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/lock.h> 51 #include <sys/mutex.h> 52 #include <sys/proc.h> 53 #include <sys/pioctl.h> 54 #include <sys/tty.h> 55 #include <sys/wait.h> 56 #include <sys/vnode.h> 57 #include <sys/resourcevar.h> 58 #include <sys/signalvar.h> 59 #include <sys/sx.h> 60 #include <sys/ptrace.h> 61 #include <sys/acct.h> /* for acct_process() function prototype */ 62 #include <sys/filedesc.h> 63 #include <sys/shm.h> 64 #include <sys/sem.h> 65 #include <sys/aio.h> 66 #include <sys/jail.h> 67 68 #include <vm/vm.h> 69 #include <vm/vm_param.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 */ 76 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 77 78 static MALLOC_DEFINE(M_ATEXIT, "atexit", "atexit callback"); 79 80 static int wait1 __P((struct proc *, struct wait_args *, int)); 81 82 /* 83 * callout list for things to do at exit time 84 */ 85 struct exitlist { 86 exitlist_fn function; 87 TAILQ_ENTRY(exitlist) next; 88 }; 89 90 TAILQ_HEAD(exit_list_head, exitlist); 91 static struct exit_list_head exit_list = TAILQ_HEAD_INITIALIZER(exit_list); 92 93 /* 94 * exit -- 95 * Death of process. 96 * 97 * MPSAFE 98 */ 99 void 100 sys_exit(p, uap) 101 struct proc *p; 102 struct sys_exit_args /* { 103 int rval; 104 } */ *uap; 105 { 106 mtx_lock(&Giant); 107 exit1(p, W_EXITCODE(uap->rval, 0)); 108 /* NOTREACHED */ 109 } 110 111 /* 112 * Exit: deallocate address space and other resources, change proc state 113 * to zombie, and unlink proc from allproc and parent's lists. Save exit 114 * status and rusage for wait(). Check for child processes and orphan them. 115 */ 116 void 117 exit1(p, rv) 118 register struct proc *p; 119 int rv; 120 { 121 register struct proc *q, *nq; 122 register struct vmspace *vm; 123 struct exitlist *ep; 124 125 GIANT_REQUIRED; 126 127 if (p->p_pid == 1) { 128 printf("init died (signal %d, exit %d)\n", 129 WTERMSIG(rv), WEXITSTATUS(rv)); 130 panic("Going nowhere without my init!"); 131 } 132 133 aio_proc_rundown(p); 134 135 /* are we a task leader? */ 136 PROC_LOCK(p); 137 if(p == p->p_leader) { 138 q = p->p_peers; 139 while (q != NULL) { 140 PROC_LOCK(q); 141 psignal(q, SIGKILL); 142 PROC_UNLOCK(q); 143 q = q->p_peers; 144 } 145 while (p->p_peers) 146 msleep((caddr_t)p, &p->p_mtx, PWAIT, "exit1", 0); 147 } 148 PROC_UNLOCK(p); 149 150 #ifdef PGINPROF 151 vmsizmon(); 152 #endif 153 STOPEVENT(p, S_EXIT, rv); 154 wakeup(&p->p_stype); /* Wakeup anyone in procfs' PIOCWAIT */ 155 156 /* 157 * Check if any loadable modules need anything done at process exit. 158 * e.g. SYSV IPC stuff 159 * XXX what if one of these generates an error? 160 */ 161 TAILQ_FOREACH(ep, &exit_list, next) 162 (*ep->function)(p); 163 164 stopprofclock(p); 165 166 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage), 167 M_ZOMBIE, M_WAITOK); 168 /* 169 * If parent is waiting for us to exit or exec, 170 * P_PPWAIT is set; we will wakeup the parent below. 171 */ 172 PROC_LOCK(p); 173 p->p_flag &= ~(P_TRACED | P_PPWAIT); 174 p->p_flag |= P_WEXIT; 175 SIGEMPTYSET(p->p_siglist); 176 PROC_UNLOCK(p); 177 if (timevalisset(&p->p_realtimer.it_value)) 178 callout_stop(&p->p_itcallout); 179 180 /* 181 * Reset any sigio structures pointing to us as a result of 182 * F_SETOWN with our pid. 183 */ 184 funsetownlst(&p->p_sigiolst); 185 186 /* 187 * Close open files and release open-file table. 188 * This may block! 189 */ 190 fdfree(p); 191 192 /* 193 * Remove ourself from our leader's peer list and wake our leader. 194 */ 195 PROC_LOCK(p->p_leader); 196 if(p->p_leader->p_peers) { 197 q = p->p_leader; 198 while(q->p_peers != p) 199 q = q->p_peers; 200 q->p_peers = p->p_peers; 201 wakeup((caddr_t)p->p_leader); 202 } 203 PROC_UNLOCK(p->p_leader); 204 205 /* 206 * XXX Shutdown SYSV semaphores 207 */ 208 semexit(p); 209 210 /* The next two chunks should probably be moved to vmspace_exit. */ 211 vm = p->p_vmspace; 212 /* 213 * Release user portion of address space. 214 * This releases references to vnodes, 215 * which could cause I/O if the file has been unlinked. 216 * Need to do this early enough that we can still sleep. 217 * Can't free the entire vmspace as the kernel stack 218 * may be mapped within that space also. 219 */ 220 if (vm->vm_refcnt == 1) { 221 if (vm->vm_shm) 222 shmexit(p); 223 pmap_remove_pages(vmspace_pmap(vm), VM_MIN_ADDRESS, 224 VM_MAXUSER_ADDRESS); 225 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS, 226 VM_MAXUSER_ADDRESS); 227 } 228 229 PROC_LOCK(p); 230 if (SESS_LEADER(p)) { 231 register struct session *sp = p->p_session; 232 233 PROC_UNLOCK(p); 234 if (sp->s_ttyvp) { 235 /* 236 * Controlling process. 237 * Signal foreground pgrp, 238 * drain controlling terminal 239 * and revoke access to controlling terminal. 240 */ 241 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 242 if (sp->s_ttyp->t_pgrp) 243 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 244 (void) ttywait(sp->s_ttyp); 245 /* 246 * The tty could have been revoked 247 * if we blocked. 248 */ 249 if (sp->s_ttyvp) 250 VOP_REVOKE(sp->s_ttyvp, REVOKEALL); 251 } 252 if (sp->s_ttyvp) 253 vrele(sp->s_ttyvp); 254 sp->s_ttyvp = NULL; 255 /* 256 * s_ttyp is not zero'd; we use this to indicate 257 * that the session once had a controlling terminal. 258 * (for logging and informational purposes) 259 */ 260 } 261 sp->s_leader = NULL; 262 } else 263 PROC_UNLOCK(p); 264 fixjobc(p, p->p_pgrp, 0); 265 (void)acct_process(p); 266 #ifdef KTRACE 267 /* 268 * release trace file 269 */ 270 p->p_traceflag = 0; /* don't trace the vrele() */ 271 if (p->p_tracep) 272 vrele(p->p_tracep); 273 #endif 274 /* 275 * Remove proc from allproc queue and pidhash chain. 276 * Place onto zombproc. Unlink from parent's child list. 277 */ 278 sx_xlock(&allproc_lock); 279 LIST_REMOVE(p, p_list); 280 LIST_INSERT_HEAD(&zombproc, p, p_list); 281 LIST_REMOVE(p, p_hash); 282 sx_xunlock(&allproc_lock); 283 284 sx_xlock(&proctree_lock); 285 q = LIST_FIRST(&p->p_children); 286 if (q != NULL) /* only need this if any child is S_ZOMB */ 287 wakeup((caddr_t) initproc); 288 for (; q != NULL; q = nq) { 289 nq = LIST_NEXT(q, p_sibling); 290 PROC_LOCK(q); 291 proc_reparent(q, initproc); 292 q->p_sigparent = SIGCHLD; 293 /* 294 * Traced processes are killed 295 * since their existence means someone is screwing up. 296 */ 297 if (q->p_flag & P_TRACED) { 298 q->p_flag &= ~P_TRACED; 299 psignal(q, SIGKILL); 300 } 301 PROC_UNLOCK(q); 302 } 303 304 /* 305 * Save exit status and final rusage info, adding in child rusage 306 * info and self times. 307 */ 308 p->p_xstat = rv; 309 *p->p_ru = p->p_stats->p_ru; 310 mtx_lock_spin(&sched_lock); 311 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 312 mtx_unlock_spin(&sched_lock); 313 ruadd(p->p_ru, &p->p_stats->p_cru); 314 315 /* 316 * Pretend that an mi_switch() to the next process occurs now. We 317 * must set `switchtime' directly since we will call cpu_switch() 318 * directly. Set it now so that the rest of the exit time gets 319 * counted somewhere if possible. 320 */ 321 mtx_lock_spin(&sched_lock); 322 microuptime(PCPU_PTR(switchtime)); 323 PCPU_SET(switchticks, ticks); 324 mtx_unlock_spin(&sched_lock); 325 326 /* 327 * notify interested parties of our demise. 328 */ 329 PROC_LOCK(p); 330 KNOTE(&p->p_klist, NOTE_EXIT); 331 332 /* 333 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 334 * flag set, notify process 1 instead (and hope it will handle 335 * this situation). 336 */ 337 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) { 338 struct proc *pp = p->p_pptr; 339 proc_reparent(p, initproc); 340 /* 341 * If this was the last child of our parent, notify 342 * parent, so in case he was wait(2)ing, he will 343 * continue. 344 */ 345 if (LIST_EMPTY(&pp->p_children)) 346 wakeup((caddr_t)pp); 347 } 348 349 PROC_LOCK(p->p_pptr); 350 if (p->p_sigparent && p->p_pptr != initproc) 351 psignal(p->p_pptr, p->p_sigparent); 352 else 353 psignal(p->p_pptr, SIGCHLD); 354 PROC_UNLOCK(p->p_pptr); 355 356 /* 357 * If this is a kthread, then wakeup anyone waiting for it to exit. 358 */ 359 if (p->p_flag & P_KTHREAD) 360 wakeup((caddr_t)p); 361 PROC_UNLOCK(p); 362 sx_xunlock(&proctree_lock); 363 364 /* 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 /* 394 * MPSAFE, the dirty work is handled by wait1(). 395 */ 396 int 397 owait(p, uap) 398 struct proc *p; 399 register struct owait_args /* { 400 int dummy; 401 } */ *uap; 402 { 403 struct wait_args w; 404 405 w.options = 0; 406 w.rusage = NULL; 407 w.pid = WAIT_ANY; 408 w.status = NULL; 409 return (wait1(p, &w, 1)); 410 } 411 #endif /* COMPAT_43 */ 412 413 /* 414 * MPSAFE, the dirty work is handled by wait1(). 415 */ 416 int 417 wait4(p, uap) 418 struct proc *p; 419 struct wait_args *uap; 420 { 421 return (wait1(p, uap, 0)); 422 } 423 424 /* 425 * MPSAFE 426 */ 427 static int 428 wait1(q, uap, compat) 429 register struct proc *q; 430 register struct wait_args /* { 431 int pid; 432 int *status; 433 int options; 434 struct rusage *rusage; 435 } */ *uap; 436 int compat; 437 { 438 register int nfound; 439 register struct proc *p, *t; 440 int status, error; 441 442 mtx_lock(&Giant); 443 if (uap->pid == 0) 444 uap->pid = -q->p_pgid; 445 if (uap->options &~ (WUNTRACED|WNOHANG|WLINUXCLONE)) { 446 error = EINVAL; 447 goto done2; 448 } 449 loop: 450 nfound = 0; 451 sx_slock(&proctree_lock); 452 LIST_FOREACH(p, &q->p_children, p_sibling) { 453 if (uap->pid != WAIT_ANY && 454 p->p_pid != uap->pid && p->p_pgid != -uap->pid) 455 continue; 456 457 /* 458 * This special case handles a kthread spawned by linux_clone 459 * (see linux_misc.c). The linux_wait4 and linux_waitpid 460 * functions need to be able to distinguish between waiting 461 * on a process and waiting on a thread. It is a thread if 462 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option 463 * signifies we want to wait for threads and not processes. 464 */ 465 PROC_LOCK(p); 466 if ((p->p_sigparent != SIGCHLD) ^ 467 ((uap->options & WLINUXCLONE) != 0)) { 468 PROC_UNLOCK(p); 469 continue; 470 } 471 472 nfound++; 473 mtx_lock_spin(&sched_lock); 474 if (p->p_stat == SZOMB) { 475 /* charge childs scheduling cpu usage to parent */ 476 if (curproc->p_pid != 1) { 477 curproc->p_estcpu = 478 ESTCPULIM(curproc->p_estcpu + p->p_estcpu); 479 } 480 481 mtx_unlock_spin(&sched_lock); 482 PROC_UNLOCK(p); 483 sx_sunlock(&proctree_lock); 484 485 q->p_retval[0] = p->p_pid; 486 #ifdef COMPAT_43 487 if (compat) 488 q->p_retval[1] = p->p_xstat; 489 else 490 #endif 491 if (uap->status) { 492 status = p->p_xstat; /* convert to int */ 493 if ((error = copyout((caddr_t)&status, 494 (caddr_t)uap->status, sizeof(status)))) { 495 goto done2; 496 } 497 } 498 if (uap->rusage && (error = copyout((caddr_t)p->p_ru, 499 (caddr_t)uap->rusage, sizeof (struct rusage)))) { 500 goto done2; 501 } 502 /* 503 * If we got the child via a ptrace 'attach', 504 * we need to give it back to the old parent. 505 */ 506 sx_xlock(&proctree_lock); 507 if (p->p_oppid) { 508 if ((t = pfind(p->p_oppid)) != NULL) { 509 PROC_LOCK(p); 510 p->p_oppid = 0; 511 proc_reparent(p, t); 512 PROC_UNLOCK(p); 513 psignal(t, SIGCHLD); 514 wakeup((caddr_t)t); 515 PROC_UNLOCK(t); 516 sx_xunlock(&proctree_lock); 517 error = 0; 518 goto done2; 519 } 520 } 521 sx_xunlock(&proctree_lock); 522 PROC_LOCK(p); 523 p->p_xstat = 0; 524 PROC_UNLOCK(p); 525 ruadd(&q->p_stats->p_cru, p->p_ru); 526 FREE(p->p_ru, M_ZOMBIE); 527 p->p_ru = NULL; 528 529 /* 530 * Decrement the count of procs running with this uid. 531 */ 532 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 533 534 /* 535 * Release reference to text vnode 536 */ 537 if (p->p_textvp) 538 vrele(p->p_textvp); 539 540 /* 541 * Finally finished with old proc entry. 542 * Unlink it from its process group and free it. 543 */ 544 leavepgrp(p); 545 546 sx_xlock(&allproc_lock); 547 LIST_REMOVE(p, p_list); /* off zombproc */ 548 sx_xunlock(&allproc_lock); 549 550 sx_xlock(&proctree_lock); 551 LIST_REMOVE(p, p_sibling); 552 sx_xunlock(&proctree_lock); 553 554 /* 555 * Free up credentials. 556 */ 557 crfree(p->p_ucred); 558 p->p_ucred = NULL; 559 560 /* 561 * Remove unused arguments 562 */ 563 if (p->p_args && --p->p_args->ar_ref == 0) 564 FREE(p->p_args, M_PARGS); 565 566 if (--p->p_procsig->ps_refcnt == 0) { 567 if (p->p_sigacts != &p->p_addr->u_sigacts) 568 FREE(p->p_sigacts, M_SUBPROC); 569 FREE(p->p_procsig, M_SUBPROC); 570 p->p_procsig = NULL; 571 } 572 573 /* 574 * Give machine-dependent layer a chance 575 * to free anything that cpu_exit couldn't 576 * release while still running in process context. 577 */ 578 cpu_wait(p); 579 mtx_destroy(&p->p_mtx); 580 zfree(proc_zone, p); 581 nprocs--; 582 error = 0; 583 goto done2; 584 } 585 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 586 (p->p_flag & P_TRACED || uap->options & WUNTRACED)) { 587 mtx_unlock_spin(&sched_lock); 588 p->p_flag |= P_WAITED; 589 PROC_UNLOCK(p); 590 sx_sunlock(&proctree_lock); 591 q->p_retval[0] = p->p_pid; 592 #ifdef COMPAT_43 593 if (compat) { 594 q->p_retval[1] = W_STOPCODE(p->p_xstat); 595 error = 0; 596 } else 597 #endif 598 if (uap->status) { 599 status = W_STOPCODE(p->p_xstat); 600 error = copyout((caddr_t)&status, 601 (caddr_t)uap->status, sizeof(status)); 602 } else 603 error = 0; 604 goto done2; 605 } 606 mtx_unlock_spin(&sched_lock); 607 PROC_UNLOCK(p); 608 } 609 sx_sunlock(&proctree_lock); 610 if (nfound == 0) { 611 error = ECHILD; 612 goto done2; 613 } 614 if (uap->options & WNOHANG) { 615 q->p_retval[0] = 0; 616 error = 0; 617 goto done2; 618 } 619 if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0)) != 0) 620 goto done2; 621 goto loop; 622 done2: 623 mtx_unlock(&Giant); 624 return(error); 625 } 626 627 /* 628 * Make process 'parent' the new parent of process 'child'. 629 * Must be called with an exclusive hold of proctree lock. 630 */ 631 void 632 proc_reparent(child, parent) 633 register struct proc *child; 634 register struct proc *parent; 635 { 636 637 SX_ASSERT_XLOCKED(&proctree_lock); 638 PROC_LOCK_ASSERT(child, MA_OWNED); 639 if (child->p_pptr == parent) 640 return; 641 642 LIST_REMOVE(child, p_sibling); 643 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 644 child->p_pptr = parent; 645 } 646 647 /* 648 * The next two functions are to handle adding/deleting items on the 649 * exit callout list 650 * 651 * at_exit(): 652 * Take the arguments given and put them onto the exit callout list, 653 * However first make sure that it's not already there. 654 * returns 0 on success. 655 */ 656 657 int 658 at_exit(function) 659 exitlist_fn function; 660 { 661 struct exitlist *ep; 662 663 #ifdef INVARIANTS 664 /* Be noisy if the programmer has lost track of things */ 665 if (rm_at_exit(function)) 666 printf("WARNING: exit callout entry (%p) already present\n", 667 function); 668 #endif 669 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT); 670 if (ep == NULL) 671 return (ENOMEM); 672 ep->function = function; 673 TAILQ_INSERT_TAIL(&exit_list, ep, next); 674 return (0); 675 } 676 677 /* 678 * Scan the exit callout list for the given item and remove it. 679 * Returns the number of items removed (0 or 1) 680 */ 681 int 682 rm_at_exit(function) 683 exitlist_fn function; 684 { 685 struct exitlist *ep; 686 687 TAILQ_FOREACH(ep, &exit_list, next) { 688 if (ep->function == function) { 689 TAILQ_REMOVE(&exit_list, ep, next); 690 free(ep, M_ATEXIT); 691 return(1); 692 } 693 } 694 return (0); 695 } 696