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 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 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(AUDIT_SYSCALL_ENABLED() || 145 SYSTRACE_ENABLED() || !sy_thr_static)) { 146 if (!sy_thr_static) { 147 error = syscall_thread_enter(td, &se); 148 sy_thr_static = (se->sy_thrcnt & SY_THR_STATIC) != 0; 149 if (error != 0) { 150 td->td_errno = error; 151 goto retval; 152 } 153 } 154 155 #ifdef KDTRACE_HOOKS 156 /* Give the syscall:::entry DTrace probe a chance to fire. */ 157 if (__predict_false(se->sy_entry != 0)) 158 (*systrace_probe_func)(sa, SYSTRACE_ENTRY, 0); 159 #endif 160 161 AUDIT_SYSCALL_ENTER(sa->code, td); 162 163 error = (se->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_pflags &= ~TDP_NERRNO; 167 else 168 td->td_errno = error; 169 170 /* 171 * Note that some syscall implementations (e.g., sys_execve) 172 * will commit the audit record just before their final return. 173 * These were done under the assumption that nothing of interest 174 * would happen between their return and here, where we would 175 * normally commit the audit record. These assumptions will 176 * need to be revisited should any substantial logic be added 177 * above. 178 */ 179 AUDIT_SYSCALL_EXIT(error, td); 180 181 #ifdef KDTRACE_HOOKS 182 /* Give the syscall:::return DTrace probe a chance to fire. */ 183 if (__predict_false(se->sy_return != 0)) 184 (*systrace_probe_func)(sa, SYSTRACE_RETURN, 185 error ? -1 : td->td_retval[0]); 186 #endif 187 188 if (!sy_thr_static) 189 syscall_thread_exit(td, se); 190 } else { 191 error = (se->sy_call)(td, sa->args); 192 /* Save the latest error return value. */ 193 if (__predict_false((td->td_pflags & TDP_NERRNO) != 0)) 194 td->td_pflags &= ~TDP_NERRNO; 195 else 196 td->td_errno = error; 197 } 198 199 retval: 200 KTR_STOP4(KTR_SYSC, "syscall", syscallname(p, sa->code), 201 (uintptr_t)td, "pid:%d", td->td_proc->p_pid, "error:%d", error, 202 "retval0:%#lx", td->td_retval[0], "retval1:%#lx", 203 td->td_retval[1]); 204 if (__predict_false(traced)) { 205 PROC_LOCK(p); 206 td->td_dbgflags &= ~(TDB_SCE | TDB_BOUNDARY); 207 PROC_UNLOCK(p); 208 } 209 (p->p_sysent->sv_set_syscall_retval)(td, error); 210 } 211 212 static inline void 213 syscallret(struct thread *td) 214 { 215 struct proc *p; 216 struct syscall_args *sa; 217 ksiginfo_t ksi; 218 int traced; 219 220 KASSERT(td->td_errno != ERELOOKUP, 221 ("ERELOOKUP not consumed syscall %d", td->td_sa.code)); 222 223 p = td->td_proc; 224 sa = &td->td_sa; 225 if (__predict_false(td->td_errno == ENOTCAPABLE || 226 td->td_errno == ECAPMODE)) { 227 if ((trap_enotcap || 228 (p->p_flag2 & P2_TRAPCAP) != 0) && IN_CAPABILITY_MODE(td)) { 229 ksiginfo_init_trap(&ksi); 230 ksi.ksi_signo = SIGTRAP; 231 ksi.ksi_errno = td->td_errno; 232 ksi.ksi_code = TRAP_CAP; 233 ksi.ksi_info.si_syscall = sa->original_code; 234 trapsignal(td, &ksi); 235 } 236 } 237 238 /* 239 * Handle reschedule and other end-of-syscall issues 240 */ 241 userret(td, td->td_frame); 242 243 #ifdef KTRACE 244 if (KTRPOINT(td, KTR_SYSRET)) { 245 ktrsysret(sa->code, td->td_errno, td->td_retval[0]); 246 } 247 #endif 248 249 traced = 0; 250 if (__predict_false(p->p_flag & P_TRACED)) { 251 traced = 1; 252 PROC_LOCK(p); 253 td->td_dbgflags |= TDB_SCX; 254 PROC_UNLOCK(p); 255 } 256 if (__predict_false(traced || 257 (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0)) { 258 PROC_LOCK(p); 259 /* 260 * Linux debuggers expect an additional stop for exec, 261 * between the usual syscall entry and exit. Raise 262 * the exec event now and then clear TDB_EXEC so that 263 * the next stop is reported as a syscall exit by 264 * linux_ptrace_status(). 265 * 266 * We are accessing p->p_pptr without any additional 267 * locks here: it cannot change while p is kept locked; 268 * while the debugger could in theory change its ABI 269 * while tracing another process, the outcome of such 270 * a race wouln't be deterministic anyway. 271 */ 272 if (traced && (td->td_dbgflags & TDB_EXEC) != 0 && 273 SV_PROC_ABI(p->p_pptr) == SV_ABI_LINUX) { 274 ptracestop(td, SIGTRAP, NULL); 275 td->td_dbgflags &= ~TDB_EXEC; 276 } 277 /* 278 * If tracing the execed process, trap to the debugger 279 * so that breakpoints can be set before the program 280 * executes. If debugger requested tracing of syscall 281 * returns, do it now too. 282 */ 283 if (traced && 284 ((td->td_dbgflags & (TDB_FORK | TDB_EXEC)) != 0 || 285 (p->p_ptevents & PTRACE_SCX) != 0)) { 286 MPASS((td->td_dbgflags & TDB_BOUNDARY) == 0); 287 td->td_dbgflags |= TDB_BOUNDARY; 288 ptracestop(td, SIGTRAP, NULL); 289 } 290 td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK | 291 TDB_BOUNDARY); 292 PROC_UNLOCK(p); 293 } 294 } 295