1 /* $OpenBSD: kern_fork.c,v 1.103 2009/04/14 09:13:25 art Exp $ */ 2 /* $NetBSD: kern_fork.c,v 1.29 1996/02/09 18:59:34 christos Exp $ */ 3 4 /* 5 * Copyright (c) 1982, 1986, 1989, 1991, 1993 6 * The Regents of the University of California. All rights reserved. 7 * (c) UNIX System Laboratories, Inc. 8 * All or some portions of this file are derived from material licensed 9 * to the University of California by American Telephone and Telegraph 10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 11 * the permission of UNIX System Laboratories, Inc. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 3. Neither the name of the University nor the names of its contributors 22 * may be used to endorse or promote products derived from this software 23 * without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35 * SUCH DAMAGE. 36 * 37 * @(#)kern_fork.c 8.6 (Berkeley) 4/8/94 38 */ 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/filedesc.h> 43 #include <sys/kernel.h> 44 #include <sys/malloc.h> 45 #include <sys/mount.h> 46 #include <sys/proc.h> 47 #include <sys/exec.h> 48 #include <sys/resourcevar.h> 49 #include <sys/signalvar.h> 50 #include <sys/vnode.h> 51 #include <sys/file.h> 52 #include <sys/acct.h> 53 #include <sys/ktrace.h> 54 #include <sys/sched.h> 55 #include <dev/rndvar.h> 56 #include <sys/pool.h> 57 #include <sys/mman.h> 58 #include <sys/ptrace.h> 59 60 #include <sys/syscallargs.h> 61 62 #include "systrace.h" 63 #include <dev/systrace.h> 64 65 #include <uvm/uvm_extern.h> 66 #include <uvm/uvm_map.h> 67 68 int nprocs = 1; /* process 0 */ 69 int randompid; /* when set to 1, pid's go random */ 70 pid_t lastpid; 71 struct forkstat forkstat; 72 73 void fork_return(void *); 74 int pidtaken(pid_t); 75 76 void process_new(struct proc *, struct proc *); 77 78 void 79 fork_return(void *arg) 80 { 81 struct proc *p = (struct proc *)arg; 82 83 if (p->p_flag & P_TRACED) 84 psignal(p, SIGTRAP); 85 86 child_return(p); 87 } 88 89 /*ARGSUSED*/ 90 int 91 sys_fork(struct proc *p, void *v, register_t *retval) 92 { 93 int flags; 94 95 flags = FORK_FORK; 96 if (p->p_ptmask & PTRACE_FORK) 97 flags |= FORK_PTRACE; 98 return (fork1(p, SIGCHLD, flags, NULL, 0, 99 fork_return, NULL, retval, NULL)); 100 } 101 102 /*ARGSUSED*/ 103 int 104 sys_vfork(struct proc *p, void *v, register_t *retval) 105 { 106 return (fork1(p, SIGCHLD, FORK_VFORK|FORK_PPWAIT, NULL, 0, NULL, 107 NULL, retval, NULL)); 108 } 109 110 int 111 sys_rfork(struct proc *p, void *v, register_t *retval) 112 { 113 struct sys_rfork_args /* { 114 syscallarg(int) flags; 115 } */ *uap = v; 116 117 int rforkflags; 118 int flags; 119 120 flags = FORK_RFORK; 121 rforkflags = SCARG(uap, flags); 122 123 if ((rforkflags & RFPROC) == 0) 124 return (EINVAL); 125 126 switch(rforkflags & (RFFDG|RFCFDG)) { 127 case (RFFDG|RFCFDG): 128 return EINVAL; 129 case RFCFDG: 130 flags |= FORK_CLEANFILES; 131 break; 132 case RFFDG: 133 break; 134 default: 135 flags |= FORK_SHAREFILES; 136 break; 137 } 138 139 if (rforkflags & RFNOWAIT) 140 flags |= FORK_NOZOMBIE; 141 142 if (rforkflags & RFMEM) 143 flags |= FORK_SHAREVM; 144 #ifdef RTHREADS 145 if (rforkflags & RFTHREAD) 146 flags |= FORK_THREAD | FORK_SIGHAND; 147 #endif 148 149 return (fork1(p, SIGCHLD, flags, NULL, 0, NULL, NULL, retval, NULL)); 150 } 151 152 /* 153 * Allocate and initialize a new process. 154 */ 155 void 156 process_new(struct proc *newproc, struct proc *parent) 157 { 158 struct process *pr; 159 160 pr = pool_get(&process_pool, PR_WAITOK); 161 pr->ps_mainproc = newproc; 162 TAILQ_INIT(&pr->ps_threads); 163 TAILQ_INSERT_TAIL(&pr->ps_threads, newproc, p_thr_link); 164 pr->ps_refcnt = 1; 165 newproc->p_p = pr; 166 } 167 168 /* print the 'table full' message once per 10 seconds */ 169 struct timeval fork_tfmrate = { 10, 0 }; 170 171 int 172 fork1(struct proc *p1, int exitsig, int flags, void *stack, size_t stacksize, 173 void (*func)(void *), void *arg, register_t *retval, 174 struct proc **rnewprocp) 175 { 176 struct proc *p2; 177 uid_t uid; 178 struct vmspace *vm; 179 int count; 180 vaddr_t uaddr; 181 int s; 182 extern void endtsleep(void *); 183 extern void realitexpire(void *); 184 struct ptrace_state *newptstat = NULL; 185 #if NSYSTRACE > 0 186 void *newstrp = NULL; 187 #endif 188 189 /* 190 * Although process entries are dynamically created, we still keep 191 * a global limit on the maximum number we will create. We reserve 192 * the last 5 processes to root. The variable nprocs is the current 193 * number of processes, maxproc is the limit. 194 */ 195 uid = p1->p_cred->p_ruid; 196 if ((nprocs >= maxproc - 5 && uid != 0) || nprocs >= maxproc) { 197 static struct timeval lasttfm; 198 199 if (ratecheck(&lasttfm, &fork_tfmrate)) 200 tablefull("proc"); 201 return (EAGAIN); 202 } 203 nprocs++; 204 205 /* 206 * Increment the count of procs running with this uid. Don't allow 207 * a nonprivileged user to exceed their current limit. 208 */ 209 count = chgproccnt(uid, 1); 210 if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) { 211 (void)chgproccnt(uid, -1); 212 nprocs--; 213 return (EAGAIN); 214 } 215 216 uaddr = uvm_km_alloc1(kernel_map, USPACE, USPACE_ALIGN, 1); 217 if (uaddr == 0) { 218 chgproccnt(uid, -1); 219 nprocs--; 220 return (ENOMEM); 221 } 222 223 /* 224 * From now on, we're committed to the fork and cannot fail. 225 */ 226 227 /* Allocate new proc. */ 228 p2 = pool_get(&proc_pool, PR_WAITOK); 229 230 p2->p_stat = SIDL; /* protect against others */ 231 p2->p_exitsig = exitsig; 232 p2->p_flag = 0; 233 234 #ifdef RTHREADS 235 if (flags & FORK_THREAD) { 236 atomic_setbits_int(&p2->p_flag, P_THREAD); 237 p2->p_p = p1->p_p; 238 TAILQ_INSERT_TAIL(&p2->p_p->ps_threads, p2, p_thr_link); 239 p2->p_p->ps_refcnt++; 240 } else { 241 process_new(p2, p1); 242 } 243 #else 244 process_new(p2, p1); 245 #endif 246 247 /* 248 * Make a proc table entry for the new process. 249 * Start by zeroing the section of proc that is zero-initialized, 250 * then copy the section that is copied directly from the parent. 251 */ 252 bzero(&p2->p_startzero, 253 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); 254 bcopy(&p1->p_startcopy, &p2->p_startcopy, 255 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); 256 257 /* 258 * Initialize the timeouts. 259 */ 260 timeout_set(&p2->p_sleep_to, endtsleep, p2); 261 timeout_set(&p2->p_realit_to, realitexpire, p2); 262 263 /* 264 * Duplicate sub-structures as needed. 265 * Increase reference counts on shared objects. 266 * The p_stats and p_sigacts substructs are set in vm_fork. 267 */ 268 p2->p_emul = p1->p_emul; 269 if (p1->p_flag & P_PROFIL) 270 startprofclock(p2); 271 atomic_setbits_int(&p2->p_flag, p1->p_flag & (P_SUGID | P_SUGIDEXEC)); 272 if (flags & FORK_PTRACE) 273 atomic_setbits_int(&p2->p_flag, p1->p_flag & P_TRACED); 274 #ifdef RTHREADS 275 if (flags & FORK_THREAD) { 276 /* nothing */ 277 } else 278 #endif 279 { 280 p2->p_p->ps_cred = pool_get(&pcred_pool, PR_WAITOK); 281 bcopy(p1->p_p->ps_cred, p2->p_p->ps_cred, sizeof(*p2->p_p->ps_cred)); 282 p2->p_p->ps_cred->p_refcnt = 1; 283 crhold(p1->p_ucred); 284 } 285 286 /* bump references to the text vnode (for procfs) */ 287 p2->p_textvp = p1->p_textvp; 288 if (p2->p_textvp) 289 VREF(p2->p_textvp); 290 291 if (flags & FORK_CLEANFILES) 292 p2->p_fd = fdinit(p1); 293 else if (flags & FORK_SHAREFILES) 294 p2->p_fd = fdshare(p1); 295 else 296 p2->p_fd = fdcopy(p1); 297 298 /* 299 * If ps_limit is still copy-on-write, bump refcnt, 300 * otherwise get a copy that won't be modified. 301 * (If PL_SHAREMOD is clear, the structure is shared 302 * copy-on-write.) 303 */ 304 #ifdef RTHREADS 305 if (flags & FORK_THREAD) { 306 /* nothing */ 307 } else 308 #endif 309 { 310 if (p1->p_p->ps_limit->p_lflags & PL_SHAREMOD) 311 p2->p_p->ps_limit = limcopy(p1->p_p->ps_limit); 312 else { 313 p2->p_p->ps_limit = p1->p_p->ps_limit; 314 p2->p_p->ps_limit->p_refcnt++; 315 } 316 } 317 318 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) 319 atomic_setbits_int(&p2->p_flag, P_CONTROLT); 320 if (flags & FORK_PPWAIT) 321 atomic_setbits_int(&p2->p_flag, P_PPWAIT); 322 p2->p_pptr = p1; 323 if (flags & FORK_NOZOMBIE) 324 atomic_setbits_int(&p2->p_flag, P_NOZOMBIE); 325 LIST_INIT(&p2->p_children); 326 327 #ifdef KTRACE 328 /* 329 * Copy traceflag and tracefile if enabled. 330 * If not inherited, these were zeroed above. 331 */ 332 if (p1->p_traceflag & KTRFAC_INHERIT) { 333 p2->p_traceflag = p1->p_traceflag; 334 if ((p2->p_tracep = p1->p_tracep) != NULL) 335 VREF(p2->p_tracep); 336 } 337 #endif 338 339 /* 340 * set priority of child to be that of parent 341 * XXX should move p_estcpu into the region of struct proc which gets 342 * copied. 343 */ 344 scheduler_fork_hook(p1, p2); 345 346 /* 347 * Create signal actions for the child process. 348 */ 349 if (flags & FORK_SIGHAND) 350 sigactsshare(p1, p2); 351 else 352 p2->p_sigacts = sigactsinit(p1); 353 354 /* 355 * If emulation has process fork hook, call it now. 356 */ 357 if (p2->p_emul->e_proc_fork) 358 (*p2->p_emul->e_proc_fork)(p2, p1); 359 360 p2->p_addr = (struct user *)uaddr; 361 362 /* 363 * Finish creating the child process. It will return through a 364 * different path later. 365 */ 366 uvm_fork(p1, p2, ((flags & FORK_SHAREVM) ? TRUE : FALSE), stack, 367 stacksize, func ? func : child_return, arg ? arg : p2); 368 369 timeout_set(&p2->p_stats->p_virt_to, virttimer_trampoline, p2); 370 timeout_set(&p2->p_stats->p_prof_to, proftimer_trampoline, p2); 371 372 vm = p2->p_vmspace; 373 374 if (flags & FORK_FORK) { 375 forkstat.cntfork++; 376 forkstat.sizfork += vm->vm_dsize + vm->vm_ssize; 377 } else if (flags & FORK_VFORK) { 378 forkstat.cntvfork++; 379 forkstat.sizvfork += vm->vm_dsize + vm->vm_ssize; 380 } else if (flags & FORK_RFORK) { 381 forkstat.cntrfork++; 382 forkstat.sizrfork += vm->vm_dsize + vm->vm_ssize; 383 } else { 384 forkstat.cntkthread++; 385 forkstat.sizkthread += vm->vm_dsize + vm->vm_ssize; 386 } 387 388 if (p2->p_flag & P_TRACED && flags & FORK_FORK) 389 newptstat = malloc(sizeof(*newptstat), M_SUBPROC, M_WAITOK); 390 #if NSYSTRACE > 0 391 if (ISSET(p1->p_flag, P_SYSTRACE)) 392 newstrp = systrace_getproc(); 393 #endif 394 395 /* Find an unused pid satisfying 1 <= lastpid <= PID_MAX */ 396 do { 397 lastpid = 1 + (randompid ? arc4random() : lastpid) % PID_MAX; 398 } while (pidtaken(lastpid)); 399 p2->p_pid = lastpid; 400 401 LIST_INSERT_HEAD(&allproc, p2, p_list); 402 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash); 403 LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling); 404 LIST_INSERT_AFTER(p1, p2, p_pglist); 405 if (p2->p_flag & P_TRACED) { 406 p2->p_oppid = p1->p_pid; 407 if (p2->p_pptr != p1->p_pptr) 408 proc_reparent(p2, p1->p_pptr); 409 410 /* 411 * Set ptrace status. 412 */ 413 if (flags & FORK_FORK) { 414 p2->p_ptstat = newptstat; 415 newptstat = NULL; 416 p1->p_ptstat->pe_report_event = PTRACE_FORK; 417 p2->p_ptstat->pe_report_event = PTRACE_FORK; 418 p1->p_ptstat->pe_other_pid = p2->p_pid; 419 p2->p_ptstat->pe_other_pid = p1->p_pid; 420 } 421 } 422 423 #if NSYSTRACE > 0 424 if (newstrp) 425 systrace_fork(p1, p2, newstrp); 426 #endif 427 428 /* 429 * Make child runnable, set start time, and add to run queue. 430 */ 431 SCHED_LOCK(s); 432 getmicrotime(&p2->p_stats->p_start); 433 p2->p_acflag = AFORK; 434 p2->p_stat = SRUN; 435 p2->p_cpu = sched_choosecpu_fork(p1, flags); 436 setrunqueue(p2); 437 SCHED_UNLOCK(s); 438 439 if (newptstat) 440 free(newptstat, M_SUBPROC); 441 442 /* 443 * Notify any interested parties about the new process. 444 */ 445 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid); 446 447 /* 448 * Update stats now that we know the fork was successfull. 449 */ 450 uvmexp.forks++; 451 if (flags & FORK_PPWAIT) 452 uvmexp.forks_ppwait++; 453 if (flags & FORK_SHAREVM) 454 uvmexp.forks_sharevm++; 455 456 /* 457 * Pass a pointer to the new process to the caller. 458 */ 459 if (rnewprocp != NULL) 460 *rnewprocp = p2; 461 462 /* 463 * Preserve synchronization semantics of vfork. If waiting for 464 * child to exec or exit, set P_PPWAIT on child, and sleep on our 465 * proc (in case of exit). 466 */ 467 if (flags & FORK_PPWAIT) 468 while (p2->p_flag & P_PPWAIT) 469 tsleep(p1, PWAIT, "ppwait", 0); 470 471 /* 472 * If we're tracing the child, alert the parent too. 473 */ 474 if ((flags & FORK_PTRACE) && (p1->p_flag & P_TRACED)) 475 psignal(p1, SIGTRAP); 476 477 /* 478 * Return child pid to parent process, 479 * marking us as parent via retval[1]. 480 */ 481 if (retval != NULL) { 482 retval[0] = p2->p_pid; 483 retval[1] = 0; 484 } 485 return (0); 486 } 487 488 /* 489 * Checks for current use of a pid, either as a pid or pgid. 490 */ 491 int 492 pidtaken(pid_t pid) 493 { 494 struct proc *p; 495 496 if (pfind(pid) != NULL) 497 return (1); 498 if (pgfind(pid) != NULL) 499 return (1); 500 LIST_FOREACH(p, &zombproc, p_list) 501 if (p->p_pid == pid || p->p_pgid == pid) 502 return (1); 503 return (0); 504 } 505 506 #if defined(MULTIPROCESSOR) 507 /* 508 * XXX This is a slight hack to get newly-formed processes to 509 * XXX acquire the kernel lock as soon as they run. 510 */ 511 void 512 proc_trampoline_mp(void) 513 { 514 struct proc *p; 515 516 p = curproc; 517 518 SCHED_ASSERT_LOCKED(); 519 __mp_unlock(&sched_lock); 520 spl0(); 521 SCHED_ASSERT_UNLOCKED(); 522 KASSERT(__mp_lock_held(&kernel_lock) == 0); 523 524 KERNEL_PROC_LOCK(p); 525 } 526 #endif 527