1 /* $NetBSD: kern_fork.c,v 1.232 2025/07/16 19:14:13 kre 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 1.232 2025/07/16 19:14:13 kre Exp $"); 72 73 #include "opt_ktrace.h" 74 #include "opt_dtrace.h" 75 76 #include <sys/param.h> 77 #include <sys/systm.h> 78 #include <sys/filedesc.h> 79 #include <sys/kernel.h> 80 #include <sys/pool.h> 81 #include <sys/mount.h> 82 #include <sys/proc.h> 83 #include <sys/ras.h> 84 #include <sys/resourcevar.h> 85 #include <sys/vnode.h> 86 #include <sys/file.h> 87 #include <sys/acct.h> 88 #include <sys/ktrace.h> 89 #include <sys/sched.h> 90 #include <sys/signalvar.h> 91 #include <sys/syscall.h> 92 #include <sys/kauth.h> 93 #include <sys/atomic.h> 94 #include <sys/syscallargs.h> 95 #include <sys/uidinfo.h> 96 #include <sys/sdt.h> 97 #include <sys/ptrace.h> 98 99 /* 100 * DTrace SDT provider definitions 101 */ 102 SDT_PROVIDER_DECLARE(proc); 103 SDT_PROBE_DEFINE3(proc, kernel, , create, 104 "struct proc *", /* new process */ 105 "struct proc *", /* parent process */ 106 "int" /* flags */); 107 108 u_int nprocs __cacheline_aligned = 1; /* process 0 */ 109 110 /* 111 * Number of ticks to sleep if fork() would fail due to process hitting 112 * limits. Exported in milliseconds to userland via sysctl. 113 */ 114 int forkfsleep = 0; 115 116 int 117 sys_fork(struct lwp *l, const void *v, register_t *retval) 118 { 119 120 return fork1(l, 0, SIGCHLD, NULL, 0, NULL, NULL, retval); 121 } 122 123 /* 124 * vfork(2) system call compatible with 4.4BSD (i.e. BSD with Mach VM). 125 * Address space is not shared, but parent is blocked until child exit. 126 */ 127 int 128 sys_vfork(struct lwp *l, const void *v, register_t *retval) 129 { 130 131 return fork1(l, FORK_PPWAIT, SIGCHLD, NULL, 0, NULL, NULL, 132 retval); 133 } 134 135 /* 136 * New vfork(2) system call for NetBSD, which implements original 3BSD vfork(2) 137 * semantics. Address space is shared, and parent is blocked until child exit. 138 */ 139 int 140 sys___vfork14(struct lwp *l, const void *v, register_t *retval) 141 { 142 143 return fork1(l, FORK_PPWAIT|FORK_SHAREVM, SIGCHLD, NULL, 0, 144 NULL, NULL, retval); 145 } 146 147 /* 148 * Linux-compatible __clone(2) system call. 149 */ 150 int 151 sys___clone(struct lwp *l, const struct sys___clone_args *uap, 152 register_t *retval) 153 { 154 /* { 155 syscallarg(int) flags; 156 syscallarg(void *) stack; 157 } */ 158 int flags, sig; 159 160 /* 161 * We don't support the CLONE_PTRACE flag. 162 */ 163 if (SCARG(uap, flags) & (CLONE_PTRACE)) 164 return EINVAL; 165 166 /* 167 * Linux enforces CLONE_VM with CLONE_SIGHAND, do same. 168 */ 169 if (SCARG(uap, flags) & CLONE_SIGHAND 170 && (SCARG(uap, flags) & CLONE_VM) == 0) 171 return EINVAL; 172 173 flags = 0; 174 175 if (SCARG(uap, flags) & CLONE_VM) 176 flags |= FORK_SHAREVM; 177 if (SCARG(uap, flags) & CLONE_FS) 178 flags |= FORK_SHARECWD; 179 if (SCARG(uap, flags) & CLONE_FILES) 180 flags |= FORK_SHAREFILES; 181 if (SCARG(uap, flags) & CLONE_SIGHAND) 182 flags |= FORK_SHARESIGS; 183 if (SCARG(uap, flags) & CLONE_VFORK) 184 flags |= FORK_PPWAIT; 185 186 sig = SCARG(uap, flags) & CLONE_CSIGNAL; 187 if (sig < 0 || sig >= _NSIG) 188 return EINVAL; 189 190 /* 191 * Linux doesn't have close-on-fork yet, so we don't 192 * know what they will do combining CLONE_FILES with 193 * close-on-fork (which are not really compatible). 194 * This might need to be changed in the future (another 195 * option would be to just disable FORK_SHAREFILES) 196 */ 197 if ((flags & FORK_SHAREFILES) != 0) { 198 if (l->l_fd != NULL && l->l_fd->fd_foclose) 199 return EINVAL; 200 } 201 202 /* 203 * Note that the Linux API does not provide a portable way of 204 * specifying the stack area; the caller must know if the stack 205 * grows up or down. So, we pass a stack size of 0, so that the 206 * code that makes this adjustment is a noop. 207 */ 208 return fork1(l, flags, sig, SCARG(uap, stack), 0, 209 NULL, NULL, retval); 210 } 211 212 /* 213 * Print the 'table full' message once per 10 seconds. 214 */ 215 static struct timeval fork_tfmrate = { 10, 0 }; 216 217 /* 218 * Check if a process is traced and shall inform about FORK events. 219 */ 220 static inline bool 221 tracefork(struct proc *p, int flags) 222 { 223 224 return (p->p_slflag & (PSL_TRACEFORK|PSL_TRACED)) == 225 (PSL_TRACEFORK|PSL_TRACED) && (flags & FORK_PPWAIT) == 0; 226 } 227 228 /* 229 * Check if a process is traced and shall inform about VFORK events. 230 */ 231 static inline bool 232 tracevfork(struct proc *p, int flags) 233 { 234 235 return (p->p_slflag & (PSL_TRACEVFORK|PSL_TRACED)) == 236 (PSL_TRACEVFORK|PSL_TRACED) && (flags & FORK_PPWAIT) != 0; 237 } 238 239 /* 240 * Check if a process is traced and shall inform about VFORK_DONE events. 241 */ 242 static inline bool 243 tracevforkdone(struct proc *p, int flags) 244 { 245 246 return (p->p_slflag & (PSL_TRACEVFORK_DONE|PSL_TRACED)) == 247 (PSL_TRACEVFORK_DONE|PSL_TRACED) && (flags & FORK_PPWAIT); 248 } 249 250 /* 251 * General fork call. Note that another LWP in the process may call exec() 252 * or exit() while we are forking. It's safe to continue here, because 253 * neither operation will complete until all LWPs have exited the process. 254 */ 255 int 256 fork1(struct lwp *l1, int flags, int exitsig, void *stack, size_t stacksize, 257 void (*func)(void *), void *arg, register_t *retval) 258 { 259 struct proc *p1, *p2, *parent; 260 struct plimit *p1_lim; 261 uid_t uid; 262 struct lwp *l2; 263 int count; 264 vaddr_t uaddr; 265 int tnprocs; 266 int error = 0; 267 268 p1 = l1->l_proc; 269 uid = kauth_cred_getuid(l1->l_cred); 270 tnprocs = atomic_inc_uint_nv(&nprocs); 271 272 /* 273 * Although process entries are dynamically created, we still keep 274 * a global limit on the maximum number we will create. 275 */ 276 if (__predict_false(tnprocs >= maxproc)) 277 error = -1; 278 else 279 error = kauth_authorize_process(l1->l_cred, 280 KAUTH_PROCESS_FORK, p1, KAUTH_ARG(tnprocs), NULL, NULL); 281 282 if (error) { 283 static struct timeval lasttfm; 284 atomic_dec_uint(&nprocs); 285 if (ratecheck(&lasttfm, &fork_tfmrate)) 286 tablefull("proc", "increase kern.maxproc or NPROC"); 287 if (forkfsleep) 288 kpause("forkmx", false, forkfsleep, NULL); 289 return EAGAIN; 290 } 291 292 /* 293 * Enforce limits. 294 */ 295 count = chgproccnt(uid, 1); 296 if (__predict_false(count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur)) { 297 if (kauth_authorize_process(l1->l_cred, KAUTH_PROCESS_RLIMIT, 298 p1, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_BYPASS), 299 &p1->p_rlimit[RLIMIT_NPROC], KAUTH_ARG(RLIMIT_NPROC)) != 0) { 300 (void)chgproccnt(uid, -1); 301 atomic_dec_uint(&nprocs); 302 if (forkfsleep) 303 kpause("forkulim", false, forkfsleep, NULL); 304 return EAGAIN; 305 } 306 } 307 308 /* 309 * Allocate virtual address space for the U-area now, while it 310 * is still easy to abort the fork operation if we're out of 311 * kernel virtual address space. 312 */ 313 uaddr = uvm_uarea_alloc(); 314 if (__predict_false(uaddr == 0)) { 315 (void)chgproccnt(uid, -1); 316 atomic_dec_uint(&nprocs); 317 return ENOMEM; 318 } 319 320 /* Allocate new proc. */ 321 p2 = proc_alloc(); 322 if (p2 == NULL) { 323 /* We were unable to allocate a process ID. */ 324 uvm_uarea_free(uaddr); 325 mutex_enter(p1->p_lock); 326 uid = kauth_cred_getuid(p1->p_cred); 327 (void)chgproccnt(uid, -1); 328 mutex_exit(p1->p_lock); 329 atomic_dec_uint(&nprocs); 330 return EAGAIN; 331 } 332 333 /* 334 * We are now committed to the fork. From here on, we may 335 * block on resources, but resource allocation may NOT fail. 336 */ 337 338 /* 339 * Make a proc table entry for the new process. 340 * Start by zeroing the section of proc that is zero-initialized, 341 * then copy the section that is copied directly from the parent. 342 */ 343 memset(&p2->p_startzero, 0, 344 (unsigned) ((char *)&p2->p_endzero - (char *)&p2->p_startzero)); 345 memcpy(&p2->p_startcopy, &p1->p_startcopy, 346 (unsigned) ((char *)&p2->p_endcopy - (char *)&p2->p_startcopy)); 347 348 TAILQ_INIT(&p2->p_sigpend.sp_info); 349 350 LIST_INIT(&p2->p_lwps); 351 LIST_INIT(&p2->p_sigwaiters); 352 353 /* 354 * Duplicate sub-structures as needed. 355 * Increase reference counts on shared objects. 356 * Inherit flags we want to keep. The flags related to SIGCHLD 357 * handling are important in order to keep a consistent behaviour 358 * for the child after the fork. If we are a 32-bit process, the 359 * child will be too. 360 */ 361 p2->p_flag = 362 p1->p_flag & (PK_SUGID | PK_NOCLDWAIT | PK_CLDSIGIGN | PK_32); 363 p2->p_emul = p1->p_emul; 364 p2->p_execsw = p1->p_execsw; 365 366 if (flags & FORK_SYSTEM) { 367 /* 368 * Mark it as a system process. Set P_NOCLDWAIT so that 369 * children are reparented to init(8) when they exit. 370 * init(8) can easily wait them out for us. 371 */ 372 p2->p_flag |= (PK_SYSTEM | PK_NOCLDWAIT); 373 } 374 375 mutex_init(&p2->p_stmutex, MUTEX_DEFAULT, IPL_HIGH); 376 mutex_init(&p2->p_auxlock, MUTEX_DEFAULT, IPL_NONE); 377 rw_init(&p2->p_reflock); 378 cv_init(&p2->p_waitcv, "wait"); 379 cv_init(&p2->p_lwpcv, "lwpwait"); 380 381 /* 382 * Share a lock between the processes if they are to share signal 383 * state: we must synchronize access to it. 384 */ 385 if (flags & FORK_SHARESIGS) { 386 p2->p_lock = p1->p_lock; 387 mutex_obj_hold(p1->p_lock); 388 } else 389 p2->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 390 391 kauth_proc_fork(p1, p2); 392 393 p2->p_raslist = NULL; 394 #if defined(__HAVE_RAS) 395 ras_fork(p1, p2); 396 #endif 397 398 /* bump references to the text vnode (for procfs) */ 399 p2->p_textvp = p1->p_textvp; 400 if (p2->p_textvp) 401 vref(p2->p_textvp); 402 if (p1->p_path) 403 p2->p_path = kmem_strdupsize(p1->p_path, NULL, KM_SLEEP); 404 else 405 p2->p_path = NULL; 406 407 if (flags & FORK_SHAREFILES) 408 fd_share(p2); 409 else if (flags & FORK_CLEANFILES) 410 p2->p_fd = fd_init(NULL); 411 else 412 p2->p_fd = fd_copy(); 413 414 /* XXX racy */ 415 p2->p_mqueue_cnt = p1->p_mqueue_cnt; 416 417 if (flags & FORK_SHARECWD) 418 cwdshare(p2); 419 else 420 p2->p_cwdi = cwdinit(); 421 422 /* 423 * Note: p_limit (rlimit stuff) is copy-on-write, so normally 424 * we just need increase pl_refcnt. 425 */ 426 p1_lim = p1->p_limit; 427 if (!p1_lim->pl_writeable) { 428 lim_addref(p1_lim); 429 p2->p_limit = p1_lim; 430 } else { 431 p2->p_limit = lim_copy(p1_lim); 432 } 433 434 if (flags & FORK_PPWAIT) { 435 /* Mark ourselves as waiting for a child. */ 436 p2->p_lflag = PL_PPWAIT; 437 l1->l_vforkwaiting = true; 438 p2->p_vforklwp = l1; 439 } else { 440 p2->p_lflag = 0; 441 l1->l_vforkwaiting = false; 442 } 443 p2->p_sflag = 0; 444 p2->p_slflag = 0; 445 parent = (flags & FORK_NOWAIT) ? initproc : p1; 446 p2->p_pptr = parent; 447 p2->p_ppid = parent->p_pid; 448 LIST_INIT(&p2->p_children); 449 450 p2->p_aio = NULL; 451 452 #ifdef KTRACE 453 /* 454 * Copy traceflag and tracefile if enabled. 455 * If not inherited, these were zeroed above. 456 */ 457 if (p1->p_traceflag & KTRFAC_INHERIT) { 458 mutex_enter(&ktrace_lock); 459 p2->p_traceflag = p1->p_traceflag; 460 if ((p2->p_tracep = p1->p_tracep) != NULL) 461 ktradref(p2); 462 mutex_exit(&ktrace_lock); 463 } 464 #endif 465 466 /* 467 * Create signal actions for the child process. 468 */ 469 p2->p_sigacts = sigactsinit(p1, flags & FORK_SHARESIGS); 470 mutex_enter(p1->p_lock); 471 p2->p_sflag |= 472 (p1->p_sflag & (PS_STOPFORK | PS_STOPEXEC | PS_NOCLDSTOP)); 473 sched_proc_fork(p1, p2); 474 mutex_exit(p1->p_lock); 475 476 p2->p_stflag = p1->p_stflag; 477 478 /* 479 * p_stats. 480 * Copy parts of p_stats, and zero out the rest. 481 */ 482 p2->p_stats = pstatscopy(p1->p_stats); 483 484 /* 485 * Set up the new process address space. 486 */ 487 uvm_proc_fork(p1, p2, (flags & FORK_SHAREVM) ? true : false); 488 489 /* 490 * Finish creating the child process. 491 * It will return through a different path later. 492 */ 493 lwp_create(l1, p2, uaddr, (flags & FORK_PPWAIT) ? LWP_VFORK : 0, 494 stack, stacksize, (func != NULL) ? func : child_return, arg, &l2, 495 l1->l_class, &l1->l_sigmask, &l1->l_sigstk); 496 497 /* 498 * Inherit l_private from the parent. 499 * Note that we cannot use lwp_setprivate() here since that 500 * also sets the CPU TLS register, which is incorrect if the 501 * process has changed that without letting the kernel know. 502 */ 503 l2->l_private = l1->l_private; 504 505 /* 506 * If emulation has a process fork hook, call it now. 507 */ 508 if (p2->p_emul->e_proc_fork) 509 (*p2->p_emul->e_proc_fork)(p2, l1, flags); 510 511 /* 512 * ...and finally, any other random fork hooks that subsystems 513 * might have registered. 514 */ 515 doforkhooks(p2, p1); 516 517 SDT_PROBE(proc, kernel, , create, p2, p1, flags, 0, 0); 518 519 /* 520 * It's now safe for the scheduler and other processes to see the 521 * child process. 522 */ 523 mutex_enter(&proc_lock); 524 525 if (p1->p_session->s_ttyvp != NULL && p1->p_lflag & PL_CONTROLT) 526 p2->p_lflag |= PL_CONTROLT; 527 528 LIST_INSERT_HEAD(&parent->p_children, p2, p_sibling); 529 p2->p_exitsig = exitsig; /* signal for parent on exit */ 530 531 /* 532 * Trace fork(2) and vfork(2)-like events on demand in a debugger. 533 */ 534 if (tracefork(p1, flags) || tracevfork(p1, flags)) { 535 proc_changeparent(p2, p1->p_pptr); 536 SET(p2->p_slflag, PSL_TRACEDCHILD); 537 } 538 539 p2->p_oppid = p1->p_pid; /* Remember the original parent id. */ 540 541 LIST_INSERT_AFTER(p1, p2, p_pglist); 542 LIST_INSERT_HEAD(&allproc, p2, p_list); 543 544 p2->p_trace_enabled = trace_is_enabled(p2); 545 #ifdef __HAVE_SYSCALL_INTERN 546 (*p2->p_emul->e_syscall_intern)(p2); 547 #endif 548 549 /* 550 * Update stats now that we know the fork was successful. 551 */ 552 KPREEMPT_DISABLE(l1); 553 CPU_COUNT(CPU_COUNT_FORKS, 1); 554 if (flags & FORK_PPWAIT) 555 CPU_COUNT(CPU_COUNT_FORKS_PPWAIT, 1); 556 if (flags & FORK_SHAREVM) 557 CPU_COUNT(CPU_COUNT_FORKS_SHAREVM, 1); 558 KPREEMPT_ENABLE(l1); 559 560 if (ktrpoint(KTR_EMUL)) 561 p2->p_traceflag |= KTRFAC_TRC_EMUL; 562 563 /* 564 * Notify any interested parties about the new process. 565 */ 566 if (!SLIST_EMPTY(&p1->p_klist)) { 567 mutex_exit(&proc_lock); 568 knote_proc_fork(p1, p2); 569 mutex_enter(&proc_lock); 570 } 571 572 /* 573 * Make child runnable, set start time, and add to run queue except 574 * if the parent requested the child to start in SSTOP state. 575 */ 576 mutex_enter(p2->p_lock); 577 578 /* 579 * Start profiling. 580 */ 581 if ((p2->p_stflag & PST_PROFIL) != 0) { 582 mutex_spin_enter(&p2->p_stmutex); 583 startprofclock(p2); 584 mutex_spin_exit(&p2->p_stmutex); 585 } 586 587 getmicrotime(&p2->p_stats->p_start); 588 p2->p_acflag = AFORK; 589 lwp_lock(l2); 590 KASSERT(p2->p_nrlwps == 1); 591 KASSERT(l2->l_stat == LSIDL); 592 if (p2->p_sflag & PS_STOPFORK) { 593 p2->p_nrlwps = 0; 594 p2->p_stat = SSTOP; 595 p2->p_waited = 0; 596 p1->p_nstopchild++; 597 l2->l_stat = LSSTOP; 598 KASSERT(l2->l_wchan == NULL); 599 lwp_unlock(l2); 600 } else { 601 p2->p_nrlwps = 1; 602 p2->p_stat = SACTIVE; 603 setrunnable(l2); 604 /* LWP now unlocked */ 605 } 606 607 /* 608 * Return child pid to parent process, 609 * marking us as parent via retval[1]. 610 */ 611 if (retval != NULL) { 612 retval[0] = p2->p_pid; 613 retval[1] = 0; 614 } 615 616 mutex_exit(p2->p_lock); 617 618 /* 619 * Let the parent know that we are tracing its child. 620 */ 621 if (tracefork(p1, flags) || tracevfork(p1, flags)) { 622 mutex_enter(p1->p_lock); 623 eventswitch(TRAP_CHLD, 624 tracefork(p1, flags) ? PTRACE_FORK : PTRACE_VFORK, 625 retval[0]); 626 mutex_enter(&proc_lock); 627 } 628 629 /* 630 * Preserve synchronization semantics of vfork. If waiting for 631 * child to exec or exit, sleep until it clears p_vforkwaiting. 632 */ 633 while (l1->l_vforkwaiting) 634 cv_wait(&l1->l_waitcv, &proc_lock); 635 636 /* 637 * Let the parent know that we are tracing its child. 638 */ 639 if (tracevforkdone(p1, flags)) { 640 mutex_enter(p1->p_lock); 641 eventswitch(TRAP_CHLD, PTRACE_VFORK_DONE, retval[0]); 642 } else 643 mutex_exit(&proc_lock); 644 645 return 0; 646 } 647 648 /* 649 * MI code executed in each newly spawned process before returning to userland. 650 */ 651 void 652 child_return(void *arg) 653 { 654 struct lwp *l = curlwp; 655 struct proc *p = l->l_proc; 656 657 if ((p->p_slflag & (PSL_TRACED|PSL_TRACEDCHILD)) == 658 (PSL_TRACED|PSL_TRACEDCHILD)) { 659 eventswitchchild(p, TRAP_CHLD, 660 ISSET(p->p_lflag, PL_PPWAIT) ? PTRACE_VFORK : PTRACE_FORK); 661 } 662 663 md_child_return(l); 664 665 /* 666 * Return SYS_fork for all fork types, including vfork(2) and clone(2). 667 * 668 * This approach simplifies the code and avoids extra locking. 669 */ 670 ktrsysret(SYS_fork, 0, 0); 671 } 672
Indexes created Sat Sep 20 22:09:52 GMT 2025