| @@ -1,1549 +1,1553 @@ | | | @@ -1,1549 +1,1553 @@ |
1 | /* $NetBSD: kern_event.c,v 1.80 2014/06/24 14:42:43 maxv Exp $ */ | | 1 | /* $NetBSD: kern_event.c,v 1.80.2.1 2015/04/14 04:39:58 snj Exp $ */ |
2 | | | 2 | |
3 | /*- | | 3 | /*- |
4 | * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. | | 4 | * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. |
5 | * All rights reserved. | | 5 | * All rights reserved. |
6 | * | | 6 | * |
7 | * This code is derived from software contributed to The NetBSD Foundation | | 7 | * This code is derived from software contributed to The NetBSD Foundation |
8 | * by Andrew Doran. | | 8 | * by Andrew Doran. |
9 | * | | 9 | * |
10 | * Redistribution and use in source and binary forms, with or without | | 10 | * Redistribution and use in source and binary forms, with or without |
11 | * modification, are permitted provided that the following conditions | | 11 | * modification, are permitted provided that the following conditions |
12 | * are met: | | 12 | * are met: |
13 | * 1. Redistributions of source code must retain the above copyright | | 13 | * 1. Redistributions of source code must retain the above copyright |
14 | * notice, this list of conditions and the following disclaimer. | | 14 | * notice, this list of conditions and the following disclaimer. |
15 | * 2. Redistributions in binary form must reproduce the above copyright | | 15 | * 2. Redistributions in binary form must reproduce the above copyright |
16 | * notice, this list of conditions and the following disclaimer in the | | 16 | * notice, this list of conditions and the following disclaimer in the |
17 | * documentation and/or other materials provided with the distribution. | | 17 | * documentation and/or other materials provided with the distribution. |
18 | * | | 18 | * |
19 | * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS | | 19 | * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS |
20 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED | | 20 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED |
21 | * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR | | 21 | * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
22 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS | | 22 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS |
23 | * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR | | 23 | * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
24 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF | | 24 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
25 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS | | 25 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
26 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN | | 26 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
27 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) | | 27 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
28 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE | | 28 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
29 | * POSSIBILITY OF SUCH DAMAGE. | | 29 | * POSSIBILITY OF SUCH DAMAGE. |
30 | */ | | 30 | */ |
31 | | | 31 | |
32 | /*- | | 32 | /*- |
33 | * Copyright (c) 1999,2000,2001 Jonathan Lemon <jlemon@FreeBSD.org> | | 33 | * Copyright (c) 1999,2000,2001 Jonathan Lemon <jlemon@FreeBSD.org> |
34 | * All rights reserved. | | 34 | * All rights reserved. |
35 | * | | 35 | * |
36 | * Redistribution and use in source and binary forms, with or without | | 36 | * Redistribution and use in source and binary forms, with or without |
37 | * modification, are permitted provided that the following conditions | | 37 | * modification, are permitted provided that the following conditions |
38 | * are met: | | 38 | * are met: |
39 | * 1. Redistributions of source code must retain the above copyright | | 39 | * 1. Redistributions of source code must retain the above copyright |
40 | * notice, this list of conditions and the following disclaimer. | | 40 | * notice, this list of conditions and the following disclaimer. |
41 | * 2. Redistributions in binary form must reproduce the above copyright | | 41 | * 2. Redistributions in binary form must reproduce the above copyright |
42 | * notice, this list of conditions and the following disclaimer in the | | 42 | * notice, this list of conditions and the following disclaimer in the |
43 | * documentation and/or other materials provided with the distribution. | | 43 | * documentation and/or other materials provided with the distribution. |
44 | * | | 44 | * |
45 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND | | 45 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND |
46 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | | 46 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
47 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | | 47 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
48 | * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE | | 48 | * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE |
49 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | | 49 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
50 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | | 50 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
51 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | | 51 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
52 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | | 52 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
53 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | | 53 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
54 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | | 54 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
55 | * SUCH DAMAGE. | | 55 | * SUCH DAMAGE. |
56 | * | | 56 | * |
57 | * FreeBSD: src/sys/kern/kern_event.c,v 1.27 2001/07/05 17:10:44 rwatson Exp | | 57 | * FreeBSD: src/sys/kern/kern_event.c,v 1.27 2001/07/05 17:10:44 rwatson Exp |
58 | */ | | 58 | */ |
59 | | | 59 | |
60 | #include <sys/cdefs.h> | | 60 | #include <sys/cdefs.h> |
61 | __KERNEL_RCSID(0, "$NetBSD: kern_event.c,v 1.80 2014/06/24 14:42:43 maxv Exp $"); | | 61 | __KERNEL_RCSID(0, "$NetBSD: kern_event.c,v 1.80.2.1 2015/04/14 04:39:58 snj Exp $"); |
62 | | | 62 | |
63 | #include <sys/param.h> | | 63 | #include <sys/param.h> |
64 | #include <sys/systm.h> | | 64 | #include <sys/systm.h> |
65 | #include <sys/kernel.h> | | 65 | #include <sys/kernel.h> |
66 | #include <sys/proc.h> | | 66 | #include <sys/proc.h> |
67 | #include <sys/file.h> | | 67 | #include <sys/file.h> |
68 | #include <sys/select.h> | | 68 | #include <sys/select.h> |
69 | #include <sys/queue.h> | | 69 | #include <sys/queue.h> |
70 | #include <sys/event.h> | | 70 | #include <sys/event.h> |
71 | #include <sys/eventvar.h> | | 71 | #include <sys/eventvar.h> |
72 | #include <sys/poll.h> | | 72 | #include <sys/poll.h> |
73 | #include <sys/kmem.h> | | 73 | #include <sys/kmem.h> |
74 | #include <sys/stat.h> | | 74 | #include <sys/stat.h> |
75 | #include <sys/filedesc.h> | | 75 | #include <sys/filedesc.h> |
76 | #include <sys/syscallargs.h> | | 76 | #include <sys/syscallargs.h> |
77 | #include <sys/kauth.h> | | 77 | #include <sys/kauth.h> |
78 | #include <sys/conf.h> | | 78 | #include <sys/conf.h> |
79 | #include <sys/atomic.h> | | 79 | #include <sys/atomic.h> |
80 | | | 80 | |
81 | static int kqueue_scan(file_t *, size_t, struct kevent *, | | 81 | static int kqueue_scan(file_t *, size_t, struct kevent *, |
82 | const struct timespec *, register_t *, | | 82 | const struct timespec *, register_t *, |
83 | const struct kevent_ops *, struct kevent *, | | 83 | const struct kevent_ops *, struct kevent *, |
84 | size_t); | | 84 | size_t); |
85 | static int kqueue_ioctl(file_t *, u_long, void *); | | 85 | static int kqueue_ioctl(file_t *, u_long, void *); |
86 | static int kqueue_fcntl(file_t *, u_int, void *); | | 86 | static int kqueue_fcntl(file_t *, u_int, void *); |
87 | static int kqueue_poll(file_t *, int); | | 87 | static int kqueue_poll(file_t *, int); |
88 | static int kqueue_kqfilter(file_t *, struct knote *); | | 88 | static int kqueue_kqfilter(file_t *, struct knote *); |
89 | static int kqueue_stat(file_t *, struct stat *); | | 89 | static int kqueue_stat(file_t *, struct stat *); |
90 | static int kqueue_close(file_t *); | | 90 | static int kqueue_close(file_t *); |
91 | static int kqueue_register(struct kqueue *, struct kevent *); | | 91 | static int kqueue_register(struct kqueue *, struct kevent *); |
92 | static void kqueue_doclose(struct kqueue *, struct klist *, int); | | 92 | static void kqueue_doclose(struct kqueue *, struct klist *, int); |
93 | | | 93 | |
94 | static void knote_detach(struct knote *, filedesc_t *fdp, bool); | | 94 | static void knote_detach(struct knote *, filedesc_t *fdp, bool); |
95 | static void knote_enqueue(struct knote *); | | 95 | static void knote_enqueue(struct knote *); |
96 | static void knote_activate(struct knote *); | | 96 | static void knote_activate(struct knote *); |
97 | | | 97 | |
98 | static void filt_kqdetach(struct knote *); | | 98 | static void filt_kqdetach(struct knote *); |
99 | static int filt_kqueue(struct knote *, long hint); | | 99 | static int filt_kqueue(struct knote *, long hint); |
100 | static int filt_procattach(struct knote *); | | 100 | static int filt_procattach(struct knote *); |
101 | static void filt_procdetach(struct knote *); | | 101 | static void filt_procdetach(struct knote *); |
102 | static int filt_proc(struct knote *, long hint); | | 102 | static int filt_proc(struct knote *, long hint); |
103 | static int filt_fileattach(struct knote *); | | 103 | static int filt_fileattach(struct knote *); |
104 | static void filt_timerexpire(void *x); | | 104 | static void filt_timerexpire(void *x); |
105 | static int filt_timerattach(struct knote *); | | 105 | static int filt_timerattach(struct knote *); |
106 | static void filt_timerdetach(struct knote *); | | 106 | static void filt_timerdetach(struct knote *); |
107 | static int filt_timer(struct knote *, long hint); | | 107 | static int filt_timer(struct knote *, long hint); |
108 | | | 108 | |
109 | static const struct fileops kqueueops = { | | 109 | static const struct fileops kqueueops = { |
110 | .fo_read = (void *)enxio, | | 110 | .fo_read = (void *)enxio, |
111 | .fo_write = (void *)enxio, | | 111 | .fo_write = (void *)enxio, |
112 | .fo_ioctl = kqueue_ioctl, | | 112 | .fo_ioctl = kqueue_ioctl, |
113 | .fo_fcntl = kqueue_fcntl, | | 113 | .fo_fcntl = kqueue_fcntl, |
114 | .fo_poll = kqueue_poll, | | 114 | .fo_poll = kqueue_poll, |
115 | .fo_stat = kqueue_stat, | | 115 | .fo_stat = kqueue_stat, |
116 | .fo_close = kqueue_close, | | 116 | .fo_close = kqueue_close, |
117 | .fo_kqfilter = kqueue_kqfilter, | | 117 | .fo_kqfilter = kqueue_kqfilter, |
118 | .fo_restart = fnullop_restart, | | 118 | .fo_restart = fnullop_restart, |
119 | }; | | 119 | }; |
120 | | | 120 | |
121 | static const struct filterops kqread_filtops = | | 121 | static const struct filterops kqread_filtops = |
122 | { 1, NULL, filt_kqdetach, filt_kqueue }; | | 122 | { 1, NULL, filt_kqdetach, filt_kqueue }; |
123 | static const struct filterops proc_filtops = | | 123 | static const struct filterops proc_filtops = |
124 | { 0, filt_procattach, filt_procdetach, filt_proc }; | | 124 | { 0, filt_procattach, filt_procdetach, filt_proc }; |
125 | static const struct filterops file_filtops = | | 125 | static const struct filterops file_filtops = |
126 | { 1, filt_fileattach, NULL, NULL }; | | 126 | { 1, filt_fileattach, NULL, NULL }; |
127 | static const struct filterops timer_filtops = | | 127 | static const struct filterops timer_filtops = |
128 | { 0, filt_timerattach, filt_timerdetach, filt_timer }; | | 128 | { 0, filt_timerattach, filt_timerdetach, filt_timer }; |
129 | | | 129 | |
130 | static u_int kq_ncallouts = 0; | | 130 | static u_int kq_ncallouts = 0; |
131 | static int kq_calloutmax = (4 * 1024); | | 131 | static int kq_calloutmax = (4 * 1024); |
132 | | | 132 | |
133 | #define KN_HASHSIZE 64 /* XXX should be tunable */ | | 133 | #define KN_HASHSIZE 64 /* XXX should be tunable */ |
134 | #define KN_HASH(val, mask) (((val) ^ (val >> 8)) & (mask)) | | 134 | #define KN_HASH(val, mask) (((val) ^ (val >> 8)) & (mask)) |
135 | | | 135 | |
136 | extern const struct filterops sig_filtops; | | 136 | extern const struct filterops sig_filtops; |
137 | | | 137 | |
138 | /* | | 138 | /* |
139 | * Table for for all system-defined filters. | | 139 | * Table for for all system-defined filters. |
140 | * These should be listed in the numeric order of the EVFILT_* defines. | | 140 | * These should be listed in the numeric order of the EVFILT_* defines. |
141 | * If filtops is NULL, the filter isn't implemented in NetBSD. | | 141 | * If filtops is NULL, the filter isn't implemented in NetBSD. |
142 | * End of list is when name is NULL. | | 142 | * End of list is when name is NULL. |
143 | * | | 143 | * |
144 | * Note that 'refcnt' is meaningless for built-in filters. | | 144 | * Note that 'refcnt' is meaningless for built-in filters. |
145 | */ | | 145 | */ |
146 | struct kfilter { | | 146 | struct kfilter { |
147 | const char *name; /* name of filter */ | | 147 | const char *name; /* name of filter */ |
148 | uint32_t filter; /* id of filter */ | | 148 | uint32_t filter; /* id of filter */ |
149 | unsigned refcnt; /* reference count */ | | 149 | unsigned refcnt; /* reference count */ |
150 | const struct filterops *filtops;/* operations for filter */ | | 150 | const struct filterops *filtops;/* operations for filter */ |
151 | size_t namelen; /* length of name string */ | | 151 | size_t namelen; /* length of name string */ |
152 | }; | | 152 | }; |
153 | | | 153 | |
154 | /* System defined filters */ | | 154 | /* System defined filters */ |
155 | static struct kfilter sys_kfilters[] = { | | 155 | static struct kfilter sys_kfilters[] = { |
156 | { "EVFILT_READ", EVFILT_READ, 0, &file_filtops, 0 }, | | 156 | { "EVFILT_READ", EVFILT_READ, 0, &file_filtops, 0 }, |
157 | { "EVFILT_WRITE", EVFILT_WRITE, 0, &file_filtops, 0, }, | | 157 | { "EVFILT_WRITE", EVFILT_WRITE, 0, &file_filtops, 0, }, |
158 | { "EVFILT_AIO", EVFILT_AIO, 0, NULL, 0 }, | | 158 | { "EVFILT_AIO", EVFILT_AIO, 0, NULL, 0 }, |
159 | { "EVFILT_VNODE", EVFILT_VNODE, 0, &file_filtops, 0 }, | | 159 | { "EVFILT_VNODE", EVFILT_VNODE, 0, &file_filtops, 0 }, |
160 | { "EVFILT_PROC", EVFILT_PROC, 0, &proc_filtops, 0 }, | | 160 | { "EVFILT_PROC", EVFILT_PROC, 0, &proc_filtops, 0 }, |
161 | { "EVFILT_SIGNAL", EVFILT_SIGNAL, 0, &sig_filtops, 0 }, | | 161 | { "EVFILT_SIGNAL", EVFILT_SIGNAL, 0, &sig_filtops, 0 }, |
162 | { "EVFILT_TIMER", EVFILT_TIMER, 0, &timer_filtops, 0 }, | | 162 | { "EVFILT_TIMER", EVFILT_TIMER, 0, &timer_filtops, 0 }, |
163 | { NULL, 0, 0, NULL, 0 }, | | 163 | { NULL, 0, 0, NULL, 0 }, |
164 | }; | | 164 | }; |
165 | | | 165 | |
166 | /* User defined kfilters */ | | 166 | /* User defined kfilters */ |
167 | static struct kfilter *user_kfilters; /* array */ | | 167 | static struct kfilter *user_kfilters; /* array */ |
168 | static int user_kfilterc; /* current offset */ | | 168 | static int user_kfilterc; /* current offset */ |
169 | static int user_kfiltermaxc; /* max size so far */ | | 169 | static int user_kfiltermaxc; /* max size so far */ |
170 | static size_t user_kfiltersz; /* size of allocated memory */ | | 170 | static size_t user_kfiltersz; /* size of allocated memory */ |
171 | | | 171 | |
172 | /* Locks */ | | 172 | /* Locks */ |
173 | static krwlock_t kqueue_filter_lock; /* lock on filter lists */ | | 173 | static krwlock_t kqueue_filter_lock; /* lock on filter lists */ |
174 | static kmutex_t kqueue_misc_lock; /* miscellaneous */ | | 174 | static kmutex_t kqueue_misc_lock; /* miscellaneous */ |
175 | | | 175 | |
176 | static kauth_listener_t kqueue_listener; | | 176 | static kauth_listener_t kqueue_listener; |
177 | | | 177 | |
178 | static int | | 178 | static int |
179 | kqueue_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, | | 179 | kqueue_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, |
180 | void *arg0, void *arg1, void *arg2, void *arg3) | | 180 | void *arg0, void *arg1, void *arg2, void *arg3) |
181 | { | | 181 | { |
182 | struct proc *p; | | 182 | struct proc *p; |
183 | int result; | | 183 | int result; |
184 | | | 184 | |
185 | result = KAUTH_RESULT_DEFER; | | 185 | result = KAUTH_RESULT_DEFER; |
186 | p = arg0; | | 186 | p = arg0; |
187 | | | 187 | |
188 | if (action != KAUTH_PROCESS_KEVENT_FILTER) | | 188 | if (action != KAUTH_PROCESS_KEVENT_FILTER) |
189 | return result; | | 189 | return result; |
190 | | | 190 | |
191 | if ((kauth_cred_getuid(p->p_cred) != kauth_cred_getuid(cred) || | | 191 | if ((kauth_cred_getuid(p->p_cred) != kauth_cred_getuid(cred) || |
192 | ISSET(p->p_flag, PK_SUGID))) | | 192 | ISSET(p->p_flag, PK_SUGID))) |
193 | return result; | | 193 | return result; |
194 | | | 194 | |
195 | result = KAUTH_RESULT_ALLOW; | | 195 | result = KAUTH_RESULT_ALLOW; |
196 | | | 196 | |
197 | return result; | | 197 | return result; |
198 | } | | 198 | } |
199 | | | 199 | |
200 | /* | | 200 | /* |
201 | * Initialize the kqueue subsystem. | | 201 | * Initialize the kqueue subsystem. |
202 | */ | | 202 | */ |
203 | void | | 203 | void |
204 | kqueue_init(void) | | 204 | kqueue_init(void) |
205 | { | | 205 | { |
206 | | | 206 | |
207 | rw_init(&kqueue_filter_lock); | | 207 | rw_init(&kqueue_filter_lock); |
208 | mutex_init(&kqueue_misc_lock, MUTEX_DEFAULT, IPL_NONE); | | 208 | mutex_init(&kqueue_misc_lock, MUTEX_DEFAULT, IPL_NONE); |
209 | | | 209 | |
210 | kqueue_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, | | 210 | kqueue_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, |
211 | kqueue_listener_cb, NULL); | | 211 | kqueue_listener_cb, NULL); |
212 | } | | 212 | } |
213 | | | 213 | |
214 | /* | | 214 | /* |
215 | * Find kfilter entry by name, or NULL if not found. | | 215 | * Find kfilter entry by name, or NULL if not found. |
216 | */ | | 216 | */ |
217 | static struct kfilter * | | 217 | static struct kfilter * |
218 | kfilter_byname_sys(const char *name) | | 218 | kfilter_byname_sys(const char *name) |
219 | { | | 219 | { |
220 | int i; | | 220 | int i; |
221 | | | 221 | |
222 | KASSERT(rw_lock_held(&kqueue_filter_lock)); | | 222 | KASSERT(rw_lock_held(&kqueue_filter_lock)); |
223 | | | 223 | |
224 | for (i = 0; sys_kfilters[i].name != NULL; i++) { | | 224 | for (i = 0; sys_kfilters[i].name != NULL; i++) { |
225 | if (strcmp(name, sys_kfilters[i].name) == 0) | | 225 | if (strcmp(name, sys_kfilters[i].name) == 0) |
226 | return &sys_kfilters[i]; | | 226 | return &sys_kfilters[i]; |
227 | } | | 227 | } |
228 | return NULL; | | 228 | return NULL; |
229 | } | | 229 | } |
230 | | | 230 | |
231 | static struct kfilter * | | 231 | static struct kfilter * |
232 | kfilter_byname_user(const char *name) | | 232 | kfilter_byname_user(const char *name) |
233 | { | | 233 | { |
234 | int i; | | 234 | int i; |
235 | | | 235 | |
236 | KASSERT(rw_lock_held(&kqueue_filter_lock)); | | 236 | KASSERT(rw_lock_held(&kqueue_filter_lock)); |
237 | | | 237 | |
238 | /* user filter slots have a NULL name if previously deregistered */ | | 238 | /* user filter slots have a NULL name if previously deregistered */ |
239 | for (i = 0; i < user_kfilterc ; i++) { | | 239 | for (i = 0; i < user_kfilterc ; i++) { |
240 | if (user_kfilters[i].name != NULL && | | 240 | if (user_kfilters[i].name != NULL && |
241 | strcmp(name, user_kfilters[i].name) == 0) | | 241 | strcmp(name, user_kfilters[i].name) == 0) |
242 | return &user_kfilters[i]; | | 242 | return &user_kfilters[i]; |
243 | } | | 243 | } |
244 | return NULL; | | 244 | return NULL; |
245 | } | | 245 | } |
246 | | | 246 | |
247 | static struct kfilter * | | 247 | static struct kfilter * |
248 | kfilter_byname(const char *name) | | 248 | kfilter_byname(const char *name) |
249 | { | | 249 | { |
250 | struct kfilter *kfilter; | | 250 | struct kfilter *kfilter; |
251 | | | 251 | |
252 | KASSERT(rw_lock_held(&kqueue_filter_lock)); | | 252 | KASSERT(rw_lock_held(&kqueue_filter_lock)); |
253 | | | 253 | |
254 | if ((kfilter = kfilter_byname_sys(name)) != NULL) | | 254 | if ((kfilter = kfilter_byname_sys(name)) != NULL) |
255 | return kfilter; | | 255 | return kfilter; |
256 | | | 256 | |
257 | return kfilter_byname_user(name); | | 257 | return kfilter_byname_user(name); |
258 | } | | 258 | } |
259 | | | 259 | |
260 | /* | | 260 | /* |
261 | * Find kfilter entry by filter id, or NULL if not found. | | 261 | * Find kfilter entry by filter id, or NULL if not found. |
262 | * Assumes entries are indexed in filter id order, for speed. | | 262 | * Assumes entries are indexed in filter id order, for speed. |
263 | */ | | 263 | */ |
264 | static struct kfilter * | | 264 | static struct kfilter * |
265 | kfilter_byfilter(uint32_t filter) | | 265 | kfilter_byfilter(uint32_t filter) |
266 | { | | 266 | { |
267 | struct kfilter *kfilter; | | 267 | struct kfilter *kfilter; |
268 | | | 268 | |
269 | KASSERT(rw_lock_held(&kqueue_filter_lock)); | | 269 | KASSERT(rw_lock_held(&kqueue_filter_lock)); |
270 | | | 270 | |
271 | if (filter < EVFILT_SYSCOUNT) /* it's a system filter */ | | 271 | if (filter < EVFILT_SYSCOUNT) /* it's a system filter */ |
272 | kfilter = &sys_kfilters[filter]; | | 272 | kfilter = &sys_kfilters[filter]; |
273 | else if (user_kfilters != NULL && | | 273 | else if (user_kfilters != NULL && |
274 | filter < EVFILT_SYSCOUNT + user_kfilterc) | | 274 | filter < EVFILT_SYSCOUNT + user_kfilterc) |
275 | /* it's a user filter */ | | 275 | /* it's a user filter */ |
276 | kfilter = &user_kfilters[filter - EVFILT_SYSCOUNT]; | | 276 | kfilter = &user_kfilters[filter - EVFILT_SYSCOUNT]; |
277 | else | | 277 | else |
278 | return (NULL); /* out of range */ | | 278 | return (NULL); /* out of range */ |
279 | KASSERT(kfilter->filter == filter); /* sanity check! */ | | 279 | KASSERT(kfilter->filter == filter); /* sanity check! */ |
280 | return (kfilter); | | 280 | return (kfilter); |
281 | } | | 281 | } |
282 | | | 282 | |
283 | /* | | 283 | /* |
284 | * Register a new kfilter. Stores the entry in user_kfilters. | | 284 | * Register a new kfilter. Stores the entry in user_kfilters. |
285 | * Returns 0 if operation succeeded, or an appropriate errno(2) otherwise. | | 285 | * Returns 0 if operation succeeded, or an appropriate errno(2) otherwise. |
286 | * If retfilter != NULL, the new filterid is returned in it. | | 286 | * If retfilter != NULL, the new filterid is returned in it. |
287 | */ | | 287 | */ |
288 | int | | 288 | int |
289 | kfilter_register(const char *name, const struct filterops *filtops, | | 289 | kfilter_register(const char *name, const struct filterops *filtops, |
290 | int *retfilter) | | 290 | int *retfilter) |
291 | { | | 291 | { |
292 | struct kfilter *kfilter; | | 292 | struct kfilter *kfilter; |
293 | size_t len; | | 293 | size_t len; |
294 | int i; | | 294 | int i; |
295 | | | 295 | |
296 | if (name == NULL || name[0] == '\0' || filtops == NULL) | | 296 | if (name == NULL || name[0] == '\0' || filtops == NULL) |
297 | return (EINVAL); /* invalid args */ | | 297 | return (EINVAL); /* invalid args */ |
298 | | | 298 | |
299 | rw_enter(&kqueue_filter_lock, RW_WRITER); | | 299 | rw_enter(&kqueue_filter_lock, RW_WRITER); |
300 | if (kfilter_byname(name) != NULL) { | | 300 | if (kfilter_byname(name) != NULL) { |
301 | rw_exit(&kqueue_filter_lock); | | 301 | rw_exit(&kqueue_filter_lock); |
302 | return (EEXIST); /* already exists */ | | 302 | return (EEXIST); /* already exists */ |
303 | } | | 303 | } |
304 | if (user_kfilterc > 0xffffffff - EVFILT_SYSCOUNT) { | | 304 | if (user_kfilterc > 0xffffffff - EVFILT_SYSCOUNT) { |
305 | rw_exit(&kqueue_filter_lock); | | 305 | rw_exit(&kqueue_filter_lock); |
306 | return (EINVAL); /* too many */ | | 306 | return (EINVAL); /* too many */ |
307 | } | | 307 | } |
308 | | | 308 | |
309 | for (i = 0; i < user_kfilterc; i++) { | | 309 | for (i = 0; i < user_kfilterc; i++) { |
310 | kfilter = &user_kfilters[i]; | | 310 | kfilter = &user_kfilters[i]; |
311 | if (kfilter->name == NULL) { | | 311 | if (kfilter->name == NULL) { |
312 | /* Previously deregistered slot. Reuse. */ | | 312 | /* Previously deregistered slot. Reuse. */ |
313 | goto reuse; | | 313 | goto reuse; |
314 | } | | 314 | } |
315 | } | | 315 | } |
316 | | | 316 | |
317 | /* check if need to grow user_kfilters */ | | 317 | /* check if need to grow user_kfilters */ |
318 | if (user_kfilterc + 1 > user_kfiltermaxc) { | | 318 | if (user_kfilterc + 1 > user_kfiltermaxc) { |
319 | /* Grow in KFILTER_EXTENT chunks. */ | | 319 | /* Grow in KFILTER_EXTENT chunks. */ |
320 | user_kfiltermaxc += KFILTER_EXTENT; | | 320 | user_kfiltermaxc += KFILTER_EXTENT; |
321 | len = user_kfiltermaxc * sizeof(*kfilter); | | 321 | len = user_kfiltermaxc * sizeof(*kfilter); |
322 | kfilter = kmem_alloc(len, KM_SLEEP); | | 322 | kfilter = kmem_alloc(len, KM_SLEEP); |
323 | memset((char *)kfilter + user_kfiltersz, 0, len - user_kfiltersz); | | 323 | memset((char *)kfilter + user_kfiltersz, 0, len - user_kfiltersz); |
324 | if (user_kfilters != NULL) { | | 324 | if (user_kfilters != NULL) { |
325 | memcpy(kfilter, user_kfilters, user_kfiltersz); | | 325 | memcpy(kfilter, user_kfilters, user_kfiltersz); |
326 | kmem_free(user_kfilters, user_kfiltersz); | | 326 | kmem_free(user_kfilters, user_kfiltersz); |
327 | } | | 327 | } |
328 | user_kfiltersz = len; | | 328 | user_kfiltersz = len; |
329 | user_kfilters = kfilter; | | 329 | user_kfilters = kfilter; |
330 | } | | 330 | } |
331 | /* Adding new slot */ | | 331 | /* Adding new slot */ |
332 | kfilter = &user_kfilters[user_kfilterc++]; | | 332 | kfilter = &user_kfilters[user_kfilterc++]; |
333 | reuse: | | 333 | reuse: |
334 | kfilter->namelen = strlen(name) + 1; | | 334 | kfilter->namelen = strlen(name) + 1; |
335 | kfilter->name = kmem_alloc(kfilter->namelen, KM_SLEEP); | | 335 | kfilter->name = kmem_alloc(kfilter->namelen, KM_SLEEP); |
336 | memcpy(__UNCONST(kfilter->name), name, kfilter->namelen); | | 336 | memcpy(__UNCONST(kfilter->name), name, kfilter->namelen); |
337 | | | 337 | |
338 | kfilter->filter = (kfilter - user_kfilters) + EVFILT_SYSCOUNT; | | 338 | kfilter->filter = (kfilter - user_kfilters) + EVFILT_SYSCOUNT; |
339 | | | 339 | |
340 | kfilter->filtops = kmem_alloc(sizeof(*filtops), KM_SLEEP); | | 340 | kfilter->filtops = kmem_alloc(sizeof(*filtops), KM_SLEEP); |
341 | memcpy(__UNCONST(kfilter->filtops), filtops, sizeof(*filtops)); | | 341 | memcpy(__UNCONST(kfilter->filtops), filtops, sizeof(*filtops)); |
342 | | | 342 | |
343 | if (retfilter != NULL) | | 343 | if (retfilter != NULL) |
344 | *retfilter = kfilter->filter; | | 344 | *retfilter = kfilter->filter; |
345 | rw_exit(&kqueue_filter_lock); | | 345 | rw_exit(&kqueue_filter_lock); |
346 | | | 346 | |
347 | return (0); | | 347 | return (0); |
348 | } | | 348 | } |
349 | | | 349 | |
350 | /* | | 350 | /* |
351 | * Unregister a kfilter previously registered with kfilter_register. | | 351 | * Unregister a kfilter previously registered with kfilter_register. |
352 | * This retains the filter id, but clears the name and frees filtops (filter | | 352 | * This retains the filter id, but clears the name and frees filtops (filter |
353 | * operations), so that the number isn't reused during a boot. | | 353 | * operations), so that the number isn't reused during a boot. |
354 | * Returns 0 if operation succeeded, or an appropriate errno(2) otherwise. | | 354 | * Returns 0 if operation succeeded, or an appropriate errno(2) otherwise. |
355 | */ | | 355 | */ |
356 | int | | 356 | int |
357 | kfilter_unregister(const char *name) | | 357 | kfilter_unregister(const char *name) |
358 | { | | 358 | { |
359 | struct kfilter *kfilter; | | 359 | struct kfilter *kfilter; |
360 | | | 360 | |
361 | if (name == NULL || name[0] == '\0') | | 361 | if (name == NULL || name[0] == '\0') |
362 | return (EINVAL); /* invalid name */ | | 362 | return (EINVAL); /* invalid name */ |
363 | | | 363 | |
364 | rw_enter(&kqueue_filter_lock, RW_WRITER); | | 364 | rw_enter(&kqueue_filter_lock, RW_WRITER); |
365 | if (kfilter_byname_sys(name) != NULL) { | | 365 | if (kfilter_byname_sys(name) != NULL) { |
366 | rw_exit(&kqueue_filter_lock); | | 366 | rw_exit(&kqueue_filter_lock); |
367 | return (EINVAL); /* can't detach system filters */ | | 367 | return (EINVAL); /* can't detach system filters */ |
368 | } | | 368 | } |
369 | | | 369 | |
370 | kfilter = kfilter_byname_user(name); | | 370 | kfilter = kfilter_byname_user(name); |
371 | if (kfilter == NULL) { | | 371 | if (kfilter == NULL) { |
372 | rw_exit(&kqueue_filter_lock); | | 372 | rw_exit(&kqueue_filter_lock); |
373 | return (ENOENT); | | 373 | return (ENOENT); |
374 | } | | 374 | } |
375 | if (kfilter->refcnt != 0) { | | 375 | if (kfilter->refcnt != 0) { |
376 | rw_exit(&kqueue_filter_lock); | | 376 | rw_exit(&kqueue_filter_lock); |
377 | return (EBUSY); | | 377 | return (EBUSY); |
378 | } | | 378 | } |
379 | | | 379 | |
380 | /* Cast away const (but we know it's safe. */ | | 380 | /* Cast away const (but we know it's safe. */ |
381 | kmem_free(__UNCONST(kfilter->name), kfilter->namelen); | | 381 | kmem_free(__UNCONST(kfilter->name), kfilter->namelen); |
382 | kfilter->name = NULL; /* mark as `not implemented' */ | | 382 | kfilter->name = NULL; /* mark as `not implemented' */ |
383 | | | 383 | |
384 | if (kfilter->filtops != NULL) { | | 384 | if (kfilter->filtops != NULL) { |
385 | /* Cast away const (but we know it's safe. */ | | 385 | /* Cast away const (but we know it's safe. */ |
386 | kmem_free(__UNCONST(kfilter->filtops), | | 386 | kmem_free(__UNCONST(kfilter->filtops), |
387 | sizeof(*kfilter->filtops)); | | 387 | sizeof(*kfilter->filtops)); |
388 | kfilter->filtops = NULL; /* mark as `not implemented' */ | | 388 | kfilter->filtops = NULL; /* mark as `not implemented' */ |
389 | } | | 389 | } |
390 | rw_exit(&kqueue_filter_lock); | | 390 | rw_exit(&kqueue_filter_lock); |
391 | | | 391 | |
392 | return (0); | | 392 | return (0); |
393 | } | | 393 | } |
394 | | | 394 | |
395 | | | 395 | |
396 | /* | | 396 | /* |
397 | * Filter attach method for EVFILT_READ and EVFILT_WRITE on normal file | | 397 | * Filter attach method for EVFILT_READ and EVFILT_WRITE on normal file |
398 | * descriptors. Calls fileops kqfilter method for given file descriptor. | | 398 | * descriptors. Calls fileops kqfilter method for given file descriptor. |
399 | */ | | 399 | */ |
400 | static int | | 400 | static int |
401 | filt_fileattach(struct knote *kn) | | 401 | filt_fileattach(struct knote *kn) |
402 | { | | 402 | { |
403 | file_t *fp; | | 403 | file_t *fp; |
404 | | | 404 | |
405 | fp = kn->kn_obj; | | 405 | fp = kn->kn_obj; |
406 | | | 406 | |
407 | return (*fp->f_ops->fo_kqfilter)(fp, kn); | | 407 | return (*fp->f_ops->fo_kqfilter)(fp, kn); |
408 | } | | 408 | } |
409 | | | 409 | |
410 | /* | | 410 | /* |
411 | * Filter detach method for EVFILT_READ on kqueue descriptor. | | 411 | * Filter detach method for EVFILT_READ on kqueue descriptor. |
412 | */ | | 412 | */ |
413 | static void | | 413 | static void |
414 | filt_kqdetach(struct knote *kn) | | 414 | filt_kqdetach(struct knote *kn) |
415 | { | | 415 | { |
416 | struct kqueue *kq; | | 416 | struct kqueue *kq; |
417 | | | 417 | |
418 | kq = ((file_t *)kn->kn_obj)->f_data; | | 418 | kq = ((file_t *)kn->kn_obj)->f_data; |
419 | | | 419 | |
420 | mutex_spin_enter(&kq->kq_lock); | | 420 | mutex_spin_enter(&kq->kq_lock); |
421 | SLIST_REMOVE(&kq->kq_sel.sel_klist, kn, knote, kn_selnext); | | 421 | SLIST_REMOVE(&kq->kq_sel.sel_klist, kn, knote, kn_selnext); |
422 | mutex_spin_exit(&kq->kq_lock); | | 422 | mutex_spin_exit(&kq->kq_lock); |
423 | } | | 423 | } |
424 | | | 424 | |
425 | /* | | 425 | /* |
426 | * Filter event method for EVFILT_READ on kqueue descriptor. | | 426 | * Filter event method for EVFILT_READ on kqueue descriptor. |
427 | */ | | 427 | */ |
428 | /*ARGSUSED*/ | | 428 | /*ARGSUSED*/ |
429 | static int | | 429 | static int |
430 | filt_kqueue(struct knote *kn, long hint) | | 430 | filt_kqueue(struct knote *kn, long hint) |
431 | { | | 431 | { |
432 | struct kqueue *kq; | | 432 | struct kqueue *kq; |
433 | int rv; | | 433 | int rv; |
434 | | | 434 | |
435 | kq = ((file_t *)kn->kn_obj)->f_data; | | 435 | kq = ((file_t *)kn->kn_obj)->f_data; |
436 | | | 436 | |
437 | if (hint != NOTE_SUBMIT) | | 437 | if (hint != NOTE_SUBMIT) |
438 | mutex_spin_enter(&kq->kq_lock); | | 438 | mutex_spin_enter(&kq->kq_lock); |
439 | kn->kn_data = kq->kq_count; | | 439 | kn->kn_data = kq->kq_count; |
440 | rv = (kn->kn_data > 0); | | 440 | rv = (kn->kn_data > 0); |
441 | if (hint != NOTE_SUBMIT) | | 441 | if (hint != NOTE_SUBMIT) |
442 | mutex_spin_exit(&kq->kq_lock); | | 442 | mutex_spin_exit(&kq->kq_lock); |
443 | | | 443 | |
444 | return rv; | | 444 | return rv; |
445 | } | | 445 | } |
446 | | | 446 | |
447 | /* | | 447 | /* |
448 | * Filter attach method for EVFILT_PROC. | | 448 | * Filter attach method for EVFILT_PROC. |
449 | */ | | 449 | */ |
450 | static int | | 450 | static int |
451 | filt_procattach(struct knote *kn) | | 451 | filt_procattach(struct knote *kn) |
452 | { | | 452 | { |
453 | struct proc *p; | | 453 | struct proc *p; |
454 | struct lwp *curl; | | 454 | struct lwp *curl; |
455 | | | 455 | |
456 | curl = curlwp; | | 456 | curl = curlwp; |
457 | | | 457 | |
458 | mutex_enter(proc_lock); | | 458 | mutex_enter(proc_lock); |
459 | if (kn->kn_flags & EV_FLAG1) { | | 459 | if (kn->kn_flags & EV_FLAG1) { |
460 | /* | | 460 | /* |
461 | * NOTE_TRACK attaches to the child process too early | | 461 | * NOTE_TRACK attaches to the child process too early |
462 | * for proc_find, so do a raw look up and check the state | | 462 | * for proc_find, so do a raw look up and check the state |
463 | * explicitly. | | 463 | * explicitly. |
464 | */ | | 464 | */ |
465 | p = proc_find_raw(kn->kn_id); | | 465 | p = proc_find_raw(kn->kn_id); |
466 | if (p != NULL && p->p_stat != SIDL) | | 466 | if (p != NULL && p->p_stat != SIDL) |
467 | p = NULL; | | 467 | p = NULL; |
468 | } else { | | 468 | } else { |
469 | p = proc_find(kn->kn_id); | | 469 | p = proc_find(kn->kn_id); |
470 | } | | 470 | } |
471 | | | 471 | |
472 | if (p == NULL) { | | 472 | if (p == NULL) { |
473 | mutex_exit(proc_lock); | | 473 | mutex_exit(proc_lock); |
474 | return ESRCH; | | 474 | return ESRCH; |
475 | } | | 475 | } |
476 | | | 476 | |
477 | /* | | 477 | /* |
478 | * Fail if it's not owned by you, or the last exec gave us | | 478 | * Fail if it's not owned by you, or the last exec gave us |
479 | * setuid/setgid privs (unless you're root). | | 479 | * setuid/setgid privs (unless you're root). |
480 | */ | | 480 | */ |
481 | mutex_enter(p->p_lock); | | 481 | mutex_enter(p->p_lock); |
482 | mutex_exit(proc_lock); | | 482 | mutex_exit(proc_lock); |
483 | if (kauth_authorize_process(curl->l_cred, KAUTH_PROCESS_KEVENT_FILTER, | | 483 | if (kauth_authorize_process(curl->l_cred, KAUTH_PROCESS_KEVENT_FILTER, |
484 | p, NULL, NULL, NULL) != 0) { | | 484 | p, NULL, NULL, NULL) != 0) { |
485 | mutex_exit(p->p_lock); | | 485 | mutex_exit(p->p_lock); |
486 | return EACCES; | | 486 | return EACCES; |
487 | } | | 487 | } |
488 | | | 488 | |
489 | kn->kn_obj = p; | | 489 | kn->kn_obj = p; |
490 | kn->kn_flags |= EV_CLEAR; /* automatically set */ | | 490 | kn->kn_flags |= EV_CLEAR; /* automatically set */ |
491 | | | 491 | |
492 | /* | | 492 | /* |
493 | * internal flag indicating registration done by kernel | | 493 | * internal flag indicating registration done by kernel |
494 | */ | | 494 | */ |
495 | if (kn->kn_flags & EV_FLAG1) { | | 495 | if (kn->kn_flags & EV_FLAG1) { |
496 | kn->kn_data = kn->kn_sdata; /* ppid */ | | 496 | kn->kn_data = kn->kn_sdata; /* ppid */ |
497 | kn->kn_fflags = NOTE_CHILD; | | 497 | kn->kn_fflags = NOTE_CHILD; |
498 | kn->kn_flags &= ~EV_FLAG1; | | 498 | kn->kn_flags &= ~EV_FLAG1; |
499 | } | | 499 | } |
500 | SLIST_INSERT_HEAD(&p->p_klist, kn, kn_selnext); | | 500 | SLIST_INSERT_HEAD(&p->p_klist, kn, kn_selnext); |
501 | mutex_exit(p->p_lock); | | 501 | mutex_exit(p->p_lock); |
502 | | | 502 | |
503 | return 0; | | 503 | return 0; |
504 | } | | 504 | } |
505 | | | 505 | |
506 | /* | | 506 | /* |
507 | * Filter detach method for EVFILT_PROC. | | 507 | * Filter detach method for EVFILT_PROC. |
508 | * | | 508 | * |
509 | * The knote may be attached to a different process, which may exit, | | 509 | * The knote may be attached to a different process, which may exit, |
510 | * leaving nothing for the knote to be attached to. So when the process | | 510 | * leaving nothing for the knote to be attached to. So when the process |
511 | * exits, the knote is marked as DETACHED and also flagged as ONESHOT so | | 511 | * exits, the knote is marked as DETACHED and also flagged as ONESHOT so |
512 | * it will be deleted when read out. However, as part of the knote deletion, | | 512 | * it will be deleted when read out. However, as part of the knote deletion, |
513 | * this routine is called, so a check is needed to avoid actually performing | | 513 | * this routine is called, so a check is needed to avoid actually performing |
514 | * a detach, because the original process might not exist any more. | | 514 | * a detach, because the original process might not exist any more. |
515 | */ | | 515 | */ |
516 | static void | | 516 | static void |
517 | filt_procdetach(struct knote *kn) | | 517 | filt_procdetach(struct knote *kn) |
518 | { | | 518 | { |
519 | struct proc *p; | | 519 | struct proc *p; |
520 | | | 520 | |
521 | if (kn->kn_status & KN_DETACHED) | | 521 | if (kn->kn_status & KN_DETACHED) |
522 | return; | | 522 | return; |
523 | | | 523 | |
524 | p = kn->kn_obj; | | 524 | p = kn->kn_obj; |
525 | | | 525 | |
526 | mutex_enter(p->p_lock); | | 526 | mutex_enter(p->p_lock); |
527 | SLIST_REMOVE(&p->p_klist, kn, knote, kn_selnext); | | 527 | SLIST_REMOVE(&p->p_klist, kn, knote, kn_selnext); |
528 | mutex_exit(p->p_lock); | | 528 | mutex_exit(p->p_lock); |
529 | } | | 529 | } |
530 | | | 530 | |
531 | /* | | 531 | /* |
532 | * Filter event method for EVFILT_PROC. | | 532 | * Filter event method for EVFILT_PROC. |
533 | */ | | 533 | */ |
534 | static int | | 534 | static int |
535 | filt_proc(struct knote *kn, long hint) | | 535 | filt_proc(struct knote *kn, long hint) |
536 | { | | 536 | { |
537 | u_int event, fflag; | | 537 | u_int event, fflag; |
538 | struct kevent kev; | | 538 | struct kevent kev; |
539 | struct kqueue *kq; | | 539 | struct kqueue *kq; |
540 | int error; | | 540 | int error; |
541 | | | 541 | |
542 | event = (u_int)hint & NOTE_PCTRLMASK; | | 542 | event = (u_int)hint & NOTE_PCTRLMASK; |
543 | kq = kn->kn_kq; | | 543 | kq = kn->kn_kq; |
544 | fflag = 0; | | 544 | fflag = 0; |
545 | | | 545 | |
546 | /* If the user is interested in this event, record it. */ | | 546 | /* If the user is interested in this event, record it. */ |
547 | if (kn->kn_sfflags & event) | | 547 | if (kn->kn_sfflags & event) |
548 | fflag |= event; | | 548 | fflag |= event; |
549 | | | 549 | |
550 | if (event == NOTE_EXIT) { | | 550 | if (event == NOTE_EXIT) { |
| | | 551 | struct proc *p = kn->kn_obj; |
| | | 552 | |
| | | 553 | if (p != NULL) |
| | | 554 | kn->kn_data = p->p_xstat; |
551 | /* | | 555 | /* |
552 | * Process is gone, so flag the event as finished. | | 556 | * Process is gone, so flag the event as finished. |
553 | * | | 557 | * |
554 | * Detach the knote from watched process and mark | | 558 | * Detach the knote from watched process and mark |
555 | * it as such. We can't leave this to kqueue_scan(), | | 559 | * it as such. We can't leave this to kqueue_scan(), |
556 | * since the process might not exist by then. And we | | 560 | * since the process might not exist by then. And we |
557 | * have to do this now, since psignal KNOTE() is called | | 561 | * have to do this now, since psignal KNOTE() is called |
558 | * also for zombies and we might end up reading freed | | 562 | * also for zombies and we might end up reading freed |
559 | * memory if the kevent would already be picked up | | 563 | * memory if the kevent would already be picked up |
560 | * and knote g/c'ed. | | 564 | * and knote g/c'ed. |
561 | */ | | 565 | */ |
562 | filt_procdetach(kn); | | 566 | filt_procdetach(kn); |
563 | | | 567 | |
564 | mutex_spin_enter(&kq->kq_lock); | | 568 | mutex_spin_enter(&kq->kq_lock); |
565 | kn->kn_status |= KN_DETACHED; | | 569 | kn->kn_status |= KN_DETACHED; |
566 | /* Mark as ONESHOT, so that the knote it g/c'ed when read */ | | 570 | /* Mark as ONESHOT, so that the knote it g/c'ed when read */ |
567 | kn->kn_flags |= (EV_EOF | EV_ONESHOT); | | 571 | kn->kn_flags |= (EV_EOF | EV_ONESHOT); |
568 | kn->kn_fflags |= fflag; | | 572 | kn->kn_fflags |= fflag; |
569 | mutex_spin_exit(&kq->kq_lock); | | 573 | mutex_spin_exit(&kq->kq_lock); |
570 | | | 574 | |
571 | return 1; | | 575 | return 1; |
572 | } | | 576 | } |
573 | | | 577 | |
574 | mutex_spin_enter(&kq->kq_lock); | | 578 | mutex_spin_enter(&kq->kq_lock); |
575 | if ((event == NOTE_FORK) && (kn->kn_sfflags & NOTE_TRACK)) { | | 579 | if ((event == NOTE_FORK) && (kn->kn_sfflags & NOTE_TRACK)) { |
576 | /* | | 580 | /* |
577 | * Process forked, and user wants to track the new process, | | 581 | * Process forked, and user wants to track the new process, |
578 | * so attach a new knote to it, and immediately report an | | 582 | * so attach a new knote to it, and immediately report an |
579 | * event with the parent's pid. Register knote with new | | 583 | * event with the parent's pid. Register knote with new |
580 | * process. | | 584 | * process. |
581 | */ | | 585 | */ |
582 | kev.ident = hint & NOTE_PDATAMASK; /* pid */ | | 586 | kev.ident = hint & NOTE_PDATAMASK; /* pid */ |
583 | kev.filter = kn->kn_filter; | | 587 | kev.filter = kn->kn_filter; |
584 | kev.flags = kn->kn_flags | EV_ADD | EV_ENABLE | EV_FLAG1; | | 588 | kev.flags = kn->kn_flags | EV_ADD | EV_ENABLE | EV_FLAG1; |
585 | kev.fflags = kn->kn_sfflags; | | 589 | kev.fflags = kn->kn_sfflags; |
586 | kev.data = kn->kn_id; /* parent */ | | 590 | kev.data = kn->kn_id; /* parent */ |
587 | kev.udata = kn->kn_kevent.udata; /* preserve udata */ | | 591 | kev.udata = kn->kn_kevent.udata; /* preserve udata */ |
588 | mutex_spin_exit(&kq->kq_lock); | | 592 | mutex_spin_exit(&kq->kq_lock); |
589 | error = kqueue_register(kq, &kev); | | 593 | error = kqueue_register(kq, &kev); |
590 | mutex_spin_enter(&kq->kq_lock); | | 594 | mutex_spin_enter(&kq->kq_lock); |
591 | if (error != 0) | | 595 | if (error != 0) |
592 | kn->kn_fflags |= NOTE_TRACKERR; | | 596 | kn->kn_fflags |= NOTE_TRACKERR; |
593 | } | | 597 | } |
594 | kn->kn_fflags |= fflag; | | 598 | kn->kn_fflags |= fflag; |
595 | fflag = kn->kn_fflags; | | 599 | fflag = kn->kn_fflags; |
596 | mutex_spin_exit(&kq->kq_lock); | | 600 | mutex_spin_exit(&kq->kq_lock); |
597 | | | 601 | |
598 | return fflag != 0; | | 602 | return fflag != 0; |
599 | } | | 603 | } |
600 | | | 604 | |
601 | static void | | 605 | static void |
602 | filt_timerexpire(void *knx) | | 606 | filt_timerexpire(void *knx) |
603 | { | | 607 | { |
604 | struct knote *kn = knx; | | 608 | struct knote *kn = knx; |
605 | int tticks; | | 609 | int tticks; |
606 | | | 610 | |
607 | mutex_enter(&kqueue_misc_lock); | | 611 | mutex_enter(&kqueue_misc_lock); |
608 | kn->kn_data++; | | 612 | kn->kn_data++; |
609 | knote_activate(kn); | | 613 | knote_activate(kn); |
610 | if ((kn->kn_flags & EV_ONESHOT) == 0) { | | 614 | if ((kn->kn_flags & EV_ONESHOT) == 0) { |
611 | tticks = mstohz(kn->kn_sdata); | | 615 | tticks = mstohz(kn->kn_sdata); |
612 | if (tticks <= 0) | | 616 | if (tticks <= 0) |
613 | tticks = 1; | | 617 | tticks = 1; |
614 | callout_schedule((callout_t *)kn->kn_hook, tticks); | | 618 | callout_schedule((callout_t *)kn->kn_hook, tticks); |
615 | } | | 619 | } |
616 | mutex_exit(&kqueue_misc_lock); | | 620 | mutex_exit(&kqueue_misc_lock); |
617 | } | | 621 | } |
618 | | | 622 | |
619 | /* | | 623 | /* |
620 | * data contains amount of time to sleep, in milliseconds | | 624 | * data contains amount of time to sleep, in milliseconds |
621 | */ | | 625 | */ |
622 | static int | | 626 | static int |
623 | filt_timerattach(struct knote *kn) | | 627 | filt_timerattach(struct knote *kn) |
624 | { | | 628 | { |
625 | callout_t *calloutp; | | 629 | callout_t *calloutp; |
626 | struct kqueue *kq; | | 630 | struct kqueue *kq; |
627 | int tticks; | | 631 | int tticks; |
628 | | | 632 | |
629 | tticks = mstohz(kn->kn_sdata); | | 633 | tticks = mstohz(kn->kn_sdata); |
630 | | | 634 | |
631 | /* if the supplied value is under our resolution, use 1 tick */ | | 635 | /* if the supplied value is under our resolution, use 1 tick */ |
632 | if (tticks == 0) { | | 636 | if (tticks == 0) { |
633 | if (kn->kn_sdata == 0) | | 637 | if (kn->kn_sdata == 0) |
634 | return EINVAL; | | 638 | return EINVAL; |
635 | tticks = 1; | | 639 | tticks = 1; |
636 | } | | 640 | } |
637 | | | 641 | |
638 | if (atomic_inc_uint_nv(&kq_ncallouts) >= kq_calloutmax || | | 642 | if (atomic_inc_uint_nv(&kq_ncallouts) >= kq_calloutmax || |
639 | (calloutp = kmem_alloc(sizeof(*calloutp), KM_NOSLEEP)) == NULL) { | | 643 | (calloutp = kmem_alloc(sizeof(*calloutp), KM_NOSLEEP)) == NULL) { |
640 | atomic_dec_uint(&kq_ncallouts); | | 644 | atomic_dec_uint(&kq_ncallouts); |
641 | return ENOMEM; | | 645 | return ENOMEM; |
642 | } | | 646 | } |
643 | callout_init(calloutp, CALLOUT_MPSAFE); | | 647 | callout_init(calloutp, CALLOUT_MPSAFE); |
644 | | | 648 | |
645 | kq = kn->kn_kq; | | 649 | kq = kn->kn_kq; |
646 | mutex_spin_enter(&kq->kq_lock); | | 650 | mutex_spin_enter(&kq->kq_lock); |
647 | kn->kn_flags |= EV_CLEAR; /* automatically set */ | | 651 | kn->kn_flags |= EV_CLEAR; /* automatically set */ |
648 | kn->kn_hook = calloutp; | | 652 | kn->kn_hook = calloutp; |
649 | mutex_spin_exit(&kq->kq_lock); | | 653 | mutex_spin_exit(&kq->kq_lock); |
650 | | | 654 | |
651 | callout_reset(calloutp, tticks, filt_timerexpire, kn); | | 655 | callout_reset(calloutp, tticks, filt_timerexpire, kn); |
652 | | | 656 | |
653 | return (0); | | 657 | return (0); |
654 | } | | 658 | } |
655 | | | 659 | |
656 | static void | | 660 | static void |
657 | filt_timerdetach(struct knote *kn) | | 661 | filt_timerdetach(struct knote *kn) |
658 | { | | 662 | { |
659 | callout_t *calloutp; | | 663 | callout_t *calloutp; |
660 | | | 664 | |
661 | calloutp = (callout_t *)kn->kn_hook; | | 665 | calloutp = (callout_t *)kn->kn_hook; |
662 | callout_halt(calloutp, NULL); | | 666 | callout_halt(calloutp, NULL); |
663 | callout_destroy(calloutp); | | 667 | callout_destroy(calloutp); |
664 | kmem_free(calloutp, sizeof(*calloutp)); | | 668 | kmem_free(calloutp, sizeof(*calloutp)); |
665 | atomic_dec_uint(&kq_ncallouts); | | 669 | atomic_dec_uint(&kq_ncallouts); |
666 | } | | 670 | } |
667 | | | 671 | |
668 | static int | | 672 | static int |
669 | filt_timer(struct knote *kn, long hint) | | 673 | filt_timer(struct knote *kn, long hint) |
670 | { | | 674 | { |
671 | int rv; | | 675 | int rv; |
672 | | | 676 | |
673 | mutex_enter(&kqueue_misc_lock); | | 677 | mutex_enter(&kqueue_misc_lock); |
674 | rv = (kn->kn_data != 0); | | 678 | rv = (kn->kn_data != 0); |
675 | mutex_exit(&kqueue_misc_lock); | | 679 | mutex_exit(&kqueue_misc_lock); |
676 | | | 680 | |
677 | return rv; | | 681 | return rv; |
678 | } | | 682 | } |
679 | | | 683 | |
680 | /* | | 684 | /* |
681 | * filt_seltrue: | | 685 | * filt_seltrue: |
682 | * | | 686 | * |
683 | * This filter "event" routine simulates seltrue(). | | 687 | * This filter "event" routine simulates seltrue(). |
684 | */ | | 688 | */ |
685 | int | | 689 | int |
686 | filt_seltrue(struct knote *kn, long hint) | | 690 | filt_seltrue(struct knote *kn, long hint) |
687 | { | | 691 | { |
688 | | | 692 | |
689 | /* | | 693 | /* |
690 | * We don't know how much data can be read/written, | | 694 | * We don't know how much data can be read/written, |
691 | * but we know that it *can* be. This is about as | | 695 | * but we know that it *can* be. This is about as |
692 | * good as select/poll does as well. | | 696 | * good as select/poll does as well. |
693 | */ | | 697 | */ |
694 | kn->kn_data = 0; | | 698 | kn->kn_data = 0; |
695 | return (1); | | 699 | return (1); |
696 | } | | 700 | } |
697 | | | 701 | |
698 | /* | | 702 | /* |
699 | * This provides full kqfilter entry for device switch tables, which | | 703 | * This provides full kqfilter entry for device switch tables, which |
700 | * has same effect as filter using filt_seltrue() as filter method. | | 704 | * has same effect as filter using filt_seltrue() as filter method. |
701 | */ | | 705 | */ |
702 | static void | | 706 | static void |
703 | filt_seltruedetach(struct knote *kn) | | 707 | filt_seltruedetach(struct knote *kn) |
704 | { | | 708 | { |
705 | /* Nothing to do */ | | 709 | /* Nothing to do */ |
706 | } | | 710 | } |
707 | | | 711 | |
708 | const struct filterops seltrue_filtops = | | 712 | const struct filterops seltrue_filtops = |
709 | { 1, NULL, filt_seltruedetach, filt_seltrue }; | | 713 | { 1, NULL, filt_seltruedetach, filt_seltrue }; |
710 | | | 714 | |
711 | int | | 715 | int |
712 | seltrue_kqfilter(dev_t dev, struct knote *kn) | | 716 | seltrue_kqfilter(dev_t dev, struct knote *kn) |
713 | { | | 717 | { |
714 | switch (kn->kn_filter) { | | 718 | switch (kn->kn_filter) { |
715 | case EVFILT_READ: | | 719 | case EVFILT_READ: |
716 | case EVFILT_WRITE: | | 720 | case EVFILT_WRITE: |
717 | kn->kn_fop = &seltrue_filtops; | | 721 | kn->kn_fop = &seltrue_filtops; |
718 | break; | | 722 | break; |
719 | default: | | 723 | default: |
720 | return (EINVAL); | | 724 | return (EINVAL); |
721 | } | | 725 | } |
722 | | | 726 | |
723 | /* Nothing more to do */ | | 727 | /* Nothing more to do */ |
724 | return (0); | | 728 | return (0); |
725 | } | | 729 | } |
726 | | | 730 | |
727 | /* | | 731 | /* |
728 | * kqueue(2) system call. | | 732 | * kqueue(2) system call. |
729 | */ | | 733 | */ |
730 | static int | | 734 | static int |
731 | kqueue1(struct lwp *l, int flags, register_t *retval) | | 735 | kqueue1(struct lwp *l, int flags, register_t *retval) |
732 | { | | 736 | { |
733 | struct kqueue *kq; | | 737 | struct kqueue *kq; |
734 | file_t *fp; | | 738 | file_t *fp; |
735 | int fd, error; | | 739 | int fd, error; |
736 | | | 740 | |
737 | if ((error = fd_allocfile(&fp, &fd)) != 0) | | 741 | if ((error = fd_allocfile(&fp, &fd)) != 0) |
738 | return error; | | 742 | return error; |
739 | fp->f_flag = FREAD | FWRITE | (flags & (FNONBLOCK|FNOSIGPIPE)); | | 743 | fp->f_flag = FREAD | FWRITE | (flags & (FNONBLOCK|FNOSIGPIPE)); |
740 | fp->f_type = DTYPE_KQUEUE; | | 744 | fp->f_type = DTYPE_KQUEUE; |
741 | fp->f_ops = &kqueueops; | | 745 | fp->f_ops = &kqueueops; |
742 | kq = kmem_zalloc(sizeof(*kq), KM_SLEEP); | | 746 | kq = kmem_zalloc(sizeof(*kq), KM_SLEEP); |
743 | mutex_init(&kq->kq_lock, MUTEX_DEFAULT, IPL_SCHED); | | 747 | mutex_init(&kq->kq_lock, MUTEX_DEFAULT, IPL_SCHED); |
744 | cv_init(&kq->kq_cv, "kqueue"); | | 748 | cv_init(&kq->kq_cv, "kqueue"); |
745 | selinit(&kq->kq_sel); | | 749 | selinit(&kq->kq_sel); |
746 | TAILQ_INIT(&kq->kq_head); | | 750 | TAILQ_INIT(&kq->kq_head); |
747 | fp->f_data = kq; | | 751 | fp->f_data = kq; |
748 | *retval = fd; | | 752 | *retval = fd; |
749 | kq->kq_fdp = curlwp->l_fd; | | 753 | kq->kq_fdp = curlwp->l_fd; |
750 | fd_set_exclose(l, fd, (flags & O_CLOEXEC) != 0); | | 754 | fd_set_exclose(l, fd, (flags & O_CLOEXEC) != 0); |
751 | fd_affix(curproc, fp, fd); | | 755 | fd_affix(curproc, fp, fd); |
752 | return error; | | 756 | return error; |
753 | } | | 757 | } |
754 | | | 758 | |
755 | /* | | 759 | /* |
756 | * kqueue(2) system call. | | 760 | * kqueue(2) system call. |
757 | */ | | 761 | */ |
758 | int | | 762 | int |
759 | sys_kqueue(struct lwp *l, const void *v, register_t *retval) | | 763 | sys_kqueue(struct lwp *l, const void *v, register_t *retval) |
760 | { | | 764 | { |
761 | return kqueue1(l, 0, retval); | | 765 | return kqueue1(l, 0, retval); |
762 | } | | 766 | } |
763 | | | 767 | |
764 | int | | 768 | int |
765 | sys_kqueue1(struct lwp *l, const struct sys_kqueue1_args *uap, | | 769 | sys_kqueue1(struct lwp *l, const struct sys_kqueue1_args *uap, |
766 | register_t *retval) | | 770 | register_t *retval) |
767 | { | | 771 | { |
768 | /* { | | 772 | /* { |
769 | syscallarg(int) flags; | | 773 | syscallarg(int) flags; |
770 | } */ | | 774 | } */ |
771 | return kqueue1(l, SCARG(uap, flags), retval); | | 775 | return kqueue1(l, SCARG(uap, flags), retval); |
772 | } | | 776 | } |
773 | | | 777 | |
774 | /* | | 778 | /* |
775 | * kevent(2) system call. | | 779 | * kevent(2) system call. |
776 | */ | | 780 | */ |
777 | int | | 781 | int |
778 | kevent_fetch_changes(void *private, const struct kevent *changelist, | | 782 | kevent_fetch_changes(void *private, const struct kevent *changelist, |
779 | struct kevent *changes, size_t index, int n) | | 783 | struct kevent *changes, size_t index, int n) |
780 | { | | 784 | { |
781 | | | 785 | |
782 | return copyin(changelist + index, changes, n * sizeof(*changes)); | | 786 | return copyin(changelist + index, changes, n * sizeof(*changes)); |
783 | } | | 787 | } |
784 | | | 788 | |
785 | int | | 789 | int |
786 | kevent_put_events(void *private, struct kevent *events, | | 790 | kevent_put_events(void *private, struct kevent *events, |
787 | struct kevent *eventlist, size_t index, int n) | | 791 | struct kevent *eventlist, size_t index, int n) |
788 | { | | 792 | { |
789 | | | 793 | |
790 | return copyout(events, eventlist + index, n * sizeof(*events)); | | 794 | return copyout(events, eventlist + index, n * sizeof(*events)); |
791 | } | | 795 | } |
792 | | | 796 | |
793 | static const struct kevent_ops kevent_native_ops = { | | 797 | static const struct kevent_ops kevent_native_ops = { |
794 | .keo_private = NULL, | | 798 | .keo_private = NULL, |
795 | .keo_fetch_timeout = copyin, | | 799 | .keo_fetch_timeout = copyin, |
796 | .keo_fetch_changes = kevent_fetch_changes, | | 800 | .keo_fetch_changes = kevent_fetch_changes, |
797 | .keo_put_events = kevent_put_events, | | 801 | .keo_put_events = kevent_put_events, |
798 | }; | | 802 | }; |
799 | | | 803 | |
800 | int | | 804 | int |
801 | sys___kevent50(struct lwp *l, const struct sys___kevent50_args *uap, | | 805 | sys___kevent50(struct lwp *l, const struct sys___kevent50_args *uap, |
802 | register_t *retval) | | 806 | register_t *retval) |
803 | { | | 807 | { |
804 | /* { | | 808 | /* { |
805 | syscallarg(int) fd; | | 809 | syscallarg(int) fd; |
806 | syscallarg(const struct kevent *) changelist; | | 810 | syscallarg(const struct kevent *) changelist; |
807 | syscallarg(size_t) nchanges; | | 811 | syscallarg(size_t) nchanges; |
808 | syscallarg(struct kevent *) eventlist; | | 812 | syscallarg(struct kevent *) eventlist; |
809 | syscallarg(size_t) nevents; | | 813 | syscallarg(size_t) nevents; |
810 | syscallarg(const struct timespec *) timeout; | | 814 | syscallarg(const struct timespec *) timeout; |
811 | } */ | | 815 | } */ |
812 | | | 816 | |
813 | return kevent1(retval, SCARG(uap, fd), SCARG(uap, changelist), | | 817 | return kevent1(retval, SCARG(uap, fd), SCARG(uap, changelist), |
814 | SCARG(uap, nchanges), SCARG(uap, eventlist), SCARG(uap, nevents), | | 818 | SCARG(uap, nchanges), SCARG(uap, eventlist), SCARG(uap, nevents), |
815 | SCARG(uap, timeout), &kevent_native_ops); | | 819 | SCARG(uap, timeout), &kevent_native_ops); |
816 | } | | 820 | } |
817 | | | 821 | |
818 | int | | 822 | int |
819 | kevent1(register_t *retval, int fd, | | 823 | kevent1(register_t *retval, int fd, |
820 | const struct kevent *changelist, size_t nchanges, | | 824 | const struct kevent *changelist, size_t nchanges, |
821 | struct kevent *eventlist, size_t nevents, | | 825 | struct kevent *eventlist, size_t nevents, |
822 | const struct timespec *timeout, | | 826 | const struct timespec *timeout, |
823 | const struct kevent_ops *keops) | | 827 | const struct kevent_ops *keops) |
824 | { | | 828 | { |
825 | struct kevent *kevp; | | 829 | struct kevent *kevp; |
826 | struct kqueue *kq; | | 830 | struct kqueue *kq; |
827 | struct timespec ts; | | 831 | struct timespec ts; |
828 | size_t i, n, ichange; | | 832 | size_t i, n, ichange; |
829 | int nerrors, error; | | 833 | int nerrors, error; |
830 | struct kevent kevbuf[KQ_NEVENTS]; /* approx 300 bytes on 64-bit */ | | 834 | struct kevent kevbuf[KQ_NEVENTS]; /* approx 300 bytes on 64-bit */ |
831 | file_t *fp; | | 835 | file_t *fp; |
832 | | | 836 | |
833 | /* check that we're dealing with a kq */ | | 837 | /* check that we're dealing with a kq */ |
834 | fp = fd_getfile(fd); | | 838 | fp = fd_getfile(fd); |
835 | if (fp == NULL) | | 839 | if (fp == NULL) |
836 | return (EBADF); | | 840 | return (EBADF); |
837 | | | 841 | |
838 | if (fp->f_type != DTYPE_KQUEUE) { | | 842 | if (fp->f_type != DTYPE_KQUEUE) { |
839 | fd_putfile(fd); | | 843 | fd_putfile(fd); |
840 | return (EBADF); | | 844 | return (EBADF); |
841 | } | | 845 | } |
842 | | | 846 | |
843 | if (timeout != NULL) { | | 847 | if (timeout != NULL) { |
844 | error = (*keops->keo_fetch_timeout)(timeout, &ts, sizeof(ts)); | | 848 | error = (*keops->keo_fetch_timeout)(timeout, &ts, sizeof(ts)); |
845 | if (error) | | 849 | if (error) |
846 | goto done; | | 850 | goto done; |
847 | timeout = &ts; | | 851 | timeout = &ts; |
848 | } | | 852 | } |
849 | | | 853 | |
850 | kq = (struct kqueue *)fp->f_data; | | 854 | kq = (struct kqueue *)fp->f_data; |
851 | nerrors = 0; | | 855 | nerrors = 0; |
852 | ichange = 0; | | 856 | ichange = 0; |
853 | | | 857 | |
854 | /* traverse list of events to register */ | | 858 | /* traverse list of events to register */ |
855 | while (nchanges > 0) { | | 859 | while (nchanges > 0) { |
856 | n = MIN(nchanges, __arraycount(kevbuf)); | | 860 | n = MIN(nchanges, __arraycount(kevbuf)); |
857 | error = (*keops->keo_fetch_changes)(keops->keo_private, | | 861 | error = (*keops->keo_fetch_changes)(keops->keo_private, |
858 | changelist, kevbuf, ichange, n); | | 862 | changelist, kevbuf, ichange, n); |
859 | if (error) | | 863 | if (error) |
860 | goto done; | | 864 | goto done; |
861 | for (i = 0; i < n; i++) { | | 865 | for (i = 0; i < n; i++) { |
862 | kevp = &kevbuf[i]; | | 866 | kevp = &kevbuf[i]; |
863 | kevp->flags &= ~EV_SYSFLAGS; | | 867 | kevp->flags &= ~EV_SYSFLAGS; |
864 | /* register each knote */ | | 868 | /* register each knote */ |
865 | error = kqueue_register(kq, kevp); | | 869 | error = kqueue_register(kq, kevp); |
866 | if (error) { | | 870 | if (error) { |
867 | if (nevents != 0) { | | 871 | if (nevents != 0) { |
868 | kevp->flags = EV_ERROR; | | 872 | kevp->flags = EV_ERROR; |
869 | kevp->data = error; | | 873 | kevp->data = error; |
870 | error = (*keops->keo_put_events) | | 874 | error = (*keops->keo_put_events) |
871 | (keops->keo_private, kevp, | | 875 | (keops->keo_private, kevp, |
872 | eventlist, nerrors, 1); | | 876 | eventlist, nerrors, 1); |
873 | if (error) | | 877 | if (error) |
874 | goto done; | | 878 | goto done; |
875 | nevents--; | | 879 | nevents--; |
876 | nerrors++; | | 880 | nerrors++; |
877 | } else { | | 881 | } else { |
878 | goto done; | | 882 | goto done; |
879 | } | | 883 | } |
880 | } | | 884 | } |
881 | } | | 885 | } |
882 | nchanges -= n; /* update the results */ | | 886 | nchanges -= n; /* update the results */ |
883 | ichange += n; | | 887 | ichange += n; |
884 | } | | 888 | } |
885 | if (nerrors) { | | 889 | if (nerrors) { |
886 | *retval = nerrors; | | 890 | *retval = nerrors; |
887 | error = 0; | | 891 | error = 0; |
888 | goto done; | | 892 | goto done; |
889 | } | | 893 | } |
890 | | | 894 | |
891 | /* actually scan through the events */ | | 895 | /* actually scan through the events */ |
892 | error = kqueue_scan(fp, nevents, eventlist, timeout, retval, keops, | | 896 | error = kqueue_scan(fp, nevents, eventlist, timeout, retval, keops, |
893 | kevbuf, __arraycount(kevbuf)); | | 897 | kevbuf, __arraycount(kevbuf)); |
894 | done: | | 898 | done: |
895 | fd_putfile(fd); | | 899 | fd_putfile(fd); |
896 | return (error); | | 900 | return (error); |
897 | } | | 901 | } |
898 | | | 902 | |
899 | /* | | 903 | /* |
900 | * Register a given kevent kev onto the kqueue | | 904 | * Register a given kevent kev onto the kqueue |
901 | */ | | 905 | */ |
902 | static int | | 906 | static int |
903 | kqueue_register(struct kqueue *kq, struct kevent *kev) | | 907 | kqueue_register(struct kqueue *kq, struct kevent *kev) |
904 | { | | 908 | { |
905 | struct kfilter *kfilter; | | 909 | struct kfilter *kfilter; |
906 | filedesc_t *fdp; | | 910 | filedesc_t *fdp; |
907 | file_t *fp; | | 911 | file_t *fp; |
908 | fdfile_t *ff; | | 912 | fdfile_t *ff; |
909 | struct knote *kn, *newkn; | | 913 | struct knote *kn, *newkn; |
910 | struct klist *list; | | 914 | struct klist *list; |
911 | int error, fd, rv; | | 915 | int error, fd, rv; |
912 | | | 916 | |
913 | fdp = kq->kq_fdp; | | 917 | fdp = kq->kq_fdp; |
914 | fp = NULL; | | 918 | fp = NULL; |
915 | kn = NULL; | | 919 | kn = NULL; |
916 | error = 0; | | 920 | error = 0; |
917 | fd = 0; | | 921 | fd = 0; |
918 | | | 922 | |
919 | newkn = kmem_zalloc(sizeof(*newkn), KM_SLEEP); | | 923 | newkn = kmem_zalloc(sizeof(*newkn), KM_SLEEP); |
920 | | | 924 | |
921 | rw_enter(&kqueue_filter_lock, RW_READER); | | 925 | rw_enter(&kqueue_filter_lock, RW_READER); |
922 | kfilter = kfilter_byfilter(kev->filter); | | 926 | kfilter = kfilter_byfilter(kev->filter); |
923 | if (kfilter == NULL || kfilter->filtops == NULL) { | | 927 | if (kfilter == NULL || kfilter->filtops == NULL) { |
924 | /* filter not found nor implemented */ | | 928 | /* filter not found nor implemented */ |
925 | rw_exit(&kqueue_filter_lock); | | 929 | rw_exit(&kqueue_filter_lock); |
926 | kmem_free(newkn, sizeof(*newkn)); | | 930 | kmem_free(newkn, sizeof(*newkn)); |
927 | return (EINVAL); | | 931 | return (EINVAL); |
928 | } | | 932 | } |
929 | | | 933 | |
930 | /* search if knote already exists */ | | 934 | /* search if knote already exists */ |
931 | if (kfilter->filtops->f_isfd) { | | 935 | if (kfilter->filtops->f_isfd) { |
932 | /* monitoring a file descriptor */ | | 936 | /* monitoring a file descriptor */ |
933 | fd = kev->ident; | | 937 | fd = kev->ident; |
934 | if ((fp = fd_getfile(fd)) == NULL) { | | 938 | if ((fp = fd_getfile(fd)) == NULL) { |
935 | rw_exit(&kqueue_filter_lock); | | 939 | rw_exit(&kqueue_filter_lock); |
936 | kmem_free(newkn, sizeof(*newkn)); | | 940 | kmem_free(newkn, sizeof(*newkn)); |
937 | return EBADF; | | 941 | return EBADF; |
938 | } | | 942 | } |
939 | mutex_enter(&fdp->fd_lock); | | 943 | mutex_enter(&fdp->fd_lock); |
940 | ff = fdp->fd_dt->dt_ff[fd]; | | 944 | ff = fdp->fd_dt->dt_ff[fd]; |
941 | if (fd <= fdp->fd_lastkqfile) { | | 945 | if (fd <= fdp->fd_lastkqfile) { |
942 | SLIST_FOREACH(kn, &ff->ff_knlist, kn_link) { | | 946 | SLIST_FOREACH(kn, &ff->ff_knlist, kn_link) { |
943 | if (kq == kn->kn_kq && | | 947 | if (kq == kn->kn_kq && |
944 | kev->filter == kn->kn_filter) | | 948 | kev->filter == kn->kn_filter) |
945 | break; | | 949 | break; |
946 | } | | 950 | } |
947 | } | | 951 | } |
948 | } else { | | 952 | } else { |
949 | /* | | 953 | /* |
950 | * not monitoring a file descriptor, so | | 954 | * not monitoring a file descriptor, so |
951 | * lookup knotes in internal hash table | | 955 | * lookup knotes in internal hash table |
952 | */ | | 956 | */ |
953 | mutex_enter(&fdp->fd_lock); | | 957 | mutex_enter(&fdp->fd_lock); |
954 | if (fdp->fd_knhashmask != 0) { | | 958 | if (fdp->fd_knhashmask != 0) { |
955 | list = &fdp->fd_knhash[ | | 959 | list = &fdp->fd_knhash[ |
956 | KN_HASH((u_long)kev->ident, fdp->fd_knhashmask)]; | | 960 | KN_HASH((u_long)kev->ident, fdp->fd_knhashmask)]; |
957 | SLIST_FOREACH(kn, list, kn_link) { | | 961 | SLIST_FOREACH(kn, list, kn_link) { |
958 | if (kev->ident == kn->kn_id && | | 962 | if (kev->ident == kn->kn_id && |
959 | kq == kn->kn_kq && | | 963 | kq == kn->kn_kq && |
960 | kev->filter == kn->kn_filter) | | 964 | kev->filter == kn->kn_filter) |
961 | break; | | 965 | break; |
962 | } | | 966 | } |
963 | } | | 967 | } |
964 | } | | 968 | } |
965 | | | 969 | |
966 | /* | | 970 | /* |
967 | * kn now contains the matching knote, or NULL if no match | | 971 | * kn now contains the matching knote, or NULL if no match |
968 | */ | | 972 | */ |
969 | if (kev->flags & EV_ADD) { | | 973 | if (kev->flags & EV_ADD) { |
970 | if (kn == NULL) { | | 974 | if (kn == NULL) { |
971 | /* create new knote */ | | 975 | /* create new knote */ |
972 | kn = newkn; | | 976 | kn = newkn; |
973 | newkn = NULL; | | 977 | newkn = NULL; |
974 | kn->kn_obj = fp; | | 978 | kn->kn_obj = fp; |
975 | kn->kn_id = kev->ident; | | 979 | kn->kn_id = kev->ident; |
976 | kn->kn_kq = kq; | | 980 | kn->kn_kq = kq; |
977 | kn->kn_fop = kfilter->filtops; | | 981 | kn->kn_fop = kfilter->filtops; |
978 | kn->kn_kfilter = kfilter; | | 982 | kn->kn_kfilter = kfilter; |
979 | kn->kn_sfflags = kev->fflags; | | 983 | kn->kn_sfflags = kev->fflags; |
980 | kn->kn_sdata = kev->data; | | 984 | kn->kn_sdata = kev->data; |
981 | kev->fflags = 0; | | 985 | kev->fflags = 0; |
982 | kev->data = 0; | | 986 | kev->data = 0; |
983 | kn->kn_kevent = *kev; | | 987 | kn->kn_kevent = *kev; |
984 | | | 988 | |
985 | /* | | 989 | /* |
986 | * apply reference count to knote structure, and | | 990 | * apply reference count to knote structure, and |
987 | * do not release it at the end of this routine. | | 991 | * do not release it at the end of this routine. |
988 | */ | | 992 | */ |
989 | fp = NULL; | | 993 | fp = NULL; |
990 | | | 994 | |
991 | if (!kn->kn_fop->f_isfd) { | | 995 | if (!kn->kn_fop->f_isfd) { |
992 | /* | | 996 | /* |
993 | * If knote is not on an fd, store on | | 997 | * If knote is not on an fd, store on |
994 | * internal hash table. | | 998 | * internal hash table. |
995 | */ | | 999 | */ |
996 | if (fdp->fd_knhashmask == 0) { | | 1000 | if (fdp->fd_knhashmask == 0) { |
997 | /* XXXAD can block with fd_lock held */ | | 1001 | /* XXXAD can block with fd_lock held */ |
998 | fdp->fd_knhash = hashinit(KN_HASHSIZE, | | 1002 | fdp->fd_knhash = hashinit(KN_HASHSIZE, |
999 | HASH_LIST, true, | | 1003 | HASH_LIST, true, |
1000 | &fdp->fd_knhashmask); | | 1004 | &fdp->fd_knhashmask); |
1001 | } | | 1005 | } |
1002 | list = &fdp->fd_knhash[KN_HASH(kn->kn_id, | | 1006 | list = &fdp->fd_knhash[KN_HASH(kn->kn_id, |
1003 | fdp->fd_knhashmask)]; | | 1007 | fdp->fd_knhashmask)]; |
1004 | } else { | | 1008 | } else { |
1005 | /* Otherwise, knote is on an fd. */ | | 1009 | /* Otherwise, knote is on an fd. */ |
1006 | list = (struct klist *) | | 1010 | list = (struct klist *) |
1007 | &fdp->fd_dt->dt_ff[kn->kn_id]->ff_knlist; | | 1011 | &fdp->fd_dt->dt_ff[kn->kn_id]->ff_knlist; |
1008 | if ((int)kn->kn_id > fdp->fd_lastkqfile) | | 1012 | if ((int)kn->kn_id > fdp->fd_lastkqfile) |
1009 | fdp->fd_lastkqfile = kn->kn_id; | | 1013 | fdp->fd_lastkqfile = kn->kn_id; |
1010 | } | | 1014 | } |
1011 | SLIST_INSERT_HEAD(list, kn, kn_link); | | 1015 | SLIST_INSERT_HEAD(list, kn, kn_link); |
1012 | | | 1016 | |
1013 | KERNEL_LOCK(1, NULL); /* XXXSMP */ | | 1017 | KERNEL_LOCK(1, NULL); /* XXXSMP */ |
1014 | error = (*kfilter->filtops->f_attach)(kn); | | 1018 | error = (*kfilter->filtops->f_attach)(kn); |
1015 | KERNEL_UNLOCK_ONE(NULL); /* XXXSMP */ | | 1019 | KERNEL_UNLOCK_ONE(NULL); /* XXXSMP */ |
1016 | if (error != 0) { | | 1020 | if (error != 0) { |
1017 | #ifdef DIAGNOSTIC | | 1021 | #ifdef DIAGNOSTIC |
1018 | printf("%s: event not supported for file type" | | 1022 | printf("%s: event not supported for file type" |
1019 | " %d\n", __func__, fp ? fp->f_type : -1); | | 1023 | " %d\n", __func__, fp ? fp->f_type : -1); |
1020 | #endif | | 1024 | #endif |
1021 | /* knote_detach() drops fdp->fd_lock */ | | 1025 | /* knote_detach() drops fdp->fd_lock */ |
1022 | knote_detach(kn, fdp, false); | | 1026 | knote_detach(kn, fdp, false); |
1023 | goto done; | | 1027 | goto done; |
1024 | } | | 1028 | } |
1025 | atomic_inc_uint(&kfilter->refcnt); | | 1029 | atomic_inc_uint(&kfilter->refcnt); |
1026 | } else { | | 1030 | } else { |
1027 | /* | | 1031 | /* |
1028 | * The user may change some filter values after the | | 1032 | * The user may change some filter values after the |
1029 | * initial EV_ADD, but doing so will not reset any | | 1033 | * initial EV_ADD, but doing so will not reset any |
1030 | * filter which have already been triggered. | | 1034 | * filter which have already been triggered. |
1031 | */ | | 1035 | */ |
1032 | kn->kn_sfflags = kev->fflags; | | 1036 | kn->kn_sfflags = kev->fflags; |
1033 | kn->kn_sdata = kev->data; | | 1037 | kn->kn_sdata = kev->data; |
1034 | kn->kn_kevent.udata = kev->udata; | | 1038 | kn->kn_kevent.udata = kev->udata; |
1035 | } | | 1039 | } |
1036 | /* | | 1040 | /* |
1037 | * We can get here if we are trying to attach | | 1041 | * We can get here if we are trying to attach |
1038 | * an event to a file descriptor that does not | | 1042 | * an event to a file descriptor that does not |
1039 | * support events, and the attach routine is | | 1043 | * support events, and the attach routine is |
1040 | * broken and does not return an error. | | 1044 | * broken and does not return an error. |
1041 | */ | | 1045 | */ |
1042 | KASSERT(kn->kn_fop->f_event != NULL); | | 1046 | KASSERT(kn->kn_fop->f_event != NULL); |
1043 | KERNEL_LOCK(1, NULL); /* XXXSMP */ | | 1047 | KERNEL_LOCK(1, NULL); /* XXXSMP */ |
1044 | rv = (*kn->kn_fop->f_event)(kn, 0); | | 1048 | rv = (*kn->kn_fop->f_event)(kn, 0); |
1045 | KERNEL_UNLOCK_ONE(NULL); /* XXXSMP */ | | 1049 | KERNEL_UNLOCK_ONE(NULL); /* XXXSMP */ |
1046 | if (rv) | | 1050 | if (rv) |
1047 | knote_activate(kn); | | 1051 | knote_activate(kn); |
1048 | } else { | | 1052 | } else { |
1049 | if (kn == NULL) { | | 1053 | if (kn == NULL) { |
1050 | error = ENOENT; | | 1054 | error = ENOENT; |
1051 | mutex_exit(&fdp->fd_lock); | | 1055 | mutex_exit(&fdp->fd_lock); |
1052 | goto done; | | 1056 | goto done; |
1053 | } | | 1057 | } |
1054 | if (kev->flags & EV_DELETE) { | | 1058 | if (kev->flags & EV_DELETE) { |
1055 | /* knote_detach() drops fdp->fd_lock */ | | 1059 | /* knote_detach() drops fdp->fd_lock */ |
1056 | knote_detach(kn, fdp, true); | | 1060 | knote_detach(kn, fdp, true); |
1057 | goto done; | | 1061 | goto done; |
1058 | } | | 1062 | } |
1059 | } | | 1063 | } |
1060 | | | 1064 | |
1061 | /* disable knote */ | | 1065 | /* disable knote */ |
1062 | if ((kev->flags & EV_DISABLE)) { | | 1066 | if ((kev->flags & EV_DISABLE)) { |
1063 | mutex_spin_enter(&kq->kq_lock); | | 1067 | mutex_spin_enter(&kq->kq_lock); |
1064 | if ((kn->kn_status & KN_DISABLED) == 0) | | 1068 | if ((kn->kn_status & KN_DISABLED) == 0) |
1065 | kn->kn_status |= KN_DISABLED; | | 1069 | kn->kn_status |= KN_DISABLED; |
1066 | mutex_spin_exit(&kq->kq_lock); | | 1070 | mutex_spin_exit(&kq->kq_lock); |
1067 | } | | 1071 | } |
1068 | | | 1072 | |
1069 | /* enable knote */ | | 1073 | /* enable knote */ |
1070 | if ((kev->flags & EV_ENABLE)) { | | 1074 | if ((kev->flags & EV_ENABLE)) { |
1071 | knote_enqueue(kn); | | 1075 | knote_enqueue(kn); |
1072 | } | | 1076 | } |
1073 | mutex_exit(&fdp->fd_lock); | | 1077 | mutex_exit(&fdp->fd_lock); |
1074 | done: | | 1078 | done: |
1075 | rw_exit(&kqueue_filter_lock); | | 1079 | rw_exit(&kqueue_filter_lock); |
1076 | if (newkn != NULL) | | 1080 | if (newkn != NULL) |
1077 | kmem_free(newkn, sizeof(*newkn)); | | 1081 | kmem_free(newkn, sizeof(*newkn)); |
1078 | if (fp != NULL) | | 1082 | if (fp != NULL) |
1079 | fd_putfile(fd); | | 1083 | fd_putfile(fd); |
1080 | return (error); | | 1084 | return (error); |
1081 | } | | 1085 | } |
1082 | | | 1086 | |
1083 | #if defined(DEBUG) | | 1087 | #if defined(DEBUG) |
1084 | static void | | 1088 | static void |
1085 | kq_check(struct kqueue *kq) | | 1089 | kq_check(struct kqueue *kq) |
1086 | { | | 1090 | { |
1087 | const struct knote *kn; | | 1091 | const struct knote *kn; |
1088 | int count; | | 1092 | int count; |
1089 | int nmarker; | | 1093 | int nmarker; |
1090 | | | 1094 | |
1091 | KASSERT(mutex_owned(&kq->kq_lock)); | | 1095 | KASSERT(mutex_owned(&kq->kq_lock)); |
1092 | KASSERT(kq->kq_count >= 0); | | 1096 | KASSERT(kq->kq_count >= 0); |
1093 | | | 1097 | |
1094 | count = 0; | | 1098 | count = 0; |
1095 | nmarker = 0; | | 1099 | nmarker = 0; |
1096 | TAILQ_FOREACH(kn, &kq->kq_head, kn_tqe) { | | 1100 | TAILQ_FOREACH(kn, &kq->kq_head, kn_tqe) { |
1097 | if ((kn->kn_status & (KN_MARKER | KN_QUEUED)) == 0) { | | 1101 | if ((kn->kn_status & (KN_MARKER | KN_QUEUED)) == 0) { |
1098 | panic("%s: kq=%p kn=%p inconsist 1", __func__, kq, kn); | | 1102 | panic("%s: kq=%p kn=%p inconsist 1", __func__, kq, kn); |
1099 | } | | 1103 | } |
1100 | if ((kn->kn_status & KN_MARKER) == 0) { | | 1104 | if ((kn->kn_status & KN_MARKER) == 0) { |
1101 | if (kn->kn_kq != kq) { | | 1105 | if (kn->kn_kq != kq) { |
1102 | panic("%s: kq=%p kn=%p inconsist 2", | | 1106 | panic("%s: kq=%p kn=%p inconsist 2", |
1103 | __func__, kq, kn); | | 1107 | __func__, kq, kn); |
1104 | } | | 1108 | } |
1105 | if ((kn->kn_status & KN_ACTIVE) == 0) { | | 1109 | if ((kn->kn_status & KN_ACTIVE) == 0) { |
1106 | panic("%s: kq=%p kn=%p: not active", | | 1110 | panic("%s: kq=%p kn=%p: not active", |
1107 | __func__, kq, kn); | | 1111 | __func__, kq, kn); |
1108 | } | | 1112 | } |
1109 | count++; | | 1113 | count++; |
1110 | if (count > kq->kq_count) { | | 1114 | if (count > kq->kq_count) { |
1111 | goto bad; | | 1115 | goto bad; |
1112 | } | | 1116 | } |
1113 | } else { | | 1117 | } else { |
1114 | nmarker++; | | 1118 | nmarker++; |
1115 | #if 0 | | 1119 | #if 0 |
1116 | if (nmarker > 10000) { | | 1120 | if (nmarker > 10000) { |
1117 | panic("%s: kq=%p too many markers: %d != %d, " | | 1121 | panic("%s: kq=%p too many markers: %d != %d, " |
1118 | "nmarker=%d", | | 1122 | "nmarker=%d", |
1119 | __func__, kq, kq->kq_count, count, nmarker); | | 1123 | __func__, kq, kq->kq_count, count, nmarker); |
1120 | } | | 1124 | } |
1121 | #endif | | 1125 | #endif |
1122 | } | | 1126 | } |
1123 | } | | 1127 | } |
1124 | if (kq->kq_count != count) { | | 1128 | if (kq->kq_count != count) { |
1125 | bad: | | 1129 | bad: |
1126 | panic("%s: kq=%p inconsist 3: %d != %d, nmarker=%d", | | 1130 | panic("%s: kq=%p inconsist 3: %d != %d, nmarker=%d", |
1127 | __func__, kq, kq->kq_count, count, nmarker); | | 1131 | __func__, kq, kq->kq_count, count, nmarker); |
1128 | } | | 1132 | } |
1129 | } | | 1133 | } |
1130 | #else /* defined(DEBUG) */ | | 1134 | #else /* defined(DEBUG) */ |
1131 | #define kq_check(a) /* nothing */ | | 1135 | #define kq_check(a) /* nothing */ |
1132 | #endif /* defined(DEBUG) */ | | 1136 | #endif /* defined(DEBUG) */ |
1133 | | | 1137 | |
1134 | /* | | 1138 | /* |
1135 | * Scan through the list of events on fp (for a maximum of maxevents), | | 1139 | * Scan through the list of events on fp (for a maximum of maxevents), |
1136 | * returning the results in to ulistp. Timeout is determined by tsp; if | | 1140 | * returning the results in to ulistp. Timeout is determined by tsp; if |
1137 | * NULL, wait indefinitely, if 0 valued, perform a poll, otherwise wait | | 1141 | * NULL, wait indefinitely, if 0 valued, perform a poll, otherwise wait |
1138 | * as appropriate. | | 1142 | * as appropriate. |
1139 | */ | | 1143 | */ |
1140 | static int | | 1144 | static int |
1141 | kqueue_scan(file_t *fp, size_t maxevents, struct kevent *ulistp, | | 1145 | kqueue_scan(file_t *fp, size_t maxevents, struct kevent *ulistp, |
1142 | const struct timespec *tsp, register_t *retval, | | 1146 | const struct timespec *tsp, register_t *retval, |
1143 | const struct kevent_ops *keops, struct kevent *kevbuf, | | 1147 | const struct kevent_ops *keops, struct kevent *kevbuf, |
1144 | size_t kevcnt) | | 1148 | size_t kevcnt) |
1145 | { | | 1149 | { |
1146 | struct kqueue *kq; | | 1150 | struct kqueue *kq; |
1147 | struct kevent *kevp; | | 1151 | struct kevent *kevp; |
1148 | struct timespec ats, sleepts; | | 1152 | struct timespec ats, sleepts; |
1149 | struct knote *kn, *marker; | | 1153 | struct knote *kn, *marker; |
1150 | size_t count, nkev, nevents; | | 1154 | size_t count, nkev, nevents; |
1151 | int timeout, error, rv; | | 1155 | int timeout, error, rv; |
1152 | filedesc_t *fdp; | | 1156 | filedesc_t *fdp; |
1153 | | | 1157 | |
1154 | fdp = curlwp->l_fd; | | 1158 | fdp = curlwp->l_fd; |
1155 | kq = fp->f_data; | | 1159 | kq = fp->f_data; |
1156 | count = maxevents; | | 1160 | count = maxevents; |
1157 | nkev = nevents = error = 0; | | 1161 | nkev = nevents = error = 0; |
1158 | if (count == 0) { | | 1162 | if (count == 0) { |
1159 | *retval = 0; | | 1163 | *retval = 0; |
1160 | return 0; | | 1164 | return 0; |
1161 | } | | 1165 | } |
1162 | | | 1166 | |
1163 | if (tsp) { /* timeout supplied */ | | 1167 | if (tsp) { /* timeout supplied */ |
1164 | ats = *tsp; | | 1168 | ats = *tsp; |
1165 | if (inittimeleft(&ats, &sleepts) == -1) { | | 1169 | if (inittimeleft(&ats, &sleepts) == -1) { |
1166 | *retval = maxevents; | | 1170 | *retval = maxevents; |
1167 | return EINVAL; | | 1171 | return EINVAL; |
1168 | } | | 1172 | } |
1169 | timeout = tstohz(&ats); | | 1173 | timeout = tstohz(&ats); |
1170 | if (timeout <= 0) | | 1174 | if (timeout <= 0) |
1171 | timeout = -1; /* do poll */ | | 1175 | timeout = -1; /* do poll */ |
1172 | } else { | | 1176 | } else { |
1173 | /* no timeout, wait forever */ | | 1177 | /* no timeout, wait forever */ |
1174 | timeout = 0; | | 1178 | timeout = 0; |
1175 | } | | 1179 | } |
1176 | | | 1180 | |
1177 | marker = kmem_zalloc(sizeof(*marker), KM_SLEEP); | | 1181 | marker = kmem_zalloc(sizeof(*marker), KM_SLEEP); |
1178 | marker->kn_status = KN_MARKER; | | 1182 | marker->kn_status = KN_MARKER; |
1179 | mutex_spin_enter(&kq->kq_lock); | | 1183 | mutex_spin_enter(&kq->kq_lock); |
1180 | retry: | | 1184 | retry: |
1181 | kevp = kevbuf; | | 1185 | kevp = kevbuf; |
1182 | if (kq->kq_count == 0) { | | 1186 | if (kq->kq_count == 0) { |
1183 | if (timeout >= 0) { | | 1187 | if (timeout >= 0) { |
1184 | error = cv_timedwait_sig(&kq->kq_cv, | | 1188 | error = cv_timedwait_sig(&kq->kq_cv, |
1185 | &kq->kq_lock, timeout); | | 1189 | &kq->kq_lock, timeout); |
1186 | if (error == 0) { | | 1190 | if (error == 0) { |
1187 | if (tsp == NULL || (timeout = | | 1191 | if (tsp == NULL || (timeout = |
1188 | gettimeleft(&ats, &sleepts)) > 0) | | 1192 | gettimeleft(&ats, &sleepts)) > 0) |
1189 | goto retry; | | 1193 | goto retry; |
1190 | } else { | | 1194 | } else { |
1191 | /* don't restart after signals... */ | | 1195 | /* don't restart after signals... */ |
1192 | if (error == ERESTART) | | 1196 | if (error == ERESTART) |
1193 | error = EINTR; | | 1197 | error = EINTR; |
1194 | if (error == EWOULDBLOCK) | | 1198 | if (error == EWOULDBLOCK) |
1195 | error = 0; | | 1199 | error = 0; |
1196 | } | | 1200 | } |
1197 | } | | 1201 | } |
1198 | } else { | | 1202 | } else { |
1199 | /* mark end of knote list */ | | 1203 | /* mark end of knote list */ |
1200 | TAILQ_INSERT_TAIL(&kq->kq_head, marker, kn_tqe); | | 1204 | TAILQ_INSERT_TAIL(&kq->kq_head, marker, kn_tqe); |
1201 | | | 1205 | |
1202 | while (count != 0) { | | 1206 | while (count != 0) { |
1203 | kn = TAILQ_FIRST(&kq->kq_head); /* get next knote */ | | 1207 | kn = TAILQ_FIRST(&kq->kq_head); /* get next knote */ |
1204 | while ((kn->kn_status & KN_MARKER) != 0) { | | 1208 | while ((kn->kn_status & KN_MARKER) != 0) { |
1205 | if (kn == marker) { | | 1209 | if (kn == marker) { |
1206 | /* it's our marker, stop */ | | 1210 | /* it's our marker, stop */ |
1207 | TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); | | 1211 | TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); |
1208 | if (count < maxevents || (tsp != NULL && | | 1212 | if (count < maxevents || (tsp != NULL && |
1209 | (timeout = gettimeleft(&ats, | | 1213 | (timeout = gettimeleft(&ats, |
1210 | &sleepts)) <= 0)) | | 1214 | &sleepts)) <= 0)) |
1211 | goto done; | | 1215 | goto done; |
1212 | goto retry; | | 1216 | goto retry; |
1213 | } | | 1217 | } |
1214 | /* someone else's marker. */ | | 1218 | /* someone else's marker. */ |
1215 | kn = TAILQ_NEXT(kn, kn_tqe); | | 1219 | kn = TAILQ_NEXT(kn, kn_tqe); |
1216 | } | | 1220 | } |
1217 | kq_check(kq); | | 1221 | kq_check(kq); |
1218 | TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); | | 1222 | TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); |
1219 | kq->kq_count--; | | 1223 | kq->kq_count--; |
1220 | kn->kn_status &= ~KN_QUEUED; | | 1224 | kn->kn_status &= ~KN_QUEUED; |
1221 | kq_check(kq); | | 1225 | kq_check(kq); |
1222 | if (kn->kn_status & KN_DISABLED) { | | 1226 | if (kn->kn_status & KN_DISABLED) { |
1223 | /* don't want disabled events */ | | 1227 | /* don't want disabled events */ |
1224 | continue; | | 1228 | continue; |
1225 | } | | 1229 | } |
1226 | if ((kn->kn_flags & EV_ONESHOT) == 0) { | | 1230 | if ((kn->kn_flags & EV_ONESHOT) == 0) { |
1227 | mutex_spin_exit(&kq->kq_lock); | | 1231 | mutex_spin_exit(&kq->kq_lock); |
1228 | KERNEL_LOCK(1, NULL); /* XXXSMP */ | | 1232 | KERNEL_LOCK(1, NULL); /* XXXSMP */ |
1229 | rv = (*kn->kn_fop->f_event)(kn, 0); | | 1233 | rv = (*kn->kn_fop->f_event)(kn, 0); |
1230 | KERNEL_UNLOCK_ONE(NULL); /* XXXSMP */ | | 1234 | KERNEL_UNLOCK_ONE(NULL); /* XXXSMP */ |
1231 | mutex_spin_enter(&kq->kq_lock); | | 1235 | mutex_spin_enter(&kq->kq_lock); |
1232 | /* Re-poll if note was re-enqueued. */ | | 1236 | /* Re-poll if note was re-enqueued. */ |
1233 | if ((kn->kn_status & KN_QUEUED) != 0) | | 1237 | if ((kn->kn_status & KN_QUEUED) != 0) |
1234 | continue; | | 1238 | continue; |
1235 | if (rv == 0) { | | 1239 | if (rv == 0) { |
1236 | /* | | 1240 | /* |
1237 | * non-ONESHOT event that hasn't | | 1241 | * non-ONESHOT event that hasn't |
1238 | * triggered again, so de-queue. | | 1242 | * triggered again, so de-queue. |
1239 | */ | | 1243 | */ |
1240 | kn->kn_status &= ~KN_ACTIVE; | | 1244 | kn->kn_status &= ~KN_ACTIVE; |
1241 | continue; | | 1245 | continue; |
1242 | } | | 1246 | } |
1243 | } | | 1247 | } |
1244 | /* XXXAD should be got from f_event if !oneshot. */ | | 1248 | /* XXXAD should be got from f_event if !oneshot. */ |
1245 | *kevp++ = kn->kn_kevent; | | 1249 | *kevp++ = kn->kn_kevent; |
1246 | nkev++; | | 1250 | nkev++; |
1247 | if (kn->kn_flags & EV_ONESHOT) { | | 1251 | if (kn->kn_flags & EV_ONESHOT) { |
1248 | /* delete ONESHOT events after retrieval */ | | 1252 | /* delete ONESHOT events after retrieval */ |
1249 | mutex_spin_exit(&kq->kq_lock); | | 1253 | mutex_spin_exit(&kq->kq_lock); |
1250 | mutex_enter(&fdp->fd_lock); | | 1254 | mutex_enter(&fdp->fd_lock); |
1251 | knote_detach(kn, fdp, true); | | 1255 | knote_detach(kn, fdp, true); |
1252 | mutex_spin_enter(&kq->kq_lock); | | 1256 | mutex_spin_enter(&kq->kq_lock); |
1253 | } else if (kn->kn_flags & EV_CLEAR) { | | 1257 | } else if (kn->kn_flags & EV_CLEAR) { |
1254 | /* clear state after retrieval */ | | 1258 | /* clear state after retrieval */ |
1255 | kn->kn_data = 0; | | 1259 | kn->kn_data = 0; |
1256 | kn->kn_fflags = 0; | | 1260 | kn->kn_fflags = 0; |
1257 | kn->kn_status &= ~KN_ACTIVE; | | 1261 | kn->kn_status &= ~KN_ACTIVE; |
1258 | } else { | | 1262 | } else { |
1259 | /* add event back on list */ | | 1263 | /* add event back on list */ |
1260 | kq_check(kq); | | 1264 | kq_check(kq); |
1261 | TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe); | | 1265 | TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe); |
1262 | kq->kq_count++; | | 1266 | kq->kq_count++; |
1263 | kn->kn_status |= KN_QUEUED; | | 1267 | kn->kn_status |= KN_QUEUED; |
1264 | kq_check(kq); | | 1268 | kq_check(kq); |
1265 | } | | 1269 | } |
1266 | if (nkev == kevcnt) { | | 1270 | if (nkev == kevcnt) { |
1267 | /* do copyouts in kevcnt chunks */ | | 1271 | /* do copyouts in kevcnt chunks */ |
1268 | mutex_spin_exit(&kq->kq_lock); | | 1272 | mutex_spin_exit(&kq->kq_lock); |
1269 | error = (*keops->keo_put_events) | | 1273 | error = (*keops->keo_put_events) |
1270 | (keops->keo_private, | | 1274 | (keops->keo_private, |
1271 | kevbuf, ulistp, nevents, nkev); | | 1275 | kevbuf, ulistp, nevents, nkev); |
1272 | mutex_spin_enter(&kq->kq_lock); | | 1276 | mutex_spin_enter(&kq->kq_lock); |
1273 | nevents += nkev; | | 1277 | nevents += nkev; |
1274 | nkev = 0; | | 1278 | nkev = 0; |
1275 | kevp = kevbuf; | | 1279 | kevp = kevbuf; |
1276 | } | | 1280 | } |
1277 | count--; | | 1281 | count--; |
1278 | if (error != 0 || count == 0) { | | 1282 | if (error != 0 || count == 0) { |
1279 | /* remove marker */ | | 1283 | /* remove marker */ |
1280 | TAILQ_REMOVE(&kq->kq_head, marker, kn_tqe); | | 1284 | TAILQ_REMOVE(&kq->kq_head, marker, kn_tqe); |
1281 | break; | | 1285 | break; |
1282 | } | | 1286 | } |
1283 | } | | 1287 | } |
1284 | } | | 1288 | } |
1285 | done: | | 1289 | done: |
1286 | mutex_spin_exit(&kq->kq_lock); | | 1290 | mutex_spin_exit(&kq->kq_lock); |
1287 | if (marker != NULL) | | 1291 | if (marker != NULL) |
1288 | kmem_free(marker, sizeof(*marker)); | | 1292 | kmem_free(marker, sizeof(*marker)); |
1289 | if (nkev != 0) { | | 1293 | if (nkev != 0) { |
1290 | /* copyout remaining events */ | | 1294 | /* copyout remaining events */ |
1291 | error = (*keops->keo_put_events)(keops->keo_private, | | 1295 | error = (*keops->keo_put_events)(keops->keo_private, |
1292 | kevbuf, ulistp, nevents, nkev); | | 1296 | kevbuf, ulistp, nevents, nkev); |
1293 | } | | 1297 | } |
1294 | *retval = maxevents - count; | | 1298 | *retval = maxevents - count; |
1295 | | | 1299 | |
1296 | return error; | | 1300 | return error; |
1297 | } | | 1301 | } |
1298 | | | 1302 | |
1299 | /* | | 1303 | /* |
1300 | * fileops ioctl method for a kqueue descriptor. | | 1304 | * fileops ioctl method for a kqueue descriptor. |
1301 | * | | 1305 | * |
1302 | * Two ioctls are currently supported. They both use struct kfilter_mapping: | | 1306 | * Two ioctls are currently supported. They both use struct kfilter_mapping: |
1303 | * KFILTER_BYNAME find name for filter, and return result in | | 1307 | * KFILTER_BYNAME find name for filter, and return result in |
1304 | * name, which is of size len. | | 1308 | * name, which is of size len. |
1305 | * KFILTER_BYFILTER find filter for name. len is ignored. | | 1309 | * KFILTER_BYFILTER find filter for name. len is ignored. |
1306 | */ | | 1310 | */ |
1307 | /*ARGSUSED*/ | | 1311 | /*ARGSUSED*/ |
1308 | static int | | 1312 | static int |
1309 | kqueue_ioctl(file_t *fp, u_long com, void *data) | | 1313 | kqueue_ioctl(file_t *fp, u_long com, void *data) |
1310 | { | | 1314 | { |
1311 | struct kfilter_mapping *km; | | 1315 | struct kfilter_mapping *km; |
1312 | const struct kfilter *kfilter; | | 1316 | const struct kfilter *kfilter; |
1313 | char *name; | | 1317 | char *name; |
1314 | int error; | | 1318 | int error; |
1315 | | | 1319 | |
1316 | km = data; | | 1320 | km = data; |
1317 | error = 0; | | 1321 | error = 0; |
1318 | name = kmem_alloc(KFILTER_MAXNAME, KM_SLEEP); | | 1322 | name = kmem_alloc(KFILTER_MAXNAME, KM_SLEEP); |
1319 | | | 1323 | |
1320 | switch (com) { | | 1324 | switch (com) { |
1321 | case KFILTER_BYFILTER: /* convert filter -> name */ | | 1325 | case KFILTER_BYFILTER: /* convert filter -> name */ |
1322 | rw_enter(&kqueue_filter_lock, RW_READER); | | 1326 | rw_enter(&kqueue_filter_lock, RW_READER); |
1323 | kfilter = kfilter_byfilter(km->filter); | | 1327 | kfilter = kfilter_byfilter(km->filter); |
1324 | if (kfilter != NULL) { | | 1328 | if (kfilter != NULL) { |
1325 | strlcpy(name, kfilter->name, KFILTER_MAXNAME); | | 1329 | strlcpy(name, kfilter->name, KFILTER_MAXNAME); |
1326 | rw_exit(&kqueue_filter_lock); | | 1330 | rw_exit(&kqueue_filter_lock); |
1327 | error = copyoutstr(name, km->name, km->len, NULL); | | 1331 | error = copyoutstr(name, km->name, km->len, NULL); |
1328 | } else { | | 1332 | } else { |
1329 | rw_exit(&kqueue_filter_lock); | | 1333 | rw_exit(&kqueue_filter_lock); |
1330 | error = ENOENT; | | 1334 | error = ENOENT; |
1331 | } | | 1335 | } |
1332 | break; | | 1336 | break; |
1333 | | | 1337 | |
1334 | case KFILTER_BYNAME: /* convert name -> filter */ | | 1338 | case KFILTER_BYNAME: /* convert name -> filter */ |
1335 | error = copyinstr(km->name, name, KFILTER_MAXNAME, NULL); | | 1339 | error = copyinstr(km->name, name, KFILTER_MAXNAME, NULL); |
1336 | if (error) { | | 1340 | if (error) { |
1337 | break; | | 1341 | break; |
1338 | } | | 1342 | } |
1339 | rw_enter(&kqueue_filter_lock, RW_READER); | | 1343 | rw_enter(&kqueue_filter_lock, RW_READER); |
1340 | kfilter = kfilter_byname(name); | | 1344 | kfilter = kfilter_byname(name); |
1341 | if (kfilter != NULL) | | 1345 | if (kfilter != NULL) |
1342 | km->filter = kfilter->filter; | | 1346 | km->filter = kfilter->filter; |
1343 | else | | 1347 | else |
1344 | error = ENOENT; | | 1348 | error = ENOENT; |
1345 | rw_exit(&kqueue_filter_lock); | | 1349 | rw_exit(&kqueue_filter_lock); |
1346 | break; | | 1350 | break; |
1347 | | | 1351 | |
1348 | default: | | 1352 | default: |
1349 | error = ENOTTY; | | 1353 | error = ENOTTY; |
1350 | break; | | 1354 | break; |
1351 | | | 1355 | |
1352 | } | | 1356 | } |
1353 | kmem_free(name, KFILTER_MAXNAME); | | 1357 | kmem_free(name, KFILTER_MAXNAME); |
1354 | return (error); | | 1358 | return (error); |
1355 | } | | 1359 | } |
1356 | | | 1360 | |
1357 | /* | | 1361 | /* |
1358 | * fileops fcntl method for a kqueue descriptor. | | 1362 | * fileops fcntl method for a kqueue descriptor. |
1359 | */ | | 1363 | */ |
1360 | static int | | 1364 | static int |
1361 | kqueue_fcntl(file_t *fp, u_int com, void *data) | | 1365 | kqueue_fcntl(file_t *fp, u_int com, void *data) |
1362 | { | | 1366 | { |
1363 | | | 1367 | |
1364 | return (ENOTTY); | | 1368 | return (ENOTTY); |
1365 | } | | 1369 | } |
1366 | | | 1370 | |
1367 | /* | | 1371 | /* |
1368 | * fileops poll method for a kqueue descriptor. | | 1372 | * fileops poll method for a kqueue descriptor. |
1369 | * Determine if kqueue has events pending. | | 1373 | * Determine if kqueue has events pending. |
1370 | */ | | 1374 | */ |
1371 | static int | | 1375 | static int |
1372 | kqueue_poll(file_t *fp, int events) | | 1376 | kqueue_poll(file_t *fp, int events) |
1373 | { | | 1377 | { |
1374 | struct kqueue *kq; | | 1378 | struct kqueue *kq; |
1375 | int revents; | | 1379 | int revents; |
1376 | | | 1380 | |
1377 | kq = fp->f_data; | | 1381 | kq = fp->f_data; |
1378 | | | 1382 | |
1379 | revents = 0; | | 1383 | revents = 0; |
1380 | if (events & (POLLIN | POLLRDNORM)) { | | 1384 | if (events & (POLLIN | POLLRDNORM)) { |
1381 | mutex_spin_enter(&kq->kq_lock); | | 1385 | mutex_spin_enter(&kq->kq_lock); |
1382 | if (kq->kq_count != 0) { | | 1386 | if (kq->kq_count != 0) { |
1383 | revents |= events & (POLLIN | POLLRDNORM); | | 1387 | revents |= events & (POLLIN | POLLRDNORM); |
1384 | } else { | | 1388 | } else { |
1385 | selrecord(curlwp, &kq->kq_sel); | | 1389 | selrecord(curlwp, &kq->kq_sel); |
1386 | } | | 1390 | } |
1387 | kq_check(kq); | | 1391 | kq_check(kq); |
1388 | mutex_spin_exit(&kq->kq_lock); | | 1392 | mutex_spin_exit(&kq->kq_lock); |
1389 | } | | 1393 | } |
1390 | | | 1394 | |
1391 | return revents; | | 1395 | return revents; |
1392 | } | | 1396 | } |
1393 | | | 1397 | |
1394 | /* | | 1398 | /* |
1395 | * fileops stat method for a kqueue descriptor. | | 1399 | * fileops stat method for a kqueue descriptor. |
1396 | * Returns dummy info, with st_size being number of events pending. | | 1400 | * Returns dummy info, with st_size being number of events pending. |
1397 | */ | | 1401 | */ |
1398 | static int | | 1402 | static int |
1399 | kqueue_stat(file_t *fp, struct stat *st) | | 1403 | kqueue_stat(file_t *fp, struct stat *st) |
1400 | { | | 1404 | { |
1401 | struct kqueue *kq; | | 1405 | struct kqueue *kq; |
1402 | | | 1406 | |
1403 | kq = fp->f_data; | | 1407 | kq = fp->f_data; |
1404 | | | 1408 | |
1405 | memset(st, 0, sizeof(*st)); | | 1409 | memset(st, 0, sizeof(*st)); |
1406 | st->st_size = kq->kq_count; | | 1410 | st->st_size = kq->kq_count; |
1407 | st->st_blksize = sizeof(struct kevent); | | 1411 | st->st_blksize = sizeof(struct kevent); |
1408 | st->st_mode = S_IFIFO; | | 1412 | st->st_mode = S_IFIFO; |
1409 | | | 1413 | |
1410 | return 0; | | 1414 | return 0; |
1411 | } | | 1415 | } |
1412 | | | 1416 | |
1413 | static void | | 1417 | static void |
1414 | kqueue_doclose(struct kqueue *kq, struct klist *list, int fd) | | 1418 | kqueue_doclose(struct kqueue *kq, struct klist *list, int fd) |
1415 | { | | 1419 | { |
1416 | struct knote *kn; | | 1420 | struct knote *kn; |
1417 | filedesc_t *fdp; | | 1421 | filedesc_t *fdp; |
1418 | | | 1422 | |
1419 | fdp = kq->kq_fdp; | | 1423 | fdp = kq->kq_fdp; |
1420 | | | 1424 | |
1421 | KASSERT(mutex_owned(&fdp->fd_lock)); | | 1425 | KASSERT(mutex_owned(&fdp->fd_lock)); |
1422 | | | 1426 | |
1423 | for (kn = SLIST_FIRST(list); kn != NULL;) { | | 1427 | for (kn = SLIST_FIRST(list); kn != NULL;) { |
1424 | if (kq != kn->kn_kq) { | | 1428 | if (kq != kn->kn_kq) { |
1425 | kn = SLIST_NEXT(kn, kn_link); | | 1429 | kn = SLIST_NEXT(kn, kn_link); |
1426 | continue; | | 1430 | continue; |
1427 | } | | 1431 | } |
1428 | knote_detach(kn, fdp, true); | | 1432 | knote_detach(kn, fdp, true); |
1429 | mutex_enter(&fdp->fd_lock); | | 1433 | mutex_enter(&fdp->fd_lock); |
1430 | kn = SLIST_FIRST(list); | | 1434 | kn = SLIST_FIRST(list); |
1431 | } | | 1435 | } |
1432 | } | | 1436 | } |
1433 | | | 1437 | |
1434 | | | 1438 | |
1435 | /* | | 1439 | /* |
1436 | * fileops close method for a kqueue descriptor. | | 1440 | * fileops close method for a kqueue descriptor. |
1437 | */ | | 1441 | */ |
1438 | static int | | 1442 | static int |
1439 | kqueue_close(file_t *fp) | | 1443 | kqueue_close(file_t *fp) |
1440 | { | | 1444 | { |
1441 | struct kqueue *kq; | | 1445 | struct kqueue *kq; |
1442 | filedesc_t *fdp; | | 1446 | filedesc_t *fdp; |
1443 | fdfile_t *ff; | | 1447 | fdfile_t *ff; |
1444 | int i; | | 1448 | int i; |
1445 | | | 1449 | |
1446 | kq = fp->f_data; | | 1450 | kq = fp->f_data; |
1447 | fp->f_data = NULL; | | 1451 | fp->f_data = NULL; |
1448 | fp->f_type = 0; | | 1452 | fp->f_type = 0; |
1449 | fdp = curlwp->l_fd; | | 1453 | fdp = curlwp->l_fd; |
1450 | | | 1454 | |
1451 | mutex_enter(&fdp->fd_lock); | | 1455 | mutex_enter(&fdp->fd_lock); |
1452 | for (i = 0; i <= fdp->fd_lastkqfile; i++) { | | 1456 | for (i = 0; i <= fdp->fd_lastkqfile; i++) { |
1453 | if ((ff = fdp->fd_dt->dt_ff[i]) == NULL) | | 1457 | if ((ff = fdp->fd_dt->dt_ff[i]) == NULL) |
1454 | continue; | | 1458 | continue; |
1455 | kqueue_doclose(kq, (struct klist *)&ff->ff_knlist, i); | | 1459 | kqueue_doclose(kq, (struct klist *)&ff->ff_knlist, i); |
1456 | } | | 1460 | } |
1457 | if (fdp->fd_knhashmask != 0) { | | 1461 | if (fdp->fd_knhashmask != 0) { |
1458 | for (i = 0; i < fdp->fd_knhashmask + 1; i++) { | | 1462 | for (i = 0; i < fdp->fd_knhashmask + 1; i++) { |
1459 | kqueue_doclose(kq, &fdp->fd_knhash[i], -1); | | 1463 | kqueue_doclose(kq, &fdp->fd_knhash[i], -1); |
1460 | } | | 1464 | } |
1461 | } | | 1465 | } |
1462 | mutex_exit(&fdp->fd_lock); | | 1466 | mutex_exit(&fdp->fd_lock); |
1463 | | | 1467 | |
1464 | KASSERT(kq->kq_count == 0); | | 1468 | KASSERT(kq->kq_count == 0); |
1465 | mutex_destroy(&kq->kq_lock); | | 1469 | mutex_destroy(&kq->kq_lock); |
1466 | cv_destroy(&kq->kq_cv); | | 1470 | cv_destroy(&kq->kq_cv); |
1467 | seldestroy(&kq->kq_sel); | | 1471 | seldestroy(&kq->kq_sel); |
1468 | kmem_free(kq, sizeof(*kq)); | | 1472 | kmem_free(kq, sizeof(*kq)); |
1469 | | | 1473 | |
1470 | return (0); | | 1474 | return (0); |
1471 | } | | 1475 | } |
1472 | | | 1476 | |
1473 | /* | | 1477 | /* |
1474 | * struct fileops kqfilter method for a kqueue descriptor. | | 1478 | * struct fileops kqfilter method for a kqueue descriptor. |
1475 | * Event triggered when monitored kqueue changes. | | 1479 | * Event triggered when monitored kqueue changes. |
1476 | */ | | 1480 | */ |
1477 | static int | | 1481 | static int |
1478 | kqueue_kqfilter(file_t *fp, struct knote *kn) | | 1482 | kqueue_kqfilter(file_t *fp, struct knote *kn) |
1479 | { | | 1483 | { |
1480 | struct kqueue *kq; | | 1484 | struct kqueue *kq; |
1481 | | | 1485 | |
1482 | kq = ((file_t *)kn->kn_obj)->f_data; | | 1486 | kq = ((file_t *)kn->kn_obj)->f_data; |
1483 | | | 1487 | |
1484 | KASSERT(fp == kn->kn_obj); | | 1488 | KASSERT(fp == kn->kn_obj); |
1485 | | | 1489 | |
1486 | if (kn->kn_filter != EVFILT_READ) | | 1490 | if (kn->kn_filter != EVFILT_READ) |
1487 | return 1; | | 1491 | return 1; |
1488 | | | 1492 | |
1489 | kn->kn_fop = &kqread_filtops; | | 1493 | kn->kn_fop = &kqread_filtops; |
1490 | mutex_enter(&kq->kq_lock); | | 1494 | mutex_enter(&kq->kq_lock); |
1491 | SLIST_INSERT_HEAD(&kq->kq_sel.sel_klist, kn, kn_selnext); | | 1495 | SLIST_INSERT_HEAD(&kq->kq_sel.sel_klist, kn, kn_selnext); |
1492 | mutex_exit(&kq->kq_lock); | | 1496 | mutex_exit(&kq->kq_lock); |
1493 | | | 1497 | |
1494 | return 0; | | 1498 | return 0; |
1495 | } | | 1499 | } |
1496 | | | 1500 | |
1497 | | | 1501 | |
1498 | /* | | 1502 | /* |
1499 | * Walk down a list of knotes, activating them if their event has | | 1503 | * Walk down a list of knotes, activating them if their event has |
1500 | * triggered. The caller's object lock (e.g. device driver lock) | | 1504 | * triggered. The caller's object lock (e.g. device driver lock) |
1501 | * must be held. | | 1505 | * must be held. |
1502 | */ | | 1506 | */ |
1503 | void | | 1507 | void |
1504 | knote(struct klist *list, long hint) | | 1508 | knote(struct klist *list, long hint) |
1505 | { | | 1509 | { |
1506 | struct knote *kn, *tmpkn; | | 1510 | struct knote *kn, *tmpkn; |
1507 | | | 1511 | |
1508 | SLIST_FOREACH_SAFE(kn, list, kn_selnext, tmpkn) { | | 1512 | SLIST_FOREACH_SAFE(kn, list, kn_selnext, tmpkn) { |
1509 | if ((*kn->kn_fop->f_event)(kn, hint)) | | 1513 | if ((*kn->kn_fop->f_event)(kn, hint)) |
1510 | knote_activate(kn); | | 1514 | knote_activate(kn); |
1511 | } | | 1515 | } |
1512 | } | | 1516 | } |
1513 | | | 1517 | |
1514 | /* | | 1518 | /* |
1515 | * Remove all knotes referencing a specified fd | | 1519 | * Remove all knotes referencing a specified fd |
1516 | */ | | 1520 | */ |
1517 | void | | 1521 | void |
1518 | knote_fdclose(int fd) | | 1522 | knote_fdclose(int fd) |
1519 | { | | 1523 | { |
1520 | struct klist *list; | | 1524 | struct klist *list; |
1521 | struct knote *kn; | | 1525 | struct knote *kn; |
1522 | filedesc_t *fdp; | | 1526 | filedesc_t *fdp; |
1523 | | | 1527 | |
1524 | fdp = curlwp->l_fd; | | 1528 | fdp = curlwp->l_fd; |
1525 | list = (struct klist *)&fdp->fd_dt->dt_ff[fd]->ff_knlist; | | 1529 | list = (struct klist *)&fdp->fd_dt->dt_ff[fd]->ff_knlist; |
1526 | mutex_enter(&fdp->fd_lock); | | 1530 | mutex_enter(&fdp->fd_lock); |
1527 | while ((kn = SLIST_FIRST(list)) != NULL) { | | 1531 | while ((kn = SLIST_FIRST(list)) != NULL) { |
1528 | knote_detach(kn, fdp, true); | | 1532 | knote_detach(kn, fdp, true); |
1529 | mutex_enter(&fdp->fd_lock); | | 1533 | mutex_enter(&fdp->fd_lock); |
1530 | } | | 1534 | } |
1531 | mutex_exit(&fdp->fd_lock); | | 1535 | mutex_exit(&fdp->fd_lock); |
1532 | } | | 1536 | } |
1533 | | | 1537 | |
1534 | /* | | 1538 | /* |
1535 | * Drop knote. Called with fdp->fd_lock held, and will drop before | | 1539 | * Drop knote. Called with fdp->fd_lock held, and will drop before |
1536 | * returning. | | 1540 | * returning. |
1537 | */ | | 1541 | */ |
1538 | static void | | 1542 | static void |
1539 | knote_detach(struct knote *kn, filedesc_t *fdp, bool dofop) | | 1543 | knote_detach(struct knote *kn, filedesc_t *fdp, bool dofop) |
1540 | { | | 1544 | { |
1541 | struct klist *list; | | 1545 | struct klist *list; |
1542 | struct kqueue *kq; | | 1546 | struct kqueue *kq; |
1543 | | | 1547 | |
1544 | kq = kn->kn_kq; | | 1548 | kq = kn->kn_kq; |
1545 | | | 1549 | |
1546 | KASSERT((kn->kn_status & KN_MARKER) == 0); | | 1550 | KASSERT((kn->kn_status & KN_MARKER) == 0); |
1547 | KASSERT(mutex_owned(&fdp->fd_lock)); | | 1551 | KASSERT(mutex_owned(&fdp->fd_lock)); |
1548 | | | 1552 | |
1549 | /* Remove from monitored object. */ | | 1553 | /* Remove from monitored object. */ |