kern_exit.c revision 72200
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 72200 2001-02-09 06:11:45Z bmilekic $ 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/mutex.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 <sys/user.h> 73 74/* Required to be non-static for SysVR4 emulator */ 75MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 76 77static MALLOC_DEFINE(M_ATEXIT, "atexit", "atexit callback"); 78 79static int wait1 __P((struct proc *, struct wait_args *, int)); 80 81/* 82 * callout list for things to do at exit time 83 */ 84struct exitlist { 85 exitlist_fn function; 86 TAILQ_ENTRY(exitlist) next; 87}; 88 89TAILQ_HEAD(exit_list_head, exitlist); 90static struct exit_list_head exit_list = TAILQ_HEAD_INITIALIZER(exit_list); 91 92/* 93 * exit -- 94 * Death of process. 95 */ 96void 97sys_exit(p, uap) 98 struct proc *p; 99 struct sys_exit_args /* { 100 int rval; 101 } */ *uap; 102{ 103 104 exit1(p, W_EXITCODE(uap->rval, 0)); 105 /* NOTREACHED */ 106} 107 108/* 109 * Exit: deallocate address space and other resources, change proc state 110 * to zombie, and unlink proc from allproc and parent's lists. Save exit 111 * status and rusage for wait(). Check for child processes and orphan them. 112 */ 113void 114exit1(p, rv) 115 register struct proc *p; 116 int rv; 117{ 118 register struct proc *q, *nq; 119 register struct vmspace *vm; 120 struct exitlist *ep; 121 122 if (p->p_pid == 1) { 123 printf("init died (signal %d, exit %d)\n", 124 WTERMSIG(rv), WEXITSTATUS(rv)); 125 panic("Going nowhere without my init!"); 126 } 127 128 aio_proc_rundown(p); 129 130 /* are we a task leader? */ 131 PROC_LOCK(p); 132 if(p == p->p_leader) { 133 struct kill_args killArgs; 134 135 killArgs.signum = SIGKILL; 136 q = p->p_peers; 137 while(q) { 138 killArgs.pid = q->p_pid; 139 /* 140 * The interface for kill is better 141 * than the internal signal 142 */ 143 PROC_UNLOCK(p); 144 kill(p, &killArgs); 145 PROC_LOCK(p); 146 nq = q; 147 q = q->p_peers; 148 } 149 while (p->p_peers) 150 msleep((caddr_t)p, &p->p_mtx, PWAIT, "exit1", 0); 151 } 152 PROC_UNLOCK(p); 153 154#ifdef PGINPROF 155 vmsizmon(); 156#endif 157 STOPEVENT(p, S_EXIT, rv); 158 wakeup(&p->p_stype); /* Wakeup anyone in procfs' PIOCWAIT */ 159 160 /* 161 * Check if any loadable modules need anything done at process exit. 162 * e.g. SYSV IPC stuff 163 * XXX what if one of these generates an error? 164 */ 165 TAILQ_FOREACH(ep, &exit_list, next) 166 (*ep->function)(p); 167 168 stopprofclock(p); 169 170 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage), 171 M_ZOMBIE, M_WAITOK); 172 /* 173 * If parent is waiting for us to exit or exec, 174 * P_PPWAIT is set; we will wakeup the parent below. 175 */ 176 PROC_LOCK(p); 177 p->p_flag &= ~(P_TRACED | P_PPWAIT); 178 p->p_flag |= P_WEXIT; 179 SIGEMPTYSET(p->p_siglist); 180 PROC_UNLOCK(p); 181 if (timevalisset(&p->p_realtimer.it_value)) 182 callout_stop(&p->p_itcallout); 183 184 /* 185 * Reset any sigio structures pointing to us as a result of 186 * F_SETOWN with our pid. 187 */ 188 funsetownlst(&p->p_sigiolst); 189 190 /* 191 * Close open files and release open-file table. 192 * This may block! 193 */ 194 fdfree(p); 195 196 /* 197 * Remove ourself from our leader's peer list and wake our leader. 198 */ 199 PROC_LOCK(p); 200 if(p->p_leader->p_peers) { 201 q = p->p_leader; 202 while(q->p_peers != p) 203 q = q->p_peers; 204 q->p_peers = p->p_peers; 205 wakeup((caddr_t)p->p_leader); 206 } 207 PROC_UNLOCK(p); 208 209 /* 210 * XXX Shutdown SYSV semaphores 211 */ 212 semexit(p); 213 214 /* The next two chunks should probably be moved to vmspace_exit. */ 215 vm = p->p_vmspace; 216 /* 217 * Release user portion of address space. 218 * This releases references to vnodes, 219 * which could cause I/O if the file has been unlinked. 220 * Need to do this early enough that we can still sleep. 221 * Can't free the entire vmspace as the kernel stack 222 * may be mapped within that space also. 223 */ 224 if (vm->vm_refcnt == 1) { 225 if (vm->vm_shm) 226 shmexit(p); 227 pmap_remove_pages(vmspace_pmap(vm), VM_MIN_ADDRESS, 228 VM_MAXUSER_ADDRESS); 229 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS, 230 VM_MAXUSER_ADDRESS); 231 } 232 233 PROC_LOCK(p); 234 if (SESS_LEADER(p)) { 235 register struct session *sp = p->p_session; 236 237 PROC_UNLOCK(p); 238 if (sp->s_ttyvp) { 239 /* 240 * Controlling process. 241 * Signal foreground pgrp, 242 * drain controlling terminal 243 * and revoke access to controlling terminal. 244 */ 245 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 246 if (sp->s_ttyp->t_pgrp) 247 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 248 (void) ttywait(sp->s_ttyp); 249 /* 250 * The tty could have been revoked 251 * if we blocked. 252 */ 253 if (sp->s_ttyvp) 254 VOP_REVOKE(sp->s_ttyvp, REVOKEALL); 255 } 256 if (sp->s_ttyvp) 257 vrele(sp->s_ttyvp); 258 sp->s_ttyvp = NULL; 259 /* 260 * s_ttyp is not zero'd; we use this to indicate 261 * that the session once had a controlling terminal. 262 * (for logging and informational purposes) 263 */ 264 } 265 sp->s_leader = NULL; 266 } else 267 PROC_UNLOCK(p); 268 fixjobc(p, p->p_pgrp, 0); 269 (void)acct_process(p); 270#ifdef KTRACE 271 /* 272 * release trace file 273 */ 274 p->p_traceflag = 0; /* don't trace the vrele() */ 275 if (p->p_tracep) 276 vrele(p->p_tracep); 277#endif 278 /* 279 * Remove proc from allproc queue and pidhash chain. 280 * Place onto zombproc. Unlink from parent's child list. 281 */ 282 ALLPROC_LOCK(AP_EXCLUSIVE); 283 LIST_REMOVE(p, p_list); 284 LIST_INSERT_HEAD(&zombproc, p, p_list); 285 LIST_REMOVE(p, p_hash); 286 ALLPROC_LOCK(AP_RELEASE); 287 288 PROCTREE_LOCK(PT_EXCLUSIVE); 289 q = LIST_FIRST(&p->p_children); 290 if (q != NULL) /* only need this if any child is S_ZOMB */ 291 wakeup((caddr_t) initproc); 292 for (; q != NULL; q = nq) { 293 nq = LIST_NEXT(q, p_sibling); 294 LIST_REMOVE(q, p_sibling); 295 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling); 296 q->p_pptr = initproc; 297 PROC_LOCK(q); 298 q->p_sigparent = SIGCHLD; 299 /* 300 * Traced processes are killed 301 * since their existence means someone is screwing up. 302 */ 303 if (q->p_flag & P_TRACED) { 304 q->p_flag &= ~P_TRACED; 305 PROC_UNLOCK(q); 306 psignal(q, SIGKILL); 307 } else 308 PROC_UNLOCK(q); 309 } 310 311 /* 312 * Save exit status and final rusage info, adding in child rusage 313 * info and self times. 314 */ 315 p->p_xstat = rv; 316 *p->p_ru = p->p_stats->p_ru; 317 mtx_lock_spin(&sched_lock); 318 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 319 mtx_unlock_spin(&sched_lock); 320 ruadd(p->p_ru, &p->p_stats->p_cru); 321 322 /* 323 * Pretend that an mi_switch() to the next process occurs now. We 324 * must set `switchtime' directly since we will call cpu_switch() 325 * directly. Set it now so that the rest of the exit time gets 326 * counted somewhere if possible. 327 */ 328 microuptime(PCPU_PTR(switchtime)); 329 PCPU_SET(switchticks, ticks); 330 331 /* 332 * notify interested parties of our demise. 333 */ 334 PROC_LOCK(p); 335 KNOTE(&p->p_klist, NOTE_EXIT); 336 PROC_UNLOCK(p); 337 338 /* 339 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 340 * flag set, notify process 1 instead (and hope it will handle 341 * this situation). 342 */ 343 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) { 344 struct proc *pp = p->p_pptr; 345 proc_reparent(p, initproc); 346 /* 347 * If this was the last child of our parent, notify 348 * parent, so in case he was wait(2)ing, he will 349 * continue. 350 */ 351 if (LIST_EMPTY(&pp->p_children)) 352 wakeup((caddr_t)pp); 353 } 354 355 if (p->p_sigparent && p->p_pptr != initproc) { 356 psignal(p->p_pptr, p->p_sigparent); 357 } else { 358 psignal(p->p_pptr, SIGCHLD); 359 } 360 361 PROCTREE_LOCK(PT_RELEASE); 362 /* 363 * Clear curproc after we've done all operations 364 * that could block, and before tearing down the rest 365 * of the process state that might be used from clock, etc. 366 * Also, can't clear curproc while we're still runnable, 367 * as we're not on a run queue (we are current, just not 368 * a proper proc any longer!). 369 * 370 * Other substructures are freed from wait(). 371 */ 372 mtx_assert(&Giant, MA_OWNED); 373 if (--p->p_limit->p_refcnt == 0) { 374 FREE(p->p_limit, M_SUBPROC); 375 p->p_limit = NULL; 376 } 377 378 /* 379 * Finally, call machine-dependent code to release the remaining 380 * resources including address space, the kernel stack and pcb. 381 * The address space is released by "vmspace_free(p->p_vmspace)"; 382 * This is machine-dependent, as we may have to change stacks 383 * or ensure that the current one isn't reallocated before we 384 * finish. cpu_exit will end with a call to cpu_switch(), finishing 385 * our execution (pun intended). 386 */ 387 cpu_exit(p); 388} 389 390#ifdef COMPAT_43 391int 392owait(p, uap) 393 struct proc *p; 394 register struct owait_args /* { 395 int dummy; 396 } */ *uap; 397{ 398 struct wait_args w; 399 400 w.options = 0; 401 w.rusage = NULL; 402 w.pid = WAIT_ANY; 403 w.status = NULL; 404 return (wait1(p, &w, 1)); 405} 406#endif /* COMPAT_43 */ 407 408int 409wait4(p, uap) 410 struct proc *p; 411 struct wait_args *uap; 412{ 413 414 return (wait1(p, uap, 0)); 415} 416 417static int 418wait1(q, uap, compat) 419 register struct proc *q; 420 register struct wait_args /* { 421 int pid; 422 int *status; 423 int options; 424 struct rusage *rusage; 425 } */ *uap; 426 int compat; 427{ 428 register int nfound; 429 register struct proc *p, *t; 430 int status, error; 431 432 if (uap->pid == 0) 433 uap->pid = -q->p_pgid; 434 if (uap->options &~ (WUNTRACED|WNOHANG|WLINUXCLONE)) 435 return (EINVAL); 436loop: 437 nfound = 0; 438 PROCTREE_LOCK(PT_SHARED); 439 LIST_FOREACH(p, &q->p_children, p_sibling) { 440 if (uap->pid != WAIT_ANY && 441 p->p_pid != uap->pid && p->p_pgid != -uap->pid) 442 continue; 443 444 /* 445 * This special case handles a kthread spawned by linux_clone 446 * (see linux_misc.c). The linux_wait4 and linux_waitpid 447 * functions need to be able to distinguish between waiting 448 * on a process and waiting on a thread. It is a thread if 449 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option 450 * signifies we want to wait for threads and not processes. 451 */ 452 PROC_LOCK(p); 453 if ((p->p_sigparent != SIGCHLD) ^ 454 ((uap->options & WLINUXCLONE) != 0)) { 455 PROC_UNLOCK(p); 456 continue; 457 } 458 459 nfound++; 460 mtx_lock_spin(&sched_lock); 461 if (p->p_stat == SZOMB) { 462 mtx_unlock_spin(&sched_lock); 463 PROC_UNLOCK(p); 464 PROCTREE_LOCK(PT_RELEASE); 465 466 /* charge childs scheduling cpu usage to parent */ 467 if (curproc->p_pid != 1) { 468 curproc->p_estcpu = 469 ESTCPULIM(curproc->p_estcpu + p->p_estcpu); 470 } 471 472 q->p_retval[0] = p->p_pid; 473#ifdef COMPAT_43 474 if (compat) 475 q->p_retval[1] = p->p_xstat; 476 else 477#endif 478 if (uap->status) { 479 status = p->p_xstat; /* convert to int */ 480 if ((error = copyout((caddr_t)&status, 481 (caddr_t)uap->status, sizeof(status)))) 482 return (error); 483 } 484 if (uap->rusage && (error = copyout((caddr_t)p->p_ru, 485 (caddr_t)uap->rusage, sizeof (struct rusage)))) 486 return (error); 487 /* 488 * If we got the child via a ptrace 'attach', 489 * we need to give it back to the old parent. 490 */ 491 if (p->p_oppid) { 492 PROCTREE_LOCK(PT_EXCLUSIVE); 493 if ((t = pfind(p->p_oppid)) != NULL) { 494 p->p_oppid = 0; 495 proc_reparent(p, t); 496 psignal(t, SIGCHLD); 497 wakeup((caddr_t)t); 498 PROCTREE_LOCK(PT_RELEASE); 499 return (0); 500 } 501 PROCTREE_LOCK(PT_RELEASE); 502 } 503 p->p_xstat = 0; 504 ruadd(&q->p_stats->p_cru, p->p_ru); 505 FREE(p->p_ru, M_ZOMBIE); 506 p->p_ru = NULL; 507 508 /* 509 * Decrement the count of procs running with this uid. 510 */ 511 (void)chgproccnt(p->p_cred->p_uidinfo, -1, 0); 512 513 /* 514 * Release reference to text vnode 515 */ 516 if (p->p_textvp) 517 vrele(p->p_textvp); 518 519 /* 520 * Free up credentials. 521 */ 522 PROC_LOCK(p); 523 if (--p->p_cred->p_refcnt == 0) { 524 crfree(p->p_ucred); 525 uifree(p->p_cred->p_uidinfo); 526 FREE(p->p_cred, M_SUBPROC); 527 p->p_cred = NULL; 528 } 529 530 /* 531 * Destroy empty prisons 532 */ 533 if (p->p_prison && !--p->p_prison->pr_ref) { 534 if (p->p_prison->pr_linux != NULL) 535 FREE(p->p_prison->pr_linux, M_PRISON); 536 FREE(p->p_prison, M_PRISON); 537 } 538 539 /* 540 * Remove unused arguments 541 */ 542 if (p->p_args && --p->p_args->ar_ref == 0) 543 FREE(p->p_args, M_PARGS); 544 PROC_UNLOCK(p); 545 546 /* 547 * Finally finished with old proc entry. 548 * Unlink it from its process group and free it. 549 */ 550 leavepgrp(p); 551 552 ALLPROC_LOCK(AP_EXCLUSIVE); 553 LIST_REMOVE(p, p_list); /* off zombproc */ 554 ALLPROC_LOCK(AP_RELEASE); 555 556 PROCTREE_LOCK(PT_EXCLUSIVE); 557 LIST_REMOVE(p, p_sibling); 558 PROCTREE_LOCK(PT_RELEASE); 559 560 PROC_LOCK(p); 561 if (--p->p_procsig->ps_refcnt == 0) { 562 if (p->p_sigacts != &p->p_addr->u_sigacts) 563 FREE(p->p_sigacts, M_SUBPROC); 564 FREE(p->p_procsig, M_SUBPROC); 565 p->p_procsig = NULL; 566 } 567 PROC_UNLOCK(p); 568 569 /* 570 * Give machine-dependent layer a chance 571 * to free anything that cpu_exit couldn't 572 * release while still running in process context. 573 */ 574 cpu_wait(p); 575 mtx_destroy(&p->p_mtx); 576 zfree(proc_zone, p); 577 nprocs--; 578 return (0); 579 } 580 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 581 (p->p_flag & P_TRACED || uap->options & WUNTRACED)) { 582 mtx_unlock_spin(&sched_lock); 583 p->p_flag |= P_WAITED; 584 PROC_UNLOCK(p); 585 PROCTREE_LOCK(PT_RELEASE); 586 q->p_retval[0] = p->p_pid; 587#ifdef COMPAT_43 588 if (compat) { 589 q->p_retval[1] = W_STOPCODE(p->p_xstat); 590 error = 0; 591 } else 592#endif 593 if (uap->status) { 594 status = W_STOPCODE(p->p_xstat); 595 error = copyout((caddr_t)&status, 596 (caddr_t)uap->status, sizeof(status)); 597 } else 598 error = 0; 599 return (error); 600 } 601 mtx_unlock_spin(&sched_lock); 602 PROC_UNLOCK(p); 603 } 604 PROCTREE_LOCK(PT_RELEASE); 605 if (nfound == 0) 606 return (ECHILD); 607 if (uap->options & WNOHANG) { 608 q->p_retval[0] = 0; 609 return (0); 610 } 611 if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0))) 612 return (error); 613 goto loop; 614} 615 616/* 617 * Make process 'parent' the new parent of process 'child'. 618 * Must be called with an exclusive hold of proctree lock. 619 */ 620void 621proc_reparent(child, parent) 622 register struct proc *child; 623 register struct proc *parent; 624{ 625 626 PROCTREE_ASSERT(PT_EXCLUSIVE); 627 if (child->p_pptr == parent) 628 return; 629 630 LIST_REMOVE(child, p_sibling); 631 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 632 child->p_pptr = parent; 633} 634 635/* 636 * The next two functions are to handle adding/deleting items on the 637 * exit callout list 638 * 639 * at_exit(): 640 * Take the arguments given and put them onto the exit callout list, 641 * However first make sure that it's not already there. 642 * returns 0 on success. 643 */ 644 645int 646at_exit(function) 647 exitlist_fn function; 648{ 649 struct exitlist *ep; 650 651#ifdef INVARIANTS 652 /* Be noisy if the programmer has lost track of things */ 653 if (rm_at_exit(function)) 654 printf("WARNING: exit callout entry (%p) already present\n", 655 function); 656#endif 657 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT); 658 if (ep == NULL) 659 return (ENOMEM); 660 ep->function = function; 661 TAILQ_INSERT_TAIL(&exit_list, ep, next); 662 return (0); 663} 664 665/* 666 * Scan the exit callout list for the given item and remove it. 667 * Returns the number of items removed (0 or 1) 668 */ 669int 670rm_at_exit(function) 671 exitlist_fn function; 672{ 673 struct exitlist *ep; 674 675 TAILQ_FOREACH(ep, &exit_list, next) { 676 if (ep->function == function) { 677 TAILQ_REMOVE(&exit_list, ep, next); 678 free(ep, M_ATEXIT); 679 return(1); 680 } 681 } 682 return (0); 683} 684 685void check_sigacts (void) 686{ 687 struct proc *p = curproc; 688 struct sigacts *pss; 689 int s; 690 691 PROC_LOCK(p); 692 if (p->p_procsig->ps_refcnt == 1 && 693 p->p_sigacts != &p->p_addr->u_sigacts) { 694 pss = p->p_sigacts; 695 s = splhigh(); 696 p->p_addr->u_sigacts = *pss; 697 p->p_sigacts = &p->p_addr->u_sigacts; 698 splx(s); 699 FREE(pss, M_SUBPROC); 700 } 701 PROC_UNLOCK(p); 702} 703 704