1 /* $OpenBSD: kern_fork.c,v 1.69 2004/06/13 21:49:26 niklas 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, NULL, retval)); 79 } 80 81 /*ARGSUSED*/ 82 int 83 sys_vfork(struct proc *p, void *v, register_t *retval) 84 { 85 return (fork1(p, SIGCHLD, FORK_VFORK|FORK_PPWAIT, NULL, 0, NULL, 86 NULL, retval)); 87 } 88 89 int 90 sys_rfork(struct proc *p, void *v, register_t *retval) 91 { 92 struct sys_rfork_args /* { 93 syscallarg(int) flags; 94 } */ *uap = v; 95 96 int rforkflags; 97 int flags; 98 99 flags = FORK_RFORK; 100 rforkflags = SCARG(uap, flags); 101 102 if ((rforkflags & RFPROC) == 0) 103 return (EINVAL); 104 105 switch(rforkflags & (RFFDG|RFCFDG)) { 106 case (RFFDG|RFCFDG): 107 return EINVAL; 108 case RFCFDG: 109 flags |= FORK_CLEANFILES; 110 break; 111 case RFFDG: 112 break; 113 default: 114 flags |= FORK_SHAREFILES; 115 break; 116 } 117 118 if (rforkflags & RFNOWAIT) 119 flags |= FORK_NOZOMBIE; 120 121 if (rforkflags & RFMEM) 122 flags |= FORK_SHAREVM; 123 124 return (fork1(p, SIGCHLD, flags, NULL, 0, NULL, NULL, retval)); 125 } 126 127 /* print the 'table full' message once per 10 seconds */ 128 struct timeval fork_tfmrate = { 10, 0 }; 129 130 int 131 fork1(struct proc *p1, int exitsig, int flags, void *stack, size_t stacksize, 132 void (*func)(void *), void *arg, register_t *retval) 133 { 134 struct proc *p2; 135 uid_t uid; 136 struct vmspace *vm; 137 int count; 138 vaddr_t uaddr; 139 int s; 140 extern void endtsleep(void *); 141 extern void realitexpire(void *); 142 143 /* 144 * Although process entries are dynamically created, we still keep 145 * a global limit on the maximum number we will create. We reserve 146 * the last 5 processes to root. The variable nprocs is the current 147 * number of processes, maxproc is the limit. 148 */ 149 uid = p1->p_cred->p_ruid; 150 if ((nprocs >= maxproc - 5 && uid != 0) || nprocs >= maxproc) { 151 static struct timeval lasttfm; 152 153 if (ratecheck(&lasttfm, &fork_tfmrate)) 154 tablefull("proc"); 155 return (EAGAIN); 156 } 157 nprocs++; 158 159 /* 160 * Increment the count of procs running with this uid. Don't allow 161 * a nonprivileged user to exceed their current limit. 162 */ 163 count = chgproccnt(uid, 1); 164 if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) { 165 (void)chgproccnt(uid, -1); 166 nprocs--; 167 return (EAGAIN); 168 } 169 170 /* 171 * Allocate a pcb and kernel stack for the process 172 */ 173 uaddr = uvm_km_valloc(kernel_map, USPACE); 174 if (uaddr == 0) { 175 chgproccnt(uid, -1); 176 nprocs--; 177 return (ENOMEM); 178 } 179 180 /* 181 * From now on, we're committed to the fork and cannot fail. 182 */ 183 184 /* Allocate new proc. */ 185 p2 = pool_get(&proc_pool, PR_WAITOK); 186 187 p2->p_stat = SIDL; /* protect against others */ 188 p2->p_exitsig = exitsig; 189 p2->p_forw = p2->p_back = NULL; 190 191 /* 192 * Make a proc table entry for the new process. 193 * Start by zeroing the section of proc that is zero-initialized, 194 * then copy the section that is copied directly from the parent. 195 */ 196 bzero(&p2->p_startzero, 197 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); 198 bcopy(&p1->p_startcopy, &p2->p_startcopy, 199 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); 200 201 /* 202 * Initialize the timeouts. 203 */ 204 timeout_set(&p2->p_sleep_to, endtsleep, p2); 205 timeout_set(&p2->p_realit_to, realitexpire, p2); 206 207 #if defined(__HAVE_CPUINFO) 208 p2->p_cpu = NULL; 209 #endif 210 211 /* 212 * Duplicate sub-structures as needed. 213 * Increase reference counts on shared objects. 214 * The p_stats and p_sigacts substructs are set in vm_fork. 215 */ 216 p2->p_flag = P_INMEM; 217 p2->p_emul = p1->p_emul; 218 if (p1->p_flag & P_PROFIL) 219 startprofclock(p2); 220 p2->p_flag |= (p1->p_flag & (P_SUGID | P_SUGIDEXEC)); 221 p2->p_cred = pool_get(&pcred_pool, PR_WAITOK); 222 bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred)); 223 p2->p_cred->p_refcnt = 1; 224 crhold(p1->p_ucred); 225 226 /* bump references to the text vnode (for procfs) */ 227 p2->p_textvp = p1->p_textvp; 228 if (p2->p_textvp) 229 VREF(p2->p_textvp); 230 231 if (flags & FORK_CLEANFILES) 232 p2->p_fd = fdinit(p1); 233 else if (flags & FORK_SHAREFILES) 234 p2->p_fd = fdshare(p1); 235 else 236 p2->p_fd = fdcopy(p1); 237 238 /* 239 * If p_limit is still copy-on-write, bump refcnt, 240 * otherwise get a copy that won't be modified. 241 * (If PL_SHAREMOD is clear, the structure is shared 242 * copy-on-write.) 243 */ 244 if (p1->p_limit->p_lflags & PL_SHAREMOD) 245 p2->p_limit = limcopy(p1->p_limit); 246 else { 247 p2->p_limit = p1->p_limit; 248 p2->p_limit->p_refcnt++; 249 } 250 251 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) 252 p2->p_flag |= P_CONTROLT; 253 if (flags & FORK_PPWAIT) 254 p2->p_flag |= P_PPWAIT; 255 LIST_INSERT_AFTER(p1, p2, p_pglist); 256 p2->p_pptr = p1; 257 if (flags & FORK_NOZOMBIE) 258 p2->p_flag |= P_NOZOMBIE; 259 LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling); 260 LIST_INIT(&p2->p_children); 261 262 #ifdef KTRACE 263 /* 264 * Copy traceflag and tracefile if enabled. 265 * If not inherited, these were zeroed above. 266 */ 267 if (p1->p_traceflag & KTRFAC_INHERIT) { 268 p2->p_traceflag = p1->p_traceflag; 269 if ((p2->p_tracep = p1->p_tracep) != NULL) 270 VREF(p2->p_tracep); 271 } 272 #endif 273 274 /* 275 * set priority of child to be that of parent 276 * XXX should move p_estcpu into the region of struct proc which gets 277 * copied. 278 */ 279 scheduler_fork_hook(p1, p2); 280 281 /* 282 * Create signal actions for the child process. 283 */ 284 if (flags & FORK_SIGHAND) 285 sigactsshare(p1, p2); 286 else 287 p2->p_sigacts = sigactsinit(p1); 288 289 /* 290 * If emulation has process fork hook, call it now. 291 */ 292 if (p2->p_emul->e_proc_fork) 293 (*p2->p_emul->e_proc_fork)(p2, p1); 294 /* 295 * This begins the section where we must prevent the parent 296 * from being swapped. 297 */ 298 PHOLD(p1); 299 300 p2->p_addr = (struct user *)uaddr; 301 302 /* 303 * Finish creating the child process. It will return through a 304 * different path later. 305 */ 306 uvm_fork(p1, p2, ((flags & FORK_SHAREVM) ? TRUE : FALSE), stack, 307 stacksize, func ? func : child_return, arg ? arg : p2); 308 309 vm = p2->p_vmspace; 310 311 if (flags & FORK_FORK) { 312 forkstat.cntfork++; 313 forkstat.sizfork += vm->vm_dsize + vm->vm_ssize; 314 } else if (flags & FORK_VFORK) { 315 forkstat.cntvfork++; 316 forkstat.sizvfork += vm->vm_dsize + vm->vm_ssize; 317 } else if (flags & FORK_RFORK) { 318 forkstat.cntrfork++; 319 forkstat.sizrfork += vm->vm_dsize + vm->vm_ssize; 320 } else { 321 forkstat.cntkthread++; 322 forkstat.sizkthread += vm->vm_dsize + vm->vm_ssize; 323 } 324 325 /* Find an unused pid satisfying 1 <= lastpid <= PID_MAX */ 326 do { 327 lastpid = 1 + (randompid ? arc4random() : lastpid) % PID_MAX; 328 } while (pidtaken(lastpid)); 329 p2->p_pid = lastpid; 330 331 LIST_INSERT_HEAD(&allproc, p2, p_list); 332 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash); 333 334 #if NSYSTRACE > 0 335 if (ISSET(p1->p_flag, P_SYSTRACE)) 336 systrace_fork(p1, p2); 337 #endif 338 339 /* 340 * Make child runnable, set start time, and add to run queue. 341 */ 342 SCHED_LOCK(s); 343 p2->p_stats->p_start = time; 344 p2->p_acflag = AFORK; 345 p2->p_stat = SRUN; 346 setrunqueue(p2); 347 SCHED_UNLOCK(s); 348 349 /* 350 * Now can be swapped. 351 */ 352 PRELE(p1); 353 354 uvmexp.forks++; 355 if (flags & FORK_PPWAIT) 356 uvmexp.forks_ppwait++; 357 if (flags & FORK_SHAREVM) 358 uvmexp.forks_sharevm++; 359 360 /* 361 * tell any interested parties about the new process 362 */ 363 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid); 364 365 /* 366 * Preserve synchronization semantics of vfork. If waiting for 367 * child to exec or exit, set P_PPWAIT on child, and sleep on our 368 * proc (in case of exit). 369 */ 370 if (flags & FORK_PPWAIT) 371 while (p2->p_flag & P_PPWAIT) 372 tsleep(p1, PWAIT, "ppwait", 0); 373 374 /* 375 * Return child pid to parent process, 376 * marking us as parent via retval[1]. 377 */ 378 if (retval != NULL) { 379 retval[0] = p2->p_pid; 380 retval[1] = 0; 381 } 382 return (0); 383 } 384 385 /* 386 * Checks for current use of a pid, either as a pid or pgid. 387 */ 388 int 389 pidtaken(pid_t pid) 390 { 391 struct proc *p; 392 393 if (pfind(pid) != NULL) 394 return (1); 395 if (pgfind(pid) != NULL) 396 return (1); 397 LIST_FOREACH(p, &zombproc, p_list) 398 if (p->p_pid == pid || p->p_pgid == pid) 399 return (1); 400 return (0); 401 } 402 403 #if defined(MULTIPROCESSOR) 404 /* 405 * XXX This is a slight hack to get newly-formed processes to 406 * XXX acquire the kernel lock as soon as they run. 407 */ 408 void 409 proc_trampoline_mp(void) 410 { 411 struct proc *p; 412 413 p = curproc; 414 415 SCHED_ASSERT_UNLOCKED(); 416 KERNEL_PROC_LOCK(p); 417 } 418 #endif 419