kern_fork.c revision 1.51
1 /* $NetBSD: kern_fork.c,v 1.51 1999/01/23 17:02:35 sommerfe Exp $ */ 2 3 /* 4 * Copyright (c) 1982, 1986, 1989, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. All advertising materials mentioning features or use of this software 21 * must display the following acknowledgement: 22 * This product includes software developed by the University of 23 * California, Berkeley and its contributors. 24 * 4. Neither the name of the University nor the names of its contributors 25 * may be used to endorse or promote products derived from this software 26 * without specific prior written permission. 27 * 28 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 31 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 38 * SUCH DAMAGE. 39 * 40 * @(#)kern_fork.c 8.8 (Berkeley) 2/14/95 41 */ 42 43 #include "opt_ktrace.h" 44 #include "opt_uvm.h" 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/map.h> 49 #include <sys/filedesc.h> 50 #include <sys/kernel.h> 51 #include <sys/malloc.h> 52 #include <sys/pool.h> 53 #include <sys/mount.h> 54 #include <sys/proc.h> 55 #include <sys/resourcevar.h> 56 #include <sys/vnode.h> 57 #include <sys/file.h> 58 #include <sys/acct.h> 59 #include <sys/ktrace.h> 60 #include <sys/vmmeter.h> 61 62 #include <sys/syscallargs.h> 63 64 #include <vm/vm.h> 65 #include <vm/vm_kern.h> 66 67 #if defined(UVM) 68 #include <uvm/uvm_extern.h> 69 #endif 70 71 int nprocs = 1; /* process 0 */ 72 73 /*ARGSUSED*/ 74 int 75 sys_fork(p, v, retval) 76 struct proc *p; 77 void *v; 78 register_t *retval; 79 { 80 81 return (fork1(p, 0, retval, NULL)); 82 } 83 84 /* 85 * vfork(2) system call compatible with 4.4BSD (i.e. BSD with Mach VM). 86 * Address space is not shared, but parent is blocked until child exit. 87 */ 88 /*ARGSUSED*/ 89 int 90 sys_vfork(p, v, retval) 91 struct proc *p; 92 void *v; 93 register_t *retval; 94 { 95 96 return (fork1(p, FORK_PPWAIT, retval, NULL)); 97 } 98 99 /* 100 * New vfork(2) system call for NetBSD, which implements original 3BSD vfork(2) 101 * semantics. Address space is shared, and parent is blocked until child exit. 102 */ 103 /*ARGSUSED*/ 104 int 105 sys___vfork14(p, v, retval) 106 struct proc *p; 107 void *v; 108 register_t *retval; 109 { 110 111 return (fork1(p, FORK_PPWAIT|FORK_SHAREVM, retval, NULL)); 112 } 113 114 int slowchild = 1; 115 116 int 117 fork1(p1, flags, retval, rnewprocp) 118 register struct proc *p1; 119 int flags; 120 register_t *retval; 121 struct proc **rnewprocp; 122 { 123 register struct proc *p2; 124 register uid_t uid; 125 struct proc *newproc; 126 int count, s; 127 vaddr_t uaddr; 128 static int nextpid, pidchecked = 0; 129 130 /* 131 * Although process entries are dynamically created, we still keep 132 * a global limit on the maximum number we will create. Don't allow 133 * a nonprivileged user to use the last process; don't let root 134 * exceed the limit. The variable nprocs is the current number of 135 * processes, maxproc is the limit. 136 */ 137 uid = p1->p_cred->p_ruid; 138 if ((nprocs >= maxproc - 1 && uid != 0) || nprocs >= maxproc) { 139 tablefull("proc"); 140 return (EAGAIN); 141 } 142 143 /* 144 * Increment the count of procs running with this uid. Don't allow 145 * a nonprivileged user to exceed their current limit. 146 */ 147 count = chgproccnt(uid, 1); 148 if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) { 149 (void)chgproccnt(uid, -1); 150 return (EAGAIN); 151 } 152 153 /* 154 * Allocate virtual address space for the U-area now, while it 155 * is still easy to abort the fork operation if we're out of 156 * kernel virtual address space. The actual U-area pages will 157 * be allocated and wired in vm_fork(). 158 */ 159 #if defined(UVM) 160 uaddr = uvm_km_valloc(kernel_map, USPACE); 161 #else 162 uaddr = kmem_alloc_pageable(kernel_map, USPACE); 163 #endif 164 if (uaddr == 0) { 165 (void)chgproccnt(uid, -1); 166 return (ENOMEM); 167 } 168 169 /* 170 * We are now committed to the fork. From here on, we may 171 * block on resources, but resource allocation may NOT fail. 172 */ 173 174 /* Allocate new proc. */ 175 newproc = pool_get(&proc_pool, PR_WAITOK); 176 177 /* 178 * BEGIN PID ALLOCATION. (Lock PID allocation variables eventually). 179 */ 180 181 /* 182 * Find an unused process ID. We remember a range of unused IDs 183 * ready to use (from nextpid+1 through pidchecked-1). 184 */ 185 nextpid++; 186 retry: 187 /* 188 * If the process ID prototype has wrapped around, 189 * restart somewhat above 0, as the low-numbered procs 190 * tend to include daemons that don't exit. 191 */ 192 if (nextpid >= PID_MAX) { 193 nextpid = 100; 194 pidchecked = 0; 195 } 196 if (nextpid >= pidchecked) { 197 const struct proclist_desc *pd; 198 199 pidchecked = PID_MAX; 200 /* 201 * Scan the process lists to check whether this pid 202 * is in use. Remember the lowest pid that's greater 203 * than nextpid, so we can avoid checking for a while. 204 */ 205 pd = proclists; 206 again: 207 for (p2 = LIST_FIRST(pd->pd_list); p2 != 0; 208 p2 = LIST_NEXT(p2, p_list)) { 209 while (p2->p_pid == nextpid || 210 p2->p_pgrp->pg_id == nextpid || 211 p2->p_session->s_sid == nextpid) { 212 nextpid++; 213 if (nextpid >= pidchecked) 214 goto retry; 215 } 216 if (p2->p_pid > nextpid && pidchecked > p2->p_pid) 217 pidchecked = p2->p_pid; 218 219 if (p2->p_pgrp->pg_id > nextpid && 220 pidchecked > p2->p_pgrp->pg_id) 221 pidchecked = p2->p_pgrp->pg_id; 222 223 if (p2->p_session->s_sid > nextpid && 224 pidchecked > p2->p_session->s_sid) 225 pidchecked = p2->p_session->s_sid; 226 } 227 228 /* 229 * If there's another list, scan it. If we have checked 230 * them all, we've found one! 231 */ 232 pd++; 233 if (pd->pd_list != NULL) 234 goto again; 235 } 236 237 nprocs++; 238 p2 = newproc; 239 240 /* Record the pid we've allocated. */ 241 p2->p_pid = nextpid; 242 243 /* 244 * Put the proc on allproc before unlocking PID allocation 245 * so that waiters won't grab it as soon as we unlock. 246 */ 247 LIST_INSERT_HEAD(&allproc, p2, p_list); 248 249 /* 250 * END PID ALLOCATION. (Unlock PID allocation variables). 251 */ 252 253 p2->p_stat = SIDL; /* protect against others */ 254 p2->p_forw = p2->p_back = NULL; /* shouldn't be necessary */ 255 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash); 256 257 /* 258 * Make a proc table entry for the new process. 259 * Start by zeroing the section of proc that is zero-initialized, 260 * then copy the section that is copied directly from the parent. 261 */ 262 memset(&p2->p_startzero, 0, 263 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); 264 memcpy(&p2->p_startcopy, &p1->p_startcopy, 265 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); 266 267 /* 268 * Duplicate sub-structures as needed. 269 * Increase reference counts on shared objects. 270 * The p_stats and p_sigacts substructs are set in vm_fork. 271 */ 272 p2->p_flag = P_INMEM | (p1->p_flag & P_SUGID); 273 p2->p_emul = p1->p_emul; 274 if (p1->p_flag & P_PROFIL) 275 startprofclock(p2); 276 p2->p_cred = pool_get(&pcred_pool, PR_WAITOK); 277 memcpy(p2->p_cred, p1->p_cred, sizeof(*p2->p_cred)); 278 p2->p_cred->p_refcnt = 1; 279 crhold(p1->p_ucred); 280 281 /* 282 * slow us down if parent was cpu-bound 283 */ 284 if (slowchild) 285 p2->p_estcpu = p1->p_estcpu; 286 287 /* bump references to the text vnode (for procfs) */ 288 p2->p_textvp = p1->p_textvp; 289 if (p2->p_textvp) 290 VREF(p2->p_textvp); 291 292 p2->p_fd = fdcopy(p1); 293 /* 294 * If p_limit is still copy-on-write, bump refcnt, 295 * otherwise get a copy that won't be modified. 296 * (If PL_SHAREMOD is clear, the structure is shared 297 * copy-on-write.) 298 */ 299 if (p1->p_limit->p_lflags & PL_SHAREMOD) 300 p2->p_limit = limcopy(p1->p_limit); 301 else { 302 p2->p_limit = p1->p_limit; 303 p2->p_limit->p_refcnt++; 304 } 305 306 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) 307 p2->p_flag |= P_CONTROLT; 308 if (flags & FORK_PPWAIT) 309 p2->p_flag |= P_PPWAIT; 310 LIST_INSERT_AFTER(p1, p2, p_pglist); 311 p2->p_pptr = p1; 312 LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling); 313 LIST_INIT(&p2->p_children); 314 315 #ifdef KTRACE 316 /* 317 * Copy traceflag and tracefile if enabled. 318 * If not inherited, these were zeroed above. 319 */ 320 if (p1->p_traceflag&KTRFAC_INHERIT) { 321 p2->p_traceflag = p1->p_traceflag; 322 if ((p2->p_tracep = p1->p_tracep) != NULL) 323 ktradref(p2); 324 } 325 #endif 326 327 /* 328 * This begins the section where we must prevent the parent 329 * from being swapped. 330 */ 331 PHOLD(p1); 332 333 /* 334 * Finish creating the child process. It will return through a 335 * different path later. 336 */ 337 p2->p_addr = (struct user *)uaddr; 338 #if defined(UVM) 339 uvm_fork(p1, p2, (flags & FORK_SHAREVM) ? TRUE : FALSE); 340 #else 341 vm_fork(p1, p2, (flags & FORK_SHAREVM) ? TRUE : FALSE); 342 #endif 343 344 /* 345 * Make child runnable, set start time, and add to run queue. 346 */ 347 s = splstatclock(); 348 p2->p_stats->p_start = time; 349 p2->p_acflag = AFORK; 350 p2->p_stat = SRUN; 351 setrunqueue(p2); 352 splx(s); 353 354 /* 355 * Now can be swapped. 356 */ 357 PRELE(p1); 358 359 /* 360 * Update stats now that we know the fork was successful. 361 */ 362 #if defined(UVM) 363 uvmexp.forks++; 364 if (flags & FORK_PPWAIT) 365 uvmexp.forks_ppwait++; 366 if (flags & FORK_SHAREVM) 367 uvmexp.forks_sharevm++; 368 #else 369 cnt.v_forks++; 370 if (flags & FORK_PPWAIT) 371 cnt.v_forks_ppwait++; 372 if (flags & FORK_SHAREVM) 373 cnt.v_forks_sharevm++; 374 #endif 375 376 /* 377 * Pass a pointer to the new process to the caller. 378 */ 379 if (rnewprocp != NULL) 380 *rnewprocp = p2; 381 382 /* 383 * Preserve synchronization semantics of vfork. If waiting for 384 * child to exec or exit, set P_PPWAIT on child, and sleep on our 385 * proc (in case of exit). 386 */ 387 if (flags & FORK_PPWAIT) 388 while (p2->p_flag & P_PPWAIT) 389 tsleep(p1, PWAIT, "ppwait", 0); 390 391 /* 392 * Return child pid to parent process, 393 * marking us as parent via retval[1]. 394 */ 395 if (retval != NULL) { 396 retval[0] = p2->p_pid; 397 retval[1] = 0; 398 } 399 return (0); 400 } 401
Indexes created Mon Sep 22 13:09:51 GMT 2025