1 /* $OpenBSD: kern_exit.c,v 1.135 2014/02/12 05:47:36 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 74 #include <uvm/uvm_extern.h> 75 76 /* 77 * exit -- 78 * Death of process. 79 */ 80 int 81 sys_exit(struct proc *p, void *v, register_t *retval) 82 { 83 struct sys_exit_args /* { 84 syscallarg(int) rval; 85 } */ *uap = v; 86 87 exit1(p, W_EXITCODE(SCARG(uap, rval), 0), EXIT_NORMAL); 88 /* NOTREACHED */ 89 return (0); 90 } 91 92 int 93 sys___threxit(struct proc *p, void *v, register_t *retval) 94 { 95 struct sys___threxit_args /* { 96 syscallarg(pid_t *) notdead; 97 } */ *uap = v; 98 99 if (SCARG(uap, notdead) != NULL) { 100 pid_t zero = 0; 101 if (copyout(&zero, SCARG(uap, notdead), sizeof(zero))) 102 psignal(p, SIGSEGV); 103 } 104 exit1(p, 0, EXIT_THREAD); 105 106 return (0); 107 } 108 109 /* 110 * Exit: deallocate address space and other resources, change proc state 111 * to zombie, and unlink proc from allproc and parent's lists. Save exit 112 * status and rusage for wait(). Check for child processes and orphan them. 113 */ 114 void 115 exit1(struct proc *p, int rv, int flags) 116 { 117 struct process *pr, *qr, *nqr; 118 struct rusage *rup; 119 struct vnode *ovp; 120 121 if (p->p_pid == 1) 122 panic("init died (signal %d, exit %d)", 123 WTERMSIG(rv), WEXITSTATUS(rv)); 124 125 atomic_setbits_int(&p->p_flag, P_WEXIT); 126 127 pr = p->p_p; 128 129 /* single-threaded? */ 130 if (TAILQ_FIRST(&pr->ps_threads) == p && 131 TAILQ_NEXT(p, p_thr_link) == NULL) { 132 flags = EXIT_NORMAL; 133 } else { 134 /* nope, multi-threaded */ 135 if (flags == EXIT_NORMAL) 136 single_thread_set(p, SINGLE_EXIT, 0); 137 else if (flags == EXIT_THREAD) 138 single_thread_check(p, 0); 139 } 140 141 if (flags == EXIT_NORMAL) { 142 atomic_setbits_int(&pr->ps_flags, PS_EXITING); 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 if (pr->ps_flags & PS_PROFIL) 164 stopprofclock(pr); 165 } else if (TAILQ_EMPTY(&pr->ps_threads)) { 166 wakeup(&pr->ps_threads); 167 } 168 169 rup = pr->ps_ru; 170 if (rup == NULL) { 171 rup = pool_get(&rusage_pool, PR_WAITOK | PR_ZERO); 172 if (pr->ps_ru == NULL) { 173 pr->ps_ru = rup; 174 } else { 175 pool_put(&rusage_pool, rup); 176 rup = pr->ps_ru; 177 } 178 } 179 p->p_siglist = 0; 180 181 /* 182 * Close open files and release open-file table. 183 */ 184 fdfree(p); 185 186 if ((p->p_flag & P_THREAD) == 0) { 187 timeout_del(&pr->ps_realit_to); 188 #ifdef SYSVSEM 189 semexit(pr); 190 #endif 191 if (SESS_LEADER(pr)) { 192 struct session *sp = pr->ps_session; 193 194 if (sp->s_ttyvp) { 195 /* 196 * Controlling process. 197 * Signal foreground pgrp, 198 * drain controlling terminal 199 * and revoke access to controlling terminal. 200 */ 201 if (sp->s_ttyp->t_session == sp) { 202 if (sp->s_ttyp->t_pgrp) 203 pgsignal(sp->s_ttyp->t_pgrp, 204 SIGHUP, 1); 205 ttywait(sp->s_ttyp); 206 /* 207 * The tty could have been revoked 208 * if we blocked. 209 */ 210 if (sp->s_ttyvp) 211 VOP_REVOKE(sp->s_ttyvp, 212 REVOKEALL); 213 } 214 ovp = sp->s_ttyvp; 215 sp->s_ttyvp = NULL; 216 if (ovp) 217 vrele(ovp); 218 /* 219 * s_ttyp is not zero'd; we use this to 220 * indicate that the session once had a 221 * controlling terminal. (for logging and 222 * informational purposes) 223 */ 224 } 225 sp->s_leader = NULL; 226 } 227 fixjobc(pr, pr->ps_pgrp, 0); 228 229 #ifdef ACCOUNTING 230 acct_process(p); 231 #endif 232 233 #ifdef KTRACE 234 /* release trace file */ 235 if (pr->ps_tracevp) 236 ktrcleartrace(pr); 237 #endif 238 239 /* 240 * If parent has the SAS_NOCLDWAIT flag set, we're not 241 * going to become a zombie. 242 */ 243 if (pr->ps_pptr->ps_mainproc->p_sigacts->ps_flags & 244 SAS_NOCLDWAIT) 245 atomic_setbits_int(&pr->ps_flags, PS_NOZOMBIE); 246 } 247 248 #if NSYSTRACE > 0 249 if (ISSET(p->p_flag, P_SYSTRACE)) 250 systrace_exit(p); 251 #endif 252 253 /* 254 * If emulation has process exit hook, call it now. 255 */ 256 if (p->p_emul->e_proc_exit) 257 (*p->p_emul->e_proc_exit)(p); 258 259 /* 260 * Remove proc from pidhash chain and allproc so looking 261 * it up won't work. We will put the proc on the 262 * deadproc list later (using the p_hash member), and 263 * wake up the reaper when we do. If this is the last 264 * thread of a process that isn't PS_NOZOMBIE, we'll put 265 * the process on the zombprocess list below. 266 */ 267 /* 268 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP! 269 */ 270 p->p_stat = SDEAD; 271 272 LIST_REMOVE(p, p_hash); 273 LIST_REMOVE(p, p_list); 274 275 if ((p->p_flag & P_THREAD) == 0) { 276 LIST_REMOVE(pr, ps_list); 277 278 if ((pr->ps_flags & PS_NOZOMBIE) == 0) 279 LIST_INSERT_HEAD(&zombprocess, pr, ps_list); 280 else { 281 /* 282 * Not going to be a zombie, so it's now off all 283 * the lists scanned by ispidtaken(), so block 284 * fast reuse of the pid now. 285 */ 286 freepid(p->p_pid); 287 } 288 289 /* 290 * Give orphaned children to init(8). 291 */ 292 qr = LIST_FIRST(&pr->ps_children); 293 if (qr) /* only need this if any child is S_ZOMB */ 294 wakeup(initproc->p_p); 295 for (; qr != 0; qr = nqr) { 296 nqr = LIST_NEXT(qr, ps_sibling); 297 proc_reparent(qr, initproc->p_p); 298 /* 299 * Traced processes are killed since their 300 * existence means someone is screwing up. 301 */ 302 if (qr->ps_flags & PS_TRACED && 303 !(qr->ps_flags & PS_EXITING)) { 304 atomic_clearbits_int(&qr->ps_flags, PS_TRACED); 305 /* 306 * If single threading is active, 307 * direct the signal to the active 308 * thread to avoid deadlock. 309 */ 310 if (qr->ps_single) 311 ptsignal(qr->ps_single, SIGKILL, 312 STHREAD); 313 else 314 prsignal(qr, SIGKILL); 315 } 316 } 317 } 318 319 /* add thread's accumulated rusage into the process's total */ 320 ruadd(rup, &p->p_ru); 321 tuagg(pr, p); 322 323 /* 324 * clear %cpu usage during swap 325 */ 326 p->p_pctcpu = 0; 327 328 if ((p->p_flag & P_THREAD) == 0) { 329 /* 330 * Final thread has died, so add on our children's rusage 331 * and calculate the total times 332 */ 333 calcru(&pr->ps_tu, &rup->ru_utime, &rup->ru_stime, NULL); 334 ruadd(rup, &pr->ps_cru); 335 336 /* notify interested parties of our demise and clean up */ 337 knote_processexit(pr); 338 339 /* 340 * Notify parent that we're gone. If we're not going to 341 * become a zombie, reparent to process 1 (init) so that 342 * we can wake our original parent to possibly unblock 343 * wait4() to return ECHILD. 344 */ 345 if (pr->ps_flags & PS_NOZOMBIE) { 346 struct process *ppr = pr->ps_pptr; 347 proc_reparent(pr, initproc->p_p); 348 wakeup(ppr); 349 } 350 } 351 352 /* 353 * Release the process's signal state. 354 */ 355 sigactsfree(p); 356 357 /* just a thread? detach it from its process */ 358 if (p->p_flag & P_THREAD) { 359 /* scheduler_wait_hook(pr->ps_mainproc, p); XXX */ 360 --pr->ps_refcnt; 361 KASSERT(pr->ps_refcnt > 0); 362 } 363 364 /* 365 * Other substructures are freed from reaper and wait(). 366 */ 367 368 /* 369 * Finally, call machine-dependent code to switch to a new 370 * context (possibly the idle context). Once we are no longer 371 * using the dead process's vmspace and stack, exit2() will be 372 * called to schedule those resources to be released by the 373 * reaper thread. 374 * 375 * Note that cpu_exit() will end with a call equivalent to 376 * cpu_switch(), finishing our execution (pun intended). 377 */ 378 uvmexp.swtch++; 379 cpu_exit(p); 380 panic("cpu_exit returned"); 381 } 382 383 /* 384 * Locking of this proclist is special; it's accessed in a 385 * critical section of process exit, and thus locking it can't 386 * modify interrupt state. We use a simple spin lock for this 387 * proclist. We use the p_hash member to linkup to deadproc. 388 */ 389 struct mutex deadproc_mutex = MUTEX_INITIALIZER(IPL_NONE); 390 struct proclist deadproc = LIST_HEAD_INITIALIZER(deadproc); 391 392 /* 393 * We are called from cpu_exit() once it is safe to schedule the 394 * dead process's resources to be freed. 395 * 396 * NOTE: One must be careful with locking in this routine. It's 397 * called from a critical section in machine-dependent code, so 398 * we should refrain from changing any interrupt state. 399 * 400 * We lock the deadproc list, place the proc on that list (using 401 * the p_hash member), and wake up the reaper. 402 */ 403 void 404 exit2(struct proc *p) 405 { 406 mtx_enter(&deadproc_mutex); 407 LIST_INSERT_HEAD(&deadproc, p, p_hash); 408 mtx_leave(&deadproc_mutex); 409 410 wakeup(&deadproc); 411 } 412 413 void 414 proc_free(struct proc *p) 415 { 416 pool_put(&proc_pool, p); 417 nthreads--; 418 } 419 420 /* 421 * Process reaper. This is run by a kernel thread to free the resources 422 * of a dead process. Once the resources are free, the process becomes 423 * a zombie, and the parent is allowed to read the undead's status. 424 */ 425 void 426 reaper(void) 427 { 428 struct proc *p; 429 430 KERNEL_UNLOCK(); 431 432 SCHED_ASSERT_UNLOCKED(); 433 434 for (;;) { 435 mtx_enter(&deadproc_mutex); 436 while ((p = LIST_FIRST(&deadproc)) == NULL) 437 msleep(&deadproc, &deadproc_mutex, PVM, "reaper", 0); 438 439 /* Remove us from the deadproc list. */ 440 LIST_REMOVE(p, p_hash); 441 mtx_leave(&deadproc_mutex); 442 443 KERNEL_LOCK(); 444 445 /* 446 * Free the VM resources we're still holding on to. 447 * We must do this from a valid thread because doing 448 * so may block. 449 */ 450 uvm_exit(p); 451 452 if (p->p_flag & P_THREAD) { 453 /* Just a thread */ 454 proc_free(p); 455 } else { 456 struct process *pr = p->p_p; 457 458 if ((pr->ps_flags & PS_NOZOMBIE) == 0) { 459 /* Process is now a true zombie. */ 460 p->p_stat = SZOMB; 461 prsignal(pr->ps_pptr, SIGCHLD); 462 463 /* Wake up the parent so it can get exit status. */ 464 wakeup(pr->ps_pptr); 465 } else { 466 /* No one will wait for us. Just zap the process now */ 467 process_zap(pr); 468 } 469 } 470 471 KERNEL_UNLOCK(); 472 } 473 } 474 475 int 476 sys_wait4(struct proc *q, void *v, register_t *retval) 477 { 478 struct sys_wait4_args /* { 479 syscallarg(pid_t) pid; 480 syscallarg(int *) status; 481 syscallarg(int) options; 482 syscallarg(struct rusage *) rusage; 483 } */ *uap = v; 484 struct rusage ru; 485 int status, error; 486 487 error = dowait4(q, SCARG(uap, pid), 488 SCARG(uap, status) ? &status : NULL, 489 SCARG(uap, options), SCARG(uap, rusage) ? &ru : NULL, retval); 490 if (error == 0 && SCARG(uap, status)) { 491 error = copyout(&status, SCARG(uap, status), sizeof(status)); 492 } 493 if (error == 0 && SCARG(uap, rusage)) { 494 error = copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 495 #ifdef KTRACE 496 if (error == 0 && KTRPOINT(q, KTR_STRUCT)) 497 ktrrusage(q, &ru); 498 #endif 499 } 500 return (error); 501 } 502 503 int 504 dowait4(struct proc *q, pid_t pid, int *statusp, int options, 505 struct rusage *rusage, register_t *retval) 506 { 507 int nfound; 508 struct process *pr; 509 struct proc *p; 510 int error; 511 512 if (pid == 0) 513 pid = -q->p_p->ps_pgid; 514 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED)) 515 return (EINVAL); 516 517 loop: 518 nfound = 0; 519 LIST_FOREACH(pr, &q->p_p->ps_children, ps_sibling) { 520 p = pr->ps_mainproc; 521 if ((pr->ps_flags & PS_NOZOMBIE) || 522 (pid != WAIT_ANY && 523 p->p_pid != pid && 524 pr->ps_pgid != -pid)) 525 continue; 526 527 nfound++; 528 if (p->p_stat == SZOMB) { 529 retval[0] = p->p_pid; 530 531 if (statusp != NULL) 532 *statusp = p->p_xstat; /* convert to int */ 533 if (rusage != NULL) 534 memcpy(rusage, pr->ps_ru, sizeof(*rusage)); 535 proc_finish_wait(q, p); 536 return (0); 537 } 538 if (pr->ps_flags & PS_TRACED && 539 (pr->ps_flags & PS_WAITED) == 0 && pr->ps_single && 540 pr->ps_single->p_stat == SSTOP && 541 (pr->ps_single->p_flag & P_SUSPSINGLE) == 0) { 542 single_thread_wait(pr); 543 544 atomic_setbits_int(&pr->ps_flags, PS_WAITED); 545 retval[0] = p->p_pid; 546 547 if (statusp != NULL) 548 *statusp = W_STOPCODE(pr->ps_single->p_xstat); 549 return (0); 550 } 551 if (p->p_stat == SSTOP && 552 (pr->ps_flags & PS_WAITED) == 0 && 553 (p->p_flag & P_SUSPSINGLE) == 0 && 554 (pr->ps_flags & PS_TRACED || 555 options & WUNTRACED)) { 556 atomic_setbits_int(&pr->ps_flags, PS_WAITED); 557 retval[0] = p->p_pid; 558 559 if (statusp != NULL) 560 *statusp = W_STOPCODE(p->p_xstat); 561 return (0); 562 } 563 if ((options & WCONTINUED) && (p->p_flag & P_CONTINUED)) { 564 atomic_clearbits_int(&p->p_flag, P_CONTINUED); 565 retval[0] = p->p_pid; 566 567 if (statusp != NULL) 568 *statusp = _WCONTINUED; 569 return (0); 570 } 571 } 572 if (nfound == 0) 573 return (ECHILD); 574 if (options & WNOHANG) { 575 retval[0] = 0; 576 return (0); 577 } 578 if ((error = tsleep(q->p_p, PWAIT | PCATCH, "wait", 0)) != 0) 579 return (error); 580 goto loop; 581 } 582 583 void 584 proc_finish_wait(struct proc *waiter, struct proc *p) 585 { 586 struct process *pr, *tr; 587 struct rusage *rup; 588 589 /* 590 * If we got the child via a ptrace 'attach', 591 * we need to give it back to the old parent. 592 */ 593 pr = p->p_p; 594 if (pr->ps_oppid && (tr = prfind(pr->ps_oppid))) { 595 atomic_clearbits_int(&pr->ps_flags, PS_TRACED); 596 pr->ps_oppid = 0; 597 proc_reparent(pr, tr); 598 prsignal(tr, SIGCHLD); 599 wakeup(tr); 600 } else { 601 scheduler_wait_hook(waiter, p); 602 p->p_xstat = 0; 603 rup = &waiter->p_p->ps_cru; 604 ruadd(rup, pr->ps_ru); 605 LIST_REMOVE(pr, ps_list); /* off zombprocess */ 606 freepid(p->p_pid); 607 process_zap(pr); 608 } 609 } 610 611 /* 612 * make process 'parent' the new parent of process 'child'. 613 */ 614 void 615 proc_reparent(struct process *child, struct process *parent) 616 { 617 618 if (child->ps_pptr == parent) 619 return; 620 621 LIST_REMOVE(child, ps_sibling); 622 LIST_INSERT_HEAD(&parent->ps_children, child, ps_sibling); 623 child->ps_pptr = parent; 624 } 625 626 void 627 process_zap(struct process *pr) 628 { 629 struct vnode *otvp; 630 struct proc *p = pr->ps_mainproc; 631 632 /* 633 * Finally finished with old proc entry. 634 * Unlink it from its process group and free it. 635 */ 636 leavepgrp(pr); 637 LIST_REMOVE(pr, ps_sibling); 638 639 /* 640 * Decrement the count of procs running with this uid. 641 */ 642 (void)chgproccnt(pr->ps_cred->p_ruid, -1); 643 644 /* 645 * Release reference to text vnode 646 */ 647 otvp = pr->ps_textvp; 648 pr->ps_textvp = NULL; 649 if (otvp) 650 vrele(otvp); 651 652 KASSERT(pr->ps_refcnt == 1); 653 if (pr->ps_ptstat != NULL) 654 free(pr->ps_ptstat, M_SUBPROC); 655 pool_put(&rusage_pool, pr->ps_ru); 656 KASSERT(TAILQ_EMPTY(&pr->ps_threads)); 657 limfree(pr->ps_limit); 658 crfree(pr->ps_cred->pc_ucred); 659 pool_put(&pcred_pool, pr->ps_cred); 660 pool_put(&process_pool, pr); 661 nprocesses--; 662 663 proc_free(p); 664 } 665