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/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 */ 75 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 76 77 static MALLOC_DEFINE(M_ATEXIT, "atexit", "atexit callback"); 78 79 static int wait1 __P((struct proc *, struct wait_args *, int)); 80 81 /* 82 * callout list for things to do at exit time 83 */ 84 struct exitlist { 85 exitlist_fn function; 86 TAILQ_ENTRY(exitlist) next; 87 }; 88 89 TAILQ_HEAD(exit_list_head, exitlist); 90 static struct exit_list_head exit_list = TAILQ_HEAD_INITIALIZER(exit_list); 91 92 /* 93 * exit -- 94 * Death of process. 95 */ 96 void 97 sys_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 */ 113 void 114 exit1(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 mtx_lock_spin(&sched_lock); 329 microuptime(PCPU_PTR(switchtime)); 330 PCPU_SET(switchticks, ticks); 331 mtx_unlock_spin(&sched_lock); 332 333 /* 334 * notify interested parties of our demise. 335 */ 336 PROC_LOCK(p); 337 KNOTE(&p->p_klist, NOTE_EXIT); 338 PROC_UNLOCK(p); 339 340 /* 341 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 342 * flag set, notify process 1 instead (and hope it will handle 343 * this situation). 344 */ 345 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) { 346 struct proc *pp = p->p_pptr; 347 proc_reparent(p, initproc); 348 /* 349 * If this was the last child of our parent, notify 350 * parent, so in case he was wait(2)ing, he will 351 * continue. 352 */ 353 if (LIST_EMPTY(&pp->p_children)) 354 wakeup((caddr_t)pp); 355 } 356 357 if (p->p_sigparent && p->p_pptr != initproc) { 358 psignal(p->p_pptr, p->p_sigparent); 359 } else { 360 psignal(p->p_pptr, SIGCHLD); 361 } 362 363 PROCTREE_LOCK(PT_RELEASE); 364 /* 365 * Clear curproc after we've done all operations 366 * that could block, and before tearing down the rest 367 * of the process state that might be used from clock, etc. 368 * Also, can't clear curproc while we're still runnable, 369 * as we're not on a run queue (we are current, just not 370 * a proper proc any longer!). 371 * 372 * Other substructures are freed from wait(). 373 */ 374 mtx_assert(&Giant, MA_OWNED); 375 if (--p->p_limit->p_refcnt == 0) { 376 FREE(p->p_limit, M_SUBPROC); 377 p->p_limit = NULL; 378 } 379 380 /* 381 * Finally, call machine-dependent code to release the remaining 382 * resources including address space, the kernel stack and pcb. 383 * The address space is released by "vmspace_free(p->p_vmspace)"; 384 * This is machine-dependent, as we may have to change stacks 385 * or ensure that the current one isn't reallocated before we 386 * finish. cpu_exit will end with a call to cpu_switch(), finishing 387 * our execution (pun intended). 388 */ 389 cpu_exit(p); 390 } 391 392 #ifdef COMPAT_43 393 int 394 owait(p, uap) 395 struct proc *p; 396 register struct owait_args /* { 397 int dummy; 398 } */ *uap; 399 { 400 struct wait_args w; 401 402 w.options = 0; 403 w.rusage = NULL; 404 w.pid = WAIT_ANY; 405 w.status = NULL; 406 return (wait1(p, &w, 1)); 407 } 408 #endif /* COMPAT_43 */ 409 410 int 411 wait4(p, uap) 412 struct proc *p; 413 struct wait_args *uap; 414 { 415 416 return (wait1(p, uap, 0)); 417 } 418 419 static int 420 wait1(q, uap, compat) 421 register struct proc *q; 422 register struct wait_args /* { 423 int pid; 424 int *status; 425 int options; 426 struct rusage *rusage; 427 } */ *uap; 428 int compat; 429 { 430 register int nfound; 431 register struct proc *p, *t; 432 int status, error; 433 434 if (uap->pid == 0) 435 uap->pid = -q->p_pgid; 436 if (uap->options &~ (WUNTRACED|WNOHANG|WLINUXCLONE)) 437 return (EINVAL); 438 loop: 439 nfound = 0; 440 PROCTREE_LOCK(PT_SHARED); 441 LIST_FOREACH(p, &q->p_children, p_sibling) { 442 if (uap->pid != WAIT_ANY && 443 p->p_pid != uap->pid && p->p_pgid != -uap->pid) 444 continue; 445 446 /* 447 * This special case handles a kthread spawned by linux_clone 448 * (see linux_misc.c). The linux_wait4 and linux_waitpid 449 * functions need to be able to distinguish between waiting 450 * on a process and waiting on a thread. It is a thread if 451 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option 452 * signifies we want to wait for threads and not processes. 453 */ 454 PROC_LOCK(p); 455 if ((p->p_sigparent != SIGCHLD) ^ 456 ((uap->options & WLINUXCLONE) != 0)) { 457 PROC_UNLOCK(p); 458 continue; 459 } 460 461 nfound++; 462 mtx_lock_spin(&sched_lock); 463 if (p->p_stat == SZOMB) { 464 mtx_unlock_spin(&sched_lock); 465 PROC_UNLOCK(p); 466 PROCTREE_LOCK(PT_RELEASE); 467 468 /* charge childs scheduling cpu usage to parent */ 469 if (curproc->p_pid != 1) { 470 curproc->p_estcpu = 471 ESTCPULIM(curproc->p_estcpu + p->p_estcpu); 472 } 473 474 q->p_retval[0] = p->p_pid; 475 #ifdef COMPAT_43 476 if (compat) 477 q->p_retval[1] = p->p_xstat; 478 else 479 #endif 480 if (uap->status) { 481 status = p->p_xstat; /* convert to int */ 482 if ((error = copyout((caddr_t)&status, 483 (caddr_t)uap->status, sizeof(status)))) 484 return (error); 485 } 486 if (uap->rusage && (error = copyout((caddr_t)p->p_ru, 487 (caddr_t)uap->rusage, sizeof (struct rusage)))) 488 return (error); 489 /* 490 * If we got the child via a ptrace 'attach', 491 * we need to give it back to the old parent. 492 */ 493 if (p->p_oppid) { 494 PROCTREE_LOCK(PT_EXCLUSIVE); 495 if ((t = pfind(p->p_oppid)) != NULL) { 496 p->p_oppid = 0; 497 proc_reparent(p, t); 498 psignal(t, SIGCHLD); 499 wakeup((caddr_t)t); 500 PROCTREE_LOCK(PT_RELEASE); 501 return (0); 502 } 503 PROCTREE_LOCK(PT_RELEASE); 504 } 505 p->p_xstat = 0; 506 ruadd(&q->p_stats->p_cru, p->p_ru); 507 FREE(p->p_ru, M_ZOMBIE); 508 p->p_ru = NULL; 509 510 /* 511 * Decrement the count of procs running with this uid. 512 */ 513 (void)chgproccnt(p->p_cred->p_uidinfo, -1, 0); 514 515 /* 516 * Release reference to text vnode 517 */ 518 if (p->p_textvp) 519 vrele(p->p_textvp); 520 521 /* 522 * Free up credentials. 523 */ 524 PROC_LOCK(p); 525 if (--p->p_cred->p_refcnt == 0) { 526 crfree(p->p_ucred); 527 uifree(p->p_cred->p_uidinfo); 528 FREE(p->p_cred, M_SUBPROC); 529 p->p_cred = NULL; 530 } 531 532 /* 533 * Remove unused arguments 534 */ 535 if (p->p_args && --p->p_args->ar_ref == 0) 536 FREE(p->p_args, M_PARGS); 537 PROC_UNLOCK(p); 538 539 /* 540 * Finally finished with old proc entry. 541 * Unlink it from its process group and free it. 542 */ 543 leavepgrp(p); 544 545 ALLPROC_LOCK(AP_EXCLUSIVE); 546 LIST_REMOVE(p, p_list); /* off zombproc */ 547 ALLPROC_LOCK(AP_RELEASE); 548 549 PROCTREE_LOCK(PT_EXCLUSIVE); 550 LIST_REMOVE(p, p_sibling); 551 PROCTREE_LOCK(PT_RELEASE); 552 553 PROC_LOCK(p); 554 if (--p->p_procsig->ps_refcnt == 0) { 555 if (p->p_sigacts != &p->p_addr->u_sigacts) 556 FREE(p->p_sigacts, M_SUBPROC); 557 FREE(p->p_procsig, M_SUBPROC); 558 p->p_procsig = NULL; 559 } 560 PROC_UNLOCK(p); 561 562 /* 563 * Give machine-dependent layer a chance 564 * to free anything that cpu_exit couldn't 565 * release while still running in process context. 566 */ 567 cpu_wait(p); 568 mtx_destroy(&p->p_mtx); 569 zfree(proc_zone, p); 570 nprocs--; 571 return (0); 572 } 573 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 574 (p->p_flag & P_TRACED || uap->options & WUNTRACED)) { 575 mtx_unlock_spin(&sched_lock); 576 p->p_flag |= P_WAITED; 577 PROC_UNLOCK(p); 578 PROCTREE_LOCK(PT_RELEASE); 579 q->p_retval[0] = p->p_pid; 580 #ifdef COMPAT_43 581 if (compat) { 582 q->p_retval[1] = W_STOPCODE(p->p_xstat); 583 error = 0; 584 } else 585 #endif 586 if (uap->status) { 587 status = W_STOPCODE(p->p_xstat); 588 error = copyout((caddr_t)&status, 589 (caddr_t)uap->status, sizeof(status)); 590 } else 591 error = 0; 592 return (error); 593 } 594 mtx_unlock_spin(&sched_lock); 595 PROC_UNLOCK(p); 596 } 597 PROCTREE_LOCK(PT_RELEASE); 598 if (nfound == 0) 599 return (ECHILD); 600 if (uap->options & WNOHANG) { 601 q->p_retval[0] = 0; 602 return (0); 603 } 604 if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0))) 605 return (error); 606 goto loop; 607 } 608 609 /* 610 * Make process 'parent' the new parent of process 'child'. 611 * Must be called with an exclusive hold of proctree lock. 612 */ 613 void 614 proc_reparent(child, parent) 615 register struct proc *child; 616 register struct proc *parent; 617 { 618 619 PROCTREE_ASSERT(PT_EXCLUSIVE); 620 if (child->p_pptr == parent) 621 return; 622 623 LIST_REMOVE(child, p_sibling); 624 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 625 child->p_pptr = parent; 626 } 627 628 /* 629 * The next two functions are to handle adding/deleting items on the 630 * exit callout list 631 * 632 * at_exit(): 633 * Take the arguments given and put them onto the exit callout list, 634 * However first make sure that it's not already there. 635 * returns 0 on success. 636 */ 637 638 int 639 at_exit(function) 640 exitlist_fn function; 641 { 642 struct exitlist *ep; 643 644 #ifdef INVARIANTS 645 /* Be noisy if the programmer has lost track of things */ 646 if (rm_at_exit(function)) 647 printf("WARNING: exit callout entry (%p) already present\n", 648 function); 649 #endif 650 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT); 651 if (ep == NULL) 652 return (ENOMEM); 653 ep->function = function; 654 TAILQ_INSERT_TAIL(&exit_list, ep, next); 655 return (0); 656 } 657 658 /* 659 * Scan the exit callout list for the given item and remove it. 660 * Returns the number of items removed (0 or 1) 661 */ 662 int 663 rm_at_exit(function) 664 exitlist_fn function; 665 { 666 struct exitlist *ep; 667 668 TAILQ_FOREACH(ep, &exit_list, next) { 669 if (ep->function == function) { 670 TAILQ_REMOVE(&exit_list, ep, next); 671 free(ep, M_ATEXIT); 672 return(1); 673 } 674 } 675 return (0); 676 } 677 678 void check_sigacts (void) 679 { 680 struct proc *p = curproc; 681 struct sigacts *pss; 682 int s; 683 684 PROC_LOCK(p); 685 if (p->p_procsig->ps_refcnt == 1 && 686 p->p_sigacts != &p->p_addr->u_sigacts) { 687 pss = p->p_sigacts; 688 s = splhigh(); 689 p->p_addr->u_sigacts = *pss; 690 p->p_sigacts = &p->p_addr->u_sigacts; 691 splx(s); 692 FREE(pss, M_SUBPROC); 693 } 694 PROC_UNLOCK(p); 695 } 696 697