init_main.c revision 330897
1/*- 2 * SPDX-License-Identifier: BSD-4-Clause 3 * 4 * Copyright (c) 1995 Terrence R. Lambert 5 * All rights reserved. 6 * 7 * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993 8 * The Regents of the University of California. All rights reserved. 9 * (c) UNIX System Laboratories, Inc. 10 * All or some portions of this file are derived from material licensed 11 * to the University of California by American Telephone and Telegraph 12 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 13 * the permission of UNIX System Laboratories, Inc. 14 * 15 * Redistribution and use in source and binary forms, with or without 16 * modification, are permitted provided that the following conditions 17 * are met: 18 * 1. Redistributions of source code must retain the above copyright 19 * notice, this list of conditions and the following disclaimer. 20 * 2. Redistributions in binary form must reproduce the above copyright 21 * notice, this list of conditions and the following disclaimer in the 22 * documentation and/or other materials provided with the distribution. 23 * 3. All advertising materials mentioning features or use of this software 24 * must display the following acknowledgement: 25 * This product includes software developed by the University of 26 * California, Berkeley and its contributors. 27 * 4. Neither the name of the University nor the names of its contributors 28 * may be used to endorse or promote products derived from this software 29 * without specific prior written permission. 30 * 31 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 32 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 33 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 34 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 35 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 36 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 37 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 38 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 39 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 40 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 41 * SUCH DAMAGE. 42 * 43 * @(#)init_main.c 8.9 (Berkeley) 1/21/94 44 */ 45 46#include <sys/cdefs.h> 47__FBSDID("$FreeBSD: stable/11/sys/kern/init_main.c 330897 2018-03-14 03:19:51Z eadler $"); 48 49#include "opt_ddb.h" 50#include "opt_init_path.h" 51#include "opt_verbose_sysinit.h" 52 53#include <sys/param.h> 54#include <sys/kernel.h> 55#include <sys/exec.h> 56#include <sys/file.h> 57#include <sys/filedesc.h> 58#include <sys/jail.h> 59#include <sys/ktr.h> 60#include <sys/lock.h> 61#include <sys/loginclass.h> 62#include <sys/mount.h> 63#include <sys/mutex.h> 64#include <sys/syscallsubr.h> 65#include <sys/sysctl.h> 66#include <sys/proc.h> 67#include <sys/racct.h> 68#include <sys/resourcevar.h> 69#include <sys/systm.h> 70#include <sys/signalvar.h> 71#include <sys/vnode.h> 72#include <sys/sysent.h> 73#include <sys/reboot.h> 74#include <sys/sched.h> 75#include <sys/sx.h> 76#include <sys/sysproto.h> 77#include <sys/vmmeter.h> 78#include <sys/unistd.h> 79#include <sys/malloc.h> 80#include <sys/conf.h> 81#include <sys/cpuset.h> 82 83#include <machine/cpu.h> 84 85#include <security/audit/audit.h> 86#include <security/mac/mac_framework.h> 87 88#include <vm/vm.h> 89#include <vm/vm_param.h> 90#include <vm/pmap.h> 91#include <vm/vm_map.h> 92#include <vm/vm_domain.h> 93#include <sys/copyright.h> 94 95#include <ddb/ddb.h> 96#include <ddb/db_sym.h> 97 98void mi_startup(void); /* Should be elsewhere */ 99 100/* Components of the first process -- never freed. */ 101static struct session session0; 102static struct pgrp pgrp0; 103struct proc proc0; 104struct thread0_storage thread0_st __aligned(32); 105struct vmspace vmspace0; 106struct proc *initproc; 107 108#ifndef BOOTHOWTO 109#define BOOTHOWTO 0 110#endif 111int boothowto = BOOTHOWTO; /* initialized so that it can be patched */ 112SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, 113 "Boot control flags, passed from loader"); 114 115#ifndef BOOTVERBOSE 116#define BOOTVERBOSE 0 117#endif 118int bootverbose = BOOTVERBOSE; 119SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, 120 "Control the output of verbose kernel messages"); 121 122#ifdef INVARIANTS 123FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance"); 124#endif 125 126/* 127 * This ensures that there is at least one entry so that the sysinit_set 128 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never 129 * executed. 130 */ 131SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL); 132 133/* 134 * The sysinit table itself. Items are checked off as the are run. 135 * If we want to register new sysinit types, add them to newsysinit. 136 */ 137SET_DECLARE(sysinit_set, struct sysinit); 138struct sysinit **sysinit, **sysinit_end; 139struct sysinit **newsysinit, **newsysinit_end; 140 141/* 142 * Merge a new sysinit set into the current set, reallocating it if 143 * necessary. This can only be called after malloc is running. 144 */ 145void 146sysinit_add(struct sysinit **set, struct sysinit **set_end) 147{ 148 struct sysinit **newset; 149 struct sysinit **sipp; 150 struct sysinit **xipp; 151 int count; 152 153 count = set_end - set; 154 if (newsysinit) 155 count += newsysinit_end - newsysinit; 156 else 157 count += sysinit_end - sysinit; 158 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT); 159 if (newset == NULL) 160 panic("cannot malloc for sysinit"); 161 xipp = newset; 162 if (newsysinit) 163 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 164 *xipp++ = *sipp; 165 else 166 for (sipp = sysinit; sipp < sysinit_end; sipp++) 167 *xipp++ = *sipp; 168 for (sipp = set; sipp < set_end; sipp++) 169 *xipp++ = *sipp; 170 if (newsysinit) 171 free(newsysinit, M_TEMP); 172 newsysinit = newset; 173 newsysinit_end = newset + count; 174} 175 176#if defined (DDB) && defined(VERBOSE_SYSINIT) 177static const char * 178symbol_name(vm_offset_t va, db_strategy_t strategy) 179{ 180 const char *name; 181 c_db_sym_t sym; 182 db_expr_t offset; 183 184 if (va == 0) 185 return (NULL); 186 sym = db_search_symbol(va, strategy, &offset); 187 if (offset != 0) 188 return (NULL); 189 db_symbol_values(sym, &name, NULL); 190 return (name); 191} 192#endif 193 194/* 195 * System startup; initialize the world, create process 0, mount root 196 * filesystem, and fork to create init and pagedaemon. Most of the 197 * hard work is done in the lower-level initialization routines including 198 * startup(), which does memory initialization and autoconfiguration. 199 * 200 * This allows simple addition of new kernel subsystems that require 201 * boot time initialization. It also allows substitution of subsystem 202 * (for instance, a scheduler, kernel profiler, or VM system) by object 203 * module. Finally, it allows for optional "kernel threads". 204 */ 205void 206mi_startup(void) 207{ 208 209 register struct sysinit **sipp; /* system initialization*/ 210 register struct sysinit **xipp; /* interior loop of sort*/ 211 register struct sysinit *save; /* bubble*/ 212 213#if defined(VERBOSE_SYSINIT) 214 int last; 215 int verbose; 216#endif 217 218 if (boothowto & RB_VERBOSE) 219 bootverbose++; 220 221 if (sysinit == NULL) { 222 sysinit = SET_BEGIN(sysinit_set); 223 sysinit_end = SET_LIMIT(sysinit_set); 224 } 225 226restart: 227 /* 228 * Perform a bubble sort of the system initialization objects by 229 * their subsystem (primary key) and order (secondary key). 230 */ 231 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 232 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 233 if ((*sipp)->subsystem < (*xipp)->subsystem || 234 ((*sipp)->subsystem == (*xipp)->subsystem && 235 (*sipp)->order <= (*xipp)->order)) 236 continue; /* skip*/ 237 save = *sipp; 238 *sipp = *xipp; 239 *xipp = save; 240 } 241 } 242 243#if defined(VERBOSE_SYSINIT) 244 last = SI_SUB_COPYRIGHT; 245 verbose = 0; 246#if !defined(DDB) 247 printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n"); 248#endif 249#endif 250 251 /* 252 * Traverse the (now) ordered list of system initialization tasks. 253 * Perform each task, and continue on to the next task. 254 */ 255 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 256 257 if ((*sipp)->subsystem == SI_SUB_DUMMY) 258 continue; /* skip dummy task(s)*/ 259 260 if ((*sipp)->subsystem == SI_SUB_DONE) 261 continue; 262 263#if defined(VERBOSE_SYSINIT) 264 if ((*sipp)->subsystem > last) { 265 verbose = 1; 266 last = (*sipp)->subsystem; 267 printf("subsystem %x\n", last); 268 } 269 if (verbose) { 270#if defined(DDB) 271 const char *func, *data; 272 273 func = symbol_name((vm_offset_t)(*sipp)->func, 274 DB_STGY_PROC); 275 data = symbol_name((vm_offset_t)(*sipp)->udata, 276 DB_STGY_ANY); 277 if (func != NULL && data != NULL) 278 printf(" %s(&%s)... ", func, data); 279 else if (func != NULL) 280 printf(" %s(%p)... ", func, (*sipp)->udata); 281 else 282#endif 283 printf(" %p(%p)... ", (*sipp)->func, 284 (*sipp)->udata); 285 } 286#endif 287 288 /* Call function */ 289 (*((*sipp)->func))((*sipp)->udata); 290 291#if defined(VERBOSE_SYSINIT) 292 if (verbose) 293 printf("done.\n"); 294#endif 295 296 /* Check off the one we're just done */ 297 (*sipp)->subsystem = SI_SUB_DONE; 298 299 /* Check if we've installed more sysinit items via KLD */ 300 if (newsysinit != NULL) { 301 if (sysinit != SET_BEGIN(sysinit_set)) 302 free(sysinit, M_TEMP); 303 sysinit = newsysinit; 304 sysinit_end = newsysinit_end; 305 newsysinit = NULL; 306 newsysinit_end = NULL; 307 goto restart; 308 } 309 } 310 311 mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED); 312 mtx_unlock(&Giant); 313 314 /* 315 * Now hand over this thread to swapper. 316 */ 317 swapper(); 318 /* NOTREACHED*/ 319} 320 321 322/* 323 *************************************************************************** 324 **** 325 **** The following SYSINIT's belong elsewhere, but have not yet 326 **** been moved. 327 **** 328 *************************************************************************** 329 */ 330static void 331print_caddr_t(void *data) 332{ 333 printf("%s", (char *)data); 334} 335 336static void 337print_version(void *data __unused) 338{ 339 int len; 340 341 /* Strip a trailing newline from version. */ 342 len = strlen(version); 343 while (len > 0 && version[len - 1] == '\n') 344 len--; 345 printf("%.*s %s\n", len, version, machine); 346 printf("%s\n", compiler_version); 347} 348 349SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, 350 copyright); 351SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, 352 trademark); 353SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL); 354 355#ifdef WITNESS 356static char wit_warn[] = 357 "WARNING: WITNESS option enabled, expect reduced performance.\n"; 358SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1, 359 print_caddr_t, wit_warn); 360SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 1, 361 print_caddr_t, wit_warn); 362#endif 363 364#ifdef DIAGNOSTIC 365static char diag_warn[] = 366 "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n"; 367SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2, 368 print_caddr_t, diag_warn); 369SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 2, 370 print_caddr_t, diag_warn); 371#endif 372 373static int 374null_fetch_syscall_args(struct thread *td __unused) 375{ 376 377 panic("null_fetch_syscall_args"); 378} 379 380static void 381null_set_syscall_retval(struct thread *td __unused, int error __unused) 382{ 383 384 panic("null_set_syscall_retval"); 385} 386 387struct sysentvec null_sysvec = { 388 .sv_size = 0, 389 .sv_table = NULL, 390 .sv_mask = 0, 391 .sv_errsize = 0, 392 .sv_errtbl = NULL, 393 .sv_transtrap = NULL, 394 .sv_fixup = NULL, 395 .sv_sendsig = NULL, 396 .sv_sigcode = NULL, 397 .sv_szsigcode = NULL, 398 .sv_name = "null", 399 .sv_coredump = NULL, 400 .sv_imgact_try = NULL, 401 .sv_minsigstksz = 0, 402 .sv_pagesize = PAGE_SIZE, 403 .sv_minuser = VM_MIN_ADDRESS, 404 .sv_maxuser = VM_MAXUSER_ADDRESS, 405 .sv_usrstack = USRSTACK, 406 .sv_psstrings = PS_STRINGS, 407 .sv_stackprot = VM_PROT_ALL, 408 .sv_copyout_strings = NULL, 409 .sv_setregs = NULL, 410 .sv_fixlimit = NULL, 411 .sv_maxssiz = NULL, 412 .sv_flags = 0, 413 .sv_set_syscall_retval = null_set_syscall_retval, 414 .sv_fetch_syscall_args = null_fetch_syscall_args, 415 .sv_syscallnames = NULL, 416 .sv_schedtail = NULL, 417 .sv_thread_detach = NULL, 418 .sv_trap = NULL, 419}; 420 421/* 422 *************************************************************************** 423 **** 424 **** The two following SYSINIT's are proc0 specific glue code. I am not 425 **** convinced that they can not be safely combined, but their order of 426 **** operation has been maintained as the same as the original init_main.c 427 **** for right now. 428 **** 429 **** These probably belong in init_proc.c or kern_proc.c, since they 430 **** deal with proc0 (the fork template process). 431 **** 432 *************************************************************************** 433 */ 434/* ARGSUSED*/ 435static void 436proc0_init(void *dummy __unused) 437{ 438 struct proc *p; 439 struct thread *td; 440 struct ucred *newcred; 441 vm_paddr_t pageablemem; 442 int i; 443 444 GIANT_REQUIRED; 445 p = &proc0; 446 td = &thread0; 447 448 /* 449 * Initialize magic number and osrel. 450 */ 451 p->p_magic = P_MAGIC; 452 p->p_osrel = osreldate; 453 454 /* 455 * Initialize thread and process structures. 456 */ 457 procinit(); /* set up proc zone */ 458 threadinit(); /* set up UMA zones */ 459 460 /* 461 * Initialise scheduler resources. 462 * Add scheduler specific parts to proc, thread as needed. 463 */ 464 schedinit(); /* scheduler gets its house in order */ 465 466 /* 467 * Create process 0 (the swapper). 468 */ 469 LIST_INSERT_HEAD(&allproc, p, p_list); 470 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash); 471 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 472 p->p_pgrp = &pgrp0; 473 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 474 LIST_INIT(&pgrp0.pg_members); 475 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 476 477 pgrp0.pg_session = &session0; 478 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF); 479 refcount_init(&session0.s_count, 1); 480 session0.s_leader = p; 481 482 p->p_sysent = &null_sysvec; 483 p->p_flag = P_SYSTEM | P_INMEM | P_KPROC; 484 p->p_flag2 = 0; 485 p->p_state = PRS_NORMAL; 486 p->p_klist = knlist_alloc(&p->p_mtx); 487 STAILQ_INIT(&p->p_ktr); 488 p->p_nice = NZERO; 489 /* pid_max cannot be greater than PID_MAX */ 490 td->td_tid = PID_MAX + 1; 491 LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash); 492 td->td_state = TDS_RUNNING; 493 td->td_pri_class = PRI_TIMESHARE; 494 td->td_user_pri = PUSER; 495 td->td_base_user_pri = PUSER; 496 td->td_lend_user_pri = PRI_MAX; 497 td->td_priority = PVM; 498 td->td_base_pri = PVM; 499 td->td_oncpu = 0; 500 td->td_flags = TDF_INMEM; 501 td->td_pflags = TDP_KTHREAD; 502 td->td_cpuset = cpuset_thread0(); 503 vm_domain_policy_init(&td->td_vm_dom_policy); 504 vm_domain_policy_set(&td->td_vm_dom_policy, VM_POLICY_NONE, -1); 505 vm_domain_policy_init(&p->p_vm_dom_policy); 506 vm_domain_policy_set(&p->p_vm_dom_policy, VM_POLICY_NONE, -1); 507 prison0_init(); 508 p->p_peers = 0; 509 p->p_leader = p; 510 p->p_reaper = p; 511 LIST_INIT(&p->p_reaplist); 512 513 strncpy(p->p_comm, "kernel", sizeof (p->p_comm)); 514 strncpy(td->td_name, "swapper", sizeof (td->td_name)); 515 516 callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0); 517 callout_init_mtx(&p->p_limco, &p->p_mtx, 0); 518 callout_init(&td->td_slpcallout, 1); 519 520 /* Create credentials. */ 521 newcred = crget(); 522 newcred->cr_ngroups = 1; /* group 0 */ 523 newcred->cr_uidinfo = uifind(0); 524 newcred->cr_ruidinfo = uifind(0); 525 newcred->cr_prison = &prison0; 526 newcred->cr_loginclass = loginclass_find("default"); 527 proc_set_cred_init(p, newcred); 528#ifdef AUDIT 529 audit_cred_kproc0(newcred); 530#endif 531#ifdef MAC 532 mac_cred_create_swapper(newcred); 533#endif 534 /* Create sigacts. */ 535 p->p_sigacts = sigacts_alloc(); 536 537 /* Initialize signal state for process 0. */ 538 siginit(&proc0); 539 540 /* Create the file descriptor table. */ 541 p->p_fd = fdinit(NULL, false); 542 p->p_fdtol = NULL; 543 544 /* Create the limits structures. */ 545 p->p_limit = lim_alloc(); 546 for (i = 0; i < RLIM_NLIMITS; i++) 547 p->p_limit->pl_rlimit[i].rlim_cur = 548 p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY; 549 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur = 550 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 551 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur = 552 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 553 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz; 554 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz; 555 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz; 556 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz; 557 /* Cast to avoid overflow on i386/PAE. */ 558 pageablemem = ptoa((vm_paddr_t)vm_cnt.v_free_count); 559 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur = 560 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem; 561 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3; 562 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem; 563 p->p_cpulimit = RLIM_INFINITY; 564 565 PROC_LOCK(p); 566 thread_cow_get_proc(td, p); 567 PROC_UNLOCK(p); 568 569 /* Initialize resource accounting structures. */ 570 racct_create(&p->p_racct); 571 572 p->p_stats = pstats_alloc(); 573 574 /* Allocate a prototype map so we have something to fork. */ 575 p->p_vmspace = &vmspace0; 576 vmspace0.vm_refcnt = 1; 577 pmap_pinit0(vmspace_pmap(&vmspace0)); 578 579 /* 580 * proc0 is not expected to enter usermode, so there is no special 581 * handling for sv_minuser here, like is done for exec_new_vmspace(). 582 */ 583 vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0), 584 p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser); 585 586 /* 587 * Call the init and ctor for the new thread and proc. We wait 588 * to do this until all other structures are fairly sane. 589 */ 590 EVENTHANDLER_INVOKE(process_init, p); 591 EVENTHANDLER_INVOKE(thread_init, td); 592 EVENTHANDLER_INVOKE(process_ctor, p); 593 EVENTHANDLER_INVOKE(thread_ctor, td); 594 595 /* 596 * Charge root for one process. 597 */ 598 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0); 599 PROC_LOCK(p); 600 racct_add_force(p, RACCT_NPROC, 1); 601 PROC_UNLOCK(p); 602} 603SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL); 604 605/* ARGSUSED*/ 606static void 607proc0_post(void *dummy __unused) 608{ 609 struct timespec ts; 610 struct proc *p; 611 struct rusage ru; 612 struct thread *td; 613 614 /* 615 * Now we can look at the time, having had a chance to verify the 616 * time from the filesystem. Pretend that proc0 started now. 617 */ 618 sx_slock(&allproc_lock); 619 FOREACH_PROC_IN_SYSTEM(p) { 620 microuptime(&p->p_stats->p_start); 621 PROC_STATLOCK(p); 622 rufetch(p, &ru); /* Clears thread stats */ 623 PROC_STATUNLOCK(p); 624 p->p_rux.rux_runtime = 0; 625 p->p_rux.rux_uticks = 0; 626 p->p_rux.rux_sticks = 0; 627 p->p_rux.rux_iticks = 0; 628 FOREACH_THREAD_IN_PROC(p, td) { 629 td->td_runtime = 0; 630 } 631 } 632 sx_sunlock(&allproc_lock); 633 PCPU_SET(switchtime, cpu_ticks()); 634 PCPU_SET(switchticks, ticks); 635 636 /* 637 * Give the ``random'' number generator a thump. 638 */ 639 nanotime(&ts); 640 srandom(ts.tv_sec ^ ts.tv_nsec); 641} 642SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL); 643 644static void 645random_init(void *dummy __unused) 646{ 647 648 /* 649 * After CPU has been started we have some randomness on most 650 * platforms via get_cyclecount(). For platforms that don't 651 * we will reseed random(9) in proc0_post() as well. 652 */ 653 srandom(get_cyclecount()); 654} 655SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL); 656 657/* 658 *************************************************************************** 659 **** 660 **** The following SYSINIT's and glue code should be moved to the 661 **** respective files on a per subsystem basis. 662 **** 663 *************************************************************************** 664 */ 665 666 667/* 668 *************************************************************************** 669 **** 670 **** The following code probably belongs in another file, like 671 **** kern/init_init.c. 672 **** 673 *************************************************************************** 674 */ 675 676/* 677 * List of paths to try when searching for "init". 678 */ 679static char init_path[MAXPATHLEN] = 680#ifdef INIT_PATH 681 __XSTRING(INIT_PATH); 682#else 683 "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init"; 684#endif 685SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, 686 "Path used to search the init process"); 687 688/* 689 * Shutdown timeout of init(8). 690 * Unused within kernel, but used to control init(8), hence do not remove. 691 */ 692#ifndef INIT_SHUTDOWN_TIMEOUT 693#define INIT_SHUTDOWN_TIMEOUT 120 694#endif 695static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT; 696SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout, 697 CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). " 698 "Unused within kernel, but used to control init(8)"); 699 700/* 701 * Start the initial user process; try exec'ing each pathname in init_path. 702 * The program is invoked with one argument containing the boot flags. 703 */ 704static void 705start_init(void *dummy) 706{ 707 vm_offset_t addr; 708 struct execve_args args; 709 int options, error; 710 char *var, *path, *next, *s; 711 char *ucp, **uap, *arg0, *arg1; 712 struct thread *td; 713 struct proc *p; 714 715 mtx_lock(&Giant); 716 717 GIANT_REQUIRED; 718 719 td = curthread; 720 p = td->td_proc; 721 722 vfs_mountroot(); 723 724 /* Wipe GELI passphrase from the environment. */ 725 kern_unsetenv("kern.geom.eli.passphrase"); 726 727 /* 728 * Need just enough stack to hold the faked-up "execve()" arguments. 729 */ 730 addr = p->p_sysent->sv_usrstack - PAGE_SIZE; 731 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 0, 732 VMFS_NO_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0) 733 panic("init: couldn't allocate argument space"); 734 p->p_vmspace->vm_maxsaddr = (caddr_t)addr; 735 p->p_vmspace->vm_ssize = 1; 736 737 if ((var = kern_getenv("init_path")) != NULL) { 738 strlcpy(init_path, var, sizeof(init_path)); 739 freeenv(var); 740 } 741 742 for (path = init_path; *path != '\0'; path = next) { 743 while (*path == ':') 744 path++; 745 if (*path == '\0') 746 break; 747 for (next = path; *next != '\0' && *next != ':'; next++) 748 /* nothing */ ; 749 if (bootverbose) 750 printf("start_init: trying %.*s\n", (int)(next - path), 751 path); 752 753 /* 754 * Move out the boot flag argument. 755 */ 756 options = 0; 757 ucp = (char *)p->p_sysent->sv_usrstack; 758 (void)subyte(--ucp, 0); /* trailing zero */ 759 if (boothowto & RB_SINGLE) { 760 (void)subyte(--ucp, 's'); 761 options = 1; 762 } 763#ifdef notyet 764 if (boothowto & RB_FASTBOOT) { 765 (void)subyte(--ucp, 'f'); 766 options = 1; 767 } 768#endif 769 770#ifdef BOOTCDROM 771 (void)subyte(--ucp, 'C'); 772 options = 1; 773#endif 774 775 if (options == 0) 776 (void)subyte(--ucp, '-'); 777 (void)subyte(--ucp, '-'); /* leading hyphen */ 778 arg1 = ucp; 779 780 /* 781 * Move out the file name (also arg 0). 782 */ 783 (void)subyte(--ucp, 0); 784 for (s = next - 1; s >= path; s--) 785 (void)subyte(--ucp, *s); 786 arg0 = ucp; 787 788 /* 789 * Move out the arg pointers. 790 */ 791 uap = (char **)rounddown2((intptr_t)ucp, sizeof(intptr_t)); 792 (void)suword((caddr_t)--uap, (long)0); /* terminator */ 793 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1); 794 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0); 795 796 /* 797 * Point at the arguments. 798 */ 799 args.fname = arg0; 800 args.argv = uap; 801 args.envv = NULL; 802 803 /* 804 * Now try to exec the program. If can't for any reason 805 * other than it doesn't exist, complain. 806 * 807 * Otherwise, return via fork_trampoline() all the way 808 * to user mode as init! 809 */ 810 if ((error = sys_execve(td, &args)) == 0) { 811 mtx_unlock(&Giant); 812 return; 813 } 814 if (error != ENOENT) 815 printf("exec %.*s: error %d\n", (int)(next - path), 816 path, error); 817 } 818 printf("init: not found in path %s\n", init_path); 819 panic("no init"); 820} 821 822/* 823 * Like kproc_create(), but runs in it's own address space. 824 * We do this early to reserve pid 1. 825 * 826 * Note special case - do not make it runnable yet. Other work 827 * in progress will change this more. 828 */ 829static void 830create_init(const void *udata __unused) 831{ 832 struct fork_req fr; 833 struct ucred *newcred, *oldcred; 834 struct thread *td; 835 int error; 836 837 bzero(&fr, sizeof(fr)); 838 fr.fr_flags = RFFDG | RFPROC | RFSTOPPED; 839 fr.fr_procp = &initproc; 840 error = fork1(&thread0, &fr); 841 if (error) 842 panic("cannot fork init: %d\n", error); 843 KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1")); 844 /* divorce init's credentials from the kernel's */ 845 newcred = crget(); 846 sx_xlock(&proctree_lock); 847 PROC_LOCK(initproc); 848 initproc->p_flag |= P_SYSTEM | P_INMEM; 849 initproc->p_treeflag |= P_TREE_REAPER; 850 LIST_INSERT_HEAD(&initproc->p_reaplist, &proc0, p_reapsibling); 851 oldcred = initproc->p_ucred; 852 crcopy(newcred, oldcred); 853#ifdef MAC 854 mac_cred_create_init(newcred); 855#endif 856#ifdef AUDIT 857 audit_cred_proc1(newcred); 858#endif 859 proc_set_cred(initproc, newcred); 860 td = FIRST_THREAD_IN_PROC(initproc); 861 crfree(td->td_ucred); 862 td->td_ucred = crhold(initproc->p_ucred); 863 PROC_UNLOCK(initproc); 864 sx_xunlock(&proctree_lock); 865 crfree(oldcred); 866 cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc), 867 start_init, NULL); 868} 869SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL); 870 871/* 872 * Make it runnable now. 873 */ 874static void 875kick_init(const void *udata __unused) 876{ 877 struct thread *td; 878 879 td = FIRST_THREAD_IN_PROC(initproc); 880 thread_lock(td); 881 TD_SET_CAN_RUN(td); 882 sched_add(td, SRQ_BORING); 883 thread_unlock(td); 884} 885SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL); 886