xref: /freebsd/sys/kern/subr_syscall.c (revision b0d29bc47dba79f6f38e67eabadfb4b32ffd9390)
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 	STOPEVENT(p, S_SCE, sa->narg);
94 	if (__predict_false((p->p_flag & P_TRACED) != 0)) {
95 		PROC_LOCK(p);
96 		if (p->p_ptevents & PTRACE_SCE)
97 			ptracestop((td), SIGTRAP, NULL);
98 		PROC_UNLOCK(p);
99 	}
100 	if (__predict_false((td->td_dbgflags & TDB_USERWR) != 0)) {
101 		/*
102 		 * Reread syscall number and arguments if debugger
103 		 * modified registers or memory.
104 		 */
105 		error = (p->p_sysent->sv_fetch_syscall_args)(td);
106 #ifdef KTRACE
107 		if (KTRPOINT(td, KTR_SYSCALL))
108 			ktrsyscall(sa->code, sa->narg, sa->args);
109 #endif
110 		if (error != 0) {
111 			td->td_errno = error;
112 			goto retval;
113 		}
114 	}
115 
116 #ifdef CAPABILITY_MODE
117 	/*
118 	 * In capability mode, we only allow access to system calls
119 	 * flagged with SYF_CAPENABLED.
120 	 */
121 	if (__predict_false(IN_CAPABILITY_MODE(td) &&
122 	    !(sa->callp->sy_flags & SYF_CAPENABLED))) {
123 		td->td_errno = error = ECAPMODE;
124 		goto retval;
125 	}
126 #endif
127 
128 	error = syscall_thread_enter(td, sa->callp);
129 	if (error != 0) {
130 		td->td_errno = error;
131 		goto retval;
132 	}
133 
134 	/*
135 	 * Fetch fast sigblock value at the time of syscall entry to
136 	 * handle sleepqueue primitives which might call cursig().
137 	 */
138 	if (__predict_false(sigfastblock_fetch_always))
139 		(void)sigfastblock_fetch(td);
140 
141 	/* Let system calls set td_errno directly. */
142 	td->td_pflags &= ~TDP_NERRNO;
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_errno = error;
155 		AUDIT_SYSCALL_EXIT(error, td);
156 #ifdef KDTRACE_HOOKS
157 		/* Give the syscall:::return DTrace probe a chance to fire. */
158 		if (__predict_false(sa->callp->sy_return != 0))
159 			(*systrace_probe_func)(sa, SYSTRACE_RETURN,
160 			    error ? -1 : td->td_retval[0]);
161 #endif
162 	} else {
163 		error = (sa->callp->sy_call)(td, sa->args);
164 		/* Save the latest error return value. */
165 		if (__predict_false((td->td_pflags & TDP_NERRNO) == 0))
166 			td->td_errno = error;
167 	}
168 	syscall_thread_exit(td, sa->callp);
169 
170  retval:
171 	KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code),
172 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error,
173 	    "retval0:%#lx", td->td_retval[0], "retval1:%#lx",
174 	    td->td_retval[1]);
175 	if (__predict_false(traced)) {
176 		PROC_LOCK(p);
177 		td->td_dbgflags &= ~TDB_SCE;
178 		PROC_UNLOCK(p);
179 	}
180 	(p->p_sysent->sv_set_syscall_retval)(td, error);
181 }
182 
183 static inline void
184 syscallret(struct thread *td)
185 {
186 	struct proc *p;
187 	struct syscall_args *sa;
188 	ksiginfo_t ksi;
189 	int traced;
190 
191 	KASSERT((td->td_pflags & TDP_FORKING) == 0,
192 	    ("fork() did not clear TDP_FORKING upon completion"));
193 
194 	p = td->td_proc;
195 	sa = &td->td_sa;
196 	if (__predict_false(td->td_errno == ENOTCAPABLE ||
197 	    td->td_errno == ECAPMODE)) {
198 		if ((trap_enotcap ||
199 		    (p->p_flag2 & P2_TRAPCAP) != 0) && IN_CAPABILITY_MODE(td)) {
200 			ksiginfo_init_trap(&ksi);
201 			ksi.ksi_signo = SIGTRAP;
202 			ksi.ksi_errno = td->td_errno;
203 			ksi.ksi_code = TRAP_CAP;
204 			trapsignal(td, &ksi);
205 		}
206 	}
207 
208 	/*
209 	 * Handle reschedule and other end-of-syscall issues
210 	 */
211 	userret(td, td->td_frame);
212 
213 #ifdef KTRACE
214 	if (KTRPOINT(td, KTR_SYSRET)) {
215 		ktrsysret(sa->code, td->td_errno, td->td_retval[0]);
216 	}
217 #endif
218 
219 	traced = 0;
220 	if (__predict_false(p->p_flag & P_TRACED)) {
221 		traced = 1;
222 		PROC_LOCK(p);
223 		td->td_dbgflags |= TDB_SCX;
224 		PROC_UNLOCK(p);
225 	}
226 	/*
227 	 * This works because errno is findable through the
228 	 * register set.  If we ever support an emulation where this
229 	 * is not the case, this code will need to be revisited.
230 	 */
231 	STOPEVENT(p, S_SCX, sa->code);
232 	if (__predict_false(traced ||
233 	    (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0)) {
234 		PROC_LOCK(p);
235 		/*
236 		 * If tracing the execed process, trap to the debugger
237 		 * so that breakpoints can be set before the program
238 		 * executes.  If debugger requested tracing of syscall
239 		 * returns, do it now too.
240 		 */
241 		if (traced &&
242 		    ((td->td_dbgflags & (TDB_FORK | TDB_EXEC)) != 0 ||
243 		    (p->p_ptevents & PTRACE_SCX) != 0))
244 			ptracestop(td, SIGTRAP, NULL);
245 		td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK);
246 		PROC_UNLOCK(p);
247 	}
248 
249 	if (__predict_false(td->td_pflags & TDP_RFPPWAIT))
250 		fork_rfppwait(td);
251 }
252