xref: /freebsd/sys/kern/kern_syscalls.c (revision f6a3b357e9be4c6423c85eff9a847163a0d307c8)
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_DRAINING | SY_THR_ABSENT)) != 0)
90 			return (ENOSYS);
91 		cnt = oldcnt + SY_THR_INCR;
92 	} while (atomic_cmpset_acq_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
93 	return (0);
94 }
95 
96 void
97 _syscall_thread_exit(struct thread *td, struct sysent *se)
98 {
99 	u_int32_t cnt, oldcnt;
100 
101 	do {
102 		oldcnt = se->sy_thrcnt;
103 		cnt = oldcnt - SY_THR_INCR;
104 	} while (atomic_cmpset_rel_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
105 }
106 
107 int
108 kern_syscall_register(struct sysent *sysents, int *offset,
109     struct sysent *new_sysent, struct sysent *old_sysent, int flags)
110 {
111 	int i;
112 
113 	if ((flags & ~SY_THR_STATIC) != 0)
114 		return (EINVAL);
115 
116 	if (*offset == NO_SYSCALL) {
117 		for (i = 1; i < SYS_MAXSYSCALL; ++i)
118 			if (sysents[i].sy_call == (sy_call_t *)lkmnosys)
119 				break;
120 		if (i == SYS_MAXSYSCALL)
121 			return (ENFILE);
122 		*offset = i;
123 	} else if (*offset < 0 || *offset >= SYS_MAXSYSCALL)
124 		return (EINVAL);
125 	else if (sysents[*offset].sy_call != (sy_call_t *)lkmnosys &&
126 	    sysents[*offset].sy_call != (sy_call_t *)lkmressys)
127 		return (EEXIST);
128 
129 	KASSERT(sysents[*offset].sy_thrcnt == SY_THR_ABSENT,
130 	    ("dynamic syscall is not protected"));
131 	*old_sysent = sysents[*offset];
132 	new_sysent->sy_thrcnt = SY_THR_ABSENT;
133 	sysents[*offset] = *new_sysent;
134 	atomic_store_rel_32(&sysents[*offset].sy_thrcnt, flags);
135 	return (0);
136 }
137 
138 int
139 kern_syscall_deregister(struct sysent *sysents, int offset,
140     const struct sysent *old_sysent)
141 {
142 	struct sysent *se;
143 
144 	if (offset == 0)
145 		return (0); /* XXX? */
146 
147 	se = &sysents[offset];
148 	if ((se->sy_thrcnt & SY_THR_STATIC) != 0)
149 		return (EINVAL);
150 	syscall_thread_drain(se);
151 	sysents[offset] = *old_sysent;
152 	return (0);
153 }
154 
155 int
156 syscall_module_handler(struct module *mod, int what, void *arg)
157 {
158 
159 	return (kern_syscall_module_handler(sysent, mod, what, arg));
160 }
161 
162 int
163 kern_syscall_module_handler(struct sysent *sysents, struct module *mod,
164     int what, void *arg)
165 {
166 	struct syscall_module_data *data = arg;
167 	modspecific_t ms;
168 	int error;
169 
170 	switch (what) {
171 	case MOD_LOAD:
172 		error = kern_syscall_register(sysents, data->offset,
173 		    data->new_sysent, &data->old_sysent, data->flags);
174 		if (error) {
175 			/* Leave a mark so we know to safely unload below. */
176 			data->offset = NULL;
177 			return (error);
178 		}
179 		ms.intval = *data->offset;
180 		MOD_XLOCK;
181 		module_setspecific(mod, &ms);
182 		MOD_XUNLOCK;
183 		if (data->chainevh)
184 			error = data->chainevh(mod, what, data->chainarg);
185 		return (error);
186 	case MOD_UNLOAD:
187 		/*
188 		 * MOD_LOAD failed, so just return without calling the
189 		 * chained handler since we didn't pass along the MOD_LOAD
190 		 * event.
191 		 */
192 		if (data->offset == NULL)
193 			return (0);
194 		if (data->chainevh) {
195 			error = data->chainevh(mod, what, data->chainarg);
196 			if (error)
197 				return error;
198 		}
199 		error = kern_syscall_deregister(sysents, *data->offset,
200 		    &data->old_sysent);
201 		return (error);
202 	default:
203 		if (data->chainevh)
204 			return (data->chainevh(mod, what, data->chainarg));
205 		return (EOPNOTSUPP);
206 	}
207 
208 	/* NOTREACHED */
209 }
210 
211 int
212 syscall_helper_register(struct syscall_helper_data *sd, int flags)
213 {
214 
215 	return (kern_syscall_helper_register(sysent, sd, flags));
216 }
217 
218 int
219 kern_syscall_helper_register(struct sysent *sysents,
220     struct syscall_helper_data *sd, int flags)
221 {
222 	struct syscall_helper_data *sd1;
223 	int error;
224 
225 	for (sd1 = sd; sd1->syscall_no != NO_SYSCALL; sd1++) {
226 		error = kern_syscall_register(sysents, &sd1->syscall_no,
227 		    &sd1->new_sysent, &sd1->old_sysent, flags);
228 		if (error != 0) {
229 			kern_syscall_helper_unregister(sysents, sd);
230 			return (error);
231 		}
232 		sd1->registered = 1;
233 	}
234 	return (0);
235 }
236 
237 int
238 syscall_helper_unregister(struct syscall_helper_data *sd)
239 {
240 
241 	return (kern_syscall_helper_unregister(sysent, sd));
242 }
243 
244 int
245 kern_syscall_helper_unregister(struct sysent *sysents,
246     struct syscall_helper_data *sd)
247 {
248 	struct syscall_helper_data *sd1;
249 
250 	for (sd1 = sd; sd1->registered != 0; sd1++) {
251 		kern_syscall_deregister(sysents, sd1->syscall_no,
252 		    &sd1->old_sysent);
253 		sd1->registered = 0;
254 	}
255 	return (0);
256 }
257