man7/epoll.7

   1 .\"
   2 .\"  epoll by Davide Libenzi ( efficient event notification retrieval )
   3 .\"  Copyright (C) 2003  Davide Libenzi
   4 .\"
   5 .\"  This program is free software; you can redistribute it and/or modify
   6 .\"  it under the terms of the GNU General Public License as published by
   7 .\"  the Free Software Foundation; either version 2 of the License, or
   8 .\"  (at your option) any later version.
   9 .\"
  10 .\"  This program is distributed in the hope that it will be useful,
  11 .\"  but WITHOUT ANY WARRANTY; without even the implied warranty of
  12 .\"  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13 .\"  GNU General Public License for more details.
  14 .\"
  15 .\"  You should have received a copy of the GNU General Public License
  16 .\"  along with this program; if not, write to the Free Software
  17 .\"  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  18 .\"
  19 .\"  Davide Libenzi <davidel@xmailserver.org>
  20 .\"
  21 .\"
  22 .TH EPOLL 7 "2002-10-23" Linux "Linux Programmer's Manual"
  23 .SH NAME
  24 epoll \- I/O event notification facility
  25 .SH SYNOPSIS
  26 .B #include <sys/epoll.h>
  27 .SH DESCRIPTION
  28 .B epoll
  29 is a variant of
  30 .BR poll (2)
  31 that can be used either as Edge or Level Triggered interface and scales
  32 well to large numbers of watched fds.
  33 Three system calls are provided to
  34 set up and control an
  35 .B epoll
  36 set:
  37 .BR epoll_create (2),
  38 .BR epoll_ctl (2),
  39 .BR epoll_wait (2).
  40
  41 An
  42 .B epoll
  43 set is connected to a file descriptor created by
  44 .BR epoll_create (2).
  45 Interest for certain file descriptors is then registered via
  46 .BR epoll_ctl (2).
  47 Finally, the actual wait is started by
  48 .BR epoll_wait (2).
  49 .SH NOTES
  50 The
  51 .B epoll
  52 event distribution interface is able to behave both as Edge Triggered
  53 ( ET ) and Level Triggered ( LT ).
  54 The difference between ET and LT
  55 event distribution mechanism can be described as follows.
  56 Suppose that
  57 this scenario happens :
  58 .TP
  59 .B 1
  60 The file descriptor that represents the read side of a pipe (
  61 .B RFD
  62 ) is added inside the
  63 .B epoll
  64 device.
  65 .TP
  66 .B 2
  67 Pipe writer writes 2Kb of data on the write side of the pipe.
  68 .TP
  69 .B 3
  70 A call to
  71 .BR epoll_wait (2)
  72 is done that will return
  73 .B RFD
  74 as ready file descriptor.
  75 .TP
  76 .B 4
  77 The pipe reader reads 1Kb of data from
  78 .BR RFD .
  79 .TP
  80 .B 5
  81 A call to
  82 .BR epoll_wait (2)
  83 is done.
  84 .PP
  85 If the
  86 .B RFD
  87 file descriptor has been added to the
  88 .B epoll
  89 interface using the
  90 .B EPOLLET
  91 flag, the call to
  92 .BR epoll_wait (2)
  93 done in step
  94 .B 5
  95 will probably hang because of the available data still present in the file
  96 input buffers and the remote peer might be expecting a response based on the
  97 data it already sent.
  98 The reason for this is that Edge Triggered event
  99 distribution delivers events only when events happens on the monitored file.
 100 So, in step
 101 .B 5
 102 the caller might end up waiting for some data that is already present inside
 103 the input buffer.
 104 In the above example, an event on
 105 .B RFD
 106 will be generated because of the write done in
 107 .BR 2
 108 and the event is consumed in
 109 .BR 3 .
 110 Since the read operation done in
 111 .B 4
 112 does not consume the whole buffer data, the call to
 113 .BR epoll_wait (2)
 114 done in step
 115 .B 5
 116 might lock indefinitely.
 117 The
 118 .B epoll
 119 interface, when used with the
 120 .B EPOLLET
 121 flag ( Edge Triggered )
 122 should use non-blocking file descriptors to avoid having a blocking
 123 read or write starve the task that is handling multiple file descriptors.
 124 The suggested way to use
 125 .B epoll
 126 as an Edge Triggered
 127 .RB ( EPOLLET )
 128 interface is below, and possible pitfalls to avoid follow.
 129 .RS
 130 .TP
 131 .B i
 132 with non-blocking file descriptors
 133 .TP
 134 .B ii
 135 by going to wait for an event only after
 136 .BR read (2)
 137 or
 138 .BR write (2)
 139 return EAGAIN
 140 .RE
 141 .PP
 142 On the contrary, when used as a Level Triggered interface,
 143 .B epoll
 144 is by all means a faster
 145 .BR poll (2),
 146 and can be used wherever the latter is used since it shares the
 147 same semantics.
 148 Since even with the Edge Triggered
 149 .B epoll
 150 multiple events can be generated up on receipt of multiple chunks of data,
 151 the caller has the option to specify the
 152 .B EPOLLONESHOT
 153 flag, to tell
 154 .B epoll
 155 to disable the associated file descriptor after the receipt of an event with
 156 .BR epoll_wait (2).
 157 When the
 158 .B EPOLLONESHOT
 159 flag is specified,
 160 it is caller responsibility to rearm the file descriptor using
 161 .BR epoll_ctl (2)
 162 with
 163 .BR EPOLL_CTL_MOD .
 164 .SH EXAMPLE FOR SUGGESTED USAGE
 165 While the usage of
 166 .B epoll
 167 when employed like a Level Triggered interface does have the same
 168 semantics of
 169 .BR poll (2),
 170 an Edge Triggered usage requires more clarification to avoid stalls
 171 in the application event loop.
 172 In this example, listener is a
 173 non-blocking socket on which
 174 .BR listen (2)
 175 has been called.
 176 The function do_use_fd() uses the new ready
 177 file descriptor until EAGAIN is returned by either
 178 .BR read (2)
 179 or
 180 .BR write (2).
 181 An event driven state machine application should, after having received
 182 EAGAIN, record its current state so that at the next call to do_use_fd()
 183 it will continue to
 184 .BR read (2)
 185 or
 186 .BR write (2)
 187 from where it stopped before.
 188
 189 .nf
 190 struct epoll_event ev, *events;
 191
 192 for(;;) {
 193     nfds = epoll_wait(kdpfd, events, maxevents, \-1);
 194
 195     for (n = 0; n < nfds; ++n) {
 196         if (events[n].data.fd == listener) {
 197             client = accept(listener, (struct sockaddr *) &local,
 198                             &addrlen);
 199             if(client < 0){
 200                 perror("accept");
 201                 continue;
 202             }
 203             setnonblocking(client);
 204             ev.events = EPOLLIN | EPOLLET;
 205             ev.data.fd = client;
 206             if (epoll_ctl(kdpfd, EPOLL_CTL_ADD, client, &ev) < 0) {
 207                 fprintf(stderr, "epoll set insertion error: fd=%d\\n",
 208                         client);
 209                 return \-1;
 210             }
 211         } else {
 212             do_use_fd(events[n].data.fd);
 213         }
 214     }
 215 }
 216 .fi
 217
 218 When used as an Edge triggered interface, for performance reasons, it is
 219 possible to add the file descriptor inside the epoll interface (
 220 .B EPOLL_CTL_ADD
 221 ) once by specifying (
 222 .BR EPOLLIN | EPOLLOUT
 223 ).
 224 This allows you to avoid
 225 continuously switching between
 226 .B EPOLLIN
 227 and
 228 .B EPOLLOUT
 229 calling
 230 .BR epoll_ctl (2)
 231 with
 232 .BR EPOLL_CTL_MOD .
 233 .SH QUESTIONS AND ANSWERS
 234 .TP
 235 .B Q1
 236 What happens if you add the same fd to an epoll_set twice?
 237 .TP
 238 .B A1
 239 You will probably get EEXIST.
 240 However, it is possible that two
 241 threads may add the same fd twice.
 242 This is a harmless condition.
 243 .TP
 244 .B Q2
 245 Can two
 246 .B epoll
 247 sets wait for the same fd? If so, are events reported
 248 to both
 249 .B epoll
 250 sets fds?
 251 .TP
 252 .B A2
 253 Yes.
 254 However, it is not recommended.
 255 Yes it would be reported to both.
 256 .TP
 257 .B Q3
 258 Is the
 259 .B epoll
 260 fd itself poll/epoll/selectable?
 261 .TP
 262 .B A3
 263 Yes.
 264 .TP
 265 .B Q4
 266 What happens if the
 267 .B epoll
 268 fd is put into its own fd set?
 269 .TP
 270 .B A4
 271 It will fail.
 272 However, you can add an
 273 .B epoll
 274 fd inside another epoll fd set.
 275 .TP
 276 .B Q5
 277 Can I send the
 278 .B epoll
 279 fd over a unix-socket to another process?
 280 .TP
 281 .B A5
 282 No.
 283 .TP
 284 .B Q6
 285 Will the close of an fd cause it to be removed from all
 286 .B epoll
 287 sets automatically?
 288 .TP
 289 .B A6
 290 Yes.
 291 .TP
 292 .B Q7
 293 If more than one event comes in between
 294 .BR epoll_wait (2)
 295 calls, are they combined or reported separately?
 296 .TP
 297 .B A7
 298 They will be combined.
 299 .TP
 300 .B Q8
 301 Does an operation on an fd affect the already collected but not yet reported
 302 events?
 303 .TP
 304 .B A8
 305 You can do two operations on an existing fd.
 306 Remove would be meaningless for
 307 this case.
 308 Modify will re-read available I/O.
 309 .TP
 310 .B Q9
 311 Do I need to continuously read/write an fd until EAGAIN when using the
 312 .B EPOLLET
 313 flag ( Edge Triggered behaviour ) ?
 314 .TP
 315 .B A9
 316 No you don't.
 317 Receiving an event from
 318 .BR epoll_wait (2)
 319 should suggest to you that such file descriptor is ready
 320 for the requested I/O operation.
 321 You have simply to consider it ready until you will receive the
 322 next EAGAIN.
 323 When and how you will use such file descriptor is entirely up
 324 to you.
 325 Also, the condition that the read/write I/O space is exhausted can
 326 be detected by checking the amount of data read/write from/to the target
 327 file descriptor.
 328 For example, if you call
 329 .BR read (2)
 330 by asking to read a certain amount of data and
 331 .BR read (2)
 332 returns a lower number of bytes, you can be sure to have exhausted the read
 333 I/O space for such file descriptor.
 334 Same is valid when writing using the
 335 .BR write (2)
 336 function.
 337 .SH POSSIBLE PITFALLS AND WAYS TO AVOID THEM
 338 .TP
 339 .B o Starvation ( Edge Triggered )
 340 .PP
 341 If there is a large amount of I/O space,
 342 it is possible that by trying to drain
 343 it the other files will not get processed causing starvation.
 344 This is not specific to
 345 .BR epoll .
 346 .PP
 347 The solution is to maintain a ready list
 348 and mark the file descriptor as ready
 349 in its associated data structure, thereby allowing the application to
 350 remember which files need to be processed but still round robin amongst
 351 all the ready files.
 352 This also supports ignoring subsequent events you
 353 receive for fd's that are already ready.
 354 .TP
 355 .B o If using an event cache...
 356 .PP
 357 If you use an event cache or store all the fd's returned from
 358 .BR epoll_wait (2),
 359 then make sure to provide a way to mark
 360 its closure dynamically (ie- caused by
 361 a previous event's processing).
 362 Suppose you receive 100 events from
 363 .BR epoll_wait (2),
 364 and in event #47 a condition causes event #13 to be closed.
 365 If you remove the structure and
 366 .BR close ()
 367 the fd for event #13, then your
 368 event cache might still say there are events waiting for that fd causing
 369 confusion.
 370 .PP
 371 One solution for this is to call, during the processing of event 47,
 372 .BR epoll_ctl ( EPOLL_CTL_DEL )
 373 to delete fd 13 and
 374 .BR close (),
 375 then mark its associated
 376 data structure as removed and link it to a cleanup list.
 377 If you find another
 378 event for fd 13 in your batch processing, you will discover the fd had been
 379 previously removed and there will be no confusion.
 380 .SH CONFORMING TO
 381 The epoll API is Linux specific.
 382 Some other systems provide similar
 383 mechanisms, e.g., FreeBSD has
 384 .IR kqueue ,
 385 and Solaris has
 386 .IR /dev/poll .
 387 .SH VERSIONS
 388 .BR epoll (7)
 389 is a new API introduced in Linux kernel 2.5.44.
 390 Its interface should be finalized in Linux kernel 2.5.66.
 391 .SH "SEE ALSO"
 392 .BR epoll_create (2),
 393 .BR epoll_ctl (2),
 394 .BR epoll_wait (2)