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 1.34 2004/05/13 17:40:15 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 caps_exit(p->p_thread); 128 129 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); 189 190 if(p->p_leader->p_peers) { 191 q = p->p_leader; 192 while(q->p_peers != p) 193 q = q->p_peers; 194 q->p_peers = p->p_peers; 195 wakeup((caddr_t)p->p_leader); 196 } 197 198 /* 199 * XXX Shutdown SYSV semaphores 200 */ 201 semexit(p); 202 203 KKASSERT(p->p_numposixlocks == 0); 204 205 /* The next two chunks should probably be moved to vmspace_exit. */ 206 vm = p->p_vmspace; 207 208 /* 209 * Release upcalls associated with this process 210 */ 211 if (vm->vm_upcalls) 212 upc_release(vm, p); 213 214 /* 215 * Release user portion of address space. 216 * This releases references to vnodes, 217 * which could cause I/O if the file has been unlinked. 218 * Need to do this early enough that we can still sleep. 219 * Can't free the entire vmspace as the kernel stack 220 * may be mapped within that space also. 221 * 222 * Processes sharing the same vmspace may exit in one order, and 223 * get cleaned up by vmspace_exit() in a different order. The 224 * last exiting process to reach this point releases as much of 225 * the environment as it can, and the last process cleaned up 226 * by vmspace_exit() (which decrements exitingcnt) cleans up the 227 * remainder. 228 */ 229 ++vm->vm_exitingcnt; 230 if (--vm->vm_refcnt == 0) { 231 shmexit(vm); 232 pmap_remove_pages(vmspace_pmap(vm), VM_MIN_ADDRESS, 233 VM_MAXUSER_ADDRESS); 234 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS, 235 VM_MAXUSER_ADDRESS); 236 } 237 238 if (SESS_LEADER(p)) { 239 struct session *sp = p->p_session; 240 241 if (sp->s_ttyvp) { 242 /* 243 * Controlling process. 244 * Signal foreground pgrp, 245 * drain controlling terminal 246 * and revoke access to controlling terminal. 247 */ 248 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 249 if (sp->s_ttyp->t_pgrp) 250 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 251 (void) ttywait(sp->s_ttyp); 252 /* 253 * The tty could have been revoked 254 * if we blocked. 255 */ 256 if (sp->s_ttyvp) 257 VOP_REVOKE(sp->s_ttyvp, REVOKEALL); 258 } 259 if (sp->s_ttyvp) 260 vrele(sp->s_ttyvp); 261 sp->s_ttyvp = NULL; 262 /* 263 * s_ttyp is not zero'd; we use this to indicate 264 * that the session once had a controlling terminal. 265 * (for logging and informational purposes) 266 */ 267 } 268 sp->s_leader = NULL; 269 } 270 fixjobc(p, p->p_pgrp, 0); 271 (void)acct_process(p); 272 #ifdef KTRACE 273 /* 274 * release trace file 275 */ 276 p->p_traceflag = 0; /* don't trace the vrele() */ 277 if ((vtmp = p->p_tracep) != NULL) { 278 p->p_tracep = NULL; 279 vrele(vtmp); 280 } 281 #endif 282 /* 283 * Release reference to text vnode 284 */ 285 if ((vtmp = p->p_textvp) != NULL) { 286 p->p_textvp = NULL; 287 vrele(vtmp); 288 } 289 290 /* 291 * Once we set SZOMB the process can get reaped. The wait1 code 292 * will also wait for TDF_RUNNING to be cleared in the thread's flags, 293 * indicating that it has been completely switched out. 294 */ 295 296 /* 297 * Remove proc from allproc queue and pidhash chain. 298 * Place onto zombproc. Unlink from parent's child list. 299 */ 300 LIST_REMOVE(p, p_list); 301 LIST_INSERT_HEAD(&zombproc, p, p_list); 302 p->p_stat = SZOMB; 303 304 LIST_REMOVE(p, p_hash); 305 306 q = LIST_FIRST(&p->p_children); 307 if (q) /* only need this if any child is S_ZOMB */ 308 wakeup((caddr_t) initproc); 309 for (; q != 0; q = nq) { 310 nq = LIST_NEXT(q, p_sibling); 311 LIST_REMOVE(q, p_sibling); 312 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling); 313 q->p_pptr = initproc; 314 q->p_sigparent = SIGCHLD; 315 /* 316 * Traced processes are killed 317 * since their existence means someone is screwing up. 318 */ 319 if (q->p_flag & P_TRACED) { 320 q->p_flag &= ~P_TRACED; 321 psignal(q, SIGKILL); 322 } 323 } 324 325 /* 326 * Save exit status and final rusage info, adding in child rusage 327 * info and self times. 328 */ 329 p->p_xstat = rv; 330 *p->p_ru = p->p_stats->p_ru; 331 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 332 ruadd(p->p_ru, &p->p_stats->p_cru); 333 334 /* 335 * notify interested parties of our demise. 336 */ 337 KNOTE(&p->p_klist, NOTE_EXIT); 338 339 /* 340 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 341 * flag set, notify process 1 instead (and hope it will handle 342 * this situation). 343 */ 344 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) { 345 struct proc *pp = p->p_pptr; 346 proc_reparent(p, initproc); 347 /* 348 * If this was the last child of our parent, notify 349 * parent, so in case he was wait(2)ing, he will 350 * continue. 351 */ 352 if (LIST_EMPTY(&pp->p_children)) 353 wakeup((caddr_t)pp); 354 } 355 356 if (p->p_sigparent && p->p_pptr != initproc) { 357 psignal(p->p_pptr, p->p_sigparent); 358 } else { 359 psignal(p->p_pptr, SIGCHLD); 360 } 361 362 wakeup((caddr_t)p->p_pptr); 363 #if defined(tahoe) 364 /* move this to cpu_exit */ 365 p->p_thread->td_pcb->pcb_saveacc.faddr = (float *)NULL; 366 #endif 367 /* 368 * cpu_exit is responsible for clearing curproc, since 369 * it is heavily integrated with the thread/switching sequence. 370 * 371 * Other substructures are freed from wait(). 372 */ 373 if (--p->p_limit->p_refcnt == 0) { 374 FREE(p->p_limit, M_SUBPROC); 375 p->p_limit = NULL; 376 } 377 378 /* 379 * Release the current user process designation on the process so 380 * the userland scheduler can work in someone else. 381 */ 382 release_curproc(p); 383 384 /* 385 * Finally, call machine-dependent code to release the remaining 386 * resources including address space, the kernel stack and pcb. 387 * The address space is released by "vmspace_free(p->p_vmspace)"; 388 * This is machine-dependent, as we may have to change stacks 389 * or ensure that the current one isn't reallocated before we 390 * finish. cpu_exit will end with a call to cpu_switch(), finishing 391 * our execution (pun intended). 392 */ 393 cpu_proc_exit(); 394 } 395 396 int 397 wait4(struct wait_args *uap) 398 { 399 struct rusage rusage; 400 int error, status; 401 402 error = kern_wait(uap->pid, uap->status ? &status : NULL, 403 uap->options, uap->rusage ? &rusage : NULL, &uap->sysmsg_fds[0]); 404 405 if (error == 0 && uap->status) 406 error = copyout(&status, uap->status, sizeof(*uap->status)); 407 if (error == 0 && uap->rusage) 408 error = copyout(&rusage, uap->rusage, sizeof(*uap->rusage)); 409 return (error); 410 } 411 412 /* 413 * wait1() 414 * 415 * wait_args(int pid, int *status, int options, struct rusage *rusage) 416 */ 417 int 418 kern_wait(pid_t pid, int *status, int options, struct rusage *rusage, int *res) 419 { 420 struct thread *td = curthread; 421 struct proc *q = td->td_proc; 422 struct proc *p, *t; 423 int nfound, error; 424 425 if (pid == 0) 426 pid = -q->p_pgid; 427 if (options &~ (WUNTRACED|WNOHANG|WLINUXCLONE)) 428 return (EINVAL); 429 loop: 430 nfound = 0; 431 LIST_FOREACH(p, &q->p_children, p_sibling) { 432 if (pid != WAIT_ANY && 433 p->p_pid != pid && p->p_pgid != -pid) 434 continue; 435 436 /* This special case handles a kthread spawned by linux_clone 437 * (see linux_misc.c). The linux_wait4 and linux_waitpid functions 438 * need to be able to distinguish between waiting on a process and 439 * waiting on a thread. It is a thread if p_sigparent is not SIGCHLD, 440 * and the WLINUXCLONE option signifies we want to wait for threads 441 * and not processes. 442 */ 443 if ((p->p_sigparent != SIGCHLD) ^ ((options & WLINUXCLONE) != 0)) 444 continue; 445 446 nfound++; 447 if (p->p_stat == SZOMB) { 448 /* 449 * The process's thread may still be in the middle 450 * of switching away, we can't rip its stack out from 451 * under it until TDF_RUNNING clears! 452 * 453 * YYY no wakeup occurs so we depend on the timeout. 454 */ 455 if ((p->p_thread->td_flags & TDF_RUNNING) != 0) { 456 tsleep(p->p_thread, 0, "reap", 1); 457 goto loop; 458 } 459 460 /* 461 * Other kernel threads may be in the middle of 462 * accessing the proc. For example, kern/kern_proc.c 463 * could be blocked writing proc data to a sysctl. 464 * At the moment, if this occurs, we are not woken 465 * up and rely on a one-second retry. 466 */ 467 if (p->p_lock) { 468 while (p->p_lock) 469 tsleep(p, 0, "reap2", hz); 470 } 471 lwkt_wait_free(p->p_thread); 472 473 /* 474 * Charge the parent for the child's change in 475 * estimated cpu as of when the child exits to 476 * account for batch scripts, large make's, etc. 477 */ 478 if (q->p_pid != 1) { 479 if (p->p_estcpu > p->p_estcpu_fork) { 480 q->p_estcpu = ESTCPULIM(q->p_estcpu + 481 p->p_estcpu - p->p_estcpu_fork); 482 } 483 } 484 485 /* Take care of our return values. */ 486 *res = p->p_pid; 487 if (status) 488 *status = p->p_xstat; 489 if (rusage) 490 *rusage = *p->p_ru; 491 /* 492 * If we got the child via a ptrace 'attach', 493 * we need to give it back to the old parent. 494 */ 495 if (p->p_oppid && (t = pfind(p->p_oppid))) { 496 p->p_oppid = 0; 497 proc_reparent(p, t); 498 psignal(t, SIGCHLD); 499 wakeup((caddr_t)t); 500 return (0); 501 } 502 p->p_xstat = 0; 503 ruadd(&q->p_stats->p_cru, p->p_ru); 504 FREE(p->p_ru, M_ZOMBIE); 505 p->p_ru = NULL; 506 507 /* 508 * Decrement the count of procs running with this uid. 509 */ 510 chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 511 512 /* 513 * Free up credentials. 514 */ 515 crfree(p->p_ucred); 516 p->p_ucred = NULL; 517 518 /* 519 * Remove unused arguments 520 */ 521 if (p->p_args && --p->p_args->ar_ref == 0) 522 FREE(p->p_args, M_PARGS); 523 524 /* 525 * Finally finished with old proc entry. 526 * Unlink it from its process group and free it. 527 */ 528 leavepgrp(p); 529 LIST_REMOVE(p, p_list); /* off zombproc */ 530 LIST_REMOVE(p, p_sibling); 531 532 if (--p->p_procsig->ps_refcnt == 0) { 533 if (p->p_sigacts != &p->p_addr->u_sigacts) 534 FREE(p->p_sigacts, M_SUBPROC); 535 FREE(p->p_procsig, M_SUBPROC); 536 p->p_procsig = NULL; 537 } 538 539 vm_waitproc(p); 540 zfree(proc_zone, p); 541 nprocs--; 542 return (0); 543 } 544 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 545 (p->p_flag & P_TRACED || options & WUNTRACED)) { 546 p->p_flag |= P_WAITED; 547 548 *res = p->p_pid; 549 if (status) 550 *status = W_STOPCODE(p->p_xstat); 551 /* Zero rusage so we get something consistent. */ 552 if (rusage) 553 bzero(rusage, sizeof(rusage)); 554 return (0); 555 } 556 } 557 if (nfound == 0) 558 return (ECHILD); 559 if (options & WNOHANG) { 560 *res = 0; 561 return (0); 562 } 563 error = tsleep((caddr_t)q, PCATCH, "wait", 0); 564 if (error) 565 return (error); 566 goto loop; 567 } 568 569 /* 570 * make process 'parent' the new parent of process 'child'. 571 */ 572 void 573 proc_reparent(child, parent) 574 struct proc *child; 575 struct proc *parent; 576 { 577 578 if (child->p_pptr == parent) 579 return; 580 581 LIST_REMOVE(child, p_sibling); 582 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 583 child->p_pptr = parent; 584 } 585 586 /* 587 * The next two functions are to handle adding/deleting items on the 588 * exit callout list 589 * 590 * at_exit(): 591 * Take the arguments given and put them onto the exit callout list, 592 * However first make sure that it's not already there. 593 * returns 0 on success. 594 */ 595 596 int 597 at_exit(function) 598 exitlist_fn function; 599 { 600 struct exitlist *ep; 601 602 #ifdef INVARIANTS 603 /* Be noisy if the programmer has lost track of things */ 604 if (rm_at_exit(function)) 605 printf("WARNING: exit callout entry (%p) already present\n", 606 function); 607 #endif 608 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT); 609 if (ep == NULL) 610 return (ENOMEM); 611 ep->function = function; 612 TAILQ_INSERT_TAIL(&exit_list, ep, next); 613 return (0); 614 } 615 616 /* 617 * Scan the exit callout list for the given item and remove it. 618 * Returns the number of items removed (0 or 1) 619 */ 620 int 621 rm_at_exit(function) 622 exitlist_fn function; 623 { 624 struct exitlist *ep; 625 626 TAILQ_FOREACH(ep, &exit_list, next) { 627 if (ep->function == function) { 628 TAILQ_REMOVE(&exit_list, ep, next); 629 free(ep, M_ATEXIT); 630 return(1); 631 } 632 } 633 return (0); 634 } 635 636 void check_sigacts (void) 637 { 638 struct proc *p = curproc; 639 struct sigacts *pss; 640 int s; 641 642 if (p->p_procsig->ps_refcnt == 1 && 643 p->p_sigacts != &p->p_addr->u_sigacts) { 644 pss = p->p_sigacts; 645 s = splhigh(); 646 p->p_addr->u_sigacts = *pss; 647 p->p_sigacts = &p->p_addr->u_sigacts; 648 splx(s); 649 FREE(pss, M_SUBPROC); 650 } 651 } 652 653