ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libev/ev_epoll.c
(Generate patch)

Comparing libev/ev_epoll.c (file contents):
Revision 1.28 by root, Wed Dec 5 13:54:36 2007 UTC vs.
Revision 1.49 by root, Tue Nov 24 06:37:23 2009 UTC

1/* 1/*
2 * libev epoll fd activity backend 2 * libev epoll fd activity backend
3 * 3 *
4 * Copyright (c) 2007 Marc Alexander Lehmann <libev@schmorp.de> 4 * Copyright (c) 2007,2008,2009 Marc Alexander Lehmann <libev@schmorp.de>
5 * All rights reserved. 5 * All rights reserved.
6 * 6 *
7 * Redistribution and use in source and binary forms, with or without 7 * Redistribution and use in source and binary forms, with or without modifica-
8 * modification, are permitted provided that the following conditions are 8 * tion, are permitted provided that the following conditions are met:
9 * met:
10 * 9 *
11 * * Redistributions of source code must retain the above copyright 10 * 1. Redistributions of source code must retain the above copyright notice,
11 * this list of conditions and the following disclaimer.
12 *
13 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer. 14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
13 * 16 *
14 * * Redistributions in binary form must reproduce the above 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED
15 * copyright notice, this list of conditions and the following 18 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER-
16 * disclaimer in the documentation and/or other materials provided 19 * CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO
17 * with the distribution. 20 * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE-
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
22 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
23 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 21 * CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
25 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 22 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
26 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 23 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
27 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH-
28 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 25 * ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
29 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 * OF THE POSSIBILITY OF SUCH DAMAGE.
27 *
28 * Alternatively, the contents of this file may be used under the terms of
29 * the GNU General Public License ("GPL") version 2 or any later version,
30 * in which case the provisions of the GPL are applicable instead of
31 * the above. If you wish to allow the use of your version of this file
32 * only under the terms of the GPL and not to allow others to use your
33 * version of this file under the BSD license, indicate your decision
34 * by deleting the provisions above and replace them with the notice
35 * and other provisions required by the GPL. If you do not delete the
36 * provisions above, a recipient may use your version of this file under
37 * either the BSD or the GPL.
38 */
39
40/*
41 * general notes about epoll:
42 *
43 * a) epoll silently removes fds from the fd set. as nothing tells us
44 * that an fd has been removed otherwise, we have to continually
45 * "rearm" fds that we suspect *might* have changed (same
46 * problem with kqueue, but much less costly there).
47 * b) the fact that ADD != MOD creates a lot of extra syscalls due to a)
48 * and seems not to have any advantage.
49 * c) the inability to handle fork or file descriptors (think dup)
50 * limits the applicability over poll, so this is not a generic
51 * poll replacement.
52 *
53 * lots of "weird code" and complication handling in this file is due
54 * to these design problems with epoll, as we try very hard to avoid
55 * epoll_ctl syscalls for common usage patterns and handle the breakage
56 * ensuing from receiving events for closed and otherwise long gone
57 * file descriptors.
30 */ 58 */
31 59
32#include <sys/epoll.h> 60#include <sys/epoll.h>
33 61
34static void 62static void
35epoll_modify (EV_P_ int fd, int oev, int nev) 63epoll_modify (EV_P_ int fd, int oev, int nev)
36{ 64{
37 int mode = nev ? oev ? EPOLL_CTL_MOD : EPOLL_CTL_ADD : EPOLL_CTL_DEL;
38
39 struct epoll_event ev; 65 struct epoll_event ev;
40 ev.data.u64 = fd; /* use u64 to fully initialise the struct, for nicer strace etc. */ 66 unsigned char oldmask;
41 ev.events = 67
68 /*
69 * we handle EPOLL_CTL_DEL by ignoring it here
70 * on the assumption that the fd is gone anyways
71 * if that is wrong, we have to handle the spurious
72 * event in epoll_poll.
73 * if the fd is added again, we try to ADD it, and, if that
74 * fails, we assume it still has the same eventmask.
75 */
76 if (!nev)
77 return;
78
79 oldmask = anfds [fd].emask;
80 anfds [fd].emask = nev;
81
82 /* store the generation counter in the upper 32 bits, the fd in the lower 32 bits */
83 ev.data.u64 = (uint64_t)(uint32_t)fd
84 | ((uint64_t)(uint32_t)++anfds [fd].egen << 32);
42 (nev & EV_READ ? EPOLLIN : 0) 85 ev.events = (nev & EV_READ ? EPOLLIN : 0)
43 | (nev & EV_WRITE ? EPOLLOUT : 0); 86 | (nev & EV_WRITE ? EPOLLOUT : 0);
44 87
45 if (epoll_ctl (backend_fd, mode, fd, &ev)) 88 if (expect_true (!epoll_ctl (backend_fd, oev ? EPOLL_CTL_MOD : EPOLL_CTL_ADD, fd, &ev)))
89 return;
90
91 if (expect_true (errno == ENOENT))
92 {
46 if (errno != ENOENT /* on ENOENT the fd went away, so try to do the right thing */ 93 /* if ENOENT then the fd went away, so try to do the right thing */
94 if (!nev)
95 goto dec_egen;
96
47 || (nev && epoll_ctl (backend_fd, EPOLL_CTL_ADD, fd, &ev))) 97 if (!epoll_ctl (backend_fd, EPOLL_CTL_ADD, fd, &ev))
98 return;
99 }
100 else if (expect_true (errno == EEXIST))
101 {
102 /* EEXIST means we ignored a previous DEL, but the fd is still active */
103 /* if the kernel mask is the same as the new mask, we assume it hasn't changed */
104 if (oldmask == nev)
105 goto dec_egen;
106
107 if (!epoll_ctl (backend_fd, EPOLL_CTL_MOD, fd, &ev))
108 return;
109 }
110
48 fd_kill (EV_A_ fd); 111 fd_kill (EV_A_ fd);
112
113dec_egen:
114 /* we didn't successfully call epoll_ctl, so decrement the generation counter again */
115 --anfds [fd].egen;
49} 116}
50 117
51static void 118static void
52epoll_poll (EV_P_ ev_tstamp timeout) 119epoll_poll (EV_P_ ev_tstamp timeout)
53{ 120{
54 int i; 121 int i;
122 int eventcnt;
123
124 /* epoll wait times cannot be larger than (LONG_MAX - 999UL) / HZ msecs, which is below */
125 /* the default libev max wait time, however. */
126 EV_RELEASE_CB;
55 int eventcnt = epoll_wait (backend_fd, epoll_events, epoll_eventmax, (int)ceil (timeout * 1000.)); 127 eventcnt = epoll_wait (backend_fd, epoll_events, epoll_eventmax, (int)ceil (timeout * 1000.));
128 EV_ACQUIRE_CB;
56 129
57 if (eventcnt < 0) 130 if (expect_false (eventcnt < 0))
58 { 131 {
59 if (errno != EINTR) 132 if (errno != EINTR)
60 syserr ("(libev) epoll_wait"); 133 ev_syserr ("(libev) epoll_wait");
61 134
62 return; 135 return;
63 } 136 }
64 137
65 for (i = 0; i < eventcnt; ++i) 138 for (i = 0; i < eventcnt; ++i)
66 fd_event ( 139 {
67 EV_A_ 140 struct epoll_event *ev = epoll_events + i;
68 epoll_events [i].data.u64, 141
142 int fd = (uint32_t)ev->data.u64; /* mask out the lower 32 bits */
143 int want = anfds [fd].events;
69 (epoll_events [i].events & (EPOLLOUT | EPOLLERR | EPOLLHUP) ? EV_WRITE : 0) 144 int got = (ev->events & (EPOLLOUT | EPOLLERR | EPOLLHUP) ? EV_WRITE : 0)
70 | (epoll_events [i].events & (EPOLLIN | EPOLLERR | EPOLLHUP) ? EV_READ : 0) 145 | (ev->events & (EPOLLIN | EPOLLERR | EPOLLHUP) ? EV_READ : 0);
71 ); 146
147 /* check for spurious notification */
148 if (expect_false ((uint32_t)anfds [fd].egen != (uint32_t)(ev->data.u64 >> 32)))
149 {
150 /* recreate kernel state */
151 postfork = 1;
152 continue;
153 }
154
155 if (expect_false (got & ~want))
156 {
157 anfds [fd].emask = want;
158
159 /* we received an event but are not interested in it, try mod or del */
160 /* I don't think we ever need MOD, but let's handle it anyways */
161 ev->events = (want & EV_READ ? EPOLLIN : 0)
162 | (want & EV_WRITE ? EPOLLOUT : 0);
163
164 /* pre-2.6.9 kernels require a non-null pointer with EPOLL_CTL_DEL, */
165 /* which is fortunately easy to do for us. */
166 if (epoll_ctl (backend_fd, want ? EPOLL_CTL_MOD : EPOLL_CTL_DEL, fd, ev))
167 {
168 postfork = 1; /* an error occured, recreate kernel state */
169 continue;
170 }
171 }
172
173 fd_event (EV_A_ fd, got);
174 }
72 175
73 /* if the receive array was full, increase its size */ 176 /* if the receive array was full, increase its size */
74 if (expect_false (eventcnt == epoll_eventmax)) 177 if (expect_false (eventcnt == epoll_eventmax))
75 { 178 {
76 ev_free (epoll_events); 179 ev_free (epoll_events);
… …
80} 183}
81 184
82int inline_size 185int inline_size
83epoll_init (EV_P_ int flags) 186epoll_init (EV_P_ int flags)
84{ 187{
188#ifdef EPOLL_CLOEXEC
189 backend_fd = epoll_create1 (EPOLL_CLOEXEC);
190
191 if (backend_fd <= 0)
192#endif
85 backend_fd = epoll_create (256); 193 backend_fd = epoll_create (256);
86 194
87 if (backend_fd < 0) 195 if (backend_fd < 0)
88 return 0; 196 return 0;
89 197
90 fcntl (backend_fd, F_SETFD, FD_CLOEXEC); 198 fcntl (backend_fd, F_SETFD, FD_CLOEXEC);
91 199
92 backend_fudge = 1e-3; /* needed to compensate for epoll returning early */ 200 backend_fudge = 0.; /* kernel sources seem to indicate this to be zero */
93 backend_modify = epoll_modify; 201 backend_modify = epoll_modify;
94 backend_poll = epoll_poll; 202 backend_poll = epoll_poll;
95 203
96 epoll_eventmax = 64; /* intiial number of events receivable per poll */ 204 epoll_eventmax = 64; /* initial number of events receivable per poll */
97 epoll_events = (struct epoll_event *)ev_malloc (sizeof (struct epoll_event) * epoll_eventmax); 205 epoll_events = (struct epoll_event *)ev_malloc (sizeof (struct epoll_event) * epoll_eventmax);
98 206
99 return EVBACKEND_EPOLL; 207 return EVBACKEND_EPOLL;
100} 208}
101 209
… …
109epoll_fork (EV_P) 217epoll_fork (EV_P)
110{ 218{
111 close (backend_fd); 219 close (backend_fd);
112 220
113 while ((backend_fd = epoll_create (256)) < 0) 221 while ((backend_fd = epoll_create (256)) < 0)
114 syserr ("(libev) epoll_create"); 222 ev_syserr ("(libev) epoll_create");
115 223
116 fcntl (backend_fd, F_SETFD, FD_CLOEXEC); 224 fcntl (backend_fd, F_SETFD, FD_CLOEXEC);
117 225
118 fd_rearm_all (EV_A); 226 fd_rearm_all (EV_A);
119} 227}

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines