1 /* $OpenBSD: kern_fork.c,v 1.37 2001/02/13 21:00:48 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/vnode.h> 54 #include <sys/file.h> 55 #include <sys/acct.h> 56 #include <sys/ktrace.h> 57 #include <sys/sched.h> 58 #include <dev/rndvar.h> 59 60 #include <sys/syscallargs.h> 61 62 #include <vm/vm.h> 63 #include <vm/vm_kern.h> 64 65 #if defined(UVM) 66 #include <uvm/uvm_extern.h> 67 #include <uvm/uvm_map.h> 68 #endif 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, 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, 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 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, flags, NULL, 0, retval)); 134 } 135 136 int 137 fork1(p1, flags, stack, stacksize, retval) 138 register struct proc *p1; 139 int flags; 140 void *stack; 141 size_t stacksize; 142 register_t *retval; 143 { 144 struct proc *p2; 145 uid_t uid; 146 struct proc *newproc; 147 struct vmspace *vm; 148 int count; 149 static int pidchecked = 0; 150 vaddr_t uaddr; 151 int s; 152 extern void endtsleep __P((void *)); 153 extern void realitexpire __P((void *)); 154 155 /* 156 * Although process entries are dynamically created, we still keep 157 * a global limit on the maximum number we will create. We reserve 158 * the last 5 processes to root. The variable nprocs is the current 159 * number of processes, maxproc is the limit. 160 */ 161 uid = p1->p_cred->p_ruid; 162 if ((nprocs >= maxproc - 5 && uid != 0) || nprocs >= maxproc) { 163 tablefull("proc"); 164 return (EAGAIN); 165 } 166 167 #if !defined(UVM) 168 if (flags & FORK_SHAREVM) 169 return (EINVAL); 170 #endif 171 172 /* 173 * Increment the count of procs running with this uid. Don't allow 174 * a nonprivileged user to exceed their current limit. 175 */ 176 count = chgproccnt(uid, 1); 177 if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) { 178 (void)chgproccnt(uid, -1); 179 return (EAGAIN); 180 } 181 182 /* 183 * Allocate a pcb and kernel stack for the process 184 */ 185 #if defined(arc) || defined(mips_cachealias) 186 uaddr = kmem_alloc_upage(kernel_map, USPACE); 187 #else 188 #if defined(UVM) 189 uaddr = uvm_km_valloc(kernel_map, USPACE); 190 #else 191 uaddr = kmem_alloc_pageable(kernel_map, USPACE); 192 #endif 193 #endif 194 if (uaddr == 0) 195 return ENOMEM; 196 197 /* Allocate new proc. */ 198 MALLOC(newproc, struct proc *, sizeof(struct proc), M_PROC, M_WAITOK); 199 200 lastpid++; 201 if (randompid) 202 lastpid = PID_MAX; 203 retry: 204 /* 205 * If the process ID prototype has wrapped around, 206 * restart somewhat above 0, as the low-numbered procs 207 * tend to include daemons that don't exit. 208 */ 209 if (lastpid >= PID_MAX) { 210 lastpid = arc4random() % PID_MAX; 211 pidchecked = 0; 212 } 213 if (lastpid >= pidchecked) { 214 int doingzomb = 0; 215 216 pidchecked = PID_MAX; 217 /* 218 * Scan the active and zombie procs to check whether this pid 219 * is in use. Remember the lowest pid that's greater 220 * than lastpid, so we can avoid checking for a while. 221 */ 222 p2 = LIST_FIRST(&allproc); 223 again: 224 for (; p2 != 0; p2 = LIST_NEXT(p2, p_list)) { 225 while (p2->p_pid == lastpid || 226 p2->p_pgrp->pg_id == lastpid) { 227 lastpid++; 228 if (lastpid >= pidchecked) 229 goto retry; 230 } 231 if (p2->p_pid > lastpid && pidchecked > p2->p_pid) 232 pidchecked = p2->p_pid; 233 if (p2->p_pgrp->pg_id > lastpid && 234 pidchecked > p2->p_pgrp->pg_id) 235 pidchecked = p2->p_pgrp->pg_id; 236 } 237 if (!doingzomb) { 238 doingzomb = 1; 239 p2 = LIST_FIRST(&zombproc); 240 goto again; 241 } 242 } 243 244 nprocs++; 245 p2 = newproc; 246 p2->p_stat = SIDL; /* protect against others */ 247 p2->p_pid = lastpid; 248 LIST_INSERT_HEAD(&allproc, p2, p_list); 249 p2->p_forw = p2->p_back = NULL; /* shouldn't be necessary */ 250 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash); 251 252 /* 253 * Make a proc table entry for the new process. 254 * Start by zeroing the section of proc that is zero-initialized, 255 * then copy the section that is copied directly from the parent. 256 */ 257 bzero(&p2->p_startzero, 258 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); 259 bcopy(&p1->p_startcopy, &p2->p_startcopy, 260 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); 261 262 /* 263 * Initialize the timeouts. 264 */ 265 timeout_set(&p2->p_sleep_to, endtsleep, p2); 266 timeout_set(&p2->p_realit_to, realitexpire, p2); 267 268 /* 269 * Duplicate sub-structures as needed. 270 * Increase reference counts on shared objects. 271 * The p_stats and p_sigacts substructs are set in vm_fork. 272 */ 273 p2->p_flag = P_INMEM; 274 p2->p_emul = p1->p_emul; 275 if (p1->p_flag & P_PROFIL) 276 startprofclock(p2); 277 p2->p_flag |= (p1->p_flag & (P_SUGID | P_SUGIDEXEC)); 278 MALLOC(p2->p_cred, struct pcred *, sizeof(struct pcred), 279 M_SUBPROC, M_WAITOK); 280 bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred)); 281 p2->p_cred->p_refcnt = 1; 282 crhold(p1->p_ucred); 283 284 /* bump references to the text vnode (for procfs) */ 285 p2->p_textvp = p1->p_textvp; 286 if (p2->p_textvp) 287 VREF(p2->p_textvp); 288 289 if (flags & FORK_CLEANFILES) 290 p2->p_fd = fdinit(p1); 291 else if (flags & FORK_SHAREFILES) 292 p2->p_fd = fdshare(p1); 293 else 294 p2->p_fd = fdcopy(p1); 295 296 /* 297 * If p_limit is still copy-on-write, bump refcnt, 298 * otherwise get a copy that won't be modified. 299 * (If PL_SHAREMOD is clear, the structure is shared 300 * copy-on-write.) 301 */ 302 if (p1->p_limit->p_lflags & PL_SHAREMOD) 303 p2->p_limit = limcopy(p1->p_limit); 304 else { 305 p2->p_limit = p1->p_limit; 306 p2->p_limit->p_refcnt++; 307 } 308 309 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) 310 p2->p_flag |= P_CONTROLT; 311 if (flags & FORK_PPWAIT) 312 p2->p_flag |= P_PPWAIT; 313 LIST_INSERT_AFTER(p1, p2, p_pglist); 314 p2->p_pptr = p1; 315 if (flags & FORK_NOZOMBIE) 316 p2->p_flag |= P_NOZOMBIE; 317 LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling); 318 LIST_INIT(&p2->p_children); 319 320 #ifdef KTRACE 321 /* 322 * Copy traceflag and tracefile if enabled. 323 * If not inherited, these were zeroed above. 324 */ 325 if (p1->p_traceflag & KTRFAC_INHERIT) { 326 p2->p_traceflag = p1->p_traceflag; 327 if ((p2->p_tracep = p1->p_tracep) != NULL) 328 VREF(p2->p_tracep); 329 } 330 #endif 331 332 /* 333 * set priority of child to be that of parent 334 * XXX should move p_estcpu into the region of struct proc which gets 335 * copied. 336 */ 337 scheduler_fork_hook(p1, p2); 338 339 /* 340 * This begins the section where we must prevent the parent 341 * from being swapped. 342 */ 343 PHOLD(p1); 344 345 #if defined(UVM) 346 if (flags & FORK_VMNOSTACK) { 347 /* share as much address space as possible */ 348 (void) uvm_map_inherit(&p1->p_vmspace->vm_map, 349 VM_MIN_ADDRESS, VM_MAXUSER_ADDRESS - MAXSSIZ, 350 VM_INHERIT_SHARE); 351 } 352 #else 353 if (flags & FORK_VMNOSTACK) { 354 /* share as much address space as possible */ 355 (void) vm_map_inherit(&p1->p_vmspace->vm_map, 356 VM_MIN_ADDRESS, VM_MAXUSER_ADDRESS - MAXSSIZ, 357 VM_INHERIT_SHARE); 358 } 359 #endif 360 361 p2->p_addr = (struct user *)uaddr; 362 363 #ifdef __FORK_BRAINDAMAGE 364 /* 365 * Set return values for child before vm_fork, 366 * so they can be copied to child stack. 367 * We return 0, rather than the traditional behaviour of modifying the 368 * return value in the system call stub. 369 * NOTE: the kernel stack may be at a different location in the child 370 * process, and thus addresses of automatic variables (including retval) 371 * may be invalid after vm_fork returns in the child process. 372 */ 373 retval[0] = 0; 374 retval[1] = 1; 375 if (vm_fork(p1, p2, stack, stacksize)) 376 return (0); 377 #else 378 /* 379 * Finish creating the child process. It will return through a 380 * different path later. 381 */ 382 #if defined(UVM) 383 uvm_fork(p1, p2, ((flags & FORK_SHAREVM) ? TRUE : FALSE), stack, 384 stacksize); 385 #else /* UVM */ 386 vm_fork(p1, p2, stack, stacksize); 387 #endif /* UVM */ 388 #endif 389 vm = p2->p_vmspace; 390 391 if (flags & FORK_FORK) { 392 forkstat.cntfork++; 393 forkstat.sizfork += vm->vm_dsize + vm->vm_ssize; 394 } else if (flags & FORK_VFORK) { 395 forkstat.cntvfork++; 396 forkstat.sizvfork += vm->vm_dsize + vm->vm_ssize; 397 } else if (flags & FORK_RFORK) { 398 forkstat.cntrfork++; 399 forkstat.sizrfork += vm->vm_dsize + vm->vm_ssize; 400 } else { 401 forkstat.cntkthread++; 402 forkstat.sizkthread += vm->vm_dsize + vm->vm_ssize; 403 } 404 405 /* 406 * Make child runnable, set start time, and add to run queue. 407 */ 408 s = splstatclock(); 409 p2->p_stats->p_start = time; 410 p2->p_acflag = AFORK; 411 p2->p_stat = SRUN; 412 setrunqueue(p2); 413 splx(s); 414 415 /* 416 * Now can be swapped. 417 */ 418 PRELE(p1); 419 420 #if defined(UVM) 421 uvmexp.forks++; 422 if (flags & FORK_PPWAIT) 423 uvmexp.forks_ppwait++; 424 if (flags & FORK_SHAREVM) 425 uvmexp.forks_sharevm++; 426 #endif 427 428 /* 429 * tell any interested parties about the new process 430 */ 431 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid); 432 433 /* 434 * Preserve synchronization semantics of vfork. If waiting for 435 * child to exec or exit, set P_PPWAIT on child, and sleep on our 436 * proc (in case of exit). 437 */ 438 if (flags & FORK_PPWAIT) 439 while (p2->p_flag & P_PPWAIT) 440 tsleep(p1, PWAIT, "ppwait", 0); 441 442 /* 443 * Return child pid to parent process, 444 * marking us as parent via retval[1]. 445 */ 446 retval[0] = p2->p_pid; 447 retval[1] = 0; 448 return (0); 449 } 450 451