\n| <\/td>\n | \n \u2022<\/p>\n<\/td>\n | <\/td>\n | \n The interest<\/i> list (sometimes also called the epoll<\/b> set): the set of file descriptors that the process has registered an interest in monitoring.<\/p>\n<\/td>\n<\/tr>\n\n| <\/td>\n | \n \u2022<\/p>\n<\/td>\n | <\/td>\n | \n The ready<\/i> list: the set of file descriptors that are “ready” for I\/O. The ready list is a subset of (or, more precisely, a set of references to) the file descriptors in the interest list. The ready list is dynamically populated by the kernel as a result of I\/O activity on those file descriptors.<\/p>\n<\/td>\n<\/tr>\n<\/table>\n The following system calls are provided to create and manage an epoll<\/b> instance:<\/p>\n\n\n| <\/td>\n | \n \u2022<\/p>\n<\/td>\n | <\/td>\n | \n epoll_create<\/b>(2) creates a new epoll<\/b> instance and returns a file descriptor referring to that instance. (The more recent epoll_create1<\/b>(2) extends the functionality of epoll_create<\/b>(2).)<\/p>\n<\/td>\n<\/tr>\n\n| <\/td>\n | \n \u2022<\/p>\n<\/td>\n | <\/td>\n | \n Interest in particular file descriptors is then registered via epoll_ctl<\/b>(2), which adds items to the interest list of the epoll<\/b> instance.<\/p>\n<\/td>\n<\/tr>\n\n| <\/td>\n | \n \u2022<\/p>\n<\/td>\n | <\/td>\n | \n epoll_wait<\/b>(2) waits for I\/O events, blocking the calling thread if no events are currently available. (This system call can be thought of as fetching items from the ready list of the epoll<\/b> instance.)<\/p>\n<\/td>\n<\/tr>\n<\/table>\n Level-triggered and edge-triggered<\/b>
The epoll<\/b> event distribution interface is able to behave both as edge-triggered (ET) and as level-triggered (LT). The difference between the two mechanisms can be described as follows. Suppose that this scenario happens:<\/p>\n\n\n| <\/td>\n | \n 1.<\/p>\n<\/td>\n | <\/td>\n | \n The file descriptor that represents the read side of a pipe (rfd<\/i>) is registered on the epoll<\/b> instance.<\/p>\n<\/td>\n<\/tr>\n\n| <\/td>\n | \n 2.<\/p>\n<\/td>\n | <\/td>\n | \n A pipe writer writes 2\u00a0kB of data on the write side of the pipe.<\/p>\n<\/td>\n<\/tr>\n | \n| <\/td>\n | \n 3.<\/p>\n<\/td>\n | <\/td>\n | \n A call to epoll_wait<\/b>(2) is done that will return rfd<\/i> as a ready file descriptor.<\/p>\n<\/td>\n<\/tr>\n\n| <\/td>\n | \n 4.<\/p>\n<\/td>\n | <\/td>\n | \n The pipe reader reads 1\u00a0kB of data from rfd<\/i>.<\/p>\n<\/td>\n<\/tr>\n\n| <\/td>\n | \n 5.<\/p>\n<\/td>\n | <\/td>\n | \n A call to epoll_wait<\/b>(2) is done.<\/p>\n<\/td>\n<\/tr>\n<\/table>\n If the rfd<\/i> file descriptor has been added to the epoll<\/b> interface using the EPOLLET<\/b> (edge-triggered) flag, the call to epoll_wait<\/b>(2) done in step 5<\/b> will probably hang despite the available data still present in the file input buffer; meanwhile the remote peer might be expecting a response based on the data it already sent. The reason for this is that edge-triggered mode delivers events only when changes occur on the monitored file descriptor. So, in step 5<\/b> the caller might end up waiting for some data that is already present inside the input buffer. In the above example, an event on rfd<\/i> will be generated because of the write done in 2<\/b> and the event is consumed in 3<\/b>. Since the read operation done in 4<\/b> does not consume the whole buffer data, the call to epoll_wait<\/b>(2) done in step 5<\/b> might block indefinitely.<\/p>\n An application that employs the EPOLLET<\/b> flag should use nonblocking file descriptors to avoid having a blocking read or write starve a task that is handling multiple file descriptors. The suggested way to use epoll<\/b> as an edge-triggered (EPOLLET<\/b>) interface is as follows:<\/p>\n\n\n| <\/td>\n | \n a)<\/p>\n<\/td>\n | <\/td>\n | \n with nonblocking file descriptors; and<\/p>\n<\/td>\n<\/tr>\n | \n| <\/td>\n | \n b)<\/p>\n<\/td>\n | <\/td>\n | \n by waiting for an event only after read<\/b>(2) or write<\/b>(2) return EAGAIN<\/b>.<\/p>\n<\/td>\n<\/tr>\n<\/table>\n By contrast, when used as a level-triggered interface (the default, when EPOLLET<\/b> is not specified), epoll<\/b> is simply a faster poll<\/b>(2), and can be used wherever the latter is used since it shares the same semantics.<\/p>\n Since even with edge-triggered epoll<\/b>, multiple events can be generated upon receipt of multiple chunks of data, the caller has the option to specify the EPOLLONESHOT<\/b> flag, to tell epoll<\/b> to disable the associated file descriptor after the receipt of an event with epoll_wait<\/b>(2). When the EPOLLONESHOT<\/b> flag is specified, it is the caller\u2019s responsibility to rearm the file descriptor using epoll_ctl<\/b>(2) with EPOLL_CTL_MOD<\/b>.<\/p>\n If multiple threads (or processes, if child processes have inherited the epoll<\/b> file descriptor across fork<\/b>(2)) are blocked in epoll_wait<\/b>(2) waiting on the same epoll file descriptor and a file descriptor in the interest list that is marked for edge-triggered (EPOLLET<\/b>) notification becomes ready, just one of the threads (or processes) is awoken from epoll_wait<\/b>(2). This provides a useful optimization for avoiding “thundering herd” wake-ups in some scenarios.<\/p>\n Interaction with autosleep<\/b>
If the system is in autosleep<\/b> mode via \/sys\/power\/autosleep<\/i> and an event happens which wakes the device from sleep, the device driver will keep the device awake only until that event is queued. To keep the device awake until the event has been processed, it is necessary to use the epoll_ctl<\/b>(2) EPOLLWAKEUP<\/b> flag.<\/p>\n When the EPOLLWAKEUP<\/b> flag is set in the events<\/b> field for a struct epoll_event<\/i>, the system will be kept awake from the moment the event is queued, through the epoll_wait<\/b>(2) call which returns the event until the subsequent epoll_wait<\/b>(2) call. If the event should keep the system awake beyond that time, then a separate wake_lock<\/i> should be taken before the second epoll_wait<\/b>(2) call.<\/p>\n \/proc interfaces<\/b>
The following interfaces can be used to limit the amount of kernel memory consumed by epoll:
\/proc\/sys\/fs\/epoll\/max_user_watches<\/i> (since Linux 2.6.28)<\/p>\n This specifies a limit on the total number of file descriptors that a user can register across all epoll instances on the system. The limit is per real user ID. Each registered file descriptor costs roughly 90 bytes on a 32-bit kernel, and roughly 160 bytes on a 64-bit kernel. Currently, the default value for max_user_watches<\/i> is 1\/25 (4%) of the available low memory, divided by the registration cost in bytes.<\/p>\n Example for suggested usage<\/b>
While the usage of epoll<\/b> when employed as a level-triggered interface does have the same semantics as poll<\/b>(2), the edge-triggered usage requires more clarification to avoid stalls in the application event loop. In this example, listener is a nonblocking socket on which listen<\/b>(2) has been called. The function do_use_fd()<\/i> uses the new ready file descriptor until EAGAIN<\/b> is returned by either read<\/b>(2) or write<\/b>(2). An event-driven state machine application should, after having received EAGAIN<\/b>, record its current state so that at the next call to do_use_fd()<\/i> it will continue to read<\/b>(2) or write<\/b>(2) from where it stopped before.<\/p>\n #define MAX_EVENTS 10
struct epoll_event ev, events[MAX_EVENTS];
int listen_sock, conn_sock, nfds, epollfd;<\/p>\n \/* Code to set up listening socket, ‘listen_sock’,
(socket(), bind(), listen()) omitted *\/<\/p>\n epollfd = epoll_create1(0);
if (epollfd == \u22121) {
perror(“epoll_create1”);
exit(EXIT_FAILURE);
}<\/p>\n ev.events = EPOLLIN;
ev.data.fd = listen_sock;
if (epoll_ctl(epollfd, EPOLL_CTL_ADD, listen_sock, &ev) == \u22121) {
perror(“epoll_ctl: listen_sock”);
exit(EXIT_FAILURE);
}<\/p>\n for (;;) {
nfds = epoll_wait(epollfd, events, MAX_EVENTS, \u22121);
if (nfds == \u22121) {
perror(“epoll_wait”);
exit(EXIT_FAILURE);
}<\/p>\n for (n = 0; n < nfds; ++n) {
if (events[n].data.fd == listen_sock) {
conn_sock = accept(listen_sock,
(struct sockaddr *) &addr, &addrlen);
if (conn_sock == \u22121) {
perror(“accept”);
exit(EXIT_FAILURE);
}
setnonblocking(conn_sock);
ev.events = EPOLLIN | EPOLLET;
ev.data.fd = conn_sock;
if (epoll_ctl(epollfd, EPOLL_CTL_ADD, conn_sock,
&ev) == \u22121) {
perror(“epoll_ctl: conn_sock”);
exit(EXIT_FAILURE);
}
} else {
do_use_fd(events[n].data.fd);
}
}
}<\/p>\n When used as an edge-triggered interface, for performance reasons, it is possible to add the file descriptor inside the epoll<\/b> interface (EPOLL_CTL_ADD<\/b>) once by specifying (EPOLLIN<\/b>|EPOLLOUT<\/b>). This allows you to avoid continuously switching between EPOLLIN<\/b> and EPOLLOUT<\/b> calling epoll_ctl<\/b>(2) with EPOLL_CTL_MOD<\/b>.<\/p>\n Questions and answers<\/b><\/p>\n | | | | | | | | | |