kernel - Major signal path adjustments to fix races, tsleep race fixes, +more
[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>
19bfc8ab 68#include <sys/refcount.h>
e43a034f 69#include <sys/thread2.h>
b1b4e5a6 70#include <sys/signal2.h>
8f1f6170 71#include <sys/spinlock2.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
167e6ecb 121 error = fork1(lp, RFFDG | RFPROC | RFPGLOCK, &p2);
984263bc 122 if (error == 0) {
553ea3c8 123 start_forked_proc(lp, p2);
c7114eea
MD
124 uap->sysmsg_fds[0] = p2->p_pid;
125 uap->sysmsg_fds[1] = 0;
984263bc
MD
126 }
127 return error;
128}
129
3919ced0
MD
130/*
131 * MPALMOSTSAFE
132 */
984263bc 133int
753fd850 134sys_vfork(struct vfork_args *uap)
984263bc 135{
553ea3c8 136 struct lwp *lp = curthread->td_lwp;
984263bc 137 struct proc *p2;
41c20dac 138 int error;
984263bc 139
167e6ecb 140 error = fork1(lp, RFFDG | RFPROC | RFPPWAIT | RFMEM | RFPGLOCK, &p2);
984263bc 141 if (error == 0) {
553ea3c8 142 start_forked_proc(lp, p2);
c7114eea
MD
143 uap->sysmsg_fds[0] = p2->p_pid;
144 uap->sysmsg_fds[1] = 0;
984263bc
MD
145 }
146 return error;
147}
148
f61c1ff1
MD
149/*
150 * Handle rforks. An rfork may (1) operate on the current process without
151 * creating a new, (2) create a new process that shared the current process's
152 * vmspace, signals, and/or descriptors, or (3) create a new process that does
153 * not share these things (normal fork).
154 *
155 * Note that we only call start_forked_proc() if a new process is actually
156 * created.
157 *
158 * rfork { int flags }
3919ced0
MD
159 *
160 * MPALMOSTSAFE
f61c1ff1 161 */
984263bc 162int
753fd850 163sys_rfork(struct rfork_args *uap)
984263bc 164{
553ea3c8 165 struct lwp *lp = curthread->td_lwp;
984263bc 166 struct proc *p2;
41c20dac 167 int error;
984263bc 168
6654fbcb
MD
169 if ((uap->flags & RFKERNELONLY) != 0)
170 return (EINVAL);
171
167e6ecb 172 error = fork1(lp, uap->flags | RFPGLOCK, &p2);
984263bc 173 if (error == 0) {
f61c1ff1 174 if (p2)
553ea3c8 175 start_forked_proc(lp, p2);
c7114eea
MD
176 uap->sysmsg_fds[0] = p2 ? p2->p_pid : 0;
177 uap->sysmsg_fds[1] = 0;
984263bc
MD
178 }
179 return error;
180}
181
3919ced0
MD
182/*
183 * MPALMOSTSAFE
184 */
91bd9c1e
SS
185int
186sys_lwp_create(struct lwp_create_args *uap)
187{
188 struct proc *p = curproc;
189 struct lwp *lp;
190 struct lwp_params params;
191 int error;
192
193 error = copyin(uap->params, &params, sizeof(params));
194 if (error)
195 goto fail2;
196
b5c4d81f 197 lwkt_gettoken(&p->p_token);
8f1f6170 198 plimit_lwp_fork(p); /* force exclusive access */
91bd9c1e
SS
199 lp = lwp_fork(curthread->td_lwp, p, RFPROC);
200 error = cpu_prepare_lwp(lp, &params);
7b925b71
MD
201 if (error)
202 goto fail;
91bd9c1e
SS
203 if (params.tid1 != NULL &&
204 (error = copyout(&lp->lwp_tid, params.tid1, sizeof(lp->lwp_tid))))
205 goto fail;
206 if (params.tid2 != NULL &&
207 (error = copyout(&lp->lwp_tid, params.tid2, sizeof(lp->lwp_tid))))
208 goto fail;
209
210 /*
8f1f6170 211 * Now schedule the new lwp.
91bd9c1e
SS
212 */
213 p->p_usched->resetpriority(lp);
214 crit_enter();
215 lp->lwp_stat = LSRUN;
216 p->p_usched->setrunqueue(lp);
217 crit_exit();
b5c4d81f 218 lwkt_reltoken(&p->p_token);
91bd9c1e
SS
219
220 return (0);
221
222fail:
3e291793 223 lwp_rb_tree_RB_REMOVE(&p->p_lwp_tree, lp);
0b26dde3 224 --p->p_nthreads;
e3161323 225 /* lwp_dispose expects an exited lwp, and a held proc */
4643740a 226 atomic_set_int(&lp->lwp_mpflags, LWP_MP_WEXIT);
e3161323 227 lp->lwp_thread->td_flags |= TDF_EXITING;
7b925b71 228 lwkt_remove_tdallq(lp->lwp_thread);
e3161323 229 PHOLD(p);
18af8f55
MD
230 biosched_done(lp->lwp_thread);
231 dsched_exit_thread(lp->lwp_thread);
91bd9c1e 232 lwp_dispose(lp);
b5c4d81f 233 lwkt_reltoken(&p->p_token);
91bd9c1e
SS
234fail2:
235 return (error);
236}
984263bc
MD
237
238int nprocs = 1; /* process 0 */
984263bc
MD
239
240int
553ea3c8 241fork1(struct lwp *lp1, int flags, struct proc **procp)
984263bc 242{
553ea3c8 243 struct proc *p1 = lp1->lwp_proc;
984263bc 244 struct proc *p2, *pptr;
58c2553a
MD
245 struct pgrp *p1grp;
246 struct pgrp *plkgrp;
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
b5c4d81f 257 lwkt_gettoken(&p1->p_token);
58c2553a 258 plkgrp = NULL;
b5c4d81f 259
984263bc
MD
260 /*
261 * Here we don't create a new process, but we divorce
262 * certain parts of a process from itself.
263 */
264 if ((flags & RFPROC) == 0) {
13d13d89
SS
265 /*
266 * This kind of stunt does not work anymore if
267 * there are native threads (lwps) running
268 */
b5c4d81f
MD
269 if (p1->p_nthreads != 1) {
270 error = EINVAL;
271 goto done;
272 }
13d13d89
SS
273
274 vm_fork(p1, 0, flags);
984263bc
MD
275
276 /*
277 * Close all file descriptors.
278 */
279 if (flags & RFCFDG) {
280 struct filedesc *fdtmp;
281 fdtmp = fdinit(p1);
0a4a9c77 282 fdfree(p1, fdtmp);
984263bc
MD
283 }
284
285 /*
286 * Unshare file descriptors (from parent.)
287 */
288 if (flags & RFFDG) {
289 if (p1->p_fd->fd_refcnt > 1) {
290 struct filedesc *newfd;
2994659f
VS
291 error = fdcopy(p1, &newfd);
292 if (error != 0) {
293 error = ENOMEM;
294 goto done;
295 }
0a4a9c77 296 fdfree(p1, newfd);
984263bc
MD
297 }
298 }
299 *procp = NULL;
b5c4d81f
MD
300 error = 0;
301 goto done;
984263bc
MD
302 }
303
304 /*
167e6ecb
MD
305 * Interlock against process group signal delivery. If signals
306 * are pending after the interlock is obtained we have to restart
307 * the system call to process the signals. If we don't the child
308 * can miss a pgsignal (such as ^C) sent during the fork.
309 *
310 * We can't use CURSIG() here because it will process any STOPs
311 * and cause the process group lock to be held indefinitely. If
312 * a STOP occurs, the fork will be restarted after the CONT.
313 */
58c2553a
MD
314 p1grp = p1->p_pgrp;
315 if ((flags & RFPGLOCK) && (plkgrp = p1->p_pgrp) != NULL) {
316 pgref(plkgrp);
317 lockmgr(&plkgrp->pg_lock, LK_SHARED);
f6e73860 318 if (CURSIG_NOBLOCK(lp1)) {
167e6ecb
MD
319 error = ERESTART;
320 goto done;
321 }
322 }
323
324 /*
984263bc
MD
325 * Although process entries are dynamically created, we still keep
326 * a global limit on the maximum number we will create. Don't allow
327 * a nonprivileged user to use the last ten processes; don't let root
328 * exceed the limit. The variable nprocs is the current number of
329 * processes, maxproc is the limit.
330 */
9910d07b 331 uid = lp1->lwp_thread->td_ucred->cr_ruid;
984263bc 332 if ((nprocs >= maxproc - 10 && uid != 0) || nprocs >= maxproc) {
5bc7cd8d 333 if (ppsratecheck(&lastfail, &curfail, 1))
6ea70f76 334 kprintf("maxproc limit exceeded by uid %d, please "
5bc7cd8d 335 "see tuning(7) and login.conf(5).\n", uid);
377d4740 336 tsleep(&forksleep, 0, "fork", hz / 2);
167e6ecb
MD
337 error = EAGAIN;
338 goto done;
984263bc 339 }
8c2bce60 340
984263bc
MD
341 /*
342 * Increment the nprocs resource before blocking can occur. There
343 * are hard-limits as to the number of processes that can run.
344 */
8c2bce60 345 atomic_add_int(&nprocs, 1);
984263bc
MD
346
347 /*
348 * Increment the count of procs running with this uid. Don't allow
349 * a nonprivileged user to exceed their current limit.
350 */
9910d07b 351 ok = chgproccnt(lp1->lwp_thread->td_ucred->cr_ruidinfo, 1,
984263bc
MD
352 (uid != 0) ? p1->p_rlimit[RLIMIT_NPROC].rlim_cur : 0);
353 if (!ok) {
354 /*
355 * Back out the process count
356 */
8c2bce60 357 atomic_add_int(&nprocs, -1);
5bc7cd8d 358 if (ppsratecheck(&lastfail, &curfail, 1))
6ea70f76 359 kprintf("maxproc limit exceeded by uid %d, please "
5bc7cd8d 360 "see tuning(7) and login.conf(5).\n", uid);
377d4740 361 tsleep(&forksleep, 0, "fork", hz / 2);
167e6ecb
MD
362 error = EAGAIN;
363 goto done;
984263bc
MD
364 }
365
366 /* Allocate new proc. */
37733243 367 p2 = kmalloc(sizeof(struct proc), M_PROC, M_WAITOK|M_ZERO);
984263bc
MD
368
369 /*
ef09c3ed 370 * Setup linkage for kernel based threading XXX lwp
984263bc 371 */
cb74210d 372 if (flags & RFTHREAD) {
51e64ff2
MD
373 p2->p_peers = p1->p_peers;
374 p1->p_peers = p2;
375 p2->p_leader = p1->p_leader;
984263bc 376 } else {
51e64ff2 377 p2->p_leader = p2;
984263bc
MD
378 }
379
3e291793 380 RB_INIT(&p2->p_lwp_tree);
8f1f6170 381 spin_init(&p2->p_spin);
8c2bce60 382 lwkt_token_init(&p2->p_token, "proc");
3e291793 383 p2->p_lasttid = -1; /* first tid will be 0 */
ef09c3ed 384
984263bc 385 /*
51e64ff2
MD
386 * Setting the state to SIDL protects the partially initialized
387 * process once it starts getting hooked into the rest of the system.
984263bc 388 */
51e64ff2
MD
389 p2->p_stat = SIDL;
390 proc_add_allproc(p2);
984263bc
MD
391
392 /*
393 * Make a proc table entry for the new process.
f9fe34b6
SS
394 * The whole structure was zeroed above, so copy the section that is
395 * copied directly from the parent.
984263bc 396 */
984263bc
MD
397 bcopy(&p1->p_startcopy, &p2->p_startcopy,
398 (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
399
984263bc 400 /*
88072e3b
MD
401 * Duplicate sub-structures as needed. Increase reference counts
402 * on shared objects.
403 *
404 * NOTE: because we are now on the allproc list it is possible for
405 * other consumers to gain temporary references to p2
406 * (p2->p_lock can change).
984263bc 407 */
4643740a 408 if (p1->p_flags & P_PROFIL)
984263bc 409 startprofclock(p2);
9910d07b 410 p2->p_ucred = crhold(lp1->lwp_thread->td_ucred);
984263bc 411
b40e316c 412 if (jailed(p2->p_ucred))
4643740a 413 p2->p_flags |= P_JAILED;
984263bc
MD
414
415 if (p2->p_args)
19bfc8ab 416 refcount_acquire(&p2->p_args->ar_ref);
984263bc 417
13d13d89 418 p2->p_usched = p1->p_usched;
8c72e3d5
AH
419 /* XXX: verify copy of the secondary iosched stuff */
420 dsched_new_proc(p2);
13d13d89 421
984263bc 422 if (flags & RFSIGSHARE) {
b1b4e5a6 423 p2->p_sigacts = p1->p_sigacts;
6fa9e71a 424 refcount_acquire(&p2->p_sigacts->ps_refcnt);
984263bc 425 } else {
6fa9e71a
MD
426 p2->p_sigacts = kmalloc(sizeof(*p2->p_sigacts),
427 M_SUBPROC, M_WAITOK);
b1b4e5a6 428 bcopy(p1->p_sigacts, p2->p_sigacts, sizeof(*p2->p_sigacts));
6fa9e71a 429 refcount_init(&p2->p_sigacts->ps_refcnt, 1);
984263bc
MD
430 }
431 if (flags & RFLINUXTHPN)
432 p2->p_sigparent = SIGUSR1;
433 else
434 p2->p_sigparent = SIGCHLD;
435
436 /* bump references to the text vnode (for procfs) */
437 p2->p_textvp = p1->p_textvp;
438 if (p2->p_textvp)
597aea93 439 vref(p2->p_textvp);
984263bc 440
8ba5f7ef
AH
441 /* copy namecache handle to the text file */
442 if (p1->p_textnch.mount)
443 cache_copy(&p1->p_textnch, &p2->p_textnch);
444
0daa37a5
MD
445 /*
446 * Handle file descriptors
447 */
984263bc
MD
448 if (flags & RFCFDG) {
449 p2->p_fd = fdinit(p1);
450 fdtol = NULL;
451 } else if (flags & RFFDG) {
2994659f
VS
452 error = fdcopy(p1, &p2->p_fd);
453 if (error != 0) {
454 error = ENOMEM;
455 goto done;
456 }
984263bc
MD
457 fdtol = NULL;
458 } else {
459 p2->p_fd = fdshare(p1);
b5c4d81f
MD
460 if (p1->p_fdtol == NULL) {
461 lwkt_gettoken(&p1->p_token);
984263bc
MD
462 p1->p_fdtol =
463 filedesc_to_leader_alloc(NULL,
464 p1->p_leader);
b5c4d81f
MD
465 lwkt_reltoken(&p1->p_token);
466 }
984263bc
MD
467 if ((flags & RFTHREAD) != 0) {
468 /*
469 * Shared file descriptor table and
470 * shared process leaders.
471 */
472 fdtol = p1->p_fdtol;
473 fdtol->fdl_refcount++;
474 } else {
475 /*
476 * Shared file descriptor table, and
477 * different process leaders
478 */
98a7f915 479 fdtol = filedesc_to_leader_alloc(p1->p_fdtol, p2);
984263bc
MD
480 }
481 }
482 p2->p_fdtol = fdtol;
8f1f6170 483 p2->p_limit = plimit_fork(p1);
984263bc
MD
484
485 /*
486 * Preserve some more flags in subprocess. P_PROFIL has already
487 * been preserved.
488 */
4643740a
MD
489 p2->p_flags |= p1->p_flags & P_SUGID;
490 if (p1->p_session->s_ttyvp != NULL && p1->p_flags & P_CONTROLT)
491 p2->p_flags |= P_CONTROLT;
984263bc 492 if (flags & RFPPWAIT)
4643740a 493 p2->p_flags |= P_PPWAIT;
984263bc 494
5fd012e0 495 /*
0daa37a5
MD
496 * Inherit the virtual kernel structure (allows a virtual kernel
497 * to fork to simulate multiple cpus).
498 */
4a22e893
MD
499 if (p1->p_vkernel)
500 vkernel_inherit(p1, p2);
0daa37a5
MD
501
502 /*
5fd012e0
MD
503 * Once we are on a pglist we may receive signals. XXX we might
504 * race a ^C being sent to the process group by not receiving it
505 * at all prior to this line.
506 */
58c2553a
MD
507 pgref(p1grp);
508 lwkt_gettoken(&p1grp->pg_token);
984263bc 509 LIST_INSERT_AFTER(p1, p2, p_pglist);
58c2553a 510 lwkt_reltoken(&p1grp->pg_token);
984263bc
MD
511
512 /*
513 * Attach the new process to its parent.
514 *
515 * If RFNOWAIT is set, the newly created process becomes a child
516 * of init. This effectively disassociates the child from the
517 * parent.
518 */
519 if (flags & RFNOWAIT)
520 pptr = initproc;
521 else
522 pptr = p1;
523 p2->p_pptr = pptr;
984263bc 524 LIST_INIT(&p2->p_children);
b5c4d81f
MD
525
526 lwkt_gettoken(&pptr->p_token);
527 LIST_INSERT_HEAD(&pptr->p_children, p2, p_sibling);
528 lwkt_reltoken(&pptr->p_token);
529
98a7f915 530 varsymset_init(&p2->p_varsymset, &p1->p_varsymset);
8c2bce60 531 callout_init_mp(&p2->p_ithandle);
984263bc
MD
532
533#ifdef KTRACE
534 /*
535 * Copy traceflag and tracefile if enabled. If not inherited,
536 * these were zeroed above but we still could have a trace race
29f58392 537 * so make sure p2's p_tracenode is NULL.
984263bc 538 */
29f58392 539 if ((p1->p_traceflag & KTRFAC_INHERIT) && p2->p_tracenode == NULL) {
984263bc 540 p2->p_traceflag = p1->p_traceflag;
29f58392 541 p2->p_tracenode = ktrinherit(p1->p_tracenode);
984263bc
MD
542 }
543#endif
544
545 /*
984263bc
MD
546 * This begins the section where we must prevent the parent
547 * from being swapped.
13d13d89
SS
548 *
549 * Gets PRELE'd in the caller in start_forked_proc().
984263bc
MD
550 */
551 PHOLD(p1);
552
13d13d89
SS
553 vm_fork(p1, p2, flags);
554
984263bc 555 /*
13d13d89
SS
556 * Create the first lwp associated with the new proc.
557 * It will return via a different execution path later, directly
558 * into userland, after it was put on the runq by
559 * start_forked_proc().
984263bc 560 */
13d13d89 561 lwp_fork(lp1, p2, flags);
984263bc 562
6b72a0c2 563 if (flags == (RFFDG | RFPROC | RFPGLOCK)) {
12e4aaff
MD
564 mycpu->gd_cnt.v_forks++;
565 mycpu->gd_cnt.v_forkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
6b72a0c2 566 } else if (flags == (RFFDG | RFPROC | RFPPWAIT | RFMEM | RFPGLOCK)) {
12e4aaff
MD
567 mycpu->gd_cnt.v_vforks++;
568 mycpu->gd_cnt.v_vforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 569 } else if (p1 == &proc0) {
12e4aaff
MD
570 mycpu->gd_cnt.v_kthreads++;
571 mycpu->gd_cnt.v_kthreadpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc 572 } else {
12e4aaff
MD
573 mycpu->gd_cnt.v_rforks++;
574 mycpu->gd_cnt.v_rforkpages += p2->p_vmspace->vm_dsize + p2->p_vmspace->vm_ssize;
984263bc
MD
575 }
576
577 /*
578 * Both processes are set up, now check if any loadable modules want
579 * to adjust anything.
580 * What if they have an error? XXX
581 */
582 TAILQ_FOREACH(ep, &fork_list, next) {
583 (*ep->function)(p1, p2, flags);
584 }
585
586 /*
a77ac49d
MD
587 * Set the start time. Note that the process is not runnable. The
588 * caller is responsible for making it runnable.
984263bc 589 */
d9fa5f67 590 microtime(&p2->p_start);
984263bc 591 p2->p_acflag = AFORK;
984263bc
MD
592
593 /*
594 * tell any interested parties about the new process
595 */
596 KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid);
597
598 /*
984263bc
MD
599 * Return child proc pointer to parent.
600 */
601 *procp = p2;
b5c4d81f 602 error = 0;
167e6ecb 603done:
b5c4d81f 604 lwkt_reltoken(&p1->p_token);
58c2553a
MD
605 if (plkgrp) {
606 lockmgr(&plkgrp->pg_lock, LK_RELEASE);
607 pgrel(plkgrp);
608 }
167e6ecb 609 return (error);
984263bc
MD
610}
611
13d13d89
SS
612static struct lwp *
613lwp_fork(struct lwp *origlp, struct proc *destproc, int flags)
614{
d2d8515b 615 globaldata_t gd = mycpu;
13d13d89
SS
616 struct lwp *lp;
617 struct thread *td;
13d13d89 618
f6c36234 619 lp = kmalloc(sizeof(struct lwp), M_LWP, M_WAITOK|M_ZERO);
3e291793 620
13d13d89 621 lp->lwp_proc = destproc;
287ebb09 622 lp->lwp_vmspace = destproc->p_vmspace;
13d13d89 623 lp->lwp_stat = LSRUN;
13d13d89
SS
624 bcopy(&origlp->lwp_startcopy, &lp->lwp_startcopy,
625 (unsigned) ((caddr_t)&lp->lwp_endcopy -
626 (caddr_t)&lp->lwp_startcopy));
4643740a 627 lp->lwp_flags |= origlp->lwp_flags & LWP_ALTSTACK;
13d13d89
SS
628 /*
629 * Set cpbase to the last timeout that occured (not the upcoming
630 * timeout).
631 *
632 * A critical section is required since a timer IPI can update
633 * scheduler specific data.
634 */
635 crit_enter();
d2d8515b 636 lp->lwp_cpbase = gd->gd_schedclock.time - gd->gd_schedclock.periodic;
13d13d89
SS
637 destproc->p_usched->heuristic_forking(origlp, lp);
638 crit_exit();
a5ae2446 639 lp->lwp_cpumask &= usched_mastermask;
e2b148c6 640 lwkt_token_init(&lp->lwp_token, "lwp_token");
94f98873 641 spin_init(&lp->lwp_spin);
13d13d89 642
d2d8515b
MD
643 /*
644 * Assign the thread to the current cpu to begin with so we
645 * can manipulate it.
646 */
647 td = lwkt_alloc_thread(NULL, LWKT_THREAD_STACK, gd->gd_cpuid, 0);
13d13d89
SS
648 lp->lwp_thread = td;
649 td->td_proc = destproc;
650 td->td_lwp = lp;
651 td->td_switch = cpu_heavy_switch;
13d13d89
SS
652 lwkt_setpri(td, TDPRI_KERN_USER);
653 lwkt_set_comm(td, "%s", destproc->p_comm);
654
655 /*
656 * cpu_fork will copy and update the pcb, set up the kernel stack,
657 * and make the child ready to run.
658 */
659 cpu_fork(origlp, lp, flags);
660 caps_fork(origlp->lwp_thread, lp->lwp_thread);
a591f597 661 kqueue_init(&lp->lwp_kqueue, destproc->p_fd);
13d13d89 662
0d78b86e
MD
663 /*
664 * Assign a TID to the lp. Loop until the insert succeeds (returns
665 * NULL).
666 */
667 lp->lwp_tid = destproc->p_lasttid;
668 do {
669 if (++lp->lwp_tid < 0)
670 lp->lwp_tid = 1;
671 } while (lwp_rb_tree_RB_INSERT(&destproc->p_lwp_tree, lp) != NULL);
672 destproc->p_lasttid = lp->lwp_tid;
673 destproc->p_nthreads++;
674
13d13d89
SS
675 return (lp);
676}
677
984263bc
MD
678/*
679 * The next two functionms are general routines to handle adding/deleting
680 * items on the fork callout list.
681 *
682 * at_fork():
683 * Take the arguments given and put them onto the fork callout list,
684 * However first make sure that it's not already there.
685 * Returns 0 on success or a standard error number.
686 */
984263bc 687int
303c76d5 688at_fork(forklist_fn function)
984263bc
MD
689{
690 struct forklist *ep;
691
692#ifdef INVARIANTS
693 /* let the programmer know if he's been stupid */
303c76d5 694 if (rm_at_fork(function)) {
6ea70f76 695 kprintf("WARNING: fork callout entry (%p) already present\n",
984263bc 696 function);
303c76d5 697 }
984263bc 698#endif
efda3bd0 699 ep = kmalloc(sizeof(*ep), M_ATFORK, M_WAITOK|M_ZERO);
984263bc
MD
700 ep->function = function;
701 TAILQ_INSERT_TAIL(&fork_list, ep, next);
702 return (0);
703}
704
705/*
706 * Scan the exit callout list for the given item and remove it..
707 * Returns the number of items removed (0 or 1)
708 */
984263bc 709int
303c76d5 710rm_at_fork(forklist_fn function)
984263bc
MD
711{
712 struct forklist *ep;
713
714 TAILQ_FOREACH(ep, &fork_list, next) {
715 if (ep->function == function) {
716 TAILQ_REMOVE(&fork_list, ep, next);
efda3bd0 717 kfree(ep, M_ATFORK);
984263bc
MD
718 return(1);
719 }
720 }
721 return (0);
722}
7d0bac62
MD
723
724/*
725 * Add a forked process to the run queue after any remaining setup, such
726 * as setting the fork handler, has been completed.
727 */
7d0bac62 728void
553ea3c8 729start_forked_proc(struct lwp *lp1, struct proc *p2)
7d0bac62 730{
08f2f1bb 731 struct lwp *lp2 = ONLY_LWP_IN_PROC(p2);
553ea3c8 732
7d0bac62 733 /*
26a0694b
MD
734 * Move from SIDL to RUN queue, and activate the process's thread.
735 * Activation of the thread effectively makes the process "a"
736 * current process, so we do not setrunqueue().
8ec60c3f
MD
737 *
738 * YYY setrunqueue works here but we should clean up the trampoline
739 * code so we just schedule the LWKT thread and let the trampoline
740 * deal with the userland scheduler on return to userland.
7d0bac62 741 */
553ea3c8 742 KASSERT(p2->p_stat == SIDL,
7d0bac62 743 ("cannot start forked process, bad status: %p", p2));
553ea3c8 744 p2->p_usched->resetpriority(lp2);
e43a034f 745 crit_enter();
164b8401
SS
746 p2->p_stat = SACTIVE;
747 lp2->lwp_stat = LSRUN;
553ea3c8 748 p2->p_usched->setrunqueue(lp2);
e43a034f 749 crit_exit();
7d0bac62
MD
750
751 /*
752 * Now can be swapped.
753 */
553ea3c8 754 PRELE(lp1->lwp_proc);
7d0bac62
MD
755
756 /*
757 * Preserve synchronization semantics of vfork. If waiting for
758 * child to exec or exit, set P_PPWAIT on child, and sleep on our
ee934fe9
MD
759 * proc (in case of exec or exit).
760 *
761 * We must hold our p_token to interlock the flag/tsleep
7d0bac62 762 */
ee934fe9 763 lwkt_gettoken(&p2->p_token);
4643740a 764 while (p2->p_flags & P_PPWAIT)
553ea3c8 765 tsleep(lp1->lwp_proc, 0, "ppwait", 0);
ee934fe9 766 lwkt_reltoken(&p2->p_token);
7d0bac62 767}