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/mac.h> 69 #include <sys/shm.h> 70 #include <sys/sem.h> 71 #ifdef KTRACE 72 #include <sys/ktrace.h> 73 #endif 74 75 #include <security/audit/audit.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 -- 93 * Death of process. 94 * 95 * MPSAFE 96 */ 97 void 98 sys_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 */ 110 void 111 exit1(struct thread *td, int rv) 112 { 113 uint64_t new_switchtime; 114 struct proc *p, *nq, *q; 115 struct tty *tp; 116 struct vnode *ttyvp; 117 struct vmspace *vm; 118 struct vnode *vtmp; 119 #ifdef KTRACE 120 struct vnode *tracevp; 121 struct ucred *tracecred; 122 #endif 123 struct plimit *plim; 124 int locked, refcnt; 125 126 /* 127 * Drop Giant if caller has it. Eventually we should warn about 128 * being called with Giant held. 129 */ 130 while (mtx_owned(&Giant)) 131 mtx_unlock(&Giant); 132 133 p = td->td_proc; 134 if (p == initproc) { 135 printf("init died (signal %d, exit %d)\n", 136 WTERMSIG(rv), WEXITSTATUS(rv)); 137 panic("Going nowhere without my init!"); 138 } 139 140 /* 141 * MUST abort all other threads before proceeding past here. 142 */ 143 PROC_LOCK(p); 144 if (p->p_flag & P_HADTHREADS) { 145 retry: 146 /* 147 * First check if some other thread got here before us.. 148 * if so, act apropriatly, (exit or suspend); 149 */ 150 thread_suspend_check(0); 151 152 /* 153 * Kill off the other threads. This requires 154 * some co-operation from other parts of the kernel 155 * so it may not be instantaneous. With this state set 156 * any thread entering the kernel from userspace will 157 * thread_exit() in trap(). Any thread attempting to 158 * sleep will return immediately with EINTR or EWOULDBLOCK 159 * which will hopefully force them to back out to userland 160 * freeing resources as they go. Any thread attempting 161 * to return to userland will thread_exit() from userret(). 162 * thread_exit() will unsuspend us when the last of the 163 * other threads exits. 164 * If there is already a thread singler after resumption, 165 * calling thread_single will fail; in that case, we just 166 * re-check all suspension request, the thread should 167 * either be suspended there or exit. 168 */ 169 if (thread_single(SINGLE_EXIT)) 170 goto retry; 171 172 /* 173 * All other activity in this process is now stopped. 174 * Threading support has been turned off. 175 */ 176 } 177 178 /* 179 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold 180 * on our vmspace, so we should block below until they have 181 * released their reference to us. Note that if they have 182 * requested S_EXIT stops we will block here until they ack 183 * via PIOCCONT. 184 */ 185 _STOPEVENT(p, S_EXIT, rv); 186 187 /* 188 * Note that we are exiting and do another wakeup of anyone in 189 * PIOCWAIT in case they aren't listening for S_EXIT stops or 190 * decided to wait again after we told them we are exiting. 191 */ 192 p->p_flag |= P_WEXIT; 193 wakeup(&p->p_stype); 194 195 /* 196 * Wait for any processes that have a hold on our vmspace to 197 * release their reference. 198 */ 199 while (p->p_lock > 0) 200 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0); 201 202 PROC_LOCK(p->p_pptr); 203 sigqueue_take(p->p_ksi); 204 PROC_UNLOCK(p->p_pptr); 205 206 PROC_UNLOCK(p); 207 208 #ifdef AUDIT 209 /* 210 * The Sun BSM exit token contains two components: an exit status as 211 * passed to exit(), and a return value to indicate what sort of exit 212 * it was. The exit status is WEXITSTATUS(rv), but it's not clear 213 * what the return value is. 214 */ 215 AUDIT_ARG(exit, WEXITSTATUS(rv), 0); 216 AUDIT_SYSCALL_EXIT(0, td); 217 #endif 218 219 /* Are we a task leader? */ 220 if (p == p->p_leader) { 221 mtx_lock(&ppeers_lock); 222 q = p->p_peers; 223 while (q != NULL) { 224 PROC_LOCK(q); 225 psignal(q, SIGKILL); 226 PROC_UNLOCK(q); 227 q = q->p_peers; 228 } 229 while (p->p_peers != NULL) 230 msleep(p, &ppeers_lock, PWAIT, "exit1", 0); 231 mtx_unlock(&ppeers_lock); 232 } 233 234 /* 235 * Check if any loadable modules need anything done at process exit. 236 * E.g. SYSV IPC stuff 237 * XXX what if one of these generates an error? 238 */ 239 EVENTHANDLER_INVOKE(process_exit, p); 240 241 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage), 242 M_ZOMBIE, M_WAITOK); 243 /* 244 * If parent is waiting for us to exit or exec, 245 * P_PPWAIT is set; we will wakeup the parent below. 246 */ 247 PROC_LOCK(p); 248 stopprofclock(p); 249 p->p_flag &= ~(P_TRACED | P_PPWAIT); 250 251 /* 252 * Stop the real interval timer. If the handler is currently 253 * executing, prevent it from rearming itself and let it finish. 254 */ 255 if (timevalisset(&p->p_realtimer.it_value) && 256 callout_stop(&p->p_itcallout) == 0) { 257 timevalclear(&p->p_realtimer.it_interval); 258 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0); 259 KASSERT(!timevalisset(&p->p_realtimer.it_value), 260 ("realtime timer is still armed")); 261 } 262 sigqueue_flush(&p->p_sigqueue); 263 sigqueue_flush(&td->td_sigqueue); 264 PROC_UNLOCK(p); 265 266 /* 267 * Reset any sigio structures pointing to us as a result of 268 * F_SETOWN with our pid. 269 */ 270 mtx_lock(&Giant); /* XXX: not sure if needed */ 271 funsetownlst(&p->p_sigiolst); 272 mtx_unlock(&Giant); 273 274 /* 275 * If this process has an nlminfo data area (for lockd), release it 276 */ 277 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL) 278 (*nlminfo_release_p)(p); 279 280 /* 281 * Close open files and release open-file table. 282 * This may block! 283 */ 284 fdfree(td); 285 286 /* 287 * If this thread tickled GEOM, we need to wait for the giggling to 288 * stop before we return to userland 289 */ 290 if (td->td_pflags & TDP_GEOM) 291 g_waitidle(); 292 293 /* 294 * Remove ourself from our leader's peer list and wake our leader. 295 */ 296 mtx_lock(&ppeers_lock); 297 if (p->p_leader->p_peers) { 298 q = p->p_leader; 299 while (q->p_peers != p) 300 q = q->p_peers; 301 q->p_peers = p->p_peers; 302 wakeup(p->p_leader); 303 } 304 mtx_unlock(&ppeers_lock); 305 306 /* The next two chunks should probably be moved to vmspace_exit. */ 307 vm = p->p_vmspace; 308 /* 309 * Release user portion of address space. 310 * This releases references to vnodes, 311 * which could cause I/O if the file has been unlinked. 312 * Need to do this early enough that we can still sleep. 313 * Can't free the entire vmspace as the kernel stack 314 * may be mapped within that space also. 315 * 316 * Processes sharing the same vmspace may exit in one order, and 317 * get cleaned up by vmspace_exit() in a different order. The 318 * last exiting process to reach this point releases as much of 319 * the environment as it can, and the last process cleaned up 320 * by vmspace_exit() (which decrements exitingcnt) cleans up the 321 * remainder. 322 */ 323 atomic_add_int(&vm->vm_exitingcnt, 1); 324 do 325 refcnt = vm->vm_refcnt; 326 while (!atomic_cmpset_int(&vm->vm_refcnt, refcnt, refcnt - 1)); 327 if (refcnt == 1) { 328 shmexit(vm); 329 pmap_remove_pages(vmspace_pmap(vm), vm_map_min(&vm->vm_map), 330 vm_map_max(&vm->vm_map)); 331 (void) vm_map_remove(&vm->vm_map, vm_map_min(&vm->vm_map), 332 vm_map_max(&vm->vm_map)); 333 } 334 335 sx_xlock(&proctree_lock); 336 if (SESS_LEADER(p)) { 337 struct session *sp; 338 339 sp = p->p_session; 340 if (sp->s_ttyvp) { 341 locked = VFS_LOCK_GIANT(sp->s_ttyvp->v_mount); 342 /* 343 * Controlling process. 344 * Signal foreground pgrp, 345 * drain controlling terminal 346 * and revoke access to controlling terminal. 347 */ 348 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 349 tp = sp->s_ttyp; 350 if (sp->s_ttyp->t_pgrp) { 351 PGRP_LOCK(sp->s_ttyp->t_pgrp); 352 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 353 PGRP_UNLOCK(sp->s_ttyp->t_pgrp); 354 } 355 /* XXX tp should be locked. */ 356 sx_xunlock(&proctree_lock); 357 (void) ttywait(tp); 358 sx_xlock(&proctree_lock); 359 /* 360 * The tty could have been revoked 361 * if we blocked. 362 */ 363 if (sp->s_ttyvp) { 364 ttyvp = sp->s_ttyvp; 365 SESS_LOCK(p->p_session); 366 sp->s_ttyvp = NULL; 367 SESS_UNLOCK(p->p_session); 368 sx_xunlock(&proctree_lock); 369 VOP_LOCK(ttyvp, LK_EXCLUSIVE, td); 370 VOP_REVOKE(ttyvp, REVOKEALL); 371 vput(ttyvp); 372 sx_xlock(&proctree_lock); 373 } 374 } 375 if (sp->s_ttyvp) { 376 ttyvp = sp->s_ttyvp; 377 SESS_LOCK(p->p_session); 378 sp->s_ttyvp = NULL; 379 SESS_UNLOCK(p->p_session); 380 vrele(ttyvp); 381 } 382 /* 383 * s_ttyp is not zero'd; we use this to indicate 384 * that the session once had a controlling terminal. 385 * (for logging and informational purposes) 386 */ 387 VFS_UNLOCK_GIANT(locked); 388 } 389 SESS_LOCK(p->p_session); 390 sp->s_leader = NULL; 391 SESS_UNLOCK(p->p_session); 392 } 393 fixjobc(p, p->p_pgrp, 0); 394 sx_xunlock(&proctree_lock); 395 (void)acct_process(td); 396 #ifdef KTRACE 397 /* 398 * Drain any pending records on the thread and release the trace 399 * file. It might be better if drain-and-clear were atomic. 400 */ 401 ktrprocexit(td); 402 PROC_LOCK(p); 403 mtx_lock(&ktrace_mtx); 404 p->p_traceflag = 0; /* don't trace the vrele() */ 405 tracevp = p->p_tracevp; 406 p->p_tracevp = NULL; 407 tracecred = p->p_tracecred; 408 p->p_tracecred = NULL; 409 mtx_unlock(&ktrace_mtx); 410 PROC_UNLOCK(p); 411 if (tracevp != NULL) { 412 locked = VFS_LOCK_GIANT(tracevp->v_mount); 413 vrele(tracevp); 414 VFS_UNLOCK_GIANT(locked); 415 } 416 if (tracecred != NULL) 417 crfree(tracecred); 418 #endif 419 /* 420 * Release reference to text vnode 421 */ 422 if ((vtmp = p->p_textvp) != NULL) { 423 p->p_textvp = NULL; 424 locked = VFS_LOCK_GIANT(vtmp->v_mount); 425 vrele(vtmp); 426 VFS_UNLOCK_GIANT(locked); 427 } 428 429 /* 430 * Release our limits structure. 431 */ 432 PROC_LOCK(p); 433 plim = p->p_limit; 434 p->p_limit = NULL; 435 PROC_UNLOCK(p); 436 lim_free(plim); 437 438 /* 439 * Remove proc from allproc queue and pidhash chain. 440 * Place onto zombproc. Unlink from parent's child list. 441 */ 442 sx_xlock(&allproc_lock); 443 LIST_REMOVE(p, p_list); 444 LIST_INSERT_HEAD(&zombproc, p, p_list); 445 LIST_REMOVE(p, p_hash); 446 sx_xunlock(&allproc_lock); 447 448 /* 449 * Reparent all of our children to init. 450 */ 451 sx_xlock(&proctree_lock); 452 q = LIST_FIRST(&p->p_children); 453 if (q != NULL) /* only need this if any child is S_ZOMB */ 454 wakeup(initproc); 455 for (; q != NULL; q = nq) { 456 nq = LIST_NEXT(q, p_sibling); 457 PROC_LOCK(q); 458 proc_reparent(q, initproc); 459 q->p_sigparent = SIGCHLD; 460 /* 461 * Traced processes are killed 462 * since their existence means someone is screwing up. 463 */ 464 if (q->p_flag & P_TRACED) { 465 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE); 466 psignal(q, SIGKILL); 467 } 468 PROC_UNLOCK(q); 469 } 470 471 /* 472 * Save exit status and finalize rusage info except for times, 473 * adding in child rusage info later when our time is locked. 474 */ 475 PROC_LOCK(p); 476 p->p_xstat = rv; 477 p->p_xthread = td; 478 p->p_stats->p_ru.ru_nvcsw++; 479 *p->p_ru = p->p_stats->p_ru; 480 481 /* 482 * Notify interested parties of our demise. 483 */ 484 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT); 485 486 /* 487 * Just delete all entries in the p_klist. At this point we won't 488 * report any more events, and there are nasty race conditions that 489 * can beat us if we don't. 490 */ 491 knlist_clear(&p->p_klist, 1); 492 493 /* 494 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 495 * flag set, or if the handler is set to SIG_IGN, notify process 496 * 1 instead (and hope it will handle this situation). 497 */ 498 PROC_LOCK(p->p_pptr); 499 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx); 500 if (p->p_pptr->p_sigacts->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) { 501 struct proc *pp; 502 503 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx); 504 pp = p->p_pptr; 505 PROC_UNLOCK(pp); 506 proc_reparent(p, initproc); 507 p->p_sigparent = SIGCHLD; 508 PROC_LOCK(p->p_pptr); 509 /* 510 * If this was the last child of our parent, notify 511 * parent, so in case he was wait(2)ing, he will 512 * continue. 513 */ 514 if (LIST_EMPTY(&pp->p_children)) 515 wakeup(pp); 516 } else 517 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx); 518 519 if (p->p_pptr == initproc) 520 psignal(p->p_pptr, SIGCHLD); 521 else if (p->p_sigparent != 0) { 522 if (p->p_sigparent == SIGCHLD) 523 childproc_exited(p); 524 else /* LINUX thread */ 525 psignal(p->p_pptr, p->p_sigparent); 526 } 527 PROC_UNLOCK(p->p_pptr); 528 PROC_UNLOCK(p); 529 530 /* 531 * Finally, call machine-dependent code to release the remaining 532 * resources including address space. 533 * The address space is released by "vmspace_exitfree(p)" in 534 * vm_waitproc(). 535 */ 536 cpu_exit(td); 537 538 WITNESS_WARN(WARN_PANIC, &proctree_lock.sx_object, 539 "process (pid %d) exiting", p->p_pid); 540 541 PROC_LOCK(p); 542 PROC_LOCK(p->p_pptr); 543 sx_xunlock(&proctree_lock); 544 545 /* 546 * We have to wait until after acquiring all locks before 547 * changing p_state. We need to avoid all possible context 548 * switches (including ones from blocking on a mutex) while 549 * marked as a zombie. We also have to set the zombie state 550 * before we release the parent process' proc lock to avoid 551 * a lost wakeup. So, we first call wakeup, then we grab the 552 * sched lock, update the state, and release the parent process' 553 * proc lock. 554 */ 555 wakeup(p->p_pptr); 556 mtx_lock_spin(&sched_lock); 557 p->p_state = PRS_ZOMBIE; 558 PROC_UNLOCK(p->p_pptr); 559 560 /* Do the same timestamp bookkeeping that mi_switch() would do. */ 561 new_switchtime = cpu_ticks(); 562 p->p_rux.rux_runtime += (new_switchtime - PCPU_GET(switchtime)); 563 p->p_rux.rux_uticks += td->td_uticks; 564 p->p_rux.rux_sticks += td->td_sticks; 565 p->p_rux.rux_iticks += td->td_iticks; 566 PCPU_SET(switchtime, new_switchtime); 567 PCPU_SET(switchticks, ticks); 568 cnt.v_swtch++; 569 570 /* Add our usage into the usage of all our children. */ 571 ruadd(p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux); 572 573 sched_exit(p->p_pptr, td); 574 575 /* 576 * Hopefully no one will try to deliver a signal to the process this 577 * late in the game. 578 */ 579 knlist_destroy(&p->p_klist); 580 581 /* 582 * Make sure the scheduler takes this thread out of its tables etc. 583 * This will also release this thread's reference to the ucred. 584 * Other thread parts to release include pcb bits and such. 585 */ 586 thread_exit(); 587 } 588 589 590 #ifndef _SYS_SYSPROTO_H_ 591 struct abort2_args { 592 char *why; 593 int nargs; 594 void **args; 595 }; 596 #endif 597 598 /* 599 * MPSAFE. 600 */ 601 int 602 abort2(struct thread *td, struct abort2_args *uap) 603 { 604 struct proc *p = td->td_proc; 605 struct sbuf *sb; 606 void *uargs[16]; 607 int error, i, sig; 608 609 error = 0; /* satisfy compiler */ 610 611 /* 612 * Do it right now so we can log either proper call of abort2(), or 613 * note, that invalid argument was passed. 512 is big enough to 614 * handle 16 arguments' descriptions with additional comments. 615 */ 616 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN); 617 sbuf_clear(sb); 618 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ", 619 p->p_comm, p->p_pid, td->td_ucred->cr_uid); 620 /* 621 * Since we can't return from abort2(), send SIGKILL in cases, where 622 * abort2() was called improperly 623 */ 624 sig = SIGKILL; 625 /* Prevent from DoSes from user-space. */ 626 if (uap->nargs < 0 || uap->nargs > 16) 627 goto out; 628 if (uap->args == NULL) 629 goto out; 630 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *)); 631 if (error != 0) 632 goto out; 633 /* 634 * Limit size of 'reason' string to 128. Will fit even when 635 * maximal number of arguments was chosen to be logged. 636 */ 637 if (uap->why != NULL) { 638 error = sbuf_copyin(sb, uap->why, 128); 639 if (error < 0) 640 goto out; 641 } else { 642 sbuf_printf(sb, "(null)"); 643 } 644 if (uap->nargs) { 645 sbuf_printf(sb, "("); 646 for (i = 0;i < uap->nargs; i++) 647 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]); 648 sbuf_printf(sb, ")"); 649 } 650 /* 651 * Final stage: arguments were proper, string has been 652 * successfully copied from userspace, and copying pointers 653 * from user-space succeed. 654 */ 655 sig = SIGABRT; 656 out: 657 if (sig == SIGKILL) { 658 sbuf_trim(sb); 659 sbuf_printf(sb, " (Reason text inaccessible)"); 660 } 661 sbuf_cat(sb, "\n"); 662 sbuf_finish(sb); 663 log(LOG_INFO, "%s", sbuf_data(sb)); 664 sbuf_delete(sb); 665 exit1(td, W_EXITCODE(0, sig)); 666 return (0); 667 } 668 669 670 #ifdef COMPAT_43 671 /* 672 * The dirty work is handled by kern_wait(). 673 * 674 * MPSAFE. 675 */ 676 int 677 owait(struct thread *td, struct owait_args *uap __unused) 678 { 679 int error, status; 680 681 error = kern_wait(td, WAIT_ANY, &status, 0, NULL); 682 if (error == 0) 683 td->td_retval[1] = status; 684 return (error); 685 } 686 #endif /* COMPAT_43 */ 687 688 /* 689 * The dirty work is handled by kern_wait(). 690 * 691 * MPSAFE. 692 */ 693 int 694 wait4(struct thread *td, struct wait_args *uap) 695 { 696 struct rusage ru, *rup; 697 int error, status; 698 699 if (uap->rusage != NULL) 700 rup = &ru; 701 else 702 rup = NULL; 703 error = kern_wait(td, uap->pid, &status, uap->options, rup); 704 if (uap->status != NULL && error == 0) 705 error = copyout(&status, uap->status, sizeof(status)); 706 if (uap->rusage != NULL && error == 0) 707 error = copyout(&ru, uap->rusage, sizeof(struct rusage)); 708 return (error); 709 } 710 711 int 712 kern_wait(struct thread *td, pid_t pid, int *status, int options, 713 struct rusage *rusage) 714 { 715 struct proc *p, *q, *t; 716 int error, nfound; 717 718 AUDIT_ARG(pid, pid); 719 720 q = td->td_proc; 721 if (pid == 0) { 722 PROC_LOCK(q); 723 pid = -q->p_pgid; 724 PROC_UNLOCK(q); 725 } 726 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED|WLINUXCLONE)) 727 return (EINVAL); 728 loop: 729 if (q->p_flag & P_STATCHILD) { 730 PROC_LOCK(q); 731 q->p_flag &= ~P_STATCHILD; 732 PROC_UNLOCK(q); 733 } 734 nfound = 0; 735 sx_xlock(&proctree_lock); 736 LIST_FOREACH(p, &q->p_children, p_sibling) { 737 PROC_LOCK(p); 738 if (pid != WAIT_ANY && 739 p->p_pid != pid && p->p_pgid != -pid) { 740 PROC_UNLOCK(p); 741 continue; 742 } 743 if (p_canwait(td, p)) { 744 PROC_UNLOCK(p); 745 continue; 746 } 747 748 /* 749 * This special case handles a kthread spawned by linux_clone 750 * (see linux_misc.c). The linux_wait4 and linux_waitpid 751 * functions need to be able to distinguish between waiting 752 * on a process and waiting on a thread. It is a thread if 753 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option 754 * signifies we want to wait for threads and not processes. 755 */ 756 if ((p->p_sigparent != SIGCHLD) ^ 757 ((options & WLINUXCLONE) != 0)) { 758 PROC_UNLOCK(p); 759 continue; 760 } 761 762 nfound++; 763 if (p->p_state == PRS_ZOMBIE) { 764 765 /* 766 * It is possible that the last thread of this 767 * process is still running on another CPU 768 * in thread_exit() after having dropped the process 769 * lock via PROC_UNLOCK() but before it has completed 770 * cpu_throw(). In that case, the other thread must 771 * still hold sched_lock, so simply by acquiring 772 * sched_lock once we will wait long enough for the 773 * thread to exit in that case. 774 */ 775 mtx_lock_spin(&sched_lock); 776 mtx_unlock_spin(&sched_lock); 777 778 td->td_retval[0] = p->p_pid; 779 if (status) 780 *status = p->p_xstat; /* convert to int */ 781 if (rusage) { 782 *rusage = *p->p_ru; 783 calcru(p, &rusage->ru_utime, &rusage->ru_stime); 784 } 785 786 PROC_LOCK(q); 787 sigqueue_take(p->p_ksi); 788 PROC_UNLOCK(q); 789 790 /* 791 * If we got the child via a ptrace 'attach', 792 * we need to give it back to the old parent. 793 */ 794 PROC_UNLOCK(p); 795 if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) { 796 PROC_LOCK(p); 797 p->p_oppid = 0; 798 proc_reparent(p, t); 799 PROC_UNLOCK(p); 800 tdsignal(t, NULL, SIGCHLD, p->p_ksi); 801 wakeup(t); 802 PROC_UNLOCK(t); 803 sx_xunlock(&proctree_lock); 804 return (0); 805 } 806 807 /* 808 * Remove other references to this process to ensure 809 * we have an exclusive reference. 810 */ 811 sx_xlock(&allproc_lock); 812 LIST_REMOVE(p, p_list); /* off zombproc */ 813 sx_xunlock(&allproc_lock); 814 LIST_REMOVE(p, p_sibling); 815 leavepgrp(p); 816 sx_xunlock(&proctree_lock); 817 818 /* 819 * As a side effect of this lock, we know that 820 * all other writes to this proc are visible now, so 821 * no more locking is needed for p. 822 */ 823 PROC_LOCK(p); 824 p->p_xstat = 0; /* XXX: why? */ 825 PROC_UNLOCK(p); 826 PROC_LOCK(q); 827 ruadd(&q->p_stats->p_cru, &q->p_crux, p->p_ru, 828 &p->p_rux); 829 PROC_UNLOCK(q); 830 FREE(p->p_ru, M_ZOMBIE); 831 p->p_ru = NULL; 832 833 /* 834 * Decrement the count of procs running with this uid. 835 */ 836 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 837 838 /* 839 * Free credentials, arguments, and sigacts. 840 */ 841 crfree(p->p_ucred); 842 p->p_ucred = NULL; 843 pargs_drop(p->p_args); 844 p->p_args = NULL; 845 sigacts_free(p->p_sigacts); 846 p->p_sigacts = NULL; 847 848 /* 849 * Do any thread-system specific cleanups. 850 */ 851 thread_wait(p); 852 853 /* 854 * Give vm and machine-dependent layer a chance 855 * to free anything that cpu_exit couldn't 856 * release while still running in process context. 857 */ 858 vm_waitproc(p); 859 #ifdef MAC 860 mac_destroy_proc(p); 861 #endif 862 #ifdef AUDIT 863 audit_proc_free(p); 864 #endif 865 KASSERT(FIRST_THREAD_IN_PROC(p), 866 ("kern_wait: no residual thread!")); 867 uma_zfree(proc_zone, p); 868 sx_xlock(&allproc_lock); 869 nprocs--; 870 sx_xunlock(&allproc_lock); 871 return (0); 872 } 873 mtx_lock_spin(&sched_lock); 874 if ((p->p_flag & P_STOPPED_SIG) && 875 (p->p_suspcount == p->p_numthreads) && 876 (p->p_flag & P_WAITED) == 0 && 877 (p->p_flag & P_TRACED || options & WUNTRACED)) { 878 mtx_unlock_spin(&sched_lock); 879 p->p_flag |= P_WAITED; 880 sx_xunlock(&proctree_lock); 881 td->td_retval[0] = p->p_pid; 882 if (status) 883 *status = W_STOPCODE(p->p_xstat); 884 PROC_UNLOCK(p); 885 886 PROC_LOCK(q); 887 sigqueue_take(p->p_ksi); 888 PROC_UNLOCK(q); 889 890 return (0); 891 } 892 mtx_unlock_spin(&sched_lock); 893 if (options & WCONTINUED && (p->p_flag & P_CONTINUED)) { 894 sx_xunlock(&proctree_lock); 895 td->td_retval[0] = p->p_pid; 896 p->p_flag &= ~P_CONTINUED; 897 PROC_UNLOCK(p); 898 899 PROC_LOCK(q); 900 sigqueue_take(p->p_ksi); 901 PROC_UNLOCK(q); 902 903 if (status) 904 *status = SIGCONT; 905 return (0); 906 } 907 PROC_UNLOCK(p); 908 } 909 if (nfound == 0) { 910 sx_xunlock(&proctree_lock); 911 return (ECHILD); 912 } 913 if (options & WNOHANG) { 914 sx_xunlock(&proctree_lock); 915 td->td_retval[0] = 0; 916 return (0); 917 } 918 PROC_LOCK(q); 919 sx_xunlock(&proctree_lock); 920 if (q->p_flag & P_STATCHILD) { 921 q->p_flag &= ~P_STATCHILD; 922 error = 0; 923 } else 924 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0); 925 PROC_UNLOCK(q); 926 if (error) 927 return (error); 928 goto loop; 929 } 930 931 /* 932 * Make process 'parent' the new parent of process 'child'. 933 * Must be called with an exclusive hold of proctree lock. 934 */ 935 void 936 proc_reparent(struct proc *child, struct proc *parent) 937 { 938 939 sx_assert(&proctree_lock, SX_XLOCKED); 940 PROC_LOCK_ASSERT(child, MA_OWNED); 941 if (child->p_pptr == parent) 942 return; 943 944 LIST_REMOVE(child, p_sibling); 945 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 946 child->p_pptr = parent; 947 } 948