kern_fork.c revision 1.84.2.13
1 /* $NetBSD: kern_fork.c,v 1.84.2.13 2002/07/17 18:03:06 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 1.84.2.13 2002/07/17 18:03:06 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/proc.h> 96 #include <sys/resourcevar.h> 97 #include <sys/vnode.h> 98 #include <sys/file.h> 99 #include <sys/acct.h> 100 #include <sys/ktrace.h> 101 #include <sys/vmmeter.h> 102 #include <sys/sched.h> 103 #include <sys/signalvar.h> 104 #include <sys/systrace.h> 105 106 #include <sys/sa.h> 107 #include <sys/syscallargs.h> 108 109 #include <uvm/uvm_extern.h> 110 111 112 int nprocs = 1; /* process 0 */ 113 114 /*ARGSUSED*/ 115 int 116 sys_fork(struct lwp *l, void *v, register_t *retval) 117 { 118 119 return (fork1(l, 0, SIGCHLD, NULL, 0, NULL, NULL, retval, NULL)); 120 } 121 122 /* 123 * vfork(2) system call compatible with 4.4BSD (i.e. BSD with Mach VM). 124 * Address space is not shared, but parent is blocked until child exit. 125 */ 126 /*ARGSUSED*/ 127 int 128 sys_vfork(struct lwp *l, void *v, register_t *retval) 129 { 130 131 return (fork1(l, FORK_PPWAIT, SIGCHLD, NULL, 0, NULL, NULL, 132 retval, NULL)); 133 } 134 135 /* 136 * New vfork(2) system call for NetBSD, which implements original 3BSD vfork(2) 137 * semantics. Address space is shared, and parent is blocked until child exit. 138 */ 139 /*ARGSUSED*/ 140 int 141 sys___vfork14(struct lwp *l, void *v, register_t *retval) 142 { 143 144 return (fork1(l, FORK_PPWAIT|FORK_SHAREVM, SIGCHLD, NULL, 0, 145 NULL, NULL, retval, NULL)); 146 } 147 148 /* 149 * Linux-compatible __clone(2) system call. 150 */ 151 int 152 sys___clone(struct lwp *l, void *v, register_t *retval) 153 { 154 struct sys___clone_args /* { 155 syscallarg(int) flags; 156 syscallarg(void *) stack; 157 } */ *uap = v; 158 int flags, sig; 159 160 /* 161 * We don't support the CLONE_PID or CLONE_PTRACE flags. 162 */ 163 if (SCARG(uap, flags) & (CLONE_PID|CLONE_PTRACE)) 164 return (EINVAL); 165 166 flags = 0; 167 168 if (SCARG(uap, flags) & CLONE_VM) 169 flags |= FORK_SHAREVM; 170 if (SCARG(uap, flags) & CLONE_FS) 171 flags |= FORK_SHARECWD; 172 if (SCARG(uap, flags) & CLONE_FILES) 173 flags |= FORK_SHAREFILES; 174 if (SCARG(uap, flags) & CLONE_SIGHAND) 175 flags |= FORK_SHARESIGS; 176 if (SCARG(uap, flags) & CLONE_VFORK) 177 flags |= FORK_PPWAIT; 178 179 sig = SCARG(uap, flags) & CLONE_CSIGNAL; 180 if (sig < 0 || sig >= _NSIG) 181 return (EINVAL); 182 183 /* 184 * Note that the Linux API does not provide a portable way of 185 * specifying the stack area; the caller must know if the stack 186 * grows up or down. So, we pass a stack size of 0, so that the 187 * code that makes this adjustment is a noop. 188 */ 189 return (fork1(l, flags, sig, SCARG(uap, stack), 0, 190 NULL, NULL, retval, NULL)); 191 } 192 193 int 194 fork1(struct lwp *l1, int flags, int exitsig, void *stack, size_t stacksize, 195 void (*func)(void *), void *arg, register_t *retval, 196 struct proc **rnewprocp) 197 { 198 struct proc *p1, *p2, *tp; 199 uid_t uid; 200 struct lwp *l2; 201 int count, s; 202 vaddr_t uaddr; 203 static int nextpid, pidchecked; 204 205 /* 206 * Although process entries are dynamically created, we still keep 207 * a global limit on the maximum number we will create. Don't allow 208 * a nonprivileged user to use the last process; don't let root 209 * exceed the limit. The variable nprocs is the current number of 210 * processes, maxproc is the limit. 211 */ 212 p1 = l1->l_proc; 213 uid = p1->p_cred->p_ruid; 214 if (__predict_false((nprocs >= maxproc - 1 && uid != 0) || 215 nprocs >= maxproc)) { 216 tablefull("proc", "increase kern.maxproc or NPROC"); 217 return (EAGAIN); 218 } 219 nprocs++; 220 221 /* 222 * Increment the count of procs running with this uid. Don't allow 223 * a nonprivileged user to exceed their current limit. 224 */ 225 count = chgproccnt(uid, 1); 226 if (__predict_false(uid != 0 && count > 227 p1->p_rlimit[RLIMIT_NPROC].rlim_cur)) { 228 (void)chgproccnt(uid, -1); 229 nprocs--; 230 return (EAGAIN); 231 } 232 233 /* 234 * Allocate virtual address space for the U-area now, while it 235 * is still easy to abort the fork operation if we're out of 236 * kernel virtual address space. The actual U-area pages will 237 * be allocated and wired in vm_fork(). 238 */ 239 240 #ifndef USPACE_ALIGN 241 #define USPACE_ALIGN 0 242 #endif 243 244 uaddr = uvm_km_valloc_align(kernel_map, USPACE, USPACE_ALIGN); 245 if (__predict_false(uaddr == 0)) { 246 (void)chgproccnt(uid, -1); 247 nprocs--; 248 return (ENOMEM); 249 } 250 251 /* 252 * We are now committed to the fork. From here on, we may 253 * block on resources, but resource allocation may NOT fail. 254 */ 255 256 /* Allocate new proc. */ 257 p2 = pool_get(&proc_pool, PR_WAITOK); 258 259 /* 260 * BEGIN PID ALLOCATION. 261 */ 262 s = proclist_lock_write(); 263 264 /* 265 * Find an unused process ID. We remember a range of unused IDs 266 * ready to use (from nextpid+1 through pidchecked-1). 267 */ 268 nextpid++; 269 retry: 270 /* 271 * If the process ID prototype has wrapped around, 272 * restart somewhat above 0, as the low-numbered procs 273 * tend to include daemons that don't exit. 274 */ 275 if (nextpid >= PID_MAX) { 276 nextpid = 500; 277 pidchecked = 0; 278 } 279 if (nextpid >= pidchecked) { 280 const struct proclist_desc *pd; 281 282 pidchecked = PID_MAX; 283 /* 284 * Scan the process lists to check whether this pid 285 * is in use. Remember the lowest pid that's greater 286 * than nextpid, so we can avoid checking for a while. 287 */ 288 pd = proclists; 289 again: 290 LIST_FOREACH(tp, pd->pd_list, p_list) { 291 while (tp->p_pid == nextpid || 292 tp->p_pgrp->pg_id == nextpid || 293 tp->p_session->s_sid == nextpid) { 294 nextpid++; 295 if (nextpid >= pidchecked) 296 goto retry; 297 } 298 if (tp->p_pid > nextpid && pidchecked > tp->p_pid) 299 pidchecked = tp->p_pid; 300 301 if (tp->p_pgrp->pg_id > nextpid && 302 pidchecked > tp->p_pgrp->pg_id) 303 pidchecked = tp->p_pgrp->pg_id; 304 305 if (tp->p_session->s_sid > nextpid && 306 pidchecked > tp->p_session->s_sid) 307 pidchecked = tp->p_session->s_sid; 308 } 309 310 /* 311 * If there's another list, scan it. If we have checked 312 * them all, we've found one! 313 */ 314 pd++; 315 if (pd->pd_list != NULL) 316 goto again; 317 } 318 319 /* 320 * Put the proc on allproc before unlocking PID allocation 321 * so that waiters won't grab it as soon as we unlock. 322 */ 323 324 p2->p_stat = SIDL; /* protect against others */ 325 p2->p_pid = nextpid; 326 p2->p_exitsig = exitsig; /* signal for parent on exit */ 327 328 LIST_INSERT_HEAD(&allproc, p2, p_list); 329 330 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash); 331 332 /* 333 * END PID ALLOCATION. 334 */ 335 proclist_unlock_write(s); 336 337 /* 338 * Make a proc table entry for the new process. 339 * Start by zeroing the section of proc that is zero-initialized, 340 * then copy the section that is copied directly from the parent. 341 */ 342 memset(&p2->p_startzero, 0, 343 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); 344 memcpy(&p2->p_startcopy, &p1->p_startcopy, 345 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); 346 347 simple_lock_init(&p2->p_lwplock); 348 LIST_INIT(&p2->p_lwps); 349 350 /* 351 * Duplicate sub-structures as needed. 352 * Increase reference counts on shared objects. 353 * The p_stats and p_sigacts substructs are set in uvm_fork(). 354 */ 355 p2->p_flag = p1->p_flag & (P_SUGID); 356 p2->p_emul = p1->p_emul; 357 p2->p_execsw = p1->p_execsw; 358 359 if (p1->p_flag & P_PROFIL) 360 startprofclock(p2); 361 p2->p_cred = pool_get(&pcred_pool, PR_WAITOK); 362 memcpy(p2->p_cred, p1->p_cred, sizeof(*p2->p_cred)); 363 p2->p_cred->p_refcnt = 1; 364 crhold(p1->p_ucred); 365 366 /* bump references to the text vnode (for procfs) */ 367 p2->p_textvp = p1->p_textvp; 368 if (p2->p_textvp) 369 VREF(p2->p_textvp); 370 371 if (flags & FORK_SHAREFILES) 372 fdshare(p1, p2); 373 else 374 p2->p_fd = fdcopy(p1); 375 376 if (flags & FORK_SHARECWD) 377 cwdshare(p1, p2); 378 else 379 p2->p_cwdi = cwdinit(p1); 380 381 /* 382 * If p_limit is still copy-on-write, bump refcnt, 383 * otherwise get a copy that won't be modified. 384 * (If PL_SHAREMOD is clear, the structure is shared 385 * copy-on-write.) 386 */ 387 if (p1->p_limit->p_lflags & PL_SHAREMOD) 388 p2->p_limit = limcopy(p1->p_limit); 389 else { 390 p2->p_limit = p1->p_limit; 391 p2->p_limit->p_refcnt++; 392 } 393 394 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) 395 p2->p_flag |= P_CONTROLT; 396 if (flags & FORK_PPWAIT) 397 p2->p_flag |= P_PPWAIT; 398 LIST_INSERT_AFTER(p1, p2, p_pglist); 399 p2->p_pptr = p1; 400 LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling); 401 LIST_INIT(&p2->p_children); 402 403 #ifdef KTRACE 404 /* 405 * Copy traceflag and tracefile if enabled. 406 * If not inherited, these were zeroed above. 407 */ 408 if (p1->p_traceflag & KTRFAC_INHERIT) { 409 p2->p_traceflag = p1->p_traceflag; 410 if ((p2->p_tracep = p1->p_tracep) != NULL) 411 ktradref(p2); 412 } 413 #endif 414 #ifdef SYSTRACE 415 /* Tell systrace what's happening. */ 416 if (ISSET(p1->p_flag, P_SYSTRACE)) 417 systrace_sys_fork(p1, p2); 418 #endif 419 420 421 #ifdef __HAVE_SYSCALL_INTERN 422 (*p2->p_emul->e_syscall_intern)(p2); 423 #endif 424 425 scheduler_fork_hook(p1, p2); 426 427 /* 428 * Create signal actions for the child process. 429 */ 430 sigactsinit(p2, p1, flags & FORK_SHARESIGS); 431 432 /* 433 * p_stats. 434 * Copy parts of p_stats, and zero out the rest. 435 */ 436 p2->p_stats = pstatscopy(p1->p_stats); 437 438 /* 439 * If emulation has process fork hook, call it now. 440 */ 441 if (p2->p_emul->e_proc_fork) 442 (*p2->p_emul->e_proc_fork)(p2, p1); 443 444 /* 445 * This begins the section where we must prevent the parent 446 * from being swapped. 447 */ 448 PHOLD(l1); 449 450 uvm_proc_fork(p1, p2, (flags & FORK_SHAREVM) ? TRUE : FALSE); 451 452 /* 453 * Finish creating the child process. It will return through a 454 * different path later. 455 */ 456 newlwp(l1, p2, uaddr, 0, stack, stacksize, 457 (func != NULL) ? func : child_return, 458 arg, &l2); 459 460 /* 461 * Make child runnable, set start time, and add to run queue. 462 */ 463 SCHED_LOCK(s); 464 p2->p_stats->p_start = time; 465 p2->p_acflag = AFORK; 466 p2->p_stat = SACTIVE; 467 p2->p_nrlwps = 1; 468 l2->l_stat = LSRUN; 469 setrunqueue(l2); 470 SCHED_UNLOCK(s); 471 472 /* 473 * Now can be swapped. 474 */ 475 PRELE(l1); 476 477 /* 478 * Update stats now that we know the fork was successful. 479 */ 480 uvmexp.forks++; 481 if (flags & FORK_PPWAIT) 482 uvmexp.forks_ppwait++; 483 if (flags & FORK_SHAREVM) 484 uvmexp.forks_sharevm++; 485 486 /* 487 * Pass a pointer to the new process to the caller. 488 */ 489 if (rnewprocp != NULL) 490 *rnewprocp = p2; 491 492 #ifdef KTRACE 493 if (KTRPOINT(p2, KTR_EMUL)) 494 ktremul(p2); 495 #endif 496 497 /* 498 * Preserve synchronization semantics of vfork. If waiting for 499 * child to exec or exit, set P_PPWAIT on child, and sleep on our 500 * proc (in case of exit). 501 */ 502 if (flags & FORK_PPWAIT) 503 while (p2->p_flag & P_PPWAIT) 504 tsleep(p1, PWAIT, "ppwait", 0); 505 506 /* 507 * Return child pid to parent process, 508 * marking us as parent via retval[1]. 509 */ 510 if (retval != NULL) { 511 retval[0] = p2->p_pid; 512 retval[1] = 0; 513 } 514 515 return (0); 516 } 517 518 #if defined(MULTIPROCESSOR) 519 /* 520 * XXX This is a slight hack to get newly-formed processes to 521 * XXX acquire the kernel lock as soon as they run. 522 */ 523 void 524 proc_trampoline_mp(void) 525 { 526 struct lwp *l; 527 528 l = curlwp; 529 530 SCHED_ASSERT_UNLOCKED(); 531 KERNEL_PROC_LOCK(l); 532 } 533 #endif 534
Indexes created Thu Sep 25 16:09:42 GMT 2025