xref: /dflybsd-src/sys/kern/init_main.c (revision 5b991541a99aa38e5ca17ac8e6abee49bd57ac56)
1 /*
2  * Copyright (c) 1995 Terrence R. Lambert
3  * All rights reserved.
4  *
5  * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  * (c) UNIX System Laboratories, Inc.
8  * All or some portions of this file are derived from material licensed
9  * to the University of California by American Telephone and Telegraph
10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11  * the permission of UNIX System Laboratories, Inc.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. All advertising materials mentioning features or use of this software
22  *    must display the following acknowledgement:
23  *	This product includes software developed by the University of
24  *	California, Berkeley and its contributors.
25  * 4. Neither the name of the University nor the names of its contributors
26  *    may be used to endorse or promote products derived from this software
27  *    without specific prior written permission.
28  *
29  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
30  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
31  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
32  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
33  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
34  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
35  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
36  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
37  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
38  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39  * SUCH DAMAGE.
40  *
41  *	@(#)init_main.c	8.9 (Berkeley) 1/21/94
42  * $FreeBSD: src/sys/kern/init_main.c,v 1.134.2.8 2003/06/06 20:21:32 tegge Exp $
43  * $DragonFly: src/sys/kern/init_main.c,v 1.87 2008/06/07 11:37:23 mneumann Exp $
44  */
45 
46 #include "opt_init_path.h"
47 
48 #include <sys/param.h>
49 #include <sys/file.h>
50 #include <sys/filedesc.h>
51 #include <sys/kernel.h>
52 #include <sys/mount.h>
53 #include <sys/sysctl.h>
54 #include <sys/proc.h>
55 #include <sys/resourcevar.h>
56 #include <sys/signalvar.h>
57 #include <sys/systm.h>
58 #include <sys/vnode.h>
59 #include <sys/sysent.h>
60 #include <sys/reboot.h>
61 #include <sys/sysproto.h>
62 #include <sys/vmmeter.h>
63 #include <sys/unistd.h>
64 #include <sys/malloc.h>
65 #include <sys/machintr.h>
66 
67 #include <sys/file2.h>
68 #include <sys/thread2.h>
69 #include <sys/sysref2.h>
70 #include <sys/spinlock2.h>
71 #include <sys/mplock2.h>
72 
73 #include <machine/cpu.h>
74 
75 #include <vm/vm.h>
76 #include <vm/vm_param.h>
77 #include <sys/lock.h>
78 #include <vm/pmap.h>
79 #include <vm/vm_map.h>
80 #include <vm/vm_extern.h>
81 #include <sys/user.h>
82 #include <sys/copyright.h>
83 
84 int vfs_mountroot_devfs(void);
85 
86 /* Components of the first process -- never freed. */
87 static struct session session0;
88 static struct pgrp pgrp0;
89 static struct sigacts sigacts0;
90 static struct filedesc filedesc0;
91 static struct plimit limit0;
92 static struct vmspace vmspace0;
93 struct proc *initproc;
94 struct proc proc0;
95 struct lwp lwp0;
96 struct thread thread0;
97 
98 int cmask = CMASK;
99 extern	struct user *proc0paddr;
100 extern int fallback_elf_brand;
101 
102 int	boothowto = 0;		/* initialized so that it can be patched */
103 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
104 
105 /*
106  * This ensures that there is at least one entry so that the sysinit_set
107  * symbol is not undefined.  A subsystem ID of SI_SPECIAL_DUMMY is never
108  * executed.
109  */
110 SYSINIT(placeholder, SI_SPECIAL_DUMMY, SI_ORDER_ANY, NULL, NULL)
111 
112 /*
113  * The sysinit table itself.  Items are checked off as the are run.
114  * If we want to register new sysinit types, add them to newsysinit.
115  */
116 SET_DECLARE(sysinit_set, struct sysinit);
117 struct sysinit **sysinit, **sysinit_end;
118 struct sysinit **newsysinit, **newsysinit_end;
119 
120 
121 /*
122  * Merge a new sysinit set into the current set, reallocating it if
123  * necessary.  This can only be called after malloc is running.
124  */
125 void
126 sysinit_add(struct sysinit **set, struct sysinit **set_end)
127 {
128 	struct sysinit **newset;
129 	struct sysinit **sipp;
130 	struct sysinit **xipp;
131 	int count;
132 
133 	count = set_end - set;
134 	if (newsysinit)
135 		count += newsysinit_end - newsysinit;
136 	else
137 		count += sysinit_end - sysinit;
138 	newset = kmalloc(count * sizeof(*sipp), M_TEMP, M_WAITOK);
139 	xipp = newset;
140 	if (newsysinit) {
141 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
142 			*xipp++ = *sipp;
143 	} else {
144 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
145 			*xipp++ = *sipp;
146 	}
147 	for (sipp = set; sipp < set_end; sipp++)
148 		*xipp++ = *sipp;
149 	if (newsysinit)
150 		kfree(newsysinit, M_TEMP);
151 	newsysinit = newset;
152 	newsysinit_end = newset + count;
153 }
154 
155 /*
156  * Callbacks from machine-dependant startup code (e.g. init386) to set
157  * up low level entities related to cpu #0's globaldata.
158  *
159  * Called from very low level boot code.
160  */
161 void
162 mi_proc0init(struct globaldata *gd, struct user *proc0paddr)
163 {
164 	lwkt_init_thread(&thread0, proc0paddr, LWKT_THREAD_STACK, 0, gd);
165 	lwkt_set_comm(&thread0, "thread0");
166 #ifdef SMP
167 	thread0.td_mpcount = 1;	/* will hold mplock initially */
168 #endif
169 	RB_INIT(&proc0.p_lwp_tree);
170 	spin_init(&proc0.p_spin);
171 	proc0.p_lasttid = 0;	/* +1 = next TID */
172 	lwp_rb_tree_RB_INSERT(&proc0.p_lwp_tree, &lwp0);
173 	lwp0.lwp_thread = &thread0;
174 	lwp0.lwp_proc = &proc0;
175 	proc0.p_usched = usched_init();
176 	lwp0.lwp_cpumask = 0xFFFFFFFF;
177 	varsymset_init(&proc0.p_varsymset, NULL);
178 	thread0.td_flags |= TDF_RUNNING;
179 	thread0.td_proc = &proc0;
180 	thread0.td_lwp = &lwp0;
181 	thread0.td_switch = cpu_lwkt_switch;
182 	lwkt_schedule_self(curthread);
183 }
184 
185 /*
186  * System startup; initialize the world, create process 0, mount root
187  * filesystem, and fork to create init and pagedaemon.  Most of the
188  * hard work is done in the lower-level initialization routines including
189  * startup(), which does memory initialization and autoconfiguration.
190  *
191  * This allows simple addition of new kernel subsystems that require
192  * boot time initialization.  It also allows substitution of subsystem
193  * (for instance, a scheduler, kernel profiler, or VM system) by object
194  * module.  Finally, it allows for optional "kernel threads".
195  */
196 void
197 mi_startup(void)
198 {
199 	struct sysinit *sip;		/* system initialization*/
200 	struct sysinit **sipp;		/* system initialization*/
201 	struct sysinit **xipp;		/* interior loop of sort*/
202 	struct sysinit *save;		/* bubble*/
203 
204 	if (sysinit == NULL) {
205 		sysinit = SET_BEGIN(sysinit_set);
206 #if defined(__amd64__) && defined(_KERNEL_VIRTUAL)
207 		/*
208 		 * XXX For whatever reason, on 64-bit vkernels
209 		 * the value of sysinit obtained from the
210 		 * linker set is wrong.
211 		 */
212 		if ((long)sysinit % 8 != 0) {
213 			kprintf("Fixing sysinit value...\n");
214 			sysinit = (long)sysinit + 4;
215 		}
216 #endif
217 		sysinit_end = SET_LIMIT(sysinit_set);
218 	}
219 #if defined(__amd64__) && defined(_KERNEL_VIRTUAL)
220 	KKASSERT((long)sysinit % 8 == 0);
221 #endif
222 
223 restart:
224 	/*
225 	 * Perform a bubble sort of the system initialization objects by
226 	 * their subsystem (primary key) and order (secondary key).
227 	 */
228 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
229 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
230 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
231 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
232 			      (*sipp)->order <= (*xipp)->order))
233 				continue;	/* skip*/
234 			save = *sipp;
235 			*sipp = *xipp;
236 			*xipp = save;
237 		}
238 	}
239 
240 	/*
241 	 * Traverse the (now) ordered list of system initialization tasks.
242 	 * Perform each task, and continue on to the next task.
243 	 *
244 	 * The last item on the list is expected to be the scheduler,
245 	 * which will not return.
246 	 */
247 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
248 		sip = *sipp;
249 		if (sip->subsystem == SI_SPECIAL_DUMMY)
250 			continue;	/* skip dummy task(s)*/
251 
252 		if (sip->subsystem == SI_SPECIAL_DONE)
253 			continue;
254 
255 		/* Call function */
256 		(*(sip->func))(sip->udata);
257 
258 		/* Check off the one we're just done */
259 		sip->subsystem = SI_SPECIAL_DONE;
260 
261 		/* Check if we've installed more sysinit items via KLD */
262 		if (newsysinit != NULL) {
263 			if (sysinit != SET_BEGIN(sysinit_set))
264 				kfree(sysinit, M_TEMP);
265 			sysinit = newsysinit;
266 			sysinit_end = newsysinit_end;
267 			newsysinit = NULL;
268 			newsysinit_end = NULL;
269 			goto restart;
270 		}
271 	}
272 
273 	panic("Shouldn't get here!");
274 	/* NOTREACHED*/
275 }
276 
277 
278 /*
279  ***************************************************************************
280  ****
281  **** The following SYSINIT's belong elsewhere, but have not yet
282  **** been moved.
283  ****
284  ***************************************************************************
285  */
286 static void
287 print_caddr_t(void *data __unused)
288 {
289 	kprintf("%s", (char *)data);
290 }
291 SYSINIT(announce, SI_BOOT1_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
292 
293 /*
294  * Leave the critical section that protected us from spurious interrupts
295  * so device probes work.
296  */
297 static void
298 leavecrit(void *dummy __unused)
299 {
300 	MachIntrABI.finalize();
301 	cpu_enable_intr();
302 	MachIntrABI.cleanup();
303 	crit_exit();
304 	KKASSERT(!IN_CRITICAL_SECT(curthread));
305 
306 	if (bootverbose)
307 		kprintf("Leaving critical section, allowing interrupts\n");
308 }
309 SYSINIT(leavecrit, SI_BOOT2_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL)
310 
311 /*
312  * This is called after the threading system is up and running,
313  * including the softclock, clock interrupts, and SMP.
314  */
315 static void
316 tsleepworks(void *dummy __unused)
317 {
318 	tsleep_now_works = 1;
319 }
320 SYSINIT(tsleepworks, SI_BOOT2_FINISH_SMP, SI_ORDER_SECOND, tsleepworks, NULL)
321 
322 /*
323  * This is called after devices have configured.  Tell the kernel we are
324  * no longer in cold boot.
325  */
326 static void
327 endofcoldboot(void *dummy __unused)
328 {
329 	cold = 0;
330 }
331 SYSINIT(endofcoldboot, SI_SUB_ISWARM, SI_ORDER_ANY, endofcoldboot, NULL)
332 
333 /*
334  ***************************************************************************
335  ****
336  **** The two following SYSINT's are proc0 specific glue code.  I am not
337  **** convinced that they can not be safely combined, but their order of
338  **** operation has been maintained as the same as the original init_main.c
339  **** for right now.
340  ****
341  **** These probably belong in init_proc.c or kern_proc.c, since they
342  **** deal with proc0 (the fork template process).
343  ****
344  ***************************************************************************
345  */
346 /* ARGSUSED*/
347 static void
348 proc0_init(void *dummy __unused)
349 {
350 	struct proc *p;
351 	struct lwp *lp;
352 
353 	p = &proc0;
354 	lp = &lwp0;
355 
356 	/*
357 	 * Initialize process and pgrp structures.
358 	 */
359 	procinit();
360 
361 	/*
362 	 * additional VM structures
363 	 */
364 	vm_init2();
365 
366 	/*
367 	 * Create process 0 (the swapper).
368 	 */
369 	LIST_INSERT_HEAD(&allproc, p, p_list);
370 	p->p_pgrp = &pgrp0;
371 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
372 	LIST_INIT(&pgrp0.pg_members);
373 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
374 
375 	pgrp0.pg_session = &session0;
376 	session0.s_count = 1;
377 	session0.s_leader = p;
378 
379 	p->p_sysent = &aout_sysvec;
380 
381 	p->p_flag = P_SYSTEM;
382 	p->p_stat = SACTIVE;
383 	lp->lwp_stat = LSRUN;
384 	p->p_nice = NZERO;
385 	p->p_rtprio.type = RTP_PRIO_NORMAL;
386 	p->p_rtprio.prio = 0;
387 	lp->lwp_rtprio = p->p_rtprio;
388 
389 	p->p_peers = 0;
390 	p->p_leader = p;
391 
392 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
393 	bcopy("swapper", thread0.td_comm, sizeof ("swapper"));
394 
395 	/* Create credentials. */
396 	p->p_ucred = crget();
397 	p->p_ucred->cr_ruidinfo = uifind(0);
398 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
399 	p->p_ucred->cr_uidinfo = uifind(0);
400 	thread0.td_ucred = crhold(p->p_ucred);	/* bootstrap fork1() */
401 
402 	/* Don't jail it */
403 	p->p_ucred->cr_prison = NULL;
404 
405 	/* Create sigacts. */
406 	p->p_sigacts = &sigacts0;
407 	p->p_sigacts->ps_refcnt = 1;
408 
409 	/* Initialize signal state for process 0. */
410 	siginit(p);
411 
412 	/* Create the file descriptor table. */
413 	fdinit_bootstrap(p, &filedesc0, cmask);
414 
415 	/* Create the limits structures. */
416 	plimit_init0(&limit0);
417 	p->p_limit = &limit0;
418 
419 	/* Allocate a prototype map so we have something to fork. */
420 	pmap_pinit0(vmspace_pmap(&vmspace0));
421 	p->p_vmspace = &vmspace0;
422 	lp->lwp_vmspace = p->p_vmspace;
423 	sysref_init(&vmspace0.vm_sysref, &vmspace_sysref_class);
424 	vm_map_init(&vmspace0.vm_map,
425 		    round_page(VM_MIN_USER_ADDRESS),
426 		    trunc_page(VM_MAX_USER_ADDRESS),
427 		    vmspace_pmap(&vmspace0));
428 	sysref_activate(&vmspace0.vm_sysref);
429 
430 	kqueue_init(&lwp0.lwp_kqueue, &filedesc0);
431 
432 	/*
433 	 * Charge root for one process.
434 	 */
435 	(void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0);
436 	vm_init_limits(p);
437 }
438 SYSINIT(p0init, SI_BOOT2_PROC0, SI_ORDER_FIRST, proc0_init, NULL)
439 
440 static int proc0_post_callback(struct proc *p, void *data __unused);
441 
442 /* ARGSUSED*/
443 static void
444 proc0_post(void *dummy __unused)
445 {
446 	struct timespec ts;
447 
448 	/*
449 	 * Now we can look at the time, having had a chance to verify the
450 	 * time from the file system.  Pretend that proc0 started now.
451 	 */
452 	allproc_scan(proc0_post_callback, NULL);
453 
454 	/*
455 	 * Give the ``random'' number generator a thump.
456 	 * XXX: Does read_random() contain enough bits to be used here ?
457 	 */
458 	nanotime(&ts);
459 	skrandom(ts.tv_sec ^ ts.tv_nsec);
460 }
461 
462 static int
463 proc0_post_callback(struct proc *p, void *data __unused)
464 {
465 	microtime(&p->p_start);
466 	return(0);
467 }
468 
469 SYSINIT(p0post, SI_SUB_PROC0_POST, SI_ORDER_FIRST, proc0_post, NULL)
470 
471 /*
472  ***************************************************************************
473  ****
474  **** The following SYSINIT's and glue code should be moved to the
475  **** respective files on a per subsystem basis.
476  ****
477  ***************************************************************************
478  */
479 
480 
481 /*
482  ***************************************************************************
483  ****
484  **** The following code probably belongs in another file, like
485  **** kern/init_init.c.
486  ****
487  ***************************************************************************
488  */
489 
490 /*
491  * List of paths to try when searching for "init".
492  */
493 static char init_path[MAXPATHLEN] =
494 #ifdef	INIT_PATH
495     __XSTRING(INIT_PATH);
496 #else
497     "/sbin/init:/sbin/oinit:/sbin/init.bak";
498 #endif
499 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
500 
501 /*
502  * Start the initial user process; try exec'ing each pathname in init_path.
503  * The program is invoked with one argument containing the boot flags.
504  */
505 static void
506 start_init(void *dummy, struct trapframe *frame)
507 {
508 	vm_offset_t addr;
509 	struct execve_args args;
510 	int options, error;
511 	char *var, *path, *next, *s;
512 	char *ucp, **uap, *arg0, *arg1;
513 	struct proc *p;
514 	struct lwp *lp;
515 	struct mount *mp;
516 	struct vnode *vp;
517 	char *env;
518 
519         /*
520 	 * This is passed in by the bootloader
521          */
522 	env = kgetenv("kernelname");
523 	if (env != NULL)
524 		strlcpy(kernelname, env, sizeof(kernelname));
525 
526 	/*
527 	 * The MP lock is not held on entry.  We release it before
528 	 * returning to userland.
529 	 */
530 	get_mplock();
531 	p = curproc;
532 
533 	lp = ONLY_LWP_IN_PROC(p);
534 
535 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
536 	mp = mountlist_boot_getfirst();
537 	if (VFS_ROOT(mp, &vp))
538 		panic("cannot find root vnode");
539 	if (mp->mnt_ncmountpt.ncp == NULL) {
540 		cache_allocroot(&mp->mnt_ncmountpt, mp, vp);
541 		cache_unlock(&mp->mnt_ncmountpt);	/* leave ref intact */
542 	}
543 	p->p_fd->fd_cdir = vp;
544 	vref(p->p_fd->fd_cdir);
545 	p->p_fd->fd_rdir = vp;
546 	vref(p->p_fd->fd_rdir);
547 	vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt));
548 	vn_unlock(vp);			/* leave ref intact */
549 	cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir);
550 	cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir);
551 
552 	kprintf("Mounting devfs\n");
553 	vfs_mountroot_devfs();
554 
555 	/*
556 	 * Need just enough stack to hold the faked-up "execve()" arguments.
557 	 */
558 	addr = trunc_page(USRSTACK - PAGE_SIZE);
559 	error = vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr,
560 			    PAGE_SIZE, PAGE_SIZE,
561 			    FALSE, VM_MAPTYPE_NORMAL,
562 			    VM_PROT_ALL, VM_PROT_ALL,
563 			    0);
564 	if (error)
565 		panic("init: couldn't allocate argument space");
566 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
567 	p->p_vmspace->vm_ssize = 1;
568 
569 	if ((var = kgetenv("init_path")) != NULL) {
570 		strncpy(init_path, var, sizeof init_path);
571 		init_path[sizeof init_path - 1] = 0;
572 	}
573 	if ((var = kgetenv("kern.fallback_elf_brand")) != NULL)
574 		fallback_elf_brand = strtol(var, NULL, 0);
575 
576 	for (path = init_path; *path != '\0'; path = next) {
577 		while (*path == ':')
578 			path++;
579 		if (*path == '\0')
580 			break;
581 		for (next = path; *next != '\0' && *next != ':'; next++)
582 			/* nothing */ ;
583 		if (bootverbose)
584 			kprintf("start_init: trying %.*s\n", (int)(next - path),
585 			    path);
586 
587 		/*
588 		 * Move out the boot flag argument.
589 		 */
590 		options = 0;
591 		ucp = (char *)USRSTACK;
592 		(void)subyte(--ucp, 0);		/* trailing zero */
593 		if (boothowto & RB_SINGLE) {
594 			(void)subyte(--ucp, 's');
595 			options = 1;
596 		}
597 #ifdef notyet
598                 if (boothowto & RB_FASTBOOT) {
599 			(void)subyte(--ucp, 'f');
600 			options = 1;
601 		}
602 #endif
603 
604 #ifdef BOOTCDROM
605 		(void)subyte(--ucp, 'C');
606 		options = 1;
607 #endif
608 		if (options == 0)
609 			(void)subyte(--ucp, '-');
610 		(void)subyte(--ucp, '-');		/* leading hyphen */
611 		arg1 = ucp;
612 
613 		/*
614 		 * Move out the file name (also arg 0).
615 		 */
616 		(void)subyte(--ucp, 0);
617 		for (s = next - 1; s >= path; s--)
618 			(void)subyte(--ucp, *s);
619 		arg0 = ucp;
620 
621 		/*
622 		 * Move out the arg pointers.
623 		 */
624 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
625 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
626 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
627 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
628 
629 		/*
630 		 * Point at the arguments.
631 		 */
632 		args.fname = arg0;
633 		args.argv = uap;
634 		args.envv = NULL;
635 
636 		/*
637 		 * Now try to exec the program.  If can't for any reason
638 		 * other than it doesn't exist, complain.
639 		 *
640 		 * Otherwise, return via fork_trampoline() all the way
641 		 * to user mode as init!
642 		 *
643 		 * WARNING!  We may have been moved to another cpu after
644 		 * acquiring the current user process designation.  The
645 		 * MP lock will migrate with us though so we still have to
646 		 * release it.
647 		 */
648 		if ((error = sys_execve(&args)) == 0) {
649 			rel_mplock();
650 			lp->lwp_proc->p_usched->acquire_curproc(lp);
651 			return;
652 		}
653 		if (error != ENOENT)
654 			kprintf("exec %.*s: error %d\n", (int)(next - path),
655 			    path, error);
656 	}
657 	kprintf("init: not found in path %s\n", init_path);
658 	panic("no init");
659 }
660 
661 /*
662  * Like kthread_create(), but runs in it's own address space.
663  * We do this early to reserve pid 1.
664  *
665  * Note special case - do not make it runnable yet.  Other work
666  * in progress will change this more.
667  */
668 static void
669 create_init(const void *udata __unused)
670 {
671 	int error;
672 	struct lwp *lp;
673 
674 	crit_enter();
675 	error = fork1(&lwp0, RFFDG | RFPROC, &initproc);
676 	if (error)
677 		panic("cannot fork init: %d", error);
678 	initproc->p_flag |= P_SYSTEM;
679 	lp = ONLY_LWP_IN_PROC(initproc);
680 	cpu_set_fork_handler(lp, start_init, NULL);
681 	crit_exit();
682 }
683 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
684 
685 /*
686  * Make it runnable now.
687  */
688 static void
689 kick_init(const void *udata __unused)
690 {
691 	start_forked_proc(&lwp0, initproc);
692 }
693 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
694 
695 /*
696  * Machine independant globaldata initialization
697  *
698  * WARNING!  Called from early boot, 'mycpu' may not work yet.
699  */
700 void
701 mi_gdinit(struct globaldata *gd, int cpuid)
702 {
703 	TAILQ_INIT(&gd->gd_systimerq);
704 	gd->gd_sysid_alloc = cpuid;	/* prime low bits for cpu lookup */
705 	gd->gd_cpuid = cpuid;
706 	gd->gd_cpumask = (cpumask_t)1 << cpuid;
707 	lwkt_gdinit(gd);
708 	vm_map_entry_reserve_cpu_init(gd);
709 	sleep_gdinit(gd);
710 }
711 
712