Add sc_newstate_arg to ural_softc, so ieee80211_new_state() 'arg' can be
[dragonfly.git] / sys / kern / kern_fork.c
CommitLineData
984263bc
MD
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
5bc7cd8d 39 * $FreeBSD: src/sys/kern/kern_fork.c,v 1.72.2.14 2003/06/26 04:15:10 silby Exp $
13d13d89 40 * $DragonFly: src/sys/kern/kern_fork.c,v 1.65 2007/02/26 21:41:08 corecode Exp $
984263bc
MD
41 */
42
43#include "opt_ktrace.h"
44
45#include <sys/param.h>
46#include <sys/systm.h>
47#include <sys/sysproto.h>
48#include <sys/filedesc.h>
49#include <sys/kernel.h>
50#include <sys/sysctl.h>
51#include <sys/malloc.h>
52#include <sys/proc.h>
53#include <sys/resourcevar.h>
54#include <sys/vnode.h>
55#include <sys/acct.h>
56#include <sys/ktrace.h>
dfc1fc13
EN
57#include <sys/unistd.h>
58#include <sys/jail.h>
59#include <sys/caps.h>
984263bc
MD
60
61#include <vm/vm.h>
62#include <sys/lock.h>
63#include <vm/pmap.h>
64#include <vm/vm_map.h>
65#include <vm/vm_extern.h>
66#include <vm/vm_zone.h>
67
68#include <sys/vmmeter.h>
e43a034f 69#include <sys/thread2.h>
b1b4e5a6 70#include <sys/signal2.h>
984263bc
MD
71
72static MALLOC_DEFINE(M_ATFORK, "atfork", "atfork callback");
73
74/*
75 * These are the stuctures used to create a callout list for things to do
76 * when forking a process
77 */
78struct forklist {
79 forklist_fn function;
80 TAILQ_ENTRY(forklist) next;
81};
82
83TAILQ_HEAD(forklist_head, forklist);
84static struct forklist_head fork_list = TAILQ_HEAD_INITIALIZER(fork_list);
85
13d13d89
SS
86static struct lwp *lwp_fork(struct lwp *, struct proc *, int flags);
87
984263bc
MD
88int forksleep; /* Place for fork1() to sleep on. */
89
90/* ARGSUSED */
91int
753fd850 92sys_fork(struct fork_args *uap)
984263bc 93{
553ea3c8 94 struct lwp *lp = curthread->td_lwp;
984263bc 95 struct proc *p2;
41c20dac 96 int error;
984263bc 97
167e6ecb 98 error = fork1(lp, RFFDG | RFPROC | RFPGLOCK, &p2);
984263bc 99 if (error == 0) {
553ea3c8 100 start_forked_proc(lp, p2);
c7114eea
MD
101 uap->sysmsg_fds[0] = p2->p_pid;
102 uap->sysmsg_fds[1] = 0;
984263bc
MD
103 }
104 return error;
105}
106
107/* ARGSUSED */
108int
753fd850 109sys_vfork(struct vfork_args *uap)
984263bc 110{
553ea3c8 111 struct lwp *lp = curthread->td_lwp;
984263bc 112 struct proc *p2;
41c20dac 113 int error;
984263bc 114
167e6ecb 115 error = fork1(lp, RFFDG | RFPROC | RFPPWAIT | RFMEM | RFPGLOCK, &p2);
984263bc 116 if (error == 0) {
553ea3c8 117 start_forked_proc(lp, p2);
c7114eea
MD
118 uap->sysmsg_fds[0] = p2->p_pid;
119 uap->sysmsg_fds[1] = 0;
984263bc
MD
120 }
121 return error;
122}
123
f61c1ff1
MD
124/*
125 * Handle rforks. An rfork may (1) operate on the current process without
126 * creating a new, (2) create a new process that shared the current process's
127 * vmspace, signals, and/or descriptors, or (3) create a new process that does
128 * not share these things (normal fork).
129 *
130 * Note that we only call start_forked_proc() if a new process is actually
131 * created.
132 *
133 * rfork { int flags }
134 */
984263bc 135int
753fd850 136sys_rfork(struct rfork_args *uap)
984263bc 137{
553ea3c8 138 struct lwp *lp = curthread->td_lwp;
984263bc 139 struct proc *p2;
41c20dac 140 int error;
984263bc 141
6654fbcb
MD
142 if ((uap->flags & RFKERNELONLY) != 0)
143 return (EINVAL);
144
167e6ecb 145 error = fork1(lp, uap->flags | RFPGLOCK, &p2);
984263bc 146 if (error == 0) {
f61c1ff1 147 if (p2)
553ea3c8 148 start_forked_proc(lp, p2);
c7114eea
MD
149 uap->sysmsg_fds[0] = p2 ? p2->p_pid : 0;
150 uap->sysmsg_fds[1] = 0;
984263bc
MD
151 }
152 return error;
153}
154
155
156int nprocs = 1; /* process 0 */
984263bc
MD
157
158int
553ea3c8 159fork1(struct lwp *lp1, int flags, struct proc **procp)
984263bc 160{
553ea3c8 161 struct proc *p1 = lp1->lwp_proc;
984263bc 162 struct proc *p2, *pptr;
167e6ecb 163 struct pgrp *pgrp;
984263bc 164 uid_t uid;
167e6ecb 165 int ok, error;
51e64ff2 166 static int curfail = 0;
5bc7cd8d 167 static struct timeval lastfail;
984263bc
MD
168 struct forklist *ep;
169 struct filedesc_to_leader *fdtol;
170
171 if ((flags & (RFFDG|RFCFDG)) == (RFFDG|RFCFDG))
172 return (EINVAL);
173
174 /*
175 * Here we don't create a new process, but we divorce
176 * certain parts of a process from itself.
177 */
178 if ((flags & RFPROC) == 0) {
179
13d13d89
SS
180 /*
181 * This kind of stunt does not work anymore if
182 * there are native threads (lwps) running
183 */
184 if (p1->p_nthreads != 1)
185 return (EINVAL);
186
187 vm_fork(p1, 0, flags);
984263bc
MD
188
189 /*
190 * Close all file descriptors.
191 */
192 if (flags & RFCFDG) {
193 struct filedesc *fdtmp;
194 fdtmp = fdinit(p1);
195 fdfree(p1);
196 p1->p_fd = fdtmp;
197 }
198
199 /*
200 * Unshare file descriptors (from parent.)
201 */
202 if (flags & RFFDG) {
203 if (p1->p_fd->fd_refcnt > 1) {
204 struct filedesc *newfd;
205 newfd = fdcopy(p1);
206 fdfree(p1);
207 p1->p_fd = newfd;
208 }
209 }
210 *procp = NULL;
211 return (0);
212 }
213
167e6ecb
MD
214 /*
215 * Interlock against process group signal delivery. If signals
216 * are pending after the interlock is obtained we have to restart
217 * the system call to process the signals. If we don't the child
218 * can miss a pgsignal (such as ^C) sent during the fork.
219 *
220 * We can't use CURSIG() here because it will process any STOPs
221 * and cause the process group lock to be held indefinitely. If
222 * a STOP occurs, the fork will be restarted after the CONT.
223 */
224 error = 0;
225 pgrp = NULL;
226 if ((flags & RFPGLOCK) && (pgrp = p1->p_pgrp) != NULL) {
227 lockmgr(&pgrp->pg_lock, LK_SHARED);
08f2f1bb 228 if (CURSIGNB(lp1)) {
167e6ecb
MD
229 error = ERESTART;
230 goto done;
231 }
232 }
233
984263bc
MD
234 /*
235 * Although process entries are dynamically created, we still keep
236 * a global limit on the maximum number we will create. Don't allow
237 * a nonprivileged user to use the last ten processes; don't let root
238 * exceed the limit. The variable nprocs is the current number of
239 * processes, maxproc is the limit.
240 */
41c20dac 241 uid = p1->p_ucred->cr_ruid;
984263bc 242 if ((nprocs >= maxproc - 10 && uid != 0) || nprocs >= maxproc) {
5bc7cd8d 243 if (ppsratecheck(&lastfail, &curfail, 1))
6ea70f76 244 kprintf("maxproc limit exceeded by uid %d, please "
5bc7cd8d 245 "see tuning(7) and login.conf(5).\n", uid);
377d4740 246 tsleep(&forksleep, 0, "fork", hz / 2);
167e6ecb
MD
247 error = EAGAIN;
248 goto done;
984263bc
MD
249 }
250 /*
251 * Increment the nprocs resource before blocking can occur. There
252 * are hard-limits as to the number of processes that can run.
253 */
254 nprocs++;
255
256 /*
257 * Increment the count of procs running with this uid. Don't allow
258 * a nonprivileged user to exceed their current limit.
259 */
41c20dac 260 ok = chgproccnt(p1->p_ucred->cr_ruidinfo, 1,
984263bc
MD
261 (uid != 0) ? p1->p_rlimit[RLIMIT_NPROC].rlim_cur : 0);
262 if (!ok) {
263 /*
264 * Back out the process count
265 */
266 nprocs--;
5bc7cd8d 267 if (ppsratecheck(&lastfail, &curfail, 1))
6ea70f76 268 kprintf("maxproc limit exceeded by uid %d, please "
5bc7cd8d 269 "see tuning(7) and login.conf(5).\n", uid);
377d4740 270 tsleep(&forksleep, 0, "fork", hz / 2);
167e6ecb
MD
271 error = EAGAIN;
272 goto done;
984263bc
MD
273 }
274
275 /* Allocate new proc. */
51e64ff2 276 p2 = zalloc(proc_zone);
984263bc
MD
277
278 /*
ef09c3ed 279 * Setup linkage for kernel based threading XXX lwp
984263bc 280 */
cb74210d 281 if (flags & RFTHREAD) {
51e64ff2
MD
282 p2->p_peers = p1->p_peers;
283 p1->p_peers = p2;
284 p2->p_leader = p1->p_leader;
984263bc 285 } else {
51e64ff2
MD
286 p2->p_peers = NULL;
287 p2->p_leader = p2;
984263bc
MD
288 }
289
51e64ff2
MD
290 p2->p_wakeup = 0;
291 p2->p_vmspace = NULL;
292 p2->p_numposixlocks = 0;
293 p2->p_emuldata = NULL;
51e64ff2 294 LIST_INIT(&p2->p_lwps);
ef09c3ed 295
13d13d89 296 p2->p_nthreads = 0;
51e64ff2
MD
297 p2->p_nstopped = 0;
298 p2->p_lasttid = 0;
984263bc
MD
299
300 /*
51e64ff2
MD
301 * Setting the state to SIDL protects the partially initialized
302 * process once it starts getting hooked into the rest of the system.
984263bc 303 */
51e64ff2
MD
304 p2->p_stat = SIDL;
305 proc_add_allproc(p2);
984263bc
MD
306
307 /*
308 * Make a proc table entry for the new process.
309 * Start by zeroing the section of proc that is zero-initialized,
310 * then copy the section that is copied directly from the parent.
311 */
312 bzero(&p2->p_startzero,
313 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero));
314 bcopy(&p1->p_startcopy, &p2->p_startcopy,
315 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
316
317 p2->p_aioinfo = NULL;
318
319 /*
320 * Duplicate sub-structures as needed.
321 * Increase reference counts on shared objects.
3a72e617 322 * p_lock is in the copy area and must be cleared.
984263bc 323 */
344ad853 324 p2->p_flag = 0;
3a72e617 325 p2->p_lock = 0;
13d13d89 326
984263bc
MD
327 if (p1->p_flag & P_PROFIL)
328 startprofclock(p2);
dadab5e9 329 p2->p_ucred = crhold(p1->p_ucred);
984263bc 330
b40e316c 331 if (jailed(p2->p_ucred))
984263bc 332 p2->p_flag |= P_JAILED;
984263bc
MD
333
334 if (p2->p_args)
335 p2->p_args->ar_ref++;
336
13d13d89
SS
337 p2->p_usched = p1->p_usched;
338
984263bc 339 if (flags & RFSIGSHARE) {
b1b4e5a6
SS
340 p2->p_sigacts = p1->p_sigacts;
341 p2->p_sigacts->ps_refcnt++;
984263bc 342 } else {
b1b4e5a6 343 p2->p_sigacts = (struct sigacts *)kmalloc(sizeof(*p2->p_sigacts),
984263bc 344 M_SUBPROC, M_WAITOK);
b1b4e5a6
SS
345 bcopy(p1->p_sigacts, p2->p_sigacts, sizeof(*p2->p_sigacts));
346 p2->p_sigacts->ps_refcnt = 1;
984263bc
MD
347 }
348 if (flags & RFLINUXTHPN)
349 p2->p_sigparent = SIGUSR1;
350 else
351 p2->p_sigparent = SIGCHLD;
352
353 /* bump references to the text vnode (for procfs) */
354 p2->p_textvp = p1->p_textvp;
355 if (p2->p_textvp)
597aea93 356 vref(p2->p_textvp);
984263bc 357
0daa37a5
MD
358 /*
359 * Handle file descriptors
360 */
984263bc
MD
361 if (flags & RFCFDG) {
362 p2->p_fd = fdinit(p1);
363 fdtol = NULL;
364 } else if (flags & RFFDG) {
365 p2->p_fd = fdcopy(p1);
366 fdtol = NULL;
367 } else {
368 p2->p_fd = fdshare(p1);
369 if (p1->p_fdtol == NULL)
370 p1->p_fdtol =
371 filedesc_to_leader_alloc(NULL,
372 p1->p_leader);
373 if ((flags & RFTHREAD) != 0) {
374 /*
375 * Shared file descriptor table and
376 * shared process leaders.
377 */
378 fdtol = p1->p_fdtol;
379 fdtol->fdl_refcount++;
380 } else {
381 /*
382 * Shared file descriptor table, and
383 * different process leaders
384 */
98a7f915 385 fdtol = filedesc_to_leader_alloc(p1->p_fdtol, p2);
984263bc
MD
386 }
387 }
388 p2->p_fdtol = fdtol;
c0b8a06d 389 p2->p_limit = plimit_fork(p1->p_limit);
984263bc
MD
390
391 /*
392 * Preserve some more flags in subprocess. P_PROFIL has already
393 * been preserved.
394 */
08f2f1bb 395 p2->p_flag |= p1->p_flag & P_SUGID;
984263bc
MD
396 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
397 p2->p_flag |= P_CONTROLT;
398 if (flags & RFPPWAIT)
399 p2->p_flag |= P_PPWAIT;
400
0daa37a5
MD
401 /*
402 * Inherit the virtual kernel structure (allows a virtual kernel
403 * to fork to simulate multiple cpus).
404 */
4a22e893
MD
405 p2->p_vkernel = NULL;
406 if (p1->p_vkernel)
407 vkernel_inherit(p1, p2);
0daa37a5 408
5fd012e0
MD
409 /*
410 * Once we are on a pglist we may receive signals. XXX we might
411 * race a ^C being sent to the process group by not receiving it
412 * at all prior to this line.
413 */
984263bc
MD
414 LIST_INSERT_AFTER(p1, p2, p_pglist);
415
416 /*
417 * Attach the new process to its parent.
418 *
419 * If RFNOWAIT is set, the newly created process becomes a child
420 * of init. This effectively disassociates the child from the
421 * parent.
422 */
423 if (flags & RFNOWAIT)
424 pptr = initproc;
425 else
426 pptr = p1;
427 p2->p_pptr = pptr;
428 LIST_INSERT_HEAD(&pptr->p_children, p2, p_sibling);
429 LIST_INIT(&p2->p_children);
98a7f915 430 varsymset_init(&p2->p_varsymset, &p1->p_varsymset);
8fbf9130 431 callout_init(&p2->p_ithandle);
984263bc
MD
432
433#ifdef KTRACE
434 /*
435 * Copy traceflag and tracefile if enabled. If not inherited,
436 * these were zeroed above but we still could have a trace race
29f58392 437 * so make sure p2's p_tracenode is NULL.
984263bc 438 */
29f58392 439 if ((p1->p_traceflag & KTRFAC_INHERIT) && p2->p_tracenode == NULL) {
984263bc 440 p2->p_traceflag = p1->p_traceflag;
29f58392 441 p2->p_tracenode = ktrinherit(p1->p_tracenode);
984263bc
MD
442 }
443#endif
444
984263bc
MD
445 /*
446 * This begins the section where we must prevent the parent
447 * from being swapped.
13d13d89
SS
448 *
449 * Gets PRELE'd in the caller in start_forked_proc().
984263bc
MD
450 */
451 PHOLD(p1);
452
13d13d89
SS
453 vm_fork(p1, p2, flags);
454
984263bc 455 /*
13d13d89
SS
456 * Create the first lwp associated with the new proc.
457 * It will return via a different execution path later, directly
458 * into userland, after it was put on the runq by
459 * start_forked_proc().
984263bc 460 */
13d13d89 461 lwp_fork(lp1, p2, flags);
984263bc
MD
462
463 if (flags == (RFFDG | RFPROC)) {
12e4aaff
MD
464 mycpu->gd_cnt.v_forks++;
465 mycpu->gd_cnt.v_forkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 466 } else if (flags == (RFFDG | RFPROC | RFPPWAIT | RFMEM)) {
12e4aaff
MD
467 mycpu->gd_cnt.v_vforks++;
468 mycpu->gd_cnt.v_vforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 469 } else if (p1 == &proc0) {
12e4aaff
MD
470 mycpu->gd_cnt.v_kthreads++;
471 mycpu->gd_cnt.v_kthreadpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 472 } else {
12e4aaff
MD
473 mycpu->gd_cnt.v_rforks++;
474 mycpu->gd_cnt.v_rforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc
MD
475 }
476
477 /*
478 * Both processes are set up, now check if any loadable modules want
479 * to adjust anything.
480 * What if they have an error? XXX
481 */
482 TAILQ_FOREACH(ep, &fork_list, next) {
483 (*ep->function)(p1, p2, flags);
484 }
485
486 /*
a77ac49d
MD
487 * Set the start time. Note that the process is not runnable. The
488 * caller is responsible for making it runnable.
984263bc 489 */
d9fa5f67 490 microtime(&p2->p_start);
984263bc 491 p2->p_acflag = AFORK;
984263bc
MD
492
493 /*
494 * tell any interested parties about the new process
495 */
496 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid);
497
984263bc
MD
498 /*
499 * Return child proc pointer to parent.
500 */
501 *procp = p2;
167e6ecb
MD
502done:
503 if (pgrp)
504 lockmgr(&pgrp->pg_lock, LK_RELEASE);
505 return (error);
984263bc
MD
506}
507
13d13d89
SS
508static struct lwp *
509lwp_fork(struct lwp *origlp, struct proc *destproc, int flags)
510{
511 struct lwp *lp;
512 struct thread *td;
513 lwpid_t tid;
514
515 /*
516 * We need to prevent wrap-around collisions.
517 * Until we have a nice tid allocator, we need to
518 * start searching for free tids once we wrap around.
519 *
520 * XXX give me a nicer allocator
521 */
522 if (destproc->p_lasttid + 1 <= 0) {
523 tid = 0;
524restart:
525 FOREACH_LWP_IN_PROC(lp, destproc) {
526 if (lp->lwp_tid != tid)
527 continue;
528 /* tids match, search next. */
529 tid++;
530 /*
531 * Wait -- the whole tid space is depleted?
532 * Impossible.
533 */
534 if (tid <= 0)
535 panic("lwp_fork: All tids depleted?!");
536 goto restart;
537 }
538 /* When we come here, the tid is not occupied */
539 } else {
540 tid = destproc->p_lasttid++;
541 }
542
543 lp = zalloc(lwp_zone);
544 lp->lwp_proc = destproc;
545 lp->lwp_tid = tid;
546 LIST_INSERT_HEAD(&destproc->p_lwps, lp, lwp_list);
547 destproc->p_nthreads++;
548 lp->lwp_stat = LSRUN;
549 bzero(&lp->lwp_startzero,
550 (unsigned) ((caddr_t)&lp->lwp_endzero -
551 (caddr_t)&lp->lwp_startzero));
552 bcopy(&origlp->lwp_startcopy, &lp->lwp_startcopy,
553 (unsigned) ((caddr_t)&lp->lwp_endcopy -
554 (caddr_t)&lp->lwp_startcopy));
555 lp->lwp_lock = 0;
556 lp->lwp_flag |= origlp->lwp_flag & LWP_ALTSTACK;
557 /*
558 * Set cpbase to the last timeout that occured (not the upcoming
559 * timeout).
560 *
561 * A critical section is required since a timer IPI can update
562 * scheduler specific data.
563 */
564 crit_enter();
565 lp->lwp_cpbase = mycpu->gd_schedclock.time -
566 mycpu->gd_schedclock.periodic;
567 destproc->p_usched->heuristic_forking(origlp, lp);
568 crit_exit();
569
570 td = lwkt_alloc_thread(NULL, LWKT_THREAD_STACK, -1, 0);
571 lp->lwp_thread = td;
572 td->td_proc = destproc;
573 td->td_lwp = lp;
574 td->td_switch = cpu_heavy_switch;
575#ifdef SMP
576 KKASSERT(td->td_mpcount == 1);
577#endif
578 lwkt_setpri(td, TDPRI_KERN_USER);
579 lwkt_set_comm(td, "%s", destproc->p_comm);
580
581 /*
582 * cpu_fork will copy and update the pcb, set up the kernel stack,
583 * and make the child ready to run.
584 */
585 cpu_fork(origlp, lp, flags);
586 caps_fork(origlp->lwp_thread, lp->lwp_thread);
587
588 return (lp);
589}
590
984263bc
MD
591/*
592 * The next two functionms are general routines to handle adding/deleting
593 * items on the fork callout list.
594 *
595 * at_fork():
596 * Take the arguments given and put them onto the fork callout list,
597 * However first make sure that it's not already there.
598 * Returns 0 on success or a standard error number.
599 */
984263bc 600int
303c76d5 601at_fork(forklist_fn function)
984263bc
MD
602{
603 struct forklist *ep;
604
605#ifdef INVARIANTS
606 /* let the programmer know if he's been stupid */
303c76d5 607 if (rm_at_fork(function)) {
6ea70f76 608 kprintf("WARNING: fork callout entry (%p) already present\n",
984263bc 609 function);
303c76d5 610 }
984263bc 611#endif
efda3bd0 612 ep = kmalloc(sizeof(*ep), M_ATFORK, M_WAITOK|M_ZERO);
984263bc
MD
613 ep->function = function;
614 TAILQ_INSERT_TAIL(&fork_list, ep, next);
615 return (0);
616}
617
618/*
619 * Scan the exit callout list for the given item and remove it..
620 * Returns the number of items removed (0 or 1)
621 */
984263bc 622int
303c76d5 623rm_at_fork(forklist_fn function)
984263bc
MD
624{
625 struct forklist *ep;
626
627 TAILQ_FOREACH(ep, &fork_list, next) {
628 if (ep->function == function) {
629 TAILQ_REMOVE(&fork_list, ep, next);
efda3bd0 630 kfree(ep, M_ATFORK);
984263bc
MD
631 return(1);
632 }
633 }
634 return (0);
635}
7d0bac62
MD
636
637/*
638 * Add a forked process to the run queue after any remaining setup, such
639 * as setting the fork handler, has been completed.
640 */
7d0bac62 641void
553ea3c8 642start_forked_proc(struct lwp *lp1, struct proc *p2)
7d0bac62 643{
08f2f1bb 644 struct lwp *lp2 = ONLY_LWP_IN_PROC(p2);
553ea3c8 645
7d0bac62 646 /*
26a0694b
MD
647 * Move from SIDL to RUN queue, and activate the process's thread.
648 * Activation of the thread effectively makes the process "a"
649 * current process, so we do not setrunqueue().
8ec60c3f
MD
650 *
651 * YYY setrunqueue works here but we should clean up the trampoline
652 * code so we just schedule the LWKT thread and let the trampoline
653 * deal with the userland scheduler on return to userland.
7d0bac62 654 */
553ea3c8 655 KASSERT(p2->p_stat == SIDL,
7d0bac62 656 ("cannot start forked process, bad status: %p", p2));
553ea3c8 657 p2->p_usched->resetpriority(lp2);
e43a034f 658 crit_enter();
164b8401
SS
659 p2->p_stat = SACTIVE;
660 lp2->lwp_stat = LSRUN;
553ea3c8 661 p2->p_usched->setrunqueue(lp2);
e43a034f 662 crit_exit();
7d0bac62
MD
663
664 /*
665 * Now can be swapped.
666 */
553ea3c8 667 PRELE(lp1->lwp_proc);
7d0bac62
MD
668
669 /*
670 * Preserve synchronization semantics of vfork. If waiting for
671 * child to exec or exit, set P_PPWAIT on child, and sleep on our
672 * proc (in case of exit).
673 */
674 while (p2->p_flag & P_PPWAIT)
553ea3c8 675 tsleep(lp1->lwp_proc, 0, "ppwait", 0);
7d0bac62 676}