xref: /freebsd/sys/kern/kern_syscalls.c (revision d0b2dbfa)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
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 #include <sys/param.h>
31 #include <sys/kernel.h>
32 #include <sys/lock.h>
33 #include <sys/module.h>
34 #include <sys/mutex.h>
35 #include <sys/proc.h>
36 #include <sys/resourcevar.h>
37 #include <sys/sx.h>
38 #include <sys/syscall.h>
39 #include <sys/sysent.h>
40 #include <sys/sysproto.h>
41 #include <sys/systm.h>
42 #include <machine/atomic.h>
43 
44 /*
45  * Acts like "nosys" but can be identified in sysent for dynamic call
46  * number assignment for a limited number of calls.
47  *
48  * Place holder for system call slots reserved for loadable modules.
49  */
50 int
51 lkmnosys(struct thread *td, struct nosys_args *args)
52 {
53 
54 	return (nosys(td, args));
55 }
56 
57 int
58 lkmressys(struct thread *td, struct nosys_args *args)
59 {
60 
61 	return (nosys(td, args));
62 }
63 
64 static void
65 syscall_thread_drain(struct sysent *se)
66 {
67 	uint32_t cnt, oldcnt;
68 
69 	do {
70 		oldcnt = se->sy_thrcnt;
71 		KASSERT((oldcnt & SY_THR_STATIC) == 0,
72 		    ("drain on static syscall"));
73 		cnt = oldcnt | SY_THR_DRAINING;
74 	} while (atomic_cmpset_acq_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
75 	while (atomic_cmpset_32(&se->sy_thrcnt, SY_THR_DRAINING,
76 	    SY_THR_ABSENT) == 0)
77 		pause("scdrn", hz/2);
78 }
79 
80 int
81 syscall_thread_enter(struct thread *td, struct sysent *se)
82 {
83 	uint32_t cnt, oldcnt;
84 
85 	KASSERT((se->sy_thrcnt & SY_THR_STATIC) == 0,
86 	    ("%s: not a static syscall", __func__));
87 
88 	do {
89 		oldcnt = se->sy_thrcnt;
90 		if ((oldcnt & (SY_THR_DRAINING | SY_THR_ABSENT)) != 0)
91 			return (ENOSYS);
92 		cnt = oldcnt + SY_THR_INCR;
93 	} while (atomic_cmpset_acq_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
94 	return (0);
95 }
96 
97 void
98 syscall_thread_exit(struct thread *td, struct sysent *se)
99 {
100 	uint32_t cnt, oldcnt;
101 
102 	KASSERT((se->sy_thrcnt & SY_THR_STATIC) == 0,
103 	    ("%s: not a static syscall", __func__));
104 
105 	do {
106 		oldcnt = se->sy_thrcnt;
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 		KASSERT(sysents[*offset].sy_call != NULL,
132 		    ("undefined syscall %d", *offset));
133 		return (EEXIST);
134 	}
135 
136 	KASSERT(sysents[*offset].sy_thrcnt == SY_THR_ABSENT,
137 	    ("dynamic syscall is not protected"));
138 	*old_sysent = sysents[*offset];
139 	new_sysent->sy_thrcnt = SY_THR_ABSENT;
140 	sysents[*offset] = *new_sysent;
141 	atomic_store_rel_32(&sysents[*offset].sy_thrcnt, flags);
142 	return (0);
143 }
144 
145 int
146 kern_syscall_deregister(struct sysent *sysents, int offset,
147     const struct sysent *old_sysent)
148 {
149 	struct sysent *se;
150 
151 	if (offset == 0)
152 		return (0); /* XXX? */
153 
154 	se = &sysents[offset];
155 	if ((se->sy_thrcnt & SY_THR_STATIC) != 0)
156 		return (EINVAL);
157 	syscall_thread_drain(se);
158 	sysents[offset] = *old_sysent;
159 	return (0);
160 }
161 
162 int
163 syscall_module_handler(struct module *mod, int what, void *arg)
164 {
165 
166 	return (kern_syscall_module_handler(sysent, mod, what, arg));
167 }
168 
169 int
170 kern_syscall_module_handler(struct sysent *sysents, struct module *mod,
171     int what, void *arg)
172 {
173 	struct syscall_module_data *data = arg;
174 	modspecific_t ms;
175 	int error;
176 
177 	bzero(&ms, sizeof(ms));
178 	switch (what) {
179 	case MOD_LOAD:
180 		error = kern_syscall_register(sysents, data->offset,
181 		    data->new_sysent, &data->old_sysent, data->flags);
182 		if (error) {
183 			/* Leave a mark so we know to safely unload below. */
184 			data->offset = NULL;
185 			return (error);
186 		}
187 		ms.intval = *data->offset;
188 		MOD_XLOCK;
189 		module_setspecific(mod, &ms);
190 		MOD_XUNLOCK;
191 		if (data->chainevh)
192 			error = data->chainevh(mod, what, data->chainarg);
193 		return (error);
194 	case MOD_UNLOAD:
195 		/*
196 		 * MOD_LOAD failed, so just return without calling the
197 		 * chained handler since we didn't pass along the MOD_LOAD
198 		 * event.
199 		 */
200 		if (data->offset == NULL)
201 			return (0);
202 		if (data->chainevh) {
203 			error = data->chainevh(mod, what, data->chainarg);
204 			if (error)
205 				return error;
206 		}
207 		error = kern_syscall_deregister(sysents, *data->offset,
208 		    &data->old_sysent);
209 		return (error);
210 	default:
211 		if (data->chainevh)
212 			return (data->chainevh(mod, what, data->chainarg));
213 		return (EOPNOTSUPP);
214 	}
215 
216 	/* NOTREACHED */
217 }
218 
219 int
220 syscall_helper_register(struct syscall_helper_data *sd, int flags)
221 {
222 
223 	return (kern_syscall_helper_register(sysent, sd, flags));
224 }
225 
226 int
227 kern_syscall_helper_register(struct sysent *sysents,
228     struct syscall_helper_data *sd, int flags)
229 {
230 	struct syscall_helper_data *sd1;
231 	int error;
232 
233 	for (sd1 = sd; sd1->syscall_no != NO_SYSCALL; sd1++) {
234 		error = kern_syscall_register(sysents, &sd1->syscall_no,
235 		    &sd1->new_sysent, &sd1->old_sysent, flags);
236 		if (error != 0) {
237 			kern_syscall_helper_unregister(sysents, sd);
238 			return (error);
239 		}
240 		sd1->registered = 1;
241 	}
242 	return (0);
243 }
244 
245 int
246 syscall_helper_unregister(struct syscall_helper_data *sd)
247 {
248 
249 	return (kern_syscall_helper_unregister(sysent, sd));
250 }
251 
252 int
253 kern_syscall_helper_unregister(struct sysent *sysents,
254     struct syscall_helper_data *sd)
255 {
256 	struct syscall_helper_data *sd1;
257 
258 	for (sd1 = sd; sd1->registered != 0; sd1++) {
259 		kern_syscall_deregister(sysents, sd1->syscall_no,
260 		    &sd1->old_sysent);
261 		sd1->registered = 0;
262 	}
263 	return (0);
264 }
265