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 1.51 2005/11/14 18:50:05 dillon 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 void mi_startup(void); /* Should be elsewhere */ 79 80 /* Components of the first process -- never freed. */ 81 static struct session session0; 82 static struct pgrp pgrp0; 83 static struct procsig procsig0; 84 static struct filedesc filedesc0; 85 static struct plimit limit0; 86 static struct vmspace vmspace0; 87 struct proc *initproc; 88 struct proc proc0; 89 struct thread thread0; 90 91 int cmask = CMASK; 92 extern struct user *proc0paddr; 93 extern int fallback_elf_brand; 94 95 int boothowto = 0; /* initialized so that it can be patched */ 96 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, ""); 97 98 /* 99 * This ensures that there is at least one entry so that the sysinit_set 100 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never 101 * executed. 102 */ 103 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL) 104 105 /* 106 * The sysinit table itself. Items are checked off as the are run. 107 * If we want to register new sysinit types, add them to newsysinit. 108 */ 109 SET_DECLARE(sysinit_set, struct sysinit); 110 struct sysinit **sysinit, **sysinit_end; 111 struct sysinit **newsysinit, **newsysinit_end; 112 113 114 /* 115 * Merge a new sysinit set into the current set, reallocating it if 116 * necessary. This can only be called after malloc is running. 117 */ 118 void 119 sysinit_add(struct sysinit **set, struct sysinit **set_end) 120 { 121 struct sysinit **newset; 122 struct sysinit **sipp; 123 struct sysinit **xipp; 124 int count; 125 126 count = set_end - set; 127 if (newsysinit) 128 count += newsysinit_end - newsysinit; 129 else 130 count += sysinit_end - sysinit; 131 newset = malloc(count * sizeof(*sipp), M_TEMP, M_WAITOK); 132 if (newset == NULL) 133 panic("cannot malloc for sysinit"); 134 xipp = newset; 135 if (newsysinit) { 136 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 137 *xipp++ = *sipp; 138 } else { 139 for (sipp = sysinit; sipp < sysinit_end; sipp++) 140 *xipp++ = *sipp; 141 } 142 for (sipp = set; sipp < set_end; sipp++) 143 *xipp++ = *sipp; 144 if (newsysinit) 145 free(newsysinit, M_TEMP); 146 newsysinit = newset; 147 newsysinit_end = newset + count; 148 } 149 150 /* 151 * System startup; initialize the world, create process 0, mount root 152 * filesystem, and fork to create init and pagedaemon. Most of the 153 * hard work is done in the lower-level initialization routines including 154 * startup(), which does memory initialization and autoconfiguration. 155 * 156 * This allows simple addition of new kernel subsystems that require 157 * boot time initialization. It also allows substitution of subsystem 158 * (for instance, a scheduler, kernel profiler, or VM system) by object 159 * module. Finally, it allows for optional "kernel threads". 160 */ 161 void 162 mi_startup(void) 163 { 164 struct sysinit *sip; /* system initialization*/ 165 struct sysinit **sipp; /* system initialization*/ 166 struct sysinit **xipp; /* interior loop of sort*/ 167 struct sysinit *save; /* bubble*/ 168 169 if (sysinit == NULL) { 170 sysinit = SET_BEGIN(sysinit_set); 171 sysinit_end = SET_LIMIT(sysinit_set); 172 } 173 174 restart: 175 /* 176 * Perform a bubble sort of the system initialization objects by 177 * their subsystem (primary key) and order (secondary key). 178 */ 179 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 180 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 181 if ((*sipp)->subsystem < (*xipp)->subsystem || 182 ((*sipp)->subsystem == (*xipp)->subsystem && 183 (*sipp)->order <= (*xipp)->order)) 184 continue; /* skip*/ 185 save = *sipp; 186 *sipp = *xipp; 187 *xipp = save; 188 } 189 } 190 191 /* 192 * Traverse the (now) ordered list of system initialization tasks. 193 * Perform each task, and continue on to the next task. 194 * 195 * The last item on the list is expected to be the scheduler, 196 * which will not return. 197 */ 198 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 199 sip = *sipp; 200 if (sip->subsystem == SI_SUB_DUMMY) 201 continue; /* skip dummy task(s)*/ 202 203 if (sip->subsystem == SI_SUB_DONE) 204 continue; 205 206 /* Call function */ 207 (*(sip->func))(sip->udata); 208 209 /* Check off the one we're just done */ 210 sip->subsystem = SI_SUB_DONE; 211 212 /* Check if we've installed more sysinit items via KLD */ 213 if (newsysinit != NULL) { 214 if (sysinit != SET_BEGIN(sysinit_set)) 215 free(sysinit, M_TEMP); 216 sysinit = newsysinit; 217 sysinit_end = newsysinit_end; 218 newsysinit = NULL; 219 newsysinit_end = NULL; 220 goto restart; 221 } 222 } 223 224 panic("Shouldn't get here!"); 225 /* NOTREACHED*/ 226 } 227 228 229 /* 230 *************************************************************************** 231 **** 232 **** The following SYSINIT's belong elsewhere, but have not yet 233 **** been moved. 234 **** 235 *************************************************************************** 236 */ 237 static void 238 print_caddr_t(void *data __unused) 239 { 240 printf("%s", (char *)data); 241 } 242 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright) 243 244 /* 245 * Leave the critical section that protected us from spurious interrupts 246 * so device probes work. 247 */ 248 static void 249 leavecrit(void *dummy __unused) 250 { 251 crit_exit(); 252 KKASSERT(!IN_CRITICAL_SECT(curthread)); 253 if (bootverbose) 254 printf("Leaving critical section, allowing interrupts\n"); 255 } 256 SYSINIT(leavecrit, SI_SUB_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL) 257 258 /* 259 *************************************************************************** 260 **** 261 **** The two following SYSINT's are proc0 specific glue code. I am not 262 **** convinced that they can not be safely combined, but their order of 263 **** operation has been maintained as the same as the original init_main.c 264 **** for right now. 265 **** 266 **** These probably belong in init_proc.c or kern_proc.c, since they 267 **** deal with proc0 (the fork template process). 268 **** 269 *************************************************************************** 270 */ 271 /* ARGSUSED*/ 272 static void 273 proc0_init(void *dummy __unused) 274 { 275 struct filedesc *fdp; 276 struct proc *p; 277 struct lwp *lp; 278 unsigned i; 279 280 p = &proc0; 281 lp = &proc0.p_lwp; /* XXX lwp to be: lwp0 */ 282 283 /* 284 * Initialize process and pgrp structures. 285 */ 286 procinit(); 287 288 /* 289 * additional VM structures 290 */ 291 vm_init2(); 292 293 /* 294 * Create process 0 (the swapper). 295 */ 296 LIST_INSERT_HEAD(&allproc, p, p_list); 297 p->p_pgrp = &pgrp0; 298 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 299 LIST_INIT(&pgrp0.pg_members); 300 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 301 302 pgrp0.pg_session = &session0; 303 session0.s_count = 1; 304 session0.s_leader = p; 305 306 p->p_sysent = &aout_sysvec; 307 TAILQ_INIT(&lp->lwp_sysmsgq); 308 309 p->p_flag = P_SYSTEM; 310 p->p_stat = SRUN; 311 p->p_nice = NZERO; 312 p->p_rtprio.type = RTP_PRIO_NORMAL; 313 p->p_rtprio.prio = 0; 314 p->p_lwp.lwp_rtprio = p->p_rtprio; 315 316 p->p_peers = 0; 317 p->p_leader = p; 318 319 bcopy("swapper", p->p_comm, sizeof ("swapper")); 320 bcopy("swapper", thread0.td_comm, sizeof ("swapper")); 321 322 /* Create credentials. */ 323 p->p_ucred = crget(); 324 p->p_ucred->cr_ruidinfo = uifind(0); 325 p->p_ucred->cr_ngroups = 1; /* group 0 */ 326 p->p_ucred->cr_uidinfo = uifind(0); 327 328 /* Don't jail it */ 329 p->p_ucred->cr_prison = NULL; 330 331 /* Create procsig. */ 332 p->p_procsig = &procsig0; 333 p->p_procsig->ps_refcnt = 1; 334 335 /* Initialize signal state for process 0. */ 336 siginit(&proc0); 337 338 /* Create the file descriptor table. */ 339 fdp = &filedesc0; 340 p->p_fd = fdp; 341 p->p_fdtol = NULL; 342 fdp->fd_refcnt = 1; 343 fdp->fd_cmask = cmask; 344 fdp->fd_files = fdp->fd_builtin_files; 345 fdp->fd_nfiles = NDFILE; 346 347 /* Create the limits structures. */ 348 p->p_limit = &limit0; 349 for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++) 350 limit0.pl_rlimit[i].rlim_cur = 351 limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY; 352 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = 353 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 354 limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = 355 limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 356 i = ptoa(vmstats.v_free_count); 357 limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i; 358 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i; 359 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3; 360 limit0.p_cpulimit = RLIM_INFINITY; 361 limit0.p_refcnt = 1; 362 363 /* Allocate a prototype map so we have something to fork. */ 364 pmap_pinit0(vmspace_pmap(&vmspace0)); 365 p->p_vmspace = &vmspace0; 366 vmspace0.vm_refcnt = 1; 367 vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS), 368 trunc_page(VM_MAXUSER_ADDRESS)); 369 vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0); 370 371 /* 372 * We continue to place resource usage info and signal 373 * actions in the user struct so they're pageable. 374 */ 375 p->p_stats = &p->p_addr->u_stats; 376 p->p_sigacts = &p->p_addr->u_sigacts; 377 378 /* 379 * Charge root for one process. 380 */ 381 (void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0); 382 383 } 384 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL) 385 386 /* ARGSUSED*/ 387 static void 388 proc0_post(void *dummy __unused) 389 { 390 struct timespec ts; 391 struct proc *p; 392 393 /* 394 * Now we can look at the time, having had a chance to verify the 395 * time from the file system. Pretend that proc0 started now. 396 */ 397 FOREACH_PROC_IN_SYSTEM(p) { 398 microtime(&p->p_start); 399 } 400 401 /* 402 * Give the ``random'' number generator a thump. 403 * XXX: Does read_random() contain enough bits to be used here ? 404 */ 405 nanotime(&ts); 406 srandom(ts.tv_sec ^ ts.tv_nsec); 407 } 408 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL) 409 410 /* 411 *************************************************************************** 412 **** 413 **** The following SYSINIT's and glue code should be moved to the 414 **** respective files on a per subsystem basis. 415 **** 416 *************************************************************************** 417 */ 418 419 420 /* 421 *************************************************************************** 422 **** 423 **** The following code probably belongs in another file, like 424 **** kern/init_init.c. 425 **** 426 *************************************************************************** 427 */ 428 429 /* 430 * List of paths to try when searching for "init". 431 */ 432 static char init_path[MAXPATHLEN] = 433 #ifdef INIT_PATH 434 __XSTRING(INIT_PATH); 435 #else 436 "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall"; 437 #endif 438 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, ""); 439 440 /* 441 * Start the initial user process; try exec'ing each pathname in init_path. 442 * The program is invoked with one argument containing the boot flags. 443 * 444 * The MP lock is held on entry. 445 */ 446 static void 447 start_init(void *dummy) 448 { 449 vm_offset_t addr; 450 struct execve_args args; 451 int options, error; 452 char *var, *path, *next, *s; 453 char *ucp, **uap, *arg0, *arg1; 454 struct proc *p; 455 struct lwp *lp; 456 struct mount *mp; 457 struct vnode *vp; 458 459 p = curproc; 460 461 KKASSERT(p->p_nthreads == 1); 462 463 lp = LIST_FIRST(&p->p_lwps); 464 465 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */ 466 mp = mountlist_boot_getfirst(); 467 if (VFS_ROOT(mp, &vp)) 468 panic("cannot find root vnode"); 469 if (mp->mnt_ncp == NULL) { 470 mp->mnt_ncp = cache_allocroot(mp, vp); 471 cache_unlock(mp->mnt_ncp); /* leave ref intact */ 472 } 473 p->p_fd->fd_cdir = vp; 474 vref(p->p_fd->fd_cdir); 475 p->p_fd->fd_rdir = vp; 476 vref(p->p_fd->fd_rdir); 477 vfs_cache_setroot(vp, cache_hold(mp->mnt_ncp)); 478 VOP_UNLOCK(vp, 0, curthread); /* leave ref intact */ 479 p->p_fd->fd_ncdir = cache_hold(mp->mnt_ncp); 480 p->p_fd->fd_nrdir = cache_hold(mp->mnt_ncp); 481 482 /* 483 * Need just enough stack to hold the faked-up "execve()" arguments. 484 */ 485 addr = trunc_page(USRSTACK - PAGE_SIZE); 486 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 487 FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0) 488 panic("init: couldn't allocate argument space"); 489 p->p_vmspace->vm_maxsaddr = (caddr_t)addr; 490 p->p_vmspace->vm_ssize = 1; 491 492 if ((var = getenv("init_path")) != NULL) { 493 strncpy(init_path, var, sizeof init_path); 494 init_path[sizeof init_path - 1] = 0; 495 } 496 if ((var = getenv("kern.fallback_elf_brand")) != NULL) 497 fallback_elf_brand = strtol(var, NULL, 0); 498 499 for (path = init_path; *path != '\0'; path = next) { 500 while (*path == ':') 501 path++; 502 if (*path == '\0') 503 break; 504 for (next = path; *next != '\0' && *next != ':'; next++) 505 /* nothing */ ; 506 if (bootverbose) 507 printf("start_init: trying %.*s\n", (int)(next - path), 508 path); 509 510 /* 511 * Move out the boot flag argument. 512 */ 513 options = 0; 514 ucp = (char *)USRSTACK; 515 (void)subyte(--ucp, 0); /* trailing zero */ 516 if (boothowto & RB_SINGLE) { 517 (void)subyte(--ucp, 's'); 518 options = 1; 519 } 520 #ifdef notyet 521 if (boothowto & RB_FASTBOOT) { 522 (void)subyte(--ucp, 'f'); 523 options = 1; 524 } 525 #endif 526 527 #ifdef BOOTCDROM 528 (void)subyte(--ucp, 'C'); 529 options = 1; 530 #endif 531 if (options == 0) 532 (void)subyte(--ucp, '-'); 533 (void)subyte(--ucp, '-'); /* leading hyphen */ 534 arg1 = ucp; 535 536 /* 537 * Move out the file name (also arg 0). 538 */ 539 (void)subyte(--ucp, 0); 540 for (s = next - 1; s >= path; s--) 541 (void)subyte(--ucp, *s); 542 arg0 = ucp; 543 544 /* 545 * Move out the arg pointers. 546 */ 547 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1)); 548 (void)suword((caddr_t)--uap, (long)0); /* terminator */ 549 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1); 550 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0); 551 552 /* 553 * Point at the arguments. 554 */ 555 args.fname = arg0; 556 args.argv = uap; 557 args.envv = NULL; 558 559 /* 560 * Now try to exec the program. If can't for any reason 561 * other than it doesn't exist, complain. 562 * 563 * Otherwise, return via fork_trampoline() all the way 564 * to user mode as init! 565 * 566 * WARNING! We may have been moved to another cpu after 567 * acquiring the current user process designation. The 568 * MP lock will migrate with us though so we still have to 569 * release it. 570 */ 571 if ((error = execve(&args)) == 0) { 572 if (lp->lwp_thread->td_gd->gd_uschedcp != lp) 573 lp->lwp_proc->p_usched->acquire_curproc(lp); 574 rel_mplock(); 575 return; 576 } 577 if (error != ENOENT) 578 printf("exec %.*s: error %d\n", (int)(next - path), 579 path, error); 580 } 581 printf("init: not found in path %s\n", init_path); 582 panic("no init"); 583 } 584 585 /* 586 * Like kthread_create(), but runs in it's own address space. 587 * We do this early to reserve pid 1. 588 * 589 * Note special case - do not make it runnable yet. Other work 590 * in progress will change this more. 591 */ 592 static void 593 create_init(const void *udata __unused) 594 { 595 int error; 596 597 crit_enter(); 598 error = fork1(&proc0.p_lwp, RFFDG | RFPROC, &initproc); 599 if (error) 600 panic("cannot fork init: %d", error); 601 initproc->p_flag |= P_SYSTEM; 602 cpu_set_fork_handler(initproc, start_init, NULL); 603 crit_exit(); 604 } 605 SYSINIT(init,SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL) 606 607 /* 608 * Make it runnable now. 609 */ 610 static void 611 kick_init(const void *udata __unused) 612 { 613 start_forked_proc(&proc0.p_lwp, initproc); 614 } 615 SYSINIT(kickinit,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL) 616 617 /* 618 * Machine independant globaldata initialization 619 * 620 * WARNING! Called from early boot, 'mycpu' may not work yet. 621 */ 622 void 623 mi_gdinit(struct globaldata *gd, int cpuid) 624 { 625 TAILQ_INIT(&gd->gd_tdfreeq); /* for pmap_{new,dispose}_thread() */ 626 TAILQ_INIT(&gd->gd_systimerq); 627 gd->gd_cpuid = cpuid; 628 gd->gd_cpumask = (cpumask_t)1 << cpuid; 629 lwkt_gdinit(gd); 630 vm_map_entry_reserve_cpu_init(gd); 631 sleep_gdinit(gd); 632 } 633 634 635