xref: /freebsd/sys/kern/subr_syscall.c (revision 7c1b51d6dc2e165ae7333373513b080f17cf79bd)
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 #include <sys/vmmeter.h>
49 #ifdef KTRACE
50 #include <sys/uio.h>
51 #include <sys/ktrace.h>
52 #endif
53 #include <security/audit/audit.h>
54 
55 static inline int
56 syscallenter(struct thread *td)
57 {
58 	struct proc *p;
59 	struct syscall_args *sa;
60 	int error, traced;
61 
62 	VM_CNT_INC(v_syscall);
63 	p = td->td_proc;
64 	sa = &td->td_sa;
65 
66 	td->td_pticks = 0;
67 	if (td->td_cowgen != p->p_cowgen)
68 		thread_cow_update(td);
69 	traced = (p->p_flag & P_TRACED) != 0;
70 	if (traced || td->td_dbgflags & TDB_USERWR) {
71 		PROC_LOCK(p);
72 		td->td_dbgflags &= ~TDB_USERWR;
73 		if (traced)
74 			td->td_dbgflags |= TDB_SCE;
75 		PROC_UNLOCK(p);
76 	}
77 	error = (p->p_sysent->sv_fetch_syscall_args)(td);
78 #ifdef KTRACE
79 	if (KTRPOINT(td, KTR_SYSCALL))
80 		ktrsyscall(sa->code, sa->narg, sa->args);
81 #endif
82 	KTR_START4(KTR_SYSC, "syscall", syscallname(p, sa->code),
83 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "arg0:%p", sa->args[0],
84 	    "arg1:%p", sa->args[1], "arg2:%p", sa->args[2]);
85 
86 	if (error == 0) {
87 
88 		STOPEVENT(p, S_SCE, sa->narg);
89 		if (p->p_flag & P_TRACED) {
90 			PROC_LOCK(p);
91 			if (p->p_ptevents & PTRACE_SCE)
92 				ptracestop((td), SIGTRAP, NULL);
93 			PROC_UNLOCK(p);
94 		}
95 		if (td->td_dbgflags & TDB_USERWR) {
96 			/*
97 			 * Reread syscall number and arguments if
98 			 * debugger modified registers or memory.
99 			 */
100 			error = (p->p_sysent->sv_fetch_syscall_args)(td);
101 #ifdef KTRACE
102 			if (KTRPOINT(td, KTR_SYSCALL))
103 				ktrsyscall(sa->code, sa->narg, sa->args);
104 #endif
105 			if (error != 0)
106 				goto retval;
107 		}
108 
109 #ifdef CAPABILITY_MODE
110 		/*
111 		 * In capability mode, we only allow access to system calls
112 		 * flagged with SYF_CAPENABLED.
113 		 */
114 		if (IN_CAPABILITY_MODE(td) &&
115 		    !(sa->callp->sy_flags & SYF_CAPENABLED)) {
116 			error = ECAPMODE;
117 			goto retval;
118 		}
119 #endif
120 
121 		error = syscall_thread_enter(td, sa->callp);
122 		if (error != 0)
123 			goto retval;
124 
125 #ifdef KDTRACE_HOOKS
126 		/* Give the syscall:::entry DTrace probe a chance to fire. */
127 		if (systrace_probe_func != NULL && sa->callp->sy_entry != 0)
128 			(*systrace_probe_func)(sa, SYSTRACE_ENTRY, 0);
129 #endif
130 
131 		AUDIT_SYSCALL_ENTER(sa->code, td);
132 		error = (sa->callp->sy_call)(td, sa->args);
133 		AUDIT_SYSCALL_EXIT(error, td);
134 
135 		/* Save the latest error return value. */
136 		if ((td->td_pflags & TDP_NERRNO) == 0)
137 			td->td_errno = error;
138 
139 #ifdef KDTRACE_HOOKS
140 		/* Give the syscall:::return DTrace probe a chance to fire. */
141 		if (systrace_probe_func != NULL && sa->callp->sy_return != 0)
142 			(*systrace_probe_func)(sa, SYSTRACE_RETURN,
143 			    error ? -1 : td->td_retval[0]);
144 #endif
145 		syscall_thread_exit(td, sa->callp);
146 	}
147  retval:
148 	KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code),
149 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error,
150 	    "retval0:%#lx", td->td_retval[0], "retval1:%#lx",
151 	    td->td_retval[1]);
152 	if (traced) {
153 		PROC_LOCK(p);
154 		td->td_dbgflags &= ~TDB_SCE;
155 		PROC_UNLOCK(p);
156 	}
157 	(p->p_sysent->sv_set_syscall_retval)(td, error);
158 	return (error);
159 }
160 
161 static inline void
162 syscallret(struct thread *td, int error)
163 {
164 	struct proc *p, *p2;
165 	struct syscall_args *sa;
166 	ksiginfo_t ksi;
167 	int traced, error1;
168 
169 	KASSERT((td->td_pflags & TDP_FORKING) == 0,
170 	    ("fork() did not clear TDP_FORKING upon completion"));
171 
172 	p = td->td_proc;
173 	sa = &td->td_sa;
174 	if ((trap_enotcap || (p->p_flag2 & P2_TRAPCAP) != 0) &&
175 	    IN_CAPABILITY_MODE(td)) {
176 		error1 = (td->td_pflags & TDP_NERRNO) == 0 ? error :
177 		    td->td_errno;
178 		if (error1 == ENOTCAPABLE || error1 == ECAPMODE) {
179 			ksiginfo_init_trap(&ksi);
180 			ksi.ksi_signo = SIGTRAP;
181 			ksi.ksi_errno = error1;
182 			ksi.ksi_code = TRAP_CAP;
183 			trapsignal(td, &ksi);
184 		}
185 	}
186 
187 	/*
188 	 * Handle reschedule and other end-of-syscall issues
189 	 */
190 	userret(td, td->td_frame);
191 
192 #ifdef KTRACE
193 	if (KTRPOINT(td, KTR_SYSRET)) {
194 		ktrsysret(sa->code, (td->td_pflags & TDP_NERRNO) == 0 ?
195 		    error : td->td_errno, td->td_retval[0]);
196 	}
197 #endif
198 	td->td_pflags &= ~TDP_NERRNO;
199 
200 	if (p->p_flag & P_TRACED) {
201 		traced = 1;
202 		PROC_LOCK(p);
203 		td->td_dbgflags |= TDB_SCX;
204 		PROC_UNLOCK(p);
205 	} else
206 		traced = 0;
207 	/*
208 	 * This works because errno is findable through the
209 	 * register set.  If we ever support an emulation where this
210 	 * is not the case, this code will need to be revisited.
211 	 */
212 	STOPEVENT(p, S_SCX, sa->code);
213 	if (traced || (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0) {
214 		PROC_LOCK(p);
215 		/*
216 		 * If tracing the execed process, trap to the debugger
217 		 * so that breakpoints can be set before the program
218 		 * executes.  If debugger requested tracing of syscall
219 		 * returns, do it now too.
220 		 */
221 		if (traced &&
222 		    ((td->td_dbgflags & (TDB_FORK | TDB_EXEC)) != 0 ||
223 		    (p->p_ptevents & PTRACE_SCX) != 0))
224 			ptracestop(td, SIGTRAP, NULL);
225 		td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK);
226 		PROC_UNLOCK(p);
227 	}
228 
229 	if (td->td_pflags & TDP_RFPPWAIT) {
230 		/*
231 		 * Preserve synchronization semantics of vfork.  If
232 		 * waiting for child to exec or exit, fork set
233 		 * P_PPWAIT on child, and there we sleep on our proc
234 		 * (in case of exit).
235 		 *
236 		 * Do it after the ptracestop() above is finished, to
237 		 * not block our debugger until child execs or exits
238 		 * to finish vfork wait.
239 		 */
240 		td->td_pflags &= ~TDP_RFPPWAIT;
241 		p2 = td->td_rfppwait_p;
242 again:
243 		PROC_LOCK(p2);
244 		while (p2->p_flag & P_PPWAIT) {
245 			PROC_LOCK(p);
246 			if (thread_suspend_check_needed()) {
247 				PROC_UNLOCK(p2);
248 				thread_suspend_check(0);
249 				PROC_UNLOCK(p);
250 				goto again;
251 			} else {
252 				PROC_UNLOCK(p);
253 			}
254 			cv_timedwait(&p2->p_pwait, &p2->p_mtx, hz);
255 		}
256 		PROC_UNLOCK(p2);
257 
258 		if (td->td_dbgflags & TDB_VFORK) {
259 			PROC_LOCK(p);
260 			if (p->p_ptevents & PTRACE_VFORK)
261 				ptracestop(td, SIGTRAP, NULL);
262 			td->td_dbgflags &= ~TDB_VFORK;
263 			PROC_UNLOCK(p);
264 		}
265 	}
266 }
267