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("$FreeBSD$"); 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 98 void mi_startup(void); /* Should be elsewhere */ 99 100 /* Components of the first process -- never freed. */ 101 static struct session session0; 102 static struct pgrp pgrp0; 103 struct proc proc0; 104 struct thread0_storage thread0_st __aligned(32); 105 struct vmspace vmspace0; 106 struct proc *initproc; 107 108 #ifndef BOOTHOWTO 109 #define BOOTHOWTO 0 110 #endif 111 int boothowto = BOOTHOWTO; /* initialized so that it can be patched */ 112 SYSCTL_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 118 int bootverbose = BOOTVERBOSE; 119 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, 120 "Control the output of verbose kernel messages"); 121 122 #ifdef INVARIANTS 123 FEATURE(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 */ 131 SYSINIT(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 */ 137 SET_DECLARE(sysinit_set, struct sysinit); 138 struct sysinit **sysinit, **sysinit_end; 139 struct sysinit **newsysinit, **newsysinit_end; 140 141 EVENTHANDLER_LIST_DECLARE(process_init); 142 EVENTHANDLER_LIST_DECLARE(thread_init); 143 EVENTHANDLER_LIST_DECLARE(process_ctor); 144 EVENTHANDLER_LIST_DECLARE(thread_ctor); 145 146 /* 147 * Merge a new sysinit set into the current set, reallocating it if 148 * necessary. This can only be called after malloc is running. 149 */ 150 void 151 sysinit_add(struct sysinit **set, struct sysinit **set_end) 152 { 153 struct sysinit **newset; 154 struct sysinit **sipp; 155 struct sysinit **xipp; 156 int count; 157 158 count = set_end - set; 159 if (newsysinit) 160 count += newsysinit_end - newsysinit; 161 else 162 count += sysinit_end - sysinit; 163 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT); 164 if (newset == NULL) 165 panic("cannot malloc for sysinit"); 166 xipp = newset; 167 if (newsysinit) 168 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 169 *xipp++ = *sipp; 170 else 171 for (sipp = sysinit; sipp < sysinit_end; sipp++) 172 *xipp++ = *sipp; 173 for (sipp = set; sipp < set_end; sipp++) 174 *xipp++ = *sipp; 175 if (newsysinit) 176 free(newsysinit, M_TEMP); 177 newsysinit = newset; 178 newsysinit_end = newset + count; 179 } 180 181 #if defined (DDB) && defined(VERBOSE_SYSINIT) 182 static const char * 183 symbol_name(vm_offset_t va, db_strategy_t strategy) 184 { 185 const char *name; 186 c_db_sym_t sym; 187 db_expr_t offset; 188 189 if (va == 0) 190 return (NULL); 191 sym = db_search_symbol(va, strategy, &offset); 192 if (offset != 0) 193 return (NULL); 194 db_symbol_values(sym, &name, NULL); 195 return (name); 196 } 197 #endif 198 199 /* 200 * System startup; initialize the world, create process 0, mount root 201 * filesystem, and fork to create init and pagedaemon. Most of the 202 * hard work is done in the lower-level initialization routines including 203 * startup(), which does memory initialization and autoconfiguration. 204 * 205 * This allows simple addition of new kernel subsystems that require 206 * boot time initialization. It also allows substitution of subsystem 207 * (for instance, a scheduler, kernel profiler, or VM system) by object 208 * module. Finally, it allows for optional "kernel threads". 209 */ 210 void 211 mi_startup(void) 212 { 213 214 struct sysinit **sipp; /* system initialization*/ 215 struct sysinit **xipp; /* interior loop of sort*/ 216 struct sysinit *save; /* bubble*/ 217 218 #if defined(VERBOSE_SYSINIT) 219 int last; 220 int verbose; 221 #endif 222 223 TSENTER(); 224 225 if (boothowto & RB_VERBOSE) 226 bootverbose++; 227 228 if (sysinit == NULL) { 229 sysinit = SET_BEGIN(sysinit_set); 230 sysinit_end = SET_LIMIT(sysinit_set); 231 } 232 233 restart: 234 /* 235 * Perform a bubble sort of the system initialization objects by 236 * their subsystem (primary key) and order (secondary key). 237 */ 238 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 239 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 240 if ((*sipp)->subsystem < (*xipp)->subsystem || 241 ((*sipp)->subsystem == (*xipp)->subsystem && 242 (*sipp)->order <= (*xipp)->order)) 243 continue; /* skip*/ 244 save = *sipp; 245 *sipp = *xipp; 246 *xipp = save; 247 } 248 } 249 250 #if defined(VERBOSE_SYSINIT) 251 last = SI_SUB_COPYRIGHT; 252 verbose = 0; 253 #if !defined(DDB) 254 printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n"); 255 #endif 256 #endif 257 258 /* 259 * Traverse the (now) ordered list of system initialization tasks. 260 * Perform each task, and continue on to the next task. 261 */ 262 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 263 264 if ((*sipp)->subsystem == SI_SUB_DUMMY) 265 continue; /* skip dummy task(s)*/ 266 267 if ((*sipp)->subsystem == SI_SUB_DONE) 268 continue; 269 270 #if defined(VERBOSE_SYSINIT) 271 if ((*sipp)->subsystem > last) { 272 verbose = 1; 273 last = (*sipp)->subsystem; 274 printf("subsystem %x\n", last); 275 } 276 if (verbose) { 277 #if defined(DDB) 278 const char *func, *data; 279 280 func = symbol_name((vm_offset_t)(*sipp)->func, 281 DB_STGY_PROC); 282 data = symbol_name((vm_offset_t)(*sipp)->udata, 283 DB_STGY_ANY); 284 if (func != NULL && data != NULL) 285 printf(" %s(&%s)... ", func, data); 286 else if (func != NULL) 287 printf(" %s(%p)... ", func, (*sipp)->udata); 288 else 289 #endif 290 printf(" %p(%p)... ", (*sipp)->func, 291 (*sipp)->udata); 292 } 293 #endif 294 295 /* Call function */ 296 (*((*sipp)->func))((*sipp)->udata); 297 298 #if defined(VERBOSE_SYSINIT) 299 if (verbose) 300 printf("done.\n"); 301 #endif 302 303 /* Check off the one we're just done */ 304 (*sipp)->subsystem = SI_SUB_DONE; 305 306 /* Check if we've installed more sysinit items via KLD */ 307 if (newsysinit != NULL) { 308 if (sysinit != SET_BEGIN(sysinit_set)) 309 free(sysinit, M_TEMP); 310 sysinit = newsysinit; 311 sysinit_end = newsysinit_end; 312 newsysinit = NULL; 313 newsysinit_end = NULL; 314 goto restart; 315 } 316 } 317 318 TSEXIT(); /* Here so we don't overlap with start_init. */ 319 320 mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED); 321 mtx_unlock(&Giant); 322 323 /* 324 * Now hand over this thread to swapper. 325 */ 326 swapper(); 327 /* NOTREACHED*/ 328 } 329 330 static void 331 print_caddr_t(void *data) 332 { 333 printf("%s", (char *)data); 334 } 335 336 static void 337 print_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 } 348 349 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, 350 copyright); 351 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, 352 trademark); 353 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL); 354 355 #ifdef WITNESS 356 static char wit_warn[] = 357 "WARNING: WITNESS option enabled, expect reduced performance.\n"; 358 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1, 359 print_caddr_t, wit_warn); 360 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 1, 361 print_caddr_t, wit_warn); 362 #endif 363 364 #ifdef DIAGNOSTIC 365 static char diag_warn[] = 366 "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n"; 367 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2, 368 print_caddr_t, diag_warn); 369 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 2, 370 print_caddr_t, diag_warn); 371 #endif 372 373 static int 374 null_fetch_syscall_args(struct thread *td __unused) 375 { 376 377 panic("null_fetch_syscall_args"); 378 } 379 380 static void 381 null_set_syscall_retval(struct thread *td __unused, int error __unused) 382 { 383 384 panic("null_set_syscall_retval"); 385 } 386 387 struct 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 * The two following SYSINIT's are proc0 specific glue code. I am not 423 * convinced that they can not be safely combined, but their order of 424 * operation has been maintained as the same as the original init_main.c 425 * for right now. 426 */ 427 /* ARGSUSED*/ 428 static void 429 proc0_init(void *dummy __unused) 430 { 431 struct proc *p; 432 struct thread *td; 433 struct ucred *newcred; 434 struct uidinfo tmpuinfo; 435 struct loginclass tmplc = { 436 .lc_name = "", 437 }; 438 vm_paddr_t pageablemem; 439 int i; 440 441 GIANT_REQUIRED; 442 p = &proc0; 443 td = &thread0; 444 445 /* 446 * Initialize magic number and osrel. 447 */ 448 p->p_magic = P_MAGIC; 449 p->p_osrel = osreldate; 450 451 /* 452 * Initialize thread and process structures. 453 */ 454 procinit(); /* set up proc zone */ 455 threadinit(); /* set up UMA zones */ 456 457 /* 458 * Initialise scheduler resources. 459 * Add scheduler specific parts to proc, thread as needed. 460 */ 461 schedinit(); /* scheduler gets its house in order */ 462 463 /* 464 * Create process 0 (the swapper). 465 */ 466 LIST_INSERT_HEAD(&allproc, p, p_list); 467 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash); 468 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 469 p->p_pgrp = &pgrp0; 470 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 471 LIST_INIT(&pgrp0.pg_members); 472 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 473 474 pgrp0.pg_session = &session0; 475 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF); 476 refcount_init(&session0.s_count, 1); 477 session0.s_leader = p; 478 479 p->p_sysent = &null_sysvec; 480 p->p_flag = P_SYSTEM | P_INMEM | P_KPROC; 481 p->p_flag2 = 0; 482 p->p_state = PRS_NORMAL; 483 p->p_klist = knlist_alloc(&p->p_mtx); 484 STAILQ_INIT(&p->p_ktr); 485 p->p_nice = NZERO; 486 /* pid_max cannot be greater than PID_MAX */ 487 td->td_tid = PID_MAX + 1; 488 LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash); 489 td->td_state = TDS_RUNNING; 490 td->td_pri_class = PRI_TIMESHARE; 491 td->td_user_pri = PUSER; 492 td->td_base_user_pri = PUSER; 493 td->td_lend_user_pri = PRI_MAX; 494 td->td_priority = PVM; 495 td->td_base_pri = PVM; 496 td->td_oncpu = curcpu; 497 td->td_flags = TDF_INMEM; 498 td->td_pflags = TDP_KTHREAD; 499 td->td_cpuset = cpuset_thread0(); 500 vm_domain_policy_init(&td->td_vm_dom_policy); 501 vm_domain_policy_set(&td->td_vm_dom_policy, VM_POLICY_NONE, -1); 502 vm_domain_policy_init(&p->p_vm_dom_policy); 503 vm_domain_policy_set(&p->p_vm_dom_policy, VM_POLICY_NONE, -1); 504 prison0_init(); 505 p->p_peers = 0; 506 p->p_leader = p; 507 p->p_reaper = p; 508 LIST_INIT(&p->p_reaplist); 509 510 strncpy(p->p_comm, "kernel", sizeof (p->p_comm)); 511 strncpy(td->td_name, "swapper", sizeof (td->td_name)); 512 513 callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0); 514 callout_init_mtx(&p->p_limco, &p->p_mtx, 0); 515 callout_init(&td->td_slpcallout, 1); 516 517 /* Create credentials. */ 518 newcred = crget(); 519 newcred->cr_ngroups = 1; /* group 0 */ 520 /* A hack to prevent uifind from tripping over NULL pointers. */ 521 curthread->td_ucred = newcred; 522 tmpuinfo.ui_uid = 1; 523 newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo; 524 newcred->cr_uidinfo = uifind(0); 525 newcred->cr_ruidinfo = uifind(0); 526 newcred->cr_loginclass = &tmplc; 527 newcred->cr_loginclass = loginclass_find("default"); 528 /* End hack. creds get properly set later with thread_cow_get_proc */ 529 curthread->td_ucred = NULL; 530 newcred->cr_prison = &prison0; 531 proc_set_cred_init(p, newcred); 532 #ifdef AUDIT 533 audit_cred_kproc0(newcred); 534 #endif 535 #ifdef MAC 536 mac_cred_create_swapper(newcred); 537 #endif 538 /* Create sigacts. */ 539 p->p_sigacts = sigacts_alloc(); 540 541 /* Initialize signal state for process 0. */ 542 siginit(&proc0); 543 544 /* Create the file descriptor table. */ 545 p->p_fd = fdinit(NULL, false); 546 p->p_fdtol = NULL; 547 548 /* Create the limits structures. */ 549 p->p_limit = lim_alloc(); 550 for (i = 0; i < RLIM_NLIMITS; i++) 551 p->p_limit->pl_rlimit[i].rlim_cur = 552 p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY; 553 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur = 554 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 555 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur = 556 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 557 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz; 558 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz; 559 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz; 560 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz; 561 /* Cast to avoid overflow on i386/PAE. */ 562 pageablemem = ptoa((vm_paddr_t)vm_cnt.v_free_count); 563 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur = 564 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem; 565 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3; 566 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem; 567 p->p_cpulimit = RLIM_INFINITY; 568 569 PROC_LOCK(p); 570 thread_cow_get_proc(td, p); 571 PROC_UNLOCK(p); 572 573 /* Initialize resource accounting structures. */ 574 racct_create(&p->p_racct); 575 576 p->p_stats = pstats_alloc(); 577 578 /* Allocate a prototype map so we have something to fork. */ 579 p->p_vmspace = &vmspace0; 580 vmspace0.vm_refcnt = 1; 581 pmap_pinit0(vmspace_pmap(&vmspace0)); 582 583 /* 584 * proc0 is not expected to enter usermode, so there is no special 585 * handling for sv_minuser here, like is done for exec_new_vmspace(). 586 */ 587 vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0), 588 p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser); 589 590 /* 591 * Call the init and ctor for the new thread and proc. We wait 592 * to do this until all other structures are fairly sane. 593 */ 594 EVENTHANDLER_DIRECT_INVOKE(process_init, p); 595 EVENTHANDLER_DIRECT_INVOKE(thread_init, td); 596 EVENTHANDLER_DIRECT_INVOKE(process_ctor, p); 597 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td); 598 599 /* 600 * Charge root for one process. 601 */ 602 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0); 603 PROC_LOCK(p); 604 racct_add_force(p, RACCT_NPROC, 1); 605 PROC_UNLOCK(p); 606 } 607 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL); 608 609 /* ARGSUSED*/ 610 static void 611 proc0_post(void *dummy __unused) 612 { 613 struct timespec ts; 614 struct proc *p; 615 struct rusage ru; 616 struct thread *td; 617 618 /* 619 * Now we can look at the time, having had a chance to verify the 620 * time from the filesystem. Pretend that proc0 started now. 621 */ 622 sx_slock(&allproc_lock); 623 FOREACH_PROC_IN_SYSTEM(p) { 624 microuptime(&p->p_stats->p_start); 625 PROC_STATLOCK(p); 626 rufetch(p, &ru); /* Clears thread stats */ 627 PROC_STATUNLOCK(p); 628 p->p_rux.rux_runtime = 0; 629 p->p_rux.rux_uticks = 0; 630 p->p_rux.rux_sticks = 0; 631 p->p_rux.rux_iticks = 0; 632 FOREACH_THREAD_IN_PROC(p, td) { 633 td->td_runtime = 0; 634 } 635 } 636 sx_sunlock(&allproc_lock); 637 PCPU_SET(switchtime, cpu_ticks()); 638 PCPU_SET(switchticks, ticks); 639 640 /* 641 * Give the ``random'' number generator a thump. 642 */ 643 nanotime(&ts); 644 srandom(ts.tv_sec ^ ts.tv_nsec); 645 } 646 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL); 647 648 static void 649 random_init(void *dummy __unused) 650 { 651 652 /* 653 * After CPU has been started we have some randomness on most 654 * platforms via get_cyclecount(). For platforms that don't 655 * we will reseed random(9) in proc0_post() as well. 656 */ 657 srandom(get_cyclecount()); 658 } 659 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL); 660 661 /* 662 *************************************************************************** 663 **** 664 **** The following SYSINIT's and glue code should be moved to the 665 **** respective files on a per subsystem basis. 666 **** 667 *************************************************************************** 668 */ 669 670 /* 671 * List of paths to try when searching for "init". 672 */ 673 static char init_path[MAXPATHLEN] = 674 #ifdef INIT_PATH 675 __XSTRING(INIT_PATH); 676 #else 677 "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init"; 678 #endif 679 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, 680 "Path used to search the init process"); 681 682 /* 683 * Shutdown timeout of init(8). 684 * Unused within kernel, but used to control init(8), hence do not remove. 685 */ 686 #ifndef INIT_SHUTDOWN_TIMEOUT 687 #define INIT_SHUTDOWN_TIMEOUT 120 688 #endif 689 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT; 690 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout, 691 CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). " 692 "Unused within kernel, but used to control init(8)"); 693 694 /* 695 * Start the initial user process; try exec'ing each pathname in init_path. 696 * The program is invoked with one argument containing the boot flags. 697 */ 698 static void 699 start_init(void *dummy) 700 { 701 vm_offset_t addr; 702 struct execve_args args; 703 int options, error; 704 char *var, *path, *next, *s; 705 char *ucp, **uap, *arg0, *arg1; 706 struct thread *td; 707 struct proc *p; 708 709 mtx_lock(&Giant); 710 711 GIANT_REQUIRED; 712 713 TSENTER(); /* Here so we don't overlap with mi_startup. */ 714 715 td = curthread; 716 p = td->td_proc; 717 718 vfs_mountroot(); 719 720 /* Wipe GELI passphrase from the environment. */ 721 kern_unsetenv("kern.geom.eli.passphrase"); 722 723 /* 724 * Need just enough stack to hold the faked-up "execve()" arguments. 725 */ 726 addr = p->p_sysent->sv_usrstack - PAGE_SIZE; 727 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 0, 728 VMFS_NO_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0) 729 panic("init: couldn't allocate argument space"); 730 p->p_vmspace->vm_maxsaddr = (caddr_t)addr; 731 p->p_vmspace->vm_ssize = 1; 732 733 if ((var = kern_getenv("init_path")) != NULL) { 734 strlcpy(init_path, var, sizeof(init_path)); 735 freeenv(var); 736 } 737 738 for (path = init_path; *path != '\0'; path = next) { 739 while (*path == ':') 740 path++; 741 if (*path == '\0') 742 break; 743 for (next = path; *next != '\0' && *next != ':'; next++) 744 /* nothing */ ; 745 if (bootverbose) 746 printf("start_init: trying %.*s\n", (int)(next - path), 747 path); 748 749 /* 750 * Move out the boot flag argument. 751 */ 752 options = 0; 753 ucp = (char *)p->p_sysent->sv_usrstack; 754 (void)subyte(--ucp, 0); /* trailing zero */ 755 if (boothowto & RB_SINGLE) { 756 (void)subyte(--ucp, 's'); 757 options = 1; 758 } 759 #ifdef notyet 760 if (boothowto & RB_FASTBOOT) { 761 (void)subyte(--ucp, 'f'); 762 options = 1; 763 } 764 #endif 765 766 #ifdef BOOTCDROM 767 (void)subyte(--ucp, 'C'); 768 options = 1; 769 #endif 770 771 if (options == 0) 772 (void)subyte(--ucp, '-'); 773 (void)subyte(--ucp, '-'); /* leading hyphen */ 774 arg1 = ucp; 775 776 /* 777 * Move out the file name (also arg 0). 778 */ 779 (void)subyte(--ucp, 0); 780 for (s = next - 1; s >= path; s--) 781 (void)subyte(--ucp, *s); 782 arg0 = ucp; 783 784 /* 785 * Move out the arg pointers. 786 */ 787 uap = (char **)rounddown2((intptr_t)ucp, sizeof(intptr_t)); 788 (void)suword((caddr_t)--uap, (long)0); /* terminator */ 789 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1); 790 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0); 791 792 /* 793 * Point at the arguments. 794 */ 795 args.fname = arg0; 796 args.argv = uap; 797 args.envv = NULL; 798 799 /* 800 * Now try to exec the program. If can't for any reason 801 * other than it doesn't exist, complain. 802 * 803 * Otherwise, return via fork_trampoline() all the way 804 * to user mode as init! 805 */ 806 if ((error = sys_execve(td, &args)) == EJUSTRETURN) { 807 mtx_unlock(&Giant); 808 TSEXIT(); 809 return; 810 } 811 if (error != ENOENT) 812 printf("exec %.*s: error %d\n", (int)(next - path), 813 path, error); 814 } 815 printf("init: not found in path %s\n", init_path); 816 panic("no init"); 817 } 818 819 /* 820 * Like kproc_create(), but runs in its own address space. 821 * We do this early to reserve pid 1. 822 * 823 * Note special case - do not make it runnable yet. Other work 824 * in progress will change this more. 825 */ 826 static void 827 create_init(const void *udata __unused) 828 { 829 struct fork_req fr; 830 struct ucred *newcred, *oldcred; 831 struct thread *td; 832 int error; 833 834 bzero(&fr, sizeof(fr)); 835 fr.fr_flags = RFFDG | RFPROC | RFSTOPPED; 836 fr.fr_procp = &initproc; 837 error = fork1(&thread0, &fr); 838 if (error) 839 panic("cannot fork init: %d\n", error); 840 KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1")); 841 /* divorce init's credentials from the kernel's */ 842 newcred = crget(); 843 sx_xlock(&proctree_lock); 844 PROC_LOCK(initproc); 845 initproc->p_flag |= P_SYSTEM | P_INMEM; 846 initproc->p_treeflag |= P_TREE_REAPER; 847 LIST_INSERT_HEAD(&initproc->p_reaplist, &proc0, p_reapsibling); 848 oldcred = initproc->p_ucred; 849 crcopy(newcred, oldcred); 850 #ifdef MAC 851 mac_cred_create_init(newcred); 852 #endif 853 #ifdef AUDIT 854 audit_cred_proc1(newcred); 855 #endif 856 proc_set_cred(initproc, newcred); 857 td = FIRST_THREAD_IN_PROC(initproc); 858 crfree(td->td_ucred); 859 td->td_ucred = crhold(initproc->p_ucred); 860 PROC_UNLOCK(initproc); 861 sx_xunlock(&proctree_lock); 862 crfree(oldcred); 863 cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc), 864 start_init, NULL); 865 } 866 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL); 867 868 /* 869 * Make it runnable now. 870 */ 871 static void 872 kick_init(const void *udata __unused) 873 { 874 struct thread *td; 875 876 td = FIRST_THREAD_IN_PROC(initproc); 877 thread_lock(td); 878 TD_SET_CAN_RUN(td); 879 sched_add(td, SRQ_BORING); 880 thread_unlock(td); 881 } 882 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL); 883