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 1.71 2007/01/01 22:51:17 corecode 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/file2.h> 66 #include <sys/thread2.h> 67 68 #include <machine/cpu.h> 69 70 #include <vm/vm.h> 71 #include <vm/vm_param.h> 72 #include <sys/lock.h> 73 #include <vm/pmap.h> 74 #include <vm/vm_map.h> 75 #include <sys/user.h> 76 #include <sys/copyright.h> 77 78 /* Components of the first process -- never freed. */ 79 static struct session session0; 80 static struct pgrp pgrp0; 81 static struct procsig procsig0; 82 static struct filedesc filedesc0; 83 static struct plimit limit0; 84 static struct vmspace vmspace0; 85 struct proc *initproc; 86 struct proc proc0; 87 struct thread thread0; 88 89 int cmask = CMASK; 90 extern struct user *proc0paddr; 91 extern int fallback_elf_brand; 92 93 int boothowto = 0; /* initialized so that it can be patched */ 94 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, ""); 95 96 /* 97 * This ensures that there is at least one entry so that the sysinit_set 98 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never 99 * executed. 100 */ 101 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL) 102 103 /* 104 * The sysinit table itself. Items are checked off as the are run. 105 * If we want to register new sysinit types, add them to newsysinit. 106 */ 107 SET_DECLARE(sysinit_set, struct sysinit); 108 struct sysinit **sysinit, **sysinit_end; 109 struct sysinit **newsysinit, **newsysinit_end; 110 111 112 /* 113 * Merge a new sysinit set into the current set, reallocating it if 114 * necessary. This can only be called after malloc is running. 115 */ 116 void 117 sysinit_add(struct sysinit **set, struct sysinit **set_end) 118 { 119 struct sysinit **newset; 120 struct sysinit **sipp; 121 struct sysinit **xipp; 122 int count; 123 124 count = set_end - set; 125 if (newsysinit) 126 count += newsysinit_end - newsysinit; 127 else 128 count += sysinit_end - sysinit; 129 newset = kmalloc(count * sizeof(*sipp), M_TEMP, M_WAITOK); 130 if (newset == NULL) 131 panic("cannot malloc for sysinit"); 132 xipp = newset; 133 if (newsysinit) { 134 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 135 *xipp++ = *sipp; 136 } else { 137 for (sipp = sysinit; sipp < sysinit_end; sipp++) 138 *xipp++ = *sipp; 139 } 140 for (sipp = set; sipp < set_end; sipp++) 141 *xipp++ = *sipp; 142 if (newsysinit) 143 kfree(newsysinit, M_TEMP); 144 newsysinit = newset; 145 newsysinit_end = newset + count; 146 } 147 148 /* 149 * Callbacks from machine-dependant startup code (e.g. init386) to set 150 * up low level entities related to cpu #0's globaldata. 151 * 152 * Called from very low level boot code. 153 */ 154 void 155 mi_proc0init(struct globaldata *gd, struct user *proc0paddr) 156 { 157 lwkt_init_thread(&thread0, proc0paddr, LWKT_THREAD_STACK, 0, gd); 158 lwkt_set_comm(&thread0, "thread0"); 159 proc0.p_addr = (void *)thread0.td_kstack; 160 LIST_INIT(&proc0.p_lwps); 161 LIST_INSERT_HEAD(&proc0.p_lwps, &proc0.p_lwp, lwp_list); 162 proc0.p_lwp.lwp_thread = &thread0; 163 proc0.p_lwp.lwp_proc = &proc0; 164 proc0.p_usched = usched_init(); 165 proc0.p_lwp.lwp_cpumask = 0xFFFFFFFF; 166 varsymset_init(&proc0.p_varsymset, NULL); 167 thread0.td_flags |= TDF_RUNNING; 168 thread0.td_proc = &proc0; 169 thread0.td_lwp = &proc0.p_lwp; 170 thread0.td_switch = cpu_heavy_switch; /* YYY eventually LWKT */ 171 } 172 173 /* 174 * System startup; initialize the world, create process 0, mount root 175 * filesystem, and fork to create init and pagedaemon. Most of the 176 * hard work is done in the lower-level initialization routines including 177 * startup(), which does memory initialization and autoconfiguration. 178 * 179 * This allows simple addition of new kernel subsystems that require 180 * boot time initialization. It also allows substitution of subsystem 181 * (for instance, a scheduler, kernel profiler, or VM system) by object 182 * module. Finally, it allows for optional "kernel threads". 183 */ 184 void 185 mi_startup(void) 186 { 187 struct sysinit *sip; /* system initialization*/ 188 struct sysinit **sipp; /* system initialization*/ 189 struct sysinit **xipp; /* interior loop of sort*/ 190 struct sysinit *save; /* bubble*/ 191 192 if (sysinit == NULL) { 193 sysinit = SET_BEGIN(sysinit_set); 194 sysinit_end = SET_LIMIT(sysinit_set); 195 } 196 197 restart: 198 /* 199 * Perform a bubble sort of the system initialization objects by 200 * their subsystem (primary key) and order (secondary key). 201 */ 202 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 203 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 204 if ((*sipp)->subsystem < (*xipp)->subsystem || 205 ((*sipp)->subsystem == (*xipp)->subsystem && 206 (*sipp)->order <= (*xipp)->order)) 207 continue; /* skip*/ 208 save = *sipp; 209 *sipp = *xipp; 210 *xipp = save; 211 } 212 } 213 214 /* 215 * Traverse the (now) ordered list of system initialization tasks. 216 * Perform each task, and continue on to the next task. 217 * 218 * The last item on the list is expected to be the scheduler, 219 * which will not return. 220 */ 221 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 222 sip = *sipp; 223 if (sip->subsystem == SI_SUB_DUMMY) 224 continue; /* skip dummy task(s)*/ 225 226 if (sip->subsystem == SI_SUB_DONE) 227 continue; 228 229 /* Call function */ 230 (*(sip->func))(sip->udata); 231 232 /* Check off the one we're just done */ 233 sip->subsystem = SI_SUB_DONE; 234 235 /* Check if we've installed more sysinit items via KLD */ 236 if (newsysinit != NULL) { 237 if (sysinit != SET_BEGIN(sysinit_set)) 238 kfree(sysinit, M_TEMP); 239 sysinit = newsysinit; 240 sysinit_end = newsysinit_end; 241 newsysinit = NULL; 242 newsysinit_end = NULL; 243 goto restart; 244 } 245 } 246 247 panic("Shouldn't get here!"); 248 /* NOTREACHED*/ 249 } 250 251 252 /* 253 *************************************************************************** 254 **** 255 **** The following SYSINIT's belong elsewhere, but have not yet 256 **** been moved. 257 **** 258 *************************************************************************** 259 */ 260 static void 261 print_caddr_t(void *data __unused) 262 { 263 kprintf("%s", (char *)data); 264 } 265 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright) 266 267 /* 268 * Leave the critical section that protected us from spurious interrupts 269 * so device probes work. 270 */ 271 static void 272 leavecrit(void *dummy __unused) 273 { 274 crit_exit(); 275 KKASSERT(!IN_CRITICAL_SECT(curthread)); 276 if (bootverbose) 277 kprintf("Leaving critical section, allowing interrupts\n"); 278 } 279 SYSINIT(leavecrit, SI_SUB_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL) 280 281 /* 282 *************************************************************************** 283 **** 284 **** The two following SYSINT's are proc0 specific glue code. I am not 285 **** convinced that they can not be safely combined, but their order of 286 **** operation has been maintained as the same as the original init_main.c 287 **** for right now. 288 **** 289 **** These probably belong in init_proc.c or kern_proc.c, since they 290 **** deal with proc0 (the fork template process). 291 **** 292 *************************************************************************** 293 */ 294 /* ARGSUSED*/ 295 static void 296 proc0_init(void *dummy __unused) 297 { 298 struct proc *p; 299 struct lwp *lp; 300 301 p = &proc0; 302 lp = &proc0.p_lwp; /* XXX lwp to be: lwp0 */ 303 304 /* 305 * Initialize process and pgrp structures. 306 */ 307 procinit(); 308 309 /* 310 * additional VM structures 311 */ 312 vm_init2(); 313 314 /* 315 * Create process 0 (the swapper). 316 */ 317 LIST_INSERT_HEAD(&allproc, p, p_list); 318 p->p_pgrp = &pgrp0; 319 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 320 LIST_INIT(&pgrp0.pg_members); 321 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 322 323 pgrp0.pg_session = &session0; 324 session0.s_count = 1; 325 session0.s_leader = p; 326 327 p->p_sysent = &aout_sysvec; 328 329 p->p_flag = P_SYSTEM; 330 p->p_stat = SRUN; 331 p->p_nice = NZERO; 332 p->p_rtprio.type = RTP_PRIO_NORMAL; 333 p->p_rtprio.prio = 0; 334 p->p_lwp.lwp_rtprio = p->p_rtprio; 335 336 p->p_peers = 0; 337 p->p_leader = p; 338 339 bcopy("swapper", p->p_comm, sizeof ("swapper")); 340 bcopy("swapper", thread0.td_comm, sizeof ("swapper")); 341 342 /* Create credentials. */ 343 p->p_ucred = crget(); 344 p->p_ucred->cr_ruidinfo = uifind(0); 345 p->p_ucred->cr_ngroups = 1; /* group 0 */ 346 p->p_ucred->cr_uidinfo = uifind(0); 347 348 /* Don't jail it */ 349 p->p_ucred->cr_prison = NULL; 350 351 /* Create procsig. */ 352 p->p_procsig = &procsig0; 353 p->p_procsig->ps_refcnt = 1; 354 355 /* Initialize signal state for process 0. */ 356 siginit(p); 357 358 /* Create the file descriptor table. */ 359 fdinit_bootstrap(p, &filedesc0, cmask); 360 361 /* Create the limits structures. */ 362 plimit_init0(&limit0); 363 p->p_limit = &limit0; 364 365 /* Allocate a prototype map so we have something to fork. */ 366 pmap_pinit0(vmspace_pmap(&vmspace0)); 367 p->p_vmspace = &vmspace0; 368 vmspace0.vm_refcnt = 1; 369 vm_map_init(&vmspace0.vm_map, 370 round_page(VM_MIN_USER_ADDRESS), 371 trunc_page(VM_MAX_USER_ADDRESS), 372 vmspace_pmap(&vmspace0)); 373 374 /* 375 * We continue to place signal 376 * actions in the user struct so they're pageable. 377 * 378 * XXX old + cufty. will be removed 379 */ 380 p->p_sigacts = &p->p_addr->u_sigacts; 381 382 /* 383 * Charge root for one process. 384 */ 385 (void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0); 386 387 } 388 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL) 389 390 static int proc0_post_callback(struct proc *p, void *data __unused); 391 392 /* ARGSUSED*/ 393 static void 394 proc0_post(void *dummy __unused) 395 { 396 struct timespec ts; 397 398 /* 399 * Now we can look at the time, having had a chance to verify the 400 * time from the file system. Pretend that proc0 started now. 401 */ 402 allproc_scan(proc0_post_callback, NULL); 403 404 /* 405 * Give the ``random'' number generator a thump. 406 * XXX: Does read_random() contain enough bits to be used here ? 407 */ 408 nanotime(&ts); 409 skrandom(ts.tv_sec ^ ts.tv_nsec); 410 } 411 412 static int 413 proc0_post_callback(struct proc *p, void *data __unused) 414 { 415 microtime(&p->p_start); 416 return(0); 417 } 418 419 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL) 420 421 /* 422 *************************************************************************** 423 **** 424 **** The following SYSINIT's and glue code should be moved to the 425 **** respective files on a per subsystem basis. 426 **** 427 *************************************************************************** 428 */ 429 430 431 /* 432 *************************************************************************** 433 **** 434 **** The following code probably belongs in another file, like 435 **** kern/init_init.c. 436 **** 437 *************************************************************************** 438 */ 439 440 /* 441 * List of paths to try when searching for "init". 442 */ 443 static char init_path[MAXPATHLEN] = 444 #ifdef INIT_PATH 445 __XSTRING(INIT_PATH); 446 #else 447 "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall"; 448 #endif 449 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, ""); 450 451 /* 452 * Start the initial user process; try exec'ing each pathname in init_path. 453 * The program is invoked with one argument containing the boot flags. 454 * 455 * The MP lock is held on entry. 456 */ 457 static void 458 start_init(void *dummy) 459 { 460 vm_offset_t addr; 461 struct execve_args args; 462 int options, error; 463 char *var, *path, *next, *s; 464 char *ucp, **uap, *arg0, *arg1; 465 struct proc *p; 466 struct lwp *lp; 467 struct mount *mp; 468 struct vnode *vp; 469 470 p = curproc; 471 472 KKASSERT(p->p_nthreads == 1); 473 474 lp = LIST_FIRST(&p->p_lwps); 475 476 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */ 477 mp = mountlist_boot_getfirst(); 478 if (VFS_ROOT(mp, &vp)) 479 panic("cannot find root vnode"); 480 if (mp->mnt_ncmountpt.ncp == NULL) { 481 cache_allocroot(&mp->mnt_ncmountpt, mp, vp); 482 cache_unlock(&mp->mnt_ncmountpt); /* leave ref intact */ 483 } 484 p->p_fd->fd_cdir = vp; 485 vref(p->p_fd->fd_cdir); 486 p->p_fd->fd_rdir = vp; 487 vref(p->p_fd->fd_rdir); 488 vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt)); 489 vn_unlock(vp); /* leave ref intact */ 490 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir); 491 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir); 492 493 /* 494 * Need just enough stack to hold the faked-up "execve()" arguments. 495 */ 496 addr = trunc_page(USRSTACK - PAGE_SIZE); 497 error = vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 498 FALSE, 499 VM_MAPTYPE_NORMAL, 500 VM_PROT_ALL, VM_PROT_ALL, 501 0); 502 if (error) 503 panic("init: couldn't allocate argument space"); 504 p->p_vmspace->vm_maxsaddr = (caddr_t)addr; 505 p->p_vmspace->vm_ssize = 1; 506 507 if ((var = kgetenv("init_path")) != NULL) { 508 strncpy(init_path, var, sizeof init_path); 509 init_path[sizeof init_path - 1] = 0; 510 } 511 if ((var = kgetenv("kern.fallback_elf_brand")) != NULL) 512 fallback_elf_brand = strtol(var, NULL, 0); 513 514 for (path = init_path; *path != '\0'; path = next) { 515 while (*path == ':') 516 path++; 517 if (*path == '\0') 518 break; 519 for (next = path; *next != '\0' && *next != ':'; next++) 520 /* nothing */ ; 521 if (bootverbose) 522 kprintf("start_init: trying %.*s\n", (int)(next - path), 523 path); 524 525 /* 526 * Move out the boot flag argument. 527 */ 528 options = 0; 529 ucp = (char *)USRSTACK; 530 (void)subyte(--ucp, 0); /* trailing zero */ 531 if (boothowto & RB_SINGLE) { 532 (void)subyte(--ucp, 's'); 533 options = 1; 534 } 535 #ifdef notyet 536 if (boothowto & RB_FASTBOOT) { 537 (void)subyte(--ucp, 'f'); 538 options = 1; 539 } 540 #endif 541 542 #ifdef BOOTCDROM 543 (void)subyte(--ucp, 'C'); 544 options = 1; 545 #endif 546 if (options == 0) 547 (void)subyte(--ucp, '-'); 548 (void)subyte(--ucp, '-'); /* leading hyphen */ 549 arg1 = ucp; 550 551 /* 552 * Move out the file name (also arg 0). 553 */ 554 (void)subyte(--ucp, 0); 555 for (s = next - 1; s >= path; s--) 556 (void)subyte(--ucp, *s); 557 arg0 = ucp; 558 559 /* 560 * Move out the arg pointers. 561 */ 562 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1)); 563 (void)suword((caddr_t)--uap, (long)0); /* terminator */ 564 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1); 565 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0); 566 567 /* 568 * Point at the arguments. 569 */ 570 args.fname = arg0; 571 args.argv = uap; 572 args.envv = NULL; 573 574 /* 575 * Now try to exec the program. If can't for any reason 576 * other than it doesn't exist, complain. 577 * 578 * Otherwise, return via fork_trampoline() all the way 579 * to user mode as init! 580 * 581 * WARNING! We may have been moved to another cpu after 582 * acquiring the current user process designation. The 583 * MP lock will migrate with us though so we still have to 584 * release it. 585 */ 586 if ((error = sys_execve(&args)) == 0) { 587 rel_mplock(); 588 lp->lwp_proc->p_usched->acquire_curproc(lp); 589 return; 590 } 591 if (error != ENOENT) 592 kprintf("exec %.*s: error %d\n", (int)(next - path), 593 path, error); 594 } 595 kprintf("init: not found in path %s\n", init_path); 596 panic("no init"); 597 } 598 599 /* 600 * Like kthread_create(), but runs in it's own address space. 601 * We do this early to reserve pid 1. 602 * 603 * Note special case - do not make it runnable yet. Other work 604 * in progress will change this more. 605 */ 606 static void 607 create_init(const void *udata __unused) 608 { 609 int error; 610 struct lwp *lp; 611 612 crit_enter(); 613 error = fork1(&proc0.p_lwp, RFFDG | RFPROC, &initproc); 614 if (error) 615 panic("cannot fork init: %d", error); 616 initproc->p_flag |= P_SYSTEM; 617 lp = LIST_FIRST(&initproc->p_lwps); 618 cpu_set_fork_handler(lp, start_init, NULL); 619 crit_exit(); 620 } 621 SYSINIT(init,SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL) 622 623 /* 624 * Make it runnable now. 625 */ 626 static void 627 kick_init(const void *udata __unused) 628 { 629 start_forked_proc(&proc0.p_lwp, initproc); 630 } 631 SYSINIT(kickinit,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL) 632 633 /* 634 * Machine independant globaldata initialization 635 * 636 * WARNING! Called from early boot, 'mycpu' may not work yet. 637 */ 638 void 639 mi_gdinit(struct globaldata *gd, int cpuid) 640 { 641 TAILQ_INIT(&gd->gd_tdfreeq); /* for pmap_{new,dispose}_thread() */ 642 TAILQ_INIT(&gd->gd_systimerq); 643 gd->gd_cpuid = cpuid; 644 gd->gd_cpumask = (cpumask_t)1 << cpuid; 645 lwkt_gdinit(gd); 646 vm_map_entry_reserve_cpu_init(gd); 647 sleep_gdinit(gd); 648 } 649 650 651