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: src/sys/kern/kern_exit.c,v 1.92.2.11 2003/01/13 22:51:16 dillon Exp $ 40 * $DragonFly: src/sys/kern/kern_exit.c,v 1.27 2003/11/03 15:57:33 daver Exp $ 41 */ 42 43 #include "opt_compat.h" 44 #include "opt_ktrace.h" 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/sysproto.h> 49 #include <sys/kernel.h> 50 #include <sys/malloc.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 #include <sys/kern_syscall.h> 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 <vm/vm_extern.h> 74 #include <sys/user.h> 75 76 /* Required to be non-static for SysVR4 emulator */ 77 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 78 79 static MALLOC_DEFINE(M_ATEXIT, "atexit", "atexit callback"); 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 * SYS_EXIT_ARGS(int rval) 97 */ 98 void 99 sys_exit(struct sys_exit_args *uap) 100 { 101 exit1(W_EXITCODE(uap->rval, 0)); 102 /* NOTREACHED */ 103 } 104 105 /* 106 * Exit: deallocate address space and other resources, change proc state 107 * to zombie, and unlink proc from allproc and parent's lists. Save exit 108 * status and rusage for wait(). Check for child processes and orphan them. 109 */ 110 void 111 exit1(int rv) 112 { 113 struct proc *p = curproc; 114 struct proc *q, *nq; 115 struct vmspace *vm; 116 struct vnode *vtmp; 117 struct exitlist *ep; 118 119 if (p->p_pid == 1) { 120 printf("init died (signal %d, exit %d)\n", 121 WTERMSIG(rv), WEXITSTATUS(rv)); 122 panic("Going nowhere without my init!"); 123 } 124 125 aio_proc_rundown(p); 126 127 /* are we a task leader? */ 128 if(p == p->p_leader) { 129 struct kill_args killArgs; 130 killArgs.signum = SIGKILL; 131 q = p->p_peers; 132 while(q) { 133 killArgs.pid = q->p_pid; 134 /* 135 * The interface for kill is better 136 * than the internal signal 137 */ 138 kill(&killArgs); 139 nq = q; 140 q = q->p_peers; 141 } 142 while (p->p_peers) 143 tsleep((caddr_t)p, 0, "exit1", 0); 144 } 145 146 #ifdef PGINPROF 147 vmsizmon(); 148 #endif 149 STOPEVENT(p, S_EXIT, rv); 150 wakeup(&p->p_stype); /* Wakeup anyone in procfs' PIOCWAIT */ 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 TAILQ_FOREACH(ep, &exit_list, next) 158 (*ep->function)(p->p_thread); 159 160 if (p->p_flag & P_PROFIL) 161 stopprofclock(p); 162 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage), 163 M_ZOMBIE, M_WAITOK); 164 /* 165 * If parent is waiting for us to exit or exec, 166 * P_PPWAIT is set; we will wakeup the parent below. 167 */ 168 p->p_flag &= ~(P_TRACED | P_PPWAIT); 169 p->p_flag |= P_WEXIT; 170 SIGEMPTYSET(p->p_siglist); 171 if (timevalisset(&p->p_realtimer.it_value)) 172 untimeout(realitexpire, (caddr_t)p, p->p_ithandle); 173 174 /* 175 * Reset any sigio structures pointing to us as a result of 176 * F_SETOWN with our pid. 177 */ 178 funsetownlst(&p->p_sigiolst); 179 180 /* 181 * Close open files and release open-file table. 182 * This may block! 183 */ 184 fdfree(p); 185 186 if(p->p_leader->p_peers) { 187 q = p->p_leader; 188 while(q->p_peers != p) 189 q = q->p_peers; 190 q->p_peers = p->p_peers; 191 wakeup((caddr_t)p->p_leader); 192 } 193 194 /* 195 * XXX Shutdown SYSV semaphores 196 */ 197 semexit(p); 198 199 /* The next two chunks should probably be moved to vmspace_exit. */ 200 vm = p->p_vmspace; 201 /* 202 * Release user portion of address space. 203 * This releases references to vnodes, 204 * which could cause I/O if the file has been unlinked. 205 * Need to do this early enough that we can still sleep. 206 * Can't free the entire vmspace as the kernel stack 207 * may be mapped within that space also. 208 * 209 * Processes sharing the same vmspace may exit in one order, and 210 * get cleaned up by vmspace_exit() in a different order. The 211 * last exiting process to reach this point releases as much of 212 * the environment as it can, and the last process cleaned up 213 * by vmspace_exit() (which decrements exitingcnt) cleans up the 214 * remainder. 215 */ 216 ++vm->vm_exitingcnt; 217 if (--vm->vm_refcnt == 0) { 218 shmexit(vm); 219 pmap_remove_pages(vmspace_pmap(vm), VM_MIN_ADDRESS, 220 VM_MAXUSER_ADDRESS); 221 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS, 222 VM_MAXUSER_ADDRESS); 223 } 224 225 if (SESS_LEADER(p)) { 226 struct session *sp = p->p_session; 227 228 if (sp->s_ttyvp) { 229 /* 230 * Controlling process. 231 * Signal foreground pgrp, 232 * drain controlling terminal 233 * and revoke access to controlling terminal. 234 */ 235 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 236 if (sp->s_ttyp->t_pgrp) 237 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 238 (void) ttywait(sp->s_ttyp); 239 /* 240 * The tty could have been revoked 241 * if we blocked. 242 */ 243 if (sp->s_ttyvp) 244 VOP_REVOKE(sp->s_ttyvp, REVOKEALL); 245 } 246 if (sp->s_ttyvp) 247 vrele(sp->s_ttyvp); 248 sp->s_ttyvp = NULL; 249 /* 250 * s_ttyp is not zero'd; we use this to indicate 251 * that the session once had a controlling terminal. 252 * (for logging and informational purposes) 253 */ 254 } 255 sp->s_leader = NULL; 256 } 257 fixjobc(p, p->p_pgrp, 0); 258 (void)acct_process(p); 259 #ifdef KTRACE 260 /* 261 * release trace file 262 */ 263 p->p_traceflag = 0; /* don't trace the vrele() */ 264 if ((vtmp = p->p_tracep) != NULL) { 265 p->p_tracep = NULL; 266 vrele(vtmp); 267 } 268 #endif 269 /* 270 * Release reference to text vnode 271 */ 272 if ((vtmp = p->p_textvp) != NULL) { 273 p->p_textvp = NULL; 274 vrele(vtmp); 275 } 276 277 /* 278 * Once we set SZOMB the process can get reaped. The wait1 code 279 * will also wait for TDF_RUNNING to be cleared in the thread's flags, 280 * indicating that it has been completely switched out. 281 */ 282 283 /* 284 * Remove proc from allproc queue and pidhash chain. 285 * Place onto zombproc. Unlink from parent's child list. 286 */ 287 LIST_REMOVE(p, p_list); 288 LIST_INSERT_HEAD(&zombproc, p, p_list); 289 p->p_stat = SZOMB; 290 291 LIST_REMOVE(p, p_hash); 292 293 q = LIST_FIRST(&p->p_children); 294 if (q) /* only need this if any child is S_ZOMB */ 295 wakeup((caddr_t) initproc); 296 for (; q != 0; q = nq) { 297 nq = LIST_NEXT(q, p_sibling); 298 LIST_REMOVE(q, p_sibling); 299 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling); 300 q->p_pptr = initproc; 301 q->p_sigparent = SIGCHLD; 302 /* 303 * Traced processes are killed 304 * since their existence means someone is screwing up. 305 */ 306 if (q->p_flag & P_TRACED) { 307 q->p_flag &= ~P_TRACED; 308 psignal(q, SIGKILL); 309 } 310 } 311 312 /* 313 * Save exit status and final rusage info, adding in child rusage 314 * info and self times. 315 */ 316 p->p_xstat = rv; 317 *p->p_ru = p->p_stats->p_ru; 318 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 319 ruadd(p->p_ru, &p->p_stats->p_cru); 320 321 /* 322 * notify interested parties of our demise. 323 */ 324 KNOTE(&p->p_klist, NOTE_EXIT); 325 326 /* 327 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 328 * flag set, notify process 1 instead (and hope it will handle 329 * this situation). 330 */ 331 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) { 332 struct proc *pp = p->p_pptr; 333 proc_reparent(p, initproc); 334 /* 335 * If this was the last child of our parent, notify 336 * parent, so in case he was wait(2)ing, he will 337 * continue. 338 */ 339 if (LIST_EMPTY(&pp->p_children)) 340 wakeup((caddr_t)pp); 341 } 342 343 if (p->p_sigparent && p->p_pptr != initproc) { 344 psignal(p->p_pptr, p->p_sigparent); 345 } else { 346 psignal(p->p_pptr, SIGCHLD); 347 } 348 349 wakeup((caddr_t)p->p_pptr); 350 #if defined(tahoe) 351 /* move this to cpu_exit */ 352 p->p_thread->td_pcb->pcb_saveacc.faddr = (float *)NULL; 353 #endif 354 /* 355 * cpu_exit is responsible for clearing curproc, since 356 * it is heavily integrated with the thread/switching sequence. 357 * 358 * Other substructures are freed from wait(). 359 */ 360 if (--p->p_limit->p_refcnt == 0) { 361 FREE(p->p_limit, M_SUBPROC); 362 p->p_limit = NULL; 363 } 364 365 /* 366 * Release the P_CURPROC designation on the process so the userland 367 * scheduler can work in someone else. 368 */ 369 release_curproc(p); 370 371 /* 372 * Finally, call machine-dependent code to release the remaining 373 * resources including address space, the kernel stack and pcb. 374 * The address space is released by "vmspace_free(p->p_vmspace)"; 375 * This is machine-dependent, as we may have to change stacks 376 * or ensure that the current one isn't reallocated before we 377 * finish. cpu_exit will end with a call to cpu_switch(), finishing 378 * our execution (pun intended). 379 */ 380 cpu_proc_exit(); 381 } 382 383 int 384 wait4(struct wait_args *uap) 385 { 386 struct rusage rusage; 387 int error, status; 388 389 error = kern_wait(uap->pid, uap->status ? &status : NULL, 390 uap->options, uap->rusage ? &rusage : NULL, &uap->sysmsg_fds[0]); 391 392 if (error == 0 && uap->status) 393 error = copyout(&status, uap->status, sizeof(*uap->status)); 394 if (error == 0 && uap->rusage) 395 error = copyout(&rusage, uap->rusage, sizeof(*uap->rusage)); 396 return (error); 397 } 398 399 /* 400 * wait1() 401 * 402 * wait_args(int pid, int *status, int options, struct rusage *rusage) 403 */ 404 int 405 kern_wait(pid_t pid, int *status, int options, struct rusage *rusage, int *res) 406 { 407 struct thread *td = curthread; 408 struct proc *q = td->td_proc; 409 struct proc *p, *t; 410 int nfound, error; 411 412 if (pid == 0) 413 pid = -q->p_pgid; 414 if (options &~ (WUNTRACED|WNOHANG|WLINUXCLONE)) 415 return (EINVAL); 416 loop: 417 nfound = 0; 418 LIST_FOREACH(p, &q->p_children, p_sibling) { 419 if (pid != WAIT_ANY && 420 p->p_pid != pid && p->p_pgid != -pid) 421 continue; 422 423 /* This special case handles a kthread spawned by linux_clone 424 * (see linux_misc.c). The linux_wait4 and linux_waitpid functions 425 * need to be able to distinguish between waiting on a process and 426 * waiting on a thread. It is a thread if p_sigparent is not SIGCHLD, 427 * and the WLINUXCLONE option signifies we want to wait for threads 428 * and not processes. 429 */ 430 if ((p->p_sigparent != SIGCHLD) ^ ((options & WLINUXCLONE) != 0)) 431 continue; 432 433 nfound++; 434 if (p->p_stat == SZOMB) { 435 /* 436 * The process's thread may still be in the middle 437 * of switching away, we can't rip its stack out from 438 * under it until TDF_RUNNING clears! 439 * 440 * YYY no wakeup occurs so we depend on the timeout. 441 */ 442 if ((p->p_thread->td_flags & TDF_RUNNING) != 0) { 443 tsleep(p->p_thread, 0, "reap", 1); 444 goto loop; 445 } 446 447 /* 448 * Other kernel threads may be in the middle of 449 * accessing the proc. For example, kern/kern_proc.c 450 * could be blocked writing proc data to a sysctl. 451 * At the moment, if this occurs, we are not woken 452 * up and rely on a one-second retry. 453 */ 454 if (p->p_lock) { 455 while (p->p_lock) 456 tsleep(p, 0, "reap2", hz); 457 } 458 lwkt_wait_free(p->p_thread); 459 460 /* charge childs scheduling cpu usage to parent */ 461 if (curproc->p_pid != 1) { 462 curproc->p_estcpu = 463 ESTCPULIM(curproc->p_estcpu + p->p_estcpu); 464 } 465 466 /* Take care of our return values. */ 467 *res = p->p_pid; 468 if (status) 469 *status = p->p_xstat; 470 if (rusage) 471 *rusage = *p->p_ru; 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 chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 492 493 /* 494 * Free up credentials. 495 */ 496 crfree(p->p_ucred); 497 p->p_ucred = NULL; 498 499 /* 500 * Remove unused arguments 501 */ 502 if (p->p_args && --p->p_args->ar_ref == 0) 503 FREE(p->p_args, M_PARGS); 504 505 /* 506 * Finally finished with old proc entry. 507 * Unlink it from its process group and free it. 508 */ 509 leavepgrp(p); 510 LIST_REMOVE(p, p_list); /* off zombproc */ 511 LIST_REMOVE(p, p_sibling); 512 513 if (--p->p_procsig->ps_refcnt == 0) { 514 if (p->p_sigacts != &p->p_addr->u_sigacts) 515 FREE(p->p_sigacts, M_SUBPROC); 516 FREE(p->p_procsig, M_SUBPROC); 517 p->p_procsig = NULL; 518 } 519 520 vm_waitproc(p); 521 zfree(proc_zone, p); 522 nprocs--; 523 return (0); 524 } 525 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 526 (p->p_flag & P_TRACED || options & WUNTRACED)) { 527 p->p_flag |= P_WAITED; 528 529 *res = p->p_pid; 530 if (status) 531 *status = p->p_xstat; 532 /* Zero rusage so we get something consistent. */ 533 if (rusage) 534 bzero(rusage, sizeof(rusage)); 535 return (0); 536 } 537 } 538 if (nfound == 0) 539 return (ECHILD); 540 if (options & WNOHANG) { 541 *res = 0; 542 return (0); 543 } 544 error = tsleep((caddr_t)q, PCATCH, "wait", 0); 545 if (error) 546 return (error); 547 goto loop; 548 } 549 550 /* 551 * make process 'parent' the new parent of process 'child'. 552 */ 553 void 554 proc_reparent(child, parent) 555 struct proc *child; 556 struct proc *parent; 557 { 558 559 if (child->p_pptr == parent) 560 return; 561 562 LIST_REMOVE(child, p_sibling); 563 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 564 child->p_pptr = parent; 565 } 566 567 /* 568 * The next two functions are to handle adding/deleting items on the 569 * exit callout list 570 * 571 * at_exit(): 572 * Take the arguments given and put them onto the exit callout list, 573 * However first make sure that it's not already there. 574 * returns 0 on success. 575 */ 576 577 int 578 at_exit(function) 579 exitlist_fn function; 580 { 581 struct exitlist *ep; 582 583 #ifdef INVARIANTS 584 /* Be noisy if the programmer has lost track of things */ 585 if (rm_at_exit(function)) 586 printf("WARNING: exit callout entry (%p) already present\n", 587 function); 588 #endif 589 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT); 590 if (ep == NULL) 591 return (ENOMEM); 592 ep->function = function; 593 TAILQ_INSERT_TAIL(&exit_list, ep, next); 594 return (0); 595 } 596 597 /* 598 * Scan the exit callout list for the given item and remove it. 599 * Returns the number of items removed (0 or 1) 600 */ 601 int 602 rm_at_exit(function) 603 exitlist_fn function; 604 { 605 struct exitlist *ep; 606 607 TAILQ_FOREACH(ep, &exit_list, next) { 608 if (ep->function == function) { 609 TAILQ_REMOVE(&exit_list, ep, next); 610 free(ep, M_ATEXIT); 611 return(1); 612 } 613 } 614 return (0); 615 } 616 617 void check_sigacts (void) 618 { 619 struct proc *p = curproc; 620 struct sigacts *pss; 621 int s; 622 623 if (p->p_procsig->ps_refcnt == 1 && 624 p->p_sigacts != &p->p_addr->u_sigacts) { 625 pss = p->p_sigacts; 626 s = splhigh(); 627 p->p_addr->u_sigacts = *pss; 628 p->p_sigacts = &p->p_addr->u_sigacts; 629 splx(s); 630 FREE(pss, M_SUBPROC); 631 } 632 } 633 634