1/* $NetBSD: kern_syscall.c,v 1.18 2019/05/06 08:05:03 kamil Exp $ */
2
3/*-
4 * Copyright (c) 2008 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software developed for The NetBSD Foundation
8 * by Andrew Doran.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29 * POSSIBILITY OF SUCH DAMAGE.
30 */
31
32#include <sys/cdefs.h>
33__KERNEL_RCSID(0, "$NetBSD: kern_syscall.c,v 1.18 2019/05/06 08:05:03 kamil Exp $");
34
35#ifdef _KERNEL_OPT
36#include "opt_modular.h"
37#include "opt_syscall_debug.h"
38#include "opt_ktrace.h"
39#include "opt_ptrace.h"
40#include "opt_dtrace.h"
41#endif
42
43/* XXX To get syscall prototypes. */
44#define SYSVSHM
45#define SYSVSEM
46#define SYSVMSG
47
48#include <sys/param.h>
49#include <sys/module.h>
50#include <sys/sched.h>
51#include <sys/syscall.h>
52#include <sys/syscallargs.h>
53#include <sys/syscallvar.h>
54#include <sys/systm.h>
55#include <sys/xcall.h>
56#include <sys/ktrace.h>
57#include <sys/ptrace.h>
58
59int
60sys_nomodule(struct lwp *l, const void *v, register_t *retval)
61{
62#ifdef MODULAR
63
64 const struct sysent *sy;
65 const struct emul *em;
66 const struct sc_autoload *auto_list;
67 u_int code;
68
69 /*
70 * Restart the syscall if we interrupted a module unload that
71 * failed. Acquiring kernconfig_lock delays us until any unload
72 * has been completed or rolled back.
73 */
74 kernconfig_lock();
75 sy = l->l_sysent;
76 if (sy->sy_call != sys_nomodule) {
77 kernconfig_unlock();
78 return ERESTART;
79 }
80 /*
81 * Try to autoload a module to satisfy the request. If it
82 * works, retry the request.
83 */
84 em = l->l_proc->p_emul;
85 code = sy - em->e_sysent;
86
87 if ((auto_list = em->e_sc_autoload) != NULL)
88 for (; auto_list->al_code > 0; auto_list++) {
89 if (auto_list->al_code != code) {
90 continue;
91 }
92 if (module_autoload(auto_list->al_module,
93 MODULE_CLASS_ANY) != 0 ||
94 sy->sy_call == sys_nomodule) {
95 break;
96 }
97 kernconfig_unlock();
98 return ERESTART;
99 }
100 kernconfig_unlock();
101#endif /* MODULAR */
102
103 return sys_nosys(l, v, retval);
104}
105
106int
107syscall_establish(const struct emul *em, const struct syscall_package *sp)
108{
109 struct sysent *sy;
110 int i;
111
112 KASSERT(kernconfig_is_held());
113
114 if (em == NULL) {
115 em = &emul_netbsd;
116 }
117 sy = em->e_sysent;
118
119 /*
120 * Ensure that all preconditions are valid, since this is
121 * an all or nothing deal. Once a system call is entered,
122 * it can become busy and we could be unable to remove it
123 * on error.
124 */
125 for (i = 0; sp[i].sp_call != NULL; i++) {
126 if (sp[i].sp_code >= SYS_NSYSENT)
127 return EINVAL;
128 if (sy[sp[i].sp_code].sy_call != sys_nomodule &&
129 sy[sp[i].sp_code].sy_call != sys_nosys) {
130#ifdef DIAGNOSTIC
131 printf("syscall %d is busy\n", sp[i].sp_code);
132#endif
133 return EBUSY;
134 }
135 }
136 /* Everything looks good, patch them in. */
137 for (i = 0; sp[i].sp_call != NULL; i++) {
138 sy[sp[i].sp_code].sy_call = sp[i].sp_call;
139 }
140
141 return 0;
142}
143
144int
145syscall_disestablish(const struct emul *em, const struct syscall_package *sp)
146{
147 struct sysent *sy;
148 const uint32_t *sb;
149 uint64_t where;
150 lwp_t *l;
151 int i;
152
153 KASSERT(kernconfig_is_held());
154
155 if (em == NULL) {
156 em = &emul_netbsd;
157 }
158 sy = em->e_sysent;
159 sb = em->e_nomodbits;
160
161 /*
162 * First, patch the system calls to sys_nomodule or sys_nosys
163 * to gate further activity.
164 */
165 for (i = 0; sp[i].sp_call != NULL; i++) {
166 KASSERT(sy[sp[i].sp_code].sy_call == sp[i].sp_call);
167 sy[sp[i].sp_code].sy_call =
168 sb[sp[i].sp_code / 32] & (1 << (sp[i].sp_code % 32)) ?
169 sys_nomodule : sys_nosys;
170 }
171
172 /*
173 * Run a cross call to cycle through all CPUs. This does two
174 * things: lock activity provides a barrier and makes our update
175 * of sy_call visible to all CPUs, and upon return we can be sure
176 * that we see pertinent values of l_sysent posted by remote CPUs.
177 */
178 where = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL);
179 xc_wait(where);
180
181 /*
182 * Now it's safe to check l_sysent. Run through all LWPs and see
183 * if anyone is still using the system call.
184 */
185 for (i = 0; sp[i].sp_call != NULL; i++) {
186 mutex_enter(proc_lock);
187 LIST_FOREACH(l, &alllwp, l_list) {
188 if (l->l_sysent == &sy[sp[i].sp_code]) {
189 break;
190 }
191 }
192 mutex_exit(proc_lock);
193 if (l == NULL) {
194 continue;
195 }
196 /*
197 * We lose: one or more calls are still in use. Put back
198 * the old entrypoints and act like nothing happened.
199 * When we drop kernconfig_lock, any system calls held in
200 * sys_nomodule() will be restarted.
201 */
202 for (i = 0; sp[i].sp_call != NULL; i++) {
203 sy[sp[i].sp_code].sy_call = sp[i].sp_call;
204 }
205 return EBUSY;
206 }
207
208 return 0;
209}
210
211/*
212 * Return true if system call tracing is enabled for the specified process.
213 */
214bool
215trace_is_enabled(struct proc *p)
216{
217#ifdef SYSCALL_DEBUG
218 return (true);
219#endif
220#ifdef KTRACE
221 if (ISSET(p->p_traceflag, (KTRFAC_SYSCALL | KTRFAC_SYSRET)))
222 return (true);
223#endif
224#ifdef PTRACE
225 if (ISSET(p->p_slflag, PSL_SYSCALL))
226 return (true);
227#endif
228
229 return (false);
230}
231
232/*
233 * Start trace of particular system call. If process is being traced,
234 * this routine is called by MD syscall dispatch code just before
235 * a system call is actually executed.
236 */
237int
238trace_enter(register_t code, const struct sysent *sy, const void *args)
239{
240 int error = 0;
241
242#ifdef KDTRACE_HOOKS
243 if (sy->sy_entry) {
244 struct emul *e = curlwp->l_proc->p_emul;
245 (*e->e_dtrace_syscall)(sy->sy_entry, code, sy, args, NULL, 0);
246 }
247#endif
248
249#ifdef SYSCALL_DEBUG
250 scdebug_call(code, args);
251#endif /* SYSCALL_DEBUG */
252
253 ktrsyscall(code, args, sy->sy_narg);
254
255#ifdef PTRACE
256 if ((curlwp->l_proc->p_slflag & (PSL_SYSCALL|PSL_TRACED)) ==
257 (PSL_SYSCALL|PSL_TRACED)) {
258 proc_stoptrace(TRAP_SCE, code, args, NULL, 0);
259 if (curlwp->l_proc->p_slflag & PSL_SYSCALLEMU) {
260 /* tracer will emulate syscall for us */
261 error = EJUSTRETURN;
262 }
263 }
264#endif
265 return error;
266}
267
268/*
269 * End trace of particular system call. If process is being traced,
270 * this routine is called by MD syscall dispatch code just after
271 * a system call finishes.
272 * MD caller guarantees the passed 'code' is within the supported
273 * system call number range for emulation the process runs under.
274 */
275void
276trace_exit(register_t code, const struct sysent *sy, const void *args,
277 register_t rval[], int error)
278{
279#if defined(PTRACE) || defined(KDTRACE_HOOKS)
280 struct proc *p = curlwp->l_proc;
281#endif
282
283#ifdef KDTRACE_HOOKS
284 if (sy->sy_return) {
285 (*p->p_emul->e_dtrace_syscall)(sy->sy_return, code, sy, args,
286 rval, error);
287 }
288#endif
289
290#ifdef SYSCALL_DEBUG
291 scdebug_ret(code, error, rval);
292#endif /* SYSCALL_DEBUG */
293
294 ktrsysret(code, error, rval);
295
296#ifdef PTRACE
297 if ((p->p_slflag & (PSL_SYSCALL|PSL_TRACED|PSL_SYSCALLEMU)) ==
298 (PSL_SYSCALL|PSL_TRACED)) {
299 proc_stoptrace(TRAP_SCX, code, args, rval, error);
300 }
301 CLR(p->p_slflag, PSL_SYSCALLEMU);
302#endif
303}
304