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
52static inline void
53syscallenter(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	error = (p->p_sysent->sv_fetch_syscall_args)(td);
78	se = sa->callp;
79#ifdef KTRACE
80	if (KTRPOINT(td, KTR_SYSCALL))
81		ktrsyscall(sa->code, se->sy_narg, sa->args);
82#endif
83	KTR_START4(KTR_SYSC, "syscall", syscallname(p, sa->code),
84	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "arg0:%p", sa->args[0],
85	    "arg1:%p", sa->args[1], "arg2:%p", sa->args[2]);
86
87	if (__predict_false(error != 0)) {
88		td->td_errno = error;
89		goto retval;
90	}
91
92	if (__predict_false(traced)) {
93		PROC_LOCK(p);
94		if (p->p_ptevents & PTRACE_SCE)
95			ptracestop((td), SIGTRAP, NULL);
96		PROC_UNLOCK(p);
97
98		if ((td->td_dbgflags & TDB_USERWR) != 0) {
99			/*
100			 * Reread syscall number and arguments if debugger
101			 * modified registers or memory.
102			 */
103			error = (p->p_sysent->sv_fetch_syscall_args)(td);
104			se = sa->callp;
105#ifdef KTRACE
106			if (KTRPOINT(td, KTR_SYSCALL))
107				ktrsyscall(sa->code, se->sy_narg, sa->args);
108#endif
109			if (error != 0) {
110				td->td_errno = error;
111				goto retval;
112			}
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 ((se->sy_flags & SYF_CAPENABLED) == 0) {
122		if (CAP_TRACING(td))
123			ktrcapfail(CAPFAIL_SYSCALL, NULL);
124		if (IN_CAPABILITY_MODE(td)) {
125			td->td_errno = error = ECAPMODE;
126			goto retval;
127		}
128	}
129#endif
130
131	/*
132	 * Fetch fast sigblock value at the time of syscall entry to
133	 * handle sleepqueue primitives which might call cursig().
134	 */
135	if (__predict_false(sigfastblock_fetch_always))
136		(void)sigfastblock_fetch(td);
137
138	/* Let system calls set td_errno directly. */
139	KASSERT((td->td_pflags & TDP_NERRNO) == 0,
140	    ("%s: TDP_NERRNO set", __func__));
141
142	sy_thr_static = (se->sy_thrcnt & SY_THR_STATIC) != 0;
143
144	if (__predict_false(SYSTRACE_ENABLED() ||
145	    AUDIT_SYSCALL_ENTER(sa->code, td) ||
146	    !sy_thr_static)) {
147		if (!sy_thr_static) {
148			error = syscall_thread_enter(td, &se);
149			sy_thr_static = (se->sy_thrcnt & SY_THR_STATIC) != 0;
150			if (error != 0) {
151				td->td_errno = error;
152				goto retval;
153			}
154		}
155
156#ifdef KDTRACE_HOOKS
157		/* Give the syscall:::entry DTrace probe a chance to fire. */
158		if (__predict_false(se->sy_entry != 0))
159			(*systrace_probe_func)(sa, SYSTRACE_ENTRY, 0);
160#endif
161		error = (se->sy_call)(td, sa->args);
162		/* Save the latest error return value. */
163		if (__predict_false((td->td_pflags & TDP_NERRNO) != 0))
164			td->td_pflags &= ~TDP_NERRNO;
165		else
166			td->td_errno = error;
167
168		/*
169		 * Note that some syscall implementations (e.g., sys_execve)
170		 * will commit the audit record just before their final return.
171		 * These were done under the assumption that nothing of interest
172		 * would happen between their return and here, where we would
173		 * normally commit the audit record.  These assumptions will
174		 * need to be revisited should any substantial logic be added
175		 * above.
176		 */
177		AUDIT_SYSCALL_EXIT(error, td);
178
179#ifdef KDTRACE_HOOKS
180		/* Give the syscall:::return DTrace probe a chance to fire. */
181		if (__predict_false(se->sy_return != 0))
182			(*systrace_probe_func)(sa, SYSTRACE_RETURN,
183			    error ? -1 : td->td_retval[0]);
184#endif
185
186		if (!sy_thr_static)
187			syscall_thread_exit(td, se);
188	} else {
189		error = (se->sy_call)(td, sa->args);
190		/* Save the latest error return value. */
191		if (__predict_false((td->td_pflags & TDP_NERRNO) != 0))
192			td->td_pflags &= ~TDP_NERRNO;
193		else
194			td->td_errno = error;
195	}
196
197 retval:
198	KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code),
199	    (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error,
200	    "retval0:%#lx", td->td_retval[0], "retval1:%#lx",
201	    td->td_retval[1]);
202	if (__predict_false(traced)) {
203		PROC_LOCK(p);
204		td->td_dbgflags &= ~(TDB_SCE | TDB_BOUNDARY);
205		PROC_UNLOCK(p);
206	}
207	(p->p_sysent->sv_set_syscall_retval)(td, error);
208}
209
210static inline void
211syscallret(struct thread *td)
212{
213	struct proc *p;
214	struct syscall_args *sa;
215	ksiginfo_t ksi;
216	int traced;
217
218	KASSERT(td->td_errno != ERELOOKUP,
219	    ("ERELOOKUP not consumed syscall %d", td->td_sa.code));
220
221	p = td->td_proc;
222	sa = &td->td_sa;
223	if (__predict_false(td->td_errno == ENOTCAPABLE ||
224	    td->td_errno == ECAPMODE)) {
225		if ((trap_enotcap ||
226		    (p->p_flag2 & P2_TRAPCAP) != 0) && IN_CAPABILITY_MODE(td)) {
227			ksiginfo_init_trap(&ksi);
228			ksi.ksi_signo = SIGTRAP;
229			ksi.ksi_errno = td->td_errno;
230			ksi.ksi_code = TRAP_CAP;
231			ksi.ksi_info.si_syscall = sa->original_code;
232			trapsignal(td, &ksi);
233		}
234	}
235
236	/*
237	 * Handle reschedule and other end-of-syscall issues
238	 */
239	userret(td, td->td_frame);
240
241#ifdef KTRACE
242	if (KTRPOINT(td, KTR_SYSRET)) {
243		ktrsysret(sa->code, td->td_errno, td->td_retval[0]);
244	}
245#endif
246
247	traced = 0;
248	if (__predict_false(p->p_flag & P_TRACED)) {
249		traced = 1;
250		PROC_LOCK(p);
251		td->td_dbgflags |= TDB_SCX;
252		PROC_UNLOCK(p);
253	}
254	if (__predict_false(traced ||
255	    (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0)) {
256		PROC_LOCK(p);
257		/*
258		 * Linux debuggers expect an additional stop for exec,
259		 * between the usual syscall entry and exit.  Raise
260		 * the exec event now and then clear TDB_EXEC so that
261		 * the next stop is reported as a syscall exit by
262		 * linux_ptrace_status().
263		 *
264		 * We are accessing p->p_pptr without any additional
265		 * locks here: it cannot change while p is kept locked;
266		 * while the debugger could in theory change its ABI
267		 * while tracing another process, the outcome of such
268		 * a race wouln't be deterministic anyway.
269		 */
270		if (traced && (td->td_dbgflags & TDB_EXEC) != 0 &&
271		    SV_PROC_ABI(p->p_pptr) == SV_ABI_LINUX) {
272			ptracestop(td, SIGTRAP, NULL);
273			td->td_dbgflags &= ~TDB_EXEC;
274		}
275		/*
276		 * If tracing the execed process, trap to the debugger
277		 * so that breakpoints can be set before the program
278		 * executes.  If debugger requested tracing of syscall
279		 * returns, do it now too.
280		 */
281		if (traced &&
282		    ((td->td_dbgflags & (TDB_FORK | TDB_EXEC)) != 0 ||
283		    (p->p_ptevents & PTRACE_SCX) != 0)) {
284			MPASS((td->td_dbgflags & TDB_BOUNDARY) == 0);
285			td->td_dbgflags |= TDB_BOUNDARY;
286			ptracestop(td, SIGTRAP, NULL);
287		}
288		td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK |
289		    TDB_BOUNDARY);
290		PROC_UNLOCK(p);
291	}
292}
293