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 * $FreeBSD$ 43 */ 44 45 #include "opt_init_path.h" 46 47 #include <sys/param.h> 48 #include <sys/kernel.h> 49 #include <sys/file.h> 50 #include <sys/filedesc.h> 51 #include <sys/ktr.h> 52 #include <sys/lock.h> 53 #include <sys/mount.h> 54 #include <sys/mutex.h> 55 #include <sys/sysctl.h> 56 #include <sys/proc.h> 57 #include <sys/resourcevar.h> 58 #include <sys/systm.h> 59 #include <sys/signalvar.h> 60 #include <sys/vnode.h> 61 #include <sys/sysent.h> 62 #include <sys/reboot.h> 63 #include <sys/sx.h> 64 #include <sys/sysproto.h> 65 #include <sys/vmmeter.h> 66 #include <sys/unistd.h> 67 #include <sys/malloc.h> 68 #include <sys/conf.h> 69 70 #include <machine/cpu.h> 71 72 #include <vm/vm.h> 73 #include <vm/vm_param.h> 74 #include <vm/pmap.h> 75 #include <vm/vm_map.h> 76 #include <sys/user.h> 77 #include <sys/copyright.h> 78 79 void mi_startup(void); /* Should be elsewhere */ 80 81 /* Components of the first process -- never freed. */ 82 static struct session session0; 83 static struct pgrp pgrp0; 84 struct proc proc0; 85 struct thread *thread0; 86 static struct procsig procsig0; 87 static struct filedesc0 filedesc0; 88 static struct plimit limit0; 89 static struct vmspace vmspace0; 90 struct proc *initproc; 91 92 int cmask = CMASK; 93 extern int fallback_elf_brand; 94 95 struct vnode *rootvp; 96 int boothowto = 0; /* initialized so that it can be patched */ 97 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, ""); 98 int bootverbose; 99 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, ""); 100 101 /* 102 * This ensures that there is at least one entry so that the sysinit_set 103 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never 104 * executed. 105 */ 106 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL) 107 108 /* 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. 111 */ 112 SET_DECLARE(sysinit_set, struct sysinit); 113 struct sysinit **sysinit, **sysinit_end; 114 struct sysinit **newsysinit, **newsysinit_end; 115 116 /* 117 * Merge a new sysinit set into the current set, reallocating it if 118 * necessary. This can only be called after malloc is running. 119 */ 120 void 121 sysinit_add(struct sysinit **set, struct sysinit **set_end) 122 { 123 struct sysinit **newset; 124 struct sysinit **sipp; 125 struct sysinit **xipp; 126 int count; 127 128 count = set_end - set; 129 if (newsysinit) 130 count += newsysinit_end - newsysinit; 131 else 132 count += sysinit_end - sysinit; 133 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT); 134 if (newset == NULL) 135 panic("cannot malloc for sysinit"); 136 xipp = newset; 137 if (newsysinit) 138 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 139 *xipp++ = *sipp; 140 else 141 for (sipp = sysinit; sipp < sysinit_end; sipp++) 142 *xipp++ = *sipp; 143 for (sipp = set; sipp < set_end; sipp++) 144 *xipp++ = *sipp; 145 if (newsysinit) 146 free(newsysinit, M_TEMP); 147 newsysinit = newset; 148 newsysinit_end = newset + count; 149 } 150 151 /* 152 * System startup; initialize the world, create process 0, mount root 153 * filesystem, and fork to create init and pagedaemon. Most of the 154 * hard work is done in the lower-level initialization routines including 155 * startup(), which does memory initialization and autoconfiguration. 156 * 157 * This allows simple addition of new kernel subsystems that require 158 * boot time initialization. It also allows substitution of subsystem 159 * (for instance, a scheduler, kernel profiler, or VM system) by object 160 * module. Finally, it allows for optional "kernel threads". 161 */ 162 void 163 mi_startup(void) 164 { 165 166 register struct sysinit **sipp; /* system initialization*/ 167 register struct sysinit **xipp; /* interior loop of sort*/ 168 register struct sysinit *save; /* bubble*/ 169 170 if (sysinit == NULL) { 171 sysinit = SET_BEGIN(sysinit_set); 172 sysinit_end = SET_LIMIT(sysinit_set); 173 } 174 175 restart: 176 /* 177 * Perform a bubble sort of the system initialization objects by 178 * their subsystem (primary key) and order (secondary key). 179 */ 180 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 181 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 182 if ((*sipp)->subsystem < (*xipp)->subsystem || 183 ((*sipp)->subsystem == (*xipp)->subsystem && 184 (*sipp)->order <= (*xipp)->order)) 185 continue; /* skip*/ 186 save = *sipp; 187 *sipp = *xipp; 188 *xipp = save; 189 } 190 } 191 192 /* 193 * Traverse the (now) ordered list of system initialization tasks. 194 * Perform each task, and continue on to the next task. 195 * 196 * The last item on the list is expected to be the scheduler, 197 * which will not return. 198 */ 199 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 200 201 if ((*sipp)->subsystem == SI_SUB_DUMMY) 202 continue; /* skip dummy task(s)*/ 203 204 if ((*sipp)->subsystem == SI_SUB_DONE) 205 continue; 206 207 /* Call function */ 208 (*((*sipp)->func))((*sipp)->udata); 209 210 /* Check off the one we're just done */ 211 (*sipp)->subsystem = SI_SUB_DONE; 212 213 /* Check if we've installed more sysinit items via KLD */ 214 if (newsysinit != NULL) { 215 if (sysinit != SET_BEGIN(sysinit_set)) 216 free(sysinit, M_TEMP); 217 sysinit = newsysinit; 218 sysinit_end = newsysinit_end; 219 newsysinit = NULL; 220 newsysinit_end = NULL; 221 goto restart; 222 } 223 } 224 225 panic("Shouldn't get here!"); 226 /* NOTREACHED*/ 227 } 228 229 230 /* 231 *************************************************************************** 232 **** 233 **** The following SYSINIT's belong elsewhere, but have not yet 234 **** been moved. 235 **** 236 *************************************************************************** 237 */ 238 static void 239 print_caddr_t(void *data __unused) 240 { 241 printf("%s", (char *)data); 242 } 243 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright) 244 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, version) 245 246 static void 247 set_boot_verbose(void *data __unused) 248 { 249 250 if (boothowto & RB_VERBOSE) 251 bootverbose++; 252 } 253 SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL) 254 255 /* 256 *************************************************************************** 257 **** 258 **** The two following SYSINT's are proc0 specific glue code. I am not 259 **** convinced that they can not be safely combined, but their order of 260 **** operation has been maintained as the same as the original init_main.c 261 **** for right now. 262 **** 263 **** These probably belong in init_proc.c or kern_proc.c, since they 264 **** deal with proc0 (the fork template process). 265 **** 266 *************************************************************************** 267 */ 268 /* ARGSUSED*/ 269 static void 270 proc0_init(void *dummy __unused) 271 { 272 register struct proc *p; 273 register struct filedesc0 *fdp; 274 register unsigned i; 275 struct thread *td; 276 277 GIANT_REQUIRED; 278 /* 279 * This assumes the proc0 struct has already been linked 280 * using proc_linkup() in the machine specific initialisation 281 * e.g. i386_init() 282 */ 283 p = &proc0; 284 td = thread0; 285 286 /* 287 * Initialize magic number. 288 */ 289 p->p_magic = P_MAGIC; 290 291 /* 292 * Initialize process and pgrp structures. 293 */ 294 procinit(); 295 296 /* 297 * Initialize sleep queue hash table 298 */ 299 sleepinit(); 300 301 /* 302 * additional VM structures 303 */ 304 vm_init2(); 305 306 /* 307 * Create process 0 (the swapper). 308 */ 309 LIST_INSERT_HEAD(&allproc, p, p_list); 310 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash); 311 p->p_pgrp = &pgrp0; 312 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 313 LIST_INIT(&pgrp0.pg_members); 314 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 315 316 pgrp0.pg_session = &session0; 317 session0.s_count = 1; 318 session0.s_leader = p; 319 320 #ifdef __ELF__ 321 p->p_sysent = &elf_freebsd_sysvec; 322 #else 323 p->p_sysent = &aout_sysvec; 324 #endif 325 326 p->p_flag = P_SYSTEM; 327 p->p_sflag = PS_INMEM; 328 p->p_stat = SRUN; 329 p->p_ksegrp.kg_nice = NZERO; 330 p->p_ksegrp.kg_pri.pri_class = PRI_TIMESHARE; 331 p->p_ksegrp.kg_pri.pri_level = PVM; 332 p->p_ksegrp.kg_pri.pri_native = PUSER; 333 p->p_ksegrp.kg_pri.pri_user = PUSER; 334 335 p->p_peers = 0; 336 p->p_leader = p; 337 338 bcopy("swapper", p->p_comm, sizeof ("swapper")); 339 340 callout_init(&p->p_itcallout, 0); 341 callout_init(&td->td_slpcallout, 1); 342 343 /* Create credentials. */ 344 p->p_ucred = crget(); 345 p->p_ucred->cr_ngroups = 1; /* group 0 */ 346 p->p_ucred->cr_uidinfo = uifind(0); 347 p->p_ucred->cr_ruidinfo = uifind(0); 348 p->p_ucred->cr_prison = NULL; /* Don't jail it. */ 349 td->td_ucred = crhold(p->p_ucred); 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(&proc0); 357 358 /* Create the file descriptor table. */ 359 fdp = &filedesc0; 360 p->p_fd = &fdp->fd_fd; 361 mtx_init(&fdp->fd_fd.fd_mtx, "struct filedesc", MTX_DEF); 362 fdp->fd_fd.fd_refcnt = 1; 363 fdp->fd_fd.fd_cmask = cmask; 364 fdp->fd_fd.fd_ofiles = fdp->fd_dfiles; 365 fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags; 366 fdp->fd_fd.fd_nfiles = NDFILE; 367 368 /* Create the limits structures. */ 369 p->p_limit = &limit0; 370 for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++) 371 limit0.pl_rlimit[i].rlim_cur = 372 limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY; 373 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = 374 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 375 limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = 376 limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 377 i = ptoa(cnt.v_free_count); 378 limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i; 379 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i; 380 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3; 381 limit0.p_cpulimit = RLIM_INFINITY; 382 limit0.p_refcnt = 1; 383 384 /* Allocate a prototype map so we have something to fork. */ 385 pmap_pinit0(vmspace_pmap(&vmspace0)); 386 p->p_vmspace = &vmspace0; 387 vmspace0.vm_refcnt = 1; 388 vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS), 389 trunc_page(VM_MAXUSER_ADDRESS)); 390 vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0); 391 392 /* 393 * We continue to place resource usage info and signal 394 * actions in the user struct so they're pageable. 395 */ 396 p->p_stats = &p->p_uarea->u_stats; 397 p->p_sigacts = &p->p_uarea->u_sigacts; 398 399 /* 400 * Charge root for one process. 401 */ 402 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0); 403 } 404 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL) 405 406 /* ARGSUSED*/ 407 static void 408 proc0_post(void *dummy __unused) 409 { 410 struct timespec ts; 411 struct proc *p; 412 413 /* 414 * Now we can look at the time, having had a chance to verify the 415 * time from the file system. Pretend that proc0 started now. 416 */ 417 sx_slock(&allproc_lock); 418 LIST_FOREACH(p, &allproc, p_list) { 419 microtime(&p->p_stats->p_start); 420 p->p_runtime = 0; 421 } 422 sx_sunlock(&allproc_lock); 423 microuptime(PCPU_PTR(switchtime)); 424 PCPU_SET(switchticks, ticks); 425 426 /* 427 * Give the ``random'' number generator a thump. 428 */ 429 nanotime(&ts); 430 srandom(ts.tv_sec ^ ts.tv_nsec); 431 } 432 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL) 433 434 /* 435 *************************************************************************** 436 **** 437 **** The following SYSINIT's and glue code should be moved to the 438 **** respective files on a per subsystem basis. 439 **** 440 *************************************************************************** 441 */ 442 443 444 /* 445 *************************************************************************** 446 **** 447 **** The following code probably belongs in another file, like 448 **** kern/init_init.c. 449 **** 450 *************************************************************************** 451 */ 452 453 /* 454 * List of paths to try when searching for "init". 455 */ 456 static char init_path[MAXPATHLEN] = 457 #ifdef INIT_PATH 458 __XSTRING(INIT_PATH); 459 #else 460 "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall"; 461 #endif 462 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, 463 "Path used to search the init process"); 464 465 /* 466 * Start the initial user process; try exec'ing each pathname in init_path. 467 * The program is invoked with one argument containing the boot flags. 468 */ 469 static void 470 start_init(void *dummy) 471 { 472 vm_offset_t addr; 473 struct execve_args args; 474 int options, error; 475 char *var, *path, *next, *s; 476 char *ucp, **uap, *arg0, *arg1; 477 struct thread *td; 478 struct proc *p; 479 int init_does_devfs = 0; 480 481 mtx_lock(&Giant); 482 483 GIANT_REQUIRED; 484 485 td = curthread; 486 p = td->td_proc; 487 488 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */ 489 if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode)) 490 panic("cannot find root vnode"); 491 FILEDESC_LOCK(p->p_fd); 492 p->p_fd->fd_cdir = rootvnode; 493 VREF(p->p_fd->fd_cdir); 494 p->p_fd->fd_rdir = rootvnode; 495 VREF(p->p_fd->fd_rdir); 496 FILEDESC_UNLOCK(p->p_fd); 497 VOP_UNLOCK(rootvnode, 0, td); 498 499 if (devfs_present) { 500 /* 501 * For disk based systems, we probably cannot do this yet 502 * since the fs will be read-only. But a NFS root 503 * might be ok. It is worth a shot. 504 */ 505 error = vn_mkdir("/dev", 0700, UIO_SYSSPACE, td); 506 if (error == EEXIST) 507 error = 0; 508 if (error == 0) 509 error = vfs_mount(td, "devfs", "/dev", 0, 0); 510 if (error != 0) 511 init_does_devfs = 1; 512 } 513 514 /* 515 * Need just enough stack to hold the faked-up "execve()" arguments. 516 */ 517 addr = trunc_page(USRSTACK - PAGE_SIZE); 518 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 519 FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0) 520 panic("init: couldn't allocate argument space"); 521 p->p_vmspace->vm_maxsaddr = (caddr_t)addr; 522 p->p_vmspace->vm_ssize = 1; 523 524 if ((var = getenv("init_path")) != NULL) { 525 strncpy(init_path, var, sizeof init_path); 526 init_path[sizeof init_path - 1] = 0; 527 } 528 if ((var = getenv("kern.fallback_elf_brand")) != NULL) 529 fallback_elf_brand = strtol(var, NULL, 0); 530 531 for (path = init_path; *path != '\0'; path = next) { 532 while (*path == ':') 533 path++; 534 if (*path == '\0') 535 break; 536 for (next = path; *next != '\0' && *next != ':'; next++) 537 /* nothing */ ; 538 if (bootverbose) 539 printf("start_init: trying %.*s\n", (int)(next - path), 540 path); 541 542 /* 543 * Move out the boot flag argument. 544 */ 545 options = 0; 546 ucp = (char *)USRSTACK; 547 (void)subyte(--ucp, 0); /* trailing zero */ 548 if (boothowto & RB_SINGLE) { 549 (void)subyte(--ucp, 's'); 550 options = 1; 551 } 552 #ifdef notyet 553 if (boothowto & RB_FASTBOOT) { 554 (void)subyte(--ucp, 'f'); 555 options = 1; 556 } 557 #endif 558 559 #ifdef BOOTCDROM 560 (void)subyte(--ucp, 'C'); 561 options = 1; 562 #endif 563 if (init_does_devfs) { 564 (void)subyte(--ucp, 'd'); 565 options = 1; 566 } 567 568 if (options == 0) 569 (void)subyte(--ucp, '-'); 570 (void)subyte(--ucp, '-'); /* leading hyphen */ 571 arg1 = ucp; 572 573 /* 574 * Move out the file name (also arg 0). 575 */ 576 (void)subyte(--ucp, 0); 577 for (s = next - 1; s >= path; s--) 578 (void)subyte(--ucp, *s); 579 arg0 = ucp; 580 581 /* 582 * Move out the arg pointers. 583 */ 584 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1)); 585 (void)suword((caddr_t)--uap, (long)0); /* terminator */ 586 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1); 587 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0); 588 589 /* 590 * Point at the arguments. 591 */ 592 args.fname = arg0; 593 args.argv = uap; 594 args.envv = NULL; 595 596 /* 597 * Now try to exec the program. If can't for any reason 598 * other than it doesn't exist, complain. 599 * 600 * Otherwise, return via fork_trampoline() all the way 601 * to user mode as init! 602 */ 603 if ((error = execve(td, &args)) == 0) { 604 mtx_unlock(&Giant); 605 return; 606 } 607 if (error != ENOENT) 608 printf("exec %.*s: error %d\n", (int)(next - path), 609 path, error); 610 } 611 printf("init: not found in path %s\n", init_path); 612 panic("no init"); 613 } 614 615 /* 616 * Like kthread_create(), but runs in it's own address space. 617 * We do this early to reserve pid 1. 618 * 619 * Note special case - do not make it runnable yet. Other work 620 * in progress will change this more. 621 */ 622 static void 623 create_init(const void *udata __unused) 624 { 625 int error; 626 627 error = fork1(thread0, RFFDG | RFPROC | RFSTOPPED, &initproc); 628 if (error) 629 panic("cannot fork init: %d\n", error); 630 PROC_LOCK(initproc); 631 initproc->p_flag |= P_SYSTEM; 632 PROC_UNLOCK(initproc); 633 mtx_lock_spin(&sched_lock); 634 initproc->p_sflag |= PS_INMEM; 635 mtx_unlock_spin(&sched_lock); 636 cpu_set_fork_handler(&initproc->p_thread, start_init, NULL); 637 } 638 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL) 639 640 /* 641 * Make it runnable now. 642 */ 643 static void 644 kick_init(const void *udata __unused) 645 { 646 647 mtx_lock_spin(&sched_lock); 648 initproc->p_stat = SRUN; 649 setrunqueue(&initproc->p_thread); /* XXXKSE */ 650 mtx_unlock_spin(&sched_lock); 651 } 652 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL) 653