xref: /freebsd/sys/kern/subr_syscall.c (revision f0cfa1b168014f56c02b83e5f28412cc5f78d117)
1 /*-
2  * SPDX-License-Identifier: BSD-4-Clause
3  *
4  * Copyright (C) 1994, David Greenman
5  * Copyright (c) 1990, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  * Copyright (C) 2010 Konstantin Belousov <kib@freebsd.org>
8  *
9  * This code is derived from software contributed to Berkeley by
10  * the University of Utah, and William Jolitz.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. All advertising materials mentioning features or use of this software
21  *    must display the following acknowledgement:
22  *	This product includes software developed by the University of
23  *	California, Berkeley and its contributors.
24  * 4. Neither the name of the University nor the names of its contributors
25  *    may be used to endorse or promote products derived from this software
26  *    without specific prior written permission.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38  * SUCH DAMAGE.
39  *
40  *	from: @(#)trap.c	7.4 (Berkeley) 5/13/91
41  */
42 
43 #include "opt_capsicum.h"
44 #include "opt_ktrace.h"
45 
46 __FBSDID("$FreeBSD$");
47 
48 #include <sys/capsicum.h>
49 #include <sys/ktr.h>
50 #include <sys/vmmeter.h>
51 #ifdef KTRACE
52 #include <sys/uio.h>
53 #include <sys/ktrace.h>
54 #endif
55 #include <security/audit/audit.h>
56 
57 static inline int
58 syscallenter(struct thread *td)
59 {
60 	struct proc *p;
61 	struct syscall_args *sa;
62 	int error, traced;
63 
64 	VM_CNT_INC(v_syscall);
65 	p = td->td_proc;
66 	sa = &td->td_sa;
67 
68 	td->td_pticks = 0;
69 	if (td->td_cowgen != p->p_cowgen)
70 		thread_cow_update(td);
71 	traced = (p->p_flag & P_TRACED) != 0;
72 	if (traced || td->td_dbgflags & TDB_USERWR) {
73 		PROC_LOCK(p);
74 		td->td_dbgflags &= ~TDB_USERWR;
75 		if (traced)
76 			td->td_dbgflags |= TDB_SCE;
77 		PROC_UNLOCK(p);
78 	}
79 	error = (p->p_sysent->sv_fetch_syscall_args)(td);
80 #ifdef KTRACE
81 	if (KTRPOINT(td, KTR_SYSCALL))
82 		ktrsyscall(sa->code, sa->narg, sa->args);
83 #endif
84 	KTR_START4(KTR_SYSC, "syscall", syscallname(p, sa->code),
85 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "arg0:%p", sa->args[0],
86 	    "arg1:%p", sa->args[1], "arg2:%p", sa->args[2]);
87 
88 	if (error == 0) {
89 
90 		STOPEVENT(p, S_SCE, sa->narg);
91 		if (p->p_flag & P_TRACED) {
92 			PROC_LOCK(p);
93 			if (p->p_ptevents & PTRACE_SCE)
94 				ptracestop((td), SIGTRAP, NULL);
95 			PROC_UNLOCK(p);
96 		}
97 		if (td->td_dbgflags & TDB_USERWR) {
98 			/*
99 			 * Reread syscall number and arguments if
100 			 * debugger modified registers or memory.
101 			 */
102 			error = (p->p_sysent->sv_fetch_syscall_args)(td);
103 #ifdef KTRACE
104 			if (KTRPOINT(td, KTR_SYSCALL))
105 				ktrsyscall(sa->code, sa->narg, sa->args);
106 #endif
107 			if (error != 0)
108 				goto retval;
109 		}
110 
111 #ifdef CAPABILITY_MODE
112 		/*
113 		 * In capability mode, we only allow access to system calls
114 		 * flagged with SYF_CAPENABLED.
115 		 */
116 		if (IN_CAPABILITY_MODE(td) &&
117 		    !(sa->callp->sy_flags & SYF_CAPENABLED)) {
118 			error = ECAPMODE;
119 			goto retval;
120 		}
121 #endif
122 
123 		error = syscall_thread_enter(td, sa->callp);
124 		if (error != 0)
125 			goto retval;
126 
127 #ifdef KDTRACE_HOOKS
128 		/* Give the syscall:::entry DTrace probe a chance to fire. */
129 		if (systrace_probe_func != NULL && sa->callp->sy_entry != 0)
130 			(*systrace_probe_func)(sa, SYSTRACE_ENTRY, 0);
131 #endif
132 
133 		AUDIT_SYSCALL_ENTER(sa->code, td);
134 		error = (sa->callp->sy_call)(td, sa->args);
135 		AUDIT_SYSCALL_EXIT(error, td);
136 
137 		/* Save the latest error return value. */
138 		if ((td->td_pflags & TDP_NERRNO) == 0)
139 			td->td_errno = error;
140 
141 #ifdef KDTRACE_HOOKS
142 		/* Give the syscall:::return DTrace probe a chance to fire. */
143 		if (systrace_probe_func != NULL && sa->callp->sy_return != 0)
144 			(*systrace_probe_func)(sa, SYSTRACE_RETURN,
145 			    error ? -1 : td->td_retval[0]);
146 #endif
147 		syscall_thread_exit(td, sa->callp);
148 	}
149  retval:
150 	KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code),
151 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error,
152 	    "retval0:%#lx", td->td_retval[0], "retval1:%#lx",
153 	    td->td_retval[1]);
154 	if (traced) {
155 		PROC_LOCK(p);
156 		td->td_dbgflags &= ~TDB_SCE;
157 		PROC_UNLOCK(p);
158 	}
159 	(p->p_sysent->sv_set_syscall_retval)(td, error);
160 	return (error);
161 }
162 
163 static inline void
164 syscallret(struct thread *td, int error)
165 {
166 	struct proc *p, *p2;
167 	struct syscall_args *sa;
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 	sa = &td->td_sa;
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