1 /* $OpenBSD: kern_exit.c,v 1.162 2017/11/28 06:09:44 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/sysctl.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 #include <sys/pledge.h> 64 #ifdef SYSVSEM 65 #include <sys/sem.h> 66 #endif 67 #include <sys/witness.h> 68 69 #include <sys/mount.h> 70 #include <sys/syscallargs.h> 71 72 #include <uvm/uvm_extern.h> 73 74 void proc_finish_wait(struct proc *, struct proc *); 75 void process_zap(struct process *); 76 void proc_free(struct proc *); 77 78 /* 79 * exit -- 80 * Death of process. 81 */ 82 int 83 sys_exit(struct proc *p, void *v, register_t *retval) 84 { 85 struct sys_exit_args /* { 86 syscallarg(int) rval; 87 } */ *uap = v; 88 89 exit1(p, W_EXITCODE(SCARG(uap, rval), 0), EXIT_NORMAL); 90 /* NOTREACHED */ 91 return (0); 92 } 93 94 int 95 sys___threxit(struct proc *p, void *v, register_t *retval) 96 { 97 struct sys___threxit_args /* { 98 syscallarg(pid_t *) notdead; 99 } */ *uap = v; 100 101 if (SCARG(uap, notdead) != NULL) { 102 pid_t zero = 0; 103 if (copyout(&zero, SCARG(uap, notdead), sizeof(zero))) 104 psignal(p, SIGSEGV); 105 } 106 exit1(p, 0, EXIT_THREAD); 107 108 return (0); 109 } 110 111 /* 112 * Exit: deallocate address space and other resources, change proc state 113 * to zombie, and unlink proc from allproc and parent's lists. Save exit 114 * status and rusage for wait(). Check for child processes and orphan them. 115 */ 116 void 117 exit1(struct proc *p, int rv, int flags) 118 { 119 struct process *pr, *qr, *nqr; 120 struct rusage *rup; 121 struct vnode *ovp; 122 123 atomic_setbits_int(&p->p_flag, P_WEXIT); 124 125 pr = p->p_p; 126 127 /* single-threaded? */ 128 if (!P_HASSIBLING(p)) { 129 flags = EXIT_NORMAL; 130 } else { 131 /* nope, multi-threaded */ 132 if (flags == EXIT_NORMAL) 133 single_thread_set(p, SINGLE_EXIT, 0); 134 else if (flags == EXIT_THREAD) 135 single_thread_check(p, 0); 136 } 137 138 if (flags == EXIT_NORMAL) { 139 if (pr->ps_pid == 1) 140 panic("init died (signal %d, exit %d)", 141 WTERMSIG(rv), WEXITSTATUS(rv)); 142 143 atomic_setbits_int(&pr->ps_flags, PS_EXITING); 144 pr->ps_mainproc->p_xstat = rv; 145 146 /* 147 * If parent is waiting for us to exit or exec, PS_PPWAIT 148 * is set; we wake up the parent early to avoid deadlock. 149 */ 150 if (pr->ps_flags & PS_PPWAIT) { 151 atomic_clearbits_int(&pr->ps_flags, PS_PPWAIT); 152 atomic_clearbits_int(&pr->ps_pptr->ps_flags, 153 PS_ISPWAIT); 154 wakeup(pr->ps_pptr); 155 } 156 } 157 158 /* unlink ourselves from the active threads */ 159 TAILQ_REMOVE(&pr->ps_threads, p, p_thr_link); 160 if ((p->p_flag & P_THREAD) == 0) { 161 /* main thread gotta wait because it has the pid, et al */ 162 while (pr->ps_refcnt > 1) 163 tsleep(&pr->ps_threads, PUSER, "thrdeath", 0); 164 if (pr->ps_flags & PS_PROFIL) 165 stopprofclock(pr); 166 } 167 168 rup = pr->ps_ru; 169 if (rup == NULL) { 170 rup = pool_get(&rusage_pool, PR_WAITOK | PR_ZERO); 171 if (pr->ps_ru == NULL) { 172 pr->ps_ru = rup; 173 } else { 174 pool_put(&rusage_pool, rup); 175 rup = pr->ps_ru; 176 } 177 } 178 p->p_siglist = 0; 179 180 if ((p->p_flag & P_THREAD) == 0) { 181 /* close open files and release open-file table */ 182 fdfree(p); 183 184 timeout_del(&pr->ps_realit_to); 185 #ifdef SYSVSEM 186 semexit(pr); 187 #endif 188 if (SESS_LEADER(pr)) { 189 struct session *sp = pr->ps_session; 190 191 if (sp->s_ttyvp) { 192 /* 193 * Controlling process. 194 * Signal foreground pgrp, 195 * drain controlling terminal 196 * and revoke access to controlling terminal. 197 */ 198 if (sp->s_ttyp->t_session == sp) { 199 if (sp->s_ttyp->t_pgrp) 200 pgsignal(sp->s_ttyp->t_pgrp, 201 SIGHUP, 1); 202 ttywait(sp->s_ttyp); 203 /* 204 * The tty could have been revoked 205 * if we blocked. 206 */ 207 if (sp->s_ttyvp) 208 VOP_REVOKE(sp->s_ttyvp, 209 REVOKEALL); 210 } 211 ovp = sp->s_ttyvp; 212 sp->s_ttyvp = NULL; 213 if (ovp) 214 vrele(ovp); 215 /* 216 * s_ttyp is not zero'd; we use this to 217 * indicate that the session once had a 218 * controlling terminal. (for logging and 219 * informational purposes) 220 */ 221 } 222 sp->s_leader = NULL; 223 } 224 fixjobc(pr, pr->ps_pgrp, 0); 225 226 #ifdef ACCOUNTING 227 acct_process(p); 228 #endif 229 230 #ifdef KTRACE 231 /* release trace file */ 232 if (pr->ps_tracevp) 233 ktrcleartrace(pr); 234 #endif 235 236 /* 237 * If parent has the SAS_NOCLDWAIT flag set, we're not 238 * going to become a zombie. 239 */ 240 if (pr->ps_pptr->ps_sigacts->ps_flags & SAS_NOCLDWAIT) 241 atomic_setbits_int(&pr->ps_flags, PS_NOZOMBIE); 242 } 243 244 p->p_fd = NULL; /* zap the thread's copy */ 245 246 /* 247 * Remove proc from pidhash chain and allproc so looking 248 * it up won't work. We will put the proc on the 249 * deadproc list later (using the p_hash member), and 250 * wake up the reaper when we do. If this is the last 251 * thread of a process that isn't PS_NOZOMBIE, we'll put 252 * the process on the zombprocess list below. 253 */ 254 /* 255 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP! 256 */ 257 p->p_stat = SDEAD; 258 259 LIST_REMOVE(p, p_hash); 260 LIST_REMOVE(p, p_list); 261 262 if ((p->p_flag & P_THREAD) == 0) { 263 LIST_REMOVE(pr, ps_hash); 264 LIST_REMOVE(pr, ps_list); 265 266 if ((pr->ps_flags & PS_NOZOMBIE) == 0) 267 LIST_INSERT_HEAD(&zombprocess, pr, ps_list); 268 else { 269 /* 270 * Not going to be a zombie, so it's now off all 271 * the lists scanned by ispidtaken(), so block 272 * fast reuse of the pid now. 273 */ 274 freepid(pr->ps_pid); 275 } 276 277 /* 278 * Give orphaned children to init(8). 279 */ 280 qr = LIST_FIRST(&pr->ps_children); 281 if (qr) /* only need this if any child is S_ZOMB */ 282 wakeup(initprocess); 283 for (; qr != 0; qr = nqr) { 284 nqr = LIST_NEXT(qr, ps_sibling); 285 proc_reparent(qr, initprocess); 286 /* 287 * Traced processes are killed since their 288 * existence means someone is screwing up. 289 */ 290 if (qr->ps_flags & PS_TRACED && 291 !(qr->ps_flags & PS_EXITING)) { 292 atomic_clearbits_int(&qr->ps_flags, PS_TRACED); 293 /* 294 * If single threading is active, 295 * direct the signal to the active 296 * thread to avoid deadlock. 297 */ 298 if (qr->ps_single) 299 ptsignal(qr->ps_single, SIGKILL, 300 STHREAD); 301 else 302 prsignal(qr, SIGKILL); 303 } 304 } 305 } 306 307 /* add thread's accumulated rusage into the process's total */ 308 ruadd(rup, &p->p_ru); 309 tuagg(pr, p); 310 311 /* 312 * clear %cpu usage during swap 313 */ 314 p->p_pctcpu = 0; 315 316 if ((p->p_flag & P_THREAD) == 0) { 317 /* 318 * Final thread has died, so add on our children's rusage 319 * and calculate the total times 320 */ 321 calcru(&pr->ps_tu, &rup->ru_utime, &rup->ru_stime, NULL); 322 ruadd(rup, &pr->ps_cru); 323 324 /* notify interested parties of our demise and clean up */ 325 knote_processexit(p); 326 327 /* 328 * Notify parent that we're gone. If we're not going to 329 * become a zombie, reparent to process 1 (init) so that 330 * we can wake our original parent to possibly unblock 331 * wait4() to return ECHILD. 332 */ 333 if (pr->ps_flags & PS_NOZOMBIE) { 334 struct process *ppr = pr->ps_pptr; 335 proc_reparent(pr, initprocess); 336 wakeup(ppr); 337 } 338 339 /* 340 * Release the process's signal state. 341 */ 342 sigactsfree(pr); 343 } 344 345 /* just a thread? detach it from its process */ 346 if (p->p_flag & P_THREAD) { 347 /* scheduler_wait_hook(pr->ps_mainproc, p); XXX */ 348 if (--pr->ps_refcnt == 1) 349 wakeup(&pr->ps_threads); 350 KASSERT(pr->ps_refcnt > 0); 351 } 352 353 /* 354 * Other substructures are freed from reaper and wait(). 355 */ 356 357 /* 358 * Finally, call machine-dependent code to switch to a new 359 * context (possibly the idle context). Once we are no longer 360 * using the dead process's vmspace and stack, exit2() will be 361 * called to schedule those resources to be released by the 362 * reaper thread. 363 * 364 * Note that cpu_exit() will end with a call equivalent to 365 * cpu_switch(), finishing our execution (pun intended). 366 */ 367 uvmexp.swtch++; 368 cpu_exit(p); 369 panic("cpu_exit returned"); 370 } 371 372 /* 373 * Locking of this proclist is special; it's accessed in a 374 * critical section of process exit, and thus locking it can't 375 * modify interrupt state. We use a simple spin lock for this 376 * proclist. We use the p_hash member to linkup to deadproc. 377 */ 378 struct mutex deadproc_mutex = 379 MUTEX_INITIALIZER_FLAGS(IPL_NONE, NULL, MTX_NOWITNESS); 380 struct proclist deadproc = LIST_HEAD_INITIALIZER(deadproc); 381 382 /* 383 * We are called from cpu_exit() once it is safe to schedule the 384 * dead process's resources to be freed. 385 * 386 * NOTE: One must be careful with locking in this routine. It's 387 * called from a critical section in machine-dependent code, so 388 * we should refrain from changing any interrupt state. 389 * 390 * We lock the deadproc list, place the proc on that list (using 391 * the p_hash member), and wake up the reaper. 392 */ 393 void 394 exit2(struct proc *p) 395 { 396 mtx_enter(&deadproc_mutex); 397 LIST_INSERT_HEAD(&deadproc, p, p_hash); 398 mtx_leave(&deadproc_mutex); 399 400 wakeup(&deadproc); 401 } 402 403 void 404 proc_free(struct proc *p) 405 { 406 crfree(p->p_ucred); 407 pool_put(&proc_pool, p); 408 nthreads--; 409 } 410 411 /* 412 * Process reaper. This is run by a kernel thread to free the resources 413 * of a dead process. Once the resources are free, the process becomes 414 * a zombie, and the parent is allowed to read the undead's status. 415 */ 416 void 417 reaper(void) 418 { 419 struct proc *p; 420 421 KERNEL_UNLOCK(); 422 423 SCHED_ASSERT_UNLOCKED(); 424 425 for (;;) { 426 mtx_enter(&deadproc_mutex); 427 while ((p = LIST_FIRST(&deadproc)) == NULL) 428 msleep(&deadproc, &deadproc_mutex, PVM, "reaper", 0); 429 430 /* Remove us from the deadproc list. */ 431 LIST_REMOVE(p, p_hash); 432 mtx_leave(&deadproc_mutex); 433 434 WITNESS_THREAD_EXIT(p); 435 436 KERNEL_LOCK(); 437 438 /* 439 * Free the VM resources we're still holding on to. 440 * We must do this from a valid thread because doing 441 * so may block. 442 */ 443 uvm_uarea_free(p); 444 p->p_vmspace = NULL; /* zap the thread's copy */ 445 446 if (p->p_flag & P_THREAD) { 447 /* Just a thread */ 448 proc_free(p); 449 } else { 450 struct process *pr = p->p_p; 451 452 /* Release the rest of the process's vmspace */ 453 uvm_exit(pr); 454 455 if ((pr->ps_flags & PS_NOZOMBIE) == 0) { 456 /* Process is now a true zombie. */ 457 atomic_setbits_int(&pr->ps_flags, PS_ZOMBIE); 458 prsignal(pr->ps_pptr, SIGCHLD); 459 460 /* Wake up the parent so it can get exit status. */ 461 wakeup(pr->ps_pptr); 462 } else { 463 /* No one will wait for us. Just zap the process now */ 464 process_zap(pr); 465 } 466 } 467 468 KERNEL_UNLOCK(); 469 } 470 } 471 472 int 473 sys_wait4(struct proc *q, void *v, register_t *retval) 474 { 475 struct sys_wait4_args /* { 476 syscallarg(pid_t) pid; 477 syscallarg(int *) status; 478 syscallarg(int) options; 479 syscallarg(struct rusage *) rusage; 480 } */ *uap = v; 481 struct rusage ru; 482 int status, error; 483 484 error = dowait4(q, SCARG(uap, pid), 485 SCARG(uap, status) ? &status : NULL, 486 SCARG(uap, options), SCARG(uap, rusage) ? &ru : NULL, retval); 487 if (error == 0 && retval[0] > 0 && SCARG(uap, status)) { 488 error = copyout(&status, SCARG(uap, status), sizeof(status)); 489 } 490 if (error == 0 && retval[0] > 0 && SCARG(uap, rusage)) { 491 error = copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 492 #ifdef KTRACE 493 if (error == 0 && KTRPOINT(q, KTR_STRUCT)) 494 ktrrusage(q, &ru); 495 #endif 496 } 497 return (error); 498 } 499 500 int 501 dowait4(struct proc *q, pid_t pid, int *statusp, int options, 502 struct rusage *rusage, register_t *retval) 503 { 504 int nfound; 505 struct process *pr; 506 struct proc *p; 507 int error; 508 509 if (pid == 0) 510 pid = -q->p_p->ps_pgid; 511 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED)) 512 return (EINVAL); 513 514 loop: 515 nfound = 0; 516 LIST_FOREACH(pr, &q->p_p->ps_children, ps_sibling) { 517 if ((pr->ps_flags & PS_NOZOMBIE) || 518 (pid != WAIT_ANY && 519 pr->ps_pid != pid && 520 pr->ps_pgid != -pid)) 521 continue; 522 523 p = pr->ps_mainproc; 524 525 nfound++; 526 if (pr->ps_flags & PS_ZOMBIE) { 527 retval[0] = pr->ps_pid; 528 529 if (statusp != NULL) 530 *statusp = p->p_xstat; /* convert to int */ 531 if (rusage != NULL) 532 memcpy(rusage, pr->ps_ru, sizeof(*rusage)); 533 proc_finish_wait(q, p); 534 return (0); 535 } 536 if (pr->ps_flags & PS_TRACED && 537 (pr->ps_flags & PS_WAITED) == 0 && pr->ps_single && 538 pr->ps_single->p_stat == SSTOP && 539 (pr->ps_single->p_flag & P_SUSPSINGLE) == 0) { 540 single_thread_wait(pr); 541 542 atomic_setbits_int(&pr->ps_flags, PS_WAITED); 543 retval[0] = pr->ps_pid; 544 545 if (statusp != NULL) 546 *statusp = W_STOPCODE(pr->ps_single->p_xstat); 547 if (rusage != NULL) 548 memset(rusage, 0, sizeof(*rusage)); 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] = pr->ps_pid; 558 559 if (statusp != NULL) 560 *statusp = W_STOPCODE(p->p_xstat); 561 if (rusage != NULL) 562 memset(rusage, 0, sizeof(*rusage)); 563 return (0); 564 } 565 if ((options & WCONTINUED) && (p->p_flag & P_CONTINUED)) { 566 atomic_clearbits_int(&p->p_flag, P_CONTINUED); 567 retval[0] = pr->ps_pid; 568 569 if (statusp != NULL) 570 *statusp = _WCONTINUED; 571 if (rusage != NULL) 572 memset(rusage, 0, sizeof(*rusage)); 573 return (0); 574 } 575 } 576 if (nfound == 0) 577 return (ECHILD); 578 if (options & WNOHANG) { 579 retval[0] = 0; 580 return (0); 581 } 582 if ((error = tsleep(q->p_p, PWAIT | PCATCH, "wait", 0)) != 0) 583 return (error); 584 goto loop; 585 } 586 587 void 588 proc_finish_wait(struct proc *waiter, struct proc *p) 589 { 590 struct process *pr, *tr; 591 struct rusage *rup; 592 593 /* 594 * If we got the child via a ptrace 'attach', 595 * we need to give it back to the old parent. 596 */ 597 pr = p->p_p; 598 if (pr->ps_oppid && (tr = prfind(pr->ps_oppid))) { 599 atomic_clearbits_int(&pr->ps_flags, PS_TRACED); 600 pr->ps_oppid = 0; 601 proc_reparent(pr, tr); 602 prsignal(tr, SIGCHLD); 603 wakeup(tr); 604 } else { 605 scheduler_wait_hook(waiter, p); 606 p->p_xstat = 0; 607 rup = &waiter->p_p->ps_cru; 608 ruadd(rup, pr->ps_ru); 609 LIST_REMOVE(pr, ps_list); /* off zombprocess */ 610 freepid(pr->ps_pid); 611 process_zap(pr); 612 } 613 } 614 615 /* 616 * make process 'parent' the new parent of process 'child'. 617 */ 618 void 619 proc_reparent(struct process *child, struct process *parent) 620 { 621 622 if (child->ps_pptr == parent) 623 return; 624 625 LIST_REMOVE(child, ps_sibling); 626 LIST_INSERT_HEAD(&parent->ps_children, child, ps_sibling); 627 child->ps_pptr = parent; 628 } 629 630 void 631 process_zap(struct process *pr) 632 { 633 struct vnode *otvp; 634 struct proc *p = pr->ps_mainproc; 635 636 /* 637 * Finally finished with old proc entry. 638 * Unlink it from its process group and free it. 639 */ 640 leavepgrp(pr); 641 LIST_REMOVE(pr, ps_sibling); 642 643 /* 644 * Decrement the count of procs running with this uid. 645 */ 646 (void)chgproccnt(pr->ps_ucred->cr_ruid, -1); 647 648 /* 649 * Release reference to text vnode 650 */ 651 otvp = pr->ps_textvp; 652 pr->ps_textvp = NULL; 653 if (otvp) 654 vrele(otvp); 655 656 KASSERT(pr->ps_refcnt == 1); 657 if (pr->ps_ptstat != NULL) 658 free(pr->ps_ptstat, M_SUBPROC, sizeof(*pr->ps_ptstat)); 659 pool_put(&rusage_pool, pr->ps_ru); 660 KASSERT(TAILQ_EMPTY(&pr->ps_threads)); 661 limfree(pr->ps_limit); 662 crfree(pr->ps_ucred); 663 pool_put(&process_pool, pr); 664 nprocesses--; 665 666 proc_free(p); 667 } 668