The man page of epoll has a sample code for edge triggered like the following :
for (;;) {
nfds = epoll_wait(epollfd, events, MAX_EVENTS, -1);
if (nfds == -1) {
perror("epoll_pwait");
exit(EXIT_FAILURE);
}
for (n = 0; n < nfds; ++n) {
if (events[n].data.fd == listen_sock) {
conn_sock = accept(listen_sock,
(struct sockaddr *) &local, &addrlen);
if (conn_sock == -1) {
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) == -1) {
perror("epoll_ctl: conn_sock");
exit(EXIT_FAILURE);
}
} else {
do_use_fd(events[n].data.fd);
}
}
}
In function do_use_fd , i call nonblocked recv in while loop until EAGAIN, the sample code works fine.
I have a question about this sample code, suppose now I have 50 socket clients connections , suddenly 10 clients writes data at the same time, so epoll_wait() will return 10 and then go to for loop :
for (n = 0; n < nfds; ++n)
it will call do_use_fd(events[n].data.fd);
for those 10 clients , suppose n=5 is done , and n=6 is not yet finished , suddenly the file description of event n= 3 has receive new data , after all of those 10 events are done and back to epoll_wait , will I get the event inform me that there is a client has new data to read ? or I will miss it because when event happened , the code not in epoll_wait !!
As long as you read until you get an EAGAIN
error, you will get the event the next time you are calling epoll_wait
.
The event is only triggered when there is a change between empty and non-empty (or full and non-full for EPOLLOUT
), but that status then remains until the event is delivered via epoll_wait
.
On a somewhat related note: if you register for EPOLLIN
and EPOLLOUT
events and assuming you never fill up the send buffer, you still get the EPOLLOUT
flag set in the event returned by epoll_wait
each time EPOLLIN
is triggered - see https://lkml.org/lkml/2011/11/17/234 for a more detailed explanation.
And finally, the exact behaviour of edge-triggered mode actually depends on the socket type used and isn't really documented anywhere. I did some tests some time ago and documented my findings here: http://cmeerw.org/blog/753.html#753 - in short, for datagram sockets you might get more events than you would expect.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With