Migrate allocation of proc structures from zones to kmalloc(). This gives us
[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 $
fb2a331e 40 * $DragonFly: src/sys/kern/kern_fork.c,v 1.72 2008/04/01 18:06:34 nth 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
3e291793
MD
90/*
91 * Red-Black tree support for LWPs
92 */
93
94static int
95rb_lwp_compare(struct lwp *lp1, struct lwp *lp2)
96{
97 if (lp1->lwp_tid < lp2->lwp_tid)
98 return(-1);
99 if (lp1->lwp_tid > lp2->lwp_tid)
100 return(1);
101 return(0);
102}
103
104RB_GENERATE2(lwp_rb_tree, lwp, u.lwp_rbnode, rb_lwp_compare, lwpid_t, lwp_tid);
105
106
984263bc
MD
107/* ARGSUSED */
108int
753fd850 109sys_fork(struct fork_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 | 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
124/* ARGSUSED */
125int
753fd850 126sys_vfork(struct vfork_args *uap)
984263bc 127{
553ea3c8 128 struct lwp *lp = curthread->td_lwp;
984263bc 129 struct proc *p2;
41c20dac 130 int error;
984263bc 131
167e6ecb 132 error = fork1(lp, RFFDG | RFPROC | RFPPWAIT | RFMEM | RFPGLOCK, &p2);
984263bc 133 if (error == 0) {
553ea3c8 134 start_forked_proc(lp, p2);
c7114eea
MD
135 uap->sysmsg_fds[0] = p2->p_pid;
136 uap->sysmsg_fds[1] = 0;
984263bc
MD
137 }
138 return error;
139}
140
f61c1ff1
MD
141/*
142 * Handle rforks. An rfork may (1) operate on the current process without
143 * creating a new, (2) create a new process that shared the current process's
144 * vmspace, signals, and/or descriptors, or (3) create a new process that does
145 * not share these things (normal fork).
146 *
147 * Note that we only call start_forked_proc() if a new process is actually
148 * created.
149 *
150 * rfork { int flags }
151 */
984263bc 152int
753fd850 153sys_rfork(struct rfork_args *uap)
984263bc 154{
553ea3c8 155 struct lwp *lp = curthread->td_lwp;
984263bc 156 struct proc *p2;
41c20dac 157 int error;
984263bc 158
6654fbcb
MD
159 if ((uap->flags & RFKERNELONLY) != 0)
160 return (EINVAL);
161
167e6ecb 162 error = fork1(lp, uap->flags | RFPGLOCK, &p2);
984263bc 163 if (error == 0) {
f61c1ff1 164 if (p2)
553ea3c8 165 start_forked_proc(lp, p2);
c7114eea
MD
166 uap->sysmsg_fds[0] = p2 ? p2->p_pid : 0;
167 uap->sysmsg_fds[1] = 0;
984263bc
MD
168 }
169 return error;
170}
171
91bd9c1e
SS
172int
173sys_lwp_create(struct lwp_create_args *uap)
174{
175 struct proc *p = curproc;
176 struct lwp *lp;
177 struct lwp_params params;
178 int error;
179
180 error = copyin(uap->params, &params, sizeof(params));
181 if (error)
182 goto fail2;
183
184 lp = lwp_fork(curthread->td_lwp, p, RFPROC);
185 error = cpu_prepare_lwp(lp, &params);
186 if (params.tid1 != NULL &&
187 (error = copyout(&lp->lwp_tid, params.tid1, sizeof(lp->lwp_tid))))
188 goto fail;
189 if (params.tid2 != NULL &&
190 (error = copyout(&lp->lwp_tid, params.tid2, sizeof(lp->lwp_tid))))
191 goto fail;
192
193 /*
194 * Now schedule the new lwp.
195 */
196 p->p_usched->resetpriority(lp);
197 crit_enter();
198 lp->lwp_stat = LSRUN;
199 p->p_usched->setrunqueue(lp);
200 crit_exit();
201
202 return (0);
203
204fail:
205 --p->p_nthreads;
3e291793 206 lwp_rb_tree_RB_REMOVE(&p->p_lwp_tree, lp);
e3161323
MD
207 /* lwp_dispose expects an exited lwp, and a held proc */
208 lp->lwp_flag |= LWP_WEXIT;
209 lp->lwp_thread->td_flags |= TDF_EXITING;
210 PHOLD(p);
91bd9c1e
SS
211 lwp_dispose(lp);
212fail2:
213 return (error);
214}
984263bc
MD
215
216int nprocs = 1; /* process 0 */
984263bc
MD
217
218int
553ea3c8 219fork1(struct lwp *lp1, int flags, struct proc **procp)
984263bc 220{
553ea3c8 221 struct proc *p1 = lp1->lwp_proc;
984263bc 222 struct proc *p2, *pptr;
167e6ecb 223 struct pgrp *pgrp;
984263bc 224 uid_t uid;
167e6ecb 225 int ok, error;
51e64ff2 226 static int curfail = 0;
5bc7cd8d 227 static struct timeval lastfail;
984263bc
MD
228 struct forklist *ep;
229 struct filedesc_to_leader *fdtol;
230
231 if ((flags & (RFFDG|RFCFDG)) == (RFFDG|RFCFDG))
232 return (EINVAL);
233
234 /*
235 * Here we don't create a new process, but we divorce
236 * certain parts of a process from itself.
237 */
238 if ((flags & RFPROC) == 0) {
13d13d89
SS
239 /*
240 * This kind of stunt does not work anymore if
241 * there are native threads (lwps) running
242 */
243 if (p1->p_nthreads != 1)
244 return (EINVAL);
245
246 vm_fork(p1, 0, flags);
984263bc
MD
247
248 /*
249 * Close all file descriptors.
250 */
251 if (flags & RFCFDG) {
252 struct filedesc *fdtmp;
253 fdtmp = fdinit(p1);
254 fdfree(p1);
255 p1->p_fd = fdtmp;
256 }
257
258 /*
259 * Unshare file descriptors (from parent.)
260 */
261 if (flags & RFFDG) {
262 if (p1->p_fd->fd_refcnt > 1) {
263 struct filedesc *newfd;
264 newfd = fdcopy(p1);
265 fdfree(p1);
266 p1->p_fd = newfd;
267 }
268 }
269 *procp = NULL;
270 return (0);
271 }
272
167e6ecb
MD
273 /*
274 * Interlock against process group signal delivery. If signals
275 * are pending after the interlock is obtained we have to restart
276 * the system call to process the signals. If we don't the child
277 * can miss a pgsignal (such as ^C) sent during the fork.
278 *
279 * We can't use CURSIG() here because it will process any STOPs
280 * and cause the process group lock to be held indefinitely. If
281 * a STOP occurs, the fork will be restarted after the CONT.
282 */
283 error = 0;
284 pgrp = NULL;
285 if ((flags & RFPGLOCK) && (pgrp = p1->p_pgrp) != NULL) {
286 lockmgr(&pgrp->pg_lock, LK_SHARED);
08f2f1bb 287 if (CURSIGNB(lp1)) {
167e6ecb
MD
288 error = ERESTART;
289 goto done;
290 }
291 }
292
984263bc
MD
293 /*
294 * Although process entries are dynamically created, we still keep
295 * a global limit on the maximum number we will create. Don't allow
296 * a nonprivileged user to use the last ten processes; don't let root
297 * exceed the limit. The variable nprocs is the current number of
298 * processes, maxproc is the limit.
299 */
41c20dac 300 uid = p1->p_ucred->cr_ruid;
984263bc 301 if ((nprocs >= maxproc - 10 && uid != 0) || nprocs >= maxproc) {
5bc7cd8d 302 if (ppsratecheck(&lastfail, &curfail, 1))
6ea70f76 303 kprintf("maxproc limit exceeded by uid %d, please "
5bc7cd8d 304 "see tuning(7) and login.conf(5).\n", uid);
377d4740 305 tsleep(&forksleep, 0, "fork", hz / 2);
167e6ecb
MD
306 error = EAGAIN;
307 goto done;
984263bc
MD
308 }
309 /*
310 * Increment the nprocs resource before blocking can occur. There
311 * are hard-limits as to the number of processes that can run.
312 */
313 nprocs++;
314
315 /*
316 * Increment the count of procs running with this uid. Don't allow
317 * a nonprivileged user to exceed their current limit.
318 */
41c20dac 319 ok = chgproccnt(p1->p_ucred->cr_ruidinfo, 1,
984263bc
MD
320 (uid != 0) ? p1->p_rlimit[RLIMIT_NPROC].rlim_cur : 0);
321 if (!ok) {
322 /*
323 * Back out the process count
324 */
325 nprocs--;
5bc7cd8d 326 if (ppsratecheck(&lastfail, &curfail, 1))
6ea70f76 327 kprintf("maxproc limit exceeded by uid %d, please "
5bc7cd8d 328 "see tuning(7) and login.conf(5).\n", uid);
377d4740 329 tsleep(&forksleep, 0, "fork", hz / 2);
167e6ecb
MD
330 error = EAGAIN;
331 goto done;
984263bc
MD
332 }
333
334 /* Allocate new proc. */
fb2a331e 335 p2 = kmalloc(sizeof(struct proc), M_PROC, M_WAITOK);
f9fe34b6 336 bzero(p2, sizeof(*p2));
984263bc
MD
337
338 /*
ef09c3ed 339 * Setup linkage for kernel based threading XXX lwp
984263bc 340 */
cb74210d 341 if (flags & RFTHREAD) {
51e64ff2
MD
342 p2->p_peers = p1->p_peers;
343 p1->p_peers = p2;
344 p2->p_leader = p1->p_leader;
984263bc 345 } else {
51e64ff2 346 p2->p_leader = p2;
984263bc
MD
347 }
348
3e291793
MD
349 RB_INIT(&p2->p_lwp_tree);
350 p2->p_lasttid = -1; /* first tid will be 0 */
ef09c3ed 351
984263bc 352 /*
51e64ff2
MD
353 * Setting the state to SIDL protects the partially initialized
354 * process once it starts getting hooked into the rest of the system.
984263bc 355 */
51e64ff2
MD
356 p2->p_stat = SIDL;
357 proc_add_allproc(p2);
984263bc
MD
358
359 /*
360 * Make a proc table entry for the new process.
f9fe34b6
SS
361 * The whole structure was zeroed above, so copy the section that is
362 * copied directly from the parent.
984263bc 363 */
984263bc
MD
364 bcopy(&p1->p_startcopy, &p2->p_startcopy,
365 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
366
984263bc
MD
367 /*
368 * Duplicate sub-structures as needed.
369 * Increase reference counts on shared objects.
984263bc 370 */
984263bc
MD
371 if (p1->p_flag & P_PROFIL)
372 startprofclock(p2);
dadab5e9 373 p2->p_ucred = crhold(p1->p_ucred);
984263bc 374
b40e316c 375 if (jailed(p2->p_ucred))
984263bc 376 p2->p_flag |= P_JAILED;
984263bc
MD
377
378 if (p2->p_args)
379 p2->p_args->ar_ref++;
380
13d13d89
SS
381 p2->p_usched = p1->p_usched;
382
984263bc 383 if (flags & RFSIGSHARE) {
b1b4e5a6
SS
384 p2->p_sigacts = p1->p_sigacts;
385 p2->p_sigacts->ps_refcnt++;
984263bc 386 } else {
b1b4e5a6 387 p2->p_sigacts = (struct sigacts *)kmalloc(sizeof(*p2->p_sigacts),
984263bc 388 M_SUBPROC, M_WAITOK);
b1b4e5a6
SS
389 bcopy(p1->p_sigacts, p2->p_sigacts, sizeof(*p2->p_sigacts));
390 p2->p_sigacts->ps_refcnt = 1;
984263bc
MD
391 }
392 if (flags & RFLINUXTHPN)
393 p2->p_sigparent = SIGUSR1;
394 else
395 p2->p_sigparent = SIGCHLD;
396
397 /* bump references to the text vnode (for procfs) */
398 p2->p_textvp = p1->p_textvp;
399 if (p2->p_textvp)
597aea93 400 vref(p2->p_textvp);
984263bc 401
0daa37a5
MD
402 /*
403 * Handle file descriptors
404 */
984263bc
MD
405 if (flags & RFCFDG) {
406 p2->p_fd = fdinit(p1);
407 fdtol = NULL;
408 } else if (flags & RFFDG) {
409 p2->p_fd = fdcopy(p1);
410 fdtol = NULL;
411 } else {
412 p2->p_fd = fdshare(p1);
413 if (p1->p_fdtol == NULL)
414 p1->p_fdtol =
415 filedesc_to_leader_alloc(NULL,
416 p1->p_leader);
417 if ((flags & RFTHREAD) != 0) {
418 /*
419 * Shared file descriptor table and
420 * shared process leaders.
421 */
422 fdtol = p1->p_fdtol;
423 fdtol->fdl_refcount++;
424 } else {
425 /*
426 * Shared file descriptor table, and
427 * different process leaders
428 */
98a7f915 429 fdtol = filedesc_to_leader_alloc(p1->p_fdtol, p2);
984263bc
MD
430 }
431 }
432 p2->p_fdtol = fdtol;
c0b8a06d 433 p2->p_limit = plimit_fork(p1->p_limit);
984263bc
MD
434
435 /*
436 * Preserve some more flags in subprocess. P_PROFIL has already
437 * been preserved.
438 */
08f2f1bb 439 p2->p_flag |= p1->p_flag & P_SUGID;
984263bc
MD
440 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
441 p2->p_flag |= P_CONTROLT;
442 if (flags & RFPPWAIT)
443 p2->p_flag |= P_PPWAIT;
444
0daa37a5
MD
445 /*
446 * Inherit the virtual kernel structure (allows a virtual kernel
447 * to fork to simulate multiple cpus).
448 */
4a22e893
MD
449 if (p1->p_vkernel)
450 vkernel_inherit(p1, p2);
0daa37a5 451
5fd012e0
MD
452 /*
453 * Once we are on a pglist we may receive signals. XXX we might
454 * race a ^C being sent to the process group by not receiving it
455 * at all prior to this line.
456 */
984263bc
MD
457 LIST_INSERT_AFTER(p1, p2, p_pglist);
458
459 /*
460 * Attach the new process to its parent.
461 *
462 * If RFNOWAIT is set, the newly created process becomes a child
463 * of init. This effectively disassociates the child from the
464 * parent.
465 */
466 if (flags & RFNOWAIT)
467 pptr = initproc;
468 else
469 pptr = p1;
470 p2->p_pptr = pptr;
471 LIST_INSERT_HEAD(&pptr->p_children, p2, p_sibling);
472 LIST_INIT(&p2->p_children);
98a7f915 473 varsymset_init(&p2->p_varsymset, &p1->p_varsymset);
8fbf9130 474 callout_init(&p2->p_ithandle);
984263bc
MD
475
476#ifdef KTRACE
477 /*
478 * Copy traceflag and tracefile if enabled. If not inherited,
479 * these were zeroed above but we still could have a trace race
29f58392 480 * so make sure p2's p_tracenode is NULL.
984263bc 481 */
29f58392 482 if ((p1->p_traceflag & KTRFAC_INHERIT) && p2->p_tracenode == NULL) {
984263bc 483 p2->p_traceflag = p1->p_traceflag;
29f58392 484 p2->p_tracenode = ktrinherit(p1->p_tracenode);
984263bc
MD
485 }
486#endif
487
984263bc
MD
488 /*
489 * This begins the section where we must prevent the parent
490 * from being swapped.
13d13d89
SS
491 *
492 * Gets PRELE'd in the caller in start_forked_proc().
984263bc
MD
493 */
494 PHOLD(p1);
495
13d13d89
SS
496 vm_fork(p1, p2, flags);
497
984263bc 498 /*
13d13d89
SS
499 * Create the first lwp associated with the new proc.
500 * It will return via a different execution path later, directly
501 * into userland, after it was put on the runq by
502 * start_forked_proc().
984263bc 503 */
13d13d89 504 lwp_fork(lp1, p2, flags);
984263bc
MD
505
506 if (flags == (RFFDG | RFPROC)) {
12e4aaff
MD
507 mycpu->gd_cnt.v_forks++;
508 mycpu->gd_cnt.v_forkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 509 } else if (flags == (RFFDG | RFPROC | RFPPWAIT | RFMEM)) {
12e4aaff
MD
510 mycpu->gd_cnt.v_vforks++;
511 mycpu->gd_cnt.v_vforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 512 } else if (p1 == &proc0) {
12e4aaff
MD
513 mycpu->gd_cnt.v_kthreads++;
514 mycpu->gd_cnt.v_kthreadpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 515 } else {
12e4aaff
MD
516 mycpu->gd_cnt.v_rforks++;
517 mycpu->gd_cnt.v_rforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc
MD
518 }
519
520 /*
521 * Both processes are set up, now check if any loadable modules want
522 * to adjust anything.
523 * What if they have an error? XXX
524 */
525 TAILQ_FOREACH(ep, &fork_list, next) {
526 (*ep->function)(p1, p2, flags);
527 }
528
529 /*
a77ac49d
MD
530 * Set the start time. Note that the process is not runnable. The
531 * caller is responsible for making it runnable.
984263bc 532 */
d9fa5f67 533 microtime(&p2->p_start);
984263bc 534 p2->p_acflag = AFORK;
984263bc
MD
535
536 /*
537 * tell any interested parties about the new process
538 */
539 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid);
540
984263bc
MD
541 /*
542 * Return child proc pointer to parent.
543 */
544 *procp = p2;
167e6ecb
MD
545done:
546 if (pgrp)
547 lockmgr(&pgrp->pg_lock, LK_RELEASE);
548 return (error);
984263bc
MD
549}
550
13d13d89
SS
551static struct lwp *
552lwp_fork(struct lwp *origlp, struct proc *destproc, int flags)
553{
554 struct lwp *lp;
555 struct thread *td;
13d13d89
SS
556
557 lp = zalloc(lwp_zone);
f9fe34b6 558 bzero(lp, sizeof(*lp));
3e291793 559
13d13d89 560 lp->lwp_proc = destproc;
287ebb09 561 lp->lwp_vmspace = destproc->p_vmspace;
13d13d89 562 lp->lwp_stat = LSRUN;
13d13d89
SS
563 bcopy(&origlp->lwp_startcopy, &lp->lwp_startcopy,
564 (unsigned) ((caddr_t)&lp->lwp_endcopy -
565 (caddr_t)&lp->lwp_startcopy));
13d13d89
SS
566 lp->lwp_flag |= origlp->lwp_flag & LWP_ALTSTACK;
567 /*
568 * Set cpbase to the last timeout that occured (not the upcoming
569 * timeout).
570 *
571 * A critical section is required since a timer IPI can update
572 * scheduler specific data.
573 */
574 crit_enter();
575 lp->lwp_cpbase = mycpu->gd_schedclock.time -
576 mycpu->gd_schedclock.periodic;
577 destproc->p_usched->heuristic_forking(origlp, lp);
578 crit_exit();
a5ae2446 579 lp->lwp_cpumask &= usched_mastermask;
13d13d89 580
3e291793
MD
581 /*
582 * Assign a TID to the lp. Loop until the insert succeeds (returns
583 * NULL).
584 */
585 lp->lwp_tid = destproc->p_lasttid;
586 do {
587 if (++lp->lwp_tid < 0)
588 lp->lwp_tid = 1;
589 } while (lwp_rb_tree_RB_INSERT(&destproc->p_lwp_tree, lp) != NULL);
590 destproc->p_lasttid = lp->lwp_tid;
591 destproc->p_nthreads++;
592
13d13d89
SS
593 td = lwkt_alloc_thread(NULL, LWKT_THREAD_STACK, -1, 0);
594 lp->lwp_thread = td;
595 td->td_proc = destproc;
596 td->td_lwp = lp;
597 td->td_switch = cpu_heavy_switch;
598#ifdef SMP
599 KKASSERT(td->td_mpcount == 1);
600#endif
601 lwkt_setpri(td, TDPRI_KERN_USER);
602 lwkt_set_comm(td, "%s", destproc->p_comm);
603
604 /*
605 * cpu_fork will copy and update the pcb, set up the kernel stack,
606 * and make the child ready to run.
607 */
608 cpu_fork(origlp, lp, flags);
609 caps_fork(origlp->lwp_thread, lp->lwp_thread);
610
611 return (lp);
612}
613
984263bc
MD
614/*
615 * The next two functionms are general routines to handle adding/deleting
616 * items on the fork callout list.
617 *
618 * at_fork():
619 * Take the arguments given and put them onto the fork callout list,
620 * However first make sure that it's not already there.
621 * Returns 0 on success or a standard error number.
622 */
984263bc 623int
303c76d5 624at_fork(forklist_fn function)
984263bc
MD
625{
626 struct forklist *ep;
627
628#ifdef INVARIANTS
629 /* let the programmer know if he's been stupid */
303c76d5 630 if (rm_at_fork(function)) {
6ea70f76 631 kprintf("WARNING: fork callout entry (%p) already present\n",
984263bc 632 function);
303c76d5 633 }
984263bc 634#endif
efda3bd0 635 ep = kmalloc(sizeof(*ep), M_ATFORK, M_WAITOK|M_ZERO);
984263bc
MD
636 ep->function = function;
637 TAILQ_INSERT_TAIL(&fork_list, ep, next);
638 return (0);
639}
640
641/*
642 * Scan the exit callout list for the given item and remove it..
643 * Returns the number of items removed (0 or 1)
644 */
984263bc 645int
303c76d5 646rm_at_fork(forklist_fn function)
984263bc
MD
647{
648 struct forklist *ep;
649
650 TAILQ_FOREACH(ep, &fork_list, next) {
651 if (ep->function == function) {
652 TAILQ_REMOVE(&fork_list, ep, next);
efda3bd0 653 kfree(ep, M_ATFORK);
984263bc
MD
654 return(1);
655 }
656 }
657 return (0);
658}
7d0bac62
MD
659
660/*
661 * Add a forked process to the run queue after any remaining setup, such
662 * as setting the fork handler, has been completed.
663 */
7d0bac62 664void
553ea3c8 665start_forked_proc(struct lwp *lp1, struct proc *p2)
7d0bac62 666{
08f2f1bb 667 struct lwp *lp2 = ONLY_LWP_IN_PROC(p2);
553ea3c8 668
7d0bac62 669 /*
26a0694b
MD
670 * Move from SIDL to RUN queue, and activate the process's thread.
671 * Activation of the thread effectively makes the process "a"
672 * current process, so we do not setrunqueue().
8ec60c3f
MD
673 *
674 * YYY setrunqueue works here but we should clean up the trampoline
675 * code so we just schedule the LWKT thread and let the trampoline
676 * deal with the userland scheduler on return to userland.
7d0bac62 677 */
553ea3c8 678 KASSERT(p2->p_stat == SIDL,
7d0bac62 679 ("cannot start forked process, bad status: %p", p2));
553ea3c8 680 p2->p_usched->resetpriority(lp2);
e43a034f 681 crit_enter();
164b8401
SS
682 p2->p_stat = SACTIVE;
683 lp2->lwp_stat = LSRUN;
553ea3c8 684 p2->p_usched->setrunqueue(lp2);
e43a034f 685 crit_exit();
7d0bac62
MD
686
687 /*
688 * Now can be swapped.
689 */
553ea3c8 690 PRELE(lp1->lwp_proc);
7d0bac62
MD
691
692 /*
693 * Preserve synchronization semantics of vfork. If waiting for
694 * child to exec or exit, set P_PPWAIT on child, and sleep on our
695 * proc (in case of exit).
696 */
697 while (p2->p_flag & P_PPWAIT)
553ea3c8 698 tsleep(lp1->lwp_proc, 0, "ppwait", 0);
7d0bac62 699}