xref: /freebsd/sys/kern/subr_syscall.c (revision f56f82e0)
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, struct syscall_args *sa)
57 {
58 	struct proc *p;
59 	int error, traced;
60 
61 	VM_CNT_INC(v_syscall);
62 	p = td->td_proc;
63 
64 	td->td_pticks = 0;
65 	if (td->td_cowgen != p->p_cowgen)
66 		thread_cow_update(td);
67 	traced = (p->p_flag & P_TRACED) != 0;
68 	if (traced || td->td_dbgflags & TDB_USERWR) {
69 		PROC_LOCK(p);
70 		td->td_dbgflags &= ~TDB_USERWR;
71 		if (traced)
72 			td->td_dbgflags |= TDB_SCE;
73 		PROC_UNLOCK(p);
74 	}
75 	error = (p->p_sysent->sv_fetch_syscall_args)(td, sa);
76 #ifdef KTRACE
77 	if (KTRPOINT(td, KTR_SYSCALL))
78 		ktrsyscall(sa->code, sa->narg, sa->args);
79 #endif
80 	KTR_START4(KTR_SYSC, "syscall", syscallname(p, sa->code),
81 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "arg0:%p", sa->args[0],
82 	    "arg1:%p", sa->args[1], "arg2:%p", sa->args[2]);
83 
84 	if (error == 0) {
85 
86 		STOPEVENT(p, S_SCE, sa->narg);
87 		if (p->p_flag & P_TRACED) {
88 			PROC_LOCK(p);
89 			td->td_dbg_sc_code = sa->code;
90 			td->td_dbg_sc_narg = sa->narg;
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, sa);
101 			PROC_LOCK(p);
102 			td->td_dbg_sc_code = sa->code;
103 			td->td_dbg_sc_narg = sa->narg;
104 			PROC_UNLOCK(p);
105 #ifdef KTRACE
106 			if (KTRPOINT(td, KTR_SYSCALL))
107 				ktrsyscall(sa->code, sa->narg, sa->args);
108 #endif
109 			if (error != 0)
110 				goto retval;
111 		}
112 
113 #ifdef CAPABILITY_MODE
114 		/*
115 		 * In capability mode, we only allow access to system calls
116 		 * flagged with SYF_CAPENABLED.
117 		 */
118 		if (IN_CAPABILITY_MODE(td) &&
119 		    !(sa->callp->sy_flags & SYF_CAPENABLED)) {
120 			error = ECAPMODE;
121 			goto retval;
122 		}
123 #endif
124 
125 		error = syscall_thread_enter(td, sa->callp);
126 		if (error != 0)
127 			goto retval;
128 
129 #ifdef KDTRACE_HOOKS
130 		/* Give the syscall:::entry DTrace probe a chance to fire. */
131 		if (systrace_probe_func != NULL && sa->callp->sy_entry != 0)
132 			(*systrace_probe_func)(sa, SYSTRACE_ENTRY, 0);
133 #endif
134 
135 		AUDIT_SYSCALL_ENTER(sa->code, td);
136 		error = (sa->callp->sy_call)(td, sa->args);
137 		AUDIT_SYSCALL_EXIT(error, td);
138 
139 		/* Save the latest error return value. */
140 		if ((td->td_pflags & TDP_NERRNO) == 0)
141 			td->td_errno = error;
142 
143 #ifdef KDTRACE_HOOKS
144 		/* Give the syscall:::return DTrace probe a chance to fire. */
145 		if (systrace_probe_func != NULL && sa->callp->sy_return != 0)
146 			(*systrace_probe_func)(sa, SYSTRACE_RETURN,
147 			    error ? -1 : td->td_retval[0]);
148 #endif
149 		syscall_thread_exit(td, sa->callp);
150 	}
151  retval:
152 	KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code),
153 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error,
154 	    "retval0:%#lx", td->td_retval[0], "retval1:%#lx",
155 	    td->td_retval[1]);
156 	if (traced) {
157 		PROC_LOCK(p);
158 		td->td_dbgflags &= ~TDB_SCE;
159 		PROC_UNLOCK(p);
160 	}
161 	(p->p_sysent->sv_set_syscall_retval)(td, error);
162 	return (error);
163 }
164 
165 static inline void
166 syscallret(struct thread *td, int error, struct syscall_args *sa)
167 {
168 	struct proc *p, *p2;
169 	ksiginfo_t ksi;
170 	int traced, error1;
171 
172 	KASSERT((td->td_pflags & TDP_FORKING) == 0,
173 	    ("fork() did not clear TDP_FORKING upon completion"));
174 
175 	p = td->td_proc;
176 	if ((trap_enotcap || (p->p_flag2 & P2_TRAPCAP) != 0) &&
177 	    IN_CAPABILITY_MODE(td)) {
178 		error1 = (td->td_pflags & TDP_NERRNO) == 0 ? error :
179 		    td->td_errno;
180 		if (error1 == ENOTCAPABLE || error1 == ECAPMODE) {
181 			ksiginfo_init_trap(&ksi);
182 			ksi.ksi_signo = SIGTRAP;
183 			ksi.ksi_errno = error1;
184 			ksi.ksi_code = TRAP_CAP;
185 			trapsignal(td, &ksi);
186 		}
187 	}
188 
189 	/*
190 	 * Handle reschedule and other end-of-syscall issues
191 	 */
192 	userret(td, td->td_frame);
193 
194 #ifdef KTRACE
195 	if (KTRPOINT(td, KTR_SYSRET)) {
196 		ktrsysret(sa->code, (td->td_pflags & TDP_NERRNO) == 0 ?
197 		    error : td->td_errno, td->td_retval[0]);
198 	}
199 #endif
200 	td->td_pflags &= ~TDP_NERRNO;
201 
202 	if (p->p_flag & P_TRACED) {
203 		traced = 1;
204 		PROC_LOCK(p);
205 		td->td_dbgflags |= TDB_SCX;
206 		PROC_UNLOCK(p);
207 	} else
208 		traced = 0;
209 	/*
210 	 * This works because errno is findable through the
211 	 * register set.  If we ever support an emulation where this
212 	 * is not the case, this code will need to be revisited.
213 	 */
214 	STOPEVENT(p, S_SCX, sa->code);
215 	if (traced || (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0) {
216 		PROC_LOCK(p);
217 		/*
218 		 * If tracing the execed process, trap to the debugger
219 		 * so that breakpoints can be set before the program
220 		 * executes.  If debugger requested tracing of syscall
221 		 * returns, do it now too.
222 		 */
223 		if (traced &&
224 		    ((td->td_dbgflags & (TDB_FORK | TDB_EXEC)) != 0 ||
225 		    (p->p_ptevents & PTRACE_SCX) != 0))
226 			ptracestop(td, SIGTRAP, NULL);
227 		td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK);
228 		PROC_UNLOCK(p);
229 	}
230 
231 	if (td->td_pflags & TDP_RFPPWAIT) {
232 		/*
233 		 * Preserve synchronization semantics of vfork.  If
234 		 * waiting for child to exec or exit, fork set
235 		 * P_PPWAIT on child, and there we sleep on our proc
236 		 * (in case of exit).
237 		 *
238 		 * Do it after the ptracestop() above is finished, to
239 		 * not block our debugger until child execs or exits
240 		 * to finish vfork wait.
241 		 */
242 		td->td_pflags &= ~TDP_RFPPWAIT;
243 		p2 = td->td_rfppwait_p;
244 again:
245 		PROC_LOCK(p2);
246 		while (p2->p_flag & P_PPWAIT) {
247 			PROC_LOCK(p);
248 			if (thread_suspend_check_needed()) {
249 				PROC_UNLOCK(p2);
250 				thread_suspend_check(0);
251 				PROC_UNLOCK(p);
252 				goto again;
253 			} else {
254 				PROC_UNLOCK(p);
255 			}
256 			cv_timedwait(&p2->p_pwait, &p2->p_mtx, hz);
257 		}
258 		PROC_UNLOCK(p2);
259 
260 		if (td->td_dbgflags & TDB_VFORK) {
261 			PROC_LOCK(p);
262 			if (p->p_ptevents & PTRACE_VFORK)
263 				ptracestop(td, SIGTRAP, NULL);
264 			td->td_dbgflags &= ~TDB_VFORK;
265 			PROC_UNLOCK(p);
266 		}
267 	}
268 }
269