	Stop	Stop the process. Cannot be caught or ignored.	147
20	SIGTSTP	Stop	Stop typed at the terminal (Ctrl-Z).	148
21	SIGTTIN	Stop	Background process tried to read from the terminal.	149
22	SIGTTOU	Stop	Background process tried to write to the terminal.	150
23	SIGURG	Ignore	Urgent data on a socket.	151
24	SIGXCPU	Core dump	CPU time limit exceeded.	152
25	SIGXFSZ	Core dump	File size limit exceeded.	153
26	SIGVTALRM	Terminate	Virtual alarm clock.	154
27	SIGPROF	Terminate	Profiling timer expired.	155
28	SIGWINCH	Ignore	Terminal window size changed.	156
29	SIGIO	Terminate	I/O now possible (asynchronous I/O).	157
30	SIGPWR	Terminate	Power failure.	158
31	SIGSYS	Core dump	Bad system call: the seccomp filter rejected a syscall (systemd SystemCallFilter=, Docker default profile).	159

What a signal is

A signal is a one-word message the kernel delivers to a process: “the terminal closed”, “you touched bad memory”, “please stop”. The process may install a handler for most of them, ignore them, or accept the default action, which is usually to terminate (sometimes with a core dump). Two cannot be caught or ignored: SIGKILL ends the process on the spot and SIGSTOP freezes it.

TERM versus KILL

kill PID sends SIGTERM: a request. A well-written program catches it, finishes the current request, flushes files, closes connections and exits; that is what systemd, Docker and Kubernetes send first. kill -9 sends SIGKILL: the kernel removes the process without running any of its code, so locks stay held, temp files stay, and in-flight writes are lost. Use it only after TERM has been given time (systemd waits TimeoutStopSec, Docker 10 seconds by default) and failed.

Signals as an API

SIGHUP: daemons reload their configuration (nginx, sshd, HAProxy with -sf semantics via systemd reload). For a foreground process it means the terminal went away; nohup ignores it.
SIGUSR1 / SIGUSR2: program-defined. nginx reopens logs on USR1 and upgrades its binary on USR2; dd prints progress on USR1; Gunicorn and Unicorn use them for graceful restarts.
SIGPIPE: writing to a pipe whose reader has exited. Its default action, terminating, is why yes | head ends and why servers must ignore it or handle write errors.
SIGCHLD: a child exited; parents reap children with wait(), and a parent that does not leaves zombies.
SIGWINCH: the terminal was resized; editors and multiplexers redraw.

Sending signals

kill -TERM 1234        # by name (or kill -15 1234)
kill -s USR1 $(pidof nginx)
pkill -HUP sshd        # by process name
killall -9 runaway     # all processes with that name
systemctl kill -s HUP myservice
docker kill --signal=SIGUSR1 container

Numbers 1–15 are the same on every Unix; above that they differ by architecture (SIGUSR1 is 10 on x86 and 30 on SPARC/MIPS), which is why scripts should use names.

Questions

What is the difference between SIGTERM and SIGKILL?

SIGTERM asks the process to exit and can be handled; SIGKILL ends it immediately and cannot be caught. Always try TERM first; KILL is for processes that are stuck.

What is signal 11 / SIGSEGV?

A segmentation fault: the program read or wrote memory it does not own, usually a null or dangling pointer in C/C++ code or a bug in a native extension. The shell reports exit code 139.

Why did my process exit with 137 when nobody ran kill -9?

The kernel did: the out-of-memory killer sends SIGKILL to the largest process when memory runs out, and container runtimes do the same when a memory limit is hit. Check dmesg or the container’s OOMKilled flag.

Can I ignore SIGINT so Ctrl-C does not stop my script?

Yes: trap "" INT in bash, or signal.signal(SIGINT, SIG_IGN) in Python. Usually it is better to trap it and clean up before exiting, so the user can still stop the script.