kern_exit.c revision 195509
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 195509 2009-07-09 18:54:38Z kib $"); 39 40#include "opt_compat.h" 41#include "opt_kdtrace.h" 42#include "opt_ktrace.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/jail.h> 55#include <sys/tty.h> 56#include <sys/wait.h> 57#include <sys/vmmeter.h> 58#include <sys/vnode.h> 59#include <sys/resourcevar.h> 60#include <sys/sbuf.h> 61#include <sys/signalvar.h> 62#include <sys/sched.h> 63#include <sys/sx.h> 64#include <sys/syscallsubr.h> 65#include <sys/syslog.h> 66#include <sys/ptrace.h> 67#include <sys/acct.h> /* for acct_process() function prototype */ 68#include <sys/filedesc.h> 69#include <sys/sdt.h> 70#include <sys/shm.h> 71#include <sys/sem.h> 72#include <sys/vimage.h> 73#ifdef KTRACE 74#include <sys/ktrace.h> 75#endif 76 77#include <security/audit/audit.h> 78#include <security/mac/mac_framework.h> 79 80#include <vm/vm.h> 81#include <vm/vm_extern.h> 82#include <vm/vm_param.h> 83#include <vm/pmap.h> 84#include <vm/vm_map.h> 85#include <vm/vm_page.h> 86#include <vm/uma.h> 87 88#ifdef KDTRACE_HOOKS 89#include <sys/dtrace_bsd.h> 90dtrace_execexit_func_t dtrace_fasttrap_exit; 91#endif 92 93SDT_PROVIDER_DECLARE(proc); 94SDT_PROBE_DEFINE(proc, kernel, , exit); 95SDT_PROBE_ARGTYPE(proc, kernel, , exit, 0, "int"); 96 97/* Required to be non-static for SysVR4 emulator */ 98MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 99 100/* Hook for NFS teardown procedure. */ 101void (*nlminfo_release_p)(struct proc *p); 102 103/* 104 * exit -- death of process. 105 */ 106void 107sys_exit(struct thread *td, struct sys_exit_args *uap) 108{ 109 110 exit1(td, W_EXITCODE(uap->rval, 0)); 111 /* NOTREACHED */ 112} 113 114/* 115 * Exit: deallocate address space and other resources, change proc state to 116 * zombie, and unlink proc from allproc and parent's lists. Save exit status 117 * and rusage for wait(). Check for child processes and orphan them. 118 */ 119void 120exit1(struct thread *td, int rv) 121{ 122 struct proc *p, *nq, *q; 123 struct vnode *vtmp; 124 struct vnode *ttyvp = NULL; 125#ifdef KTRACE 126 struct vnode *tracevp; 127 struct ucred *tracecred; 128#endif 129 struct plimit *plim; 130 int locked; 131 132 mtx_assert(&Giant, MA_NOTOWNED); 133 134 p = td->td_proc; 135 if (p == initproc) { 136 printf("init died (signal %d, exit %d)\n", 137 WTERMSIG(rv), WEXITSTATUS(rv)); 138 panic("Going nowhere without my init!"); 139 } 140 141 /* 142 * MUST abort all other threads before proceeding past here. 143 */ 144 PROC_LOCK(p); 145 while (p->p_flag & P_HADTHREADS) { 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 break; 171 172 /* 173 * All other activity in this process is now stopped. 174 * Threading support has been turned off. 175 */ 176 } 177 KASSERT(p->p_numthreads == 1, 178 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads)); 179 /* 180 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold 181 * on our vmspace, so we should block below until they have 182 * released their reference to us. Note that if they have 183 * requested S_EXIT stops we will block here until they ack 184 * via PIOCCONT. 185 */ 186 _STOPEVENT(p, S_EXIT, rv); 187 188 /* 189 * Note that we are exiting and do another wakeup of anyone in 190 * PIOCWAIT in case they aren't listening for S_EXIT stops or 191 * decided to wait again after we told them we are exiting. 192 */ 193 p->p_flag |= P_WEXIT; 194 wakeup(&p->p_stype); 195 196 /* 197 * Wait for any processes that have a hold on our vmspace to 198 * release their reference. 199 */ 200 while (p->p_lock > 0) 201 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0); 202 203 PROC_UNLOCK(p); 204 /* Drain the limit callout while we don't have the proc locked */ 205 callout_drain(&p->p_limco); 206 207#ifdef AUDIT 208 /* 209 * The Sun BSM exit token contains two components: an exit status as 210 * passed to exit(), and a return value to indicate what sort of exit 211 * it was. The exit status is WEXITSTATUS(rv), but it's not clear 212 * what the return value is. 213 */ 214 AUDIT_ARG_EXIT(WEXITSTATUS(rv), 0); 215 AUDIT_SYSCALL_EXIT(0, td); 216#endif 217 218 /* Are we a task leader? */ 219 if (p == p->p_leader) { 220 mtx_lock(&ppeers_lock); 221 q = p->p_peers; 222 while (q != NULL) { 223 PROC_LOCK(q); 224 psignal(q, SIGKILL); 225 PROC_UNLOCK(q); 226 q = q->p_peers; 227 } 228 while (p->p_peers != NULL) 229 msleep(p, &ppeers_lock, PWAIT, "exit1", 0); 230 mtx_unlock(&ppeers_lock); 231 } 232 233 /* 234 * Check if any loadable modules need anything done at process exit. 235 * E.g. SYSV IPC stuff 236 * XXX what if one of these generates an error? 237 */ 238 EVENTHANDLER_INVOKE(process_exit, p); 239 240 /* 241 * If parent is waiting for us to exit or exec, 242 * P_PPWAIT is set; we will wakeup the parent below. 243 */ 244 PROC_LOCK(p); 245 stopprofclock(p); 246 p->p_flag &= ~(P_TRACED | P_PPWAIT); 247 248 /* 249 * Stop the real interval timer. If the handler is currently 250 * executing, prevent it from rearming itself and let it finish. 251 */ 252 if (timevalisset(&p->p_realtimer.it_value) && 253 callout_stop(&p->p_itcallout) == 0) { 254 timevalclear(&p->p_realtimer.it_interval); 255 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0); 256 KASSERT(!timevalisset(&p->p_realtimer.it_value), 257 ("realtime timer is still armed")); 258 } 259 PROC_UNLOCK(p); 260 261 /* 262 * Reset any sigio structures pointing to us as a result of 263 * F_SETOWN with our pid. 264 */ 265 funsetownlst(&p->p_sigiolst); 266 267 /* 268 * If this process has an nlminfo data area (for lockd), release it 269 */ 270 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL) 271 (*nlminfo_release_p)(p); 272 273 /* 274 * Close open files and release open-file table. 275 * This may block! 276 */ 277 fdfree(td); 278 279 /* 280 * If this thread tickled GEOM, we need to wait for the giggling to 281 * stop before we return to userland 282 */ 283 if (td->td_pflags & TDP_GEOM) 284 g_waitidle(); 285 286 /* 287 * Remove ourself from our leader's peer list and wake our leader. 288 */ 289 mtx_lock(&ppeers_lock); 290 if (p->p_leader->p_peers) { 291 q = p->p_leader; 292 while (q->p_peers != p) 293 q = q->p_peers; 294 q->p_peers = p->p_peers; 295 wakeup(p->p_leader); 296 } 297 mtx_unlock(&ppeers_lock); 298 299 vmspace_exit(td); 300 301 sx_xlock(&proctree_lock); 302 if (SESS_LEADER(p)) { 303 struct session *sp = p->p_session; 304 struct tty *tp; 305 306 /* 307 * s_ttyp is not zero'd; we use this to indicate that 308 * the session once had a controlling terminal. (for 309 * logging and informational purposes) 310 */ 311 SESS_LOCK(sp); 312 ttyvp = sp->s_ttyvp; 313 tp = sp->s_ttyp; 314 sp->s_ttyvp = NULL; 315 sp->s_leader = NULL; 316 SESS_UNLOCK(sp); 317 318 /* 319 * Signal foreground pgrp and revoke access to 320 * controlling terminal if it has not been revoked 321 * already. 322 * 323 * Because the TTY may have been revoked in the mean 324 * time and could already have a new session associated 325 * with it, make sure we don't send a SIGHUP to a 326 * foreground process group that does not belong to this 327 * session. 328 */ 329 330 if (tp != NULL) { 331 tty_lock(tp); 332 if (tp->t_session == sp) 333 tty_signal_pgrp(tp, SIGHUP); 334 tty_unlock(tp); 335 } 336 337 if (ttyvp != NULL) { 338 sx_xunlock(&proctree_lock); 339 vn_lock(ttyvp, LK_EXCLUSIVE | LK_RETRY); 340 if (ttyvp->v_type != VBAD) 341 VOP_REVOKE(ttyvp, REVOKEALL); 342 VOP_UNLOCK(ttyvp, 0); 343 sx_xlock(&proctree_lock); 344 } 345 } 346 fixjobc(p, p->p_pgrp, 0); 347 sx_xunlock(&proctree_lock); 348 (void)acct_process(td); 349 350 /* Release the TTY now we've unlocked everything. */ 351 if (ttyvp != NULL) 352 vrele(ttyvp); 353#ifdef KTRACE 354 /* 355 * Disable tracing, then drain any pending records and release 356 * the trace file. 357 */ 358 if (p->p_traceflag != 0) { 359 PROC_LOCK(p); 360 mtx_lock(&ktrace_mtx); 361 p->p_traceflag = 0; 362 mtx_unlock(&ktrace_mtx); 363 PROC_UNLOCK(p); 364 ktrprocexit(td); 365 PROC_LOCK(p); 366 mtx_lock(&ktrace_mtx); 367 tracevp = p->p_tracevp; 368 p->p_tracevp = NULL; 369 tracecred = p->p_tracecred; 370 p->p_tracecred = NULL; 371 mtx_unlock(&ktrace_mtx); 372 PROC_UNLOCK(p); 373 if (tracevp != NULL) { 374 locked = VFS_LOCK_GIANT(tracevp->v_mount); 375 vrele(tracevp); 376 VFS_UNLOCK_GIANT(locked); 377 } 378 if (tracecred != NULL) 379 crfree(tracecred); 380 } 381#endif 382 /* 383 * Release reference to text vnode 384 */ 385 if ((vtmp = p->p_textvp) != NULL) { 386 p->p_textvp = NULL; 387 locked = VFS_LOCK_GIANT(vtmp->v_mount); 388 vrele(vtmp); 389 VFS_UNLOCK_GIANT(locked); 390 } 391 392 /* 393 * Release our limits structure. 394 */ 395 PROC_LOCK(p); 396 plim = p->p_limit; 397 p->p_limit = NULL; 398 PROC_UNLOCK(p); 399 lim_free(plim); 400 401 /* 402 * Remove proc from allproc queue and pidhash chain. 403 * Place onto zombproc. Unlink from parent's child list. 404 */ 405 sx_xlock(&allproc_lock); 406 LIST_REMOVE(p, p_list); 407 LIST_INSERT_HEAD(&zombproc, p, p_list); 408 LIST_REMOVE(p, p_hash); 409 sx_xunlock(&allproc_lock); 410 411 /* 412 * Call machine-dependent code to release any 413 * machine-dependent resources other than the address space. 414 * The address space is released by "vmspace_exitfree(p)" in 415 * vm_waitproc(). 416 */ 417 cpu_exit(td); 418 419 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid); 420 421 /* 422 * Reparent all of our children to init. 423 */ 424 sx_xlock(&proctree_lock); 425 q = LIST_FIRST(&p->p_children); 426 if (q != NULL) /* only need this if any child is S_ZOMB */ 427 wakeup(initproc); 428 for (; q != NULL; q = nq) { 429 nq = LIST_NEXT(q, p_sibling); 430 PROC_LOCK(q); 431 proc_reparent(q, initproc); 432 q->p_sigparent = SIGCHLD; 433 /* 434 * Traced processes are killed 435 * since their existence means someone is screwing up. 436 */ 437 if (q->p_flag & P_TRACED) { 438 struct thread *temp; 439 440 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE); 441 FOREACH_THREAD_IN_PROC(q, temp) 442 temp->td_dbgflags &= ~TDB_SUSPEND; 443 psignal(q, SIGKILL); 444 } 445 PROC_UNLOCK(q); 446 } 447 448 /* Save exit status. */ 449 PROC_LOCK(p); 450 p->p_xstat = rv; 451 p->p_xthread = td; 452 453 /* Tell the prison that we are gone. */ 454 prison_proc_free(p->p_ucred->cr_prison); 455 456#ifdef KDTRACE_HOOKS 457 /* 458 * Tell the DTrace fasttrap provider about the exit if it 459 * has declared an interest. 460 */ 461 if (dtrace_fasttrap_exit) 462 dtrace_fasttrap_exit(p); 463#endif 464 465 /* 466 * Notify interested parties of our demise. 467 */ 468 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT); 469 470#ifdef KDTRACE_HOOKS 471 int reason = CLD_EXITED; 472 if (WCOREDUMP(rv)) 473 reason = CLD_DUMPED; 474 else if (WIFSIGNALED(rv)) 475 reason = CLD_KILLED; 476 SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0); 477#endif 478 479 /* 480 * Just delete all entries in the p_klist. At this point we won't 481 * report any more events, and there are nasty race conditions that 482 * can beat us if we don't. 483 */ 484 knlist_clear(&p->p_klist, 1); 485 486 /* 487 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 488 * flag set, or if the handler is set to SIG_IGN, notify process 489 * 1 instead (and hope it will handle this situation). 490 */ 491 PROC_LOCK(p->p_pptr); 492 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx); 493 if (p->p_pptr->p_sigacts->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) { 494 struct proc *pp; 495 496 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx); 497 pp = p->p_pptr; 498 PROC_UNLOCK(pp); 499 proc_reparent(p, initproc); 500 p->p_sigparent = SIGCHLD; 501 PROC_LOCK(p->p_pptr); 502 503 /* 504 * Notify parent, so in case he was wait(2)ing or 505 * executing waitpid(2) with our pid, he will 506 * continue. 507 */ 508 wakeup(pp); 509 } else 510 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx); 511 512 if (p->p_pptr == initproc) 513 psignal(p->p_pptr, SIGCHLD); 514 else if (p->p_sigparent != 0) { 515 if (p->p_sigparent == SIGCHLD) 516 childproc_exited(p); 517 else /* LINUX thread */ 518 psignal(p->p_pptr, p->p_sigparent); 519 } 520 sx_xunlock(&proctree_lock); 521 522 /* 523 * The state PRS_ZOMBIE prevents other proesses from sending 524 * signal to the process, to avoid memory leak, we free memory 525 * for signal queue at the time when the state is set. 526 */ 527 sigqueue_flush(&p->p_sigqueue); 528 sigqueue_flush(&td->td_sigqueue); 529 530 /* 531 * We have to wait until after acquiring all locks before 532 * changing p_state. We need to avoid all possible context 533 * switches (including ones from blocking on a mutex) while 534 * marked as a zombie. We also have to set the zombie state 535 * before we release the parent process' proc lock to avoid 536 * a lost wakeup. So, we first call wakeup, then we grab the 537 * sched lock, update the state, and release the parent process' 538 * proc lock. 539 */ 540 wakeup(p->p_pptr); 541 cv_broadcast(&p->p_pwait); 542 sched_exit(p->p_pptr, td); 543 PROC_SLOCK(p); 544 p->p_state = PRS_ZOMBIE; 545 PROC_UNLOCK(p->p_pptr); 546 547 /* 548 * Hopefully no one will try to deliver a signal to the process this 549 * late in the game. 550 */ 551 knlist_destroy(&p->p_klist); 552 553 /* 554 * Save our children's rusage information in our exit rusage. 555 */ 556 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux); 557 558 /* 559 * Make sure the scheduler takes this thread out of its tables etc. 560 * This will also release this thread's reference to the ucred. 561 * Other thread parts to release include pcb bits and such. 562 */ 563 thread_exit(); 564} 565 566 567#ifndef _SYS_SYSPROTO_H_ 568struct abort2_args { 569 char *why; 570 int nargs; 571 void **args; 572}; 573#endif 574 575int 576abort2(struct thread *td, struct abort2_args *uap) 577{ 578 struct proc *p = td->td_proc; 579 struct sbuf *sb; 580 void *uargs[16]; 581 int error, i, sig; 582 583 /* 584 * Do it right now so we can log either proper call of abort2(), or 585 * note, that invalid argument was passed. 512 is big enough to 586 * handle 16 arguments' descriptions with additional comments. 587 */ 588 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN); 589 sbuf_clear(sb); 590 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ", 591 p->p_comm, p->p_pid, td->td_ucred->cr_uid); 592 /* 593 * Since we can't return from abort2(), send SIGKILL in cases, where 594 * abort2() was called improperly 595 */ 596 sig = SIGKILL; 597 /* Prevent from DoSes from user-space. */ 598 if (uap->nargs < 0 || uap->nargs > 16) 599 goto out; 600 if (uap->nargs > 0) { 601 if (uap->args == NULL) 602 goto out; 603 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *)); 604 if (error != 0) 605 goto out; 606 } 607 /* 608 * Limit size of 'reason' string to 128. Will fit even when 609 * maximal number of arguments was chosen to be logged. 610 */ 611 if (uap->why != NULL) { 612 error = sbuf_copyin(sb, uap->why, 128); 613 if (error < 0) 614 goto out; 615 } else { 616 sbuf_printf(sb, "(null)"); 617 } 618 if (uap->nargs > 0) { 619 sbuf_printf(sb, "("); 620 for (i = 0;i < uap->nargs; i++) 621 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]); 622 sbuf_printf(sb, ")"); 623 } 624 /* 625 * Final stage: arguments were proper, string has been 626 * successfully copied from userspace, and copying pointers 627 * from user-space succeed. 628 */ 629 sig = SIGABRT; 630out: 631 if (sig == SIGKILL) { 632 sbuf_trim(sb); 633 sbuf_printf(sb, " (Reason text inaccessible)"); 634 } 635 sbuf_cat(sb, "\n"); 636 sbuf_finish(sb); 637 log(LOG_INFO, "%s", sbuf_data(sb)); 638 sbuf_delete(sb); 639 exit1(td, W_EXITCODE(0, sig)); 640 return (0); 641} 642 643 644#ifdef COMPAT_43 645/* 646 * The dirty work is handled by kern_wait(). 647 */ 648int 649owait(struct thread *td, struct owait_args *uap __unused) 650{ 651 int error, status; 652 653 error = kern_wait(td, WAIT_ANY, &status, 0, NULL); 654 if (error == 0) 655 td->td_retval[1] = status; 656 return (error); 657} 658#endif /* COMPAT_43 */ 659 660/* 661 * The dirty work is handled by kern_wait(). 662 */ 663int 664wait4(struct thread *td, struct wait_args *uap) 665{ 666 struct rusage ru, *rup; 667 int error, status; 668 669 if (uap->rusage != NULL) 670 rup = &ru; 671 else 672 rup = NULL; 673 error = kern_wait(td, uap->pid, &status, uap->options, rup); 674 if (uap->status != NULL && error == 0) 675 error = copyout(&status, uap->status, sizeof(status)); 676 if (uap->rusage != NULL && error == 0) 677 error = copyout(&ru, uap->rusage, sizeof(struct rusage)); 678 return (error); 679} 680 681/* 682 * Reap the remains of a zombie process and optionally return status and 683 * rusage. Asserts and will release both the proctree_lock and the process 684 * lock as part of its work. 685 */ 686static void 687proc_reap(struct thread *td, struct proc *p, int *status, int options, 688 struct rusage *rusage) 689{ 690 INIT_VPROCG(P_TO_VPROCG(p)); 691 struct proc *q, *t; 692 693 sx_assert(&proctree_lock, SA_XLOCKED); 694 PROC_LOCK_ASSERT(p, MA_OWNED); 695 PROC_SLOCK_ASSERT(p, MA_OWNED); 696 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE")); 697 698 q = td->td_proc; 699 if (rusage) { 700 *rusage = p->p_ru; 701 calcru(p, &rusage->ru_utime, &rusage->ru_stime); 702 } 703 PROC_SUNLOCK(p); 704 td->td_retval[0] = p->p_pid; 705 if (status) 706 *status = p->p_xstat; /* convert to int */ 707 if (options & WNOWAIT) { 708 /* 709 * Only poll, returning the status. Caller does not wish to 710 * release the proc struct just yet. 711 */ 712 PROC_UNLOCK(p); 713 sx_xunlock(&proctree_lock); 714 return; 715 } 716 717 PROC_LOCK(q); 718 sigqueue_take(p->p_ksi); 719 PROC_UNLOCK(q); 720 PROC_UNLOCK(p); 721 722 /* 723 * If we got the child via a ptrace 'attach', we need to give it back 724 * to the old parent. 725 */ 726 if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) { 727 PROC_LOCK(p); 728 p->p_oppid = 0; 729 proc_reparent(p, t); 730 PROC_UNLOCK(p); 731 tdsignal(t, NULL, SIGCHLD, p->p_ksi); 732 wakeup(t); 733 cv_broadcast(&p->p_pwait); 734 PROC_UNLOCK(t); 735 sx_xunlock(&proctree_lock); 736 return; 737 } 738 739 /* 740 * Remove other references to this process to ensure we have an 741 * exclusive reference. 742 */ 743 sx_xlock(&allproc_lock); 744 LIST_REMOVE(p, p_list); /* off zombproc */ 745 sx_xunlock(&allproc_lock); 746 LIST_REMOVE(p, p_sibling); 747 leavepgrp(p); 748 sx_xunlock(&proctree_lock); 749 750 /* 751 * As a side effect of this lock, we know that all other writes to 752 * this proc are visible now, so no more locking is needed for p. 753 */ 754 PROC_LOCK(p); 755 p->p_xstat = 0; /* XXX: why? */ 756 PROC_UNLOCK(p); 757 PROC_LOCK(q); 758 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux); 759 PROC_UNLOCK(q); 760 761 /* 762 * Decrement the count of procs running with this uid. 763 */ 764 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 765 766 /* 767 * Free credentials, arguments, and sigacts. 768 */ 769 crfree(p->p_ucred); 770 p->p_ucred = NULL; 771 pargs_drop(p->p_args); 772 p->p_args = NULL; 773 sigacts_free(p->p_sigacts); 774 p->p_sigacts = NULL; 775 776 /* 777 * Do any thread-system specific cleanups. 778 */ 779 thread_wait(p); 780 781 /* 782 * Give vm and machine-dependent layer a chance to free anything that 783 * cpu_exit couldn't release while still running in process context. 784 */ 785 vm_waitproc(p); 786#ifdef MAC 787 mac_proc_destroy(p); 788#endif 789 KASSERT(FIRST_THREAD_IN_PROC(p), 790 ("proc_reap: no residual thread!")); 791 uma_zfree(proc_zone, p); 792 sx_xlock(&allproc_lock); 793 nprocs--; 794#ifdef VIMAGE 795 vprocg->nprocs--; 796#endif 797 sx_xunlock(&allproc_lock); 798} 799 800int 801kern_wait(struct thread *td, pid_t pid, int *status, int options, 802 struct rusage *rusage) 803{ 804 struct proc *p, *q; 805 int error, nfound; 806 807 AUDIT_ARG_PID(pid); 808 AUDIT_ARG_VALUE(options); 809 810 q = td->td_proc; 811 if (pid == 0) { 812 PROC_LOCK(q); 813 pid = -q->p_pgid; 814 PROC_UNLOCK(q); 815 } 816 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED|WNOWAIT|WLINUXCLONE)) 817 return (EINVAL); 818loop: 819 if (q->p_flag & P_STATCHILD) { 820 PROC_LOCK(q); 821 q->p_flag &= ~P_STATCHILD; 822 PROC_UNLOCK(q); 823 } 824 nfound = 0; 825 sx_xlock(&proctree_lock); 826 LIST_FOREACH(p, &q->p_children, p_sibling) { 827 PROC_LOCK(p); 828 if (pid != WAIT_ANY && 829 p->p_pid != pid && p->p_pgid != -pid) { 830 PROC_UNLOCK(p); 831 continue; 832 } 833 if (p_canwait(td, p)) { 834 PROC_UNLOCK(p); 835 continue; 836 } 837 838 /* 839 * This special case handles a kthread spawned by linux_clone 840 * (see linux_misc.c). The linux_wait4 and linux_waitpid 841 * functions need to be able to distinguish between waiting 842 * on a process and waiting on a thread. It is a thread if 843 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option 844 * signifies we want to wait for threads and not processes. 845 */ 846 if ((p->p_sigparent != SIGCHLD) ^ 847 ((options & WLINUXCLONE) != 0)) { 848 PROC_UNLOCK(p); 849 continue; 850 } 851 852 nfound++; 853 PROC_SLOCK(p); 854 if (p->p_state == PRS_ZOMBIE) { 855 proc_reap(td, p, status, options, rusage); 856 return (0); 857 } 858 if ((p->p_flag & P_STOPPED_SIG) && 859 (p->p_suspcount == p->p_numthreads) && 860 (p->p_flag & P_WAITED) == 0 && 861 (p->p_flag & P_TRACED || options & WUNTRACED)) { 862 PROC_SUNLOCK(p); 863 p->p_flag |= P_WAITED; 864 sx_xunlock(&proctree_lock); 865 td->td_retval[0] = p->p_pid; 866 if (status) 867 *status = W_STOPCODE(p->p_xstat); 868 869 PROC_LOCK(q); 870 sigqueue_take(p->p_ksi); 871 PROC_UNLOCK(q); 872 PROC_UNLOCK(p); 873 874 return (0); 875 } 876 PROC_SUNLOCK(p); 877 if (options & WCONTINUED && (p->p_flag & P_CONTINUED)) { 878 sx_xunlock(&proctree_lock); 879 td->td_retval[0] = p->p_pid; 880 p->p_flag &= ~P_CONTINUED; 881 882 PROC_LOCK(q); 883 sigqueue_take(p->p_ksi); 884 PROC_UNLOCK(q); 885 PROC_UNLOCK(p); 886 887 if (status) 888 *status = SIGCONT; 889 return (0); 890 } 891 PROC_UNLOCK(p); 892 } 893 if (nfound == 0) { 894 sx_xunlock(&proctree_lock); 895 return (ECHILD); 896 } 897 if (options & WNOHANG) { 898 sx_xunlock(&proctree_lock); 899 td->td_retval[0] = 0; 900 return (0); 901 } 902 PROC_LOCK(q); 903 sx_xunlock(&proctree_lock); 904 if (q->p_flag & P_STATCHILD) { 905 q->p_flag &= ~P_STATCHILD; 906 error = 0; 907 } else 908 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0); 909 PROC_UNLOCK(q); 910 if (error) 911 return (error); 912 goto loop; 913} 914 915/* 916 * Make process 'parent' the new parent of process 'child'. 917 * Must be called with an exclusive hold of proctree lock. 918 */ 919void 920proc_reparent(struct proc *child, struct proc *parent) 921{ 922 923 sx_assert(&proctree_lock, SX_XLOCKED); 924 PROC_LOCK_ASSERT(child, MA_OWNED); 925 if (child->p_pptr == parent) 926 return; 927 928 PROC_LOCK(child->p_pptr); 929 sigqueue_take(child->p_ksi); 930 PROC_UNLOCK(child->p_pptr); 931 LIST_REMOVE(child, p_sibling); 932 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 933 child->p_pptr = parent; 934} 935