1epoll_wait(2) System Calls Manual epoll_wait(2)
2
6 epoll_wait, epoll_pwait, epoll_pwait2 - wait for an I/O event on an
7 epoll file descriptor
8
10 Standard C library (libc, -lc)
11
13 #include <sys/epoll.h>
14 int epoll_wait(int epfd, struct epoll_event *events,
16 int maxevents, int timeout);
17 int epoll_pwait(int epfd, struct epoll_event *events,
18 int maxevents, int timeout,
19 const sigset_t *_Nullable sigmask);
20 int epoll_pwait2(int epfd, struct epoll_event *events,
21 int maxevents, const struct timespec *_Nullable timeout,
22 const sigset_t *_Nullable sigmask);
23
25 The epoll_wait() system call waits for events on the epoll(7) instance
26 referred to by the file descriptor epfd. The buffer pointed to by
27 events is used to return information from the ready list about file de‐
28 scriptors in the interest list that have some events available. Up to
29 maxevents are returned by epoll_wait(). The maxevents argument must be
30 greater than zero.
31 The timeout argument specifies the number of milliseconds that
33 epoll_wait() will block. Time is measured against the CLOCK_MONOTONIC
34 clock.
35 A call to epoll_wait() will block until either:
37 • a file descriptor delivers an event;
39 • the call is interrupted by a signal handler; or
41 • the timeout expires.
43 Note that the timeout interval will be rounded up to the system clock
45 granularity, and kernel scheduling delays mean that the blocking inter‐
46 val may overrun by a small amount. Specifying a timeout of -1 causes
47 epoll_wait() to block indefinitely, while specifying a timeout equal to
48 zero cause epoll_wait() to return immediately, even if no events are
49 available.
50 The struct epoll_event is described in epoll_event(3type).
52 The data field of each returned epoll_event structure contains the same
54 data as was specified in the most recent call to epoll_ctl(2)
55 (EPOLL_CTL_ADD, EPOLL_CTL_MOD) for the corresponding open file descrip‐
56 tor.
57 The events field is a bit mask that indicates the events that have oc‐
59 curred for the corresponding open file description. See epoll_ctl(2)
60 for a list of the bits that may appear in this mask.
61 epoll_pwait()
63 The relationship between epoll_wait() and epoll_pwait() is analogous to
64 the relationship between select(2) and pselect(2): like pselect(2),
65 epoll_pwait() allows an application to safely wait until either a file
66 descriptor becomes ready or until a signal is caught.
67 The following epoll_pwait() call:
69 ready = epoll_pwait(epfd, &events, maxevents, timeout, &sigmask);
71 is equivalent to atomically executing the following calls:
73 sigset_t origmask;
75 pthread_sigmask(SIG_SETMASK, &sigmask, &origmask);
77 ready = epoll_wait(epfd, &events, maxevents, timeout);
78 pthread_sigmask(SIG_SETMASK, &origmask, NULL);
79 The sigmask argument may be specified as NULL, in which case
81 epoll_pwait() is equivalent to epoll_wait().
82 epoll_pwait2()
84 The epoll_pwait2() system call is equivalent to epoll_pwait() except
85 for the timeout argument. It takes an argument of type timespec to be
86 able to specify nanosecond resolution timeout. This argument functions
87 the same as in pselect(2) and ppoll(2). If timeout is NULL, then
88 epoll_pwait2() can block indefinitely.
89
91 On success, epoll_wait() returns the number of file descriptors ready
92 for the requested I/O, or zero if no file descriptor became ready dur‐
93 ing the requested timeout milliseconds. On failure, epoll_wait() re‐
94 turns -1 and errno is set to indicate the error.
95
97 EBADF epfd is not a valid file descriptor.
98 EFAULT The memory area pointed to by events is not accessible with
100 write permissions.
101 EINTR The call was interrupted by a signal handler before either (1)
103 any of the requested events occurred or (2) the timeout expired;
104 see signal(7).
105 EINVAL epfd is not an epoll file descriptor, or maxevents is less than
107 or equal to zero.
108
110 Linux.
111
113 epoll_wait()
114 Linux 2.6, glibc 2.3.2.
115 epoll_pwait()
117 Linux 2.6.19, glibc 2.6.
118 epoll_pwait2()
120 Linux 5.11.
121
123 While one thread is blocked in a call to epoll_wait(), it is possible
124 for another thread to add a file descriptor to the waited-upon epoll
125 instance. If the new file descriptor becomes ready, it will cause the
126 epoll_wait() call to unblock.
127 If more than maxevents file descriptors are ready when epoll_wait() is
129 called, then successive epoll_wait() calls will round robin through the
130 set of ready file descriptors. This behavior helps avoid starvation
131 scenarios, where a process fails to notice that additional file de‐
132 scriptors are ready because it focuses on a set of file descriptors
133 that are already known to be ready.
134 Note that it is possible to call epoll_wait() on an epoll instance
136 whose interest list is currently empty (or whose interest list becomes
137 empty because file descriptors are closed or removed from the interest
138 in another thread). The call will block until some file descriptor is
139 later added to the interest list (in another thread) and that file de‐
140 scriptor becomes ready.
141 C library/kernel differences
143 The raw epoll_pwait() and epoll_pwait2() system calls have a sixth ar‐
144 gument, size_t sigsetsize, which specifies the size in bytes of the
145 sigmask argument. The glibc epoll_pwait() wrapper function specifies
146 this argument as a fixed value (equal to sizeof(sigset_t)).
147
149 Before Linux 2.6.37, a timeout value larger than approximately LONG_MAX
150 / HZ milliseconds is treated as -1 (i.e., infinity). Thus, for exam‐
151 ple, on a system where sizeof(long) is 4 and the kernel HZ value is
152 1000, this means that timeouts greater than 35.79 minutes are treated
153 as infinity.
154
156 epoll_create(2), epoll_ctl(2), epoll(7)
157Linux man-pages 6.04 2023-03-30 epoll_wait(2)