1EPOLL(7) Linux Programmer's Manual EPOLL(7)
2
6 epoll - I/O event notification facility
7
9 #include <sys/epoll.h>
10
12 The epoll API performs a similar task to poll(2): monitoring multiple
13 file descriptors to see if I/O is possible on any of them. The epoll
14 API can be used either as an edge-triggered or a level-triggered inter‐
15 face and scales well to large numbers of watched file descriptors. The
16 following system calls are provided to create and manage an epoll
17 instance:
18 * epoll_create(2) creates a new epoll instance and returns a file
20 descriptor referring to that instance. (The more recent epoll_cre‐
21 ate1(2) extends the functionality of epoll_create(2).)
22 * Interest in particular file descriptors is then registered via
24 epoll_ctl(2). The set of file descriptors currently registered on
25 an epoll instance is sometimes called an epoll set.
26 * epoll_wait(2) waits for I/O events, blocking the calling thread if
28 no events are currently available.
29 Level-triggered and edge-triggered
31 The epoll event distribution interface is able to behave both as edge-
32 triggered (ET) and as level-triggered (LT). The difference between the
33 two mechanisms can be described as follows. Suppose that this scenario
34 happens:
35 1. The file descriptor that represents the read side of a pipe (rfd) is
37 registered on the epoll instance.
38 2. A pipe writer writes 2 kB of data on the write side of the pipe.
40 3. A call to epoll_wait(2) is done that will return rfd as a ready file
42 descriptor.
43 4. The pipe reader reads 1 kB of data from rfd.
45 5. A call to epoll_wait(2) is done.
47 If the rfd file descriptor has been added to the epoll interface using
49 the EPOLLET (edge-triggered) flag, the call to epoll_wait(2) done in
50 step 5 will probably hang despite the available data still present in
51 the file input buffer; meanwhile the remote peer might be expecting a
52 response based on the data it already sent. The reason for this is
53 that edge-triggered mode delivers events only when changes occur on the
54 monitored file descriptor. So, in step 5 the caller might end up wait‐
55 ing for some data that is already present inside the input buffer. In
56 the above example, an event on rfd will be generated because of the
57 write done in 2 and the event is consumed in 3. Since the read opera‐
58 tion done in 4 does not consume the whole buffer data, the call to
59 epoll_wait(2) done in step 5 might block indefinitely.
60 An application that employs the EPOLLET flag should use nonblocking
62 file descriptors to avoid having a blocking read or write starve a task
63 that is handling multiple file descriptors. The suggested way to use
64 epoll as an edge-triggered (EPOLLET) interface is as follows:
65 i with nonblocking file descriptors; and
67 ii by waiting for an event only after read(2) or write(2)
69 return EAGAIN.
70 By contrast, when used as a level-triggered interface (the default,
72 when EPOLLET is not specified), epoll is simply a faster poll(2), and
73 can be used wherever the latter is used since it shares the same seman‐
74 tics.
75 Since even with edge-triggered epoll, multiple events can be generated
77 upon receipt of multiple chunks of data, the caller has the option to
78 specify the EPOLLONESHOT flag, to tell epoll to disable the associated
79 file descriptor after the receipt of an event with epoll_wait(2). When
80 the EPOLLONESHOT flag is specified, it is the caller's responsibility
81 to rearm the file descriptor using epoll_ctl(2) with EPOLL_CTL_MOD.
82 Interaction with autosleep
84 If the system is in autosleep mode via /sys/power/autosleep and an
85 event happens which wakes the device from sleep, the device driver will
86 keep the device awake only until that event is queued. To keep the
87 device awake until the event has been processed, it is necessary to use
88 the epoll_ctl(2) EPOLLWAKEUP flag.
89 When the EPOLLWAKEUP flag is set in the events field for a struct
91 epoll_event, the system will be kept awake from the moment the event is
92 queued, through the epoll_wait(2) call which returns the event until
93 the subsequent epoll_wait(2) call. If the event should keep the system
94 awake beyond that time, then a separate wake_lock should be taken
95 before the second epoll_wait(2) call.
96 /proc interfaces
98 The following interfaces can be used to limit the amount of kernel mem‐
99 ory consumed by epoll:
100 /proc/sys/fs/epoll/max_user_watches (since Linux 2.6.28)
102 This specifies a limit on the total number of file descriptors
103 that a user can register across all epoll instances on the sys‐
104 tem. The limit is per real user ID. Each registered file
105 descriptor costs roughly 90 bytes on a 32-bit kernel, and
106 roughly 160 bytes on a 64-bit kernel. Currently, the default
107 value for max_user_watches is 1/25 (4%) of the available low
108 memory, divided by the registration cost in bytes.
109 Example for suggested usage
111 While the usage of epoll when employed as a level-triggered interface
112 does have the same semantics as poll(2), the edge-triggered usage
113 requires more clarification to avoid stalls in the application event
114 loop. In this example, listener is a nonblocking socket on which lis‐
115 ten(2) has been called. The function do_use_fd() uses the new ready
116 file descriptor until EAGAIN is returned by either read(2) or write(2).
117 An event-driven state machine application should, after having received
118 EAGAIN, record its current state so that at the next call to
119 do_use_fd() it will continue to read(2) or write(2) from where it
120 stopped before.
121 #define MAX_EVENTS 10
123 struct epoll_event ev, events[MAX_EVENTS];
124 int listen_sock, conn_sock, nfds, epollfd;
125 /* Code to set up listening socket, 'listen_sock',
127 (socket(), bind(), listen()) omitted */
128 epollfd = epoll_create1(0);
130 if (epollfd == -1) {
131 perror("epoll_create1");
132 exit(EXIT_FAILURE);
133 }
134 ev.events = EPOLLIN;
136 ev.data.fd = listen_sock;
137 if (epoll_ctl(epollfd, EPOLL_CTL_ADD, listen_sock, &ev) == -1) {
138 perror("epoll_ctl: listen_sock");
139 exit(EXIT_FAILURE);
140 }
141 for (;;) {
143 nfds = epoll_wait(epollfd, events, MAX_EVENTS, -1);
144 if (nfds == -1) {
145 perror("epoll_wait");
146 exit(EXIT_FAILURE);
147 }
148 for (n = 0; n < nfds; ++n) {
150 if (events[n].data.fd == listen_sock) {
151 conn_sock = accept(listen_sock,
152 (struct sockaddr *) &addr, &addrlen);
153 if (conn_sock == -1) {
154 perror("accept");
155 exit(EXIT_FAILURE);
156 }
157 setnonblocking(conn_sock);
158 ev.events = EPOLLIN | EPOLLET;
159 ev.data.fd = conn_sock;
160 if (epoll_ctl(epollfd, EPOLL_CTL_ADD, conn_sock,
161 &ev) == -1) {
162 perror("epoll_ctl: conn_sock");
163 exit(EXIT_FAILURE);
164 }
165 } else {
166 do_use_fd(events[n].data.fd);
167 }
168 }
169 }
170 When used as an edge-triggered interface, for performance reasons, it
172 is possible to add the file descriptor inside the epoll interface
173 (EPOLL_CTL_ADD) once by specifying (EPOLLIN|EPOLLOUT). This allows you
174 to avoid continuously switching between EPOLLIN and EPOLLOUT calling
175 epoll_ctl(2) with EPOLL_CTL_MOD.
176 Questions and answers
178 Q0 What is the key used to distinguish the file descriptors registered
179 in an epoll set?
180 A0 The key is the combination of the file descriptor number and the
182 open file description (also known as an "open file handle", the
183 kernel's internal representation of an open file).
184 Q1 What happens if you register the same file descriptor on an epoll
186 instance twice?
187 A1 You will probably get EEXIST. However, it is possible to add a
189 duplicate (dup(2), dup2(2), fcntl(2) F_DUPFD) file descriptor to
190 the same epoll instance. This can be a useful technique for fil‐
191 tering events, if the duplicate file descriptors are registered
192 with different events masks.
193 Q2 Can two epoll instances wait for the same file descriptor? If so,
195 are events reported to both epoll file descriptors?
196 A2 Yes, and events would be reported to both. However, careful pro‐
198 gramming may be needed to do this correctly.
199 Q3 Is the epoll file descriptor itself poll/epoll/selectable?
201 A3 Yes. If an epoll file descriptor has events waiting, then it will
203 indicate as being readable.
204 Q4 What happens if one attempts to put an epoll file descriptor into
206 its own file descriptor set?
207 A4 The epoll_ctl(2) call fails (EINVAL). However, you can add an
209 epoll file descriptor inside another epoll file descriptor set.
210 Q5 Can I send an epoll file descriptor over a UNIX domain socket to
212 another process?
213 A5 Yes, but it does not make sense to do this, since the receiving
215 process would not have copies of the file descriptors in the epoll
216 set.
217 Q6 Will closing a file descriptor cause it to be removed from all
219 epoll sets automatically?
220 A6 Yes, but be aware of the following point. A file descriptor is a
222 reference to an open file description (see open(2)). Whenever a
223 file descriptor is duplicated via dup(2), dup2(2), fcntl(2)
224 F_DUPFD, or fork(2), a new file descriptor referring to the same
225 open file description is created. An open file description contin‐
226 ues to exist until all file descriptors referring to it have been
227 closed. A file descriptor is removed from an epoll set only after
228 all the file descriptors referring to the underlying open file
229 description have been closed (or before if the file descriptor is
230 explicitly removed using epoll_ctl(2) EPOLL_CTL_DEL). This means
231 that even after a file descriptor that is part of an epoll set has
232 been closed, events may be reported for that file descriptor if
233 other file descriptors referring to the same underlying file
234 description remain open.
235 Q7 If more than one event occurs between epoll_wait(2) calls, are they
237 combined or reported separately?
238 A7 They will be combined.
240 Q8 Does an operation on a file descriptor affect the already collected
242 but not yet reported events?
243 A8 You can do two operations on an existing file descriptor. Remove
245 would be meaningless for this case. Modify will reread available
246 I/O.
247 Q9 Do I need to continuously read/write a file descriptor until EAGAIN
249 when using the EPOLLET flag (edge-triggered behavior) ?
250 A9 Receiving an event from epoll_wait(2) should suggest to you that
252 such file descriptor is ready for the requested I/O operation. You
253 must consider it ready until the next (nonblocking) read/write
254 yields EAGAIN. When and how you will use the file descriptor is
255 entirely up to you.
256 For packet/token-oriented files (e.g., datagram socket, terminal in
258 canonical mode), the only way to detect the end of the read/write
259 I/O space is to continue to read/write until EAGAIN.
260 For stream-oriented files (e.g., pipe, FIFO, stream socket), the
262 condition that the read/write I/O space is exhausted can also be
263 detected by checking the amount of data read from / written to the
264 target file descriptor. For example, if you call read(2) by asking
265 to read a certain amount of data and read(2) returns a lower number
266 of bytes, you can be sure of having exhausted the read I/O space
267 for the file descriptor. The same is true when writing using
268 write(2). (Avoid this latter technique if you cannot guarantee
269 that the monitored file descriptor always refers to a stream-ori‐
270 ented file.)
271 Possible pitfalls and ways to avoid them
273 o Starvation (edge-triggered)
274 If there is a large amount of I/O space, it is possible that by trying
276 to drain it the other files will not get processed causing starvation.
277 (This problem is not specific to epoll.)
278 The solution is to maintain a ready list and mark the file descriptor
280 as ready in its associated data structure, thereby allowing the appli‐
281 cation to remember which files need to be processed but still round
282 robin amongst all the ready files. This also supports ignoring subse‐
283 quent events you receive for file descriptors that are already ready.
284 o If using an event cache...
286 If you use an event cache or store all the file descriptors returned
288 from epoll_wait(2), then make sure to provide a way to mark its closure
289 dynamically (i.e., caused by a previous event's processing). Suppose
290 you receive 100 events from epoll_wait(2), and in event #47 a condition
291 causes event #13 to be closed. If you remove the structure and
292 close(2) the file descriptor for event #13, then your event cache might
293 still say there are events waiting for that file descriptor causing
294 confusion.
295 One solution for this is to call, during the processing of event 47,
297 epoll_ctl(EPOLL_CTL_DEL) to delete file descriptor 13 and close(2),
298 then mark its associated data structure as removed and link it to a
299 cleanup list. If you find another event for file descriptor 13 in your
300 batch processing, you will discover the file descriptor had been previ‐
301 ously removed and there will be no confusion.
302
304 The epoll API was introduced in Linux kernel 2.5.44. Support was added
305 to glibc in version 2.3.2.
306
308 The epoll API is Linux-specific. Some other systems provide similar
309 mechanisms, for example, FreeBSD has kqueue, and Solaris has /dev/poll.
310
312 The set of file descriptors that is being monitored via an epoll file
313 descriptor can be viewed via the entry for the epoll file descriptor in
314 the process's /proc/[pid]/fdinfo directory. See proc(5) for further
315 details.
316 The kcmp(2) KCMP_EPOLL_TFD operation can be used to test whether a file
318 descriptor is present in an epoll instance.
319
321 epoll_create(2), epoll_create1(2), epoll_ctl(2), epoll_wait(2),
322 poll(2), select(2)
323
325 This page is part of release 4.16 of the Linux man-pages project. A
326 description of the project, information about reporting bugs, and the
327 latest version of this page, can be found at
328 https://www.kernel.org/doc/man-pages/.
329Linux 2017-09-15 EPOLL(7)