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 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. |
30 | */ |
38 | */ |
31 | |
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. |
|
|
58 | */ |
|
|
59 | |
32 | #include <sys/epoll.h> |
60 | #include <sys/epoll.h> |
33 | |
61 | |
34 | static int epoll_fd = -1; |
|
|
35 | |
|
|
36 | static void |
62 | static void |
37 | epoll_modify (int fd, int oev, int nev) |
63 | epoll_modify (EV_P_ int fd, int oev, int nev) |
38 | { |
64 | { |
39 | int mode = nev ? oev ? EPOLL_CTL_MOD : EPOLL_CTL_ADD : EPOLL_CTL_DEL; |
|
|
40 | |
|
|
41 | struct epoll_event ev; |
65 | struct epoll_event ev; |
42 | ev.data.u64 = fd; /* use u64 to fully initialise the struct, for nicer strace etc. */ |
66 | unsigned char oldmask; |
43 | 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 | * the fd is later added, 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); |
44 | (nev & EV_READ ? EPOLLIN : 0) |
85 | ev.events = (nev & EV_READ ? EPOLLIN : 0) |
45 | | (nev & EV_WRITE ? EPOLLOUT : 0); |
86 | | (nev & EV_WRITE ? EPOLLOUT : 0); |
46 | |
87 | |
47 | epoll_ctl (epoll_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 | { |
|
|
93 | /* if ENOENT then the fd went away, so try to do the right thing */ |
|
|
94 | if (!nev) |
|
|
95 | goto dec_egen; |
|
|
96 | |
|
|
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 | |
|
|
111 | fd_kill (EV_A_ fd); |
|
|
112 | |
|
|
113 | dec_egen: |
|
|
114 | /* we didn't successfully call epoll_ctl, so decrement the generation counter again */ |
|
|
115 | --anfds [fd].egen; |
48 | } |
116 | } |
49 | |
117 | |
50 | static void |
118 | static void |
51 | epoll_postfork_child (void) |
|
|
52 | { |
|
|
53 | int fd; |
|
|
54 | |
|
|
55 | epoll_fd = epoll_create (256); |
|
|
56 | fcntl (epoll_fd, F_SETFD, FD_CLOEXEC); |
|
|
57 | |
|
|
58 | /* re-register interest in fds */ |
|
|
59 | for (fd = 0; fd < anfdmax; ++fd) |
|
|
60 | if (anfds [fd].events)//D |
|
|
61 | epoll_modify (fd, EV_NONE, anfds [fd].events); |
|
|
62 | } |
|
|
63 | |
|
|
64 | static struct epoll_event *events; |
|
|
65 | static int eventmax; |
|
|
66 | |
|
|
67 | static void |
|
|
68 | epoll_poll (ev_tstamp timeout) |
119 | epoll_poll (EV_P_ ev_tstamp timeout) |
69 | { |
120 | { |
70 | int eventcnt = epoll_wait (epoll_fd, events, eventmax, ceil (timeout * 1000.)); |
|
|
71 | int i; |
121 | int i; |
|
|
122 | int eventcnt = epoll_wait (backend_fd, epoll_events, epoll_eventmax, (int)ceil (timeout * 1000.)); |
72 | |
123 | |
73 | if (eventcnt < 0) |
124 | if (expect_false (eventcnt < 0)) |
|
|
125 | { |
|
|
126 | if (errno != EINTR) |
|
|
127 | ev_syserr ("(libev) epoll_wait"); |
|
|
128 | |
74 | return; |
129 | return; |
|
|
130 | } |
75 | |
131 | |
76 | for (i = 0; i < eventcnt; ++i) |
132 | for (i = 0; i < eventcnt; ++i) |
77 | fd_event ( |
133 | { |
78 | events [i].data.fd, |
134 | struct epoll_event *ev = epoll_events + i; |
|
|
135 | |
|
|
136 | int fd = (uint32_t)ev->data.u64; /* mask out the lower 32 bits */ |
|
|
137 | int want = anfds [fd].events; |
79 | (events [i].events & (EPOLLOUT | EPOLLERR | EPOLLHUP) ? EV_WRITE : 0) |
138 | int got = (ev->events & (EPOLLOUT | EPOLLERR | EPOLLHUP) ? EV_WRITE : 0) |
80 | | (events [i].events & (EPOLLIN | EPOLLERR | EPOLLHUP) ? EV_READ : 0) |
139 | | (ev->events & (EPOLLIN | EPOLLERR | EPOLLHUP) ? EV_READ : 0); |
81 | ); |
140 | |
|
|
141 | /* check for spurious notification */ |
|
|
142 | if (expect_false ((uint32_t)anfds [fd].egen != (uint32_t)(ev->data.u64 >> 32))) |
|
|
143 | { |
|
|
144 | /* recreate kernel state */ |
|
|
145 | postfork = 1; |
|
|
146 | continue; |
|
|
147 | } |
|
|
148 | |
|
|
149 | if (expect_false (got & ~want)) |
|
|
150 | { |
|
|
151 | anfds [fd].emask = want; |
|
|
152 | |
|
|
153 | /* we received an event but are not interested in it, try mod or del */ |
|
|
154 | /* I don't think we ever need MOD, but let's handle it anyways */ |
|
|
155 | ev->events = (want & EV_READ ? EPOLLIN : 0) |
|
|
156 | | (want & EV_WRITE ? EPOLLOUT : 0); |
|
|
157 | |
|
|
158 | if (epoll_ctl (backend_fd, want ? EPOLL_CTL_MOD : EPOLL_CTL_DEL, fd, ev)) |
|
|
159 | { |
|
|
160 | postfork = 1; /* an error occured, recreate kernel state */ |
|
|
161 | continue; |
|
|
162 | } |
|
|
163 | } |
|
|
164 | |
|
|
165 | fd_event (EV_A_ fd, got); |
|
|
166 | } |
82 | |
167 | |
83 | /* if the receive array was full, increase its size */ |
168 | /* if the receive array was full, increase its size */ |
84 | if (expect_false (eventcnt == eventmax)) |
169 | if (expect_false (eventcnt == epoll_eventmax)) |
85 | { |
170 | { |
86 | free (events); |
171 | ev_free (epoll_events); |
87 | eventmax = array_roundsize (events, eventmax << 1); |
172 | epoll_eventmax = array_nextsize (sizeof (struct epoll_event), epoll_eventmax, epoll_eventmax + 1); |
88 | events = malloc (sizeof (struct epoll_event) * eventmax); |
173 | epoll_events = (struct epoll_event *)ev_malloc (sizeof (struct epoll_event) * epoll_eventmax); |
89 | } |
174 | } |
90 | } |
175 | } |
91 | |
176 | |
92 | static void |
177 | int inline_size |
93 | epoll_init (int flags) |
178 | epoll_init (EV_P_ int flags) |
94 | { |
179 | { |
95 | epoll_fd = epoll_create (256); |
180 | backend_fd = epoll_create (256); |
96 | |
181 | |
97 | if (epoll_fd < 0) |
182 | if (backend_fd < 0) |
98 | return; |
183 | return 0; |
99 | |
184 | |
100 | fcntl (epoll_fd, F_SETFD, FD_CLOEXEC); |
185 | fcntl (backend_fd, F_SETFD, FD_CLOEXEC); |
101 | |
186 | |
102 | ev_method = EVMETHOD_EPOLL; |
187 | backend_fudge = 0.; /* kernel sources seem to indicate this to be zero */ |
103 | method_fudge = 1e-3; /* needed to compensate for epoll returning early */ |
|
|
104 | method_modify = epoll_modify; |
188 | backend_modify = epoll_modify; |
105 | method_poll = epoll_poll; |
189 | backend_poll = epoll_poll; |
106 | |
190 | |
107 | eventmax = 64; /* intiial number of events receivable per poll */ |
191 | epoll_eventmax = 64; /* initial number of events receivable per poll */ |
108 | events = malloc (sizeof (struct epoll_event) * eventmax); |
192 | epoll_events = (struct epoll_event *)ev_malloc (sizeof (struct epoll_event) * epoll_eventmax); |
109 | } |
|
|
110 | |
193 | |
|
|
194 | return EVBACKEND_EPOLL; |
|
|
195 | } |
|
|
196 | |
|
|
197 | void inline_size |
|
|
198 | epoll_destroy (EV_P) |
|
|
199 | { |
|
|
200 | ev_free (epoll_events); |
|
|
201 | } |
|
|
202 | |
|
|
203 | void inline_size |
|
|
204 | epoll_fork (EV_P) |
|
|
205 | { |
|
|
206 | close (backend_fd); |
|
|
207 | |
|
|
208 | while ((backend_fd = epoll_create (256)) < 0) |
|
|
209 | ev_syserr ("(libev) epoll_create"); |
|
|
210 | |
|
|
211 | fcntl (backend_fd, F_SETFD, FD_CLOEXEC); |
|
|
212 | |
|
|
213 | fd_rearm_all (EV_A); |
|
|
214 | } |
|
|
215 | |