1 /* $OpenBSD: kern_fork.c,v 1.47 2001/11/06 19:53:20 miod 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. All advertising materials mentioning features or use of this software 22 * must display the following acknowledgement: 23 * This product includes software developed by the University of 24 * California, Berkeley and its contributors. 25 * 4. Neither the name of the University nor the names of its contributors 26 * may be used to endorse or promote products derived from this software 27 * without specific prior written permission. 28 * 29 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 32 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 39 * SUCH DAMAGE. 40 * 41 * @(#)kern_fork.c 8.6 (Berkeley) 4/8/94 42 */ 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/map.h> 47 #include <sys/filedesc.h> 48 #include <sys/kernel.h> 49 #include <sys/malloc.h> 50 #include <sys/mount.h> 51 #include <sys/proc.h> 52 #include <sys/resourcevar.h> 53 #include <sys/signalvar.h> 54 #include <sys/vnode.h> 55 #include <sys/file.h> 56 #include <sys/acct.h> 57 #include <sys/ktrace.h> 58 #include <sys/sched.h> 59 #include <dev/rndvar.h> 60 #include <sys/pool.h> 61 #include <sys/mman.h> 62 63 #include <sys/syscallargs.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 74 /*ARGSUSED*/ 75 int 76 sys_fork(p, v, retval) 77 struct proc *p; 78 void *v; 79 register_t *retval; 80 { 81 return (fork1(p, SIGCHLD, FORK_FORK, NULL, 0, NULL, NULL, retval)); 82 } 83 84 /*ARGSUSED*/ 85 int 86 sys_vfork(p, v, retval) 87 struct proc *p; 88 void *v; 89 register_t *retval; 90 { 91 return (fork1(p, SIGCHLD, FORK_VFORK|FORK_PPWAIT, NULL, 0, NULL, 92 NULL, retval)); 93 } 94 95 int 96 sys_rfork(p, v, retval) 97 struct proc *p; 98 void *v; 99 register_t *retval; 100 { 101 struct sys_rfork_args /* { 102 syscallarg(int) flags; 103 } */ *uap = v; 104 105 int rforkflags; 106 int flags; 107 108 flags = FORK_RFORK; 109 rforkflags = SCARG(uap, flags); 110 111 if ((rforkflags & RFPROC) == 0) 112 return (EINVAL); 113 114 switch(rforkflags & (RFFDG|RFCFDG)) { 115 case (RFFDG|RFCFDG): 116 return EINVAL; 117 case RFCFDG: 118 flags |= FORK_CLEANFILES; 119 break; 120 case RFFDG: 121 break; 122 default: 123 flags |= FORK_SHAREFILES; 124 break; 125 } 126 127 if (rforkflags & RFNOWAIT) 128 flags |= FORK_NOZOMBIE; 129 130 if (rforkflags & RFMEM) 131 flags |= FORK_VMNOSTACK; 132 133 return (fork1(p, SIGCHLD, flags, NULL, 0, NULL, NULL, retval)); 134 } 135 136 int 137 fork1(p1, exitsig, flags, stack, stacksize, func, arg, retval) 138 struct proc *p1; 139 int exitsig; 140 int flags; 141 void *stack; 142 size_t stacksize; 143 void (*func)(void *); 144 void *arg; 145 register_t *retval; 146 { 147 struct proc *p2; 148 uid_t uid; 149 struct proc *newproc; 150 struct vmspace *vm; 151 int count; 152 static int pidchecked = 0; 153 vaddr_t uaddr; 154 int s; 155 extern void endtsleep __P((void *)); 156 extern void realitexpire __P((void *)); 157 158 #ifndef RFORK_FDSHARE 159 /* XXX - Too dangerous right now. */ 160 if (flags & FORK_SHAREFILES) { 161 return (EOPNOTSUPP); 162 } 163 #endif 164 165 /* 166 * Although process entries are dynamically created, we still keep 167 * a global limit on the maximum number we will create. We reserve 168 * the last 5 processes to root. The variable nprocs is the current 169 * number of processes, maxproc is the limit. 170 */ 171 uid = p1->p_cred->p_ruid; 172 if ((nprocs >= maxproc - 5 && uid != 0) || nprocs >= maxproc) { 173 tablefull("proc"); 174 return (EAGAIN); 175 } 176 177 /* 178 * Increment the count of procs running with this uid. Don't allow 179 * a nonprivileged user to exceed their current limit. 180 */ 181 count = chgproccnt(uid, 1); 182 if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) { 183 (void)chgproccnt(uid, -1); 184 return (EAGAIN); 185 } 186 187 /* 188 * Allocate a pcb and kernel stack for the process 189 */ 190 uaddr = uvm_km_valloc(kernel_map, USPACE); 191 if (uaddr == 0) 192 return ENOMEM; 193 194 /* Allocate new proc. */ 195 newproc = pool_get(&proc_pool, PR_WAITOK); 196 197 lastpid++; 198 if (randompid) 199 lastpid = PID_MAX; 200 retry: 201 /* 202 * If the process ID prototype has wrapped around, 203 * restart somewhat above 0, as the low-numbered procs 204 * tend to include daemons that don't exit. 205 */ 206 if (lastpid >= PID_MAX) { 207 lastpid = arc4random() % PID_MAX; 208 pidchecked = 0; 209 } 210 if (lastpid >= pidchecked) { 211 int doingzomb = 0; 212 213 pidchecked = PID_MAX; 214 /* 215 * Scan the active and zombie procs to check whether this pid 216 * is in use. Remember the lowest pid that's greater 217 * than lastpid, so we can avoid checking for a while. 218 */ 219 p2 = LIST_FIRST(&allproc); 220 again: 221 for (; p2 != 0; p2 = LIST_NEXT(p2, p_list)) { 222 while (p2->p_pid == lastpid || 223 p2->p_pgrp->pg_id == lastpid) { 224 lastpid++; 225 if (lastpid >= pidchecked) 226 goto retry; 227 } 228 if (p2->p_pid > lastpid && pidchecked > p2->p_pid) 229 pidchecked = p2->p_pid; 230 if (p2->p_pgrp->pg_id > lastpid && 231 pidchecked > p2->p_pgrp->pg_id) 232 pidchecked = p2->p_pgrp->pg_id; 233 } 234 if (!doingzomb) { 235 doingzomb = 1; 236 p2 = LIST_FIRST(&zombproc); 237 goto again; 238 } 239 } 240 241 nprocs++; 242 p2 = newproc; 243 p2->p_stat = SIDL; /* protect against others */ 244 p2->p_pid = lastpid; 245 p2->p_exitsig = exitsig; 246 LIST_INSERT_HEAD(&allproc, p2, p_list); 247 p2->p_forw = p2->p_back = NULL; /* shouldn't be necessary */ 248 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash); 249 250 /* 251 * Make a proc table entry for the new process. 252 * Start by zeroing the section of proc that is zero-initialized, 253 * then copy the section that is copied directly from the parent. 254 */ 255 bzero(&p2->p_startzero, 256 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); 257 bcopy(&p1->p_startcopy, &p2->p_startcopy, 258 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); 259 260 /* 261 * Initialize the timeouts. 262 */ 263 timeout_set(&p2->p_sleep_to, endtsleep, p2); 264 timeout_set(&p2->p_realit_to, realitexpire, p2); 265 266 /* 267 * Duplicate sub-structures as needed. 268 * Increase reference counts on shared objects. 269 * The p_stats and p_sigacts substructs are set in vm_fork. 270 */ 271 p2->p_flag = P_INMEM; 272 p2->p_emul = p1->p_emul; 273 if (p1->p_flag & P_PROFIL) 274 startprofclock(p2); 275 p2->p_flag |= (p1->p_flag & (P_SUGID | P_SUGIDEXEC)); 276 MALLOC(p2->p_cred, struct pcred *, sizeof(struct pcred), 277 M_SUBPROC, M_WAITOK); 278 bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred)); 279 p2->p_cred->p_refcnt = 1; 280 crhold(p1->p_ucred); 281 282 /* bump references to the text vnode (for procfs) */ 283 p2->p_textvp = p1->p_textvp; 284 if (p2->p_textvp) 285 VREF(p2->p_textvp); 286 287 if (flags & FORK_CLEANFILES) 288 p2->p_fd = fdinit(p1); 289 else if (flags & FORK_SHAREFILES) 290 p2->p_fd = fdshare(p1); 291 else 292 p2->p_fd = fdcopy(p1); 293 294 /* 295 * If p_limit is still copy-on-write, bump refcnt, 296 * otherwise get a copy that won't be modified. 297 * (If PL_SHAREMOD is clear, the structure is shared 298 * copy-on-write.) 299 */ 300 if (p1->p_limit->p_lflags & PL_SHAREMOD) 301 p2->p_limit = limcopy(p1->p_limit); 302 else { 303 p2->p_limit = p1->p_limit; 304 p2->p_limit->p_refcnt++; 305 } 306 307 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) 308 p2->p_flag |= P_CONTROLT; 309 if (flags & FORK_PPWAIT) 310 p2->p_flag |= P_PPWAIT; 311 LIST_INSERT_AFTER(p1, p2, p_pglist); 312 p2->p_pptr = p1; 313 if (flags & FORK_NOZOMBIE) 314 p2->p_flag |= P_NOZOMBIE; 315 LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling); 316 LIST_INIT(&p2->p_children); 317 318 #ifdef KTRACE 319 /* 320 * Copy traceflag and tracefile if enabled. 321 * If not inherited, these were zeroed above. 322 */ 323 if (p1->p_traceflag & KTRFAC_INHERIT) { 324 p2->p_traceflag = p1->p_traceflag; 325 if ((p2->p_tracep = p1->p_tracep) != NULL) 326 VREF(p2->p_tracep); 327 } 328 #endif 329 330 /* 331 * set priority of child to be that of parent 332 * XXX should move p_estcpu into the region of struct proc which gets 333 * copied. 334 */ 335 scheduler_fork_hook(p1, p2); 336 337 /* 338 * Create signal actions for the child process. 339 */ 340 if (flags & FORK_SIGHAND) 341 sigactsshare(p1, p2); 342 else 343 p2->p_sigacts = sigactsinit(p1); 344 345 /* 346 * This begins the section where we must prevent the parent 347 * from being swapped. 348 */ 349 PHOLD(p1); 350 351 if (flags & FORK_VMNOSTACK) { 352 /* share as much address space as possible */ 353 (void) uvm_map_inherit(&p1->p_vmspace->vm_map, 354 VM_MIN_ADDRESS, VM_MAXUSER_ADDRESS - MAXSSIZ, 355 MAP_INHERIT_SHARE); 356 } 357 358 p2->p_addr = (struct user *)uaddr; 359 360 /* 361 * Finish creating the child process. It will return through a 362 * different path later. 363 */ 364 uvm_fork(p1, p2, ((flags & FORK_SHAREVM) ? TRUE : FALSE), stack, 365 stacksize, func ? func : child_return, arg ? arg : p2); 366 367 vm = p2->p_vmspace; 368 369 if (flags & FORK_FORK) { 370 forkstat.cntfork++; 371 forkstat.sizfork += vm->vm_dsize + vm->vm_ssize; 372 } else if (flags & FORK_VFORK) { 373 forkstat.cntvfork++; 374 forkstat.sizvfork += vm->vm_dsize + vm->vm_ssize; 375 } else if (flags & FORK_RFORK) { 376 forkstat.cntrfork++; 377 forkstat.sizrfork += vm->vm_dsize + vm->vm_ssize; 378 } else { 379 forkstat.cntkthread++; 380 forkstat.sizkthread += vm->vm_dsize + vm->vm_ssize; 381 } 382 383 /* 384 * Make child runnable, set start time, and add to run queue. 385 */ 386 s = splstatclock(); 387 p2->p_stats->p_start = time; 388 p2->p_acflag = AFORK; 389 p2->p_stat = SRUN; 390 setrunqueue(p2); 391 splx(s); 392 393 /* 394 * Now can be swapped. 395 */ 396 PRELE(p1); 397 398 uvmexp.forks++; 399 if (flags & FORK_PPWAIT) 400 uvmexp.forks_ppwait++; 401 if (flags & FORK_SHAREVM) 402 uvmexp.forks_sharevm++; 403 404 /* 405 * tell any interested parties about the new process 406 */ 407 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid); 408 409 /* 410 * Preserve synchronization semantics of vfork. If waiting for 411 * child to exec or exit, set P_PPWAIT on child, and sleep on our 412 * proc (in case of exit). 413 */ 414 if (flags & FORK_PPWAIT) 415 while (p2->p_flag & P_PPWAIT) 416 tsleep(p1, PWAIT, "ppwait", 0); 417 418 /* 419 * Return child pid to parent process, 420 * marking us as parent via retval[1]. 421 */ 422 retval[0] = p2->p_pid; 423 retval[1] = 0; 424 return (0); 425 } 426 427