1 /* $OpenBSD: kern_fork.c,v 1.76 2005/05/29 03:20:41 deraadt 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 59 #include <sys/syscallargs.h> 60 61 #include "systrace.h" 62 #include <dev/systrace.h> 63 64 #include <uvm/uvm_extern.h> 65 #include <uvm/uvm_map.h> 66 67 int nprocs = 1; /* process 0 */ 68 int randompid; /* when set to 1, pid's go random */ 69 pid_t lastpid; 70 struct forkstat forkstat; 71 72 int pidtaken(pid_t); 73 74 /*ARGSUSED*/ 75 int 76 sys_fork(struct proc *p, void *v, register_t *retval) 77 { 78 return (fork1(p, SIGCHLD, FORK_FORK, NULL, 0, NULL, 79 NULL, retval, NULL)); 80 } 81 82 /*ARGSUSED*/ 83 int 84 sys_vfork(struct proc *p, void *v, register_t *retval) 85 { 86 return (fork1(p, SIGCHLD, FORK_VFORK|FORK_PPWAIT, NULL, 0, NULL, 87 NULL, retval, NULL)); 88 } 89 90 int 91 sys_rfork(struct proc *p, void *v, register_t *retval) 92 { 93 struct sys_rfork_args /* { 94 syscallarg(int) flags; 95 } */ *uap = v; 96 97 int rforkflags; 98 int flags; 99 100 flags = FORK_RFORK; 101 rforkflags = SCARG(uap, flags); 102 103 if ((rforkflags & RFPROC) == 0) 104 return (EINVAL); 105 106 switch(rforkflags & (RFFDG|RFCFDG)) { 107 case (RFFDG|RFCFDG): 108 return EINVAL; 109 case RFCFDG: 110 flags |= FORK_CLEANFILES; 111 break; 112 case RFFDG: 113 break; 114 default: 115 flags |= FORK_SHAREFILES; 116 break; 117 } 118 119 if (rforkflags & RFNOWAIT) 120 flags |= FORK_NOZOMBIE; 121 122 if (rforkflags & RFMEM) 123 flags |= FORK_SHAREVM; 124 125 return (fork1(p, SIGCHLD, flags, NULL, 0, NULL, NULL, retval, NULL)); 126 } 127 128 /* print the 'table full' message once per 10 seconds */ 129 struct timeval fork_tfmrate = { 10, 0 }; 130 131 int 132 fork1(struct proc *p1, int exitsig, int flags, void *stack, size_t stacksize, 133 void (*func)(void *), void *arg, register_t *retval, 134 struct proc **rnewprocp) 135 { 136 struct proc *p2; 137 uid_t uid; 138 struct vmspace *vm; 139 int count; 140 vaddr_t uaddr; 141 int s; 142 extern void endtsleep(void *); 143 extern void realitexpire(void *); 144 145 /* 146 * Although process entries are dynamically created, we still keep 147 * a global limit on the maximum number we will create. We reserve 148 * the last 5 processes to root. The variable nprocs is the current 149 * number of processes, maxproc is the limit. 150 */ 151 uid = p1->p_cred->p_ruid; 152 if ((nprocs >= maxproc - 5 && uid != 0) || nprocs >= maxproc) { 153 static struct timeval lasttfm; 154 155 if (ratecheck(&lasttfm, &fork_tfmrate)) 156 tablefull("proc"); 157 return (EAGAIN); 158 } 159 nprocs++; 160 161 /* 162 * Increment the count of procs running with this uid. Don't allow 163 * a nonprivileged user to exceed their current limit. 164 */ 165 count = chgproccnt(uid, 1); 166 if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) { 167 (void)chgproccnt(uid, -1); 168 nprocs--; 169 return (EAGAIN); 170 } 171 172 /* 173 * Allocate a pcb and kernel stack for the process 174 */ 175 uaddr = uvm_km_valloc_align(kernel_map, USPACE, USPACE_ALIGN); 176 if (uaddr == 0) { 177 chgproccnt(uid, -1); 178 nprocs--; 179 return (ENOMEM); 180 } 181 182 /* 183 * From now on, we're committed to the fork and cannot fail. 184 */ 185 186 /* Allocate new proc. */ 187 p2 = pool_get(&proc_pool, PR_WAITOK); 188 189 p2->p_stat = SIDL; /* protect against others */ 190 p2->p_exitsig = exitsig; 191 p2->p_forw = p2->p_back = NULL; 192 193 /* 194 * Make a proc table entry for the new process. 195 * Start by zeroing the section of proc that is zero-initialized, 196 * then copy the section that is copied directly from the parent. 197 */ 198 bzero(&p2->p_startzero, 199 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); 200 bcopy(&p1->p_startcopy, &p2->p_startcopy, 201 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); 202 203 /* 204 * Initialize the timeouts. 205 */ 206 timeout_set(&p2->p_sleep_to, endtsleep, p2); 207 timeout_set(&p2->p_realit_to, realitexpire, p2); 208 209 #if defined(__HAVE_CPUINFO) 210 p2->p_cpu = p1->p_cpu; 211 #endif 212 213 /* 214 * Duplicate sub-structures as needed. 215 * Increase reference counts on shared objects. 216 * The p_stats and p_sigacts substructs are set in vm_fork. 217 */ 218 p2->p_flag = P_INMEM; 219 p2->p_emul = p1->p_emul; 220 if (p1->p_flag & P_PROFIL) 221 startprofclock(p2); 222 p2->p_flag |= (p1->p_flag & (P_SUGID | P_SUGIDEXEC)); 223 p2->p_cred = pool_get(&pcred_pool, PR_WAITOK); 224 bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred)); 225 p2->p_cred->p_refcnt = 1; 226 crhold(p1->p_ucred); 227 228 /* bump references to the text vnode (for procfs) */ 229 p2->p_textvp = p1->p_textvp; 230 if (p2->p_textvp) 231 VREF(p2->p_textvp); 232 233 if (flags & FORK_CLEANFILES) 234 p2->p_fd = fdinit(p1); 235 else if (flags & FORK_SHAREFILES) 236 p2->p_fd = fdshare(p1); 237 else 238 p2->p_fd = fdcopy(p1); 239 240 /* 241 * If p_limit is still copy-on-write, bump refcnt, 242 * otherwise get a copy that won't be modified. 243 * (If PL_SHAREMOD is clear, the structure is shared 244 * copy-on-write.) 245 */ 246 if (p1->p_limit->p_lflags & PL_SHAREMOD) 247 p2->p_limit = limcopy(p1->p_limit); 248 else { 249 p2->p_limit = p1->p_limit; 250 p2->p_limit->p_refcnt++; 251 } 252 253 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) 254 p2->p_flag |= P_CONTROLT; 255 if (flags & FORK_PPWAIT) 256 p2->p_flag |= P_PPWAIT; 257 LIST_INSERT_AFTER(p1, p2, p_pglist); 258 p2->p_pptr = p1; 259 if (flags & FORK_NOZOMBIE) 260 p2->p_flag |= P_NOZOMBIE; 261 LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling); 262 LIST_INIT(&p2->p_children); 263 264 #ifdef KTRACE 265 /* 266 * Copy traceflag and tracefile if enabled. 267 * If not inherited, these were zeroed above. 268 */ 269 if (p1->p_traceflag & KTRFAC_INHERIT) { 270 p2->p_traceflag = p1->p_traceflag; 271 if ((p2->p_tracep = p1->p_tracep) != NULL) 272 VREF(p2->p_tracep); 273 } 274 #endif 275 276 /* 277 * set priority of child to be that of parent 278 * XXX should move p_estcpu into the region of struct proc which gets 279 * copied. 280 */ 281 scheduler_fork_hook(p1, p2); 282 283 /* 284 * Create signal actions for the child process. 285 */ 286 if (flags & FORK_SIGHAND) 287 sigactsshare(p1, p2); 288 else 289 p2->p_sigacts = sigactsinit(p1); 290 291 /* 292 * If emulation has process fork hook, call it now. 293 */ 294 if (p2->p_emul->e_proc_fork) 295 (*p2->p_emul->e_proc_fork)(p2, p1); 296 /* 297 * This begins the section where we must prevent the parent 298 * from being swapped. 299 */ 300 PHOLD(p1); 301 302 p2->p_addr = (struct user *)uaddr; 303 304 /* 305 * Finish creating the child process. It will return through a 306 * different path later. 307 */ 308 uvm_fork(p1, p2, ((flags & FORK_SHAREVM) ? TRUE : FALSE), stack, 309 stacksize, func ? func : child_return, arg ? arg : p2); 310 311 vm = p2->p_vmspace; 312 313 if (flags & FORK_FORK) { 314 forkstat.cntfork++; 315 forkstat.sizfork += vm->vm_dsize + vm->vm_ssize; 316 } else if (flags & FORK_VFORK) { 317 forkstat.cntvfork++; 318 forkstat.sizvfork += vm->vm_dsize + vm->vm_ssize; 319 } else if (flags & FORK_RFORK) { 320 forkstat.cntrfork++; 321 forkstat.sizrfork += vm->vm_dsize + vm->vm_ssize; 322 } else { 323 forkstat.cntkthread++; 324 forkstat.sizkthread += vm->vm_dsize + vm->vm_ssize; 325 } 326 327 /* Find an unused pid satisfying 1 <= lastpid <= PID_MAX */ 328 do { 329 lastpid = 1 + (randompid ? arc4random() : lastpid) % PID_MAX; 330 } while (pidtaken(lastpid)); 331 p2->p_pid = lastpid; 332 333 LIST_INSERT_HEAD(&allproc, p2, p_list); 334 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash); 335 336 #if NSYSTRACE > 0 337 if (ISSET(p1->p_flag, P_SYSTRACE)) 338 systrace_fork(p1, p2); 339 #endif 340 341 /* 342 * Make child runnable, set start time, and add to run queue. 343 */ 344 SCHED_LOCK(s); 345 getmicrotime(&p2->p_stats->p_start); 346 p2->p_acflag = AFORK; 347 p2->p_stat = SRUN; 348 setrunqueue(p2); 349 SCHED_UNLOCK(s); 350 351 timeout_set(&p2->p_stats->p_virt_to, virttimer_trampoline, p2); 352 timeout_set(&p2->p_stats->p_prof_to, proftimer_trampoline, p2); 353 354 /* 355 * Now can be swapped. 356 */ 357 PRELE(p1); 358 359 /* 360 * Notify any interested parties about the new process. 361 */ 362 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid); 363 364 /* 365 * Update stats now that we know the fork was successfull. 366 */ 367 uvmexp.forks++; 368 if (flags & FORK_PPWAIT) 369 uvmexp.forks_ppwait++; 370 if (flags & FORK_SHAREVM) 371 uvmexp.forks_sharevm++; 372 373 /* 374 * Pass a pointer to the new process to the caller. 375 */ 376 if (rnewprocp != NULL) 377 *rnewprocp = p2; 378 379 /* 380 * Preserve synchronization semantics of vfork. If waiting for 381 * child to exec or exit, set P_PPWAIT on child, and sleep on our 382 * proc (in case of exit). 383 */ 384 if (flags & FORK_PPWAIT) 385 while (p2->p_flag & P_PPWAIT) 386 tsleep(p1, PWAIT, "ppwait", 0); 387 388 /* 389 * Return child pid to parent process, 390 * marking us as parent via retval[1]. 391 */ 392 if (retval != NULL) { 393 retval[0] = p2->p_pid; 394 retval[1] = 0; 395 } 396 return (0); 397 } 398 399 /* 400 * Checks for current use of a pid, either as a pid or pgid. 401 */ 402 int 403 pidtaken(pid_t pid) 404 { 405 struct proc *p; 406 407 if (pfind(pid) != NULL) 408 return (1); 409 if (pgfind(pid) != NULL) 410 return (1); 411 LIST_FOREACH(p, &zombproc, p_list) 412 if (p->p_pid == pid || p->p_pgid == pid) 413 return (1); 414 return (0); 415 } 416 417 #if defined(MULTIPROCESSOR) 418 /* 419 * XXX This is a slight hack to get newly-formed processes to 420 * XXX acquire the kernel lock as soon as they run. 421 */ 422 void 423 proc_trampoline_mp(void) 424 { 425 struct proc *p; 426 427 p = curproc; 428 429 SCHED_ASSERT_UNLOCKED(); 430 KERNEL_PROC_LOCK(p); 431 } 432 #endif 433