xref: /freebsd/sys/kern/subr_syscall.c (revision b740c88bfb6453416926271c089262e7164dace3)
1 /*-
2  * Copyright (C) 1994, David Greenman
3  * Copyright (c) 1990, 1993
4  *	The Regents of the University of California.  All rights reserved.
5  * Copyright (C) 2010 Konstantin Belousov <kib@freebsd.org>
6  *
7  * This code is derived from software contributed to Berkeley by
8  * the University of Utah, and William Jolitz.
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  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the University of
21  *	California, Berkeley and its contributors.
22  * 4. Neither the name of the University nor the names of its contributors
23  *    may be used to endorse or promote products derived from this software
24  *    without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36  * SUCH DAMAGE.
37  *
38  *	from: @(#)trap.c	7.4 (Berkeley) 5/13/91
39  */
40 
41 #include "opt_capsicum.h"
42 #include "opt_ktrace.h"
43 
44 __FBSDID("$FreeBSD$");
45 
46 #include <sys/capsicum.h>
47 #include <sys/ktr.h>
48 #ifdef KTRACE
49 #include <sys/uio.h>
50 #include <sys/ktrace.h>
51 #endif
52 #include <security/audit/audit.h>
53 
54 static inline int
55 syscallenter(struct thread *td, struct syscall_args *sa)
56 {
57 	struct proc *p;
58 	int error, traced;
59 
60 	PCPU_INC(cnt.v_syscall);
61 	p = td->td_proc;
62 
63 	td->td_pticks = 0;
64 	if (td->td_ucred != p->p_ucred)
65 		cred_update_thread(td);
66 	if (p->p_flag & P_TRACED) {
67 		traced = 1;
68 		PROC_LOCK(p);
69 		td->td_dbgflags &= ~TDB_USERWR;
70 		td->td_dbgflags |= TDB_SCE;
71 		PROC_UNLOCK(p);
72 	} else
73 		traced = 0;
74 	error = (p->p_sysent->sv_fetch_syscall_args)(td, sa);
75 #ifdef KTRACE
76 	if (KTRPOINT(td, KTR_SYSCALL))
77 		ktrsyscall(sa->code, sa->narg, sa->args);
78 #endif
79 	KTR_START4(KTR_SYSC, "syscall", syscallname(p, sa->code),
80 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "arg0:%p", sa->args[0],
81 	    "arg1:%p", sa->args[1], "arg2:%p", sa->args[2]);
82 
83 	if (error == 0) {
84 
85 		STOPEVENT(p, S_SCE, sa->narg);
86 		if (p->p_flag & P_TRACED && p->p_stops & S_PT_SCE) {
87 			PROC_LOCK(p);
88 			ptracestop((td), SIGTRAP);
89 			PROC_UNLOCK(p);
90 		}
91 		if (td->td_dbgflags & TDB_USERWR) {
92 			/*
93 			 * Reread syscall number and arguments if
94 			 * debugger modified registers or memory.
95 			 */
96 			error = (p->p_sysent->sv_fetch_syscall_args)(td, sa);
97 #ifdef KTRACE
98 			if (KTRPOINT(td, KTR_SYSCALL))
99 				ktrsyscall(sa->code, sa->narg, sa->args);
100 #endif
101 			if (error != 0)
102 				goto retval;
103 		}
104 
105 #ifdef CAPABILITY_MODE
106 		/*
107 		 * In capability mode, we only allow access to system calls
108 		 * flagged with SYF_CAPENABLED.
109 		 */
110 		if (IN_CAPABILITY_MODE(td) &&
111 		    !(sa->callp->sy_flags & SYF_CAPENABLED)) {
112 			error = ECAPMODE;
113 			goto retval;
114 		}
115 #endif
116 
117 		error = syscall_thread_enter(td, sa->callp);
118 		if (error != 0)
119 			goto retval;
120 
121 #ifdef KDTRACE_HOOKS
122 		/*
123 		 * If the systrace module has registered it's probe
124 		 * callback and if there is a probe active for the
125 		 * syscall 'entry', process the probe.
126 		 */
127 		if (systrace_probe_func != NULL && sa->callp->sy_entry != 0)
128 			(*systrace_probe_func)(sa->callp->sy_entry, sa->code,
129 			    sa->callp, sa->args, 0);
130 #endif
131 
132 		AUDIT_SYSCALL_ENTER(sa->code, td);
133 		error = (sa->callp->sy_call)(td, sa->args);
134 		AUDIT_SYSCALL_EXIT(error, td);
135 
136 		/* Save the latest error return value. */
137 		if ((td->td_pflags & TDP_NERRNO) == 0)
138 			td->td_errno = error;
139 
140 #ifdef KDTRACE_HOOKS
141 		/*
142 		 * If the systrace module has registered it's probe
143 		 * callback and if there is a probe active for the
144 		 * syscall 'return', process the probe.
145 		 */
146 		if (systrace_probe_func != NULL && sa->callp->sy_return != 0)
147 			(*systrace_probe_func)(sa->callp->sy_return, sa->code,
148 			    sa->callp, NULL, (error) ? -1 : td->td_retval[0]);
149 #endif
150 		syscall_thread_exit(td, sa->callp);
151 	}
152  retval:
153 	KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code),
154 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error,
155 	    "retval0:%#lx", td->td_retval[0], "retval1:%#lx",
156 	    td->td_retval[1]);
157 	if (traced) {
158 		PROC_LOCK(p);
159 		td->td_dbgflags &= ~TDB_SCE;
160 		PROC_UNLOCK(p);
161 	}
162 	(p->p_sysent->sv_set_syscall_retval)(td, error);
163 	return (error);
164 }
165 
166 static inline void
167 syscallret(struct thread *td, int error, struct syscall_args *sa __unused)
168 {
169 	struct proc *p, *p2;
170 	int traced;
171 
172 	p = td->td_proc;
173 
174 	/*
175 	 * Handle reschedule and other end-of-syscall issues
176 	 */
177 	userret(td, td->td_frame);
178 
179 #ifdef KTRACE
180 	if (KTRPOINT(td, KTR_SYSRET)) {
181 		ktrsysret(sa->code, (td->td_pflags & TDP_NERRNO) == 0 ?
182 		    error : td->td_errno, td->td_retval[0]);
183 	}
184 #endif
185 	td->td_pflags &= ~TDP_NERRNO;
186 
187 	if (p->p_flag & P_TRACED) {
188 		traced = 1;
189 		PROC_LOCK(p);
190 		td->td_dbgflags |= TDB_SCX;
191 		PROC_UNLOCK(p);
192 	} else
193 		traced = 0;
194 	/*
195 	 * This works because errno is findable through the
196 	 * register set.  If we ever support an emulation where this
197 	 * is not the case, this code will need to be revisited.
198 	 */
199 	STOPEVENT(p, S_SCX, sa->code);
200 	if (traced || (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0) {
201 		PROC_LOCK(p);
202 		/*
203 		 * If tracing the execed process, trap to the debugger
204 		 * so that breakpoints can be set before the program
205 		 * executes.  If debugger requested tracing of syscall
206 		 * returns, do it now too.
207 		 */
208 		if (traced &&
209 		    ((td->td_dbgflags & (TDB_FORK | TDB_EXEC)) != 0 ||
210 		    (p->p_stops & S_PT_SCX) != 0))
211 			ptracestop(td, SIGTRAP);
212 		td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK);
213 		PROC_UNLOCK(p);
214 	}
215 
216 	if (td->td_pflags & TDP_RFPPWAIT) {
217 		/*
218 		 * Preserve synchronization semantics of vfork.  If
219 		 * waiting for child to exec or exit, fork set
220 		 * P_PPWAIT on child, and there we sleep on our proc
221 		 * (in case of exit).
222 		 *
223 		 * Do it after the ptracestop() above is finished, to
224 		 * not block our debugger until child execs or exits
225 		 * to finish vfork wait.
226 		 */
227 		td->td_pflags &= ~TDP_RFPPWAIT;
228 		p2 = td->td_rfppwait_p;
229 again:
230 		PROC_LOCK(p2);
231 		while (p2->p_flag & P_PPWAIT) {
232 			PROC_LOCK(p);
233 			if (thread_suspend_check_needed()) {
234 				PROC_UNLOCK(p2);
235 				thread_suspend_check(0);
236 				PROC_UNLOCK(p);
237 				goto again;
238 			} else {
239 				PROC_UNLOCK(p);
240 			}
241 			cv_timedwait(&p2->p_pwait, &p2->p_mtx, hz);
242 		}
243 		PROC_UNLOCK(p2);
244 	}
245 }
246