kern_exit.c revision 27465
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 1.50 1997/07/06 02:40:40 dyson 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/sysent.h> 48#include <sys/proc.h> 49#include <sys/tty.h> 50#include <sys/time.h> 51#include <sys/resource.h> 52#include <sys/kernel.h> 53#include <sys/buf.h> 54#include <sys/wait.h> 55#include <sys/file.h> 56#include <sys/vnode.h> 57#include <sys/syslog.h> 58#include <sys/malloc.h> 59#include <sys/resourcevar.h> 60#include <sys/signalvar.h> 61#include <sys/ptrace.h> 62#include <sys/acct.h> /* for acct_process() function prototype */ 63#include <sys/filedesc.h> 64#include <sys/shm.h> 65#include <sys/sem.h> 66#include <sys/aio.h> 67 68#ifdef COMPAT_43 69#include <machine/reg.h> 70#include <machine/psl.h> 71#endif 72 73#include <vm/vm.h> 74#include <vm/vm_param.h> 75#include <vm/vm_prot.h> 76#include <sys/lock.h> 77#include <vm/pmap.h> 78#include <vm/vm_map.h> 79#include <vm/vm_kern.h> 80 81static int wait1 __P((struct proc *, struct wait_args *, int [], int)); 82 83/* 84 * callout list for things to do at exit time 85 */ 86typedef struct exit_list_element { 87 struct exit_list_element *next; 88 exitlist_fn function; 89} *ele_p; 90 91static ele_p exit_list; 92 93/* 94 * exit -- 95 * Death of process. 96 */ 97void 98exit(p, uap, retval) 99 struct proc *p; 100 struct rexit_args /* { 101 int rval; 102 } */ *uap; 103 int *retval; 104{ 105 106 exit1(p, W_EXITCODE(uap->rval, 0)); 107 /* NOTREACHED */ 108} 109 110/* 111 * Exit: deallocate address space and other resources, change proc state 112 * to zombie, and unlink proc from allproc and parent's lists. Save exit 113 * status and rusage for wait(). Check for child processes and orphan them. 114 */ 115void 116exit1(p, rv) 117 register struct proc *p; 118 int rv; 119{ 120 register struct proc *q, *nq; 121 register struct vmspace *vm; 122 ele_p ep = exit_list; 123 124 if (p->p_pid == 1) { 125 printf("init died (signal %d, exit %d)\n", 126 WTERMSIG(rv), WEXITSTATUS(rv)); 127 panic("Going nowhere without my init!"); 128 } 129 130 aio_proc_rundown(p); 131 132 /* are we a task leader? */ 133 if(p == p->p_leader) { 134 struct kill_args killArgs; 135 killArgs.signum = SIGKILL; 136 q = p->p_peers; 137 while(q) { 138 killArgs.pid = q->p_pid; 139 /* 140 * The interface for kill is better 141 * than the internal signal 142 */ 143 kill(p, &killArgs, &rv); 144 nq = q; 145 q = q->p_peers; 146 /* 147 * orphan the threads so we don't mess up 148 * when they call exit 149 */ 150 nq->p_peers = 0; 151 nq->p_leader = nq; 152 } 153 154 /* otherwise are we a peer? */ 155 } else if(p->p_peers) { 156 q = p->p_leader; 157 while(q->p_peers != p) 158 q = q->p_peers; 159 q->p_peers = p->p_peers; 160 } 161 162#ifdef PGINPROF 163 vmsizmon(); 164#endif 165 /* 166 * Check if any LKMs need anything done at process exit. 167 * e.g. SYSV IPC stuff 168 * XXX what if one of these generates an error? 169 */ 170 while (ep) { 171 (*ep->function)(p); 172 ep = ep->next; 173 } 174 175 if (p->p_flag & P_PROFIL) 176 stopprofclock(p); 177 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage), 178 M_ZOMBIE, M_WAITOK); 179 /* 180 * If parent is waiting for us to exit or exec, 181 * P_PPWAIT is set; we will wakeup the parent below. 182 */ 183 p->p_flag &= ~(P_TRACED | P_PPWAIT); 184 p->p_flag |= P_WEXIT; 185 p->p_sigignore = ~0; 186 p->p_siglist = 0; 187 untimeout(realitexpire, (caddr_t)p); 188 189 /* 190 * Close open files and release open-file table. 191 * This may block! 192 */ 193 fdfree(p); 194 195 /* 196 * XXX Shutdown SYSV semaphores 197 */ 198 semexit(p); 199 200 /* The next two chunks should probably be moved to vmspace_exit. */ 201 vm = p->p_vmspace; 202 /* 203 * Release user portion of address space. 204 * This releases references to vnodes, 205 * which could cause I/O if the file has been unlinked. 206 * Need to do this early enough that we can still sleep. 207 * Can't free the entire vmspace as the kernel stack 208 * may be mapped within that space also. 209 */ 210 if (vm->vm_refcnt == 1) { 211 if (vm->vm_shm) 212 shmexit(p); 213 pmap_remove_pages(&vm->vm_pmap, VM_MIN_ADDRESS, 214 VM_MAXUSER_ADDRESS); 215 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS, 216 VM_MAXUSER_ADDRESS); 217 } 218 219 if (SESS_LEADER(p)) { 220 register struct session *sp = p->p_session; 221 222 if (sp->s_ttyvp) { 223 /* 224 * Controlling process. 225 * Signal foreground pgrp, 226 * drain controlling terminal 227 * and revoke access to controlling terminal. 228 */ 229 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) { 230 if (sp->s_ttyp->t_pgrp) 231 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 232 (void) ttywait(sp->s_ttyp); 233 /* 234 * The tty could have been revoked 235 * if we blocked. 236 */ 237 if (sp->s_ttyvp) 238 VOP_REVOKE(sp->s_ttyvp, REVOKEALL); 239 } 240 if (sp->s_ttyvp) 241 vrele(sp->s_ttyvp); 242 sp->s_ttyvp = NULL; 243 /* 244 * s_ttyp is not zero'd; we use this to indicate 245 * that the session once had a controlling terminal. 246 * (for logging and informational purposes) 247 */ 248 } 249 sp->s_leader = NULL; 250 } 251 fixjobc(p, p->p_pgrp, 0); 252 if (p->p_limit->p_refcnt > 1 && 253 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) { 254 p->p_limit->p_refcnt--; 255 p->p_limit = limcopy(p->p_limit); 256 } 257 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 258 (void)acct_process(p); 259#ifdef KTRACE 260 /* 261 * release trace file 262 */ 263 p->p_traceflag = 0; /* don't trace the vrele() */ 264 if (p->p_tracep) 265 vrele(p->p_tracep); 266#endif 267 /* 268 * Remove proc from allproc queue and pidhash chain. 269 * Place onto zombproc. Unlink from parent's child list. 270 */ 271 LIST_REMOVE(p, p_list); 272 LIST_INSERT_HEAD(&zombproc, p, p_list); 273 p->p_stat = SZOMB; 274 275 LIST_REMOVE(p, p_hash); 276 277 q = p->p_children.lh_first; 278 if (q) /* only need this if any child is S_ZOMB */ 279 wakeup((caddr_t) initproc); 280 for (; q != 0; q = nq) { 281 nq = q->p_sibling.le_next; 282 LIST_REMOVE(q, p_sibling); 283 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling); 284 q->p_pptr = initproc; 285 /* 286 * Traced processes are killed 287 * since their existence means someone is screwing up. 288 */ 289 if (q->p_flag & P_TRACED) { 290 q->p_flag &= ~P_TRACED; 291 psignal(q, SIGKILL); 292 } 293 } 294 295 /* 296 * Save exit status and final rusage info, adding in child rusage 297 * info and self times. 298 */ 299 p->p_xstat = rv; 300 *p->p_ru = p->p_stats->p_ru; 301 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 302 ruadd(p->p_ru, &p->p_stats->p_cru); 303 304 /* 305 * Notify parent that we're gone. 306 */ 307 psignal(p->p_pptr, SIGCHLD); 308 wakeup((caddr_t)p->p_pptr); 309#if defined(tahoe) 310 /* move this to cpu_exit */ 311 p->p_addr->u_pcb.pcb_savacc.faddr = (float *)NULL; 312#endif 313 /* 314 * Clear curproc after we've done all operations 315 * that could block, and before tearing down the rest 316 * of the process state that might be used from clock, etc. 317 * Also, can't clear curproc while we're still runnable, 318 * as we're not on a run queue (we are current, just not 319 * a proper proc any longer!). 320 * 321 * Other substructures are freed from wait(). 322 */ 323 curproc = NULL; 324 if (--p->p_limit->p_refcnt == 0) { 325 FREE(p->p_limit, M_SUBPROC); 326 p->p_limit = NULL; 327 } 328 329 /* 330 * Finally, call machine-dependent code to release the remaining 331 * resources including address space, the kernel stack and pcb. 332 * The address space is released by "vmspace_free(p->p_vmspace)"; 333 * This is machine-dependent, as we may have to change stacks 334 * or ensure that the current one isn't reallocated before we 335 * finish. cpu_exit will end with a call to cpu_switch(), finishing 336 * our execution (pun intended). 337 */ 338 cpu_exit(p); 339} 340 341#ifdef COMPAT_43 342#if defined(hp300) || defined(luna68k) 343#include <machine/frame.h> 344#define GETPS(rp) ((struct frame *)(rp))->f_sr 345#else 346#define GETPS(rp) (rp)[PS] 347#endif 348 349int 350owait(p, uap, retval) 351 struct proc *p; 352 register struct owait_args /* { 353 int dummy; 354 } */ *uap; 355 int *retval; 356{ 357 struct wait_args w; 358 359#ifdef PSL_ALLCC 360 if ((GETPS(p->p_md.md_regs) & PSL_ALLCC) != PSL_ALLCC) { 361 w.options = 0; 362 w.rusage = NULL; 363 } else { 364 w.options = p->p_md.md_regs[R0]; 365 w.rusage = (struct rusage *)p->p_md.md_regs[R1]; 366 } 367#else 368 w.options = 0; 369 w.rusage = NULL; 370#endif 371 w.pid = WAIT_ANY; 372 w.status = NULL; 373 return (wait1(p, &w, retval, 1)); 374} 375#endif /* COMPAT_43 */ 376 377int 378wait4(p, uap, retval) 379 struct proc *p; 380 struct wait_args *uap; 381 int *retval; 382{ 383 384 return (wait1(p, uap, retval, 0)); 385} 386 387static int 388wait1(q, uap, retval, compat) 389 register struct proc *q; 390 register struct wait_args /* { 391 int pid; 392 int *status; 393 int options; 394 struct rusage *rusage; 395 } */ *uap; 396 int retval[]; 397 int compat; 398{ 399 register int nfound; 400 register struct proc *p, *t; 401 int status, error; 402 403 if (uap->pid == 0) 404 uap->pid = -q->p_pgid; 405#ifdef notyet 406 if (uap->options &~ (WUNTRACED|WNOHANG)) 407 return (EINVAL); 408#endif 409loop: 410 nfound = 0; 411 for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) { 412 if (uap->pid != WAIT_ANY && 413 p->p_pid != uap->pid && p->p_pgid != -uap->pid) 414 continue; 415 nfound++; 416 if (p->p_stat == SZOMB) { 417 /* charge childs scheduling cpu usage to parent */ 418 if (curproc->p_pid != 1) { 419 curproc->p_estcpu = min(curproc->p_estcpu + 420 p->p_estcpu, UCHAR_MAX); 421 } 422 423 retval[0] = p->p_pid; 424#ifdef COMPAT_43 425 if (compat) 426 retval[1] = p->p_xstat; 427 else 428#endif 429 if (uap->status) { 430 status = p->p_xstat; /* convert to int */ 431 if ((error = copyout((caddr_t)&status, 432 (caddr_t)uap->status, sizeof(status)))) 433 return (error); 434 } 435 if (uap->rusage && (error = copyout((caddr_t)p->p_ru, 436 (caddr_t)uap->rusage, sizeof (struct rusage)))) 437 return (error); 438 /* 439 * If we got the child via a ptrace 'attach', 440 * we need to give it back to the old parent. 441 */ 442 if (p->p_oppid && (t = pfind(p->p_oppid))) { 443 p->p_oppid = 0; 444 proc_reparent(p, t); 445 psignal(t, SIGCHLD); 446 wakeup((caddr_t)t); 447 return (0); 448 } 449 p->p_xstat = 0; 450 ruadd(&q->p_stats->p_cru, p->p_ru); 451 FREE(p->p_ru, M_ZOMBIE); 452 p->p_ru = NULL; 453 454 /* 455 * Decrement the count of procs running with this uid. 456 */ 457 (void)chgproccnt(p->p_cred->p_ruid, -1); 458 459 /* 460 * Release reference to text vnode 461 */ 462 if (p->p_textvp) 463 vrele(p->p_textvp); 464 465 /* 466 * Free up credentials. 467 */ 468 if (--p->p_cred->p_refcnt == 0) { 469 crfree(p->p_cred->pc_ucred); 470 FREE(p->p_cred, M_SUBPROC); 471 p->p_cred = NULL; 472 } 473 474 /* 475 * Finally finished with old proc entry. 476 * Unlink it from its process group and free it. 477 */ 478 leavepgrp(p); 479 LIST_REMOVE(p, p_list); /* off zombproc */ 480 LIST_REMOVE(p, p_sibling); 481 482 /* 483 * Give machine-dependent layer a chance 484 * to free anything that cpu_exit couldn't 485 * release while still running in process context. 486 */ 487 cpu_wait(p); 488 FREE(p, M_PROC); 489 nprocs--; 490 return (0); 491 } 492 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 493 (p->p_flag & P_TRACED || uap->options & WUNTRACED)) { 494 p->p_flag |= P_WAITED; 495 retval[0] = p->p_pid; 496#ifdef COMPAT_43 497 if (compat) { 498 retval[1] = W_STOPCODE(p->p_xstat); 499 error = 0; 500 } else 501#endif 502 if (uap->status) { 503 status = W_STOPCODE(p->p_xstat); 504 error = copyout((caddr_t)&status, 505 (caddr_t)uap->status, sizeof(status)); 506 } else 507 error = 0; 508 return (error); 509 } 510 } 511 if (nfound == 0) 512 return (ECHILD); 513 if (uap->options & WNOHANG) { 514 retval[0] = 0; 515 return (0); 516 } 517 if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0))) 518 return (error); 519 goto loop; 520} 521 522/* 523 * make process 'parent' the new parent of process 'child'. 524 */ 525void 526proc_reparent(child, parent) 527 register struct proc *child; 528 register struct proc *parent; 529{ 530 531 if (child->p_pptr == parent) 532 return; 533 534 LIST_REMOVE(child, p_sibling); 535 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 536 child->p_pptr = parent; 537} 538 539/* 540 * The next two functions are to handle adding/deleting items on the 541 * exit callout list 542 * 543 * at_exit(): 544 * Take the arguments given and put them onto the exit callout list, 545 * However first make sure that it's not already there. 546 * returns 0 on success. 547 */ 548int 549at_exit(exitlist_fn function) 550{ 551 ele_p ep; 552 553 /* Be noisy if the programmer has lost track of things */ 554 if (rm_at_exit(function)) 555 printf("exit callout entry already present\n"); 556 ep = malloc(sizeof(*ep), M_TEMP, M_NOWAIT); 557 if (ep == NULL) 558 return (ENOMEM); 559 ep->next = exit_list; 560 ep->function = function; 561 exit_list = ep; 562 return (0); 563} 564/* 565 * Scan the exit callout list for the given items and remove them. 566 * Returns the number of items removed. 567 * Logically this can only be 0 or 1. 568 */ 569int 570rm_at_exit(exitlist_fn function) 571{ 572 ele_p *epp, ep; 573 int count; 574 575 count = 0; 576 epp = &exit_list; 577 ep = *epp; 578 while (ep) { 579 if (ep->function == function) { 580 *epp = ep->next; 581 free(ep, M_TEMP); 582 count++; 583 } else { 584 epp = &ep->next; 585 } 586 ep = *epp; 587 } 588 return (count); 589} 590 591 592