1EVENTFD(2) Linux Programmer's Manual EVENTFD(2)
2
6 eventfd - create a file descriptor for event notification
7
9 #include <sys/eventfd.h>
10 int eventfd(unsigned int initval, int flags);
12
14 eventfd() creates an "eventfd object" that can be used as an event
15 wait/notify mechanism by user-space applications, and by the kernel to
16 notify user-space applications of events. The object contains an un‐
17 signed 64-bit integer (uint64_t) counter that is maintained by the ker‐
18 nel. This counter is initialized with the value specified in the argu‐
19 ment initval.
20 As its return value, eventfd() returns a new file descriptor that can
22 be used to refer to the eventfd object.
23 The following values may be bitwise ORed in flags to change the behav‐
25 ior of eventfd():
26 EFD_CLOEXEC (since Linux 2.6.27)
28 Set the close-on-exec (FD_CLOEXEC) flag on the new file descrip‐
29 tor. See the description of the O_CLOEXEC flag in open(2) for
30 reasons why this may be useful.
31 EFD_NONBLOCK (since Linux 2.6.27)
33 Set the O_NONBLOCK file status flag on the open file description
34 (see open(2)) referred to by the new file descriptor. Using
35 this flag saves extra calls to fcntl(2) to achieve the same re‐
36 sult.
37 EFD_SEMAPHORE (since Linux 2.6.30)
39 Provide semaphore-like semantics for reads from the new file de‐
40 scriptor. See below.
41 In Linux up to version 2.6.26, the flags argument is unused, and must
43 be specified as zero.
44 The following operations can be performed on the file descriptor re‐
46 turned by eventfd():
47 read(2)
49 Each successful read(2) returns an 8-byte integer. A read(2)
50 fails with the error EINVAL if the size of the supplied buffer
51 is less than 8 bytes.
52 The value returned by read(2) is in host byte order—that is, the
54 native byte order for integers on the host machine.
55 The semantics of read(2) depend on whether the eventfd counter
57 currently has a nonzero value and whether the EFD_SEMAPHORE flag
58 was specified when creating the eventfd file descriptor:
59 * If EFD_SEMAPHORE was not specified and the eventfd counter
61 has a nonzero value, then a read(2) returns 8 bytes contain‐
62 ing that value, and the counter's value is reset to zero.
63 * If EFD_SEMAPHORE was specified and the eventfd counter has a
65 nonzero value, then a read(2) returns 8 bytes containing the
66 value 1, and the counter's value is decremented by 1.
67 * If the eventfd counter is zero at the time of the call to
69 read(2), then the call either blocks until the counter be‐
70 comes nonzero (at which time, the read(2) proceeds as de‐
71 scribed above) or fails with the error EAGAIN if the file de‐
72 scriptor has been made nonblocking.
73 write(2)
75 A write(2) call adds the 8-byte integer value supplied in its
76 buffer to the counter. The maximum value that may be stored in
77 the counter is the largest unsigned 64-bit value minus 1 (i.e.,
78 0xfffffffffffffffe). If the addition would cause the counter's
79 value to exceed the maximum, then the write(2) either blocks un‐
80 til a read(2) is performed on the file descriptor, or fails with
81 the error EAGAIN if the file descriptor has been made nonblock‐
82 ing.
83 A write(2) fails with the error EINVAL if the size of the sup‐
85 plied buffer is less than 8 bytes, or if an attempt is made to
86 write the value 0xffffffffffffffff.
87 poll(2), select(2) (and similar)
89 The returned file descriptor supports poll(2) (and analogously
90 epoll(7)) and select(2), as follows:
91 * The file descriptor is readable (the select(2) readfds argu‐
93 ment; the poll(2) POLLIN flag) if the counter has a value
94 greater than 0.
95 * The file descriptor is writable (the select(2) writefds argu‐
97 ment; the poll(2) POLLOUT flag) if it is possible to write a
98 value of at least "1" without blocking.
99 * If an overflow of the counter value was detected, then se‐
101 lect(2) indicates the file descriptor as being both readable
102 and writable, and poll(2) returns a POLLERR event. As noted
103 above, write(2) can never overflow the counter. However an
104 overflow can occur if 2^64 eventfd "signal posts" were per‐
105 formed by the KAIO subsystem (theoretically possible, but
106 practically unlikely). If an overflow has occurred, then
107 read(2) will return that maximum uint64_t value (i.e.,
108 0xffffffffffffffff).
109 The eventfd file descriptor also supports the other file-de‐
111 scriptor multiplexing APIs: pselect(2) and ppoll(2).
112 close(2)
114 When the file descriptor is no longer required it should be
115 closed. When all file descriptors associated with the same
116 eventfd object have been closed, the resources for object are
117 freed by the kernel.
118 A copy of the file descriptor created by eventfd() is inherited by the
120 child produced by fork(2). The duplicate file descriptor is associated
121 with the same eventfd object. File descriptors created by eventfd()
122 are preserved across execve(2), unless the close-on-exec flag has been
123 set.
124
126 On success, eventfd() returns a new eventfd file descriptor. On error,
127 -1 is returned and errno is set to indicate the error.
128
130 EINVAL An unsupported value was specified in flags.
131 EMFILE The per-process limit on the number of open file descriptors has
133 been reached.
134 ENFILE The system-wide limit on the total number of open files has been
136 reached.
137 ENODEV Could not mount (internal) anonymous inode device.
139 ENOMEM There was insufficient memory to create a new eventfd file de‐
141 scriptor.
142
144 eventfd() is available on Linux since kernel 2.6.22. Working support
145 is provided in glibc since version 2.8. The eventfd2() system call
146 (see NOTES) is available on Linux since kernel 2.6.27. Since version
147 2.9, the glibc eventfd() wrapper will employ the eventfd2() system
148 call, if it is supported by the kernel.
149
151 For an explanation of the terms used in this section, see at‐
152 tributes(7).
153 ┌──────────┬───────────────┬─────────┐
155 │Interface │ Attribute │ Value │
156 ├──────────┼───────────────┼─────────┤
157 │eventfd() │ Thread safety │ MT-Safe │
158 └──────────┴───────────────┴─────────┘
159
161 eventfd() and eventfd2() are Linux-specific.
162
164 Applications can use an eventfd file descriptor instead of a pipe (see
165 pipe(2)) in all cases where a pipe is used simply to signal events.
166 The kernel overhead of an eventfd file descriptor is much lower than
167 that of a pipe, and only one file descriptor is required (versus the
168 two required for a pipe).
169 When used in the kernel, an eventfd file descriptor can provide a
171 bridge from kernel to user space, allowing, for example, functionali‐
172 ties like KAIO (kernel AIO) to signal to a file descriptor that some
173 operation is complete.
174 A key point about an eventfd file descriptor is that it can be moni‐
176 tored just like any other file descriptor using select(2), poll(2), or
177 epoll(7). This means that an application can simultaneously monitor
178 the readiness of "traditional" files and the readiness of other kernel
179 mechanisms that support the eventfd interface. (Without the eventfd()
180 interface, these mechanisms could not be multiplexed via select(2),
181 poll(2), or epoll(7).)
182 The current value of an eventfd counter can be viewed via the entry for
184 the corresponding file descriptor in the process's /proc/[pid]/fdinfo
185 directory. See proc(5) for further details.
186 C library/kernel differences
188 There are two underlying Linux system calls: eventfd() and the more re‐
189 cent eventfd2(). The former system call does not implement a flags ar‐
190 gument. The latter system call implements the flags values described
191 above. The glibc wrapper function will use eventfd2() where it is
192 available.
193 Additional glibc features
195 The GNU C library defines an additional type, and two functions that
196 attempt to abstract some of the details of reading and writing on an
197 eventfd file descriptor:
198 typedef uint64_t eventfd_t;
200 int eventfd_read(int fd, eventfd_t *value);
202 int eventfd_write(int fd, eventfd_t value);
203 The functions perform the read and write operations on an eventfd file
205 descriptor, returning 0 if the correct number of bytes was transferred,
206 or -1 otherwise.
207
209 The following program creates an eventfd file descriptor and then forks
210 to create a child process. While the parent briefly sleeps, the child
211 writes each of the integers supplied in the program's command-line ar‐
212 guments to the eventfd file descriptor. When the parent has finished
213 sleeping, it reads from the eventfd file descriptor.
214 The following shell session shows a sample run of the program:
216 $ ./a.out 1 2 4 7 14
218 Child writing 1 to efd
219 Child writing 2 to efd
220 Child writing 4 to efd
221 Child writing 7 to efd
222 Child writing 14 to efd
223 Child completed write loop
224 Parent about to read
225 Parent read 28 (0x1c) from efd
226 Program source
228 #include <sys/eventfd.h>
230 #include <unistd.h>
231 #include <inttypes.h> /* Definition of PRIu64 & PRIx64 */
232 #include <stdlib.h>
233 #include <stdio.h>
234 #include <stdint.h> /* Definition of uint64_t */
235 #define handle_error(msg) \
237 do { perror(msg); exit(EXIT_FAILURE); } while (0)
238 int
240 main(int argc, char *argv[])
241 {
242 int efd;
243 uint64_t u;
244 ssize_t s;
245 if (argc < 2) {
247 fprintf(stderr, "Usage: %s <num>...\n", argv[0]);
248 exit(EXIT_FAILURE);
249 }
250 efd = eventfd(0, 0);
252 if (efd == -1)
253 handle_error("eventfd");
254 switch (fork()) {
256 case 0:
257 for (int j = 1; j < argc; j++) {
258 printf("Child writing %s to efd\n", argv[j]);
259 u = strtoull(argv[j], NULL, 0);
260 /* strtoull() allows various bases */
261 s = write(efd, &u, sizeof(uint64_t));
262 if (s != sizeof(uint64_t))
263 handle_error("write");
264 }
265 printf("Child completed write loop\n");
266 exit(EXIT_SUCCESS);
268 default:
270 sleep(2);
271 printf("Parent about to read\n");
273 s = read(efd, &u, sizeof(uint64_t));
274 if (s != sizeof(uint64_t))
275 handle_error("read");
276 printf("Parent read %"PRIu64" (%#"PRIx64") from efd\n", u, u);
277 exit(EXIT_SUCCESS);
278 case -1:
280 handle_error("fork");
281 }
282 }
283
285 futex(2), pipe(2), poll(2), read(2), select(2), signalfd(2),
286 timerfd_create(2), write(2), epoll(7), sem_overview(7)
287
289 This page is part of release 5.10 of the Linux man-pages project. A
290 description of the project, information about reporting bugs, and the
291 latest version of this page, can be found at
292 https://www.kernel.org/doc/man-pages/.
293Linux 2020-11-01 EVENTFD(2)