1 /* $OpenBSD: kern_exit.c,v 1.178 2019/06/21 09:39:48 visa 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/vnode.h> 51 #include <sys/syslog.h> 52 #include <sys/malloc.h> 53 #include <sys/resourcevar.h> 54 #include <sys/ptrace.h> 55 #include <sys/acct.h> 56 #include <sys/filedesc.h> 57 #include <sys/signalvar.h> 58 #include <sys/sched.h> 59 #include <sys/ktrace.h> 60 #include <sys/pool.h> 61 #include <sys/mutex.h> 62 #include <sys/pledge.h> 63 #ifdef SYSVSEM 64 #include <sys/sem.h> 65 #endif 66 #include <sys/witness.h> 67 68 #include <sys/mount.h> 69 #include <sys/syscallargs.h> 70 71 #include <uvm/uvm_extern.h> 72 73 #include "kcov.h" 74 #if NKCOV > 0 75 #include <sys/kcov.h> 76 #endif 77 78 void proc_finish_wait(struct proc *, struct proc *); 79 void process_zap(struct process *); 80 void proc_free(struct proc *); 81 void unveil_destroy(struct process *ps); 82 83 /* 84 * exit -- 85 * Death of process. 86 */ 87 int 88 sys_exit(struct proc *p, void *v, register_t *retval) 89 { 90 struct sys_exit_args /* { 91 syscallarg(int) rval; 92 } */ *uap = v; 93 94 exit1(p, W_EXITCODE(SCARG(uap, rval), 0), EXIT_NORMAL); 95 /* NOTREACHED */ 96 return (0); 97 } 98 99 int 100 sys___threxit(struct proc *p, void *v, register_t *retval) 101 { 102 struct sys___threxit_args /* { 103 syscallarg(pid_t *) notdead; 104 } */ *uap = v; 105 106 if (SCARG(uap, notdead) != NULL) { 107 pid_t zero = 0; 108 if (copyout(&zero, SCARG(uap, notdead), sizeof(zero))) 109 psignal(p, SIGSEGV); 110 } 111 exit1(p, 0, EXIT_THREAD); 112 113 return (0); 114 } 115 116 /* 117 * Exit: deallocate address space and other resources, change proc state 118 * to zombie, and unlink proc from allproc and parent's lists. Save exit 119 * status and rusage for wait(). Check for child processes and orphan them. 120 */ 121 void 122 exit1(struct proc *p, int rv, int flags) 123 { 124 struct process *pr, *qr, *nqr; 125 struct rusage *rup; 126 127 atomic_setbits_int(&p->p_flag, P_WEXIT); 128 129 pr = p->p_p; 130 131 /* single-threaded? */ 132 if (!P_HASSIBLING(p)) { 133 flags = EXIT_NORMAL; 134 } else { 135 /* nope, multi-threaded */ 136 if (flags == EXIT_NORMAL) 137 single_thread_set(p, SINGLE_EXIT, 0); 138 else if (flags == EXIT_THREAD) 139 single_thread_check(p, 0); 140 } 141 142 if (flags == EXIT_NORMAL) { 143 if (pr->ps_pid == 1) 144 panic("init died (signal %d, exit %d)", 145 WTERMSIG(rv), WEXITSTATUS(rv)); 146 147 atomic_setbits_int(&pr->ps_flags, PS_EXITING); 148 pr->ps_mainproc->p_xstat = rv; 149 150 /* 151 * If parent is waiting for us to exit or exec, PS_PPWAIT 152 * is set; we wake up the parent early to avoid deadlock. 153 */ 154 if (pr->ps_flags & PS_PPWAIT) { 155 atomic_clearbits_int(&pr->ps_flags, PS_PPWAIT); 156 atomic_clearbits_int(&pr->ps_pptr->ps_flags, 157 PS_ISPWAIT); 158 wakeup(pr->ps_pptr); 159 } 160 } 161 162 /* unlink ourselves from the active threads */ 163 TAILQ_REMOVE(&pr->ps_threads, p, p_thr_link); 164 if ((p->p_flag & P_THREAD) == 0) { 165 /* main thread gotta wait because it has the pid, et al */ 166 while (pr->ps_refcnt > 1) 167 tsleep(&pr->ps_threads, PWAIT, "thrdeath", 0); 168 if (pr->ps_flags & PS_PROFIL) 169 stopprofclock(pr); 170 } 171 172 rup = pr->ps_ru; 173 if (rup == NULL) { 174 rup = pool_get(&rusage_pool, PR_WAITOK | PR_ZERO); 175 if (pr->ps_ru == NULL) { 176 pr->ps_ru = rup; 177 } else { 178 pool_put(&rusage_pool, rup); 179 rup = pr->ps_ru; 180 } 181 } 182 p->p_siglist = 0; 183 if ((p->p_flag & P_THREAD) == 0) 184 pr->ps_siglist = 0; 185 186 #if NKCOV > 0 187 kcov_exit(p); 188 #endif 189 190 if ((p->p_flag & P_THREAD) == 0) { 191 sigio_freelist(&pr->ps_sigiolst); 192 193 /* close open files and release open-file table */ 194 fdfree(p); 195 196 timeout_del(&pr->ps_realit_to); 197 timeout_del(&pr->ps_rucheck_to); 198 #ifdef SYSVSEM 199 semexit(pr); 200 #endif 201 killjobc(pr); 202 #ifdef ACCOUNTING 203 acct_process(p); 204 #endif 205 206 #ifdef KTRACE 207 /* release trace file */ 208 if (pr->ps_tracevp) 209 ktrcleartrace(pr); 210 #endif 211 212 unveil_destroy(pr); 213 214 /* 215 * If parent has the SAS_NOCLDWAIT flag set, we're not 216 * going to become a zombie. 217 */ 218 if (pr->ps_pptr->ps_sigacts->ps_flags & SAS_NOCLDWAIT) 219 atomic_setbits_int(&pr->ps_flags, PS_NOZOMBIE); 220 } 221 222 p->p_fd = NULL; /* zap the thread's copy */ 223 224 /* 225 * Remove proc from pidhash chain and allproc so looking 226 * it up won't work. We will put the proc on the 227 * deadproc list later (using the p_hash member), and 228 * wake up the reaper when we do. If this is the last 229 * thread of a process that isn't PS_NOZOMBIE, we'll put 230 * the process on the zombprocess list below. 231 */ 232 /* 233 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP! 234 */ 235 p->p_stat = SDEAD; 236 237 LIST_REMOVE(p, p_hash); 238 LIST_REMOVE(p, p_list); 239 240 if ((p->p_flag & P_THREAD) == 0) { 241 LIST_REMOVE(pr, ps_hash); 242 LIST_REMOVE(pr, ps_list); 243 244 if ((pr->ps_flags & PS_NOZOMBIE) == 0) 245 LIST_INSERT_HEAD(&zombprocess, pr, ps_list); 246 else { 247 /* 248 * Not going to be a zombie, so it's now off all 249 * the lists scanned by ispidtaken(), so block 250 * fast reuse of the pid now. 251 */ 252 freepid(pr->ps_pid); 253 } 254 255 /* 256 * Give orphaned children to init(8). 257 */ 258 qr = LIST_FIRST(&pr->ps_children); 259 if (qr) /* only need this if any child is S_ZOMB */ 260 wakeup(initprocess); 261 for (; qr != 0; qr = nqr) { 262 nqr = LIST_NEXT(qr, ps_sibling); 263 proc_reparent(qr, initprocess); 264 /* 265 * Traced processes are killed since their 266 * existence means someone is screwing up. 267 */ 268 if (qr->ps_flags & PS_TRACED && 269 !(qr->ps_flags & PS_EXITING)) { 270 atomic_clearbits_int(&qr->ps_flags, PS_TRACED); 271 /* 272 * If single threading is active, 273 * direct the signal to the active 274 * thread to avoid deadlock. 275 */ 276 if (qr->ps_single) 277 ptsignal(qr->ps_single, SIGKILL, 278 STHREAD); 279 else 280 prsignal(qr, SIGKILL); 281 } 282 } 283 } 284 285 /* add thread's accumulated rusage into the process's total */ 286 ruadd(rup, &p->p_ru); 287 tuagg(pr, p); 288 289 /* 290 * clear %cpu usage during swap 291 */ 292 p->p_pctcpu = 0; 293 294 if ((p->p_flag & P_THREAD) == 0) { 295 /* 296 * Final thread has died, so add on our children's rusage 297 * and calculate the total times 298 */ 299 calcru(&pr->ps_tu, &rup->ru_utime, &rup->ru_stime, NULL); 300 ruadd(rup, &pr->ps_cru); 301 302 /* notify interested parties of our demise and clean up */ 303 knote_processexit(p); 304 305 /* 306 * Notify parent that we're gone. If we're not going to 307 * become a zombie, reparent to process 1 (init) so that 308 * we can wake our original parent to possibly unblock 309 * wait4() to return ECHILD. 310 */ 311 if (pr->ps_flags & PS_NOZOMBIE) { 312 struct process *ppr = pr->ps_pptr; 313 proc_reparent(pr, initprocess); 314 wakeup(ppr); 315 } 316 317 /* 318 * Release the process's signal state. 319 */ 320 sigactsfree(pr); 321 } 322 323 /* just a thread? detach it from its process */ 324 if (p->p_flag & P_THREAD) { 325 /* scheduler_wait_hook(pr->ps_mainproc, p); XXX */ 326 if (--pr->ps_refcnt == 1) 327 wakeup(&pr->ps_threads); 328 KASSERT(pr->ps_refcnt > 0); 329 } 330 331 /* Release the thread's read reference of resource limit structure. */ 332 if (p->p_limit != NULL) { 333 struct plimit *limit; 334 335 limit = p->p_limit; 336 p->p_limit = NULL; 337 lim_free(limit); 338 } 339 340 /* 341 * Other substructures are freed from reaper and wait(). 342 */ 343 344 /* 345 * Finally, call machine-dependent code to switch to a new 346 * context (possibly the idle context). Once we are no longer 347 * using the dead process's vmspace and stack, exit2() will be 348 * called to schedule those resources to be released by the 349 * reaper thread. 350 * 351 * Note that cpu_exit() will end with a call equivalent to 352 * cpu_switch(), finishing our execution (pun intended). 353 */ 354 uvmexp.swtch++; 355 cpu_exit(p); 356 panic("cpu_exit returned"); 357 } 358 359 /* 360 * Locking of this proclist is special; it's accessed in a 361 * critical section of process exit, and thus locking it can't 362 * modify interrupt state. We use a simple spin lock for this 363 * proclist. We use the p_hash member to linkup to deadproc. 364 */ 365 struct mutex deadproc_mutex = 366 MUTEX_INITIALIZER_FLAGS(IPL_NONE, "deadproc", MTX_NOWITNESS); 367 struct proclist deadproc = LIST_HEAD_INITIALIZER(deadproc); 368 369 /* 370 * We are called from cpu_exit() once it is safe to schedule the 371 * dead process's resources to be freed. 372 * 373 * NOTE: One must be careful with locking in this routine. It's 374 * called from a critical section in machine-dependent code, so 375 * we should refrain from changing any interrupt state. 376 * 377 * We lock the deadproc list, place the proc on that list (using 378 * the p_hash member), and wake up the reaper. 379 */ 380 void 381 exit2(struct proc *p) 382 { 383 mtx_enter(&deadproc_mutex); 384 LIST_INSERT_HEAD(&deadproc, p, p_hash); 385 mtx_leave(&deadproc_mutex); 386 387 wakeup(&deadproc); 388 } 389 390 void 391 proc_free(struct proc *p) 392 { 393 crfree(p->p_ucred); 394 pool_put(&proc_pool, p); 395 nthreads--; 396 } 397 398 /* 399 * Process reaper. This is run by a kernel thread to free the resources 400 * of a dead process. Once the resources are free, the process becomes 401 * a zombie, and the parent is allowed to read the undead's status. 402 */ 403 void 404 reaper(void *arg) 405 { 406 struct proc *p; 407 408 KERNEL_UNLOCK(); 409 410 SCHED_ASSERT_UNLOCKED(); 411 412 for (;;) { 413 mtx_enter(&deadproc_mutex); 414 while ((p = LIST_FIRST(&deadproc)) == NULL) 415 msleep(&deadproc, &deadproc_mutex, PVM, "reaper", 0); 416 417 /* Remove us from the deadproc list. */ 418 LIST_REMOVE(p, p_hash); 419 mtx_leave(&deadproc_mutex); 420 421 WITNESS_THREAD_EXIT(p); 422 423 KERNEL_LOCK(); 424 425 /* 426 * Free the VM resources we're still holding on to. 427 * We must do this from a valid thread because doing 428 * so may block. 429 */ 430 uvm_uarea_free(p); 431 p->p_vmspace = NULL; /* zap the thread's copy */ 432 433 if (p->p_flag & P_THREAD) { 434 /* Just a thread */ 435 proc_free(p); 436 } else { 437 struct process *pr = p->p_p; 438 439 /* Release the rest of the process's vmspace */ 440 uvm_exit(pr); 441 442 if ((pr->ps_flags & PS_NOZOMBIE) == 0) { 443 /* Process is now a true zombie. */ 444 atomic_setbits_int(&pr->ps_flags, PS_ZOMBIE); 445 prsignal(pr->ps_pptr, SIGCHLD); 446 447 /* Wake up the parent so it can get exit status. */ 448 wakeup(pr->ps_pptr); 449 } else { 450 /* No one will wait for us. Just zap the process now */ 451 process_zap(pr); 452 } 453 } 454 455 KERNEL_UNLOCK(); 456 } 457 } 458 459 int 460 sys_wait4(struct proc *q, void *v, register_t *retval) 461 { 462 struct sys_wait4_args /* { 463 syscallarg(pid_t) pid; 464 syscallarg(int *) status; 465 syscallarg(int) options; 466 syscallarg(struct rusage *) rusage; 467 } */ *uap = v; 468 struct rusage ru; 469 int status, error; 470 471 error = dowait4(q, SCARG(uap, pid), 472 SCARG(uap, status) ? &status : NULL, 473 SCARG(uap, options), SCARG(uap, rusage) ? &ru : NULL, retval); 474 if (error == 0 && retval[0] > 0 && SCARG(uap, status)) { 475 error = copyout(&status, SCARG(uap, status), sizeof(status)); 476 } 477 if (error == 0 && retval[0] > 0 && SCARG(uap, rusage)) { 478 error = copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 479 #ifdef KTRACE 480 if (error == 0 && KTRPOINT(q, KTR_STRUCT)) 481 ktrrusage(q, &ru); 482 #endif 483 } 484 return (error); 485 } 486 487 int 488 dowait4(struct proc *q, pid_t pid, int *statusp, int options, 489 struct rusage *rusage, register_t *retval) 490 { 491 int nfound; 492 struct process *pr; 493 struct proc *p; 494 int error; 495 496 if (pid == 0) 497 pid = -q->p_p->ps_pgid; 498 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED)) 499 return (EINVAL); 500 501 loop: 502 nfound = 0; 503 LIST_FOREACH(pr, &q->p_p->ps_children, ps_sibling) { 504 if ((pr->ps_flags & PS_NOZOMBIE) || 505 (pid != WAIT_ANY && 506 pr->ps_pid != pid && 507 pr->ps_pgid != -pid)) 508 continue; 509 510 p = pr->ps_mainproc; 511 512 nfound++; 513 if (pr->ps_flags & PS_ZOMBIE) { 514 retval[0] = pr->ps_pid; 515 516 if (statusp != NULL) 517 *statusp = p->p_xstat; /* convert to int */ 518 if (rusage != NULL) 519 memcpy(rusage, pr->ps_ru, sizeof(*rusage)); 520 proc_finish_wait(q, p); 521 return (0); 522 } 523 if (pr->ps_flags & PS_TRACED && 524 (pr->ps_flags & PS_WAITED) == 0 && pr->ps_single && 525 pr->ps_single->p_stat == SSTOP && 526 (pr->ps_single->p_flag & P_SUSPSINGLE) == 0) { 527 single_thread_wait(pr); 528 529 atomic_setbits_int(&pr->ps_flags, PS_WAITED); 530 retval[0] = pr->ps_pid; 531 532 if (statusp != NULL) 533 *statusp = W_STOPCODE(pr->ps_single->p_xstat); 534 if (rusage != NULL) 535 memset(rusage, 0, sizeof(*rusage)); 536 return (0); 537 } 538 if (p->p_stat == SSTOP && 539 (pr->ps_flags & PS_WAITED) == 0 && 540 (p->p_flag & P_SUSPSINGLE) == 0 && 541 (pr->ps_flags & PS_TRACED || 542 options & WUNTRACED)) { 543 atomic_setbits_int(&pr->ps_flags, PS_WAITED); 544 retval[0] = pr->ps_pid; 545 546 if (statusp != NULL) 547 *statusp = W_STOPCODE(p->p_xstat); 548 if (rusage != NULL) 549 memset(rusage, 0, sizeof(*rusage)); 550 return (0); 551 } 552 if ((options & WCONTINUED) && (p->p_flag & P_CONTINUED)) { 553 atomic_clearbits_int(&p->p_flag, P_CONTINUED); 554 retval[0] = pr->ps_pid; 555 556 if (statusp != NULL) 557 *statusp = _WCONTINUED; 558 if (rusage != NULL) 559 memset(rusage, 0, sizeof(*rusage)); 560 return (0); 561 } 562 } 563 if (nfound == 0) 564 return (ECHILD); 565 if (options & WNOHANG) { 566 retval[0] = 0; 567 return (0); 568 } 569 if ((error = tsleep(q->p_p, PWAIT | PCATCH, "wait", 0)) != 0) 570 return (error); 571 goto loop; 572 } 573 574 void 575 proc_finish_wait(struct proc *waiter, struct proc *p) 576 { 577 struct process *pr, *tr; 578 struct rusage *rup; 579 580 /* 581 * If we got the child via a ptrace 'attach', 582 * we need to give it back to the old parent. 583 */ 584 pr = p->p_p; 585 if (pr->ps_oppid && (tr = prfind(pr->ps_oppid))) { 586 atomic_clearbits_int(&pr->ps_flags, PS_TRACED); 587 pr->ps_oppid = 0; 588 proc_reparent(pr, tr); 589 prsignal(tr, SIGCHLD); 590 wakeup(tr); 591 } else { 592 scheduler_wait_hook(waiter, p); 593 p->p_xstat = 0; 594 rup = &waiter->p_p->ps_cru; 595 ruadd(rup, pr->ps_ru); 596 LIST_REMOVE(pr, ps_list); /* off zombprocess */ 597 freepid(pr->ps_pid); 598 process_zap(pr); 599 } 600 } 601 602 /* 603 * make process 'parent' the new parent of process 'child'. 604 */ 605 void 606 proc_reparent(struct process *child, struct process *parent) 607 { 608 609 if (child->ps_pptr == parent) 610 return; 611 612 LIST_REMOVE(child, ps_sibling); 613 LIST_INSERT_HEAD(&parent->ps_children, child, ps_sibling); 614 child->ps_pptr = parent; 615 } 616 617 void 618 process_zap(struct process *pr) 619 { 620 struct vnode *otvp; 621 struct proc *p = pr->ps_mainproc; 622 623 /* 624 * Finally finished with old proc entry. 625 * Unlink it from its process group and free it. 626 */ 627 leavepgrp(pr); 628 LIST_REMOVE(pr, ps_sibling); 629 630 /* 631 * Decrement the count of procs running with this uid. 632 */ 633 (void)chgproccnt(pr->ps_ucred->cr_ruid, -1); 634 635 /* 636 * Release reference to text vnode 637 */ 638 otvp = pr->ps_textvp; 639 pr->ps_textvp = NULL; 640 if (otvp) 641 vrele(otvp); 642 643 KASSERT(pr->ps_refcnt == 1); 644 if (pr->ps_ptstat != NULL) 645 free(pr->ps_ptstat, M_SUBPROC, sizeof(*pr->ps_ptstat)); 646 pool_put(&rusage_pool, pr->ps_ru); 647 KASSERT(TAILQ_EMPTY(&pr->ps_threads)); 648 lim_free(pr->ps_limit); 649 crfree(pr->ps_ucred); 650 pool_put(&process_pool, pr); 651 nprocesses--; 652 653 proc_free(p); 654 } 655