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 */ 43 44 #include <sys/cdefs.h> 45 __FBSDID("$FreeBSD$"); 46 47 #include "opt_ddb.h" 48 #include "opt_init_path.h" 49 50 #include <sys/param.h> 51 #include <sys/kernel.h> 52 #include <sys/exec.h> 53 #include <sys/file.h> 54 #include <sys/filedesc.h> 55 #include <sys/jail.h> 56 #include <sys/ktr.h> 57 #include <sys/lock.h> 58 #include <sys/mount.h> 59 #include <sys/mutex.h> 60 #include <sys/syscallsubr.h> 61 #include <sys/sysctl.h> 62 #include <sys/proc.h> 63 #include <sys/resourcevar.h> 64 #include <sys/systm.h> 65 #include <sys/signalvar.h> 66 #include <sys/vnode.h> 67 #include <sys/sysent.h> 68 #include <sys/reboot.h> 69 #include <sys/sched.h> 70 #include <sys/sx.h> 71 #include <sys/sysproto.h> 72 #include <sys/vmmeter.h> 73 #include <sys/unistd.h> 74 #include <sys/malloc.h> 75 #include <sys/conf.h> 76 #include <sys/cpuset.h> 77 78 #include <machine/cpu.h> 79 80 #include <security/audit/audit.h> 81 #include <security/mac/mac_framework.h> 82 83 #include <vm/vm.h> 84 #include <vm/vm_param.h> 85 #include <vm/pmap.h> 86 #include <vm/vm_map.h> 87 #include <sys/copyright.h> 88 89 #include <ddb/ddb.h> 90 #include <ddb/db_sym.h> 91 92 void mi_startup(void); /* Should be elsewhere */ 93 94 /* Components of the first process -- never freed. */ 95 static struct session session0; 96 static struct pgrp pgrp0; 97 struct proc proc0; 98 struct thread thread0 __aligned(16); 99 struct vmspace vmspace0; 100 struct proc *initproc; 101 102 int boothowto = 0; /* initialized so that it can be patched */ 103 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, ""); 104 int bootverbose; 105 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, ""); 106 107 /* 108 * This ensures that there is at least one entry so that the sysinit_set 109 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never 110 * executed. 111 */ 112 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL); 113 114 /* 115 * The sysinit table itself. Items are checked off as the are run. 116 * If we want to register new sysinit types, add them to newsysinit. 117 */ 118 SET_DECLARE(sysinit_set, struct sysinit); 119 struct sysinit **sysinit, **sysinit_end; 120 struct sysinit **newsysinit, **newsysinit_end; 121 122 /* 123 * Merge a new sysinit set into the current set, reallocating it if 124 * necessary. This can only be called after malloc is running. 125 */ 126 void 127 sysinit_add(struct sysinit **set, struct sysinit **set_end) 128 { 129 struct sysinit **newset; 130 struct sysinit **sipp; 131 struct sysinit **xipp; 132 int count; 133 134 count = set_end - set; 135 if (newsysinit) 136 count += newsysinit_end - newsysinit; 137 else 138 count += sysinit_end - sysinit; 139 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT); 140 if (newset == NULL) 141 panic("cannot malloc for sysinit"); 142 xipp = newset; 143 if (newsysinit) 144 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 145 *xipp++ = *sipp; 146 else 147 for (sipp = sysinit; sipp < sysinit_end; sipp++) 148 *xipp++ = *sipp; 149 for (sipp = set; sipp < set_end; sipp++) 150 *xipp++ = *sipp; 151 if (newsysinit) 152 free(newsysinit, M_TEMP); 153 newsysinit = newset; 154 newsysinit_end = newset + count; 155 } 156 157 /* 158 * System startup; initialize the world, create process 0, mount root 159 * filesystem, and fork to create init and pagedaemon. Most of the 160 * hard work is done in the lower-level initialization routines including 161 * startup(), which does memory initialization and autoconfiguration. 162 * 163 * This allows simple addition of new kernel subsystems that require 164 * boot time initialization. It also allows substitution of subsystem 165 * (for instance, a scheduler, kernel profiler, or VM system) by object 166 * module. Finally, it allows for optional "kernel threads". 167 */ 168 void 169 mi_startup(void) 170 { 171 172 register struct sysinit **sipp; /* system initialization*/ 173 register struct sysinit **xipp; /* interior loop of sort*/ 174 register struct sysinit *save; /* bubble*/ 175 176 #if defined(VERBOSE_SYSINIT) 177 int last; 178 int verbose; 179 #endif 180 181 if (sysinit == NULL) { 182 sysinit = SET_BEGIN(sysinit_set); 183 sysinit_end = SET_LIMIT(sysinit_set); 184 } 185 186 restart: 187 /* 188 * Perform a bubble sort of the system initialization objects by 189 * their subsystem (primary key) and order (secondary key). 190 */ 191 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 192 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 193 if ((*sipp)->subsystem < (*xipp)->subsystem || 194 ((*sipp)->subsystem == (*xipp)->subsystem && 195 (*sipp)->order <= (*xipp)->order)) 196 continue; /* skip*/ 197 save = *sipp; 198 *sipp = *xipp; 199 *xipp = save; 200 } 201 } 202 203 #if defined(VERBOSE_SYSINIT) 204 last = SI_SUB_COPYRIGHT; 205 verbose = 0; 206 #if !defined(DDB) 207 printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n"); 208 #endif 209 #endif 210 211 /* 212 * Traverse the (now) ordered list of system initialization tasks. 213 * Perform each task, and continue on to the next task. 214 * 215 * The last item on the list is expected to be the scheduler, 216 * which will not return. 217 */ 218 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 219 220 if ((*sipp)->subsystem == SI_SUB_DUMMY) 221 continue; /* skip dummy task(s)*/ 222 223 if ((*sipp)->subsystem == SI_SUB_DONE) 224 continue; 225 226 #if defined(VERBOSE_SYSINIT) 227 if ((*sipp)->subsystem > last) { 228 verbose = 1; 229 last = (*sipp)->subsystem; 230 printf("subsystem %x\n", last); 231 } 232 if (verbose) { 233 #if defined(DDB) 234 const char *name; 235 c_db_sym_t sym; 236 db_expr_t offset; 237 238 sym = db_search_symbol((vm_offset_t)(*sipp)->func, 239 DB_STGY_PROC, &offset); 240 db_symbol_values(sym, &name, NULL); 241 if (name != NULL) 242 printf(" %s(%p)... ", name, (*sipp)->udata); 243 else 244 #endif 245 printf(" %p(%p)... ", (*sipp)->func, 246 (*sipp)->udata); 247 } 248 #endif 249 250 /* Call function */ 251 (*((*sipp)->func))((*sipp)->udata); 252 253 #if defined(VERBOSE_SYSINIT) 254 if (verbose) 255 printf("done.\n"); 256 #endif 257 258 /* Check off the one we're just done */ 259 (*sipp)->subsystem = SI_SUB_DONE; 260 261 /* Check if we've installed more sysinit items via KLD */ 262 if (newsysinit != NULL) { 263 if (sysinit != SET_BEGIN(sysinit_set)) 264 free(sysinit, M_TEMP); 265 sysinit = newsysinit; 266 sysinit_end = newsysinit_end; 267 newsysinit = NULL; 268 newsysinit_end = NULL; 269 goto restart; 270 } 271 } 272 273 panic("Shouldn't get here!"); 274 /* NOTREACHED*/ 275 } 276 277 278 /* 279 *************************************************************************** 280 **** 281 **** The following SYSINIT's belong elsewhere, but have not yet 282 **** been moved. 283 **** 284 *************************************************************************** 285 */ 286 static void 287 print_caddr_t(void *data) 288 { 289 printf("%s", (char *)data); 290 } 291 292 static void 293 print_version(void *data __unused) 294 { 295 int len; 296 297 /* Strip a trailing newline from version. */ 298 len = strlen(version); 299 while (len > 0 && version[len - 1] == '\n') 300 len--; 301 printf("%.*s %s\n", len, version, machine); 302 } 303 304 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, 305 copyright); 306 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, 307 trademark); 308 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL); 309 310 #ifdef WITNESS 311 static char wit_warn[] = 312 "WARNING: WITNESS option enabled, expect reduced performance.\n"; 313 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1, 314 print_caddr_t, wit_warn); 315 SYSINIT(witwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 1, 316 print_caddr_t, wit_warn); 317 #endif 318 319 #ifdef DIAGNOSTIC 320 static char diag_warn[] = 321 "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n"; 322 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2, 323 print_caddr_t, diag_warn); 324 SYSINIT(diagwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 2, 325 print_caddr_t, diag_warn); 326 #endif 327 328 static void 329 set_boot_verbose(void *data __unused) 330 { 331 332 if (boothowto & RB_VERBOSE) 333 bootverbose++; 334 } 335 SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL); 336 337 static int 338 null_fetch_syscall_args(struct thread *td __unused, 339 struct syscall_args *sa __unused) 340 { 341 342 panic("null_fetch_syscall_args"); 343 } 344 345 static void 346 null_set_syscall_retval(struct thread *td __unused, int error __unused) 347 { 348 349 panic("null_set_syscall_retval"); 350 } 351 352 struct sysentvec null_sysvec = { 353 .sv_size = 0, 354 .sv_table = NULL, 355 .sv_mask = 0, 356 .sv_sigsize = 0, 357 .sv_sigtbl = NULL, 358 .sv_errsize = 0, 359 .sv_errtbl = NULL, 360 .sv_transtrap = NULL, 361 .sv_fixup = NULL, 362 .sv_sendsig = NULL, 363 .sv_sigcode = NULL, 364 .sv_szsigcode = NULL, 365 .sv_prepsyscall = NULL, 366 .sv_name = "null", 367 .sv_coredump = NULL, 368 .sv_imgact_try = NULL, 369 .sv_minsigstksz = 0, 370 .sv_pagesize = PAGE_SIZE, 371 .sv_minuser = VM_MIN_ADDRESS, 372 .sv_maxuser = VM_MAXUSER_ADDRESS, 373 .sv_usrstack = USRSTACK, 374 .sv_psstrings = PS_STRINGS, 375 .sv_stackprot = VM_PROT_ALL, 376 .sv_copyout_strings = NULL, 377 .sv_setregs = NULL, 378 .sv_fixlimit = NULL, 379 .sv_maxssiz = NULL, 380 .sv_flags = 0, 381 .sv_set_syscall_retval = null_set_syscall_retval, 382 .sv_fetch_syscall_args = null_fetch_syscall_args, 383 .sv_syscallnames = NULL, 384 }; 385 386 /* 387 *************************************************************************** 388 **** 389 **** The two following SYSINIT's are proc0 specific glue code. I am not 390 **** convinced that they can not be safely combined, but their order of 391 **** operation has been maintained as the same as the original init_main.c 392 **** for right now. 393 **** 394 **** These probably belong in init_proc.c or kern_proc.c, since they 395 **** deal with proc0 (the fork template process). 396 **** 397 *************************************************************************** 398 */ 399 /* ARGSUSED*/ 400 static void 401 proc0_init(void *dummy __unused) 402 { 403 struct proc *p; 404 struct thread *td; 405 vm_paddr_t pageablemem; 406 int i; 407 408 GIANT_REQUIRED; 409 p = &proc0; 410 td = &thread0; 411 412 /* 413 * Initialize magic number and osrel. 414 */ 415 p->p_magic = P_MAGIC; 416 p->p_osrel = osreldate; 417 418 /* 419 * Initialize thread and process structures. 420 */ 421 procinit(); /* set up proc zone */ 422 threadinit(); /* set up UMA zones */ 423 424 /* 425 * Initialise scheduler resources. 426 * Add scheduler specific parts to proc, thread as needed. 427 */ 428 schedinit(); /* scheduler gets its house in order */ 429 /* 430 * Initialize sleep queue hash table 431 */ 432 sleepinit(); 433 434 /* 435 * additional VM structures 436 */ 437 vm_init2(); 438 439 /* 440 * Create process 0 (the swapper). 441 */ 442 LIST_INSERT_HEAD(&allproc, p, p_list); 443 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash); 444 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 445 p->p_pgrp = &pgrp0; 446 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 447 LIST_INIT(&pgrp0.pg_members); 448 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 449 450 pgrp0.pg_session = &session0; 451 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF); 452 refcount_init(&session0.s_count, 1); 453 session0.s_leader = p; 454 455 p->p_sysent = &null_sysvec; 456 p->p_flag = P_SYSTEM | P_INMEM; 457 p->p_state = PRS_NORMAL; 458 knlist_init_mtx(&p->p_klist, &p->p_mtx); 459 STAILQ_INIT(&p->p_ktr); 460 p->p_nice = NZERO; 461 td->td_tid = PID_MAX + 1; 462 td->td_state = TDS_RUNNING; 463 td->td_pri_class = PRI_TIMESHARE; 464 td->td_user_pri = PUSER; 465 td->td_base_user_pri = PUSER; 466 td->td_priority = PVM; 467 td->td_base_pri = PUSER; 468 td->td_oncpu = 0; 469 td->td_flags = TDF_INMEM|TDP_KTHREAD; 470 td->td_cpuset = cpuset_thread0(); 471 prison0.pr_cpuset = cpuset_ref(td->td_cpuset); 472 p->p_peers = 0; 473 p->p_leader = p; 474 475 476 strncpy(p->p_comm, "kernel", sizeof (p->p_comm)); 477 strncpy(td->td_name, "swapper", sizeof (td->td_name)); 478 479 callout_init(&p->p_itcallout, CALLOUT_MPSAFE); 480 callout_init_mtx(&p->p_limco, &p->p_mtx, 0); 481 callout_init(&td->td_slpcallout, CALLOUT_MPSAFE); 482 483 /* Create credentials. */ 484 p->p_ucred = crget(); 485 p->p_ucred->cr_ngroups = 1; /* group 0 */ 486 p->p_ucred->cr_uidinfo = uifind(0); 487 p->p_ucred->cr_ruidinfo = uifind(0); 488 p->p_ucred->cr_prison = &prison0; 489 #ifdef AUDIT 490 audit_cred_kproc0(p->p_ucred); 491 #endif 492 #ifdef MAC 493 mac_cred_create_swapper(p->p_ucred); 494 #endif 495 td->td_ucred = crhold(p->p_ucred); 496 497 /* Create sigacts. */ 498 p->p_sigacts = sigacts_alloc(); 499 500 /* Initialize signal state for process 0. */ 501 siginit(&proc0); 502 503 /* Create the file descriptor table. */ 504 p->p_fd = fdinit(NULL); 505 p->p_fdtol = NULL; 506 507 /* Create the limits structures. */ 508 p->p_limit = lim_alloc(); 509 for (i = 0; i < RLIM_NLIMITS; i++) 510 p->p_limit->pl_rlimit[i].rlim_cur = 511 p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY; 512 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur = 513 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 514 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur = 515 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 516 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz; 517 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz; 518 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz; 519 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz; 520 /* Cast to avoid overflow on i386/PAE. */ 521 pageablemem = ptoa((vm_paddr_t)cnt.v_free_count); 522 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur = 523 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem; 524 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3; 525 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem; 526 p->p_cpulimit = RLIM_INFINITY; 527 528 p->p_stats = pstats_alloc(); 529 530 /* Allocate a prototype map so we have something to fork. */ 531 pmap_pinit0(vmspace_pmap(&vmspace0)); 532 p->p_vmspace = &vmspace0; 533 vmspace0.vm_refcnt = 1; 534 535 /* 536 * proc0 is not expected to enter usermode, so there is no special 537 * handling for sv_minuser here, like is done for exec_new_vmspace(). 538 */ 539 vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0), 540 p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser); 541 542 /*- 543 * call the init and ctor for the new thread and proc 544 * we wait to do this until all other structures 545 * are fairly sane. 546 */ 547 EVENTHANDLER_INVOKE(process_init, p); 548 EVENTHANDLER_INVOKE(thread_init, td); 549 EVENTHANDLER_INVOKE(process_ctor, p); 550 EVENTHANDLER_INVOKE(thread_ctor, td); 551 552 /* 553 * Charge root for one process. 554 */ 555 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0); 556 } 557 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL); 558 559 /* ARGSUSED*/ 560 static void 561 proc0_post(void *dummy __unused) 562 { 563 struct timespec ts; 564 struct proc *p; 565 struct rusage ru; 566 struct thread *td; 567 568 /* 569 * Now we can look at the time, having had a chance to verify the 570 * time from the filesystem. Pretend that proc0 started now. 571 */ 572 sx_slock(&allproc_lock); 573 FOREACH_PROC_IN_SYSTEM(p) { 574 microuptime(&p->p_stats->p_start); 575 PROC_SLOCK(p); 576 rufetch(p, &ru); /* Clears thread stats */ 577 PROC_SUNLOCK(p); 578 p->p_rux.rux_runtime = 0; 579 p->p_rux.rux_uticks = 0; 580 p->p_rux.rux_sticks = 0; 581 p->p_rux.rux_iticks = 0; 582 FOREACH_THREAD_IN_PROC(p, td) { 583 td->td_runtime = 0; 584 } 585 } 586 sx_sunlock(&allproc_lock); 587 PCPU_SET(switchtime, cpu_ticks()); 588 PCPU_SET(switchticks, ticks); 589 590 /* 591 * Give the ``random'' number generator a thump. 592 */ 593 nanotime(&ts); 594 srandom(ts.tv_sec ^ ts.tv_nsec); 595 } 596 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL); 597 598 static void 599 random_init(void *dummy __unused) 600 { 601 602 /* 603 * After CPU has been started we have some randomness on most 604 * platforms via get_cyclecount(). For platforms that don't 605 * we will reseed random(9) in proc0_post() as well. 606 */ 607 srandom(get_cyclecount()); 608 } 609 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL); 610 611 /* 612 *************************************************************************** 613 **** 614 **** The following SYSINIT's and glue code should be moved to the 615 **** respective files on a per subsystem basis. 616 **** 617 *************************************************************************** 618 */ 619 620 621 /* 622 *************************************************************************** 623 **** 624 **** The following code probably belongs in another file, like 625 **** kern/init_init.c. 626 **** 627 *************************************************************************** 628 */ 629 630 /* 631 * List of paths to try when searching for "init". 632 */ 633 static char init_path[MAXPATHLEN] = 634 #ifdef INIT_PATH 635 __XSTRING(INIT_PATH); 636 #else 637 "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init:/stand/sysinstall"; 638 #endif 639 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, 640 "Path used to search the init process"); 641 642 /* 643 * Shutdown timeout of init(8). 644 * Unused within kernel, but used to control init(8), hence do not remove. 645 */ 646 #ifndef INIT_SHUTDOWN_TIMEOUT 647 #define INIT_SHUTDOWN_TIMEOUT 120 648 #endif 649 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT; 650 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout, 651 CTLFLAG_RW, &init_shutdown_timeout, 0, ""); 652 653 /* 654 * Start the initial user process; try exec'ing each pathname in init_path. 655 * The program is invoked with one argument containing the boot flags. 656 */ 657 static void 658 start_init(void *dummy) 659 { 660 vm_offset_t addr; 661 struct execve_args args; 662 int options, error; 663 char *var, *path, *next, *s; 664 char *ucp, **uap, *arg0, *arg1; 665 struct thread *td; 666 struct proc *p; 667 668 mtx_lock(&Giant); 669 670 GIANT_REQUIRED; 671 672 td = curthread; 673 p = td->td_proc; 674 675 vfs_mountroot(); 676 677 /* 678 * Need just enough stack to hold the faked-up "execve()" arguments. 679 */ 680 addr = p->p_sysent->sv_usrstack - PAGE_SIZE; 681 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 682 FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0) 683 panic("init: couldn't allocate argument space"); 684 p->p_vmspace->vm_maxsaddr = (caddr_t)addr; 685 p->p_vmspace->vm_ssize = 1; 686 687 if ((var = getenv("init_path")) != NULL) { 688 strlcpy(init_path, var, sizeof(init_path)); 689 freeenv(var); 690 } 691 692 for (path = init_path; *path != '\0'; path = next) { 693 while (*path == ':') 694 path++; 695 if (*path == '\0') 696 break; 697 for (next = path; *next != '\0' && *next != ':'; next++) 698 /* nothing */ ; 699 if (bootverbose) 700 printf("start_init: trying %.*s\n", (int)(next - path), 701 path); 702 703 /* 704 * Move out the boot flag argument. 705 */ 706 options = 0; 707 ucp = (char *)p->p_sysent->sv_usrstack; 708 (void)subyte(--ucp, 0); /* trailing zero */ 709 if (boothowto & RB_SINGLE) { 710 (void)subyte(--ucp, 's'); 711 options = 1; 712 } 713 #ifdef notyet 714 if (boothowto & RB_FASTBOOT) { 715 (void)subyte(--ucp, 'f'); 716 options = 1; 717 } 718 #endif 719 720 #ifdef BOOTCDROM 721 (void)subyte(--ucp, 'C'); 722 options = 1; 723 #endif 724 725 if (options == 0) 726 (void)subyte(--ucp, '-'); 727 (void)subyte(--ucp, '-'); /* leading hyphen */ 728 arg1 = ucp; 729 730 /* 731 * Move out the file name (also arg 0). 732 */ 733 (void)subyte(--ucp, 0); 734 for (s = next - 1; s >= path; s--) 735 (void)subyte(--ucp, *s); 736 arg0 = ucp; 737 738 /* 739 * Move out the arg pointers. 740 */ 741 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1)); 742 (void)suword((caddr_t)--uap, (long)0); /* terminator */ 743 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1); 744 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0); 745 746 /* 747 * Point at the arguments. 748 */ 749 args.fname = arg0; 750 args.argv = uap; 751 args.envv = NULL; 752 753 /* 754 * Now try to exec the program. If can't for any reason 755 * other than it doesn't exist, complain. 756 * 757 * Otherwise, return via fork_trampoline() all the way 758 * to user mode as init! 759 */ 760 if ((error = execve(td, &args)) == 0) { 761 mtx_unlock(&Giant); 762 return; 763 } 764 if (error != ENOENT) 765 printf("exec %.*s: error %d\n", (int)(next - path), 766 path, error); 767 } 768 printf("init: not found in path %s\n", init_path); 769 panic("no init"); 770 } 771 772 /* 773 * Like kproc_create(), but runs in it's own address space. 774 * We do this early to reserve pid 1. 775 * 776 * Note special case - do not make it runnable yet. Other work 777 * in progress will change this more. 778 */ 779 static void 780 create_init(const void *udata __unused) 781 { 782 struct ucred *newcred, *oldcred; 783 int error; 784 785 error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc); 786 if (error) 787 panic("cannot fork init: %d\n", error); 788 KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1")); 789 /* divorce init's credentials from the kernel's */ 790 newcred = crget(); 791 PROC_LOCK(initproc); 792 initproc->p_flag |= P_SYSTEM | P_INMEM; 793 oldcred = initproc->p_ucred; 794 crcopy(newcred, oldcred); 795 #ifdef MAC 796 mac_cred_create_init(newcred); 797 #endif 798 #ifdef AUDIT 799 audit_cred_proc1(newcred); 800 #endif 801 initproc->p_ucred = newcred; 802 PROC_UNLOCK(initproc); 803 crfree(oldcred); 804 cred_update_thread(FIRST_THREAD_IN_PROC(initproc)); 805 cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL); 806 } 807 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL); 808 809 /* 810 * Make it runnable now. 811 */ 812 static void 813 kick_init(const void *udata __unused) 814 { 815 struct thread *td; 816 817 td = FIRST_THREAD_IN_PROC(initproc); 818 thread_lock(td); 819 TD_SET_CAN_RUN(td); 820 sched_add(td, SRQ_BORING); 821 thread_unlock(td); 822 } 823 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL); 824