1 /* $OpenBSD: kern_exit.c,v 1.147 2014/07/12 18:43:32 tedu 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/sysctl.h> 49 #include <sys/buf.h> 50 #include <sys/wait.h> 51 #include <sys/file.h> 52 #include <sys/vnode.h> 53 #include <sys/syslog.h> 54 #include <sys/malloc.h> 55 #include <sys/resourcevar.h> 56 #include <sys/ptrace.h> 57 #include <sys/acct.h> 58 #include <sys/filedesc.h> 59 #include <sys/signalvar.h> 60 #include <sys/sched.h> 61 #include <sys/ktrace.h> 62 #include <sys/pool.h> 63 #include <sys/mutex.h> 64 #ifdef SYSVSEM 65 #include <sys/sem.h> 66 #endif 67 68 #include "systrace.h" 69 #include <dev/systrace.h> 70 71 #include <sys/mount.h> 72 #include <sys/syscallargs.h> 73 74 /* 75 * exit -- 76 * Death of process. 77 */ 78 int 79 sys_exit(struct proc *p, void *v, register_t *retval) 80 { 81 struct sys_exit_args /* { 82 syscallarg(int) rval; 83 } */ *uap = v; 84 85 exit1(p, W_EXITCODE(SCARG(uap, rval), 0), EXIT_NORMAL); 86 /* NOTREACHED */ 87 return (0); 88 } 89 90 int 91 sys___threxit(struct proc *p, void *v, register_t *retval) 92 { 93 struct sys___threxit_args /* { 94 syscallarg(pid_t *) notdead; 95 } */ *uap = v; 96 97 if (SCARG(uap, notdead) != NULL) { 98 pid_t zero = 0; 99 if (copyout(&zero, SCARG(uap, notdead), sizeof(zero))) 100 psignal(p, SIGSEGV); 101 } 102 exit1(p, 0, EXIT_THREAD); 103 104 return (0); 105 } 106 107 /* 108 * Exit: deallocate address space and other resources, change proc state 109 * to zombie, and unlink proc from allproc and parent's lists. Save exit 110 * status and rusage for wait(). Check for child processes and orphan them. 111 */ 112 void 113 exit1(struct proc *p, int rv, int flags) 114 { 115 struct process *pr, *qr, *nqr; 116 struct rusage *rup; 117 struct vnode *ovp; 118 119 atomic_setbits_int(&p->p_flag, P_WEXIT); 120 121 pr = p->p_p; 122 123 /* single-threaded? */ 124 if (TAILQ_FIRST(&pr->ps_threads) == p && 125 TAILQ_NEXT(p, p_thr_link) == NULL) { 126 flags = EXIT_NORMAL; 127 } else { 128 /* nope, multi-threaded */ 129 if (flags == EXIT_NORMAL) 130 single_thread_set(p, SINGLE_EXIT, 0); 131 else if (flags == EXIT_THREAD) 132 single_thread_check(p, 0); 133 } 134 135 if (flags == EXIT_NORMAL) { 136 if (pr->ps_pid == 1) 137 panic("init died (signal %d, exit %d)", 138 WTERMSIG(rv), WEXITSTATUS(rv)); 139 140 atomic_setbits_int(&pr->ps_flags, PS_EXITING); 141 pr->ps_mainproc->p_xstat = rv; 142 143 /* 144 * If parent is waiting for us to exit or exec, PS_PPWAIT 145 * is set; we wake up the parent early to avoid deadlock. 146 */ 147 if (pr->ps_flags & PS_PPWAIT) { 148 atomic_clearbits_int(&pr->ps_flags, PS_PPWAIT); 149 atomic_clearbits_int(&pr->ps_pptr->ps_flags, 150 PS_ISPWAIT); 151 wakeup(pr->ps_pptr); 152 } 153 } 154 155 /* unlink ourselves from the active threads */ 156 TAILQ_REMOVE(&pr->ps_threads, p, p_thr_link); 157 if ((p->p_flag & P_THREAD) == 0) { 158 /* main thread gotta wait because it has the pid, et al */ 159 while (pr->ps_refcnt > 1) 160 tsleep(&pr->ps_threads, PUSER, "thrdeath", 0); 161 if (pr->ps_flags & PS_PROFIL) 162 stopprofclock(pr); 163 } 164 165 rup = pr->ps_ru; 166 if (rup == NULL) { 167 rup = pool_get(&rusage_pool, PR_WAITOK | PR_ZERO); 168 if (pr->ps_ru == NULL) { 169 pr->ps_ru = rup; 170 } else { 171 pool_put(&rusage_pool, rup); 172 rup = pr->ps_ru; 173 } 174 } 175 p->p_siglist = 0; 176 177 if ((p->p_flag & P_THREAD) == 0) { 178 /* close open files and release open-file table */ 179 fdfree(p); 180 181 timeout_del(&pr->ps_realit_to); 182 #ifdef SYSVSEM 183 semexit(pr); 184 #endif 185 if (SESS_LEADER(pr)) { 186 struct session *sp = pr->ps_session; 187 188 if (sp->s_ttyvp) { 189 /* 190 * Controlling process. 191 * Signal foreground pgrp, 192 * drain controlling terminal 193 * and revoke access to controlling terminal. 194 */ 195 if (sp->s_ttyp->t_session == sp) { 196 if (sp->s_ttyp->t_pgrp) 197 pgsignal(sp->s_ttyp->t_pgrp, 198 SIGHUP, 1); 199 ttywait(sp->s_ttyp); 200 /* 201 * The tty could have been revoked 202 * if we blocked. 203 */ 204 if (sp->s_ttyvp) 205 VOP_REVOKE(sp->s_ttyvp, 206 REVOKEALL); 207 } 208 ovp = sp->s_ttyvp; 209 sp->s_ttyvp = NULL; 210 if (ovp) 211 vrele(ovp); 212 /* 213 * s_ttyp is not zero'd; we use this to 214 * indicate that the session once had a 215 * controlling terminal. (for logging and 216 * informational purposes) 217 */ 218 } 219 sp->s_leader = NULL; 220 } 221 fixjobc(pr, pr->ps_pgrp, 0); 222 223 #ifdef ACCOUNTING 224 acct_process(p); 225 #endif 226 227 #ifdef KTRACE 228 /* release trace file */ 229 if (pr->ps_tracevp) 230 ktrcleartrace(pr); 231 #endif 232 233 /* 234 * If parent has the SAS_NOCLDWAIT flag set, we're not 235 * going to become a zombie. 236 */ 237 if (pr->ps_pptr->ps_sigacts->ps_flags & SAS_NOCLDWAIT) 238 atomic_setbits_int(&pr->ps_flags, PS_NOZOMBIE); 239 } 240 241 p->p_fd = NULL; /* zap the thread's copy */ 242 243 #if NSYSTRACE > 0 244 if (ISSET(p->p_flag, P_SYSTRACE)) 245 systrace_exit(p); 246 #endif 247 248 /* 249 * If emulation has thread exit hook, call it now. 250 */ 251 if (pr->ps_emul->e_proc_exit) 252 (*pr->ps_emul->e_proc_exit)(p); 253 254 /* 255 * Remove proc from pidhash chain and allproc so looking 256 * it up won't work. We will put the proc on the 257 * deadproc list later (using the p_hash member), and 258 * wake up the reaper when we do. If this is the last 259 * thread of a process that isn't PS_NOZOMBIE, we'll put 260 * the process on the zombprocess list below. 261 */ 262 /* 263 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP! 264 */ 265 p->p_stat = SDEAD; 266 267 LIST_REMOVE(p, p_hash); 268 LIST_REMOVE(p, p_list); 269 270 if ((p->p_flag & P_THREAD) == 0) { 271 LIST_REMOVE(pr, ps_list); 272 273 if ((pr->ps_flags & PS_NOZOMBIE) == 0) 274 LIST_INSERT_HEAD(&zombprocess, pr, ps_list); 275 else { 276 /* 277 * Not going to be a zombie, so it's now off all 278 * the lists scanned by ispidtaken(), so block 279 * fast reuse of the pid now. 280 */ 281 freepid(p->p_pid); 282 } 283 284 /* 285 * Give orphaned children to init(8). 286 */ 287 qr = LIST_FIRST(&pr->ps_children); 288 if (qr) /* only need this if any child is S_ZOMB */ 289 wakeup(initprocess); 290 for (; qr != 0; qr = nqr) { 291 nqr = LIST_NEXT(qr, ps_sibling); 292 proc_reparent(qr, initprocess); 293 /* 294 * Traced processes are killed since their 295 * existence means someone is screwing up. 296 */ 297 if (qr->ps_flags & PS_TRACED && 298 !(qr->ps_flags & PS_EXITING)) { 299 atomic_clearbits_int(&qr->ps_flags, PS_TRACED); 300 /* 301 * If single threading is active, 302 * direct the signal to the active 303 * thread to avoid deadlock. 304 */ 305 if (qr->ps_single) 306 ptsignal(qr->ps_single, SIGKILL, 307 STHREAD); 308 else 309 prsignal(qr, SIGKILL); 310 } 311 } 312 } 313 314 /* add thread's accumulated rusage into the process's total */ 315 ruadd(rup, &p->p_ru); 316 tuagg(pr, p); 317 318 /* 319 * clear %cpu usage during swap 320 */ 321 p->p_pctcpu = 0; 322 323 if ((p->p_flag & P_THREAD) == 0) { 324 /* 325 * Final thread has died, so add on our children's rusage 326 * and calculate the total times 327 */ 328 calcru(&pr->ps_tu, &rup->ru_utime, &rup->ru_stime, NULL); 329 ruadd(rup, &pr->ps_cru); 330 331 /* notify interested parties of our demise and clean up */ 332 knote_processexit(p); 333 334 /* 335 * Notify parent that we're gone. If we're not going to 336 * become a zombie, reparent to process 1 (init) so that 337 * we can wake our original parent to possibly unblock 338 * wait4() to return ECHILD. 339 */ 340 if (pr->ps_flags & PS_NOZOMBIE) { 341 struct process *ppr = pr->ps_pptr; 342 proc_reparent(pr, initprocess); 343 wakeup(ppr); 344 } 345 346 /* 347 * Release the process's signal state. 348 */ 349 sigactsfree(pr); 350 } 351 352 /* just a thread? detach it from its process */ 353 if (p->p_flag & P_THREAD) { 354 /* scheduler_wait_hook(pr->ps_mainproc, p); XXX */ 355 if (--pr->ps_refcnt == 1) 356 wakeup(&pr->ps_threads); 357 KASSERT(pr->ps_refcnt > 0); 358 } 359 360 /* 361 * Other substructures are freed from reaper and wait(). 362 */ 363 364 /* 365 * Finally, call machine-dependent code to switch to a new 366 * context (possibly the idle context). Once we are no longer 367 * using the dead process's vmspace and stack, exit2() will be 368 * called to schedule those resources to be released by the 369 * reaper thread. 370 * 371 * Note that cpu_exit() will end with a call equivalent to 372 * cpu_switch(), finishing our execution (pun intended). 373 */ 374 uvmexp.swtch++; 375 cpu_exit(p); 376 panic("cpu_exit returned"); 377 } 378 379 /* 380 * Locking of this proclist is special; it's accessed in a 381 * critical section of process exit, and thus locking it can't 382 * modify interrupt state. We use a simple spin lock for this 383 * proclist. We use the p_hash member to linkup to deadproc. 384 */ 385 struct mutex deadproc_mutex = MUTEX_INITIALIZER(IPL_NONE); 386 struct proclist deadproc = LIST_HEAD_INITIALIZER(deadproc); 387 388 /* 389 * We are called from cpu_exit() once it is safe to schedule the 390 * dead process's resources to be freed. 391 * 392 * NOTE: One must be careful with locking in this routine. It's 393 * called from a critical section in machine-dependent code, so 394 * we should refrain from changing any interrupt state. 395 * 396 * We lock the deadproc list, place the proc on that list (using 397 * the p_hash member), and wake up the reaper. 398 */ 399 void 400 exit2(struct proc *p) 401 { 402 mtx_enter(&deadproc_mutex); 403 LIST_INSERT_HEAD(&deadproc, p, p_hash); 404 mtx_leave(&deadproc_mutex); 405 406 wakeup(&deadproc); 407 } 408 409 void 410 proc_free(struct proc *p) 411 { 412 crfree(p->p_ucred); 413 pool_put(&proc_pool, p); 414 nthreads--; 415 } 416 417 /* 418 * Process reaper. This is run by a kernel thread to free the resources 419 * of a dead process. Once the resources are free, the process becomes 420 * a zombie, and the parent is allowed to read the undead's status. 421 */ 422 void 423 reaper(void) 424 { 425 struct proc *p; 426 427 KERNEL_UNLOCK(); 428 429 SCHED_ASSERT_UNLOCKED(); 430 431 for (;;) { 432 mtx_enter(&deadproc_mutex); 433 while ((p = LIST_FIRST(&deadproc)) == NULL) 434 msleep(&deadproc, &deadproc_mutex, PVM, "reaper", 0); 435 436 /* Remove us from the deadproc list. */ 437 LIST_REMOVE(p, p_hash); 438 mtx_leave(&deadproc_mutex); 439 440 KERNEL_LOCK(); 441 442 /* 443 * Free the VM resources we're still holding on to. 444 * We must do this from a valid thread because doing 445 * so may block. 446 */ 447 uvm_uarea_free(p); 448 p->p_vmspace = NULL; /* zap the thread's copy */ 449 450 if (p->p_flag & P_THREAD) { 451 /* Just a thread */ 452 proc_free(p); 453 } else { 454 struct process *pr = p->p_p; 455 456 /* Release the rest of the process's vmspace */ 457 uvm_exit(pr); 458 459 if ((pr->ps_flags & PS_NOZOMBIE) == 0) { 460 /* Process is now a true zombie. */ 461 atomic_setbits_int(&pr->ps_flags, PS_ZOMBIE); 462 prsignal(pr->ps_pptr, SIGCHLD); 463 464 /* Wake up the parent so it can get exit status. */ 465 wakeup(pr->ps_pptr); 466 } else { 467 /* No one will wait for us. Just zap the process now */ 468 process_zap(pr); 469 } 470 } 471 472 KERNEL_UNLOCK(); 473 } 474 } 475 476 int 477 sys_wait4(struct proc *q, void *v, register_t *retval) 478 { 479 struct sys_wait4_args /* { 480 syscallarg(pid_t) pid; 481 syscallarg(int *) status; 482 syscallarg(int) options; 483 syscallarg(struct rusage *) rusage; 484 } */ *uap = v; 485 struct rusage ru; 486 int status, error; 487 488 error = dowait4(q, SCARG(uap, pid), 489 SCARG(uap, status) ? &status : NULL, 490 SCARG(uap, options), SCARG(uap, rusage) ? &ru : NULL, retval); 491 if (error == 0 && retval[0] > 0 && SCARG(uap, status)) { 492 error = copyout(&status, SCARG(uap, status), sizeof(status)); 493 } 494 if (error == 0 && retval[0] > 0 && SCARG(uap, rusage)) { 495 error = copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 496 #ifdef KTRACE 497 if (error == 0 && KTRPOINT(q, KTR_STRUCT)) 498 ktrrusage(q, &ru); 499 #endif 500 } 501 return (error); 502 } 503 504 int 505 dowait4(struct proc *q, pid_t pid, int *statusp, int options, 506 struct rusage *rusage, register_t *retval) 507 { 508 int nfound; 509 struct process *pr; 510 struct proc *p; 511 int error; 512 513 if (pid == 0) 514 pid = -q->p_p->ps_pgid; 515 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED)) 516 return (EINVAL); 517 518 loop: 519 nfound = 0; 520 LIST_FOREACH(pr, &q->p_p->ps_children, ps_sibling) { 521 p = pr->ps_mainproc; 522 if ((pr->ps_flags & PS_NOZOMBIE) || 523 (pid != WAIT_ANY && 524 p->p_pid != pid && 525 pr->ps_pgid != -pid)) 526 continue; 527 528 nfound++; 529 if (pr->ps_flags & PS_ZOMBIE) { 530 retval[0] = p->p_pid; 531 532 if (statusp != NULL) 533 *statusp = p->p_xstat; /* convert to int */ 534 if (rusage != NULL) 535 memcpy(rusage, pr->ps_ru, sizeof(*rusage)); 536 proc_finish_wait(q, p); 537 return (0); 538 } 539 if (pr->ps_flags & PS_TRACED && 540 (pr->ps_flags & PS_WAITED) == 0 && pr->ps_single && 541 pr->ps_single->p_stat == SSTOP && 542 (pr->ps_single->p_flag & P_SUSPSINGLE) == 0) { 543 single_thread_wait(pr); 544 545 atomic_setbits_int(&pr->ps_flags, PS_WAITED); 546 retval[0] = p->p_pid; 547 548 if (statusp != NULL) 549 *statusp = W_STOPCODE(pr->ps_single->p_xstat); 550 if (rusage != NULL) 551 memset(rusage, 0, sizeof(*rusage)); 552 return (0); 553 } 554 if (p->p_stat == SSTOP && 555 (pr->ps_flags & PS_WAITED) == 0 && 556 (p->p_flag & P_SUSPSINGLE) == 0 && 557 (pr->ps_flags & PS_TRACED || 558 options & WUNTRACED)) { 559 atomic_setbits_int(&pr->ps_flags, PS_WAITED); 560 retval[0] = p->p_pid; 561 562 if (statusp != NULL) 563 *statusp = W_STOPCODE(p->p_xstat); 564 if (rusage != NULL) 565 memset(rusage, 0, sizeof(*rusage)); 566 return (0); 567 } 568 if ((options & WCONTINUED) && (p->p_flag & P_CONTINUED)) { 569 atomic_clearbits_int(&p->p_flag, P_CONTINUED); 570 retval[0] = p->p_pid; 571 572 if (statusp != NULL) 573 *statusp = _WCONTINUED; 574 if (rusage != NULL) 575 memset(rusage, 0, sizeof(*rusage)); 576 return (0); 577 } 578 } 579 if (nfound == 0) 580 return (ECHILD); 581 if (options & WNOHANG) { 582 retval[0] = 0; 583 return (0); 584 } 585 if ((error = tsleep(q->p_p, PWAIT | PCATCH, "wait", 0)) != 0) 586 return (error); 587 goto loop; 588 } 589 590 void 591 proc_finish_wait(struct proc *waiter, struct proc *p) 592 { 593 struct process *pr, *tr; 594 struct rusage *rup; 595 596 /* 597 * If we got the child via a ptrace 'attach', 598 * we need to give it back to the old parent. 599 */ 600 pr = p->p_p; 601 if (pr->ps_oppid && (tr = prfind(pr->ps_oppid))) { 602 atomic_clearbits_int(&pr->ps_flags, PS_TRACED); 603 pr->ps_oppid = 0; 604 proc_reparent(pr, tr); 605 prsignal(tr, SIGCHLD); 606 wakeup(tr); 607 } else { 608 scheduler_wait_hook(waiter, p); 609 p->p_xstat = 0; 610 rup = &waiter->p_p->ps_cru; 611 ruadd(rup, pr->ps_ru); 612 LIST_REMOVE(pr, ps_list); /* off zombprocess */ 613 freepid(p->p_pid); 614 process_zap(pr); 615 } 616 } 617 618 /* 619 * make process 'parent' the new parent of process 'child'. 620 */ 621 void 622 proc_reparent(struct process *child, struct process *parent) 623 { 624 625 if (child->ps_pptr == parent) 626 return; 627 628 LIST_REMOVE(child, ps_sibling); 629 LIST_INSERT_HEAD(&parent->ps_children, child, ps_sibling); 630 child->ps_pptr = parent; 631 } 632 633 void 634 process_zap(struct process *pr) 635 { 636 struct vnode *otvp; 637 struct proc *p = pr->ps_mainproc; 638 639 /* 640 * Finally finished with old proc entry. 641 * Unlink it from its process group and free it. 642 */ 643 leavepgrp(pr); 644 LIST_REMOVE(pr, ps_sibling); 645 646 /* 647 * Decrement the count of procs running with this uid. 648 */ 649 (void)chgproccnt(pr->ps_ucred->cr_ruid, -1); 650 651 /* 652 * Release reference to text vnode 653 */ 654 otvp = pr->ps_textvp; 655 pr->ps_textvp = NULL; 656 if (otvp) 657 vrele(otvp); 658 659 KASSERT(pr->ps_refcnt == 1); 660 if (pr->ps_ptstat != NULL) 661 free(pr->ps_ptstat, M_SUBPROC, 0); 662 pool_put(&rusage_pool, pr->ps_ru); 663 KASSERT(TAILQ_EMPTY(&pr->ps_threads)); 664 limfree(pr->ps_limit); 665 crfree(pr->ps_ucred); 666 pool_put(&process_pool, pr); 667 nprocesses--; 668 669 proc_free(p); 670 } 671