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 #include "opt_verbose_sysinit.h" 50 51 #include <sys/param.h> 52 #include <sys/kernel.h> 53 #include <sys/exec.h> 54 #include <sys/file.h> 55 #include <sys/filedesc.h> 56 #include <sys/jail.h> 57 #include <sys/ktr.h> 58 #include <sys/lock.h> 59 #include <sys/loginclass.h> 60 #include <sys/mount.h> 61 #include <sys/mutex.h> 62 #include <sys/syscallsubr.h> 63 #include <sys/sysctl.h> 64 #include <sys/proc.h> 65 #include <sys/racct.h> 66 #include <sys/resourcevar.h> 67 #include <sys/systm.h> 68 #include <sys/signalvar.h> 69 #include <sys/vnode.h> 70 #include <sys/sysent.h> 71 #include <sys/reboot.h> 72 #include <sys/sched.h> 73 #include <sys/sx.h> 74 #include <sys/sysproto.h> 75 #include <sys/vmmeter.h> 76 #include <sys/unistd.h> 77 #include <sys/malloc.h> 78 #include <sys/conf.h> 79 #include <sys/cpuset.h> 80 81 #include <machine/cpu.h> 82 83 #include <security/audit/audit.h> 84 #include <security/mac/mac_framework.h> 85 86 #include <vm/vm.h> 87 #include <vm/vm_param.h> 88 #include <vm/pmap.h> 89 #include <vm/vm_map.h> 90 #include <vm/vm_domain.h> 91 #include <sys/copyright.h> 92 93 #include <ddb/ddb.h> 94 #include <ddb/db_sym.h> 95 96 void mi_startup(void); /* Should be elsewhere */ 97 98 /* Components of the first process -- never freed. */ 99 static struct session session0; 100 static struct pgrp pgrp0; 101 struct proc proc0; 102 struct thread0_storage thread0_st __aligned(32); 103 struct vmspace vmspace0; 104 struct proc *initproc; 105 106 #ifndef BOOTHOWTO 107 #define BOOTHOWTO 0 108 #endif 109 int boothowto = BOOTHOWTO; /* initialized so that it can be patched */ 110 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, 111 "Boot control flags, passed from loader"); 112 113 #ifndef BOOTVERBOSE 114 #define BOOTVERBOSE 0 115 #endif 116 int bootverbose = BOOTVERBOSE; 117 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, 118 "Control the output of verbose kernel messages"); 119 120 #ifdef INVARIANTS 121 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance"); 122 #endif 123 124 /* 125 * This ensures that there is at least one entry so that the sysinit_set 126 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never 127 * executed. 128 */ 129 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL); 130 131 /* 132 * The sysinit table itself. Items are checked off as the are run. 133 * If we want to register new sysinit types, add them to newsysinit. 134 */ 135 SET_DECLARE(sysinit_set, struct sysinit); 136 struct sysinit **sysinit, **sysinit_end; 137 struct sysinit **newsysinit, **newsysinit_end; 138 139 /* 140 * Merge a new sysinit set into the current set, reallocating it if 141 * necessary. This can only be called after malloc is running. 142 */ 143 void 144 sysinit_add(struct sysinit **set, struct sysinit **set_end) 145 { 146 struct sysinit **newset; 147 struct sysinit **sipp; 148 struct sysinit **xipp; 149 int count; 150 151 count = set_end - set; 152 if (newsysinit) 153 count += newsysinit_end - newsysinit; 154 else 155 count += sysinit_end - sysinit; 156 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT); 157 if (newset == NULL) 158 panic("cannot malloc for sysinit"); 159 xipp = newset; 160 if (newsysinit) 161 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 162 *xipp++ = *sipp; 163 else 164 for (sipp = sysinit; sipp < sysinit_end; sipp++) 165 *xipp++ = *sipp; 166 for (sipp = set; sipp < set_end; sipp++) 167 *xipp++ = *sipp; 168 if (newsysinit) 169 free(newsysinit, M_TEMP); 170 newsysinit = newset; 171 newsysinit_end = newset + count; 172 } 173 174 #if defined (DDB) && defined(VERBOSE_SYSINIT) 175 static const char * 176 symbol_name(vm_offset_t va, db_strategy_t strategy) 177 { 178 const char *name; 179 c_db_sym_t sym; 180 db_expr_t offset; 181 182 if (va == 0) 183 return (NULL); 184 sym = db_search_symbol(va, strategy, &offset); 185 if (offset != 0) 186 return (NULL); 187 db_symbol_values(sym, &name, NULL); 188 return (name); 189 } 190 #endif 191 192 /* 193 * System startup; initialize the world, create process 0, mount root 194 * filesystem, and fork to create init and pagedaemon. Most of the 195 * hard work is done in the lower-level initialization routines including 196 * startup(), which does memory initialization and autoconfiguration. 197 * 198 * This allows simple addition of new kernel subsystems that require 199 * boot time initialization. It also allows substitution of subsystem 200 * (for instance, a scheduler, kernel profiler, or VM system) by object 201 * module. Finally, it allows for optional "kernel threads". 202 */ 203 void 204 mi_startup(void) 205 { 206 207 struct sysinit **sipp; /* system initialization*/ 208 struct sysinit **xipp; /* interior loop of sort*/ 209 struct sysinit *save; /* bubble*/ 210 211 #if defined(VERBOSE_SYSINIT) 212 int last; 213 int verbose; 214 #endif 215 216 if (boothowto & RB_VERBOSE) 217 bootverbose++; 218 219 if (sysinit == NULL) { 220 sysinit = SET_BEGIN(sysinit_set); 221 sysinit_end = SET_LIMIT(sysinit_set); 222 } 223 224 restart: 225 /* 226 * Perform a bubble sort of the system initialization objects by 227 * their subsystem (primary key) and order (secondary key). 228 */ 229 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 230 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 231 if ((*sipp)->subsystem < (*xipp)->subsystem || 232 ((*sipp)->subsystem == (*xipp)->subsystem && 233 (*sipp)->order <= (*xipp)->order)) 234 continue; /* skip*/ 235 save = *sipp; 236 *sipp = *xipp; 237 *xipp = save; 238 } 239 } 240 241 #if defined(VERBOSE_SYSINIT) 242 last = SI_SUB_COPYRIGHT; 243 verbose = 0; 244 #if !defined(DDB) 245 printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n"); 246 #endif 247 #endif 248 249 /* 250 * Traverse the (now) ordered list of system initialization tasks. 251 * Perform each task, and continue on to the next task. 252 */ 253 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 254 255 if ((*sipp)->subsystem == SI_SUB_DUMMY) 256 continue; /* skip dummy task(s)*/ 257 258 if ((*sipp)->subsystem == SI_SUB_DONE) 259 continue; 260 261 #if defined(VERBOSE_SYSINIT) 262 if ((*sipp)->subsystem > last) { 263 verbose = 1; 264 last = (*sipp)->subsystem; 265 printf("subsystem %x\n", last); 266 } 267 if (verbose) { 268 #if defined(DDB) 269 const char *func, *data; 270 271 func = symbol_name((vm_offset_t)(*sipp)->func, 272 DB_STGY_PROC); 273 data = symbol_name((vm_offset_t)(*sipp)->udata, 274 DB_STGY_ANY); 275 if (func != NULL && data != NULL) 276 printf(" %s(&%s)... ", func, data); 277 else if (func != NULL) 278 printf(" %s(%p)... ", func, (*sipp)->udata); 279 else 280 #endif 281 printf(" %p(%p)... ", (*sipp)->func, 282 (*sipp)->udata); 283 } 284 #endif 285 286 /* Call function */ 287 (*((*sipp)->func))((*sipp)->udata); 288 289 #if defined(VERBOSE_SYSINIT) 290 if (verbose) 291 printf("done.\n"); 292 #endif 293 294 /* Check off the one we're just done */ 295 (*sipp)->subsystem = SI_SUB_DONE; 296 297 /* Check if we've installed more sysinit items via KLD */ 298 if (newsysinit != NULL) { 299 if (sysinit != SET_BEGIN(sysinit_set)) 300 free(sysinit, M_TEMP); 301 sysinit = newsysinit; 302 sysinit_end = newsysinit_end; 303 newsysinit = NULL; 304 newsysinit_end = NULL; 305 goto restart; 306 } 307 } 308 309 mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED); 310 mtx_unlock(&Giant); 311 312 /* 313 * Now hand over this thread to swapper. 314 */ 315 swapper(); 316 /* NOTREACHED*/ 317 } 318 319 static void 320 print_caddr_t(void *data) 321 { 322 printf("%s", (char *)data); 323 } 324 325 static void 326 print_version(void *data __unused) 327 { 328 int len; 329 330 /* Strip a trailing newline from version. */ 331 len = strlen(version); 332 while (len > 0 && version[len - 1] == '\n') 333 len--; 334 printf("%.*s %s\n", len, version, machine); 335 printf("%s\n", compiler_version); 336 } 337 338 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, 339 copyright); 340 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, 341 trademark); 342 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL); 343 344 #ifdef WITNESS 345 static char wit_warn[] = 346 "WARNING: WITNESS option enabled, expect reduced performance.\n"; 347 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1, 348 print_caddr_t, wit_warn); 349 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 1, 350 print_caddr_t, wit_warn); 351 #endif 352 353 #ifdef DIAGNOSTIC 354 static char diag_warn[] = 355 "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n"; 356 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2, 357 print_caddr_t, diag_warn); 358 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 2, 359 print_caddr_t, diag_warn); 360 #endif 361 362 static int 363 null_fetch_syscall_args(struct thread *td __unused) 364 { 365 366 panic("null_fetch_syscall_args"); 367 } 368 369 static void 370 null_set_syscall_retval(struct thread *td __unused, int error __unused) 371 { 372 373 panic("null_set_syscall_retval"); 374 } 375 376 struct sysentvec null_sysvec = { 377 .sv_size = 0, 378 .sv_table = NULL, 379 .sv_mask = 0, 380 .sv_errsize = 0, 381 .sv_errtbl = NULL, 382 .sv_transtrap = NULL, 383 .sv_fixup = NULL, 384 .sv_sendsig = NULL, 385 .sv_sigcode = NULL, 386 .sv_szsigcode = NULL, 387 .sv_name = "null", 388 .sv_coredump = NULL, 389 .sv_imgact_try = NULL, 390 .sv_minsigstksz = 0, 391 .sv_pagesize = PAGE_SIZE, 392 .sv_minuser = VM_MIN_ADDRESS, 393 .sv_maxuser = VM_MAXUSER_ADDRESS, 394 .sv_usrstack = USRSTACK, 395 .sv_psstrings = PS_STRINGS, 396 .sv_stackprot = VM_PROT_ALL, 397 .sv_copyout_strings = NULL, 398 .sv_setregs = NULL, 399 .sv_fixlimit = NULL, 400 .sv_maxssiz = NULL, 401 .sv_flags = 0, 402 .sv_set_syscall_retval = null_set_syscall_retval, 403 .sv_fetch_syscall_args = null_fetch_syscall_args, 404 .sv_syscallnames = NULL, 405 .sv_schedtail = NULL, 406 .sv_thread_detach = NULL, 407 .sv_trap = NULL, 408 }; 409 410 /* 411 * The two following SYSINIT's are proc0 specific glue code. I am not 412 * convinced that they can not be safely combined, but their order of 413 * operation has been maintained as the same as the original init_main.c 414 * for right now. 415 */ 416 /* ARGSUSED*/ 417 static void 418 proc0_init(void *dummy __unused) 419 { 420 struct proc *p; 421 struct thread *td; 422 struct ucred *newcred; 423 vm_paddr_t pageablemem; 424 int i; 425 426 GIANT_REQUIRED; 427 p = &proc0; 428 td = &thread0; 429 430 /* 431 * Initialize magic number and osrel. 432 */ 433 p->p_magic = P_MAGIC; 434 p->p_osrel = osreldate; 435 436 /* 437 * Initialize thread and process structures. 438 */ 439 procinit(); /* set up proc zone */ 440 threadinit(); /* set up UMA zones */ 441 442 /* 443 * Initialise scheduler resources. 444 * Add scheduler specific parts to proc, thread as needed. 445 */ 446 schedinit(); /* scheduler gets its house in order */ 447 448 /* 449 * Create process 0 (the swapper). 450 */ 451 LIST_INSERT_HEAD(&allproc, p, p_list); 452 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash); 453 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 454 p->p_pgrp = &pgrp0; 455 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 456 LIST_INIT(&pgrp0.pg_members); 457 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 458 459 pgrp0.pg_session = &session0; 460 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF); 461 refcount_init(&session0.s_count, 1); 462 session0.s_leader = p; 463 464 p->p_sysent = &null_sysvec; 465 p->p_flag = P_SYSTEM | P_INMEM | P_KPROC; 466 p->p_flag2 = 0; 467 p->p_state = PRS_NORMAL; 468 p->p_klist = knlist_alloc(&p->p_mtx); 469 STAILQ_INIT(&p->p_ktr); 470 p->p_nice = NZERO; 471 /* pid_max cannot be greater than PID_MAX */ 472 td->td_tid = PID_MAX + 1; 473 LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash); 474 td->td_state = TDS_RUNNING; 475 td->td_pri_class = PRI_TIMESHARE; 476 td->td_user_pri = PUSER; 477 td->td_base_user_pri = PUSER; 478 td->td_lend_user_pri = PRI_MAX; 479 td->td_priority = PVM; 480 td->td_base_pri = PVM; 481 td->td_oncpu = curcpu; 482 td->td_flags = TDF_INMEM; 483 td->td_pflags = TDP_KTHREAD; 484 td->td_cpuset = cpuset_thread0(); 485 vm_domain_policy_init(&td->td_vm_dom_policy); 486 vm_domain_policy_set(&td->td_vm_dom_policy, VM_POLICY_NONE, -1); 487 vm_domain_policy_init(&p->p_vm_dom_policy); 488 vm_domain_policy_set(&p->p_vm_dom_policy, VM_POLICY_NONE, -1); 489 prison0_init(); 490 p->p_peers = 0; 491 p->p_leader = p; 492 p->p_reaper = p; 493 LIST_INIT(&p->p_reaplist); 494 495 strncpy(p->p_comm, "kernel", sizeof (p->p_comm)); 496 strncpy(td->td_name, "swapper", sizeof (td->td_name)); 497 498 callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0); 499 callout_init_mtx(&p->p_limco, &p->p_mtx, 0); 500 callout_init(&td->td_slpcallout, 1); 501 502 /* Create credentials. */ 503 newcred = crget(); 504 newcred->cr_ngroups = 1; /* group 0 */ 505 newcred->cr_uidinfo = uifind(0); 506 newcred->cr_ruidinfo = uifind(0); 507 newcred->cr_prison = &prison0; 508 newcred->cr_loginclass = loginclass_find("default"); 509 proc_set_cred_init(p, newcred); 510 #ifdef AUDIT 511 audit_cred_kproc0(newcred); 512 #endif 513 #ifdef MAC 514 mac_cred_create_swapper(newcred); 515 #endif 516 /* Create sigacts. */ 517 p->p_sigacts = sigacts_alloc(); 518 519 /* Initialize signal state for process 0. */ 520 siginit(&proc0); 521 522 /* Create the file descriptor table. */ 523 p->p_fd = fdinit(NULL, false); 524 p->p_fdtol = NULL; 525 526 /* Create the limits structures. */ 527 p->p_limit = lim_alloc(); 528 for (i = 0; i < RLIM_NLIMITS; i++) 529 p->p_limit->pl_rlimit[i].rlim_cur = 530 p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY; 531 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur = 532 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 533 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur = 534 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 535 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz; 536 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz; 537 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz; 538 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz; 539 /* Cast to avoid overflow on i386/PAE. */ 540 pageablemem = ptoa((vm_paddr_t)vm_cnt.v_free_count); 541 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur = 542 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem; 543 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3; 544 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem; 545 p->p_cpulimit = RLIM_INFINITY; 546 547 PROC_LOCK(p); 548 thread_cow_get_proc(td, p); 549 PROC_UNLOCK(p); 550 551 /* Initialize resource accounting structures. */ 552 racct_create(&p->p_racct); 553 554 p->p_stats = pstats_alloc(); 555 556 /* Allocate a prototype map so we have something to fork. */ 557 p->p_vmspace = &vmspace0; 558 vmspace0.vm_refcnt = 1; 559 pmap_pinit0(vmspace_pmap(&vmspace0)); 560 561 /* 562 * proc0 is not expected to enter usermode, so there is no special 563 * handling for sv_minuser here, like is done for exec_new_vmspace(). 564 */ 565 vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0), 566 p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser); 567 568 /* 569 * Call the init and ctor for the new thread and proc. We wait 570 * to do this until all other structures are fairly sane. 571 */ 572 EVENTHANDLER_INVOKE(process_init, p); 573 EVENTHANDLER_INVOKE(thread_init, td); 574 EVENTHANDLER_INVOKE(process_ctor, p); 575 EVENTHANDLER_INVOKE(thread_ctor, td); 576 577 /* 578 * Charge root for one process. 579 */ 580 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0); 581 PROC_LOCK(p); 582 racct_add_force(p, RACCT_NPROC, 1); 583 PROC_UNLOCK(p); 584 } 585 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL); 586 587 /* ARGSUSED*/ 588 static void 589 proc0_post(void *dummy __unused) 590 { 591 struct timespec ts; 592 struct proc *p; 593 struct rusage ru; 594 struct thread *td; 595 596 /* 597 * Now we can look at the time, having had a chance to verify the 598 * time from the filesystem. Pretend that proc0 started now. 599 */ 600 sx_slock(&allproc_lock); 601 FOREACH_PROC_IN_SYSTEM(p) { 602 microuptime(&p->p_stats->p_start); 603 PROC_STATLOCK(p); 604 rufetch(p, &ru); /* Clears thread stats */ 605 PROC_STATUNLOCK(p); 606 p->p_rux.rux_runtime = 0; 607 p->p_rux.rux_uticks = 0; 608 p->p_rux.rux_sticks = 0; 609 p->p_rux.rux_iticks = 0; 610 FOREACH_THREAD_IN_PROC(p, td) { 611 td->td_runtime = 0; 612 } 613 } 614 sx_sunlock(&allproc_lock); 615 PCPU_SET(switchtime, cpu_ticks()); 616 PCPU_SET(switchticks, ticks); 617 618 /* 619 * Give the ``random'' number generator a thump. 620 */ 621 nanotime(&ts); 622 srandom(ts.tv_sec ^ ts.tv_nsec); 623 } 624 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL); 625 626 static void 627 random_init(void *dummy __unused) 628 { 629 630 /* 631 * After CPU has been started we have some randomness on most 632 * platforms via get_cyclecount(). For platforms that don't 633 * we will reseed random(9) in proc0_post() as well. 634 */ 635 srandom(get_cyclecount()); 636 } 637 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL); 638 639 /* 640 *************************************************************************** 641 **** 642 **** The following SYSINIT's and glue code should be moved to the 643 **** respective files on a per subsystem basis. 644 **** 645 *************************************************************************** 646 */ 647 648 /* 649 * List of paths to try when searching for "init". 650 */ 651 static char init_path[MAXPATHLEN] = 652 #ifdef INIT_PATH 653 __XSTRING(INIT_PATH); 654 #else 655 "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init"; 656 #endif 657 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, 658 "Path used to search the init process"); 659 660 /* 661 * Shutdown timeout of init(8). 662 * Unused within kernel, but used to control init(8), hence do not remove. 663 */ 664 #ifndef INIT_SHUTDOWN_TIMEOUT 665 #define INIT_SHUTDOWN_TIMEOUT 120 666 #endif 667 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT; 668 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout, 669 CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). " 670 "Unused within kernel, but used to control init(8)"); 671 672 /* 673 * Start the initial user process; try exec'ing each pathname in init_path. 674 * The program is invoked with one argument containing the boot flags. 675 */ 676 static void 677 start_init(void *dummy) 678 { 679 vm_offset_t addr; 680 struct execve_args args; 681 int options, error; 682 char *var, *path, *next, *s; 683 char *ucp, **uap, *arg0, *arg1; 684 struct thread *td; 685 struct proc *p; 686 687 mtx_lock(&Giant); 688 689 GIANT_REQUIRED; 690 691 td = curthread; 692 p = td->td_proc; 693 694 vfs_mountroot(); 695 696 /* Wipe GELI passphrase from the environment. */ 697 kern_unsetenv("kern.geom.eli.passphrase"); 698 699 /* 700 * Need just enough stack to hold the faked-up "execve()" arguments. 701 */ 702 addr = p->p_sysent->sv_usrstack - PAGE_SIZE; 703 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 0, 704 VMFS_NO_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0) 705 panic("init: couldn't allocate argument space"); 706 p->p_vmspace->vm_maxsaddr = (caddr_t)addr; 707 p->p_vmspace->vm_ssize = 1; 708 709 if ((var = kern_getenv("init_path")) != NULL) { 710 strlcpy(init_path, var, sizeof(init_path)); 711 freeenv(var); 712 } 713 714 for (path = init_path; *path != '\0'; path = next) { 715 while (*path == ':') 716 path++; 717 if (*path == '\0') 718 break; 719 for (next = path; *next != '\0' && *next != ':'; next++) 720 /* nothing */ ; 721 if (bootverbose) 722 printf("start_init: trying %.*s\n", (int)(next - path), 723 path); 724 725 /* 726 * Move out the boot flag argument. 727 */ 728 options = 0; 729 ucp = (char *)p->p_sysent->sv_usrstack; 730 (void)subyte(--ucp, 0); /* trailing zero */ 731 if (boothowto & RB_SINGLE) { 732 (void)subyte(--ucp, 's'); 733 options = 1; 734 } 735 #ifdef notyet 736 if (boothowto & RB_FASTBOOT) { 737 (void)subyte(--ucp, 'f'); 738 options = 1; 739 } 740 #endif 741 742 #ifdef BOOTCDROM 743 (void)subyte(--ucp, 'C'); 744 options = 1; 745 #endif 746 747 if (options == 0) 748 (void)subyte(--ucp, '-'); 749 (void)subyte(--ucp, '-'); /* leading hyphen */ 750 arg1 = ucp; 751 752 /* 753 * Move out the file name (also arg 0). 754 */ 755 (void)subyte(--ucp, 0); 756 for (s = next - 1; s >= path; s--) 757 (void)subyte(--ucp, *s); 758 arg0 = ucp; 759 760 /* 761 * Move out the arg pointers. 762 */ 763 uap = (char **)rounddown2((intptr_t)ucp, sizeof(intptr_t)); 764 (void)suword((caddr_t)--uap, (long)0); /* terminator */ 765 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1); 766 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0); 767 768 /* 769 * Point at the arguments. 770 */ 771 args.fname = arg0; 772 args.argv = uap; 773 args.envv = NULL; 774 775 /* 776 * Now try to exec the program. If can't for any reason 777 * other than it doesn't exist, complain. 778 * 779 * Otherwise, return via fork_trampoline() all the way 780 * to user mode as init! 781 */ 782 if ((error = sys_execve(td, &args)) == 0) { 783 mtx_unlock(&Giant); 784 return; 785 } 786 if (error != ENOENT) 787 printf("exec %.*s: error %d\n", (int)(next - path), 788 path, error); 789 } 790 printf("init: not found in path %s\n", init_path); 791 panic("no init"); 792 } 793 794 /* 795 * Like kproc_create(), but runs in its own address space. 796 * We do this early to reserve pid 1. 797 * 798 * Note special case - do not make it runnable yet. Other work 799 * in progress will change this more. 800 */ 801 static void 802 create_init(const void *udata __unused) 803 { 804 struct fork_req fr; 805 struct ucred *newcred, *oldcred; 806 struct thread *td; 807 int error; 808 809 bzero(&fr, sizeof(fr)); 810 fr.fr_flags = RFFDG | RFPROC | RFSTOPPED; 811 fr.fr_procp = &initproc; 812 error = fork1(&thread0, &fr); 813 if (error) 814 panic("cannot fork init: %d\n", error); 815 KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1")); 816 /* divorce init's credentials from the kernel's */ 817 newcred = crget(); 818 sx_xlock(&proctree_lock); 819 PROC_LOCK(initproc); 820 initproc->p_flag |= P_SYSTEM | P_INMEM; 821 initproc->p_treeflag |= P_TREE_REAPER; 822 LIST_INSERT_HEAD(&initproc->p_reaplist, &proc0, p_reapsibling); 823 oldcred = initproc->p_ucred; 824 crcopy(newcred, oldcred); 825 #ifdef MAC 826 mac_cred_create_init(newcred); 827 #endif 828 #ifdef AUDIT 829 audit_cred_proc1(newcred); 830 #endif 831 proc_set_cred(initproc, newcred); 832 td = FIRST_THREAD_IN_PROC(initproc); 833 crfree(td->td_ucred); 834 td->td_ucred = crhold(initproc->p_ucred); 835 PROC_UNLOCK(initproc); 836 sx_xunlock(&proctree_lock); 837 crfree(oldcred); 838 cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc), 839 start_init, NULL); 840 } 841 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL); 842 843 /* 844 * Make it runnable now. 845 */ 846 static void 847 kick_init(const void *udata __unused) 848 { 849 struct thread *td; 850 851 td = FIRST_THREAD_IN_PROC(initproc); 852 thread_lock(td); 853 TD_SET_CAN_RUN(td); 854 sched_add(td, SRQ_BORING); 855 thread_unlock(td); 856 } 857 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL); 858