1 /* $OpenBSD: kern_exit.c,v 1.107 2012/02/20 22:23:39 guenther Exp $ */ 2 /* $NetBSD: kern_exit.c,v 1.39 1996/04/22 01:38:25 christos Exp $ */ 3 4 /* 5 * Copyright (c) 1982, 1986, 1989, 1991, 1993 6 * The Regents of the University of California. All rights reserved. 7 * (c) UNIX System Laboratories, Inc. 8 * All or some portions of this file are derived from material licensed 9 * to the University of California by American Telephone and Telegraph 10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 11 * the permission of UNIX System Laboratories, 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. Neither the name of the University nor the names of its contributors 22 * may be used to endorse or promote products derived from this software 23 * without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35 * SUCH DAMAGE. 36 * 37 * @(#)kern_exit.c 8.7 (Berkeley) 2/12/94 38 */ 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/ioctl.h> 43 #include <sys/proc.h> 44 #include <sys/tty.h> 45 #include <sys/time.h> 46 #include <sys/resource.h> 47 #include <sys/kernel.h> 48 #include <sys/buf.h> 49 #include <sys/wait.h> 50 #include <sys/file.h> 51 #include <sys/vnode.h> 52 #include <sys/syslog.h> 53 #include <sys/malloc.h> 54 #include <sys/resourcevar.h> 55 #include <sys/ptrace.h> 56 #include <sys/acct.h> 57 #include <sys/filedesc.h> 58 #include <sys/signalvar.h> 59 #include <sys/sched.h> 60 #include <sys/ktrace.h> 61 #include <sys/pool.h> 62 #include <sys/mutex.h> 63 #ifdef SYSVSEM 64 #include <sys/sem.h> 65 #endif 66 67 #include "systrace.h" 68 #include <dev/systrace.h> 69 70 #include <sys/mount.h> 71 #include <sys/syscallargs.h> 72 73 #include <machine/cpu.h> 74 75 #include <uvm/uvm_extern.h> 76 77 /* 78 * exit -- 79 * Death of process. 80 */ 81 int 82 sys_exit(struct proc *p, void *v, register_t *retval) 83 { 84 struct sys_exit_args /* { 85 syscallarg(int) rval; 86 } */ *uap = v; 87 88 exit1(p, W_EXITCODE(SCARG(uap, rval), 0), EXIT_NORMAL); 89 /* NOTREACHED */ 90 return (0); 91 } 92 93 int 94 sys___threxit(struct proc *p, void *v, register_t *retval) 95 { 96 struct sys___threxit_args /* { 97 syscallarg(pid_t *) notdead; 98 } */ *uap = v; 99 100 if (!rthreads_enabled) 101 return (EINVAL); 102 103 if (SCARG(uap, notdead) != NULL) { 104 pid_t zero = 0; 105 if (copyout(&zero, SCARG(uap, notdead), sizeof(zero))) { 106 psignal(p, SIGSEGV); 107 } 108 } 109 exit1(p, 0, EXIT_THREAD); 110 111 return (0); 112 } 113 114 /* 115 * Exit: deallocate address space and other resources, change proc state 116 * to zombie, and unlink proc from allproc and parent's lists. Save exit 117 * status and rusage for wait(). Check for child processes and orphan them. 118 */ 119 void 120 exit1(struct proc *p, int rv, int flags) 121 { 122 struct process *pr, *qr, *nqr; 123 124 if (p->p_pid == 1) 125 panic("init died (signal %d, exit %d)", 126 WTERMSIG(rv), WEXITSTATUS(rv)); 127 128 atomic_setbits_int(&p->p_flag, P_WEXIT); 129 130 pr = p->p_p; 131 132 /* single-threaded? */ 133 if (TAILQ_FIRST(&pr->ps_threads) == p && 134 TAILQ_NEXT(p, p_thr_link) == NULL) 135 flags = EXIT_NORMAL; 136 else { 137 /* nope, multi-threaded */ 138 if (flags == EXIT_NORMAL) 139 single_thread_set(p, SINGLE_EXIT, 0); 140 } 141 142 if (flags == EXIT_NORMAL) { 143 pr->ps_mainproc->p_xstat = rv; 144 145 /* 146 * If parent is waiting for us to exit or exec, PS_PPWAIT 147 * is set; we wake up the parent early to avoid deadlock. 148 */ 149 if (pr->ps_flags & PS_PPWAIT) { 150 atomic_clearbits_int(&pr->ps_flags, PS_PPWAIT); 151 atomic_clearbits_int(&pr->ps_pptr->ps_flags, 152 PS_ISPWAIT); 153 wakeup(pr->ps_pptr); 154 } 155 } 156 157 /* unlink ourselves from the active threads */ 158 TAILQ_REMOVE(&pr->ps_threads, p, p_thr_link); 159 if ((p->p_flag & P_THREAD) == 0) { 160 /* main thread gotta wait because it has the pid, et al */ 161 while (! TAILQ_EMPTY(&pr->ps_threads)) 162 tsleep(&pr->ps_threads, PUSER, "thrdeath", 0); 163 } else if (TAILQ_EMPTY(&pr->ps_threads)) 164 wakeup(&pr->ps_threads); 165 166 if (p->p_flag & P_PROFIL) 167 stopprofclock(p); 168 p->p_ru = pool_get(&rusage_pool, PR_WAITOK); 169 p->p_siglist = 0; 170 timeout_del(&p->p_realit_to); 171 timeout_del(&p->p_stats->p_virt_to); 172 timeout_del(&p->p_stats->p_prof_to); 173 174 /* 175 * Close open files and release open-file table. 176 */ 177 fdfree(p); 178 179 if ((p->p_flag & P_THREAD) == 0) { 180 #ifdef SYSVSEM 181 semexit(pr); 182 #endif 183 if (SESS_LEADER(pr)) { 184 struct session *sp = pr->ps_session; 185 186 if (sp->s_ttyvp) { 187 /* 188 * Controlling process. 189 * Signal foreground pgrp, 190 * drain controlling terminal 191 * and revoke access to controlling terminal. 192 */ 193 if (sp->s_ttyp->t_session == sp) { 194 if (sp->s_ttyp->t_pgrp) 195 pgsignal(sp->s_ttyp->t_pgrp, 196 SIGHUP, 1); 197 (void) ttywait(sp->s_ttyp); 198 /* 199 * The tty could have been revoked 200 * if we blocked. 201 */ 202 if (sp->s_ttyvp) 203 VOP_REVOKE(sp->s_ttyvp, 204 REVOKEALL); 205 } 206 if (sp->s_ttyvp) 207 vrele(sp->s_ttyvp); 208 sp->s_ttyvp = NULL; 209 /* 210 * s_ttyp is not zero'd; we use this to 211 * indicate that the session once had a 212 * controlling terminal. (for logging and 213 * informational purposes) 214 */ 215 } 216 sp->s_leader = NULL; 217 } 218 fixjobc(pr, pr->ps_pgrp, 0); 219 220 #ifdef ACCOUNTING 221 (void)acct_process(p); 222 #endif 223 224 #ifdef KTRACE 225 /* release trace file */ 226 if (pr->ps_tracevp) 227 ktrcleartrace(pr); 228 #endif 229 } 230 231 #if NSYSTRACE > 0 232 if (ISSET(p->p_flag, P_SYSTRACE)) 233 systrace_exit(p); 234 #endif 235 /* 236 * Remove proc from pidhash chain so looking it up won't 237 * work. Move it from allproc to zombproc, but do not yet 238 * wake up the reaper. We will put the proc on the 239 * deadproc list later (using the p_hash member), and 240 * wake up the reaper when we do. 241 */ 242 /* 243 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP! 244 */ 245 p->p_stat = SDEAD; 246 247 LIST_REMOVE(p, p_hash); 248 LIST_REMOVE(p, p_list); 249 LIST_INSERT_HEAD(&zombproc, p, p_list); 250 251 /* 252 * Give orphaned children to init(8). 253 */ 254 if ((p->p_flag & P_THREAD) == 0) { 255 qr = LIST_FIRST(&pr->ps_children); 256 if (qr) /* only need this if any child is S_ZOMB */ 257 wakeup(initproc->p_p); 258 for (; qr != 0; qr = nqr) { 259 nqr = LIST_NEXT(qr, ps_sibling); 260 proc_reparent(qr, initproc->p_p); 261 /* 262 * Traced processes are killed 263 * since their existence means someone is screwing up. 264 */ 265 if (qr->ps_flags & PS_TRACED) { 266 atomic_clearbits_int(&qr->ps_flags, PS_TRACED); 267 prsignal(qr, SIGKILL); 268 } 269 } 270 } 271 272 273 /* 274 * Save exit status and final rusage info, adding in child rusage 275 * info and self times. 276 */ 277 *p->p_ru = p->p_stats->p_ru; 278 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 279 ruadd(p->p_ru, &p->p_stats->p_cru); 280 281 /* 282 * clear %cpu usage during swap 283 */ 284 p->p_pctcpu = 0; 285 286 if ((p->p_flag & P_THREAD) == 0) { 287 /* notify interested parties of our demise and clean up */ 288 knote_processexit(pr); 289 290 /* 291 * Notify parent that we're gone. If we have P_NOZOMBIE 292 * or parent has the SAS_NOCLDWAIT flag set, notify process 1 293 * instead (and hope it will handle this situation). 294 */ 295 if ((p->p_flag & P_NOZOMBIE) || 296 (pr->ps_pptr->ps_mainproc->p_sigacts->ps_flags & 297 SAS_NOCLDWAIT)) { 298 struct process *ppr = pr->ps_pptr; 299 proc_reparent(pr, initproc->p_p); 300 /* 301 * If this was the last child of our parent, notify 302 * parent, so in case he was wait(2)ing, he will 303 * continue. 304 */ 305 if (LIST_EMPTY(&ppr->ps_children)) 306 wakeup(ppr); 307 } 308 } 309 310 /* 311 * Release the process's signal state. 312 */ 313 sigactsfree(p); 314 315 /* 316 * Other substructures are freed from reaper and wait(). 317 */ 318 319 /* 320 * If emulation has process exit hook, call it now. 321 */ 322 if (p->p_emul->e_proc_exit) 323 (*p->p_emul->e_proc_exit)(p); 324 325 /* 326 * Finally, call machine-dependent code to switch to a new 327 * context (possibly the idle context). Once we are no longer 328 * using the dead process's vmspace and stack, exit2() will be 329 * called to schedule those resources to be released by the 330 * reaper thread. 331 * 332 * Note that cpu_exit() will end with a call equivalent to 333 * cpu_switch(), finishing our execution (pun intended). 334 */ 335 uvmexp.swtch++; 336 cpu_exit(p); 337 panic("cpu_exit returned"); 338 } 339 340 /* 341 * Locking of this proclist is special; it's accessed in a 342 * critical section of process exit, and thus locking it can't 343 * modify interrupt state. We use a simple spin lock for this 344 * proclist. Processes on this proclist are also on zombproc; 345 * we use the p_hash member to linkup to deadproc. 346 */ 347 struct mutex deadproc_mutex = MUTEX_INITIALIZER(IPL_NONE); 348 struct proclist deadproc = LIST_HEAD_INITIALIZER(deadproc); 349 350 /* 351 * We are called from cpu_exit() once it is safe to schedule the 352 * dead process's resources to be freed. 353 * 354 * NOTE: One must be careful with locking in this routine. It's 355 * called from a critical section in machine-dependent code, so 356 * we should refrain from changing any interrupt state. 357 * 358 * We lock the deadproc list, place the proc on that list (using 359 * the p_hash member), and wake up the reaper. 360 */ 361 void 362 exit2(struct proc *p) 363 { 364 mtx_enter(&deadproc_mutex); 365 LIST_INSERT_HEAD(&deadproc, p, p_hash); 366 mtx_leave(&deadproc_mutex); 367 368 wakeup(&deadproc); 369 } 370 371 /* 372 * Process reaper. This is run by a kernel thread to free the resources 373 * of a dead process. Once the resources are free, the process becomes 374 * a zombie, and the parent is allowed to read the undead's status. 375 */ 376 void 377 reaper(void) 378 { 379 struct proc *p; 380 381 KERNEL_UNLOCK(); 382 383 SCHED_ASSERT_UNLOCKED(); 384 385 for (;;) { 386 mtx_enter(&deadproc_mutex); 387 while ((p = LIST_FIRST(&deadproc)) == NULL) 388 msleep(&deadproc, &deadproc_mutex, PVM, "reaper", 0); 389 390 /* Remove us from the deadproc list. */ 391 LIST_REMOVE(p, p_hash); 392 mtx_leave(&deadproc_mutex); 393 394 KERNEL_LOCK(); 395 396 /* 397 * Free the VM resources we're still holding on to. 398 * We must do this from a valid thread because doing 399 * so may block. 400 */ 401 uvm_exit(p); 402 403 /* Process is now a true zombie. */ 404 if ((p->p_flag & P_NOZOMBIE) == 0) { 405 p->p_stat = SZOMB; 406 407 if (P_EXITSIG(p) != 0) 408 prsignal(p->p_p->ps_pptr, P_EXITSIG(p)); 409 /* Wake up the parent so it can get exit status. */ 410 wakeup(p->p_p->ps_pptr); 411 } else { 412 /* Noone will wait for us. Just zap the process now */ 413 proc_zap(p); 414 } 415 416 KERNEL_UNLOCK(); 417 } 418 } 419 420 int 421 sys_wait4(struct proc *q, void *v, register_t *retval) 422 { 423 struct sys_wait4_args /* { 424 syscallarg(pid_t) pid; 425 syscallarg(int *) status; 426 syscallarg(int) options; 427 syscallarg(struct rusage *) rusage; 428 } */ *uap = v; 429 int nfound; 430 struct process *pr; 431 struct proc *p; 432 int status, error; 433 434 if (SCARG(uap, pid) == 0) 435 SCARG(uap, pid) = -q->p_p->ps_pgid; 436 if (SCARG(uap, options) &~ (WUNTRACED|WNOHANG|WALTSIG|WCONTINUED)) 437 return (EINVAL); 438 439 loop: 440 nfound = 0; 441 LIST_FOREACH(pr, &q->p_p->ps_children, ps_sibling) { 442 p = pr->ps_mainproc; 443 if ((p->p_flag & P_NOZOMBIE) || 444 (SCARG(uap, pid) != WAIT_ANY && 445 p->p_pid != SCARG(uap, pid) && 446 pr->ps_pgid != -SCARG(uap, pid))) 447 continue; 448 449 /* 450 * Wait for processes with p_exitsig != SIGCHLD processes only 451 * if WALTSIG is set; wait for processes with pexitsig == 452 * SIGCHLD only if WALTSIG is clear. 453 */ 454 if ((SCARG(uap, options) & WALTSIG) ? 455 (p->p_exitsig == SIGCHLD) : (P_EXITSIG(p) != SIGCHLD)) 456 continue; 457 458 nfound++; 459 if (p->p_stat == SZOMB) { 460 retval[0] = p->p_pid; 461 462 if (SCARG(uap, status)) { 463 status = p->p_xstat; /* convert to int */ 464 error = copyout(&status, 465 SCARG(uap, status), sizeof(status)); 466 if (error) 467 return (error); 468 } 469 if (SCARG(uap, rusage) && 470 (error = copyout(p->p_ru, 471 SCARG(uap, rusage), sizeof(struct rusage)))) 472 return (error); 473 proc_finish_wait(q, p); 474 return (0); 475 } 476 if (p->p_stat == SSTOP && 477 (p->p_flag & (P_WAITED|P_SUSPSINGLE)) == 0 && 478 (pr->ps_flags & PS_TRACED || 479 SCARG(uap, options) & WUNTRACED)) { 480 atomic_setbits_int(&p->p_flag, P_WAITED); 481 retval[0] = p->p_pid; 482 483 if (SCARG(uap, status)) { 484 status = W_STOPCODE(p->p_xstat); 485 error = copyout(&status, SCARG(uap, status), 486 sizeof(status)); 487 } else 488 error = 0; 489 return (error); 490 } 491 if ((SCARG(uap, options) & WCONTINUED) && (p->p_flag & P_CONTINUED)) { 492 atomic_clearbits_int(&p->p_flag, P_CONTINUED); 493 retval[0] = p->p_pid; 494 495 if (SCARG(uap, status)) { 496 status = _WCONTINUED; 497 error = copyout(&status, SCARG(uap, status), 498 sizeof(status)); 499 } else 500 error = 0; 501 return (error); 502 } 503 } 504 if (nfound == 0) 505 return (ECHILD); 506 if (SCARG(uap, options) & WNOHANG) { 507 retval[0] = 0; 508 return (0); 509 } 510 if ((error = tsleep(q->p_p, PWAIT | PCATCH, "wait", 0)) != 0) 511 return (error); 512 goto loop; 513 } 514 515 void 516 proc_finish_wait(struct proc *waiter, struct proc *p) 517 { 518 struct process *pr, *tr; 519 520 /* 521 * If we got the child via a ptrace 'attach', 522 * we need to give it back to the old parent. 523 */ 524 pr = p->p_p; 525 if ((p->p_flag & P_THREAD) == 0 && pr->ps_oppid && 526 (tr = prfind(pr->ps_oppid))) { 527 atomic_clearbits_int(&pr->ps_flags, PS_TRACED); 528 pr->ps_oppid = 0; 529 proc_reparent(pr, tr); 530 if (p->p_exitsig != 0) 531 prsignal(tr, p->p_exitsig); 532 wakeup(tr); 533 } else { 534 scheduler_wait_hook(waiter, p); 535 p->p_xstat = 0; 536 ruadd(&waiter->p_stats->p_cru, p->p_ru); 537 proc_zap(p); 538 } 539 } 540 541 /* 542 * make process 'parent' the new parent of process 'child'. 543 */ 544 void 545 proc_reparent(struct process *child, struct process *parent) 546 { 547 548 if (child->ps_pptr == parent) 549 return; 550 551 if (parent == initproc->p_p) 552 child->ps_mainproc->p_exitsig = SIGCHLD; 553 554 LIST_REMOVE(child, ps_sibling); 555 LIST_INSERT_HEAD(&parent->ps_children, child, ps_sibling); 556 child->ps_pptr = parent; 557 } 558 559 void 560 proc_zap(struct proc *p) 561 { 562 struct process *pr = p->p_p; 563 564 pool_put(&rusage_pool, p->p_ru); 565 if ((p->p_flag & P_THREAD) == 0 && pr->ps_ptstat) 566 free(pr->ps_ptstat, M_SUBPROC); 567 568 /* 569 * Finally finished with old proc entry. 570 * Unlink it from its process group and free it. 571 */ 572 if ((p->p_flag & P_THREAD) == 0) 573 leavepgrp(pr); 574 LIST_REMOVE(p, p_list); /* off zombproc */ 575 if ((p->p_flag & P_THREAD) == 0) 576 LIST_REMOVE(pr, ps_sibling); 577 578 /* 579 * Decrement the count of procs running with this uid. 580 */ 581 (void)chgproccnt(p->p_cred->p_ruid, -1); 582 583 /* 584 * Release reference to text vnode 585 */ 586 if (p->p_textvp) 587 vrele(p->p_textvp); 588 589 /* 590 * Remove us from our process list, possibly killing the process 591 * in the process (pun intended). 592 */ 593 if (--pr->ps_refcnt == 0) { 594 KASSERT(TAILQ_EMPTY(&pr->ps_threads)); 595 limfree(pr->ps_limit); 596 crfree(pr->ps_cred->pc_ucred); 597 pool_put(&pcred_pool, pr->ps_cred); 598 pool_put(&process_pool, pr); 599 } 600 601 pool_put(&proc_pool, p); 602 nprocs--; 603 } 604