1eventfd(2) System Calls Manual eventfd(2)
2
6 eventfd - create a file descriptor for event notification
7
9 Standard C library (libc, -lc)
10
12 #include <sys/eventfd.h>
13 int eventfd(unsigned int initval, int flags);
15
17 eventfd() creates an "eventfd object" that can be used as an event
18 wait/notify mechanism by user-space applications, and by the kernel to
19 notify user-space applications of events. The object contains an un‐
20 signed 64-bit integer (uint64_t) counter that is maintained by the ker‐
21 nel. This counter is initialized with the value specified in the argu‐
22 ment initval.
23 As its return value, eventfd() returns a new file descriptor that can
25 be used to refer to the eventfd object.
26 The following values may be bitwise ORed in flags to change the behav‐
28 ior of eventfd():
29 EFD_CLOEXEC (since Linux 2.6.27)
31 Set the close-on-exec (FD_CLOEXEC) flag on the new file descrip‐
32 tor. See the description of the O_CLOEXEC flag in open(2) for
33 reasons why this may be useful.
34 EFD_NONBLOCK (since Linux 2.6.27)
36 Set the O_NONBLOCK file status flag on the open file description
37 (see open(2)) referred to by the new file descriptor. Using
38 this flag saves extra calls to fcntl(2) to achieve the same re‐
39 sult.
40 EFD_SEMAPHORE (since Linux 2.6.30)
42 Provide semaphore-like semantics for reads from the new file de‐
43 scriptor. See below.
44 Up to Linux 2.6.26, the flags argument is unused, and must be specified
46 as zero.
47 The following operations can be performed on the file descriptor re‐
49 turned by eventfd():
50 read(2)
52 Each successful read(2) returns an 8-byte integer. A read(2)
53 fails with the error EINVAL if the size of the supplied buffer
54 is less than 8 bytes.
55 The value returned by read(2) is in host byte order—that is, the
57 native byte order for integers on the host machine.
58 The semantics of read(2) depend on whether the eventfd counter
60 currently has a nonzero value and whether the EFD_SEMAPHORE flag
61 was specified when creating the eventfd file descriptor:
62 • If EFD_SEMAPHORE was not specified and the eventfd counter
64 has a nonzero value, then a read(2) returns 8 bytes contain‐
65 ing that value, and the counter's value is reset to zero.
66 • If EFD_SEMAPHORE was specified and the eventfd counter has a
68 nonzero value, then a read(2) returns 8 bytes containing the
69 value 1, and the counter's value is decremented by 1.
70 • If the eventfd counter is zero at the time of the call to
72 read(2), then the call either blocks until the counter be‐
73 comes nonzero (at which time, the read(2) proceeds as de‐
74 scribed above) or fails with the error EAGAIN if the file de‐
75 scriptor has been made nonblocking.
76 write(2)
78 A write(2) call adds the 8-byte integer value supplied in its
79 buffer to the counter. The maximum value that may be stored in
80 the counter is the largest unsigned 64-bit value minus 1 (i.e.,
81 0xfffffffffffffffe). If the addition would cause the counter's
82 value to exceed the maximum, then the write(2) either blocks un‐
83 til a read(2) is performed on the file descriptor, or fails with
84 the error EAGAIN if the file descriptor has been made nonblock‐
85 ing.
86 A write(2) fails with the error EINVAL if the size of the sup‐
88 plied buffer is less than 8 bytes, or if an attempt is made to
89 write the value 0xffffffffffffffff.
90 poll(2), select(2) (and similar)
92 The returned file descriptor supports poll(2) (and analogously
93 epoll(7)) and select(2), as follows:
94 • The file descriptor is readable (the select(2) readfds argu‐
96 ment; the poll(2) POLLIN flag) if the counter has a value
97 greater than 0.
98 • The file descriptor is writable (the select(2) writefds argu‐
100 ment; the poll(2) POLLOUT flag) if it is possible to write a
101 value of at least "1" without blocking.
102 • If an overflow of the counter value was detected, then se‐
104 lect(2) indicates the file descriptor as being both readable
105 and writable, and poll(2) returns a POLLERR event. As noted
106 above, write(2) can never overflow the counter. However an
107 overflow can occur if 2^64 eventfd "signal posts" were per‐
108 formed by the KAIO subsystem (theoretically possible, but
109 practically unlikely). If an overflow has occurred, then
110 read(2) will return that maximum uint64_t value (i.e.,
111 0xffffffffffffffff).
112 The eventfd file descriptor also supports the other file-de‐
114 scriptor multiplexing APIs: pselect(2) and ppoll(2).
115 close(2)
117 When the file descriptor is no longer required it should be
118 closed. When all file descriptors associated with the same
119 eventfd object have been closed, the resources for object are
120 freed by the kernel.
121 A copy of the file descriptor created by eventfd() is inherited by the
123 child produced by fork(2). The duplicate file descriptor is associated
124 with the same eventfd object. File descriptors created by eventfd()
125 are preserved across execve(2), unless the close-on-exec flag has been
126 set.
127
129 On success, eventfd() returns a new eventfd file descriptor. On error,
130 -1 is returned and errno is set to indicate the error.
131
133 EINVAL An unsupported value was specified in flags.
134 EMFILE The per-process limit on the number of open file descriptors has
136 been reached.
137 ENFILE The system-wide limit on the total number of open files has been
139 reached.
140 ENODEV Could not mount (internal) anonymous inode device.
142 ENOMEM There was insufficient memory to create a new eventfd file de‐
144 scriptor.
145
147 For an explanation of the terms used in this section, see at‐
148 tributes(7).
149 ┌────────────────────────────────────────────┬───────────────┬─────────┐
151 │Interface │ Attribute │ Value │
152 ├────────────────────────────────────────────┼───────────────┼─────────┤
153 │eventfd() │ Thread safety │ MT-Safe │
154 └────────────────────────────────────────────┴───────────────┴─────────┘
155
157 C library/kernel differences
158 There are two underlying Linux system calls: eventfd() and the more
159 recent eventfd2(). The former system call does not implement a flags
160 argument. The latter system call implements the flags values described
161 above. The glibc wrapper function will use eventfd2() where it is
162 available.
163 Additional glibc features
165 The GNU C library defines an additional type, and two functions that
166 attempt to abstract some of the details of reading and writing on an
167 eventfd file descriptor:
168 typedef uint64_t eventfd_t;
170 int eventfd_read(int fd, eventfd_t *value);
172 int eventfd_write(int fd, eventfd_t value);
173 The functions perform the read and write operations on an eventfd file
175 descriptor, returning 0 if the correct number of bytes was transferred,
176 or -1 otherwise.
177
179 Linux, GNU.
180
182 eventfd()
183 Linux 2.6.22, glibc 2.8.
184 eventfd2()
186 Linux 2.6.27 (see VERSIONS). Since glibc 2.9, the eventfd()
187 wrapper will employ the eventfd2() system call, if it is sup‐
188 ported by the kernel.
189
191 Applications can use an eventfd file descriptor instead of a pipe (see
192 pipe(2)) in all cases where a pipe is used simply to signal events.
193 The kernel overhead of an eventfd file descriptor is much lower than
194 that of a pipe, and only one file descriptor is required (versus the
195 two required for a pipe).
196 When used in the kernel, an eventfd file descriptor can provide a
198 bridge from kernel to user space, allowing, for example, functionali‐
199 ties like KAIO (kernel AIO) to signal to a file descriptor that some
200 operation is complete.
201 A key point about an eventfd file descriptor is that it can be moni‐
203 tored just like any other file descriptor using select(2), poll(2), or
204 epoll(7). This means that an application can simultaneously monitor
205 the readiness of "traditional" files and the readiness of other kernel
206 mechanisms that support the eventfd interface. (Without the eventfd()
207 interface, these mechanisms could not be multiplexed via select(2),
208 poll(2), or epoll(7).)
209 The current value of an eventfd counter can be viewed via the entry for
211 the corresponding file descriptor in the process's /proc/pid/fdinfo di‐
212 rectory. See proc(5) for further details.
213
215 The following program creates an eventfd file descriptor and then forks
216 to create a child process. While the parent briefly sleeps, the child
217 writes each of the integers supplied in the program's command-line ar‐
218 guments to the eventfd file descriptor. When the parent has finished
219 sleeping, it reads from the eventfd file descriptor.
220 The following shell session shows a sample run of the program:
222 $ ./a.out 1 2 4 7 14
224 Child writing 1 to efd
225 Child writing 2 to efd
226 Child writing 4 to efd
227 Child writing 7 to efd
228 Child writing 14 to efd
229 Child completed write loop
230 Parent about to read
231 Parent read 28 (0x1c) from efd
232 Program source
234 #include <err.h>
236 #include <inttypes.h>
237 #include <stdio.h>
238 #include <stdlib.h>
239 #include <sys/eventfd.h>
240 #include <unistd.h>
241 int
243 main(int argc, char *argv[])
244 {
245 int efd;
246 uint64_t u;
247 ssize_t s;
248 if (argc < 2) {
250 fprintf(stderr, "Usage: %s <num>...\n", argv[0]);
251 exit(EXIT_FAILURE);
252 }
253 efd = eventfd(0, 0);
255 if (efd == -1)
256 err(EXIT_FAILURE, "eventfd");
257 switch (fork()) {
259 case 0:
260 for (size_t j = 1; j < argc; j++) {
261 printf("Child writing %s to efd\n", argv[j]);
262 u = strtoull(argv[j], NULL, 0);
263 /* strtoull() allows various bases */
264 s = write(efd, &u, sizeof(uint64_t));
265 if (s != sizeof(uint64_t))
266 err(EXIT_FAILURE, "write");
267 }
268 printf("Child completed write loop\n");
269 exit(EXIT_SUCCESS);
271 default:
273 sleep(2);
274 printf("Parent about to read\n");
276 s = read(efd, &u, sizeof(uint64_t));
277 if (s != sizeof(uint64_t))
278 err(EXIT_FAILURE, "read");
279 printf("Parent read %"PRIu64" (%#"PRIx64") from efd\n", u, u);
280 exit(EXIT_SUCCESS);
281 case -1:
283 err(EXIT_FAILURE, "fork");
284 }
285 }
286
288 futex(2), pipe(2), poll(2), read(2), select(2), signalfd(2),
289 timerfd_create(2), write(2), epoll(7), sem_overview(7)
290Linux man-pages 6.05 2023-07-20 eventfd(2)