1 /* $NetBSD: kern_exit.c,v 1.221 2009/05/08 13:32:59 yamt Exp $ */ 2 3 /*- 4 * Copyright (c) 1998, 1999, 2006, 2007, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility, 9 * NASA Ames Research Center, and by Andrew Doran. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1989, 1991, 1993 35 * The Regents of the University of California. All rights reserved. 36 * (c) UNIX System Laboratories, Inc. 37 * All or some portions of this file are derived from material licensed 38 * to the University of California by American Telephone and Telegraph 39 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 40 * the permission of UNIX System Laboratories, Inc. 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 3. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * @(#)kern_exit.c 8.10 (Berkeley) 2/23/95 67 */ 68 69 #include <sys/cdefs.h> 70 __KERNEL_RCSID(0, "$NetBSD: kern_exit.c,v 1.221 2009/05/08 13:32:59 yamt Exp $"); 71 72 #include "opt_ktrace.h" 73 #include "opt_perfctrs.h" 74 #include "opt_sa.h" 75 #include "opt_sysv.h" 76 77 #include <sys/param.h> 78 #include <sys/systm.h> 79 #include <sys/ioctl.h> 80 #include <sys/tty.h> 81 #include <sys/time.h> 82 #include <sys/resource.h> 83 #include <sys/kernel.h> 84 #include <sys/proc.h> 85 #include <sys/buf.h> 86 #include <sys/wait.h> 87 #include <sys/file.h> 88 #include <sys/vnode.h> 89 #include <sys/syslog.h> 90 #include <sys/pool.h> 91 #include <sys/uidinfo.h> 92 #if defined(PERFCTRS) 93 #include <sys/pmc.h> 94 #endif 95 #include <sys/ptrace.h> 96 #include <sys/acct.h> 97 #include <sys/filedesc.h> 98 #include <sys/ras.h> 99 #include <sys/signalvar.h> 100 #include <sys/sched.h> 101 #include <sys/sa.h> 102 #include <sys/savar.h> 103 #include <sys/mount.h> 104 #include <sys/syscallargs.h> 105 #include <sys/kauth.h> 106 #include <sys/sleepq.h> 107 #include <sys/lockdebug.h> 108 #include <sys/ktrace.h> 109 #include <sys/cpu.h> 110 #include <sys/lwpctl.h> 111 #include <sys/atomic.h> 112 113 #include <uvm/uvm_extern.h> 114 115 #ifdef DEBUG_EXIT 116 int debug_exit = 0; 117 #define DPRINTF(x) if (debug_exit) printf x 118 #else 119 #define DPRINTF(x) 120 #endif 121 122 static int find_stopped_child(struct proc *, pid_t, int, struct proc **, int *); 123 static void proc_free(struct proc *, struct rusage *); 124 125 /* 126 * Fill in the appropriate signal information, and signal the parent. 127 */ 128 static void 129 exit_psignal(struct proc *p, struct proc *pp, ksiginfo_t *ksi) 130 { 131 132 KSI_INIT(ksi); 133 if ((ksi->ksi_signo = P_EXITSIG(p)) == SIGCHLD) { 134 if (WIFSIGNALED(p->p_xstat)) { 135 if (WCOREDUMP(p->p_xstat)) 136 ksi->ksi_code = CLD_DUMPED; 137 else 138 ksi->ksi_code = CLD_KILLED; 139 } else { 140 ksi->ksi_code = CLD_EXITED; 141 } 142 } 143 /* 144 * We fill those in, even for non-SIGCHLD. 145 * It's safe to access p->p_cred unlocked here. 146 */ 147 ksi->ksi_pid = p->p_pid; 148 ksi->ksi_uid = kauth_cred_geteuid(p->p_cred); 149 ksi->ksi_status = p->p_xstat; 150 /* XXX: is this still valid? */ 151 ksi->ksi_utime = p->p_stats->p_ru.ru_utime.tv_sec; 152 ksi->ksi_stime = p->p_stats->p_ru.ru_stime.tv_sec; 153 } 154 155 /* 156 * exit -- 157 * Death of process. 158 */ 159 int 160 sys_exit(struct lwp *l, const struct sys_exit_args *uap, register_t *retval) 161 { 162 /* { 163 syscallarg(int) rval; 164 } */ 165 struct proc *p = l->l_proc; 166 167 /* Don't call exit1() multiple times in the same process. */ 168 mutex_enter(p->p_lock); 169 if (p->p_sflag & PS_WEXIT) { 170 mutex_exit(p->p_lock); 171 lwp_exit(l); 172 } 173 174 /* exit1() will release the mutex. */ 175 exit1(l, W_EXITCODE(SCARG(uap, rval), 0)); 176 /* NOTREACHED */ 177 return (0); 178 } 179 180 /* 181 * Exit: deallocate address space and other resources, change proc state 182 * to zombie, and unlink proc from allproc and parent's lists. Save exit 183 * status and rusage for wait(). Check for child processes and orphan them. 184 * 185 * Must be called with p->p_lock held. Does not return. 186 */ 187 void 188 exit1(struct lwp *l, int rv) 189 { 190 struct proc *p, *q, *nq; 191 struct pgrp *pgrp; 192 ksiginfo_t ksi; 193 ksiginfoq_t kq; 194 int wakeinit, sa; 195 196 p = l->l_proc; 197 198 KASSERT(mutex_owned(p->p_lock)); 199 200 if (__predict_false(p == initproc)) 201 panic("init died (signal %d, exit %d)", 202 WTERMSIG(rv), WEXITSTATUS(rv)); 203 204 /* 205 * Disable scheduler activation upcalls. We're trying to get out of 206 * here. 207 */ 208 sa = 0; 209 #ifdef KERN_SA 210 if ((p->p_sa != NULL)) { 211 l->l_pflag |= LP_SA_NOBLOCK; 212 sa = 1; 213 } 214 #endif 215 216 p->p_sflag |= PS_WEXIT; 217 218 /* 219 * Force all other LWPs to exit before we do. Only then can we 220 * begin to tear down the rest of the process state. 221 */ 222 if (sa || p->p_nlwps > 1) 223 exit_lwps(l); 224 225 ksiginfo_queue_init(&kq); 226 227 /* 228 * If we have been asked to stop on exit, do so now. 229 */ 230 if (__predict_false(p->p_sflag & PS_STOPEXIT)) { 231 KERNEL_UNLOCK_ALL(l, &l->l_biglocks); 232 sigclearall(p, &contsigmask, &kq); 233 p->p_waited = 0; 234 membar_producer(); 235 p->p_stat = SSTOP; 236 lwp_lock(l); 237 p->p_nrlwps--; 238 l->l_stat = LSSTOP; 239 mutex_exit(p->p_lock); 240 mi_switch(l); 241 KERNEL_LOCK(l->l_biglocks, l); 242 mutex_enter(p->p_lock); 243 } 244 245 /* 246 * Bin any remaining signals and mark the process as dying so it will 247 * not be found for, e.g. signals. 248 */ 249 sigfillset(&p->p_sigctx.ps_sigignore); 250 sigclearall(p, NULL, &kq); 251 p->p_stat = SDYING; 252 mutex_exit(p->p_lock); 253 ksiginfo_queue_drain(&kq); 254 255 /* Destroy any lwpctl info. */ 256 if (p->p_lwpctl != NULL) 257 lwp_ctl_exit(); 258 259 /* 260 * Drain all remaining references that procfs, ptrace and others may 261 * have on the process. 262 */ 263 rw_enter(&p->p_reflock, RW_WRITER); 264 265 DPRINTF(("exit1: %d.%d exiting.\n", p->p_pid, l->l_lid)); 266 267 timers_free(p, TIMERS_ALL); 268 #if defined(__HAVE_RAS) 269 ras_purgeall(); 270 #endif 271 272 /* 273 * Close open files, release open-file table and free signal 274 * actions. This may block! 275 */ 276 fd_free(); 277 cwdfree(p->p_cwdi); 278 p->p_cwdi = NULL; 279 doexithooks(p); 280 sigactsfree(p->p_sigacts); 281 282 /* 283 * Write out accounting data. 284 */ 285 (void)acct_process(l); 286 287 #ifdef KTRACE 288 /* 289 * Release trace file. 290 */ 291 if (p->p_tracep != NULL) { 292 mutex_enter(&ktrace_lock); 293 ktrderef(p); 294 mutex_exit(&ktrace_lock); 295 } 296 #endif 297 298 /* 299 * If emulation has process exit hook, call it now. 300 * Set the exit status now so that the exit hook has 301 * an opportunity to tweak it (COMPAT_LINUX requires 302 * this for thread group emulation) 303 */ 304 p->p_xstat = rv; 305 if (p->p_emul->e_proc_exit) 306 (*p->p_emul->e_proc_exit)(p); 307 308 /* 309 * Free the VM resources we're still holding on to. 310 * We must do this from a valid thread because doing 311 * so may block. This frees vmspace, which we don't 312 * need anymore. The only remaining lwp is the one 313 * we run at this moment, nothing runs in userland 314 * anymore. 315 */ 316 uvm_proc_exit(p); 317 318 /* 319 * Stop profiling. 320 */ 321 if (__predict_false((p->p_stflag & PST_PROFIL) != 0)) { 322 mutex_spin_enter(&p->p_stmutex); 323 stopprofclock(p); 324 mutex_spin_exit(&p->p_stmutex); 325 } 326 327 /* 328 * If parent is waiting for us to exit or exec, PL_PPWAIT is set; we 329 * wake up the parent early to avoid deadlock. We can do this once 330 * the VM resources are released. 331 */ 332 mutex_enter(proc_lock); 333 if (p->p_lflag & PL_PPWAIT) { 334 p->p_lflag &= ~PL_PPWAIT; 335 cv_broadcast(&p->p_pptr->p_waitcv); 336 } 337 338 if (SESS_LEADER(p)) { 339 struct vnode *vprele = NULL, *vprevoke = NULL; 340 struct session *sp = p->p_session; 341 struct tty *tp; 342 343 if (sp->s_ttyvp) { 344 /* 345 * Controlling process. 346 * Signal foreground pgrp, 347 * drain controlling terminal 348 * and revoke access to controlling terminal. 349 */ 350 tp = sp->s_ttyp; 351 mutex_spin_enter(&tty_lock); 352 if (tp->t_session == sp) { 353 /* we can't guarantee the revoke will do this */ 354 pgrp = tp->t_pgrp; 355 tp->t_pgrp = NULL; 356 tp->t_session = NULL; 357 mutex_spin_exit(&tty_lock); 358 if (pgrp != NULL) { 359 pgsignal(pgrp, SIGHUP, 1); 360 } 361 mutex_exit(proc_lock); 362 (void) ttywait(tp); 363 mutex_enter(proc_lock); 364 365 /* The tty could have been revoked. */ 366 vprevoke = sp->s_ttyvp; 367 } else 368 mutex_spin_exit(&tty_lock); 369 vprele = sp->s_ttyvp; 370 sp->s_ttyvp = NULL; 371 /* 372 * s_ttyp is not zero'd; we use this to indicate 373 * that the session once had a controlling terminal. 374 * (for logging and informational purposes) 375 */ 376 } 377 sp->s_leader = NULL; 378 379 if (vprevoke != NULL || vprele != NULL) { 380 if (vprevoke != NULL) { 381 /* Releases proc_lock. */ 382 proc_sessrele(sp); 383 VOP_REVOKE(vprevoke, REVOKEALL); 384 } else 385 mutex_exit(proc_lock); 386 if (vprele != NULL) 387 vrele(vprele); 388 mutex_enter(proc_lock); 389 } 390 } 391 fixjobc(p, p->p_pgrp, 0); 392 393 /* 394 * Finalize the last LWP's specificdata, as well as the 395 * specificdata for the proc itself. 396 */ 397 lwp_finispecific(l); 398 proc_finispecific(p); 399 400 /* 401 * Notify interested parties of our demise. 402 */ 403 KNOTE(&p->p_klist, NOTE_EXIT); 404 405 #if PERFCTRS 406 /* 407 * Save final PMC information in parent process & clean up. 408 */ 409 if (PMC_ENABLED(p)) { 410 pmc_save_context(p); 411 pmc_accumulate(p->p_pptr, p); 412 pmc_process_exit(p); 413 } 414 #endif 415 416 /* 417 * Reset p_opptr pointer of all former children which got 418 * traced by another process and were reparented. We reset 419 * it to NULL here; the trace detach code then reparents 420 * the child to initproc. We only check allproc list, since 421 * eventual former children on zombproc list won't reference 422 * p_opptr anymore. 423 */ 424 if (__predict_false(p->p_slflag & PSL_CHTRACED)) { 425 PROCLIST_FOREACH(q, &allproc) { 426 if ((q->p_flag & PK_MARKER) != 0) 427 continue; 428 if (q->p_opptr == p) 429 q->p_opptr = NULL; 430 } 431 } 432 433 /* 434 * Give orphaned children to init(8). 435 */ 436 q = LIST_FIRST(&p->p_children); 437 wakeinit = (q != NULL); 438 for (; q != NULL; q = nq) { 439 nq = LIST_NEXT(q, p_sibling); 440 441 /* 442 * Traced processes are killed since their existence 443 * means someone is screwing up. Since we reset the 444 * trace flags, the logic in sys_wait4() would not be 445 * triggered to reparent the process to its 446 * original parent, so we must do this here. 447 */ 448 if (__predict_false(q->p_slflag & PSL_TRACED)) { 449 mutex_enter(p->p_lock); 450 q->p_slflag &= ~(PSL_TRACED|PSL_FSTRACE|PSL_SYSCALL); 451 mutex_exit(p->p_lock); 452 if (q->p_opptr != q->p_pptr) { 453 struct proc *t = q->p_opptr; 454 proc_reparent(q, t ? t : initproc); 455 q->p_opptr = NULL; 456 } else 457 proc_reparent(q, initproc); 458 killproc(q, "orphaned traced process"); 459 } else 460 proc_reparent(q, initproc); 461 } 462 463 /* 464 * Move proc from allproc to zombproc, it's now nearly ready to be 465 * collected by parent. 466 */ 467 LIST_REMOVE(l, l_list); 468 LIST_REMOVE(p, p_list); 469 LIST_INSERT_HEAD(&zombproc, p, p_list); 470 471 /* 472 * Mark the process as dead. We must do this before we signal 473 * the parent. 474 */ 475 p->p_stat = SDEAD; 476 477 /* Put in front of parent's sibling list for parent to collect it */ 478 q = p->p_pptr; 479 q->p_nstopchild++; 480 if (LIST_FIRST(&q->p_children) != p) { 481 /* Put child where it can be found quickly */ 482 LIST_REMOVE(p, p_sibling); 483 LIST_INSERT_HEAD(&q->p_children, p, p_sibling); 484 } 485 486 /* 487 * Notify parent that we're gone. If parent has the P_NOCLDWAIT 488 * flag set, notify init instead (and hope it will handle 489 * this situation). 490 */ 491 if (q->p_flag & (PK_NOCLDWAIT|PK_CLDSIGIGN)) { 492 proc_reparent(p, initproc); 493 wakeinit = 1; 494 495 /* 496 * If this was the last child of our parent, notify 497 * parent, so in case he was wait(2)ing, he will 498 * continue. 499 */ 500 if (LIST_FIRST(&q->p_children) == NULL) 501 cv_broadcast(&q->p_waitcv); 502 } 503 504 /* Reload parent pointer, since p may have been reparented above */ 505 q = p->p_pptr; 506 507 if (__predict_false((p->p_slflag & PSL_FSTRACE) == 0 && p->p_exitsig != 0)) { 508 exit_psignal(p, q, &ksi); 509 kpsignal(q, &ksi, NULL); 510 } 511 512 /* Calculate the final rusage info. */ 513 calcru(p, &p->p_stats->p_ru.ru_utime, &p->p_stats->p_ru.ru_stime, 514 NULL, NULL); 515 516 if (wakeinit) 517 cv_broadcast(&initproc->p_waitcv); 518 519 callout_destroy(&l->l_timeout_ch); 520 521 /* 522 * Remaining lwp resources will be freed in lwp_exit2() once we've 523 * switch to idle context; at that point, we will be marked as a 524 * full blown zombie. 525 */ 526 mutex_enter(p->p_lock); 527 lwp_drainrefs(l); 528 lwp_lock(l); 529 l->l_prflag &= ~LPR_DETACHED; 530 l->l_stat = LSZOMB; 531 lwp_unlock(l); 532 KASSERT(curlwp == l); 533 KASSERT(p->p_nrlwps == 1); 534 KASSERT(p->p_nlwps == 1); 535 p->p_stat = SZOMB; 536 p->p_nrlwps--; 537 p->p_nzlwps++; 538 p->p_ndlwps = 0; 539 mutex_exit(p->p_lock); 540 541 /* 542 * Signal the parent to collect us, and drop the proclist lock. 543 * Drop debugger/procfs lock; no new references can be gained. 544 */ 545 cv_broadcast(&p->p_pptr->p_waitcv); 546 rw_exit(&p->p_reflock); 547 mutex_exit(proc_lock); 548 549 /* Verify that we hold no locks other than the kernel lock. */ 550 LOCKDEBUG_BARRIER(&kernel_lock, 0); 551 552 /* 553 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP! 554 */ 555 556 /* 557 * Give machine-dependent code a chance to free any MD LWP 558 * resources. This must be done before uvm_lwp_exit(), in 559 * case these resources are in the PCB. 560 */ 561 #ifndef __NO_CPU_LWP_FREE 562 cpu_lwp_free(l, 1); 563 #endif 564 pmap_deactivate(l); 565 566 /* This process no longer needs to hold the kernel lock. */ 567 #ifdef notyet 568 /* XXXSMP hold in lwp_userret() */ 569 KERNEL_UNLOCK_LAST(l); 570 #else 571 KERNEL_UNLOCK_ALL(l, NULL); 572 #endif 573 574 lwp_exit_switchaway(l); 575 } 576 577 void 578 exit_lwps(struct lwp *l) 579 { 580 struct proc *p; 581 struct lwp *l2; 582 int error; 583 lwpid_t waited; 584 int nlocks; 585 586 KERNEL_UNLOCK_ALL(l, &nlocks); 587 588 p = l->l_proc; 589 KASSERT(mutex_owned(p->p_lock)); 590 591 #ifdef KERN_SA 592 if (p->p_sa != NULL) { 593 struct sadata_vp *vp; 594 SLIST_FOREACH(vp, &p->p_sa->sa_vps, savp_next) { 595 /* 596 * Make SA-cached LWPs normal process interruptable 597 * so that the exit code can wake them. Locking 598 * savp_mutex locks all the lwps on this vp that 599 * we need to adjust. 600 */ 601 mutex_enter(&vp->savp_mutex); 602 DPRINTF(("exit_lwps: Making cached LWPs of %d on " 603 "VP %d interruptable: ", p->p_pid, vp->savp_id)); 604 TAILQ_FOREACH(l2, &vp->savp_lwpcache, l_sleepchain) { 605 l2->l_flag |= LW_SINTR; 606 DPRINTF(("%d ", l2->l_lid)); 607 } 608 DPRINTF(("\n")); 609 610 DPRINTF(("exit_lwps: Making unblocking LWPs of %d on " 611 "VP %d interruptable: ", p->p_pid, vp->savp_id)); 612 TAILQ_FOREACH(l2, &vp->savp_woken, l_sleepchain) { 613 vp->savp_woken_count--; 614 l2->l_flag |= LW_SINTR; 615 DPRINTF(("%d ", l2->l_lid)); 616 } 617 DPRINTF(("\n")); 618 mutex_exit(&vp->savp_mutex); 619 } 620 } 621 #endif 622 623 retry: 624 /* 625 * Interrupt LWPs in interruptable sleep, unsuspend suspended 626 * LWPs and then wait for everyone else to finish. 627 */ 628 LIST_FOREACH(l2, &p->p_lwps, l_sibling) { 629 if (l2 == l) 630 continue; 631 lwp_lock(l2); 632 l2->l_flag &= ~LW_SA; 633 l2->l_flag |= LW_WEXIT; 634 if ((l2->l_stat == LSSLEEP && (l2->l_flag & LW_SINTR)) || 635 l2->l_stat == LSSUSPENDED || l2->l_stat == LSSTOP) { 636 /* setrunnable() will release the lock. */ 637 setrunnable(l2); 638 DPRINTF(("exit_lwps: Made %d.%d runnable\n", 639 p->p_pid, l2->l_lid)); 640 continue; 641 } 642 lwp_unlock(l2); 643 } 644 while (p->p_nlwps > 1) { 645 DPRINTF(("exit_lwps: waiting for %d LWPs (%d zombies)\n", 646 p->p_nlwps, p->p_nzlwps)); 647 error = lwp_wait1(l, 0, &waited, LWPWAIT_EXITCONTROL); 648 if (p->p_nlwps == 1) 649 break; 650 if (error == EDEADLK) { 651 /* 652 * LWPs can get suspended/slept behind us. 653 * (eg. sa_setwoken) 654 * kick them again and retry. 655 */ 656 goto retry; 657 } 658 if (error) 659 panic("exit_lwps: lwp_wait1 failed with error %d", 660 error); 661 DPRINTF(("exit_lwps: Got LWP %d from lwp_wait1()\n", waited)); 662 } 663 664 KERNEL_LOCK(nlocks, l); 665 KASSERT(p->p_nlwps == 1); 666 } 667 668 int 669 do_sys_wait(struct lwp *l, int *pid, int *status, int options, 670 struct rusage *ru, int *was_zombie) 671 { 672 struct proc *child; 673 int error; 674 675 mutex_enter(proc_lock); 676 error = find_stopped_child(l->l_proc, *pid, options, &child, status); 677 678 if (child == NULL) { 679 mutex_exit(proc_lock); 680 *pid = 0; 681 return error; 682 } 683 684 *pid = child->p_pid; 685 686 if (child->p_stat == SZOMB) { 687 /* proc_free() will release the proc_lock. */ 688 *was_zombie = 1; 689 if (options & WNOWAIT) 690 mutex_exit(proc_lock); 691 else { 692 proc_free(child, ru); 693 } 694 } else { 695 /* Child state must have been SSTOP. */ 696 *was_zombie = 0; 697 mutex_exit(proc_lock); 698 *status = W_STOPCODE(*status); 699 } 700 701 return 0; 702 } 703 704 int 705 sys___wait450(struct lwp *l, const struct sys___wait450_args *uap, register_t *retval) 706 { 707 /* { 708 syscallarg(int) pid; 709 syscallarg(int *) status; 710 syscallarg(int) options; 711 syscallarg(struct rusage *) rusage; 712 } */ 713 int status, error; 714 int was_zombie; 715 struct rusage ru; 716 int pid = SCARG(uap, pid); 717 718 error = do_sys_wait(l, &pid, &status, SCARG(uap, options), 719 SCARG(uap, rusage) != NULL ? &ru : NULL, &was_zombie); 720 721 retval[0] = pid; 722 if (pid == 0) 723 return error; 724 725 if (SCARG(uap, rusage)) 726 error = copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 727 728 if (error == 0 && SCARG(uap, status)) 729 error = copyout(&status, SCARG(uap, status), sizeof(status)); 730 731 return error; 732 } 733 734 /* 735 * Scan list of child processes for a child process that has stopped or 736 * exited. Used by sys_wait4 and 'compat' equivalents. 737 * 738 * Must be called with the proc_lock held, and may release while waiting. 739 */ 740 static int 741 find_stopped_child(struct proc *parent, pid_t pid, int options, 742 struct proc **child_p, int *status_p) 743 { 744 struct proc *child, *dead; 745 int error; 746 747 KASSERT(mutex_owned(proc_lock)); 748 749 if (options & ~(WUNTRACED|WNOHANG|WALTSIG|WALLSIG) 750 && !(options & WOPTSCHECKED)) { 751 *child_p = NULL; 752 return EINVAL; 753 } 754 755 if (pid == 0 && !(options & WOPTSCHECKED)) 756 pid = -parent->p_pgid; 757 758 for (;;) { 759 error = ECHILD; 760 dead = NULL; 761 762 LIST_FOREACH(child, &parent->p_children, p_sibling) { 763 if (pid >= 0) { 764 if (child->p_pid != pid) { 765 child = p_find(pid, PFIND_ZOMBIE | 766 PFIND_LOCKED); 767 if (child == NULL || 768 child->p_pptr != parent) { 769 child = NULL; 770 break; 771 } 772 } 773 } else if (pid != WAIT_ANY && child->p_pgid != -pid) { 774 /* Child not in correct pgrp */ 775 continue; 776 } 777 778 /* 779 * Wait for processes with p_exitsig != SIGCHLD 780 * processes only if WALTSIG is set; wait for 781 * processes with p_exitsig == SIGCHLD only 782 * if WALTSIG is clear. 783 */ 784 if (((options & WALLSIG) == 0) && 785 (options & WALTSIG ? child->p_exitsig == SIGCHLD 786 : P_EXITSIG(child) != SIGCHLD)){ 787 if (child->p_pid == pid) { 788 child = NULL; 789 break; 790 } 791 continue; 792 } 793 794 error = 0; 795 if ((options & WNOZOMBIE) == 0) { 796 if (child->p_stat == SZOMB) 797 break; 798 if (child->p_stat == SDEAD) { 799 /* 800 * We may occasionally arrive here 801 * after receiving a signal, but 802 * immediatley before the child 803 * process is zombified. The wait 804 * will be short, so avoid returning 805 * to userspace. 806 */ 807 dead = child; 808 } 809 } 810 811 if (child->p_stat == SSTOP && 812 child->p_waited == 0 && 813 (child->p_slflag & PSL_TRACED || 814 options & WUNTRACED)) { 815 if ((options & WNOWAIT) == 0) { 816 child->p_waited = 1; 817 parent->p_nstopchild--; 818 } 819 break; 820 } 821 if (parent->p_nstopchild == 0 || child->p_pid == pid) { 822 child = NULL; 823 break; 824 } 825 } 826 827 if (child != NULL || error != 0 || 828 ((options & WNOHANG) != 0 && dead == NULL)) { 829 if (child != NULL) { 830 *status_p = child->p_xstat; 831 } 832 *child_p = child; 833 return error; 834 } 835 836 /* 837 * Wait for another child process to stop. 838 */ 839 error = cv_wait_sig(&parent->p_waitcv, proc_lock); 840 841 if (error != 0) { 842 *child_p = NULL; 843 return error; 844 } 845 } 846 } 847 848 /* 849 * Free a process after parent has taken all the state info. Must be called 850 * with the proclist lock held, and will release before returning. 851 * 852 * *ru is returned to the caller, and must be freed by the caller. 853 */ 854 static void 855 proc_free(struct proc *p, struct rusage *ru) 856 { 857 struct proc *parent = p->p_pptr; 858 struct lwp *l; 859 ksiginfo_t ksi; 860 kauth_cred_t cred1, cred2; 861 uid_t uid; 862 863 KASSERT(mutex_owned(proc_lock)); 864 KASSERT(p->p_nlwps == 1); 865 KASSERT(p->p_nzlwps == 1); 866 KASSERT(p->p_nrlwps == 0); 867 KASSERT(p->p_stat == SZOMB); 868 869 /* 870 * If we got the child via ptrace(2) or procfs, and 871 * the parent is different (meaning the process was 872 * attached, rather than run as a child), then we need 873 * to give it back to the old parent, and send the 874 * parent the exit signal. The rest of the cleanup 875 * will be done when the old parent waits on the child. 876 */ 877 if ((p->p_slflag & PSL_TRACED) != 0 && p->p_opptr != parent) { 878 mutex_enter(p->p_lock); 879 p->p_slflag &= ~(PSL_TRACED|PSL_FSTRACE|PSL_SYSCALL); 880 mutex_exit(p->p_lock); 881 parent = (p->p_opptr == NULL) ? initproc : p->p_opptr; 882 proc_reparent(p, parent); 883 p->p_opptr = NULL; 884 if (p->p_exitsig != 0) { 885 exit_psignal(p, parent, &ksi); 886 kpsignal(parent, &ksi, NULL); 887 } 888 cv_broadcast(&parent->p_waitcv); 889 mutex_exit(proc_lock); 890 return; 891 } 892 893 sched_proc_exit(parent, p); 894 895 /* 896 * Add child times of exiting process onto its own times. 897 * This cannot be done any earlier else it might get done twice. 898 */ 899 l = LIST_FIRST(&p->p_lwps); 900 p->p_stats->p_ru.ru_nvcsw += (l->l_ncsw - l->l_nivcsw); 901 p->p_stats->p_ru.ru_nivcsw += l->l_nivcsw; 902 ruadd(&p->p_stats->p_ru, &l->l_ru); 903 ruadd(&p->p_stats->p_ru, &p->p_stats->p_cru); 904 ruadd(&parent->p_stats->p_cru, &p->p_stats->p_ru); 905 if (ru != NULL) 906 *ru = p->p_stats->p_ru; 907 p->p_xstat = 0; 908 909 /* Release any SA state. */ 910 #ifdef KERN_SA 911 if (p->p_sa) 912 sa_release(p); 913 #endif 914 915 /* 916 * At this point we are going to start freeing the final resources. 917 * If anyone tries to access the proc structure after here they will 918 * get a shock - bits are missing. Attempt to make it hard! We 919 * don't bother with any further locking past this point. 920 */ 921 p->p_stat = SIDL; /* not even a zombie any more */ 922 LIST_REMOVE(p, p_list); /* off zombproc */ 923 parent->p_nstopchild--; 924 LIST_REMOVE(p, p_sibling); 925 926 /* 927 * Let pid be reallocated. 928 */ 929 proc_free_pid(p); 930 931 /* 932 * Unlink process from its process group. 933 * Releases the proc_lock. 934 */ 935 proc_leavepgrp(p); 936 937 /* 938 * Delay release until after lwp_free. 939 */ 940 cred2 = l->l_cred; 941 942 /* 943 * Free the last LWP's resources. 944 * 945 * lwp_free ensures the LWP is no longer running on another CPU. 946 */ 947 lwp_free(l, false, true); 948 949 /* 950 * Now no one except us can reach the process p. 951 */ 952 953 /* 954 * Decrement the count of procs running with this uid. 955 */ 956 cred1 = p->p_cred; 957 uid = kauth_cred_getuid(cred1); 958 (void)chgproccnt(uid, -1); 959 960 /* 961 * Release substructures. 962 */ 963 964 limfree(p->p_limit); 965 pstatsfree(p->p_stats); 966 kauth_cred_free(cred1); 967 kauth_cred_free(cred2); 968 969 /* 970 * Release reference to text vnode 971 */ 972 if (p->p_textvp) 973 vrele(p->p_textvp); 974 975 mutex_destroy(&p->p_auxlock); 976 mutex_obj_free(p->p_lock); 977 mutex_destroy(&p->p_stmutex); 978 cv_destroy(&p->p_waitcv); 979 cv_destroy(&p->p_lwpcv); 980 rw_destroy(&p->p_reflock); 981 982 proc_free_mem(p); 983 } 984 985 /* 986 * make process 'parent' the new parent of process 'child'. 987 * 988 * Must be called with proc_lock held. 989 */ 990 void 991 proc_reparent(struct proc *child, struct proc *parent) 992 { 993 994 KASSERT(mutex_owned(proc_lock)); 995 996 if (child->p_pptr == parent) 997 return; 998 999 if (child->p_stat == SZOMB || 1000 (child->p_stat == SSTOP && !child->p_waited)) { 1001 child->p_pptr->p_nstopchild--; 1002 parent->p_nstopchild++; 1003 } 1004 if (parent == initproc) 1005 child->p_exitsig = SIGCHLD; 1006 1007 LIST_REMOVE(child, p_sibling); 1008 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 1009 child->p_pptr = parent; 1010 child->p_ppid = parent->p_pid; 1011 } 1012