1 /* $OpenBSD: kern_fork.c,v 1.55 2002/05/16 16:16:51 provos 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/filedesc.h> 47 #include <sys/kernel.h> 48 #include <sys/malloc.h> 49 #include <sys/mount.h> 50 #include <sys/proc.h> 51 #include <sys/resourcevar.h> 52 #include <sys/signalvar.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 #include <sys/pool.h> 60 #include <sys/mman.h> 61 62 #include <sys/syscallargs.h> 63 64 #include "systrace.h" 65 #include <dev/systrace.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 int pidtaken(pid_t); 76 77 /*ARGSUSED*/ 78 int 79 sys_fork(p, v, retval) 80 struct proc *p; 81 void *v; 82 register_t *retval; 83 { 84 return (fork1(p, SIGCHLD, FORK_FORK, NULL, 0, NULL, NULL, retval)); 85 } 86 87 /*ARGSUSED*/ 88 int 89 sys_vfork(p, v, retval) 90 struct proc *p; 91 void *v; 92 register_t *retval; 93 { 94 return (fork1(p, SIGCHLD, FORK_VFORK|FORK_PPWAIT, NULL, 0, NULL, 95 NULL, retval)); 96 } 97 98 int 99 sys_rfork(p, v, retval) 100 struct proc *p; 101 void *v; 102 register_t *retval; 103 { 104 struct sys_rfork_args /* { 105 syscallarg(int) flags; 106 } */ *uap = v; 107 108 int rforkflags; 109 int flags; 110 111 flags = FORK_RFORK; 112 rforkflags = SCARG(uap, flags); 113 114 if ((rforkflags & RFPROC) == 0) 115 return (EINVAL); 116 117 switch(rforkflags & (RFFDG|RFCFDG)) { 118 case (RFFDG|RFCFDG): 119 return EINVAL; 120 case RFCFDG: 121 flags |= FORK_CLEANFILES; 122 break; 123 case RFFDG: 124 break; 125 default: 126 flags |= FORK_SHAREFILES; 127 break; 128 } 129 130 if (rforkflags & RFNOWAIT) 131 flags |= FORK_NOZOMBIE; 132 133 if (rforkflags & RFMEM) 134 flags |= FORK_VMNOSTACK; 135 136 return (fork1(p, SIGCHLD, flags, NULL, 0, NULL, NULL, retval)); 137 } 138 139 int 140 fork1(p1, exitsig, flags, stack, stacksize, func, arg, retval) 141 struct proc *p1; 142 int exitsig; 143 int flags; 144 void *stack; 145 size_t stacksize; 146 void (*func)(void *); 147 void *arg; 148 register_t *retval; 149 { 150 struct proc *p2; 151 uid_t uid; 152 struct proc *newproc; 153 struct vmspace *vm; 154 int count; 155 vaddr_t uaddr; 156 int s; 157 extern void endtsleep(void *); 158 extern void realitexpire(void *); 159 160 /* 161 * Although process entries are dynamically created, we still keep 162 * a global limit on the maximum number we will create. We reserve 163 * the last 5 processes to root. The variable nprocs is the current 164 * number of processes, maxproc is the limit. 165 */ 166 uid = p1->p_cred->p_ruid; 167 if ((nprocs >= maxproc - 5 && uid != 0) || nprocs >= maxproc) { 168 tablefull("proc"); 169 return (EAGAIN); 170 } 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 uaddr = uvm_km_valloc(kernel_map, USPACE); 186 if (uaddr == 0) 187 return ENOMEM; 188 189 /* Allocate new proc. */ 190 newproc = pool_get(&proc_pool, PR_WAITOK); 191 192 /* Find an unused pid satisfying 1 <= lastpid <= PID_MAX */ 193 do { 194 lastpid = 1 + (randompid ? arc4random() : lastpid) % PID_MAX; 195 } while (pidtaken(lastpid)); 196 197 nprocs++; 198 p2 = newproc; 199 p2->p_stat = SIDL; /* protect against others */ 200 p2->p_pid = lastpid; 201 p2->p_exitsig = exitsig; 202 LIST_INSERT_HEAD(&allproc, p2, p_list); 203 p2->p_forw = p2->p_back = NULL; /* shouldn't be necessary */ 204 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash); 205 206 /* 207 * Make a proc table entry for the new process. 208 * Start by zeroing the section of proc that is zero-initialized, 209 * then copy the section that is copied directly from the parent. 210 */ 211 bzero(&p2->p_startzero, 212 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); 213 bcopy(&p1->p_startcopy, &p2->p_startcopy, 214 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); 215 216 /* 217 * Initialize the timeouts. 218 */ 219 timeout_set(&p2->p_sleep_to, endtsleep, p2); 220 timeout_set(&p2->p_realit_to, realitexpire, p2); 221 222 /* 223 * Duplicate sub-structures as needed. 224 * Increase reference counts on shared objects. 225 * The p_stats and p_sigacts substructs are set in vm_fork. 226 */ 227 p2->p_flag = P_INMEM; 228 p2->p_emul = p1->p_emul; 229 if (p1->p_flag & P_PROFIL) 230 startprofclock(p2); 231 p2->p_flag |= (p1->p_flag & (P_SUGID | P_SUGIDEXEC)); 232 p2->p_cred = pool_get(&pcred_pool, PR_WAITOK); 233 bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred)); 234 p2->p_cred->p_refcnt = 1; 235 crhold(p1->p_ucred); 236 237 /* bump references to the text vnode (for procfs) */ 238 p2->p_textvp = p1->p_textvp; 239 if (p2->p_textvp) 240 VREF(p2->p_textvp); 241 242 if (flags & FORK_CLEANFILES) 243 p2->p_fd = fdinit(p1); 244 else if (flags & FORK_SHAREFILES) 245 p2->p_fd = fdshare(p1); 246 else 247 p2->p_fd = fdcopy(p1); 248 249 /* 250 * If p_limit is still copy-on-write, bump refcnt, 251 * otherwise get a copy that won't be modified. 252 * (If PL_SHAREMOD is clear, the structure is shared 253 * copy-on-write.) 254 */ 255 if (p1->p_limit->p_lflags & PL_SHAREMOD) 256 p2->p_limit = limcopy(p1->p_limit); 257 else { 258 p2->p_limit = p1->p_limit; 259 p2->p_limit->p_refcnt++; 260 } 261 262 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) 263 p2->p_flag |= P_CONTROLT; 264 if (flags & FORK_PPWAIT) 265 p2->p_flag |= P_PPWAIT; 266 LIST_INSERT_AFTER(p1, p2, p_pglist); 267 p2->p_pptr = p1; 268 if (flags & FORK_NOZOMBIE) 269 p2->p_flag |= P_NOZOMBIE; 270 LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling); 271 LIST_INIT(&p2->p_children); 272 273 #ifdef KTRACE 274 /* 275 * Copy traceflag and tracefile if enabled. 276 * If not inherited, these were zeroed above. 277 */ 278 if (p1->p_traceflag & KTRFAC_INHERIT) { 279 p2->p_traceflag = p1->p_traceflag; 280 if ((p2->p_tracep = p1->p_tracep) != NULL) 281 VREF(p2->p_tracep); 282 } 283 #endif 284 #if NSYSTRACE > 0 285 if (ISSET(p1->p_flag, P_SYSTRACE)) 286 systrace_fork(p1, p2); 287 #endif 288 289 /* 290 * set priority of child to be that of parent 291 * XXX should move p_estcpu into the region of struct proc which gets 292 * copied. 293 */ 294 scheduler_fork_hook(p1, p2); 295 296 /* 297 * Create signal actions for the child process. 298 */ 299 if (flags & FORK_SIGHAND) 300 sigactsshare(p1, p2); 301 else 302 p2->p_sigacts = sigactsinit(p1); 303 304 /* 305 * This begins the section where we must prevent the parent 306 * from being swapped. 307 */ 308 PHOLD(p1); 309 310 if (flags & FORK_VMNOSTACK) { 311 /* share everything, but ... */ 312 uvm_map_inherit(&p1->p_vmspace->vm_map, 313 VM_MIN_ADDRESS, VM_MAXUSER_ADDRESS, 314 MAP_INHERIT_SHARE); 315 /* ... don't share stack */ 316 #ifdef MACHINE_STACK_GROWS_UP 317 uvm_map_inherit(&p1->p_vmspace->vm_map, 318 USRSTACK, USRSTACK + MAXSSIZ, 319 MAP_INHERIT_COPY); 320 #else 321 uvm_map_inherit(&p1->p_vmspace->vm_map, 322 USRSTACK - MAXSSIZ, USRSTACK, 323 MAP_INHERIT_COPY); 324 #endif 325 } 326 327 p2->p_addr = (struct user *)uaddr; 328 329 /* 330 * Finish creating the child process. It will return through a 331 * different path later. 332 */ 333 uvm_fork(p1, p2, ((flags & FORK_SHAREVM) ? TRUE : FALSE), stack, 334 stacksize, func ? func : child_return, arg ? arg : p2); 335 336 vm = p2->p_vmspace; 337 338 if (flags & FORK_FORK) { 339 forkstat.cntfork++; 340 forkstat.sizfork += vm->vm_dsize + vm->vm_ssize; 341 } else if (flags & FORK_VFORK) { 342 forkstat.cntvfork++; 343 forkstat.sizvfork += vm->vm_dsize + vm->vm_ssize; 344 } else if (flags & FORK_RFORK) { 345 forkstat.cntrfork++; 346 forkstat.sizrfork += vm->vm_dsize + vm->vm_ssize; 347 } else { 348 forkstat.cntkthread++; 349 forkstat.sizkthread += vm->vm_dsize + vm->vm_ssize; 350 } 351 352 /* 353 * Make child runnable, set start time, and add to run queue. 354 */ 355 s = splstatclock(); 356 p2->p_stats->p_start = time; 357 p2->p_acflag = AFORK; 358 p2->p_stat = SRUN; 359 setrunqueue(p2); 360 splx(s); 361 362 /* 363 * Now can be swapped. 364 */ 365 PRELE(p1); 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 * tell any interested parties about the new process 375 */ 376 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid); 377 378 /* 379 * Preserve synchronization semantics of vfork. If waiting for 380 * child to exec or exit, set P_PPWAIT on child, and sleep on our 381 * proc (in case of exit). 382 */ 383 if (flags & FORK_PPWAIT) 384 while (p2->p_flag & P_PPWAIT) 385 tsleep(p1, PWAIT, "ppwait", 0); 386 387 /* 388 * Return child pid to parent process, 389 * marking us as parent via retval[1]. 390 */ 391 retval[0] = p2->p_pid; 392 retval[1] = 0; 393 return (0); 394 } 395 396 /* 397 * Checks for current use of a pid, either as a pid or pgid. 398 */ 399 int 400 pidtaken(pid_t pid) 401 { 402 struct proc *p; 403 404 if (pfind(pid) != NULL) 405 return (1); 406 if (pgfind(pid) != NULL) 407 return (1); 408 LIST_FOREACH(p, &zombproc, p_list) 409 if (p->p_pid == pid || p->p_pgid == pid) 410 return (1); 411 return (0); 412 } 413