1 /* $NetBSD: kern_fork.c,v 1.234 2026/01/04 01:33:56 riastradh Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2001, 2004, 2006, 2007, 2008, 2019 5 * The NetBSD Foundation, Inc. 6 * All rights reserved. 7 * 8 * This code is derived from software contributed to The NetBSD Foundation 9 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility, 10 * NASA Ames Research Center, by Charles M. Hannum, and by Andrew Doran. 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 * 21 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 23 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 24 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 25 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 28 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 29 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 30 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 31 * POSSIBILITY OF SUCH DAMAGE. 32 */ 33 34 /* 35 * Copyright (c) 1982, 1986, 1989, 1991, 1993 36 * The Regents of the University of California. All rights reserved. 37 * (c) UNIX System Laboratories, Inc. 38 * All or some portions of this file are derived from material licensed 39 * to the University of California by American Telephone and Telegraph 40 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 41 * the permission of UNIX System Laboratories, Inc. 42 * 43 * Redistribution and use in source and binary forms, with or without 44 * modification, are permitted provided that the following conditions 45 * are met: 46 * 1. Redistributions of source code must retain the above copyright 47 * notice, this list of conditions and the following disclaimer. 48 * 2. Redistributions in binary form must reproduce the above copyright 49 * notice, this list of conditions and the following disclaimer in the 50 * documentation and/or other materials provided with the distribution. 51 * 3. Neither the name of the University nor the names of its contributors 52 * may be used to endorse or promote products derived from this software 53 * without specific prior written permission. 54 * 55 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 56 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 57 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 58 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 59 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 60 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 61 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 62 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 63 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 64 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 65 * SUCH DAMAGE. 66 * 67 * @(#)kern_fork.c 8.8 (Berkeley) 2/14/95 68 */ 69 70 #include <sys/cdefs.h> 71 __KERNEL_RCSID(0, "$NetBSD: kern_fork.c,v 1.234 2026/01/04 01:33:56 riastradh Exp $"); 72 73 #include "opt_ktrace.h" 74 #include "opt_dtrace.h" 75 76 #include <sys/param.h> 77 #include <sys/types.h> 78 79 #include <sys/acct.h> 80 #include <sys/atomic.h> 81 #include <sys/file.h> 82 #include <sys/filedesc.h> 83 #include <sys/kauth.h> 84 #include <sys/kernel.h> 85 #include <sys/ktrace.h> 86 #include <sys/mount.h> 87 #include <sys/pool.h> 88 #include <sys/proc.h> 89 #include <sys/ptrace.h> 90 #include <sys/ras.h> 91 #include <sys/resourcevar.h> 92 #include <sys/sched.h> 93 #include <sys/sdt.h> 94 #include <sys/signalvar.h> 95 #include <sys/syscall.h> 96 #include <sys/syscallargs.h> 97 #include <sys/systm.h> 98 #include <sys/uidinfo.h> 99 #include <sys/vnode.h> 100 101 /* 102 * DTrace SDT provider definitions 103 */ 104 SDT_PROVIDER_DECLARE(proc); 105 SDT_PROBE_DEFINE3(proc, kernel, , create, 106 "struct proc *", /* new process */ 107 "struct proc *", /* parent process */ 108 "int" /* flags */); 109 110 u_int nprocs __cacheline_aligned = 1; /* process 0 */ 111 112 /* 113 * Number of ticks to sleep if fork() would fail due to process hitting 114 * limits. Exported in milliseconds to userland via sysctl. 115 */ 116 int forkfsleep = 0; 117 118 int 119 sys_fork(struct lwp *l, const void *v, register_t *retval) 120 { 121 122 return fork1(l, 0, SIGCHLD, NULL, 0, NULL, NULL, retval); 123 } 124 125 /* 126 * vfork(2) system call compatible with 4.4BSD (i.e. BSD with Mach VM). 127 * Address space is not shared, but parent is blocked until child exit. 128 */ 129 int 130 sys_vfork(struct lwp *l, const void *v, register_t *retval) 131 { 132 133 return fork1(l, FORK_PPWAIT, SIGCHLD, NULL, 0, NULL, NULL, 134 retval); 135 } 136 137 /* 138 * New vfork(2) system call for NetBSD, which implements original 3BSD vfork(2) 139 * semantics. Address space is shared, and parent is blocked until child exit. 140 */ 141 int 142 sys___vfork14(struct lwp *l, const void *v, register_t *retval) 143 { 144 145 return fork1(l, FORK_PPWAIT|FORK_SHAREVM, SIGCHLD, NULL, 0, 146 NULL, NULL, retval); 147 } 148 149 /* 150 * Linux-compatible __clone(2) system call. 151 */ 152 int 153 sys___clone(struct lwp *l, const struct sys___clone_args *uap, 154 register_t *retval) 155 { 156 /* { 157 syscallarg(int) flags; 158 syscallarg(void *) stack; 159 } */ 160 int flags, sig; 161 162 /* 163 * We don't support the CLONE_PTRACE flag. 164 */ 165 if (SCARG(uap, flags) & CLONE_PTRACE) 166 return SET_ERROR(EINVAL); 167 168 /* 169 * Linux enforces CLONE_VM with CLONE_SIGHAND, do same. 170 */ 171 if (SCARG(uap, flags) & CLONE_SIGHAND 172 && (SCARG(uap, flags) & CLONE_VM) == 0) 173 return SET_ERROR(EINVAL); 174 175 flags = 0; 176 177 if (SCARG(uap, flags) & CLONE_VM) 178 flags |= FORK_SHAREVM; 179 if (SCARG(uap, flags) & CLONE_FS) 180 flags |= FORK_SHARECWD; 181 if (SCARG(uap, flags) & CLONE_FILES) 182 flags |= FORK_SHAREFILES; 183 if (SCARG(uap, flags) & CLONE_SIGHAND) 184 flags |= FORK_SHARESIGS; 185 if (SCARG(uap, flags) & CLONE_VFORK) 186 flags |= FORK_PPWAIT; 187 188 sig = SCARG(uap, flags) & CLONE_CSIGNAL; 189 if (sig < 0 || sig >= _NSIG) 190 return SET_ERROR(EINVAL); 191 192 /* 193 * Linux doesn't have close-on-fork yet, so we don't 194 * know what they will do combining CLONE_FILES with 195 * close-on-fork (which are not really compatible). 196 * This might need to be changed in the future (another 197 * option would be to just disable FORK_SHAREFILES) 198 */ 199 if ((flags & FORK_SHAREFILES) != 0) { 200 if (l->l_fd != NULL && l->l_fd->fd_foclose) 201 return SET_ERROR(EINVAL); 202 } 203 204 /* 205 * Note that the Linux API does not provide a portable way of 206 * specifying the stack area; the caller must know if the stack 207 * grows up or down. So, we pass a stack size of 0, so that the 208 * code that makes this adjustment is a noop. 209 */ 210 return fork1(l, flags, sig, SCARG(uap, stack), 0, 211 NULL, NULL, retval); 212 } 213 214 /* 215 * Print the 'table full' message once per 10 seconds. 216 */ 217 static struct timeval fork_tfmrate = { 10, 0 }; 218 219 /* 220 * Check if a process is traced and shall inform about FORK events. 221 */ 222 static inline bool 223 tracefork(struct proc *p, int flags) 224 { 225 226 return (p->p_slflag & (PSL_TRACEFORK|PSL_TRACED)) == 227 (PSL_TRACEFORK|PSL_TRACED) && (flags & FORK_PPWAIT) == 0; 228 } 229 230 /* 231 * Check if a process is traced and shall inform about VFORK events. 232 */ 233 static inline bool 234 tracevfork(struct proc *p, int flags) 235 { 236 237 return (p->p_slflag & (PSL_TRACEVFORK|PSL_TRACED)) == 238 (PSL_TRACEVFORK|PSL_TRACED) && (flags & FORK_PPWAIT) != 0; 239 } 240 241 /* 242 * Check if a process is traced and shall inform about VFORK_DONE events. 243 */ 244 static inline bool 245 tracevforkdone(struct proc *p, int flags) 246 { 247 248 return (p->p_slflag & (PSL_TRACEVFORK_DONE|PSL_TRACED)) == 249 (PSL_TRACEVFORK_DONE|PSL_TRACED) && (flags & FORK_PPWAIT); 250 } 251 252 /* 253 * General fork call. Note that another LWP in the process may call exec() 254 * or exit() while we are forking. It's safe to continue here, because 255 * neither operation will complete until all LWPs have exited the process. 256 */ 257 int 258 fork1(struct lwp *l1, int flags, int exitsig, void *stack, size_t stacksize, 259 void (*func)(void *), void *arg, register_t *retval) 260 { 261 struct proc *p1, *p2, *parent; 262 struct plimit *p1_lim; 263 uid_t uid; 264 struct lwp *l2; 265 int count; 266 vaddr_t uaddr; 267 int tnprocs; 268 int error = 0; 269 270 p1 = l1->l_proc; 271 uid = kauth_cred_getuid(l1->l_cred); 272 tnprocs = atomic_inc_uint_nv(&nprocs); 273 274 /* 275 * Although process entries are dynamically created, we still keep 276 * a global limit on the maximum number we will create. 277 */ 278 if (__predict_false(tnprocs >= maxproc)) 279 error = -1; 280 else 281 error = kauth_authorize_process(l1->l_cred, 282 KAUTH_PROCESS_FORK, p1, KAUTH_ARG(tnprocs), NULL, NULL); 283 284 if (error) { 285 static struct timeval lasttfm; 286 atomic_dec_uint(&nprocs); 287 if (ratecheck(&lasttfm, &fork_tfmrate)) 288 tablefull("proc", "increase kern.maxproc or NPROC"); 289 if (forkfsleep) 290 kpause("forkmx", false, forkfsleep, NULL); 291 return SET_ERROR(EAGAIN); 292 } 293 294 /* 295 * Enforce limits. 296 */ 297 count = chgproccnt(uid, 1); 298 if (__predict_false(count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur)) { 299 if (kauth_authorize_process(l1->l_cred, KAUTH_PROCESS_RLIMIT, 300 p1, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_BYPASS), 301 &p1->p_rlimit[RLIMIT_NPROC], KAUTH_ARG(RLIMIT_NPROC)) != 0) { 302 (void)chgproccnt(uid, -1); 303 atomic_dec_uint(&nprocs); 304 if (forkfsleep) 305 kpause("forkulim", false, forkfsleep, NULL); 306 return SET_ERROR(EAGAIN); 307 } 308 } 309 310 /* 311 * Allocate virtual address space for the U-area now, while it 312 * is still easy to abort the fork operation if we're out of 313 * kernel virtual address space. 314 */ 315 uaddr = uvm_uarea_alloc(); 316 if (__predict_false(uaddr == 0)) { 317 (void)chgproccnt(uid, -1); 318 atomic_dec_uint(&nprocs); 319 return SET_ERROR(ENOMEM); 320 } 321 322 /* Allocate new proc. */ 323 p2 = proc_alloc(); 324 if (p2 == NULL) { 325 /* We were unable to allocate a process ID. */ 326 uvm_uarea_free(uaddr); 327 mutex_enter(p1->p_lock); 328 uid = kauth_cred_getuid(p1->p_cred); 329 (void)chgproccnt(uid, -1); 330 mutex_exit(p1->p_lock); 331 atomic_dec_uint(&nprocs); 332 return SET_ERROR(EAGAIN); 333 } 334 335 /* 336 * We are now committed to the fork. From here on, we may 337 * block on resources, but resource allocation may NOT fail. 338 */ 339 340 /* 341 * Make a proc table entry for the new process. 342 * Start by zeroing the section of proc that is zero-initialized, 343 * then copy the section that is copied directly from the parent. 344 */ 345 memset(&p2->p_startzero, 0, 346 (unsigned) ((char *)&p2->p_endzero - (char *)&p2->p_startzero)); 347 memcpy(&p2->p_startcopy, &p1->p_startcopy, 348 (unsigned) ((char *)&p2->p_endcopy - (char *)&p2->p_startcopy)); 349 350 TAILQ_INIT(&p2->p_sigpend.sp_info); 351 352 LIST_INIT(&p2->p_lwps); 353 LIST_INIT(&p2->p_sigwaiters); 354 355 /* 356 * Duplicate sub-structures as needed. 357 * Increase reference counts on shared objects. 358 * Inherit flags we want to keep. The flags related to SIGCHLD 359 * handling are important in order to keep a consistent behaviour 360 * for the child after the fork. If we are a 32-bit process, the 361 * child will be too. 362 */ 363 p2->p_flag = 364 p1->p_flag & (PK_SUGID | PK_NOCLDWAIT | PK_CLDSIGIGN | PK_32); 365 p2->p_emul = p1->p_emul; 366 p2->p_execsw = p1->p_execsw; 367 368 if (flags & FORK_SYSTEM) { 369 /* 370 * Mark it as a system process. Set P_NOCLDWAIT so that 371 * children are reparented to init(8) when they exit. 372 * init(8) can easily wait them out for us. 373 */ 374 p2->p_flag |= (PK_SYSTEM | PK_NOCLDWAIT); 375 } 376 377 mutex_init(&p2->p_stmutex, MUTEX_DEFAULT, IPL_HIGH); 378 mutex_init(&p2->p_auxlock, MUTEX_DEFAULT, IPL_NONE); 379 rw_init(&p2->p_reflock); 380 cv_init(&p2->p_waitcv, "wait"); 381 cv_init(&p2->p_lwpcv, "lwpwait"); 382 383 /* 384 * Share a lock between the processes if they are to share signal 385 * state: we must synchronize access to it. 386 */ 387 if (flags & FORK_SHARESIGS) { 388 p2->p_lock = p1->p_lock; 389 mutex_obj_hold(p1->p_lock); 390 } else 391 p2->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 392 393 kauth_proc_fork(p1, p2); 394 395 p2->p_raslist = NULL; 396 #if defined(__HAVE_RAS) 397 ras_fork(p1, p2); 398 #endif 399 400 /* bump references to the text vnode (for procfs) */ 401 p2->p_textvp = p1->p_textvp; 402 if (p2->p_textvp) 403 vref(p2->p_textvp); 404 if (p1->p_path) 405 p2->p_path = kmem_strdupsize(p1->p_path, NULL, KM_SLEEP); 406 else 407 p2->p_path = NULL; 408 409 if (flags & FORK_SHAREFILES) 410 fd_share(p2); 411 else if (flags & FORK_CLEANFILES) 412 p2->p_fd = fd_init(NULL); 413 else 414 p2->p_fd = fd_copy(); 415 416 /* XXX racy */ 417 p2->p_mqueue_cnt = p1->p_mqueue_cnt; 418 419 if (flags & FORK_SHARECWD) 420 cwdshare(p2); 421 else 422 p2->p_cwdi = cwdinit(); 423 424 /* 425 * Note: p_limit (rlimit stuff) is copy-on-write, so normally 426 * we just need increase pl_refcnt. 427 */ 428 p1_lim = p1->p_limit; 429 if (!p1_lim->pl_writeable) { 430 lim_addref(p1_lim); 431 p2->p_limit = p1_lim; 432 } else { 433 p2->p_limit = lim_copy(p1_lim); 434 } 435 436 if (flags & FORK_PPWAIT) { 437 /* Mark ourselves as waiting for a child. */ 438 p2->p_lflag = PL_PPWAIT; 439 l1->l_vforkwaiting = true; 440 p2->p_vforklwp = l1; 441 } else { 442 p2->p_lflag = 0; 443 l1->l_vforkwaiting = false; 444 } 445 p2->p_sflag = 0; 446 p2->p_slflag = 0; 447 parent = (flags & FORK_NOWAIT) ? initproc : p1; 448 p2->p_pptr = parent; 449 p2->p_ppid = parent->p_pid; 450 LIST_INIT(&p2->p_children); 451 452 p2->p_aio = NULL; 453 454 #ifdef KTRACE 455 /* 456 * Copy traceflag and tracefile if enabled. 457 * If not inherited, these were zeroed above. 458 */ 459 if (p1->p_traceflag & KTRFAC_INHERIT) { 460 mutex_enter(&ktrace_lock); 461 p2->p_traceflag = p1->p_traceflag; 462 if ((p2->p_tracep = p1->p_tracep) != NULL) 463 ktradref(p2); 464 mutex_exit(&ktrace_lock); 465 } 466 #endif 467 468 /* 469 * Create signal actions for the child process. 470 */ 471 p2->p_sigacts = sigactsinit(p1, flags & FORK_SHARESIGS); 472 mutex_enter(p1->p_lock); 473 p2->p_sflag |= 474 (p1->p_sflag & (PS_STOPFORK | PS_STOPEXEC | PS_NOCLDSTOP)); 475 sched_proc_fork(p1, p2); 476 mutex_exit(p1->p_lock); 477 478 p2->p_stflag = p1->p_stflag; 479 480 /* 481 * p_stats. 482 * Copy parts of p_stats, and zero out the rest. 483 */ 484 p2->p_stats = pstatscopy(p1->p_stats); 485 486 /* 487 * Set up the new process address space. 488 */ 489 uvm_proc_fork(p1, p2, (flags & FORK_SHAREVM) ? true : false); 490 491 /* 492 * Finish creating the child process. 493 * It will return through a different path later. 494 */ 495 lwp_create(l1, p2, uaddr, (flags & FORK_PPWAIT) ? LWP_VFORK : 0, 496 stack, stacksize, (func != NULL) ? func : child_return, arg, &l2, 497 l1->l_class, &l1->l_sigmask, &l1->l_sigstk); 498 499 /* 500 * Inherit l_private from the parent. 501 * Note that we cannot use lwp_setprivate() here since that 502 * also sets the CPU TLS register, which is incorrect if the 503 * process has changed that without letting the kernel know. 504 */ 505 l2->l_private = l1->l_private; 506 507 /* 508 * If emulation has a process fork hook, call it now. 509 */ 510 if (p2->p_emul->e_proc_fork) 511 (*p2->p_emul->e_proc_fork)(p2, l1, flags); 512 513 /* 514 * ...and finally, any other random fork hooks that subsystems 515 * might have registered. 516 */ 517 doforkhooks(p2, p1); 518 519 SDT_PROBE(proc, kernel, , create, p2, p1, flags, 0, 0); 520 521 /* 522 * It's now safe for the scheduler and other processes to see the 523 * child process. 524 */ 525 mutex_enter(&proc_lock); 526 527 if (p1->p_session->s_ttyvp != NULL && p1->p_lflag & PL_CONTROLT) 528 p2->p_lflag |= PL_CONTROLT; 529 530 LIST_INSERT_HEAD(&parent->p_children, p2, p_sibling); 531 p2->p_exitsig = exitsig; /* signal for parent on exit */ 532 533 /* 534 * Trace fork(2) and vfork(2)-like events on demand in a debugger. 535 */ 536 if (tracefork(p1, flags) || tracevfork(p1, flags)) { 537 proc_changeparent(p2, p1->p_pptr); 538 SET(p2->p_slflag, PSL_TRACEDCHILD); 539 } 540 541 p2->p_oppid = p1->p_pid; /* Remember the original parent id. */ 542 543 LIST_INSERT_AFTER(p1, p2, p_pglist); 544 LIST_INSERT_HEAD(&allproc, p2, p_list); 545 546 p2->p_trace_enabled = trace_is_enabled(p2); 547 #ifdef __HAVE_SYSCALL_INTERN 548 (*p2->p_emul->e_syscall_intern)(p2); 549 #endif 550 551 /* 552 * Update stats now that we know the fork was successful. 553 */ 554 KPREEMPT_DISABLE(l1); 555 CPU_COUNT(CPU_COUNT_FORKS, 1); 556 if (flags & FORK_PPWAIT) 557 CPU_COUNT(CPU_COUNT_FORKS_PPWAIT, 1); 558 if (flags & FORK_SHAREVM) 559 CPU_COUNT(CPU_COUNT_FORKS_SHAREVM, 1); 560 KPREEMPT_ENABLE(l1); 561 562 if (ktrpoint(KTR_EMUL)) 563 p2->p_traceflag |= KTRFAC_TRC_EMUL; 564 565 /* 566 * Notify any interested parties about the new process. 567 */ 568 if (!SLIST_EMPTY(&p1->p_klist)) { 569 mutex_exit(&proc_lock); 570 knote_proc_fork(p1, p2); 571 mutex_enter(&proc_lock); 572 } 573 574 /* 575 * Make child runnable, set start time, and add to run queue except 576 * if the parent requested the child to start in SSTOP state. 577 */ 578 mutex_enter(p2->p_lock); 579 580 /* 581 * Start profiling. 582 */ 583 if ((p2->p_stflag & PST_PROFIL) != 0) { 584 mutex_spin_enter(&p2->p_stmutex); 585 startprofclock(p2); 586 mutex_spin_exit(&p2->p_stmutex); 587 } 588 589 getmicrotime(&p2->p_stats->p_start); 590 p2->p_acflag = AFORK; 591 lwp_lock(l2); 592 KASSERT(p2->p_nrlwps == 1); 593 KASSERT(l2->l_stat == LSIDL); 594 if (p2->p_sflag & PS_STOPFORK) { 595 p2->p_nrlwps = 0; 596 p2->p_stat = SSTOP; 597 p2->p_waited = 0; 598 p1->p_nstopchild++; 599 l2->l_stat = LSSTOP; 600 KASSERT(l2->l_wchan == NULL); 601 lwp_unlock(l2); 602 } else { 603 p2->p_nrlwps = 1; 604 p2->p_stat = SACTIVE; 605 setrunnable(l2); 606 /* LWP now unlocked */ 607 } 608 609 /* 610 * Return child pid to parent process, 611 * marking us as parent via retval[1]. 612 */ 613 if (retval != NULL) { 614 retval[0] = p2->p_pid; 615 retval[1] = 0; 616 } 617 618 mutex_exit(p2->p_lock); 619 620 /* 621 * Let the parent know that we are tracing its child. 622 */ 623 if (tracefork(p1, flags) || tracevfork(p1, flags)) { 624 mutex_enter(p1->p_lock); 625 eventswitch(TRAP_CHLD, 626 tracefork(p1, flags) ? PTRACE_FORK : PTRACE_VFORK, 627 retval[0]); 628 mutex_enter(&proc_lock); 629 } 630 631 /* 632 * Preserve synchronization semantics of vfork. If waiting for 633 * child to exec or exit, sleep until it clears p_vforkwaiting. 634 */ 635 while (l1->l_vforkwaiting) 636 cv_wait(&l1->l_waitcv, &proc_lock); 637 638 /* 639 * Let the parent know that we are tracing its child. 640 */ 641 if (tracevforkdone(p1, flags)) { 642 mutex_enter(p1->p_lock); 643 eventswitch(TRAP_CHLD, PTRACE_VFORK_DONE, retval[0]); 644 } else 645 mutex_exit(&proc_lock); 646 647 return 0; 648 } 649 650 /* 651 * MI code executed in each newly spawned process before returning to userland. 652 */ 653 void 654 child_return(void *arg) 655 { 656 struct lwp *l = curlwp; 657 struct proc *p = l->l_proc; 658 659 if ((p->p_slflag & (PSL_TRACED|PSL_TRACEDCHILD)) == 660 (PSL_TRACED|PSL_TRACEDCHILD)) { 661 eventswitchchild(p, TRAP_CHLD, 662 ISSET(p->p_lflag, PL_PPWAIT) ? PTRACE_VFORK : PTRACE_FORK); 663 } 664 665 md_child_return(l); 666 667 /* 668 * Return SYS_fork for all fork types, including vfork(2) and clone(2). 669 * 670 * This approach simplifies the code and avoids extra locking. 671 */ 672 ktrsysret(SYS_fork, 0, 0); 673 } 674
Indexes created Fri Feb 20 16:20:20 UTC 2026