xref: /freebsd/sys/kern/subr_syscall.c (revision b19cdab3456b361e1ef79651fe1437d8cab4de19)
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 	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 (__predict_false(td->td_cowgen != p->p_cowgen))
70 		thread_cow_update(td);
71 	traced = (p->p_flag & P_TRACED) != 0;
72 	if (__predict_false(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->callp->sy_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 (__predict_false(error != 0)) {
89 		td->td_errno = error;
90 		goto retval;
91 	}
92 
93 	if (__predict_false(traced)) {
94 		PROC_LOCK(p);
95 		if (p->p_ptevents & PTRACE_SCE)
96 			ptracestop((td), SIGTRAP, NULL);
97 		PROC_UNLOCK(p);
98 	}
99 	if (__predict_false((td->td_dbgflags & TDB_USERWR) != 0)) {
100 		/*
101 		 * Reread syscall number and arguments if debugger
102 		 * modified registers or memory.
103 		 */
104 		error = (p->p_sysent->sv_fetch_syscall_args)(td);
105 #ifdef KTRACE
106 		if (KTRPOINT(td, KTR_SYSCALL))
107 			ktrsyscall(sa->code, sa->callp->sy_narg, sa->args);
108 #endif
109 		if (error != 0) {
110 			td->td_errno = error;
111 			goto retval;
112 		}
113 	}
114 
115 #ifdef CAPABILITY_MODE
116 	/*
117 	 * In capability mode, we only allow access to system calls
118 	 * flagged with SYF_CAPENABLED.
119 	 */
120 	if (__predict_false(IN_CAPABILITY_MODE(td) &&
121 	    !(sa->callp->sy_flags & SYF_CAPENABLED))) {
122 		td->td_errno = error = ECAPMODE;
123 		goto retval;
124 	}
125 #endif
126 
127 	error = syscall_thread_enter(td, sa->callp);
128 	if (error != 0) {
129 		td->td_errno = error;
130 		goto retval;
131 	}
132 
133 	/*
134 	 * Fetch fast sigblock value at the time of syscall entry to
135 	 * handle sleepqueue primitives which might call cursig().
136 	 */
137 	if (__predict_false(sigfastblock_fetch_always))
138 		(void)sigfastblock_fetch(td);
139 
140 	/* Let system calls set td_errno directly. */
141 	KASSERT((td->td_pflags & TDP_NERRNO) == 0,
142 	    ("%s: TDP_NERRNO set", __func__));
143 
144 	if (__predict_false(SYSTRACE_ENABLED() ||
145 	    AUDIT_SYSCALL_ENTER(sa->code, td))) {
146 #ifdef KDTRACE_HOOKS
147 		/* Give the syscall:::entry DTrace probe a chance to fire. */
148 		if (__predict_false(sa->callp->sy_entry != 0))
149 			(*systrace_probe_func)(sa, SYSTRACE_ENTRY, 0);
150 #endif
151 		error = (sa->callp->sy_call)(td, sa->args);
152 		/* Save the latest error return value. */
153 		if (__predict_false((td->td_pflags & TDP_NERRNO) != 0))
154 			td->td_pflags &= ~TDP_NERRNO;
155 		else
156 			td->td_errno = error;
157 
158 		/*
159 		 * Note that some syscall implementations (e.g., sys_execve)
160 		 * will commit the audit record just before their final return.
161 		 * These were done under the assumption that nothing of interest
162 		 * would happen between their return and here, where we would
163 		 * normally commit the audit record.  These assumptions will
164 		 * need to be revisited should any substantial logic be added
165 		 * above.
166 		 */
167 		AUDIT_SYSCALL_EXIT(error, td);
168 
169 #ifdef KDTRACE_HOOKS
170 		/* Give the syscall:::return DTrace probe a chance to fire. */
171 		if (__predict_false(sa->callp->sy_return != 0))
172 			(*systrace_probe_func)(sa, SYSTRACE_RETURN,
173 			    error ? -1 : td->td_retval[0]);
174 #endif
175 	} else {
176 		error = (sa->callp->sy_call)(td, sa->args);
177 		/* Save the latest error return value. */
178 		if (__predict_false((td->td_pflags & TDP_NERRNO) != 0))
179 			td->td_pflags &= ~TDP_NERRNO;
180 		else
181 			td->td_errno = error;
182 	}
183 	syscall_thread_exit(td, sa->callp);
184 
185  retval:
186 	KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code),
187 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error,
188 	    "retval0:%#lx", td->td_retval[0], "retval1:%#lx",
189 	    td->td_retval[1]);
190 	if (__predict_false(traced)) {
191 		PROC_LOCK(p);
192 		td->td_dbgflags &= ~TDB_SCE;
193 		PROC_UNLOCK(p);
194 	}
195 	(p->p_sysent->sv_set_syscall_retval)(td, error);
196 }
197 
198 static inline void
199 syscallret(struct thread *td)
200 {
201 	struct proc *p;
202 	struct syscall_args *sa;
203 	ksiginfo_t ksi;
204 	int traced;
205 
206 	KASSERT((td->td_pflags & TDP_FORKING) == 0,
207 	    ("fork() did not clear TDP_FORKING upon completion"));
208 
209 	p = td->td_proc;
210 	sa = &td->td_sa;
211 	if (__predict_false(td->td_errno == ENOTCAPABLE ||
212 	    td->td_errno == ECAPMODE)) {
213 		if ((trap_enotcap ||
214 		    (p->p_flag2 & P2_TRAPCAP) != 0) && IN_CAPABILITY_MODE(td)) {
215 			ksiginfo_init_trap(&ksi);
216 			ksi.ksi_signo = SIGTRAP;
217 			ksi.ksi_errno = td->td_errno;
218 			ksi.ksi_code = TRAP_CAP;
219 			trapsignal(td, &ksi);
220 		}
221 	}
222 
223 	/*
224 	 * Handle reschedule and other end-of-syscall issues
225 	 */
226 	userret(td, td->td_frame);
227 
228 #ifdef KTRACE
229 	if (KTRPOINT(td, KTR_SYSRET)) {
230 		ktrsysret(sa->code, td->td_errno, td->td_retval[0]);
231 	}
232 #endif
233 
234 	traced = 0;
235 	if (__predict_false(p->p_flag & P_TRACED)) {
236 		traced = 1;
237 		PROC_LOCK(p);
238 		td->td_dbgflags |= TDB_SCX;
239 		PROC_UNLOCK(p);
240 	}
241 	if (__predict_false(traced ||
242 	    (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0)) {
243 		PROC_LOCK(p);
244 		/*
245 		 * If tracing the execed process, trap to the debugger
246 		 * so that breakpoints can be set before the program
247 		 * executes.  If debugger requested tracing of syscall
248 		 * returns, do it now too.
249 		 */
250 		if (traced &&
251 		    ((td->td_dbgflags & (TDB_FORK | TDB_EXEC)) != 0 ||
252 		    (p->p_ptevents & PTRACE_SCX) != 0))
253 			ptracestop(td, SIGTRAP, NULL);
254 		td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK);
255 		PROC_UNLOCK(p);
256 	}
257 
258 	if (__predict_false(td->td_pflags & TDP_RFPPWAIT))
259 		fork_rfppwait(td);
260 }
261