xref: /freebsd/sys/kern/kern_syscalls.c (revision 7029da5c36f2d3cf6bb6c81bf551229f416399e8)
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 		KASSERT(sysents[*offset].sy_call != NULL,
128 		    ("undefined syscall %d", *offset));
129 		return (EEXIST);
130 	}
131 
132 	KASSERT(sysents[*offset].sy_thrcnt == SY_THR_ABSENT,
133 	    ("dynamic syscall is not protected"));
134 	*old_sysent = sysents[*offset];
135 	new_sysent->sy_thrcnt = SY_THR_ABSENT;
136 	sysents[*offset] = *new_sysent;
137 	atomic_store_rel_32(&sysents[*offset].sy_thrcnt, flags);
138 	return (0);
139 }
140 
141 int
142 kern_syscall_deregister(struct sysent *sysents, int offset,
143     const struct sysent *old_sysent)
144 {
145 	struct sysent *se;
146 
147 	if (offset == 0)
148 		return (0); /* XXX? */
149 
150 	se = &sysents[offset];
151 	if ((se->sy_thrcnt & SY_THR_STATIC) != 0)
152 		return (EINVAL);
153 	syscall_thread_drain(se);
154 	sysents[offset] = *old_sysent;
155 	return (0);
156 }
157 
158 int
159 syscall_module_handler(struct module *mod, int what, void *arg)
160 {
161 
162 	return (kern_syscall_module_handler(sysent, mod, what, arg));
163 }
164 
165 int
166 kern_syscall_module_handler(struct sysent *sysents, struct module *mod,
167     int what, void *arg)
168 {
169 	struct syscall_module_data *data = arg;
170 	modspecific_t ms;
171 	int error;
172 
173 	switch (what) {
174 	case MOD_LOAD:
175 		error = kern_syscall_register(sysents, data->offset,
176 		    data->new_sysent, &data->old_sysent, data->flags);
177 		if (error) {
178 			/* Leave a mark so we know to safely unload below. */
179 			data->offset = NULL;
180 			return (error);
181 		}
182 		ms.intval = *data->offset;
183 		MOD_XLOCK;
184 		module_setspecific(mod, &ms);
185 		MOD_XUNLOCK;
186 		if (data->chainevh)
187 			error = data->chainevh(mod, what, data->chainarg);
188 		return (error);
189 	case MOD_UNLOAD:
190 		/*
191 		 * MOD_LOAD failed, so just return without calling the
192 		 * chained handler since we didn't pass along the MOD_LOAD
193 		 * event.
194 		 */
195 		if (data->offset == NULL)
196 			return (0);
197 		if (data->chainevh) {
198 			error = data->chainevh(mod, what, data->chainarg);
199 			if (error)
200 				return error;
201 		}
202 		error = kern_syscall_deregister(sysents, *data->offset,
203 		    &data->old_sysent);
204 		return (error);
205 	default:
206 		if (data->chainevh)
207 			return (data->chainevh(mod, what, data->chainarg));
208 		return (EOPNOTSUPP);
209 	}
210 
211 	/* NOTREACHED */
212 }
213 
214 int
215 syscall_helper_register(struct syscall_helper_data *sd, int flags)
216 {
217 
218 	return (kern_syscall_helper_register(sysent, sd, flags));
219 }
220 
221 int
222 kern_syscall_helper_register(struct sysent *sysents,
223     struct syscall_helper_data *sd, int flags)
224 {
225 	struct syscall_helper_data *sd1;
226 	int error;
227 
228 	for (sd1 = sd; sd1->syscall_no != NO_SYSCALL; sd1++) {
229 		error = kern_syscall_register(sysents, &sd1->syscall_no,
230 		    &sd1->new_sysent, &sd1->old_sysent, flags);
231 		if (error != 0) {
232 			kern_syscall_helper_unregister(sysents, sd);
233 			return (error);
234 		}
235 		sd1->registered = 1;
236 	}
237 	return (0);
238 }
239 
240 int
241 syscall_helper_unregister(struct syscall_helper_data *sd)
242 {
243 
244 	return (kern_syscall_helper_unregister(sysent, sd));
245 }
246 
247 int
248 kern_syscall_helper_unregister(struct sysent *sysents,
249     struct syscall_helper_data *sd)
250 {
251 	struct syscall_helper_data *sd1;
252 
253 	for (sd1 = sd; sd1->registered != 0; sd1++) {
254 		kern_syscall_deregister(sysents, sd1->syscall_no,
255 		    &sd1->old_sysent);
256 		sd1->registered = 0;
257 	}
258 	return (0);
259 }
260