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/loginclass.h> 59 #include <sys/mount.h> 60 #include <sys/mutex.h> 61 #include <sys/syscallsubr.h> 62 #include <sys/sysctl.h> 63 #include <sys/proc.h> 64 #include <sys/racct.h> 65 #include <sys/resourcevar.h> 66 #include <sys/systm.h> 67 #include <sys/signalvar.h> 68 #include <sys/vnode.h> 69 #include <sys/sysent.h> 70 #include <sys/reboot.h> 71 #include <sys/sched.h> 72 #include <sys/sx.h> 73 #include <sys/sysproto.h> 74 #include <sys/vmmeter.h> 75 #include <sys/unistd.h> 76 #include <sys/malloc.h> 77 #include <sys/conf.h> 78 #include <sys/cpuset.h> 79 80 #include <machine/cpu.h> 81 82 #include <security/audit/audit.h> 83 #include <security/mac/mac_framework.h> 84 85 #include <vm/vm.h> 86 #include <vm/vm_param.h> 87 #include <vm/pmap.h> 88 #include <vm/vm_map.h> 89 #include <sys/copyright.h> 90 91 #include <ddb/ddb.h> 92 #include <ddb/db_sym.h> 93 94 void mi_startup(void); /* Should be elsewhere */ 95 96 /* Components of the first process -- never freed. */ 97 static struct session session0; 98 static struct pgrp pgrp0; 99 struct proc proc0; 100 struct thread thread0 __aligned(16); 101 struct vmspace vmspace0; 102 struct proc *initproc; 103 104 #ifndef BOOTHOWTO 105 #define BOOTHOWTO 0 106 #endif 107 int boothowto = BOOTHOWTO; /* initialized so that it can be patched */ 108 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, 109 "Boot control flags, passed from loader"); 110 111 #ifndef BOOTVERBOSE 112 #define BOOTVERBOSE 0 113 #endif 114 int bootverbose = BOOTVERBOSE; 115 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, 116 "Control the output of verbose kernel messages"); 117 118 /* 119 * This ensures that there is at least one entry so that the sysinit_set 120 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never 121 * executed. 122 */ 123 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL); 124 125 /* 126 * The sysinit table itself. Items are checked off as the are run. 127 * If we want to register new sysinit types, add them to newsysinit. 128 */ 129 SET_DECLARE(sysinit_set, struct sysinit); 130 struct sysinit **sysinit, **sysinit_end; 131 struct sysinit **newsysinit, **newsysinit_end; 132 133 /* 134 * Merge a new sysinit set into the current set, reallocating it if 135 * necessary. This can only be called after malloc is running. 136 */ 137 void 138 sysinit_add(struct sysinit **set, struct sysinit **set_end) 139 { 140 struct sysinit **newset; 141 struct sysinit **sipp; 142 struct sysinit **xipp; 143 int count; 144 145 count = set_end - set; 146 if (newsysinit) 147 count += newsysinit_end - newsysinit; 148 else 149 count += sysinit_end - sysinit; 150 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT); 151 if (newset == NULL) 152 panic("cannot malloc for sysinit"); 153 xipp = newset; 154 if (newsysinit) 155 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 156 *xipp++ = *sipp; 157 else 158 for (sipp = sysinit; sipp < sysinit_end; sipp++) 159 *xipp++ = *sipp; 160 for (sipp = set; sipp < set_end; sipp++) 161 *xipp++ = *sipp; 162 if (newsysinit) 163 free(newsysinit, M_TEMP); 164 newsysinit = newset; 165 newsysinit_end = newset + count; 166 } 167 168 #if defined (DDB) && defined(VERBOSE_SYSINIT) 169 static const char * 170 symbol_name(vm_offset_t va, db_strategy_t strategy) 171 { 172 const char *name; 173 c_db_sym_t sym; 174 db_expr_t offset; 175 176 if (va == 0) 177 return (NULL); 178 sym = db_search_symbol(va, strategy, &offset); 179 if (offset != 0) 180 return (NULL); 181 db_symbol_values(sym, &name, NULL); 182 return (name); 183 } 184 #endif 185 186 /* 187 * System startup; initialize the world, create process 0, mount root 188 * filesystem, and fork to create init and pagedaemon. Most of the 189 * hard work is done in the lower-level initialization routines including 190 * startup(), which does memory initialization and autoconfiguration. 191 * 192 * This allows simple addition of new kernel subsystems that require 193 * boot time initialization. It also allows substitution of subsystem 194 * (for instance, a scheduler, kernel profiler, or VM system) by object 195 * module. Finally, it allows for optional "kernel threads". 196 */ 197 void 198 mi_startup(void) 199 { 200 201 register struct sysinit **sipp; /* system initialization*/ 202 register struct sysinit **xipp; /* interior loop of sort*/ 203 register struct sysinit *save; /* bubble*/ 204 205 #if defined(VERBOSE_SYSINIT) 206 int last; 207 int verbose; 208 #endif 209 210 if (boothowto & RB_VERBOSE) 211 bootverbose++; 212 213 if (sysinit == NULL) { 214 sysinit = SET_BEGIN(sysinit_set); 215 sysinit_end = SET_LIMIT(sysinit_set); 216 } 217 218 restart: 219 /* 220 * Perform a bubble sort of the system initialization objects by 221 * their subsystem (primary key) and order (secondary key). 222 */ 223 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 224 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 225 if ((*sipp)->subsystem < (*xipp)->subsystem || 226 ((*sipp)->subsystem == (*xipp)->subsystem && 227 (*sipp)->order <= (*xipp)->order)) 228 continue; /* skip*/ 229 save = *sipp; 230 *sipp = *xipp; 231 *xipp = save; 232 } 233 } 234 235 #if defined(VERBOSE_SYSINIT) 236 last = SI_SUB_COPYRIGHT; 237 verbose = 0; 238 #if !defined(DDB) 239 printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n"); 240 #endif 241 #endif 242 243 /* 244 * Traverse the (now) ordered list of system initialization tasks. 245 * Perform each task, and continue on to the next task. 246 */ 247 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 248 249 if ((*sipp)->subsystem == SI_SUB_DUMMY) 250 continue; /* skip dummy task(s)*/ 251 252 if ((*sipp)->subsystem == SI_SUB_DONE) 253 continue; 254 255 #if defined(VERBOSE_SYSINIT) 256 if ((*sipp)->subsystem > last) { 257 verbose = 1; 258 last = (*sipp)->subsystem; 259 printf("subsystem %x\n", last); 260 } 261 if (verbose) { 262 #if defined(DDB) 263 const char *func, *data; 264 265 func = symbol_name((vm_offset_t)(*sipp)->func, 266 DB_STGY_PROC); 267 data = symbol_name((vm_offset_t)(*sipp)->udata, 268 DB_STGY_ANY); 269 if (func != NULL && data != NULL) 270 printf(" %s(&%s)... ", func, data); 271 else if (func != NULL) 272 printf(" %s(%p)... ", func, (*sipp)->udata); 273 else 274 #endif 275 printf(" %p(%p)... ", (*sipp)->func, 276 (*sipp)->udata); 277 } 278 #endif 279 280 /* Call function */ 281 (*((*sipp)->func))((*sipp)->udata); 282 283 #if defined(VERBOSE_SYSINIT) 284 if (verbose) 285 printf("done.\n"); 286 #endif 287 288 /* Check off the one we're just done */ 289 (*sipp)->subsystem = SI_SUB_DONE; 290 291 /* Check if we've installed more sysinit items via KLD */ 292 if (newsysinit != NULL) { 293 if (sysinit != SET_BEGIN(sysinit_set)) 294 free(sysinit, M_TEMP); 295 sysinit = newsysinit; 296 sysinit_end = newsysinit_end; 297 newsysinit = NULL; 298 newsysinit_end = NULL; 299 goto restart; 300 } 301 } 302 303 mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED); 304 mtx_unlock(&Giant); 305 306 /* 307 * Now hand over this thread to swapper. 308 */ 309 swapper(); 310 /* NOTREACHED*/ 311 } 312 313 314 /* 315 *************************************************************************** 316 **** 317 **** The following SYSINIT's belong elsewhere, but have not yet 318 **** been moved. 319 **** 320 *************************************************************************** 321 */ 322 static void 323 print_caddr_t(void *data) 324 { 325 printf("%s", (char *)data); 326 } 327 328 static void 329 print_version(void *data __unused) 330 { 331 int len; 332 333 /* Strip a trailing newline from version. */ 334 len = strlen(version); 335 while (len > 0 && version[len - 1] == '\n') 336 len--; 337 printf("%.*s %s\n", len, version, machine); 338 printf("%s\n", compiler_version); 339 } 340 341 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, 342 copyright); 343 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, 344 trademark); 345 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL); 346 347 #ifdef WITNESS 348 static char wit_warn[] = 349 "WARNING: WITNESS option enabled, expect reduced performance.\n"; 350 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1, 351 print_caddr_t, wit_warn); 352 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 1, 353 print_caddr_t, wit_warn); 354 #endif 355 356 #ifdef DIAGNOSTIC 357 static char diag_warn[] = 358 "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n"; 359 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2, 360 print_caddr_t, diag_warn); 361 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 2, 362 print_caddr_t, diag_warn); 363 #endif 364 365 static int 366 null_fetch_syscall_args(struct thread *td __unused, 367 struct syscall_args *sa __unused) 368 { 369 370 panic("null_fetch_syscall_args"); 371 } 372 373 static void 374 null_set_syscall_retval(struct thread *td __unused, int error __unused) 375 { 376 377 panic("null_set_syscall_retval"); 378 } 379 380 struct sysentvec null_sysvec = { 381 .sv_size = 0, 382 .sv_table = NULL, 383 .sv_mask = 0, 384 .sv_sigsize = 0, 385 .sv_sigtbl = NULL, 386 .sv_errsize = 0, 387 .sv_errtbl = NULL, 388 .sv_transtrap = NULL, 389 .sv_fixup = NULL, 390 .sv_sendsig = NULL, 391 .sv_sigcode = NULL, 392 .sv_szsigcode = NULL, 393 .sv_prepsyscall = NULL, 394 .sv_name = "null", 395 .sv_coredump = NULL, 396 .sv_imgact_try = NULL, 397 .sv_minsigstksz = 0, 398 .sv_pagesize = PAGE_SIZE, 399 .sv_minuser = VM_MIN_ADDRESS, 400 .sv_maxuser = VM_MAXUSER_ADDRESS, 401 .sv_usrstack = USRSTACK, 402 .sv_psstrings = PS_STRINGS, 403 .sv_stackprot = VM_PROT_ALL, 404 .sv_copyout_strings = NULL, 405 .sv_setregs = NULL, 406 .sv_fixlimit = NULL, 407 .sv_maxssiz = NULL, 408 .sv_flags = 0, 409 .sv_set_syscall_retval = null_set_syscall_retval, 410 .sv_fetch_syscall_args = null_fetch_syscall_args, 411 .sv_syscallnames = NULL, 412 .sv_schedtail = NULL, 413 }; 414 415 /* 416 *************************************************************************** 417 **** 418 **** The two following SYSINIT's are proc0 specific glue code. I am not 419 **** convinced that they can not be safely combined, but their order of 420 **** operation has been maintained as the same as the original init_main.c 421 **** for right now. 422 **** 423 **** These probably belong in init_proc.c or kern_proc.c, since they 424 **** deal with proc0 (the fork template process). 425 **** 426 *************************************************************************** 427 */ 428 /* ARGSUSED*/ 429 static void 430 proc0_init(void *dummy __unused) 431 { 432 struct proc *p; 433 struct thread *td; 434 vm_paddr_t pageablemem; 435 int i; 436 437 GIANT_REQUIRED; 438 p = &proc0; 439 td = &thread0; 440 441 /* 442 * Initialize magic number and osrel. 443 */ 444 p->p_magic = P_MAGIC; 445 p->p_osrel = osreldate; 446 447 /* 448 * Initialize thread and process structures. 449 */ 450 procinit(); /* set up proc zone */ 451 threadinit(); /* set up UMA zones */ 452 453 /* 454 * Initialise scheduler resources. 455 * Add scheduler specific parts to proc, thread as needed. 456 */ 457 schedinit(); /* scheduler gets its house in order */ 458 /* 459 * Initialize sleep queue hash table 460 */ 461 sleepinit(); 462 463 /* 464 * additional VM structures 465 */ 466 vm_init2(); 467 468 /* 469 * Create process 0 (the swapper). 470 */ 471 LIST_INSERT_HEAD(&allproc, p, p_list); 472 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash); 473 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 474 p->p_pgrp = &pgrp0; 475 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 476 LIST_INIT(&pgrp0.pg_members); 477 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 478 479 pgrp0.pg_session = &session0; 480 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF); 481 refcount_init(&session0.s_count, 1); 482 session0.s_leader = p; 483 484 p->p_sysent = &null_sysvec; 485 p->p_flag = P_SYSTEM | P_INMEM; 486 p->p_state = PRS_NORMAL; 487 knlist_init_mtx(&p->p_klist, &p->p_mtx); 488 STAILQ_INIT(&p->p_ktr); 489 p->p_nice = NZERO; 490 /* pid_max cannot be greater than PID_MAX */ 491 td->td_tid = PID_MAX + 1; 492 LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash); 493 td->td_state = TDS_RUNNING; 494 td->td_pri_class = PRI_TIMESHARE; 495 td->td_user_pri = PUSER; 496 td->td_base_user_pri = PUSER; 497 td->td_lend_user_pri = PRI_MAX; 498 td->td_priority = PVM; 499 td->td_base_pri = PVM; 500 td->td_oncpu = 0; 501 td->td_flags = TDF_INMEM; 502 td->td_pflags = TDP_KTHREAD; 503 td->td_cpuset = cpuset_thread0(); 504 prison0.pr_cpuset = cpuset_ref(td->td_cpuset); 505 p->p_peers = 0; 506 p->p_leader = p; 507 508 509 strncpy(p->p_comm, "kernel", sizeof (p->p_comm)); 510 strncpy(td->td_name, "swapper", sizeof (td->td_name)); 511 512 callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0); 513 callout_init_mtx(&p->p_limco, &p->p_mtx, 0); 514 callout_init(&td->td_slpcallout, CALLOUT_MPSAFE); 515 516 /* Create credentials. */ 517 p->p_ucred = crget(); 518 p->p_ucred->cr_ngroups = 1; /* group 0 */ 519 p->p_ucred->cr_uidinfo = uifind(0); 520 p->p_ucred->cr_ruidinfo = uifind(0); 521 p->p_ucred->cr_prison = &prison0; 522 p->p_ucred->cr_loginclass = loginclass_find("default"); 523 #ifdef AUDIT 524 audit_cred_kproc0(p->p_ucred); 525 #endif 526 #ifdef MAC 527 mac_cred_create_swapper(p->p_ucred); 528 #endif 529 td->td_ucred = crhold(p->p_ucred); 530 531 /* Create sigacts. */ 532 p->p_sigacts = sigacts_alloc(); 533 534 /* Initialize signal state for process 0. */ 535 siginit(&proc0); 536 537 /* Create the file descriptor table. */ 538 p->p_fd = fdinit(NULL); 539 p->p_fdtol = NULL; 540 541 /* Create the limits structures. */ 542 p->p_limit = lim_alloc(); 543 for (i = 0; i < RLIM_NLIMITS; i++) 544 p->p_limit->pl_rlimit[i].rlim_cur = 545 p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY; 546 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur = 547 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 548 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur = 549 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 550 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz; 551 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz; 552 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz; 553 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz; 554 /* Cast to avoid overflow on i386/PAE. */ 555 pageablemem = ptoa((vm_paddr_t)cnt.v_free_count); 556 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur = 557 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem; 558 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3; 559 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem; 560 p->p_cpulimit = RLIM_INFINITY; 561 562 /* Initialize resource accounting structures. */ 563 racct_create(&p->p_racct); 564 565 p->p_stats = pstats_alloc(); 566 567 /* Allocate a prototype map so we have something to fork. */ 568 pmap_pinit0(vmspace_pmap(&vmspace0)); 569 p->p_vmspace = &vmspace0; 570 vmspace0.vm_refcnt = 1; 571 572 /* 573 * proc0 is not expected to enter usermode, so there is no special 574 * handling for sv_minuser here, like is done for exec_new_vmspace(). 575 */ 576 vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0), 577 p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser); 578 579 /* 580 * Call the init and ctor for the new thread and proc. We wait 581 * to do this until all other structures are fairly sane. 582 */ 583 EVENTHANDLER_INVOKE(process_init, p); 584 EVENTHANDLER_INVOKE(thread_init, td); 585 EVENTHANDLER_INVOKE(process_ctor, p); 586 EVENTHANDLER_INVOKE(thread_ctor, td); 587 588 /* 589 * Charge root for one process. 590 */ 591 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0); 592 PROC_LOCK(p); 593 racct_add_force(p, RACCT_NPROC, 1); 594 PROC_UNLOCK(p); 595 } 596 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL); 597 598 /* ARGSUSED*/ 599 static void 600 proc0_post(void *dummy __unused) 601 { 602 struct timespec ts; 603 struct proc *p; 604 struct rusage ru; 605 struct thread *td; 606 607 /* 608 * Now we can look at the time, having had a chance to verify the 609 * time from the filesystem. Pretend that proc0 started now. 610 */ 611 sx_slock(&allproc_lock); 612 FOREACH_PROC_IN_SYSTEM(p) { 613 microuptime(&p->p_stats->p_start); 614 PROC_SLOCK(p); 615 rufetch(p, &ru); /* Clears thread stats */ 616 PROC_SUNLOCK(p); 617 p->p_rux.rux_runtime = 0; 618 p->p_rux.rux_uticks = 0; 619 p->p_rux.rux_sticks = 0; 620 p->p_rux.rux_iticks = 0; 621 FOREACH_THREAD_IN_PROC(p, td) { 622 td->td_runtime = 0; 623 } 624 } 625 sx_sunlock(&allproc_lock); 626 PCPU_SET(switchtime, cpu_ticks()); 627 PCPU_SET(switchticks, ticks); 628 629 /* 630 * Give the ``random'' number generator a thump. 631 */ 632 nanotime(&ts); 633 srandom(ts.tv_sec ^ ts.tv_nsec); 634 } 635 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL); 636 637 static void 638 random_init(void *dummy __unused) 639 { 640 641 /* 642 * After CPU has been started we have some randomness on most 643 * platforms via get_cyclecount(). For platforms that don't 644 * we will reseed random(9) in proc0_post() as well. 645 */ 646 srandom(get_cyclecount()); 647 } 648 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL); 649 650 /* 651 *************************************************************************** 652 **** 653 **** The following SYSINIT's and glue code should be moved to the 654 **** respective files on a per subsystem basis. 655 **** 656 *************************************************************************** 657 */ 658 659 660 /* 661 *************************************************************************** 662 **** 663 **** The following code probably belongs in another file, like 664 **** kern/init_init.c. 665 **** 666 *************************************************************************** 667 */ 668 669 /* 670 * List of paths to try when searching for "init". 671 */ 672 static char init_path[MAXPATHLEN] = 673 #ifdef INIT_PATH 674 __XSTRING(INIT_PATH); 675 #else 676 "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init"; 677 #endif 678 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, 679 "Path used to search the init process"); 680 681 /* 682 * Shutdown timeout of init(8). 683 * Unused within kernel, but used to control init(8), hence do not remove. 684 */ 685 #ifndef INIT_SHUTDOWN_TIMEOUT 686 #define INIT_SHUTDOWN_TIMEOUT 120 687 #endif 688 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT; 689 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout, 690 CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). " 691 "Unused within kernel, but used to control init(8)"); 692 693 /* 694 * Start the initial user process; try exec'ing each pathname in init_path. 695 * The program is invoked with one argument containing the boot flags. 696 */ 697 static void 698 start_init(void *dummy) 699 { 700 vm_offset_t addr; 701 struct execve_args args; 702 int options, error; 703 char *var, *path, *next, *s; 704 char *ucp, **uap, *arg0, *arg1; 705 struct thread *td; 706 struct proc *p; 707 708 mtx_lock(&Giant); 709 710 GIANT_REQUIRED; 711 712 td = curthread; 713 p = td->td_proc; 714 715 vfs_mountroot(); 716 717 /* 718 * Need just enough stack to hold the faked-up "execve()" arguments. 719 */ 720 addr = p->p_sysent->sv_usrstack - PAGE_SIZE; 721 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 722 FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0) 723 panic("init: couldn't allocate argument space"); 724 p->p_vmspace->vm_maxsaddr = (caddr_t)addr; 725 p->p_vmspace->vm_ssize = 1; 726 727 if ((var = getenv("init_path")) != NULL) { 728 strlcpy(init_path, var, sizeof(init_path)); 729 freeenv(var); 730 } 731 732 for (path = init_path; *path != '\0'; path = next) { 733 while (*path == ':') 734 path++; 735 if (*path == '\0') 736 break; 737 for (next = path; *next != '\0' && *next != ':'; next++) 738 /* nothing */ ; 739 if (bootverbose) 740 printf("start_init: trying %.*s\n", (int)(next - path), 741 path); 742 743 /* 744 * Move out the boot flag argument. 745 */ 746 options = 0; 747 ucp = (char *)p->p_sysent->sv_usrstack; 748 (void)subyte(--ucp, 0); /* trailing zero */ 749 if (boothowto & RB_SINGLE) { 750 (void)subyte(--ucp, 's'); 751 options = 1; 752 } 753 #ifdef notyet 754 if (boothowto & RB_FASTBOOT) { 755 (void)subyte(--ucp, 'f'); 756 options = 1; 757 } 758 #endif 759 760 #ifdef BOOTCDROM 761 (void)subyte(--ucp, 'C'); 762 options = 1; 763 #endif 764 765 if (options == 0) 766 (void)subyte(--ucp, '-'); 767 (void)subyte(--ucp, '-'); /* leading hyphen */ 768 arg1 = ucp; 769 770 /* 771 * Move out the file name (also arg 0). 772 */ 773 (void)subyte(--ucp, 0); 774 for (s = next - 1; s >= path; s--) 775 (void)subyte(--ucp, *s); 776 arg0 = ucp; 777 778 /* 779 * Move out the arg pointers. 780 */ 781 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1)); 782 (void)suword((caddr_t)--uap, (long)0); /* terminator */ 783 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1); 784 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0); 785 786 /* 787 * Point at the arguments. 788 */ 789 args.fname = arg0; 790 args.argv = uap; 791 args.envv = NULL; 792 793 /* 794 * Now try to exec the program. If can't for any reason 795 * other than it doesn't exist, complain. 796 * 797 * Otherwise, return via fork_trampoline() all the way 798 * to user mode as init! 799 */ 800 if ((error = sys_execve(td, &args)) == 0) { 801 mtx_unlock(&Giant); 802 return; 803 } 804 if (error != ENOENT) 805 printf("exec %.*s: error %d\n", (int)(next - path), 806 path, error); 807 } 808 printf("init: not found in path %s\n", init_path); 809 panic("no init"); 810 } 811 812 /* 813 * Like kproc_create(), but runs in it's own address space. 814 * We do this early to reserve pid 1. 815 * 816 * Note special case - do not make it runnable yet. Other work 817 * in progress will change this more. 818 */ 819 static void 820 create_init(const void *udata __unused) 821 { 822 struct ucred *newcred, *oldcred; 823 int error; 824 825 error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc, 826 NULL, 0); 827 if (error) 828 panic("cannot fork init: %d\n", error); 829 KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1")); 830 /* divorce init's credentials from the kernel's */ 831 newcred = crget(); 832 PROC_LOCK(initproc); 833 initproc->p_flag |= P_SYSTEM | P_INMEM; 834 oldcred = initproc->p_ucred; 835 crcopy(newcred, oldcred); 836 #ifdef MAC 837 mac_cred_create_init(newcred); 838 #endif 839 #ifdef AUDIT 840 audit_cred_proc1(newcred); 841 #endif 842 initproc->p_ucred = newcred; 843 PROC_UNLOCK(initproc); 844 crfree(oldcred); 845 cred_update_thread(FIRST_THREAD_IN_PROC(initproc)); 846 cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL); 847 } 848 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL); 849 850 /* 851 * Make it runnable now. 852 */ 853 static void 854 kick_init(const void *udata __unused) 855 { 856 struct thread *td; 857 858 td = FIRST_THREAD_IN_PROC(initproc); 859 thread_lock(td); 860 TD_SET_CAN_RUN(td); 861 sched_add(td, SRQ_BORING); 862 thread_unlock(td); 863 } 864 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL); 865