1 /* $OpenBSD: kern_exit.c,v 1.172 2019/01/06 12:59:45 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, PUSER, "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; 183 184 #if NKCOV > 0 185 kcov_exit(p); 186 #endif 187 188 if ((p->p_flag & P_THREAD) == 0) { 189 sigio_freelist(&pr->ps_sigiolst); 190 191 /* close open files and release open-file table */ 192 fdfree(p); 193 194 timeout_del(&pr->ps_realit_to); 195 timeout_del(&pr->ps_rucheck_to); 196 #ifdef SYSVSEM 197 semexit(pr); 198 #endif 199 killjobc(pr); 200 #ifdef ACCOUNTING 201 acct_process(p); 202 #endif 203 204 #ifdef KTRACE 205 /* release trace file */ 206 if (pr->ps_tracevp) 207 ktrcleartrace(pr); 208 #endif 209 210 unveil_destroy(pr); 211 212 /* 213 * If parent has the SAS_NOCLDWAIT flag set, we're not 214 * going to become a zombie. 215 */ 216 if (pr->ps_pptr->ps_sigacts->ps_flags & SAS_NOCLDWAIT) 217 atomic_setbits_int(&pr->ps_flags, PS_NOZOMBIE); 218 } 219 220 p->p_fd = NULL; /* zap the thread's copy */ 221 222 /* 223 * Remove proc from pidhash chain and allproc so looking 224 * it up won't work. We will put the proc on the 225 * deadproc list later (using the p_hash member), and 226 * wake up the reaper when we do. If this is the last 227 * thread of a process that isn't PS_NOZOMBIE, we'll put 228 * the process on the zombprocess list below. 229 */ 230 /* 231 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP! 232 */ 233 p->p_stat = SDEAD; 234 235 LIST_REMOVE(p, p_hash); 236 LIST_REMOVE(p, p_list); 237 238 if ((p->p_flag & P_THREAD) == 0) { 239 LIST_REMOVE(pr, ps_hash); 240 LIST_REMOVE(pr, ps_list); 241 242 if ((pr->ps_flags & PS_NOZOMBIE) == 0) 243 LIST_INSERT_HEAD(&zombprocess, pr, ps_list); 244 else { 245 /* 246 * Not going to be a zombie, so it's now off all 247 * the lists scanned by ispidtaken(), so block 248 * fast reuse of the pid now. 249 */ 250 freepid(pr->ps_pid); 251 } 252 253 /* 254 * Give orphaned children to init(8). 255 */ 256 qr = LIST_FIRST(&pr->ps_children); 257 if (qr) /* only need this if any child is S_ZOMB */ 258 wakeup(initprocess); 259 for (; qr != 0; qr = nqr) { 260 nqr = LIST_NEXT(qr, ps_sibling); 261 proc_reparent(qr, initprocess); 262 /* 263 * Traced processes are killed since their 264 * existence means someone is screwing up. 265 */ 266 if (qr->ps_flags & PS_TRACED && 267 !(qr->ps_flags & PS_EXITING)) { 268 atomic_clearbits_int(&qr->ps_flags, PS_TRACED); 269 /* 270 * If single threading is active, 271 * direct the signal to the active 272 * thread to avoid deadlock. 273 */ 274 if (qr->ps_single) 275 ptsignal(qr->ps_single, SIGKILL, 276 STHREAD); 277 else 278 prsignal(qr, SIGKILL); 279 } 280 } 281 } 282 283 /* add thread's accumulated rusage into the process's total */ 284 ruadd(rup, &p->p_ru); 285 tuagg(pr, p); 286 287 /* 288 * clear %cpu usage during swap 289 */ 290 p->p_pctcpu = 0; 291 292 if ((p->p_flag & P_THREAD) == 0) { 293 /* 294 * Final thread has died, so add on our children's rusage 295 * and calculate the total times 296 */ 297 calcru(&pr->ps_tu, &rup->ru_utime, &rup->ru_stime, NULL); 298 ruadd(rup, &pr->ps_cru); 299 300 /* notify interested parties of our demise and clean up */ 301 knote_processexit(p); 302 303 /* 304 * Notify parent that we're gone. If we're not going to 305 * become a zombie, reparent to process 1 (init) so that 306 * we can wake our original parent to possibly unblock 307 * wait4() to return ECHILD. 308 */ 309 if (pr->ps_flags & PS_NOZOMBIE) { 310 struct process *ppr = pr->ps_pptr; 311 proc_reparent(pr, initprocess); 312 wakeup(ppr); 313 } 314 315 /* 316 * Release the process's signal state. 317 */ 318 sigactsfree(pr); 319 } 320 321 /* just a thread? detach it from its process */ 322 if (p->p_flag & P_THREAD) { 323 /* scheduler_wait_hook(pr->ps_mainproc, p); XXX */ 324 if (--pr->ps_refcnt == 1) 325 wakeup(&pr->ps_threads); 326 KASSERT(pr->ps_refcnt > 0); 327 } 328 329 /* 330 * Other substructures are freed from reaper and wait(). 331 */ 332 333 /* 334 * Finally, call machine-dependent code to switch to a new 335 * context (possibly the idle context). Once we are no longer 336 * using the dead process's vmspace and stack, exit2() will be 337 * called to schedule those resources to be released by the 338 * reaper thread. 339 * 340 * Note that cpu_exit() will end with a call equivalent to 341 * cpu_switch(), finishing our execution (pun intended). 342 */ 343 uvmexp.swtch++; 344 cpu_exit(p); 345 panic("cpu_exit returned"); 346 } 347 348 /* 349 * Locking of this proclist is special; it's accessed in a 350 * critical section of process exit, and thus locking it can't 351 * modify interrupt state. We use a simple spin lock for this 352 * proclist. We use the p_hash member to linkup to deadproc. 353 */ 354 struct mutex deadproc_mutex = 355 MUTEX_INITIALIZER_FLAGS(IPL_NONE, NULL, MTX_NOWITNESS); 356 struct proclist deadproc = LIST_HEAD_INITIALIZER(deadproc); 357 358 /* 359 * We are called from cpu_exit() once it is safe to schedule the 360 * dead process's resources to be freed. 361 * 362 * NOTE: One must be careful with locking in this routine. It's 363 * called from a critical section in machine-dependent code, so 364 * we should refrain from changing any interrupt state. 365 * 366 * We lock the deadproc list, place the proc on that list (using 367 * the p_hash member), and wake up the reaper. 368 */ 369 void 370 exit2(struct proc *p) 371 { 372 mtx_enter(&deadproc_mutex); 373 LIST_INSERT_HEAD(&deadproc, p, p_hash); 374 mtx_leave(&deadproc_mutex); 375 376 wakeup(&deadproc); 377 } 378 379 void 380 proc_free(struct proc *p) 381 { 382 crfree(p->p_ucred); 383 pool_put(&proc_pool, p); 384 nthreads--; 385 } 386 387 /* 388 * Process reaper. This is run by a kernel thread to free the resources 389 * of a dead process. Once the resources are free, the process becomes 390 * a zombie, and the parent is allowed to read the undead's status. 391 */ 392 void 393 reaper(void *arg) 394 { 395 struct proc *p; 396 397 KERNEL_UNLOCK(); 398 399 SCHED_ASSERT_UNLOCKED(); 400 401 for (;;) { 402 mtx_enter(&deadproc_mutex); 403 while ((p = LIST_FIRST(&deadproc)) == NULL) 404 msleep(&deadproc, &deadproc_mutex, PVM, "reaper", 0); 405 406 /* Remove us from the deadproc list. */ 407 LIST_REMOVE(p, p_hash); 408 mtx_leave(&deadproc_mutex); 409 410 WITNESS_THREAD_EXIT(p); 411 412 KERNEL_LOCK(); 413 414 /* 415 * Free the VM resources we're still holding on to. 416 * We must do this from a valid thread because doing 417 * so may block. 418 */ 419 uvm_uarea_free(p); 420 p->p_vmspace = NULL; /* zap the thread's copy */ 421 422 if (p->p_flag & P_THREAD) { 423 /* Just a thread */ 424 proc_free(p); 425 } else { 426 struct process *pr = p->p_p; 427 428 /* Release the rest of the process's vmspace */ 429 uvm_exit(pr); 430 431 if ((pr->ps_flags & PS_NOZOMBIE) == 0) { 432 /* Process is now a true zombie. */ 433 atomic_setbits_int(&pr->ps_flags, PS_ZOMBIE); 434 prsignal(pr->ps_pptr, SIGCHLD); 435 436 /* Wake up the parent so it can get exit status. */ 437 wakeup(pr->ps_pptr); 438 } else { 439 /* No one will wait for us. Just zap the process now */ 440 process_zap(pr); 441 } 442 } 443 444 KERNEL_UNLOCK(); 445 } 446 } 447 448 int 449 sys_wait4(struct proc *q, void *v, register_t *retval) 450 { 451 struct sys_wait4_args /* { 452 syscallarg(pid_t) pid; 453 syscallarg(int *) status; 454 syscallarg(int) options; 455 syscallarg(struct rusage *) rusage; 456 } */ *uap = v; 457 struct rusage ru; 458 int status, error; 459 460 error = dowait4(q, SCARG(uap, pid), 461 SCARG(uap, status) ? &status : NULL, 462 SCARG(uap, options), SCARG(uap, rusage) ? &ru : NULL, retval); 463 if (error == 0 && retval[0] > 0 && SCARG(uap, status)) { 464 error = copyout(&status, SCARG(uap, status), sizeof(status)); 465 } 466 if (error == 0 && retval[0] > 0 && SCARG(uap, rusage)) { 467 error = copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 468 #ifdef KTRACE 469 if (error == 0 && KTRPOINT(q, KTR_STRUCT)) 470 ktrrusage(q, &ru); 471 #endif 472 } 473 return (error); 474 } 475 476 int 477 dowait4(struct proc *q, pid_t pid, int *statusp, int options, 478 struct rusage *rusage, register_t *retval) 479 { 480 int nfound; 481 struct process *pr; 482 struct proc *p; 483 int error; 484 485 if (pid == 0) 486 pid = -q->p_p->ps_pgid; 487 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED)) 488 return (EINVAL); 489 490 loop: 491 nfound = 0; 492 LIST_FOREACH(pr, &q->p_p->ps_children, ps_sibling) { 493 if ((pr->ps_flags & PS_NOZOMBIE) || 494 (pid != WAIT_ANY && 495 pr->ps_pid != pid && 496 pr->ps_pgid != -pid)) 497 continue; 498 499 p = pr->ps_mainproc; 500 501 nfound++; 502 if (pr->ps_flags & PS_ZOMBIE) { 503 retval[0] = pr->ps_pid; 504 505 if (statusp != NULL) 506 *statusp = p->p_xstat; /* convert to int */ 507 if (rusage != NULL) 508 memcpy(rusage, pr->ps_ru, sizeof(*rusage)); 509 proc_finish_wait(q, p); 510 return (0); 511 } 512 if (pr->ps_flags & PS_TRACED && 513 (pr->ps_flags & PS_WAITED) == 0 && pr->ps_single && 514 pr->ps_single->p_stat == SSTOP && 515 (pr->ps_single->p_flag & P_SUSPSINGLE) == 0) { 516 single_thread_wait(pr); 517 518 atomic_setbits_int(&pr->ps_flags, PS_WAITED); 519 retval[0] = pr->ps_pid; 520 521 if (statusp != NULL) 522 *statusp = W_STOPCODE(pr->ps_single->p_xstat); 523 if (rusage != NULL) 524 memset(rusage, 0, sizeof(*rusage)); 525 return (0); 526 } 527 if (p->p_stat == SSTOP && 528 (pr->ps_flags & PS_WAITED) == 0 && 529 (p->p_flag & P_SUSPSINGLE) == 0 && 530 (pr->ps_flags & PS_TRACED || 531 options & WUNTRACED)) { 532 atomic_setbits_int(&pr->ps_flags, PS_WAITED); 533 retval[0] = pr->ps_pid; 534 535 if (statusp != NULL) 536 *statusp = W_STOPCODE(p->p_xstat); 537 if (rusage != NULL) 538 memset(rusage, 0, sizeof(*rusage)); 539 return (0); 540 } 541 if ((options & WCONTINUED) && (p->p_flag & P_CONTINUED)) { 542 atomic_clearbits_int(&p->p_flag, P_CONTINUED); 543 retval[0] = pr->ps_pid; 544 545 if (statusp != NULL) 546 *statusp = _WCONTINUED; 547 if (rusage != NULL) 548 memset(rusage, 0, sizeof(*rusage)); 549 return (0); 550 } 551 } 552 if (nfound == 0) 553 return (ECHILD); 554 if (options & WNOHANG) { 555 retval[0] = 0; 556 return (0); 557 } 558 if ((error = tsleep(q->p_p, PWAIT | PCATCH, "wait", 0)) != 0) 559 return (error); 560 goto loop; 561 } 562 563 void 564 proc_finish_wait(struct proc *waiter, struct proc *p) 565 { 566 struct process *pr, *tr; 567 struct rusage *rup; 568 569 /* 570 * If we got the child via a ptrace 'attach', 571 * we need to give it back to the old parent. 572 */ 573 pr = p->p_p; 574 if (pr->ps_oppid && (tr = prfind(pr->ps_oppid))) { 575 atomic_clearbits_int(&pr->ps_flags, PS_TRACED); 576 pr->ps_oppid = 0; 577 proc_reparent(pr, tr); 578 prsignal(tr, SIGCHLD); 579 wakeup(tr); 580 } else { 581 scheduler_wait_hook(waiter, p); 582 p->p_xstat = 0; 583 rup = &waiter->p_p->ps_cru; 584 ruadd(rup, pr->ps_ru); 585 LIST_REMOVE(pr, ps_list); /* off zombprocess */ 586 freepid(pr->ps_pid); 587 process_zap(pr); 588 } 589 } 590 591 /* 592 * make process 'parent' the new parent of process 'child'. 593 */ 594 void 595 proc_reparent(struct process *child, struct process *parent) 596 { 597 598 if (child->ps_pptr == parent) 599 return; 600 601 LIST_REMOVE(child, ps_sibling); 602 LIST_INSERT_HEAD(&parent->ps_children, child, ps_sibling); 603 child->ps_pptr = parent; 604 } 605 606 void 607 process_zap(struct process *pr) 608 { 609 struct vnode *otvp; 610 struct proc *p = pr->ps_mainproc; 611 612 /* 613 * Finally finished with old proc entry. 614 * Unlink it from its process group and free it. 615 */ 616 leavepgrp(pr); 617 LIST_REMOVE(pr, ps_sibling); 618 619 /* 620 * Decrement the count of procs running with this uid. 621 */ 622 (void)chgproccnt(pr->ps_ucred->cr_ruid, -1); 623 624 /* 625 * Release reference to text vnode 626 */ 627 otvp = pr->ps_textvp; 628 pr->ps_textvp = NULL; 629 if (otvp) 630 vrele(otvp); 631 632 KASSERT(pr->ps_refcnt == 1); 633 if (pr->ps_ptstat != NULL) 634 free(pr->ps_ptstat, M_SUBPROC, sizeof(*pr->ps_ptstat)); 635 pool_put(&rusage_pool, pr->ps_ru); 636 KASSERT(TAILQ_EMPTY(&pr->ps_threads)); 637 limfree(pr->ps_limit); 638 crfree(pr->ps_ucred); 639 pool_put(&process_pool, pr); 640 nprocesses--; 641 642 proc_free(p); 643 } 644