1 | /* |
1 | /* |
|
|
2 | * libev epoll fd activity backend |
|
|
3 | * |
2 | * Copyright (c) 2007 Marc Alexander Lehmann <libev@schmorp.de> |
4 | * Copyright (c) 2007,2008,2009,2010 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 | * d) epoll doesn't work the same as select with many file descriptors |
|
|
53 | * (such as files). while not critical, no other advanced interface |
|
|
54 | * seems to share this (rather non-unixy) limitation. |
|
|
55 | * e) epoll claims to be embeddable, but in practise you never get |
|
|
56 | * a ready event for the epoll fd. |
|
|
57 | * |
|
|
58 | * lots of "weird code" and complication handling in this file is due |
|
|
59 | * to these design problems with epoll, as we try very hard to avoid |
|
|
60 | * epoll_ctl syscalls for common usage patterns and handle the breakage |
|
|
61 | * ensuing from receiving events for closed and otherwise long gone |
|
|
62 | * file descriptors. |
|
|
63 | */ |
|
|
64 | |
30 | #include <sys/epoll.h> |
65 | #include <sys/epoll.h> |
31 | |
66 | |
32 | static int epoll_fd = -1; |
|
|
33 | |
|
|
34 | static void |
67 | static void |
35 | epoll_modify (int fd, int oev, int nev) |
68 | epoll_modify (EV_P_ int fd, int oev, int nev) |
36 | { |
69 | { |
37 | int mode = nev ? oev ? EPOLL_CTL_MOD : EPOLL_CTL_ADD : EPOLL_CTL_DEL; |
|
|
38 | |
|
|
39 | struct epoll_event ev; |
70 | struct epoll_event ev; |
40 | ev.data.fd = fd; |
71 | unsigned char oldmask; |
41 | ev.events = |
72 | |
|
|
73 | /* |
|
|
74 | * we handle EPOLL_CTL_DEL by ignoring it here |
|
|
75 | * on the assumption that the fd is gone anyways |
|
|
76 | * if that is wrong, we have to handle the spurious |
|
|
77 | * event in epoll_poll. |
|
|
78 | * if the fd is added again, we try to ADD it, and, if that |
|
|
79 | * fails, we assume it still has the same eventmask. |
|
|
80 | */ |
|
|
81 | if (!nev) |
|
|
82 | return; |
|
|
83 | |
|
|
84 | oldmask = anfds [fd].emask; |
|
|
85 | anfds [fd].emask = nev; |
|
|
86 | |
|
|
87 | /* store the generation counter in the upper 32 bits, the fd in the lower 32 bits */ |
|
|
88 | ev.data.u64 = (uint64_t)(uint32_t)fd |
|
|
89 | | ((uint64_t)(uint32_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 && oldmask != nev ? EPOLL_CTL_MOD : EPOLL_CTL_ADD, fd, &ev))) |
46 | } |
94 | return; |
47 | |
95 | |
48 | void epoll_postfork_child (void) |
96 | if (expect_true (errno == ENOENT)) |
49 | { |
97 | { |
50 | int fd; |
98 | /* if ENOENT then the fd went away, so try to do the right thing */ |
|
|
99 | if (!nev) |
|
|
100 | goto dec_egen; |
51 | |
101 | |
52 | epoll_fd = epoll_create (256); |
102 | if (!epoll_ctl (backend_fd, EPOLL_CTL_ADD, fd, &ev)) |
53 | fcntl (epoll_fd, F_SETFD, FD_CLOEXEC); |
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; |
54 | |
111 | |
55 | /* re-register interest in fds */ |
112 | if (!epoll_ctl (backend_fd, EPOLL_CTL_MOD, fd, &ev)) |
56 | for (fd = 0; fd < anfdmax; ++fd) |
113 | return; |
57 | if (anfds [fd].wev) |
114 | } |
58 | epoll_modify (fd, EV_NONE, anfds [fd].wev); |
|
|
59 | } |
|
|
60 | |
115 | |
61 | static struct epoll_event *events; |
116 | fd_kill (EV_A_ fd); |
62 | static int eventmax; |
|
|
63 | |
117 | |
|
|
118 | dec_egen: |
|
|
119 | /* we didn't successfully call epoll_ctl, so decrement the generation counter again */ |
|
|
120 | --anfds [fd].egen; |
|
|
121 | } |
|
|
122 | |
|
|
123 | static void |
64 | static void epoll_poll (ev_tstamp timeout) |
124 | epoll_poll (EV_P_ ev_tstamp timeout) |
65 | { |
125 | { |
66 | int eventcnt = epoll_wait (epoll_fd, events, eventmax, ceil (timeout * 1000.)); |
|
|
67 | int i; |
126 | int i; |
|
|
127 | int eventcnt; |
|
|
128 | |
|
|
129 | /* epoll wait times cannot be larger than (LONG_MAX - 999UL) / HZ msecs, which is below */ |
|
|
130 | /* the default libev max wait time, however. */ |
|
|
131 | EV_RELEASE_CB; |
|
|
132 | eventcnt = epoll_wait (backend_fd, epoll_events, epoll_eventmax, (int)ceil (timeout * 1000.)); |
|
|
133 | EV_ACQUIRE_CB; |
68 | |
134 | |
69 | if (eventcnt < 0) |
135 | if (expect_false (eventcnt < 0)) |
|
|
136 | { |
|
|
137 | if (errno != EINTR) |
|
|
138 | ev_syserr ("(libev) epoll_wait"); |
|
|
139 | |
70 | return; |
140 | return; |
|
|
141 | } |
71 | |
142 | |
72 | for (i = 0; i < eventcnt; ++i) |
143 | for (i = 0; i < eventcnt; ++i) |
73 | fd_event ( |
144 | { |
74 | events [i].data.fd, |
145 | struct epoll_event *ev = epoll_events + i; |
|
|
146 | |
|
|
147 | int fd = (uint32_t)ev->data.u64; /* mask out the lower 32 bits */ |
|
|
148 | int want = anfds [fd].events; |
75 | (events [i].events & (EPOLLOUT | EPOLLERR | EPOLLHUP) ? EV_WRITE : 0) |
149 | int got = (ev->events & (EPOLLOUT | EPOLLERR | EPOLLHUP) ? EV_WRITE : 0) |
76 | | (events [i].events & (EPOLLIN | EPOLLERR | EPOLLHUP) ? EV_READ : 0) |
150 | | (ev->events & (EPOLLIN | EPOLLERR | EPOLLHUP) ? EV_READ : 0); |
77 | ); |
151 | |
|
|
152 | /* check for spurious notification */ |
|
|
153 | /* we assume that fd is always in range, as we never shrink the anfds arraay */ |
|
|
154 | if (expect_false ((uint32_t)anfds [fd].egen != (uint32_t)(ev->data.u64 >> 32))) |
|
|
155 | { |
|
|
156 | /* recreate kernel state */ |
|
|
157 | postfork = 1; |
|
|
158 | continue; |
|
|
159 | } |
|
|
160 | |
|
|
161 | if (expect_false (got & ~want)) |
|
|
162 | { |
|
|
163 | anfds [fd].emask = want; |
|
|
164 | |
|
|
165 | /* we received an event but are not interested in it, try mod or del */ |
|
|
166 | /* I don't think we ever need MOD, but let's handle it anyways */ |
|
|
167 | ev->events = (want & EV_READ ? EPOLLIN : 0) |
|
|
168 | | (want & EV_WRITE ? EPOLLOUT : 0); |
|
|
169 | |
|
|
170 | /* pre-2.6.9 kernels require a non-null pointer with EPOLL_CTL_DEL, */ |
|
|
171 | /* which is fortunately easy to do for us. */ |
|
|
172 | if (epoll_ctl (backend_fd, want ? EPOLL_CTL_MOD : EPOLL_CTL_DEL, fd, ev)) |
|
|
173 | { |
|
|
174 | postfork = 1; /* an error occurred, recreate kernel state */ |
|
|
175 | continue; |
|
|
176 | } |
|
|
177 | } |
|
|
178 | |
|
|
179 | fd_event (EV_A_ fd, got); |
|
|
180 | } |
78 | |
181 | |
79 | /* if the receive array was full, increase its size */ |
182 | /* if the receive array was full, increase its size */ |
80 | if (eventcnt == eventmax) |
183 | if (expect_false (eventcnt == epoll_eventmax)) |
81 | { |
184 | { |
82 | free (events); |
185 | ev_free (epoll_events); |
83 | eventmax += eventmax >> 1; |
186 | epoll_eventmax = array_nextsize (sizeof (struct epoll_event), epoll_eventmax, epoll_eventmax + 1); |
84 | events = malloc (sizeof (struct epoll_event) * eventmax); |
187 | epoll_events = (struct epoll_event *)ev_malloc (sizeof (struct epoll_event) * epoll_eventmax); |
85 | } |
188 | } |
86 | } |
189 | } |
87 | |
190 | |
|
|
191 | int inline_size |
88 | void epoll_init (int flags) |
192 | epoll_init (EV_P_ int flags) |
89 | { |
193 | { |
|
|
194 | #ifdef EPOLL_CLOEXEC |
|
|
195 | backend_fd = epoll_create1 (EPOLL_CLOEXEC); |
|
|
196 | |
|
|
197 | if (backend_fd <= 0) |
|
|
198 | #endif |
90 | epoll_fd = epoll_create (256); |
199 | backend_fd = epoll_create (256); |
91 | |
200 | |
92 | if (epoll_fd < 0) |
201 | if (backend_fd < 0) |
93 | return; |
202 | return 0; |
94 | |
203 | |
95 | fcntl (epoll_fd, F_SETFD, FD_CLOEXEC); |
204 | fcntl (backend_fd, F_SETFD, FD_CLOEXEC); |
96 | |
205 | |
97 | ev_method = EVMETHOD_EPOLL; |
206 | backend_fudge = 0.; /* kernel sources seem to indicate this to be zero */ |
98 | method_fudge = 1e-3; /* needed to compensate for epoll returning early */ |
|
|
99 | method_modify = epoll_modify; |
207 | backend_modify = epoll_modify; |
100 | method_poll = epoll_poll; |
208 | backend_poll = epoll_poll; |
101 | |
209 | |
102 | eventmax = 64; /* intiial number of events receivable per poll */ |
210 | epoll_eventmax = 64; /* initial number of events receivable per poll */ |
103 | events = malloc (sizeof (struct epoll_event) * eventmax); |
211 | epoll_events = (struct epoll_event *)ev_malloc (sizeof (struct epoll_event) * epoll_eventmax); |
|
|
212 | |
|
|
213 | return EVBACKEND_EPOLL; |
104 | } |
214 | } |
|
|
215 | |
|
|
216 | void inline_size |
|
|
217 | epoll_destroy (EV_P) |
|
|
218 | { |
|
|
219 | ev_free (epoll_events); |
|
|
220 | } |
|
|
221 | |
|
|
222 | void inline_size |
|
|
223 | epoll_fork (EV_P) |
|
|
224 | { |
|
|
225 | close (backend_fd); |
|
|
226 | |
|
|
227 | while ((backend_fd = epoll_create (256)) < 0) |
|
|
228 | ev_syserr ("(libev) epoll_create"); |
|
|
229 | |
|
|
230 | fcntl (backend_fd, F_SETFD, FD_CLOEXEC); |
|
|
231 | |
|
|
232 | fd_rearm_all (EV_A); |
|
|
233 | } |
|
|
234 | |