1EPOLL_CTL(2) Linux Programmer's Manual EPOLL_CTL(2)
2
6 epoll_ctl - control interface for an epoll file descriptor
7
9 #include <sys/epoll.h>
10 int epoll_ctl(int epfd, int op, int fd, struct epoll_event *event);
12
14 This system call performs control operations on the epoll(7) instance
15 referred to by the file descriptor epfd. It requests that the opera‐
16 tion op be performed for the target file descriptor, fd.
17 Valid values for the op argument are:
19 EPOLL_CTL_ADD
21 Register the target file descriptor fd on the epoll instance
22 referred to by the file descriptor epfd and associate the event
23 event with the internal file linked to fd.
24 EPOLL_CTL_MOD
26 Change the event event associated with the target file descrip‐
27 tor fd.
28 EPOLL_CTL_DEL
30 Remove (deregister) the target file descriptor fd from the epoll
31 instance referred to by epfd. The event is ignored and can be
32 NULL (but see BUGS below).
33 The event argument describes the object linked to the file descriptor
35 fd. The struct epoll_event is defined as:
36 typedef union epoll_data {
38 void *ptr;
39 int fd;
40 uint32_t u32;
41 uint64_t u64;
42 } epoll_data_t;
43 struct epoll_event {
45 uint32_t events; /* Epoll events */
46 epoll_data_t data; /* User data variable */
47 };
48 The events member is a bit mask composed by ORing together zero or more
50 of the following available event types:
51 EPOLLIN
53 The associated file is available for read(2) operations.
54 EPOLLOUT
56 The associated file is available for write(2) operations.
57 EPOLLRDHUP (since Linux 2.6.17)
59 Stream socket peer closed connection, or shut down writing half
60 of connection. (This flag is especially useful for writing sim‐
61 ple code to detect peer shutdown when using Edge Triggered moni‐
62 toring.)
63 EPOLLPRI
65 There is an exceptional condition on the file descriptor. See
66 the discussion of POLLPRI in poll(2).
67 EPOLLERR
69 Error condition happened on the associated file descriptor.
70 This event is also reported for the write end of a pipe when the
71 read end has been closed. epoll_wait(2) will always report for
72 this event; it is not necessary to set it in events.
73 EPOLLHUP
75 Hang up happened on the associated file descriptor.
76 epoll_wait(2) will always wait for this event; it is not neces‐
77 sary to set it in events.
78 Note that when reading from a channel such as a pipe or a stream
80 socket, this event merely indicates that the peer closed its end
81 of the channel. Subsequent reads from the channel will return 0
82 (end of file) only after all outstanding data in the channel has
83 been consumed.
84 EPOLLET
86 Sets the Edge Triggered behavior for the associated file
87 descriptor. The default behavior for epoll is Level Triggered.
88 See epoll(7) for more detailed information about Edge and Level
89 Triggered event distribution architectures.
90 EPOLLONESHOT (since Linux 2.6.2)
92 Sets the one-shot behavior for the associated file descriptor.
93 This means that after an event is pulled out with epoll_wait(2)
94 the associated file descriptor is internally disabled and no
95 other events will be reported by the epoll interface. The user
96 must call epoll_ctl() with EPOLL_CTL_MOD to rearm the file
97 descriptor with a new event mask.
98 EPOLLWAKEUP (since Linux 3.5)
100 If EPOLLONESHOT and EPOLLET are clear and the process has the
101 CAP_BLOCK_SUSPEND capability, ensure that the system does not
102 enter "suspend" or "hibernate" while this event is pending or
103 being processed. The event is considered as being "processed"
104 from the time when it is returned by a call to epoll_wait(2)
105 until the next call to epoll_wait(2) on the same epoll(7) file
106 descriptor, the closure of that file descriptor, the removal of
107 the event file descriptor with EPOLL_CTL_DEL, or the clearing of
108 EPOLLWAKEUP for the event file descriptor with EPOLL_CTL_MOD.
109 See also BUGS.
110 EPOLLEXCLUSIVE (since Linux 4.5)
112 Sets an exclusive wakeup mode for the epoll file descriptor that
113 is being attached to the target file descriptor, fd. When a
114 wakeup event occurs and multiple epoll file descriptors are
115 attached to the same target file using EPOLLEXCLUSIVE, one or
116 more of the epoll file descriptors will receive an event with
117 epoll_wait(2). The default in this scenario (when EPOLLEXCLU‐
118 SIVE is not set) is for all epoll file descriptors to receive an
119 event. EPOLLEXCLUSIVE is thus useful for avoiding thundering
120 herd problems in certain scenarios.
121 If the same file descriptor is in multiple epoll instances, some
123 with the EPOLLEXCLUSIVE flag, and others without, then events
124 will be provided to all epoll instances that did not specify
125 EPOLLEXCLUSIVE, and at least one of the epoll instances that did
126 specify EPOLLEXCLUSIVE.
127 The following values may be specified in conjunction with
129 EPOLLEXCLUSIVE: EPOLLIN, EPOLLOUT, EPOLLWAKEUP, and EPOLLET.
130 EPOLLHUP and EPOLLERR can also be specified, but this is not
131 required: as usual, these events are always reported if they
132 occur, regardless of whether they are specified in events.
133 Attempts to specify other values in events yield an error.
134 EPOLLEXCLUSIVE may be used only in an EPOLL_CTL_ADD operation;
135 attempts to employ it with EPOLL_CTL_MOD yield an error. If
136 EPOLLEXCLUSIVE has been set using epoll_ctl(), then a subsequent
137 EPOLL_CTL_MOD on the same epfd, fd pair yields an error. A call
138 to epoll_ctl() that specifies EPOLLEXCLUSIVE in events and spec‐
139 ifies the target file descriptor fd as an epoll instance will
140 likewise fail. The error in all of these cases is EINVAL.
141
143 When successful, epoll_ctl() returns zero. When an error occurs,
144 epoll_ctl() returns -1 and errno is set appropriately.
145
147 EBADF epfd or fd is not a valid file descriptor.
148 EEXIST op was EPOLL_CTL_ADD, and the supplied file descriptor fd is
150 already registered with this epoll instance.
151 EINVAL epfd is not an epoll file descriptor, or fd is the same as epfd,
153 or the requested operation op is not supported by this inter‐
154 face.
155 EINVAL An invalid event type was specified along with EPOLLEXCLUSIVE in
157 events.
158 EINVAL op was EPOLL_CTL_MOD and events included EPOLLEXCLUSIVE.
160 EINVAL op was EPOLL_CTL_MOD and the EPOLLEXCLUSIVE flag has previously
162 been applied to this epfd, fd pair.
163 EINVAL EPOLLEXCLUSIVE was specified in event and fd refers to an epoll
165 instance.
166 ELOOP fd refers to an epoll instance and this EPOLL_CTL_ADD operation
168 would result in a circular loop of epoll instances monitoring
169 one another.
170 ENOENT op was EPOLL_CTL_MOD or EPOLL_CTL_DEL, and fd is not registered
172 with this epoll instance.
173 ENOMEM There was insufficient memory to handle the requested op control
175 operation.
176 ENOSPC The limit imposed by /proc/sys/fs/epoll/max_user_watches was
178 encountered while trying to register (EPOLL_CTL_ADD) a new file
179 descriptor on an epoll instance. See epoll(7) for further
180 details.
181 EPERM The target file fd does not support epoll. This error can occur
183 if fd refers to, for example, a regular file or a directory.
184
186 epoll_ctl() was added to the kernel in version 2.6.
187
189 epoll_ctl() is Linux-specific. Library support is provided in glibc
190 starting with version 2.3.2.
191
193 The epoll interface supports all file descriptors that support poll(2).
194
196 In kernel versions before 2.6.9, the EPOLL_CTL_DEL operation required a
197 non-null pointer in event, even though this argument is ignored. Since
198 Linux 2.6.9, event can be specified as NULL when using EPOLL_CTL_DEL.
199 Applications that need to be portable to kernels before 2.6.9 should
200 specify a non-null pointer in event.
201 If EPOLLWAKEUP is specified in flags, but the caller does not have the
203 CAP_BLOCK_SUSPEND capability, then the EPOLLWAKEUP flag is silently
204 ignored. This unfortunate behavior is necessary because no validity
205 checks were performed on the flags argument in the original implementa‐
206 tion, and the addition of the EPOLLWAKEUP with a check that caused the
207 call to fail if the caller did not have the CAP_BLOCK_SUSPEND capabil‐
208 ity caused a breakage in at least one existing user-space application
209 that happened to randomly (and uselessly) specify this bit. A robust
210 application should therefore double check that it has the
211 CAP_BLOCK_SUSPEND capability if attempting to use the EPOLLWAKEUP flag.
212
214 epoll_create(2), epoll_wait(2), poll(2), epoll(7)
215
217 This page is part of release 4.16 of the Linux man-pages project. A
218 description of the project, information about reporting bugs, and the
219 latest version of this page, can be found at
220 https://www.kernel.org/doc/man-pages/.
221Linux 2017-09-15 EPOLL_CTL(2)