kern_syscall.c revision 1.5.12.2 1 /* $NetBSD: kern_syscall.c,v 1.5.12.2 2012/06/02 11:09:33 mrg Exp $ */
2
3 /*-
4 * Copyright (c) 2008 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software developed for The NetBSD Foundation
8 * by Andrew Doran.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 *
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
21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
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
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
29 * POSSIBILITY OF SUCH DAMAGE.
30 */
31
32 #include <sys/cdefs.h>
33 __KERNEL_RCSID(0, "$NetBSD: kern_syscall.c,v 1.5.12.2 2012/06/02 11:09:33 mrg Exp $");
34
35 #include "opt_modular.h"
36
37 /* XXX To get syscall prototypes. */
38 #define SYSVSHM
39 #define SYSVSEM
40 #define SYSVMSG
41
42 #include <sys/param.h>
43 #include <sys/module.h>
44 #include <sys/sched.h>
45 #include <sys/syscall.h>
46 #include <sys/syscallargs.h>
47 #include <sys/syscallvar.h>
48 #include <sys/systm.h>
49 #include <sys/xcall.h>
50
51 int
52 sys_nomodule(struct lwp *l, const void *v, register_t *retval)
53 {
54 #ifdef MODULAR
55 static struct {
56 u_int al_code;
57 const char *al_module;
58 } const autoload[] = {
59 { SYS_aio_cancel, "aio" },
60 { SYS_aio_error, "aio" },
61 { SYS_aio_fsync, "aio" },
62 { SYS_aio_read, "aio" },
63 { SYS_aio_return, "aio" },
64 { SYS___aio_suspend50, "aio" },
65 { SYS_aio_write, "aio" },
66 { SYS_lio_listio, "aio" },
67 { SYS_mq_open, "mqueue" },
68 { SYS_mq_close, "mqueue" },
69 { SYS_mq_unlink, "mqueue" },
70 { SYS_mq_getattr, "mqueue" },
71 { SYS_mq_setattr, "mqueue" },
72 { SYS_mq_notify, "mqueue" },
73 { SYS_mq_send, "mqueue" },
74 { SYS_mq_receive, "mqueue" },
75 { SYS___mq_timedsend50, "mqueue" },
76 { SYS___mq_timedreceive50, "mqueue" },
77 { SYS_compat_43_fstat43, "compat" },
78 { SYS_compat_43_lstat43, "compat" },
79 { SYS_compat_43_oaccept, "compat" },
80 { SYS_compat_43_ocreat, "compat" },
81 { SYS_compat_43_oftruncate, "compat" },
82 { SYS_compat_43_ogetdirentries, "compat" },
83 { SYS_compat_43_ogetdtablesize, "compat" },
84 { SYS_compat_43_ogethostid, "compat" },
85 { SYS_compat_43_ogethostname, "compat" },
86 { SYS_compat_43_ogetkerninfo, "compat" },
87 { SYS_compat_43_ogetpagesize, "compat" },
88 { SYS_compat_43_ogetpeername, "compat" },
89 { SYS_compat_43_ogetrlimit, "compat" },
90 { SYS_compat_43_ogetsockname, "compat" },
91 { SYS_compat_43_okillpg, "compat" },
92 { SYS_compat_43_olseek, "compat" },
93 { SYS_compat_43_ommap, "compat" },
94 { SYS_compat_43_oquota, "compat" },
95 { SYS_compat_43_orecv, "compat" },
96 { SYS_compat_43_orecvfrom, "compat" },
97 { SYS_compat_43_orecvmsg, "compat" },
98 { SYS_compat_43_osend, "compat" },
99 { SYS_compat_43_osendmsg, "compat" },
100 { SYS_compat_43_osethostid, "compat" },
101 { SYS_compat_43_osethostname, "compat" },
102 { SYS_compat_43_osetrlimit, "compat" },
103 { SYS_compat_43_osigblock, "compat" },
104 { SYS_compat_43_osigsetmask, "compat" },
105 { SYS_compat_43_osigstack, "compat" },
106 { SYS_compat_43_osigvec, "compat" },
107 { SYS_compat_43_otruncate, "compat" },
108 { SYS_compat_43_owait, "compat" },
109 { SYS_compat_43_stat43, "compat" },
110 { SYS_compat_09_ogetdomainname, "compat" },
111 { SYS_compat_09_osetdomainname, "compat" },
112 { SYS_compat_09_ouname, "compat" },
113 #ifndef _LP64
114 { SYS_compat_10_omsgsys, "compat" },
115 { SYS_compat_10_osemsys, "compat" },
116 { SYS_compat_10_oshmsys, "compat" },
117 #endif
118 { SYS_compat_12_fstat12, "compat" },
119 { SYS_compat_12_getdirentries, "compat" },
120 { SYS_compat_12_lstat12, "compat" },
121 { SYS_compat_12_msync, "compat" },
122 { SYS_compat_12_oreboot, "compat" },
123 { SYS_compat_12_oswapon, "compat" },
124 { SYS_compat_12_stat12, "compat" },
125 { SYS_compat_13_sigaction13, "compat" },
126 { SYS_compat_13_sigaltstack13, "compat" },
127 { SYS_compat_13_sigpending13, "compat" },
128 { SYS_compat_13_sigprocmask13, "compat" },
129 { SYS_compat_13_sigreturn13, "compat" },
130 { SYS_compat_13_sigsuspend13, "compat" },
131 { SYS_compat_14___semctl, "compat" },
132 { SYS_compat_14_msgctl, "compat" },
133 { SYS_compat_14_shmctl, "compat" },
134 { SYS_compat_16___sigaction14, "compat" },
135 { SYS_compat_16___sigreturn14, "compat" },
136 { SYS_compat_20_fhstatfs, "compat" },
137 { SYS_compat_20_fstatfs, "compat" },
138 { SYS_compat_20_getfsstat, "compat" },
139 { SYS_compat_20_statfs, "compat" },
140 { SYS_compat_30___fhstat30, "compat" },
141 { SYS_compat_30___fstat13, "compat" },
142 { SYS_compat_30___lstat13, "compat" },
143 { SYS_compat_30___stat13, "compat" },
144 { SYS_compat_30_fhopen, "compat" },
145 { SYS_compat_30_fhstat, "compat" },
146 { SYS_compat_30_fhstatvfs1, "compat" },
147 { SYS_compat_30_getdents, "compat" },
148 { SYS_compat_30_getfh, "compat" },
149 { SYS_compat_30_socket, "compat" },
150 { SYS_compat_40_mount, "compat" },
151 { SYS_compat_50_wait4, "compat" },
152 { SYS_compat_50_mknod, "compat" },
153 { SYS_compat_50_setitimer, "compat" },
154 { SYS_compat_50_getitimer, "compat" },
155 { SYS_compat_50_select, "compat" },
156 { SYS_compat_50_gettimeofday, "compat" },
157 { SYS_compat_50_getrusage, "compat" },
158 { SYS_compat_50_settimeofday, "compat" },
159 { SYS_compat_50_utimes, "compat" },
160 { SYS_compat_50_adjtime, "compat" },
161 { SYS_compat_50_lfs_segwait, "compat" },
162 { SYS_compat_50_futimes, "compat" },
163 { SYS_compat_50_clock_gettime, "compat" },
164 { SYS_compat_50_clock_settime, "compat" },
165 { SYS_compat_50_clock_getres, "compat" },
166 { SYS_compat_50_timer_settime, "compat" },
167 { SYS_compat_50_timer_gettime, "compat" },
168 { SYS_compat_50_nanosleep, "compat" },
169 { SYS_compat_50___sigtimedwait, "compat" },
170 { SYS_compat_50_mq_timedsend, "compat" },
171 { SYS_compat_50_mq_timedreceive, "compat" },
172 { SYS_compat_50_lutimes, "compat" },
173 { SYS_compat_50_____semctl13, "compat" },
174 { SYS_compat_50___msgctl13, "compat" },
175 { SYS_compat_50___shmctl13, "compat" },
176 { SYS_compat_50__lwp_park, "compat" },
177 { SYS_compat_50_kevent, "compat" },
178 { SYS_compat_50_pselect, "compat" },
179 { SYS_compat_50_pollts, "compat" },
180 { SYS_compat_50___stat30, "compat" },
181 { SYS_compat_50___fstat30, "compat" },
182 { SYS_compat_50___lstat30, "compat" },
183 { SYS_compat_50___ntp_gettime30, "compat" },
184 { SYS_compat_50___fhstat40, "compat" },
185 { SYS_compat_50_aio_suspend, "compat" },
186 { SYS__ksem_init, "ksem" },
187 { SYS__ksem_open, "ksem" },
188 { SYS__ksem_unlink, "ksem" },
189 { SYS__ksem_close, "ksem" },
190 { SYS__ksem_post, "ksem" },
191 { SYS__ksem_wait, "ksem" },
192 { SYS__ksem_trywait, "ksem" },
193 { SYS__ksem_getvalue, "ksem" },
194 { SYS__ksem_destroy, "ksem" },
195 { SYS__ksem_timedwait, "ksem" },
196 { SYS_nfssvc, "nfsserver" },
197 { SYS_afssys, "openafs" },
198 };
199 const struct sysent *sy;
200 const struct emul *em;
201 int code, i;
202
203 /*
204 * Restart the syscall if we interrupted a module unload that
205 * failed. Acquiring kernconfig_lock delays us until any unload
206 * has been completed or rolled back.
207 */
208 kernconfig_lock();
209 sy = l->l_sysent;
210 if (sy->sy_call != sys_nomodule) {
211 kernconfig_unlock();
212 return ERESTART;
213 }
214 /*
215 * Try to autoload a module to satisfy the request. If it
216 * works, retry the request.
217 */
218 em = l->l_proc->p_emul;
219 if (em == &emul_netbsd) {
220 code = sy - em->e_sysent;
221 for (i = 0; i < __arraycount(autoload); i++) {
222 if (autoload[i].al_code != code) {
223 continue;
224 }
225 if (module_autoload(autoload[i].al_module,
226 MODULE_CLASS_ANY) != 0 ||
227 sy->sy_call == sys_nomodule) {
228 break;
229 }
230 kernconfig_unlock();
231 return ERESTART;
232 }
233 }
234 kernconfig_unlock();
235 #endif /* MODULAR */
236
237 return sys_nosys(l, v, retval);
238 }
239
240 int
241 syscall_establish(const struct emul *em, const struct syscall_package *sp)
242 {
243 struct sysent *sy;
244 int i;
245
246 KASSERT(kernconfig_is_held());
247
248 if (em == NULL) {
249 em = &emul_netbsd;
250 }
251 sy = em->e_sysent;
252
253 /*
254 * Ensure that all preconditions are valid, since this is
255 * an all or nothing deal. Once a system call is entered,
256 * it can become busy and we could be unable to remove it
257 * on error.
258 */
259 for (i = 0; sp[i].sp_call != NULL; i++) {
260 if (sy[sp[i].sp_code].sy_call != sys_nomodule) {
261 #ifdef DIAGNOSTIC
262 printf("syscall %d is busy\n", sp[i].sp_code);
263 #endif
264 return EBUSY;
265 }
266 }
267 /* Everything looks good, patch them in. */
268 for (i = 0; sp[i].sp_call != NULL; i++) {
269 sy[sp[i].sp_code].sy_call = sp[i].sp_call;
270 }
271
272 return 0;
273 }
274
275 int
276 syscall_disestablish(const struct emul *em, const struct syscall_package *sp)
277 {
278 struct sysent *sy;
279 uint64_t where;
280 lwp_t *l;
281 int i;
282
283 KASSERT(kernconfig_is_held());
284
285 if (em == NULL) {
286 em = &emul_netbsd;
287 }
288 sy = em->e_sysent;
289
290 /*
291 * First, patch the system calls to sys_nomodule to gate further
292 * activity.
293 */
294 for (i = 0; sp[i].sp_call != NULL; i++) {
295 KASSERT(sy[sp[i].sp_code].sy_call == sp[i].sp_call);
296 sy[sp[i].sp_code].sy_call = sys_nomodule;
297 }
298
299 /*
300 * Run a cross call to cycle through all CPUs. This does two
301 * things: lock activity provides a barrier and makes our update
302 * of sy_call visible to all CPUs, and upon return we can be sure
303 * that we see pertinent values of l_sysent posted by remote CPUs.
304 */
305 where = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL);
306 xc_wait(where);
307
308 /*
309 * Now it's safe to check l_sysent. Run through all LWPs and see
310 * if anyone is still using the system call.
311 */
312 for (i = 0; sp[i].sp_call != NULL; i++) {
313 mutex_enter(proc_lock);
314 LIST_FOREACH(l, &alllwp, l_list) {
315 if (l->l_sysent == &sy[sp[i].sp_code]) {
316 break;
317 }
318 }
319 mutex_exit(proc_lock);
320 if (l == NULL) {
321 continue;
322 }
323 /*
324 * We lose: one or more calls are still in use. Put back
325 * the old entrypoints and act like nothing happened.
326 * When we drop kernconfig_lock, any system calls held in
327 * sys_nomodule() will be restarted.
328 */
329 for (i = 0; sp[i].sp_call != NULL; i++) {
330 sy[sp[i].sp_code].sy_call = sp[i].sp_call;
331 }
332 return EBUSY;
333 }
334
335 return 0;
336 }
337