kern_exit.c revision 31610
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 * $Id: kern_exit.c,v 1.61 1997/12/06 04:11:10 sef Exp $ 40 */ 41 42#include "opt_ktrace.h" 43 44#include <sys/param.h> 45#include <sys/systm.h> 46#include <sys/sysproto.h> 47#include <sys/malloc.h> 48#include <sys/proc.h> 49#include <sys/pioctl.h> 50#include <sys/tty.h> 51#include <sys/wait.h> 52#include <sys/vnode.h> 53#include <sys/resourcevar.h> 54#include <sys/signalvar.h> 55#include <sys/ptrace.h> 56#include <sys/acct.h> /* for acct_process() function prototype */ 57#include <sys/filedesc.h> 58#include <sys/shm.h> 59#include <sys/sem.h> 60#include <sys/aio.h> 61 62#ifdef COMPAT_43 63#include <machine/reg.h> 64#include <machine/psl.h> 65#endif 66#include <machine/limits.h> /* for UCHAR_MAX = typeof(p_priority)_MAX */ 67 68#include <vm/vm.h> 69#include <vm/vm_param.h> 70#include <sys/lock.h> 71#include <vm/pmap.h> 72#include <vm/vm_map.h> 73 74static MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status"); 75 76static int wait1 __P((struct proc *, struct wait_args *, int)); 77 78/* 79 * callout list for things to do at exit time 80 */ 81typedef struct exit_list_element { 82 struct exit_list_element *next; 83 exitlist_fn function; 84} *ele_p; 85 86static ele_p exit_list; 87 88/* 89 * exit -- 90 * Death of process. 91 */ 92void 93exit(p, uap) 94 struct proc *p; 95 struct rexit_args /* { 96 int rval; 97 } */ *uap; 98{ 99 100 exit1(p, W_EXITCODE(uap->rval, 0)); 101 /* NOTREACHED */ 102} 103 104/* 105 * Exit: deallocate address space and other resources, change proc state 106 * to zombie, and unlink proc from allproc and parent's lists. Save exit 107 * status and rusage for wait(). Check for child processes and orphan them. 108 */ 109void 110exit1(p, rv) 111 register struct proc *p; 112 int rv; 113{ 114 register struct proc *q, *nq; 115 register struct vmspace *vm; 116 ele_p ep = exit_list; 117#ifdef PROCFS 118 extern void procfs_exit(pid_t); 119#endif 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 aio_proc_rundown(p); 128 129 /* are we a task leader? */ 130 if(p == p->p_leader) { 131 struct kill_args killArgs; 132 killArgs.signum = SIGKILL; 133 q = p->p_peers; 134 while(q) { 135 killArgs.pid = q->p_pid; 136 /* 137 * The interface for kill is better 138 * than the internal signal 139 */ 140 kill(p, &killArgs); 141 nq = q; 142 q = q->p_peers; 143 /* 144 * orphan the threads so we don't mess up 145 * when they call exit 146 */ 147 nq->p_peers = 0; 148 nq->p_leader = nq; 149 } 150 151 /* otherwise are we a peer? */ 152 } else if(p->p_peers) { 153 q = p->p_leader; 154 while(q->p_peers != p) 155 q = q->p_peers; 156 q->p_peers = p->p_peers; 157 } 158 159#ifdef PGINPROF 160 vmsizmon(); 161#endif 162 STOPEVENT(p, S_EXIT, rv); 163 164#ifdef PROCFS 165 /* 166 * Now that we're back from stopevent(), force a close 167 * of all open procfs files for this process. 168 */ 169 procfs_exit(p->p_pid); 170#endif 171 172 /* 173 * Check if any LKMs need anything done at process exit. 174 * e.g. SYSV IPC stuff 175 * XXX what if one of these generates an error? 176 */ 177 while (ep) { 178 (*ep->function)(p); 179 ep = ep->next; 180 } 181 182 if (p->p_flag & P_PROFIL) 183 stopprofclock(p); 184 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage), 185 M_ZOMBIE, M_WAITOK); 186 /* 187 * If parent is waiting for us to exit or exec, 188 * P_PPWAIT is set; we will wakeup the parent below. 189 */ 190 p->p_flag &= ~(P_TRACED | P_PPWAIT); 191 p->p_flag |= P_WEXIT; 192 p->p_sigignore = ~0; 193 p->p_siglist = 0; 194 if (timerisset(&p->p_realtimer.it_value)) 195 untimeout(realitexpire, (caddr_t)p, p->p_ithandle); 196 197 /* 198 * Close open files and release open-file table. 199 * This may block! 200 */ 201 fdfree(p); 202 203 /* 204 * XXX Shutdown SYSV semaphores 205 */ 206 semexit(p); 207 208 /* The next two chunks should probably be moved to vmspace_exit. */ 209 vm = p->p_vmspace; 210 /* 211 * Release user portion of address space. 212 * This releases references to vnodes, 213 * which could cause I/O if the file has been unlinked. 214 * Need to do this early enough that we can still sleep. 215 * Can't free the entire vmspace as the kernel stack 216 * may be mapped within that space also. 217 */ 218 if (vm->vm_refcnt == 1) { 219 if (vm->vm_shm) 220 shmexit(p); 221 pmap_remove_pages(&vm->vm_pmap, VM_MIN_ADDRESS, 222 VM_MAXUSER_ADDRESS); 223 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS, 224 VM_MAXUSER_ADDRESS); 225 } 226 227 if (SESS_LEADER(p)) { 228 register struct session *sp = p->p_session; 229 230 if (sp->s_ttyvp) { 231 /* 232 * Controlling process. 233 * Signal foreground pgrp, 234 * drain controlling terminal 235 * and revoke access to controlling terminal. 236 */ 237 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 238 if (sp->s_ttyp->t_pgrp) 239 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 240 (void) ttywait(sp->s_ttyp); 241 /* 242 * The tty could have been revoked 243 * if we blocked. 244 */ 245 if (sp->s_ttyvp) 246 VOP_REVOKE(sp->s_ttyvp, REVOKEALL); 247 } 248 if (sp->s_ttyvp) 249 vrele(sp->s_ttyvp); 250 sp->s_ttyvp = NULL; 251 /* 252 * s_ttyp is not zero'd; we use this to indicate 253 * that the session once had a controlling terminal. 254 * (for logging and informational purposes) 255 */ 256 } 257 sp->s_leader = NULL; 258 } 259 fixjobc(p, p->p_pgrp, 0); 260 if (p->p_limit->p_refcnt > 1 && 261 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) { 262 p->p_limit->p_refcnt--; 263 p->p_limit = limcopy(p->p_limit); 264 } 265 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 266 (void)acct_process(p); 267#ifdef KTRACE 268 /* 269 * release trace file 270 */ 271 p->p_traceflag = 0; /* don't trace the vrele() */ 272 if (p->p_tracep) 273 vrele(p->p_tracep); 274#endif 275 /* 276 * Remove proc from allproc queue and pidhash chain. 277 * Place onto zombproc. Unlink from parent's child list. 278 */ 279 LIST_REMOVE(p, p_list); 280 LIST_INSERT_HEAD(&zombproc, p, p_list); 281 p->p_stat = SZOMB; 282 283 LIST_REMOVE(p, p_hash); 284 285 q = p->p_children.lh_first; 286 if (q) /* only need this if any child is S_ZOMB */ 287 wakeup((caddr_t) initproc); 288 for (; q != 0; q = nq) { 289 nq = q->p_sibling.le_next; 290 LIST_REMOVE(q, p_sibling); 291 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling); 292 q->p_pptr = initproc; 293 /* 294 * Traced processes are killed 295 * since their existence means someone is screwing up. 296 */ 297 if (q->p_flag & P_TRACED) { 298 q->p_flag &= ~P_TRACED; 299 psignal(q, SIGKILL); 300 } 301 } 302 303 /* 304 * Save exit status and final rusage info, adding in child rusage 305 * info and self times. 306 */ 307 p->p_xstat = rv; 308 *p->p_ru = p->p_stats->p_ru; 309 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 310 ruadd(p->p_ru, &p->p_stats->p_cru); 311 312 /* 313 * Notify parent that we're gone. If parent has the P_NOCLDWAIT 314 * flag set, notify process 1 instead (and hope it will handle 315 * this situation). 316 */ 317 if (p->p_pptr->p_flag & P_NOCLDWAIT) { 318 struct proc *pp = p->p_pptr; 319 proc_reparent(p, initproc); 320 /* 321 * If this was the last child of our parent, notify 322 * parent, so in case he was wait(2)ing, he will 323 * continue. 324 */ 325 if (LIST_EMPTY(&pp->p_children)) 326 wakeup((caddr_t)pp); 327 } 328 329 psignal(p->p_pptr, SIGCHLD); 330 wakeup((caddr_t)p->p_pptr); 331#if defined(tahoe) 332 /* move this to cpu_exit */ 333 p->p_addr->u_pcb.pcb_savacc.faddr = (float *)NULL; 334#endif 335 /* 336 * Clear curproc after we've done all operations 337 * that could block, and before tearing down the rest 338 * of the process state that might be used from clock, etc. 339 * Also, can't clear curproc while we're still runnable, 340 * as we're not on a run queue (we are current, just not 341 * a proper proc any longer!). 342 * 343 * Other substructures are freed from wait(). 344 */ 345 curproc = NULL; 346 if (--p->p_limit->p_refcnt == 0) { 347 FREE(p->p_limit, M_SUBPROC); 348 p->p_limit = NULL; 349 } 350 351 /* 352 * Finally, call machine-dependent code to release the remaining 353 * resources including address space, the kernel stack and pcb. 354 * The address space is released by "vmspace_free(p->p_vmspace)"; 355 * This is machine-dependent, as we may have to change stacks 356 * or ensure that the current one isn't reallocated before we 357 * finish. cpu_exit will end with a call to cpu_switch(), finishing 358 * our execution (pun intended). 359 */ 360 cpu_exit(p); 361} 362 363#ifdef COMPAT_43 364#if defined(hp300) || defined(luna68k) 365#include <machine/frame.h> 366#define GETPS(rp) ((struct frame *)(rp))->f_sr 367#else 368#define GETPS(rp) (rp)[PS] 369#endif 370 371int 372owait(p, uap) 373 struct proc *p; 374 register struct owait_args /* { 375 int dummy; 376 } */ *uap; 377{ 378 struct wait_args w; 379 380#ifdef PSL_ALLCC 381 if ((GETPS(p->p_md.md_regs) & PSL_ALLCC) != PSL_ALLCC) { 382 w.options = 0; 383 w.rusage = NULL; 384 } else { 385 w.options = p->p_md.md_regs[R0]; 386 w.rusage = (struct rusage *)p->p_md.md_regs[R1]; 387 } 388#else 389 w.options = 0; 390 w.rusage = NULL; 391#endif 392 w.pid = WAIT_ANY; 393 w.status = NULL; 394 return (wait1(p, &w, 1)); 395} 396#endif /* COMPAT_43 */ 397 398int 399wait4(p, uap) 400 struct proc *p; 401 struct wait_args *uap; 402{ 403 404 return (wait1(p, uap, 0)); 405} 406 407static int 408wait1(q, uap, compat) 409 register struct proc *q; 410 register struct wait_args /* { 411 int pid; 412 int *status; 413 int options; 414 struct rusage *rusage; 415 } */ *uap; 416 int compat; 417{ 418 register int nfound; 419 register struct proc *p, *t; 420 int status, error; 421 422 if (uap->pid == 0) 423 uap->pid = -q->p_pgid; 424 if (uap->options &~ (WUNTRACED|WNOHANG)) 425 return (EINVAL); 426loop: 427 nfound = 0; 428 for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) { 429 if (uap->pid != WAIT_ANY && 430 p->p_pid != uap->pid && p->p_pgid != -uap->pid) 431 continue; 432 nfound++; 433 if (p->p_stat == SZOMB) { 434 /* charge childs scheduling cpu usage to parent */ 435 if (curproc->p_pid != 1) { 436 curproc->p_estcpu = min(curproc->p_estcpu + 437 p->p_estcpu, UCHAR_MAX); 438 } 439 440 q->p_retval[0] = p->p_pid; 441#ifdef COMPAT_43 442 if (compat) 443 q->p_retval[1] = p->p_xstat; 444 else 445#endif 446 if (uap->status) { 447 status = p->p_xstat; /* convert to int */ 448 if ((error = copyout((caddr_t)&status, 449 (caddr_t)uap->status, sizeof(status)))) 450 return (error); 451 } 452 if (uap->rusage && (error = copyout((caddr_t)p->p_ru, 453 (caddr_t)uap->rusage, sizeof (struct rusage)))) 454 return (error); 455 /* 456 * If we got the child via a ptrace 'attach', 457 * we need to give it back to the old parent. 458 */ 459 if (p->p_oppid && (t = pfind(p->p_oppid))) { 460 p->p_oppid = 0; 461 proc_reparent(p, t); 462 psignal(t, SIGCHLD); 463 wakeup((caddr_t)t); 464 return (0); 465 } 466 p->p_xstat = 0; 467 ruadd(&q->p_stats->p_cru, p->p_ru); 468 FREE(p->p_ru, M_ZOMBIE); 469 p->p_ru = NULL; 470 471 /* 472 * Decrement the count of procs running with this uid. 473 */ 474 (void)chgproccnt(p->p_cred->p_ruid, -1); 475 476 /* 477 * Release reference to text vnode 478 */ 479 if (p->p_textvp) 480 vrele(p->p_textvp); 481 482 /* 483 * Free up credentials. 484 */ 485 if (--p->p_cred->p_refcnt == 0) { 486 crfree(p->p_cred->pc_ucred); 487 FREE(p->p_cred, M_SUBPROC); 488 p->p_cred = NULL; 489 } 490 491 /* 492 * Finally finished with old proc entry. 493 * Unlink it from its process group and free it. 494 */ 495 leavepgrp(p); 496 LIST_REMOVE(p, p_list); /* off zombproc */ 497 LIST_REMOVE(p, p_sibling); 498 499 /* 500 * Give machine-dependent layer a chance 501 * to free anything that cpu_exit couldn't 502 * release while still running in process context. 503 */ 504 cpu_wait(p); 505 FREE(p, M_PROC); 506 nprocs--; 507 return (0); 508 } 509 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 510 (p->p_flag & P_TRACED || uap->options & WUNTRACED)) { 511 p->p_flag |= P_WAITED; 512 q->p_retval[0] = p->p_pid; 513#ifdef COMPAT_43 514 if (compat) { 515 q->p_retval[1] = W_STOPCODE(p->p_xstat); 516 error = 0; 517 } else 518#endif 519 if (uap->status) { 520 status = W_STOPCODE(p->p_xstat); 521 error = copyout((caddr_t)&status, 522 (caddr_t)uap->status, sizeof(status)); 523 } else 524 error = 0; 525 return (error); 526 } 527 } 528 if (nfound == 0) 529 return (ECHILD); 530 if (uap->options & WNOHANG) { 531 q->p_retval[0] = 0; 532 return (0); 533 } 534 if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0))) 535 return (error); 536 goto loop; 537} 538 539/* 540 * make process 'parent' the new parent of process 'child'. 541 */ 542void 543proc_reparent(child, parent) 544 register struct proc *child; 545 register struct proc *parent; 546{ 547 548 if (child->p_pptr == parent) 549 return; 550 551 LIST_REMOVE(child, p_sibling); 552 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 553 child->p_pptr = parent; 554} 555 556/* 557 * The next two functions are to handle adding/deleting items on the 558 * exit callout list 559 * 560 * at_exit(): 561 * Take the arguments given and put them onto the exit callout list, 562 * However first make sure that it's not already there. 563 * returns 0 on success. 564 */ 565int 566at_exit(function) 567 exitlist_fn function; 568{ 569 ele_p ep; 570 571 /* Be noisy if the programmer has lost track of things */ 572 if (rm_at_exit(function)) 573 printf("exit callout entry already present\n"); 574 ep = malloc(sizeof(*ep), M_TEMP, M_NOWAIT); 575 if (ep == NULL) 576 return (ENOMEM); 577 ep->next = exit_list; 578 ep->function = function; 579 exit_list = ep; 580 return (0); 581} 582/* 583 * Scan the exit callout list for the given items and remove them. 584 * Returns the number of items removed. 585 * Logically this can only be 0 or 1. 586 */ 587int 588rm_at_exit(function) 589 exitlist_fn function; 590{ 591 ele_p *epp, ep; 592 int count; 593 594 count = 0; 595 epp = &exit_list; 596 ep = *epp; 597 while (ep) { 598 if (ep->function == function) { 599 *epp = ep->next; 600 free(ep, M_TEMP); 601 count++; 602 } else { 603 epp = &ep->next; 604 } 605 ep = *epp; 606 } 607 return (count); 608} 609