1 /* $NetBSD: kern_fork.c,v 1.212 2019/05/03 22:34:21 kamil Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2001, 2004, 2006, 2007, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility, 9 * NASA Ames Research Center, by Charles M. Hannum, and by Andrew Doran. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1989, 1991, 1993 35 * The Regents of the University of California. All rights reserved. 36 * (c) UNIX System Laboratories, Inc. 37 * All or some portions of this file are derived from material licensed 38 * to the University of California by American Telephone and Telegraph 39 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 40 * the permission of UNIX System Laboratories, Inc. 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 3. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * @(#)kern_fork.c 8.8 (Berkeley) 2/14/95 67 */ 68 69 #include <sys/cdefs.h> 70 __KERNEL_RCSID(0, "$NetBSD: kern_fork.c,v 1.212 2019/05/03 22:34:21 kamil Exp $"); 71 72 #include "opt_ktrace.h" 73 #include "opt_dtrace.h" 74 75 #include <sys/param.h> 76 #include <sys/systm.h> 77 #include <sys/filedesc.h> 78 #include <sys/kernel.h> 79 #include <sys/pool.h> 80 #include <sys/mount.h> 81 #include <sys/proc.h> 82 #include <sys/ras.h> 83 #include <sys/resourcevar.h> 84 #include <sys/vnode.h> 85 #include <sys/file.h> 86 #include <sys/acct.h> 87 #include <sys/ktrace.h> 88 #include <sys/sched.h> 89 #include <sys/signalvar.h> 90 #include <sys/syscall.h> 91 #include <sys/kauth.h> 92 #include <sys/atomic.h> 93 #include <sys/syscallargs.h> 94 #include <sys/uidinfo.h> 95 #include <sys/sdt.h> 96 #include <sys/ptrace.h> 97 98 #include <uvm/uvm_extern.h> 99 100 /* 101 * DTrace SDT provider definitions 102 */ 103 SDT_PROVIDER_DECLARE(proc); 104 SDT_PROBE_DEFINE3(proc, kernel, , create, 105 "struct proc *", /* new process */ 106 "struct proc *", /* parent process */ 107 "int" /* flags */); 108 109 u_int nprocs __cacheline_aligned = 1; /* process 0 */ 110 111 /* 112 * Number of ticks to sleep if fork() would fail due to process hitting 113 * limits. Exported in miliseconds to userland via sysctl. 114 */ 115 int forkfsleep = 0; 116 117 int 118 sys_fork(struct lwp *l, const void *v, register_t *retval) 119 { 120 121 return fork1(l, 0, SIGCHLD, NULL, 0, NULL, NULL, retval); 122 } 123 124 /* 125 * vfork(2) system call compatible with 4.4BSD (i.e. BSD with Mach VM). 126 * Address space is not shared, but parent is blocked until child exit. 127 */ 128 int 129 sys_vfork(struct lwp *l, const void *v, register_t *retval) 130 { 131 132 return fork1(l, FORK_PPWAIT, SIGCHLD, NULL, 0, NULL, NULL, 133 retval); 134 } 135 136 /* 137 * New vfork(2) system call for NetBSD, which implements original 3BSD vfork(2) 138 * semantics. Address space is shared, and parent is blocked until child exit. 139 */ 140 int 141 sys___vfork14(struct lwp *l, const void *v, register_t *retval) 142 { 143 144 return fork1(l, FORK_PPWAIT|FORK_SHAREVM, SIGCHLD, NULL, 0, 145 NULL, NULL, retval); 146 } 147 148 /* 149 * Linux-compatible __clone(2) system call. 150 */ 151 int 152 sys___clone(struct lwp *l, const struct sys___clone_args *uap, 153 register_t *retval) 154 { 155 /* { 156 syscallarg(int) flags; 157 syscallarg(void *) stack; 158 } */ 159 int flags, sig; 160 161 /* 162 * We don't support the CLONE_PID or CLONE_PTRACE flags. 163 */ 164 if (SCARG(uap, flags) & (CLONE_PID|CLONE_PTRACE)) 165 return EINVAL; 166 167 /* 168 * Linux enforces CLONE_VM with CLONE_SIGHAND, do same. 169 */ 170 if (SCARG(uap, flags) & CLONE_SIGHAND 171 && (SCARG(uap, flags) & CLONE_VM) == 0) 172 return EINVAL; 173 174 flags = 0; 175 176 if (SCARG(uap, flags) & CLONE_VM) 177 flags |= FORK_SHAREVM; 178 if (SCARG(uap, flags) & CLONE_FS) 179 flags |= FORK_SHARECWD; 180 if (SCARG(uap, flags) & CLONE_FILES) 181 flags |= FORK_SHAREFILES; 182 if (SCARG(uap, flags) & CLONE_SIGHAND) 183 flags |= FORK_SHARESIGS; 184 if (SCARG(uap, flags) & CLONE_VFORK) 185 flags |= FORK_PPWAIT; 186 187 sig = SCARG(uap, flags) & CLONE_CSIGNAL; 188 if (sig < 0 || sig >= _NSIG) 189 return EINVAL; 190 191 /* 192 * Note that the Linux API does not provide a portable way of 193 * specifying the stack area; the caller must know if the stack 194 * grows up or down. So, we pass a stack size of 0, so that the 195 * code that makes this adjustment is a noop. 196 */ 197 return fork1(l, flags, sig, SCARG(uap, stack), 0, 198 NULL, NULL, retval); 199 } 200 201 /* 202 * Print the 'table full' message once per 10 seconds. 203 */ 204 static struct timeval fork_tfmrate = { 10, 0 }; 205 206 /* 207 * Check if a process is traced and shall inform about FORK events. 208 */ 209 static inline bool 210 tracefork(struct proc *p, int flags) 211 { 212 213 return (p->p_slflag & (PSL_TRACEFORK|PSL_TRACED)) == 214 (PSL_TRACEFORK|PSL_TRACED) && (flags & FORK_PPWAIT) == 0; 215 } 216 217 /* 218 * Check if a process is traced and shall inform about VFORK events. 219 */ 220 static inline bool 221 tracevfork(struct proc *p, int flags) 222 { 223 224 return (p->p_slflag & (PSL_TRACEVFORK|PSL_TRACED)) == 225 (PSL_TRACEVFORK|PSL_TRACED) && (flags & FORK_PPWAIT) != 0; 226 } 227 228 /* 229 * Check if a process is traced and shall inform about VFORK_DONE events. 230 */ 231 static inline bool 232 tracevforkdone(struct proc *p, int flags) 233 { 234 235 return (p->p_slflag & (PSL_TRACEVFORK_DONE|PSL_TRACED)) == 236 (PSL_TRACEVFORK_DONE|PSL_TRACED) && (flags & FORK_PPWAIT); 237 } 238 239 /* 240 * General fork call. Note that another LWP in the process may call exec() 241 * or exit() while we are forking. It's safe to continue here, because 242 * neither operation will complete until all LWPs have exited the process. 243 */ 244 int 245 fork1(struct lwp *l1, int flags, int exitsig, void *stack, size_t stacksize, 246 void (*func)(void *), void *arg, register_t *retval) 247 { 248 struct proc *p1, *p2, *parent; 249 struct plimit *p1_lim; 250 uid_t uid; 251 struct lwp *l2; 252 int count; 253 vaddr_t uaddr; 254 int tnprocs; 255 bool trace_fork, trace_vfork; 256 int error = 0; 257 258 p1 = l1->l_proc; 259 uid = kauth_cred_getuid(l1->l_cred); 260 tnprocs = atomic_inc_uint_nv(&nprocs); 261 262 /* 263 * Although process entries are dynamically created, we still keep 264 * a global limit on the maximum number we will create. 265 */ 266 if (__predict_false(tnprocs >= maxproc)) 267 error = -1; 268 else 269 error = kauth_authorize_process(l1->l_cred, 270 KAUTH_PROCESS_FORK, p1, KAUTH_ARG(tnprocs), NULL, NULL); 271 272 if (error) { 273 static struct timeval lasttfm; 274 atomic_dec_uint(&nprocs); 275 if (ratecheck(&lasttfm, &fork_tfmrate)) 276 tablefull("proc", "increase kern.maxproc or NPROC"); 277 if (forkfsleep) 278 kpause("forkmx", false, forkfsleep, NULL); 279 return EAGAIN; 280 } 281 282 /* 283 * Enforce limits. 284 */ 285 count = chgproccnt(uid, 1); 286 if (__predict_false(count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur)) { 287 if (kauth_authorize_process(l1->l_cred, KAUTH_PROCESS_RLIMIT, 288 p1, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_BYPASS), 289 &p1->p_rlimit[RLIMIT_NPROC], KAUTH_ARG(RLIMIT_NPROC)) != 0) { 290 (void)chgproccnt(uid, -1); 291 atomic_dec_uint(&nprocs); 292 if (forkfsleep) 293 kpause("forkulim", false, forkfsleep, NULL); 294 return EAGAIN; 295 } 296 } 297 298 /* 299 * Allocate virtual address space for the U-area now, while it 300 * is still easy to abort the fork operation if we're out of 301 * kernel virtual address space. 302 */ 303 uaddr = uvm_uarea_alloc(); 304 if (__predict_false(uaddr == 0)) { 305 (void)chgproccnt(uid, -1); 306 atomic_dec_uint(&nprocs); 307 return ENOMEM; 308 } 309 310 /* 311 * We are now committed to the fork. From here on, we may 312 * block on resources, but resource allocation may NOT fail. 313 */ 314 315 /* Allocate new proc. */ 316 p2 = proc_alloc(); 317 318 /* 319 * Make a proc table entry for the new process. 320 * Start by zeroing the section of proc that is zero-initialized, 321 * then copy the section that is copied directly from the parent. 322 */ 323 memset(&p2->p_startzero, 0, 324 (unsigned) ((char *)&p2->p_endzero - (char *)&p2->p_startzero)); 325 memcpy(&p2->p_startcopy, &p1->p_startcopy, 326 (unsigned) ((char *)&p2->p_endcopy - (char *)&p2->p_startcopy)); 327 328 TAILQ_INIT(&p2->p_sigpend.sp_info); 329 330 LIST_INIT(&p2->p_lwps); 331 LIST_INIT(&p2->p_sigwaiters); 332 333 /* 334 * Duplicate sub-structures as needed. 335 * Increase reference counts on shared objects. 336 * Inherit flags we want to keep. The flags related to SIGCHLD 337 * handling are important in order to keep a consistent behaviour 338 * for the child after the fork. If we are a 32-bit process, the 339 * child will be too. 340 */ 341 p2->p_flag = 342 p1->p_flag & (PK_SUGID | PK_NOCLDWAIT | PK_CLDSIGIGN | PK_32); 343 p2->p_emul = p1->p_emul; 344 p2->p_execsw = p1->p_execsw; 345 346 if (flags & FORK_SYSTEM) { 347 /* 348 * Mark it as a system process. Set P_NOCLDWAIT so that 349 * children are reparented to init(8) when they exit. 350 * init(8) can easily wait them out for us. 351 */ 352 p2->p_flag |= (PK_SYSTEM | PK_NOCLDWAIT); 353 } 354 355 mutex_init(&p2->p_stmutex, MUTEX_DEFAULT, IPL_HIGH); 356 mutex_init(&p2->p_auxlock, MUTEX_DEFAULT, IPL_NONE); 357 rw_init(&p2->p_reflock); 358 cv_init(&p2->p_waitcv, "wait"); 359 cv_init(&p2->p_lwpcv, "lwpwait"); 360 361 /* 362 * Share a lock between the processes if they are to share signal 363 * state: we must synchronize access to it. 364 */ 365 if (flags & FORK_SHARESIGS) { 366 p2->p_lock = p1->p_lock; 367 mutex_obj_hold(p1->p_lock); 368 } else 369 p2->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 370 371 kauth_proc_fork(p1, p2); 372 373 p2->p_raslist = NULL; 374 #if defined(__HAVE_RAS) 375 ras_fork(p1, p2); 376 #endif 377 378 /* bump references to the text vnode (for procfs) */ 379 p2->p_textvp = p1->p_textvp; 380 if (p2->p_textvp) 381 vref(p2->p_textvp); 382 if (p1->p_path) 383 p2->p_path = kmem_strdupsize(p1->p_path, NULL, KM_SLEEP); 384 else 385 p2->p_path = NULL; 386 387 if (flags & FORK_SHAREFILES) 388 fd_share(p2); 389 else if (flags & FORK_CLEANFILES) 390 p2->p_fd = fd_init(NULL); 391 else 392 p2->p_fd = fd_copy(); 393 394 /* XXX racy */ 395 p2->p_mqueue_cnt = p1->p_mqueue_cnt; 396 397 if (flags & FORK_SHARECWD) 398 cwdshare(p2); 399 else 400 p2->p_cwdi = cwdinit(); 401 402 /* 403 * Note: p_limit (rlimit stuff) is copy-on-write, so normally 404 * we just need increase pl_refcnt. 405 */ 406 p1_lim = p1->p_limit; 407 if (!p1_lim->pl_writeable) { 408 lim_addref(p1_lim); 409 p2->p_limit = p1_lim; 410 } else { 411 p2->p_limit = lim_copy(p1_lim); 412 } 413 414 if (flags & FORK_PPWAIT) { 415 /* Mark ourselves as waiting for a child. */ 416 l1->l_pflag |= LP_VFORKWAIT; 417 p2->p_lflag = PL_PPWAIT; 418 p2->p_vforklwp = l1; 419 } else { 420 p2->p_lflag = 0; 421 } 422 p2->p_sflag = 0; 423 p2->p_slflag = 0; 424 parent = (flags & FORK_NOWAIT) ? initproc : p1; 425 p2->p_pptr = parent; 426 p2->p_ppid = parent->p_pid; 427 LIST_INIT(&p2->p_children); 428 429 p2->p_aio = NULL; 430 431 #ifdef KTRACE 432 /* 433 * Copy traceflag and tracefile if enabled. 434 * If not inherited, these were zeroed above. 435 */ 436 if (p1->p_traceflag & KTRFAC_INHERIT) { 437 mutex_enter(&ktrace_lock); 438 p2->p_traceflag = p1->p_traceflag; 439 if ((p2->p_tracep = p1->p_tracep) != NULL) 440 ktradref(p2); 441 mutex_exit(&ktrace_lock); 442 } 443 #endif 444 445 /* 446 * Create signal actions for the child process. 447 */ 448 p2->p_sigacts = sigactsinit(p1, flags & FORK_SHARESIGS); 449 mutex_enter(p1->p_lock); 450 p2->p_sflag |= 451 (p1->p_sflag & (PS_STOPFORK | PS_STOPEXEC | PS_NOCLDSTOP)); 452 sched_proc_fork(p1, p2); 453 mutex_exit(p1->p_lock); 454 455 p2->p_stflag = p1->p_stflag; 456 457 /* 458 * p_stats. 459 * Copy parts of p_stats, and zero out the rest. 460 */ 461 p2->p_stats = pstatscopy(p1->p_stats); 462 463 /* 464 * Set up the new process address space. 465 */ 466 uvm_proc_fork(p1, p2, (flags & FORK_SHAREVM) ? true : false); 467 468 /* 469 * Finish creating the child process. 470 * It will return through a different path later. 471 */ 472 lwp_create(l1, p2, uaddr, (flags & FORK_PPWAIT) ? LWP_VFORK : 0, 473 stack, stacksize, (func != NULL) ? func : child_return, arg, &l2, 474 l1->l_class, &l1->l_sigmask, &l1->l_sigstk); 475 476 /* 477 * Inherit l_private from the parent. 478 * Note that we cannot use lwp_setprivate() here since that 479 * also sets the CPU TLS register, which is incorrect if the 480 * process has changed that without letting the kernel know. 481 */ 482 l2->l_private = l1->l_private; 483 484 /* 485 * If emulation has a process fork hook, call it now. 486 */ 487 if (p2->p_emul->e_proc_fork) 488 (*p2->p_emul->e_proc_fork)(p2, l1, flags); 489 490 /* 491 * ...and finally, any other random fork hooks that subsystems 492 * might have registered. 493 */ 494 doforkhooks(p2, p1); 495 496 SDT_PROBE(proc, kernel, , create, p2, p1, flags, 0, 0); 497 498 /* 499 * It's now safe for the scheduler and other processes to see the 500 * child process. 501 */ 502 mutex_enter(proc_lock); 503 504 if (p1->p_session->s_ttyvp != NULL && p1->p_lflag & PL_CONTROLT) 505 p2->p_lflag |= PL_CONTROLT; 506 507 LIST_INSERT_HEAD(&parent->p_children, p2, p_sibling); 508 p2->p_exitsig = exitsig; /* signal for parent on exit */ 509 510 /* 511 * Trace fork(2) and vfork(2)-like events on demand in a debugger. 512 */ 513 trace_fork = tracefork(p1, flags); 514 trace_vfork = tracevfork(p1, flags); 515 if (trace_fork || trace_vfork) 516 proc_changeparent(p2, p1->p_pptr); 517 if (trace_fork) { 518 p1->p_fpid = p2->p_pid; 519 p2->p_fpid = p1->p_pid; 520 } 521 if (trace_vfork) { 522 p1->p_vfpid = p2->p_pid; 523 p2->p_vfpid = p1->p_pid; 524 } 525 526 LIST_INSERT_AFTER(p1, p2, p_pglist); 527 LIST_INSERT_HEAD(&allproc, p2, p_list); 528 529 p2->p_trace_enabled = trace_is_enabled(p2); 530 #ifdef __HAVE_SYSCALL_INTERN 531 (*p2->p_emul->e_syscall_intern)(p2); 532 #endif 533 534 /* 535 * Update stats now that we know the fork was successful. 536 */ 537 uvmexp.forks++; 538 if (flags & FORK_PPWAIT) 539 uvmexp.forks_ppwait++; 540 if (flags & FORK_SHAREVM) 541 uvmexp.forks_sharevm++; 542 543 if (ktrpoint(KTR_EMUL)) 544 p2->p_traceflag |= KTRFAC_TRC_EMUL; 545 546 /* 547 * Notify any interested parties about the new process. 548 */ 549 if (!SLIST_EMPTY(&p1->p_klist)) { 550 mutex_exit(proc_lock); 551 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid); 552 mutex_enter(proc_lock); 553 } 554 555 /* 556 * Make child runnable, set start time, and add to run queue except 557 * if the parent requested the child to start in SSTOP state. 558 */ 559 mutex_enter(p2->p_lock); 560 561 /* 562 * Start profiling. 563 */ 564 if ((p2->p_stflag & PST_PROFIL) != 0) { 565 mutex_spin_enter(&p2->p_stmutex); 566 startprofclock(p2); 567 mutex_spin_exit(&p2->p_stmutex); 568 } 569 570 getmicrotime(&p2->p_stats->p_start); 571 p2->p_acflag = AFORK; 572 lwp_lock(l2); 573 KASSERT(p2->p_nrlwps == 1); 574 if (p2->p_sflag & PS_STOPFORK) { 575 struct schedstate_percpu *spc = &l2->l_cpu->ci_schedstate; 576 p2->p_nrlwps = 0; 577 p2->p_stat = SSTOP; 578 p2->p_waited = 0; 579 p1->p_nstopchild++; 580 l2->l_stat = LSSTOP; 581 KASSERT(l2->l_wchan == NULL); 582 lwp_unlock_to(l2, spc->spc_lwplock); 583 } else { 584 p2->p_nrlwps = 1; 585 p2->p_stat = SACTIVE; 586 l2->l_stat = LSRUN; 587 sched_enqueue(l2, false); 588 lwp_unlock(l2); 589 } 590 591 /* 592 * Return child pid to parent process, 593 * marking us as parent via retval[1]. 594 */ 595 if (retval != NULL) { 596 retval[0] = p2->p_pid; 597 retval[1] = 0; 598 } 599 600 mutex_exit(p2->p_lock); 601 602 /* 603 * Let the parent know that we are tracing its child. 604 */ 605 if (tracefork(p1, flags) || tracevfork(p1, flags)) { 606 mutex_enter(p1->p_lock); 607 eventswitch(TRAP_CHLD); 608 mutex_enter(proc_lock); 609 } 610 611 /* 612 * Preserve synchronization semantics of vfork. If waiting for 613 * child to exec or exit, sleep until it clears LP_VFORKWAIT. 614 */ 615 while (p2->p_lflag & PL_PPWAIT) // XXX: p2 can go invalid 616 cv_wait(&p1->p_waitcv, proc_lock); 617 618 /* 619 * Let the parent know that we are tracing its child. 620 */ 621 if (tracevforkdone(p1, flags)) { 622 mutex_enter(p1->p_lock); 623 p1->p_vfpid_done = retval[0]; 624 eventswitch(TRAP_CHLD); 625 } else 626 mutex_exit(proc_lock); 627 628 return 0; 629 } 630 631 /* 632 * MI code executed in each newly spawned process before returning to userland. 633 */ 634 void 635 child_return(void *arg) 636 { 637 struct lwp *l = arg; 638 struct proc *p = l->l_proc; 639 640 if (p->p_slflag & PSL_TRACED) { 641 /* Paranoid check */ 642 mutex_enter(proc_lock); 643 if (!(p->p_slflag & PSL_TRACED)) { 644 mutex_exit(proc_lock); 645 goto my_tracer_is_gone; 646 } 647 648 mutex_enter(p->p_lock); 649 eventswitch(TRAP_CHLD); 650 } 651 652 my_tracer_is_gone: 653 md_child_return(l); 654 655 /* 656 * Return SYS_fork for all fork types, including vfork(2) and clone(2). 657 * 658 * This approach simplifies the code and avoids extra locking. 659 */ 660 ktrsysret(SYS_fork, 0, 0); 661 } 662