1 /* $OpenBSD: kern_fork.c,v 1.31 2000/03/23 16:54:44 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_SHAREVM; 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 extern void endtsleep __P((void *)); 152 extern void realitexpire __P((void *)); 153 154 /* 155 * Although process entries are dynamically created, we still keep 156 * a global limit on the maximum number we will create. We reserve 157 * the last 5 processes to root. The variable nprocs is the current 158 * number of processes, maxproc is the limit. 159 */ 160 uid = p1->p_cred->p_ruid; 161 if ((nprocs >= maxproc - 5 && uid != 0) || nprocs >= maxproc) { 162 tablefull("proc"); 163 return (EAGAIN); 164 } 165 166 /* 167 * Increment the count of procs running with this uid. Don't allow 168 * a nonprivileged user to exceed their current limit. 169 */ 170 count = chgproccnt(uid, 1); 171 if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) { 172 (void)chgproccnt(uid, -1); 173 return (EAGAIN); 174 } 175 176 /* 177 * Allocate a pcb and kernel stack for the process 178 */ 179 #if defined(arc) || defined(mips_cachealias) 180 uaddr = kmem_alloc_upage(kernel_map, USPACE); 181 #else 182 #if defined(UVM) 183 uaddr = uvm_km_valloc(kernel_map, USPACE); 184 #else 185 uaddr = kmem_alloc_pageable(kernel_map, USPACE); 186 #endif 187 #endif 188 if (uaddr == 0) 189 return ENOMEM; 190 191 /* Allocate new proc. */ 192 MALLOC(newproc, struct proc *, sizeof(struct proc), M_PROC, M_WAITOK); 193 194 lastpid++; 195 if (randompid) 196 lastpid = PID_MAX; 197 retry: 198 /* 199 * If the process ID prototype has wrapped around, 200 * restart somewhat above 0, as the low-numbered procs 201 * tend to include daemons that don't exit. 202 */ 203 if (lastpid >= PID_MAX) { 204 lastpid = arc4random() % PID_MAX; 205 pidchecked = 0; 206 } 207 if (lastpid >= pidchecked) { 208 int doingzomb = 0; 209 210 pidchecked = PID_MAX; 211 /* 212 * Scan the active and zombie procs to check whether this pid 213 * is in use. Remember the lowest pid that's greater 214 * than lastpid, so we can avoid checking for a while. 215 */ 216 p2 = LIST_FIRST(&allproc); 217 again: 218 for (; p2 != 0; p2 = LIST_NEXT(p2, p_list)) { 219 while (p2->p_pid == lastpid || 220 p2->p_pgrp->pg_id == lastpid) { 221 lastpid++; 222 if (lastpid >= pidchecked) 223 goto retry; 224 } 225 if (p2->p_pid > lastpid && pidchecked > p2->p_pid) 226 pidchecked = p2->p_pid; 227 if (p2->p_pgrp->pg_id > lastpid && 228 pidchecked > p2->p_pgrp->pg_id) 229 pidchecked = p2->p_pgrp->pg_id; 230 } 231 if (!doingzomb) { 232 doingzomb = 1; 233 p2 = LIST_FIRST(&zombproc); 234 goto again; 235 } 236 } 237 238 nprocs++; 239 p2 = newproc; 240 p2->p_stat = SIDL; /* protect against others */ 241 p2->p_pid = lastpid; 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 * This begins the section where we must prevent the parent 335 * from being swapped. 336 */ 337 p1->p_holdcnt++; 338 339 #if !defined(UVM) /* We do this later for UVM */ 340 if (flags & FORK_SHAREVM) { 341 /* share as much address space as possible */ 342 (void) vm_map_inherit(&p1->p_vmspace->vm_map, 343 VM_MIN_ADDRESS, VM_MAXUSER_ADDRESS - MAXSSIZ, 344 VM_INHERIT_SHARE); 345 } 346 #endif 347 348 p2->p_addr = (struct user *)uaddr; 349 350 #ifdef __FORK_BRAINDAMAGE 351 /* 352 * Set return values for child before vm_fork, 353 * so they can be copied to child stack. 354 * We return 0, rather than the traditional behaviour of modifying the 355 * return value in the system call stub. 356 * NOTE: the kernel stack may be at a different location in the child 357 * process, and thus addresses of automatic variables (including retval) 358 * may be invalid after vm_fork returns in the child process. 359 */ 360 retval[0] = 0; 361 retval[1] = 1; 362 if (vm_fork(p1, p2, stack, stacksize)) 363 return (0); 364 #else 365 /* 366 * Finish creating the child process. It will return through a 367 * different path later. 368 */ 369 #if defined(UVM) 370 uvm_fork(p1, p2, ((flags & FORK_SHAREVM) ? TRUE : FALSE), stack, 371 stacksize); 372 #else /* UVM */ 373 vm_fork(p1, p2, stack, stacksize); 374 #endif /* UVM */ 375 #endif 376 vm = p2->p_vmspace; 377 378 if (flags & FORK_FORK) { 379 forkstat.cntfork++; 380 forkstat.sizfork += vm->vm_dsize + vm->vm_ssize; 381 } else if (flags & FORK_VFORK) { 382 forkstat.cntvfork++; 383 forkstat.sizvfork += vm->vm_dsize + vm->vm_ssize; 384 } else if (flags & FORK_RFORK) { 385 forkstat.cntrfork++; 386 forkstat.sizrfork += vm->vm_dsize + vm->vm_ssize; 387 } 388 389 /* 390 * Make child runnable, set start time, and add to run queue. 391 */ 392 (void) splstatclock(); 393 p2->p_stats->p_start = time; 394 p2->p_acflag = AFORK; 395 p2->p_stat = SRUN; 396 setrunqueue(p2); 397 (void) spl0(); 398 399 /* 400 * Now can be swapped. 401 */ 402 p1->p_holdcnt--; 403 404 #if defined(UVM) 405 uvmexp.forks++; 406 if (flags & FORK_PPWAIT) 407 uvmexp.forks_ppwait++; 408 if (flags & FORK_SHAREVM) 409 uvmexp.forks_sharevm++; 410 #endif 411 412 /* 413 * Preserve synchronization semantics of vfork. If waiting for 414 * child to exec or exit, set P_PPWAIT on child, and sleep on our 415 * proc (in case of exit). 416 */ 417 if (flags & FORK_PPWAIT) 418 while (p2->p_flag & P_PPWAIT) 419 tsleep(p1, PWAIT, "ppwait", 0); 420 421 /* 422 * Return child pid to parent process, 423 * marking us as parent via retval[1]. 424 */ 425 retval[0] = p2->p_pid; 426 retval[1] = 0; 427 return (0); 428 } 429 430