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 1.39 2004/10/12 19:20:46 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 callout_stop(&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 ttyclosesession(sp, 0); 264 if (vx_lock(vp) == 0) { 265 VOP_REVOKE(vp, REVOKEALL); 266 vx_unlock(vp); 267 } 268 vrele(vp); /* s_ttyvp ref */ 269 } 270 } 271 /* 272 * Release the tty. If someone has it open via 273 * /dev/tty then close it (since they no longer can 274 * once we've NULL'd it out). 275 */ 276 if (sp->s_ttyvp) 277 ttyclosesession(sp, 1); 278 /* 279 * s_ttyp is not zero'd; we use this to indicate 280 * that the session once had a controlling terminal. 281 * (for logging and informational purposes) 282 */ 283 } 284 sp->s_leader = NULL; 285 } 286 fixjobc(p, p->p_pgrp, 0); 287 (void)acct_process(p); 288 #ifdef KTRACE 289 /* 290 * release trace file 291 */ 292 p->p_traceflag = 0; /* don't trace the vrele() */ 293 if ((vtmp = p->p_tracep) != NULL) { 294 p->p_tracep = NULL; 295 vrele(vtmp); 296 } 297 #endif 298 /* 299 * Release reference to text vnode 300 */ 301 if ((vtmp = p->p_textvp) != NULL) { 302 p->p_textvp = NULL; 303 vrele(vtmp); 304 } 305 306 /* 307 * Once we set SZOMB the process can get reaped. The wait1 code 308 * will also wait for TDF_RUNNING to be cleared in the thread's flags, 309 * indicating that it has been completely switched out. 310 */ 311 312 /* 313 * Remove proc from allproc queue and pidhash chain. 314 * Place onto zombproc. Unlink from parent's child list. 315 */ 316 LIST_REMOVE(p, p_list); 317 LIST_INSERT_HEAD(&zombproc, p, p_list); 318 p->p_stat = SZOMB; 319 320 LIST_REMOVE(p, p_hash); 321 322 q = LIST_FIRST(&p->p_children); 323 if (q) /* only need this if any child is S_ZOMB */ 324 wakeup((caddr_t) initproc); 325 for (; q != 0; q = nq) { 326 nq = LIST_NEXT(q, p_sibling); 327 LIST_REMOVE(q, p_sibling); 328 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling); 329 q->p_pptr = initproc; 330 q->p_sigparent = SIGCHLD; 331 /* 332 * Traced processes are killed 333 * since their existence means someone is screwing up. 334 */ 335 if (q->p_flag & P_TRACED) { 336 q->p_flag &= ~P_TRACED; 337 psignal(q, SIGKILL); 338 } 339 } 340 341 /* 342 * Save exit status and final rusage info, adding in child rusage 343 * info and self times. 344 */ 345 p->p_xstat = rv; 346 *p->p_ru = p->p_stats->p_ru; 347 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 348 ruadd(p->p_ru, &p->p_stats->p_cru); 349 350 /* 351 * notify interested parties of our demise. 352 */ 353 KNOTE(&p->p_klist, NOTE_EXIT); 354 355 /* 356 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 357 * flag set, notify process 1 instead (and hope it will handle 358 * this situation). 359 */ 360 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) { 361 struct proc *pp = p->p_pptr; 362 proc_reparent(p, initproc); 363 /* 364 * If this was the last child of our parent, notify 365 * parent, so in case he was wait(2)ing, he will 366 * continue. 367 */ 368 if (LIST_EMPTY(&pp->p_children)) 369 wakeup((caddr_t)pp); 370 } 371 372 if (p->p_sigparent && p->p_pptr != initproc) { 373 psignal(p->p_pptr, p->p_sigparent); 374 } else { 375 psignal(p->p_pptr, SIGCHLD); 376 } 377 378 wakeup((caddr_t)p->p_pptr); 379 #if defined(tahoe) 380 /* move this to cpu_exit */ 381 p->p_thread->td_pcb->pcb_saveacc.faddr = (float *)NULL; 382 #endif 383 /* 384 * cpu_exit is responsible for clearing curproc, since 385 * it is heavily integrated with the thread/switching sequence. 386 * 387 * Other substructures are freed from wait(). 388 */ 389 if (--p->p_limit->p_refcnt == 0) { 390 FREE(p->p_limit, M_SUBPROC); 391 p->p_limit = NULL; 392 } 393 394 /* 395 * Release the current user process designation on the process so 396 * the userland scheduler can work in someone else. 397 */ 398 release_curproc(p); 399 400 /* 401 * Finally, call machine-dependent code to release the remaining 402 * resources including address space, the kernel stack and pcb. 403 * The address space is released by "vmspace_free(p->p_vmspace)"; 404 * This is machine-dependent, as we may have to change stacks 405 * or ensure that the current one isn't reallocated before we 406 * finish. cpu_exit will end with a call to cpu_switch(), finishing 407 * our execution (pun intended). 408 */ 409 cpu_proc_exit(); 410 } 411 412 int 413 wait4(struct wait_args *uap) 414 { 415 struct rusage rusage; 416 int error, status; 417 418 error = kern_wait(uap->pid, uap->status ? &status : NULL, 419 uap->options, uap->rusage ? &rusage : NULL, &uap->sysmsg_fds[0]); 420 421 if (error == 0 && uap->status) 422 error = copyout(&status, uap->status, sizeof(*uap->status)); 423 if (error == 0 && uap->rusage) 424 error = copyout(&rusage, uap->rusage, sizeof(*uap->rusage)); 425 return (error); 426 } 427 428 /* 429 * wait1() 430 * 431 * wait_args(int pid, int *status, int options, struct rusage *rusage) 432 */ 433 int 434 kern_wait(pid_t pid, int *status, int options, struct rusage *rusage, int *res) 435 { 436 struct thread *td = curthread; 437 struct proc *q = td->td_proc; 438 struct proc *p, *t; 439 int nfound, error; 440 441 if (pid == 0) 442 pid = -q->p_pgid; 443 if (options &~ (WUNTRACED|WNOHANG|WLINUXCLONE)) 444 return (EINVAL); 445 loop: 446 nfound = 0; 447 LIST_FOREACH(p, &q->p_children, p_sibling) { 448 if (pid != WAIT_ANY && 449 p->p_pid != pid && p->p_pgid != -pid) 450 continue; 451 452 /* This special case handles a kthread spawned by linux_clone 453 * (see linux_misc.c). The linux_wait4 and linux_waitpid functions 454 * need to be able to distinguish between waiting on a process and 455 * waiting on a thread. It is a thread if p_sigparent is not SIGCHLD, 456 * and the WLINUXCLONE option signifies we want to wait for threads 457 * and not processes. 458 */ 459 if ((p->p_sigparent != SIGCHLD) ^ ((options & WLINUXCLONE) != 0)) 460 continue; 461 462 nfound++; 463 if (p->p_stat == SZOMB) { 464 /* 465 * The process's thread may still be in the middle 466 * of switching away, we can't rip its stack out from 467 * under it until TDF_RUNNING clears! 468 * 469 * YYY no wakeup occurs so we depend on the timeout. 470 */ 471 if ((p->p_thread->td_flags & TDF_RUNNING) != 0) { 472 tsleep(p->p_thread, 0, "reap", 1); 473 goto loop; 474 } 475 476 /* 477 * Other kernel threads may be in the middle of 478 * accessing the proc. For example, kern/kern_proc.c 479 * could be blocked writing proc data to a sysctl. 480 * At the moment, if this occurs, we are not woken 481 * up and rely on a one-second retry. 482 */ 483 if (p->p_lock) { 484 while (p->p_lock) 485 tsleep(p, 0, "reap2", hz); 486 } 487 lwkt_wait_free(p->p_thread); 488 489 /* 490 * Charge the parent for the child's change in 491 * estimated cpu as of when the child exits to 492 * account for batch scripts, large make's, etc. 493 */ 494 if (q->p_pid != 1) { 495 if (p->p_estcpu > p->p_estcpu_fork) { 496 q->p_estcpu = ESTCPULIM(q->p_estcpu + 497 p->p_estcpu - p->p_estcpu_fork); 498 } 499 } 500 501 /* Take care of our return values. */ 502 *res = p->p_pid; 503 if (status) 504 *status = p->p_xstat; 505 if (rusage) 506 *rusage = *p->p_ru; 507 /* 508 * If we got the child via a ptrace 'attach', 509 * we need to give it back to the old parent. 510 */ 511 if (p->p_oppid && (t = pfind(p->p_oppid))) { 512 p->p_oppid = 0; 513 proc_reparent(p, t); 514 psignal(t, SIGCHLD); 515 wakeup((caddr_t)t); 516 return (0); 517 } 518 p->p_xstat = 0; 519 ruadd(&q->p_stats->p_cru, p->p_ru); 520 FREE(p->p_ru, M_ZOMBIE); 521 p->p_ru = NULL; 522 523 /* 524 * Decrement the count of procs running with this uid. 525 */ 526 chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 527 528 /* 529 * Free up credentials. 530 */ 531 crfree(p->p_ucred); 532 p->p_ucred = NULL; 533 534 /* 535 * Remove unused arguments 536 */ 537 if (p->p_args && --p->p_args->ar_ref == 0) 538 FREE(p->p_args, M_PARGS); 539 540 /* 541 * Finally finished with old proc entry. 542 * Unlink it from its process group and free it. 543 */ 544 leavepgrp(p); 545 LIST_REMOVE(p, p_list); /* off zombproc */ 546 LIST_REMOVE(p, p_sibling); 547 548 if (--p->p_procsig->ps_refcnt == 0) { 549 if (p->p_sigacts != &p->p_addr->u_sigacts) 550 FREE(p->p_sigacts, M_SUBPROC); 551 FREE(p->p_procsig, M_SUBPROC); 552 p->p_procsig = NULL; 553 } 554 555 vm_waitproc(p); 556 zfree(proc_zone, p); 557 nprocs--; 558 return (0); 559 } 560 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 561 (p->p_flag & P_TRACED || options & WUNTRACED)) { 562 p->p_flag |= P_WAITED; 563 564 *res = p->p_pid; 565 if (status) 566 *status = W_STOPCODE(p->p_xstat); 567 /* Zero rusage so we get something consistent. */ 568 if (rusage) 569 bzero(rusage, sizeof(rusage)); 570 return (0); 571 } 572 } 573 if (nfound == 0) 574 return (ECHILD); 575 if (options & WNOHANG) { 576 *res = 0; 577 return (0); 578 } 579 error = tsleep((caddr_t)q, PCATCH, "wait", 0); 580 if (error) 581 return (error); 582 goto loop; 583 } 584 585 /* 586 * make process 'parent' the new parent of process 'child'. 587 */ 588 void 589 proc_reparent(child, parent) 590 struct proc *child; 591 struct proc *parent; 592 { 593 594 if (child->p_pptr == parent) 595 return; 596 597 LIST_REMOVE(child, p_sibling); 598 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 599 child->p_pptr = parent; 600 } 601 602 /* 603 * The next two functions are to handle adding/deleting items on the 604 * exit callout list 605 * 606 * at_exit(): 607 * Take the arguments given and put them onto the exit callout list, 608 * However first make sure that it's not already there. 609 * returns 0 on success. 610 */ 611 612 int 613 at_exit(function) 614 exitlist_fn function; 615 { 616 struct exitlist *ep; 617 618 #ifdef INVARIANTS 619 /* Be noisy if the programmer has lost track of things */ 620 if (rm_at_exit(function)) 621 printf("WARNING: exit callout entry (%p) already present\n", 622 function); 623 #endif 624 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT); 625 if (ep == NULL) 626 return (ENOMEM); 627 ep->function = function; 628 TAILQ_INSERT_TAIL(&exit_list, ep, next); 629 return (0); 630 } 631 632 /* 633 * Scan the exit callout list for the given item and remove it. 634 * Returns the number of items removed (0 or 1) 635 */ 636 int 637 rm_at_exit(function) 638 exitlist_fn function; 639 { 640 struct exitlist *ep; 641 642 TAILQ_FOREACH(ep, &exit_list, next) { 643 if (ep->function == function) { 644 TAILQ_REMOVE(&exit_list, ep, next); 645 free(ep, M_ATEXIT); 646 return(1); 647 } 648 } 649 return (0); 650 } 651 652 void check_sigacts (void) 653 { 654 struct proc *p = curproc; 655 struct sigacts *pss; 656 int s; 657 658 if (p->p_procsig->ps_refcnt == 1 && 659 p->p_sigacts != &p->p_addr->u_sigacts) { 660 pss = p->p_sigacts; 661 s = splhigh(); 662 p->p_addr->u_sigacts = *pss; 663 p->p_sigacts = &p->p_addr->u_sigacts; 664 splx(s); 665 FREE(pss, M_SUBPROC); 666 } 667 } 668 669