1:1 Userland threading stage 2.3/4:
[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 $
addbf0b7 40 * $DragonFly: src/sys/kern/kern_fork.c,v 1.39 2005/10/06 20:01:29 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>
69#include <sys/user.h>
e43a034f 70#include <sys/thread2.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
984263bc
MD
86int forksleep; /* Place for fork1() to sleep on. */
87
88/* ARGSUSED */
89int
41c20dac 90fork(struct fork_args *uap)
984263bc 91{
41c20dac 92 struct proc *p = curproc;
984263bc 93 struct proc *p2;
41c20dac 94 int error;
984263bc
MD
95
96 error = fork1(p, RFFDG | RFPROC, &p2);
97 if (error == 0) {
7d0bac62 98 start_forked_proc(p, p2);
c7114eea
MD
99 uap->sysmsg_fds[0] = p2->p_pid;
100 uap->sysmsg_fds[1] = 0;
984263bc
MD
101 }
102 return error;
103}
104
105/* ARGSUSED */
106int
41c20dac 107vfork(struct vfork_args *uap)
984263bc 108{
41c20dac 109 struct proc *p = curproc;
984263bc 110 struct proc *p2;
41c20dac 111 int error;
984263bc
MD
112
113 error = fork1(p, RFFDG | RFPROC | RFPPWAIT | RFMEM, &p2);
114 if (error == 0) {
7d0bac62 115 start_forked_proc(p, p2);
c7114eea
MD
116 uap->sysmsg_fds[0] = p2->p_pid;
117 uap->sysmsg_fds[1] = 0;
984263bc
MD
118 }
119 return error;
120}
121
f61c1ff1
MD
122/*
123 * Handle rforks. An rfork may (1) operate on the current process without
124 * creating a new, (2) create a new process that shared the current process's
125 * vmspace, signals, and/or descriptors, or (3) create a new process that does
126 * not share these things (normal fork).
127 *
128 * Note that we only call start_forked_proc() if a new process is actually
129 * created.
130 *
131 * rfork { int flags }
132 */
984263bc 133int
41c20dac 134rfork(struct rfork_args *uap)
984263bc 135{
41c20dac 136 struct proc *p = curproc;
984263bc 137 struct proc *p2;
41c20dac 138 int error;
984263bc 139
6654fbcb
MD
140 if ((uap->flags & RFKERNELONLY) != 0)
141 return (EINVAL);
142
984263bc
MD
143 error = fork1(p, uap->flags, &p2);
144 if (error == 0) {
f61c1ff1
MD
145 if (p2)
146 start_forked_proc(p, p2);
c7114eea
MD
147 uap->sysmsg_fds[0] = p2 ? p2->p_pid : 0;
148 uap->sysmsg_fds[1] = 0;
984263bc
MD
149 }
150 return error;
151}
152
153
154int nprocs = 1; /* process 0 */
155static int nextpid = 0;
156
157/*
158 * Random component to nextpid generation. We mix in a random factor to make
159 * it a little harder to predict. We sanity check the modulus value to avoid
160 * doing it in critical paths. Don't let it be too small or we pointlessly
161 * waste randomness entropy, and don't let it be impossibly large. Using a
162 * modulus that is too big causes a LOT more process table scans and slows
163 * down fork processing as the pidchecked caching is defeated.
164 */
165static int randompid = 0;
166
167static int
168sysctl_kern_randompid(SYSCTL_HANDLER_ARGS)
169{
170 int error, pid;
171
172 pid = randompid;
173 error = sysctl_handle_int(oidp, &pid, 0, req);
174 if (error || !req->newptr)
175 return (error);
176 if (pid < 0 || pid > PID_MAX - 100) /* out of range */
177 pid = PID_MAX - 100;
178 else if (pid < 2) /* NOP */
179 pid = 0;
180 else if (pid < 100) /* Make it reasonable */
181 pid = 100;
182 randompid = pid;
183 return (error);
184}
185
186SYSCTL_PROC(_kern, OID_AUTO, randompid, CTLTYPE_INT|CTLFLAG_RW,
187 0, 0, sysctl_kern_randompid, "I", "Random PID modulus");
188
189int
303c76d5 190fork1(struct proc *p1, int flags, struct proc **procp)
984263bc
MD
191{
192 struct proc *p2, *pptr;
193 uid_t uid;
194 struct proc *newproc;
195 int ok;
5bc7cd8d
HP
196 static int curfail = 0, pidchecked = 0;
197 static struct timeval lastfail;
984263bc
MD
198 struct forklist *ep;
199 struct filedesc_to_leader *fdtol;
200
201 if ((flags & (RFFDG|RFCFDG)) == (RFFDG|RFCFDG))
202 return (EINVAL);
203
204 /*
205 * Here we don't create a new process, but we divorce
206 * certain parts of a process from itself.
207 */
208 if ((flags & RFPROC) == 0) {
209
210 vm_fork(p1, 0, flags);
211
212 /*
213 * Close all file descriptors.
214 */
215 if (flags & RFCFDG) {
216 struct filedesc *fdtmp;
217 fdtmp = fdinit(p1);
218 fdfree(p1);
219 p1->p_fd = fdtmp;
220 }
221
222 /*
223 * Unshare file descriptors (from parent.)
224 */
225 if (flags & RFFDG) {
226 if (p1->p_fd->fd_refcnt > 1) {
227 struct filedesc *newfd;
228 newfd = fdcopy(p1);
229 fdfree(p1);
230 p1->p_fd = newfd;
231 }
232 }
233 *procp = NULL;
234 return (0);
235 }
236
237 /*
238 * Although process entries are dynamically created, we still keep
239 * a global limit on the maximum number we will create. Don't allow
240 * a nonprivileged user to use the last ten processes; don't let root
241 * exceed the limit. The variable nprocs is the current number of
242 * processes, maxproc is the limit.
243 */
41c20dac 244 uid = p1->p_ucred->cr_ruid;
984263bc 245 if ((nprocs >= maxproc - 10 && uid != 0) || nprocs >= maxproc) {
5bc7cd8d
HP
246 if (ppsratecheck(&lastfail, &curfail, 1))
247 printf("maxproc limit exceeded by uid %d, please "
248 "see tuning(7) and login.conf(5).\n", uid);
377d4740 249 tsleep(&forksleep, 0, "fork", hz / 2);
984263bc
MD
250 return (EAGAIN);
251 }
252 /*
253 * Increment the nprocs resource before blocking can occur. There
254 * are hard-limits as to the number of processes that can run.
255 */
256 nprocs++;
257
258 /*
259 * Increment the count of procs running with this uid. Don't allow
260 * a nonprivileged user to exceed their current limit.
261 */
41c20dac 262 ok = chgproccnt(p1->p_ucred->cr_ruidinfo, 1,
984263bc
MD
263 (uid != 0) ? p1->p_rlimit[RLIMIT_NPROC].rlim_cur : 0);
264 if (!ok) {
265 /*
266 * Back out the process count
267 */
268 nprocs--;
5bc7cd8d
HP
269 if (ppsratecheck(&lastfail, &curfail, 1))
270 printf("maxproc limit exceeded by uid %d, please "
271 "see tuning(7) and login.conf(5).\n", uid);
377d4740 272 tsleep(&forksleep, 0, "fork", hz / 2);
984263bc
MD
273 return (EAGAIN);
274 }
275
276 /* Allocate new proc. */
277 newproc = zalloc(proc_zone);
278
279 /*
ef09c3ed 280 * Setup linkage for kernel based threading XXX lwp
984263bc 281 */
303c76d5 282 if ((flags & RFTHREAD) != 0) {
984263bc
MD
283 newproc->p_peers = p1->p_peers;
284 p1->p_peers = newproc;
285 newproc->p_leader = p1->p_leader;
286 } else {
287 newproc->p_peers = 0;
288 newproc->p_leader = newproc;
289 }
290
291 newproc->p_wakeup = 0;
984263bc 292 newproc->p_vmspace = NULL;
1fa2b4b4 293 TAILQ_INIT(&newproc->p_sysmsgq);
ef09c3ed
SS
294 LIST_INIT(&newproc->p_lwps);
295
296 /* XXX lwp */
297 LIST_INSERT_HEAD(&newproc->p_lwps, &newproc->p_lwp, lwp_list);
f5329d64 298 newproc->p_lwp.lwp_proc = newproc;
ef09c3ed 299 newproc->p_lwp.lwp_tid = 0;
984263bc
MD
300
301 /*
302 * Find an unused process ID. We remember a range of unused IDs
303 * ready to use (from nextpid+1 through pidchecked-1).
304 */
305 nextpid++;
306 if (randompid)
307 nextpid += arc4random() % randompid;
308retry:
309 /*
310 * If the process ID prototype has wrapped around,
311 * restart somewhat above 0, as the low-numbered procs
312 * tend to include daemons that don't exit.
313 */
314 if (nextpid >= PID_MAX) {
315 nextpid = nextpid % PID_MAX;
316 if (nextpid < 100)
317 nextpid += 100;
318 pidchecked = 0;
319 }
320 if (nextpid >= pidchecked) {
321 int doingzomb = 0;
322
323 pidchecked = PID_MAX;
324 /*
325 * Scan the active and zombie procs to check whether this pid
326 * is in use. Remember the lowest pid that's greater
327 * than nextpid, so we can avoid checking for a while.
328 */
329 p2 = LIST_FIRST(&allproc);
330again:
331 for (; p2 != 0; p2 = LIST_NEXT(p2, p_list)) {
332 while (p2->p_pid == nextpid ||
333 p2->p_pgrp->pg_id == nextpid ||
334 p2->p_session->s_sid == nextpid) {
335 nextpid++;
336 if (nextpid >= pidchecked)
337 goto retry;
338 }
339 if (p2->p_pid > nextpid && pidchecked > p2->p_pid)
340 pidchecked = p2->p_pid;
341 if (p2->p_pgrp->pg_id > nextpid &&
342 pidchecked > p2->p_pgrp->pg_id)
343 pidchecked = p2->p_pgrp->pg_id;
344 if (p2->p_session->s_sid > nextpid &&
345 pidchecked > p2->p_session->s_sid)
346 pidchecked = p2->p_session->s_sid;
347 }
348 if (!doingzomb) {
349 doingzomb = 1;
350 p2 = LIST_FIRST(&zombproc);
351 goto again;
352 }
353 }
354
355 p2 = newproc;
356 p2->p_stat = SIDL; /* protect against others */
357 p2->p_pid = nextpid;
358 LIST_INSERT_HEAD(&allproc, p2, p_list);
359 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash);
360
361 /*
362 * Make a proc table entry for the new process.
363 * Start by zeroing the section of proc that is zero-initialized,
364 * then copy the section that is copied directly from the parent.
365 */
366 bzero(&p2->p_startzero,
367 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero));
ef09c3ed
SS
368 bzero(&p2->p_lwp.lwp_startzero,
369 (unsigned) ((caddr_t)&p2->p_lwp.lwp_endzero -
370 (caddr_t)&p2->p_lwp.lwp_startzero));
984263bc
MD
371 bcopy(&p1->p_startcopy, &p2->p_startcopy,
372 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
ef09c3ed
SS
373 bcopy(&p1->p_lwp.lwp_startcopy, &p2->p_lwp.lwp_startcopy,
374 (unsigned) ((caddr_t)&p2->p_lwp.lwp_endcopy -
375 (caddr_t)&p2->p_lwp.lwp_startcopy));
984263bc
MD
376
377 p2->p_aioinfo = NULL;
378
379 /*
380 * Duplicate sub-structures as needed.
381 * Increase reference counts on shared objects.
382 * The p_stats and p_sigacts substructs are set in vm_fork.
383 */
8ec60c3f 384 p2->p_flag = P_INMEM;
984263bc
MD
385 if (p1->p_flag & P_PROFIL)
386 startprofclock(p2);
dadab5e9 387 p2->p_ucred = crhold(p1->p_ucred);
984263bc 388
b40e316c 389 if (jailed(p2->p_ucred))
984263bc 390 p2->p_flag |= P_JAILED;
984263bc
MD
391
392 if (p2->p_args)
393 p2->p_args->ar_ref++;
394
395 if (flags & RFSIGSHARE) {
396 p2->p_procsig = p1->p_procsig;
397 p2->p_procsig->ps_refcnt++;
398 if (p1->p_sigacts == &p1->p_addr->u_sigacts) {
399 struct sigacts *newsigacts;
984263bc
MD
400
401 /* Create the shared sigacts structure */
402 MALLOC(newsigacts, struct sigacts *,
403 sizeof(struct sigacts), M_SUBPROC, M_WAITOK);
e43a034f 404 crit_enter();
984263bc
MD
405 /*
406 * Set p_sigacts to the new shared structure.
407 * Note that this is updating p1->p_sigacts at the
408 * same time, since p_sigacts is just a pointer to
409 * the shared p_procsig->ps_sigacts.
410 */
411 p2->p_sigacts = newsigacts;
412 bcopy(&p1->p_addr->u_sigacts, p2->p_sigacts,
413 sizeof(*p2->p_sigacts));
414 *p2->p_sigacts = p1->p_addr->u_sigacts;
e43a034f 415 crit_exit();
984263bc
MD
416 }
417 } else {
418 MALLOC(p2->p_procsig, struct procsig *, sizeof(struct procsig),
419 M_SUBPROC, M_WAITOK);
420 bcopy(p1->p_procsig, p2->p_procsig, sizeof(*p2->p_procsig));
421 p2->p_procsig->ps_refcnt = 1;
422 p2->p_sigacts = NULL; /* finished in vm_fork() */
423 }
424 if (flags & RFLINUXTHPN)
425 p2->p_sigparent = SIGUSR1;
426 else
427 p2->p_sigparent = SIGCHLD;
428
429 /* bump references to the text vnode (for procfs) */
430 p2->p_textvp = p1->p_textvp;
431 if (p2->p_textvp)
597aea93 432 vref(p2->p_textvp);
984263bc
MD
433
434 if (flags & RFCFDG) {
435 p2->p_fd = fdinit(p1);
436 fdtol = NULL;
437 } else if (flags & RFFDG) {
438 p2->p_fd = fdcopy(p1);
439 fdtol = NULL;
440 } else {
441 p2->p_fd = fdshare(p1);
442 if (p1->p_fdtol == NULL)
443 p1->p_fdtol =
444 filedesc_to_leader_alloc(NULL,
445 p1->p_leader);
446 if ((flags & RFTHREAD) != 0) {
447 /*
448 * Shared file descriptor table and
449 * shared process leaders.
450 */
451 fdtol = p1->p_fdtol;
452 fdtol->fdl_refcount++;
453 } else {
454 /*
455 * Shared file descriptor table, and
456 * different process leaders
457 */
98a7f915 458 fdtol = filedesc_to_leader_alloc(p1->p_fdtol, p2);
984263bc
MD
459 }
460 }
461 p2->p_fdtol = fdtol;
462
463 /*
464 * If p_limit is still copy-on-write, bump refcnt,
465 * otherwise get a copy that won't be modified.
466 * (If PL_SHAREMOD is clear, the structure is shared
467 * copy-on-write.)
468 */
303c76d5 469 if (p1->p_limit->p_lflags & PL_SHAREMOD) {
984263bc 470 p2->p_limit = limcopy(p1->p_limit);
303c76d5 471 } else {
984263bc
MD
472 p2->p_limit = p1->p_limit;
473 p2->p_limit->p_refcnt++;
474 }
475
476 /*
477 * Preserve some more flags in subprocess. P_PROFIL has already
478 * been preserved.
479 */
480 p2->p_flag |= p1->p_flag & (P_SUGID | P_ALTSTACK);
481 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
482 p2->p_flag |= P_CONTROLT;
483 if (flags & RFPPWAIT)
484 p2->p_flag |= P_PPWAIT;
485
5fd012e0
MD
486 /*
487 * Once we are on a pglist we may receive signals. XXX we might
488 * race a ^C being sent to the process group by not receiving it
489 * at all prior to this line.
490 */
984263bc
MD
491 LIST_INSERT_AFTER(p1, p2, p_pglist);
492
493 /*
494 * Attach the new process to its parent.
495 *
496 * If RFNOWAIT is set, the newly created process becomes a child
497 * of init. This effectively disassociates the child from the
498 * parent.
499 */
500 if (flags & RFNOWAIT)
501 pptr = initproc;
502 else
503 pptr = p1;
504 p2->p_pptr = pptr;
505 LIST_INSERT_HEAD(&pptr->p_children, p2, p_sibling);
506 LIST_INIT(&p2->p_children);
98a7f915 507 varsymset_init(&p2->p_varsymset, &p1->p_varsymset);
8fbf9130 508 callout_init(&p2->p_ithandle);
984263bc
MD
509
510#ifdef KTRACE
511 /*
512 * Copy traceflag and tracefile if enabled. If not inherited,
513 * these were zeroed above but we still could have a trace race
514 * so make sure p2's p_tracep is NULL.
515 */
516 if ((p1->p_traceflag & KTRFAC_INHERIT) && p2->p_tracep == NULL) {
517 p2->p_traceflag = p1->p_traceflag;
518 if ((p2->p_tracep = p1->p_tracep) != NULL)
597aea93 519 vref(p2->p_tracep);
984263bc
MD
520 }
521#endif
522
523 /*
352f5709 524 * Inherit the scheduler and initialize scheduler-related fields.
dcc99b62
MD
525 * Set cpbase to the last timeout that occured (not the upcoming
526 * timeout).
a77ac49d
MD
527 */
528 p2->p_usched = p1->p_usched;
dcc99b62
MD
529 p2->p_cpbase = mycpu->gd_schedclock.time -
530 mycpu->gd_schedclock.periodic;
352f5709 531 p2->p_usched->heuristic_forking(p1, p2);
a77ac49d 532
984263bc
MD
533 /*
534 * This begins the section where we must prevent the parent
535 * from being swapped.
536 */
537 PHOLD(p1);
538
539 /*
540 * Finish creating the child process. It will return via a different
541 * execution path later. (ie: directly into user mode)
542 */
543 vm_fork(p1, p2, flags);
e30f9e2c 544 caps_fork(p1, p2, flags);
984263bc
MD
545
546 if (flags == (RFFDG | RFPROC)) {
12e4aaff
MD
547 mycpu->gd_cnt.v_forks++;
548 mycpu->gd_cnt.v_forkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 549 } else if (flags == (RFFDG | RFPROC | RFPPWAIT | RFMEM)) {
12e4aaff
MD
550 mycpu->gd_cnt.v_vforks++;
551 mycpu->gd_cnt.v_vforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 552 } else if (p1 == &proc0) {
12e4aaff
MD
553 mycpu->gd_cnt.v_kthreads++;
554 mycpu->gd_cnt.v_kthreadpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 555 } else {
12e4aaff
MD
556 mycpu->gd_cnt.v_rforks++;
557 mycpu->gd_cnt.v_rforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc
MD
558 }
559
560 /*
561 * Both processes are set up, now check if any loadable modules want
562 * to adjust anything.
563 * What if they have an error? XXX
564 */
565 TAILQ_FOREACH(ep, &fork_list, next) {
566 (*ep->function)(p1, p2, flags);
567 }
568
569 /*
a77ac49d
MD
570 * Set the start time. Note that the process is not runnable. The
571 * caller is responsible for making it runnable.
984263bc 572 */
41f3429e 573 microtime(&p2->p_thread->td_start);
984263bc 574 p2->p_acflag = AFORK;
984263bc
MD
575
576 /*
577 * tell any interested parties about the new process
578 */
579 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid);
580
984263bc
MD
581 /*
582 * Return child proc pointer to parent.
583 */
584 *procp = p2;
585 return (0);
586}
587
588/*
589 * The next two functionms are general routines to handle adding/deleting
590 * items on the fork callout list.
591 *
592 * at_fork():
593 * Take the arguments given and put them onto the fork callout list,
594 * However first make sure that it's not already there.
595 * Returns 0 on success or a standard error number.
596 */
984263bc 597int
303c76d5 598at_fork(forklist_fn function)
984263bc
MD
599{
600 struct forklist *ep;
601
602#ifdef INVARIANTS
603 /* let the programmer know if he's been stupid */
303c76d5 604 if (rm_at_fork(function)) {
984263bc
MD
605 printf("WARNING: fork callout entry (%p) already present\n",
606 function);
303c76d5 607 }
984263bc 608#endif
679246c3 609 ep = malloc(sizeof(*ep), M_ATFORK, M_WAITOK|M_ZERO);
984263bc
MD
610 ep->function = function;
611 TAILQ_INSERT_TAIL(&fork_list, ep, next);
612 return (0);
613}
614
615/*
616 * Scan the exit callout list for the given item and remove it..
617 * Returns the number of items removed (0 or 1)
618 */
984263bc 619int
303c76d5 620rm_at_fork(forklist_fn function)
984263bc
MD
621{
622 struct forklist *ep;
623
624 TAILQ_FOREACH(ep, &fork_list, next) {
625 if (ep->function == function) {
626 TAILQ_REMOVE(&fork_list, ep, next);
627 free(ep, M_ATFORK);
628 return(1);
629 }
630 }
631 return (0);
632}
7d0bac62
MD
633
634/*
635 * Add a forked process to the run queue after any remaining setup, such
636 * as setting the fork handler, has been completed.
637 */
7d0bac62
MD
638void
639start_forked_proc(struct proc *p1, struct proc *p2)
640{
641 /*
26a0694b
MD
642 * Move from SIDL to RUN queue, and activate the process's thread.
643 * Activation of the thread effectively makes the process "a"
644 * current process, so we do not setrunqueue().
8ec60c3f
MD
645 *
646 * YYY setrunqueue works here but we should clean up the trampoline
647 * code so we just schedule the LWKT thread and let the trampoline
648 * deal with the userland scheduler on return to userland.
7d0bac62 649 */
f61c1ff1 650 KASSERT(p2 && p2->p_stat == SIDL,
7d0bac62 651 ("cannot start forked process, bad status: %p", p2));
a77ac49d 652 p2->p_usched->resetpriority(p2);
e43a034f 653 crit_enter();
7d0bac62 654 p2->p_stat = SRUN;
a77ac49d 655 p2->p_usched->setrunqueue(p2);
e43a034f 656 crit_exit();
7d0bac62
MD
657
658 /*
659 * Now can be swapped.
660 */
661 PRELE(p1);
662
663 /*
664 * Preserve synchronization semantics of vfork. If waiting for
665 * child to exec or exit, set P_PPWAIT on child, and sleep on our
666 * proc (in case of exit).
667 */
668 while (p2->p_flag & P_PPWAIT)
377d4740 669 tsleep(p1, 0, "ppwait", 0);
7d0bac62
MD
670}
671