1 /* $NetBSD: kern_fork.c,v 1.230 2023/02/25 08:22:00 skrll 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 1.230 2023/02/25 08:22:00 skrll 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 miliseconds to userland via sysctl.
113 */
114 int forkfsleep = 0;
115
116 int
sys_fork(struct lwp * l,const void * v,register_t * retval)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
sys_vfork(struct lwp * l,const void * v,register_t * retval)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
sys___vfork14(struct lwp * l,const void * v,register_t * retval)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
sys___clone(struct lwp * l,const struct sys___clone_args * uap,register_t * retval)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 * Note that the Linux API does not provide a portable way of
192 * specifying the stack area; the caller must know if the stack
193 * grows up or down. So, we pass a stack size of 0, so that the
194 * code that makes this adjustment is a noop.
195 */
196 return fork1(l, flags, sig, SCARG(uap, stack), 0,
197 NULL, NULL, retval);
198 }
199
200 /*
201 * Print the 'table full' message once per 10 seconds.
202 */
203 static struct timeval fork_tfmrate = { 10, 0 };
204
205 /*
206 * Check if a process is traced and shall inform about FORK events.
207 */
208 static inline bool
tracefork(struct proc * p,int flags)209 tracefork(struct proc *p, int flags)
210 {
211
212 return (p->p_slflag & (PSL_TRACEFORK|PSL_TRACED)) ==
213 (PSL_TRACEFORK|PSL_TRACED) && (flags & FORK_PPWAIT) == 0;
214 }
215
216 /*
217 * Check if a process is traced and shall inform about VFORK events.
218 */
219 static inline bool
tracevfork(struct proc * p,int flags)220 tracevfork(struct proc *p, int flags)
221 {
222
223 return (p->p_slflag & (PSL_TRACEVFORK|PSL_TRACED)) ==
224 (PSL_TRACEVFORK|PSL_TRACED) && (flags & FORK_PPWAIT) != 0;
225 }
226
227 /*
228 * Check if a process is traced and shall inform about VFORK_DONE events.
229 */
230 static inline bool
tracevforkdone(struct proc * p,int flags)231 tracevforkdone(struct proc *p, int flags)
232 {
233
234 return (p->p_slflag & (PSL_TRACEVFORK_DONE|PSL_TRACED)) ==
235 (PSL_TRACEVFORK_DONE|PSL_TRACED) && (flags & FORK_PPWAIT);
236 }
237
238 /*
239 * General fork call. Note that another LWP in the process may call exec()
240 * or exit() while we are forking. It's safe to continue here, because
241 * neither operation will complete until all LWPs have exited the process.
242 */
243 int
fork1(struct lwp * l1,int flags,int exitsig,void * stack,size_t stacksize,void (* func)(void *),void * arg,register_t * retval)244 fork1(struct lwp *l1, int flags, int exitsig, void *stack, size_t stacksize,
245 void (*func)(void *), void *arg, register_t *retval)
246 {
247 struct proc *p1, *p2, *parent;
248 struct plimit *p1_lim;
249 uid_t uid;
250 struct lwp *l2;
251 int count;
252 vaddr_t uaddr;
253 int tnprocs;
254 int error = 0;
255
256 p1 = l1->l_proc;
257 uid = kauth_cred_getuid(l1->l_cred);
258 tnprocs = atomic_inc_uint_nv(&nprocs);
259
260 /*
261 * Although process entries are dynamically created, we still keep
262 * a global limit on the maximum number we will create.
263 */
264 if (__predict_false(tnprocs >= maxproc))
265 error = -1;
266 else
267 error = kauth_authorize_process(l1->l_cred,
268 KAUTH_PROCESS_FORK, p1, KAUTH_ARG(tnprocs), NULL, NULL);
269
270 if (error) {
271 static struct timeval lasttfm;
272 atomic_dec_uint(&nprocs);
273 if (ratecheck(&lasttfm, &fork_tfmrate))
274 tablefull("proc", "increase kern.maxproc or NPROC");
275 if (forkfsleep)
276 kpause("forkmx", false, forkfsleep, NULL);
277 return EAGAIN;
278 }
279
280 /*
281 * Enforce limits.
282 */
283 count = chgproccnt(uid, 1);
284 if (__predict_false(count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur)) {
285 if (kauth_authorize_process(l1->l_cred, KAUTH_PROCESS_RLIMIT,
286 p1, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_BYPASS),
287 &p1->p_rlimit[RLIMIT_NPROC], KAUTH_ARG(RLIMIT_NPROC)) != 0) {
288 (void)chgproccnt(uid, -1);
289 atomic_dec_uint(&nprocs);
290 if (forkfsleep)
291 kpause("forkulim", false, forkfsleep, NULL);
292 return EAGAIN;
293 }
294 }
295
296 /*
297 * Allocate virtual address space for the U-area now, while it
298 * is still easy to abort the fork operation if we're out of
299 * kernel virtual address space.
300 */
301 uaddr = uvm_uarea_alloc();
302 if (__predict_false(uaddr == 0)) {
303 (void)chgproccnt(uid, -1);
304 atomic_dec_uint(&nprocs);
305 return ENOMEM;
306 }
307
308 /* Allocate new proc. */
309 p2 = proc_alloc();
310 if (p2 == NULL) {
311 /* We were unable to allocate a process ID. */
312 uvm_uarea_free(uaddr);
313 mutex_enter(p1->p_lock);
314 uid = kauth_cred_getuid(p1->p_cred);
315 (void)chgproccnt(uid, -1);
316 mutex_exit(p1->p_lock);
317 atomic_dec_uint(&nprocs);
318 return EAGAIN;
319 }
320
321 /*
322 * We are now committed to the fork. From here on, we may
323 * block on resources, but resource allocation may NOT fail.
324 */
325
326 /*
327 * Make a proc table entry for the new process.
328 * Start by zeroing the section of proc that is zero-initialized,
329 * then copy the section that is copied directly from the parent.
330 */
331 memset(&p2->p_startzero, 0,
332 (unsigned) ((char *)&p2->p_endzero - (char *)&p2->p_startzero));
333 memcpy(&p2->p_startcopy, &p1->p_startcopy,
334 (unsigned) ((char *)&p2->p_endcopy - (char *)&p2->p_startcopy));
335
336 TAILQ_INIT(&p2->p_sigpend.sp_info);
337
338 LIST_INIT(&p2->p_lwps);
339 LIST_INIT(&p2->p_sigwaiters);
340
341 /*
342 * Duplicate sub-structures as needed.
343 * Increase reference counts on shared objects.
344 * Inherit flags we want to keep. The flags related to SIGCHLD
345 * handling are important in order to keep a consistent behaviour
346 * for the child after the fork. If we are a 32-bit process, the
347 * child will be too.
348 */
349 p2->p_flag =
350 p1->p_flag & (PK_SUGID | PK_NOCLDWAIT | PK_CLDSIGIGN | PK_32);
351 p2->p_emul = p1->p_emul;
352 p2->p_execsw = p1->p_execsw;
353
354 if (flags & FORK_SYSTEM) {
355 /*
356 * Mark it as a system process. Set P_NOCLDWAIT so that
357 * children are reparented to init(8) when they exit.
358 * init(8) can easily wait them out for us.
359 */
360 p2->p_flag |= (PK_SYSTEM | PK_NOCLDWAIT);
361 }
362
363 mutex_init(&p2->p_stmutex, MUTEX_DEFAULT, IPL_HIGH);
364 mutex_init(&p2->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
365 rw_init(&p2->p_reflock);
366 cv_init(&p2->p_waitcv, "wait");
367 cv_init(&p2->p_lwpcv, "lwpwait");
368
369 /*
370 * Share a lock between the processes if they are to share signal
371 * state: we must synchronize access to it.
372 */
373 if (flags & FORK_SHARESIGS) {
374 p2->p_lock = p1->p_lock;
375 mutex_obj_hold(p1->p_lock);
376 } else
377 p2->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
378
379 kauth_proc_fork(p1, p2);
380
381 p2->p_raslist = NULL;
382 #if defined(__HAVE_RAS)
383 ras_fork(p1, p2);
384 #endif
385
386 /* bump references to the text vnode (for procfs) */
387 p2->p_textvp = p1->p_textvp;
388 if (p2->p_textvp)
389 vref(p2->p_textvp);
390 if (p1->p_path)
391 p2->p_path = kmem_strdupsize(p1->p_path, NULL, KM_SLEEP);
392 else
393 p2->p_path = NULL;
394
395 if (flags & FORK_SHAREFILES)
396 fd_share(p2);
397 else if (flags & FORK_CLEANFILES)
398 p2->p_fd = fd_init(NULL);
399 else
400 p2->p_fd = fd_copy();
401
402 /* XXX racy */
403 p2->p_mqueue_cnt = p1->p_mqueue_cnt;
404
405 if (flags & FORK_SHARECWD)
406 cwdshare(p2);
407 else
408 p2->p_cwdi = cwdinit();
409
410 /*
411 * Note: p_limit (rlimit stuff) is copy-on-write, so normally
412 * we just need increase pl_refcnt.
413 */
414 p1_lim = p1->p_limit;
415 if (!p1_lim->pl_writeable) {
416 lim_addref(p1_lim);
417 p2->p_limit = p1_lim;
418 } else {
419 p2->p_limit = lim_copy(p1_lim);
420 }
421
422 if (flags & FORK_PPWAIT) {
423 /* Mark ourselves as waiting for a child. */
424 p2->p_lflag = PL_PPWAIT;
425 l1->l_vforkwaiting = true;
426 p2->p_vforklwp = l1;
427 } else {
428 p2->p_lflag = 0;
429 l1->l_vforkwaiting = false;
430 }
431 p2->p_sflag = 0;
432 p2->p_slflag = 0;
433 parent = (flags & FORK_NOWAIT) ? initproc : p1;
434 p2->p_pptr = parent;
435 p2->p_ppid = parent->p_pid;
436 LIST_INIT(&p2->p_children);
437
438 p2->p_aio = NULL;
439
440 #ifdef KTRACE
441 /*
442 * Copy traceflag and tracefile if enabled.
443 * If not inherited, these were zeroed above.
444 */
445 if (p1->p_traceflag & KTRFAC_INHERIT) {
446 mutex_enter(&ktrace_lock);
447 p2->p_traceflag = p1->p_traceflag;
448 if ((p2->p_tracep = p1->p_tracep) != NULL)
449 ktradref(p2);
450 mutex_exit(&ktrace_lock);
451 }
452 #endif
453
454 /*
455 * Create signal actions for the child process.
456 */
457 p2->p_sigacts = sigactsinit(p1, flags & FORK_SHARESIGS);
458 mutex_enter(p1->p_lock);
459 p2->p_sflag |=
460 (p1->p_sflag & (PS_STOPFORK | PS_STOPEXEC | PS_NOCLDSTOP));
461 sched_proc_fork(p1, p2);
462 mutex_exit(p1->p_lock);
463
464 p2->p_stflag = p1->p_stflag;
465
466 /*
467 * p_stats.
468 * Copy parts of p_stats, and zero out the rest.
469 */
470 p2->p_stats = pstatscopy(p1->p_stats);
471
472 /*
473 * Set up the new process address space.
474 */
475 uvm_proc_fork(p1, p2, (flags & FORK_SHAREVM) ? true : false);
476
477 /*
478 * Finish creating the child process.
479 * It will return through a different path later.
480 */
481 lwp_create(l1, p2, uaddr, (flags & FORK_PPWAIT) ? LWP_VFORK : 0,
482 stack, stacksize, (func != NULL) ? func : child_return, arg, &l2,
483 l1->l_class, &l1->l_sigmask, &l1->l_sigstk);
484
485 /*
486 * Inherit l_private from the parent.
487 * Note that we cannot use lwp_setprivate() here since that
488 * also sets the CPU TLS register, which is incorrect if the
489 * process has changed that without letting the kernel know.
490 */
491 l2->l_private = l1->l_private;
492
493 /*
494 * If emulation has a process fork hook, call it now.
495 */
496 if (p2->p_emul->e_proc_fork)
497 (*p2->p_emul->e_proc_fork)(p2, l1, flags);
498
499 /*
500 * ...and finally, any other random fork hooks that subsystems
501 * might have registered.
502 */
503 doforkhooks(p2, p1);
504
505 SDT_PROBE(proc, kernel, , create, p2, p1, flags, 0, 0);
506
507 /*
508 * It's now safe for the scheduler and other processes to see the
509 * child process.
510 */
511 mutex_enter(&proc_lock);
512
513 if (p1->p_session->s_ttyvp != NULL && p1->p_lflag & PL_CONTROLT)
514 p2->p_lflag |= PL_CONTROLT;
515
516 LIST_INSERT_HEAD(&parent->p_children, p2, p_sibling);
517 p2->p_exitsig = exitsig; /* signal for parent on exit */
518
519 /*
520 * Trace fork(2) and vfork(2)-like events on demand in a debugger.
521 */
522 if (tracefork(p1, flags) || tracevfork(p1, flags)) {
523 proc_changeparent(p2, p1->p_pptr);
524 SET(p2->p_slflag, PSL_TRACEDCHILD);
525 }
526
527 p2->p_oppid = p1->p_pid; /* Remember the original parent id. */
528
529 LIST_INSERT_AFTER(p1, p2, p_pglist);
530 LIST_INSERT_HEAD(&allproc, p2, p_list);
531
532 p2->p_trace_enabled = trace_is_enabled(p2);
533 #ifdef __HAVE_SYSCALL_INTERN
534 (*p2->p_emul->e_syscall_intern)(p2);
535 #endif
536
537 /*
538 * Update stats now that we know the fork was successful.
539 */
540 KPREEMPT_DISABLE(l1);
541 CPU_COUNT(CPU_COUNT_FORKS, 1);
542 if (flags & FORK_PPWAIT)
543 CPU_COUNT(CPU_COUNT_FORKS_PPWAIT, 1);
544 if (flags & FORK_SHAREVM)
545 CPU_COUNT(CPU_COUNT_FORKS_SHAREVM, 1);
546 KPREEMPT_ENABLE(l1);
547
548 if (ktrpoint(KTR_EMUL))
549 p2->p_traceflag |= KTRFAC_TRC_EMUL;
550
551 /*
552 * Notify any interested parties about the new process.
553 */
554 if (!SLIST_EMPTY(&p1->p_klist)) {
555 mutex_exit(&proc_lock);
556 knote_proc_fork(p1, p2);
557 mutex_enter(&proc_lock);
558 }
559
560 /*
561 * Make child runnable, set start time, and add to run queue except
562 * if the parent requested the child to start in SSTOP state.
563 */
564 mutex_enter(p2->p_lock);
565
566 /*
567 * Start profiling.
568 */
569 if ((p2->p_stflag & PST_PROFIL) != 0) {
570 mutex_spin_enter(&p2->p_stmutex);
571 startprofclock(p2);
572 mutex_spin_exit(&p2->p_stmutex);
573 }
574
575 getmicrotime(&p2->p_stats->p_start);
576 p2->p_acflag = AFORK;
577 lwp_lock(l2);
578 KASSERT(p2->p_nrlwps == 1);
579 KASSERT(l2->l_stat == LSIDL);
580 if (p2->p_sflag & PS_STOPFORK) {
581 p2->p_nrlwps = 0;
582 p2->p_stat = SSTOP;
583 p2->p_waited = 0;
584 p1->p_nstopchild++;
585 l2->l_stat = LSSTOP;
586 KASSERT(l2->l_wchan == NULL);
587 lwp_unlock(l2);
588 } else {
589 p2->p_nrlwps = 1;
590 p2->p_stat = SACTIVE;
591 setrunnable(l2);
592 /* LWP now unlocked */
593 }
594
595 /*
596 * Return child pid to parent process,
597 * marking us as parent via retval[1].
598 */
599 if (retval != NULL) {
600 retval[0] = p2->p_pid;
601 retval[1] = 0;
602 }
603
604 mutex_exit(p2->p_lock);
605
606 /*
607 * Let the parent know that we are tracing its child.
608 */
609 if (tracefork(p1, flags) || tracevfork(p1, flags)) {
610 mutex_enter(p1->p_lock);
611 eventswitch(TRAP_CHLD,
612 tracefork(p1, flags) ? PTRACE_FORK : PTRACE_VFORK,
613 retval[0]);
614 mutex_enter(&proc_lock);
615 }
616
617 /*
618 * Preserve synchronization semantics of vfork. If waiting for
619 * child to exec or exit, sleep until it clears p_vforkwaiting.
620 */
621 while (l1->l_vforkwaiting)
622 cv_wait(&l1->l_waitcv, &proc_lock);
623
624 /*
625 * Let the parent know that we are tracing its child.
626 */
627 if (tracevforkdone(p1, flags)) {
628 mutex_enter(p1->p_lock);
629 eventswitch(TRAP_CHLD, PTRACE_VFORK_DONE, retval[0]);
630 } else
631 mutex_exit(&proc_lock);
632
633 return 0;
634 }
635
636 /*
637 * MI code executed in each newly spawned process before returning to userland.
638 */
639 void
child_return(void * arg)640 child_return(void *arg)
641 {
642 struct lwp *l = curlwp;
643 struct proc *p = l->l_proc;
644
645 if ((p->p_slflag & (PSL_TRACED|PSL_TRACEDCHILD)) ==
646 (PSL_TRACED|PSL_TRACEDCHILD)) {
647 eventswitchchild(p, TRAP_CHLD,
648 ISSET(p->p_lflag, PL_PPWAIT) ? PTRACE_VFORK : PTRACE_FORK);
649 }
650
651 md_child_return(l);
652
653 /*
654 * Return SYS_fork for all fork types, including vfork(2) and clone(2).
655 *
656 * This approach simplifies the code and avoids extra locking.
657 */
658 ktrsysret(SYS_fork, 0, 0);
659 }
660