xref: /freebsd/sys/kern/kern_syscalls.c (revision 83823d063ab57db8d3954c1530d036f1ccdceb41)
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 struct sysent nosys_sysent = {
65 	.sy_call =	(sy_call_t *)nosys,
66 	.sy_systrace_args_func = NULL,
67 	.sy_narg =	0,
68 	.sy_flags =	SYF_CAPENABLED,
69 	.sy_auevent =	AUE_NULL,
70 	.sy_entry =	0, /* DTRACE_IDNONE */
71 	.sy_return =	0,
72 	.sy_thrcnt =	SY_THR_STATIC,
73 };
74 
75 static void
76 syscall_thread_drain(struct sysent *se)
77 {
78 	uint32_t cnt, oldcnt;
79 
80 	do {
81 		oldcnt = se->sy_thrcnt;
82 		KASSERT((oldcnt & SY_THR_STATIC) == 0,
83 		    ("drain on static syscall"));
84 		cnt = oldcnt | SY_THR_DRAINING;
85 	} while (atomic_cmpset_acq_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
86 	while (atomic_cmpset_32(&se->sy_thrcnt, SY_THR_DRAINING,
87 	    SY_THR_ABSENT) == 0)
88 		pause("scdrn", hz/2);
89 }
90 
91 int
92 syscall_thread_enter(struct thread *td, struct sysent **se)
93 {
94 	uint32_t cnt, oldcnt;
95 
96 	KASSERT(((*se)->sy_thrcnt & SY_THR_STATIC) == 0,
97 	    ("%s: not a static syscall", __func__));
98 
99 	do {
100 		oldcnt = (*se)->sy_thrcnt;
101 		if ((oldcnt & (SY_THR_DRAINING | SY_THR_ABSENT)) != 0) {
102 			*se = &nosys_sysent;
103 			return (0);
104 		}
105 		cnt = oldcnt + SY_THR_INCR;
106 	} while (atomic_cmpset_acq_32(&(*se)->sy_thrcnt, oldcnt, cnt) == 0);
107 	return (0);
108 }
109 
110 void
111 syscall_thread_exit(struct thread *td, struct sysent *se)
112 {
113 	uint32_t cnt, oldcnt;
114 
115 	KASSERT((se->sy_thrcnt & SY_THR_STATIC) == 0,
116 	    ("%s: not a static syscall", __func__));
117 
118 	do {
119 		oldcnt = se->sy_thrcnt;
120 		cnt = oldcnt - SY_THR_INCR;
121 	} while (atomic_cmpset_rel_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
122 }
123 
124 int
125 kern_syscall_register(struct sysent *sysents, int *offset,
126     struct sysent *new_sysent, struct sysent *old_sysent, int flags)
127 {
128 	int i;
129 
130 	if ((flags & ~SY_THR_STATIC) != 0)
131 		return (EINVAL);
132 
133 	if (*offset == NO_SYSCALL) {
134 		for (i = 1; i < SYS_MAXSYSCALL; ++i)
135 			if (sysents[i].sy_call == (sy_call_t *)lkmnosys)
136 				break;
137 		if (i == SYS_MAXSYSCALL)
138 			return (ENFILE);
139 		*offset = i;
140 	} else if (*offset < 0 || *offset >= SYS_MAXSYSCALL) {
141 		return (EINVAL);
142 	} else if (sysents[*offset].sy_call != (sy_call_t *)lkmnosys &&
143 	    sysents[*offset].sy_call != (sy_call_t *)lkmressys) {
144 		KASSERT(sysents[*offset].sy_call != NULL,
145 		    ("undefined syscall %d", *offset));
146 		return (EEXIST);
147 	}
148 
149 	KASSERT(sysents[*offset].sy_thrcnt == SY_THR_ABSENT,
150 	    ("dynamic syscall is not protected"));
151 	*old_sysent = sysents[*offset];
152 	new_sysent->sy_thrcnt = SY_THR_ABSENT;
153 	sysents[*offset] = *new_sysent;
154 	atomic_store_rel_32(&sysents[*offset].sy_thrcnt, flags);
155 	return (0);
156 }
157 
158 int
159 kern_syscall_deregister(struct sysent *sysents, int offset,
160     const struct sysent *old_sysent)
161 {
162 	struct sysent *se;
163 
164 	if (offset == 0)
165 		return (0); /* XXX? */
166 
167 	se = &sysents[offset];
168 	if ((se->sy_thrcnt & SY_THR_STATIC) != 0)
169 		return (EINVAL);
170 	syscall_thread_drain(se);
171 	sysents[offset] = *old_sysent;
172 	return (0);
173 }
174 
175 int
176 syscall_module_handler(struct module *mod, int what, void *arg)
177 {
178 
179 	return (kern_syscall_module_handler(sysent, mod, what, arg));
180 }
181 
182 int
183 kern_syscall_module_handler(struct sysent *sysents, struct module *mod,
184     int what, void *arg)
185 {
186 	struct syscall_module_data *data = arg;
187 	modspecific_t ms;
188 	int error;
189 
190 	bzero(&ms, sizeof(ms));
191 	switch (what) {
192 	case MOD_LOAD:
193 		error = kern_syscall_register(sysents, data->offset,
194 		    data->new_sysent, &data->old_sysent, data->flags);
195 		if (error) {
196 			/* Leave a mark so we know to safely unload below. */
197 			data->offset = NULL;
198 			return (error);
199 		}
200 		ms.intval = *data->offset;
201 		MOD_XLOCK;
202 		module_setspecific(mod, &ms);
203 		MOD_XUNLOCK;
204 		if (data->chainevh)
205 			error = data->chainevh(mod, what, data->chainarg);
206 		return (error);
207 	case MOD_UNLOAD:
208 		/*
209 		 * MOD_LOAD failed, so just return without calling the
210 		 * chained handler since we didn't pass along the MOD_LOAD
211 		 * event.
212 		 */
213 		if (data->offset == NULL)
214 			return (0);
215 		if (data->chainevh) {
216 			error = data->chainevh(mod, what, data->chainarg);
217 			if (error)
218 				return error;
219 		}
220 		error = kern_syscall_deregister(sysents, *data->offset,
221 		    &data->old_sysent);
222 		return (error);
223 	default:
224 		if (data->chainevh)
225 			return (data->chainevh(mod, what, data->chainarg));
226 		return (EOPNOTSUPP);
227 	}
228 
229 	/* NOTREACHED */
230 }
231 
232 int
233 syscall_helper_register(struct syscall_helper_data *sd, int flags)
234 {
235 
236 	return (kern_syscall_helper_register(sysent, sd, flags));
237 }
238 
239 int
240 kern_syscall_helper_register(struct sysent *sysents,
241     struct syscall_helper_data *sd, int flags)
242 {
243 	struct syscall_helper_data *sd1;
244 	int error;
245 
246 	for (sd1 = sd; sd1->syscall_no != NO_SYSCALL; sd1++) {
247 		error = kern_syscall_register(sysents, &sd1->syscall_no,
248 		    &sd1->new_sysent, &sd1->old_sysent, flags);
249 		if (error != 0) {
250 			kern_syscall_helper_unregister(sysents, sd);
251 			return (error);
252 		}
253 		sd1->registered = 1;
254 	}
255 	return (0);
256 }
257 
258 int
259 syscall_helper_unregister(struct syscall_helper_data *sd)
260 {
261 
262 	return (kern_syscall_helper_unregister(sysent, sd));
263 }
264 
265 int
266 kern_syscall_helper_unregister(struct sysent *sysents,
267     struct syscall_helper_data *sd)
268 {
269 	struct syscall_helper_data *sd1;
270 
271 	for (sd1 = sd; sd1->registered != 0; sd1++) {
272 		kern_syscall_deregister(sysents, sd1->syscall_no,
273 		    &sd1->old_sysent);
274 		sd1->registered = 0;
275 	}
276 	return (0);
277 }
278