1 /* $NetBSD: kern_exit.c,v 1.272 2018/07/12 10:46:48 maxv Exp $ */ 2 3 /*- 4 * Copyright (c) 1998, 1999, 2006, 2007, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility, 9 * NASA Ames Research Center, and by Andrew Doran. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1989, 1991, 1993 35 * The Regents of the University of California. All rights reserved. 36 * (c) UNIX System Laboratories, Inc. 37 * All or some portions of this file are derived from material licensed 38 * to the University of California by American Telephone and Telegraph 39 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 40 * the permission of UNIX System Laboratories, Inc. 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 3. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * @(#)kern_exit.c 8.10 (Berkeley) 2/23/95 67 */ 68 69 #include <sys/cdefs.h> 70 __KERNEL_RCSID(0, "$NetBSD: kern_exit.c,v 1.272 2018/07/12 10:46:48 maxv Exp $"); 71 72 #include "opt_ktrace.h" 73 #include "opt_dtrace.h" 74 #include "opt_sysv.h" 75 76 #include <sys/param.h> 77 #include <sys/systm.h> 78 #include <sys/ioctl.h> 79 #include <sys/tty.h> 80 #include <sys/time.h> 81 #include <sys/resource.h> 82 #include <sys/kernel.h> 83 #include <sys/proc.h> 84 #include <sys/buf.h> 85 #include <sys/wait.h> 86 #include <sys/file.h> 87 #include <sys/vnode.h> 88 #include <sys/syslog.h> 89 #include <sys/pool.h> 90 #include <sys/uidinfo.h> 91 #include <sys/ptrace.h> 92 #include <sys/acct.h> 93 #include <sys/filedesc.h> 94 #include <sys/ras.h> 95 #include <sys/signalvar.h> 96 #include <sys/sched.h> 97 #include <sys/mount.h> 98 #include <sys/syscallargs.h> 99 #include <sys/kauth.h> 100 #include <sys/sleepq.h> 101 #include <sys/lockdebug.h> 102 #include <sys/ktrace.h> 103 #include <sys/cpu.h> 104 #include <sys/lwpctl.h> 105 #include <sys/atomic.h> 106 #include <sys/sdt.h> 107 108 #include <uvm/uvm_extern.h> 109 110 #ifdef DEBUG_EXIT 111 int debug_exit = 0; 112 #define DPRINTF(x) if (debug_exit) printf x 113 #else 114 #define DPRINTF(x) 115 #endif 116 117 static int find_stopped_child(struct proc *, idtype_t, id_t, int, 118 struct proc **, struct wrusage *, siginfo_t *); 119 static void proc_free(struct proc *, struct wrusage *); 120 121 /* 122 * DTrace SDT provider definitions 123 */ 124 SDT_PROVIDER_DECLARE(proc); 125 SDT_PROBE_DEFINE1(proc, kernel, , exit, "int"); 126 127 /* 128 * Fill in the appropriate signal information, and signal the parent. 129 */ 130 /* XXX noclone works around a gcc 4.5 bug on arm */ 131 static void __noclone 132 exit_psignal(struct proc *p, struct proc *pp, ksiginfo_t *ksi) 133 { 134 135 KSI_INIT(ksi); 136 if ((ksi->ksi_signo = P_EXITSIG(p)) == SIGCHLD) { 137 if (p->p_xsig) { 138 if (p->p_sflag & PS_COREDUMP) 139 ksi->ksi_code = CLD_DUMPED; 140 else 141 ksi->ksi_code = CLD_KILLED; 142 ksi->ksi_status = p->p_xsig; 143 } else { 144 ksi->ksi_code = CLD_EXITED; 145 ksi->ksi_status = p->p_xexit; 146 } 147 } else { 148 ksi->ksi_code = SI_USER; 149 ksi->ksi_status = p->p_xsig; 150 } 151 /* 152 * We fill those in, even for non-SIGCHLD. 153 * It's safe to access p->p_cred unlocked here. 154 */ 155 ksi->ksi_pid = p->p_pid; 156 ksi->ksi_uid = kauth_cred_geteuid(p->p_cred); 157 /* XXX: is this still valid? */ 158 ksi->ksi_utime = p->p_stats->p_ru.ru_utime.tv_sec; 159 ksi->ksi_stime = p->p_stats->p_ru.ru_stime.tv_sec; 160 } 161 162 /* 163 * exit -- 164 * Death of process. 165 */ 166 int 167 sys_exit(struct lwp *l, const struct sys_exit_args *uap, register_t *retval) 168 { 169 /* { 170 syscallarg(int) rval; 171 } */ 172 struct proc *p = l->l_proc; 173 174 /* Don't call exit1() multiple times in the same process. */ 175 mutex_enter(p->p_lock); 176 if (p->p_sflag & PS_WEXIT) { 177 mutex_exit(p->p_lock); 178 lwp_exit(l); 179 } 180 181 /* exit1() will release the mutex. */ 182 exit1(l, SCARG(uap, rval), 0); 183 /* NOTREACHED */ 184 return (0); 185 } 186 187 /* 188 * Exit: deallocate address space and other resources, change proc state 189 * to zombie, and unlink proc from allproc and parent's lists. Save exit 190 * status and rusage for wait(). Check for child processes and orphan them. 191 * 192 * Must be called with p->p_lock held. Does not return. 193 */ 194 void 195 exit1(struct lwp *l, int exitcode, int signo) 196 { 197 struct proc *p, *child, *next_child, *old_parent, *new_parent; 198 struct pgrp *pgrp; 199 ksiginfo_t ksi; 200 ksiginfoq_t kq; 201 int wakeinit; 202 203 p = l->l_proc; 204 205 KASSERT(mutex_owned(p->p_lock)); 206 KASSERT(p->p_vmspace != NULL); 207 208 if (__predict_false(p == initproc)) { 209 panic("init died (signal %d, exit %d)", signo, exitcode); 210 } 211 212 p->p_sflag |= PS_WEXIT; 213 214 /* 215 * Force all other LWPs to exit before we do. Only then can we 216 * begin to tear down the rest of the process state. 217 */ 218 if (p->p_nlwps > 1) { 219 exit_lwps(l); 220 } 221 222 ksiginfo_queue_init(&kq); 223 224 /* 225 * If we have been asked to stop on exit, do so now. 226 */ 227 if (__predict_false(p->p_sflag & PS_STOPEXIT)) { 228 KERNEL_UNLOCK_ALL(l, &l->l_biglocks); 229 sigclearall(p, &contsigmask, &kq); 230 231 if (!mutex_tryenter(proc_lock)) { 232 mutex_exit(p->p_lock); 233 mutex_enter(proc_lock); 234 mutex_enter(p->p_lock); 235 } 236 p->p_waited = 0; 237 p->p_pptr->p_nstopchild++; 238 p->p_stat = SSTOP; 239 mutex_exit(proc_lock); 240 lwp_lock(l); 241 p->p_nrlwps--; 242 l->l_stat = LSSTOP; 243 lwp_unlock(l); 244 mutex_exit(p->p_lock); 245 lwp_lock(l); 246 mi_switch(l); 247 KERNEL_LOCK(l->l_biglocks, l); 248 mutex_enter(p->p_lock); 249 } 250 251 /* 252 * Bin any remaining signals and mark the process as dying so it will 253 * not be found for, e.g. signals. 254 */ 255 sigfillset(&p->p_sigctx.ps_sigignore); 256 sigclearall(p, NULL, &kq); 257 p->p_stat = SDYING; 258 mutex_exit(p->p_lock); 259 ksiginfo_queue_drain(&kq); 260 261 /* Destroy any lwpctl info. */ 262 if (p->p_lwpctl != NULL) 263 lwp_ctl_exit(); 264 265 /* 266 * Drain all remaining references that procfs, ptrace and others may 267 * have on the process. 268 */ 269 rw_enter(&p->p_reflock, RW_WRITER); 270 271 DPRINTF(("%s: %d.%d exiting.\n", __func__, p->p_pid, l->l_lid)); 272 273 timers_free(p, TIMERS_ALL); 274 #if defined(__HAVE_RAS) 275 ras_purgeall(); 276 #endif 277 278 /* 279 * Close open files, release open-file table and free signal 280 * actions. This may block! 281 */ 282 fd_free(); 283 cwdfree(p->p_cwdi); 284 p->p_cwdi = NULL; 285 doexithooks(p); 286 sigactsfree(p->p_sigacts); 287 288 /* 289 * Write out accounting data. 290 */ 291 (void)acct_process(l); 292 293 #ifdef KTRACE 294 /* 295 * Release trace file. 296 */ 297 if (p->p_tracep != NULL) { 298 mutex_enter(&ktrace_lock); 299 ktrderef(p); 300 mutex_exit(&ktrace_lock); 301 } 302 #endif 303 304 p->p_xexit = exitcode; 305 p->p_xsig = signo; 306 307 /* 308 * If emulation has process exit hook, call it now. 309 * Set the exit status now so that the exit hook has 310 * an opportunity to tweak it (COMPAT_LINUX requires 311 * this for thread group emulation) 312 */ 313 if (p->p_emul->e_proc_exit) 314 (*p->p_emul->e_proc_exit)(p); 315 316 /* 317 * Free the VM resources we're still holding on to. 318 * We must do this from a valid thread because doing 319 * so may block. This frees vmspace, which we don't 320 * need anymore. The only remaining lwp is the one 321 * we run at this moment, nothing runs in userland 322 * anymore. 323 */ 324 ruspace(p); /* Update our vm resource use */ 325 uvm_proc_exit(p); 326 327 /* 328 * Stop profiling. 329 */ 330 if (__predict_false((p->p_stflag & PST_PROFIL) != 0)) { 331 mutex_spin_enter(&p->p_stmutex); 332 stopprofclock(p); 333 mutex_spin_exit(&p->p_stmutex); 334 } 335 336 /* 337 * If parent is waiting for us to exit or exec, PL_PPWAIT is set; we 338 * wake up the parent early to avoid deadlock. We can do this once 339 * the VM resources are released. 340 */ 341 mutex_enter(proc_lock); 342 if (p->p_lflag & PL_PPWAIT) { 343 l->l_lwpctl = NULL; /* was on loan from blocked parent */ 344 p->p_lflag &= ~PL_PPWAIT; 345 cv_broadcast(&p->p_pptr->p_waitcv); 346 } 347 348 if (SESS_LEADER(p)) { 349 struct vnode *vprele = NULL, *vprevoke = NULL; 350 struct session *sp = p->p_session; 351 struct tty *tp; 352 353 if (sp->s_ttyvp) { 354 /* 355 * Controlling process. 356 * Signal foreground pgrp, 357 * drain controlling terminal 358 * and revoke access to controlling terminal. 359 */ 360 tp = sp->s_ttyp; 361 mutex_spin_enter(&tty_lock); 362 if (tp->t_session == sp) { 363 /* we can't guarantee the revoke will do this */ 364 pgrp = tp->t_pgrp; 365 tp->t_pgrp = NULL; 366 tp->t_session = NULL; 367 mutex_spin_exit(&tty_lock); 368 if (pgrp != NULL) { 369 pgsignal(pgrp, SIGHUP, 1); 370 } 371 mutex_exit(proc_lock); 372 (void) ttywait(tp); 373 mutex_enter(proc_lock); 374 375 /* The tty could have been revoked. */ 376 vprevoke = sp->s_ttyvp; 377 } else 378 mutex_spin_exit(&tty_lock); 379 vprele = sp->s_ttyvp; 380 sp->s_ttyvp = NULL; 381 /* 382 * s_ttyp is not zero'd; we use this to indicate 383 * that the session once had a controlling terminal. 384 * (for logging and informational purposes) 385 */ 386 } 387 sp->s_leader = NULL; 388 389 if (vprevoke != NULL || vprele != NULL) { 390 if (vprevoke != NULL) { 391 /* Releases proc_lock. */ 392 proc_sessrele(sp); 393 VOP_REVOKE(vprevoke, REVOKEALL); 394 } else 395 mutex_exit(proc_lock); 396 if (vprele != NULL) 397 vrele(vprele); 398 mutex_enter(proc_lock); 399 } 400 } 401 fixjobc(p, p->p_pgrp, 0); 402 403 /* 404 * Finalize the last LWP's specificdata, as well as the 405 * specificdata for the proc itself. 406 */ 407 lwp_finispecific(l); 408 proc_finispecific(p); 409 410 /* 411 * Notify interested parties of our demise. 412 */ 413 KNOTE(&p->p_klist, NOTE_EXIT); 414 415 SDT_PROBE(proc, kernel, , exit, 416 ((p->p_sflag & PS_COREDUMP) ? CLD_DUMPED : 417 (p->p_xsig ? CLD_KILLED : CLD_EXITED)), 418 0,0,0,0); 419 420 /* 421 * Reset p_opptr pointer of all former children which got 422 * traced by another process and were reparented. We reset 423 * it to NULL here; the trace detach code then reparents 424 * the child to initproc. We only check allproc list, since 425 * eventual former children on zombproc list won't reference 426 * p_opptr anymore. 427 */ 428 if (__predict_false(p->p_slflag & PSL_CHTRACED)) { 429 struct proc *q; 430 PROCLIST_FOREACH(q, &allproc) { 431 if (q->p_opptr == p) 432 q->p_opptr = NULL; 433 } 434 PROCLIST_FOREACH(q, &zombproc) { 435 if (q->p_opptr == p) 436 q->p_opptr = NULL; 437 } 438 } 439 440 /* 441 * Give orphaned children to init(8). 442 */ 443 child = LIST_FIRST(&p->p_children); 444 wakeinit = (child != NULL); 445 for (; child != NULL; child = next_child) { 446 next_child = LIST_NEXT(child, p_sibling); 447 448 /* 449 * Traced processes are killed since their existence 450 * means someone is screwing up. Since we reset the 451 * trace flags, the logic in sys_wait4() would not be 452 * triggered to reparent the process to its 453 * original parent, so we must do this here. 454 */ 455 if (__predict_false(child->p_slflag & PSL_TRACED)) { 456 mutex_enter(p->p_lock); 457 child->p_slflag &= 458 ~(PSL_TRACED|PSL_SYSCALL); 459 mutex_exit(p->p_lock); 460 if (child->p_opptr != child->p_pptr) { 461 struct proc *t = child->p_opptr; 462 proc_reparent(child, t ? t : initproc); 463 child->p_opptr = NULL; 464 } else 465 proc_reparent(child, initproc); 466 killproc(child, "orphaned traced process"); 467 } else 468 proc_reparent(child, initproc); 469 } 470 471 /* 472 * Move proc from allproc to zombproc, it's now nearly ready to be 473 * collected by parent. 474 */ 475 LIST_REMOVE(l, l_list); 476 LIST_REMOVE(p, p_list); 477 LIST_INSERT_HEAD(&zombproc, p, p_list); 478 479 /* 480 * Mark the process as dead. We must do this before we signal 481 * the parent. 482 */ 483 p->p_stat = SDEAD; 484 485 /* Put in front of parent's sibling list for parent to collect it */ 486 old_parent = p->p_pptr; 487 old_parent->p_nstopchild++; 488 if (LIST_FIRST(&old_parent->p_children) != p) { 489 /* Put child where it can be found quickly */ 490 LIST_REMOVE(p, p_sibling); 491 LIST_INSERT_HEAD(&old_parent->p_children, p, p_sibling); 492 } 493 494 /* 495 * Notify parent that we're gone. If parent has the P_NOCLDWAIT 496 * flag set, notify init instead (and hope it will handle 497 * this situation). 498 */ 499 if (old_parent->p_flag & (PK_NOCLDWAIT|PK_CLDSIGIGN)) { 500 proc_reparent(p, initproc); 501 wakeinit = 1; 502 503 /* 504 * If this was the last child of our parent, notify 505 * parent, so in case he was wait(2)ing, he will 506 * continue. 507 */ 508 if (LIST_FIRST(&old_parent->p_children) == NULL) 509 cv_broadcast(&old_parent->p_waitcv); 510 } 511 512 /* Reload parent pointer, since p may have been reparented above */ 513 new_parent = p->p_pptr; 514 515 if (__predict_false(p->p_exitsig != 0)) { 516 exit_psignal(p, new_parent, &ksi); 517 kpsignal(new_parent, &ksi, NULL); 518 } 519 520 /* Calculate the final rusage info. */ 521 calcru(p, &p->p_stats->p_ru.ru_utime, &p->p_stats->p_ru.ru_stime, 522 NULL, NULL); 523 524 if (wakeinit) 525 cv_broadcast(&initproc->p_waitcv); 526 527 callout_destroy(&l->l_timeout_ch); 528 529 /* 530 * Release any PCU resources before becoming a zombie. 531 */ 532 pcu_discard_all(l); 533 534 mutex_enter(p->p_lock); 535 /* Free the linux lwp id */ 536 if ((l->l_pflag & LP_PIDLID) != 0 && l->l_lid != p->p_pid) 537 proc_free_pid(l->l_lid); 538 lwp_drainrefs(l); 539 lwp_lock(l); 540 l->l_prflag &= ~LPR_DETACHED; 541 l->l_stat = LSZOMB; 542 lwp_unlock(l); 543 KASSERT(curlwp == l); 544 KASSERT(p->p_nrlwps == 1); 545 KASSERT(p->p_nlwps == 1); 546 p->p_stat = SZOMB; 547 p->p_nrlwps--; 548 p->p_nzlwps++; 549 p->p_ndlwps = 0; 550 mutex_exit(p->p_lock); 551 552 /* 553 * Signal the parent to collect us, and drop the proclist lock. 554 * Drop debugger/procfs lock; no new references can be gained. 555 */ 556 cv_broadcast(&p->p_pptr->p_waitcv); 557 rw_exit(&p->p_reflock); 558 mutex_exit(proc_lock); 559 560 /* Verify that we hold no locks other than the kernel lock. */ 561 LOCKDEBUG_BARRIER(&kernel_lock, 0); 562 563 /* 564 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP! 565 */ 566 567 /* 568 * Give machine-dependent code a chance to free any MD LWP 569 * resources. This must be done before uvm_lwp_exit(), in 570 * case these resources are in the PCB. 571 */ 572 cpu_lwp_free(l, 1); 573 574 pmap_deactivate(l); 575 576 /* This process no longer needs to hold the kernel lock. */ 577 #ifdef notyet 578 /* XXXSMP hold in lwp_userret() */ 579 KERNEL_UNLOCK_LAST(l); 580 #else 581 KERNEL_UNLOCK_ALL(l, NULL); 582 #endif 583 584 lwp_exit_switchaway(l); 585 } 586 587 void 588 exit_lwps(struct lwp *l) 589 { 590 proc_t *p = l->l_proc; 591 lwp_t *l2; 592 int nlocks; 593 594 KERNEL_UNLOCK_ALL(l, &nlocks); 595 retry: 596 KASSERT(mutex_owned(p->p_lock)); 597 598 /* 599 * Interrupt LWPs in interruptable sleep, unsuspend suspended 600 * LWPs and then wait for everyone else to finish. 601 */ 602 LIST_FOREACH(l2, &p->p_lwps, l_sibling) { 603 if (l2 == l) 604 continue; 605 lwp_lock(l2); 606 l2->l_flag |= LW_WEXIT; 607 if ((l2->l_stat == LSSLEEP && (l2->l_flag & LW_SINTR)) || 608 l2->l_stat == LSSUSPENDED || l2->l_stat == LSSTOP) { 609 /* setrunnable() will release the lock. */ 610 setrunnable(l2); 611 continue; 612 } 613 lwp_unlock(l2); 614 } 615 616 /* 617 * Wait for every LWP to exit. Note: LWPs can get suspended/slept 618 * behind us or there may even be new LWPs created. Therefore, a 619 * full retry is required on error. 620 */ 621 while (p->p_nlwps > 1) { 622 if (lwp_wait(l, 0, NULL, true)) { 623 goto retry; 624 } 625 } 626 627 KERNEL_LOCK(nlocks, l); 628 KASSERT(p->p_nlwps == 1); 629 } 630 631 int 632 do_sys_waitid(idtype_t idtype, id_t id, int *pid, int *status, int options, 633 struct wrusage *wru, siginfo_t *si) 634 { 635 proc_t *child; 636 int error; 637 638 639 if (wru != NULL) 640 memset(wru, 0, sizeof(*wru)); 641 if (si != NULL) 642 memset(si, 0, sizeof(*si)); 643 644 mutex_enter(proc_lock); 645 error = find_stopped_child(curproc, idtype, id, options, &child, 646 wru, si); 647 if (child == NULL) { 648 mutex_exit(proc_lock); 649 *pid = 0; 650 return error; 651 } 652 *pid = child->p_pid; 653 654 if (child->p_stat == SZOMB) { 655 /* Child is exiting */ 656 *status = P_WAITSTATUS(child); 657 /* proc_free() will release the proc_lock. */ 658 if (options & WNOWAIT) { 659 mutex_exit(proc_lock); 660 } else { 661 proc_free(child, wru); 662 } 663 } else { 664 /* Don't mark SIGCONT if we are being stopped */ 665 *status = (child->p_xsig == SIGCONT && child->p_stat != SSTOP) ? 666 W_CONTCODE() : W_STOPCODE(child->p_xsig); 667 mutex_exit(proc_lock); 668 } 669 return 0; 670 } 671 672 int 673 do_sys_wait(int *pid, int *status, int options, struct rusage *ru) 674 { 675 idtype_t idtype; 676 id_t id; 677 int ret; 678 struct wrusage wru; 679 680 /* 681 * Translate the special pid values into the (idtype, pid) 682 * pair for wait6. The WAIT_MYPGRP case is handled by 683 * find_stopped_child() on its own. 684 */ 685 if (*pid == WAIT_ANY) { 686 idtype = P_ALL; 687 id = 0; 688 } else if (*pid < 0) { 689 idtype = P_PGID; 690 id = (id_t)-*pid; 691 } else { 692 idtype = P_PID; 693 id = (id_t)*pid; 694 } 695 options |= WEXITED | WTRAPPED; 696 ret = do_sys_waitid(idtype, id, pid, status, options, ru ? &wru : NULL, 697 NULL); 698 if (ru) 699 *ru = wru.wru_self; 700 return ret; 701 } 702 703 int 704 sys___wait450(struct lwp *l, const struct sys___wait450_args *uap, 705 register_t *retval) 706 { 707 /* { 708 syscallarg(int) pid; 709 syscallarg(int *) status; 710 syscallarg(int) options; 711 syscallarg(struct rusage *) rusage; 712 } */ 713 int error, status, pid = SCARG(uap, pid); 714 struct rusage ru; 715 716 error = do_sys_wait(&pid, &status, SCARG(uap, options), 717 SCARG(uap, rusage) != NULL ? &ru : NULL); 718 719 retval[0] = pid; 720 if (pid == 0) { 721 return error; 722 } 723 if (SCARG(uap, status)) { 724 error = copyout(&status, SCARG(uap, status), sizeof(status)); 725 } 726 if (SCARG(uap, rusage) && error == 0) { 727 error = copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 728 } 729 return error; 730 } 731 732 int 733 sys_wait6(struct lwp *l, const struct sys_wait6_args *uap, register_t *retval) 734 { 735 /* { 736 syscallarg(idtype_t) idtype; 737 syscallarg(id_t) id; 738 syscallarg(int *) status; 739 syscallarg(int) options; 740 syscallarg(struct wrusage *) wru; 741 syscallarg(siginfo_t *) si; 742 } */ 743 struct wrusage wru, *wrup; 744 siginfo_t si, *sip; 745 idtype_t idtype; 746 int pid; 747 id_t id; 748 int error, status; 749 750 idtype = SCARG(uap, idtype); 751 id = SCARG(uap, id); 752 753 if (SCARG(uap, wru) != NULL) 754 wrup = &wru; 755 else 756 wrup = NULL; 757 758 if (SCARG(uap, info) != NULL) 759 sip = &si; 760 else 761 sip = NULL; 762 763 /* 764 * We expect all callers of wait6() to know about WEXITED and 765 * WTRAPPED. 766 */ 767 error = do_sys_waitid(idtype, id, &pid, &status, SCARG(uap, options), 768 wrup, sip); 769 770 retval[0] = pid; /* tell userland who it was */ 771 772 #if 0 773 /* 774 * should we copyout if there was no process, hence no useful data? 775 * We don't for an old sytle wait4() (etc) but I believe 776 * FreeBSD does for wait6(), so a tossup... Go with FreeBSD for now. 777 */ 778 if (pid == 0) 779 return error; 780 #endif 781 782 if (SCARG(uap, status) != NULL && error == 0) 783 error = copyout(&status, SCARG(uap, status), sizeof(status)); 784 if (SCARG(uap, wru) != NULL && error == 0) 785 error = copyout(&wru, SCARG(uap, wru), sizeof(wru)); 786 if (SCARG(uap, info) != NULL && error == 0) 787 error = copyout(&si, SCARG(uap, info), sizeof(si)); 788 return error; 789 } 790 791 792 /* 793 * Find a process that matches the provided criteria, and fill siginfo 794 * and resources if found. 795 * Returns: 796 * -1: Not found, abort early 797 * 0: Not matched 798 * 1: Matched, there might be more matches 799 * 2: This is the only match 800 */ 801 static int 802 match_process(const struct proc *pp, struct proc **q, idtype_t idtype, id_t id, 803 int options, struct wrusage *wrusage, siginfo_t *siginfo) 804 { 805 struct rusage *rup; 806 struct proc *p = *q; 807 int rv = 1; 808 809 mutex_enter(p->p_lock); 810 switch (idtype) { 811 case P_ALL: 812 break; 813 case P_PID: 814 if (p->p_pid != (pid_t)id) { 815 mutex_exit(p->p_lock); 816 p = *q = proc_find_raw((pid_t)id); 817 if (p == NULL || p->p_stat == SIDL || p->p_pptr != pp) { 818 *q = NULL; 819 return -1; 820 } 821 mutex_enter(p->p_lock); 822 } 823 rv++; 824 break; 825 case P_PGID: 826 if (p->p_pgid != (pid_t)id) 827 goto out; 828 break; 829 case P_SID: 830 if (p->p_session->s_sid != (pid_t)id) 831 goto out; 832 break; 833 case P_UID: 834 if (kauth_cred_geteuid(p->p_cred) != (uid_t)id) 835 goto out; 836 break; 837 case P_GID: 838 if (kauth_cred_getegid(p->p_cred) != (gid_t)id) 839 goto out; 840 break; 841 case P_CID: 842 case P_PSETID: 843 case P_CPUID: 844 /* XXX: Implement me */ 845 default: 846 out: 847 mutex_exit(p->p_lock); 848 return 0; 849 } 850 851 if ((options & WEXITED) == 0 && p->p_stat == SZOMB) 852 goto out; 853 854 if (siginfo != NULL) { 855 siginfo->si_errno = 0; 856 857 /* 858 * SUSv4 requires that the si_signo value is always 859 * SIGCHLD. Obey it despite the rfork(2) interface 860 * allows to request other signal for child exit 861 * notification. 862 */ 863 siginfo->si_signo = SIGCHLD; 864 865 /* 866 * This is still a rough estimate. We will fix the 867 * cases TRAPPED, STOPPED, and CONTINUED later. 868 */ 869 if (p->p_sflag & PS_COREDUMP) { 870 siginfo->si_code = CLD_DUMPED; 871 siginfo->si_status = p->p_xsig; 872 } else if (p->p_xsig) { 873 siginfo->si_code = CLD_KILLED; 874 siginfo->si_status = p->p_xsig; 875 } else { 876 siginfo->si_code = CLD_EXITED; 877 siginfo->si_status = p->p_xexit; 878 } 879 880 siginfo->si_pid = p->p_pid; 881 siginfo->si_uid = kauth_cred_geteuid(p->p_cred); 882 siginfo->si_utime = p->p_stats->p_ru.ru_utime.tv_sec; 883 siginfo->si_stime = p->p_stats->p_ru.ru_stime.tv_sec; 884 } 885 886 /* 887 * There should be no reason to limit resources usage info to 888 * exited processes only. A snapshot about any resources used 889 * by a stopped process may be exactly what is needed. 890 */ 891 if (wrusage != NULL) { 892 rup = &wrusage->wru_self; 893 *rup = p->p_stats->p_ru; 894 calcru(p, &rup->ru_utime, &rup->ru_stime, NULL, NULL); 895 896 rup = &wrusage->wru_children; 897 *rup = p->p_stats->p_cru; 898 calcru(p, &rup->ru_utime, &rup->ru_stime, NULL, NULL); 899 } 900 901 mutex_exit(p->p_lock); 902 return rv; 903 } 904 905 /* 906 * Determine if there are existing processes being debugged 907 * that used to be (and sometime later will be again) children 908 * of a specific parent (while matching wait criteria) 909 */ 910 static bool 911 debugged_child_exists(idtype_t idtype, id_t id, int options, siginfo_t *si, 912 const struct proc *parent) 913 { 914 struct proc *pp; 915 916 /* 917 * If we are searching for a specific pid, we can optimise a little 918 */ 919 if (idtype == P_PID) { 920 /* 921 * Check the specific process to see if its real parent is us 922 */ 923 pp = proc_find_raw((pid_t)id); 924 if (pp != NULL && pp->p_stat != SIDL && pp->p_opptr == parent) { 925 /* 926 * using P_ALL here avoids match_process() doing the 927 * same work that we just did, but incorrectly for 928 * this scenario. 929 */ 930 if (match_process(parent, &pp, P_ALL, id, options, 931 NULL, si)) 932 return true; 933 } 934 return false; 935 } 936 937 /* 938 * For the hard cases, just look everywhere to see if some 939 * stolen (reparented) process is really our lost child. 940 * Then check if that process could satisfy the wait conditions. 941 */ 942 943 /* 944 * XXX inefficient, but hopefully fairly rare. 945 * XXX should really use a list of reparented processes. 946 */ 947 PROCLIST_FOREACH(pp, &allproc) { 948 if (pp->p_stat == SIDL) /* XXX impossible ?? */ 949 continue; 950 if (pp->p_opptr == parent && 951 match_process(parent, &pp, idtype, id, options, NULL, si)) 952 return true; 953 } 954 PROCLIST_FOREACH(pp, &zombproc) { 955 if (pp->p_stat == SIDL) /* XXX impossible ?? */ 956 continue; 957 if (pp->p_opptr == parent && 958 match_process(parent, &pp, idtype, id, options, NULL, si)) 959 return true; 960 } 961 962 return false; 963 } 964 965 /* 966 * Scan list of child processes for a child process that has stopped or 967 * exited. Used by sys_wait4 and 'compat' equivalents. 968 * 969 * Must be called with the proc_lock held, and may release while waiting. 970 */ 971 static int 972 find_stopped_child(struct proc *parent, idtype_t idtype, id_t id, int options, 973 struct proc **child_p, struct wrusage *wru, siginfo_t *si) 974 { 975 struct proc *child, *dead; 976 int error; 977 978 KASSERT(mutex_owned(proc_lock)); 979 980 if (options & ~WALLOPTS) { 981 *child_p = NULL; 982 return EINVAL; 983 } 984 985 if ((options & WSELECTOPTS) == 0) { 986 /* 987 * We will be unable to find any matching processes, 988 * because there are no known events to look for. 989 * Prefer to return error instead of blocking 990 * indefinitely. 991 */ 992 *child_p = NULL; 993 return EINVAL; 994 } 995 996 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) { 997 mutex_enter(parent->p_lock); 998 id = (id_t)parent->p_pgid; 999 mutex_exit(parent->p_lock); 1000 idtype = P_PGID; 1001 } 1002 1003 for (;;) { 1004 error = ECHILD; 1005 dead = NULL; 1006 1007 LIST_FOREACH(child, &parent->p_children, p_sibling) { 1008 int rv = match_process(parent, &child, idtype, id, 1009 options, wru, si); 1010 if (rv == -1) 1011 break; 1012 if (rv == 0) 1013 continue; 1014 1015 /* 1016 * Wait for processes with p_exitsig != SIGCHLD 1017 * processes only if WALTSIG is set; wait for 1018 * processes with p_exitsig == SIGCHLD only 1019 * if WALTSIG is clear. 1020 */ 1021 if (((options & WALLSIG) == 0) && 1022 (options & WALTSIG ? child->p_exitsig == SIGCHLD 1023 : P_EXITSIG(child) != SIGCHLD)){ 1024 if (rv == 2) { 1025 child = NULL; 1026 break; 1027 } 1028 continue; 1029 } 1030 1031 error = 0; 1032 if ((options & WNOZOMBIE) == 0) { 1033 if (child->p_stat == SZOMB) 1034 break; 1035 if (child->p_stat == SDEAD) { 1036 /* 1037 * We may occasionally arrive here 1038 * after receiving a signal, but 1039 * immediately before the child 1040 * process is zombified. The wait 1041 * will be short, so avoid returning 1042 * to userspace. 1043 */ 1044 dead = child; 1045 } 1046 } 1047 1048 if ((options & WCONTINUED) != 0 && 1049 child->p_xsig == SIGCONT && 1050 (child->p_sflag & PS_CONTINUED)) { 1051 if ((options & WNOWAIT) == 0) { 1052 child->p_sflag &= ~PS_CONTINUED; 1053 child->p_waited = 1; 1054 parent->p_nstopchild--; 1055 } 1056 if (si) { 1057 si->si_status = child->p_xsig; 1058 si->si_code = CLD_CONTINUED; 1059 } 1060 break; 1061 } 1062 1063 if ((options & (WTRAPPED|WSTOPPED)) != 0 && 1064 child->p_stat == SSTOP && 1065 child->p_waited == 0 && 1066 ((child->p_slflag & PSL_TRACED) || 1067 options & (WUNTRACED|WSTOPPED))) { 1068 if ((options & WNOWAIT) == 0) { 1069 child->p_waited = 1; 1070 parent->p_nstopchild--; 1071 } 1072 if (si) { 1073 si->si_status = child->p_xsig; 1074 si->si_code = 1075 (child->p_slflag & PSL_TRACED) ? 1076 CLD_TRAPPED : CLD_STOPPED; 1077 } 1078 break; 1079 } 1080 if (parent->p_nstopchild == 0 || rv == 2) { 1081 child = NULL; 1082 break; 1083 } 1084 } 1085 1086 /* 1087 * If we found nothing, but we are the bereaved parent 1088 * of a stolen child, look and see if that child (or 1089 * one of them) meets our search criteria. If so, then 1090 * we cannot succeed, but we can hang (wait...), 1091 * or if WNOHANG, return 0 instead of ECHILD 1092 */ 1093 if (child == NULL && error == ECHILD && 1094 (parent->p_slflag & PSL_CHTRACED) && 1095 debugged_child_exists(idtype, id, options, si, parent)) 1096 error = 0; 1097 1098 if (child != NULL || error != 0 || 1099 ((options & WNOHANG) != 0 && dead == NULL)) { 1100 *child_p = child; 1101 return error; 1102 } 1103 1104 /* 1105 * Wait for another child process to stop. 1106 */ 1107 error = cv_wait_sig(&parent->p_waitcv, proc_lock); 1108 1109 if (error != 0) { 1110 *child_p = NULL; 1111 return error; 1112 } 1113 } 1114 } 1115 1116 /* 1117 * Free a process after parent has taken all the state info. Must be called 1118 * with the proclist lock held, and will release before returning. 1119 * 1120 * *ru is returned to the caller, and must be freed by the caller. 1121 */ 1122 static void 1123 proc_free(struct proc *p, struct wrusage *wru) 1124 { 1125 struct proc *parent = p->p_pptr; 1126 struct lwp *l; 1127 ksiginfo_t ksi; 1128 kauth_cred_t cred1, cred2; 1129 uid_t uid; 1130 1131 KASSERT(mutex_owned(proc_lock)); 1132 KASSERT(p->p_nlwps == 1); 1133 KASSERT(p->p_nzlwps == 1); 1134 KASSERT(p->p_nrlwps == 0); 1135 KASSERT(p->p_stat == SZOMB); 1136 1137 /* 1138 * If we got the child via ptrace(2) or procfs, and 1139 * the parent is different (meaning the process was 1140 * attached, rather than run as a child), then we need 1141 * to give it back to the old parent, and send the 1142 * parent the exit signal. The rest of the cleanup 1143 * will be done when the old parent waits on the child. 1144 */ 1145 if ((p->p_slflag & PSL_TRACED) != 0 && p->p_opptr != parent) { 1146 mutex_enter(p->p_lock); 1147 p->p_slflag &= ~(PSL_TRACED|PSL_SYSCALL); 1148 mutex_exit(p->p_lock); 1149 parent = (p->p_opptr == NULL) ? initproc : p->p_opptr; 1150 proc_reparent(p, parent); 1151 p->p_opptr = NULL; 1152 if (p->p_exitsig != 0) { 1153 exit_psignal(p, parent, &ksi); 1154 kpsignal(parent, &ksi, NULL); 1155 } 1156 cv_broadcast(&parent->p_waitcv); 1157 mutex_exit(proc_lock); 1158 return; 1159 } 1160 1161 sched_proc_exit(parent, p); 1162 1163 /* 1164 * Add child times of exiting process onto its own times. 1165 * This cannot be done any earlier else it might get done twice. 1166 */ 1167 l = LIST_FIRST(&p->p_lwps); 1168 p->p_stats->p_ru.ru_nvcsw += (l->l_ncsw - l->l_nivcsw); 1169 p->p_stats->p_ru.ru_nivcsw += l->l_nivcsw; 1170 ruadd(&p->p_stats->p_ru, &l->l_ru); 1171 ruadd(&p->p_stats->p_ru, &p->p_stats->p_cru); 1172 ruadd(&parent->p_stats->p_cru, &p->p_stats->p_ru); 1173 if (wru != NULL) { 1174 wru->wru_self = p->p_stats->p_ru; 1175 wru->wru_children = p->p_stats->p_cru; 1176 } 1177 p->p_xsig = 0; 1178 p->p_xexit = 0; 1179 1180 /* 1181 * At this point we are going to start freeing the final resources. 1182 * If anyone tries to access the proc structure after here they will 1183 * get a shock - bits are missing. Attempt to make it hard! We 1184 * don't bother with any further locking past this point. 1185 */ 1186 p->p_stat = SIDL; /* not even a zombie any more */ 1187 LIST_REMOVE(p, p_list); /* off zombproc */ 1188 parent->p_nstopchild--; 1189 LIST_REMOVE(p, p_sibling); 1190 1191 /* 1192 * Let pid be reallocated. 1193 */ 1194 proc_free_pid(p->p_pid); 1195 1196 /* 1197 * Unlink process from its process group. 1198 * Releases the proc_lock. 1199 */ 1200 proc_leavepgrp(p); 1201 1202 /* 1203 * Delay release until after lwp_free. 1204 */ 1205 cred2 = l->l_cred; 1206 1207 /* 1208 * Free the last LWP's resources. 1209 * 1210 * lwp_free ensures the LWP is no longer running on another CPU. 1211 */ 1212 lwp_free(l, false, true); 1213 1214 /* 1215 * Now no one except us can reach the process p. 1216 */ 1217 1218 /* 1219 * Decrement the count of procs running with this uid. 1220 */ 1221 cred1 = p->p_cred; 1222 uid = kauth_cred_getuid(cred1); 1223 (void)chgproccnt(uid, -1); 1224 1225 /* 1226 * Release substructures. 1227 */ 1228 1229 lim_free(p->p_limit); 1230 pstatsfree(p->p_stats); 1231 kauth_cred_free(cred1); 1232 kauth_cred_free(cred2); 1233 1234 /* 1235 * Release reference to text vnode 1236 */ 1237 if (p->p_textvp) 1238 vrele(p->p_textvp); 1239 kmem_strfree(p->p_path); 1240 1241 mutex_destroy(&p->p_auxlock); 1242 mutex_obj_free(p->p_lock); 1243 mutex_destroy(&p->p_stmutex); 1244 cv_destroy(&p->p_waitcv); 1245 cv_destroy(&p->p_lwpcv); 1246 rw_destroy(&p->p_reflock); 1247 1248 proc_free_mem(p); 1249 } 1250 1251 /* 1252 * Change the parent of a process for tracing purposes. 1253 */ 1254 void 1255 proc_changeparent(struct proc *t, struct proc *p) 1256 { 1257 SET(t->p_slflag, PSL_TRACED); 1258 t->p_opptr = t->p_pptr; 1259 if (t->p_pptr == p) 1260 return; 1261 struct proc *parent = t->p_pptr; 1262 1263 if (parent->p_lock < t->p_lock) { 1264 if (!mutex_tryenter(parent->p_lock)) { 1265 mutex_exit(t->p_lock); 1266 mutex_enter(parent->p_lock); 1267 mutex_enter(t->p_lock); 1268 } 1269 } else if (parent->p_lock > t->p_lock) { 1270 mutex_enter(parent->p_lock); 1271 } 1272 parent->p_slflag |= PSL_CHTRACED; 1273 proc_reparent(t, p); 1274 if (parent->p_lock != t->p_lock) 1275 mutex_exit(parent->p_lock); 1276 } 1277 1278 /* 1279 * make process 'parent' the new parent of process 'child'. 1280 * 1281 * Must be called with proc_lock held. 1282 */ 1283 void 1284 proc_reparent(struct proc *child, struct proc *parent) 1285 { 1286 1287 KASSERT(mutex_owned(proc_lock)); 1288 1289 if (child->p_pptr == parent) 1290 return; 1291 1292 if (child->p_stat == SZOMB || child->p_stat == SDEAD || 1293 (child->p_stat == SSTOP && !child->p_waited)) { 1294 child->p_pptr->p_nstopchild--; 1295 parent->p_nstopchild++; 1296 } 1297 if (parent == initproc) { 1298 child->p_exitsig = SIGCHLD; 1299 child->p_ppid = parent->p_pid; 1300 } 1301 1302 LIST_REMOVE(child, p_sibling); 1303 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 1304 child->p_pptr = parent; 1305 } 1306