kern_exit.c revision 41931
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 * $Id: kern_exit.c,v 1.69 1998/11/11 10:03:54 truckman Exp $ 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 64#ifdef COMPAT_43 65#include <machine/reg.h> 66#include <machine/psl.h> 67#endif 68#include <machine/limits.h> /* for UCHAR_MAX = typeof(p_priority)_MAX */ 69 70#include <vm/vm.h> 71#include <vm/vm_param.h> 72#include <sys/lock.h> 73#include <vm/pmap.h> 74#include <vm/vm_map.h> 75#include <vm/vm_zone.h> 76 77static MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 78 79static int wait1 __P((struct proc *, struct wait_args *, int)); 80 81/* 82 * callout list for things to do at exit time 83 */ 84typedef struct exit_list_element { 85 struct exit_list_element *next; 86 exitlist_fn function; 87} *ele_p; 88 89static ele_p exit_list; 90 91/* 92 * exit -- 93 * Death of process. 94 */ 95void 96exit(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 */ 112void 113exit1(p, rv) 114 register struct proc *p; 115 int rv; 116{ 117 register struct proc *q, *nq; 118 register struct vmspace *vm; 119 ele_p ep = exit_list; 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 * orphan the threads so we don't mess up 145 * when they call exit 146 */ 147 nq->p_peers = 0; 148 nq->p_leader = nq; 149 } 150 151 /* otherwise are we a peer? */ 152 } else if(p->p_peers) { 153 q = p->p_leader; 154 while(q->p_peers != p) 155 q = q->p_peers; 156 q->p_peers = p->p_peers; 157 } 158 159#ifdef PGINPROF 160 vmsizmon(); 161#endif 162 STOPEVENT(p, S_EXIT, rv); 163 164 /* 165 * Check if any LKMs need anything done at process exit. 166 * e.g. SYSV IPC stuff 167 * XXX what if one of these generates an error? 168 */ 169 while (ep) { 170 (*ep->function)(p); 171 ep = ep->next; 172 } 173 174 if (p->p_flag & P_PROFIL) 175 stopprofclock(p); 176 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage), 177 M_ZOMBIE, M_WAITOK); 178 /* 179 * If parent is waiting for us to exit or exec, 180 * P_PPWAIT is set; we will wakeup the parent below. 181 */ 182 p->p_flag &= ~(P_TRACED | P_PPWAIT); 183 p->p_flag |= P_WEXIT; 184#ifndef COMPAT_LINUX_THREADS 185 p->p_sigignore = ~0; 186#endif /* COMPAT_LINUX_THREADS */ 187 p->p_siglist = 0; 188 if (timevalisset(&p->p_realtimer.it_value)) 189 untimeout(realitexpire, (caddr_t)p, p->p_ithandle); 190 191 /* 192 * Reset any sigio structures pointing to us as a result of 193 * F_SETOWN with our pid. 194 */ 195 funsetownlst(&p->p_sigiolst); 196 197 /* 198 * Close open files and release open-file table. 199 * This may block! 200 */ 201 fdfree(p); 202 203 /* 204 * XXX Shutdown SYSV semaphores 205 */ 206 semexit(p); 207 208 /* The next two chunks should probably be moved to vmspace_exit. */ 209 vm = p->p_vmspace; 210 /* 211 * Release user portion of address space. 212 * This releases references to vnodes, 213 * which could cause I/O if the file has been unlinked. 214 * Need to do this early enough that we can still sleep. 215 * Can't free the entire vmspace as the kernel stack 216 * may be mapped within that space also. 217 */ 218 if (vm->vm_refcnt == 1) { 219 if (vm->vm_shm) 220 shmexit(p); 221 pmap_remove_pages(&vm->vm_pmap, VM_MIN_ADDRESS, 222 VM_MAXUSER_ADDRESS); 223 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS, 224 VM_MAXUSER_ADDRESS); 225 } 226 227 if (SESS_LEADER(p)) { 228 register struct session *sp = p->p_session; 229 230 if (sp->s_ttyvp) { 231 /* 232 * Controlling process. 233 * Signal foreground pgrp, 234 * drain controlling terminal 235 * and revoke access to controlling terminal. 236 */ 237 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 238 if (sp->s_ttyp->t_pgrp) 239 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 240 (void) ttywait(sp->s_ttyp); 241 /* 242 * The tty could have been revoked 243 * if we blocked. 244 */ 245 if (sp->s_ttyvp) 246 VOP_REVOKE(sp->s_ttyvp, REVOKEALL); 247 } 248 if (sp->s_ttyvp) 249 vrele(sp->s_ttyvp); 250 sp->s_ttyvp = NULL; 251 /* 252 * s_ttyp is not zero'd; we use this to indicate 253 * that the session once had a controlling terminal. 254 * (for logging and informational purposes) 255 */ 256 } 257 sp->s_leader = NULL; 258 } 259 fixjobc(p, p->p_pgrp, 0); 260 (void)acct_process(p); 261#ifdef KTRACE 262 /* 263 * release trace file 264 */ 265 p->p_traceflag = 0; /* don't trace the vrele() */ 266 if (p->p_tracep) 267 vrele(p->p_tracep); 268#endif 269 /* 270 * Remove proc from allproc queue and pidhash chain. 271 * Place onto zombproc. Unlink from parent's child list. 272 */ 273 LIST_REMOVE(p, p_list); 274 LIST_INSERT_HEAD(&zombproc, p, p_list); 275 p->p_stat = SZOMB; 276 277 LIST_REMOVE(p, p_hash); 278 279 q = p->p_children.lh_first; 280 if (q) /* only need this if any child is S_ZOMB */ 281 wakeup((caddr_t) initproc); 282 for (; q != 0; q = nq) { 283 nq = q->p_sibling.le_next; 284 LIST_REMOVE(q, p_sibling); 285 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling); 286 q->p_pptr = initproc; 287#ifdef COMPAT_LINUX_THREADS 288 q->p_sigparent = 0; 289#endif /* COMPAT_LINUX_THREADS */ 290 /* 291 * Traced processes are killed 292 * since their existence means someone is screwing up. 293 */ 294 if (q->p_flag & P_TRACED) { 295 q->p_flag &= ~P_TRACED; 296 psignal(q, SIGKILL); 297 } 298 } 299 300 /* 301 * Save exit status and final rusage info, adding in child rusage 302 * info and self times. 303 */ 304 p->p_xstat = rv; 305 *p->p_ru = p->p_stats->p_ru; 306 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 307 ruadd(p->p_ru, &p->p_stats->p_cru); 308 309 /* 310 * Notify parent that we're gone. If parent has the P_NOCLDWAIT 311 * flag set, notify process 1 instead (and hope it will handle 312 * this situation). 313 */ 314#ifndef COMPAT_LINUX_THREADS 315 if (p->p_pptr->p_flag & P_NOCLDWAIT) { 316#else 317 if (p->p_pptr->p_procsig->ps_flag & P_NOCLDWAIT) { 318#endif /* COMPAT_LINUX_THREADS */ 319 struct proc *pp = p->p_pptr; 320 proc_reparent(p, initproc); 321 /* 322 * If this was the last child of our parent, notify 323 * parent, so in case he was wait(2)ing, he will 324 * continue. 325 */ 326 if (LIST_EMPTY(&pp->p_children)) 327 wakeup((caddr_t)pp); 328 } 329 330#ifndef COMPAT_LINUX_THREADS 331 psignal(p->p_pptr, SIGCHLD); 332#else 333 if (p->p_sigparent && p->p_pptr != initproc) { 334 psignal(p->p_pptr, p->p_sigparent); 335 } else { 336 psignal(p->p_pptr, SIGCHLD); 337 } 338#endif /* COMPAT_LINUX_THREADS */ 339 wakeup((caddr_t)p->p_pptr); 340#if defined(tahoe) 341 /* move this to cpu_exit */ 342 p->p_addr->u_pcb.pcb_savacc.faddr = (float *)NULL; 343#endif 344 /* 345 * Clear curproc after we've done all operations 346 * that could block, and before tearing down the rest 347 * of the process state that might be used from clock, etc. 348 * Also, can't clear curproc while we're still runnable, 349 * as we're not on a run queue (we are current, just not 350 * a proper proc any longer!). 351 * 352 * Other substructures are freed from wait(). 353 */ 354 curproc = NULL; 355 if (--p->p_limit->p_refcnt == 0) { 356 FREE(p->p_limit, M_SUBPROC); 357 p->p_limit = NULL; 358 } 359 360 /* 361 * Finally, call machine-dependent code to release the remaining 362 * resources including address space, the kernel stack and pcb. 363 * The address space is released by "vmspace_free(p->p_vmspace)"; 364 * This is machine-dependent, as we may have to change stacks 365 * or ensure that the current one isn't reallocated before we 366 * finish. cpu_exit will end with a call to cpu_switch(), finishing 367 * our execution (pun intended). 368 */ 369 cpu_exit(p); 370} 371 372#ifdef COMPAT_43 373#if defined(hp300) || defined(luna68k) 374#include <machine/frame.h> 375#define GETPS(rp) ((struct frame *)(rp))->f_sr 376#else 377#define GETPS(rp) (rp)[PS] 378#endif 379 380int 381owait(p, uap) 382 struct proc *p; 383 register struct owait_args /* { 384 int dummy; 385 } */ *uap; 386{ 387 struct wait_args w; 388 389#ifdef PSL_ALLCC 390 if ((GETPS(p->p_md.md_regs) & PSL_ALLCC) != PSL_ALLCC) { 391 w.options = 0; 392 w.rusage = NULL; 393 } else { 394 w.options = p->p_md.md_regs[R0]; 395 w.rusage = (struct rusage *)p->p_md.md_regs[R1]; 396 } 397#else 398 w.options = 0; 399 w.rusage = NULL; 400#endif 401 w.pid = WAIT_ANY; 402 w.status = NULL; 403 return (wait1(p, &w, 1)); 404} 405#endif /* COMPAT_43 */ 406 407int 408wait4(p, uap) 409 struct proc *p; 410 struct wait_args *uap; 411{ 412 413 return (wait1(p, uap, 0)); 414} 415 416static int 417wait1(q, uap, compat) 418 register struct proc *q; 419 register struct wait_args /* { 420 int pid; 421 int *status; 422 int options; 423 struct rusage *rusage; 424 } */ *uap; 425 int compat; 426{ 427 register int nfound; 428 register struct proc *p, *t; 429 int status, error; 430 431 if (uap->pid == 0) 432 uap->pid = -q->p_pgid; 433 if (uap->options &~ (WUNTRACED|WNOHANG)) 434 return (EINVAL); 435loop: 436 nfound = 0; 437 for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) { 438 if (uap->pid != WAIT_ANY && 439 p->p_pid != uap->pid && p->p_pgid != -uap->pid) 440 continue; 441#ifdef COMPAT_LINUX_THREADS 442 #if 0 443 if ((p->p_sigparent != 0) ^ ((uap->options & WLINUXCLONE) != 0)) { 444 continue; 445 } 446 #endif 447 448#endif /* COMPAT_LINUX_THREADS */ 449 nfound++; 450 if (p->p_stat == SZOMB) { 451 /* charge childs scheduling cpu usage to parent */ 452 if (curproc->p_pid != 1) { 453 curproc->p_estcpu = min(curproc->p_estcpu + 454 p->p_estcpu, UCHAR_MAX); 455 } 456 457 q->p_retval[0] = p->p_pid; 458#ifdef COMPAT_43 459 if (compat) 460 q->p_retval[1] = p->p_xstat; 461 else 462#endif 463 if (uap->status) { 464 status = p->p_xstat; /* convert to int */ 465 if ((error = copyout((caddr_t)&status, 466 (caddr_t)uap->status, sizeof(status)))) 467 return (error); 468 } 469 if (uap->rusage && (error = copyout((caddr_t)p->p_ru, 470 (caddr_t)uap->rusage, sizeof (struct rusage)))) 471 return (error); 472 /* 473 * If we got the child via a ptrace 'attach', 474 * we need to give it back to the old parent. 475 */ 476 if (p->p_oppid && (t = pfind(p->p_oppid))) { 477 p->p_oppid = 0; 478 proc_reparent(p, t); 479 psignal(t, SIGCHLD); 480 wakeup((caddr_t)t); 481 return (0); 482 } 483 p->p_xstat = 0; 484 ruadd(&q->p_stats->p_cru, p->p_ru); 485 FREE(p->p_ru, M_ZOMBIE); 486 p->p_ru = NULL; 487 488 /* 489 * Decrement the count of procs running with this uid. 490 */ 491 (void)chgproccnt(p->p_cred->p_ruid, -1); 492 493 /* 494 * Release reference to text vnode 495 */ 496 if (p->p_textvp) 497 vrele(p->p_textvp); 498 499 /* 500 * Free up credentials. 501 */ 502 if (--p->p_cred->p_refcnt == 0) { 503 crfree(p->p_cred->pc_ucred); 504 FREE(p->p_cred, M_SUBPROC); 505 p->p_cred = NULL; 506 } 507 508 /* 509 * Finally finished with old proc entry. 510 * Unlink it from its process group and free it. 511 */ 512 leavepgrp(p); 513 LIST_REMOVE(p, p_list); /* off zombproc */ 514 LIST_REMOVE(p, p_sibling); 515 516#ifdef COMPAT_LINUX_THREADS 517 if (--p->p_procsig->ps_refcnt == 0) { 518 free(p->p_procsig, M_TEMP); 519 p->p_procsig = NULL; 520 p->p_sigacts = NULL; 521 } 522 523#endif /* COMPAT_LINUX_THREADS */ 524 /* 525 * Give machine-dependent layer a chance 526 * to free anything that cpu_exit couldn't 527 * release while still running in process context. 528 */ 529 cpu_wait(p); 530 zfree(proc_zone, p); 531 nprocs--; 532 return (0); 533 } 534 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 535 (p->p_flag & P_TRACED || uap->options & WUNTRACED)) { 536 p->p_flag |= P_WAITED; 537 q->p_retval[0] = p->p_pid; 538#ifdef COMPAT_43 539 if (compat) { 540 q->p_retval[1] = W_STOPCODE(p->p_xstat); 541 error = 0; 542 } else 543#endif 544 if (uap->status) { 545 status = W_STOPCODE(p->p_xstat); 546 error = copyout((caddr_t)&status, 547 (caddr_t)uap->status, sizeof(status)); 548 } else 549 error = 0; 550 return (error); 551 } 552 } 553 if (nfound == 0) 554 return (ECHILD); 555 if (uap->options & WNOHANG) { 556 q->p_retval[0] = 0; 557 return (0); 558 } 559 if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0))) 560 return (error); 561 goto loop; 562} 563 564/* 565 * make process 'parent' the new parent of process 'child'. 566 */ 567void 568proc_reparent(child, parent) 569 register struct proc *child; 570 register struct proc *parent; 571{ 572 573 if (child->p_pptr == parent) 574 return; 575 576 LIST_REMOVE(child, p_sibling); 577 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 578 child->p_pptr = parent; 579#ifdef COMPAT_LINUX_THREADS 580 #if 0 581 child->p_sigparent = 0; 582 #endif 583#endif /* COMPAT_LINUX_THREADS */ 584} 585 586/* 587 * The next two functions are to handle adding/deleting items on the 588 * exit callout list 589 * 590 * at_exit(): 591 * Take the arguments given and put them onto the exit callout list, 592 * However first make sure that it's not already there. 593 * returns 0 on success. 594 */ 595int 596at_exit(function) 597 exitlist_fn function; 598{ 599 ele_p ep; 600 601 /* Be noisy if the programmer has lost track of things */ 602 if (rm_at_exit(function)) 603 printf("exit callout entry already present\n"); 604 ep = malloc(sizeof(*ep), M_TEMP, M_NOWAIT); 605 if (ep == NULL) 606 return (ENOMEM); 607 ep->next = exit_list; 608 ep->function = function; 609 exit_list = ep; 610 return (0); 611} 612/* 613 * Scan the exit callout list for the given items and remove them. 614 * Returns the number of items removed. 615 * Logically this can only be 0 or 1. 616 */ 617int 618rm_at_exit(function) 619 exitlist_fn function; 620{ 621 ele_p *epp, ep; 622 int count; 623 624 count = 0; 625 epp = &exit_list; 626 ep = *epp; 627 while (ep) { 628 if (ep->function == function) { 629 *epp = ep->next; 630 free(ep, M_TEMP); 631 count++; 632 } else { 633 epp = &ep->next; 634 } 635 ep = *epp; 636 } 637 return (count); 638} 639