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 <vm/vm.h> 76 #include <vm/vm_extern.h> 77 #include <vm/vm_param.h> 78 #include <vm/pmap.h> 79 #include <vm/vm_map.h> 80 #include <vm/vm_page.h> 81 #include <vm/uma.h> 82 83 /* Required to be non-static for SysVR4 emulator */ 84 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 85 86 /* Hook for NFS teardown procedure. */ 87 void (*nlminfo_release_p)(struct proc *p); 88 89 /* 90 * exit -- 91 * Death of process. 92 * 93 * MPSAFE 94 */ 95 void 96 sys_exit(struct thread *td, struct sys_exit_args *uap) 97 { 98 99 exit1(td, W_EXITCODE(uap->rval, 0)); 100 /* NOTREACHED */ 101 } 102 103 /* 104 * Exit: deallocate address space and other resources, change proc state 105 * to zombie, and unlink proc from allproc and parent's lists. Save exit 106 * status and rusage for wait(). Check for child processes and orphan them. 107 */ 108 void 109 exit1(struct thread *td, int rv) 110 { 111 struct bintime new_switchtime; 112 struct proc *p, *nq, *q; 113 struct tty *tp; 114 struct vnode *ttyvp; 115 struct vmspace *vm; 116 struct vnode *vtmp; 117 #ifdef KTRACE 118 struct vnode *tracevp; 119 struct ucred *tracecred; 120 #endif 121 struct plimit *plim; 122 int locked, refcnt; 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) { 143 retry: 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 p->p_flag |= P_WEXIT; 177 178 PROC_LOCK(p->p_pptr); 179 sigqueue_take(p->p_ksi); 180 PROC_UNLOCK(p->p_pptr); 181 182 PROC_UNLOCK(p); 183 184 /* Are we a task leader? */ 185 if (p == p->p_leader) { 186 mtx_lock(&ppeers_lock); 187 q = p->p_peers; 188 while (q != NULL) { 189 PROC_LOCK(q); 190 psignal(q, SIGKILL); 191 PROC_UNLOCK(q); 192 q = q->p_peers; 193 } 194 while (p->p_peers != NULL) 195 msleep(p, &ppeers_lock, PWAIT, "exit1", 0); 196 mtx_unlock(&ppeers_lock); 197 } 198 199 PROC_LOCK(p); 200 _STOPEVENT(p, S_EXIT, rv); 201 wakeup(&p->p_stype); /* Wakeup anyone in procfs' PIOCWAIT */ 202 PROC_UNLOCK(p); 203 204 /* 205 * Check if any loadable modules need anything done at process exit. 206 * E.g. SYSV IPC stuff 207 * XXX what if one of these generates an error? 208 */ 209 EVENTHANDLER_INVOKE(process_exit, p); 210 211 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage), 212 M_ZOMBIE, M_WAITOK); 213 /* 214 * If parent is waiting for us to exit or exec, 215 * P_PPWAIT is set; we will wakeup the parent below. 216 */ 217 PROC_LOCK(p); 218 stopprofclock(p); 219 p->p_flag &= ~(P_TRACED | P_PPWAIT); 220 221 /* 222 * Stop the real interval timer. If the handler is currently 223 * executing, prevent it from rearming itself and let it finish. 224 */ 225 if (timevalisset(&p->p_realtimer.it_value) && 226 callout_stop(&p->p_itcallout) == 0) { 227 timevalclear(&p->p_realtimer.it_interval); 228 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0); 229 KASSERT(!timevalisset(&p->p_realtimer.it_value), 230 ("realtime timer is still armed")); 231 } 232 sigqueue_flush(&p->p_sigqueue); 233 sigqueue_flush(&td->td_sigqueue); 234 PROC_UNLOCK(p); 235 236 /* 237 * Reset any sigio structures pointing to us as a result of 238 * F_SETOWN with our pid. 239 */ 240 mtx_lock(&Giant); /* XXX: not sure if needed */ 241 funsetownlst(&p->p_sigiolst); 242 mtx_unlock(&Giant); 243 244 /* 245 * If this process has an nlminfo data area (for lockd), release it 246 */ 247 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL) 248 (*nlminfo_release_p)(p); 249 250 /* 251 * Close open files and release open-file table. 252 * This may block! 253 */ 254 fdfree(td); 255 256 /* 257 * If this thread tickled GEOM, we need to wait for the giggling to 258 * stop before we return to userland 259 */ 260 if (td->td_pflags & TDP_GEOM) 261 g_waitidle(); 262 263 /* 264 * Remove ourself from our leader's peer list and wake our leader. 265 */ 266 mtx_lock(&ppeers_lock); 267 if (p->p_leader->p_peers) { 268 q = p->p_leader; 269 while (q->p_peers != p) 270 q = q->p_peers; 271 q->p_peers = p->p_peers; 272 wakeup(p->p_leader); 273 } 274 mtx_unlock(&ppeers_lock); 275 276 /* The next two chunks should probably be moved to vmspace_exit. */ 277 vm = p->p_vmspace; 278 /* 279 * Release user portion of address space. 280 * This releases references to vnodes, 281 * which could cause I/O if the file has been unlinked. 282 * Need to do this early enough that we can still sleep. 283 * Can't free the entire vmspace as the kernel stack 284 * may be mapped within that space also. 285 * 286 * Processes sharing the same vmspace may exit in one order, and 287 * get cleaned up by vmspace_exit() in a different order. The 288 * last exiting process to reach this point releases as much of 289 * the environment as it can, and the last process cleaned up 290 * by vmspace_exit() (which decrements exitingcnt) cleans up the 291 * remainder. 292 */ 293 atomic_add_int(&vm->vm_exitingcnt, 1); 294 do 295 refcnt = vm->vm_refcnt; 296 while (!atomic_cmpset_int(&vm->vm_refcnt, refcnt, refcnt - 1)); 297 if (refcnt == 1) { 298 shmexit(vm); 299 pmap_remove_pages(vmspace_pmap(vm), vm_map_min(&vm->vm_map), 300 vm_map_max(&vm->vm_map)); 301 (void) vm_map_remove(&vm->vm_map, vm_map_min(&vm->vm_map), 302 vm_map_max(&vm->vm_map)); 303 } 304 305 sx_xlock(&proctree_lock); 306 if (SESS_LEADER(p)) { 307 struct session *sp; 308 309 sp = p->p_session; 310 if (sp->s_ttyvp) { 311 locked = VFS_LOCK_GIANT(sp->s_ttyvp->v_mount); 312 /* 313 * Controlling process. 314 * Signal foreground pgrp, 315 * drain controlling terminal 316 * and revoke access to controlling terminal. 317 */ 318 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 319 tp = sp->s_ttyp; 320 if (sp->s_ttyp->t_pgrp) { 321 PGRP_LOCK(sp->s_ttyp->t_pgrp); 322 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 323 PGRP_UNLOCK(sp->s_ttyp->t_pgrp); 324 } 325 /* XXX tp should be locked. */ 326 sx_xunlock(&proctree_lock); 327 (void) ttywait(tp); 328 sx_xlock(&proctree_lock); 329 /* 330 * The tty could have been revoked 331 * if we blocked. 332 */ 333 if (sp->s_ttyvp) { 334 ttyvp = sp->s_ttyvp; 335 SESS_LOCK(p->p_session); 336 sp->s_ttyvp = NULL; 337 SESS_UNLOCK(p->p_session); 338 sx_xunlock(&proctree_lock); 339 VOP_LOCK(ttyvp, LK_EXCLUSIVE, td); 340 VOP_REVOKE(ttyvp, REVOKEALL); 341 vput(ttyvp); 342 sx_xlock(&proctree_lock); 343 } 344 } 345 if (sp->s_ttyvp) { 346 ttyvp = sp->s_ttyvp; 347 SESS_LOCK(p->p_session); 348 sp->s_ttyvp = NULL; 349 SESS_UNLOCK(p->p_session); 350 vrele(ttyvp); 351 } 352 /* 353 * s_ttyp is not zero'd; we use this to indicate 354 * that the session once had a controlling terminal. 355 * (for logging and informational purposes) 356 */ 357 VFS_UNLOCK_GIANT(locked); 358 } 359 SESS_LOCK(p->p_session); 360 sp->s_leader = NULL; 361 SESS_UNLOCK(p->p_session); 362 } 363 fixjobc(p, p->p_pgrp, 0); 364 sx_xunlock(&proctree_lock); 365 (void)acct_process(td); 366 #ifdef KTRACE 367 /* 368 * Drain any pending records on the thread and release the trace 369 * file. It might be better if drain-and-clear were atomic. 370 */ 371 ktrprocexit(td); 372 PROC_LOCK(p); 373 mtx_lock(&ktrace_mtx); 374 p->p_traceflag = 0; /* don't trace the vrele() */ 375 tracevp = p->p_tracevp; 376 p->p_tracevp = NULL; 377 tracecred = p->p_tracecred; 378 p->p_tracecred = NULL; 379 mtx_unlock(&ktrace_mtx); 380 PROC_UNLOCK(p); 381 if (tracevp != NULL) { 382 locked = VFS_LOCK_GIANT(tracevp->v_mount); 383 vrele(tracevp); 384 VFS_UNLOCK_GIANT(locked); 385 } 386 if (tracecred != NULL) 387 crfree(tracecred); 388 #endif 389 /* 390 * Release reference to text vnode 391 */ 392 if ((vtmp = p->p_textvp) != NULL) { 393 p->p_textvp = NULL; 394 locked = VFS_LOCK_GIANT(vtmp->v_mount); 395 vrele(vtmp); 396 VFS_UNLOCK_GIANT(locked); 397 } 398 399 /* 400 * Release our limits structure. 401 */ 402 PROC_LOCK(p); 403 plim = p->p_limit; 404 p->p_limit = NULL; 405 PROC_UNLOCK(p); 406 lim_free(plim); 407 408 /* 409 * Remove proc from allproc queue and pidhash chain. 410 * Place onto zombproc. Unlink from parent's child list. 411 */ 412 sx_xlock(&allproc_lock); 413 LIST_REMOVE(p, p_list); 414 LIST_INSERT_HEAD(&zombproc, p, p_list); 415 LIST_REMOVE(p, p_hash); 416 sx_xunlock(&allproc_lock); 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. 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 ruadd(p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux); 448 449 /* 450 * Notify interested parties of our demise. 451 */ 452 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT); 453 454 /* 455 * Just delete all entries in the p_klist. At this point we won't 456 * report any more events, and there are nasty race conditions that 457 * can beat us if we don't. 458 */ 459 knlist_clear(&p->p_klist, 1); 460 461 /* 462 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 463 * flag set, or if the handler is set to SIG_IGN, notify process 464 * 1 instead (and hope it will handle this situation). 465 */ 466 PROC_LOCK(p->p_pptr); 467 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx); 468 if (p->p_pptr->p_sigacts->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) { 469 struct proc *pp; 470 471 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx); 472 pp = p->p_pptr; 473 PROC_UNLOCK(pp); 474 proc_reparent(p, initproc); 475 p->p_sigparent = SIGCHLD; 476 PROC_LOCK(p->p_pptr); 477 /* 478 * If this was the last child of our parent, notify 479 * parent, so in case he was wait(2)ing, he will 480 * continue. 481 */ 482 if (LIST_EMPTY(&pp->p_children)) 483 wakeup(pp); 484 } else 485 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx); 486 487 if (p->p_pptr == initproc) 488 psignal(p->p_pptr, SIGCHLD); 489 else if (p->p_sigparent != 0) { 490 if (p->p_sigparent == SIGCHLD) 491 childproc_exited(p); 492 else /* LINUX thread */ 493 psignal(p->p_pptr, p->p_sigparent); 494 } 495 PROC_UNLOCK(p->p_pptr); 496 497 /* 498 * If this is a kthread, then wakeup anyone waiting for it to exit. 499 */ 500 if (p->p_flag & P_KTHREAD) 501 wakeup(p); 502 PROC_UNLOCK(p); 503 504 /* 505 * Finally, call machine-dependent code to release the remaining 506 * resources including address space. 507 * The address space is released by "vmspace_exitfree(p)" in 508 * vm_waitproc(). 509 */ 510 cpu_exit(td); 511 512 WITNESS_WARN(WARN_PANIC, &proctree_lock.sx_object, 513 "process (pid %d) exiting", p->p_pid); 514 515 PROC_LOCK(p); 516 PROC_LOCK(p->p_pptr); 517 sx_xunlock(&proctree_lock); 518 519 /* 520 * We have to wait until after acquiring all locks before 521 * changing p_state. We need to avoid all possible context 522 * switches (including ones from blocking on a mutex) while 523 * marked as a zombie. We also have to set the zombie state 524 * before we release the parent process' proc lock to avoid 525 * a lost wakeup. So, we first call wakeup, then we grab the 526 * sched lock, update the state, and release the parent process' 527 * proc lock. 528 */ 529 wakeup(p->p_pptr); 530 mtx_lock_spin(&sched_lock); 531 p->p_state = PRS_ZOMBIE; 532 PROC_UNLOCK(p->p_pptr); 533 534 /* Do the same timestamp bookkeeping that mi_switch() would do. */ 535 binuptime(&new_switchtime); 536 bintime_add(&p->p_rux.rux_runtime, &new_switchtime); 537 bintime_sub(&p->p_rux.rux_runtime, PCPU_PTR(switchtime)); 538 PCPU_SET(switchtime, new_switchtime); 539 PCPU_SET(switchticks, ticks); 540 cnt.v_swtch++; 541 542 sched_exit(p->p_pptr, td); 543 544 /* 545 * Hopefully no one will try to deliver a signal to the process this 546 * late in the game. 547 */ 548 knlist_destroy(&p->p_klist); 549 550 /* 551 * Make sure the scheduler takes this thread out of its tables etc. 552 * This will also release this thread's reference to the ucred. 553 * Other thread parts to release include pcb bits and such. 554 */ 555 thread_exit(); 556 } 557 558 559 #ifndef _SYS_SYSPROTO_H_ 560 struct abort2_args { 561 char *why; 562 int nargs; 563 void **args; 564 }; 565 #endif 566 567 /* 568 * MPSAFE. 569 */ 570 int 571 abort2(struct thread *td, struct abort2_args *uap) 572 { 573 struct proc *p = td->td_proc; 574 struct sbuf *sb; 575 void *uargs[16]; 576 int error, i, sig; 577 578 error = 0; /* satisfy compiler */ 579 580 /* 581 * Do it right now so we can log either proper call of abort2(), or 582 * note, that invalid argument was passed. 512 is big enough to 583 * handle 16 arguments' descriptions with additional comments. 584 */ 585 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN); 586 sbuf_clear(sb); 587 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ", 588 p->p_comm, p->p_pid, td->td_ucred->cr_uid); 589 /* 590 * Since we can't return from abort2(), send SIGKILL in cases, where 591 * abort2() was called improperly 592 */ 593 sig = SIGKILL; 594 /* Prevent from DoSes from user-space. */ 595 if (uap->nargs < 0 || uap->nargs > 16) 596 goto out; 597 if (uap->args == NULL) 598 goto out; 599 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *)); 600 if (error != 0) 601 goto out; 602 /* 603 * Limit size of 'reason' string to 128. Will fit even when 604 * maximal number of arguments was chosen to be logged. 605 */ 606 if (uap->why != NULL) { 607 error = sbuf_copyin(sb, uap->why, 128); 608 if (error < 0) 609 goto out; 610 } else { 611 sbuf_printf(sb, "(null)"); 612 } 613 if (uap->nargs) { 614 sbuf_printf(sb, "("); 615 for (i = 0;i < uap->nargs; i++) 616 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]); 617 sbuf_printf(sb, ")"); 618 } 619 /* 620 * Final stage: arguments were proper, string has been 621 * successfully copied from userspace, and copying pointers 622 * from user-space succeed. 623 */ 624 sig = SIGABRT; 625 out: 626 if (sig == SIGKILL) { 627 sbuf_trim(sb); 628 sbuf_printf(sb, " (Reason text inaccessible)"); 629 } 630 sbuf_cat(sb, "\n"); 631 sbuf_finish(sb); 632 log(LOG_INFO, "%s", sbuf_data(sb)); 633 sbuf_delete(sb); 634 exit1(td, W_EXITCODE(0, sig)); 635 return (0); 636 } 637 638 639 #ifdef COMPAT_43 640 /* 641 * The dirty work is handled by kern_wait(). 642 * 643 * MPSAFE. 644 */ 645 int 646 owait(struct thread *td, struct owait_args *uap __unused) 647 { 648 int error, status; 649 650 error = kern_wait(td, WAIT_ANY, &status, 0, NULL); 651 if (error == 0) 652 td->td_retval[1] = status; 653 return (error); 654 } 655 #endif /* COMPAT_43 */ 656 657 /* 658 * The dirty work is handled by kern_wait(). 659 * 660 * MPSAFE. 661 */ 662 int 663 wait4(struct thread *td, struct wait_args *uap) 664 { 665 struct rusage ru, *rup; 666 int error, status; 667 668 if (uap->rusage != NULL) 669 rup = &ru; 670 else 671 rup = NULL; 672 error = kern_wait(td, uap->pid, &status, uap->options, rup); 673 if (uap->status != NULL && error == 0) 674 error = copyout(&status, uap->status, sizeof(status)); 675 if (uap->rusage != NULL && error == 0) 676 error = copyout(&ru, uap->rusage, sizeof(struct rusage)); 677 return (error); 678 } 679 680 int 681 kern_wait(struct thread *td, pid_t pid, int *status, int options, 682 struct rusage *rusage) 683 { 684 struct proc *p, *q, *t; 685 int error, nfound; 686 687 q = td->td_proc; 688 if (pid == 0) { 689 PROC_LOCK(q); 690 pid = -q->p_pgid; 691 PROC_UNLOCK(q); 692 } 693 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED|WLINUXCLONE)) 694 return (EINVAL); 695 loop: 696 if (q->p_flag & P_STATCHILD) { 697 PROC_LOCK(q); 698 q->p_flag &= ~P_STATCHILD; 699 PROC_UNLOCK(q); 700 } 701 nfound = 0; 702 sx_xlock(&proctree_lock); 703 LIST_FOREACH(p, &q->p_children, p_sibling) { 704 PROC_LOCK(p); 705 if (pid != WAIT_ANY && 706 p->p_pid != pid && p->p_pgid != -pid) { 707 PROC_UNLOCK(p); 708 continue; 709 } 710 if (p_canwait(td, p)) { 711 PROC_UNLOCK(p); 712 continue; 713 } 714 715 /* 716 * This special case handles a kthread spawned by linux_clone 717 * (see linux_misc.c). The linux_wait4 and linux_waitpid 718 * functions need to be able to distinguish between waiting 719 * on a process and waiting on a thread. It is a thread if 720 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option 721 * signifies we want to wait for threads and not processes. 722 */ 723 if ((p->p_sigparent != SIGCHLD) ^ 724 ((options & WLINUXCLONE) != 0)) { 725 PROC_UNLOCK(p); 726 continue; 727 } 728 729 nfound++; 730 if (p->p_state == PRS_ZOMBIE) { 731 732 /* 733 * It is possible that the last thread of this 734 * process is still running on another CPU 735 * in thread_exit() after having dropped the process 736 * lock via PROC_UNLOCK() but before it has completed 737 * cpu_throw(). In that case, the other thread must 738 * still hold sched_lock, so simply by acquiring 739 * sched_lock once we will wait long enough for the 740 * thread to exit in that case. 741 */ 742 mtx_lock_spin(&sched_lock); 743 mtx_unlock_spin(&sched_lock); 744 745 td->td_retval[0] = p->p_pid; 746 if (status) 747 *status = p->p_xstat; /* convert to int */ 748 if (rusage) { 749 *rusage = *p->p_ru; 750 calcru(p, &rusage->ru_utime, &rusage->ru_stime); 751 } 752 753 PROC_LOCK(q); 754 sigqueue_take(p->p_ksi); 755 PROC_UNLOCK(q); 756 757 /* 758 * If we got the child via a ptrace 'attach', 759 * we need to give it back to the old parent. 760 */ 761 PROC_UNLOCK(p); 762 if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) { 763 PROC_LOCK(p); 764 p->p_oppid = 0; 765 proc_reparent(p, t); 766 PROC_UNLOCK(p); 767 tdsignal(t, NULL, SIGCHLD, p->p_ksi); 768 wakeup(t); 769 PROC_UNLOCK(t); 770 sx_xunlock(&proctree_lock); 771 return (0); 772 } 773 774 /* 775 * Remove other references to this process to ensure 776 * we have an exclusive reference. 777 */ 778 sx_xlock(&allproc_lock); 779 LIST_REMOVE(p, p_list); /* off zombproc */ 780 sx_xunlock(&allproc_lock); 781 LIST_REMOVE(p, p_sibling); 782 leavepgrp(p); 783 sx_xunlock(&proctree_lock); 784 785 /* 786 * As a side effect of this lock, we know that 787 * all other writes to this proc are visible now, so 788 * no more locking is needed for p. 789 */ 790 PROC_LOCK(p); 791 p->p_xstat = 0; /* XXX: why? */ 792 PROC_UNLOCK(p); 793 PROC_LOCK(q); 794 ruadd(&q->p_stats->p_cru, &q->p_crux, p->p_ru, 795 &p->p_rux); 796 PROC_UNLOCK(q); 797 FREE(p->p_ru, M_ZOMBIE); 798 p->p_ru = NULL; 799 800 /* 801 * Decrement the count of procs running with this uid. 802 */ 803 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 804 805 /* 806 * Free credentials, arguments, and sigacts. 807 */ 808 crfree(p->p_ucred); 809 p->p_ucred = NULL; 810 pargs_drop(p->p_args); 811 p->p_args = NULL; 812 sigacts_free(p->p_sigacts); 813 p->p_sigacts = NULL; 814 815 /* 816 * Do any thread-system specific cleanups. 817 */ 818 thread_wait(p); 819 820 /* 821 * Give vm and machine-dependent layer a chance 822 * to free anything that cpu_exit couldn't 823 * release while still running in process context. 824 */ 825 vm_waitproc(p); 826 #ifdef MAC 827 mac_destroy_proc(p); 828 #endif 829 KASSERT(FIRST_THREAD_IN_PROC(p), 830 ("kern_wait: no residual thread!")); 831 uma_zfree(proc_zone, p); 832 sx_xlock(&allproc_lock); 833 nprocs--; 834 sx_xunlock(&allproc_lock); 835 return (0); 836 } 837 mtx_lock_spin(&sched_lock); 838 if ((p->p_flag & P_STOPPED_SIG) && 839 (p->p_suspcount == p->p_numthreads) && 840 (p->p_flag & P_WAITED) == 0 && 841 (p->p_flag & P_TRACED || options & WUNTRACED)) { 842 mtx_unlock_spin(&sched_lock); 843 p->p_flag |= P_WAITED; 844 sx_xunlock(&proctree_lock); 845 td->td_retval[0] = p->p_pid; 846 if (status) 847 *status = W_STOPCODE(p->p_xstat); 848 PROC_UNLOCK(p); 849 850 PROC_LOCK(q); 851 sigqueue_take(p->p_ksi); 852 PROC_UNLOCK(q); 853 854 return (0); 855 } 856 mtx_unlock_spin(&sched_lock); 857 if (options & WCONTINUED && (p->p_flag & P_CONTINUED)) { 858 sx_xunlock(&proctree_lock); 859 td->td_retval[0] = p->p_pid; 860 p->p_flag &= ~P_CONTINUED; 861 PROC_UNLOCK(p); 862 863 PROC_LOCK(q); 864 sigqueue_take(p->p_ksi); 865 PROC_UNLOCK(q); 866 867 if (status) 868 *status = SIGCONT; 869 return (0); 870 } 871 PROC_UNLOCK(p); 872 } 873 if (nfound == 0) { 874 sx_xunlock(&proctree_lock); 875 return (ECHILD); 876 } 877 if (options & WNOHANG) { 878 sx_xunlock(&proctree_lock); 879 td->td_retval[0] = 0; 880 return (0); 881 } 882 PROC_LOCK(q); 883 sx_xunlock(&proctree_lock); 884 if (q->p_flag & P_STATCHILD) { 885 q->p_flag &= ~P_STATCHILD; 886 error = 0; 887 } else 888 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0); 889 PROC_UNLOCK(q); 890 if (error) 891 return (error); 892 goto loop; 893 } 894 895 /* 896 * Make process 'parent' the new parent of process 'child'. 897 * Must be called with an exclusive hold of proctree lock. 898 */ 899 void 900 proc_reparent(struct proc *child, struct proc *parent) 901 { 902 903 sx_assert(&proctree_lock, SX_XLOCKED); 904 PROC_LOCK_ASSERT(child, MA_OWNED); 905 if (child->p_pptr == parent) 906 return; 907 908 LIST_REMOVE(child, p_sibling); 909 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 910 child->p_pptr = parent; 911 } 912