HAMMER 61E/Many: Stabilization, Performance
[dragonfly.git] / sys / kern / init_main.c
blob0075c19bf0ca2f4fe1fe9621cd4999abf1eebe02
1 /*
2 * Copyright (c) 1995 Terrence R. Lambert
3 * All rights reserved.
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.
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.
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.
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 $
46 #include "opt_init_path.h"
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 #include <sys/file2.h>
67 #include <sys/thread2.h>
68 #include <sys/sysref2.h>
69 #include <sys/spinlock2.h>
71 #include <machine/cpu.h>
73 #include <vm/vm.h>
74 #include <vm/vm_param.h>
75 #include <sys/lock.h>
76 #include <vm/pmap.h>
77 #include <vm/vm_map.h>
78 #include <vm/vm_extern.h>
79 #include <sys/user.h>
80 #include <sys/copyright.h>
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;
94 int cmask = CMASK;
95 extern struct user *proc0paddr;
96 extern int fallback_elf_brand;
98 int boothowto = 0; /* initialized so that it can be patched */
99 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
102 * This ensures that there is at least one entry so that the sysinit_set
103 * symbol is not undefined. A subsystem ID of SI_SPECIAL_DUMMY is never
104 * executed.
106 SYSINIT(placeholder, SI_SPECIAL_DUMMY, SI_ORDER_ANY, NULL, NULL)
109 * The sysinit table itself. Items are checked off as the are run.
110 * If we want to register new sysinit types, add them to newsysinit.
112 SET_DECLARE(sysinit_set, struct sysinit);
113 struct sysinit **sysinit, **sysinit_end;
114 struct sysinit **newsysinit, **newsysinit_end;
118 * Merge a new sysinit set into the current set, reallocating it if
119 * necessary. This can only be called after malloc is running.
121 void
122 sysinit_add(struct sysinit **set, struct sysinit **set_end)
124 struct sysinit **newset;
125 struct sysinit **sipp;
126 struct sysinit **xipp;
127 int count;
129 count = set_end - set;
130 if (newsysinit)
131 count += newsysinit_end - newsysinit;
132 else
133 count += sysinit_end - sysinit;
134 newset = kmalloc(count * sizeof(*sipp), M_TEMP, M_WAITOK);
135 xipp = newset;
136 if (newsysinit) {
137 for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
138 *xipp++ = *sipp;
139 } else {
140 for (sipp = sysinit; sipp < sysinit_end; sipp++)
141 *xipp++ = *sipp;
143 for (sipp = set; sipp < set_end; sipp++)
144 *xipp++ = *sipp;
145 if (newsysinit)
146 kfree(newsysinit, M_TEMP);
147 newsysinit = newset;
148 newsysinit_end = newset + count;
152 * Callbacks from machine-dependant startup code (e.g. init386) to set
153 * up low level entities related to cpu #0's globaldata.
155 * Called from very low level boot code.
157 void
158 mi_proc0init(struct globaldata *gd, struct user *proc0paddr)
160 lwkt_init_thread(&thread0, proc0paddr, LWKT_THREAD_STACK, 0, gd);
161 lwkt_set_comm(&thread0, "thread0");
162 RB_INIT(&proc0.p_lwp_tree);
163 spin_init(&proc0.p_spin);
164 proc0.p_lasttid = 0; /* +1 = next TID */
165 lwp_rb_tree_RB_INSERT(&proc0.p_lwp_tree, &lwp0);
166 lwp0.lwp_thread = &thread0;
167 lwp0.lwp_proc = &proc0;
168 proc0.p_usched = usched_init();
169 lwp0.lwp_cpumask = 0xFFFFFFFF;
170 varsymset_init(&proc0.p_varsymset, NULL);
171 thread0.td_flags |= TDF_RUNNING;
172 thread0.td_proc = &proc0;
173 thread0.td_lwp = &lwp0;
174 thread0.td_switch = cpu_heavy_switch; /* YYY eventually LWKT */
175 lwkt_schedule_self(curthread);
179 * System startup; initialize the world, create process 0, mount root
180 * filesystem, and fork to create init and pagedaemon. Most of the
181 * hard work is done in the lower-level initialization routines including
182 * startup(), which does memory initialization and autoconfiguration.
184 * This allows simple addition of new kernel subsystems that require
185 * boot time initialization. It also allows substitution of subsystem
186 * (for instance, a scheduler, kernel profiler, or VM system) by object
187 * module. Finally, it allows for optional "kernel threads".
189 void
190 mi_startup(void)
192 struct sysinit *sip; /* system initialization*/
193 struct sysinit **sipp; /* system initialization*/
194 struct sysinit **xipp; /* interior loop of sort*/
195 struct sysinit *save; /* bubble*/
197 if (sysinit == NULL) {
198 sysinit = SET_BEGIN(sysinit_set);
199 sysinit_end = SET_LIMIT(sysinit_set);
202 restart:
204 * Perform a bubble sort of the system initialization objects by
205 * their subsystem (primary key) and order (secondary key).
207 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
208 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
209 if ((*sipp)->subsystem < (*xipp)->subsystem ||
210 ((*sipp)->subsystem == (*xipp)->subsystem &&
211 (*sipp)->order <= (*xipp)->order))
212 continue; /* skip*/
213 save = *sipp;
214 *sipp = *xipp;
215 *xipp = save;
220 * Traverse the (now) ordered list of system initialization tasks.
221 * Perform each task, and continue on to the next task.
223 * The last item on the list is expected to be the scheduler,
224 * which will not return.
226 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
227 sip = *sipp;
228 if (sip->subsystem == SI_SPECIAL_DUMMY)
229 continue; /* skip dummy task(s)*/
231 if (sip->subsystem == SI_SPECIAL_DONE)
232 continue;
234 /* Call function */
235 (*(sip->func))(sip->udata);
237 /* Check off the one we're just done */
238 sip->subsystem = SI_SPECIAL_DONE;
240 /* Check if we've installed more sysinit items via KLD */
241 if (newsysinit != NULL) {
242 if (sysinit != SET_BEGIN(sysinit_set))
243 kfree(sysinit, M_TEMP);
244 sysinit = newsysinit;
245 sysinit_end = newsysinit_end;
246 newsysinit = NULL;
247 newsysinit_end = NULL;
248 goto restart;
252 panic("Shouldn't get here!");
253 /* NOTREACHED*/
258 ***************************************************************************
259 ****
260 **** The following SYSINIT's belong elsewhere, but have not yet
261 **** been moved.
262 ****
263 ***************************************************************************
265 static void
266 print_caddr_t(void *data __unused)
268 kprintf("%s", (char *)data);
270 SYSINIT(announce, SI_BOOT1_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
273 * Leave the critical section that protected us from spurious interrupts
274 * so device probes work.
276 static void
277 leavecrit(void *dummy __unused)
279 MachIntrABI.finalize();
280 cpu_enable_intr();
281 MachIntrABI.cleanup();
282 crit_exit();
283 KKASSERT(!IN_CRITICAL_SECT(curthread));
285 if (bootverbose)
286 kprintf("Leaving critical section, allowing interrupts\n");
288 SYSINIT(leavecrit, SI_BOOT2_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL)
291 * This is called after the threading system is up and running,
292 * including the softclock, clock interrupts, and SMP.
294 static void
295 tsleepworks(void *dummy __unused)
297 tsleep_now_works = 1;
299 SYSINIT(tsleepworks, SI_BOOT2_FINISH_SMP, SI_ORDER_SECOND, tsleepworks, NULL)
302 * This is called after devices have configured. Tell the kernel we are
303 * no longer in cold boot.
305 static void
306 endofcoldboot(void *dummy __unused)
308 cold = 0;
310 SYSINIT(endofcoldboot, SI_SUB_ISWARM, SI_ORDER_ANY, endofcoldboot, NULL)
313 ***************************************************************************
314 ****
315 **** The two following SYSINT's are proc0 specific glue code. I am not
316 **** convinced that they can not be safely combined, but their order of
317 **** operation has been maintained as the same as the original init_main.c
318 **** for right now.
319 ****
320 **** These probably belong in init_proc.c or kern_proc.c, since they
321 **** deal with proc0 (the fork template process).
322 ****
323 ***************************************************************************
325 /* ARGSUSED*/
326 static void
327 proc0_init(void *dummy __unused)
329 struct proc *p;
330 struct lwp *lp;
332 p = &proc0;
333 lp = &lwp0;
336 * Initialize process and pgrp structures.
338 procinit();
341 * additional VM structures
343 vm_init2();
346 * Create process 0 (the swapper).
348 LIST_INSERT_HEAD(&allproc, p, p_list);
349 p->p_pgrp = &pgrp0;
350 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
351 LIST_INIT(&pgrp0.pg_members);
352 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
354 pgrp0.pg_session = &session0;
355 session0.s_count = 1;
356 session0.s_leader = p;
358 p->p_sysent = &aout_sysvec;
360 p->p_flag = P_SYSTEM;
361 p->p_stat = SACTIVE;
362 lp->lwp_stat = LSRUN;
363 p->p_nice = NZERO;
364 p->p_rtprio.type = RTP_PRIO_NORMAL;
365 p->p_rtprio.prio = 0;
366 lp->lwp_rtprio = p->p_rtprio;
368 p->p_peers = 0;
369 p->p_leader = p;
371 bcopy("swapper", p->p_comm, sizeof ("swapper"));
372 bcopy("swapper", thread0.td_comm, sizeof ("swapper"));
374 /* Create credentials. */
375 p->p_ucred = crget();
376 p->p_ucred->cr_ruidinfo = uifind(0);
377 p->p_ucred->cr_ngroups = 1; /* group 0 */
378 p->p_ucred->cr_uidinfo = uifind(0);
380 /* Don't jail it */
381 p->p_ucred->cr_prison = NULL;
383 /* Create sigacts. */
384 p->p_sigacts = &sigacts0;
385 p->p_sigacts->ps_refcnt = 1;
387 /* Initialize signal state for process 0. */
388 siginit(p);
390 /* Create the file descriptor table. */
391 fdinit_bootstrap(p, &filedesc0, cmask);
393 /* Create the limits structures. */
394 plimit_init0(&limit0);
395 p->p_limit = &limit0;
397 /* Allocate a prototype map so we have something to fork. */
398 pmap_pinit0(vmspace_pmap(&vmspace0));
399 p->p_vmspace = &vmspace0;
400 lp->lwp_vmspace = p->p_vmspace;
401 sysref_init(&vmspace0.vm_sysref, &vmspace_sysref_class);
402 vm_map_init(&vmspace0.vm_map,
403 round_page(VM_MIN_USER_ADDRESS),
404 trunc_page(VM_MAX_USER_ADDRESS),
405 vmspace_pmap(&vmspace0));
406 sysref_activate(&vmspace0.vm_sysref);
409 * Charge root for one process.
411 (void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0);
412 vm_init_limits(p);
414 SYSINIT(p0init, SI_BOOT2_PROC0, SI_ORDER_FIRST, proc0_init, NULL)
416 static int proc0_post_callback(struct proc *p, void *data __unused);
418 /* ARGSUSED*/
419 static void
420 proc0_post(void *dummy __unused)
422 struct timespec ts;
425 * Now we can look at the time, having had a chance to verify the
426 * time from the file system. Pretend that proc0 started now.
428 allproc_scan(proc0_post_callback, NULL);
431 * Give the ``random'' number generator a thump.
432 * XXX: Does read_random() contain enough bits to be used here ?
434 nanotime(&ts);
435 skrandom(ts.tv_sec ^ ts.tv_nsec);
438 static int
439 proc0_post_callback(struct proc *p, void *data __unused)
441 microtime(&p->p_start);
442 return(0);
445 SYSINIT(p0post, SI_SUB_PROC0_POST, SI_ORDER_FIRST, proc0_post, NULL)
448 ***************************************************************************
449 ****
450 **** The following SYSINIT's and glue code should be moved to the
451 **** respective files on a per subsystem basis.
452 ****
453 ***************************************************************************
458 ***************************************************************************
459 ****
460 **** The following code probably belongs in another file, like
461 **** kern/init_init.c.
462 ****
463 ***************************************************************************
467 * List of paths to try when searching for "init".
469 static char init_path[MAXPATHLEN] =
470 #ifdef INIT_PATH
471 __XSTRING(INIT_PATH);
472 #else
473 "/sbin/init:/sbin/oinit:/sbin/init.bak";
474 #endif
475 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
478 * Start the initial user process; try exec'ing each pathname in init_path.
479 * The program is invoked with one argument containing the boot flags.
481 * The MP lock is held on entry.
483 static void
484 start_init(void *dummy, struct trapframe *frame)
486 vm_offset_t addr;
487 struct execve_args args;
488 int options, error;
489 char *var, *path, *next, *s;
490 char *ucp, **uap, *arg0, *arg1;
491 struct proc *p;
492 struct lwp *lp;
493 struct mount *mp;
494 struct vnode *vp;
496 p = curproc;
498 lp = ONLY_LWP_IN_PROC(p);
500 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */
501 mp = mountlist_boot_getfirst();
502 if (VFS_ROOT(mp, &vp))
503 panic("cannot find root vnode");
504 if (mp->mnt_ncmountpt.ncp == NULL) {
505 cache_allocroot(&mp->mnt_ncmountpt, mp, vp);
506 cache_unlock(&mp->mnt_ncmountpt); /* leave ref intact */
508 p->p_fd->fd_cdir = vp;
509 vref(p->p_fd->fd_cdir);
510 p->p_fd->fd_rdir = vp;
511 vref(p->p_fd->fd_rdir);
512 vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt));
513 vn_unlock(vp); /* leave ref intact */
514 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir);
515 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir);
518 * Need just enough stack to hold the faked-up "execve()" arguments.
520 addr = trunc_page(USRSTACK - PAGE_SIZE);
521 error = vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
522 FALSE,
523 VM_MAPTYPE_NORMAL,
524 VM_PROT_ALL, VM_PROT_ALL,
526 if (error)
527 panic("init: couldn't allocate argument space");
528 p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
529 p->p_vmspace->vm_ssize = 1;
531 if ((var = kgetenv("init_path")) != NULL) {
532 strncpy(init_path, var, sizeof init_path);
533 init_path[sizeof init_path - 1] = 0;
535 if ((var = kgetenv("kern.fallback_elf_brand")) != NULL)
536 fallback_elf_brand = strtol(var, NULL, 0);
538 for (path = init_path; *path != '\0'; path = next) {
539 while (*path == ':')
540 path++;
541 if (*path == '\0')
542 break;
543 for (next = path; *next != '\0' && *next != ':'; next++)
544 /* nothing */ ;
545 if (bootverbose)
546 kprintf("start_init: trying %.*s\n", (int)(next - path),
547 path);
550 * Move out the boot flag argument.
552 options = 0;
553 ucp = (char *)USRSTACK;
554 (void)subyte(--ucp, 0); /* trailing zero */
555 if (boothowto & RB_SINGLE) {
556 (void)subyte(--ucp, 's');
557 options = 1;
559 #ifdef notyet
560 if (boothowto & RB_FASTBOOT) {
561 (void)subyte(--ucp, 'f');
562 options = 1;
564 #endif
566 #ifdef BOOTCDROM
567 (void)subyte(--ucp, 'C');
568 options = 1;
569 #endif
570 if (options == 0)
571 (void)subyte(--ucp, '-');
572 (void)subyte(--ucp, '-'); /* leading hyphen */
573 arg1 = ucp;
576 * Move out the file name (also arg 0).
578 (void)subyte(--ucp, 0);
579 for (s = next - 1; s >= path; s--)
580 (void)subyte(--ucp, *s);
581 arg0 = ucp;
584 * Move out the arg pointers.
586 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
587 (void)suword((caddr_t)--uap, (long)0); /* terminator */
588 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
589 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
592 * Point at the arguments.
594 args.fname = arg0;
595 args.argv = uap;
596 args.envv = NULL;
599 * Now try to exec the program. If can't for any reason
600 * other than it doesn't exist, complain.
602 * Otherwise, return via fork_trampoline() all the way
603 * to user mode as init!
605 * WARNING! We may have been moved to another cpu after
606 * acquiring the current user process designation. The
607 * MP lock will migrate with us though so we still have to
608 * release it.
610 if ((error = sys_execve(&args)) == 0) {
611 rel_mplock();
612 lp->lwp_proc->p_usched->acquire_curproc(lp);
613 return;
615 if (error != ENOENT)
616 kprintf("exec %.*s: error %d\n", (int)(next - path),
617 path, error);
619 kprintf("init: not found in path %s\n", init_path);
620 panic("no init");
624 * Like kthread_create(), but runs in it's own address space.
625 * We do this early to reserve pid 1.
627 * Note special case - do not make it runnable yet. Other work
628 * in progress will change this more.
630 static void
631 create_init(const void *udata __unused)
633 int error;
634 struct lwp *lp;
636 crit_enter();
637 error = fork1(&lwp0, RFFDG | RFPROC, &initproc);
638 if (error)
639 panic("cannot fork init: %d", error);
640 initproc->p_flag |= P_SYSTEM;
641 lp = ONLY_LWP_IN_PROC(initproc);
642 cpu_set_fork_handler(lp, start_init, NULL);
643 crit_exit();
645 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
648 * Make it runnable now.
650 static void
651 kick_init(const void *udata __unused)
653 start_forked_proc(&lwp0, initproc);
655 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
658 * Machine independant globaldata initialization
660 * WARNING! Called from early boot, 'mycpu' may not work yet.
662 void
663 mi_gdinit(struct globaldata *gd, int cpuid)
665 TAILQ_INIT(&gd->gd_systimerq);
666 gd->gd_sysid_alloc = cpuid; /* prime low bits for cpu lookup */
667 gd->gd_cpuid = cpuid;
668 gd->gd_cpumask = (cpumask_t)1 << cpuid;
669 lwkt_gdinit(gd);
670 vm_map_entry_reserve_cpu_init(gd);
671 sleep_gdinit(gd);