xref: /freebsd/sys/kern/subr_syscall.c (revision 6419bb52)
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->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((p->p_flag & P_TRACED) != 0)) {
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->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 	td->td_pflags &= ~TDP_NERRNO;
142 
143 	if (__predict_false(SYSTRACE_ENABLED() ||
144 	    AUDIT_SYSCALL_ENTER(sa->code, td))) {
145 #ifdef KDTRACE_HOOKS
146 		/* Give the syscall:::entry DTrace probe a chance to fire. */
147 		if (__predict_false(sa->callp->sy_entry != 0))
148 			(*systrace_probe_func)(sa, SYSTRACE_ENTRY, 0);
149 #endif
150 		error = (sa->callp->sy_call)(td, sa->args);
151 		/* Save the latest error return value. */
152 		if (__predict_false((td->td_pflags & TDP_NERRNO) == 0))
153 			td->td_errno = error;
154 		AUDIT_SYSCALL_EXIT(error, td);
155 #ifdef KDTRACE_HOOKS
156 		/* Give the syscall:::return DTrace probe a chance to fire. */
157 		if (__predict_false(sa->callp->sy_return != 0))
158 			(*systrace_probe_func)(sa, SYSTRACE_RETURN,
159 			    error ? -1 : td->td_retval[0]);
160 #endif
161 	} else {
162 		error = (sa->callp->sy_call)(td, sa->args);
163 		/* Save the latest error return value. */
164 		if (__predict_false((td->td_pflags & TDP_NERRNO) == 0))
165 			td->td_errno = error;
166 	}
167 	syscall_thread_exit(td, sa->callp);
168 
169  retval:
170 	KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code),
171 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error,
172 	    "retval0:%#lx", td->td_retval[0], "retval1:%#lx",
173 	    td->td_retval[1]);
174 	if (__predict_false(traced)) {
175 		PROC_LOCK(p);
176 		td->td_dbgflags &= ~TDB_SCE;
177 		PROC_UNLOCK(p);
178 	}
179 	(p->p_sysent->sv_set_syscall_retval)(td, error);
180 }
181 
182 static inline void
183 syscallret(struct thread *td)
184 {
185 	struct proc *p;
186 	struct syscall_args *sa;
187 	ksiginfo_t ksi;
188 	int traced;
189 
190 	KASSERT((td->td_pflags & TDP_FORKING) == 0,
191 	    ("fork() did not clear TDP_FORKING upon completion"));
192 
193 	p = td->td_proc;
194 	sa = &td->td_sa;
195 	if (__predict_false(td->td_errno == ENOTCAPABLE ||
196 	    td->td_errno == ECAPMODE)) {
197 		if ((trap_enotcap ||
198 		    (p->p_flag2 & P2_TRAPCAP) != 0) && IN_CAPABILITY_MODE(td)) {
199 			ksiginfo_init_trap(&ksi);
200 			ksi.ksi_signo = SIGTRAP;
201 			ksi.ksi_errno = td->td_errno;
202 			ksi.ksi_code = TRAP_CAP;
203 			trapsignal(td, &ksi);
204 		}
205 	}
206 
207 	/*
208 	 * Handle reschedule and other end-of-syscall issues
209 	 */
210 	userret(td, td->td_frame);
211 
212 #ifdef KTRACE
213 	if (KTRPOINT(td, KTR_SYSRET)) {
214 		ktrsysret(sa->code, td->td_errno, td->td_retval[0]);
215 	}
216 #endif
217 
218 	traced = 0;
219 	if (__predict_false(p->p_flag & P_TRACED)) {
220 		traced = 1;
221 		PROC_LOCK(p);
222 		td->td_dbgflags |= TDB_SCX;
223 		PROC_UNLOCK(p);
224 	}
225 	if (__predict_false(traced ||
226 	    (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0)) {
227 		PROC_LOCK(p);
228 		/*
229 		 * If tracing the execed process, trap to the debugger
230 		 * so that breakpoints can be set before the program
231 		 * executes.  If debugger requested tracing of syscall
232 		 * returns, do it now too.
233 		 */
234 		if (traced &&
235 		    ((td->td_dbgflags & (TDB_FORK | TDB_EXEC)) != 0 ||
236 		    (p->p_ptevents & PTRACE_SCX) != 0))
237 			ptracestop(td, SIGTRAP, NULL);
238 		td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK);
239 		PROC_UNLOCK(p);
240 	}
241 
242 	if (__predict_false(td->td_pflags & TDP_RFPPWAIT))
243 		fork_rfppwait(td);
244 }
245