xref: /dragonfly/sys/kern/init_main.c (revision 81c11cd3)
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 u_int cpu_mi_feature;
100 extern	struct user *proc0paddr;
101 extern int fallback_elf_brand;
102 
103 int	boothowto = 0;		/* initialized so that it can be patched */
104 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
105     "Reboot flags, from console subsystem");
106 
107 /*
108  * This ensures that there is at least one entry so that the sysinit_set
109  * symbol is not undefined.  A subsystem ID of SI_SPECIAL_DUMMY is never
110  * executed.
111  */
112 SYSINIT(placeholder, SI_SPECIAL_DUMMY, SI_ORDER_ANY, NULL, NULL)
113 
114 /*
115  * The sysinit table itself.  Items are checked off as the are run.
116  * If we want to register new sysinit types, add them to newsysinit.
117  */
118 SET_DECLARE(sysinit_set, struct sysinit);
119 struct sysinit **sysinit, **sysinit_end;
120 struct sysinit **newsysinit, **newsysinit_end;
121 
122 
123 /*
124  * Merge a new sysinit set into the current set, reallocating it if
125  * necessary.  This can only be called after malloc is running.
126  */
127 void
128 sysinit_add(struct sysinit **set, struct sysinit **set_end)
129 {
130 	struct sysinit **newset;
131 	struct sysinit **sipp;
132 	struct sysinit **xipp;
133 	int count;
134 
135 	count = set_end - set;
136 	if (newsysinit)
137 		count += newsysinit_end - newsysinit;
138 	else
139 		count += sysinit_end - sysinit;
140 	newset = kmalloc(count * sizeof(*sipp), M_TEMP, M_WAITOK);
141 	xipp = newset;
142 	if (newsysinit) {
143 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
144 			*xipp++ = *sipp;
145 	} else {
146 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
147 			*xipp++ = *sipp;
148 	}
149 	for (sipp = set; sipp < set_end; sipp++)
150 		*xipp++ = *sipp;
151 	if (newsysinit)
152 		kfree(newsysinit, M_TEMP);
153 	newsysinit = newset;
154 	newsysinit_end = newset + count;
155 }
156 
157 /*
158  * Callbacks from machine-dependant startup code (e.g. init386) to set
159  * up low level entities related to cpu #0's globaldata.
160  *
161  * Called from very low level boot code.
162  */
163 void
164 mi_proc0init(struct globaldata *gd, struct user *proc0paddr)
165 {
166 	lwkt_init_thread(&thread0, proc0paddr, LWKT_THREAD_STACK, 0, gd);
167 	lwkt_set_comm(&thread0, "thread0");
168 	RB_INIT(&proc0.p_lwp_tree);
169 	spin_init(&proc0.p_spin);
170 	proc0.p_lasttid = 0;	/* +1 = next TID */
171 	lwp_rb_tree_RB_INSERT(&proc0.p_lwp_tree, &lwp0);
172 	lwp0.lwp_thread = &thread0;
173 	lwp0.lwp_proc = &proc0;
174 	proc0.p_usched = usched_init();
175 	lwp0.lwp_cpumask = (cpumask_t)-1;
176 	varsymset_init(&proc0.p_varsymset, NULL);
177 	thread0.td_flags |= TDF_RUNNING;
178 	thread0.td_proc = &proc0;
179 	thread0.td_lwp = &lwp0;
180 	thread0.td_switch = cpu_lwkt_switch;
181 	lwkt_schedule_self(curthread);
182 }
183 
184 /*
185  * System startup; initialize the world, create process 0, mount root
186  * filesystem, and fork to create init and pagedaemon.  Most of the
187  * hard work is done in the lower-level initialization routines including
188  * startup(), which does memory initialization and autoconfiguration.
189  *
190  * This allows simple addition of new kernel subsystems that require
191  * boot time initialization.  It also allows substitution of subsystem
192  * (for instance, a scheduler, kernel profiler, or VM system) by object
193  * module.  Finally, it allows for optional "kernel threads".
194  */
195 void
196 mi_startup(void)
197 {
198 	struct sysinit *sip;		/* system initialization*/
199 	struct sysinit **sipp;		/* system initialization*/
200 	struct sysinit **xipp;		/* interior loop of sort*/
201 	struct sysinit *save;		/* bubble*/
202 
203 	if (sysinit == NULL) {
204 		sysinit = SET_BEGIN(sysinit_set);
205 #if defined(__amd64__) && defined(_KERNEL_VIRTUAL)
206 		/*
207 		 * XXX For whatever reason, on 64-bit vkernels
208 		 * the value of sysinit obtained from the
209 		 * linker set is wrong.
210 		 */
211 		if ((long)sysinit % 8 != 0) {
212 			kprintf("Fixing sysinit value...\n");
213 			sysinit = (long)sysinit + 4;
214 		}
215 #endif
216 		sysinit_end = SET_LIMIT(sysinit_set);
217 	}
218 #if defined(__amd64__) && defined(_KERNEL_VIRTUAL)
219 	KKASSERT((long)sysinit % 8 == 0);
220 #endif
221 
222 restart:
223 	/*
224 	 * Perform a bubble sort of the system initialization objects by
225 	 * their subsystem (primary key) and order (secondary key).
226 	 */
227 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
228 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
229 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
230 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
231 			      (*sipp)->order <= (*xipp)->order))
232 				continue;	/* skip*/
233 			save = *sipp;
234 			*sipp = *xipp;
235 			*xipp = save;
236 		}
237 	}
238 
239 	/*
240 	 * Traverse the (now) ordered list of system initialization tasks.
241 	 * Perform each task, and continue on to the next task.
242 	 *
243 	 * The last item on the list is expected to be the scheduler,
244 	 * which will not return.
245 	 */
246 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
247 		sip = *sipp;
248 		if (sip->subsystem == SI_SPECIAL_DUMMY)
249 			continue;	/* skip dummy task(s)*/
250 
251 		if (sip->subsystem == SI_SPECIAL_DONE)
252 			continue;
253 
254 		/* Call function */
255 		(*(sip->func))(sip->udata);
256 
257 		/* Check off the one we're just done */
258 		sip->subsystem = SI_SPECIAL_DONE;
259 
260 		/* Check if we've installed more sysinit items via KLD */
261 		if (newsysinit != NULL) {
262 			if (sysinit != SET_BEGIN(sysinit_set))
263 				kfree(sysinit, M_TEMP);
264 			sysinit = newsysinit;
265 			sysinit_end = newsysinit_end;
266 			newsysinit = NULL;
267 			newsysinit_end = NULL;
268 			goto restart;
269 		}
270 	}
271 
272 	panic("Shouldn't get here!");
273 	/* NOTREACHED*/
274 }
275 
276 
277 /*
278  ***************************************************************************
279  ****
280  **** The following SYSINIT's belong elsewhere, but have not yet
281  **** been moved.
282  ****
283  ***************************************************************************
284  */
285 static void
286 print_caddr_t(void *data __unused)
287 {
288 	kprintf("%s", (char *)data);
289 }
290 SYSINIT(announce, SI_BOOT1_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
291 
292 /*
293  * Leave the critical section that protected us from spurious interrupts
294  * so device probes work.
295  */
296 static void
297 leavecrit(void *dummy __unused)
298 {
299 	MachIntrABI.finalize();
300 	cpu_enable_intr();
301 	MachIntrABI.cleanup();
302 	crit_exit();
303 	KKASSERT(!IN_CRITICAL_SECT(curthread));
304 
305 	if (bootverbose)
306 		kprintf("Leaving critical section, allowing interrupts\n");
307 }
308 SYSINIT(leavecrit, SI_BOOT2_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL)
309 
310 /*
311  * This is called after the threading system is up and running,
312  * including the softclock, clock interrupts, and SMP.
313  */
314 static void
315 tsleepworks(void *dummy __unused)
316 {
317 	tsleep_now_works = 1;
318 }
319 SYSINIT(tsleepworks, SI_BOOT2_FINISH_SMP, SI_ORDER_SECOND, tsleepworks, NULL)
320 
321 /*
322  * This is called after devices have configured.  Tell the kernel we are
323  * no longer in cold boot.
324  */
325 static void
326 endofcoldboot(void *dummy __unused)
327 {
328 	cold = 0;
329 }
330 SYSINIT(endofcoldboot, SI_SUB_ISWARM, SI_ORDER_ANY, endofcoldboot, NULL)
331 
332 /*
333  ***************************************************************************
334  ****
335  **** The two following SYSINT's are proc0 specific glue code.  I am not
336  **** convinced that they can not be safely combined, but their order of
337  **** operation has been maintained as the same as the original init_main.c
338  **** for right now.
339  ****
340  **** These probably belong in init_proc.c or kern_proc.c, since they
341  **** deal with proc0 (the fork template process).
342  ****
343  ***************************************************************************
344  */
345 /* ARGSUSED*/
346 static void
347 proc0_init(void *dummy __unused)
348 {
349 	struct proc *p;
350 	struct lwp *lp;
351 
352 	p = &proc0;
353 	lp = &lwp0;
354 
355 	/*
356 	 * Initialize process and pgrp structures.
357 	 */
358 	procinit();
359 
360 	/*
361 	 * additional VM structures
362 	 */
363 	vm_init2();
364 
365 	/*
366 	 * Create process 0 (the swapper).
367 	 */
368 	LIST_INSERT_HEAD(&allproc, p, p_list);
369 	p->p_pgrp = &pgrp0;
370 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
371 	LIST_INIT(&pgrp0.pg_members);
372 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
373 
374 	pgrp0.pg_session = &session0;
375 	session0.s_count = 1;
376 	session0.s_leader = p;
377 
378 	p->p_sysent = &aout_sysvec;
379 
380 	p->p_flag = P_SYSTEM;
381 	p->p_stat = SACTIVE;
382 	lp->lwp_stat = LSRUN;
383 	p->p_nice = NZERO;
384 	p->p_rtprio.type = RTP_PRIO_NORMAL;
385 	p->p_rtprio.prio = 0;
386 	lp->lwp_rtprio = p->p_rtprio;
387 
388 	p->p_peers = 0;
389 	p->p_leader = p;
390 
391 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
392 	bcopy("swapper", thread0.td_comm, sizeof ("swapper"));
393 
394 	/* Create credentials. */
395 	p->p_ucred = crget();
396 	p->p_ucred->cr_ruidinfo = uifind(0);
397 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
398 	p->p_ucred->cr_uidinfo = uifind(0);
399 	thread0.td_ucred = crhold(p->p_ucred);	/* bootstrap fork1() */
400 
401 	/* Don't jail it */
402 	p->p_ucred->cr_prison = NULL;
403 
404 	/* Create sigacts. */
405 	p->p_sigacts = &sigacts0;
406 	p->p_sigacts->ps_refcnt = 1;
407 
408 	/* Initialize signal state for process 0. */
409 	siginit(p);
410 
411 	/* Create the file descriptor table. */
412 	fdinit_bootstrap(p, &filedesc0, cmask);
413 
414 	/* Create the limits structures. */
415 	plimit_init0(&limit0);
416 	p->p_limit = &limit0;
417 
418 	/* Allocate a prototype map so we have something to fork. */
419 	pmap_pinit0(vmspace_pmap(&vmspace0));
420 	p->p_vmspace = &vmspace0;
421 	lp->lwp_vmspace = p->p_vmspace;
422 	sysref_init(&vmspace0.vm_sysref, &vmspace_sysref_class);
423 	vm_map_init(&vmspace0.vm_map,
424 		    round_page(VM_MIN_USER_ADDRESS),
425 		    trunc_page(VM_MAX_USER_ADDRESS),
426 		    vmspace_pmap(&vmspace0));
427 	sysref_activate(&vmspace0.vm_sysref);
428 
429 	kqueue_init(&lwp0.lwp_kqueue, &filedesc0);
430 
431 	/*
432 	 * Charge root for one process.
433 	 */
434 	(void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0);
435 	vm_init_limits(p);
436 }
437 SYSINIT(p0init, SI_BOOT2_PROC0, SI_ORDER_FIRST, proc0_init, NULL)
438 
439 static int proc0_post_callback(struct proc *p, void *data __unused);
440 
441 /* ARGSUSED*/
442 static void
443 proc0_post(void *dummy __unused)
444 {
445 	struct timespec ts;
446 
447 	/*
448 	 * Now we can look at the time, having had a chance to verify the
449 	 * time from the file system.  Pretend that proc0 started now.
450 	 */
451 	allproc_scan(proc0_post_callback, NULL);
452 
453 	/*
454 	 * Give the ``random'' number generator a thump.
455 	 * XXX: Does read_random() contain enough bits to be used here ?
456 	 */
457 	nanotime(&ts);
458 	skrandom(ts.tv_sec ^ ts.tv_nsec);
459 }
460 
461 static int
462 proc0_post_callback(struct proc *p, void *data __unused)
463 {
464 	microtime(&p->p_start);
465 	return(0);
466 }
467 
468 SYSINIT(p0post, SI_SUB_PROC0_POST, SI_ORDER_FIRST, proc0_post, NULL)
469 
470 /*
471  ***************************************************************************
472  ****
473  **** The following SYSINIT's and glue code should be moved to the
474  **** respective files on a per subsystem basis.
475  ****
476  ***************************************************************************
477  */
478 
479 
480 /*
481  ***************************************************************************
482  ****
483  **** The following code probably belongs in another file, like
484  **** kern/init_init.c.
485  ****
486  ***************************************************************************
487  */
488 
489 /*
490  * List of paths to try when searching for "init".
491  */
492 static char init_path[MAXPATHLEN] =
493 #ifdef	INIT_PATH
494     __XSTRING(INIT_PATH);
495 #else
496     "/sbin/init:/sbin/oinit:/sbin/init.bak";
497 #endif
498 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
499 
500 /*
501  * Start the initial user process; try exec'ing each pathname in init_path.
502  * The program is invoked with one argument containing the boot flags.
503  */
504 static void
505 start_init(void *dummy, struct trapframe *frame)
506 {
507 	vm_offset_t addr;
508 	struct execve_args args;
509 	int options, error;
510 	char *var, *path, *next, *s;
511 	char *ucp, **uap, *arg0, *arg1;
512 	struct proc *p;
513 	struct lwp *lp;
514 	struct mount *mp;
515 	struct vnode *vp;
516 	char *env;
517 
518         /*
519 	 * This is passed in by the bootloader
520          */
521 	env = kgetenv("kernelname");
522 	if (env != NULL)
523 		strlcpy(kernelname, env, sizeof(kernelname));
524 
525 	/*
526 	 * The MP lock is not held on entry.  We release it before
527 	 * returning to userland.
528 	 */
529 	get_mplock();
530 	p = curproc;
531 
532 	lp = ONLY_LWP_IN_PROC(p);
533 
534 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
535 	mp = mountlist_boot_getfirst();
536 	if (VFS_ROOT(mp, &vp))
537 		panic("cannot find root vnode");
538 	if (mp->mnt_ncmountpt.ncp == NULL) {
539 		cache_allocroot(&mp->mnt_ncmountpt, mp, vp);
540 		cache_unlock(&mp->mnt_ncmountpt);	/* leave ref intact */
541 	}
542 	p->p_fd->fd_cdir = vp;
543 	vref(p->p_fd->fd_cdir);
544 	p->p_fd->fd_rdir = vp;
545 	vref(p->p_fd->fd_rdir);
546 	vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt));
547 	vn_unlock(vp);			/* leave ref intact */
548 	cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir);
549 	cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir);
550 
551 	kprintf("Mounting devfs\n");
552 	vfs_mountroot_devfs();
553 
554 	/*
555 	 * Need just enough stack to hold the faked-up "execve()" arguments.
556 	 */
557 	addr = trunc_page(USRSTACK - PAGE_SIZE);
558 	error = vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr,
559 			    PAGE_SIZE, PAGE_SIZE,
560 			    FALSE, VM_MAPTYPE_NORMAL,
561 			    VM_PROT_ALL, VM_PROT_ALL,
562 			    0);
563 	if (error)
564 		panic("init: couldn't allocate argument space");
565 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
566 	p->p_vmspace->vm_ssize = 1;
567 
568 	if ((var = kgetenv("init_path")) != NULL) {
569 		strncpy(init_path, var, sizeof init_path);
570 		init_path[sizeof init_path - 1] = 0;
571 	}
572 	if ((var = kgetenv("kern.fallback_elf_brand")) != NULL)
573 		fallback_elf_brand = strtol(var, NULL, 0);
574 
575 	for (path = init_path; *path != '\0'; path = next) {
576 		while (*path == ':')
577 			path++;
578 		if (*path == '\0')
579 			break;
580 		for (next = path; *next != '\0' && *next != ':'; next++)
581 			/* nothing */ ;
582 		if (bootverbose)
583 			kprintf("start_init: trying %.*s\n", (int)(next - path),
584 			    path);
585 
586 		/*
587 		 * Move out the boot flag argument.
588 		 */
589 		options = 0;
590 		ucp = (char *)USRSTACK;
591 		(void)subyte(--ucp, 0);		/* trailing zero */
592 		if (boothowto & RB_SINGLE) {
593 			(void)subyte(--ucp, 's');
594 			options = 1;
595 		}
596 #ifdef notyet
597                 if (boothowto & RB_FASTBOOT) {
598 			(void)subyte(--ucp, 'f');
599 			options = 1;
600 		}
601 #endif
602 
603 #ifdef BOOTCDROM
604 		(void)subyte(--ucp, 'C');
605 		options = 1;
606 #endif
607 		if (options == 0)
608 			(void)subyte(--ucp, '-');
609 		(void)subyte(--ucp, '-');		/* leading hyphen */
610 		arg1 = ucp;
611 
612 		/*
613 		 * Move out the file name (also arg 0).
614 		 */
615 		(void)subyte(--ucp, 0);
616 		for (s = next - 1; s >= path; s--)
617 			(void)subyte(--ucp, *s);
618 		arg0 = ucp;
619 
620 		/*
621 		 * Move out the arg pointers.
622 		 */
623 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
624 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
625 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
626 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
627 
628 		/*
629 		 * Point at the arguments.
630 		 */
631 		args.fname = arg0;
632 		args.argv = uap;
633 		args.envv = NULL;
634 
635 		/*
636 		 * Now try to exec the program.  If can't for any reason
637 		 * other than it doesn't exist, complain.
638 		 *
639 		 * Otherwise, return via fork_trampoline() all the way
640 		 * to user mode as init!
641 		 *
642 		 * WARNING!  We may have been moved to another cpu after
643 		 * acquiring the current user process designation.  The
644 		 * MP lock will migrate with us though so we still have to
645 		 * release it.
646 		 */
647 		if ((error = sys_execve(&args)) == 0) {
648 			rel_mplock();
649 			lp->lwp_proc->p_usched->acquire_curproc(lp);
650 			return;
651 		}
652 		if (error != ENOENT)
653 			kprintf("exec %.*s: error %d\n", (int)(next - path),
654 			    path, error);
655 	}
656 	kprintf("init: not found in path %s\n", init_path);
657 	panic("no init");
658 }
659 
660 /*
661  * Like kthread_create(), but runs in it's own address space.
662  * We do this early to reserve pid 1.
663  *
664  * Note special case - do not make it runnable yet.  Other work
665  * in progress will change this more.
666  */
667 static void
668 create_init(const void *udata __unused)
669 {
670 	int error;
671 	struct lwp *lp;
672 
673 	crit_enter();
674 	error = fork1(&lwp0, RFFDG | RFPROC, &initproc);
675 	if (error)
676 		panic("cannot fork init: %d", error);
677 	initproc->p_flag |= P_SYSTEM;
678 	lp = ONLY_LWP_IN_PROC(initproc);
679 	cpu_set_fork_handler(lp, start_init, NULL);
680 	crit_exit();
681 }
682 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
683 
684 /*
685  * Make it runnable now.
686  */
687 static void
688 kick_init(const void *udata __unused)
689 {
690 	start_forked_proc(&lwp0, initproc);
691 }
692 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
693 
694 /*
695  * Machine independant globaldata initialization
696  *
697  * WARNING!  Called from early boot, 'mycpu' may not work yet.
698  */
699 void
700 mi_gdinit(struct globaldata *gd, int cpuid)
701 {
702 	TAILQ_INIT(&gd->gd_systimerq);
703 	gd->gd_sysid_alloc = cpuid;	/* prime low bits for cpu lookup */
704 	gd->gd_cpuid = cpuid;
705 	gd->gd_cpumask = CPUMASK(cpuid);
706 	lwkt_gdinit(gd);
707 	vm_map_entry_reserve_cpu_init(gd);
708 	sleep_gdinit(gd);
709 }
710 
711