EPOLL(4) Linux Programmer's Manual EPOLL(4)
NAME
epoll - I/O event notification facility
SYNOPSIS
#include <sys/epoll.h>
DESCRIPTION
epoll is a variant of poll(2) that can be used either as Edge or Level
Triggered interface and scales well to large numbers of watched fds.
Three system calls are provided to set(7,n,1 builtins) up and control an epoll set:
epoll_create(2), epoll_ctl(2), epoll_wait(2).
An epoll set(7,n,1 builtins) is connected to a file(1,n) descriptor created by epoll_cre-
ate(2). Interest for certain file(1,n) descriptors is then registered via
epoll_ctl(2). Finally, the actual wait is started by epoll_wait(2).
NOTES
The epoll event distribution interface is able to behave both as Edge
Triggered ( ET ) and Level Triggered ( LT ). The difference between ET
and LT event distribution mechanism can be described as follows. Sup-
pose that this scenario happens :
1 The file(1,n) descriptor that represents the read(2,n,1 builtins) side of a pipe(2,8) (
RFD ) is added inside the epoll device.
2 Pipe writer writes 2Kb of data on the write(1,2) side of the pipe.
3 A call to epoll_wait(2) is done that will return RFD as ready
file(1,n) descriptor.
4 The pipe(2,8) reader reads 1Kb of data from RFD.
5 A call to epoll_wait(2) is done.
If the RFD file(1,n) descriptor has been added to the epoll interface using
the EPOLLET flag, the call to epoll_wait(2) done in(1,8) step 5 will proba-
bly hang because of the available data still present in(1,8) the file(1,n) input
buffers and the remote peer might be expecting a response based on the
data it already sent. The reason for this is that Edge Triggered event
distribution delivers events only when events happens on the monitored
file. So, in(1,8) step 5 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 will be generated because of the write(1,2) done in(1,8) 2 , and the
event is consumed in(1,8) 3. Since the read(2,n,1 builtins) operation done in(1,8) 4 does not
consume the whole buffer data, the call to epoll_wait(2) done in(1,8) step 5
might lock indefinitely. The epoll interface, when used with the EPOL-
LET flag ( Edge Triggered ) should use non-blocking file(1,n) descriptors to
avoid having a blocking read(2,n,1 builtins) or write(1,2) starve the task that is handling
multiple file(1,n) descriptors. The suggested way to use epoll as an Edge
Triggered ( EPOLLET ) interface is below, and possible pitfalls to
avoid follow.
i with non-blocking file(1,n) descriptors
ii by going to wait for an event only after read(2,n,1 builtins)(2) or
write(1,2)(2) return EAGAIN
On the contrary, when used as a Level Triggered interface, epoll is by
all means a faster poll(2), and can be used wherever the latter is used
since it shares the same semantics. Since even with the Edge Triggered
epoll multiple events can be generated up on receival of multiple
chunks of data, the caller has the option to specify the EPOLLONESHOT
flag, to tell epoll to disable the associated file(1,n) descriptor after the
receival of an event with epoll_wait(2). When the EPOLLONESHOT flag is
specified, it is caller responsibility to rearm the file(1,n) descriptor
using epoll_ctl(2) with EPOLL_CTL_MOD.
EXAMPLE FOR SUGGESTED USAGE
While the usage of epoll when employed like a Level Triggered interface
does have the same semantics of poll(2), an Edge Triggered usage
requires more clarifiction to avoid stalls in(1,8) the application event
loop. In this example, listener is a non-blocking socket(2,7,n) on which lis-
ten(2) has been called. The function do_use_fd() uses the new ready
file(1,n) descriptor until EAGAIN is returned by either read(2,n,1 builtins)(2) or write(1,2)(2).
An event driven state machine application should, after having received
EAGAIN, record its current state so that at the next call to
do_use_fd() it will continue to read(2,n,1 builtins)(2) or write(1,2)(2) from where it
stopped before.
struct epoll_event ev, *events;
for(;;) {
nfds = epoll_wait(kdpfd, events, maxevents, -1);
for(n = 0; n < nfds; ++n) {
if(3,n)(events[n].data.fd == listener) {
client = accept(2,8)(listener, (struct sockaddr *) &local,
&addrlen);
if(3,n)(client < 0){
perror(1,3)("accept(2,8)");
continue;
}
setnonblocking(client);
ev.events = EPOLLIN | EPOLLET;
ev.data.fd = client;
if(3,n) (epoll_ctl(kdpfd, EPOLL_CTL_ADD, client, &ev) < 0) {
fprintf(stderr, "epoll set(7,n,1 builtins) insertion error: fd=%d0,
client);
return -1;
}
}
else
do_use_fd(events[n].data.fd);
}
}
When used as an Edge triggered interface, for performance reasons, it
is possible to add the file(1,n) descriptor inside the epoll interface (
EPOLL_CTL_ADD ) once by specifying ( EPOLLIN|EPOLLOUT ). This allows
you to avoid continuously switching between EPOLLIN and EPOLLOUT call-
ing epoll_ctl(2) with EPOLL_CTL_MOD.
QUESTIONS AND ANSWERS (from linux-kernel)
Q1 What happens if(3,n) you add the same fd to an epoll_set
twice?
A1 You will probably get EEXIST. However, it is possible
that two threads may add the same fd twice. This is a
harmless condition.
Q2 Can two epoll sets wait for the same fd? If so, are
events reported to both epoll sets fds?
A2 Yes. However, it is not recommended. Yes it would be
reported to both.
Q3 Is the epoll fd itself poll/epoll/selectable?
A3 Yes.
Q4 What happens if(3,n) the epoll fd is put into its own fd set(7,n,1 builtins)?
A4 It will fail. However, you can add an epoll fd inside
another epoll fd set.
Q5 Can I send(2,n) the epoll fd over a unix-socket to another
process?
A5 No.
Q6 Will the close(2,7,n) of an fd cause it to be removed from all
epoll sets automatically?
A6 Yes.
Q7 If more than one event comes in(1,8) between epoll_wait(2)
calls, are they combined or reported separately?
A7 They will be combined.
Q8 Does an operation on an fd affect the already collected
but not yet reported events?
A8 You can do two operations on an existing fd. Remove would
be meaningless for this case. Modify will re-read avail-
able I/O.
Q9 Do I need to continuously read(2,n,1 builtins)/write(1,2) an fd until EAGAIN
when using the EPOLLET flag ( Edge Triggered behaviour )
?
A9 No you don't. Receiving an event from epoll_wait(2)
should suggest to you that such file(1,n) descriptor is ready
for the requested I/O operation. You have simply to con-
sider it ready until you will receive the next EAGAIN.
When and how you will use such file(1,n) descriptor is
entirely up to you. Also, the condition that the
read(2,n,1 builtins)/write(1,2) I/O space is exhausted can be detected by
checking the amount of data read(2,n,1 builtins)/write(1,2) from/to the target
file(1,n) descriptor. For example, if(3,n) you call read(2,n,1 builtins)(2) by ask-
ing to read(2,n,1 builtins) a certain amount of data and read(2,n,1 builtins)(2) returns
a lower number of bytes, you can be sure to have
exhausted the read(2,n,1 builtins) I/O space for such file(1,n) descriptor.
Same is valid when writing using the write(1,2)(2) function.
POSSIBLE PITFALLS AND WAYS TO AVOID THEM
o Starvation ( Edge Triggered )
If there is a large amount of I/O space, it is possible that by
trying to drain it the other files will not get processed caus-
ing starvation. This is not specific to epoll.
The solution is to maintain a ready list and mark the file(1,n)
descriptor as ready in(1,8) its associated data structure, thereby
allowing the application to remember which files need to be pro-
cessed but still round robin amongst all the ready files. This
also supports ignoring subsequent events you receive for fd's
that are already ready.
o If using an event cache...
If you use an event cache or store all the fd's returned from
epoll_wait(2), then make sure to provide a way to mark its clo-
sure dynamically (ie- caused by a previous event's processing).
Suppose you receive 100 events from epoll_wait(2), and in(1,8) eventi
#47 a condition causes event #13 to be closed. If you remove
the structure and close(2,7,n)() the fd for event #13, then your event
cache might still say there are events waiting for that fd caus-
ing confusion.
One solution for this is to call, during the processing of event
47, epoll_ctl(EPOLL_CTL_DEL) to delete fd 13 and close(2,7,n)(), then
mark its associated data structure as removed and link(1,2) it to a
cleanup list. If you find another event for fd 13 in(1,8) your batch
processing, you will discover(1,3,5) the fd had been previously removed
and there will be no confusion.
CONFORMING TO
epoll(4) is a new API introduced in(1,8) Linux kernel 2.5.44. Its interface
should be finalized in(1,8) Linux kernel 2.5.66.
SEE ALSO
epoll_create(2), epoll_ctl(2), epoll_wait(2)
Linux 2002-10-23 EPOLL(4)