1 /* $OpenBSD: kern_fork.c,v 1.40 2001/06/27 04:49:42 art 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 62 #include <sys/syscallargs.h> 63 64 #include <vm/vm.h> 65 #include <vm/vm_kern.h> 66 67 #include <uvm/uvm_extern.h> 68 #include <uvm/uvm_map.h> 69 70 int nprocs = 1; /* process 0 */ 71 int randompid; /* when set to 1, pid's go random */ 72 pid_t lastpid; 73 struct forkstat forkstat; 74 75 76 /*ARGSUSED*/ 77 int 78 sys_fork(p, v, retval) 79 struct proc *p; 80 void *v; 81 register_t *retval; 82 { 83 return (fork1(p, SIGCHLD, FORK_FORK, NULL, 0, retval)); 84 } 85 86 /*ARGSUSED*/ 87 int 88 sys_vfork(p, v, retval) 89 struct proc *p; 90 void *v; 91 register_t *retval; 92 { 93 return (fork1(p, SIGCHLD, FORK_VFORK|FORK_PPWAIT, NULL, 0, retval)); 94 } 95 96 int 97 sys_rfork(p, v, retval) 98 struct proc *p; 99 void *v; 100 register_t *retval; 101 { 102 struct sys_rfork_args /* { 103 syscallarg(int) flags; 104 } */ *uap = v; 105 106 int rforkflags; 107 int flags; 108 109 flags = FORK_RFORK; 110 rforkflags = SCARG(uap, flags); 111 112 if ((rforkflags & RFPROC) == 0) 113 return (EINVAL); 114 115 switch(rforkflags & (RFFDG|RFCFDG)) { 116 case (RFFDG|RFCFDG): 117 return EINVAL; 118 case RFCFDG: 119 flags |= FORK_CLEANFILES; 120 break; 121 case RFFDG: 122 break; 123 default: 124 flags |= FORK_SHAREFILES; 125 break; 126 } 127 128 if (rforkflags & RFNOWAIT) 129 flags |= FORK_NOZOMBIE; 130 131 if (rforkflags & RFMEM) 132 flags |= FORK_VMNOSTACK; 133 134 return (fork1(p, SIGCHLD, flags, NULL, 0, retval)); 135 } 136 137 int 138 fork1(p1, exitsig, flags, stack, stacksize, retval) 139 register struct proc *p1; 140 int exitsig; 141 int flags; 142 void *stack; 143 size_t stacksize; 144 register_t *retval; 145 { 146 struct proc *p2; 147 uid_t uid; 148 struct proc *newproc; 149 struct vmspace *vm; 150 int count; 151 static int pidchecked = 0; 152 vaddr_t uaddr; 153 int s; 154 extern void endtsleep __P((void *)); 155 extern void realitexpire __P((void *)); 156 157 /* 158 * Although process entries are dynamically created, we still keep 159 * a global limit on the maximum number we will create. We reserve 160 * the last 5 processes to root. The variable nprocs is the current 161 * number of processes, maxproc is the limit. 162 */ 163 uid = p1->p_cred->p_ruid; 164 if ((nprocs >= maxproc - 5 && uid != 0) || nprocs >= maxproc) { 165 tablefull("proc"); 166 return (EAGAIN); 167 } 168 169 /* 170 * Increment the count of procs running with this uid. Don't allow 171 * a nonprivileged user to exceed their current limit. 172 */ 173 count = chgproccnt(uid, 1); 174 if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) { 175 (void)chgproccnt(uid, -1); 176 return (EAGAIN); 177 } 178 179 /* 180 * Allocate a pcb and kernel stack for the process 181 */ 182 #if defined(arc) || defined(mips_cachealias) 183 uaddr = kmem_alloc_upage(kernel_map, USPACE); 184 #else 185 uaddr = uvm_km_valloc(kernel_map, USPACE); 186 #endif 187 if (uaddr == 0) 188 return ENOMEM; 189 190 /* Allocate new proc. */ 191 newproc = pool_get(&proc_pool, PR_WAITOK); 192 193 lastpid++; 194 if (randompid) 195 lastpid = PID_MAX; 196 retry: 197 /* 198 * If the process ID prototype has wrapped around, 199 * restart somewhat above 0, as the low-numbered procs 200 * tend to include daemons that don't exit. 201 */ 202 if (lastpid >= PID_MAX) { 203 lastpid = arc4random() % PID_MAX; 204 pidchecked = 0; 205 } 206 if (lastpid >= pidchecked) { 207 int doingzomb = 0; 208 209 pidchecked = PID_MAX; 210 /* 211 * Scan the active and zombie procs to check whether this pid 212 * is in use. Remember the lowest pid that's greater 213 * than lastpid, so we can avoid checking for a while. 214 */ 215 p2 = LIST_FIRST(&allproc); 216 again: 217 for (; p2 != 0; p2 = LIST_NEXT(p2, p_list)) { 218 while (p2->p_pid == lastpid || 219 p2->p_pgrp->pg_id == lastpid) { 220 lastpid++; 221 if (lastpid >= pidchecked) 222 goto retry; 223 } 224 if (p2->p_pid > lastpid && pidchecked > p2->p_pid) 225 pidchecked = p2->p_pid; 226 if (p2->p_pgrp->pg_id > lastpid && 227 pidchecked > p2->p_pgrp->pg_id) 228 pidchecked = p2->p_pgrp->pg_id; 229 } 230 if (!doingzomb) { 231 doingzomb = 1; 232 p2 = LIST_FIRST(&zombproc); 233 goto again; 234 } 235 } 236 237 nprocs++; 238 p2 = newproc; 239 p2->p_stat = SIDL; /* protect against others */ 240 p2->p_pid = lastpid; 241 p2->p_exitsig = exitsig; 242 LIST_INSERT_HEAD(&allproc, p2, p_list); 243 p2->p_forw = p2->p_back = NULL; /* shouldn't be necessary */ 244 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash); 245 246 /* 247 * Make a proc table entry for the new process. 248 * Start by zeroing the section of proc that is zero-initialized, 249 * then copy the section that is copied directly from the parent. 250 */ 251 bzero(&p2->p_startzero, 252 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); 253 bcopy(&p1->p_startcopy, &p2->p_startcopy, 254 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); 255 256 /* 257 * Initialize the timeouts. 258 */ 259 timeout_set(&p2->p_sleep_to, endtsleep, p2); 260 timeout_set(&p2->p_realit_to, realitexpire, p2); 261 262 /* 263 * Duplicate sub-structures as needed. 264 * Increase reference counts on shared objects. 265 * The p_stats and p_sigacts substructs are set in vm_fork. 266 */ 267 p2->p_flag = P_INMEM; 268 p2->p_emul = p1->p_emul; 269 if (p1->p_flag & P_PROFIL) 270 startprofclock(p2); 271 p2->p_flag |= (p1->p_flag & (P_SUGID | P_SUGIDEXEC)); 272 MALLOC(p2->p_cred, struct pcred *, sizeof(struct pcred), 273 M_SUBPROC, M_WAITOK); 274 bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred)); 275 p2->p_cred->p_refcnt = 1; 276 crhold(p1->p_ucred); 277 278 /* bump references to the text vnode (for procfs) */ 279 p2->p_textvp = p1->p_textvp; 280 if (p2->p_textvp) 281 VREF(p2->p_textvp); 282 283 if (flags & FORK_CLEANFILES) 284 p2->p_fd = fdinit(p1); 285 else if (flags & FORK_SHAREFILES) 286 p2->p_fd = fdshare(p1); 287 else 288 p2->p_fd = fdcopy(p1); 289 290 /* 291 * If p_limit is still copy-on-write, bump refcnt, 292 * otherwise get a copy that won't be modified. 293 * (If PL_SHAREMOD is clear, the structure is shared 294 * copy-on-write.) 295 */ 296 if (p1->p_limit->p_lflags & PL_SHAREMOD) 297 p2->p_limit = limcopy(p1->p_limit); 298 else { 299 p2->p_limit = p1->p_limit; 300 p2->p_limit->p_refcnt++; 301 } 302 303 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) 304 p2->p_flag |= P_CONTROLT; 305 if (flags & FORK_PPWAIT) 306 p2->p_flag |= P_PPWAIT; 307 LIST_INSERT_AFTER(p1, p2, p_pglist); 308 p2->p_pptr = p1; 309 if (flags & FORK_NOZOMBIE) 310 p2->p_flag |= P_NOZOMBIE; 311 LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling); 312 LIST_INIT(&p2->p_children); 313 314 #ifdef KTRACE 315 /* 316 * Copy traceflag and tracefile if enabled. 317 * If not inherited, these were zeroed above. 318 */ 319 if (p1->p_traceflag & KTRFAC_INHERIT) { 320 p2->p_traceflag = p1->p_traceflag; 321 if ((p2->p_tracep = p1->p_tracep) != NULL) 322 VREF(p2->p_tracep); 323 } 324 #endif 325 326 /* 327 * set priority of child to be that of parent 328 * XXX should move p_estcpu into the region of struct proc which gets 329 * copied. 330 */ 331 scheduler_fork_hook(p1, p2); 332 333 /* 334 * Create signal actions for the child process. 335 */ 336 if (flags & FORK_SIGHAND) 337 sigactsshare(p1, p2); 338 else 339 p2->p_sigacts = sigactsinit(p1); 340 341 /* 342 * This begins the section where we must prevent the parent 343 * from being swapped. 344 */ 345 PHOLD(p1); 346 347 if (flags & FORK_VMNOSTACK) { 348 /* share as much address space as possible */ 349 (void) uvm_map_inherit(&p1->p_vmspace->vm_map, 350 VM_MIN_ADDRESS, VM_MAXUSER_ADDRESS - MAXSSIZ, 351 VM_INHERIT_SHARE); 352 } 353 354 p2->p_addr = (struct user *)uaddr; 355 356 #ifdef __FORK_BRAINDAMAGE 357 /* 358 * Set return values for child before vm_fork, 359 * so they can be copied to child stack. 360 * We return 0, rather than the traditional behaviour of modifying the 361 * return value in the system call stub. 362 * NOTE: the kernel stack may be at a different location in the child 363 * process, and thus addresses of automatic variables (including retval) 364 * may be invalid after vm_fork returns in the child process. 365 */ 366 retval[0] = 0; 367 retval[1] = 1; 368 if (vm_fork(p1, p2, stack, stacksize)) 369 return (0); 370 #else 371 /* 372 * Finish creating the child process. It will return through a 373 * different path later. 374 */ 375 uvm_fork(p1, p2, ((flags & FORK_SHAREVM) ? TRUE : FALSE), stack, 376 stacksize); 377 #endif 378 vm = p2->p_vmspace; 379 380 if (flags & FORK_FORK) { 381 forkstat.cntfork++; 382 forkstat.sizfork += vm->vm_dsize + vm->vm_ssize; 383 } else if (flags & FORK_VFORK) { 384 forkstat.cntvfork++; 385 forkstat.sizvfork += vm->vm_dsize + vm->vm_ssize; 386 } else if (flags & FORK_RFORK) { 387 forkstat.cntrfork++; 388 forkstat.sizrfork += vm->vm_dsize + vm->vm_ssize; 389 } else { 390 forkstat.cntkthread++; 391 forkstat.sizkthread += vm->vm_dsize + vm->vm_ssize; 392 } 393 394 /* 395 * Make child runnable, set start time, and add to run queue. 396 */ 397 s = splstatclock(); 398 p2->p_stats->p_start = time; 399 p2->p_acflag = AFORK; 400 p2->p_stat = SRUN; 401 setrunqueue(p2); 402 splx(s); 403 404 /* 405 * Now can be swapped. 406 */ 407 PRELE(p1); 408 409 uvmexp.forks++; 410 if (flags & FORK_PPWAIT) 411 uvmexp.forks_ppwait++; 412 if (flags & FORK_SHAREVM) 413 uvmexp.forks_sharevm++; 414 415 /* 416 * tell any interested parties about the new process 417 */ 418 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid); 419 420 /* 421 * Preserve synchronization semantics of vfork. If waiting for 422 * child to exec or exit, set P_PPWAIT on child, and sleep on our 423 * proc (in case of exit). 424 */ 425 if (flags & FORK_PPWAIT) 426 while (p2->p_flag & P_PPWAIT) 427 tsleep(p1, PWAIT, "ppwait", 0); 428 429 /* 430 * Return child pid to parent process, 431 * marking us as parent via retval[1]. 432 */ 433 retval[0] = p2->p_pid; 434 retval[1] = 0; 435 return (0); 436 } 437 438