xref: /dflybsd-src/sys/kern/init_main.c (revision a62226e46c982d037de05e1bb0894805c0b7a32f)
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. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  *
37  *	@(#)init_main.c	8.9 (Berkeley) 1/21/94
38  * $FreeBSD: src/sys/kern/init_main.c,v 1.134.2.8 2003/06/06 20:21:32 tegge Exp $
39  */
40 
41 #include "opt_init_path.h"
42 
43 #include <sys/param.h>
44 #include <sys/file.h>
45 #include <sys/filedesc.h>
46 #include <sys/kernel.h>
47 #include <sys/mount.h>
48 #include <sys/sysctl.h>
49 #include <sys/proc.h>
50 #include <sys/resourcevar.h>
51 #include <sys/signalvar.h>
52 #include <sys/systm.h>
53 #include <sys/vnode.h>
54 #include <sys/sysent.h>
55 #include <sys/reboot.h>
56 #include <sys/sysproto.h>
57 #include <sys/vmmeter.h>
58 #include <sys/unistd.h>
59 #include <sys/malloc.h>
60 #include <sys/machintr.h>
61 
62 #include <sys/refcount.h>
63 #include <sys/file2.h>
64 #include <sys/thread2.h>
65 #include <sys/sysref2.h>
66 #include <sys/spinlock2.h>
67 #include <sys/mplock2.h>
68 
69 #include <machine/cpu.h>
70 
71 #include <vm/vm.h>
72 #include <vm/vm_param.h>
73 #include <sys/lock.h>
74 #include <vm/pmap.h>
75 #include <vm/vm_map.h>
76 #include <vm/vm_extern.h>
77 #include <sys/user.h>
78 #include <sys/copyright.h>
79 
80 int vfs_mountroot_devfs(void);
81 
82 /* Components of the first process -- never freed. */
83 static struct session session0;
84 static struct pgrp pgrp0;
85 static struct sigacts sigacts0;
86 static struct filedesc filedesc0;
87 static struct plimit limit0;
88 static struct vmspace vmspace0;
89 struct proc *initproc;
90 struct proc proc0;
91 struct lwp lwp0;
92 struct thread thread0;
93 
94 int cmask = CMASK;
95 u_int cpu_mi_feature;
96 cpumask_t usched_global_cpumask;
97 extern	struct user *proc0paddr;
98 
99 int	boothowto = 0;		/* initialized so that it can be patched */
100 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
101     "Reboot flags, from console subsystem");
102 SYSCTL_ULONG(_kern, OID_AUTO, usched_global_cpumask, CTLFLAG_RW,
103     &usched_global_cpumask, 0, "global user scheduler cpumask");
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 	RB_INIT(&proc0.p_lwp_tree);
167 	spin_init(&proc0.p_spin);
168 	lwkt_token_init(&proc0.p_token, "iproc");
169 	proc0.p_lasttid = 0;	/* +1 = next TID */
170 	lwp_rb_tree_RB_INSERT(&proc0.p_lwp_tree, &lwp0);
171 	lwp0.lwp_thread = &thread0;
172 	lwp0.lwp_proc = &proc0;
173 	proc0.p_usched = usched_init();
174 	CPUMASK_ASSALLONES(lwp0.lwp_cpumask);
175 	lwkt_token_init(&lwp0.lwp_token, "lwp_token");
176 	spin_init(&lwp0.lwp_spin);
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(__x86_64__) && 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 = (void *)((long)(intptr_t)sysinit + 4);
215 		}
216 #endif
217 		sysinit_end = SET_LIMIT(sysinit_set);
218 	}
219 #if defined(__x86_64__) && 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 #if 0
256 		if (bootverbose)
257 			kprintf("(%08x-%p)\n", sip->subsystem, sip->func);
258 #endif
259 
260 		/* Call function */
261 		(*(sip->func))(sip->udata);
262 
263 		/* Check off the one we're just done */
264 		sip->subsystem = SI_SPECIAL_DONE;
265 
266 		/* Check if we've installed more sysinit items via KLD */
267 		if (newsysinit != NULL) {
268 			if (sysinit != SET_BEGIN(sysinit_set))
269 				kfree(sysinit, M_TEMP);
270 			sysinit = newsysinit;
271 			sysinit_end = newsysinit_end;
272 			newsysinit = NULL;
273 			newsysinit_end = NULL;
274 			goto restart;
275 		}
276 	}
277 
278 	panic("Shouldn't get here!");
279 	/* NOTREACHED*/
280 }
281 
282 
283 /*
284  ***************************************************************************
285  ****
286  **** The following SYSINIT's belong elsewhere, but have not yet
287  **** been moved.
288  ****
289  ***************************************************************************
290  */
291 static void
292 print_caddr_t(void *data)
293 {
294 	kprintf("%s", (char *)data);
295 }
296 SYSINIT(announce, SI_BOOT1_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
297 
298 /*
299  * Leave the critical section that protected us from spurious interrupts
300  * so device probes work.
301  */
302 static void
303 leavecrit(void *dummy __unused)
304 {
305 	MachIntrABI.stabilize();
306 	cpu_enable_intr();
307 	MachIntrABI.cleanup();
308 	crit_exit();
309 	KKASSERT(!IN_CRITICAL_SECT(curthread));
310 
311 	if (bootverbose)
312 		kprintf("Leaving critical section, allowing interrupts\n");
313 }
314 SYSINIT(leavecrit, SI_BOOT2_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL)
315 
316 /*
317  * This is called after the threading system is up and running,
318  * including the softclock, clock interrupts, and SMP.
319  */
320 static void
321 tsleepworks(void *dummy __unused)
322 {
323 	tsleep_now_works = 1;
324 }
325 SYSINIT(tsleepworks, SI_BOOT2_FINISH_SMP, SI_ORDER_SECOND, tsleepworks, NULL)
326 
327 /*
328  * This is called after devices have configured.  Tell the kernel we are
329  * no longer in cold boot.
330  */
331 static void
332 endofcoldboot(void *dummy __unused)
333 {
334 	cold = 0;
335 }
336 SYSINIT(endofcoldboot, SI_SUB_ISWARM, SI_ORDER_ANY, endofcoldboot, NULL)
337 
338 /*
339  ***************************************************************************
340  ****
341  **** The two following SYSINT's are proc0 specific glue code.  I am not
342  **** convinced that they can not be safely combined, but their order of
343  **** operation has been maintained as the same as the original init_main.c
344  **** for right now.
345  ****
346  **** These probably belong in init_proc.c or kern_proc.c, since they
347  **** deal with proc0 (the fork template process).
348  ****
349  ***************************************************************************
350  */
351 /* ARGSUSED*/
352 static void
353 proc0_init(void *dummy __unused)
354 {
355 	struct proc *p;
356 	struct lwp *lp;
357 
358 	p = &proc0;
359 	lp = &lwp0;
360 
361 	/*
362 	 * Initialize osrel
363 	 */
364 	p->p_osrel = osreldate;
365 
366 	/*
367 	 * Initialize process and pgrp structures.
368 	 */
369 	procinit();
370 
371 	/*
372 	 * additional VM structures
373 	 */
374 	vm_init2();
375 
376 	/*
377 	 * Create process 0 (the swapper).
378 	 */
379 	procinsertinit(p);
380 	pgrpinsertinit(&pgrp0);
381 	LIST_INIT(&pgrp0.pg_members);
382 	lwkt_token_init(&pgrp0.pg_token, "pgrp0");
383 	refcount_init(&pgrp0.pg_refs, 1);
384 	lockinit(&pgrp0.pg_lock, "pgwt0", 0, 0);
385 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
386 
387 	pgrp0.pg_session = &session0;
388 	session0.s_count = 1;
389 	session0.s_leader = p;
390 	sessinsertinit(&session0);
391 
392 	pgref(&pgrp0);
393 	p->p_pgrp = &pgrp0;
394 
395 	p->p_sysent = &aout_sysvec;
396 
397 	p->p_flags = P_SYSTEM;
398 	p->p_stat = SACTIVE;
399 	lp->lwp_stat = LSRUN;
400 	p->p_nice = NZERO;
401 	p->p_rtprio.type = RTP_PRIO_NORMAL;
402 	p->p_rtprio.prio = 0;
403 	lp->lwp_rtprio = p->p_rtprio;
404 
405 	p->p_peers = NULL;
406 	p->p_leader = p;
407 
408 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
409 	bcopy("swapper", thread0.td_comm, sizeof ("swapper"));
410 
411 	/* Create credentials. */
412 	p->p_ucred = crget();
413 	p->p_ucred->cr_ruidinfo = uifind(0);
414 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
415 	p->p_ucred->cr_uidinfo = uifind(0);
416 	thread0.td_ucred = crhold(p->p_ucred);	/* bootstrap fork1() */
417 
418 	/* Don't jail it */
419 	p->p_ucred->cr_prison = NULL;
420 
421 	/* Create sigacts. */
422 	p->p_sigacts = &sigacts0;
423 	refcount_init(&p->p_sigacts->ps_refcnt, 1);
424 
425 	/* Initialize signal state for process 0. */
426 	siginit(p);
427 
428 	/* Create the file descriptor table. */
429 	fdinit_bootstrap(p, &filedesc0, cmask);
430 
431 	/* Create the limits structures. */
432 	plimit_init0(&limit0);
433 	p->p_limit = &limit0;
434 
435 	/* Allocate a prototype map so we have something to fork. */
436 	pmap_pinit0(vmspace_pmap(&vmspace0));
437 	p->p_vmspace = &vmspace0;
438 	lp->lwp_vmspace = p->p_vmspace;
439 	sysref_init(&vmspace0.vm_sysref, &vmspace_sysref_class);
440 	vm_map_init(&vmspace0.vm_map,
441 		    round_page(VM_MIN_USER_ADDRESS),
442 		    trunc_page(VM_MAX_USER_ADDRESS),
443 		    vmspace_pmap(&vmspace0));
444 	sysref_activate(&vmspace0.vm_sysref);
445 
446 	kqueue_init(&lwp0.lwp_kqueue, &filedesc0);
447 
448 	/*
449 	 * Charge root for one process.
450 	 */
451 	(void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0);
452 	vm_init_limits(p);
453 }
454 SYSINIT(p0init, SI_BOOT2_PROC0, SI_ORDER_FIRST, proc0_init, NULL)
455 
456 static int proc0_post_callback(struct proc *p, void *data __unused);
457 
458 /* ARGSUSED*/
459 static void
460 proc0_post(void *dummy __unused)
461 {
462 	struct timespec ts;
463 
464 	/*
465 	 * Now we can look at the time, having had a chance to verify the
466 	 * time from the file system.  Pretend that proc0 started now.
467 	 */
468 	allproc_scan(proc0_post_callback, NULL);
469 
470 	/*
471 	 * Give the ``random'' number generator a thump.
472 	 * XXX: Does read_random() contain enough bits to be used here ?
473 	 */
474 	nanotime(&ts);
475 	skrandom(ts.tv_sec ^ ts.tv_nsec);
476 }
477 
478 static int
479 proc0_post_callback(struct proc *p, void *data __unused)
480 {
481 	microtime(&p->p_start);
482 	return(0);
483 }
484 
485 SYSINIT(p0post, SI_SUB_PROC0_POST, SI_ORDER_FIRST, proc0_post, NULL)
486 
487 /*
488  ***************************************************************************
489  ****
490  **** The following SYSINIT's and glue code should be moved to the
491  **** respective files on a per subsystem basis.
492  ****
493  ***************************************************************************
494  */
495 
496 
497 /*
498  ***************************************************************************
499  ****
500  **** The following code probably belongs in another file, like
501  **** kern/init_init.c.
502  ****
503  ***************************************************************************
504  */
505 
506 /*
507  * List of paths to try when searching for "init".
508  */
509 static char init_path[MAXPATHLEN] =
510 #ifdef	INIT_PATH
511     __XSTRING(INIT_PATH);
512 #else
513     "/sbin/init:/sbin/oinit:/sbin/init.bak";
514 #endif
515 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
516 
517 /*
518  * Shutdown timeout of init(8).
519  * Unused within kernel, but used to control init(8), hence do not remove.
520  */
521 #ifndef INIT_SHUTDOWN_TIMEOUT
522 #define INIT_SHUTDOWN_TIMEOUT 120
523 #endif
524 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
525 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
526 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
527 	"Unused within kernel, but used to control init(8)");
528 
529 /*
530  * Start the initial user process; try exec'ing each pathname in init_path.
531  * The program is invoked with one argument containing the boot flags.
532  */
533 static void
534 start_init(void *dummy, struct trapframe *frame)
535 {
536 	vm_offset_t addr;
537 	struct execve_args args;
538 	int options, error;
539 	char *var, *path, *next, *s;
540 	char *ucp, **uap, *arg0, *arg1;
541 	struct proc *p;
542 	struct lwp *lp;
543 	struct mount *mp;
544 	struct vnode *vp;
545 	char *env;
546 
547         /*
548 	 * This is passed in by the bootloader
549          */
550 	env = kgetenv("kernelname");
551 	if (env != NULL)
552 		strlcpy(kernelname, env, sizeof(kernelname));
553 
554 	/*
555 	 * The MP lock is not held on entry.  We release it before
556 	 * returning to userland.
557 	 */
558 	get_mplock();
559 	p = curproc;
560 
561 	lp = ONLY_LWP_IN_PROC(p);
562 
563 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
564 	mp = mountlist_boot_getfirst();
565 	if (VFS_ROOT(mp, &vp))
566 		panic("cannot find root vnode");
567 	if (mp->mnt_ncmountpt.ncp == NULL) {
568 		cache_allocroot(&mp->mnt_ncmountpt, mp, vp);
569 		cache_unlock(&mp->mnt_ncmountpt);	/* leave ref intact */
570 	}
571 	p->p_fd->fd_cdir = vp;
572 	vref(p->p_fd->fd_cdir);
573 	p->p_fd->fd_rdir = vp;
574 	vref(p->p_fd->fd_rdir);
575 	vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt));
576 	vn_unlock(vp);			/* leave ref intact */
577 	cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir);
578 	cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir);
579 
580 	kprintf("Mounting devfs\n");
581 	vfs_mountroot_devfs();
582 
583 	/*
584 	 * Need just enough stack to hold the faked-up "execve()" arguments.
585 	 */
586 	addr = trunc_page(USRSTACK - PAGE_SIZE);
587 	error = vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr,
588 			    PAGE_SIZE, PAGE_SIZE,
589 			    FALSE, VM_MAPTYPE_NORMAL,
590 			    VM_PROT_ALL, VM_PROT_ALL,
591 			    0);
592 	if (error)
593 		panic("init: couldn't allocate argument space");
594 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
595 	p->p_vmspace->vm_ssize = 1;
596 
597 	if ((var = kgetenv("init_path")) != NULL) {
598 		strncpy(init_path, var, sizeof init_path);
599 		init_path[sizeof init_path - 1] = 0;
600 	}
601 
602 	for (path = init_path; *path != '\0'; path = next) {
603 		while (*path == ':')
604 			path++;
605 		if (*path == '\0')
606 			break;
607 		for (next = path; *next != '\0' && *next != ':'; next++)
608 			/* nothing */ ;
609 		if (bootverbose)
610 			kprintf("start_init: trying %.*s\n", (int)(next - path),
611 			    path);
612 
613 		/*
614 		 * Move out the boot flag argument.
615 		 */
616 		options = 0;
617 		ucp = (char *)USRSTACK;
618 		(void)subyte(--ucp, 0);		/* trailing zero */
619 		if (boothowto & RB_SINGLE) {
620 			(void)subyte(--ucp, 's');
621 			options = 1;
622 		}
623 #ifdef notyet
624                 if (boothowto & RB_FASTBOOT) {
625 			(void)subyte(--ucp, 'f');
626 			options = 1;
627 		}
628 #endif
629 
630 #ifdef BOOTCDROM
631 		(void)subyte(--ucp, 'C');
632 		options = 1;
633 #endif
634 		if (options == 0)
635 			(void)subyte(--ucp, '-');
636 		(void)subyte(--ucp, '-');		/* leading hyphen */
637 		arg1 = ucp;
638 
639 		/*
640 		 * Move out the file name (also arg 0).
641 		 */
642 		(void)subyte(--ucp, 0);
643 		for (s = next - 1; s >= path; s--)
644 			(void)subyte(--ucp, *s);
645 		arg0 = ucp;
646 
647 		/*
648 		 * Move out the arg pointers.
649 		 */
650 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
651 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
652 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
653 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
654 
655 		/*
656 		 * Point at the arguments.
657 		 */
658 		args.fname = arg0;
659 		args.argv = uap;
660 		args.envv = NULL;
661 
662 		/*
663 		 * Now try to exec the program.  If can't for any reason
664 		 * other than it doesn't exist, complain.
665 		 *
666 		 * Otherwise, return via fork_trampoline() all the way
667 		 * to user mode as init!
668 		 *
669 		 * WARNING!  We may have been moved to another cpu after
670 		 * acquiring the current user process designation.  The
671 		 * MP lock will migrate with us though so we still have to
672 		 * release it.
673 		 */
674 		if ((error = sys_execve(&args)) == 0) {
675 			rel_mplock();
676 			lp->lwp_proc->p_usched->acquire_curproc(lp);
677 			return;
678 		}
679 		if (error != ENOENT)
680 			kprintf("exec %.*s: error %d\n", (int)(next - path),
681 			    path, error);
682 	}
683 	kprintf("init: not found in path %s\n", init_path);
684 	panic("no init");
685 }
686 
687 /*
688  * Like kthread_create(), but runs in it's own address space.
689  * We do this early to reserve pid 1.
690  *
691  * Note special case - do not make it runnable yet.  Other work
692  * in progress will change this more.
693  */
694 static void
695 create_init(const void *udata __unused)
696 {
697 	int error;
698 	struct lwp *lp;
699 
700 	crit_enter();
701 	error = fork1(&lwp0, RFFDG | RFPROC, &initproc);
702 	if (error)
703 		panic("cannot fork init: %d", error);
704 	initproc->p_flags |= P_SYSTEM;
705 	lp = ONLY_LWP_IN_PROC(initproc);
706 	cpu_set_fork_handler(lp, start_init, NULL);
707 	crit_exit();
708 }
709 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
710 
711 /*
712  * Make it runnable now.
713  */
714 static void
715 kick_init(const void *udata __unused)
716 {
717 	start_forked_proc(&lwp0, initproc);
718 }
719 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
720 
721 /*
722  * Machine independant globaldata initialization
723  *
724  * WARNING!  Called from early boot, 'mycpu' may not work yet.
725  */
726 void
727 mi_gdinit(struct globaldata *gd, int cpuid)
728 {
729 	TAILQ_INIT(&gd->gd_systimerq);
730 	gd->gd_sysid_alloc = cpuid;	/* prime low bits for cpu lookup */
731 	gd->gd_cpuid = cpuid;
732 	CPUMASK_ASSBIT(gd->gd_cpumask, cpuid);
733 	lwkt_gdinit(gd);
734 	vm_map_entry_reserve_cpu_init(gd);
735 	sleep_gdinit(gd);
736 	ATOMIC_CPUMASK_ORBIT(usched_global_cpumask, cpuid);
737 }
738 
739