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 1.73 2007/02/16 23:11:39 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 lwp lwp0; 88 struct thread thread0; 89 90 int cmask = CMASK; 91 extern struct user *proc0paddr; 92 extern int fallback_elf_brand; 93 94 int boothowto = 0; /* initialized so that it can be patched */ 95 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, ""); 96 97 /* 98 * This ensures that there is at least one entry so that the sysinit_set 99 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never 100 * executed. 101 */ 102 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL) 103 104 /* 105 * The sysinit table itself. Items are checked off as the are run. 106 * If we want to register new sysinit types, add them to newsysinit. 107 */ 108 SET_DECLARE(sysinit_set, struct sysinit); 109 struct sysinit **sysinit, **sysinit_end; 110 struct sysinit **newsysinit, **newsysinit_end; 111 112 113 /* 114 * Merge a new sysinit set into the current set, reallocating it if 115 * necessary. This can only be called after malloc is running. 116 */ 117 void 118 sysinit_add(struct sysinit **set, struct sysinit **set_end) 119 { 120 struct sysinit **newset; 121 struct sysinit **sipp; 122 struct sysinit **xipp; 123 int count; 124 125 count = set_end - set; 126 if (newsysinit) 127 count += newsysinit_end - newsysinit; 128 else 129 count += sysinit_end - sysinit; 130 newset = kmalloc(count * sizeof(*sipp), M_TEMP, M_WAITOK); 131 if (newset == NULL) 132 panic("cannot malloc for sysinit"); 133 xipp = newset; 134 if (newsysinit) { 135 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 136 *xipp++ = *sipp; 137 } else { 138 for (sipp = sysinit; sipp < sysinit_end; sipp++) 139 *xipp++ = *sipp; 140 } 141 for (sipp = set; sipp < set_end; sipp++) 142 *xipp++ = *sipp; 143 if (newsysinit) 144 kfree(newsysinit, M_TEMP); 145 newsysinit = newset; 146 newsysinit_end = newset + count; 147 } 148 149 /* 150 * Callbacks from machine-dependant startup code (e.g. init386) to set 151 * up low level entities related to cpu #0's globaldata. 152 * 153 * Called from very low level boot code. 154 */ 155 void 156 mi_proc0init(struct globaldata *gd, struct user *proc0paddr) 157 { 158 lwkt_init_thread(&thread0, proc0paddr, LWKT_THREAD_STACK, 0, gd); 159 lwkt_set_comm(&thread0, "thread0"); 160 proc0.p_addr = (void *)thread0.td_kstack; 161 LIST_INIT(&proc0.p_lwps); 162 LIST_INSERT_HEAD(&proc0.p_lwps, &lwp0, lwp_list); 163 lwp0.lwp_thread = &thread0; 164 lwp0.lwp_proc = &proc0; 165 proc0.p_usched = usched_init(); 166 lwp0.lwp_cpumask = 0xFFFFFFFF; 167 varsymset_init(&proc0.p_varsymset, NULL); 168 thread0.td_flags |= TDF_RUNNING; 169 thread0.td_proc = &proc0; 170 thread0.td_lwp = &lwp0; 171 thread0.td_switch = cpu_heavy_switch; /* YYY eventually LWKT */ 172 } 173 174 /* 175 * System startup; initialize the world, create process 0, mount root 176 * filesystem, and fork to create init and pagedaemon. Most of the 177 * hard work is done in the lower-level initialization routines including 178 * startup(), which does memory initialization and autoconfiguration. 179 * 180 * This allows simple addition of new kernel subsystems that require 181 * boot time initialization. It also allows substitution of subsystem 182 * (for instance, a scheduler, kernel profiler, or VM system) by object 183 * module. Finally, it allows for optional "kernel threads". 184 */ 185 void 186 mi_startup(void) 187 { 188 struct sysinit *sip; /* system initialization*/ 189 struct sysinit **sipp; /* system initialization*/ 190 struct sysinit **xipp; /* interior loop of sort*/ 191 struct sysinit *save; /* bubble*/ 192 193 if (sysinit == NULL) { 194 sysinit = SET_BEGIN(sysinit_set); 195 sysinit_end = SET_LIMIT(sysinit_set); 196 } 197 198 restart: 199 /* 200 * Perform a bubble sort of the system initialization objects by 201 * their subsystem (primary key) and order (secondary key). 202 */ 203 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 204 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 205 if ((*sipp)->subsystem < (*xipp)->subsystem || 206 ((*sipp)->subsystem == (*xipp)->subsystem && 207 (*sipp)->order <= (*xipp)->order)) 208 continue; /* skip*/ 209 save = *sipp; 210 *sipp = *xipp; 211 *xipp = save; 212 } 213 } 214 215 /* 216 * Traverse the (now) ordered list of system initialization tasks. 217 * Perform each task, and continue on to the next task. 218 * 219 * The last item on the list is expected to be the scheduler, 220 * which will not return. 221 */ 222 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 223 sip = *sipp; 224 if (sip->subsystem == SI_SUB_DUMMY) 225 continue; /* skip dummy task(s)*/ 226 227 if (sip->subsystem == SI_SUB_DONE) 228 continue; 229 230 /* Call function */ 231 (*(sip->func))(sip->udata); 232 233 /* Check off the one we're just done */ 234 sip->subsystem = SI_SUB_DONE; 235 236 /* Check if we've installed more sysinit items via KLD */ 237 if (newsysinit != NULL) { 238 if (sysinit != SET_BEGIN(sysinit_set)) 239 kfree(sysinit, M_TEMP); 240 sysinit = newsysinit; 241 sysinit_end = newsysinit_end; 242 newsysinit = NULL; 243 newsysinit_end = NULL; 244 goto restart; 245 } 246 } 247 248 panic("Shouldn't get here!"); 249 /* NOTREACHED*/ 250 } 251 252 253 /* 254 *************************************************************************** 255 **** 256 **** The following SYSINIT's belong elsewhere, but have not yet 257 **** been moved. 258 **** 259 *************************************************************************** 260 */ 261 static void 262 print_caddr_t(void *data __unused) 263 { 264 kprintf("%s", (char *)data); 265 } 266 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright) 267 268 /* 269 * Leave the critical section that protected us from spurious interrupts 270 * so device probes work. 271 */ 272 static void 273 leavecrit(void *dummy __unused) 274 { 275 crit_exit(); 276 KKASSERT(!IN_CRITICAL_SECT(curthread)); 277 if (bootverbose) 278 kprintf("Leaving critical section, allowing interrupts\n"); 279 } 280 SYSINIT(leavecrit, SI_SUB_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL) 281 282 /* 283 *************************************************************************** 284 **** 285 **** The two following SYSINT's are proc0 specific glue code. I am not 286 **** convinced that they can not be safely combined, but their order of 287 **** operation has been maintained as the same as the original init_main.c 288 **** for right now. 289 **** 290 **** These probably belong in init_proc.c or kern_proc.c, since they 291 **** deal with proc0 (the fork template process). 292 **** 293 *************************************************************************** 294 */ 295 /* ARGSUSED*/ 296 static void 297 proc0_init(void *dummy __unused) 298 { 299 struct proc *p; 300 struct lwp *lp; 301 302 p = &proc0; 303 lp = &lwp0; 304 305 /* 306 * Initialize process and pgrp structures. 307 */ 308 procinit(); 309 310 /* 311 * additional VM structures 312 */ 313 vm_init2(); 314 315 /* 316 * Create process 0 (the swapper). 317 */ 318 LIST_INSERT_HEAD(&allproc, p, p_list); 319 p->p_pgrp = &pgrp0; 320 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 321 LIST_INIT(&pgrp0.pg_members); 322 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 323 324 pgrp0.pg_session = &session0; 325 session0.s_count = 1; 326 session0.s_leader = p; 327 328 p->p_sysent = &aout_sysvec; 329 330 p->p_flag = P_SYSTEM; 331 p->p_stat = SACTIVE; 332 lp->lwp_stat = LSRUN; 333 p->p_nice = NZERO; 334 p->p_rtprio.type = RTP_PRIO_NORMAL; 335 p->p_rtprio.prio = 0; 336 lp->lwp_rtprio = p->p_rtprio; 337 338 p->p_peers = 0; 339 p->p_leader = p; 340 341 bcopy("swapper", p->p_comm, sizeof ("swapper")); 342 bcopy("swapper", thread0.td_comm, sizeof ("swapper")); 343 344 /* Create credentials. */ 345 p->p_ucred = crget(); 346 p->p_ucred->cr_ruidinfo = uifind(0); 347 p->p_ucred->cr_ngroups = 1; /* group 0 */ 348 p->p_ucred->cr_uidinfo = uifind(0); 349 350 /* Don't jail it */ 351 p->p_ucred->cr_prison = NULL; 352 353 /* Create procsig. */ 354 p->p_procsig = &procsig0; 355 p->p_procsig->ps_refcnt = 1; 356 357 /* Initialize signal state for process 0. */ 358 siginit(p); 359 360 /* Create the file descriptor table. */ 361 fdinit_bootstrap(p, &filedesc0, cmask); 362 363 /* Create the limits structures. */ 364 plimit_init0(&limit0); 365 p->p_limit = &limit0; 366 367 /* Allocate a prototype map so we have something to fork. */ 368 pmap_pinit0(vmspace_pmap(&vmspace0)); 369 p->p_vmspace = &vmspace0; 370 vmspace0.vm_refcnt = 1; 371 vm_map_init(&vmspace0.vm_map, 372 round_page(VM_MIN_USER_ADDRESS), 373 trunc_page(VM_MAX_USER_ADDRESS), 374 vmspace_pmap(&vmspace0)); 375 376 /* 377 * We continue to place signal 378 * actions in the user struct so they're pageable. 379 * 380 * XXX old + cufty. will be removed 381 */ 382 p->p_sigacts = &p->p_addr->u_sigacts; 383 384 /* 385 * Charge root for one process. 386 */ 387 (void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0); 388 389 } 390 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL) 391 392 static int proc0_post_callback(struct proc *p, void *data __unused); 393 394 /* ARGSUSED*/ 395 static void 396 proc0_post(void *dummy __unused) 397 { 398 struct timespec ts; 399 400 /* 401 * Now we can look at the time, having had a chance to verify the 402 * time from the file system. Pretend that proc0 started now. 403 */ 404 allproc_scan(proc0_post_callback, NULL); 405 406 /* 407 * Give the ``random'' number generator a thump. 408 * XXX: Does read_random() contain enough bits to be used here ? 409 */ 410 nanotime(&ts); 411 skrandom(ts.tv_sec ^ ts.tv_nsec); 412 } 413 414 static int 415 proc0_post_callback(struct proc *p, void *data __unused) 416 { 417 microtime(&p->p_start); 418 return(0); 419 } 420 421 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL) 422 423 /* 424 *************************************************************************** 425 **** 426 **** The following SYSINIT's and glue code should be moved to the 427 **** respective files on a per subsystem basis. 428 **** 429 *************************************************************************** 430 */ 431 432 433 /* 434 *************************************************************************** 435 **** 436 **** The following code probably belongs in another file, like 437 **** kern/init_init.c. 438 **** 439 *************************************************************************** 440 */ 441 442 /* 443 * List of paths to try when searching for "init". 444 */ 445 static char init_path[MAXPATHLEN] = 446 #ifdef INIT_PATH 447 __XSTRING(INIT_PATH); 448 #else 449 "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall"; 450 #endif 451 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, ""); 452 453 /* 454 * Start the initial user process; try exec'ing each pathname in init_path. 455 * The program is invoked with one argument containing the boot flags. 456 * 457 * The MP lock is held on entry. 458 */ 459 static void 460 start_init(void *dummy) 461 { 462 vm_offset_t addr; 463 struct execve_args args; 464 int options, error; 465 char *var, *path, *next, *s; 466 char *ucp, **uap, *arg0, *arg1; 467 struct proc *p; 468 struct lwp *lp; 469 struct mount *mp; 470 struct vnode *vp; 471 472 p = curproc; 473 474 lp = ONLY_LWP_IN_PROC(p); 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(&lwp0, RFFDG | RFPROC, &initproc); 614 if (error) 615 panic("cannot fork init: %d", error); 616 initproc->p_flag |= P_SYSTEM; 617 lp = ONLY_LWP_IN_PROC(initproc); 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(&lwp0, 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