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_kdb.h" 51 #include "opt_init_path.h" 52 #include "opt_verbose_sysinit.h" 53 54 #include <sys/param.h> 55 #include <sys/kernel.h> 56 #include <sys/epoch.h> 57 #include <sys/eventhandler.h> 58 #include <sys/exec.h> 59 #include <sys/file.h> 60 #include <sys/filedesc.h> 61 #include <sys/imgact.h> 62 #include <sys/jail.h> 63 #include <sys/ktr.h> 64 #include <sys/lock.h> 65 #include <sys/loginclass.h> 66 #include <sys/mount.h> 67 #include <sys/mutex.h> 68 #include <sys/dtrace_bsd.h> 69 #include <sys/syscallsubr.h> 70 #include <sys/sysctl.h> 71 #include <sys/proc.h> 72 #include <sys/racct.h> 73 #include <sys/resourcevar.h> 74 #include <sys/systm.h> 75 #include <sys/signalvar.h> 76 #include <sys/vnode.h> 77 #include <sys/sysent.h> 78 #include <sys/reboot.h> 79 #include <sys/sched.h> 80 #include <sys/sx.h> 81 #include <sys/sysproto.h> 82 #include <sys/vmmeter.h> 83 #include <sys/unistd.h> 84 #include <sys/malloc.h> 85 #include <sys/conf.h> 86 #include <sys/cpuset.h> 87 88 #include <machine/cpu.h> 89 90 #include <security/audit/audit.h> 91 #include <security/mac/mac_framework.h> 92 93 #include <vm/vm.h> 94 #include <vm/vm_param.h> 95 #include <vm/vm_extern.h> 96 #include <vm/pmap.h> 97 #include <vm/vm_map.h> 98 #include <sys/copyright.h> 99 100 #include <ddb/ddb.h> 101 #include <ddb/db_sym.h> 102 103 void mi_startup(void); /* Should be elsewhere */ 104 105 /* Components of the first process -- never freed. */ 106 static struct session session0; 107 static struct pgrp pgrp0; 108 struct proc proc0; 109 struct thread0_storage thread0_st __aligned(32); 110 struct vmspace vmspace0; 111 struct proc *initproc; 112 113 int 114 linux_alloc_current_noop(struct thread *td __unused, int flags __unused) 115 { 116 return (0); 117 } 118 int (*lkpi_alloc_current)(struct thread *, int) = linux_alloc_current_noop; 119 120 #ifndef BOOTHOWTO 121 #define BOOTHOWTO 0 122 #endif 123 int boothowto = BOOTHOWTO; /* initialized so that it can be patched */ 124 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, 125 "Boot control flags, passed from loader"); 126 127 #ifndef BOOTVERBOSE 128 #define BOOTVERBOSE 0 129 #endif 130 int bootverbose = BOOTVERBOSE; 131 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, 132 "Control the output of verbose kernel messages"); 133 134 #ifdef VERBOSE_SYSINIT 135 /* 136 * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to 137 * dictate the default VERBOSE_SYSINIT behavior. Significant values for this 138 * option and associated tunable are: 139 * - 0, 'compiled in but silent by default' 140 * - 1, 'compiled in but verbose by default' (default) 141 */ 142 int verbose_sysinit = VERBOSE_SYSINIT; 143 TUNABLE_INT("debug.verbose_sysinit", &verbose_sysinit); 144 #endif 145 146 #ifdef INVARIANTS 147 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance"); 148 #endif 149 150 /* 151 * This ensures that there is at least one entry so that the sysinit_set 152 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never 153 * executed. 154 */ 155 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL); 156 157 /* 158 * The sysinit table itself. Items are checked off as the are run. 159 * If we want to register new sysinit types, add them to newsysinit. 160 */ 161 SET_DECLARE(sysinit_set, struct sysinit); 162 struct sysinit **sysinit, **sysinit_end; 163 struct sysinit **newsysinit, **newsysinit_end; 164 165 /* 166 * Merge a new sysinit set into the current set, reallocating it if 167 * necessary. This can only be called after malloc is running. 168 */ 169 void 170 sysinit_add(struct sysinit **set, struct sysinit **set_end) 171 { 172 struct sysinit **newset; 173 struct sysinit **sipp; 174 struct sysinit **xipp; 175 int count; 176 177 count = set_end - set; 178 if (newsysinit) 179 count += newsysinit_end - newsysinit; 180 else 181 count += sysinit_end - sysinit; 182 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT); 183 if (newset == NULL) 184 panic("cannot malloc for sysinit"); 185 xipp = newset; 186 if (newsysinit) 187 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 188 *xipp++ = *sipp; 189 else 190 for (sipp = sysinit; sipp < sysinit_end; sipp++) 191 *xipp++ = *sipp; 192 for (sipp = set; sipp < set_end; sipp++) 193 *xipp++ = *sipp; 194 if (newsysinit) 195 free(newsysinit, M_TEMP); 196 newsysinit = newset; 197 newsysinit_end = newset + count; 198 } 199 200 #if defined (DDB) && defined(VERBOSE_SYSINIT) 201 static const char * 202 symbol_name(vm_offset_t va, db_strategy_t strategy) 203 { 204 const char *name; 205 c_db_sym_t sym; 206 db_expr_t offset; 207 208 if (va == 0) 209 return (NULL); 210 sym = db_search_symbol(va, strategy, &offset); 211 if (offset != 0) 212 return (NULL); 213 db_symbol_values(sym, &name, NULL); 214 return (name); 215 } 216 #endif 217 218 /* 219 * System startup; initialize the world, create process 0, mount root 220 * filesystem, and fork to create init and pagedaemon. Most of the 221 * hard work is done in the lower-level initialization routines including 222 * startup(), which does memory initialization and autoconfiguration. 223 * 224 * This allows simple addition of new kernel subsystems that require 225 * boot time initialization. It also allows substitution of subsystem 226 * (for instance, a scheduler, kernel profiler, or VM system) by object 227 * module. Finally, it allows for optional "kernel threads". 228 */ 229 void 230 mi_startup(void) 231 { 232 233 struct sysinit **sipp; /* system initialization*/ 234 struct sysinit **xipp; /* interior loop of sort*/ 235 struct sysinit *save; /* bubble*/ 236 237 #if defined(VERBOSE_SYSINIT) 238 int last; 239 int verbose; 240 #endif 241 242 TSENTER(); 243 244 if (boothowto & RB_VERBOSE) 245 bootverbose++; 246 247 if (sysinit == NULL) { 248 sysinit = SET_BEGIN(sysinit_set); 249 sysinit_end = SET_LIMIT(sysinit_set); 250 } 251 252 restart: 253 /* 254 * Perform a bubble sort of the system initialization objects by 255 * their subsystem (primary key) and order (secondary key). 256 */ 257 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 258 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 259 if ((*sipp)->subsystem < (*xipp)->subsystem || 260 ((*sipp)->subsystem == (*xipp)->subsystem && 261 (*sipp)->order <= (*xipp)->order)) 262 continue; /* skip*/ 263 save = *sipp; 264 *sipp = *xipp; 265 *xipp = save; 266 } 267 } 268 269 #if defined(VERBOSE_SYSINIT) 270 last = SI_SUB_COPYRIGHT; 271 verbose = 0; 272 #if !defined(DDB) 273 printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n"); 274 #endif 275 #endif 276 277 /* 278 * Traverse the (now) ordered list of system initialization tasks. 279 * Perform each task, and continue on to the next task. 280 */ 281 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 282 if ((*sipp)->subsystem == SI_SUB_DUMMY) 283 continue; /* skip dummy task(s)*/ 284 285 if ((*sipp)->subsystem == SI_SUB_DONE) 286 continue; 287 288 #if defined(VERBOSE_SYSINIT) 289 if ((*sipp)->subsystem > last && verbose_sysinit != 0) { 290 verbose = 1; 291 last = (*sipp)->subsystem; 292 printf("subsystem %x\n", last); 293 } 294 if (verbose) { 295 #if defined(DDB) 296 const char *func, *data; 297 298 func = symbol_name((vm_offset_t)(*sipp)->func, 299 DB_STGY_PROC); 300 data = symbol_name((vm_offset_t)(*sipp)->udata, 301 DB_STGY_ANY); 302 if (func != NULL && data != NULL) 303 printf(" %s(&%s)... ", func, data); 304 else if (func != NULL) 305 printf(" %s(%p)... ", func, (*sipp)->udata); 306 else 307 #endif 308 printf(" %p(%p)... ", (*sipp)->func, 309 (*sipp)->udata); 310 } 311 #endif 312 313 /* Call function */ 314 (*((*sipp)->func))((*sipp)->udata); 315 316 #if defined(VERBOSE_SYSINIT) 317 if (verbose) 318 printf("done.\n"); 319 #endif 320 321 /* Check off the one we're just done */ 322 (*sipp)->subsystem = SI_SUB_DONE; 323 324 /* Check if we've installed more sysinit items via KLD */ 325 if (newsysinit != NULL) { 326 if (sysinit != SET_BEGIN(sysinit_set)) 327 free(sysinit, M_TEMP); 328 sysinit = newsysinit; 329 sysinit_end = newsysinit_end; 330 newsysinit = NULL; 331 newsysinit_end = NULL; 332 goto restart; 333 } 334 } 335 336 TSEXIT(); /* Here so we don't overlap with start_init. */ 337 338 mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED); 339 mtx_unlock(&Giant); 340 341 /* 342 * Now hand over this thread to swapper. 343 */ 344 swapper(); 345 /* NOTREACHED*/ 346 } 347 348 static void 349 print_caddr_t(void *data) 350 { 351 printf("%s", (char *)data); 352 } 353 354 static void 355 print_version(void *data __unused) 356 { 357 int len; 358 359 /* Strip a trailing newline from version. */ 360 len = strlen(version); 361 while (len > 0 && version[len - 1] == '\n') 362 len--; 363 printf("%.*s %s\n", len, version, machine); 364 printf("%s\n", compiler_version); 365 } 366 367 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, 368 copyright); 369 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, 370 trademark); 371 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL); 372 373 #ifdef WITNESS 374 static char wit_warn[] = 375 "WARNING: WITNESS option enabled, expect reduced performance.\n"; 376 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH, 377 print_caddr_t, wit_warn); 378 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH, 379 print_caddr_t, wit_warn); 380 #endif 381 382 #ifdef DIAGNOSTIC 383 static char diag_warn[] = 384 "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n"; 385 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH, 386 print_caddr_t, diag_warn); 387 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH, 388 print_caddr_t, diag_warn); 389 #endif 390 391 static int 392 null_fetch_syscall_args(struct thread *td __unused) 393 { 394 395 panic("null_fetch_syscall_args"); 396 } 397 398 static void 399 null_set_syscall_retval(struct thread *td __unused, int error __unused) 400 { 401 402 panic("null_set_syscall_retval"); 403 } 404 405 struct sysentvec null_sysvec = { 406 .sv_size = 0, 407 .sv_table = NULL, 408 .sv_transtrap = NULL, 409 .sv_fixup = NULL, 410 .sv_sendsig = NULL, 411 .sv_sigcode = NULL, 412 .sv_szsigcode = NULL, 413 .sv_name = "null", 414 .sv_coredump = NULL, 415 .sv_imgact_try = NULL, 416 .sv_minsigstksz = 0, 417 .sv_minuser = VM_MIN_ADDRESS, 418 .sv_maxuser = VM_MAXUSER_ADDRESS, 419 .sv_usrstack = USRSTACK, 420 .sv_psstrings = PS_STRINGS, 421 .sv_stackprot = VM_PROT_ALL, 422 .sv_copyout_strings = NULL, 423 .sv_setregs = NULL, 424 .sv_fixlimit = NULL, 425 .sv_maxssiz = NULL, 426 .sv_flags = 0, 427 .sv_set_syscall_retval = null_set_syscall_retval, 428 .sv_fetch_syscall_args = null_fetch_syscall_args, 429 .sv_syscallnames = NULL, 430 .sv_schedtail = NULL, 431 .sv_thread_detach = NULL, 432 .sv_trap = NULL, 433 }; 434 435 /* 436 * The two following SYSINIT's are proc0 specific glue code. I am not 437 * convinced that they can not be safely combined, but their order of 438 * operation has been maintained as the same as the original init_main.c 439 * for right now. 440 */ 441 /* ARGSUSED*/ 442 static void 443 proc0_init(void *dummy __unused) 444 { 445 struct proc *p; 446 struct thread *td; 447 struct ucred *newcred; 448 struct uidinfo tmpuinfo; 449 struct loginclass tmplc = { 450 .lc_name = "", 451 }; 452 vm_paddr_t pageablemem; 453 int i; 454 455 GIANT_REQUIRED; 456 p = &proc0; 457 td = &thread0; 458 459 /* 460 * Initialize magic number and osrel. 461 */ 462 p->p_magic = P_MAGIC; 463 p->p_osrel = osreldate; 464 465 /* 466 * Initialize thread and process structures. 467 */ 468 procinit(); /* set up proc zone */ 469 threadinit(); /* set up UMA zones */ 470 471 /* 472 * Initialise scheduler resources. 473 * Add scheduler specific parts to proc, thread as needed. 474 */ 475 schedinit(); /* scheduler gets its house in order */ 476 477 /* 478 * Create process 0 (the swapper). 479 */ 480 LIST_INSERT_HEAD(&allproc, p, p_list); 481 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash); 482 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 483 p->p_pgrp = &pgrp0; 484 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 485 LIST_INIT(&pgrp0.pg_members); 486 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 487 488 pgrp0.pg_session = &session0; 489 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF); 490 refcount_init(&session0.s_count, 1); 491 session0.s_leader = p; 492 493 p->p_sysent = &null_sysvec; 494 p->p_flag = P_SYSTEM | P_INMEM | P_KPROC; 495 p->p_flag2 = 0; 496 p->p_state = PRS_NORMAL; 497 p->p_klist = knlist_alloc(&p->p_mtx); 498 STAILQ_INIT(&p->p_ktr); 499 p->p_nice = NZERO; 500 td->td_tid = THREAD0_TID; 501 tidhash_add(td); 502 TD_SET_STATE(td, TDS_RUNNING); 503 td->td_pri_class = PRI_TIMESHARE; 504 td->td_user_pri = PUSER; 505 td->td_base_user_pri = PUSER; 506 td->td_lend_user_pri = PRI_MAX; 507 td->td_priority = PVM; 508 td->td_base_pri = PVM; 509 td->td_oncpu = curcpu; 510 td->td_flags = TDF_INMEM; 511 td->td_pflags = TDP_KTHREAD; 512 td->td_cpuset = cpuset_thread0(); 513 td->td_domain.dr_policy = td->td_cpuset->cs_domain; 514 prison0_init(); 515 p->p_peers = 0; 516 p->p_leader = p; 517 p->p_reaper = p; 518 p->p_treeflag |= P_TREE_REAPER; 519 LIST_INIT(&p->p_reaplist); 520 521 strncpy(p->p_comm, "kernel", sizeof (p->p_comm)); 522 strncpy(td->td_name, "swapper", sizeof (td->td_name)); 523 524 callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0); 525 callout_init_mtx(&p->p_limco, &p->p_mtx, 0); 526 callout_init(&td->td_slpcallout, 1); 527 TAILQ_INIT(&p->p_kqtim_stop); 528 529 /* Create credentials. */ 530 newcred = crget(); 531 newcred->cr_ngroups = 1; /* group 0 */ 532 /* A hack to prevent uifind from tripping over NULL pointers. */ 533 curthread->td_ucred = newcred; 534 tmpuinfo.ui_uid = 1; 535 newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo; 536 newcred->cr_uidinfo = uifind(0); 537 newcred->cr_ruidinfo = uifind(0); 538 newcred->cr_loginclass = &tmplc; 539 newcred->cr_loginclass = loginclass_find("default"); 540 /* End hack. creds get properly set later with thread_cow_get_proc */ 541 curthread->td_ucred = NULL; 542 newcred->cr_prison = &prison0; 543 newcred->cr_users++; /* avoid assertion failure */ 544 proc_set_cred_init(p, newcred); 545 newcred->cr_users--; 546 crfree(newcred); 547 #ifdef AUDIT 548 audit_cred_kproc0(newcred); 549 #endif 550 #ifdef MAC 551 mac_cred_create_swapper(newcred); 552 #endif 553 /* Create sigacts. */ 554 p->p_sigacts = sigacts_alloc(); 555 556 /* Initialize signal state for process 0. */ 557 siginit(&proc0); 558 559 /* Create the file descriptor table. */ 560 p->p_pd = pdinit(NULL, false); 561 p->p_fd = fdinit(NULL, false, NULL); 562 p->p_fdtol = NULL; 563 564 /* Create the limits structures. */ 565 p->p_limit = lim_alloc(); 566 for (i = 0; i < RLIM_NLIMITS; i++) 567 p->p_limit->pl_rlimit[i].rlim_cur = 568 p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY; 569 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur = 570 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 571 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur = 572 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 573 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz; 574 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz; 575 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz; 576 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz; 577 /* Cast to avoid overflow on i386/PAE. */ 578 pageablemem = ptoa((vm_paddr_t)vm_free_count()); 579 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur = 580 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem; 581 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3; 582 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem; 583 p->p_cpulimit = RLIM_INFINITY; 584 585 PROC_LOCK(p); 586 thread_cow_get_proc(td, p); 587 PROC_UNLOCK(p); 588 589 /* Initialize resource accounting structures. */ 590 racct_create(&p->p_racct); 591 592 p->p_stats = pstats_alloc(); 593 594 /* Allocate a prototype map so we have something to fork. */ 595 p->p_vmspace = &vmspace0; 596 refcount_init(&vmspace0.vm_refcnt, 1); 597 pmap_pinit0(vmspace_pmap(&vmspace0)); 598 599 /* 600 * proc0 is not expected to enter usermode, so there is no special 601 * handling for sv_minuser here, like is done for exec_new_vmspace(). 602 */ 603 vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0), 604 p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser); 605 606 /* 607 * Call the init and ctor for the new thread and proc. We wait 608 * to do this until all other structures are fairly sane. 609 */ 610 EVENTHANDLER_DIRECT_INVOKE(process_init, p); 611 EVENTHANDLER_DIRECT_INVOKE(thread_init, td); 612 #ifdef KDTRACE_HOOKS 613 kdtrace_proc_ctor(p); 614 kdtrace_thread_ctor(td); 615 #endif 616 EVENTHANDLER_DIRECT_INVOKE(process_ctor, p); 617 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td); 618 619 /* 620 * Charge root for one process. 621 */ 622 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0); 623 PROC_LOCK(p); 624 racct_add_force(p, RACCT_NPROC, 1); 625 PROC_UNLOCK(p); 626 } 627 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL); 628 629 /* ARGSUSED*/ 630 static void 631 proc0_post(void *dummy __unused) 632 { 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 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL); 666 667 /* 668 *************************************************************************** 669 **** 670 **** The following SYSINIT's and glue code should be moved to the 671 **** respective files on a per subsystem basis. 672 **** 673 *************************************************************************** 674 */ 675 676 /* 677 * List of paths to try when searching for "init". 678 */ 679 static 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 685 SYSCTL_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 695 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT; 696 SYSCTL_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 */ 704 static void 705 start_init(void *dummy) 706 { 707 struct image_args args; 708 int error; 709 char *var, *path; 710 char *free_init_path, *tmp_init_path; 711 struct thread *td; 712 struct proc *p; 713 struct vmspace *oldvmspace; 714 715 TSENTER(); /* Here so we don't overlap with mi_startup. */ 716 717 td = curthread; 718 p = td->td_proc; 719 720 vfs_mountroot(); 721 722 /* Wipe GELI passphrase from the environment. */ 723 kern_unsetenv("kern.geom.eli.passphrase"); 724 725 /* For Multicons, report which console is primary to both */ 726 if (boothowto & RB_MULTIPLE) { 727 if (boothowto & RB_SERIAL) 728 printf("Dual Console: Serial Primary, Video Secondary\n"); 729 else 730 printf("Dual Console: Video Primary, Serial Secondary\n"); 731 } 732 733 if ((var = kern_getenv("init_path")) != NULL) { 734 strlcpy(init_path, var, sizeof(init_path)); 735 freeenv(var); 736 } 737 free_init_path = tmp_init_path = strdup(init_path, M_TEMP); 738 739 while ((path = strsep(&tmp_init_path, ":")) != NULL) { 740 if (bootverbose) 741 printf("start_init: trying %s\n", path); 742 743 memset(&args, 0, sizeof(args)); 744 error = exec_alloc_args(&args); 745 if (error != 0) 746 panic("%s: Can't allocate space for init arguments %d", 747 __func__, error); 748 749 error = exec_args_add_fname(&args, path, UIO_SYSSPACE); 750 if (error != 0) 751 panic("%s: Can't add fname %d", __func__, error); 752 error = exec_args_add_arg(&args, path, UIO_SYSSPACE); 753 if (error != 0) 754 panic("%s: Can't add argv[0] %d", __func__, error); 755 if (boothowto & RB_SINGLE) 756 error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE); 757 if (error != 0) 758 panic("%s: Can't add argv[0] %d", __func__, error); 759 760 /* 761 * Now try to exec the program. If can't for any reason 762 * other than it doesn't exist, complain. 763 * 764 * Otherwise, return via fork_trampoline() all the way 765 * to user mode as init! 766 */ 767 KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0, 768 ("nested execve")); 769 oldvmspace = td->td_proc->p_vmspace; 770 error = kern_execve(td, &args, NULL, oldvmspace); 771 KASSERT(error != 0, 772 ("kern_execve returned success, not EJUSTRETURN")); 773 if (error == EJUSTRETURN) { 774 exec_cleanup(td, oldvmspace); 775 free(free_init_path, M_TEMP); 776 TSEXIT(); 777 return; 778 } 779 if (error != ENOENT) 780 printf("exec %s: error %d\n", path, error); 781 } 782 free(free_init_path, M_TEMP); 783 printf("init: not found in path %s\n", init_path); 784 panic("no init"); 785 } 786 787 /* 788 * Like kproc_create(), but runs in its own address space. We do this 789 * early to reserve pid 1. Note special case - do not make it 790 * runnable yet, init execution is started when userspace can be served. 791 */ 792 static void 793 create_init(const void *udata __unused) 794 { 795 struct fork_req fr; 796 struct ucred *newcred, *oldcred; 797 struct thread *td; 798 int error; 799 800 bzero(&fr, sizeof(fr)); 801 fr.fr_flags = RFFDG | RFPROC | RFSTOPPED; 802 fr.fr_procp = &initproc; 803 error = fork1(&thread0, &fr); 804 if (error) 805 panic("cannot fork init: %d\n", error); 806 KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1")); 807 /* divorce init's credentials from the kernel's */ 808 newcred = crget(); 809 sx_xlock(&proctree_lock); 810 PROC_LOCK(initproc); 811 initproc->p_flag |= P_SYSTEM | P_INMEM; 812 initproc->p_treeflag |= P_TREE_REAPER; 813 oldcred = initproc->p_ucred; 814 crcopy(newcred, oldcred); 815 #ifdef MAC 816 mac_cred_create_init(newcred); 817 #endif 818 #ifdef AUDIT 819 audit_cred_proc1(newcred); 820 #endif 821 proc_set_cred(initproc, newcred); 822 td = FIRST_THREAD_IN_PROC(initproc); 823 crcowfree(td); 824 td->td_realucred = crcowget(initproc->p_ucred); 825 td->td_ucred = td->td_realucred; 826 PROC_UNLOCK(initproc); 827 sx_xunlock(&proctree_lock); 828 crfree(oldcred); 829 cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc), 830 start_init, NULL); 831 } 832 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL); 833 834 /* 835 * Make it runnable now. 836 */ 837 static void 838 kick_init(const void *udata __unused) 839 { 840 struct thread *td; 841 842 td = FIRST_THREAD_IN_PROC(initproc); 843 thread_lock(td); 844 TD_SET_CAN_RUN(td); 845 sched_add(td, SRQ_BORING); 846 } 847 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL); 848 849 /* 850 * DDB(4). 851 */ 852 #ifdef DDB 853 static void 854 db_show_print_syinit(struct sysinit *sip, bool ddb) 855 { 856 const char *sname, *funcname; 857 c_db_sym_t sym; 858 db_expr_t offset; 859 860 #define xprint(...) \ 861 if (ddb) \ 862 db_printf(__VA_ARGS__); \ 863 else \ 864 printf(__VA_ARGS__) 865 866 if (sip == NULL) { 867 xprint("%s: no sysinit * given\n", __func__); 868 return; 869 } 870 871 sym = db_search_symbol((vm_offset_t)sip, DB_STGY_ANY, &offset); 872 db_symbol_values(sym, &sname, NULL); 873 sym = db_search_symbol((vm_offset_t)sip->func, DB_STGY_PROC, &offset); 874 db_symbol_values(sym, &funcname, NULL); 875 xprint("%s(%p)\n", (sname != NULL) ? sname : "", sip); 876 xprint(" %#08x %#08x\n", sip->subsystem, sip->order); 877 xprint(" %p(%s)(%p)\n", 878 sip->func, (funcname != NULL) ? funcname : "", sip->udata); 879 #undef xprint 880 } 881 882 DB_SHOW_COMMAND(sysinit, db_show_sysinit) 883 { 884 struct sysinit **sipp; 885 886 db_printf("SYSINIT vs Name(Ptr)\n"); 887 db_printf(" Subsystem Order\n"); 888 db_printf(" Function(Name)(Arg)\n"); 889 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 890 db_show_print_syinit(*sipp, true); 891 if (db_pager_quit) 892 break; 893 } 894 } 895 #endif /* DDB */ 896