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