xref: /freebsd/sys/kern/kern_syscalls.c (revision 3982006ed5587cfd83cc1955186e0aa4b92b3fcd)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 1999 Assar Westerlund
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 #include <sys/param.h>
33 #include <sys/kernel.h>
34 #include <sys/lock.h>
35 #include <sys/module.h>
36 #include <sys/mutex.h>
37 #include <sys/proc.h>
38 #include <sys/resourcevar.h>
39 #include <sys/sx.h>
40 #include <sys/syscall.h>
41 #include <sys/sysent.h>
42 #include <sys/sysproto.h>
43 #include <sys/systm.h>
44 #include <machine/atomic.h>
45 
46 /*
47  * Acts like "nosys" but can be identified in sysent for dynamic call
48  * number assignment for a limited number of calls.
49  *
50  * Place holder for system call slots reserved for loadable modules.
51  */
52 int
53 lkmnosys(struct thread *td, struct nosys_args *args)
54 {
55 
56 	return (nosys(td, args));
57 }
58 
59 int
60 lkmressys(struct thread *td, struct nosys_args *args)
61 {
62 
63 	return (nosys(td, args));
64 }
65 
66 static void
67 syscall_thread_drain(struct sysent *se)
68 {
69 	u_int32_t cnt, oldcnt;
70 
71 	do {
72 		oldcnt = se->sy_thrcnt;
73 		KASSERT((oldcnt & SY_THR_STATIC) == 0,
74 		    ("drain on static syscall"));
75 		cnt = oldcnt | SY_THR_DRAINING;
76 	} while (atomic_cmpset_acq_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
77 	while (atomic_cmpset_32(&se->sy_thrcnt, SY_THR_DRAINING,
78 	    SY_THR_ABSENT) == 0)
79 		pause("scdrn", hz/2);
80 }
81 
82 int
83 syscall_thread_enter(struct thread *td, struct sysent *se)
84 {
85 	u_int32_t cnt, oldcnt;
86 
87 	do {
88 		oldcnt = se->sy_thrcnt;
89 		if ((oldcnt & SY_THR_STATIC) != 0)
90 			return (0);
91 		if ((oldcnt & (SY_THR_DRAINING | SY_THR_ABSENT)) != 0)
92 			return (ENOSYS);
93 		cnt = oldcnt + SY_THR_INCR;
94 	} while (atomic_cmpset_acq_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
95 	return (0);
96 }
97 
98 void
99 syscall_thread_exit(struct thread *td, struct sysent *se)
100 {
101 	u_int32_t cnt, oldcnt;
102 
103 	do {
104 		oldcnt = se->sy_thrcnt;
105 		if ((oldcnt & SY_THR_STATIC) != 0)
106 			return;
107 		cnt = oldcnt - SY_THR_INCR;
108 	} while (atomic_cmpset_rel_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
109 }
110 
111 int
112 kern_syscall_register(struct sysent *sysents, int *offset,
113     struct sysent *new_sysent, struct sysent *old_sysent, int flags)
114 {
115 	int i;
116 
117 	if ((flags & ~SY_THR_STATIC) != 0)
118 		return (EINVAL);
119 
120 	if (*offset == NO_SYSCALL) {
121 		for (i = 1; i < SYS_MAXSYSCALL; ++i)
122 			if (sysents[i].sy_call == (sy_call_t *)lkmnosys)
123 				break;
124 		if (i == SYS_MAXSYSCALL)
125 			return (ENFILE);
126 		*offset = i;
127 	} else if (*offset < 0 || *offset >= SYS_MAXSYSCALL)
128 		return (EINVAL);
129 	else if (sysents[*offset].sy_call != (sy_call_t *)lkmnosys &&
130 	    sysents[*offset].sy_call != (sy_call_t *)lkmressys)
131 		return (EEXIST);
132 
133 	KASSERT(sysents[*offset].sy_thrcnt == SY_THR_ABSENT,
134 	    ("dynamic syscall is not protected"));
135 	*old_sysent = sysents[*offset];
136 	new_sysent->sy_thrcnt = SY_THR_ABSENT;
137 	sysents[*offset] = *new_sysent;
138 	atomic_store_rel_32(&sysents[*offset].sy_thrcnt, flags);
139 	return (0);
140 }
141 
142 int
143 kern_syscall_deregister(struct sysent *sysents, int offset,
144     const struct sysent *old_sysent)
145 {
146 	struct sysent *se;
147 
148 	if (offset == 0)
149 		return (0); /* XXX? */
150 
151 	se = &sysents[offset];
152 	if ((se->sy_thrcnt & SY_THR_STATIC) != 0)
153 		return (EINVAL);
154 	syscall_thread_drain(se);
155 	sysent[offset] = *old_sysent;
156 	return (0);
157 }
158 
159 int
160 syscall_module_handler(struct module *mod, int what, void *arg)
161 {
162 
163 	return (kern_syscall_module_handler(sysent, mod, what, arg));
164 }
165 
166 int
167 kern_syscall_module_handler(struct sysent *sysents, struct module *mod,
168     int what, void *arg)
169 {
170 	struct syscall_module_data *data = arg;
171 	modspecific_t ms;
172 	int error;
173 
174 	switch (what) {
175 	case MOD_LOAD:
176 		error = kern_syscall_register(sysents, data->offset,
177 		    data->new_sysent, &data->old_sysent, data->flags);
178 		if (error) {
179 			/* Leave a mark so we know to safely unload below. */
180 			data->offset = NULL;
181 			return (error);
182 		}
183 		ms.intval = *data->offset;
184 		MOD_XLOCK;
185 		module_setspecific(mod, &ms);
186 		MOD_XUNLOCK;
187 		if (data->chainevh)
188 			error = data->chainevh(mod, what, data->chainarg);
189 		return (error);
190 	case MOD_UNLOAD:
191 		/*
192 		 * MOD_LOAD failed, so just return without calling the
193 		 * chained handler since we didn't pass along the MOD_LOAD
194 		 * event.
195 		 */
196 		if (data->offset == NULL)
197 			return (0);
198 		if (data->chainevh) {
199 			error = data->chainevh(mod, what, data->chainarg);
200 			if (error)
201 				return error;
202 		}
203 		error = kern_syscall_deregister(sysents, *data->offset,
204 		    &data->old_sysent);
205 		return (error);
206 	default:
207 		if (data->chainevh)
208 			return (data->chainevh(mod, what, data->chainarg));
209 		return (EOPNOTSUPP);
210 	}
211 
212 	/* NOTREACHED */
213 }
214 
215 int
216 syscall_helper_register(struct syscall_helper_data *sd, int flags)
217 {
218 
219 	return (kern_syscall_helper_register(sysent, sd, flags));
220 }
221 
222 int
223 kern_syscall_helper_register(struct sysent *sysents,
224     struct syscall_helper_data *sd, int flags)
225 {
226 	struct syscall_helper_data *sd1;
227 	int error;
228 
229 	for (sd1 = sd; sd1->syscall_no != NO_SYSCALL; sd1++) {
230 		error = kern_syscall_register(sysents, &sd1->syscall_no,
231 		    &sd1->new_sysent, &sd1->old_sysent, flags);
232 		if (error != 0) {
233 			kern_syscall_helper_unregister(sysents, sd);
234 			return (error);
235 		}
236 		sd1->registered = 1;
237 	}
238 	return (0);
239 }
240 
241 int
242 syscall_helper_unregister(struct syscall_helper_data *sd)
243 {
244 
245 	return (kern_syscall_helper_unregister(sysent, sd));
246 }
247 
248 int
249 kern_syscall_helper_unregister(struct sysent *sysents,
250     struct syscall_helper_data *sd)
251 {
252 	struct syscall_helper_data *sd1;
253 
254 	for (sd1 = sd; sd1->registered != 0; sd1++) {
255 		kern_syscall_deregister(sysents, sd1->syscall_no,
256 		    &sd1->old_sysent);
257 		sd1->registered = 0;
258 	}
259 	return (0);
260 }
261