kern_exit.c revision 53212
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: head/sys/kern/kern_exit.c 53212 1999-11-16 10:56:05Z phk $ 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 <machine/limits.h> /* for UCHAR_MAX = typeof(p_priority)_MAX */ 66 67#include <vm/vm.h> 68#include <vm/vm_param.h> 69#include <sys/lock.h> 70#include <vm/pmap.h> 71#include <vm/vm_map.h> 72#include <vm/vm_zone.h> 73#include <sys/user.h> 74 75/* Required to be non-static for SysVR4 emulator */ 76MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 77 78static int wait1 __P((struct proc *, struct wait_args *, int)); 79 80/* 81 * callout list for things to do at exit time 82 */ 83typedef struct exit_list_element { 84 struct exit_list_element *next; 85 exitlist_fn function; 86} *ele_p; 87 88static ele_p exit_list; 89 90/* 91 * exit -- 92 * Death of process. 93 */ 94void 95exit(p, uap) 96 struct proc *p; 97 struct rexit_args /* { 98 int rval; 99 } */ *uap; 100{ 101 102 exit1(p, 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 */ 111void 112exit1(p, rv) 113 register struct proc *p; 114 int rv; 115{ 116 register struct proc *q, *nq; 117 register struct vmspace *vm; 118 ele_p ep = exit_list; 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(p, &killArgs); 140 nq = q; 141 q = q->p_peers; 142 } 143 while (p->p_peers) 144 tsleep((caddr_t)p, PWAIT, "exit1", 0); 145 } 146 147#ifdef PGINPROF 148 vmsizmon(); 149#endif 150 STOPEVENT(p, S_EXIT, rv); 151 152 /* 153 * Check if any loadable modules need anything done at process exit. 154 * e.g. SYSV IPC stuff 155 * XXX what if one of these generates an error? 156 */ 157 while (ep) { 158 (*ep->function)(p); 159 ep = ep->next; 160 } 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 parent that we're gone. If parent has the PS_NOCLDWAIT 311 * flag set, notify process 1 instead (and hope it will handle 312 * this situation). 313 */ 314 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) { 315 struct proc *pp = p->p_pptr; 316 proc_reparent(p, initproc); 317 /* 318 * If this was the last child of our parent, notify 319 * parent, so in case he was wait(2)ing, he will 320 * continue. 321 */ 322 if (LIST_EMPTY(&pp->p_children)) 323 wakeup((caddr_t)pp); 324 } 325 326 if (p->p_sigparent && p->p_pptr != initproc) { 327 psignal(p->p_pptr, p->p_sigparent); 328 } else { 329 psignal(p->p_pptr, SIGCHLD); 330 } 331 332 wakeup((caddr_t)p->p_pptr); 333#if defined(tahoe) 334 /* move this to cpu_exit */ 335 p->p_addr->u_pcb.pcb_savacc.faddr = (float *)NULL; 336#endif 337 /* 338 * Clear curproc after we've done all operations 339 * that could block, and before tearing down the rest 340 * of the process state that might be used from clock, etc. 341 * Also, can't clear curproc while we're still runnable, 342 * as we're not on a run queue (we are current, just not 343 * a proper proc any longer!). 344 * 345 * Other substructures are freed from wait(). 346 */ 347 SET_CURPROC(NULL); 348 if (--p->p_limit->p_refcnt == 0) { 349 FREE(p->p_limit, M_SUBPROC); 350 p->p_limit = NULL; 351 } 352 353 /* 354 * Finally, call machine-dependent code to release the remaining 355 * resources including address space, the kernel stack and pcb. 356 * The address space is released by "vmspace_free(p->p_vmspace)"; 357 * This is machine-dependent, as we may have to change stacks 358 * or ensure that the current one isn't reallocated before we 359 * finish. cpu_exit will end with a call to cpu_switch(), finishing 360 * our execution (pun intended). 361 */ 362 cpu_exit(p); 363} 364 365#ifdef COMPAT_43 366int 367owait(p, uap) 368 struct proc *p; 369 register struct owait_args /* { 370 int dummy; 371 } */ *uap; 372{ 373 struct wait_args w; 374 375 w.options = 0; 376 w.rusage = NULL; 377 w.pid = WAIT_ANY; 378 w.status = NULL; 379 return (wait1(p, &w, 1)); 380} 381#endif /* COMPAT_43 */ 382 383int 384wait4(p, uap) 385 struct proc *p; 386 struct wait_args *uap; 387{ 388 389 return (wait1(p, uap, 0)); 390} 391 392static int 393wait1(q, uap, compat) 394 register struct proc *q; 395 register struct wait_args /* { 396 int pid; 397 int *status; 398 int options; 399 struct rusage *rusage; 400 } */ *uap; 401 int compat; 402{ 403 register int nfound; 404 register struct proc *p, *t; 405 int status, error; 406 407 if (uap->pid == 0) 408 uap->pid = -q->p_pgid; 409 if (uap->options &~ (WUNTRACED|WNOHANG|WLINUXCLONE)) 410 return (EINVAL); 411loop: 412 nfound = 0; 413 LIST_FOREACH(p, &q->p_children, p_sibling) { 414 if (uap->pid != WAIT_ANY && 415 p->p_pid != uap->pid && p->p_pgid != -uap->pid) 416 continue; 417 418 /* This special case handles a kthread spawned by linux_clone 419 * (see linux_misc.c). The linux_wait4 and linux_waitpid functions 420 * need to be able to distinguish between waiting on a process and 421 * waiting on a thread. It is a thread if p_sigparent is not SIGCHLD, 422 * and the WLINUXCLONE option signifies we want to wait for threads 423 * and not processes. 424 */ 425 if ((p->p_sigparent != SIGCHLD) ^ ((uap->options & WLINUXCLONE) != 0)) 426 continue; 427 428 nfound++; 429 if (p->p_stat == SZOMB) { 430 /* charge childs scheduling cpu usage to parent */ 431 if (curproc->p_pid != 1) { 432 curproc->p_estcpu = min(curproc->p_estcpu + 433 p->p_estcpu, UCHAR_MAX); 434 } 435 436 q->p_retval[0] = p->p_pid; 437#ifdef COMPAT_43 438 if (compat) 439 q->p_retval[1] = p->p_xstat; 440 else 441#endif 442 if (uap->status) { 443 status = p->p_xstat; /* convert to int */ 444 if ((error = copyout((caddr_t)&status, 445 (caddr_t)uap->status, sizeof(status)))) 446 return (error); 447 } 448 if (uap->rusage && (error = copyout((caddr_t)p->p_ru, 449 (caddr_t)uap->rusage, sizeof (struct rusage)))) 450 return (error); 451 /* 452 * If we got the child via a ptrace 'attach', 453 * we need to give it back to the old parent. 454 */ 455 if (p->p_oppid && (t = pfind(p->p_oppid))) { 456 p->p_oppid = 0; 457 proc_reparent(p, t); 458 psignal(t, SIGCHLD); 459 wakeup((caddr_t)t); 460 return (0); 461 } 462 p->p_xstat = 0; 463 ruadd(&q->p_stats->p_cru, p->p_ru); 464 FREE(p->p_ru, M_ZOMBIE); 465 p->p_ru = NULL; 466 467 /* 468 * Decrement the count of procs running with this uid. 469 */ 470 (void)chgproccnt(p->p_cred->p_ruid, -1); 471 472 /* 473 * Release reference to text vnode 474 */ 475 if (p->p_textvp) 476 vrele(p->p_textvp); 477 478 /* 479 * Free up credentials. 480 */ 481 if (--p->p_cred->p_refcnt == 0) { 482 crfree(p->p_cred->pc_ucred); 483 FREE(p->p_cred, M_SUBPROC); 484 p->p_cred = NULL; 485 } 486 487 /* 488 * Destroy empty prisons 489 */ 490 if (p->p_prison && !--p->p_prison->pr_ref) { 491 if (p->p_prison->pr_linux != NULL) 492 FREE(p->p_prison->pr_linux, M_PRISON); 493 FREE(p->p_prison, M_PRISON); 494 } 495 496 /* 497 * Finally finished with old proc entry. 498 * Unlink it from its process group and free it. 499 */ 500 leavepgrp(p); 501 LIST_REMOVE(p, p_list); /* off zombproc */ 502 LIST_REMOVE(p, p_sibling); 503 504 if (--p->p_procsig->ps_refcnt == 0) { 505 if (p->p_sigacts != &p->p_addr->u_sigacts) 506 FREE(p->p_sigacts, M_SUBPROC); 507 FREE(p->p_procsig, M_SUBPROC); 508 p->p_procsig = NULL; 509 } 510 511 /* 512 * Give machine-dependent layer a chance 513 * to free anything that cpu_exit couldn't 514 * release while still running in process context. 515 */ 516 cpu_wait(p); 517 zfree(proc_zone, p); 518 nprocs--; 519 return (0); 520 } 521 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 522 (p->p_flag & P_TRACED || uap->options & WUNTRACED)) { 523 p->p_flag |= P_WAITED; 524 q->p_retval[0] = p->p_pid; 525#ifdef COMPAT_43 526 if (compat) { 527 q->p_retval[1] = W_STOPCODE(p->p_xstat); 528 error = 0; 529 } else 530#endif 531 if (uap->status) { 532 status = W_STOPCODE(p->p_xstat); 533 error = copyout((caddr_t)&status, 534 (caddr_t)uap->status, sizeof(status)); 535 } else 536 error = 0; 537 return (error); 538 } 539 } 540 if (nfound == 0) 541 return (ECHILD); 542 if (uap->options & WNOHANG) { 543 q->p_retval[0] = 0; 544 return (0); 545 } 546 if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0))) 547 return (error); 548 goto loop; 549} 550 551/* 552 * make process 'parent' the new parent of process 'child'. 553 */ 554void 555proc_reparent(child, parent) 556 register struct proc *child; 557 register struct proc *parent; 558{ 559 560 if (child->p_pptr == parent) 561 return; 562 563 LIST_REMOVE(child, p_sibling); 564 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 565 child->p_pptr = parent; 566} 567 568/* 569 * The next two functions are to handle adding/deleting items on the 570 * exit callout list 571 * 572 * at_exit(): 573 * Take the arguments given and put them onto the exit callout list, 574 * However first make sure that it's not already there. 575 * returns 0 on success. 576 */ 577int 578at_exit(function) 579 exitlist_fn function; 580{ 581 ele_p ep; 582 583 /* Be noisy if the programmer has lost track of things */ 584 if (rm_at_exit(function)) 585 printf("exit callout entry already present\n"); 586 ep = malloc(sizeof(*ep), M_TEMP, M_NOWAIT); 587 if (ep == NULL) 588 return (ENOMEM); 589 ep->next = exit_list; 590 ep->function = function; 591 exit_list = ep; 592 return (0); 593} 594/* 595 * Scan the exit callout list for the given items and remove them. 596 * Returns the number of items removed. 597 * Logically this can only be 0 or 1. 598 */ 599int 600rm_at_exit(function) 601 exitlist_fn function; 602{ 603 ele_p *epp, ep; 604 int count; 605 606 count = 0; 607 epp = &exit_list; 608 ep = *epp; 609 while (ep) { 610 if (ep->function == function) { 611 *epp = ep->next; 612 free(ep, M_TEMP); 613 count++; 614 } else { 615 epp = &ep->next; 616 } 617 ep = *epp; 618 } 619 return (count); 620} 621 622void check_sigacts (void) 623{ 624 struct proc *p = curproc; 625 struct sigacts *pss; 626 int s; 627 628 if (p->p_procsig->ps_refcnt == 1 && 629 p->p_sigacts != &p->p_addr->u_sigacts) { 630 pss = p->p_sigacts; 631 s = splhigh(); 632 p->p_addr->u_sigacts = *pss; 633 p->p_sigacts = &p->p_addr->u_sigacts; 634 splx(s); 635 FREE(pss, M_SUBPROC); 636 } 637} 638 639