xref: /freebsd/sys/kern/subr_syscall.c (revision 792bbaba989533a1fc93823df1720c8c4aaf0442)
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_cowgen != p->p_cowgen)
65 		thread_cow_update(td);
66 	traced = (p->p_flag & P_TRACED) != 0;
67 	if (traced || td->td_dbgflags & TDB_USERWR) {
68 		PROC_LOCK(p);
69 		td->td_dbgflags &= ~TDB_USERWR;
70 		if (traced)
71 			td->td_dbgflags |= TDB_SCE;
72 		PROC_UNLOCK(p);
73 	}
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) {
87 			PROC_LOCK(p);
88 			td->td_dbg_sc_code = sa->code;
89 			td->td_dbg_sc_narg = sa->narg;
90 			if (p->p_ptevents & PTRACE_SCE)
91 				ptracestop((td), SIGTRAP);
92 			PROC_UNLOCK(p);
93 		}
94 		if (td->td_dbgflags & TDB_USERWR) {
95 			/*
96 			 * Reread syscall number and arguments if
97 			 * debugger modified registers or memory.
98 			 */
99 			error = (p->p_sysent->sv_fetch_syscall_args)(td, sa);
100 			PROC_LOCK(p);
101 			td->td_dbg_sc_code = sa->code;
102 			td->td_dbg_sc_narg = sa->narg;
103 			PROC_UNLOCK(p);
104 #ifdef KTRACE
105 			if (KTRPOINT(td, KTR_SYSCALL))
106 				ktrsyscall(sa->code, sa->narg, sa->args);
107 #endif
108 			if (error != 0)
109 				goto retval;
110 		}
111 
112 #ifdef CAPABILITY_MODE
113 		/*
114 		 * In capability mode, we only allow access to system calls
115 		 * flagged with SYF_CAPENABLED.
116 		 */
117 		if (IN_CAPABILITY_MODE(td) &&
118 		    !(sa->callp->sy_flags & SYF_CAPENABLED)) {
119 			error = ECAPMODE;
120 			goto retval;
121 		}
122 #endif
123 
124 		error = syscall_thread_enter(td, sa->callp);
125 		if (error != 0)
126 			goto retval;
127 
128 #ifdef KDTRACE_HOOKS
129 		/* Give the syscall:::entry DTrace probe a chance to fire. */
130 		if (systrace_probe_func != NULL && sa->callp->sy_entry != 0)
131 			(*systrace_probe_func)(sa, SYSTRACE_ENTRY, 0);
132 #endif
133 
134 		AUDIT_SYSCALL_ENTER(sa->code, td);
135 		error = (sa->callp->sy_call)(td, sa->args);
136 		AUDIT_SYSCALL_EXIT(error, td);
137 
138 		/* Save the latest error return value. */
139 		if ((td->td_pflags & TDP_NERRNO) == 0)
140 			td->td_errno = error;
141 
142 #ifdef KDTRACE_HOOKS
143 		/* Give the syscall:::return DTrace probe a chance to fire. */
144 		if (systrace_probe_func != NULL && sa->callp->sy_return != 0)
145 			(*systrace_probe_func)(sa, SYSTRACE_RETURN,
146 			    error ? -1 : td->td_retval[0]);
147 #endif
148 		syscall_thread_exit(td, sa->callp);
149 	}
150  retval:
151 	KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code),
152 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error,
153 	    "retval0:%#lx", td->td_retval[0], "retval1:%#lx",
154 	    td->td_retval[1]);
155 	if (traced) {
156 		PROC_LOCK(p);
157 		td->td_dbgflags &= ~TDB_SCE;
158 		PROC_UNLOCK(p);
159 	}
160 	(p->p_sysent->sv_set_syscall_retval)(td, error);
161 	return (error);
162 }
163 
164 static inline void
165 syscallret(struct thread *td, int error, struct syscall_args *sa)
166 {
167 	struct proc *p, *p2;
168 	ksiginfo_t ksi;
169 	int traced, error1;
170 
171 	KASSERT((td->td_pflags & TDP_FORKING) == 0,
172 	    ("fork() did not clear TDP_FORKING upon completion"));
173 
174 	p = td->td_proc;
175 	if ((trap_enotcap || (p->p_flag2 & P2_TRAPCAP) != 0) &&
176 	    IN_CAPABILITY_MODE(td)) {
177 		error1 = (td->td_pflags & TDP_NERRNO) == 0 ? error :
178 		    td->td_errno;
179 		if (error1 == ENOTCAPABLE || error1 == ECAPMODE) {
180 			ksiginfo_init_trap(&ksi);
181 			ksi.ksi_signo = SIGTRAP;
182 			ksi.ksi_errno = error1;
183 			ksi.ksi_code = TRAP_CAP;
184 			trapsignal(td, &ksi);
185 		}
186 	}
187 
188 	/*
189 	 * Handle reschedule and other end-of-syscall issues
190 	 */
191 	userret(td, td->td_frame);
192 
193 #ifdef KTRACE
194 	if (KTRPOINT(td, KTR_SYSRET)) {
195 		ktrsysret(sa->code, (td->td_pflags & TDP_NERRNO) == 0 ?
196 		    error : td->td_errno, td->td_retval[0]);
197 	}
198 #endif
199 	td->td_pflags &= ~TDP_NERRNO;
200 
201 	if (p->p_flag & P_TRACED) {
202 		traced = 1;
203 		PROC_LOCK(p);
204 		td->td_dbgflags |= TDB_SCX;
205 		PROC_UNLOCK(p);
206 	} else
207 		traced = 0;
208 	/*
209 	 * This works because errno is findable through the
210 	 * register set.  If we ever support an emulation where this
211 	 * is not the case, this code will need to be revisited.
212 	 */
213 	STOPEVENT(p, S_SCX, sa->code);
214 	if (traced || (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0) {
215 		PROC_LOCK(p);
216 		/*
217 		 * If tracing the execed process, trap to the debugger
218 		 * so that breakpoints can be set before the program
219 		 * executes.  If debugger requested tracing of syscall
220 		 * returns, do it now too.
221 		 */
222 		if (traced &&
223 		    ((td->td_dbgflags & (TDB_FORK | TDB_EXEC)) != 0 ||
224 		    (p->p_ptevents & PTRACE_SCX) != 0))
225 			ptracestop(td, SIGTRAP);
226 		td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK);
227 		PROC_UNLOCK(p);
228 	}
229 
230 	if (td->td_pflags & TDP_RFPPWAIT) {
231 		/*
232 		 * Preserve synchronization semantics of vfork.  If
233 		 * waiting for child to exec or exit, fork set
234 		 * P_PPWAIT on child, and there we sleep on our proc
235 		 * (in case of exit).
236 		 *
237 		 * Do it after the ptracestop() above is finished, to
238 		 * not block our debugger until child execs or exits
239 		 * to finish vfork wait.
240 		 */
241 		td->td_pflags &= ~TDP_RFPPWAIT;
242 		p2 = td->td_rfppwait_p;
243 again:
244 		PROC_LOCK(p2);
245 		while (p2->p_flag & P_PPWAIT) {
246 			PROC_LOCK(p);
247 			if (thread_suspend_check_needed()) {
248 				PROC_UNLOCK(p2);
249 				thread_suspend_check(0);
250 				PROC_UNLOCK(p);
251 				goto again;
252 			} else {
253 				PROC_UNLOCK(p);
254 			}
255 			cv_timedwait(&p2->p_pwait, &p2->p_mtx, hz);
256 		}
257 		PROC_UNLOCK(p2);
258 
259 		if (td->td_dbgflags & TDB_VFORK) {
260 			PROC_LOCK(p);
261 			if (p->p_ptevents & PTRACE_VFORK)
262 				ptracestop(td, SIGTRAP);
263 			td->td_dbgflags &= ~TDB_VFORK;
264 			PROC_UNLOCK(p);
265 		}
266 	}
267 }
268