1 | /* |
1 | /* |
|
|
2 | * libev kqueue backend |
|
|
3 | * |
2 | * Copyright 2007 Marc Alexander Lehmann <libev@schmorp.de> |
4 | * Copyright (c) 2007,2008,2009,2010,2011,2012,2013 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 modifica- |
7 | * modification, are permitted provided that the following conditions |
8 | * tion, are permitted provided that the following conditions are met: |
8 | * are met: |
|
|
9 | * |
9 | * |
10 | * 1. Redistributions of source code must retain the above copyright |
10 | * 1. Redistributions of source code must retain the above copyright notice, |
11 | * notice, this list of conditions and the following disclaimer. |
11 | * this list of conditions and the following disclaimer. |
|
|
12 | * |
12 | * 2. Redistributions in binary form must reproduce the above copyright |
13 | * 2. Redistributions in binary form must reproduce the above copyright |
13 | * notice, this list of conditions and the following disclaimer in the |
14 | * notice, this list of conditions and the following disclaimer in the |
14 | * documentation and/or other materials provided with the distribution. |
15 | * 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 | * |
16 | * |
18 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR |
17 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED |
19 | * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES |
18 | * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER- |
20 | * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. |
19 | * CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO |
21 | * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, |
20 | * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE- |
22 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT |
21 | * CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
23 | * NOT 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 OF |
25 | * ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED |
27 | * 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 | |
30 | #include <sys/types.h> |
40 | #include <sys/types.h> |
31 | #include <sys/time.h> |
41 | #include <sys/time.h> |
32 | #include <sys/queue.h> |
|
|
33 | #include <sys/event.h> |
42 | #include <sys/event.h> |
34 | #include <string.h> |
43 | #include <string.h> |
35 | #include <errno.h> |
44 | #include <errno.h> |
36 | |
45 | |
37 | static void |
46 | void inline_speed |
38 | kqueue_change (EV_P_ int fd, int filter, int flags, int fflags) |
47 | kqueue_change (EV_P_ int fd, int filter, int flags, int fflags) |
39 | { |
48 | { |
40 | struct kevent *ke; |
49 | ++kqueue_changecnt; |
41 | |
|
|
42 | array_needsize (struct kevent, kqueue_changes, kqueue_changemax, ++kqueue_changecnt, ); |
50 | array_needsize (struct kevent, kqueue_changes, kqueue_changemax, kqueue_changecnt, EMPTY2); |
43 | |
51 | |
44 | ke = &kqueue_changes [kqueue_changecnt - 1]; |
52 | 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 | } |
53 | } |
51 | |
54 | |
|
|
55 | /* OS X at least needs this */ |
|
|
56 | #ifndef EV_ENABLE |
|
|
57 | # define EV_ENABLE 0 |
|
|
58 | #endif |
52 | #ifndef NOTE_EOF |
59 | #ifndef NOTE_EOF |
53 | # define NOTE_EOF 0 |
60 | # define NOTE_EOF 0 |
54 | #endif |
61 | #endif |
55 | |
62 | |
56 | static void |
63 | static void |
57 | kqueue_modify (EV_P_ int fd, int oev, int nev) |
64 | kqueue_modify (EV_P_ int fd, int oev, int nev) |
58 | { |
65 | { |
59 | if ((oev ^ nev) & EV_READ) |
66 | if (oev != nev) |
60 | { |
67 | { |
61 | if (nev & EV_READ) |
68 | if (oev & EV_READ) |
62 | kqueue_change (EV_A_ fd, EVFILT_READ, EV_ADD, NOTE_EOF); |
|
|
63 | else |
|
|
64 | kqueue_change (EV_A_ fd, EVFILT_READ, EV_DELETE, 0); |
69 | kqueue_change (EV_A_ fd, EVFILT_READ , EV_DELETE, 0); |
65 | } |
|
|
66 | |
70 | |
67 | if ((oev ^ nev) & EV_WRITE) |
|
|
68 | { |
|
|
69 | if (nev & EV_WRITE) |
71 | if (oev & EV_WRITE) |
70 | kqueue_change (EV_A_ fd, EVFILT_WRITE, EV_ADD, NOTE_EOF); |
|
|
71 | else |
|
|
72 | kqueue_change (EV_A_ fd, EVFILT_WRITE, EV_DELETE, 0); |
72 | kqueue_change (EV_A_ fd, EVFILT_WRITE, EV_DELETE, 0); |
73 | } |
73 | } |
|
|
74 | |
|
|
75 | /* to detect close/reopen reliably, we have to re-add */ |
|
|
76 | /* event requests even when oev == nev */ |
|
|
77 | |
|
|
78 | if (nev & EV_READ) |
|
|
79 | kqueue_change (EV_A_ fd, EVFILT_READ , EV_ADD | EV_ENABLE, NOTE_EOF); |
|
|
80 | |
|
|
81 | if (nev & EV_WRITE) |
|
|
82 | kqueue_change (EV_A_ fd, EVFILT_WRITE, EV_ADD | EV_ENABLE, NOTE_EOF); |
74 | } |
83 | } |
75 | |
84 | |
76 | static void |
85 | static void |
77 | kqueue_poll (EV_P_ ev_tstamp timeout) |
86 | kqueue_poll (EV_P_ ev_tstamp timeout) |
78 | { |
87 | { |
79 | int res, i; |
88 | int res, i; |
80 | struct timespec ts; |
89 | struct timespec ts; |
81 | |
90 | |
82 | ts.tv_sec = (time_t)timeout; |
91 | /* need to resize so there is enough space for errors */ |
83 | ts.tv_nsec = (long)(timeout - (ev_tstamp)ts.tv_sec) * 1e9; |
92 | if (kqueue_changecnt > kqueue_eventmax) |
|
|
93 | { |
|
|
94 | ev_free (kqueue_events); |
|
|
95 | kqueue_eventmax = array_nextsize (sizeof (struct kevent), kqueue_eventmax, kqueue_changecnt); |
|
|
96 | kqueue_events = (struct kevent *)ev_malloc (sizeof (struct kevent) * kqueue_eventmax); |
|
|
97 | } |
|
|
98 | |
|
|
99 | EV_RELEASE_CB; |
|
|
100 | EV_TS_SET (ts, timeout); |
84 | res = kevent (kqueue_fd, kqueue_changes, kqueue_changecnt, kqueue_events, kqueue_eventmax, &ts); |
101 | res = kevent (backend_fd, kqueue_changes, kqueue_changecnt, kqueue_events, kqueue_eventmax, &ts); |
|
|
102 | EV_ACQUIRE_CB; |
85 | kqueue_changecnt = 0; |
103 | kqueue_changecnt = 0; |
86 | |
104 | |
87 | if (res < 0) |
105 | if (expect_false (res < 0)) |
88 | { |
106 | { |
89 | if (errno != EINTR) |
107 | if (errno != EINTR) |
90 | syserr ("(libev) kevent"); |
108 | ev_syserr ("(libev) kevent"); |
91 | |
109 | |
92 | return; |
110 | return; |
93 | } |
111 | } |
94 | |
112 | |
95 | for (i = 0; i < res; ++i) |
113 | for (i = 0; i < res; ++i) |
96 | { |
114 | { |
|
|
115 | int fd = kqueue_events [i].ident; |
|
|
116 | |
97 | if (kqueue_events [i].flags & EV_ERROR) |
117 | if (expect_false (kqueue_events [i].flags & EV_ERROR)) |
98 | { |
118 | { |
99 | /* |
119 | int err = kqueue_events [i].data; |
100 | * Error messages that can happen, when a delete fails. |
120 | |
101 | * EBADF happens when the file discriptor has been |
121 | /* we are only interested in errors for fds that we are interested in :) */ |
102 | * closed, |
122 | if (anfds [fd].events) |
103 | * ENOENT when the file discriptor was closed and |
|
|
104 | * then reopened. |
|
|
105 | * EINVAL for some reasons not understood; EINVAL |
|
|
106 | * should not be returned ever; but FreeBSD does :-\ |
|
|
107 | * An error is also indicated when a callback deletes |
|
|
108 | * an event we are still processing. In that case |
|
|
109 | * the data field is set to ENOENT. |
|
|
110 | */ |
123 | { |
111 | if (kqueue_events [i].data == EBADF) |
124 | if (err == ENOENT) /* resubmit changes on ENOENT */ |
112 | fd_kill (EV_A_ kqueue_events [i].ident); |
125 | kqueue_modify (EV_A_ fd, 0, anfds [fd].events); |
|
|
126 | else if (err == EBADF) /* on EBADF, we re-check the fd */ |
|
|
127 | { |
|
|
128 | if (fd_valid (fd)) |
|
|
129 | kqueue_modify (EV_A_ fd, 0, anfds [fd].events); |
|
|
130 | else |
|
|
131 | fd_kill (EV_A_ fd); |
|
|
132 | } |
|
|
133 | else /* on all other errors, we error out on the fd */ |
|
|
134 | fd_kill (EV_A_ fd); |
|
|
135 | } |
113 | } |
136 | } |
114 | else |
137 | else |
115 | fd_event ( |
138 | fd_event ( |
116 | EV_A_ |
139 | EV_A_ |
117 | kqueue_events [i].ident, |
140 | fd, |
118 | kqueue_events [i].filter == EVFILT_READ ? EV_READ |
141 | kqueue_events [i].filter == EVFILT_READ ? EV_READ |
119 | : kqueue_events [i].filter == EVFILT_WRITE ? EV_WRITE |
142 | : kqueue_events [i].filter == EVFILT_WRITE ? EV_WRITE |
120 | : 0 |
143 | : 0 |
121 | ); |
144 | ); |
122 | } |
145 | } |
123 | |
146 | |
124 | if (expect_false (res == kqueue_eventmax)) |
147 | if (expect_false (res == kqueue_eventmax)) |
125 | { |
148 | { |
126 | ev_free (kqueue_events); |
149 | ev_free (kqueue_events); |
127 | kqueue_eventmax = array_roundsize (kqueue_events, kqueue_eventmax << 1); |
150 | kqueue_eventmax = array_nextsize (sizeof (struct kevent), kqueue_eventmax, kqueue_eventmax + 1); |
128 | kqueue_events = ev_malloc (sizeof (struct kevent) * kqueue_eventmax); |
151 | kqueue_events = (struct kevent *)ev_malloc (sizeof (struct kevent) * kqueue_eventmax); |
129 | } |
152 | } |
130 | } |
153 | } |
131 | |
154 | |
132 | static int |
155 | int inline_size |
133 | kqueue_init (EV_P_ int flags) |
156 | kqueue_init (EV_P_ int flags) |
134 | { |
157 | { |
135 | struct kevent ch, ev; |
|
|
136 | |
|
|
137 | /* Initalize the kernel queue */ |
158 | /* initialize the kernel queue */ |
138 | if ((kqueue_fd = kqueue ()) < 0) |
159 | kqueue_fd_pid = getpid (); |
|
|
160 | if ((backend_fd = kqueue ()) < 0) |
139 | return 0; |
161 | return 0; |
140 | |
162 | |
141 | fcntl (kqueue_fd, F_SETFD, FD_CLOEXEC); /* not sure if necessary, hopefully doesn't hurt */ |
163 | fcntl (backend_fd, F_SETFD, FD_CLOEXEC); /* not sure if necessary, hopefully doesn't hurt */ |
142 | |
164 | |
143 | /* Check for Mac OS X kqueue bug. */ |
165 | backend_mintime = 1e-9; /* apparently, they did the right thing in freebsd */ |
144 | ch.ident = -1; |
|
|
145 | ch.filter = EVFILT_READ; |
|
|
146 | ch.flags = EV_ADD; |
|
|
147 | |
|
|
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; |
166 | backend_modify = kqueue_modify; |
164 | method_poll = kqueue_poll; |
167 | backend_poll = kqueue_poll; |
165 | |
168 | |
166 | kqueue_eventmax = 64; /* intiial number of events receivable per poll */ |
169 | kqueue_eventmax = 64; /* initial number of events receivable per poll */ |
167 | kqueue_events = ev_malloc (sizeof (struct kevent) * kqueue_eventmax); |
170 | kqueue_events = (struct kevent *)ev_malloc (sizeof (struct kevent) * kqueue_eventmax); |
168 | |
171 | |
169 | kqueue_changes = 0; |
172 | kqueue_changes = 0; |
170 | kqueue_changemax = 0; |
173 | kqueue_changemax = 0; |
171 | kqueue_changecnt = 0; |
174 | kqueue_changecnt = 0; |
172 | |
175 | |
173 | return EVMETHOD_KQUEUE; |
176 | return EVBACKEND_KQUEUE; |
174 | } |
177 | } |
175 | |
178 | |
176 | static void |
179 | void inline_size |
177 | kqueue_destroy (EV_P) |
180 | kqueue_destroy (EV_P) |
178 | { |
181 | { |
179 | close (kqueue_fd); |
|
|
180 | |
|
|
181 | ev_free (kqueue_events); |
182 | ev_free (kqueue_events); |
182 | ev_free (kqueue_changes); |
183 | ev_free (kqueue_changes); |
183 | } |
184 | } |
184 | |
185 | |
185 | static void |
186 | void inline_size |
186 | kqueue_fork (EV_P) |
187 | kqueue_fork (EV_P) |
187 | { |
188 | { |
188 | close (kqueue_fd); |
189 | /* some BSD kernels don't just destroy the kqueue itself, |
|
|
190 | * but also close the fd, which isn't documented, and |
|
|
191 | * impossible to support properly. |
|
|
192 | * we remember the pid of the kqueue call and only close |
|
|
193 | * the fd if the pid is still the same. |
|
|
194 | * this leaks fds on sane kernels, but BSD interfaces are |
|
|
195 | * notoriously buggy and rarely get fixed. |
|
|
196 | */ |
|
|
197 | pid_t newpid = getpid (); |
189 | |
198 | |
190 | while ((kqueue_fd = kqueue ()) < 0) |
199 | if (newpid == kqueue_fd_pid) |
|
|
200 | close (backend_fd); |
|
|
201 | |
|
|
202 | kqueue_fd_pid = newpid; |
|
|
203 | while ((backend_fd = kqueue ()) < 0) |
191 | syserr ("(libev) kqueue"); |
204 | ev_syserr ("(libev) kqueue"); |
192 | |
205 | |
193 | fcntl (kqueue_fd, F_SETFD, FD_CLOEXEC); |
206 | fcntl (backend_fd, F_SETFD, FD_CLOEXEC); |
194 | |
207 | |
195 | /* re-register interest in fds */ |
208 | /* re-register interest in fds */ |
196 | fd_rearm_all (EV_A); |
209 | fd_rearm_all (EV_A); |
197 | } |
210 | } |
198 | |
211 | |
|
|
212 | /* sys/event.h defines EV_ERROR */ |
|
|
213 | #undef EV_ERROR |
|
|
214 | |