init_main.c revision 50477 
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: head/sys/kern/init_main.c 50477 1999-08-28 01:08:13Z peter $
43 */
44
45#include "opt_init_path.h"
46
47#include <sys/param.h>
48#include <sys/file.h>
49#include <sys/filedesc.h>
50#include <sys/kernel.h>
51#include <sys/mount.h>
52#include <sys/sysctl.h>
53#include <sys/proc.h>
54#include <sys/kthread.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
66#include <machine/cpu.h>
67
68#include <vm/vm.h>
69#include <vm/vm_param.h>
70#include <vm/vm_prot.h>
71#include <sys/lock.h>
72#include <vm/pmap.h>
73#include <vm/vm_map.h>
74#include <sys/user.h>
75#include <sys/copyright.h>
76
77extern struct linker_set	sysinit_set;	/* XXX */
78
79extern void mi_startup __P((void *framep));
80
81/* Components of the first process -- never freed. */
82static struct session session0;
83static struct pgrp pgrp0;
84struct	proc proc0;
85static struct pcred cred0;
86static struct procsig procsig0;
87static struct filedesc0 filedesc0;
88static struct plimit limit0;
89static struct vmspace vmspace0;
90struct	proc *initproc;
91
92int cmask = CMASK;
93extern	struct user *proc0paddr;
94
95struct	vnode *rootvp;
96int	boothowto = 0;		/* initialized so that it can be patched */
97
98struct	timeval boottime;
99SYSCTL_STRUCT(_kern, KERN_BOOTTIME, boottime, CTLFLAG_RD,
100    &boottime, timeval, "System boottime");
101
102/*
103 * Promiscuous argument pass for start_init()
104 *
105 * This is a kludge because we use a return from mi_startup() rather than a call
106 * to a new routine in locore.s to kick the kernel alive from locore.s.
107 */
108static void	*init_framep;
109
110/*
111 * This ensures that there is at least one entry so that the sysinit_set
112 * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
113 * executed.
114 */
115SYSINIT(placeholder, SI_SUB_DUMMY,SI_ORDER_ANY, NULL, NULL)
116
117/*
118 * The sysinit table itself.  Items are checked off as the are run.
119 * If we want to register new sysinit types, add them to newsysinit.
120 */
121struct sysinit **sysinit = (struct sysinit **)sysinit_set.ls_items;
122struct sysinit **newsysinit;
123
124/*
125 * Merge a new sysinit set into the current set, reallocating it if
126 * necessary.  This can only be called after malloc is running.
127 */
128void
129sysinit_add(set)
130	struct sysinit **set;
131{
132	struct sysinit **newset;
133	struct sysinit **sipp;
134	struct sysinit **xipp;
135	int count = 0;
136
137	if (newsysinit)
138		for (sipp = newsysinit; *sipp; sipp++)
139			count++;
140	else
141		for (sipp = sysinit; *sipp; sipp++)
142			count++;
143	for (sipp = set; *sipp; sipp++)
144		count++;
145	count++;		/* Trailing NULL */
146	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
147	if (newset == NULL)
148		panic("cannot malloc for sysinit");
149	xipp = newset;
150	if (newsysinit)
151		for (sipp = newsysinit; *sipp; sipp++)
152			*xipp++ = *sipp;
153	else
154		for (sipp = sysinit; *sipp; sipp++)
155			*xipp++ = *sipp;
156	for (sipp = set; *sipp; sipp++)
157		*xipp++ = *sipp;
158	*xipp = NULL;
159	if (newsysinit)
160		free(newsysinit, M_TEMP);
161	newsysinit = newset;
162}
163
164/*
165 * System startup; initialize the world, create process 0, mount root
166 * filesystem, and fork to create init and pagedaemon.  Most of the
167 * hard work is done in the lower-level initialization routines including
168 * startup(), which does memory initialization and autoconfiguration.
169 *
170 * This allows simple addition of new kernel subsystems that require
171 * boot time initialization.  It also allows substitution of subsystem
172 * (for instance, a scheduler, kernel profiler, or VM system) by object
173 * module.  Finally, it allows for optional "kernel threads".
174 */
175void
176mi_startup(framep)
177	void *framep;
178{
179
180	register struct sysinit **sipp;		/* system initialization*/
181	register struct sysinit **xipp;		/* interior loop of sort*/
182	register struct sysinit *save;		/* bubble*/
183
184	/*
185	 * Copy the locore.s frame pointer for proc0, this is forked into
186	 * all other processes.
187	 */
188	init_framep = framep;
189
190restart:
191	/*
192	 * Perform a bubble sort of the system initialization objects by
193	 * their subsystem (primary key) and order (secondary key).
194	 */
195	for (sipp = sysinit; *sipp; sipp++) {
196		for (xipp = sipp + 1; *xipp; xipp++) {
197			if ((*sipp)->subsystem < (*xipp)->subsystem ||
198			     ((*sipp)->subsystem == (*xipp)->subsystem &&
199			      (*sipp)->order < (*xipp)->order))
200				continue;	/* skip*/
201			save = *sipp;
202			*sipp = *xipp;
203			*xipp = save;
204		}
205	}
206
207	/*
208	 * Traverse the (now) ordered list of system initialization tasks.
209	 * Perform each task, and continue on to the next task.
210	 *
211	 * The last item on the list is expected to be the scheduler,
212	 * which will not return.
213	 */
214	for (sipp = sysinit; *sipp; sipp++) {
215
216		if ((*sipp)->subsystem == SI_SUB_DUMMY)
217			continue;	/* skip dummy task(s)*/
218
219		if ((*sipp)->subsystem == SI_SUB_DONE)
220			continue;
221
222		/* Call function */
223		(*((*sipp)->func))((*sipp)->udata);
224
225		/* Check off the one we're just done */
226		(*sipp)->subsystem = SI_SUB_DONE;
227
228		/* Check if we've installed more sysinit items via KLD */
229		if (newsysinit != NULL) {
230			if (sysinit != (struct sysinit **)sysinit_set.ls_items)
231				free(sysinit, M_TEMP);
232			sysinit = newsysinit;
233			newsysinit = NULL;
234			goto restart;
235		}
236	}
237
238	panic("Shouldn't get here!");
239	/* NOTREACHED*/
240}
241
242
243/*
244 ***************************************************************************
245 ****
246 **** The following SYSINIT's belong elsewhere, but have not yet
247 **** been moved.
248 ****
249 ***************************************************************************
250 */
251#ifdef OMIT
252/*
253 * Handled by vfs_mountroot (bad idea) at this time... should be
254 * done the same as 4.4Lite2.
255 */
256SYSINIT(swapinit, SI_SUB_SWAP, SI_ORDER_FIRST, swapinit, NULL)
257#endif	/* OMIT*/
258
259static void print_caddr_t __P((void *data));
260static void
261print_caddr_t(data)
262	void *data;
263{
264	printf("%s", (char *)data);
265}
266SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
267
268
269/*
270 ***************************************************************************
271 ****
272 **** The two following SYSINT's are proc0 specific glue code.  I am not
273 **** convinced that they can not be safely combined, but their order of
274 **** operation has been maintained as the same as the original init_main.c
275 **** for right now.
276 ****
277 **** These probably belong in init_proc.c or kern_proc.c, since they
278 **** deal with proc0 (the fork template process).
279 ****
280 ***************************************************************************
281 */
282/* ARGSUSED*/
283static void proc0_init __P((void *dummy));
284static void
285proc0_init(dummy)
286	void *dummy;
287{
288	register struct proc		*p;
289	register struct filedesc0	*fdp;
290	register unsigned i;
291
292	p = &proc0;
293
294	/*
295	 * Initialize process and pgrp structures.
296	 */
297	procinit();
298
299	/*
300	 * Initialize sleep queue hash table
301	 */
302	sleepinit();
303
304	/*
305	 * additional VM structures
306	 */
307	vm_init2();
308
309	/*
310	 * Create process 0 (the swapper).
311	 */
312	LIST_INSERT_HEAD(&allproc, p, p_list);
313	p->p_pgrp = &pgrp0;
314	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
315	LIST_INIT(&pgrp0.pg_members);
316	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
317
318	pgrp0.pg_session = &session0;
319	session0.s_count = 1;
320	session0.s_leader = p;
321
322	p->p_sysent = &aout_sysvec;
323
324	p->p_flag = P_INMEM | P_SYSTEM;
325	p->p_stat = SRUN;
326	p->p_nice = NZERO;
327	p->p_rtprio.type = RTP_PRIO_NORMAL;
328	p->p_rtprio.prio = 0;
329
330/*
331 * Link for kernel based threads
332 */
333	p->p_peers = 0;
334	p->p_leader = p;
335
336	bcopy("swapper", p->p_comm, sizeof ("swapper"));
337
338	/* Create credentials. */
339	cred0.p_refcnt = 1;
340	p->p_cred = &cred0;
341	p->p_ucred = crget();
342	p->p_ucred->cr_ngroups = 1;	/* group 0 */
343
344	/* Don't jail it */
345	p->p_prison = 0;
346
347	/* Create procsig. */
348	p->p_procsig = &procsig0;
349	p->p_procsig->ps_refcnt = 1;
350
351	/* Create the file descriptor table. */
352	fdp = &filedesc0;
353	p->p_fd = &fdp->fd_fd;
354	fdp->fd_fd.fd_refcnt = 1;
355	fdp->fd_fd.fd_cmask = cmask;
356	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
357	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
358	fdp->fd_fd.fd_nfiles = NDFILE;
359
360	/* Create the limits structures. */
361	p->p_limit = &limit0;
362	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
363		limit0.pl_rlimit[i].rlim_cur =
364		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
365	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
366	    limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
367	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
368	    limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
369	i = ptoa(cnt.v_free_count);
370	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
371	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
372	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
373	limit0.p_cpulimit = RLIM_INFINITY;
374	limit0.p_refcnt = 1;
375
376
377	/* Allocate a prototype map so we have something to fork. */
378	pmap_pinit0(vmspace_pmap(&vmspace0));
379	p->p_vmspace = &vmspace0;
380	vmspace0.vm_refcnt = 1;
381	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS),
382	    trunc_page(VM_MAXUSER_ADDRESS));
383	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
384	p->p_addr = proc0paddr;				/* XXX */
385
386#ifdef cpu_set_init_frame
387	/*
388	 * proc0 needs to have a coherent frame base in its stack.
389	 */
390	cpu_set_init_frame(p, init_framep);			/* XXX! */
391#endif
392
393	/*
394	 * We continue to place resource usage info and signal
395	 * actions in the user struct so they're pageable.
396	 */
397	p->p_stats = &p->p_addr->u_stats;
398	p->p_sigacts = &p->p_addr->u_sigacts;
399
400	/*
401	 * Charge root for one process.
402	 */
403	(void)chgproccnt(0, 1);
404
405	/*
406	 * Initialize the current process pointer (curproc) before
407	 * any possible traps/probes to simplify trap processing.
408	 */
409	SET_CURPROC(p);
410
411}
412SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
413
414/* ARGSUSED*/
415static void proc0_post __P((void *dummy));
416static void
417proc0_post(dummy)
418	void *dummy;
419{
420	struct timespec ts;
421
422	/*
423	 * Now we can look at the time, having had a chance to verify the
424	 * time from the file system.  Pretend that proc0 started now.
425	 */
426	microtime(&proc0.p_stats->p_start);
427	proc0.p_runtime = 0;
428	microuptime(&switchtime);
429	switchticks = ticks;
430
431	/*
432	 * Give the ``random'' number generator a thump.
433	 * XXX: Does read_random() contain enough bits to be used here ?
434	 */
435	nanotime(&ts);
436	srandom(ts.tv_sec ^ ts.tv_nsec);
437
438	/* Initialize signal state for process 0. */
439	siginit(&proc0);
440}
441SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
442
443
444
445
446/*
447 ***************************************************************************
448 ****
449 **** The following SYSINIT's and glue code should be moved to the
450 **** respective files on a per subsystem basis.
451 ****
452 ***************************************************************************
453 */
454
455/* ARGSUSED */
456static void root_conf __P((void *dummy));
457static void
458root_conf(dummy)
459	void *dummy;
460{
461	cpu_rootconf();
462}
463SYSINIT(root_conf, SI_SUB_ROOT_CONF, SI_ORDER_FIRST, root_conf, NULL)
464
465/* ARGSUSED*/
466static void xxx_vfs_root_fdtab __P((void *dummy));
467static void
468xxx_vfs_root_fdtab(dummy)
469	void *dummy;
470{
471	register struct filedesc0	*fdp = &filedesc0;
472
473	/* Get the vnode for '/'.  Set fdp->fd_fd.fd_cdir to reference it. */
474	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
475		panic("cannot find root vnode");
476	fdp->fd_fd.fd_cdir = rootvnode;
477	VREF(fdp->fd_fd.fd_cdir);
478	VOP_UNLOCK(rootvnode, 0, &proc0);
479	fdp->fd_fd.fd_rdir = rootvnode;
480}
481SYSINIT(retrofit, SI_SUB_ROOT_FDTAB, SI_ORDER_FIRST, xxx_vfs_root_fdtab, NULL)
482
483
484/*
485 ***************************************************************************
486 ****
487 **** The following code probably belongs in another file, like
488 **** kern/init_init.c.  It is here for two reasons only:
489 ****
490 ****	1)	This code returns to startup the system; this is
491 ****		abnormal for a kernel thread.
492 ****	2)	This code promiscuously uses init_frame
493 ****
494 ***************************************************************************
495 */
496
497extern void prepare_usermode __P((void));
498static void create_init __P((const void *dummy));
499static void start_init __P((void *dummy));
500SYSINIT(init,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, create_init, NULL)
501
502/*
503 * Like kthread_create(), but runs in it's own address space.
504 */
505static void
506create_init(udata)
507	const void *udata;
508{
509	int error;
510
511	error = fork1(&proc0, RFFDG | RFPROC, &initproc);
512	if (error)
513		panic("cannot fork init: %d\n", error);
514	initproc->p_flag |= P_INMEM | P_SYSTEM;
515	cpu_set_fork_handler(initproc, start_init, NULL);
516}
517
518/*
519 * List of paths to try when searching for "init".
520 */
521static char init_path[MAXPATHLEN] =
522#ifdef	INIT_PATH
523    __XSTRING(INIT_PATH);
524#else
525    "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
526#endif
527SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
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 */
533static void
534start_init(dummy)
535	void *dummy;
536{
537	vm_offset_t addr;
538	struct execve_args args;
539	int options, error;
540	char *var, *path, *next, *s;
541	char *ucp, **uap, *arg0, *arg1;
542	struct proc *p;
543
544	p = curproc;
545
546	/*
547	 * Need just enough stack to hold the faked-up "execve()" arguments.
548	 */
549	addr = trunc_page(USRSTACK - PAGE_SIZE);
550	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
551			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
552		panic("init: couldn't allocate argument space");
553	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
554	p->p_vmspace->vm_ssize = 1;
555
556	if ((var = getenv("init_path")) != NULL) {
557		strncpy(init_path, var, sizeof init_path);
558		init_path[sizeof init_path - 1] = 0;
559	}
560
561	for (path = init_path; *path != '\0'; path = next) {
562		while (*path == ':')
563			path++;
564		if (*path == '\0')
565			break;
566		for (next = path; *next != '\0' && *next != ':'; next++)
567			/* nothing */ ;
568		if (bootverbose)
569			printf("start_init: trying %.*s\n", (int)(next - path),
570			    path);
571
572		/*
573		 * Move out the boot flag argument.
574		 */
575		options = 0;
576		ucp = (char *)USRSTACK;
577		(void)subyte(--ucp, 0);		/* trailing zero */
578		if (boothowto & RB_SINGLE) {
579			(void)subyte(--ucp, 's');
580			options = 1;
581		}
582#ifdef notyet
583                if (boothowto & RB_FASTBOOT) {
584			(void)subyte(--ucp, 'f');
585			options = 1;
586		}
587#endif
588
589#ifdef BOOTCDROM
590		(void)subyte(--ucp, 'C');
591		options = 1;
592#endif
593		if (options == 0)
594			(void)subyte(--ucp, '-');
595		(void)subyte(--ucp, '-');		/* leading hyphen */
596		arg1 = ucp;
597
598		/*
599		 * Move out the file name (also arg 0).
600		 */
601		(void)subyte(--ucp, 0);
602		for (s = next - 1; s >= path; s--)
603			(void)subyte(--ucp, *s);
604		arg0 = ucp;
605
606		/*
607		 * Move out the arg pointers.
608		 */
609		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
610		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
611		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
612		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
613
614		/*
615		 * Point at the arguments.
616		 */
617		args.fname = arg0;
618		args.argv = uap;
619		args.envv = NULL;
620
621		/*
622		 * Now try to exec the program.  If can't for any reason
623		 * other than it doesn't exist, complain.
624		 *
625		 * Otherwise, return via the fork trampoline all the way
626		 * to user mode as init!
627		 */
628		if ((error = execve(p, &args)) == 0) {
629			prepare_usermode();
630			return;
631		}
632		if (error != ENOENT)
633			printf("exec %.*s: error %d\n", (int)(next - path),
634			    path, error);
635	}
636	printf("init: not found\n");
637	panic("no init");
638}
639