kern_exit.c revision 163653
1/*- 2 * Copyright (c) 1982, 1986, 1989, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 4. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)kern_exit.c 8.7 (Berkeley) 2/12/94 35 */ 36 37#include <sys/cdefs.h> 38__FBSDID("$FreeBSD: head/sys/kern/kern_exit.c 163653 2006-10-24 12:04:21Z davidxu $"); 39 40#include "opt_compat.h" 41#include "opt_ktrace.h" 42#include "opt_mac.h" 43 44#include <sys/param.h> 45#include <sys/systm.h> 46#include <sys/sysproto.h> 47#include <sys/eventhandler.h> 48#include <sys/kernel.h> 49#include <sys/malloc.h> 50#include <sys/lock.h> 51#include <sys/mutex.h> 52#include <sys/proc.h> 53#include <sys/pioctl.h> 54#include <sys/tty.h> 55#include <sys/wait.h> 56#include <sys/vmmeter.h> 57#include <sys/vnode.h> 58#include <sys/resourcevar.h> 59#include <sys/sbuf.h> 60#include <sys/signalvar.h> 61#include <sys/sched.h> 62#include <sys/sx.h> 63#include <sys/syscallsubr.h> 64#include <sys/syslog.h> 65#include <sys/ptrace.h> 66#include <sys/acct.h> /* for acct_process() function prototype */ 67#include <sys/filedesc.h> 68#include <sys/shm.h> 69#include <sys/sem.h> 70#ifdef KTRACE 71#include <sys/ktrace.h> 72#endif 73 74#include <security/audit/audit.h> 75#include <security/mac/mac_framework.h> 76 77#include <vm/vm.h> 78#include <vm/vm_extern.h> 79#include <vm/vm_param.h> 80#include <vm/pmap.h> 81#include <vm/vm_map.h> 82#include <vm/vm_page.h> 83#include <vm/uma.h> 84 85/* Required to be non-static for SysVR4 emulator */ 86MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 87 88/* Hook for NFS teardown procedure. */ 89void (*nlminfo_release_p)(struct proc *p); 90 91/* 92 * exit -- 93 * Death of process. 94 * 95 * MPSAFE 96 */ 97void 98sys_exit(struct thread *td, struct sys_exit_args *uap) 99{ 100 101 exit1(td, W_EXITCODE(uap->rval, 0)); 102 /* NOTREACHED */ 103} 104 105/* 106 * Exit: deallocate address space and other resources, change proc state 107 * to zombie, and unlink proc from allproc and parent's lists. Save exit 108 * status and rusage for wait(). Check for child processes and orphan them. 109 */ 110void 111exit1(struct thread *td, int rv) 112{ 113 struct proc *p, *nq, *q; 114 struct tty *tp; 115 struct vnode *ttyvp; 116 struct vnode *vtmp; 117#ifdef KTRACE 118 struct vnode *tracevp; 119 struct ucred *tracecred; 120#endif 121 struct plimit *plim; 122 int locked; 123 124 /* 125 * Drop Giant if caller has it. Eventually we should warn about 126 * being called with Giant held. 127 */ 128 while (mtx_owned(&Giant)) 129 mtx_unlock(&Giant); 130 131 p = td->td_proc; 132 if (p == initproc) { 133 printf("init died (signal %d, exit %d)\n", 134 WTERMSIG(rv), WEXITSTATUS(rv)); 135 panic("Going nowhere without my init!"); 136 } 137 138 /* 139 * MUST abort all other threads before proceeding past here. 140 */ 141 PROC_LOCK(p); 142 if (p->p_flag & P_HADTHREADS) { 143retry: 144 /* 145 * First check if some other thread got here before us.. 146 * if so, act apropriatly, (exit or suspend); 147 */ 148 thread_suspend_check(0); 149 150 /* 151 * Kill off the other threads. This requires 152 * some co-operation from other parts of the kernel 153 * so it may not be instantaneous. With this state set 154 * any thread entering the kernel from userspace will 155 * thread_exit() in trap(). Any thread attempting to 156 * sleep will return immediately with EINTR or EWOULDBLOCK 157 * which will hopefully force them to back out to userland 158 * freeing resources as they go. Any thread attempting 159 * to return to userland will thread_exit() from userret(). 160 * thread_exit() will unsuspend us when the last of the 161 * other threads exits. 162 * If there is already a thread singler after resumption, 163 * calling thread_single will fail; in that case, we just 164 * re-check all suspension request, the thread should 165 * either be suspended there or exit. 166 */ 167 if (thread_single(SINGLE_EXIT)) 168 goto retry; 169 170 /* 171 * All other activity in this process is now stopped. 172 * Threading support has been turned off. 173 */ 174 } 175 176 /* 177 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold 178 * on our vmspace, so we should block below until they have 179 * released their reference to us. Note that if they have 180 * requested S_EXIT stops we will block here until they ack 181 * via PIOCCONT. 182 */ 183 _STOPEVENT(p, S_EXIT, rv); 184 185 /* 186 * Note that we are exiting and do another wakeup of anyone in 187 * PIOCWAIT in case they aren't listening for S_EXIT stops or 188 * decided to wait again after we told them we are exiting. 189 */ 190 p->p_flag |= P_WEXIT; 191 wakeup(&p->p_stype); 192 193 /* 194 * Wait for any processes that have a hold on our vmspace to 195 * release their reference. 196 */ 197 while (p->p_lock > 0) 198 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0); 199 200 PROC_LOCK(p->p_pptr); 201 sigqueue_take(p->p_ksi); 202 PROC_UNLOCK(p->p_pptr); 203 204 PROC_UNLOCK(p); 205 206#ifdef AUDIT 207 /* 208 * The Sun BSM exit token contains two components: an exit status as 209 * passed to exit(), and a return value to indicate what sort of exit 210 * it was. The exit status is WEXITSTATUS(rv), but it's not clear 211 * what the return value is. 212 */ 213 AUDIT_ARG(exit, WEXITSTATUS(rv), 0); 214 AUDIT_SYSCALL_EXIT(0, td); 215#endif 216 217 /* Are we a task leader? */ 218 if (p == p->p_leader) { 219 mtx_lock(&ppeers_lock); 220 q = p->p_peers; 221 while (q != NULL) { 222 PROC_LOCK(q); 223 psignal(q, SIGKILL); 224 PROC_UNLOCK(q); 225 q = q->p_peers; 226 } 227 while (p->p_peers != NULL) 228 msleep(p, &ppeers_lock, PWAIT, "exit1", 0); 229 mtx_unlock(&ppeers_lock); 230 } 231 232 /* 233 * Check if any loadable modules need anything done at process exit. 234 * E.g. SYSV IPC stuff 235 * XXX what if one of these generates an error? 236 */ 237 EVENTHANDLER_INVOKE(process_exit, p); 238 239 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage), 240 M_ZOMBIE, M_WAITOK); 241 /* 242 * If parent is waiting for us to exit or exec, 243 * P_PPWAIT is set; we will wakeup the parent below. 244 */ 245 PROC_LOCK(p); 246 stopprofclock(p); 247 p->p_flag &= ~(P_TRACED | P_PPWAIT); 248 249 /* 250 * Stop the real interval timer. If the handler is currently 251 * executing, prevent it from rearming itself and let it finish. 252 */ 253 if (timevalisset(&p->p_realtimer.it_value) && 254 callout_stop(&p->p_itcallout) == 0) { 255 timevalclear(&p->p_realtimer.it_interval); 256 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0); 257 KASSERT(!timevalisset(&p->p_realtimer.it_value), 258 ("realtime timer is still armed")); 259 } 260 PROC_UNLOCK(p); 261 262 /* 263 * Reset any sigio structures pointing to us as a result of 264 * F_SETOWN with our pid. 265 */ 266 funsetownlst(&p->p_sigiolst); 267 268 /* 269 * If this process has an nlminfo data area (for lockd), release it 270 */ 271 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL) 272 (*nlminfo_release_p)(p); 273 274 /* 275 * Close open files and release open-file table. 276 * This may block! 277 */ 278 fdfree(td); 279 280 /* 281 * If this thread tickled GEOM, we need to wait for the giggling to 282 * stop before we return to userland 283 */ 284 if (td->td_pflags & TDP_GEOM) 285 g_waitidle(); 286 287 /* 288 * Remove ourself from our leader's peer list and wake our leader. 289 */ 290 mtx_lock(&ppeers_lock); 291 if (p->p_leader->p_peers) { 292 q = p->p_leader; 293 while (q->p_peers != p) 294 q = q->p_peers; 295 q->p_peers = p->p_peers; 296 wakeup(p->p_leader); 297 } 298 mtx_unlock(&ppeers_lock); 299 300 vmspace_exit(td); 301 302 mtx_lock(&Giant); /* XXX TTY */ 303 sx_xlock(&proctree_lock); 304 if (SESS_LEADER(p)) { 305 struct session *sp; 306 307 sp = p->p_session; 308 if (sp->s_ttyvp) { 309 /* 310 * Controlling process. 311 * Signal foreground pgrp, 312 * drain controlling terminal 313 * and revoke access to controlling terminal. 314 */ 315 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 316 tp = sp->s_ttyp; 317 if (sp->s_ttyp->t_pgrp) { 318 PGRP_LOCK(sp->s_ttyp->t_pgrp); 319 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 320 PGRP_UNLOCK(sp->s_ttyp->t_pgrp); 321 } 322 /* XXX tp should be locked. */ 323 sx_xunlock(&proctree_lock); 324 (void) ttywait(tp); 325 sx_xlock(&proctree_lock); 326 /* 327 * The tty could have been revoked 328 * if we blocked. 329 */ 330 if (sp->s_ttyvp) { 331 ttyvp = sp->s_ttyvp; 332 SESS_LOCK(p->p_session); 333 sp->s_ttyvp = NULL; 334 SESS_UNLOCK(p->p_session); 335 sx_xunlock(&proctree_lock); 336 VOP_LOCK(ttyvp, LK_EXCLUSIVE, td); 337 VOP_REVOKE(ttyvp, REVOKEALL); 338 vput(ttyvp); 339 sx_xlock(&proctree_lock); 340 } 341 } 342 if (sp->s_ttyvp) { 343 ttyvp = sp->s_ttyvp; 344 SESS_LOCK(p->p_session); 345 sp->s_ttyvp = NULL; 346 SESS_UNLOCK(p->p_session); 347 vrele(ttyvp); 348 } 349 /* 350 * s_ttyp is not zero'd; we use this to indicate 351 * that the session once had a controlling terminal. 352 * (for logging and informational purposes) 353 */ 354 } 355 SESS_LOCK(p->p_session); 356 sp->s_leader = NULL; 357 SESS_UNLOCK(p->p_session); 358 } 359 fixjobc(p, p->p_pgrp, 0); 360 sx_xunlock(&proctree_lock); 361 (void)acct_process(td); 362 mtx_unlock(&Giant); 363#ifdef KTRACE 364 /* 365 * Drain any pending records on the thread and release the trace 366 * file. It might be better if drain-and-clear were atomic. 367 */ 368 ktrprocexit(td); 369 PROC_LOCK(p); 370 mtx_lock(&ktrace_mtx); 371 p->p_traceflag = 0; /* don't trace the vrele() */ 372 tracevp = p->p_tracevp; 373 p->p_tracevp = NULL; 374 tracecred = p->p_tracecred; 375 p->p_tracecred = NULL; 376 mtx_unlock(&ktrace_mtx); 377 PROC_UNLOCK(p); 378 if (tracevp != NULL) { 379 locked = VFS_LOCK_GIANT(tracevp->v_mount); 380 vrele(tracevp); 381 VFS_UNLOCK_GIANT(locked); 382 } 383 if (tracecred != NULL) 384 crfree(tracecred); 385#endif 386 /* 387 * Release reference to text vnode 388 */ 389 if ((vtmp = p->p_textvp) != NULL) { 390 p->p_textvp = NULL; 391 locked = VFS_LOCK_GIANT(vtmp->v_mount); 392 vrele(vtmp); 393 VFS_UNLOCK_GIANT(locked); 394 } 395 396 /* 397 * Release our limits structure. 398 */ 399 PROC_LOCK(p); 400 plim = p->p_limit; 401 p->p_limit = NULL; 402 PROC_UNLOCK(p); 403 lim_free(plim); 404 405 /* 406 * Remove proc from allproc queue and pidhash chain. 407 * Place onto zombproc. Unlink from parent's child list. 408 */ 409 sx_xlock(&allproc_lock); 410 LIST_REMOVE(p, p_list); 411 LIST_INSERT_HEAD(&zombproc, p, p_list); 412 LIST_REMOVE(p, p_hash); 413 sx_xunlock(&allproc_lock); 414 415 /* 416 * Reparent all of our children to init. 417 */ 418 sx_xlock(&proctree_lock); 419 q = LIST_FIRST(&p->p_children); 420 if (q != NULL) /* only need this if any child is S_ZOMB */ 421 wakeup(initproc); 422 for (; q != NULL; q = nq) { 423 nq = LIST_NEXT(q, p_sibling); 424 PROC_LOCK(q); 425 proc_reparent(q, initproc); 426 q->p_sigparent = SIGCHLD; 427 /* 428 * Traced processes are killed 429 * since their existence means someone is screwing up. 430 */ 431 if (q->p_flag & P_TRACED) { 432 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE); 433 psignal(q, SIGKILL); 434 } 435 PROC_UNLOCK(q); 436 } 437 438 /* 439 * Save exit status and finalize rusage info except for times, 440 * adding in child rusage info later when our time is locked. 441 */ 442 PROC_LOCK(p); 443 p->p_xstat = rv; 444 p->p_xthread = td; 445 p->p_stats->p_ru.ru_nvcsw++; 446 *p->p_ru = p->p_stats->p_ru; 447 448 /* 449 * Notify interested parties of our demise. 450 */ 451 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT); 452 453 /* 454 * Just delete all entries in the p_klist. At this point we won't 455 * report any more events, and there are nasty race conditions that 456 * can beat us if we don't. 457 */ 458 knlist_clear(&p->p_klist, 1); 459 460 /* 461 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 462 * flag set, or if the handler is set to SIG_IGN, notify process 463 * 1 instead (and hope it will handle this situation). 464 */ 465 PROC_LOCK(p->p_pptr); 466 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx); 467 if (p->p_pptr->p_sigacts->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) { 468 struct proc *pp; 469 470 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx); 471 pp = p->p_pptr; 472 PROC_UNLOCK(pp); 473 proc_reparent(p, initproc); 474 p->p_sigparent = SIGCHLD; 475 PROC_LOCK(p->p_pptr); 476 /* 477 * If this was the last child of our parent, notify 478 * parent, so in case he was wait(2)ing, he will 479 * continue. 480 */ 481 if (LIST_EMPTY(&pp->p_children)) 482 wakeup(pp); 483 } else 484 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx); 485 486 if (p->p_pptr == initproc) 487 psignal(p->p_pptr, SIGCHLD); 488 else if (p->p_sigparent != 0) { 489 if (p->p_sigparent == SIGCHLD) 490 childproc_exited(p); 491 else /* LINUX thread */ 492 psignal(p->p_pptr, p->p_sigparent); 493 } 494 PROC_UNLOCK(p->p_pptr); 495 PROC_UNLOCK(p); 496 497 /* 498 * Finally, call machine-dependent code to release the remaining 499 * resources including address space. 500 * The address space is released by "vmspace_exitfree(p)" in 501 * vm_waitproc(). 502 */ 503 cpu_exit(td); 504 505 WITNESS_WARN(WARN_PANIC, &proctree_lock.sx_object, 506 "process (pid %d) exiting", p->p_pid); 507 508 PROC_LOCK(p); 509 PROC_LOCK(p->p_pptr); 510 sx_xunlock(&proctree_lock); 511 512 /* 513 * The state PRS_ZOMBIE prevents other proesses from sending 514 * signal to the process, to avoid memory leak, we free memory 515 * for signal queue at the time when the state is set. 516 */ 517 sigqueue_flush(&p->p_sigqueue); 518 sigqueue_flush(&td->td_sigqueue); 519 520 /* 521 * We have to wait until after acquiring all locks before 522 * changing p_state. We need to avoid all possible context 523 * switches (including ones from blocking on a mutex) while 524 * marked as a zombie. We also have to set the zombie state 525 * before we release the parent process' proc lock to avoid 526 * a lost wakeup. So, we first call wakeup, then we grab the 527 * sched lock, update the state, and release the parent process' 528 * proc lock. 529 */ 530 wakeup(p->p_pptr); 531 mtx_lock_spin(&sched_lock); 532 p->p_state = PRS_ZOMBIE; 533 PROC_UNLOCK(p->p_pptr); 534 535 sched_exit(p->p_pptr, td); 536 537 /* 538 * Hopefully no one will try to deliver a signal to the process this 539 * late in the game. 540 */ 541 knlist_destroy(&p->p_klist); 542 543 /* 544 * Make sure the scheduler takes this thread out of its tables etc. 545 * This will also release this thread's reference to the ucred. 546 * Other thread parts to release include pcb bits and such. 547 */ 548 thread_exit(); 549} 550 551 552#ifndef _SYS_SYSPROTO_H_ 553struct abort2_args { 554 char *why; 555 int nargs; 556 void **args; 557}; 558#endif 559 560/* 561 * MPSAFE. 562 */ 563int 564abort2(struct thread *td, struct abort2_args *uap) 565{ 566 struct proc *p = td->td_proc; 567 struct sbuf *sb; 568 void *uargs[16]; 569 int error, i, sig; 570 571 error = 0; /* satisfy compiler */ 572 573 /* 574 * Do it right now so we can log either proper call of abort2(), or 575 * note, that invalid argument was passed. 512 is big enough to 576 * handle 16 arguments' descriptions with additional comments. 577 */ 578 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN); 579 sbuf_clear(sb); 580 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ", 581 p->p_comm, p->p_pid, td->td_ucred->cr_uid); 582 /* 583 * Since we can't return from abort2(), send SIGKILL in cases, where 584 * abort2() was called improperly 585 */ 586 sig = SIGKILL; 587 /* Prevent from DoSes from user-space. */ 588 if (uap->nargs < 0 || uap->nargs > 16) 589 goto out; 590 if (uap->args == NULL) 591 goto out; 592 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *)); 593 if (error != 0) 594 goto out; 595 /* 596 * Limit size of 'reason' string to 128. Will fit even when 597 * maximal number of arguments was chosen to be logged. 598 */ 599 if (uap->why != NULL) { 600 error = sbuf_copyin(sb, uap->why, 128); 601 if (error < 0) 602 goto out; 603 } else { 604 sbuf_printf(sb, "(null)"); 605 } 606 if (uap->nargs) { 607 sbuf_printf(sb, "("); 608 for (i = 0;i < uap->nargs; i++) 609 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]); 610 sbuf_printf(sb, ")"); 611 } 612 /* 613 * Final stage: arguments were proper, string has been 614 * successfully copied from userspace, and copying pointers 615 * from user-space succeed. 616 */ 617 sig = SIGABRT; 618out: 619 if (sig == SIGKILL) { 620 sbuf_trim(sb); 621 sbuf_printf(sb, " (Reason text inaccessible)"); 622 } 623 sbuf_cat(sb, "\n"); 624 sbuf_finish(sb); 625 log(LOG_INFO, "%s", sbuf_data(sb)); 626 sbuf_delete(sb); 627 exit1(td, W_EXITCODE(0, sig)); 628 return (0); 629} 630 631 632#ifdef COMPAT_43 633/* 634 * The dirty work is handled by kern_wait(). 635 * 636 * MPSAFE. 637 */ 638int 639owait(struct thread *td, struct owait_args *uap __unused) 640{ 641 int error, status; 642 643 error = kern_wait(td, WAIT_ANY, &status, 0, NULL); 644 if (error == 0) 645 td->td_retval[1] = status; 646 return (error); 647} 648#endif /* COMPAT_43 */ 649 650/* 651 * The dirty work is handled by kern_wait(). 652 * 653 * MPSAFE. 654 */ 655int 656wait4(struct thread *td, struct wait_args *uap) 657{ 658 struct rusage ru, *rup; 659 int error, status; 660 661 if (uap->rusage != NULL) 662 rup = &ru; 663 else 664 rup = NULL; 665 error = kern_wait(td, uap->pid, &status, uap->options, rup); 666 if (uap->status != NULL && error == 0) 667 error = copyout(&status, uap->status, sizeof(status)); 668 if (uap->rusage != NULL && error == 0) 669 error = copyout(&ru, uap->rusage, sizeof(struct rusage)); 670 return (error); 671} 672 673int 674kern_wait(struct thread *td, pid_t pid, int *status, int options, 675 struct rusage *rusage) 676{ 677 struct proc *p, *q, *t; 678 int error, nfound; 679 680 AUDIT_ARG(pid, pid); 681 682 q = td->td_proc; 683 if (pid == 0) { 684 PROC_LOCK(q); 685 pid = -q->p_pgid; 686 PROC_UNLOCK(q); 687 } 688 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED|WLINUXCLONE)) 689 return (EINVAL); 690loop: 691 if (q->p_flag & P_STATCHILD) { 692 PROC_LOCK(q); 693 q->p_flag &= ~P_STATCHILD; 694 PROC_UNLOCK(q); 695 } 696 nfound = 0; 697 sx_xlock(&proctree_lock); 698 LIST_FOREACH(p, &q->p_children, p_sibling) { 699 PROC_LOCK(p); 700 if (pid != WAIT_ANY && 701 p->p_pid != pid && p->p_pgid != -pid) { 702 PROC_UNLOCK(p); 703 continue; 704 } 705 if (p_canwait(td, p)) { 706 PROC_UNLOCK(p); 707 continue; 708 } 709 710 /* 711 * This special case handles a kthread spawned by linux_clone 712 * (see linux_misc.c). The linux_wait4 and linux_waitpid 713 * functions need to be able to distinguish between waiting 714 * on a process and waiting on a thread. It is a thread if 715 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option 716 * signifies we want to wait for threads and not processes. 717 */ 718 if ((p->p_sigparent != SIGCHLD) ^ 719 ((options & WLINUXCLONE) != 0)) { 720 PROC_UNLOCK(p); 721 continue; 722 } 723 724 nfound++; 725 if (p->p_state == PRS_ZOMBIE) { 726 727 /* 728 * It is possible that the last thread of this 729 * process is still running on another CPU 730 * in thread_exit() after having dropped the process 731 * lock via PROC_UNLOCK() but before it has completed 732 * cpu_throw(). In that case, the other thread must 733 * still hold sched_lock, so simply by acquiring 734 * sched_lock once we will wait long enough for the 735 * thread to exit in that case. 736 */ 737 mtx_lock_spin(&sched_lock); 738 mtx_unlock_spin(&sched_lock); 739 740 td->td_retval[0] = p->p_pid; 741 if (status) 742 *status = p->p_xstat; /* convert to int */ 743 if (rusage) { 744 *rusage = *p->p_ru; 745 calcru(p, &rusage->ru_utime, &rusage->ru_stime); 746 } 747 748 PROC_LOCK(q); 749 sigqueue_take(p->p_ksi); 750 PROC_UNLOCK(q); 751 752 /* 753 * If we got the child via a ptrace 'attach', 754 * we need to give it back to the old parent. 755 */ 756 PROC_UNLOCK(p); 757 if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) { 758 PROC_LOCK(p); 759 p->p_oppid = 0; 760 proc_reparent(p, t); 761 PROC_UNLOCK(p); 762 tdsignal(t, NULL, SIGCHLD, p->p_ksi); 763 wakeup(t); 764 PROC_UNLOCK(t); 765 sx_xunlock(&proctree_lock); 766 return (0); 767 } 768 769 /* 770 * Remove other references to this process to ensure 771 * we have an exclusive reference. 772 */ 773 sx_xlock(&allproc_lock); 774 LIST_REMOVE(p, p_list); /* off zombproc */ 775 sx_xunlock(&allproc_lock); 776 LIST_REMOVE(p, p_sibling); 777 leavepgrp(p); 778 sx_xunlock(&proctree_lock); 779 780 /* 781 * As a side effect of this lock, we know that 782 * all other writes to this proc are visible now, so 783 * no more locking is needed for p. 784 */ 785 PROC_LOCK(p); 786 p->p_xstat = 0; /* XXX: why? */ 787 PROC_UNLOCK(p); 788 PROC_LOCK(q); 789 ruadd(&q->p_stats->p_cru, &q->p_crux, p->p_ru, 790 &p->p_rux); 791 PROC_UNLOCK(q); 792 FREE(p->p_ru, M_ZOMBIE); 793 p->p_ru = NULL; 794 795 /* 796 * Decrement the count of procs running with this uid. 797 */ 798 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 799 800 /* 801 * Free credentials, arguments, and sigacts. 802 */ 803 crfree(p->p_ucred); 804 p->p_ucred = NULL; 805 pargs_drop(p->p_args); 806 p->p_args = NULL; 807 sigacts_free(p->p_sigacts); 808 p->p_sigacts = NULL; 809 810 /* 811 * Do any thread-system specific cleanups. 812 */ 813 thread_wait(p); 814 815 /* 816 * Give vm and machine-dependent layer a chance 817 * to free anything that cpu_exit couldn't 818 * release while still running in process context. 819 */ 820 vm_waitproc(p); 821#ifdef MAC 822 mac_destroy_proc(p); 823#endif 824#ifdef AUDIT 825 audit_proc_free(p); 826#endif 827 KASSERT(FIRST_THREAD_IN_PROC(p), 828 ("kern_wait: no residual thread!")); 829 uma_zfree(proc_zone, p); 830 sx_xlock(&allproc_lock); 831 nprocs--; 832 sx_xunlock(&allproc_lock); 833 return (0); 834 } 835 mtx_lock_spin(&sched_lock); 836 if ((p->p_flag & P_STOPPED_SIG) && 837 (p->p_suspcount == p->p_numthreads) && 838 (p->p_flag & P_WAITED) == 0 && 839 (p->p_flag & P_TRACED || options & WUNTRACED)) { 840 mtx_unlock_spin(&sched_lock); 841 p->p_flag |= P_WAITED; 842 sx_xunlock(&proctree_lock); 843 td->td_retval[0] = p->p_pid; 844 if (status) 845 *status = W_STOPCODE(p->p_xstat); 846 847 PROC_LOCK(q); 848 sigqueue_take(p->p_ksi); 849 PROC_UNLOCK(q); 850 PROC_UNLOCK(p); 851 852 return (0); 853 } 854 mtx_unlock_spin(&sched_lock); 855 if (options & WCONTINUED && (p->p_flag & P_CONTINUED)) { 856 sx_xunlock(&proctree_lock); 857 td->td_retval[0] = p->p_pid; 858 p->p_flag &= ~P_CONTINUED; 859 860 PROC_LOCK(q); 861 sigqueue_take(p->p_ksi); 862 PROC_UNLOCK(q); 863 PROC_UNLOCK(p); 864 865 if (status) 866 *status = SIGCONT; 867 return (0); 868 } 869 PROC_UNLOCK(p); 870 } 871 if (nfound == 0) { 872 sx_xunlock(&proctree_lock); 873 return (ECHILD); 874 } 875 if (options & WNOHANG) { 876 sx_xunlock(&proctree_lock); 877 td->td_retval[0] = 0; 878 return (0); 879 } 880 PROC_LOCK(q); 881 sx_xunlock(&proctree_lock); 882 if (q->p_flag & P_STATCHILD) { 883 q->p_flag &= ~P_STATCHILD; 884 error = 0; 885 } else 886 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0); 887 PROC_UNLOCK(q); 888 if (error) 889 return (error); 890 goto loop; 891} 892 893/* 894 * Make process 'parent' the new parent of process 'child'. 895 * Must be called with an exclusive hold of proctree lock. 896 */ 897void 898proc_reparent(struct proc *child, struct proc *parent) 899{ 900 901 sx_assert(&proctree_lock, SX_XLOCKED); 902 PROC_LOCK_ASSERT(child, MA_OWNED); 903 if (child->p_pptr == parent) 904 return; 905 906 LIST_REMOVE(child, p_sibling); 907 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 908 child->p_pptr = parent; 909} 910