Regen.
[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 $
91bd9c1e 40 * $DragonFly: src/sys/kern/kern_fork.c,v 1.66 2007/03/01 01:46:52 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
91bd9c1e
SS
155int
156sys_lwp_create(struct lwp_create_args *uap)
157{
158 struct proc *p = curproc;
159 struct lwp *lp;
160 struct lwp_params params;
161 int error;
162
163 error = copyin(uap->params, &params, sizeof(params));
164 if (error)
165 goto fail2;
166
167 lp = lwp_fork(curthread->td_lwp, p, RFPROC);
168 error = cpu_prepare_lwp(lp, &params);
169 if (params.tid1 != NULL &&
170 (error = copyout(&lp->lwp_tid, params.tid1, sizeof(lp->lwp_tid))))
171 goto fail;
172 if (params.tid2 != NULL &&
173 (error = copyout(&lp->lwp_tid, params.tid2, sizeof(lp->lwp_tid))))
174 goto fail;
175
176 /*
177 * Now schedule the new lwp.
178 */
179 p->p_usched->resetpriority(lp);
180 crit_enter();
181 lp->lwp_stat = LSRUN;
182 p->p_usched->setrunqueue(lp);
183 crit_exit();
184
185 return (0);
186
187fail:
188 --p->p_nthreads;
189 LIST_REMOVE(lp, lwp_list);
190 /* lwp_dispose expects a exited lwp */
191 lp->lwp_thread->td_flags = TDF_EXITING;
192 lwp_dispose(lp);
193fail2:
194 return (error);
195}
984263bc
MD
196
197int nprocs = 1; /* process 0 */
984263bc
MD
198
199int
553ea3c8 200fork1(struct lwp *lp1, int flags, struct proc **procp)
984263bc 201{
553ea3c8 202 struct proc *p1 = lp1->lwp_proc;
984263bc 203 struct proc *p2, *pptr;
167e6ecb 204 struct pgrp *pgrp;
984263bc 205 uid_t uid;
167e6ecb 206 int ok, error;
51e64ff2 207 static int curfail = 0;
5bc7cd8d 208 static struct timeval lastfail;
984263bc
MD
209 struct forklist *ep;
210 struct filedesc_to_leader *fdtol;
211
212 if ((flags & (RFFDG|RFCFDG)) == (RFFDG|RFCFDG))
213 return (EINVAL);
214
215 /*
216 * Here we don't create a new process, but we divorce
217 * certain parts of a process from itself.
218 */
219 if ((flags & RFPROC) == 0) {
220
13d13d89
SS
221 /*
222 * This kind of stunt does not work anymore if
223 * there are native threads (lwps) running
224 */
225 if (p1->p_nthreads != 1)
226 return (EINVAL);
227
228 vm_fork(p1, 0, flags);
984263bc
MD
229
230 /*
231 * Close all file descriptors.
232 */
233 if (flags & RFCFDG) {
234 struct filedesc *fdtmp;
235 fdtmp = fdinit(p1);
236 fdfree(p1);
237 p1->p_fd = fdtmp;
238 }
239
240 /*
241 * Unshare file descriptors (from parent.)
242 */
243 if (flags & RFFDG) {
244 if (p1->p_fd->fd_refcnt > 1) {
245 struct filedesc *newfd;
246 newfd = fdcopy(p1);
247 fdfree(p1);
248 p1->p_fd = newfd;
249 }
250 }
251 *procp = NULL;
252 return (0);
253 }
254
167e6ecb
MD
255 /*
256 * Interlock against process group signal delivery. If signals
257 * are pending after the interlock is obtained we have to restart
258 * the system call to process the signals. If we don't the child
259 * can miss a pgsignal (such as ^C) sent during the fork.
260 *
261 * We can't use CURSIG() here because it will process any STOPs
262 * and cause the process group lock to be held indefinitely. If
263 * a STOP occurs, the fork will be restarted after the CONT.
264 */
265 error = 0;
266 pgrp = NULL;
267 if ((flags & RFPGLOCK) && (pgrp = p1->p_pgrp) != NULL) {
268 lockmgr(&pgrp->pg_lock, LK_SHARED);
08f2f1bb 269 if (CURSIGNB(lp1)) {
167e6ecb
MD
270 error = ERESTART;
271 goto done;
272 }
273 }
274
984263bc
MD
275 /*
276 * Although process entries are dynamically created, we still keep
277 * a global limit on the maximum number we will create. Don't allow
278 * a nonprivileged user to use the last ten processes; don't let root
279 * exceed the limit. The variable nprocs is the current number of
280 * processes, maxproc is the limit.
281 */
41c20dac 282 uid = p1->p_ucred->cr_ruid;
984263bc 283 if ((nprocs >= maxproc - 10 && uid != 0) || nprocs >= maxproc) {
5bc7cd8d 284 if (ppsratecheck(&lastfail, &curfail, 1))
6ea70f76 285 kprintf("maxproc limit exceeded by uid %d, please "
5bc7cd8d 286 "see tuning(7) and login.conf(5).\n", uid);
377d4740 287 tsleep(&forksleep, 0, "fork", hz / 2);
167e6ecb
MD
288 error = EAGAIN;
289 goto done;
984263bc
MD
290 }
291 /*
292 * Increment the nprocs resource before blocking can occur. There
293 * are hard-limits as to the number of processes that can run.
294 */
295 nprocs++;
296
297 /*
298 * Increment the count of procs running with this uid. Don't allow
299 * a nonprivileged user to exceed their current limit.
300 */
41c20dac 301 ok = chgproccnt(p1->p_ucred->cr_ruidinfo, 1,
984263bc
MD
302 (uid != 0) ? p1->p_rlimit[RLIMIT_NPROC].rlim_cur : 0);
303 if (!ok) {
304 /*
305 * Back out the process count
306 */
307 nprocs--;
5bc7cd8d 308 if (ppsratecheck(&lastfail, &curfail, 1))
6ea70f76 309 kprintf("maxproc limit exceeded by uid %d, please "
5bc7cd8d 310 "see tuning(7) and login.conf(5).\n", uid);
377d4740 311 tsleep(&forksleep, 0, "fork", hz / 2);
167e6ecb
MD
312 error = EAGAIN;
313 goto done;
984263bc
MD
314 }
315
316 /* Allocate new proc. */
51e64ff2 317 p2 = zalloc(proc_zone);
984263bc
MD
318
319 /*
ef09c3ed 320 * Setup linkage for kernel based threading XXX lwp
984263bc 321 */
cb74210d 322 if (flags & RFTHREAD) {
51e64ff2
MD
323 p2->p_peers = p1->p_peers;
324 p1->p_peers = p2;
325 p2->p_leader = p1->p_leader;
984263bc 326 } else {
51e64ff2
MD
327 p2->p_peers = NULL;
328 p2->p_leader = p2;
984263bc
MD
329 }
330
51e64ff2
MD
331 p2->p_wakeup = 0;
332 p2->p_vmspace = NULL;
333 p2->p_numposixlocks = 0;
334 p2->p_emuldata = NULL;
51e64ff2 335 LIST_INIT(&p2->p_lwps);
ef09c3ed 336
13d13d89 337 p2->p_nthreads = 0;
51e64ff2
MD
338 p2->p_nstopped = 0;
339 p2->p_lasttid = 0;
984263bc
MD
340
341 /*
51e64ff2
MD
342 * Setting the state to SIDL protects the partially initialized
343 * process once it starts getting hooked into the rest of the system.
984263bc 344 */
51e64ff2
MD
345 p2->p_stat = SIDL;
346 proc_add_allproc(p2);
984263bc
MD
347
348 /*
349 * Make a proc table entry for the new process.
350 * Start by zeroing the section of proc that is zero-initialized,
351 * then copy the section that is copied directly from the parent.
352 */
353 bzero(&p2->p_startzero,
354 (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero));
355 bcopy(&p1->p_startcopy, &p2->p_startcopy,
356 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
357
358 p2->p_aioinfo = NULL;
359
360 /*
361 * Duplicate sub-structures as needed.
362 * Increase reference counts on shared objects.
3a72e617 363 * p_lock is in the copy area and must be cleared.
984263bc 364 */
344ad853 365 p2->p_flag = 0;
3a72e617 366 p2->p_lock = 0;
13d13d89 367
984263bc
MD
368 if (p1->p_flag & P_PROFIL)
369 startprofclock(p2);
dadab5e9 370 p2->p_ucred = crhold(p1->p_ucred);
984263bc 371
b40e316c 372 if (jailed(p2->p_ucred))
984263bc 373 p2->p_flag |= P_JAILED;
984263bc
MD
374
375 if (p2->p_args)
376 p2->p_args->ar_ref++;
377
13d13d89
SS
378 p2->p_usched = p1->p_usched;
379
984263bc 380 if (flags & RFSIGSHARE) {
b1b4e5a6
SS
381 p2->p_sigacts = p1->p_sigacts;
382 p2->p_sigacts->ps_refcnt++;
984263bc 383 } else {
b1b4e5a6 384 p2->p_sigacts = (struct sigacts *)kmalloc(sizeof(*p2->p_sigacts),
984263bc 385 M_SUBPROC, M_WAITOK);
b1b4e5a6
SS
386 bcopy(p1->p_sigacts, p2->p_sigacts, sizeof(*p2->p_sigacts));
387 p2->p_sigacts->ps_refcnt = 1;
984263bc
MD
388 }
389 if (flags & RFLINUXTHPN)
390 p2->p_sigparent = SIGUSR1;
391 else
392 p2->p_sigparent = SIGCHLD;
393
394 /* bump references to the text vnode (for procfs) */
395 p2->p_textvp = p1->p_textvp;
396 if (p2->p_textvp)
597aea93 397 vref(p2->p_textvp);
984263bc 398
0daa37a5
MD
399 /*
400 * Handle file descriptors
401 */
984263bc
MD
402 if (flags & RFCFDG) {
403 p2->p_fd = fdinit(p1);
404 fdtol = NULL;
405 } else if (flags & RFFDG) {
406 p2->p_fd = fdcopy(p1);
407 fdtol = NULL;
408 } else {
409 p2->p_fd = fdshare(p1);
410 if (p1->p_fdtol == NULL)
411 p1->p_fdtol =
412 filedesc_to_leader_alloc(NULL,
413 p1->p_leader);
414 if ((flags & RFTHREAD) != 0) {
415 /*
416 * Shared file descriptor table and
417 * shared process leaders.
418 */
419 fdtol = p1->p_fdtol;
420 fdtol->fdl_refcount++;
421 } else {
422 /*
423 * Shared file descriptor table, and
424 * different process leaders
425 */
98a7f915 426 fdtol = filedesc_to_leader_alloc(p1->p_fdtol, p2);
984263bc
MD
427 }
428 }
429 p2->p_fdtol = fdtol;
c0b8a06d 430 p2->p_limit = plimit_fork(p1->p_limit);
984263bc
MD
431
432 /*
433 * Preserve some more flags in subprocess. P_PROFIL has already
434 * been preserved.
435 */
08f2f1bb 436 p2->p_flag |= p1->p_flag & P_SUGID;
984263bc
MD
437 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
438 p2->p_flag |= P_CONTROLT;
439 if (flags & RFPPWAIT)
440 p2->p_flag |= P_PPWAIT;
441
0daa37a5
MD
442 /*
443 * Inherit the virtual kernel structure (allows a virtual kernel
444 * to fork to simulate multiple cpus).
445 */
4a22e893
MD
446 p2->p_vkernel = NULL;
447 if (p1->p_vkernel)
448 vkernel_inherit(p1, p2);
0daa37a5 449
5fd012e0
MD
450 /*
451 * Once we are on a pglist we may receive signals. XXX we might
452 * race a ^C being sent to the process group by not receiving it
453 * at all prior to this line.
454 */
984263bc
MD
455 LIST_INSERT_AFTER(p1, p2, p_pglist);
456
457 /*
458 * Attach the new process to its parent.
459 *
460 * If RFNOWAIT is set, the newly created process becomes a child
461 * of init. This effectively disassociates the child from the
462 * parent.
463 */
464 if (flags & RFNOWAIT)
465 pptr = initproc;
466 else
467 pptr = p1;
468 p2->p_pptr = pptr;
469 LIST_INSERT_HEAD(&pptr->p_children, p2, p_sibling);
470 LIST_INIT(&p2->p_children);
98a7f915 471 varsymset_init(&p2->p_varsymset, &p1->p_varsymset);
8fbf9130 472 callout_init(&p2->p_ithandle);
984263bc
MD
473
474#ifdef KTRACE
475 /*
476 * Copy traceflag and tracefile if enabled. If not inherited,
477 * these were zeroed above but we still could have a trace race
29f58392 478 * so make sure p2's p_tracenode is NULL.
984263bc 479 */
29f58392 480 if ((p1->p_traceflag & KTRFAC_INHERIT) && p2->p_tracenode == NULL) {
984263bc 481 p2->p_traceflag = p1->p_traceflag;
29f58392 482 p2->p_tracenode = ktrinherit(p1->p_tracenode);
984263bc
MD
483 }
484#endif
485
984263bc
MD
486 /*
487 * This begins the section where we must prevent the parent
488 * from being swapped.
13d13d89
SS
489 *
490 * Gets PRELE'd in the caller in start_forked_proc().
984263bc
MD
491 */
492 PHOLD(p1);
493
13d13d89
SS
494 vm_fork(p1, p2, flags);
495
984263bc 496 /*
13d13d89
SS
497 * Create the first lwp associated with the new proc.
498 * It will return via a different execution path later, directly
499 * into userland, after it was put on the runq by
500 * start_forked_proc().
984263bc 501 */
13d13d89 502 lwp_fork(lp1, p2, flags);
984263bc
MD
503
504 if (flags == (RFFDG | RFPROC)) {
12e4aaff
MD
505 mycpu->gd_cnt.v_forks++;
506 mycpu->gd_cnt.v_forkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 507 } else if (flags == (RFFDG | RFPROC | RFPPWAIT | RFMEM)) {
12e4aaff
MD
508 mycpu->gd_cnt.v_vforks++;
509 mycpu->gd_cnt.v_vforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 510 } else if (p1 == &proc0) {
12e4aaff
MD
511 mycpu->gd_cnt.v_kthreads++;
512 mycpu->gd_cnt.v_kthreadpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 513 } else {
12e4aaff
MD
514 mycpu->gd_cnt.v_rforks++;
515 mycpu->gd_cnt.v_rforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc
MD
516 }
517
518 /*
519 * Both processes are set up, now check if any loadable modules want
520 * to adjust anything.
521 * What if they have an error? XXX
522 */
523 TAILQ_FOREACH(ep, &fork_list, next) {
524 (*ep->function)(p1, p2, flags);
525 }
526
527 /*
a77ac49d
MD
528 * Set the start time. Note that the process is not runnable. The
529 * caller is responsible for making it runnable.
984263bc 530 */
d9fa5f67 531 microtime(&p2->p_start);
984263bc 532 p2->p_acflag = AFORK;
984263bc
MD
533
534 /*
535 * tell any interested parties about the new process
536 */
537 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid);
538
984263bc
MD
539 /*
540 * Return child proc pointer to parent.
541 */
542 *procp = p2;
167e6ecb
MD
543done:
544 if (pgrp)
545 lockmgr(&pgrp->pg_lock, LK_RELEASE);
546 return (error);
984263bc
MD
547}
548
13d13d89
SS
549static struct lwp *
550lwp_fork(struct lwp *origlp, struct proc *destproc, int flags)
551{
552 struct lwp *lp;
553 struct thread *td;
554 lwpid_t tid;
555
556 /*
557 * We need to prevent wrap-around collisions.
558 * Until we have a nice tid allocator, we need to
559 * start searching for free tids once we wrap around.
560 *
561 * XXX give me a nicer allocator
562 */
563 if (destproc->p_lasttid + 1 <= 0) {
564 tid = 0;
565restart:
566 FOREACH_LWP_IN_PROC(lp, destproc) {
567 if (lp->lwp_tid != tid)
568 continue;
569 /* tids match, search next. */
570 tid++;
571 /*
572 * Wait -- the whole tid space is depleted?
573 * Impossible.
574 */
575 if (tid <= 0)
576 panic("lwp_fork: All tids depleted?!");
577 goto restart;
578 }
579 /* When we come here, the tid is not occupied */
580 } else {
581 tid = destproc->p_lasttid++;
582 }
583
584 lp = zalloc(lwp_zone);
585 lp->lwp_proc = destproc;
586 lp->lwp_tid = tid;
587 LIST_INSERT_HEAD(&destproc->p_lwps, lp, lwp_list);
588 destproc->p_nthreads++;
589 lp->lwp_stat = LSRUN;
590 bzero(&lp->lwp_startzero,
591 (unsigned) ((caddr_t)&lp->lwp_endzero -
592 (caddr_t)&lp->lwp_startzero));
593 bcopy(&origlp->lwp_startcopy, &lp->lwp_startcopy,
594 (unsigned) ((caddr_t)&lp->lwp_endcopy -
595 (caddr_t)&lp->lwp_startcopy));
596 lp->lwp_lock = 0;
597 lp->lwp_flag |= origlp->lwp_flag & LWP_ALTSTACK;
598 /*
599 * Set cpbase to the last timeout that occured (not the upcoming
600 * timeout).
601 *
602 * A critical section is required since a timer IPI can update
603 * scheduler specific data.
604 */
605 crit_enter();
606 lp->lwp_cpbase = mycpu->gd_schedclock.time -
607 mycpu->gd_schedclock.periodic;
608 destproc->p_usched->heuristic_forking(origlp, lp);
609 crit_exit();
610
611 td = lwkt_alloc_thread(NULL, LWKT_THREAD_STACK, -1, 0);
612 lp->lwp_thread = td;
613 td->td_proc = destproc;
614 td->td_lwp = lp;
615 td->td_switch = cpu_heavy_switch;
616#ifdef SMP
617 KKASSERT(td->td_mpcount == 1);
618#endif
619 lwkt_setpri(td, TDPRI_KERN_USER);
620 lwkt_set_comm(td, "%s", destproc->p_comm);
621
622 /*
623 * cpu_fork will copy and update the pcb, set up the kernel stack,
624 * and make the child ready to run.
625 */
626 cpu_fork(origlp, lp, flags);
627 caps_fork(origlp->lwp_thread, lp->lwp_thread);
628
629 return (lp);
630}
631
984263bc
MD
632/*
633 * The next two functionms are general routines to handle adding/deleting
634 * items on the fork callout list.
635 *
636 * at_fork():
637 * Take the arguments given and put them onto the fork callout list,
638 * However first make sure that it's not already there.
639 * Returns 0 on success or a standard error number.
640 */
984263bc 641int
303c76d5 642at_fork(forklist_fn function)
984263bc
MD
643{
644 struct forklist *ep;
645
646#ifdef INVARIANTS
647 /* let the programmer know if he's been stupid */
303c76d5 648 if (rm_at_fork(function)) {
6ea70f76 649 kprintf("WARNING: fork callout entry (%p) already present\n",
984263bc 650 function);
303c76d5 651 }
984263bc 652#endif
efda3bd0 653 ep = kmalloc(sizeof(*ep), M_ATFORK, M_WAITOK|M_ZERO);
984263bc
MD
654 ep->function = function;
655 TAILQ_INSERT_TAIL(&fork_list, ep, next);
656 return (0);
657}
658
659/*
660 * Scan the exit callout list for the given item and remove it..
661 * Returns the number of items removed (0 or 1)
662 */
984263bc 663int
303c76d5 664rm_at_fork(forklist_fn function)
984263bc
MD
665{
666 struct forklist *ep;
667
668 TAILQ_FOREACH(ep, &fork_list, next) {
669 if (ep->function == function) {
670 TAILQ_REMOVE(&fork_list, ep, next);
efda3bd0 671 kfree(ep, M_ATFORK);
984263bc
MD
672 return(1);
673 }
674 }
675 return (0);
676}
7d0bac62
MD
677
678/*
679 * Add a forked process to the run queue after any remaining setup, such
680 * as setting the fork handler, has been completed.
681 */
7d0bac62 682void
553ea3c8 683start_forked_proc(struct lwp *lp1, struct proc *p2)
7d0bac62 684{
08f2f1bb 685 struct lwp *lp2 = ONLY_LWP_IN_PROC(p2);
553ea3c8 686
7d0bac62 687 /*
26a0694b
MD
688 * Move from SIDL to RUN queue, and activate the process's thread.
689 * Activation of the thread effectively makes the process "a"
690 * current process, so we do not setrunqueue().
8ec60c3f
MD
691 *
692 * YYY setrunqueue works here but we should clean up the trampoline
693 * code so we just schedule the LWKT thread and let the trampoline
694 * deal with the userland scheduler on return to userland.
7d0bac62 695 */
553ea3c8 696 KASSERT(p2->p_stat == SIDL,
7d0bac62 697 ("cannot start forked process, bad status: %p", p2));
553ea3c8 698 p2->p_usched->resetpriority(lp2);
e43a034f 699 crit_enter();
164b8401
SS
700 p2->p_stat = SACTIVE;
701 lp2->lwp_stat = LSRUN;
553ea3c8 702 p2->p_usched->setrunqueue(lp2);
e43a034f 703 crit_exit();
7d0bac62
MD
704
705 /*
706 * Now can be swapped.
707 */
553ea3c8 708 PRELE(lp1->lwp_proc);
7d0bac62
MD
709
710 /*
711 * Preserve synchronization semantics of vfork. If waiting for
712 * child to exec or exit, set P_PPWAIT on child, and sleep on our
713 * proc (in case of exit).
714 */
715 while (p2->p_flag & P_PPWAIT)
553ea3c8 716 tsleep(lp1->lwp_proc, 0, "ppwait", 0);
7d0bac62 717}