xref: /freebsd/sys/kern/subr_syscall.c (revision 4bfebc8d2c5d0e813dfdcbe7038b36ffc2bb9f1b)
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 void
58 syscallenter(struct thread *td)
59 {
60 	struct proc *p;
61 	struct syscall_args *sa;
62 	struct sysent *se;
63 	int error, traced;
64 
65 	VM_CNT_INC(v_syscall);
66 	p = td->td_proc;
67 	sa = &td->td_sa;
68 
69 	td->td_pticks = 0;
70 	if (__predict_false(td->td_cowgen != p->p_cowgen))
71 		thread_cow_update(td);
72 	traced = (p->p_flag & P_TRACED) != 0;
73 	if (__predict_false(traced || td->td_dbgflags & TDB_USERWR)) {
74 		PROC_LOCK(p);
75 		td->td_dbgflags &= ~TDB_USERWR;
76 		if (traced)
77 			td->td_dbgflags |= TDB_SCE;
78 		PROC_UNLOCK(p);
79 	}
80 	error = (p->p_sysent->sv_fetch_syscall_args)(td);
81 	se = sa->callp;
82 #ifdef KTRACE
83 	if (KTRPOINT(td, KTR_SYSCALL))
84 		ktrsyscall(sa->code, se->sy_narg, sa->args);
85 #endif
86 	KTR_START4(KTR_SYSC, "syscall", syscallname(p, sa->code),
87 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "arg0:%p", sa->args[0],
88 	    "arg1:%p", sa->args[1], "arg2:%p", sa->args[2]);
89 
90 	if (__predict_false(error != 0)) {
91 		td->td_errno = error;
92 		goto retval;
93 	}
94 
95 	if (__predict_false(traced)) {
96 		PROC_LOCK(p);
97 		if (p->p_ptevents & PTRACE_SCE)
98 			ptracestop((td), SIGTRAP, NULL);
99 		PROC_UNLOCK(p);
100 	}
101 	if (__predict_false((td->td_dbgflags & TDB_USERWR) != 0)) {
102 		/*
103 		 * Reread syscall number and arguments if debugger
104 		 * modified registers or memory.
105 		 */
106 		error = (p->p_sysent->sv_fetch_syscall_args)(td);
107 		se = sa->callp;
108 #ifdef KTRACE
109 		if (KTRPOINT(td, KTR_SYSCALL))
110 			ktrsyscall(sa->code, se->sy_narg, sa->args);
111 #endif
112 		if (error != 0) {
113 			td->td_errno = error;
114 			goto retval;
115 		}
116 	}
117 
118 #ifdef CAPABILITY_MODE
119 	/*
120 	 * In capability mode, we only allow access to system calls
121 	 * flagged with SYF_CAPENABLED.
122 	 */
123 	if (__predict_false(IN_CAPABILITY_MODE(td) &&
124 	    (se->sy_flags & SYF_CAPENABLED) == 0)) {
125 		td->td_errno = error = ECAPMODE;
126 		goto retval;
127 	}
128 #endif
129 
130 	error = syscall_thread_enter(td, se);
131 	if (error != 0) {
132 		td->td_errno = error;
133 		goto retval;
134 	}
135 
136 	/*
137 	 * Fetch fast sigblock value at the time of syscall entry to
138 	 * handle sleepqueue primitives which might call cursig().
139 	 */
140 	if (__predict_false(sigfastblock_fetch_always))
141 		(void)sigfastblock_fetch(td);
142 
143 	/* Let system calls set td_errno directly. */
144 	KASSERT((td->td_pflags & TDP_NERRNO) == 0,
145 	    ("%s: TDP_NERRNO set", __func__));
146 
147 	if (__predict_false(SYSTRACE_ENABLED() ||
148 	    AUDIT_SYSCALL_ENTER(sa->code, td))) {
149 #ifdef KDTRACE_HOOKS
150 		/* Give the syscall:::entry DTrace probe a chance to fire. */
151 		if (__predict_false(se->sy_entry != 0))
152 			(*systrace_probe_func)(sa, SYSTRACE_ENTRY, 0);
153 #endif
154 		error = (se->sy_call)(td, sa->args);
155 		/* Save the latest error return value. */
156 		if (__predict_false((td->td_pflags & TDP_NERRNO) != 0))
157 			td->td_pflags &= ~TDP_NERRNO;
158 		else
159 			td->td_errno = error;
160 
161 		/*
162 		 * Note that some syscall implementations (e.g., sys_execve)
163 		 * will commit the audit record just before their final return.
164 		 * These were done under the assumption that nothing of interest
165 		 * would happen between their return and here, where we would
166 		 * normally commit the audit record.  These assumptions will
167 		 * need to be revisited should any substantial logic be added
168 		 * above.
169 		 */
170 		AUDIT_SYSCALL_EXIT(error, td);
171 
172 #ifdef KDTRACE_HOOKS
173 		/* Give the syscall:::return DTrace probe a chance to fire. */
174 		if (__predict_false(se->sy_return != 0))
175 			(*systrace_probe_func)(sa, SYSTRACE_RETURN,
176 			    error ? -1 : td->td_retval[0]);
177 #endif
178 	} else {
179 		error = (se->sy_call)(td, sa->args);
180 		/* Save the latest error return value. */
181 		if (__predict_false((td->td_pflags & TDP_NERRNO) != 0))
182 			td->td_pflags &= ~TDP_NERRNO;
183 		else
184 			td->td_errno = error;
185 	}
186 	syscall_thread_exit(td, se);
187 
188  retval:
189 	KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code),
190 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error,
191 	    "retval0:%#lx", td->td_retval[0], "retval1:%#lx",
192 	    td->td_retval[1]);
193 	if (__predict_false(traced)) {
194 		PROC_LOCK(p);
195 		td->td_dbgflags &= ~TDB_SCE;
196 		PROC_UNLOCK(p);
197 	}
198 	(p->p_sysent->sv_set_syscall_retval)(td, error);
199 }
200 
201 static inline void
202 syscallret(struct thread *td)
203 {
204 	struct proc *p;
205 	struct syscall_args *sa;
206 	ksiginfo_t ksi;
207 	int traced;
208 
209 	KASSERT((td->td_pflags & TDP_FORKING) == 0,
210 	    ("fork() did not clear TDP_FORKING upon completion"));
211 
212 	p = td->td_proc;
213 	sa = &td->td_sa;
214 	if (__predict_false(td->td_errno == ENOTCAPABLE ||
215 	    td->td_errno == ECAPMODE)) {
216 		if ((trap_enotcap ||
217 		    (p->p_flag2 & P2_TRAPCAP) != 0) && IN_CAPABILITY_MODE(td)) {
218 			ksiginfo_init_trap(&ksi);
219 			ksi.ksi_signo = SIGTRAP;
220 			ksi.ksi_errno = td->td_errno;
221 			ksi.ksi_code = TRAP_CAP;
222 			trapsignal(td, &ksi);
223 		}
224 	}
225 
226 	/*
227 	 * Handle reschedule and other end-of-syscall issues
228 	 */
229 	userret(td, td->td_frame);
230 
231 #ifdef KTRACE
232 	if (KTRPOINT(td, KTR_SYSRET)) {
233 		ktrsysret(sa->code, td->td_errno, td->td_retval[0]);
234 	}
235 #endif
236 
237 	traced = 0;
238 	if (__predict_false(p->p_flag & P_TRACED)) {
239 		traced = 1;
240 		PROC_LOCK(p);
241 		td->td_dbgflags |= TDB_SCX;
242 		PROC_UNLOCK(p);
243 	}
244 	if (__predict_false(traced ||
245 	    (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0)) {
246 		PROC_LOCK(p);
247 		/*
248 		 * If tracing the execed process, trap to the debugger
249 		 * so that breakpoints can be set before the program
250 		 * executes.  If debugger requested tracing of syscall
251 		 * returns, do it now too.
252 		 */
253 		if (traced &&
254 		    ((td->td_dbgflags & (TDB_FORK | TDB_EXEC)) != 0 ||
255 		    (p->p_ptevents & PTRACE_SCX) != 0))
256 			ptracestop(td, SIGTRAP, NULL);
257 		td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK);
258 		PROC_UNLOCK(p);
259 	}
260 
261 	if (__predict_false(td->td_pflags & TDP_RFPPWAIT))
262 		fork_rfppwait(td);
263 }
264