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 1.43 2005/06/26 04:36:31 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 #include <sys/thread2.h> 79 80 /* Required to be non-static for SysVR4 emulator */ 81 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 82 83 static MALLOC_DEFINE(M_ATEXIT, "atexit", "atexit callback"); 84 85 /* 86 * callout list for things to do at exit time 87 */ 88 struct exitlist { 89 exitlist_fn function; 90 TAILQ_ENTRY(exitlist) next; 91 }; 92 93 TAILQ_HEAD(exit_list_head, exitlist); 94 static struct exit_list_head exit_list = TAILQ_HEAD_INITIALIZER(exit_list); 95 96 /* 97 * exit -- 98 * Death of process. 99 * 100 * SYS_EXIT_ARGS(int rval) 101 */ 102 void 103 sys_exit(struct sys_exit_args *uap) 104 { 105 exit1(W_EXITCODE(uap->rval, 0)); 106 /* NOTREACHED */ 107 } 108 109 /* 110 * Exit: deallocate address space and other resources, change proc state 111 * to zombie, and unlink proc from allproc and parent's lists. Save exit 112 * status and rusage for wait(). Check for child processes and orphan them. 113 */ 114 void 115 exit1(int rv) 116 { 117 struct proc *p = curproc; 118 struct proc *q, *nq; 119 struct vmspace *vm; 120 struct vnode *vtmp; 121 struct exitlist *ep; 122 123 if (p->p_pid == 1) { 124 printf("init died (signal %d, exit %d)\n", 125 WTERMSIG(rv), WEXITSTATUS(rv)); 126 panic("Going nowhere without my init!"); 127 } 128 129 sysmsg_rundown(p, 1); 130 caps_exit(p->p_thread); 131 aio_proc_rundown(p); 132 133 /* are we a task leader? */ 134 if(p == p->p_leader) { 135 struct kill_args killArgs; 136 killArgs.signum = SIGKILL; 137 q = p->p_peers; 138 while(q) { 139 killArgs.pid = q->p_pid; 140 /* 141 * The interface for kill is better 142 * than the internal signal 143 */ 144 kill(&killArgs); 145 nq = q; 146 q = q->p_peers; 147 } 148 while (p->p_peers) 149 tsleep((caddr_t)p, 0, "exit1", 0); 150 } 151 152 #ifdef PGINPROF 153 vmsizmon(); 154 #endif 155 STOPEVENT(p, S_EXIT, rv); 156 wakeup(&p->p_stype); /* Wakeup anyone in procfs' PIOCWAIT */ 157 158 /* 159 * Check if any loadable modules need anything done at process exit. 160 * e.g. SYSV IPC stuff 161 * XXX what if one of these generates an error? 162 */ 163 TAILQ_FOREACH(ep, &exit_list, next) 164 (*ep->function)(p->p_thread); 165 166 if (p->p_flag & P_PROFIL) 167 stopprofclock(p); 168 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage), 169 M_ZOMBIE, M_WAITOK); 170 /* 171 * If parent is waiting for us to exit or exec, 172 * P_PPWAIT is set; we will wakeup the parent below. 173 */ 174 p->p_flag &= ~(P_TRACED | P_PPWAIT); 175 p->p_flag |= P_WEXIT; 176 SIGEMPTYSET(p->p_siglist); 177 if (timevalisset(&p->p_realtimer.it_value)) 178 callout_stop(&p->p_ithandle); 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 p->p_fd = NULL; 192 193 if(p->p_leader->p_peers) { 194 q = p->p_leader; 195 while(q->p_peers != p) 196 q = q->p_peers; 197 q->p_peers = p->p_peers; 198 wakeup((caddr_t)p->p_leader); 199 } 200 201 /* 202 * XXX Shutdown SYSV semaphores 203 */ 204 semexit(p); 205 206 KKASSERT(p->p_numposixlocks == 0); 207 208 /* The next two chunks should probably be moved to vmspace_exit. */ 209 vm = p->p_vmspace; 210 211 /* 212 * Release upcalls associated with this process 213 */ 214 if (vm->vm_upcalls) 215 upc_release(vm, p); 216 217 /* 218 * Release user portion of address space. 219 * This releases references to vnodes, 220 * which could cause I/O if the file has been unlinked. 221 * Need to do this early enough that we can still sleep. 222 * Can't free the entire vmspace as the kernel stack 223 * may be mapped within that space also. 224 * 225 * Processes sharing the same vmspace may exit in one order, and 226 * get cleaned up by vmspace_exit() in a different order. The 227 * last exiting process to reach this point releases as much of 228 * the environment as it can, and the last process cleaned up 229 * by vmspace_exit() (which decrements exitingcnt) cleans up the 230 * remainder. 231 */ 232 ++vm->vm_exitingcnt; 233 if (--vm->vm_refcnt == 0) { 234 shmexit(vm); 235 pmap_remove_pages(vmspace_pmap(vm), VM_MIN_ADDRESS, 236 VM_MAXUSER_ADDRESS); 237 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS, 238 VM_MAXUSER_ADDRESS); 239 } 240 241 if (SESS_LEADER(p)) { 242 struct session *sp = p->p_session; 243 struct vnode *vp; 244 245 if (sp->s_ttyvp) { 246 /* 247 * We are the controlling process. Signal the 248 * foreground process group, drain the controlling 249 * terminal, and revoke access to the controlling 250 * terminal. 251 * 252 * NOTE: while waiting for the process group to exit 253 * it is possible that one of the processes in the 254 * group will revoke the tty, so we have to recheck. 255 */ 256 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 257 if (sp->s_ttyp->t_pgrp) 258 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 259 (void) ttywait(sp->s_ttyp); 260 /* 261 * The tty could have been revoked 262 * if we blocked. 263 */ 264 if ((vp = sp->s_ttyvp) != NULL) { 265 ttyclosesession(sp, 0); 266 if (vx_lock(vp) == 0) { 267 VOP_REVOKE(vp, REVOKEALL); 268 vx_unlock(vp); 269 } 270 vrele(vp); /* s_ttyvp ref */ 271 } 272 } 273 /* 274 * Release the tty. If someone has it open via 275 * /dev/tty then close it (since they no longer can 276 * once we've NULL'd it out). 277 */ 278 if (sp->s_ttyvp) 279 ttyclosesession(sp, 1); 280 /* 281 * s_ttyp is not zero'd; we use this to indicate 282 * that the session once had a controlling terminal. 283 * (for logging and informational purposes) 284 */ 285 } 286 sp->s_leader = NULL; 287 } 288 fixjobc(p, p->p_pgrp, 0); 289 (void)acct_process(p); 290 #ifdef KTRACE 291 /* 292 * release trace file 293 */ 294 p->p_traceflag = 0; /* don't trace the vrele() */ 295 if ((vtmp = p->p_tracep) != NULL) { 296 p->p_tracep = NULL; 297 vrele(vtmp); 298 } 299 #endif 300 /* 301 * Release reference to text vnode 302 */ 303 if ((vtmp = p->p_textvp) != NULL) { 304 p->p_textvp = NULL; 305 vrele(vtmp); 306 } 307 308 /* 309 * Once we set SZOMB the process can get reaped. The wait1 code 310 * will also wait for TDF_RUNNING to be cleared in the thread's flags, 311 * indicating that it has been completely switched out. 312 */ 313 314 /* 315 * Remove proc from allproc queue and pidhash chain. 316 * Place onto zombproc. Unlink from parent's child list. 317 */ 318 LIST_REMOVE(p, p_list); 319 LIST_INSERT_HEAD(&zombproc, p, p_list); 320 p->p_stat = SZOMB; 321 322 LIST_REMOVE(p, p_hash); 323 324 q = LIST_FIRST(&p->p_children); 325 if (q) /* only need this if any child is S_ZOMB */ 326 wakeup((caddr_t) initproc); 327 for (; q != 0; q = nq) { 328 nq = LIST_NEXT(q, p_sibling); 329 LIST_REMOVE(q, p_sibling); 330 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling); 331 q->p_pptr = initproc; 332 q->p_sigparent = SIGCHLD; 333 /* 334 * Traced processes are killed 335 * since their existence means someone is screwing up. 336 */ 337 if (q->p_flag & P_TRACED) { 338 q->p_flag &= ~P_TRACED; 339 psignal(q, SIGKILL); 340 } 341 } 342 343 /* 344 * Save exit status and final rusage info, adding in child rusage 345 * info and self times. 346 */ 347 p->p_xstat = rv; 348 *p->p_ru = p->p_stats->p_ru; 349 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 350 ruadd(p->p_ru, &p->p_stats->p_cru); 351 352 /* 353 * notify interested parties of our demise. 354 */ 355 KNOTE(&p->p_klist, NOTE_EXIT); 356 357 /* 358 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 359 * flag set, notify process 1 instead (and hope it will handle 360 * this situation). 361 */ 362 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) { 363 struct proc *pp = p->p_pptr; 364 proc_reparent(p, initproc); 365 /* 366 * If this was the last child of our parent, notify 367 * parent, so in case he was wait(2)ing, he will 368 * continue. 369 */ 370 if (LIST_EMPTY(&pp->p_children)) 371 wakeup((caddr_t)pp); 372 } 373 374 if (p->p_sigparent && p->p_pptr != initproc) { 375 psignal(p->p_pptr, p->p_sigparent); 376 } else { 377 psignal(p->p_pptr, SIGCHLD); 378 } 379 380 wakeup((caddr_t)p->p_pptr); 381 #if defined(tahoe) 382 /* move this to cpu_exit */ 383 p->p_thread->td_pcb->pcb_saveacc.faddr = (float *)NULL; 384 #endif 385 /* 386 * cpu_exit is responsible for clearing curproc, since 387 * it is heavily integrated with the thread/switching sequence. 388 * 389 * Other substructures are freed from wait(). 390 */ 391 if (--p->p_limit->p_refcnt == 0) { 392 FREE(p->p_limit, M_SUBPROC); 393 p->p_limit = NULL; 394 } 395 396 /* 397 * Release the current user process designation on the process so 398 * the userland scheduler can work in someone else. 399 */ 400 p->p_usched->release_curproc(p); 401 402 /* 403 * Finally, call machine-dependent code to release the remaining 404 * resources including address space, the kernel stack and pcb. 405 * The address space is released by "vmspace_free(p->p_vmspace)"; 406 * This is machine-dependent, as we may have to change stacks 407 * or ensure that the current one isn't reallocated before we 408 * finish. cpu_exit will end with a call to cpu_switch(), finishing 409 * our execution (pun intended). 410 */ 411 cpu_proc_exit(); 412 } 413 414 int 415 wait4(struct wait_args *uap) 416 { 417 struct rusage rusage; 418 int error, status; 419 420 error = kern_wait(uap->pid, uap->status ? &status : NULL, 421 uap->options, uap->rusage ? &rusage : NULL, &uap->sysmsg_fds[0]); 422 423 if (error == 0 && uap->status) 424 error = copyout(&status, uap->status, sizeof(*uap->status)); 425 if (error == 0 && uap->rusage) 426 error = copyout(&rusage, uap->rusage, sizeof(*uap->rusage)); 427 return (error); 428 } 429 430 /* 431 * wait1() 432 * 433 * wait_args(int pid, int *status, int options, struct rusage *rusage) 434 */ 435 int 436 kern_wait(pid_t pid, int *status, int options, struct rusage *rusage, int *res) 437 { 438 struct thread *td = curthread; 439 struct proc *q = td->td_proc; 440 struct proc *p, *t; 441 int nfound, error; 442 443 if (pid == 0) 444 pid = -q->p_pgid; 445 if (options &~ (WUNTRACED|WNOHANG|WLINUXCLONE)) 446 return (EINVAL); 447 loop: 448 nfound = 0; 449 LIST_FOREACH(p, &q->p_children, p_sibling) { 450 if (pid != WAIT_ANY && 451 p->p_pid != pid && p->p_pgid != -pid) 452 continue; 453 454 /* This special case handles a kthread spawned by linux_clone 455 * (see linux_misc.c). The linux_wait4 and linux_waitpid 456 * functions need to be able to distinguish between waiting 457 * on a process and waiting on a thread. It is a thread if 458 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option 459 * signifies we want to wait for threads and not processes. 460 */ 461 if ((p->p_sigparent != SIGCHLD) ^ 462 ((options & WLINUXCLONE) != 0)) { 463 continue; 464 } 465 466 nfound++; 467 if (p->p_stat == SZOMB) { 468 /* 469 * The process's thread may still be in the middle 470 * of switching away, we can't rip its stack out from 471 * under it until TDF_RUNNING clears! 472 * 473 * YYY no wakeup occurs so we depend on the timeout. 474 */ 475 if ((p->p_thread->td_flags & TDF_RUNNING) != 0) { 476 tsleep(p->p_thread, 0, "reap", 1); 477 goto loop; 478 } 479 480 /* 481 * Other kernel threads may be in the middle of 482 * accessing the proc. For example, kern/kern_proc.c 483 * could be blocked writing proc data to a sysctl. 484 * At the moment, if this occurs, we are not woken 485 * up and rely on a one-second retry. 486 */ 487 if (p->p_lock) { 488 while (p->p_lock) 489 tsleep(p, 0, "reap2", hz); 490 } 491 lwkt_wait_free(p->p_thread); 492 493 /* scheduling hook for heuristic */ 494 p->p_usched->heuristic_exiting(q, p); 495 496 /* Take care of our return values. */ 497 *res = p->p_pid; 498 if (status) 499 *status = p->p_xstat; 500 if (rusage) 501 *rusage = *p->p_ru; 502 /* 503 * If we got the child via a ptrace 'attach', 504 * we need to give it back to the old parent. 505 */ 506 if (p->p_oppid && (t = pfind(p->p_oppid))) { 507 p->p_oppid = 0; 508 proc_reparent(p, t); 509 psignal(t, SIGCHLD); 510 wakeup((caddr_t)t); 511 return (0); 512 } 513 p->p_xstat = 0; 514 ruadd(&q->p_stats->p_cru, p->p_ru); 515 FREE(p->p_ru, M_ZOMBIE); 516 p->p_ru = NULL; 517 518 /* 519 * Decrement the count of procs running with this uid. 520 */ 521 chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 522 523 /* 524 * Free up credentials. 525 */ 526 crfree(p->p_ucred); 527 p->p_ucred = NULL; 528 529 /* 530 * Remove unused arguments 531 */ 532 if (p->p_args && --p->p_args->ar_ref == 0) 533 FREE(p->p_args, M_PARGS); 534 535 /* 536 * Finally finished with old proc entry. 537 * Unlink it from its process group and free it. 538 */ 539 leavepgrp(p); 540 LIST_REMOVE(p, p_list); /* off zombproc */ 541 LIST_REMOVE(p, p_sibling); 542 543 if (--p->p_procsig->ps_refcnt == 0) { 544 if (p->p_sigacts != &p->p_addr->u_sigacts) 545 FREE(p->p_sigacts, M_SUBPROC); 546 FREE(p->p_procsig, M_SUBPROC); 547 p->p_procsig = NULL; 548 } 549 550 vm_waitproc(p); 551 zfree(proc_zone, p); 552 nprocs--; 553 return (0); 554 } 555 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 556 (p->p_flag & P_TRACED || options & WUNTRACED)) { 557 p->p_flag |= P_WAITED; 558 559 *res = p->p_pid; 560 if (status) 561 *status = W_STOPCODE(p->p_xstat); 562 /* Zero rusage so we get something consistent. */ 563 if (rusage) 564 bzero(rusage, sizeof(rusage)); 565 return (0); 566 } 567 } 568 if (nfound == 0) 569 return (ECHILD); 570 if (options & WNOHANG) { 571 *res = 0; 572 return (0); 573 } 574 error = tsleep((caddr_t)q, PCATCH, "wait", 0); 575 if (error) 576 return (error); 577 goto loop; 578 } 579 580 /* 581 * make process 'parent' the new parent of process 'child'. 582 */ 583 void 584 proc_reparent(struct proc *child, struct proc *parent) 585 { 586 587 if (child->p_pptr == parent) 588 return; 589 590 LIST_REMOVE(child, p_sibling); 591 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 592 child->p_pptr = parent; 593 } 594 595 /* 596 * The next two functions are to handle adding/deleting items on the 597 * exit callout list 598 * 599 * at_exit(): 600 * Take the arguments given and put them onto the exit callout list, 601 * However first make sure that it's not already there. 602 * returns 0 on success. 603 */ 604 605 int 606 at_exit(exitlist_fn function) 607 { 608 struct exitlist *ep; 609 610 #ifdef INVARIANTS 611 /* Be noisy if the programmer has lost track of things */ 612 if (rm_at_exit(function)) 613 printf("WARNING: exit callout entry (%p) already present\n", 614 function); 615 #endif 616 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT); 617 if (ep == NULL) 618 return (ENOMEM); 619 ep->function = function; 620 TAILQ_INSERT_TAIL(&exit_list, ep, next); 621 return (0); 622 } 623 624 /* 625 * Scan the exit callout list for the given item and remove it. 626 * Returns the number of items removed (0 or 1) 627 */ 628 int 629 rm_at_exit(exitlist_fn function) 630 { 631 struct exitlist *ep; 632 633 TAILQ_FOREACH(ep, &exit_list, next) { 634 if (ep->function == function) { 635 TAILQ_REMOVE(&exit_list, ep, next); 636 free(ep, M_ATEXIT); 637 return(1); 638 } 639 } 640 return (0); 641 } 642 643 void 644 check_sigacts(void) 645 { 646 struct proc *p = curproc; 647 struct sigacts *pss; 648 649 if (p->p_procsig->ps_refcnt == 1 && 650 p->p_sigacts != &p->p_addr->u_sigacts) { 651 pss = p->p_sigacts; 652 crit_enter(); 653 p->p_addr->u_sigacts = *pss; 654 p->p_sigacts = &p->p_addr->u_sigacts; 655 crit_exit(); 656 FREE(pss, M_SUBPROC); 657 } 658 } 659 660