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