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