1 /* $NetBSD: kern_exit.c,v 1.93 2002/03/17 23:41:30 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 1998, 1999 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility, 9 * NASA Ames Research Center. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the NetBSD 22 * Foundation, Inc. and its contributors. 23 * 4. Neither the name of The NetBSD Foundation nor the names of its 24 * contributors may be used to endorse or promote products derived 25 * from this software without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 28 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 29 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 30 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 31 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 32 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 33 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 34 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 35 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 36 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 37 * POSSIBILITY OF SUCH DAMAGE. 38 */ 39 40 /* 41 * Copyright (c) 1982, 1986, 1989, 1991, 1993 42 * The Regents of the University of California. All rights reserved. 43 * (c) UNIX System Laboratories, Inc. 44 * All or some portions of this file are derived from material licensed 45 * to the University of California by American Telephone and Telegraph 46 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 47 * the permission of UNIX System Laboratories, Inc. 48 * 49 * Redistribution and use in source and binary forms, with or without 50 * modification, are permitted provided that the following conditions 51 * are met: 52 * 1. Redistributions of source code must retain the above copyright 53 * notice, this list of conditions and the following disclaimer. 54 * 2. Redistributions in binary form must reproduce the above copyright 55 * notice, this list of conditions and the following disclaimer in the 56 * documentation and/or other materials provided with the distribution. 57 * 3. All advertising materials mentioning features or use of this software 58 * must display the following acknowledgement: 59 * This product includes software developed by the University of 60 * California, Berkeley and its contributors. 61 * 4. Neither the name of the University nor the names of its contributors 62 * may be used to endorse or promote products derived from this software 63 * without specific prior written permission. 64 * 65 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 66 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 67 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 68 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 69 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 70 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 71 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 72 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 73 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 74 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 75 * SUCH DAMAGE. 76 * 77 * @(#)kern_exit.c 8.10 (Berkeley) 2/23/95 78 */ 79 80 #include <sys/cdefs.h> 81 __KERNEL_RCSID(0, "$NetBSD: kern_exit.c,v 1.93 2002/03/17 23:41:30 christos Exp $"); 82 83 #include "opt_ktrace.h" 84 #include "opt_sysv.h" 85 86 #include <sys/param.h> 87 #include <sys/systm.h> 88 #include <sys/map.h> 89 #include <sys/ioctl.h> 90 #include <sys/proc.h> 91 #include <sys/tty.h> 92 #include <sys/time.h> 93 #include <sys/resource.h> 94 #include <sys/kernel.h> 95 #include <sys/ktrace.h> 96 #include <sys/proc.h> 97 #include <sys/buf.h> 98 #include <sys/wait.h> 99 #include <sys/file.h> 100 #include <sys/vnode.h> 101 #include <sys/syslog.h> 102 #include <sys/malloc.h> 103 #include <sys/pool.h> 104 #include <sys/resourcevar.h> 105 #include <sys/ptrace.h> 106 #include <sys/acct.h> 107 #include <sys/filedesc.h> 108 #include <sys/signalvar.h> 109 #include <sys/sched.h> 110 #include <sys/mount.h> 111 #include <sys/syscallargs.h> 112 113 #include <machine/cpu.h> 114 115 #include <uvm/uvm_extern.h> 116 117 /* 118 * exit -- 119 * Death of process. 120 */ 121 int 122 sys_exit(struct proc *p, void *v, register_t *retval) 123 { 124 struct sys_exit_args /* { 125 syscallarg(int) rval; 126 } */ *uap = v; 127 128 exit1(p, W_EXITCODE(SCARG(uap, rval), 0)); 129 /* NOTREACHED */ 130 return (0); 131 } 132 133 /* 134 * Exit: deallocate address space and other resources, change proc state 135 * to zombie, and unlink proc from allproc and parent's lists. Save exit 136 * status and rusage for wait(). Check for child processes and orphan them. 137 */ 138 void 139 exit1(struct proc *p, int rv) 140 { 141 struct proc *q, *nq; 142 int s; 143 144 if (__predict_false(p == initproc)) 145 panic("init died (signal %d, exit %d)", 146 WTERMSIG(rv), WEXITSTATUS(rv)); 147 148 #ifdef PGINPROF 149 vmsizmon(); 150 #endif 151 if (p->p_flag & P_PROFIL) 152 stopprofclock(p); 153 p->p_ru = pool_get(&rusage_pool, PR_WAITOK); 154 /* 155 * If parent is waiting for us to exit or exec, P_PPWAIT is set; we 156 * wake up the parent early to avoid deadlock. 157 */ 158 p->p_flag |= P_WEXIT; 159 if (p->p_flag & P_PPWAIT) { 160 p->p_flag &= ~P_PPWAIT; 161 wakeup((caddr_t)p->p_pptr); 162 } 163 sigfillset(&p->p_sigctx.ps_sigignore); 164 sigemptyset(&p->p_sigctx.ps_siglist); 165 p->p_sigctx.ps_sigcheck = 0; 166 callout_stop(&p->p_realit_ch); 167 168 /* 169 * Close open files and release open-file table. 170 * This may block! 171 */ 172 fdfree(p); 173 cwdfree(p); 174 175 doexithooks(p); 176 177 if (SESS_LEADER(p)) { 178 struct session *sp = p->p_session; 179 180 if (sp->s_ttyvp) { 181 /* 182 * Controlling process. 183 * Signal foreground pgrp, 184 * drain controlling terminal 185 * and revoke access to controlling terminal. 186 */ 187 if (sp->s_ttyp->t_session == sp) { 188 if (sp->s_ttyp->t_pgrp) 189 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1); 190 (void) ttywait(sp->s_ttyp); 191 /* 192 * The tty could have been revoked 193 * if we blocked. 194 */ 195 if (sp->s_ttyvp) 196 VOP_REVOKE(sp->s_ttyvp, REVOKEALL); 197 } 198 if (sp->s_ttyvp) 199 vrele(sp->s_ttyvp); 200 sp->s_ttyvp = NULL; 201 /* 202 * s_ttyp is not zero'd; we use this to indicate 203 * that the session once had a controlling terminal. 204 * (for logging and informational purposes) 205 */ 206 } 207 sp->s_leader = NULL; 208 } 209 fixjobc(p, p->p_pgrp, 0); 210 (void)acct_process(p); 211 #ifdef KTRACE 212 /* 213 * release trace file 214 */ 215 ktrderef(p); 216 #endif 217 /* 218 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP! 219 */ 220 p->p_stat = SDEAD; 221 222 /* 223 * Remove proc from pidhash chain so looking it up won't 224 * work. Move it from allproc to zombproc, but do not yet 225 * wake up the reaper. We will put the proc on the 226 * deadproc list later (using the p_hash member), and 227 * wake up the reaper when we do. 228 */ 229 s = proclist_lock_write(); 230 LIST_REMOVE(p, p_hash); 231 LIST_REMOVE(p, p_list); 232 LIST_INSERT_HEAD(&zombproc, p, p_list); 233 proclist_unlock_write(s); 234 235 /* 236 * Give orphaned children to init(8). 237 */ 238 q = p->p_children.lh_first; 239 if (q) /* only need this if any child is S_ZOMB */ 240 wakeup((caddr_t)initproc); 241 for (; q != 0; q = nq) { 242 nq = q->p_sibling.le_next; 243 proc_reparent(q, initproc); 244 /* 245 * Traced processes are killed 246 * since their existence means someone is screwing up. 247 */ 248 if (q->p_flag & P_TRACED) { 249 q->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE); 250 psignal(q, SIGKILL); 251 } 252 } 253 254 /* 255 * Save exit status and final rusage info, adding in child rusage 256 * info and self times. 257 */ 258 p->p_xstat = rv; 259 *p->p_ru = p->p_stats->p_ru; 260 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL); 261 ruadd(p->p_ru, &p->p_stats->p_cru); 262 263 /* 264 * Notify parent that we're gone. If parent has the P_NOCLDWAIT 265 * flag set, notify init instead (and hope it will handle 266 * this situation). 267 */ 268 if (p->p_pptr->p_flag & P_NOCLDWAIT) { 269 struct proc *pp = p->p_pptr; 270 proc_reparent(p, initproc); 271 /* 272 * If this was the last child of our parent, notify 273 * parent, so in case he was wait(2)ing, he will 274 * continue. 275 */ 276 if (pp->p_children.lh_first == NULL) 277 wakeup((caddr_t)pp); 278 } 279 280 /* 281 * Release the process's signal state. 282 */ 283 sigactsfree(p); 284 285 /* 286 * Clear curproc after we've done all operations 287 * that could block, and before tearing down the rest 288 * of the process state that might be used from clock, etc. 289 * Also, can't clear curproc while we're still runnable, 290 * as we're not on a run queue (we are current, just not 291 * a proper proc any longer!). 292 * 293 * Other substructures are freed from wait(). 294 */ 295 curproc = NULL; 296 limfree(p->p_limit); 297 p->p_limit = NULL; 298 299 /* 300 * If emulation has process exit hook, call it now. 301 */ 302 if (p->p_emul->e_proc_exit) 303 (*p->p_emul->e_proc_exit)(p); 304 305 /* This process no longer needs to hold the kernel lock. */ 306 KERNEL_PROC_UNLOCK(p); 307 308 /* 309 * Finally, call machine-dependent code to switch to a new 310 * context (possibly the idle context). Once we are no longer 311 * using the dead process's vmspace and stack, exit2() will be 312 * called to schedule those resources to be released by the 313 * reaper thread. 314 * 315 * Note that cpu_exit() will end with a call equivalent to 316 * cpu_switch(), finishing our execution (pun intended). 317 */ 318 cpu_exit(p); 319 } 320 321 /* 322 * We are called from cpu_exit() once it is safe to schedule the 323 * dead process's resources to be freed (i.e., once we've switched to 324 * the idle PCB for the current CPU). 325 * 326 * NOTE: One must be careful with locking in this routine. It's 327 * called from a critical section in machine-dependent code, so 328 * we should refrain from changing any interrupt state. 329 * 330 * We lock the deadproc list (a spin lock), place the proc on that 331 * list (using the p_hash member), and wake up the reaper. 332 */ 333 void 334 exit2(struct proc *p) 335 { 336 337 simple_lock(&deadproc_slock); 338 LIST_INSERT_HEAD(&deadproc, p, p_hash); 339 simple_unlock(&deadproc_slock); 340 341 wakeup(&deadproc); 342 } 343 344 /* 345 * Process reaper. This is run by a kernel thread to free the resources 346 * of a dead process. Once the resources are free, the process becomes 347 * a zombie, and the parent is allowed to read the undead's status. 348 */ 349 void 350 reaper(void *arg) 351 { 352 struct proc *p; 353 354 KERNEL_PROC_UNLOCK(curproc); 355 356 for (;;) { 357 simple_lock(&deadproc_slock); 358 p = LIST_FIRST(&deadproc); 359 if (p == NULL) { 360 /* No work for us; go to sleep until someone exits. */ 361 (void) ltsleep(&deadproc, PVM|PNORELOCK, 362 "reaper", 0, &deadproc_slock); 363 continue; 364 } 365 366 /* Remove us from the deadproc list. */ 367 LIST_REMOVE(p, p_hash); 368 simple_unlock(&deadproc_slock); 369 KERNEL_PROC_LOCK(curproc); 370 371 /* 372 * Give machine-dependent code a chance to free any 373 * resources it couldn't free while still running on 374 * that process's context. This must be done before 375 * uvm_exit(), in case these resources are in the PCB. 376 */ 377 cpu_wait(p); 378 379 /* 380 * Free the VM resources we're still holding on to. 381 * We must do this from a valid thread because doing 382 * so may block. 383 */ 384 uvm_exit(p); 385 386 /* Process is now a true zombie. */ 387 p->p_stat = SZOMB; 388 389 /* Wake up the parent so it can get exit status. */ 390 if ((p->p_flag & P_FSTRACE) == 0 && p->p_exitsig != 0) 391 psignal(p->p_pptr, P_EXITSIG(p)); 392 KERNEL_PROC_UNLOCK(curproc); 393 wakeup((caddr_t)p->p_pptr); 394 } 395 } 396 397 int 398 sys_wait4(struct proc *q, void *v, register_t *retval) 399 { 400 struct sys_wait4_args /* { 401 syscallarg(int) pid; 402 syscallarg(int *) status; 403 syscallarg(int) options; 404 syscallarg(struct rusage *) rusage; 405 } */ *uap = v; 406 struct proc *p, *t; 407 int nfound, status, error, s; 408 409 if (SCARG(uap, pid) == 0) 410 SCARG(uap, pid) = -q->p_pgid; 411 if (SCARG(uap, options) &~ (WUNTRACED|WNOHANG|WALTSIG)) 412 return (EINVAL); 413 414 loop: 415 nfound = 0; 416 for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) { 417 if (SCARG(uap, pid) != WAIT_ANY && 418 p->p_pid != SCARG(uap, pid) && 419 p->p_pgid != -SCARG(uap, pid)) 420 continue; 421 /* 422 * Wait for processes with p_exitsig != SIGCHLD processes only 423 * if WALTSIG is set; wait for processes with p_exitsig == 424 * SIGCHLD only if WALTSIG is clear. 425 */ 426 if (((SCARG(uap, options) & WALLSIG) == 0) && 427 ((SCARG(uap, options) & WALTSIG) ? 428 (p->p_exitsig == SIGCHLD) : (P_EXITSIG(p) != SIGCHLD))) 429 continue; 430 431 nfound++; 432 if (p->p_stat == SZOMB) { 433 retval[0] = p->p_pid; 434 435 if (SCARG(uap, status)) { 436 status = p->p_xstat; /* convert to int */ 437 error = copyout((caddr_t)&status, 438 (caddr_t)SCARG(uap, status), 439 sizeof(status)); 440 if (error) 441 return (error); 442 } 443 if (SCARG(uap, rusage) && 444 (error = copyout((caddr_t)p->p_ru, 445 (caddr_t)SCARG(uap, rusage), 446 sizeof(struct rusage)))) 447 return (error); 448 /* 449 * If we got the child via ptrace(2) or procfs, and 450 * the parent is different (meaning the process was 451 * attached, rather than run as a child), then we need 452 * to give it back to the old parent, and send the 453 * parent the exit signal. The rest of the cleanup 454 * will be done when the old parent waits on the child. 455 */ 456 if ((p->p_flag & P_TRACED) && 457 p->p_oppid != p->p_pptr->p_pid) { 458 t = pfind(p->p_oppid); 459 proc_reparent(p, t ? t : initproc); 460 p->p_oppid = 0; 461 p->p_flag &= ~(P_TRACED|P_WAITED|P_FSTRACE); 462 if (p->p_exitsig != 0) 463 psignal(p->p_pptr, P_EXITSIG(p)); 464 wakeup((caddr_t)p->p_pptr); 465 return (0); 466 } 467 scheduler_wait_hook(q, p); 468 p->p_xstat = 0; 469 ruadd(&q->p_stats->p_cru, p->p_ru); 470 pool_put(&rusage_pool, p->p_ru); 471 472 /* 473 * Finally finished with old proc entry. 474 * Unlink it from its process group and free it. 475 */ 476 leavepgrp(p); 477 478 s = proclist_lock_write(); 479 LIST_REMOVE(p, p_list); /* off zombproc */ 480 proclist_unlock_write(s); 481 482 LIST_REMOVE(p, p_sibling); 483 484 /* 485 * Decrement the count of procs running with this uid. 486 */ 487 (void)chgproccnt(p->p_cred->p_ruid, -1); 488 489 /* 490 * Free up credentials. 491 */ 492 if (--p->p_cred->p_refcnt == 0) { 493 crfree(p->p_cred->pc_ucred); 494 pool_put(&pcred_pool, p->p_cred); 495 } 496 497 /* 498 * Release reference to text vnode 499 */ 500 if (p->p_textvp) 501 vrele(p->p_textvp); 502 503 pool_put(&proc_pool, p); 504 nprocs--; 505 return (0); 506 } 507 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 && 508 (p->p_flag & P_TRACED || SCARG(uap, options) & WUNTRACED)) { 509 p->p_flag |= P_WAITED; 510 retval[0] = p->p_pid; 511 512 if (SCARG(uap, status)) { 513 status = W_STOPCODE(p->p_xstat); 514 error = copyout((caddr_t)&status, 515 (caddr_t)SCARG(uap, status), 516 sizeof(status)); 517 } else 518 error = 0; 519 return (error); 520 } 521 } 522 if (nfound == 0) 523 return (ECHILD); 524 if (SCARG(uap, options) & WNOHANG) { 525 retval[0] = 0; 526 return (0); 527 } 528 if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0)) != 0) 529 return (error); 530 goto loop; 531 } 532 533 /* 534 * make process 'parent' the new parent of process 'child'. 535 */ 536 void 537 proc_reparent(struct proc *child, struct proc *parent) 538 { 539 540 if (child->p_pptr == parent) 541 return; 542 543 if (parent == initproc) 544 child->p_exitsig = SIGCHLD; 545 546 LIST_REMOVE(child, p_sibling); 547 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); 548 child->p_pptr = parent; 549 } 550