1signalfd(2) System Calls Manual signalfd(2)
2
3
4
6 signalfd - create a file descriptor for accepting signals
7
9 Standard C library (libc, -lc)
10
12 #include <sys/signalfd.h>
13
14 int signalfd(int fd, const sigset_t *mask, int flags);
15
17 signalfd() creates a file descriptor that can be used to accept signals
18 targeted at the caller. This provides an alternative to the use of a
19 signal handler or sigwaitinfo(2), and has the advantage that the file
20 descriptor may be monitored by select(2), poll(2), and epoll(7).
21
22 The mask argument specifies the set of signals that the caller wishes
23 to accept via the file descriptor. This argument is a signal set whose
24 contents can be initialized using the macros described in sigsetops(3).
25 Normally, the set of signals to be received via the file descriptor
26 should be blocked using sigprocmask(2), to prevent the signals being
27 handled according to their default dispositions. It is not possible to
28 receive SIGKILL or SIGSTOP signals via a signalfd file descriptor;
29 these signals are silently ignored if specified in mask.
30
31 If the fd argument is -1, then the call creates a new file descriptor
32 and associates the signal set specified in mask with that file descrip‐
33 tor. If fd is not -1, then it must specify a valid existing signalfd
34 file descriptor, and mask is used to replace the signal set associated
35 with that file descriptor.
36
37 Starting with Linux 2.6.27, the following values may be bitwise ORed in
38 flags to change the behavior of signalfd():
39
40 SFD_NONBLOCK Set the O_NONBLOCK file status flag on the open file de‐
41 scription (see open(2)) referred to by the new file de‐
42 scriptor. Using this flag saves extra calls to fcntl(2)
43 to achieve the same result.
44
45 SFD_CLOEXEC Set the close-on-exec (FD_CLOEXEC) flag on the new file
46 descriptor. See the description of the O_CLOEXEC flag in
47 open(2) for reasons why this may be useful.
48
49 Up to Linux 2.6.26, the flags argument is unused, and must be specified
50 as zero.
51
52 signalfd() returns a file descriptor that supports the following opera‐
53 tions:
54
55 read(2)
56 If one or more of the signals specified in mask is pending for
57 the process, then the buffer supplied to read(2) is used to re‐
58 turn one or more signalfd_siginfo structures (see below) that
59 describe the signals. The read(2) returns information for as
60 many signals as are pending and will fit in the supplied buffer.
61 The buffer must be at least sizeof(struct signalfd_siginfo)
62 bytes. The return value of the read(2) is the total number of
63 bytes read.
64
65 As a consequence of the read(2), the signals are consumed, so
66 that they are no longer pending for the process (i.e., will not
67 be caught by signal handlers, and cannot be accepted using sig‐
68 waitinfo(2)).
69
70 If none of the signals in mask is pending for the process, then
71 the read(2) either blocks until one of the signals in mask is
72 generated for the process, or fails with the error EAGAIN if the
73 file descriptor has been made nonblocking.
74
75 poll(2), select(2) (and similar)
76 The file descriptor is readable (the select(2) readfds argument;
77 the poll(2) POLLIN flag) if one or more of the signals in mask
78 is pending for the process.
79
80 The signalfd file descriptor also supports the other file-de‐
81 scriptor multiplexing APIs: pselect(2), ppoll(2), and epoll(7).
82
83 close(2)
84 When the file descriptor is no longer required it should be
85 closed. When all file descriptors associated with the same sig‐
86 nalfd object have been closed, the resources for object are
87 freed by the kernel.
88
89 The signalfd_siginfo structure
90 The format of the signalfd_siginfo structure(s) returned by read(2)s
91 from a signalfd file descriptor is as follows:
92
93 struct signalfd_siginfo {
94 uint32_t ssi_signo; /* Signal number */
95 int32_t ssi_errno; /* Error number (unused) */
96 int32_t ssi_code; /* Signal code */
97 uint32_t ssi_pid; /* PID of sender */
98 uint32_t ssi_uid; /* Real UID of sender */
99 int32_t ssi_fd; /* File descriptor (SIGIO) */
100 uint32_t ssi_tid; /* Kernel timer ID (POSIX timers)
101 uint32_t ssi_band; /* Band event (SIGIO) */
102 uint32_t ssi_overrun; /* POSIX timer overrun count */
103 uint32_t ssi_trapno; /* Trap number that caused signal */
104 int32_t ssi_status; /* Exit status or signal (SIGCHLD) */
105 int32_t ssi_int; /* Integer sent by sigqueue(3) */
106 uint64_t ssi_ptr; /* Pointer sent by sigqueue(3) */
107 uint64_t ssi_utime; /* User CPU time consumed (SIGCHLD) */
108 uint64_t ssi_stime; /* System CPU time consumed
109 (SIGCHLD) */
110 uint64_t ssi_addr; /* Address that generated signal
111 (for hardware-generated signals) */
112 uint16_t ssi_addr_lsb; /* Least significant bit of address
113 (SIGBUS; since Linux 2.6.37) */
114 uint8_t pad[X]; /* Pad size to 128 bytes (allow for
115 additional fields in the future) */
116 };
117
118 Each of the fields in this structure is analogous to the similarly
119 named field in the siginfo_t structure. The siginfo_t structure is de‐
120 scribed in sigaction(2). Not all fields in the returned signalfd_sig‐
121 info structure will be valid for a specific signal; the set of valid
122 fields can be determined from the value returned in the ssi_code field.
123 This field is the analog of the siginfo_t si_code field; see sigac‐
124 tion(2) for details.
125
126 fork(2) semantics
127 After a fork(2), the child inherits a copy of the signalfd file de‐
128 scriptor. A read(2) from the file descriptor in the child will return
129 information about signals queued to the child.
130
131 Semantics of file descriptor passing
132 As with other file descriptors, signalfd file descriptors can be passed
133 to another process via a UNIX domain socket (see unix(7)). In the re‐
134 ceiving process, a read(2) from the received file descriptor will re‐
135 turn information about signals queued to that process.
136
137 execve(2) semantics
138 Just like any other file descriptor, a signalfd file descriptor remains
139 open across an execve(2), unless it has been marked for close-on-exec
140 (see fcntl(2)). Any signals that were available for reading before the
141 execve(2) remain available to the newly loaded program. (This is anal‐
142 ogous to traditional signal semantics, where a blocked signal that is
143 pending remains pending across an execve(2).)
144
145 Thread semantics
146 The semantics of signalfd file descriptors in a multithreaded program
147 mirror the standard semantics for signals. In other words, when a
148 thread reads from a signalfd file descriptor, it will read the signals
149 that are directed to the thread itself and the signals that are di‐
150 rected to the process (i.e., the entire thread group). (A thread will
151 not be able to read signals that are directed to other threads in the
152 process.)
153
154 epoll(7) semantics
155 If a process adds (via epoll_ctl(2)) a signalfd file descriptor to an
156 epoll(7) instance, then epoll_wait(2) returns events only for signals
157 sent to that process. In particular, if the process then uses fork(2)
158 to create a child process, then the child will be able to read(2) sig‐
159 nals that are sent to it using the signalfd file descriptor, but
160 epoll_wait(2) will not indicate that the signalfd file descriptor is
161 ready. In this scenario, a possible workaround is that after the
162 fork(2), the child process can close the signalfd file descriptor that
163 it inherited from the parent process and then create another signalfd
164 file descriptor and add it to the epoll instance. Alternatively, the
165 parent and the child could delay creating their (separate) signalfd
166 file descriptors and adding them to the epoll instance until after the
167 call to fork(2).
168
170 On success, signalfd() returns a signalfd file descriptor; this is ei‐
171 ther a new file descriptor (if fd was -1), or fd if fd was a valid sig‐
172 nalfd file descriptor. On error, -1 is returned and errno is set to
173 indicate the error.
174
176 EBADF The fd file descriptor is not a valid file descriptor.
177
178 EINVAL fd is not a valid signalfd file descriptor.
179
180 EINVAL flags is invalid; or, in Linux 2.6.26 or earlier, flags is non‐
181 zero.
182
183 EMFILE The per-process limit on the number of open file descriptors has
184 been reached.
185
186 ENFILE The system-wide limit on the total number of open files has been
187 reached.
188
189 ENODEV Could not mount (internal) anonymous inode device.
190
191 ENOMEM There was insufficient memory to create a new signalfd file de‐
192 scriptor.
193
195 C library/kernel differences
196 The underlying Linux system call requires an additional argument,
197 size_t sizemask, which specifies the size of the mask argument. The
198 glibc signalfd() wrapper function does not include this argument, since
199 it provides the required value for the underlying system call.
200
201 There are two underlying Linux system calls: signalfd() and the more
202 recent signalfd4(). The former system call does not implement a flags
203 argument. The latter system call implements the flags values described
204 above. Starting with glibc 2.9, the signalfd() wrapper function will
205 use signalfd4() where it is available.
206
208 Linux.
209
211 signalfd()
212 Linux 2.6.22, glibc 2.8.
213
214 signalfd4()
215 Linux 2.6.27.
216
218 A process can create multiple signalfd file descriptors. This makes it
219 possible to accept different signals on different file descriptors.
220 (This may be useful if monitoring the file descriptors using select(2),
221 poll(2), or epoll(7): the arrival of different signals will make dif‐
222 ferent file descriptors ready.) If a signal appears in the mask of
223 more than one of the file descriptors, then occurrences of that signal
224 can be read (once) from any one of the file descriptors.
225
226 Attempts to include SIGKILL and SIGSTOP in mask are silently ignored.
227
228 The signal mask employed by a signalfd file descriptor can be viewed
229 via the entry for the corresponding file descriptor in the process's
230 /proc/pid/fdinfo directory. See proc(5) for further details.
231
232 Limitations
233 The signalfd mechanism can't be used to receive signals that are syn‐
234 chronously generated, such as the SIGSEGV signal that results from ac‐
235 cessing an invalid memory address or the SIGFPE signal that results
236 from an arithmetic error. Such signals can be caught only via signal
237 handler.
238
239 As described above, in normal usage one blocks the signals that will be
240 accepted via signalfd(). If spawning a child process to execute a
241 helper program (that does not need the signalfd file descriptor), then,
242 after the call to fork(2), you will normally want to unblock those sig‐
243 nals before calling execve(2), so that the helper program can see any
244 signals that it expects to see. Be aware, however, that this won't be
245 possible in the case of a helper program spawned behind the scenes by
246 any library function that the program may call. In such cases, one
247 must fall back to using a traditional signal handler that writes to a
248 file descriptor monitored by select(2), poll(2), or epoll(7).
249
251 Before Linux 2.6.25, the ssi_ptr and ssi_int fields are not filled in
252 with the data accompanying a signal sent by sigqueue(3).
253
255 The program below accepts the signals SIGINT and SIGQUIT via a signalfd
256 file descriptor. The program terminates after accepting a SIGQUIT sig‐
257 nal. The following shell session demonstrates the use of the program:
258
259 $ ./signalfd_demo
260 ^C # Control-C generates SIGINT
261 Got SIGINT
262 ^C
263 Got SIGINT
264 ^\ # Control-\ generates SIGQUIT
265 Got SIGQUIT
266 $
267
268 Program source
269
270 #include <err.h>
271 #include <signal.h>
272 #include <stdio.h>
273 #include <stdlib.h>
274 #include <sys/signalfd.h>
275 #include <unistd.h>
276
277 int
278 main(void)
279 {
280 int sfd;
281 ssize_t s;
282 sigset_t mask;
283 struct signalfd_siginfo fdsi;
284
285 sigemptyset(&mask);
286 sigaddset(&mask, SIGINT);
287 sigaddset(&mask, SIGQUIT);
288
289 /* Block signals so that they aren't handled
290 according to their default dispositions. */
291
292 if (sigprocmask(SIG_BLOCK, &mask, NULL) == -1)
293 err(EXIT_FAILURE, "sigprocmask");
294
295 sfd = signalfd(-1, &mask, 0);
296 if (sfd == -1)
297 err(EXIT_FAILURE, "signalfd");
298
299 for (;;) {
300 s = read(sfd, &fdsi, sizeof(fdsi));
301 if (s != sizeof(fdsi))
302 err(EXIT_FAILURE, "read");
303
304 if (fdsi.ssi_signo == SIGINT) {
305 printf("Got SIGINT\n");
306 } else if (fdsi.ssi_signo == SIGQUIT) {
307 printf("Got SIGQUIT\n");
308 exit(EXIT_SUCCESS);
309 } else {
310 printf("Read unexpected signal\n");
311 }
312 }
313 }
314
316 eventfd(2), poll(2), read(2), select(2), sigaction(2), sigprocmask(2),
317 sigwaitinfo(2), timerfd_create(2), sigsetops(3), sigwait(3), epoll(7),
318 signal(7)
319
320
321
322Linux man-pages 6.04 2023-03-30 signalfd(2)