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 1.37 2004/06/23 16:45:23 dillon 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 #include <sys/upcall.h> 67 #include <sys/caps.h> 68 69 #include <vm/vm.h> 70 #include <vm/vm_param.h> 71 #include <sys/lock.h> 72 #include <vm/pmap.h> 73 #include <vm/vm_map.h> 74 #include <vm/vm_zone.h> 75 #include <vm/vm_extern.h> 76 #include <sys/user.h> 77 78 /* Required to be non-static for SysVR4 emulator */ 79 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 80 81 static MALLOC_DEFINE(M_ATEXIT, "atexit", "atexit callback"); 82 83 /* 84 * callout list for things to do at exit time 85 */ 86 struct exitlist { 87 exitlist_fn function; 88 TAILQ_ENTRY(exitlist) next; 89 }; 90 91 TAILQ_HEAD(exit_list_head, exitlist); 92 static struct exit_list_head exit_list = TAILQ_HEAD_INITIALIZER(exit_list); 93 94 /* 95 * exit -- 96 * Death of process. 97 * 98 * SYS_EXIT_ARGS(int rval) 99 */ 100 void 101 sys_exit(struct sys_exit_args *uap) 102 { 103 exit1(W_EXITCODE(uap->rval, 0)); 104 /* NOTREACHED */ 105 } 106 107 /* 108 * Exit: deallocate address space and other resources, change proc state 109 * to zombie, and unlink proc from allproc and parent's lists. Save exit 110 * status and rusage for wait(). Check for child processes and orphan them. 111 */ 112 void 113 exit1(int rv) 114 { 115 struct proc *p = curproc; 116 struct proc *q, *nq; 117 struct vmspace *vm; 118 struct vnode *vtmp; 119 struct exitlist *ep; 120 121 if (p->p_pid == 1) { 122 printf("init died (signal %d, exit %d)\n", 123 WTERMSIG(rv), WEXITSTATUS(rv)); 124 panic("Going nowhere without my init!"); 125 } 126 127 sysmsg_rundown(p, 1); 128 caps_exit(p->p_thread); 129 aio_proc_rundown(p); 130 131 /* are we a task leader? */ 132 if(p == p->p_leader) { 133 struct kill_args killArgs; 134 killArgs.signum = SIGKILL; 135 q = p->p_peers; 136 while(q) { 137 killArgs.pid = q->p_pid; 138 /* 139 * The interface for kill is better 140 * than the internal signal 141 */ 142 kill(&killArgs); 143 nq = q; 144 q = q->p_peers; 145 } 146 while (p->p_peers) 147 tsleep((caddr_t)p, 0, "exit1", 0); 148 } 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->p_thread); 163 164 if (p->p_flag & P_PROFIL) 165 stopprofclock(p); 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 p->p_flag &= ~(P_TRACED | P_PPWAIT); 173 p->p_flag |= P_WEXIT; 174 SIGEMPTYSET(p->p_siglist); 175 if (timevalisset(&p->p_realtimer.it_value)) 176 untimeout(realitexpire, (caddr_t)p, p->p_ithandle); 177 178 /* 179 * Reset any sigio structures pointing to us as a result of 180 * F_SETOWN with our pid. 181 */ 182 funsetownlst(&p->p_sigiolst); 183 184 /* 185 * Close open files and release open-file table. 186 * This may block! 187 */ 188 fdfree(p); 189 p->p_fd = NULL; 190 191 if(p->p_leader->p_peers) { 192 q = p->p_leader; 193 while(q->p_peers != p) 194 q = q->p_peers; 195 q->p_peers = p->p_peers; 196 wakeup((caddr_t)p->p_leader); 197 } 198 199 /* 200 * XXX Shutdown SYSV semaphores 201 */ 202 semexit(p); 203 204 KKASSERT(p->p_numposixlocks == 0); 205 206 /* The next two chunks should probably be moved to vmspace_exit. */ 207 vm = p->p_vmspace; 208 209 /* 210 * Release upcalls associated with this process 211 */ 212 if (vm->vm_upcalls) 213 upc_release(vm, p); 214 215 /* 216 * Release user portion of address space. 217 * This releases references to vnodes, 218 * which could cause I/O if the file has been unlinked. 219 * Need to do this early enough that we can still sleep. 220 * Can't free the entire vmspace as the kernel stack 221 * may be mapped within that space also. 222 * 223 * Processes sharing the same vmspace may exit in one order, and 224 * get cleaned up by vmspace_exit() in a different order. The 225 * last exiting process to reach this point releases as much of 226 * the environment as it can, and the last process cleaned up 227 * by vmspace_exit() (which decrements exitingcnt) cleans up the 228 * remainder. 229 */ 230 ++vm->vm_exitingcnt; 231 if (--vm->vm_refcnt == 0) { 232 shmexit(vm); 233 pmap_remove_pages(vmspace_pmap(vm), VM_MIN_ADDRESS, 234 VM_MAXUSER_ADDRESS); 235 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS, 236 VM_MAXUSER_ADDRESS); 237 } 238 239 if (SESS_LEADER(p)) { 240 struct session *sp = p->p_session; 241 struct vnode *vp; 242 243 if (sp->s_ttyvp) { 244 /* 245 * We are the controlling process. Signal the 246 * foreground process group, drain the controlling 247 * terminal, and revoke access to the controlling 248 * terminal. 249 * 250 * NOTE: while waiting for the process group to exit 251 * it is possible that one of the processes in the 252 * group will revoke the tty, so we have to recheck. 253 */ 254 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 255 if (sp->s_ttyp->t_pgrp) 256 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 257 (void) ttywait(sp->s_ttyp); 258 /* 259 * The tty could have been revoked 260 * if we blocked. 261 */ 262 if ((vp = sp->s_ttyvp) != NULL) { 263 sp->s_ttyvp = NULL; 264 VOP_REVOKE(vp, REVOKEALL); 265 vrele(vp); 266 } 267 } 268 if ((vp = sp->s_ttyvp) != NULL) { 269 sp->s_ttyvp = NULL; 270 vrele(vp); 271 } 272 /* 273 * s_ttyp is not zero'd; we use this to indicate 274 * that the session once had a controlling terminal. 275 * (for logging and informational purposes) 276 */ 277 } 278 sp->s_leader = NULL; 279 } 280 fixjobc(p, p->p_pgrp, 0); 281 (void)acct_process(p); 282 #ifdef KTRACE 283 /* 284 * release trace file 285 */ 286 p->p_traceflag = 0; /* don't trace the vrele() */ 287 if ((vtmp = p->p_tracep) != NULL) { 288 p->p_tracep = NULL; 289 vrele(vtmp); 290 } 291 #endif 292 /* 293 * Release reference to text vnode 294 */ 295 if ((vtmp = p->p_textvp) != NULL) { 296 p->p_textvp = NULL; 297 vrele(vtmp); 298 } 299 300 /* 301 * Once we set SZOMB the process can get reaped. The wait1 code 302 * will also wait for TDF_RUNNING to be cleared in the thread's flags, 303 * indicating that it has been completely switched out. 304 */ 305 306 /* 307 * Remove proc from allproc queue and pidhash chain. 308 * Place onto zombproc. Unlink from parent's child list. 309 */ 310 LIST_REMOVE(p, p_list); 311 LIST_INSERT_HEAD(&zombproc, p, p_list); 312 p->p_stat = SZOMB; 313 314 LIST_REMOVE(p, p_hash); 315 316 q = LIST_FIRST(&p->p_children); 317 if (q) /* only need this if any child is S_ZOMB */ 318 wakeup((caddr_t) initproc); 319 for (; q != 0; q = nq) { 320 nq = LIST_NEXT(q, p_sibling); 321 LIST_REMOVE(q, p_sibling); 322 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling); 323 q->p_pptr = initproc; 324 q->p_sigparent = SIGCHLD; 325 /* 326 * Traced processes are killed 327 * since their existence means someone is screwing up. 328 */ 329 if (q->p_flag & P_TRACED) { 330 q->p_flag &= ~P_TRACED; 331 psignal(q, SIGKILL); 332 } 333 } 334 335 /* 336 * Save exit status and final rusage info, adding in child rusage 337 * info and self times. 338 */ 339 p->p_xstat = rv; 340 *p->p_ru = p->p_stats->p_ru; 341 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 342 ruadd(p->p_ru, &p->p_stats->p_cru); 343 344 /* 345 * notify interested parties of our demise. 346 */ 347 KNOTE(&p->p_klist, NOTE_EXIT); 348 349 /* 350 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 351 * flag set, notify process 1 instead (and hope it will handle 352 * this situation). 353 */ 354 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) { 355 struct proc *pp = p->p_pptr; 356 proc_reparent(p, initproc); 357 /* 358 * If this was the last child of our parent, notify 359 * parent, so in case he was wait(2)ing, he will 360 * continue. 361 */ 362 if (LIST_EMPTY(&pp->p_children)) 363 wakeup((caddr_t)pp); 364 } 365 366 if (p->p_sigparent && p->p_pptr != initproc) { 367 psignal(p->p_pptr, p->p_sigparent); 368 } else { 369 psignal(p->p_pptr, SIGCHLD); 370 } 371 372 wakeup((caddr_t)p->p_pptr); 373 #if defined(tahoe) 374 /* move this to cpu_exit */ 375 p->p_thread->td_pcb->pcb_saveacc.faddr = (float *)NULL; 376 #endif 377 /* 378 * cpu_exit is responsible for clearing curproc, since 379 * it is heavily integrated with the thread/switching sequence. 380 * 381 * Other substructures are freed from wait(). 382 */ 383 if (--p->p_limit->p_refcnt == 0) { 384 FREE(p->p_limit, M_SUBPROC); 385 p->p_limit = NULL; 386 } 387 388 /* 389 * Release the current user process designation on the process so 390 * the userland scheduler can work in someone else. 391 */ 392 release_curproc(p); 393 394 /* 395 * Finally, call machine-dependent code to release the remaining 396 * resources including address space, the kernel stack and pcb. 397 * The address space is released by "vmspace_free(p->p_vmspace)"; 398 * This is machine-dependent, as we may have to change stacks 399 * or ensure that the current one isn't reallocated before we 400 * finish. cpu_exit will end with a call to cpu_switch(), finishing 401 * our execution (pun intended). 402 */ 403 cpu_proc_exit(); 404 } 405 406 int 407 wait4(struct wait_args *uap) 408 { 409 struct rusage rusage; 410 int error, status; 411 412 error = kern_wait(uap->pid, uap->status ? &status : NULL, 413 uap->options, uap->rusage ? &rusage : NULL, &uap->sysmsg_fds[0]); 414 415 if (error == 0 && uap->status) 416 error = copyout(&status, uap->status, sizeof(*uap->status)); 417 if (error == 0 && uap->rusage) 418 error = copyout(&rusage, uap->rusage, sizeof(*uap->rusage)); 419 return (error); 420 } 421 422 /* 423 * wait1() 424 * 425 * wait_args(int pid, int *status, int options, struct rusage *rusage) 426 */ 427 int 428 kern_wait(pid_t pid, int *status, int options, struct rusage *rusage, int *res) 429 { 430 struct thread *td = curthread; 431 struct proc *q = td->td_proc; 432 struct proc *p, *t; 433 int nfound, error; 434 435 if (pid == 0) 436 pid = -q->p_pgid; 437 if (options &~ (WUNTRACED|WNOHANG|WLINUXCLONE)) 438 return (EINVAL); 439 loop: 440 nfound = 0; 441 LIST_FOREACH(p, &q->p_children, p_sibling) { 442 if (pid != WAIT_ANY && 443 p->p_pid != pid && p->p_pgid != -pid) 444 continue; 445 446 /* This special case handles a kthread spawned by linux_clone 447 * (see linux_misc.c). The linux_wait4 and linux_waitpid functions 448 * need to be able to distinguish between waiting on a process and 449 * waiting on a thread. It is a thread if p_sigparent is not SIGCHLD, 450 * and the WLINUXCLONE option signifies we want to wait for threads 451 * and not processes. 452 */ 453 if ((p->p_sigparent != SIGCHLD) ^ ((options & WLINUXCLONE) != 0)) 454 continue; 455 456 nfound++; 457 if (p->p_stat == SZOMB) { 458 /* 459 * The process's thread may still be in the middle 460 * of switching away, we can't rip its stack out from 461 * under it until TDF_RUNNING clears! 462 * 463 * YYY no wakeup occurs so we depend on the timeout. 464 */ 465 if ((p->p_thread->td_flags & TDF_RUNNING) != 0) { 466 tsleep(p->p_thread, 0, "reap", 1); 467 goto loop; 468 } 469 470 /* 471 * Other kernel threads may be in the middle of 472 * accessing the proc. For example, kern/kern_proc.c 473 * could be blocked writing proc data to a sysctl. 474 * At the moment, if this occurs, we are not woken 475 * up and rely on a one-second retry. 476 */ 477 if (p->p_lock) { 478 while (p->p_lock) 479 tsleep(p, 0, "reap2", hz); 480 } 481 lwkt_wait_free(p->p_thread); 482 483 /* 484 * Charge the parent for the child's change in 485 * estimated cpu as of when the child exits to 486 * account for batch scripts, large make's, etc. 487 */ 488 if (q->p_pid != 1) { 489 if (p->p_estcpu > p->p_estcpu_fork) { 490 q->p_estcpu = ESTCPULIM(q->p_estcpu + 491 p->p_estcpu - p->p_estcpu_fork); 492 } 493 } 494 495 /* Take care of our return values. */ 496 *res = p->p_pid; 497 if (status) 498 *status = p->p_xstat; 499 if (rusage) 500 *rusage = *p->p_ru; 501 /* 502 * If we got the child via a ptrace 'attach', 503 * we need to give it back to the old parent. 504 */ 505 if (p->p_oppid && (t = pfind(p->p_oppid))) { 506 p->p_oppid = 0; 507 proc_reparent(p, t); 508 psignal(t, SIGCHLD); 509 wakeup((caddr_t)t); 510 return (0); 511 } 512 p->p_xstat = 0; 513 ruadd(&q->p_stats->p_cru, p->p_ru); 514 FREE(p->p_ru, M_ZOMBIE); 515 p->p_ru = NULL; 516 517 /* 518 * Decrement the count of procs running with this uid. 519 */ 520 chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 521 522 /* 523 * Free up credentials. 524 */ 525 crfree(p->p_ucred); 526 p->p_ucred = NULL; 527 528 /* 529 * Remove unused arguments 530 */ 531 if (p->p_args && --p->p_args->ar_ref == 0) 532 FREE(p->p_args, M_PARGS); 533 534 /* 535 * Finally finished with old proc entry. 536 * Unlink it from its process group and free it. 537 */ 538 leavepgrp(p); 539 LIST_REMOVE(p, p_list); /* off zombproc */ 540 LIST_REMOVE(p, p_sibling); 541 542 if (--p->p_procsig->ps_refcnt == 0) { 543 if (p->p_sigacts != &p->p_addr->u_sigacts) 544 FREE(p->p_sigacts, M_SUBPROC); 545 FREE(p->p_procsig, M_SUBPROC); 546 p->p_procsig = NULL; 547 } 548 549 vm_waitproc(p); 550 zfree(proc_zone, p); 551 nprocs--; 552 return (0); 553 } 554 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 555 (p->p_flag & P_TRACED || options & WUNTRACED)) { 556 p->p_flag |= P_WAITED; 557 558 *res = p->p_pid; 559 if (status) 560 *status = W_STOPCODE(p->p_xstat); 561 /* Zero rusage so we get something consistent. */ 562 if (rusage) 563 bzero(rusage, sizeof(rusage)); 564 return (0); 565 } 566 } 567 if (nfound == 0) 568 return (ECHILD); 569 if (options & WNOHANG) { 570 *res = 0; 571 return (0); 572 } 573 error = tsleep((caddr_t)q, PCATCH, "wait", 0); 574 if (error) 575 return (error); 576 goto loop; 577 } 578 579 /* 580 * make process 'parent' the new parent of process 'child'. 581 */ 582 void 583 proc_reparent(child, parent) 584 struct proc *child; 585 struct proc *parent; 586 { 587 588 if (child->p_pptr == parent) 589 return; 590 591 LIST_REMOVE(child, p_sibling); 592 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 593 child->p_pptr = parent; 594 } 595 596 /* 597 * The next two functions are to handle adding/deleting items on the 598 * exit callout list 599 * 600 * at_exit(): 601 * Take the arguments given and put them onto the exit callout list, 602 * However first make sure that it's not already there. 603 * returns 0 on success. 604 */ 605 606 int 607 at_exit(function) 608 exitlist_fn function; 609 { 610 struct exitlist *ep; 611 612 #ifdef INVARIANTS 613 /* Be noisy if the programmer has lost track of things */ 614 if (rm_at_exit(function)) 615 printf("WARNING: exit callout entry (%p) already present\n", 616 function); 617 #endif 618 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT); 619 if (ep == NULL) 620 return (ENOMEM); 621 ep->function = function; 622 TAILQ_INSERT_TAIL(&exit_list, ep, next); 623 return (0); 624 } 625 626 /* 627 * Scan the exit callout list for the given item and remove it. 628 * Returns the number of items removed (0 or 1) 629 */ 630 int 631 rm_at_exit(function) 632 exitlist_fn function; 633 { 634 struct exitlist *ep; 635 636 TAILQ_FOREACH(ep, &exit_list, next) { 637 if (ep->function == function) { 638 TAILQ_REMOVE(&exit_list, ep, next); 639 free(ep, M_ATEXIT); 640 return(1); 641 } 642 } 643 return (0); 644 } 645 646 void check_sigacts (void) 647 { 648 struct proc *p = curproc; 649 struct sigacts *pss; 650 int s; 651 652 if (p->p_procsig->ps_refcnt == 1 && 653 p->p_sigacts != &p->p_addr->u_sigacts) { 654 pss = p->p_sigacts; 655 s = splhigh(); 656 p->p_addr->u_sigacts = *pss; 657 p->p_sigacts = &p->p_addr->u_sigacts; 658 splx(s); 659 FREE(pss, M_SUBPROC); 660 } 661 } 662 663