xref: /freebsd/sys/kern/subr_syscall.c (revision 09dfe066f00c927e88c23265387d432e6d9f0c5e)
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 
41 #include "opt_capsicum.h"
42 #include "opt_ktrace.h"
43 #include <sys/capsicum.h>
44 #include <sys/ktr.h>
45 #include <sys/vmmeter.h>
46 #ifdef KTRACE
47 #include <sys/uio.h>
48 #include <sys/ktrace.h>
49 #endif
50 #include <security/audit/audit.h>
51 
52 static inline void
syscallenter(struct thread * td)53 syscallenter(struct thread *td)
54 {
55 	struct proc *p;
56 	struct syscall_args *sa;
57 	struct sysent *se;
58 	int error, traced;
59 	bool sy_thr_static;
60 
61 	VM_CNT_INC(v_syscall);
62 	p = td->td_proc;
63 	sa = &td->td_sa;
64 
65 	td->td_pticks = 0;
66 	if (__predict_false(td->td_cowgen != atomic_load_int(&p->p_cowgen)))
67 		thread_cow_update(td);
68 	traced = (p->p_flag & P_TRACED) != 0;
69 	if (__predict_false(traced || td->td_dbgflags & TDB_USERWR)) {
70 		PROC_LOCK(p);
71 		MPASS((td->td_dbgflags & TDB_BOUNDARY) == 0);
72 		td->td_dbgflags &= ~TDB_USERWR;
73 		if (traced)
74 			td->td_dbgflags |= TDB_SCE;
75 		PROC_UNLOCK(p);
76 	}
77 	if ((td->td_pflags2 & TDP2_UEXTERR) != 0)
78 		td->td_pflags2 &= ~TDP2_EXTERR;
79 	error = (p->p_sysent->sv_fetch_syscall_args)(td);
80 	se = sa->callp;
81 #ifdef KTRACE
82 	if (KTRPOINT(td, KTR_SYSCALL))
83 		ktrsyscall(sa->code, se->sy_narg, sa->args);
84 #endif
85 	KTR_START4(KTR_SYSC, "syscall", syscallname(p, sa->code),
86 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "arg0:%p", sa->args[0],
87 	    "arg1:%p", sa->args[1], "arg2:%p", sa->args[2]);
88 
89 	if (__predict_false(error != 0)) {
90 		td->td_errno = error;
91 		goto retval;
92 	}
93 
94 	if (__predict_false(traced)) {
95 		PROC_LOCK(p);
96 		if (p->p_ptevents & PTRACE_SCE)
97 			ptracestop((td), SIGTRAP, NULL);
98 		PROC_UNLOCK(p);
99 
100 		if ((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 			se = sa->callp;
107 #ifdef KTRACE
108 			if (KTRPOINT(td, KTR_SYSCALL))
109 				ktrsyscall(sa->code, se->sy_narg, sa->args);
110 #endif
111 			if (error != 0) {
112 				td->td_errno = error;
113 				goto retval;
114 			}
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 ((se->sy_flags & SYF_CAPENABLED) == 0) {
124 		if (CAP_TRACING(td))
125 			ktrcapfail(CAPFAIL_SYSCALL, NULL);
126 		if (IN_CAPABILITY_MODE(td)) {
127 			td->td_errno = error = ECAPMODE;
128 			goto retval;
129 		}
130 	}
131 #endif
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 	sy_thr_static = (se->sy_thrcnt & SY_THR_STATIC) != 0;
145 
146 	if (__predict_false(AUDIT_SYSCALL_ENABLED() ||
147 	    SYSTRACE_ENABLED() || !sy_thr_static)) {
148 		if (!sy_thr_static) {
149 			error = syscall_thread_enter(td, &se);
150 			sy_thr_static = (se->sy_thrcnt & SY_THR_STATIC) != 0;
151 			if (error != 0) {
152 				td->td_errno = error;
153 				goto retval;
154 			}
155 		}
156 
157 #ifdef KDTRACE_HOOKS
158 		/* Give the syscall:::entry DTrace probe a chance to fire. */
159 		if (__predict_false(se->sy_entry != 0))
160 			(*systrace_probe_func)(sa, SYSTRACE_ENTRY, 0);
161 #endif
162 
163 		AUDIT_SYSCALL_ENTER(sa->code, td);
164 
165 		error = (se->sy_call)(td, sa->args);
166 		/* Save the latest error return value. */
167 		if (__predict_false((td->td_pflags & TDP_NERRNO) != 0))
168 			td->td_pflags &= ~TDP_NERRNO;
169 		else
170 			td->td_errno = error;
171 
172 		/*
173 		 * Note that some syscall implementations (e.g., sys_execve)
174 		 * will commit the audit record just before their final return.
175 		 * These were done under the assumption that nothing of interest
176 		 * would happen between their return and here, where we would
177 		 * normally commit the audit record.  These assumptions will
178 		 * need to be revisited should any substantial logic be added
179 		 * above.
180 		 */
181 		AUDIT_SYSCALL_EXIT(error, td);
182 
183 #ifdef KDTRACE_HOOKS
184 		/* Give the syscall:::return DTrace probe a chance to fire. */
185 		if (__predict_false(se->sy_return != 0))
186 			(*systrace_probe_func)(sa, SYSTRACE_RETURN,
187 			    error ? -1 : td->td_retval[0]);
188 #endif
189 
190 		if (!sy_thr_static)
191 			syscall_thread_exit(td, se);
192 	} else {
193 		error = (se->sy_call)(td, sa->args);
194 		/* Save the latest error return value. */
195 		if (__predict_false((td->td_pflags & TDP_NERRNO) != 0))
196 			td->td_pflags &= ~TDP_NERRNO;
197 		else
198 			td->td_errno = error;
199 	}
200 
201  retval:
202 	KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code),
203 	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error,
204 	    "retval0:%#lx", td->td_retval[0], "retval1:%#lx",
205 	    td->td_retval[1]);
206 	if (__predict_false(traced)) {
207 		PROC_LOCK(p);
208 		td->td_dbgflags &= ~(TDB_SCE | TDB_BOUNDARY);
209 		PROC_UNLOCK(p);
210 	}
211 	(p->p_sysent->sv_set_syscall_retval)(td, error);
212 	if (error != 0 && (td->td_pflags2 & TDP2_UEXTERR) != 0)
213 		exterr_copyout(td);
214 }
215 
216 static inline void
syscallret(struct thread * td)217 syscallret(struct thread *td)
218 {
219 	struct proc *p;
220 	struct syscall_args *sa;
221 	ksiginfo_t ksi;
222 	int traced;
223 
224 	KASSERT(td->td_errno != ERELOOKUP,
225 	    ("ERELOOKUP not consumed syscall %d", td->td_sa.code));
226 
227 	p = td->td_proc;
228 	sa = &td->td_sa;
229 	if (__predict_false(td->td_errno == ENOTCAPABLE ||
230 	    td->td_errno == ECAPMODE)) {
231 		if ((trap_enotcap ||
232 		    (p->p_flag2 & P2_TRAPCAP) != 0) && IN_CAPABILITY_MODE(td)) {
233 			ksiginfo_init_trap(&ksi);
234 			ksi.ksi_signo = SIGTRAP;
235 			ksi.ksi_errno = td->td_errno;
236 			ksi.ksi_code = TRAP_CAP;
237 			ksi.ksi_info.si_syscall = sa->original_code;
238 			trapsignal(td, &ksi);
239 		}
240 	}
241 
242 	/*
243 	 * Handle reschedule and other end-of-syscall issues
244 	 */
245 	userret(td, td->td_frame);
246 
247 #ifdef KTRACE
248 	if (KTRPOINT(td, KTR_SYSRET)) {
249 		ktrsysret(sa->code, td->td_errno, td->td_retval[0]);
250 	}
251 #endif
252 
253 	traced = 0;
254 	if (__predict_false(p->p_flag & P_TRACED)) {
255 		traced = 1;
256 		PROC_LOCK(p);
257 		td->td_dbgflags |= TDB_SCX;
258 		PROC_UNLOCK(p);
259 	}
260 	if (__predict_false(traced ||
261 	    (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0)) {
262 		PROC_LOCK(p);
263 		/*
264 		 * Linux debuggers expect an additional stop for exec,
265 		 * between the usual syscall entry and exit.  Raise
266 		 * the exec event now and then clear TDB_EXEC so that
267 		 * the next stop is reported as a syscall exit by
268 		 * linux_ptrace_status().
269 		 *
270 		 * We are accessing p->p_pptr without any additional
271 		 * locks here: it cannot change while p is kept locked;
272 		 * while the debugger could in theory change its ABI
273 		 * while tracing another process, the outcome of such
274 		 * a race wouln't be deterministic anyway.
275 		 */
276 		if (traced && (td->td_dbgflags & TDB_EXEC) != 0 &&
277 		    SV_PROC_ABI(p->p_pptr) == SV_ABI_LINUX) {
278 			ptracestop(td, SIGTRAP, NULL);
279 			td->td_dbgflags &= ~TDB_EXEC;
280 		}
281 		/*
282 		 * If tracing the execed process, trap to the debugger
283 		 * so that breakpoints can be set before the program
284 		 * executes.  If debugger requested tracing of syscall
285 		 * returns, do it now too.
286 		 */
287 		if (traced &&
288 		    ((td->td_dbgflags & (TDB_FORK | TDB_EXEC)) != 0 ||
289 		    (p->p_ptevents & PTRACE_SCX) != 0)) {
290 			MPASS((td->td_dbgflags & TDB_BOUNDARY) == 0);
291 			td->td_dbgflags |= TDB_BOUNDARY;
292 			ptracestop(td, SIGTRAP, NULL);
293 		}
294 		td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK |
295 		    TDB_BOUNDARY);
296 		PROC_UNLOCK(p);
297 	}
298 }
299