1 | /* |
1 | /* |
|
|
2 | * libev kqueue backend |
|
|
3 | * |
2 | * Copyright 2007 Marc Alexander Lehmann <libev@schmorp.de> |
4 | * Copyright (c) 2007 Marc Alexander Lehmann <libev@schmorp.de> |
3 | * Copyright 2000-2002 Niels Provos <provos@citi.umich.edu> |
|
|
4 | * All rights reserved. |
5 | * All rights reserved. |
5 | * |
6 | * |
6 | * Redistribution and use in source and binary forms, with or without |
7 | * Redistribution and use in source and binary forms, with or without |
7 | * modification, are permitted provided that the following conditions |
8 | * modification, are permitted provided that the following conditions are |
8 | * are met: |
9 | * met: |
9 | * |
10 | * |
10 | * 1. Redistributions of source code must retain the above copyright |
11 | * * Redistributions of source code must retain the above copyright |
11 | * notice, this list of conditions and the following disclaimer. |
12 | * notice, this list of conditions and the following disclaimer. |
12 | * 2. Redistributions in binary form must reproduce the above copyright |
|
|
13 | * notice, this list of conditions and the following disclaimer in the |
|
|
14 | * documentation and/or other materials provided with the distribution. |
|
|
15 | * 3. The name of the author may not be used to endorse or promote products |
|
|
16 | * derived from this software without specific prior written permission. |
|
|
17 | * |
13 | * |
18 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR |
14 | * * Redistributions in binary form must reproduce the above |
19 | * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES |
15 | * copyright notice, this list of conditions and the following |
20 | * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. |
16 | * disclaimer in the documentation and/or other materials provided |
21 | * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, |
17 | * with the distribution. |
|
|
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, |
22 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT |
24 | * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT |
23 | * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
25 | * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
24 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
26 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
25 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
27 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
26 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF |
28 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
27 | * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
29 | * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
28 | */ |
30 | */ |
29 | |
31 | |
30 | #include <sys/types.h> |
32 | #include <sys/types.h> |
31 | #include <sys/time.h> |
33 | #include <sys/time.h> |
32 | #include <sys/queue.h> |
34 | #include <sys/queue.h> |
33 | #include <sys/event.h> |
35 | #include <sys/event.h> |
34 | #include <string.h> |
36 | #include <string.h> |
35 | #include <errno.h> |
37 | #include <errno.h> |
36 | |
38 | |
37 | static void |
39 | void inline_speed |
38 | kqueue_change (EV_P_ int fd, int filter, int flags, int fflags) |
40 | kqueue_change (EV_P_ int fd, int filter, int flags, int fflags) |
39 | { |
41 | { |
40 | struct kevent *ke; |
42 | struct kevent *ke; |
41 | |
43 | |
|
|
44 | ++kqueue_changecnt; |
42 | array_needsize (struct kevent, kqueue_changes, kqueue_changemax, ++kqueue_changecnt, ); |
45 | array_needsize (struct kevent, kqueue_changes, kqueue_changemax, kqueue_changecnt, EMPTY2); |
43 | |
46 | |
44 | ke = &kqueue_changes [kqueue_changecnt - 1]; |
47 | EV_SET (&kqueue_changes [kqueue_changecnt - 1], fd, filter, flags, fflags, 0, 0); |
45 | memset (ke, 0, sizeof (struct kevent)); |
|
|
46 | ke->ident = fd; |
|
|
47 | ke->filter = filter; |
|
|
48 | ke->flags = flags; |
|
|
49 | ke->fflags = fflags; |
|
|
50 | } |
48 | } |
51 | |
49 | |
52 | #ifndef NOTE_EOF |
50 | #ifndef NOTE_EOF |
53 | # define NOTE_EOF 0 |
51 | # define NOTE_EOF 0 |
54 | #endif |
52 | #endif |
55 | |
53 | |
56 | static void |
54 | static void |
57 | kqueue_modify (EV_P_ int fd, int oev, int nev) |
55 | kqueue_modify (EV_P_ int fd, int oev, int nev) |
58 | { |
56 | { |
59 | if ((oev ^ nev) & EV_READ) |
57 | if (oev != nev) |
60 | { |
58 | { |
61 | if (nev & EV_READ) |
59 | if (oev & EV_READ) |
62 | kqueue_change (fd, EVFILT_READ, EV_ADD, NOTE_EOF); |
|
|
63 | else |
|
|
64 | kqueue_change (fd, EVFILT_READ, EV_DELETE, 0); |
60 | kqueue_change (EV_A_ fd, EVFILT_READ , EV_DELETE, 0); |
|
|
61 | |
|
|
62 | if (oev & EV_WRITE) |
|
|
63 | kqueue_change (EV_A_ fd, EVFILT_WRITE, EV_DELETE, 0); |
65 | } |
64 | } |
66 | |
65 | |
67 | if ((oev ^ nev) & EV_WRITE) |
66 | /* to detect close/reopen reliably, we have to re-add */ |
68 | { |
67 | /* event requests even when oev == nev */ |
|
|
68 | |
|
|
69 | if (nev & EV_READ) |
|
|
70 | kqueue_change (EV_A_ fd, EVFILT_READ , EV_ADD, NOTE_EOF); |
|
|
71 | |
69 | if (nev & EV_WRITE) |
72 | if (nev & EV_WRITE) |
70 | kqueue_change (fd, EVFILT_WRITE, EV_ADD, NOTE_EOF); |
73 | kqueue_change (EV_A_ fd, EVFILT_WRITE, EV_ADD, NOTE_EOF); |
71 | else |
|
|
72 | kqueue_change (fd, EVFILT_WRITE, EV_DELETE, 0); |
|
|
73 | } |
|
|
74 | } |
74 | } |
75 | |
75 | |
76 | static void |
76 | static void |
77 | kqueue_poll (EV_P_ ev_tstamp timeout) |
77 | kqueue_poll (EV_P_ ev_tstamp timeout) |
78 | { |
78 | { |
79 | int res, i; |
79 | int res, i; |
80 | struct timespec ts; |
80 | struct timespec ts; |
81 | |
81 | |
|
|
82 | /* need to resize so there is enough space for errors */ |
|
|
83 | if (kqueue_changecnt > kqueue_eventmax) |
|
|
84 | { |
|
|
85 | ev_free (kqueue_events); |
|
|
86 | kqueue_eventmax = array_nextsize (sizeof (struct kevent), kqueue_eventmax, kqueue_changecnt); |
|
|
87 | kqueue_events = (struct kevent *)ev_malloc (sizeof (struct kevent) * kqueue_eventmax); |
|
|
88 | } |
|
|
89 | |
82 | ts.tv_sec = (time_t)timeout; |
90 | ts.tv_sec = (time_t)timeout; |
83 | ts.tv_nsec = (long)(timeout - (ev_tstamp)ts.tv_sec) * 1e9; |
91 | ts.tv_nsec = (long)((timeout - (ev_tstamp)ts.tv_sec) * 1e9); |
84 | res = kevent (kqueue_fd, kqueue_changes, kqueue_changecnt, kqueue_events, kqueue_eventmax, &ts); |
92 | res = kevent (backend_fd, kqueue_changes, kqueue_changecnt, kqueue_events, kqueue_eventmax, &ts); |
85 | kqueue_changecnt = 0; |
93 | kqueue_changecnt = 0; |
86 | |
94 | |
87 | if (res < 0) |
95 | if (expect_false (res < 0)) |
88 | { |
96 | { |
89 | if (errno != EINTR) |
97 | if (errno != EINTR) |
90 | syserr ("(libev) kevent"); |
98 | syserr ("(libev) kevent"); |
91 | |
99 | |
92 | return; |
100 | return; |
93 | } |
101 | } |
94 | |
102 | |
95 | for (i = 0; i < res; ++i) |
103 | for (i = 0; i < res; ++i) |
96 | { |
104 | { |
|
|
105 | int fd = kqueue_events [i].ident; |
|
|
106 | |
97 | if (kqueue_events [i].flags & EV_ERROR) |
107 | if (expect_false (kqueue_events [i].flags & EV_ERROR)) |
98 | { |
108 | { |
99 | /* |
109 | int err = kqueue_events [i].data; |
100 | * Error messages that can happen, when a delete fails. |
110 | |
101 | * EBADF happens when the file discriptor has been |
111 | /* we are only interested in errors for fds that we are interested in :) */ |
102 | * closed, |
112 | if (anfds [fd].events) |
103 | * ENOENT when the file discriptor was closed and |
113 | { |
104 | * then reopened. |
114 | if (err == ENOENT) /* resubmit changes on ENOENT */ |
105 | * EINVAL for some reasons not understood; EINVAL |
115 | kqueue_modify (EV_A_ fd, 0, anfds [fd].events); |
106 | * should not be returned ever; but FreeBSD does :-\ |
116 | else if (err == EBADF) /* on EBADF, we re-check the fd */ |
107 | * An error is also indicated when a callback deletes |
117 | { |
108 | * an event we are still processing. In that case |
118 | if (fd_valid (fd)) |
109 | * the data field is set to ENOENT. |
119 | kqueue_modify (EV_A_ fd, 0, anfds [fd].events); |
110 | */ |
120 | else |
111 | if (kqueue_events [i].data == EBADF) |
121 | fd_kill (EV_A_ fd); |
112 | fd_kill (EV_A_ kqueue_events [i].ident); |
122 | } |
|
|
123 | else /* on all other errors, we error out on the fd */ |
|
|
124 | fd_kill (EV_A_ fd); |
|
|
125 | } |
113 | } |
126 | } |
114 | else |
127 | else |
115 | fd_event ( |
128 | fd_event ( |
116 | EV_A_ |
129 | EV_A_ |
117 | kqueue_events [i].ident, |
130 | fd, |
118 | kqueue_events [i].filter == EVFILT_READ ? EV_READ |
131 | kqueue_events [i].filter == EVFILT_READ ? EV_READ |
119 | : kqueue_events [i].filter == EVFILT_WRITE ? EV_WRITE |
132 | : kqueue_events [i].filter == EVFILT_WRITE ? EV_WRITE |
120 | : 0 |
133 | : 0 |
121 | ); |
134 | ); |
122 | } |
135 | } |
123 | |
136 | |
124 | if (expect_false (res == kqueue_eventmax)) |
137 | if (expect_false (res == kqueue_eventmax)) |
125 | { |
138 | { |
126 | ev_free (kqueue_events); |
139 | ev_free (kqueue_events); |
127 | kqueue_eventmax = array_roundsize (kqueue_events, kqueue_eventmax << 1); |
140 | kqueue_eventmax = array_nextsize (sizeof (struct kevent), kqueue_eventmax, kqueue_eventmax + 1); |
128 | kqueue_events = ev_malloc (sizeof (struct kevent) * kqueue_eventmax); |
141 | kqueue_events = (struct kevent *)ev_malloc (sizeof (struct kevent) * kqueue_eventmax); |
129 | } |
142 | } |
130 | } |
143 | } |
131 | |
144 | |
132 | static int |
145 | int inline_size |
133 | kqueue_init (EV_P_ int flags) |
146 | kqueue_init (EV_P_ int flags) |
134 | { |
147 | { |
135 | struct kevent ch, ev; |
148 | struct kevent ch, ev; |
136 | |
149 | |
137 | /* Initalize the kernel queue */ |
150 | /* Initalize the kernel queue */ |
138 | if ((kqueue_fd = kqueue ()) < 0) |
151 | if ((backend_fd = kqueue ()) < 0) |
139 | return 0; |
152 | return 0; |
140 | |
153 | |
141 | fcntl (kqueue_fd, F_SETFD, FD_CLOEXEC); /* not sure if necessary, hopefully doesn't hurt */ |
154 | fcntl (backend_fd, F_SETFD, FD_CLOEXEC); /* not sure if necessary, hopefully doesn't hurt */ |
142 | |
155 | |
143 | /* Check for Mac OS X kqueue bug. */ |
156 | backend_fudge = 1e-3; /* needed to compensate for kevent returning early */ |
144 | ch.ident = -1; |
157 | backend_modify = kqueue_modify; |
145 | ch.filter = EVFILT_READ; |
158 | backend_poll = kqueue_poll; |
146 | ch.flags = EV_ADD; |
|
|
147 | |
159 | |
148 | /* |
|
|
149 | * If kqueue works, then kevent will succeed, and it will |
|
|
150 | * stick an error in ev. If kqueue is broken, then |
|
|
151 | * kevent will fail. |
|
|
152 | */ |
|
|
153 | if (kevent (kqueue_fd, &ch, 1, &ev, 1, 0) != 1 |
|
|
154 | || ev.ident != -1 |
|
|
155 | || ev.flags != EV_ERROR) |
|
|
156 | { |
|
|
157 | /* detected broken kqueue */ |
|
|
158 | close (kqueue_fd); |
|
|
159 | return 0; |
|
|
160 | } |
|
|
161 | |
|
|
162 | method_fudge = 1e-3; /* needed to compensate for kevent returning early */ |
|
|
163 | method_modify = kqueue_modify; |
|
|
164 | method_poll = kqueue_poll; |
|
|
165 | |
|
|
166 | kqueue_eventmax = 64; /* intiial number of events receivable per poll */ |
160 | kqueue_eventmax = 64; /* initial number of events receivable per poll */ |
167 | kqueue_events = ev_malloc (sizeof (struct kevent) * kqueue_eventmax); |
161 | kqueue_events = (struct kevent *)ev_malloc (sizeof (struct kevent) * kqueue_eventmax); |
168 | |
162 | |
169 | kqueue_changes = 0; |
163 | kqueue_changes = 0; |
170 | kqueue_changemax = 0; |
164 | kqueue_changemax = 0; |
171 | kqueue_changecnt = 0; |
165 | kqueue_changecnt = 0; |
172 | |
166 | |
173 | return EVMETHOD_KQUEUE; |
167 | return EVBACKEND_KQUEUE; |
174 | } |
168 | } |
175 | |
169 | |
176 | static void |
170 | void inline_size |
177 | kqueue_destroy (EV_P) |
171 | kqueue_destroy (EV_P) |
178 | { |
172 | { |
179 | close (kqueue_fd); |
|
|
180 | |
|
|
181 | ev_free (kqueue_events); |
173 | ev_free (kqueue_events); |
182 | ev_free (kqueue_changes); |
174 | ev_free (kqueue_changes); |
183 | } |
175 | } |
184 | |
176 | |
185 | static void |
177 | void inline_size |
186 | kqueue_fork (EV_P) |
178 | kqueue_fork (EV_P) |
187 | { |
179 | { |
188 | close (kqueue_fd); |
180 | close (backend_fd); |
189 | |
181 | |
190 | while ((kqueue_fd = kqueue ()) < 0) |
182 | while ((backend_fd = kqueue ()) < 0) |
191 | syserr ("(libev) kqueue"); |
183 | syserr ("(libev) kqueue"); |
192 | |
184 | |
193 | fcntl (kqueue_fd, F_SETFD, FD_CLOEXEC); |
185 | fcntl (backend_fd, F_SETFD, FD_CLOEXEC); |
194 | |
186 | |
195 | /* re-register interest in fds */ |
187 | /* re-register interest in fds */ |
196 | fd_rearm_all (EV_A); |
188 | fd_rearm_all (EV_A); |
197 | } |
189 | } |
198 | |
190 | |