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 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)kern_exit.c 8.7 (Berkeley) 2/12/94 39 * $FreeBSD: src/sys/kern/kern_exit.c,v 1.92.2.11 2003/01/13 22:51:16 dillon Exp $ 40 * $DragonFly: src/sys/kern/kern_exit.c,v 1.22 2003/07/30 00:19:14 dillon Exp $ 41 */ 42 43 #include "opt_compat.h" 44 #include "opt_ktrace.h" 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/sysproto.h> 49 #include <sys/kernel.h> 50 #include <sys/malloc.h> 51 #include <sys/proc.h> 52 #include <sys/pioctl.h> 53 #include <sys/tty.h> 54 #include <sys/wait.h> 55 #include <sys/vnode.h> 56 #include <sys/resourcevar.h> 57 #include <sys/signalvar.h> 58 #include <sys/ptrace.h> 59 #include <sys/acct.h> /* for acct_process() function prototype */ 60 #include <sys/filedesc.h> 61 #include <sys/shm.h> 62 #include <sys/sem.h> 63 #include <sys/aio.h> 64 #include <sys/jail.h> 65 66 #include <vm/vm.h> 67 #include <vm/vm_param.h> 68 #include <sys/lock.h> 69 #include <vm/pmap.h> 70 #include <vm/vm_map.h> 71 #include <vm/vm_zone.h> 72 #include <vm/vm_extern.h> 73 #include <sys/user.h> 74 75 /* Required to be non-static for SysVR4 emulator */ 76 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 77 78 static MALLOC_DEFINE(M_ATEXIT, "atexit", "atexit callback"); 79 80 static int wait1 __P((struct wait_args *, int)); 81 82 /* 83 * callout list for things to do at exit time 84 */ 85 struct exitlist { 86 exitlist_fn function; 87 TAILQ_ENTRY(exitlist) next; 88 }; 89 90 TAILQ_HEAD(exit_list_head, exitlist); 91 static struct exit_list_head exit_list = TAILQ_HEAD_INITIALIZER(exit_list); 92 93 /* 94 * exit -- 95 * Death of process. 96 * 97 * SYS_EXIT_ARGS(int rval) 98 */ 99 void 100 sys_exit(struct sys_exit_args *uap) 101 { 102 exit1(W_EXITCODE(uap->rval, 0)); 103 /* NOTREACHED */ 104 } 105 106 /* 107 * Exit: deallocate address space and other resources, change proc state 108 * to zombie, and unlink proc from allproc and parent's lists. Save exit 109 * status and rusage for wait(). Check for child processes and orphan them. 110 */ 111 void 112 exit1(int rv) 113 { 114 struct proc *p = curproc; 115 struct proc *q, *nq; 116 struct vmspace *vm; 117 struct vnode *vtmp; 118 struct exitlist *ep; 119 120 if (p->p_pid == 1) { 121 printf("init died (signal %d, exit %d)\n", 122 WTERMSIG(rv), WEXITSTATUS(rv)); 123 panic("Going nowhere without my init!"); 124 } 125 126 aio_proc_rundown(p); 127 128 /* are we a task leader? */ 129 if(p == p->p_leader) { 130 struct kill_args killArgs; 131 killArgs.signum = SIGKILL; 132 q = p->p_peers; 133 while(q) { 134 killArgs.pid = q->p_pid; 135 /* 136 * The interface for kill is better 137 * than the internal signal 138 */ 139 kill(&killArgs); 140 nq = q; 141 q = q->p_peers; 142 } 143 while (p->p_peers) 144 tsleep((caddr_t)p, 0, "exit1", 0); 145 } 146 147 #ifdef PGINPROF 148 vmsizmon(); 149 #endif 150 STOPEVENT(p, S_EXIT, rv); 151 wakeup(&p->p_stype); /* Wakeup anyone in procfs' PIOCWAIT */ 152 153 /* 154 * Check if any loadable modules need anything done at process exit. 155 * e.g. SYSV IPC stuff 156 * XXX what if one of these generates an error? 157 */ 158 TAILQ_FOREACH(ep, &exit_list, next) 159 (*ep->function)(p->p_thread); 160 161 if (p->p_flag & P_PROFIL) 162 stopprofclock(p); 163 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage), 164 M_ZOMBIE, M_WAITOK); 165 /* 166 * If parent is waiting for us to exit or exec, 167 * P_PPWAIT is set; we will wakeup the parent below. 168 */ 169 p->p_flag &= ~(P_TRACED | P_PPWAIT); 170 p->p_flag |= P_WEXIT; 171 SIGEMPTYSET(p->p_siglist); 172 if (timevalisset(&p->p_realtimer.it_value)) 173 untimeout(realitexpire, (caddr_t)p, p->p_ithandle); 174 175 /* 176 * Reset any sigio structures pointing to us as a result of 177 * F_SETOWN with our pid. 178 */ 179 funsetownlst(&p->p_sigiolst); 180 181 /* 182 * Close open files and release open-file table. 183 * This may block! 184 */ 185 fdfree(p); 186 187 if(p->p_leader->p_peers) { 188 q = p->p_leader; 189 while(q->p_peers != p) 190 q = q->p_peers; 191 q->p_peers = p->p_peers; 192 wakeup((caddr_t)p->p_leader); 193 } 194 195 /* 196 * XXX Shutdown SYSV semaphores 197 */ 198 semexit(p); 199 200 /* The next two chunks should probably be moved to vmspace_exit. */ 201 vm = p->p_vmspace; 202 /* 203 * Release user portion of address space. 204 * This releases references to vnodes, 205 * which could cause I/O if the file has been unlinked. 206 * Need to do this early enough that we can still sleep. 207 * Can't free the entire vmspace as the kernel stack 208 * may be mapped within that space also. 209 * 210 * Processes sharing the same vmspace may exit in one order, and 211 * get cleaned up by vmspace_exit() in a different order. The 212 * last exiting process to reach this point releases as much of 213 * the environment as it can, and the last process cleaned up 214 * by vmspace_exit() (which decrements exitingcnt) cleans up the 215 * remainder. 216 */ 217 ++vm->vm_exitingcnt; 218 if (--vm->vm_refcnt == 0) { 219 shmexit(vm); 220 pmap_remove_pages(vmspace_pmap(vm), VM_MIN_ADDRESS, 221 VM_MAXUSER_ADDRESS); 222 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS, 223 VM_MAXUSER_ADDRESS); 224 } 225 226 if (SESS_LEADER(p)) { 227 struct session *sp = p->p_session; 228 229 if (sp->s_ttyvp) { 230 /* 231 * Controlling process. 232 * Signal foreground pgrp, 233 * drain controlling terminal 234 * and revoke access to controlling terminal. 235 */ 236 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 237 if (sp->s_ttyp->t_pgrp) 238 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 239 (void) ttywait(sp->s_ttyp); 240 /* 241 * The tty could have been revoked 242 * if we blocked. 243 */ 244 if (sp->s_ttyvp) 245 VOP_REVOKE(sp->s_ttyvp, REVOKEALL); 246 } 247 if (sp->s_ttyvp) 248 vrele(sp->s_ttyvp); 249 sp->s_ttyvp = NULL; 250 /* 251 * s_ttyp is not zero'd; we use this to indicate 252 * that the session once had a controlling terminal. 253 * (for logging and informational purposes) 254 */ 255 } 256 sp->s_leader = NULL; 257 } 258 fixjobc(p, p->p_pgrp, 0); 259 (void)acct_process(p); 260 #ifdef KTRACE 261 /* 262 * release trace file 263 */ 264 p->p_traceflag = 0; /* don't trace the vrele() */ 265 if ((vtmp = p->p_tracep) != NULL) { 266 p->p_tracep = NULL; 267 vrele(vtmp); 268 } 269 #endif 270 /* 271 * Release reference to text vnode 272 */ 273 if ((vtmp = p->p_textvp) != NULL) { 274 p->p_textvp = NULL; 275 vrele(vtmp); 276 } 277 278 /* 279 * Once we set SZOMB the process can get reaped. The wait1 code 280 * will also wait for TDF_RUNNING to be cleared in the thread's flags, 281 * indicating that it has been completely switched out. 282 */ 283 284 /* 285 * Remove proc from allproc queue and pidhash chain. 286 * Place onto zombproc. Unlink from parent's child list. 287 */ 288 LIST_REMOVE(p, p_list); 289 LIST_INSERT_HEAD(&zombproc, p, p_list); 290 p->p_stat = SZOMB; 291 292 LIST_REMOVE(p, p_hash); 293 294 q = LIST_FIRST(&p->p_children); 295 if (q) /* only need this if any child is S_ZOMB */ 296 wakeup((caddr_t) initproc); 297 for (; q != 0; q = nq) { 298 nq = LIST_NEXT(q, p_sibling); 299 LIST_REMOVE(q, p_sibling); 300 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling); 301 q->p_pptr = initproc; 302 q->p_sigparent = SIGCHLD; 303 /* 304 * Traced processes are killed 305 * since their existence means someone is screwing up. 306 */ 307 if (q->p_flag & P_TRACED) { 308 q->p_flag &= ~P_TRACED; 309 psignal(q, SIGKILL); 310 } 311 } 312 313 /* 314 * Save exit status and final rusage info, adding in child rusage 315 * info and self times. 316 */ 317 p->p_xstat = rv; 318 *p->p_ru = p->p_stats->p_ru; 319 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 320 ruadd(p->p_ru, &p->p_stats->p_cru); 321 322 /* 323 * notify interested parties of our demise. 324 */ 325 KNOTE(&p->p_klist, NOTE_EXIT); 326 327 /* 328 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 329 * flag set, notify process 1 instead (and hope it will handle 330 * this situation). 331 */ 332 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) { 333 struct proc *pp = p->p_pptr; 334 proc_reparent(p, initproc); 335 /* 336 * If this was the last child of our parent, notify 337 * parent, so in case he was wait(2)ing, he will 338 * continue. 339 */ 340 if (LIST_EMPTY(&pp->p_children)) 341 wakeup((caddr_t)pp); 342 } 343 344 if (p->p_sigparent && p->p_pptr != initproc) { 345 psignal(p->p_pptr, p->p_sigparent); 346 } else { 347 psignal(p->p_pptr, SIGCHLD); 348 } 349 350 wakeup((caddr_t)p->p_pptr); 351 #if defined(tahoe) 352 /* move this to cpu_exit */ 353 p->p_thread->td_pcb->pcb_saveacc.faddr = (float *)NULL; 354 #endif 355 /* 356 * cpu_exit is responsible for clearing curproc, since 357 * it is heavily integrated with the thread/switching sequence. 358 * 359 * Other substructures are freed from wait(). 360 */ 361 if (--p->p_limit->p_refcnt == 0) { 362 FREE(p->p_limit, M_SUBPROC); 363 p->p_limit = NULL; 364 } 365 366 /* 367 * Release the P_CURPROC designation on the process so the userland 368 * scheduler can work in someone else. 369 */ 370 release_curproc(p); 371 372 /* 373 * Finally, call machine-dependent code to release the remaining 374 * resources including address space, the kernel stack and pcb. 375 * The address space is released by "vmspace_free(p->p_vmspace)"; 376 * This is machine-dependent, as we may have to change stacks 377 * or ensure that the current one isn't reallocated before we 378 * finish. cpu_exit will end with a call to cpu_switch(), finishing 379 * our execution (pun intended). 380 */ 381 cpu_proc_exit(); 382 } 383 384 #ifdef COMPAT_43 385 /* 386 * owait() 387 * 388 * owait_args(int dummy) 389 */ 390 int 391 owait(struct owait_args *uap) 392 { 393 struct wait_args w; 394 395 w.options = 0; 396 w.rusage = NULL; 397 w.pid = WAIT_ANY; 398 w.status = NULL; 399 return (wait1(&w, 1)); 400 } 401 #endif /* COMPAT_43 */ 402 403 int 404 wait4(struct wait_args *uap) 405 { 406 return (wait1(uap, 0)); 407 } 408 409 /* 410 * wait1() 411 * 412 * wait_args(int pid, int *status, int options, struct rusage *rusage) 413 */ 414 static int 415 wait1(struct wait_args *uap, int compat) 416 { 417 struct proc *q = curproc; 418 struct proc *p, *t; 419 int status, nfound, error; 420 421 if (uap->pid == 0) 422 uap->pid = -q->p_pgid; 423 if (uap->options &~ (WUNTRACED|WNOHANG|WLINUXCLONE)) 424 return (EINVAL); 425 loop: 426 nfound = 0; 427 LIST_FOREACH(p, &q->p_children, p_sibling) { 428 if (uap->pid != WAIT_ANY && 429 p->p_pid != uap->pid && p->p_pgid != -uap->pid) 430 continue; 431 432 /* This special case handles a kthread spawned by linux_clone 433 * (see linux_misc.c). The linux_wait4 and linux_waitpid functions 434 * need to be able to distinguish between waiting on a process and 435 * waiting on a thread. It is a thread if p_sigparent is not SIGCHLD, 436 * and the WLINUXCLONE option signifies we want to wait for threads 437 * and not processes. 438 */ 439 if ((p->p_sigparent != SIGCHLD) ^ ((uap->options & WLINUXCLONE) != 0)) 440 continue; 441 442 nfound++; 443 if (p->p_stat == SZOMB) { 444 /* 445 * The process's thread may still be in the middle 446 * of switching away, we can't rip its stack out from 447 * under it until TDF_RUNNING clears! 448 * 449 * YYY no wakeup occurs so we depend on the timeout. 450 */ 451 if ((p->p_thread->td_flags & TDF_RUNNING) != 0) { 452 tsleep(p->p_thread, 0, "reap", 1); 453 goto loop; 454 } 455 456 /* 457 * Other kernel threads may be in the middle of 458 * accessing the proc. For example, kern/kern_proc.c 459 * could be blocked writing proc data to a sysctl. 460 * At the moment, if this occurs, we are not woken 461 * up and rely on a one-second retry. 462 */ 463 if (p->p_lock) { 464 printf("Diagnostic: waiting for p_lock\n"); 465 while (p->p_lock) 466 tsleep(p, 0, "reap2", hz); 467 } 468 lwkt_wait_free(p->p_thread); 469 470 /* charge childs scheduling cpu usage to parent */ 471 if (curproc->p_pid != 1) { 472 curproc->p_estcpu = 473 ESTCPULIM(curproc->p_estcpu + p->p_estcpu); 474 } 475 476 uap->sysmsg_fds[0] = p->p_pid; 477 #ifdef COMPAT_43 478 if (compat) 479 uap->sysmsg_fds[1] = p->p_xstat; 480 else 481 #endif 482 if (uap->status) { 483 status = p->p_xstat; /* convert to int */ 484 if ((error = copyout((caddr_t)&status, 485 (caddr_t)uap->status, sizeof(status)))) 486 return (error); 487 } 488 if (uap->rusage && (error = copyout((caddr_t)p->p_ru, 489 (caddr_t)uap->rusage, sizeof (struct rusage)))) 490 return (error); 491 /* 492 * If we got the child via a ptrace 'attach', 493 * we need to give it back to the old parent. 494 */ 495 if (p->p_oppid && (t = pfind(p->p_oppid))) { 496 p->p_oppid = 0; 497 proc_reparent(p, t); 498 psignal(t, SIGCHLD); 499 wakeup((caddr_t)t); 500 return (0); 501 } 502 p->p_xstat = 0; 503 ruadd(&q->p_stats->p_cru, p->p_ru); 504 FREE(p->p_ru, M_ZOMBIE); 505 p->p_ru = NULL; 506 507 /* 508 * Decrement the count of procs running with this uid. 509 */ 510 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 511 512 /* 513 * Free up credentials. 514 */ 515 crfree(p->p_ucred); 516 p->p_ucred = NULL; 517 518 /* 519 * Remove unused arguments 520 */ 521 if (p->p_args && --p->p_args->ar_ref == 0) 522 FREE(p->p_args, M_PARGS); 523 524 /* 525 * Finally finished with old proc entry. 526 * Unlink it from its process group and free it. 527 */ 528 leavepgrp(p); 529 LIST_REMOVE(p, p_list); /* off zombproc */ 530 LIST_REMOVE(p, p_sibling); 531 532 if (--p->p_procsig->ps_refcnt == 0) { 533 if (p->p_sigacts != &p->p_addr->u_sigacts) 534 FREE(p->p_sigacts, M_SUBPROC); 535 FREE(p->p_procsig, M_SUBPROC); 536 p->p_procsig = NULL; 537 } 538 539 vm_waitproc(p); 540 zfree(proc_zone, p); 541 nprocs--; 542 return (0); 543 } 544 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 545 (p->p_flag & P_TRACED || uap->options & WUNTRACED)) { 546 p->p_flag |= P_WAITED; 547 uap->sysmsg_fds[0] = p->p_pid; 548 #ifdef COMPAT_43 549 if (compat) { 550 uap->sysmsg_fds[1] = W_STOPCODE(p->p_xstat); 551 error = 0; 552 } else 553 #endif 554 if (uap->status) { 555 status = W_STOPCODE(p->p_xstat); 556 error = copyout((caddr_t)&status, 557 (caddr_t)uap->status, sizeof(status)); 558 } else 559 error = 0; 560 return (error); 561 } 562 } 563 if (nfound == 0) 564 return (ECHILD); 565 if (uap->options & WNOHANG) { 566 uap->sysmsg_fds[0] = 0; 567 return (0); 568 } 569 if ((error = tsleep((caddr_t)q, PCATCH, "wait", 0))) 570 return (error); 571 goto loop; 572 } 573 574 /* 575 * make process 'parent' the new parent of process 'child'. 576 */ 577 void 578 proc_reparent(child, parent) 579 struct proc *child; 580 struct proc *parent; 581 { 582 583 if (child->p_pptr == parent) 584 return; 585 586 LIST_REMOVE(child, p_sibling); 587 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 588 child->p_pptr = parent; 589 } 590 591 /* 592 * The next two functions are to handle adding/deleting items on the 593 * exit callout list 594 * 595 * at_exit(): 596 * Take the arguments given and put them onto the exit callout list, 597 * However first make sure that it's not already there. 598 * returns 0 on success. 599 */ 600 601 int 602 at_exit(function) 603 exitlist_fn function; 604 { 605 struct exitlist *ep; 606 607 #ifdef INVARIANTS 608 /* Be noisy if the programmer has lost track of things */ 609 if (rm_at_exit(function)) 610 printf("WARNING: exit callout entry (%p) already present\n", 611 function); 612 #endif 613 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT); 614 if (ep == NULL) 615 return (ENOMEM); 616 ep->function = function; 617 TAILQ_INSERT_TAIL(&exit_list, ep, next); 618 return (0); 619 } 620 621 /* 622 * Scan the exit callout list for the given item and remove it. 623 * Returns the number of items removed (0 or 1) 624 */ 625 int 626 rm_at_exit(function) 627 exitlist_fn function; 628 { 629 struct exitlist *ep; 630 631 TAILQ_FOREACH(ep, &exit_list, next) { 632 if (ep->function == function) { 633 TAILQ_REMOVE(&exit_list, ep, next); 634 free(ep, M_ATEXIT); 635 return(1); 636 } 637 } 638 return (0); 639 } 640 641 void check_sigacts (void) 642 { 643 struct proc *p = curproc; 644 struct sigacts *pss; 645 int s; 646 647 if (p->p_procsig->ps_refcnt == 1 && 648 p->p_sigacts != &p->p_addr->u_sigacts) { 649 pss = p->p_sigacts; 650 s = splhigh(); 651 p->p_addr->u_sigacts = *pss; 652 p->p_sigacts = &p->p_addr->u_sigacts; 653 splx(s); 654 FREE(pss, M_SUBPROC); 655 } 656 } 657 658