1 .\" Copyright (C) 2008 Michael Kerrisk <mtk.manpages@gmail.com>
2 .\" starting from a version by Davide Libenzi <davidel@xmailserver.org>
4 .\" %%%LICENSE_START(GPLv2+_SW_3_PARA)
5 .\" This program is free software; you can redistribute it and/or modify
6 .\" it under the terms of the GNU General Public License as published by
7 .\" the Free Software Foundation; either version 2 of the License, or
8 .\" (at your option) any later version.
10 .\" This program is distributed in the hope that it will be useful,
11 .\" but WITHOUT ANY WARRANTY; without even the implied warranty of
12 .\" MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 .\" GNU General Public License for more details.
15 .\" You should have received a copy of the GNU General Public
16 .\" License along with this manual; if not, see
17 .\" <http://www.gnu.org/licenses/>.
20 .TH SIGNALFD 2 2021-03-22 Linux "Linux Programmer's Manual"
22 signalfd \- create a file descriptor for accepting signals
25 .B #include <sys/signalfd.h>
27 .BI "int signalfd(int " fd ", const sigset_t *" mask ", int " flags );
31 creates a file descriptor that can be used to accept signals
32 targeted at the caller.
33 This provides an alternative to the use of a signal handler or
35 and has the advantage that the file descriptor may be monitored by
43 argument specifies the set of signals that the caller
44 wishes to accept via the file descriptor.
45 This argument is a signal set whose contents can be initialized
46 using the macros described in
48 Normally, the set of signals to be received via the
49 file descriptor should be blocked using
51 to prevent the signals being handled according to their default
53 It is not possible to receive
57 signals via a signalfd file descriptor;
58 these signals are silently ignored if specified in
64 then the call creates a new file descriptor and associates the
65 signal set specified in
67 with that file descriptor.
71 then it must specify a valid existing signalfd file descriptor, and
73 is used to replace the signal set associated with that file descriptor.
75 Starting with Linux 2.6.27, the following values may be bitwise ORed in
77 to change the behavior of
83 file status flag on the open file description (see
85 referred to by the new file descriptor.
86 Using this flag saves extra calls to
88 to achieve the same result.
93 flag on the new file descriptor.
94 See the description of the
98 for reasons why this may be useful.
100 In Linux up to version 2.6.26, the
102 argument is unused, and must be specified as zero.
105 returns a file descriptor that supports the following operations:
108 If one or more of the signals specified in
110 is pending for the process, then the buffer supplied to
112 is used to return one or more
114 structures (see below) that describe the signals.
117 returns information for as many signals as are pending and will
118 fit in the supplied buffer.
119 The buffer must be at least
120 .I "sizeof(struct signalfd_siginfo)"
122 The return value of the
124 is the total number of bytes read.
126 As a consequence of the
128 the signals are consumed,
129 so that they are no longer pending for the process
130 (i.e., will not be caught by signal handlers,
131 and cannot be accepted using
132 .BR sigwaitinfo (2)).
134 If none of the signals in
136 is pending for the process, then the
138 either blocks until one of the signals in
140 is generated for the process,
141 or fails with the error
143 if the file descriptor has been made nonblocking.
145 .BR poll "(2), " select "(2) (and similar)"
146 The file descriptor is readable
154 if one or more of the signals in
156 is pending for the process.
158 The signalfd file descriptor also supports the other file-descriptor
166 When the file descriptor is no longer required it should be closed.
167 When all file descriptors associated with the same signalfd object
168 have been closed, the resources for object are freed by the kernel.
169 .SS The signalfd_siginfo structure
172 structure(s) returned by
174 from a signalfd file descriptor is as follows:
178 struct signalfd_siginfo {
179 uint32_t ssi_signo; /* Signal number */
180 int32_t ssi_errno; /* Error number (unused) */
181 int32_t ssi_code; /* Signal code */
182 uint32_t ssi_pid; /* PID of sender */
183 uint32_t ssi_uid; /* Real UID of sender */
184 int32_t ssi_fd; /* File descriptor (SIGIO) */
185 uint32_t ssi_tid; /* Kernel timer ID (POSIX timers)
186 uint32_t ssi_band; /* Band event (SIGIO) */
187 uint32_t ssi_overrun; /* POSIX timer overrun count */
188 uint32_t ssi_trapno; /* Trap number that caused signal */
189 .\" ssi_trapno is unused on most arches
190 int32_t ssi_status; /* Exit status or signal (SIGCHLD) */
191 int32_t ssi_int; /* Integer sent by sigqueue(3) */
192 uint64_t ssi_ptr; /* Pointer sent by sigqueue(3) */
193 uint64_t ssi_utime; /* User CPU time consumed (SIGCHLD) */
194 uint64_t ssi_stime; /* System CPU time consumed
196 uint64_t ssi_addr; /* Address that generated signal
197 (for hardware\-generated signals) */
198 uint16_t ssi_addr_lsb; /* Least significant bit of address
199 (SIGBUS; since Linux 2.6.37) */
200 .\" ssi_addr_lsb: commit b8aeec34175fc8fe8b0d40efea4846dfc1ba663e
201 uint8_t pad[\fIX\fP]; /* Pad size to 128 bytes (allow for
202 additional fields in the future) */
207 Each of the fields in this structure
208 is analogous to the similarly named field in the
213 structure is described in
215 Not all fields in the returned
217 structure will be valid for a specific signal;
218 the set of valid fields can be determined from the value returned in the
221 This field is the analog of the
227 .SS fork(2) semantics
230 the child inherits a copy of the signalfd file descriptor.
233 from the file descriptor in the child will return information
234 about signals queued to the child.
235 .SS Semantics of file descriptor passing
236 As with other file descriptors,
237 signalfd file descriptors can be passed to another process
238 via a UNIX domain socket (see
240 In the receiving process, a
242 from the received file descriptor will return information
243 about signals queued to that process.
244 .SS execve(2) semantics
245 Just like any other file descriptor,
246 a signalfd file descriptor remains open across an
248 unless it has been marked for close-on-exec (see
250 Any signals that were available for reading before the
252 remain available to the newly loaded program.
253 (This is analogous to traditional signal semantics,
254 where a blocked signal that is pending remains pending across an
257 The semantics of signalfd file descriptors in a multithreaded program
258 mirror the standard semantics for signals.
260 when a thread reads from a signalfd file descriptor,
261 it will read the signals that are directed to the thread
262 itself and the signals that are directed to the process
263 (i.e., the entire thread group).
264 (A thread will not be able to read signals that are directed
265 to other threads in the process.)
267 .SS epoll(7) semantics
268 If a process adds (via
270 a signalfd file descriptor to an
274 returns events only for signals sent to that process.
275 In particular, if the process then uses
277 to create a child process, then the child will be able to
279 signals that are sent to it using the signalfd file descriptor, but
283 indicate that the signalfd file descriptor is ready.
284 In this scenario, a possible workaround is that after the
286 the child process can close the signalfd file descriptor that it inherited
287 from the parent process and then create another signalfd file descriptor
288 and add it to the epoll instance.
289 Alternatively, the parent and the child could delay creating their
290 (separate) signalfd file descriptors and adding them to the
291 epoll instance until after the call to
296 returns a signalfd file descriptor;
297 this is either a new file descriptor (if
303 was a valid signalfd file descriptor.
304 On error, \-1 is returned and
306 is set to indicate the error.
312 file descriptor is not a valid file descriptor.
316 is not a valid signalfd file descriptor.
319 .\" argument is not equal to
320 .\" .IR sizeof(sigset_t) ;
325 or, in Linux 2.6.26 or earlier,
330 The per-process limit on the number of open file descriptors has been reached.
333 The system-wide limit on the total number of open files has been
337 Could not mount (internal) anonymous inode device.
340 There was insufficient memory to create a new signalfd file descriptor.
343 is available on Linux since kernel 2.6.22.
344 Working support is provided in glibc since version 2.8.
345 .\" signalfd() is in glibc 2.7, but reportedly does not build
348 system call (see NOTES) is available on Linux since kernel 2.6.27.
355 A process can create multiple signalfd file descriptors.
356 This makes it possible to accept different signals
357 on different file descriptors.
358 (This may be useful if monitoring the file descriptors using
363 the arrival of different signals will make different file descriptors ready.)
364 If a signal appears in the
366 of more than one of the file descriptors, then occurrences
367 of that signal can be read (once) from any one of the file descriptors.
375 are silently ignored.
377 The signal mask employed by a signalfd file descriptor can be viewed
378 via the entry for the corresponding file descriptor in the process's
379 .IR /proc/[pid]/fdinfo
386 The signalfd mechanism can't be used to receive signals that
387 are synchronously generated, such as the
389 signal that results from accessing an invalid memory address
392 signal that results from an arithmetic error.
393 Such signals can be caught only via signal handler.
396 in normal usage one blocks the signals that will be accepted via
398 If spawning a child process to execute a helper program
399 (that does not need the signalfd file descriptor),
400 then, after the call to
402 you will normally want to unblock those signals before calling
404 so that the helper program can see any signals that it expects to see.
406 that this won't be possible in the case of a helper program spawned
407 behind the scenes by any library function that the program may call.
408 In such cases, one must fall back to using a traditional signal
409 handler that writes to a file descriptor monitored by
415 .SS C library/kernel differences
416 The underlying Linux system call requires an additional argument,
417 .IR "size_t sizemask" ,
418 which specifies the size of the
423 wrapper function does not include this argument,
424 since it provides the required value for the underlying system call.
426 There are two underlying Linux system calls:
430 The former system call does not implement a
433 The latter system call implements the
435 values described above.
436 Starting with glibc 2.9, the
438 wrapper function will use
440 where it is available.
442 In kernels before 2.6.25, the
446 fields are not filled in with the data accompanying a signal sent by
448 .\" The fix also was put into 2.6.24.5
450 The program below accepts the signals
454 via a signalfd file descriptor.
455 The program terminates after accepting a
458 The following shell session demonstrates the use of the program:
462 .RB "$" " ./signalfd_demo"
463 .BR "\(haC" " # Control\-C generates SIGINT"
467 \fB\(ha\e\fP # Control\-\e generates SIGQUIT
475 #include <sys/signalfd.h>
481 #define handle_error(msg) \e
482 do { perror(msg); exit(EXIT_FAILURE); } while (0)
485 main(int argc, char *argv[])
489 struct signalfd_siginfo fdsi;
493 sigaddset(&mask, SIGINT);
494 sigaddset(&mask, SIGQUIT);
496 /* Block signals so that they aren\(aqt handled
497 according to their default dispositions. */
499 if (sigprocmask(SIG_BLOCK, &mask, NULL) == \-1)
500 handle_error("sigprocmask");
502 sfd = signalfd(\-1, &mask, 0);
504 handle_error("signalfd");
507 s = read(sfd, &fdsi, sizeof(fdsi));
508 if (s != sizeof(fdsi))
509 handle_error("read");
511 if (fdsi.ssi_signo == SIGINT) {
512 printf("Got SIGINT\en");
513 } else if (fdsi.ssi_signo == SIGQUIT) {
514 printf("Got SIGQUIT\en");
517 printf("Read unexpected signal\en");
530 .BR timerfd_create (2),