Start consolidating process related code into kern_proc.c. Implement
[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 $
c0b8a06d 40 * $DragonFly: src/sys/kern/kern_fork.c,v 1.48 2006/05/23 20:35:10 dillon 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{
553ea3c8 92 struct lwp *lp = curthread->td_lwp;
984263bc 93 struct proc *p2;
41c20dac 94 int error;
984263bc 95
553ea3c8 96 error = fork1(lp, RFFDG | RFPROC, &p2);
984263bc 97 if (error == 0) {
553ea3c8 98 start_forked_proc(lp, 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{
553ea3c8 109 struct lwp *lp = curthread->td_lwp;
984263bc 110 struct proc *p2;
41c20dac 111 int error;
984263bc 112
553ea3c8 113 error = fork1(lp, RFFDG | RFPROC | RFPPWAIT | RFMEM, &p2);
984263bc 114 if (error == 0) {
553ea3c8 115 start_forked_proc(lp, 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{
553ea3c8 136 struct lwp *lp = curthread->td_lwp;
984263bc 137 struct proc *p2;
41c20dac 138 int error;
984263bc 139
6654fbcb
MD
140 if ((uap->flags & RFKERNELONLY) != 0)
141 return (EINVAL);
142
553ea3c8 143 error = fork1(lp, uap->flags, &p2);
984263bc 144 if (error == 0) {
f61c1ff1 145 if (p2)
553ea3c8 146 start_forked_proc(lp, 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
553ea3c8 190fork1(struct lwp *lp1, int flags, struct proc **procp)
984263bc 191{
553ea3c8 192 struct proc *p1 = lp1->lwp_proc;
984263bc 193 struct proc *p2, *pptr;
553ea3c8 194 struct lwp *lp2;
984263bc
MD
195 uid_t uid;
196 struct proc *newproc;
197 int ok;
5bc7cd8d
HP
198 static int curfail = 0, pidchecked = 0;
199 static struct timeval lastfail;
984263bc
MD
200 struct forklist *ep;
201 struct filedesc_to_leader *fdtol;
202
203 if ((flags & (RFFDG|RFCFDG)) == (RFFDG|RFCFDG))
204 return (EINVAL);
205
206 /*
207 * Here we don't create a new process, but we divorce
208 * certain parts of a process from itself.
209 */
210 if ((flags & RFPROC) == 0) {
211
212 vm_fork(p1, 0, flags);
213
214 /*
215 * Close all file descriptors.
216 */
217 if (flags & RFCFDG) {
218 struct filedesc *fdtmp;
219 fdtmp = fdinit(p1);
220 fdfree(p1);
221 p1->p_fd = fdtmp;
222 }
223
224 /*
225 * Unshare file descriptors (from parent.)
226 */
227 if (flags & RFFDG) {
228 if (p1->p_fd->fd_refcnt > 1) {
229 struct filedesc *newfd;
230 newfd = fdcopy(p1);
231 fdfree(p1);
232 p1->p_fd = newfd;
233 }
234 }
235 *procp = NULL;
236 return (0);
237 }
238
239 /*
240 * Although process entries are dynamically created, we still keep
241 * a global limit on the maximum number we will create. Don't allow
242 * a nonprivileged user to use the last ten processes; don't let root
243 * exceed the limit. The variable nprocs is the current number of
244 * processes, maxproc is the limit.
245 */
41c20dac 246 uid = p1->p_ucred->cr_ruid;
984263bc 247 if ((nprocs >= maxproc - 10 && uid != 0) || nprocs >= maxproc) {
5bc7cd8d
HP
248 if (ppsratecheck(&lastfail, &curfail, 1))
249 printf("maxproc limit exceeded by uid %d, please "
250 "see tuning(7) and login.conf(5).\n", uid);
377d4740 251 tsleep(&forksleep, 0, "fork", hz / 2);
984263bc
MD
252 return (EAGAIN);
253 }
254 /*
255 * Increment the nprocs resource before blocking can occur. There
256 * are hard-limits as to the number of processes that can run.
257 */
258 nprocs++;
259
260 /*
261 * Increment the count of procs running with this uid. Don't allow
262 * a nonprivileged user to exceed their current limit.
263 */
41c20dac 264 ok = chgproccnt(p1->p_ucred->cr_ruidinfo, 1,
984263bc
MD
265 (uid != 0) ? p1->p_rlimit[RLIMIT_NPROC].rlim_cur : 0);
266 if (!ok) {
267 /*
268 * Back out the process count
269 */
270 nprocs--;
5bc7cd8d
HP
271 if (ppsratecheck(&lastfail, &curfail, 1))
272 printf("maxproc limit exceeded by uid %d, please "
273 "see tuning(7) and login.conf(5).\n", uid);
377d4740 274 tsleep(&forksleep, 0, "fork", hz / 2);
984263bc
MD
275 return (EAGAIN);
276 }
277
278 /* Allocate new proc. */
279 newproc = zalloc(proc_zone);
280
281 /*
ef09c3ed 282 * Setup linkage for kernel based threading XXX lwp
984263bc 283 */
cb74210d 284 if (flags & RFTHREAD) {
984263bc
MD
285 newproc->p_peers = p1->p_peers;
286 p1->p_peers = newproc;
287 newproc->p_leader = p1->p_leader;
288 } else {
cb74210d 289 newproc->p_peers = NULL;
984263bc
MD
290 newproc->p_leader = newproc;
291 }
292
293 newproc->p_wakeup = 0;
984263bc 294 newproc->p_vmspace = NULL;
cb74210d
MD
295 newproc->p_numposixlocks = 0;
296 newproc->p_emuldata = NULL;
81eea9f4 297 TAILQ_INIT(&newproc->p_lwp.lwp_sysmsgq);
ef09c3ed
SS
298 LIST_INIT(&newproc->p_lwps);
299
300 /* XXX lwp */
553ea3c8
SS
301 lp2 = &newproc->p_lwp;
302 lp2->lwp_proc = newproc;
303 lp2->lwp_tid = 0;
304 LIST_INSERT_HEAD(&newproc->p_lwps, lp2, lwp_list);
ac33e87f 305 newproc->p_nthreads = 1;
cb74210d
MD
306 newproc->p_nstopped = 0;
307 newproc->p_lasttid = 0;
984263bc
MD
308
309 /*
310 * Find an unused process ID. We remember a range of unused IDs
311 * ready to use (from nextpid+1 through pidchecked-1).
312 */
313 nextpid++;
314 if (randompid)
315 nextpid += arc4random() % randompid;
316retry:
317 /*
318 * If the process ID prototype has wrapped around,
319 * restart somewhat above 0, as the low-numbered procs
320 * tend to include daemons that don't exit.
321 */
322 if (nextpid >= PID_MAX) {
323 nextpid = nextpid % PID_MAX;
324 if (nextpid < 100)
325 nextpid += 100;
326 pidchecked = 0;
327 }
328 if (nextpid >= pidchecked) {
329 int doingzomb = 0;
330
331 pidchecked = PID_MAX;
332 /*
333 * Scan the active and zombie procs to check whether this pid
334 * is in use. Remember the lowest pid that's greater
335 * than nextpid, so we can avoid checking for a while.
336 */
337 p2 = LIST_FIRST(&allproc);
338again:
339 for (; p2 != 0; p2 = LIST_NEXT(p2, p_list)) {
340 while (p2->p_pid == nextpid ||
341 p2->p_pgrp->pg_id == nextpid ||
342 p2->p_session->s_sid == nextpid) {
343 nextpid++;
344 if (nextpid >= pidchecked)
345 goto retry;
346 }
347 if (p2->p_pid > nextpid && pidchecked > p2->p_pid)
348 pidchecked = p2->p_pid;
349 if (p2->p_pgrp->pg_id > nextpid &&
350 pidchecked > p2->p_pgrp->pg_id)
351 pidchecked = p2->p_pgrp->pg_id;
352 if (p2->p_session->s_sid > nextpid &&
353 pidchecked > p2->p_session->s_sid)
354 pidchecked = p2->p_session->s_sid;
355 }
356 if (!doingzomb) {
357 doingzomb = 1;
358 p2 = LIST_FIRST(&zombproc);
359 goto again;
360 }
361 }
362
363 p2 = newproc;
364 p2->p_stat = SIDL; /* protect against others */
365 p2->p_pid = nextpid;
366 LIST_INSERT_HEAD(&allproc, p2, p_list);
367 LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash);
368
369 /*
370 * Make a proc table entry for the new process.
371 * Start by zeroing the section of proc that is zero-initialized,
372 * then copy the section that is copied directly from the parent.
373 */
374 bzero(&p2->p_startzero,
375 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero));
553ea3c8
SS
376 bzero(&lp2->lwp_startzero,
377 (unsigned) ((caddr_t)&lp2->lwp_endzero -
378 (caddr_t)&lp2->lwp_startzero));
984263bc
MD
379 bcopy(&p1->p_startcopy, &p2->p_startcopy,
380 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
553ea3c8
SS
381 bcopy(&p1->p_lwp.lwp_startcopy, &lp2->lwp_startcopy,
382 (unsigned) ((caddr_t)&lp2->lwp_endcopy -
383 (caddr_t)&lp2->lwp_startcopy));
984263bc
MD
384
385 p2->p_aioinfo = NULL;
386
387 /*
388 * Duplicate sub-structures as needed.
389 * Increase reference counts on shared objects.
390 * The p_stats and p_sigacts substructs are set in vm_fork.
3a72e617 391 * p_lock is in the copy area and must be cleared.
984263bc 392 */
344ad853 393 p2->p_flag = 0;
3a72e617 394 p2->p_lock = 0;
984263bc
MD
395 if (p1->p_flag & P_PROFIL)
396 startprofclock(p2);
dadab5e9 397 p2->p_ucred = crhold(p1->p_ucred);
984263bc 398
b40e316c 399 if (jailed(p2->p_ucred))
984263bc 400 p2->p_flag |= P_JAILED;
984263bc
MD
401
402 if (p2->p_args)
403 p2->p_args->ar_ref++;
404
405 if (flags & RFSIGSHARE) {
406 p2->p_procsig = p1->p_procsig;
407 p2->p_procsig->ps_refcnt++;
408 if (p1->p_sigacts == &p1->p_addr->u_sigacts) {
409 struct sigacts *newsigacts;
984263bc
MD
410
411 /* Create the shared sigacts structure */
412 MALLOC(newsigacts, struct sigacts *,
413 sizeof(struct sigacts), M_SUBPROC, M_WAITOK);
e43a034f 414 crit_enter();
984263bc
MD
415 /*
416 * Set p_sigacts to the new shared structure.
417 * Note that this is updating p1->p_sigacts at the
418 * same time, since p_sigacts is just a pointer to
419 * the shared p_procsig->ps_sigacts.
420 */
421 p2->p_sigacts = newsigacts;
422 bcopy(&p1->p_addr->u_sigacts, p2->p_sigacts,
423 sizeof(*p2->p_sigacts));
424 *p2->p_sigacts = p1->p_addr->u_sigacts;
e43a034f 425 crit_exit();
984263bc
MD
426 }
427 } else {
428 MALLOC(p2->p_procsig, struct procsig *, sizeof(struct procsig),
429 M_SUBPROC, M_WAITOK);
430 bcopy(p1->p_procsig, p2->p_procsig, sizeof(*p2->p_procsig));
431 p2->p_procsig->ps_refcnt = 1;
432 p2->p_sigacts = NULL; /* finished in vm_fork() */
433 }
434 if (flags & RFLINUXTHPN)
435 p2->p_sigparent = SIGUSR1;
436 else
437 p2->p_sigparent = SIGCHLD;
438
439 /* bump references to the text vnode (for procfs) */
440 p2->p_textvp = p1->p_textvp;
441 if (p2->p_textvp)
597aea93 442 vref(p2->p_textvp);
984263bc
MD
443
444 if (flags & RFCFDG) {
445 p2->p_fd = fdinit(p1);
446 fdtol = NULL;
447 } else if (flags & RFFDG) {
448 p2->p_fd = fdcopy(p1);
449 fdtol = NULL;
450 } else {
451 p2->p_fd = fdshare(p1);
452 if (p1->p_fdtol == NULL)
453 p1->p_fdtol =
454 filedesc_to_leader_alloc(NULL,
455 p1->p_leader);
456 if ((flags & RFTHREAD) != 0) {
457 /*
458 * Shared file descriptor table and
459 * shared process leaders.
460 */
461 fdtol = p1->p_fdtol;
462 fdtol->fdl_refcount++;
463 } else {
464 /*
465 * Shared file descriptor table, and
466 * different process leaders
467 */
98a7f915 468 fdtol = filedesc_to_leader_alloc(p1->p_fdtol, p2);
984263bc
MD
469 }
470 }
471 p2->p_fdtol = fdtol;
c0b8a06d 472 p2->p_limit = plimit_fork(p1->p_limit);
984263bc
MD
473
474 /*
475 * Preserve some more flags in subprocess. P_PROFIL has already
476 * been preserved.
477 */
478 p2->p_flag |= p1->p_flag & (P_SUGID | P_ALTSTACK);
479 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
480 p2->p_flag |= P_CONTROLT;
481 if (flags & RFPPWAIT)
482 p2->p_flag |= P_PPWAIT;
483
5fd012e0
MD
484 /*
485 * Once we are on a pglist we may receive signals. XXX we might
486 * race a ^C being sent to the process group by not receiving it
487 * at all prior to this line.
488 */
984263bc
MD
489 LIST_INSERT_AFTER(p1, p2, p_pglist);
490
491 /*
492 * Attach the new process to its parent.
493 *
494 * If RFNOWAIT is set, the newly created process becomes a child
495 * of init. This effectively disassociates the child from the
496 * parent.
497 */
498 if (flags & RFNOWAIT)
499 pptr = initproc;
500 else
501 pptr = p1;
502 p2->p_pptr = pptr;
503 LIST_INSERT_HEAD(&pptr->p_children, p2, p_sibling);
504 LIST_INIT(&p2->p_children);
98a7f915 505 varsymset_init(&p2->p_varsymset, &p1->p_varsymset);
8fbf9130 506 callout_init(&p2->p_ithandle);
984263bc
MD
507
508#ifdef KTRACE
509 /*
510 * Copy traceflag and tracefile if enabled. If not inherited,
511 * these were zeroed above but we still could have a trace race
29f58392 512 * so make sure p2's p_tracenode is NULL.
984263bc 513 */
29f58392 514 if ((p1->p_traceflag & KTRFAC_INHERIT) && p2->p_tracenode == NULL) {
984263bc 515 p2->p_traceflag = p1->p_traceflag;
29f58392 516 p2->p_tracenode = ktrinherit(p1->p_tracenode);
984263bc
MD
517 }
518#endif
519
520 /*
352f5709 521 * Inherit the scheduler and initialize scheduler-related fields.
dcc99b62
MD
522 * Set cpbase to the last timeout that occured (not the upcoming
523 * timeout).
a77ac49d
MD
524 */
525 p2->p_usched = p1->p_usched;
553ea3c8 526 lp2->lwp_cpbase = mycpu->gd_schedclock.time -
dcc99b62 527 mycpu->gd_schedclock.periodic;
553ea3c8 528 p2->p_usched->heuristic_forking(&p1->p_lwp, lp2);
a77ac49d 529
984263bc
MD
530 /*
531 * This begins the section where we must prevent the parent
532 * from being swapped.
533 */
534 PHOLD(p1);
535
536 /*
537 * Finish creating the child process. It will return via a different
538 * execution path later. (ie: directly into user mode)
539 */
540 vm_fork(p1, p2, flags);
e30f9e2c 541 caps_fork(p1, p2, flags);
984263bc
MD
542
543 if (flags == (RFFDG | RFPROC)) {
12e4aaff
MD
544 mycpu->gd_cnt.v_forks++;
545 mycpu->gd_cnt.v_forkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 546 } else if (flags == (RFFDG | RFPROC | RFPPWAIT | RFMEM)) {
12e4aaff
MD
547 mycpu->gd_cnt.v_vforks++;
548 mycpu->gd_cnt.v_vforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 549 } else if (p1 == &proc0) {
12e4aaff
MD
550 mycpu->gd_cnt.v_kthreads++;
551 mycpu->gd_cnt.v_kthreadpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 552 } else {
12e4aaff
MD
553 mycpu->gd_cnt.v_rforks++;
554 mycpu->gd_cnt.v_rforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc
MD
555 }
556
557 /*
558 * Both processes are set up, now check if any loadable modules want
559 * to adjust anything.
560 * What if they have an error? XXX
561 */
562 TAILQ_FOREACH(ep, &fork_list, next) {
563 (*ep->function)(p1, p2, flags);
564 }
565
566 /*
a77ac49d
MD
567 * Set the start time. Note that the process is not runnable. The
568 * caller is responsible for making it runnable.
984263bc 569 */
d9fa5f67 570 microtime(&p2->p_start);
984263bc 571 p2->p_acflag = AFORK;
984263bc
MD
572
573 /*
574 * tell any interested parties about the new process
575 */
576 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid);
577
984263bc
MD
578 /*
579 * Return child proc pointer to parent.
580 */
581 *procp = p2;
582 return (0);
583}
584
585/*
586 * The next two functionms are general routines to handle adding/deleting
587 * items on the fork callout list.
588 *
589 * at_fork():
590 * Take the arguments given and put them onto the fork callout list,
591 * However first make sure that it's not already there.
592 * Returns 0 on success or a standard error number.
593 */
984263bc 594int
303c76d5 595at_fork(forklist_fn function)
984263bc
MD
596{
597 struct forklist *ep;
598
599#ifdef INVARIANTS
600 /* let the programmer know if he's been stupid */
303c76d5 601 if (rm_at_fork(function)) {
984263bc
MD
602 printf("WARNING: fork callout entry (%p) already present\n",
603 function);
303c76d5 604 }
984263bc 605#endif
679246c3 606 ep = malloc(sizeof(*ep), M_ATFORK, M_WAITOK|M_ZERO);
984263bc
MD
607 ep->function = function;
608 TAILQ_INSERT_TAIL(&fork_list, ep, next);
609 return (0);
610}
611
612/*
613 * Scan the exit callout list for the given item and remove it..
614 * Returns the number of items removed (0 or 1)
615 */
984263bc 616int
303c76d5 617rm_at_fork(forklist_fn function)
984263bc
MD
618{
619 struct forklist *ep;
620
621 TAILQ_FOREACH(ep, &fork_list, next) {
622 if (ep->function == function) {
623 TAILQ_REMOVE(&fork_list, ep, next);
624 free(ep, M_ATFORK);
625 return(1);
626 }
627 }
628 return (0);
629}
7d0bac62
MD
630
631/*
632 * Add a forked process to the run queue after any remaining setup, such
633 * as setting the fork handler, has been completed.
634 */
7d0bac62 635void
553ea3c8 636start_forked_proc(struct lwp *lp1, struct proc *p2)
7d0bac62 637{
553ea3c8
SS
638 struct lwp *lp2;
639
640 KKASSERT(p2 != NULL && p2->p_nthreads == 1);
641
642 lp2 = LIST_FIRST(&p2->p_lwps);
643
7d0bac62 644 /*
26a0694b
MD
645 * Move from SIDL to RUN queue, and activate the process's thread.
646 * Activation of the thread effectively makes the process "a"
647 * current process, so we do not setrunqueue().
8ec60c3f
MD
648 *
649 * YYY setrunqueue works here but we should clean up the trampoline
650 * code so we just schedule the LWKT thread and let the trampoline
651 * deal with the userland scheduler on return to userland.
7d0bac62 652 */
553ea3c8 653 KASSERT(p2->p_stat == SIDL,
7d0bac62 654 ("cannot start forked process, bad status: %p", p2));
553ea3c8 655 p2->p_usched->resetpriority(lp2);
e43a034f 656 crit_enter();
7d0bac62 657 p2->p_stat = SRUN;
553ea3c8 658 p2->p_usched->setrunqueue(lp2);
e43a034f 659 crit_exit();
7d0bac62
MD
660
661 /*
662 * Now can be swapped.
663 */
553ea3c8 664 PRELE(lp1->lwp_proc);
7d0bac62
MD
665
666 /*
667 * Preserve synchronization semantics of vfork. If waiting for
668 * child to exec or exit, set P_PPWAIT on child, and sleep on our
669 * proc (in case of exit).
670 */
671 while (p2->p_flag & P_PPWAIT)
553ea3c8 672 tsleep(lp1->lwp_proc, 0, "ppwait", 0);
7d0bac62 673}