kern_fork.c revision 1.84.2.9
1 /* $NetBSD: kern_fork.c,v 1.84.2.9 2002/06/20 03:47:13 nathanw Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2001 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_fork.c 8.8 (Berkeley) 2/14/95 78 */ 79 80 #include <sys/cdefs.h> 81 __KERNEL_RCSID(0, "$NetBSD: kern_fork.c,v 1.84.2.9 2002/06/20 03:47:13 nathanw Exp $"); 82 83 #include "opt_ktrace.h" 84 #include "opt_systrace.h" 85 #include "opt_multiprocessor.h" 86 87 #include <sys/param.h> 88 #include <sys/systm.h> 89 #include <sys/map.h> 90 #include <sys/filedesc.h> 91 #include <sys/kernel.h> 92 #include <sys/malloc.h> 93 #include <sys/pool.h> 94 #include <sys/mount.h> 95 #include <sys/lwp.h> 96 #include <sys/proc.h> 97 #include <sys/resourcevar.h> 98 #include <sys/vnode.h> 99 #include <sys/file.h> 100 #include <sys/acct.h> 101 #include <sys/ktrace.h> 102 #include <sys/vmmeter.h> 103 #include <sys/sched.h> 104 #include <sys/signalvar.h> 105 #include <sys/systrace.h> 106 107 #include <sys/sa.h> 108 #include <sys/syscallargs.h> 109 110 #include <uvm/uvm_extern.h> 111 112 113 int nprocs = 1; /* process 0 */ 114 115 /*ARGSUSED*/ 116 int 117 sys_fork(struct lwp *l, void *v, register_t *retval) 118 { 119 120 return (fork1(l, 0, SIGCHLD, NULL, 0, NULL, NULL, retval, NULL)); 121 } 122 123 /* 124 * vfork(2) system call compatible with 4.4BSD (i.e. BSD with Mach VM). 125 * Address space is not shared, but parent is blocked until child exit. 126 */ 127 /*ARGSUSED*/ 128 int 129 sys_vfork(struct lwp *l, void *v, register_t *retval) 130 { 131 132 return (fork1(l, FORK_PPWAIT, SIGCHLD, NULL, 0, NULL, NULL, 133 retval, NULL)); 134 } 135 136 /* 137 * New vfork(2) system call for NetBSD, which implements original 3BSD vfork(2) 138 * semantics. Address space is shared, and parent is blocked until child exit. 139 */ 140 /*ARGSUSED*/ 141 int 142 sys___vfork14(struct lwp *l, void *v, register_t *retval) 143 { 144 145 return (fork1(l, FORK_PPWAIT|FORK_SHAREVM, SIGCHLD, NULL, 0, 146 NULL, NULL, retval, NULL)); 147 } 148 149 /* 150 * Linux-compatible __clone(2) system call. 151 */ 152 int 153 sys___clone(struct lwp *l, void *v, register_t *retval) 154 { 155 struct sys___clone_args /* { 156 syscallarg(int) flags; 157 syscallarg(void *) stack; 158 } */ *uap = v; 159 int flags, sig; 160 161 /* 162 * We don't support the CLONE_PID or CLONE_PTRACE flags. 163 */ 164 if (SCARG(uap, flags) & (CLONE_PID|CLONE_PTRACE)) 165 return (EINVAL); 166 167 flags = 0; 168 169 if (SCARG(uap, flags) & CLONE_VM) 170 flags |= FORK_SHAREVM; 171 if (SCARG(uap, flags) & CLONE_FS) 172 flags |= FORK_SHARECWD; 173 if (SCARG(uap, flags) & CLONE_FILES) 174 flags |= FORK_SHAREFILES; 175 if (SCARG(uap, flags) & CLONE_SIGHAND) 176 flags |= FORK_SHARESIGS; 177 if (SCARG(uap, flags) & CLONE_VFORK) 178 flags |= FORK_PPWAIT; 179 180 sig = SCARG(uap, flags) & CLONE_CSIGNAL; 181 if (sig < 0 || sig >= _NSIG) 182 return (EINVAL); 183 184 /* 185 * Note that the Linux API does not provide a portable way of 186 * specifying the stack area; the caller must know if the stack 187 * grows up or down. So, we pass a stack size of 0, so that the 188 * code that makes this adjustment is a noop. 189 */ 190 return (fork1(l, flags, sig, SCARG(uap, stack), 0, 191 NULL, NULL, retval, NULL)); 192 } 193 194 int 195 fork1(struct lwp *l1, int flags, int exitsig, void *stack, size_t stacksize, 196 void (*func)(void *), void *arg, register_t *retval, 197 struct proc **rnewprocp) 198 { 199 struct proc *p1, *p2, *tp; 200 uid_t uid; 201 struct lwp *l2; 202 int count, s; 203 vaddr_t uaddr; 204 static int nextpid, pidchecked; 205 206 /* 207 * Although process entries are dynamically created, we still keep 208 * a global limit on the maximum number we will create. Don't allow 209 * a nonprivileged user to use the last process; don't let root 210 * exceed the limit. The variable nprocs is the current number of 211 * processes, maxproc is the limit. 212 */ 213 p1 = l1->l_proc; 214 uid = p1->p_cred->p_ruid; 215 if (__predict_false((nprocs >= maxproc - 1 && uid != 0) || 216 nprocs >= maxproc)) { 217 tablefull("proc", "increase kern.maxproc or NPROC"); 218 return (EAGAIN); 219 } 220 nprocs++; 221 222 /* 223 * Increment the count of procs running with this uid. Don't allow 224 * a nonprivileged user to exceed their current limit. 225 */ 226 count = chgproccnt(uid, 1); 227 if (__predict_false(uid != 0 && count > 228 p1->p_rlimit[RLIMIT_NPROC].rlim_cur)) { 229 (void)chgproccnt(uid, -1); 230 nprocs--; 231 return (EAGAIN); 232 } 233 234 /* 235 * Allocate virtual address space for the U-area now, while it 236 * is still easy to abort the fork operation if we're out of 237 * kernel virtual address space. The actual U-area pages will 238 * be allocated and wired in vm_fork(). 239 */ 240 241 #ifndef USPACE_ALIGN 242 #define USPACE_ALIGN 0 243 #endif 244 245 uaddr = uvm_km_valloc_align(kernel_map, USPACE, USPACE_ALIGN); 246 if (__predict_false(uaddr == 0)) { 247 (void)chgproccnt(uid, -1); 248 nprocs--; 249 return (ENOMEM); 250 } 251 252 /* 253 * We are now committed to the fork. From here on, we may 254 * block on resources, but resource allocation may NOT fail. 255 */ 256 257 /* Allocate new proc. */ 258 p2 = pool_get(&proc_pool, PR_WAITOK); 259 260 /* 261 * BEGIN PID ALLOCATION. 262 */ 263 s = proclist_lock_write(); 264 265 /* 266 * Find an unused process ID. We remember a range of unused IDs 267 * ready to use (from nextpid+1 through pidchecked-1). 268 */ 269 nextpid++; 270 retry: 271 /* 272 * If the process ID prototype has wrapped around, 273 * restart somewhat above 0, as the low-numbered procs 274 * tend to include daemons that don't exit. 275 */ 276 if (nextpid >= PID_MAX) { 277 nextpid = 500; 278 pidchecked = 0; 279 } 280 if (nextpid >= pidchecked) { 281 const struct proclist_desc *pd; 282 283 pidchecked = PID_MAX; 284 /* 285 * Scan the process lists to check whether this pid 286 * is in use. Remember the lowest pid that's greater 287 * than nextpid, so we can avoid checking for a while. 288 */ 289 pd = proclists; 290 again: 291 LIST_FOREACH(tp, pd->pd_list, p_list) { 292 while (tp->p_pid == nextpid || 293 tp->p_pgrp->pg_id == nextpid || 294 tp->p_session->s_sid == nextpid) { 295 nextpid++; 296 if (nextpid >= pidchecked) 297 goto retry; 298 } 299 if (tp->p_pid > nextpid && pidchecked > tp->p_pid) 300 pidchecked = tp->p_pid; 301 302 if (tp->p_pgrp->pg_id > nextpid && 303 pidchecked > tp->p_pgrp->pg_id) 304 pidchecked = tp->p_pgrp->pg_id; 305 306 if (tp->p_session->s_sid > nextpid && 307 pidchecked > tp->p_session->s_sid) 308 pidchecked = tp->p_session->s_sid; 309 } 310 311 /* 312 * If there's another list, scan it. If we have checked 313 * them all, we've found one! 314 */ 315 pd++; 316 if (pd->pd_list != NULL) 317 goto again; 318 } 319 320 /* 321 * Put the proc on allproc before unlocking PID allocation 322 * so that waiters won't grab it as soon as we unlock. 323 */ 324 325 p2->p_stat = SIDL; /* protect against others */ 326 p2->p_pid = nextpid; 327 p2->p_exitsig = exitsig; /* signal for parent on exit */ 328 p2->p_forw = p2->p_back = NULL; /* shouldn't be necessary */ 329 330 LIST_INSERT_HEAD(&allproc, p2, p_list); 331 332 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash); 333 334 /* 335 * END PID ALLOCATION. 336 */ 337 proclist_unlock_write(s); 338 339 /* 340 * Make a proc table entry for the new process. 341 * Start by zeroing the section of proc that is zero-initialized, 342 * then copy the section that is copied directly from the parent. 343 */ 344 memset(&p2->p_startzero, 0, 345 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); 346 memcpy(&p2->p_startcopy, &p1->p_startcopy, 347 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); 348 349 simple_lock_init(&p2->p_lwplock); 350 LIST_INIT(&p2->p_lwps); 351 352 /* 353 * Duplicate sub-structures as needed. 354 * Increase reference counts on shared objects. 355 * The p_stats and p_sigacts substructs are set in uvm_fork(). 356 */ 357 p2->p_flag = p1->p_flag & (P_SUGID); 358 p2->p_emul = p1->p_emul; 359 p2->p_execsw = p1->p_execsw; 360 361 if (p1->p_flag & P_PROFIL) 362 startprofclock(p2); 363 p2->p_cred = pool_get(&pcred_pool, PR_WAITOK); 364 memcpy(p2->p_cred, p1->p_cred, sizeof(*p2->p_cred)); 365 p2->p_cred->p_refcnt = 1; 366 crhold(p1->p_ucred); 367 368 369 /* bump references to the text vnode (for procfs) */ 370 p2->p_textvp = p1->p_textvp; 371 if (p2->p_textvp) 372 VREF(p2->p_textvp); 373 374 if (flags & FORK_SHAREFILES) 375 fdshare(p1, p2); 376 else 377 p2->p_fd = fdcopy(p1); 378 379 if (flags & FORK_SHARECWD) 380 cwdshare(p1, p2); 381 else 382 p2->p_cwdi = cwdinit(p1); 383 384 /* 385 * If p_limit is still copy-on-write, bump refcnt, 386 * otherwise get a copy that won't be modified. 387 * (If PL_SHAREMOD is clear, the structure is shared 388 * copy-on-write.) 389 */ 390 if (p1->p_limit->p_lflags & PL_SHAREMOD) 391 p2->p_limit = limcopy(p1->p_limit); 392 else { 393 p2->p_limit = p1->p_limit; 394 p2->p_limit->p_refcnt++; 395 } 396 397 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) 398 p2->p_flag |= P_CONTROLT; 399 if (flags & FORK_PPWAIT) 400 p2->p_flag |= P_PPWAIT; 401 LIST_INSERT_AFTER(p1, p2, p_pglist); 402 p2->p_pptr = p1; 403 LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling); 404 LIST_INIT(&p2->p_children); 405 406 #ifdef KTRACE 407 /* 408 * Copy traceflag and tracefile if enabled. 409 * If not inherited, these were zeroed above. 410 */ 411 if (p1->p_traceflag & KTRFAC_INHERIT) { 412 p2->p_traceflag = p1->p_traceflag; 413 if ((p2->p_tracep = p1->p_tracep) != NULL) 414 ktradref(p2); 415 } 416 #endif 417 #ifdef SYSTRACE 418 /* Tell systrace what's happening. */ 419 if (ISSET(p1->p_flag, P_SYSTRACE)) 420 systrace_sys_fork(p1, p2); 421 #endif 422 423 424 #ifdef __HAVE_SYSCALL_INTERN 425 (*p2->p_emul->e_syscall_intern)(p2); 426 #endif 427 428 scheduler_fork_hook(p1, p2); 429 430 /* 431 * Create signal actions for the child process. 432 */ 433 sigactsinit(p2, p1, flags & FORK_SHARESIGS); 434 435 /* 436 * p_stats. 437 * Copy parts of p_stats, and zero out the rest. 438 */ 439 p2->p_stats = pstatscopy(p1->p_stats); 440 441 /* 442 * If emulation has process fork hook, call it now. 443 */ 444 if (p2->p_emul->e_proc_fork) 445 (*p2->p_emul->e_proc_fork)(p2, p1); 446 447 /* 448 * This begins the section where we must prevent the parent 449 * from being swapped. 450 */ 451 PHOLD(l1); 452 453 uvm_proc_fork(p1, p2, (flags & FORK_SHAREVM) ? TRUE : FALSE); 454 455 /* 456 * Finish creating the child process. It will return through a 457 * different path later. 458 */ 459 newlwp(l1, p2, uaddr, 0, stack, stacksize, 460 (func != NULL) ? func : child_return, 461 arg, &l2); 462 463 /* 464 * Make child runnable, set start time, and add to run queue. 465 */ 466 SCHED_LOCK(s); 467 p2->p_stats->p_start = time; 468 p2->p_acflag = AFORK; 469 p2->p_stat = SACTIVE; 470 p2->p_nrlwps = 1; 471 l2->l_stat = LSRUN; 472 setrunqueue(l2); 473 SCHED_UNLOCK(s); 474 475 /* 476 * Now can be swapped. 477 */ 478 PRELE(l1); 479 480 /* 481 * Update stats now that we know the fork was successful. 482 */ 483 uvmexp.forks++; 484 if (flags & FORK_PPWAIT) 485 uvmexp.forks_ppwait++; 486 if (flags & FORK_SHAREVM) 487 uvmexp.forks_sharevm++; 488 489 /* 490 * Pass a pointer to the new process to the caller. 491 */ 492 if (rnewprocp != NULL) 493 *rnewprocp = p2; 494 495 #ifdef KTRACE 496 if (KTRPOINT(p2, KTR_EMUL)) 497 ktremul(p2); 498 #endif 499 500 /* 501 * Preserve synchronization semantics of vfork. If waiting for 502 * child to exec or exit, set P_PPWAIT on child, and sleep on our 503 * proc (in case of exit). 504 */ 505 if (flags & FORK_PPWAIT) 506 while (p2->p_flag & P_PPWAIT) 507 tsleep(p1, PWAIT, "ppwait", 0); 508 509 /* 510 * Return child pid to parent process, 511 * marking us as parent via retval[1]. 512 */ 513 if (retval != NULL) { 514 retval[0] = p2->p_pid; 515 retval[1] = 0; 516 } 517 518 return (0); 519 } 520 521 #if defined(MULTIPROCESSOR) 522 /* 523 * XXX This is a slight hack to get newly-formed processes to 524 * XXX acquire the kernel lock as soon as they run. 525 */ 526 void 527 proc_trampoline_mp(void) 528 { 529 struct lwp *l; 530 531 l = curproc; 532 533 SCHED_ASSERT_UNLOCKED(); 534 KERNEL_PROC_LOCK(l); 535 } 536 #endif 537
Indexes created Fri Sep 26 20:09:58 GMT 2025