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 1.61 2006/09/05 00:55:45 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/ktrace.h> 53 #include <sys/pioctl.h> 54 #include <sys/tty.h> 55 #include <sys/wait.h> 56 #include <sys/vnode.h> 57 #include <sys/resourcevar.h> 58 #include <sys/signalvar.h> 59 #include <sys/ptrace.h> 60 #include <sys/acct.h> /* for acct_process() function prototype */ 61 #include <sys/filedesc.h> 62 #include <sys/shm.h> 63 #include <sys/sem.h> 64 #include <sys/aio.h> 65 #include <sys/jail.h> 66 #include <sys/kern_syscall.h> 67 #include <sys/upcall.h> 68 #include <sys/caps.h> 69 70 #include <vm/vm.h> 71 #include <vm/vm_param.h> 72 #include <sys/lock.h> 73 #include <vm/pmap.h> 74 #include <vm/vm_map.h> 75 #include <vm/vm_zone.h> 76 #include <vm/vm_extern.h> 77 #include <sys/user.h> 78 79 #include <sys/thread2.h> 80 81 static MALLOC_DEFINE(M_ATEXIT, "atexit", "atexit callback"); 82 static MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 83 84 /* 85 * callout list for things to do at exit time 86 */ 87 struct exitlist { 88 exitlist_fn function; 89 TAILQ_ENTRY(exitlist) next; 90 }; 91 92 TAILQ_HEAD(exit_list_head, exitlist); 93 static struct exit_list_head exit_list = TAILQ_HEAD_INITIALIZER(exit_list); 94 95 /* 96 * exit -- 97 * Death of process. 98 * 99 * SYS_EXIT_ARGS(int rval) 100 */ 101 int 102 sys_exit(struct exit_args *uap) 103 { 104 exit1(W_EXITCODE(uap->rval, 0)); 105 /* NOTREACHED */ 106 } 107 108 /* 109 * Exit: deallocate address space and other resources, change proc state 110 * to zombie, and unlink proc from allproc and parent's lists. Save exit 111 * status and rusage for wait(). Check for child processes and orphan them. 112 */ 113 void 114 exit1(int rv) 115 { 116 struct proc *p = curproc; 117 struct lwp *lp; 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 lp = &p->p_lwp; /* XXX lwp kill other threads */ 130 131 caps_exit(lp->lwp_thread); 132 aio_proc_rundown(p); 133 134 /* are we a task leader? */ 135 if(p == p->p_leader) { 136 struct kill_args killArgs; 137 killArgs.signum = SIGKILL; 138 q = p->p_peers; 139 while(q) { 140 killArgs.pid = q->p_pid; 141 /* 142 * The interface for kill is better 143 * than the internal signal 144 */ 145 sys_kill(&killArgs); 146 nq = q; 147 q = q->p_peers; 148 } 149 while (p->p_peers) 150 tsleep((caddr_t)p, 0, "exit1", 0); 151 } 152 153 #ifdef PGINPROF 154 vmsizmon(); 155 #endif 156 STOPEVENT(p, S_EXIT, rv); 157 wakeup(&p->p_stype); /* Wakeup anyone in procfs' PIOCWAIT */ 158 159 /* 160 * Check if any loadable modules need anything done at process exit. 161 * e.g. SYSV IPC stuff 162 * XXX what if one of these generates an error? 163 */ 164 TAILQ_FOREACH(ep, &exit_list, next) 165 (*ep->function)(p->p_thread); 166 167 if (p->p_flag & P_PROFIL) 168 stopprofclock(p); 169 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage), 170 M_ZOMBIE, M_WAITOK); 171 /* 172 * If parent is waiting for us to exit or exec, 173 * P_PPWAIT is set; we will wakeup the parent below. 174 */ 175 p->p_flag &= ~(P_TRACED | P_PPWAIT); 176 p->p_flag |= P_WEXIT; 177 SIGEMPTYSET(p->p_siglist); 178 if (timevalisset(&p->p_realtimer.it_value)) 179 callout_stop(&p->p_ithandle); 180 181 /* 182 * Reset any sigio structures pointing to us as a result of 183 * F_SETOWN with our pid. 184 */ 185 funsetownlst(&p->p_sigiolst); 186 187 /* 188 * Close open files and release open-file table. 189 * This may block! 190 */ 191 fdfree(p); 192 p->p_fd = NULL; 193 194 if(p->p_leader->p_peers) { 195 q = p->p_leader; 196 while(q->p_peers != p) 197 q = q->p_peers; 198 q->p_peers = p->p_peers; 199 wakeup((caddr_t)p->p_leader); 200 } 201 202 /* 203 * XXX Shutdown SYSV semaphores 204 */ 205 semexit(p); 206 207 KKASSERT(p->p_numposixlocks == 0); 208 209 /* The next two chunks should probably be moved to vmspace_exit. */ 210 vm = p->p_vmspace; 211 212 /* 213 * Release upcalls associated with this process 214 */ 215 if (vm->vm_upcalls) 216 upc_release(vm, &p->p_lwp); 217 218 /* 219 * Release user portion of address space. 220 * This releases references to vnodes, 221 * which could cause I/O if the file has been unlinked. 222 * Need to do this early enough that we can still sleep. 223 * Can't free the entire vmspace as the kernel stack 224 * may be mapped within that space also. 225 * 226 * Processes sharing the same vmspace may exit in one order, and 227 * get cleaned up by vmspace_exit() in a different order. The 228 * last exiting process to reach this point releases as much of 229 * the environment as it can, and the last process cleaned up 230 * by vmspace_exit() (which decrements exitingcnt) cleans up the 231 * remainder. 232 */ 233 ++vm->vm_exitingcnt; 234 if (--vm->vm_refcnt == 0) { 235 shmexit(vm); 236 pmap_remove_pages(vmspace_pmap(vm), VM_MIN_ADDRESS, 237 VM_MAXUSER_ADDRESS); 238 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS, 239 VM_MAXUSER_ADDRESS); 240 } 241 242 if (SESS_LEADER(p)) { 243 struct session *sp = p->p_session; 244 struct vnode *vp; 245 246 if (sp->s_ttyvp) { 247 /* 248 * We are the controlling process. Signal the 249 * foreground process group, drain the controlling 250 * terminal, and revoke access to the controlling 251 * terminal. 252 * 253 * NOTE: while waiting for the process group to exit 254 * it is possible that one of the processes in the 255 * group will revoke the tty, so we have to recheck. 256 */ 257 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 258 if (sp->s_ttyp->t_pgrp) 259 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 260 (void) ttywait(sp->s_ttyp); 261 /* 262 * The tty could have been revoked 263 * if we blocked. 264 */ 265 if ((vp = sp->s_ttyvp) != NULL) { 266 ttyclosesession(sp, 0); 267 vx_lock(vp); 268 VOP_REVOKE(vp, REVOKEALL); 269 vx_unlock(vp); 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 if (p->p_tracenode) 295 ktrdestroy(&p->p_tracenode); 296 p->p_traceflag = 0; 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 * Move the process to the zombie list. This will block 308 * until the process p_lock count reaches 0. The process will 309 * not be reaped until TDF_EXITING is set by cpu_thread_exit(), 310 * which is called from cpu_proc_exit(). 311 */ 312 proc_move_allproc_zombie(p); 313 314 q = LIST_FIRST(&p->p_children); 315 if (q) /* only need this if any child is S_ZOMB */ 316 wakeup((caddr_t) initproc); 317 for (; q != 0; q = nq) { 318 nq = LIST_NEXT(q, p_sibling); 319 LIST_REMOVE(q, p_sibling); 320 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling); 321 q->p_pptr = initproc; 322 q->p_sigparent = SIGCHLD; 323 /* 324 * Traced processes are killed 325 * since their existence means someone is screwing up. 326 */ 327 if (q->p_flag & P_TRACED) { 328 q->p_flag &= ~P_TRACED; 329 ksignal(q, SIGKILL); 330 } 331 } 332 333 /* 334 * Save exit status and final rusage info, adding in child rusage 335 * info and self times. 336 */ 337 p->p_xstat = rv; 338 *p->p_ru = p->p_stats->p_ru; 339 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 340 ruadd(p->p_ru, &p->p_stats->p_cru); 341 342 /* 343 * notify interested parties of our demise. 344 */ 345 KNOTE(&p->p_klist, NOTE_EXIT); 346 347 /* 348 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT 349 * flag set, notify process 1 instead (and hope it will handle 350 * this situation). 351 */ 352 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) { 353 struct proc *pp = p->p_pptr; 354 proc_reparent(p, initproc); 355 /* 356 * If this was the last child of our parent, notify 357 * parent, so in case he was wait(2)ing, he will 358 * continue. 359 */ 360 if (LIST_EMPTY(&pp->p_children)) 361 wakeup((caddr_t)pp); 362 } 363 364 if (p->p_sigparent && p->p_pptr != initproc) { 365 ksignal(p->p_pptr, p->p_sigparent); 366 } else { 367 ksignal(p->p_pptr, SIGCHLD); 368 } 369 370 wakeup((caddr_t)p->p_pptr); 371 /* 372 * cpu_exit is responsible for clearing curproc, since 373 * it is heavily integrated with the thread/switching sequence. 374 * 375 * Other substructures are freed from wait(). 376 */ 377 plimit_free(&p->p_limit); 378 379 /* 380 * Release the current user process designation on the process so 381 * the userland scheduler can work in someone else. 382 */ 383 p->p_usched->release_curproc(lp); 384 385 /* 386 * Finally, call machine-dependent code to release the remaining 387 * resources including address space, the kernel stack and pcb. 388 * The address space is released by "vmspace_free(p->p_vmspace)"; 389 * This is machine-dependent, as we may have to change stacks 390 * or ensure that the current one isn't reallocated before we 391 * finish. cpu_exit will end with a call to cpu_switch(), finishing 392 * our execution (pun intended). 393 */ 394 cpu_proc_exit(); 395 } 396 397 int 398 sys_wait4(struct wait_args *uap) 399 { 400 struct rusage rusage; 401 int error, status; 402 403 error = kern_wait(uap->pid, uap->status ? &status : NULL, 404 uap->options, uap->rusage ? &rusage : NULL, &uap->sysmsg_fds[0]); 405 406 if (error == 0 && uap->status) 407 error = copyout(&status, uap->status, sizeof(*uap->status)); 408 if (error == 0 && uap->rusage) 409 error = copyout(&rusage, uap->rusage, sizeof(*uap->rusage)); 410 return (error); 411 } 412 413 /* 414 * wait1() 415 * 416 * wait_args(int pid, int *status, int options, struct rusage *rusage) 417 */ 418 int 419 kern_wait(pid_t pid, int *status, int options, struct rusage *rusage, int *res) 420 { 421 struct thread *td = curthread; 422 struct proc *q = td->td_proc; 423 struct proc *p, *t; 424 int nfound, error; 425 426 if (pid == 0) 427 pid = -q->p_pgid; 428 if (options &~ (WUNTRACED|WNOHANG|WLINUXCLONE)) 429 return (EINVAL); 430 loop: 431 /* 432 * Hack for backwards compatibility with badly written user code. 433 * Or perhaps we have to do this anyway, it is unclear. XXX 434 * 435 * The problem is that if a process group is stopped and the parent 436 * is doing a wait*(..., WUNTRACED, ...), it will see the STOP 437 * of the child and then stop itself when it tries to return from the 438 * system call. When the process group is resumed the parent will 439 * then get the STOP status even though the child has now resumed 440 * (a followup wait*() will get the CONT status). 441 * 442 * Previously the CONT would overwrite the STOP because the tstop 443 * was handled within tsleep(), and the parent would only see 444 * the CONT when both are stopped and continued together. This litte 445 * two-line hack restores this effect. 446 */ 447 while (q->p_flag & P_STOPPED) 448 tstop(q); 449 450 nfound = 0; 451 LIST_FOREACH(p, &q->p_children, p_sibling) { 452 if (pid != WAIT_ANY && 453 p->p_pid != pid && p->p_pgid != -pid) 454 continue; 455 456 /* This special case handles a kthread spawned by linux_clone 457 * (see linux_misc.c). The linux_wait4 and linux_waitpid 458 * functions need to be able to distinguish between waiting 459 * on a process and waiting on a thread. It is a thread if 460 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option 461 * signifies we want to wait for threads and not processes. 462 */ 463 if ((p->p_sigparent != SIGCHLD) ^ 464 ((options & WLINUXCLONE) != 0)) { 465 continue; 466 } 467 468 nfound++; 469 if (p->p_flag & P_ZOMBIE) { 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, "reap3", hz); 480 } 481 lwkt_wait_free(p->p_thread); 482 483 /* 484 * The process's thread may still be in the middle 485 * of switching away, we can't rip its stack out from 486 * under it until TDF_EXITING is set and both 487 * TDF_RUNNING and TDF_PREEMPT_LOCK are clear. 488 * TDF_PREEMPT_LOCK must be checked because TDF_RUNNING 489 * will be cleared temporarily if a thread gets 490 * preempted. 491 * 492 * YYY no wakeup occurs so we depend on the timeout. 493 */ 494 if ((p->p_thread->td_flags & (TDF_RUNNING|TDF_PREEMPT_LOCK|TDF_EXITING)) != TDF_EXITING) { 495 tsleep(p->p_thread, 0, "reap2", 1); 496 goto loop; 497 } 498 499 /* scheduling hook for heuristic */ 500 p->p_usched->heuristic_exiting(td->td_lwp, &p->p_lwp); 501 502 /* Take care of our return values. */ 503 *res = p->p_pid; 504 if (status) 505 *status = p->p_xstat; 506 if (rusage) 507 *rusage = *p->p_ru; 508 /* 509 * If we got the child via a ptrace 'attach', 510 * we need to give it back to the old parent. 511 */ 512 if (p->p_oppid && (t = pfind(p->p_oppid))) { 513 p->p_oppid = 0; 514 proc_reparent(p, t); 515 ksignal(t, SIGCHLD); 516 wakeup((caddr_t)t); 517 return (0); 518 } 519 p->p_xstat = 0; 520 ruadd(&q->p_stats->p_cru, p->p_ru); 521 FREE(p->p_ru, M_ZOMBIE); 522 p->p_ru = NULL; 523 524 /* 525 * Decrement the count of procs running with this uid. 526 */ 527 chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0); 528 529 /* 530 * Free up credentials. 531 */ 532 crfree(p->p_ucred); 533 p->p_ucred = NULL; 534 535 /* 536 * Remove unused arguments 537 */ 538 if (p->p_args && --p->p_args->ar_ref == 0) 539 FREE(p->p_args, M_PARGS); 540 541 /* 542 * Finally finished with old proc entry. 543 * Unlink it from its process group and free it. 544 */ 545 leavepgrp(p); 546 proc_remove_zombie(p); 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_flag & P_STOPPED) && (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(struct proc *child, struct proc *parent) 590 { 591 592 if (child->p_pptr == parent) 593 return; 594 595 LIST_REMOVE(child, p_sibling); 596 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 597 child->p_pptr = parent; 598 } 599 600 /* 601 * The next two functions are to handle adding/deleting items on the 602 * exit callout list 603 * 604 * at_exit(): 605 * Take the arguments given and put them onto the exit callout list, 606 * However first make sure that it's not already there. 607 * returns 0 on success. 608 */ 609 610 int 611 at_exit(exitlist_fn function) 612 { 613 struct exitlist *ep; 614 615 #ifdef INVARIANTS 616 /* Be noisy if the programmer has lost track of things */ 617 if (rm_at_exit(function)) 618 printf("WARNING: exit callout entry (%p) already present\n", 619 function); 620 #endif 621 ep = kmalloc(sizeof(*ep), M_ATEXIT, M_NOWAIT); 622 if (ep == NULL) 623 return (ENOMEM); 624 ep->function = function; 625 TAILQ_INSERT_TAIL(&exit_list, ep, next); 626 return (0); 627 } 628 629 /* 630 * Scan the exit callout list for the given item and remove it. 631 * Returns the number of items removed (0 or 1) 632 */ 633 int 634 rm_at_exit(exitlist_fn function) 635 { 636 struct exitlist *ep; 637 638 TAILQ_FOREACH(ep, &exit_list, next) { 639 if (ep->function == function) { 640 TAILQ_REMOVE(&exit_list, ep, next); 641 kfree(ep, M_ATEXIT); 642 return(1); 643 } 644 } 645 return (0); 646 } 647 648 void 649 check_sigacts(void) 650 { 651 struct proc *p = curproc; 652 struct sigacts *pss; 653 654 if (p->p_procsig->ps_refcnt == 1 && 655 p->p_sigacts != &p->p_addr->u_sigacts) { 656 pss = p->p_sigacts; 657 crit_enter(); 658 p->p_addr->u_sigacts = *pss; 659 p->p_sigacts = &p->p_addr->u_sigacts; 660 crit_exit(); 661 FREE(pss, M_SUBPROC); 662 } 663 } 664 665