2 * SPDX-License-Identifier: BSD-4-Clause
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>
9 * This code is derived from software contributed to Berkeley by
10 * the University of Utah, and William Jolitz.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
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.
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
41 #include "opt_capsicum.h"
42 #include "opt_ktrace.h"
43 #include <sys/capsicum.h>
45 #include <sys/vmmeter.h>
48 #include <sys/ktrace.h>
50 #include <security/audit/audit.h>
53 syscallenter(struct thread *td)
56 struct syscall_args *sa;
61 VM_CNT_INC(v_syscall);
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)) {
71 MPASS((td->td_dbgflags & TDB_BOUNDARY) == 0);
72 td->td_dbgflags &= ~TDB_USERWR;
74 td->td_dbgflags |= TDB_SCE;
77 error = (p->p_sysent->sv_fetch_syscall_args)(td);
80 if (KTRPOINT(td, KTR_SYSCALL))
81 ktrsyscall(sa->code, se->sy_narg, sa->args);
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]);
87 if (__predict_false(error != 0)) {
92 if (__predict_false(traced)) {
94 if (p->p_ptevents & PTRACE_SCE)
95 ptracestop((td), SIGTRAP, NULL);
98 if ((td->td_dbgflags & TDB_USERWR) != 0) {
100 * Reread syscall number and arguments if debugger
101 * modified registers or memory.
103 error = (p->p_sysent->sv_fetch_syscall_args)(td);
106 if (KTRPOINT(td, KTR_SYSCALL))
107 ktrsyscall(sa->code, se->sy_narg, sa->args);
110 td->td_errno = error;
116 #ifdef CAPABILITY_MODE
118 * In capability mode, we only allow access to system calls
119 * flagged with SYF_CAPENABLED.
121 if (__predict_false(IN_CAPABILITY_MODE(td) &&
122 (se->sy_flags & SYF_CAPENABLED) == 0)) {
123 td->td_errno = error = ECAPMODE;
129 * Fetch fast sigblock value at the time of syscall entry to
130 * handle sleepqueue primitives which might call cursig().
132 if (__predict_false(sigfastblock_fetch_always))
133 (void)sigfastblock_fetch(td);
135 /* Let system calls set td_errno directly. */
136 KASSERT((td->td_pflags & TDP_NERRNO) == 0,
137 ("%s: TDP_NERRNO set", __func__));
139 sy_thr_static = (se->sy_thrcnt & SY_THR_STATIC) != 0;
141 if (__predict_false(SYSTRACE_ENABLED() ||
142 AUDIT_SYSCALL_ENTER(sa->code, td) ||
144 if (!sy_thr_static) {
145 error = syscall_thread_enter(td, &se);
146 sy_thr_static = (se->sy_thrcnt & SY_THR_STATIC) != 0;
148 td->td_errno = error;
154 /* Give the syscall:::entry DTrace probe a chance to fire. */
155 if (__predict_false(se->sy_entry != 0))
156 (*systrace_probe_func)(sa, SYSTRACE_ENTRY, 0);
158 error = (se->sy_call)(td, sa->args);
159 /* Save the latest error return value. */
160 if (__predict_false((td->td_pflags & TDP_NERRNO) != 0))
161 td->td_pflags &= ~TDP_NERRNO;
163 td->td_errno = error;
166 * Note that some syscall implementations (e.g., sys_execve)
167 * will commit the audit record just before their final return.
168 * These were done under the assumption that nothing of interest
169 * would happen between their return and here, where we would
170 * normally commit the audit record. These assumptions will
171 * need to be revisited should any substantial logic be added
174 AUDIT_SYSCALL_EXIT(error, td);
177 /* Give the syscall:::return DTrace probe a chance to fire. */
178 if (__predict_false(se->sy_return != 0))
179 (*systrace_probe_func)(sa, SYSTRACE_RETURN,
180 error ? -1 : td->td_retval[0]);
184 syscall_thread_exit(td, se);
186 error = (se->sy_call)(td, sa->args);
187 /* Save the latest error return value. */
188 if (__predict_false((td->td_pflags & TDP_NERRNO) != 0))
189 td->td_pflags &= ~TDP_NERRNO;
191 td->td_errno = error;
195 KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code),
196 (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error,
197 "retval0:%#lx", td->td_retval[0], "retval1:%#lx",
199 if (__predict_false(traced)) {
201 td->td_dbgflags &= ~(TDB_SCE | TDB_BOUNDARY);
204 (p->p_sysent->sv_set_syscall_retval)(td, error);
208 syscallret(struct thread *td)
211 struct syscall_args *sa;
215 KASSERT(td->td_errno != ERELOOKUP,
216 ("ERELOOKUP not consumed syscall %d", td->td_sa.code));
220 if (__predict_false(td->td_errno == ENOTCAPABLE ||
221 td->td_errno == ECAPMODE)) {
223 (p->p_flag2 & P2_TRAPCAP) != 0) && IN_CAPABILITY_MODE(td)) {
224 ksiginfo_init_trap(&ksi);
225 ksi.ksi_signo = SIGTRAP;
226 ksi.ksi_errno = td->td_errno;
227 ksi.ksi_code = TRAP_CAP;
228 ksi.ksi_info.si_syscall = sa->original_code;
229 trapsignal(td, &ksi);
234 * Handle reschedule and other end-of-syscall issues
236 userret(td, td->td_frame);
239 if (KTRPOINT(td, KTR_SYSRET)) {
240 ktrsysret(sa->code, td->td_errno, td->td_retval[0]);
245 if (__predict_false(p->p_flag & P_TRACED)) {
248 td->td_dbgflags |= TDB_SCX;
251 if (__predict_false(traced ||
252 (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0)) {
255 * Linux debuggers expect an additional stop for exec,
256 * between the usual syscall entry and exit. Raise
257 * the exec event now and then clear TDB_EXEC so that
258 * the next stop is reported as a syscall exit by
259 * linux_ptrace_status().
261 * We are accessing p->p_pptr without any additional
262 * locks here: it cannot change while p is kept locked;
263 * while the debugger could in theory change its ABI
264 * while tracing another process, the outcome of such
265 * a race wouln't be deterministic anyway.
267 if (traced && (td->td_dbgflags & TDB_EXEC) != 0 &&
268 SV_PROC_ABI(p->p_pptr) == SV_ABI_LINUX) {
269 ptracestop(td, SIGTRAP, NULL);
270 td->td_dbgflags &= ~TDB_EXEC;
273 * If tracing the execed process, trap to the debugger
274 * so that breakpoints can be set before the program
275 * executes. If debugger requested tracing of syscall
276 * returns, do it now too.
279 ((td->td_dbgflags & (TDB_FORK | TDB_EXEC)) != 0 ||
280 (p->p_ptevents & PTRACE_SCX) != 0)) {
281 MPASS((td->td_dbgflags & TDB_BOUNDARY) == 0);
282 td->td_dbgflags |= TDB_BOUNDARY;
283 ptracestop(td, SIGTRAP, NULL);
285 td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK |