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 1.8 by root, Thu Nov 1 11:11:22 2007 UTC vs.
Revision 1.36 by root, Mon Oct 27 11:08:29 2008 UTC

1/* 1/*
2 * libev epoll fd activity backend
3 *
2 * Copyright (c) 2007 Marc Alexander Lehmann <libev@schmorp.de> 4 * Copyright (c) 2007,2008 Marc Alexander Lehmann <libev@schmorp.de>
3 * All rights reserved. 5 * All rights reserved.
4 * 6 *
5 * Redistribution and use in source and binary forms, with or without 7 * Redistribution and use in source and binary forms, with or without modifica-
6 * modification, are permitted provided that the following conditions are 8 * tion, are permitted provided that the following conditions are met:
7 * met:
8 * 9 *
9 * * 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
10 * 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.
11 * 16 *
12 * * Redistributions in binary form must reproduce the above 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED
13 * copyright notice, this list of conditions and the following 18 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER-
14 * disclaimer in the documentation and/or other materials provided 19 * CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO
15 * with the distribution. 20 * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE-
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
18 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
19 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
20 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
21 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
22 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 21 * CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
23 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 22 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
24 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 23 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
25 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH-
26 * (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
27 * 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.
28 */ 38 */
29 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.
56 */
57
30#include <sys/epoll.h> 58#include <sys/epoll.h>
31 59
32static int epoll_fd = -1; 60void inline_size
61unsigned_char_init (unsigned char *base, int count)
62{
63 /* memset might be overkill */
64 while (count--)
65 *base++ = 0;
66}
33 67
34static void 68static void
35epoll_modify (int fd, int oev, int nev) 69epoll_modify (EV_P_ int fd, int oev, int nev)
36{ 70{
37 int mode = nev ? oev ? EPOLL_CTL_MOD : EPOLL_CTL_ADD : EPOLL_CTL_DEL;
38
39 struct epoll_event ev; 71 struct epoll_event ev;
40 ev.data.fd = fd; 72 unsigned char oldmask;
41 ev.events = 73
74 /*
75 * we handle EPOLL_CTL_DEL by ignoring it here
76 * on the assumption that the fd is gone anyways
77 * if that is wrong, we have to handle the spurious
78 * event in epoll_poll.
79 * the fd is later added, we try to ADD it, and, if that
80 * fails, we assume it still has the same eventmask.
81 */
82 if (!nev)
83 return;
84
85 oldmask = anfds [fd].emask;
86 anfds [fd].emask = nev;
87
88 /* store the generation counter in the upper 32 bits */
89 ev.data.u64 = fd | ((uint64_t)++anfds [fd].egen << 32);
42 (nev & EV_READ ? EPOLLIN : 0) 90 ev.events = (nev & EV_READ ? EPOLLIN : 0)
43 | (nev & EV_WRITE ? EPOLLOUT : 0); 91 | (nev & EV_WRITE ? EPOLLOUT : 0);
44 92
45 epoll_ctl (epoll_fd, mode, fd, &ev); 93 if (expect_true (!epoll_ctl (backend_fd, oev ? EPOLL_CTL_MOD : EPOLL_CTL_ADD, fd, &ev)))
94 return;
95
96 if (expect_true (errno == ENOENT))
97 {
98 /* if ENOENT then the fd went away, so try to do the right thing */
99 if (!nev)
100 goto dec_egen;
101
102 if (!epoll_ctl (backend_fd, EPOLL_CTL_ADD, fd, &ev))
103 return;
104 }
105 else if (expect_true (errno == EEXIST))
106 {
107 /* EEXIST means we ignored a previous DEL, but the fd is still active */
108 /* if the kernel mask is the same as the new mask, we assume it hasn't changed */
109 if (oldmask == nev)
110 goto dec_egen;
111
112 if (!epoll_ctl (backend_fd, EPOLL_CTL_MOD, fd, &ev))
113 return;
114 }
115
116 fd_kill (EV_A_ fd);
117
118dec_egen:
119 /* we didn't successfully call epoll_ctl, so decrement the generation counter again */
120 --anfds [fd].egen;
46} 121}
47 122
48static void 123static void
49epoll_postfork_child (void)
50{
51 int fd;
52
53 epoll_fd = epoll_create (256);
54 fcntl (epoll_fd, F_SETFD, FD_CLOEXEC);
55
56 /* re-register interest in fds */
57 for (fd = 0; fd < anfdmax; ++fd)
58 if (anfds [fd].events)//D
59 epoll_modify (fd, EV_NONE, anfds [fd].events);
60}
61
62static struct epoll_event *events;
63static int eventmax;
64
65static void
66epoll_poll (ev_tstamp timeout) 124epoll_poll (EV_P_ ev_tstamp timeout)
67{ 125{
68 int eventcnt = epoll_wait (epoll_fd, events, eventmax, ceil (timeout * 1000.));
69 int i; 126 int i;
127 int eventcnt = epoll_wait (backend_fd, epoll_events, epoll_eventmax, (int)ceil (timeout * 1000.));
70 128
71 if (eventcnt < 0) 129 if (expect_false (eventcnt < 0))
130 {
131 if (errno != EINTR)
132 syserr ("(libev) epoll_wait");
133
72 return; 134 return;
135 }
73 136
74 for (i = 0; i < eventcnt; ++i) 137 for (i = 0; i < eventcnt; ++i)
75 fd_event ( 138 {
76 events [i].data.fd, 139 struct epoll_event *ev = epoll_events + i;
140
141 int fd = (uint32_t)ev->data.u64; /* mask out the lower 32 bits */
77 (events [i].events & (EPOLLOUT | EPOLLERR | EPOLLHUP) ? EV_WRITE : 0) 142 int got = (ev->events & (EPOLLOUT | EPOLLERR | EPOLLHUP) ? EV_WRITE : 0)
78 | (events [i].events & (EPOLLIN | EPOLLERR | EPOLLHUP) ? EV_READ : 0) 143 | (ev->events & (EPOLLIN | EPOLLERR | EPOLLHUP) ? EV_READ : 0);
79 ); 144 int want = anfds [fd].events;
145
146 if (anfds [fd].egen != (unsigned char)(ev->data.u64 >> 32))
147 /*fprintf (stderr, "spurious notification fd %d, %d vs %d\n", fd, (int)(ev->data.u64 >> 32), anfds [fd].egen);*/
148 continue;
149
150 if (expect_false (got & ~want))
151 {
152 anfds [fd].emask = want;
153
154 /* we received an event but are not interested in it, try mod or del */
155 /* I don't think we ever need MOD, but let's handle it anyways */
156 ev->events = (want & EV_READ ? EPOLLIN : 0)
157 | (want & EV_WRITE ? EPOLLOUT : 0);
158
159 epoll_ctl (backend_fd, want ? EPOLL_CTL_MOD : EPOLL_CTL_DEL, fd, ev);
160 }
161
162 fd_event (EV_A_ fd, got);
163 }
80 164
81 /* if the receive array was full, increase its size */ 165 /* if the receive array was full, increase its size */
82 if (eventcnt == eventmax) 166 if (expect_false (eventcnt == epoll_eventmax))
83 { 167 {
84 free (events); 168 ev_free (epoll_events);
85 eventmax += eventmax >> 1; 169 epoll_eventmax = array_nextsize (sizeof (struct epoll_event), epoll_eventmax, epoll_eventmax + 1);
86 events = malloc (sizeof (struct epoll_event) * eventmax); 170 epoll_events = (struct epoll_event *)ev_malloc (sizeof (struct epoll_event) * epoll_eventmax);
87 } 171 }
88} 172}
89 173
90static void 174int inline_size
91epoll_init (int flags) 175epoll_init (EV_P_ int flags)
92{ 176{
93 epoll_fd = epoll_create (256); 177 backend_fd = epoll_create (256);
94 178
95 if (epoll_fd < 0) 179 if (backend_fd < 0)
96 return; 180 return 0;
97 181
98 fcntl (epoll_fd, F_SETFD, FD_CLOEXEC); 182 fcntl (backend_fd, F_SETFD, FD_CLOEXEC);
99 183
100 ev_method = EVMETHOD_EPOLL; 184 backend_fudge = 0.; /* kernel sources seem to indicate this to be zero */
101 method_fudge = 1e-3; /* needed to compensate for epoll returning early */
102 method_modify = epoll_modify; 185 backend_modify = epoll_modify;
103 method_poll = epoll_poll; 186 backend_poll = epoll_poll;
104 187
105 eventmax = 64; /* intiial number of events receivable per poll */ 188 epoll_eventmax = 64; /* intiial number of events receivable per poll */
106 events = malloc (sizeof (struct epoll_event) * eventmax); 189 epoll_events = (struct epoll_event *)ev_malloc (sizeof (struct epoll_event) * epoll_eventmax);
107}
108 190
191 return EVBACKEND_EPOLL;
192}
193
194void inline_size
195epoll_destroy (EV_P)
196{
197 ev_free (epoll_events);
198}
199
200void inline_size
201epoll_fork (EV_P)
202{
203 close (backend_fd);
204
205 while ((backend_fd = epoll_create (256)) < 0)
206 syserr ("(libev) epoll_create");
207
208 fcntl (backend_fd, F_SETFD, FD_CLOEXEC);
209
210 fd_rearm_all (EV_A);
211}
212

Diff Legend

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