xref: /freebsd/sys/kern/kern_syscalls.c (revision 4b9d6057)
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/param.h>
30 #include <sys/kernel.h>
31 #include <sys/lock.h>
32 #include <sys/module.h>
33 #include <sys/mutex.h>
34 #include <sys/proc.h>
35 #include <sys/resourcevar.h>
36 #include <sys/sx.h>
37 #include <sys/syscall.h>
38 #include <sys/sysent.h>
39 #include <sys/sysproto.h>
40 #include <sys/systm.h>
41 #include <machine/atomic.h>
42 
43 /*
44  * Acts like "nosys" but can be identified in sysent for dynamic call
45  * number assignment for a limited number of calls.
46  *
47  * Place holder for system call slots reserved for loadable modules.
48  */
49 int
50 lkmnosys(struct thread *td, struct nosys_args *args)
51 {
52 
53 	return (nosys(td, args));
54 }
55 
56 int
57 lkmressys(struct thread *td, struct nosys_args *args)
58 {
59 
60 	return (nosys(td, args));
61 }
62 
63 struct sysent nosys_sysent = {
64 	.sy_call =	(sy_call_t *)nosys,
65 	.sy_systrace_args_func = NULL,
66 	.sy_narg =	0,
67 	.sy_flags =	SYF_CAPENABLED,
68 	.sy_auevent =	AUE_NULL,
69 	.sy_entry =	0, /* DTRACE_IDNONE */
70 	.sy_return =	0,
71 	.sy_thrcnt =	SY_THR_STATIC,
72 };
73 
74 static void
75 syscall_thread_drain(struct sysent *se)
76 {
77 	uint32_t cnt, oldcnt;
78 
79 	do {
80 		oldcnt = se->sy_thrcnt;
81 		KASSERT((oldcnt & SY_THR_STATIC) == 0,
82 		    ("drain on static syscall"));
83 		cnt = oldcnt | SY_THR_DRAINING;
84 	} while (atomic_cmpset_acq_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
85 	while (atomic_cmpset_32(&se->sy_thrcnt, SY_THR_DRAINING,
86 	    SY_THR_ABSENT) == 0)
87 		pause("scdrn", hz/2);
88 }
89 
90 int
91 syscall_thread_enter(struct thread *td, struct sysent **se)
92 {
93 	uint32_t cnt, oldcnt;
94 
95 	KASSERT(((*se)->sy_thrcnt & SY_THR_STATIC) == 0,
96 	    ("%s: not a static syscall", __func__));
97 
98 	do {
99 		oldcnt = (*se)->sy_thrcnt;
100 		if ((oldcnt & (SY_THR_DRAINING | SY_THR_ABSENT)) != 0) {
101 			*se = &nosys_sysent;
102 			return (0);
103 		}
104 		cnt = oldcnt + SY_THR_INCR;
105 	} while (atomic_cmpset_acq_32(&(*se)->sy_thrcnt, oldcnt, cnt) == 0);
106 	return (0);
107 }
108 
109 void
110 syscall_thread_exit(struct thread *td, struct sysent *se)
111 {
112 	uint32_t cnt, oldcnt;
113 
114 	KASSERT((se->sy_thrcnt & SY_THR_STATIC) == 0,
115 	    ("%s: not a static syscall", __func__));
116 
117 	do {
118 		oldcnt = se->sy_thrcnt;
119 		cnt = oldcnt - SY_THR_INCR;
120 	} while (atomic_cmpset_rel_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
121 }
122 
123 int
124 kern_syscall_register(struct sysent *sysents, int *offset,
125     struct sysent *new_sysent, struct sysent *old_sysent, int flags)
126 {
127 	int i;
128 
129 	if ((flags & ~SY_THR_STATIC) != 0)
130 		return (EINVAL);
131 
132 	if (*offset == NO_SYSCALL) {
133 		for (i = 1; i < SYS_MAXSYSCALL; ++i)
134 			if (sysents[i].sy_call == (sy_call_t *)lkmnosys)
135 				break;
136 		if (i == SYS_MAXSYSCALL)
137 			return (ENFILE);
138 		*offset = i;
139 	} else if (*offset < 0 || *offset >= SYS_MAXSYSCALL) {
140 		return (EINVAL);
141 	} else if (sysents[*offset].sy_call != (sy_call_t *)lkmnosys &&
142 	    sysents[*offset].sy_call != (sy_call_t *)lkmressys) {
143 		KASSERT(sysents[*offset].sy_call != NULL,
144 		    ("undefined syscall %d", *offset));
145 		return (EEXIST);
146 	}
147 
148 	KASSERT(sysents[*offset].sy_thrcnt == SY_THR_ABSENT,
149 	    ("dynamic syscall is not protected"));
150 	*old_sysent = sysents[*offset];
151 	new_sysent->sy_thrcnt = SY_THR_ABSENT;
152 	sysents[*offset] = *new_sysent;
153 	atomic_store_rel_32(&sysents[*offset].sy_thrcnt, flags);
154 	return (0);
155 }
156 
157 int
158 kern_syscall_deregister(struct sysent *sysents, int offset,
159     const struct sysent *old_sysent)
160 {
161 	struct sysent *se;
162 
163 	if (offset == 0)
164 		return (0); /* XXX? */
165 
166 	se = &sysents[offset];
167 	if ((se->sy_thrcnt & SY_THR_STATIC) != 0)
168 		return (EINVAL);
169 	syscall_thread_drain(se);
170 	sysents[offset] = *old_sysent;
171 	return (0);
172 }
173 
174 int
175 syscall_module_handler(struct module *mod, int what, void *arg)
176 {
177 
178 	return (kern_syscall_module_handler(sysent, mod, what, arg));
179 }
180 
181 int
182 kern_syscall_module_handler(struct sysent *sysents, struct module *mod,
183     int what, void *arg)
184 {
185 	struct syscall_module_data *data = arg;
186 	modspecific_t ms;
187 	int error;
188 
189 	bzero(&ms, sizeof(ms));
190 	switch (what) {
191 	case MOD_LOAD:
192 		error = kern_syscall_register(sysents, data->offset,
193 		    data->new_sysent, &data->old_sysent, data->flags);
194 		if (error) {
195 			/* Leave a mark so we know to safely unload below. */
196 			data->offset = NULL;
197 			return (error);
198 		}
199 		ms.intval = *data->offset;
200 		MOD_XLOCK;
201 		module_setspecific(mod, &ms);
202 		MOD_XUNLOCK;
203 		if (data->chainevh)
204 			error = data->chainevh(mod, what, data->chainarg);
205 		return (error);
206 	case MOD_UNLOAD:
207 		/*
208 		 * MOD_LOAD failed, so just return without calling the
209 		 * chained handler since we didn't pass along the MOD_LOAD
210 		 * event.
211 		 */
212 		if (data->offset == NULL)
213 			return (0);
214 		if (data->chainevh) {
215 			error = data->chainevh(mod, what, data->chainarg);
216 			if (error)
217 				return error;
218 		}
219 		error = kern_syscall_deregister(sysents, *data->offset,
220 		    &data->old_sysent);
221 		return (error);
222 	default:
223 		if (data->chainevh)
224 			return (data->chainevh(mod, what, data->chainarg));
225 		return (EOPNOTSUPP);
226 	}
227 
228 	/* NOTREACHED */
229 }
230 
231 int
232 syscall_helper_register(struct syscall_helper_data *sd, int flags)
233 {
234 
235 	return (kern_syscall_helper_register(sysent, sd, flags));
236 }
237 
238 int
239 kern_syscall_helper_register(struct sysent *sysents,
240     struct syscall_helper_data *sd, int flags)
241 {
242 	struct syscall_helper_data *sd1;
243 	int error;
244 
245 	for (sd1 = sd; sd1->syscall_no != NO_SYSCALL; sd1++) {
246 		error = kern_syscall_register(sysents, &sd1->syscall_no,
247 		    &sd1->new_sysent, &sd1->old_sysent, flags);
248 		if (error != 0) {
249 			kern_syscall_helper_unregister(sysents, sd);
250 			return (error);
251 		}
252 		sd1->registered = 1;
253 	}
254 	return (0);
255 }
256 
257 int
258 syscall_helper_unregister(struct syscall_helper_data *sd)
259 {
260 
261 	return (kern_syscall_helper_unregister(sysent, sd));
262 }
263 
264 int
265 kern_syscall_helper_unregister(struct sysent *sysents,
266     struct syscall_helper_data *sd)
267 {
268 	struct syscall_helper_data *sd1;
269 
270 	for (sd1 = sd; sd1->registered != 0; sd1++) {
271 		kern_syscall_deregister(sysents, sd1->syscall_no,
272 		    &sd1->old_sysent);
273 		sd1->registered = 0;
274 	}
275 	return (0);
276 }
277