1SIGNALFD(2) Linux Programmer's Manual SIGNALFD(2)
2
6 signalfd - create a file descriptor for accepting signals
7
9 #include <sys/signalfd.h>
10 int signalfd(int fd, const sigset_t *mask, int flags);
12
14 signalfd() creates a file descriptor that can be used to accept signals
15 targeted at the caller. This provides an alternative to the use of a
16 signal handler or sigwaitinfo(2), and has the advantage that the file
17 descriptor may be monitored by select(2), poll(2), and epoll(7).
18 The mask argument specifies the set of signals that the caller wishes
20 to accept via the file descriptor. This argument is a signal set whose
21 contents can be initialized using the macros described in sigsetops(3).
22 Normally, the set of signals to be received via the file descriptor
23 should be blocked using sigprocmask(2), to prevent the signals being
24 handled according to their default dispositions. It is not possible to
25 receive SIGKILL or SIGSTOP signals via a signalfd file descriptor;
26 these signals are silently ignored if specified in mask.
27 If the fd argument is -1, then the call creates a new file descriptor
29 and associates the signal set specified in mask with that file descrip‐
30 tor. If fd is not -1, then it must specify a valid existing signalfd
31 file descriptor, and mask is used to replace the signal set associated
32 with that file descriptor.
33 Starting with Linux 2.6.27, the following values may be bitwise ORed in
35 flags to change the behavior of signalfd():
36 SFD_NONBLOCK Set the O_NONBLOCK file status flag on the open file de‐
38 scription (see open(2)) referred to by the new file de‐
39 scriptor. Using this flag saves extra calls to fcntl(2)
40 to achieve the same result.
41 SFD_CLOEXEC Set the close-on-exec (FD_CLOEXEC) flag on the new file
43 descriptor. See the description of the O_CLOEXEC flag in
44 open(2) for reasons why this may be useful.
45 In Linux up to version 2.6.26, the flags argument is unused, and must
47 be specified as zero.
48 signalfd() returns a file descriptor that supports the following opera‐
50 tions:
51 read(2)
53 If one or more of the signals specified in mask is pending for
54 the process, then the buffer supplied to read(2) is used to re‐
55 turn one or more signalfd_siginfo structures (see below) that
56 describe the signals. The read(2) returns information for as
57 many signals as are pending and will fit in the supplied buffer.
58 The buffer must be at least sizeof(struct signalfd_siginfo)
59 bytes. The return value of the read(2) is the total number of
60 bytes read.
61 As a consequence of the read(2), the signals are consumed, so
63 that they are no longer pending for the process (i.e., will not
64 be caught by signal handlers, and cannot be accepted using sig‐
65 waitinfo(2)).
66 If none of the signals in mask is pending for the process, then
68 the read(2) either blocks until one of the signals in mask is
69 generated for the process, or fails with the error EAGAIN if the
70 file descriptor has been made nonblocking.
71 poll(2), select(2) (and similar)
73 The file descriptor is readable (the select(2) readfds argument;
74 the poll(2) POLLIN flag) if one or more of the signals in mask
75 is pending for the process.
76 The signalfd file descriptor also supports the other file-de‐
78 scriptor multiplexing APIs: pselect(2), ppoll(2), and epoll(7).
79 close(2)
81 When the file descriptor is no longer required it should be
82 closed. When all file descriptors associated with the same sig‐
83 nalfd object have been closed, the resources for object are
84 freed by the kernel.
85 The signalfd_siginfo structure
87 The format of the signalfd_siginfo structure(s) returned by read(2)s
88 from a signalfd file descriptor is as follows:
89 struct signalfd_siginfo {
91 uint32_t ssi_signo; /* Signal number */
92 int32_t ssi_errno; /* Error number (unused) */
93 int32_t ssi_code; /* Signal code */
94 uint32_t ssi_pid; /* PID of sender */
95 uint32_t ssi_uid; /* Real UID of sender */
96 int32_t ssi_fd; /* File descriptor (SIGIO) */
97 uint32_t ssi_tid; /* Kernel timer ID (POSIX timers)
98 uint32_t ssi_band; /* Band event (SIGIO) */
99 uint32_t ssi_overrun; /* POSIX timer overrun count */
100 uint32_t ssi_trapno; /* Trap number that caused signal */
101 int32_t ssi_status; /* Exit status or signal (SIGCHLD) */
102 int32_t ssi_int; /* Integer sent by sigqueue(3) */
103 uint64_t ssi_ptr; /* Pointer sent by sigqueue(3) */
104 uint64_t ssi_utime; /* User CPU time consumed (SIGCHLD) */
105 uint64_t ssi_stime; /* System CPU time consumed
106 (SIGCHLD) */
107 uint64_t ssi_addr; /* Address that generated signal
108 (for hardware-generated signals) */
109 uint16_t ssi_addr_lsb; /* Least significant bit of address
110 (SIGBUS; since Linux 2.6.37) */
111 uint8_t pad[X]; /* Pad size to 128 bytes (allow for
112 additional fields in the future) */
113 };
114 Each of the fields in this structure is analogous to the similarly
116 named field in the siginfo_t structure. The siginfo_t structure is de‐
117 scribed in sigaction(2). Not all fields in the returned signalfd_sig‐
118 info structure will be valid for a specific signal; the set of valid
119 fields can be determined from the value returned in the ssi_code field.
120 This field is the analog of the siginfo_t si_code field; see sigac‐
121 tion(2) for details.
122 fork(2) semantics
124 After a fork(2), the child inherits a copy of the signalfd file de‐
125 scriptor. A read(2) from the file descriptor in the child will return
126 information about signals queued to the child.
127 Semantics of file descriptor passing
129 As with other file descriptors, signalfd file descriptors can be passed
130 to another process via a UNIX domain socket (see unix(7)). In the re‐
131 ceiving process, a read(2) from the received file descriptor will re‐
132 turn information about signals queued to that process.
133 execve(2) semantics
135 Just like any other file descriptor, a signalfd file descriptor remains
136 open across an execve(2), unless it has been marked for close-on-exec
137 (see fcntl(2)). Any signals that were available for reading before the
138 execve(2) remain available to the newly loaded program. (This is anal‐
139 ogous to traditional signal semantics, where a blocked signal that is
140 pending remains pending across an execve(2).)
141 Thread semantics
143 The semantics of signalfd file descriptors in a multithreaded program
144 mirror the standard semantics for signals. In other words, when a
145 thread reads from a signalfd file descriptor, it will read the signals
146 that are directed to the thread itself and the signals that are di‐
147 rected to the process (i.e., the entire thread group). (A thread will
148 not be able to read signals that are directed to other threads in the
149 process.)
150 epoll(7) semantics
152 If a process adds (via epoll_ctl(2)) a signalfd file descriptor to an
153 epoll(7) instance, then epoll_wait(2) returns events only for signals
154 sent to that process. In particular, if the process then uses fork(2)
155 to create a child process, then the child will be able to read(2) sig‐
156 nals that are sent to it using the signalfd file descriptor, but
157 epoll_wait(2) will not indicate that the signalfd file descriptor is
158 ready. In this scenario, a possible workaround is that after the
159 fork(2), the child process can close the signalfd file descriptor that
160 it inherited from the parent process and then create another signalfd
161 file descriptor and add it to the epoll instance. Alternatively, the
162 parent and the child could delay creating their (separate) signalfd
163 file descriptors and adding them to the epoll instance until after the
164 call to fork(2).
165
167 On success, signalfd() returns a signalfd file descriptor; this is ei‐
168 ther a new file descriptor (if fd was -1), or fd if fd was a valid sig‐
169 nalfd file descriptor. On error, -1 is returned and errno is set to
170 indicate the error.
171
173 EBADF The fd file descriptor is not a valid file descriptor.
174 EINVAL fd is not a valid signalfd file descriptor.
176 EINVAL flags is invalid; or, in Linux 2.6.26 or earlier, flags is non‐
178 zero.
179 EMFILE The per-process limit on the number of open file descriptors has
181 been reached.
182 ENFILE The system-wide limit on the total number of open files has been
184 reached.
185 ENODEV Could not mount (internal) anonymous inode device.
187 ENOMEM There was insufficient memory to create a new signalfd file de‐
189 scriptor.
190
192 signalfd() is available on Linux since kernel 2.6.22. Working support
193 is provided in glibc since version 2.8. The signalfd4() system call
194 (see NOTES) is available on Linux since kernel 2.6.27.
195
197 signalfd() and signalfd4() are Linux-specific.
198
200 A process can create multiple signalfd file descriptors. This makes it
201 possible to accept different signals on different file descriptors.
202 (This may be useful if monitoring the file descriptors using select(2),
203 poll(2), or epoll(7): the arrival of different signals will make dif‐
204 ferent file descriptors ready.) If a signal appears in the mask of
205 more than one of the file descriptors, then occurrences of that signal
206 can be read (once) from any one of the file descriptors.
207 Attempts to include SIGKILL and SIGSTOP in mask are silently ignored.
209 The signal mask employed by a signalfd file descriptor can be viewed
211 via the entry for the corresponding file descriptor in the process's
212 /proc/[pid]/fdinfo directory. See proc(5) for further details.
213 Limitations
215 The signalfd mechanism can't be used to receive signals that are syn‐
216 chronously generated, such as the SIGSEGV signal that results from ac‐
217 cessing an invalid memory address or the SIGFPE signal that results
218 from an arithmetic error. Such signals can be caught only via signal
219 handler.
220 As described above, in normal usage one blocks the signals that will be
222 accepted via signalfd(). If spawning a child process to execute a
223 helper program (that does not need the signalfd file descriptor), then,
224 after the call to fork(2), you will normally want to unblock those sig‐
225 nals before calling execve(2), so that the helper program can see any
226 signals that it expects to see. Be aware, however, that this won't be
227 possible in the case of a helper program spawned behind the scenes by
228 any library function that the program may call. In such cases, one
229 must fall back to using a traditional signal handler that writes to a
230 file descriptor monitored by select(2), poll(2), or epoll(7).
231 C library/kernel differences
233 The underlying Linux system call requires an additional argument,
234 size_t sizemask, which specifies the size of the mask argument. The
235 glibc signalfd() wrapper function does not include this argument, since
236 it provides the required value for the underlying system call.
237 There are two underlying Linux system calls: signalfd() and the more
239 recent signalfd4(). The former system call does not implement a flags
240 argument. The latter system call implements the flags values described
241 above. Starting with glibc 2.9, the signalfd() wrapper function will
242 use signalfd4() where it is available.
243
245 In kernels before 2.6.25, the ssi_ptr and ssi_int fields are not filled
246 in with the data accompanying a signal sent by sigqueue(3).
247
249 The program below accepts the signals SIGINT and SIGQUIT via a signalfd
250 file descriptor. The program terminates after accepting a SIGQUIT sig‐
251 nal. The following shell session demonstrates the use of the program:
252 $ ./signalfd_demo
254 ^C # Control-C generates SIGINT
255 Got SIGINT
256 ^C
257 Got SIGINT
258 ^\ # Control-\ generates SIGQUIT
259 Got SIGQUIT
260 $
261 Program source
263 #include <sys/signalfd.h>
265 #include <signal.h>
266 #include <unistd.h>
267 #include <stdlib.h>
268 #include <stdio.h>
269 #define handle_error(msg) \
271 do { perror(msg); exit(EXIT_FAILURE); } while (0)
272 int
274 main(int argc, char *argv[])
275 {
276 sigset_t mask;
277 int sfd;
278 struct signalfd_siginfo fdsi;
279 ssize_t s;
280 sigemptyset(&mask);
282 sigaddset(&mask, SIGINT);
283 sigaddset(&mask, SIGQUIT);
284 /* Block signals so that they aren't handled
286 according to their default dispositions. */
287 if (sigprocmask(SIG_BLOCK, &mask, NULL) == -1)
289 handle_error("sigprocmask");
290 sfd = signalfd(-1, &mask, 0);
292 if (sfd == -1)
293 handle_error("signalfd");
294 for (;;) {
296 s = read(sfd, &fdsi, sizeof(fdsi));
297 if (s != sizeof(fdsi))
298 handle_error("read");
299 if (fdsi.ssi_signo == SIGINT) {
301 printf("Got SIGINT\n");
302 } else if (fdsi.ssi_signo == SIGQUIT) {
303 printf("Got SIGQUIT\n");
304 exit(EXIT_SUCCESS);
305 } else {
306 printf("Read unexpected signal\n");
307 }
308 }
309 }
310
312 eventfd(2), poll(2), read(2), select(2), sigaction(2), sigprocmask(2),
313 sigwaitinfo(2), timerfd_create(2), sigsetops(3), sigwait(3), epoll(7),
314 signal(7)
315
317 This page is part of release 5.13 of the Linux man-pages project. A
318 description of the project, information about reporting bugs, and the
319 latest version of this page, can be found at
320 https://www.kernel.org/doc/man-pages/.
321Linux 2021-03-22 SIGNALFD(2)