1 /* 2 * Copyright (c) 1982, 1986, 1989, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)kern_fork.c 8.6 (Berkeley) 4/8/94 39 * $FreeBSD$ 40 */ 41 42 #include "opt_ktrace.h" 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/sysproto.h> 47 #include <sys/filedesc.h> 48 #include <sys/kernel.h> 49 #include <sys/sysctl.h> 50 #include <sys/lock.h> 51 #include <sys/malloc.h> 52 #include <sys/mutex.h> 53 #include <sys/proc.h> 54 #include <sys/resourcevar.h> 55 #include <sys/syscall.h> 56 #include <sys/vnode.h> 57 #include <sys/acct.h> 58 #include <sys/ktr.h> 59 #include <sys/ktrace.h> 60 #include <sys/kthread.h> 61 #include <sys/unistd.h> 62 #include <sys/jail.h> 63 #include <sys/sx.h> 64 65 #include <vm/vm.h> 66 #include <vm/pmap.h> 67 #include <vm/vm_map.h> 68 #include <vm/vm_extern.h> 69 #include <vm/vm_zone.h> 70 71 #include <sys/vmmeter.h> 72 #include <sys/user.h> 73 74 static MALLOC_DEFINE(M_ATFORK, "atfork", "atfork callback"); 75 76 static int fast_vfork = 1; 77 SYSCTL_INT(_kern, OID_AUTO, fast_vfork, CTLFLAG_RW, &fast_vfork, 0, 78 "flag to indicate whether we have a fast vfork()"); 79 80 /* 81 * These are the stuctures used to create a callout list for things to do 82 * when forking a process 83 */ 84 struct forklist { 85 forklist_fn function; 86 TAILQ_ENTRY(forklist) next; 87 }; 88 89 static struct sx fork_list_lock; 90 91 TAILQ_HEAD(forklist_head, forklist); 92 static struct forklist_head fork_list = TAILQ_HEAD_INITIALIZER(fork_list); 93 94 #ifndef _SYS_SYSPROTO_H_ 95 struct fork_args { 96 int dummy; 97 }; 98 #endif 99 100 static void 101 init_fork_list(void *data __unused) 102 { 103 104 sx_init(&fork_list_lock, "fork list"); 105 } 106 SYSINIT(fork_list, SI_SUB_INTRINSIC, SI_ORDER_ANY, init_fork_list, NULL); 107 108 /* 109 * MPSAFE 110 */ 111 /* ARGSUSED */ 112 int 113 fork(p, uap) 114 struct proc *p; 115 struct fork_args *uap; 116 { 117 int error; 118 struct proc *p2; 119 120 mtx_lock(&Giant); 121 error = fork1(p, RFFDG | RFPROC, &p2); 122 if (error == 0) { 123 p->p_retval[0] = p2->p_pid; 124 p->p_retval[1] = 0; 125 } 126 mtx_unlock(&Giant); 127 return error; 128 } 129 130 /* 131 * MPSAFE 132 */ 133 /* ARGSUSED */ 134 int 135 vfork(p, uap) 136 struct proc *p; 137 struct vfork_args *uap; 138 { 139 int error; 140 struct proc *p2; 141 142 mtx_lock(&Giant); 143 error = fork1(p, RFFDG | RFPROC | RFPPWAIT | RFMEM, &p2); 144 if (error == 0) { 145 p->p_retval[0] = p2->p_pid; 146 p->p_retval[1] = 0; 147 } 148 mtx_unlock(&Giant); 149 return error; 150 } 151 152 /* 153 * MPSAFE 154 */ 155 int 156 rfork(p, uap) 157 struct proc *p; 158 struct rfork_args *uap; 159 { 160 int error; 161 struct proc *p2; 162 163 /* mask kernel only flags out of the user flags */ 164 mtx_lock(&Giant); 165 error = fork1(p, uap->flags & ~RFKERNELONLY, &p2); 166 if (error == 0) { 167 p->p_retval[0] = p2 ? p2->p_pid : 0; 168 p->p_retval[1] = 0; 169 } 170 mtx_unlock(&Giant); 171 return error; 172 } 173 174 175 int nprocs = 1; /* process 0 */ 176 int lastpid = 0; 177 SYSCTL_INT(_kern, OID_AUTO, lastpid, CTLFLAG_RD, &lastpid, 0, 178 "Last used PID"); 179 180 /* 181 * Random component to lastpid generation. We mix in a random factor to make 182 * it a little harder to predict. We sanity check the modulus value to avoid 183 * doing it in critical paths. Don't let it be too small or we pointlessly 184 * waste randomness entropy, and don't let it be impossibly large. Using a 185 * modulus that is too big causes a LOT more process table scans and slows 186 * down fork processing as the pidchecked caching is defeated. 187 */ 188 static int randompid = 0; 189 190 static int 191 sysctl_kern_randompid(SYSCTL_HANDLER_ARGS) 192 { 193 int error, pid; 194 195 pid = randompid; 196 error = sysctl_handle_int(oidp, &pid, 0, req); 197 if (error || !req->newptr) 198 return (error); 199 if (pid < 0 || pid > PID_MAX - 100) /* out of range */ 200 pid = PID_MAX - 100; 201 else if (pid < 2) /* NOP */ 202 pid = 0; 203 else if (pid < 100) /* Make it reasonable */ 204 pid = 100; 205 randompid = pid; 206 return (error); 207 } 208 209 SYSCTL_PROC(_kern, OID_AUTO, randompid, CTLTYPE_INT|CTLFLAG_RW, 210 0, 0, sysctl_kern_randompid, "I", "Random PID modulus"); 211 212 int 213 fork1(p1, flags, procp) 214 struct proc *p1; /* parent proc */ 215 int flags; 216 struct proc **procp; /* child proc */ 217 { 218 struct proc *p2, *pptr; 219 uid_t uid; 220 struct proc *newproc; 221 int trypid; 222 int ok; 223 static int pidchecked = 0; 224 struct forklist *ep; 225 struct filedesc *fd; 226 227 GIANT_REQUIRED; 228 229 /* Can't copy and clear */ 230 if ((flags & (RFFDG|RFCFDG)) == (RFFDG|RFCFDG)) 231 return (EINVAL); 232 233 /* 234 * Here we don't create a new process, but we divorce 235 * certain parts of a process from itself. 236 */ 237 if ((flags & RFPROC) == 0) { 238 vm_fork(p1, 0, flags); 239 240 /* 241 * Close all file descriptors. 242 */ 243 if (flags & RFCFDG) { 244 struct filedesc *fdtmp; 245 fdtmp = fdinit(p1); 246 PROC_LOCK(p1); 247 fdfree(p1); 248 p1->p_fd = fdtmp; 249 PROC_UNLOCK(p1); 250 } 251 252 /* 253 * Unshare file descriptors (from parent.) 254 */ 255 if (flags & RFFDG) { 256 if (p1->p_fd->fd_refcnt > 1) { 257 struct filedesc *newfd; 258 newfd = fdcopy(p1); 259 PROC_LOCK(p1); 260 fdfree(p1); 261 p1->p_fd = newfd; 262 PROC_UNLOCK(p1); 263 } 264 } 265 *procp = NULL; 266 return (0); 267 } 268 269 /* 270 * Although process entries are dynamically created, we still keep 271 * a global limit on the maximum number we will create. Don't allow 272 * a nonprivileged user to use the last process; don't let root 273 * exceed the limit. The variable nprocs is the current number of 274 * processes, maxproc is the limit. 275 */ 276 uid = p1->p_ucred->cr_ruid; 277 if ((nprocs >= maxproc - 1 && uid != 0) || nprocs >= maxproc) { 278 tablefull("proc"); 279 return (EAGAIN); 280 } 281 /* 282 * Increment the nprocs resource before blocking can occur. There 283 * are hard-limits as to the number of processes that can run. 284 */ 285 nprocs++; 286 287 /* 288 * Increment the count of procs running with this uid. Don't allow 289 * a nonprivileged user to exceed their current limit. 290 */ 291 ok = chgproccnt(p1->p_ucred->cr_ruidinfo, 1, 292 (uid != 0) ? p1->p_rlimit[RLIMIT_NPROC].rlim_cur : 0); 293 if (!ok) { 294 /* 295 * Back out the process count 296 */ 297 nprocs--; 298 return (EAGAIN); 299 } 300 301 /* Allocate new proc. */ 302 newproc = zalloc(proc_zone); 303 304 /* 305 * Setup linkage for kernel based threading 306 */ 307 if((flags & RFTHREAD) != 0) { 308 newproc->p_peers = p1->p_peers; 309 p1->p_peers = newproc; 310 newproc->p_leader = p1->p_leader; 311 } else { 312 newproc->p_peers = NULL; 313 newproc->p_leader = newproc; 314 } 315 316 newproc->p_vmspace = NULL; 317 318 /* 319 * Find an unused process ID. We remember a range of unused IDs 320 * ready to use (from lastpid+1 through pidchecked-1). 321 * 322 * If RFHIGHPID is set (used during system boot), do not allocate 323 * low-numbered pids. 324 */ 325 sx_xlock(&allproc_lock); 326 trypid = lastpid + 1; 327 if (flags & RFHIGHPID) { 328 if (trypid < 10) { 329 trypid = 10; 330 } 331 } else { 332 if (randompid) 333 trypid += arc4random() % randompid; 334 } 335 retry: 336 /* 337 * If the process ID prototype has wrapped around, 338 * restart somewhat above 0, as the low-numbered procs 339 * tend to include daemons that don't exit. 340 */ 341 if (trypid >= PID_MAX) { 342 trypid = trypid % PID_MAX; 343 if (trypid < 100) 344 trypid += 100; 345 pidchecked = 0; 346 } 347 if (trypid >= pidchecked) { 348 int doingzomb = 0; 349 350 pidchecked = PID_MAX; 351 /* 352 * Scan the active and zombie procs to check whether this pid 353 * is in use. Remember the lowest pid that's greater 354 * than trypid, so we can avoid checking for a while. 355 */ 356 p2 = LIST_FIRST(&allproc); 357 again: 358 for (; p2 != NULL; p2 = LIST_NEXT(p2, p_list)) { 359 while (p2->p_pid == trypid || 360 p2->p_pgrp->pg_id == trypid || 361 p2->p_session->s_sid == trypid) { 362 trypid++; 363 if (trypid >= pidchecked) 364 goto retry; 365 } 366 if (p2->p_pid > trypid && pidchecked > p2->p_pid) 367 pidchecked = p2->p_pid; 368 if (p2->p_pgrp->pg_id > trypid && 369 pidchecked > p2->p_pgrp->pg_id) 370 pidchecked = p2->p_pgrp->pg_id; 371 if (p2->p_session->s_sid > trypid && 372 pidchecked > p2->p_session->s_sid) 373 pidchecked = p2->p_session->s_sid; 374 } 375 if (!doingzomb) { 376 doingzomb = 1; 377 p2 = LIST_FIRST(&zombproc); 378 goto again; 379 } 380 } 381 382 /* 383 * RFHIGHPID does not mess with the lastpid counter during boot. 384 */ 385 if (flags & RFHIGHPID) 386 pidchecked = 0; 387 else 388 lastpid = trypid; 389 390 p2 = newproc; 391 p2->p_stat = SIDL; /* protect against others */ 392 p2->p_pid = trypid; 393 LIST_INSERT_HEAD(&allproc, p2, p_list); 394 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash); 395 sx_xunlock(&allproc_lock); 396 397 /* 398 * Make a proc table entry for the new process. 399 * Start by zeroing the section of proc that is zero-initialized, 400 * then copy the section that is copied directly from the parent. 401 */ 402 bzero(&p2->p_startzero, 403 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); 404 PROC_LOCK(p1); 405 bcopy(&p1->p_startcopy, &p2->p_startcopy, 406 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); 407 PROC_UNLOCK(p1); 408 409 mtx_init(&p2->p_mtx, "process lock", MTX_DEF); 410 PROC_LOCK(p2); 411 412 /* 413 * Duplicate sub-structures as needed. 414 * Increase reference counts on shared objects. 415 * The p_stats and p_sigacts substructs are set in vm_fork. 416 */ 417 p2->p_flag = 0; 418 mtx_lock_spin(&sched_lock); 419 p2->p_sflag = PS_INMEM; 420 if (p1->p_sflag & PS_PROFIL) 421 startprofclock(p2); 422 mtx_unlock_spin(&sched_lock); 423 /* 424 * We start off holding one spinlock after fork: sched_lock. 425 */ 426 PROC_LOCK(p1); 427 crhold(p1->p_ucred); 428 p2->p_ucred = p1->p_ucred; 429 430 if (p2->p_args) 431 p2->p_args->ar_ref++; 432 433 if (flags & RFSIGSHARE) { 434 p2->p_procsig = p1->p_procsig; 435 p2->p_procsig->ps_refcnt++; 436 if (p1->p_sigacts == &p1->p_addr->u_sigacts) { 437 struct sigacts *newsigacts; 438 439 PROC_UNLOCK(p1); 440 PROC_UNLOCK(p2); 441 /* Create the shared sigacts structure */ 442 MALLOC(newsigacts, struct sigacts *, 443 sizeof(struct sigacts), M_SUBPROC, M_WAITOK); 444 PROC_LOCK(p2); 445 PROC_LOCK(p1); 446 /* 447 * Set p_sigacts to the new shared structure. 448 * Note that this is updating p1->p_sigacts at the 449 * same time, since p_sigacts is just a pointer to 450 * the shared p_procsig->ps_sigacts. 451 */ 452 p2->p_sigacts = newsigacts; 453 *p2->p_sigacts = p1->p_addr->u_sigacts; 454 } 455 } else { 456 PROC_UNLOCK(p1); 457 PROC_UNLOCK(p2); 458 MALLOC(p2->p_procsig, struct procsig *, sizeof(struct procsig), 459 M_SUBPROC, M_WAITOK); 460 PROC_LOCK(p2); 461 PROC_LOCK(p1); 462 bcopy(p1->p_procsig, p2->p_procsig, sizeof(*p2->p_procsig)); 463 p2->p_procsig->ps_refcnt = 1; 464 p2->p_sigacts = NULL; /* finished in vm_fork() */ 465 } 466 if (flags & RFLINUXTHPN) 467 p2->p_sigparent = SIGUSR1; 468 else 469 p2->p_sigparent = SIGCHLD; 470 471 /* bump references to the text vnode (for procfs) */ 472 p2->p_textvp = p1->p_textvp; 473 PROC_UNLOCK(p1); 474 PROC_UNLOCK(p2); 475 if (p2->p_textvp) 476 VREF(p2->p_textvp); 477 478 if (flags & RFCFDG) 479 fd = fdinit(p1); 480 else if (flags & RFFDG) 481 fd = fdcopy(p1); 482 else 483 fd = fdshare(p1); 484 PROC_LOCK(p2); 485 p2->p_fd = fd; 486 487 /* 488 * If p_limit is still copy-on-write, bump refcnt, 489 * otherwise get a copy that won't be modified. 490 * (If PL_SHAREMOD is clear, the structure is shared 491 * copy-on-write.) 492 */ 493 PROC_LOCK(p1); 494 if (p1->p_limit->p_lflags & PL_SHAREMOD) 495 p2->p_limit = limcopy(p1->p_limit); 496 else { 497 p2->p_limit = p1->p_limit; 498 p2->p_limit->p_refcnt++; 499 } 500 501 /* 502 * Preserve some more flags in subprocess. PS_PROFIL has already 503 * been preserved. 504 */ 505 p2->p_flag |= p1->p_flag & (P_SUGID | P_ALTSTACK); 506 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) 507 p2->p_flag |= P_CONTROLT; 508 if (flags & RFPPWAIT) 509 p2->p_flag |= P_PPWAIT; 510 511 LIST_INSERT_AFTER(p1, p2, p_pglist); 512 PROC_UNLOCK(p1); 513 PROC_UNLOCK(p2); 514 515 /* 516 * Attach the new process to its parent. 517 * 518 * If RFNOWAIT is set, the newly created process becomes a child 519 * of init. This effectively disassociates the child from the 520 * parent. 521 */ 522 if (flags & RFNOWAIT) 523 pptr = initproc; 524 else 525 pptr = p1; 526 sx_xlock(&proctree_lock); 527 PROC_LOCK(p2); 528 p2->p_pptr = pptr; 529 PROC_UNLOCK(p2); 530 LIST_INSERT_HEAD(&pptr->p_children, p2, p_sibling); 531 sx_xunlock(&proctree_lock); 532 PROC_LOCK(p2); 533 LIST_INIT(&p2->p_children); 534 LIST_INIT(&p2->p_contested); 535 536 callout_init(&p2->p_itcallout, 0); 537 callout_init(&p2->p_slpcallout, 1); 538 539 PROC_LOCK(p1); 540 #ifdef KTRACE 541 /* 542 * Copy traceflag and tracefile if enabled. 543 * If not inherited, these were zeroed above. 544 */ 545 if (p1->p_traceflag & KTRFAC_INHERIT) { 546 p2->p_traceflag = p1->p_traceflag; 547 if ((p2->p_tracep = p1->p_tracep) != NULL) { 548 PROC_UNLOCK(p1); 549 PROC_UNLOCK(p2); 550 VREF(p2->p_tracep); 551 PROC_LOCK(p2); 552 PROC_LOCK(p1); 553 } 554 } 555 #endif 556 557 /* 558 * set priority of child to be that of parent 559 */ 560 mtx_lock_spin(&sched_lock); 561 p2->p_estcpu = p1->p_estcpu; 562 mtx_unlock_spin(&sched_lock); 563 564 /* 565 * This begins the section where we must prevent the parent 566 * from being swapped. 567 */ 568 _PHOLD(p1); 569 PROC_UNLOCK(p1); 570 PROC_UNLOCK(p2); 571 572 /* 573 * Finish creating the child process. It will return via a different 574 * execution path later. (ie: directly into user mode) 575 */ 576 vm_fork(p1, p2, flags); 577 578 if (flags == (RFFDG | RFPROC)) { 579 cnt.v_forks++; 580 cnt.v_forkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize; 581 } else if (flags == (RFFDG | RFPROC | RFPPWAIT | RFMEM)) { 582 cnt.v_vforks++; 583 cnt.v_vforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize; 584 } else if (p1 == &proc0) { 585 cnt.v_kthreads++; 586 cnt.v_kthreadpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize; 587 } else { 588 cnt.v_rforks++; 589 cnt.v_rforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize; 590 } 591 592 /* 593 * Both processes are set up, now check if any loadable modules want 594 * to adjust anything. 595 * What if they have an error? XXX 596 */ 597 sx_slock(&fork_list_lock); 598 TAILQ_FOREACH(ep, &fork_list, next) { 599 (*ep->function)(p1, p2, flags); 600 } 601 sx_sunlock(&fork_list_lock); 602 603 /* 604 * If RFSTOPPED not requested, make child runnable and add to 605 * run queue. 606 */ 607 microtime(&(p2->p_stats->p_start)); 608 p2->p_acflag = AFORK; 609 if ((flags & RFSTOPPED) == 0) { 610 mtx_lock_spin(&sched_lock); 611 p2->p_stat = SRUN; 612 setrunqueue(p2); 613 mtx_unlock_spin(&sched_lock); 614 } 615 616 /* 617 * Now can be swapped. 618 */ 619 PROC_LOCK(p1); 620 _PRELE(p1); 621 622 /* 623 * tell any interested parties about the new process 624 */ 625 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid); 626 PROC_UNLOCK(p1); 627 628 /* 629 * Preserve synchronization semantics of vfork. If waiting for 630 * child to exec or exit, set P_PPWAIT on child, and sleep on our 631 * proc (in case of exit). 632 */ 633 PROC_LOCK(p2); 634 while (p2->p_flag & P_PPWAIT) 635 msleep(p1, &p2->p_mtx, PWAIT, "ppwait", 0); 636 PROC_UNLOCK(p2); 637 638 /* 639 * Return child proc pointer to parent. 640 */ 641 *procp = p2; 642 return (0); 643 } 644 645 /* 646 * The next two functionms are general routines to handle adding/deleting 647 * items on the fork callout list. 648 * 649 * at_fork(): 650 * Take the arguments given and put them onto the fork callout list, 651 * However first make sure that it's not already there. 652 * Returns 0 on success or a standard error number. 653 */ 654 655 int 656 at_fork(function) 657 forklist_fn function; 658 { 659 struct forklist *ep; 660 661 #ifdef INVARIANTS 662 /* let the programmer know if he's been stupid */ 663 if (rm_at_fork(function)) 664 printf("WARNING: fork callout entry (%p) already present\n", 665 function); 666 #endif 667 ep = malloc(sizeof(*ep), M_ATFORK, M_NOWAIT); 668 if (ep == NULL) 669 return (ENOMEM); 670 ep->function = function; 671 sx_xlock(&fork_list_lock); 672 TAILQ_INSERT_TAIL(&fork_list, ep, next); 673 sx_xunlock(&fork_list_lock); 674 return (0); 675 } 676 677 /* 678 * Scan the exit callout list for the given item and remove it.. 679 * Returns the number of items removed (0 or 1) 680 */ 681 682 int 683 rm_at_fork(function) 684 forklist_fn function; 685 { 686 struct forklist *ep; 687 688 sx_xlock(&fork_list_lock); 689 TAILQ_FOREACH(ep, &fork_list, next) { 690 if (ep->function == function) { 691 TAILQ_REMOVE(&fork_list, ep, next); 692 sx_xunlock(&fork_list_lock); 693 free(ep, M_ATFORK); 694 return(1); 695 } 696 } 697 sx_xunlock(&fork_list_lock); 698 return (0); 699 } 700 701 /* 702 * Handle the return of a child process from fork1(). This function 703 * is called from the MD fork_trampoline() entry point. 704 */ 705 void 706 fork_exit(callout, arg, frame) 707 void (*callout)(void *, struct trapframe *); 708 void *arg; 709 struct trapframe *frame; 710 { 711 struct proc *p; 712 713 p = curproc; 714 715 /* 716 * Setup the sched_lock state so that we can release it. 717 */ 718 sched_lock.mtx_lock = (uintptr_t)p; 719 sched_lock.mtx_recurse = 0; 720 /* 721 * XXX: We really shouldn't have to do this. 722 */ 723 mtx_intr_enable(&sched_lock); 724 mtx_unlock_spin(&sched_lock); 725 726 #ifdef SMP 727 if (PCPU_GET(switchtime.tv_sec) == 0) 728 microuptime(PCPU_PTR(switchtime)); 729 PCPU_SET(switchticks, ticks); 730 #endif 731 732 /* 733 * cpu_set_fork_handler intercepts this function call to 734 * have this call a non-return function to stay in kernel mode. 735 * initproc has its own fork handler, but it does return. 736 */ 737 KASSERT(callout != NULL, ("NULL callout in fork_exit")); 738 callout(arg, frame); 739 740 /* 741 * Check if a kernel thread misbehaved and returned from its main 742 * function. 743 */ 744 PROC_LOCK(p); 745 if (p->p_flag & P_KTHREAD) { 746 PROC_UNLOCK(p); 747 mtx_lock(&Giant); 748 printf("Kernel thread \"%s\" (pid %d) exited prematurely.\n", 749 p->p_comm, p->p_pid); 750 kthread_exit(0); 751 } 752 PROC_UNLOCK(p); 753 mtx_assert(&Giant, MA_NOTOWNED); 754 } 755 756 /* 757 * Simplified back end of syscall(), used when returning from fork() 758 * directly into user mode. Giant is not held on entry, and must not 759 * be held on return. This function is passed in to fork_exit() as the 760 * first parameter and is called when returning to a new userland process. 761 */ 762 void 763 fork_return(p, frame) 764 struct proc *p; 765 struct trapframe *frame; 766 { 767 768 userret(p, frame, 0); 769 #ifdef KTRACE 770 if (KTRPOINT(p, KTR_SYSRET)) { 771 ktrsysret(p->p_tracep, SYS_fork, 0, 0); 772 } 773 #endif 774 mtx_assert(&Giant, MA_NOTOWNED); 775 } 776