kernel - Add per-process token, adjust signal code to use it.
[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 $
f6c36234 40 * $DragonFly: src/sys/kern/kern_fork.c,v 1.77 2008/05/18 20:02:02 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>
984263bc
MD
66
67#include <sys/vmmeter.h>
e43a034f 68#include <sys/thread2.h>
b1b4e5a6 69#include <sys/signal2.h>
8f1f6170 70#include <sys/spinlock2.h>
684a93c4 71#include <sys/mplock2.h>
984263bc 72
8c72e3d5
AH
73#include <sys/dsched.h>
74
984263bc
MD
75static MALLOC_DEFINE(M_ATFORK, "atfork", "atfork callback");
76
77/*
78 * These are the stuctures used to create a callout list for things to do
79 * when forking a process
80 */
81struct forklist {
82 forklist_fn function;
83 TAILQ_ENTRY(forklist) next;
84};
85
86TAILQ_HEAD(forklist_head, forklist);
87static struct forklist_head fork_list = TAILQ_HEAD_INITIALIZER(fork_list);
88
13d13d89
SS
89static struct lwp *lwp_fork(struct lwp *, struct proc *, int flags);
90
984263bc
MD
91int forksleep; /* Place for fork1() to sleep on. */
92
3e291793
MD
93/*
94 * Red-Black tree support for LWPs
95 */
96
97static int
98rb_lwp_compare(struct lwp *lp1, struct lwp *lp2)
99{
100 if (lp1->lwp_tid < lp2->lwp_tid)
101 return(-1);
102 if (lp1->lwp_tid > lp2->lwp_tid)
103 return(1);
104 return(0);
105}
106
107RB_GENERATE2(lwp_rb_tree, lwp, u.lwp_rbnode, rb_lwp_compare, lwpid_t, lwp_tid);
108
3919ced0
MD
109/*
110 * Fork system call
111 *
112 * MPALMOSTSAFE
113 */
984263bc 114int
753fd850 115sys_fork(struct fork_args *uap)
984263bc 116{
553ea3c8 117 struct lwp *lp = curthread->td_lwp;
984263bc 118 struct proc *p2;
41c20dac 119 int error;
984263bc 120
3919ced0 121 get_mplock();
167e6ecb 122 error = fork1(lp, RFFDG | RFPROC | RFPGLOCK, &p2);
984263bc 123 if (error == 0) {
553ea3c8 124 start_forked_proc(lp, p2);
c7114eea
MD
125 uap->sysmsg_fds[0] = p2->p_pid;
126 uap->sysmsg_fds[1] = 0;
984263bc 127 }
3919ced0 128 rel_mplock();
984263bc
MD
129 return error;
130}
131
3919ced0
MD
132/*
133 * MPALMOSTSAFE
134 */
984263bc 135int
753fd850 136sys_vfork(struct vfork_args *uap)
984263bc 137{
553ea3c8 138 struct lwp *lp = curthread->td_lwp;
984263bc 139 struct proc *p2;
41c20dac 140 int error;
984263bc 141
3919ced0 142 get_mplock();
167e6ecb 143 error = fork1(lp, RFFDG | RFPROC | RFPPWAIT | RFMEM | RFPGLOCK, &p2);
984263bc 144 if (error == 0) {
553ea3c8 145 start_forked_proc(lp, p2);
c7114eea
MD
146 uap->sysmsg_fds[0] = p2->p_pid;
147 uap->sysmsg_fds[1] = 0;
984263bc 148 }
3919ced0 149 rel_mplock();
984263bc
MD
150 return error;
151}
152
f61c1ff1
MD
153/*
154 * Handle rforks. An rfork may (1) operate on the current process without
155 * creating a new, (2) create a new process that shared the current process's
156 * vmspace, signals, and/or descriptors, or (3) create a new process that does
157 * not share these things (normal fork).
158 *
159 * Note that we only call start_forked_proc() if a new process is actually
160 * created.
161 *
162 * rfork { int flags }
3919ced0
MD
163 *
164 * MPALMOSTSAFE
f61c1ff1 165 */
984263bc 166int
753fd850 167sys_rfork(struct rfork_args *uap)
984263bc 168{
553ea3c8 169 struct lwp *lp = curthread->td_lwp;
984263bc 170 struct proc *p2;
41c20dac 171 int error;
984263bc 172
6654fbcb
MD
173 if ((uap->flags & RFKERNELONLY) != 0)
174 return (EINVAL);
175
3919ced0 176 get_mplock();
167e6ecb 177 error = fork1(lp, uap->flags | RFPGLOCK, &p2);
984263bc 178 if (error == 0) {
f61c1ff1 179 if (p2)
553ea3c8 180 start_forked_proc(lp, p2);
c7114eea
MD
181 uap->sysmsg_fds[0] = p2 ? p2->p_pid : 0;
182 uap->sysmsg_fds[1] = 0;
984263bc 183 }
3919ced0 184 rel_mplock();
984263bc
MD
185 return error;
186}
187
3919ced0
MD
188/*
189 * MPALMOSTSAFE
190 */
91bd9c1e
SS
191int
192sys_lwp_create(struct lwp_create_args *uap)
193{
194 struct proc *p = curproc;
195 struct lwp *lp;
196 struct lwp_params params;
197 int error;
198
199 error = copyin(uap->params, &params, sizeof(params));
200 if (error)
201 goto fail2;
202
3919ced0 203 get_mplock();
8f1f6170 204 plimit_lwp_fork(p); /* force exclusive access */
91bd9c1e
SS
205 lp = lwp_fork(curthread->td_lwp, p, RFPROC);
206 error = cpu_prepare_lwp(lp, &params);
207 if (params.tid1 != NULL &&
208 (error = copyout(&lp->lwp_tid, params.tid1, sizeof(lp->lwp_tid))))
209 goto fail;
210 if (params.tid2 != NULL &&
211 (error = copyout(&lp->lwp_tid, params.tid2, sizeof(lp->lwp_tid))))
212 goto fail;
213
214 /*
8f1f6170 215 * Now schedule the new lwp.
91bd9c1e
SS
216 */
217 p->p_usched->resetpriority(lp);
218 crit_enter();
219 lp->lwp_stat = LSRUN;
220 p->p_usched->setrunqueue(lp);
221 crit_exit();
3919ced0 222 rel_mplock();
91bd9c1e
SS
223
224 return (0);
225
226fail:
3e291793 227 lwp_rb_tree_RB_REMOVE(&p->p_lwp_tree, lp);
0b26dde3 228 --p->p_nthreads;
e3161323
MD
229 /* lwp_dispose expects an exited lwp, and a held proc */
230 lp->lwp_flag |= LWP_WEXIT;
231 lp->lwp_thread->td_flags |= TDF_EXITING;
232 PHOLD(p);
91bd9c1e 233 lwp_dispose(lp);
3919ced0 234 rel_mplock();
91bd9c1e
SS
235fail2:
236 return (error);
237}
984263bc
MD
238
239int nprocs = 1; /* process 0 */
984263bc
MD
240
241int
553ea3c8 242fork1(struct lwp *lp1, int flags, struct proc **procp)
984263bc 243{
553ea3c8 244 struct proc *p1 = lp1->lwp_proc;
984263bc 245 struct proc *p2, *pptr;
167e6ecb 246 struct pgrp *pgrp;
984263bc 247 uid_t uid;
167e6ecb 248 int ok, error;
51e64ff2 249 static int curfail = 0;
5bc7cd8d 250 static struct timeval lastfail;
984263bc
MD
251 struct forklist *ep;
252 struct filedesc_to_leader *fdtol;
253
254 if ((flags & (RFFDG|RFCFDG)) == (RFFDG|RFCFDG))
255 return (EINVAL);
256
257 /*
258 * Here we don't create a new process, but we divorce
259 * certain parts of a process from itself.
260 */
261 if ((flags & RFPROC) == 0) {
13d13d89
SS
262 /*
263 * This kind of stunt does not work anymore if
264 * there are native threads (lwps) running
265 */
266 if (p1->p_nthreads != 1)
267 return (EINVAL);
268
269 vm_fork(p1, 0, flags);
984263bc
MD
270
271 /*
272 * Close all file descriptors.
273 */
274 if (flags & RFCFDG) {
275 struct filedesc *fdtmp;
276 fdtmp = fdinit(p1);
0a4a9c77 277 fdfree(p1, fdtmp);
984263bc
MD
278 }
279
280 /*
281 * Unshare file descriptors (from parent.)
282 */
283 if (flags & RFFDG) {
284 if (p1->p_fd->fd_refcnt > 1) {
285 struct filedesc *newfd;
286 newfd = fdcopy(p1);
0a4a9c77 287 fdfree(p1, newfd);
984263bc
MD
288 }
289 }
290 *procp = NULL;
291 return (0);
292 }
293
294 /*
167e6ecb
MD
295 * Interlock against process group signal delivery. If signals
296 * are pending after the interlock is obtained we have to restart
297 * the system call to process the signals. If we don't the child
298 * can miss a pgsignal (such as ^C) sent during the fork.
299 *
300 * We can't use CURSIG() here because it will process any STOPs
301 * and cause the process group lock to be held indefinitely. If
302 * a STOP occurs, the fork will be restarted after the CONT.
303 */
304 error = 0;
305 pgrp = NULL;
306 if ((flags & RFPGLOCK) && (pgrp = p1->p_pgrp) != NULL) {
307 lockmgr(&pgrp->pg_lock, LK_SHARED);
f6e73860 308 if (CURSIG_NOBLOCK(lp1)) {
167e6ecb
MD
309 error = ERESTART;
310 goto done;
311 }
312 }
313
314 /*
984263bc
MD
315 * Although process entries are dynamically created, we still keep
316 * a global limit on the maximum number we will create. Don't allow
317 * a nonprivileged user to use the last ten processes; don't let root
318 * exceed the limit. The variable nprocs is the current number of
319 * processes, maxproc is the limit.
320 */
9910d07b 321 uid = lp1->lwp_thread->td_ucred->cr_ruid;
984263bc 322 if ((nprocs >= maxproc - 10 && uid != 0) || nprocs >= maxproc) {
5bc7cd8d 323 if (ppsratecheck(&lastfail, &curfail, 1))
6ea70f76 324 kprintf("maxproc limit exceeded by uid %d, please "
5bc7cd8d 325 "see tuning(7) and login.conf(5).\n", uid);
377d4740 326 tsleep(&forksleep, 0, "fork", hz / 2);
167e6ecb
MD
327 error = EAGAIN;
328 goto done;
984263bc
MD
329 }
330 /*
331 * Increment the nprocs resource before blocking can occur. There
332 * are hard-limits as to the number of processes that can run.
333 */
334 nprocs++;
335
336 /*
337 * Increment the count of procs running with this uid. Don't allow
338 * a nonprivileged user to exceed their current limit.
339 */
9910d07b 340 ok = chgproccnt(lp1->lwp_thread->td_ucred->cr_ruidinfo, 1,
984263bc
MD
341 (uid != 0) ? p1->p_rlimit[RLIMIT_NPROC].rlim_cur : 0);
342 if (!ok) {
343 /*
344 * Back out the process count
345 */
346 nprocs--;
5bc7cd8d 347 if (ppsratecheck(&lastfail, &curfail, 1))
6ea70f76 348 kprintf("maxproc limit exceeded by uid %d, please "
5bc7cd8d 349 "see tuning(7) and login.conf(5).\n", uid);
377d4740 350 tsleep(&forksleep, 0, "fork", hz / 2);
167e6ecb
MD
351 error = EAGAIN;
352 goto done;
984263bc
MD
353 }
354
355 /* Allocate new proc. */
37733243 356 p2 = kmalloc(sizeof(struct proc), M_PROC, M_WAITOK|M_ZERO);
984263bc
MD
357
358 /*
ef09c3ed 359 * Setup linkage for kernel based threading XXX lwp
984263bc 360 */
cb74210d 361 if (flags & RFTHREAD) {
51e64ff2
MD
362 p2->p_peers = p1->p_peers;
363 p1->p_peers = p2;
364 p2->p_leader = p1->p_leader;
984263bc 365 } else {
51e64ff2 366 p2->p_leader = p2;
984263bc
MD
367 }
368
3e291793 369 RB_INIT(&p2->p_lwp_tree);
8f1f6170 370 spin_init(&p2->p_spin);
5686ec5a 371 lwkt_token_init(&p2->p_token, "iproc");
3e291793 372 p2->p_lasttid = -1; /* first tid will be 0 */
ef09c3ed 373
984263bc 374 /*
51e64ff2
MD
375 * Setting the state to SIDL protects the partially initialized
376 * process once it starts getting hooked into the rest of the system.
984263bc 377 */
51e64ff2
MD
378 p2->p_stat = SIDL;
379 proc_add_allproc(p2);
984263bc
MD
380
381 /*
382 * Make a proc table entry for the new process.
f9fe34b6
SS
383 * The whole structure was zeroed above, so copy the section that is
384 * copied directly from the parent.
984263bc 385 */
984263bc
MD
386 bcopy(&p1->p_startcopy, &p2->p_startcopy,
387 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
388
984263bc
MD
389 /*
390 * Duplicate sub-structures as needed.
391 * Increase reference counts on shared objects.
984263bc 392 */
984263bc
MD
393 if (p1->p_flag & P_PROFIL)
394 startprofclock(p2);
9910d07b 395 p2->p_ucred = crhold(lp1->lwp_thread->td_ucred);
0d355d3b 396 KKASSERT(p2->p_lock == 0);
984263bc 397
b40e316c 398 if (jailed(p2->p_ucred))
984263bc 399 p2->p_flag |= P_JAILED;
984263bc
MD
400
401 if (p2->p_args)
402 p2->p_args->ar_ref++;
403
13d13d89 404 p2->p_usched = p1->p_usched;
8c72e3d5
AH
405 /* XXX: verify copy of the secondary iosched stuff */
406 dsched_new_proc(p2);
13d13d89 407
984263bc 408 if (flags & RFSIGSHARE) {
b1b4e5a6
SS
409 p2->p_sigacts = p1->p_sigacts;
410 p2->p_sigacts->ps_refcnt++;
984263bc 411 } else {
b1b4e5a6 412 p2->p_sigacts = (struct sigacts *)kmalloc(sizeof(*p2->p_sigacts),
984263bc 413 M_SUBPROC, M_WAITOK);
b1b4e5a6
SS
414 bcopy(p1->p_sigacts, p2->p_sigacts, sizeof(*p2->p_sigacts));
415 p2->p_sigacts->ps_refcnt = 1;
984263bc
MD
416 }
417 if (flags & RFLINUXTHPN)
418 p2->p_sigparent = SIGUSR1;
419 else
420 p2->p_sigparent = SIGCHLD;
421
422 /* bump references to the text vnode (for procfs) */
423 p2->p_textvp = p1->p_textvp;
424 if (p2->p_textvp)
597aea93 425 vref(p2->p_textvp);
984263bc 426
8ba5f7ef
AH
427 /* copy namecache handle to the text file */
428 if (p1->p_textnch.mount)
429 cache_copy(&p1->p_textnch, &p2->p_textnch);
430
0daa37a5
MD
431 /*
432 * Handle file descriptors
433 */
984263bc
MD
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;
8f1f6170 462 p2->p_limit = plimit_fork(p1);
984263bc
MD
463
464 /*
465 * Preserve some more flags in subprocess. P_PROFIL has already
466 * been preserved.
467 */
08f2f1bb 468 p2->p_flag |= p1->p_flag & P_SUGID;
984263bc
MD
469 if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
470 p2->p_flag |= P_CONTROLT;
471 if (flags & RFPPWAIT)
472 p2->p_flag |= P_PPWAIT;
473
5fd012e0 474 /*
0daa37a5
MD
475 * Inherit the virtual kernel structure (allows a virtual kernel
476 * to fork to simulate multiple cpus).
477 */
4a22e893
MD
478 if (p1->p_vkernel)
479 vkernel_inherit(p1, p2);
0daa37a5
MD
480
481 /*
5fd012e0
MD
482 * Once we are on a pglist we may receive signals. XXX we might
483 * race a ^C being sent to the process group by not receiving it
484 * at all prior to this line.
485 */
984263bc
MD
486 LIST_INSERT_AFTER(p1, p2, p_pglist);
487
488 /*
489 * Attach the new process to its parent.
490 *
491 * If RFNOWAIT is set, the newly created process becomes a child
492 * of init. This effectively disassociates the child from the
493 * parent.
494 */
495 if (flags & RFNOWAIT)
496 pptr = initproc;
497 else
498 pptr = p1;
499 p2->p_pptr = pptr;
500 LIST_INSERT_HEAD(&pptr->p_children, p2, p_sibling);
501 LIST_INIT(&p2->p_children);
98a7f915 502 varsymset_init(&p2->p_varsymset, &p1->p_varsymset);
8fbf9130 503 callout_init(&p2->p_ithandle);
984263bc
MD
504
505#ifdef KTRACE
506 /*
507 * Copy traceflag and tracefile if enabled. If not inherited,
508 * these were zeroed above but we still could have a trace race
29f58392 509 * so make sure p2's p_tracenode is NULL.
984263bc 510 */
29f58392 511 if ((p1->p_traceflag & KTRFAC_INHERIT) && p2->p_tracenode == NULL) {
984263bc 512 p2->p_traceflag = p1->p_traceflag;
29f58392 513 p2->p_tracenode = ktrinherit(p1->p_tracenode);
984263bc
MD
514 }
515#endif
516
517 /*
984263bc
MD
518 * This begins the section where we must prevent the parent
519 * from being swapped.
13d13d89
SS
520 *
521 * Gets PRELE'd in the caller in start_forked_proc().
984263bc
MD
522 */
523 PHOLD(p1);
524
13d13d89
SS
525 vm_fork(p1, p2, flags);
526
984263bc 527 /*
13d13d89
SS
528 * Create the first lwp associated with the new proc.
529 * It will return via a different execution path later, directly
530 * into userland, after it was put on the runq by
531 * start_forked_proc().
984263bc 532 */
13d13d89 533 lwp_fork(lp1, p2, flags);
984263bc 534
6b72a0c2 535 if (flags == (RFFDG | RFPROC | RFPGLOCK)) {
12e4aaff
MD
536 mycpu->gd_cnt.v_forks++;
537 mycpu->gd_cnt.v_forkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
6b72a0c2 538 } else if (flags == (RFFDG | RFPROC | RFPPWAIT | RFMEM | RFPGLOCK)) {
12e4aaff
MD
539 mycpu->gd_cnt.v_vforks++;
540 mycpu->gd_cnt.v_vforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 541 } else if (p1 == &proc0) {
12e4aaff
MD
542 mycpu->gd_cnt.v_kthreads++;
543 mycpu->gd_cnt.v_kthreadpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 544 } else {
12e4aaff
MD
545 mycpu->gd_cnt.v_rforks++;
546 mycpu->gd_cnt.v_rforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc
MD
547 }
548
549 /*
550 * Both processes are set up, now check if any loadable modules want
551 * to adjust anything.
552 * What if they have an error? XXX
553 */
554 TAILQ_FOREACH(ep, &fork_list, next) {
555 (*ep->function)(p1, p2, flags);
556 }
557
558 /*
a77ac49d
MD
559 * Set the start time. Note that the process is not runnable. The
560 * caller is responsible for making it runnable.
984263bc 561 */
d9fa5f67 562 microtime(&p2->p_start);
984263bc 563 p2->p_acflag = AFORK;
984263bc
MD
564
565 /*
566 * tell any interested parties about the new process
567 */
568 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid);
569
570 /*
984263bc
MD
571 * Return child proc pointer to parent.
572 */
573 *procp = p2;
167e6ecb
MD
574done:
575 if (pgrp)
576 lockmgr(&pgrp->pg_lock, LK_RELEASE);
577 return (error);
984263bc
MD
578}
579
13d13d89
SS
580static struct lwp *
581lwp_fork(struct lwp *origlp, struct proc *destproc, int flags)
582{
583 struct lwp *lp;
584 struct thread *td;
13d13d89 585
f6c36234 586 lp = kmalloc(sizeof(struct lwp), M_LWP, M_WAITOK|M_ZERO);
3e291793 587
13d13d89 588 lp->lwp_proc = destproc;
287ebb09 589 lp->lwp_vmspace = destproc->p_vmspace;
13d13d89 590 lp->lwp_stat = LSRUN;
13d13d89
SS
591 bcopy(&origlp->lwp_startcopy, &lp->lwp_startcopy,
592 (unsigned) ((caddr_t)&lp->lwp_endcopy -
593 (caddr_t)&lp->lwp_startcopy));
13d13d89
SS
594 lp->lwp_flag |= origlp->lwp_flag & LWP_ALTSTACK;
595 /*
596 * Set cpbase to the last timeout that occured (not the upcoming
597 * timeout).
598 *
599 * A critical section is required since a timer IPI can update
600 * scheduler specific data.
601 */
602 crit_enter();
603 lp->lwp_cpbase = mycpu->gd_schedclock.time -
604 mycpu->gd_schedclock.periodic;
605 destproc->p_usched->heuristic_forking(origlp, lp);
606 crit_exit();
a5ae2446 607 lp->lwp_cpumask &= usched_mastermask;
13d13d89 608
3e291793
MD
609 /*
610 * Assign a TID to the lp. Loop until the insert succeeds (returns
611 * NULL).
612 */
613 lp->lwp_tid = destproc->p_lasttid;
614 do {
615 if (++lp->lwp_tid < 0)
616 lp->lwp_tid = 1;
617 } while (lwp_rb_tree_RB_INSERT(&destproc->p_lwp_tree, lp) != NULL);
618 destproc->p_lasttid = lp->lwp_tid;
619 destproc->p_nthreads++;
620
fdce8919 621 td = lwkt_alloc_thread(NULL, LWKT_THREAD_STACK, -1, 0);
13d13d89
SS
622 lp->lwp_thread = td;
623 td->td_proc = destproc;
624 td->td_lwp = lp;
625 td->td_switch = cpu_heavy_switch;
13d13d89
SS
626 lwkt_setpri(td, TDPRI_KERN_USER);
627 lwkt_set_comm(td, "%s", destproc->p_comm);
628
629 /*
630 * cpu_fork will copy and update the pcb, set up the kernel stack,
631 * and make the child ready to run.
632 */
633 cpu_fork(origlp, lp, flags);
634 caps_fork(origlp->lwp_thread, lp->lwp_thread);
a591f597 635 kqueue_init(&lp->lwp_kqueue, destproc->p_fd);
13d13d89
SS
636
637 return (lp);
638}
639
984263bc
MD
640/*
641 * The next two functionms are general routines to handle adding/deleting
642 * items on the fork callout list.
643 *
644 * at_fork():
645 * Take the arguments given and put them onto the fork callout list,
646 * However first make sure that it's not already there.
647 * Returns 0 on success or a standard error number.
648 */
984263bc 649int
303c76d5 650at_fork(forklist_fn function)
984263bc
MD
651{
652 struct forklist *ep;
653
654#ifdef INVARIANTS
655 /* let the programmer know if he's been stupid */
303c76d5 656 if (rm_at_fork(function)) {
6ea70f76 657 kprintf("WARNING: fork callout entry (%p) already present\n",
984263bc 658 function);
303c76d5 659 }
984263bc 660#endif
efda3bd0 661 ep = kmalloc(sizeof(*ep), M_ATFORK, M_WAITOK|M_ZERO);
984263bc
MD
662 ep->function = function;
663 TAILQ_INSERT_TAIL(&fork_list, ep, next);
664 return (0);
665}
666
667/*
668 * Scan the exit callout list for the given item and remove it..
669 * Returns the number of items removed (0 or 1)
670 */
984263bc 671int
303c76d5 672rm_at_fork(forklist_fn function)
984263bc
MD
673{
674 struct forklist *ep;
675
676 TAILQ_FOREACH(ep, &fork_list, next) {
677 if (ep->function == function) {
678 TAILQ_REMOVE(&fork_list, ep, next);
efda3bd0 679 kfree(ep, M_ATFORK);
984263bc
MD
680 return(1);
681 }
682 }
683 return (0);
684}
7d0bac62
MD
685
686/*
687 * Add a forked process to the run queue after any remaining setup, such
688 * as setting the fork handler, has been completed.
689 */
7d0bac62 690void
553ea3c8 691start_forked_proc(struct lwp *lp1, struct proc *p2)
7d0bac62 692{
08f2f1bb 693 struct lwp *lp2 = ONLY_LWP_IN_PROC(p2);
553ea3c8 694
7d0bac62 695 /*
26a0694b
MD
696 * Move from SIDL to RUN queue, and activate the process's thread.
697 * Activation of the thread effectively makes the process "a"
698 * current process, so we do not setrunqueue().
8ec60c3f
MD
699 *
700 * YYY setrunqueue works here but we should clean up the trampoline
701 * code so we just schedule the LWKT thread and let the trampoline
702 * deal with the userland scheduler on return to userland.
7d0bac62 703 */
553ea3c8 704 KASSERT(p2->p_stat == SIDL,
7d0bac62 705 ("cannot start forked process, bad status: %p", p2));
553ea3c8 706 p2->p_usched->resetpriority(lp2);
e43a034f 707 crit_enter();
164b8401
SS
708 p2->p_stat = SACTIVE;
709 lp2->lwp_stat = LSRUN;
553ea3c8 710 p2->p_usched->setrunqueue(lp2);
e43a034f 711 crit_exit();
7d0bac62
MD
712
713 /*
714 * Now can be swapped.
715 */
553ea3c8 716 PRELE(lp1->lwp_proc);
7d0bac62
MD
717
718 /*
719 * Preserve synchronization semantics of vfork. If waiting for
720 * child to exec or exit, set P_PPWAIT on child, and sleep on our
721 * proc (in case of exit).
722 */
723 while (p2->p_flag & P_PPWAIT)
553ea3c8 724 tsleep(lp1->lwp_proc, 0, "ppwait", 0);
7d0bac62 725}