xref: /freebsd/sys/kern/kern_syscalls.c (revision f7c32ed617858bcd22f8d1b03199099d50125721)
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 	KASSERT((se->sy_thrcnt & SY_THR_STATIC) == 0,
88 	    ("%s: not a static syscall", __func__));
89 
90 	do {
91 		oldcnt = se->sy_thrcnt;
92 		if ((oldcnt & (SY_THR_DRAINING | SY_THR_ABSENT)) != 0)
93 			return (ENOSYS);
94 		cnt = oldcnt + SY_THR_INCR;
95 	} while (atomic_cmpset_acq_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
96 	return (0);
97 }
98 
99 void
100 syscall_thread_exit(struct thread *td, struct sysent *se)
101 {
102 	u_int32_t cnt, oldcnt;
103 
104 	KASSERT((se->sy_thrcnt & SY_THR_STATIC) == 0,
105 	    ("%s: not a static syscall", __func__));
106 
107 	do {
108 		oldcnt = se->sy_thrcnt;
109 		cnt = oldcnt - SY_THR_INCR;
110 	} while (atomic_cmpset_rel_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
111 }
112 
113 int
114 kern_syscall_register(struct sysent *sysents, int *offset,
115     struct sysent *new_sysent, struct sysent *old_sysent, int flags)
116 {
117 	int i;
118 
119 	if ((flags & ~SY_THR_STATIC) != 0)
120 		return (EINVAL);
121 
122 	if (*offset == NO_SYSCALL) {
123 		for (i = 1; i < SYS_MAXSYSCALL; ++i)
124 			if (sysents[i].sy_call == (sy_call_t *)lkmnosys)
125 				break;
126 		if (i == SYS_MAXSYSCALL)
127 			return (ENFILE);
128 		*offset = i;
129 	} else if (*offset < 0 || *offset >= SYS_MAXSYSCALL) {
130 		return (EINVAL);
131 	} else if (sysents[*offset].sy_call != (sy_call_t *)lkmnosys &&
132 	    sysents[*offset].sy_call != (sy_call_t *)lkmressys) {
133 		KASSERT(sysents[*offset].sy_call != NULL,
134 		    ("undefined syscall %d", *offset));
135 		return (EEXIST);
136 	}
137 
138 	KASSERT(sysents[*offset].sy_thrcnt == SY_THR_ABSENT,
139 	    ("dynamic syscall is not protected"));
140 	*old_sysent = sysents[*offset];
141 	new_sysent->sy_thrcnt = SY_THR_ABSENT;
142 	sysents[*offset] = *new_sysent;
143 	atomic_store_rel_32(&sysents[*offset].sy_thrcnt, flags);
144 	return (0);
145 }
146 
147 int
148 kern_syscall_deregister(struct sysent *sysents, int offset,
149     const struct sysent *old_sysent)
150 {
151 	struct sysent *se;
152 
153 	if (offset == 0)
154 		return (0); /* XXX? */
155 
156 	se = &sysents[offset];
157 	if ((se->sy_thrcnt & SY_THR_STATIC) != 0)
158 		return (EINVAL);
159 	syscall_thread_drain(se);
160 	sysents[offset] = *old_sysent;
161 	return (0);
162 }
163 
164 int
165 syscall_module_handler(struct module *mod, int what, void *arg)
166 {
167 
168 	return (kern_syscall_module_handler(sysent, mod, what, arg));
169 }
170 
171 int
172 kern_syscall_module_handler(struct sysent *sysents, struct module *mod,
173     int what, void *arg)
174 {
175 	struct syscall_module_data *data = arg;
176 	modspecific_t ms;
177 	int error;
178 
179 	switch (what) {
180 	case MOD_LOAD:
181 		error = kern_syscall_register(sysents, data->offset,
182 		    data->new_sysent, &data->old_sysent, data->flags);
183 		if (error) {
184 			/* Leave a mark so we know to safely unload below. */
185 			data->offset = NULL;
186 			return (error);
187 		}
188 		ms.intval = *data->offset;
189 		MOD_XLOCK;
190 		module_setspecific(mod, &ms);
191 		MOD_XUNLOCK;
192 		if (data->chainevh)
193 			error = data->chainevh(mod, what, data->chainarg);
194 		return (error);
195 	case MOD_UNLOAD:
196 		/*
197 		 * MOD_LOAD failed, so just return without calling the
198 		 * chained handler since we didn't pass along the MOD_LOAD
199 		 * event.
200 		 */
201 		if (data->offset == NULL)
202 			return (0);
203 		if (data->chainevh) {
204 			error = data->chainevh(mod, what, data->chainarg);
205 			if (error)
206 				return error;
207 		}
208 		error = kern_syscall_deregister(sysents, *data->offset,
209 		    &data->old_sysent);
210 		return (error);
211 	default:
212 		if (data->chainevh)
213 			return (data->chainevh(mod, what, data->chainarg));
214 		return (EOPNOTSUPP);
215 	}
216 
217 	/* NOTREACHED */
218 }
219 
220 int
221 syscall_helper_register(struct syscall_helper_data *sd, int flags)
222 {
223 
224 	return (kern_syscall_helper_register(sysent, sd, flags));
225 }
226 
227 int
228 kern_syscall_helper_register(struct sysent *sysents,
229     struct syscall_helper_data *sd, int flags)
230 {
231 	struct syscall_helper_data *sd1;
232 	int error;
233 
234 	for (sd1 = sd; sd1->syscall_no != NO_SYSCALL; sd1++) {
235 		error = kern_syscall_register(sysents, &sd1->syscall_no,
236 		    &sd1->new_sysent, &sd1->old_sysent, flags);
237 		if (error != 0) {
238 			kern_syscall_helper_unregister(sysents, sd);
239 			return (error);
240 		}
241 		sd1->registered = 1;
242 	}
243 	return (0);
244 }
245 
246 int
247 syscall_helper_unregister(struct syscall_helper_data *sd)
248 {
249 
250 	return (kern_syscall_helper_unregister(sysent, sd));
251 }
252 
253 int
254 kern_syscall_helper_unregister(struct sysent *sysents,
255     struct syscall_helper_data *sd)
256 {
257 	struct syscall_helper_data *sd1;
258 
259 	for (sd1 = sd; sd1->registered != 0; sd1++) {
260 		kern_syscall_deregister(sysents, sd1->syscall_no,
261 		    &sd1->old_sysent);
262 		sd1->registered = 0;
263 	}
264 	return (0);
265 }
266