subr_syscall.c revision 218688
1/*- 2 * Copyright (C) 1994, David Greenman 3 * Copyright (c) 1990, 1993 4 * The Regents of the University of California. All rights reserved. 5 * Copyright (c) 2007 The FreeBSD Foundation 6 * 7 * This code is derived from software contributed to Berkeley by 8 * the University of Utah, and William Jolitz. 9 * 10 * Portions of this software were developed by A. Joseph Koshy under 11 * sponsorship from the FreeBSD Foundation and Google, Inc. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 3. All advertising materials mentioning features or use of this software 22 * must display the following acknowledgement: 23 * This product includes software developed by the University of 24 * California, Berkeley and its contributors. 25 * 4. Neither the name of the University nor the names of its contributors 26 * may be used to endorse or promote products derived from this software 27 * without specific prior written permission. 28 * 29 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 32 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 39 * SUCH DAMAGE. 40 * 41 * from: @(#)trap.c 7.4 (Berkeley) 5/13/91 42 */ 43 44#include <sys/cdefs.h> 45__FBSDID("$FreeBSD: head/sys/kern/subr_trap.c 218688 2011-02-14 20:49:37Z bz $"); 46 47#include "opt_ktrace.h" 48#include "opt_kdtrace.h" 49#include "opt_sched.h" 50 51#include <sys/param.h> 52#include <sys/bus.h> 53#include <sys/kernel.h> 54#include <sys/lock.h> 55#include <sys/mutex.h> 56#include <sys/pmckern.h> 57#include <sys/proc.h> 58#include <sys/ktr.h> 59#include <sys/pioctl.h> 60#include <sys/ptrace.h> 61#include <sys/resourcevar.h> 62#include <sys/sched.h> 63#include <sys/signalvar.h> 64#include <sys/syscall.h> 65#include <sys/syscallsubr.h> 66#include <sys/sysent.h> 67#include <sys/systm.h> 68#include <sys/vmmeter.h> 69#ifdef KTRACE 70#include <sys/uio.h> 71#include <sys/ktrace.h> 72#endif 73#include <security/audit/audit.h> 74 75#include <machine/cpu.h> 76 77#ifdef VIMAGE 78#include <net/vnet.h> 79#endif 80 81#ifdef XEN 82#include <vm/vm.h> 83#include <vm/vm_param.h> 84#include <vm/pmap.h> 85#endif 86 87#include <security/mac/mac_framework.h> 88 89/* 90 * Define the code needed before returning to user mode, for trap and 91 * syscall. 92 */ 93void 94userret(struct thread *td, struct trapframe *frame) 95{ 96 struct proc *p = td->td_proc; 97 98 CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid, 99 td->td_name); 100#if 0 101#ifdef DIAGNOSTIC 102 /* Check that we called signotify() enough. */ 103 PROC_LOCK(p); 104 thread_lock(td); 105 if (SIGPENDING(td) && ((td->td_flags & TDF_NEEDSIGCHK) == 0 || 106 (td->td_flags & TDF_ASTPENDING) == 0)) 107 printf("failed to set signal flags properly for ast()\n"); 108 thread_unlock(td); 109 PROC_UNLOCK(p); 110#endif 111#endif 112#ifdef KTRACE 113 KTRUSERRET(td); 114#endif 115 /* 116 * If this thread tickled GEOM, we need to wait for the giggling to 117 * stop before we return to userland 118 */ 119 if (td->td_pflags & TDP_GEOM) 120 g_waitidle(); 121 122 /* 123 * Charge system time if profiling. 124 */ 125 if (p->p_flag & P_PROFIL) 126 addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio); 127 /* 128 * Let the scheduler adjust our priority etc. 129 */ 130 sched_userret(td); 131 KASSERT(td->td_locks == 0, 132 ("userret: Returning with %d locks held.", td->td_locks)); 133#ifdef VIMAGE 134 /* Unfortunately td_vnet_lpush needs VNET_DEBUG. */ 135 VNET_ASSERT(curvnet == NULL, 136 ("%s: Returning on td %p (pid %d, %s) with vnet %p set in %s", 137 __func__, td, p->p_pid, td->td_name, curvnet, 138 (td->td_vnet_lpush != NULL) ? td->td_vnet_lpush : "N/A")); 139#endif 140#ifdef XEN 141 PT_UPDATES_FLUSH(); 142#endif 143} 144 145/* 146 * Process an asynchronous software trap. 147 * This is relatively easy. 148 * This function will return with preemption disabled. 149 */ 150void 151ast(struct trapframe *framep) 152{ 153 struct thread *td; 154 struct proc *p; 155 int flags; 156 int sig; 157 158 td = curthread; 159 p = td->td_proc; 160 161 CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid, 162 p->p_comm); 163 KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode")); 164 WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode"); 165 mtx_assert(&Giant, MA_NOTOWNED); 166 THREAD_LOCK_ASSERT(td, MA_NOTOWNED); 167 td->td_frame = framep; 168 td->td_pticks = 0; 169 170 /* 171 * This updates the td_flag's for the checks below in one 172 * "atomic" operation with turning off the astpending flag. 173 * If another AST is triggered while we are handling the 174 * AST's saved in flags, the astpending flag will be set and 175 * ast() will be called again. 176 */ 177 thread_lock(td); 178 flags = td->td_flags; 179 td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK | 180 TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND); 181 thread_unlock(td); 182 PCPU_INC(cnt.v_trap); 183 184 if (td->td_ucred != p->p_ucred) 185 cred_update_thread(td); 186 if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) { 187 addupc_task(td, td->td_profil_addr, td->td_profil_ticks); 188 td->td_profil_ticks = 0; 189 td->td_pflags &= ~TDP_OWEUPC; 190 } 191 if (flags & TDF_ALRMPEND) { 192 PROC_LOCK(p); 193 psignal(p, SIGVTALRM); 194 PROC_UNLOCK(p); 195 } 196 if (flags & TDF_PROFPEND) { 197 PROC_LOCK(p); 198 psignal(p, SIGPROF); 199 PROC_UNLOCK(p); 200 } 201#ifdef MAC 202 if (flags & TDF_MACPEND) 203 mac_thread_userret(td); 204#endif 205 if (flags & TDF_NEEDRESCHED) { 206#ifdef KTRACE 207 if (KTRPOINT(td, KTR_CSW)) 208 ktrcsw(1, 1); 209#endif 210 thread_lock(td); 211 sched_prio(td, td->td_user_pri); 212 mi_switch(SW_INVOL | SWT_NEEDRESCHED, NULL); 213 thread_unlock(td); 214#ifdef KTRACE 215 if (KTRPOINT(td, KTR_CSW)) 216 ktrcsw(0, 1); 217#endif 218 } 219 220 /* 221 * Check for signals. Unlocked reads of p_pendingcnt or 222 * p_siglist might cause process-directed signal to be handled 223 * later. 224 */ 225 if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 || 226 !SIGISEMPTY(p->p_siglist)) { 227 PROC_LOCK(p); 228 mtx_lock(&p->p_sigacts->ps_mtx); 229 while ((sig = cursig(td, SIG_STOP_ALLOWED)) != 0) 230 postsig(sig); 231 mtx_unlock(&p->p_sigacts->ps_mtx); 232 PROC_UNLOCK(p); 233 } 234 /* 235 * We need to check to see if we have to exit or wait due to a 236 * single threading requirement or some other STOP condition. 237 */ 238 if (flags & TDF_NEEDSUSPCHK) { 239 PROC_LOCK(p); 240 thread_suspend_check(0); 241 PROC_UNLOCK(p); 242 } 243 244 if (td->td_pflags & TDP_OLDMASK) { 245 td->td_pflags &= ~TDP_OLDMASK; 246 kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0); 247 } 248 249 userret(td, framep); 250 mtx_assert(&Giant, MA_NOTOWNED); 251} 252 253#ifdef HAVE_SYSCALL_ARGS_DEF 254const char * 255syscallname(struct proc *p, u_int code) 256{ 257 static const char unknown[] = "unknown"; 258 struct sysentvec *sv; 259 260 sv = p->p_sysent; 261 if (sv->sv_syscallnames == NULL || code >= sv->sv_size) 262 return (unknown); 263 return (sv->sv_syscallnames[code]); 264} 265 266int 267syscallenter(struct thread *td, struct syscall_args *sa) 268{ 269 struct proc *p; 270 int error, traced; 271 272 PCPU_INC(cnt.v_syscall); 273 p = td->td_proc; 274 275 td->td_pticks = 0; 276 if (td->td_ucred != p->p_ucred) 277 cred_update_thread(td); 278 if (p->p_flag & P_TRACED) { 279 traced = 1; 280 PROC_LOCK(p); 281 td->td_dbgflags &= ~TDB_USERWR; 282 td->td_dbgflags |= TDB_SCE; 283 PROC_UNLOCK(p); 284 } else 285 traced = 0; 286 error = (p->p_sysent->sv_fetch_syscall_args)(td, sa); 287#ifdef KTRACE 288 if (KTRPOINT(td, KTR_SYSCALL)) 289 ktrsyscall(sa->code, sa->narg, sa->args); 290#endif 291 292 CTR6(KTR_SYSC, 293"syscall: td=%p pid %d %s (%#lx, %#lx, %#lx)", 294 td, td->td_proc->p_pid, syscallname(p, sa->code), 295 sa->args[0], sa->args[1], sa->args[2]); 296 297 if (error == 0) { 298 STOPEVENT(p, S_SCE, sa->narg); 299 PTRACESTOP_SC(p, td, S_PT_SCE); 300 if (td->td_dbgflags & TDB_USERWR) { 301 /* 302 * Reread syscall number and arguments if 303 * debugger modified registers or memory. 304 */ 305 error = (p->p_sysent->sv_fetch_syscall_args)(td, sa); 306#ifdef KTRACE 307 if (KTRPOINT(td, KTR_SYSCALL)) 308 ktrsyscall(sa->code, sa->narg, sa->args); 309#endif 310 if (error != 0) 311 goto retval; 312 } 313 error = syscall_thread_enter(td, sa->callp); 314 if (error != 0) 315 goto retval; 316 317#ifdef KDTRACE_HOOKS 318 /* 319 * If the systrace module has registered it's probe 320 * callback and if there is a probe active for the 321 * syscall 'entry', process the probe. 322 */ 323 if (systrace_probe_func != NULL && sa->callp->sy_entry != 0) 324 (*systrace_probe_func)(sa->callp->sy_entry, sa->code, 325 sa->callp, sa->args, 0); 326#endif 327 328 AUDIT_SYSCALL_ENTER(sa->code, td); 329 error = (sa->callp->sy_call)(td, sa->args); 330 AUDIT_SYSCALL_EXIT(error, td); 331 332 /* Save the latest error return value. */ 333 td->td_errno = error; 334 335#ifdef KDTRACE_HOOKS 336 /* 337 * If the systrace module has registered it's probe 338 * callback and if there is a probe active for the 339 * syscall 'return', process the probe. 340 */ 341 if (systrace_probe_func != NULL && sa->callp->sy_return != 0) 342 (*systrace_probe_func)(sa->callp->sy_return, sa->code, 343 sa->callp, NULL, (error) ? -1 : td->td_retval[0]); 344#endif 345 syscall_thread_exit(td, sa->callp); 346 CTR4(KTR_SYSC, "syscall: p=%p error=%d return %#lx %#lx", 347 p, error, td->td_retval[0], td->td_retval[1]); 348 } 349 retval: 350 if (traced) { 351 PROC_LOCK(p); 352 td->td_dbgflags &= ~TDB_SCE; 353 PROC_UNLOCK(p); 354 } 355 (p->p_sysent->sv_set_syscall_retval)(td, error); 356 return (error); 357} 358 359void 360syscallret(struct thread *td, int error, struct syscall_args *sa __unused) 361{ 362 struct proc *p; 363 int traced; 364 365 p = td->td_proc; 366 367 /* 368 * Check for misbehavior. 369 */ 370 WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning", 371 syscallname(p, sa->code)); 372 KASSERT(td->td_critnest == 0, 373 ("System call %s returning in a critical section", 374 syscallname(p, sa->code))); 375 KASSERT(td->td_locks == 0, 376 ("System call %s returning with %d locks held", 377 syscallname(p, sa->code), td->td_locks)); 378 379 /* 380 * Handle reschedule and other end-of-syscall issues 381 */ 382 userret(td, td->td_frame); 383 384 CTR4(KTR_SYSC, "syscall %s exit thread %p pid %d proc %s", 385 syscallname(p, sa->code), td, td->td_proc->p_pid, td->td_name); 386 387#ifdef KTRACE 388 if (KTRPOINT(td, KTR_SYSRET)) 389 ktrsysret(sa->code, error, td->td_retval[0]); 390#endif 391 392 if (p->p_flag & P_TRACED) { 393 traced = 1; 394 PROC_LOCK(p); 395 td->td_dbgflags |= TDB_SCX; 396 PROC_UNLOCK(p); 397 } else 398 traced = 0; 399 /* 400 * This works because errno is findable through the 401 * register set. If we ever support an emulation where this 402 * is not the case, this code will need to be revisited. 403 */ 404 STOPEVENT(p, S_SCX, sa->code); 405 PTRACESTOP_SC(p, td, S_PT_SCX); 406 if (traced || (td->td_dbgflags & (TDB_EXEC | TDB_FORK)) != 0) { 407 PROC_LOCK(p); 408 td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC | TDB_FORK); 409 PROC_UNLOCK(p); 410 } 411} 412#endif /* HAVE_SYSCALL_ARGS_DEF */ 413