kern_exit.c revision 197942
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 197942 2009-10-10 21:17:30Z 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#ifdef KTRACE 73#include <sys/ktrace.h> 74#endif 75 76#include <security/audit/audit.h> 77#include <security/mac/mac_framework.h> 78 79#include <vm/vm.h> 80#include <vm/vm_extern.h> 81#include <vm/vm_param.h> 82#include <vm/pmap.h> 83#include <vm/vm_map.h> 84#include <vm/vm_page.h> 85#include <vm/uma.h> 86 87#ifdef KDTRACE_HOOKS 88#include <sys/dtrace_bsd.h> 89dtrace_execexit_func_t dtrace_fasttrap_exit; 90#endif 91 92SDT_PROVIDER_DECLARE(proc); 93SDT_PROBE_DEFINE(proc, kernel, , exit); 94SDT_PROBE_ARGTYPE(proc, kernel, , exit, 0, "int"); 95 96/* Required to be non-static for SysVR4 emulator */ 97MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 98 99/* Hook for NFS teardown procedure. */ 100void (*nlminfo_release_p)(struct proc *p); 101 102/* 103 * exit -- death of process. 104 */ 105void 106sys_exit(struct thread *td, struct sys_exit_args *uap) 107{ 108 109 exit1(td, W_EXITCODE(uap->rval, 0)); 110 /* NOTREACHED */ 111} 112 113/* 114 * Exit: deallocate address space and other resources, change proc state to 115 * zombie, and unlink proc from allproc and parent's lists. Save exit status 116 * and rusage for wait(). Check for child processes and orphan them. 117 */ 118void 119exit1(struct thread *td, int rv) 120{ 121 struct proc *p, *nq, *q; 122 struct vnode *vtmp; 123 struct vnode *ttyvp = NULL; 124#ifdef KTRACE 125 struct vnode *tracevp; 126 struct ucred *tracecred; 127#endif 128 struct plimit *plim; 129 int locked; 130 131 mtx_assert(&Giant, MA_NOTOWNED); 132 133 p = td->td_proc; 134 /* 135 * XXX in case we're rebooting we just let init die in order to 136 * work around an unsolved stack overflow seen very late during 137 * shutdown on sparc64 when the gmirror worker process exists. 138 */ 139 if (p == initproc && rebooting == 0) { 140 printf("init died (signal %d, exit %d)\n", 141 WTERMSIG(rv), WEXITSTATUS(rv)); 142 panic("Going nowhere without my init!"); 143 } 144 145 /* 146 * MUST abort all other threads before proceeding past here. 147 */ 148 PROC_LOCK(p); 149 while (p->p_flag & P_HADTHREADS) { 150 /* 151 * First check if some other thread got here before us.. 152 * if so, act apropriatly, (exit or suspend); 153 */ 154 thread_suspend_check(0); 155 156 /* 157 * Kill off the other threads. This requires 158 * some co-operation from other parts of the kernel 159 * so it may not be instantaneous. With this state set 160 * any thread entering the kernel from userspace will 161 * thread_exit() in trap(). Any thread attempting to 162 * sleep will return immediately with EINTR or EWOULDBLOCK 163 * which will hopefully force them to back out to userland 164 * freeing resources as they go. Any thread attempting 165 * to return to userland will thread_exit() from userret(). 166 * thread_exit() will unsuspend us when the last of the 167 * other threads exits. 168 * If there is already a thread singler after resumption, 169 * calling thread_single will fail; in that case, we just 170 * re-check all suspension request, the thread should 171 * either be suspended there or exit. 172 */ 173 if (! thread_single(SINGLE_EXIT)) 174 break; 175 176 /* 177 * All other activity in this process is now stopped. 178 * Threading support has been turned off. 179 */ 180 } 181 KASSERT(p->p_numthreads == 1, 182 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads)); 183 /* 184 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold 185 * on our vmspace, so we should block below until they have 186 * released their reference to us. Note that if they have 187 * requested S_EXIT stops we will block here until they ack 188 * via PIOCCONT. 189 */ 190 _STOPEVENT(p, S_EXIT, rv); 191 192 /* 193 * Note that we are exiting and do another wakeup of anyone in 194 * PIOCWAIT in case they aren't listening for S_EXIT stops or 195 * decided to wait again after we told them we are exiting. 196 */ 197 p->p_flag |= P_WEXIT; 198 wakeup(&p->p_stype); 199 200 /* 201 * Wait for any processes that have a hold on our vmspace to 202 * release their reference. 203 */ 204 while (p->p_lock > 0) 205 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0); 206 207 PROC_UNLOCK(p); 208 /* Drain the limit callout while we don't have the proc locked */ 209 callout_drain(&p->p_limco); 210 211#ifdef AUDIT 212 /* 213 * The Sun BSM exit token contains two components: an exit status as 214 * passed to exit(), and a return value to indicate what sort of exit 215 * it was. The exit status is WEXITSTATUS(rv), but it's not clear 216 * what the return value is. 217 */ 218 AUDIT_ARG_EXIT(WEXITSTATUS(rv), 0); 219 AUDIT_SYSCALL_EXIT(0, td); 220#endif 221 222 /* Are we a task leader? */ 223 if (p == p->p_leader) { 224 mtx_lock(&ppeers_lock); 225 q = p->p_peers; 226 while (q != NULL) { 227 PROC_LOCK(q); 228 psignal(q, SIGKILL); 229 PROC_UNLOCK(q); 230 q = q->p_peers; 231 } 232 while (p->p_peers != NULL) 233 msleep(p, &ppeers_lock, PWAIT, "exit1", 0); 234 mtx_unlock(&ppeers_lock); 235 } 236 237 /* 238 * Check if any loadable modules need anything done at process exit. 239 * E.g. SYSV IPC stuff 240 * XXX what if one of these generates an error? 241 */ 242 EVENTHANDLER_INVOKE(process_exit, p); 243 244 /* 245 * If parent is waiting for us to exit or exec, 246 * P_PPWAIT is set; we will wakeup the parent below. 247 */ 248 PROC_LOCK(p); 249 stopprofclock(p); 250 p->p_flag &= ~(P_TRACED | P_PPWAIT); 251 252 /* 253 * Stop the real interval timer. If the handler is currently 254 * executing, prevent it from rearming itself and let it finish. 255 */ 256 if (timevalisset(&p->p_realtimer.it_value) && 257 callout_stop(&p->p_itcallout) == 0) { 258 timevalclear(&p->p_realtimer.it_interval); 259 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0); 260 KASSERT(!timevalisset(&p->p_realtimer.it_value), 261 ("realtime timer is still armed")); 262 } 263 PROC_UNLOCK(p); 264 265 /* 266 * Reset any sigio structures pointing to us as a result of 267 * F_SETOWN with our pid. 268 */ 269 funsetownlst(&p->p_sigiolst); 270 271 /* 272 * If this process has an nlminfo data area (for lockd), release it 273 */ 274 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL) 275 (*nlminfo_release_p)(p); 276 277 /* 278 * Close open files and release open-file table. 279 * This may block! 280 */ 281 fdfree(td); 282 283 /* 284 * If this thread tickled GEOM, we need to wait for the giggling to 285 * stop before we return to userland 286 */ 287 if (td->td_pflags & TDP_GEOM) 288 g_waitidle(); 289 290 /* 291 * Remove ourself from our leader's peer list and wake our leader. 292 */ 293 mtx_lock(&ppeers_lock); 294 if (p->p_leader->p_peers) { 295 q = p->p_leader; 296 while (q->p_peers != p) 297 q = q->p_peers; 298 q->p_peers = p->p_peers; 299 wakeup(p->p_leader); 300 } 301 mtx_unlock(&ppeers_lock); 302 303 vmspace_exit(td); 304 305 sx_xlock(&proctree_lock); 306 if (SESS_LEADER(p)) { 307 struct session *sp = p->p_session; 308 struct tty *tp; 309 310 /* 311 * s_ttyp is not zero'd; we use this to indicate that 312 * the session once had a controlling terminal. (for 313 * logging and informational purposes) 314 */ 315 SESS_LOCK(sp); 316 ttyvp = sp->s_ttyvp; 317 tp = sp->s_ttyp; 318 sp->s_ttyvp = NULL; 319 sp->s_leader = NULL; 320 SESS_UNLOCK(sp); 321 322 /* 323 * Signal foreground pgrp and revoke access to 324 * controlling terminal if it has not been revoked 325 * already. 326 * 327 * Because the TTY may have been revoked in the mean 328 * time and could already have a new session associated 329 * with it, make sure we don't send a SIGHUP to a 330 * foreground process group that does not belong to this 331 * session. 332 */ 333 334 if (tp != NULL) { 335 tty_lock(tp); 336 if (tp->t_session == sp) 337 tty_signal_pgrp(tp, SIGHUP); 338 tty_unlock(tp); 339 } 340 341 if (ttyvp != NULL) { 342 sx_xunlock(&proctree_lock); 343 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) { 344 VOP_REVOKE(ttyvp, REVOKEALL); 345 VOP_UNLOCK(ttyvp, 0); 346 } 347 sx_xlock(&proctree_lock); 348 } 349 } 350 fixjobc(p, p->p_pgrp, 0); 351 sx_xunlock(&proctree_lock); 352 (void)acct_process(td); 353 354 /* Release the TTY now we've unlocked everything. */ 355 if (ttyvp != NULL) 356 vrele(ttyvp); 357#ifdef KTRACE 358 /* 359 * Disable tracing, then drain any pending records and release 360 * the trace file. 361 */ 362 if (p->p_traceflag != 0) { 363 PROC_LOCK(p); 364 mtx_lock(&ktrace_mtx); 365 p->p_traceflag = 0; 366 mtx_unlock(&ktrace_mtx); 367 PROC_UNLOCK(p); 368 ktrprocexit(td); 369 PROC_LOCK(p); 370 mtx_lock(&ktrace_mtx); 371 tracevp = p->p_tracevp; 372 p->p_tracevp = NULL; 373 tracecred = p->p_tracecred; 374 p->p_tracecred = NULL; 375 mtx_unlock(&ktrace_mtx); 376 PROC_UNLOCK(p); 377 if (tracevp != NULL) { 378 locked = VFS_LOCK_GIANT(tracevp->v_mount); 379 vrele(tracevp); 380 VFS_UNLOCK_GIANT(locked); 381 } 382 if (tracecred != NULL) 383 crfree(tracecred); 384 } 385#endif 386 /* 387 * Release reference to text vnode 388 */ 389 if ((vtmp = p->p_textvp) != NULL) { 390 p->p_textvp = NULL; 391 locked = VFS_LOCK_GIANT(vtmp->v_mount); 392 vrele(vtmp); 393 VFS_UNLOCK_GIANT(locked); 394 } 395 396 /* 397 * Release our limits structure. 398 */ 399 PROC_LOCK(p); 400 plim = p->p_limit; 401 p->p_limit = NULL; 402 PROC_UNLOCK(p); 403 lim_free(plim); 404 405 /* 406 * Remove proc from allproc queue and pidhash chain. 407 * Place onto zombproc. Unlink from parent's child list. 408 */ 409 sx_xlock(&allproc_lock); 410 LIST_REMOVE(p, p_list); 411 LIST_INSERT_HEAD(&zombproc, p, p_list); 412 LIST_REMOVE(p, p_hash); 413 sx_xunlock(&allproc_lock); 414 415 /* 416 * Call machine-dependent code to release any 417 * machine-dependent resources other than the address space. 418 * The address space is released by "vmspace_exitfree(p)" in 419 * vm_waitproc(). 420 */ 421 cpu_exit(td); 422 423 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid); 424 425 /* 426 * Reparent all of our children to init. 427 */ 428 sx_xlock(&proctree_lock); 429 q = LIST_FIRST(&p->p_children); 430 if (q != NULL) /* only need this if any child is S_ZOMB */ 431 wakeup(initproc); 432 for (; q != NULL; q = nq) { 433 nq = LIST_NEXT(q, p_sibling); 434 PROC_LOCK(q); 435 proc_reparent(q, initproc); 436 q->p_sigparent = SIGCHLD; 437 /* 438 * Traced processes are killed 439 * since their existence means someone is screwing up. 440 */ 441 if (q->p_flag & P_TRACED) { 442 struct thread *temp; 443 444 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE); 445 FOREACH_THREAD_IN_PROC(q, temp) 446 temp->td_dbgflags &= ~TDB_SUSPEND; 447 psignal(q, SIGKILL); 448 } 449 PROC_UNLOCK(q); 450 } 451 452 /* Save exit status. */ 453 PROC_LOCK(p); 454 p->p_xstat = rv; 455 p->p_xthread = td; 456 457 /* Tell the prison that we are gone. */ 458 prison_proc_free(p->p_ucred->cr_prison); 459 460#ifdef KDTRACE_HOOKS 461 /* 462 * Tell the DTrace fasttrap provider about the exit if it 463 * has declared an interest. 464 */ 465 if (dtrace_fasttrap_exit) 466 dtrace_fasttrap_exit(p); 467#endif 468 469 /* 470 * Notify interested parties of our demise. 471 */ 472 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT); 473 474#ifdef KDTRACE_HOOKS 475 int reason = CLD_EXITED; 476 if (WCOREDUMP(rv)) 477 reason = CLD_DUMPED; 478 else if (WIFSIGNALED(rv)) 479 reason = CLD_KILLED; 480 SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0); 481#endif 482 483 /* 484 * Just delete all entries in the p_klist. At this point we won't 485 * report any more events, and there are nasty race conditions that 486 * can beat us if we don't. 487 */ 488 knlist_clear(&p->p_klist, 1); 489 490 /* 491 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 492 * flag set, or if the handler is set to SIG_IGN, notify process 493 * 1 instead (and hope it will handle this situation). 494 */ 495 PROC_LOCK(p->p_pptr); 496 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx); 497 if (p->p_pptr->p_sigacts->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) { 498 struct proc *pp; 499 500 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx); 501 pp = p->p_pptr; 502 PROC_UNLOCK(pp); 503 proc_reparent(p, initproc); 504 p->p_sigparent = SIGCHLD; 505 PROC_LOCK(p->p_pptr); 506 507 /* 508 * Notify parent, so in case he was wait(2)ing or 509 * executing waitpid(2) with our pid, he will 510 * continue. 511 */ 512 wakeup(pp); 513 } else 514 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx); 515 516 if (p->p_pptr == initproc) 517 psignal(p->p_pptr, SIGCHLD); 518 else if (p->p_sigparent != 0) { 519 if (p->p_sigparent == SIGCHLD) 520 childproc_exited(p); 521 else /* LINUX thread */ 522 psignal(p->p_pptr, p->p_sigparent); 523 } 524 sx_xunlock(&proctree_lock); 525 526 /* 527 * The state PRS_ZOMBIE prevents other proesses from sending 528 * signal to the process, to avoid memory leak, we free memory 529 * for signal queue at the time when the state is set. 530 */ 531 sigqueue_flush(&p->p_sigqueue); 532 sigqueue_flush(&td->td_sigqueue); 533 534 /* 535 * We have to wait until after acquiring all locks before 536 * changing p_state. We need to avoid all possible context 537 * switches (including ones from blocking on a mutex) while 538 * marked as a zombie. We also have to set the zombie state 539 * before we release the parent process' proc lock to avoid 540 * a lost wakeup. So, we first call wakeup, then we grab the 541 * sched lock, update the state, and release the parent process' 542 * proc lock. 543 */ 544 wakeup(p->p_pptr); 545 cv_broadcast(&p->p_pwait); 546 sched_exit(p->p_pptr, td); 547 PROC_SLOCK(p); 548 p->p_state = PRS_ZOMBIE; 549 PROC_UNLOCK(p->p_pptr); 550 551 /* 552 * Hopefully no one will try to deliver a signal to the process this 553 * late in the game. 554 */ 555 knlist_destroy(&p->p_klist); 556 557 /* 558 * Save our children's rusage information in our exit rusage. 559 */ 560 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux); 561 562 /* 563 * Make sure the scheduler takes this thread out of its tables etc. 564 * This will also release this thread's reference to the ucred. 565 * Other thread parts to release include pcb bits and such. 566 */ 567 thread_exit(); 568} 569 570 571#ifndef _SYS_SYSPROTO_H_ 572struct abort2_args { 573 char *why; 574 int nargs; 575 void **args; 576}; 577#endif 578 579int 580abort2(struct thread *td, struct abort2_args *uap) 581{ 582 struct proc *p = td->td_proc; 583 struct sbuf *sb; 584 void *uargs[16]; 585 int error, i, sig; 586 587 /* 588 * Do it right now so we can log either proper call of abort2(), or 589 * note, that invalid argument was passed. 512 is big enough to 590 * handle 16 arguments' descriptions with additional comments. 591 */ 592 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN); 593 sbuf_clear(sb); 594 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ", 595 p->p_comm, p->p_pid, td->td_ucred->cr_uid); 596 /* 597 * Since we can't return from abort2(), send SIGKILL in cases, where 598 * abort2() was called improperly 599 */ 600 sig = SIGKILL; 601 /* Prevent from DoSes from user-space. */ 602 if (uap->nargs < 0 || uap->nargs > 16) 603 goto out; 604 if (uap->nargs > 0) { 605 if (uap->args == NULL) 606 goto out; 607 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *)); 608 if (error != 0) 609 goto out; 610 } 611 /* 612 * Limit size of 'reason' string to 128. Will fit even when 613 * maximal number of arguments was chosen to be logged. 614 */ 615 if (uap->why != NULL) { 616 error = sbuf_copyin(sb, uap->why, 128); 617 if (error < 0) 618 goto out; 619 } else { 620 sbuf_printf(sb, "(null)"); 621 } 622 if (uap->nargs > 0) { 623 sbuf_printf(sb, "("); 624 for (i = 0;i < uap->nargs; i++) 625 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]); 626 sbuf_printf(sb, ")"); 627 } 628 /* 629 * Final stage: arguments were proper, string has been 630 * successfully copied from userspace, and copying pointers 631 * from user-space succeed. 632 */ 633 sig = SIGABRT; 634out: 635 if (sig == SIGKILL) { 636 sbuf_trim(sb); 637 sbuf_printf(sb, " (Reason text inaccessible)"); 638 } 639 sbuf_cat(sb, "\n"); 640 sbuf_finish(sb); 641 log(LOG_INFO, "%s", sbuf_data(sb)); 642 sbuf_delete(sb); 643 exit1(td, W_EXITCODE(0, sig)); 644 return (0); 645} 646 647 648#ifdef COMPAT_43 649/* 650 * The dirty work is handled by kern_wait(). 651 */ 652int 653owait(struct thread *td, struct owait_args *uap __unused) 654{ 655 int error, status; 656 657 error = kern_wait(td, WAIT_ANY, &status, 0, NULL); 658 if (error == 0) 659 td->td_retval[1] = status; 660 return (error); 661} 662#endif /* COMPAT_43 */ 663 664/* 665 * The dirty work is handled by kern_wait(). 666 */ 667int 668wait4(struct thread *td, struct wait_args *uap) 669{ 670 struct rusage ru, *rup; 671 int error, status; 672 673 if (uap->rusage != NULL) 674 rup = &ru; 675 else 676 rup = NULL; 677 error = kern_wait(td, uap->pid, &status, uap->options, rup); 678 if (uap->status != NULL && error == 0) 679 error = copyout(&status, uap->status, sizeof(status)); 680 if (uap->rusage != NULL && error == 0) 681 error = copyout(&ru, uap->rusage, sizeof(struct rusage)); 682 return (error); 683} 684 685/* 686 * Reap the remains of a zombie process and optionally return status and 687 * rusage. Asserts and will release both the proctree_lock and the process 688 * lock as part of its work. 689 */ 690static void 691proc_reap(struct thread *td, struct proc *p, int *status, int options, 692 struct rusage *rusage) 693{ 694 struct proc *q, *t; 695 696 sx_assert(&proctree_lock, SA_XLOCKED); 697 PROC_LOCK_ASSERT(p, MA_OWNED); 698 PROC_SLOCK_ASSERT(p, MA_OWNED); 699 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE")); 700 701 q = td->td_proc; 702 if (rusage) { 703 *rusage = p->p_ru; 704 calcru(p, &rusage->ru_utime, &rusage->ru_stime); 705 } 706 PROC_SUNLOCK(p); 707 td->td_retval[0] = p->p_pid; 708 if (status) 709 *status = p->p_xstat; /* convert to int */ 710 if (options & WNOWAIT) { 711 /* 712 * Only poll, returning the status. Caller does not wish to 713 * release the proc struct just yet. 714 */ 715 PROC_UNLOCK(p); 716 sx_xunlock(&proctree_lock); 717 return; 718 } 719 720 PROC_LOCK(q); 721 sigqueue_take(p->p_ksi); 722 PROC_UNLOCK(q); 723 PROC_UNLOCK(p); 724 725 /* 726 * If we got the child via a ptrace 'attach', we need to give it back 727 * to the old parent. 728 */ 729 if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) { 730 PROC_LOCK(p); 731 p->p_oppid = 0; 732 proc_reparent(p, t); 733 PROC_UNLOCK(p); 734 tdsignal(t, NULL, SIGCHLD, p->p_ksi); 735 wakeup(t); 736 cv_broadcast(&p->p_pwait); 737 PROC_UNLOCK(t); 738 sx_xunlock(&proctree_lock); 739 return; 740 } 741 742 /* 743 * Remove other references to this process to ensure we have an 744 * exclusive reference. 745 */ 746 sx_xlock(&allproc_lock); 747 LIST_REMOVE(p, p_list); /* off zombproc */ 748 sx_xunlock(&allproc_lock); 749 LIST_REMOVE(p, p_sibling); 750 leavepgrp(p); 751 sx_xunlock(&proctree_lock); 752 753 /* 754 * As a side effect of this lock, we know that all other writes to 755 * this proc are visible now, so no more locking is needed for p. 756 */ 757 PROC_LOCK(p); 758 p->p_xstat = 0; /* XXX: why? */ 759 PROC_UNLOCK(p); 760 PROC_LOCK(q); 761 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux); 762 PROC_UNLOCK(q); 763 764 /* 765 * Decrement the count of procs running with this uid. 766 */ 767 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 768 769 /* 770 * Free credentials, arguments, and sigacts. 771 */ 772 crfree(p->p_ucred); 773 p->p_ucred = NULL; 774 pargs_drop(p->p_args); 775 p->p_args = NULL; 776 sigacts_free(p->p_sigacts); 777 p->p_sigacts = NULL; 778 779 /* 780 * Do any thread-system specific cleanups. 781 */ 782 thread_wait(p); 783 784 /* 785 * Give vm and machine-dependent layer a chance to free anything that 786 * cpu_exit couldn't release while still running in process context. 787 */ 788 vm_waitproc(p); 789#ifdef MAC 790 mac_proc_destroy(p); 791#endif 792 KASSERT(FIRST_THREAD_IN_PROC(p), 793 ("proc_reap: no residual thread!")); 794 uma_zfree(proc_zone, p); 795 sx_xlock(&allproc_lock); 796 nprocs--; 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