kern_fork.c revision 1.85
1 /* $NetBSD: kern_fork.c,v 1.85 2001/07/01 18:06:11 thorpej 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 "opt_ktrace.h" 81 #include "opt_multiprocessor.h" 82 83 #include <sys/param.h> 84 #include <sys/systm.h> 85 #include <sys/map.h> 86 #include <sys/filedesc.h> 87 #include <sys/kernel.h> 88 #include <sys/malloc.h> 89 #include <sys/pool.h> 90 #include <sys/mount.h> 91 #include <sys/proc.h> 92 #include <sys/resourcevar.h> 93 #include <sys/vnode.h> 94 #include <sys/file.h> 95 #include <sys/acct.h> 96 #include <sys/ktrace.h> 97 #include <sys/vmmeter.h> 98 #include <sys/sched.h> 99 #include <sys/signalvar.h> 100 101 #include <sys/syscallargs.h> 102 103 #include <uvm/uvm_extern.h> 104 105 int nprocs = 1; /* process 0 */ 106 107 /*ARGSUSED*/ 108 int 109 sys_fork(struct proc *p, void *v, register_t *retval) 110 { 111 112 return (fork1(p, 0, SIGCHLD, NULL, 0, NULL, NULL, retval, NULL)); 113 } 114 115 /* 116 * vfork(2) system call compatible with 4.4BSD (i.e. BSD with Mach VM). 117 * Address space is not shared, but parent is blocked until child exit. 118 */ 119 /*ARGSUSED*/ 120 int 121 sys_vfork(struct proc *p, void *v, register_t *retval) 122 { 123 124 return (fork1(p, FORK_PPWAIT, SIGCHLD, NULL, 0, NULL, NULL, 125 retval, NULL)); 126 } 127 128 /* 129 * New vfork(2) system call for NetBSD, which implements original 3BSD vfork(2) 130 * semantics. Address space is shared, and parent is blocked until child exit. 131 */ 132 /*ARGSUSED*/ 133 int 134 sys___vfork14(struct proc *p, void *v, register_t *retval) 135 { 136 137 return (fork1(p, FORK_PPWAIT|FORK_SHAREVM, SIGCHLD, NULL, 0, 138 NULL, NULL, retval, NULL)); 139 } 140 141 /* 142 * Linux-compatible __clone(2) system call. 143 */ 144 int 145 sys___clone(struct proc *p, void *v, register_t *retval) 146 { 147 struct sys___clone_args /* { 148 syscallarg(int) flags; 149 syscallarg(void *) stack; 150 } */ *uap = v; 151 int flags, sig; 152 153 /* 154 * We don't support the CLONE_PID or CLONE_PTRACE flags. 155 */ 156 if (SCARG(uap, flags) & (CLONE_PID|CLONE_PTRACE)) 157 return (EINVAL); 158 159 if (SCARG(uap, flags) & CLONE_VM) 160 flags |= FORK_SHAREVM; 161 if (SCARG(uap, flags) & CLONE_FS) 162 flags |= FORK_SHARECWD; 163 if (SCARG(uap, flags) & CLONE_FILES) 164 flags |= FORK_SHAREFILES; 165 if (SCARG(uap, flags) & CLONE_SIGHAND) 166 flags |= FORK_SHARESIGS; 167 if (SCARG(uap, flags) & CLONE_VFORK) 168 flags |= FORK_PPWAIT; 169 170 sig = SCARG(uap, flags) & CLONE_CSIGNAL; 171 if (sig < 0 || sig >= _NSIG) 172 return (EINVAL); 173 174 /* 175 * Note that the Linux API does not provide a portable way of 176 * specifying the stack area; the caller must know if the stack 177 * grows up or down. So, we pass a stack size of 0, so that the 178 * code that makes this adjustment is a noop. 179 */ 180 return (fork1(p, flags, sig, SCARG(uap, stack), 0, 181 NULL, NULL, retval, NULL)); 182 } 183 184 int 185 fork1(struct proc *p1, int flags, int exitsig, void *stack, size_t stacksize, 186 void (*func)(void *), void *arg, register_t *retval, 187 struct proc **rnewprocp) 188 { 189 struct proc *p2, *tp; 190 uid_t uid; 191 int count, s; 192 vaddr_t uaddr; 193 static int nextpid, pidchecked; 194 195 /* 196 * Although process entries are dynamically created, we still keep 197 * a global limit on the maximum number we will create. Don't allow 198 * a nonprivileged user to use the last process; don't let root 199 * exceed the limit. The variable nprocs is the current number of 200 * processes, maxproc is the limit. 201 */ 202 uid = p1->p_cred->p_ruid; 203 if (__predict_false((nprocs >= maxproc - 1 && uid != 0) || 204 nprocs >= maxproc)) { 205 tablefull("proc", "increase kern.maxproc or NPROC"); 206 return (EAGAIN); 207 } 208 nprocs++; 209 210 /* 211 * Increment the count of procs running with this uid. Don't allow 212 * a nonprivileged user to exceed their current limit. 213 */ 214 count = chgproccnt(uid, 1); 215 if (__predict_false(uid != 0 && count > 216 p1->p_rlimit[RLIMIT_NPROC].rlim_cur)) { 217 (void)chgproccnt(uid, -1); 218 nprocs--; 219 return (EAGAIN); 220 } 221 222 /* 223 * Allocate virtual address space for the U-area now, while it 224 * is still easy to abort the fork operation if we're out of 225 * kernel virtual address space. The actual U-area pages will 226 * be allocated and wired in vm_fork(). 227 */ 228 229 #ifndef USPACE_ALIGN 230 #define USPACE_ALIGN 0 231 #endif 232 233 uaddr = uvm_km_valloc_align(kernel_map, USPACE, USPACE_ALIGN); 234 if (__predict_false(uaddr == 0)) { 235 (void)chgproccnt(uid, -1); 236 nprocs--; 237 return (ENOMEM); 238 } 239 240 /* 241 * We are now committed to the fork. From here on, we may 242 * block on resources, but resource allocation may NOT fail. 243 */ 244 245 /* Allocate new proc. */ 246 p2 = pool_get(&proc_pool, PR_WAITOK); 247 248 /* 249 * Make a proc table entry for the new process. 250 * Start by zeroing the section of proc that is zero-initialized, 251 * then copy the section that is copied directly from the parent. 252 */ 253 memset(&p2->p_startzero, 0, 254 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); 255 memcpy(&p2->p_startcopy, &p1->p_startcopy, 256 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); 257 258 #if !defined(MULTIPROCESSOR) 259 /* 260 * In the single-processor case, all processes will always run 261 * on the same CPU. So, initialize the child's CPU to the parent's 262 * now. In the multiprocessor case, the child's CPU will be 263 * initialized in the low-level context switch code when the 264 * process runs. 265 */ 266 p2->p_cpu = p1->p_cpu; 267 #else 268 /* 269 * zero child's cpu pointer so we don't get trash. 270 */ 271 p2->p_cpu = NULL; 272 #endif /* ! MULTIPROCESSOR */ 273 274 /* 275 * Duplicate sub-structures as needed. 276 * Increase reference counts on shared objects. 277 * The p_stats and p_sigacts substructs are set in uvm_fork(). 278 */ 279 p2->p_flag = P_INMEM | (p1->p_flag & P_SUGID); 280 p2->p_emul = p1->p_emul; 281 282 if (p1->p_flag & P_PROFIL) 283 startprofclock(p2); 284 p2->p_cred = pool_get(&pcred_pool, PR_WAITOK); 285 memcpy(p2->p_cred, p1->p_cred, sizeof(*p2->p_cred)); 286 p2->p_cred->p_refcnt = 1; 287 crhold(p1->p_ucred); 288 289 /* bump references to the text vnode (for procfs) */ 290 p2->p_textvp = p1->p_textvp; 291 if (p2->p_textvp) 292 VREF(p2->p_textvp); 293 294 if (flags & FORK_SHAREFILES) 295 fdshare(p1, p2); 296 else 297 p2->p_fd = fdcopy(p1); 298 299 if (flags & FORK_SHARECWD) 300 cwdshare(p1, p2); 301 else 302 p2->p_cwdi = cwdinit(p1); 303 304 /* 305 * If p_limit is still copy-on-write, bump refcnt, 306 * otherwise get a copy that won't be modified. 307 * (If PL_SHAREMOD is clear, the structure is shared 308 * copy-on-write.) 309 */ 310 if (p1->p_limit->p_lflags & PL_SHAREMOD) 311 p2->p_limit = limcopy(p1->p_limit); 312 else { 313 p2->p_limit = p1->p_limit; 314 p2->p_limit->p_refcnt++; 315 } 316 317 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) 318 p2->p_flag |= P_CONTROLT; 319 if (flags & FORK_PPWAIT) 320 p2->p_flag |= P_PPWAIT; 321 LIST_INSERT_AFTER(p1, p2, p_pglist); 322 p2->p_pptr = p1; 323 LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling); 324 LIST_INIT(&p2->p_children); 325 326 callout_init(&p2->p_realit_ch); 327 callout_init(&p2->p_tsleep_ch); 328 329 #ifdef KTRACE 330 /* 331 * Copy traceflag and tracefile if enabled. 332 * If not inherited, these were zeroed above. 333 */ 334 if (p1->p_traceflag & KTRFAC_INHERIT) { 335 p2->p_traceflag = p1->p_traceflag; 336 if ((p2->p_tracep = p1->p_tracep) != NULL) 337 ktradref(p2); 338 } 339 #endif 340 341 #ifdef __HAVE_SYSCALL_INTERN 342 (*p2->p_emul->e_syscall_intern)(p2); 343 #endif 344 345 scheduler_fork_hook(p1, p2); 346 347 /* 348 * Create signal actions for the child process. 349 */ 350 sigactsinit(p2, p1, flags & FORK_SHARESIGS); 351 352 /* 353 * If emulation has process fork hook, call it now. 354 */ 355 if (p2->p_emul->e_proc_fork) 356 (*p2->p_emul->e_proc_fork)(p2, p1); 357 358 /* 359 * This begins the section where we must prevent the parent 360 * from being swapped. 361 */ 362 PHOLD(p1); 363 364 /* 365 * Finish creating the child process. It will return through a 366 * different path later. 367 */ 368 p2->p_addr = (struct user *)uaddr; 369 uvm_fork(p1, p2, (flags & FORK_SHAREVM) ? TRUE : FALSE, 370 stack, stacksize, 371 (func != NULL) ? func : child_return, 372 (arg != NULL) ? arg : p2); 373 374 /* 375 * BEGIN PID ALLOCATION. 376 */ 377 s = proclist_lock_write(); 378 379 /* 380 * Find an unused process ID. We remember a range of unused IDs 381 * ready to use (from nextpid+1 through pidchecked-1). 382 */ 383 nextpid++; 384 retry: 385 /* 386 * If the process ID prototype has wrapped around, 387 * restart somewhat above 0, as the low-numbered procs 388 * tend to include daemons that don't exit. 389 */ 390 if (nextpid >= PID_MAX) { 391 nextpid = 500; 392 pidchecked = 0; 393 } 394 if (nextpid >= pidchecked) { 395 const struct proclist_desc *pd; 396 397 pidchecked = PID_MAX; 398 /* 399 * Scan the process lists to check whether this pid 400 * is in use. Remember the lowest pid that's greater 401 * than nextpid, so we can avoid checking for a while. 402 */ 403 pd = proclists; 404 again: 405 LIST_FOREACH(tp, pd->pd_list, p_list) { 406 while (tp->p_pid == nextpid || 407 tp->p_pgrp->pg_id == nextpid || 408 tp->p_session->s_sid == nextpid) { 409 nextpid++; 410 if (nextpid >= pidchecked) 411 goto retry; 412 } 413 if (tp->p_pid > nextpid && pidchecked > tp->p_pid) 414 pidchecked = tp->p_pid; 415 416 if (tp->p_pgrp->pg_id > nextpid && 417 pidchecked > tp->p_pgrp->pg_id) 418 pidchecked = tp->p_pgrp->pg_id; 419 420 if (tp->p_session->s_sid > nextpid && 421 pidchecked > tp->p_session->s_sid) 422 pidchecked = tp->p_session->s_sid; 423 } 424 425 /* 426 * If there's another list, scan it. If we have checked 427 * them all, we've found one! 428 */ 429 pd++; 430 if (pd->pd_list != NULL) 431 goto again; 432 } 433 434 /* Record the pid we've allocated. */ 435 p2->p_pid = nextpid; 436 437 /* Record the signal to be delivered to the parent on exit. */ 438 p2->p_exitsig = exitsig; 439 440 /* 441 * Put the proc on allproc before unlocking PID allocation 442 * so that waiters won't grab it as soon as we unlock. 443 */ 444 445 p2->p_stat = SIDL; /* protect against others */ 446 p2->p_forw = p2->p_back = NULL; /* shouldn't be necessary */ 447 448 LIST_INSERT_HEAD(&allproc, p2, p_list); 449 450 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash); 451 452 /* 453 * END PID ALLOCATION. 454 */ 455 proclist_unlock_write(s); 456 457 /* 458 * Make child runnable, set start time, and add to run queue. 459 */ 460 SCHED_LOCK(s); 461 p2->p_stats->p_start = time; 462 p2->p_acflag = AFORK; 463 p2->p_stat = SRUN; 464 setrunqueue(p2); 465 SCHED_UNLOCK(s); 466 467 /* 468 * Now can be swapped. 469 */ 470 PRELE(p1); 471 472 /* 473 * Update stats now that we know the fork was successful. 474 */ 475 uvmexp.forks++; 476 if (flags & FORK_PPWAIT) 477 uvmexp.forks_ppwait++; 478 if (flags & FORK_SHAREVM) 479 uvmexp.forks_sharevm++; 480 481 /* 482 * Pass a pointer to the new process to the caller. 483 */ 484 if (rnewprocp != NULL) 485 *rnewprocp = p2; 486 487 #ifdef KTRACE 488 if (KTRPOINT(p2, KTR_EMUL)) 489 ktremul(p2); 490 #endif 491 492 /* 493 * Preserve synchronization semantics of vfork. If waiting for 494 * child to exec or exit, set P_PPWAIT on child, and sleep on our 495 * proc (in case of exit). 496 */ 497 if (flags & FORK_PPWAIT) 498 while (p2->p_flag & P_PPWAIT) 499 tsleep(p1, PWAIT, "ppwait", 0); 500 501 /* 502 * Return child pid to parent process, 503 * marking us as parent via retval[1]. 504 */ 505 if (retval != NULL) { 506 retval[0] = p2->p_pid; 507 retval[1] = 0; 508 } 509 510 return (0); 511 } 512 513 #if defined(MULTIPROCESSOR) 514 /* 515 * XXX This is a slight hack to get newly-formed processes to 516 * XXX acquire the kernel lock as soon as they run. 517 */ 518 void 519 proc_trampoline_mp(void) 520 { 521 struct proc *p; 522 523 p = curproc; 524 525 SCHED_ASSERT_UNLOCKED(); 526 KERNEL_PROC_LOCK(p); 527 } 528 #endif 529
Indexes created Mon Sep 22 05:09:51 GMT 2025